Genes contributing to minimal residual disease in childhood acute lymphoblastic leukemia: prognostic significance of CASP8AP2

Response to therapy in childhood acute lymphoblastic leukemia (ALL) is ultimately linked to the expression of genes that control cellular drug sensitivity and propensity to apoptosis. The discovery of such genes is important because it could provide a means to enhance classification systems based on relapse hazard and to identify signaling pathways that could be productively targeted with novel therapies. Genome-wide expression profiling technology promises to significantly facilitate these discoveries, as shown by studies determining gene expression changes in response to methotrexate and/or mercaptopurine,^1,2  and by correlative studies based on drug sensitivity findings in vitro.^3-5 

Minimal residual disease (MRD) assays provide a direct measure of treatment response in vivo that is likely to depend not only on the resistance of leukemic cells to individual drugs, but on other factors as well, including drug interactions and pharmacokinetic/pharmacogenetic variables.^6-8  Such assays have revealed considerable heterogeneity in the response of childhood ALL patients to remission induction therapy, which was not appreciated from conventional microscopic analyses.^9-14  While some patients can show profound reductions in their leukemia cell counts (to less than one leukemic cell among 10 000 normal bone marrow cells) after only 2 weeks of remission induction chemotherapy,^13,15  others require additional remission induction chemotherapy to achieve a similar level of leukemia cytoreduction, or retain detectable MRD beyond the completion of remission induction and consolidation treatment.^9-13  Thus, treatment response measured by MRD assays has consistently been the most powerful prognostic indicator in childhood ALL.^7,16 

In this study, we sought to identify genes whose expression is closely associated with the in vivo response to multiagent chemotherapy. We therefore compared the gene expression profiles of ALL cells obtained at diagnosis from 189 children with MRD findings on days 19 and 46 of remission induction chemotherapy. Seventeen genes whose expression was specifically associated with MRD at both time points were identified, including CASP8AP2 (caspase 8–associated protein 2), which encodes a key mediator of apoptosis and glucocorticoid signaling,^17-20  and whose expression in this study was inversely related to persistent MRD during remission induction therapy. In a separate cohort of 99 children with ALL, CASP8AP2 expression was a strong and independent predictor of treatment outcome.

Bone marrow samples were collected at diagnosis from 288 children with ALL enrolled in St Jude Total Therapy Studies XIII, XIV, or XV. Samples were also collected during remission induction chemotherapy from 189 patients: MRD was studied on day 19 in 187 and on day 46 in 188 of these patients. At diagnosis, the immunophenotypic and karyotypic features of the leukemic cells were determined according to standard techniques.^21,22  The presence of BCR-ABL, E2A-PBX1, and TEL-AML1 fusions or MLL gene rearrangements was detected by reverse transcriptase–polymerase chain reaction (RT-PCR).²³  Among the 288 ALL cases studied (189 to identify genes associated with MRD and 99 to test the clinical significance of the identified genes), 47 were classified as T-lineage ALL and 241 as B-lineage ALL. The latter included 16 cases with BCR-ABL, 22 with E2A-PBX1, 18 with MLL rearrangements, 57 with TEL-AML1, 52 with hyperdiploidy (> 50 chromosomes), and 76 with other features.

Initial treatment consisted of methotrexate alone followed 4 days later by 6 weeks of remission induction therapy with prednisone, vincristine, daunorubicin, asparaginase, and etoposide plus cytarabine.^24-26  Once they attained a complete clinical remission, all patients received 2 weeks of consolidation therapy with high-dose methotrexate and mercaptopurine, followed by risk-directed continuation therapy. The studies were approved by the St Jude institutional review board, with informed consent obtained from the parents or guardians of each child.

Gene expression profiling studies were performed as previously described.^27,28  Briefly, bone marrow mononuclear cells obtained at diagnosis were enriched with a density gradient, washed twice, and cryopreserved. We isolated total RNA from bone marrow mononuclear cells using the Trizol reagent (Invitrogen, Carlsbad, CA). After generating cDNA, we prepared biotin-labeled cRNA hybridization solutions according to the protocols of Affymetrix (Santa Clara, CA). The solutions were hybridized to HG-U133A oligonucleotide microarrays (Affymetrix). After staining with phycoerythrin-conjugated streptavidin, the arrays were read with a laser confocal scanner (Agilent, Palo Alto, CA). Signal values were computed from the image files using Affymetrix GeneChip Operating Software. Signal intensities were normalized to a standard target value of 500. Detection calls (present, marginal, or absent) were determined by default parameters. Intensity values for a total of 22 283 probe sets on the U133A microarray were obtained.

Studies of MRD were performed by flow cytometry as previously described.^9,12-14,29  Bone marrow mononuclear cells were labeled with various combinations of monoclonal antibodies conjugated to fluorescein isothiocyanate, phycoerythrin, peridinin chlorophyll protein, and allophycocyanin. For each case, optimal marker combinations specific for the leukemic clone were selected by labeling bone marrow mononuclear cells at diagnosis with antibody combinations previously shown to distinguish leukemic from normal cells; the combinations were then applied during clinical remission. Cell staining was analyzed using a dual laser FACSCalibur flow cytometer with Cell Quest software (Becton Dickinson, San Jose, CA). The flow cytometry protocol used for MRD detection has been described in detail elsewhere.^12,13,30  In all samples, the data represent all mononuclear cells in each test tube (> 1 × 10⁵). These detection methods allow the identification of one leukemic cell among 10 000 or more normal bone marrow cells,³⁰  produce results that are highly concordant with those obtained by PCR analysis of antigen receptor genes,³¹  and are currently applicable to more than 95% of patients.

For detection of CASP8AP2 expression by RT-PCR, we obtained cDNA by reverse transcription of RNA of established ALL cell lines and primary ALL samples with random hexamers and amplified it using the primers 5′-GAAGGTAATCATCCTGCATT-3′ (sense) and 5′-GAGCTTCATTAGCTGCTGGA-3′ (antisense). PCR amplification was performed for 30 cycles (95°C for 45 seconds, 56°C for 45 seconds, and 72°C for 60 seconds). The PCR products were separated on a 2% agarose gel, and the DNA was visualized by ethidium bromide staining.

For flow cytometry, we used a specific rabbit polyclonal antibody anti-CASP8AP2 from ProSCI (Poway, CA). Cells were permeabilized with 8E, a reagent developed in our laboratory, and incubated with the antibody or with nonreactive rabbit immunoglobulin (as a control). After 2 washes in phosphate-buffered saline containing 0.2% serum albumin and 0.2% sodium azide, cells were incubated with a goat anti–rabbit Ig antibody conjugated to phycoerythrin (Jackson Laboratories, West Grove, PA). Tests in which the anti-CASP8AP2 antibody was preincubated with the immunizing peptide (ProSCI) were performed to ensure specificity of staining. Cells were analyzed with a FACSCalibur flow cytometer.

In vitro cultures of leukemic lymphoblasts on bone marrow mesenchymal cells were performed as previously described.³²  Briefly, the leukemic cells were resuspended in AIM-V medium (Gibco, Grand Island, NY) at a final concentration of 1.5 × 10⁶/mL. Of the suspension, 200 μL was then placed in a 96-well tissue culture plate or seeded onto confluent bone marrow mesenchymal cell layers. In all samples, cell viability exceeded 80% by trypan-blue dye exclusion. All cell cultures were performed at 37°C under 5% CO₂.

At the termination of cultures, cells were harvested by vigorous pipetting. B-lineage ALL samples were incubated with CD19 monoclonal antibody conjugated to FITC; T-ALL samples were incubated with FITC-conjugated CD7. All antibodies were from Becton Dickinson. Samples were analyzed with a FACScan flow cytometer with Cell Quest software, as previously described.^32,33  After 7 days of culture, the percentage of cell recovery was calculated as follows: (no. of CD19⁺ or CD7⁺ lymphoblasts after 7 days of culture) × 100/(no. of CD19⁺ or CD7⁺ lymphoblasts after 1 hour of culture). Results are reported as the means of at least duplicate experiments. Leukemic cells were counted without knowledge of the patient's level of CASP8AP2 expression.

Individual genes associated with MRD adjusted for lineage and genetic subtypes were identified with an analysis-of-variance (ANOVA) model; t test analysis was used to identify individual genes associated with MRD without adjustment for other factors. Statistical significance and false discovery rate (FDR) estimates in this part of the analysis were determined using the profile information criterion and the FDR estimator, as described.³⁴ 

Correlations between gene expression, clinicobiologic features of ALL, and MRD status were performed with the Kruskal-Wallis test for multiple samples or Wilcoxon–Mann-Whitney tests for 2 samples. Event-free survival and cumulative incidence of relapse (where death in remission and second malignancy were treated as competing risks) were analyzed by a proportional hazard regression model, as well as by log rank and Gray test, respectively. Cumulative incidence of relapse in relation to competing known prognostic factors of childhood was analyzed with a Fine and Gray model.³⁵  All analyses were performed with the R (The R Project, http://www.r-project.org/), SAS (SAS Institute, Cary, NC), and S-plus (Insightful, Seattle, WA) programs.

We compared gene expression data for the diagnostic bone marrow samples of 189 patients with results of MRD measurements obtained in 187 of these patients on day 19, and in 188 on day 46 of remission induction therapy. MRD positivity was defined as 0.01% or more cells expressing the leukemia-associated immunophenotype identified at diagnosis among bone marrow mononuclear cells. By this criterion, 109 (58.3%) of 187 patients were MRD positive on day 19, and 43 (22.9%) of 188 on day 46, in agreement with our previous findings in different patient cohorts.^12,13  After eliminating the possible confounding influence of genetic subtypes known to be associated with treatment response (BCR-ABL, MLL gene rearrangements, TEL-AML1, and hyperdiploidy > 50 chromosomes), and applying a P value threshold of .001 by t test (estimated FDR, 17.1%),³⁴  we identified 105 probe sets whose expression was associated with MRD on day 46. The probe sets corresponded to 85 named genes; 53 were overexpressed in diagnostic samples from patients with MRD on day 46, and 32 were underexpressed (Table 1). Expression of 17 of the 85 genes was also significantly (P < .02) related to the presence or absence of MRD at day 19: 10 such genes were overexpressed in diagnostic samples of patients who had MRD at days 19 and 46, whereas 7 were underexpressed (Table 1).

A review of the reported functions of the 17 genes associated with MRD at both time points led us to select CASP8AP2 (caspase 8–associated protein 2), also known as FLASH (FLICE-associated huge protein), for further study. This gene encodes a protein that interacts with caspase 8, a key mediator of apoptosis,^17,18  and has been shown to be a determinant of glucocorticoid signaling as well.^19,20  The reported function of CASP8AP2 together with its underexpression in leukemic cells from patients with a poor initial early response to remission induction therapy (as demonstrated by the presence of MRD on days 19 and 46) provided a compelling rationale for assessing its clinical significance in childhood ALL.

CASP8AP2 transcripts were detectable in ALL cells by RT-PCR. In all 9 ALL cell lines (B-lineage: 380, OP-1, NALM6, RS4;11, REH, and 697; T-lineage: CEM-C1, CEM-C7, and Jurkat) and in 10 primary ALL samples (not included in the gene expression array study), a transcript of approximately 700 kb corresponding to CASP8AP2 was clearly detectable (Figure 1A). To demonstrate that the CASP8AP2 transcript was translated into the encoded protein in ALL cells, we used a rabbit polyclonal antibody anti-CASP8AP2. Labeling of the ALL cell line REH with this antibody after cell membrane permeabilization stained virtually all cells; staining was prevented by preincubating the antibody with the immunizing peptide corresponding to amino acids 1966 to 1981 of the human CASP8AP2 protein (not shown). In general, levels of protein expression correlated with those of the CASP8AP2 transcripts, as shown by staining of 12 primary ALL samples selected among those studied by microarray (r = 0.839, P < .001 by Spearman correlation test; Figure 1B-C). We noted, however, that differences in levels of transcript expression were generally higher than those measured by flow cytometry. It is unclear whether this was due to the low affinity of the polyclonal antibody used or to posttranslational regulatory mechanisms. Such a discrepancy has been reported for other molecules expressed in ALL cells.³⁶ 

To assess the relationship of CASP8AP2 expression at diagnosis to the clinical and biologic features of ALL in 288 children with ALL, we divided the patients into 3 groups of 96 patients each according to level of CASP8AP2 expression measured by gene array. Among clinical features, low levels of expression were significantly more prevalent among patients younger than 1 year of age, a known adverse prognostic feature (P = .016),⁶  whereas patients with hyperdiploidy (> 50 chromosomes), a feature associated with favorable outcome,⁶  generally had higher levels of CASP8AP2 expression (P < .001) (Table 2). Using CASP8AP2 expression as a continuous variable, we found that low levels of expression were again significantly associated with age younger than 1 year (P = .012) and less favorable genetic subtypes (P < .001). No significant associations between CASP8AP2 and other presenting clinicobiologic features were apparent (Table 2). Levels of CASP8AP2 were not different among patients enrolled in the 3 sequential treatment protocols (Total XIII, XIV, and XV; P = .16 by Kruskal-Wallis test).

As expected, low CASP8AP2 expression was significantly associated with slow early treatment response as defined by the presence of MRD at day 19 (P = .006) and at day 46 (P < .001; Wilcoxon–Mann-Whitney test; data not shown). Figure 2 illustrates the prevalence of MRD on days 19 and 46 in 2 groups of patients with the lowest and highest levels of CASP8AP2 expression, respectively. On day 19, MRD was detected in 36 of the 50 patients with low expression of this gene versus 20 of the 50 with high expression; on day 46, these prevalence rates were 22 of 50 versus 5 of 50, respectively. Among the positive samples, higher levels of MRD were found primarily in cases with low CASP8AP2 expression. For example, 7 patients with low CASP8AP2 expression, compared with 1 with high expression, had an MRD level higher than 1% at day 46, a feature typically associated with a dismal outcome and an indication for hematopoietic stem cell transplantation.¹²  Indeed, all but 1 of the 7 patients with this finding and low CASP8AP2 expression have relapsed or shown persistent MRD after day 46 of remission induction therapy.

To test the suggestive relationship of CASP8AP2 level with clinical outcome in Figure 2, we focused our analysis on a separate group of 99 patients enrolled in St Jude Total Therapy Study XIII.²⁴  As a continuous variable, CASP8AP2 expression measured by gene array at diagnosis was significantly associated with event-free survival (P = .023) and with cumulative incidence of ALL relapse (P = .013) in a proportional hazard regression model. When the patients were divided into 3 groups of 33 each according to level of CASP8AP2 expression, those with high levels of expression had a significantly better event-free survival rate than those with intermediate or low levels (P = .011 by log rank test; Figure 3A), and a lower cumulative incidence of relapse (P = .043 by Gray test; Figure 3B). In a cumulative incidence regression model including all major presenting features associated with prognosis in childhood ALL, expression of CASP8AP2 remained a significant predictor of outcome (Table 3). Note that age older than 10 years was the only factor that increased the relapse hazard more than low CASP8AP2 expression did in this analysis. The results indicate that CASP8AP2 levels measured in diagnostic samples of leukemic blasts are a powerful predictor of treatment response in childhood ALL.

We performed similar analyses with 2 other genes included in the group of 17 genes associated with MRD: integrin α6 and tissue matrix remodeling-like gene (MXRA7; Table 1). Both genes were overexpressed in patients with a positive MRD assay at days 19 and 46. However, we found that expression of neither gene was significantly associated with ALL relapse in the independent cohort of 99 patients (not shown).

To begin to define the role of CASP8AP2 in ALL cell biology, we determined the association between CASP8AP2 expression and the capacity of leukemic cells to survive and grow in vitro. For this purpose, we used a 7-day culture assay in which the survival of leukemic cells is supported by bone marrow mesenchymal cells.³⁷  This assay is well suited to test the growth potential of leukemic cells, a feature that correlates with treatment outcome.³⁸  We previously found that cells from approximately 50% of ALL cases expand in vitro when grown on mesenchymal cell layers, whereas in the remaining cases, the leukemic cells undergo apoptosis.³² 

After dividing 24 cases of ALL studied by expression array into 2 equal groups based on the expression of CASP8AP2, we compared recovery of leukemic cells after 7 days of culture. As shown in Figure 4, the recovery of lymphoblasts was significantly lower in cases with higher CASP8AP2 expression (P = .018), in agreement with morphologic and flow cytometric evidence of apoptosis (not shown). This result together with the reported function of CASP8AP2 suggests that the lower prevalence of MRD and better outcome in patients with high CASP8AP2 expression could be related to a higher propensity of the cells to undergo apoptosis and a lower capacity for expansion.

The wealth of information generated by microarray studies provides unprecedented opportunities for identifying molecules that influence the propensity of leukemic cells to undergo apoptosis and hence their susceptibility to multiagent chemotherapy in vivo. We postulated that genes whose expression is associated with the presence of MRD during remission induction should have significant prognostic impact. After adjusting for associations with known ALL subtypes and using P values below .001 as a cutoff, we found 85 genes whose expression level was associated with MRD on day 46. Comparison of the reported functions of these genes led us to select CASP8AP2 (FLASH), a member of the apoptosis signaling complex that activates caspase 8 and facilitates Fas-induced apoptosis,¹⁷  as a candidate for further study. CASP8AP2 participates in apoptosis,^17,18  and its enforced expression in brain cells increases glucocorticoid receptor–mediated transactivation.²⁰  The proapoptotic function of CASP8AP2, together with the down-regulation of CASP8AP2 in the leukemic lymphoblasts of patients with persistent MRD during remission induction therapy, strongly suggested that this gene may be an important prognostic factor in ALL.

In an independent cohort of children with ALL, we found a striking association between low levels of CASP8AP2 expression and a high rate of leukemia relapse. In this cohort, low expression of CASP8AP2 was a very strong predictor of ALL relapse, second only to age older than 10 years. Low CASP8AP2 expression was more prevalent among patients younger than 1 year of age, a subgroup with generally poor response to therapy,³⁹  whereas high CASP8AP2 expression was more prevalent among patients with hyperdiploidy, a favorable genetic abnormality.^40-42  However, the relation between CASP8AP2 expression with age and chromosome number was not absolute, and, in a multivariate analysis, it remained a significant predictor of outcome.

Recently, Holleman et al⁴  reported correlations between expression of apoptosis-related genes and drug sensitivity in vitro in samples of childhood ALL cells. Their study focused on the prognostic significance of another apoptosis regulator, BCL2L13, but we noted that in their analysis CASP8AP2 expression was, on average, twice as high in prednisone-sensitive cases and was associated with higher sensitivity to asparaginase and daunorubicin. In our study, higher CASP8AP2 levels were related to a reduced capacity of leukemic lymphoblasts to grow in vitro, suggesting that this feature confers a general propensity to undergo apoptosis. CASP8AP2 was not detected in 2 other screenings of genes associated with early treatment response in ALL. The study by Chiaretti et al⁴³  analyzed genes associated with early treatment response in 33 adult patients with T-ALL and was performed with the HG-U95A Affymetrix GeneChip, which does not include probes for CASP8AP2. The study of Cario et al⁴⁴  analyzed gene expression in 51 patients classified as either “MRD standard risk” or “MRD high risk” according to the BFM2000 criteria but was performed on a different microarray platform.⁴⁵  Both groups of investigators noted that low expression of TTK (a gene encoding a kinase involved in cell-cycle regulation)⁴⁶  was associated with poorer treatment response. Remarkably, low expression of TTK was also associated with the presence of MRD on both days 19 and 46 in our series, although its association with MRD was weaker than that of CASP8AP2 and of the other genes listed in Table 1. Nevertheless, the consistent association of this gene with treatment response clearly merits further investigation.

Further progress in the treatment of childhood ALL will require optimization of risk assignment to avoid overtreatment and under-treatment of patients as well as the development of new antileukemic agents capable of overcoming drug resistance.^6,47  We suggest that measurements of CASP8AP2 expression could help to identify patients whose leukemic cells are highly susceptible or highly resistant to chemotherapy. Indeed, this gene is a strong candidate for inclusion in gene arrays specifically designed for leukemia diagnosis. Since patients with detectable MRD during remission induction therapy appeared to have a much higher risk of relapse if their leukemic cells had low levels of CASP8AP2, measurements of this gene's levels could also be used to augment the informative power of MRD studies.

We thank Chris Clark, Peixin Liu, and Mo Mehrpooya for technical assistance, and John Gilbert for critical review of the paper.