HOX11L2 expression defines a clinical subtype of pediatric T-ALL associated with poor prognosis

Extensive characterization of chromosomal abnormalities specific for T-ALL led to the identification of several oncogenes whose expression was up-regulated under the influence of the transcriptional regulation elements of genes normally expressed during T-cell differentiation. These T-cell–specific oncogenes code for transcription factors, and their oncogenicity has been well demonstrated in mouse models.1,2 However, though they have been proven useful for monitoring residual disease, these T-ALL–specific molecular abnormalities are present in only 20% to 30% of samples and do not allow improvement of risk-based therapies.3,4 

The HOX11 gene is an orphan homeobox gene isolated because of its transcriptional activation as a result of the t(10;14)(q24;q11) or t(7;10)(q35;q24) translocation of T-ALL.5 We have recently characterized a T-ALL–specific cryptic chromosomal translocation, t(5;14)(q35;q32), which is present in approximately 20% of childhood T-ALL samples and is associated with strong expression ofHOX11L2, a HOX11-related gene.6,7 To gain insight into the relation between HOX11L2 expression and the other known T-ALL abnormal molecular features, we analyzed a series of 28 children with T-ALL for the presence of other frequent oncogenic events. We analyzed those samples for the expression ofTAL1/SCL, LYL1, LMO28,9and the 3 HOX11 family genes.10 We also investigated for SIL-TAL1 fusion and deletion of the CDKN2A/CDKN2D locus.11 Relationships with immunologic and clinical data are presented.

Patients were children with T-cell malignancies diagnosed at Trousseau Hospital (Paris, France) from March 1996 to September 2001. Diagnoses were based on standard morphologic and histochemical parameters of leukemic cells and on the expression the T-cell antigens, cytoplasmic CD3, CD2, CD5, CD7, and the absence of B-cell antigens. Thirty-two patients were given diagnoses of T-ALL during that time, and samples for 28 were available. Informed consent was obtained from patients and their parents according to the Declaration of Helsinki. Clinical data of T-ALL patients are summarized in Table1. Samples from children with T-lymphoblastic lymphoma (5 patients), B-ALL (52 patients), and acute myeloblastic leukemia (AML, 15 patients) diagnosed during the same period of time were also used. Bone marrow (BM) or peripheral blood (PB) samples were obtained at the time of diagnosis and during cytologic remission or at relapse and were cryopreserved.

Patients were treated according to the French multicenter risk-adapted protocol FRALLE.12 T-ALL patients were assigned to the high-risk arm. Corticoresistance—defined as the persistence of more than 1000 circulating blast cells per microliter after 8-day treatment (with 60 mg/m² prednisone per day and 1 intrathecal injection of methotrexate–cytosine–arabinoside–prednisolone) and/or M3 status13—defined as the persistence of more than 25% blasts in BM aspiration at day 21—were indications that patients should be moved to the very high-risk arm and should undergo allogeneic or autologous bone marrow transplantation during the first complete remission (CR). Clinical data of the patients are presented in Table 1.

Mononuclear cells isolated from bone marrow aspirates were stained with a standard panel of antibodies against CD2, CD5, CD7, CD3, CD1a, TCR αβ, TCR γδ, CD19, CD20, CD22, CD24, CD79α, IgM, CD10, CD34, HLA-DR, CD45, CD13, CD15, and CD33. Results were classified as positive when a given monoclonal antibody stained more than 20% of leukemic cells. At least 5000 cells per sample were analyzed using a FACSort cytometer and CELLQuest software (both from Becton Dickinson, Le Pont de Claix, France).

Total RNA was extracted from frozen patient samples using RNAble (Eurobio, Les Ulis, France) according to the manufacturer's instructions. RNA from normal fetal spleen, fetal thymus, and adult spleen were obtained from commercial sources. Adult thymus RNA was a kind gift from Karen Leroy (Hôpital Henri Mondor, Créteil, France), and 4 normal fetal thymi were a kind gift from Jelena Martinovic (Hôpital Necker–Enfants Malades, Paris, France). Normal peripheral blood lymphocyte RNA was extracted from healthy donor samples, and 5 normal bone marrow RNA samples were obtained from B-ALL patients in long-term remission (CR longer than 5 years). Genomic DNA was extracted using standard methods.

RNA was reverse transcribed from 1 μg total RNA in a final volume of 20 μL containing reverse transcription–polymerase chain reaction (RT-PCR) buffer (1 mM each dNTP, 3 mM MgCl₂, 75 mM KCl, 50 mM Tris-HCl pH 8.3), 10 U RNAsin (Promega, Madison, WI), 100 mM dithiothreitol, 100 U Superscript II (Gibco-BRL, Cergy Pontoise, France), and 25 μM random hexamers.

One hundred nanograms cDNA equivalent of RNA was analyzed in each PCR experiment. PCR was carried out in a final volume of 50 μL with 0.5 U AmpliGold polymerase (PE Applied Biosystems, Foster City, CA), 200 μM each dATP, dCTP, dGTP, 400 μM dUTP, 25 pmol each primer (see below), and 2.5 mM MgCl₂. Cycle parameters were set for 10 minutes at 95°C and for 15 seconds at 95°C, 40 seconds at 60°C, and 40 seconds at 72°C for 35 cycles. Specificity of each HOX11,HOX11L1, and HOX11L2 RT-PCR product was confirmed by direct nucleotide sequence analysis.

Screening for HOX11L2, HOX11L1, andHOX11 expression was carried out by standard RT-PCR using the following primers: HOX11L2 2Fo, GCGCATCGGCCACCCCTACCAGA;HOX11L2 3Rw, CCGCTCCGCCTCCCGCTCCTC;HOX11L1-286Fo, AGCACCTGTGAGCGGGAGAAG;HOX11L1-413Rw, GYGCCTGGGCCCTCGGGTTTG;HOX11-712Fo, CTGGCCAAGGCGCTCAAAATG; andHOX11-810Rw, GGCCTCCCGTTCCTCCGCAGTC.

Primers and probe sequences used for real-time quantitative PCR amplification (RQ-PCR) of HOX11L2 and HOX11 were selected with the assistance of the computer program Primer Expres (PE Applied Biosystem). They were designed on different exons to avoid amplification from residual genomic RQ-HOX11L2-3, CAAGACCTGGTTCCAAAACCG; RQ-HOX11 L2-4, AGGCTGGATGGAGTCGTTGA; probe, FAM-CAGCTGCAACACGACGCCTTCCAA-TAMRA; RQ-HOX11-1F, AAATGACCGATGCGCAGGT; RQ-HOX11-2R, GTTCGCTTGCTGCCTCTCG; and probe, FAM-AACCGGCGGACAAAGTGGAGACG-TAMRA. Expression analyses ofTAL1, LYL1, and LMO2 were performed using the primers described by Ferrando et al.14 

The CDKN2A/CDKN2D copy number was estimated using an RQ-PCR assay starting from genomic DNA (adapted from15). SIL-TAL1 fusion RNA was detected as recommended by the BIOMED consortium for minimal residual studies in acute leukemia.16 Detection of theABL transcript was performed as described.17 

Theoretical and practical aspects of RQ-PCR have already been described.18 Figure 1A shows changes in reporter fluorescence during PCR reactions starting from the known copy number (10⁸ to 10²) of theHOX11L2 cDNA. Fluorescence was expressed as normalized F1/F2 against background fluorescence accumulated during the first 15 PCR cycles, whereas cycle number indicated the point of the reaction at which the fluorescence generated by the cleavage of the probe crossed a fixed threshold above the baseline (crossing point or Cp). The standard curve was obtained by correlation of the standard concentration versus the Cp value, as shown in Figure 1B. A strong correlation between the Cp and the HOX11L2 copy number (r > 0.99) was found over a range of at least 6 orders of magnitude, with a PCR efficiency value of 90%. Similar results were obtained withHOX11 standard and the endogenous control ABL(data not shown).

For each patient and control sample, the quality and quantity of RNA were assessed by the amplification of ABL gene transcripts in independent RQ-PCR runs. Samples were considered eligible for testing only when the Cp of the internal referenceABL was lower than 30. Quantitative results were thus expressed as 1000 times the normalized copy number of the target gene against the copy number of the endogenous ABLgene.

PCR reactions were performed using the Light Cycler System (Hoffman-LaRoche, Grenoble, France). For each reaction, 100 ng reverse-transcribed RNA sample was added to 15 μL vol PCR mix containing 1× LC master mix, 5 mM MgCl₂, 300 μM each primer, and 200 μM probe. Thermal cycling conditions consisted of an initial denaturation step at 95°C for 10 minutes followed by 50 cycles at 95°C for 15 seconds and 60°C for 1 minute. Experiments were performed in duplicate for each data point. Each PCR run included the standard curve, a control without reverse transcriptase, and a control without template.

Cytogenetic studies were performed in Hôpital Saint Antoine on bone marrow, blood cells, or both after short-term cultures for 17 and 24 hours. RHG banding technique was applied. Karyotypes are summarized in Table 2. Fluorescence in situ hybridization (FISH) was carried out using the usual techniques.19 In addition to whole chromosome 5 and 14 painting probes (INSERM U301 and Appligene Oncor [Illkirch, France], respectively), chromosome 5 YAC clone (885a6) and BAC clones (already described or selected from available human sequence data bases) were used as probes as follows: HOX11L2 (5q35)—593F7, 2248N14, 45L16; 14q32—2576L4, 1082A3; CDKN2A/CDKN2D (9p21)—145E5, 70L8; HOX11 (10q24)—31L3; and HOX11L1(2p13)—140K4.

Qualitative and categorized quantitative variables were compared to each other using χ² analysis with Yates correction. For event-free survival (EFS) time, the period taken into account was the interval between the diagnosis and an event or the last examination if no events occurred. Survival curves were calculated according to the Kaplan-Meier method with Statview 4.5 software (Abacus Concept, Berkeley, CA), and differences were assessed using the log-rank test.

We used RT-PCR to detect HOX11L2 expression in a panel of 28 pediatric patients with T-ALL, 5 with T-lymphoblastic lymphoma, 52 with B-ALL, and 15 with AML. No specific fragment could be amplified from the T-lymphoblastic lymphoma, B-ALL, or AML samples, whereas a single 244-nucleotide fragment was observed in 6 of 28 (21.4%) T-ALL samples (Figure 2; Table 2).

To evaluate the expression level of the HOX11L2 gene, we developed a real-time quantitative RQ-PCR assay that was used to analyze the patients expressing HOX11L2. In these samples, the expression was homogeneously high, with HOX11L2normalized copy number ranging from 3984 to 21 680. In contrast, when investigated in normal hematopoietic tissue, HOX11L2expression was not detected in adult spleen, adult peripheral blood lymphocytes, or control bone marrow. Using the same assay,HOX11L2 was found to be expressed at a very low level in fetal thymus (1.2 normalized copy number), fetal spleen (0.57 normalized copy number), and adult thymus (0.7 normalized copy number). Taken together, these results are consistent with a restricted expression pattern of the HOX11L2 gene and a specific association between high HOX11L2 expression and T-ALL within leukemic samples.

The expression of HOX11 was observed in 3 of 28 (10.7%) samples, though 1 (UNPT5) exhibited a very low level of expression when compared to the 2 others. Virtually no expression was detected in thymus or bone marrow controls. Results are summarized in Table 2. All patients except UNPT2 were searched for the expression ofHOX11L1, the third member of the HOX11 family. One (UPNT1) exhibited low expression of HOX11L1 in addition to HOX11L2 (data not shown).

We next searched for the expression of other T-cell oncogenes and for TAL1, LYL1, and LMO2 genes. Because those genes are expressed during normal hematopoietic differentiation, we used RQ-PCR analyses to accurately estimate the level of gene expression. For LYL1 and LMO2, samples expressing more than the bone marrow controls were considered positive.

For TAL1, samples expressing more than the weaker expressingSIL-TAL1–positive patient (UPNT28, see below) were considered positive. As shown in Table 2, using these thresholds, 14 patients were found positive for LYL1, 9 forLMO2, and 14 for TAL1. Of note, χ²analysis suggests a significant association between LYL1 andLMO2 expression (P = .002).

We next examined the T-ALL samples for the presence of other known T-ALL frequent, specific oncogenic events. TheSIL-TAL1 fusion is specific for T-ALL and is known to be due to an infra-microscopic (approximately 80-kb) deletion that leads to the transcription of the first exon of SILfused to the coding sequences of TAL1. In our T-ALL patients, the SIL-TAL1 fusion transcript was detected in 6 of 28 (21.4%) samples (Table 2), in keeping with previous reports.8 

In one instance (UPNT5), HOX11 expression was observed in a patient with SIL-TAL1 fusion. No co-expression ofHOX11L2 with SIL-TAL1 orHOX11 was observed in any patient.

A frequent but nonspecific event observed in T-ALL is the inactivation of the CDKN2A/CDKN2D genes, occurring mainly through the deletion of 1 or 2 gene copies.11 When genomic DNA was available, the number of CDKN2A/CDKN2D copy was estimated by RQ-PCR according to the described procedure.15 The status of this locus was also investigated by FISH analysis using BAC clones specific for theCDKN2A/CDKN2D locus on 9 patients. Taken together, homozygous deletion of CDKN2A/CDKN2D copies was found in 13 of 24 samples (Table 2), in keeping with previous reports,11 but does not appear to be obviously associated with the other specific lesions investigated here—HOX11, HOX11L2, andSIL-TAL1 expression.

We next wanted to compare our molecular results with cytogenetic data. Karyotype analysis of the patient samples is summarized in Table2. Two patients exhibited a recognized chromosomal translocation: UPNT3 exhibited t(10;11)(p14;q14), expected to result in aCALM-AF10 fusion gene,20 and UNPT9 showed t(8;14)(q24;q11), expected to lead to transcriptional activation of the c-MYC gene.21 No cytogenetic abnormality of theLYL1(19p13), TAL1 (1p32), or LMO2(11p13) loci was observed in these samples.

Because it detects an SIL-TAL1 fusion transcript, the RT-PCR assay allows direct detection of the rearrangedTAL1 copy, which can be considered a bona fide oncogenic event. In contrast, because the known abnormalities of HOX11and HOX11L2 genes result in transcriptional activation of these genes without any kind of fusion at the RNA level, RT-PCR analysis demonstrated only the expression of these genes, without any clue to the underlying molecular reasons. To establish whether the expression of these 2 genes was associated with a structural abnormality of the corresponding locus, we investigated the structure of chromosome 10, on which the HOX11 gene lies, and of chromosome 5, on which the HOX11L2 gene lies, using classical cytogenetic and FISH analyses.

Three samples show HOX11 expression. No material for cytogenetic studies was available for UNPT5 or UPNT6. No 10q24 abnormality could be uncovered in UPNT27 through FISH using a specific BAC clone (data not shown). Based on our RQ-PCR data, HOX11was not detectably expressed during normal bone marrow differentiation, and the molecular reasons for its expression in patient samples without 10q24 abnormalities remain to be established. Similarly, no abnormality of the HOX11L1 locus could be detected through FISH analysis using a specific BAC clone on UNPT1 material (data not shown).

HOX11L2 expression was observed in 6 patients, but material for cytogenetic studies was available for further investigation in only 4 patients. In 2 of them (UPNT11 and UPNT24), FISH analysis uncovered t(5;14)(q35;q32), known to be associated with HOX11L2expression. The involvement of the HOX11L2 locus was further established using specific BAC probes (data not shown). In the 2 other patients, conventional karyotype was normal in UNPT1 and showed a der(5) in UNPT8 (Table 2). The involvement of the HOX11L2locus was assessed in these 2 patients using samples obtained in relapse through the use of chromosome 5 YAC and BAC probes. These probes gave a split signal on each patient's metaphase chromosomes, between a der(5) and a submetacentric chromosome identified by whole chromosome painting as chromosome 7 in each instance (data not shown).

Because HOX11L2 expression is barely detected in normal hematopoietic tissues, we used RQ-PCR to monitor HOX11L2expression and to follow the disease course in 2 patients.HOX11L2 level of expression was estimated in bone marrow samples from UPNT18 collected at complete remission (CR1) and 4 months after diagnosis. Samples were also analyzed at the time of patient UNPT18 relapse, which occurred 10 months after diagnosis, and were collected 13 and 14 months after diagnosis during CR2. A second relapse occurred 1 month after the last sample analyzed.

Similar analysis was performed on samples from UNPT24, collected at diagnosis and at CR1. Two samples, collected 4 and 9 months after diagnosis while the patient was still in CR1, were also investigated. As shown in Figure 3, samples from UPNT18 showed low but consistently detectable levels of HOX11L2expression, even in samples obtained after remission appeared as complete, based on cytologic and immunologic data. This high level ofHOX11L2 expression is likely to reflect the persistence of leukemic cells. The patient had a relapse 9 months after CR1, and againHOX11L2 expression remained detectable in CR2. Furthermore, a dramatic increase of HOX11L2 expression was observed 1 month before the occurrence of a second relapse. In contrast, in UPNT24 patient samples in CR1, HOX11L2 expression rapidly decreased and remained at very low levels. These data illustrate the potential use of HOX11L2 expression as a marker for minimal residual disease monitoring.

We next searched for an association between a specific immunophenotype and one of the T-ALL subgroup defined by the expression of HOX11, HOX11L2, or the presence of theSIL-TAL1 fusion. Data are summarized in Table3. As previously described, the SIL-TAL1–positive samples were not associated with a precise phenotype but are in keeping with a late cortical phenotype (CD3⁺, CD8⁺).

All the HOX11L2-positive samples were positive for CD1a and CD4; the association was close to significance with CD4 (P = .055). Most of the samples were also positive for CD10 (4 of 6; 66.6%), CD8 (4 of 6; 66.6%), and CD34 (4 of 6; 66.6%). When compared with HOX11-positive samples, the prominent differences are a less frequent expression of CD10 and CD34 in the latter samples. Taken together, these data indicate a slightly different immunophenotype for HOX11- andHOX11L2-expressing samples. Nevertheless, each reflected an early cortical stage of thymocyte differentiation.

Six patients in this series had relapses. Among them, 4 had been shown to express HOX11L2 at diagnosis. Only one patient withSIL-TAL1 fusion (and t(8;14)) and none withHOX11 expression had relapses. Association of the clinical outcome (24-month EFS) with clinical and genetic data were submitted to statistical analysis. Selected results are shown in Table 4. WBC andHOX11L2 expression were shown to be significantly associated with a relapse event (P = .05 and P = .02, respectively). Interestingly, as shown in Figure4, when adjusted for WBC,HOX11L2 expression remained significantly predictive of relapse in this series (P = .03).

The high frequency of ectopic expression of the HOX11L2gene in childhood T-ALL prompted us to draw a broad molecular picture of this disease. We investigated a panel of 28 pediatric patients with T-ALL for the presence and association of the frequent known oncogenic activation, expression of the related HOX11 andHOX11L2 genes, SIL-TAL1 fusion, and gross alteration of the CDKN2A/CDKN2D locus.

A surprisingly high number of samples was found to express at least 1 of 3 other T-cell oncogenes tested—TAL1,LYL1, and LMO2—in the absence of obvious abnormalities of their loci. In this series, theLMO2-positive samples also expressed LYL1, a feature of immature hematopoietic cells. TAL1 expression in the absence of obvious genomic alteration of its locus has been reported in T-ALL samples, though the proportions are still controversial.14,22-24 Whether expression of those genes reflects bona fide oncogenic activation of a transcriptional regulatory cascade or merely a block in differentiation remains to be established.

The expression of HOX11 and HOX11L2 and the presence of SIL-TAL1 fusion was observed in 3 (10%), 6 (21%), and 6 (21%) patients, respectively, in keeping with previous reports.2,8 As expected, a high frequency ofCDKN2A/CDKN2D alterations was observed, that were not obviously associated with the other genetic features studied here. Similarly, HOX11L2 expression was not associated with another known oncogenic activation. In contrast,SIL-TAL1 fusion was observed in association with a c-MYC activation in one patient and with HOX11expression in another (UNPT5), indicating a possible collaboration between TAL1/SCL and these oncogenes in leukemogenesis.

However, HOX11 was expressed at a very low level in UPNT5 in comparison with the 2 other patients expressing HOX11, challenging a direct role in leukemogenesis. In addition, HOX11 and HOX11L1 were expressed in 2 samples in the absence of any obvious chromosomal abnormality of the corresponding locus, a situation already reported for HOX11.14,25 The mechanism and the meaning of these expressions are yet to be established.

HOX11L2 expression was detected in T-ALL but not in a series of B-ALL and AML patients, underscoring the specificity of the HOX11L2 ectopic expression. Cytogenetic and FISH analyses could be performed in 4 of 6 T-ALL patients expressing HOX11L2 and unraveled a t(5;14) translocation in 2 patients and a t(5;7) translocation in the 2 others. These data show that HOX11L2 expression in T-ALL is dependent on a molecular abnormality of 5q35. In addition, FISH analysis of 14 samples did not identify additional patients with a 5q35 abnormality (data not shown), confirming a tight relation between 5q35 abnormality and expression of HOX11L2. Because no or very low expression of HOX11L2 is detected in normal samples, we applied this quantitative assay to follow the outcome of 2 patients. Interestingly, the expression level of HOX11L2 was similar at diagnosis in both patients but differed markedly in the early times of treatment. One patient who maintained a relatively high level of HOX11L2 expression had a relapse, whereas the other patient showed a quick drop in the expression level of HOX11L2 and did not have a relapse for 10 months. These data suggest that monitoring HOX11L2 expression could be useful to follow the clearance of leukemic cells during the early phases of treatment, which is thought to be important for risk assessment.26,27 

SIL-TAL1 fusion has not been associated with a distinct T-ALL clinical subgroup.8 On the contrary, HOX11L2-expressing samples might define a T-ALL subgroup which shows a constant expression of CD1a and CD4 because all patients analyzed to date express both antigens.6,7 

Successful treatment is achieved in approximately 60% to 75% of T-ALL patients. Interestingly, 4 of 6 patients in this series who had relapses were HOX11L2-expressing patients, whereas only one exhibitedSIL-TAL1 fusion and none expressed HOX11. All relapse events were observed in the group of patients with high WBC counts (greater than 50 × 10⁹/L) at diagnosis. Interestingly, all patients of this group who expressed HOX11L2 had relapses. Although the small size of this series prevents any definitive conclusion, this observation indicates that HOX11L2 patients have poor prognoses. Additional studies are needed to validate these conclusions for clinical application.

We thank Fréderique Siotto for assistance in writing this manuscript, Brigitte Godard for help in collecting clinical data, Pierre Laurent-Puig for help in statistical analyses, and A. T. Look and A. A. Ferrando for primer sequences.