- © 2012 by American Society of Clinical Oncology
Biallelic ATM Inactivation Significantly Reduces Survival in Patients Treated on the United Kingdom Leukemia Research Fund Chronic Lymphocytic Leukemia 4 Trial
- Anna Skowronska,
- Anton Parker,
- Gulshanara Ahmed,
- Ceri Oldreive,
- Zadie Davis,
- Sue Richards,
- Martin Dyer,
- Estella Matutes,
- David Gonzalez,
- A. Malcolm R. Taylor,
- Paul Moss,
- Peter Thomas,
- David Oscier and
- Tatjana Stankovic⇓
- Anna Skowronska, Gulshanara Ahmed, Ceri Oldreive, A.M.R. Taylor, Paul Moss, and Tatjana Stankovic, University of Birmingham, Birmingham; Anton Parker, Zadie Davis, and David Oscier, Royal Bournemouth Hospital and Bournemouth University; Peter Thomas, Bournemouth University, Bournemouth; Sue Richards, University of Oxford, Oxford; Martin Dyer, Leicester University, Leicester; and Estella Matutes and David Gonzalez, Institute for Cancer Research and the Royal Marsden National Health Service, London, United Kingdom.
- Corresponding author: Tatjana Stankovic, MD, PhD, School of Cancer Sciences, University of Birmingham, Vincent Dr, Edgbaston, Birmingham B15 2TT, United Kingdom; e-mail: t.stankovic{at}bham.ac.uk.
-
Presented at the 53rd Annual Meeting and Exposition of the American Society of Hematology, San Diego, CA,December 10-13, 2011.
-
A.S. and A.P. contributed equally to this study.
Abstract
Purpose The prognostic significance of ATM mutations in chronic lymphocytic leukemia (CLL) is unclear. We assessed their impact in the context of a prospective randomized trial.
Patients and Methods We analyzed the ATM gene in 224 patients treated on the Leukemia Research Fund Chronic Lymphocytic Leukemia 4 (LRF-CLL4) trial with chlorambucil or fludarabine with and without cyclophosphamide. ATM status was analyzed by denaturing high-performance liquid chromatography and was related to treatment response, survival, and the impact of TP53 alterations for the same patient cohort.
Results We identified 36 ATM mutations in 33 tumors, 16 with and 17 without 11q deletion. Mutations were associated with advanced disease stage and involvement of multiple lymphoid sites. Patients with both ATM mutation and 11q deletion showed significantly reduced progression-free survival (median, 7.4 months) compared with those with ATM wild type (28.6 months), 11q deletion alone (17.1 months), or ATM mutation alone (30.8 months), but survival was similar to that in patients with monoallelic (6.7 months) or biallelic (3.4 months) TP53 alterations. This effect was independent of treatment, immunoglobulin heavy chain variable gene (IGHV) status, age, sex, or disease stage. Overall survival for patients with biallelic ATM alterations was also significantly reduced compared with those with ATM wild type or ATM mutation alone (median, 42.2 v 85.5 v 77.6 months, respectively).
Conclusion The combination of 11q deletion and ATM mutation in CLL is associated with significantly shorter progression-free and overall survival following first-line treatment with alkylating agents and purine analogs. Assessment of ATM mutation status in patients with 11q deletion may influence the choice of subsequent therapy.
INTRODUCTION
The clinical course of chronic lymphocytic leukemia (CLL) is extremely variable, ranging from an indolent disease to a rapidly progressive leukemia that is resistant to therapy with standard DNA-damaging agents. Several features of CLL tumor cells influence clinical outcome, including the functional integrity of the DNA damage response (DDR) genes ATM and TP53.1–8
ATM and p53 function may be impaired either through physical deletion of the genes on chromosome 11q and 17p, respectively, or the development of inactivating mutations. TP53 gene alterations are observed in 3% to 8% patients at diagnosis or during first-line treatment and in up to 30% of patients with refractory CLL.5 Several recent studies6–8 confirmed that TP53 gene alterations represent the single most important parameter of CLL chemoresistance.
ATM is a protein kinase that, following induction of DNA double-strand breaks, synchronizes cellular responses that include DNA repair, activation of cell cycle checkpoints, and induction of apoptosis.9 The activation of p53 by ATM leads to transcriptional activation of proapoptotic genes and elimination of cells with excessive DNA damage. The inherited inactivation of ATM causes the ataxia telangiectasia (AT) syndrome characterized by frequent lymphoid tumors.9
The ATM gene is located at 11q23 within the minimal region that is deleted in 20% to 30% of patients with CLL. Thirty-six percent of CLL tumors with 11q deletion have mutations in the remaining ATM allele, and a smaller proportion of tumors exhibit the presence of ATM mutations in the absence of 11q deletion.1,2 The loss of both ATM alleles in CLL tumors leads to loss of ATM function.2
We have previously shown that ATM mutations are associated with shorter overall survival (OS) and treatment-free survival in an unselected CLL cohort, although many prospective and retrospective studies1–4 showed inferior OS and progression-free survival (PFS) in patients with an 11q deletion who have been treated with alkylating agents and/or purine analogs.
Evidence that patients with biallelic ATM inactivation may have an inferior outcome compared with those with monoallelic ATM loss or mutation comes from our retrospective study of 72 patients with CLL who had an 11q deletion and who were treated with alkylating agents and/or fludarabine.2 Patients with an ATM mutation had a significantly shorter OS from diagnosis than those with a residual wild-type ATM allele because of impaired apoptosis in response to DNA damage. ATM mutations were either detected concurrently with the 11q deletion or arose following disease progression, suggesting that stepwise loss of ATM function may be associated with a poorer outcome and that stratification of CLL tumors with respect to ATM status may have clinical relevance.
To address this question further, we have performed a retrospective analysis of the impact of an ATM mutation at trial entry in patients receiving first-line therapy in the United Kingdom Leukemia Research Fund Chronic Lymphocytic Leukemia 4 (LRF-CLL4) trial.10,11 A recent study6 of patients on this trial revealed that both monoallelic and biallelic TP53 gene alterations confer a significantly shorter PFS and OS, irrespective of whether they received treatment with single or combined DNA purine analog/alkylating agents. Therefore, it is important to determine whether complete functional loss of the ATM gene that acts in the same DDR pathway as p53 has a similar clinical impact.
PATIENTS AND METHODS
Patients and Samples
The United Kingdom LRF-CLL4 trial randomly assigned 777 patients to first-line treatment with chlorambucil, fludarabine alone, or fludarabine plus cyclophosphamide.10 The study was performed on a subset of 224 patients enrolled onto the trial (Fig 1). All patients with an 11q deletion (from the cohort of 777 patients) for whom sufficient material was available were included in the study, with the remaining patients being randomly selected from those without 11q deletion. This approach was taken to facilitate identification of a sufficient number of patients with an ATM mutation and allow valid comparisons. Other than significant enrichment for patients with 11q deletion (P < .001) and a significant reduction in cases with trisomy 12 (P = .016), the 224-patient cohort did not differ significantly from the remaining 553 patients (Data Supplement).
CONSORT diagram. FC, fludarabine plus cyclophosphamide.
ATM Mutation Screening
The entire ATM coding region, consisting of 62 exons, was screened by using high-performance liquid chromatography.1 Sequence changes were assigned into two groups: (1) truncating mutations predicted to cause premature termination of the protein or small in-frame deletion or (2) missense mutations, either reported in patients with AT or predicted to cause an amino acid substitution in the residue located within the region encoding the functional domain of the ATM protein and/or conserved between man and mouse and not observed in 200 alleles of healthy controls.
Statistical Analysis
The recorded response was the best achieved at any time in accordance with the LRF-CLL4 trial protocol. OS was calculated from trial entry (random assignment) to death as a result of any cause, and PFS was expressed as time from random assignment to relapse needing further therapy, progression, or death as a result of any cause. For nonresponders and progressive disease, date of progression was when no response or progressive disease was recorded.
Clinical and biologic characteristics were compared between selected and nonselected cohorts and patients with and without a mutated ATM allele (Table 1) by using Fisher's exact test, except for age for which the Mann-Whitney U test was used (Data Supplement). Logistic regression analysis was applied to model the odds of multiple tissue site involvement (Data Supplement). Anemia and thrombocytopenia (involved in Binet staging) and deletion of 11q (implicated in bulky disease) were also included in the model. The odds of response to therapy was modeled by using logistic regression (for 212 patients who had response data). Given the number of nonresponders (n = 44) and potential model complexity, a careful step-by-step approach was used. ATM and 11q status were the focus of the analysis and were included in all models. By adding variables singly and in combination, a final parsimonious model emphasizing significant factors was developed. OS and PFS included censored data; therefore, comparison between ATM mutation/11q deletion groups was initially conducted by using Kaplan and Meier survival analysis and log-rank test; subsequent multivariable analysis used Cox proportional hazards regression. The modeling process was the same as described earlier for response to treatment. To address the issue of whether ATM mutation and 11q deletion have a synergistic or antagonistic effect on survival, the statistical interaction between ATM mutation and 11q was tested.
Associations Between ATM Mutation Status and Clinical Characteristics of the Patients in the LRF-CLL4 Trial
To compare the lower survival times resulting from 11q deletion (with and without ATM mutation) with those resulting from 17p and TP53 alterations, we returned to the full trial data and categorized patients in order of decreasing priority to the following five categories: TP53 mutation and 17p deletion; TP53 deletion or 17p deletion (but not both); no TP53 abnormality and 11q deletion with ATM mutation; no TP53 abnormality and 11q deletion without ATM mutation; and no TP53 abnormality and no 11q deletion. By using the third category as a reference, other categories were compared by using Kaplan and Meier analysis for survival data and Fisher's exact test for proportions. All statistical analyses were performed by using PASW statistical software, version 18.0 (www.ibm.com/software/analytics/spss) and a two-sided 5% significance level.
RESULTS
Frequency, Type, and Distribution of ATM Mutations
After excluding polymorphisms (Data Supplement), we identified 36 sequence changes in 33 (14.7%) of 224 patients that fulfilled our criteria for mutations (Table 2). Twenty mutations (55.5%) qualified as truncating and sixteen mutations (44.5%) qualified as missense. Three patients without 11q deletion displayed both a truncating and a nontruncating mutation and one truncating mutation, c.1402_1403del2, was detected in two different tumor samples.
ATM Mutations Detected in 224 CLL Tumors With and Without 11q Deletion
Consistent with previous observations, missense mutations clustered in the 3′ terminal region of the gene encoding FAT and PI3 kinase functional domains (Appendix Fig A1, online only). Of note, we observed a higher frequency of truncating mutations in this study compared with our previous reports (20 [55.5%] of 36 v seven [30%] of 23 and 11 [40.7%] of 27, respectively).1,2 These were equally distributed across the whole ATM gene and were more frequently associated with 11q deletion status: 12 (75%) of 16 mutations were truncating in patients with 11q deletion, and only eight (40%) of 20 were truncating in patients without 11q deletion (P = .049).
Associations Between ATM Mutations, 11q Deletions, and Other Clinical and Laboratory Variables
Associations between the presence of an ATM mutation and other clinical and biologic variables are presented in Table 1. As expected, ATM mutations were most strongly associated with the presence of an 11q deletion (P = .012). Overall, 17 (7.6%) of 224 patients had an ATM mutation alone, 16 (7.1%) of 224 had both ATM mutation and 11q deletion, 51 (22.8%) of 224 had 11q deletion alone, and both ATM alleles were wild type in the remaining 140 (62.5%) of 224 patients. ATM mutations were also significantly more frequent in patients with Binet disease stage B/C compared with those who had stage A (30 [17.9%] of 168 v three [5.4%] of 56; P = .027) and in patients with three or more enlarged tissue sites (26 [19%] of 137 v seven [8%] of 87; P = .032). Further analysis (Data Supplement) showed that the presence of an ATM mutation was associated with increased odds of having three or more tissue sites involved, regardless of the presence of thrombocytopenia, anemia, or an 11q deletion (odds ratio, 3.13; 95% CI, 1.26 to 7.80; P = .014).6
Interestingly, ATM mutation was not significantly associated with any other clinical or biologic variable, including expression of ZAP70, CD38, or IGHV status. Of note, ATM and TP53 mutations were not detected together in any of the 224 patients with available data for both genes. Furthermore, only one patient with an ATM mutation carried a 17p deletion.
ATM Abnormalities and Response to Therapy
The overall response rate, defined as complete response, nodular partial response, or partial response, was observed in seven (46.7%) of 15 patients with ATM mutation and 11q deletion, 112 (84.2%) of 133 patients with ATM wild type, 35 (72.9%) of 48 with 11q deletion alone, and 14 (87.5%) of 16 patients with ATM mutation alone (Data Supplement). The only significant difference in overall response rate occurred between patients carrying ATM mutation and 11q deletion and those with ATM wild type (P = .002). Notably, response rates were significantly different between treatment arms. In the group of 65 patients treated with fludarabine-cyclophosphamide, all 25 patients with ATM mutations and/or 11q deletions responded to therapy, and the only two nonresponders carried a 17p deletion and/or TP53 mutation. Among the 47 patients treated with fludarabine alone, 11 were nonresponders: of these, three had a TP53 abnormality, three had an ATM mutation (two with 11q deletion, one without), and three had 11q deletion only. Thirty-one of the 100 patients treated with chlorambucil were nonresponders, and of the 26 patients with available genotype data, only six had detectable TP53 abnormalities. Of the remaining 20 TP53 wild-type nonresponders, six had an ATM mutation and 11q deletion and 10 had 11q deletion alone (Data Supplement). Overall, in the patients treated with chlorambucil, the combination of an ATM mutation with 11q deletion was associated with significantly reduced odds of response to therapy compared with patients who had no ATM mutation or 11q deletion. In logistic regression analysis, this was independent of age, sex, disease stage, TP53 mutation, 17p deletion, and IGHV status (odds ratio, 0.09; 95% CI, 0.02 to 0.50; P = .003; Table 3).
Bivariate Regression Analysis for Odds of Response to Treatment With Chlorambucil
ATM Abnormalities and Outcome
Univariate analysis revealed that although the median PFS and OS were shorter for patients with an ATM mutation compared with those who had ATM wild type, the differences did not reach significance (P = .055 and P = .083, respectively; Appendix Fig A2, online only). However, when stratified for the presence of 11q deletion, ATM mutations were associated with a significantly shorter median PFS (7.4 v 17.1 months; P = .002) in patients with 11q deletion but not in those without 11q deletion (30.8 v 28.6 months; P = .983). Two-year PFS was also significantly reduced by the presence of an ATM mutation in patients with 11q deletion (one [6.3%] of 16 v 17 [33.3%] of 51; P = .05). In contrast, the presence of an ATM mutation had no significant effect on OS in patients who had either 11q deletion or no 11q deletion. Notably, patients with both ATM mutation and 11q deletion did show inferior OS compared with patients without 11q deletion or ATM mutation (Fig 2and Data Supplement).
Impact of ATM mutation and 11q deletion on (A) progression-free survival (PFS) and (B) overall survival (OS) in patients on Leukemia Research Fund Chronic Lymphocytic Leukemia 4 trial. Kaplan-Meier plots and proportions at risk for cases stratified by presence of 11q deletion and/or ATM mutation. P values shown are for log-rank analysis of difference in median survival for no ATM mutation and no 11q deletion v ATM mutation and for 11q deletion v both ATM mutation and 11q deletion. ATM mutations were associated with a significantly shorter median PFS (7.4 v 17.1 months; P = .002) in patients with 11q deletion but not in patients without 11q deletion (30.8 v 28.6 months; P = .983). In contrast, the presence of an ATM mutation had no significant effect on OS in patients who had 11q deletion or patients who had no 11q deletion (P = .284 and P = .433, respectively).
We subsequently performed multivariate analysis and compared ATM mutation and 11q deletion status with sex, age, Binet stage, treatment, 17p deletion, presence of TP53 mutation, and IGHV status or IGHV3-21 use (Data Supplement). We first addressed the presence of ATM mutation and 11q deletion as separate covariants and observed that 11q deletion but not ATM mutation showed significantly increased hazard ratios (HRs) for both PFS and OS (Data Supplement). We noted a significant interaction effect between 11q deletion and ATM mutation in PFS analysis (HR, 2.45; 95% CI, 1.08 to 5.56; P = .031) but not in OS analysis (HR, 1.04; 95% CI, 0.41 to 2.63; P = .936; Data Supplement). This was supportive of ATM mutations having a different impact on outcome in patients with 11q deletion compared with those with no 11q deletion.
We next analyzed ATM mutation and 11q deletion status in interaction. Compared with ATM wild-type patients with no 11q deletion, the presence of 11q deletion alone (HR, 1.83; 95% CI, 1.29 to 2.60; P = .001) or 11q deletion and ATM mutation (HR, 4.39; 95% CI, 2.48 to 7.77; P < .001) was independently associated with a significantly increased risk of progression (HR, 1.92; 95% CI, 1.25 to 2.96; P = .003) and death (HR, 2.43; 95% CI, 1.24 to 4.74; P = .009), whereas the presence of ATM mutation alone was not (P = .859 and P = .529; Table 4).
Multivariate Analysis of PFS and OS
Further analysis of this model showed that when compared with ATM mutation and 11q deletion combined, presence of both ATM mutation alone (HR, 0.22; 95% CI, 0.10 to 0.46; P ≤ .001) and 11q deletion alone (HR, 0.42; 95% CI, 0.23 to 0.76; P = .004) were associated with significantly lower risk of progression after therapy but not significantly lower risk of death (P = .123 and P = .514, respectively; Data Supplement).
ATM Versus TP53 Abnormalities
To determine whether ATM and TP53 abnormalities have comparable prognostic effect, we analyzed 501 patients from the United Kingdom LRF-CLL4 trial in which data enabled the application of a hierarchical model. We included those with monoallelic (23 patients) or biallelic (25 patients) TP53 abnormalities, TP53 wild-type patients with 11q deletion with ATM mutation (16 patients) or without ATM mutation (48 patients), and 389 patients with no TP53 or ATM abnormality. The univariate analysis of PFS is shown in Figure 3. Of note, although all four groups (11q deletion, ATM mutation and 11q deletion, TP53 monoalleic abnormalities, and TP53 biallelic abnormalities) showed inferior median PFS compared with patients who had no TP53 or ATM abnormality, only patients with biallelic ATM or biallelic TP53 abnormalities revealed inferior outcome when compared with patients with 11q deletion alone. Furthermore, there was no significant difference in median PFS or 2-year PFS between patients with monoallelic TP53 abnormalities, biallelic TP53 abnormalities, and biallelic ATM abnormalities (Data Supplement).
Hierarchical model of the impact of TP53 and ATM abnormalities on (A) progression-free survival (PFS) and (B) overall survival (OS) in patients on Leukemia Research Fund Chronic Lymphocytic Leukemia 4 trial. Kaplan-Meier plots and proportions at risk for 501 patients stratified first by presence of TP53 mutation and/or 17p deletion, then by presence or absence of 11q deletion with or without ATM mutation. Each of four groups (11q deletion alone, ATM mutation and 11q deletion, TP53 monoallelic abnormalities, and TP53 biallelic abnormalities) showed inferior median PFS and OS compared with patients who had no TP53 or ATM abnormality (P < .003 for each). Only patients with biallelic ATM or biallelic TP53 abnormalities showed inferior PFS when compared with patients who had 11q deletion alone (P = .001 and P < .001, respectively). The differences in median PFS or 2-year PFS between patients with monoallelic TP53 abnormalities, biallelic TP53 abnormalities, and biallelic ATM abnormalities did not reach significance.
DISCUSSION
In this study, we have shown that ATM genetic alterations affect patient responses to DNA-damaging agents in a stepwise manner. The presence of an ATM mutation in combination with an 11q deletion confers an increased risk for disease progression above the risk caused by 11q deletion or ATM mutation alone, and this effect is independent of other markers of poor prognosis. Strikingly, the PFS of patients with biallelic ATM inactivation was similar to that of patients with TP53 loss and/or mutation.
Interestingly, this study highlighted differences in clinical outcome among the tumors with functional loss of a single ATM allele. Notably, the PFS of patients with an 11q deletion was shorter than the PFS for those with an ATM mutation. There are several possibilities that could account for this observation.
First, in tumors with 11q deletion, loss of an ATM allele is coupled with monoallelic loss of several genes, including those that are implicated in DDR. These comprise MRE11, a component of the MRN complex that participates in sensing and processing DNA double-strand breaks and directly interacts with ATM; H2AX, a histone subtype directly involved in chromatin relaxation and DNA repair; and MLL and NPAT, which are involved in DDR through their roles in the regulation of replication and cell cycle, respectively.12,13 Thus, the collective impact of haploinsufficiency of several functionally related genes might exceed the impact of the loss of a single ATM allele.
Second, both 11q deletion and an ATM mutation are often later events in the pathogenesis of CLL, and molecular events that precede 11q deletion or ATM mutation may differ causing differences in treatment response between the two subgroups. Indeed, a recent report14 showed differences in gene expression profiles between tumors with ATM mutations and those with isolated 11q deletion.
Third, some of the ATM missense mutations identified in patients on the LRF-CLL4 trial may retain residual ATM function and some degree of DDR. Consistent with this possibility is the milder phenotype observed in patients with AT who carry missense ATM mutations and express ATM protein with residual function.15–17
In this study, we observed the association of ATM mutations with involvement of multiple lymphoid sites. Interestingly, this was not a feature of tumors with isolated 11q deletion. Our observations are consistent with previous data showing the connection between downregulation of ATM mRNA and bulky lymphadenopathy, irrespective of 11q status.18 Further studies are necessary to determine the mechanism by which ATM mutations might have an impact on homing of CLL tumor cells.
Recent reports19,20 suggest that patients with 11q deletion significantly benefit from addition of rituximab to DNA-damage–based therapies. In the CLL8 trial (Fludarabine and Cyclophosphamide With or Without Rituximab in Patients With Previously Untreated Chronic B-Cell Lymphocytic Leukemia), the 3-year PFS for patients with 11q deletion was 64% in the fludarabine, cyclophosphamide, and rituximab arm and 32% in the fludarabine plus cyclophosphamide arm (univariate analysis).19 It will be important to determine whether the incidence of ATM mutations in patients with an 11q deletion differ between those with a short or longer PFS and whether relapse is associated with the acquisition of an ATM mutation. If the benefit of adding anti-CD20 antibodies to chemotherapy is less pronounced in patients with 11q deletion and ATM mutation, the option of avoiding the use of DNA-damaging agents, similar to that adopted for patients with a TP53 abnormality, should be considered.
In summary, our results suggest that patients with biallelic ATM alterations represent a distinctive cohort with a particularly poor response to DNA-damaging agents. Consequently, assessment of ATM status by ATM mutation analysis might be important for the accurate stratification of patients with CLL for tailored treatments and might be particularly important in tumors with 11q deletion.
AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
The author(s) indicated no potential conflicts of interest.
AUTHOR CONTRIBUTIONS
Conception and design: A.M.R. Taylor, Paul Moss, David Oscier, Tatjana Stankovic
Provision of study materials or patients: Martin Dyer, Estella Matutes, Paul Moss, David Oscier
Collection and assembly of data: Gulshanara Ahmed, Ceri Oldreive, Zadie Davis
Data analysis and interpretation: Anna Skowronska, Anton Parker, Sue Richards, Martin Dyer, Estella Matutes, David Gonzalez, Peter Thomas, David Oscier, Tatjana Stankovic
Manuscript writing: All authors
Final approval of manuscript: All authors
Appendix
Impact of ATM mutation on (A) progression-free survival (PFS) and (B) overall survival (OS) in patients on Leukemia Research Fund Chronic Lymphocytic Leukemia 4 trial. Kaplan-Meier plots and proportions at risk for patients stratified by presence of ATM mutation. (P values shown are for log-rank analysis of difference in median survival).
Distribution of all ATM mutations detected in selected cohort of 224 chronic lymphocytic leukemia samples across the ATM gene. Schematic representation of the coding region of the ATM gene with indicated regions encoding conserved domains of the ATM protein. Red arrows indicate truncating mutations, blue arrows indicate missense mutations, arrows with solid lines indicate mutations detected in tumors with 11q deletion, and arrows with dashed lines indicate mutations detected in tumors with no 11q deletion. Localization of the ATM protein domains is given according to National Center for Biotechnology Information Conserved Domain Database: http://www.ncbi.nlm.nih.gov/cdd. FAT, FRAP, ATM, and TRRAP; FATC, FRAP, ATM, TRRAP C-terminal; PI3Kc, phosphoinositide 3-kinase, catalytic domain; TAN, Tel1/ATM N-terminal motif.
Footnotes
-
Supported by Leukemia Lymphoma Research United Kingdom.
-
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
-
Clinical trial information: NCT00004218.
- Received December 7, 2011.
- Accepted August 21, 2012.
