Dr Mensink Haematology, University Hospital Nijmegen, P.O. Box 9101, 6500HB Nijmegen, The Netherlands.
We used a recently developed system for real-time quantitative polymerase chain reaction (PCR) to determine residual disease in patients with chronic myeloid leukaemia. The expression of the Bcr-Abl hybrid oncogene was determined and normalized by using the PBGD housekeeping gene product as endogenous reference. The sensitivity and reproducibility of the assay was tested on cell line K562. A dilution of Bcr-Abl-positive cell line K562 remained positive at up to 250 fg of RNA. 10 copies of Bcr-Abl DNA in water could still be detected. The dynamic range of the method spanned six orders of magnitude. Analysis of 10 identical assays on K562 RNA resulted in a variation of 15%. To test the feasibility of normalization of Bcr-Abl dosage by the PBGD product, we compared the efficiencies of the RT-PCRs in 150 patient analyses. We concluded that PBGD was a suitable and stringent quality control standard. Three patients who were treated with donor leucocyte infusions for chronic myeloid leukaemia who had relapsed after bone marrow transplantation were followed over time. The normalized Bcr-Abl dosage was compared to the results of cytogenetics. Cytogenetic analysis was negative below a normalized Bcr-Abl dose of about 3 × 10−2. This semi-automated method is fast, sensitive and accurate and enables a high throughput of samples.
Molecular diagnostic techniques based on the amplification of leukaemia-specific mRNA sequences by reverse transcriptase polymerase chain reaction (RT-PCR) can effectively monitor the presence of leukaemic cells. In chronic myeloid leukaemia (CML) the presence of the hybrid oncogenic Bcr-Abl transcript enables detection of one malignant cell in a background of at least 105 normal cells ( Kawasaki et al, 1988 ). Patients in whom the presence of bcr-abl gene rearrangements was sustained had a higher probability of relapse ( Delage et al, 1991 ). The mere detection of residual leukaemic cells carrying the characteristic chromosomal translocation in individual patient samples is completely dependent on the sensitivity of the assay and is of limited value in predicting disease progression. Small numbers of leukaemic cells often persist in CML after allogeneic transplantation. Radich et al (1995 ) showed that a positive PCR assay of bcr-abl fusion transcripts 6–12 months post-BMT was an independent predictor of subsequent relapse, but PCR positivity at 3 months post-BMT was not. Lin et al (1996 ) demonstrated the importance of the kinetics of increasing numbers of BCR-ABL transcripts in predicting relapse after bone marrow transplantation.
Current methods for quantitation use competitive PCR assays in which a Bcr-Abl cDNA fragment of different size is used as an internal standard ( Gilliland et al, 1990 ; Lion et al, 1992 ; Thompson et al, 1992 ; Cross et al, 1993 ). These methods are very time-consuming and cannot be used in a routine setting. Reproducibility and variations in the final results are hard to assess and the use of competitor DNA molecules is a potential source of contamination.
We here describe a new method for the routine quantitative analysis of nucleic acids. We used a recently developed integrated system for thermal cycling, real-time fluorescence detection and subsequent analysis (Perkin Elmer-ABI PRISMTM 7700 Sequence Detector System). DNA amplification is detected in a closed tube, based upon a 5′ nuclease assay ( Holland et al, 1991 ; Livak et al, 1995 ) and no post-PCR sample handling is necessary, thus minimizing cross sample contamination. The instrument provides real-time quantitative information and enables a high throughput of patient samples ( Heid et al, 1996 ; Gibson et al, 1996 ).
In a large number of patient samples we checked normalization of Bcr-Abl dosage by the expression of the PBGD housekeeping gene. We determined normalized Bcr-Abl dosage in consecutive bone marrow samples from three patients who were given donor leucocyte infusions (DLI) for CML and who had relapsed after allogeneic bone marrow transplantation. In paired bone marrow samples we compared the outcome of quantitative Bcr-Abl RNA analysis with results of cytogenetics or in situ hybridization (ISH). Samples were taken before DLI and at several time points after DLI.
MATERIALS AND METHODS
Paired bone marrow samples were available from three patients at consecutive time points. Patients received donor leucocyte infusion (DLI) for relapse of Philadelphia chromosome positive CML after lymphocyte-depleted transplantation ( Bär et al, 1993 ; Schattenberg et al, 1997 ). UPN 51 and UPN 182 who had not achieved a haematological remission within 2 and 3 months after the first DLI respectively, were given a second DLI at 70 and 112 d after the first infusion.
A second set of 126 independent patient samples was used to test the normalization of the quantitative procedure as discussed below.
Cytogenetic analysis and in situ hybridization (ISH)
For the detection of the Ph translocation, bone marrow cells were harvested directly and/or after a 24 h culture period in RPMI 1640 medium (Gibco, Paisley, Scotland) using an IL3, G-CSF and GM-CSF growth factor mix and processed for GTG banding. As a standard, 32 or 20 metaphases were analysed ( Schattenberg et al, 1993 ). In the present analysis, results of ISH only are given where no GTG banding was performed: in UPN 132 at 1467 d after DLI and in UPN 182 at 49 and 401 d after the first DLI. In an ongoing study on chimaerism, ISH was performed on 400 cells in interphase using a combination of X- and Y-chromosome specific probes ( Lau, 1985). A signal that detected one to five cells inclusive out of 400 interphases was considered background. Accordingly, the percentage of Ph-chromosome positive cells was defined zero and the patient was considered to be in cytogenetic remission. 21/24 bone marrow samples (88%) were processed for GTG banding and three (12%) were analysed by ISH.
RNA isolation and cDNA synthesis
RNA was extracted from 106 cells using the guanidinium–thiocyanate acid–phenol–chloroform procedure with minor modifications ( Puissant & Houdebine, 1990). Briefly, 500 μl of guanidinium thiocyanate solution containing 100 μg/ml yeast tRNA (Boehringer Mannheim, Germany) as a carrier was added to the cells. After phenol/chloroform extraction and two isopropanol precipitations, the pellet was washed in 75% ethanol solution and RNA was dissolved in 50 μl of DEPC-treated distilled water for 10 min at 65°C. Subsequently the samples were stored at −80°C. To minimize quantitative differences due to decay of mRNA, samples were thawed immediately before further processing. One fifth of the total amount of RNA was used for cDNA synthesis.
The cDNA synthesis reaction was performed in a total volume of 20 μl, containing 50 m M Tris-HCl (pH 8.3), 75 m M KCl, 3 m M MgCl2, 10 m M DTT, 625 μM dNTPs, 5 μM random hexamers (Pharmacia, Uppsala, Sweden), 20 U RNAsin (Promega, Madison, Wis., U.S.A.) and 200 U Mo-MuLV reverse transcriptase (Life Technologies, Gaithersburg, Md., U.S.A.). To this, 10 μl of RNA or dH2O was added. The reaction was performed for 10 min at 20°C, followed by 42°C for 45 min and a 10 min incubation step at 95°C. Of the total RT volume of 20 μl, 1 μl was used for each PCR.
PCR conditions, controls
Using the Primer Express software program (Perkin-Elmer, Foster City, Calif., Demo version 1.0 ppd), we designed PCR primers for the amplification of cDNA derived from the Bcr-Abl transcript and PBGD transcript (see below). Bcr-Abl forward and reverse primers in 5′ to 3′ orientation: CGGGAGCAGCAGAAGAAGTGT and AAAGGTTGGGGTCATTTTCAC. PBGD forward and reverse primers CTGGTAACGGCAATGCGGCT and GCAGATGGCTCCGATGGTGA, respectively.
The TaqmanTM probes carried a 5′ TET reporter label and a 3′ TAMRA quencher group and were synthesized according to Lee et al (1993 ) by PE-Applied Biosystems (Warrington, U.K.). The following sequences were used as probes: Bcr-Abl probe 5′ (TET)-TCAGCGGCCAGTAGCATCTGACTT-(TAMRA)-3′ and the PBGD probe 5′-(TET)-CGAATCACTCTCATCTTTGGGCT-(TAMRA)-3′.
As positive controls for the PCR we used KW-3 construct, a plasmid containing the Bcr exon 3–Abl exon 2 breakpoint (kindly provided by Dr G. Grosveld, Memphis, U.S.A.) and cDNA derived from the t(9;22) positive cell line K562. Negative control cDNA was generated from RNA of cell line HL-60. The positive controls were used to optimize the conditions for the different PCRs. We used 1.25 U AmpliTaq GoldTM (Perkin Elmer/Applied Biosystems), 250 μM dNTP (Pharmacia, Uppsala, Sweden) and 15 pmol forward and reverse primer in a total reaction volume of 50 μl. The enzyme was activated by heating for 10 min at 95°C. We used a two-step PCR procedure, 90 s at 60°C and 30 s at 95°C for 60 cycles and used 6 m M MgCl2 and 160 n M probe for the Bcr-Abl PCR and 4 m M MgCl2 and 120 n M probe for the PBGD. We initially performed 60 cycles, but as samples never showed positivity at a cycle threshold higher than 40, the total number of cycles was limited to 45.
Real-time quantitation using TaqmanTM assay and threshold cycle
The PCR, TaqmanTM analysis and subsequent calculations were performed in the ABI/PrismTM 7700 Sequence Detector System (ABI/PE, Foster City, Calif., U.S.A.). Real-time quantitation was based upon the TaqmanTM assay and used a fluorogenic oligonucleotide probe labelled with both a fluorescent reporter dye and a quencher dye. In the intact TaqmanTM probe the 5′ fluorescent reporter dye was quenched by the 3′ quencher dye through Förster-type energy transfer (FRET) ( Förster, 1948; Lankowicz, 1989). Fluorigenic DNA probes (TaqmanTM probes), were hydrolysed during PCR upon hybridization to the template DNA by the 5′ secondary structure dependent nuclease activity of the Taq DNA polymerase ( Holland et al, 1991 , 1992). After hydrolysis the release of the reporter signal caused an increase in fluorescence intensity that was proportional to the accumulation of PCR product. The fluorescence intensity of the reporter label was normalized using the rhodamin derivative ROX as a passive reference label present in the buffer solution. The system generates a real-time amplification plot based upon the normalized fluorescence signal. Subsequently the threshold cycle (CT) was determined, i.e. the fractional cycle number at which the amount of amplified target reached a fixed threshold. This threshold was defined as 10 times the standard deviation of the baseline fluorescent signal, i.e. the normalized fluorescence signal of the first few PCR cycles. After reaching the threshold, the sample was designated positive. The CT was then used in the kinetic analysis and was proportional to the initial number of target copies in the sample ( Heid et al, 1996 ; Gibson et al, 1996 ; Higuchi et al, 1993 ). This relation was expressed as CT = A log(N) + B, in which ‘N’ is the starting quantity, ‘A’ is the slope of the curve and ‘B’ is the Y-intercept (i.e. the CT at starting quantity 1). The starting quantity in a patient sample was calculated after comparison of the CT of the sample with the CTs of a serial dilution of a positive control.
For the construction of standard curves of positive controls, RNA of the t(9;22)-positive cell line K562 was reverse transcribed into cDNA and serially diluted in six log steps into cDNA derived from cell line HL60 at a total amount of 50 ng or into water. This cDNA serial dilution was prepared once for all tests performed in this study and stored at −20°C. A standard curve that displayed the linear relation between the CT and the logarithm of the initial template concentration was established ( Heid et al, 1996 ; Gibson et al, 1996 ). The Bcr-Abl mRNA expression in patient samples was related to this standard curve and presented as the number of ngs of K562 RNA with the same level of expression. We used the expression of porphobillinogen dehydrogenase gene (PBGD) to normalize the Bcr-Abl expression. PBGD was used as an active and endogenous reference to correct for differences in the amount of total RNA added to a reaction and to compensate for different levels of inhibition during reverse transcription of RNA into cDNA and during PCR. With the PBGD primer set only cDNA derived from the PBGD housekeeping gene is amplified, not the tissue-specific transcript that is present in erythroid cells ( Chretien et al, 1988 ; Finke et al, 1993 ). PBGD expression was related to a standard curve derived from a serial dilution of K562 cDNA into dH2O. Also the PBGD quantity was expressed as the number of ngs of K562 RNA with the same level of expression. Subsequently, the normalized Bcr-Abl dose was defined as the expression level of Bcr-Abl divided by the PBGD expression level.
Quality standards: inclusion and exclusion criteria for the quantitation procedure
All standard dilutions and patient samples were tested in duplicate. For quantitation, the average value of both duplicates was used. Samples that had a more than a hundred-fold difference for both PBGD values were excluded from further analysis. Samples negative for PBGD but positive for Bcr-Abl were excluded from quantitation. When samples were positive for both PBGD duplicates and negative for both Bcr-Abl duplicates, the outcome was designated as ‘0’, i.e. below detection level. In all serial dilutions of K562-positive controls the regression coefficient of the standard curve that expresses CT as a function of the logarithm of the initial amount of template was at least 0.93.
Haematological remission was defined as the disappearance of all signs and symptoms and normalization of blood counts and bone marrow cellularity. The percentage of blast cells in the bone marrow had to be < 5%. Cytogenetic remission was defined as a haematological remission associated with disappearance of the Ph chromosome. If the Bcr-Abl normalized dose was zero, the patient was considered to be in molecular remission.
The RT-PCR assay
We determined the sensitivity of the TaqmanTM assay in PCR and RT-PCR. At least 10 copies of the Bcr-Abl DNA KW3 construct dissolved in water could be detected and a serial dilution of RNA derived from cell line K562 remained positive for Bcr-Abl to about 250 fg (not shown). The dynamic range of quantitation was six orders of magnitude. The sensitivity of the method was comparable to that of a conventional RT-PCR procedure we had used previously in which one malignant cell in 105 normal cells could be detected ( Bär et al, 1993 ). Below an amount of about 10 pg K562 RNA the reactions became negative for Bcr-Abl in about 50% of the cases.
We used a single mixture of reagents in 10 RT-PCR assays to investigate the reproduciblity of the assay. A cDNA amount derived from 25 ng of K562 RNA was used to calculate the variation in the CT. This resulted in average CT values of 21.56 (SD = 0.21) and 26.40 (SD = 0.23) for Bcr-Abl and PBGD tests respectively. The relationship between CT and starting quantity is expressed as CT =−3.0 log(N) + 25.4 and CT = −3.3 log(N) + 32.3 for the Bcr-Abl and PBGD standard curves. A standard deviation of 0.2 in the CT therefore reflected a 15% variation in the starting quantity expressed as the amount of K562 RNA in ngs.
The day-to-day variation was investigated on 10 different days. On each day a new mixture of reagents was composed and added to a reference sample containing 250 pg K562 RNA into 50 ng HL60 RNA and to a reference sample of 250 pg K562 RNA and 250 pg HL60 RNA (total 500 pg) into dH2O. The calculated result for 10 different days was 257 pg (SD 110) for the Bcr-Abl and 572 pg (SD 316) for the PBGD assay.
Use of PBGD expression for normalization
To investigate the efficiency of both Bcr-Abl and PBGD RT-PCR procedures, we used 50 ng of K562 RNA which had been serially diluted, as indicated in Fig 1, and subsequently determined the CT. These reactions were performed five times. The average difference in CT was calculated for both PCRs and found to be 4.7 (SD 0.2) from 50 000 pg to 16 pg (Fig 1). This constant difference in CT indicated that the efficiency of both RT-PCR procedures was the same within this range of template concentration.
We used PBGD as an endogenous reference for RNA quality and quantity. The expression of PBGD as a housekeeping gene was low in all tissues. It was therefore a very stringent standard for the quality of the RNA samples investigated ( Finke et al, 1993 ). PBGD expression was determined in the 24 samples of DLI patients and in a second set of 126 independent samples of CML patients. A total of 16% (n = 24) of the 150 patient samples tested was repeatedly negative for PBGD expression and were excluded from further analysis. In 4% (n = 6) of cases PBGD was negative whereas Bcr-Abl was positive, demonstrating the stringency of PBGD as a quality control. 46 of the 126 samples that were positive for PBGD had no detectable Bcr-Abl expression.
In our test we extracted RNA from 106 cells. Since 20% of RNA and 5% of cDNA mixture were used per reaction, the amount of RNA that was used in the assay therefore equalled the amount present in 104 cells. The mean of the PBGD value for 126 independent patient samples equalled PBGD expression as present in 2.3 ng K562 RNA. There was a 3 × 103-fold variation in the PBGD amount between patient samples derived from 104 cell equivalents (maximum 31, minimum 8 × 10−3, median 0.43). The differences in the PBGD expression between the different samples was the result of combined variation in the quality of the bone marrow sample, RNA extraction and storage, cDNA synthesis and the real-time PCR assay, as well as pipetting errors during the whole procedure. This demonstrated the need for a relative quantitation procedure and normalization of Bcr-Abl expression.
The quantitative assay in three patients
UPN 51 did not show any haematological response within 2 months after DLI and a second DLI was given. 92 d after the second DLI, the Ph chromosome had disappeared and the Bcr-Abl normalized dose had decreased ( Table I, Fig 2A). Bcr-Abl normalized dose was negative at 540 d after the second DLI. The patient relapsed molecularly at 1072 d after the second infusion. However, the Bcr-Abl normalized dose was low and gradually decreased below detection level.
Table 1. Table I. Comparison of cytogenetic analysis and ISH with BCR-ABL normalized dose.
UPN: unique patient number; DLI: donor lymphocyte infusion; (0)*: if the number of autologous cells in interphase (ISH) was ≤ 5, GTG banding was omitted and the percentage of Ph chromosome positive cells was defined as zero.
This is the first report regarding the measurement of Bcr-Abl dose by a real-time quantitative assay. In contrast to end-point quantitation, the use of threshold cycle (CT) for quantitation was accurate because it was established in the early exponential phase of the PCR where no competition for reagents exists and the inhibitory effects on the reaction are minimal. As demonstrated previously, a direct relationship exists between initial template concentration and CT ( Heid et al, 1996 ; Gibson et al, 1996 ; Higuchi et al, 1993 ). The new assay appeared to be sensitive and highly reproducible. The closed system minimized the risk of contamination since there was no need for post-PCR sample handling. Within the 96-well format a total of 32 tubes were used for duplicates of negative controls and serial dilutions of positive controls. Therefore quantitation of 16 different patient samples as duplicates for the Bcr-Abl target as well as the PBGD reference could be performed in one run. Compared to conventional quantitative PCR procedures the assay was fast, taking about 3 h ( Gilliland et al, 1990 ; Lion et al, 1992 ; Thompson et al, 1992 ; Cross et al, 1993 ). The numerical characteristics of the molecular diagnostic test, such as sensitivity and reproducibility or efficiency of different PCRs, can easily be provided. The PCR primers were located in exon BCR2 and ABL exon 2. This enabled detection and quantitation of the most common B2A2 and B3A2 fusion transcripts. To also detect the B1A2 splicing variant one additional B1 primer should be included.
We use a single cDNA preparation derived from patient RNA as the source for both target and reference PCR in duplicate. The efficiency of PBGD PCR assay and Bcr-Abl assay were shown to be similar and PBGD expression was regarded as a suitable endogenous reference for normalization of the Bcr-Abl dose. The PBGD housekeeping gene was demonstrated to be a very stringent quality control. Since we had a 16% drop-out of samples because of PBGD negativity, we are currently investigating other less stringent endogenous references that could be compared to PBGD. The use of a two-colour assay in which target and reference PCR are performed in one tube is expected to increase the reproducibility and throughput of the system. Reproducibility would be further increased by the use of one stock solution of reagent mixture to be added to patient samples in different assays on different days. If the amount of material allows, two independent RNA isolations and cDNA syntheses would further increase accuracy. The feasibility of these improvements is currently under investigation.
In the present analysis the quantitative assay was used in three patients treated with DLI for relapsed Philadelphia-positive CML. They attained cytogenetic remission and in all three cases the Bcr-Abl normalized dose had become zero at the end of follow-up. After DLI, Kolb et al (1995 ) confirmed complete haematological and molecular remission by the absence of Bcr-Abl RNA transcripts in 42/44 CML patients studied using PCR analysis. Collins et al (1997 ) did not demonstrate Bcr-Abl transcripts in all 23 CML patients in cytogenetic remission after DLI, who were also evaluated with PCR studies for Bcr-Abl rearrangement. In both studies no details were given for the methods used for PCR but were probably qualitative and not quantitative.
In the three patients described here, cytogenetic remission was attained at 49–97 d (median 92), and the first molecular remission was found at 401–732 d (median 540) after last DLI. This means that in these three patients the infused donor lymphocytes exerted an ongoing and long-lasting anti-leukaemia reaction that ultimately suppressed the Bcr-Abl breakpoint molecules under the detection limit of the sensitive real-time PCR analysis.
We present a sensitive and accurate semi-automated method for the quantitation of residual disease of CML and possibly for other malignancies beyond the clinical and cytogenetic level. The method enables detailed study of the biology of the disease. In particular the quantitative information, the increase and decrease of residual disease, i.e. the kinetics of the process, could be of value for monitoring of therapy and early intervention with a low-risk therapeutic approach. The importance of the kinetics of increasing BCR-ABL transcript numbers in predicting relapse in patients that were treated for CML by bone marrow transplantation has been demonstrated previously ( Lin et al, 1996 ). The three patients presented here all showed gradually decreasing transcript numbers after donor leucocyte infusions. Further studies will be necessary to determine threshold levels of Bcr-Abl normalized dose for the prediction of cytogenetic relapse and to study the influence of GVHD and immunosuppressive drugs on Bcr-Abl normalized dose.
We are indebted to Dr Chris Noordanus and colleagues of Perkin-Elmer Nederland and to Drs K. Livak, F. Goodsaid and J. Snider of PE/Applied Biosystems Foster City, California, for their technical support and to Drs T. Smetsers and J. Meijerink for fruitful discussions.