Leukemia (2001) 15, 1721–1728  2001 Nature Publishing Group All rights reserved 0887-6924/01 $15.00 www.nature.com/leu Distinct expression profiling in chronic lymphocytic leukemia with 11q23 deletion Y Aalto1, W El-Rifai1,2, L Vilpo3, J Ollila4, B Nagy1,5, M Vihinen6, J Vilpo3 and S Knuutila1

1Department of Medical Genetics, Haartman Institute and Helsinki University Central Hospital, University of Helsinki, Helsinki, Finland; 2Department of Medicine, University of Virginia Health Systems, VA, USA; 3Department of Clinical Chemistry, Tampere University Hospital and University of Tampere, Tampere, Finland; 4Department of Biosciences, Division of Biochemistry, University of Helsinki, Helsinki, Finland; 5Genetic Laboratory, Faculty of Medicine, Semmelweis University, Budapest, Hungary; and 6Department of Bioinformatics, Institute of Medical Technology, University of Tampere, Tampere, Finland

Chronic lymphocytic leukemia (CLL) is a heterogeneous dis- sands of in a single experiment. The information ease with regard to its clinical course. The limitations of the obtained from cDNA array analyses provides new diagnostic methods currently available for prognostic assessment in CLL 10,11 do not allow accurate prediction of the risk of disease pro- and prognostic parameters for cancer patients. We used gression in individual patients. The recently developed cDNA this technology to investigate the profiles of array technique provides a unique opportunity to study gene patients at different stages of CLL and to assess the association expression in various malignancies. To identify new molecular between the gene expression profile and the clinical behavior markers for prognostication of CLL patients, we analyzed cDNA of the disease. arrays by using hierarchical clustering and standard statistic t-test on 34 CLL patients. We found significant expression dif- ferences in 78 genes compared to the reference tonsillar B lym- phocytes. A cluster of genes, LCP1, PARP, BLR1, DEK, NPM, Materials and methods MCL1, SLP76, STAM, HIVEP1, EVI2B, CD25, HTLF, HIVEP2, BCL2, MNDA, PBX3, EBI2, TCF1, CGRP, CD14, IL8, GZMK, Patients GPR17 and CD79B, was associated (P Ͻ 0.05) with the unfavor- able 11q deletion and also with the unfavorable Binet stages B Peripheral blood specimens were collected from 34 CLL and C. We present here gene expression profiling that is asso- ciated with CLL patients with the 11q23 deletion. Many of the patients referred to the CLL outpatient clinic at the Tampere genes in the cluster have not previously been shown to be University Hospital. The selection criterion was a blood lym- related to the initiation or progression of CLL. These novel fin- phocyte count of 30 × 109 or higher.1 The median age of the dings provide fundamental information for further attempts to patients (nine females and 25 males) was 64.8 years (range understand the interaction of the clustered genes in the leuko- 48–79). The patients were diagnosed and staged according mogenesis of CLL in order to better design treatments aimed to standard clinical, morphological and immunophenotyping at specific molecular target(s). Leukemia (2001) 15, 1721–1728. criteria and the Binet system.1,12,13 All patients had a Keywords: chronic lymphocytic leukemia; Binet clinical staging; + + + 11q23 deletion; cDNA array; molecular marker CD19 /CD5 /CD23 immunophenotype. Fifteen patients presented with stage A, 10 with stage B and nine with stage C disease.1 The clinical data are shown in Table 1. Introduction

Chronic lymphocytic leukemia (CLL) is the most common Reference form of leukemia in the Western world. The prognosis for patients with CLL is highly variable. Some patients survive To obtain pure reference material for the expression compari- with indolent disease for many years without any treatment son, CD19-positive B cells were purified from human adenoid and eventually succumb to other diseases. Other patients palatine tonsil samples from six healthy children. The B cells develop a more aggressive malignancy from which they die. were purified using microbeads conjugated to a monoclonal The median survival is approximately 10 years.1 In order to CD19 antibody (Miltenyi Biotec, Bergisch Gladbach, reduce CLL-related mortality, it is essential to identify patients Germany). The proportion of T lymphocytes was less than 5%, at risk and prevent progression of their disease. Cytogenetic indicating that 95% of the isolated cells represented B-lym- studies have shown aberrations in 50–80% of phocyte population. CLL patients and certain genetic changes have been reported repeatedly, such as trisomy 12, deletions in 13q and 14q, and losses in 6q, 11q and 13q.2–7 Losses in 11q appear to be one RNA isolation of the most common structural chromosome aberrations in CLL and they are associated with both the development of the Lymphocytic cells were isolated from the patient’s blood using disease and survival.3,4,7–9 Despite the relatively large number one-step density gradient centrifugation in Ficoll–Paque of patients available for study, little is known about the genes (Pharmacia Fine Chemicals, Uppsala, Sweden), and the total involved in the etiology and progression of CLL. RNA was extracted using the Trizol Reagent (Gibco BRL, cDNA microarray is a new powerful technology capable of Grand Island, NY, USA). The proportion of monocytes and profiling the gene expression patterns of hundreds to thou- polyclonal T and B lymphocytes was 1–13%, indicating that 87–99% of the isolated cells represented the leukemic popu- lation. The RNA was then treated with DNase I (Boehringer Mannheim, Mannheim, Germany) for purification, according Correspondence: S Knuutila, Department of Medical Genetics, Hel- sinki University Central Hospital, PO Box 400 (Haartmaninkatu 3, 4th to the protocol of the Atlas cDNA Expression Array’s user floor), FIN-00029 HUS, Helsinki, Finland; Fax: +358-9-191.26788 manual (Clontech Laboratories, Palo Alto, CA, USA). The Received 13 June 2001; accepted 19 July 2001 quality and integrity of the RNA were checked using 1% aga- cDNA array in chronic lymphocytic leukemia Y Aalto et al 1722 Table 1 Clinical, histological and cytogenetic data for 34 patients with CLL

Patient Age/ FAB Stage Progressionc Blood 11q23 Karyotype No Sex diagnosisa (Binet)b lymphocytes deletion (×109/l) found by FISH

1 61/F CLL A S 86 No 46,XY,del(6)(q?22)/46,XX[18] 2 69/M CLL A I 51 No 46,XY,del(13)(q?14q?22)[2]/46, XY[18] 3 50/F CLL A S 130 No 46,XX[20] 4 54/M CLL A S 63 Yes 46,XY,del(11)(q14q24)[1]/46,idem, del(13)(q?14q?32)[1]/46,idem, t(3;10)(p13–21;q25–26), del(13)(q?14q?32)[6]/46,XY[6] 5 64/M CLL A T 46 Yes 46,XY,del(7)(q32q36), del(11)(q21q23)[6]/46,idem, del(6)(q?21q?24)[4]/46,XY[7] 6 57/M CLL A F 63 No 46,XY[20] 47XY,+3[1]/46,XY[19] 7 57/M CLL A F 120 No 46,XY[20] 8 55/M CLL/PL A F/T 36 No 48,XY,+6,add(11)(p?15),+12[4]/46, XY[8] 9 48/M CLL A S 93 No 46,XY[20] 10 67/M CLL A I 67 No 46,XY[20] 11 57/M CLL A S 68 No 46,XY[10] 12 59/M CLL A I 98 No 46,XY[20] 13 79/F CLL A S 155 No 46XX 14 71/M CLL A F 134 No 46,XY[20] 15 70/F CLL A F 69 No 43–44,XX,−4,del(6)(q?22),−16,−17, −18,+2mar[11]/46,XX[4] 16 68/M CLL B S 67 No 46,XY,del(6)(q?15q22), del(13)(q?14q?31)[15/46,XY[1]] 17 72/M CLL B F 84 Yes 47,XY,+12[2]/46,XY[25] 18 73/M CLL/mix B I 142 Yes 46,XY,del(11)(q?22q?24)[2]/46,XY[18] 19 55/M CLL/PL B I 59 No 46,XY,del(6)(q?23)[2]/46,XY[18] 20 63/M CLL B S 59 Yes 46,XY,del(11)(q?13),del(13)(q?14) [2]/46,XY[15] 21 79/F CLL/mix B T 97 No 46,XX[15] 22 55/M CLL/mix B T 216 No 47,XY,+12[2]/46,XY[10] 23 70/M CLL B F 92 No 46,XY[20] 24 76/F CLL B I 173 No 46,XY[20] 25 78/M CLL B F 224 No 46,XY,t(4;11)(p?15;q?21)[3]/ 46,XY[17] 26 58/M CLL/PL C T 310 No 46,XY,+der(12)t(12;15)(p12;q13–15), der(14;17)(q10;q10)[17]/46,idem, add(1)(q21)[2]/46,XY[1] 27 69/F CLL C T 88 Yes 46,XX,del(11)(q1323), del(13)(q12q21)[7]/46,XX[9] 28 65/M CLL C T 132 No 46,XY,del(13)(q?14q?22)[2]/ 46,XY[18] 29 67/F CLL/PL C T 188 No 45–46,XX,add(1)(p33–35), del(5)(q?31),−10,−17,+mar, inc[cp10]/46,XX[10] 30 73/F CLL/mix C F/T 78 No 47,XX,+12[5]/46,XX[15] 31 68/M CLL C F 114 Yes 46,XY,+8,del(6)(q;23),del(11)(q?14), del(13)(q?14q?22),−20[cp9]/46,XY[5] 32 77/F CLL C F 206 Yes 46,XX[2] 33 67/M CLL C F/T 178 Yes 46,XY[20] 34 63/M CLL/PL C F 430 No

aFAB diagnosis: according to Bennett et al.1,12,13 bStage (Binet): according to Binet et al.1,12,13 cProgression: S (slow), blood lymphocytes increased less than 20% within a year; F (fast), lymphocyte doubling time 1 year; I (intermediate), between S and F; T, chemotherapy given afterward; if alone, natural disease progression non-evaluable.

Leukemia cDNA array in chronic lymphocytic leukemia Y Aalto et al 1723 rose gel electrophoresis. Only high-quality RNA was used for 0.5 for underexpression. We raised the threshold ratio values gene expression profiling study. to 3 for overexpression and to 0.33 for underexpression. cDNA array hybridization Real-time quantitative reverse transcription polymerase chain reaction (QRT-PCR) To screen the samples for gene expression, we used the Atlas Human Hematology array (Clontech; 7737-1). Each filter con- Quantitative RT-PCR was performed to confirm the expression tains 406duplicate spots representing cDNAs of known and alteration of seven genes (MCL1, BAX, BCL2L1, EVI2B, sequence-verified genes (for information about genes see SLP76, HIVEP1 and BCL2). The First Strand cDNA Synthesis www.clontech.com). The total RNA (3–4 ␮g) from 34 CLL Kit for RT-PCR (Roche Diagnostic, Mannheim, Germany) was patients and from reference, CD19-positive B cells purified used to obtain cDNAs from 30 CLL samples and from refer- human adenoid palatine tonsil samples, was converted into ence B cells applying random hexamer primers. Gene-specific cDNA and labeled with 33PdATP using the Clontech cDNA PCR primer sequence information corresponding to the PCR array labeling kit. After hybridization, the filters were washed targets on the Atlas Human Hematology/Immunology cDNA following the manufacturer’s instructions. The filters were expression array filters (Clontech) was acquired from TIB exposed to imaging plates (BAS-MP 2040S; Fuji, Kanagawa, MOLBIOL Syntheselabor (Berlin, Germany). Primers for the Japan) and scanned with a phosphorimager (Bio-Imaging Ana- following genes were used, MCL1, BAX, BCL2L1, EVI2B, lyzer, BAS-2500; Fuji) to obtain high-resolution (16bit) tiff- SLP76, HIVEP1 and BCL2 (Table 2). PCRs were performed in format images. the LightCycler thermal cycler (Roche). Standard curves were obtained by using serial dilutions of the ␤-globulin gene (DNA Control kit; Roche) according to the supplier’s instructions in each LightCycler run. The PCR reaction consisted of 2 ␮lof cDNA array data analysis DNA Master SYBR Green I mix (LightCycler-FastStart DNA Master SYBR Green I kit, Roche; containing Taq DNA poly- The intensity ratio was quantified for each gene and it ␮ merase, dNTP, MgCl2, and SYBR Green I dye), 2 l of cDNA reflected the relative abundance of the gene in each experi- and 4–10 pmol of each primer. The amplification program mental RNA sample compared with RNA samples pooled included an initial denaturation at 95°C with 8 min hold, fol- from six adenoids. The intensities of spots were obtained with lowed by 45 cycles at 95°C with 10 s hold, annealing tem- Atlas Image analysis software (Clontech). Local background perature of 60°C with 5 s hold, and at 72°C with 20 s hold. was subtracted at each point. To minimize the possible effect Amplifications were followed by melting curve analysis using of differences in the amounts of RNA and efficiencies in the the program run for one cycle at 95°C with 0 s hold, 65°C hybridizations, intensities were normalized so that the sums with 10 s hold, and 95°C with 0 s hold at the acquisition step of intensities between patients were equal. To obtain the ratio mode. A negative control without cDNA template was run value of expression, intensity values of genes from each simultaneously with every assay. Each PCR from each cDNA patient were compared to values from reference sample. sample was run in duplicate. The concentration of each gene Ratios were log2 transformed and genes, in which the product was determined on the basis of kinetic approach expression ratio altered significantly in at least three patients, using the LightCycler software (Roche). Results are expressed were taken to further analysis. Hierarchical clustering was as ratio value of expression obtained from comparison of the applied to both axes with the Cluster program (Michael Eisen, concentration values of genes from each patient with the aver- Stanford University, USA) using uncentered correlation as a age value of expression from reference sample. similarity metric. Results were visualized with the TreeView program (Michael Eisen, Stanford University). To identify the subsets of genes that may be associated with a poor clinical Detection of 11q23 deletion by fluorescence in situ outcome in patients with CLL, the Mann–Whitney U test was hybridization used to compare the differences in gene expression between different groups. Statistical analyses were performed using the YAC clone from DEPH-Ge´ne´thon (Paris, France) was labeled SPSS statistical package (SPSS, Chicago, IL, USA). The statisti- Ͻ and applied for the analysis of 11q23 deletion as described cal significance was accepted when P 0.05. earlier.14

Thresholds and controls for cDNA array analysis Results

All the filters used were from same printing batch to minimize The analysis revealed 78 genes that were differentially variation from one lot to another. To ensure the optimum expressed as compared to the gene expression profile of the quality of the hybridizations, the filters were used for a reference pool sample, in at least three of the cases. Of the maximum of three times. Five samples were hybridized twice, differentially expressed genes, 37 were down-regulated and each in a different filter in a different experiment, to confirm 41 were up-regulated. These genes included surface antigens, the reproducibility of the results. The two independent transcription factors, oncogenes and kinases, chemo- hybridization images of the same sample were then compared kines and chemokine receptors, and other genes involved in to each other using Atlas Image software (Clontech). In our blood disorders. The profiles for all the patients are shown in analysis, the differences between any two hybridizations Figure 1, in which a hierarchical clustering was used to group obtained from the same sample were less than 2% using a the genes on the basis of the similarity in their expression. The ratio of 1.5 for overexpression and 0.65 for underexpression, same clustering method was also used to group the 34 patient and less than 1% using a ratio of 2.0 for overexpression and samples on the basis of similarities in their expression of these

Leukemia cDNA array in chronic lymphocytic leukemia Y Aalto et al 1724 Table 2 Sequence of primers used and the corresponding Tm values

Gene Forward primer Tm Reverse primer Tm (°C) (°C)

MCL1 gATgATCCATgTTTTCAgCgAC 56.4 CTCCACAAACCCATCCCAg 57.7 BAX TgCTTCAgggTTTCATCCAg 56.9 ggCggCAATCATCCTCTg 57.3 BCL2L1 ATggCAgCAgTAAAgCAAgC 57.6 CggAAgAgTTCATTCACTACCTgT 57.3 EV12B AgAAgCAgTCACAgCCTACCTTA 57.2 ggTAgTTgTCTggCAgAggTg 56.7 SLP76 AAACCACCATTTTCTgACAAgC 57.1 CTCATAggAAgTAgTgCTggCTg 57.0 HIVEP1 gCACACATTCCAggTCTCCA 57.7 TgAgTTCAggCTTgggCTT 57.7 BCL2 CgCATCAggAAggCTAgAgTT 57.2 TCggTCTCCTAAAAgCAggC 58.2

genes. Although no information on the identity of the sample underexpressed genes in monoclonal neoplastic B cells in was used in the clustering, the algorithm segregated all the CLL: (a) normal tonsillar B cells are CD5-negative unless they patients with 11q deletion into the same class. However, the are strongly activated; and (b) mantle zone B cells, germinal algorithm failed to aggregate the patients in stages A, B and center B cells and memory B cells show different gene C into definite classes. expression profiling. Furthermore, we did not purify B cells To identify a subgroup of genes that may be associated with from mononuclear fraction from CLL patients, and the pro- poor clinical outcome in CLL patients, we performed standard portions of monocytes and/or non-clonal B and T cells in the statistics (t-test) to compare the gene expression levels in the specimens varied between 1% and 13%. This may have pro- different groups. The expression of eight down-regulated duced a level of background expression that can be difficult genes, LCP1, PARP, BLR1, DEK, NPM and MCL1, was signifi- to interpret. Even when the cases with high frequency of non- cantly different between the groups with and without 11q23 clonal B cells were equally distributed among patients with deletion (P Ͻ 0.05), whereas the expression of BLR1 and EBI2 and without the 11q23 deletion, as well as within different differed between stage B or C disease and stage A disease (P Binet stages (data not shown), it is still possible that the back- Ͻ 0.05) (Figure 1). Ten genes, SLP76, STAM, HIVEP1, EVI2B, ground expression causes, for example, overexpression of CD25, HTLF, HIVEP2, BCL2, MNDA and PBX3, were more CD3 and CD14, which are not normally expressed in CLL). frequently up-regulated in patients with 11q23 deletion than Similarly, overexpression of SLP76should be interpreted in those without the deletion (P Ͻ 0.05). The expression of with caution. TCF1, CGRP, CD14, IL8, and GZMK was significantly higher Several up-regulated and down-regulated genes were com- in stage B or C disease than in stage A disease (P Ͻ 0.05). monly expressed in most CLL patients irrespective of the Binet The expression of CD79B, TCF1, CD14, GPR17 and GZMK stage or the 11q status. These up-regulated genes are surface associated with the disease progression (P Ͻ 0.05) (Figure 1). antigens CD5, CD23, CD55 and CD44H, oncogenes AF17 We used quantitative RT-PCR to validate the expression dif- and ZAP70, B cell-specific activator protein PAX5, and P1. ferences for seven genes. Relative expression levels initially Down-regulated genes are CD21, CD22, CD10, CD54, RGS, determined with the cDNA array were correlated with Light- RHOM2, TOP2A, AHPD1, ADA and RAC1. These genes are Cycler results for 30 samples tested (lack of sufficient RNA probably not related to CLL progression but may be more prevented testing of four cases). Figure 2 illustrates the important for the initiation of this disease or may reflect lin- concordant results by cDNA array and LightCycler. eage specificity (as discussed under Results). Loss of genomic material in 11q is one of the most common structural chromosome aberrations in CLL and the finding is Discussion associated with disease progression and poor survival.3,4,7–9 Our results showed that gene expression in patients with We used a cDNA array to profile gene expression in CLL 11q23 deletions, observed by high-resolution FISH mapping, patients with different stages/status. Our results show that all irrespective of their Binet clinical stage, varied in a larger the patients with 11q deletion segregated into the same class number of genes than in other patients. Sixteen genes had stat- based on global similarities in gene expression patterns. The istically significant differences in expression between the clustering of the patients based on their expression signatures patients with 11q23 deletion and without it. Of these genes, did not correlate with their Binet classification. This finding 10 were up-regulated and six down-regulated. Among the further supports the notion that CLL characterized by 11q down-regulated genes, apoptosis-associated MCL1 deletion defines a distinct disease subgroup and the clinical and PARP, and oncogenes DEK and NPM showed signifi- heterogeneity of CLL is due to unrecognized molecular het- cantly higher average values in expression in patients with erogeneity of CLL in the patients. Another finding of this study 11q23 deletion than in patients without 11q23 deletion, is that subsets of genes can be identified using standard stat- whereas the value of cell surface antigen BLR1 was signifi- istics, the t-test. These subsets can be used for grouping and cantly lower in the presence of 11q deletion. These genes indicating a poor prognosis. Our results also show that together with 10 up-regulated genes, SLP76, STAM, HIVEP1, patients with poor prognostic parameters expressed a higher EVI2B, CD25, HTLF, HIVEP2, BCl2, MNDA and PBX3, indi- number of disease-related genes than patients with a more cated that 11q23 deletion is associated with the activation of favorable prognosis. This may indicate that increased gene a number of genes. The significantly higher level of expression expression is the underlying mechanism of CLL progression. in a number of genes in these patients may contribute to the CD19-positive B cells were pooled from six normal aden- pathogenesis of the CLL subgroup characterized by 11q23 oids and used as reference material. The following issues were deletion and poor prognosis. Conceivably, we suggest that the taken into consideration when interpreting the over- and loss of putative tumor suppression genes in the 11q23 deletion

Leukemia cDNA array in chronic lymphocytic leukemia Y Aalto et al 1725 n; N, no 11q23 deletion. Red color represents up-regulated genes, green 2.5 to 2.5 in log (real ratios are 5.66 over/under). − , significant expression differences in stage B or C and stage A disease; §, expression associated with # 0.05); Ͻ P test). U 0.05, Mann–Whitney Ͻ P Summary of the gene expression profiles for 34 CLL patients. At the top are listed the samples studied including patient’s code number: D, 11q23 deletio Figure 1 Progression status is showndown-regulated as; genes. S, The slow; brightness F,*, of fast; significant the I, color expression intermediate; correlates differences T, with chemotherapy in the given degree cases afterwards. of with Names expression and of as without shown the at genes 11q23 the appear deletion bottom on ( of the the right-hand image. side. The scale is from disease progression (

Leukemia cDNA array in chronic lymphocytic leukemia Y Aalto et al 1726

Figure 2 Comparison of the results from cDNA array and RT-PCR in 30 CLL samples. Red color represents up-regulated genes, green down- regulated genes. Names of the genes appear on the right-hand side of the image. a, detected by cDNA array; r, detected by quantitative RT- PCR. The brightness of the color correlates with the degree of expression as shown at the bottom of the image. The scale is from −7to7in log. The average ratios of relative gene expression detected by cDNA array and real-time quantitative RT-PCR are on the right-hand side.

represents a general removal of suppression that leads to acti- tein is regarded as a major mechanism in the genesis of CLL. vation of several cellular pathways in CLL. High-level overexpression of BCL2 was found in our patients Overexpression of STAM and SLP76 was significant in all with 11q23 deletion. MNDA and PBX3, both involved in tran- patients with 11q23 deletion. STAM and SLP76 both act as scription and act as transcription activators, were also fre- adaptor molecules and are involved in intracellular signal quently seen in the patients with 11q23 deletion. MNDA is a transduction mediated by a variety of cytokines. STAM has nuclear antigen expressed specifically in maturing cells of the been implicated in the signaling pathways for cell growth and myelomonocytic lineage and in monocytes and granulocytes. c- induction.15 The role of STAM in CLL, as well as in Examination of lymphoid tissues showed a low level of other types of hematological malignancy, is unknown. expression in a population of normal mantle B lymphocytes Another adaptor protein SLP76 is normally expressed in T and in a subset of B cell neoplasms including small lympho- lymphocytes and myeloid cells, and it is a substrate for cytic lymphoma.24 MNDA is considered an interferon- ZAP70. It has been reported that the SLP76protein was not inducible gene, participating in blood cell-specific responses detected in a number of B cell lines or in normal mouse B to interferons.25 cells.16 The involvement of SLP76 has been limited to T cell When we compared the gene expression in the Binet clini- development.17 cal staging groups A, B and C, we found that two down-regu- Overexpression of EVI2B in the patients with 11q23 lated genes, BLR1 and EBI2, and five up-regulated genes, deletion was more frequent than in the patients without 11q23 TCF1, CGRP, CD14, IL8 and GZMK, displayed statistically deletion. The possible role of DNA viruses, such as ecotropic significant differences in expression between stage A and viruses, in the development of CLL, is unknown. However, it stage B or stage C (P Ͻ 0.05). Furthermore, three out of the has been suggested that the Epstein–Barr virus affects blast and five above-mentioned up-regulated genes, TCF1, CD14, growth transformation in infected B lymphocytes and the pro- GZMK together with CD79B and GPR17, were associated gression of CLL/small lymphocytic lymphoma to Hodgkin-like with fast lymphocyte doubling time (P Ͻ 0.05). This makes lymphoma.18,19 EBI2B lies within an intron of the neurofib- these genes particularly interesting because they may be romatosis type 1 tumor suppressor gene and is expressed in a related to CLL progression and could subsequently be useful number of different cell types, including myeloid cells.20,21 as prognostic markers for CLL patients. Of these up-regulated EVI2B is also located within EVI2, a common viral integration genes, cell surface antigen CD14 has been demonstrated to site identified in retrovirus-induced myeloid tumors.22 It has be associated with more advanced B-CLL patients and shorter been speculated that viral integration at EVI2 alters the overall survival.26 CGRP is a sensory neuropeptide with expression of EVI2B and that this altered expression predis- inflammatory and immunoregulatory activities. CGRP played poses mice to myeloid tumor development.20,22 Furthermore, an inhibitory role in early B cell differentiation27 and inhibited human T cell leukemia virus enhancer factor (HTLF) and interleukin-7-induced pre-B cell colony formation.28 Interleu- human immunodeficiency virus type I enhancer-binding pro- kin-8 (IL-8) has been shown to induce the accumulation of B teins 1 and 2 (HIVEP1 and HIVEP2) were frequently up-regu- cell chronic lymphocytic leukemia cells by prolonging sur- lated in patients with 11q23 deletion. Whether DNA viruses vival in an autocrine fashion.29 The role of GRP17, a potential are involved in the development of CLL patients with 11q chemokine , in CLL is not known. The level of CD79 deletion is an issue that has to be further studied. expression was significantly higher in the group of fast disease Apoptosis dysregulation is a major feature of CLL.23 Accord- progression. This result is consistent with an immunohistoch- ing to the role of BCL2, increased expression of the bcl-2 pro- emical study on CLL cases in which the advanced clinical

Leukemia cDNA array in chronic lymphocytic leukemia Y Aalto et al 1727 stage (Binet stages B and C) was most frequently observed in HF, Sole´ F, Caballı´n MR, Woessner S, Knuutila S. Trisomy 12 in CD79b+ cases than in CD79b− cases.30 Two down-regulated chronic lymphocytic leukemia: an interphase cytogenetic study. genes, BLR1 and EBI2 that encode G protein-coupled recep- Blood 1991; 78: 775–779. 7 Dohner H, Stilgenbauer S, Benner A, Leupolt E, Krober A, Bul- tors, have been suggested to have a role in B cell migration linger L, Dohner K, Bentz M, Lichter P. Genomic aberrations and 31,32 or normal lymphocyte functions. survival in chronic lymphocytic leukemia. N Engl J Med 2000; Our study was not designed to evaluate the prognostic 343: 1910–1916. value of the findings. For this purpose a larger number of 8 Zhu Y, Monni O, El-Rifai W, Siitonen SM, Vilpo L, Vilpo J, Knuut- patients and longer follow-up period are needed. In a recent ila S. Discontinuous deletions at 11q23 in B cell chronic lympho- report Stratowa et al33 were able to identify genes that indicate cytic leukemia. Leukemia 1999; 13: 708–712. 9 Dohner H, Stilgenbauer S, James MR, Benner A, Weilguni T, Bentz poor prognosis. Many of the genes encode cell adhesion-asso- M, Fischer K, Hunstein W, Lichter P. 11q deletions identify a new ciated molecules (L-selectin, integrin-B2, ILIB, IL8 and EGRI). subset of B-cell chronic lymphocytic leukemia characterized by In their comprehensive study, Stratowa et al analyzed the extensive nodal involvement and inferior prognosis. Blood 1997; expression levels of 1024 selected genes in 54 CLL cases. 89: 2516–2522. They did not find correlation between the gene expression 10 DeRisi J, Penland L, Brown PO, Bittner ML, Meltzer PS, Ray M, profiles and the 11q23 deletion. Different gene selections in Chen Y, Su YA, Trent JM. Use of a cDNA microarray to analyse the arrays may partly explain why our findings are not in line gene expression patterns in human cancer. Nat Genet 1996; 14: 457–460. with theirs. 11 Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesi- To conclude, our most prominent findings are significant rov JP, Coller H, Loh ML, Downing JR, Caligiuri MA, Bloomfield differences in the gene expression profiles of patients with CD, Lander ES. Molecular classification of cancer: class discovery indicators of poor prognosis (stages B and C, or 11q deletion). and class prediction by gene expression monitoring. Science These differences point to several genes that could be utilized 1999; 286: 531–537. for prognostication of CLL patients. The genes that were com- 12 Rai KR, Sawitsky A, Cronkite EP, Chanana AD, Levy RN, Pas- ternack BS. Clinical staging of chronic lymphocytic leukemia. monly expressed in patients irrespective of their ABC stage or Blood 1975; 46: 219–234. chromosomal aberration are probably related to the initiation 13 Bennett JM, Catovsky D, Daniel MT, Flandrin G, Galton DA, Gral- of CLL. Our novel results are the first milestone on the road nick HR, Sultan C. Proposals for the classification of chronic to finding molecular prognostic markers for CLL. Further study (mature) B and T lymphoid leukaemias. French–American–British should be carried out in a larger patient material to ascertain (FAB) Cooperative Group. J Clin Pathol 1989; 42: 567–584. whether gene expression clustering can reveal molecular 14 Zhu Y, Monni O, Franssila K, Elonen E, Vilpo J, Joensuu H, Knuut- prognostic markers in CLL. ila S. Deletions at 11q23 in different lymphoma subtypes. Haema- tologica 2000; 85: 908–912. 15 Takeshita T, Arita T, Higuchi M, Asao H, Endo K, Kuroda H, Tanaka N, Murata K, Ishii N, Sugamura K. STAM, signal transduc- Acknowledgements ing adaptor molecule, is associated with Janus kinases and involved in signaling for cell growth and c-myc induction. Immun- ity 1997; 6: 449–457. This work was supported by grants from the Medical Research 16Fu C, Chan AC. Identification of two tyrosine phosphoproteins, Fund of Tampere University Hospital, the Finnish Foundation pp70 and pp68, which interact with phospholipase Cgamma, for Cancer Research, the Sigrid Juse´lius Foundation, Paulo Grb2, and Vav after B cell antigen receptor activation. J Biol Chem Foundation, Finnish Cultural Foundation and Helsinki Univer- 1997; 272: 27362–27368. sity Central Hospital Research Funds. We thank the Tampere 17 Pivniouk V, Tsitsikov E, Swinton P, Rathbun G, Alt FW, Geha RS. CLL Group for collaboration, and Leena Pankko and Merja Impaired viability and profound block in thymocyte development Suoranta for their skilful technical assistance. in mice lacking the adaptor protein SLP-76. Cell 1998; 94: 229–238. 18 Klein E, Teramoto N, Gogolak P, Nagy N, Bjorkholm M. LMP-1, the Epstein–Barr virus-encoded oncogene with a B cell activating References mechanism similar to CD40. Immunol Lett 1999; 68: 147–154. 19 Petrella T, Yaziji N, Collin F, Rifle G, Morlevat F, Arnould L, Farg- 1 Binet JL, Auquier A, Dighiero G, Chastang C, Piguet H, Goasguen eot P, Depret O. Implication of the Epstein–Barr virus in the pro- J, Vaugier G, Potron G, Colona P, Oberling F, Thomas M, Tchernia gression of chronic lymphocytic leukaemia/small lymphocytic G, Jacquillat C, Boivin P, Lesty C, Duault MT, Monconduit M, lymphoma to Hodgkin-like lymphomas. Anticancer Res 1997; 17: Belabbes S, Gremy F. A new prognostic classification of chronic 3907–3913. lymphocytic leukemia derived from a multivariate survival analy- 20 Cawthon RM, Andersen LB, Buchberg AM, Xu GF, O’Connell P, sis. Cancer 1981; 48: 198–206. Viskochil D, Weiss RB, Wallace MR, Marchuk DA, Culver M, Ste- 2 Knuutila S, Elonen E, Teerenhovi L, Rossi L, Leskinen R, Bloom- vens J, Jenkins NA, Copeland NG, Collins FS, White R. cDNA field CD, de la Chapelle A. Trisomy 12 in B cells of patients with sequence and genomic structure of EV12B, a gene lying within B-cell chronic lymphocytic leukemia. N Engl J Med 1986; 314: an intron of the neurofibromatosis type 1 gene. Genomics 1991; 865–869. 9: 446–460. 3 Juliusson G, Oscier DG, Fitchett M, Ross FM, Stockdill G, Mackie 21 Kaufmann D, Gruener S, Braun F, Stark M, Griesser J, Hoffmeyer MJ, Parker AC, Castoldi GL, Guneo A, Knuutila S, Elonen E, S, Bartelt B. EVI2B, a gene lying in an intron of the neurofibrom- Gahrton G. Prognostic subgroups in B cell chronic lymphocytic atosis type 1 (NF1) gene, is as the NF1 gene involved in differen- leukemia defined by specific chromosomal abnormalities. N Engl tiation of melanocytes and keratinocytes and is overexpressed in JMed1990; 323: 720–724. cells derived from NF1 neurofibromas. DNA Cell Biol 1999; 18: 4 Karhu R, Knuutila S, Kallioniemi O-P, Siitonen S, Aine R, Vilpo L, 345–356. Vilpo J. Frequent loss of the 11q14–24 region in chronic lympho- 22 Buchberg AM, Bedigian HG, Jenkins NA, Copeland NG. Evi-2, a cytic leukemia: a study by comparative genomic hybridization. common integration site involved in murine myeloid leukemogen- Genes Chromos Cancer 1997; 19: 286–290. esis. Mol Cell Biol 1990; 10: 4658–4666. 5 Larramendy ML, Siitonen SM, Zhu Y, Hurme M, Vilpo L, Vilpo JA, 23 Keating MJ. Chronic lymphocytic leukemia. Semin Oncol 1999; Knuutila S. Optimized mitogen stimulation induces proliferation of 26: 107–114. neoplastic B cells in chronic lymphocytic leukemia: significance 24 Miranda RN, Briggs RC, Shults K, Kinney MC, Jensen RA, Cousar for cytogenetic analysis. Cytogenet Cell Genet 1998; 82: 215–221. JB. Immunocytochemical analysis of MNDA in tissue sections and 6Pe´rez Losada A, Wessman M, Tiainen M, Hopman AHN, Willard sorted normal bone marrow cells documents expression only in

Leukemia cDNA array in chronic lymphocytic leukemia Y Aalto et al 1728 maturing normal and neoplastic myelomonocytic cells and a sub- 29 Francia di Celle P, Mariani S, Riera L, Stacchini A, Reato G, Foa set of normal and neoplastic B lymphocytes. Hum Pathol 1999; R. Interleukin-8 induces the accumulation of B-cell chronic lym- 30: 1040–1049. phocytic leukemia cells by prolonging survival in an autocrine 25 Briggs RC, Briggs JA, Ozer J, Sealy L, Dworkin LL, Kingsmore SF, fashion. Blood 1996; 87: 4382–4389. Seldin MF, Kaur GP, Athwal RS, Dessypris EN. The human 30 Garcia Vela J, Delgado I, Benito L, Monteserin M, Garcia Alonso myeloid cell nuclear differentiation antigen gene is one of at least L, Somolinos N, Andreu M, Ona F. CD79b expression in B cell two related interferon-inducible genes located on chromosome 1q chronic lymphocytic leukemia: its implication for minimal that are expressed specifically in hematopoietic cells. Blood 1994; residual disease detection. Leukemia 1999; 13: 1501–1505. 83: 2153–2162. 31 Forster R, Mattis AE, Kremmer E, Wolf E, Brem G, Lipp M. A puta- 26Callea V, Morabito F, Oliva BM, Stelitano C, Levato D, Dattilo A, tive chemokine receptor, BLR1, directs B cell migration to defined Gangemi F, Iorfida A, Iacopino P, Nobile F, Molica S, Brugiatelli lymphoid organs and specific anatomic compartments of the M. Surface CD14 positivity in B-cell chronic lymphocytic leu- spleen. Cell 1996; 87: 1037–1047. kaemia is related to clinical outcome. Br J Haematol 1999; 107: 32 Birkenbach M, Josefsen K, Yalamanchili R, Lenoir G, Kieff E. 347–352. Epstein–Barr virus-induced genes: first lymphocyte-specific G pro- 27 McGillis JP, Rangnekar V, Cialella JR. A role for calcitonin gene tein-coupled peptide receptors. J Virol 1993; 67: 2209–2220. related peptide (CGRP) in the regulation of early B lymphocyte 33 Stratowa C, Lo¨ffler G, Lichter P, Stilgenbauer S, Haberl P, differentiation. Can J Physiol Pharmacol 1995; 73: 1057–1064. Schweifer N, Do¨hner H, Wilgenbus KK. cDNA microarray gene 28 Fernandez S, Knopf MA, McGillis JP. Calcitonin-gene related pep- expression analysis of B-cell chronic lymphocytic leukemia pro- tide (CGRP) inhibits interleukin-7-induced pre-B cell colony for- poses potential new prognostic markers involved in lymphocyte mation. J Leukoc Biol 2000; 67: 669–676. trafficking. Int J Cancer 2001; 91: 474–480.

Leukemia