Phase I and Pharmacokinetic Study of Bortezomib in Combination with Idarubicin and Cytarabine in Patients with Acute Myelogenous Leukemia

Purpose: Proteasome inhibition results in cytotoxicity to the leukemia stem cell in vitro. We conducted this phase I study to determine if the proteasome inhibitor bortezomib could be safely added to induction chemotherapy in patients with acute myelogenous leukemia (AML).

Experimental Design: Bortezomib was given on days 1, 4, 8, and 11 at doses of 0.7, 1.0, 1.3, or 1.5 mg/m² with idarubicin 12 mg/m² on days 1 to 3 and cytarabine 100 mg/m²/day on days 1 to 7.

Results: A total of 31 patients were enrolled. The median age was 62 years, and 16 patients were male. Nine patients had relapsed AML (ages, 18-59 years, n = 4 and ≥60 years, n = 5). There were 22 patients of ≥60 years with previously untreated AML (eight with prior myelodysplasia/myeloproliferative disorder or cytotoxic therapy). All doses of bortezomib, up to and including 1.5 mg/m², were tolerable. Nonhematologic grade 3 or greater toxicities included 12 hypoxia (38%; 11 were grade 3), 4 hyperbilirubinemia (13%), and 6 elevated aspartate aminotransferase (19%). Overall, 19 patients (61%) achieved complete remission (CR) and three had CR with incomplete platelet recovery. Pharmacokinetic studies revealed that the total body clearance of bortezomib decreased significantly (P < 0.01, N = 26) between the first (mean ± SD, 41.9 ± 17.1 L/h/m²) and third (18.4 ± 7.0 L/h/m²) doses. Increased bone marrow expression of CD74 was associated with CR.

Conclusions: The combination of bortezomib, idarubicin, and cytarabine showed a good safety profile. The recommended dose of bortezomib for phase II studies with idarubicin and cytarabine is 1.5 mg/m².

Current standard induction chemotherapy for AML involves an anthracycline combined with 1-β-d-arabinofuranosylcytosine (cytarabine, Ara-C), an antimetabolite cytotoxic primarily to cells in active cell cycle (1). The remission rate in previously untreated adults over 60 years old is ∼50%, whereas responses for patients with relapsed disease are considerably lower. The majority of older patients with AML ultimately relapse and die from their disease (2–4). It is possible that relapse is due to the regrowth of leukemia from quiescent leukemia stem cells (LSC) resistant to chemotherapy, and therefore targeting the quiescent LSC may result in more durable complete remissions (CR) in AML.

Both normal and malignant HSCs share common molecular pathways that promote self-renewal (5, 6). However, LSCs also have molecular features distinct from normal HSCs. For example, LSCs have elevated levels of the active form of nuclear factor-κB (NF-κB), a transcription factor with antiapoptotic activity in human cancer (7–9). One strategy to target NF-κB is to use proteasome inhibition, which prevents degradation of IκB, an endogenous negative regulator of NF-κB. Apart from its role in LSCs, NF-κB may mediate chemotherapy resistance. Both NF-κB and proteasome activity increase after treatment of leukemia cell lines with chemotherapy (10–12). Because the ubiquitin-proteasome pathway plays an important role in regulating the cell cycle, gene transcription, cellular adhesion, and other important neoplastic processes, proteasome inhibition impairs tumor growth through a variety of mechanisms (13, 14).

Data from in vitro experiments with the proteasome inhibitor MG-132 suggest that inhibition of the NF-κB pathway may result in LSC cytotoxicity (8). LSCs undergo apoptosis after exposure to MG-132, whereas normal HSCs remain viable. Furthermore, LSCs undergo increased apoptosis after exposure to MG-132 compared with when exposed to cytarabine. Most importantly, treatment of human AML bone marrow with the combination of MG-132 and idarubicin, at doses that spare normal HSCs, results in decreased leukemia cell engraftment in a mouse LSC xenograft model (15).

Bortezomib is a potent, reversible, and specific inhibitor of the proteasome approved for use in multiple myeloma. Whereas the effect of bortezomib on LSCs has not been explored in functional LSC assays, cytotoxicity to primary CD34+ AML cells has been shown (16). However, bortezomib has limited single-agent activity in patients with AML (17). We and others have reported that bortezomib interacts synergistically with anthracyclines and/or cytarabine to induce leukemic cell death in vitro (15, 18–22). We conducted a phase I study to define the safety and tolerability of bortezomib in combination with standard induction chemotherapy in adult patients with AML. We applied the bortezomib administration schedule used in multiple myeloma to standard AML induction chemotherapy consisting of idarubicin and cytarabine. Bortezomib was the first agent given to assess for possible infusional reactions.

Patient selection. All patients were at least 18 years of age and required to read and sign an informed consent document. Patients with AML by WHO criteria, excluding acute promyelocytic leukemia [t(15;17)], were eligible for this study (23). Patients with first and subsequently relapsed AML were eligible if their most recent documented remission period was at least 3 months. Patients with previously untreated AML were eligible if they were ≥60 years old. Patients may have had prior myelodysplasia or myeloproliferative disorders (MPD) except chronic myelogenous leukemia. Prior treatment for myelodysplasia or MPD was allowed. Prior chemotherapy and/or radiation therapy for another malignancy was permitted, as was prior autologous and/or allogeneic stem cell transplantation for AML. All patients were required to have a performance status of 0 to 3, total bilirubin of ≤1.5 × upper limit of normal (ULN), alanine aminotransferase and aspartate aminotransferase of ≤2.5 × ULN, and creatinine of ≤2.0 mg/dL. A left ventricular ejection fraction of at least 40%, as assessed by radionuclide ventriculogram or echocardiogram, was required. Patients were excluded if they had severe pulmonary or cardiac disease, ≥grade 2 peripheral neuropathy within 21 days before enrollment, or known current or previous central nervous system leukemia. This protocol was reviewed and approved by the Dana-Farber/Harvard Cancer Center Scientific Review Committee and Institutional Review Board, and informed consent was obtained from all patients. The study was done according to the tenets of the Helsinki protocol.

Treatment and study design. All patients received idarubicin (12 mg/m²) by rapid i.v. infusion on days 1 to 3 and cytarabine (100 mg/m²) by continuous i.v. infusion days 1 to 7. Bortezomib was given by rapid i.v. infusion on days 1, 4, 8, and 11 at approximately (±2 h) the same time each day. Dose levels evaluated were 0.7, 1.0, 1.3, and 1.5 mg/m². Bortezomib was given 1 h before idarubicin and cytarabine on day 1. The protocol specified reductions in subsequent doses of bortezomib if patients experienced bortezomib-induced neuropathy. Bortezomib was provided by Millennium Pharmaceuticals, Inc. Idarubicin and cytarabine were obtained through commercial sources. A bone marrow biopsy was done 15 days after the start of therapy, and patients with residual leukemia defined as ≥5% myeloblasts with a cellularity of >20% received reinduction with idarubicin (12 mg/m² for 2 days) and cytarabine (100 mg/m²/day for 5 days by continuous i.v. infusion). Bortezomib was not given during reinduction due to concerns of causing excessive thrombocytopenia.

Toxicities were assessed using the National Cancer Institute Common Toxicity Criteria scale, version 2.0. Dose-limiting toxicity (DLT) was defined as grade 3 or grade 4 sensory or autonomic neuropathy, grade 4 thrombocytopenia 42 days beyond the start of the most recent chemotherapy, grade 4 neutropenia 42 days beyond the start of the most recent chemotherapy, any grade 4 nonhematologic toxicity at any time (excluding toxicities secondary to neutropenia and sepsis), or any grade 3 nonhematologic toxicity (excluding neuropathy and toxicities secondary to neutropenia and sepsis) that did not resolve to grade 2 by 42 days beyond the start of the most recent chemotherapy. Infections due to neutropenia (including pneumonia with respiratory failure, hypotension with renal failure, and death when clearly secondary to sepsis) up to 42 days after the start of treatment were not considered DLT. Because hyperbilirubinemia, anorexia, and fatigue are common toxicities associated with standard induction chemotherapy for AML, the following were defined in the protocol as not being DLT events: (a) anorexia requiring TPN, (b) fatigue requiring bed rest, and (c) grades 2, 3, and 4 hyperbilirubinemia redefined as 1.5 to <10× ULN, 10.0 to 20.0× ULN, and >20× ULN, respectively. If a DLT was experienced, the study drug was held and the patient followed clinically. Postremission therapy was at the discretion of the treating physician.

Cohorts consisted of three to six patients. If no DLTs were observed within the first three patients in a cohort, dose escalation proceeded. If a single DLT was experienced in the first three patients, an additional three patients were enrolled. If <2 DLTs were observed in these six patients, dose escalation was permitted. If ≥2 DLTs were observed in a cohort, the prior dose was declared the MTD and six additional patients were enrolled to provide a greater degree of assurance that this dose was tolerable.

Response criteria. A bone marrow aspiration and biopsy was required 28 days after the start of the most recent induction cycle and, then, every 2 weeks until treatment response could be determined. All patients who received at least 1 day of treatment were evaluable for efficacy. Response was defined according to standard criteria with minor modifications (24, 25). CR was defined as absolute neutrophil count of ≥1500/μL, platelet count of ≥100,000/μL (independent of transfusions), and bone marrow blasts of <5%, regardless of bone marrow cellularity. CR without recovery of platelets (CRp) was defined as CR except that the untransfused platelet count ranged from 25,000 to 99,000/μL. Partial response was defined as meeting all criteria for CR except that the bone marrow contained 5% to 24% bone marrow blasts. All other patients were considered nonresponders.

Pharmacokinetic studies. Sampling to characterize the plasma pharmacokinetics of bortezomib was done. Blood samples were drawn from a peripheral vein before dosing and at the following times relative to the start of the bortezomib injection on days 1 and 8: 5, 15, 30, and 60 min and 2, 4, 6, and 24 h. Sample tubes were mixed by inversion and placed over ice until centrifuged (1,200-1,500 × g, 15 min, 4°C) within 10 min. Plasma was stored in polypropylene cryovials at −70°C until assayed.

The concentration of bortezomib in human plasma was determined by high-performance liquid chromatography with tandem mass spectrometric detection at Millennium Pharmaceuticals, Inc., as previously reported (26). During analysis of samples from this clinical trial, the interday accuracy of the assay for measuring quality control samples of bortezomib in human plasma at concentrations of 0.3, 10, and 16 ng/mL ranged from 95.7% to 98.7% of the known concentrations and the precision, calculated as the coefficient of variation, was 8.3% to 9.3%. Plasma concentration-time profiles of the drug for individual patients were fit to model-independent equations for bolus i.v. drug input with either biexponential or triexponential first-order elimination by weighted nonlinear regression using WinNonlin Professional version 5.0 software (Pharsight Corp.), as previously described in detail (27). Final values of the iterated variables in the best-fit equation were used to calculate all pharmacokinetic variables according to standard equations (28). Linear regression was used to assess dose-dependent trends in pharmacokinetic variables. The paired two-tailed t test was used to compare mean pharmacokinetic variables between the doses given on days 1 and 8 after logarithmic transformation of the data. P < 0.05 was the criterion for significance.

Microarray analysis. Pretreatment bone marrow pellets were collected from 27 of 31 patients enrolled in the study. High-quality RNA for gene expression profiling was obtained from 23 of these samples. RNA was extracted, and the quality assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies). Biotinylated probe was generated and hybridized to Affymetrix whole-genome expression microarrays (Human Genome U133 Plus 2.0 Array). After hybridization, a GeneChip Scanner 3000 with GeneChip Operating Software was used to assess the probe intensity. DNA chip analyzer (29) was used to normalize the Affymetrix gene array data and to obtain perfect match–only model–based expression intensities. An array with a median overall intensity was chosen as the baseline array against which other arrays were normalized at probe intensity level. DNA chip analyzer was used to perform an unsupervised analysis that consisted of gene filtering, that is, excluding genes that lacked sufficient variability across samples and hierarchical clustering of genes and samples.

Patient samples were then categorized into two groups according to clinical response. One class consisted of patients achieving a CR (n = 14), whereas the other included patients who achieved a CRp, partial response, or nonresponder (n = 9). To identify the genes whose expression patterns best distinguished patients achieving a CR from patients not achieving a CR, the permutation distribution of the maximum t statistic was analyzed using the Permax test (30). The customized program written for R is available online.⁶

To control the overall error rate, the Permax P value was calculated by comparing the observed t statistics for each gene from their log values to the permutation distribution of the largest t statistic obtained over the 54,674 genes. For each gene, the P value is the proportion of permutations with the maximum t statistics over all genes greater or equal to the observed value for a particular gene. A test declaring as significant any genes with P of <0.05 guarantees that the chance of any false positives being selected is <5%.

Patients. A total of 31 patients were enrolled between November 2003 and September 2005. Demographic characteristics and type of AML are depicted in Table 1. The median age for all patients was 62 years. The median age among patients with previously untreated AML was 65 years (22 patients, range 60-70 years). Seven of these 22 patients had an antecedent hematologic disorder, and one patient previously received radiotherapy for breast cancer. Of these seven patients, six had prior myelodysplasia (age range, 60-70 years; median, 5.5 years) and one had prior myeloproliferative disease (polycythemia vera, age 62 years). The median age among patients with relapsed AML was 61 years (9 patients, range 42-75 years). The median WBC count was 3.15 × 10³/μL (range, 0.49-62.61), and median lactate dehydrogenase was 308 (range, 131-866) for all patients.

The nine patients with relapsed AML had received a median of three prior cycles of chemotherapy (range, 1-9), including consolidation cycles of chemotherapy; eight were in first relapse and one was in third relapse. The median duration of the most recent CR in relapsed patients was 23 months (range, 4-48 months). Four of the nine patients in this group had previously undergone stem cell transplantation (SCT; three autologous, one ablative allogeneic). Of the three patients previously treated with autologous SCT, one received autologous SCT in first relapse followed by matched unrelated donor nonmyeloablative SCT in second relapse. This patient was in third relapse upon enrolling to the current protocol.

Among patients with prior myelodysplasia, only one received prior chemotherapy (azacitidine for three cycles). One patient with relapsed AML was previously treated with five cycles of chemotherapy consisting of induction, reinduction, followed by three cycles of intermediate dose cytarabine consolidation. The patient with prior polycythemia vera was treated with hydroxyurea for ∼20 years before developing AML.

Dose escalation and DLTs. We treated 9 total of 6 patients with 0.7 mg/m², 6 with 1.0 mg/m², 9 with 1.3 mg/m², and 10 with 1.5 mg/m² bortezomib. At the 0.7 mg/m² dose of bortezomib, one of the patients experienced a DLT consisting of prolonged, grade 4 thrombocytopenia beyond day 42. This occurred in a 65-year-old man with preceding myelodysplasia who had previously received three cycles of azacitidine. Although his bone marrow blasts cleared, he had persistent grade 4 neutropenia and thrombocytopenia 42 days beyond induction chemotherapy without evidence of disease. His response was classified as nonresponder on the basis of his low peripheral blood counts. Myelodysplasia ultimately returned, and his thrombocytopenia did not improve. An additional three patients were enrolled at this dose, and none experienced DLTs.

There was one DLT consisting of prolonged, grade 4 neutropenia and thrombocytopenia lasting 42 days beyond the most recent induction in patients treated at 1.0 mg/m² bortezomib. This occurred in a 68-year-old woman diagnosed with breast cancer 2 years before development of AML. Her breast cancer was treated with lumpectomy, radiotherapy, and hormonal blockade but without chemotherapy. Persistent AML was noted on the day 15 bone marrow biopsy, and she received reinduction therapy. She had myelosuppression lasting 42 days after reinduction without evidence of disease, which was considered a DLT. Then, a bone marrow biopsy conducted 6 days later, 48 days after reinduction, indicated persistent AML, and she was considered a nonresponder. Three additional patients were enrolled at this dose level, and no further DLTs were noted.

Three patients were then enrolled at 1.3 mg/m² of bortezomib, and because there were no DLTs observed, an additional six patients were enrolled and treated at 1.3 mg/m² bortezomib. No DLTs were noted in any of the nine patients treated at this dose level.

Because there were no DLTs at 1.3 mg/m² bortezomib, the protocol was modified to evaluate bortezomib at 1.5 mg/m². We did not plan to test doses of bortezomib above 1.5 mg/m² in this regimen due to safety concerns, as higher doses of bortezomib given twice weekly have resulted in dose-limiting thrombocytopenia (31). A total of 10 patients were treated at 1.5 mg/m² bortezomib. Among the initial three patients, one patient developed febrile neutropenia and died on the 10th day of treatment. This patient had previously undergone SCT for AML 4 years earlier. The adverse event was not considered a DLT because the autopsy revealed disseminated fungal infection thought related to previous immunosuppressive therapy. The patient was replaced. Within the first three evaluable patients treated at 1.5 mg/m² bortezomib, a 57-year-old man with AML in fourth relapse with prior regimens, including autologous SCT and nonmyeloablative unrelated donor SCT, was noted to have prolonged fatigue lasting beyond day 90 of treatment. The fatigue was considered to be a DLT despite ongoing graft-versus-host disease and hepatosplenic candidiasis, and an additional three patients were enrolled. No DLTs were noted within the second cohort of three patients nor the six patients who were subsequently enrolled. Thus, among 10 patients treated with bortezomib (1.5 mg/m²), a single DLT consisting of prolonged fatigue lasting beyond day 42 occurred in a patient with graft-versus-host disease and hepatosplenic candidiasis. An MTD was not formally established based upon the definition set forth in the protocol as doses of bortezomib higher than 1.5 mg/m² were not evaluated in this study.

Adverse events. Adverse events were documented in all patients at all dose levels. Grade 3 and greater toxicities are presented in Table 2. Grade 3 or greater hematologic toxicities consisting of neutropenia and thrombocytopenia characteristic of induction chemotherapy for AML were observed in all patients. Two patients who had previously undergone allogeneic stem cell transplantation experienced graft-versus-host disease, one grade 3 and one grade 4. One patient died on day 10 from sepsis as described above. This was not thought to be related to bortezomib.

In the 17 patients who achieved CR without requirement for reinduction, the median time from day 1 of induction chemotherapy to absolute neutrophil count of ≥1,000/μL was 26 days (range, 20-46) and to platelets of ≥100,000/μL was 27 days (range 21-56). In the two patients who required reinduction and achieved CR, the median time to absolute neutrophil count of ≥1,000/μL was 26.5 days (range, 25-28) and the median time to achieve platelets of ≥100,000/μL was 37.5 days (range, 34-41) from date of reinduction.

Nonhematologic grade 3 or greater toxicities included hypoxia 12 (38%), hyperbilirubinemia 4 (13%), and elevated aspartate aminotransferase 6 (19%), and alanine aminotransferase 3 (9%). There was one episode of National Cancer Institute Common Toxicity Criteria grade 4 hyperbilirubinemia (>10× ULN) which occurred in a patient treated at 1.3 mg/m² bortezomib. There were seven patients (23%) with grade 3 rashes. One patient treated with 1.3 mg/m² bortezomib had grade 4 sinus tachycardia, which resolved and, in this setting, had a slight elevation in cardiac troponin. Another patient at this level had a grade 4 nonmalignant pleural effusion.

Clinical response. Of the 31 patients treated on this study, 19 (61%) achieved CR, 3 (9%) achieved remission without CRp, 2 achieved partial remission, and 7 (23%) patients failed to respond.

The clinical response achieved in patients enrolled on this study is indicated in Tables 3A and B. Of 22 patients with previously untreated AML, 13 (59%) achieved CR. Of the seven patients previously diagnosed with prior myelodysplasia (MDS) and MPD, two (29%) achieved CR. A total of six of nine patients (67%) with relapsed AML achieved CR. Within this group, three of the four (75%) patients with prior SCT achieved CR.

In nine patients who underwent allogeneic stem cell transplantation (seven reduced intensity and two myeloablative) at some time after treatment on this study, the median number of days to neutrophil engraftment (absolute neutrophil count, >500/μL) was 13 days (range, 0-24) and median time to platelet engraftment (untransfused platelet count, >20,000/μL) was 14 days (range, 0-55).

The median overall survival for patients on this study was 12.4 months (range, 10 days to 36.5 months). For the 22 patients achieving CR/CRp, the median disease-free survival was 15.3 months (range, 2.0-36.5+ months) and median overall survival was 17.6 months (range, 3.2-36.5+ months). For the nine patients who proceeded to SCT, the median disease-free survival and overall survival were 17.6 months (range, 4.6-33.3 and 5.1-33.3 months, respectively).

Pharmacokinetics. The pharmacodynamics of bortezomib on 20S proteasome activity in whole-blood lysates have been studied extensively during its clinical development (32). In contrast, information on the pharmacokinetic behavior of bortezomib is extremely limited (26, 33). We therefore sought to define the pharmacokinetics of bortezomib when given in this regimen. Because anthracyclines are metabolized by an aldoketoreductase (34) and cytarabine by cytidine deaminase (35), their pharmacokinetics are unlikely to be altered by bortezomib (36), which is hepatically metabolized by cytochrome P450 enzymes (37, 38). Therefore, the pharmacokinetic studies were limited to that of bortezomib.

Samples to define the plasma pharmacokinetics of bortezomib were obtained from 30 patients after administration of the first dose of drug and from 29 patients during treatment with the third dose. Pharmacokinetic variables could not be estimated for three patients receiving the initial dose of 0.7 mg/m² because bortezomib was undetectable in plasma samples obtained 24 h after dosing, precluding definition of the terminal disposition phase. Mean plasma concentration-time profiles of bortezomib for the group of 10 patients receiving doses of 1.5 mg/m² on days 1 and 8 of the first cycle of therapy are shown in Fig. 1. Plasma levels of the drug decayed in a triexponential manner in the large majority of patients. Plasma profiles were best described by a biexponential equation for only four patients after administration of the first dose and two patients for the dose given on day 8. Mean values of the pharmacokinetic variables for the groups of patients evaluated at each dose level are presented in Table 4. Linear pharmacokinetic behavior was supported by the absence of a relationship between the given dose of bortezomib and the total body clearance, as illustrated in Fig. 2 for the first and third doses, as well as all other pharmacokinetic variables. Overall mean values of the pharmacokinetic variables for both doses of the drug in the entire group of patients are listed in Table 5 together with results from their statistical comparison.

As evident in the mean plasma profiles shown in Fig. 1, there was a distinct difference in the pharmacokinetic behavior of bortezomib when given alone on day 8 compared with the initial dose that was given 1 h before the bolus i.v. injection of idarubicin on day 1. Specifically, the overall mean total body clearance for the third dose of bortezomib (18.4 ± 7.0 L/h/m²) was 56.1% less than that for the first dose (41.9 ± 17.1 L/h/m²; P < 0.01), resulting in a comparable increase in the mean residence time from 11.4 ± 5.7 h for the first dose to 16.2 ± 6.9 h for the third dose (P = 0.02). The decreased clearance was not associated with a change in the peak concentration of drug in plasma, the apparent volume of distribution of the central compartment, or even the half-lives of the initial and terminal disposition phases (Table 5). The steady-state and total body apparent volumes of distribution were the only other variables exhibiting a significant difference between the two doses, which were respectively 37.2% and 51.1% lower for the third dose than the first, thus suggesting that the effect was associated with decreased uptake of the drug by slowly perfused tissues. The increased concentration of the drug in plasma beginning ∼1 h after dosing cannot be attributed to classic drug accumulation. This was conclusively shown by simulating the plasma concentration-time course of bortezomib for the days 1, 4, 8, and 11 repeated dosing schedule using variables describing the best-fit equation for the mean plasma profile for the first dose in the 1.5 mg/m² treatment group. The predicted time course after administration of the third dose was almost perfectly superimposable with that for the initial dose of the drug.

The median body surface area of the patients was 1.95 m² (range, 1.44-2.27 m²), and the mean total body clearance of bortezomib (dose 3) was 32.8 ± 13.0 L/h. Unnormalized total body clearance was weakly correlated with body surface area (r = 0.33, P = 0.084). There was no association between the total body clearance (unnormalized) and pretreatment creatinine clearance estimated by the Cockroft-Gault equation, which ranged from 29 to 136 mL/min (r = 0.10). The total body clearance (normalized, dose 3) was not associated with patient age, although there were only three patients of ages 70 years or older. In addition, there was no statistically significant difference between the mean total body clearance of the drug (dose 3) in patients grouped according to gender (males, 20.1 ± 7.4 L/h/m²; females, 17.0 ± 6.6 L/h/m²; P = 0.24).

Assessment of gene expression using microarray analysis. To define the existence of molecular markers that predict response to the combination of idarubicin/cytarabine/bortezomib, pretreatment bone marrow samples from patients treated on the current study were collected and stored. Microarray analysis of gene expression was conducted, and hierarchical clustering analysis was done. Hierarchical clustering failed to categorize patient samples according to clinical response (data not shown). Patient samples were then categorized into two groups according to clinical response [CR (n = 14) in one class, CRp, partial response, and nonresponder (n = 9) in another]. There were no statistically significant (P < 0.05) differences in gene expression profiles between bone marrow responders and nonresponders. However, we identified one particular gene warranting further study, namely CD74, which was expressed at a 6-fold increase in patients achieving a CR compared with patients that did not achieve a CR (P = 0.06).

Proteasome inhibition represents a novel anticancer therapy to which LSCs seem particularly sensitive in in vitro and in vivo models of leukemia. Bortezomib has shown cytotoxicity in primitive leukemia cells (16). In addition, bortezomib has additive-synergistic cytotoxic activity when combined with anthracyclines and cytarabine in primary AML cells and cell lines (16, 19, 39). In this phase I dose escalation study, we show the safety of combining bortezomib with standard induction chemotherapy in patients with AML.

Bortezomib was well tolerated in this combination, including the highest dose level evaluated, 1.5 mg/m². Higher doses of bortezomib were not tested due to concerns for severe thrombocytopenia (31). We were unable to determine the MTD in this study as the highest studied dose of bortezomib (1.5 mg/m²) was not associated with significant toxicity. In particular, there were no irreversible cardiopulmonary toxicities attributable to bortezomib. Furthermore, there were no neurologic toxicities despite coadministration with the potentially neurotoxic agent, cytarabine.

Thrombocytopenia has been frequently observed after multiple administrations of bortezomib. In this study, there were only two cases of prolonged myelosuppression categorized as DLTs; both were at low doses of bortezomib and both patients had reemergence of their underlying hematologic malignancies shortly after protocol treatment, making bortezomib unlikely to be the cause of the prolonged thrombocytopenia. Also in this study were three patients who achieved CRp. One patient had been treated at 0.7 mg/m², the second at 1.0 mg/m², and the third at 1.3 mg/m² bortezomib. Two of these patients had a prior history of myelodysplasia. One relapsed 79 days after induction, whereas the second proceeded to nonmyeloablative double cord blood transplantation and is alive and disease-free 1,000 days after induction. The third patient had relapsed AML before study entry. The disease-free survival for this patient was 202 days. These cases make it difficult to determine whether bortezomib contributes to delayed platelet recovery in patients who would have otherwise achieved CR. However, the incidence of CRp did not seem to increase with the dose of bortezomib.

Pulmonary toxicities have been reported with bortezomib (40), although there were only four episodes of grade 4 pulmonary toxicities in this study. Two episodes involved the development of pulmonary infiltrates, whereas there was one episode of grade 4 pleural effusion and one grade 4 hypoxia. There were 11 episodes of transient grade 3 hypoxia defined as oxygen desaturation requiring supplementary oxygen at rest; however, there were no pulmonary DLTs. Whereas hypoxia is not frequently reported in AML induction studies, the frequency of dyspnea was similar to those seen in other studies (41, 42). Therefore, it is difficult to determine if bortezomib contributed to the pulmonary toxicities experienced by patients on this study above the hypoxia caused by fluid shifts and infections in patients undergoing induction chemotherapy for AML.

In both prior investigations of bortezomib pharmacokinetics, pharmacokinetic sampling was only done for the first dose of the first cycle of therapy (26, 33). The mean total body clearance of the drug in 13 adult patients receiving the maximum tolerated dose of 1.6 mg/m² for the weekly administration schedule was 31.7 ± 26.3 L/h/m² with an apparent biological half-life of 11.9 ± 10.4 h (33). In five children treated with doses of 1.3 or 1.7 mg/m², bortezomib exhibited a mean total body clearance of 44.6 ± 7.7 L/h/m² and a biological half-life of 14.8 ± 6.2 h (26). The apparent volume of distribution was large, with mean values of 545 ± 266 L/m² in adult patients and 920 ± 283 L/m² in children. In our study, overall mean values of the total body clearance, biological half-life, and apparent volume of distribution were 41.9 ± 17.1 L/h/m², 14.2 ± 5.5 h, and 856 ± 506 L/m², respectively, and were in good agreement with prior studies. Whereas there was no correlation between the total body clearance of bortezomib and patient body surface area in the initial phase I study of the drug (33), we found a weak correlation for the first dose of bortezomib, supporting the rationale for adjusting the dose of this drug to patient body surface area. Consistent with an evaluation based upon the extent of 20S proteasome inhibition in blood (32), the total body clearance of bortezomib and creatinine clearance were uncorrelated, indicating dose modification of bortezomib is unnecessary in patients with moderate renal dysfunction.

Sampling was also done to obtain comparative pharmacokinetic data from 26 of these patients during treatment with the third dose of bortezomib. The overall mean total body clearance for the third dose of bortezomib (18.4 ± 7.0 L/h/m²) was 56% less than that for the first dose (41.9 ± 17.1 L/h/m²; P < 0.01). This occurred without a significant change in either the peak concentration of drug in plasma or the apparent terminal phase half-life, whereas the total body apparent volume of distribution was significantly lower for the third dose than the first (419 ± 227 versus 856 ± 506 L/m²; P < 0.01). This finding suggests that tissue binding sites become saturated upon repeated dosing every 3 or 4 days, with a greater fraction of the dose remaining in systemic circulation compared with the initial dose, resulting in higher plasma concentrations after distribution equilibrium is achieved.

Due to the lack of a control population of patients undergoing treatment with idarubicin and cytarabine but without bortezomib, pharmacodynamic studies in AML cells were not conducted. However, peripheral blood CD34+ myeloblasts obtained from one patient treated with 1.3 mg/m² bortezomib showed a decrease in nuclear NF-κB within 1 h of initiating therapy, with significant reductions seen at 6 and 24 h (courtesy of Dr. Craig Jordan; data not shown). Additionally, peripheral blood myeloblasts showed a 50% decrease in proteasome activity within 1 h of bortezomib administration (data not shown).

A CR rate of 59% was observed in previously untreated older adults, in line with published response rates for previously untreated older adults with AML (41, 42). A CR rate of 67% was observed in patients with relapsed AML, a group with expected CR between 20% and 40% depending on disease-free interval (43). Further studies will be required to determine if the addition of bortezomib to induction chemotherapy improves the response rate over chemotherapy alone.

There was an association between expression of CD74 in bone marrow cells and CR in this study. CD74 is a type II integral membrane protein located on chromosome 5q32 and functions as a chaperone molecule for the class II MHC. CD74 is expressed in numerous tumor types (44). CD74 is also a high-affinity binding protein for the proinflammatory cytokine, macrophage migration inhibitory factor (45). Stimulation of CD74 by migration inhibitory factor results in cell proliferation, activation of NF-κB, and cell survival (46, 47). Therapies which block CD74 have shown therapeutic potential in non–Hodgkin's lymphoma and multiple myeloma (48, 49). Therefore, we hypothesize that elevated bone marrow CD74 expression identifies a subset of leukemia in which NF-κB is operative and that these patients have increased sensitivity to NF-κB inhibition via proteasome inhibition. Future studies will define the potential role of CD74 in myeloid malignancies and evaluate its relevance to bortezomib-induced response.

In summary, our study suggests that bortezomib is well tolerated in combination with idarubicin and cytarabine. The recommended dose of bortezomib in future phase II studies combining bortezomib with idarubicin and cytarabine is 1.5 mg/m².

The authors wish to thank Drs. Monica Guzman, Craig Jordan, Diana Howard, George Mulligan, and Bill Trepicchio for helpful discussions.