Detection of minimal residual disease in acute myeloid leukaemia: methodologies, clinical and biological significance

At the time of diagnosis, patients with acute leukaemia may have tumour burdens in excess of 1012 malignant cells (Hagenbeek Pratt et al, 1994). It would therefore appear logical to adopt a strategy aimed at achieving the maximum reduction, if not eradication, of residual leukaemia. In chronic myeloid leukaemia (CML), quantitation of MRD, by measuring the BCR-ABL transcript using polymerase chain reaction (PCR) methods has had a major impact on the management of relapsed disease following allografting (Cross et al, 1993). Similarly, although many questions still remain unanswered, studies of MRD in acute lymphoblastic leukaemia (ALL) and acute myeloid leukaemia (AML) have to date proved highly informative and clinically relevant. Such studies are also providing new insights into the biology of acute leukaemia and challenging our previously held concept of what constitutes a ‘cure’ in leukaemia. This review will focus on AML, describe the methodological advances in the detection of MRD, and discuss its clinical and biological significance. Four approaches which are currently used for the detection and monitoring of residual disease in patients with AML will be described. These include (1) conventional cytogenetics, (2) fluorescence in situ hybridization (FISH), (3) multiparameter flow cytometry, and (4) nucleic acid-based amplification (polymerase chain reaction, PCR). Routine karyotyping at diagnosis is essential for detecting clonal abnormalities such as numerical and structural changes specific to the leukaemic cells. It can be applied during remission to detect residual disease (Hart et al, 1971; Testa et al, 1979; Freireich et al, 1992). In early studies of patients with acute leukaemia the abnormal karyotype generally disappeared with the achievement of morphological remission but re-appeared with the onset of relapse (Hart et al, 1971). An additional benefit of this technique is its ability to detect the acquisition of new, prognostically important, clonal abnormalities with disease progression. Although an abnormal karyotype is present in some 70% of AML cases, its use as a marker of MRD is limited by its insensitivity (relatively few cells are examined in a labour-intensive procedure) and requirement for proliferation (variable number and quality of metaphases) of the leukaemic cells (Secker-Walker, 1988). Thus, techniques have been developed to facilitate the detection of aneuploidy and structural rearrangements in both metaphase and interphase cells. FISH has proved in recent years to be a valuable technique for the identification of chromosomal abnormalities either on metaphase spreads or interphase cells (Le Beau, 1993; Martinet et al, 1997; Tanaka et al, 1997; Van-Lom et al, 1997). DNA probes used for FISH analysis are either specific for the chromosomal breakpoint regions or specific for whole chromosomes (chromosome paint). FISH probes are usually labelled with either biotin or digoxigenin, which can then be detected by antibodies conjugated to fluorescent dye. They can also be labelled with fluorescent molecules directly. For the detection of chromosomal translocations, two probes (one for each chromosomal area involved with the translocation) are used, labelled with different molecules to produce two different colour signals. The major advantages of FISH technology compared to conventional cytogenetics are its improved power to detect structural rearrangements, which may be missed with traditional banding techniques, and its ability to detect chromosome abnormalities in interphase cells. Although FISH can be applied to detect MRD in acute leukaemia, its sensitivity approaches only 1% (detection of one abnormal cell in 100 normal cells), because of the presence of aneuploid cells and other technical artefacts. However, in one small study of 22 AML patients, metaphase-FISH proved to be a useful and reliable technique for monitoring MRD where the clonal chromosomal abnormality was a trisomy, deletion or a translocation. Residual leukaemic cells were detected in 41% of patients. All patients who showed a persistent or increasing level of abnormal cells in two or more subsequent samples during remission, with one exception, relapsed (El-Rifai et al, 1997). Techniques to detect structural chromosomal abnormalities in interphase cells using locus-specific probes, have also been developed. One good example is the detection of the (9;22) translocation in CML using probes directed against the BCR and ABL genes (Tkachuk et al, 1990). Although interphase-FISH, which obviates the need for good-quality metaphase preparations, is a potentially useful technique for MRD detection, its sensitivity is limited by the presence of false positive cells, due to artefactual co-localization of two signals. Such a ‘background’ rate of detection may be observed in some 5% of normal lymphocytes (Gray et al, 1990; Lion, 1996). However, there have been attempts to improve the overall sensitivity of FISH methods, for example the application of the technique to specific cell populations e.g. FACS sorted CD34+ cells and more recently the development of a triple-probe three-colour FISH system to detect the BCR-ABL gene (Sinclair et al, 1997). The relative usefulness of FISH technology versus other assays in detecting and monitoring MRD in AML remains yet to be established in prospective studies involving large numbers of patients. Although in AML there are no leukaemia-specific antigens which can be detected by immunophenotypic methods, AML cells frequently display aberrant or uncommon phenotypes that allow their distinction from normal haemopoietic cells and therefore can be used as markers of MRD (Coustan-Smith et al, 1993; Reading et al, 1993; Campana Macedo et al, 1995). Initial studies of immunophenotypic detection of MRD were based on double antigen staining of leukaemic cells, analysed by fluorescent microscopy (Campana et al, 1990; Adriaansen et al, 1993). However, a more sophisticated approach, based on multiparameter flow cytometry to detect multiple antigens, has been successfully applied to detect MRD (Reading et al, 1993; San Miguel et al, 1997). The use of multiparameter/multidimensional flow cytometry enables simultaneous analysis of two light scatter and up to four fluorescent signals, and, by appropriate gating, analysis can be restricted to the blast cell population, thus excluding the majority of unwanted mononuclear cells. Using such an approach the frequency and distribution of aberrant antigen expression on AML cells has been described. Comparison with normal bone marrow cells reveals four patterns of aberrant expression: (a) expression of non-myeloid antigens, (b) asynchronous expression of myeloid antigens, (c) over-expression of myeloid antigens, and (d) absence of expression of myeloid antigens (Terstappen et al, 1992). Reading et al (1993) have also reported their immunophenotypic results of 272 patients with AML, using a panel of 22 antibodies. Unusual co-expressions were found in 85% of cases. Multiparameter fluorescence analysis enabled detection of unusual phenotypes in patients who were in remission, and in a small number of patients studied, increased levels of the unusual presentation phenotypes could predict relapse. Although studies of MRD in AML using flow cytometry methods are still limited, as compared to acute lymphoblastic leukaemia (Campana (2) quantitative amplification by (a) manual quantitation (such as competitive PCR/RT-PCR), (b) automated quantitation (such as real time PCR). PCR is a technique by which a DNA fragment can be amplified if the flanking sequences are known. If the breakpoints in the target lesion are clustered within a small area of the gene, then a PCR method can be developed to detect this lesion from genomic DNA. An example of this DNA-PCR is the detection of monoclonality in B-lineage lymphoproliferative diseases including childhood ALL, by the amplification of the complementarity determining region-III (CDR-III) (Potter et al, 1992). Amplification of DNA target sequences enables the detection of mutations, translocations and polymorphisms. On the other hand, if the breakpoints are spread over a large area of the gene, then such a lesion can be detected by the amplification of the target gene transcripts (mRNA), in which large areas of the gene (introns) are deleted. This is performed by RT-PCR. The first step in RT-PCR is the conversion of RNA sequences into complementary DNA (cDNA). This cDNA is then subjected to PCR amplification. Examples of RT-PCR in AML are those developed to amplify the PML-RARA, AML1-MTG8 (ETO) and CBFB-MYH11 fusion transcripts. Most of these RT-PCR methods are very sensitive and can detect one leukaemic cell in 104–106 normal cells and are therefore very useful in detecting and monitoring MRD. However, the sensitivity of RT-PCR procedures may be significantly reduced by RNA degradation and inefficiency in the reverse transcription of mRNA to the cDNA step. PCR/RT-PCR amplification can generate a very large number of copies of the amplified gene. Therefore extreme care must be taken to avoid contamination between samples which would lead to false positive results. It is essential to include suitable control (positive and negative) samples with each test. All samples must also be tested for the amplification of a normal gene, e.g. ABL and β-actin, to assess the suitability of isolated nucleic acids (DNA and RNA) for PCR amplification. Initial studies of MRD in AML relied on the RT-PCR assay being either positive or negative, the latter result being dependent also on the sensitivity level of the test. A negative RT-PCR result for a particular fusion transcript does not necessarily mean absence of residual disease, but more specifically an absence of disease that can be detected by the assay used. Failure to appreciate this point can lead to the erroneous interpretation that RT-PCR negativity for MRD implies eradication of a leukaemic clone. Undoubtedly, qualitative RT-PCR methods are still very useful in monitoring MRD in specific leukaemias, for example in acute promyelocytic leukaemia (APL) (see later section), but they do have inherent limitations and are less useful in assessing residual disease in AML with t(8;21) (see later section). For these reasons, estimating the level of MRD by quantitating target genes provides a better tool for monitoring MRD. Two of the best examples of the benefit of quantitative RT-PCR for monitoring MRD are in CML and AML with t(8;21) (Cross et al, 1993; Tobal 21) and AML with inv(16) (Cross, 1995; Tobal Evans et al, 1997) Different considerations are required when choosing suitable control genes for qualitative or quantitave RT-PCR. The choice of control genes for quantitative RT-PCR requires two major considerations: (1) transcript levels of the control gene must not be affected by the disease; (2) degradation rate of the control gene must be equal to that of the gene of interest. This is important for accurately assessing the level of amplifiable RNA present in a given sample, and in particular when samples are received from a number of centres or when the quantitation is carried out in different centres (Lion 21) AML, we have tested the suitability of AML1 and ABL genes as controls for the quantitative RT-PCR amplification of AML1-MTG8. The level of the AML1 gene was found to vary between samples at different phases of the leukaemia with t(8;21), unlike ABL which was not affected by t(8;21). The degradation rate of ABL was then compared with that of AML1-MTG8. Samples were left at room temperature for 0, 24 and 48 h, and the levels of ABL and AML1-MTG8 fusion gene transcripts were assessed. Our data show that the level of degradation of ABL transcripts (one log over 48 h) is equivalent to that of AML1-MTG8. These data showed that ABL is a suitable control gene for quantitation of AML1-MTG8 (Newton et al, 1997). The principle behind the real-time techniques is to estimate the level of PCR products as they accumulate rather than using the usual approach of estimating the level of the final products. The idea of estimating the level of PCR products as they accumulate was first devised and published by Higuchi et al (1992, 1993). This technique involved the addition of intercalator ethidium bromide in each PCR reaction, coupled with an adapted thermal cycler to irradiate the samples with UV light, and the detection of the resulting fluorescence with a computer-controlled cooled camera. By plotting the increase in fluorescence versus cycle number, the starting copy number of the target gene could be estimated. The main drawback of this methodology is that both specific and non-specific products generate indistinguishable fluorescent signals. Another approach devised for real-time PCR is the 5′ nuclease assay, which is based on the 5′ nuclease activity of Taq DNA polymerase. The use of this 5′ nuclease activity for the detection of amplified products was first demonstrated by Holland et al (1991). The usefulness of this approach was enhanced by the development of fluorogenic probes by Lee et al (1993). This made it possible to eliminate post-PCR processing for the analysis of probe degradation. The probes used are oligonucleotide with a reporter fluorescent dye attached at the 5′ end and a quencher dye at the 3′ end. While the probe is intact, the proximity of the quencher greatly reduces the fluorescence emitted by the reporter dye, due to Forster copy of the gene amplified by this method will cause the of a reporter are by the number of which the fluorescent is first detected to the of reporter molecules rather than the amount of PCR products a number of This approach to an and useful tool for the rapid and quantitation of nucleic acid sequences et al, 1996; et al, 1996). A number of real-time RT-PCR methods have been developed to quantitate genes which include and AML1-MTG8. of MRD using real-time PCR in both acute and chronic are currently in et al, et al, et al, et al, et al, 1998). most quantitative studies of MRD by RT-PCR assays have been performed with some using manual competitive RT-PCR. For these methods to be used by a number of it would be to sensitivity levels between the different and to out of these Although real-time RT-PCR studies for monitoring MRD in leukaemia are still in the this strategy may approach which will reduce and time used for the up of manual competitive RT-PCR manual may to produce real time RT-PCR (such as can produce results in and may be less to However, the of real-time to predict relapse would depend on these being based on sensitive PCR if the basis of the real-time PCR is a which the this approach will to detect small changes in the level of MRD and thus to predict relapse. In this we that an RT-PCR similar to the one we established for the amplification of AML1-MTG8 with a sensitivity of et al, 1991). As a two on chromosome and on chromosome are et al, et al, 1993). for were detected in patients with transcripts were present in only 70% of cases et al, 1992). the chromosome breakpoints in the of the gene, different genomic breakpoints may in the gene on chromosome to types of fusion transcripts, the and et al, 1995). The and of fusion transcripts The fusion provides a very useful molecular marker which can be used both for diagnosis and for detecting residual disease et al, 1993; et al, 1995). detection of in is highly correlated with of therapy with acid et al, 1993). is a of AML by a and chromosomal with and response to Furthermore, it is also very to chemotherapy and a relatively approaches the use of and chemotherapy, the majority of patients to with may remission and be of their disease et al, 1995; et al, 1996; et al, 1997; et al, 1997; et al, et al, 1998). However, relapse of leukaemia remains the major cause of treatment occurring in some of patients chemotherapy et al, 1998). disease is with a frequency of leukaemia and survival et al, 1992). of the use of RT-PCR for transcripts have demonstrated the of detecting MRD during clinical remission. This was first by et al and by several other et al (1993) showed of transcripts in patients who achieved complete remission with that therapy does not eradicate the leukaemic clone. Furthermore, patients on treatment with RT-PCR positive and In most patients who received chemotherapy RT-PCR negative et al, 1993). Using RT-PCR assays for the with a sensitivity level of it has been that PCR negative of chemotherapy are with remission, patients who remain or to PCR positivity were very likely to relapse within a short period of time. However, negative PCR were not with remission, as patients could still relapse negative, or PCR positive within of being negative et al, et al, et al, et al, et al, 1993; et al, 1993; et al, et al, et al, 1995; et al, 1996; et al, 1996; et al, 1997). et al reported the results of a prospective RT-PCR monitoring study in patients in the by the of and for induction by courses as level of the RT-PCR assay for was All patients were in remission and tested PCR negative at the end of patients PCR of relapsed at a time of from the first PCR positive test. of PCR within the first patients who tested PCR negative in two or more have Using a the relative risk of relapse for patients who to PCR positivity at any time during compared to patients who PCR negative, was et al, 1998). These results that conversion to PCR positivity for during remission is highly predictive of subsequent relapse and the need for molecular monitoring of MRD, during the early such a strategy may the an to patients in molecular relapse additional or treatment the onset of relapse. based on its the has recently its and treatment at the time of molecular relapse. therapy for such patients standard of for by chemotherapy and data that was very and could be given on an in the majority of cases a molecular remission was with a result which is only obtained when is given for leukaemic relapse et al, 1998). It is that the qualitative RT-PCR assay for with its relatively low has For in the some of patients with PCR negativity at the end of chemotherapy within et al, 1998). This that the current methods lack the sensitivity to detect low but significant levels of MRD during remission. It has been that there are technical with the RT-PCR assay for the transcript which may be to the quality or of resulting in a less sensitive compared to that which can be achieved for other fusion transcripts such as BCR-ABL where RT-PCR assays have a sensitivity level of et al, et al, 1993; et al, 1996). attempts to improve the sensitivity of the RT-PCR assay for have been made et al, 1996; et al, 1996; Tobal & Liu Yin, 1998). The RT-PCR assay for the transcript was to be more sensitive than that for by at 1 being able to detect one leukaemic cell in normal cells et al, Using this assay, we were able to the of transcripts in out of patients in remission of APL, following intensive chemotherapy or bone marrow et al, have a number of technical to increase the sensitivity of the transcript amplification by including the use of PCR & Liu Yin, 1998). The new method was some log more sensitive than the standard assays for and was able to detect one leukaemic cell in cells. have applied this new method to assess MRD in patients in to remission of and found patients to be & Liu Yin, 1998). These results our of the of cells the translocation in patients clinically of their disease. The of these is (see