In vitro and in vivo irinotecan-induced changes in expression profiles of cell cycle and apoptosis-associated genes in acute myeloid leukemia cells

Irinotecan (CPT-11) is metabolized by carboxylesterases to its active metabolite 7-ethyl-10-hydroxycamptothecin (SN-38). SN-38 acts as a classic topoisomerase I inhibitor: it stabilizes the topoisomerase I/DNA cleavable complex, which blocks DNA replication and causes DNA strand breaks. Although CPT-11 has activity as a single agent in diverse malignancies (1–7), its major role seems to be in combination with other chemotherapy drugs.

The optimal sequence of administration of CPT-11 with other chemotherapy drugs has not been clearly established because mechanisms of synergy are incompletely understood. These mechanisms may be drug specific. For example, synergy of CPT-11 with thymidylate synthetase inhibitors, such as 5-fluorouracil, may be explained by prolonged inhibition of thymidylate synthetase and increased incorporation of 5-fluorouracil metabolites into DNA (8, 9). On the other hand, synergistic effects of CPT-11 may be associated with a more general cellular response. In this regard, CPT-11 has been reported to alter cell cycle kinetics (10) and also to augment the apoptotic response (11–17), for example, increasing apoptosis induced by tumor necrosis factor (TNF)–related apoptosis-inducing ligand/Apo2L by up-regulating death receptor 5 [TNF receptor superfamily 10B (TNFRSF10B); refs. 11, 13]. Preclinical data favor administration of CPT-11 prior to antimetabolites (14), which is a schedule consistent with both of these mechanisms of synergy.

Established determinants of response to CPT-11 include topoisomerase I expression levels, mutations of the topoisomerase I gene, and carboxylesterase activity. It has been suggested that differential drug sensitivity may also reflect molecular responses to DNA damage (18, 19), including cell cycle perturbations, apoptotic response, and DNA repair. Thus, evaluation of kinetic changes in molecular responses to CPT-11 or SN-38 may provide a rationale for optimizing scheduling of CPT-11 combination chemotherapy. The current study explores changes in gene transcription in HL60 cells exposed to SN-38 studied by cDNA microarray analysis using a customized gene array comprising 3,011 cancer-related genes. Kinetic changes in gene expression in blasts from patients with acute myeloid leukemia (AML) and chronic myeloid leukemia (CML) in myeloid blast transformation following in vivo therapy with a single dose of CPT-11 were then compared with the changes observed in HL60 cells following short-term in vitro SN-38 exposure.

HL60 cells were maintained in exponential growth in RPMI 1640 (Life Technologies, Grand Island, NY) supplemented with 10% heat-inactivated fetal bovine serum, 2 mmol/L l-glutamine, 20 units/mL penicillin, and 20 μg/mL streptomycin (Life Technologies) buffered in 5% CO₂ in air. The in vitro studies were done on asynchronously growing HL60 cells to ensure that changes in gene transcription could be attributed to exposure to SN-38 rather than to prior cell synchronization. A stock solution of 5 mmol/L SN-38 (Pharmacia Co., Kalamazoo, MI) was prepared in 100% DMSO and stored in aliquots at −20°C. SN-38 from the frozen stock was added to cells cultured at 1 × 10⁶/mL to achieve the desired final concentrations. Two separate series of drug exposures were done. In the first series of experiments, HL60 cells were exposed to 0, 0.1, and 0.3 μmol/L SN-38 for 2 hours. In the second series of experiments, cells were exposed to 0, 0.1, 0.3, and 1.0 μmol/L SN-38 for 2 hours to establish reproducibility of results and to extend the drug concentration range. Cells were incubated with drug at 37°C for 2 hours, washed twice with PBS, resuspended in supplemented RPMI 1640 at 0.25 × 10⁶ cells/mL, and incubated at 37°C in a fully humidified atmosphere of 5% CO₂ in air.

Cell aliquots required for each assay were removed from the cultures at the time intervals indicated in Results. For cDNA microarray analysis, 1 × 10⁷ cells per sample point were cryopreserved at −80°C in 20% DMSO in supplemented RPMI 1640 using controlled cooling (Nalgene Cryo 1°C freezing container, Nalge Nunc International, Naperville, IL). For flow cytometric DNA distribution analyses, 1 × 10⁶ cells per sample point were fixed in 70% ice-cold ethanol.

A patient with refractory AML (WBC count 24 × 10⁹ cells/L, with 88% blasts) and a patient with refractory CML in myeloid blast transformation (WBC count 42 × 10⁹ cells/L, with 89% blasts) received 10 and 15 mg/m² CPT-11, respectively, over 90 minutes daily for 5 days, with 1 g/m² cytarabine (Ara-C) 12 hours after each CPT-11 dose, on a Roswell Park Cancer Institute phase I protocol (RPC 9901). Peripheral blood was sampled following the first CPT-11 dose at the times detailed in Results. Samples were enriched for nucleated cells by ammonium chloride lysis and then cryopreserved in 20% DMSO. The study was approved by the Roswell Park Cancer Institute Scientific Review Committee and Institutional Review Board.

HL60 cells were cultured without drug (control) and with SN-38 at 0.1, 0.3, and 1.0 μmol/L for 2 hours. The microarray analysis compared gene transcription profiles of drug-treated cells to those of control cells sampled at the same time points. Time points of blood samples obtained from the two patients after the start of the CPT-11 infusion were the same as the HL60 sampling time points. For the patient samples, the microarray analysis compared gene transcription profiles of cells sampled following drug exposure to that of a sample taken immediately before the start of the infusion.

Total RNA was extracted using RNeasy Midi kits (Qiagen, Inc., Valencia, CA) according to manufacturer's instructions. After elution, RNA samples were concentrated by ethanol precipitation at −20°C overnight and resuspended in nuclease-free water. Before labeling, RNA was measured with a Genequant spectrophotometer (Amersham Biosciences, Piscataway, NJ) and evaluated for degradation with a 2100 Bioanalyzer (Agilent, Palo Alto, CA). Only samples with A_{260/280 nm} ratios of 1.9–2.0 and 28S:18S ribosomal band ratios of >1.5 were analyzed.

The cancer-specific arrays used in this study were produced in the Roswell Park Cancer Institute Microarray and Genomics Core Facility. cDNA clones (n = 3,011; Research Genetics, Huntsville, AL) were selected based on their association with oncogenesis. Each clone was amplified from 100 ng bacterial DNA by PCR amplification of the insert using M13 universal primers for the plasmids represented in the clone set (5′-TGAGCGGATAACAATTTCACACAG-3′ and 5′-GTTTTCCCAGTCACGACGTTG-3′). Each PCR product (75 μL) was purified by ethanol precipitation, resuspended in 25% DMSO, and adjusted to 200 ng/μL. Printing solutions were spotted in triplicate on type A glass slides (Schott Nexterion, Duryea, PA) using a MicroGrid II TAS arrayer and MicroSpot 2500 split pins (Genomic Solutions, Inc., Ann Arbor, MI).

cDNAs were synthesized and indirectly labeled using the Atlas Powerscript Fluorescent Labeling kit (BD Biosciences Palo Alto, CA). Total RNA from drug-exposed cells and untreated (no drug) reference cells was labeled with Cy5 and Cy3, respectively. For each reverse transcription reaction, total RNA (2.5 μg) was mixed with 2 μL random primers (Invitrogen, Carlsbad, CA) in a total volume of 10 μL, heated to 70°C for 5 minutes, and cooled to 42°C. To this sample was added an equal volume of reaction mix (4 μL of 5× First-Strand buffer, 2 μL of 10× deoxynucleotide triphosphate mix, 2 μL DTT, 1 μL deionized H₂O, and 1 μL Powerscript Reverse Transcriptase) per manufacturer's instructions. After 1-hour incubation at 42°C, the RNA was degraded by 70°C incubation for 5 minutes. The mixture was cooled to 37°C and incubated for 15 minutes with 0.2 μL RNase H (10 units/μL). The resultant amino-modified cDNA was purified, precipitated, and fluoresceinated as described by the manufacturer. Uncoupled dye was removed from the labeled probe by washing thrice using a Qiaquick PCR Purification kit (Qiagen). The probe was then eluted in 60 μL elution buffer and dried down to completion in a SpeedVac.

Because of low RNA yields, the RNA extracted from the AML patient samples was amplified and labeled using the Atlas SMART Fluorescent Probe Amplification kit (Becton Dickinson, Franklin Lakes, NJ; refs. 20, 21) according to the manufacturer's instructions. This amplification was shown not to alter the relative gene expressions by comparing the expression profiles of amplified and unamplified genes in HL60 cells over the entire time course following exposure to 0.3 μmol/L SN-38 (data not shown). Amplified cDNA (2 μg) was then amino-modified and purified before coupling to Cy3- or Cy5-reactive dye (Amersham Biosciences) with the reagents supplied in the Atlas kit. The labeled probe was purified, eluted, and dried as described previously for the unamplified samples.

Before hybridization, the two separate probes were resuspended in 10 μL deionized H₂O, combined, and mixed with 2 μL human Cot-1 (20 μg/μL, Invitrogen) and 2 μL poly(A) (20 μg/μL, Sigma, St. Louis, MO). The probe mixture was denatured at 95°C for 5 minutes, placed on ice for 1 minute, and prepared for hybridization with addition of 110 μL preheated (65°C) SlideHyb # 3 buffer (Ambion, Inc., Austin, TX). After a 5-minute incubation at 65°C, the probe solution was placed on the array in an assembled GeneTAC hybridization station module (Genomic Solutions). The slides were incubated at 55°C for 16 to 18 hours with occasional pulsation of the hybridization solution. After hybridization, the slides were washed in the automated GeneTAC station with reducing concentrations of SSC and SDS. The final wash was 30 seconds in 0.1× SSC followed by a 5-second 100% ethanol dip. The slides were spun dry and scanned immediately on an Affymetrix 428 laser scanner to generate two 10 μmol/L images, one for Cy3 and one for Cy5. Two hybridizations were done for each RNA sample, switching the dyes in the second hybridization to account for possible dye bias.

The hybridized slides were scanned using an Affymetrix 428 scanner to generate high-resolution (10 μm) images for both Cy3 and Cy5 channels. Image analysis was done on the raw image files using ImaGene (version 4.1, BioDiscovery, Inc., El Segundo, CA). Each cDNA spot was defined by a circular region. The size of the region was programmatically adjusted to match the size of the spot. Local background for a spot was determined by ignoring a 2- to 3-pixel buffer region around the spot and then measuring signal intensity in a 2- to 3-pixel area outside the buffer region. Raw signal intensity values for each spot and its background region were segmented using a proprietary optimized segmentation algorithm, which excludes pixels that are not representative of the majority of pixels in that region. The background-corrected signal for each cDNA spot is the mean signal (of all the pixels in the region) minus the mean local background. The output of the image analysis is two tab-delimited files, one for each channel, containing all of the raw fluorescence data.

The output of the image analysis was processed by a Perl program developed at Roswell Park Cancer Institute. Spots that were not significantly above background or had a poor coefficient of variance were excluded. For each spot, the ratio of the background-subtracted mean signals of the two channels was calculated. The ratios were then normalized on the log scale across the entire slide. For each slide, the expression ratios were defined as the log₂ mean of all replicate spots. The results from the two slides that make up the dye swap were then averaged on the log scale and became the final expression ratio of that clone. Hierarchical clustering analysis was done with Stanford's Genecluster software version 2.12. The genes were filtered to only include those that appeared in at least 50% of the time points and had at least a 2-fold differential expression (up or down) at at least one time point. The raw and processed electronic files associated with all of the data presented are available.³

The mRNA levels of survivin and TNF (ligand) superfamily member 9 (TNFSF9) in HL60 cells following exposure to 0, 0.1, 0.3, and 1.0 μmol/L SN-38 were validated using real-time quantitative reverse transcription-PCR (Taqman assay) with a Perkin-Elmer (Boston, MA) ABI PRISM Sequence Detection System. The mRNA level of each gene is expressed relative to that of the endogenous standard (β-actin) measured concurrently from the same RNA extracts and cDNA preparations. cDNA for these experiments was synthesized using SuperScript II reverse transcriptase. The comparative C_T method of quantitation was used as described previously (22). The results are presented as mRNA expression values relative to that of β-actin. The primers and probes for survivin, TNFSF9 and β-actin were purchased from Applied Biosystems, Inc. (Foster City, CA) as ready-to-use kits.

Aliquots of 1 × 10⁶ cells were harvested at the indicated time intervals, fixed in ice-cold 70% (v/v) ethanol/distilled water, and stored at −20°C. Cell pellets were rehydrated by washing twice in PBS and then resuspended for 30 minutes in 1 mL propidium iodide staining buffer [0.05 mg/mL propidium iodide (Molecular Probes, Eugene, OR), 0.1% sodium citrate, 0.02 mg/mL RNase A, 0.37% NP40 (pH 7)] on ice. Cell pellets were washed twice with PBS and then analyzed on a FACScan flow cytometer to assess DNA content. DNA distribution histograms from the flow cytometry data were produced with the WinList software program (Verity Software House, Topsham, ME).

A separate set of experiments was set up to study cells that were in S phase during drug exposure. Cells were labeled with bromodeoxyuridine (BrdUrd; 10 μmol/L, 20 minutes; Roche, Mannheim, Germany), washed with PBS, and then exposed to SN-38 for 2 hours. Following the 2-hour exposure to SN-38, cells were washed twice and cell aliquots (1 × 10⁶) were harvested at the indicated time intervals, fixed in ice-cold 70% (v/v) ethanol/distilled water, and stored at −20°C. BrdUrd labeling was visualized with a fluorescein-labeled anti-BrdUrd monoclonal antibody (Roche Diagnostics Corporation/Roche Applied Science, Indianapolis, IN) according to the manufacturer's instructions. Bivariate analysis of BrdUrd/propidium iodide staining was done as described previously (23).

The technical reproducibility of the cDNA microarray analysis was first assessed by a dye-swap study in which the Cy3 and Cy5 labels were interchanged in two preparations of the same sample to show lack of dye bias. Dye swaps in three independently prepared samples of HL60 cells in logarithmic phase yielded r² correlation coefficients of 0.92, 0.89, and 0.95, indicating excellent correlation.

Biological reproducibility was assessed by comparing three independently prepared samples of HL60 cells in logarithmic growth. The gene transcription profile of each preparation was compared with that of the other two; the three possible hybridization combinations yielded r² correlation coefficients of 0.92, 0.96, and 0.94, indicating excellent correlation.

Hierarchical clustering, applying an arbitrary filter to only include those genes that appeared in at least 50% of the time points and had at least a 2-fold differential expression (up or down) at at least one time point, identified groups of genes with similar kinetic patterns of gene expression in drug-treated cells (Fig. 1). However, the function of the genes comprising each group was heterogeneous (data not shown), indicating that hierarchical clustering of gene expression ratios may not be the optimal approach to identifying biologically relevant associations.

In addition to the hierarchical clustering analysis, which groups genes according to similarities in kinetic changes in gene expression, the data were also analyzed by grouping genes according to biological function. Diverse gene transcription profiles, including both increases and decreases in expression, were observed within biological function groups. In addition, among genes with a single type of response (increase or decrease), different time kinetics were observed.

Notable changes in transcription of cell cycle–associated genes (arbitrary threshold >1.87-fold increase or decrease compared with control) included down-regulation of cyclin A2, cell division cycle (CDC) 2, CDC7-like 1, cyclin F, CDC27, centromere protein E, and kinesin-like 5, with recovery to control values 12 hours following drug exposure (Fig. 2). v-abl Abelson murine leukemia viral oncogene homologue 1 and cyclin-dependent kinase (CDK) 4 were also down-regulated, but for a shorter period. BTG family member 2 was up-regulated for at least 12 hours after drug removal. In contrast, B-cell chronic lymphocytic leukemia (CLL)/lymphoma (BCL) 6 and CDK inhibitor 2A (p16) were also up-regulated but quickly returned to control levels.

With the same arbitrary threshold of >1.87-fold increase or decrease compared with control, the apoptosis-associated genes BCL10, BCL2/adenovirus E1B 19-kDa interacting protein 3, TNFSF9, and caspase-6 apoptosis-related cysteine protease were all rapidly up-regulated (within 2 hours of drug removal) and returned to control levels within 8 to 12 hours after drug removal (Fig. 3). TNFRSF10B (death receptor 5) and BCL2-related protein A1 were also up-regulated, but transcription peaked 4 to 6 hours after drug removal. Three different patterns of down-regulation of genes associated with apoptosis were also observed. BCL2 and TNF receptor–associated factor (TRAF) family member–associated nuclear factor-κB activator were down-regulated immediately after drug removal and gradually recovered to control levels over the sampling period. Baculoviral IAP repeat–containing 5 [BIRC5 (survivin)] transcription decreased gradually, nadired 8 hours after drug removal, and then recovered over the next 6 hours. BCL2-like 2 was quickly down-regulated but returned to control levels within 2 hours.

More genes were affected by incubation with 0.3 μmol/L compared with 0.1 μmol/L SN-38. Using the arbitrary threshold of a 1.87-fold difference in transcription, 488 genes were affected by incubation with 0.3 μmol/L SN-38 compared with 368 with 0.1 μmol/L SN-38, but the number of genes affected was not greater with 1 μmol/L SN-38 (453 genes). Using the same arbitrary threshold of a 1.87-fold difference in transcription, 65 genes were affected by exposure to SN-38 at all three concentrations: 28 were up-regulated (Table 1) and 37 were down-regulated (Table 2). There were no appreciable differences in the extent or duration of transcriptional changes of these genes at the different drug concentrations used (data not shown).

To determine whether the gene transcription changes induced by SN-38 in vitro were also induced in vivo, gene transcription was studied in peripheral blasts from two leukemia patients treated with CPT-11. Using the same arbitrary threshold as used for the HL60 data (1.87-fold difference in transcription), 58 genes were found to be up-regulated (Table 3) and 64 genes were down-regulated (Table 4) in cells of the CML patient treated with 15 mg/m² dose, whereas in the AML patient treated with the 10 mg/m² dose ∼1,230 genes were up-regulated and 777 genes were down-regulated (not filtered for overlapping clones identifying the same gene). It is unclear why the number of affected genes in the AML patient were much higher than those observed in the HL60 cells in vitro and the CML patient in vivo. Although preliminary data have shown that the RNA amplification method used for this patient's samples did not affect relative gene expression levels in HL60 cells, a contribution of the amplification to these results cannot be ruled out definitively.

Kinetic gene expression data were compared among seven data sets: two data sets each for HL60 cells exposed to 0.1 and 0.3 μmol/L SN-38, one data set for HL60 cells exposed to 1.0 μmol/L SN-38, one data set for the AML patient treated with 10 mg/m² CPT-11, and one data set for the CML patient treated with 15 mg/m² CPT-11. These data sets were correlated with regard to the extent of treatment-induced changes using Microsoft Access software. Table 5 (up-regulation) and Table 6 (down-regulation) show the results of queries for similarities in changes in gene expression, applying a range of threshold values. When the arbitrary threshold of >1.87-fold difference was applied (as used for the analysis of the HL60 data), only eight genes were similarly affected in all seven data sets. The number of genes similarly affected increased rapidly with decreasing threshold values.

Notably, one of the genes that were down-regulated both in vitro and in vivo was survivin, which is one of only four recognized transcriptomes (genes selectively expressed in human tumors but undetectable or found at very low levels in normal tissue isolated from the same organ; ref. 22). We therefore analyzed the transcription of survivin pathway-associated genes following in vitro SN-38 (Fig. 4) and in vivo CPT-11 (Fig. 5) exposure and found that all were down-regulated both in vitro and in vivo. Some differences were observed between the two patients (Fig. 5), notably with regard to the degree of down-regulation of serine/threonine kinase 12 transcription and the duration of down-regulation of the other genes. The observed differences may reflect interpatient variability and/or different doses received (10 versus 15 mg/m²). Nevertheless, the data in Figs. 4 and 5 show compelling similarities in the kinetic changes in transcription of genes in this specific biological pathway in vitro and in vivo.

mRNA expression of representative down-regulated (survivin) and up-regulated (TNFSF9) genes in HL60 cells as detected by microarray analysis was validated by quantitative reverse transcription-PCR. Expression levels were determined 6 hours following a 2-hour exposure to 0, 0.1, 0.3, or 1.0 μmol/L SN-38 and are presented as ratios to the endogenous β-actin control. The data in Table 7 confirm the SN-38-induced changes in expression levels as observed with the microarray studies. However, whereas no appreciable dose-dependent effects in extent of the changes were observed with the microarray studies, the reverse transcription-PCR results did show a dose-dependent effect, indicating that the latter has a superior quantitative detection sensitivity.

We sought to correlate the observed changes in gene expression with biological effects. To this end, analysis of DNA distribution was done, providing information on both cell cycle kinetics and apoptosis. The DNA distribution histograms of HL60 cells exposed to SN-38 for 2 hours are shown in Fig. 6. Although the DNA distribution of the untreated (control) cells did not change over time, a dose-dependent loss of cells with S-phase DNA content was seen following treatment, accompanied by a gain of cells with a sub-G₁ (apoptotic) DNA content. The loss of cells with S-phase DNA content resolved over time due to recruitment of cells from the G₁ phase.

BrdUrd labeling of S-phase cells before SN-38 exposure allowed determination of the fate of cells in S phase during SN-38 exposure. The bivariate plots of BrdUrd/propidium iodide labeling (Fig. 7) unequivocally show that, whereas the BrdUrd-labeled cells in the control cultures traverse the cell cycle, the BrdUrd-labeled cells following SN-38 exposure appear in the sub-G₁ (apoptotic) compartment of the DNA distribution histograms. These data are consistent with our previous report demonstrating both rapid induction of apoptosis in HL60 cells in S phase during camptothecin treatment and recruitment of cells from the G₁ compartment using BrdUrd labeling (24).

We sought to elucidate whether CPT-11 alters expression of genes associated with cell cycle or apoptosis as a basis for its synergy with other chemotherapy drugs. Gene expression profiles were analyzed in human leukemia cells following in vitro exposure to SN-38, the active metabolite of CPT-11, and following in vivo therapy with CPT-11. Gene transcription profiles observed in vitro were compatible with temporary delay of G₁-S cell cycle transition, loss of S-phase cells, unperturbed G₂-M-G₁ cycle transition, and enhanced apoptotic response, and similar drug-induced changes were detected for selected genes following treatment in vivo. To our knowledge, this is the first demonstration that specific pharmacogenomic effects detected by gene expression profiling in vitro following exposure to SN-38 also occur in vivo following administration of CPT-11.

Cell cycle–related changes included down-regulation o fcyclin A2 and CDC2 and subsequent restoration to control values at 12 hours. These changes are consistent with the loss of S-phase cells, temporary G₁-S block, and subsequent reentry into S phase shown in the DNA distribution studies following exposure to SN-38. Cyclin A2 regulates the G₁-S- and G₂-M-phase transitions in the eukaryotic cell cycle by binding to and activating CDC2 (25, 26). The temporary down-regulation of genes with constant expression throughout the cell cycle, such as CDC7-like 1 (CDC7) and v-abl Abelson murine leukemia viral oncogene homologue 1, is also consistent with SN-38-induced loss of S-phase cells. Cyclin F is thought to be involved in the G₂ cell cycle phase based on the accumulation of cyclin F protein during interphase and its subsequent rapid degradation (27). The observed down-regulation of cyclin F following exposure of HL60 cells to SN-38 and its slow recovery to control levels are consistent with a temporary lack of recruitment of cells into G₂. This is further supported by the temporary reduction of expression of genes encoding G₂-M-phase–specific proteins, such as CDC27 (28), centromere protein E (29), and kinesin-like 5 (CHO1; ref. 30). CDC27 is a component of the so-called anaphase-promoting complex that plays an essential role in the metaphase-to-anaphase transition. Centromere protein E and kinesin-like 5 are kinesin-like motor proteins that are involved in transport of organelles within cells and movement of chromosomes during cell division.

Other changes consistent with a temporary SN-38-induced block in G₁-S transition include temporary down-regulation of CDK4 and increased expression of BTG family member 2 and BCL6. CDK4 is a catalytic subunit of the protein kinase complex that is important for cell cycle G₁-phase progression (31). The activity of this kinase is restricted to the G₁-S phase. CDK4 is required for cell cycle progression; it phosphorylates the retinoblastoma gene product (Rb) and inactivates its repressor function. Following SN-38 exposure, CDK4 expression in HL60 decreases slightly, reaching its nadir 1 hour following drug exposure and subsequently returning to normal levels at 12 hours. The expression kinetics of CDK inhibitor 2A (p16), which inhibits CDK-4 (32), are opposite to those observed for CDK-4, suggesting a causal relationship. SN-38 exposure also leads to rapid up-regulation of BTG family member 2 and BCL6, with subsequent slow decrease to control levels. BTG family member 2 belongs to a family of structurally related proteins that seem to have antiproliferative properties and are involved in the regulation of the G₁-S transition of the cell cycle (33). BCL6 is a zinc finger transcription factor that acts as a sequence-specific repressor of transcription (34).

Proapoptotic changes in gene expression following SN-38 exposure included down-regulation of BCL2, an integral inner mitochondrial membrane protein that blocks apoptotic death (35), and up-regulation of BCL10, which encodes a protein with a caspase recruitment domain and has been shown to induce apoptosis (36). Increased expression was also observed for BCL2/adenovirus E1B 19-kDa interacting protein 3, TNFRSF10B (death receptor 5), TNFSF2, TNFSF9 (4-1 BB-L), caspase-6 apoptosis-related cysteine protease, and TRAF family member–associated nuclear factor-κB activator/TRAF. BCL2/adenovirus E1B 19-kDa interacting protein 3 is a member of the BCL2/adenovirus E1B 19-kDa interacting protein family and has been associated with proapoptotic function (37). The dimeric mitochondrial protein encoded by this gene is known to induce apoptosis even in the presence of BCL2. It binds antiapoptotic viral E1B 19-kDa protein and cellular Bcl2 protein. The protein encoded by TNFRSF10B is a member of the TNFRSF and is involved in apoptosis signal transduction (38). The TNF gene encodes a multifunctional proinflammatory cytokine that belongs to the TNFSF and is involved in the regulation of a wide spectrum of biological processes, including cell proliferation and apoptosis (38). The protein encoded by TNFSF9 is a TNF-related ligand that induces activated T lymphocytes to proliferate and is involved in antigen presentation (39). Caspase-6 apoptosis-related cysteine protease encodes a protein that is a member of the cysteine-aspartic acid protease (caspase) family. Sequential activation of caspases plays a central role in the execution phase of cell apoptosis, and caspase-6 apoptosis-related cysteine protease is thought to function as a downstream enzyme in the caspase activation cascade (40). The TRAF family of proteins associate with and transduce signals from members of the TNFRSF and are required for TNF receptor signaling (41). These proapoptotic molecular responses observed in our study are in line with published CPT-11/SN-38 augmentation of extrinsic apoptotic response signal transduction pathways (12–17).

Down-regulation of BIRC5 (survivin) was also observed following CPT-11/SN-38 exposure in vitro and in vivo. The protein encoded by this gene is an apoptosis inhibitor that is expressed during the G₂-M phase of the cell cycle. BIRC5 associates with the microtubules of the mitotic spindle, and any disruption results in the loss of apoptosis activity (42). Inhibition of signal transducers and activators of transcription 3 (STAT3) signaling has been described to induce apoptosis and decrease survivin expression in primary effusion lymphoma (43). Downstream, survivin stimulates Aurora-B kinase (serine/threonine kinase) activity and helps correctly target Aurora-B to its substrates during the cell cycle (44). One of the substrates targeted by Aurora-B is topoisomerase IIα (45). Recently, it has been suggested that an approach to targeting survivin in cancer therapy is the use of CDC2 antagonists, because CDC2 phosphorylates survivin at the Thr³⁴ position (42). The microarray analysis showed that SN-38 down-regulates STAT3, Aurora-B kinase, topoisomerase IIα, and CDC2, suggesting that CPT-11 may be an interesting candidate for pharmacologic targeting of survivin.

In addition to proapoptotic changes in gene expression, antiapoptotic changes were also observed. Of note, SN-38 exposure induced up-regulation of BCL2-related protein A1 (BCL2A) and BCL2-like 2 (bcl-w). The protein encoded by BCL2-related protein A1 reduces cytochrome c release from mitochondria and blocks caspase activation (46). Expression of BCL2-like 2 in cells has been shown to reduce apoptosis (47). Biological responses are commonly controlled by a balance between stimulatory and inhibitory mechanisms, and the preponderance of proapoptotic signals following SN-38 exposure suggests that the antiapoptotic changes may be relatively inconsequential.

It should be noted that although some of the fold changes in expression detected were small (e.g., in the survivin-associated genes in vivo) the patterns shown by the kinetic analysis gave support to the concept that the observed increases and decreases in expression were true time-dependent changes rather than background noise, while this conclusion could not have been drawn from studying only a single posttreatment sample. However, it will be necessary to determine what levels of changes are associated with biologically significant responses.

The gene expression data reported here have implications for the use of CPT-11 in combination regimens. Demonstration of changes in expression of cell cycle–related genes induced by SN-38, associated with a temporary delay of G₁-S transition, corroborates a sequence-dependent synergistic interaction of this drug with S-phase–specific agents. In addition, demonstration of proapoptotic changes in gene expression corroborates the sequence-dependent synergistic interaction of SN-38 with inducers of extrinsic apoptotic pathways, such as TNF-related apoptosis-inducing ligand/Apo2L. In addition, the in vitro and in vivo effects of SN-38/CPT-11 on the survivin pathway suggest that the proapoptotic effects of SN-38 may not be restricted to the extrinsic apoptotic pathway. Further studies may be helpful in optimizing the timing of combination regimens.

This is, to our knowledge, the first study comparing changes in gene transcription profiles induced by SN-38 in vitro and CPT-11 in vivo. Future studies will seek to increase the sophistication of analysis of changes in gene expression. Moreover, to fully comprehend the significance of the observed gene transcription changes, correlative studies of protein expression and phosphorylation are required and additional correlative studies are required to associate changes in gene and protein expression with biological responses. The continual enhancement of bioinformatic analysis programs and the further development of high-throughput techniques in these areas are expected to facilitate these comprehensive studies in the very near future.