Glucocorticoid Resistance in Paediatric B-Cell Precursor Acute Lymphoblastic Leukaemia

Glucocorticoid Resistance in Paediatric B-cell Precursor Acute Lymphoblastic Leukaemia

A thesis submitted in fulfilment of the requirements for the degree of DOCTOR OF PHILOSOPHY 2012

VIVEK BHADRI

Faculty of Medicine School of Women’s and Children’s Health University of New South Wales Sydney, Australia

ORIGINALITY STATEMENT ‘I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project’s design or conception or in style, presentation and linguistic expression is acknowledged.’

Signed …………………………………………………………

Date ………………………………………………………… Acknowledgements

Firstly, I wish to acknowledge my supervisor, Prof Richard Lock, whose support and guidance was integral to the successful completion of this project. He was always available for advice and encouraged excellence in all aspects of research and publication. I would also like thank all the scientists in the Leukaemia Biology group at CCIA who were always willing to answer any questions. In particular I am especially grateful to Rachael Papa and Rosemary O’Brien, who trained me in all laboratory techniques, and to Dr Hernan Carol, Ingrid Boehm and Kathryn Evans who helped at times with the mousework. I would also like to thank Dr Rosemary Sutton for help with identifying patients from the Study VIII database, and Dr Michelle Henderson for help with the TLDA cards. I also must thank Dr Mark Cowley, a bioinformatician from the Garvan Institute who was a wealth of information and expertise, and an invaluable help in microarray analysis.

Secondly, I wish to thank Prof Glenn Marshall for giving me the opportunity to do a research fellowship at Sydney Children’s Hospital, and Dr Toby Trahair, my co-supervisor, for advice on how to combine a clinical and research career. I am also grateful to the Leukaemia Foundation, the Steven Walter Foundation and the NHMRC for funding this work.

Thirdly, I wish to acknowledge the support of my fellow contemporary PhD students Guy Klamer, Laura High and Scott Brown. We all learned from each other, and understood the difficulties faced by each of us. The light-hearted banter certainly lifted the mood when circumstances were not going to plan.

Finally, and most importantly, I wish to thank my wife Lissa and my two daughters Sabena and Priyanka, for all the love, encouragement and support given to me daily. I am truly blessed to have such a wonderful family, and I thank them for everything.

Bibliography

Publications  Evaluation of the NOD/SCID xenograft model for glucocorticoid-regulated gene expression in childhood B-cell precursor acute lymphoblastic leukemia. Bhadri VA, Cowley MJ, Kaplan W, Trahair TN, Lock RB. BMC Genomics (2011) 12:565

 Glucocorticoid resistance in paediatric acute lymphoblastic leukaemia. Bhadri VA, Trahair TN, Lock RB. Journal of Paediatrics and Child Health epub 2011 Nov 3

Presentations  Dexamethasone-induced Gene Expression Identifies HSP90 Inhibitors as Glucocorticoid-resistance Reversing Agents in Childhood B-cell Precursor Acute Lymphoblastic Leukaemia. New Directions in Leukaemia Research Meeting 2012, Australia (Oral)

 The NOD/SCID Xenograft Model Provides Clinically-Relevant Insights Into Glucocorticoid-Induced Gene Expression In Childhood B-Cell Precursor Acute Lymphoblastic Leukemia. American Society of Hematology Annual Meeting 2010, Orlando (Poster)

Abstract

Acute lymphoblastic leukaemia (ALL) is the most common cancer in children, and is a major cause of non-accidental death in childhood. Glucocorticoids, such as prednisolone and dexamethasone, are critical agents in the treatment of ALL. Early response to prednisolone monotherapy is highly prognostic of outcome, with poor responders having significantly inferior survival rates. However the mechanisms of glucocorticoid resistance are poorly understood, and identification of novel agents to reverse resistance could improve the outcomes for this high-risk group of patients.

The aim of this study was to investigate glucocorticoid resistance using an unbiased genome-wide approach. A panel of 10 ALL xenografts derived from children classified as prednisolone poor responders (PPRs) and matched prednisolone good responders (PGRs) was created using the NOD/SCID xenograft mouse model, and the in vitro and in vivo responses to dexamethasone determined. A pilot study was undertaken which established the optimal experimental design to investigate in vivo glucocorticoid-induced gene expression, and using this setup the in vivo dexamethasone-induced gene expression profiles of the xenograft panel was determined.

Two gene signatures of glucocorticoid resistance were generated from this data, and a third signature associated with in vivo glucocorticoid sensitivity of a separate xenograft cohort was derived using historical data. Interrogation of the Connectivity Map with each signature identified HSP90 inhibitors as agents which induced a complementary gene expression signature and thus could potentially reverse glucocorticoid resistance. Testing of the HSP90 inhibitor 17- DMAG demonstrated a moderate degree of synergy with dexamethasone in 3 PPR xenografts tested in vitro, but no synergy in vivo, potentially due to the toxicity of the vehicle to the mice. Glucocorticoid resistance in ALL can potentially be overcome with HSP90 inhibitors, and next generation inhibitors warrant further evaluation in an extended panel of glucocorticoid-resistant xenografts. Table of Contents

1 Introduction ...... 1 1.1 Acute lymphoblastic leukaemia (ALL) ...... 2 1.2 Diagnosis of ALL ...... 2 1.3 Risk stratification ...... 4 1.3.1 Clinical features ...... 4 1.3.2 Cytogenetics ...... 4 1.3.3 Response to initial treatment ...... 5 1.4 Treatment of ALL ...... 6 1.5 Glucocorticoids in the treatment of ALL ...... 7 1.6 Experimental models of ALL ...... 9 1.6.1 Cell lines ...... 9 1.6.2 Primary patient cells ...... 10 1.6.3 Genetically engineered mouse models ...... 10 1.6.4 Xenografts ...... 11 1.7 Glucocorticoid action on lymphocytes ...... 13 1.7.1 Apoptosis ...... 15 1.7.2 Autophagy ...... 16 1.8 Mechanisms of glucocorticoid resistance in ALL ...... 20 1.8.1 The role of the Glucocorticoid Receptor (GR) ...... 20 1.8.2 The role of co-chaperone molecules of the GR ...... 24 1.8.3 The role of the BCL-2 family of proteins ...... 25 1.9 Gene expression analyses ...... 29 1.9.1 Glucocorticoid-regulated gene expression ...... 29 1.9.2 Gene expression associated with glucocorticoid resistance ...... 31 1.10 Therapeutic attempts to overcome glucocorticoid resistance ...... 34 1.10.1 Arsenic Trioxide...... 34 1.10.2 Rapamycin ...... 34 1.10.3 MEK inhibitors ...... 35 1.10.4 Glycolysis inhibitors ...... 35 1.10.5 NOTCH inhibitors ...... 36 1.10.6 MCL-1 inhibitors ...... 36 1.11 Aims of investigation ...... 37 2 Materials and Methods ...... 38 2.1 Tissue Culture ...... 39 2.1.1 Reagents and equipment ...... 39 2.1.2 In vitro cell culture ...... 39 2.1.3 Trypan blue exclusion assay ...... 40 2.2 Cytotoxicity Assays ...... 41 2.2.1 Reagents and equipment ...... 41 2.2.2 Cytotoxic drugs...... 41 2.2.3 MTT cytotoxic assay ...... 41 2.3 Immunoblotting and Analysis of Protein Expression ...... 44 2.3.1 Reagents and equipment ...... 44 2.3.2 Isolation of total cellular protein ...... 46 2.3.3 Protein electrophoresis, transfer and western blotting ...... 46 2.4 Isolation and Processing of RNA ...... 47 2.4.1 Reagents and equipment ...... 47 2.4.2 RNA extraction and verification ...... 47 2.4.3 RNA verification...... 48 2.4.4 RNA amplification ...... 49 2.4.5 RNA concentration and purification ...... 51 2.5 Illumina Gene Expression ...... 52 2.5.1 Reagents and Equipment ...... 52 2.5.2 Direct hybridisation assay ...... 52 2.6 Analysis of Gene Expression ...... 54 2.6.1 Normalisation and transformation...... 54 2.6.2 Differential gene expression ...... 54 2.6.3 Hierarchical clustering and correlation ...... 54 2.6.4 Functional analysis ...... 54 2.6.5 Replicate analysis ...... 55 2.6.6 Connectivity Map analysis ...... 55 2.7 cDNA Synthesis and Real-Time Quantitative PCR ...... 56 2.7.1 Reagents and equipment ...... 56 2.7.2 cDNA synthesis ...... 56 2.7.3 Real-time quantitative PCR ...... 58 2.8 Xenograft Mouse Model ...... 60 2.8.1 Reagents and equipment ...... 60 2.8.2 Inoculation of human leukaemia cells into mice ...... 60 2.8.3 Monitoring of engraftment ...... 61 2.8.4 Harvesting of cells from engrafted mice ...... 62 2.8.5 Analysis of cell surface markers ...... 62 2.8.6 Assessment of in vivo drug sensitivity ...... 62 2.9 Ethics ...... 65 2.9.1 Human Ethics ...... 65 2.9.2 Animal Ethics ...... 65 3 Establishment and Characterisation of Xenograft Panel ...... 66 3.1 Introduction ...... 67 3.2 Patient characteristics ...... 67 3.3 Engraftment characteristics ...... 70 3.3.1 ALL-26 ...... 70 3.3.2 ALL-28 ...... 70 3.3.3 ALL-50 ...... 73 3.3.4 ALL-51 ...... 73 3.3.5 ALL-52 ...... 73 3.3.6 ALL-53 ...... 78 3.3.7 ALL-54 ...... 78 3.3.8 ALL-55 ...... 82 3.3.9 ALL-56 ...... 85 3.3.10 ALL-57 ...... 85 3.3.11 Summary and discussion of engraftment characteristics ...... 88 3.4 Xenograft immunophenotyping by serial passage ...... 95 3.5 Gene expression analysis ...... 97 3.5.1 Clustering and correlation ...... 97 3.5.2 Xenograft-associated genes ...... 97 3.5.3 Summary and discussion ...... 106 3.6 Characterisation of xenograft glucocorticoid responses ...... 107 3.6.1 In vitro assessment of glucocorticoid responses ...... 107 3.6.2 In vivo assessment of glucocorticoid responses ...... 110 3.6.3 Summary and discussion of glucocorticoid responses ...... 118 3.7 Conclusions ...... 120

4 Pilot study of in vivo Glucocorticoid-Induced Gene Expression ...... 121 4.1 Introduction ...... 122 4.2 Experimental design ...... 123 4.3 Generation of in vivo gene expression ...... 126 4.4 Identification of the optimal time point ...... 127 4.5 Analysis of identified genes ...... 130 4.5.1 Early 8 hour time point ...... 130 4.5.2 Later 24 hour and 48 hour time points ...... 138 4.6 Functional analysis ...... 141 4.7 Comparison of models ...... 142 4.8 Search for GRE motifs ...... 145 4.9 Replicate analysis ...... 146 4.10 Validation of results ...... 148 4.11 Conclusions ...... 156 5 Evaluation of Glucocorticoid Resistance Reversing Agents ...... 157 5.1 Introduction ...... 158 5.2 Generation of xenograft in vivo gene expression ...... 158 5.3 Gene expression analysis ...... 159 5.3.1 Hierarchical clustering ...... 159 5.3.2 Differentially expressed genes PPR v PGR ...... 162 5.3.3 Validation of BIM ...... 168 5.4 Connectivity Map results ...... 171 5.5 Comparison with in vitro gene expression ...... 177 5.6 Evaluation of the HSP90 inhibitor 17-DMAG ...... 182 5.6.1 In vitro evaluation ...... 182 5.6.2 In vivo evaluation...... 191 5.6.3 Summary ...... 200 5.7 Evaluation of a BIM-BH3 stapled peptide ...... 206 5.8 Discussion ...... 209 6 Conclusions and Future Directions ...... 213 7 Appendices ...... 219 7.1 Appendix A – metaGSEA results ...... 220 7.2 Appendix B – Top leading edge genes from metaGSEA ...... 225 7.3 Appendix C – Gene Ontology of ALL-55/56 with dexamethasone ..... 226 8 References ...... 232 List of Figures

Figure 1.1. EFS and OS in 2628 Children with Newly Diagnosed ALL...... 3 Figure 1.2. EFS of PPR patients in the trials ALL-BFM 86, 90, and 95…...... 8 Figure 1.3. The GC-GR mechanism of action...... 14 Figure 1.4. Signaling pathways that mediate apoptosis...... 18 Figure 1.5. The BCL-2 family of proteins and BH-domains...... 19 Figure 3.1. Engraftment of ALL-26...... 71 Figure 3.2. Infiltration of organs with ALL-26 at 1o harvest...... 71 Figure 3.3. Engraftment of ALL-28...... 72 Figure 3.4. Infiltration of organs with ALL-28 at 1o harvest ...... 72 Figure 3.5. Engraftment of ALL-50...... 74 Figure 3.6. Infiltration of organs with ALL-50 at 1o harvest ...... 74 Figure 3.7. Engraftment of ALL-51...... 75 Figure 3.8. Infiltration of organs with ALL-51 at 1o harvest ...... 75 Figure 3.9. Images of ALL-52 NOD/SCID lymphadenopathy...... 76 Figure 3.10. Engraftment of ALL-52 in NS mice ...... 77 Figure 3.11. Infiltration of organs with ALL-52 in NS mice at 1o harvest...... 77 Figure 3.12. Engraftment of ALL-52 in NSG mice ...... 79 Figure 3.13. Infiltration of organs with ALL-52 in NSG mice at 1o harvest...... 79 Figure 3.14. Engraftment of ALL-53...... 80 Figure 3.15. Infiltration of organs with ALL-53 at 1o harvest ...... 80 Figure 3.16. Engraftment of ALL-54...... 81 Figure 3.17. Infiltration of organs with ALL-54 at 1o harvest ...... 81 Figure 3.18. Engraftment of ALL-55 in NS mice...... 83 Figure 3.19. Infiltration of organs with ALL-55 in NS mice at 1o harvest...... 83 Figure 3.20. Engraftment of ALL-55 in NSG mice ...... 84 Figure 3.21. Infiltration of organs with ALL-55 in NSG mice at 1o harvest...... 84 Figure 3.22. Engraftment of ALL-56...... 86 Figure 3.23. Infiltration of organs with ALL-56 at 1o harvest ...... 86 Figure 3.24. Engraftment of ALL-57...... 87 Figure 3.25. Infiltration of organs with ALL-57 at 1o harvest ...... 87 Figure 3.26. Xenograft engraftment rates by serial passage ...... 92 Figure 3.27. Engraftment time and prednisolone response ...... 94 Figure 3.28. Engraftment time and diagnostic white cell count ...... 94 Figure 3.29. Clustering of xenografts with diagnostic patient samples ...... 98 Figure 3.30. Correlation of PPR xenografts with patient samples ...... 99 Figure 3.31. Correlation of PGR xenografts with patient samples...... 100 Figure 3.32. Heatmap of pro-apoptotic genes in patient and 1o xenografts .... 104 Figure 3.33. GSEA of Clappier xenograft gene set in ALL-50 and ALL-56 ..... 105 Figure 3.34. In vitro assessment of xenograft glucocorticoid responses ...... 108 Figure 3.35. Viability after 48 hours incubation with dexamethasone...... 109 Figure 3.36. In vivo assessment of PPR response to dexamethasone ...... 111 Figure 3.37. In vivo assessment of PGR response to dexamethasone ...... 114 Figure 3.38. In vivo objective response to dexamethasone ...... 115 Figure 3.39. In vitro assessment of glucocorticoid sensitivity of ALL-56 ...... 117 Figure 4.1. Time course of NALM-6 on exposure to dexamethasone...... 124 Figure 4.2. Volcano plots of significantly differentially expressed genes ...... 129 Figure 4.3. ALL-3 engraftment and harvest data...... 131 Figure 4.4. Parametric GSEA of top 100 glucocorticoid-induced gene sets .... 144 Figure 4.5. Recovery scores at 8 hours and 24 hours ...... 147 Figure 4.6. mRNA expression of BMF ...... 150 Figure 4.7. BMF protein expression in ALL-3 treated with dexamethasone. ... 150 Figure 4.8. mRNA expression of pro-apoptotic BCL-2 family members ...... 151 Figure 4.9. BIM protein expression in ALL-3 treated with dexamethasone. .... 152 Figure 4.10. mRNA expression of pro-survival BCL-2 family members ...... 153 Figure 4.11. BCL-2 protein expression in ALL-3 following dexamethasone .... 154 Figure 4.12. mRNA expression of the GR and downstream targets ...... 155 Figure 5.1. Clustering of dexamethasone-induced gene expression profiles .. 161 Figure 5.2. Genes differentially expressed between PGRs and PPRs ...... 163 Figure 5.3. Genes regulated by dexamethasone in MAP kinase pathways .... 165 Figure 5.4. BIM and BMF mRNA induction by dexamethasone ...... 167 Figure 5.5. BIM protein induction following dexamethasone ...... 169 Figure 5.6. Quantification of BIM mRNA and protein induction ...... 170 Figure 5.7. mRNA expression of HSP90 ...... 176 Figure 5.8. In vivo/vitro dexamethasone responses in ALL-55 and ALL-56. ... 179 Figure 5.9. In vitro assessment of 17-DMAG in PPRs ...... 184 Figure 5.10. In vitro assessment of 17-DMAG in ALL-2/19 and cell lines ...... 185 Figure 5.11. In vitro assessment of 17-DMAG ± dexamethasone ...... 187 Figure 5.12. HSP70/AKT in ALL-55 and ALL-57 following 17-DMAG in vitro. 189 Figure 5.13. Quantification of HSP70/AKT in ALL-55 and ALL-57 ...... 190 Figure 5.14. Toxicity study I of 17-DMAG ± dexamethasone ...... 192 Figure 5.15. Toxicity study II of 17-DMAG ± dexamethasone ...... 195 Figure 5.16. Regulation of HSP70/AKT by 17-DMAG in ALL-2 ...... 197 Figure 5.17. Quantification of HSP70/AKT in ALL-2...... 198 Figure 5.18. Efficacy of 17-DMAG ± dexamethasone in ALL-55...... 201 Figure 5.19. Efficacy of 17-DMAG ± dexamethasone in ALL-57 ...... 202 Figure 5.20. HSP70/AKT in ALL-55 and ALL-57 following 17-DMAG in vivo. . 204 Figure 5.21. Quantification of HSP70/AKT in ALL-55 and ALL-57 ...... 205 Figure 5.22. SAHB structures ...... 206 Figure 5.23. Evaluation of SAHBs in vitro ...... 208 Figure 5.24. The role of HSP90 in the hallmarks of cancer ...... 210 List of Tables

Table 2.1. Optimised cell densities for MTT cytotoxicity assay ...... 43 Table 2.2. Antibodies used for immunoblotting...... 45 Table 2.3. Reverse Transcription Master Mix ...... 50 Table 2.4. Second Strand Master Mix ...... 50 Table 2.5. IVT Master Mix ...... 50 Table 2.6. cDNA Synthesis Mix 1 ...... 57 Table 2.7. cDNA Synthesis Mix 2 ...... 57 Table 2.8. TLDA Card of BCL-2 family members ...... 59 Table 2.9. PPTP Objective Response Measure scoring method ...... 64 Table 3.1. Patient characteristics of xenograft panel ...... 69 Table 3.2. Summary of xenograft engraftment kinetics...... 89 Table 3.3. Comparison of engraftment in NOD/SCID and NSG mice...... 90 Table 3.4. Xenograft immunophenotyping by serial passage...... 96 Table 3.5. Gene Ontology terms upregulated in 1o xenografts ...... 102 Table 3.6. Gene Ontology terms downregulated in 1o xenografts ...... 103 Table 3.7. Blast reduction and MRD levels...... 112 Table 3.8. Summary of dexamethasone responses ...... 119 Table 4.1 Number of differentially expressed genes by FDR...... 128 Table 4.2 Number of differentially expressed genes by FC ...... 128 Table 4.3. Genes upregulated 8 hours following dexamethasone ...... 133 Table 4.4. Genes downregulated 8 hours following dexamethasone ...... 137 Table 4.5. Genes regulated 24 hours following dexamethasone ...... 139 Table 4.6. Genes regulated 48 hours following dexamethasone ...... 140 Table 5.1. Number of differentially expressed probes in vivo...... 160 Table 5.2. Numbers of significantly differentially expressed genes...... 164 Table 5.3. CMap results comparing ALL-28 and ALL-57 with PGRs...... 173 Table 5.4. CMap results comparing ALL-55 with ALL-56...... 173 Table 5.5. CMap results of original xenograft panel...... 174 Table 5.6. Number of differentially expressed probes in vivo vs in vitro ...... 178 Table 5.7. CMap results comparing in vitro ALL-55 with ALL-56...... 181 Table 5.8. Summary of 17-DMAG in vitro responses...... 186 Table 5.9. Combination indices of 17-DMAG ± dexamethasone...... 188 Table 5.10. Toxicity studies - haematology and biochemistry results...... 193 Table 5.11. LGD/ORM of ALL-55/ALL-57 with 17-DMAG ± dexamethasone. . 203

1 Introduction

Page | 1

1.1 Acute lymphoblastic leukaemia (ALL) Acute lymphoblastic leukaemia (ALL), a haematological disorder arising from the bone marrow characterised by the uncontrolled proliferation of lymphoid progenitor cells, is the most common cancer affecting the paediatric population. Approximately 170 new cases in children under the age of 14 are diagnosed each year in Australia (AIHW: McDermid 2005) of which 80 are diagnosed in New South Wales (Tracey 2008). Despite significant advances in therapeutics cancer remains the major cause of non-accidental mortality in childhood (ABS 2007). The clinical outcomes for ALL have shown a steady improvement over the last 4 decades through the development of serial clinical trials in multicentre co-operative groups, and event-free survival (EFS) now exceeds 80% and overall survival (OS) exceeds 85% (Pui and Evans 2006), Figure 1.1. Despite these figures, nearly 20% of children with ALL will relapse, and survival after relapse is poor, particularly in high-risk patients (Roy et al. 2005).

1.2 Diagnosis of ALL There are no pathognomonic symptoms or signs which indicate that a child has ALL. Common presenting symptoms include non-specific lethargy and malaise, loss of appetite, bone pain or limp, unexplained fever, bleeding or bruising. On examination the child may look pale or listless, and may have petechiae or bruising on the skin, palpable lymphadenopathy or hepatosplenomegaly, and evidence of papilloedema (from central nervous system (CNS) disease) or testicular enlargement (from testicular infiltration).

The diagnosis of ALL is made by peripheral blood examination, which may reveal signs of bone marrow failure (anaemia, neutropaenia, thrombocytopaenia) and hyperleucocytosis from circulating leukaemic blasts. Absolute confirmation is through a bone marrow biopsy, both for morphological assessment of the leukaemia, and for obtaining sufficient diagnostic material for immunophenotyping, cytogenetics and molecular testing. Children also undergo a lumbar puncture to assess for the presence of leukaemic blasts in the cerebrospinal fluid (CSF) indicating CNS disease.

Page | 2

Figure 1.1. Event-free and overall survival in 2628 Children with Newly Diagnosed ALL. The patients participated in 15 consecutive studies conducted at St. Jude Children's Research Hospital from 1962 to 2005. The five-year event-free (A) and overall survival (B) estimates (±SE) are shown, except for Study 15, for which preliminary results at four years are provided (Pui and Evans 2006).

Page | 3

1.3 Risk stratification ALL is a heterogeneous disease, and optimal treatment is based on stringent application of prognostic factors for risk-directed therapy in clinical trials. Assessment of risk of relapse is based on a combination of clinical and genetic factors at diagnosis and more importantly, response-based criteria.

1.3.1 Clinical features A number of adverse prognostic factors were identified in early clinical trials, including age <1 year or >10 years, male sex, T-ALL immunophenotype, and high diagnostic white cell count (WCC). Young age remains a significant adverse factor and children <12 months of age are treated on separate infant ALL protocols. In the US, National Cancer Institute (NCI) criteria stratify children aged >10 years or with a diagnostic WCC >50 x 109/L as high risk, whereas these criteria are no longer used in contemporary European protocols.

1.3.2 Cytogenetics A number of recurring genetic abnormalities of leukaemic blasts have important prognostic significance, particularly in B-cell precursor (BCP) ALL. Abnormalities of chromosomal number associated with a favourable outcome include high hyperdiploidy (51-65 chromosomes per cell) (Paulsson and Johansson 2009), and trisomies 4, 10 and 17 (Sutcliffe et al. 2005). In contrast, an inferior outcome is seen with decreasing chromosome number, or hypodiploidy (<44 chromosomes per cell) (Nachman et al. 2007).

There are three main recognised chromosomal translocations with prognostic significance. The ETV6-RUNX1 t(12;21) fusion is seen in 20-25% of cases of BCP-ALL and is generally associated with a favourable prognosis, but its impact is modified by response and treatment regimen (Madzo et al. 2003). The BCR- ABL1 t(9;22) Philadelphia chromosome, which results in a constitutively active fusion protein with tyrosine kinase activity, is seen in approximately 3% of cases of BCP-ALL. The presence of this translocation has historically been associated with a dismal prognosis and is considered an indication for allogeneic haemopoietic stem cell transplantation in first remission. However the

Page | 4

introduction of tyrosine kinase inhibitors such as imatinib has dramatically improved disease free survival (Schultz et al. 2009), although whether allogeneic transplantation can be avoided remains to be determined. Translocations involving the MLL gene at chromosomal location 11q23 occur in up to 5% of childhood ALL cases, but occur in over 50% of infant ALL cases, and also confer an inferior prognosis (Pieters et al. 2007).

Recent application of genome-wide analysis of gene expression and DNA copy number has identified a specific subset of children with high-risk BCP-ALL and poor outcomes. These patients have a gene expression signature similar to that of patients with Philadelphia positive (Ph+) ALL but lack the translocation. IKZF1 deletions were found in 30% of these high-risk patients and were significantly associated with a very poor outcome (Mullighan et al. 2009). The same group identified activating JAK mutations in 10% of Ph- high-risk BCP- ALL, the majority of which were associated with IKZF1 deletions and a poor prognosis (Mullighan et al. 2009). IKZF1 codes for the lymphoid transcription factor Ikaros, and JAK mutations lead to activated JAK-STAT signalling and growth factor independence. Translocations resulting in overexpression of cytokine receptor-like factor 2 (CRLF2) have also been identified as an adverse factor in high-risk BCP-ALL, and are also associated with JAK mutations (Harvey et al. 2010). Together, these observations suggest that CRLF2 overexpression, JAK mutations and alterations in IKZF1 co-operate to promote B-cell leukaemogenesis.

1.3.3 Response to initial treatment A number of early clinical trials have identified that the rate of clearance of leukaemia in the early stages of treatment is highly prognostic, reflecting the chemosensitivity of the disease. The timepoints of assessment vary, but include peripheral blood assessment for clearance of circulating blasts, and bone marrow examination for residual leukaemia after 7 and/or 14 days of treatment with multi-agent chemotherapy, and after 30-35 days at the end of induction.

Page | 5

Morphological assessment of residual leukaemia in the bone marrow by conventional light microscopy is relatively insensitive, and traditionally a cutoff of <5% (1:20) leukaemic blasts is used to define remission status. To detect lower levels of leukaemia in peripheral blood or bone marrow, specialised techniques are required, usually PCR- or flow cytometry-based methods to detect minimal residual disease (MRD). PCR-based assays detect unique immunoglobulin and/or T-cell receptor (Ig/TCR) rearrangements, whereas flow- cytometry based assays detect cells with the leukaemia-specific immunophenotype. These techniques can detect leukaemia cells at the level of 1:10000 to 1:100000, significantly increasing the sensitivity of detection of residual disease. MRD has been shown to be highly prognostic (van Dongen et al. 1998; Conter et al. 2010), and is now incorporated into most contemporary childhood ALL protocols for risk stratification and treatment modification.

1.4 Treatment of ALL Protocols used to treat ALL are made up of distinct phases of treatment comprising multiple chemotherapeutic agents with different mechanisms of action, for a total duration of 2 years. Treatment begins with a 3 or 4 drug induction, with the aim of achieving clinical remission within the first 4 to 5 weeks. Induction comprises of a glucocorticoid (prednisolone or dexamethasone), vincristine and L-asparaginase, with many protocols adding an anthracycline such as daunorubicin. This is followed by consolidation, consisting of cyclophosphamide, cytarabine and 6-mercaptopurine, a CNS- directed phase with high-dose methotrexate, re-induction and re-consolidation (with agents closely related to those used initially) and then maintenance with 6- mercaptopurine and methotrexate. Throughout treatment, children also receive intrathecal injections of methotrexate for treatment and prophylaxis of CNS- disease, and those with T-ALL or overt CNS-disease at diagnosis may also receive 12 Gy cranial radiation therapy. In total up to 11 different agents are used with the aim of eliminating residual leukaemic blasts and effecting cure.

Page | 6

1.5 Glucocorticoids in the treatment of ALL Glucocorticoids such as prednisolone and dexamethasone are critical components of multi-agent chemotherapy regimens used in the treatment of ALL (Gaynon and Carrel 1999), and form the backbone of the induction and re- induction phases of treatment. It has been known for over 50 years that glucocorticoids can induce remission in childhood ALL (Hyman and Sturgeon 1956) though even today their mechanism of action is not completely understood. The Berlin-Frankfurt-Münster (BFM) group has identified that the initial response to prednisolone is a major prognostic factor (Riehm et al. 1987; Dordelmann et al. 1999; Schrappe et al. 2000). BFM-based ALL therapy involves an initial 7 day treatment with prednisolone 60 mg/m2/day and a single age-related dose of intrathecal methotrexate. Those patients with a day 8 peripheral blast count of <1 x 109/L are termed Prednisolone Good Responders (PGRs), whilst those with a day 8 peripheral blast count of ≥1 x 109/L are termed Prednisolone Poor Responders (PPRs) and account for approximately 10% of children with ALL. PPRs are stratified into high-risk arms of the protocols and may undergo allogeneic stem cell transplantation in first remission.

Figure 1.2 shows that the outcomes for PPRs in sequential trials for ALL (Moricke et al. 2008) are poor, with EFS significantly inferior to those seen in Figure 1.1. At relapse, de novo or acquired resistance to glucocorticoids is common and disproportionate to other chemotherapeutic drugs (Klumper et al. 1995). Early intervention strategies to reverse glucocorticoid resistance could significantly improve the chances of cure for these patients.

Page | 7

Figure 1.2. Kaplan-Meier estimate of event-free survival of patients with prednisone poor-response and/or non-remission on day 33 in the trials ALL-BFM 86, 90, and 95. Log-rank tests: ALL-BFM 86 versus 90, P=0.029; ALL-BFM 86 versus 95, P=0.14; ALL-BFM 90 versus 95, P<0.001. 6y-pEFS indicates probability of event-free survival at 6 years; SE, standard error (Moricke et al. 2008).

Page | 8

1.6 Experimental models of ALL There are four main experimental models used to investigate mechanisms and processes in ALL – cell lines, primary patient cells, genetically engineered mouse models and xenografts created in immunodeficient mice.

1.6.1 Cell lines The vast majority of research in ALL has been undertaken using cultured cell lines. Initially derived from patients, cell lines have undergone multiple passages and transformations to become immortalised. They have the advantage of survival and proliferation in vitro, an unlimited supply and availability of identical material, and can be stored indefinitely in liquid nitrogen and recovered. Cell lines are also readily transduced by lentiviral or retroviral techniques, allowing for the study of the biological function of individual genes.

The first solid tumour-derived cell line, HeLa, was established from a uterine cervix carcinoma in 1951 (Scherer et al. 1953), and the first haematopoietic cell lines, derived from Nigerian patients with Burkitts Lymphoma, were established in 1963 (Pulvertaft 1964). There are now hundreds of well characterised leukaemia and lymphoma cell lines, covering almost all lymphoid and myeloid disease subtypes. The common characteristics of cell lines are (1) they are monoclonal, but subclones may emerge during extended culture; (2) they have arrested maturation at a discrete stage; (3) they undergo sustained autonomous proliferation in culture; (4) they harbour a number of genetic alterations which provide the cell with a survival and proliferative advantage and (5) the salient features remain stable in long-term culture (Drexler et al. 2000).

A number of limitations exist with cell line-based studies. (1) Cultured cells can be easily contaminated with mycoplasmas, which can produce extensive changes and growth arrest in the cultures they infect, and cultures need to be regularly screened; (2) laboratory cell lines can often be cross-contaminated with other cell lines and/or misidentified, and it is now a prerequisite for publication that studies use cell lines authenticated by DNA single tandem repeat ‘fingerprinting’; (3) cell lines may have genetic instability, and acquire

Page | 9

mutations during in vitro culture, particularly in p53, which may be a prerequisite for immortalisation (Drexler et al. 2000). Further, cell line studies will always suffer from the lack of fundamental interactions with the host that provide stromal support to cancer cells. Thus mechanisms elucidated from cell lines may not accurately represent those seen in patients.

1.6.2 Primary patient cells Primary patient cells extracted from bone marrow aspirates are the most direct representation of human leukaemia, but these cells have a finite supply, rarely survive ex vivo for more than a few days and are thus used primarily for baseline genetic/genomic analysis or short term in vitro cytotoxicity experiments.

1.6.3 Genetically engineered mouse models The development of many cancers through the activation of oncogenes or the loss of tumour suppressor genes is well characterised. By the late 1980s, it became possible to inactive genes in the mouse germline through homologous recombination in embryonic stem cells, resulting in a cancer-prone strain in which one or both copies of a tumour suppressor gene were mutated. An early example was the development of a p53-/- strain in which the mice were predisposed to lymphomas and sarcomas (Donehower et al. 1992). Over the last 20 years transgenic and knockout alleles have been combined to yield numerous well-defined murine models, including the Eµ-Myc transgenic model of lymphoma and the N-Myc transgenic model of neuroblastoma, which have contributed greatly to the understanding of cancer biology. Current techniques using tetracycline-regulation and CRE-inducible alleles allow the timing, duration and tissue compartment of gene expression or inactivation to be further controlled, and thus potential targets can be validated. Genetically engineered mouse models (GEMMs) can also be used to evaluate the ability of novel agents to induce regression of established tumours.

There are certain limitations to GEMMs, as important biological differences exist between murine and human malignancies. For example, mice have a different cytogenetic structure, and co-deletion of adjacent genes may influence tumour

Page | 10

development. Mice also have different telomere biology which could affect the development of secondary cytogenetic events during tumour progression (Artandi et al. 2000). However GEMMs are likely to become increasingly important in preclinical drug testing.

1.6.4 Xenografts A widely used experimental model is the non-obese diabetic/severe combined immunodeficient (NOD/SCID) mouse model. These mice lack functional T and B cells, have low natural killer (NK)-cell function and an absence of circulating complement, and are highly receptive to engraftment of primary childhood ALL samples (Baersch et al. 1997). In a series of experiments, Lock et al inoculated a cohort of 19 patient ALL samples from both T and B lineage, all disease stages and risk categories into NOD/SCID mice, and created a panel of xenografts. Engraftment was monitored using serial blood sampling by tail-vein injection, and cells were harvested from spleen and bone marrow. They demonstrated that when compared to the primary patient cells, the engrafted cells retained the same morphological characteristics, do not acquire or require p53 mutations, and undergo minimal changes in immunophenotype. They also showed that cells from patients in relapse engrafted marginally more quickly than those taken at diagnosis, and that the in vivo sensitivity to vincristine was significantly correlated to the patients length of first remission (CR1). Together this suggests that the xenograft model reflects the experience of the patients, providing a clinically relevant model of ALL (Lock et al. 2002).

This group subsequently further characterised the in vivo and in vitro responses of a sub-panel of 10 xenografts to the established drugs vincristine, dexamethasone and methotrexate. The xenografts demonstrated a broad range of sensitivity, but when stratified according to outcome (alive more than 4.5 years from diagnosis or died of disease) there was a significant correlation between sensitivity to vincristine and dexamethasone individually (but not methotrexate) and clinical outcome. This model thus has the strengths of (1) the development of a panel of continuous ALL xenografts from diverse patient subtypes; (2) their propagation as models of systemic disease; (3) their

Page | 11

retention of the fundamental biologic characteristics of the original disease; (4) the ability to monitor engraftment and response to therapy in ‘real-time’; and (5) their spectrum of sensitivity to established drugs providing an overall reflection of clinical outcome, and is thus a clinically relevant model for the study of ALL.

The NOD/SCID xenograft mouse model also provides a robust model for the in vivo evaluation of drug responses, and has been selected by the Pediatric Preclinical Testing Program (PPTP) for the evaluation of novel therapies in ALL (Houghton et al. 2007). Molecular characterisation of the entire PPTP panel of multiple tumour types has demonstrated that ALL xenografts cluster with ALL cell lines and primary samples from patients with ALL, reflecting that ALL xenografts accurately represent the disease (Neale et al. 2008).

Although the NOD/SCID strain is highly receptive to engraftment with human ALL, some samples can take up to 6 months to engraft (Lock et al. 2002) and such xenografts are impractical for experimental purposes. Furthermore, this strain has a high frequency to develop spontaneous thymic lymphomas by the age of 40 weeks, resulting in early death (Prochazka et al. 1992). To overcome these problems, a number of NOD/SCID-derived mice strains have been developed, including the NOD/SCID/IL-2Rγnull (NSG) strain. NSG mice lack the IL-2R gamma-chain, the absence of which blocks NK-cell differentiation removing a major obstacle preventing efficient engraftment of primary human cells (Shultz et al. 2005), and are now increasingly used in multiple human malignant and non-malignant xenograft models.

There are, however, a number of limitations to the use of xenografts in mice. (1) Current murine models retain some ability to reject human tumours, and are thus inadequate for the study of tumour initiating cells (Kelly et al. 2007); (2) the use of immunocompromised mice eliminates the ability to study modulators of anti-tumour immunity; (3) xenografts are made up of homogeneous cells and do not necessarily replicate the complex nature of human malignancies and dynamic host-tumour interactions; (4) the limited number of xenografts in any one study represents only a fraction of the genetic and epigenetic mutations Page | 12

seen in the patients with the disease; (5) inter-species pharmacokinetic differences may mean that drug doses used in mice may not represent clinically achievable drug concentrations in humans. Further, it has proved technically challenging to lenti- or retrovirally transduce ALL xenograft cells, and studies to validate genes of interest in ALL xenografts have so far proved impossible. However, they remain a crucial part of preclinical studies in ALL, particularly as xenografts can be created from specific patient groups with known high risk features or genetic mutations, enabling the evaluation of targeted therapies.

1.7 Glucocorticoid action on lymphocytes Glucocorticoids have diverse effects on multiple cell types, and can act as transcriptional activators, repressors or both depending on the target cell, schematically shown in Figure 1.3. Glucocorticoids are steroid hormones that act on their target cell through a specific cytosolic glucocorticoid receptor (GR) (Baxter et al. 1971). The GR is maintained in a complex with co-chaperone molecules. Heat shock protein 70 (HSP70) and HSP40 bind to the cytosolic GR, with BAG-1 (BCL-2-associated gene product-1) and HIP (HSP70-interacting protein) acting as positive and negative regulators, respectively. HSP90 and HOP (HSP70/HSP90-organising protein) bind to this complex leading to opening of the glucocorticoid (GC)-binding cleft of the GR (Pratt and Toft 2003). This complex is stabilised by p23 (protein-23), FKBP-51 and FKBP-52 (FK506- binding proteins-51 and 52), and CYP-40 (cyclophilin D) (Ratajczak et al. 2003).

Glucocorticoids enter the cell through passive diffusion. Upon ligand binding, the GC-GR dissociates from chaperone complex, dimerises and is then transported to the nucleus. This translocation is dependent on HSP90 and FKBP52 which facilitate binding of the complex to dynein, part of the microtubule-based movement machinery (Galigniana et al. 2002). FKBP51 inhibits this translocation, maintaining the cytoplasmic location of the complex (Davies et al. 2002). HSP90 stabilises binding of the GC-GR complex to the promoter of GR-responsive genes whereas p23 induces removal of the complex (Stavreva et al. 2004).

Page | 13

Nuclear translocation of the activated GR results in transactivation of target genes via a direct interaction with specific palindromic DNA sequences known as Glucocorticoid Response Elements (GREs) (Yamamoto and Alberts 1976), transrepression of gene activation via interaction with negative GREs (Drouin et al. 1987), and repression of gene activation via interactions between the GR and other transcription factor complexes such as activator protein-1 (AP-1) and nuclear factor κB (NF-κB) (Ray and Prefontaine 1994). These changes in gene expression induce a number of often competing biological processes on a wide range of cells, including effects on metabolism, proliferation, cell survival and cell death.

Figure 1.3. The GC-GR mechanism of action (Tissing et al. 2005). See text for details.

Page | 14

1.7.1 Apoptosis It has been recognised for many years that glucocorticoids induce lymphocyte and lymphoblast cell death primarily through apoptosis, or programmed cell death, and this forms the basis of their inclusion in all contemporary ALL protocols worldwide. The morphological changes of apoptosis are characterised by blebbing, cell shrinkage, nuclear fragmentation, chromatin condensation and chromosomal DNA fragmentation. These changes were recognised in glucocorticoid-treated lymphoid cells over 30 years ago (Robertson et al. 1978) and subsequent reports using actinomycin-D and cycloheximide concluded that glucocorticoid-induced death of rat thymocytes was dependent on de novo transcription and protein synthesis (Wyllie et al. 1984).

Mammals have two distinct signaling pathways that mediate apoptosis, and these converge on activation of effector caspases (Figure 1.4). The extrinsic pathway is activated when death ligands, such as CD95L/FAS ligand, bind their cognate death receptors. It requires activation of caspase-8 by the CD95 adaptor FAS-associated protein with death domain (FADD) to cause effector caspase activation and cell death. The intrinsic pathway, regulated by the BCL- 2 family is triggered by many developmental cues, including cytokine withdrawal or cytotoxic stimuli. The BCL-2 family of proteins possess either pro-apoptotic or pro-survival properties and are related through conserved sequence motifs known as BCL-2 homology (BH) domains (Figure 1.5). BH3-only proteins are a group of pro-apoptotic proteins related by the single 9-16 amino acid BH3 region of homology. The BH3 domain forms an amphipathic α-helix that binds with high-affinity to a hydrophobic cleft on pro-survival members formed by the BH1 and BH2 domains (Huang and Strasser 2000).

Death stimuli activate BH3-only proteins which initiates apoptosis by a mechanism that requires BAX/BAK (Zong et al. 2001). This pathway leads to disruption of the outer mitochondrial membrane, resulting in release of cytochrome c and other apoptogenic proteins. After its release into the cytosol, cytochrome c promotes APAF1-mediated activation of caspase-9, which in turn leads to activation of effector caspases. In addition, the BCL-2 family regulates Page | 15

additional pathways to apoptosis, such as activation of other caspases and/or caspase-independent cell-death processes (Strasser 2005). Glucocorticoids induce apoptosis in lymphocytes through the intrinsic pathway, and a significant body of literature has attempted to elucidate the roles of the various proteins involved in this process (Section 1.8.3).

1.7.2 Autophagy The majority of work in glucocorticoid-induced cell death has focused on the stimuli and mechanisms that induce cell death by apoptosis. Evidence is accumulating that non-apoptotic programmed cell death (PCD), such as mitotic catastrophe, senescence, necrosis-like PCD and autophagy, may also play a role in the tumour response to anti-cancer drugs (Gozuacik and Kimchi 2004; Jaattela 2004). Autophagy is a well-conserved process, regulating normal cytoplasmic and organelle turnover, and is characterised by the formation of autophagosomes: double- or multiple-membrane-surrounded cytoplasmic vesicles engulfing cytoplasmic organelles such as mitochondria and endoplasmic reticulum. Subsequently autophagosomes fuse with lysosomes and their contents are degraded by lysosomal enzymes (Gozuacik and Kimchi 2004).

Laane et al studied the role of autophagy in dexamethasone-induced cell death. RS4;11 cells were cultured in the presence or absence of dexamethasone, and the appearance of autophagosomes at 24 hours post-treatment was demonstrated by electron microscopy. They also showed a change in the pattern of microtubule-associated protein light chain 3, LC3, consistent with the initiation of autophagy, and demonstrated that the formation of autophagic vacuoles in response to dexamethasone precedes the activation of BAX and apoptosis. Detailed experiments further showed that chemical and genetic inhibition of autophagosomes formation led to a decrease in the apoptotic response, showing that dexamethasone-induced autophagy not only lies upstream of apoptosis but is also required for the latter to occur (Laane et al. 2009).

Page | 16

There is evidence that the BCL-2 family members can engage in cross-talk between the apoptotic and autophagic pathways, as they were shown to associate with the autophagy-regulator BECLIN-1 (Pattingre et al. 2005). Bonapace et al demonstrated that subcytotoxic concentrations of the small molecule MCL-1 inhibitor obatoclax restored the response to dexamethasone in glucocorticoid-resistant ALL by triggering a non-apoptotic cell death pathway. Obatoclax induced disruption of BECLIN-1 from MCL-1 and the combination of dexamethasone with obatoclax was associated with inhibition of mTOR activity. They provide detailed genetic and pharmacological evidence that sensitisation to dexamethasone occurred via autophagy-dependent cell death which bypassed the block in apoptosis. They showed that inhibition of autophagy in glucocorticoid-sensitive ALL cell lines and patient samples did not affect the response to glucocorticoids, suggesting that autophagy is not required for glucocorticoid-induced cell death in glucocorticoid-sensitive cells (Bonapace et al. 2010).

Page | 17

Figure 1.4. Signaling pathways that mediate apoptosis. The BCL-2 regulated (intrinsic) pathway shown on the left; death-receptor (extrinsic) pathway shown on the right (Strasser 2005).

Page | 18

Figure 1.5. The BCL-2 family of proteins and BH-domains (Strasser 2005).

Page | 19

1.8 Mechanisms of glucocorticoid resistance in ALL The mechanism of resistance to glucocorticoids in ALL has been extensively studied, using both cell line and patient-derived xenograft models, with varying concordance with results obtained from primary patient samples. Studies using cultured cell lines have shown that glucocorticoid resistance is almost invariably associated with defects at the level of the GR leading to impaired ligand- receptor interactions, but this has rarely been demonstrated in clinically relevant models.

1.8.1 The role of the Glucocorticoid Receptor (GR) Gruber et al generated an in vitro model using the glucocorticoid-sensitive T- ALL cell line CEM-C7H2-2C8. Under selection culture in 1µM dexamethasone, a glucocorticoid-resistant clone was derived and transduced with a lentiviral vector for tetracycline-dependent expression of human GR, termed CEM-C7H2- 2C8-R42D9-1B11. This clone was then treated with combinations of doxycycline and dexamethasone, and showed a dependence of the glucocorticoid dose required for apoptosis induction on the GR level. At low doses of doxycycline, the extent of apoptosis was minimal despite supra- physiological doses of dexamethasone, but as the doxycycline dose increased the clone demonstrated sensitivity to dexamethasone at concentrations as low as 10 nM, suggesting that in this model glucocorticoid-sensitivity is dependent on GR levels (Gruber et al. 2009).

Schmidt et al generated glucocorticoid-resistant subclones from CCRF-CEM- C7H2 T-ALL and B-cell precursor PreB697 cell lines by limiting dilution in the presence of glucocorticoid, and compared responses to glucocorticoid with the response of the glucocorticoid-sensitive parental lines. The glucocorticoid- sensitive lines showed marked increase in expression levels of GR and the downstream target GILZ, but in the resistant lines this was absent or marginal. When two resistant subclones were transduced with a retroviral construct expressing human wild-type GRα glucocorticoid sensitivity was restored. Further analysis of the GR in this model revealed a high frequency of GR mutations, particularly in the CCRF-CEM-C7H2 cell line (Schmidt et al. 2006).

Page | 20

In a model using CEM-C1, a glucocorticoid-resistant T-ALL cell line and its glucocorticoid-sensitive sister CEM-C7H2, Geley et al demonstrated that the CEM-C1 cells demonstrated markedly reduced glucocorticoid binding by radioligand assay. It was also shown that both glucocorticoid-sensitive and glucocorticoid-resistant strains had identical GR alleles, being compound heterozygotes for the L753F mutation and wild type. When CEM-C1 were stably transfected with rat GR that is known to function in human cells, glucocorticoid resistance was reversed and GR overexpressing clones underwent apoptosis roughly corresponding to the level of GR (Geley et al. 1996).

Glucocorticoid resistance has also been associated with resistance to other chemotherapeutic agents. In CEM-MTX-R3 cells, a methotrexate-resistant cell line derived from parental CEM cells, there was a 19,000 fold increase in dexamethasone resistance compared to wild-type. It was further shown that the CEM-MTX-R3 cells had less than half the number of GR binding sites, a nearly 40% decrease in GR protein expression when compared to wild-type, and expressed only the L753F mutant receptor allele (Catts et al. 2001).

In contrast, a recent study of primary patient samples demonstrated no significant differences in GR mRNA expression levels or GR isoforms at baseline or on exposure to glucocorticoid when comparing in vitro glucocorticoid-resistant and -sensitive leukaemic blasts (Tissing et al. 2006). It has also been reported that there was no significant difference in isoform non- specific GR protein expression when comparing diagnostic primary patient leukaemic blasts obtained from 40 matched PPRs and PGRs (Lauten et al. 2003).

There is conflicting evidence that glucocorticoid sensitivity is correlated with isoforms of the GR. Alternate splicing of the GR gene results in five isomers, GRα, GRβ, GRγ, GR-P and GR-A, of which GRα and GRβ are the most well characterised isoforms. GRα is the functional isomer (Hollenberg et al. 1985), whereas GRβ does not bind glucocorticoid, is transcriptionally inactive and potentially functions as a dominant negative inhibitor of GRα activity (Oakley et Page | 21

al. 1996). In a study of GR isoform expression in primary patient derived leukaemic blasts, there was a non-significant correlation between GRα mRNA expression and in vitro glucocorticoid sensitivity, and a significant inverse correlation between the GRα/GRβ ratio and the proportion of apoptotic blasts on exposure to glucocorticoid (Koga et al. 2005). In contrast, a study also using primary patient derived leukemic blasts found no correlation between GRα, GRβ or GRα/GRβ ratios at the mRNA or protein level with in vitro glucocorticoid resistance (Haarman et al. 2004).

In a study using 9 T-ALL cell lines grown in the absence of drug selection pressure, there was no correlation between resting levels of GRα or GRβ and glucocorticoid resistance. Single nucleotide polymorphisms (SNP) analysis of the GRα coding region identified no mutations known to be linked with glucocorticoid resistance (Beesley et al. 2009a). Thus, in contrast to most other studies using cell lines, this group demonstrated that GR mutation was not a common mechanism of naturally occurring glucocorticoid resistance in their cell line model.

Using patient derived xenografts, Bachmann et al demonstrated that in ALL-2, ALL-4, ALL-7, ALL-10, ALL-18 and ALL-19, xenografts that exhibit high-level in vitro glucocorticoid resistance, GR downregulation was not observed. Further, these xenografts also demonstrated normal numbers of binding sites and receptor affinity for ligand (Bachmann et al. 2005). This group also showed that GR translocation to the nucleus on exposure to dexamethasone was comparable in sensitive (ALL-3 and ALL-16) and resistant (ALL-7 and ALL-19) xenografts. Together this suggests that in this model glucocorticoid resistance was not occurring at the level of the GR.

Page | 22

Early studies using the CEM T-ALL cell line identified GR mutations in glucocorticoid-resistant derivatives. The most well-studied of these is the L753F mutation which encodes a labile receptor that does not bind ligand under physiological conditions (Powers et al. 1993). The same group studied archival material obtained from the patient CEM during the course of her therapy (Foley et al. 1965) and identified the presence of the same mutation in a small proportion of cells, establishing the presence of GR mutation in a patient with ALL (Hillmann et al. 2000).

Tissing et al analysed reported genetic SNPs in the GR in diagnostic patient samples from 42 PPRs and 15 PGRs. No somatic mutations were detected in the GR gene coding region, and although six SNPs were identified in the GR gene coding region in leukaemic blasts, none were significantly associated with either in vivo or in vitro glucocorticoid resistance (Tissing et al. 2005).

In contrast, a recent study of 242 childhood ALL cases using SNP analysis and genomic DNA sequencing identified deletions and/or amplifications in cytoband 5q31.3 in 9 (4.7%) B-ALL cases and 3 (6.0%) T-ALL cases (Mullighan et al. 2007). The genes in this region include NR3C1 (which encodes for the GR) and LOC389335 which encodes a hypothetical protein. Six of the B-ALL cases occurred in patients known to have the ETV6-RUNX1 translocation, which is associated with a favourable prognosis, and correlation with outcome was not reported. Deletion of the GR locus could lead to glucocorticoid-insensitivity, whereas amplification could lead to increased sensitivity and an improved outcome. This suggests that in a small subset of patients genomic abnormalities of the GR locus may be present, but the clinical significance is not clear.

Page | 23

1.8.2 The role of co-chaperone molecules of the GR Kojika et al investigated the role of HSP70 and HSP90 in glucocorticoid resistance using a panel of lymphoid and non-lymphoid cell lines. The glucocorticoid sensitivity of each cell line was determined using 3H thymidine incorporation, and HSP70 and HSP90 protein expression determined. All of the lymphoid cell lines, regardless of glucocorticoid-sensitivity, had normal HSP70 and HSP90 expression, although two glucocorticoid-resistant myeloid cell lines showed aberrant expression of both proteins (Kojika et al. 1996).

Lauten et al investigated the role of HSP90 in glucocorticoid sensitivity in ALL. Leukaemic blasts were extracted from diagnostic bone marrow and peripheral blood samples from paediatric patients with a poor in vivo prednisolone response and matched with controls from good responders. They demonstrated no significant differences in HSP90 expression at the mRNA or protein level between the two groups. Further subgroup analysis including other known high- risk criteria also showed no difference in expression (Lauten et al. 2003).

Tissing et al investigated the mRNA expression levels of a panel of co- chaperone molecules including HSP40, HSP70 and HSP90, HIP, BAG-1, HOP, p23, CYP-40, FKBP-51 and FKBP-52. Diagnostic samples were obtained from

20 paediatric ALL patients with in vitro sensitivity to prednisolone (LC50 <0.1 µg/ml) and each matched to a patient with in vitro resistance to prednisolone

(LC50 >150 µg/ml). Using matched pair analysis, the study found no differences in mRNA expression between in vitro sensitive and resistant patients for any of the HSP molecules, the ratio of BAG-1/HIP, the ratio of FKBP-51/FKBP-52 or the ratio of HSP90/p23 (Tissing et al. 2005).

Interestingly, although HSP90 and other co-chaperones are intimately involved with the GR and the glucocorticoid response, there are no reports detailing the silencing or overexpressing of the co-chaperones and the effects on glucocorticoid sensitivity. There is, however, a body of literature regarding the pharmacological inhibition of HSP90 in leukaemia and other cancers (Section 5.8) Page | 24

1.8.3 The role of the BCL-2 family of proteins Salomons et al investigated the BCL-2 family members BCL-2, BAX, BAK, BAD,

BCL-XL and MCL-1 at presentation in 78 childhood ALL samples. There was a large variation in baseline expression levels for all the measured proteins. A significant correlation between BCL-2 protein level and immunophenotype and high WCC was seen, but there was no correlation with clinical features for the other measured proteins. There was no relationship found between BCL-2 or other protein levels with in vitro drug resistance for prednisolone, vincristine, or asparaginase as single agents or combined. No differences in protein expression levels between PPRs and PGRs were found, and no evidence was found for a prognostic value of BCL-2 or other proteins (Salomons et al. 1999).

BIM is a pro-apoptotic BH3-only member of the BCL-2 family of proteins, and is the only BCL-2 family member consistently shown in microarray analysis to be upregulated by glucocorticoids in lymphoid cells (Medh et al. 2003; Planey et al. 2003; Wang et al. 2003; Schmidt et al. 2006). A number of experiments have investigated the potentially critical role of this protein in glucocorticoid-evoked apoptosis. Bachmann et al demonstrated that BIM induction at the mRNA and protein level was evident within 8 hours of exposure to dexamethasone of ALL- 3 and ALL-16, two dexamethasone-sensitive xenografts, but this response was significantly attenuated in ALL-7 and ALL-19, two dexamethasone-resistant xenografts (Bachmann et al. 2005; Bachmann et al. 2007). This group subsequently showed that all xenografts, regardless of dexamethasone- sensitivity, showed transcriptional induction of GILZ, a primary downstream target of a functional GR. In contrast glucocorticoid-resistant MTX-R3 and Jurkat cell-lines were defective in both GILZ and BIM transcriptional induction (Bachmann et al. 2007), suggesting divergent mechanisms underlying glucocorticoid resistance in cell lines compared to patient-derived xenografts.

Abrams et al investigated the functional role of BIM in B-precursor ALL cell lines on exposure to the glucocorticoid triamcinolone. They demonstrated that all three isoforms of the BIM protein increased after 24 hours in the glucocorticoid- sensitive 697 cell line but not in the resistant KASUMI-2 and KOPN-8 lines. The Page | 25

697 cells were then transfected with BIM-targeting siRNAs and following BIM- suppression 697 cells retained viability on exposure to triamcinolone and were protected from glucocorticoid-induced apoptosis. They also targeted BCL-2 using siRNA and showed that in drug treated cells silencing of BCL-2 did not significantly modulate the level of glucocorticoid resistance caused by BIM shRNA. They concluded that BIM may function independently of BCL-2 in glucocorticoid-induced apoptosis (Abrams et al. 2004).

In a similar experiment, Lu et al demonstrated BIM protein induction in the glucocorticoid-sensitive CEM T-ALL cell line on exposure to dexamethasone, and showed significantly reduced cell death following BIM silencing by transfection with a BIM shRNA (Lu et al. 2006).

A clinical application of the role of BIM was reported by Jiang et al. This group showed that in a cohort of 30 children with BCP-ALL treated with 7 days of prednisolone monotherapy, BIM protein levels in paired day 0/day 8 bone marrow samples were upregulated in the 25 PGRs but not in the 5 PPRs, and that lower BIM expression was associated with an inferior EFS (Jiang et al. 2011). This is further evidence of the critical role of BIM in glucocorticoid- induced apoptosis.

Ploner et al performed gene expression profiling of ALL peripheral blood samples taken from 13 children at baseline, then at 6-8 hours and 24 hours following systemic therapy with prednisolone. From these data, the mRNA changes in the BCL-2 family members were extracted. They demonstrated that BMF was the only significantly regulated BCL-2 family member at the early timepoint, and at the later timepoint BMF regulation become more pronounced and BIM regulation also reached significance. Paradoxically, the pro-apoptotic PMAIP1/NOXA was repressed.

Page | 26

To functionally validate the significance of BMF and BIM induction, lentiviral constructs expressing shRNAs directed against BIM or BMF in a tetracycline- dependent manner were transduced into CCRF-CEM cells expressing the tetracycline-responsive transrepressor tetR-KRAB. BIM knockdown was associated with a reduction in glucocorticoid-induced apoptosis at 48 hours, and BMF knockdown had a similar, though less pronounced, protective effect. They then generated stable derivatives of CEM-C7H2-2C8 cells by lentiviral transduction with constructs enabling tetracycline-induced expression of transgenic BIMEL and BMF, and demonstrated that overexpression of both

BIMEL and BMF led to significant cell death in a dose-dependent manner.

Ploner et al further investigated the BCL-2 rheostat by generating CCRF-CEM derivatives with conditional knockdown of the pro-survival BCL-2, BCL-XL or MCL-1. Knockdown alone had no detectable effect on cell viability, but in all three instances the cells were more sensitive to glucocorticoid-induced apoptosis and the kinetics of the response was accelerated, an effect most pronounced with MCL-1. Using tetracycline-depending lentiviral constructs, they showed that transgenic expression of BCL-2 delayed glucocorticoid-induced cell death by 24 hours, with similar results for both BCL-XL and MCL-1 (Ploner et al. 2008).

In a follow-up publication, the same group investigated the role of PMAIP1/Noxa. They showed that in several glucocorticoid-sensitive cell lines glucocorticoid induced moderate mRNA repression of NOXA but more pronounced protein repression. When exposed to dexamethasone, transgenic CEM-C7H2-2C8 cells conditionally expressing NOXA showed accelerated kinetics in glucocorticoid- induced apoptosis, similar to that seen with BIMEL, suggesting that NOXA repression might contribute to the 24 hour lag phase in glucocorticoid-induced apoptosis. These experiments suggest that glucocorticoids simultaneously induce pro- and anti-apoptotic signals in ALL cell lines (Ploner et al. 2009).

Page | 27

There is also evidence that Puma has an important role in glucocorticoid- induced apoptosis – Villunger et al generated Puma-/- mice and showed that thymocytes remained viable following exposure to dexamethasone (Villunger et al. 2003). This group subsequently performed experiments using PUMA-/- and BIM-/- knockout mice and BCL-2 transgenic (tg) mice. The mice received intraperitoneal treatment with graded doses of dexamethasone, and cells were harvested after 20 hours. They demonstrated that double positive CD4+8+ PUMA-/-, BIM-/- and BCL-2 tg thymocytes were partially protected from apoptosis compared to wild-type. Interestingly, BIM-/- immature B-cells isolated from the bone marrow were potently protected, whereas absence of PUMA had no protective effect, suggesting different roles of these proteins in glucocorticoid-induced apoptosis depending on cell type (Erlacher et al. 2005). This group then proceeded to investigate the role of BMF and by generating BMF-/- knockout mice showed that B-precursor cells from these mice were partially protected from glucocorticoid-induced apoptosis both in vitro and in vivo when compared to wild type (Labi et al. 2008).

Page | 28

1.9 Gene expression analyses Rapid advancements in high-throughput microarray analysis have allowed groups to investigate genome-wide changes in gene expression. This unbiased approach can potentially identify new mechanisms and targets for treatment.

1.9.1 Glucocorticoid-regulated gene expression Wang et al reported microarray analysis of glucocorticoid-induced apoptosis in two murine T-cell lymphoma cell lines, S49.A2 and WEHI7.2. Cells were incubated in culture and RNA extracted from cells at 6, 12, 18 and 24hrs following treatment with 1µM dexamethasone or vehicle control. Expression profiles were determined using Affymetrix MG-U74Av2 GeneChips. The data were divided into three groups by hierarchical clustering – Group I for genes repressed in both cell lines, and included phosphofructokinase and C-MYC, a pro-proliferative gene; Group II for genes with delayed induction, including IκB-α and the regulatory subunit of PI3K, both inhibitors of pro-survival factors; and Group III for genes which were rapidly induced, including BIM. The study further focused on BIM, and BIM induction was confirmed by northern blotting and immunoblot analysis, with similar results also seen in the CEM-C7 human T- ALL cell line (Wang et al. 2003).

Medh et al reported expression profiles for three CEM clones, CEM-C7-14 (glucocorticoid-sensitive), CEM-C1-15 (glucocorticoid-resistant), and CEM-C1-6, a resistant clone which had reverted to sensitivity. RNA was extracted from cells after 20 hours of exposure to 1µM dexamethasone or vehicle control, and hybridised to Affymetrix MG-U95Av2 GeneChips. Analysis focusing on genes induced >2.5 fold or repressed >2 fold revealed 39 induced and 21 repressed genes in the glucocorticoid-sensitive clones. This included BIM (induced) and C-MYC (repressed). No obvious pro-survival genes were induced in the CEM- C1-15 resistant clone. They concluded with a brief discussion of the other genes identified (Medh et al. 2003).

Page | 29

Schmidt et al investigated glucocorticoid response genes by gene expression profiling of lymphoblasts taken from 13 newly diagnosed children with ALL prior to, 6-8 hours and 24 hours after initiation of therapy with prednisolone. All except one of the children were assessed on day 8 of therapy as PGRs. Whole genome arrays were performed on the Affymetrix U133A V2.0 GeneChips, and differential expression of 0 v 6-8 hours and 0 v 24 hours determined. The top candidate genes were defined as those that showed a ≥ 2 fold regulation in at least 7 out of 13 patients, resulting in 25 induced and 37 repressed probes corresponding to 19 and 30 genes respectively. A discussion of data regarding previously identified genes (including MYC, GR, BIM, FKBP51, SOCS1) as well as putative newly identified genes PFKFB2 (a regulator of glucose metabolism), the transcription factor ZBTB16, and SNF1LK (a protein kinase implicated in cell-cycle regulation) followed (Schmidt et al. 2006).

Tissing et al investigated prednisolone-responsive genes in freshly isolated cells from 13 paediatric patients with ALL. The cells were incubated in culture and exposed to prednisolone, and cells extracted after 3 and 8 hours for profiling on Affymetrix U133A GeneChips. The results were compared to untreated controls with paired results available for 9/13 at 3 hours and 10/13 at 8 hours. Very few significantly differentially expressed genes were seen at 3 hours, and 57 differentially expressed probe sets were seen at 8 hours, including upregulation of FKBP5, the product of which is a co-chaperone of the GR. The most significantly downregulated gene was CYP1A1 which is involved in drug metabolism and detoxification. They also identified genes involved in signal transduction pathways, including TXNIP and ZBTB16 whose proteins form a complex that induces cell-cycle arrest. Other genes identified included those involved in carbohydrate metabolism and NF-κB signalling. Although the sample cohort included in vitro prednisolone-sensitive and prednisolone-resistant cells, the numbers of each were too small to permit statistical analysis (Tissing et al. 2007).

Page | 30

1.9.2 Gene expression associated with glucocorticoid resistance Holleman et al reported gene-expression patterns in drug resistant ALL cells. Primary diagnostic patient ALL cells were assessed by MTT assay for their in vitro sensitivity to prednisolone, vincristine, asparaginase and daunorubicin, with the LC50 used to define cells as sensitive or resistant. RNA of untreated cells was isolated and hybridised to Affymetrix U133A GeneChips. Using supervised clustering, principal-component analysis, and statistical significance (p<0.001) 42 gene-probe sets were identified to be differentially expressed between prednisolone-sensitive and prednisolone-resistant cells, including an overexpression of MCL-1 in prednisolone resistant cells (Holleman et al. 2004).

Rhein et al compared gene expression profiles of initial precursor-B leukaemic blasts with those persisting on day 8 of therapy from 18 children (10 PGRs and 8 PPRs) enrolled on ALL-BFM 2000. Normal mature B cells on day 0 and day 8 were used as a direct comparison. Using the Affymetrix U133A GeneChips, 457 genes were identified as differentially expressed between day 0 and day 8 blasts with a false discovery rate (FDR) <0.05, and the majority of genes were assigned to the following functional groups: persisting blasts demonstrated gene changes consistent with inhibition of cell cycle progression, impairment of metabolism, upregulation of apoptosis and a shift towards a mature B-cell phenotype. When the PPRs and PGRs were compared, the expression of anti- apoptotic BCL-2 was higher in PPRs, and the inhibition of proliferation and metabolism more pronounced in PGRs (Rhein et al. 2007).

Infant ALL is characterised by an exceptionally high incidence of chromosomal translocations affecting the MLL gene, and 30% of MLL-rearranged infant ALL patients are classified as PPRs (Dordelmann et al. 1999). Stam et al investigated gene expression profiles of 25 infant and 27 non-infant ALL and compared the profiles of PPRs and PGRs, and identified MCL-1 as a gene significantly associated with prednisolone resistance in both patient groups. Leukaemic cells with high levels of MCL-1 mRNA expression were significantly more resistant in vitro to prednisolone and dexamethasone, and there was a significant association with a poor in vivo prednisolone response. Page | 31

Downregulation of MCL-1 using shRNA in MLL-rearranged prednisolone- resistant cell lines SEMK2 and MV4-11 showed a moderate sensitisation to prednisolone as assessed by MTT assay. Interestingly, however, the patients with the highest levels of MCL-1 expression were PGRs, and there was no clear relationship between levels of MCL-1 expression and clinical outcome. This suggests that up-regulated MCL-1 is not the only mechanism maintaining prednisolone resistance (Stam et al. 2009).

Cario et al compared expression profiles in diagnostic B-cell precursor paediatric ALL samples obtained from patients subsequently classified as PPRs or PGRs. The study cohort consisted of 20 PPRs, negative for TEL-AML1, BCR-ABL, and MLL-AF4 with a matched cohort of 20 PGRs. Unsupervised cluster analysis of the expression profiles did not reveal a clear separation of PPRs from PGRs. Supervised analysis identified 121 differentially expressed genes, all except 1 (GSTA4) expressed at lower levels in PPRs. Genes identified were mainly involved in cell cycle regulation, DNA replication, cell division and apoptosis. Gene Set Enrichment Analysis (GSEA) showed an enrichment of cell cycle- and apoptosis-associated gene sets in PGRs, and genes important for normal B-cell development (LEF1, TCF4, IL7R) were expressed at lower levels in PPRs (Cario et al. 2008).

Beesley et al described the results of baseline gene expression profiling of a local panel of 9 T-ALL cell lines grown in the absence of drug selection plus 6 additional external T-ALL cell lines. The cell lines were grouped as demonstrating high, intermediate or low in vitro resistance to dexamethasone or methylprednisolone. RNA was extracted and hybridised onto Affymetrix U133A GeneChips, and the glucocorticoid resistance profiles correlated with the gene expression profiles. The expression profiles were analysed by GSEA, using the publicly available databases plus an additional manually curated database of 40 ALL-related publications not captured by the GSEA databases. Although the biological pathways identified by GSEA were highly similar for both glucocorticoids, the results were statistically more robust for methylprednisolone, and further analysis focused on the methylprednisolone signatures. Page | 32

The predominant upregulated pathways associated with methylprednisolone- resistance were those involving cellular respiration (oxidative phosphorylation, the electron transport chain and antioxidant defence), metabolic programs (starvation signalling, glycolysis and gluconeogenesis, cholesterol and steroid biosynthesis), proliferation and regulation by the gene MYC. Analysis of the genes downregulated in glucocorticoid-resistant cells revealed a complementary biological pattern consistent with an alteration in amino-acid metabolism and downregulation of fatty acid β-oxidation. They conclude that together these signatures are consistent with the activation of biosynthetic and metabolic pathways to support a proliferative phenotype in glucocorticoid- resistant cells. They hypothesise that the proliferative phenotype confers resistance through the metabolic suppression of apoptotic potential, or indirectly through offsetting the cytostatic and metabolic suppression induced by glucocorticoids (Beesley et al. 2009b).

Although these studies are heterogeneous and use different models (cell lines in vitro or patient cells in vivo/vitro), a number of consistent themes emerge. Glucocorticoids induce an array of genes in ALL, particularly associated with apoptosis, cell cycle regulation and a number of metabolic pathways. In untreated cells glucocorticoid resistance is associated with upregulation of pro- survival genes, proliferation and impairment of cell cycle progression.

Page | 33

1.10 Therapeutic attempts to overcome glucocorticoid resistance A number of groups have attempted to reverse glucocorticoid resistance using both empirical and bioinformatics-based methods, and a number of agents are now in clinical use or in early phase trials in various cancers.

1.10.1 Arsenic Trioxide Bornhauser et al investigated the potential role of arsenic trioxide (ATO) in glucocorticoid-resistant ALL. ATO is widely used in the treatment of acute promyelocytic leukaemia (APML) (Soignet et al. 1998) and has been shown to have cytotoxic activity in T-ALL mediated through inhibition of the PI3K/AKT cellular survival and proliferation pathway (Tabellini et al. 2005). They showed that the glucocorticoid-sensitive CEM-C7 and glucocorticoid-resistant CEM-C1,

MOLT-4 and Jurkat T-ALL cell lines were sensitive to ATO with MTT IC50 values ranging from 0.8 to 2 µM. Following treatment with low dose ATO at 0.25 µM, the glucocorticoid-resistant cells were sensitised to dexamethasone. They subsequently showed that exposure to 1 µM ATO as single agent or at low dose combination with dexamethasone resulted in a decrease in AKT phosphorylation, an increase in BAD protein and a decrease in XIAP protein, both known targets of the AKT pathway. There was no effect on the protein levels of PUMA, BIM, BID, BCL-2, BCL-XL or MCL-1. The findings were validated in primary patient samples – low dose ATO increased the in vitro sensitivity to dexamethasone in 4 PPRs, but not in cells obtained from 3 PGRs, suggesting that combination with low dose ATO might be beneficial for glucocorticoid-resistant patients (Bornhauser et al. 2007).

1.10.2 Rapamycin Wei et al determined the in vitro sensitivity to prednisolone of 29 primary diagnostic patient ALL samples, and used gene expression profiling to derive a signature of glucocorticoid resistance. This signature was then used to interrogate the Connectivity Map (CMap), a database that uses gene expression signatures derived from the action of small molecule perturbagens on a variety of cancer cell lines (Lamb et al. 2006). They reported that the 10 instances of rapamycin were ranked near the top of the CMap list, indicating

Page | 34

that the expression signature of rapamycin was associated with the sensitive versus resistant signature. Rapamycin inhibits mTOR, which is activated by the PI3K/AKT pathway, and GSEA indicated that the AKT pathway was the most highly enriched gene set in the resistant samples. To further investigate the role of rapamycin, the group demonstrated that as a sole agent rapamycin did not induce apoptosis in a number of ALL or lymphoma cell lines, but that pre- treatment with rapamycin enhanced the efficacy of dexamethasone as evidenced by a decrease in the IC50 determined by MTT assay (Wei et al. 2006).

Gu et al have also showed that 10 nM rapamycin enhanced dexamethasone- induced apoptosis in a number of glucocorticoid-resistant T-ALL cell lines through the inhibition of mTOR signalling (Gu et al. 2010). This is further evidence for the potential role of the PI3K/AKT/mTOR pathway in glucocorticoid resistance in ALL.

1.10.3 MEK inhibitors Rambal et al published a recent report in which inhibition of MEK/ERK-mediated phosphorylation and degradation of BIM protein resulted in synergism with dexamethasone-induced apoptosis in ALL. The experiments were conducted using CCRF-CEM T-ALL cells and the MEK inhibitor PD184352 – as single agents dexamethasone and PD184352 induced minimal apoptosis, but combined showed marked synergism, a result replicated in RS4;11, NALM-6 and highly glucocorticoid-resistant MOLT-4 cells. They showed that the combination resulted in a decrease in phosphorylated BIM and that this effect was inhibited in CEM clones transfected with shRNA constructs against BIM. This suggests a potential role for MEK/ERK inhibitors in ALL and related malignancies (Rambal et al. 2009).

1.10.4 Glycolysis inhibitors Glucocorticoids have marked effects on glucose and carbohydrate metabolism, and glucocorticoid resistance in ALL has been associated with increased glucose consumption (Holleman et al. 2004). Thus these pathways are potential therapeutic targets. Two recent publications have investigated the role of

Page | 35

inhibitors of glycolysis. Using a number of ALL cell lines, they showed that co- incubation with prednisolone and 2-deoxy-D-glucose (2-DG) resulted in a reduction in glucose uptake and markedly increased cell death in prednisolone- resistant cells when compared to single agents. There was no additional cytotoxicity observed in prednisolone-sensitive cells. This was also demonstrated in primary patient cells (Hulleman et al. 2009). Similar experiments carried out by Eberhart et al showed that the combination of 2-DG and dexamethasone inhibited glucose uptake and potentiated cell death in glucocorticoid-resistant cells (Eberhart et al. 2009). These studies demonstrate the potential therapeutic targeting of these pathways involved in glucocorticoid resistance.

1.10.5 NOTCH inhibitors Patients with T-ALL have a less favourable prognosis than BCP-ALL, and activating mutations in NOTCH1 are present in over 50% of cases (Weng et al. 2004). Small molecule γ-secretase inhibitors (GSIs) effectively block NOTCH1 activity in T-ALL cell lines, but as single agents have only weak anti-leukaemic activity and significant gastrointestinal toxicity. Combination treatment with GSIs and dexamethasone resulted in a synergistic interaction in glucocorticoid- resistant cells, an effect which did not extend to other commonly used chemotherapeutic drugs. It was also demonstrated that this was mediated through the GR, and that glucocorticoid treatment had a protective effect on GSI-induced gut toxicity in mice (Real et al. 2009).

1.10.6 MCL-1 inhibitors Inhibition of the pro-survival BCL-2 family members is a potential therapeutic target to overcome glucocorticoid resistance. Two groups have reported that obatoclax, a small molecule inhibitor of MCL-1, can sensitise a number of glucocorticoid-resistant ALL cell lines to dexamethasone-induced apoptosis and autophagy (Bonapace et al. 2010; Heidari et al. 2010).

Page | 36

1.11 Aims of investigation The mechanisms of glucocorticoid resistance are being unravelled, and evidence has accumulated that a number of mechanisms and pathways are likely to be involved. The majority of studies have focused on cell lines, or on individual genes/pathways using in vitro models that may not represent clinically relevant mechanisms of resistance. With this project I aimed to take an unbiased approach using genome-wide glucocorticoid-induced gene expression in the NOD/SCID xenograft mouse model, and to identify clinically relevant mechanisms of glucocorticoid resistance. I further aimed to identify novel therapies to overcome glucocorticoid resistance which could then be rapidly translated into clinical trials, potentially improving the chances of cure for this high-risk group of patients.

My specific objectives were: 1. To create a panel of xenografts derived from children classified as Prednisolone Poor Responders (PPRs) and Prednisolone Good Responders (PGRs), and to characterise the in vitro and in vivo responses of the xenografts to dexamethasone. 2. To undertake a pilot study using a single glucocorticoid-sensitive xenograft to determine the optimal experimental design for investigating in vivo glucocorticoid-induced gene expression. 3. To investigate the differential gene expression of PPRs and PGRs on exposure to dexamethasone in vivo, and to identify and validate novel therapeutic agents aimed at overcoming glucocorticoid resistance.

Page | 37

2 Materials and Methods

Page | 38

2.1 Tissue Culture

2.1.1 Reagents and equipment Foetal calf serum (FCS) was purchased from Thermo Trace (Noble Park, VIC, AUS). Roswell Park Memorial Institute 1640 medium (RPMI1640), Penicillin (10,000 U/ml)- Streptomycin (10,000 U/ml)- L-glutamine (29.2 mg/ml) liquid (PSG) was supplied by Invitrogen (Carlsbad, CA, USA). Quality Biologicals Serum Free-60 (QBSF-60) media was supplied by Quality Biologicals (Gaithersburg, MD, USA), and FMS-like tyrosine kinase-3 ligand (FLT3L) was purchased from ProSpec, (East Brunswick, NJ, USA). Recombinant interleukin- 7 (IL-7) was purchased from Jomar BioSciences (Kensington, SA, AUS). Trypan blue 0.4% was provided by Sigma-Aldrich (St Louis, MO, USA). The NALM-6 human BCP-ALL cell line was purchased from the RIKEN BioResource Centre (Ibaraki, Japan). The Jurkat, CEM, REH and K562 leukaemia cell lines were obtained from CCIA laboratory stocks. Tissue culture flasks and plates were purchased from Greiner Bio-One. Neubauer haemocytometers were purchased from Dutec Diagnostics (Sydney, NSW, AUS). All cell culture procedures were performed in a Biological Safety Cabinet Class II (AES Environmental Pty Ltd, Sydney, AUS). All in vitro experiments were maintained o at 37 C/5% CO2 in humidified incubators.

2.1.2 In vitro cell culture Leukaemia cell lines were cultured in RPMI1640 supplemented with heat inactivated (30 min at 56°C) FCS (10% v/v) and PSG to a final concentration of 100 U/ml penicillin, 100 µg/ml streptomycin, and 2 mM L-glutamine, referred to as RPMI1640+. Cell lines were maintained in vitro by passaging every 3-4 days at a cell density of approximately 2 x 105 cells/ml.

Xenograft cells were retrieved from liquid nitrogen cryostorage and thawed quickly at 37oC in a water bath. Cells were transferred to a 15 ml falcon tube and washed twice with pre-warmed RPMI1640+ to remove traces of the cryopreservative dimethyl sulfoxide (DMSO). Trypan blue exclusion (section 2.1.3) was used to assess the viability and number of cells. Xenograft cells were

Page | 39

resuspended at the required concentration in QBSF-60 media supplemented with 20 ng/ml FLT3L and PSG to a final concentration of 100 U/ml penicillin, 100 µg/ml streptomycin, and 2 mM L-glutamine, referred to as QBSF-60+, an optimised culture condition (Liem et al. 2004). For experiments using the xenograft ALL-56, QBSF-60+ was supplemented with IL-7 (10 ng/ml).

2.1.3 Trypan blue exclusion assay Cells were counted and cell viability was assessed by trypan blue exclusion, a measure of cell membrane integrity. Cells were mixed at a ratio of 1:1 (v/v) with trypan blue solution, loaded onto a haemocytometer and visualised using a light microscope. Viable cells (unstained) and dead cells (stained) were counted from at least 2 individual 1 mm2 fields of the chambers, and then averaged to give the number of cells per 1 mm2 counting chamber, which is equivalent to a volume of 0.1 µL. The concentration of cells/ml was calculated by multiplying the number of cells per 1 mm2 field by the trypan blue dilution factor and by a factor of 104. Cell viability was calculated as:

total number viable cells cell viability = 100 total number of cells

Page | 40

2.2 Cytotoxicity Assays

2.2.1 Reagents and equipment MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide) was purchased from Sigma-Aldrich. MTT labelling reagent was prepared by dissolving power to a concentration of 5 mg/ml in sterile pre-warmed phosphate-buffered saline (PBS). The solution was filter sterilised and stored at 4°C for a maximum of 4 weeks. Hydrochloric acid (HCl) was purchased from Fronine Pty Ltd. (Riverstone, NSW, AUS). Sodium dodecyl sulphate (SDS) was obtained from BDH Laboratory Supplies (Kilsyth, VIC, AUS). All other common chemicals and reagents were of the highest commercial grade and were purchased from local suppliers. The BioTrak II plate reader was purchased from Amersham Biosciences (Buckinghamshire, UK).

2.2.2 Cytotoxic drugs Dexamethasone was purchased from Sigma-Aldrich. 17- dimethylaminoethylamino-17-demethoxygeldanamycin (17-DMAG) was purchased from Sigma-Aldrich and LC Laboratories (Woburn, MA, USA). BIM- BH3 and BID-BH3 hydrocarbon-stapled peptides were synthesised by Dr Kate Jolliffe at University of Sydney.

2.2.3 MTT cytotoxic assay In vitro drug sensitivity to dexamethasone and 17-DMAG was assessed using the colorimetric MTT assay, a short-term drug sensitivity assay (Mosmann 1983). Xenograft cells were retrieved from cryostorage and thawed as described in section 2.1.2, while leukaemia cell lines, which remain in long-term culture, were resuspended in RPMI1640+ at the required cell density. Xenograft cells and cell lines were plated in 100 µl aliquots in 96-well U-bottom tissue culture plates and equilibrated overnight at 37°C/5% CO2. Cell densities for xenograft cells and leukaemia cell lines are summarised in Table 2.1.

Page | 41

The cytotoxic drug was added in 20 µl of the appropriate media such that the final concentration of the drug ranged from 10-12 to 10-5 M and each drug concentration was prepared in triplicate. Controls consisted of cells with media plus vehicle equivalent to the highest concentration of drug used, and media with no cells (background). Plates were incubated for 48 hours at 37°C/5% CO2, prior to the addition of 12 µl MTT labelling reagent. The yellow coloured MTT solution was metabolised by viable cells to produce a dark-purple formazan precipitate. Following a further 4-6 hour incubation at 37°C/5% CO2, 100 µl solubilisation solution (10% SDS in 0.01M HCl) was added to each well to dissolve the formazan precipitate. Cells were incubated overnight at 37°C/5%

CO2, and the optical density (OD) was measured at 570 nM with reference to 655 nM using a BioTrak II microplate reader. The OD is linearly related to the number of viable cells. After subtracting the background, cell survival of the drug treated cells was calculated as a percentage of the vehicle treated control:

mean OD (drug treated cells) survival = 100 mean OD (control cells)

IC50 values (the drug concentration that is lethal to 50% of the cells) were determined from the dose-response curves generated using GraphPad Prism (version 5.02).

Page | 42

Table 2.1. Optimised cell densities for MTT cytotoxicity assay

Xenograft Cell density x105/100 µl ALL-2 2.5 ALL-3 4 ALL-19 3 ALL-26 5 ALL-28 4 ALL-50 5 ALL-51 4 ALL-52 3 ALL-53 4 ALL-54 4 ALL-55 2 ALL-56 3 ALL-57 4 Cell line Cell density x104/100 µl NALM-6 2 REH 2 Jurkat 2 CEM 2 K562 2

Page | 43

2.3 Immunoblotting and Analysis of Protein Expression

2.3.1 Reagents and equipment NuPage 4-12% Bis-Tris gradient polyacrylamide gels, NuPage 20X MES SDS running buffer, NuPage 20X transfer buffer, NuPage 4X LDS Sample Buffer, NuPage antioxidant, NuPage 10X reducing agent, SeeBlue Plus 2 protein marker and the XCell SureLock mini-electrophoresis cell were purchased from Invitrogen. Ponceau S sodium salt, 100X protease inhibitor (PI) cocktail, Tris- HCl, sodium chloride (NaCl), ethylenediaminetetraacetic acid (EDTA), sodium fluoride (NaF) and sodium orthovanadate (Na3VO4) were all purchased from Sigma-Aldrich. Nonidet P (NP)-40 was purchased from Fluka (Buchs, Switzerland). Methanol and acetic acid were purchased from Univar (Ajax, Seven Hills, AUS). Tween-20 was purchased from ICN Biochemicals Inc (Aurora, OH, USA). Polyvinylidene difluoride (PVDF) membrane, Immobilon-P and Immobilon™ Western Chemiluminescent HRP Substrate were purchased from Millipore (Bedford, MA, USA). Three mm chromatography blotting paper was purchased from Whatman (Maidstone, UK). The bicinchoninic acid (BCA) protein assay kit was purchased from Pierce (Rockford, USA). Non-fat dairy milk (NFDM) used was Diploma brand skim milk powder. The autoradiography films were purchased from Fijifilm Corporation (Tokyo, Japan). The source and optimised experimental conditions of the primary and secondary antibodies used in this study are summarised in Table 2.2.

Chemiluminescent images of proteins were captured using the VersaDoc Imagine System Model 5000 (Bio-Rad Laboratories, Hercules, CA, USA), and images were quantified using Quantity One software, Version 4.3.1 (Bio-Rad Laboratories). X-ray films were developed using a Konica X-ray Medical Film Processor, Model SRX-101A (Taipei, Taiwan). A Hewlett-Packard (HP) Scanjet 2200c (Hewlett-Packard Co.; Palo Alto, CA, USA), with HP PrecisionScan LTX Scanner software (Version 1.2), was used to collect the images.

Page | 44

Table 2.2. Antibodies used for immunoblotting. Antibody Dilution Species Condition Supplier

ACTIN 1:5000 Rabbit 3 hr; RT Sigma-Aldrich

BCL2 1:1333 Mouse 3 hr; RT BD Pharmingen

BIM 1:500 Rabbit 3 hr; RT Sigma-Aldrich

HSP70 1:1000 Rabbit ON; 4°C Cell Signaling

AKT 1:5000 Rabbit ON; 4°C Cell Signaling

Anti-rabbit 1:5000 Goat 1 hr; RT Amersham

Anti-mouse 1:5000 Goat 1 hr; RT Amersham

RT, room temperature; ON, overnight.

Page | 45

2.3.2 Isolation of total cellular protein Whole cell lysates of xenograft cells were prepared for immunoblot analysis. Cells were washed twice with ice-cold PBS, to remove traces of FCS, and then lysed by incubation with lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.2%

NP-40, 5 mM EDTA, 50mM NaF, 0.1mM Na3VO4) supplemented with PI, added immediately prior to use. Cell lysates were incubated on ice for 30 min, vortexed every 10 min, and then centrifuged at 10000 x g for 10 min at 4°C. Protein concentration of the supernatant was quantified by the BCA assay method using a BSA standard, according to the manufacturer’s instructions.

2.3.3 Protein electrophoresis, transfer and western blotting For each sample, whole cell lysates (a minimum 10 µg protein) were prepared with Sample Buffer, Reducing Agent, and made up to a final volume of 20 µl with deionised water. Samples were heated at 70°C for 10 min and then loaded into a 4-12% Bis-Tris gel in a XCell SureLock Mini-Cell. The chamber was filled with SDS running buffer (950 ml deionised water, 50 ml MES SDS running buffer and 500 µl antioxidant). Samples were separated by electrophoresis parallel to SeeBlue Plus2 Prestained Marker at 120 V for approximately 45-60 min.

The separated proteins were electro-transferred to a PVDF membrane for 2 hours at 30 V using the XCell II Blot module according to the manufacturer’s instructions. The chamber was filled with transfer buffer (1275 ml deionised water, 75 ml Transfer Buffer, 150 ml methanol and 1.5 ml antioxidant). To verify equal loading and transfer efficiency, PVDF membranes were stained with 0.5% Ponceau S prior to immunoblotting.

Membranes were blocked for 2-3 hours at room temperature (RT) or overnight (ON) at 4oC in 5% (w/v) NFDM in Tris-buffered saline (TBS; 20 mM Tris, 500 mM NaCl, pH 7.5) with 0.05% (v/v) Tween-20 (TBST) to prevent non-specific binding of antibodies. Proteins of interest were detected using commercially available antibodies and dilutions used are detailed in Table 2.2. Briefly, membranes were incubated with primary antibodies for 3 hours at RT or

Page | 46

overnight at 4oC in 0.5% NFDM in TBST. This was followed by 3 x 5 min washes in TBST and subsequent incubation of the membrane with the appropriate horseradish peroxidase (HRP)-conjugated IgG secondary antibody (in 0.5% NFDM in TBST) for 1 hour at RT. After further washing (3 x 5 min TBST), bound secondary antibody was detected by chemiluminescence using Immobilon™ Western Chemiluminescent HRP Substrate according to the manufacturer’s instructions. Results were visualised by autoradiography detection and phosphoimages obtained on the VersaDoc. Quantification of acquired VersaDoc images was performed using Quantity One software. The protein expression was measured relative to a loading control, for this study actin was used.

2.4 Isolation and Processing of RNA

2.4.1 Reagents and equipment Trizol reagent and RNase/DNase free water were purchased from Invitrogen. Nucleic acid grade absolute ethanol, chloroform and sodium acetate were purchased from Sigma-Aldrich. The RNeasy RNA isolation kit was purchased from QIAGEN (Valencia, CA, USA). Total RNA was amplified using Illumina TotalPrep RNA Amplification Kit (Ambion, Foster City, CA, USA). Quantitation of RNA was performed on a Nanodrop ND-1000 spectrophotometer supplied by Nanodrop Technologies (Wilmington, DE, USA). RNA quality and integrity were measured on an Agilent Bioanalyzer 2100, purchased from Agilent Biotechnologies (Santa Clara, CA, USA). Data were analysed and interpreted using Agilent 2100 Expert Software by the Ramaciotti Centre (UNSW, Kensington, NSW, AUS).

2.4.2 RNA extraction and verification Following aspiration of medium, xenograft cells were lysed by adding 1 ml Trizol reagent per 10-15 x106 cells and the sample was incubated for 5 min at RT. The sample was vigorously shaken for 15 sec following the addition of 200 µl chloroform and incubated at RT for 2 min. The samples were centrifuged at 12000 x g for 15 min at 4ºC and the aqueous phase transferred to a new tube Page | 47

and an equal volume of 70% ethanol added. The samples were then transferred to an RNeasy spin column and total RNA was extracted and purified using the RNeasy RNA isolation kit according to manufacturer instructions, summarised briefly below.

The RNeasy spin columns containing the aqueous phase/ethanol mix were centrifuged at 8000 x g for 30 sec at RT. The flow-through was discarded, 700 µl of RWI buffer added and the columns centrifuged at 8000 x g for 30 sec. The flow-through and collection tubes were discarded. The spin column was placed in a new collection tube, 500 µl of RPE buffer was added the columns centrifuged at 8000 x g for 30 sec. This step was repeated and the flow-through and collection tube discarded. The spin column was placed in a new collection tube and 30 µl nuclease-free water was added to the spin column membrane to elute the RNA. The tubes were incubated for 2-3 min at RT after which they were centrifuged at 8000 x g for 1 min. The eluate containing RNA was collected and analysed.

2.4.3 RNA verification The purity and concentration of extracted RNA were assessed by the Optical Density (OD) 260/280 and 260/230 ratios measured on the Nanodrop Spectrophotometer. ‘Pure’ RNA has OD 260/280 and 260/230 ratios around 2.0, and low values generally indicate contamination by phenol or salts. If samples had undesirable ratios, the RNA was precipitated and resuspended in RNase- free water (section 2.4.5). RNA integrity, which assesses for degradation products, was evaluated by using RNA 6000 LabChip kits and an Agilent 2100 Bioanalyzer by the Ramaciotti Centre. Only high quality RNA with OD 260/280 and 260/230 ratios greater than 1.80, and an RNA integrity number (RIN) of greater than 8.0 was subsequently amplified for microarray experiments.

Page | 48

2.4.4 RNA amplification The Illumina TotalPrep RNA Amplification Kit was used to generate biotinylated, amplified RNA for use with Illumina arrays. The procedure, summarised below, consists of reverse transcription to synthesise first strand cDNA, second strand cDNA synthesis, cDNA purification, in vitro transcription to synthesise amplified labelled cRNA, followed by cRNA purification.

For each sample 250-500 ng of total RNA in 11 µl RNase-free water was placed in a non-stick, sterile microcentrifuge tube. A Reverse Transcription Master Mix (Table 2.3) was prepared at RT and 9 µl added to each RNA sample. The samples were placed in a thermal cycler at 42oC for 2 hours. Following incubation, the tubes were centrifuged briefly and placed on ice. A Second Strand Master Mix (Table 2.4) was prepared on ice and 80 µl added to each tube. The samples were placed in a thermal cycler at 16oC (no lid) for 2 hours.

Following incubation, the reactions were placed on ice, and 250 µl of cDNA Binding Buffer was added to each sample. The mixture was transferred to a cDNA Filter Cartridge and centrifuged at 10000 x g for 1 min at RT. The flow- through was discarded, 500 µl Wash Buffer applied to each cartridge and centrifuged at 10000 x g for 1 min at RT. Nuclease-free water was preheated to 55oC and 20 µl applied to the centre of each cartridge. After 2 min at RT, the cartridges were centrifuged at 10000 x g for 1 min and the eluted double- stranded cDNA collected (~17.5 µl) and transferred to a 0.5 ml non-stick sterile PCR tube. An IVT Master Mix (Table 2.5) was prepared at RT and 7.5 µl added to each cDNA sample. The samples were placed in a thermal cycler at 37oC for 14 hours to maximise cRNA yield.

Page | 49

Table 2.3. Reverse Transcription Master Mix (for a single 20 µl reaction).

Amount Component

1 µl T7 Oligo(dT) Primer

2 µl 10X First Strand Buffer

4 µl dNTP Mix

1 µl RNase Inhibitor

1 µl Array Script

Table 2.4. Second Strand Master Mix (for a single 100 µl reaction).

Amount Component

63 µl Nuclease-free Water

10 µl 10X Second Strand Buffer

4 µl dNTP Mix

2 µl DNA Polymerase

1 µl RNase H

Table 2.5. IVT Master Mix (for a single 25 µl reaction).

Amount Component

2.5 µl T7 10X Reaction Buffer

2.5 µl T7 Enzyme Mix

2.5 µl Biotin-NTP Mix

Page | 50

Following incubation, the reaction was stopped by adding 75 µl nuclease-free water to each cRNA sample. The samples were transferred to a sterile 1.5 ml Eppendorf tube and 350 µl cRNA binding buffer and 250 µl 100% ethanol was added. Each sample mixture was transferred to a cRNA Filter Cartridge, and the cartridges were centrifuged at 10000 x g for 1 min at RT. The flow-through discarded, 650 µl Wash Buffer was added, and the cartridges centrifuged at 10000 x g for 1 min. The flow-through was discarded and the cartridges transferred to a fresh cRNA Collection Tube. Nuclease-free water was preheated to 55oC and 100 µl applied to the centre of the filter to elute the cRNA. The samples were incubated at RT for 2 min, and then centrifuged at 10000 x g for 1.5 min. The eluate containing the cRNA was collected and analysed on the Nanodrop spectrophotometer as in section 2.4.3.

2.4.5 RNA concentration and purification If the cRNA yield was low, or the Nanodrop values revealed contamination, the cRNA samples were concentrated and purified using sodium acetate and ethanol. To each sample, 1/10th volume 3M sodium acetate and 2.5 volumes 100% ethanol were added and incubated at -20oC for 30 min. The samples were centrifuged at top speed for 15 min at 4oC and the supernatant removed and discarded. The pellet was washed with 500 µl cold 70% ethanol, and then centrifuged again at top speed for 15 min at 4oC. The ethanol was removed, the pellet air-dried and resuspended in the desired volume of nuclease-free water.

Page | 51

2.5 Illumina Gene Expression

2.5.1 Reagents and Equipment Microarray studies were performed using Illumina Direct Hybridisation Gene Expression Arrays (Illumina, San Diego, CA, USA). The majority of studies were performed using HumanWG-6 v3 Expression BeadChips (6 samples/chip), with one later experiment using HumanHT-12 v4 BeadChips (12 samples/chip). All chips (with associated reagents) were purchased from Illumina, and scanned on the Illumina BeadArray Reader according to manufacturer’s instructions. Nucleic acid grade absolute ethanol was purchased from Sigma-Aldrich. Streptavidin-cy3 was purchased from GE Healthcare (Rydalmere, NSW, AUS).

2.5.2 Direct hybridisation assay For 6-sample chips, 1.5 µg cRNA was made up to 10 µl with RNase-free water. For 12-sample chips, 750 ng cRNA was made up to 5 µl with RNase-free water. The HYB and HCB tubes were heated at 58oC for 10 min to dissolve any precipitated salts, and 20 µl (6-sample chips) or 10 µl (12-sample chips) of HYB added to each cRNA sample. Hyb chamber gaskets were placed in the BeadChip Hyb Chamber and 200 µl HCB dispensed into the humidifying buffer reservoirs. The BeadChips were removed from their packages and placed in the Hyb Chamber. The assay samples were heated at 65oC for 5 min, allowed to cool to RT, and then loaded onto the arrays. The Hyb Chamber was sealed and incubated for 16-20 hours at 58oC with rocker speed set at 5. A 1X High-Temp Wash buffer was prepared (50 ml 10X stock plus 450 ml RNase-free water) and pre-warmed in a heat block to 55oC overnight.

The next day, Wash E1BC solution was prepared by adding 4.5 ml E1BC buffer to 1500 ml RNase-free water. The BeadChips were removed from the Hyb Chamber one at a time and submerged in 500 ml Wash E1BC solution. The coverseal was removed whilst ensuring the BeadChip remained submerged, and the BeadChip transferred to a slide rack in 250 ml Wash E1BC solution. Once all BeadChips had been removed, the slide rack and BeadChips were transferred to the High-Temp Wash buffer for 10 min. The slide rack was then

Page | 52

transferred back to Wash E1BC solution, briefly agitated and then placed on an orbital shaker for 5 min. The rack was then transferred to 250 ml 100% ethanol, briefly agitated and placed on the orbital shaker for 10 min. The rack was transferred to 250 ml fresh Wash E1BC solution, briefly agitated and placed on the orbital shaker for 2 min.

Each BeadChip was transferred to a wash tray containing 4 ml Block E1 buffer, and placed on a rocker for 10 min. Each BeadChip was then transferred to a separate wash tray containing 2 ml Block E1 buffer and 2 µl streptavidin-Cy3, covered and placed on a rocker for 10 min. The BeadChips were transferred back to a slide rack in 250ml fresh Wash E1BC solution, briefly agitated and placed on the orbital shaker for 5 min. The BeadChips in the slide racks were dried by centrifugation at 275 x g for 4 mins at RT, and then scanned on the BeadArray Reader.

Page | 53

2.6 Analysis of Gene Expression

2.6.1 Normalisation and transformation The sample probe profiles with no normalisation or background correction were exported from BeadStudio (version 3.0.14, Illumina, San Diego, CA). The data were pre-processed by either variance stabilising transformation (Lin et al. 2008) and robust spline normalisation in lumi (Du et al. 2008) (Section 4.2), or by log2 transformation and quantile normalization in R (Bolstad et al. 2003) (Sections 3.5 and 5.3) by Dr Mark Cowley at the Peter Wills Bioinformatics Centre (Garvan Institute of Medical Research, Darlinghurst, NSW, AUS).

2.6.2 Differential gene expression Differential gene expression was determined by limma (Smyth 2004) with the positive False Discovery Rate (FDR) correction for multiple testing (Storey and Tibshirani 2003). Limma (linear model for microarray analysis) is a moderated t- statistic with greater statistical power than a regular t-test, and leads to an increased ability to detect truly differentially expressed genes. Limma analyses were performed using the ‘limmaGP’ module of GenePattern version 3.2.3 (Reich et al. 2006) at http://pwbc.garvan.unsw.edu.au/gp/pages/login.jsf.

2.6.3 Hierarchical clustering and correlation Complete linkage hierarchical clustering using Euclidian distance of the top 500 probes showing the greatest variance across all samples was used to compare samples to each other. Analyses were performed using the ‘hierarchical clustering’ module of GenePattern. Paired samples were compared by Pearson correlation in GraphPad Prism or Microsoft Excel 2010.

2.6.4 Functional analysis Functional analysis was performed using gene set enrichment analysis (GSEA) version 3.7 (Subramanian et al. 2005), comparing the limma moderated t- statistic for each probe in a pre-ranked file, against the c2_all collection of gene sets from the Molecular Signatures Database (Subramanian et al. 2005) version 3.0 with 1000 permutations. GSEA is a computational method that determines whether an a priori defined set of genes shows statistically significant, Page | 54

concordant differences between two biological states. GSEA was performed using the ‘GSEApreranked’ module of GenePattern or at http://www.broadinstitute.org/gsea/index.jsp. The similarity of the top up- and down-regulated genesets was assessed by comparing leading edge genes using meta-GSEA module of GenePattern. Gene Ontology analysis was performed using tools of The Database for Annotation, Visualization and Integrated Discovery (DAVID) version 6.7 (Huang da et al. 2009) at http://david.abcc.ncifcrf.gov. Comparison of data from multiple microarray models was performed using parametric analysis of gene set enrichment (Kim and Volsky 2005) implemented in the PGSEA package (version 1.20.1, Furge and Dykema) from the Bioconductor project (Gentleman et al. 2004).

2.6.5 Replicate analysis The stability of gene expression results when reducing the number of replicates was assessed using the Recovery Score method (Pavlidis et al. 2003) from the GeneSelector package (version 1.4.0) of the Bioconductor project.

2.6.6 Connectivity Map analysis Lists of significantly up- and down-regulated genes were extracted from limma analyses and the Illumina probe IDs converted to Unigene IDs using ReMOAT at http://remoat.sysbiol.cam.ac.uk/search.php. The Unigene IDs were imported into NetAffx at http://www.affymetrix.com/analysis/index.affx and were converted to Affymetrix HG-U133A probe IDs. The resulting lists were used to interrogate the Connectivity Map (CMap) build 02 at http://www.broadinstitute.org/cmap. CMap is a collection of genome-wide transcriptional expression data from cultured human cells treated with bioactive small molecules and simple pattern-matching algorithms that together enable the discovery of functional connections between drugs, genes and diseases through the transitory feature of common gene-expression changes (Lamb et al. 2006).

Page | 55

2.7 cDNA Synthesis and Real-Time Quantitative PCR

2.7.1 Reagents and equipment MMLV Reverse Transcriptase, 5X buffer [250 mM Tris-HCl (pH 8.3), 375 mM

KCl, 15 mM MgCl2], 0.1M DTT and RNase-free water were purchased from Invitrogen. Random primers (diluted to 0.1 mg/ml) and dNTPs (diluted to 2.5 mM) were purchased from Amersham Biosciences. RNasin Plus Ribonuclease Inhibitor was purchased from Promega (Madison, WI, USA). Taqman Master Mix and Low-Density Array cards were purchased from Applied Biosystems (Mulgrave, VIC, AUS). Real-time quantitative PCR (RQ-PCR) was performed using an ABI 7900HT PCR machine purchased from Applied Biosystems.

2.7.2 cDNA synthesis For step 1, 2 µg RNA (extracted as per section 2.4.2) in 18.75 µl RNase-free water, was added to a thin-walled 0.2 ml PCR tube, and 6.25 µl of cDNA Mix 1 (Table 2.6) added to each sample. The tubes were incubated at 70oC for 10 min to denature RNA secondary structure, and then for 5 min on ice to prevent the secondary structure reforming.

Following incubation, 25 µl of cDNA Mix 2 (Table 2.7) was added to each sample. The tubes were incubated at 37oC for 1 hr, then 70oC for 15 min to inactivate the enzyme. The resulting cDNA concentration was 40 ng/µl, and was stored at -20oC.

Page | 56

Table 2.6. cDNA Synthesis Mix 1

Amount Component

5 µl 0.1 mg/ml Random Primers

1.25 µl RNasin Plus

Table 2.7. cDNA Synthesis Mix 2

Amount Component

10 µl T7 Oligo(dT) Primer

5 µl 10X First Strand Buffer

5 µl dNTP Mix

5 µl RNase Inhibitor

Page | 57

2.7.3 Real-time quantitative PCR RQ-PCR was performed using custom Taqman Low-Density Array (TLDA) cards using primers for genes of the BCL-2 family members, the Glucocorticoid Receptor (GR) and TSC22D3 and SOCS1, known downstream targets of the GR. Endogenous controls included on the cards were 18S, EEF1A1, ACTB and B2M. Up to 8 samples could be analysed in duplicate on a single card and a list of the Taqman probes selected for inclusion is shown in Table 2.8.

Each sample was made up to 101 µl with 6 µl cDNA (40 ng/ml) as synthesised in section 2.7.2, 44.5 µl nuclease-free water and 50.5 µl Taqman Master Mix. Of this, 97 µl was loaded into the large port on the card. Once all the samples were loaded, the card was centrifuged twice at 1200 rpm for 1 min to disperse the samples into all wells. The card was then sealed and the loading ports removed. The card was then placed in the ABI 7900HT PCR machine and run according to the manufacturer-prescribed TLDA-optimised program.

For each gene, ΔCt was calculated by subtracting the mean cycle threshold of endogenous control from mean cycle threshold of the gene. ΔΔCt was calculated by subtracting ΔCt of control samples from ΔCt of treated samples. Fold change was calculated as 2-ΔΔCt.

Page | 58

Table 2.8. TLDA Card of BCL-2 family members

Gene Description Taqman Probe BAK1 BCL2-antagonist/killer 1 Hs00832876_g1 BAD BCL2-associated agonist of cell death Hs00188930_m1 BCL2L11 BCL2-like 11 (apoptosis facilitator) (BIM) Hs00708019_s1

BAX BCL2-associated X protein Hs00751844_s1 HRK Harakiri Hs00705213_s1 BID BH3 interacting domain death agonist Hs00609632_m1 Apoptotic - BBC3 BCL2 binding component 3 (PUMA) Hs00248075_m1

Pro BIK BCL2-interacting killer (apoptosis-inducing) Hs00154189_m1 BOK BCL2-related ovarian killer Hs00261296_m1 BMF BCL2 modifying factor Hs00372937_m1 PMAIP1 PMA-induced protein 1 (NOXA) Hs00560402_m1 BCL2 B-cell CLL/lymphoma 2 Hs00608023_m1

BCL2L1 BCL2-like 1 (BCL-XL) Hs00169141_m1 BCL2L2 BCL2-like 2 (BCL-W) Hs00187848_m1

Survival BCL2A1 BCL2-related protein A1 Hs00187845_m1 -

Pro BCL2L10 BCL2-like 10 (apoptosis facilitator) (BOO/DIVA) Hs00368095_m1 MCL1 Myeloid cell leukemia sequence 1 (BCL2-related) Hs00172036_m1 NR3C1 Glucocorticoid receptor Hs00230813_m1

TSC22D3 TSC22 domain family, member 3 (GILZ) Hs00608272_m1 GR SOCS1 Suppressor of cytokine signaling 1 Hs00864158_g1 18S 18S Hs99999901_s1

B2M Beta-2-microglobulin Hs99999907_m1 ACTB Actin, beta Hs99999903_m1 Controls EEF1A1 Eukaryotic translation elongation factor 1 alpha 1 Hs00265885_g1

Page | 59

2.8 Xenograft Mouse Model

2.8.1 Reagents and equipment Five to seven week old female non-obese diabetic/severe combined immune deficient (NOD/SCID) or NOD/SCID/IL-2Rγnull (NSG) mice were purchased from Australian BioResources (Moss Vale, NSW, AUS). Mice were housed in a pathogen-free environment for a minimum of one week for acclimatisation prior to any experimental work. Fluorescein isothiocyanate (FITC)-conjugated anti- murine CD45, allophycocyanin (APC)-conjugated anti-human CD45 and APC- conjugated anti-human CD19 were purchased from BioLegend (San Diego, CA, USA). FACS lysing solution was purchased from BD (San Jose, CA, USA). LymphoPrep was purchased from Axis-Shield (Oslo, Norway). Fluorescence Activated Cell Sorting (FACS) analysis was performed using a FACSCanto multiparametric flow cytometer purchased from BD and acquired samples were analysed using BD FACSDiva software version 6.1.2. Patient samples were obtained from the Tumour Bank at Children’s Cancer Institute Australia for Medical Research (CCIA, UNSW, Kensington, NSW, AUS). Cells from ALL-3, a previously established xenograft (Lock et al. 2002), were retrieved from laboratory liquid nitrogen cryostorage.

2.8.2 Inoculation of human leukaemia cells into mice A cohort of BCP-ALL xenografts was established using previously described methods (Lock et al. 2002), with the exception that mice were not irradiated prior to transplantation. Leukaemia cells were retrieved from cryostorage as described in section 2.1.2. Cells were resuspended in sterile PBS at a maximum concentration of 5 x106 cells per 100 µl and placed on ice. Prior to inoculation, mice were warmed by infrared lamps for approximately 5 min until the tail veins were dilated. Each mouse was placed in a Perspex restrainer in a biological safety cabinet and inoculated with 100 µL of the cell suspension via lateral tail vein injection. The puncture site was compressed using a sterile tissue until bleeding had ceased and the mouse was returned to its cage and monitored daily for general well-being.

Page | 60

2.8.3 Monitoring of engraftment Leukaemia engraftment was monitored weekly commencing 2-3 weeks post- inoculation. Each mouse was warmed by placing it in proximity to an infrared lamp for approximately 5 min until the tail veins were dilated. Each mouse was placed in a Perspex restrainer and approximately 50 µl (3 to 4 drops) of peripheral blood was collected from the lateral tail vein using a 23 gauge needle. The puncture site was compressed using a sterile tissue until bleeding had ceased and the mouse was returned to its box.

The peripheral blood samples were stained with 100 µl of a mix consisting of 94 µl PBS, 1 µl FITC-conjugated anti-murine CD45 and 5 µl APC-conjugated anti- human CD45 or CD19. The tubes were incubated away from light for 30 min at RT. Erythrocytes were lysed with 750 µl FACS lysing solution for 15-30 min at RT or at 37oC away from light. Three ml of FACS Buffer (0.2% BSA, 0.1% sodium azide in PBS) was added and the samples centrifuged at 1500 rpm x 5 min at RT. The supernatant was discarded and the antibody complex resuspended in 100 µl FACS Buffer. The samples were analysed by multiparametric flow cytometry on a FACSCanto cytometer with appropriate compensation settings.

Cells that stained positive for FITC-conjugated anti-mouse CD45 were detected in the FL1 fluorescence channel at an emission wavelength of 530 nm and cells that stained positive with APC-conjugated anti-human CD45 or CD19 were detected in the FL4 channel at an emission wavelength of 670 nm. A minimum of 10000 events were acquired per sample. Engraftment was calculated as the proportion of human versus total CD45+ cells (see below), as this parameter has been shown to accurately reflect the overall leukaemic burden (Rice et al. 2000). Previous studies have shown that this method reliably detects as low as 0.1% human CD45+ cells in murine peripheral blood (Nijmeijer et al. 2001).

number human CD45 cells engraftment = 100 total number (human m urine CD45 ) cells

Page | 61

2.8.4 Harvesting of cells from engrafted mice When high level (>70%) engraftment was achieved in the peripheral blood, or if the mice became morbid, the mice were culled by CO2 asphyxiation. Spleen and bone marrow, collected by flushing femurs with 1 ml of RPMI1640+, were taken to harvest human leukaemia cells. To determine the extent of leukaemia infiltration in other organs, samples of the liver, lung, kidney, brain, and enlarged lymph nodes and thymus were also taken. Spleen cell suspensions were prepared by placing individual organs into a sterile tea strainer and homogenising with RPMI1640+ using the plunger of a 10 ml syringe. Subsequently, single-cell suspensions were prepared by filtering the homogenates through a 40 µm cell strainer into 50 ml tubes containing RPMI1640+ to a final volume of 35 ml. Mononuclear cells were purified to >97% human by density gradient centrifugation (800 x g for 30 min at RT) underlaid with 15 ml of LymphoPrep. The mononuclear cell layer was aspirated and washed twice with RPMI1640+. Purified cells were used for serial expansion in mice, or for further in vitro or in vivo experiments.

2.8.5 Analysis of cell surface markers Expression of a range of lineage-specific and differentiation markers on the surface of cells harvested from the spleens of engrafted mice was analysed in the Flow Cytometry Laboratory (Department of Haematology, Prince of Wales Hospital, Randwick, NSW, AUS) using standard procedures as previously described (Lock et al. 2002).

2.8.6 Assessment of in vivo drug sensitivity To assess single agent in vivo drug sensitivity, groups of 16 mice were inoculated with 3-5 x106 mononuclear cells purified from the spleens of primary or secondary recipient mice, and engraftment monitored by flow cytometry as described in section 2.8.3. For combination drug treatments, 32 mice were inoculated (8 each for control, single agent 1, single agent 2, and combination). When the proportion of human CD45+ cells in the peripheral blood (%huCD45+) reached 1% mice were randomised to receive either drug or vehicle control by intraperitoneal injection. The %huCD45+ was monitored throughout and

Page | 62

following the course of treatment, and the event-free survival (EFS) was calculated from the initiation of treatment. An event was defined a priori to be when the %huCD45+ reached 25%, or when animals exhibited clinical signs of disease (weight loss, lethargy, ruffled fur) associated with high level leukaemic infiltration of bone marrow and spleen at autopsy. Mouse EFS was graphically represented by Kaplan-Meier analysis (Kaplan and Meier 1958).

The efficacy of the drug/combination was determined by the Leukemia Growth Delay (LGD) and the Pediatric Preclinical Testing Program (PPTP) Objective Response Measure (ORM) (Houghton et al. 2007). The LGD was calculated as the difference between the median EFS of the treated mice and the median EFS of the control mice, and statistical significance determined by the log-rank (Mantel-Cox) test (Mantel 1966). The ORM was determined by a scoring system determined on day 42 following treatment initiation (‘end of study’) as shown in Table 2.9. Each mouse was scored individually and the median score used for the group score.

Page | 63

Table 2.9. PPTP Objective Response Measure (ORM) scoring method

ORM Description Score

%huCD45 never drops below 1% Progressive PD1 Event before the end of study 0 Disease 1 EFS value of <1.5 times EFS of control group %huCD45 never drops below 1% Progressive PD2 Event before the end of study 2 Disease 2 EFS value of ≥1.5 times EFS of control group

Stable %huCD45 never drops below 1% SD 4 Disease No event before the end of study

Partial PR %huCD45 below 1% for one week only 6 Response

Complete %huCD45 below 1% for two consecutive CR 8 Response weeks

Maintained %huCD45 below 1% for the last three weeks Complete MCR 10 of study Response

Page | 64

2.9 Ethics All experimental studies were approved by the Human Research Ethics Committee and the Animal Care and Ethics Committee of the University of New South Wales under the following applications:

2.9.1 Human Ethics

 05/173 Development of xenograft models of acute leukaemia in immune-deficient mice for preclinical drug testing.  10/114 Development of xenograft models of acute leukaemia in immune-deficient mice for preclinical experimentation.

2.9.2 Animal Ethics

 08/74B A preclinical model of relapse in acute lymphoblastic leukaemia.  08/86B Glucocorticoid-regulated genes in paediatric acute Lymphoblastic leukaemia; a pilot study.  09/78B Comparison of engraftment parameters and drug responses exhibited by human ALL and AML xenografts between NOD/SCID and NSG mouse strains.  10/33A Glucocorticoid-regulated genes in paediatric acute lymphoblastic leukaemia.  11/9A Pediatric Preclinical Testing Program – Leukaemia.

Page | 65

3 Establishment and Characterisation of Xenograft Panel

Page | 66

3.1 Introduction The NOD/SCID xenograft mouse model has been extensively used in our program for many years, but the xenografts that have already been created are from patients at varied clinical stages of disease (diagnosis, relapse), of heterogeneous immunophenotype (BCP-ALL, T-ALL, acute biphenotypic leukaemia), and from an era with diverse treatment protocols. To allow a more directed study of glucocorticoid resistance, I decided to focus on BCP-ALL and created a new panel of xenografts derived from a well-defined population of patients treated according to the same protocol. I then analysed changes in gene expression, and characterised the xenograft responses to dexamethasone both in vitro and in vivo.

3.2 Patient characteristics From 2003-2011, children with ALL in Sydney and other centres across Australia have been treated according to the Australia and New Zealand Children’s Haematology and Oncology Group (ANZCHOG) Study VIII protocol. This protocol commences with a 7-day pre-phase of single agent prednisolone 60 mg/m2/day (plus an age-related dose of intrathecal methotrexate). On day 8 the peripheral blast count is measured and the patients are classified as either a Prednisolone Good Responder (PGR, day 8 blast count <1.0 x109/L) or Prednisolone Poor Responder (PPR, day 8 blast count ≥1.0 x109/L), an indirect assessment of glucocorticoid (and by inference multi-drug) sensitivity. From the Study VIII database I identified 7 BCP-ALL PPRs and 7 PGRs matched by age, diagnostic white cell count and cytogenetics. Of the PPRs, one had already been established as ALL-28, one had been used by another group and was known to engraft poorly, and a third did not engraft by 20 weeks in either NOD/SCID or NSG mice. Of the PGRs, one had already been established as ALL-26, but 2 of the remaining 6 samples did not engraft by 20 weeks in either NOD/SCID or NSG mice. Thus in total my xenograft panel consisted of 5 PPRs (ALL-28, ALL-50, ALL-54, ALL-55 and ALL-57) and 5 PGRs (ALL-26, ALL-51, ALL-52, ALL-53 and ALL-56).

Page | 67

The characteristics of the patients from whom the xenograft panel is derived are shown in Table 3.1. In each cohort there were 4 (80%) males and 1 (20%) female. The median ages were 89 months (PPR) and 87 months (PGR), with median diagnostic WCC of 34.6 x109/L (PPR) and 20.3 x109/L (PGR). No patients had evidence of CNS disease at diagnosis. Each cohort had one Ph+ patient. By definition all the PPRs had a day 8 blast count of ≥1.0 x109/L, and all PGRs had a day 8 blast count of <1.0 x109/L. All PPRs received protocol- directed high-risk chemotherapy, with the patient of ALL-55 undergoing subsequent allogeneic transplantation. All remain in first complete remission (CR1) except for the patient of ALL-50 who relapsed and subsequently died of disease. The PGRs received standard or medium risk therapy and all remain in CR1 except for the patient of ALL-56, who was classified as high risk and underwent allogeneic transplant, but then suffered a late relapse and is now in second complete remission (CR2).

Page | 68

Table 3.1. Patient characteristics of xenograft panel

Xenograft Sex Age Cyto- Dx Dx D8 CR1 Current (mo) genetics WCC Blasts Blasts (mo) Status (x109/L) (x109/L) (x109/L) ALL-28 M 20 Hyperdiploid 15.0 11.8 1.9 +94 CR1 ALL-50 M 131 Normal 34.6 26.1 5.5 30 DOD

ALL-54 M 89 Normal 185.0 174.8 1.2 +93 CR1 PPR ALL-55 M 176 t(9;22) 422.5 388.7 22.6 +36 CR1 ALL-57 F 72 t(1;19) 15.9 7.2 1.6 +27 CR1

ALL-26 F 43 t(12;21) 89.4 80.5 0.0 +96 CR1 ALL-51 M 19 dic(7;9) 90.5 76.9 0.0 +92 CR1

ALL-52 M 138 t(7;15) 14.4 4.0 0.0 +80 CR1 PGR ALL-53 M 87 t(12;21) 20.3 13.8 0.1 +102 CR1 ALL-56 M 120 t(9;22) 8.5 0.1 0.0 55 CR2

PPR, Prednisolone Poor Responder; PGR, Prednisolone Good Responder; mo, months; Dx, diagnosis; WCC, white cell count; D8, Day 8; CR1, first complete remission; CR2, second complete remission; DOD, dead of disease; and +, no event.

Page | 69

3.3 Engraftment characteristics Cryopreserved diagnostic patient samples were obtained from the CCIA Tumour Bank, thawed and washed, and inoculated by intravenous (IV) injection into NOD/SCID mice. Engraftment, as measured by %huCD45+ in the peripheral blood, was monitored weekly and when the %huCD45+ exceeded 50% or the mice became morbid the primary (1o) passage cells were harvested from mouse spleens. Samples were also taken from bone marrow, liver, kidney, lung and brain to assess for leukaemic infiltration (expressed as %huCD45+ of total cells). An aliquot of 1o cells was then injected into mice and secondary (2o) passage cells harvested when the %huCD45+ exceeded 50% or the mice became morbid. Additional cells of samples that either did not engraft or engrafted poorly in NOD/SCID mice were inoculated into NSG mice, a strain more receptive to engraftment with human haematopoietic cells (Agliano et al. 2008). Rates of engraftment were expressed as median time to 1% huCD45+.

3.3.1 ALL-26 This PGR sample engrafted well in NOD/SCID mice at 1o passage, and demonstrated significantly faster kinetics at 2o passage (Figure 3.1). Assessment of organs at 1o harvest revealed high engraftment levels in peripheral blood, bone marrow, spleen and liver, with low level infiltration of kidney, lung and brain (Figure 3.2).

3.3.2 ALL-28 This PPR sample demonstrated prolonged time to engraft in NOD/SCID mice at 1o passage, but engrafted faster at 2o passage in NOD/SCID mice but not in NSG mice (Figure 3.3). Organ infiltration data at 1o harvest showed variable levels in peripheral blood and spleen, but consistently high levels in bone marrow. There was low level engraftment in liver, lung and kidney with minimal engraftment in brain (Figure 3.4).

Page | 70

90 80 70 60 PB in

+ o 50 1 40 2o 30 20 %huCD45 10 0 0 2 4 6 8 10 12 14 16 18 20 Weeks following inoculation

Time (wks) to 1%: 1o NS 13.3 (13.1-14.8) (median (range)) 2o NS 7.0 (3.7-8.1)

Figure 3.1. Engraftment of ALL-26. NS, NOD/SCID; PB, peripheral blood.

100

60 (% of total) + 40

20 huCD45 0

PB LU BM SPL LIV KID BR

Figure 3.2. Infiltration of organs with ALL-26 at 1o harvest. PB, peripheral blood; BM, bone marrow; SPL, spleen; LIV, liver; KID, kidney; LU, lung and BR, brain. Error bars represent mean ± SEM.

Page | 71

90 80

70 o 60 1 PB in o + 50 2 o 40 2 NSG 30 20 %huCD45 10 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Weeks following inoculation

Time (wks) to 1%: 1o NS 13.5 (12.3-14.0) (median (range)) 2o NS 8.0 (7.3-11.5) 2o NSG 12.1 (10.2-16.2)

Figure 3.3. Engraftment of ALL-28. NS, NOD/SCID; PB, peripheral blood.

100

60 (% of total) + 40

20 huCD45 0

PB LU BM SPL LIV KID BR

Figure 3.4. Infiltration of organs with ALL-28 at 1o harvest. PB, peripheral blood; BM, bone marrow; SPL, spleen; LIV, liver; KID, kidney; LU, lung and BR, brain. Error bars represent mean ± SEM.

Page | 72

3.3.3 ALL-50 This PPR sample engrafted well in NOD/SCID mice at both 1o and 2o passage (Figure 3.5). Organ infiltration data at 1o harvest revealed high engraftment levels in peripheral blood and bone marrow, with low levels in spleen and minimal in liver, kidney, lung and brain (Figure 3.6).

3.3.4 ALL-51 This PGR sample engrafted well in NOD/SCID mice, at both 1o and 2o passage (Figure 3.7). Organ infiltration data at 1o harvest revealed high engraftment levels in peripheral blood, bone marrow and spleen, with minimal evidence of human leukaemia cells in liver, kidney, lung and brain (Figure 3.8).

3.3.5 ALL-52 This PGR sample engrafted into NOD/SCID mice, but at huCD45+ levels over 20% in 1o passage the mice became morbid with significant palpable lymphadenopathy (Figure 3.9), similar to the clinical presentation of the child from whom this xenograft was derived. At 2o passage the mice became morbid at much lower engraftment levels (Figure 3.10). Organ infiltration data at 1o harvest revealed that with moderately high peripheral blood levels there was high engraftment in bone marrow and spleen, and minimal evidence of leukaemia in liver, kidney, lung and brain. Analysis of a representative lymph node revealed infiltration with moderately high levels of human cells (Figure 3.11), confirming that the lymphadenopathy was of human rather than murine origin.

Page | 73

70 60

50 1o PB in + 40 2o 30 20 %huCD45 10 0 0 2 4 6 8 10 12 Weeks following inoculation

Time (wks) to 1%: 1o NS 6.1 (6.0-6.4) (median (range)) 2o NS 5.7 (4.2-6.4)

Figure 3.5. Engraftment of ALL-50. NS, NOD/SCID; PB, peripheral blood.

100

60 (% of total) + 40

20 huCD45 0

PB LU BM SPL LIV KID BR

Figure 3.6. Infiltration of organs with ALL-50 at 1o harvest. PB, peripheral blood; BM, bone marrow; SPL, spleen; LIV, liver; KID, kidney; LU, lung and BR, brain. Error bars represent mean ± SEM.

Page | 74

90 80 70 o 60 1 PB in

+ o 50 2 40 30 20 %huCD45 10 0 0 2 4 6 8 10 12 14 16 18 Weeks following inoculation

Time (wks) to 1%: 1o NS 9.9 (7.8-11.8) (median (range)) 2o NS 3.7 (3.2-4.8)

Figure 3.7. Engraftment of ALL-51. NS, NOD/SCID; PB, peripheral blood.

100

60 (% of total) + 40

20 huCD45 0

PB LU BM SPL LIV KID BR

Figure 3.8. Infiltration of organs with ALL-51 at 1o harvest. PB, peripheral blood; BM, bone marrow; SPL, spleen; LIV, liver; KID, kidney; LU, lung and BR, brain. Error bars represent mean ± SEM.

Page | 75

Figure 3.9. Images of ALL-52 NOD/SCID lymphadenopathy. Representative enlarged lymph nodes marked with arrows.

Page | 76

40 1o

PB in o + 30 2

%huCD45 10

0 0 2 4 6 8 10 12 14 16 18 20 22 24 Weeks following inoculation

Time (wks) to 1%: 1o NS 6.4 (4.8-6.7) (median (range)) 2o NS 7.8 (4.7-10.8)

Figure 3.10. Engraftment of ALL-52 in NS mice. NS, NOD/SCID; PB, peripheral blood.

100

60 (% of total) + 40

20 huCD45 0

PB LU LN BM SPL LIV KID BR

Figure 3.11. Infiltration of organs with ALL-52 in NS mice at 1o harvest. PB, peripheral blood; BM, bone marrow; SPL, spleen; LIV, liver; KID, kidney; LU, lung; BR, brain and LN, lymph node. Error bars represent mean ± SEM.

Page | 77

I then obtained additional patient diagnostic cells and inoculated them into NSG mice. In this strain, engraftment was more robust at both 1o and 2o passage (Figure 3.12), without any lymphadenopathy or significant unexpected murine morbidity, and thus subsequent in vivo experiments were done in this strain. Organ infiltration patterns at 1o harvest were broadly comparable to that seen in NOD/SCID mice (Figure 3.13), suggesting that in this xenograft, other than the development of lymphadenopathy, the choice of mouse strain had minimal effect on leukaemia dissemination.

3.3.6 ALL-53 This PGR sample engrafted slowly in NOD/SCID mice at both 1o and 2o passage, with no appreciable improvement when 1o cells were injected into secondary recipient NSG mice (Figure 3.14). Organ infiltration data at 1o harvest revealed high level engraftment in peripheral blood, bone marrow and spleen, with low levels in liver and minimal in kidney, lung and brain (Figure 3.15).

3.3.7 ALL-54 This PPR sample engrafted slowly at 1o passage but appreciably quicker at 2o passage in NOD/SCID mice (Figure 3.16). Organ infiltration data at 1o harvest revealed moderately high levels in peripheral blood and spleen, with minimal evidence of leukaemia in liver, kidney, lung or brain (Figure 3.17).

Page | 78

80 70 60 1o PB in 50 + 2o 40 30 20 %huCD45 10 0 0 2 4 6 8 10 12 Weeks following inoculation

Time (wks) to 1%: 1o NSG 3.3 (3.0-3.6) (median (range)) 2o NSG 2.8 (2.1-3.0)

Figure 3.12. Engraftment of ALL-52 in NSG mice. PB, peripheral blood.

100

60 (% of total) + 40

20 huCD45 0

PB LU BM SPL LIV KID BR

Figure 3.13. Infiltration of organs with ALL-52 in NSG mice at 1o harvest. PB, peripheral blood; BM, bone marrow; SPL, spleen; LIV, liver; KID, kidney; LU, lung and BR, brain. Error bars represent mean ± SEM.

Page | 79

40 1o PB in

+ 30 2o o 20 2 NSG

%huCD45 10

0 0 2 4 6 8 10 12 14 16 18 20 22 24 Weeks following inoculation

Time (wks) to 1%: 1o NS 12.6 (10.8-14.3) (median (range)) 2o NS 9.3 (8.2-17.2) 2o NSG 13.4 (11.4-15.2)

Figure 3.14. Engraftment of ALL-53. NS, NOD/SCID; PB, peripheral blood.

100

60 (% of total) + 40

20 huCD45 0

PB LU BM SPL LIV KID BR

Figure 3.15. Infiltration of organs with ALL-53 at 1o harvest. PB, peripheral blood; BM, bone marrow; SPL, spleen; LIV, liver; KID, kidney; LU, lung and BR, brain. Error bars represent mean ± SEM.

Page | 80

60 1o 50 2o PB in + 40 30 20 %huCD45 10 0 0 2 4 6 8 10 12 14 16 18 20 22 Weeks following inoculation

Time (wks) to 1%: 1o NS 16.7 (median (range)) 2o NS 5.3 (4.9-6.0)

Figure 3.16. Engraftment of ALL-54. NS, NOD/SCID; PB, peripheral blood.

100

60 (% of total) + 40

20 huCD45 0

PB LU BM SPL LIV KID BR

Figure 3.17. Infiltration of organs with ALL-54 at 1o harvest. PB, peripheral blood; BM, bone marrow; SPL, spleen; LIV, liver; KID, kidney; LU, lung and BR, brain. No BM data available, and no error bars shown as data from sole engrafted mice.

Page | 81

3.3.8 ALL-55 This PPR sample engrafted extremely poorly in NOD/SCID mice, barely reaching 5% after over 20 weeks at 1o passage, and demonstrated minimal engraftment at 2o passage (Figure 3.18). Organ infiltration data at 1o harvest revealed that despite very low peripheral blood engraftment, moderately high levels were seen in bone marrow and spleen. Minimal engraftment was observed in liver, kidney or lung, but interestingly this xenograft demonstrated the highest engraftment into the CNS (Figure 3.19), without overt murine neurological morbidity.

I then obtained additional diagnostic cells and inoculated them into NSG mice. In this strain, engraftment was more robust at both 1o and 2o passage (Figure 3.20), and thus subsequent in vivo experiments were done in this strain. Organ infiltration data at 1o harvest revealed a similar pattern to that seen in NOD/SCID mice, although absolute engraftment levels in peripheral blood, bone marrow, spleen and brain were higher (Figure 3.21). This again suggests that the choice of mouse strain has minimal effect on the pattern of leukaemia dissemination.

Page | 82

1o PB in + 2o 5 %huCD45

0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 Weeks following inoculation

Time (wks) to 1%: 1o NS 19.1 (18.9-20.7) (median (range)) 2o NS n/a

Figure 3.18. Engraftment of ALL-55 in NS mice. NS, NOD/SCID; PB, peripheral blood.

100

60 (% of total) + 40

20 huCD45 0

PB LU BM SPL LIV KID BR

Figure 3.19. Infiltration of organs with ALL-55 in NS mice at 1o harvest. PB, peripheral blood; BM, bone marrow; SPL, spleen; LIV, liver; KID, kidney; LU, lung and BR, brain. Error bars represent mean ± SEM.

Page | 83

40 PB in + 30 1o 2o 20

%huCD45 10

0 0 2 4 6 8 10 12 14 16 18 Weeks following inoculation

Time (wks) to 1%: 1o NSG 8.1 (7.9-9.0) (median (range)) 2o NSG 4.8 (4.1-6.4)

Figure 3.20. Engraftment of ALL-55 in NSG mice. PB, peripheral blood.

100

60 (% of total) + 40

20 huCD45 0

PB LU BM SPL LIV KID BR

Figure 3.21. Infiltration of organs with ALL-55 in NSG mice at 1o harvest. PB, peripheral blood; BM, bone marrow; SPL, spleen; LIV, liver; KID, kidney; LU, lung and BR, brain. Error bars represent mean ± SEM.

Page | 84

3.3.9 ALL-56 This PGR sample engrafted slowly at 1o passage but significantly quicker at 2o passage in NOD/SCID mice (Figure 3.22). Organ infiltration data at 1o harvest revealed that at moderate peripheral blood engraftment, high levels were seen in bone marrow and spleen, minimal in liver, kidney and lung, and moderate in brain (Figure 3.23).

3.3.10 ALL-57 This PPR sample engrafted slowly in NOD/SCID mice but appreciably quicker when 1o cells were injected into secondary recipient NSG mice (Figure 3.24), and thus subsequent in vivo experiments were done in this strain. Organ infiltration data at 1o harvest revealed high level engraftment in peripheral blood and spleen, moderate level in bone marrow, liver and brain, and minimal in kidney and lung (Figure 3.25).

Page | 85

100 90 80 70 o

PB in 1 + 60 o 50 2 40 30

%huCD45 20 10 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 Weeks following inoculation

Time (wks) to 1%: 1o NS 16.2 (16.1-16.2) (median (range)) 2o NS 4.9 (3.5-7.2)

Figure 3.22. Engraftment of ALL-56. NS, NOD/SCID; PB, peripheral blood.

100

60 (% of total) + 40

20 huCD45 0

PB LU BM SPL LIV KID BR

Figure 3.23. Infiltration of organs with ALL-56 at 1o harvest. PB, peripheral blood; BM, bone marrow; SPL, spleen; LIV, liver; KID, kidney; LU, lung and BR, brain. Error bars represent mean ± SEM.

Page | 86

70 60 1o 50 2o in PB in + 40 2o NSG 30 20 %huCD45 10 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 Weeks following inoculation

Time (wks) to 1%: 1o NS 15.6 (13.0-18.3) (median (range)) 2o NS 14.0 (13.2-15.2) 2o NSG 4.5 (3.3-5.5)

Figure 3.24. Engraftment of ALL-57. NS, NOD/SCID; PB, peripheral blood.

100

60 (% of total) + 40

20 huCD45 0

PB LU BM SPL LIV KID BR

Figure 3.25. Infiltration of organs with ALL-57 at 1o harvest. PB, peripheral blood; BM, bone marrow; SPL, spleen; LIV, liver; KID, kidney; LU, lung and BR, brain. Error bars represent mean ± SEM.

Page | 87

3.3.11 Summary and discussion of engraftment characteristics The establishment of the xenograft panel was a complex procedure over many months in 2 different mouse strains, and the engraftment kinetics are summarised in Table 3.2. Five of the xenografts (ALL-26, ALL-50, ALL-51, ALL- 54 and ALL-56) expanded well in NOD/SCID mice, 2 xenografts (ALL-52 and ALL-55) expanded adequately only in NSG mice, and 2 xenografts demonstrated slow engraftment in either strain (ALL-28 and ALL-53). When engraftment in NOD/SCID and NSG mice was compared (Table 3.3), the rates of engraftment were significantly faster in NSG mice in ALL-52 at 1o and 2o passage, and in ALL-57 at 2o passage, with a trend to faster engraftment in NSG mice in ALL-55 at 1o passage. One xenograft, ALL-28, engrafted faster in NOD/SCID mice at 2o passage.

Of the mice inoculated, 40/47 (85%) engrafted at 1o passage and 47/51 (92%) at 2o passage (Table 3.2). Of the 1o non-engrafters, 3 succumbed to thymoma, a well-recognised complication in NOD/SCID mice (Prochazka et al. 1992), 1 to hind-limb paralysis and 3 remained well without evidence of engraftment. Of the 2o non-engrafters, 1 succumbed to thymoma and the others showed no evidence of engraftment. Although this demonstrates reliable and reproducible engraftment from patient samples destined to engraft, there were an additional 3 patient samples that failed to engraft in 11 NOD/SCID mice (2 thymomas) or 3 NSG mice. This is a potential limitation of intravenous inoculation which could be overcome by direct intrafemoral inoculation (Mazurier et al. 2003; McKenzie et al. 2005).

Page | 88

Table 3.2. Summary of xenograft engraftment kinetics.

Xenograft Mouse Weeks to 1% huCD45+ (median (range)) P Strain 1o N 2o n ALL-28 NOD/SCID 13.5 (12.3-14.0) 4/4 8.0 (7.3-11.5) 5/5 0.016 ALL-50 NOD/SCID 6.1 (6.0-6.4) 3/3 5.7 (4.2-6.4) 4/5 0.229

ALL-54 NOD/SCID 16.7 1/3 5.3 (4.9-6.0) 6/6 n/a

PPR ALL-55 NOD/SCID 19.1 (18.9-20.7) 3/4 n/a 0/2 n/a NSG 8.1 (7.9-9.0) 4/4 4.8 (4.1-6.4) 6/6 0.010 ALL-57 NOD/SCID 15.6 (13.0-18.3) 2/3 14.0 (13.2-15.2) 3/4 n/a

ALL-26 NOD/SCID 13.3 (13.1-14.8) 4/4 7.0 (3.7-8.1) 4/4 0.029 ALL-51 NOD/SCID 9.9 (7.8-11.8) 3/3 3.7 (3.2-4.8) 4/4 0.057

ALL-52 NOD/SCID 6.4 (4.8-6.7) 7/7 7.8 (4.7-10.8) 4/4 0.230

PGR NSG 3.3 (3.0-3.5) 3/3 2.8 (2.1-3.0) 7/7 0.016 ALL-53 NOD/SCID 12.6 (10.8-14.3) 2/3 9.3 (8.2-17.2) 4/4 n/a ALL-56 NOD/SCID 16.2 (16.1-16.2) 4/6 4.9 (3.5-7.2) 4/4 0.029 n, number of mice engrafted / total number of mice inoculated. P value calculated by Mann-Whitney test (n/a if <3 engrafted mice in any group).

Page | 89

Table 3.3. Comparison of engraftment in NOD/SCID and NSG mice.

Xenograft Weeks to 1% huCD45+ (median (range)) P NOD/SCID n NSG n ALL-52 1o 6.4 (4.8-6.7) 7/7 3.3 (3.0-3.5) 3/3 0.017

ALL-52 2o 7.8 (4.7-10.8) 4/4 2.8 (2.1-3.0) 7/7 0.006

ALL-55 1o 19.1 (18.9-20.7) 3/4 8.1 (7.9-9.0) 4/4 0.057

ALL-28 2o 8.0 (7.3-11.5) 5/5 12.1 (10.2-16.2) 5/6 0.032

ALL-53 2o 9.3 (8.2-17.2) 4/4 13.4 (11.4-15.2) 8/8 0.214

ALL-57 2o 14.0 (13.2-15.2) 3/4 4.5 (3.5-5.5) 8/8 0.012

n, number of mice engrafted / total number of mice inoculated. P value calculated by Mann-Whitney test.

Page | 90

Intraclonal heterogeneity at diagnosis of ALL has been reported (Choi et al. 2007; Mullighan et al. 2008; Anderson et al. 2011), and T-ALL xenografts created in immunodeficient mice mirror the process of clonal selection and subsequent relapse (Clappier et al. 2011). Xenograft cells re-establish leukaemia more rapidly in subsequent recipient mice than diagnostic patient cells in primary recipient mice (Lock et al. 2002; Meyer et al. 2011; Schmitz et al. 2011), suggesting an enrichment of the xenograft leukaemia in leukaemia initiating cells (Clappier et al. 2011). In this panel, all xenografts (with the exception of ALL-52 in NOD/SCID mice) engrafted more rapidly at secondary passage, with 5 xenografts [ALL-26, ALL-28, ALL-52 (NSG), ALL-55 (NSG) and ALL-56] showing significant acceleration and ALL-51 showing a trend to acceleration (Table 3.2 and Figure 3.26).

It has previously been reported that time to clinically assessed leukaemia is related to CR1 duration (Meyer et al. 2011), but in other reports there has been no correlation for diagnostic samples inoculated into mice (Lock et al. 2002). I objectively defined engraftment time as the time to reach 1% huCD45+ cells in the peripheral blood. In this cohort, the CR1 duration could only be determined for ALL-50 and ALL-56, the two patients who relapsed. The xenograft ALL-50, derived from the patient who relapsed early (30 months) reached 1% engraftment in 6.2 weeks, whereas the xenograft ALL-56, derived from the patient who relapsed late (55 months) reached 1% after 16.2 weeks, suggesting a trend to increasing engraftment time with increasing CR1 duration. However ALL-51 and ALL-52 both reached 1% within 10 weeks and both are long term survivors, evidence against any such a correlation.

Page | 91

ALL-26* ALL-28* + 20 ALL-50 16 ALL-51 ALL-52 12 ALL-52 NSG* 8 ALL-53 ALL-54 4 ALL-55 Weeks to 1% huCD45 0 ALL-55 NSG* 1 2 ALL-56* Passage ALL-57

Figure 3.26. Comparison of xenograft engraftment rates by passage. Statistically significant difference in time to 1% denoted by *.

Page | 92

I then investigated the association of engraftment time with conventional clinical high risk criteria (prednisolone poor response, high diagnostic white cell count). The median time to 1% huCD45+ was 15.6 weeks in the PPR cohort and 12.6 weeks in the PGR cohort, with no statistically significant difference (Figure 3.27). There was also no significant correlation between engraftment time and diagnostic white cell count (Figure 3.28). Thus these conventional high risk criteria are not associated with faster engraftment rates, consistent with previous reports (Lock et al. 2002; Meyer et al. 2011). However a recent study has shown that xenografts derived from patients with MRD-very high risk BCP- ALL engrafted faster than standard risk ALL (Schmitz et al. 2011), suggesting that intrinsically drug-resistant cells are more able to adapt to the xenograft environment.

There are however a number of limitations in comparing such studies, as cell source, host conditioning, transplantation procedure, mouse strain and definition of engraftment can vary. Cryopreservation, storage and reprocessing of cells can result in an increased latency to leukaemia compared to freshly isolated cells (Malaise et al. 2011). Host conditioning by sub-lethal irradiation can improve engraftment but can have a negative impact on homing (Spiegel et al. 2004). Direct intraosseous transfer of cells, circumventing the homing process, can result in higher rates of engraftment compared to systemic intravenous injection (Mazurier et al. 2003; McKenzie et al. 2005). It is clear that mouse strain has a significant impact on engraftment, with cells demonstrating improved engraftment in the more permissive NSG strain compared to SCID or NOD/SCID mice (Agliano et al. 2008). Interestingly, superior engraftment is seen in female versus male mice (Notta et al. 2010). Perhaps most crucially, the endpoint of observation (subjective clinical assessment versus objective definition) to determine engraftment time is of the utmost importance and will significantly affect the experimental results.

Page | 93

P=0.310 + 20

4 Weeks to 1% huCD45 0

PPR PGR

Figure 3.27. Engraftment time and prednisolone response. Median time of each xenograft to reach 1% huCD45+ shown. Groups compared by Mann- Whitney test. Error bar represents median ± interquartile range.

+ 20

8 r2=0.333 4 P=0.081 Weeks to 1% huCD45 0 0 100 200 300 400 Diagnostic WCC (x109/l)

Figure 3.28. Engraftment time and diagnostic white cell count (WCC). Median time of each xenograft to reach 1% huCD45+ shown. r, Pearson correlation.

Page | 94

3.4 Xenograft immunophenotyping by serial passage To confirm that xenograft cells express an immunophenotype concordant with diagnostic patient samples, expression of a range of lineage-specific and differentiation markers on the surface of cells harvested from the spleens of engrafted mice was analysed by flow cytometry. The blast immunophenotype was essentially preserved throughout serial passaging, and only minor changes seen with gain of CD34 in ALL-50, gain of CD45 in ALL-28 and ALL-56 (Table 3.4). ALL-52 and ALL-55 demonstrated myeloid co-expression (CD13 and CD33) at diagnosis, and other than the loss of CD33 in ALL-55 this was preserved throughout serial passaging (Table 3.4).

Table 3.4 legend: 1o, primary engraftment; 2o, secondary engraftment; pos, unquantified positive (>30%); neg, unquantified negative (<30%); and nd, not done. Numbers refer to the percentage of cells staining more intensely than the isotype control antibody.

Page | 95

Table 3.4. Xenograft immunophenotyping by serial passage.

Xenograft Sample CD45 DR CD19 CD22 CD10 CD3 CD13 CD33 CD34 ALL-26 Diagnosis pos 99 98 97 99 <1 2 1 53 NOD/SCID 1o 91 98 97 98 98 <1 <1 <1 82 NOD/SCID 2o 81 84 82 85 84 2 1 1 35 ALL-28 Diagnosis neg 95 88 nd 89 nd 1 31 86 NOD/SCID 1o 86 99 92 99 98 <1 <1 <1 94 NOD/SCID 2o 82 98 98 98 98 <1 <1 <1 95 ALL-50 Diagnosis 88 78 98 nd 97 <1 <1 <1 14 NOD/SCID 1o 17 95 99 91 84 <1 <1 <1 58 NOD/SCID 2o 36 99 99 99 99 <1 <1 <1 84 ALL-51 Diagnosis 73 97 98 95 98 <1 <1 <1 5 NOD/SCID 1o 73 100 100 99 100 <1 <1 <1 8 NOD/SCID 2o 62 98 98 98 99 <1 <1 <1 26 ALL-52 Diagnosis pos 99 98 98 5 2 95 94 9 NSG 1o 100 100 100 99 <1 <1 91 99 31 NSG 2o 100 99 100 100 <1 <1 95 100 37 ALL-53 Diagnosis pos pos pos nd pos nd nd nd nd NOD/SCID 1o 94 99 99 99 99 <1 <1 5 86 NOD/SCID 2o 100 100 99 100 100 <1 <1 <1 52 ALL-54 Diagnosis 25 66 75 nd 82 nd <1 <1 30 NOD/SCID 1o 43 80 94 97 98 3 <1 <1 36 NOD/SCID 2o 10 88 94 99 99 1 <1 <1 22 ALL-55 Diagnosis 100 98 99 nd 96 nd 95 67 96 NSG 1o 100 100 100 78 99 <1 78 7 96 NSG 2o 100 100 100 88 99 <1 95 6 99 ALL-56 Diagnosis 11 89 73 nd 73 nd 8 3 85 NOD/SCID 1o 68 99 100 99 99 <1 <1 <1 99 NOD/SCID 2o 100 100 100 97 99 <1 <1 <1 90 ALL-57 Diagnosis 86 pos 54 nd 42 nd 6 9 5 NOD/SCID 1o 100 100 100 100 89 <1 <1 <1 <1 NOD/SCID 2o 100 100 100 100 100 <1 <1 <1 <1

Page | 96

3.5 Gene expression analysis

3.5.1 Clustering and correlation For each xenograft, gene expression analysis of diagnostic patient sample, 1o and 2o xenograft samples was performed. Unsupervised hierarchical clustering revealed that each primary and secondary xenograft clustered with the patient sample from which the xenograft was derived (Figure 3.29), demonstrating the excellent overall relevance of the xenograft model. Within each patient- xenograft cluster, the primary and secondary xenograft samples clustered together, reflecting selection of a specific clone by the xenografting process which is then maintained in subsequent passages. Both the primary and secondary xenograft gene expression patterns showed highly significant correlation with the patient sample (Figure 3.30 and Figure 3.31). There was no cluster separation in gene expression profile by patient prednisolone response status.

3.5.2 Xenograft-associated genes The expression profiles from primary recipient xenograft samples were compared to the corresponding patient sample by paired limma, and a xenograft-associated list of 748 upregulated and 411 downregulated genes (FDR <0.005) was generated. The list of upregulated genes was enriched in processes related to DNA repair and cation transport, particularly potassium ion channels (Table 3.5). Potassium ion channels are important contributors to the malignant phenotype through the modulation of cell cycle progression to increase proliferation, and through cytoskeletal remodelling to enhance invasion and migration [reviewed in (Fiske et al. 2006)]. This indicates that the xenografts have an intrinsically more malignant phenotype than the patient samples from which they were derived.

Page | 97

Figure 3.29. Hierarchical clustering of xenografts with diagnostic patient samples. Unsupervised cluster dendrogram of the top 500 probes with the greatest variance across all samples. No primary xenograft sample available for ALL-28. P, patient sample; X1, 1o xenograft, and X2, 2o xenograft. Red, PPRs and blue, PGRs.

Page | 98

Figure 3.30. Correlation of gene expression of primary and secondary PPR xenografts with original diagnostic patient sample. Axes show normalised expression values. R2, square of Pearson correlation coefficient. No primary xenograft data available for ALL-28. Page | 99

Figure 3.31. Correlation of gene expression of primary and secondary PGR xenografts with original diagnostic patient sample. Axes show normalised expression values. R2, square of Pearson correlation coefficient.

Page | 100

Of the downregulated genes, most striking was the enrichment of genes involved in oxygen binding and transport (Table 3.6). Diagnostic bone marrow samples are frequently contaminated by erythrocytes, and the loss of these cells in the xenografting process is the likely explanation for the relative downregulation of oxygen carrying activity. Pathways related to the positive regulation of apoptosis were also significantly enriched (Table 3.6), evidence that pro-apoptotic pathways are downregulated in xenografts compared to the matched patient sample (Figure 3.32). This has potential implications for assessment of drug responses, as a xenograft may be intrinsically more resistant to (drug-induced) apoptosis than the patient sample from which it was derived. In a recent report Clappier et al demonstrated that although T-ALL xenograft cells were still inhibited by glucocorticoids in vitro in a dose- dependent manner, higher doses compared to the diagnostic patient samples and the addition of γ-secretase inhibitors were required to induce apoptosis (Clappier et al. 2011).

This group further suggested that clonal selection in xenografted T-ALL mirrors the process seen in relapse and that xenograft-associated gene signatures are significantly enriched in relapse samples (Clappier et al. 2011). From a series of 9 T-ALL diagnosis-xenograft pairs, they generated an ‘overexpressed in T-ALL xenograft’ geneset. This list was enriched in a cluster of genes linked to cell cycle progression and mitosis, suggesting that increased proliferative potential selectively promotes engraftment. To validate this, I obtained the corresponding relapse samples for ALL-50 and ALL-56 from the CCIA Tumour Bank and performed gene expression profiling. For each xenograft, I generated a list of genes differentially expressed between relapse and diagnosis. GSEA revealed enrichment of the Clappier xenograft-associated signature in both relapse samples when compared to diagnosis, strong confirmation that in both T-ALL and BCP-ALL xenograft and relapse leukaemias share profiles of increased proliferation (Figure 3.33).

Page | 101

Table 3.5. Gene Ontology terms upregulated in 1o xenografts compared to matched diagnostic patient samples.

Gene Ontology Term P Enrichment FDR GO:0030529~ribonucleoprotein complex 0.000 2.074 0.386 GO:0008076~voltage-gated potassium channel complex 0.001 4.140 0.911 GO:0034705~potassium channel complex 0.001 4.140 0.911 GO:0034702~ion channel complex 0.002 2.605 2.612 GO:0006813~potassium ion transport 0.002 2.833 3.679 GO:0005242~inward rectifier potassium channel activity 0.003 8.335 3.840 GO:0006310~DNA recombination 0.003 3.321 5.253 GO:0005267~potassium channel activity 0.003 3.033 4.876 GO:0034703~cation channel complex 0.004 2.967 5.367 GO:0005216~ion channel activity 0.004 1.995 6.379 GO:0022803~passive transmembrane transporter activity 0.005 1.954 6.405 GO:0005694~chromosome 0.005 1.858 7.094 GO:0022838~substrate specific channel activity 0.006 1.935 8.867 GO:0006281~DNA repair 0.008 2.087 12.56 GO:0005654~nucleoplasm 0.009 1.534 11.65 GO:0015267~channel activity 0.009 1.869 12.42 GO:0051427~hormone receptor binding 0.010 3.334 13.70

FDR, false discovery rate.

Page | 102

Table 3.6. Gene Ontology terms downregulated in 1o xenografts compared to matched diagnostic patient samples.

Gene Ontology Term P Enrichment FDR GO:0015669~gas transport 0.000 30.833 0.000 GO:0005833~hemoglobin complex 0.000 42.852 0.000 GO:0015671~oxygen transport 0.000 32.019 0.001 GO:0005344~oxygen transporter activity 0.000 29.664 0.001 GO:0044445~cytosolic part 0.000 5.799 0.009 GO:0019825~oxygen binding 0.000 10.463 0.067 GO:0005506~iron ion binding 0.000 3.130 0.439 GO:0046914~transition metal ion binding 0.001 1.477 0.880 GO:0043065~positive regulation of apoptosis 0.001 2.581 2.243 GO:0043068~positive regulation of programmed cell death 0.001 2.563 2.400 GO:0010942~positive regulation of cell death 0.002 2.552 2.510 GO:0006917~induction of apoptosis 0.002 2.818 3.683 GO:0042110~T cell activation 0.002 4.405 3.684 GO:0012502~induction of programmed cell death 0.002 2.810 3.786 GO:0020037~heme binding 0.003 4.249 3.756 GO:0002697~regulation of immune effector process 0.003 4.808 5.369 GO:0005375~copper ion transporter activity 0.003 32.136 4.683 GO:0046906~tetrapyrrole binding 0.004 3.986 5.322 GO:0006915~apoptosis 0.006 2.074 9.515 GO:0051291~protein hetero-oligomerisation 0.006 6.671 10.29 GO:0010941~regulation of cell death 0.007 1.873 10.47 GO:0012501~programmed cell death 0.007 2.044 10.90 GO:0033279~ribosomal subunit 0.008 4.017 9.755 GO:0046649~lymphocyte activation 0.008 3.138 12.53 GO:0046915~transition metal ion transporter activity 0.008 9.522 10.76 GO:0020027~hemoglobin metabolic process 0.009 20.812 13.32 GO:0005840~ribosome 0.009 3.075 11.00

FDR, false discovery rate.

Page | 103

Figure 3.32. Heatmap of pro-apoptotic genes in patient and 1o xenografts. Significant gene set identified by Gene Ontology from list of genes differentially expressed in 1o xenografts compared to the matched diagnostic patient sample. ALL-28 excluded as no X1 sample. Red denotes relative upregulation, blue denotes relative downregulation. P, patient; and X1, 1o xenograft.

Page | 104

Figure 3.33. GSEA of Clappier xenograft gene set in relapse versus diagnosis samples of ALL-50 (A) and ALL-56 (B).

Page | 105

3.5.3 Summary and discussion In summary, the gene expression profiling of diagnosis, xenograft and relapse samples confirms the overall relevance of the xenograft mouse model, with xenografts clustering with their corresponding diagnosis sample with highly significant correlation. Although potential contamination with murine cells is an inherent risk in this model, it has been shown that the presence of mouse cells does not significantly skew expression profiles in samples highly engrafted with human cells (Samuels et al. 2010).

The major strength of the xenograft mouse model is the ability to establish a systemic human leukaemia which retains the fundamental characteristics of the original disease. Using this model, it is presumed that data obtained regarding in vivo drug responses would provide an accurate pre-clinical assessment of the efficacy of a drug, enabling translation of promising drugs into clinical trials. However, gene expression analyses of xenograft samples compared to diagnostic patient samples suggest that the xenografts, with which the majority of in vivo experiments are performed, may be intrinsically more proliferative and resistant to apoptosis. This has implications for assessment of drug sensitivity, and the efficacy of drugs may be underestimated. However this is not a straightforward issue to resolve, as it would not be practical to perform all experiments solely on diagnostic patient samples (either in vitro or in vivo).

Page | 106

3.6 Characterisation of xenograft glucocorticoid responses Although the xenografts were established from patients whose clinical prednisolone response was known, experimental assessment of glucocorticoid sensitivity is model-dependent and may not always be concordant. For this panel, I assessed the sensitivity of all 10 xenografts in vitro by MTT assay using 2o passage cells, and 8 of the xenografts in vivo using the mouse model using 2o or 3o passage cells. No in vivo results for ALL-28 and ALL-53 could be obtained as mice inoculated with these xenografts did not reach 1% huCD45+ (the threshold for starting treatment) in NOD/SCID mice even when observed for up to 16 weeks from injection, and neither showed appreciable acceleration in NSG mice. In vivo experiments for ALL-26, ALL-50, ALL-51, ALL-54 and ALL- 57 were performed in NOD/SCID mice; in vivo experiments for ALL-52, ALL-55 and ALL-57 were performed in NSG mice. Unpublished data from our lab has shown that drug responses are not significantly different between the two mouse strains and supports the use of either NOD/SCID or NSG mice for in vivo efficacy studies.

3.6.1 In vitro assessment of glucocorticoid responses

All the PPR xenografts were resistant to dexamethasone in vitro, with IC50 values >10 µM, except for ALL-54 which had an IC50 of 59 nM (Figure 3.34A). All the PGR xenografts demonstrated varying degrees of in vitro sensitivity, except for ALL-51 which was resistant (Figure 3.34B). However, it is clear that glucocorticoid sensitivity is a continuum, and IC50 values are an arbitrary method to categorise xenografts as resistant or sensitive. For example, a comparison of the cytotoxicity curves of ALL-54 and ALL-57 revealed that the difference in cell viability at the highest concentration of dexamethasone is only approximately 10%, yet the IC50 values are profoundly different. An alternative method of analysing drug sensitivity is to determine cell viability at certain fixed drug concentrations. I plotted cell viability after 48 hours incubation with dexamethasone at concentrations of 10 nM and 100 nM, and it is clear that there is a trend towards lower cell viability in PGR xenografts at both concentrations (Figure 3.35), suggesting increased glucocorticoid sensitivity.

Page | 107

120 A 100 ALL-28 IC50 >10M ALL-50 IC >10M 80 50 ALL-54 IC50 59nM 60 ALL-55 IC50 >10M 40 ALL-57 IC50 >10M

Viability (% control) Viability 20

0 -12 -11 -10 -9 -8 -7 -6 -5 log [dexamethasone]

120 B 100 ALL-26 IC50 1.9nM ALL-51 IC50 >10M 80 ALL-52 IC50 142nM 60 ALL-53 IC50 1.0nM

40 ALL-56 IC50 9.0nM

Viability (% control) Viability 20

0 -12 -11 -10 -9 -8 -7 -6 -5 log [dexamethasone]

Figure 3.34. In vitro assessment of glucocorticoid responses. Secondary passage cells from PPRs (A) and PGRs (B) were treated with increasing concentrations of dexamethasone, and viability assessed by MTT assay after 48 hours incubation. Values expressed as a percentage of untreated control. Each data point represents the mean ± SEM from 3 independent experiments.

Page | 108

P=0.151 120 A

100

20 Viability (% control) Viability

PPR PGR

120 B P=0.095 100

20 Viability (% control) Viability

PPR PGR

Figure 3.35. Xenograft viability after 48 hours incubation with dexamethasone 10 nM (A) and 100 nM (B). Viability assessed by MTT assay and expressed as % control. Error bar represent the median. Groups compared by Mann-Whitney test.

Page | 109

3.6.2 In vivo assessment of glucocorticoid responses For in vivo assessment, xenografts were inoculated into groups of 16 mice, and when huCD45+ reached 1%, the mice were randomised into either treated (dexamethasone 15 mg/kg IP Monday-Friday for 4 weeks) or vehicle control groups. Mice were monitored weekly and culled at event (huCD45+ ≥ 25 in peripheral blood or internal organs at autopsy) or if they demonstrated significant morbidity. Glucocorticoid response was assessed by leukaemia growth delay (LGD) and by the PPTP Objective Response Measure (ORM) on day 42 from the start of treatment.

The engraftment and survival curves of the 4 PPR xenografts profiled are shown in Figure 3.36. ALL-57 showed the highest level of glucocorticoid resistance, with treated mice growing through treatment resulting in a LGD of 7.5 days. With ALL-50, treated mice became progressively sicker through treatment without high peripheral blood huCD45+, but on autopsy all had evidence of high level engraftment in bone marrow and spleen. ALL-55 had a LGD of 30.7 days, and ALL-54 showed the greatest response with an LGD of 46.3 days, consistent with this PPR xenograft showing the greatest sensitivity in vitro - interestingly, although ALL-54 is a PPR by definition, the absolute reduction in blast count for this patient from diagnosis is 99.3%, which is in the same range as for PGRs and markedly higher than the other PPRs (Table 3.7). ALL-57, the PPR xenograft with the shortest LGD, demonstrates the least blast reduction. Thus for the PPRs it is evident that LGD correlates with blast reduction, consistent with the clinical courses of the patients from whom the xenografts are derived.

Two PPRs, ALL-50 and ALL-55 had high levels of MRD detected on day 79 of treatment, whereas ALL-57, which is highly resistant in vivo had negative MRD (Table 3.7). This suggests that ALL-50 and ALL-55 are not only intrinsically resistant to glucocorticoids but also to multi-agent chemotherapy.

Page | 110

A ALL-50 B 60 100

50 80 Control + 40 Control Dex Dex 60 30 40 LGD 21.1 days

%huCD45 20 P=0.0001 10 20 Survival probabililty (%) probabililty Survival 0 -10 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Days post-randomisation Days post-randomisation

C ALL-54 D 70 100 60 80 Control + 50 Control Dex Dex 40 60 30 40 LGD = 46.3 days %huCD45 20 P<0.0001 20 10 Survival probabililty (%) probabililty Survival 0 -10 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Days post randomisation Days post-randomisation

E ALL-55 F 60 100

50 80 Control + 40 Control 60 Dex 30 Dex 40 LGD = 30.7 days

%huCD45 20 P=0.001 10 20 Survival probabililty (%) probabililty Survival 0 -10 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Days post randomisation Days post-randomisation

G ALL-57 H 100 100 90 80 80 Control Control + 70 Dex 60 Dex 60 50 40 40 LGD = 7.5 days

%huCD45 30 P<0.0001 20 20 10 Survival probabililty (%) probabililty Survival 0 -10 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Days post randomisation Days post-randomisation

Figure 3.36. In vivo assessment of PPR response to dexamethasone. Engraftment (A, C, E, and G) and Kaplan-Meier curves of event-free survival (EFS) (B, D, F, and H) following treatment with dexamethasone 15 mg/kg IP Mon-Fri for 4 weeks (red) or vehicle control (blue). Leukemia Growth Delay (LGD) calculated as the difference between median EFS of treated and control groups. Solid black bar represents treatment period. Dex, dexamethasone.

Page | 111

Table 3.7. Blast reduction and MRD levels.

Xenograft Diagnosis Day 8 Blasts Blast MRD Blasts (x109/L) (x109/L) Reduction (%) TP2 ALL-28 11.8 1.9 83.9 neg ALL-50 26.1 5.5 78.8 pos

ALL-54 174.8 1.2 99.3 n/a PPR ALL-55 388.7 22.6 94.2 pos ALL-57 7.2 1.6 77.8 neg

ALL-26 80.5 0.0 100 neg ALL-51 76.9 0.0 100 neg

ALL-52 4.0 0.0 100 neg PGR ALL-53 13.8 0.1 99.3 neg ALL-56 0.1 0.0 100 pos

MRD, minimal residual disease; TP2, time point 2 (day 79); neg <5x10-4; pos >5x10-4; and n/a, not available.

Page | 112

The engraftment and survival curves of the 4 PGR xenografts profiled are shown in Figure 3.37. ALL-26 and ALL-51 were highly glucocorticoid sensitive, with clearance of blasts and LGD values of 70 days and 69.7 days respectively. Several treated mice in these experiments were found dead early prior to reaching event coinciding with the move of the animal facility to new premises, and three mice were lost to thymoma. With ALL-52, engraftment in both treated and control mice appeared to plateau at huCD45+ levels around 15%, but on autopsy when the mice became morbid, there was evidence of high level engraftment in bone marrow and spleen. ALL-56 demonstrated surprising results, as although a PGR which had demonstrated in vitro sensitivity to dexamethasone, in the in vivo mouse model this xenograft is highly resistant, growing through dexamethasone treatment with a LGD of only 4.4 days. However this in vivo response is consistent with the clinical course of the patient, as high MRD levels were detected at time point 2 (Table 3.7) indicating intrinsic resistance to multi-agent chemotherapy.

The objective response measures (ORMs) scored on day 42 from the initiation of treatment are shown in heatmap and COMPARE-like format in Figure 3.38. The ORMs confirmed a mixed response amongst the PPRs, with low or intermediate activity seen in ALL-50, ALL-55 and ALL-57 but good activity seen in ALL-54. Amongst the PPRs, all showed excellent activity except ALL-56 which demonstrated low activity.

Page | 113

A ALL-26 B 70 100 60 80 Control Control + 50 Dex Dex 40 60 30 40 LGD = 70 days %huCD45 20 P<0.0001 20 10 Survival probabililty (%) probabililty Survival 0 -10 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Days post randomisation Days post-randomisation

C ALL-51 D 60 100

50 80 Control Controls + 40 Dex Dex 60 30 40

%huCD45 20 LGD 69.7 days P=0.0038 10 20 Survival probabililty (%) probabililty Survival 0 -10 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Days post-randomisation Days post-randomisation

E ALL-52 F 30 100

Control 80 Control + 20 Dex Dex 60

40 LGD = 67 days

%huCD45 10 p=0.0003 20 Survival probabililty (%) probabililty Survival 0 0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160 Days post-randomisation Days post randomisation

G ALL-56 H 80 100 70 80 Control 60 Control + Treated 50 Treated 60 40 30 40 LGD = 4.4 days %huCD45 P=0.006 20 20 10 Survival probabililty (%) probabililty Survival 0 -10 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Days post randomisation Days post-randomisation

Figure 3.37. In vivo assessment of PGR response to dexamethasone. Engraftment (A, C, E, and G) and Kaplan-Meier curves of event-free survival (EFS) (B, D, F, and H) following treatment with dexamethasone 15 mg/kg IP Mon-Fri for 4 weeks (red) or vehicle control (blue). Leukemia Growth Delay (LGD) calculated as the difference between median EFS of treated and control groups. Non-event censored mice shown by ticks. Solid black bar represents treatment period. Dex, dexamethasone. Page | 114

PPRs PGRs ALL-50 ALL-54 ALL-55 ALL-57 ALL-26 ALL-51 ALL-52 ALL-56 2 8 4 0 10 10 10 0

ALL-56 ALL-52 ALL-51 ALL-26 ALL-57 ALL-55 ALL-54 ALL-50 -5.0 -2.5 0.0 2.5 5.0 Mid-point Difference

Figure 3.38. In vivo objective response to dexamethasone. Heatmap (top) depicting the Objective Response Measure (ORM) for each xenograft – high level activity ORM ≥6, intermediate activity ORM ≥2 and <6, low activity ORM <2. COMPARE-like graph (bottom) represents the difference of xenograft response from the mid-point response (stable disease). Bars to the right of the median represent xenografts that are more sensitive, bars to the left represent those that are less sensitive. All responses were statistically significant by the log-rank (Mantel-Cox) test. Red, PPRs and blue, PGRs.

Page | 115

To investigate the disparity between the in vitro and in vivo results for ALL-56, I harvested ALL-56 cells from mice after dexamethasone treatment, and compared the in vitro dexamethasone responses of patient diagnostic and relapse cells, 1o and 2o xenograft cells, and in vivo dexamethasone-treated 3o recipient cells - Interestingly, the diagnostic, relapse and 1o cells were all resistant, yet the 2o and dexamethasone-treated cells were both sensitive (Figure 3.39), indicating that the xenografting process could potentially result in epigenetic changes which significantly affect drug response, and that the NOD/SCID host environment allows outgrowth of a dexamethasone-sensitive clone.

One explanation of why in vitro sensitive 2o cells (Figure 3.39, blue) are highly resistant in vivo is related to the differential regulation of pro-apoptotic and anti- apoptotic pathways by dexamethasone. In Section 5.5 I have shown that in vitro, pathways related to the positive and negative regulation of apoptosis were downregulated, but in vivo only pathways positively related to apoptosis were downregulated. This suggests that in vivo the competing balance favours anti- apoptotic pathways, and is a potential explanation of the discordant vivo/vitro responses seen in this xenograft.

Page | 116

120

100 Patient - diagnosis IC50 >10M Xenograft - 1o IC >10M 80 50 Xenograft - 2o IC 9.0nM 60 50 Xenograft - DexRx IC 9.1nM 40 50 Patient - relapse IC50 >10M

Viability (% control) Viability 20

0 -12 -11 -10 -9 -8 -7 -6 -5 log [dexamethasone]

Figure 3.39. In vitro assessment of glucocorticoid sensitivity of ALL-56. Cells were treated with increasing concentrations of dexamethasone, and viability assessed by MTT assay after 48 hours incubation. Values expressed as a percentage of untreated control. Each data point represents the mean ± SEM from 3 independent experiments (except relapse sample – single experiment only due to limited cell availability). DexRx, in vivo dexamethasone- treated 3o recipient cells.

Page | 117

3.6.3 Summary and discussion of glucocorticoid responses I have shown that in most instances PPR xenografts are resistant to dexamethasone both in vitro and in vivo, and that PGR xenografts are sensitive in both models (Table 3.8). However there are a number of exceptions. Of the PPR xenografts, ALL-54 shows a degree of glucocorticoid sensitivity in keeping with the clinical course of the patient from whom this xenograft was derived, and of the PGR xenografts ALL-51 and ALL-56 have discordant in vitro/in vivo responses.

Previous reports have determined in vitro glucocorticoid sensitivity of primary ALL patient samples (Holleman et al. 2004; Wei et al. 2006; Stam et al. 2009) and noted that there is not an absolute correlation with the clinical patient prednisolone response and in vitro glucocorticoid sensitivity (Stam et al. 2009). These differences may be related to the artificial cytokine-supported in vitro culture environment. There are also significant differences in drug exposure between in vitro and in vivo models. In vitro the cells are continuously exposed to drug, but in vivo drug metabolism and pharmacokinetics will determine systemic exposure. Another factor is that patients are treated with oral prednisolone and for experimental purposes I have used the closely related injectable glucocorticoid dexamethasone to assess glucocorticoid response.

In the case of ALL-56, the change from in vitro resistant to in vitro sensitive with increasing xenograft passage is potentially explained by a sub-population with a reversible drug-tolerant state, as has been reported in non-small cell lung cancer cells (Sharma et al. 2010). In general, however, our lab has not found significant variations in drug responses by xenograft passage, although due to limited cell stocks it is not routine to assess drug responses (either in vitro or in vivo) of early passage cells. It has also been reported that in vitro drug responses of xenograft cells co-cultured on immortalised human mesenchymal stromal calls remain mostly stable across serial passages (Schmitz et al. 2011). The change of ALL-56 from in vitro sensitive to in vivo resistant is potentially related to differential regulation of pathways related to the positive and negative regulation of apoptosis. Page | 118

Table 3.8. Summary of dexamethasone responses

Xenograft in vitro in vivo in vitro in vivo in vivo in vivo

IC50 LGD stratification stratification Score ORM ALL-28 >10 µM nd Resistant nd nd ALL-50 >10 µM 21.1 Resistant Resistant 2 PD2

ALL-54 59 nM 46.3 Sensitive Intermediate 8 CR PPR ALL-55 >10 µM 30.7 Resistant Intermediate 4 SD ALL-57 >10 µM 7.5 Resistant Resistant 0 PD1

ALL-26 2 nM 70.0 Sensitive Sensitive 10 MCR ALL-51 >10 µM 69.7 Resistant Sensitive 10 MCR

ALL-52 142 nM 67.0 Intermediate Sensitive 10 MCR PGR ALL-53 1 nM nd Sensitive nd nd ALL-56 9 nM 4.4 Sensitive Resistant 0 PD1

LGD, leukemia growth delay (days); ORM, objective response measure;

Resistant, IC50 >1 µM, LGD <25 days; Intermediate, IC50 100nM – 1µM; LGD

25-50 days; Sensitive, IC50 <100 nM, LGD >50 days; and nd, not done.

Page | 119

3.7 Conclusions I successfully established and characterised a panel of 5 PPR and 5 PGR xenografts, assessed and analysed gene expression changes associated with the xenografting process, and determined the in vitro and in vivo experimental responses to dexamethasone. This well characterised panel was then used to investigate in vivo dexamethasone-induced gene expression (Section 5.2) using the optimal experimental setup determined by a pilot study (Chapter 4). This panel of xenografts will also provide a valuable resource for future researchers.

Page | 120

4 Pilot of in vivo Glucocorticoid-Induced Gene Expression

Page | 121

4.1 Introduction Genome-wide gene expression studies are powerful tools to investigate mechanisms of disease and advances in high-throughput technology have significantly reduced the costs associated with such projects. A number of studies have published microarray data of glucocorticoid-induced genes in lymphoid cells, but comparison of the data is complicated by differences in experimental model, and technical issues involving platform and chip type. Previous studies of glucocorticoid-induced genes in ALL have been carried out using in vitro cell-line models (Obexer et al. 2001; Tonko et al. 2001; Yoshida et al. 2002; Medh et al. 2003; Wang et al. 2003; Webb et al. 2003; Rainer et al. 2009) and patient-derived cells, both in vivo (Schmidt et al. 2006) and in vitro (Tissing et al. 2007). However to date there have been no reports of glucocorticoid-induced gene expression in ALL using the NOD/SCID xenograft mouse model, and the optimal experimental design for such studies has not been determined.

In this pilot study I investigated glucocorticoid-induced gene expression in BCP- ALL using the NOD/SCID xenograft model, with the aim of optimising the experimental design for future large-scale studies. I then compared the microarray data to publicly available datasets of glucocorticoid-induced genes in other experimental models, and validated a number of important genes (Bhadri et al. 2011).

Page | 122

4.2 Experimental design ALL-3, a glucocorticoid-sensitive xenograft derived from a 12 year old girl with mixed lineage leukaemia (MLL)-rearranged BCP-ALL, was chosen for this study (Lock et al. 2002). MLL-rearranged ALL is associated with a poor prednisolone response and an inferior outcome (Dordelmann et al. 1999), and although this patient suffered a late bone marrow relapse, she is currently a long-term survivor after re-treatment with chemotherapy. ALL-3 demonstrates in vitro glucocorticoid sensitivity, with an IC50 of 4.7 nM by MTT assay after 48 hours exposure to dexamethasone (Bachmann et al. 2005). In the in vivo NOD/SCID xenograft mouse model, ALL-3 is highly responsive to treatment with single agent dexamethasone, with rapid clearance of leukaemic blasts from the peripheral blood and recurrence of leukaemia delayed by 63.4 days compared to vehicle-treated control (Liem et al. 2004).

The first aim of this study was to determine which time point(s) would provide the greatest biological insights into in vivo glucocorticoid-induced leukaemic cell death. It has been demonstrated that changes in gene expression can be detected as early as 6 hours after treatment of ALL with glucocorticoids in vivo (Schmidt et al. 2006) and in vitro (Wang et al. 2003), although earlier time points show few, if any, significantly differentially expressed genes (Tissing et al. 2007).

However despite these early gene expression changes, I demonstrated that in the BCP-ALL NALM-6 cell line, significant cell death in vitro does not occur until at least 24 hours following exposure to glucocorticoids. NALM-6 cells were treated with 1 µM dexamethasone or control and cell count and viability assessed every 24 hours. Whilst a small fall in viability was seen at 24 hours, the most significant falls occurred over the subsequent 24 hours, indicating that the majority of cell death was occurring in the period 24-48 hours following exposure (Figure 4.1).

Page | 123

30 100

25 90 Viability(%) /ml 5 20 80 Control - Cell count Control - Viability 15 70 Treated - Cell count 10 60 Treated - Viability

Cell count x10 count Cell 5 50

0 40 0 24 48 72 96 Time (hours)

Figure 4.1. Time course of NALM-6 cell proliferation and death following exposure to dexamethasone. NALM-6 cells were cultured in RPMI1640+ and 1 µM dexamethasone or vehicle control was added. Cell number and viability were assessed every 24 hours.

Page | 124

This phenomenon has also been previously reported in ALL-3 – when cells from this xenograft were exposed in vitro to 1 µM dexamethasone, little cell death was seen in the first 24 hours, followed by a substantial drop in cell viability in the second 24 hours (Bachmann et al. 2005). I thus chose to investigate gene expression at an early time point (8 hours) and 2 later time points (24 and 48 hours) following dexamethasone. It was also important to determine whether the manual handling of mice and the IP injections induced changes in gene expression. Thus all groups of dexamethasone-treated mice were matched to an IP vehicle-treated control group.

The final aim of the pilot study was to determine the optimal number of replicates required to maintain the statistical integrity of the results. Experiments involving mice are complex, costly and time-consuming, and attempts should always be made to minimise the number of animals used in a study. However a reduction in replicates adversely affects statistical power, and thus starting from 4 mice per group, I determined if using the data from fewer replicates maintained the overall stability of results.

Page | 125

4.3 Generation of in vivo gene expression

Cells from ALL-3 were inoculated by tail-vein injection into 28 NOD/SCID mice, and engraftment, calculated as the %huCD45+ in the peripheral blood, was monitored weekly. When high level (>70%) engraftment was achieved, between 8 and 10 weeks post-transplantation, the mice were randomised and split into groups of 4 to receive either dexamethasone 15 mg/kg or vehicle control by intra-peritoneal (IP) injection. Mice were culled by CO2 asphyxiation at 0 hours (pre-treatment, group 1), 8 hours (groups 2 and 3), 24 hours (groups 4 and 5) or 48 hours (groups 6 and 7) following treatment. The mice in groups 6 and 7 received a second dose of dexamethasone or vehicle control at 24 hours. Two mice succumbed to thymoma, resulting in 3 mice in each of groups 6 and 7.

Cell suspensions of spleens were prepared and mononuclear cells enriched and purified to >97% human by density gradient centrifugation and cell viability assessed by trypan blue exclusion. RNA was extracted and the RNA integrity number (RIN) verified. All samples had RINs >8.0 and were thus suitable for processing. The RNA was amplified and hybridised onto Illumina WG-6 v3 chips. The chips were scanned on the Illumina Bead Array Reader and gene expression analysed. The sample probe profiles with no normalisation or background correction were exported from BeadStudio version 3.0.14. The data were pre-processed using variance stabilising transformation (Lin et al. 2008) and robust spline normalisation in lumi (Du et al. 2008) which takes advantage of each probe being represented by >25 beads.

Page | 126

4.4 Identification of the optimal time point To identify the optimal time point for investigating glucocorticoid-regulated gene expression, I used limma (Smyth 2004) to determine the differential gene expression for each group (treated or control) compared to time 0 hours, with the positive False Discovery Rate (FDR) correction for multiple testing (Storey and Tibshirani 2003). Of the three time points investigated in this study, the 8 hour dexamethasone-treated time point demonstrated the greatest number of differentially expressed genes compared to baseline control, with far fewer observed at the 24 and 48 hour dexamethasone-treated time points (Table 4.1, Table 4.2, and Figure 4.2).

Whilst a similar proportion of up-regulated and down-regulated genes were identified at the 8 hour dexamethasone-treated time point (1158 vs 1072 respectively, FDR<0.05), of those with large fold changes (FC>2 or FC<0.5, red dots in Figure 4.2A), 75% were up regulated (199 vs 65 respectively), consistent with the predominant role of glucocorticoids as transcriptional activators. The large number of statistically differentially expressed genes (FDR<0.05) with small fold changes (0.5

This demonstrated that in the xenograft mouse model, the 8 hour time point provided the greatest information, and that these changes were not sustained over later time points. The handling of the mice and IP vehicle control injections had minimal effect on gene expression, and thus time-matched controls were redundant.

Page | 127

Table 4.1 Number of differentially expressed genes by False Discovery Rate (FDR), compared to time 0 hours.

Time Point FDR <0.25 FDR <0.1 FDR <0.05 (hours) + - + - + - 8 2313 2434 1470 1423 1158 1072

24 970 1087 273 421 75 195 Dex 48 321 327 41 95 12 44 8 0 0 0 0 0 0

24 0 1 0 1 0 1 Con 48 0 1 0 1 0 1

+ upregulated; - downregulated; Dex, dexamethasone treated and Con, control.

Table 4.2 Number of differentially expressed genes by Fold Change (FC), compared to time 0 hours.

Time Point FC >1.5 FC >2 FC >4 (hours) + - + - + - 8 501 429 201 68 38 0

24 137 341 15 90 0 0 Dex 48 79 234 5 69 0 3 8 1 37 1 2 0 0

24 1 5 0 0 0 0 Con 48 7 34 0 2 0 0

+ upregulated; - downregulated; Dex, dexamethasone treated and Con, control.

Page | 128

Figure 4.2. Volcano plots of significantly differentially expressed genes following treatment with dexamethasone at 8 hours (A), 24 hours (B), 48 hours (C). Significance was defined as log2 Fold Change >1 or <-1 with False Discovery Rate (FDR) <0.05. Each dot represents a single gene, and significant genes indicated by red dots.

Page | 129

4.5 Analysis of identified genes

4.5.1 Early 8 hour time point At the 8 hour time point, there were 173 genes upregulated with a t-statistic (the ratio of fold change to standard error) >10 and 25 genes downregulated with a t-statistic <-10 (corresponding to P <1.74 x10-9 and FDR <2.95 x10-7, Table 4.3). None of these genes had sustained expression changes at 24 or 48 hours, and although this could potentially reflect the early death of sensitive cells, there was no significant fall in peripheral blood engraftment levels with treatment (Figure 4.3A) and all splenic harvests had a cell viability of ≥96 . There was also no significant difference in the total number of cells harvested from mice spleens at any dexamethasone-treated time point compared to either time- matched controls or other dexamethasone-treated groups (Figure 4.3B).

The most significantly differentially expressed gene at the 8 hour dexamethasone-treated time point was ZBTB16, a known transcriptional repressor and glucocorticoid response gene, which has been shown to modulate glucocorticoid sensitivity in CEM T-ALL cells (Wasim et al. 2010). Other known glucocorticoid response genes upregulated included TSC22D3 (D'Adamio et al. 1997) and SOCS1 (Yoshida et al. 2002), both downstream targets of the glucocorticoid receptor, FKBP5 (Vermeer et al. 2003), a co- chaperone of the glucocorticoid receptor, and MAP kinases 5, 6 and 8 (Lu et al. 2006). Downregulated genes at 8 hours included BCL-2 (Laane et al. 2007) and C-MYC (Thulasi et al. 1993), both previously described, but also HSP90B1, a glucocorticoid receptor co-chaperone (Table 4.4).

Page | 130

100 A 80

% PB in huCD45 20

8 hr dex 24 hr dex48 hr dex 0 hr control8 hr control 24 hr control48 hr control

500 B ** * )

6 400

300

200

Cell count (x10 count Cell 100

8 hr dex 24 hr dex48 hr dex 0 hr control8 hr control 24 hr control48 hr control

Figure 4.3. Engraftment and harvest data. Peripheral blood engraftment (A) and splenic harvest cell count (B) for each treated and control group. Error bars represent mean ± SEM. *, P=0.057, **, P=0.229 by Mann-Whitney test. Dex, dexamethasone treated and PB, peripheral blood.

Page | 131

The only pro-apoptotic gene consistently upregulated across multiple microarray analyses is the BH3-only BCL-2 family member BIM, and it has been shown that BIM has a critical role in glucocorticoid sensitivity and resistance (Bachmann et al. 2005). However in this study BIM was only induced 1.3 fold. Thus these identified genes are consistent with previous reports of glucocorticoid-induced genes in ALL. Within these experimental systems however there are significant potential differences in glucocorticoid exposure between in vitro and in vivo models – a crucial one is that cells in vitro are continuously exposed to glucocorticoid whereas in in vivo models the glucocorticoid is subject to pharmacokinetic and pharmacodynamic changes which more accurately reflect changes in patients.

Page | 132

Table 4.3. Genes upregulated 8 hours following dexamethasone treatment.

Probe ID Gene t P FDR Description

ILMN_5080450 ZBTB16 83.77 <2.2E-16 <2.2E-16 zinc finger/BTB domain containing 16 ILMN_3800088 MMP7 53.22 <2.2E-16 <2.2E-16 matrix metallopeptidase 7 ILMN_1770593 CH25H 53.14 <2.2E-16 <2.2E-16 cholesterol 25-hydroxylase ILMN_6560328 C6orf85 44.60 <2.2E-16 <2.2E-16 chromosome 6 orf 85 ILMN_7570561 TSC22D3 39.16 <2.2E-16 <2.2E-16 TSC22 domain family, member 3 ILMN_580187 PDE8B 33.88 <2.2E-16 3.90E-16 phosphodiesterase 8B ILMN_5130066 C8orf61 33.82 <2.2E-16 3.90E-16 chromosome 8 orf 61 ILMN_4120431 TMEM100 31.38 <2.2E-16 1.64E-15 transmembrane protein 100 ILMN_650553 BIN1 29.76 <2.2E-16 4.43E-15 bridging integrator 1 ILMN_1400373 SLA 29.57 <2.2E-16 4.63E-15 Src-like-adaptor ILMN_6330593 PTHR1 29.28 <2.2E-16 5.22E-15 parathyroid hormone receptor 1 ILMN_6110037 LILRA3 29.04 <2.2E-16 5.75E-15 leukocyte Ig-like receptor A3 ILMN_4150477 LOXL4 28.67 <2.2E-16 6.66E-15 lysyl oxidase-like 4 ILMN_2680079 OGFRL1 28.65 <2.2E-16 6.66E-15 opioid growth factor receptor-like 1 ILMN_4210411 NDRG2 28.20 <2.2E-16 8.62E-15 NDRG family member 2 ILMN_3780093 LILRA1 27.86 <2.2E-16 1.05E-14 leukocyte Ig-like receptor A1 ILMN_240441 IL1R2 27.46 <2.2E-16 1.33E-14 interleukin 1 receptor, type II ILMN_4730315 MERTK 26.14 <2.2E-16 3.31E-14 c-mer tyrosine kinase ILMN_3800538 ACPL2 25.90 <2.2E-16 3.72E-14 acid phosphatase-like 2 ILMN_6860392 UGT2B17 25.83 <2.2E-16 3.72E-14 UGT2 family, polypeptide B17 ILMN_4730541 SLC44A1 25.82 <2.2E-16 3.72E-14 solute carrier family 44, member 1 ILMN_4860546 CTHRC1 25.64 <2.2E-16 4.10E-14 collagen triple helix repeat containing 1 ILMN_3460270 ZHX3 24.56 <2.2E-16 8.79E-14 zinc fingers and homeoboxes 3 ILMN_10639 RASSF4 23.21 <2.2E-16 2.57E-13 Ras association domain family 4 ILMN_1190064 UGT2B7 23.13 <2.2E-16 2.67E-13 UGT2 family, polypeptide B7 ILMN_6400603 MGC2463 23.06 <2.2E-16 2.71E-13 poliovirus receptor related Ig domain ILMN_3450187 IRGM 23.04 <2.2E-16 2.71E-13 immunity-related GTPase family, M ILMN_6620528 MT1X 22.95 2.40E-16 2.85E-13 metallothionein 1X ILMN_1260341 IL13RA1 22.47 3.67E-16 4.13E-13 interleukin 13 receptor, alpha 1 ILMN_2650112 SLC16A2 22.25 4.48E-16 4.91E-13 solute carrier family 16, member 2 ILMN_5570170 PNMT 22.01 5.59E-16 5.95E-13 phenylethanolamine methyltransferase ILMN_870376 C9orf152 21.93 6.02E-16 6.25E-13 chromosome 9 orf 152 ILMN_3190379 TGFBR3 21.52 8.78E-16 8.89E-13 transforming growth factor, beta III ILMN_1780142 DSCR1 21.08 1.33E-15 1.31E-12 Down syndrome critical region gene 1 ILMN_2640341 FKBP5 20.63 2.05E-15 1.89E-12 FK506 binding protein 5 ILMN_7610136 LOC652626 20.43 2.48E-15 2.23E-12 Leukocyte Ig-like receptor B2 ILMN_1410609 CORO2A 20.34 2.72E-15 2.35E-12 coronin, actin binding protein, 2A ILMN_1780088 TBXA2R 20.29 2.84E-15 2.40E-12 thromboxane A2 receptor ILMN_270431 BAALC 20.23 3.02E-15 2.50E-12 brain and acute leukemia, cytoplasmic ILMN_6280176 GBE1 20.02 3.72E-15 3.01E-12 glucan branching enzyme 1 ILMN_6060113 TBX15 19.81 4.62E-15 3.67E-12 T-box 15

Page | 133

ILMN_4890743 IQSEC1 19.71 5.09E-15 3.97E-12 IQ motif and Sec7 domain 1 ILMN_150056 DPEP1 19.65 5.41E-15 4.13E-12 dipeptidase 1 ILMN_2060364 BTNL9 19.26 8.04E-15 5.91E-12 butyrophilin-like 9 ILMN_3830735 UPB1 19.23 8.30E-15 5.91E-12 ureidopropionase, beta ILMN_5670377 STYK1 19.15 9.09E-15 6.35E-12 serine/threonine/tyrosine kinase 1 ILMN_4390630 STAG3 18.72 1.42E-14 9.39E-12 stromal antigen 3 ILMN_4070048 NPHP4 18.44 1.91E-14 1.25E-11 nephronophthisis 4 ILMN_4220474 C6orf81 18.31 2.16E-14 1.39E-11 chromosome 6 open reading frame 81 ILMN_1470746 PTPN3 18.30 2.23E-14 1.41E-11 protein tyrosine phosphatase type 3 ILMN_5860576 C20orf133 18.25 2.36E-14 1.47E-11 MACRO domain containing 2 ILMN_6020468 PPP1R14A 18.18 2.52E-14 1.55E-11 protein phosphatase 1 subunit 14A ILMN_1400634 MT1M 18.10 2.76E-14 1.64E-11 metallothionein 1M ILMN_4250315 ITGA9 17.90 3.46E-14 2.03E-11 integrin, alpha 9 ILMN_5080471 MAP3K6 17.40 6.02E-14 3.44E-11 mitogen-activated protein kinase 6 ILMN_5360242 FLJ42461 17.36 6.28E-14 3.53E-11 smoothelin-like 2 ILMN_6620402 NUDT16 17.33 6.50E-14 3.60E-11 nudix-type motif 16 ILMN_3360112 TMEM2 17.26 7.04E-14 3.85E-11 transmembrane protein 2 ILMN_6840743 PER1 17.22 7.41E-14 3.99E-11 period homolog 1 ILMN_4220347 LRRC1 17.12 8.29E-14 4.33E-11 leucine rich repeat containing 1 ILMN_4850592 P2RY14 17.11 8.35E-14 4.33E-11 purinergic receptor P2Y, 14 ILMN_6560300 SLC31A2 16.91 1.05E-13 5.39E-11 solute carrier family 31 member 2 ILMN_4060091 DKFZ 16.87 1.11E-13 5.62E-11 DKFZp451A211 protein ILMN_6770370 LOC92196 16.28 2.23E-13 1.11E-10 death associated protein-like 1 ILMN_580487 IL9R 16.21 2.40E-13 1.18E-10 interleukin 9 receptor ILMN_1990300 SOCS1 16.18 2.49E-13 1.21E-10 suppressor of cytokine signaling 1 ILMN_5720424 NRP1 16.17 2.54E-13 1.22E-10 neuropilin 1 ILMN_4180427 CIB4 16.11 2.74E-13 1.30E-10 calcium and integrin binding 4 ILMN_4180544 ROPN1L 16.08 2.81E-13 1.32E-10 ropporin 1-like ILMN_4250167 SOX13 16.04 2.95E-13 1.37E-10 SRY (sex determining region Y)-box 13 ILMN_6330170 CHKA 15.81 3.94E-13 1.81E-10 choline kinase alpha, 3 ILMN_4560192 SFXN5 15.62 4.95E-13 2.25E-10 sideroflexin 5 ILMN_2810136 CAPN11 15.56 5.33E-13 2.40E-10 calpain 11 ILMN_2690709 VIPR1 15.38 6.68E-13 2.91E-10 vasoactive intestinal peptide receptor 1 ILMN_630091 NCOA7 15.38 6.69E-13 2.91E-10 nuclear receptor coactivator 7 ILMN_5390730 MGC17330 15.21 8.25E-13 3.55E-10 PI-3-K interacting protein 1 ILMN_130364 MST150 15.19 8.49E-13 3.62E-10 MSTP150 ILMN_3450241 KIAA0774 14.95 1.16E-12 4.77E-10 KIAA0774 ILMN_2230678 ACACB 14.80 1.41E-12 5.76E-10 acetyl-Coenzyme A carboxylase beta ILMN_5870307 LOC440359 14.78 1.44E-12 5.83E-10 similar to muscle Y-box protein YB2 ILMN_3840554 SPOCK2 14.76 1.49E-12 5.95E-10 sparc/osteonectin, cwcv and kazal 2 ILMN_5810600 MAP3K5 14.69 1.63E-12 6.47E-10 mitogen-activated protein kinase 5 ILMN_2360719 IRAK3 14.65 1.71E-12 6.65E-10 IL-1 receptor-associated kinase 3 ILMN_1510121 MTSS1 14.64 1.73E-12 6.66E-10 metastasis suppressor 1 ILMN_540671 LILRB2 14.54 1.98E-12 7.41E-10 leukocyte Ig-like receptor B2

Page | 134

ILMN_6980377 MTMR15 14.44 2.26E-12 8.39E-10 myotubularin related protein 15 ILMN_6220288 PRDM1 14.43 2.28E-12 8.39E-10 PR domain containing 1 ILMN_7330739 NDRG4 14.42 2.30E-12 8.39E-10 NDRG family member 4 ILMN_2600470 WDR60 14.20 3.10E-12 1.12E-09 WD repeat domain 60 ILMN_4050441 SH3MD4 14.16 3.27E-12 1.17E-09 SH3 multiple domains 4 ILMN_6760546 TIPARP 13.89 4.74E-12 1.64E-09 TCDD-inducible polymerase ILMN_2760537 MTE 13.89 4.75E-12 1.64E-09 metallothionein E ILMN_160019 SORT1 13.79 5.44E-12 1.83E-09 sortilin 1 ILMN_6330132 ISG20 13.60 7.00E-12 2.32E-09 IFN-stimulated exonuclease gene 20 ILMN_1510685 DOK4 13.52 7.86E-12 2.58E-09 docking protein 4 ILMN_1240228 PAG1 13.47 8.50E-12 2.77E-09 protein associated glycosphingolipid 1 ILMN_580592 CPNE8 13.32 1.04E-11 3.31E-09 copine VIII ILMN_5870301 KIAA0513 13.32 1.05E-11 3.31E-09 KIAA0513 ILMN_20129 CD52 13.32 1.05E-11 3.31E-09 CD52 molecule ILMN_1820386 PARVB 13.31 1.06E-11 3.31E-09 parvin, beta ILMN_6200402 MT1A 13.24 1.17E-11 3.64E-09 metallothionein 1A ILMN_290661 CLN8 13.10 1.43E-11 4.36E-09 ceroid-lipofuscinosis, neuronal 8 ILMN_670082 GNA12 13.08 1.47E-11 4.43E-09 guanine binding protein alpha 12 ILMN_5570286 TACC2 12.99 1.67E-11 5.00E-09 transforming, acidic coiled-coil protein 2 ILMN_3190411 STARD13 12.93 1.81E-11 5.32E-09 START domain containing 13 ILMN_4540138 NGB 12.92 1.85E-11 5.39E-09 neuroglobin ILMN_2000646 B4GALT4 12.83 2.10E-11 6.07E-09 UDP-Gal-T, polypeptide 4 ILMN_7100731 CYGB 12.81 2.17E-11 6.17E-09 cytoglobin ILMN_7050113 NTRK1 12.71 2.52E-11 7.09E-09 neurotrophic tyrosine kinase receptor 1 ILMN_2490670 GNPTAB 12.66 2.71E-11 7.52E-09 GlcNAc-1-PT, alpha and beta ILMN_20170 ZNF385 12.48 3.55E-11 9.72E-09 zinc finger protein 385 ILMN_2630687 CHPT1 12.43 3.80E-11 1.02E-08 choline phosphotransferase 1 ILMN_4120215 WASF2 12.43 3.81E-11 1.02E-08 WAS protein family, member 2 ILMN_5260494 TMLHE 12.39 4.06E-11 1.08E-08 trimethyllysine hydroxylase, epsilon ILMN_5220333 C14orf139 12.31 4.54E-11 1.20E-08 chromosome 14 orf 139 ILMN_3850440 FCER1G 12.12 6.07E-11 1.60E-08 Fc fragment of IgE, receptor ILMN_1030008 TGFB3 12.11 6.21E-11 1.63E-08 transforming growth factor, beta 3 ILMN_1450468 MYT1 12.02 7.04E-11 1.81E-08 myelin transcription factor 1 ILMN_7560541 SLC2A5 12.01 7.19E-11 1.83E-08 solute carrier family 2 member 5 ILMN_2030438 GBA2 12.01 7.21E-11 1.83E-08 glucosidase, beta (bile acid) 2 ILMN_6840328 SMAD3 12.00 7.35E-11 1.86E-08 SMAD family member 3 ILMN_3930390 SMAP1L 11.91 8.40E-11 2.11E-08 stromal membrane-assoc protein 1 ILMN_7570196 TSPAN9 11.90 8.54E-11 2.12E-08 tetraspanin 9 ILMN_6980546 CACNA1I 11.90 8.56E-11 2.12E-08 calcium channel, T type, alpha 1I ILMN_1710364 LCN6 11.89 8.72E-11 2.15E-08 lipocalin 6 ILMN_5360424 RPS6KA2 11.77 1.04E-10 2.54E-08 ribosomal protein S6 kinase ILMN_5890193 MS4A4A 11.72 1.14E-10 2.75E-08 membrane-spanning 4-domains A4 ILMN_3390292 KLF9 11.66 1.24E-10 2.98E-08 Kruppel-like factor 9 ILMN_5720059 GFOD1 11.65 1.26E-10 3.02E-08 glucose-fructose oxidoreductase 1

Page | 135

ILMN_7650523 TMEM46 11.57 1.43E-10 3.39E-08 transmembrane protein 46 ILMN_5700392 LOC654000 11.46 1.70E-10 3.95E-08 ribosome biogenesis protein BMS1 ILMN_4810348 C1orf188 11.40 1.88E-10 4.33E-08 chromosome 1 orf 188 ILMN_4280180 CHRNA3 11.39 1.91E-10 4.37E-08 cholinergic receptor, nicotinic, alpha 3 ILMN_270458 CRISPLD1 11.37 1.96E-10 4.45E-08 cysteine-rich secretory protein LCCL ILMN_450615 MT2A 11.37 1.97E-10 4.46E-08 metallothionein 2A ILMN_20470 GRASP 11.35 2.02E-10 4.51E-08 GRP1-associated scaffold protein ILMN_3370594 LILRA2 11.35 2.03E-10 4.51E-08 leukocyte Ig-like receptor A2 ILMN_5220397 RREB1 11.34 2.05E-10 4.53E-08 ras response element binding protein 1 ILMN_1410192 TDRD9 11.34 2.07E-10 4.56E-08 tudor domain containing 9 ILMN_4070259 LOC653133 11.27 2.30E-10 4.99E-08 guanine binding protein alpha 12 ILMN_5960682 RBPMS2 11.24 2.41E-10 5.21E-08 RNA binding protein multiple splicing 2 ILMN_1440300 SLC27A3 11.22 2.50E-10 5.37E-08 solute carrier family 27, member 3 ILMN_5050768 LONRF1 11.20 2.58E-10 5.53E-08 LON peptidase N-terminal ring finger 1 ILMN_6270273 KHDRBS3 11.18 2.67E-10 5.68E-08 KH domain, RNA binding 3 ILMN_7100603 KCNK3 11.17 2.70E-10 5.72E-08 potassium channel K3 ILMN_2320129 CSDA 11.03 3.38E-10 7.08E-08 cold shock domain protein A ILMN_3930022 LOC644739 10.99 3.63E-10 7.54E-08 Wiskott-Aldrich syndrome protein 4 ILMN_7400133 CUGBP2 10.90 4.20E-10 8.63E-08 CUG repeat, RNA binding protein 2 ILMN_3290301 FZD8 10.88 4.33E-10 8.76E-08 frizzled homolog 8 ILMN_7320520 MTUS1 10.88 4.33E-10 8.76E-08 mitochondrial tumor suppressor 1 ILMN_3780053 PALLD 10.82 4.79E-10 9.60E-08 palladin associated protein ILMN_6860162 LOC441019 10.74 5.49E-10 1.09E-07 hypothetical LOC441019 ILMN_5810154 ALOX15B 10.74 5.50E-10 1.09E-07 arachidonate 15-lipoxygenase, type B ILMN_3930736 CHST3 10.73 5.59E-10 1.09E-07 chondroitin 6 sulfotransferase 3 ILMN_60470 STX11 10.72 5.68E-10 1.10E-07 syntaxin 11 ILMN_3390484 SERINC2 10.69 5.95E-10 1.15E-07 serine incorporator 2 ILMN_1430647 TAX1BP3 10.61 6.82E-10 1.31E-07 Tax1 binding protein 3 ILMN_5960440 VDR 10.60 6.99E-10 1.34E-07 1,25- dihydroxyvitamin D3 receptor ILMN_6290735 EPHB3 10.51 8.10E-10 1.53E-07 EPH receptor B3 ILMN_2680372 SH2D4A 10.46 8.78E-10 1.64E-07 SH2 domain containing 4A ILMN_2480050 SOX7 10.44 9.13E-10 1.69E-07 SRY (sex determining region Y)-box 7 ILMN_130128 LOC285016 10.41 9.61E-10 1.76E-07 hypothetical protein LOC285016 ILMN_4890451 GRAMD3 10.39 9.87E-10 1.80E-07 GRAM domain containing 3 ILMN_770161 C10orf73 10.39 9.92E-10 1.81E-07 chromosome 10 open reading frame 73 ILMN_2450202 KIF3C 10.35 1.05E-09 1.88E-07 kinesin family member 3C ILMN_6840468 HAL 10.35 1.06E-09 1.89E-07 histidine ammonia-lyase ILMN_2470070 TBL1X 10.30 1.15E-09 2.04E-07 transducin (beta)-like 1X-linked ILMN_2320114 KLF13 10.27 1.22E-09 2.15E-07 Kruppel-like factor 13 ILMN_6380112 DIP 10.23 1.31E-09 2.27E-07 death-inducing-protein ILMN_2470358 IFNGR1 10.22 1.32E-09 2.30E-07 interferon gamma receptor 1 ILMN_4250735 IL27RA 10.07 1.70E-09 2.91E-07 interleukin 27 receptor, alpha ILMN_1470215 MAP3K8 10.07 1.72E-09 2.91E-07 mitogen-activated protein kinase 8 ILMN_2940373 TACC1 10.06 1.74E-09 2.94E-07 transforming, acidic coiled-coil protein 1

Page | 136

Table 4.4. Genes downregulated 8 hours following dexamethasone treatment.

Probe ID Gene t P FDR Description

ILMN_770538 LYSMD2 -15.49 5.81E-13 2.58E-10 LysM domain 2 ILMN_7150059 STAMBPL1 -14.61 1.79E-12 6.84E-10 STAM binding protein-like 1 ILMN_5340692 STRBP -14.56 1.93E-12 7.31E-10 spermatid RNA binding protein ILMN_4210397 GLDC -14.05 3.80E-12 1.34E-09 glycine dehydrogenase ILMN_6980327 DKC1 -13.79 5.44E-12 1.83E-09 dyskeratosis congenita 1, dyskerin ILMN_50086 TCF12 -13.23 1.19E-11 3.69E-09 transcription factor 12 ILMN_4860356 BYSL -12.81 2.17E-11 6.17E-09 bystin-like ILMN_4280228 IVNS1ABP -12.70 2.55E-11 7.12E-09 influenza virus NS1A binding protein ILMN_1990379 SLFN11 -11.82 9.63E-11 2.36E-08 schlafen family member 11 ILMN_5220338 MPEG1 -11.64 1.27E-10 3.03E-08 macrophage expressed gene 1 ILMN_450168 SFRS7 -11.50 1.60E-10 3.74E-08 splicing factor, arginine/serine-rich 7 ILMN_3460687 KIAA0690 -11.42 1.81E-10 4.19E-08 ribosomal RNA processing 12 homolog ILMN_3400360 MAPRE2 -11.36 1.99E-10 4.48E-08 microtubule-associated protein RP/EB2 ILMN_4010414 PPFIBP1 -11.12 2.92E-10 6.16E-08 PTPRF interacting binding protein 1 ILMN_1190139 UGT3A2 -10.99 3.61E-10 7.54E-08 UGT3 family, polypeptide A2 ILMN_4150201 BCL2 -10.93 3.99E-10 8.24E-08 B-cell CLL/lymphoma 2 ILMN_780240 C12orf24 -10.85 4.53E-10 9.13E-08 chromosome 12 open reading frame 24 ILMN_6760167 MARCH3 -10.73 5.60E-10 1.09E-07 membrane-associated ring finger 3 ILMN_3940615 PUS7 -10.52 7.99E-10 1.52E-07 pseudouridylate synthase 7 homolog ILMN_20544 GART -10.41 9.53E-10 1.76E-07 ribosylglycinamide formyltransferase ILMN_2480326 HSP90B1 -10.36 1.05E-09 1.88E-07 heat shock protein 90kDa beta 1 ILMN_5270367 CTSC -10.25 1.26E-09 2.20E-07 cathepsin C ILMN_5420095 MYC -10.21 1.36E-09 2.34E-07 v-myc viral oncogene homolog ILMN_4610180 PIK3C2B -10.20 1.38E-09 2.37E-07 PI-3-K, class 2, beta polypeptide ILMN_6450300 GEMIN4 -10.00 1.95E-09 3.27E-07 gem associated protein 4

Page | 137

4.5.2 Later 24 hour and 48 hour time points At the later time points, significant differential gene expression was much less marked and predominantly downregulated. At 24 hours 5 genes were upregulated (t-statistic >6) and 10 genes downregulated (t-statistic <-6, Table 4.5), and at 48 hours 1 gene was upregulated (t-statistic >6) and 15 genes downregulated (t-statistic <-6, Table 4.6). At 24 hours, upregulated genes included NFKBIA, an inhibitor of NF-κB, and TRIM74, which was sustained at 48 hours, the significance of which is uncertain. Downregulated genes were those involved in cell cycle progression, including CCNF at 24 hours, and CCNF, CDC20 and AURKA at 48 hours, consistent with growth arrest.

Page | 138

Table 4.5. Genes regulated 24 hours following dexamethasone treatment.

Probe ID Gene t P FDR Description

Upregulated ILMN_3930687 FAM112A 6.67 1.32E-06 0.0091 family 112A ILMN_6620255 TRIM74 6.29 3.06E-06 0.0132 tripartite motif-containing 74 ILMN_4280113 NFKBIA 6.23 3.48E-06 0.0138 nuclear factor kappa B inhibitor, alpha ILMN_2140136 EMR2 6.10 4.65E-06 0.0149 egf-like, mucin-like, receptor-like 2 ILMN_7000397 ANKRD15 6.08 4.91E-06 0.0149 ankyrin repeat domain 15 Downregulated ILMN_870524 HOXB8 -8.60 2.53E-08 0.0011 homeobox B8 ILMN_4830520 LOC144501 -6.72 1.19E-06 0.0091 hypothetical protein LOC144501 ILMN_6110332 ARHGAP19 -6.70 1.24E-06 0.0091 Rho GTPase activating protein 19 ILMN_2970619 ESPL1 -6.65 1.38E-06 0.0091 extra spindle pole bodies homolog 1 ILMN_3130541 CCNF -6.64 1.43E-06 0.0091 cyclin F ILMN_4760577 CENPA -6.62 1.46E-06 0.0091 centromere protein A ILMN_4810646 PIF1 -6.54 1.76E-06 0.0095 PIF1 5'-to-3' DNA helicase homolog ILMN_1070762 PSRC1 -6.40 2.38E-06 0.0114 proline/serine-rich coiled-coil 1 ILMN_4860703 LOC648695 -6.19 3.82E-06 0.0138 retinoblastoma binding protein 4 ILMN_1110538 INCENP -6.05 5.19E-06 0.0149 inner centromere protein antigens

Page | 139

Table 4.6. Genes regulated 48 hours following dexamethasone treatment.

Probe ID Gene t P FDR Description

Upregulated ILMN_6620255 TRIM74 6.30 3.01E-06 0.0089 tripartite motif-containing 74 Downregulated ILMN_4810646 PIF1 -8.85 1.58E-08 0.0004 PIF1 5'-to-3' DNA helicase homolog ILMN_870524 HOXB8 -8.66 2.26E-08 0.0004 homeobox B8 ILMN_1450193 LGALS1 -8.57 2.66E-08 0.0004 lectin, galactoside-binding, soluble 1 ILMN_4760577 CENPA -7.64 1.71E-07 0.0018 centromere protein A ILMN_4730605 AURKA -7.47 2.42E-07 0.0021 aurora kinase A ILMN_1500010 CDC20 -6.84 9.09E-07 0.0053 CDC20 cell division cycle 20 homolog ILMN_4060064 HMMR -6.82 9.61E-07 0.0053 hyaluronan-mediated motility receptor ILMN_2070408 MID1 -6.80 9.97E-07 0.0053 midline 1 (Opitz/BBB syndrome) ILMN_2070288 MT1E -6.66 1.36E-06 0.0065 metallothionein 1E ILMN_1070762 PSRC1 -6.60 1.55E-06 0.0067 proline/serine-rich coiled-coil 1 ILMN_150543 C20orf129 -6.46 2.12E-06 0.0077 chromosome 20 orf 129 ILMN_5870193 FAM64A -6.45 2.14E-06 0.0077 family 64A ILMN_2810201 KIF14 -6.34 2.77E-06 0.0089 kinesin family member 14 ILMN_1050195 KIF20A -6.28 3.11E-06 0.0089 kinesin family member 20A ILMN_3130541 CCNF -6.05 5.21E-06 0.0131 cyclin F

Page | 140

4.6 Functional analysis The interpretation of lists of differentially expressed genes is challenging given the numbers of genes involved, and is often subject to arbitrary statistical cut- offs (P value, FDR or FC). This approach has several limitations – no individual gene may meet the threshold for statistical significance, or alternatively there may be many statistically significant genes identified without a unifying biological theme. Further, single gene analysis may miss important effects on pathways - a 20% increase in expression across all genes in a pathway may be biologically more important than a 20-fold increase in a single gene.

I thus performed functional, or pathway analysis using Gene Set Enrichment Analysis (GSEA) version 2.04 that focuses on groups of genes that share common biological function, chromosomal location or regulation (Subramanian et al. 2005). I compared the limma moderated t-statistic for each probe in a pre- ranked file, against the c2_all collection of gene sets from the Molecular Signatures Database (Subramanian et al. 2005) version 2.5 with 1000 permutations. The c2_all collection is made up of gene sets curated from online pathway databases, publications in PubMed and knowledge of domain experts. The similarity of the leading edges of the top 100 up-regulated and downregulated genesets was assessed using meta-GSEA.

Functional analysis of the expression profiles obtained at 8 hours and 24 hours after dexamethasone treatment (Appendix A), revealed a significant upregulation of metabolic pathways, particularly adipogenesis at 8 hours, and a marked effect on pathways associated with cell cycling and proliferation, particularly downregulation of C-MYC at 8 hours and NF-κB at 24 hours, and upregulation of apoptotic pathways at 24 hours. Glucocorticoids are known to have effects on multiple cellular metabolic pathways, including glucose and carbohydrate metabolism, and have pro-adipogenic effects (Pantoja et al. 2008). Suppression of C-MYC is a critical step prior to the initiation of apoptosis by dexamethasone in ALL (Medh et al. 2001) and suppression of NF-κB has been described (Auphan et al. 1995). Page | 141

4.7 Comparison of models I then proceeded to determine whether the molecular response to glucocorticoids in this xenograft model of ALL mimicked the effects seen in either glucocorticoid-treated patients with ALL (Schmidt et al. 2006) or a cell-line model of ALL (Rainer et al. 2009). Comparing gene expression profiles from multiple experiments is notoriously difficult and typically any true similarities are swamped by technical differences in microarray vendor, normalisation strategies and analytical approach. By summarising genes at the gene set level (such as genes in the same pathway), these technical differences are mitigated, allowing comparison of samples from multiple studies.

I applied parametric analysis of gene set enrichment (PGSEA) (Kim and Volsky 2005) implemented in the PGSEA package (version 1.20.1, Furge and Dykema) from the Bioconductor project (Gentleman et al. 2004), with some modifications to the algorithm to assess significance of the genes that are in the geneset and represented on the microarray, and to allow more control over sample specification. Expression levels of each gene in each 6-8 hour glucocorticoid- treated sample were converted to expression ratios relative to patient matched controls before glucocorticoid treatment (Schmidt et al), time-matched controls (Rainer et al), or time 0 hours (xenografts). Within each dataset, these gene- level ratios were summarised into geneset-level Z-scores, using PGSEA with genesets from the c2_all collection. The Z-scores from each sample from the 3 studies were combined and then compared by two-dimensional hierarchical clustering of the top 100 gene sets demonstrating the greatest variance across the combined studies.

The hierarchical cluster dendrogram used to identify the relationships between the glucocorticoid responses in the different ALL models is shown in Figure 4.4. This revealed considerable heterogeneity in the molecular response within patients, split into at least 2, and possibly 4 different groups (green bars), which may represent different biological responses. Relative to this inter-patient heterogeneity, both cell lines (purple bars) and xenografts (black bars) were remarkably reproducible. Further, in Section 5.3.1 I have shown that the Page | 142

addition of extra ALL xenografts mirrored the heterogeneity seen in patients from whom they were derived. It is also evident that overall, glucocorticoid- treated xenografts co-clustered with a group of 3 patients (B-ALL-37, B-ALL-38, and B-ALL-40), all of whom had BCP-ALL and a good early prednisolone response, with varied cytogenetics (hyperploidy, t(12;21), and normal respectively). At more relaxed clustering thresholds, the glucocorticoid-treated xenografts clustered with 4 other patients with BCP-ALL (B-ALL-24, B-ALL-31, B-ALL-33 and B-ALL-43) and the cell lines.

I identified 5 clusters of gene sets with distinct expression profiles, each behaving differently in the 3 models of ALL. Cluster 1 demonstrated the markedly heterogeneous patterns seen in patient samples, with the xenograft samples showing a pattern similar to 8 of the patients; cluster 2 showed genesets that showed strong enrichment in the cell line study, and included a number of genesets associated with cell proliferation; cluster 3 did not show any striking differences across the three ALL models; cluster 4 showed genesets downregulated in both xenografts and cell lines compared to the patient samples, and included a number genesets associated with cell cycle progression, nucleic acid replication and MYC; cluster 5 showed genesets strongly downregulated in the xenograft and cell line models, and included genesets associated with MYC and metabolic processes, particularly catabolism and energy production. In this limited comparison, it is clear that glucocorticoid-induced gene expression patterns seen in ALL are dependent on the experimental model, and that the patterns derived from the xenograft model show a greater similarity to patterns derived from a subset of the patient-derived data than to cell lines.

Page | 143

Figure 4.4. Parametric GSEA of combined top 100 glucocorticoid-induced gene sets with greatest variance from xenograft, patient and cell line models. Hierarchical clustering with gene sets in rows, samples in columns (xenografts – black, patient – green, cell line – purple). Each colour of each cell represents the Z-score (see legend). Boxes 1-5 represent defined clusters.

Page | 144

4.8 Search for GRE motifs Glucocorticoids enter the cell by passive diffusion and bind to the cytosolic GR. The activated GC-GR complex is translocated to the nucleus where it binds to palindromic DNA sequences known as Glucocorticoid Response Elements (GREs), resulting in transactivation and transrepression of target genes (detailed in Section 1.7). I proceeded to determine whether GRE motifs (sequence GGTACAnnnTGTTCT) could be identified within the DNA sequences upstream or downstream of the top genes upregulated by dexamethasone. A search of the TRANSFAC database v8.3 (Matys et al. 2003) of CoMoDis (Donaldson and Gottgens 2007) identified GRE motifs (within 100 kb either side of the gene) in only 25 (14.5%) of the top 173 upregulated genes at the 8 hour time point in this study, and no GRE motifs were identified in the upregulated genes at 24 or 48 hours. This supports accumulating evidence that glucocorticoids exert long-range effects through very distal steroid receptor binding sites (Paakinaho et al. 2010).

Analysis of significantly differentially expressed glucocorticoid-induced genes in an in vitro cell line study (Rainer et al. 2009) revealed a similar number of early response genes after 6 hours of exposure [60 upregulated (t-stat >10) and 27 downregulated (t-stat <-10)] but a significantly greater number of genes after 24 hours [593 upregulated (t-stat >10) and 782 downregulated (t-stat <-10)]. All but 2 of the genes upregulated at 6 hours remained significantly upregulated at 24 hours, and 17 of the downregulated genes at 6 hours remained downregulated at 24 hours. GRE motifs were identified in 15 (25.0%) of the top 60 upregulated genes at 6 hours, and 87 (14.6%) of the top 593 genes at 24 hours.

This observed difference at later time points is consistent with continuous rather than physiological glucocorticoid exposure. In addition, in the cell line study, the Glucocorticoid Receptor (GR, NR3C1) underwent highly significant early and sustained auto-upregulation, which in the continuous presence of ligand drives downstream gene expression. In contrast, in the xenograft model minimal GR upregulation was seen at the early time point but no significant change in GR expression was seen at either of the later time points. Page | 145

4.9 Replicate analysis Given the good statistical power observed in Figure 4.2, I proceeded to determine whether fewer replicates could be used whilst still identifying a majority of the differentially expressed genes. The stability of results when reducing the number of replicates was assessed using the Recovery Score method (Pavlidis et al. 2003) from the GeneSelector package (version 1.4.0) of the Bioconductor project (Gentleman et al. 2004).

Replicate analysis (Figure 4.5) revealed that at the 8 hour dexamethasone- treated time point, a dataset with high signal and differential expression, using data from any 3 randomly chosen biological replicates instead of 4 resulted in excellent recovery scores of >0.9. That is, on average, >90% of the differentially expressed genes identified from all 4 samples were also identified in any combination of 3 arrays. At 24 hours, a time point with less signal, the average recovery score was 0.85 with 3 replicates, but was more variable than at 8 hours. Using just 2 biological replicates recovered 88% of the list of differentially expressed genes at 8 hours, which dropped to 14% at 24 hours. This confirms that the 8 hour time point has the strongest signal, which is reproducible across different subsets of biological replicates. I concluded that it was important to use a minimum of 3 biological replicates, since fewer replicates destabilised the ability to identify differentially expressed genes. This has important considerations for experimental design, and has significant implications on cost and animal numbers.

Page | 146

Figure 4.5. Recovery scores at 8 hours and 24 hours when randomly selecting all combinations of 3 replicates (3rep) or 2 replicates (2rep) from the set of 4 biological replicates. The Recovery Score represents the proportion of differentially expressed genes from all 4 replicates recovered when using fewer replicates. Error bars represent mean ± SEM.

Page | 147

4.10 Validation of results Although genome-wide gene expression is an increasingly reliable tool, it is preferable to validate candidate genes and/or proteins. The primary aim of this pilot study was to determine the optimal experimental design to investigate in vivo glucocorticoid-induced gene expression in the NOD/SCID xenograft mouse model, and thus validation was limited to a number of genes from the BCL-2 family. These proteins are key regulators of apoptosis, and have a critical role in glucocorticoid-induced cell death (Section 1.8.3). For validation, cDNA was synthesised from extracted RNA and RQ-PCR of the BCL-2 family members performed using a Taqman LDA card. Immunoblotting for candidate proteins was performed by western blot.

Analysis of the pro-apoptotic members revealed an almost 50 fold increase in BMF mRNA expression at 8 hours following dexamethasone treatment, which returned to baseline by the later time points (Figure 4.6). This is in contrast to the microarray data in which minimal changes in BMF mRNA expression were seen. However despite several separate immunoblots with 2 different BMF antibodies, I was unable to demonstrate any BMF protein expression either at baseline or up to 24 hours following exposure to dexamethasone (Figure 4.7). Interestingly, a report analysing glucocorticoid-induced changes in the BCL-2 family members in ALL cells extracted from children after glucocorticoid monotherapy demonstrated induction of BMF mRNA at 6-8 hours, which was sustained at 24 hours. However correlation with BMF protein expression in the patient cells was not reported, and although BMF protein was induced in CEM T-ALL cells, this was only evident after 24-36 hours of continuous in vitro exposure to dexamethasone (Ploner et al. 2008). Although the possibility of ineffective antibodies cannot be excluded by these data, it is likely that BMF protein exists at undetectable levels and induction by pro-apoptotic stimuli is delayed compared to early transcriptional mRNA expression.

Page | 148

Expression of BAD was increased approximately 5 fold at 8 hours whilst the majority of the other pro-apoptotic members, including BIM (BCL2L11) showed an approximate 2 fold increase at 8 hours returning to baseline by 24 and 48 hours (Figure 4.8). It has been previously shown that BIM is induced at the mRNA and protein level in spleen-derived glucocorticoid-sensitive ALL xenografts in vitro (Bachmann et al. 2005). Here I demonstrated that in vivo BIM protein is induced at 8 hours and remains elevated relative to control at 24 hours (Figure 4.9), indicating stable protein expression despite mRNA levels returning to baseline.

The genome-wide data had revealed significant downregulation of BCL-2 at 8 hours following dexamethasone, but by RQ-PCR BCL-2 mRNA expression, along with other pro-survival members MCL-1 and BCL2A1 was increased approximately 2 fold (Figure 4.10) similar to previous reports (Ploner et al. 2008). There was however minimal change in BCL-2 protein expression (Figure 4.11). These results together indicate that glucocorticoids exert competing effects on leukaemia cells, and the BCL-2 family members are subject to complex transcriptional and translational control.

RQ-PCR of the GR (NR3C1) revealed minimal changes in GR mRNA expression levels at the time points studied and upregulation of 2 known downstream targets SOCS1 and GILZ (TSC22D3) (Figure 4.12) demonstrating concordance with the microarray data and confirming intact drug-receptor signalling pathways.

Page | 149

60 BMF 50

Fold change Fold 20

0 0 8 16 24 32 40 48 Time (hours) post Dex

Figure 4.6. mRNA expression of BMF by RQ-PCR following in vivo treatment with dexamethasone. Fold change was calculated relative to untreated control. Error bar represents mean ± SD.

Figure 4.7. BMF protein expression in ALL-3 treated with dexamethasone. Region shown corresponds to MW 18-25 kDa.

Page | 150

BIK HRK 6 BIM (BCL2L11) NOXA (PMAIP1) 4 BID BOK PUMA (BBC3) Fold change Fold 2 BAX BAK (BAK1) 0 BAD 0 8 16 24 32 40 48 Time (hours) post Dex

Figure 4.8. mRNA expression of pro-apoptotic BCL-2 family members (minus BMF) by RQ-PCR following in vivo treatment with dexamethasone. Fold change was calculated relative to untreated control. Error bars represent mean ± SD.

Page | 151

B 250

200

150

100

50 Total Bim (% Bim Total Actin)

0 Control 8 h Dex 24 h Dex

Figure 4.9. BIM protein expression in ALL-3 treated with dexamethasone. Immunoblots (A) and quantification (B) from 4 mice in each group. Expression values normalised to Actin (ACT). Error bars represent mean ± SEM.

Page | 152

BCL2 2 BCL-xL (BCL2L1) BCL2A1 MCL-1 BOO/DIVA (BCL2L10) 1 Fold change Fold

0 0 8 16 24 32 40 48 Time (hours) post Dex

Figure 4.10. mRNA expression of pro-survival BCL-2 family members by RQ-PCR following in vivo treatment with dexamethasone. Fold change was calculated relative to untreated control. No values for BCL-w (BCL2L2) are shown as the cycle threshold was not reached after 40 cycles. Error bars represent mean ± SD.

Page | 153

B 200

150

100

BCL2 (%BCL2 Actin) 50

0 Control 8 h Dex 24 h Dex

Figure 4.11. BCL-2 protein expression in ALL-3 treated with dexamethasone. Immunoblots (A) and quantification (B) from 4 mice in each group. Values were normalised to Actin (ACT). Error bars represent mean ± SEM.

Page | 154

8 GR (NR3C1) SOCS1 6 GILZ (TSC22D3) 4 Fold change Fold 2

0 0 8 16 24 32 40 48 Time (hours) post Dex

Figure 4.12. mRNA expression of the GR and downstream targets by RQ- PCR following in vivo treatment with dexamethasone. Fold change was calculated relative to untreated control. No values for GILZ (TSC22D3) at 48 hours available. Error bars represent mean ± SD.

Page | 155

4.11 Conclusions The results of the pilot study demonstrated that the 8 hour time point provides the highest number of significantly differentially expressed genes, that time- matched controls are redundant and excellent recovery scores can be obtained with 3 replicates. I was thus able to establish the optimal experimental design for future gene expression studies, and used this for the experiments detailed in Chapter 5. Validation of the results was strongly concordant with the gene expression studies, and I concluded that the NOD/SCID ALL xenograft mouse model provides biologically relevant insights into glucocorticoid-induced gene expression, in a consistent, reproducible and clinically relevant model system.

Page | 156

5 Evaluation of Glucocorticoid Resistance Reversing Agents

Page | 157

5.1 Introduction The previous chapters detailed the establishment of the xenograft panel and the optimal experimental design to investigate in vivo glucocorticoid gene expression in ALL. Here I describe how I obtained in vivo dexamethasone- induced gene expression data from all 10 xenografts, generated a number of gene signatures associated with glucocorticoid resistance to interrogate the Connectivity Map (CMap), and identified heat shock protein 90 kDa (HSP90) inhibitors as agents which could potentially reverse glucocorticoid resistance. I then performed a number of in vitro and in vivo experiments assessing the efficacy of an established HSP90 inhibitor, 17-dimethylaminoethylamino-17- demethoxygeldanamycin (17-DMAG, alvespimycin) as a single agent and in combination with dexamethasone in a number of glucocorticoid-resistant xenografts. I also evaluated a novel hydrocarbon-stapled BIM-BH3 peptide synthesised by collaborators at the University of Sydney.

5.2 Generation of xenograft in vivo gene expression Primary xenograft cells were inoculated into NOD/SCID (ALL-26, ALL-50, ALL- 51, ALL-54, and ALL-56) or NSG mice (ALL-28, ALL-52, ALL-53, ALL-55 and ALL-57) and huCD45+ monitored weekly over a total period of 23 weeks. From the pilot study, I had determined that the optimal experimental design was to investigate gene expression at 8 hours following dexamethasone treatment in groups of 3 mice. Thus, using this experimental setup, when high levels of engraftment were evident, mice were randomised into groups of 3 and treated with a single dose of dexamethasone (15 mg/kg IP). Cells were harvested from spleens at 0 hours (pre-treatment control), 8 hours and 24 hours following treatment. I extracted RNA and verified the RNA integrity number (RIN). All samples had RINs ≥ 8.0 and so were suitable for further processing. RNA was amplified, hybridised onto Illumina WG-6 BeadChips, and scanned on the Bead Array Reader. Raw gene expression data were pre-processed using log transformation and quantile normalisation, and specific analyses performed as detailed below.

Page | 158

5.3 Gene expression analysis

5.3.1 Hierarchical clustering For each xenograft, a list of dexamethasone-regulated genes was generated by comparing by limma the gene expression profiles of the 3 replicates at 8 hours following dexamethasone treatment with the 3 (untreated) controls. Each xenograft demonstrated a number of significantly differentially expressed genes (Table 5.1) confirming experimental integrity. I then compared the individual dexamethasone-induced gene expression profiles by unsupervised hierarchical clustering (Figure 5.1). Surprisingly, there was no cluster separation between PPRs and PGRs, suggesting that the dexamethasone responses are heterogeneous.

Page | 159

Table 5.1. Number of differentially expressed gene probes in vivo at 8 hours following dexamethasone treatment compared to controls.

P value False Discovery Rate (FDR) Xenograft <0.05 <0.001 <0.0001 <0.25 <0.1 <0.05 ALL-28 10634 3444 1743 14718 8551 6414 ALL-50 6289 1604 792 4895 2888 2061

ALL-54 8011 2532 1406 8197 5171 3921 PPR ALL-55 6183 1467 728 4709 2646 1839 ALL-57 10244 3026 1426 14048 7764 5629 ALL-26 5176 706 200 2791 1128 531 ALL-51 8328 2251 1028 8887 5214 3785

ALL-52 6605 1396 509 5593 2914 1930 PGR ALL-53 7249 1778 821 6768 3699 2659 ALL-56 8092 3003 1755 7926 5526 4479

Page | 160

Figure 5.1. Hierarchical clustering of dexamethasone-induced gene expression profiles. Unsupervised cluster dendrogram of the top 500 probes with the greatest variance across all samples. X, ALL xenograft. Red, PPRs and blue, PGRs.

Page | 161

5.3.2 Differentially expressed genes PPR v PGR I then used limma to identify dexamethasone-induced genes significantly differentially expressed between the 5 PPR profiles and the 5 PGR profiles. Although a number of genes with significant P values were identified (Figure 5.2), none were significant by FDR after correction for multiple hypothesis testing (Table 5.2), indicating that it was not possible to exclude the possibility that these genes had been identified by chance. When ALL-54, a PPR which has biological responses more characteristic of a PGR was reclassified as a PGR, and the 4 remaining PPR profiles compared to the 5 PGR profiles plus ALL-54 (PGR54), again a number of genes were identified, but none significant by FDR (Table 5.2). When ALL-56 was reclassified as a PPR as it is resistant in vivo, and this PPR56 group compared to PGR54, the results were essentially similar (Table 5.2). This frustrating situation is related to small sample size, and thus I was not able to identify and pursue individual genes as potential markers of glucocorticoid resistance.

I then attempted to identify biologically relevant pathways from the PPR v PGR comparison by performing GSEA, extracting the leading edge genes of the top 50 genesets by meta-GSEA, and refining the list to include only those genes that appeared in the leading edges of at least 5 of the top 50 genesets (resulting in 208 genes upregulated and 28 genes downregulated in PPRs, Appendix B). Analysis of Gene Ontology revealed a significant enrichment in PPRs of upregulated pathways involved in cell cycle progression and DNA replication, and of downregulated pathways involved in MAP kinase signaling (Figure 5.3).

Page | 162

Figure 5.2. Heatmap of genes differentially expressed between PGRs and PPRs. Genes with a P-value <0.001 shown, none significant by FDR. PPRs denoted by red bar, PGRs by blue bar. Red denotes relative upregulation, blue denotes relative downregulation.

Page | 163

Table 5.2. Numbers of significantly differentially expressed genes.

P value False Discovery Rate (FDR)

<0.05 <0.001 <0.0001 <0.25 <0.1 <0.05 5 PPR v 5 PGR 2821 53 3 0 0 0 4 PPR v PGR54 2981 70 8 0 0 0 PPR56 v PGR54 2428 65 4 0 0 0

PGR54, PGR xenografts plus ALL-54, and PPR56, PPR xenografts with ALL-54 replaced by ALL-56.

Page | 164

Figure 5.3. Genes downregulated in PPRs on exposure to dexamethasone in MAP kinase pathways. Schematic representation of the MAP kinase pathways (http://www.biocarta.com/pathfiles/h_mapkPathway.asp). Genes identified to be downregulated in PPRs on exposure to dexamethasone are indicated by a white circle.

Page | 165

MAP kinase pathways are activated in the response to a number of extrinsic trophic factors and stresses, and have a role in the response to glucocorticoids. There have been reports on the use of MEK/ERK inhibitors to overcome glucocorticoid resistance in ALL (Rambal et al. 2009), although this is counter to the finding of downregulation of MAP kinase pathways in PPRs. However further exploration of the role of this pathway in the glucocorticoid response is warranted.

From previous reports the BH3-only BCL2 family members BIM and to a lesser extent BMF are critical proteins in the glucocorticoid-evoked apoptotic response in ALL (Section 1.8.3). I extracted the t-statistic values for the BIM and BMF genes from the individual xenograft dexamethasone-induced profiles, and compared the same groups as before by the Mann-Whitney test. Interestingly, as the groups were refined, the more statistically significant the difference in BIM (but not BMF) gene induction between PPRs and PGRs became (Figure 5.4). Thus by the less stringent test and without multiple comparison testing, there is a true difference in BIM gene induction between glucocorticoid sensitive and resistant ALL xenografts, although the magnitude of induction is small and of questionable biological relevance.

Page | 166

A PPR v PGR 20 p=0.222

10 p=0.222 5 t-statistic

-5

BIM - PGR BIM - PPR BMF - PGR BMF - PPR

B PPR v PGR54 20 p=0.762

10 p=0.038* 5 t-statistic

-5

BIM - PPR BMF - PPR BIM - PGR54 BMF - PGR54

C PPR56 v PGR54 20 p=0.843 15

10 p=0.016* 5 t-statistic

-5

BIM - PGR54 BIM - PPR56 BMF - PGR54 BMF - PPR56

Figure 5.4. BIM and BMF mRNA induction. The t-statistics from each xenograft dexamethasone-induced profile were extracted and groups compared by Mann-Whitney test: A, PPR v PGR; B, PPR v PGR54, and C, PPR56 v PGR54. Error bars represent mean ± SEM.

Page | 167

5.3.3 Validation of BIM I proceeded to investigate BIM protein induction following dexamethasone in vivo. Protein lysates were prepared from cells harvested from mice at baseline (0 hours) and at 8 hours and 24 hours following dexamethasone (15 mg/kg IP). The immunoblots are shown in Figure 5.5 and quantified in Figure 5.6. In general, across the whole xenograft panel, the magnitude of mRNA and protein changes was small. For the PPRs, ALL-28, ALL-55 and ALL-57 showed a lack of induction of BIM at the mRNA and protein level, whereas ALL-54 demonstrated mRNA induction only, and in ALL-50 both were suppressed. For the PGRs, ALL-26, ALL-52 and ALL-53 showed approximate 1.3 fold induction at both the mRNA and protein level, whereas in ALL-51 and ALL-56 BIM protein was suppressed. For ALL-56, this is one potential explanation why this xenograft is highly dexamethasone-resistant in vivo. In general mRNA induction correlated with protein induction at 8 hours, although the magnitude of induction in the xenografts is small, with the majority of samples demonstrating less than 1.5 fold induction or suppression. Hence this data is subject to small variations in experimental technique, and it is not possible to conclude there is a significant biological difference in BIM regulation to explain differences in glucocorticoid response.

Page | 168

Figure 5.5. BIM protein induction following dexamethasone. Representative immunoblots of BIM protein at baseline, 8 hours and 24 hours following dexamethasone 15 mg/kg in vivo in (A) PPRs and (B) PGRs. ACT, actin.

Page | 169

ALL-28 ALL-26 2.0 2.0

1.5 1.5

1.0 1.0 Induction Induction 0.5 0.5

0.0 0.0 0 8 16 24 0 8 16 24 Hours post-dexamethasone Hours post-dexamethasone

ALL-50 ALL-51 2.0 2.0

1.5 1.5

1.0 1.0 Induction Induction 0.5 0.5

0.0 0.0 0 8 16 24 0 8 16 24 Hours post-dexamethasone Hours post-dexamethasone

ALL-54 ALL-52 2.0 2.0

1.5 1.5

1.0 1.0 Induction Induction 0.5 0.5

0.0 0.0 0 8 16 24 0 8 16 24 Hours post-dexamethasone Hours post-dexamethasone

ALL-55 ALL-53 2.0 2.0

1.5 1.5

1.0 1.0 Induction Induction 0.5 0.5

0.0 0.0 0 8 16 24 0 8 16 24 Hours post-dexamethasone Hours post-dexamethasone

ALL-57 ALL-56 2.0 2.0

1.5 1.5

1.0 1.0 Induction Induction 0.5 0.5

0.0 0.0 0 8 16 24 0 8 16 24 Hours post-dexamethasone Hours post-dexamethasone

Figure 5.6. Quantification of BIM mRNA and protein induction following dexamethasone. Mice were treated with dexamethasone and cells harvested at baseline (0 hours) and at 8 and 24 hours following treatment. mRNA induction at 8 hours indicated by dotted lines and compared to 0 hours; protein induction indicated by solid lines. Each protein sample was normalised to actin, and induction expressed relative to 0 hours. Left panels PPRs, right panels PGRs. Error bars represent mean ± SD.

Page | 170

5.4 Connectivity Map results As the above comparisons had not yielded any statistically significant differentially expressed genes, I decided to perform more focused analyses. The cluster dendrogram (Figure 5.1) demonstrated that the dexamethasone- induced profiles of ALL-28 and ALL-57 clustered separately from the other 8 xenografts, suggesting that in these two glucocorticoid-resistant xenografts dexamethasone may be activating distinctive genes and pathways. A second analysis that could reveal important mechanisms was a comparison of the dexamethasone-induced profiles of the glucocorticoid-resistant Ph+ ALL-55 with the (albeit in vitro) glucocorticoid-sensitive Ph+ ALL-56.

A list of differentially expressed genes was generated by limma comparing the dexamethasone-induced profiles of ALL-28 and ALL-57 with all 5 PGRs, and after taking the genes with FDR <0.05 I generated a signature associated with glucocorticoid resistance consisting of 427 upregulated and 250 downregulated genes. A second list of genes was generated by limma comparing the dexamethasone-induced profiles of ALL-55 and ALL-56, and after taking the genes with FDR <0.0001 I generated a signature consisting of 297 upregulated and 373 downregulated genes.

A third signature was generated using historical Affymetrix data of our laboratory’s original xenograft panel. In this analysis, performed by Dr Geoff Neale at St Judes Children’s Research Hospital through PPTP, each xenograft (untreated) had been profiled and the Pearson correlation of each gene with the in vivo LGD to dexamethasone determined. Using this correlation data, I generated a signature associated with glucocorticoid resistance consisting of 375 upregulated and 398 downregulated genes (P <0.04). The statistical thresholds were chosen to generate approximately 300-500 genes in each list suitable for analysis in the Connectivity Map (CMap).

Page | 171

For each list, the Illumina IDs were converted to Affymetrix U133A IDs, and the signatures used to interrogate CMap. CMap is a collection of genome-wide transcriptional gene expression data from cultured human cancer cell lines treated for 6-12 hours with bioactive drugs and small molecules. Interrogation with custom signatures and simple pattern-matching algorithms enables the discovery of functional connections between drugs, genes and diseases through the transitory feature of common early gene-expression changes (Lamb et al. 2006).

The top ranked CMap agents from the 3 signatures are shown in Table 5.3, Table 5.4 and Table 5.5. Most striking was the recurrent appearance of the HSP90 inhibitor geldanamycin and its derivatives tanespimycin (17-AAG) and alvespimycin (17-DMAG). All 3 drugs had highly significant negative enrichment scores, indicating that these agents induce complementary gene expression profiles in the CMap experimental model and thus could potentially reverse glucocorticoid resistance. The only other agent to appear in more than one result was doxylamine, a sedating antihistamine with no known cytotoxic effect.

Page | 172

Table 5.3. CMap results comparing ALL-28 and ALL-57 with PGRs.

Rank CMap Name ES P 1 GW-8510 0.945 0 2 alvespimycin -0.64 0 3 doxylamine -0.87 8E-05 4 geldanamycin -0.55 0.0001 5 NS-398 -0.93 0.0005 6 irinotecan 0.933 0.0005 7 econazole 0.85 0.0007 8 CP-944629 0.846 0.0008 9 15-delta PG J2 -0.48 0.001 10 Prestwick-1082 -0.91 0.0013

ES, enrichment score. HSP90 inhibitors highlighted.

Table 5.4. CMap results comparing ALL-55 with ALL-56.

Rank CMap Name ES P 1 phenoxybenzamine -0.96 0 2 helveticoside -0.95 0 3 astemizole -0.93 0 4 mefloquine -0.92 0 5 geldanamycin -0.59 0 6 tanespimycin -0.54 0 7 disulfiram -0.89 6E-05 8 8-azaguanine -0.92 0.0001 9 alvespimycin -0.57 0.0004 10 digoxin -0.88 0.0005

ES, enrichment score. HSP90 inhibitors highlighted.

Page | 173

Table 5.5. CMap results of original xenograft panel.

Rank CMap Name ES P 1 geldanamycin -0.655 0 2 tanespimycin -0.461 0 3 CP-690334-01 -0.679 0.0004 4 doxylamine -0.81 0.0006 5 camptothecin 0.924 0.0007 6 lorglumide 0.785 0.0008 7 mitoxantrone 0.904 0.0017 8 cinchonine -0.797 0.0033 9 tiapride -0.709 0.0049 10 nisoxetine 0.77 0.0057 11 withaferin A 0.768 0.0058 12 naftifine -0.764 0.0059 13 buflomedil 0.766 0.0061 14 trihexyphenidyl -0.851 0.0062 15 calycanthine 0.765 0.0063 16 alvespimycin -0.464 0.0075

ES, enrichment score. HSP90 inhibitors highlighted.

Page | 174

After reviewing the literature and the CMap results, I chose to proceed with testing 17-DMAG as a potential glucocorticoid-resistance reversing agent. The reasons for this were: (1) it appeared in the top ranked drugs from all 3 signatures; (2) it was water soluble and had the most favourable toxicity profile; (3) it was readily available and (4) it had already been shown to be safe to use in the xenograft mouse model (Smith et al. 2008). Although 17-DMAG had shown little activity as a single agent in preclinical in vivo models of ALL (Smith et al. 2008), the aim of future experiments would be to use this drug to sensitise resistant cells to dexamethasone, and so its limited single agent activity was not a major issue. The CMap database build 02 was established in 2008 prior to the development of non-geldanamycin derived HSP90 inhibitors, and although one of these novel agents could be potentially more potent, I had no experimental data to direct a choice or research collaboration to enable drug access to evaluate one or more of these compounds.

Of particular relevance to my study, HSP90 is a co-chaperone molecule of the glucocorticoid receptor (GR), and maintains the cytosolic GR in a state capable of binding ligand. It is also responsible for binding immunophilins such as FKBP52 that attach the GR complex to the dynein motor protein trafficking pathway that translocates the activated GR to the nucleus (Harrell et al. 2004). A number of studies, however, have reported that glucocorticoid resistance in ALL is not related to HSP90 levels (Section 1.8.2). In my study, there was no difference in HSP90 mRNA levels at baseline or on exposure to dexamethasone when comparing the PPR and PGR groups (Figure 5.7).

Page | 175

13 p=1.000 p=1.000

11 mRNA expression mRNA

PGR PPR

PGR+Dex PPR+Dex

Figure 5.7. mRNA expression of HSP90. Normalised mRNA expression levels expressed as the mean of 3 probes at baseline (PGR/PPR) and 8 hours after in vivo treatment with dexamethasone (PGR+Dex/PPR+Dex). Error bars represent mean ± SEM.

Page | 176

5.5 Comparison with in vitro gene expression I have previously demonstrated that glucocorticoid-induced gene expression is model dependent, with a number of differences between xenografts, in vivo patient data and in vitro cell line data (Section 4.7). I have also shown that the in vitro and in vivo dexamethasone responses in ALL-56 at secondary passage are not concordant (Section 3.6). To investigate this, I analysed gene expression of the in vitro response of ALL-55 and ALL-56 after 8 hours exposure to 1 µM dexamethasone, and compared the results to the in vivo response. The experiment was setup in triplicate, with 3 controls and 3 dexamethasone treated samples for each xenograft. The microarray data was obtained using Illumina HT-12_V4 BeadChips, the raw data was pre-processed as previously, and differentially expressed genes determined using limma.

In ALL-55, there was an approximate two fold increase in the number of significantly differentially expressed probes in vitro compared to in vivo, whereas for ALL-56 the numbers were very similar (Table 5.6). When the significant genes were compared, in ALL-55 the majority of genes expressed in vivo were also seen in vitro, whereas in ALL-56 the two profiles were quite different (Figure 5.8). An analysis of Gene Ontology revealed that for ALL-55, pathways associated with apoptosis and protein kinase signalling networks were upregulated both in vivo and in vitro, whereas downregulated pathways included those related to apoptosis in vitro, with few pathways in vivo (Appendix C). This is evidence that glucocorticoids exert competing effects on cells, with both positive and negative regulation of apoptotic pathways.

Page | 177

Table 5.6. Number of differentially expressed probes in vivo vs in vitro in ALL-55 and ALL-56 on exposure to dexamethasone.

P value False Discovery Rate (FDR)

<0.05 <0.001 <0.0001 <0.25 <0.1 <0.05 ALL-55 in vivo 6183 1467 728 4709 2646 1839 ALL-55 in vitro 8866 3228 2118 10033 6201 4798

ALL-56 in vivo 8092 3003 1755 7926 5526 4479 ALL-56 in vitro 7488 2327 1421 7395 4451 3358

Page | 178

Figure 5.8. Comparison of dexamethasone responses in ALL-55 and ALL- 56. In vivo and in vitro dexamethasone-induced gene expression profiles were compared using genes significantly differentially expressed with FDR <0.0001 in ALL-55 (A, B) and ALL-56 (C,D).

Page | 179

For ALL-56, the upregulated responses were similar to ALL-55, with apoptotic and protein kinase signalling pathways activated in both models. However an analysis of the in vitro downregulated pathways revealed a number related to the positive and negative regulation of apoptosis, but in vivo only pathways positively related to apoptosis were downregulated (Appendix C). This suggests that in vivo the competing balance favours anti-apoptotic pathways, and is a potential explanation of the discordant vivo/vitro responses seen in this xenograft.

I generated a list of differentially expressed genes comparing the in vitro dexamethasone-induce profiles of ALL-55 and ALL-56 and created a signature of resistance as in Section 5.4 to interrogate CMap. Interestingly, tanespimycin (17-AAG) was the only HSP90 inhibitor identified in the list of top ranked agents (Table 5.7), and other than this there were only four agents (helveticoside and the cardiac glycosides ouabain, lanatoside C and digoxin) common to both the in vivo and in vitro CMap lists (Table 5.4 and Table 5.7). This is further evidence to suggest that cells have distinct mechanisms of drug response when continuously exposed to drug in vitro versus drug pharmacokinetic and pharmacodynamic changes in vivo.

Page | 180

Table 5.7. CMap results comparing in vitro ALL-55 with ALL-56.

Rank CMap Name ES P 1 anisomycin -0.986 0 2 helveticoside -0.984 0 3 ouabain -0.981 0 4 digitoxigenin -0.979 0 5 H-7 0.966 0 6 lanatoside C -0.96 0 7 digoxigenin -0.951 0 8 digoxin -0.944 0 9 cephaeline -0.928 0 10 wortmannin -0.799 0 11 sirolimus -0.752 0 12 prochlorperazine -0.72 0 13 trifluoperazine -0.717 0 14 thioridazine -0.704 0 15 LY-294002 -0.639 0 16 tanespimycin -0.443 0

ES, enrichment score. HSP90 inhibitors highlighted.

Page | 181

5.6 Evaluation of the HSP90 inhibitor 17-DMAG 17-DMAG is a dark purple powder that was purchased from Sigma and LC Laboratories and stored at -20oC. For in vitro experiments, the drug was made up in 100% DMSO and stored in aliquots at -20oC. For in vivo work, the drug was made up daily in a sodium citrate/citric acid buffered pH 3.2 vehicle previously optimised by PPTP.

5.6.1 In vitro evaluation The efficacy of 17-DMAG as a single agent and in combination with dexamethasone in glucocorticoid-resistant xenografts was assessed by MTT assay. Two different experimental setups were used. In the first set of experiments, cells were treated with 17-DMAG as a single agent and in combination with a fixed 1 µM dose of dexamethasone. In the second set of experiments, cells were treated with the both drugs at fixed IC50 ratios and combination indices determined using CalcuSyn (Biosoft, UK). The first experiments were performed in the newly established glucocorticoid-resistant xenografts ALL-28, ALL-50, ALL-55 and ALL-57, plus 2 existing glucocorticoid- resistant xenografts ALL-2 and ALL-19 (Bachmann et al. 2005), and 2 glucocorticoid-resistant ALL cell lines, Jurkat and REH. The second experiments were done in ALL-50, ALL-55 and ALL-57.

Anti-tumour activity was seen in all xenografts and cell lines using 17-DMAG as a single agent (Figure 5.9 and Figure 5.10) with IC50 values in the nanomolar range, except for ALL-2 which required significantly higher concentrations (Table 5.8). No appreciable difference was seen when 17-DMAG was combined with 1 µM dexamethasone. In the second set of experiments, there was evidence of synergism, defined as a combination index (CI) <1, in all three xenografts (Figure 5.11 and Table 5.9). In ALL-50, synergistic CIs were seen at

4 out of 5 experimental values and also at the derived ED50; in ALL-55 and ALL- 57, synergistic CIs were seen at 2 out of 5 experimental values but not at the derived values. These experiments confirm that in vitro 17-DMAG demonstrates anti-leukaemic cytotoxicity and shows a degree of synergism in combination with dexamethasone in a subset of glucocorticoid-resistant xenografts.

Page | 182

To investigate whether demonstrated cytotoxicity was due to HSP90 inhibition or off-target effects, I treated ALL-55 and ALL-57 cells with 50 nM or 500 nM 17-DMAG in vitro, and harvested the cells after incubation for 8 hours or 24 hours. I then analysed changes in HSP70 and AKT protein levels, relevant pharmacodynamic endpoints in monitoring HSP90 inhibition (Johnson et al. 2007), compared to untreated controls. The results demonstrated a dose- and time-dependent increase in HSP70 and suppression of AKT in ALL-55, with similar though not as pronounced results in ALL-57 (Figure 5.12 and Figure 5.13), confirming that 17-DMAG was hitting its target.

Page | 183

ALL-28 A B ALL-50 120 17-DMAG IC 794nM 50 120 17-DMAG IC 21nM 17-DMAG + 1M Dex 50 100 100 17-DMAG + 1M Dex

80 80

60 60

40 40

Viability (% control) Viability 20

0 0 -12 -11 -10 -9 -8 -7 -6 -5 -12 -11 -10 -9 -8 -7 -6 -5 log [17-DMAG] log [17-DMAG]

C ALL-55 D ALL-57 120 120 17-DMAG IC50 65nM 17-DMAG IC50 26nM 100 17-DMAG + 1M Dex 100 17-DMAG + 1M Dex

80 80

60 60

40 40

Viability (% control) Viability 20 (% control) Viability 20

0 0 -12 -11 -10 -9 -8 -7 -6 -5 -12 -11 -10 -9 -8 -7 -6 -5 log [17-DMAG] log [17-DMAG]

Figure 5.9. In vitro assessment of 17-DMAG in new glucocorticoid- resistant xenografts. Secondary passage cells of (A) ALL-28, (B) ALL-50, (C) ALL-55 and (D) ALL-57 were treated with increasing concentrations of 17- DMAG ± 1 µM dexamethasone, and viability assessed by MTT assay after 48 hours incubation. Values expressed as a percentage of untreated control. Each data point represents the mean ± SEM from 3 independent experiments. Dex, dexamethasone.

Page | 184

A ALL-2 B ALL-19 120 120 17-DMAG IC 50 5M 17-DMAG IC 50 465nM 100 17-DMAG + 1M Dex 100 17-DMAG + 1 M Dex

80 80

60 60

40 40 Viability (% control) Viability 20 (% control) Viability 20

0 0 -12 -11 -10 -9 -8 -7 -6 -5 -12 -11 -10 -9 -8 -7 -6 -5 log [17-DMAG] log [17-DMAG]

C Jurkat D REH 120 120 17-DMAG IC 50 607nM 17-DMAG IC 50 47nM 100 17-DMAG + 1M Dex 100 17-DMAG + 1 M Dex

80 80

60 60

40 40 Viability (% control) Viability 20 (% control) Viability 20

0 0 -12 -11 -10 -9 -8 -7 -6 -5 -12 -11 -10 -9 -8 -7 -6 -5 log [17-DMAG] log [17-DMAG]

Figure 5.10. In vitro assessment of 17-DMAG in existing glucocorticoid- resistant xenografts and cell lines. Quaternary passage cells of (A) ALL-2 and (B) ALL-19, and cultured cells of (C) Jurkat and (D) REH were treated with increasing concentrations of 17-DMAG ± 1 µM dexamethasone, and viability assessed by MTT assay after 48 hours incubation. Values expressed as a percentage of untreated control. Each data point represents the mean ± SEM from 3 independent experiments. Dex, dexamethasone.

Page | 185

Table 5.8. Summary of 17-DMAG in vitro responses.

Xenograft IC50 ALL-28 794 nM ALL-50 21 nM ALL-55 65 nM ALL-57 26 nM ALL-2 5 µM ALL-19 465 nM

Cell line IC50 Jurkat 607 nM REH 47 nM

Page | 186

A ALL-50 120

100 Dex 80 17-DMAG

60 Dex + 17-DMAG

Viability (% control) Viability 20

0 0.25 0.5 1 2 4

Ratio of IC50

ALL-55 B 120

100 Dex 80 17-DMAG

60 Dex + 17-DMAG

Viability (% control) Viability 20

0 0.25 0.5 1 2 4

Ratio of IC50

C ALL-57 120

100 Dex 80 17-DMAG

60 Dex + 17-DMAG

Viability (% control) Viability 20

0 0.25 0.5 1 2 4

Ratio of IC50

Figure 5.11. In vitro assessment of 17-DMAG ± dexamethasone at fixed

IC50 ratios. Secondary passage cells of (A) ALL-50, (B) ALL-55 and (C) ALL-57 were treated with 17-DMAG ± dexamethasone at fixed IC50 ratios, and viability assessed by MTT assay after 48 hours incubation. Values expressed as a percentage of untreated control. Each data point represents the mean ± SEM from 3 independent experiments. Dex, dexamethasone.

Page | 187

Table 5.9. Combination indices of 17-DMAG ± dexamethasone in glucocorticoid-resistant xenografts.

Derived Values Experimental Values (Ratio IC50) Xenograft ED50 ED75 ED90 0.25 0.5 1 2 4 ALL-50 0.71 1.24 2.17 0.21 0.34 0.54 0.95 1.51 ALL-55 1.03 1.02 1.02 3.82 0.48 0.72 1.01 1.52 ALL-57 1.15 1.02 1.00 10.41 3.08 0.83 0.90 1.23

Page | 188

Figure 5.12. HSP70 and AKT in ALL-55 and ALL-57 following 17-DMAG in vitro. Representative immunoblot from cultured xenograft cells at control (con), 8 or 24 hours following treatment with 50 nM or 500 nM 17-DMAG. Jkt, Jurkat positive control. ACT, actin.

Page | 189

B ALL-55 - AKT A ALL-55 - HSP70 30 40

30 20

10 expression normalised normalised expression normalised

0 0

50nM 8h 50nM 8h 50nM 24h 500nM 8h 50nM 24h 500nM 8h 0h control 500nM 24h 0h control 500nM 24h

C ALL-57- HSP70 D ALL-57 - AKT 40 40

30 30

20 20

10 10 normalised expression normalised expression normalised

0 0

50nM 8h 50nM 8h 50nM 24h 500nM 8h 50nM 24h 500nM 8h 0h control 500nM 24h 0h control 500nM 24h

Figure 5.13. Quantification of HSP70 and AKT in ALL-55 and ALL-57 post- 17-DMAG in vitro. Xenograft cells were treated with 50 nM or 500 nM 17- DMAG and cells harvested at 0, 8 or 24 hours following treatment. ALL-55 values for HSP70 (A) and AKT (B); ALL-57 values for HSP70 (C) and AKT (D). Expression values normalised to actin. Error bars represent mean ± SD.

Page | 190

5.6.2 In vivo evaluation When assessing drugs in vivo, a toxicity study must first be performed to determine the maximum tolerated dose (MTD). Unengrafted mice are treated with the drug(s) as single agents and in combination at fixed or attenuated doses for 2 weeks. The mice are monitored daily for general wellbeing, and weight is monitored throughout the experiment and for up to 2 weeks after cessation of drug(s) to assess for delayed effects. In the second week, one mouse from any group showing morbidity is bled and routine haematology and biochemistry assessed, repeated in the third week if abnormalities are detected. At the end of the experiment, the mice are culled and the internal organs assessed. Any weight loss ≥20 or other significant morbidity is evidence that the MTD has been exceeded. Care must be taken in determining the dose to take forward for efficacy testing, as leukaemia-bearing mice have a lower physiological tolerance to drugs as a result of disseminated disease.

When PPTP evaluated 17-DMAG in 2005, the recommended dose schedule was 50 mg/kg BID (twice daily) on 2 days per week, but the toxicity study revealed that NOD/SCID mice could only tolerate a 50% reduced dose of 25 mg/kg BID on 2 days per week. One of the main issues was the intolerance to the acidic vehicle, as within a minute of injection of 17-DMAG or vehicle alone, the mice would become very still, extend their bodies with an exaggerated lordosis and look very unwell. After close observation for 5-10 minutes, the mice would make a full recovery with no apparent residual ongoing morbidity.

I performed 2 toxicity experiments. In the first study I used a fixed dose of 17- DMAG (25 mg/kg BID twice a week) with daily Monday-Friday doses of dexamethasone at 15 mg/kg (100%), 11.25 mg/kg (75%), 7.5 mg/kg (50%) and 3.75 mg/kg (25%). Other than the expected temporary toxicity seen immediately following injection, there was no other evidence of morbidity and no significant weight loss in any group (Figure 5.14). There were no haematological or biochemical abnormalities detected, and in particular there was no elevation of hepatic enzymes (Table 5.10). On autopsy, all internal organs had a healthy appearance. Page | 191

Dex 15mg/kg 5 17-DMAG + Dex 15mg/kg (100%) 17-DMAG + Dex 11.25mg/kg (75%) 0 17-DMAG + Dex 7.5mg/kg (50%) 17-DMAG + Dex 3.75mg/kg (25%) -5 17-DMAG % Weight change Control -10 0 7 14 21 Days following treatment initation

Figure 5.14. Toxicity study I of 17-DMAG ± dexamethasone. NOD/SCID mice were treated with 17-DMAG 25 mg/kg BD twice a week ± dexamethasone, or vehicle control. Weight change calculated relative to day 0. Solid black bars represent treatment period. Error bars represent mean ± SD from 4 individual mice. Dex, dexamethasone.

Page | 192

Table 5.10. Toxicity studies - haematology and biochemistry results.

Toxicity Study I Toxicity Study II

Control Treated Control Treated RBC (x106/µl) 9.9 9.3 10.8 11.3 HGB (g/L) 163 166 172 184 PLT (x103/µl) 899 843 1060 997 WBC (x103/µl) 8.9 4.9 6.9 1.9 ALB (g/L) 40 41 45 42 ALP (U/L) 40 108 51 56 ALT (U/L) 28 26 42 52 AMY (U/L) 992 983 1693 1781 TBIL (µmol/L) 5 6 6 7 BUN (mmol/L) 7.8 6.9 10.2 11.7 CA (mmol/L) 2.50 2.40 2.74 2.53 PHOS (mmol/L) 1.93 1.74 1.59 1.60 CRE (µmol/L) <18 18 21 <18 GLU (mmol/L) 6.4 7.7 7.9 8.7 Na (mmol/L) 154 149 149 148 K (mmol/L) 6.2 6.0 3.0 5.9 TP (g/L) 60 58 62 58

Page | 193

A major difficulty in comparing reports using HSP90 inhibitors has been the great variety of doses and schedules used both in preclinical and clinical studies. I identified HSP90 inhibitors from CMap using gene expression data generated 8 hours following treatment with dexamethasone, and the CMap database consists of HSP90 inhibitor data generated after 6 hours incubation with the drug. I thus postulated that the optimal schedule would be to treat with both drugs concomitantly, to achieve simultaneous dexamethasone-induced gene expression with 17-DMAG-induced gene expression to reverse resistance.

I thus performed a second toxicity study in which I used attenuated daily Monday-Friday doses of 17-DMAG (50 mg/kg, 30 mg/kg, 20 mg/kg, 15 mg/kg, and 10 mg/kg) ± fixed MTD of dexamethasone (15 mg/kg). Again, other than the expected temporary toxicity seen immediately following injection, there was no other evidence of morbidity. The highest dose group demonstrated weight loss approaching 10%, but there was no significant weight loss in any other group (Figure 5.15). There were no haematological abnormalities detected, and as before there was no elevation of hepatic enzymes (Table 5.10). The observed increase in asymptomatic amylase levels was attributed to the fact that the mice were bled 2 hours after the daily IP injection. On autopsy, all internal organs had a healthy appearance.

The results of the toxicity experiments were both reassuring and concerning, as no expected toxicities had been seen potentially indicating a lack of drug activity. In a review of previous reports of the use of 17-DMAG in preclinical mouse xenograft studies revealed that several different doses and schedules had been used. In a SCID mouse xenograft model of chronic lymphocytic leukaemia (CLL), 17-DMAG was used at a dose of 10 mg/kg IP daily for 5 days per week, although the rationale behind this choice of dose was not detailed (Hertlein et al. 2010). In a study using nude mice bearing solid tumour xenografts, oral 17- DMAG doses of 15 mg/kg or higher BID for 5 days per week were associated with significant toxicity and mortality, but the mice could tolerate 15 mg/kg daily for 5 days per week orally or 15 mg/kg IV daily for 3 days per week (Hollingshead et al. 2005). Page | 194

10 17-DMAG 50mg/kg 5 17-DMAG 50mg/kg + Dex 15mg/kg 17-DMAG 30mg/kg 17-DMAG 30mg/kg + Dex 15mg/kg 0 17-DMAG 20mg/kg 17-DMAG 20mg/kg + Dex 15mg/kg -5

% Weight change Dex 15mg/kg Control -10 0 7 14 21 Days following treatment initation

Figure 5.15. Toxicity study II of 17-DMAG ± dexamethasone. NOD/SCID mice were treated with daily Monday-Friday 17-DMAG ± dexamethasone, or vehicle control. Weight change calculated relative to day 0. Solid black bars represent treatment period. Error bars represent mean ± SD from 4 individual mice. Results from the lowest 17-DMAG doses (15 mg/kg and 10 mg/kg ± dexamethasone) not shown. Dex, dexamethasone.

Page | 195

To determine whether this particular batch and formulation of 17-DMAG was active, I evaluated the drug in an in vivo pharmacodynamic study. NOD/SCID mice highly engrafted with ALL-2 (Liem et al. 2004) were treated with a single dose of 17-DMAG at 30 mg/kg or 50 mg/kg and xenograft cells harvested from spleens at 0 (control), 8 or 24 hours after drug. I prepared protein lysates and analysed changes in HSP70 and AKT protein expression. The immunoblots suggest a dose- and time-dependent increase in HSP70 levels compared to controls, and a reduction in AKT levels at all doses and timepoints (Figure 5.16 and Figure 5.17), confirming that in this xenograft 17-DMAG demonstrated in vivo pharmacodynamic activity.

Page | 196

Figure 5.16. Regulation of HSP70 and AKT by 17-DMAG in ALL-2. Representative immunoblot of ALL-2 cells harvested from mice at control, 8 or 24 hours following treatment with 17-DMAG 30 mg/kg IP or 50 mg/kg IP. K562, positive control. ACT, actin.

Page | 197

A ALL-2 - HSP70

200

150

100

50 normalised expression normalised 0

K562 control

30 mg/kg 8 h 50 mg/kg 8 h 30 mg/kg 24 h 50 mg/kg 24 h

B ALL-2 - AKT 200

150

100

50 normalised expression normalised 0

K562 control

30 mg/kg 8 h 50 mg/kg 8 h 30 mg/kg 24 h 50 mg/kg 24 h

Figure 5.17. Quantification of HSP70 and AKT in ALL-2 following 17-DMAG. Expression values of (A) HSP70 and (B) AKT protein levels normalised to K562 cells. Error bars represent mean ± SD of 2 immunoblots.

Page | 198

I was thus satisfied that the drug was active and that unengrafted mice could tolerate 17-DMAG at a dose of 50 mg/kg, and decided to use this dose in an in vivo combination study with dexamethasone in the glucocorticoid-resistant PPR xenografts ALL-55 and ALL-57. For each xenograft, when huCD45+ reached 1%, mice were randomised into groups of 8 and treated with vehicle control (pH 3.2), single agent dexamethasone 15 mg/kg IP daily Monday–Friday, single agent 17-DMAG 50 mg/kg IP daily Monday-Friday, or combination with both drugs given concomitantly IP daily Monday-Friday.

It was clear within the first week of treatment that the 17-DMAG 50 mg/kg dose was toxic in engrafted mice, with a number of mice experiencing unacceptable toxicity in the 17-DMAG single agent group at low levels of engraftment. In particular, I noted that these mice had diarrhoea and lost weight rapidly, but had no other abnormalities noted on autopsy. Interestingly, the mice treated in combination appeared to be protected by dexamethasone, with less morbidity seen. Thus the 17-DMAG dose was attenuated to 30 mg/kg for weeks 2-4 of treatment in ALL-55, and for weeks 3-4 in ALL-57, and all mice receiving 17- DMAG as either single agent or in combination were given soggy food with additional subcutaneous fluid boluses if required.

As before, mice were bled weekly and an event was defined a priori as huCD45 ≥ 25 . However in several ALL-57 mice, engraftment levels would be low at the time of the weekly bleed, but the mice would experience significant morbidity 2-3 days later with very large highly engrafted spleens and were therefore classified as having reached event at the time of euthanasia. The engraftment and survival curves are shown in Figure 5.18 and Figure 5.19.

With ALL-55, both single agents and combination produced statistically significant LGDs compared to control, but none produced an objective response (Table 5.11). In fact the combination therapy was antagonistic when compared to single agent dexamethasone. With ALL-57, again both single agents and combination produced statistically significant LGDs compared to control, and this xenograft also demonstrated an objective response to 17-DMAG. However Page | 199

the combination therapy did not demonstrate therapeutic enhancement, defined as a significant improvement in EFS between the combination and each single agent separately (Table 5.11). It is possible that 17-DMAG causes downregulation of the GR in vivo which could explain the antagonism seen with dexamethasone in ALL-55.

To further investigate this result, I performed an in vivo pharmacodynamic analysis of 17-DMAG in these two xenografts. Mice highly engrafted with ALL- 55 were treated with 17-DMAG 30 mg/kg and cells harvested at 0, 8 or 24 hours after drug. Mice highly engrafted with ALL-57 were treated with 30 mg/kg or 50 mg/kg and cells harvested at 0, 4-6 or 24 hours after drug. In this second experiment, 2 mice were euthanised earlier than the planned 8 hour timepoint due to significant morbidity from the high leukaemic burden at the time of treatment. Analysis of pharmacodynamic markers revealed a subtle time- and dose-dependent increase in HSP70 levels more pronounced in ALL-55, but no significant change in AKT levels in either xenograft (Figure 5.20 and Figure 5.21). Thus in these 2 xenografts, 17-DMAG does not induce HSP90-related pharmacodynamic changes in vivo in as predictable or efficacious manner as ALL-2 in vivo or as ALL-55 or ALL-57 in vitro. Interestingly, it is also apparent that pharmacodynamic efficacy is not related to in vivo cytotoxicity as ALL-57 did show a response to single agent 17-DMAG, a fact that has been previously reported (Smith et al. 2008).

5.6.3 Summary The gene expression and CMap-derived list of potential glucocorticoid- resistance reversing agents strongly pointed towards HSP90 inhibitors. Evaluation of the HSP90 inhibitor 17-DMAG demonstrated a degree of synergy with dexamethasone in glucocorticoid-resistant xenografts in vitro, but no synergy was seen in vivo, potentially due to the toxicity of the 17-DMAG vehicle.

Page | 200

A 60

50 Dex + 40 17-DMAG 30 17-DMAG+Dex Control

%huCD45 20

-7 0 7 14 21 28 35 42 49 56 63 70 77 84

Days post-randomisation

B 100

80 Dex 60 17-DMAG 17-DMAG+Dex 40 Control 20 Survival probabililty (%) probabililty Survival 0 0 7 14 21 28 35 42 49 56 Days post randomisation

Figure 5.18. Efficacy of 17-DMAG ± dexamethasone in ALL-55. Engraftment (A) and Kaplan-Meier curves of event-free survival (B) of ALL-55 treated with vehicle control, dexamethasone 15 mg/kg IP Mon-Fri, 17-DMAG 30-50 mg/kg IP Mon-Fri or combination. Non-event censored mice shown by ticks. Solid black bar represents treatment period. Dex, dexamethasone.

Page | 201

A 100 90 80 Dex + 70 60 17-DMAG 50 17-DMAG+Dex 40 Control %huCD45 30 20 10

-7 0 7 14 21 28 35 42

Days post-randomisation

B 100

80 Dex 60 17-DMAG 17-DMAG+Dex 40 Control 20 Survival probabililty (%) probabililty Survival 0 0 7 14 21 28 35 42 Days post randomisation

Figure 5.19. Efficacy of 17-DMAG ± dexamethasone in ALL-57. Engraftment (A) and Kaplan-Meier curves of event-free survival (B) of ALL-57 treated with vehicle control, dexamethasone 15 mg/kg IP Mon-Fri, 17-DMAG 30-50 mg/kg IP Mon-Fri or combination. Non-event censored mice shown by ticks. Solid black bar represents treatment period. Dex, dexamethasone.

Page | 202

Table 5.11. Leukaemia Growth Delays and Objective Response Measures of ALL-55 and ALL-57 treated with 17-DMAG ± dexamethasone.

17-DMAG Dex Comb P (Comb v P (Comb TE LGD ORM LGD ORM LGD ORM 17-DMAG) v Dex) x55 16.1 PD2 30.6 SD 20.8 PD2 0.109 0.005 NO x57 20.0 CR 6.8 PD1 19.0 CR 0.809 0.001 NO

LGD, leukaemia growth delay (days); ORM, objective response measure; Dex, dexamethasone; Comb, combination (17-DMAG + dexamethasone) and TE, therapeutic enhancement. All LGDs were statistically significant compared to controls. Combination v Dex for ALL-55 was significantly antagonistic.

Page | 203

Figure 5.20. HSP70 and AKT in ALL-55 and ALL-57 following 17-DMAG in vivo. Representative immunoblots from ALL-55 (A) or ALL-57 (B) cells harvested from mice at control, 4-8 and 24 hours following treatment with 17- DMAG at 30 mg/kg or 50 mg/kg.

Page | 204

A ALL-55 - HSP70 B ALL-55 - AKT 3 6

2 4

1 2 normalised expression normalised expression normalised 0 0

control control

30 mg/kg 8 hr 30 mg/kg 8 hr 30 mg/kg 24 hr 30 mg/kg 24 hr

C ALL-57 - HSP70 D ALL-57 - AKT 1.0 8 0.8 6 0.6 4 0.4

0.2 2 normalised expression normalised expression normalised 0.0 0

control control

30 mg/kg 6 hr 50 mg/kg 4 hr 30 mg/kg 6 hr 50 mg/kg 4 hr 30 mg/kg 24 hr 50 mg/kg 24 hr 30 mg/kg 24 hr 50 mg/kg 24 hr

Figure 5.21. Quantification of HSP70 and AKT in ALL-55 and ALL-57 post- 17-DMAG in vivo. Mice engrafted with ALL-55 or ALL-57 were treated with 17- DMAG at 30 mg/kg or 50 mg/kg and cells harvested at control, 4-8 or 24 hours thereafter. ALL-55 values for HSP70 (A) and AKT (B); ALL-57 values for HSP70 (C) and AKT (D). Expression values normalised to actin. Error bars represent mean ± SD of 2 immunoblots.

Page | 205

5.7 Evaluation of a BIM-BH3 stapled peptide BIM is a BH3-only pro-apoptotic member of the BCL2 family of proteins, and has a critical role in glucocorticoid-evoked apoptosis (Section 1.8.3). A small molecule replicated from the amphipathic α-helical BIM-BH3 sequence could potentially act as a BIM-mimetic, and induce apoptosis in leukaemic and other cancer cells. However biologically active helical motifs within proteins typically have little structure when taken out of context and placed in solution, and the in vivo efficacy of these peptides is compromised by loss of secondary structure, susceptibility to proteolytic degradation and inability to penetrate intact cells (Walensky et al. 2004).

To overcome these limitations, Walensky et al developed a method using hydrocarbon ‘staples’ to maintain the conformation of the α-helix, termed stabilised alpha-helix of BCL2 domains (SAHBs). They generated a BID-SAHB, and demonstrated in vitro and in vivo cytotoxicity against a number of leukaemic cell lines (Walensky et al. 2004). The group subsequently developed a BIM- SAHB (Gavathiotis et al. 2008) and the method of hydrocarbon stapling was further promoted by Aileron Therapeutics (Wolfson 2009). Attempts to collaborate with Aileron to evaluate the BIM-SAHB in glucocorticoid-resistant ALL were unsuccessful, and I therefore entered into a collaboration with Dr Kate Jolliffe, a medicinal chemist, who synthesised a BIM-SAHB and BID-SAHB (Figure 5.22) according to published methods (Bird et al. 2008).

Figure 5.22. SAHB structures. Amino acid structures of the BIM-SAHB and BID-SAHB. Arc represents the hydrocarbon stapling.

Page | 206

The SAHBs were delivered as white powders, and a working 10 mM stock made by dissolving the powder in 100% DMSO for storage at -20oC. Initial solubility testing revealed significant peptide precipitation if the working stock was diluted with cell culture media, but no issues if dissolved in water, and no obvious precipitation was seen in culture media at concentrations used in in vitro cytotoxicity assays. I evaluated the BIM-SAHB in vitro by MTT assay against the glucocorticoid-sensitive xenograft ALL-3, the glucocorticoid-resistant xenograft ALL-19 (Bachmann et al. 2005), and both SAHBs against a panel of leukaemia cell lines reported to be sensitive to the BID-SAHB (Walensky et al. 2004). No cytotoxicity was seen with either SAHB in either xenografts or cell lines (Figure 5.23), and no glucocorticoid-sensitisation seen using the BIM- SAHB in combination with dexamethasone in ALL-19 (Figure 5.23A).

The reasons for the lack of efficacy were not immediately clear, as potential issues could be related to peptide synthesis, solubility and ability of the SAHBs to enter the target cell. I was not able to further investigate these molecules, but if they are commercially released in a bioavailable form in the future would certainly warrant evaluation in glucocorticoid-resistant ALL.

Page | 207

120 A ALL-3+Dex 100 ALL-3+BIM-SAHB 80 ALL-19+Dex ALL-19+BIM-SAHB 60 ALL-19+Dex 1M+BIM-SAHB 40 Jurkat+BIM-SAHB

20 Viability (% control) Viability

0 -12 -11 -10 -9 -8 -7 -6 -5 Log [dexamethasone] or [BIM-SAHB]

120 B 100 Jurkat 80 CEM K562 60 REH 40

Viability (% control) Viability 20 0 -12 -11 -10 -9 -8 -7 -6 -5 Log [BIM-SAHB]

C 120 100 Jurkat 80 CEM K562 60 REH 40

Viability (% control) Viability 20 0 -12 -11 -10 -9 -8 -7 -6 -5 Log [BID-SAHB]

Figure 5.23. Evaluation of SAHBs in vitro. Cells of xenografts (ALL-3/19) or leukaemic cell lines (Jurkat, CEM, K562, REH) were incubated for 48 hours with BIM-SAHB ± dexamethasone (A), BIM-SAHB (B) or BID-SAHB (C) and viability assessed by MTT assay. Error bars represent mean ± SEM of 2 independent experiments.

Page | 208

5.8 Discussion In this study, dexamethasone-induced gene expression identified HSP90 inhibitors as potential glucocorticoid-resistance reversing agents, although no synergy was seen in combination with dexamethasone in vivo. HSP90 is an essential molecular chaperone that is induced by cellular stresses, and serves as a buffer that regulates the activation and stability of a number of client proteins, many of which are involved in key pathways and processes in cancer (Figure 5.24). The cause of failure of therapies that selectively target a single receptor tyrosine kinase is attributed to activation of parallel pathways, and inhibition of HSP90 can lead to combinatorial inhibition of multiple signal transduction pathways which could overcome this problem (Workman et al. 2007). Thus HSP90 is a biologically attractive therapeutic target in cancer therapy.

Geldanamycin, a benzoquinone ansamycin antibiotic, was the first HSP90 inhibitor identified. Originally isolated from the organism Streptomyces hygroscopicus for its weak antibiotic activity (DeBoer et al. 1970), it was subsequently shown to have potent anti-tumour activity (Sasaki et al. 1979). However it was not until some years later that geldanamycin was shown to inhibit HSP90 (Whitesell et al. 1994). Preclinical assessment revealed significant hepatotoxicity (Supko et al. 1995), but it was discovered that modifications at position 17 resulted in a geldanamycin derivative, 17-N- allylamino-17-demethoxygeldanamycin (17-AAG, tanespimycin) with reduced hepatotoxicity (Schulte and Neckers 1998). Further modifications led to the development of 17-dimethylaminoethylamino-17-demethoxygeldanamycin (17- DMAG, alvespimycin), a water-soluble analogue of 17-AAG (Egorin et al. 2002).

In recent years, as the structure of HSP90 has been elucidated a number of novel synthetic small molecule inhibitors unrelated to geldanamycin have been developed with the aim of increasing potency of inhibition whilst minimising toxicity. These novel drugs include SNX7081 (Best et al. 2010), BJ-B11 (Ju et al. 2011), STA-9090 (Proia et al. 2011) and AUY922 (Weigert et al. 2012).

Page | 209

Figure 5.24. The role of HSP90 in the hallmarks of cancer. Client proteins shown on the right. From (Neckers and Workman 2012).

Page | 210

The agents 17-AAG and 17-DMAG, along with other HSP90 inhibitors, were assessed in a number of Phase I/II clinical trials both as single agents and in combination with established therapies from 2003-2007 (reviewed in (Neckers and Workman 2012). These were predominantly adult trials in advanced solid tumours, with the most impressive activity for 17-AAG seen in HER2+ trastuzumab-refractory breast cancer (Modi et al. 2007). This activity was attributed to the extreme sensitivity of HER2 as a HSP90 client protein. Although stable disease was attained in a number of Phase I trials in a variety of tumour types, no partial or complete responses were seen. This was attributed to suboptimal inhibition of client proteins from insufficient exposure to drug. Further, in preclinical testing 17-AAG was shown to be predominantly cytostatic with cell cycle arrest predominating over apoptosis (Hostein et al. 2001). A complete response to 17-DMAG was seen in castrate-resistant prostate cancer (Pacey et al. 2011), attributed to depletion of the androgen receptor, a HSP90 client protein (Solit et al. 2002). Stable disease was seen in other tumour types.

HSP90 inhibitors have also been studied in haematological malignancies, primarily acute myeloid leukaemia (AML), chronic myeloid leukaemia (CML) and chronic lymphocytic leukaemia (CLL). AML is frequently associated with activating mutations of the HSP90 client protein FMS-like tyrosine kinase 3 (FLT3). 17-AAG has demonstrated in vitro cytotoxicity to primary AML cells expressing mutant FLT3 and synergy with cytarabine (Al Shaer et al. 2008). Complete remissions have been obtained with 17-AAG in combination with cytarabine (Kaufmann et al. 2011) and 17-DMAG (Lancet et al. 2010) in Phase I trials of patients with advanced AML

Aggressive CLL is associated with expression of ZAP-70, a HSP90 client, and preclinical reports have demonstrated efficacy of 17-AAG in ZAP-70+ CLL cells (Castro et al. 2005; Best et al. 2010). It has also been reported that 17-DMAG can target the NF-κB family of proteins to induce apoptosis in CLL cells (Hertlein et al. 2010).

Page | 211

The majority of cases of CML and up to 25% of adult ALL and 3-5% of childhood ALL are associated with the t(9;22) Philadelphia (Ph+) translocation, which results in the oncogenic BCR-ABL fusion protein, a client of HSP90. Acquired resistance to tyrosine kinase inhibitors such as imatinib is a common clinical problem, and HSP90 inhibitors have demonstrated preclinical activity in mice bearing CML with the BCR-ABL-T315l mutant (Peng et al. 2007). Several studies have reported promising preclinical activity of other novel HSP90 inhibitors in CML (Lu et al. 2010; Ju et al. 2011) and Ph+ ALL (Tong et al. 2011).

Novel HSP90 inhibitors have also shown activity in cancer cells with activated JAK/STAT signalling. Janus-associated kinase 2 (JAK2) is a non-receptor tyrosine kinase and a HSP90 client protein. Activating JAK2 mutations have been described with high frequency in myeloproliferative disorders (Kralovics et al. 2005) and in a subset of high-risk BCP-ALL with rearrangements of cytokine receptor-like factor 2 (CRLF2) (Mullighan et al. 2009). Preclinical data of novel HSP90 inhibitors has demonstrated potent activity in a number of JAK2 mutant cancer cells with significant advantages over JAK-specific inhibitors (Proia et al. 2011; Weigert et al. 2012).

The Pediatric Preclinical Testing Program evaluated 17-DMAG as a single agent and reported that although activity was uniformly seen against the cancer cell lines of the in vitro panel, little activity was seen against the in vivo panel of paediatric cancer xenografts (Smith et al. 2008). In particular, although statistically significant growth delay was seen, there were no objective responses in the panel of ALL xenografts.

Thus, although the current study failed to show an ability of 17-DMAG to reverse glucocorticoid-resistance, HSP90 inhibitors have shown promise in a number of preclinical models, and next generation HSP90 inhibitors warrant evaluation in an extended panel of glucocorticoid-resistant xenografts. If successful, these novel agents could be introduced into upfront ‘windows’ in clinical trials of relapsed/refractory ALL, particularly if the patient was classified as a PPR as part of initial therapy. Page | 212

6 Conclusions and Future Directions

Page | 213

ALL is the most common cancer in childhood, and although cure rates now approach 85% (Pui and Evans 2006) a significant proportion of children will relapse, and subsequent survival, particularly in high-risk patients, is poor (Roy et al. 2005). Children classified on day 8 of treatment as a PPR have inferior outcomes (Riehm et al. 1987; Dordelmann et al. 1999; Schrappe et al. 2000) despite stratification to the high-risk arms of protocols and allogeneic transplantation. Early intervention to reverse glucocorticoid resistance could improve the chances of cure for this group of patients.

In Chapter 3, I used the xenograft mouse model to establish a panel of xenografts derived from patients classified as PPRs and matched PGRs. I showed that the xenografts demonstrated minimal changes in immunophenotype compared to the diagnostic patient samples, and that the xenograft gene expression profiles clustered with the corresponding patient samples. This is strong evidence that the xenografts are an accurate representation of the patients from which they were derived, although the paired xenograft expression profiles were enriched in genes associated with proliferation and survival, suggesting a more malignant phenotype. The xenografts were then characterised by glucocorticoid sensitivity both in vitro and in vivo, with some xenografts demonstrating discordant responses, indicating that results obtained were model-specific as has been previously described (Stam et al. 2009).

A number of studies have used baseline gene expression profiling of large cohorts of ALL patient samples to determine signatures associated with poor outcome (Flotho et al. 2007; Bhojwani et al. 2008; Den Boer et al. 2009; Harvey et al. 2010; Kang et al. 2010). In this study, due to the small number of patients in each cohort, I was unable to identify a robust signature associated with prednisolone response, and as the majority of both PPR and PGR patients from this cohort are in long-term CR1, correlation with poor outcome was not possible.

Page | 214

I used our laboratory’s well-described method (Lock et al. 2002) using NOD/SCID mice but without irradiation to establish the xenografts. Although some xenografts were created successfully, a number of diagnostic patient samples either failed to engraft or engrafted poorly and were thus unsuitable for experimental purposes. This issue was partly overcome by the use of NSG mice, and I was thus able to establish 8 new xenografts in addition to the 2 that had already been created, resulting in a panel of 10 xenografts which will be a valuable resource for researchers in the future. However, 3 samples failed to engraft in either mouse strain, a limitation of intravenous inoculation which could potentially be overcome by direct intrafemoral inoculation (Mazurier et al. 2003; McKenzie et al. 2005).

The use of xenograft models will remain a significant tool for the preclinical evaluation of new therapeutic agents, and I believe future success will depend on the ability to establish larger xenograft panels encompassing specific patient groups. In ALL, the establishment of xenografts is time consuming, and efforts to optimise the methods used will significantly improve the ability to create new xenografts. I would suggest future xenografts be created in NSG mice using direct intrafemoral inoculation.

Gene expression profiling is a powerful tool to investigate mechanisms of disease, and advances in high-throughput technology have improved the accessibility to these platforms. Gene expression studies need to be sufficiently powered with adequate sample sizes, as differential genes identified from underpowered studies are likely to be non-significant when corrected for multiple hypothesis testing, but this needs to be balanced with often limited sample availability and cost.

Page | 215

In Chapter 4, I investigated glucocorticoid-induced gene expression using the NOD/SCID xenograft mouse model (Bhadri et al. 2011). I determined that the 8 hour timepoint following dexamethasone treatment identified the greatest number of significant differentially expressed genes, and that time-matched controls were redundant. I demonstrated that robust and reproducible results could be obtained from 3 replicates instead of 4, and I thus determined the optimal experimental design to investigate in vivo glucocorticoid-induced gene expression.

A number of publications have reported glucocorticoid-induced gene expression in ALL, primarily in vitro using cell-line models (Obexer et al. 2001; Tonko et al. 2001; Yoshida et al. 2002; Medh et al. 2003; Wang et al. 2003; Webb et al. 2003; Rainer et al. 2009), but also using patient-derived cells, both in vitro (Tissing et al. 2007) and directly from patients in vivo (Schmidt et al. 2006). In this study I showed that the glucocorticoid-induced gene expression profiles using the xenograft mouse model were similar to those obtained directly from treated patients in vivo. Thus, although more complex and time-consuming than in vitro cell line studies, future studies of drug-induced gene expression using the xenograft model are likely to generate reliable and clinically relevant data.

In Chapter 5, I obtained the dexamethasone-regulated gene expression profiles of the entire xenograft cohort. I derived 2 signatures associated with glucocorticoid resistance and an additional third signature of genes associated with in vivo glucocorticoid sensitivity from a separate xenograft panel. Interrogation of CMap with these 3 signatures identified HSP90 inhibitors as agents that induced the complementary gene expression profile and thus could potentially overcome glucocorticoid resistance. Although a moderate degree of synergism was seen with 17-DMAG and dexamethasone in vitro, no synergism was seen in vivo in the tested glucocorticoid resistant xenografts. The acidic vehicle used to make up 17-DMAG for in vivo use was particularly toxic to the mice, and may have contributed to its perceived lack of efficacy.

Page | 216

One of the main issues surrounding the use of CMap is that the CMap database build 02 was established in 2008 by exposing the cancer cell lines MCF7 (breast cancer), SKMEL5 (melanoma), PC3 (prostate cancer) and HL60 (myeloid leukaemia) to drugs and small molecules in vitro (Lamb 2007), whereas 2 of the signatures generated to interrogate CMap were derived from the in vivo treatment of acute lymphoblastic leukaemia xenografts. I have shown that glucocorticoid-induced gene expression is model specific (Sections 4.7 and 5.5) and this may explain why the HSP90 inhibitors identified had some efficacy in vitro, but this was not translated to the in vivo setting.

HSP90 is a conceptually attractive target, and as a co-chaperone of the GR has particularly relevance to glucocorticoid action. In recent years, as the structure of HSP90 has been elucidated a number of novel synthetic small molecule inhibitors unrelated to geldanamycin have been developed with the aim of increasing potency of inhibition whilst minimising toxicity. These novel drugs include SNX7081 (Best et al. 2010), BJ-B11 (Ju et al. 2011), STA-9090 (Proia et al. 2011) and AUY922 (Weigert et al. 2012). None of these molecules featured in CMap build 02, but as HSP90 inhibitors were so consistently identified, I believe one or more of these agents warrants evaluation in combination with dexamethasone in an extended panel of glucocorticoid resistant xenografts.

A significant body of literature has linked glucocorticoid response and glucocorticoid resistance in ALL to the BCL-2 family of proteins, particularly pro- apoptotic BIM/BMF, and pro-survival BCL-2/MCL-1 (Section 1.8.3). In this study, I demonstrated that there were minimal changes in BIM mRNA or protein in any of the PPR or PGR xenografts in vivo. This is intriguing, and in direct contrast to the majority of in vitro literature (Abrams et al. 2004; Bachmann et al. 2005; Lu et al. 2006; Bachmann et al. 2007; Ploner et al. 2008; Bachmann et al. 2010) and even to the results obtained with ALL-3 in vivo in Chapter 4. The reasons for this are not clear, but suggest that (lack of) BIM induction is not the mechanism underlying glucocorticoid resistance in this panel of xenografts.

Page | 217

This study established a well-defined panel of BCP-ALL xenografts, determined the optimal experimental design to investigate in vivo glucocorticoid-induced gene expression, and identified HSP90 inhibitors as agents to potentially reverse glucocorticoid resistance. Further studies should include the evaluation of novel HSP90 inhibitors, and the correlation of gene expression profiles with changes in methylation patterns and proteomic analysis to identify other targets potentially amenable to pharmacological intervention. If successful, these novel agents could be introduced into upfront ‘windows’ in clinical trials of relapsed/refractory ALL, particularly if the patient was classified as a PPR as part of initial therapy.

Page | 218

7 Appendices

Page | 219

7.1 Appendix A – metaGSEA results

A. Top 100 genesets upregulated in ALL-3 8 hours following dexamethasone treatment in vivo. B. Top 100 genesets downregulated in ALL-3 8 hours following dexamethasone treatment in vivo. C. Top 100 genesets upregulated in ALL-3 24 hours following dexamethasone treatment in vivo. D. Top 100 genesets downregulated in ALL-3 24 hours following dexamethasone treatment in vivo.

Page | 220

Page | 221

Page | 222

Page | 223

Page | 224

7.2 Appendix B – Top leading edge genes from metaGSEA Genes identified in the leading edges of at least 5 out of 50 genesets upregulated in PPRs v PGRs following treatment with dexamethasone in vivo.

Upregulated in PPRs ACAT2,ANLN,ANP32E,APOBEC3B,ASPM,AURKA,BIRC5,BRCA1,BRCA2,BUB1,BUB1B,BUB 3,CASC5,CBX5,CCDC99,CCNA2,CCNB1,CCNB2,CCND3,CCNE1,CDC2,CDC25C,CDC7,CDC A3,CDCA8,CDK2,CDKN2C,CDKN3,CDT1,CENPA,CENPC1,CENPE,CENPF,CENPH,CENPK, CENPL,CENPM,CENPN,CENPQ,CEP55,CEP57,CHAF1B,CHEK1,CKAP2,CKS1B,CKS2,COX 7B,DEK,DHFR,DLGAP5,DNA2,DNAJC9,DSCC1,DSN1,DTL,DTYMK,DUT,E2F8,EBP,ECT2,ER CC6L,EXO1,EZH2,FABP5,FANCI,FBL,FBXO5,FOXM1,GINS1,GINS2,GMNN,GPSM2,H2AFV, H2AFZ,HAT1,HELLS,HMGA2,HMGB2,HMGB3,HMGN2,HMMR,HN1,HSPA14,ILF3,IQGAP3,IT GB3BP,KIAA0101,KIF11,KIF14,KIF18A,KIF20A,KIF20B,KIF23,KIF2C,KIF4A,KPNA2,LMNB1,L SM5,MAD2L1,MAPRE1,MCM10,MCM4,MCM6,MELK,MKI67,MLF1IP,MRE11A,MSH6,MTF2,M THFD2,NASP,NCAPD2,NCAPG,NCAPG2,NDC80,NEK2,NME1,NUDT1,NUF2,NUP107,NUSA P1,OIP5,ORC5L,ORC6L,PAICS,PBK,PCNA,PGK1,PLK1,PLK4,POLA1,POLD3,POLE2,PPP1C C,PPP2R5C,PRIM1,PRKDC,PRPS1,PSIP1,PSMA1,PSMA3,PSMA4,PSMA5,PSMA6,PSMB1,P SMB10,PSMB2,PSMB3,PSMB7,PSMB8,PSMB9,PSMC3,PSMC3IP,PSMC5,PSMD10,PSMD13 ,PSMD14,PSMD5,PSMD8,PSMF1,PTTG1,RACGAP1,RAD21,RAD51,RAD51AP1,RAD54L,RB BP4,RBL1,RFC1,RFC2,RFC3,RFC4,RFC5,RIF1,RPA1,RPA3,RRM1,RRM2,SGOL1,SHCBP1,S HMT1,SKA1,SKA2,SKP2,SMC2,SMC4,SNRPD1,SPAG5,SPC24,SPC25,STIL,STMN1,SYNCRI P,TIMELESS,TMEM97,TMPO,TOP2A,TOPBP1,TPX2,TTK,TUBG1,TYMS,UBE2T,USP1,VRK1, WDHD1,WHSC1,ZWILCH,ZWINT

Downregulated in PPRs ACTN1,BAT2,DUSP6,DUSP7,ELK1,EP300,GRB2,HRAS,MAP2K1,MAP2K2,MAP2K6,MAP3K7, MAPK1,MAPK11,MAPK14,MAPK3,MAPK7,MAPKAPK2,MEF2A,MYO9B,PPP2CB,PPP2R1A,P PP2R5D,RELA,RPS6KA2,SRC,TICAM1,UBAP2L

Page | 225

7.3 Appendix C – Gene Ontology of ALL-55/56 with dexamethasone

ALL-55 upregulated in vitro Gene Ontology Term P Enrichment FDR GO:0043549~regulation of kinase activity 0.000 2.644 0.244 GO:0033674~positive regulation of kinase activity 0.001 2.918 1.122 GO:0045859~regulation of protein kinase activity 0.001 2.475 1.212 GO:0004672~protein kinase activity 0.001 2.029 1.931 GO:0042981~regulation of apoptosis 0.002 1.789 3.232 GO:0006915~apoptosis 0.002 1.941 3.343 GO:0043067~regulation of programmed cell death 0.002 1.771 3.781 GO:0010941~regulation of cell death 0.002 1.765 3.967 GO:0012501~programmed cell death 0.002 1.912 4.054 GO:0045860~positive regulation of protein kinase activity 0.004 2.620 6.967 GO:0043066~negative regulation of apoptosis 0.006 2.158 9.371 GO:0006916~anti-apoptosis 0.006 2.618 10.41 GO:0043069~negative regulation of programmed cell death 0.006 2.128 10.57 GO:0004674~protein serine/threonine kinase activity 0.007 2.059 9.372 GO:0060548~negative regulation of cell death 0.007 2.122 10.99 GO:0043549~regulation of kinase activity 0.000 2.644 0.244 GO:0033674~positive regulation of kinase activity 0.001 2.918 1.122 GO:0045859~regulation of protein kinase activity 0.001 2.475 1.212 GO:0004672~protein kinase activity 0.001 2.029 1.931 GO:0042981~regulation of apoptosis 0.002 1.789 3.232 GO:0006915~apoptosis 0.002 1.941 3.343 GO:0043067~regulation of programmed cell death 0.002 1.771 3.781 GO:0010941~regulation of cell death 0.002 1.765 3.967 GO:0012501~programmed cell death 0.002 1.912 4.054 GO:0045860~positive regulation of protein kinase activity 0.004 2.620 6.967 GO:0043066~negative regulation of apoptosis 0.006 2.158 9.371

Page | 226

ALL-55 downregulated in vitro Gene Ontology Term P Enrichment FDR GO:0042981~regulation of apoptosis 0.000 2.041 0.039 GO:0043067~regulation of programmed cell death 0.000 2.021 0.048 GO:0010941~regulation of cell death 0.000 2.013 0.052 GO:0043065~positive regulation of apoptosis 0.000 2.327 0.346 GO:0043068~positive regulation of programmed cell death 0.000 2.311 0.383 GO:0010942~positive regulation of cell death 0.000 2.300 0.410 GO:0006917~induction of apoptosis 0.003 2.251 4.769 GO:0012502~induction of programmed cell death 0.003 2.244 4.922 GO:0043066~negative regulation of apoptosis 0.003 2.148 5.794 GO:0043069~negative regulation of programmed cell death 0.004 2.118 6.590 GO:0060548~negative regulation of cell death 0.004 2.112 6.818 GO:0042981~regulation of apoptosis 0.000 2.041 0.039 GO:0043067~regulation of programmed cell death 0.000 2.021 0.048 GO:0010941~regulation of cell death 0.000 2.013 0.052 GO:0043065~positive regulation of apoptosis 0.000 2.327 0.346 GO:0043068~positive regulation of programmed cell death 0.000 2.311 0.383 GO:0010942~positive regulation of cell death 0.000 2.300 0.410 GO:0006917~induction of apoptosis 0.003 2.251 4.769 GO:0012502~induction of programmed cell death 0.003 2.244 4.922 GO:0043066~negative regulation of apoptosis 0.003 2.148 5.794 GO:0043069~negative regulation of programmed cell death 0.004 2.118 6.590 GO:0060548~negative regulation of cell death 0.004 2.112 6.818 GO:0042981~regulation of apoptosis 0.000 2.041 0.039 GO:0043067~regulation of programmed cell death 0.000 2.021 0.048 GO:0010941~regulation of cell death 0.000 2.013 0.052 GO:0043065~positive regulation of apoptosis 0.000 2.327 0.346 GO:0043068~positive regulation of programmed cell death 0.000 2.311 0.383

Page | 227

ALL-55 upregulated in vivo Gene Ontology Term P Enrichment FDR GO:0043405~regulation of MAP kinase activity 0.000 5.578 0.124 GO:0006915~apoptosis 0.001 2.482 1.011 GO:0012501~programmed cell death 0.001 2.446 1.200 GO:0043406~positive regulation of MAP kinase activity 0.002 5.398 3.019 GO:0043549~regulation of kinase activity 0.002 2.864 3.177 GO:0042981~regulation of apoptosis 0.003 2.054 4.639 GO:0043067~regulation of programmed cell death 0.003 2.034 5.178 GO:0010941~regulation of cell death 0.003 2.027 5.393 GO:0000187~activation of MAPK activity 0.004 5.755 6.099 GO:0008219~cell death 0.004 2.078 6.956 GO:0045859~regulation of protein kinase activity 0.004 2.736 7.110 GO:0033674~positive regulation of kinase activity 0.009 3.064 14.13 GO:0043405~regulation of MAP kinase activity 0.000 5.578 0.124

ALL-55 downregulated in vivo Gene Ontology Term P Enrichment FDR GO:0046983~protein dimerization activity 0.001 3.992 1.823 GO:0032496~response to lipopolysaccharide 0.005 11.335 7.591 GO:0005615~extracellular space 0.005 3.230 6.056 GO:0016023~cytoplasmic membrane-bounded vesicle 0.006 3.575 6.602 GO:0002237~response to molecule of bacterial origin 0.007 10.149 10.19 GO:0031988~membrane-bounded vesicle 0.007 3.462 7.796 GO:0005625~soluble fraction 0.008 4.712 8.856 GO:0005125~cytokine activity 0.008 6.165 9.662 GO:0000267~cell fraction 0.009 2.497 10.45 GO:0046983~protein dimerization activity 0.001 3.992 1.823 GO:0032496~response to lipopolysaccharide 0.005 11.335 7.591 GO:0005615~extracellular space 0.005 3.230 6.056 GO:0016023~cytoplasmic membrane-bounded vesicle 0.006 3.575 6.602

Page | 228

ALL-56 upregulated in vitro Gene Ontology Term P Enrichment FDR GO:0045859~regulation of protein kinase activity 0.000 3.465 0.016 GO:0043405~regulation of MAP kinase activity 0.000 5.355 0.025 GO:0006915~apoptosis 0.000 2.404 0.375 GO:0012501~programmed cell death 0.000 2.369 0.461 GO:0008219~cell death 0.001 2.100 1.722 GO:0016265~death 0.001 2.086 1.886 GO:0033674~positive regulation of kinase activity 0.001 3.269 1.896 GO:0006469~negative regulation of protein kinase activity 0.003 5.063 4.191 GO:0000165~MAPKKK cascade 0.003 3.420 4.349 GO:0045860~positive regulation of protein kinase activity 0.003 3.104 4.807 GO:0033673~negative regulation of kinase activity 0.003 4.894 4.952 GO:0046328~regulation of JNK cascade 0.004 5.808 5.907 GO:0004672~protein kinase activity 0.005 2.096 7.355 GO:0043406~positive regulation of MAP kinase activity 0.006 4.318 8.983 GO:0045859~regulation of protein kinase activity 0.000 3.465 0.016 GO:0043405~regulation of MAP kinase activity 0.000 5.355 0.025 GO:0006915~apoptosis 0.000 2.404 0.375 GO:0012501~programmed cell death 0.000 2.369 0.461 GO:0008219~cell death 0.001 2.100 1.722 GO:0016265~death 0.001 2.086 1.886 GO:0033674~positive regulation of kinase activity 0.001 3.269 1.896 GO:0006469~negative regulation of protein kinase activity 0.003 5.063 4.191 GO:0000165~MAPKKK cascade 0.003 3.420 4.349 GO:0045860~positive regulation of protein kinase activity 0.003 3.104 4.807 GO:0033673~negative regulation of kinase activity 0.003 4.894 4.952 GO:0046328~regulation of JNK cascade 0.004 5.808 5.907

Page | 229

ALL-56 downregulated in vitro Gene Ontology Term P Enrichment FDR GO:0042981~regulation of apoptosis 0.000 2.749 0.018 GO:0043067~regulation of programmed cell death 0.000 2.722 0.021 GO:0010941~regulation of cell death 0.000 2.712 0.022 GO:0051384~response to glucocorticoid stimulus 0.000 7.935 0.410 GO:0009986~cell surface 0.000 3.485 0.416 GO:0043065~positive regulation of apoptosis 0.000 3.084 0.593 GO:0043068~positive regulation of programmed cell death 0.000 3.063 0.636 GO:0031960~response to corticosteroid stimulus 0.000 7.282 0.653 GO:0010942~positive regulation of cell death 0.000 3.049 0.666 GO:0043066~negative regulation of apoptosis 0.002 2.997 3.662 GO:0043069~negative regulation of programmed cell death 0.002 2.955 4.074 GO:0060548~negative regulation of cell death 0.002 2.947 4.161 GO:0051726~regulation of cell cycle 0.004 2.938 6.934 GO:0006916~anti-apoptosis 0.009 3.434 13.80 GO:0042981~regulation of apoptosis 0.000 2.749 0.018 GO:0043067~regulation of programmed cell death 0.000 2.722 0.021 GO:0010941~regulation of cell death 0.000 2.712 0.022 GO:0051384~response to glucocorticoid stimulus 0.000 7.935 0.410 GO:0009986~cell surface 0.000 3.485 0.416 GO:0043065~positive regulation of apoptosis 0.000 3.084 0.593 GO:0043068~positive regulation of programmed cell death 0.000 3.063 0.636 GO:0031960~response to corticosteroid stimulus 0.000 7.282 0.653 GO:0010942~positive regulation of cell death 0.000 3.049 0.666 GO:0043066~negative regulation of apoptosis 0.002 2.997 3.662 GO:0043069~negative regulation of programmed cell death 0.002 2.955 4.074 GO:0060548~negative regulation of cell death 0.002 2.947 4.161

Page | 230

ALL-56 upregulated in vivo Gene Ontology Term P Enrichment FDR GO:0004672~protein kinase activity 0.000 2.430 0.378 GO:0043549~regulation of kinase activity 0.004 2.479 6.925 GO:0007243~protein kinase cascade 0.006 2.392 9.191 GO:0006915~apoptosis 0.007 1.995 10.77 GO:0043405~regulation of MAP kinase activity 0.007 3.587 10.85 GO:0012501~programmed cell death 0.008 1.966 12.36 GO:0042981~regulation of apoptosis 0.008 1.808 12.59 GO:0045859~regulation of protein kinase activity 0.009 2.382 13.24 GO:0000165~MAPKKK cascade 0.009 3.092 13.38 GO:0043067~regulation of programmed cell death 0.009 1.791 13.94 GO:0000187~activation of MAPK activity 0.009 4.625 14.48 GO:0010941~regulation of cell death 0.009 1.784 14.48

ALL-56 downregulated in vivo Gene Ontology Term P Enrichment FDR GO:0042981~regulation of apoptosis 0.000 2.345 0.470 GO:0043067~regulation of programmed cell death 0.000 2.322 0.539 GO:0010941~regulation of cell death 0.000 2.314 0.567 GO:0016265~death 0.003 2.152 4.876 GO:0012501~programmed cell death 0.003 2.281 5.017 GO:0008219~cell death 0.006 2.053 10.16 GO:0006915~apoptosis 0.006 2.179 10.29 GO:0042981~regulation of apoptosis 0.000 2.345 0.470 GO:0043067~regulation of programmed cell death 0.000 2.322 0.539 GO:0010941~regulation of cell death 0.000 2.314 0.567 GO:0016265~death 0.003 2.152 4.876 GO:0012501~programmed cell death 0.003 2.281 5.017

Page | 231

8 References

Page | 232

Abrams, M. T., N. M. Robertson, et al. (2004). "Inhibition of glucocorticoid- induced apoptosis by targeting the major splice variants of BIM mRNA with small interfering RNA and short hairpin RNA." J Biol Chem 279(53): 55809-17. ABS (2007). "Health of children in Australia: A Snapshot, 2004-5." Australian Bureau of Statistics. Agliano, A., I. Martin-Padura, et al. (2008). "Human acute leukemia cells injected in NOD/LtSz-scid/IL-2Rgamma null mice generate a faster and more efficient disease compared to other NOD/scid-related strains." Int J Cancer 123(9): 2222-7. AIHW: McDermid, I. (2005). Cancer incidence projections Australia 2002 to 2011 Australian Institute of Health and Welfare. Al Shaer, L., E. Walsby, et al. (2008). "Heat shock protein 90 inhibition is cytotoxic to primary AML cells expressing mutant FLT3 and results in altered downstream signalling." Br J Haematol 141(4): 483-93. Anderson, K., C. Lutz, et al. (2011). "Genetic variegation of clonal architecture and propagating cells in leukaemia." Nature 469(7330): 356-61. Artandi, S. E., S. Chang, et al. (2000). "Telomere dysfunction promotes non- reciprocal translocations and epithelial cancers in mice." Nature 406(6796): 641-5. Auphan, N., J. A. DiDonato, et al. (1995). "Immunosuppression by glucocorticoids: inhibition of NF-kappa B activity through induction of I kappa B synthesis." Science 270(5234): 286-90. Bachmann, P. S., R. Gorman, et al. (2005). "Dexamethasone resistance in B- cell precursor childhood acute lymphoblastic leukemia occurs downstream of ligand-induced nuclear translocation of the glucocorticoid receptor." Blood 105(6): 2519-26. Bachmann, P. S., R. Gorman, et al. (2007). "Divergent mechanisms of glucocorticoid resistance in experimental models of pediatric acute lymphoblastic leukemia." Cancer Res 67(9): 4482-90. Bachmann, P. S., R. G. Piazza, et al. (2010). "Epigenetic silencing of BIM in glucocorticoid poor-responsive pediatric acute lymphoblastic leukemia, and its reversal by histone deacetylase inhibition." Blood 116(16): 3013- 22. Baersch, G., T. Mollers, et al. (1997). "Good engraftment of B-cell precursor ALL in NOD-SCID mice." Klin Padiatr 209(4): 178-85. Baxter, J. D., A. W. Harris, et al. (1971). "Glucocorticoid receptors in lymphoma cells in culture: relationship to glucocorticoid killing activity." Science 171(967): 189-91.

Page | 233

Beesley, A. H., R. E. Weller, et al. (2009a). "Receptor mutation is not a common mechanism of naturally occurring glucocorticoid resistance in leukaemia cell lines." Leuk Res 33(2): 321-5. Beesley, A. H., M. J. Firth, et al. (2009b). "Glucocorticoid resistance in T-lineage acute lymphoblastic leukaemia is associated with a proliferative metabolism." Br J Cancer 100(12): 1926-36. Best, O. G., N. Singh, et al. (2010). "The novel Hsp-90 inhibitor SNX7081 is significantly more potent than 17-AAG against primary CLL cells and a range of haematological cell lines, irrespective of lesions in the TP53 pathway." Br J Haematol 151(2): 185-8. Bhadri, V. A., M. J. Cowley, et al. (2011). "Evaluation of the NOD/SCID xenograft model for glucocorticoid-regulated gene expression in childhood B-cell precursor acute lymphoblastic leukemia." BMC Genomics 12: 565. Bhojwani, D., H. Kang, et al. (2008). "Gene expression signatures predictive of early response and outcome in high-risk childhood acute lymphoblastic leukemia: A Children's Oncology Group Study [corrected]." J Clin Oncol 26(27): 4376-84. Bird, G. H., F. Bernal, et al. (2008). "Synthesis and biophysical characterization of stabilized alpha-helices of BCL-2 domains." Methods Enzymol 446: 369-86. Bolstad, B. M., R. A. Irizarry, et al. (2003). "A comparison of normalization methods for high density oligonucleotide array data based on variance and bias." Bioinformatics 19(2): 185-93. Bonapace, L., B. C. Bornhauser, et al. (2010). "Induction of autophagy- dependent necroptosis is required for childhood acute lymphoblastic leukemia cells to overcome glucocorticoid resistance." J Clin Invest 120(4): 1310-23. Bornhauser, B. C., L. Bonapace, et al. (2007). "Low-dose arsenic trioxide sensitizes glucocorticoid-resistant acute lymphoblastic leukemia cells to dexamethasone via an Akt-dependent pathway." Blood 110(6): 2084-91. Cario, G., A. Fetz, et al. (2008). "Initial leukemic gene expression profiles of patients with poor in vivo prednisone response are similar to those of blasts persisting under prednisone treatment in childhood acute lymphoblastic leukemia." Ann Hematol 87(9): 709-16. Castro, J. E., C. E. Prada, et al. (2005). "ZAP-70 is a novel conditional heat shock protein 90 (Hsp90) client: inhibition of Hsp90 leads to ZAP-70 degradation, apoptosis, and impaired signaling in chronic lymphocytic leukemia." Blood 106(7): 2506-12.

Page | 234

Catts, V. S., M. L. Farnsworth, et al. (2001). "High level resistance to glucocorticoids, associated with a dysfunctional glucocorticoid receptor, in childhood acute lymphoblastic leukemia cells selected for methotrexate resistance." Leukemia 15(6): 929-35. Choi, S., M. J. Henderson, et al. (2007). "Relapse in children with acute lymphoblastic leukemia involving selection of a preexisting drug-resistant subclone." Blood 110(2): 632-9. Clappier, E., B. Gerby, et al. (2011). "Clonal selection in xenografted human T cell acute lymphoblastic leukemia recapitulates gain of malignancy at relapse." J Exp Med 208(4): 653-61. Conter, V., C. R. Bartram, et al. (2010). "Molecular response to treatment redefines all prognostic factors in children and adolescents with B-cell precursor acute lymphoblastic leukemia: results in 3184 patients of the AIEOP-BFM ALL 2000 study." Blood 115(16): 3206-14. D'Adamio, F., O. Zollo, et al. (1997). "A new dexamethasone-induced gene of the leucine zipper family protects T lymphocytes from TCR/CD3- activated cell death." Immunity 7(6): 803-12. Davies, T. H., Y. M. Ning, et al. (2002). "A new first step in activation of steroid receptors: hormone-induced switching of FKBP51 and FKBP52 immunophilins." J Biol Chem 277(7): 4597-600. DeBoer, C., P. A. Meulman, et al. (1970). "Geldanamycin, a new antibiotic." J Antibiot (Tokyo) 23(9): 442-7. Den Boer, M. L., M. van Slegtenhorst, et al. (2009). "A subtype of childhood acute lymphoblastic leukaemia with poor treatment outcome: a genome- wide classification study." Lancet Oncol 10(2): 125-34. Donaldson, I. J. and B. Gottgens (2007). "CoMoDis: composite motif discovery in mammalian genomes." Nucleic Acids Res 35(1): e1. Donehower, L. A., M. Harvey, et al. (1992). "Mice deficient for p53 are developmentally normal but susceptible to spontaneous tumours." Nature 356(6366): 215-21. Dordelmann, M., A. Reiter, et al. (1999). "Prednisone response is the strongest predictor of treatment outcome in infant acute lymphoblastic leukemia." Blood 94(4): 1209-17. Drexler, H. G., S. Fombonne, et al. (2000). "p53 alterations in human leukemia- lymphoma cell lines: in vitro artefact or prerequisite for cell immortalization?" Leukemia 14(1): 198-206. Drexler, H. G., A. Y. Matsuo, et al. (2000). "Continuous hematopoietic cell lines as model systems for leukemia-lymphoma research." Leuk Res 24(11): 881-911.

Page | 235

Drouin, J., J. Charron, et al. (1987). "Pro-opiomelanocortin gene: a model for negative regulation of transcription by glucocorticoids." J Cell Biochem 35(4): 293-304. Du, P., W. A. Kibbe, et al. (2008). "lumi: a pipeline for processing Illumina microarray." Bioinformatics 24(13): 1547-8. Eberhart, K., K. Renner, et al. (2009). "Low doses of 2-deoxy-glucose sensitize acute lymphoblastic leukemia cells to glucocorticoid-induced apoptosis." Leukemia 23(11): 2167-70. Egorin, M. J., T. F. Lagattuta, et al. (2002). "Pharmacokinetics, tissue distribution, and metabolism of NSC 707545 (17- (dimethylaminoethylamino)-17-demethoxygeldanamycin) in CD2F1 mice and Fischer 344 rats." Cancer Chemother Pharmacol 49(1): 7-19. Erlacher, M., E. M. Michalak, et al. (2005). "BH3-only proteins Puma and Bim are rate-limiting for gamma-radiation- and glucocorticoid-induced apoptosis of lymphoid cells in vivo." Blood 106(13): 4131-8. Fiske, J. L., V. P. Fomin, et al. (2006). "Voltage-sensitive ion channels and cancer." Cancer Metastasis Rev 25(3): 493-500. Flotho, C., E. Coustan-Smith, et al. (2007). "A set of genes that regulate cell proliferation predicts treatment outcome in childhood acute lymphoblastic leukemia." Blood 110(4): 1271-7. Foley, G. E., H. Lazarus, et al. (1965). "Continuous Culture of Human Lymphoblasts from Peripheral Blood of a Child with Acute Leukemia." Cancer 18: 522-9. Galigniana, M. D., J. M. Harrell, et al. (2002). "Binding of hsp90-associated immunophilins to cytoplasmic dynein: direct binding and in vivo evidence that the peptidylprolyl isomerase domain is a dynein interaction domain." Biochemistry 41(46): 13602-10. Gavathiotis, E., M. Suzuki, et al. (2008). "BAX activation is initiated at a novel interaction site." Nature 455(7216): 1076-81. Gaynon, P. S. and A. L. Carrel (1999). "Glucocorticosteroid therapy in childhood acute lymphoblastic leukemia." Adv Exp Med Biol 457: 593-605. Geley, S., B. L. Hartmann, et al. (1996). "Resistance to glucocorticoid-induced apoptosis in human T-cell acute lymphoblastic leukemia CEM-C1 cells is due to insufficient glucocorticoid receptor expression." Cancer Res 56(21): 5033-8. Gentleman, R. C., V. J. Carey, et al. (2004). "Bioconductor: open software development for computational biology and bioinformatics." Genome Biol 5(10): R80.

Page | 236

Gruber, G., M. Carlet, et al. (2009). "Levels of glucocorticoid receptor and its ligand determine sensitivity and kinetics of glucocorticoid-induced leukemia apoptosis." Leukemia 23(4): 820-3. Gu, L., C. Zhou, et al. (2010). "Rapamycin sensitizes T-ALL cells to dexamethasone-induced apoptosis." J Exp Clin Cancer Res 29: 150. Haarman, E. G., G. J. Kaspers, et al. (2004). "Glucocorticoid receptor alpha, beta and gamma expression vs in vitro glucocorticoid resistance in childhood leukemia." Leukemia 18(3): 530-7. Harrell, J. M., P. J. Murphy, et al. (2004). "Evidence for glucocorticoid receptor transport on microtubules by dynein." J Biol Chem 279(52): 54647-54. Harvey, R. C., C. G. Mullighan, et al. (2010). "Rearrangement of CRLF2 is associated with mutation of JAK kinases, alteration of IKZF1, Hispanic/Latino ethnicity, and a poor outcome in pediatric B-progenitor acute lymphoblastic leukemia." Blood 115(26): 5312-21. Harvey, R. C., C. G. Mullighan, et al. (2010). "Identification of novel cluster groups in pediatric high-risk B-precursor acute lymphoblastic leukemia with gene expression profiling: correlation with genome-wide DNA copy number alterations, clinical characteristics, and outcome." Blood 116(23): 4874-84. Heidari, N., M. A. Hicks, et al. (2010). "GX15-070 (obatoclax) overcomes glucocorticoid resistance in acute lymphoblastic leukemia through induction of apoptosis and autophagy." Cell Death Dis 1: e76. Hertlein, E., A. J. Wagner, et al. (2010). "17-DMAG targets the nuclear factor- kappaB family of proteins to induce apoptosis in chronic lymphocytic leukemia: clinical implications of HSP90 inhibition." Blood 116(1): 45-53. Hillmann, A. G., J. Ramdas, et al. (2000). "Glucocorticoid receptor gene mutations in leukemic cells acquired in vitro and in vivo." Cancer Res 60(7): 2056-62. Holleman, A., M. H. Cheok, et al. (2004). "Gene-expression patterns in drug- resistant acute lymphoblastic leukemia cells and response to treatment." N Engl J Med 351(6): 533-42. Hollenberg, S. M., C. Weinberger, et al. (1985). "Primary structure and expression of a functional human glucocorticoid receptor cDNA." Nature 318(6047): 635-41. Hollingshead, M., M. Alley, et al. (2005). "In vivo antitumor efficacy of 17-DMAG (17-dimethylaminoethylamino-17-demethoxygeldanamycin hydrochloride), a water-soluble geldanamycin derivative." Cancer Chemother Pharmacol 56(2): 115-25.

Page | 237

Hostein, I., D. Robertson, et al. (2001). "Inhibition of signal transduction by the Hsp90 inhibitor 17-allylamino-17-demethoxygeldanamycin results in cytostasis and apoptosis." Cancer Res 61(10): 4003-9. Houghton, P. J., C. L. Morton, et al. (2007). "The pediatric preclinical testing program: description of models and early testing results." Pediatr Blood Cancer 49(7): 928-40. Huang da, W., B. T. Sherman, et al. (2009). "Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources." Nat Protoc 4(1): 44-57. Huang, D. C. and A. Strasser (2000). "BH3-Only proteins-essential initiators of apoptotic cell death." Cell 103(6): 839-42. Hulleman, E., K. M. Kazemier, et al. (2009). "Inhibition of glycolysis modulates prednisolone resistance in acute lymphoblastic leukemia cells." Blood 113(9): 2014-21. Hyman, C. B. and P. Sturgeon (1956). "Prednisone therapy of acute lymphatic leukemia in children." Cancer 9(5): 965-70. Jiang, N., G. S. Koh, et al. (2011). "BIM is a prognostic biomarker for early prednisolone response in pediatric acute lymphoblastic leukemia." Exp Hematol 39(3): 321-9, 329 e1-3. Johnson, A. J., A. J. Wagner, et al. (2007). "Rituximab and 17-allylamino-17- demethoxygeldanamycin induce synergistic apoptosis in B-cell chronic lymphocytic leukaemia." Br J Haematol 139(5): 837-44. Ju, H. Q., S. X. Wang, et al. (2011). "BJ-B11, a novel Hsp90 inhibitor, induces apoptosis in human chronic myeloid leukemia K562 cells through the mitochondria-dependent pathway." Eur J Pharmacol 666(1-3): 26-34. Kang, H., I. M. Chen, et al. (2010). "Gene expression classifiers for relapse-free survival and minimal residual disease improve risk classification and outcome prediction in pediatric B-precursor acute lymphoblastic leukemia." Blood 115(7): 1394-405. Kaplan, E. L. and P. Meier (1958). "Nonparametric estimation from incomplete observations." American Statistical Association Journal: 457-481. Kaufmann, S. H., J. E. Karp, et al. (2011). "Phase I and pharmacological study of cytarabine and tanespimycin in relapsed and refractory acute leukemia." Haematologica 96(11): 1619-26. Kelly, P. N., A. Dakic, et al. (2007). "Tumor growth need not be driven by rare cancer stem cells." Science 317(5836): 337. Kim, S. Y. and D. J. Volsky (2005). "PAGE: parametric analysis of gene set enrichment." BMC Bioinformatics 6: 144.

Page | 238

Klumper, E., R. Pieters, et al. (1995). "In vitro cellular drug resistance in children with relapsed/refractory acute lymphoblastic leukemia." Blood 86(10): 3861-8. Koga, Y., A. Matsuzaki, et al. (2005). "Differential mRNA expression of glucocorticoid receptor alpha and beta is associated with glucocorticoid sensitivity of acute lymphoblastic leukemia in children." Pediatr Blood Cancer 45(2): 121-7. Kojika, S., K. Sugita, et al. (1996). "Mechanisms of glucocorticoid resistance in human leukemic cells: implication of abnormal 90 and 70 kDa heat shock proteins." Leukemia 10(6): 994-9. Kralovics, R., F. Passamonti, et al. (2005). "A gain-of-function mutation of JAK2 in myeloproliferative disorders." N Engl J Med 352(17): 1779-90. Laane, E., T. Panaretakis, et al. (2007). "Dexamethasone-induced apoptosis in acute lymphoblastic leukemia involves differential regulation of Bcl-2 family members." Haematologica 92(11): 1460-9. Labi, V., M. Erlacher, et al. (2008). "Loss of the BH3-only protein Bmf impairs B cell homeostasis and accelerates gamma irradiation-induced thymic lymphoma development." J Exp Med 205(3): 641-55. Lamb, J. (2007). "The Connectivity Map: a new tool for biomedical research." Nat Rev Cancer 7(1): 54-60. Lamb, J., E. D. Crawford, et al. (2006). "The Connectivity Map: using gene- expression signatures to connect small molecules, genes, and disease." Science 313(5795): 1929-35. Lancet, J. E., I. Gojo, et al. (2010). "Phase I study of the heat shock protein 90 inhibitor alvespimycin (KOS-1022, 17-DMAG) administered intravenously twice weekly to patients with acute myeloid leukemia." Leukemia 24(4): 699-705. Lauten, M., C. Beger, et al. (2003). "Expression of heat-shock protein 90 in glucocorticoid-sensitive and -resistant childhood acute lymphoblastic leukaemia." Leukemia 17(8): 1551-6. Lauten, M., G. Cario, et al. (2003). "Protein expression of the glucocorticoid receptor in childhood acute lymphoblastic leukemia." Haematologica 88(11): 1253-8. Liem, N. L., R. A. Papa, et al. (2004). "Characterization of childhood acute lymphoblastic leukemia xenograft models for the preclinical evaluation of new therapies." Blood 103(10): 3905-14. Lin, S. M., P. Du, et al. (2008). "Model-based variance-stabilizing transformation for Illumina microarray data." Nucleic Acids Res 36(2): e11.

Page | 239

Lock, R. B., N. Liem, et al. (2002). "The nonobese diabetic/severe combined immunodeficient (NOD/SCID) mouse model of childhood acute lymphoblastic leukemia reveals intrinsic differences in biologic characteristics at diagnosis and relapse." Blood 99(11): 4100-8. Lu, J., B. Quearry, et al. (2006). "p38-MAP kinase activation followed by BIM induction is essential for glucocorticoid-induced apoptosis in lymphoblastic leukemia cells." FEBS Lett 580(14): 3539-44. Lu, Z., Y. Jin, et al. (2010). "Celastrol, a novel HSP90 inhibitor, depletes Bcr-Abl and induces apoptosis in imatinib-resistant chronic myelogenous leukemia cells harboring T315I mutation." Cancer Lett 290(2): 182-91. Madzo, J., J. Zuna, et al. (2003). "Slower molecular response to treatment predicts poor outcome in patients with TEL/AML1 positive acute lymphoblastic leukemia: prospective real-time quantitative reverse transcriptase-polymerase chain reaction study." Cancer 97(1): 105-13. Malaise, M., M. Neumeier, et al. (2011). "Stable and reproducible engraftment of primary adult and pediatric acute myeloid leukemia in NSG mice." Leukemia 25(10): 1635-9. Mantel, N. (1966). "Evaluation of survival data and two new rank order statistics arising in its consideration." Cancer Chemother Rep 50(3): 163-70. Matys, V., E. Fricke, et al. (2003). "TRANSFAC: transcriptional regulation, from patterns to profiles." Nucleic Acids Res 31(1): 374-8. Mazurier, F., M. Doedens, et al. (2003). "Rapid myeloerythroid repopulation after intrafemoral transplantation of NOD-SCID mice reveals a new class of human stem cells." Nat Med 9(7): 959-63. McKenzie, J. L., O. I. Gan, et al. (2005). "Human short-term repopulating stem cells are efficiently detected following intrafemoral transplantation into NOD/SCID recipients depleted of CD122+ cells." Blood 106(4): 1259-61. Medh, R. D., A. Wang, et al. (2001). "Constitutive expression of ectopic c-Myc delays glucocorticoid-evoked apoptosis of human leukemic CEM-C7 cells." Oncogene 20(34): 4629-39. Medh, R. D., M. S. Webb, et al. (2003). "Gene expression profile of human lymphoid CEM cells sensitive and resistant to glucocorticoid-evoked apoptosis." Genomics 81(6): 543-55. Meyer, L. H., S. M. Eckhoff, et al. (2011). "Early relapse in all is identified by time to leukemia in NOD/SCID mice and is characterized by a gene signature involving survival pathways." Cancer Cell 19(2): 206-17. Modi, S., A. T. Stopeck, et al. (2007). "Combination of trastuzumab and tanespimycin (17-AAG, KOS-953) is safe and active in trastuzumab-

Page | 240

refractory HER-2 overexpressing breast cancer: a phase I dose- escalation study." J Clin Oncol 25(34): 5410-7. Moricke, A., A. Reiter, et al. (2008). "Risk-adjusted therapy of acute lymphoblastic leukemia can decrease treatment burden and improve survival: treatment results of 2169 unselected pediatric and adolescent patients enrolled in the trial ALL-BFM 95." Blood 111(9): 4477-4489. Mosmann, T. (1983). "Rapid colorimetric assay for cellular growth and survival: application to proliferation and cytotoxicity assays." J Immunol Methods 65(1-2): 55-63. Mullighan, C. G., J. R. Collins-Underwood, et al. (2009). "Rearrangement of CRLF2 in B-progenitor- and Down syndrome-associated acute lymphoblastic leukemia." Nat Genet 41(11): 1243-6. Mullighan, C. G., S. Goorha, et al. (2007). "Genome-wide analysis of genetic alterations in acute lymphoblastic leukaemia." Nature 446(7137): 758-64. Mullighan, C. G., L. A. Phillips, et al. (2008). "Genomic analysis of the clonal origins of relapsed acute lymphoblastic leukemia." Science 322(5906): 1377-80. Mullighan, C. G., X. Su, et al. (2009). "Deletion of IKZF1 and prognosis in acute lymphoblastic leukemia." N Engl J Med 360(5): 470-80. Mullighan, C. G., J. Zhang, et al. (2009). "JAK mutations in high-risk childhood acute lymphoblastic leukemia." Proc Natl Acad Sci U S A 106(23): 9414- 8. Nachman, J. B., N. A. Heerema, et al. (2007). "Outcome of treatment in children with hypodiploid acute lymphoblastic leukemia." Blood 110(4): 1112-5. Neale, G., X. Su, et al. (2008). "Molecular characterization of the pediatric preclinical testing panel." Clin Cancer Res 14(14): 4572-83. Neckers, L. and P. Workman (2012). "Hsp90 molecular chaperone inhibitors: are we there yet?" Clin Cancer Res 18(1): 64-76. Nijmeijer, B. A., P. Mollevanger, et al. (2001). "Monitoring of engraftment and progression of acute lymphoblastic leukemia in individual NOD/SCID mice." Exp Hematol 29(3): 322-9. Notta, F., S. Doulatov, et al. (2010). "Engraftment of human hematopoietic stem cells is more efficient in female NOD/SCID/IL-2Rgc-null recipients." Blood 115(18): 3704-7. Oakley, R. H., M. Sar, et al. (1996). "The human glucocorticoid receptor beta isoform. Expression, biochemical properties, and putative function." J Biol Chem 271(16): 9550-9.

Page | 241

Obexer, P., U. Certa, et al. (2001). "Expression profiling of glucocorticoid- treated T-ALL cell lines: rapid repression of multiple genes involved in RNA-, protein- and nucleotide synthesis." Oncogene 20(32): 4324-36. Paakinaho, V., H. Makkonen, et al. (2010). "Glucocorticoid receptor activates poised FKBP51 locus through long-distance interactions." Mol Endocrinol 24(3): 511-25. Pacey, S., R. H. Wilson, et al. (2011). "A phase I study of the heat shock protein 90 inhibitor alvespimycin (17-DMAG) given intravenously to patients with advanced solid tumors." Clin Cancer Res 17(6): 1561-70. Pantoja, C., J. T. Huff, et al. (2008). "Glucocorticoid signaling defines a novel commitment state during adipogenesis in vitro." Mol Biol Cell 19(10): 4032-41. Paulsson, K. and B. Johansson (2009). "High hyperdiploid childhood acute lymphoblastic leukemia." Genes Chromosomes Cancer 48(8): 637-60. Pavlidis, P., Q. Li, et al. (2003). "The effect of replication on gene expression microarray experiments." Bioinformatics 19(13): 1620-7. Peng, C., J. Brain, et al. (2007). "Inhibition of heat shock protein 90 prolongs survival of mice with BCR-ABL-T315I-induced leukemia and suppresses leukemic stem cells." Blood 110(2): 678-85. Pieters, R., M. Schrappe, et al. (2007). "A treatment protocol for infants younger than 1 year with acute lymphoblastic leukaemia (Interfant-99): an observational study and a multicentre randomised trial." Lancet 370(9583): 240-50. Planey, S. L., M. T. Abrams, et al. (2003). "Role of apical caspases and glucocorticoid-regulated genes in glucocorticoid-induced apoptosis of pre-B leukemic cells." Cancer Res 63(1): 172-8. Ploner, C., J. Rainer, et al. (2009). "Repression of the BH3-only molecule PMAIP1/Noxa impairs glucocorticoid sensitivity of acute lymphoblastic leukemia cells." Apoptosis 14(6): 821-8. Ploner, C., J. Rainer, et al. (2008). "The BCL2 rheostat in glucocorticoid- induced apoptosis of acute lymphoblastic leukemia." Leukemia 22(2): 370-7. Powers, J. H., A. G. Hillmann, et al. (1993). "Cloning and expression of mutant glucocorticoid receptors from glucocorticoid-sensitive and -resistant human leukemic cells." Cancer Res 53(17): 4059-65. Pratt, W. B. and D. O. Toft (2003). "Regulation of signaling protein function and trafficking by the hsp90/hsp70-based chaperone machinery." Exp Biol Med 228(2): 111-33.

Page | 242

Prochazka, M., H. R. Gaskins, et al. (1992). "The nonobese diabetic scid mouse: model for spontaneous thymomagenesis associated with immunodeficiency." Proc Natl Acad Sci U S A 89(8): 3290-4. Proia, D. A., K. P. Foley, et al. (2011). "Multifaceted intervention by the Hsp90 inhibitor ganetespib (STA-9090) in cancer cells with activated JAK/STAT signaling." PLoS One 6(4): e18552. Pui, C. H. and W. E. Evans (2006). "Treatment of acute lymphoblastic leukemia." N Engl J Med 354(2): 166-78. Pulvertaft, J. V. (1964). "Cytology of Burkitt's Tumour (African Lymphoma)." Lancet 1(7327): 238-40. Rainer, J., C. Ploner, et al. (2009). "Glucocorticoid-regulated microRNAs and mirtrons in acute lymphoblastic leukemia." Leukemia 23(4): 746-52. Rambal, A. A., Z. L. Panaguiton, et al. (2009). "MEK inhibitors potentiate dexamethasone lethality in acute lymphoblastic leukemia cells through the pro-apoptotic molecule BIM." Leukemia 23(10): 1744-54. Ratajczak, T., B. K. Ward, et al. (2003). "Immunophilin chaperones in steroid receptor signalling." Curr Top Med Chem 3(12): 1348-57. Ray, A. and K. E. Prefontaine (1994). "Physical association and functional antagonism between the p65 subunit of transcription factor NF-kappa B and the glucocorticoid receptor." Proc Natl Acad Sci U S A 91(2): 752-6. Real, P. J., V. Tosello, et al. (2009). "Gamma-secretase inhibitors reverse glucocorticoid resistance in T cell acute lymphoblastic leukemia." Nat Med 15(1): 50-8. Reich, M., T. Liefeld, et al. (2006). "GenePattern 2.0." Nat Genet 38(5): 500-1. Rhein, P., S. Scheid, et al. (2007). "Gene expression shift towards normal B cells, decreased proliferative capacity and distinct surface receptors characterize leukemic blasts persisting during induction therapy in childhood acute lymphoblastic leukemia." Leukemia 21(5): 897-905. Rice, A. M., J. A. Wood, et al. (2000). "Conditions that enable human hematopoietic stem cell engraftment in all NOD-SCID mice." Transplantation 69(5): 927-35. Riehm, H., A. Reiter, et al. (1987). "[Corticosteroid-dependent reduction of leukocyte count in blood as a prognostic factor in acute lymphoblastic leukemia in childhood (therapy study ALL-BFM 83)]." Klin Padiatr 199(3): 151-60. Robertson, A. M., C. C. Bird, et al. (1978). "Morphological aspects of glucocorticoid-induced cell death in human lymphoblastoid cells." J Pathol 126(3): 181-7.

Page | 243

Roy, A., A. Cargill, et al. (2005). "Outcome after first relapse in childhood acute lymphoblastic leukaemia - lessons from the United Kingdom R2 trial." Br J Haematol 130(1): 67-75. Salomons, G. S., L. A. Smets, et al. (1999). "Bcl-2 family members in childhood acute lymphoblastic leukemia: relationships with features at presentation, in vitro and in vivo drug response and long-term clinical outcome." Leukemia 13(10): 1574-80. Samuels, A. L., V. K. Peeva, et al. (2010). "Validation of a mouse xenograft model system for gene expression analysis of human acute lymphoblastic leukaemia." BMC Genomics 11: 256. Sasaki, K., H. Yasuda, et al. (1979). "Growth inhibition of virus transformed cells in vitro and antitumor activity in vivo of geldanamycin and its derivatives." J Antibiot (Tokyo) 32(8): 849-51. Scherer, W. F., J. T. Syverton, et al. (1953). "Studies on the propagation in vitro of poliomyelitis viruses. IV. Viral multiplication in a stable strain of human malignant epithelial cells (strain HeLa) derived from an epidermoid carcinoma of the cervix." J Exp Med 97(5): 695-710. Schmidt, S., J. A. Irving, et al. (2006). "Glucocorticoid resistance in two key models of acute lymphoblastic leukemia occurs at the level of the glucocorticoid receptor." Faseb J 20(14): 2600-2. Schmidt, S., J. Rainer, et al. (2006). "Identification of glucocorticoid-response genes in children with acute lymphoblastic leukemia." Blood 107(5): 2061-9. Schmitz, M., P. Breithaupt, et al. (2011). "Xenografts of highly resistant leukemia recapitulate the clonal composition of the leukemogenic compartment." Blood 118(7): 1854-64. Schrappe, M., A. Reiter, et al. (2000). "Long-term results of four consecutive trials in childhood ALL performed by the ALL-BFM study group from 1981 to 1995. Berlin-Frankfurt-Munster." Leukemia 14(12): 2205-22. Schulte, T. W. and L. M. Neckers (1998). "The benzoquinone ansamycin 17- allylamino-17-demethoxygeldanamycin binds to HSP90 and shares important biologic activities with geldanamycin." Cancer Chemother Pharmacol 42(4): 273-9. Schultz, K. R., W. P. Bowman, et al. (2009). "Improved early event-free survival with imatinib in Philadelphia chromosome-positive acute lymphoblastic leukemia: a children's oncology group study." J Clin Oncol 27(31): 5175- 81. Sharma, S. V., D. Y. Lee, et al. (2010). "A chromatin-mediated reversible drug- tolerant state in cancer cell subpopulations." Cell 141(1): 69-80.

Page | 244

Shultz, L. D., B. L. Lyons, et al. (2005). "Human lymphoid and myeloid cell development in NOD/LtSz-scid IL2R gamma null mice engrafted with mobilized human hemopoietic stem cells." J Immunol 174(10): 6477-89. Smith, M. A., C. L. Morton, et al. (2008). "Stage 1 testing and pharmacodynamic evaluation of the HSP90 inhibitor alvespimycin (17-DMAG, KOS-1022) by the pediatric preclinical testing program." Pediatr Blood Cancer 51(1): 34-41. Smyth, G. K. (2004). "Linear models and empirical bayes methods for assessing differential expression in microarray experiments." Stat Appl Genet Mol Biol 3: Article3. Soignet, S. L., P. Maslak, et al. (1998). "Complete remission after treatment of acute promyelocytic leukemia with arsenic trioxide." N Engl J Med 339(19): 1341-8. Solit, D. B., F. F. Zheng, et al. (2002). "17-DMAG induces the degradation of androgen receptor and HER-2/neu and inhibits the growth of prostate cancer xenografts." Clin Cancer Res 8(5): 986-93. Spiegel, A., O. Kollet, et al. (2004). "Unique SDF-1-induced activation of human precursor-B ALL cells as a result of altered CXCR4 expression and signaling." Blood 103(8): 2900-7. Stam, R. W., M. L. Den Boer, et al. (2009). "Association of high-level MCL-1 expression with in vitro and in vivo prednisone resistance in MLL- rearranged infant acute lymphoblastic leukemia." Blood 115(5): 1018-25. Stavreva, D. A., W. G. Muller, et al. (2004). "Rapid glucocorticoid receptor exchange at a promoter is coupled to transcription and regulated by chaperones and proteasomes." Mol Cell Biol 24(7): 2682-97. Storey, J. D. and R. Tibshirani (2003). "Statistical significance for genomewide studies." Proc Natl Acad Sci U S A 100(16): 9440-5. Strasser, A. (2005). "The role of BH3-only proteins in the immune system." Nat Rev Immunol 5(3): 189-200. Subramanian, A., P. Tamayo, et al. (2005). "Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles." Proc Natl Acad Sci U S A 102(43): 15545-50. Supko, J. G., R. L. Hickman, et al. (1995). "Preclinical pharmacologic evaluation of geldanamycin as an antitumor agent." Cancer Chemother Pharmacol 36(4): 305-15. Sutcliffe, M. J., J. J. Shuster, et al. (2005). "High concordance from independent studies by the Children's Cancer Group (CCG) and Pediatric Oncology Group (POG) associating favorable prognosis with combined trisomies 4, 10, and 17 in children with NCI Standard-Risk B-precursor Acute

Page | 245

Lymphoblastic Leukemia: a Children's Oncology Group (COG) initiative." Leukemia 19(5): 734-40. Tabellini, G., P. L. Tazzari, et al. (2005). "Phosphoinositide 3-kinase/Akt inhibition increases arsenic trioxide-induced apoptosis of acute promyelocytic and T-cell leukaemias." Br J Haematol 130(5): 716-25. Thulasi, R., D. V. Harbour, et al. (1993). "Suppression of c-myc is a critical step in glucocorticoid-induced human leukemic cell lysis." J Biol Chem 268(24): 18306-12. Tissing, W. J., M. L. den Boer, et al. (2007). "Genomewide identification of prednisolone-responsive genes in acute lymphoblastic leukemia cells." Blood 109(9): 3929-35. Tissing, W. J., J. P. Meijerink, et al. (2006). "Glucocorticoid-induced glucocorticoid-receptor expression and promoter usage is not linked to glucocorticoid resistance in childhood ALL." Blood 108(3): 1045-9. Tissing, W. J., J. P. Meijerink, et al. (2005). "mRNA expression levels of (co)chaperone molecules of the glucocorticoid receptor are not involved in glucocorticoid resistance in pediatric ALL." Leukemia 19(5): 727-33. Tissing, W. J., J. P. Meijerink, et al. (2005). "Genetic variations in the glucocorticoid receptor gene are not related to glucocorticoid resistance in childhood acute lymphoblastic leukemia." Clin Cancer Res 11(16): 6050-6. Tong, W. G., Z. Estrov, et al. (2011). "The synthetic heat shock protein 90 (Hsp90) inhibitor EC141 induces degradation of Bcr-Abl p190 protein and apoptosis of Ph-positive acute lymphoblastic leukemia cells." Invest New Drugs 29(6): 1206-12. Tonko, M., M. J. Ausserlechner, et al. (2001). "Gene expression profiles of proliferating vs. G1/G0 arrested human leukemia cells suggest a mechanism for glucocorticoid-induced apoptosis." Faseb J 15(3): 693-9. Tracey, E. A., N; Chen, W; Bishop, J (2008). "Cancer in New South Wales: Incidence and Mortality, 2006." Cancer Institute NSW: 106. van Dongen, J. J., T. Seriu, et al. (1998). "Prognostic value of minimal residual disease in acute lymphoblastic leukaemia in childhood." Lancet 352(9142): 1731-8. Vermeer, H., B. I. Hendriks-Stegeman, et al. (2003). "Glucocorticoid-induced increase in lymphocytic FKBP51 messenger ribonucleic acid expression: a potential marker for glucocorticoid sensitivity, potency, and bioavailability." J Clin Endocrinol Metab 88(1): 277-84.

Page | 246

Villunger, A., E. M. Michalak, et al. (2003). "p53- and drug-induced apoptotic responses mediated by BH3-only proteins puma and noxa." Science 302(5647): 1036-8. Walensky, L. D., A. L. Kung, et al. (2004). "Activation of apoptosis in vivo by a hydrocarbon-stapled BH3 helix." Science 305(5689): 1466-70. Wang, Z., M. H. Malone, et al. (2003). "Microarray analysis uncovers the induction of the proapoptotic BH3-only protein Bim in multiple models of glucocorticoid-induced apoptosis." J Biol Chem 278(26): 23861-7. Wasim, M., M. Carlet, et al. (2010). "PLZF/ZBTB16, a glucocorticoid response gene in acute lymphoblastic leukemia, interferes with glucocorticoid- induced apoptosis." J Steroid Biochem Mol Biol 120(4-5): 218-27. Webb, M. S., A. L. Miller, et al. (2003). "Gene networks in glucocorticoid-evoked apoptosis of leukemic cells." J Steroid Biochem Mol Biol 85(2-5): 183-93. Wei, G., D. Twomey, et al. (2006). "Gene expression-based chemical genomics identifies rapamycin as a modulator of MCL1 and glucocorticoid resistance." Cancer Cell 10(4): 331-42. Weigert, O., A. A. Lane, et al. (2012). "Genetic resistance to JAK2 enzymatic inhibitors is overcome by HSP90 inhibition." J Exp Med 209(2): 259-73. Weng, A. P., A. A. Ferrando, et al. (2004). "Activating mutations of NOTCH1 in human T cell acute lymphoblastic leukemia." Science 306(5694): 269-71. Whitesell, L., E. G. Mimnaugh, et al. (1994). "Inhibition of heat shock protein HSP90-pp60v-src heteroprotein complex formation by benzoquinone ansamycins: essential role for stress proteins in oncogenic transformation." Proc Natl Acad Sci U S A 91(18): 8324-8. Wolfson, W. (2009). "Aileron staples peptides." Chem Biol 16(9): 910-2. Workman, P., F. Burrows, et al. (2007). "Drugging the cancer chaperone HSP90: combinatorial therapeutic exploitation of oncogene addiction and tumor stress." Ann N Y Acad Sci 1113: 202-16. Wyllie, A. H., R. G. Morris, et al. (1984). "Chromatin cleavage in apoptosis: association with condensed chromatin morphology and dependence on macromolecular synthesis." J Pathol 142(1): 67-77. Yamamoto, K. R. and B. M. Alberts (1976). "Steroid receptors: elements for modulation of eukaryotic transcription." Annu Rev Biochem 45: 721-46. Yoshida, N. L., T. Miyashita, et al. (2002). "Analysis of gene expression patterns during glucocorticoid-induced apoptosis using oligonucleotide arrays." Biochem Biophys Res Commun 293(4): 1254-61. Zong, W. X., T. Lindsten, et al. (2001). "BH3-only proteins that bind pro-survival Bcl-2 family members fail to induce apoptosis in the absence of Bax and Bak." Genes Dev 15(12): 1481-6. Page | 247