Prognostic Value of Isocitrate Dehydrogenase Mutations in Myelodysplastic Syndromes: A Retrospective Cohort Study and Meta-Analysis

Background Recent genomic sequencing efforts have identified a number of recurrent mutations in myelodysplastic syndromes (MDS) that may contribute to disease progression and overall survival, including mutations in isocitrate dehydrogenases 1 and 2 (IDH1 and IDH2). Methods Pretreatment bone marrow (BM) samples were acquired from mononuclear cells in 146 adult patients with de novo MDS from January 2006 to June 2013. Polymerase chain reaction (PCR) and direct sequencing were performed on exon 4 of IDH1/2 genes and mutation status was correlated with overall survival (OS) and leukemia-free survival (LFS). We then performed a meta-analysis combining previously published and current studies to explore the effect of IDH mutations on OS and LFS in MDS. Results In our study, somatic mutations of either IDH gene were discovered in 11 MDS patients (7.53%) and were significantly correlated with poorer OS (P = 0.007). IDH mutations were specifically associated with a poorer OS in the intermediate-1 risk group by the International Prognostic Scoring System (IPSS) (P = 0.039). In addition, we discovered decitabine achieved a better therapeutic effect compared to other treatments in IDH mutation-positive patients (P = 0.023). We identified six previous studies of IDH mutations in MDS. A meta-analysis of these studies included 111 MDS patients IDH mutations and 1671 MDS patients with wild-type IDH1/2. The hazard ratios (HRs) of OS and LFS for patients with IDH mutations were 1.62 (95% CI, 1.27–2.09) and 2.21 (95% CI, 1.48–3.30), respectively. Conclusion The results from our study and the meta-analysis provide firm evidence that IDH mutations are significantly associated with poorer clinical outcomes in MDS. Identification of IDH mutations may be pivotal for better risk stratification in MDS patients and improving IPSS score. Additionally, hypomethylating agents may be an effective treatment option for MDS patients with IDH mutations.


Introduction
Myelodysplastic syndromes (MDS) comprise a heterogeneous group of hematological disorders defined by blood cytopenias due to ineffective hematopoiesis and an increased risk of developing acute myeloid leukemia (AML) [1,2]. Despite recent advances in therapeutic methods, treatments for MDS are currently tailored to individual patient needs, making the precise forecast of the prognosis an important component of treating patients [3]. Current prognostic scoring systems for patients with MDS are mainly based on karyotypic abnormalities and certain clinical features that are used to stratify risk. Although existing systems such as the IPSS [4], Revised-IPSS [5] and WHO-classificationbased Prognostic Scoring System (WPSS) [6] help to estimate patient outcomes and guide treatment decisions, there remains significant variability in prognosis. Hence, novel molecular markers may offer more precise cancer phenotypes and more accurate estimation of prognosis for MDS patients.
Until now, the pathogenesis of MDS has not been clearly identified, but it is generally acknowledged that genetic mutations and dysfunction of gene contribute to the development and progression of this preleukemic disease [7,8]. Genetic mutations are not currently used in estimating prognosis in MDS but are likely key determinants of overall survival and clinical phenotypes [9]. Therefore, contributing gene mutations may supplement current prognostic systems to improve the prediction of prognosis for MDS patients.
IDH 1/2 are key metabolic enzymes that convert isocitrate to aketoglutarate (a-KG or 2-oxoglutarate, 2-OG), which is an essential cofactor for a-KG dependent dioxygenases [10,11]. These enzymes are associated with diverse cellular processes such as adapting to histone deacetylation, hypoxia, and DNA demethylation [12]. Therefore, IDH mutations may be causally linked to the clinical impacts of patients with MDS. We identified 146 patients with primary MDS and analysed IDH mutation status with OS and LFS. We then performed a meta-analysis combining our data with those of the published literature to furnish a more accurate estimation of the relationship between IDH mutations and MDS.

Patients
One hundred and forty-six adult patients with de novo MDS diagnosed according to World Health Organization (WHO) 2001 criteria [13] were recruited at the department of hematology, the First Affiliated Hospital of Zhejiang University. MDS patients were stratified by cytogenetic risk according to IPSS protocols [4]. All of the subjects were well-informed about the study and provided written informed consent to participate this study. This study was approved by the Institutional Review boards of the First Affiliated Hospital of Zhejiang University. Follow-up data were obtained by telephoning and reviewing patients' medical records. 7 of 146 patients (4.79%) were lost to follow-up. Treatments were performed for patients including chemotherapy regimens (the GAA regimen (granulocyte-colony stimulating factor (G-CSF) 200 mg/m 2 per day on days 1-14, aclacinomycin 10 mg per day on days 1-14; cytarabine 10 mg/m 2 , days 1-14; n = 2); the GHA regimen (G-CSF 200 mg/m 2 per day on days 1-14, homoharringtonine 1 mg/m 2 per day on days 1-14, cytarabine 10 mg/m 2 per day on days 1-14; n = 6); the DA or IA regimen (daunorubicin 40-45 mg/m 2 per day on days 1-3 or idarubicin 8-12 mg/m 2 per day on days 1-3, cytarabine 100 mg/m 2 per day on days 1-7; n = 8); decitabine (20 mg/m 2 /day, days 1-5 or 15 mg/m 2 , q8 h, days 1-3; n = 44)) and supportive care (antibiotics, androgen, alltrans retinoic acid, blood product transfusion and iron chelation therapy; n = 86).

Mutational Analyses for the IDH1 and IDH2 Genes
Pretreatment BM specimens were enriched for mononuclear cells using Ficoll density gradient centrifugation. Genomic DNA was extracted from cryopreserved mononuclear cells using the DNA Kit (Sangon, Shanghai, China) according to the manufacturer's instructions. Approximately 100 ng of DNA was used for each PCR reaction. The primer pairs were the same as those designed by Patnaik et al [14]. The PCR amplification conditions were as follows: 95uC for 5 minutes; followed by 40 cycles of 95uC for 30 seconds, 60uC for 30 seconds, and 72uC for 30 seconds; and finally, 72uC for 5 minutes. PCR products were directly sequenced on both strands using an ABI 3730 automatic sequencer by Sangon.

Statistical Analysis
OS end-points were defined as the time from diagnosis of MDS to death due to any cause or to the time of last follow-up. LFS endpoints were defined as the time from MDS diagnosis to either AML progression or death or failure or alive without disease progression at the date of most recent follow-up. Length of survival comparisons were analyzed using the Kaplan-Meier method. For categorical parameters, overall group differences were compared with the x 2 or Fisher exact test. For continuous variables, overall group differences were evaluated with the Mann-Whitney U test. A Cox proportional hazards model was performed to evaluate the effect of endpoint on OS and LFS for multivariate analysis. Statistical analysis was performed with SPSS 16.0 software package (SPSS, Chicago, USA). All tests were 2-tailed, and a Pvalue of less than 0.05 was considered statistically significant.

Meta-analysis of IDH1/2 Mutations in MDS
To further assess the relationship between IDH1/2 mutations and MDS risk, we conducted a meta-analysis combining our study data with published studies on IDH mutations in MDS [3,14,15,16,17,18]. Two independent reviewers (CH and MXY) performed a systematic literature search using ISI Web of Science, PubMed and the Cochrane Library for relevant papers published before December 2013 by the search term ''(MDS OR myelodysplastic syndrome OR preleukemia OR myelodysplasia) AND (IDH1 OR IDH2).'' Reviews and references of related articles were checked for missing information. Eligible papers met all the following criteria: (1) assessed the association between IDH1/2 mutations and outcomes in MDS; (2) detailed survival information of patients with IDH1 or IDH2 mutations; (3) reported the study in English. Animal studies, letters to the editor without original data, reviews and case reports were excluded. In the event of multiple publications from overlapping study populations or the same study, only the one with the largest sample size was selected ( Figure 1).
The following data were extracted from each article: first author's name, year of publication, country of origin, participant gender, participant age, sample size, MDS subtype, criteria for classification of MDS, karyotypes and IPSS classification. If the required data for the meta-analysis were not available in the published study, we contacted the corresponding authors for missing data.
A general variance-based method and a mathematical HR approximation method [19] in this meta-analysis were simultaneously used to estimate the summary HRs and their 95% CIs for the combined large sample set. Assessing heterogeneity and choosing fixed-effect or random-effect were performed as described previously [20]. Sensitivity analysis was conducted by sequential omission of individual studies and evaluated influence of each study on the stability of the results. Cumulative analysis was performed by assortment of publication time. Publication bias was assessed by funnel plot and Egger's test [21,22]. All statistical analyses were carried out in STATA 11.0 statistical software (Stata Corporation, College Station, Texas), and a P-value less than 0.05 was considered significant.

Meta-analysis Results
As shown in Figure 1, six studies and our data covering a total of 1782 subjects (111 with IDH1/2 mutations, 1671 with wild-type IDH) were included in the meta-analysis. Two of them were from United States [3,14], one from Germany [15] and three from Asia [16,17,18] (Table 2). Two of these studies found a correlation between IDH1/2 mutations and adverse prognosis in MDS [14,15]. For all studies in this meta-analysis, MDS were diagnosed by the WHO [13] or FAB (French-American-British) criteria [24].
Cumulative analysis of the relationship between IDH mutations and MDS was performed via the assortment of studies by publication time. Inclinations toward significant association were evident over time. Moreover, the 95% CI became increasingly narrow with accumulation of more data, indicating the exactness of estimates was progressively boosted by the addition of more subjects ( Figure 4F).

Discussion
Due to the heterogeneity that still exists in the current prognostic scoring systems of MDS, the inclusion of novel molecular markers in these systems may enhance prognostic information. Although single gene mutations are not currently included in prognostic scoring systems, they may be vital to clinical phenotypes and overall survival in MDS. Actually, a great number of single gene mutations including EZH2, SF3B1, TET2, ASXL1 and TP53 have been associated with the development of MDS [25,26]. The illumination of new gene mutations may therefore improve the prevention, diagnosis, prognosis and treatment of MDS.
IDH is a key cytosolic enzyme in the Krebs cycle. It catalyzes the decarboxylation of isocitrate to a-KG, leading to the production of nicotinamide adenine dinucleotide phosphate (NADP) [27,28]. IDH mutations were first reported in a metastatic colon cancer in 2006 [29], and then since 2010 recurring IDH mutations were successively found in MDS (3.42%,12.27%) [3,14,17,30,31,32]. IDH mutations impair the normal enzymes' function, which may be associated with poor prognosis in MDS. However, prior studies have not provided a definitive link between IDH mutations and MDS. Meta-analysis is a useful statistical method for integrating results from independent studies for a specified outcome. Combining the relevant studies increases statistical power and thus makes it possible to detect effects that may be missed by individual studies. Therefore, we summarized here the current data available regarding this potential relationship and revealed several valuable points.
Firstly we discovered a significant relationship between IDH1/2 mutations status and MDS prognosis in the Chinese population, IDH mutations predicted more adverse OS for patients with MDS (P = 0.007). Furthermore, a meta-analysis combining the current and six previously published studies on IDH1/2 mutations and MDS indicated IDH1/2 mutations negatively affected OS (HR, 1.62; 95% CI, 1.27-2.09) and LFS 2.21 (95% CI, 1.48-3.30). Cumulative analysis further confirmed the significant correlation, demonstrating the effect of the variant became progressively significant with each accumulation of more data over time. In Finally, we found IDH1/2 mutation-positive patients with MDS who were treated with decitibine had a significantly longer OS (P = 0.023) suggesting hypomethylating agents might be an effective treatment option for these patients. There are several mechanisms by which IDH1/2 mutations can worsen the prognosis of patients with MDS. (1) IDH mutations occur at low frequency (3.42%-12.27%) in MDS, but IDH1/2 mutations are more frequent in both de novo AML (7.5%-31%) and AML arising from MDS (7.5%) [15,33,34,35,36,37], indicating a role for IDH mutations in leukemic transformation of MDS.
(2) At the cytogenetic level, Caramazza et al. [38] showed a likely association between IDH1/2 mutations and trisomy 8 in MDS, and our results demonstrated 75% (3/4) of IDH1/2 mutants with abnormal karyotypes carried a 27/7qkaryotype. In MDS, +8 and 27/7qkaryotypes were categorized in the intermediate-risk and high-risk cytogenetic group, respectively, suggesting they were linked to poor outcome in MDS. (3) The mutant IDH proteins displayed a gain of function as they could convert the a-KG that was generated by wild-type IDH proteins into 2-hydroxyglutarate (2-HG). Recent studies [39,40] reported that 2-HG was closely related to therapeutic response and relapse in AML. Since MDS and AML share many similar characteristics [41], it is possible that 2-HG is an oncogenic factor in MDS. (4) DNA hypermethylation played a vital role in MDS pathogenesis [42]. Dang et al [43] reported that mutant IDH1/2 proteins produced 2-HG which competitively inhibited a-KG-dependent enzymes, such as the DNA demethylating protein TET2 (Ten-eleven translocation 2) resulting in DNA hypermethylation. Indeed, Figueroa et al. [44]   found that AML patients with IDH1/2 mutations shared a similar methylation profile to those with TET2 mutations, and both mutations led to a block in myeloid differentiation and leukemogenesis. This might also be a potential reason for affecting outcomes in MDS. (5) Accumulation of 2-HG might lead to DNA damage by generating reactive oxygen species (ROS) [45] and inhibit EGLN (Egg-laying defective Nine) with subsequent stabilization of hypoxia-inducible factor 1a (HIF-1a) [46]. DNA damage and HIF-1a stabilization have been reported to be closely linked to MDS pathogenesis [47,48]. There are thus several mechanisms by which IDH1/2 mutations may contribute to MDS pathophysiology but further research is needed to elucidate their exact contributions to the disease. While the findings of this study are largely consistent with previous studies on IDH1/2 mutations in MDS, several limitations should be addressed. First, analyses were based on observational rather than experimental studies. Cohort studies are prone to several types of bias including selection bias and lossto-follow-up [49]. Second, we did not uncover unpublished studies and chose to collect only published articles in English, which could bring publication bias, despite there being no significant evidence of publication bias observed in Egger's test. Third, our study did not assess the potential effects of gene-gene interactions known to influence outcome in MDS such as TET2 mutation-associated hypermethylation [44]. Similarly, we did not account for other known genetic contributions to leukemic transformation in MDS such as ASXL1 loss-of-function [50]. This could lead to possible confounding in our study results. However, since IDH1/2 mutations and TET2 mutations were previously found to be mutually exclusive in patients with AML [44], it is not likely this particular interaction significantly contributed to our results. In conclusion, we screened exon 4 of the IDH1/2 gene in a large cohort of Chinese patients with MDS. Consistent with previous observations, we found that IDH mutations were present in some patients with MDS. IDH1 mutations rather than IDH2 mutations were significantly associated with shorter OS and LFS in patients with MDS. Further studies with larger sample sizes and functional assays of mutant IDH proteins are essential to decipher the role of IDH mutations in the development of MDS. Given that IDH mutations may adversely affect outcome in MDS are relatively easy to assess at diagnosis, examining IDH mutations in MDS may enhance the current prognostic scoring systems and guide patient-specific treatment in MDS. Finally, the identification of IDH mutations in the development and progression of MDS offers the promise of ameliorating the disease using targeted therapeutics against this biochemical pathway.