A Pipeline with Multiplex Reverse Transcription Polymerase Chain Reaction and Microarray for Screening of Chromosomal Translocations in Leukemia

Chromosome rearrangements and fusion genes present major portion of leukemogenesis and contribute to leukemic subtypes. It is practical and helpful to detect the fusion genes in clinic diagnosis of leukemia. Present application of reverse transcription polymerase chain reaction (RT-PCR) method to detect the fusion gene transcripts is effective, but time- and labor-consuming. To set up a simple and rapid system, we established a method that combined multiplex RT-PCR and microarray. We selected 15 clinically most frequently observed chromosomal rearrangements generating more than 50 fusion gene variants. Chimeric reverse primers and chimeric PCR primers containing both gene-specific and universal sequences were applied in the procedure of multiplex RT-PCR, and then the PCR products hybridized with a designed microarray. With this approach, among 200 clinic samples, 63 samples were detected to have gene rearrangements. All the detected fusion genes positive and negative were validated with RT-PCR and Sanger sequencing. Our data suggested that the RT-PCR-microarray pipeline could screen 15 partner gene pairs simultaneously at the same accuracy of the fusion gene detection with regular RT-PCR. The pipeline showed effectiveness in multiple fusion genes screening in clinic samples.

PCR method, especially reverse transcription-PCR (RT-PCR), nowadays has been shown to be a sensitive tool in the clinical evaluation of leukemia. As there are many distinct genetic alterations in various leukemia subtypes, it would be extremely labor intensive to evaluate specific fusions via a panel of individual monoplex assays. This can be avoided by the use of multiplex RT-PCR assays with various downstream 2 BioMed Research International detection methods, such as gel-based techniques and bead array [7][8][9][10]. Microarray is another useful detection assay. Two biochip-based diagnostic systems were reported: a gelbased biochip by Nasedkina et al. [11,12] and MLLFusionChip and AMLFusionChip from France [13,14]. In these previous works, the gel-based biochip only targeted 7 chromosomal translocations, addressing 13 fusion variants in sum, while other two chips covered certain leukemia group. That was far from translocation types needed for the initial screening stage. In addition, the procedure of PCR was very complicated, with at least two parallel nested multiplex reactions. Therefore, we planned to make RT-PCR-microarray assay much easier, hoping to detect the frequently occurring and well-defined translocations in leukemia.
In our study, we described (a) the improvement of multiplex RT-PCR in combination with microarrays analysis system that facilitated the simultaneous detection of 15 chromosomal aberrations, including more than 50 mRNA splice variants with prognostic value; (b) the sensitivity level of each fusion gene in cell lines or clinical patients unique translocations; (c) the application of this method to test 200 leukemia clinical patients; and (d) the potential diagnostic value of this procedure for detection of unusual fusion genes or fusion junctions.

Cell Lines and Patient
Samples. The 15 chromosomal translocations studied and the GenBank database references for the genes involved were given in Table 1. Cell lines and some patient samples with unique translocations as positive controls were also used in the study. The leukemic cell line HL-60 served as a negative control. Cells were maintained in RPMI 1640 (Gibco, Grand Island, NY, USA) supplemented with 10% fetal bovine serum (Gibco). Meanwhile, we also constructed fusion fraction RNAs for translocations (e.g., PLZF-RARA, NPM1-RARA, CBFB-MYH11 type D, SIL-TAL1, and BCR-ABL p230) that had neither cell lines nor positive samples. We constructed expression plasmids using pcDNA3.0 vector (Invitrogen, Carlsbad, CA, USA), then transfected the constructed expression plasmids into 293 T cells using Lipofectamine 2000 (Invitrogen), and finally collected cells in TRIzol (Invitrogen) 48 h after transfection.
200 bone marrow samples of ALL, AML, and CML patients were also used in this work for clinical screening. They were received from the Department of Hematology of Shanghai Children's Medical Center (Shanghai, China), Ruijin Hospital (Shanghai, China), and Changhai Hospital (Shanghai, China).

RNA Preparation.
Leukemia cell extract was prepared by the TRIzol (Invitrogen) homogenization method according to the manufacturer's recommendations. Total RNA was resuspended in RNase-free water, and its concentration and quality were determined by NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) and by 2% agarose gel electrophoresis. Then the RNA was stored at −80 ∘ C.

Multiplex RT-PCR Primers.
We planned to use only one multiplex RT-PCR reaction to detect more than 50 fusion gene variants of 15 chromosomal rearrangements. To reduce the number of primers used in multiple PCR, we designed chimeric forward primers and chimeric reverse primers containing both gene-specific and universal sequences.
The gene-specific portions of the chimeric forward primers for fusion genes in leukemia were designed or modified with the primer analysis software Primer Premier 5.0 (Premier Biosoft International, Palo Alto, CA, USA), based on sequence data of fusion partner genes deposited in GenBank database or from publications [7] (Table 2). We added bacteriophage promoter T7 and SP6 to the 5 end of the PCR amplifying forward and reverse primers, respectively, as the universal parts. Thus in the first reverse transcription step, the synthesized cDNAs could be linked to SP6 sequence at their 5 end, so that only one reverse universal primer SP6 was required in the following multiple PCR step. In addition, the universal primer, SP6, was 5 -labeled with biotin so that amplicons could be analyzed with microarrays ( Figure 2). We applied GUS gene (beta-glucuronidase, M15182) [15] as internal positive control to avoid false-negative results cased by varying RNA quality and handling errors. Oligonucleotide primers were provided by Generay Biotechnology (Shanghai, China).

Microarray Probes.
Oligonucleotide probes for translocations in leukemia were designed with the primer analysis software Primer Premier 5.0 (Premier Biosoft International), based on cDNA sequences deposited in GenBank. We designed 29 chimeric junction probes for each splice variant already known for fusion gene transcripts of interest, except the MLL rearrangements. In addition, 20 unique probes for each fusion genes partner were selected as positive controls in each gene fusion event, which located upstream or downstream of the described breakpoint. This microarray also included one probe for the GUS gene served as an internal positive control for the quality of DNA target prepared for hybridization. The sequences of probes were shown in Table 3. All probes were included in triplicate. The probes were 20-30 nucleotides (nt) long, and their specificity and sensitivity were tested by positive samples to ensure the accuracy of the microarray. The probes and biotinylated control probes were spotted onto CapitalBio optical grade aldehyde slides (CapitalBio Corporation, Beijing, China) ( Figure 1). Oligonucleotide probes were synthesized in Sangon Biotech (Shanghai, China).  the cDNA reaction mixture was heated at 70 ∘ C 15 min to inactivate reverse transcriptase. For the PCR step, a two-round PCR amplification was carried out. 1 L cDNA reaction mixture was added to 20 L of multiplex mixture containing a mixture of chimeric PCR primers (3 pmol of each), 4 pmol the universal primer SP6, 11 mM Tris-HCl, pH 8.3, 55 mM KCl, 1.5 mM MgCl 2 , 15% DMSO, 0.4 mM each of dNTPs, and 1.25 U of ExTaq HS polymerase (Takara, Dalian, LN, China). PCR cycles included 30 cycles of 95 ∘ C for 30 s, 55 ∘ C for 40 seconds, 72 ∘ C for 1 min, and finally 72 ∘ C for 7 minutes. After the first PCR, 1 uL of the first round product was added to 20 uL secondround multiplex mixtures that contained 10 mM Tris-HCl, pH 8.3, 50 mM KCl, 1.5 mM MgCl 2 , 0.2 mM of each dNTP, 0.4 pmol of forward primers T7 and 4 pmol of reverse primer SP6 (labeled with biotin), and 1 U of ExTaq HS polymerase (Takara). The second PCR consisted of 30 cycles of 95 ∘ C for 30 s, 50 ∘ C for 30 seconds, 72 ∘ C for 1 min, and finally 7 minutes of extension at 72 ∘ C. 3 L of the biotinylated-complexes was electrophoresed on an ethidium bromide stained 2% agarose gel. Negative controls without cDNA template were included for all PCR reaction mixtures.

Hybridization and Image Analysis.
Microarray firstly was blocked in blocking buffer (2x SSC, 1 mg/mL BSA, 0.2% SDS) for 1 h at room temperature and then washed by pure water. Hybridization was performed in 25 L hybridization solution (5x SSPE, 0.1 mg/mL Salmon Sperm DNA, 0.5x Denhardt's solution, all from Invitrogen) and 20 L biotin labeled PCR products. Before hybridization,  T7  TAATACGACTCACTATAGGGA  SP6 ATTTAGGTGACACTATAGA SP6-biotin biotin-ATTTAGGTGACACTATAGA R: reverse primer, F: forward primer. a Primer composition was given with gene accession number plus universal gene (T7 or SP6), the starting position on the gene, and total length in brackets.
the PCR product was treated with SAP (Takara) and ExoI (Takara) to eliminate primer dimers, then denatured at 94 ∘ C, and briefly cooled on ice. Then the hybridization mixture was applied on the hybridization area covered with hybridization chamber (CapitalBio Corporation) and 12-14 h at 46 ∘ C. After incubation, the slide was washed in washing solution A (2x SSC, 1% SDS), washing solution B (1x SSC, 0.4% SDS), and washing solution C (0.6x SSC) for 5 min each, followed by incubation in Cy3-Streptavidin (SA) (Invitrogen) 1 h at room temperature. After washing, the fluorescence signal was detected on a GenePix 4000B scanner (Axon Instruments Inc., Union City, CA, USA) using 532 nm excitation. Data processing and image analysis were performed using GenePixPro6.0 software (Axon Instruments Inc.). When the averaged signal intensity from hybridization was higher than both two standard values, background value (negative samples and blank control plus 3x SD) and cutoff value (15% of biotinylated control value), we considered the probe as true signal.     Each fusion gene included one probe for each partner and chimeric junction probes specific for each splice variant already known. The partner's probe was as positive control in each gene fusion event, which located upstream or downstream of the described breakpoint. MLL rearrangements only had each partner's probe. We spotted probes for each fusion gene in horizontal row, first chimeric junction probes then specific partner's probes, with continuous triplicate repeats. The biotinylated control probe was arranged in the top row and left column to position the probes and also to control fluorescence signal hybridization. After incubating in Cy3-SA, the fluorescence signal was detected on certain location.

MLL-ELL
The signal intensities of every three spotted probes were averaged. By trial and error, we confirmed that for each probe the average signal intensity from negative samples and blank control plus 3x SD was a background value of this probe. To remove the false positive signals, another standard value is established. We finally chose 15% of biotinylated control probe signal intensity as the second standard, the cutoff value, through the statistics on over 100 samples. When the signal intensity from hybridization was higher than both two standard values, the background value and cutoff value, we believed the probe as true signal. In addition, when the true signal included one chimeric junction probe and two partner's probes of the same fusion gene group, the result was considered to be positive; on the contrary, the result was considered to be negative.
Furthermore we constructed a series of plasmids for each splice variant of fusion genes studied here for optimizing all the probes. In addition, the microarray was also tested with eighteen control samples (six cell lines, eight patient positive samples, and five constructed fusion fractions) of known genotypes to check the reliability of the results (Figure 3 and Supplementary Figure 1). Figure 3 showed four examples, cell line KASUMI-1 with AML1-ETO, K-562 with BCR-ABL b3a2 splice variant, AML patient with PML-RARA S-form, and ALL patient with E2A-PBX1(I), respectively. As could be seen, the signal from the GUS-gene was present in all samples (row L).
We also carried out the precision of our approach. Repeatability was tested by total RNA from one positive cell line K-562 bearing a t(9;22) BCR-ABL b3a2 splice variant. The RNA was divided into three parts. Each part of RNA was processed for reverse transcription and multiplex RT-PCR. Finally, the three reaction productions were hybridized onto three individual microarrays. The three hybridization results were identical, positive for the BCR-ABL b3a2 translocation (data not shown). Reproducibility was blindly evaluated by three different operators in three independent experiments using total RNA from cell lines (three positive, Kasumi-1, THP-1, and REH, and one negative, HL-60). The mean reproducibility of operators and experiments was 100 and 100%, respectively (Table 4). These repeatability and reproducibility results demonstrated that the multiplex RT-PCR-microarrays method had good intra-and interassay precision.

Patient Samples
Testing. Bone marrow samples from 200 patients (74 AML, 115 ALL, and 11 CML) were analyzed blindly using the multiplex RT-PCR combined with microarrays method. Some examples were showed in Figure 4(a).
In AML patients, 37.8% chromosomal aberrations were found. Most frequent translocations were t(15;17) PML-RARA (13.5%), t(8;21) AML-ETO (10.8%), and inv(16)(p13q22) CBFB-MYH11 (6.8%). Among ten patients with PML-RARA, there were five long (L)-form (bcr1) and four short (S)-form (bcr3) transcripts. It was interesting that the one remaining patient number 57 might be V-form transcript, because of the only two signals of probes PML-L and RARA besides GUS (Figure 4(a), patient number 57). The breakpoint of PML was located within exon 6 (base position 1683), belonging to bcr2, which was referred to as "variant" or V-form. Five CBFB-MYH11 patients were all type A, which was consistent with previous findings that type A transcript was the major type in inv(16) positive patients.
A relatively large proportion of CML patients were found t(9;22) BCR-ABL p210 about 72.7% including two transcripts with the b3a2 (45.4%) and b2a2 (27.2%). Perhaps, the amount of CML patients studied here was only 11, resulting in the fact that the positive portion was lower than expected.
Validation of the multiplex-microarray method was done either by cytogenetics, fluorescence in situ hybridization (FISH), or RT-PCR analysis for all 200 patient samples in the Department of Hematology of Shanghai Children's Medical Center, Ruijin Hospital, Changhai Hospital of Shanghai (Shanghai, China). The detailed information was listed in Supplementary Text 1 and Supplementary Table 1. The comparison of the two methods was shown in Table 5, and the concordance was about 96.5%, except four in AML and three in ALL. It was noted that these seven samples were not involved in the assay. We also performed additional individual PCR reactions for all positive samples (Figure 4(b)). The results of both methods were almost identical.

Discussion
Monoplex or multiplex RT-PCR techniques have been increasingly used to characterize chromosomal translocations found in leukemic cells, with numerous advantages over traditional cytogenetics and FISH, including shorter turnaround time, no requirement for dividing cells, and detection of cryptic translocations. A multiplex RT-PCR protocol to simultaneously detect 29 translocations had been shown to be effective for clinical screening, followed by series of identifying split-out analysis with primers specific for individual translocations [7]. It was still time-and labor-consuming. Here we facilitated the detection in a single PCR reaction (two rounds), in combination with the following microarrays to simultaneously detect a great variety of different fusion transcripts.

TEL-AML1ex3
No. 178 No. 303 BCR-ABL b2a2   Compared with the previous reports [11][12][13][14]25], the procedure established here had many advantages. First, the procedure here targeted more chromosomal translocations. We focused on about 15 chromosomal translocations, including more than 50 mRNA splice variants. However, the previous reports only had allowed the analysis of only a few rearrangements and one split variant in each fusion gene. For instance, the AMLFusionChip [14] only included three major types of rearrangements in AML: AML1-ETO, CBFB-MYH11, and 11q23/MLL abnormalities (MLL-AF9, MLL-ENL, MLL-AF6, and MLL-AF10); the gel-based biochip developed by Nasedkina et al. [12] addressed 13 fusion variants in 7 leukemia translocations. Second, the range of detection was wider, covering AML, ALL, and CML. AMLFusionChip was only limited to one leukemia group. Third, the operating steps were easier. The assay here used chimeric primers, amplifying different fusion transcripts in the same condition. In previous reports, it was necessary to perform the standard multiplex RT-PCR with multiple parallel nested PCR reactions.
A control gene was also used as an internal control to evaluate the RT-PCR reaction and hybridization process. ABL, B2M, and GUS were three candidate control genes for multiplex RT-PCR [15]. ABL is a more reliable control gene to compare diagnostic and MRD samples, yet there may be competition between BCR-ABL fusion transcripts and ABL gene when amplifying samples with t(9;22). However GUS gene is also similarly expressed in normal and diagnostic samples, and variation of GUS and fusion gene transcript expression are correlated. Based on these, we chose GUS as control gene in this study.
We also tested the sensitivity of positive controls. It was the fact that the sensitivity was weak in multiplex RT-PCR compared to single PCR reactions, especially for detection of minimal residual diseases. We could observe in this report that the sensitivity of our reactions was decreased one to two orders of magnitude, due to the existence of sixteen primers in first reaction round. However, in incipient patient diagnosed with leukemia, the cell source used for RNA preparations was usually greater than 90% leukemic blasts. Moreover, the use of microarrays could improve the efficiency of discriminating splice variants and minimize the risk of contamination. Thus this multiplex PCR-microarray assay was suitable for diagnosing de novo leukemia.
We had scaled up in this method of analysis 200 patients with 74 AML, 115 ALL, and 11 CML. In 6 AML and 9 ALL cases, the internal control band using multiplex-microarray could not be detected. This could be ascribed to insufficient amount or quantity of RNA. Four in AML and 3 in ALL samples were beyond the detection scope of the assay. With regard to AML, t(15;17)PML-RARA, t(8;21)AML-ETO and inv(16)(p13q22)CBFB-MYH11 were mainly identified. In ALL, t(12,21)TEL-AML1 was predominantly observed, while t(4,11)MLL-AF4, t(1,19)E2A-PBX1, and t(9,22)BCR-ABL p190 were less frequently identified. Two forms of BCR-ABL p210 in CML were found: b3a2 and b2a2 transcripts, and the proportion of b3a2 was a little higher than that of b2a2. The majority of splice variants in PML-RARA, E2A-PBX1, and TEL-AML1 were PML-RARA L-form (50%), E2A-PBX1 (I) (66.7%), and TEL-AML1ex2 (94.4%), respectively. In addition, we also detected a novel form of PML-RARA, besides Lform and S-form transcripts, classified as V-form transcripts. The statistics on the frequency of different splice variants might provide more profound and additional molecular characteristics of the majority fusion transcripts styles. This would eventually allow a more precise clinical diagnosis and optimization of therapy.
The multiplex PCR-microarray assay could also be applied to detect rare rearrangements, such as PLZF-RARA, NPM1-RARA, and SIL-TAL1. For CBFB-MYH11 translocation, type A transcript accounted for more than 85% of the positive patients; two other transcripts (D and E) represented nearly 5% each, whereas all others represented unique cases [4]. We designed two reverse primers and MYH11 probes for type A and types D and E, respectively. It was indicated that there were other seven or novel transcripts though signal intensity of wild probes CBFB, MYH11-A, and MYH11-DE.