Gene expression profiling of CD34+ cells in patients with the 5q− syndrome

Authors


Dr Jacqueline Boultwood, University Reader at University of Oxford, Co-director of the LRF Molecular Haematology Unit, Nuffield Department of Clinical Laboratory Sciences, John Radcliffe Hospital, Oxford OX3 9DU, UK. E-mail: jacqueline.boultwood@ndcls.ox.ac.uk

Summary

The transcriptome of the CD34+ cells was determined in a group of 10 patients with the 5q− syndrome using a comprehensive array platform, and was compared with the transcriptome of CD34+ cells from 16 healthy control subjects and 14 patients with refractory anaemia and a normal karyotype. The majority of the genes assigned to the commonly deleted region (CDR) of the 5q− syndrome at 5q31–q32 showed a reduction in expression levels in patients with the 5q− syndrome, consistent with the loss of one allele. Candidate genes showing haploinsufficiency in the 5q− syndrome included the tumour suppressor gene SPARC and RPS14, a component of the 40S ribosomal subunit. Two genes mapping to the CDR, RBM22 and CSNK1A1, showed a >50% reduction in gene expression, consistent with the downregulation of the remaining allele. This study identified several significantly deregulated gene pathways in patients with the 5q− syndrome and gene pathway analysis data supports the proposal that SPARC may play a role in the pathogenesis of the 5q− syndrome. This study suggests that several of the genes mapping to the CDR of the 5q− syndrome play a role in the pathogenesis of this disorder.

The del(5q) is the most commonly reported deletion in de novo myelodysplastic syndromes (MDS) and is found in 10–15% of all patients. The del(5q) occurs as the sole karyotypic abnormality in the 5q− syndrome, which is the most distinct of the MDS (Boultwood et al, 1994; Giagounidis et al, 2004). The 5q− syndrome was first described by Van den Berghe et al (1974) who reported the consistent association of the 5q− abnormality with the following haematological features: macrocytosis, anaemia, normal or high platelet count and hypolobulated megakaryocytes in the bone marrow. From the earliest series a female preponderance and a good prognosis have been noted (Boultwood et al, 1994; Giagounidis et al, 2004). The 5q− syndrome is characterized by a clear genotype–phenotype relationship which is not seen in most other chromosomal deletions in MDS and acute myeloid leukaemia (AML). Indeed, the 5q− syndrome is now recognized as a distinct clinical entity according to the World Health Organization classification and is defined as patients with a medullary blast count of <5% and the presence of the del(5q) as the sole karyotypic abnormality (Vardiman et al, 2002). Until recently, most patients with the 5q− syndrome were treated with best supportive care only. However, the immunomodulatory drug lenalidomide, which has been shown to have dramatic therapeutic effects in patients with MDS and a del(5q), is now the treatment of choice (List et al, 2006).

The del(5q) in the 5q− syndrome is widely believed to mark the location for a tumour suppressor gene(s), the loss of which may affect important processes such as growth control and normal haematopoiesis. We identified the commonly deleted region (CDR) or critical region of the 5q− syndrome at 5q31–q32 as the 1·5-Mb interval flanked by the marker D5S413 and the GLRA1 gene (Boultwood et al, 2002). We have been involved in the generation of a transcription map of the CDR (Boultwood et al, 1997, 2000a,b) and have used the Ensembl gene prediction program for the complete genomic annotation of this region (Boultwood et al, 2002). The CDR is gene rich and contains 44 genes, including the putative tumour suppressor genes FAT2 (also known as MEGF1) and SPARC (Boultwood et al, 2002). Several promising candidate genes have been identified although whether these act according to Knudson’s ‘two hit’ mechanism or through the ‘one hit’ mechanism of haploinsufficiency (a dosage effect resulting from the loss of a single allele of a gene; Largaespada, 2001) remains unknown. Mutation analysis of all 44 genes mapping to the CDR in a group of patients with the 5q− syndrome is now essentially complete and no mutations have been identified (unpublished data). These data offer support for the proposal that haploinsufficiency for one or more of the genes mapping to the CDR may be the pathogenetic basis of the 5q− syndrome. A detailed assessment of the expression levels of all the genes mapping to the CDR is now clearly warranted.

The relationship between the 5q− syndrome and the other myeloid malignancies with the del(5q) is a complex question. When the del(5q) occurs as the sole karyotypic abnormality in the 5q− syndrome it carries a good prognosis, otherwise it is among the worst prognostic indicators. Moreover, the del(5q) in AML, particularly secondary AML, invariably occurs together with other karyotypic abnormalities and frequently as part of a complex karyotype. The 1–1·5 Mb CDR at 5q31 identified in AML and the more aggressive forms of MDS by Lai et al (2001) is distinct from the CDR of the 5q− syndrome. This might be expected given the very different clinical features and prognosis observed in patients with the 5q− syndrome and patients with AML and a del(5q).

The identification of genes and gene pathways deregulated in cancer can lead to a better understanding of the molecular pathogenesis of the disease. We have performed gene expression profiling in a group of patients with the 5q− syndrome, patients with refractory anaemia (RA) with a normal karyotype and healthy controls in order to gain insight into those genes/gene pathways that are deregulated in the 5q− syndrome.

Materials and methods

Sample collection and cell separation

Ten patients with MDS 5q− syndrome, 14 MDS patients with RA and a normal karyotype and 16 healthy controls were included in the study (Table SI). Classification of MDS patients was according to the French–American–British (FAB) criteria (Bennett et al, 1982) and the patients were selected solely on the basis of having MDS. Patients with the 5q− syndrome had a 5q− deletion as the sole chromosomal abnormality and characteristic clinical morphological features (Giagounidis et al, 2004). The MDS patient samples were obtained from several sources: Oxford and Bournemouth (UK), Duisburg (Germany), Stockholm (Sweden) and Pavia (Italy). The study was approved by the ethics committees for each institution (Oxford C00·196, Bournemouth 9991/03/E, Duisburg 2283/03, Stockholm 410/03, Pavia 26264/2002) and informed consent was obtained. Bone marrow samples were obtained and CD34+ cells isolated from MDS patients and healthy controls. Mononuclear cells were purified by Histopaque (Sigma-Aldrich, Gillingham, UK) density gradient centrifugation, labelled with CD34 MicroBeads and CD34+ cells were isolated using MACS magnetic cell separation columns (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer’s recommendations. CD34+ cell purity was evaluated with FACS and was ≥90%.

Affymetrix experiments

Total RNA was extracted using TRIZOL (Invitrogen, Paisley, UK) following the protocol supplied by the manufacturer. An aliquot of the RNA samples was used for evaluation of quality using Agilent Bioanalyzer 2100 (Agilent Technologies, Palo Alto, CA, USA). For each sample, 50 ng of total RNA were amplified and labelled with the Two-Cycle cDNA Synthesis and the Two-Cycle Target Labelling and Control Reagent packages (Affymetrix, Santa Clara, CA, USA) following the manufacturer’s recommendations. 10 μg of biotin-labelled fragmented cRNA was hybridized to GeneChip Human Genome U133 Plus 2·0 arrays (Affymetrix), covering over 47 000 transcripts representing 39 000 human genes. Hybridization occurred at 45°C for 16 h in Hybridization Oven 640 (Affymetrix). Chips were then washed and stained in a Fluidics Station 450 (Affymetrix) and scanned with a GeneChip Scanner 3000 (Affymetrix).

Microarray data analysis

Cell intensity calculation and scaling was performed using GeneChip Operating Software (GCOS). Data analysis was performed using GeneSpring 7·3 (Agilent Technologies) and the R environment for statistical computing (R Development Core Team, 2005) using Affy and Limma packages (Smyth, 2004). Quality control was performed within the GCOS software after scaling the signal intensities of all arrays to a target of 100. Scale factors, background levels, percentage of present calls, 3′/5′GAPDH ratio and intensities of spike hybridization controls were within the acceptable range for all samples. Affymetrix CEL files were preprocessed using Robust MultiChip Analysis (RMA) (Irizarry et al, 2003). Differentially expressed genes (< 0·05) between conditions were identified using Limma and used for pathway analysis using Ingenuity 5·0. Hierarchical clustering was performed with GeneSpring software using Pearson correlation.

Real-time quantitative polymerase chain reaction

Real-time quantitative polymerase chain reaction (PCR) (Heid et al, 1996; Holland et al, 1991) was used to validate expression data for selected genes. The expression level of the beta-2-microglobulin gene (B2M) was used to normalize for differences in input cDNA. Predeveloped TaqMan Assays were used (Assays-on-Demand; Applied Biosystems, Foster City, CA, USA) and reactions were run on a LightCycler 480 Real-Time PCR System (Roche Diagnostics, Lewes, UK). Each sample was performed in triplicate and a reverse-transcriptase negative control was also tested to exclude any contaminating DNA amplification. The expression ratio was calculated using the △△CT method (Livak & Schmittgen, 2001).

MicroRNA quantitation

Quantitative reverse transcription polymerase chain reaction (qRT-PCR) was carried out using TaqMan microRNA probes as described by the manufacturer (Applied Biosystems, Warrington, UK) using 10 ng of RNA per reaction in a Roche LightCycler 480 machine. Triplicate samples were used throughout. The mean Ct value of each triplicate was used for analysis, by the △Ct method (△Ct = mean Ct of U6–mean Ct of microRNA of interest).

Results

The gene expression profiles of the CD34+ cells obtained from ten MDS 5q− syndrome patients were compared to those of the CD34+ cells from 14 MDS patients with RA and a normal karyotype and to those of the CD34+ cells from 16 healthy individuals. We identified several genes that were significantly differentially expressed between the 5q− syndrome patient group and the other two groups.

5q− syndrome versus healthy controls

In the comparison of the gene expression profiles of the CD34+ cells from 5q− syndrome patients with those of healthy controls, 536 significantly differentially expressed genes (adjusted P-value <0·05) were identified (Table I). 200 of these genes map to chromosome 5q (Table I). Hierarchical clustering performed using these 536 probe sets grouped the 5q− syndrome patients together (Fig 1).

Table I.   The most significant differentially expressed genes between MDS 5q− syndrome patients and healthy controls.
Probe set IDGene symbolMapFold changeAdj. P-value
  1. Genes are ranked by decreasing P-value after adjustment for multiple testing. The top 20 genes in each group are given. Genes in bold are common with Table II.

Genes expressed at lower levels in 5q− syndrome, mapping to 5q
 201345_s_atUBE2D25q31·2−2·017·81 × 10−8
 224068_x_atRBM225q33·1−2·201·60 × 10−7
 1555961_a_atHINT15q31·2−1·532·26 × 10−7
 202042_atHARS5q31·3−1·928·58 × 10−7
 201023_atTAF75q31−2·169·61 × 10−7
 208646_atRPS145q31–q33−1·462·49 × 10−6
 202163_s_atCNOT85q31–q33−1·822·93 × 10−6
 225698_atC5orf265q21–q22−2·584·04 × 10−6
 225326_atRBM275q32−1·854·18 × 10−6
 1568611_atP4HA25q31−3·614·41 × 10−6
 214919_s_atEIF4EBP3 /// MASK-BP35q31·3−2·014·83 × 10−6
 206562_s_atCSNK1A15q32−1·615·70 × 10−6
 209190_s_atDIAPH15q31−1·749·08 × 10−6
 205335_s_atSRP195q21–q22−1·781·05 × 10−5
 213026_atATG125q21–q22−1·771·05 × 10−5
 208773_s_atANKHD15q31·3−2·171·69 × 10−5
 203538_atCAMLG5q23−1·891·75 × 10−5
 209404_s_atTMED75q22·3−1·562·01 × 10−5
 220495_s_atC5orf145q31·1−1·803·12 × 10−5
 221193_s_atZCCHC105q31·1−2·034·19 × 10−5
Genes expressed at lower levels in 5q− syndrome, not mapping to 5q
 200066_atIK2p15-p14−1·725·29 × 10−7
 1556599_s_atARPP-213p22·3−7·571·69 × 10−5
 208285_atOR7A519p13·1−3·185·16 × 10−5
 225007_at−1·805·29 × 10−5
 227530_atAKAP126q24–q25−7·526·93 × 10−5
 227846_atGPR17615q14–q15·1−2·169·12 × 10−5
 231067_s_at−6·023·11 × 10−4
 215117_atRAG211p13−5·393·36 × 10−4
 210450_atLOC9092514q32·33−4·413·93 × 10−4
 224060_s_atDPH51p21·2−2·144·07 × 10−4
 238798_atTAPT14p15·32−1·914·57 × 10−4
 225541_atRPL22L13q26·2−1·744·67 × 10−4
 1563849_atSH2D4B10q23·1−2·876·38 × 10−4
 230672_at−1·758·24 × 10−4
 235502_at−1·828·78 × 10−4
 202908_atWFS14p16−1·958·96 × 10−4
 238071_atLCN109q34·3−1·939·30 × 10−4
 1552652_atHPS422cen-q12·3−1·929·57 × 10−4
 1557030_atGAB14q31·21−1·721·01 × 10−3
 212382_atTCF418q21·1−2·481·12 × 10−3
Genes expressed at higher levels in 5q− syndrome, not mapping to 5q
 219628_atZMAT3 (also known as WIG1)3q26·3–q272·936·62 × 10−7
 218007_s_atRPS27L15q22·21·893·19 × 10−6
 227221_at2·834·69 × 10−6
 228315_at4·115·23 × 10−6
 225725_at3·291·69 × 10−5
 235571_at1·572·17 × 10−4
 209295_atTNFRSF10B8p22-p212·712·28 × 10−4
 210033_s_atSPAG610p12·24·895·41 × 10−4
 235534_at1·521·08 × 10−3
 207813_s_atFDXR17q24–q251·991·35 × 10−3
 227372_s_atBAIAP2L17q21·32·171·43 × 10−3
 218634_atPHLDA31q311·381·82 × 10−3
 233302_atBCL11B14q32·21·821·88 × 10−3
 221081_s_atDENND2D1p13·32·221·91 × 10−3
 216252_x_atFAS10q24·11·463·42 × 10−3
 211833_s_atBAX19q13·3–q13·41·975·01 × 10−3
 217428_s_atCOL10A16q21–q221·695·25 × 10−3
 210648_x_atSNX36q211·337·68 × 10−3
 229418_atC17orf6317q11·21·428·75 × 10−3
 228502_at1·159·52 × 10−3
Figure 1.

 Hierarchical clustering of 536 differentially expressed genes in 5q− syndrome patients and healthy controls. Each row represents a single Affy probe set and each column a separate CD34+ sample.

5q− syndrome versus RA with normal karyotype

In the comparison of the gene expression profiles of the CD34+ cells from 5q− syndrome patients with those of MDS patients with RA and a normal karyotype, 170 significantly differentially expressed genes (adjusted P-value <0·05) were identified (Table II). 106 of these genes map to chromosome 5q (Table II). Hierarchical clustering performed using these 170 probe sets could separate the 5q− syndrome patients and the MDS patients with RA and a normal karyotype (Fig 2).

Table II.   The most significant differentially genes expressed genes between MDS 5q− syndrome patients and MDS patients with RA and a normal karyotype.
Probe set IDGene symbolMapFold changeAdj. P-value
  1. Genes are ranked by decreasing P-value after adjustment for multiple testing. The top 20 genes in each group are given. Genes in bold are common with Table I.

Genes expressed at lower levels in 5q− syndrome, mapping to 5q
 1555961_a_atHINT15q31·2−1·563·32 × 10−7
 206562_s_atCSNK1A15q32−1·746·12 × 10−7
 202042_atHARS5q31·3−2·006·12 × 10−7
 225698_atC5orf265q21–q22−2·878·39 × 10−7
 214919_s_atEIF4EBP3 /// MASK-BP35q31·3−2·142·11 × 10−6
 208646_atRPS145q31–q33−1·474·90 × 10−6
 203538_atCAMLG5q23−2·007·29 × 10−6
 221734_atPRRC15q23·2−1·732·46 × 10−5
 222527_s_atRBM225q33·1−1·802·75 × 10−5
 220495_s_atC5orf145q31·1−1·872·75 × 10−5
 207196_s_atTNIP15q32–q33·1−1·923·46 × 10−5
 201345_s_atUBE2D25q31·2−1·734·77 × 10−5
 208645_s_atRPS145q31–q33−1·435·00 × 10−5
 205335_s_atSRP195q21–q22−1·749·00 × 10−5
 218339_atMRPL225q33·1–q33·3−2·049·32 × 10−5
 209224_s_atNDUFA25q31−2·041·20 × 10−4
 225326_atRBM275q32−1·722·23 × 10−4
 201023_atTAF75q31−1·883·34 × 10−4
 213026_atATG125q21–q22−1,663·99 × 10−4
 218436_atSIL15q31−1·953·99 × 10−4
Genes expressed at lower levels in 5q− syndrome, not mapping to 5q
 200066_atIK2p15-p14−1·684·31 × 10−6
 230672_at−1·844·08 × 10−4
 224060_s_atDPH51p21·2−2·169·05 × 10−4
 225541_atRPL22L13q26·2−1·759·83 × 10−4
 219548_atZNF168q24−1·471·33 × 10−3
 214394_x_atEEF1D /// LOC1260378q24·3 /// 19p13·12−1·262·90 × 10−3
 212085_atSLC25A6Xp22·32−1·594·85 × 10−3
 230404_at−1·754·85 × 10−3
 200715_x_atRPL13A19q13·3−1·395·06 × 10−3
 201871_s_atLOC5103511q12·3−1·626·76 × 10−3
 212307_s_atOGTXq13−1·976·76 × 10−3
 238026_atRPL35A3q29–qter−1·748·25 × 10−3
 211073_x_atRPL322q13−1·139·70 × 10−3
 223774_atC1orf791p35·3−1·931·03 × 10−2
 234339_s_atGLTSCR219q13·3−1·651·33 × 10−2
 224719_s_atC12orf5712p13·31−1·921·43 × 10−2
 225007_at−1·561·85 × 10−2
 229590_atRPL1316q24·3|17p11·2−1·771·97 × 10−2
 214317_x_atRPS919q13·4−1·212·20 × 10−2
 233781_s_atRIF12q23·3−1·342·76 × 10−2
Genes expressed at higher levels in 5q− syndrome, not mapping to 5q
 219628_atZMAT3 (also known as WIG1)3q26·3–q272·431·76 × 10−4
 210033_s_atSPAG610p12·25·069·83 × 10−4
 225725_at2·691·88 × 10−3
 209845_atMKRN17q341·873·48 × 10−3
 228315_at2·914·04 × 10−3
 201938_atCDK2AP112q24·311·494·57 × 10−3
 227221_at2·145·20 × 10−3
 201312_s_atSH3BGRLXq13·31·286·40 × 10−3
 202265_atBMI110p11·231·821·03 × 10−2
 235571_at1·451·22 × 10−2
 233302_atBCL11B14q32·21·731·64 × 10−2
 202368_s_atTRAM26p21·1-p121·451·94 × 10−2
 224044_atRHOT117q11·21·302·12 × 10−2
 210706_s_atRNF2420p13-p12·11·552·40 × 10−2
 243007_at1·392·59 × 10−2
 217727_x_atVPS3516q121·292·68 × 10−2
 219956_atGALNT612q131·573·10 × 10−2
 227372_s_atBAIAP2L17q21·31·923·44 × 10−2
 1555691_a_atKLRC4 /// KLRK112p13·2-p12·31·313·76 × 10−2
 206110_atHIST1H3H6p22-p21·34·444·29 × 10−2
Figure 2.

 Hierarchical clustering of 170 differentially expressed genes in 5q− syndrome patients and MDS patients with RA and a normal karyotype. Each row represents a single Affy probe set and each column a separate patient CD34+ sample.

Analysis of genes mapping to the CDR

All but three of the 44 genes mapping to the CDR of the 5q− syndrome were present on the Affymetrix U133 Plus2·0 arrays used in this study. The majority of the genes assigned to the CDR showed a reduction in expression levels in the 5q− syndrome patient group (median expression ratio 0·70; Fig 3). No consistent reduction was seen in the RA with a normal karyotype patient group (median ratio 0·99) and in the healthy control group (median ratio 0·99). The observed result is consistent with a gene dosage effect due to the loss of one allele in patients with the 5q− syndrome. A small number of genes mapping to the CDR showed expression levels in the 5q− syndrome patients that were within the range of the healthy controls, including RPL7, SYNPO and CDX1.

Figure 3.

 Box–plot showing the expression ratios of the genes mapping to the CDR in patients with the 5q− syndrome, in patients with RA and a normal karyotype and in healthy controls. The boxes represent the 25th, 50th and 75th percentiles; the bars correspond to genes within a 0·5 interquartile distance.

Pathway analysis and gene networks

To identify significantly deregulated pathways, the lists of significantly differentially expressed genes between 5q− syndrome and healthy controls, and between 5q− syndrome and RA with a normal karyotype were mapped to canonical pathways using the Ingenuity software. This analysis found several pathways deregulated in 5q− syndrome, including the Wnt/β-catenin signalling, protein ubiquitination and actin cytoskeleton signalling pathways (Table III).

Table III.   Significantly deregulated pathways in 5q− syndrome vs. healthy controls and in 5q− syndrome vs. RA with a normal karyotype.
PathwayGenesP-value
  1. Genes in bold are expressed at higher level in the 5q− syndrome patient group.

5q− syndrome vs. healthy controls
 Wnt/β-catenin signallingAPC, PPP2CA, TCF3, PPP2R2C, CSNK1G3, CSNK1E, TLE1, CSNK1A11·36 × 10−3
 Cardiac β-adrenergic signallingGNG11, PPP2CA, AKAP12, AKAP2, GNG7, PPP2R2C8·19 × 10−3
 Protein ubiquitination pathwayPSME1, CDC23, UBE2B, UBE2D2, PSMB10, SKP1A, UBE2F1·23 × 10−2
 Huntington disease signallingHDAC3, HSPA4, TCERG1, GNG11, BAX, PRKCE, GNG7, HSPA91·24 × 10−2
 Actin cytoskeleton signallingARPC2, PIP5K1B, APC, ITGB1, ARHGEF7, DIAPH1, VIL2, ARPC1B1·5 × 10−2
5q− syndrome vs. RA with normal karyotype
 Estrogen receptor signallingHDAC3, SRA1, NR3C1, CCNH, TAF71·22 × 10−4
 Protein ubiquitination pathwayPSME1, CDC23, UBE2D2, PSMB10, SKP1A1·39 × 10−3
 Wnt/β-catenin signallingAPC, PPP2CA, CSNK1G3, CSNK1A15·54 × 10−3
 Aminoacyl-tRNA biosynthesisQARS, HARS, LARS6·72 × 10−3
 Cell cycle: G1/S checkpoint regulationHDAC3, SKP1A2·72 × 10−2

Differentially expressed genes between 5q− syndrome and healthy controls were also used to generate gene networks using the Ingenuity software (Fig S1). A graph was produced that identified direct and indirect interactions between genes.

Real-time quantitative PCR

The expression levels of some deregulated genes identified by microarray analysis were validated using real-time quantitative PCR in the samples of patients and healthy controls from which enough material was left. Two downregulated genes (RBM22 and CSNK1A1) and two upregulated genes (SPAG6 and ZMAT3, also known as WIG1) in the patients with 5q− syndrome were selected for validation (Fig 4). Real-time quantitative PCR experiments confirmed the downregulation of RBM22 and CSNK1A1 and the upregulation of SPAG6 and WIG1 in the majority of the 5q− syndrome patients.

Figure 4.

 Expression ratios obtained from real-time quantitative PCR experiments for selected genes in patients with 5q− syndrome (n = 7, black bars), patients with RA and a normal karyotype (n = 11, grey bars) and healthy controls (n = 11, white bars). For the downregulated genes (RBM22 and CSNK1A1), the dashed line indicates the twofold downregulation level (expression ratio = 0·5). For the upregulated genes (SPAG6 and ZMAT3), the dashed line indicates the twofold upregulation level (expression ratio = 2).

Real-time quantitative PCR was also used to evaluate the expression of the microRNA genes MIRN143, MIRN145, MIRN378 (also known as miR-143, miR-145, miR-378 respectively) that map within the CDR of the 5q− syndrome. No significant difference was observed in patients with the 5q− syndrome and patients with RA and a normal karyotype and healthy controls (Fig S2).

Discussion

There is considerable evidence to suggest that the 5q− syndrome is a clonal disease of the haematopoietic stem cell (Nilsson et al, 2000). We have used the Affymetrix GeneChip U133 Plus 2·0 platform, on which most human genes are represented, to determine the transcriptome of the CD34+ cells of a group of patients with the 5q− syndrome. We have compared these results with the corresponding transcriptome of CD34+ cells from both healthy control subjects and MDS patients with RA and a normal karyotype. In particular, we sought to identify genes/gene pathways deregulated in the 5q− syndrome and also to determine whether the 5q− syndrome has a distinct gene expression profile.

In the comparison of 5q− syndrome patients with healthy controls, 536 significantly differentially expressed genes were identified. Approximately 40% of the significant probe sets showing lower expression levels in patients with the 5q− syndrome mapped to chromosome 5q, suggesting that the loss of one allele in these patients has a gene dosage effect. Hierarchical clustering grouped patients with the 5q− syndrome into a distinct cluster, separate from healthy controls. The demonstration that the 5q− syndrome has a distinct gene expression profile suggests a common underlying pathophysiological basis to the disease.

One or more of the 44 genes assigned to the CDR of the 5q− syndrome may represent the tumour suppressor gene(s) associated with the development of this disorder. This study sought to determine the expression levels of these potential candidate genes in the CD34+ cells of patients with the 5q− syndrome. The majority of the genes assigned to the CDR showed a reduction in expression levels within the range 0·6–0·8 in patients with the 5q− syndrome, consistent with the loss of one allele. Candidate genes identified as showing haploinsufficiency in the 5q− syndrome in this study included the tumour suppressor gene SPARC (Yiu et al, 2001) and RPS14, a component of the 40S ribosomal subunit. Interestingly, haploinsufficiency for RPS19, also encoding a ribosomal protein, causes Diamond–Blackfan anaemia (Gazda et al, 2004). We identified two genes assigned to the CDR, RBM22 and CSNK1A1, which showed a marked reduction in expression levels in some patients with the 5q− syndrome. Indeed, the RBM22 gene represents the most significantly downregulated gene mapping to the CDR of the 5q− syndrome and is the second most downregulated gene mapping to the whole long arm of chromosome 5. The gene expression results for RBM22 and CSNK1A1 were validated using quantitative real-time PCR and the majority of patients with the 5q− syndrome studied showed a >50% reduction in the gene expression level of RBM22 and CSNK1A1, consistent with the downregulation of the remaining allele.

RBM22 is a highly conserved RNA-binding protein. Recently, several laboratories have used genome-wide screening to identify essential proteins involved in the regulation of alternative splicing (Park et al, 2004), in cell cycle (Kittler et al, 2004) and in zebrafish development (Amsterdam et al, 2004). All these studies have identified homologues of RBM22 as being essential genes. RBM22 has recently been shown to play a critical role in the nuclear translocation of the Apoptosis-linked gene 2 (ALG-2; Montaville et al, 2006), a member of the family of Ca(2+)-binding proteins essential for the execution of apoptosis by various signals including Fas activation (Jung et al, 2001). ALG-2 translocates from the cytoplasmic membrane to the cytosol during FAS-induced apoptosis and interacts with FAS. However, in the presence of RBM22 the cytosolic protein ALG-2 is translocated to the nucleus (Montaville et al, 2006). The effects of the marked downregulation of RBM22 in the CD34+ cells of patients with the 5q− syndrome are presently unknown, but may include impaired splicing and/or impaired translocation of ALG-2, leading to deregulated apoptosis. It is well recognized that impaired splicing/apoptosis can play a role in the development of cancer (Venables, 2004) and we suggest that RBM22 represents a candidate gene for the 5q− syndrome.

The CK1 proteins represent a family of serine/threonine kinases that have been shown to play several different roles in the development of cancer (Knippschild et al, 2005). CSNK1A1 has been recently shown to play a role in the regulation of Hedgehog (Hh) signalling which governs cell growth and patterning in animal development (Ingham & McMahon, 2001). CSNK1A1 promotes Hh signalling via phosphorylation of the Smo protein (Jia et al, 2004). Hh signalling has been shown to be deregulated in a variety of human cancers (Chen et al, 2007; Liu et al, 2006). The effects of the marked downregulation of CSNK1A1 in the CD34+ cells of patients with the 5q− syndrome are unknown but may include impaired Hh signalling and deregulation of the Wnt pathway (Dejmek et al, 2006; Hammerlein et al, 2005).

We have previously used a cDNA array platform comprising 6,000 human genes to investigate differences in gene expression profiles in the neutrophils of patients with the 5q− syndrome and patients with RA and a normal karyotype (Pellagatti et al, 2004). The array platform used in this earlier study contained only a small proportion of those genes mapping to the long arm of chromosome 5. In agreement with the study on neutrophils (Pellagatti et al, 2004), we found lower expression levels of TMED7 (also known as CGI-109), MED7 (also known as CRSP9), GLRX (also known as GRX) and ATOX1 (all mapping to chromosome 5q and present on both arrays) in the CD34+ cells of patients with the 5q− syndrome compared with patients with RA and a normal karyotype.

There is considerable interest in the role of miRNAs in cancer (Calin et al, 2005) and three miRNA genes map within the CDR of the 5q− syndrome. We have investigated the expression levels of these miRNA genes but no significant difference was observed between patients with the 5q− syndrome and healthy controls or patients with RA and a normal karyotype. However, the expression levels of MIRN145 were slightly increased in the 5q− syndrome patients, probably reflecting increased expression from the retained allele.

Many genes were identified that were significantly differentially expressed in the CD34+ cells from patients with the 5q− syndrome compared with both normal individuals and to patients with RA and a normal karyotype. As expected, many of the genes downregulated in the 5q− syndrome mapped to the long arm of chromosome 5, consistent with a gene dosage effect. Differentially expressed genes mapping to other chromosomes include SPAG6, ZMAT3, BMI1 (all upregulated in 5q− syndrome) and DPH5 (downregulated in 5q− syndrome). SPAG6 has recently been shown to be markedly overexpressed in paediatric AML (Steinbach et al, 2006). ZMAT3, a p53-induced gene, encodes a growth inhibitory protein (Hellborg et al, 2001). ZMAT3 plays a role in gene regulation mediated by siRNAs and miRNAs (Mendez Vidal et al, 2006). Over-expression of ZMAT3 has been demonstrated in lung cancer (Varmeh-Ziaie et al, 2001). BMI1 is a member of the Polycomb group of transcriptional repressor genes (Orlando, 2003). Interestingly, BMI1 is required for maintenance of adult self-renewing haematopoietic stem cells (Park et al, 2003).

To identify significantly deregulated pathways, the lists of significantly differentially expressed genes between 5q− syndrome and healthy controls, and between 5q− syndrome and RA with normal karyotype were mapped to canonical pathway using the Ingenuity software. Two pathways are highlighted in both comparisons: the Wnt/β-catenin signalling pathway and the protein ubiquitination pathway. The Wnt/β-catenin signalling pathway is a critical regulator of stem cells (Reya & Clevers, 2005), and may play a role in many cancers including AML (Simon et al, 2005). It is now known that the deregulation of ubiquitin pathways results in the development of several human diseases, including many types of tumours (Hoeller et al, 2006). The importance of further investigation of this pathway in MDS is that there is now an active research field of drug discovery targeting the ubiquitin-proteasome system in cancer (Nalepa et al, 2006).

Next we sought to determine whether any of the proposed candidate genes for the 5q− syndrome play a role in the pathways found to be deregulated in the 5q− syndrome. We have shown marked downregulation of CSNK1A1 in patients with the 5q− syndrome. The Wnt pathway represents one of the most significantly deregulated pathways in the 5q− syndrome and interestingly the CK1 proteins (including CSNK1A1) have been shown to play a role in the regulation of the Wnt pathway (Dejmek et al, 2006; Hammerlein et al, 2005). We have recently identified the tumour suppressor gene SPARC, mapping to the CDR (Boultwood et al, 2002), as a major target of lenalidomide in the response of MDS 5q− cells to the drug and have suggested that this gene may play a role in the pathogenesis of the 5q− syndrome (Pellagatti et al, 2007). In the present study, we found a reduction in expression level of SPARC within the range 0·3–0·8 consistent with the loss of one allele in the majority of patients with the 5q− syndrome. One case with the 5q− syndrome, however, did not show downregulation of the SPARC gene. SPARC acts via the integrin-linked actin cytoskeleton signalling pathway (Barker et al, 2005; Said et al, 2007). The integrins are linked to the actin cytoskeleton and mediate cellular signalling. They provide a bridge between the cytoskeleton and the extracellular matrix (ECM) and regulate cell adhesion (Guo & Giancotti, 2004). Intriguingly, actin cytoskeleton signalling (including integrin signalling) was amongst the most significantly deregulated pathways in the 5q− syndrome. We previously reported deregulation of genes related to the actin cytoskeleton in MDS patients with the del(5q) (Pellagatti et al, 2006). These data give further support to the proposal that SPARC may play a role in the pathogenesis of the 5q− syndrome.

This study has identified several genes and gene pathways that are deregulated in the CD34+ cells of patients with the 5q− syndrome using gene expression profiling. We have identified two genes assigned to the CDR of the 5q− syndrome, RBM22 and CSNK1A1, which show marked downregulation in this disorder. We suggest that RBM22 and CSNK1A1 represent candidate genes for the 5q− syndrome. Pathway analysis data supports the proposal that SPARC may also play a role in the 5q− syndrome. We conclude that several of the genes mapping to the CDR of the 5q− syndrome may play a role in the pathogenesis of this disorder. This is the first paper to address the question of deregulated pathways in the 5q− syndrome. We believe that such analysis will form the basis of a much better understanding of the pathophysiology of the 5q− syndrome.

Ancillary