Bexarotene (LGD1069, Targretin), a Selective Retinoid X Receptor Agonist, Prevents and Reverses Gemcitabine Resistance in NSCLC Cells by Modulating Gene Amplification

Lung cancer is the leading cause of cancer death for both men and women (1). The American Cancer Society estimated for 2004 ∼173,770 new cases of lung cancer in the United States, and ∼160,440 people dying of this disease. Non–small-cell lung cancer (NSCLC) is the most common type of lung cancer, accounting for almost 80% of all cases. Chemotherapy remains the only treatment option for patients with unresectable NSCLC. However, the current 5-year survival rate has not improved with this therapy (2). Thus, identification of effective treatment regimens for late-stage disease is urgently needed.

Gemcitabine (2′,2′-difluorodeoxycytidine) is a nucleoside analogue and has shown antitumor activity in various cancers, including lung cancer (3–6). The activity of gemcitabine is based primarily on the formation of gemcitabine triphosphate (dFdCTP), which incorporates into the DNA and acts as a DNA synthesis chain terminator (7). In addition, the diphosphate metabolite of gemcitabine (dFdCDP) was also reported to inhibit ribonucleotide reductase causing a decrease in dCTP, decreased feedback inhibition of deoxycytidine kinase, and enhanced phosphorylation of gemcitabine (8–10). Recent studies have suggested that ribonucleotide reductase large subunit 1 (RRM1) and deoxycytidine kinase play important roles in resistance to gemcitabine (11, 12). Developing new treatment strategies to prevent and overcome gemcitabine resistance will greatly improve the clinical utility of this compound.

We previously showed that the selective retinoid X receptor agonist bexarotene (LGD1069, Targretin) is an efficacious chemopreventive and chemotherapeutic agent in a number of preclinical rodent models of breast cancer (13–16). We further showed that the combination of bexarotene and cytotoxic agents produces synergistic growth inhibition as well as prevents and overcomes acquired drug resistance in several preclinical cancer models (17–21). The encouraging preclinical results with bexarotene, alone or in combination with various cytotoxic agents, led us to further examine its role in treatment of solid tumors. The purpose of this study was to evaluate the influence of bexarotene on the acquisition and maintenance of gemcitabine resistance in human NSCLC. Our results show that inclusion of bexarotene into the gemcitabine treatment regimen prevented and overcame acquired gemcitabine resistance in NSCLC Calu3 cells. This activity was associated with prevention or reversal of rrm1 gene amplification. Moreover, xenograft tumors generated from the gemcitabine-resistant Calu3-GemR cells were unresponsive to gemcitabine or bexarotene as single agents but showed a statistically significant decrease in growth when treated with the bexarotene/gemcitabine combination.

Chemicals and reagents. RPMI 1640, fetal bovine serum, glutamine, and gentamicin were obtained from Life Technologies. Bexarotene was synthesized at Ligand Pharmaceuticals, Inc. Gemcitabine (2′,2′-difluorodeoxycytidine) was purchased through Pharmaceutical Buyers, Inc. Cytosine arabinoside (ara-C), doxorubicin, cisplatin, and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolum bromide were obtained from Sigma Chemicals. Gemcitabine and bexarotene stock solutions in sterile water or DMSO, respectively, were at concentrations that limited the final concentration of the solvent in the culture medium to <0.1%.

Cell lines. The human NSCLC cell line Calu3 was obtained from the American Type Culture Collection. A gemcitabine-resistant derivative of Calu3 cells (Calu3-GemR) was generated as described in Results. All cells were routinely cultured in RPMI 1640 supplemented with 10% fetal bovine serum and 2 mmol/L glutamine in 95% air, 5% CO₂.

In vitro treatment regimens. To determine the effect of bexarotene and/or gemcitabine on the development of gemcitabine resistance, Calu3 cells were seeded at 2 × 10⁶ in T-225 flasks. The treatment schemes are illustrated in Fig. 1A. Briefly, the cells were exposed to the regimens on a 10-day cycle. After a 3-day treatment with gemcitabine alone or in combination with bexarotene the cells were washed, counted, and replated followed by either a 7-day exposure to bexarotene or to control medium. At the end of each treatment cycle, the cells were trypsinized, when possible counted, and then replated onto a new flask and again exposed to the same treatment regimen. This procedure was repeated 10 times. For the gemcitabine single-agent regimen (Fig. 1A,, regimen 1), the cells were exposed to 50 nmol/L gemcitabine for 3 days followed by 7 days in control medium. For the combination of intermittent gemcitabine with continuous bexarotene (Fig. 1A,, regimen 3), the cells were exposed to 50 nmol/L gemcitabine and 1 μmol/L bexarotene for 3 days followed by 1 μmol/L bexarotene for 7 days. Control cells were treated similarly with fresh medium containing 0.1% solvent or 1 μmol/L bexarotene given continuously (Fig. 1A , regimen 2). Cell numbers were monitored by trypan blue exclusion.

In vitro drug sensitivity assay. To determine the sensitivity of parental Calu3 cells and their gemcitabine-resistant variants to gemcitabine, ara-C, doxorubicin, and cisplatin, cells were seeded in 96-well tissue culture plates and treated with each agent at various concentrations for 3 days. Drug effect was measured by WST-1 assay (Roche). Briefly, 10 μL of WST-1 reagent were added to each well and incubated for 2 h at 37°C; the absorbance was measured using a microplate reader (BioTek Instruments) at a wavelength of 420 nm.

RNA isolation and cDNA preparation. Total RNA was isolated from 2 × 10⁶ to 5 × 10⁶ cells using RNeasy Mini kits (Qiagen) according to the manufacturer's instructions. Total RNA was eluted in RNase-free water and stored at −80°C. Total RNA (200 ng) was reverse transcribed into cDNA in a 50 μL reaction volume containing 1× reverse transcription buffer, 5.5 mmol/L MgCl₂, 2 mmol/L deoxynucleotide triphosphates, 2.5 μmol/L random hexamers, 0.4 unit of RNase inhibitor, and 1.25 units of MuLV reverse transcriptase (Applied Biosystems). Human reference cDNA, prepared from commercial RNA pooled from various tissues (Clontech), was used as a standard. All cDNAs were stored at −80°C.

DNA isolation and purification. Total DNA was isolated from 1 × 10⁶ to 5 × 10⁶ cells using PureLink genomic DNA purification kits (Invitrogen) according to the manufacturer's instructions. Genomic DNA was eluted with Tris-HCl (pH 8.5) and stored at −20°C until use.

Quantitative real-time PCR. Real-time PCR assays were done using specific primers and a specific dual-labeled (5′-6FAM/3′-TAMRA) probes for each target. For gene expression analyses, cDNAs corresponding to 10 ng total RNA were analyzed, for relative DNA copy number determinations 10 ng of genomic DNA were used per assay. Reaction volumes were 50 μL containing 1× Universal Taqman buffer (Applied Biosystems), 300 nmol/L of each forward and reverse primer and 100 nmol/L probe. Reactions were carried out in an ABI PRISM 7700 sequence detection system (Applied Biosystems) for 40 cycles of 15 s at 95°C and 60 s at 60°C. RNA expression of the target genes was normalized to the expression level of the housekeeping genes 36b4 or gapdh. DNA copy numbers were normalized by gapdh DNA signal to account for input variation. The following primers and probes were used for RNA analyses: 36b4 (M17885) forward 5′-GCAGATCCGCATGTCCCTT-3′, reverse 5′-TGTTTTCCAGGTGCCCTCG-3′, probe 5′-6FAM-AGGCTGTGGTGCTGATGGCCAAGAAC-TAMRA-3′; gapdh (NM_002046) forward 5′-ACCACAGTCCATGCCATCACT-3′, reverse 5′-CATCACGCCACAGTTTCCC-3′, probe 5′-6FAM-ACCCAGAAGACTGTGGATGGCCCCT-TAMRA-3′; rrm1 (NM_001033) forward 5′-CAGCTTTGGTATGCCATCATTG-3′, reverse 5′-GCTCTTTCGATTACAGGAATCTTTG-3′, probe 5′-6FAM-CTCAGACGGAAACAGGCACCCCG-TAMRA-3′. Expression of dck, cda, ctps, and rrm2 RNA was measured by commercial assays obtained from Applied Biosystems (assay Ids Hs00176127_m1, Hs00156401_m1, Hs00157163_m1, and Hs00357247_g1, respectively). The following primers and probes were used to measure relative gene copy numbers: rrm1 (NC_000011.8, intron 5) forward 5′-AAGAGATTGCCTTATTGTGGATGAT-3′, reverse 5′-GCGTCTAGAAAGAGAGAAACAACAAA-3′, probe 5′-6FAM-TATCCACTGCGTCTTTAAAAACGGTGCTCTG-TAMRA-3′; gapdh (NM_002046) forward 5′-ACCACAGTCCATGCCATCACT-3′, reverse 5′-CATCACGCCACAGTTTCCC-3′, probe 5′-6FAM-ACCCAGAAGACTGTGGATGGCCCCT-TAMRA-3′. The probe for 36b4 was obtained from Integrated DNA Technologies. All other probes were purchased from Applied Biosystems.

Oligonucleotide array hybridization and data analysis. Sample processing and hybridization was done as described in the Affymetrix GeneChip Expression Analysis Manual (Affymetrix). Briefly, double-stranded cDNA was synthesized from 2 μg total RNA using the T7-(dT)24 primer [5′-GGCCAGTGAATTGTAATACGACTCACTATAGGGAGGCGG(T)₂₄-3′]. From the cDNA, biotinylated cRNA was synthesized using the ENZO High Yield BioArray IVT Kit (Affymetrix). Following chemical fragmentation, 10 μg of the fragmented cRNA probe were mixed with a hybridization cocktail containing control oligonucleotide B2 and hybridization controls BioB, BioC, BioD, and cre (Affymetrix), herring sperm DNA (Fisher Scientific), and acetylated bovine serum albumin (Invitrogen). All samples were hybridized to human HG-U133A GeneChip Arrays for 16 h at 45°C with rotation (60 rpm). After washing and staining (using R-phycoerythrin streptavidin with amplification by biotinylated antistreptavidin antibody) on the Affymetrix Microfluidics Workstation according to the manufacturer's instruction (Microfluidics Protocol EukGE-WS2v4), the arrays were scanned on an Agilent scanner at a wavelength of 570 nm. Expression intensity signals for each probe set on each of the microarrays were generated with the Affymetrix MAS5.0 software using the software's global scaling procedure to normalize between arrays. From each of the different Calu3 cultures (parental, gemcitabine-resistant treated with vehicle, bexarotene, gemcitabine, or with the gemcitabine/bexarotene combination for 10 cycles), four replicate RNA samples were prepared and individually subjected to microarray analysis. The expression signals were averaged and transformed into log 2 ratios. Genes not called consistently “present” by the MAS5.0 software in all replicates of at least one of the five conditions were excluded from further analysis. The complete microarray data set has been deposited in National Center for Biotechnology Informations Gene Expression Omnibus¹

Fluorescence in situ hybridization on metaphase chromosomes. Confluent cultures of parental, resistant, or resensitized Calu3 cells were split (1:5) and 24 h after passage these semisynchronized cultures were treated for 2 h with 0.1 μg/mL colcemid for metaphase arrest. Cells were harvested, treated with a hypotonic solution of 0.075 mol/L KCl and fixed in methanol/acetic acid (3:1). The suspensions were dropped onto microscope slides and metaphase spreads were stored overnight at room temperature. Afterward, slides were either processed immediately for hybridization or stored at −20°C. Fluorescence in situ hybridization (FISH) analysis was done using a probe for rrm1 that was generated from the bacterial artificial chromosome clone CTD-2253D11 (Invitrogen) and labeled using the FISH Tag DNA Green Kit (Invitrogen) following the manufacturer's protocol. A chromosome 11 centromeric probe, purchased prelabeled with Direct Red (Qbiogene), was included as control. Metaphase chromosomes were washed twice with 2× SSC and digested with pepsin (100 μg/mL in 10 mmol/L HCl) for 5 min at 37°C. After two more washes with 2× SSC, slides were consecutively treated with PBS/50 mmol/L MgCl₂ (5 min), PBS/50 mmol/L MgCl₂/1% formaldehyde (10 min), and PBS/50 mmol/L MgCl₂ (5 min), followed by ethanol dehydration and air drying. Before hybridization, both probes were combined in Hybrisol VII hybridization buffer (Qbiogene), denatured (96°C for 5 min), and immediately placed on ice. The hybridization mixture was added onto the slides, sealed under a coverslip, and the slides were incubated at 75°C for 5 min. Hybridization was done for 18 h at 37°C followed by washes in 2× SSC/0.1% NP40 (5 min at 37°C) and 0.4× SSC/0.3% NP40 (2 min at 73°C). Because chromosomes were visible due to signal background in the green channel, counterstaining was not necessary. Slides were evaluated using a Leitz DMRB fluorescence microscope using filters for FITC (rrm1) and Texas red (chromosome 11).

In vivo animal studies. For the human xenograft tumor model, parental or gemcitabine-resistant Calu3 cells were harvested in log-phase growth and resuspended in a 1:1 (v/v) mixture of culture medium and Matrigel (BD Biosciences). Tumor cells were implanted s.c. into the right and left axial regions of 6-week-old female athymic nude mice (Harlan) with a 25-gauge needle containing 0.5 × 10⁶ cells in 100 μL. Animals were randomized and treatment began when tumors were palpable (4–5 days after tumor injection). Each group consisted of 10 animals bearing two tumors per animal. Bexarotene was suspended in an aqueous solution containing 10% (v/v) polyethylene glycol (M_r 400) / Tween 80 (99.5:0.5) and 90% of 1% (w/v) carboxymethylcellulose (Sigma Chemical Co.) and dosed p.o. once daily at 100 mg/kg. This dose of bexarotene was previously determined as the maximum tolerated dose, the dose that caused <10% weight loss over the course of the study (15, 22). Gemcitabine was prepared fresh each time in sterile saline and was administered at 100 mg/kg i.p. weekly. Animals receiving no drugs were given vehicle for bexarotene p.o. daily and saline i.p. weekly. Animals receiving bexarotene only were given saline i.p. weekly. Animals receiving gemcitabine only were given vehicle for bexarotene p.o. daily. The treatment continued for 6 weeks. Tumor growth was measured with an electronic caliper (Mitutoyo, Inc.) twice weekly. Tumor volumes were calculated using the formula V = 1/2AB², where A and B represent the longest and the shortest axis of the tumor, respectively. Animal weights were recorded once weekly. The animals used in this study were housed in a U.S. Department of Agriculture–registered facility in accordance with NIH guidelines for the care and use of laboratory animals.

Data analysis. Dose-response curves for growth inhibition were generated and were plotted as a percentage of untreated control. Values for IC₅₀ (the drug concentration needed to produce 50% growth inhibition) were determined by nonlinear least square regression (JMP). Differences in mean values between groups were analyzed by unpaired Student's t test with two-tailed comparison. Multiple comparisons used the one-way ANOVA test with post hoc t test comparison. Differences of P < 0.05 are considered significantly different. Software for statistical analysis was by SigmaStat (SPSS, Inc.).

In vitro growth inhibition by bexarotene and gemcitabine. When used as a single agent, gemcitabine produced a sigmoidal concentration-dependent growth inhibition of Calu3 cells (data not shown). The concentration needed to inhibit 50% of cell growth, IC₅₀, was ∼3 nmol/L. On the other hand, bexarotene showed limited growth inhibitory activity up to 10 μmol/L. Bexarotene did not interfere with or enhance gemcitabine activity in the combination with gemcitabine after single exposure or multiple exposures (Fig. 1B). Furthermore, after repeated treatment for five cycles, gemcitabine activity was neither enhanced nor inhibited by the combination with bexarotene when compared with cells treated with gemcitabine alone (Fig. 1B). Cells treated with continuous bexarotene grew similarly as vehicle-treated controls (Fig. 1B) and continued to do so until the end of the 10-cycle regimen (data not shown). However, when the gemcitabine single-agent regimen (Fig. 1A,, regimen 1) was continued beyond five cycles, the cells lost sensitivity to the agent and the cultures regrew within 100 days (Fig. 1C). In contrast, when gemcitabine was combined with bexarotene (Fig. 1A,, regimen 3), the cultures stayed sensitive and only recovered after treatment was stopped (Fig. 1C).

Characterization of gemcitabine-resistant cells. Surviving cells obtained at the end of treatment from cultures treated with gemcitabine alone, Calu3-GemR, or from cultures treated with the combination regimen were evaluated for their sensitivity to gemcitabine and other cytotoxic agents. The surviving cells isolated from gemcitabine-treated cultures were highly resistant to gemcitabine with a resistance factor of >50. They also showed some cross-resistance to a compound of same class, ara-C, but were sensitive to other cytotoxic agents with different mechanisms of action (Table 1A). On the other hand, cells recovered after treatment with the bexarotene/gemcitabine combination remained chemosensitive toward all compounds tested, including gemcitabine (Table 1A).

To elucidate the mechanism of gemcitabine resistance in the Calu3-GemR cells, real-time PCR was used to analyze mRNA levels for genes known to be involved in gemcitabine resistance: rrm1 (ribonucleotide reductase large subunit 1), rrm2 (ribonucleotide reductase large subunit 2), dck (deoxycytidine kinase), cda (cytidine deaminase), and ctps (CTP synthase). Our data showed that rrm1 mRNA expression was low in Calu3 parental cells, in cells treated with bexarotene alone or with the combination regimen, but was increased ∼20-fold (P < 0.0001) in the gemcitabine-resistant cells (Fig. 2A). In addition, there was a small but significant (P < 0.01) change in the expression of cda (60% decrease) and ctps (60% increase) in the resistant cells when compared with the parental cells. The expression of rrm2 and dck was not significantly different between parental and gemcitabine-resistant cells. When the gene expression profiles of gemcitabine-resistant and parental Calu3 cells were compared by microarray analysis, the p15.5 locus on chromosome 11 showed a large number of genes overexpressed in the resistant cells (Fig. 2B). Included in the center of the overexpression locus was rrm1 (Table 2). These data suggested that increased expression of rrm1 and surrounding genes may be the result of gene amplification. This was confirmed for rrm1 using a DNA-specific quantitative PCR assay. The copy number of the rrm1 gene was ∼12-fold higher (P < 0.0001) in the resistant cells compared with the parental cells (Fig. 2C). In general, gene amplification is observed in two forms: intrachromosomal as extended chromosomal regions or extrachromosomal in the form of episomes or double minutes (23). To distinguish between these mechanisms, metaphase chromosomes of parental and resistant cells were analyzed by FISH (Fig. 2D). Single copies of rrm1 on chromosome 11 (identified by a chromosome-specific centromeric probe) in parental cells could not be detected by FISH, apparently due to insufficient sensitivity of the probe (Fig. 2D,, top). However, in the resistant cells, a very strong rrm1 signal was observed on one of the number 11 chromosomes (Fig. 2D , bottom). These data indicate that the rrm1 gene is amplified intrachromosomally as extended chromosomal region within one of the number 11 chromosomes. No signal indicative of extrachromosomal elements was observed. Taken together, these results showed that intrachromosomal amplification of the rrm1 gene and its concomitant overexpression was most likely responsible for the phenotype in gemcitabine-resistant Calu3-GemR cells.

Effects of bexarotene on gemcitabine-resistant cells. Because addition of bexarotene to the gemcitabine regimen successfully prevented the development of resistance, we investigated whether bexarotene can also overcome preexisting gemcitabine resistance. Calu3-GemR cells were repeatedly treated either with vehicle, with bexarotene or gemcitabine as single agents, or with a combination of both agents in the same regimens used in the prevention study (Fig. 1A). After four treatment cycles, the cultures treated with the combination regimen showed a loss in cell numbers, which continued until the end of the 10th cycle (Fig. 3A). This indicated a resensitization of the Calu3-GemR cells to the cytotoxic activity of gemcitabine. The cultures treated with the single drug regimens expanded as the vehicle controls. However, analysis of RNA and DNA samples collected over the treatment course revealed that, whereas with vehicle and gemcitabine treatment rrm1 gene amplification and expression was still increasing over time, in the cultures treated with bexarotene alone they were gradually reduced, reaching parental levels at the end of the study (Fig. 3B and C). As expected from these data, the Calu3-GemR cells, after treatment with bexarotene alone for 10 cycles, were again sensitive to gemcitabine (Table 1B). In addition, FISH analysis of metaphase chromosomes of the resensitized cells resembled those of the parental cells (data not shown), indicating a complete loss of the rrm1 gene amplification. Comparison of the microarray gene expression profiles of the resensitized cells and the resistant Calu3-GemR cells showed that not only rrm1 but the entire chr11p15.5 amplicon was repressed by exposure to bexarotene (Table 2). In addition, most of the genes outside of this locus, whose expression was altered concomitantly with the acquisition of gemcitabine resistance, were reverted as well (Supplementary Fig. S1 and Table S1). In summary, bexarotene over the course of 10 treatment cycles reversed the phenotype of the gemcitabine-resistant cells almost completely to that of the parental cells, with regard to drug sensitivity, intrachromosomal rrm1 amplification, as well as their molecular profiles.

Effect of bexarotene/gemcitabine combination in vivo. To evaluate the antitumor efficacy of the bexarotene/gemcitabine combination in vivo, Calu3 cells were established as xenograft tumors in athymic nude mice. As seen in Fig. 4A, parental Calu3 tumors grew continuously throughout the course of study in both vehicle-treated control and drug-treated animals. When compared with vehicle control, bexarotene given alone at 100 mg/kg daily had no significant effect on tumor growth after the 4-week treatment period. In contrast, gemcitabine at 100 mg/kg once weekly decreased tumor volume by 63% (P < 0.05 versus control). More importantly, the combination of both agents decreased tumor volumes by 80% relative to gemcitabine alone (P < 0.05) and by 93% relative to vehicle control (P < 0.05; Fig. 4A). To determine whether the combination regimen can overcome gemcitabine resistance in vivo, animals were implanted with the gemcitabine-resistant Calu3 cells (Calu3-GemR) and were treated with single agents or with the combination. Gemcitabine-resistant tumor growth in animals treated with single agents was similar to that of vehicle controls, whereas the combination regimen produced a 38% decrease in tumor growth compared with gemcitabine-treated animals (Fig. 4B; P < 0.05 versus gemcitabine alone). The benefit of the bexarotene/gemcitabine combination became apparent beyond 30 days as the tumor volume in mice receiving the combination therapy began to diverge from the tumor volume in mice treated with gemcitabine alone. Collectively, these results showed that the bexarotene/gemcitabine combination produced a greater antitumor effect than the single agents in the Calu3 xenograft model. Furthermore, the data showed that, as in vitro, inclusion of bexarotene into the treatment regimen could overcome gemcitabine resistance in vivo.

Acquired drug resistance is one of the major obstacles to successful cancer therapy. Whatever the class of the cytotoxic agent and its mechanism of action, most solid tumors eventually become resistant to treatment. Therefore, the inclusion of agents that can interfere with the development of acquired resistance to cytotoxic drugs may provide a substantial clinical benefit. In this study, we show that bexarotene can prevent and reverse resistance to gemcitabine in the Calu3 NSCLC model. Parental Calu3 cells that were treated with repetitive cycles of gemcitabine acquired resistance to the cytotoxic agent within ∼8 weeks of treatment. This was completely prevented when bexarotene was included in the treatment regimen. Parental Calu3 cells and cells recovered after treatment with the combination regimen were ∼50-fold more sensitive to gemcitabine than the Calu3-GemR–resistant cells. This resistance was accompanied by a strong increase in rrm1 expression, one of the two polypeptides that assemble to form the tetrameric RNase reductase enzyme. Other genes that play roles in the activity and metabolism of gemcitabine and have been associated with development of resistance were either unaffected (rrm2, dck) or the effects were too small (cda, ctps) to contribute substantially to the strong increase in resistance [resistant factor (RF), 55]. Therefore, the majority of the resistance can probably be attributed to the strong amplification and overexpression of rrm1. RNase reductase is the key enzyme for the synthesis of deoxyribonucleotides, including dCTP, and has been implicated in gemcitabine resistance in various models (11, 12, 24). Overexpression of this enzyme leads to resistance on two levels, directly through competition of dCTP with the triphosphate of gemcitabine (dFdCTP) for incorporation into the DNA and indirectly through inhibition of the conversion of gemcitabine into the active metabolites. Because RNase reductase itself is also inhibited by one of these metabolites (dFdCDP), resistance through overexpression of the enzyme is potentiated. This activity also accounts for the observed cross-resistance (Table 1A) with ara-C.

Increased expression of rrm1 coincided with a strong increase in rrm1 gene copy number. The rrm1 gene is located in the chr11p15.5 locus. Microarray analysis showed that, in addition to rrm1, the expression of many genes in close vicinity of this gene was increased in the resistant cells (Fig. 2B; Table 2). Although we have not directly determined whether, as for rrm1, this is also due to increased DNA copy number, the high concentration of overexpressed genes together with the measured increase in rrm1 DNA locus suggests a general amplification of this region. This is also in agreement with amplification of the mouse 7E1 locus, as measured by comparative genomic hybridization and observed with gemcitabine resistance in a mouse colon tumor model (25). The mouse 7E1 region corresponds to the human chr11p15.5 locus and also contains rrm1 as well as several other homologues contained in the human chr11p15.5 locus. Other than rrm1 itself, chr11p15.5 genes with the highest induction of expression in Calu3-GemR–resistant cells are several hemoglobins as well as members of the TRIM (tripartite motif) cluster. Based on current knowledge, there is no functional association of these genes with drug resistance, chromosomal instability, growth, or other activities that favor them for selection. Thus, their overexpression is likely due to bystander coamplification with the rrm1 locus. In addition to the genes at the chr11p15.5 locus, microarray analysis showed that several hundred other genes throughout the genome are also significantly higher or lower expressed in the resistant cells when compared with the parental cells (Supplementary Table S1). Some of these genes are involved in DNA replication and repair and one might be tempted to speculate whether these genes might causatively be involved in the development of resistance (or its reversal; see below). However, at this point, it is difficult to distinguish between cause and effect. The observed expression changes for most of these genes are most likely a consequence of the resistance and adaptation to the molecular changes caused by amplification of a large gene set. Further studies are necessary to identify the genes that are the active players and possible mediators of the activity of bexarotene in preventing development of drug resistance.

Treatment of the Calu3-GemR cells with bexarotene resensitizes them to the cytotoxic activity of gemcitabine. This was not only observed in vitro but also in vivo. Xenograft tumors generated from gemcitabine-resistant cells started to respond to the combination treatment after ∼30 days of therapy, whereas no effect on tumor growth was observed with single agent. We have not tested whether the tumors treated with bexarotene as single agent for an extended period regained sensitivity to gemcitabine. However, Calu3-GemR cells treated in vitro with bexarotene alone were as sensitive as parental cells at the end of the treatment period. This was not simply due to lack of selective pressure because it did not occur in cultures treated with vehicle over the same time course. In addition, even in the presence of selective pressure (i.e., in the combination treatment), bexarotene caused resensitization, as indicated by the dramatic cell loss in these cultures. Not surprisingly, due to selective pressure by the cytotoxic agent, the small number of cells remaining alive in the cultures treated with the combination regimen represented the not yet resensitized fraction, which still contained amplified rrm1 (Fig. 3) and showed the resistant phenotype (Table 1B). The resensitization by bexarotene was accompanied by a gradual loss of rrm1 gene copy number and expression over the treatment time course, resulting in a parental rrm1 genotype and phenotype after 3 months of treatment. Furthermore, genes that are colocalized with rrm1 on chr11p15.5 were also normalized.

Gene amplification has been observed either intrachromosomally as extended chromosomal regions or extrachromosomally as episomes or double minutes. Conceptually, reversal of not only rrm1 amplification but of the whole locus could be most easily explained by loss or removal of double minutes containing the amplified genes. Such a mechanism is independent of the amplified genes, does not require any sequence-specific recombination events, and therefore could also account for the broad activities of bexarotene in drug resistance, including resistance to paclitaxel in models of breast and lung cancer, which involve the overexpression and its reversal of the mdr1 multidrug resistance gene (18, 19). Studies have shown that double minutes can be eliminated from cells through a process that involves increased micronuclei formation and that this process can be induced by the differentiating agent hydroxyurea (26–28). However, not all differentiating agents show such activity and therefore differentiation per se is insufficient for the reversal of amplification. Our FISH studies, however, indicate that the rrm1 overexpression is caused by intrachromosomal amplification of the rrm1 locus in one of the number 11 chromosomes (Fig. 2D). At this point, we do not know whether this intrachromosomal amplification had been generated through a multistep mechanism that involves initial formation of episomes (29), in which case extrachromosomal amplification may still play a role in the prevention of drug resistance by bexarotene. However, the reversal of gene amplification in the form of intrachromosomally extended regions would involve more complex processes and no studies have been published that specifically address such mechanisms. Conceptually, the amplified locus or even the whole chromosome carrying the amplification could be replaced by the normal alleles. This would resemble mechanisms described for processes leading to loss of heterozygosity. Studies of RNA-guided sequence-specific DNA rearrangements in Tetrahymena thermophila have suggested a role for targeted DNA deletion in genome surveillance (30) and the observed activity of bexarotene may involve similar processes. Whatever the mechanism, bexarotene could either induce cellular changes that selectively favor those cells that reverse the gene amplification or, alternatively, specifically induce such processes involved in genome surveillance. The latter would also be in concordance with previous data from fluctuation analysis showing that bexarotene can maintain/increase genomic integrity of tumor cells by interfering with the acquisition of spontaneous mutations resulting in drug resistance (18, 19).

As mentioned above, bexarotene also reversed the resistance-associated changes for the vast majority of the genes outside the chr11p15.5 locus (Supplementary Fig. S1). Some of these may be directly or indirectly involved in the reversal or prevention of the DNA amplification. A potential candidate is XRCC4, which is reduced in the resistant cells and normalized by treatment with bexarotene (Supplementary Table S1). This gene is involved in nonhomologous end-joining of DNA, which is an important process in DNA repair and maintenance of genome integrity. We have observed similar effects of bexarotene on XRCC4 in other resistance models.²

In summary, we have shown that bexarotene can prevent and overcome amplification of rrm1 and associated resistance to gemcitabine, in vitro and in vivo. It is likely that the mechanism of action is related for both prevention and reversal of drug resistance, and future research is directed toward understanding of the molecular events involved. Hopefully, these findings can serve as a basis for innovative therapeutic strategies in the treatment of cancer.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Rene Prudente, Tracy Cooke, Manny Corpus, Jen Sanders, Jorge Valencia, and Wen Luo for technical assistance.