Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

A genomic analysis of adult T-cell leukemia


Adult T-cell leukemia (ATL) is an intractable malignancy of CD4+ T cells that is etiologically associated with infection by human T-cell leukemia virus-type I. Most individuals in the chronic stage of ATL eventually undergo progression to a highly aggressive acute stage. To clarify the mechanism responsible for this stage progression, we isolated CD4+ cells from individuals in the chronic (n=19) or acute (n=22) stages of ATL and subjected them to profiling of gene expression with DNA microarrays containing >44 000 probe sets. Changes in chromosome copy number were also examined for 24 cell specimens with the use of microarrays harboring 50 000 probe sets. Stage-dependent changes in gene expression profile and chromosome copy number were apparent. Furthermore, expression of the gene for MET, a receptor tyrosine kinase for hepatocyte growth factor (HGF), was shown to be specific to the acute stage of ATL, and the plasma concentration of HGF was increased in individuals in either the acute or chronic stage. HGF induced proliferation of a MET-positive ATL cell line, and this effect was blocked by antibodies to HGF. The HGF-MET signaling pathway is thus a potential therapeutic target for ATL.


Adult T-cell leukemia (ATL) is an intractable malignancy of CD4+ T cells that is etiologically associated with infection by human T-cell leukemia virus-type I (HTLV-I) (Uchiyama et al., 1977; Poiesz et al., 1980; Yoshida et al., 1982). Virally encoded proteins such as Tax trigger polyclonal growth of T cells in infected individuals, and there are an estimated 15–20 million such carriers worldwide (Edlich et al., 2000). After a latency period of decades, a small proportion of carriers (2%) develop ATL. Many ATL patients initially manifest only monoclonal (or oligoclonal) growth of leukemic clones without apparent clinical symptoms, a condition referred to as the chronic or smoldering stages (Shimoyama, 1991). Most individuals in the chronic stage, however, eventually undergo progression to a highly aggressive acute stage (Tajima, 1990). Given that the prognosis of individuals at the acute stage remains very poor, it is important to clarify the molecular mechanism that underlies stage progression.

Homozygous deletion or epigenetic silencing of the gene for the cyclin-dependent kinase inhibitor p16 (Hatta et al., 1995; Yamada et al., 1997; Nosaka et al., 2000) as well as altered expression of other genes related to cell proliferation (Cesarman et al., 1992; Tamiya et al., 1998) have been detected in ATL cells at the acute stage. However, such genetic or epigenetic changes may be infrequent (Matsuoka, 2003), and the transforming events responsible for chronic to acute stage progression remain largely unknown.

DNA microarray analysis allows simultaneous comparison of the expression intensities of tens of thousands of genes. Such analysis of the transcriptomes of ATL cells at the chronic and acute stages might thus be expected to provide insight into the mechanism of stage progression in this disease. With the use of this approach, Sasaki et al. (2005) recently compared transcriptomes between normal CD4+ T cells (n=5) and mononuclear cells (MNCs) isolated from individuals in the acute stage of ATL (n=8). Tsukasaki et al. (2004) also compared transcriptomes between MNCs from patients in the chronic or acute stages of ATL (n=4 for each). However, the significance of these data may be limited by the small number of study subjects and by the use of unfractionated MNCs that contain various proportions of non-ATL cells.

In addition to changes in gene expression, ATL cells frequently manifest various karyotype anomalies. Comparative genomic hybridization (CGH) has thus revealed recurrent gains in chromosomes 2p, 3p, 7q and 14q as well as losses in 6q in ATL cells (Ariyama et al., 1999; Tsukasaki et al., 2001). However, CGH or its successor, bacterial artificial chromosome (BAC) array-based CGH, is able to analyse chromosome copy number alterations (CNAs) at a resolution of only several hundred kilobase pairs (Lockwood et al., 2005). High-density oligonucleotide microarrays originally designed for genotyping of single nucleotide polymorphisms (SNPs) have recently been adapted for CNA analysis (Lin et al., 2004; Nannya et al., 2005). With this approach, chromosome copy number is inferred from the signal intensity of SNP probe sets distributed throughout the human genome. For instance, with Affymetrix GeneChip Mapping 100K arrays developed for genotyping of 100 000 SNPs, it is possible to determine CNAs at a mean resolution of 23.6 kbp, which is substantially greater than that achievable with BAC array-based technologies.

With both microarray-based gene expression profiling and SNP array-based CNA profiling, we have now performed a comprehensive genomic analysis of ATL in order to investigate the mechanism of stage progression from chronic to acute. Given that the CD4+CD8 fraction of peripheral blood (PB) cells of individuals with chronic or acute ATL is composed predominantly of ATL cells, we purified this fraction from ATL patients. We then subjected the isolated cells to gene expression profiling with microarrays containing >44 000 probe sets and to CNA analysis with microarrays harboring 50 000 probe sets. The gene expression data indicate that the transcriptomes for the chronic and acute stages of ATL are distinct, and the CNA data reveal frequent amplification or deletion of genomic fragments of various sizes in each ATL stage.


Transcriptomes of ATL cells

To characterize the transcriptomes of ATL cells, we purified CD4+ cells from PB of ATL patients at either the chronic (n=19) or the acute (n=22) stage. The clinical characteristics of the patients are summarized in Supplementary Table 1. The CD4+ fraction was also purified from healthy volunteers (n=3) and was either activated with phytohemagglutinin (PHA) or not.

A simple, one-step column purification with antibodies to CD4 yielded a highly pure CD3+CD4+ T-cell fraction. For example, whereas the CD3+CD4+ fraction constituted only 29.1% of PB MNCs of one healthy individual, it constituted 98.8% of the corresponding column eluate (Figure 1a). Similarly, CD3+CD4+ cells constituted 25.7% of MNCs from one ATL patient at the acute stage, but accounted for 97.5% of cells in the corresponding column eluate (data not shown).

Figure 1

Purification and gene expression profiling of ATL cells. (a) MNCs isolated from the PB of a healthy individual were subjected to staining with Wright–Giemsa solution before (Pre) and after (Post) purification by affinity chromatography with antibodies to CD4 (upper panels). Scale bar, 10 μm. The same fractions were also subjected to flow cytometry with antibodies to CD3 and to CD4 (lower panels). The proportion of double-positive cells is indicated. (b) Subject tree generated by hierarchical clustering analysis of the expression profiles for 15 121 probe sets. Normal T cells (Nor 1–3) stimulated (+) or not (−) with PHA (8 μg/ml) clustered together, separate from the ATL samples from patients in the chronic (green) or acute (red) stage. (c) Subject tree generated by two-way clustering analysis with 21 probe sets that contrasted the two clinical conditions (Student's t-test with the Benjamini and Hochberg false discovery rate of 0.01, and effect size of 100 U). Each column corresponds to a separate sample, and each row to a probe set whose expression is color-coded according to the indicated scale. Gene symbols are shown on the right; 224664_at and 211933_s_at are the expressed sequence tag IDs designated by Affymetrix ( Annotations and expression intensities for the probe sets are presented in Supplementary Table 2.

All of the ATL and normal CD4+ cell specimens were then subjected to expression profiling with 44 000 probe sets (corresponding to 33 000 transcripts) on Affymetrix HGU133 microarrays. To eliminate from the analysis genes that were transcriptionally silent in the ATL specimens, we first selected probe sets that received the ‘Present’ call by Microarray Suite 5.0 software (Affymetrix) in at least 30% (n=13) of the ATL samples. A total of 15 121 probe sets fulfilled this criterion. On the basis of the similarity of the expression profiles for these probe sets, all 47 samples were subjected to hierarchical two-way clustering (Alon et al., 1999), yielding a dendrogram of the subjects (Figure 1b). All six normal samples, irrespective of PHA stimulation, formed a distinct branch separated from the ATL specimens, indicating that the overall gene expression profiles differed between normal and transformed T cells. However, samples corresponding to patients with chronic or acute ATL were not clearly separated from each other in this tree.

To compare the transcriptomes of ATL cells between chronic and acute stages, we conducted Student's t-test on the gene expression intensity for the 15 121 probe sets with the Benjamini and Hochberg false discovery rate (Reiner et al., 2003) of 0.01, leading to the isolation of 84 probe sets (data not shown). To enrich probe sets whose expression level was high in at least one of the stages, we adopted another selection window, effect size (absolute difference in mean expression intensity) (Dhanasekaran et al., 2001). We extensively compared the expression level of given probe sets determined by DNA microarray and by quantitative real-time reverse transcription–polymerase chain reaction (RT–PCR). With our normalization procedure (see Materials and methods), expression of genes with an array data of 100 units (U) was almost always detected by real-time RT–PCR (data not shown). Thus, we chose 100 U as the threshold value for the effect size.

A total of 21 probe sets (corresponding to 21 independent genes) whose expression level contrasted the two clinical conditions were finally identified. Hierarchical two-way clustering analysis of the expression profiles of these stage-associated genes revealed that only two gene were preferentially expressed at the chronic stage, whereas the other 19 genes were preferentially expressed in the acute stage (Figure 1c and Supplementary Table 2). Interestingly, the latter gene cluster contains several genes encoding for growth-related proteins, such as nuclear receptor coactivator 3 (NCOA3, GenBank accession no. NM_006534), heat-shock 60-kDa protein 1 (HSPD1, GenBank accession no. NM_002156) and general transcription factor IIIA (GTF3A, GenBank accession no. BE542815).

Gene expression-based prediction of ATL stage

We next attempted to develop a microarray-based class prediction algorithm for ATL. Among several approaches examined, an artificial neural network (ANN) provided the highest accuracy for prediction (O'Neill and Song, 2003). ANNs are computer-based algorithms modeled on the structure and behavior of neurons in human brain. Pattern recognition by ANNs is accomplished by training the networks for multiple times in a supervised mode. ANNs adjust continuously their internal weighted connections to reduce the observed errors in matching input to output.

Here, the 15 121 probe sets originally selected in Figure 1b were divided into three nonoverlapping groups, each of which was used as the input for 10 ANNs (Figure 2a). We performed a 10% crossvalidation rotation with 37 samples, training with 33 samples and testing of the remaining four samples. We then reduced the weight of one input in the first layer (one at a time by 15%), and the network was run again to evaluate the difference in the result from the original output. The same procedure was performed in turn for every input, in order to identify 44 ‘predictor’ genes whose expression markedly influenced the prediction accuracy in each set of ANNs (Figure 2b and Supplementary Table 3). Such predictor set contains only one gene (UBE2E1) shared with the stage-associated probe sets shown in Figure 1c. As demonstrated previously, ANN and other approaches (such as t-test or clustering analysis) frequently isolate distinct sets of predictor genes (O'Neill and Song, 2003).

Figure 2

Schematic of the ANN analysis used for class prediction of ATL. (a) The 15 121 probe sets originally selected in Figure 1b were divided into three nonoverlapping groups, each of which was used as the input for 10 ANNs. We performed a 10% crossvalidation rotation with 37 samples, training with 33 samples and testing of the remaining four samples. On the basis of the differentiation process with the three sets of 10 ANNs, we selected 44 ‘predictor’ genes whose expression markedly influenced the prediction accuracy in each set of ANNs. (b) Subject tree generated by two-way clustering analysis with the 44 predictor genes selected in (a) is demonstrated as in Figure 1c. 241843_at and 212371_at are the expressed sequence tag IDs designated by Affymetrix. Annotations and expression intensities for the probe sets are presented in Supplementary Table 3.

Another nine ANNs were then trained and tested with the 44 predictor genes in the same 10% crossvalidation round, yielding one error of prediction for the 37 samples. Finally, the withheld four samples were tested with the trained ANN, resulting in the correct prediction of the class of each. Given that diagnosis of the stage of ATL patients is sometimes problematic, especially when an individual is undergoing stage transition, our analysis offers the possibility of a microarray diagnostic system based on the expression profile of a small number of genes.

Copy number analysis of the ATL genome

To analyse chromosomal gain or loss in ATL cells, we subjected genomic DNA to hybridization with genotyping arrays that represent 50 000 human SNPs and allow determination of copy number at an average resolution of 47.2 kbp. We first examined whether MNCs and purified CD4+ ATL cells may yield similar CNA profiles by analyzing genomic DNA from such cell fractions of a single individual (patient ID6) at the acute stage of ATL. Flow cytometry revealed that CD3+CD4+ T cells constituted 58.9 and 98.0% of MNCs and purified CD4+ cells of this individual, respectively (data not shown).

As shown in Figure 3a, gain of chromosomal content (3n) was apparent at 1q, 3q, 5p, 7q, 18q and 21q, whereas loss of genomic content (1n) was observed at 2p, 12p, 13q, 14q and 18p. In addition to changes affecting such large chromosomal regions, numerous CNAs too small to be detected by conventional methods were apparent at various positions (hospital karyotyping of MNCs from this patient indicated a karyotype of 46,XY). We also identified many chromosomal regions whose copy number differed between the unfractionated MNCs and purified CD4+ cells (Figure 3a). These data indicate that purification of CD4+ cells increases the sensitivity of copy number measurement.

Figure 3

CNA analysis of purified ATL cells. (a) Inferred copy number for all SNP sites analysed in MNCs (left) and CD4+ cells (right) isolated from an ATL patient in the acute stage (ID6). Copy number is color-coded according to the indicated scheme. Chromosomal regions with an aberrant copy number detected only in CD4+ cells are indicated by blue underlines. An inset below demonstrates the raw signal (log2 ratio) for every SNP locus by the array hybridization (red dots) and corresponding inferred copy number (green lines) on chromosome 1 for the MNC sample, along with the chromosome cytobands. (b) Inferred copy number for all SNP sites (chromosomes 1–22) in all subjects studied (n=9 for chronic stage, n=15 for acute stage) shown according to the color scheme at the bottom. SNP sites are ordered by their physical position from top to bottom (shown on the right), and the borders between chromosomes are indicated by small bars. The most frequently deleted region at 14q11.2 is indicated by the arrowhead. (c) Inferred copy number (yellow columns) for a region of chromosome 6 (nucleotides 16 651 304–16 651 533) compared with the relative amount (blue line) of genomic DNA corresponding to this region (expressed relative to the amount of GAPDH genomic DNA). An ATL cell line (KK-1) and CD4+ cells isolated from a male (46XY) or female (46XX) volunteer were also analysed. (d) Subject tree based on inferred copy number of chromosomal regions that showed a statistically significant difference in copy number for at least two consecutive SNP sites between the acute and chronic stages of ATL (P<0.01, Student's t-test). (e) Comparison of expression intensities of the genes assigned to two chromosomal regions (region #1 and region #2) indicated in (d) between the subjects with or without a gain in chromosome copy number (Chr#) for the corresponding region. Data are means+s.d. The P-values were calculated by Student's t-test.

Among the ATL specimens subjected to gene expression profiling, all those for which CD4+ cells were available for preparation of genomic DNA were analysed for CNAs (n=24; 15 specimens for the acute stage, nine specimens for the chronic stage). Assessment of copy number revealed frequent anomalies of various sizes, ranging from amplification of an entire chromosome to small deletions spanning only a few probe sets, in the ATL genome (Figure 3b). The most frequent gain or loss in our data set was a small deletion at 14q11.2, which was identified in 22 of the 24 patients tested; the core deleted region spans five probe sets, encompassing as little as 30 857 bp at the locus of TRD (encoding T-cell receptor delta locus) and TRA (encoding T-cell receptor alpha constant). These deletions likely reflect genomic rearrangement at the T-cell receptor locus in ATL cells and support the high sensitivity of the method.

Further, a high-grade amplification of genome could be found in a region spanning 14 Mbp at 3p (nucleotide 10 672 576–24 556 563) among the ATL patients, especially at the acute stage. A chromosome copy number of four in this region was inferred for three patients at the acute stage (ID 3, 15 and 70), and that of three was inferred for seven patients. Interestingly, expression level of the genes mapped on this 3p region was significantly higher in the patients with a chromosome copy number of four compared to those with a copy number of two (P=0.03, Student's t-test), and marginally higher to those with a copy number of three (P=0.051) (data not shown).

To confirm the inferred copy numbers in our data set, we subjected genomic DNA at a locus with marked variation in copy number (chromosome 6, nucleotides 16 651 304–16 651 533) to quantitative real-time PCR analysis. Such analysis of the 24 patients, two healthy volunteers (one male, one female), and a cell line (KK-1) (Imaizumi et al., 2003) derived from a patient at the acute stage of ATL revealed that the inferred copy number was highly correlated with DNA content measured by PCR (Figure 3c).

Stage-dependent CNAs

To screen for CNA patterns linked to stage progression in ATL, we applied Student's t-test (P<0.01) to the obtained data set. Subsequent application of a selection window specifying that at least two contiguous probes show the same CNA pattern led to the isolation of 330 probe sets that corresponded to 3p, 3q, 14q and 19p (Figure 3d). Segmental amplification of chromosome 3 was detected only in the ATL patients at the acute stage, consistent with previous results obtained by CGH analysis (Tsukasaki et al., 2001; Oshiro et al., 2006).

To examine the effect of gene dosage on mRNA abundance, we analysed our gene expression data set for the expression level of genes assigned to a segment (region #1, nucleotides 114 092 369–119 769 881) of chromosome 3 (Figure 3d and e). The mean expression intensity of genes in this region was significantly greater for the patients with a corresponding gain of DNA content than for those without such a gain (P=0.00015, Student's t-test). Similarly, the expression level of genes on a segment (region #2; nucleotide 8 782 486–12 322 072) of chromosome 19 was greater in cells with a gain of DNA content in this region than in those without such a gain (P=0.0357). These data indicate that gene dosage indeed affects transcript abundance in ATL cells. The large standard deviations apparent in the data shown in Figure 3e, however, suggest that other mechanisms (mediated by transcription factors or epigenetic regulation, for example) have also a large impact on gene expression level.

The hepatocyte growth factor-MET pathway in ATL cells

The long latency period for ATL in HTLV-I carriers suggests that the molecular pathogenesis of ATL and its stage progression might be highly heterogeneous. To identify molecular events that might contribute to transition to the acute stage, we next attempted to isolate ‘acute stage-specific genes,’ defined by their silence (expression level of <10 U) in all normal T cells and chronic ATL specimens and their activation (>100 U) in at least one of the acute-stage samples. We isolated six probe sets that fulfilled such criteria (Figure 4a and Supplementary Table 4).

Figure 4

Acute stage-specific expression of MET in ATL. (a) Expression profiles of the six acute stage-specific probe sets. The expression level of each probe set is colored according to the indicated scale. Annotations and expression intensities for the probe sets are presented in Supplementary Table 4. (b) Comparison of the abundance of MET mRNA in the study specimens as determined by microarray and RT–PCR analyses. For the latter, the amount of MET mRNA is expressed relative to that of GAPDH mRNA. Pearson's correlation coefficient (r) for the comparison is indicated. (c) Cell surface expression of MET was examined by flow cytometry in four ATL samples corresponding to the acute stage. The solid and dashed traces were obtained with antibodies to MET and control antibodies, respectively. The proportion of MET+ cells determined by flow cytometry is indicated together with the corresponding amount of MET mRNA determined by microarray analysis.

Among these acute stage-specific genes, we focused on MET (GenBank accession no. NM_000245), given that we recently found, in an independent study, that the amount of MET mRNA was specifically increased in ATL cells that manifested liver invasion (Imaizumi et al., 2003). MET encodes a transmembrane protein tyrosine kinase that is the receptor for hepatocyte growth factor (HGF) (Bottaro et al., 1991). The expression level of MET in the study specimens as determined by microarray analysis was highly correlated with that determined by quantitative RT–PCR analysis (Figure 4b), as revealed by a Pearson's correlation coefficient (r) of 0.851 (P<0.001). (Also see Supplementary Table 5 for verification of microarray data by RT–PCR.) Flow cytometry revealed that the expression of MET at the cell surface reflected the abundance of the corresponding mRNA in ATL samples (Figure 4c).

The acute stage-specific expression of MET at both the mRNA and protein levels suggested that ATL cells might acquire mitogenic potential as a result of activation of a MET-linked signaling pathway. To examine the possible operation of an HGF-MET autocrine loop, we quantitated HGF mRNA in ATL cells by both microarray and quantitative RT–PCR analyses. No substantial amounts of HGF mRNA were detected in ATL specimens, however (data not shown).

We therefore next measured the plasma concentration of HGF in the study subjects. High levels of HGF were detected in the plasma of ATL patients, especially in that of individuals in the acute stage (Figure 5a), compared with the previously determined values for healthy adults (0.27±0.08 ng/ml, mean±s.d.) and some cancer patients (1–2 ng/ml) (Funakoshi and Nakamura, 2003). To test directly whether activation of the HGF-MET signaling pathway is able to promote the proliferation of ATL cells, we examined the MET-positive ATL cell line KK-1. HGF induced both the tyrosine phosphorylation of MET and proliferation in KK-1 cells (Figure 5b and c). The addition of antibodies to HGF (Montesano et al., 1991) could abolish both effects.

Figure 5

HGF-MET activity induces proliferation in ATL cells. (a) Comparison of the plasma concentration of HGF in ATL patients (as measured by enzyme-linked immunosorbent assay) with the corresponding amount of MET mRNA in leukemic cells (as determined by RT–PCR). (b) Tyrosine phosphorylation of MET in KK-1 cells incubated for 10 min in the absence (−) or presence of human HGF (50 ng/ml), alone or together with antibodies to HGF (10 μg/ml), was examined by immunoblot analysis of cell lysates with antibodies to phosphotyrosine (Anti-P-Tyr). The blot was also probed with antibodies to MET. (c) The proliferation of KK-1 cells (3 × 104) incubated for 1 day in the absence (−) or presence of HGF (50 ng/ml), alone or together with antibodies to HGF (10 μg/ml), was evaluated by the MTS assay. Data are expressed in absorbance units and are means+s.d. of triplicates from a representative experiment. The P-values for the indicated comparisons were determined by Student's t-test.


We have analysed gene expression and CNA profiles in leukemic cell-enriched fractions of individuals with ATL. We found that both types of profile differ markedly between the chronic and acute stages of ATL, and that the level of gene expression is influenced by the copy number of genomic DNA in ATL cells. Although ATL cells have previously been shown to manifest multiple genomic gains or losses (Ariyama et al., 1999; Tsukasaki et al., 2001), our data have revealed that the ATL genome is more unstable than has been appreciated. Similar complex CNAs have been identified by SNP array-based methods for other types of cancer (Zhao et al., 2005). We detected 3386.9±2254.0 (mean±s.d.) and 4678.5±2874.6 SNP sites showing 3n ploidy as well as 1039.9±2310.0 and 927.1±1137.9 SNP sites with 1n ploidy in samples corresponding to the chronic and acute stages of ATL, respectively. The numbers of probe sets showing gain or loss of DNA content did not differ significantly (P>0.05) between the chronic and acute stages of ATL. Given the large numbers of probe sets with an aberrant DNA content in the ATL genome, a large-scale study will likely be required to pinpoint the bona fide disease-dependent or stage-dependent CNAs.

Recently, with the use of BAC array-based CGH (with a mean resolution of 1.3 Mbp), Oshiro et al. (2006) have compared CNA of MNCs for 17 patients at the acute stage of ATL to that of lymph nodes for 42 cases with the lymphoma type of ATL. The recurrent gain of chromosomes was found at 3/3p among individuals with the acute stage of ATL, which is in good agreement with our results.

We found that an increased plasma concentration of HGF coexists with an increased expression of MET in ATL cells from some individuals at the acute stage of the disease. Together with the demonstrated mitogenic effect of HGF in ATL cells, these data suggest a novel scenario for stage progression in ATL. The mechanism responsible for the increased plasma level of HGF in ATL patients is unclear. Given that ATL is a malignancy of activated mature T cells, the leukemic cells secrete a variety of cytokines, including tumor necrosis factor-α and interleukin-1β (Wano et al., 1987; Yamada et al., 1996). Both of these cytokines are potent inducers of HGF expression in fibroblasts (Matsumoto et al., 1992; Tamura et al., 1993), suggesting that ATL cells may indirectly increase the plasma HGF level through secretion of these cytokines and consequent activation of fibroblasts.

Our data indicate that the plasma concentration of HGF in ATL patients increases before the onset of expression of MET in the leukemic cells (Figure 5a). The increased concentration of HGF might therefore confer a growth advantage on ATL cells after they upregulate the expression of MET. Given that the JAK-STAT signaling pathway is activated in the leukemic cells of patients with advanced ATL (Migone et al., 1995), it may be relevant that binding sites for STAT1 or STAT3 are present in the promoter regions of five (including MET) of the six acute stage-specific genes identified in the present study (Figure 4a). We did not detect a significant difference in DNA content (in our data set) for the MET locus between chronic and acute stages of ATL. It is thus possible that JAK-STAT signaling contributes to transcriptional activation of MET.

Given that our data are derived from purified ATL cells, they can be further used to characterize ATL in various ways. For instance, we attempted to isolate genes whose expression was linked to the presence of hypercalcemia in the study patients (data not shown); such genes included that for parathyroid hormone-like hormone (GenBank accession no. BC005961), which has been shown to be responsible for many instances of humoral hypercalcemia in individuals with cancer including ATL (Broadus et al., 1988; Motokura et al., 1988).

We have demonstrated the existence of an HGF-MET paracrine loop specific to the acute stage of ATL. Given that ligation of MET by HGF promoted the proliferation of ATL cells, activation of the HGF-MET signaling pathway is a candidate molecular mechanism for stage progression in ATL. Furthermore, our observation that this mitogenic effect was blocked by antibodies to HGF provides potential new targets for ATL therapy.

Materials and methods

Expression profiling

All clinical specimens were obtained with written informed consent, and the study was approved by the ethics committees of both Jichi Medical University and Nagasaki University. The diagnosis of ATL in all cases was based on clinical features, immunophenotypes of leukemic cells, and the monoclonal integration of HTLV-1 proviral DNA into the genome of leukemic cells (Shimoyama, 1991). MNCs isolated from PB were labeled with magnetic bead-conjugated mouse monoclonal antibodies to CD4 (CD4 MicroBeads, Miltenyi Biotec, Auburn, CA, USA). For PHA stimulation, purified CD4+ cells from healthy individuals were incubated for 48 h in Rosewell's Park Memorial Institute media (RPMI) 1640 medium (Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (FBS) and PHA-P (8 μg/ml) (Sigma, St Louis, MO, USA).

Column fractionation of MNCs, RNA preparation, and hybridization with HGU133A & B microarrays (Affymetrix, Santa Clara, CA, USA) were performed as described previously (Choi et al., 2004). The mean expression intensity of the internal positive control probe sets ( was set to 500 U in each hybridization, and the fluorescence intensity of each test gene was normalized accordingly. All HGU133A & B microarray data are available at the Gene Expression Omnibus web site ( under the accession number GSE1466.

Student's t-test with the Benjamini and Hochberg false discovery rate option was performed with GeneSpring 7.0 software (Silicon Genetics). Effect size was defined as an absolute difference in mean expression intensity between a given pair of classes (Dhanasekaran et al., 2001). Education of and prediction by our ANN were performed with NeuralWorks Professional II Plus v.5.3 software (NeuralWare, Carnegie, PA, USA) as described previously (O'Neill and Song, 2003).

Analysis of CNA

Genomic DNA was obtained from purified CD4+ ATL cells (n=24) and from MNCs of patient ID6 with the use of a QIAmp DNA Mini Kit (Qiagen, Valencia, CA, USA). Each DNA sample (250 ng) was digested with HindIII, ligated to Adaptor-Hind (Affymetrix), amplified by PCR, and subjected to hybridization with Mapping 50K Hind 240 arrays (Affymetrix). Chromosome copy number at each SNP site was inferred from hybridization signal intensity on the arrays with the use of CNAG software ( (Nannya et al., 2005). For a normal reference, we used array data of PB MNCs isolated from four healthy volunteers. Assessment of copy number for all SNP sites is demonstrated in Supplementary Table 6. The raw data of Mapping 50K Hind 240 arrays is available upon request. Statistical analysis of copy number was performed with GeneSpring 7.0. Alterations in the amount of genomic DNA were confirmed by quantitative real-time PCR with an ABI PRISM 7700 sequence detection system (PE Applied Biosystems, Foster City, CA, USA). The oligonucleotide primers were 5′-IndexTermAGCATGTCCACAAATGGCCTTTGG-3′ and 5′-IndexTermCAGTTTTCCTGTCATGGGAAAGGG-3′ for a region of chromosome 6, and 5′-IndexTermCTGACCTGCCGTCTAGAAAAACCT-3′ and 5′-IndexTermCAGGAAATGAGCTTGACAAAGTGG-3′ for the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene.

MET expression

ATL cells were stained with rabbit polyclonal antibodies to MET (Santa Cruz Biotechnology, Santa Cruz, CA, USA) or with mouse monoclonal antibodies to CD3 and to CD4 (both from BD Biosciences, San Diego, CA, USA) and were then subjected to flow cytometry with a FACScan instrument (BD Biosciences). The concentration of HGF in plasma was determined with a Quantikine ELISA kit for human HGF (R&D Systems, Minneapolis, MN, USA).

For quantitative RT–PCR analysis of MET expression, portions of nonamplified cDNA were subjected to PCR with a QuantiTect SYBR Green PCR kit (Qiagen). The oligonucleotide primers for PCR were 5′-IndexTermGTCAGTGGTGGACCTGACCT-3′ and 5′-IndexTermTGAGCTTGACAAAGTGGTCG-3′ for GAPDH cDNA, and 5′-IndexTermACTTGTTGCAAGGGAGAAGACTCC-3′ and 5′-IndexTermAGCGTTCACATGGACATAGTGCTC-3′ for MET cDNA.

KK-1 cell experiments

KK-1 cells were maintained in RPMI 1640 medium supplemented with 10% FBS and recombinant human interleukin-2 (10 ng/ml) (Sigma). For immunoblot analysis, cells were cultured for 48 h without FBS and interleukin-2 and then incubated for 10 min with recombinant human HGF (50 ng/ml) (Sigma) either alone or together with rabbit polyclonal antibodies to HGF (10 μg/ml) (Montesano et al., 1991). The cells were then lysed and subjected to immunoblot analysis with mouse monoclonal antibodies to phosphotyrosine (4G10, Upstate Biotechnology, Charlottesville, VA, USA) or with rabbit polyclonal antibodies to MET (#05–237, Upstate Biotechnology) as described previously (Yamashita et al., 2001). For assay of cell proliferation, serum-deprived KK-1 cells were cultured for 24 h at a density of 5 × 105/ml with HGF (50 ng/ml) either alone or together with antibodies to HGF (10 μg/ml) and were then mixed with MTS [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt]. Cell proliferation was measured with a CellTiter 96 Aqueous One Solution Cell Proliferation Assay (Promega, Madison, WI, USA).

Accession codes





Adult T-cell leukemia


polymerase chain reaction




artificial neural network


copy number alterations


hepatocyte growth factor


  1. Alon U, Barkai N, Notterman DA, Gish K, Ybarra S, Mack D et al. (1999). Broad patterns of gene expression revealed by clustering analysis of tumor and normal colon tissues probed by oligonucleotide arrays. Proc Natl Acad Sci USA 96: 6745–6750.

    CAS  Article  Google Scholar 

  2. Ariyama Y, Mori T, Shinomiya T, Sakabe T, Fukuda Y, Kanamaru A et al. (1999). Chromosomal imbalances in adult T-cell leukemia revealed by comparative genomic hybridization: gains at 14q32 and 2p16–22 in cell lines. J Hum Genet 44: 357–363.

    CAS  Article  Google Scholar 

  3. Bottaro DP, Rubin JS, Faletto DL, Chan AM, Kmiecik TE, Vande Woude GF et al. (1991). Identification of the hepatocyte growth factor receptor as the c-met proto-oncogene product. Science 251: 802–804.

    CAS  Article  Google Scholar 

  4. Broadus AE, Mangin M, Ikeda K, Insogna KL, Weir EC, Burtis WJ et al. (1988). Humoral hypercalcemia of cancer. Identification of a novel parathyroid hormone-like peptide. N Engl J Med 319: 556–563.

    CAS  Article  Google Scholar 

  5. Cesarman E, Chadburn A, Inghirami G, Gaidano G, Knowles DM . (1992). Structural and functional analysis of oncogenes and tumor suppressor genes in adult T-cell leukemia/lymphoma shows frequent p53 mutations. Blood 80: 3205–3216.

    CAS  PubMed  Google Scholar 

  6. Choi YL, Makishima H, Ohashi J, Yamashita Y, Ohki R, Koinuma K et al. (2004). DNA microarray analysis of natural killer cell-type lymphoproliferative disease of granular lymphocytes with purified CD3(−)CD56(+) fractions. Leukemia 18: 556–565.

    CAS  Article  Google Scholar 

  7. Dhanasekaran SM, Barrette TR, Ghosh D, Shah R, Varambally S, Kurachi K et al. (2001). Delineation of prognostic biomarkers in prostate cancer. Nature 412: 822–826.

    CAS  Article  Google Scholar 

  8. Edlich RF, Arnette JA, Williams FM . (2000). Global epidemic of human T-cell lymphotropic virus type-I (HTLV-I). J Emerg Med 18: 109–119.

    CAS  Article  Google Scholar 

  9. Funakoshi H, Nakamura T . (2003). Hepatocyte growth factor: from diagnosis to clinical applications. Clin Chim Acta 327: 1–23.

    CAS  Article  Google Scholar 

  10. Hatta Y, Hirama T, Miller CW, Yamada Y, Tomonaga M, Koeffler HP . (1995). Homozygous deletions of the p15 (MTS2) and p16 (CDKN2/MTS1) genes in adult T-cell leukemia. Blood 85: 2699–2704.

    CAS  PubMed  PubMed Central  Google Scholar 

  11. Imaizumi Y, Murota H, Kanda S, Hishikawa Y, Koji T, Taguchi T et al. (2003). Expression of the c-Met proto-oncogene and its possible involvement in liver invasion in adult T-cell leukemia. Clin Cancer Res 9: 181–187.

    CAS  PubMed  Google Scholar 

  12. Lin M, Wei LJ, Sellers WR, Lieberfarb M, Wong WH, Li C . (2004). dChipSNP: significance curve and clustering of SNP-array-based loss-of-heterozygosity data. Bioinformatics 20: 1233–1240.

    CAS  Article  Google Scholar 

  13. Lockwood WW, Chari R, Chi B, Lam WL . (2005). Recent advances in array comparative genomic hybridization technologies and their applications in human genetics. Eur J Hum Genet 14: 139–148.

    Article  Google Scholar 

  14. Matsumoto K, Okazaki H, Nakamura T . (1992). Up-regulation of hepatocyte growth factor gene expression by interleukin-1 in human skin fibroblasts. Biochem Biophys Res Commun 188: 235–243.

    CAS  Article  Google Scholar 

  15. Matsuoka M . (2003). Human T-cell leukemia virus type I and adult T-cell leukemia. Oncogene 22: 5131–5140.

    CAS  Article  Google Scholar 

  16. Migone TS, Lin JX, Cereseto A, Mulloy JC, O'Shea JJ, Franchini G et al. (1995). Constitutively activated Jak-STAT pathway in T cells transformed with HTLV-I. Science 269: 79–81.

    CAS  Article  Google Scholar 

  17. Montesano R, Matsumoto K, Nakamura T, Orci L . (1991). Identification of a fibroblast-derived epithelial morphogen as hepatocyte growth factor. Cell 67: 901–908.

    CAS  Article  Google Scholar 

  18. Motokura T, Fukumoto S, Takahashi S, Watanabe T, Matsumoto T, Igarashi T et al. (1988). Expression of parathyroid hormone-related protein in a human T cell lymphotrophic virus type I-infected T cell line. Biochem Biophys Res Commun 154: 1182–1188.

    CAS  Article  Google Scholar 

  19. Nannya Y, Sanada M, Nakazaki K, Hosoya N, Wang L, Hangaishi A et al. (2005). A robust algorithm for copy number detection using high-density oligonucleotide single nucleotide polymorphism genotyping arrays. Cancer Res 65: 6071–6079.

    CAS  Article  Google Scholar 

  20. Nosaka K, Maeda M, Tamiya S, Sakai T, Mitsuya H, Matsuoka M . (2000). Increasing methylation of the CDKN2A gene is associated with the progression of adult T-cell leukemia. Cancer Res 60: 1043–1048.

    CAS  PubMed  PubMed Central  Google Scholar 

  21. O'Neill MC, Song L . (2003). Neural network analysis of lymphoma microarray data: prognosis and diagnosis near-perfect. BMC Bioinformatics 4: 13.

    Article  Google Scholar 

  22. Oshiro A, Tagawa H, Ohshima K, Karube K, Uike N, Tashiro Y et al. (2006). Identification of subtype-specific genomic alterations in aggressive adult T-cell leukemia/lymphoma. Blood 107: 4500–4507.

    CAS  Article  Google Scholar 

  23. Poiesz BJ, Ruscetti FW, Gazdar AF, Bunn PA, Minna JD, Gallo RC . (1980). Detection and isolation of type C retrovirus particles from fresh and cultured lymphocytes of a patient with cutaneous T-cell lymphoma. Proc Natl Acad Sci USA 77: 7415–7419.

    CAS  Article  Google Scholar 

  24. Reiner A, Yekutieli D, Benjamini Y . (2003). Identifying differentially expressed genes using false discovery rate controlling procedures. Bioinformatics 19: 368–375.

    CAS  Article  Google Scholar 

  25. Sasaki H, Nishikata I, Shiraga T, Akamatsu E, Fukami T, Hidaka T et al. (2005). Overexpression of a cell adhesion molecule, TSLC1, as a possible molecular marker for acute-type adult T-cell leukemia. Blood 105: 1204–1213.

    CAS  Article  Google Scholar 

  26. Shimoyama M . (1991). Diagnostic criteria and classification of clinical subtypes of adult T-cell leukaemia-lymphoma. A report from the Lymphoma Study Group (1984–1987). Br J Haematol 79: 428–437.

    CAS  Article  Google Scholar 

  27. Tajima K . (1990). The 4th nation-wide study of adult T-cell leukemia/lymphoma (ATL) in Japan: estimates of risk of ATL and its geographical and clinical features. The T- and B-cell Malignancy Study Group. Int J Cancer 45: 237–243.

    CAS  Article  Google Scholar 

  28. Tamiya S, Etoh K, Suzushima H, Takatsuki K, Matsuoka M . (1998). Mutation of CD95 (Fas/Apo-1) gene in adult T-cell leukemia cells. Blood 91: 3935–3942.

    CAS  Google Scholar 

  29. Tamura M, Arakaki N, Tsubouchi H, Takada H, Daikuhara Y . (1993). Enhancement of human hepatocyte growth factor production by interleukin-1 alpha and -1 beta and tumor necrosis factor-alpha by fibroblasts in culture. J Biol Chem 268: 8140–8145.

    CAS  PubMed  Google Scholar 

  30. Tsukasaki K, Krebs J, Nagai K, Tomonaga M, Koeffler HP, Bartram CR et al. (2001). Comparative genomic hybridization analysis in adult T-cell leukemia/lymphoma: correlation with clinical course. Blood 97: 3875–3881.

    CAS  Article  Google Scholar 

  31. Tsukasaki K, Tanosaki S, DeVos S, Hofmann WK, Wachsman W, Gombart AF et al. (2004). Identifying progression-associated genes in adult T-cell leukemia/lymphoma by using oligonucleotide microarrays. Int J Cancer 109: 875–881.

    CAS  Article  Google Scholar 

  32. Uchiyama T, Yodoi J, Sagawa K, Takatsuki K, Uchino H . (1977). Adult T-cell leukemia: clinical and hematologic features of 16 cases. Blood 50: 481–492.

    CAS  PubMed  PubMed Central  Google Scholar 

  33. Wano Y, Hattori T, Matsuoka M, Takatsuki K, Chua AO, Gubler U et al. (1987). Interleukin 1 gene expression in adult T cell leukemia. J Clin Invest 80: 911–916.

    CAS  Article  Google Scholar 

  34. Yamada Y, Hatta Y, Murata K, Sugawara K, Ikeda S, Mine M et al. (1997). Deletions of p15 and/or p16 genes as a poor-prognosis factor in adult T-cell leukemia. J Clin Oncol 15: 1778–1785.

    CAS  Article  Google Scholar 

  35. Yamada Y, Ohmoto Y, Hata T, Yamamura M, Murata K, Tsukasaki K et al. (1996). Features of the cytokines secreted by adult T cell leukemia (ATL) cells. Leuk Lymphoma 21: 443–447.

    CAS  Article  Google Scholar 

  36. Yamashita Y, Kajigaya S, Yoshida K, Ueno S, Ota J, Ohmine K et al. (2001). Sak serine/threonine kinase acts as an effector of Tec tyrosine kinase. J Biol Chem 276: 39012–39020.

    CAS  Article  Google Scholar 

  37. Yoshida M, Miyoshi I, Hinuma Y . (1982). Isolation and characterization of retrovirus from cell lines of human adult T-cell leukemia and its implication in the disease. Proc Natl Acad Sci USA 79: 2031–2035.

    CAS  Article  Google Scholar 

  38. Zhao X, Weir BA, LaFramboise T, Lin M, Beroukhim R, Garraway L et al. (2005). Homozygous deletions and chromosome amplifications in human lung carcinomas revealed by single nucleotide polymorphism array analysis. Cancer Res 65: 5561–5570.

    CAS  Article  Google Scholar 

Download references


This study was supported in-part by a grant for Third-Term Comprehensive Control Research for Cancer from the Ministry of Health, Labor and Welfare of Japan, and by a grant for Scientific Research on Priority Areas ‘Applied Genomics’ from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

Author information



Corresponding author

Correspondence to H Mano.

Additional information

Supplementary Information accompanies the paper on the Oncogene website (

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Choi, Y., Tsukasaki, K., O'Neill, M. et al. A genomic analysis of adult T-cell leukemia. Oncogene 26, 1245–1255 (2007).

Download citation


  • adult T-cell leukemia
  • DNA microarray
  • MET
  • artificial neural network

Further reading


Quick links