Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Doxorubicin is important to the treatment of a variety of cancers; however, the development of drug resistance reduces its effectiveness. Factors reported to be involved in doxorubicin resistance include altered expression of topoisomerase II (Nielsen et al., 1996) and integrins (Damiano et al., 1999), changes in glutathione levels (Sinha et al., 1989), and expression of membrane-associated pumps such as P-glycoprotein encoded by the multidrug resistance gene MDR1 (Dalton, 1997). In the human multiple myeloma cell line, RPMI 8226, doxorubicin selection at 60 nM resulted in a resistant variant, 8226/Dox6. Further selection of 8226/Dox6 with 400 nM doxorubicin led to the highly resistant 8226/Dox40 cell line. Both 8226/Dox cell lines possess a multidrug-resistant phenotype. In association with this phenotype, 8226/Dox6 expresses MDR1 from a single gene copy while 8226/Dox40 expresses higher levels of P-glycoprotein from amplified MDR1 and in turn displays higher levels of multidrug resistance (Futscher et al., 1993).

While MDR1 expression plays a significant role in the multidrug resistance phenotype in the RPMI 8226, several observations indicate that additional factors are involved. For example, MCF-7 cells transfected with MDR1 compared with MCF-7 cells that expressed MDR1 as a result of selection with doxorubicin were 15-fold more resistant than the transfectants, even though the MDR1 transfectants produced comparable levels of P-glycoprotein (Zyad et al., 1994). Another study compared cell lines that expressed P-glycoprotein from amplified MDR1 and found that the cell line expressing the most P-glycoprotein was 10-fold less resistant than a cell line with lower MDR1 expression (Dolci et al., 1993). Similarly suggestive results were found using the 8226 cell lines. When 8226/Dox40 was grown out of doxorubicin selection for 53 weeks, the 8226/Dox40 cell line lost only 22% of its P-glycoprotein expression while losing 87% of its drug resistance. Finally, the 8226/Dox40 cell line is cross-resistant to the apoptotic stimulus of Fas ligand, even though Fas ligand is not a substrate for P-glycoprotein (Landowski et al., 1997).

One explanation for the cross-resistance of 8226/D40 to Fas ligand would be defect(s) in the apoptotic signaling that begins upon Fas ligation. Evidence supporting an anti-apoptotic role for MDR1 in addition to its drug efflux activity has been reported. Studies have found evidence that P-glycoprotein prevents formation of the proapoptotic ceramide in response to chemotherapeutics including doxorubicin (Cabot et al., 1999; Côme et al., 1999; Sietsma et al., 2000). Ceramide is cleaved from sphingomyelin residing in the inner leaflet by sphingomyelinases following stimulation by chemotherapeutics and the death receptor ligands Fas tumor necrosis factor- (TNF-). P-glycoprotein has been reported to act as a broad specificity lipid translocase that reduces the availability of inner leaflet sphingomyelin and thus prevents production of ceramide (van Helvoort et al., 1996). In addition, it has been recently shown that P-glycoprotein-positive samples from patients with acute myeloblastic leukemia were more resistant to induction of ceramide-mediated apoptosis than P-glycoprotein-negative samples (Pallis and Russell, 2000). Doxorubicin-triggered apoptosis has gained attention as evidence has accumulated that many chemotherapeutics kill by activating the apoptotic processes (Dive and Hickman, 1991; Hannun, 1997; Los et al., 1997). Thus defects in apoptotic signaling could explain the cross-resistance to a wide variety of apoptosis inducers displayed by cells selected for resistance to doxorubicin and other chemotherapeutics (Smyth et al., 1998; Johnstone et al., 1999).

Evidence indicates ceramide is a key molecule in signaling apoptosis in some cell types following treatment with TNF-, Fas ligand, or chemotherapeutics such as doxorubicin (Hannun, 1996; Jaffrézou et al., 1996; Mathias et al., 1998). Furthermore, inhibition of ceramide production and metabolism reduces apoptotic death in MCF-7 cells following doxorubicin exposure (Lucci et al., 1999). Evidence exists for many downstream targets of ceramide, all of which can mediate the apoptotic/survival balance within cells (Kolesnick and Krönke, 1998).

With this background in mind, we have analyzed the RPMI 8226 cell line and its multidrug-resistant variants, 8226/Dox6 and 8226/Dox40, using 5760-element cDNA microarrays to identify differential gene expression. The cDNA microarray analysis indicated that 7% of genes had some degree of altered expression as a result of doxorubicin selection. A subset of the differentially expressed genes was of particular interest because of the genes' roles in apoptotic signaling. Defects in apoptosis signaling provide an explanation of the 8226/Dox cells' cross-resistance to Fas and staurosporine, indicating the functional significance of the altered gene expression. These results show that genes are affected at the expression level by doxorubicin selection, which contributes to the multidrug-resistant phenotype by reducing apoptotic signaling.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Probe Preparation. Human cDNA bacterial clones were purchased from Research Genetics (Huntsville, AL). Referred to as gf200, the set consists of 5184 sequence-validated I.M.A.G.E. consortium bacterial clones. Approximately 3000 of these clones represent known genes, while the remaining clones represent expressed sequence tags. cDNA targets were produced by PCR amplification of the cDNA inserts directly from bacterial cultures. In brief, individual I.M.A.G.E. clones were grown in 96-well plates at 37°C for 6 h. One microliter of the bacterial culture was added to a 96-well plate containing 45 µl of premixed PCR reaction (Marsh BioProducts, Rochester, NY) and 4 µl of primer (2 µM, catalog number: GF200.primer, Research Genetics). Primers and unincorporated nucleotides were removed following the PCR amplification (initial denaturation step was 96°C for 30 s, followed by 40 cycles of 94°C for 30 s, 55°C for 45 s, and 72°C for 2.5 min) using a 96-well PCR clean up kit from Qiagen (Valencia, CA). PCR amplification and purification were verified by agarose gel electrophoresis, and PCR product yield was determined using a PicoGreen-based fluorescence assay (Molecular Probes, Eugene, OR) in a 96-well format. Typical yields ranged from 1 to 5 µg. After quantitation, the purified PCR products were dried and resuspended in 10 µl of 2× SSC for printing onto slides.

Microarray Fabrication. cDNAs were printed onto chemically activated glass slides (CEL industries, Houston, TX) using four quill-type pins (catalog no. SMP4, Telechem International, San Jose, CA) mounted onto an OmniGrid robot (GeneMachines, San Carlos, CA). In addition to the 5184 I.M.A.G.E. consortium clones, a set of 88 human housekeeping genes (Research Genetics), a set of 8 Mesembryanthemum crystallinum genes, and Cy3 or Cy5 end-labeled oligonucleotides were placed strategically into the array to aid in data normalization, the measurement of nonspecific hybridization, and the identification of the corners of the array, respectively. Additionally, a set of 103 I.M.A.G.E. clones representing known genes of interest not found in gf200 were purchased from Research Genetics and included in the microarray. A list of the clones on the arrays can be found at http://microarray.azcc.arizona.edu/HUMAN5KLIST. After printing, the slides were placed in a humidity chamber overnight in the dark. The following day, slides were washed 1 min in 0.1% sodium dodecyl sulfate and 1 min in double distilled water at room temperature with mild agitation. Slides were then submerged in 240 ml of 75% (v/v) water/ethanol solution into which 0.6 g of sodium cyanoborohydride had been freshly dissolved. After 5 min at room temperature, slides were washed four times in double distilled water for 2 min with mild agitation and spun dry at 500g for 1 min. Slides were stored in the dark at room temperature and <40% humidity until use.

Cell Culture. RPMI 8226, 8226/Dox6, and 8226/D40 cells were cultured at 37°C in 95% air/5% CO₂ and RPMI 1640 medium supplemented with 5% fetal bovine serum, 50 I.U./ml penicillin, 50 mg/ml streptozotocin, and 2 mM L-glutamine. 8226/Dox6 was maintained in 60 nM doxorubicin, while 8226/D40 was maintained in 400 nM doxorubicin (Sigma Chemical Co., St. Louis, MO). Prior to mRNA isolation, 8226/Dox6 and 8226/D40 were cultured 1 week without doxorubicin to eliminate acute effects of doxorubicin from confounding the microarray results.

Target Preparation. Fluorescent first strand cDNA was made from 4 µg of poly(A⁺) RNA in the presence of 50 µM Cy5-dCTP or Cy3-dCTP (Amersham Pharmacia Biotech Inc., Piscataway, NJ) in a 25-µl volume containing the following: 500 ng of oligo(dT)_12-18, 1× Superscript buffer, 400 U of Superscript II, 3.3 U of RNase inhibitor (all from Invitrogen, Grand Island, NY), 400 µM each of dGTP, dATP, dTTP, 100 µM dCTP, and 10 mM dithiothreitol. All reagents except the Superscript II were mixed on ice, placed at 65°C for 5 min and then 25°C for 5 min, at which point the Superscript was added and the mixture heated to 42°C for 2 h. The mRNA template was hydrolyzed by heating the reaction for 5 min at 95°C, adding 6.25 µl of 1 M NaOH, and incubating for 10 min at 37°C. Neutralization was achieved by the addition of 6.25 µl of 1 M HCl. Labeled cDNA from two reactions (one Cy3-labeled, one Cy5-labeled) was combined and purified on a microcon-50 column using four buffer exchanges (the first three were double-distilled water, the final exchange was 10 mM Tris-HCl, pH 7.5). After elution from the column, probe was lyophilized to dryness, and resuspended in 10 µl of hybridization buffer (2× SSC, 0.1% SDS, 100 ng/µl Cot1 DNA, 100 ng/µl oligo dA), denatured by boiling for 2.5 min, and added to a denatured (2-min boil slide in double distilled water, plunge into room temperature ethanol, spin dry at 500g) microarray. A cover slip (22 × 22 mm) was applied, and the array was placed in a hybridization chamber (catalog number HYB-03, GeneMachines) at 62°C for 18 h. Following hybridization, slides were washed by placing them into 50-ml conical tubes containing 2× SSC, 0.1% SDS for 5 min, 0.06× SSC, 0.1% SDS for 5 min, and 0.06× SSC for 2 min all at room temperature. Slides were scanned for Cy3 and Cy5 fluorescence using a Axon GenePix 4000 microarray reader (Axon Instruments, Foster City, CA) and quantitated using GenePix software. The RPMI 8226 versus 8226/Dox6 hybridizations were performed in triplicate, and the RPMI 8226 versus 8226/Dox40 hybridizations were performed seven times.

Northern Blot Analysis. Selected results obtained by cDNA microarray analysis were confirmed experimentally by Northern analysis. Some genes were selected based on their potential role in doxorubicin resistance or apoptosis signaling. Other genes were selected based simply on their expression ratios; genes were analyzed that were underexpressed, overexpressed, or unchanged in 8226/Dox6 and 8226/D40 relative to RPMI 8226. Briefly, 10 µg of total RNA from RPMI 8226 and 8226/Dox40 or RPMI 8226, 8226/Dox6, and 8226/Dox40 was size separated on a 1% denaturing formaldehyde agarose gel, transferred to a nylon membrane, and hybridized with radiolabeled probes specific to the gene to be analyzed. Probe templates were obtained by PCR amplification of the cDNA insert from the respective I.M.A.G.E. Consortium clone used on the cDNA microarrays.

Data Analysis. Fluorescence intensity measurements from each array element were compared with local background and divided into two groups: those in which signal was more than 1.4 times the background in both the Cy3 and Cy5 channels, and those in which signal was 1.4-fold background in only one channel. Based on the variation of the data within each hybridization, a 90% confidence limit was determined for elements with signal over 1.4-fold background in both channels. Elements with signal below 1.4-fold background in one channel were designated as possible cases of a gene being turned on or off. A gene found to be differentially expressed in four of the seven hybridizations that compared RPMI 8226 with 8226/Dox40 was considered further for relevance to the multidrug-resistant phenotype. Next, a gene differentially expressed between RPMI 8226 and 8226/Dox40 was checked for evidence of differential expression in the RPMI 8226 versus 8226/Dox6 hybridizations.

Apoptosis Analysis. RPMI 8226 and 8226/D40 cells grown in log phase in the absence of doxorubicin for 1 week were treated with 100 nM or 500 nM staurosporine for 23 h. Cells were washed and incubated with propidium iodide and fluorescein isothiocyanate-conjugated Annexin-V according to the manufacturer's instructions (CLONTECH, Palo Alto, CA) and compared with untreated cells using flow cytometry.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

cDNA Microarray Analysis of a Multidrug-Resistant Model. To identify changes in gene expression that contribute to the multidrug-resistant phenotype, the doxorubicin selected, multidrug-resistant cell lines 8226/Dox6 and 8226/Dox40 were compared with their doxorubicin-sensitive parent RPMI 8226. 8226/Dox6, which was selected with 60 nM doxorubicin, represents an intermediate level of resistance while 8226/Dox40 (selected at 400 nM doxorubicin) is highly drug resistant (Futscher et al., 1993). Following identification of genes that could contribute to the multidrug-resistant phenotype in 8226/Dox40, the 8226/Dox6 cell line was analyzed as a way to observe the emergence of the altered gene expression seen in the highly resistant 8226/Dox40 cell line. 5287 unique human genes were examined for differential expression using two-color fluorescence hybridization to glass cDNA microarrays manufactured on site. Cy3-labeled cDNA from either 8226/Dox6 or 8226/Dox40 was simultaneously hybridized with Cy5-labeled cDNA from RPMI 8226. RPMI 8226 was compared with 8226/Dox40 seven times and compared with 8226/Dox6 three times.

View larger version (93K): [in this window] [in a new window]   Fig. 1.   Identification of genes potentially involved in the multidrug-resistant phenotype. A false color representation of 8226/Dox40 cDNA (labeled with Cy3, shown in red) cohybridized against drug-sensitive RPMI 8226 cDNA (labeled with Cy5, shown in green) is shown. The 5760-element array consists of four subarrays delineated at their corners by Cy3-labeled oligomers, seen as bright red elements in duplicate; each subarray contains test elements interspersed with 8 nonhuman negative controls and 88 human housekeeping gene positive controls. A yellow element represents equal expression of a particular gene in the two cell lines analyzed. Differential gene expression is seen as deviation of an element's color toward either the red (Cy3) or the green (Cy5) channel. A portion of the cDNA microarray is magnified at right to show the element representing the MDR1 gene, which is expressed only in the multidrug-resistant 8226/Dox40 cells. Note that the bottom half of each subarray contains ESTs and thus represents few expressed elements.

Confirmation of Differential Gene Expression by Northern Analysis. To verify the microarray results, genes with increased, decreased, or unchanged expression in response to doxorubicin selection were analyzed by Northern blot analysis. Figure 2 shows a comparison of the expression ratios determined by cDNA microarray analysis and Northern blot analysis for 30 genes. Overall, good agreement was seen between the two techniques for analyzing gene expression, except for glutathione S-transferase, a facilitated glucose transporter, and fibronectin (labeled gene numbers 25, 27, and 28, respectively, in Fig. 2).

View larger version (48K): [in this window] [in a new window]   Fig. 2.   Comparison of cDNA microarray expression ratios (8226/Dox40 expression over RPMI 8226 expression) with those determined by Northern analyses for 30 genes. The genes chosen for analysis represent levels of expression that increased, decreased, or remained the same as a result of doxorubicin selection. For genes that were turned on or off as a result of doxorubicin selection, expression is plotted as 10 or 0.1, respectively.

View larger version (22K): [in this window] [in a new window]   Fig. 3.   Northern analysis confirmation of cDNA microarray results. Equal amounts of RNA from RPMI 8226, 8226/Dox6, and 8226/D40 were size-separated and hybridized with radiolabeled probes specific to the genes indicated. The Northern analyses confirmed the microarray results that showed that expression of some genes progressively increased (a, c, f) while others progressively decreased (b, g) as a result of doxorubicin selection. PTP, protein tyrosine phosphatases; PP1, protein phosphatase 1; SSAT, spermine/spermidine acetyltransferase.

Identification of Genes with Potential Roles in the Multidrug Resistance Phenotype. Once differentially expressed genes were identified and confirmed by Northern blot analyses, the significance of the changes in gene expression were viewed in terms of their potential effect on doxorubicin toxicity and the multidrug resistance phenotype. As a starting point, it was noted that cellular death in response to doxorubicin is dose-dependent, with apoptosis being the predominant form of death at clinically relevant doses (Muller et al., 1997). Exposing cells to clinically relevant doses of doxorubicin stimulates the production of ceramide by neutral sphingomyelinase (N-SMase) and ceramide produced in this manner can then act as a second messenger signaling apoptosis (Jaffrézou et al., 1996). Additionally, exogenous ceramide can mimic the effects of doxorubicin, and inhibition of ceramide production or protease activation can reduce induction of apoptosis by doxorubicin treatment (Mansat et al., 1997). Two additional observations specific to the 8226/Dox cell lines pointed to defects in apoptotic signaling. First, the antiapoptotic BCLxL is overexpressed in the 8226/Dox40 cell line compared with RPMI 8226 (Tu et al., 1998); and second, 8226/Dox6 and 8226/Dox40 are cross-resistant to Fas ligand but are Fas-receptor-positive (Landowski et al., 1997). In light of these observations, we searched our microarray results for differentially expressed genes with roles in apoptotic signaling and execution. Twenty-nine genes with potential involvement in apoptotic signaling or growth and survival were identified in the list of genes differentially expressed between RPMI 8226 and 8226/Dox40. Additionally, 20 of the 29 genes identified in 8226/D40 were also found in the comparison of RPMI 8226 and 8226/Dox6 (Table 1).

Altered Apoptotic Response to Staurosporine in 8226/D40. The microarray results indicated that multiple genes involved in apoptosis signaling were altered in the doxorubicin-selected cell lines. These results led us to predict that the doxorubicin-selected cell lines would possess a broad resistance to apoptosisa phenotype already hinted at by their resistance to Fas ligand despite expression of the Fas receptor. To test this hypothesis, we treated RPMI 8226 and 8226/D40 with staurosporine and measured the apoptotic response. The results graphed in Fig. 4 show that 8226/D40 had an apoptotic response within 20% of control at both doses of staurosporine, even the high dose of 500 nM. In contrast, RPMI 8226 shows a progressive increase in the number of apoptotic cells up to 200% of control at the 500 nM dose of staurosporine. These results confirm that doxorubicin selection has led to defects in apoptosis signaling suggested by the changes in gene expression found using microarrays.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Development of multidrug resistance prevents successful chemotherapy of many tumors, including multiple myeloma. Although drug efflux mechanisms have been shown to play an important role in the multidrug-resistant phenotype, evidence for additional contributing factors led us to use cDNA microarrays to search for differentially expressed genes in the 8226 cell line model of multidrug resistance. By using a model of progressive multidrug resistance, we were able to increase confidence in the microarray analyses by requiring genes to lose or gain expression progressively with increasing doxorubicin resistance. The microarray results identified a subset of differentially expressed genes in the 8226 cell line model that included MDR1. Within the subset of differentially expressed genes, many were of particular interest due to their role in apoptotic signaling, their potential to provide resistance to doxorubicin, and their ability to explain the cross-resistance to Fas ligand seen in the 8226/Dox cells. To confirm the defects in apoptotic signaling suggested by the microarray results, the 8226/D40 cells were compared with RPMI 8226 in their response to staurosporine. 8226/D40 cells were found to be resistant to apoptotic doses of staurosporine as compared with RPMI 8226.

To confirm the microarray results, 30 Northern analyses were performed; of these, 3 could not be interpreted and cast doubt on the differential expression found by microarray analysis. The genes with inconclusive Northern results were glutathione S-transferase, a facilitated glucose transporter, and fibronectin (25, 27, and 28, respectively, in Fig. 2). The Northern analyses of these three genes produced more than three bands of hybridization; the fact that the genes came from large gene families makes cross-hybridization of the probe to multiple homologous sequences a likely explanation. Overall, however, the results of the Northern analyses support the microarray results and indicate that multiple changes in gene expression occurred in response to doxorubicin selection. Furthermore, the Northern analyses that included 8226/Dox6 showed that the changes in gene expression progressed with increasing drug resistance. This correlation between increasing drug resistance and progressive increases or decreases in gene expression provides support for their involvement in the multidrug-resistant phenotype.

The progressive nature of the changes in gene expression suggests they either play a functional role in doxorubicin resistance directly or are coordinately regulated with a gene that plays a functional role in doxorubicin resistance. Previous reports demonstrating that doxorubicin stimulates apoptosis within the dose range used to select the 8226/Dox cells (Muller et al., 1997), that BCLxL expression is altered (Tu et al., 1998), and that ceramide is involved in doxorubicin toxicity (Kolesnick and Krönke, 1998) suggested that we should find genes involved in apoptosis and cell survival in the list of differentially expressed genes. Indeed, 29 genes with potential roles in apoptotic signaling and execution, especially as mediated by ceramide, were identified between RPMI 8226 and 8226/D40, 20 of the 29 genes were also differentially expressed between RPMI 8226 and 8226/Dox6. When viewed as a whole, the roles the genes listed in Table 1 play in apoptosis and cell survival suggest that, rather than simply being coordinately regulated with a resistance gene such as MDR-1, they play a functional role in the multidrug resistance phenotype. Figure 5 shows a model illustrating how changes in expression of the genes listed in Table 1 could provide resistance to doxorubicin-induced apoptosis.

View larger version (27K): [in this window] [in a new window]   Fig. 4.   Altered gene expression associated with resistance to apoptosis in 8226/D40. RPMI 8226 and 8226/D40 cells in log phase growth were treated with 100 and 500 nM staurosporine. Twenty-three hours later, cell flow cytometry was used to analyze cells using Annexin-V-conjugated fluorescein isothiocyanate and propidium iodide to detect apoptotic and dead cells.

Defects in ceramide-mediated apoptotic signaling within the 8226/Dox cells likely begin with lower availability of inner leaflet sphingomyelin due to lipid translocase activity of MDR1-encoded P-glycoprotein (van Helvoort et al., 1996). Next, the doxorubicin-selected cells display a progressive loss of FAN expression (Fig. 3b; Table 1) that would decrease ligand-bound Fas receptor-induced N-SMase activation (Adam-Klages et al., 1996; Kreder et al., 1999). Finally, the 8226/Dox cells have altered their metabolism of arachidonic acid through increased expression of cyclooxygenase 1. A previous report showed that N-SMase was stimulated by arachidonic acid when L929 cells were treated with TNF- (Jayadev et al., 1994). The link between arachidonic acid and N-SMase was further supported when selection for TNF- resistance led to a loss of arachidonic acid production (Jayadev et al., 1997). Thus, the increase in cyclooxygenase 1 expression observed with increasing multidrug suggests the 8226/Dox cells increased arachidonic acid metabolism, thus reducing its stimulatory effect on ceramide production. Taken together, the 8226/Dox cells have altered expression of three genes to reduce intracellular levels of the proapoptotic messenger molecule ceramide.

Scaffidi et al. (1998) have proposed a categorization for cells' apoptotic response that provides a framework for viewing the changes in 8226/Dox ceramide signaling pathway. They defined two cell types based on formation of the death-inducing signaling complex (DISC) following Fas ligand binding and involvement of the mitochondria in Fas ligand-induced apoptosis. Type I cells activate caspase 8, and subsequently caspase 3, via DISC formation within seconds to minutes of Fas ligand binding. In type I cells, a ceramide-activated, BCL-2/BCLxL-inhibitable mitochondrial permeability transition is secondary. Conversely, type II cells do not appear to activate caspases directly. Instead, the apoptotic signal must travel through the ceramide-triggered mitochondrial permeability transition, with the resulting release of cytochrome c and activation of caspase 3. In addition, activation of caspase 3 in type II cells is delayed at least 30 min or more. When RPMI 8226 cells were treated with 5 µM doxorubicin, activated caspase 8 and caspase 3 fragments did not appear for 4 to 8 h, an indication that doxorubicin likely uses a DISC-independent mechanism of apoptosis induction in 8226 cells (Landowski et al., 1999). Thus, 8226 cells appear to be in the type II category of slow caspase activation that proceeds via a ceramide-signaled mitochondrial permeability transition. Based on this model for doxorubicin-induced apoptosis in 8226 cells, altered expression of genes involved with ceramide signaling takes on additional significance.

In addition to genes whose products alter ceramide production, multiple downstream targets of ceramide were affected by doxorubicin selection. Ceramide signaling has been shown to proceed through a ceramide-activated kinase pathway involving the sequential activation of the ceramide target KSR/CAPK, Ras, Raf, and MEK and finally, activation of BAD via Akt (Basu et al., 1998). Two members of this ceramide-activated kinase pathway, Raf and BAD (Fig. 5), had lower expression in the doxorubicin resistant cells. BAD works to promote apoptosis by forming a heterodimer with, and inactivating, the antiapoptotic protein BCLxL, which has been previously reported to be up-regulated in the 8226/Dox40 cell line (Tu et al., 1998). Taken together, the simultaneous reduction of Raf and BAD in coordination with increased BCLxL expression could provide a potent dampening of ceramide-mediated apoptosis signaling and further indicates a role for mitochondria permeability transition-dependent (type II) apoptosis in doxorubicin toxicity.

In addition to the ceramide arm of apoptotic signaling, there were changes in expression of two caspases responsible for initiation and execution of apoptosis, caspase 8/MACH/FLICE and caspase 3/CPP32, respectively. Caspase 8/Mach/FLICE binds the Fas/CD95 receptor via Fas-associated death domain and is activated by DISC formation. Caspase 8/Mach/FLICE is the induction caspase in Fas-induced apoptosis and is important to the activation of the Fas-signaled caspase cascade (Susin et al., 1997). Our microarray results also indicated decreased expression of the proapoptotic molecule caspase-like apoptosis-regulatory protein (CLARP) (Inohara et al., 1997), and the CED-4 homologous protein FLASH in the 8226/D40 cell line. Both CLARP and FLASH are thought to be involved in formation of the DISC and subsequent activation of caspase 8; thus, their decreased expression in conjunction with the previously reported reduction of caspase 8 expression in 8226/Dox40 (Landowski et al., 1999) could operate to reduce activation of the caspase cascade in the 8226/Dox cells. Decreased expression of DISC components argues against the designation of 8226 cells as type II; however, it should be kept in mind that the cell types defined by Scaffidi et al. (1998) refer to induction of apoptosis by Fas ligand. Nonphysiological inducers of apoptosis such as doxorubicin may trigger both type I and type II signaling from multiple points within the signaling cascades. This possibility is even more intriguing in light of the numerous disputed reports of doxorubicin and other chemotherapeutics triggering Fas/CD95 receptor and Fas ligand transcription and subsequent apoptosis in an autocrine fashion. Alternatively, the reduction in CLARP and FLASH expression may reflect their coregulation with caspase 8.

The downstream effector caspase critical to execution of either type I or type II apoptosis appears to be caspase 3, which can be activated by caspase 8 or as a result of the mitochondrial permeability transition. With the convergence of both type I and type II apoptotic signaling at caspase 3, the progressive loss of caspase 3 expression (Fig. 3g) with increasing doxorubicin resistance may represent a critical defect in the apoptotic pathway in cells with a multidrug-resistant phenotype. The importance of caspase 3 is supported by the finding that doxorubicin-treated CEM cells activated caspases 1 and 3 and subsequently cleaved poly(ADP-ribose) polymerase. In addition, CEM cells selected with 172 nM doxorubicin are analogous to the 82226/D40 cells in that they are also cross-resistant to Fas ligand (Los et al., 1997). The exact mechanism(s) by which doxorubicin stimulates cells in a manner similar to natural death receptor ligands such as TNF- and Fas ligand is unknown; however, it should be noted that in both the CEM cells used by Los et al. and the 8226 cells of this study apoptosis was significantly delayed as compared with Fas ligand induction (Los et al., 1997; Landowski et al., 1999). The delayed cell death and caspase activation produced by doxorubicin in comparison with Fas ligand serves to warn of a limit to which the analogy of doxorubicin with Fas ligand can be used. Regardless of how doxorubicin or other chemotherapeutics affected by multidrug resistance activate endogenous apoptotic signaling pathways, the downstream defects in ceramide and caspase activation found in the 8226/Dox cells likely limit their ability to cause cell death via apoptosis (Fig. 5).

View larger version (32K): [in this window] [in a new window]   Fig. 5.   Scheme showing changes in gene expression as a result of doxorubicin selection. In many cell types, doxorubicin stimulates the formation of ceramide, activation of caspases, and finally, apoptosis. The changes in gene expression shown could work together to reduce these signaling events in the apoptotic process and increase resistance to doxorubicin toxicity in the 8226/Dox6 and 8226/D40 cell lines. Genes shown in boxes had increased expression as a result of doxorubicin selection while those shown in ovals had decreased expression. CAPP, ceramide-activated protein phosphatase; GDI, GDP dissociation inhibitor.

Recent work has supported a vital role for apoptosis in chemotherapy. The results presented here using a multidrug-resistant cell line model indicate that along with drug efflux mechanisms, multiple changes to apoptosis signaling occurred as a result of doxorubicin selection. Future studies into the commonality of the changes observed in the 8226 cell line model should lead to a more refined multidrug resistance fingerprint and new commonly affected targets for reversing of the multidrug-resistant phenotype.