Targeting Treatment in AML

There is no single marker that is specific for leukemia, but aberrant combinations can be defined that are not present in normal hemopoiesis. Newer knowledge suggests that leukemic stem cells reside in the CD38⁺ CD33⁻ population^1,2 and are potentially distinguished from normal cells by the expression of CD123, which represents the phenotype for interleukin-3 receptor α (IL3-Rα).³ For antibody-directed treatment, CD33 has become the epitope to be targeted based on the rationale that it is widely expressed in patients with acute myeloid leukemia (AML) and is restricted in normal tissues to hemopoietic precursors. If the leukemic stem cell was CD33⁻, then the potential of using CD33 could be limited. Recent data, however, suggest that the leukemic stem cell, as defined in a non-obese dibetic–severe combined immunodeficiency (NOD-SCID) model, does express CD33,⁴ but whether this has clinical relevance is unclear.

CD33 is ubiquitously expressed on hemopoietic precursors and is expressed on more than 90% of AML blasts, and has become the main target for antibody-directed treatment. Initial experience using a humanized unconjugated antibody, I195, showed that the antibody had modest efficacy against recurring disease as did monotherapy, but it was able to convert most patients in a small series of acute promyelocytic leukemias, who were in hematologic but not molecular remission, to molecular negativity.⁵ A randomized trial combining I195 with chemotherapy versus chemotherapy alone in recurring disease showed no benefit from the addition of antibody in remission rate, disease-free survival (DFS), or overall survival (OS).⁶ New studies with this antibody are being planned in older patients as monotherapy combined with courses of low-dose Ara-C on a continuous basis. New enthusiasm for this approach emerged with the development of antibody-directed chemotherapy also targeting the CD33 epitope. The immuno-conjugate gemtuzumab ozogamicin (GO) combines the IgG₄ humanized antibody to the powerful intercalating agent, calicheamicin.⁷ Calicheamicin is too toxic to be administered as a free drug, and therefore it must be delivered when coupled to the antibody. Since the CD33 antigen/antibody complex is rapidly internalized, this is a potentially attractive mechanism of drug delivery. A pivotal study of GO monotherapy in older patients with AML in first relapse showed an overall remission rate of 28% (13% complete response [CR], 15% CRp),⁸ with subsequent US Food and Drug Administration (FDA) approval being granted for treatment of patients aged 60 years or older with recurring CD33⁺ AML. Several prospective trials have been initiated to assess the therapy’s potential in different settings in the disease (Table 1 ). The MRC AML15 and SWOG106 trial, are assessing the value of adding GO simultaneously to conventional induction treatment in younger patients and also in consolidation (MRC AML15) or as maintenance (SWOG106). The feasibility of combining GO with chemotherapy was in doubt based on an early study using the licenced 9-mg dose, which resulted in excess liver toxicity. The Eastern Cooperative Oncology Group (ECOG) E1900 trial is evaluating whether the administration of GO before autologous stem cell transplantation can reduce relapse risk. In Europe, the Grupo Italiano Malattie Ematologiche Maligne dell’Adulto–European Organization for Research on Treatment of Cancer (GIMEMA-EORTC) has developed a study in older patients (AML17). In this trial, the GO and chemotherapy are given sequentially. In a preparatory study in untreated patients older than 65 years with performance scores of 0 to 1, at a standard dose of 9 mg/m², CR was achieved in 35%. Patients were then scheduled to proceed to chemotherapy. At the end of this total sequence, 54% of patients entered CR.⁹ This experience has been developed into a randomized trial (AML17) with GO plus chemo versus chemo in induction and consolidation chemotherapy with or without GO.

In older frail patients, the GIMEMA-EORTC are comparing 2 dose sequences of GO with a view to comparing GO monotherapy with best supportive care, while the MRC Group is comparing low-dose Ara-C versus low-dose Ara-C plus GO in older, less fit patients.

The first of these major trials to report is MRC AML15, where preliminary analysis on 1113 randomized patients showed a significant reduction in relapse risk and improvement in DFS in patients with favorable or intermediate-risk cytogenetics but not in patients with adverse cytogenetics.¹⁰ Subsequent follow-up has shown a significant overall survival benefit in the favorable and intermediate groups.

Other interesting approaches such as fractionated dosing with induction chemotherapy or as part of reduced-intensity conditioning for allograft are in development. It is now apparent from these studies that CD33 expression on blasts is not a precondition for benefit. There has been concern that, since calicheamicin is a substrate for P-glycoprotein (Pgp), that this may compromise efficacy. This is implied by the demonstration of a reciprocal expression of CD33 to Pgp; however, in the MRC AML15 trial, Pgp expression appeared to have little effect on outcome. Similarly, concerns about the combination of GO at the licenced dose with chemotherapy in relation to liver toxicity have been alleviated by preliminary studies that established a safe dose of GO.¹¹ At the lower dose in combination, the risk of hepatotoxicity in patients who subsequently go on to transplantation seems to be unfounded based on the experience of the MRC AML15 trial, in which patients received a transplant as the third treatment course.

Molecular and immunophenotypic techniques are capable of detecting evidence of the leukemic clone in about 80% of patients at lower levels than are possible by morphology, standard metaphase cytogenetics, or interphase analysis (FISH). From a molecular point of view, the confirmation of detection by a single timepoint assay with a sensitivity of 1 × 10⁴ at the end of consolidation is associated with a high risk of subsequent relapse; however, the absence of detectable disease at this level does not predict cure—paradoxically, in acute promyelocytic luekemia (APL) most relapses that occurred happened in patients who were negative at this timepoint. This raises the issue of whether sequential monitoring using quantitative techniques would be more informative. Most work has been undertaken in patients with APL or core binding factor (CBF) leukemia who have a lower relapse risk and who may still be chemosensitive when relapse occurs. Monitoring by RQ-PCR in APL is technically and logistically demanding and expensive but is capable of predicting clinical relapse 3 to 6 months before it happens,¹² although this requires vigorous monitoring. Preliminary data in CBF leukemias suggest that a greater than 3 log reduction in transcript level after the first induction course is highly predictive of subsequent outcome.¹³ These data may become less relevant as treatment developments, such arsenic trioxide in APL, are introduced. The more frequent FLT-3 mutation theoretically could be a more ubiquitous minimal residual disease marker; however, studies of diagnostic and relapse paired samples indicate that it may be absent in relapse in patients with mutations at diagnosis, and vice versa. A more useful marker may be NPM1, which is frequently mutated in patients with a normal karyotype.¹⁴ This appears to be a stable marker that may be a more practical target for monitoring studies.^15,16 

Several studies have shown that a leukemia-associated phenotype can be characterized in a high proportion (> 85%) of patients with AML. There is little phenotypic shift so an individualized target can be defined for most patients. At least six studies in recent years in adult and pediatric AML have suggested that immunophenotypic detection of residual disease can predict relapse; however, several questions remain. Will a single timepoint assay be sufficient? What level of reduction at what timepoint is most useful? What additional prognostic value does this provide beyond that provided by other factors? To what extent is the technology influenced by treatment?

With the arrival of validated quantitation techniques of MRD detection, it becomes important to establish if this information provides prognostic information beyond that currently provided by existing factors, and whether knowledge of this information can improve the treatment outcome. MRD detection can identify individual patients at risk, but it has yet to be clearly demonstrated that clinical intervention when MRD is detected is more effective than treating hematologic relapse.

As discussed earlier, there is widespread evidence that the presence of a FLT-3 mutation is a powerful predictor of relapse,^17,18 although this may be modified by the coexistence of a mutation of NPM1^14,19,20 and the remaining uncertainty of the prognostic impact of the 7% of patients with a point mutation. This has two clinical implications: does this constitute an indication for transplantation? and what do FLT-3 inhibitors have to offer?

No consensus has yet emerged from two retrospective studies of the effect of transplantation on FLT-3–mutated patients. The retrospective study conducted by Gale and colleagues reports on the MRC experience²¹ (35 patients with a mutation and 135 with no mutation were analyzed on a donor vs no-donor basis). It was clear that a mutation, like all other adverse prognostic factors, is also an adverse prognostic factor in the recipients of transplants. In this “intent-to-treat” analysis, (donor vs no-donor) transplantation reduced the relapse risk and improved the DFS, but the OS was not significantly improved. These observations were not different from the FLT-3–negative group, and there was no significance in tests for heterogeneity. On this basis this study could not find evidence to recommend transplantation based on FLT-3 status alone. In a similar study by the Ulm group,²² the negative impact of FLT-3 mutation status was seen in patients who received chemotherapy or autologous transplantation, but not in the recipients of allografts. A major difference in the analysis was that the assessments were based on treatments given, rather than on an intent to treat that attempts to reduce selection of better-risk patients entering the transplantation option.

Because of the negative impact of FLT-3 mutations in AML, several agents have been identified by high-throughput screening that have FLT-3 inhibitory activity against cell lines that contain transfected mutants. They are effective in cytotoxicity assays against primary cells, and rodent models provide impressive evidence of in vivo activity. Whether advantageous or not, these agents are not FLT-3 specific and inhibit other tyrosine kinases (Table 2 ).

Phase 1 clinical studies have indicated tolerability at doses that result in in vivo activity, i.e., dephosphorylation of the FLT-3 kinase and receptor. In many cases the maximum tolerated dose has not been reached or is well above that required for in vivo activity. The two agents that we have in phase 3 development—PKC-412 (Novartis, Basel, Switzerland) and lestaurtinib (CEP-701; Cephalon, Frazer, PA)^23,–25—are indolocarbazoles, which are derivatives of staurosporine and a fermentation product of Nonomurea longicatena, respectively. PKC-412 (N-benzoyl-staurosporine) was developed several years ago as a protein kinase C inhibitor, but it also has PDGF-R and KIT inhibitory properties. It performed well in preclinical in vitro and in vivo models. In a phase 2 monotherapy study in recurring AML or high-risk MDS, modest clinical activity was demonstrated for this single agent.²³ Of 20 patients, 2 had a bone marrow blast reduction to less than 5% without complete peripheral blood recovery; 14 patients had a greater than 50% reduction in peripheral blood blasts. Inhibition of phosphorylation of FLT-3 was demonstrated in responding patients. Pharmacokinetic information suggests that the activity correlates with levels of the active metabolite CGP52541 rather than of PKC-412 itself. The clinical responses were typically 3 to 4 months in duration. Lestaurtinib developed from neurologic research as a TrkA inhibitor that was found to have potent FLT-3 inhibitory activity, but with some activity against KIT, PDGF-R, and FMS at higher concentrations and, more recently, against JAK-2. Like PKC-412, it performed well against mutated cell lines, primary samples and in mouse models, and was well tolerated in phase 1 development, with the maximum tolerated dose not being reached at a dose where there is clear evidence of inhibition of FLT3 phosphorylation. Two phase 1/2 studies have been undertaken. In the phase 1/2 trial conducted by Smith and colleagues,²⁴ 14 relapsed patients were treated with doses of 60 to 80 mg twice a day. One patient had a reduction in marrow blasts to less than 5%, while 5 patients had a reduction in peripheral blood blasts, but again responses were brief. A second phase 2 study adopted the same dosing schedule in a series of 29 untreated elderly patients without selecting by mutant status.²⁵ Treatment was administered for 56 days. In 3 of the 5 patients who had a mutation and in 5 of 22 with wild-type FLT-3, responses were seen with transient reduction in bone marrow or peripheral blood blasts. In general, the oral dosing schedule was well tolerated, which suggests that higher doses could be feasible. There are a number of reasons for the discrepancy between the preclinical models and the modest clinical activity. It is clear that some patients have disease that is insensitive in vitro, and in cases where this has been demonstrated, no clinical responses have been observed if there is insufficient plasma activity and/or it is not sustained. A feature of CEP-701 is that it is extensively protein bound in humans and not in mice, which may explain the contrasting responses in mouse and man. To get around this problem, Levis and colleagues developed a “plasma inhibitory assay” that tests samples of patient plasma against a cell line containing mutant FLT-3.²⁶ The extent to which the plasma contains inhibitory activity is directly correlated with the extent to which the FLT-3 receptor is dephosphorylated. Correlations with clinical response suggest that an adequate plasma inhibitory level can be defined as that which is capable of exceeding 85% inhibition of the receptor. It appears that the combination of in vitro sensitivity of the patients blasts combined with more than 85% inhibiton of FLT-3 phosphorylation in the bioassay indicate likely clinical response. Further limitations may be the inability to sustain plasma activity or the emergence of resistance due to mutations in the binding domain.

These two agents have now been taken forward for assessment in combination with standard chemotherapy. Sequencing of inhibitor and chemotherapy may be crucial both in terms of achieving synergy or avoidance of side effects. Initial results in relapsed patients look promising,^27,28 and first-line trials are underway. More recently, similar clinical activity has been seen with MLN-518.²⁹ 

Farnesylation catalyzes the post-translational modification of RAS and other transcriptionally active proteins by adding a fatty acid chain to the C-terminal–CAAX motif, which is necessary for attachment to the inner surface of the cell membrane. Apart from RAS, this farnesylation occurs with other proteins (e.g., RhoB, RAC, and TGF) that participate in signal transduction leading to cell proliferation. Although not prognostic, mutations of RAS proteins are present in 10% to 15% of patients with AML and therefore represent a molecular target. Subsequent clinical studies have demonstrated that a RAS mutation is not a prerequisite for response. Of the four classes of inhibitor compounds, the most advanced in development is tipifarnib, which is a member of the nonpeptidomimetic class. In preclinical assessment, it inhibited most cell lines tested. Phase 1 experiences suggested that when orally administered, non-hematologic toxicity is acceptable.³⁰ In a large phase 2 study in patients with recurring (n = 135) or refractory (n = 117) AML, oral administration of 600 mg twice a day for 21 days was tolerable, but the resulting 7% remission rate was disappointing.³¹ However, the pivotal unrandomized study in untreated patients demonstrated a 14% CR in older patients who were considered to be unfit for conventional chemotherapy.³² This is similar to what can be achieved with low-dose Ara-C. A prospective randomized trial in older patients versus best supportive care is now fully recruited. The favorable toxicity profile has raised the prospects of evaluating different schedules or combinations in ongoing studies, including the US Intergroup Study (E2902), which randomizes older adults in first CR or any patient in CR2 or beyond, to Tipifarnib, or not, as maintenance treatment.

Although CBF leukemias have a favorable prognosis and do not require stem cell transplantation, relapses do occur, particularly in those with inv16. Approximately 30% of these leukemias have a cKIT mutation that increases the risk of relapse.^33,34 This raises the specter of introducing agents that have KIT inhibitory activity, such as dasatinib or PKC-412. Of interest, in the MRC-AML15 trial the patients with CBF leukemia who had received GO had an 87% OS at 4 years, although it is not clear what proportion of KIT mutants were included,¹⁰ which may suggest that the adverse influence of c-KIT in this subset can be overcome.

Preliminary studies with agents inducing demethylation or histone deacetylase inhibition alone or in combination are under investigation and some activity has been demonstrated.

One explanation for treatment failure is that the treatments currently in use are cytotoxic to the tumor bulk, but not to the leukemic stem cells. Elegant experiments in NOD-SCID mice provide compelling evidence for the existence of a leukemic stem cell that is distinct from the normal stem cell.^1,2 This resides in the CD34⁺ CD38⁻ subpopulation, but may be distinguished from normal by expression of CD123, which is the IL3-Rα receptor.³ This has led to the development of a treatment schedule of the administration of IL-3 and a diphtheria toxin (DR383 IL-3), which satisfies the necessary selective critical in preclinical studies³⁵ and is now in phase 1 trials.³⁶ Monoclonal antibody therapy directed against IL3-Rα is entering clinical trial.

Jordan et al demonstrated that NF-κB was activated in the leukemic stem cell, but not the normal stem cell compartment. A parthenalide compound with NF-κB activity has been shown to be cytotoxic against AML cells and will shortly enter clinical trial.³⁷ Indirect inhibition of NF-κB by the inhibitors of IκB kinase are also in early-stage trials. Inhibition of mTOR has recently become a candidate target. In vivo models suggest that inhibition selectively inhibits self renewal of leukemic stem cells and may also synergise with Ara-C. Some clinical data on using mTOR inhibitors alone or in combination with chemotherapy are promising and feasible.^38,–40 

The next generation of clinical trials will exploit the emerging knowledge of the biology of AML, and several mechanisms now present themselves, either for refining the application of currently available schedules or by incorporating agents that appear to counteract the specific effects of known abnormalities. While it is perhaps overly optimistic to expect a “Glivec-like” impact from any single targeted agent in AML, careful clinical studies of these new agents, probably in combination with standard chemotherapy, may result in improved outcomes.

The only randomized trial so far reported of the areas discussed is the addition of GO to chemotherapy in younger patients. If longer follow-up confirms the benefit, this would establish a new standard of care. The other possibilities discussed have yet to be supported by randomized trial evidence.