|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 25, No 10 (April 1), 2007: pp. 1209-1215 © 2007 American Society of Clinical Oncology. DOI: 10.1200/JCO.2006.08.4061 Activated Intrinsic Apoptosis Pathway Is a Key Related Prognostic Parameter in Acute Myeloid Leukemia
From the Departments of Hematology, Clinical Epidemiology and Biostatistics, and Clinical Pathology, VU University Medical Center; and MRC-Holland B.V., Amsterdam, the Netherlands Address reprint requests to Gerrit J. Schuurhuis, PhD, Department of Hematology, CCA Building Room 224, VU University Medical Center, De Boelelaan 1117, 1081 HV Amsterdam, the Netherlands; e-mail: gj.schuurhuis{at}vumc.nl
Purpose By parallel assessment of multiple apoptosis-related transcripts, we aimed to refine the current concept of apoptosis resistance in acute myeloid leukemia (AML) and identify the combination of genes best predicting overall survival (OS). Patients and Methods The reverse transcriptase multiplex ligation–dependent probe amplification technique was used for simultaneous quantification of 31 apoptosis-related transcripts in viable (7AAD–/AnnexinV–) blasts (CD45dim) from bone marrow aspirates of 120 newly diagnosed AML patients. By forward selection, a prognosis-predicting gene expression profile was constructed. The predictive validity of this profile was assessed by cross validation. Results High transcript levels were associated with poor OS for seven of 31 genes, three of which were proapoptotic. The average expression of all 12 antiapoptotic genes was associated with poor OS (P = .029). A similar association with poor OS was found for the average expression of all 19 proapoptotic genes (P = .009). Forward selection and cross validation revealed the antiapoptotic gene BIRC3 and the proapoptotic genes BAX-(l) and BMF to optimally predict OS. Three equally sized patient groups, constructed by ranking the cross-validated prognoses of the patients, were clearly distinct (median OS times were 8.2, 16.7, and 85.6 months). Conclusion High expression of both pro- and antiapoptotic genes predicted poor OS, which postulates a mechanism of activation of the apoptosis pathway as a whole. This mechanism, which culminates in a three-gene expression signature, allows accurate clinical outcome prediction in AML and puts efforts to target single antiapoptosis genes in a new perspective.
It has become evident that apoptosis-related molecules are useful as prognostic markers and can serve as therapeutic tools in acute myeloid leukemia (AML). Defects in the apoptosis machinery also play an important role in chemotherapy resistance in AML, resulting in primary refractoriness or high levels of minimal residual disease cells, the latter eventually leading to relapse.1 At present, the general concept describes a balance between proapoptotic and antiapoptotic family members, leading to execution of apoptosis if proapoptotic family members are in excess and apoptotic resistance if the antiapoptotic members are dominantly present.2,3 Because apoptosis is regulated by a process involving complex protein-protein interactions between the pro- and antiapoptotic family members via shared functional motifs, indexes combining pro-and antiapoptotic parameters in an inverse relationship, such as BAX/BCL-2 or BCL-XS/BCL-XL, have been suggested to further improve predictive strength.4,5 We hypothesized that determination of the expression of multiple genes involved in the apoptosis pathway would allow the construction of a more accurate prediction model for patient survival and, moreover, possibly guide therapeutic modalities such as downregulation of BCL-2 expression using antisense approach.6,7 The novel and highly sensitive reverse transcriptase multiplex ligation–dependent probe amplification (RT-MLPA) technique8 allows simultaneous monitoring of the expression of 31 apoptosis-related genes. Using this approach, we report a novel prognosis-determining concept of pathway-wide involvement of pro- and antiapoptotic genes in AML.
Patient Characteristics This study included 120 AML patients (median age, 54 years; range, 16 to 77 years) treated at the Hematology Department of the VU University Medical Center (Amsterdam, the Netherlands). The diagnoses were established according to morphology and immunophenotype, following the French-American-British (FAB) classification criteria,9 and were as follows: M0 (n = 6), M1 (n = 16), M2 (n = 22), M3 (n = 3), M4 (n = 24), M5 (n = 34), M6 (n = 6), and unknown FAB type (n = 9). Eight of 101 patients were diagnosed with AML after a myelodysplastic syndrome (MDS). Median WBC at diagnosis was 39.7 x 109 cells/L (range, 0.4 to 300 x 109 cells/L). Patients younger than 60 years old were treated according to protocols of the Dutch haemato-oncology association (Stichting Hemato-Oncologic voor Volwassenen, the Netherlands) HOVON-29 (1998 to 2000) and HOVON-42 (2001 to 2006) protocols, both combining cytarabine and an anthracycline (idarubicin in the first cycle, amsacrine in the second cycle) for induction. Consolidation chemotherapy consisted of one cycle of etoposide and mitoxantrone or, for eligible patients, a myeloablative regimen of cyclophosphamide and busulfan, followed by autologous (n = 24) or allogeneic (n = 7) stem-cell transplantation. Patients  60 years old were treated either according to the HOVON-3210 (1996 to 1999) or HOVON-43 (2000 to 2006). HOVON-32 consisted of cytarabine alone (arm A) or the combination of cytarabine and fludarabine (arm B) in two induction cycles, supplemented with granulocyte colony-stimulating factor (filgrastim, 5 µg/kg, day 0 until neutrophil recovery > 0.5 x 109/L) in both arms. HOVON-43 protocol combined cytarabine and daunorubicin in the first cycle followed by cytarabine as a single agent in the second cycle. Three patients with acute promyelocytic leukemia were additionally treated with all-trans-retinoic acid (45 mg/m2/d). A more detailed description of the protocols can be found at http://www.hovon.nl. Early/toxic deaths were defined as all deaths occurring within 7 days after completion of the first induction cycle or death during therapy-induced bone marrow hypoplasia. Complete remission (CR) or resistance to induction treatment was assessed by bone marrow evaluation at the day of hematologic recovery according to standard criteria. Informed consent was obtained from all patients according to institutional guidelines. CR was defined as a bone marrow with normal hematopoiesis of all cell types, less than 5% blast cells, and a peripheral-blood count of at least 1,500/µL (1.5 x 109/L) neutrophils and 100,000/µL (100 x 109/L) platelets. Relapse was defined as reinfiltration of the bone marrow by 5% leukemic blasts or a proven infiltration of leukemic blasts at any other site. Overall survival (OS) was defined as the time period from inclusion to death; relapse-free survival (RFS) was defined as the time period from achievement of CR to relapse.
Sample Characteristics
RT-MLPA
Statistical Analyses
The effects of gene expression levels alone or in combination with clinical/pathologic factors on achievement of CR (data available for 94 patients) were estimated using logistic regression, whereas the effects on OS and RFS (data available for 117 and 83 patients, respectively) were estimated using Cox regression. The total number of events for OS (death) and RFS (relapse) was 69 and 49, respectively. Adjustment for the clinical/pathologic factors, including WBC at diagnosis (expressed as logarithm), CD34 percentage on CD45dim blasts, treatment protocol (< 60 or A forward selection approach was applied to obtain a combination of genes that best predicts OS. The optimal number of genes in the prediction equation was determined by maximizing the cross-validated fit.12 The cross-validation method mimics prediction by leaving out one patient at a time and constructing a gene prediction equation with a certain number of genes on the basis of the remaining patients. This prediction equation and the gene expressions of the left out patient are used to obtain a cross-validated prognosis for the left out patient. After having left out all patients one at a time, the individual prognoses can be summed into a cross-validated model fit. The cross-validation procedure was repeated for models with one, two, three, and so on, genes, and the model with the highest cross-validated fit was selected. After estimating the prediction equation, the patients were ranked according to their cross-validated prognoses and classified in three equally sized groups. Kaplan-Meier survival curves were constructed for these three groups. Differences in the distribution of clinical parameters across the three groups were assessed using Fisher's exact test and the two-sided t test or Mann-Whitney U test. Comparison of gene expression levels in AML samples with normal controls was performed using the two-sided t test. All univariate and multivariate analyses as well as the forward selection procedure have also been executed while censoring for the time of autologous or allogeneic stem-cell transplantation. For all analyses, the significance level was set at P = .05. The reported P values are two sided. Calculations were performed using SPSS 11.0.1 (SPSS Inc, Chicago, IL) and Matlab 7.1 (Mathworks, Natick, MA) software.
Associations Between Both Expression of Single Genes and Clinical/Pathologic Factors and CR Rate, OS, and RFS Univariate Cox regression analyses yielded seven genes with predictive value for OS; in all cases, high expression was associated with poor OS. Similarly, high expression of seven other genes was associated with poor RFS. For six genes with predictive value for CR, in all cases, high expression was associated with a low CR rate. Most remarkably, the genes with prognostic value for OS included both proapoptotic (n = 4) and antiapoptotic (n = 3) genes. For achievement of CR, three of six genes were proapoptotic, and for RFS, six of seven genes were proapoptotic (Fig 1). Four genes were found to be predictive for both OS and achievement of CR. For RFS and OS, such overlap in prediction was observed for only one gene, whereas for CR and RFS, no overlap was found (compare asterisks in Fig 1). Censoring for time of transplantation resulted in identification of the same individual genes with a significant effect on OS, now also including BMF (hazard ratio [HR], 1.574; 95% CI, 1.021 to 2.427; P = .04).
Assessment of the impact of clinical/pathologic factors on achievement of CR did not yield significant effects. For OS, the presence of an FLT3-ITD (HR, 2.11; 95% CI, 1.14 to 3.90; P = .018) and achievement of CR (HR, 0.037; 95% CI, 0.01 to 0.13; P < .0001) had a significant effect, whereas the effect of age-related treatment protocol was borderline significant (HR, 1.49; 95% CI, 0.93 to 2.40; P = .09). Cytogenetics as categoric predictor did not yield a significant effect on OS (P = .52), most likely because of the relatively low numbers of patients with either good or adverse cytogenetics. Neither a known history of MDS (P = .44) nor CD34 percentage as a continuous variable (P = .50) or as a dichotomous variable using cutoff values of 10% (P = .19) and 20% (P = .42) had impact on OS. None of the clinical/pathologic factors showed effects on RFS. After adjustment for the two relevant diagnostic factors (ie, age category and FLT3 status), we found that the same individual genes had predictive impact on CR, OS, and RFS.
Gene Expression Characteristics Mean gene expression levels were comparable between normal controls and AML blasts; however, the variance was much higher in AML blasts. As an example, the mean BAX-(l) expression in controls (–3.81) was comparable with that in AML patients (–3.79; P = .74). However, gene expression variance was significantly lower in controls compared with AML patients (0.07 v 0.18, respectively; P = .02).
Expression of Multiple Genes and Prediction of OS BIRC3, BMF, and BAX-(l) were selected by forward selection for the prediction profile (three-gene signature; a linear combination of genes that is estimated by Cox regression) best predictive for OS. The OS HRs of the three genes deducted from the regression equation profile were 1.48 for BIRC3 (95% CI, 1.16 to 1.89; P = .002), 1.35 for BMF (95% CI, 1.08 to 1.67; P = .008), and 1.45 for BAX-(l) (95% CI, 1.13 to 1.87; P = .004). After adjustment for age category and FLT3 status, which are the baseline parameters at diagnosis with a significant effect on OS, the effect of BIRC3, BMF, and BAX-(l) remained predictive (P = .0002, P = .0005, and P = .014, respectively). In multivariate analysis that included age category and FLT3 status, the three-gene signature also remained predictive for OS (P < .0001). According to their cross-validated OS prognosis, patients (n = 117) were ordered and split into three equally sized groups (Fig 2A) of good prognosis patients (group 1), intermediate prognosis patients (group 2), and poor prognosis patients (group 3). The median OS time was 8.2 months for poor prognosis patients, 16.7 months for intermediate prognosis patients, and 85.6 months for good prognosis patients. Characteristics of the three groups are listed in Table 1. Only 21 (53%) of 39 poor prognosis patients achieved CR, whereas 89% of the patients with intermediate or good prognosis achieved CR (34 of 38 patients and 33 of 37 patients, respectively). A number of patients experienced early/toxic death (ie, death before remission evaluation). Although the factors that underlie early death represent treatment resistance characteristics that are unknown, the three-gene signature stratified 13 of the 21 patients who experienced early death to the poor prognosis patient subgroup.
The gene selection procedure was repeated in a model in which age category was included as a covariate. This resulted in the selection of the same genes (data not shown), which further demonstrates that the selected genes predict OS independent of treatment protocol, albeit with borderline significance (< 60 years, P = .08; and 60 years, P = .06) likely because of the relatively low patient numbers (71 patients < 60 years and 46 patients 60 years, Table 1). However, the distribution of patients over the three prognostic groups was different for patients younger than 60 years old and patients 60 years old, with the latter more often classified as poor prognosis patients (Table 1). Consistent with their poor prognosis, five of eight patients with a history of pre-existent MDS were stratified in the poor prognosis group. Patients with favorable cytogenetics were mainly present in groups 1 and 2, whereas the patients with adverse cytogenetics were in groups 2 and 3. If only patients with intermediate cytogenetics were selected, the BAX-(l)/BIRC3/BMF three-gene signature divided these patients almost evenly over the three prognostic patient subgroups, as depicted in Table 1. Moreover, within this intermediate cytogenetic group, a good prognosis patient subgroup could be identified using Kaplan-Meier analysis (Fig 2B).
In this article, we showed high expression of both the anti- and proapoptotic genes in AML blasts at diagnosis to have adverse prognostic impact. For the proapoptotic transcripts, this inverse correlation between expression and survival is quite unlike the current concept. This concept outlines that high expression of proapoptotic members leads to execution of apoptosis, resulting in improved patient survival, with the opposite for antiapoptotic members. Retrospectively, one can deduct similar observations from other studies, be it in a scattered way, because, in all studies, only a limited number of apoptosis-related genes have been studied. For example, in adult AML, although high BAX transcript level was a good prognostic indicator in one study,13 another study reported the opposite for BAX and BAD.14 In diffuse large B-cell lymphoma15 as well as in solid tumors like ovarian16 and pancreatic cancers,17 the same nonclassical relationship between high BAX transcript level and poor survival was observed. Studies that strongly influenced the concept of an apoptotic balance have been performed on protein level. Although BAX protein expression was reported to have favorable prognostic impact in one study,13 in another study in childhood acute lymphoblastic leukemia, it was associated with increased risk of relapse.18,19 In 35 patients in our previous study, who are included in the present study, positively directed, although nonsignificant, associations were observed between gene and protein expression for BAX, BCL-2, MCL-1, and BCL-XL (data not shown). From this, it can be argued that the predictive strength of the balance between expressions of pro- and antiapoptotic proteins (eg, BAX to BCL-2 ratio) may merely be ascribed to the antiapoptotic member. Retrospectively, such may indeed be deduced from the studies on this.4 Our observation offers an explanation for these scattered discrepancies by showing that pro- and antiapoptotic transcript levels describe complementary features of the same mechanism of altered pathway-related gene expression. This fits in a recently defined concept,20 the so-called primed to death status, which is associated with oncogenic addiction; upregulation of antiapoptotic proteins leads to increased binding of activator proapoptotic proteins, which in turn evokes upregulation of the bond-breaking sensitizer proapoptotic (BH3-only) proteins, finally resulting in apoptosis. This concept of oncogenic addiction could explain the tight association between antiapoptotic and proapoptotic expression levels of genes. The higher gene expression variance in AML blasts compared with the control CD34+ cells is well in line with clinical reality; a substantial portion of patients may respond to therapy (low gene expression in AML blasts) but may experience considerable toxicity (low gene expression in normal CD34+ cells). Another portion of the patients may have a poor response (high AML gene expression), again, all or not paralleled with toxicity. As a spin-off from our key observation, by applying forward selection, an objectively defined prediction profile emerged with independent prognostic impact and that consisted of one antiapoptotic and two proapoptotic genes (ie, BIRC3, BAX-(l), and BMF), thereby perfectly illustrating the main concept of this article. This signature allowed the identification of three distinct prognostic patient subgroups, importantly even within an otherwise homogeneous intermediate cytogenetic patient group. These unique observations have been enabled by the RT-MLPA technique, which allowed simultaneous monitoring of as many as 12 antiapoptotic and 19 proapoptotic genes, thereby revealing their communal characteristic. Further studies will be required to demonstrate the predictive ability in a prospective patient cohort. The novel transcriptional characteristic will change our notion of the role of apoptosis in cancer treatment and should be considered in the design of apoptosis-inducing targeted strategies.
Although all authors completed the disclosure declaration, the following authors or their immediate family members indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors. Employment: Jan P. Schouten, MRC-Holland B.V. Leadership: Jan P. Schouten, MRC-Holland B.V. Consultant: N/A Stock: Jan P. Schouten, MRC-Holland B.V. Honoraria: N/A Research Funds: N/A Testimony: N/A Other: N/A
Conception and design: Corine J. Hess, Johannes Berkhof, Gert J. Ossenkoppele, Quinten Waisfisz, Joost J. Oudejans, Gerrit J. Schuurhuis Financial support: Gert J. Ossenkoppele Administrative support: Fedor Denkers Provision of study materials or patients: Gert J. Ossenkoppele, Gerrit J. Schuurhuis Collection and assembly of data: Corine J. Hess, Fedor Denkers, Jan P. Schouten, Quinten Waisfisz, Gerrit J. Schuurhuis Data analysis and interpretation: Corine J. Hess, Johannes Berkhof, Quinten Waisfisz, Joost J. Oudejans, Gerrit J. Schuurhuis Manuscript writing: Corine J. Hess, Johannes Berkhof, Fedor Denkers, Gert J. Ossenkoppele, Jan P. Schouten, Quinten Waisfisz, Gerrit J. Schuurhuis Final approval of manuscript: Corine J. Hess, Johannes Berkhof, Fedor Denkers, Gert J. Ossenkoppele, Jan P. Schouten, Quinten Waisfisz, Joost J. Oudejans, Gerrit J. Schuurhuis
We thank A.H. Westra and J.W. Oberink for blast purification by means of fluorescence-activated cell sorting.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
1. van Stijn A, Feller N, Kok A, et al: Minimal residual disease in acute myeloid leukemia is predicted by an apoptosis-resistant protein profile at diagnosis. Clin Cancer Res 11:2540-2546, 2005 2. Kroemer G: The proto-oncogene Bcl-2 and its role in regulating apoptosis. Nat Med 3:614-620, 1997[CrossRef][Medline] 3. Adams JM, Cory S: Life-or-death decisions by the Bcl-2 protein family. Trends Biochem Sci 26:61-66, 2001[CrossRef][Medline] 4. Del Poeta G, Venditti A, Del Principe MI, et al: Amount of spontaneous apoptosis detected by Bax/Bcl-2 ratio predicts outcome in acute myeloid leukemia (AML). Blood 101:2125-2131, 2003 5. Deng G, Lane C, Kornblau S, et al: Ratio of bcl-xshort to bcl-xlong is different in good- and poor-prognosis subsets of acute myeloid leukemia. Mol Med 4:158-164, 1998[CrossRef][Medline] 6. Marcucci G, Byrd JC, Dai G, et al: Phase 1 and pharmacodynamic studies of G3139, a Bcl-2 antisense oligonucleotide, in combination with chemotherapy in refractory or relapsed acute leukemia. Blood 101:425-432, 2003 7. Moore J, Seiter K, Kolitz J, et al: A phase II study of Bcl-2 antisense (oblimersen sodium) combined with gemtuzumab ozogamicin in older patients with acute myeloid leukemia in first relapse. Leuk Res 30:777-783, 2006[CrossRef][Medline] 8. Hess CJ, Denkers F, Ossenkoppele GJ, et al: Gene expression profiling of minimal residual disease in acute myeloid leukaemia by novel multiplex-PCR-based method. Leukemia 18:1981-1988, 2004[CrossRef][Medline] 9. Bennett JM, Catovsky D, Daniel MT, et al: Proposals for the classification of the acute leukaemias: French-American-British (FAB) co-operative group. Br J Haematol 33:451-458, 1976[Medline] 10. Ossenkoppele GJ, Graveland WJ, Sonneveld P, et al: The value of fludarabine in addition to ARA-C and G-CSF in the treatment of patients with high-risk myelodysplastic syndromes and AML in elderly patients. Blood 103:2908-2913, 2004 11. Grimwade D, Walker H, Oliver F, et al: The importance of diagnostic cytogenetics on outcome in AML: Analysis of 1,612 patients entered into the MRC AML 10 trial—The Medical Research Council Adult and Children's Leukaemia Working Parties. Blood 92:2322-2333, 1998 12. van Houwelingen HC, Bruinsma T, Hart AA, et al: Cross-validated Cox regression on microarray gene expression data. Stat Med 25:3201-3216, 2006[CrossRef][Medline] 13. Ong YL, McMullin MF, Bailie KE, et al: High bax expression is a good prognostic indicator in acute myeloid leukaemia. Br J Haematol 111:182-189, 2000[CrossRef][Medline] 14. Kohler T, Schill C, Deininger MW, et al: High Bad and Bax mRNA expression correlate with negative outcome in acute myeloid leukemia (AML). Leukemia 16:22-29, 2002[CrossRef][Medline] 15. Bairey O, Zimra Y, Shaklai M, et al: Bcl-2, Bcl-X, Bax, and Bak expression in short- and long-lived patients with diffuse large B-cell lymphomas. Clin Cancer Res 5:2860-2866, 1999 16. Marx D, Binder C, Meden H, et al: Differential expression of apoptosis associated genes bax and bcl-2 in ovarian cancer. Anticancer Res 17:2233-2240, 1997[Medline] 17. Friess H, Lu Z, Graber HU, et al: Bax, but not bcl-2, influences the prognosis of human pancreatic cancer. Gut 43:414-421, 1998 18. Srinivas G, Kusumakumary P, Nair MK, et al: Mutant p53 protein, Bcl-2/Bax ratios and apoptosis in paediatric acute lymphoblastic leukaemia. J Cancer Res Clin Oncol 126:62-67, 2000[CrossRef][Medline] 19. Hogarth LA, Hall AG: Increased BAX expression is associated with an increased risk of relapse in childhood acute lymphocytic leukemia. Blood 93:2671-2678, 1999 20. Certo M, Del Gaizo Moore V, Nishino M, et al: Mitochondria primed by death signals determine cellular addiction to antiapoptotic BCL-2 family members. Cancer Cell 9:351-365, 2006[CrossRef][Medline] Submitted July 28, 2006; accepted December 26, 2006.
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2007 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
