Gene Expression Profiling of Acute Promyelocytic Leukaemia Identifies

Gene expression proﬁling of acute promyelocytic leukaemia identiﬁes two subtypes mainly associated with Flt3 mutational status

R Marasca1, R Maffei1, P Zucchini, I Castelli, A Saviola, S Martinelli, A Ferrari, M Fontana, S Ravanetti and G Torelli

Department of Oncology and Hematology, University of Modena and Reggio Emilia, Modena, Italy

Acute promyelocytic leukaemia (APL) is a well-defined disease vary among APL patients and which have been correlated with characterized by a typical morphology of leukaemic cells, the different clinical manifestations of the disease. First, besides presence of t(15;17) translocation and the unique sensitivity to the differentiating effect of all-trans retinoic acid. Nevertheless, cases with the classical morphologic type, characterized by some aspects are variable among APL patients, with differ- blast cells with heavy azurophilic granules, bundles of Auer rods ences substantially related to morphological variants, periph- and reniform or bilobated nuclei, also known as hypergranular eral leukocytes count, the presence of a disseminated form and identified by the FAB cooperative group as M3-AML, intravascular coagulopathy, different PML/RARa isoforms about 15–20% of cases show a hypogranular or microgranular (long, variable or short) and Fms-like tyrosine kinase 3 (Flt3) APL variant form (M3v-AML), in which the majority of blast cells mutations. In order to better define this variability, we have a typical bilobated nucleus and are devoid or contain only investigated the gene expression profiles of 18 APL cases 5 revealing, besides a high uniformity in gene expression pattern, a few cytoplasmic granules. Besides this, other aspects appear the presence of few robust differences among patients able to to vary among APL patients: peripheral white cell count at identify, by an unsupervised analysis, two major clusters of diagnosis is increased in only a minority of patients, mostly patients characterized by different phenotypes (hypogranular related to the M3v subtype and currently considered an adverse M3v vs classical M3) and by the presence or absence of Flt3 prognostic indicator. Moreover, due to the variable breakpoint internal tandem duplications (ITDs). Further supervised analysis confirmed that Flt3 status was the APL parameter best on the PML gene that is involved in the t(15;17) translocation, a associated with these two subgroups. We identified, between long, a variable and a short form of the PML/RARa chimeric Flt3 wild-type and Flt3-ITDs subsets, 147 differentially ex- transcript have been recognized and examined for their pressed genes that were involved in the cytoskeleton organiza- prognostic importance. Although, more relevant, the great tion, in the cell adhesion and migration, in the proliferation and majority of APL patients experience a long-term remission, a the coagulation/inflammation pathways as well as in differen- significant number of them, equal to 20–30%, die due to early tiation and myeloid granules constitution suggesting a role of Flt3 mutations in the pathogenesis and clinical manifestations haemorrhagic complication or to leukaemia relapse. of APL. Although APL cells invariably express aberrant fusion proteins Leukemia (2006) 20, 103–114. doi:10.1038/sj.leu.2404000; involving the RARa gene, mainly joined with the PML gene, all published online 3 November 2005 PML/RARa transgenic mice develop a nonfatal myeloprolifera- Keywords: acute promyelocytic leukaemia; Flt3 mutational status; tive disorder but only a minority of them, about 15%, develop M3 variant morphology; gene expression profiling an APL-like disease after a long latency period, suggesting that additional mutations are required to develop APL.6–8 A candidate gene to play such a role is the Fms-like tyrosine kinase 3 (Flt3) gene, a receptor tyrosine kinase (RTK) belonging Introduction to type III family, along with c-kit, c-fms and PDGFRa and b. Mutations, leading to a constitutive activation of the Flt3 Acute promyelocytic leukaemia (APL) represents approximately receptor kinase, are found in about one-quarter of all AMLs 10–15% of acute myeloid leukaemias (AMLs) and is considered and are considered the most common genetic alterations in a unique disease clearly distinguishable from the other types of human acute leukaemia. In particular, internal tandem duplica- AMLs. In fact, APL is characterized by the presence of a serious tions (ITDs) or point mutation within the activation loop (ALM) haemorrhagic syndrome due to disseminated intravascular of the tyrosine kinase domain 2 (TKD2) of Flt3 seem to be coagulopathy and fibrinolysis at the onset, a promyelocytic-like present at an even higher frequency (35–40%) in human APL morphologic aspect of leukaemic blast cells, a very specific cases.9,10 Interestingly, PML/RARa and Flt3-ITDs have been molecular lesion ever involving the RARa gene and a very shown to cooperate in the induction of an APL-like phenotype in effective therapy represented by all-trans retinoic acid (ATRA) PML/RARa transgenic mice subjected to transplantation with which is able to determine the terminal differentiation of BM cells retrovirally transfected with Flt3-ITD.11 Associations promyelocytic leukaemic cell.1 Recently, gene expression between Flt3 mutations and clinical or molecular characteristics profiling studies have also confirmed a unique expression of APL have been found, and in particular a high white blood pattern in APL cases which can be clearly identified among cell count, M3v morphology and the presence of the short PML/ 2–4 the other AMLs. Nevertheless, there are some aspects which RARa isoform were reported to be related to Flt3-ITDs.12 On the contrary, a prognostic role of Flt3 mutations was not clearly Correspondence: Dr R Marasca, Department of Oncology and defined in APL patients, although an inferior overall survival Hematology, University of Modena and Reggio Emilia, Via Del Pozzo trend was observed.13 71, 41100, Modena, Italy. In this study, we analysed the gene expression profiles of 18 E-mail: [email protected] 1Both authors contributed equally to this work. APL patients using a high density DNA-oligonucleotide micro- Received 13 July 2005; revised 1 September 2005; accepted 21 array (Agilent) representing about 20 000 genes and demon- September 2005; published online 3 November 2005 strated the presence of a very uniform expression pattern in APL Expression profiling and Flt3 status in APL R Marasca et al 104 cases among which however two main APL subtypes were followed by EcoRV digestion and confirmed by direct sequen- recognized by unsupervised analysis. Relationship between cing as reported elsewhere.17 these two subgroups and the different clinical, haematological and molecular APL features, comprising Flt3 status, were investigated. RNA reference Universal Human Reference RNA (Stratagene, Cedar Creek, TX, USA), used as reference control in all microarray gene-profiling Materials and methods experiments, consisted of equal amount of total RNA from 10 human cancer cell lines. Amplified reference cRNA was APL patients and samples labelled with cyanine 3-CTP (Perkin-Elmer, NEN Life Science, In total, 29 patients newly diagnosed with APL and treated Boston, MA, USA) in each experiment. according to the AIDA protocol14 were studied. The median age at diagnosis was 60 years with a range of 22–83 years. Bone Total RNA labelling and microarray hybridization marrow (BM) samples were obtained at diagnosis. APL was For all 18 selected RNAs for gene expression analysis, morphologically diagnosed according to the FAB classification fluorescently labelled cRNA was generated by in vitro transcrip- and confirmed by immunophenotyping, standard cytogenetic tion using Low RNA Input Fluorescent Linear Amplification Kit a and RT-polymerase chain reaction (PCR) detection of PML/RAR (Agilent Technologies, Palo Alto, CA, USA) according to the 15 a fusion transcripts. The different PML/RAR isoforms were also manufacturer’s instructions. Amplified cRNA of each patient determined: 15 cases had bcr1, 13 had bcr3 and one case had was labelled with cyanine 5-CTP (Perkin-Elmer, NEN Life bcr2 breakpoint cluster region. A total of 18 BM samples Science, Boston, MA, USA) in each experiment. Moreover, showing high level of blast cells (at least 90%) and high-quality cRNA products were purified using RNeasy columns (Qiagen). RNA were selected for gene expression analysis. Table 1 Samples had to contain 10–20 picomoles of cyanine dye/mgof summarizes the patients’ characteristics. All samples were cRNA to be considered suitable for subsequent hybridization. enriched for the blast fraction using Lymphoprep (density Cy5-labelled cRNA (1 mg) was mixed with the same amount of 1.077 g/ml Nycomed) density gradient centrifugation. Genomic Cy3-labelled reference cRNA, and then mixed cRNAs were DNA was extracted using Easy-DNA kit (Invitrogen Life fragmented to an average size of B50–100 nt by incubation at Technologies, Carlsbad, CA, USA) according to the manufac- 601C for 30 min using in situ hybridization kit-plus (Agilent). turer’s protocol. Total RNA was isolated using RNeasy Mini Kit Samples were hybridized on Agilent Human 1A Oligo (Qiagen, Valencia, CA, USA) according to the manufacturer’s Microarray (V2), ink-jet printed microarray, comprising 20 173 instructions. (60-mer) experimentally validated oligonucleotide probes (features).18 After hybridization for 17 h at 601C, slides were washed according to Agilent SSPE protocol instructions and then Analysis of the Flt3 status scanned using a confocal laser scanner (Agilent Technologies). To detect Flt3-ITDs, exons 14 and 15 covering the juxtamem- brane domain (JMD) were amplified by PCR as described elsewhere.16 ITDs were detected as an additional longer PCR Data analysis and unsupervised clustering product by 3% agarose gel electrophoresis and confirmed by Fluorescence data were analysed with Feature Extraction direct sequencing. D835 and/or I836 mutations were analyzed Software v7.5 (Agilent Technologies). Log10 ratio of the dye- by PCR amplification of exon 20, corresponding to TKD2, normalized Cy3 and Cy5 channel signals were calculated, then

Table 1 Clinical characteristics of 18 APL patients selected for gene expression analysis

No. Age/sex FLT3 WBC (109/l) Hb (g/dl) PLT (109/l) blast (%) DIC LDH (IU/l) Fibrinogen bcr FAB ATRA (mg/dl) syndrome

1 22/M Neg 1.9 12.0 53 21 No 489 178 1 M3 No 3 57/M ITD 21.2 10.4 21 84 Yes 1516 120 3 M3v Yes 4 34/M Neg 1.3 10.5 7 39 Yes 638 85 3 M3 No 8 26/M ITD 18.4 12.7 17 86 Yes 1581 85 3 M3v No 9 60/M Asp835 2.8 6.5 21 80 Yes 621 139 1 M3v No 10 25/F Neg 2.8 10.6 23 36 Yes 594 83 1 M3 No 11 68/M Neg 1.0 9.0 4 10 Yes 441 346 1 M3 No 12 65/F Neg 0.8 13.1 128 0 No 342 205 1 M3 No 13 72/M Neg 3.8 10.5 15 61 No 957 323 2 M3 No 16 70/M ITD 21.2 7.9 17 83 No 415 344 3 M3v No 18 61/M ITD 0.9 4.6 4 85 Yes 490 114 1 M3v No 19 56/F Neg 18.1 11.4 18 94 No NA 152 3 M3 No 21 82/F ITD, Asp835 12.3 10.4 15 28 Yes 711 387 1 M3 Yes 22 66/F Asp835, Iso836 1.3 8.0 6 29 Yes 623 337 1 M3v No 23 71/F ITD 10.1 7.4 8 90 Yes 1807 148 3 M3v No 26 64/F Asp835 14.0 11.5 47 74 Yes 1777 95 3 M3v Yes 27 83/M ITD 12.7 5.5 14 78 Yes 882 136 3 M3v No 28 34/F ITD 65.6 11.3 17 86 Yes 392 94 3 M3v No WBC: white blood count; Hb: haemoglobin; PLT: platelets; blast: circulant blast cells; DIC: disseminated intravascular coagulation; LDH: lactate dehydrogenase; bcr: breakpoint cluster region; FAB: French–American–British classiﬁcation; ATRA: all-trans retinoic acid; NA: not available; Neg: FLT3 wild type; ITD: internal tandem duplication; Asp835, Iso836: point mutations.

Leukemia Expression profiling and Flt3 status in APL R Marasca et al 105 a P-value and a final error were associated to establish the Statistical analysis significance of each feature. The data discussed in this publica- Data were analysed using SPSS for Windows Version 11.0. tion have been deposited in NCBIs Gene Expression Omnibus (SPSS Inc, Chicago, IL, USA). To determine significant differ- (GEO, http://www.ncbi.nlm.nih.gov/geo/) and are accessible ences in binary variables, the Pearson w2 test or the Fisher exact through GEO Series accession number GSE2550. Raw signal test were used. Furthermore, group-wise comparisons of intensities from each scan were imported into the gene expression distributions of continuous variables were performed with analysis software Luminators (Rosetta Biosoftware, Seattle, WA, Mann–Whitney nonparametric test. All were two-sided. An USA). A complete description of the statistical methods used is effect was considered statistically significant at Pp0.05. available in the technology section of the Rosetta Biosoftware website (http://www.rosettabio.com/tech/default.htm). Two-dimensional clustering analysis was performed with Rosetta Results Luminators using an agglomerative algorithm with an average 19 link heuristics and a correlation with mean subtraction. Analysis of Flt3 mutation status We assessed the Flt3 mutation status in 29 adult patients with Supervised clustering APL at the time of diagnosis. DNAs obtained from untreated BM Supervised classification of APL samples into categories samples were screened for the presence of Flt3-ITDs and for according to mutational status of Flt3 gene based on gene mutations in the activation loop (ALM) of TKD2. Overall, 12 expression profiles was performed using Bayesian classifier APL patients (41.4%) carried Flt3 mutations (ITD and/or implemented by Rosetta Biosoftware. This method includes the Asp835), whereas 17 APL patients (58.6%) showed Flt3 gene following three steps: (1) selection of candidate genes that are in a wild-type configuration (Flt3-WT group). Seven out of 29 significantly different among relevant classes based on t-tests. APL patients (24.1%) had only ITDs length mutations (Flt3-ITD Those features showing the larger t-values tend to carry more group). The size of length mutations varied from 18 to 93 bp discriminating power than others. The features are selected with (median 46.9 bp) and the reading frame was always maintained. |t|X1; (2) probability computation, performed using a set of As shown in Table 2, Flt3-ITD group was significantly associated discriminant functions (d) derived from the posterior probabil- with a higher white blood cell count (WBC) (P ¼ 0.009) as well ities of the classes. These are used to estimate the relevant as with a higher circulating blast cell percentage (P ¼ 0.021). probability from the training set; (3) classification and classifier Moreover, the presence of ITDs in Flt3 gene was related to validation, in which each observation to be classified is assigned microgranular morphology (M3v) (Po0.0001) and short-type to a class if its discriminative function value (d ), given by the PML/RARa isoform (bcr3) (P ¼ 0.042), as previously de- i i 16,22,23 selected features, is higher than the one obtained for the second scribed. Flt3 activation loop mutations were present in class. To estimate the quality of the classifier, the member of five of 29 (17.2%) APL patients (Flt3-ALM group). We found the each training set was also reclassified (self-validation) and a presence of a missense mutation at codon 835 in four cases and leave-one-out crossvalidation was then performed.20 Gene a six-base pair deletion together with a missense mutation at expression data were also analyzed using analysis of variance codon 836 in one case. Only one patient (case 21) harboured (ANOVA) implemented by Rosetta Biosoftware. To increase the both ITDs and D835 mutation. statistical power, the enhanced ANOVA uses as input data both expression level and the estimated technology error associated with the expression level. As a result, the false-positive rate is Identification of APL subsets by unsupervised reduced when the number of replicates is small and, then, hierarchical clustering sensitivity detection is increased. Our original m-dimensional A total of 18 APL patients selected from our series for the high data patterns were transformed to data patterns in a feature quality of total RNA were hybridized on Agilent microarray space with lower dimensionality using Principal Component containing approximately 20 000 human genes. Flt3-ITDs and Analysis (PCA).21 Flt3-ALMs were present in seven and three cases, respectively,

Table 2 Comparisons of clinical parameters between FLT3-ITD and FLT3-WT groups

FLT3-ITD group (n ¼ 7)a FLT3-WT group (n ¼ 17)a P-value

Male/female (n) 5/2 7/10 0.334 Age at diagnosisb 61 (26–83) 56 (22–79) 0.611 FAB M3 microgranular variant 7 (100%) 3 (17.6%) o0.0001 DIC 6 (85.7%) 8 (47.1%) 0.172 t(15;17) breakpoint cluster region Bcr1 1 (14.3%) 11 (64.7%) Bcr2 0 (0%) 1 (5.9%) 0.042 Bcr3 6 (85.7%) 5 (29.4%)

Hyperleukocytosis (41010/l) 6 (85.7%) 4 (23.5%) 0.009 Circulant blast cells % (range) 85 (78–90) 44 (0–97) 0.021 Hb (g/dl)b 7.9 (5–13) 10.6 (5–13) 0.182 PLT (109/l)b 17 (4–21) 18 (4–223) 0.143 LDH (IU/l)b 882 (392–1807) 485 (265–1228) 0.124 Fibrinogen (mg/dl)b 120 (85–344) 177 (83–376) 0.099 aNumber of cases (% inside group). bMedian (range).

Leukemia Expression profiling and Flt3 status in APL R Marasca et al 106 one patient had both Flt3 alterations and the remaining seven Moreover, we identified the differentially expressed genes cases had Flt3-WT. able to characterize and thus distinguish APL Clusters I and II To identify biologically meaningful subsets of APL, resultant patients (Po0.01). Among 1477 genes found, 382 (26%) were gene expression profiles were first analysed using an unsuper- common to ITD signature (80% of ITD signature). On the other vised, agglomerative hierarchical clustering with average link as hand, only 14% (216 genes) and 9% (136 genes) of differentially heuristic criteria and correlation with mean subtraction as expressed genes were influenced by PML/RARa isoforms and by similarity metric. Before hierarchical clustering computation, M3/M3v morphology, respectively. Furthermore, it is remark- some filters were applied on the data: only those genes showing able that when the analysis was restricted to the 349 genes with a fold change greater than 3 and a P-valuep0.001 in at least a higher statistical significance (Po0.0001) among differentially one-third of the samples were selected. The two-dimensional expressed genes in APL Clusters I and II, 171 genes (49%) were matrix clearly showed the high grade of uniformity in gene common to the ITD signature, and only 91 and 57 genes were expression pattern among APL patients, the majority of genes related, respectively, to the bcr and M3v signatures (Figure 2b). being expressed at high (red colour) or low (green colour) levels relative to reference in all APL cases analysed (Figure 1a). Concordantly, this uniformity was revealed by the high Altered expression of 147 genes mainly differentiates correlation coefficient (Pearson correlation coefficient range: Flt3-ITD from Flt3-WT patients 0.90–0.98) measured between all the samples examined. This Data analysis suggested Flt3-ITDs gene as the main source of appearance of unsupervised matrix reflects the homogeneous gene expression variability in our APL patients. To better identify nature of APL whose gene expression pattern definitely those genes distinguishing the Flt3-ITD leukaemic cells from distinguishes it from all other types of acute myeloid leukaemia, Flt3-WT, we applied a Bayesian classifier. The Bayesian as assessed previously.2 Besides the homogeneous pattern, algorithm was carried out defining two training classes, one subtle robust differences in gene expression enabled the consisted of all Flt3-ITD samples and the other contained all but identification of three clusters of APL patients (designed I, II, one Flt3-WT cases (case 19 was excluded). First of all, 147 III) (Figure 1b). Clusters I and II consisted of nine and seven genes, significantly different among classes based on t-tests, cases, respectively, whereas Cluster III had only two samples. were selected and included into the classifier. Among them, 92 Comparisons between Cluster I and Cluster II patients clearly genes were upregulated in Flt3-ITD class and 55 were down- revealed a preferential distribution of Flt3 gene-mutated cases in regulated. When the 147-gene list was applied to hierarchical Cluster I (Table 3). The majority of Cluster I patients (88.9%) had clustering of the 14 Flt3-ITD and -WT specimens, all samples mutated Flt3 gene, instead Cluster II patients mainly expressed except case 19 were correctly clustered (Figure 3a). Multi- Flt3 genes in wild-type configuration (71.4%). In particular, all dimensional visualization by PCA was also used to observe the Flt3-ITD patients clustered tightly in the same leftmost side (I), spatial distribution of sample clusters in multidimensional and all but one (case 19) Flt3-WT patients clustered in the right- space. As shown in Figure 3b, the visualization of high- hand side of the dendrogram (II and III) (P ¼ 0.007). Three out of dimensional data in three-dimensional principal components four Flt3-ALM cases were localized on the right branch of the reveals two major distant clusters, Flt3-ITD and Flt3-WT. dendrogram closer to Flt3-WT cases. Again, Cluster I was related Interestingly, two cases among Flt3-ITD patients appeared to microgranular morphology (M3v) (P ¼ 0.035) and short-type nearer to Flt3-WT cluster than others. A partial list of PML/RARa isoform (bcr3) (P ¼ 0.044). In particular, 71.4% differentially expressed genes rank-ordered on the basis of main (n ¼ 5) of Cluster II had a classic APL morphology in contrast functions is shown in Table 4. A full list is included in GEO web with only 11.1% of patients in Cluster I. Moreover, most cases site (Supplement 7). Second, the Bayesian classifier developed a harbouring t(15;17) breakpoint cluster region 3 belonged to discriminative function based on the most different genes Cluster I (80% of all bcr3 cases), whereas 80% of bcr1 belonged between Flt3-ITD and -WT classes. The performance of the to Cluster II. Furthermore, as shown in Table 3, Cluster I was constructed classifier was assessed by leave-one-out cross- significantly associated with higher circulant blast cell percen- validation testing method. Briefly, each sample is excluded from tage (P ¼ 0.030) as well as hyperleukocytosis (P ¼ 0.009). the training set one at a time and then classified on the basis of the predictor built from the data for all of the other samples. All the steps of the Bayesian classifier method underwent the cross- Gene expression variability between APL Clusters I and validation procedure.24 The 147-gene classifier assigned 12 of II is mainly associated to Flt3-ITDs the 13 samples into the correct class (accuracy ¼ 92%). The In order to identify the principal sources of gene expression fraction of incorrectly classified training observations (error rate) variability within the two main APL subsets identified by was 0.077. unsupervised clustering, we extrapolated, applying ANOVA algorithm (excluding Cluster III cases), those genes (Po0.01). able to distinguish hypogranular from hypergranular morpholo- Characterization of differentially expressed genes gical APL subsets (M3v signature), cases harbouring Flt3-ITDs between Flt3-ITD and -WT APL subsets from the others (ITD signature) and cases with different PML/ Flt3 ITDs gene expression signature consisted of several high- RARa isoforms (bcr signature). Overall, altered expression of expressed genes, among them we found a group of genes 477 and 356 genes determined the variability linked to the two involved in the inflammation and coagulation pathways (Blood genetic features (Flt3-ITD and PML/RARa isoforms, respec- Coagulation gene cluster) (Table 4).25–31 In particular, genes tively), whereas only 229 genes characterized the APL involved in vasodilatation and vascular permeability as PTGIR, morphological phenotypes. Of note, 100 genes (44% of M3v NPR1 were up-regulated in Flt3-ITDs subset. We also identified signature) were shared between M3v morphology and Flt3-ITD CFH, HFL1 and FHR-3, belonging to human Factor H protein signature, but only 37 and 25 differentially expressed genes of family, which are involved in the inhibition of complement bcr signature were shared with ITD and M3v signatures (10 and activation. Their overexpression in certain tumour cells such as 7% of bcr signature, respectively) (Figure 2a). Full ANOVA gene glioblastoma, ovarian tumour and non-small lung cancer lists are included in GEO web site (Supplements 1–6). confers an exceptional resistance to such cells against comple-

Leukemia Expression proﬁling and Flt3 status in APL R Marasca et al 107 a

Log (Ratio) b III

III Case#4 Flt3-WT Case#1 Flt3-WT Case#3 Flt3-ITD Case#8 Flt3-ITD Case#9 Flt3-ALM Case#19 Flt3-WT Case#13 Flt3-WT Case#10 Flt3-WT Case#11 Flt3-WT Case#12 Flt3-WT Case#28 Flt3-ITD Case#16 Flt3-ITD Case#23 Flt3-ITD Case#18 Flt3-ITD Case#27 Flt3-ITD Case#26 Flt3-ALM Case#22 Flt3-ALM Case#21 Flt3-ITD-ALM

Figure 1 Global hierarchical clustering of APL samples based on gene expression proﬁling. (a) Thumbnail overview of the two-way (genes against samples) hierarchical clustering of 18 samples. Data are shown in a matrix format where each row represents a gene and each column represents a sample. Ratios of gene expression are depicted by a log-transformed (on a base 10 scale) pseudo-colour scale. Green squares are low transcript levels in the sample relative to Universal Reference; red squares are high transcript levels; black squares are transcript levels equal to the reference; grey squares are technically inadequate or missing data. (b) Enlarged view of the sample dendrogram. Samples are coloured according to FLT3 gene status. Two major clusters of samples (designed I, II) are identiﬁed. A third small cluster with only two sample (designed III) is shown.

Leukemia Expression proﬁling and Flt3 status in APL R Marasca et al 108 Table 3 Comparisons of clinical parameters between APL Cluster I and Cluster II deﬁned by unsupervised hierarchical analysis

Cluster I (n ¼ 9)a Cluster II (n ¼ 7)a P-value

Male/female (n) 5/4 5/2 0.633 Age at diagnosisb 61 (26–83) 60 (22–79) 0.681

FLT3 status Wild type 1 (11.1%) 5 (71.4%) Internal tandem duplication 7 (77.8%) 0 (0%) 0.007 D835 1 (11.1%) 2 (28.6%)

FAB M3 microgranular variant 8 (88.9%) 2 (28.6%) 0.035 DIC 7 (77.8%) 5 (71.4%) 0.771

t(15;17) breakpoint cluster region Bcr1 1 (11.1%) 4 (57.1%) Bcr2 0 (0%) 1 (14.3%) 0.044 Bcr3 8 (88.9%) 2 (28.6%)

Hyperleukocytosis (41010/l) 8 (88.9%) 1 (14.3%) 0.009 Circulant blast cells, % (range) 85 (74–94) 39 (10–97) 0.030 Hb (g/dl)b 10.4 (4.6–12.7) 10.5 (4.6–12) 0.837 PLT (109/l)b 17 (4–47) 15 (4–53) 0.918 LDH (IU/l)b 1199 (392–1807) 621 (441–1228) 0.418 Fibrinogen (mg/dl)b 120 (85–344) 178 (83–346) 0.408 aNumber of cases (% inside group). bMedian (range).

ment-mediated lysis. Furthermore, we found high expression of specific gene expression patterns of acute leukaemias. In genes involved in coagulation pathway such as PTX3, which particular, AML with specific reciprocal rearrangements can be upregulates tissue factor expression in endothelial cells, H963, recognized on the basis of different gene expression patterns. platelet-activating receptor homolog and MMRN1. Another Interestingly, APL cases, characterized by t(15;17) chromosome group of genes overexpressed in APL with internal duplications translocation, have a highly homogeneous expression pattern, in the Flt3 gene with respect to the wild-type Flt3 were related to whereas cases with t(8;21) or inv(16) are less well correlated.2,3 proliferation and cell adhesion/invasiveness (Hyperleukocytosis Despite the fact that APL is a well-characterized disease, with a gene cluster) (Table 4).32–34 Briefly, we found genes for well-defined pathogenetic and clinical picture as well as an structural constituents of the cytoskeleton and nuclear lamina established therapeutic strategy based on ATRA combined with as VIM, TUBA6, tubulin b and LMNA and genes involved in chemotherapy, several aspects of variability characterize the cytoskeletal remodelling and cell adhesion/migration, usually in disease at a phenotypic or at a haematological level, that is, M3 association with the inflammatory process, such as CAPN2, or M3v morphology, cytopenia or leukocytosis, intensity of AMIGO2 and MMP19. In particular, S100A10 and S100A4, disseminated intravascular coagulation (DIC) and platelet count. genes of S100 Ca þ binding family, play an important role in Of importance, initial leukocytes and platelet counts are plasmin production, which is involved in both the stimulation of considered important factors influencing the relapse-free survival inflammation and in cancer invasiveness and metastasis. More- allowing to stratify APL patients into different prognostic over, we identified a group of genes inducing cell proliferation groups.45 Moreover, this variability is also observed at a (LDLR, LRP8, HMGCS1, CRIP1, hCRHP, LRRFIP2, SFRP4).35–39 molecular level, substantially represented by the presence of In particular, elevated levels of low-density lipoprotein receptors different PML breakpoints with different PML/RARa transcripts (LDLR and LRP8) were reported in proliferating normal cells and and by the presence of Flt3 mutations. Interestingly, Flt3-ITDs tumours. Finally, we found several genes involved in embry- appear to be frequently associated to leukocytosis, M3v form and ogenesis (MEF2C, BAPX1, GLI2, HLX1, PMP22, EMP1, to the presence of the short PML/RARa isoform (bcr3).13,16,22,46 TAGLN2) (Table 4). On the other hand, a group of 55 genes We decided to examine the gene expression profiles by were found to be downregulated in leukaemic populations microarray containing approximately 20 000 human genes of 18 harbouring Flt3-ITDs with respect to cases with Flt3 in the wild- t(15;17) APL cases looking for correlation with clinical, type configuration. Among these genes, we identified BPI, CTSF, morphological, and molecular features. As expected, APL CAMP, EPX, LTF and LCN2 coding antimicrobial proteins demonstrated a homogeneous pattern of gene expression contained within granules40 as well as a group of genes (Figure 1a) characterized by a very high correlation coefficient involved in the control of differentiation such as NELL2,41 (Pearson correlation coefficient range ¼ 0.90–0.98) among all CREG,42 PPARG,43 TLE144 and five genes expressed in the samples examined. Nevertheless, robust differences in the erythrocytes (RHD, RHCE, SPTA1, GYPC, GYPA) (Table 4 and expression levels of some genes enabled the division of APL Supplement 7). patients into two major and one minor cluster with hierarchical unsupervised analysis. Of relevance, the two principal clusters of APL patients identified by two-dimensional unsupervised Discussion clustering were characterized by a preferential distribution of the haematological and molecular parameters examined and, in Distinct features defined by cytomorphology, immunopheno- particular, Cluster I was mainly represented by cases with M3v type, cytogenetics and molecular genetics parameters identify morphology, high leukocyte count, bcr3 PML/RARa (short type)

Leukemia Expression profiling and Flt3 status in APL R Marasca et al 109 23 a features was reported. This is therefore consistent with the presence of an APL subtype, clearly identifiable with unsupervised gene expression analysis, corresponding substantially to a 352 M3v form with Flt3-ITDs. Nevertheless, some exceptions in 306 cluster constitution were present. Case 19, a classic M3 APL BCR FLT3-ITD patient with Flt3 in a wild-type configuration, but characterized signature 25 signature by a leukocyte count higher than 10.000 Â 106/l and a bcr3 (356 genes) (477 genes) PML/RARa isoform, cosegregated into Cluster I among Flt3-ITD/ 12 M3v samples suggesting the presence of some unknown factors determining a very similar expression pattern to Flt3-ITDs cases, 13 88 that is, the presence of an undetectable Flt3-activating mutation as well as an alteration of a regulatory factor downstream of or linked to the Flt3 pathway. Moreover, two M3v cases, but Flt3- 116 M3v WT, cosegregated into Cluster II among classic APL, suggesting signature that the main feature determining gene expression patterns (229 genes) could be Flt3 gene status independent of morphological variability. In order to examine the causal factors involved in gene b M3v 2% expression variability in APL, ANOVA analysis was performed. It clearly indicated that almost half genes differentially BCR expressed between the two main clusters (I and II) were related 21% unknown 28% to the Flt3 status, especially when strong statistical criteria were applied. In fact, among 396 genes that better distinguished the two APL clusters identified by unsupervised analysis, we found 171 genes (49%) of Flt3-ITD signature, and only a few genes of M3v and bcr signatures, equal to 16 and 26%, respectively. ITD-BCR-M3v Very recently, Haferlach et al47 published the gene expression 3% analysis by microarray of a cohort of APL patients identifying, by ITD-BCR a supervised analysis, a group of genes able to distinguish M3 2% from M3v cases. They also found that genes that better discriminate M3 from M3v are not substantially related to ITD-M3v Flt3-ITD. Our results are substantially in agreement with these 11% observations. In fact, as shown in Figure 2a, only a minority of genes whose expression correlates with morphology or Flt3 status belongs to both groups, suggesting that morphology and ITD FLT3 status, may, at least partly, affect gene expression 33% independently. Figure 2 Visualization of differentially expressed genes characteriz- As FLT3-ITD plays an important role in APL gene expression ing M3 morphology, FLT3-ITD and PML/RARa isoforms. (a) ANOVA profile and is supposed to play a pathogenetic role in acute algorithm was applied on 16 samples which belonged to Clusters I and leukaemia, we identified those genes that are affected by Flt3- II in order to identify genes differentially expressed among distinct APL morphological and genetic subsets (M3v, FLT-ITD and BCR). 477, 356 ITD status. A Bayesian classifier was constructed with a panel of and 229 genes formed the signatures of the two genetic features (FLT3- 147 differentially expressed genes, and in particular 92 genes ITD and BCR, respectively) and the morphological phenotypes with were upregulated and 55 were downregulated in Flt3-ITDs cases Po0.01. Intersection between M3 morphology and FLT3-ITD circles (Figure 3). Among the differentially expressed genes, some was composed of 100 genes, whereas only 37 and 25 differentially homogeneous groups were identified involving specific cellular expressed genes in BCR signature were shared with other groups of functions that could explain the phenotypic and haematological genes characterizing FLT3-ITD and M3 morphology. (b) The graph shows the percentage of genes belonging to ITD, M3v and BCR differences in APL patients correlated to the presence of Flt3- signatures common to the 349 genes differentially expressed between ITDs. In particular, we identified a group of genes, indicated as Clusters I and II with a higher statistical significance (Po0.0001). hyperleukocytosis gene cluster, related to cell adhesion and Intersections between different sets are also shown. invasiveness that could explain the observed correlation between Flt3-ITDs and leukocytosis. More specifically, it included genes involved in cytoskeleton organization, cell isoform and Flt3-ITDs, whereas Cluster II was mainly constituted adhesion as well as cancer invasiveness and metastasis. More- of APL cases with classical morphology, bcr1 (long form) PML/ over, high expression of several genes involved in the RARa transcript, leukopenia and Flt3-WT. Remarkably, all Flt3- inflammation/coagulation process (blood coagulation gene ITDs cases were grouped into the Cluster I with the exception of cluster) was found, despite the absence of statistical correlation case 21, belonging to Cluster III, in which Flt3-ITD was with the presence of DIC in our cohort. In addition, we also associated with Flt3-ALM. On the contrary, the three Flt3-ALM identified several genes involved in embryogenesis and cell cases were distributed over the three identified clusters, growth, which were upregulated in Flt3-ITDs APL cases. On the suggesting that, with the limitation of the small number of other hand, among downregulated genes in Flt3-ITDs cases, we Flt3-AML cases examined, this type of alteration does not found genes expressed along the hematopoietic differentiation consistently influence APL gene expression pattern. On the pathway, encoding for proteins present in granulocytic granules. other hand, this type of Flt3 alteration does not seem to It was perfectly concordant with the observed correlation influence phenotypic presentations of APL because no associa- existing between Flt3-ITDs and the hypogranular variant form tion with M3v or other phenotypic, clinical or molecular of APL. Acute promyelocytic leukaemia is characterized by

Leukemia Expression proﬁling and Flt3 status in APL R Marasca et al 110 a

Flt3-ITD signature Flt3-WT signature

PTGIR LRRFIP2 HFL1 NPR1 HF1 FHR 3 LRP8 CD97 S100A10 S100A4

MMP19

SFRP4

LDLR TLE1 STAR CTSF LTF CREG

NELL2 BPI HEMGN

CAMP LCN2 PPARG EPX Flt3-WT samples Flt3-ITD samples b

− 0.15

− 0.20

− 0.25 PC1

− 0.30

− 0.35

−0.40 −0.30 − 0.20 −0.10 PC3 0.00 0.10 0.20 PC2

Figure 3 Hierarchical clustering and PCA visualization of FLT3-ITD versus FLT3-WT. (a) The hierarchical clustering was performed using 147 genes differentially expressed between FLT3-ITD and FLT3-WT cases deﬁned by supervised Bayesian classiﬁer method. In total, 92 genes were upregulated in FLT3-ITD class and 55 were downregulated. All samples except case#19WT were correctly clustered. (b) Three-dimensional projection of the three principal components in a principal component analysis of all FLT3-ITD and -WT samples, with the use of the 147 variably expressed genes. The two different subsets are coloured as indicated.

Leukemia Expression proﬁling and Flt3 status in APL R Marasca et al 111 Table 4 Statistically different genes between APL FLT3-ITD and APL FLT3-WT subsets

Gene symbol Gene name Gene function (reference)

Upregulated genes in FLT3-ITD vs FLT3-WT Blood coagulation/inflammation CD97 CD97 antigen Leukocyte migration, role associated with establishment or development of inflammatory process25 ADAM17 TNF-alpha convertase Metalloendopeptidase activity, it cleaves the membrane-bound precursor of TNF-alpha to its mature soluble form. Inflammation26 PTGIR Prostaglandin I2 receptor It mediates the vasodilator, antithrombotic and antiplatelet function of prostacyclin27 NPR1 Atrionatriuretic-peptide receptor A Regulator of vasodilatation, vascular permeability and blood pressure CFH, HFL1, FHR-3 H factor 1, H factor-like 1, H factor- Complement regulatory functions. It controls formation and stability of related 3 C3 convertases, conferring resistance to glioblastoma, ovarian and lung cancer against complement lysis. Adhesion activity28 PTX3 Pentraxin 3 It increases tissue factor expression and plays a role in thrombogenesis and ischemic vascular disease29 H963 Platelet-activating receptor homolog Procoagulant activity30 MMRN1 Multimerin 1 Blood coagulation, it binds the coagulation protein factor V and its activated form, factor Va31

Cytoskeleton/cell adhesion/metastasis LGALS1, LGALS2 Galectin 1, 2 Cell–matrix adhesion, resistance to apoptosis and angiogenesis, cancer progression,32 tumour evasion of immune responses. Inflammatory mediators33 VIM Vimentin Internediate-filament protein, reorganization of the cytoskeleton TUBA6 Tubulin alpha 6 Constituent of microtubules TUBULIN b Tubulin beta Constituent of microtubules LMNA Laminin A/C Intermediate-filament protein, chromatin reorganization CAPN2 Calpain 2 Critical regulator of cell migration and of the organization of the actin cytoskeleton and focal adhesions AMIGO2 Amphoterin-induced gene 2 Cell adhesion/migration MMP19 Matrix metalloproteinase 19 Endopeptidase that degrades various components of the extracellular matrix, involved in invasion and metastasis S100A10, S100A4 S100 calcium-binding protein A10, A4 Plasmin production, extracellular matrix degradation, cancer invasiveness and metastasis34

Proliferation/cell growth LDLR, LRP8 Low-density lipoprotein receptor, It binds LDL, the major cholesterol-carrying lipoprotein of plasma, and lipoprotein receptor-related protein 8 transports it into cells by endocytosis. Elevated LDL-receptor activity is associated with sterol resistance and cell proliferation in AML35,36 HMGCS1 3-Hydroxy-3-methylglutaryl-coenzyme It condenses acetyl-CoA with acetoacetyl-CoA to form HMG-CoA, A synthase 1 which is the substrate for HMG-CoA reductase. Cholesterol biosynthesis. AML was shown to require abnormally high levels of cholesterol for their survival37 CRIP1, HCRHP Cystein-rich protein 1 It seems to have a role in zinc absorption and may function as an Cystein-rich heart protein intracellular zinc transport protein. Involved in cell proliferation LRRFIP2 Leucine-rich repeat (in FLII) interacting LRRFIP2 plays an important role in transducing Wnt signals38 protein 2 SFRP4 Secreted frizzled-related protein 4 It functions as a regulator of the Wnt-frizzled signalling pathway. It binds directly to Wnts, thereby altering their ability to bind to the Wnt receptor complex. sFRPs can also potentiate Wnt activity39

Embryogenesis MEF2C MADS box transcription enhancer- It may be involved in myogenesis, neurogenesis and in the factor 2 development of cortical architecture, MAPK signalling pathway BAPX1 Bagpipe homeobox homolog 1 Transcription factor activity. It controls development of axial skeleton, skull, and spleen GLI2 GLI-Kruppel family member Transcription factor activity. It may play a role during embryogenesis HLX1 H2.0 (Drosophila)-like homeo box 1 Transcription factor activity. Contains 1 homeobox domain. It’s a putative transcription factor involved in embryogenesis and hematopoiesis PMP22 Peripheral myelin protein 22 It might be involved in growth regulation, and in myelinization in the peripheral nervous system EMP1 Epithelial membrane protein 1 Epidermis development. Cell growth. TAGLN2 Transgelin 2 Muscle development

Downregulated genes in FLT3-ITD vs FLT3-WT Proteins of granules BPI Bactericidal/permeability-increasing Defense response to bacteria. Azurophil granules40 protein CTSF Cathepsin F Azurophil granules.

Leukemia Expression proﬁling and Flt3 status in APL R Marasca et al 112 Table 4 Continued

Gene symbol Gene name Gene function (reference) CAMP Cathelicidin It binds to bacterial lipopolysaccharides (LPS), antibacterial activity EPX Eosinophil peroxidase Eosinophil granule proteins LTF Lactoferrin Antimicrobial activity. Speciﬁc granules LCN2 Lipocalin 2 Speciﬁc granules

Differentiation NELL2 NEL-like 2 Haemopoietic cells development41 CREG Cellular repressor of E1A-stimutated It promotes differentiation42 genes PPARG Peroxisome proliferative activated Potent inducers of differentiation in combination with retinoids43 receptor TLE1 Transducin-like enhancer of split 1 Corepressor of RUNX genes and LEF-144

uncontrolled proliferation, antiapoptotic advantages, differentia- elements represented by its association with several clinical or tion failure and inflammation/coagulation stimulation. In haematologic features, by data deriving from transgenic animal particular, APL patients, who presented internal duplications APL models as well as by experiments indicating a disrupted of the Flt3 gene, seemed characterized by an accentuated control of differentiation and proliferation in Flt3-ITD myeloid proliferative promptness and by a much more undifferentiated leukaemia cell lines. In this paper, the reported gene expression phenotype. Of note, Flt3-ITD mediates cell growth and survival, profile analysis gives further support to this hypothesis also blocks myeloid differentiation and induces leukaemic transfor- identifying several interesting genes whose expression levels are mation in hematopoietic progenitor cell lines and primary significantly different on the basis of Flt3 status. mouse BM.48–50 Recently, a pivotal role of Flt3-ITDs constitutive activation was demonstrated in determining a blockage of myeloid differentiation in 32Dcl3 (32D) myeloblastic cell line.51 Acknowledgements The mechanism inducing such differentiation failure in Flt3- ITDs transduced 32D cells involves suppression of C/EBPa and This work was supported by grants from Associazione Italiana per PU.1 expression, which are essential genes needed for the full la Ricerca sul Cancro (AIRC), Milan, Italy; Associazione Italiana commitment and differentiation of granulocyte lineage.52,53 contro le Leucemie (AIL), Modena, Italy; Cassa di Risparmio of Moreover, Pim-1 and Pim-2 upregulation was suggested to Modena Foundation, Modena, Italy. contribute to proliferative and antiapoptotic pathways induced by Flt3-ITDs signalling.53,54 Both C/EBPa, PU.1 downregulation and Pim-1, Pim-2 upregulation were not found in our Flt3-ITDs References patients suggesting that other differentiation blockages and proliferative mechanisms must be hypothesized in Flt3-ITD APL 1 Melnick A, Licht JD. Deconstructing a disease: RARa, its fusion in vivo. We can not obviously exclude an alteration of their partner, and their roles in the pathogenesis of acute promyelocytic functions at protein levels. Interestingly, here we found that the leukemia. Blood 1999; 93: 3167–3215. expression of crucial genes involved in hematopoietic differ- 2 Bullinger L, Dohner K, Bair E, Frohling S, Schlenk RF, Tibshirani R entiation such as PPARG and TLE1 was suppressed in Flt3-ITD et al. Use of gene-expression profiling to identify prognostic leukaemic cells. In particular, TLE1/groucho is a crucial subclasses in adult acute myeloid leukemia. N Engl J Med 2004; 350: 1605–1616. corepressor of RUNX genes and of Wnt signalling path- 3 Schoch C, Kohlmann A, Schnittger S, Brors B, Dugas M, 44,55,56 way. Remarkably, RUNX1/AML1 gene (runt-related Mergenthaler S et al. Acute myeloid leukemias with reciprocal transcription factor I/acute myeloid leukaemia 1) plays an rearrangements can be distinguished by specific gene expression important role in the regulation of the expression level of many profiles. Proc Natl Acad Sci USA 2002; 99: 10008–10013. genes involved in hematopoietic cell development; the impair- 4 Kohlmann A, Schoch C, Dugas M, Rauhut S, Weninger F, ment of AML1 function deregulates the pathways leading to cell Schnittger S et al. Pattern robustness of diagnostic gene expression signatures in leukemia. Genes Chromosomes Cancer 2005; 42: proliferation and differentiation and thus seems to contribute to 299–307. 57–59 tumoral transformation. On the other hand, Wnt (Wingless- 5 Jaffe ES, Harris NL, Stein H, Vardiman JW. World Health type) signalling pathway has been implicated in self-renewal Organization Classification of Tumors. Pathology and Genetics and proliferation of hematopoietic stem cells and the deregula- of Tumors of Haematopoietic Tissues. IARC Press: Lyon, France, tion of its function is associated with several human cancers.60 2001. Notably, Flt3-ITD and Wnt signalling pathways have recently 6 Brown D, Kogan S, Lagasse E, Weissman I, Alcalay M, Pelicci PG 61 et al. A PMLRARalpha transgene initiates murine acute been proposed as integrated transforming events in AML. promyelocytic leukemia. Proc Natl Acad Sci USA 1997; 94: Nevertheless, additional experiments must be conducted to 2551–2556. unravel the precise role, if any, of TLE1 in APL. 7 He LZ, Triboli C, Rivi R, Peruzzi D, Pelicci PG, Soares V et al. In conclusion, different genetic and phenotypic alterations Acute leukemia with promyelocytic features in PML/RA concurrently modify the APL gene expression pattern, so Ralpha transgenic mice. Proc Natl Acad Sci USA 1997; 94: consistently as to determine the generation of two APL subtypes. 5302–5307. 8 Grisolano JL, Wesselschmidt RL, Pelicci PG, Ley TJ. Altered Nevertheless, a role of Flt3-ITDs as a second genetic lesion, myeloid development and acute leukemia in transgenic mice which cooperates with PML/RARa in leukaemic transformation expressing PML-RAR alpha under control of cathepsin G regulatory in a subset of APL patients, is suggested by several converging sequences. Blood 1997; 89: 376–387.

Leukemia Expression profiling and Flt3 status in APL R Marasca et al 113 9 Kottaridis PD, Gale RE, Frew ME, Harrison G, Langabeer SE, 28 Zipfel PF, Jokiranta TS, Hellwage J, Koistinen V, Meri S. The factor Belton AA et al. The presence of a FLT3 internal tandem H protein family. Immunopharmacology 1999; 42: 53–60. duplication in patients with acute myeloid leukemia (AML) adds 29 Napoleone E, di Santo A, Peri G, Mantovani A, de Gaetano G, important prognostic information to cytogenetic risk group and Donati MB et al. The long pentraxin PTX3 up-regulates tissue response to the first cycle of chemotherapy: analysis of 854 factor in activated monocytes: another link between inflammation patients from the United Kingdom Medical Research Council AML and clotting activation. J Leukoc Biol 2004; 76: 203–209. 10 and 12 trials. Blood 2001; 98: 1752–1759. 30 Lorant DE, Zimmerman GA, McIntyre TM, Prescott SM. Platelet- 10 Thiede C, Strudel C, Mohr B, Schaich M, Schakel U, Platzbecker U activating factor mediates procoagulant activity on the surface of et al. Analysis of FLT3-activating mutations in 979 patients with endothelial cells by promoting leukocyte adhesion. Semin Cell acute myelogenous leukemia: association with FAB subtypes and Biol 1995; 6: 295–303. identification of subgroups with poor prognosis. Blood 2002; 99: 31 Hayward CP. Multimerin: a bench-to-bedside chronology of a 4326–4335. unique platelet and endothelial cell protein – from discovery to 11 Kelly LM, Kutok JL, Williams IR, Boulton CL, Amaral SM, function to abnormalities in disease. Clin Invest Med 1997; 20: Curley DP et al. PML/RARalpha and FLT3-ITD induce an 176–187. APL-like disease in a mouse model. Proc Natl Acad Sci USA 32 Liu FT, Rabinovich GA. Galectins as modulators of tumour 2002; 99: 8283–8288. progression. Nat Rev Cancer 2005; 5: 29–41. 12 Schnittger S, Schoch C, Dugas M, Kern W, Staib P, Wuchter C 33 Almkvist J, Karlsson A. Galectins as inflammatory mediators. et al. Analysis of FLT3 length mutations in 1003 patients with acute Glycoconj J 2004; 19: 575–581. myeloid leukemia: correlation to cytogenetics, FAB subtype, and 34 Sherbet GV, Lakshmi MS. S100A4 (MTS1) calcium binding protein prognosis in the AMLCG study and usefulness as a marker for the in cancer growth, invasion and metastasis. Anticancer Res 1998; detection of minimal residual disease. Blood 2002; 100: 59–66. 18: 2415–2421. 13 Noguera NI, Breccia M, Divona M, Diverio D, Costa V, De Santis 35 Rao KN. The significance of the cholesterol biosynthetic pathway S et al. Alterations of the FLT3 gene in acute promyelocytic in cell growth and carcinogenesis. Anticancer Res 1995; 15: leukemia: association with diagnostic characteristics and analysis 309–314. of clinical outcome in patients treated with the Italian AIDA 36 Cuthbert JA, Lipsky PE. Regulation of lymphocyte proliferation by protocol. Leukemia 2002; 16: 2185–2189. cholesterol: the role of endogenous sterol metabolism and low 14 Avvisati G, Lo Coco F, Diverio D, Falda M, Ferrara F, Lazzarino M density lipoprotein receptors. Int J Tissue React 1987; 9: 447–457. et al. AIDA (all-trans retinoic acid + idarubicin) in newly 37 Banker DE, Mayer SJ, Li HY, Willman CL, Appelbaum FR, Zager diagnosed acute promyelocytic leukemia: a Gruppo Italiano RA. Cholesterol synthesis and import contribute to protective Malattie Ematologiche Maligne dell’Adulto (GIMEMA) pilot study. cholesterol increments in acute myeloid leukemia cells. Blood Blood 1996; 88: 1390–1398. 2004; 104: 1816–1824. 15 Borrow J, Goddard AD, Gibbons B, Katz F, Swirsky D, Fioretos T 38 Liu J, Bang AG, Kintner C, Orth AP, Chanda SK, Ding S et al. et al. Diagnosis of acute promyelocytic leukaemia by RT-PCR: Identification of the Wnt signaling activator leucine-rich repeat in detection of PML-RARA and RARA-PML fusion transcripts. Flightless interaction protein 2 by a genome-wide functional Br J Haematol 1992; 82: 529–540. analysis. Proc Natl Acad Sci USA 2005; 102: 1927–1932. 16 Kiyoi H, Naoe T, Yokota S, Nakano M, Minami S, Kuriyama K, 39 Kawano Y, Kypta R. Secreted antagonists of the Wnt signalling et al, and the Leukemia Study Group of Ministry of Health and pathway. J Cell Sci 2003; 116: 2627–2634. Welfare (Kohseisho). Internal tandem duplication of FLT3 asso- 40 Faurschou M, Borregaard N. Neutrophil granules and secretory ciated with leukocytosis in acute promyelocytic leukemia. vesicles in inflammation. Microbes Infect 2003; 5: 1317–1327. Leukemia 1997; 11: 1447–1452. 41 Luce MJ, Burrows PD. The neuronal EGF-related genes NELL1 and 17 Yamamoto Y, Kiyoi H, Nakano Y, Suzuki R, Kodera Y. Activating NELL2 are expressed in hemopoietic cells and developmentally mutation of D835 within the activation loop of FLT3 in human regulated in the B lineage. Gene 1999; 231: 121–126. hematologic malignancies. Blood 2001; 97: 2434–2439. 42 Veal E, Groisman R, Eisenstein M, Gill G. The secreted 18 Hughes TR, Mao M, Jones AR, Burchard J, Marton MJ, Shannon glycoprotein CREG enhances differentiation of NTERA-2 human KW et al. Expression profiling using microarrays fabricated by an embryonal carcinoma cells. Oncogene 2000; 19: 2120–2128. ink-jet oligonucleotide synthesizer. Nat Biotechnol 2001; 19: 43 Konopleva M, Elstner E, McQueen TJ, Tsao T, Sudarikov A, Hu W 342–347. et al. Peroxisome proliferator-activated receptor gamma and 19 Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis Â and display of genome-wide expression patterns. Proc Natl Acad retinoid receptor ligands are potent inducers of differentiation Sci USA 1998; 95: 14863–14868. and apoptosis in leukemias. Mol Cancer Ther 2004; 3: 1249–1262. 20 Efron B, Tibshirani R. Empirical Bayes methods and false discovery 44 Levanon D, Goldstein RE, Bernstein Y, Tang H, Goldenberg D, rates for microarrays. Genet Epidemiol 2002; 23: 70–86. Stifani S et al. Transcriptional repression by AML1 and LEF-1 is 21 Joliffe IT. Principal Component Analysis. Springer: New York, mediated by the TLE/Groucho corepressors. Proc Natl Acad Sci 1986. USA 1998; 95: 11590–11595. 22 Shih LY, Kuo MC, Liang DC, Huang CF, Lin TL, Wu JH et al. 45 Sanz MA, Lo Coco F, Martin G, Avvisati G, Rayon C, Barbui T Internal tandem duplication and Asp835 mutations of the FMS-like et al. Definition of relapse risk and role of nonanthracycline drugs tyrosine kinase 3 (FLT3) gene in acute promyelocytic leukemia. for consolidation in patients with acute promyelocytic leukemia: a Cancer 2003; 98: 1206–1216. joint study of thye GIMEMA and PETHEMA cooperative groups. 23 Au WY, Fung A, Chim CS, Lie AK, Liang R, Ma ES et al. FLT-3 Blood 2000; 96: 1247–1253. aberrations in acute promyelocytic leukaemia: clinicopathological 46 Callens C, Chevret S, Cayuela JM, Cassinat B, Raffoux E, de Botton associations and prognostic impact. Br J Haematol 2004; 125: S et al. Prognostic implication of FLT3 and Ras gene mutations in 463–469. patients with acute promyelocytic leukemia (APL): a retrospective 24 Simon R, Radmacher MD, Dobbin K, McShane LM. Pitfalls in the study from the European APL Group. Leukemia 2005; 19: use of DNA microarray data for diagnostic and prognostic 1153–1160. classification. J Natl Cancer Inst 2003; 95: 14–18. 47 Haferlach T, Kohlmann A, Schnittger S, Dugas M, Hiddemann W, 25 Hamann J, Eichler W, Hamann D, Kerstens HM, Poddighe PJ, Kern W et al. AML M3 and AML M3 variant each have a distinct Hoovers JM et al. Expression cloning and chromosomal mapping gene expression signature but also share patterns different from of the leukocyte activation antigen CD97, a new seven-span other genetically defined AML subtypes. Genes Chromosomes transmembrane molecule of the secretion receptor superfamily Cancer 2005; 43: 113–127. with an unusual extracellular domain. J Immunol 1995; 155: 48 Tse KF, Mukherjee G, Small D. Constitutive activation of FLT3 1942–1950. stimulates multiple intracellular signal transducers and results in 26 Black RA. Tumor necrosis factor-alpha converting enzyme. transformation. Leukemia 2000; 14: 1766–1776. Int J Biochem Cell Biol 2002; 34: 1–5. 49 Zheng R, Levis M, Piloto O, Brown P, Baldwin BR, Gorin NC et al. 27 Smyth EM, FitzGerald GA. Human prostacyclin receptor. Vitam FLT3 ligand causes autocrine signaling in acute myeloid leukemia Horm 2002; 65: 149–165. cells. Blood 2004; 103: 267–274.

Leukemia Expression profiling and Flt3 status in APL R Marasca et al 114 50 Levis M, Tse KF, Smith BD, Garrett E, Small D. A FLT3 tyrosine and acts as a repressor of AML1-induced transactivation. Biochem kinase inhibitor is selectively cytotoxic to acute myeloid leukemia Biophys Res Commun 1998; 252: 582–589. blasts harboring FLT3 internal tandem duplication mutations. 56 Javed A, Guo B, Hiebert S, Choi JY, Green J, Zhao SC et al. Blood 2001; 98: 885–887. Groucho/TLE/R-esp proteins associate with the nuclear matrix and 51 Zheng R, Friedman AD, Small D. Targeted inhibition of FLT3 repress RUNX (CBF(alpha)/AML/PEBP2(alpha)) dependent activa- overcomes the block to myeloid differentiation in 32Dcl3 cells tion of tissue-specific gene transcription. J Cell Sci 2000; 113: caused by expression of FLT3/ITD mutations. Blood 2002; 100: 2221–2231. 4154–4161. 57 Michaud J, Scott HS, Escher R. AML1 interconnected pathways of 52 Zheng R, Friedman AD, Levis M, Li L, Weir EG, Small D. Internal leukemogenesis. Cancer Invest 2003; 21: 105–136. tandem duplication mutation of FLT3 blocks myeloid differentia- 58 Kurokawa M, Hirai H. Role of AML1/Runx1 in the pathogenesis of tion through suppression of C/EBPalpha expression. Blood 2004; hematological malignancies. Cancer Sci 2003; 94: 841–846. 103: 1883–1890. 59 Planaguma J, Diaz-Fuertes M, Gil-Moreno A, Abal M, Monge M, 53 Mizuki M, Schwable J, Steur C, Choudhary C, Agrawal S, Sargin B Garcia A et al. A differential gene expression profile reveals et al. Suppression of myeloid transcription factors and induction of overexpression of RUNX1/AML1 in invasive endometrioid carci- STAT response genes by AML-specific Flt3 mutations. Blood 2003; noma. Cancer Res 2004; 64: 8846–8853. 101: 3164–3173. 60 Polakis P. The oncogenic activation of beta-catenin. Curr Opin 54 Kim KT, Baird K, Ahn JY, Meltzer P, Lilly M, Levis M et al. Pim-1 is Genet Dev 1999; 9: 15–21. up-regulated by constitutively activated FLT3 and plays a role in 61 Tickenbrock L, Schwable J, Wiedehage M, Steffen B, Sargin B, FLT3-mediated cell survival. Blood 2005; 105: 1759–1767. Choudhary C et al. Flt3 tandem duplication mutations cooperate 55 Imai Y, Kurokawa M, Tanaka K, Friedman AD, Ogawa S, Mitani K with Wnt signaling in leukemic signal transduction. Blood 2005; et al. TLE, the human homolog of groucho, interacts with AML1 105: 3699–3706.

Supplementary Information accompanies the paper on the Leukemia website (http://www.nature.com/leu).

Leukemia