Phase 2 study of imexon, a prooxidant molecule, in relapsed and refractory B-cell non-Hodgkin lymphoma

Cellular reactive oxygen species (ROS), the by-products of normal mitochondrial metabolism, are of fundamental importance in the regulation of multiple signal transduction pathways in normal cells.¹  However, oxidative stress, defined as a cellular imbalance of oxidants over antioxidants, can also disrupt redox signaling causing molecular damage.²  This ability for mutagenic ROS to damage DNA, enabling carcinogenesis, has been recognized as one of the hallmarks of cancer.^3,4  Cancer cells appear to exist under conditions of increased oxidative stress as compared with their normal cellular counterparts. Given the cell’s oncogenic dependence on increased oxidative stress, it relies on antioxidant defenses, including the thioredoxin (TXN) and glutathione systems, to protect against the ROS-induced cell death mediated by oxidation of cellular proteins, carbohydrates, lipids, and DNA. It is hypothesized that this differential dependence of cancer cells compared with that of normal cells on increased oxidative stress and antioxidant stress support systems can be exploited therapeutically. It is anticipated therefore that cancer cells will be more sensitive to further increases in oxidative stress and decreases in the activity of cellular antioxidant defenses. Based on this hypothesis, we propose to test this therapeutic strategy by pharmacologically increasing the cellular oxidative state.^5-7 

Imexon is a novel agent that increases oxidative stress in target cells. It is a cyanoaziridine compound investigated as an anti-neoplastic agent due to its ability to bind cellular thiols, depleting stores of cysteine and glutathione and subsequently increasing cellular ROS.⁸  This increase in oxidative stress increases outer mitochondrial membrane permeability allowing cytochrome c release and caspase 3 and 9 activation with a resultant induction of apoptosis.^9-11  Additionally, imexon-induced disruption of the redox balance in the endoplasmic reticulum has been demonstrated to reduce protein translation and inhibit cell growth.¹²  Early preclinical studies noted a selective inhibitory effect on B lymphocytes in several disease models as well as demonstrated cytotoxicity in lymphoma cell lines.^13,14  Phase 1 trials demonstrated imexon to be well tolerated, associated with symptoms of reversible fatigue and gastrointestinal side effects with a relative lack of severe myelosuppressive toxicities.^15,16  Although few lymphoma patients were enrolled in the early phase 1 studies, a partial response (PR) was noted in a patient with follicular lymphoma (FL).

The current study is a phase 2 evaluation of imexon as a single agent in patients with relapsed or refractory B-cell non-Hodgkin lymphoma (NHL) given the potential oncogenic role of cellular redox across lymphoma histologies. In addition to exploring the potential therapeutic efficacy of imexon, we aimed to determine whether a redox-specific biomarker could predict clinical activity. However, no reliable method exists for measuring oxidative stress directly and accurately in tumor specimens. Previously, we developed a gene expression–based approach as an alternative for evaluating oxidative stress in diagnostic specimens.¹⁷  A mathematically derived redox score, incorporating 3 major enzymatic systems of cellular antioxidant defenses, correlated with clinical outcomes of a well-defined diffuse large B-cell lymphoma (DLBCL) population. Laboratory correlates in the current study tested whether the redox score measured in either pretreatment tumor specimens or peripheral blood mononuclear cells (PBMCs) or levels of a urine marker for oxidative damage could predict for a response to imexon. The results detailed below demonstrate imexon to be tolerable in this patient population, suggest that modulating oxidative stress is a viable therapeutic strategy in lymphoma, and support the ability of antioxidant-related gene expression to predict for imexon responsiveness clinically.

The primary end point of this multicenter phase 2 trial was the response rate in patients with relapsed or refractory NHL following administration of imexon. Secondary end points included safety, progression-free survival, and correlative studies intended to identify a biomarker predictive of clinical activity. Institutional review boards approved the protocol at each participating site, and informed written consent was obtained from all patients prior to enrollment. The study was conducted in accordance with the Declaration of Helsinki. All authors had access to the primary clinical trial data. The study was registered prior to enrolling patients (ClinicalTrials.gov NCT01314014).

Patients aged 18 years and older were eligible if they had a diagnosis of B-cell NHL. Acceptable World Health Organization–defined histologies included DLBCL, follicular, mantle cell, Burkitt, marginal zone, lymphoplasmacytic, small lymphocytic as well as transformed FL.¹⁸  Patients must have received at least 1 prior chemotherapy regimen for lymphoma and have bidimensionally measurable disease. Baseline laboratory parameters included absolute neutrophil count (ANC) ≥1500 cells/mm³, hemoglobin ≥10 g/dL, platelet count ≥75 000 cells/mm³, and adequate renal and hepatic function. A G6PD level at least at the institutional lower limit of normal was required as individuals with G6PD deficiency might suffer excess toxicity from a prooxidant agent. Patients who had received antibody or chemotherapy in the preceding 4 weeks, or had other active malignancies, central nervous system lymphoma, HIV positivity, or other active infection were excluded.

Baseline evaluation included history and physical examination, laboratory evaluations, bone marrow biopsy, and imaging by computed tomography (CT), position emission tomography (PET), or magnetic resonance imaging (MRI). Imexon was administered at 1000 mg/m² IV over 60 minutes daily for the first 5 days of each 21-day treatment cycle. Treatment continued until disease progression or unacceptable toxicity. Adverse events were assessed at baseline, throughout the trial, and during the 28-day period after treatment discontinuation and were graded by the National Cancer Institute Common Terminology Criteria for Adverse Events (version 4.0).

Imexon was held for grade 3 or 4 hematologic or nonhematologic toxicity and resumed with recovery to at most grade 1. If treatment was delayed for >4weeks, patients were removed from protocol. Dose adjustments were mandated for grade 3 or 4 neutropenia lasting >7 days, neutropenic fever, grade 3 thrombocytopenia with clinically significant bleeding, or platelet count <25 000/mm³ at any time. For these events, doses were decreased by 25% after the first episode and by another 25% after a second episode. Antiemetic prophylaxis with 5-HT3 antagonists without glucocorticoids was administered 30 minutes prior to each imexon infusion.

CT, PET, or MRI scans for the assessment of objective tumor responses were performed at baseline, after cycle 2, and every 3 cycles thereafter until disease progression. Standard response criteria were used for classification of objective tumor responses.¹⁹  Progression-free survival was calculated from the first dose of study drug to the first documentation of disease progression, death regardless of cause, or change in therapy due to disease progression, whichever occurred first. Patients who were alive and progression-free at the time of final data analysis were censored at the time of their last assessment. If disease progression did not occur by the end of treatment, patients were evaluated with physical examination and laboratory and imaging studies every 3 months until progression.

First morning urine samples were collected on day 1 prior to imexon infusion and on day 5 during cycle 1. Urinary levels of 8-iso-prostaglandin F2alpha (8-isoP) were measured (Eicosanoid Core Laboratory, Vanderbilt University, Nashville TN) as previously described²⁰  and normalized to creatinine levels in the samples.

Messenger RNA (mRNA) transcripts were evaluated in banked pretreatment tumor tissue samples (submitted slides or formalin-fixed paraffin-embedded [FFPE] blocks) and from PBMCs isolated from peripheral blood samples collected prior to and 3 hours after the first treatment infusion and prior to the cycle 1 day 5 infusion. Additional FFPE samples were obtained from the lymphoma tissue bank at the University of Arizona to establish redox scores in nonmalignant lymphoid tissue, FL, and DLBCL. Six specimens of each diagnosis, based on the pathology reports, were obtained. Multiple cuts of each FFPE specimen were made. One cut of each tumor specimen was stained with hematoxylin and eosin (H&E) and examined by a pathologist to identify the area of tumor involvement. This area was macrodissected from the unstained cuts and used for the gene expression analyses. Twenty-two genes important for cellular protection against oxidative stress were analyzed along with 16 genes previously associated with outcome in NHL as well as 4 immune cell surface markers. Included were 13 genes used to generate a redox signature score.¹⁷  Analyses were performed at HTG Molecular Diagnostics using a quantitative nuclease protection assay (qNPA) on a microarray plate as previously described.²¹  Briefly, the unstained 5-μm-thick sections were lysed in buffer provided by the manufacturer and hybridized to target-specific probes. Single-stranded nucleic acids (unbound probe or mRNA) were removed with S1 nuclease. After alkaline hydrolysis to remove mRNA, the target probes were detected based on binding to oligonucleotides assembled on a custom HTG ArrayPlate. Gene expression values, normalized to housekeeping genes, were derived using HTG software.

The trial was originally planned with 2 parallel, truncated Simon 2-stage designs, 1 for aggressive histologies and another for FL. Each study was designed with 82% power to detect a response rate of 30% vs 10% using a 1-sided 0.10 level test. Up to 12 evaluable patients were to be enrolled in the first stage. With 2 or more observed responses, a total of 18 evaluable patients were to be enrolled in each group. Truncation implied that the first-stage futility check would be met as soon as the second response was observed, and furthermore that the study could be concluded early for efficacy as soon as the fourth response was observed. Both groups passed the first-stage futility check, each with at least 2 PRs among the first 9 patients. Due to sponsor resource constraints, the follicular cohort was initially expanded to include additional indolent histologies. Subsequently, with the closure of AmpliMed corporation, both studies were terminated early after an additional 2 patients were enrolled in each group, for a total of 22 patients: 11 aggressive and 11 indolent. The data from both groups were then pooled for final analysis via a single 0.10 level 1-sided exact binomial test providing 76% power to detect a 30% vs 10% response rate. Separate 0.10 level 1-sided exact binomial tests for each group were also performed, with the follicular patients providing 54% power and the aggressive cohort providing 69% power.

Continuous variables were descriptively summarized by the mean, standard deviation, median, minimum, and maximum. Counts and proportions were tabulated for categorical variables. Censored time-to-event data were summarized via Kaplan-Meier estimates of survival.

Comparisons of gene expression levels and urine 8-isoP differences for responders vs nonresponders were made using equal variance 2-sample t tests. Both Bonferroni corrections and Benjamini-Hochberg false discovery rates (FDRs) were considered as adjustments for multiple comparisons over genes, and they yielded identical qualitative results. The Pearson product-moment correlation was used to test for a trend in an aggregate gene expression redox score and urine 8-isoP differences across the 3 response categories.

Twenty-two patients were enrolled between June 2011 and March 2013. Baseline clinical characteristics and prior therapies are detailed in Table 1. The median age was 64 years and 59% of patients were male. Histologies included follicular grades 1/2 (n = 9), DLBCL (n = 5), mantle cell (n = 3), transformed follicular (n = 2), small lymphocytic (n = 2), and Burkitt (n = 1) lymphoma. The median number of prior therapies was 4 (range 1-8), and these included CHOP (cyclophosphamide, doxorubicin, vincristine, prednisone)–like regimens (n = 20), multiagent salvage regimens (n = 8), high-dose chemotherapy, and stem cell transplantation (n = 9). All patients had received previous anti-CD20 antibody therapy. Six patients were refractory to their most recent treatment regimen. Six and 2 patients had bulky disease as defined by the largest nodal diameter being >5 cm and 10 cm, respectively. Twelve patients had lactate dehydrogenase (LDH) greater than the institutional upper limit of normal at enrollment.

Adverse events regardless of attribution are detailed in Table 2. The most common grade 3 and higher adverse event deemed imexon-related included anemia, occurring in 7 patients (32%). Additional grade 3 or greater imexon-related events included fatigue, thrombocytopenia, vomiting, neutropenia, respiratory infection, dehydration, dyspnea, and muscle weakness in 2 patients each (9%) as well as anorexia, hypotension, weight loss, creatinine increase, and rash occurring in 1 patient (5%) each. Two patients died while on study, 1 of progressive DLBCL after receiving 2 cycles of therapy and 1 of a septic event occurring on day 3 of cycle 1. One patient with heavily pretreated FL who had previously undergone autologous stem cell transplantation 6 years earlier, developed acute myeloid leukemia 2 years after completing 6 cycles of imexon therapy.

The most common adverse event of any grade was fatigue. This occurred in 91% of patients with 64% of these events labeled treatment related, although only 9% were grade 3 in severity. Gastrointestinal events included diarrhea in 73% of patients, nausea (68%), constipation (41%), anorexia (36%), vomiting (36%), and abdominal pain (22%). Eighty-seven cycles were administered during the course of the study; a median of 2.5 cycles (range 1-13) was administered per patient. Dose changes included 10 dose omissions, 5 dose delays, and 1 dose modification.

Twenty patients were evaluable for response. Two patients were removed from protocol treatment prior to the first radiographic assessment due to sepsis and diarrhea with subsequent worsening of pretreatment renal insufficiency. The overall response rate for evaluable patients was 30% (P = .011; 90% confidence interval [CI], 14%-51%). All 6 responding patients achieved a PR; an additional 35% of patients achieved stable disease (SD). Response rates by initially intended analysis were: FL, 4 of 9 (44%, P = .008; 90% CI, 17%-75%); aggressive lymphoma, 2 of 11 (18%, P = .303; 90% CI, 3%-47%). In the aggressive cohort, both responses occurred in DLBCL patients. Kaplan-Meier estimates of progression-free survival for all patients, and for patients with FL are shown in Figure 1.

Twenty-one patients provided urine samples on days 1 and 5 of cycle 1 for measurement of 8-IsoP. The median normalized ratio of 8-IsoP to creatinine level on day 1:day 5 was 2.06:2.11 for partial responders, 1.77:1.63 for patients achieving SD, and 1.80:2.25 for those having disease progression. When comparing the day 1 to 5 changes between the response categories, the differences were not statistically significant.

To determine whether gene expression in tumors impacted response to therapy, pretreatment tumor biopsies were identified in 13 patients. The various NHL histologies enrolled were represented; follicular grade 1/2 (n = 5), DLBCL (n = 3), mantle cell (n = 2), transformed (n = 1), small lymphocytic (n = 1), and Burkitt (n = 1) lymphoma. Of these patients, 2 had achieved PRs, 5 had SD, and 6 patients experienced disease progression with imexon therapy. Transcript levels were measured from 42 genes. Results comparing gene expression for responders vs nonresponders to imexon are displayed in Table 3. After adjusting for multiple comparisons by controlling the FDR at the 5% level, the patients achieving an objective response had a significantly higher expression of CD68 (FDR = 0.0001), GPX1 (FDR = 0.013), and SOD2 (FDR = 0.013). Increasing the FDR to 10%, TXN (FDR = 0.055), GPX4 (FDR = 0.076), and GPX3 (FDR = 0.076) were also overexpressed among responders.

Results from PBMC lysates produced an extremely weak signal despite using a maximal concentration from these samples. Numerous genes were undetectable, suggesting weak expression within PBMCs. As such, meaningful analysis was precluded.

Further evaluating pretreatment tumors, transcript levels were used to determine the redox signature score for each patient, as previously described.¹⁷  The score is calculated as the sum of gene expression values of TXNIP, SOD1, SOD2, SOD3, GPX1, GPX3, GPX4, TXNRD1, TXNRD2, CAT, and GSTA1, minus the sum of expression values of TXN, TXN2, and mGST1. To address the hypothesis that malignant lymphocytes maintain a higher level of oxidation than nonmalignant lymphocytes, an additional analysis compared the redox score measured in independent sets of archived patient specimens with diagnoses of nonmalignant lymphoid tissues, FL, and DLBCL. Six specimens of each were analyzed. As shown in Table 4, the average (standard error [SE]) redox score from the nonmalignant lymphoid tissue was 20 001 ± 1425, significantly higher than the scores of FLs (10 353 ± 1427) or DLBCLs (10 627 ± 1965). The values for this set of lymphomas were within the range of redox scores calculated from lymphomas of patients on the imexon trial (Figure 2). Testing for a trend in redox scores across patients achieving a PR, having SD, and experiencing progression revealed that patients with higher redox scores were more likely to respond to imexon (r = 0.59, P = .03) (Figure 2).

The results from this multicenter study demonstrate imexon to have antitumor activity in relapsed refractory B-cell NHL. Previous investigations of modulating redox homeostasis have been limited in lymphoma. Arsenic trioxide may exert its cytotoxic effect in part by inhibiting the function of the glutathione system.²²  Although notable activity in acute promyelocytic leukemia has been observed with this agent, little benefit was demonstrated in relapsed lymphoid malignancies.²³  The proteasome inhibitor bortezomib, known to be active in mantle cell, lymphoplasmacytic, and to a lesser degree FL, may work in part by modulating the redox environment.²⁴  However, proteasome inhibition likely has several additional effects accounting for its cytotoxicity. Imexon triggers the intrinsic pathway toward apoptosis by depleting cellular thiols and inducing oxidative stress as its primary mechanism of action.^8-10,25  As such, these phase 2 results suggest that increasing oxidative stress by specifically targeting an antioxidant defense system is a viable therapeutic approach in lymphoma.

Previous phase 1 investigations of imexon have identified dose-limiting toxicities of fatigue and gastrointestinal effects including abdominal pain and diarrhea.^16,26  Myelosuppression was relatively limited allowing combinations with full-dose docetaxel, gemcitabine, and dacarbazine.^15,27,28  Consistent with these previous studies, few events of severe neutropenia and thrombocytopenia were observed in our heavily pretreated patients. The toxicity profile of imexon observed in the current study was otherwise similar, with gastrointestinal effects being common, but manageable. As a whole, 91% of patients had at least 1 event of diarrhea, nausea, constipation, anorexia, vomiting, or abdominal pain. These were largely grade 1 or 2 in severity and judged by the investigators to be treatment related in the majority of cases. Fatigue was also observed in 91% of patients, mostly grade 1 or 2. Sixty-four percent of adverse events were deemed treatment-related with the additional cases felt to be secondary to the patients’ underlying disease. Although these events were of low severity and reversible with treatment discontinuation, they accounted for the majority of dosing modifications. As such, their frequency may prevent prolonged administration of imexon.

The validation of biomarkers has become an integral part of early clinical trials.^29,30  Translating a biomarker into clinical practice could reduce patient risk by selecting those most likely to benefit as well as providing an early indicator of response. In developing agents that capitalize on the redox balance in cancer cells, the short-lived ROS molecules create a technical challenge in identifying predictive intermediates. As our previously developed gene expression–based approach represented an alternative for evaluating oxidative stress in diagnostic specimens, the current study tested whether the mathematically derived redox score could predict patient outcome following therapy.¹⁷  The score incorporates the 3 major enzymatic systems of cellular antioxidant defenses. The first system includes genes encoding proteins that intercept ROS: superoxide dismutases (SODs), catalase, and glutathione peroxidases (GPX). The second, incorporating components of the TXN system, acts to maintain redox-sensitive proteins in a reduced form; TXN substrates include ribonucleotide reductase and NF-κB.³¹  The third set of genes in the redox score has antioxidant response elements in their promoters; expression of the encoded proteins increases in response to oxidative stress (eg, microsomal glutathione S-transferase [mGST]). DLBCL gene expression data available from the Leukemia/Lymphoma Molecular Profiling Project³²  were used to calculate redox scores. Among the 240 patients included in that study, a low redox score, reflecting significantly lower expression of antioxidant defense proteins, higher TXN system function, and higher expression of mGST were associated with an inferior outcome.¹⁷  In support of these findings, immunohistochemical staining of DLBCLs from 106 patients who received CHOP-like therapy and rituximab demonstrated an association of inferior survival outcomes with increased levels of TXN, 8-hydroxydeoxyguanosine, and nitrotyrosine.³³  The latter 2 are markers for oxidative damage to DNA and protein, respectively.

A major finding from the laboratory correlates in this phase 2 trial was that responses to imexon correlated with the redox scores obtained from the pretreatment biopsies of the patients’ lymphomas. The tumor specimens from the 2 patients with PRs had higher redox scores than the 6 patients with progressive disease. These results are consistent with previous reports by our group and others indicating a worse treatment outcome for lymphomas with gene expression patterns consistent with the most oxidized redox balance.^17,33  Individual genes for which expression levels correlated with response to imexon were SOD2 and GPX1, with higher transcript levels seen in patients demonstrating an objective response. Notably, BCL2 expression did not correlate with treatment outcomes. Manganese SOD (MnSOD) is the antioxidant enzyme encoded by SOD2. It functions as the major scavenger of superoxide anion radicals in mitochondria, forming hydrogen peroxide in the process. Hydrogen peroxide is a relatively stable molecule and can diffuse across membranes. GPX1 is present in the cytoplasm and uses glutathione as a cofactor to reduce hydrogen peroxide to water. One interpretation of this finding is that imexon-sensitive lymphomas are more reliant on MnSOD and GPX1 to counter oxidative stress; depletion of glutathione by imexon effectively prevents GPX1 activity, inducing apoptosis in response to elevated levels of hydrogen peroxide in the tumor cells. This interpretation is supported by previous studies demonstrating the impact of GPX1 and MnSOD on doxorubicin and glucocorticoid-mediated apoptosis.^34,35  The lymphomas that did not respond to imexon may have mechanisms of resistance to increased oxidative stress that do not depend on the cellular glutathione pool. It is also possible that within this cohort of relapsed patients, the redox score is simply identifying patients with relatively improved outcomes. As such, we consider these results hypothesis generating and in need of further confirmation in a larger cohort of patients treated with imexon or other similar agents modifying oxidative stress.

Increased intratumoral CD68 expression significantly predicted for responses to imexon. Macrophages are known to elicit their tumor-specific cytotoxic effect in part through production of hydrogen peroxide. Previous murine investigations have demonstrated that glutathione depletion increases tumor cell susceptibility to macrophage-mediated oxidative injury.³⁶  As CD68 is a known prognostic marker in lymphoma and can be evaluated using immunohistochemistry, future investigations of imexon should investigate CD68 expression further.

In conclusion, this is the first report to our knowledge of an agent targeting cellular redox demonstrating activity in lymphoid malignancies. These data suggest that targeting oxidative stress warrants further study in lymphoma. Furthermore, this and other investigations suggest that combination therapy incorporating imexon with cytotoxic therapeutics is feasible. Intratumoral expression of genes regulating oxidative stress may identify patients most likely to respond to this novel therapeutic approach.

The authors thank the patients who participated in this study and their families as well as the clinical trial staff who made the study possible.

This work was supported by AmpliMed Pharmaceuticals, the Rochester-Arizona Specialized Programs of Research Excellence in lymphoma (CA-130805), and National Institutes of Health, National Cancer Institute P01 CA017094.

Contribution: P.M.B., T.P.M., S.H.B., and M.M.B. designed the research; P.M.B., T.P.M., J.W.F., C.M.S., H.C., D.J.R., D.O.P., C.C., M.S., S.P., T.H.L., L.M.R., R.I.F., S.H.B., and M.M.B. performed the research; D.R.P. and A.M.B. performed statistical analyses; P.M.B., M.H., J.L., S.H.B., and M.M.B. collected, analyzed, and interpreted the data; P.M.B., R.T.D., S.H.B., and M.M.B. wrote the manuscript; and all authors reviewed the draft manuscript and approved the final version for submission.

Conflict-of-interest disclosure: R.T.D. was previously employed by Amplimed Pharmaceuticals. The remaining authors declare no competing financial interests.

Correspondence: Paul M. Barr, University of Rochester, Wilmot Cancer Center, 601 Elmwood Ave, Box 704, Rochester, NY 14642; e-mail: paul_barr@urmc.rochester.edu.