CULTURE OF LEUKEMIA INITIATING CELLS

The present invention relates to the in vitro and ex vivo culture of leukemia initiating cells. In particular, the invention relates to a method for culturing leukemia initiating cells in vitro. This method has various applications, including screening for candidate agents for treating leukemia, identifying leukemia patients likely to respond to a candidate agent for treating leukemia, monitoring a leukemia patient for drug-resistance, e.g. chemoresistance, to an agent, predicting the prognosis of a leukemia patient and determining the ability of a therapeutic agent to reduce the proportion of LICs in a leukemic patient.

Leukemia is a type of cancer of the blood or bone marrow and is characterised by an abnormal increase of immature white blood cells. Myeloid leukemia is a type of leukemia affecting myeloid tissue. Myeloid leukemia can be divided into acute and chronic forms . Acute myeloid leukemia (AML) is characterised by the infiltration of leukemic myeloid blasts that have been arrested at various maturation steps into the bone marrow, whereas chronic myeloid leukemia (CML) is characterised by the excessive build up of relatively mature, but still abnormal, myeloid cells.

Myelodysplastic syndrome (MDS) refers to a heterogeneous group of closely related clonal hematopoietic disorders characterised by a hypercellular or hypocellular marrow with impaired morphology and maturation and peripheral blood cytopenias resulting from ineffective blood cell production (Besa, 1992) . CML and MDS can be considered as a premalignant condition in a subgroup of patients that often progresses to AML when additional genetic abnormalities are acquired.

Despite advances in diagnosis and clinical care over the years, wherein 50 to 75% of AML patients reach a complete remission state, AML remains incurable for many patients and only 20 to 30% of patients achieve long-term survival (Lowenberg et al., 1999). One established paradigm of leukemogenesis provides that leukemia arises from a single cell and is maintained by a small population of leukemia initiating cells (LICs) (Bonnet and Dick, 1997;

Lapidot et al . , 1994). LICs are typically chemo-resistant due to intrinsic and extrinsic factors: efficient DNA repair pathways, high level of drug efflux pumps and quiescence account for intrinsic drug resistance, whereas environment-mediated drug resistance (EM-DR) arises from a complex concomitance of soluble factors (SFM-DR) and cell adhesion-mediated drug resistant mechanisms (CAM-DR) (Damiano et al . , 1999; Meads et al., 2008). Chemo-resistant LICs are responsible for AML relapse and represent the target for future innovative therapies (Zhou et al., 2009). A number of studies also indicate the existence of LIC in CML and MDS (Chung et al . , 2008; Holyoake et al . , 1999).

Myeloid leukemias are a highly heterogeneous group of clonal leukemias, i.e. AML can be divided into 8 subtypes, with 54 cytogenetic subgroups identified with (i) different abnormalities distributed on all chromosomes and/or (ii) dozens of mutations and/or overexpressions responsible for constitutive signalling pathway activations at the molecular level (Grimwade et al . , 2010) . Our group and others have recently demonstrated that the LIC phenotype was more heterogeneous than initially thought, thus requiring long in vivo xenograft experiments before working on a true LIC subpopulation (Goardon et al., 2011; Sarry et al., 2011; Taussig et al., 2008; Taussig et al . , 2010). Several deregulated constitutively activated pathways have been described in AML, including Ras/Raf PI3K/AKT NFkappaB, due to upstream tyrosine kinase mutated receptors (Cairoli et al . , 2006; Cheong et al., 2003; Guzman et al . , 2001; Guzman et al., 2002; Illmer et al . , 2005; Martelli et al., 2007; Xu et al., 2003). The rationale of using drugs targeting those pathways is mainly based on

retrospective patient studies establishing a prognostic impact of such deregulations as well as mouse model data demonstrating their leukemogenic potential. Moreover, new studies have also established the rationale of targeting LSC-microenvironment crosstalk signalling (Liesveld et al . , 2007; Wang et al., 2011). Techniques currently available for culturing primary AML, CML and MDS samples have been adapted from standardised protocols for culturing normal hematopoietic cells. While in vitro short-term liquid culture and colony assays allow quantification of leukemic precursors, the identification of LICs (which have the ability to recapitulate the disease and self-renew) is still dependent on in vivo xenograft assays. While a powerful tool in one respect, the AML xenograft assay requires a timeframe of eight to twenty-four weeks before analysing and remains prospectively blind with little ability to monitor progress. These limitations make its use difficult for routine drug screening and investigation of the role of specific genes and pathways involved in LIC maintenance in steady state or under chemotherapy.

Characterisation and long-term maintenance of both haematopoietic stem cells (HSCs) and leukemia initiating cells (LICs) ex vivo are of crucial importance for the advancement of fundamental studies and the development and screening of new drugs.

Different co-culture systems supporting some degree of normal and leukemic human hematopoiesis over a few weeks have been

described. However, few of these studies truly demonstrate the maintenance of normal HSCs through xenotransplantation

assessment. Moreover, in the context of leukemia, although it has been shown that leukemic cobblestone area-forming cells (L- CAFC) and/or leukemic long-term culture initiating cells (L-LTC- IC) can be maintained for a few weeks in co-culture (Allies et al . , 1997; van Gosliga et al . , 2007), no evidence has been reported to date demonstrating the actual ex vivo maintenance of LICs using the xenograft model. In addition, studies

investigating EM-DR in vitro have never been applied to LICs.

A primary goal of establishing long-term cultures of HSC has involved mimicking the bone marrow microenvironment ex vivo.

Cytokines and extracellular matrix proteins play an essential role in providing a supportive environment but HSCs are also influenced by cues provided through cell-cell contact, "stem cell niche synapses", between different cell types in the bone marrow microenvironment . Pre-osteoblasts , osteoblasts, endothelial cells and mesenchymal cells, have each been shown to be

fundamental components of the HSC niche (Calvi et al., 2003;

Hooper et al . , 2009; Kiel et al., 2005; Mendez-Ferrer et al., 2010; Zhang et al., 2003) . In vitro co-cultures of HSCs with primary osteoblasts have been reported (Taichman et al., 1996), although studies have predominantly been performed with the osteosarcoma SaOS-2 cell line as an osteoblastic niche model (Gillette et al., 2009; Rodan et al . , 1987). HSC co-culture with endothelial cells from different origins have been shown to promote an increase in L-LTC-IC frequency and cobblestone area- forming cells (CAFC) (Lu et al., 1996; Rafii et al . , 1995), as well as the maintenance of repopulating cells (Li et al . , 2004). The mesenchymal cell line MS-5 has been used for expanding human hematopoietic stem/progenitor cells in vitro (Bennaceur-Griscelli et al . , 1999; Issaad et al . , 1993) . Using a co-culture system with MS-5, several groups have recently reported that normal human HSCs capable of repopulating NOD/SCID mice (SCID

repopulating cell [SRC] activity) could be expanded in vitro for up to five weeks (Amsellem et al., 2003; Vanheusden et al . , 2007) . In the leukemic context, survival and proliferative benefit have been reported in AML samples co-cultured with the SaOS-2 or HUVEC cell lines (Bruserud et al., 2004; Dias et al . , 2002; Glenjen et al., 2005; Liesveld et al., 2005). MS-5 co- cultures supplemented with IL3, G-CSF and TPO can also sustain primary AML samples over 24 weeks with successful generation of leukemic CAFC and primary leukemic LTC-ICs (van Gosliga et al., 2007) . However it is not known if any of these co-culture systems currently used for ex vivo AML studies can sustain LICs.

HSC functions are maintained through distinct physical and chemical features of the HSC niche, one of which is oxygen levels (Kubota et al., 2008; Parmar et al., 2007; Rehn et al . , 2011; Simsek et al., 2010; Takubo et al., 2010) . In vitro, hypoxia favors normal hematopoietic stem cell quiescence and maintenance, whereas normoxia induces proliferation and differentiation, followed by exhaustion (Danet et al., 2003; Hammoud et al . , 2011; Hermitte et al., 2006). In AML, recent data have demonstrated a requirement for the hypoxia-inducible factor 1 alpha (HIFla) pathway for the maintenance of LICs in vivo (Wang et al., 2011) . In vitro, AML cells cultured at a reduced oxygen level increase SDF-1/CXCR4 and PI3K signaling by promoting lipid raft formation (Fiegl et al . , 2009; Fiegl et al., 2010). The impact hypoxia has on survival in steady state conditions or under drug treatment is less clear, depending on which population and which drug is being studied (Chan et al., 2008; Comerford et al., 2002; Erler et al., 2004). The chemo-protective role of hypoxia in LICs in vitro has not been explored.

Over the past decade, oncology candidate drugs have been more likely to fail at the stage between successful preclinical experiments and attempts to translate the results into the clinic. Some call this failure the "Valley of Death" (Nathan, 2011) . There is usually no single reason why a phase III trial in oncology might fail and there are normally several different considerations. Firstly, it is important to bear in mind that a phase III trial is justified after phase I/II studies have established the drug anti-tumoural activity. On the other hand, the challenge at the phase III stage is to demonstrate patient survival benefit. Paradoxically, different studies have demonstrated that the assessment of bulk tumour shrinkage was not necessarily correlated to an improvement in patient survival

(Ardeshna et al . , 2003; Durie et al., 2004; Huff et al . , 2006). Thus, one could consider that the read-outs and the biological markers currently monitored in trials are not suitable for directly quantifying the efficiency of a therapy to target cancer cells of interest. A plausible explanation for the lack of correlation between objective tumour response rates and overall survival is that conventional cytotoxic therapies do not kill the Cancer Stem Cells (CSCs) at the root of cancer. Residual CSCs that are not eradicated may reinitiate tumour growth, leading eventually to clinical relapse and this is the case for AML.

Many studies have emphasised the importance of developing new approaches capable of monitoring the impact of the treatment at the CSC level.

To date, the majority of cell-based functional screens for candidate drugs for treating cancer have relied on probing cell lines in isolation in vitro to identify compounds that decrease cellular viability. The first problem in those approaches is that cancer cell lines do not recapitulate the hierarchical distribution of cancer cells in primary samples. Secondly, attempts to use primary samples for drug discovery would be limited by the lack of demonstration of the actual maintenance CSCs in vitro.

The present invention arises from the first direct comparison between different co-culture systems and the assessment of the impact of low oxygen availability on functionally defined early progenitors and stem cells via long-term secondary culture assays and in vivo repopulating activity. As a result of this work, the inventors have identified a reliable, easy and reproducible culture system suitable for the maintenance of both human primary HSC and LICs. This system can be used as a xenograft surrogate model for LIC handling. Particularly, the present invention provides methods to study and quantify the drug-resistance of LICs and to screen the activity of new therapeutic agents on LICs. Therefore, the method of the invention can assist in conducting or monitoring a clinical study.

In a first aspect, the invention provides a method for culturing leukemia initiating cells (LICs) in vitro, the method comprising culturing the LICs at an oxygen concentration of 6.8% per volume or lower. This method allows the long term culture (e.g. for at least three weeks) of LICs in vitro.

The oxygen concentration may be 6.8% per volume or lower, 6% per volume or lower, 5% per volume or lower, 4% per volume or lower, or 3% per volume or lower. Preferably, the oxygen concentration is between 0.1% per volume and 6.8% per volume inclusive. The method may comprise co-culturing the LICs with primary stromal cells from adult or embryonic tissues or with stromal cell derived cell lines. The stromal cells may be mesenchymal cells, such as mesenchymal stem cells (MSC) , pre-osteoblasts , osteoblasts, chondrocytes or endothelial cells. For example, the stromal cells may be primary MSC, primary pre-osteoblasts, primary osteoblasts, primary chondrocytes or primary endothelial cells. These types of stromal cells may be directly isolated from adult or embryonic tissues. Alternatively, pre-osteoblasts, osteoblasts, chondrocytes and endothelial cells, may be derived from MSC by differentiation or by other methods known to one skilled in the art. Primary mesenchymal cells may be pericytes or mesenchymal cells derived from the endothelial-to-mesenchymal transition. Primary endothelial cells may be bone marrow microvascular endothelial cells (BMEC) , endothelial progenitor cells (EPC) , umbilical vein endothelial cells (UVEC) . Stromal cells may be of mammalian origin. For example, the mammal may be a mouse or human. Preferably the stromal cells are mesenchymal cells, e.g. MS-5 cells.

The method may comprise culturing the LICs in medium supplemented with one or more cytokines. For example, the method may comprise culturing the LICs in a defined medium, such as serum free medium supplemented with cytokines. Cytokines known to one skilled in the art may promote myeloid activation, such as interleukin-3 (IL-3) , granulocyte colony-stimulating factor (G-CSF) or thrombopoietin (TPO) , or they may promote endothelial support and/or activation such as vascular endothelial growth factor (VEGF) or interleukin-1 (IL-1) . Preferably, the medium is supplemented with cytokines IL-3, G-CSF and/or TPO ("3GT" when added in combination) .

The method may comprise culturing the LICs in the presence of undefined medium, for example medium supplemented with mammalian serum, such as serum of human, bovine and/or equine origin. In another embodiment, the method may comprise culturing the LICs using conditioned media harvested from cultured primary normal or cancerous cells or cell lines. "Conditioned medium" is obtained culturing cells in a medium such that the cells release factors into the medium changing its composition. The medium is then collected as "conditioned medium".

The method may comprise culturing the LICs in defined or undefined media supplemented with one or more extracellular matrix protein such as collagen, fibronectin, and/or one or more proteoglycans. Various proteoglycans that can mimic the bone marrow extracellular matrix are known to those skilled in the art. For example, three-dimensional (3D) semi-solid media may be formed from natural molecules and/or synthetic polymers such as fibronectin, collagen type I, II, IV, laminin, methylcellulose , a Matrigel based matrix, or an alginate based hydrogel (Celebi et al., 2011; Leisten et al . , 2012) .

The LICs may be drug-resistant, e.g. chemoresistant, and the results presented herein show that, using the method of the invention, drug-resistant, e.g. chemoresistant, LICs can be maintained and indirectly quantified through long term culture in vi tro .

The LICs may have been obtained from a patient with leukemia, such as acute myeloid leukemia (AML) , myelodysplastic syndrome (MDS) or chronic myeloid leukemia (CML) . The AML may be AML de novo, secondary AML or therapy related AML. The LICs may have been obtained from a patient with leukemia at diagnosis or during different stages of the disease evolution and clinical care.

The LICs may be cultured in vitro for at least 1 week, at least 2 weeks, at least 3 weeks, at least 4 weeks, at least 5 weeks, at least 6 weeks, at least 7 weeks, at least 8 weeks, at least 9 weeks, at least 10 weeks, at least 11 weeks, at least 12 weeks, 13 weeks, at least 14 weeks, at least 15 weeks, at least 16 weeks, at least 17 weeks, at least 18 weeks, at least 19 weeks, at least 20 weeks, at least 21 weeks, at least 22 weeks, at least 23 weeks, or at least 24 weeks.

Once LICs have been cultured according to the method of the invention for at least one week and are capable of re-initiating secondary culture and ultimately giving rise to leukemic colony forming cells (L-CFC) progeny, they may be referred to as leukemia long term culture initiating cells (L-LTC-ICs) . LICs that have been cultured according to the method of the invention for at least five weeks may also be referred to as L-LTC-ICs.

The method may further comprise the step of re-plating the LICs and/or the L-LTC-ICs onto fresh stromal cells. This represents a secondary culture. Secondary culture may be maintained for at least 1 week, at least 2 weeks, at least 3 Weeks, at least 4 weeks, at least 5 weeks, at least 6 weeks, at least 7 weeks, at least 8 weeks, at least 9 weeks, at least 10 weeks, at least 11 weeks, at least 12 weeks, 13 weeks, at least 14 weeks, at least 15 weeks, at least 16 weeks, at least 17 weeks, at least 18 weeks, at least 19 weeks, at least 20 weeks, at least 21 weeks, at least 22 weeks, at least 23 weeks, or at least 24 weeks. The stromal cells used for this secondary culture may be the same type or a different type of stromal cells from the first culture. The secondary culture may use the same defined or undefined media and cytokine supplementation as the first culture, or different media and cytokine supplementation. Similarly, the secondary culture may be performed using the same oxygen concentration as the first culture or a different oxygen concentration. For example, the secondary culture may be performed at an oxygen concentration of 20% per volume (i.e. under normoxic conditions). Preferably, the secondary culture of re-plated LICs and/or L-LTC- ICs may be performed using conditions that allow clonogenic potential quantification of the LICs and/or L-LTC-ICs, such as normoxia, e.g. an oxygen concentration of about 20% per volume. Such conditions are known to a person skilled in the art. For example, long term co-cultured leukemic cells may be harvested and seeded into semi-solid methylcellulose supplemented with cytokines promoting myeloid colony formation, such as stem cell factor (SCF) , IL-3, interleukin-6 (IL-6) , macrophage colony- stimulating factor (M-CSF) , granulocyte macrophage colony- stimulating factor (GM-CSF) , or granulocyte colony-stimulating factor (G-CSF)

Following re-plating onto fresh stromal cells, L-LTC-ICs may be referred to as secondary L-LTC-ICs. However, if no step of re- plating onto fresh stromal cells is undertaken, the L-LTC-ICs may be referred to as primary L-LTC-ICs.

The method may further include the step of determining the total number or frequency of LICs and/or primary L-LTC-ICs present in the culture before the step of re-plating and/or the total number or frequency of LICs and/or secondary L-LTC-ICs present in the secondary culture following the step of re-plating.

In a second aspect, the invention provides an in vitro method of screening for a candidate agent for treating leukemia, the method comprising:

(i) culturing a test sample of LICs according to the method described in the first aspect above;

(ii) contacting the test sample of LICs with the candidate agent; and

(iii) determining the number or frequency of LICs and/or L-

LTC-ICs present in the test sample before and after step (ii), wherein a decrease in the number or frequency of LICs and/or L- LTC-ICs following step (ii) indicates that the candidate agent may be effective for treating leukemia.

The test sample may be contacted with a single candidate agent or with a combination of one or more candidate agents. The test sample may be contacted with the different agents simultaneously or sequentially.

As well as being contacted with a single candidate agent or with a combination of one or more candidate agents, the test sample may also be contacted with an agent already known for treating leukemia. The test sample may be contacted with the different agents simultaneously or sequentially. The effectiveness of the candidate agent, or combined candidate agents, may be compared to the effectiveness of the agent already known for treating leukemia .

The i n vitro method of screening may further comprise the step of re-plating the LICs and/or L-LTC-ICs onto fresh stromal cells after step (ii) . The stromal cells used in this secondary culture may be the same type or a different type from the stromal cells used in the first culture. The secondary culture may use the same defined or undefined media and cytokine supplementation as the first culture, or different media and cytokine

supplementation. Similarly, the secondary culture may be performed using the same oxygen concentration as the first culture or a different oxygen concentration. For example, the oxygen concentration may by about 20% per volume (i.e. normoxic conditions) . Preferably, the secondary culture of re-plated LICs and/or L-LTC-ICs may be performed using conditions that allow clonogenic potential quantification of the LICs and/or L-LTC-ICs. Such conditions are known to a person skilled in the art and include normoxia, i.e. an oxygen concentration of about 20% per volume .

The method may further include the step of determining the number or frequency of LICs and/or primary L-LTC-ICs present in the culture before the step of re-plating and/or the total number or frequency of LICs and/or secondary L-LTC-ICs present in the secondary culture following the step of re-plating. A decrease in the number or frequency of LICs and/or secondary L-LTC-ICs following re-plating compared to a control sample that has not been contacted with the candidate agent indicates that the candidate agent may be effective for treating leukemia. In a third aspect, the invention provides an ex vivo method for identifying a leukemia patient likely to respond to a candidate agent for treating leukemia, the method comprising:

(i) culturing a test sample of LICs obtained from the leukemia patient according to the method described in the first aspect above;

(ii) contacting the test sample with the candidate agent; and

(iii) determining the frequency or number of LICs and/or L- LTC-ICs present in the test sample before and after step (ii) , wherein a decrease in the frequency or number of LICs and/or L- LTC-ICs following step (ii) indicates that the leukemia patient is likely to respond to the candidate agent. The test sample may be contacted with a single candidate agent or with a combination of one or more candidate agents. The test sample may be contacted with the different agents simultaneously or sequentially. As well as being contacted with a single candidate agent or with a combination of one or more candidate agents, the test sample may also be contacted with an agent already known for treating leukemia. The effectiveness of the candidate agent, or combined candidate agents, may be compared to the effectiveness of the agent already known for treating leukemia.

The method may further comprise the step of re-plating the LICs and/or L-LTC-ICs onto fresh stromal cells after step (ii) . The method may further include the step of determining the number or frequency of LICs and/or primary L-LTC-ICs present in the culture before the step of re-plating and/or the number or frequency of LICs and/or secondary L-LTC-ICs present in the secondary culture following the step of re-plating. A decrease in the number or frequency of LICs and/or secondary L-LTC-ICs following re-plating compared to a control sample that has not been contacted with the candidate agent indicates that the leukemia patient is likely to respond to the candidate agent.

In a fourth aspect, the invention provides an ex vivo method for monitoring a leukemia patient for resistance to a therapeutic agent, the method comprising:

(i) culturing a test sample of LICs obtained from the leukemia patient according to the method described in the first aspect above;

(ii) contacting the test sample with the agent; and

(iii) determining the frequency or number of LICs and/or L- LTC-ICs present in the test sample before and after step (ii) , wherein a decrease in the frequency or number of LICs and/or L- LTC-ICs following step (ii) indicates that the leukemia patient is not resistant to the agent.

The method may further comprise the step of re-plating the LICs and/or L-LTC-ICs onto fresh stromal cells after step (ii) .

The method may further include the step of determining the total number or frequency of LICs and/or primary L-LTC-ICs present in the culture before the step of re-plating and/or the total number or frequency of LICs and/or secondary L-LTC-ICs present in the secondary culture following the step of re-plating. A decrease in the number or frequency of LICs and/or secondary L-LTC-ICs following re-plating compared to a control sample that has not been contacted with the agent indicates that the patient is not resistant to the agent.

In a fifth aspect, the invention provides an ex vivo method for predicting the prognosis of a leukemia patient, the method comprising :

(i) culturing a test sample of leukemic cells obtained from the leukemia patient in vitro for at least one week; and (ii) determining the proliferation rate of living leukemic cells in the test sample; wherein the proliferation rate

correlates to the prognosis of the patient. Typically, a higher proliferation rate indicates a worse

prognosis for the patient than a lower proliferation rate.

In step (i) , the test sample may be cultured according to the method described in the first aspect above.

The test sample of leukemic cells may be obtained from a patient with any type of leukemia, such as acute myeloid leukemia (AML) , myelodysplastic syndrome (MDS) or chronic myeloid leukemia (CML) . These test samples may have been obtained from peripheral blood, bone marrow or leukapheresis samples taken from these leukemia patients .

In a sixth aspect, the invention provides an ex vivo method for determining the ability of a therapeutic agent to reduce the proportion of LICs in a leukemia patient, the method comprising:

(i) culturing a sample of LICs obtained from a leukemia patient and containing LICs according to the method described in the first aspect above;

(ii) determining the proportion of LICs and/or L-LTC-ICs in the sample;

(iii) contacting the sample with the therapeutic agent;

(iv) determining the proportion of LICs and/or L-LTC-ICs in the sample;

wherein a decrease in the proportion of LICs and/or L-LTC-ICs in the sample following step (iii) indicates that the therapeutic agent is likely to reduce the proportion of LICs when used to treat the patient.

The method may further comprise the step of re-plating the LICs and/or L-LTC-ICs onto fresh stromal cells after step (iii) . The sample of LICs may be obtained from a patient with any type of leukemia, such as acute myeloid leukemia (AML) ,

myelodysplastic syndrome (MDS) or chronic myeloid leukemia (CML) . These test samples may have been obtained from peripheral blood, bone marrow or leukapheresis samples taken from these leukemia patients .

These and other aspects of the invention are described in further detail below.

BRIEF DESCRIPTION OF THE FIGURES

Figure 1 shows the CD34 CD38 phenotype of normal haematopoietic samples (cord blood lin ^" ) and acute myeloid leukemia patient samples (n=10). Sorted CD34-CD38- (lower left quadrant),

CD34+CD38- (lower right quadrant), CD34+CD38+ (upper right quadrant) and CD34-CD38+ (upper left quadrant) were injected into NOD/SCID mice. To abrogate antibody-mediated clearance of cells, all mice received a total of 1 mg/g of human immunoglobulin (IVIG) (Bio Products Laboratory, Elstree, UK) . Mice received a dose of sub-lethal radiation (330-375 cGy) from a ¹³⁷ caesium source twenty-four hours pre-transplant . Intravenous injection was the preferred route of administration, unless fewer than 10 ⁶ cells were administered where direct intra bone marrow injection was preferred. Black or grey panels indicates a negative or positive human chimerism respectively determined 12 weeks after inj ection .

Figure 2 shows the xenograft assay potential for distinguishing the drug sensitivity of LICs versus non-LICs. Thawed AML cells were co-cultured with MS-5 + 3GT at 3% 02 for 1 week in the absence (CT) or in the presence of 3 μΜ Ara-C (ARA) . An arbitrary 3-fold reduction in the bulk tumour size is displayed in the schematic representation. Xenograft potential was determined after ex vivo chemotherapy. The xenograft assay was performed for each sample in two ways: after 1-week, a single cell dose of sorted CD45+ cells was used or an input equivalent dose was injected into NOD scid gamma (NSG) mice (n=3 to 5 mice per group) . Leukemic engraftment in the recipient bone marrow was determined 12 weeks after injection. A Mann-Whitney unpaired two-tailed test was applied to establish statistically

significant differences observed between the CT and ARA groups. Depending on the sample and the way in which the xenograft assay was performed, the level of leukemic engraftment in the CT group (i) equals that in the ARA group (CT=ARA) , (ii) is superior to that in the ARA group (CT>ARA) , or (iii) is inferior to that in the ARA group (CT<ARA) . The results may be interpreted as follows: (i) if for the single dose CT=ARA and for the equivalent input dose CT>ARA, LICs are as sensitive to the treatment as non- LICs and therefore the bulk tumour shrinkage is informative of the LIC pool shrinkage; (ii) if for the single dose CT>ARA and for the equivalent input dose CT>ARA, LICs are more sensitive to the treatment than non-LICs and therefore the bulk tumour shrinkage understate the LIC pool shrinkage; and (iii) if for the single dose CKARA and for the equivalent input dose CT=ARA, LICs are less sensitive to the treatment than non-LICs thus the bulk tumor shrinkage overstate the LIC pool shrinkage.

Figure 3 shows the potential of the secondary L-LTC assay for distinguishing the drug sensitivity of LICs versus non-LICs.

Thawed AML cells were co-cultured with MS-5 + 3GT in 3% oxygen for 1 week in the absence (CT) or in the presence of 3μΜ Ara-C (ARA) . An arbitrary 3-fold reduction in the bulk tumour size is displayed in the schematic representation. L-LTC re-plating potential was determined following ex vivo chemotherapy. To determine the secondary LTC-IC frequency, LDA analysis was done using LCalc Software (Stem Cell Technologies) according to the Poisson statistics and the method of maximum likelihood.

Depending on the response to treatment of the sample's LICs versus non-LICs, three cases may be observed: secondary L-LTC-IC frequency in the CT group (i) equals that in the ARA group

(CT=ARA) ; (ii) is superior to that in the ARA group (CT>ARA) or (iii) is inferior to that in the ARA group (CKARA) . The results may be interpreted as follows: (i) if the secondary L-LTC-IC frequency in CT=ARA, LICs are as sensitive to the treatment as non-LICs and therefore the bulk tumour shrinkage is informative of the LIC pool shrinkage; (ii) if the secondary L-LTC-IC frequency in CT>ARA, LICs are more sensitive to the treatment than non-LICs and therefore the bulk tumor shrinkage understates the LIC pool shrinkage; and (iii) if the secondary L-LTC-IC frequency in CKARA, LICs are less sensitive to the treatment than non-LICs and therefore the bulk tumor shrinkage overstates the LIC pool shrinkage. Figure 4 shows that the MS-5+3GT preferential supportive

potential for normal HSC is revealed in long term culture (LTC) . Feeders and cytokines impact on CB Lin ^" cell expansion,

clonogenic efficiency and replating potential after 5 weeks of co-culture with MS-5, SaOS-2 or HUVEC. Data are shown as mean number (±SEM) . A: Fold expansion, n=3. Cell expansion was determined by a precise cell count using absolute counting beads. B: Total CFC, n=6. C: Secondary LTC-IC total number within expanded population sorted at week 5. 0: not determined because no secondary culture nor CFC positive wells could be generated. Representative data of 4 independent bulk LTC and 2 replating limiting dilution assay (LDA) experiments. ***p<0. OOland

****p<0.0001.

Figure 5 shows the effect of long term low oxygen culture on HSC read out in vitro and in vivo. CB Lin ^" cells were co-cultured for 5 weeks with MS-5+3GT under 20% or 3% 0 ₂ . A: 5 weeks output frequency of HSC (Lin ^" CD34 ⁺ CD38 ^" ) and progenitors (Lin ^"

CD34 ⁺ CD38 ⁺ ) cells at 20% 0 ₂ or 3% 0 ₂ as compared to 20% 0 ₂ (n=3 independent experiments, *p<0.05, NS p>0.05 in paired t-test) . B: Data shown represent fold expansion of 5 CB samples co-cultured at 20%O ₂ or 3%0 ₂ in parallel. *p<0.05 in paired t-test. C: Total output CFC (+SEM) for 10 ⁴ CB lin ^" cells plated at tO; ***p<0.001, n=3 independent experiments. D: Total input LTC-IC (90% CI) determined at week 5 after the first plating maintained at 20% or 3% 0 ₂ ; ***p<0.001. E: Total secondary LTC-IC (90% CI) determined at week 10 in the secondary replated culture at week 5.

Secondary LTC were performed at 20% 0 ₂ . NS p>0.05. F: Replating potential of sorted CD45+ cells at week-5 cultured at 20% 0 ₂ or 3% 0 ₂ and replated in limiting dilution analysis for 5 additional weeks at 20% 0 ₂ . G: Percentage of human chimerism in the bone marrow of mice transplanted with a single dose of 0.5xl0 ⁶ cells from 5-week co-cultures at 20% 0 ₂ (10 mice; opened diamond) or 3% 0 ₂ (9 mice; closed square); ***p<0.001. H: Limiting dilution analysis of sorted CD45+ cells co-cultured for 5 weeks at 20% 0 ₂ or 3%0 ₂ and injected into NOD/SCID- _? m ^_/" mice (n=75) . I: Total input and output SRC (90% CI) determined for 10 ⁷ CB lin ^" cells uncultured (TO) or co-cultured for 5 weeks with MS-5+3GT under 20% or 3% 0 ₂ ; NS p>0.05. For E and I, data presented take into account first culture fold expansion.

Figure 6 shows that normal HSC expanded with 20% oxygen present a defect in efficient lymphoid lineage engraftment associated with signs of oxidative stress. A: ultilineage analysis of mice injected with CB lin- with un-manipulated (TO, 16 mice; grey circle) or 5 weeks co-cultured cells under 20% 0 ₂ (10 mice;

opened diamond) or 3% 0 ₂ (12 mice; closed square) with MS-5+3GT. Data represent percent of lymphoid CD19+ cells and myeloid CD33+ cells within the human CD45+ population determined 12 weeks after xenotransplantation. Each symbol represents a transplanted mouse analysed. Lines represent mean levels of CD19+ or CD33+ cell engraftment in each group. *p<0.05. NS means non-significant. B: ROS generation in CD45+ cells co-cultured at 20%O ₂ or 3%0 ₂ for 5 weeks with MS-5+3GT. Bl: CARBOXY-H2DFFDA mean fluorescence intensity (MFI) (±SEM) determined by FACS; ***p<0.001. B2 :

Representative FACS plot for H2DFFDA fluorescence intensity for indicated percent of oxygen, thick line: negative control, thin line: 3%0 ₂ , dashed line: 20% 0 ₂ . B3 : Proportions of cells with negative, low, medium and high fluorescence intensity within gate displayed in B2 (±SEM) . C: pl6 gene expression profile was determined in expanded hematopoietic cells sorted from week 5 co- culture; **p<0.01. For B-C Data shown represent mean value

(+SEM) , derived from 4 independent LTC. Figure 7 shows the effect of different feeder cell layers on the viability, maintenance and expansion of AML cells and the requirement for 3GT for long-term maintenance of leukemic cells. A: AML cells percent viability comparison after 1-week co-culture with MS-5, SaOS-2 or HUVEC cells without cytokines. Al : MS-5 and SaOS-2 comparison n=57 samples A2: MS-5 and HUVEC comparison n=38 samples. A3: SaOS-2 and HUVEC comparison n=43 samples.

B: AML cells fold expansion comparison after 1-week co-culture with MS-5, SaOS-2 or HUVEC cells without cytokines. Bl : MS-5 and SaOS-2 comparison n=28 samples B2 : MS-5 and HUVEC comparison n=20 samples. B3 : SaOS-2 and HUVEC comparison n=20 samples. A-B:

samples were cultured in triplicate wells. Bold black lines represent putative equivalent viability and expansion. Thin black lines show experimental derived simple linear regression trend line with the 95% confidence band (dashed lines) .

Comparisons were made using a paired Student's t-test.

C: 3GT is required for long-term maintenance of leukemic cells. AML samples #12 and #1 were co-cultured with MS-5 in normoxia for 5-weeks in absence (MS-5) or in presence of 3GT (MS-5 3GT) . Data represent the mean fold expansion (±SEM, n=6 per condition) .

Comparisons were made using a paired Student's t-test.

D: Leukemic origin of NPM1 mutated AML samples #12 (mutation type 1) and #1 (mutation type 4) expanded cells after 5 weeks. NPMl total and mutated Quantitative genomic PCR on samples #12 (Dl,2) and #1 (D3,4) after 5 weeks co-culture with MS-5 3GT at 20% 02. Amplification plots and standard curves for total NPMl (Dl, 3) and NPM mutated type 1 (D2) and type 4 (D4) . Standard curves are derived from a dilution series of NPM mutated genomic DNA ranging from 200 ng to 0.02 ng from non-manipulated samples. To assess the specificity of the assay, equivalent genomic DNA from MS-5 only were subject to PCR. MS-5 were negative or below 0.02 ng level nor for human NPMl total primers nor for human NPMl mutation type 1 and 4. Amplification plots for sample #12 showing mean of 51.4 ng of NPMl total DNA (Dl) and mean of 55 ng NPMl mutated type 1 (D2) in 100 ng of DNA processed (100% mutated) . Amplification plots for sample #1 showing mean of 14 ng of NPMl total DNA (D3) and mean of 11.6 ng NP 1 mutated type 4 (D4) in 58 ng of DNA processed (82% mutated) .

Figure 8 shows that low oxygen concentrations favour long term in vitro maintenance of LICs over 3 weeks. AML samples were co- cultured for 3 weeks with MS-5+3GT at 20% 0 ₂ or 3% 0 ₂ in parallel. A: 3% 0 ₂ reduce AML expansion. Data shown represents fold expansion of 5 AML samples (samples #2, #7, #10, #3 and #6) determined in 20% 0 ₂ (open symbol) or 3% 0 ₂ (closed symbol) .

*p<0.05 in paired t-test. B: total input LTC-IC (uncultured: grey bars) and output LTC-IC (3 weeks co-cultured at 20% 0 ₂ :

black bars, or 3% 0 ₂ : white bars) per 10 ⁵ starting cells from samples #6, #2 and #3. Uncultured input LTC was determined at week-5 at 20% 0 ₂ . Output cultured cells were sorted at week-3 and CD45+ cells were replated in limiting dilution analysis for 5 additional weeks at 20% 0 ₂ . ***p<0.001; *p<0.05; NS p>0.05.

(90%CI is displayed) C: Replating potential at week-3. Sorted CD45+ AML#6, #2 and #3 cells co-cultured at 20% 0 ₂ or 3% 0 ₂ for 3- weeks were replated in limiting dilution analysis (LDA) for 5 additional weeks at 20% 0 ₂ . D: Total input and output SL-IC count (90% CI) determined for 4 10 ⁷ samples #2 cells uncultured (TO) or co-cultured for 3 weeks at 20% or 3% 0 ₂ . *p<0.05; NS p>0.05. For B-D, data presented take into account first culture fold expansion. E: In vivo LDA of AML #2 cells recovered from co- culture and injected into NOD/SCID- ₂ rrf ^7~ mice (n=52 mice) . F: Primary and secondary leukemic engraftment of AML samples co- cultured for 3 weeks. The data show the percentage of human leukemic chimerism in primary recipient mice (1st) for 6 samples #1, #2, #3, #4, #5 and #6 and secondary recipients (2nd) for samples #1, #2 and #3. Plain lines represent mean levels of leukemic cell engraftment in each group. Dashed line represent the threshold of positivity. Each symbol represents a single transplanted mouse. Figure 9 shows the effect of cytarabine treatment on various cell types. A: Dose-response curves of cytarabine using HEL, HL-60, KG-1 and MS-5 cells. Cell growth was measured using the XTT assay and was plotted as a mean percentage of the control (cells not exposed to drugs) of three separate experiments (done on different days), with each experimental point done in sextuplet (n=6) . AML cells (thin dashed lines) and MS-5 cells (thick plain line) were exposed for 72 h continuously to the designated Log concentration of cytarabine. Displayed IC50 values were computed using CalcuSyn software (Biosoft) . B; Suspension culture and MS- 5 based co-culture system comparison for assessing chemotherapy impact on normal and leukemic hematopoietic cells. CB Lin ^" cells (white bars, n=7 samples) and AML cells (dashed bars, n=6 samples, #1, #2, #3, #6, #10, #11) were cultured in suspension with 3GT (3GT) or co-cultured with MS-5 with 3GT (MS-5 3GT) for 1-week in absence (CT) or in presence of Ara-C 3μΜ (ARA-C) in normoxia. Bl : Data shown represent the mean (±SEM) normalized cellular count percent as compared to the more permissive condition (untreated CT MS-5 3GT) . B2 : Cell viability was assessed by DAPI and Annexin V exclusion. NS p>0.05; *p=0.0140 and ***p=0.0004. Figure 10 shows that low oxygen concentration enhances in vitro chemoresistance of AML cells in co-culture and allows the preservation of LIC activity. A: Data shown represent mean (±SEM) percent control untreated cells count after 1-week treatment with Ara-C 3μΜ in co-culture with MS-5+3GT at 20% 0 ₂ (black bars) or 3% 0 ₂ (white bars) determined for one CB lin ^" sample and 4 AML samples (#2, #3, #6 and #10) . B: Replating potential after Ara-C treatment. AML#6 and #2 cells treated during the first week with Ara-C 3μΜ in co-culture at 20% 0 ₂ or 3% 0 ₂ were recovered after 3 weeks and replated in limiting dilution analysis for 5 additional weeks. C: Total input uncultured and untreated LTC (grey bars) and 3 weeks co-cultured and treated with Ara-C at 20% 0 ₂ (black bars) or 3% 0 ₂ (white bars) output LTC-IC per 10 ⁵ starting cells from samples #6 and #2. Ara-C treatment occurs during the first week. Uncultured input LTC was determined at week-5 at 20% 0 ₂ . For co-cultured treated cells, output live cells were sorted at week-3 and CD45+ cells were replated in limiting dilution analysis for 5 additional weeks at 20% 0 ₂ . Data presented take into account first culture fold expansion. ***p<0.001; **p<0.01; *p<0.05; D: AML #1 cells were co-culture with MS-5+3GT (MS-5) or cultured in suspension (NO MS-5) in presence of Ara-C 3μΜ (ARA) in 3% 0 ₂ . Recovered live sorted hCD45+ cells at 3 weeks were injected into NSG mice (n=8). 5xl0 ⁴ cells were injected which represented 17% of the input cell number for MS-5 condition and 23% for the NO MS-5 condition. Data shown represent percent human leukemic engraftment in recipient bone marrow determined by FACS 12-weeks after injection. Plain lines represent mean levels of leukemic cell engraftment in each group. Dashed lines represent the threshold of positivity. Each symbol represents a single transplanted mouse. **p<0.005. Figure 11 shows that chemoresistant LIC maintenance can be assessed through in vitro replating potential. A: Histograms represent the normalized control untreated (CT, white bars) percentage of viable CD45+ cells count at 3-weeks (±SEM, n=3) for four AML samples (#3, #2, #1 and #6) co-cultured with MS-5+3GT at 3% 02 with Ara-C 3μΜ (ARA-C, black bars) . Ara-C treatment occurred during the first week only. B: CD34 and CD38 expression profile determined by FACS for CB lin- cells (n=4) and AML samples (N=5: #1, #2, #6, #7 and #10) in treated or untreated conditions. Data shown represent phenotypically defined HSC and progenitors and functionally defined LIC and non-LIC normalized frequency in Ara-C condition (black bars, +SEM) as compared to untreated condition (white bars) . Refer to the method section for population definition. NS p>0.05 C: Xenograft potential after ex vivo chemotherapy. A single cell dose (20,000 cells, #3 #2 and #6) of sorted CD45+ cells at 3-weeks was injected into NSG mice. Scatter plots represent leukemic engraftment in recipient bone marrow 12-weeks after injection with untreated cells (open symbol) or in vitro Ara-C treated cells (closed symbol) . Each symbol represents a single transplanted mouse. ***p<0.001; NS p>0.05. D: Percent of control of secondary LTC-IC frequencies. Untreated (white bars) or in vitro Ara-c treated (black bars) samples #3, #2 and #6 CD45+ sorted cells at 3 weeks were replated in Limiting dilution analysis (LDA) for 5 additional weeks. E: In vivo Limiting dilution Analysis: Untreated (open symbol) or in vitro Ara-c treated (closed symbol) samples #2 cells were injected into NSG mice (n= 22 mice) . F: Total input uncultured and output cultured SL-IC for 4x107 cells of samples #2.

***p<0.001. G: In vivo Ara-c treatment: At week 11, after establishment of samples #6 leukemia in NSG mice, recipients were treated with Ara-C 10 mg/kg for 7 days. Data show the absolute number of leukemic cells in the recipient mice treated with or without Ara-C (*p<0.05). H: Data shown represent leukemic engraftment in recipient bone marrow 12-weeks after secondary xenotransplantation. Indicated cell doses of sorted samples #6 cells recovered from primary recipient were injected into secondary NSG mice. NS p>0.05.

Figure 12 shows that AML immunodeficient mice engrafter samples (E) yield a higher number of output cells and are enriched in L- LTC-IC as compared to non-engrafter samples (NE) . (A) E versus NE in vitro expansion. Data shows live leukemic cell count determined by FACS after 5 weeks for 15 E and 13 NE co-cultured with MS-5 + 3GT at 20% oxygen. The horizontal line marked

"input" on the scatter plot indicates the initial cell input number seeded on day 0 of the co-culture. The data shown in the table at the bottom of the figure is the mean ± SEM fold expansion determined for each group. P=0.0277 determined by unpaired Student's t-test. (B) E versus NE L-LTC-IC frequency. 13 E and 12 NE were co-cultured with MS-5+3GT at 20% 0 ₂ in limiting dilution analysis for 5 weeks. After 5 weeks, LTC medium was replaced by methylcellulose and 2 weeks later, each well was scored as negative if no colonies were present. The data shown on the histogram and in the table at the bottom of the figure is the meanlSD L-LTC-IC frequency.

Figure 13 shows that the direct use of amino-reactive fluorescent probes for tracking AML cell division is complicated due to the lower cell viability and expansion after 1 week as compared to a normal hematopoietic sample. (A) Amino-reactive fluorescent probe cell division tracking principle: the amino-reactive fluorescent probe readily crosses intact cell membranes and crosslinks to intracellular proteins. Cell division can be measured as successive halving of the fluorescence intensity of the fluorescent probe. The assay can be used to separately quantify the number of cells in the culture that have not divided vs. those that have divided a specified number of times. CFSE = carboxyfluorescein diacetate, succinimidyl ester. (B)

Heterogeneous viability and proliferation of AML cells as compared to CB lin- cells. Bl: Data shown represent percent viability determined by Annexin-V and DAPI negative cells proportion after 1 week of culture for 16 CB lin- and 59 AML samples. B2 : Data shown represent fold cell expansion determined by precise cell counts using absolute counting beads after 1 week culture for 8 CB lin- and 28 AML samples . **p<0.01 determined by unpaired Student's t-test.

Figure 14 shows that the AML starting fluorescent probe intensity is over-estimated due to apoptosis. Annexin-V staining is essential for determining cell division related fluorescence intensity reliably. (A) Annexin-V staining refines starting CFSE intensity quantification. Al : gating strategy: Doublets and DAPI positive cells were excluded and debris eliminated by forward and side light scatter (i-ii-iii) to define R3 population. Then Annexin V was used (iv) to discriminate non apoptotic (R4) from apoptotic cells (R5). Representative FACS plot for CFSE

fluorescence intensity for one AML sample within gate R3 R4 and R5 (v-vi-vii) . A2 : AML CFSE intensity is over-estimated due to apoptosis. CFSE mean fluorescence intensity (MFI) was determined for gate R3 R4 and R5 for 14 AML samples 18 hours after staining. Data shown represent CFSE MFI in R4 (annexin V negative) and in R5 (Annexin-V positive) normalised to CFSE MFI in superior gate R3. * p<0.05 and ***p<0.001 determined by unpaired Student's t- test. B: Schematic representation of the use of Annexin-V/DAPI positive cells exclusion (grey bar) to focus the analysis on live dividing/non dividing cells (black bar) . Figure 15 shows that Fluorescence Dilution Factor (FDF) allows a low resolution proliferation quantification of AML cell

prolileration in one week. Al : Initial variance in CFSE fluorescence distribution for AML cells impedes high resolution generational computation modelling, (i-iv) Heterogeneous AML blast scatter is responsible for initial variance in CFSE fluorescence and cannot be bypassed through cell sorting without elimination of a morphologically specific sub-population.

Representative FACS double scatter dot plot of an AML sample (i) where a low scatter (R3A thick line) and a high scatter (R3B dashed line) gates are applied, (ii) CFSE fluorescence intensity histogram plot for R3A gated cells (thick line and arrow) and R3B gated cells (dashed line and arrow) 18h after CFSE loading, (iii) Cord Blood representative peak of division where computational model is applicable, (iv) AML representative of 14% of the samples where peak of division are still noticeable and

computation model is still applicable. A2 : variance on the fluorescence distribution within each generation. Data shown represent the coefficient of variation (CVF) for the CFSE fluorescence peaks determined with the FlowJo Proliferation Tool for 45 CB lin- samples and 27 AML samples co-cultured for 1-week with MS-5+3GT at 20% 0 ₂ . ****p<0.0001 B: Schematic

representation of the fluorescence dilution factor (FDF) defined as the ratio of initial CFSE MFI 18h after staining to the output 1-week CFSE MFI in living (Annexin-V/DAPI neg) AML cells.

Figure 16 shows the distinctive and predictive behaviour of NOD/SCID engrafting (E) and non-engrafting (NE) AML samples in vitro. A: Data shown represent Fluorescence Dilution Factor (FDF) determined for E and NE samples co-cultured for 1-week with MS-5 (E, n=19 ; NE, n=14) or with SaOS-2 (E, n=20 ; NE, n=14) or with HUVEC (E, n=13 ; NE, n=5) . E are more proliferative than NE independently of the co-culture system used ***p<0.001, **p<0.01, *p<0.05 for MS-5 SaOS-2 and HUVEC respectively. B: FDF and viability regression analysis for E and NE samples co-cultured for 1-week with MS-5 and/or with SaOS-2 and/or with HUVEC. The 95% confidence band is displayed (dashed curve) . C: A training data set was generated from (B) to derive a computational modelling for predicting E and NE status for an unknown sample. The data shows the observed False Discovery Rate (FDR) of predictions against the percentage of samples for which no prediction is made: e.g. as displayed by the dashed line in (C) , if no prediction is made for 25% of samples, then, 15% of the predictions will be false for those samples for which a

prediction is made. Figure 17 shows the overall survival (OS) of NOD/SCID engrafting and non-engrafting (A) or low proliferating and high

proliferating (B) AML samples. (A and B) The overall survival data of 30 de novo AML cases (<60 years old) that received intensive multi-agent chemotherapy. Allografted patients in first complete remission were censored. (A) AML MNCs (Ιθ') CD3+ depleted were injected into each NOD/SCID and/or B2m-/-N0D/SCID and/or N0D/SCID- m ^"/_ mice . A mouse was scored as engrafted if a distinct single human CD45 ⁺ CD33 ⁺ CD19 ^~ and murine CD45 ^~ population with a cut-off of 0.1% was detectable at 8 to 12 weeks in the recipient bone marrow. NOD/SCID engrafting AML cases had a poor overall survival that was statistically lower than NOD/SCID non- engrafting AML cases. P=0.0071 in a Mantel-Cox log rank test. (B) Fluorescence Dilution Factor (FDF) was determined for 39 de novo AML cases (<60 years old) co-cultured for 1-week with MS-5 or SaOS-2 or HUVEC . Mean FDF (FDFm) value was 2.23 with 19 samples above and 13 below that value independently of the co- culture system used. 6 samples (15%) were above or below FDFm depending on the feeder used. Those samples were censored. Data shown represent OS for the 19 high proliferating (FDF>2.23) and the 13 low proliferating (FDF<2.23) AML samples. P=0.00258 in a Mantel-Cox log rank test.

DETAILED DESCRIPTION The present invention is based on the finding that a low concentration of oxygen is able to support the long term in vitro or ex vivo culture and maintenance of leukemia initiating cells (LICs) , preferably when co-cultured with stromal cells in the presence of medium supplemented with cytokines. Furthermore, whereas the original distribution of LICs within the leukemic population is diluted at higher oxygen concentrations (e.g.

normoxic conventional culture at 20% per volume) , low

concentrations of oxygen (i.e. 0.1 to 6.8% per volume) maintain the original LIC distribution within the leukemic population.

In a first aspect, the invention provides a method for culturing leukemia initiating cells (LICs) in vitro, the method comprising culturing the LICs at an oxygen concentration of 6.8% per volume or lower.

The method represents an efficient culture system for maintaining and quantifying LICs in vitro. These culture conditions allow LICs to be maintained in vitro for at least three weeks, i.e. they allow the long term culture of LICs. In addition, the drug- resistance, e.g. chemoresistance, of these cells may be studied using the method of the invention. The method of the invention also allows for the high-throughput screening of candidate agents, which is not possible when using the xenotransplantation assay (the gold standard model for studying normal human HSCs and LICs (Anj os-Afonso and Bonnet, 2008; Bonnet, 2008; Bonnet, 2009a; Bonnet, 2009b) ) .

Normal HSCs may also be maintained in vitro long term (e.g. for at least 3 weeks) using the method of the invention.

The LICs are preferably mammalian in origin and are most preferably human LICs.

LICs are responsible for maintaining leukemia (Bonnet and Dick, 1997; Lapidot et al . , 1994) and have the ability to recapitulate leukemia and to self-renew. They are typically chemoresistant due to intrinsic and extrinsic factors: efficient DNA repair pathways, high level of drug efflux pumps and quiescence account for intrinsic drug resistance, whereas environment-mediated drug resistance (EM-DR) arises from a complex concomitance of soluble factor drug resistance (SFM-DR) and cell adhesion-mediated drug resistance (CAM-DR) (Damiano et al., 1999; Meads et al., 2008). LICs are derived from HSCs and are synonymous with leukemic stem cells (LSCs) and these terms can define the same entity and can be used interchangeably.

LICs are cells that are capable of maintaining leukemia. They have the capacity to self-renew and to differentiate into leukemic blasts, which re-initiate the leukemic cell diversity seen in leukemia patients. LICs are capable of re-initiating leukemia when transplanted into a xenotransplant model (e.g. NOD/SCID immunodeficient mice) . LICs identified in this way can also be called SCID-leukemia initiating cells (SL-IC) because of the type of mice used in the test. The frequency of LICs or L- LTC-ICs in a test sample may be evaluated by, for example, limiting dilution analysis. The frequency of LICs or L-LTC-ICs capable of engrafting mice in a test sample may be determined by injecting at least 4 different doses of cells into a minimum of four mice per dose. After 10-12 weeks, the number of positive mice are scored, i.e. the mice that have detectable CD33+CD19- leukemic engraftment. The frequency of LICs is calculated using extreme limiting dilution analysis software (available from the Bioinformatics section of the Walter and Eliza Hall Institute of Medical Research, http : //bioinf .

wehi.edu.au/software/elda/index.html; see also Hu and Smyth, 2009) .

Once LICs have been cultured in vitro for at least one week and are capable of re-initiating secondary culture and ultimately giving rise to leukemic colony forming cells (L-CFC) progeny, they may be referred to as leukemia long term culture initiating cells (L-LTC-ICs). LICs that have been cultured in vitro for at least five weeks may also be referred to as L-LTC-ICs.

LICs may be enriched using various antigens, such as CD34, CD38, TIM.l, CD123, CD33, CLL.l, CD44, CD133, CD117, CD90 and/or CD45RA. However, none of these markers is universally expressed in LICs and the phenotype of the LICs depends on the patient sample from which there were obtained. No marker that is absolutely specific for LICs has yet been identified, i.e. no marker that is solely expressed by LICs and not by non-LICs, such as HSCs, has been identified. Identification of LICs using phenotypically defined cell populations is therefore unreliable and instead requires functional confirmation of activity, e.g. using the xenotransplantation model described above or LTC (see Figure 1). The experimental results provided herein identify an in vitro functional test for LICs which mimics their

transplantation potential in xenograft assays.

LICs may be obtained from leukemia patient samples, e.g.

peripheral blood samples, leukapheresis samples or samples obtained by bone marrow aspiration. Samples may be obtained from leukemia patients at diagnosis or during different stages of the disease evolution and clinical care.

In the method of the invention, the LICs are cultured at an oxygen concentration of about 6.8% per volume or lower.

Typically, the oxygen concentration is measured at 37°C in an incubator. During in vitro cell culture, the oxygen

concentration may be varied using a sensor in the incubator that allows the level of oxygen being distributed to the cells to be selected. Hypoxia is considered to be between 0.1 to 3% oxygen per volume. A level of 6.8% oxygen per volume is the mean concentration of oxygen that has been measured directly in the bone marrow of samples from leukemia patients. However, HSCs may be retained in areas of the bone marrow where the oxygen concentration is less than 3% per volume. Re-creating this hypoxic environment in vitro is therefore a genuine feature of the "niche" in which these cells exist in vivo.

In the method of the invention, the LICs may be cultured at an oxygen concentration of about 6.8% per volume or lower, about 6% per volume or lower, about 5% per volume or lower, about 4% per volume or lower, or about 3% per volume or lower. For example, the oxygen concentration may be between about 0.1% per volume and about 6.8% per volume, between about 0.1% per volume and about 6% per volume, between about 0.1% per volume and about 5% per volume, between about 0.1% per volume and about 4% per volume, between about 0.1% per volume and about 3% per volume, between about 0.5% per volume and about 6.8% per volume, between about 0.5% per volume and about 6% per volume, between about 0.5% per volume and about 5% per volume, between about 0.5% per volume and about 4% per volume, between about 0.5% per volume and about 3% per volume, between about 1% per volume and about 6.8% per volume, between about 1% per volume and about 6% per volume, between about 1% per volume and about 5% per volume, between about 1% per volume and about 4% per volume, between about 1% per volume and about 3% per volume, between about 1.5% per volume and about 6.8% per volume, between about 1.5% per volume and about 6% per volume, between about 1.5% per volume and about 5% per volume, between about 1.5% per volume and about 4% per volume, between about 1.5% per volume and about 3% per volume, between about 2% per volume and about 6.8% per volume, between about 2% per volume and about 6% per volume, between about 2% per volume and about 5% per volume, between about 2% per volume and about 4% per volume, between about 2% per volume and about 3.5% per volume, between about 2% per volume and about 3% per volume, between about 2.5% per volume and about 6.8% per volume, between about 2.5% per volume and about 6% per volume, between about 2.5% per volume and about 5% per volume, between about 2.5% per volume and about 4% per volume, between about 2.5% per volume and about 3% per volume. In each of these ranges, the end points are included within the range. For example, a range of 0.1% - 3% per volume includes the values 0.1% per volume and 3% per volume.

Maintaining the oxygen concentration at a specified level during culture does not preclude temporary re-oxygenation up to an oxygen concentration of 20% per volume when the medium is changed . Culturing the LICs at low oxygen concentrations (i.e. 6.8% per volume or lower) allows long term maintenance of LICs in vitro (e.g. for at least 3 weeks). In the method of the invention, the LICs may be co-cultured with stromal cells.

Stromal cells are cells that are able to adhere to tissue culture dishes or flaks and are able to form a monolayer. These stromal cells are able to provide a feeder layer for the culture of other cell types. Alternatively, stromal cells may be encapsulated in three-dimensional (3D) semi-solid media through standard methods known in the art. Semi-solid media may be formed from natural molecules and/or synthetic polymers such as fibronectin, collagen type I, II, IV, laminin, methylcellulose, a Matrigel based matrix, or an alginate based hydrogel .

Examples of suitable stromal cells for use in the method of the invention include stromal cells from adult or embryonic tissues or cell lines. The stromal cells may be mesenchymal cells, such as mesenchymal stem cells (MSC) , pre-osteoblasts, osteoblasts, chondrocytes or endothelial cells. For example, the stromal cells may be primary MSC, primary pre-osteoblasts, primary osteoblasts, primary chondrocytes or primary endothelial cells. These types of stromal cells may be directly isolated from adult or embryonic tissues. Alternatively, pre-osteoblasts,

osteoblasts, chondrocytes and endothelial cells, may be derived from MSC by differentiation or by other methods known to one skilled in the art. Primary mesenchymal cells may be pericytes or mesenchymal cells derived from endothelial-to-mesenchymal transition. Primary endothelial cells may be bone marrow microvascular endothelial cells (BMEC) , endothelial progenitor cells (EPC) , umbilical vein endothelial cells (UVEC) . Stromal cells may be of mammalian origin (the mammal may be a mouse or human). Preferably the stromal cells are mesenchymal cells, e.g. MS-5 cells. Stromal cells may be cell lines, for example, the 2012 cell line, the AC6.21 cell line, the AFT024 cell line, the AGM-S3 cell line, the CAL-72 cell line, the FLS4.1 cell line, the FS-1 cell line, the HAS303 cell line, the HBMEC-33 cell line, the HBMEC-60 cell line, the HBMEC-28 cell line, the HCB1-SV40 cell line, the HESS-5 cell line, HM1-SV40 cell line, the HM2-SV40 cell line, the HS27a cell line, the HS5 cell line, the HYMEQ-5 cell line, the KM-101 cell line, the KM-102 cell line, the KM-103 cell line, the KM-104 cell line, the K -105 cell line, the L87/4 cell line, the M210-B4 cell line, the MRL104.8a cell line, the MS-5 cell line, the OP9 cell line, the PA6 cell line, the PK-2 cell line, the PU-34 cell line, the S10 cell line, the S17 cell line, the Saka cell line, the SAOS-2 cell line, the SCLl-24 cell line, the SC-MSC cell line, the SPY3-2 cell line, the SR-4987 cell line, the ST-1 cell line, the ST2 cell line, the STX3 cell line, or the TBR59 cell line. (Cicuttini et al . , 1992; Collins and Dorshkind, 1987; Fogh et al . , 1977; Hardy et al., 1987; Harigaya and Handa, 1985; Itoh et al . , 1989; Kodama et al., 1994; Kodama et al., 1982; Lemoine et al . , 1988; Miranda et al . , 1988; Ni and O'Neill, 1997; Obinata et al . , 1998; Ogata et al., 1989; Paul et al., 1991; Pessina et al . , 1992; Singer et al., 1987; Takahashi et al., 1995; Thalmeier et al., 1994; Tsai et al., 1986;

Tsuchiyama et al., 1995) .

Preferably the stromal cells are mesenchymal cells, e.g. MS-5 cells. These cells create a supportive stromal layer that helps to mimic part of the bone marrow environment ex vivo and helps to support long term (i.e. one week or more) culture of LICs.

Preferably, the stromal cells form a confluent monolayer. More than one type of stromal cell may be used. For example, different types of stromal cells may be mixed together to form a confluent monolayer. Alternatively, one type of stromal cell may adhere to the culture dish or flask, while a different type of stromal cell may be set down over the top of the adherent layer in a semi-solid medium.

The method may comprise culturing the LICs in the presence of undefined medium, for example medium supplemented with mammalian serum. The serum may, for example, be of human, bovine and/or equine origin. For example, the LICs may be cultured in the presence of about 12.5% fetal calf serum and about 12.5% horse serum. In another embodiment, the method may comprise culturing the LICs using conditioned media harvested from cultured primary normal or cancerous cells or cell lines. "Conditioned medium" is obtained culturing cells in a medium such that the cells release factors into the medium changing its composition. The medium is then collected as "conditioned medium".

In the method of the invention, the LICs may be cultured in defined or undefined medium complemented with an extracellular matrix protein such as collagen (e.g. collagen I, III, IV, V, VI, VIII and XVIII), fibronectin, and/or various proteoglycans known to those skilled in the art that can mimic the bone marrow extracellular matrix, such as chondroitin sulphate proteoglycan 1, versican and heparan sulphate proteoglycan 2.

The method of the invention may comprise culturing the LICs in medium supplemented with one or more cytokines. For example, the method may comprise culturing the LICs in a defined or an undefined medium, such as serum free medium, supplemented with cytokines . The cytokines used in the method of the invention may, for example, promote myeloid activation, such as interleukin-3 (IL- 3), granulocyte colony-stimulating factor (G-CSF) or

thrombopoietin (TPO) , or they may promote endothelial support and/or activation, such as vascular endothelial growth factor (VEGF) or interleukin-1 (IL-1) . Preferably, the medium is supplemented with cytokines IL-3, G-CSF and TPO ("3GT") .

Cytokines used in the method of the invention may, for example, be from the interferon family, the interleukin family, the tumor necrosis factor family, the colony stimulating factors family, or the transforming growth factor family. The method of the invention may comprise culturing the LICs in medium supplemented with a glycoprotein hormone or a growth factor, for example Interferon-alpha, Interferon-beta, IL-1, IL-2, IL-3, IL-4, IL-5, IL-6, IL-7, IL-8, IL-9, IL-10, IL-11, IL-12, IL-13, IL-17, IL-18, IL-23, CCL1, CCL2, CCL3 , CCL4, CCL5 , CCL6, CCL7, CCL8 ,

CCL9/CCL10, CCLll, CCL12, CCL13 , CCL14 , CCL15 , CCLl 6 , CCL17 , CCL18, CCL19, CCL20, CCL21, CCL22, CCL23, CCL24 , CCL25 , CCL26, CCL27, CCL28, CXCL1, CXCL2,CXCL3, CXCL4, CXCL5, CXCL6,CXCL7, CXCL8, CXCL9, CXCL10, CXCL11, CXCL12, CXCL13 , CXCL14 , CXCL15 , CXCL16 , CXCL17, XCLl, XCL2 , CX3CL1, TNF-a, CSF1, CSF2 , CSF3 , TGF p-l,TGF β-2, TGF β-3, Erythropoietin, Thrombopoietin,

Angiopoietin-1, Angiopoietin-2 , Angiopoietin-3 , Angiopoietin-4 , and Bone morphogenetic proteins (BMPs) such as BMP2, BMP3, BMP5, BMP7, and BMP8a. The growth factors used may be from the epidermal growth factor family (EGF) , the fibroblast growth factor family (FGF) , the Insulin-like growth factor family (IGF) , the platelet-derived growth factor family (PDGF) , or the vascular endothelial growth factor family (VEGF) . Preferably, the cytokines are IL-3, G-CSF and TPO ("3GT") . The concentration of each cytokine used may, for example, be at least 1 ng/ml, at least 5 ng/ml or at least 10 ng/ml.

Alternatively, the concentration of each cytokine used may, for example, be less than 1 g/ml, less than 500 ng/ml, less than 200 ng/ml, less than 100 ng/ml, or less than 50 ng/ml. Preferably, the concentration of each cytokine used is between 10 and 500 ng/ml. For example, the concentration of each cytokine may be about 20 ng/ml. In a preferred embodiment, the method comprises culturing LICs in the presence of the cytokines IL-3, G-CSF and TPO ("3GT") , with each cytokine present at a concentration of about 20 ng/ml.

In the method of the invention, the LICs are preferably co- cultured with MS-5 cells in the presence of the cytokines IL-3, G-CSF and TPO ("3GT") .

The LICs cultured in the method of the present invention may be drug-resistant. This means that they are resistant to the action of a specific therapeutic agent. For example, they may be refractory to treatment with a particular chemotherapeutic agent, such as cytarabine (Ara-C) , i.e. they may be chemoresistant .

This drug- or chemo-resistance may be due to either intrinsic factors or extrinsic factors or both. Intrinsic factors responsible for drug- or chemo-resistance include efficient DNA repair pathways, high levels of drug efflux pumps and quiescence. Environment-mediated drug resistance (EM-DR) arises from a complex concomitance of soluble factors (SFM-DR) and cell adhesion-mediated drug resistant mechanisms (CAM-DR) (Damiano et al . , 1999; Meads et al., 2008). Drug-resistant, e.g.

chemoresistant, LICs are responsible for AML relapse and represent the target for future innovative therapies (Zhou et al . , 2009). Therefore, being able to culture and maintain drug- resistant, e.g. chemoresistant, LICs in vitro represents an important advance for screening for candidate therapeutic agents that may be effective against drug-resistant, e.g.

chemoresistant, LICs. The results presented herein show that the method of the invention is capable of supporting functionally defined drug-resistant, e.g. chemoresistant LICs, as efficiently as in vivo. This is in contrast to conventional culture systems that are unable to maintain drug-resistant, e.g. chemoresistant, LICs ex vivo.

The LICs may have been obtained from a patient with leukemia, such as acute myeloid leukemia (AML) , myelodysplastic syndrome (MDS) or chronic myeloid leukemia (CML) . These cells may be obtained from peripheral blood, bone marrow or leukapheresis samples taken from leukemia patients.

The method of the invention allows the long term culture and maintenance of LICs in vitro or ex vivo. The LICs may be cultured according to the method of the invention for at least 1 week, at least 2 weeks, at least 3 weeks, at least 4 weeks or at least 5 weeks, at least 6 weeks, at least 7 weeks, at least 8 weeks, at least 9 weeks, at least 10 weeks, at least 11 weeks, at least 12 weeks, 13 weeks, at least 14 weeks, at least 15 weeks, at least 16 weeks, at least 17 weeks, at least 18 weeks, at least 19 weeks, at least 20 weeks, at least 21 weeks, at least 22 weeks, at least 23 weeks, or at least 24 weeks. The experimental results presented herein demonstrate that functionally defined LICs may be maintained in vitro for at least three weeks using the method of the invention.

Once LICs have been cultured according to the method of the invention for at least one week and are capable of re-initiating secondary culture and ultimately giving rise to leukemic colony forming cells (L-CFC) progeny, they may be referred to as leukemia long term culture initiating cells (L-LTC-ICs) . LICs that have been cultured according to the method of the invention for at least five weeks may also be referred to as L-LTC-ICs.

The method of the invention screening may further comprise the step of re-plating the LICs and/or L-LTC-ICs onto fresh stromal cells. This can be used as an in vitro surrogate bioassay to predict the sensitivity of LICs towards candidate therapeutic agents . The stromal cells used in this secondary culture may be the same or different from the stromal cells used in the first culture. The secondary culture may use the same defined or undefined media and cytokine supplementation as the first culture, or different media and cytokine supplementation.

Similarly, the secondary culture may be performed using the same oxygen concentration as the first culture or a different

concentration, e.g. about 20% oxygen per volume (i.e. normoxia) . Preferably, the secondary culture of re-plated LICs and/or L-LTC- ICs may be performed using conditions that allow clonogenic potential quantification of the LICs and/or L-LTC-ICs, e.g. using the colony forming assay. Such conditions include normoxia, i.e. a concentration of about 20% oxygen per volume.

Following re-plating onto fresh stromal cells, L-LTC-ICs may be referred to as secondary L-LTC-ICs. However, if no step of re- plating onto fresh stromal cells is undertaken, the L-LTC-ICs may be referred to as primary L-LTC-ICs.

The method may further include the step of determining the total number or frequency of LICs and/or primary L-LTC-ICs present in the culture before the step of re-plating and/or the total number or frequency of LICs and/or secondary L-LTC-ICs present in the secondary culture following the step of re-plating. The principle of secondary long term culture is the same as primary long term culture. The secondary long term culture may be maintained according to the method of the invention for at least 1 week, at least 2 weeks, at least 3 weeks, at least 4 weeks or at least 5 weeks, at least 6 weeks, at least 7 weeks, at least 8 weeks, at least 9 weeks, at least 10 weeks, at least 11 weeks, at least 12 weeks, 13 weeks, at least 14 weeks, at least 15 weeks, at least 15 weeks, at least 17 weeks, at least 18 weeks, at least 19 weeks, at least 20 weeks, at least 21 weeks, at least 22 weeks, at least 23 weeks, or at least 24 weeks.

The method may further comprise the step of determining the frequency of L-LTC-ICs in primary or secondary culture, e.g.

using a limiting dilution assay (Ploemacher et al., 1989). In this assay, different cell numbers or concentrations from the primary or secondary culture are seeded over the top of feeder cells. Co-culture is maintained for 5 weeks. After 5 weeks, a colony forming assay is performed. This assay has a duration of 2 weeks and is carried out in a semi-solid medium, such as methylcellulose supplemented with cytokines that promote myeloid colony formation, such as stem cell factor (SCF) , IL-3,

interleukin-6 (IL-6), macrophage colony-stimulating factor (M- CSF) , granulocyte macrophage colony-stimulating factor (GM-CSF) , or granulocyte colony-stimulating factor (G-CSF) . Colony forming cells (CFC) are the progeny of stem cells. Thus, the presence of colonies demonstrate the presence of colony forming cells at the start of the colony forming assay, i.e. 2 weeks previously. The presence of colony forming cells indicates the presence of stem cells 5 weeks before. Therefore, this assay involves a "retro- retro-observation" which allows the threshold number of starting cells required to generate a positive response to be determined. The inverse of this threshold number is the L-LTC-IC frequency within the starting population. However, a potential problem with using the colony-forming assay to determine the frequency of L-LTC-ICs is that conditions favouring the maintenance of LICs or HSCs (i.e. low oxygen concentrations and/or the presence of certain types of feeder cells) may reduce the colony forming potential of the cells. Therefore, those conditions which favour the LICs or HSCs maintenance may hamper the CFC quantification which may introduce a bias in the analysis aiming to identify the retrospective presence of LICs or HSCs.

In view of this, in the method of the invention, primary long term culture is performed in conditions favoring the maintenance of LICs (or HSCs) and unfavorable for L-CFC and CFC

quantification, i.e. low concentrations of oxygen. In contrast, secondary long term culture following re-plating may be performed in conditions favouring CFC (e.g. normoxic conditions of 20% oxygen per volume) to allow accurate quantification of L-CFC derived from secondary L-LTC-ICs. This method of secondary long term culture serves as a surrogate assay to the

xenotransplantation model.

The results presented herein demonstrate that long term re- plating potential, as determined by the total number and/or frequency of secondary L-LTC-IC, is a reliable indicator of LIC activity .

Long term maintenance of LICs in vitro and ex vivo, as achieved by the method of the invention, is very important for the development and screening of new drugs, and the method of the invention has utility for screening for agents that may be effective for treating leukemia. Therefore, in a second aspect, the invention provides an in vitro method of screening for a candidate agent for treating leukemia, the method comprising:

(i) culturing a test sample of LICs according to the method described in the first aspect above;

(ii) contacting the test sample of LICs with the candidate agent; and

(iii) determining the number or frequency of LICs and/or L- LTC-ICs present in the test sample before and after step (ii) , wherein a decrease in the frequency or number of LICs and/or L- LTC-ICs following step (ii) indicates that the candidate agent may be effective for treating leukemia.

The test sample may be contacted with a single candidate agent or with a combination of one or more candidate agents. The test sample may be contacted with the different agents simultaneously or sequentially.

As well as being contacted with a single candidate agent or with a combination of one or more candidate agents, the test sample may also be contacted with an agent already known for treating leukemia, such as Azacitidine, Decitabine, Doxorubicine,

Fludarabine, Hydroxyurea, idarubican, mercaptopurine,

Methotrexate, Mitoxantrone, vincristine, imatinib, dasatinib, or AMN107. The test sample may be contacted with the different agents simultaneously or sequentially. The effectiveness of the candidate agent, or combined candidate agents, may be compared to the effectiveness of the agent already known for treating leukemia .

The in vitro method of screening may further comprise the step of re-plating the LICs and/or L-LTC-ICs onto fresh stromal cells after step (ii) . The stromal cells used in this secondary culture may be the same type or a different type from the stromal cells used in the first culture. The secondary culture may use the same defined or undefined media and cytokine supplementation as the first culture, or different media and cytokine supplementation. Similarly, the secondary culture may be performed using the same oxygen concentration as the first culture or a different oxygen concentration, e.g. an oxygen concentration of about 20% per volume (i.e. normoxia) .

Preferably, the secondary culture of re-plated LICs and/or L-LTC- ICs may be performed using conditions that allow clonogenic potential quantification of the LICs and/or L-LTC-ICs. Such conditions are known to a person skilled in the art and include normoxic conditions, e.g. an oxygen concentration of about 20% per volume.

The method may further include the step of determining the number or frequency of LICs and/or primary L-LTC-ICs present in the culture before the step of re-plating and/or the number or frequency of LICs and/or secondary L-LTC-ICs present in the secondary culture following the step of re-plating.

The LICs may be drug-resistant to one or more other therapeutic agents. For example, the LICs may be chemoresistant to one or more chemotherapeutic agents. The results presented herein demonstrate that drug-resistant, e.g. chemoresistant, LICs can be maintained in culture long term (e.g. for at least 3 weeks) using the method of the invention. This screening method of the invention allows for high throughput drug discovery, which is not possible when using a xenograft assay (which is the current gold standard model for studying human normal HSCs and LICs) , as this assay requires a time frame of eight to twenty four weeks before analysing and remains prospectively blind with little ability to monitor progress. In contrast, the screening method of the invention allows routine, high throughput screening of candidate drugs for treating leukemia, while also allowing the progress of the assay to be monitored at intermediate time points .

The test sample of LICs may be obtained from a patient with leukemia, such as acute myeloid leukemia (AML) , myelodysplastic syndrome (MDS) or chronic myeloid leukemia (CML) . These test samples may have been obtained from peripheral blood, bone marrow or leukapheresis samples taken from these leukemia patients. The screening method may include the step of comparing the frequency or number of LICs and/or L-LTC-ICs in a test sample that has been contacted with the candidate agent with the frequency or number of LICs and/or L-LTC-ICs in a control sample of LICs and/or L-LTC-ICs that has not been contacted with the candidate agent. A decrease in frequency or number of LICs and/or L-LTC-ICs in the test sample following treatment compared to the frequency or number of LICs and/or L-LTC-ICs in the control sample indicates that the candidate agent may be effective for treating leukemia.

The screening method may also include the step of culturing a control sample of normal HSCs according to the method described in the first aspect above, contacting the control sample with the candidate agent and determining the number or frequency of normal HSCs present in the control sample before and after treatment with the candidate agent. A decrease in the number or frequency of HSCs indicates that the candidate agent may not be suitable for treating leukemia as it may not be specific for leukemic cells .

The in vitro method of screening may further comprise the step of re-plating the LICs and/or L-LTC-ICs onto fresh stromal cells after step (ii) . The stromal cells used in this secondary culture may be the same type or a different type from the stromal cells used in the first culture. The secondary culture may use the same defined or undefined media and cytokine supplementation as the first culture, or different media and cytokine

supplementation. Similarly, the secondary culture may be performed using the same oxygen concentration as the first culture or a different oxygen concentration. For example, the oxygen concentration may by about 20% per volume (i.e. normoxic conditions) . Preferably, the secondary culture of re-plated LICs and/or L-LTC-ICs may be performed using conditions that allow clonogenic potential quantification of the LICs and/or L-LTC-ICs. Such conditions are known to a person skilled in the art and include normoxia, i.e. an oxygen concentration of about 20% per volume.

The method may further include the step of determining the number or frequency of LICs and/or primary L-LTC-ICs present in the culture before the step of re-plating and/or the total number or frequency of LICs and/or secondary L-LTC-ICs present in the secondary culture following the step of re-plating. A decrease in the number or frequency of LICs and/or secondary L-LTC-ICs following re-plating compared to a control sample that has not been contacted with the candidate agent indicates that the candidate agent may be effective for treating leukemia.

The method of the invention also has utility in identifying or selecting for leukemia patients likely to respond to a candidate agent for treating leukemia.

Therefore, in a third aspect, the invention provides an ex vivo method for identifying a leukemia patient likely to respond to a candidate agent for treating leukemia, the method comprising:

(i) culturing a test sample of LICs obtained from the leukemia patient according to the method described in the first aspect above;

(ii) contacting the test sample with the candidate agent; and

(iii) determining the frequency or number of LICs and/or L- LTC-ICs present in the test sample before and after step (ii) , wherein a decrease in the frequency or number of LICs and/or L- LTC-ICs following step (ii) indicates that the leukemia patient is likely to respond to the candidate agent. The patient may be refractory to treatment with one or more other therapeutic agents, e.g. chemotherapeutic agents, i.e. the patient may not be responding to treatment with one or more other therapeutic agents, e.g. chemotherapeutic agents.

The method for identifying patients likely to respond to the candidate agent may include the step of comparing the frequency or number of LICs and/or L-LTC-ICs in a test sample that has been contacted with the candidate agent with the frequency or number of LICs and/or L-LTC-ICs in a control sample of LICs and/or L- LTC-ICs from the patient that has not been contacted with the candidate agent. A decrease in the frequency or number of LICs and/or L-LTC-ICs in the test sample following treatment compared to the frequency or number of LICs and/or L-LTC-ICs in the control sample indicates that the patient is likely to respond to the candidate agent .

The method for identifying patients likely to respond to the candidate agent may also include the step of re-plating the LICs and/or L-LTC-ICs onto fresh stromal cells after they have been contacted with the candidate agent (i.e. after step (ii) ) . The stromal cells used in this secondary culture may be the same type or a different type from the stromal cells used in the first culture. The secondary culture may use the same defined or undefined media and cytokine supplementation as the first culture, or different media and cytokine supplementation.

Similarly, the secondary culture may be performed using the same oxygen concentration as the first culture or a different oxygen concentration. Preferably, the secondary culture of re-plated LICs and/or L-LTC-ICs may be performed using conditions that allow clonogenic potential quantification of the LICs and/or L- LTC-ICs. Such conditions are known to a person skilled in the art and include normoxic conditions, i.e. an oxygen concentration of about 20% per volume.

Following re-plating, the number or frequency of secondary L-LTC- ICs may be determined. Thus, the method may further include the step of determining the total number or frequency of LICs and/or primary L-LTC-ICs present in the culture before the step of re- plating and/or the total number or frequency of LICs and/or secondary L-LTC-ICs present in the secondary culture following the step of re-plating. A decrease in the number or frequency of LICs and/or secondary L-LTC-ICs following re-plating compared to a control sample that has not been contacted with the candidate agent indicates that the leukemia patient is likely to respond to the candidate agent .

The test sample may be contacted with a single candidate agent or with a combination of one or more candidate agents. The test sample may be contacted with the different agents simultaneously or sequentially.

As well as being contacted with a single candidate agent or with a combination of one or more candidate agents, the test sample may also be contacted with an agent already known for treating leukemia .

The test sample of LICs may be obtained from a patient with leukemia, such as acute myeloid leukemia (AML) , myelodysplastic syndrome (MDS) or chronic myeloid leukemia (CML) . These test samples may have been obtained from peripheral blood, bone marrow or leukapheresis samples taken from these leukemia patients. The method of the invention also has utility in monitoring a leukemia patient for drug-resistance to an agent.

Therefore, in a fourth aspect, the invention provides an ex vivo method for monitoring a leukemia patient for resistance to a therapeutic agent, the method comprising:

(i) culturing a test sample of LICs obtained from the leukemia patient according to the method described in the first aspect above;

(ii) contacting the test sample with the agent; and

(iii) determining the frequency or number of LICs and/or L-

LTC-ICs present in the test sample before and after step (ii) , wherein a decrease in the frequency or number of LICs and/or L- LTC-ICs following step (ii) indicates that the leukemia patient is not resistant to the agent.

The method may further comprise the step of re-plating the LICs and/or L-LTC-ICs onto fresh stromal cells after step (ii) . The stromal cells used in this secondary culture may be the same type or a different type from the stromal cells used in the first culture. The secondary culture may use the same defined or undefined media and cytokine supplementation as the first culture, or different media and cytokine supplementation.

Similarly, the secondary culture may be performed using the same oxygen concentration as the first culture or a different oxygen concentration, e.g. about 20% oxygen per volume (i.e. normoxia) . Preferably, the secondary culture of re-plated LICs and/or L-LTC- ICs may be performed using conditions that allow clonogenic potential quantification of the LICs and/or L-LTC-ICs. Such conditions are known to a person skilled in the art, e.g.

normoxic conditions, i.e. about 20% oxygen per volume. The method may further include the step of determining the number or frequency of LICs and/or primary L-LTC-ICs present in the culture before the step of re-plating and/or the number or frequency of LICs and/or secondary L-LTC-ICs present in the secondary culture following the step of re-plating. A decrease in the number or frequency of LICs and/or secondary L-LTC-ICs following re-plating compared to a control sample that has not been contacted with the agent indicates that the patient is not resistant to the agent. The therapeutic agent may be a chemotherapeutic agent, such as cytarabine (Ara-C) .

The method for monitoring leukemia patients for resistance to a therapeutic agent may include the step of comparing the frequency or number of LICs and/or L-LTC-ICs in a test sample that has been contacted with the agent with the frequency or number of LICs and/or L-LT-LICs in a control sample of LICs obtained from the patient that has not been contacted with the agent. A decrease in frequency or number of LICs and/or L-LTC-ICs in the test sample following treatment compared to the frequency of LICs and/or L-LTC-ICs in the control sample indicates that the patient is not resistant to the agent.

The test sample of LICs may be obtained from a patient with leukemia, such as acute myeloid leukemia (AML) , myelodysplastic syndrome (MDS) or chronic myeloid leukemia (CML) . These test samples may have been obtained from peripheral blood, bone marrow or leukapheresis samples taken from these leukemia patients.

The inventors have shown that the engraftment potential of LICs in the xenotransplantation assay is related to the proliferation rate of the LICs in vitro, with engrafter cells displaying a higher proliferation rate than non-engrafter cells. As

engraftment in this model is predictive of leukemia prognosis (Pearce et al . , 2006), the proliferation rate of LICs in vitro may also be predictive of leukemia patient prognosis.

Therefore, in a fifth aspect, the invention provides an ex vivo method for predicting the prognosis of a leukemia patient, the method comprising:

(i) culturing a test sample of leukemic cells obtained from the leukemia patient in vitro for at least one week; and

(ii) determining the proliferation rate of living leukemic cells in the test sample; wherein the proliferation rate correlates to the prognosis of the patient. Typically, a higher proliferation rate indicates a worse prognosis for the patient than a lower proliferation rate.

In step (i) , the test sample of leukemic cells may be cultured according to the method described in the first aspect above for at least one week. Alternatively, the test sample may be cultured under normoxic conditions, i.e. at an oxygen

concentration of about 20%, during step (i) . The test sample of leukemic cells may be obtained from a patient with any type of leukemia, such as acute myeloid leukemia (AML) , myelodysplastic syndrome (MDS) or chronic myeloid leukemia (CML) . These test samples may have been obtained from peripheral blood, bone marrow or leukapheresis samples taken from these leukemia patients .

Various methods for determining the proliferation rate of cells in culture are known to a person skilled in the art. For example, the method may comprise staining the test sample of leukemic cells with a dye or probe and determining the amount of dye or probe bound to the test sample on the day of initiating culture (DO) . Following in vitro culture of the test sample of leukemic cells for at least one week, the amount of dye or probe bound to the test sample may again be determined. The ratio of the amount of dye or probe bound to the test sample following culture compared to the amount of dye or probe bound to the test sample at DO is indicative of the proliferation rate of the cells. The dye or probe may, for example, be fluorescent. For example, the probe or dye may be carboxyfluorescein diacetate, succinimidyl ester (CFSE) .

In a specific embodiment, the test sample of leukemic cells may be stained with an amine-reactive cell permeable dye or probe the day before initiating the culture (D-l) . The mean fluorescence intensity ( FI) of the bound amino-reactive dye or probe may then be determined on the day culture is initiated (DO) . The MFI of a control sample, e.g. fluorescent beads, may also be determined at DO for normalisation and calibration purposes. The test sample of leukemic cells may then be cultured in vitro for at least 7 days. The MFI of the amino-reactive dye or probe bound to the test sample may then be determined (e.g. after 7 days of culture, i.e. at D7) . The MFI of a control sample, e.g. fluorescent beads, may again be determined at D7 for normalisation and calibration purposes. The fluorescence dilution factor (FDF) may then be determined and this is indicative of the proliferation rate of living leukemic cells in the test sample. FDF = the MFI of the test sample at DO normalised to the MFI of the control sample at DO / the MFI of the test sample at D7 normalised to the MFI of the control sample at D7.

An FDF of between 1 and 2 indicates that most living leukemic cells have not divided during step (i) . An FDF of >2 indicates that most live leukemic cells have divided at least once during step (i) .

Test samples with a viability of ≥30% at D7 and with an FDF of ≥2.56 are predicted to be able to engraft immunodeficient mice, whereas test samples with a viability of ≥30% at D7 and with an FDF of ≤1.7 are predicted to be non-engrafters of immunodeficient mice, with a false discovery rate of 10% respectively. Test samples with a viability of >30% at D7 and with an FDF of ≥2.94 are predicted to be able to engraft immunodeficient mice, whereas test samples with a viability of ≥30% at D7 and with an FDF of ≤1.44 are predicted to be non-engrafters of immunodeficient mice, with a false discovery rate of 5% respectively.

Leukemia patients from which test samples predicted to be able to engraft immunodeficient mice are obtained are predicted to have a worse prognosis than leukemia patients from which test samples of LICs predicted to be non-engrafters are obtained. Therefore, if a sample obtained from a leukemia patient has a high

proliferation rate (e.g. a high FDF value, such as ≥2.56), then this patient is likely to have a worse prognosis (e.g. lower survival rate) than a leukemia patient from which a sample has a lower proliferation rate.

In a sixth aspect, the invention provides an ex vivo method for determining the ability of a therapeutic agent to reduce the proportion of LICs in a leukemia patient, the method comprising:

(i) culturing a tissue sample obtained from a leukemia patient and containing LICs according to the method described in the first aspect above; (ii) determining the proportion of LICs and/or L-LTC-ICs in the sample;

(iii) contacting the sample with the therapeutic agent;

(iv) determining the proportion of LICs and/or L-LTC-ICs in the sample;

wherein a decrease in the proportion of LICs and/or L-LTC-ICs in the sample following step (iii) indicates that therapeutic agent is likely to reduce the proportion of LICs when used to treat the patient .

The tissue sample may be obtained from a patient with leukemia, such as acute myeloid leukemia (AML) , myelodysplastic syndrome (MDS) or chronic myeloid leukemia (CML) . The tissue sample may, for example, be a peripheral blood, bone marrow or leukapheresis sample taken from a leukemia patient.

The therapeutic agent may be a therapeutic agent, such as a chemotherapeutic agent, known for the treatment of leukemia.

Alternatively, the therapeutic agent may be an agent identified in the in vitro screening method of the present invention.

This method allows the effectiveness of the therapeutic agent to be monitored more accurately than by measuring bulk tumour shrinkage, which may understate or overstate the LIC injury (see Figures 2 and 3) .

Further aspects and embodiments of the invention will be apparent to those skilled in the art given the present disclosure including the following experimental exemplification.

EXPERIMENTAL

MATERIALS AND METHODS Cells and co-cultures

Human cord bloods (CB) were obtained after informed consent at the Royal London Hospitals (UK) . Depletion of cells expressing lineage markers was performed using StemSep columns and human progenitor enrichment cocktail (StemCell Technologies, Vancouver, BC, Canada) . AML cells were obtained after informed consent at St Bartholomew's Hospital (London, UK). The protocol was approved by the East London Ethical Committee and in accordance with the Declaration of Helsinki. Samples were collected at diagnosis and screened for their ability to engraft in

immunodeficient mice. Details of the patient samples are listed in Table 3 and sample processing is detailed below. The stromal cell line MS-5 was kindly gifted by Dr John Dick and maintained in IMDM 10% FCS + 2 mM L-glutamine.

Three to five different cord bloods (CB) were pooled and mononuclear cells were obtained by density centrifugation .

Lineage markers expressing cells were depleted using StemSep columns and human progenitor enrichment cocktail (StemCell

Technologies, Vancouver, BC, Canada) . The purity of depleted fractions was assessed to ensure enrichment quality. For LTC experiments, freshly isolated CB lin cells aliquots were frozen and all LTC were performed on the same frozen batch to monitor inter-experiment consistency. AML samples were collected at untreated presentation. AML mononuclear cells were obtained by density centrifugation . Samples screened for their ability to engraft in immunodeficient mice before use in this study after immunomagnetxc T cell depletion using the Easysep T cell enrichment cocktail (StemCell Technologies) . The human

osteosarcoma cell line SaOS-2 was purchased from German

Collection of Microorganisms and Cell Cultures (DSMZ,

http://www.dsmz.de) and maintained in McCoy's 5a with 15% FCS + 2mM L-glutamine. Human umbilical vein endothelial cells (HUVEC cells were obtained from Clonetics) were propagated in

endothelial growth medium-2, EGM-2-MV (Clonetics) in culture dishes coated with type I collagen (from StemCell Technologies) . Feeders were cultured in their respective media and sub-cultured when reaching 80% confluency. Immunophenotyping was performed to determine a unique marker allowing identification of feeder cells in co-culture experiments. Sca-1, CD56 and CD31 were identified as a marker for 100% of MS-5, SaOS-2 and HUVEC respectively. All three antibodies were from BD Pharmingen, Oxford Science Park, UK. For co-culture experiments, 1 week before initiating the co- culture, feeder cells were plated in type-I collagen coated 96- well, 12-well plates (Corning) or 150 mm culture Dishes (Falcon) and allowed to reach confluency. One day before starting co- culture, culture media were removed, wells were washed twice with PBS and fresh Myelocult® LTC medium (from StemCell Technologies) was added. Each feeder viability and proliferation in LTC medium was preliminarily checked (data not shown) . Acceptable HUVEC viability over 3 weeks in LTC medium could only be obtained with VEGF supplementation. As described previously, HUVEC cells require stimulation by IL1 in order to induce bone marrow endothelial cells (B EC) similar properties for hematopoietic support (Jazwiec et al . , 1998; Schweitzer et al., 1996; Yildirim et al., 2005). HUVEC co-cultures were supplemented with recombinant rhu-ILla/VEGFa (10 and 50 ng/ml respectively) . On day 0 of the co-culture, CB lin or sorted HSC or progenitor cells were seeded at 1,500 cells/cm ² onto feeders. AML cells were plated at 80 to 200 10 ³ cells/cm ² . Cells were cultured at 37 °C in 5% C02-humidified incubators at indicated oxygen concentrations. For LTC, a half medium change was done twice a week without disrupting the established feeder monolayer. Since hematopoietic cell adhesions vary from one feeder to another and no precise semi-population could be operated by a simple half medium change a cumulative count was applied to each co-culture.

Short-Term (i.e. 3-7 days) or Long-Term Culture (i.e. 1 to 12 weeks) and secondary LTC

Co-cultures were performed in bulk culture or limiting dilution assay (LDA) on MS-5 confluent monolayer supplemented as indicated with recombinant human-IL3/G-CSF/TP0, i.e. "3GT" (20 ng/ml each Peprotech London, UK) in MyeloCult H5100 (StemCell Technologies, Vancouver, CA) in absence of Hydroxycortisone . Cells were cultured at 37 °C in 5% C0 ₂ humidified incubators at 20% or 3% 0 ₂ conditions. Low oxygen cultures were performed in a two-gas HERAcell® incubator. For Cytosine β-D-arabinofuranoside (Ara-C, Sigma-Aldrich) treatment, CB lin- or AML cells were pre-incubated in co-culture 72h prior to addition of 3 μΜ final of Ara-C. Ara- C treatment occurs for 7 days. In some experiments, co-cultures were washed after treatment 4 times with PBS, fresh MyeloCult H5100 medium was added and co-culture were maintained for 2 additional weeks. For LTC-LDA and secondary LTC-LDA, cells were plated in 20 replicates in 96 well microplates containing confluent S-5 monolayer. After 5 weeks LTC, medium was replaced by methylcellulose H4435 (Stem Cell Technologies) . After 2 weeks, each well was scored as negative if no colonies were present. To determine the LTC Initiating Cells (LTC-IC) frequency, LDA analysis was done using LCalc Software (Stem Cell Technologies) according to the Poisson statistics and method of maximum likelihood.

Flow cytometry analysis

FACS analysis was performed on BD LSRII flow cytometer or BD Biosciences FACS Aria 4 laser cytometers . Subsequent analysis was performed with FlowJo software (Tree Star, Oten,

Switzerland) . Cell sorting was performed on a BD FACS Aria (BD Biosciences, UK) . After 1 or 5 weeks co-culture, non-adherent cells were harvested through 3 gentle washes and adherent cells through trypsinisation. Recovered cells were re-suspended and stained in Annexin binding buffer (BD Biosciences) . Sca-1 was identified as specific markers of 100% of MS-5. Human

hematopoietic cells were stained with anti-CD45-APC-Cy7, anti- CD34-Percp, anti-CD38-PE-Cy7 antibodies, and Lin-FITC (or CFSE or carboxy-H2DFFDA) as well as with AlexaFluor647-conjugated-

Annexin-V (Invitrogen) and 4,6 diamidino-2-phenylindole (DAPI) . All antibodies were obtained from BD Biosciences, UK. Only viable (DAPI and Annexin-V negatives) human hematopoietic cells (CD45-APC-Cy7 positive and Sca-l-PE negative) were assessed and/or sorted for all analysis. For CB, HSC and progenitors frequency were quantified through Lin D34 ⁺ CD38 ^~ and Lin ^" CD34 ⁺ CD38 ⁺ phenotype respectively. For some AML sample, LIC and non-LIC phenotype were first pre-established through xenograft

experiments on sorted sub-population (published (Taussig et al . , 2010) and data not shown) . For precise cell counts, Count

Bright™ absolute counting beads (Molecular Probes Invitrogen, Paisley, UK) were used to assess the total number of cells following manufacturer's recommendations. For intracellular ROS analysis, 5 weeks co-cultures were loaded with 5 μΜ carboxy- difluorodihydrofluorescein diacetate ( carboxy-H2DFFDA) (SIGMA) for 30 min at 37°C before flow cytometry analysis.

Gene expression analysis

RNA was extracted using the RNeasy Micro Kit according to manufacturer's instructions (Qiagen, Hilden, Germany), and treated with DNase. Reverse transcription (RT) was done using the Sensiscript (Qiagen) kit according to the manufacturer's instructions. For quantitative real-time polymerase chain reaction (Q-PCR) , SYBR Green master mix reagent (Applied

BioSystems, Foster City, CA) was used and the amplification was performed using the ABI Prism 7900HT sequence detection system (Applied BioSystems) . To avoid possible amplification of contaminating DNA and unprocessed mRNA, primers were designed to span an exon boundary. The specificity of the PCR products was verified by running a 2% agarose gel and a dissociation curve. The primers used in this study were: CDKN2A (pi 6) F: 5'- GAAGGTCCCTCAGACATCCC-3 ' R: 5' -CCCTGTAGGACCTTCGGTGA-3 ' and GAPDH F: 5' -GGGAAGGTGAAGGTCGGAGT-3 ' R: 5 ' -GGGTCATTGATGGCAACAATA-3 ' .

Real-time quantitative polymerase chain reaction assay for NPM exon 12

DNA was extracted from co-cultures recovered content using a QIAamp DNA mini kit (Qiagen) according to the manufacturer's instructions. Real time quantitative analysis of NPM exon 12 mutations was done using following gene-specific primers: common forward primer: 5 ' GTGTTGTGGTTCCTTAACCACAT3 ' Reverse primer for total NPM1: 5' CTGTTACAGAAATGAAATAAGACGGAAA 3' Reverse primer for NPM1 mutation type 1 (TCTG) : 5' TCCTCCACTGCCAGACAGAG 3' and Reverse primer for NPMl mutation type 4 (CTTG) : 5' TCCTCCACTGCCAAGCAGAG 3' . All samples were tested in triplicate. Standard curves for total NPM and NPM mutated were established by amplifying a 10-fold serial dilution of uncultured sample genomic DNA. A standard curve was created with each run. MS-5 DNA alone was negative or below log -4 level for NPM1 total primers or NPM1 mutation type 1 and 4. Thus the percentage of mutated NPM was determined by dividing the value for NPM mutation by the total NPM value. Percentages greater than 100% were treated as 100%.

Ara-C dose calibration

AML cell lines HEL, HL60, KG1 and MS-5 feeders were seeded in 96- well plates at an initial concentration of 1 * lOVmL for AML cells and 1 ^χ lOVcm ² for MS-5, at a final volume of 200 pL/well, and were exposed to varying concentrations of Ara-C.

Cytotoxicity was assessed using the XTT Cell Proliferation Kit II (Roche Applied Science) after 72 h of drug (or control) exposure. 50 \i of XTT solution were added. After 2-h incubation at 37°C, absorbance at 490 nm was measured using a microplate reader (DYNEX Technologies, Inc.) . The IC50 values were computed using CalcuSyn software (Biosoft) .

Division history tracking

Defrosted CB Lin ^" or AML cells were stained with 0.8 μΜ

carboxyfluorescein diacetate, succinimidyl ester (CFSE)

(Invitrogen, UK) for 10 min at 37 °C in PBS the day before starting the co-culture (D-l) . Washed cells were incubated overnight in serum-free expansion medium StemSpan® SFEM (StemCell Technologies, Vancouver, CA) (Nordon et al., 1997) . Recovered cells were washed twice and stained with anti-CD34 PerCP, anti- CD38 PE-cy7, DAPI and AlexaFluor647-conjugated-Annexin-V

(Invitrogen) . The initial Mean Fluorescence Intensity (MFI) of CFSE bound to the test sample on the day of initiating culture (DO) was determined within the viable cell population (DAPI and Annexin-V negatives) . Fluorescent beads were also used for establishing precise cell count and for calibration. For cell sorting (CB lin-) , a narrow CFSE gate was applied to live CD34 ⁺ CD38 ^low/neg to obtain a tight homogenous CFSE staining. After 1-week (D7) co-culture, all cells were recovered and analyzed by FACS as specified in "Flow cytometry analysis" section. The output D7 CFSE MFI was determinined within viable human cell population as well as their count. The fluorescence dilution factor (FDF) was then calculated as the ratio of MFI of the test sample at DO normalised to the MFI of beads at DO of the MFI of the test sample at D7 normalised to the MFI of beads at D7. Statistics

In vitro data were analyzed using a paired or unpaired Student's t-test (GraphPad QuickCalcs

http://www.graphpad.com/quickcalcs/ttestl.cfm) . For

xenotransplantation analysis, a Mann-Whitney unpaired two tail test was applied. Observed differences were regarded as statistically significant if the calculated two-sided p-value was below 0.05.

Adoptive transfer of human hematopoietic cells in immunodeficient mice

All animal experiments were performed in compliance with Home Office and CRUK guidelines. NOD/SCID/ β2-microglobulin null (NOD/SCID- ₂ nf'' ^" ) and NOD/SCID/interleukin-2 receptor γ-chain null (NSG) mice were originally obtained from Dr Leonard Schultz (Jackson Laboratory, Bar Harbor, ME, USA) and bred at Charles Rivers Laboratories (Margate, United Kingdom) . Mice were irradiated at 375 cGy ( ^13, Cs source) 24 hours before xenograft. Xenografts were performed via intra-tibial injection due to previous reports on HSC homing defect described after in vitro manipulation (Szilvassy et al., 2001). MS-5 were eliminated by cell sorting prior to xenograft. Animals were sacrificed at 12 weeks and bone marrow cells were collected from the injected bone and contralateral non-injected long bones separately. For LDAs, cell doses of 10 ² to 10 ⁷ cells from 8 primary AML samples from were injected into at least 3 NOD/SCID- ₂ nf ^x~ mice per dose. A mouse was scored as engrafted if a distinct human CD45 ⁺ and murine CD45 ^~ population was detectable in the contralateral bone marrow. AML engraftment was defined by the presence of a single CD45 ⁺ CD33 ⁺ CDl9 ^" population.

In vivo chemotherapy.

NSG mice were injected intravenously with 7.10 ^s primary AML samples #6 cells. After demonstrating AML engraftment at 11- weeks through tibial bone marrow aspiration, mice were treated with cytosine arabinoside (Ara-C) given IP daily for 7 days at lOmg/kg daily. 6 hours after last injection, mice were

sacrificed by cervical dislocation. The femurs, tibias, and pelvis were dissected and flushed with PBS. Red blood cells were lysed via ammonium chloride. Cells were stained with human- specific FITC-conjugated anti-CD19, PE-conjugated anti-CD33, APC- Cy7-conjugated anti-CD45 and PERCP-conj ugated anti-murine CD45 antibodies. Dead cells and debris were excluded via DAPI staining. A BD LSR II flow cytometer was used for analysis.

More than 100 000 DAPI-negative events were collected.

Engraftment of AML was said to be present if a single population of mCD45-CD45+CD33+CDl9- cells was present without accompanying mCD45-CD45+CD33-CD19+cells .

RESULTS

Feeder layer, cytokine and oxygen concentration for maintaining normal HSCs in vitro

To define the best culture conditions for maintaining normal hematopoietic stem cells in vitro, mesenchymal related MS-5, osteoblast-derived SaOS-2 and endothelial HUVEC cell lines were first compared for their ability to support umbilical cord blood- derived HSCs/progenitors . No condition clearly surpassed the others after one week of culture in vitro (data not shown) thus, long-term culture (LTC) and colony forming cell (CFC) assays were performed at five weeks without or with addition of cytokines promoting myeloid (IL3, G-CSF and TPO "3GT") or endothelial activation (IL1 and VEGF; "IV") . The concentration of each cytokine added to the medium was 20ng/ml. Umbilical cord blood- derived lineage negative (CB lirf) co-culture with MS-5+3GT resulted in the best yield for total cells, total CFC as well as CFC frequency (Figure 4, A and B) . SaOS-2 cells are known to block CFC production (Taichman et al . , 1997) (Figure 4B) and the LTC read-out through a CFC assay might underestimate the actual maintenance of HSCs on osteoblastic cells. The hematopoietic cells were therefore sorted after 5 weeks of co-culture in the different systems and re-plated in the most permissive MS-5+3GT condition. Although some re-plating potential was observed from the SaOS-2 co-culture system, which was absent at 5 weeks from HUVEC co-cultures, MS-5 was again the most efficient co-culture feeder layer for maintaining secondary LTC-IC, resulting in the highest total number of secondary LTC-IC (Figure 4C) . These results reinforce the notion that MS-5+3GT is the most efficient system, as compared to SaOS-2 or HUVEC, for supporting early hematopoietic progenitors in vitro.

Based on the reported roles of hypoxia in regulating murine and human HSC functions, the impact of low oxygen concentration in the MS-5 co-culture system was assessed. CB lin ^" cells were cultured for 5 weeks with MS-5+3GT at 3% or 20% 0 ₂ . Phenotypic analysis shows a 3.7±1.46 fold increase in the frequency of HSC cells at 3% 0 ₂ compared to 20% 0 ₂ (p<0.05), whereas no difference was observed in the frequency of progenitors cells at 5 weeks (Figure 5A) . CB lin ^" cells cultured at 3% 0 ₂ expanded 18.7-fold less (Figure 5B) and derived 261 times fewer colonies, compared to 20% 0 ₂ (Figure 5C and Table 1, p<0.01). Moreover, a >90-fold reduction of primary LTC-ICs was observed at 3% 0 ₂ compared to 20% 0 ₂ (Figure 5D and Table 1, p<0.0001). To determine whether the dramatic reduction of CFC at 3% 0 ₂ represented HSC

extinction, we re-sorted the hematopoietic cells at week 5 and re-plated them on MS-5+3GT for an additional 5 weeks at 20% 0 ₂ . Interestingly, in secondary plating, total LTC-IC number increased from first plating under the 3% 0 ₂ condition whereas their number remain constant under 20% 0 ₂ (Figure 5D, E and Table 1) . Secondary LTC-IC were also found to be 2.8 times more frequent at 3% 0 ₂ than at 20% 0 ₂ (p<0.05) (Figure 5F, and Table 1) . Next HSC maintenance was investigated through the SRC assay (in vivo xenograft model) . When an equivalent number of cells were injected, those co-cultured at 3% 0 ₂ for 5 weeks engrafted at a higher level compared to cells maintained at 20% 0 ₂

(p<0.001) (Figure 5G) . Quantifying the SRC frequency, we confirmed also 2.8 fold increase in output SRC frequency for cells co-cultured at 3% 0 ₂ (1 SRC per 0.77 xlO ⁶ injected output cells) compared to those co-cultured at 20% 0 ₂ (1 per 2.20 xl0 ^e cells) (p<0.05) (Figure 5H and Table 1). When quantifying the total output number of SRC, we did not observe a significant difference between the two culture conditions and based on the initial input SRC number, this number is maintained in both conditions (Figure 51). Nevertheless, with respect to the multilineage engraftment potential, we observed that SRCs cultured in normoxia had a defect in lymphoid lineage engraftment (Figure 6A) . A 4.2-fold increase in ROS generation was also observed in the cells cultured for 5-weeks at 20% compared to 3% 0 ₂ (p<0.001) (Figure 6 Bl) . Furthermore, only 1.3+0.2% of ROS negative cells were observed after culture at 20% 0 ₂ , as compared to 21.510.2 % when cultured at 3% 0 ₂ (p<0.01) (Figure 6, B2 and B3) . This increase in ROS production was also concomitant with a 3.57-fold increase in pl6INK4a transcript (p<0.01) (Figure 6C) . Altogether, we were able to conclude that per cell basis, normal human HSCs were better maintained at 3% than at 20% 0 ₂ . AML-IC maintenance in vitro over 3 weeks.

AML is a heterogeneous disease and we first aimed to identify the co-culture system allowing the best viability of a wide number of samples. Between 38 to 57 AML samples were compared for their viability after being co-cultured for one week with MS-5, SaOS-2 or HUVEC cells without cytokine supplements. We observed a highly variable level of viability, ranging from 3.23% to 96.4%, and thus a comparable mean viability across the three systems (data not shown) . However, in a cross comparison analysis per sample, we observed that MS-5 cells were the most supportive co- culture system (Figure 7, Al-3, MS-5>SaOS-2>HUVEC, p<0.05 and p<0.0001 respectively, in a paired t-test) . Interestingly, the better viability achieved with MS-5 co-cultures did not correlate with a better cell count output. After 1-week In MS-5 condition we quantified a mean fold expansion of 1.22±0.32 (n=28) , suggesting the maintenance of the cell population rather than an expansion (Figure 7, Bl-3) . We next focused on the MS-5 feeder as it appeared the most efficient system for supporting normal and leukemic cells. The impact of the addition of 3GT cytokine cocktail was next investigated. 20 ng/ml of each of the components of the 3GT cocktail (i.e. IL-3, G-CSF or TPO) was added. After 3 days in suspension culture, 3GT did not increase the viability of AML cells, however MS-5+3GT increased it by 18.8+5.7%, compared to sample viability after thawing (data not shown, n=23, p<0.005) . AML long-term co-cultures performed without or with 3GT also demonstrated a requirement of cytokine supplement for the maintenance of leukemic cells over a 5-week period (Figure 7C) . In all cases, we confirmed the leukemic origin of the cells after culture (an example is shown in Figure 7D) . We then investigated the impact of 3% 0 ₂ condition on AML cells in vitro. As AML cells are more difficult to culture than CB Lin ^" cells, the LTC period was reduced from five to three weeks. Similarly to CB lin ^" cells, primary AML cells show a significant decrease in growth when cultured at 3% compared to 20% 0 ₂

(p<0.05) (Figure 8A) . We determined for three AML samples the total primary and secondary leukemic LTC-IC yielded at week-3 in both oxygen concentrations. In conventional 20% 0 ₂ co-culture, secondary LTC-IC were found either to decrease, maintain or increase compared to the input number for samples #6, #2 and #3 respectively (Figure 8B) . Low oxygen co-culture system was only found beneficial for sample #2 in term of total LTC-IC number whereas no impact was observed for samples #6 and #3. However when we quantified re-plating potential, we observed a 2.7-, 4.5- and 1.96-fold increase in secondary leukemic LTC-IC frequency for samples #6, #2 and #3, respectively, in 3% co-cultures as compared to 20% 0 ₂ (Figure 8C, p<0.005, p<0.0001 and p<0.05 respectively) . We then investigated whether the replating potential obtained was indicative of LIC activity. Through in vivo limiting dilution analysis, the total input uncultured and output co-cultured number of SL-IC for samples #2 was determined. 3% O ₂ culture condition allowed the maintenance of SL-IC compared to their initial input number whereas a significant 6.1-fold decrease was quantified with the normoxic culture system (Figure 8D, p<0.05) . Similarly to CB lin ^~ cells, the advantage of the culture at 3% 0 ₂ was more pronounced at the SL-IC frequency level. A 6.5-fold increase in SL-IC frequency at 3% compared to 20% 0) was observed (Figure 8E, p<0.0001). Following this, ten AML patient samples were co-cultured for three weeks at 3% O ₂ prior to transplantation into N0D/SCID~P2 ^ mice. In six of these samples, leukemic myeloid engraftment could be detected after 12 weeks (Figure 8F) . In three of these six samples, self- renewal capacity was investigated by performing serial

transplants and leukemic engraftment was observed in all secondary recipients (Figure 8F) . Myeloid restricted engraftment was observed in all cases (data not shown) , indicating that the residual normal HSCs within leukemic samples did not out-compete the AML cells during ex vivo culture, contrary to what was described previously in another culture systems (Ailles et al., 1997) . We believe that the failure to engraft seen in four of the ten patients tested (patients 7 to 10) was due to the low frequency of LICs present in these patients. LIC frequency before in vitro culture was determined for eight of the ten patient samples and displayed high heterogeneity, ranging from one SL-IC per 13,650 to one per 7.66 x 10 ^s blasts (Table 2).

Following in vitro culture, engraftment failure was restricted to the samples with low LIC frequency (samples 7, 8 and 10) , in which the injected cell doses fell below the initial LIC frequency observed in non-cultured cells. We were able to calculate the frequency of LICs before and after co-culture for patients 1 and 2, which demonstrated maintenance of the frequency (Table 2) . Based on the fact that the average expansion of AML was 2.31-fold (range: 0.135-8.39) after 3-weeks in co-culture (Table 2), we conclude that functionally defined LICs can be maintained in vitro over a 3 week period in MS-5+3GT at 3% 0 ₂ .

Chemoresistant LICs can be studied in vitro

We next investigated whether these culture conditions might be suitable for studying chemoresistance of LICs. Cytarabine (Ara- C) was used at a concentration >25 times higher than the IC ₅₀ dose determined after 3 days for three leukemic cell lines

(Figure 9A) . This selected dose of Ara-C was 3.8 times lower than the IC ₅₀ dose for MS-5 cells. After 7 days with Ara-C treatment at 3 μΜ in normoxic condition, we observed more than 94% eradication of both normal and leukemic cells in both suspension and MS-5 based culture systems (Figure 9 Bl and Figure 10A and 11A) . However, MS-5 co-culture improved the percentage of viable cells by 4-fold in the residual population for leukemic and normal samples as compared to the suspension system (Figure 9 B2, p<0.001 and p<0.05 for CB and AML cells, respectively). We observed that culturing the cells at 3% 0 ₂ improved the

recoverable fraction of live cells of normal and leukemic samples after Ara-C treatment (Figure 10A) . Re-plating live cells (at limiting dilution) at 3% 0 ₂ or at 20% 0 ₂ showed that secondary

LTC-ICs were enriched in the group cultured at 3% 0 ₂ as compared to 20% 0 ₂ (Figure 10B, 34 and 3 times more secondary LTC-IC per cells re-plated at 3% than at 20% 0 ₂ for sample #6 and #2, p<0.0001 and p<0.01, respectively). More importantly, low oxygen chemo-protective impact was also observed when looking at the total yield of secondary LTC for these two samples: 32 and 3 times more secondary LTC-IC were recovered in 3% 0 ₂ cultures (Figure IOC, p<0.01and p<0.05, for sample #6 and #2,

respectively) . We investigated if 3% 0 ₂ on its own, without feeder layers, could account for the increase in chemo-resistance observed. We also investigated whether re-plating potential within the residual population was indicative of LIC activity after in vitro chemotherapy. To address these questions, we compared the leukemic engraftment potential of the cells remaining after Ara-C treatment either grown in suspension culture at 3% 0 ₂ (NO MS-5 +3GT) or co-cultured with MS-5 + 3GT at 3% 0 ₂ (Figure 10D) . Only the cells grown on the MS-5 feeder layer could generate a leukemic engraftment, demonstrating that the maintenance of LICs requires stromal support for their chemo- resistance under 3% 0 ₂ .

In vitro Ara-C treatment does not enrich for LICs

Ara-C treatment at a concentration of 3μΜ for 1 week was very effective at killing the tumor burden (Figure 11A and 9B1) . We confirmed that the residual viable cells were selected for a population with a low division profile history (data not shown) . As LICs are reported to be quiescent (Guan et al., 2003), we investigated if the in vitro AraC treatment may target more efficiently the non-LICs and thus could enrich for LIC. For seven AML samples, the phenotype appeared globally unchanged despite the drastic impact of the Ara-C treatment at the cell number level (data not shown) . As LIC phenotype was functionally predetermined for 5 AML samples (using sorted cell fraction, data not shown) , we investigated whether a higher frequency of phenotypically defined LICs could be observed after Ara-C treatment. Indeed, a slight but not significant higher

proportion of LIC after Ara-C treatment could be observed, as well as for normal HSC (Figure 11B) . Using the

xenotransplantation assay, we further confirmed the persistence of LIC in the residual Ara-C treated population (Figure 11C) . By injecting the same cell dose (control and Ara-C group) , we observed a reduction in the level of engraftment in Ara-C treated compared to control in two out of the three patients tested (Figure 11C) . To quantify better the effect of Ara-C on LIC, we used limiting dilution analysis in secondary leukemia long-term culture system. This allowed us to determine the frequency of secondary L-LTC-ICs (Fig 11D) . For samples 2 and 3, we showed a decrease in secondary L-LTC-IC, whereas for patient 6 the frequency was not altered. Nevertheless as the total output of viable cells decreased significantly, we can conclude that in all cases (Fig 11A) , the total number of residual secondary L-LTC-IC decreases. To evaluate the correlation between secondary L-LTC- IC and SL-IC, we performed for patient 2 an in vivo limiting dilution analysis which allowed us to calculate the frequency and the total SL-LIC number after culture without or with Ara-C treatment (Fig HE) . We found a 16-fold reduction in the frequency post-treatment (Figure HE, p<0.01) and a 460 times reduction of their total SL-IC (Figure 11F, p<0.01).

Interestingly, a comparable 8-fold and 313-fold reduction in secondary L-LTC-IC frequency and in their total number

respectively could also be observed through secondary L-LTC-IC assay (Figure 11D, p<0.0001 and 7C, p<0.0001). Altogether, these data suggest that LICs are not enriched after Ara-C treatment, thus only a portion are chemoresistant demonstrating a functional heterogeneity of LICs. Comparative effect on LICs treated ex vivo and in vivo

To determine whether the in vitro Ara-C treatment in our culture condition would reproduce in vivo chemotherapy treatment protocol, we compared for one patient (#6) the impact of the two treatments. Twelve weeks after establishment of patient #6 leukemia in NSG mice, one group of animal was treated with Ara-C for 1 week and sacrificed at the end of the treatment. We observed a 2.53 times lower leukemic burden in the treated group (Figure 11G, p<0.05) . Compared to in vitro treatment, more AML cells survived in vivo (compare Fig HA and 11G) . To look at the impact on LICs, we performed limiting dilution analysis in a secondary transplant setting. Unfortunately, all the doses tested gave rise to leukemic engraftment in mice (Figure 11H) . Based on this data we can conclude that the frequency of LICs is lower than 1 in 30,000 cells. As the frequency of LIC at day 0 for this patient was 1 in 3,000, we can conclude that Ara-C treatment did not significantly impede on this. When compared to the results ex vivo, we again show that for this patient the secondary L-LTC-IC frequency was not affected (Figure 11D) .

Altogether, these results show a good correlation between in vitro and in vivo treatment of cells with Ara-C and allow us to propose that secondary L-LTC-IC could serve as a good surrogate assay to the xenotransplantation model allowing to screen drugs on LICs in vitro.

Correlation of LIC in vitro proliferation rate with engraftment potential and patient prognosis

Previous studies have reported that approximately 70% of AML cases will engraft in the NOD/SCID assay, but few studies have addressed the variables that affect engraftment itself (Monaco et al., 2004; Pearce et al., 2006; Vargaftig et al., 2011).

Although the various factors that may affect AML NOD/SCID engraftment could not be recognized as differentially expressed in "engrafter" (E) and "non-engrafter" (NE) samples, we

originally observed that engraftment was directly related to prognosis. We next investigated whether these culture conditions might be suitable for studying the biological differences between E and NE .

Fifteen E and thirteen NE were co-cultured for 5-weeks with MS- 5+3GT in bulk culture or in LDA. As shown in Figure 12, AML immunodeficient mice engrafter samples (E) yielded a higher number of output cells and were enriched in L-LTC-IC as compared to non-engrafter samples (NE) (p<0.05). To determine whether that cycling difference could be monitored at an early time point, we set up a fluorescent based quantification of cell proliferation applied to primary AML cells.

The day before initiating the co-culture (D-l), 2 to 10 x 10 ⁶ AML cells were thawed and contaminating T lymphocyte CD3+ cells were depleted using cell sorting or the immunomagnetic depletion method. AML cells were stained with 0.8 μΜ of the amine-reactive cell permeable dye carboxyfluorescein diacetate, succinimidyl ester (CFSE) (Invitrogen, UK) for 10 min at 37 °C in PBS. Cells were washed twice and incubated for 18h (overnight) in serum-free expansion medium StemSpan® SFEM (StemCell Technologies) in the absence of cytokines to allow unbound CFSE molecules to be released (Nordon et al . , 1997). CFSE-stained leukemic cells were incubated on an MS-5 confluent monolayer at D-l in MyeloCult H5100 (StemCell Technologies, Vancouver, CA) in the absence of cytokine and hydroxycortisone for 18h. Upon the non-specific unbound release step on Day 0 (DO), the cell subpopulation within the leukemic sample may be purified using methods known to one skilled in the art, such as fluorescence-activated cell sorting (FACS) technology according to cells surface antigen markers. Non-adherent cells were harvested through 3 gentle washes and adherent cells through trypsinization at DO. Recovered cells were re-suspended and stained in Annexin binding buffer (BD Biosciences) . Human leukemic cells were stained with anti-CD45- APC-Cy7, anti-CD34-Percp, anti-CD38-PE-Cy7 antibodies, and

AlexaFluor647-conjugated-Annexin-V (Invitrogen) and 4,6

diamidino-2-phenylindole (DAPI). MS-5 stromal cells were stained with Sca-l-PE (Sca-1 was identified as a specific marker of 100% of MS-5). All antibodies were obtained from BD Biosciences, UK. Fluorescent beads were also used for determining the precise cell count of leukemic viable cells. Stained cells were analyzed by FACS at DO. The direct use of an amino-reactive fluorescent probe for tracking AML cell division is complicated due to the lower cell viability and expansion after 1 week (see Figure 13) . Only viable (DAPI and Annexin-V negatives) were assessed for all analysis. This is because probe fluorescence intensity is over- estimated due to apoptosis (see Figure 14). The starting bounded amino-reactive probe Mean Fluorescence Intensity (DO Leukemic CFSE MFI) was determined for leukemic viable cells (DAPI and Annexin-V negatives, CD45-APC-Cy7 positive and Sca-l-PE negative cells) . The MFI of the fluorescent beads (DO beads MFI) was also determined for normalisation and calibration purpose. A narrow CFSE gate may be applied for sorting live subpopulation for example, CB lin- CD34+CD381ow/neg and CD34+CD38+ and CD34-CD38+ and CD34-CD38- cells may be purified by FACS to obtain a very tight homogenous CFSE staining. However, this classical procedure used for tracking cell division can only be used for 14% of the leukemic samples, particularly because of the heterogeneous AML blast scatter (see Figure 15) . In contrast, the determination of the FDF by the method described herein is applicable for all the leukemic samples.

Then, a known count of leukemic cells was seeded at 80 to 200xl0 ³ cells per cm ² of pre-established confluent stromal cells and incubated in vitro for 1 week. At day 7 (D7), classical division profiles may be analysed by FACS (e.g. Figure 15A-1 iii-iv) .

Non-adherent cells were harvested through 3 gentle washes and adherent cells through trypsinisation at D7. Recovered cells were re-suspended and stained in Annexin binding buffer (BD Biosciences) . Human leukemic cells were stained with anti-CD45- APC-Cy7, anti-CD34-Percp, anti-CD38-PE-Cy7 antibodies, and

AlexaFluor647-conjugated-Annexin-V (Invitrogen) and 4,6

diamidino-2-phenylindole (DAPI) and SCA-l-PE. Fluorescent beads were also used for determining the precise cell count of leukemic viable cells at D7. The MFI of bound amino-reactive probe at D7 (D7 Leukemic CFSE MFI) was determined for leukemic viable cells (DAPI and Annexin-V negatives, CD45-APC-Cy7 positive and Sca-l-PE negative cells) . The MFI of fluorescent beads at D7 (D7 bead MFI) was also determined for normalisation and calibration purposes .

Then the fluorescence dilution factor (FDF) was determined for the sample. FDF = Leukemic CFSE MFI at DO normalised to bead MFI at DO/ Leukemic CFSE MFI at D7 normalised to bead MFI at D7.

For normal CB samples, for which the higher viability across cell generations allows the expansion of the cell population, we could establish FDF measurement highly correlated with the cell population expansion or the proliferation index derived from computational analysis of the Flowjo proliferation platform (Tree Star, Oten, Switzerland) . A training data set derived from several independent experiments performed with purified HSC (Lin ^" CD34 ⁺ CD38 ^~ ) and progenitors (Lin ^" CD34 ⁺ CD38 ⁺ ) established FDF was even more accurate than a simple cell count in terms of

reproducibility (data not shown) . An FDF between 1 and 2 indicates the most live cells have not divided during 1 week of culture and an FDF>2 indicates that most live cells have divided at least once.

Measurement of FDF allows a low-resolution measurement of AML cell proliferation in one week (see Figure 15).

Figure 16 shows that NOD/SCID engrafting (E) samples have a higher FDF value than non-engrafting (NE) samples.

According to the predictions shown in Figure 16, test samples with a viability of >30% at D7 and with an FDF of >2.56 are predicted to be able to engraft immunodeficient mice, whereas test samples with a viability of ≥30% at D7 and with an FDF of ≤1.7 are predicted to be non-engrafters of immunodeficient mice, with a false discovery rate of 10% respectively. Test samples of AML cells with a viability of >30% at D7 and with an FDF of >2.94 are predicted to be able to engraft immunodeficient mice, whereas test samples with a viability of ≥30% at D7 and with an FDF of ≤1.44 are predicted to be non-engrafters of immunodeficient mice, with a false discovery rate of 5% respectively.

Figure 17A shows that the overall survival rate of patients whose AML samples were NOD/SCID engrafting was statistically lower than those whose AML samples were non-engrafting. In line with this, Figure 17B shows that the overall survival rate of patients whose AML samples had a high proliferation rate (FDF>2.23) was significantly lower than those whose AML samples had a low proliferation rate (FDF<2.23). Therefore, the proliferation rate of AML cells in vitro may be used to predict the prognosis (e.g. likely survival rate) of leukemia patients.

DISCUSSION

In this study, we have investigated surrogate models for bone marrow stroma for their capacity to maintain normal human HSCs and LICs in vitro. Only long-term cultures over five weeks could clearly show the divergent potential between the different systems. MS-5+3GT was the only condition tested that was able to expand secondary LTC-ICs. Although we cannot conclude that all human osteoblastic or endothelial cell models only weakly support long-term HSC maintenance, we obtained similar results with respect to total population expansion and CFC output when testing an additional osteoblastic cell line, CAL-72, and endothelial cell line, HBMEC-28 (data not shown) . For both normal and leukemic hematopoietic stem cells, the best in vitro stem cell read-out was the re-plating potential criteria. Indeed, in different experiments looking at different feeder layers, low oxygen or chemotherapy impact, we repeatedly observed that the secondary LTC potential mirrors the repopulating potential in vivo. Previous in vitro studies analyzing the impact of oxygen levels on HSCs have reported a drastic decrease in cell cycle

(Cipolleschi et al . , 1993; Dao et al . , 2007; Hammoud et al . , 2011; Hermitte et al., 2006). Recently, in vivo studies investigating the oxidative detoxification pathways in HSC at steady state or during transplantation correlated the generation of ROS with premature HSC senescence associated with pl6 overexpression (Ito et al . , 2004; Tothova et al., 2007; Yahata et al . , 2011). Consistent with these observations, our results show that culture under low oxygen concentrations reversibly reduce proliferation and differentiation while preserving LTC-IC and SRC frequencies, compared to standard culture conditions at 20% 0 ₂ , where we observed greater levels of ROS and pi 6 mRNA.

Interestingly, we also show that conventional 20% normoxic ex vivo cultures were evenly deleterious for the maintenance of LICs whereas co-culture of AML cells with MS-5+3GT at 3% 0 ₂ can maintain LICs. This was demonstrated by the leukemic engraftment potential after serial transplantation, indicating that LIC self- renewal potential was preserved during the three-week culture period. Based on the maintenance / slight expansion of total cells, we can conclude that the LIC can be maintained in co- culture and thus this in vitro system can serve to elucidate many aspects of LIC biology.

Finally, we demonstrate that this system is capable of supporting functionally defined human chemo-resistant LICs as efficiently as in vivo. Our one-week Ara-C treatment was considerably more stringent than the majority of studies investigating

chemotherapeutic drug-induced cell death in suspension after 48 to 72 hours. Moreover, our work focused on the residual spared population rather than on the bulk tumour population. We determined that low oxygen improved LIC chemoresistance in a feeder dependent way. Contrary to what was suspected, our results also support the notion LICs are not enriched by Ara-C treatment either in vitro or in vivo. This indicates for the first time that not all LICs are chemoresistant . For one sample we could actually correlate the in vitro anti-neoplastic treatment to an in vivo conditioning as well as the subsequent read-out comparing in vitro replating assay to the xenograft assay. Thus, we believe this in vitro system coupled with the re-plating LTC-IC assay read-out could represent the first system in which LIC sensitivity to treatment could be tested on a relatively large scale. This appears even more crucial in the context of preclinical and clinical studies, which have

demonstrated that assessment of bulk tumor shrinkage was not relevant to improvement in patient survival (Ardeshna et al., 2003; Durie et al . , 2004; Huff et al., 2006). Altogether we believe this tool could enable both pharmaceutical and academic research laboratories to design innovative therapeutic approaches to target the root of AML. Table 1: Impact of 3% 0 ₂ on primary and secondary CB LTC-IC and

SRC

Secondary plating

Primary plating 20 or 3% o ₂ ^A 20%O

1 SRC per

First

LTC-IC LTC-IC per inj ected lating Fold Total Total

Total per 0.1 0.1 10 ^s cells number 0 ₂ % expansi LTC-IC LTC-IC

CFC 10 ⁶ Cells Cells (range) ^A on (range) ( range )

(range) ( range )

7, 692 384 472 2,202,500

MS-5+3GT 42, 932 26.8 (15.8-

351±35 (5192- (259- (277- (903, 966- 20% 0 ₂ ±17, 954 45.6)

11, 280) 564) 803) 5,366,753) ^c

78.4 3.9 774,000

MS-5+3GT 18.7±1. 73.8 (36.4- 69 (34-

1641110 (48.4- (2.42- (450, 127- 3% 0 ₂ 09 149.7) 140)

127) 6.35) 1, 330, 925) ^D

Mean number +SEM of CB lin ^~ cells fold expansion; total CFC;

5 primary LTC-IC frequency. ^A determined at week 5 for indicated

conditions. ^B replated on MS-5+3GT 20% 0 ₂ and determined at week 10. SRC frequency determined in NOD/SCID-p ₂ m ^{~ ~} mice (n=25 ^c and 50 ^D ) in CD45+ sorted cells recovered after 5-weeks from 4 independent co-cultures . Representative data of 6 independent 10 primary bulk LTC and 2 replating LDA experiments.

Table 2: SL-IC frequency is retained over 3 weeks co-cultured in vitro with MS-5+3GT at 3% 0 ₂

Number of

1 SL-IC per 1 SL-IC per

Number Fold Dose Equivalent leukemic injected cells injected cells of co- expansion injected after input cells engrafted mice/ number before number after culture (±SE ) culture injected Number of mice culture (range) culture (range) analyzed

20,000 68,966 0/1

335,107 46,000 180,392 0/1 489,267

(201 ,948- 5 0.580+0.3 500,000 537,634 1/5 (279,919-

556,067) 500,000 607,118 9/9 855,183)

500,000 833,333 4/5

200 461 0/4

2,000 4,608 1/3

11,500 15,972 1/1

20,000 24,390 0/2

175,966 (76,016- 20,000 46,083 3/3 162,570

4 0.678±0.196 (80,320- 407,337) 100,000 121 ,951 3/3 329,046)

200,000 460,829 4/4

500,000 609,756 2/4

535,000 622,093 3/3

4,800,000 5,581 ,395 3/3

78,108 (5,635-

1 0.135 30,000 222,816 3/7 <222,815 33,065)

13,650 (28,609-

1 0.3 500,000 1 ,666,667 2/2 <1 ,666,666 213,248)

116,000 21 ,405 0/1

ND 1 5.42 <239,881

1 ,300,000 239,882 3/3

333 699

(16o 957- 1 1.21 794,000 656,198 4/4 <656,198

691,832)

2,229,695 8.40 500,000 59,546 0/8

(1 ,006,989- 2 ND 4,937,033) ND 1 ,000,000 ND 0/9

60,000 289,381 0/4

> 5,000,000 2 0.17+0.05 ND

100,000 737,191 0/2

ND 1 ND 100,000 ND 0/1 ND

7,663,450

(4,036,119- 1 2.775 3,000,000 1 ,081 ,081 0/6 ND 14,550,727)

ND: not determined Table 3: Characteristics of AML patient samples

All samples were obtained at diagnosis. FAB: French-American- British classification. ND: not determined. REFERENCES

All references cited anywhere in this specification, including those cited anywhere above, are hereby incorporated by reference in their entirety and for all purposes.

Ailles, L.E., B. Gerhard, and D.E. Hogge. 1997. Detection and characterization of primitive malignant and normal

progenitors in patients with acute myelogenous leukemia using long-term coculture with supportive feeder layers and cytokines. Blood. 90:2555-2564.

Amsellem, S., F. Pflumio, D. Bardinet, B. Izac, P. Charneau, P.H.

Romeo, A. Dubart-Kupperschmitt, and S. Fichelson. 2003. Ex vivo expansion of human hematopoietic stem cells by direct delivery of the HOXB homeoprotein . Nat Med. 9:1423-1427.

Anj os-Afonso, F., and D. Bonnet. 2008. Isolation, culture, and differentiation potential of mouse marrow stromal cells. Curr Protoc Stem Cell Biol. Chapter 2: Unit 2B 3.

Ardeshna, K.M., P. Smith, A. Norton, B.W. Hancock, P.J. Hoskin, K.A. MacLennan, R.E. Marcus, A. Jelliffe, G. Vaughan,

Hudson, and D.C. Linch. 2003. Long-term effect of a watch and wait policy versus immediate systemic treatment for asymptomatic advanced-stage non-Hodgkin lymphoma: a randomised controlled trial. Lancet. 362:516-522.

Bennaceur-Griscelli, A., C. Tourino, B. Izac, W. Vainchenker, and L. Coulombel. 1999. Murine stromal cells counteract the loss of long-term culture-initiating cell potential induced by cytokines in CD34 (+) CD38 (low/neg) human bone marrow cells. Blood. 94:529-538.

Bonnet, D. 2008. In vivo evaluation of leukemic stem cells

through the xenotransplantation model. Curr Protoc Stem Cell Biol. Chapter 3: Unit 3 2.

Bonnet, D. 2009a. Humanized model to study leukemic stem cells.

Methods Mol Biol. 538:247-262.

Bonnet, D. 2009b. Humanized mouse models to study the human

haematopoietic stem cell compartment. Methods Mol Biol. 506:331-340. Bonnet, D., and J.E. Dick. 1997. Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nat Med. 3:730-737.

Bruserud, 0., A. Ryningen, L. ergeland, N.I. Glenjen, and B.T.

Gjertsen. 2004. Osteoblasts increase proliferation and release of pro-angiogenic interleukin 8 by native human acute myelogenous leukemia blasts. Haematologica . 89:391- 402.

Cairoli, R . , A. Beghini, G. Grillo, G. Nadali, F. Elice, C.B.

Ripamonti, P. Colapietro, M. Nichelatti, L. Pezzetti, M.

Lunghi, A. Cuneo, A. Viola, F. Ferrara, M. Lazzarino, F.

Rodeghiero, G. Pizzolo, L. Larizza, and E. Morra. 2006.

Prognostic impact of c-KIT mutations in core binding factor leukemias : an Italian retrospective study. Blood. 107:3463- 3468.

Calvi, L.M., G.B. Adams, K.W. Weibrecht, J.M. Weber, D.P. Olson, M.C. Knight, R.P. Martin, E. Schipani, P. Divieti, F.R. Bringhurst, L.A. Milner, H.M. Kronenberg, and D.T. Scadden. 2003. Osteoblastic cells regulate the haematopoietic stem cell niche. Nature. 425:841-846.

Celebi, B., D. Mantovani, and N. Pineault. 2011. Effects of

extracellular matrix proteins on the growth of

haematopoietic progenitor cells. Biomed Mater. 6:055011.

Chan, . , M. Koritzinsky, H. Zhao, R. Bindra, P.M. Glazer, S.

Powell, A. Belmaaza, B. Wouters, and R.G. Bristow. 2008.

Chronic hypoxia decreases synthesis of homologous

recombination proteins to offset chemoresistance and radioresistance . Cancer Res. 68:605-614.

Cheong, J.W., J.I. Eom, H.Y. Maeng, S.T. Lee, J.S. Hahn, Y.W. Ko, and Y.H. Min. 2003. Phosphatase and tensin homologue phosphorylation in the C-terminal regulatory domain is frequently observed in acute myeloid leukaemia and

associated with poor clinical outcome. Br J Haematol .

122 : 454-456.

Chung, Y.J., C.W. Choi, C. Slape, T. Fry, and P.D. Apian. 2008.

Transplantation of a myelodysplastic syndrome by a long- term repopulating hematopoietic cell. Proc Natl Acad Sci U

S A. 105:14088-14093.

Cicuttini, F.M., M. Martin, E. Salvaris, L. Ashman, C.G. Begley,

J. Novotny, D. Maher, and A.W. Boyd. 1992. Support of human cord blood progenitor cells on human stromal cell lines transformed by SV40 large T antigen under the influence of an inducible (metallothionein) promoter. Blood. 80:102-112. Cipolleschi, M.G., P. Dello Sbarba, and M. Olivotto. 1993. The role of hypoxia in the maintenance of hematopoietic stem cells. Blood. 82:2031-2037.

Collins, L.S., and K. Dorshkind. 1987. A stromal cell line from myeloid long-term bone marrow cultures can support

myelopoiesis and B lymphopoiesis. J Immunol. 138:1082-1087. Comerford, K.M., T.J. Wallace, J. Karhausen, N.A. Louis, M.C.

Montalto, and S.P. Colgan. 2002. Hypoxia-inducible factor-

1-dependent regulation of the multidrug resistance (MDR1) gene. Cancer Res. 62:3387-3394.

Damiano, J.S., A.E. Cress, L.A. Hazlehurst, A.A. Shtil, and W.S.

Dalton. 1999. Cell adhesion mediated drug resistance (CAM- DR) : role of integrins and resistance to apoptosis in human myeloma cell lines. Blood. 93:1658-1667.

Danet, G.H., Y. Pan, J.L. Luongo, D.A. Bonnet, and M.C. Simon.

2003. Expansion of human SCID-repopulating cells under hypoxic conditions. J Clin Invest. 112:126-135.

Dao, M.A., M.H. Creer, J.A. Nolta, and CM. Verfaillie. 2007.

Biology of umbilical cord blood progenitors in bone marrow niches. Blood. 110:74-81.

Dias, S., M. Choy, K. Alitalo, and S. Rafii. 2002. Vascular

endothelial growth factor (VEGF) -C signaling through FLT-4 (VEGFR-3) mediates leukemic cell proliferation, survival, and resistance to chemotherapy. Blood. 99:2179-2184.

Durie, B.G., J. Jacobson, B. Barlogie, and J. Crowley. 2004.

Magnitude of response with myeloma frontline therapy does not predict outcome: importance of time to progression in southwest oncology group chemotherapy trials. J Clin Oncol.

22:1857-1863. Erler, J.T., C.J. Cawthorne, K.J. Williams, M. Koritzinsky, B.G. Wouters, C. Wilson, C. Miller, C. Demonacos, I.J.

Stratford, and C. Dive. 2004. Hypoxia-mediated down- regulation of Bid and Bax in tumors occurs via hypoxia- inducible factor 1-dependent and -independent mechanisms and contributes to drug resistance. Mol Cell Biol. 24:2875- 2889.

Fiegl, M. , I. Samudio, K. Clise-Dwyer, J.K. Burks, Z. Mnjoyan, and M. Andreeff. 2009. CXCR4 expression and biologic activity in acute myeloid leukemia are dependent on oxygen partial pressure. Blood. 113:1504-1512.

Fiegl, M . , I. Samudio, Z. Mnjoyan, B. Korchin, H. Fritsche, and M. Andreeff. 2010. Physiological hypoxia promotes lipid raft and PI3K-dependent activation of MAPK 42/44 in leukemia cells. Leukemia. 24:1364-1367.

Fogh, J., J.M. Fogh, and T. Orfeo . 1977. One hundred and twenty- seven cultured human tumor cell lines producing tumors in nude mice. J Natl Cancer Inst. 59:221-226.

Gillette, J.M., A. Larochelle, C.E. Dunbar, and J. Lippincott- Schwartz. 2009. Intercellular transfer to signalling endosomes regulates an ex vivo bone marrow niche. Wat Cell Biol. 11:303-311.

Glenjen, N.I., K. Hatfield, and 0. Bruserud. 2005. Coculture of native human acute myelogenous leukemia blasts with fibroblasts and osteoblasts results in an increase of vascular endothelial growth factor levels. Eur J Haematol. 7 :24-34.

Goardon, N . , E. Marchi, A. Atzberger, L. Quek, A. Schuh, S.

Soneji, P. Woll, A. Mead, K.A. Alford, R. Rout, S.

Chaudhury, A. Gilkes, S. Knapper, K. Beldjord, S. Begum, S.

Rose, N. Geddes, M. Griffiths, G. Standen, A. Sternberg, J.

Cavenagh, H. Hunter, D. Bowen, S. Killick, L. Robinson, A.

Price, E. Macintyre, P. Virgo, A. Burnett, C. Craddock, T.

Enver, S.E. Jacobsen, C. Porcher, and P. Vyas . 2011.

Coexistence of LMPP-like and GMP-like leukemia stem cells in acute myeloid leukemia. Cancer Cell. 19:138-152. Grimwade, D., R.K. Hills, A.V. Moorman, H. Walker, S. Chatters, A.H. Goldstone, K. Wheatley, C.J. Harrison, and A.K.

Burnett. 2010. Refinement of cytogenetic classification in acute myeloid leukemia: determination of prognostic significance of rare recurring chromosomal abnormalities among 5876 younger adult patients treated in the United Kingdom Medical Research Council trials. Blood. 116:354- 365.

Guan, Y., B. Gerhard, and D.E. Hogge . 2003. Detection, isolation, and stimulation of quiescent primitive leukemic progenitor cells from patients with acute myeloid leukemia (AML) .

Blood. 101:3142-3149.

Guzman, M.L., S.J. Neering, D. Upchurch, B. Grimes, D.S. Howard, D.A. Rizzieri, S.M. Luger, and C.T. Jordan. 2001. Nuclear factor-kappaB is constitutively activated in primitive human acute myelogenous leukemia cells. Blood. 98:2301- 2307.

Guzman, M.L., C.F. Swiderski, D.S. Howard, B.A. Grimes, R.M.

Rossi, S.J. Szilvassy, and C.T. Jordan. 2002. Preferential induction of apoptosis for primary human leukemic stem cells. Proc Natl Acad Sci U S A. 99:16220-16225.

Hammoud, M. , M. Vlaski, P. Duchez, J. Chevaleyre, X. Lafarge,

J.M. Boiron, P.B. De La Grange, and Z. Ivanovic. 2011.

Combination of low 0(2) concentration and mesenchymal stromal cells during culture of cord blood CD34 (+) cells improves the maintenance and proliferative capacity of hematopoietic stem cells. J Cell Physiol.

Hardy, R.R., T. Kishimoto, and K. Hayakawa. 1987. Differentiation of B cell progenitors in vitro: generation of surface IgM+ B cells, including Ly-1 B cells, from Thy-1- asialoGMl+ cells in newborn liver. Eur J Immunol. 17:1769-1774.

Harigaya, K., and H. Handa . 1985. Generation of functional clonal cell lines from human bone marrow stroma. Proc Natl Acad

Sci U S A. 82:3477-3480.

Hermitte, F . , P. Brunet de la Grange, F. Belloc, V. Praloran, and

Z. Ivanovic. 2006. Very low 02 concentration (0.1%) favors

GO return of dividing CD34+ cells. Stem Cells. 24:65-73. Holyoake, T., X. Jiang, C. Eaves, and A. Eaves. 1999. Isolation of a highly quiescent subpopulation of primitive leukemic cells in chronic myeloid leukemia. Blood. 94:2056-2064.

Hooper, A.T., J.M. Butler, D.J. Nolan, A. Kranz, K. Iida, M.

Kobayashi, H.G. Kopp, K. Shido, I. Petit, K. Yanger, D.

James, L. Witte, Z. Zhu, Y. Wu, B. Pytowski, Z. Rosenwaks, V. Mittal, T.N. Sato, and S. Rafii. 2009. Engraftment and reconstitution of hematopoiesis is dependent on VEGFR2- mediated regeneration of sinusoidal endothelial cells. Cell Stem Cell. 4:263-274.

Huff, C.A., W. Matsui, B.D. Smith, and R.J. Jones. 2006. The

paradox of response and survival in cancer therapeutics . Blood. 107:431-434.

Illmer, T., C. Thiede, A. Fredersdorf, S. Stadler, A. Neubauer, G. Ehninger, and M. Schaich. 2005. Activation of the RAS pathway is predictive for a chemosensitive phenotype of acute myelogenous leukemia blasts. Clin Cancer Res.

11:3217-3224.

Issaad, C., L. Croisille, A. Katz, W. Vainchenker, and L.

Coulombel . 1993. A murine stromal cell line allows the proliferation of very primitive human CD34++/CD38- progenitor cells in long-term cultures and semisolid assays. Blood. 81:2916-2924.

Ito, K. , A. Hirao, F. Arai, S. Matsuoka, K. Takubo, I. Hamaguchi, K. Nomiyama, K. Hosokawa, K. Sakurada, N. Nakagata, Y.

Ikeda, T.W. Mak, and T. Suda. 2004. Regulation of oxidative stress by ATM is required for self-renewal of

haematopoietic stem cells. Nature. 431:997-1002.

Itoh, K., H. Tezuka, H. Sakoda, M. Konno, K. Nagata, T. Uchiyama, H. Uchino, and K.J. Mori. 1989. Reproducible establishment of hemopoietic supportive stromal cell lines from murine bone marrow. Exp Hematol . 17:145-153.

Jazwiec, B., A. Solanilla, C. Grosset, F.X. Mahon, M. Dupouy, V.

Pigeonnier-Lagarde, F. Belloc, K. Schweitzer, J. Reiffers, and J. Ripoche . 1998. Endothelial cell support of

hematopoiesis is differentially altered by IL-1 and glucocorticoids. Leukemia. 12:1210-1220. Kiel, M.J., O.H. Yilmaz, T. Iwashita, C. Terhorst, and S.J.

Morrison. 2005. SLAM family receptors distinguish

hematopoietic stem and progenitor cells and reveal

endothelial niches for stem cells. Cell. 121:1109-1121. Kodama, H., M. Nose, S. Niida, and S. Nishikawa. 1994.

Involvement of the c-kit receptor in the adhesion of hematopoietic stem cells to stromal cells. Exp Heraatol . 22 : 979-984.

Kodama, H.A., Y. Amagai, H. Koyama, and S. Kasai. 1982. A new preadipose cell line derived from newborn mouse calvaria can promote the proliferation of pluripotent hemopoietic stem cells in vitro. J Cell Physiol. 112:89-95.

Kubota, Y., K. Takubo, and T. Suda. 2008. Bone marrow long label- retaining cells reside in the sinusoidal hypoxic niche. Biochem Blophys Res Commun. 366:335-339.

Lapidot, T., C. Sirard, J. Vormoor, B. Murdoch, T. Hoang, J.

Caceres-Cortes , M. Minden, B. Paterson, M.A. Caligiuri, and J.E. Dick. 1994. A cell initiating human acute myeloid leukaemia after transplantation into SCID mice. Nature. 367:645-648.

Leisten, I., R. Kramann, M.S. Ventura Ferreira, M. Bovi, S.

Neuss, P. Ziegler, W. Wagner, R. Knuchel, and R.K.

Schneider. 2012. 3D co-culture of hematopoietic stem and progenitor cells and mesenchymal stem cells in collagen scaffolds as a model of the hematopoietic niche.

Biomaterlals. 33:1736-1747.

Lemoine, F.M., R.K. Humphries, S.D. Abraham, G. Krystal, and C.J.

Eaves. 1988. Partial characterization of a novel stromal cell-derived pre-B-cell growth factor active on normal and immortalized pre-B cells. Exp Hematol. 16:718-726.

Li, W., S.A. Johnson, W.C. Shelley, and M.C. Yoder. 2004.

Hematopoietic stem cell repopulating ability can be maintained in vitro by some primary endothelial cells. Exp Hematol. 32:1226-1237.

Liesveld, J.L., J. Bechelli, K. Rosell, C. Lu, G. Bridger, G.

Phillips, 2nd, and C.N. Abboud. 2007. Effects of AMD3100 on transmigration and survival of acute myelogenous leukemia cells. Leuk Res. 31:1553-1563.

Liesveld, J.L., K.E. Rosell, C. Lu, J. Bechelli, G. Phillips, J.E. Lancet, and C.N. Abboud. 2005. Acute myelogenous leukemia—microenvironment interactions: role of

endothelial cells and proteasome inhibition. Hematology. 10:483-494.

Lowenberg, B., J.R. Downing, and A. Burnett. 1999. Acute myeloid leukemia. N Engl J Med. 341:1051-1062.

Lu, L.S., S.J. Wang, and R. Auerbach. 1996. In vitro and in vivo differentiation into B cells, T cells, and myeloid cells of primitive yolk sac hematopoietic precursor cells expanded > 100-fold by coculture with a clonal yolk sac endothelial cell line. Proc Natl Acad Scl U S A. 93:14782-14787.

Martelli, A.M., P.L. Tazzari, C. Evangelisti, F. Chiarini, W.L.

Blalock, A.M. Billi, L. Manzoli, J. A. McCubrey, and L.

Cocco. 2007. Targeting the phosphatidylinositol 3- kinase/Akt/mammalian target of rapamycin module for acute myelogenous leukemia therapy: from bench to bedside. Curr Med Chem. 14:2009-2023.

Meads, M.B., L.A. Hazlehurst, and W.S. Dalton. 2008. The bone marrow microenvironment as a tumor sanctuary and

contributor to drug resistance. Clin Cancer Res. 14:2519- 2526.

Mendez-Ferrer, S., T.V. Michurina, F. Ferraro, A.R. Mazloom, B.D.

Macarthur, S.A. Lira, D.T. Scadden, A. Ma'ayan, G.N.

Enikolopov, and P.S. Frenette. 2010. Mesenchymal and haematopoietic stem cells form a unique bone marrow niche.

Nature. 466:829-834.

Miranda, A.F., G.J. Duigou, E. Hernandez, and P.B. Fisher. 1988.

Characterization of mutant human fibroblast cultures transformed with simian virus 40. J Cell Sci. 89 ( Pt

4) :481-493.

Monaco, G., M. Konopleva, M. Munsell, C. Leysath, R.Y. Wang, C.E.

Jackson, M. Korbling, E. Estey, J. Belmont, and M.

Andreeff. 2004. Engraftment of acute myeloid leukemia in NOD/SCID mice is independent of CXCR4 and predicts poor patient survival. Stem Cells. 22:188-201.

Nathan, D.G. 2011. On Discovering Drugs, harvard magazine.

Ni, K., and H.C. O'Neill. 1997. Long-term stromal cultures

produce dendritic-like cells. Br J Haematol. 97:710-725.

Nordon, R.E., S.S. Ginsberg, and C.J. Eaves. 1997. High- resolution cell division tracking demonstrates the FLt3- ligand-dependence of human marrow CD34+CD38- cell

production in vitro. Br J Haematol. 98:528-539.

Obinata, M . , R. Okuyama, K.I. Matsuda, M. Koguma, and N. Yanai .

1998. Regulation of myeloid and lymphoid development of hematopoietic stem cells by bone marrow stromal cells. Leuk Lymphoma. 29:61-69.

Ogata, M . , H. atsubara, Y. Takai, H. Kosaka, T. Katagiri, H.

Sano, K. Ishimura, H. Fujita, T. Hamaoka, and H. Fujiwara.

1989. Capacities of a newly established thymic stromal cell clone to express la antigens and to produce interleukin-6, colony-stimulating factor, and thymic stroma-derived T-cell growth factor. J Leukoc Biol. 45:69-78.

Parmar, K., P. Mauch, J. A. Vergilio, R. Sackstein, and J.D. Down.

2007. Distribution of hematopoietic stem cells in the bone marrow according to regional hypoxia. Proc Natl Acad Sci U S A. 104:5431-5436.

Paul, S.R., Y.C. Yang, R.E. Donahue, S. Goldring, and D.A.

Williams. 1991. Stromal cell-associated hematopoiesis :

immortalization and characterization of a primate bone marrow-derived stromal cell line. Blood. 77:1723-1733.

Pearce, D.J., D. Taussig, K. Zibara, L.L. Smith, CM. Ridler, C.

Preudhomme, B.D. Young, A.Z. Rohatiner, T.A. Lister, and D. Bonnet. 2006. AML engraftment in the NOD/SCID assay reflects the outcome of AML: implications for our

understanding of the heterogeneity of AML. Blood. 107:1166- 1173.

Pessina, A., E. Mineo, M.G. Neri, L. Gribaldo, R. Colombi, P.

Brambilla, and G. Zaleskis. 1992. Establishment and characterization of a new murine cell line (SR-4987) derived from marrow stromal cells. Cytotechnology. 8:93- 102.

Ploemacher, R.E., J. P. van der Sluijs, J.S. Voerman, and N.H.

Brons . 1989. An in vitro limiting-dilution assay of long- term repopulating hematopoietic stem cells in the mouse. Blood. 74:2755-2763.

Rafii, S., F. Shapiro, R. Pettengell, B. Ferris, R.L. Nachman, M.A. Moore, and A.S. Asch. 1995. Human bone marrow

microvascular endothelial cells support long-term

proliferation and differentiation of myeloid and

megakaryocytic progenitors. Blood. 86:3353-3363.

Rehn, M., A. Olsson, K. Reckzeh, E. Diffner, P. Carmeliet, G.

Landberg, and J. Cammenga . 2011. Hypoxic induction of vascular endothelial growth factor regulates murine hematopoietic stem cell function in the low-oxygenic niche. Blood.

Rodan, S.B., Y. Imai, M.A. Thiede, G. Wesolowski, D. Thompson, Z.

Bar-Shavit, S. Shull, K. Mann, and G.A. Rodan. 1987.

Characterization of a human osteosarcoma cell line (Saos-2) with osteoblastic properties. Cancer Res. 47:4961-4966.

Sarry, J.E., K. Murphy, R. Perry, P.V. Sanchez, A. Secreto, C.

Keefer, C.R. Swider, A.C. Strzelecki, C. Cavelier, C.

Recher, V. Mansat-De Mas, E. Delabesse, G. Danet-Desnoyers , and M. Carroll. 2011. Human acute myelogenous leukemia stem cells are rare and heterogeneous when assayed in

NOD/SCID/IL2Rgammac-deficient mice. J Clin Invest. 121:384- 395.

Schweitzer, K.M., A.M. Drager, P. van der Valk, S.F. Thijsen, A.

Zevenbergen, A. P. Theijsmeijer, C.E. van der Schoot, and M.M. Langenhuij sen . 1996. Constitutive expression of E- selectin and vascular cell adhesion molecule-1 on

endothelial cells of hematopoietic tissues. Am J Pathol. 148:165-175.

Simsek, T., F. Kocabas, J. Zheng, R.J. Deberardinis, A.I.

Mahmoud, E.N. Olson, J.W. Schneider, C.C. Zhang, and H.A. Sadek. 2010. The distinct metabolic profile of hematopoietic stem cells reflects their location in a hypoxic niche. Cell Stem Cell. 7:380-390.

Singer, J.W., P. Charbord, A. Keating, J. Nemunaitis, G. Raugi,

T.N. Wight, J. A. Lopez, G.J. Roth, L.W. Dow, and P.J.

Fialkow. 1987. Simian virus 40-transformed adherent cells from human long-term marrow cultures: cloned cell lines produce cells with stromal and hematopoietic

characteristics. Blood. 70:464-474.

Szilvassy, S.J., T.E. Meyerrose, P.L. Ragland, and B. Grimes.

2001. Homing and engraftment defects in ex vivo expanded murine hematopoietic cells are associated with

downregulation of betal integrin. Exp Hematol . 29:1494-

1502.

Taichman, R.S., M.J. Reilly, and S.G. Emerson. 1996. Human

osteoblasts support human hematopoietic progenitor cells in vitro bone marrow cultures. Blood. 87:518-524.

Taichman, R.S., M.J. Reilly, and S.G. Emerson. 1997. Human

osteosarcomas inhibit hematopoietic colony formation:

partial reversal by antibody to transforming growth factor- beta 1. Bone. 21:353-361.

Takahashi, S., S.V. Reddy, M. Dallas, R. Devlin, J.Y. Chou, and G.D. Roodman. 1995. Development and characterization of a human marrow stromal cell line that enhances osteoclast- like cell formation. Endocrinology. 136:1441-1449.

Takubo, K. , N. Goda, W. Yamada, H. Iriuchishima, E. Ikeda, Y.

Kubota, H. Shima, R.S. Johnson, A. Hirao, M. Suematsu, and T. Suda. 2010. Regulation of the HIF-lalpha level is essential for hematopoietic stem cells. Cell Stem Cell. 7 : 391-402.

Taussig, D.C., F. Miraki-Moud, F. An os-Afonso, D.J. Pearce, K.

Allen, C. Ridler, D. Lillington, H. Oakervee, J. Cavenagh, S.G. Agrawal, T.A. Lister, J.G. Gribben, and D. Bonnet. 2008. Anti-CD38 antibody-mediated clearance of human repopulating cells masks the heterogeneity of leukemia- initiating cells. Blood. 112:568-575.

Taussig, D.C., J. Vargaftig, F. Miraki-Moud, E. Griessinger, K.

Sharrock, T. Luke, D. Lillington, H. Oakervee, J. Cavenagh, S.G. Agrawal, T.A. Lister, J.G. Gribben, and D. Bonnet. 2010. Leukemia-initiating cells from some acute myeloid leukemia patients with mutated nucleophosmin reside in the CD34(-) fraction. Blood. 115:1976-1984.

Thalmeier, K . , P. Meissner, G. Reisbach, M. Falk, A. Brechtel, and P. Dormer. 1994. Establishment of two permanent human bone marrow stromal cell lines with long-term post

irradiation feeder capacity. Blood. 83:1799-1807.

Tothova, Z., R. Kollipara, B.J. Huntly, B.H. Lee, D.H.

Castrillon, D.E. Cullen, E.P. McDowell, S. Lazo-Kallanian,

I.R. Williams, C. Sears, S.A. Armstrong, E. Passegue, R.A. DePinho, and D.G. Gilliland. 2007. FoxOs are critical mediators of hematopoietic stem cell resistance to

physiologic oxidative stress. Cell. 128:325-339.

Tsai, S., S.G. Emerson, C.A. Sieff, and D.G. Nathan. 1986.

Isolation of a human stromal cell strain secreting

hemopoietic growth factors. J Cell Physiol. 127:137-145. Tsuchiyama, J., M. Mori, and S. Okada. 1995. Murine spleen

stromal cell line SPY3-2 maintains long-term hematopoiesis in vitro. Blood. 85:3107-3116.

van Gosliga, D., H. Schepers, A. Rizo, D. van der Kolk, E.

Vellenga, and J.J. Schuringa. 2007. Establishing long-term cultures with self-renewing acute myeloid leukemia

stem/progenitor cells. Exp Hematol. 35:1538-1549.

Vanheusden, K. , S. Van Coppernolle, M. De Smedt, J. Plum, and B.

Vandekerckhove . 2007. In vitro expanded cells contributing to rapid severe combined immunodeficient repopulation activity are CD34+38-33+90+45RA. Stem Cells. 25:107-114. Vargaftig, J., D.C. Taussig, E. Griessinger, F. Anj os-Afonso, T.A. Lister, J. Cavenagh, H. Oakervee, J. Gribben, and D.

Bonnet. 2011. Frequency of leukemic initiating cells does not depend on the xenotransplantation model used. Leukemia. Wang, Y . , Y. Liu, S.N. Malek, and P. Zheng. 2011. Targeting

HIFlalpha eliminates cancer stem cells in hematological malignancies. Cell Stem Cell. 8:399-411. Xu, Q., S.E. Simpson, T.J. Scialla, A. Bagg, and M. Carroll.

2003. Survival of acute myeloid leukemia cells requires PI3 kinase activation. Blood. 102:972-980.

Yahata, T., T. Takanashi, Y. Muguruma, A. A. Ibrahim, H.

Matsuzawa, T. Uno, Y. Sheng, M. Onizuka, . Ito, S. Kato, and K. Ando. 2011. Accumulation of oxidative DNA damage restricts the self-renewal capacity of human hematopoietic stem cells. Blood. 118:2941-2950.

Yildirim, S., A.M. Boehmler, L. Kanz, and R. Mohle. 2005.

Expansion of cord blood CD34+ hematopoietic progenitor cells in coculture with autologous umbilical vein

endothelial cells (HUVEC) is superior to cytokine- supplemented liquid culture. Bone Marrow Transplant. 36:71- 79.

Zhang, J., C. Niu, L. Ye, H. Huang, X. He, W.G. Tong, J. Ross, J.

Haug, T. Johnson, J.Q. Feng, S. Harris, L.M. Wiedemann, Y.

Mishina, and L. Li. 2003. Identification of the

haematopoietic stem cell niche and control of the niche size. Nature. 425:836-841.

Zhou, B.B., H. Zhang, M. Damelin, K.G. Geles, J.C. Grindley, and

P.B. Dirks. 2009. Tumour-initiating cells: challenges and opportunities for anticancer drug discovery. Wat Rev Drug

Discov. 8:806-823.