Identification of Novel Genomic Aberrations in AML-M5 in a Level of Array CGH

To assess the possible existence of unbalanced chromosomal abnormalities and delineate the characterization of copy number alterations (CNAs) of acute myeloid leukemia-M5 (AML-M5), R-banding karyotype, oligonucelotide array CGH and FISH were performed in 24 patients with AML-M5. A total of 117 CNAs with size ranging from 0.004 to 146.263 Mb was recognized in 12 of 24 cases, involving all chromosomes other than chromosome 1, 4, X and Y. Cryptic CNAs with size less than 5 Mb accounted for 59.8% of all the CNAs. 12 recurrent chromosomal alterations were mapped. Seven out of them were described in the previous AML studies and five were new candidate AML-M5 associated CNAs, including gains of 3q26.2-qter and 13q31.3 as well as losses of 2q24.2, 8p12 and 14q32. Amplication of 3q26.2-qter was the sole large recurrent chromosomal anomaly and the pathogenic mechanism in AML-M5 was possibly different from the classical recurrent 3q21q26 abnormality in AML. As a tumor suppressor gene, FOXN3, was singled out from the small recurrent CNA of 14q32, however, it is proved that deletion of FOXN3 is a common marker of myeloid leukemia rather than a specific marker for AML-M5 subtype. Moreover, the concurrent amplication of MLL and deletion of CDKN2A were noted and it might be associated with AML-M5. The number of CNA did not show a significant association with clinico-biological parameters and CR number of the 22 patients received chemotherapy. This study provided the evidence that array CGH served as a complementary platform for routine cytogenetic analysis to identify those cryptic alterations in the patients with AML-M5. As a subtype of AML, AML-M5 carries both common recurrent CNAs and unique CNAs, which may harbor novel oncogenes or tumor suppressor genes. Clarifying the role of these genes will contribute to the understanding of leukemogenic network of AML-M5.


Introduction
Acute monocytic leukemia, also named acute myeloid leukemia-M5 (AML-M5), is one of the most common subtypes of AML defined by the French-American-British (FAB), which is comprised by more than 80% of monoblasts (AML-M5a) or 30-80% monoblasts with (pro)monocytic differentiation (AML-M5b). Responses to chemotherapy and prognosis in AML-M5 patients are varied. The recently proposed World Health Organization (WHO) classification attempts to associate the prognostic variance with cytogenetic abnormalities [1][2]. However, rather than t (8;21) in AML-M2, t (15;17) in AML-M3 and inv(16)/t (16;16) in AML-M4eo, specific subtype-associated translocation is lacking in AML-M5. It has been described that a part of AML-M5 patients carry 11q23 (MLL gene) rearrangement with a wide spectrum of recurrent translocation partner chromosomes [3][4]. In this sense, it is reasonable to assume that unbalanced chromosomal abnormalities might play a major role in the leukemogenesis of AML-M5. Especially those cryptic alterations, which are invisible for traditional banding analysis, are not rare in AML with normal karyotype [5][6].
It is necessary to evaluate the chromosomal changes in a whole genomic level. However, banding analysis, currently as the first choice to evaluate the genomic abnormalities of AML, is frequently to get inconclusive results due to chromosome condensation, imperfect banding, and the presence of few metaphases. FISH is a more sensitive technique, but is limited by the genomic coverage properties. Array CGH provides some distinct advantages including the potential for comprehensive genomic profiling and the ability to delimit the boundaries of specific genomic aberrations. Recently, it is considered as a powerful method to analyze genomic imbalances in hematopoietic malignancies, especially to determine cryptic and recurrent anomalies not observed by routine techniques [7][8][9][10][11].
The primary aim of present study is to assess, by means of array CGH assay, the possible existence of unbalanced chromosomal abnormalities, and delineate the copy number alterations (CNAs) in 24 cases of adult AML-M5 patients. Besides this, recurrent CNAs harboring cancer genes are also analyzed. The novel candidate oncogenes and tumor suppressor genes aberrations which might be implicated in the leukemogenesis of AML-M5 were revealed in this study.

Normal karyotype was in the majority of AML-M5 cases
Karyotypes were available for 16 AML-M5 cases (Table S1 in File S1). 11 of them (10 newly diagnosed cases and 1 relapsed case) had normal karyotypes and 5 cases carried 2 or 3 chromosomal abnormalities. t(9;11)(p22;q23) as the recurrent chromosomal change was presented in 3 cases.

Array CGH delineated the genomic imbalances of AML-M5
A total of 117 CNAs with size ranging from 0.004 to 146.263 Mb were recognized in 12 of 24 cases, involving all chromosomes other than chromosome 1, 4, X and Y ( Figure 1 and Table S1 in File S1). The cryptic chromosomal aberrations less than 5 Mb took up 59.8% of all CNAs (Figure 1). It was worth noting that in the 10 newly diagnosed cases with normal karyotype, microduplications spanning 11q23.3 carrying MLL and 5q35.3 carrying AGXT2L2 and COL23A1 were detected in case #8 and case #16, respectively. 12 recurrent CNAs were mapped in Figure 1 and listed in Table 1. Notably, in contrast to the most common recurrent CNAs reported by 8 representative genomic studies of MDS and AML using oligonucleotide array CGH or SNP array [5,[12][13][14][15][16][17][18] (Table 2), novel CNAs including gains of 3q26.2-qter and 13q31.3 as well as losses of 2q24.2, 8p12 and 14q32 were proposed in the present study. Except that gain of 3q26.2-qter is a large chromosomal abnormality, all the other recurrent CNAs were cryptic and harbored interesting genes.
Translocation and inversion rather than partial duplication of 3q26.2 were identified by FISH By array CGH, 3q26.2-qter was identified as a recurrent chromosomal anomaly. To investigate whether the gain of 3q26.2qter, was closely associated with AML-M5 or other subtypes of AML, FISH with EVI (3q26.2) break-apart probe was hybridized in bone marrow cells of 69 cases of de novo AMLs (Table S2 in File S1). Normal two fusion signals were observed in 67 cases and abnormal FISH signal pattern with two fusion signals plus an extra red signal (TEL'EVI1) was shown in another two cases. One of the cases (ID: 37) was AML-M5, containing 89% of abnormal cells in the interphase. t (3;21) in the metaphase were proven by subsequent FISH with ETV6/AML1 probe (Figure S1A-F in File S1). The other one (ID: 65) was AML-M6, which carried 93% of abnormal cells produced by inv(3)(p25;q26) (Figure S1G-I in File S1). No partial duplication of 3q involving EVI gene was observed.

Candidate oncogenes and tumor suppressor genes were inferred by array CGH
Oncogenes and tumor suppressor genes that have been reported to be implicated in carcinogenesis or leukemogenesis were further singled out from the cryptic recurrent CNAs (Table 3). Microdeletion of 0.209 Mb and large terminal deletion of 51.233 Mb encompassing FOXN3 gene at 14q were observed in two cases (case #13 and #24) ( Figure 2). We subsequently investigated the expression levels of the FOXN3 in a cohort of acute leukemia samples including 78 AMLs and 19 ALLs (Table S3 in File S1). In comparison with healthy controls (HCs) (1.8060.27), significantly lower expression of FOXN3 was found in the AML group (0. 9060.12) (p,0.05) ( Figure S2 in File S1). No significant difference was observed between AML-M5 and non-AML-M5 ( Figure S3 in File S1). A low tendency toward expression of FOXN3 was detected in the ALL group (1.1660.65), but the expression was not significantly different from the controls nor the AML groups (p.0.05) ( Figure S2 in File S1). Gain of 11q23.3 containing MLL gene was observed in 3 cases with size ranging from 0.004 Mb to 23.133 Mb. Cryptic chromosomal losses of 9p21.3 or a large deletion of 9p involving CDKN2A and MTAP were observed in 3 cases. Strikingly, gains of 11q23.3 and losses of 9p21.3 coexisted in two cases. To confirm the concurrent deletion of 9p21.3 and amplication of 11q23.3 inferred by array CGH, FISH analysis using MLL and CDKN2A probes was performed on case #18 ( Figure 3). Consistent with array CGH, amplication of MLL with three or four fusion signals was shown in 69.5% of the 200 nuclei. Homogenous and heterogenous deletion of CDKN2A, indicated by loss of double or single red signal, was observed in 71.5% and 19% of 200 examined nuclei respectively. Totally, MLL amplication and CDKN2A deletion concurred in 68.5% of 200 nuclei.

Correlation of CNAs number and clinico-biological parameters
To investigate the prognostic value of the number of array CGH aberrations, 22 cases of newly diagnosed AML-M5 were grouped based on the numbers of CNA. As the median of CNAs per case was four, cases with four or less CNAs were grouped apart from cases with five or more. The numbers of CNA in the two groups were correlated with clinico-biological parameters which suggest a poor prognosis. These parameters included age.60 years old, white blood cell (WBC) count .10610 9 /L, hemoglobin concentration ,80 g/L, platelet count ,50610 9 /L, higher percent of blast cells in bone marrow (comparing to the average blast ratio in bone marrow) and responses to treatment, such as number of CR. By statistic analysis, no significant difference was observed between the two groups (table 4).

Discussion
In the last few years, the classification of AML has developed from a single morphological level to a MICM basis. As one of the diagnostic methods, genetic abnormality examination plays an important role in further sub-classification of AML and gives more clues to therapy selection and prognosis [1][2]. The cohort of patients of our study is in a special subgroup of AML, AML-M5, which is characterized by variable therapy responses and low incidence of recurrent chromosomal abnormalities at the level of conventional cytogenetic method. The present study refined the genomic aberrations of AML-M5 by an integration of chromosome, FISH and array CGH procedure. The majority of the genomic alterations are cryptic and invisible to G-banding analysis, indicating that high-resolution array CGH approach is necessary to provide more precise description of chromosomal aberrations in AML-M5. Particularly, in the patients with normal karyotype, array CGH has been able to unveil the small CNAs. However, similar with the previous studies, the recurrence of the small CNAs is very low [19][20]. Normal genomes are observed in nearly half of the AML-M5s, inferring that additional studies, such as gene sequencing or epigenetic analysis, might be necessary to investigate the underlying mechanism of AML-M5.
Array CGH improves the detectable rate of chromosomal abnormalities, which make it easier to map the recurrent chromosomal aberrations of AML-M5. As a subtype of AML, it seems that AML-M5 carries both the common recurrent CNAs which have been reported in the previous AML studies and novel recurrent CNAs which are noted as candidate AML-M5 associated CNAs. However, whether these novel recurrent CNAs are specially associated with AML-M5 are needed to be clarified. Gain of 3q26.2-qter is the sole large recurrent chromosomal   change in the present study. There is an overlap between 3q26.2qter in case 12 and 3q21.1-qter in case 18. It is known that 3q21q26 anomaly is the most common karyotype in acute myeloid patients with 3q abnormalities [21][22]. RPN1 gene at the breakpoint of 3q21.3 juxtaposing to EVI1 gene at the breakpoint of 3q26.2 is the key mechanism of this type of AML. However, the significance of 3q26.2-qter amplication is not well known in AML although it is considered as a recurrent chromosomal abnormality in several solid tumors [23][24][25]. This study inferred a possible recurrent amplication of 3q26.2-qter, which is associated with EVI gene by array CGH. However, neither its recurrence in AML nor its specific association with AML-M5 was confirmed by FISH. There's limitation on the confirmatory FISH analysis since it only focus on EVI gene. Recent study by Hussenet and his colleagues  reported that SOX2 is an oncogene activated by recurrent 3q26.3 amplifications in human lung squamous cell carcinomas [26]. Therefore, amplication of 3q26.2-qter might serve as a novel recurrent 3q abnormalities and it possibly has a different leukemogenic mechanism from the classical 3q21q26 anomaly in AML.
By the advantages of array CGH, the candidate genes in the cryptic recurrent CNAs have been singled out. It is the first time that FOXN3 deletion is observed in leukemia. The FOXN3 acts as a transcriptional repressor and has been implicated in DNA damage-inducible cell cycle arrests in saccharomyces cerevisiae [27]. However, the physiological functions of FOXN3 in mamals are not known. Reduced expression of FOXN3 has been reported in several cancers, such as renal cell carcinoma and oral squamous cell carcinoma [28][29]. Scott et al. found that FOXN3 recruited SKIP, which played a role in cell growth and differentiation through the TGF-b pathway and repressed genes important for tumorigenesis [30]. The deletion of FOXN3 gene is recurrent in this study, suggesting that FOXN3 possibly serves as a novel TSG and contributes to the leukemogenesis. However, subsequent expression study suggested that reduced expression of FOXN3 was not only presented in AML-M5 but also other subtypes of AML, indicating that the FOXN3 encodes a myeloid marker rather than a specific leukemogenic protein of AML-M5. The MLL gene is reported to be highly associated with AML with complex karyotypes [31]. In this study, it suggests that the amplication of MLL frequently co-exists with the deletion of CDKN2A. CDKN2A encode cyclin-dependent kinase inhibitor, which have emerged as an important effecter in aging as well as a potent tumor suppressor. The main genetic alterations involving CDKN2A are deletions by bi-or monoallele or hypermethylation of 59 GpG island in its promotor [32][33][34]. In AML, hypermethylation of CDKN2A has been reported, but deletion of CDKN2A is rare [35][36][37]. It is intriguing to reveal the coexistence of MLL amplication and CDKN2A deletions in AML-M5 and the cooperation of these genes rearrangements in leukemogenesis of AML-M5 needs to be further clarified.
Due to the limitation of case number, no statistical difference was shown between the number of CNAs and clinico-biological parameters in our analysis. More cases are necessary to evaluate the clinical significance of the number of CNAs.
In summary, we genetically defined AML-M5 by karyotype, FISH and array CGH. Array CGH is necessary to complement the traditional cytogenetic analysis in revealing the cryptic genomic alterations in AML-M5. As a subtype of AML, AML-M5 carries both common recurrent CNAs coexisting in AML and novel CNAs in AML-M5. The recurrent CNAs harbor oncogenes or tumor suppressor genes. Further studies are needed to understand the role of these genes in the leukemogenic network of AML.  [38]. The ratio of males and females is 15:9 and the median age of patients was 47 years old (range, 15 to 72 years). The average value of blast count in bone marrow was 0.72 (range, 0.40 to 0.98). The clinical information of the patients is summarized in Table S4 in File S1.

Therapy
Twenty two patients were newly diagnosed and received a single course of daunorubicin plus cytarabine (DA) or idarubicin plus cytarabine (IA) regimen for induction treatment. For the patients who failed to achieve complete remission (CR) or relapsed from CR, salvage therapy was used, including low-dose cytarabine, aclarubicin and granulocyte colony-stimulating factor (G-CSF) (CAG) or fludarabine, cytarabine and G-CSF (FLAG) regimen. Intermediated dose cytarabine (IDAra-C) 6 antharcycline, mitoxantrone plus cytarabine (MA) and DA were chosen as postremission therapy for patients achieved CR and it lasted for 6-8 courses. No further therapy was applied to patients remaining in remission after postremission therapy. The details of each regimen were listed in Table S5 in File S1.

Criteria of response and definition of relapse
For response assessment, bone marrow aspiration was performed between days 21 and 28 after induction therapy. CR was defined as a bone marrow with normal hematopoiesis, blasts in bone marrow less than 5%, granulocyte count $1.0610 9 /L, platelet count $100610 9 /L, no myeloid blasts in the peripheral blood and no extramedullary disease. Patients who did not fulfill the above criteria were defined as unremission. Relapse was defined as the reappearance of more than 5% blasts in bone marrow aspirates or extramedullary leukemia in patients with a previously documented CR.

Ficoll gradient separation and cryopreservation
Bone marrow mononuclear cells (MNCs) were isolated by Ficoll-paque gradient centrifugation (TAKARA Biotechnology, Dalian, China), aliquoted into fetal calf serum (FCS) with 10% DMSO, and cryopreserved in the liquid phase of a liquid nitrogen tank.

Preparation of Sample DNA and RNA
Liquid nitrogen-cryopreserved (liquid phase) bone marrow mononuclear cells were washed and recovered by centrifugation. Following that, DNA was extracted with phenol-chloroform, and precipitated with ammonium acetate, ethanol and glycogen. RNA was prepared from approximately 2610 5 to 2610 6 blasts with Trizol (TAKARA Biotechnology, Dalian, China), and resuspended in 50 ml DEPC-treated water. Complementary DNA was made from ,20 ng of RNA using the SuperScript III first strand synthesis kit and oligo-dT priming [Promega Corporation, an affiliate of Promega (Beijing) Biotech, China].

Cytogenetics
Short-term cultures of unstimulated bone marrow samples were prepared and fixed according to our standard laboratory protocols. Karyotype analysis was performed by R banding technique. Chromosome study could not been performed on six bone marrow samples due to the limitation of cell number. The cytogenetic abnormalities were described according to the International System for Human Cytogenetic Nomenclature.

Array CGH
Array CGH analysis was performed on all DNA samples obtained from 24 patients with AML-M5 following the manufacturer's protocol with minor modifications, (Roche NimbleGen System Inc., Madison, WI). The array consisted of approximately 720,000-mer oligonucleotide probes that spanned both coding and noncoding sequences with average genome coverage of approximately 2 kilobases. A commercially available pooled normal control DNA was used (Promega Corporation, Madison, WI) for reference. The patients and the reference DNA were labeled with either Cyanine 3 (Cy-3) or Cyanine 5 (Cy-5) by random priming (Trilink Biotechnologies, San Diego, CA) and then hybridized to the chip via incubating in the MAUI hybridization system (BioMicro Systems, Salt Lake City, UT). After 40-hours hybridization at 42uC, the slides were washed and scanned using a GenePix 4000B (Molecular Devices, Sunnyvale, CA). NimbleScan version 2.4 and the SignalMap version 1.9 were applied for data analysis (NimbleGen System Inc, Madison, WI). The genomic locations were retrieved from National Center for Biotechnology Information (NCBI) build 36 (hg 18). Frequently affected regions recently detected as copy number polymorphisms were excluded from data analysis according to the Chromosome Number Variation (CNV) database in our lab and genomic variants in human genome (Build 36: Mar. 2006 Table S2 in File S1.  Table S2 in File S1. Expression of FOXN3 determined by Quantitative realtime PCR (qRT-PCR) qRT-PCR was performed to determine the expression of FOXN3 on bone marrow MNCs of 97 patients with acute leukemia and 16 normal controls who had normal myelograms (Table S3 in File S1). Complementary DNAs were made from ,20 ng of RNA using the Superscript I II first strand synthesis kit (Invitrogen) and oligo-dT priming. Primers and TaqMan-based probes were purchased from Applied Bio-systems (Applied Biosystems, Beijing, China) and listed in Table S6 in File S1. Duplicate amplification reactions were done with an ABI 7500 Real time PCR System (Applied Biosystems, Foster City, CA, USA) and the data were analyzed by the ABI 7500 system SDS software (1.4 Version).

Statistical analysis
The Fisher's exact test was employed to assess the association of CNAs numbers per case with clinico-biological parameters. The correlation of the expression levels of FOXN3 and the subtypes of acute leukemia was analyzed by the student's t-test. All statistical analyses were performed using GraphPad Prism5 and p-value #0.05 was considered as statistically significant.

Supporting Information
File S1 Figure S1 Figure S2, The comparisons of FOXN3 Levels in health control, ALL and AML. The significantly reduced FOXN3 level was observed between AML and health controls. *: p,0.05. The FOXN3 level was lower in ALL than health control whereas no significantly difference was observed. Figure S3, The comparisons of FOXN3 Levels in health control, M5 and non-M5. No significantly difference was observed on the FOXN3 levels between AML-M5 and non-AML-M5 (p.0.05).