Gene expression profiling in the myelodysplastic syndromes using cDNA microarray technology

Authors


Jacqueline Boultwood, Leukaemia Research Fund Molecular Haematology Unit, Nuffield Department of Clinical Laboratory Sciences, University of Oxford, John Radcliffe Hospital, Oxford OX3 9DU, UK.
E-mail: jacqueline.boultwood@ndcls.ox.ac.uk

Summary

The myelodysplastic syndromes (MDS) comprise a heterogeneous group of clonal disorders of the haematopoietic stem cell and primarily involve cells of the myeloid lineage. Using cDNA microarrays comprising 6000 human genes, we studied the gene expression profiles in the neutrophils of 21 MDS patients, seven of which had the 5q- syndrome, and two acute myeloid leukaemia (AML) patients when compared with the neutrophils from pooled healthy controls. Data analysis showed a high level of heterogeneity of gene expression between MDS patients, most probably reflecting the underlying karyotypic and genetic heterogeneity. Nevertheless, several genes were commonly up or down-regulated in MDS. The most up-regulated genes included RAB20, ARG1, ZNF183 and ACPL. The RAB20 gene is a member of the Ras gene superfamily and ARG1 promotes cellular proliferation. The most down-regulated genes include COX2, CD18, FOS and IL7R. COX2 is anti-apoptotic and promotes cell survival. Many genes were identified that are differentially expressed in the different MDS subtypes and AML. A subset of genes was able to discriminate patients with the 5q- syndrome from patients with refractory anaemia and a normal karyotype. The microarray expression results for several genes were confirmed by real-time quantitative polymerase chain reaction. The MDS-specific expression changes identified are likely to be biologically important in the pathophysiology of this disorder.

The myelodysplastic syndromes (MDS) are a heterogeneous group of haematopoietic malignancies, characterized by blood cytopenias, ineffective haematopoiesis and a hypercellular bone marrow (Heaney & Golde, 1999). Dysplasia of at least one lineage (myeloid, erythroid or megakaryocyte-platelet) is a characteristic feature of MDS. The MDS are preleukaemic conditions with transformation into acute myeloid leukaemia (AML) occurring in approximately 30–40% of cases (Heaney & Golde, 1999). MDS-associated leukaemia is rarely cured by chemotherapy and patients have a poor prognosis (Heaney & Golde, 1999). The Classification of MDS is based on morphological anomalies according to the French American British (FAB) Cooperative Study Group (Bennett et al, 1982). This group defined five subtypes based on the percentage of immature blasts in the bone marrow, the presence of ring sideroblasts and the degree of monocytosis. The five groups are refractory anaemia (RA), RA with ringed sideroblasts (RARS), RA with excess blasts (RAEB), RAEB in transformation (RAEB-t) and chronic myelomonocytic leukaemia (CMML) (Bennett et al, 1982). More recently, this classification has been modified by the World Health Organization (WHO) (Harris et al, 1999). According to the International Scoring System for evaluating prognosis in MDS, the major variables having an impact on disease outcome for evolution to AML are cytogenetic abnormalities, percentage of bone marrow myeloblasts, and numbers of cytopenias (Greenberg et al, 1997). Although the precise incidence of MDS is uncertain, it has become clear that MDS is at least as common as AML and there is considerable overlap between these two conditions (Steensma & Tefferi, 2003).

In MDS, clonal karyotypic abnormalities are frequently observed, occurring in approximately 40–50% of cases of primary MDS and 90% of cases in therapy-related MDS (Heaney & Golde, 1999; Olney & Le Beau, 2001). The bone marrow karyotypes of MDS are characterized by the loss of genetic material and, in contrast to the acute leukaemias, balanced translocations are rarely found (Johansson et al, 1993; Boultwood & Fidler, 1995; Olney & Le Beau, 2001). The del(5q) or -5, del(7q) or -7, del(20q), del(11q) and -Y are among the most frequently reported cytogenetic abnormalities in MDS (Johansson et al, 1993; Boultwood & Fidler, 1995; Olney & Le Beau, 2001).

The underlying causes of MDS are largely unknown. Few specific gene abnormalities have been causally implicated in the development and/or progression of MDS and the molecular pathogenesis of MDS remains poorly understood (Fenaux, 2001). Point mutation of members of the RAS family of proto-oncogenes represent the molecular abnormalities most frequently reported in MDS followed by p15 promoter hypermethylation, p53 mutations and FLT3 duplications. All of these abnormalities have been associated (to some extent) with disease progression to AML and none are specific for MDS; they all occur at a higher frequency in AML (Fenaux, 2001). Gene expression microarray technology has the power to investigate fundamental cancer biology at the molecular level (Golub et al, 1999; Gibbons et al, 2003; Pellagatti et al, 2003), and has the potential to provide novel insights into the molecular pathogenesis of MDS and to identify genes/new molecular pathways that are important in disease evolution. RNA extracted from neutrophils has been successfully used for the investigation of gene expression levels in myeloid malignancies (e.g. in the study of the PRV1 gene in polycythaemia vera) (Klippel et al, 2003). In order to investigate the molecular defects underlying MDS we have used cDNA microarray technology to determine the gene expression profile in neutrophils obtained from MDS patients.

Materials and methods

Sample collection and cell separation

Twenty-one patients with MDS were included in the study. Classification was according to the FAB criteria (Bennett et al, 1982) and the patients were selected solely on the basis of having MDS. Two patients with AML at diagnosis were also included in the study and were classified according to the FAB criteria. At the time of investigation 17 patients had RA (included seven cases with 5q- syndrome), two RARS, two RAEB and two AML (one with MDS-AML and one with M1). From the patient with MDS-AML, samples were available before and after disease progression. Peripheral blood samples were obtained from the patients with MDS or AML and seven healthy volunteers. The study was approved by the ethics committee (C00.198) and informed consent was obtained. Neutrophils were isolated using Histopaque (Sigma-Aldrich, Dorset, UK) and pelleted after hypotonic lysis of erythrocytes and two washes in phosphate-buffered saline (PBS) (Boyum, 1968). The purity of the neutrophil populations isolated from MDS and AML patients and from healthy volunteers was consistently high and was within the range 95–98% as assessed by standard morphology on Wright–Giemsa-stained cytospin preparations.

RNA extraction and pooling

Total RNA was extracted using TRIZOL (Invitrogen-Life Technologies, Paisley, UK) following the protocol supplied by the manufacturer. Briefly, neutrophil pellets were resuspended in TRIZOL, incubated at room temperature for 5 min and, after addition of chloroform and separation of the phases, RNA was precipitated with isopropanol from the recovered aqueous phase. The RNA pellet was resuspended in nuclease-free water and quantified by spectrophotometer measurement. An aliquot of each sample was conserved for gel electrophoresis and inspection of RNA quality. Equal amounts of RNA extracted from the neutrophil fraction from seven healthy volunteers were mixed to generate a pool of normal control RNA.

RNA labelling and hybridization

In each microarray experiment one patient RNA was compared with the normal RNA pool using a T7-based amplification method (MessageAmp aRNA Kit; Ambion Inc., Austin, TX, USA) (Van Gelder et al, 1990): 1 μg of total RNA was reverse-transcribed using a T7-oligodT primer and then labelled aRNAs produced via in vitro transcription in the presence of either Cy3-UTP or Cy5-UTP (Amersham Biosciences, Uppsala, Sweden). The labelled aRNAs were then purified and hybridized competitively to a microarray slide (Sanger Institute, Hinxton, Cambridge, http://www.sanger.ac.uk/Projects/Microarrays/) containing 10 000 spots representative of 6000 known full length human genes, including many housekeeping genes such as beta actin, glycerol kinase and glucose-6-phosphate dehydrogenase. cDNAs from five bacterial genes that do not cross-hybridize with human material were used as hybridization controls. Several genes were represented by more than one different clone on the array. Hybridization occurred at 47°C overnight and then the slides were washed at room temperature once in 2X saline sodium citrate (SSC) for 5 min, twice in 0·1X SSC, 0·1% sodium dodecyl sulphate (SDS) for 30 min and twice in 0·1X SSC for 5 min. Each experiment was performed in duplicate.

Scanning and analysis

Slides were scanned immediately (ScanArray 4000; PerkinElmer, Boston, MA, USA) with a 10-μm resolution and the generated tagged image file format (TIFF) images imported in QuantArray 3·0 (PerkinElmer). After image and grid alignment and spot location, the Cy5 and Cy3 intensities were normalized to the median value of the whole array and the ratio for every spot was obtained. Analysis was performed with GeneSpring 6·0 (Silicon Genetics, Redwood City, CA, USA). Intensity-dependent (Lowess) normalization was applied to all experiments. Quality control criteria were applied to exclude genes with signal intensities in the range of control negative spots; 3361 genes passed this filter. Statistical analysis was performed on log-transformed data using Welch's approximate t-test or analysis of variance (anova) and multiple testing corrections to control false discovery rate. Hierarchical clustering was performed using standard correlation.

Real-time quantitative polymerase chain reaction (PCR)

RETROscript kit (Ambion Inc.) was used to reverse transcribe 2 μg of total RNA from the 18 patients for which enough material was left and from each of the seven healthy controls included in the pool. The generated first-strand cDNA was diluted and used as a template for real-time quantitative PCR analysis (TaqMan) (Holland et al, 1991; Heid et al, 1996). The expression level of the β2-microglobulin gene was used to normalize for differences in input cDNA. For all the selected genes and the control gene, predeveloped TaqMan Assays were used (Assays-on-Demand; Applied Biosystems, Foster City, CA, USA), which have been shown to be mRNA-specific. PCR reactions occurred in a volume of 25 μl in the presence of 1·25 μl of 20X probe mix and 12·5 μl of TaqMan 2X Universal PCR mastermix. Five microlitres of diluted cDNA solution was used as a template in each reaction. At each cycle of the PCR process, the increase of the fluorescence was monitored by an ABI Prism 5700 Sequence Detection System (Applied Biosystems) (Holland et al, 1991; Heid et al, 1996); the thermal cycling programme was: a first step at 50°C for 2 min, then 95°C for 10 min, followed by 40 cycles of amplification at 95°C for 15 s and 60°C for 1 min. Each sample was performed in triplicate and a reverse-transcriptase negative control was also tested to exclude any contaminating DNA amplification.

Results

Gene expression analysis of neutrophils obtained from MDS patients compared with neutrophils obtained from healthy individuals showed a high level of heterogeneity. Nevertheless, several genes were commonly up- or down-regulated in MDS. The most up-regulated genes (ratio >2·0 in at least nine MDS patients) were RAB20, zinc finger protein 183 (ZNF183), liver arginase (ARG1) and interleukin 18 receptor accessory protein (ACPL) (Table 1). Also cellular repressor of E1A-stimulated gene (CREG) was up-regulated in approximately 20% of the MDS patients (Table 1). RAB20 and ZNF183 were up-regulated in five of the seven patients with 5q- syndrome. The most down-regulated genes (ratio <0·5 in at least 10 MDS patients) were cyclooxygenase 2 (COX2), CD18 antigen (CD18), G protein-coupled receptor 105 (KIAA0001), v-fos FBJ murine osteosarcoma viral oncogene homologue (FOS), interleukin-7 receptor (IL7R), immune activation 2 (ACT2) and interferon-induced protein 56 (IFI56) (Table 1). IFI56 was down-regulated in all seven patients with 5q- syndrome and the 2′–5′ oligoadenylate synthetase 1 gene (OAS1) in six of seven patients with 5q- syndrome.

Table I.  List of the genes most commonly up- or down-regulated in the neutrophils of MDS patients.
GeneAccession numberChromosomal locationFunctionNeutrophil gene expression
RAB20AK00043613q34GTP-bindingUp-regulated in 12/21 patients
ZNF183AC002477Xq25-q26Nucleic acid bindingUp-regulated in 10/21 patients
ARG1M145026q23Conversion of arginine to ornithineUp-regulated in 9/21 patients
ACPLAF0773462p24.3-p24.1Cell signallingUp-regulated in 9/21 patients
CREGAF0845231q24Control of cell growthUp-regulated in 4/21 patients
COX2M901001q25.2-q25.3Regulation of apoptosisDown-regulated in 15/21 patients
IFI56X0355710q25-q26UnknownDown-regulated in 13/21 patients
CD18NM_00021121q22.3Leucocyte adhesionDown-regulated in 12/21 patients
ACT2J0413017q12Immune responseDown-regulated in 11/21 patients
IL7RM296965p13Signal transductionDown-regulated in 11/21 patients
KIAA0001D136263q21-q25G-protein coupled receptorDown-regulated in 10/21 patients
FOSK0065014q24.3Gene TranscriptionDown-regulated in 10/21 patients

When investigating differences among the MDS subtypes, RA, RARS and RAEB, and AML, using 0·05 P-value cut-off and Bonferroni multiple testing correction, 27 significantly different genes were identified. Hierarchical clustering performed using these 27 genes grouped the two patients with AML together with the two MDS patients with RAEB in a separate family (Fig 1). Also, the two MDS patients with RARS clustered together and showed a very similar expression profile. Figure 1 also shows the changing pattern of gene expression in MDS neutrophils during disease evolution in one case for which samples were available before (sample labelled RA*) and after (sample labelled AML*) leukaemic transformation.

Figure 1.

Hierarchical clustering in MDS and AML. Hierarchical clustering of 27 differentially expressed genes in the different MDS subtypes and AML. Each column represents a single gene on the microarray and each row a separate patient neutrophil sample. The (*) indicates the one case in which samples were available before (sample labelled RA*) and after (sample labelled AML*) leukaemic transformation.

Statistical analysis (0·01 P-value cut-off and Benjamini and Hochberg multiple testing correction) showed that the expression level of 71 genes was significantly different between the seven patients with 5q- syndrome and six patients with RA and a normal karyotype. These 71 genes were used for hierarchical clustering and the resulting tree could separate the two groups in distinct families (Fig 2). For some of these 71 genes, the levels of expression in the 5q- syndrome patient group had no overlap with the levels of expression in the RA with normal karyotype patient group. Cyclic AMP-responsive element modulator (CREM), cylindromatosis tumour suppressor gene (CYLD) and RB1-inducible coiled-coil 1 (RB1CC1) were expressed at a higher level in the 5q- syndrome patient group, whilst antioxidant protein 1 (ATOX1) and cofactor required for SP1 transcriptional activation subunit 9 (CRSP9) were expressed at a lower level.

Figure 2.

Hierarchical clustering in patients with 5q- syndrome and patients with RA and a normal karyotype. Hierarchical clustering of 71 genes differentially expressed in the neutrophils of seven patients with 5q- syndrome (no. 1–7) compared with the neutrophils of six patients with RA and a normal karyotype (no. 8–13). Each column represents a single gene on the microarray and each row a separate patient neutrophil sample.

The array results for several genes were validated using the TaqMan 5′ nuclease fluorogenic quantitative PCR assay. Relative expression levels obtained with the microarrays were compared with the expression levels determined using real-time quantitative PCR. We chose to investigate RAB20, CREG, COX2 and FOS (Fig 3). The concordance between the two assays for all genes validated was high, indicating a good level of agreement between the two assays for the identification of deregulated genes.

Figure 3.

Confirmation of gene expression data. Comparison of the ratios obtained from real-time quantitative PCR (white bars) and microarray experiments (black bars) for four differentially expressed genes in each of the 18 patients from which enough material was available for real-time quantitative PCR.

Discussion

The MDS arise from the haematopoietic stem cell and primarily involve cells of the myeloid lineage. We have used cDNA microarray technology to determine gene expression profiles in the neutrophils of 21 MDS patients, seven of which had the 5q- syndrome, when compared with the neutrophils of healthy controls. The gene expression profiles of neutrophils from MDS patients were clearly distinct from those obtained from the neutrophils of healthy controls. However, the expression data showed a high level of heterogeneity between patients, most probably reflecting disease heterogeneity in MDS. There are probably many factors involved but clearly karyotype and disease evolution play a major part. Nevertheless, several genes were commonly up- or down-regulated. The most commonly up-regulated genes were RAB20, ZNF183, ARG1 and ACPL. The most commonly down-regulated genes were COX2, CD18, KIAA0001, FOS, IL7R, ACT2 and IFI56.

Several of the genes found to be up-regulated in a high proportion of the MDS patients are likely to be biologically important in the pathophysiology of this disorder. For example, we have shown that the RAB20 gene, which encodes a small GTP-binding protein and is a member of the Ras superfamily (Oxford & Theodorescu, 2003), was up-regulated in the neutrophils of more than 50% of the MDS patients included in this study by two to threefold. Rab proteins are believed to be essential components of the membrane-trafficking mechanism of eukaryotic cells (Oxford & Theodorescu, 2003). Other members of the Rab gene family have been shown to be up-regulated in certain solid tumours, for example, RAB10 and RAB25 are up-regulated in hepatocellular carcinoma and up-regulation of RAB2 has been associated with murine lung tumour progression (Yao et al, 2002). Similarly, the ARG1 gene is up-regulated two- to fourfold in nine of 21 MDS patients included in this study. ARG1, previously thought to be restricted to the liver, has been shown to be present in activated neutrophils and macrophages of the rat and high arginase activity may be found at inflammatory sites (Waddington et al, 1998). Arginase converts l-arginine to l-ornithine, which is the precursor of polyamines that are essential components of cell proliferation (Jenkinson et al, 1996). ARG1 has been shown to be up-regulated in several solid tumours, including breast cancer, and is associated with the development of metastases in a murine model of lung carcinoma (Singh et al, 2000; Margalit et al, 2003). We have shown that ARG1 was up-regulated in MDS where it may promote cell proliferation and matrix production. Neutrophils express interleukin-18 (IL18) receptor and IL18 plays a role in activating neutrophils (Leung et al, 2001). Interestingly the ACPL gene, encoding the interleukin-18 receptor accessory protein required for signalling by IL18 (Born et al, 1998), was up-regulated by greater than twofold in nine of 21 MDS patients included in this study. Also CREG, a cellular repressor of E1A-stimulated genes and inhibitor of cell growth (Di Bacco & Gill, 2003), was up-regulated in approximately 20% of the MDS patients by two- to fourfold.

We also found that several genes were commonly down-regulated in the neutrophils of the MDS patients included in this study. For example, COX2 was down-regulated in neutrophils of approximately 75% of the MDS patients by two- to fourfold. COX2 converts arachidonic acid to prostaglandins and is angiogenic and anti-apoptotic (Fosslien, 2001). In marked contrast to MDS, COX2 over-expression has been noted in many solid tumours, including colorectal and breast tumours and the myeloproliferative disorder, chronic myeloid leukaemia (Giles et al, 2002; Watanabe et al, 2003). Interestingly, targeted inhibition of COX2 can lead to growth inhibition and apoptosis of several solid cancers (Gasparini et al, 2003) and we speculate that the reduced levels of COX2 in MDS may contribute to the high levels of intramedullary cell apoptosis and blood cytopenias observed in this disorder.

The neutrophils of MDS patients often have a dysplastic morphology and have been shown to be dysfunctional (Mazzone et al, 1996; Heaney & Golde, 1999). Several groups have shown that the expression of CD18 on the neutrophils of MDS patients is reduced compared with controls and it has been suggested that neutrophil dysfunction and the resultant increased susceptibility to bacterial infections observed in MDS patients may be correlated with decreased expression of this surface adhesion molecule (Mazzone et al, 1996; Ohsaka et al, 1997). In agreement with these findings, we observed a two- to fourfold decrease in the expression of CD18 in >50% of the MDS patients studied. Similarly, FOS, a member of a family of transcription factors that regulate many cellular processes, including proliferation (van Dam & Castellazzi, 2001) was down-regulated in the neutrophils of approximately 50% of the MDS patients included in this study by two- to fivefold. FOS mRNA levels have been shown to be elevated during myeloid cell differentiation and in terminally differentiated myeloid cells and it has been suggested that fos/jun transcription factors play important roles in promoting myeloid differentiation (Lord et al, 1993). Interestingly, we noted that the pattern of expression of FOS and COX2 across all the MDS patients studied was very similar. The chemokine ACT2 (also known as MIP-1β) (Napolitano et al, 1991) was also down-regulated in neutrophils of approximately 50% of the MDS patients by two- to fourfold.

MDS represents an excellent model of leukaemic development with a progressive increase of blastic bone marrow involvement, but the genetic events that lead to this evolution remain poorly understood (Heaney & Golde, 1999; Fenaux, 2001). We identified two patients with AML from whose circulating neutrophils we were able to extract sufficient RNA. The MDS subtypes, RA, RARS and RAEB, and AML clustered into two major groups: RA/RARS and RAEB/AML. Within the first group the two RARS cases clustered together (Fig 1). Fig 1 also shows the changing pattern of gene expression in MDS neutrophils during disease evolution in one case in which samples were available before (sample labelled RA*) and after (sample labelled AML*) leukaemic transformation.

The del(5q) is the most frequently reported karyotypic abnormality in MDS and is observed in 10–15% of patients (Van den Berghe et al, 1985). The del(5q) occurs as the sole karyotypic abnormality in the 5q- syndrome, the most distinct of the myelodysplastic syndromes(Boultwood et al, 1994). The 5q- syndrome is characterized by a marked female preponderance, RA, hypolobulated megakaryocytes and a low risk of transformation to AML (Boultwood et al, 1994). The relationship between the aetiology of the 5q- syndrome and the other subtypes of RA remains unknown at a molecular level. We have defined and annotated the commonly deleted region (CDR) of the 5q- syndrome (Boultwood et al, 2002). The CDR of the 5q-syndrome is a 1·5-Mb region at 5q32 flanked by the genetic marker D5S413 and the GLRA1 gene and contains approximately 45 genes (Boultwood et al, 2002). The tumour suppressor gene believed to be associated with the development of the 5q- syndrome remains to be identified. In this study a subset of genes was able to discriminate patients with 5q- syndrome from the patients with RA and a normal karyotype. Statistical analysis showed that the expression level of 71 genes was significantly different between the seven 5q- syndrome patients and the six patients with RA and a normal karyotype. These 71 genes were used for hierarchical clustering and the resulting tree could separate the two groups in distinct families (Fig 2). For some of these 71 genes, the levels of expression in the 5q- syndrome patient group had no overlap with the levels of expression in the RA with normal karyotype patient group. CREM, CYLD and RB1CC1 were expressed at higher levels in the 5q- syndrome patient group, whilst ATOX1 and CRSP9 were expressed at lower levels. Of these 71 genes, only five (CRSP9, KIAA0209, GRX, ATOX1 and CGI-109) mapped to the long arm of chromosome 5 and of these five genes, only ATOX1 maps within the CDR of the 5q- syndrome and CRSP9 adjacent to it.

Recently two other groups have applied expression microarray analysis to the study of MDS. Miyazato et al (2001) compared the expression profiles of over 2000 genes in the haematopoietic stem cell population of five patients with MDS and five with AML and identified a number of genes, including delta-like protein 1 (DLK1), that were differentially expressed between the two groups. Similarly, Hofmann et al (2002) investigated expression profiles in the CD34+ cells of seven patients with low risk MDS and four with high risk MDS. This group confirmed the up-regulation of the DLK1 gene in some patients with MDS and also showed that class membership prediction analysis as well as hierarchical clustering based on the expression data of 11 selected genes enabled the discrimination of low risk from high risk MDS cases (Hofmann et al, 2002). We were unable to accurately assess the expression levels of DLK1 in the neutrophils of patients and controls in our study as the intensity of the signal on the arrays for this gene was very low.

This represents the first study to determine gene expression profiles in the neutrophils of MDS patients. We have identified many genes that are differentially expressed in the neutrophils of MDS patients. These MDS-specific expression changes are likely to be biologically important in the pathophysiology of this disorder. Many genes were identified that are differentially expressed in the different MDS subtypes and AML. Furthermore, a subset of genes was able to discriminate patients with 5q- syndrome from patients with RA and with a normal karyotype. Peripheral blood neutrophils represent a readily accessible source of material and it is possible that we may be able to follow disease progression in MDS by their study.

Acknowledgments

This study was funded by the Leukaemia Research Fund of the UK.

Ancillary