An increasing number of risk-stratifying genetic lesions in acute leukemia are being discovered and characterized. To translate this important and increasing volume of information from the research laboratory into effective clinical care, however, new, fast and comprehensive assays are needed. Toward this end, we have developed a two-stage multiplexing assay of broad applicability, which combines multiplex polymerase chain reaction with multiplex detection on spectrally addressable liquid bead microarrays. Using pediatric lymphoblastic leukemia as a model system, we demonstrate that all seven of the fusion transcripts resulting from risk-stratifying chromosomal translocations can be assayed in a single well of a 96-well multiplate with 100% specificity and sensitivity, within 6 h of specimen collection. The assay is automatic and high throughput and represents a significant improvement over previously available assays targeting the same genetic changes. We conclude that user-defined assays that multiplex both target selection and detection may have broad applicability in the management of hematological malignancies.
With the human genome sequencing project nearing completion, an ever-increasing number of risk-stratifying genetic changes are being discovered and characterized. 1,2 Our ability to translate rapidly into effective clinical cancer care the important prognostic information these lesions convey, however, has lagged behind, largely because assays for individual genetic lesions are time consuming and labor intensive. Comprehensive and cost-effective assays are not currently available. We hypothesized that cost-effective assays with optimal clinical performance characteristics can be developed by multiplexing both target selection and target identification. To test this hypothesis, we selected a set of well-characterized risk-stratifying molecular lesions in pediatric acute lymphoblastic leukemia3 as model system and developed a user-defined, automatic high-throughput assay. The assay, which we have dubbed BARCODE-ALL (for bead array coded detection in acute lymphoplastic leukemia), is capable of detecting all of the selected targets with high specificity and sensitivity. Similar highly multiplexed assays can readily be developed for acute leukemia in adults. Widespread availability of these tests should expedite and improve the accuracy and sensitivity of detection of translocations in acute leukemia in children and adults.
Materials and methods
Patient samples, cell lines, nucleic acid extraction
Cells lines, oligonucleotide primers, and RT-polymerase chain reaction (PCR) conditions were as in Scurto et al4 with modification (see Table 1). The following cell lines were used as positive controls for multiplex RT-PCR reactions: K562 for CML-type BCR/ABLP210, SupB15 for ALL-type BCR/ABLP190, RS4; 11 for ALL1/AF4, 697 for E2A/PBX, and REH for TEL/AML1. SUD-HL6, a follicle center cell lymphoma cell line, was used as a source of RNA devoid of a fusion transcript (negative control). Samples from 60 cases of acute lymphoblastic leukemia from patients 20 years of age or younger, which had been cryopreserved in our laboratory, were also studied without prior knowledge of cytogenetic or clinical features. These studies were conducted with prior approval by the Human Subjects Committee of the Internal Review Board of UMMMC and in accordance with NIH guidelines on research on human subjects.
Total RNA was extracted from cell lines and cryopreserved patient samples using Trizol® (Invitrogen, Carlsbad, CA, USA) and followed the instructions provided by the manufacturer.
Multiplex RT-PCR reactions
Half a microgram of unfractionated RNA was reverse transcribed using gene-specific decamer primers hybridizing to a region distal to the ‘downstream’ primer used for amplification (see Table 1). Fusion transcript cDNAs were specifically amplified by PCR using primers designed by Scurto et al4 with some modifications (Table 1). Both reverse transcription and PCR reactions were carried out on a 96-well format on separate 9600 Perkin-Elmer thermal cyclers. The primer priming the synthesis of the strand hybridizing to the probe/bead conjugate carried a biotin group in its 5′ end. This biotin, after reaction with streptavidin–Phycoerythrin serves as a reporter for cDNA hybridized to cognate bead/probes.
To determine the analytical sensitivity of the assay, cell mixing experiments were carried out by diluting cells carrying the relevant genetic lesion into negative control cells prior to extraction of RNA or by mixing known amounts of RNA from cell lines bearing translocations with SUD-HL6 or normal peripheral blood RNA. In most experiments, we used RNA mixing rather than cell mixing since RNA mixing is easier to do, more reproducibly quantitative and hence more accurate. To determine the sensitivity of the assay in a clinically relevant manner, cells from patients diagnosed with acute lymphoblastic leukemia carrying TEL/AML1 or BCR/ABL transcripts were serially diluted into normal peripheral blood samples prior to cell isolation and RNA extraction. Use of clinical samples avoids over-representation of target RNA because of the well-known overabundance of RNA and hybrid transcripts in leukemic cells grown in culture as compared to bone marrow cells.5
Liquid bead microarray assembly, hybridization and target identification
Multiplex PCR is a powerful way to simultaneously interrogate multiple genetic loci in a single reaction. However, as the number of potential targets in the reaction increases, it becomes increasingly difficult to identify the amplified target based on size alone (gel or capillary electrophoresis). To circumvent these difficulties, we identified amplified PCR targets using solution hybridization to a liquid bead microarray (XMAP beads, Luminex corporation, Austin, TX, USA) composed of seven bead sets each carrying a specific capture probe hybridizing specifically to one of the leukemia-specific fusion hybrids. Oligonucleotide probes were covalently attached to XMAP beads using a heterobifunctional crosslinking reagent (EDC, Pierce). XMAP beads are polystyrene beads optically encoded with varying amounts of two spectrally distinguishable ‘classification’ red fluorescent dyes so that each bead set has a unique optical signature. The optical signature can be used to unambiguously identify the bead in a mixed bead set (ie a liquid bead microarray) (Figure 1c) using a high-throughput Luminex 100 analyzer (upper diagram in Figure 1d) (Luminex). Using a second laser, the Luminex 100 instrument also measures bead-associated reporter fluorescence, which is proportional to the amount of bead-hybridized PCR product. Figure 1e displays the results of the assay for a single patient with the diagnosis of chronic myelogenous leukemia. Note that multiplate well H1 contains seven different microarray beads only one of which (b3/a2) becomes fluorescent upon hybridization to RT-PCR- amplified fusion transcripts. Seven bead sets were used in the formulation of our master liquid bead microarray (Figure 1c). As formulated, one master array was sufficient to perform 400 assays (patients) and was stable at 4°C in the dark for at least 6 months.
After completion of the multiplex PCR reaction 10 μl of PCR product and 7 μl H2O were added to single wells (one well per patient) of a multiplate containing 33 μl of liquid bead microarray per well suspended in 1.5 × TMAC6 (1 × =2 M TMAC, 0.1% sarcosyl, 50 mM Tris, 4 mM EDTA, Figure 1b). TMAC at high concentrations can change base staking in such a way that it largely reduces the differences in the strength of hybridization between GC and AT pairs.6 Hybridizations were initiated by codenaturation of bead microarray/PCR product mixes at 99°C for 10 min and were carried out at 52°C for 30 min. After hybridization, samples were washed in 1 × TMAC, resuspended in 1 × TMAC containing 10 μl/ml of streptavidin–phycoerythrin (Molecular Probes, OR, USA), for 5 min (52°C) and automatically injected into the Luminex 100 analyzer (Figure 1d) using an XY mechanized platform. A Microsoft Excel® algorithm transformed raw numerical instrument data into graphic display for a large number of patient samples simultaneously in a matter of seconds (Figure 1e).
Results and discussion.
Using dextran-assisted cell separation and Trizol® RNA extraction, PCR-ready cDNA could be obtained from whole-blood or bone marrow specimens in less than 2 h time. Optimization of a previously published multiplex amplification protocol led us to PCR conditions that accomplished amplification in approximately 2 h. The liquid bead microarray was hybridized to denatured PCR products for 30 min (Figure 1b and c) and then analyzed in the Luminex 100 instrument (Figure 1d, upper panel). This instrument, similar in concept to a flow cytometer, can measure analyte-associated fluorescence – in our case, target PCR product hybridized to probes on beads – on a bead array up to 100 members deep. Using peripheral blood and marrow samples from patients with presumptive acute lymphoblastic leukemia, the entire procedure from sample arrival in the laboratory to analysis in the Luminex 100 instrument could be carried out in 6–7 h, well within the scope of a full workday in a clinical laboratory. Figure 2 depicts an agarose gel of target-specific amplification by multiplex RT-PCR (Figure 2a) and the specificity of the capture oligonucleotide probes under the hybridization conditions developed for the BARCODED-ALL assay (Figure 2b). Note that in the agarose gel lanes 1 and 5 contain two or more bands (Figure 2a). Also note that in filter hybridization experiments there is virtually no crosshybridization of probes to noncognate target cDNA under these conditions (Figure 2b) and that robust detection occurs over a 100-fold dilution of target PCR product (Figure 2b). Figure 2c depicts the raw data obtained with positive and negative control cell lines. After subtraction of bead-specific background fluorescence, the signal-to-noise ratio for five independent hybridizations was greater than 25 for all targets.
The analytical sensitivity of the assay was determined by analyzing serial 10-fold dilutions of target RNA into background (target negative) RNA. Analysis of these samples indicated that all targets could be reliably identified when they accounted for 1% or more of the input RNA (Figure 3). This level of sensitivity is more than adequate for the intended use of the assay, which is to be used as an ancillary test in the diagnosis and management of acute lymphoblastic leukemia in children at first presentation. Sensitivity tests were also conducted by diluting cells from clinical samples carrying BCR/ABL or TEL/AML1 fusions into normal peripheral blood prior to RNA extraction. Results similar to those shown in Figure 3 were obtained (results not shown).
The intra-assay reproducibility was determined by simultaneously assaying 50 coded RNA samples that were prepared by mixing equal quantities of RNA from translocation carrying cells and cells devoid of them, thus simulating an average leukemia sample at diagnosis. This group of samples was simultaneously assayed using the 96-well multiplate format for reverse transcription, amplification, hybridization and assay readout. The entire assay was carried out in approximately 8 h. Automated instrument analysis time for 58 samples (50 unknowns, five positive controls and three negative controls) was 25–30 min (∼30 s per sample) and occurred unattended. In this set of samples, BARCODE-ALL correctly identified four of four CML-type BCR/ABL fusions, four of four ALL-type BCR/ABL fusions, eight of eight ALL-1/AF4 fusions, eight of eight E2A/PBX fusions and nine of nine TEL/AML1 fusions (Figure 4). All 17 unknowns devoid of fusion transcripts were scored negative in the assay (Figure 4) indicating that the assay is highly specific and suitable for clinical use. Intra- and interassay reproducibility was high with intra- and interassay coefficients of variance of <10% (results not shown). A hybridization temperature of 52°C was optimal, but the useful temperature range was wide (±5°C, results not shown). This in fact is advantageous since it provides some flexibility during bead manipulation.
To determine the clinical performance of the BARCODE-ALL, we analyzed 60 samples from children with previously diagnosed acute lymphoblastic leukemia at our institution, without prior knowledge of the results of cytogenetic and immunophenotypic studies obtained at diagnosis. All samples were also studied by single target PCR, that is, by carrying out separate PCR reactions targeting the seven ALL-associated hybrid mRNA including four variants of BCR/ABL. All 60 samples were assayed simultaneously using the multiwell format for reverse transcription, PCR, hybridization and analysis. Under these conditions, the assay detected nine of nine TEL/AML1 fusions (cases 3, 11, 21, 26, 28, 32, 44, 45 and 51), seven of seven BCR/ABL fusion (4, 6, 7, 9, 15, 52 and 56), five of five ALL-1/AF4 fusions (23, 24, 29, 34 and 48) and one of one E2A/PBX fusion (54) (Figure 5a) in good agreement with the results of single target RT-PCR (Figure 5b). In all, 38 samples were negative on our assay (Figure 5). Of these, four had faint bands on the single target RT-PCR assay (Figure 5b samples 5, 8, 14 and 31). All four bands failed to hybridize with specific probes and were therefore spurious single target PCR false positives. Samples testing positive by BARCODE-ALL that showed weak bands on single target PCR (cases 7, 9, 34 and 45) were among a subset of cryopreserved samples that had low RNA yields after extraction. In all four of these samples, RNA concentration was below the minimum quantifiable level by spectrophotometry (GeneQuant, Pharmacia). This was most likely because of sample deterioration before cryopreservation.
In this study, we show that combining multiplex PCR with multiplex hybridization to liquid bead microarrays results in a robust assay platform that can effectively be used to detect large number of targets with high sensitivity and specificity. Multiplex PCR amplification is well understood and was first successfully used in 1988 to screen for mutations in Duchenne's muscular dystrophy.7 It can detect more translocations than cytogenetics8 and it can be applied to diverse PCR assays (for review see Ferrando and Look9, Edwards and Gibls10 and Henegatiu et al11) including acute leukemia.4,12 For multiplex detection, we have chosen liquid bead miroarrays because of their attractive cost/sample ratio (see below) and the flexibility of the format for devising new test or modifying existing ones. The use of liquid bead microarrays solves the conundrum of multiplex PCR, that is, the increasing difficulty in unambiguously identifying a specifically amplified target as the number of possible targets increases. Since liquid bead arrays are customizable and compatible with fast flow read-through systems, they compare favorably with solid-phase arrays for these type of assay. Since the favorable signal-to-noise ratio of our assay, interpretation is unambiguous, as demonstrated by our mixing and coding experiments (Figure 4) as well as the perfect correlation between single target PCR and BARCODE-ALL in the clinical samples analyzed (Figure 5). Furthermore, hybridization kinetics in bead microarrays more closely approaches solution hybridization kinetics permitting shorter hybridization times and faster test turn-around time than possible with solid-phase microarrays. Our assay is amenable to high-throughput configuration and automation, significant advantages when studying samples at referral laboratories. Data analysis is simple and automatic, further reducing total assay time.
Our assay is superior to most of the previously available assays for the same genetic lesions. BARCODE-ALL detects all targeted genetic lesions in a single reaction (single well of a 96-well multiplate), is significantly faster, more accurate and less labor intensive than most other PCR assays targeting the same chromosome translocations. For instance, BARCODE-ALL is significantly faster and less cumbersome than similar multiplexed assays that utilize Southern blot hybridization as the readout platform. This is an important advantage of BARCODE-ALL since the results of risk-stratifying molecular tests are often needed to optimize therapy. It eliminates the use of radioactive labels and, unlike Southern blot-based assays, it can easily be automated. In addition, when compared to solid-phase microarrays, which could conceivably be constructed as a readout platform for multiplex PCR assays of this type, it is an order of magnitude more cost effective. We have recently demonstrated that using the Luminex bead microarray it is possible to subtype unambiguously all 45 relevant human genital papillomaviruses in a single well of a 96-well multiplate using 45 distinct type-specific papillomavirus probes in the same reaction (Wallace and Pihan, communicated at the annual meeting of AMP, November 2002, Dallas, TX, USA). An assay targeting AML-specific hybrid mRNA that uses beads for the detection step has been recently reported.13 This assay shares some features with ours and supports the suitability of liquid bead arrays as the detection platform for PCR assays.
Although our assay does not intend to replace cytogenetic studies, which provide a wealth of additional information, it can be used as an ancillary test to conventional cytogenetics for the purpose of risk-stratification targeting-defined genetic lesions. Use of BARCODE-ALL would ensure that translocations that are difficult to see by conventional cytogenetics, such as t(12;21)(p13;q22), complex translocations involving several chromosomes, cytogenetically silent translocations (microinsertions) or cases in which karyotype studies failed or were not submitted would still be appropriately assigned to their respective risk group. Although the cytogenetics data in our study are too incomplete to allow for in-depth analysis, excluding t(12;21)(p13;q22), banded karyotypes did not detect two of six (33%) translocations that were detected by both single target PCR and BARCODE-ALL. This rate is in agreement with that reported in the literature for a large series of acute leukemia, in which 33% of AML cases and 47% of ALL cases had PCR-detectable translocations that were not present in the original karyotype obtained from the same diagnostic marrow aspiration.8 Clearly, larger studies are needed to establish the relative value of BARCODE-ALL in comparison with cytogenetics.
An attractive feature of our assay is its short turn-around time. It typically can be accomplished in 6 h. Short turn-around time is essential when the result of a test is going to influence an emergent therapeutic decision, such as is appropriate therapy of children with acute leukemia. Another important feature of our assay is its high-throughput capacity. Given that the bead array we used was seven members deep and that the analysis time for a set of 58 samples was approximately 30 min, the maximum theoretical throughput of the assay is about 5040 targets, or 720 patient samples per work day per instrument. This calculation assumes 6 h of instrument data collection time, seven translocation targets per patient per assay-well and sufficient laboratory resources to assemble this many tests. These values are several orders of magnitude greater than throughput values achievable with single target PCR or alternative hybridization detection methods. These characteristics make our BARCODE-ALL assay equally well suited to tertiary care facilities and centralized national referral laboratories such as those of oncology cooperative groups.
BARCODE is also suitable for the investigation of other genetic lesions in leukemia such as point mutations, tandem duplications and deletions. We have used it to detect these lesions in the Flt3 receptor gene (unpublished), which occur in a high proportion of adults with acute myelogenous leukemia.14,15,16 It is likely that additional, risk-stratifying genetic lesions will be defined in the near future and as their combinatorial effect in prognosis becomes apparent, it will be important for us to include these lesions in our risk-stratifying tests as well. Liquid bead microarrays stand as unique platforms to satisfy these needs as they can accommodate assays targeting nucleic acid or proteins, including post-translational modifications, equally well.
Transcriptome-wide expression profiling using solid-phase microarrays is a powerful technique to uncover gene expression patterns associated with certain characteristics of a tumor including histogenetic origin, clinical and biological features, prognosis and potentially, response to therapy.17,18,19,20,21 Recently, very important results have been reported by Downing's group at St Jude's Hospital indicating that expression profiling in pediatric ALL can separate groups of patients with vastly different relapse-free survival.22 What is more impressive, expression profiling could predict the development of secondary leukemia with 99% accuracy.22 Transcriptome-wide screens, however, are complex and expensive and at this time cannot be used in the routine care of patients with acute leukemia. Since the transcriptional signatures of leukemias and lymphomas, obtained by genome-wide transcriptional profiling, are always comprised of less than 100 genes, liquid bead microarrays are an attractive alternative to solid-phase microarrays for clinical, targeted transcriptional profiling. Indeed, the Luminex liquid bead microarray has been successfully used for targeted transcriptional profiling in yeast cells. Whether it can be employed in leukemia as well remains to be determined and is being actively pursued in our laboratory.
Introducing any ‘new’ assay in a cost-conscious environment needs careful consideration of cost. A single BARCODE-ALL assay (one patient) costs less than US$ 5.00 in consumables. The Luminex 100 instrument itself prizes competitively with most other instruments in the molecular laboratory. Much like the flow cytometer is to immunophenotyping, the Luminex 100 can be used to expedite, automate and quantify a large number of other labor-intensive molecular assays currently performed manually by many molecular diagnostic laboratories in USA and Europe.
We believe that the assay we describe simplifies and improves genetic testing for currently known risk-stratifying genetic lesions in pediatric acute lymphoblastic leukemia and could quickly be expanded as additional risk-stratifying lesions become established. Our assay platform can also be adapted for testing of adults with acute leukemia and is sufficiently flexible to be easily adapted to a number of other nucleic and protein assays useful in the management of patients with both neoplastic and non-neoplastic hematological diseases.
We have developed a nearly fully automated, sensitive and specific single tube assay capable of detecting all risk-stratifying translocations in pediatric acute lymphoblastic leukemia within 6 h of sample collection. The assay simplifies and significantly improves testing for these translocations and can be quickly adapted to additional risk-stratifying lesions as they become established. Our assay platform is cost effective and can also be adapted for testing of adults with acute leukemia. It is sufficiently flexible to be easily adapted to a number of other nucleic and protein assays useful in the management of patients with hematopoietic malignancies.
Rabbitts TH . Chromosomal translocations in human cancer. Nature 1994; 372: 143–149.
Solomon E, Borrow J, Goddard AD . Chromosome aberrations and cancer. Science 1991; 254: 1153–1160.
Smith M, Arthur D, Camitta B, Carroll AJ, Crist W, Gaynon P et al. Uniform approach to risk classification and treatment assignment for children with acute lymphoblastic leukemia. J Clin Oncol 1996; 14: 18–24.
Scurto P, Hsu Rocha M, Kane JR, Williams WK, Haney DM, Conn WP et al. A multiplex RT-PCR assay for the detection of chimeric transcripts encoded by the risk-stratifying translocations of pediatric acute lymphoblastic leukemia. Leukemia 1998; 12: 1994–2005.
Moravcova J, Nadvornikova-Muchova S, Brezinova J . Overproduction of BCR-ABL transcripts in human leukemic cell lines K562 and BV173. Eur J Haematol 2000; 64: 135–136.
Wood WI, Gitschier J, Lasky LA, Lawn RM . Base composition-independent hybridization in tetramethylammonium chloride: a method for oligonucleotide screening of highly complex gene libraries. Proc Natl Acad Sci USA 1985; 82: 1585–1588.
Chamberlain JS, Gibbs RA, Ranier JE, Nguyen PN, Caskey CT . Deletion screening of the Duchenne muscular dystrophy locus via multiplex DNA amplification. Nucleic Acids Res 1988; 16: 11141–11156.
Pallisgaard N, Hokland P, Riishoj DC, Pedersen B, Jorgensen P . Multiplex reverse transcription-polymerase chain reaction for simultaneous screening of 29 translocations and chromosomal aberrations in acute leukemia. Blood 1998; 92: 574–588.
Ferrando AA, Look AT . Clinical implications of recurring chromosomal and associated molecular abnormalities in acute lymphoblastic leukemia. Semin Hematol 2000; 37: 381–395.
Edwards MC, Gibbs RA . Multiplex PCR: advantages, development, and applications. PCR Methods Appl 1994; 3: S65–S75.
Henegariu O, Heerema NA, Dlouhy SR, Vance GH, Vogt PH . Multiplex PCR: critical parameters and step-by-step protocol. Biotechniques 1997; 23: 504–511.
Viehmann S, Borkhardt A, Lampert F, Harbott J . Multiplex PCR – a rapid screening method for detection of gene rearrangements in childhood acute lymphoblastic leukemia. Ann Hematol 1999; 78: 157–162.
Zhang QY, Garner K, Viswanatha DS . Rapid detection of leukemia-associated translocation fusion genes using a novel combined RT-PCR and flow cytometric method. Leukemia 2002; 16: 144–149.
Dash A, Gilliland DG . Molecular genetics of acute myeloid leukaemia. Baillieres Best Pract Res Clin Haematol 2001; 14: 49–64.
Abu-Duhier FM, Goodeve AC, Wilson GA, Gari MA, Peake IR, Rees DC et al. FLT3 internal tandem duplication mutations in adult acute myeloid leukaemia define a high-risk group. Br J Haematol 2000; 111: 190–195.
Griffin JD . Point mutations in the FLT3 gene in AML. Blood 2001; 97: 2193A.
Alizadeh AA, Eisen MB, Davis RE, Ma C, Lossos IS, Rosenwald A et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 2000; 403: 503–511.
Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov JP et al. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science 1999; 286: 531–537.
Khan J, Wei JS, Ringner M, Saal LH, Ladanyi M, Westermann F et al. Classification and diagnostic prediction of cancers using gene expression profiling and artificial neural networks. Nat Med 2001; 7: 673–679.
Perou CM, Sorlie T, Eisen MB, van de Rijn M, Jeffrey SS, Rees CA et al. Molecular portraits of human breast tumours. Nature 2000; 406: 747–752.
Virtaneva K, Wright FA, Tanner SM, Yuan B, Lemon WJ, Caligiuri MA et al. Expression profiling reveals fundamental biological differences in acute myeloid leukemia with isolated trisomy 8 and normal cytogenetics. Proc Natl Acad Sci U S A 2001; 98: 1124–1129.
Yeoh EJ, Ross ME, Shurtleff SA, Williams WK, Patel D, Mahfouz R et al. Classification, subtype discovery, and prediction of outcome in pediatric acute lymphoblastic leukemia by gene expression profiling. Cancer Cell 2002; 1: 133–143.
About this article
Cite this article
Wallace, J., Zhou, Y., Usmani, G. et al. BARCODE-ALL: accelerated and cost-effective genetic risk stratification in acute leukemia using spectrally addressable liquid bead microarrays. Leukemia 17, 1404–1410 (2003). https://doi.org/10.1038/sj.leu.2402985
- lymphoblastic leukemia
- risk stratification