Supplementary Appendix 1 This appendix has been provided by the authors to give readers additional information about their work. Supplement to: Holleman A, Cheok MH, den Boer ML, et al. Gene-expression patterns in drug resistant acute lymphoblastic leukemia cells and response to treatment. N Engl J Med 2004;351:533-42. Table of Content Supplemental Figure 1: Unsupervised hierarchical clustering discriminating drug resistant and drug sensitive ALL patients.........................................................................................5 Supplemental Figure 2: Data analysis flowchart. .....................................................................6 Supplemental Figure 3: Supervised hierarchical clustering discriminating drug resistant and drug sensitive ALL patients (B- and T-lineage ALL). .......................................................11 Supplemental Figure 4: Principal component analysis of drug resistant and sensitive ALL samples for the four antileukemic agents (B- and T-lineage ALL)...................................12 Supplemental Figure 5: Supervised hierarchical clustering discriminating prednisolone resistant and sensitive B-lineage ALL..............................................................................13 Supplemental Figure 6: Supervised hierarchical clustering discriminating vincristine resistant and sensitive B-lineage ALL. ...........................................................................................13 Supplemental Figure 7: Supervised hierarchical clustering discriminating L-asparaginase resistant and sensitive B-lineage ALL..............................................................................14 Supplemental Figure 8: Supervised hierarchical clustering discriminating daunorubicin resistant and sensitive B-lineage ALL..............................................................................14 Supplemental Figure 9: Supervised hierarchical clustering discriminating prednisolone resistant and sensitive ALL (B- and T-lineage ALL). .......................................................15 Supplemental Figure 10: Supervised hierarchical clustering discriminating vincristine resistant and sensitive ALL (B- and T-lineage ALL). .......................................................15 Supplemental Figure 11: Supervised hierarchical clustering discriminating L-asparaginase resistant and sensitive ALL (B- and T-lineage ALL). .......................................................16 Supplemental Figure 12: Supervised hierarchical clustering discriminating daunorubicin resistant and sensitive ALL (B- and T-lineage ALL). .......................................................16 Supplemental Figure 13: Gene Ontology classification of genes discriminating drug resistance in patients with acute lymphoblastic leukemia (B- and T-lineage ALL). .........17 Supplemental Figure 14: Pathway analysis of genes associated with L-asparaginase resistance in B-lineage ALL. ............................................................................................19 Supplemental Table 1: LC50 values for classification of resistant and sensitive ALL for each chemotherapeutic agent. ...................................................................................................4 Supplemental Table 2: Patient characteristics. ........................................................................4 Supplemental Table 3: Permutation analysis and false discovery rate of probe sets selected by Wilcoxon rank sum test and t-test.................................................................................6 Supplemental Table 4: Number of significant probe sets discriminating drug resistance for each individual antileukemic agent. ...................................................................................8 Supplemental Table 5: Prediction accuracy using gene expression profiles for classification of drug resistant and sensitive acute lymphoblastic leukemia. ..........................................9 Supplemental Table 6: Gene expression scores for the intermediate sensitivity group, using genes selected to discriminate resistant and sensitive B-lineage ALL. ...........................10 Supplemental Table 7: Genes previously linked to drug resistance or prognosis in ALL.......18 Supplemental Table 8: Multivariate analysis of gene expression and known prognostic factors (age, WBC count) to discriminate in vitro drug resistance. ..................................20 Supplemental Table 9: Association of the LC50 scores with gene expression scores and other known prognostic variables. ...................................................................................20 Supplemental Table 10: Association of the drug resistance gene expression scores and minimal residual disease (MRD)......................................................................................21 3 Supplemental Table 1: LC50 values for classification of resistant and sensitive ALL for each chemotherapeutic agent. 1,2 LC50 by MTT as described in Pieters and Kaspers et al . Classification based on LC50- values previously associated with treatment outcome, as described in den Boer et al3. Drug Sensistive Resistant Prednisolone ≤ 0.100µg/ml ≥ 150µg/ml Vincristine ≤ 0.391µg/ml ≥ 1.758µg/ml Asparaginase ≤ 0.033IU/ml ≥ 0.912IU/ml Daunorubicin ≤ 0.075µg/ml ≥ 0.114µg/ml Supplemental Table 2: Patient characteristics. Included were 173 patients categorized as resistant or sensitive to one or more of the four drugs. The sample size and the distribution of known prognostic factors (i.e., white blood cell count (WBC) and age at diagnosis; gender, and lineage) for each antileukemic agent are shown for (a) all patients and (b) the B-lineage ALL patients. a. N Age at diagnosis WBC at diagnosis Gender Lineage Drug [median in years] [109/L] Prednisolone Sensitive 66 5.9 36M 30F 43.0 54B 12T Resistant 27 8.0 17M 10F 19.9 20B 7T P=0.007* P=0.497‡ P=0.314* P=0.408‡ Vincristine Sensitive 89 6.0 54M 35F 34.7 75B 14T Resistant 36 6.7 20M 16F 15.4 29B 7T P=0.651* P=0.689‡ P=0.172* P=0.606‡ asparaginase Sensitive 81 4.4 47M 34F 30.0 66B 15T Resistant 45 7.3 24M 21F 28.3 40B 5T P=0.004* P=0.708‡ P=0.729* P=0.320‡ Daunorubicin Sensitive 94 4.6 57M 37F 40.0 82B 12T Resistant 28 7.7 17M 11F 17.6 23B 5T ‡ ‡ P=0.031* P=1.0 P=0.378* P=0.537 b. N Age at diagnosis WBC at diagnosis Gender 9 Drug [median in years] [10 /L] Prednisolone Sensitive 54 5.8 28M 26F 33.1 Resistant 20 8.9 12M 8F 8.3 ‡ P=0.016* P=0.605 P=0.012* Vincristine Sensitive 75 6.0 46M 29F 25.6 Resistant 29 6.3 15M 14F 9.9 ‡ P=0.888* P=0.358 P=0.016* asparaginase Sensitive 66 4.2 38M 28F 23.2 Resistant 40 7.3 21M 19F 18.5 P=0.003* P=0.688‡ P=0.732* Daunorubicin Sensitive 82 4.4 49M 33F 26.8 Resistant 23 7.2 13M 10F 16.3 ‡ P=0.117* P=0.814 P=0.275* M=male, F=female; P-values determined by *Wilcoxon rank sum test ‡Fisher’s exact test Age WBC Subtype GE score 6.5 5.5 7.1 5.0 Supplemental Figure 1: Unsupervised hierarchical 4.2 4.1 4.7 4.5 4.7 4.7 clustering discriminating drug resistant and drug 5.0 5.1 4.0 4.9 5.0 5.3 sensitive ALL patients. 4.5 4.7 4.7 6.2 6.5 5.3 Unsupervised hierarchical clustering of 173 patients was 6.1 5.4 4.6 6.2 6.0 4.8 performed using 14,550 probe sets. Age (red= >10 years), 4.3 4.1 4.7 6.8 4.6 9 5.2 white blood cell count (WBC, red= >100 x10 /L), genetic or 4.3 4.2 4.4 5.9 6.6 5.3 lineage subtype (see legend) and the combined drug 5.9 4.8 5.8 5.3 4.9 4.6 resistance gene expression (GE) score for prednisolone, 4.0 4.6 4.2 4.0 5.3 5.3 vincristine, asparaginase and daunorubicin are shown for 5.4 4.0 4.8 5.0 5.0 4.0 each patient (high/resistant in red= GE score>5.58 and 4.1 4.0 4.4 4.0 4.6 4.5 low/sensitive green= GE score<4.70). Unsupervised 4.1 5.9 5.1 5.1 5.4 4.9 clustering is largely driven by ALL subtype, e.g., T- and B- 4.5 6.0 5.2 5.9 4.4 5.6 lineage ALL. None of the other known prognostic features 5.1 4.9 4.9 4.5 5.1 4.2 (patient age, WBC) clustered in an unsupervised analysis, as 5.1 4.4 5.5 4.3 6.3 5.6 was also the case for the drug resistance gene expression 5.6 4.7 5.7 6.4 5.5 4.4 score. 4.3 4.3 4.5 5.0 6.0 5.8 Legend 6.1 5.2 7.2 6.4 4.8 6.5 Age 4.8 4.7 4.4 >10 years 4.5 6.6 <10 years 5.1 4.9 6.7 4.3 4.6 4.9 WBC 4.3 4.2 9 6.3 >100 x10 /L 6.2 6.3 9 4.1 <100 x10 /L 5.8 5.3 5.9 4.1 4.5 4.1 Subtype 7.3 5.6 5.7 BCRABL 5.9 5.1 E2APBX1 5.1 6.3 6.6 MLL rearranged 7.8 5.7 5.8 T-ALL 6.4 6.6 6.0 TELAM L1 5.9 5.7 7.2 hyperdiploid (HD) 5.2 6.3 6.0 B-ALL other 7.3 6.7 5.8 5.8 4.6 GE score 6.6 6.5 5.4 <4.70 5.1 4.3 4.2 >5.58 5.3 4.7 6.1 4.8 4.0 4.3 4.3 4.8 4.7 4.6 4.2 5.7 4.4 5 Drug sensitivity based on MTT assay COALL/DCOG Gene expression (145 B-lineage, 28 T-ALL) COALL/DCOG (145 B-lineage, 28 T-ALL) PRD, VCR, ASP, DNR 14,550 gene probe sets after initial filtering Subset of resistant and sensitive patients per drug defined by (denBoer et al., JCO 2003) Focus on sensitive and resistant ALL for one drug Rank genes based on P-values obtained from Wilcoxon ranksum test and t-test between Construct a training set by randomly sensitive and resistant patients per drug selecting a group of patients with replacement Select the final gene set per drug, using P- Construct SVM based on the training values and false discovery rate (q-value). set, using the selected gene set PRD (42) VCR (59) ASP (54) DNR (22) identified for the drug (B-lineage ALL) 1000 iterations1000 Predict drug sensitivity or resistance Compute combined drug resistance gene for COALL/DCOG patients not expression scores based on bagging algorithm included in the training set; as well for patients in COALL/DCOG (n=173) as independent test set and independent test set at St Jude (n=98) The mean value of 1000 iterations was the gene expression score for Outcome analysis using combined drug each drug in each patient resistance gene expression scores Supplemental Figure 2: Data analysis flowchart.