A Phase I Pharmacokinetic and Pharmacodynamic Correlative Study of the Antisense Bcl-2 Oligonucleotide G3139, in Combination with Carboplatin and Paclitaxel, in Patients with Advanced Solid Tumors

Apoptosis is a program of cellular suicide that removes unnecessary cells throughout life. It reflects the balance of many proapoptotic and antiapoptotic regulatory signals, the sum of which determines the fate of the cell. Bcl-2 is an antiapoptotic protein that is frequently overexpressed in cancer cells and may play a large role in the development of many solid tumors, including prostate (1, 2), melanoma (3), breast (4), lung (5), renal (6), and ovarian (7) cancers. In addition, up-regulated Bcl-2 has been implicated as a possible mechanism for chemotherapy and radiotherapy resistance (8, 9).

G3139 (Genasense, oblimersen sodium) is an 18-mer antisense phosphorothioate oligodeoxynucleotide that binds to the first six codons of the human Bcl-2 mRNA. When given alone or in combination with chemotherapy to human cell lines in vitro, G3139 down-regulates Bcl-2 protein expression, resulting in a greater degree of apoptosis (10). Antitumor activity with single-agent G3139 has been reported in non–Hodgkin's lymphoma (11). Likewise, disease-specific trials in melanoma (12), small cell lung (13), colorectal (14), breast (15), and prostate (16) cancers have been done with G3139, alone or in combination with chemotherapy. The dose-limiting toxicities (DLT) seen with G3139 in the phase I non–Hodgkin's lymphoma study, which used a continuous 14-day s.c. infusion, were thrombocytopenia, hypotension, fever, and asthenia. However, when G3139 was given as a continuous i.v. infusion for 14 to 21 days in an advanced solid malignancy phase I study, the drug was well tolerated with the main treatment-related adverse events being fatigue and transaminase elevations. Steady-state plasma concentrations of G3139 were achieved ∼10 hours from start of the infusion with a terminal plasma half-life of ∼2 hours (17).

The motivation to combine G3139 with carboplatin and paclitaxel includes the wide applicability of this chemotherapy regimen to many solid tumor types, including lung, bladder, head/neck, breast, and ovarian cancers. By decreasing the tumor's resistance to apoptosis with G3139, potentially less toxic doses of carboplatin and paclitaxel could be used with equal or greater antitumor activity. In addition, there is evidence that paclitaxel induces prolonged mitotic arrest, resulting in hyperphosphorylation of Bcl-2 (18). It is suggested that the hyperphosphorylation of Bcl-2 is one important mechanism for antitumor activity of paclitaxel (19–22). If so, the potential of combining G3139 with a taxane-based regimen may be beneficial. Certainly, preclinical data supports at least potential additive benefits of combining a Bcl-2 antisense oligodeoxynucleotide with paclitaxel (23).

Whereas the principal objective of this phase I trial was to define the maximum tolerated dose (MTD) of G3139 in combination with carboplatin and paclitaxel, an extensive pharmacokinetic and pharmacodynamic analysis was done to better define the activity of G3139. These studies included pharmacokinetic analysis of G3139, carboplatin, and paclitaxel in plasma, as well as intratumoral G3139 levels from paired tumor biopsies in a planned cohort of 12 subjects. Quantitative changes in Bcl-2/Bax transcription and protein expression levels from the paired tumor biopsies were assayed by reverse transcription–PCR and immunohistochemistry. Changes in Bcl-2/Bax transcription and protein expression in peripheral blood mononuclear cells (PBMC) were assessed by reverse transcription–PCR and flow cytometry.

Eligible patients had progressive metastatic, or unresectable, solid malignancies with no standard curative therapy. All patients had an Eastern Cooperative Oncology Performance Status of <2 with a life expectancy of >3 mo, a total leukocyte count of >3,000/mm³, absolute neutrophil count of >1,500/mm³, platelet count of >100,000/mm³, normal total bilirubin, aspartate aminotransferase (AST) and alanine aminotransferase of <2.5× upper limit of normal, and normal serum creatinine (or creatinine clearance of >60 mL/min/1.73 m²). No known brain metastasis, preexisting grade of >2 neuropathy, or other chemotherapy/radiotherapy/investigational therapy within 4 wk of registration was allowed. This study was reviewed and approved by the University of Wisconsin Hospital and Clinic Institutional Review Board and other appropriate committees, and all patients gave written informed consent for clinical and research aspects of this study according to federal and institutional guidelines before screening.

All patients underwent placement of a central venous catheter. G3139 was given as a continuous i.v. infusion on days 1 to 7 through a portable continuous infusion pump, with paclitaxel given on day 4 by i.v. infusion over 3 h, followed immediately by carboplatin given via rapid i.v. infusion over 30 min. Standard chemotherapy premedications, including antiemetics, dexamethasone, and antihistamines (H1 and H2 blockers), were used per standard practice guidelines.

The starting dose of G3139 was 3 mg/kg/d and was escalated successively to the initial planned maximum of 7 mg/kg/d. During the G3139 dose escalation phase, the chemotherapy dose was fixed with paclitaxel at 150 mg/m² and carboplatin at an area under the curve (AUC) of 5. Standard cohorts of three were used with dose escalations proceeding if no DLT was encountered. If one patient experienced a DLT at a given dose level, a total of six patients were entered at the dose level. If two of a maximum of six evaluable patients experience DLT, then MTD would be exceeded and additional patients were enrolled at the dose level below. The MTD was defined as the highest dose at which less than two of six patients experienced a DLT in cycle 1. A DLT was defined as a toxicity that was probably or definitely related to G3139 (in combination with paclitaxel and carboplatin), with either grade 4 neutropenia lasting for >7 d (or febrile neutropenia), platelets at <25,000/mm³ lasting for >7 d (or <50,000/mm³ associated with severe bleeding), or any grade of >3 nonhematologic toxicity (except nausea/vomiting and diarrhea unless this occurs despite maximal supportive care). The National Cancer Institute Common Toxicity Criteria Version 2.0 was used for grading of all toxicities. At either dose level 5 (G3139 at 7 mg/kg) or the MTD, a planned additional cohort of 12 patients were to be enrolled for correlative assessment, including procurement of paired tumor biopsies on all subjects. Based on the results of these studies, a decision would be made to either escalate chemotherapy doses to define the MTD (if not reached) or increase G3139 to 9 mg/kg (if Bcl-2 suppression not evident). A summary of the dose escalation schema is shown in Table 1.

G3139 (Genta, Inc.) was supplied through the National Cancer Institute in 10-mL vials containing 300 mg, which was diluted to the appropriate volume and concentration (final solution concentration maintained between 10 and 30 mg/mL) with 0.9% saline solution. Commercial paclitaxel and carboplatin were used in this study. A cycle of therapy was defined as 21 d. There was no limit to the number of courses that could be given for patients benefiting and tolerating therapy.

Complete medical history, physical examination, and routine laboratory studies were done at baseline and before each new cycle of therapy. Routine laboratories included complete blood count, albumin, alkaline phosphatase, total bilirubin, electrolytes, creatinine, phosphate, total protein, aspartate aminotransferase, alanine aminotransferase, and lactate dehydrogenase. A baseline electrocardiogram was required, as well as pretreatment radiographs of all known sites of measurable and evaluable metastasis. Repeat radiographic evaluations for response were done after every two cycles of therapy. Response Evaluation Criteria in Solid Tumors guidelines were used for determination of tumor response (24).

G3139. Plasma concentrations of G3139 were determined by high-performance liquid chromatography (25) and tumor concentrations were evaluated by ELISA as previously described (26). A total of three blood samples were obtained for analysis of G3139 plasma concentrations. The first was obtained immediately before initiation of the G3139 infusion; the second was on day 4, before paclitaxel and carboplatin administration (96 h after initiating G3139); and the third was on day 5 (24 h after carboplatin and paclitaxel administration and 120 h after initiating the G3139 continuous infusion). Biopsies were obtained (dose level 5 only) at baseline and day 4 (before carboplatin and paclitaxel infusion). The sample used for pharmacokinetics was flash frozen and stored at −70°C until analysis.

To analyze G3139, 100 μL of plasma was placed in a microfuge tube with 300 μL of IPEGAL/PBS (1:1) added and vortexed. Subsequently, 400 μL acetonitrile was added and vortexed for 30 s. The sample was then centrifuged at 9,000 × g for 7 min. The supernatant was placed into an autosampler vial and analyzed by UV detection at 267 nm. High-performance liquid chromatography was done with a gradient on a Dionex DNA Pac PA-100, 2 × 10 mm, 13 μm, alkyl quaternary amine column.

The standard curve was linear from 0.234 to 30 μg/mL (r² = 0.994). Intraday variability was 3.96% for high (10 μg/mL) standard (n = 3) and 0.56% for low (0.039 μg/mL) standard (n = 3). Interday variability was 0.46% for high standard (n = 4) and 4.34% for the low standard (n = 4) over 7 d. The lower limit of quantitation was 0.234 μg/mL, and the absolute recovery from plasma compared with water was >95%.

G3139 in solid tumor tissue was measured by a 96-well plate quantitative sandwich immunoassay with data normalized to protein values by standard methods. This assay involves hybridization of G3139 to the 5′-end overhang of a 3′-biotinylated capture oligonucleotide, ligation to a digoxigenin-labeled probe, and detection by an anti–digoxigenin-alkaline phosphatase system (AttoPhos AP Fluorescent Substrate System, Promega Corp.). The plate was read at 430 nm (excitation) and 570 nm (emission) using a Molecular Devices SpectraMax Gemini plate reader.

The standard curve was linear from 0.18 to 10,000ng/mL G3139 (r² = 0.992). Intraday variability was <14% for high (10,000 ng/mL) standard (n = 3) and 0.56% for low (0.18 ng/mL) standard (n = 3). Interday variability was 12% for high standard (n = 4) and 11.15% for the low standard (n = 4) over 7 d. The lower limit of quantitation was 0.18 ng/mL, and the absolute recovery from plasma compared with water was >95%.

Carboplatin. Carboplatin concentrations were evaluated with a Varian SpectrAA 10/20 with GTA-96 atomic absorption spectrophotometer adjusted to detect platinum (265.8-266) as previously described (27, 28). Samples were obtained immediately before the carboplatin infusion, 40 min into the infusion, at the end of the infusion, 30 min, and 1, 2, 4, 6, 8 12, 24, and 48 h after the end of the infusion. Blood samples were collected in heparinized tubes, and plasma was separated from cells by centrifugation (1,200 × g, 4°C, 10 min) within 15 min of blood collection. Free drug was then separated from the protein-bound drug by centrifugation (1,200 × g, 4°C, 15 min) through an ultrafiltration membrane (Centrifree YM-30, Amicon Bioseparations, Millipore Corporation). Ultrafiltrate samples were stored at −70°C until analysis. Fifty microliters of each standard or sample to be analyzed were placed in a plastic atomic absorption vial, to which was then added 0.95 mL of 0.1% HNO₃. The three calibration samples of 0.05, 0.1, and 0.2 μg/mL were used to calibrate the atomic absorption spectrophotometer, and the calibration was checked by running the calibration samples after the sample runs.

Paclitaxel. Paclitaxel concentrations were evaluated with a Spectra Physics P2000 high-performance liquid chromatography, as previously described (29). Samples were obtained immediately before the paclitaxel infusion, 90 min into the infusion, at the end of the infusion, 30 min, and 1, 2, 4, 6, 8, 12, 24, and 48 h after the end of the infusion. Blood samples were collected in heparinized tubes, and plasma was separated from cells by centrifugation (1,200 × g, 4°C, 10 min) within 15 min of blood collection. Samples were stored at −70°C until analysis.

The plasma standard curve was linear from 0.05 to 5.0 μg/mL (r² = 0.999). Intraday variability was low; under 4% for 0.05 μg/mL and 2% for the high standard 5 μg/mL (n = 3). Interday variability was under 9% for the low standard and ∼14% for the high standard (n = 6) over 6 mo, and the lower limit of quantitation was determined to be 0.1 μg/mL.

Pharmacokinetic analysis was done by noncompartmental methods using the WinNonlin program, version 5.2 (Pharsight). The maximum plasma concentration (C_max) and the corresponding time of the maximum concentration were identified from the measured samples and recorded. Plasma concentration versus time data were plotted on a semilogarithmic scale, and the terminal log-linear phase was identified by best fit. The elimination rate constant (λ) was determined as the slope of the linear regression for the terminal log-linear portion of the plasma concentration-versus-time curve. A terminal half-life value was calculated as ln(2)/λ. AUC was calculated by the trapezoidal method using extrapolation to infinity.

Peripheral blood specimens were collected on day −4, day 1 (pre-G3139), and day 4 (pre-paclitaxel/carboplatin). At each time point, a PAXgene (Qiagen) tube was obtained and RNA isolated by standard methods (Ultraspec II RNA Isolation Systems; ref. 30). RNA was quantified spectrophotometrically and transcribed into cDNA by standard methods. A total of 250 ng of cDNA was used in each 25 μL PCR reaction. For tumor samples, laser capture microdissection of 1,000 cells was done as previously described (31), RNA was extracted and transcribed into cDNA by standard methods, and the total volume was used in the PCR reaction. Sample with mRNA concentrations above the standard curve were diluted and retested.

Gene expression was done using the MyIQ real-time thermocycler (Bio-Rad) with the iCycler variables as follow, for 95°C/2 min × 1 cycle, 95°C/30 s, 61.7°C/1 min, 72°C/30 s for 50 cycles, 60°C/7 s + 0.5°C × 70 cycles and analysis by the Eragen MultiCode RTx Analysis Software version 1.0.14 (Eragen). The following primers were used: Bcl-2 forward 5′-GCT CTT CAG GGA CGG G-3′, Bcl-2 reverse (probe) 5′-/56-FAM/Me-isod C/GC TCT CCA CAC ACA TGA CC-3, Bax forward (probe) 5′-/56-FAM/Me-isod C/CG GAC CCG GCG AGA-3′, Bax reverse 5′-CGC CTC TGG GCT GCT-3′ (IDT).

The Bcl-2 assay was validated using cDNA obtained from the DU145 prostate cancer cell line. Total cDNA concentration was determined spectrophotometrically, and the standard curve was generated by amplifying in 39 to 2,500 ng total DNA on 4 d over a period of 4 wk. The standard curve was linear from 39 to 2,500 ng, with an intraday variability of 3.5%. Interday variability over 4 wk was 3.3%. The Bax assay similarly was validated using cDNA obtained from the DU145 prostate cancer cell line. Total cDNA concentration was determined spectrophotometrically, and the standard curve was generated by amplifying in 6.25 to 100 ng total DNA on 6 d over a period of 6 wk. The standard curve was linear from 6.25 to 100 ng with an intraday variability of 14.1%. Interday variability over 6 wk was 2.9%.

Flow cytometry. Peripheral blood specimens were collected on day −4, day 1 (pre-G3139), and day 4 (pre-paclitaxel/carboplatin). One hundred microliters of blood were added to tubes and lysed with 3 mL of FACSLyse (BD Biosciences). Samples were incubated for 10 min at room temperature, centrifuged at 400 × g for 10 min, and washed with Dulbecco's phosphate-buffered saline (Media Tech). Cell pellets were resuspended in permeabilizing buffer (0.1% saponin and 0.1% bovine serum albumin, both from Sigma, in Dulbecco's phosphate buffered saline) and incubated for 10 min at room temperature and then centrifuged at 400 × g for 10 min. Antibodies or isotype controls were added to the cell pellets in the volumes given and incubated for 45 min at room temperature. Samples were washed with 3 mL of permeabilizing buffer and centrifuged again at 400 × g for 10 min. The following unconjugated or phycoerythrin-conjugated antibodies were added to the resulting cell pellets: anti–Bcl-2-phycoerythrin, hamster IgG, and goat anti-mouse IgG from BD Biosciences; anti-Bax was purchased from Beckman-Coulter. Three milliliters of Dulbecco's phosphate buffered saline were used to wash the samples that were centrifuged at 400 × g for 10 min. The resulting cell pellets were resuspended in 300 μL of Dulbecco's phosphate buffered saline with 1% FCS and analyzed on a LSRII cytometer (BD Biosciences) with FACSDiVa software. Raw data were acquired until 50,000 lymphocyte events were counted and identified by light scatter. At the same time, instrument settings data was acquired on Quantibrite Phycoerythrin standard beads (BD Biosciences). Data were analyzed with FloJo analysis software (Treestar). The fluorescence of the Bcl-2 was expressed as molecules of phycoerythrin. The ratio of Bax to Bcl-2 was been calculated by creating a ratio of the FITC (Bax) to the phycoerythrin (Bcl-2). All values were normalized to the isotype control.

Immunohistochemistry. Immunohistochemistry was done on tissue sections of formalin-fixed, paraffin-embedded tumor biopsies using an automated immunostainer (Ventana Medical Systems, VMS). Tissue sections underwent antigen retrieval before staining. The antigen retrieval for Bcl-2 staining was done on-line using mild heat-induced retrieval with Cell Conditioning Solution 1 (CC1, a proprietary VMS-pH buffer). Bax antigen retrieval was done by incubating slides in an EDTA buffer (pH 8.0), in an electric pressure cooker (Decloaking Chamber, Biocare Medical) for 2 min at ∼20 PSI. The slides were incubated with the prediluted primary antibody Bcl-2 (clone 124 from VMS) or a 1:50 dilution in Zymed antibody diluent of anti-Bax (clone 2D2, Zymed Laboratories) for 32 mi. Slides receiving the Bcl-2 were subsequently put through an amplification step using VMS amplification kit. After incubation with a universal secondary, target detection was by an indirect biotin avidin system (VMS), including diaminobenzidine. Endogenous peroxidase quenching and biotin blocking were done on-instrument with kit reagents (VMS). Immunostained slides were counterstained with hematoxylin on the instrument and subsequently dehydrated through a series of graded alcohols and coverslipped on a Tissue-Tek automated coverslipper (Sakura). The immunohistochemical staining pattern of each case was assessed by the same pathologist and scored according to staining intensity and proportion of positive cells.

A total of 46 patients were enrolled on the dose escalation and expanded correlative parts of this study. Four of these patients were deemed unevaluable. Three patients at dose level 1 were unevaluable for toxicity (two due G3139 infusion pump malfunctioning and one secondary to the development of spinal cord compression on day 15 requiring emergent radiotherapy). One additional patient at dose level 5 was unevaluable secondary to failure to complete cycle 1 of treatment because of the development of a recto-vaginal fistula attributed to her cancer and recent surgery. Table 2 summarizes relevant patient characteristics.

The predominant toxicity observed was myelosuppression and thrombocytopenia. One patient at dose level 1 developed grade 4 thrombocytopenia, and one patient at dose level 5.1 had grade 4 neutropenia. These two events were the only toxicities observed that met our stringent criteria for DLTs. Although other grades 3 and 4 events were seen in cycle 1, these events did not meet criteria for a DLT (see Table 3). In all, the combination of G3139 with carboplatin and paclitaxel was found to be well tolerated, with no unusual toxicities observed that were felt likely related to G3139 and not attributable to the expected toxicities of the chemotherapy agents. The median number of cycles received per patient was 3 (range, 1-14).

In total, six partial responses (four confirmed, two unconfirmed) were observed. Two confirmed responses were in dose level 5; the first patient being an individual with esophageal cancer previously treated with two prior cytotoxic chemotherapy regimens, and the second, a patient with bladder cancer treated with one prior chemotherapy regimen. At dose level 5.1, two partial responses (one confirmed, one unconfirmed) were observed. Both patients had urothelial cell cancer of the bladder. One patient at dose level 5.2 also experienced a confirmed partial response. This individual had head-neck carcinoma which was previously treated with chemotherapy and an epidermal growth factor receptor tyrosine kinase inhibitor. Lastly, one patient with papillary urothelial cell carcinoma of the bladder had an unconfirmed partial response at dose level 5.2. This patient was removed from study after two cycles of treatment by physician discretion, as it was felt that the patient poorly tolerated the chemotherapy despite maximal dose reduction during cycle two. Of note, this patient had significant difficulties tolerating a prior cytotoxic chemotherapy regimen as well.

G3139 plasma concentrations were evaluated at baseline, on day 4 of continuous infusion of G3139 (before carboplatin and paclitaxel chemotherapy), and on day 5 of continuous infusion of G3139 (after carboplatin and paclitaxel chemotherapy) in 11 subjects enrolled onto level 5 using high-performance liquid chromatography. All subjects had measurable G3139 concentrations on Day 4 and 5 after administration of G3139 with the mean G3139 concentration being 4.28 ± 1.40 μg/mL on Day 4 and 3.86 ± 1.36 μg/mL on Day 5. G3139 plasma levels were similar on day 4 and day 5, suggesting that G3139 pharmacokinetics are unaffected by coadministration of carboplatin and paclitaxel. G3139 concentrations in tumor were evaluated in 11 paired biopsies obtained before G3139 administration and on day 4. G3139 was detectable in 10 of 11 tumors after treatment, and the mean tumor concentrations of G3139 was 27.9 ± 43.1 ng/per μg of tumor protein indicating the G3139 is able to penetrate tumors (see Table 4).

Pharmacokinetic studies of paclitaxel and carboplatin were conducted in 11 subjects enrolled onto level 5 using high-performance liquid chromatography. The pharmacokinetic variables are described in Table 4 and are similar to reported values, suggesting that administration of G3139 did not alter the pharmacokinetics of paclitaxel or carboplatin. Both carboplatin and paclitaxel exhibited wide interpatient variability, and patients on average achieved ∼80% of the planned carboplatin AUC. Calculation of carboplatin doses used the Calvert formula with the GFR estimated from the creatinine clearance, which may have accounted for the inaccuracy in dosing (32).

PBMC gene expression. Bcl-2 is the target of G3139; therefore, we expected Bcl-2 gene expression in PBMC to decrease after administration of G3139. As expected, Bcl-2 gene expression had a statistically significant decline from baseline to 4 days of G3139 infusion (P < 0.001, two-sided Wilcoxin signed rank test), indicating that Bcl-2 gene expression is inhibited by G3139 in PBMCs. There was no change in detectable bax expression in PBMCs after G3139 administration (Fig. 1).

Tumor gene expression. There were a total of 13 day-0 biopsy samples and 10 day-4 biopsy specimens amenable for this analysis (one patient did not undergo second biopsy; two patients did not have enough tumor tissue for laser capture microdissection and subsequent gene transcription assessment). After laser capture microdissection of 1,000 cells, Bcl-2 and bax gene expression were also evaluated in biopsy samples. Bcl-2 gene expression declined in the majority of paired samples by day 4 of G3139 treatment, with a median ratio of day 4/Pre of 0.65, although this did not reach statistical significance. There was no change in bax expression in biopsy specimens after G3139 administration (Fig. 2).

Flow cytometry. PBMCs were collected from patients at baseline and on day 4 after starting G3139 (pre-carboplatin and paclitaxel chemotherapy). Bcl-2 and Bax protein expression were assessed by intracellular flow cytometric analysis. Bcl-2 expression had a statistically significant decline from baseline after 4 days of G3139 infusion (P = 0.008, paired t test), suggesting that Bcl-2 expression is reduced by G3139 treatment in PBMCs. There was no significant change in Bax expression in PBMCs after G3139 administration.

Immunohistochemistry. A total of 12 paired core needle tumor biopsy samples were obtained, of which 11 paired samples were adequate for immunohistochemistry analysis. One patient with melanoma had biopsies that were unable to be assessed due to extensive background melanin, making immunohistochemical interpretation difficult. In general, faint staining with Bax was seen in the background with focal regions of intense staining noted in areas where cells were presumably undergoing apoptosis (close to necrotic areas as seen on the H&E). Bcl-2 staining seemed similar between pretreatment and posttreatment samples for most subjects, but many tumors had negative pretreatment and posttreatment Bcl-2 staining. Few Bcl-2–positive small lymphocytes (tumor-infiltrating lymphocytes) were present within some tumors.

Overexpression of Bcl-2 in tumors has been reported to be associated with not only a more advanced stage at diagnosis but also treatment resistance and a shortened survival. By targeting critical proteins, such as Bcl-2, it may be possible to restore normal regulatory pathways and enhance the effectiveness of current chemotherapeutics (33). Here, we report the results of a phase I trial assessing the safety and feasibility of combining G3139 with carboplatin and paclitaxel, as well as an extensive evaluation of the pharmacodynamic effect of G3139 alone in PBMCs and in the tumor.

As patients are typically heavily pretreated in a phase I study, we had concerns that significant myelosuppression and thrombocytopenia would be seen secondary to the carboplatin and paclitaxel chemotherapy alone. Given the potential for additional hematologic toxicity, as noted with other phosphorothiolate olignucleotides (34–36), we chose to fix the chemotherapy dose (carboplatin AUC 5, paclitaxel 150 mg/m²) and dose-escalate the G3139 alone. Because some degree of hematologic toxicity was expected, we used an aggressive definition of a hematologic DLT to allow us to not only assess the safety of this combination, but also escalate the G3139 to levels that were felt necessary for appropriate Bcl-2 suppression. It was soon realized that G3139 could be safely combined with carboplatin and paclitaxel, with no appreciable increase in toxicity than what was consistent with carboplatin and paclitaxel alone. As a result, the protocol was modified to subsequently dose-escalate the chemotherapy to define the MTD. During cohort 5.2, it was obvious that all three patients had grade 4 myelosuppression. Although the neutropenia did not meet the 7-day duration criteria for a DLT, it was decided after discussion with the Cancer Therapy Evaluation Program at the National Cancer Institute to stop the dose escalations, as the next planned dose level involved a chemotherapy dose that is typically given only in the first-line setting and not in heavily pretreated patients commonly enrolled in phase I studies. Therefore, although the MTD was not defined, we feel confident reporting that G3139 can be safely combined with standard doses of carboplatin and paclitaxel. This is consistent with the conclusion derived in multiple other solid tumor trials combining G3139 with full doses of cytotoxic chemotherapy (37–39).

The planned expanded cohort of 12 patients with paired tumor biopsies was completed, and extensive correlative analyses of G3139 pharmacokinetic and pharmacodynamic effects were done. What is evident is that, by day 4, Bcl-2 RNA was significantly decreased in the PBMCs with a parallel decrease seen in the tumor samples. Although the decrease in the tumor Bcl-2 RNA did not meet statistical significance, given detectable intratumoral G3139 in 10 of 11 samples available for this evaluation and the fact that a relative decrease in Bcl-2 gene expression over baseline was observed in the majority of samples, this was felt to be highly indicative of G3139 activity.

We observed that Bcl-2 protein expression in PBMCs was also significantly decreased on day 4, although the change in intratumoral Bcl-2 expression was not appreciably different. This raises many questions regarding the utility of using PBMCs as a surrogate marker for changes within the tumor itself. One confounding factor that we retrospectively identified is a potential change in peripheral lymphocyte populations due to effects from dexamethasone, which is routinely given before paclitaxel administration. Because glucocorticoids are well known to induce lymphocyte apoptosis, as well as lymphocyte demargination, the relative subpopulations of lymphocytes measured in peripheral blood may be altered, affecting Bcl-2 determination. At present, a separate cohort of six subjects is being enrolled to evaluate specifically PBMC changes due to G3139 alone (without glucocorticoids). Complete data with regards to the immunologic correlative of this study and ongoing evaluation will be published separately once complete.

The lack of significant change in tumor Bcl-2 protein expression is also complex. This could be due to differences in the tumor types being assessed, the inconsistent nature of tumor biopsies (tumors are by nature heterogeneous with focal areas of necrosis and inflammation, and thus, specific site of biopsy can greatly affect results), as well as the fact that a repeat biopsy on day 4 biopsy might not provide adequate time to see changes in Bcl-2 protein expression. Given that the optimal Bcl-2 RNA suppression is likely not achieved until day 3 of G3139 and the reportedly long Bcl-2 half-life of ∼14 hours (40), the measurement of Bcl-2 protein expression on day 4 was probably too early to detect Bcl-2 protein changes. No significant change in Bax gene expression was noted in either PBMCs or tumor samples.

This study represents the most complete assessment of G3139 activity in solid tumors to date and includes the highest number of paired tumor biopsies obtained. What our data suggests is that a reduction in PBMC Bcl-2 RNA seems to correlate with reductions in Bcl-2 RNA within the tumors. Whether these changes will result in a significant decrease in tumor Bcl-2 protein expression, leading to clinically meaningful antitumor activity, is of course still open for debate and can only be addressed in the context of a randomized trial. Given issues related to tumor heterogeneity, reproducibility of paired needle biopsies, and increased risk for our patients, further assessment of anti–Bcl-2 activity using paired tumor biopsies is not recommended. Future assessments of new anti–Bcl-2 agents might include novel functional imaging modalities evaluating apoptosis, such as Annexin-V positron emission tomography scans (41), as it could allow for noninvasive assessments at multiple time points. By assessing changes over time, we may be better able to define drug effect and the optimum biological dose and schedule of these and other novel agents.

Lastly, oligodeoxynucleotides, like G3139, contain CpG dinucleotides within specific-sequence contexts (CpG motifs), which have been shown to activate rodent and primate immune cells via toll-like receptor 9 (42). Thus, G3139 might function as an immune stimulator in addition to its effects on Bcl-2. Future analysis of the effects of G3139 on T-cell functioning and signaling may be warranted.

Here, we report a pharmacodynamic trial assessing the reduction of Bcl-2 RNA and change in protein expression levels in PBMCs and paired tumor biopsy samples after exposure to G3139. The reduction of Bcl-2 RNA in the tumors and PBMCs was observed in conjunction with achievable intratumoral concentrations of G3139. In summary, G3139 remains an interesting agent with a variety of effects that still need exploring. Future plans for this particular regimen are uncertain, although one possible disease interest may be in urothelial cell cancers given the prognostic significance of Bcl-2 overexpression. Evaluation of G3139 in combination with other agents remains under active investigation, especially in diseases, such as melanoma and chronic lymphocytic leukemia. Likewise, there may be potential to combine G3139 with tumor vaccines, increasing immunologic activity by not only its anti–Bcl-2 effects but also potentially by effects mediated by its CpG motif.

We thank Heidi Bakken, Maria Kruse, Jessica Weiss, Molly Houston, Marcia Pomplun, Amy Dresen, and Barbara Woodhouse for their assistance in the conduct of this study.