Identification of subtype-specific genomic alterations in aggressive adult T-cell leukemia/lymphoma

Adult T-cell leukemia/lymphoma (ATLL) is a human T-lymphotropic virus type-I (HTLV-1)–mediated neoplasm.^1,3  ATLL develops tumors during a 40- to 70-year latency period from the time of HTLV-1 infection.⁴  The results of statistical analysis suggest that additional genetic alterations are essential for viral integrated T cells to become malignant.⁵ 

Four major clinical subtypes—smoldering, chronic, acute, and lymphoma—have been identified.⁶  The smoldering and chronic subtypes are characterized by an indolent clinical course, whereas the acute and lymphoma subtypes are aggressive and show poor prognosis, although their clinicopathologic features are clearly different. The acute type generally presents with hepatosplenomegaly, constitutional symptoms, elevated lactate dehydrogenase (LDH), and hypercalcemia,⁷  but these are seen less often in the lymphoma type. The lymphoma type is characterized by prominent lymphoadenopathy without peripheral blood involvement, while the acute type is characterized by abnormal lymphocytes in peripheral blood and blood circulating in the whole body.⁷  These differences suggest that these 2 subgroups develop tumors via distinct genetic pathways.

Two cytogenetic and 1 comparative genomic hybridization (CGH) study of genomic aberrations of ATLL have been reported.^8,10  The first 2 studies found that chromosome abnormalities of ATLL are complicated and occur more frequently in aggressive than in indolent ATLL. The third study, using conventional CGH analysis, showed that aggressive ATLL displayed more genomic gains and losses than the indolent type. However, despite these different clinicopathologic features of acute and lymphoma types, no comparative studies of the genomic aberrations of these subtypes have been reported. However, a comparative analysis of the genomic imbalances of these subtypes may provide important insights into the pathogenesis of aggressive ATLL.

To investigate the genomic aberrations of these subtypes, we performed high-resolution genome-wide array-based CGH (array CGH) for 17 patients with acute ATLL and 49 patients with lymphoma ATLL. For a further investigation of the relationship between genomic amplification and expression, we also performed quantitative real-time reverse transcription–polymerase chain reaction (RQ-PCR) to determine the target genes in some of the recurrent regions with high copy number gains.

Samples of peripheral blood and lymph nodes, together with clinical data, were obtained from 66 patients under a protocol approved by the Institutional Review Board of the Aichi Cancer Center. Informed consent was provided according to the Declaration of Helsinki. The patients were selected from those hospitalized between 1992 and 2004 at Fukuoka University School of Medicine, the University of the Ryukyus School of Medicine, Imamura-Bunin Hospital, and their affiliated hospitals. We identified 17 patients with acute ATLL and 49 patients with lymphoma type ATLL. The median survival from the time of diagnosis was 5.9 months for acute type and 12.5 months for lymphoma type. The diagnosis of ATLL was based on clinical features, hematologic characteristics, immunophenotype, the presence of serum antibodies to ATLL-associated antigens, and monoclonal integration of HTLV-I proviral DNA. The immunophenotype of the tumors was determined by immunohistochemistry for tissue sections and/or flow cytometry for cell suspensions (CD2, CD3, CD5, CD7, CD4, CD8, CD25, and CD45RO). Monoclonal integration of HTLV-I proviral DNA into ATLL cells was assayed by means of Southern blotting analysis. For the clinical subtypes of ATLL the classification by Shimoyama et al of acute, lymphoma, chronic, and smoldering types was adopted.⁶ 

Acute type. The acute type is characterized by a leukemic phase with a markedly elevated white blood cell count that consists of mainly abnormal lymphocytes known as “flower cells.” The acute type is also characterized by hypercalcemia and/or a high LDH value (more than twice the upper limit of the normal range). Tumor specimens from all 17 patients, comprising 9 men and 8 women, were all obtained from the peripheral blood, which contained more than 60% leukemic cells (more than 10 × 10⁹/L [10 000/μL]) at the time of diagnosis. The median age of the patients was 57 years (range, 44-78 years); 16 patients had elevated LDH, 9 had elevated serum calcium levels, and 9 had a poor performance status (Table 1).

Lymphoma type. The lymphoma type is characterized by prominent lymphoadenopathy without peripheral blood involvement. Tumor specimens were obtained from 26 men and 23 women whose median age was 64 years old (range, 36-85 years). No leukemic cells were recognized in the peripheral blood at the time of diagnosis (Table 1).

High molecular weight DNA was extracted from lymph nodes and peripheral blood by using standard proteinase K/RNase treatment and phenol-chloroform extraction. DNA and RNA samples were extracted from frozen tissue. Normal DNA was obtained from the blood of healthy male donors. RNA was prepared from lymph nodes and peripheral blood by homogenization in guanidinium thiocyanate and centrifugation through cesium chloride.

Monoclonal integration of HTLV-I provial DNA into ATLL cells was demonstrated by Southern blotting analysis. Briefly, high-molecular-weight DNA was extracted from mononuclear cells (MNCs). DNA samples (5 μg) were digested with EcoRI and size-fractionated on 0.75% agarose gels. They were then electrotransferred onto a nitrocellulose membrane and hybridized to randomly primed ³²P-labeled DNA probes specific for HTLV-I. Thereafter, the blots were washed at appropriate stringency and visualized by autoradiography.¹¹ 

The array consisted of 2304 BAC and PAC clones, covering the human genome at a resolution of roughly 1.3 Mb, from libraries RP11 and RP13 for BAC clones, and RP1, RP3, RP4 and RP5 for PAC clones. BAC and PAC clones were selected from the information from the National Center for Biotechnology Information (NCBI; http://www.ncbi.nlm.nih.gov/), Ensembl Genome Data Resources (http://www.ensembl.org/), and the University of California at Santa Cruz (UCSC) genome Bioinfomatics, and obtained from the BACPAC Resource Center at the Children's Hospital (Oakland Research Institute, Oakland, CA). Clones were ordered from chromosomes 1 to 22 and X within each chromosome on the basis of Ensembl Genome Data Resources from the Sanger Center Institute (February 2005 version). The locations of all the clones used for array CGH were confirmed by fluorescence in situ hybridization (FISH). For the array, 10 simultaneous hybridizations of healthy male versus healthy male were performed to define the normal variation for the log₂ ratio. A total of 113 clones with less than 10% of the mean fluorescence intensity of all the clones, with the most extreme average test-over-reference ratio deviations from 1.0 and with the largest standard deviation in this set of normal controls excluded from further analyses.

A total of 2235 clones (covering 2988 Mb, 1.3 Mb of resolution) were therefore subjected to further analysis. Since more than 96% of the measured fluorescence log₂ ratio values at each spot (2 × 2235 clones) ranged from +0.2 to –0.2, the thresholds for the log₂ ratio of gains and losses were set at log₂ ratios of +0.2 and –0.2, respectively. Regions of low-level gain/amplification were defined as log₂ ratio +0.2 to +1.0; those suspected of containing a heterozygous loss/deletion were defined as log₂ ratio –1.0 to –0.2, and those showing high-level gain/amplification as log₂ ratio greater than +1.0, and those suspected of containing a homozygous losses/deletion as log₂ ratio less than –1.0.^12,14 

Expression levels of mRNA were measured by means of real-time fluorescence detection using a previously described method.¹⁵  Briefly, the primers of CARMA1 were sense: 5′-CATTCTACATACCAAGGGGC-3′ and antisense: 5′-ATCTGCTGCTGCAGCTTGAT-3′; the primers of BCL11b were sense: 5′-CAAGCAGGAGAACATTGCAG-3′ and antisense: 5′-AGTGATCACGGATGAGTGAG-3′; the primers of IRF4 were sense: 5′-AGAAGAGCATCTTCCGCATC-3′ and antisense: 5′-TGCTCTTGTTCAAAGCGCAC-3′. The real-time PCR using SYBR Green I (Takara Bio, Tokyo, Japan) and the primers was performed with a Smart Cycler System (Takara Bio) according to the manufacturer's protocol. G6PDH served as an endogenous control, while the expression levels of CARMA1 mRNA in each sample were normalized on the basis of the corresponding G6PDH content and recorded as relative expression levels. All the experiments were performed in duplicate.

To analyze genomic regions for statistical differences between the 2 patient groups, the data set was constructed as follows. Genomic alterations were defined by a log₂ ratio threshold of +0.2 for copy number gain, and of –0.2 for copy number loss. Gained clones (log₂ ratio >+0.2) were inputted as “1” versus no-gained clones (log₂ ratio <+0.2) as “0” in an Excel template for each case (Microsoft, Redmond, WA). Similarly, lost clones (log₂ ratio <–0.2) were inputted as “1” versus no-lost clones (log₂ ratio >–0.2) as “0” in another Excel template. Cases showing genomic gain or loss were counted with an Excel template for each single clone (2235 clones in all) in the acute or the lymphoma group. Data analyses were then performed for the comparison of frequencies of gain or loss for each single clone between the acute and lymphoma groups (2235 tests each for gain and loss, 4470 tests in all). Fisher exact test for probability was used for the former analysis. The P value for screening of candidate clones was less than .05 for every analysis. When a candidate clone was identified, the clone's continuity with the subsequent clones was evaluated. In cases where the nth clone and subsequent k clones (k ≥ 0) were found to be candidate clones, the P value for continual association was calculated as

The Mann-Whitney U test was used for detecting significant differences in expression levels of IRF4 (6p25), CARMA1 (7p22), and BCL11B (14q32) in groups with and without the genomic amplifications. All the statistical analyses were conducted with the STATA version 8 statistical package (StataCorp, College Station, TX).

Representative individual examples of the high-resolution analysis of 2 patient samples are shown in Figure S1 (see the Supplemental Figure link at the top of the online article, at the Blood website). Array CGH detected both small and large chromosome areas of gains and deletions as well as clear amplification and deletion borders. Gains and losses of genetic material obtained from 66 patients (49 with lymphoma ATLL and 17 with acute ATLL) were subjected to data analysis.

Genomic imbalances in acute type. Of the 17 patients analyzed, 1 sample did not show any copy number change. The entire tumor set, including the case without any genomic alterations characteristic of the acute type, involved average copy number gains of 221 Mb and 4.2 regions, and average copy number losses of 133 Mb and 4.5 regions (Table 2). Regions of recurrent (> 20%) gain were detected on chromosomes 3, 8p11.21-8q13.1, 8q24.21-24.23, 9p24.3-24.1, 9q33.3-34.3, 14q31.1-32.33, and X, and regions of recurrent loss on chromosomes 2q37.1-37.3, 4q21.1-24, 6q14.1, 9q21.12-31.2, 17p13.1, and 19q13.31-13.2 (Figures 1A,2A).

Genomic imbalances in lymphoma type. Gains and losses of genetic materials obtained from 49 patient samples of the lymphoma type were subjected to data analysis. All of the cases demonstrated genomic aberrations. The entire tumor set involved average copy number gains of 356 Mb and 9.7 regions and average copy number losses of 274 Mb and 9.8 regions (Table 2). Regions of recurrent (> 20%) gain were detected on chromosomes 1p36.23-36.21, 1p21.3-23.3, 1p25.2-1q44, 2p25.1-22.3, 2p2.1-21, 3p26.3-q12.3, 3q25.2-29, 4p16.3-q12, 4q13-34.1, 6p25.3-24.1, 7p22.3-15.2, 7p14.3-14.2, 7p13-q36.3, 8q21.3-22.2, 8q24.22-24.21, 9p24.2-24.1, 9q34.3, 14q31.1-32.33, 18cen-23q, 21q21-22.2, and X, and losses at 1p31.2-12, 2q36.1-37.3, 3q13.13-24, 4q13.2-22.3, 5q15-23.1, 6q14.1-22.31, 8p23.3-21.2, 8q21.13, 9p21.3, 9q21.12-21.33, 10p15.3-cen., 13q12.3-34, 14q11.2-21.1, 16cen-16q24.1, 17p13.3-12, 18p11.32-cen, and 19p13.11-13.32 (Figures 1B,2B).

A comparative analysis of the differences in genomic frequencies between acute and lymphoma types identified several common regions for these 2 subtypes. The common regions of gains were found on 3p36-q12, 3q25-29, 8q24, 9p24, 9q34, and 14q32, and those of losses on 2q37, 4q21, 6q14, 9p21, 9q21, 17p13.1, and 19q13. Statistically significant differences were a gain of 3p26-q26 for the acute type and gains at 1q22-23, 1q41-44, 2p25, 4q21-22, 4q26-27, 7p22, and 7q11-36, and losses of 10p12, 13q21-32, 16q21-22, and 18p11 for the lymphoma type. The regions characteristic of each type as determined are listed in Table 3.

Recurrent regions (≥ 2 patients) of high-level amplification (log₂ ratio > +1.0) were found at 1p36, 6p25, 7p22.2, 7q21-q22, 7q31-q36, and 14q32 (Table 4) in the lymphoma-type samples, and no recurrent amplifications or homozygous losses in the acute type. In an attempt to identify target genes in the high-level amplification regions, those with amplification detected by a single BAC clone were further analyzed (Figure 3). The 3 regions containing candidate genes that have been implicated in lymphoid malignancies were IRF4 at 6p25.3, CARMA1 at 7p22.2, and BCL11B at 14q32 (2 patients each).^17,19  Candidate target genes of the other regions (eg, 1p36, 7q21-q22, and 7q31-q36) are currently being investigated.

A recurrent region of homozygous loss (log₂ ratio <–1.0) was found at 9p21.3. As shown in Table 4, the most frequently detected clone of homozygous loss at 9p21.3 was RP11-149I2 (2 cases: D1679 and D1687), which contains the CDKN2A and CDKN2B tumor suppressor genes. Other recurrent regions of homozygous loss are suspected to be 6q21 and 13q32.

IRF4, CARMA1, and BCL11B of the 6p25, 7p22, and 14q32 regions were examined for their potential as candidate target genes. These genes are included in the BAC clones with high copy number gains (RP11-498D18, CARMA1; and RP11-233K4, IRF4) or located very close to such BAC clone (BCL11B; 300 kb telomeric to RP11-1127D7). To identify the candidate target genes for these genomic regions, we performed RQ-PCR analysis for 21 patients either with or without this gain/amplification. Although the statistical power of this analysis is limited because of the small number of cases, the expression levels of CARMA1 of the patients with copy number gains at 7p22 were significantly higher than of those without (Mann-Whitney U test, P = .045). However, high expression of CARMA1 was found in the acute type regardless of the 7p22 gain/amplification. We therefore concluded that the CARMA1 gene is a target for 7p22 genomic amplification in the lymphoma type but not in the acute type. Repeated experiments showed similar results with a variance of less than 5%; representative data are shown in Figure 4.

With regard to the 14q32 region, RQ-PCR for BCL11B failed to demonstrate high-level expression in the cases with high copy number gains, indicating that BCL11B is not a candidate gene for the 14q32 region of the lymphoma type (Figure 5). Interestingly, high-level expression of BCL11B was noted in the acute type lacking genomic amplification. Various levels of IRF4 expression were observed regardless of genomic gains, indicating that IRF4 is not a likely candidate gene of this region (data not shown).

ATLL is categorized as a single-disease entity; the neoplasm originated from HTLV-I–infected T cells in the World Health Organization (WHO) classification, although the disease is comprised of 4 subtypes—smoldering, chronic, lymphoma, and acute—which reveal distinct clinicopathologic features.⁷ 

Previous cytogenetic analysis of ATLL indicated that this disease is characterized by complicated and miscellaneous genomic alterations. Kamada et al studied 107 patients with ATLL and reported recurrent cytogenetic abnormalities, including gains of chromosomes 3, 7, and 21, and various losses.⁸  Tsukasaki et al analyzed 46 patients with aggressive ATLL (acute and lymphoma types) by means of chromosomal CGH and reported frequent genomic gains at 3p, 7q, and 14q, and losses at 6q and 13q.¹⁰  They also demonstrated that aggressive ATLL featured significantly more frequent genomic aberrations than did indolent types such as chronic and smoldering. However, they did not perform a comparative analysis of genomic alterations in acute and lymphoma types. Our study showed that gain at chromosome 3p26-q26 was a frequent genomic aberration of the acute type, while gain at 7q and loss at 13q were characteristic of the lymphoma type.

The common genomic alterations may represent the common oncogenic changes for both types, while differences in their genome profiles may represent different clinical phenotypes. The loss of 9p21 was detected in 20% of patients with acute ATLL and 33% of patients with lymphoma ATLL. The loss of 9p21, which harbors the CDKN2A and CDKN2B tumor suppressor gene, has been frequently found in aggressive ATLL but not infrequently in indolent types.^20,21  This suggests that deregulation of tumor suppressor genes such as CDKN2A and CDKN2B is likely to lead to the transition from indolent to aggressive type. Gain of the 3p26-q26 region is characteristic of the acute type of ATLL, which in turn is characterized by a leukemic phase with flower cells. Recently, it has been demonstrated that phosphatidylinositol 3-kinase, catalytic, beta (PIK3CB2) located at 3q22.3 is associated with flower cell formation and the cellular proliferation of ATLL.²²  It is tempting to speculate that the gain of 3p26-q26 results in aberrant overexpression of the PI3-kinase, which then leads to the characteristic clinicopathologic features of the acute type. The most frequent gain characteristic of the lymphoma type is gain of 7q11-36.3 (57%), which was absent from the acute type. The most frequent loss characteristic of the lymphoma type is that of 13q21.1-q32.1 (41%), which was found in only 1 (6%) of 17 patients with the acute type. The candidate genes for these regions, as well as several other regions of lymphoma type which showed significant alterations, may well be responsible for the characteristic clinicopathologic features of the lymphoma type.

The comparative analysis of the genome profiles of acute and lymphoma types identified distinct patterns of genomic alteration, suggesting that these entities might use distinct genomic pathways to develop tumors. The acute type of ATLL may develop tumors via a different mechanism such as translocation or/and mutation with fewer genomic alterations, whereas the lymphoma type may do so through a stepwise process with accumulation of genetic alterations.

We were able to demonstrate that CARMA1 is a possible target of 7p22 amplification in the lymphoma type, making this the first report of aberrant CARMA1 expression in human cancer. CARMA1 is a lymphocyte-specific member of the membrane-associated guanylate kinase (MAGUK) family of scaffolding proteins. It interacts with BCL10 via its caspase-recruitmant domain (CARD), functions as an intermediate in the T-cell receptor (TCR) signal transduction pathway^23,25  and is a critical regulator of TCR inducing nuclear factor-κB (NF-κB) activation. Because NF-κB is known to be constitutively activated in primary adult T-cell leukemia cells, its activation may be essential for ATLL development.²⁶ 

Overexpression of CARMA1 leading to activation of the NF-κB pathway might play a crucial role in the tumorigenesis of acute and lymphoma types. In the lymphoma type, overexpression of CARMA1 was found in patients with 7p22 amplification, while in the acute type, CARMA1 was overexpressed without any 7p22 genomic gain. This makes it likely that a still unknown mechanism other than genomic amplification is responsible for CARMA1 activation in the acute type.

It is known that the viral protein Tax induces the activation and nuclear translocalization of NF-κB,^27,30  but primary ATLL cells express only a low level of Tax mRNA and protein.^31,32  These facts indicate that there must be a mechanism other than Tax for NF-κB activation. CARMA1 overexpression detected in our study is one of the mechanisms leading to constitutive activation of NF-κB in ATLL cells regardless of the level of Tax. However, the presence of patients with lymphoma-type ATLL lacking CARMA1 overexpression indicates that other pathways for NF-κB activation may exist in ATLL.

RQ-PCR examination of BCL11B gene³³  expression for acute and lymphoma types showed that genomic amplifications of 14q32 did not correlate well with the expression levels of this gene. This means that BCL11B gene is not a target for 14q32 genomic amplification. Interestingly, however, considerably higher levels of BCL11B expression irrespective of 14q32 amplification were found in the acute type, suggesting that a mechanism other than genomic amplification induces BCL11B overexpression. Since several studies have shown that inv(14)(q11;q32) and t(14;14)(q11;q32) are recurrent chromosomal aberrations in ATLL,^34,35  aberrant overexpression of BCL11B may be due to translocation of 14q.¹⁸  A candidate gene other than BCL11B must therefore exist in this 14q32 region of amplification for the lymphoma type and warrants further investigation. The difference in the expression profile of BCL11B between acute and lymphoma types also supports the concept that acute and lymphoma types are different disease entities.

In conclusion, genome profiling by array CGH demonstrated that the lymphoma type is characterize by more frequent gains and losses than is the acute type, and it was found that the patterns of genome profiles are also different for these 2 types. CARMA1 was identified as a possible target of 7p22 amplification in the lymphoma type but no such amplification was found in the acute type although the expression of 7p22 was high. BCL11B was not overexpressed in the lymphoma type with 14q32 amplification but a relatively high expression of BCL11B was found in the acute type regardless of 14q32 amplification. These results suggest that acute and lymphoma types are distinct subtypes, and may develop via distinct genomic/genetic pathways.

The outstanding technical assistance of Ms H. Suzuki and Ms Y. Kasugai is very much appreciated. We also wish to thank Dr Ryuzo Ohno, the chancellor of Aichi Cancer Center, for his overall support.