Original Manuscript

Leukemia (2005) 19, 1042–1050. doi:10.1038/sj.leu.2403749 Published online 21 April 2005

Gene Expression Profiling and Microarrays

Gene expression profiling of leukemic cell lines reveals conserved molecular signatures among subtypes with specific genetic aberrations

A Andersson1, P Edén2, D Lindgren1, J Nilsson3, C Lassen1, J Heldrup4, M Fontes3, Å Borg5, F Mitelman1, B Johansson1, M Höglund1 and T Fioretos1

  1. 1Department of Clinical Genetics, Lund University Hospital, Sweden
  2. 2Department of Complex Systems, Theoretical Physics, Lund University, Sweden
  3. 3Center for Mathematical Sciences, Lund University, Sweden
  4. 4Department of Pediatrics, Lund University Hospital, Sweden
  5. 5Department of Oncology, Lund University Hospital, Sweden

Correspondence: Dr A Andersson, Department of Clinical Genetics, University Hospital, SE-221 85 Lund, Sweden. Fax: +46 46 131061; E-mail: anna.andersson@klingen.lu.se

Received 28 October 2004; Accepted 24 February 2005; Published online 21 April 2005.

Top

Abstract

Hematologic malignancies are characterized by fusion genes of biological/clinical importance. Immortalized cell lines with such aberrations are today widely used to model different aspects of leukemogenesis. Using cDNA microarrays, we determined the gene expression profiles of 40 cell lines as well as of primary leukemias harboring 11q23/MLL rearrangements, t(1;19)[TCF3/PBX1], t(12;21)[ETV6/RUNX1], t(8;21)[RUNX1/CBFA2T1], t(8;14)[IGH@/MYC], t(8;14)[TRA@/MYC], t(9;22)[BCR/ABL1], t(10;11)[PICALM/MLLT10], t(15;17)[PML/RARA], or inv(16)[CBFB/MYH11]. Unsupervised classification revealed that hematopoietic cell lines of diverse origin, but with the same primary genetic changes, segregated together, suggesting that pathogenetically important regulatory networks remain conserved despite numerous passages. Moreover, primary leukemias cosegregated with cell lines carrying identical genetic rearrangements, further supporting that critical regulatory pathways remain intact in hematopoietic cell lines. Transcriptional signatures correlating with clinical subtypes/primary genetic changes were identified and annotated based on their biological/molecular properties and chromosomal localization. Furthermore, the expression profile of tyrosine kinase-encoding genes was investigated, identifying several differentially expressed members, segregating with primary genetic changes, which may be targeted with tyrosine kinase inhibitors. The identified conserved signatures are likely to reflect regulatory networks of importance for the transforming abilities of the primary genetic changes and offer important pathogenetic insights as well as a number of targets for future rational drug design.

Keywords:

cDNA microarray, pediatric leukemia, AML, ALL, CML-BC

Top

Introduction

Hematologic malignancies are characterized by balanced chromosomal abnormalities – translocations and inversions – that lead to deregulated expression of genes located in the proximity of the breakpoints or result in tumor-specific fusion genes.1, 2 In acute leukemias, the genes participating in such rearrangements often encode evolutionarily conserved transcription factors that are important also in normal hematopoiesis.3, 4In chronic myeloproliferate disorders, the chimeric fusion proteins more often involve genes encoding proteins with tyrosine kinase activity, the prototypic example being the ABL1 tyrosine kinase that becomes constitutively activated by its fusion to BCR in chronic myeloid leukemia (CML).5, 6

Although hematologic malignancies have been extensively studied and more than 350 recurrent balanced chromosomal abnormalities have been identified,2 many of them providing essential biological and prognostic information, our understanding of how individual fusion genes elicit their leukemogenic properties is still quite limited. However, during recent years, genome-wide gene expression analyses, performed by microarray-based technologies, have demonstrated that hematologic malignancies can be classified based on their gene expression profiles, which have been shown to be strongly associated with specific morphologic, immunophenotypic, genetic, and prognostic parameters.7, 8, 9, 10 Moreover, such studies have identified new molecular subgroups and increased our knowledge about leukemia biology.

Today, the perhaps most widely used model system for studying the biological consequences of leukemia-associated chromosomal rearrangements is based on immortalized hematopoietic cell lines. Although cell lines are known to differ from both neoplastic and normal tissues,11 they provide powerful tools for investigating basic and applied aspects of leukemia cell biology. An illustrative example is provided by Philadelphia (Ph) chromosome-positive cell lines, carrying the t(9;22)(q34;q11) and the BCR/ABL1 chimera, established from CML patients in blast transformation (BC). These cell lines remain sensitive to the ABL1 tyrosine kinase inhibitor imatinib mesylate (Gleevec)12 despite their numerous passages in vitro, indicating that the primary genetic change and its impact on downstream transcriptional networks still are indispensable for their growth.

Surprisingly, however, only a few studies have addressed systematically whether hematopoietic cell lines with characteristic cytogenetic aberrations display conserved gene expression signatures despite their diverse origin and numerous passages in vitro. Recently, Fine et al13 studied the gene expression pattern of a set of hematopoietic cell lines and B-lineage acute lymphoblastic leukemias (ALL), and showed that cell lines and leukemias segregate based on their genetic aberration upon cluster analyses and that they have a shared gene expression signature. In the present study, we confirm and further extend these observations by investigating a larger number of cell lines, derived from all hematopoietic lineages, and by analyzing a larger number of specific genetic aberrations.

Top

Patients, materials and methods

Cell lines and patient material

A total of 40 hematologic cell lines were investigated in this study and their clinical subtype, genetic features, as well as their providers are summarized in Table 1. The cell lines were cultured according to the instructions provided by the cell repositories.


Bone marrow samples from 11 children with acute myeloid leukemia (AML) or ALL were obtained at the time of diagnosis. The collection of bone marrow specimens was approved by the Research Ethics Committee of Lund University, Sweden. Cells were frozen in TRIzol (Invitrogen, Carlsbad, CA, USA) and all cases were analyzed cytogenetically at the Department of Clinical Genetics, Lund, Sweden. In addition, the ALL cases were screened molecularly for MLL rearrangements, BCR/ABL1, ETV6/RUNX1, and TCF3/PBX1 fusions. The clinical and genetic features are listed in Table 2.


Microarray design

cDNA microarray slides were obtained from the Swegene DNA Microarray Resource Center at Lund University, Sweden (http://swegene.onk.lu.se). cDNA from cell lines and primary leukemias were hybridized to 32 and 27K microarray slides, respectively, of different design, but both containing 25 648 clones representing 17 492 Unigene clusters and 10 949 Locus link entries. Gene names are according to Unigene build 165.

RNA extraction, probe labeling, and hybridization conditions

Cells from the cell lines were collected 24 h after medium exchange, and total RNA was extracted using TRIzol reagent (Invitrogen) and RNeasy (Qiagen, Valencia, CA, USA), according to the manufacturer's instructions. As a reference, the Universal Human Reference RNA (Stratagene, La Jolla, CA, USA) was used. cDNA synthesis, labeling of the poly(A) RNA, and hybridizations were performed using standard protocols (Supplementary data). Slides were scanned in the G2565AA Agilent DNA Microarray Scanner (Agilent Technologies, Palo Alto, CA, USA) and analyzed with the GenePix4.0 software (GenePix, Foster City, CA, USA).

Patient RNA was extracted as above, and one round of linear amplification was performed using the RiboAmp-RNA Amplification kit (Arcturus, Mountain View, CA, USA) as described by the manufacturer. The quality of total and amplified RNA (aRNA) was assessed using the Agilent 2100 BioAnalyzer (Agilent). As a reference, aRNA from the Universal Human Reference (Stratagene) was used. Labeling was performed as above. For prehybridization, hybridization, and posthybridization washes, the Pronto Universal Microarray Reagent System (Corning, Acton, MS, USA) was used, as described by the manufacturer.

Microarray data acquisition

The data matrix was uploaded to the BioArray Software Environment (BASE)14 and normalization was performed by a BASE implementation of the intensity-dependent normalization based on a Lowess fit.15 To correct for bad quality spots, an error model was used (Supplementary data) and a variation and presence filter was applied. To analyze the cell line and primary leukemia data sets together, which were hybridized on slides of two different designs (32 and 27 K, respectively), the cell lines and acute leukemias were mean-centered individually, and for the few reporters that occurred in duplicate, measurements were merged and the data were filtered as described (Supplementary data).

Hierarchical clustering analyses (HCA) were performed in TMEV16 and to view the similarities between samples in three dimensions, multidimensional scaling (MDS) using geodesic distances was performed in STATISTICA 6 software (Statsoft Inc., Tulsa, OK, USA).17 This analysis18 has previously been shown by us to visualize efficiently the biological relationship between samples subjected to microarray analysis.17

Four different measurements (Pearson, ROC area, Fisher statistic, and Golub score) were used for discriminatory analysis, yielding highly consistent results. For the gene lists presented herein, the Golub score7 was used to rank genes. A random permutation test was performed using 1000 sample label permutations with Pless than or equal to0.001 considered significant (Pless than or equal to0.002 if <70 reporters) (Supplemental data). The gene lists were subjected to the gene ontology (GO) software EASE,19 and an EASE score <0.05 was considered significant. We also used the EASE tool for basic pathway analysis using the same score as above as significant. Primary data are available through www.klingen.lu.se/E/projects.h
tm
.

Verification of gene expression data

To confirm the results from the cDNA microarray analyses, Northern blot analyses were performed on 10 selected genes, using standard protocols.20 The verified genes were selected based on their rank in the discriminatory lists as well as of their biological interest (Supplementary Table S4; Figure S4). As part of the gene verification process, Western- and Southern blot, and RT-PCR analyses were performed as previously described.21, 22

Top

Results

Gene expression analysis segregates cell lines according to clinical and molecular subtype

To investigate if gene expression profiling of the 40 hematologic cell lines could separate B-lineage ALL, T-cell ALL, AML, and CML BC, we first applied unsupervised HCA and MDS using geodesic distances. As outlined in Figure 1a and b, a nearly perfect segregation of the cell lines according to their lineage was observed, indicating that cell lines of diverse origin maintain a conserved gene expression pattern correlating with specific clinical subtypes. Only three cell lines (EM-3, CMK, and CTV-1) failed to segregate according to their subtype. EM-3, a Ph-positive CML BC cell line, segregated with the AMLs; CTV-1, a monocytic cell line, segregated with the T-cell ALLs; and CMK, a megakaryocytic cell line, segregated with the CML BCs. Using Western blot analysis, we showed that, in contrast to the other investigated BCR/ABL1-positive CML BCs, EM-3 did not display phosphorylation of STAT5, which may indicate that the original molecular signature correlating with BCR/ABL1 expression had been lost or significantly altered in EM-3 (data not shown). CMK, segregating with the CML BC group, was further examined by RT-PCR for the presence of a BCR/ABL1 rearrangement, but no such fusion transcript was detected (data not shown). To rule out contamination by other cell lines, the three misclassified cell lines were analyzed cytogenetically, confirming the karyotypes provided by the cell repository (data not shown). For the above reason, these cell lines were removed from subsequent discriminatory analysis.

Figure 1.
Figure 1 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Hierarchical dendogram and MDS using geodesic distances of the cell lines and primary leukemias according to clinical subtype and genetic rearrangement. (a) Unsupervised HCA, using Pearson correlation and complete clustering, of the 40 cell lines, showing a nearly perfect segregation of the cell lines according to clinical subtype. The analysis is based on 3491 reporters. The color scale above each cluster indicates the log2 ratios; green represents downregulated genes and red upregulated genes. The cell lines are color-coded based on their subtype; AML (red), CML BC (green), B-lineage ALL (blue), and T-cell ALL (orange). (b) Projection of the 40 cell lines in 3D using MDS and geodesic distance, colored according to their subtype as above. The projection clearly separates the four clinical subtypes of leukemia. Moreover, the three misclassified cell lines, as determined by unsupervised HCA, mis-segregate also in this analysis. (c) Unsupervised HCA, using Pearson correlation and complete clustering of the 40 cell lines, color-coded according to their genetic abnormality; MLL (red), P210 BCR/ABL1 (green), PICALM/MLLT10(AF10) (black), PML/RARA (orange), RUNX1/CBFA2T1 (light pink), CBFB/MYH11 (dark pink), TRA@/MYC (dark green), P190 BCR/ABL1 (light blue), IGH@/MYC (yellow), TCF3/PBX1 (blue/gray), and ETV6/RUNX1 (dark blue). (d) Visualization of the 37 cell lines and 11 acute leukemias according to genetic rearrangements. Cell lines and fresh leukemic samples were mean-centered individually and unsupervised HCA was applied on the 3841 reporters. The cell lines and patient samples are color-coded based on their genetic aberration, as above. This figure clearly shows that cell lines and all primary leukemias, apart from the two RUNX1/CBFA2T1-positive cases and a Ph-positive ALL, segregate based on their genetic rearrangement.

Full figure and legend (383K)

Using unsupervised HCA, cell lines with identical or closely related primary genetic aberrations were also found to segregate together (Figure 1c). The cosegregation of cell lines with the same primary genetic changes indicates a shared molecular signature among cell lines with identical genetic aberrations. The two cell lines harboring P190 BCR/ABL1 did not cluster together among the B-lineage ALLs. Moreover, three cell lines (OCI-AML2, SIG-M5, and HL-60) segregated with the five AMLs characterized by MLL abnormalities, but none of these have previously been reported to carry such rearrangements. Upon Southern blot analysis, we were, however, able to show that OCI-AML2 harbored a rearranged MLL gene (Supplementary Figure S1); the two other cell lines displayed germline configuration of MLL.

Primary acute leukemias and cell lines with the same primary genetic change cosegregate

Having shown that cell lines of diverse origin, but with the same primary genetic change, segregate together using unsupervised algorithms, we next investigated whether the maintained gene expression signatures identified in cell lines with identical primary genetic abnormalities also could be found in primary leukemias harboring the same aberrations. Accordingly, 11 acute leukemias with selected genetic aberrations were added to the data set (Table 2). Since only 27K arrays were available when analyzing the acute leukemias, the obtained data sets were mean-centered individually (Supplementary data). Following mean-centering, unsupervised HCA was applied and eight of the 11 acute leukemias cosegregated with the respective cell lines according to their genetic aberration (Figure 1d). This strongly suggests that the original gene expression pattern in primary leukemia samples is preserved in immortalized cell lines. Two acute leukemias with RUNX1/CBFA2T1 failed to segregate with the two RUNX1/CBFA2T1-positive cell lines. Also, the Ph-positive ALL failed to cluster with the Ph-positive ALL cell lines. We next applied the discriminatory gene lists obtained from the cell lines correlating with leukemia subtype and genetic aberrations (see below) on the combined data set, resulting in a successful clustering according to clinical and genetic subtype, indicating that the majority of the top genes discriminating the clinical subtypes in cell lines were preserved in the primary acute leukemias investigated (data not shown).

Identification of genes discriminating the clinical subtypes

To identify genes that significantly correlate with B-lineage ALL, T-cell ALL, AML, and CML BC, gene discriminatory analysis was performed. Discriminating genes were subsequently categorized based on their GO annotations to facilitate biological and functional interpretations. We also performed pathway analysis using the EASE tool to facilitate functional interpretation of the gene expression data.

Discriminatory analysis of AML and ALL (Pless than or equal to0.001), AML, and CML BC (Pless than or equal to0.001), and B-lineage ALL and T-cell ALL (Pless than or equal to0.002) revealed 367, 214, and 88 genes, respectively (Supplementary Figure S2; Supplementary Tables S5–7). Each gene list was subjected to GO and pathway analyses (Supplementary Tables S1 and S2). A selected set of the GO categories as well as the top 10 ranked genes are shown in Supplementary Table S3.

Comparing AML and ALL, we identified CEBPA, a myeloid transcription factor of importance for maturation along the neutrophilic lineage, and MEIS2, a homeobox-containing gene of the TALE-family, as upregulated in AML. In ALL, CBFB and LEF1, both involved in the CBF-transcription factor complex, were highly expressed. Using pathway analysis, five pathways were found, including proteasome degradation. Genes discriminating AML and CML BC included the transcription factor JUND and the serin/threonine kinase PIM1, upregulated in CML BC. Three pathways were enriched, including heme biosynthesis. In the B-lineage and T-cell ALL comparison, the transcription factor and lineage regulator GATA3 and the signal transducer CD3G were expressed in T-cell ALL, the latter gene involved in the T-cell receptor complex. The B-cell regulator PAX5 was upregulated in B-lineage ALL. Pathway analysis revealed five pathways, including activation of the T-cell receptor.

Genes discriminating the molecular subgroups

Having identified genes discriminating the various clinical subtypes, we next identified genes discriminating cell lines with identical genetic aberrations (Pless than or equal to0.002): RUNX1/CBFA2T1, 35 genes; CBF-leukemias, 62; TCF3/PBX1, 10; and IGH@/MYC, 36 (Supplementary Figure. S3a–d; Supplementary Tables S8–11). Three methods were used to analyze cell lines with 11q23/MLL rearrangements (Pless than or equal to0.002), the first compared B-lineage MLL and B-lineage ALL, 19 genes; the second AML-positive MLL and AML, 18 genes; and the third cell lines with MLL abnormalities and all other cell lines, 40 genes (Figure 2a–c; Supplementary Tables S12–14). All gene lists were subjected to GO analyses (Supplementary Table S1).

Figure 2.
Figure 2 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Genes discriminating cell lines with MLL rearrangements and visualization of the expression signature of tyrosine kinase-encoding genes. Discriminatory analysis (Pless than or equal to0.002) of B-lineage ALLs with or without MLL, AMLs with or without MLL, and cell lines harboring MLL compared to the entire data set, was performed, followed by HCA. The color scale above each cluster indicates the log2 ratios; green represents downregulated genes and red upregulated genes. (a) Cluster analysis using Euclidean distance and complete linkage of the 21 reporters discriminating B-lineage ALLs (blue) with or without (red) MLL. (b) HCA using Pearson correlation and complete linkage of the 24 reporters discriminating AMLs with (red) or without MLL (blue). (c) Clustering using Pearson correlation and complete linkage of the 48 reporters discriminating cell lines harboring MLL rearrangements (blue) from all other lineages (red). (d) Visualization of the expression of the 41 tyrosine kinase-encoding genes. Clustering was only applied on the genes; the order of the cell lines is according to the unsupervised HCA in Figure 1a. The cell lines are colore coded as follows: AML (red), CML BC (green), B-lineage ALL (blue), and T-cell ALL (orange).

Full figure and legend (479K)

Two cell lines harbored RUNX1/CBFA2T1 and among the upregulated genes CBFA2T1, were found, but no enriched GO categories were identified. Given the fact that the two cell lines with RUNX1/CBFA2T1 and the cell line with CBFB/MYH11 cosegregated upon unsupervised HCA (Figure 1c) and that both fusion genes converge on the CBF-transcription factor complex,3 we performed discriminatory analysis of AMLs with (herein designated CBF-leukemias) or without these two genetic rearrangements. The PIK3R1, a regulatory subunit of the PI3-kinase pathway, were upregulated and three GO categories were enriched. Comparing TCF3/PBX1 with other B-lineage ALLs revealed two enriched categories: signal transducer activity and transcription factor activity. The first category included genes such as MAP3K1 and the latter PBX1. In cell lines with IGH@/MYC, the signal transducer BMP7 was upregulated and the transcription factor DP1 was downregulated. Two categories were enriched. In the B-lineage ALL and MLL comparison, two GO categories were identified, and the transcription factor and homeobox-containing gene PBX3 was upregulated. We also compared AMLs with or without MLL rearrangements, identifying PBX3; no enriched category was found. Finally, when cell lines with MLL abnormalities were compared to all other cell lines, we again identified PBX3, suggesting an important role of this gene in MLL-positive leukemias. GO analysis revealed two enriched categories.

Correlation of deregulated gene expression with their chromosomal localization and expression pattern of tyrosine kinase-encoding genes

In order to investigate if the deregulated genes identified by discriminatory analysis were associated with a certain chromosomal region, the EASE tool was used. The calculations were performed for each of the discriminating gene lists, two of which resulted in a significant correlation. Comparing AML and ALL, 21 of the 367 deregulated genes were located on chromosome arm 17p (score 9.21 times 10-12), all but one downregulated in AMLs. Moreover, when in AMLs with rearrangements involving CBF, 15 of the 39 downregulated genes in cell lines with CBF-rearrangements were located on chromosome arm 2q (score 3.98 times 10-11).

Given the recent development of tyrosine-kinase inhibitory drugs,23 we investigated selectively the expression signature of protein tyrosine kinase-encoding genes across all cell lines. Hence, all genes on our array annotated with protein-tyrosine kinase activity (GO: 0004713, n=170) were extracted. In total, 41 genes of this family remained after stringent quality filtering. To investigate the relative expression of these genes, the data set was median-centered. In order to view the tyrosine kinase-encoding genes with respect to the segregation pattern observed using unsupervised methods, clustering was only applied on the genes with the order of the cell lines obtained by unsupervised HCA being maintained (Figure 2d). As seen in Figure 2d, PIM1 and FGFR3 were expressed in CML BC. LCK was upregulated in the majority of the B-lineage ALLs and particularly in T-cell ALL. In a large subset of the B-lineage ALLs, ZAP70 and ERBB2 were expressed. Another cluster of genes, upregulated in T-cell ALL, included FYN and ZAP70. In AML, a cluster of NEK6, PRKCD, and PINK1 was seen. In particular, AMLs with MLL abnormalities showed a high expression of these genes; the other AMLs displayed expression of MST4.

Top

Discussion

Immortalized hematopoietic cell lines have in recent years been instrumental in the cloning, characterization, and functional analyses of a large number of leukemia-associated fusion genes.11 These cell lines have also become increasingly important tools in evaluating biological effects of newly developed inhibitory drugs, capable of interfering with critical signaling pathways or with altered proteins or transcripts resulting from the primary genetic changes.6, 24 While it is generally assumed that cell lines acquire additional aberrations in culture, little is known about to what extent pathogenetically important transcriptional programs remain conserved upon establishment and passaging of individual cell lines.

By applying cDNA microarray analysis and unsupervised algorithms on 40 hematopoietic cell lines with selected primary genetic rearrangements, we here show that the cell lines segregate based on their clinical subtype and, more importantly, with few exceptions, according to their primary genetic change. Thus, cell lines harboring 11q23/MLL rearrangements, TCF3/PBX1, IGH@/MYC, RUNX1/CBFA2T1, P210 BCR/ABL1, ETV6/RUNX1, or CBFB/MYH11, were shown to segregate based on their primary genetic abnormality. This cosegregation suggests that critical transcriptional programs remain conserved in hematopoietic cell lines despite their diverse origin and numerous passages in vitro. We obtained further support for this by analyzing primary ALL and AML samples with 11q23/MLL rearrangements, TCF3/PBX1, RUNX1/CBFA2T1, P190 BCR/ABL1, or ETV6/RUNX1, showing cosegregation of most of these samples with cell lines harboring identical genetic rearrangements. In a similar approach, Fine et al13 recently determined the gene expression pattern in 19 cell lines, mainly of B-lineage origin, with various 11q23/MLL rearrangements, TCF3/PBX1, P190 and P210 BCR/ABL1, ETV6/RUNX1, or E2A/HLF, and in primary acute leukemia samples with MLL/MLLT2(AF4), P190 BCR/ABL1, or ETV6/RUNX1. They also found that cell lines and primary leukemias segregated primarily according to their genetic changes.13 In contrast to Fine et al, we found that cell lines with MLL rearrangements segregated primarily according to their morphologic subtype, although closely together within each lineage. Two AML cases with RUNX1/CBFA2T1 failed to segregate with the two cell lines harboring this abnormality, indicating that such cell lines may have acquired additional genetic changes, altering their transcriptional profiles, when adapting to in vitro conditions.

Surprisingly, the myeloid cell lines HL-60, SIG-M5, and OCI-AML2 segregated with cell lines harboring MLL rearrangements. Although this, in principle, could be a reflection of their lineage specificity, we performed Southern blot analysis to investigate if the MLL gene was rearranged in these cell lines. Interestingly, one cell line, OCI-AML2, established from a patient with AML M4, had a rearranged MLL gene, not previously reported. The SIG-M5 cell line was established from a patient with AML M5a, and given the high frequency of MLL abnormalities in this FAB type,25 it cannot be excluded that this cell line may carry an atypical MLL rearrangement. HL-60 on the other hand, established from a patient with AML M2, has been reported to carry a sideline with del(11)(q23), possibly resulting in loss of hetorozygosity of MLL. The similarity in gene expression pattern of the latter two cell lines with MLL-positive cell lines indicates that they have activated downstream genes similar to MLL target genes, warranting further investigations.

Supervised analysis of cell lines of different lineages generated a list of genes differentially expressed among each subtype. The majority of these genes reflected lineage characteristics, but genes of possible pathogenetic importance were also identified. For example, discriminatory analysis of P210 BCR/ABL1-positive CML BC and AML identified JUND and PIM1, both previously shown to become activated upon BCR/ABL1 expression.21, 26 Interestingly, GO and pathway analyses also revealed upregulation of genes involved in heme biosynthesis. Although the enrichment of erythroid genes could be a reflection of the stem cell involvement in CML BC, it could also reflect that BCR/ABL1 indeed activates erythroid pathways as was recently demonstrated in vitro by retroviral transduction of human CD34+ cord blood cells in which BCR/ABL1 causes erythroid lineage switching.27

By comparing cell lines harboring specific genetic aberrations, we identified a number of genes correlating with genetic subtypes, providing biological insights into the downstream genes deregulated by the primary genetic change. Three different discriminatory analyses were performed to identify genes correlating with MLL and all three independently identified the homeobox transcription factor PBX3 as upregulated, indicating that expression of PBX3 may be important for MLL-mediated transformation. Interestingly, PBX3 was recently shown to become upregulated after experimentally induced expression of MLL/MLLT1(ENL) in mouse hematopoietic cells.28 In our data set, two cell lines harbored RUNX1/CBFA2T1 and CBFA2T1 was upregulated, as has previously also been identified in microarray analyses of primary AML with this fusion gene.29 Analysis of TCF3/PBX1-positive cell lines revealed upregulation of PBX1, as has also been seen in primary ALLs carrying this fusion gene.8 Two cell lines harbored IGH@/MYC, and both showed downregulation of DP1, possibly mediating an antiapoptotic effect by loss of E2F1-mediated apoptosis.30

We also determined the chromosomal location of the deregulated genes in cell lines of different clinical subtypes or with different genetic abnormalities. A significant correlation was obtained comparing AML and ALL; 20 of the 21 genes located on chromosome arm 17p were downregulated in the AML cell lines. Loss of chromosome 17 material, in particular of 17p, is one of the most frequent cytogenetic alterations seen in AML31 and our results suggest that large chromosomal deletions may result in downregulation of several genes located in the lost regions. Interestingly, in AMLs with rearrangements involving the CBF-transcription factor complex, 15 genes, located on the distal part of 2q33–37, were downregulated. Loss of 2q is not a common cytogenetic abnormality in acute leukemias, but the present finding may suggest that cryptic deletions at this locus are present in CBF-leukemias or, alternatively, that deregulation of the CBF-transcription factor complex may lead to repression of genes on selected chromosomal loci, certainly warranting future investigations.

It is well known that genes encoding protein tyrosine kinases frequently are implicated in human cancer,32 and given the development of selective tyrosine kinase inhibitors,23 such genes have been targeted and extensively studied in vitro as well as in vivo. In order to investigate the expression of tyrosine kinase-encoding genes, we selectively analyzed these genes in our data set. Interestingly, FGFR3 was upregulated in CML BC cell lines. Recently, the expression of FGFR3 was found to correlate with disease relapse in a series of transplanted CML patients33 and to be expressed at increased levels in CD34+ CML cells.34 Hence, the high expression of FGFR3 in CML BC cell lines suggests that targeting of FGFR3 using available inhibitors may provide an important additional treatment modality of BCR/ABL1-positive leukemias.

Apart from demonstrating that acute leukemias and immortalized hematologic cell lines of similar clinical subtypes share a molecular signature, we also show that cell lines and primary leukemic samples harboring identical or closely related primary genetic aberrations cosegregate. Moreover, we have identified a large number of differentially expressed genes that correlate with the most frequently observed fusion genes in leukemia and provide examples of genes, for example, FGFR3, which may be evaluated as novel targets for treatment. Furthermore, the present study also provides an extensive gene expression data set of hematopoietic cell lines, offering an important source for further explorations to obtain additional insights into the molecular consequences of recurrent genetic changes in leukemia.

Top

References

  1. Rowley JD. The critical role of chromosome translocations in human leukemias. Annu Rev Genet 1998; 32: 495−519. | Article | PubMed | ISI | ChemPort |
  2. Mitelman F, Johansson B & Mertens F. Fusion genes and rearranged genes as a linear function of chromosome aberrations in cancer. Nat Genet 2004; 36: 331−334. | Article | PubMed | ChemPort |
  3. Downing JR. AML1/CBFbeta transcription complex: its role in normal hematopoiesis and leukemia. Leukemia 2001; 15: 664−665. | PubMed | ChemPort |
  4. Scandura JM, Boccuni P, Cammenga J & Nimer SD. Transcription factor fusions in acute leukemia: variations on a theme. Oncogene 2002; 21: 3422−3444. | Article | PubMed | ChemPort |
  5. Cross NC & Reiter A. Tyrosine kinase fusion genes in chronic myeloproliferative diseases. Leukemia 2002; 16: 1207−1212. | Article | PubMed | ChemPort |
  6. Deininger MW, Goldman JM & Melo JV. The molecular biology of chronic myeloid leukemia. Blood 2000; 96: 3343−3356. | PubMed | ISI | ChemPort |
  7. Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M & Mesirov JP et al.. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science 1999; 286: 531−537. | Article | PubMed | ISI | ChemPort |
  8. Yeoh EJ, Ross ME, Shurtleff SA, Williams WK, Patel D & Mahfouz R et al.. Classification, subtype discovery, and prediction of outcome in pediatric acute lymphoblastic leukemia by gene expression profiling. Cancer Cell 2002; 1: 133−143. | Article | PubMed | ISI | ChemPort |
  9. Valk PJ, Verhaak RG, Beijen MA, Erpelinck CA, Barjesteh van Waalwijk van Doorn-Khosrovani S & Boer JM et al.. Prognostically useful gene-expression profiles in acute myeloid leukemia. N Engl J Med 2004; 350: 1617−1628. | Article | PubMed | ChemPort |
  10. Bullinger L, Dohner K, Bair E, Frohling S, Schlenk RF & Tibshirani R et al.. Use of gene-expression profiling to identify prognostic subclasses in adult acute myeloid leukemia. N Engl J Med 2004; 350: 1605−1616. | Article | PubMed | ChemPort |
  11. Drexler HG, Matsuo AY & MacLeod RA. Continuous hematopoietic cell lines as model systems for leukemia-lymphoma research. Leuk Res 2000; 24: 881−911. | Article | PubMed | ChemPort |
  12. Druker BJ & Lydon NB. Lessons learned from the development of an abl tyrosine kinase inhibitor for chronic myelogenous leukemia. J Clin Invest 2000; 105: 3−7. | PubMed | ISI | ChemPort |
  13. Fine BM, Stanulla M, Schrappe M, Ho M, Viehmann S & Harbott J et al.. Gene expression patterns associated with recurrent chromosomal translocations in acute lymphoblastic leukemia. Blood 2004; 103: 1043−1049. | Article | PubMed | ChemPort |
  14. Saal LH, Troein C, Vallon-Christersson J, Gruvberger S, Borg A & Peterson C. BioArray Software Environment (BASE): a platform for comprehensive management and analysis of microarray data. Genome Biol 2002; 3 SOFTWARE0003. | ChemPort |
  15. Yang YH, Dudoit S, Luu P, Lin DM, Peng V & Ngai J et al.. Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variation. Nucleic Acids Res 2002; 30: e15. | Article | PubMed |
  16. Saeed AI, Sharov V, White J, Li J, Liang W & Bhagabati N et al.. TM4: a free, open-source system for microarray data management and analysis. Biotechniques 2003; 34: 374−378. | PubMed | ChemPort |
  17. Nilsson J, Fioretos T, Höglund M & Fontes M. Approximate geodesic distances reveal biologically relevant structures in microarray data. Bioinformatics 2004; 20: 874−880. | Article | PubMed | ChemPort |
  18. Tenenbaum JB, de Silva V & Langford JC. A global geometric framework for nonlinear dimensionality reduction. Science 2000; 290: 2319−2323. | Article | PubMed | ISI | ChemPort |
  19. Hosack DA, Dennis G, Jr, Sherman BT, Lane HC & Lempicki RA. Identifying biological themes within lists of genes with EASE. Genome Biol 2003; 4: R70. | Article | PubMed |
  20. Sambrook J, Fritsch EF & Maniatis T. Molecular Cloning: a Laboratory Manual 2nd edn. NY: Cold Spring Harbor 1989;.
  21. Håkansson P, Segal D, Lassen C, Gullberg U, Morse H & Fioretos T et al.. Identification of genes differentially regulated by the P210 BCR/ABL1 fusion oncogene using cDNA microarrays. Exp Hematol 2004; 32: 476−482. | Article | PubMed | ChemPort |
  22. Fioretos T, Nilsson PG, Aman P, Heim S, Kristoffersson U & Malm C et al.. Clinical impact of breakpoint position within M-bcr in chronic myeloid leukemia. Leukemia 1993; 7: 1225−1231. | PubMed | ChemPort |
  23. Gschwind A, Fischer OM & Ullrich A. Timeline: the discovery of receptor tyrosine kinases: targets for cancer therapy. Nat Rev Cancer 2004; 4: 361−370. | Article | PubMed | ISI | ChemPort |
  24. Stirewalt DL, Meshinchi S & Radich JP. Molecular targets in acute myelogenous leukemia. Blood Rev 2003; 17: 15−23. | Article | PubMed |
  25. Schoch C, Schnittger S, Klaus M, Kern W, Hiddemann W & Haferlach T. AML with 11q23/MLL abnormalities as defined by the WHO classification: incidence, partner chromosomes, FAB subtype, age distribution, and prognostic impact in an unselected series of 1897 cytogenetically analyzed AML cases. Blood 2003; 102: 2395−2402. | Article | PubMed | ChemPort |
  26. Nosaka T & Kitamura T. Pim-1 expression is sufficient to induce cytokine independence in murine hematopoietic cells, but is dispensable for BCR-ABL-mediated transformation. Exp Hematol 2002; 30: 697−702. | Article | PubMed | ChemPort |
  27. Chalandon Y, Jiang X, Hazlewood G, Loutet S, Conneally E & Eaves A et al.. Modulation of p210(BCR-ABL) activity in transduced primary human hematopoietic cells controls lineage programming. Blood 2002; 99: 3197−3204. | Article | PubMed | ChemPort |
  28. Zeisig BB, Milne T, Garcia-Cuellar MP, Schreiner S, Martin ME & Fuchs U et al.. Hoxa9 and Meis1 are key targets for MLL-ENL-mediated cellular immortalization. Mol Cell Biol 2004; 24: 617−628. | Article | PubMed | ChemPort |
  29. Schoch C, Kohlmann A, Schnittger S, Brors B, Dugas M & Mergenthaler S et al.. Acute myeloid leukemias with reciprocal rearrangements can be distinguished by specific gene expression profiles. Proc Natl Acad Sci USA 2002; 99: 10008−10013. | Article | PubMed | ChemPort |
  30. Hitchens MR & Robbins PD. The role of the transcription factor DP in apoptosis. Apoptosis 2003; 8: 461−468. | Article | PubMed | ChemPort |
  31. Mitelman F, Johansson B & Mertens F. Mitelman database of chromosome aberrations in cancer. http://cgap.nci.nih.gov/chromosomes/mitelman.
  32. Futreal PA, Coin L, Marshall M, Down T, Hubbard T & Wooster R et al.. A census of human cancer genes. Nat Rev Cancer 2004; 4: 177−183. | Article | PubMed | ISI | ChemPort |
  33. Dvorakova D, Krejci P, Mayer J, Fajkus J, Hampl A & Dvorak P. Changes in the expression of FGFR3 in patients with chronic myeloid leukaemia receiving transplants of allogeneic peripheral blood stem cells. Br J Haematol 2001; 113: 832−835. | Article | PubMed | ChemPort |
  34. Dvorak P, Dvorakova D, Doubek M, Faitova J, Pacholikova J & Hampl A et al.. Increased expression of fibroblast growth factor receptor 3 in CD34+ BCR-ABL+ cells from patients with chronic myeloid leukemia. Leukemia 2003; 17: 2418−2425. | Article | PubMed | ChemPort |
Top

Acknowledgements

We thank the Swegene DNA microarray resource center in Lund, supported by the Knut and Alice Wallenberg Foundation through the Swegene consortium, Eva Pålsson for cytogenetic analysis, and Srinivas Veerla for useful computational program construction. PE was supported by the Swedish Foundation for Strategic Research through the Lund Center for Stem Cell Biology and Cell Therapy. JN was supported by the Knowledge Foundation and AstraZeneca R&D Lund. This study was supported by the Swedish Cancer Society, The Swedish Children's Cancer Foundation, the Medical Faculty of Lund University, and the IngaBritt and Arne Lundberg foundation.

Supplementary Information

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