Chronic myelogenous leukemia: Laboratory diagnosis and monitoring

Abstract

Rapid developments have occurred both in laboratory medicine and in therapeutic interventions for the management of patients with chronic myelogenous leukemia (CML). With a wide array of laboratory tests available, selecting the appropriate test for a specific diagnostic or therapeutic setting has become increasingly difficult. In this review, we first discuss, from the point of view of laboratory medicine, the advantages and disadvantages of several commonly used laboratory assays, including cytogenetics, fluorescence in situ hybridization (FISH), and qualitative and quantitative reverse transcriptase-polymerase chain reaction (RT-PCR). We then discuss, from the point of view of clinical care, the test(s) of choice for the most common clinical scenarios, including diagnosis and monitoring of the therapeutic response and minimal residual disease in patients treated with different therapies. The purpose of this review is to help clinicians and laboratory physicians select appropriate tests for the diagnosis and monitoring of CML, with the ultimate goal of improving the cost-effective usage of clinical laboratories and improving patient care. © 2001 Wiley-Liss, Inc.

INTRODUCTION

Chronic myelogenous leukemia (CML) is a myeloproliferative disorder resulting from clonal expansion of the progeny of a malignant hematopoietic stem cell. It affects 1–2 people in every 100,000 population, with a male to female ratio of 1.3–1.4 to 1. CML comprises about 15–20% of adult leukemias, and its incidence increases with age (Faderl et al., 1999a; Sawyers, 1999).

CML is one of the most intensively and extensively studied human neoplasms. At the cytogenetic level, the Philadelphia chromosome (Ph), a hallmark of CML, was first discovered in 1960 (Nowell and Hungerford, 1960). Ph is a shortened derivative of chromosome 22 as a result of a reciprocal translocation between chromosomes 9 and 22 (Rowley, 1973). At the molecular level, this translocation fuses the 5′ segment of the BCR gene on chromosome 22 with the 3′ segment of the ABL gene on chromosome 9 (Groffen et al., 1984).

Traditionally, CML is diagnosed by the detection of cells with the Ph chromosome in patients with typical clinical and hematologic findings of a myeloproliferative disorder. Patients are subsequently monitored for response to therapy or disease progression by cytogenetic quantification of the percentage of Ph-positive metaphase cells in the bone marrow. In the past decade, many new technologies have been developed to replace the time-consuming and labor-intensive cytogenetic analysis. These include techniques detecting the BCR-ABL fusion at the DNA level with genomic polymerase chain reaction (PCR), Southern blot analysis, fluorescence in situ hybridization (FISH), or chromosomal painting; at the RNA level, with Northern blot analysis or reverse transcriptase-polymerase chain reaction (RT-PCR); and at the protein level, with Western blot analysis or immunoprecipitation (Hess, 1998).

Current therapies for CML include conventional chemotherapy with hydroxyurea or busulfan, interferon alpha (IFNα) alone or in combination with other chemotherapeutic drugs, and bone marrow or stem cell transplantation. Recently, STI 571 (Glivec®), a compound that inhibits the tyrosine kinase activity of BCR-ABL, has shown promising results in Phase I and II clinical trials on patients with chronic-phase CML refractory to interferon treatment. The early success of this class of compounds suggests that they are likely to become an important component of CML therapy. With many different choices of tests and therapies, the issue becomes more and more complicated as to what is (are) the best test(s) to order in a particular clinical context: diagnosis, monitoring response to a particular therapy, monitoring minimal residual disease, and monitoring disease progression.

To determine the appropriate laboratory test(s), clinical utility, test parameters such as sensitivity, specificity, ability to quantitate, ease of performing the test, as well as cost of the test need to be considered. In this review, we compare the advantages and limitations of currently widely used tests for the diagnosis and monitoring of CML. We then discuss the utility of each test in the diagnosis and monitoring of different therapies for CML. Recommendations are made based on data from the current literature.

MOLECULAR PATHOLOGY OF CML

As noted above, the t(9;22) translocation creates a fusion between the 5′ segment of the BCR gene on chromosome 22 and the 3′ segment of the ABL gene on chromosome 9 (Fig. 1; reviewed in Melo, 1996). Breakpoints in ABL occur in the first and second introns. Breakpoints in BCR occur mainly in one of three regions: the major breakpoint cluster region (M-bcr), the minor breakpoint cluster region (m-bcr), and the micro-breakpoint cluster region (μ-bcr). The corresponding fusion mRNA joins exon a2 from the ABL gene to either exon b2 or b3 when BCR is broken at M-bcr (called b2a2 or b3a2 transcripts), or to exon e1 when m-bcr is involved (called e1a2 transcript), or to exon e19 when BCR is broken at μ-bcr (called e19a2). Different fusion mRNAs translate into chimeric proteins of different sizes: both b2a2 and b3a2 produce p210, e1a2 produces p190, and e19a2 produces p230. Fusion types b3a2 and b2a2 account for the vast majority of CML cases. High levels of the e1a2 fusion transcript are seen primarily in Ph-positive acute lymphoblastic leukemia, the blastic phase of CML, and very rarely in the chronic phase of CML (Hermans et al., 1987). The few cases of chronic-phase CML in which the e1a2 fusion transcript predominates are associated with a prominent monocytic component (Kirk et al., 1996; Melo, 1996; Roumier et al., 1999). Recently, low levels of the e1a2 fusion were identified in addition to b2a2/b3a3 transcripts in the majority of CML patients in chronic phase (Saglio et al., 1996; van Rhee et al., 1996; Lichty et al., 1998). These e1a2 transcripts are thought to arise from alternative splicing of the b2a2/b3a3 fusion RNA. Although these e1a2 transcripts were hypothesized to be clinically significant, evidence to date does not favor either a clinical or a pathologic association. Fusion e19a2 is rare and is primarily associated with neutrophilic CML, a myeloproliferative disorder in which the large majority of circulating myeloid cells are mature granulocytes (Pane et al., 1996). Nevertheless, recent reports have also linked the e19a2 fusion with CML in accelerated phase and blast crisis (Haskovec et al., 1998; Kojima et al., 1999; Polak et al., 2000).

Figure 1. Structures of the BCR-ABL gene and BCR-ABL mRNA fusion products. Exons and introns are depicted as black boxes and horizontal lines, respectively. Brackets under the gene structures represent the breakpoint regions. Schematic representation of the fusion transcripts is shown on the lower half of the figure, and the protein products made are indicated.

Fusion of BCR to ABL results in constitutive activation of the ABL kinase activity. Dysregulation of multiple pathways, including JAK/STAT, RAS/RAF, JUN kinase, MYC, and phosphatidylinositol-3 kinase/AKT pathways, have been implicated in causing the excessive proliferation, resistance to apoptosis, and adhesion defects that are characteristic of BCR-ABL-transformed CML cells. Readers are referred to other comprehensive reviews for the molecular mechanisms of CML leukemogenesis (Ghaffari et al., 1999; Warmuth et al., 1999).

LABORATORY TESTS FOR CML

In addition to morphological assessment of bone marrow (BM) and peripheral blood (PB), many ancillary techniques have been developed for detecting and quantitating the number of translocation-bearing cells or fusion molecules for more precise monitoring. Among these, chromosomal painting shows promise because of its ability to detect complex chromosomal alterations not easily identified by conventional G-banding analysis, such as variant translocations (Yehuda et al., 1999; Harrison et al., 2000) and complex chromosomal abnormalities during blast transformation (Gribble et al., 1999; Su et al., 1999). To date, application of this relatively new technique has been limited by high cost. Southern blot analysis is seldom used at present due to its lengthy procedure and slow turnaround time as compared to newer molecular techniques such as RT-PCR. Among many techniques, cytogenetics, FISH, and RT-PCR have been well studied for their clinical utility and are currently the most widely used tests in clinical laboratories. This section compares the advantages and disadvantages of each of these techniques (Table 1) to provide a framework for selecting the proper test(s) for a diagnostic or a particular monitoring purpose.

Table 1. Assay Comparison

	Karyotyping	Interphase FISH	RT-PCR
dx, diagnosis; dz, disease; TAT, turnaround-time; MRD, minimal residual disease; BM, bone marrow; PB, peripheral blood.
Sensitivity for diagnosis	Least sensitive (∼90%) because limited number of cells routinely examined, ‘masked’ translocations	Most sensitive (∼95%)	Slightly lower sensitivity than FISH (∼95%) for diagnosis due to variant breakpoints not detectable by PCR primers
Sensitivity for MRD detection	1/25–30 cells	1/200–500 cells	1/10^4–6 cells, most sensitive in detecting low leukemic burden
Quantification	≥ ∼10%	≥ ∼0.5%	>1/10^4–6 with quantitative RT-PCR
Specificity	Highest	High	Lower specificity because of cross-contamination, low level of BCR-ABL in ‘healthy’ individuals
Dividing cells	Required	Not required	Not required
Specimen requirement	BM	PB usable	PB usable
Reasons for tech. failure	Lack of dividing cells	Loss of cells during processing	Degraded RNA
	Over- or underdigestion	Nonamplifiable cDNA
Routine TAT	72 hr	48 hr	24 hr
Unique advantage	Reveal additional karyotypic abnormalities other than Ph at dx. and during dz progression	Determine the breakpoints of the fusion genes

Cytogenetics

Traditionally, cytogenetics has been used to identify t(9;22)(q34;q11) and quantitate the percentage of Ph-positive cells after therapy. Cytogenetic analysis detects the Ph chromosome in approximately 90% of patients with CML at the time of diagnosis. Of the remaining 10% of CML cases, half carry “masked” translocations that can be detected only by molecular techniques such as FISH or RT-PCR for BCR-ABL (Cortes et al., 1995; Aurich et al., 1998). Of the remaining Ph-negative/BCR-ABL-negative cases, some carry translocations involving ABL and other translocation partners (Andreasson et al., 1997), some are Ph-negative/BCR-ABL-negative CML cases with atypical clinical features (reviewed in Melo, 1997), and some are CML cases with typical clinical and morphological features (Kurzrock et al., 1990; Selleri et al., 1990).

Typically, 25–30 metaphase cells are examined by cytogenetic analysis because of the time and labor involved. This limitation determines that the sensitivity of cytogenetics in detection of residual disease posttreatment is only 1 leukemic cell in 25–30 normal cells. In other words, the sensitivity of detection is 3–4% Ph-positive cells. The ability to quantitate makes cytogenetics a useful test for monitoring therapeutic response; however, because of the small number of cells being analyzed, the accuracy of quantification is poor, especially when Ph+ cells constitute less than 10% of the total.

In spite of its low sensitivity for detecting residual disease, cytogenetics is currently the most specific test, even in comparison to newer molecular techniques. It can detect both the Ph chromosome and the derivative of chromosome 9 (9q+) or other derivatives of variant translocations involving chromosomes 9 and/or 22. These cytogenetic abnormalities, in association with typical clinical and hematologic findings, are diagnostic of CML.

Because cytogenetics requires dividing cells for analysis of metaphase chromosomes, the best specimen for analysis is bone marrow, which contains more proliferating cells than does blood. The routine turnaround time on a nonrush basis for cytogenetics is about 3 days. Its cost, considering only reagents and labor, at our institution is approximately $125 per sample (Table 2).

Table 2. Cost Analysis of Different Ph/BCR-ABL Assays

	Karyotyping	Interphase FISH	RT-PCR
Hours of labor	4–5	2.5	3
Cost of labor based on $20/hour	80–100	50	60
Reagent/supply cost	16	45	34
Total cost	$116–136	$95	$94

A unique advantage of cytogenetics is that it may reveal, at the time of diagnosis, other cytogenetic abnormalities that are supportive of myeloproliferative disorders, including monosomies and partial deletions. Also, cytogenetics may reveal karyotypic abnormalities in addition to the Ph that arise during disease progression to accelerated phase or blast crisis. Such additional karyotypic changes often appear shortly before blastic transformation of CML (see below).

FISH

FISH makes use of differently labeled fluorescent DNA probes. In the first-generation FISH technique, two probes are utilized (Fig. 2, top panel; Tkachuk et al., 1990). One probe, specific for ABL, labeled orange, for example, hybridizes to the 3′ end of the ABL breakpoint region. The other probe, specific for BCR, labeled green, for example, hybridizes to the 5′ end of the BCR breakpoint. In BCR-ABL translocations, the 3′ portion of ABL joins the 5′ end of BCR, the orange signal overlies the green signal, and a yellow fusion signal is generated. This technique suffers from low specificity because of the random superimposition of orange and green in normal interphase nuclei (Fig. 2, bottom panel). This leads to false-positive results that severely limit the use of first-generation FISH for detection of minimal residual disease. The frequency of false positivity can be 3–10%, making quantification below 10% unreliable (Dewald et al., 1993; Garcia-Isidoro et al., 1997; Werner et al., 1997).

Figure 2. Schematic diagram of FISH analysis. Metaphase chromosomes and interphase nuclei are illustrated. Number 9 chromosomes are shaded, number 22 chromosomes are unshaded. ABL and BCR probes are represented by white and black dots, respectively. The top panel illustrates the first-generation FISH, and the middle panel illustrates one of the modified FISH analyses. The bottom panel depicts a normal metaphase and a normal interphase nucleus with overlapped ABL and BCR signals that cannot be distinguished from a BCR-ABL-positive nucleus by first-generation FISH.

Recent modifications, however, have greatly improved the specificity of FISH. In one of these technical modifications, two ABL probes are employed (Fig. 2, middle panel). One hybridizes to the 5′ side and the other to the 3′ side of the usual breakpoints in ABL. In normal cells, these two orange signals are juxtaposed, giving rise to one large orange signal. In the BCR-ABL translocation, these two orange signals are split. The 3′ orange fuses with the green on chromosome 22, generating a yellow signal, and the 5′ orange probe remains hybridized to chromosome 9. Therefore, in addition to the signals from normal homologs, a truly positive cell carries both a yellow and an orange signal, and a cell with random superimposition carries only a yellow signal. This modification reduces the lower limit of quantification from 9–10% to below 0.5% (Bentz et al., 1994). Other modifications of the FISH technique with similar specificity have been reported (Seong et al., 1994, 1995; Sinclair et al., 1997; Buno et al., 1998; Dewald et al., 1998; Grand et al., 1998).

FISH detects BCR-ABL in about 95% of CML cases. It is the most sensitive test for diagnosis because it detects the approximately 5% of cases with “masked” translocations that are missed by cytogenetics (Calabrese et al., 1994; Nacheva et al., 1994), and it also detects rare cases with variant breakpoints falling outside the regions covered by PCR primers (van der Plas et al., 1991; Iwata et al., 1994; Paldi-Harris et al., 1994; Hochhaus et al., 1996b; Estop et al., 1997). In addition, a FISH study routinely analyses 200 to 500 nuclei; thus, quantification generated by FISH is more accurate than cytogenetics, especially when few leukemic cells are present, as is frequently seen posttherapy. In one study correlating cytogenetics and FISH, FISH detected 2.5–8% BCR-ABL-positive cells in seven of nine specimens, from six patients, in which cytogenetic results were negative (Cuneo et al., 1998). Because of the added accuracy and sensitivity, FISH is being used increasingly to replace cytogenetics in monitoring of patients on IFNα and newer biological or chemotherapies (Buno et al., 1998; Duba et al., 1999).

FISH has several advantages over cytogenetics. The specificity of the newer split signal assay is high. Also, unlike cytogenetics, which requires dividing metaphase cells, FISH can be performed on interphase nuclei in peripheral blood. It therefore may bypass the requirement for a bone marrow specimen. However, the percentage of BCR-ABL-positive nuclei determined by FISH using peripheral blood specimens seems to be lower than that using bone marrow (Buno et al., 1998; Yanagi et al., 1999).

Occasional technical failures of FISH studies can result from loss of cells during processing. Over- or underdigestion of proteins to uncover DNA may give rise to signals that are difficult to interpret. Routine turnaround time is about 2 days, and the cost of reagents and labor at our institution is approximately $100 (Table 2).

Qualitative RT-PCR

Routine RT-PCR (not multiplexed) has a slightly lower sensitivity than FISH for the diagnosis of CML. It fails to detect rare cases in which variant breakpoints fall outside the region covered by PCR primers (Saglio et al., 1990; van der Plas et al., 1991; Iwata et al., 1994; Paldi-Haris et al., 1994; Hochhaus et al., 1996b; Estop et al., 1997). For minimal residual disease detection, it has been believed that RT-PCR is the most sensitive test, with its ability to detect 1 leukemic cell bearing the BCR-ABL message in 10⁴–10⁶ cells. However, this belief has recently been challenged. In a study of 21 patients who were in complete cytogenetic response after either IFNα or BMT therapy, FISH detected 1–12% BCR-ABL fusion genes in all of these patients, whereas RT-PCR detected fusion mRNA transcripts in only six of them, suggesting the existence of transcriptionally silent BCR-ABL cells (Brizard et al., 1998; Chomel et al., 2000). Whether these cells are clonogenic and clinically significant remains to be seen. Moreover, this result needs to be verified in further studies.

RT-PCR has a lower specificity compared to either cytogenetics or FISH. One reason is that cross-contamination of a negative specimen by a positive specimen may inadvertently occur and may be amplified during the PCR reaction. In addition, low levels of BCR-ABL transcripts have been identified in the blood of up to two-thirds of normal individuals, albeit using assays that are 1 to 2 orders of magnitude more sensitive than those used for routine diagnostic tests (Biernaux et al., 1995; Bose et al., 1998). This potential for false-positive results limits the utility of RT-PCR as a stand-alone diagnostic test for CML. Techniques for quantitative RT-PCR detection of BCR-ABL are available that circumvent some of these problems. The pros and cons of the techniques will be discussed further in the next section.

Like FISH, RT-PCR does not require dividing cells, so that PB can be utilized. Qualitative results obtained from PB have a good concordance with those from BM (Radich et al., 1995; Verschraegen et al., 1995). Occasional technical failures can result from RNA degradation. The routine turnaround time for RT-PCR is about 1 working day. The cost of reagents and labor is approximately $100 (Table 2).

A unique advantage of qualitative RT-PCR is its ability to determine specifically the types of fusion genes encoding either the p190 or the p210 protein. Chimeric protein p190 is reported to have a higher tyrosine kinase activity compared to p210 (Lugo et al., 1990). However, in practice, whether CML cases expressing predominantly e1a2/p190 have a more aggressive behavior and worse prognosis has not been definitively determined.

Determination of the type of fusion prior to allogeneic bone marrow or stem cell transplant is necessary for following any minimal residual leukemic clone after the transplant. Although there have not been any documented cases, a healthy donor might theoretically contribute some “benign” cells carrying BCR-ABL to the recipient (discussed below). For these reasons, RT-PCR should be used at the time of diagnosis to determine the fusion type.

Quantitative RT-PCR

Quantitative RT-PCR (Q-RT-PCR) has advantages and disadvantages similar to qualitative RT-PCR compared to cytogenetics and FISH, but has the obvious advantage of assessing the amount of BCR-ABL message. Several methods for PCR quantification are available, including both commercial “real-time” and noncommercial RT-PCR assays developed by individual laboratories.

In a typical noncommercial Q-RT-PCR assay, for which many variations exist (Lion et al., 1992; Malinge et al., 1992; Thompson et al., 1992; Cross et al., 1993), cDNA of an internal reference, such as β2 microglobulin or ABL, is co-amplified in the same tube as cDNA of BCR-ABL to control for variations in the procedures such as sample preparation and loading. PCR products are then analyzed by gel electrophoresis and densitometry. The amount of BCR-ABL is quantified by 1) normalization of the intensity of the BCR-ABL band to that of the internal reference to generate a ratio, and/or 2) comparison to a quantification standard curve to generate BCR-ABL in copy numbers/μg of total RNA. Quantification standards can be generated in two ways: 1) serial dilutions of a known quantity of a cloned plasmid containing the BCR-ABL fusion gene into normal DNA (Branford et al., 1999), or 2) serial dilutions of a known number of leukemia cells from K562 (b3a2) or BV173 (b2a2) cell lines into normal cells, followed by RNA preparation from the mixture of cells. Currently there is no consensus on which serves as the better standard. However, it is a legitimate concern that cell lines may vary in levels of BCR-ABL expression due to variation in culture conditions, which may lead to inaccurate results.

Noncommercial assays do not entail use of any special equipment. However, these assays are labor-intensive, time-consuming, and technically demanding. Numerous manual steps lead to poor intra- and interassay reproducibility. Moreover, the assays are difficult to standardize across laboratories because of the choice of different reference genes, reagents, primers, and reaction conditions for the reverse transcription reaction and PCR. As a result, the quantity of BCR-ABL transcripts measured at one laboratory cannot be compared to results from another laboratory.

Several commercial techniques have been developed in the last few years for more rapid, sensitive, and reliable Q-RT-PCR approaches. The major difference in methodologies between commercial techniques and noncommercial Q-RT-PCR lies in how PCR products are detected. Commercial techniques detect PCR products by measurement of fluorescence generated from stoichiometric binding of either double-strand DNA (dsDNA) dyes or fluorogenic probes (Branford et al., 1999; Eder et al., 1999; Emig et al., 1999; Kreuzer et al., 1999; Barbany et al., 2000).

DNA-binding dyes have enhanced fluorescence when bound to dsDNA. SYBR green I is preferred to ethidium bromide because SYBR green I is more sensitive and specific for dsDNA binding (Fig. 3). The fluorescence intensity increases in real time with the amount of the dsDNA synthesized during the ongoing PCR reaction. The cycle (C_T) at which fluorescence increases above the threshold signal correlates with the initial template copy number (Fig. 4). The advantage of this method is its relatively low cost because it does not require generation of sequence-specific probes. However, lack of a sequence-specific probe results in a relatively low specificity, because fluorescence is generated whenever dsDNA sequences are present, even though they may not be the specific PCR products. The specificity can be increased by performing additional analysis of product melting curves, which are unique to specific PCR products and can be used for identification.

Figure 3. Double-strand DNA binding dye. Double-strand DNA is represented by a ladder. Dyes are depicted as hourglass shapes. Light waves are represented by arrows. Free dyes have little fluorescence, and dyes bound to dsDNA have enhanced fluorescence.

Figure 4. Real-time PCR amplification plot. The increase in fluorescence signal is proportional to the increase in PCR products generated during the cycling process. The cycle (C_T) at which fluorescence increases above the threshold signal depends on the initial concentration of the template. In the diagram, sample 1 has a lower DNA concentration than sample 2.

Several chemical designs of sequence-specific probes are available, including a TaqMan single exonuclease hydrolysis probe (Perkin-Elmer Applied Biosystems, Foster City, CA), LightCycler dual hybridization probes (Roche Molecular Biochemicals, Indianapolis, IN), and the Molecular Beacons single hairpin probe (Stratagene, La Jolla, CA). Interested readers are referred elsewhere for a detailed description of each of these designs (Tyagi and Kramer, 1996; Wittwer et al., 1997). As an example to illustrate the principle, fluorescence resonance energy transfer involving dual hybridization probes is described below (Fig. 5). A pair of oligonucleotides, with complementary sequences to adjacent areas of a target, is labeled with fluorophores. One is labeled on its 3′ end, and the other is labeled with a different fluorophore on its 5′ end. When the target sequence is present, both probes bind to the target in a head-to-tail fashion. In this close proximity, the energy emitted from the 3′ end fluorophore excites the 5′ end fluorophore, and emission from the latter is detected. As in the SYBR green I method, the fluorescence intensity is proportional to the amount of the target sequences synthesized during the PCR cycling. The specificity of the techniques using sequence-specific fluorescently labeled probes is very high in comparison to that with dsDNA dyes. However, the synthesis of these probes is expensive. In addition, sequence selection for the initial probe design is often tricky because multiple criteria, such as PCR product length and equally efficient annealing of probes and primers, have to be satisfied at the same time for a successful result.

Figure 5. Dual hybridization probe. Single-strand oligonucleotides and target sequence are represented by comb-shaped objects. Fluorophores are depicted as circles with different shades. Light waves are represented by arrows. Fluorescence resonance energy transfer (FRET) occurs when the two probes are brought together by the target sequence, which is represented by an open arrow.

Real-time RT-PCR has several advantages over the noncommercial Q-RT-PCR assays. In all methods of real-time RT-PCR, the amount of PCR product is correlated with the fluorescence signal. Gel separation and staining of the PCR products are therefore not needed for analysis. The elimination of post-PCR manipulation minimizes the probability of cross-contamination and false-positive results. Moreover, the sensitivity and dynamic range of fluorescence detection are significantly higher than those for ethidium bromide staining and densitometric quantification. Copy numbers spanning five to six orders of magnitude can be detected with a lower limit of detection at 10 copies of BCR-ABL per 100 ng of cDNA (Mensink et al., 1998; Kreuzer et al., 1999). Automation and computerization significantly improve throughput and turnaround time. Complete PCR analysis using the fluorogenic probes described above takes less than 60 min (Emig et al., 1999), in contrast to several hours for the noncommercial PCR assays. In addition, easy standardization allows for comparison of results from different laboratories.

The list prices of real-time PCR instruments range from $36,000 to $96,000, depending on the manufacturer. The cost of reagents and supplies is high. Intra- and interassay reproducibility varies among different methodologies, as do the cycling speed, ease of probe design, and data analysis. Side-by-side comparison of TaqMan and LightCycler technologies for quantification of BCR-ABL has been reported in one study, showing a good correlation between the two (Kreuzer et al., 1999). Recommendation of a particular technique/instrument is difficult at the present time, until more experience is accumulated.

Q-RT-PCR is a close-to-ideal test for monitoring minimal residual disease for three reasons. It is sensitive, it is quantitative in detecting residual disease, and it can be performed on peripheral blood specimens that may allow easier and perhaps more frequent monitoring. Although it is still controversial, there is an accumulating body of evidence that serial Q-RT-PCR analysis is useful for assessment of the therapeutic response and for early detection of relapse in patients treated with allogeneic BM transplants (Delage et al., 1991; Lin et al., 1996) or IFNα therapy (Lion et al., 1995; Hochhaus et al., 2000b). It was shown that the level of BCR-ABL fusion mRNA at single time points correlates well with the cytogenetic response (Hochhaus et al., 1996a; Branford et al., 1999; Barbany et al., 2000) and disease stage (Cross et al., 1993; Elmaagacli et al., 2000). Moreover, on serial analyses, patients with high or increasing levels of BCR-ABL over the disease course have a greater probability of relapse than do those with steady-state or decreasing levels of BCR-ABL.

Because of lack of standardization, it is currently impractical to set a quantitative threshold for BCR-ABL above which patients are likely to relapse. Rather, serial quantitative analysis to examine the trend of change over time is a much more reliable way of predicting relapse. Q-RT-PCR would be expected to be clinically useful in monitoring patients undergoing other therapies, such as STI 571 treatment (see below).

LABORATORY DIAGNOSIS OF CML

Based on the studies reviewed in the preceding sections, for primary diagnosis of CML we recommend cytogenetics on BM in conjunction with qualitative RT-PCR on BM or PB. As discussed above, cytogenetics is required to document the Ph chromosome and other numerical or structural chromosomal abnormalities. RT-PCR is required to document the particular breakpoints (b2a2, b3a2, or e1a2).

It does not appear to be cost-effective to order FISH as a first-line test for CML diagnosis. FISH does not provide any more information in addition to the presence of BCR-ABL. Moreover, although FISH is the most sensitive test at the time of diagnosis, cases not detected by either cytogenetics or RT-PCR are very rare. Therefore, FISH on BM should be performed only when cytogenetics and RT-PCR both are negative and clinical features for CML are typical.

LABORATORY MONITORING OF THERAPIES

Therapies for the chronic phase of CML have been evolving. Currently, the mainstay therapies for chronic phase treatment are conventional chemotherapy with hydroxyurea or busulfan, IFNα (alone or combined with other drugs), and BM/stem-cell transplantation (reviewed in Faderl et al., 1999a). The latter two are used preferably, and which to choose depends on the patient's clinical characteristics and the availability of a matched donor. New drugs and new therapeutic strategies are under active investigation for both the chronic and the transformed phase of CML (reviewed in Faderl et al., 1999a; Thijsen et al., 1999). Although they will not be discussed specifically here except for STI 571, a principle by which test selection is made and might be extended to an investigational therapy will be suggested at the end of this section.

Laboratory monitoring of therapies serves two purposes. In the early stage, patients are monitored for their response to the therapy. With the development of new therapies for CML, nonresponders can be switched to other treatments. In the late stage of therapy, after patients achieve remission or after bone marrow transplantation, laboratory tests are used to monitor minimal residual disease and detect early relapse.

Before we discuss each therapy and the recommended approach to monitoring, the currently accepted definitions of therapeutic response will be reviewed. A hematologic remission refers to a state when patients have complete normalization of peripheral blood counts and show no signs or symptoms of disease. A cytogenetic response is further graded as complete, partial, minor, or no response depending on the percentage of Ph+/BCR-ABL+ cells determined by either cytogenetic or FISH analyses in the BM. A complete response is defined as no detectable Ph+/BCR-ABL+ cells in the BM. In a partial response, there are 1–34% Ph+/BCR-ABL+ cells. In a minor response, there are 35–94% Ph+/BCR-ABL+ cells. Patients who show no cytogenetic response have more than 95% Ph+/BCR-ABL+ cells in their bone marrow after therapy. A major response is a collective term including both complete and partial responses.

Conventional Chemotherapies

Conventional therapy with hydroxyurea or busulfan usually reduces the leukocyte count and complications. Although 50–80% of patients achieve a complete hematologic remission, complete cytogenetic responses are extremely rare. To date, conventional therapies have not been shown to alter the natural course of the disease nor prolong the overall survival (Cunningham et al., 1979; Hester et al., 1984; Kantarjian et al., 1993).

Given that complete cytogenetic responses are rare, the main goals of monitoring patients treated with conventional chemotherapy are to assess the therapeutic response and disease progression. Cytogenetic analysis, although not as sensitive as FISH or RT-PCR, is sufficient for quantification of Ph+ cells. Fifty to 80% of patients acquire one or more additional chromosomal changes during the course of their disease that may precede hematologic and clinical manifestation of blast crisis. Common changes include a second Ph, trisomy 8, isochromosome 17q, and trisomy 19 (reviewed in Heim and Mitelman, 1995). Although blast transformation may occur in the absence of cytogenetic clonal evolution, it has been found to be one of the poor prognostic factors for survival after the onset of transformation (Kantarjian et al., 1987, 1988; Nanjangud et al., 1994; Majlis et al., 1996).

Interferon-α-Based Therapy

IFNα, alone or combined with other agents, is the preferred therapy for older patients or young patients without an available bone marrow donor. Interferon therapy produces hematologic remission in 46–80% of cases and a complete cytogenetic response in 13–32% of cases (Yoffe et al., 1987; Talpaz et al., 1991; Faderl et al., 1999a). Patients on IFNα have a median survival that is 20 months longer than that of patients on conventional chemotherapy, and the 5-year survival is improved from less than 20% to 57% (Silver et al., 1999). Monitoring patients on IFNα therapy is important in assessing the efficacy of the treatment. For those who do not respond, alternative treatments may be needed.

Although cytogenetics formerly was the gold standard for monitoring IFNα therapy, FISH is currently the test of choice in this setting and may soon be replaced by Q-RT-PCR, as discussed above. It was shown that the cytogenetic response to IFNα is associated with prolonged survival (Kantarjian et al., 1995; Mahon et al., 1998). Five-year survival rates for patients with complete, partial, minor, and no cytogenetic response were 93, 88, 75, and 50%, respectively (Kantarjian et al., 1995).

FISH, compared to cytogenetics, is more sensitive, more accurate for quantification, and also faster and easier to perform, at a comparable cost. Recent studies performed to evaluate cytogenetic and FISH quantification have shown an excellent correlation between the two assays (Cuneo et al., 1998; Glassman, 1998; Kobzev et al., 1998). As noted earlier, FISH is more sensitive than cytogenetics in detecting Ph/BCR-ABL-positive cells in patients in whom the leukemic burden is low (Cuneo et al., 1998). For these reasons, we strongly recommend that cytogenetics be replaced by FISH for monitoring of IFNα therapy. Interestingly, in support of this recommendation, patients on the ongoing clinical trial of STI 571 are being monitored by FISH instead of cytogenetics (Druker et al., 1999).

Although the use of peripheral blood is more convenient and technically possible, it is unclear whether it can be used in place of BM specimens for FISH monitoring of IFNα therapy. The choice of PB or BM seems to depend on where the patients are in the course of therapy. Muhlmann et al. (1998), in a study of 30 paired PB and BM specimens obtained simultaneously from patients with different cytogenetic responses, showed an excellent correlation between quantification of BCR-ABL-positive cells from bone marrow and peripheral blood. However, two other reports, on a total of 53 paired specimens, showed that the levels of BCR-ABL cells detected in PB for most of the specimens were lower than that in the BM (Buno et al., 1998; Yanagi et al., 1999). In one of these reports, it was shown that, in 10 patients, the quantitative changes in the PB over an approximately 12-month course of IFNα therapy paralleled that of the BM, suggesting that monitoring with PB specimens could provide sufficient information as to whether patients respond to therapy (Buno et al., 1998; Yanagi et al., 1999). Given these findings, it can be inferred, however, that in the later stages of therapy when the leukemic burden is further reduced to below what can be detected in the peripheral blood, a BM specimen is still required for detection of minimal residual disease. Nevertheless, this remains an evolving and controversial area.

Qualitative RT-PCR is not as useful as FISH in monitoring residual disease after cytogenetic remission with IFNα therapy because RT-PCR remains positive in the majority of cases for as long as 68 months after a complete cytogenetic response (Lee et al., 1992). Furthermore, RT-PCR positivity does not seem to be associated with an immediate risk of relapse (Dhingra et al., 1992; Lee et al., 1992). Although dormant for unknown reasons, the BCR-ABL-positive cells in the bone marrow of such patients still have clonogenic potential (Talpaz et al., 1994; Reiter et al., 1998). On long-term follow-up, BCR-ABL fusion transcript levels in some patients do fall below what can be detected by qualitative RT-PCR (Kurzrock et al., 1998).

Q-RT-PCR shows great promise for monitoring IFNα therapy because of its effectiveness in predicting relapses. A persistently high level of the fusion transcript is correlated with a greater probability of relapse, whereas levels of residual disease fall over time in patients with cytogenetic remission (Hochhaus et al., 2000b). In a sequential analysis of 20 patients, 13 who displayed decreasing, constant, or fluctuating BCR-ABL remained in hematologic remission during a mean follow-up period of 25 months. In contrast, all seven patients who had increasing levels of the fusion transcript relapsed. The increase in BCR-ABL message preceded the hematologic or cytogenetic evidence of disease progression by a median of 6 months (Lion et al., 1995).

Allogeneic Bone Marrow / Stem Cell Transplantation

Allogeneic bone marrow / stem cell transplantation is currently the first-line therapy for young patients with an HLA-identical donor (reviewed in Sawyers, 1999), resulting in both hematologic remission and a complete cytogenetic response in the majority of patients. More than half of the patients survive for 5–10 years posttransplant (Clift and Storb, 1996; Gratwohl and Hermans, 1996; van Rhee et al., 1997; Silver et al., 1999). However, transplantation therapy is limited by donor availability and age restrictions because of transplant-related mortality in older patients.

Allogeneic bone marrow / stem cell transplantation dramatically reduces the leukemic burden in CML patients. The main goal of monitoring is to detect minimal residual disease and early relapse. In such cases, early detection of relapse may prompt clinical interventions such as withdrawal of immunosuppression, initiation of salvage therapies such as donor lymphocyte infusion, or a second BM transplant (Kolb et al., 1990; reviewed in Porter and Antin, 1999).

The test of choice for monitoring minimal residual disease posttransplant has been serial qualitative RT-PCR, although it is being replaced by serial Q-RT-PCR. Cytogenetics or FISH is not recommended because the leukemic burden posttransplantation in most cases is below the sensitivity of either of these techniques. RT-PCR, with its ability to detect 1 leukemic cell in 10⁴–10⁶ leukocytes, is extremely useful in this setting. However, a single PCR result is of limited value in predicting patients' outcome (Faderl et al., 1999b). Many studies have shown that persistently positive RT-PCR in serial samples correlates well with an increased probability of relapse (Hughes et al., 1991; Roth et al., 1992; Gaiger et al., 1993; Radich et al., 1995; Mackinnon et al., 1996; Drobyski et al., 1997). In the study by Roth et al. (1992), it was shown that the probability of relapse in eight patients with persistently positive RT-PCR was 75%, whereas the probability of relapse in 17 patients with positive RT-PCR at one time point was 41%. In contrast, none of the 23 patients with negative RT-PCR relapsed during a mean follow-up time of 372 days. Moreover, of 23 patients with mixed positive and negative RT-PCR results, 10 converted from positive to negative and 13 from negative to positive. Six of the positive-to-negative converters were in remission, four died, and none relapsed. In contrast, five of the 13 patients who converted from negative to positive on RT-PCR relapsed, three died, and five were in remission (Roth et al., 1992). Similar findings were reported in other studies (Radich et al., 1995; Mackinnon et al., 1996; Drobyski et al., 1997). Notably, all of these results were obtained by qualitative rather than quantitative methods.

Although serial qualitative RT-PCR has acceptable sensitivity and demonstrated clinical utility in monitoring of CML patients, reports have shown that the BCR-ABL fusion transcript can be detected by an extremely sensitive RT-PCR in the blood of up to two-thirds of normal individuals (Biernaux et al., 1995; Bose et al., 1998). This suggests that the BCR-ABL translocation by itself is not sufficient for leukemogenesis. Another possibility is that the BCR-ABL fusion resides in terminally differentiated cells, which have lost their clonogenic ability. Detection of such “benign” BCR-ABL-carrying cells therefore should be considered a false-positive result. Notably, the sensitivities of the RT-PCR assays in these two reports are 40- and 100-fold higher than those of routinely used clinical tests. Taking the conservative number, the probability of detecting a “benign” cell with a routine clinical test would be 1/40 (2.5%) in the adult population studied.

Two measures may be taken to reduce the false-positive rate of the qualitative RT-PCR. One is to analyze the fusion types by electrophoresis and to follow the specific transcript the patient had prior to the transplant. The vast majority of BCR-ABL in CML patients represent either b3a2 or b2a2 fusions, whereas most of the fusion types found in apparently healthy individuals are e1a2 (Bose et al., 1998). The second measure is to conduct serial testing, because the probability of two tests being falsely positive as a result of “benign” BCR-ABL- bearing cells is approximately 1/1600 (1/40 × 1/40).

Furthermore, in the posttransplant setting, Q-RT-PCR has been shown to be more useful than qualitative RT-PCR (Delage et al., 1991; Cross et al., 1993; Lion et al., 1993; Lin et al., 1996; Moravcova et al., 1998). In a large study of 98 patients, 21 of 29 (72%) patients with an increasing or persistently high level of BCR-ABL relapsed. In contrast, only one of the 69 patients with a stable or falling amount of BCR-ABL relapsed (Lin et al., 1996). Based on some of the early studies, the European Investigators on CML Group proposed a definition of molecular relapse as a 10-fold increase in the BCR-ABL expression detected by a minimum of three consecutive Q-RT-PCR analyses (Lion, 1994).

In summary, in the posttransplant setting, serial qualitative or quantitative RT-PCR is currently the test of choice for monitoring minimal residual disease and detecting molecular relapse well before cytogenetic or hematologic relapse.

Tyrosine Kinase Inhibitor

Elucidation of the molecular mechanisms of BCR-ABL-mediated transformation has led to the development of STI 571, an inhibitor of the tyrosine kinase activity of ABL (Druker and Lydon, 2000). Crystallographic study has shown that the catalytic domain of ABL, when complexed with the inhibitor, is locked in an inactive conformation (Schindler et al., 2000). A Phase I clinical trial on IFNα-refractory CML patients has shown that STI 571 is highly efficacious, with minimal toxicity. Twenty-three of 24 (96%) patients achieved a hematologic response, and 33% showed a cytogenetic response as assessed by FISH (Druker et al., 1999). Moreover, a Phase II study published recently showed that, of 388 refractory patients who had completed 6 months of therapy, 161/290 (56%) patients achieved a major cytogenetic response (Kantarjian et al., 2000). Given that the rate of cytogenetic response of STI 571 therapy is still lower than BMT, FISH rather than qualitative RT-PCR appears to be the most appropriate test for monitoring of patients on STI 571. Q-RT-PCR, which shows utility for monitoring interferon therapy as described above, is likely to be effective for monitoring patients treated with STI 571 as well. Whether STI 571 is capable of inducing molecular remission is currently under investigation (Paschka et al., 2000). Several Phase II clinical trials by the International STI 571 Study Group are ongoing to investigate the efficacy of STI 571 in patients with CML in accelerated phase or blast crisis (Hochhaus et al., 2000a; Talpaz et al., 2000).

CONCLUDING REMARKS

Recommendations for laboratory diagnosis and monitoring of CML are summarized in Table 3. In some scenarios, recommended tests are given in the order of preference, so that physicians can choose other alternatives in case the preferred tests are not available. Figure 6, in an algorithm format, provides an example of how laboratory information might be used effectively to direct the clinical management of CML patients.

Table 3. Summary of Recommendations

	Recommended tests	Purposes
> = Preference.
Diagnosis	Karyotyping	To document Ph and other chromosomal abnormalities
& Qualitative RT-PCR	To document the fusion type
Monitoring
Conventional therapies	Karyotyping	To assess therapeutic response and monitor for changes before blast crisis
IFN-α	Serial Q-RT-PCR > FISH > Karyotyping	To assess therapeutic response and detect relapse
BMT	Serial Q-RT-PCR > Serial qualitative RT-PCR	To monitor MRD and detect molecular relapse

Figure 6. Algorithm for therapeutic monitoring of CML. BMT, bone marrow/stem cell transplant; Q-RT-PCR, quantitative RT-PCR; Neg, negative; Pos, positive; f/u, follow-up; lymph., lymphocyte; Cont., continue; Rx, therapies; HU, hydroxyurea; BUS, busulfan. *Therapies for CML at these stages of the disease are controversial. The clinical interventions in the algorithm are examples of what some physicians might do in response to the particular laboratory information, and do not represent recommendations.

In conclusion, the choice of a laboratory test for CML monitoring depends on the degree to which a particular therapy reduces the number of leukemic cells. The residual leukemic burden during or posttherapy dictates the test of choice. Conventional chemotherapies, IFNα, and BM/stem-cell transplant are therapies with an increasing ability to reduce the tumor load. BM/stem cell transplant, in particular, has the potential to eradicate leukemic cell populations. Cytogenetics, FISH, and RT-PCR represent tests with increasing levels of sensitivity. A central principle in selecting a test for following a particular therapy is to match the sensitivity of the test to the therapeutic effectiveness. This same principle may help guide the initial selection of a monitoring test for clinical trials of any new therapies, as exemplified by STI 571.

In the future, it is very likely that Q-RT-PCR will supplant all other methods for monitoring the therapeutic response and residual disease, whereas cytogenetics and qualitative RT-PCR will prevail for initially establishing the diagnosis of CML.

Acknowledgements

We thank Ewa Tomczak, VMD, and Hanna Rennert, PhD, for the cost analyses on FISH, karyotyping, and RT-PCR. The authors also thank Pin Lu, PhD, at the Bristol Meyer-Squibb Pharmaceutical Research Institute for his valuable input on Q-RT-PCR.