November 2009
Feature Story

William Check, PhD

When it comes to detecting minimal residual disease in patients undergoing therapy for acute myeloid leukemia, laboratorians face a conundrum similar to our society’s dietary debate. In the dietary field, advocates can be found both for high-protein, low-carbohydrate regimens and for eating a low concentration of saturated fat and a plethora of whole grains. In the arena of minimal residual disease, or MRD, testing, some favor immunophenotypic analysis of leukemic blasts by multiparameter flow cytometry, while others prefer to use molecular techniques, chiefly PCR analysis of mutations and translocations. As with the dietary disagreement, the MRD detection dilemma persists because of a lack of decisive evidence for either choice. Unlike people seeking the optimal diet, however, laboratorians can, so to speak, have their bread and eat it, too—experienced practitioners use both immunophenotypic and molecular methods. They see the two approaches as complemen­tary rather than competitive.

A second, more serious issue arises when we consider the clinical dimension of MRD detection. In acute myeloid leukemia, or AML, the presence of MRD carries an ominous prognosis—a high risk of relapse and death—and it indicates that therapy has not had an adequate protective effect. Most important, no one knows how to alter the dire impact of MRD. Aside from chemotherapy and transplantation, clinicians have little in the way of therapeutic alternatives. Targeted therapies may someday fill this gap, but for now they are investigational.

“Many university hospitals do MRD detection in AML by molecular and flow cytometry,” says Guang Fan, MD, PhD, associate professor of pathology and medical director of the flow cytometry laboratory at Oregon Health and Science University. “For sensitivity, the difference between methods is not clinically significant.” Molecular can detect as low as one leukemic cell in 10⁵ or 10⁶ bone marrow cells, while flow cytometry can detect one leukemic blast in 10⁴ white cells. In practice, these sensitivities are equivalent. “Each technology has advantages and disadvantages,” Dr. Fan says. “They are complementary; you can’t use one technique to replace the other.” She uses both immunophenotypic and molecular analysis at initial diagnosis and for MRD detection.

A molecular method can detect a specific marker, such as an FLT3 mutation, known to be present in a patient’s leukemic cells. “However, this marker can only be detected in about 30 percent of AMLs,” Dr. Fan says. “An abnormal immunophenotypic pattern occurs in about 90 percent of AML cases, but may not be specific for leukemia blasts,” she says.

“When we find MRD, we report it,” Dr. Fan continues. “Our clinicians like to have the information but don’t know what to do with it. Weekly conferences are held with the oncologists to discuss this issue. Published evidence tells you that if you detect MRD, the relapse rate is higher and the survival rate is lower. However, there is no data telling you what to do if MRD is present,” she says.

Brent Wood, MD, PhD, of the University of Washington, says outcome studies are needed to show how MRD findings can guide clinical practice. “Most studies that have been done so far have demonstrated in­ferior outcome for AML when MRD is present,” says Dr. Wood, professor of laboratory medicine and director of the hematopathology laboratory. “But they haven’t demonstrated that we can intervene and change the outcome.”

Dario Campana, MD, PhD, who has extensive experience with childhood AML, says the general consensus is that MRD is a strong prognostic factor and clinically worthwhile for risk assessment. “Flow cytometry is really the only method we can use to study MRD in the majority of AML patients,” says Dr. Campana, a member of the oncology and pathology departments at St. Jude Children’s Research Hospital, Memphis. “Another approach is PCR of gene fusions, but they are found in only a minority of patients.”

In adult AML, clinical research groups outside the U.S. have done most of the work showing that the presence of MRD has an adverse prognosis. Gerrit J. Schuurhuis, PhD, associate professor of hematology at VU University Medical Center, Amsterdam, leads one such consortium. “This is one of the most difficult leukemias you can choose for [MRD analysis],” Dr. Schuurhuis says. “Almost every patient seems like a new disease.” Chronic myelogenous leukemia, which has a characteristic trans­location, is different in that way, as is acute lymphoblastic leukemia. “We chose immunophenotyping for our AML studies because there were so few molecular aberrancies to follow,” Dr. Schuurhuis says. He estimates that molecular techniques now make it possible to detect an abnormality in about half of AML patients, whereas immunophenotypic analysis is productive 90 percent to 95 percent of the time. His group is now investigating molecular MRD, including known translocations and new mutations like NPM1 (nucleophosmin 1). A multicenter study that Dr. Schuurhuis is leading, within clinical AML studies of the HOVON/SAKK cooperative clinical working group, will be the first to provide comparative data in a prospective way for immunophenotypic and molecular methods.

Michael C. Heinrich, MD, staff physician at Portland VA Medical Center and professor of hematology/oncology at OHSU, is providing genotyping of FLT3 status for the Dutch group. He agrees that molecular and multiparameter flow cytometry technologies are potentially complementary, depending on the degree of sophistication of the flow facility. “Are they individualizing the immunophenotype?” he asks. “As well, PCR might be excessively sensitive.” The clinicians at the VA Medical Center are not routinely using flow cytometry MRD at this time. “It is somewhat investigational but interesting,” Dr. Heinrich says.

Wolfgang Kern, MD, and colleagues at Munich Leukemia Laboratory GmbH also showed the prognostic value of MRD detection in AML (Bacher U, et al. Cancer. 2006;106:839–847). Dr. Kern, CEO of immunophenotypic analysis at the Munich lab, sees both immunophenotypic assessment and molecular techniques as being “very valuable.” Of the lack of consensus on antibody panels for MRD detection by multiparameter flow cytometry, Dr. Kern says: “Several antigens and antigen combinations are generally agreed to be informative. The composition of your panel depends on the number of colors you can apply.” The Munich lab now runs a five-color instrument and is validating a 10-color Beckman Coulter instrument. “We will be able to run fewer tubes, it will be more economical, and we will be able to characterize smaller populations,” he says. With 10 colors, it will be possible to identify a cell population using two markers and have eight colors left to identify markers in that population.

Dr. Wood, who was one of the early adopters of doing nine- and 10-color flow for MRD in AML, seconds the efficiencies gained with more colors. He also likes that having multiple markers in each tube “gives a more coordinated view of antigen expression throughout the maturation process. That pattern of expression forms the understanding of what we know normal to be.” As Michael J. Borowitz, MD, PhD, puts it, “Dr. Wood told me that if he had a 30-color flow cytometer, then he could do everything in one tube.”

In a more serious vein, Dr. Borowitz, professor of pathology and oncology at Johns Hopkins Medical Institutions, calls attention to the heterogeneity in multiparameter flow cytometry practice. “As far as I can tell,” Dr. Borowitz says, “no two groups publishing [on MFC detection of MRD] are doing the same thing. That makes it somewhat difficult to generalize.” It’s also difficult to standardize multiparameter flow cytometry. “It has a certain amount of art to it and requires understanding how to interpret pictures,” he says, and adds: “Doing flow cytometry is perfect for a pathologist because looking at dot plots to make a diagnosis is still doing morphology, just with its own unique set of artifacts. However, the lack of standardization means everyone is looking at different pictures.”

In the past few years Ian Lewis, MD, senior consultant in the Division of Haematology at the Institute of Medical and Veterinary Sciences in Adelaide, Australia, and his coworkers have published several articles on the prognostic value of immunophenotypic detection of MRD in adult AML (Al-Malawi A, et al. Am J Clin Pathol. 2009;131:16–26; Al-Malawi A, et al. Int J Lab Hematol. 2009;31:61–68). The chief advantage of flow cytometry, he says, is that it is potentially applicable to a larger number of patients than molecular techniques. “Its major disadvantage is that molecular technology is a bit more user friendly—it’s easier to standardize in a clinical laboratory.”

Daniel A. Arber, MD, professor and director of anatomic pathology and clinical laboratory services in the Department of Pathology at Stanford University Medical Center, reinforces several of the others’ observations. “Most places in the U.S. do flow cytometry for MRD in AML, but it’s not very uniform how we do it,” he says. “Also, there are not clear guidelines for what we call positive. We need to establish guidelines and cutoffs and demonstrate the efficacy of detecting MRD—the kind of work that Dr. Kern and Dr. Schuurhuis and others are doing.”

What Stanford’s clinicians do based on MRD information is not clear, other than to correlate it with molecular and cytogenetic studies, Dr. Arber says. He calls this “the hematologist’s dilemma—they want this data, but they don’t know what to do with it when they get it.”

To David Bahler, MD, PhD, associate professor of pathology at the University of Utah, “The bottom line is that molecular is good but not applicable to as many cases as flow cytometry, and flow is almost as sensitive.” Like Dr. Arber, he sees no definitive standard for what degree of difference from normal is needed to call a specimen abnormal, saying, “There is some ambiguity here.” He compares the phenotypic abnormalities seen in hematopoietic cells to gradations of morphologic abnormality in solid tumors—from minimal to high-grade dysplasia. In both cases, Dr. Bahler says, “The more abnormalities you find, the more confident you are they reflect neoplasia.”

He also reiterates an underlying theme: “Lots of clinicians are not acting on information about MRD. If we detect MRD by flow, it is not clear that they would reinstitute induction therapy, for instance.” Worse, Dr. Bahler finds that clinicians are not always clear about when to order MRD assessment and may order it too early. “Clinical use of MRD for AML is still under study,” Dr. Bahler sums up. “There is not a consensus.”

According to St. Jude’s Dr. Campana, looking for MRD by multiparameter flow cytometry was pioneered by Michael R. Loken, PhD, of the University of Washington. “Dr. Loken was one of the first individuals who described the fact that immunophenotypes of AML cells could be sufficiently different from normal bone marrow cells to distinguish MRD,” Dr. Campana says (Sievers EL, et al. J Natl Cancer Inst. 1996;88:1483–1488). Dr. Loken and Eric L. Sievers, MD, of the Fred Hutchinson Cancer Research Center, showed that the presence of MRD was an adverse prognostic factor in pediatric AML (J Pediatr Hematol Oncol. 1995;17:123–133). “Later,” Dr. Campana adds, “we reported that, in children with AML, MRD was the strongest indicator of outcome” (Coustan-Smith E, et al. Br J Haematol. 2003;123:243–252).

Researchers in Europe—Alberto Orfao and Jesus San Miguel in Salamanca, Spain, and Adriano Venditti in Rome—published large series in adult AML patients demonstrating that detection of MRD with immunophenotypic analysis in patients in complete morphologic remission was a strong indicator of adverse prognosis (San Miguel JF, et al. Blood. 2001;98:1746–1751; Venditti A, et al. Leukemia. 2003;17:2178–2182). Dr. Venditti has since suggested that the kinetics of reduction of MRD could predict response and survival (Buccisano F, et al. Leukemia. 2006;20:1783–1789) and that peripheral blood samples could be used for MRD detection (Maurillo L, et al. Haematologica. 2007;92:605–611).

Dr. Borowitz differentiates true MRD testing from a more common practice that requires a lower level of precision. “Say you have a patient who, after induction therapy, has maybe one percent or two percent or three percent blasts in the bone marrow,” he says. “Are these regenerating blasts? Or residual leukemia?” That situation doesn’t require the same kind of precision needed for MRD detection. “You are not asking whether something you don’t see is there. You see blasts and simply want to know whether they are good or bad,” Dr. Borowitz says. In MRD testing, he explains, “you are addressing a different question. You have a patient who is in morphologic remission. And you want to know, Do they still have residual disease?” Detecting leukemic cells at this level demands much greater sensitivity.”

In addition to immunophenotypic assessment, molecular methods can be used for MRD detection. “The biggest issue with molecular methods for MRD detection in AML is that compared to MFC, only a subset of patients can be monitored,” Dr. Borowitz says. “These include those with the relatively common cytogenetic abnormalities t(8;21) and inv(16). Each creates a fusion protein using part of a complex called core binding factor, and standard reverse transcriptase PCR can detect these with high sensitivity.”

“There is some more recent literature using mutations in genes such as FLT3 or NPM as targets for MRD measurement,” Dr. Borowitz says, “though each is again only applicable to a subset of patients.” In addition, several investigators have used the Wilms’ tumor gene 1 for MRD analysis (Cilloni D, et al. J Clin Oncol. 2009, Epub ahead of print). This does not detect a mutant gene, however, but rather overexpression, and there is some question about how generally applicable it is to AML. “Abnormalities are quantitative, not qualitative,” Dr. Borowitz says. “If it expressed at 1,000-fold higher levels than normal, by the time you are down to one cell in 1,000, you are in the normal range.” His opinion: “Flow is probably the most robust, generalizable technique across the board in AML.”

Dr. Kern has a more favorable view of molecular techniques to evaluate MRD in AML: “Both molecular genetics and multiparameter flow cytometry are very sensitive, and both give important prognostic information” (Kern W, et al. Cancer. 2008;112:4–16). At Munich Leukemia Laboratory, both are used—“it depends on what the customer wants.” Dr. Kern estimates that monitoring of MRD by flow cytometry is possible in about 95 percent of AML patients, though, he notes, “In about 20 percent it may not be as sensitive as molecular genetics.” His estimate for molecular genetics is almost 80 percent of AML patients. This figure is higher than some others because he includes NPM1, one of the newest molecular markers. Dr. Kern found NPM1 in 50 percent of AML cases with normal karyotypes, or 25 percent of all AML (Schnittger S, et al. Blood. 2009;114:2220–2231). He believes that molecular genetics will be used increasingly to target specific genetic lesions. “Perhaps someday we will aim to monitor MRD in all AML cases with molecular genetics,” Dr. Kern speculates.

Dr. Schuurhuis’ group in Amsterdam has shown that, for three cycles of treatment, threshold values can be established for immunophenotypic detection of MRD that predict a six- to sevenfold relative risk for relapse (Feller N, et al. Leukemia. 2004; 18:1380–1390). Cutoffs are around 0.1 percent, well within the sensitivity of the method. These values are being used in the prospective study that Dr. Schuurhuis leads, which includes 30 institutions in the Netherlands, Belgium, and Switzerland. Samples have been analyzed and are being correlated with clinical outcomes. “Very soon we will have that data,” Dr. Schuurhuis says. This trial is the first multicenter study of its type. “In adult AML, this kind of research has so far mainly been done in single institutions,” Dr. Schuurhuis says.

Once verified, threshold frequencies can be used to assess risk, in Dr. Schuurhuis’ view. “Good-prognosis” subtypes of AML are mostly based on cytogenetic data. “Ultimately, still many patients in good-prognosis subtypes have a poor outcome,” he notes. Tracking MRD during therapy could identify these patients, he suggests. Theoretically, MRD-positive patients might be considered for more-intensive therapy. On the other hand, among patients with poor-prognosis subtypes of AML, MRD tracking could find those who have a good response, so they can be considered for less-intensive therapy.

Another attractive application of MRD testing in AML may be to monitor allogeneic transplantation, a potentially curative treatment. “Allogeneic transplants have a very good graft-versus-leukemia effect, but also a strong graft-versus-host effect,” Dr. Schuurhuis says. “So they still have a fairly high mortality rate.” He and colleagues use immunosuppressive agents to reduce side effects, but they also reduce the graft-versus-leukemia effect. “What we have shown,” he explains, “is that we can monitor MRD while a patient is on immunosuppression and see an increase in MRD. If we see relapse coming, we might reduce immunosuppression or withdraw it temporarily.” On occasions when they have done that, MRD has dropped very fast. Monitoring allogeneic transplant patients in this way may allow clinicians to titrate immunosuppression and enhance the curative power of transplantation.

Dr. Lewis, too, believes that MRD status might help tailor treatment. Using ROC curves of multiparameter flow cytometry data, he showed that the optimal cutoff for MRD to predict relapse is 0.15 percent, very much like what Dr. Schuurhuis found. Even using this value, 28 percent of patients with no sign of MRD relapsed, which defines what Dr. Lewis calls the “main dilemma in AML therapy”: determining which patients will remain in remission. “The majority of adult AML patients who achieve complete remission by conventional criteria will relapse,” he says. Only about 25 percent of adult AML patients are cured by chemotherapy. “You don’t want to touch those patients,” he says.

Currently, if a patient in first remission has a sibling donor, the patient would be offered bone marrow transplantation as potentially curative treatment. “That’s all well and good if your disease will relapse,” Dr. Lewis says. “But if you’re already cured, you don’t want to go through this intensive procedure, which has its own risks and potential mortality.” For the 70 percent who don’t have a sibling donor, bone marrow transplantation has an even greater probability of complications and death. Clinicians are reluctant to offer transplantation in first remission. “But if we can say with fair certainty that this person will relapse, we would be more likely to go ahead with transplant,” Dr. Lewis says. That’s where immunophenotypic assessment of MRD might help: “If we can’t detect disease by conventional techniques, but we can detect it by flow cytometry, it might be better to transplant now than wait till relapse.” Conversely, in a patient who has no sign of MRD post-induction and post-consolidation and has no other risk factors, “even if they had a sibling donor, we would probably hold off on transplant.”

Much of the uncertainty about how to use MRD status to guide clinical decisions results from the fact that less than 10 percent of adults with cancer are treated in protocols, Dr. Campana says. On the other hand, 95 percent to 98 percent of children with cancer in the U.S. are treated within protocols. “That’s why there is much more progress with children than with adults,” he explains.

A multicenter clinical trial that was conducted recently among children with AML illustrates Dr. Campana’s point and shows how MRD quantitation might be evaluated as an aid to treatment planning in adult AML. In the pediatric trial, immunophenotypic assessment of MRD after a first induction phase was used to assign intensity of subsequent therapy. If MRD was more than one percent, the patient was a candidate for transplant. For persistent MRD at lower levels, patients could have the anti-CD33 antibody-based immunotoxin gemtuzumab ozogamicin (Mylotarg) added to their chemotherapy regimens. Results, which will be presented at the American Society of Hematology meeting next month, showed benefit from this protocol and led to a modified MRD-guided treatment algorithm now being tested.

Efficacy of immunophenotypic assessment of MRD relies on many details, such as the composition of panels. Dr. Schuurhuis lists some antibodies he uses, starting with a core of “primitive” markers—CD34, CD133, and CD117—to identify blasts or stem cells. He always includes the leukocyte marker CD45. To detect lineage aberrancies, among others, he incorporates lymphoid markers CD2 and CD7 and the NK cell marker CD56, which should not be present on myeloid cells. Nor should differentiation antigens like CD14 (monocytes) or CD15 (granulocytes) be present on primitive populations. “We always include myeloid markers CD13, CD33, and CD117,” he says. Both CD13 and CD33 are seen on normal myeloid cells, but in AML one can see AML cells that have one and lack the other, or lack both.

Timing is also important. “Should we look for MRD after induction or after consolidation?” Dr. Fan asks. “This is still a big issue.” Evidence exists for both times. Some say that serial time points are best, and Dr. Fan agrees, particularly since MRD may be present after induction and absent after consolidation. More important, the opposite situation can occur: No MRD after induction, then back after consolidation. “Induction, consolidation, and maintenance—all may need to be done to see the trend,” she concludes.

Dr. Fan cautions about other possible changes. “In AML the FLT3 status can change from diagnosis to relapse,” she says. An AML that is positive for an FLT3 mutation at diagnosis may be negative at relapse, or an FLT3 mutation can appear for the first time in a relapse marrow. “Any AML is not a pure cell population,” Dr. Fan says. “One clone can initially be small and later become dominant.” Antigen shift can also happen in immunophenotypic patterns. In perhaps 80 percent of cases, the leukemic blast immunophenotypic pattern at relapse could be different from initial diagnosis. “Fortunately,” Dr. Fan says, “we can typically see more than one abnormality with flow, and an abnormal pattern may remain at relapse.”

Dr. Fan believes that both immunophenotypic and molecular assessments for MRD will continue to be useful. “Patients are so different,” she says. “In the future, we plan to sort marrow cells to attain more than 90 percent blasts. Then we’ll use the sorted blast cells for molecular and cytogenetic studies, which will give us increased sensitivity and accuracy.” Sorting could be especially useful if an immunophenotypic profile resembles a normal population, such as regenerating blasts. “We can sort and then send the enriched blasts for further studies to distinguish regenerating blasts from MRD,” Dr. Fan says. For sorting, which she plans to start soon, the laboratory will use a Becton Dickinson FACS­Aria, which is basically an eight-color FACS­Canto flow cytometer plus a sorter.

Regarding reimbursement, Dr. Fan says, “We don’t make a separate charge for MRD, just a general flow cytometry analysis for leukemia.” Professional fee is paid at three levels: two to eight antibodies, eight to 16, and more than 16. “We generally use more than 16. That’s adequate for doing a good job on MRD,” she says.

Optimal time points for MRD assay are still being worked out, Dr. Wood agrees. Right now standard times are at the end of induction chemotherapy—day 28 or so—and sometime during consolidation. He notes that a few groups have reported that finding MRD later, during consolidation, may be more prognostic than earlier, after induction. Dr. Wood offers one possible explanation: “Some of our assays may suffer from sensitivity issues and may underestimate the level of MRD present at day 28.” A potential strategy to overcome this could be to introduce novel targets. For instance, a research group at Stanford’s Institute for Stem Cell Biology and Regenerative Medicine has shown that CD47 and CD96 appear to be differentially expressed on leukemic versus normal stem cells (Hosen N, et al. Proc Natl Acad Sci USA. 2007;104:11008–11013; Majeti R, et al. Cell. 2009;138:286–299).

Another important detail, Dr. Wood says, is to use the same full panels at diagnosis and at followup. He contrasts this with using mainly the immunophenotypic pattern identified at diagnosis to look for MRD at followup. “We know that immunophenotypic profiles can change after therapy,” Dr. Wood says, “which makes a more limited approach somewhat problematic. We want to characterize the maturation of all early cell populations and see that they conform to normal patterns of expression.”

Standardizing flow cytometry is critical, Dr. Wood says, not just for MRD detection but for all clinical applications. “If flow cytometric approaches for diagnosis and MRD assessment are not standardized, it will cripple the long-term use of this technology,” he warns. He has led one group attempting this task (Wood BL, et al. Cytometry B Clin Cytom. 2007;72 suppl 1:S14–22). As incoming president of the Clinical Cytometry Society, Dr. Wood says, “One of the things I’m interested in is approaching standardization of flow cytometry from a number of standpoints.” In his view, one reason flow has taken longer to standardize than some other assays is the way in which this technology developed and the people attracted to it. “Many pathologists used to morphologic evaluation, which is also subjective, like this technology,” he explains. “They have a strong independent streak. To get them to agree on any single aspect of this technology has been challenging.”

Dr. Bahler adds that, in addition to detecting the type of abnormality already described—“lineage infidelities,” the appearance on myeloid blasts of antigens that typically appear on T cells or NK cells or lymphocytes, such as CD7, CD2, or CD56—immunophenotypic assessment can reveal quantitative abnormalities. One example is expression of CD34 at a much higher level than on normal blasts. “Asynchronous” expression—two antigens not normally present together—is also detectable. An example would be an early antigen with a late-appearing one, such as CD34 with CD11b or CD64. To call a difference an abnormality, Dr. Bahler says, “you need a distinct population.” In his opinion, this involves a half log or a log difference in intensity levels for a quantitative abnormality or a level significantly above baseline for discrepant co-expression.

Interpreting immunophenotypic patterns in AML marrows requires a substantial knowledge base. “This becomes important mainly when picking up very low levels of disease,” Dr. Bahler says. “To familiarize yourself, you should run a number of different bone marrows, such as regenerating marrow. A helpful sample would be a posttreatment marrow from a person with ALL, in which the appearance of myeloid cells closely parallels the kind of regeneration you see in AML or following cytokine therapy.”

As useful as correctly done immunophenotypic assessment already is, Dr. Wood says, “We need to continue to define novel antigen targets and aberrant expression patterns that can further improve our ability to detect with a high level of sensitivity the presence of MRD.” An important advance would be to develop assays that identify and enumerate leukemic stem cells. “We need assays that can identify this developmentally rather primitive cell population,” he says.

Dr. Bahler calls work on leukemic stem cells “an area of hot investigation.” A recent publication reported the molecular characteristics of a possible Hodgkin lymphoma stem cell, which could theoretically be self-renewing and give rise to the Hodgkin and Reed-Sternberg cells typical of this disease (Jones RJ, et al. Blood. 2009;113:5920–5926). “If we could identify the phenotype of leukemic stem cells, we could substantially change things. In terms of monitoring MRD, this population is potentially very important,” Dr. Bahler says.

In AML patients, Dr. Schuurhuis and collaborators have described a possible leukemic stem cell population, which they also call “leukemia-initiating” cells (Moshaver B, et al. Stem Cells. 2008;26:3059–3067; van Rhenen A, et al. Blood. 2007;110:2659–2666; van Rhenen A, et al. Leukemia. 2007;21:1700–1707.). They looked at a small compartment of leukemia stem cells that they could assess with specific immunophenotypic methods. “We found that monitoring these tumor-initiating cells provides prognostic value in addition to MRD burden [this will be presented at the ASH meeting]. It may help to distinguish patients who will do poorly even with low MRD, since it is more biologically relevant.” Distinguishing the leukemic stem cell compartment from normal stem cells “will be a very important factor when it comes to clinical decisionmaking,” Dr. Schuurhuis says. Even more important, it could lead to new therapies. For instance, his group is working to develop directed treatment against the CLL-1 antigen uniquely present on leukemic stem cells.