Research Article Genetic and Epigenetic Modeling of the Origins of Multidrug-

The origin of drug-resistant cells in human cancers has been a fundamental problem of cancer pharmacology. Two major contrasting hypotheses (genetics versus epigenetics) have been proposed to elucidate the mechanisms of acquired drug resistance. In this study, we answer these fundamental questions through investigation of the genetic and epigenetic pathways that control the origin of ABCB1 (MDR1) gene activation with acquired multidrug resistance in drug-sensitive human sarcoma (MES-SA cells). The genetic and epigenetic bases of this selected activation involve the initiation of transcription at a site 112 kb upstream of the ABCB1 proximal promoter (P1) in the drug-resistant cells. This activation was associated with a chromatin-remodeling process characterized by an increase in acetylated histone H3 within a 968-bp region 5' of the ABCB1 upstream promoter. These alterations provide both genetic and epigenetic susceptibility for ABCB1 expression in drug-resistant cells. Complete activation of the ABCB1 gene through the coding region was proposed by interactions of selected trans-alterations or epigenetic changes on the ABCB1 proximal promoter, which occurred during initial drug exposure. Thus, our data provide evidence for a major genomic alteration that changes the chromatin structure of the ABCB1 upstream promoter via acetylation of histone H3 initiating ABCB1 activation, further elucidating the genetic and epigenetic bases that determine chemotherapeutic response in drug-resistant derivatives of MES-SA cells.


Introduction
Scientists have attempted to determine the origin of drugresistant cells for decades.Two hypotheses have been proposed to account for drug resistance mechanisms in cancer cells.One hypothesis is the occurrence of random mutational events by which a drug-resistant variant arises spontaneously from sensitive cells.Another hypothesis is epigenetic changes (e.g., induction of gene expression via methylation or acetylation) resulting in resistance to cytotoxic insults.Whether drug-resistant cells originate from spontaneous mutations or by induction has been a controversial topic, with, to date, no definitive resolution (1)(2)(3)(4)(5).
Fluctuation analysis provides a powerful genetic tool to differentiate a spontaneous mutational event from an epigenetic alteration (1)(2)(3)(4).The major drawback of this analysis is that conclusions are mainly based on ANOVA and mean of surviving clones.No direct molecular evidence can be obtained.Moreover, application of fluctuation analysis to human cancer cells is a major challenge due to the genetic complexity and heterogeneity of cancer cells compared with that of bacteria, all of which make the interpretation of data more difficult.
To circumvent the above obstacles and model the genetic steps of generating multidrug-resistant (MDR) cells, we have established a highly reliable and reproducible system from the human sarcoma cell line MES-SA to perform fluctuation and molecular analyses (6)(7)(8)(9).Moreover, the MDR phenotype mediated by the ATP-binding cassette transporter ABCB1/P-glycoprotein is also a well-characterized drug-resistant mechanism (reviewed in refs.10,11).Significant progress has been made to delineate the mechanisms underlying ABCB1 expression in normal and tumor cells, particularly in MDR cellular models (reviewed in ref. 12).The unique set of ABCB1 mutants derived from MES-SA cells represent a valuable model system to dissect the genetic and molecular elements that control the origins of multidrug resistance in cancer cells.
In this study, we provide direct genetic and epigenetic evidence for the origin of acquired multidrug resistance from drug-sensitive cancer cells.We have established a molecular/genetic model to dissect genetic and epigenetic steps that are required for ABCB1 activation under stringent drug selection.In this model, we propose that a major genomic alteration provides genetic bases for the initiation of ABCB1 activation (generating a partially activated ABCB1 status) via an epigenetic modification of chromatin structure.Interplay between spontaneous mutations and selected epigenetic alterations contributes to complete activation of the ABCB1 gene.

Materials and Methods
Cellular models.Cell Culture, development, and characterization of the sublines of a human uterine sarcoma (MES-SA) were previously described (Table 1; refs.6,9,13).Cells used in this work include a panel of drugresistant subclones derived by single exposures of MES-SA cells to either doxorubicin or vinblastine or etoposide .
Single-step selection of the MES-SA subline (MFS).Single-step selection of the subclonal MES-SA cells (MFS) with vinblastine has been previously described (9).At that time, MFS cells were also seeded into twelve 96-well plates (1.7 Â 10 6 cells per plate) and six plates treated with either 80 nmol/L doxorubicin or 0.5 Amol/L VP-16 for 2 weeks.These selection conditions were chosen based on the IC 50 of MES-SA cells and the need to minimize the proportion of cells surviving based on stochastic or epigenetic mechanisms.Surviving colonies were allowed to grow for another 2 to 3 weeks and counted.Wells containing single colonies were harvested and expanded in drug-free medium for further studies.
Induction or sublethal induction of ABCB1 expression in MES-SA cells.To determine whether ABCB1/P-glycoprotein could be induced in these sarcoma cells, bulk populations of MES-SA cells were either treated by 80 nmol/L doxorubicn for 24, 48, and 72 hours or treated for 24 hours in escalated doxorubicin concentrations (0.04, 0.08, 0.25, 1, 2, and 4 Amol/L).For sublethal induction, MES-SA cells were seed into two 80-cm 2 flasks, and each was treated with either 80 nmol/L doxorubicin or 0.5 Amol/L VP-16 for 6 days.Then the cells were maintained in drug-free medium for 2 weeks.The surviving cells were harvested and characterized for ABCB1/ P-glycoprotein expression, drug sensitivity, and other phenotypic changes as previously described (6,9).
Flow cytometry.P-glycoprotein expression and function were assessed by two-color flow cytometry using double staining of cells with both the monoclonal antibody UIC2 and rhodamine-123 (Rh-123) as described (6,13).
Southern blotting.Extraction of genomic DNAs and Southern blotting analysis were done according to a previously described method (14).
RNase protection assay.Total RNAs from control cells and MDR variants were hybridized with an antisense RNA probe generated by SP6 RNA polymerase from a 982-bp PstI-PstI genomic fragment of the ABCB1 promoter.The RNase protection assay was done by hybridizing total RNA with 2 Â 10 5 cpm of the antisense RNA probe as described (15).
Rapid amplification of 5VcDNA ends.The rapid amplification of 5V cDNA ends (5V-RACE) kit (Life Technologies/Bethesda Research Laboratories, Gaithersburg, MD) was used with mRNA isolated from 10B-E2 cells using the FastTrack 2.0 mRNA Isolation Kit (Invitrogen, Carlsbad, CA).
Reverse transcription from the isolated mRNAs was then carried out using a gene-specific primer in exon 1 (GSP-1, Table 2).PCR was carried out under the following conditions: 94jC for 4 minutes, 94jC for 1 minute, 55jC for 1.5 minutes, 72jC for 2 minutes (35 cycles); 72jC extension for 10 minutes using a gene-specific primer (GSP-2, Table 2) and the anchor primer from Life Technologies/Bethesda Research Laboratories.Following the initial PCR reaction, a nested PCR was carried out using the reverse primer GSP-3 (Table 2), an ABCB1-specific primer, and an abridged universal anchor primer provided with the kit.The products from nested PCR were subcloned into the pGEM-T vector (Promega Corp., Madison, WI) and sequenced at the Stanford University Sequencing Facility (Stanford, CA).
Conventional PCR.Both total RNAs and genomic DNAs were used for RCR analyses.Total RNA isolation and reverse transcriptase-PCR (RT-PCR) were done essentially as previously reported (6,13).The different batches of RNA samples were assayed by PCR without reverse transcription to rule out genomic DNA contamination (data not shown).The primers for upstream mRNA detection were designed according to both the sequences derived from the 5V -RACE experiments and the published sequences (accession no.AC002457).The EC-94 forward primer is located at the beginning of exon À1 (94 bp) of the ABCB1 gene (Table 2).The ABCB1 +82 reverse primer is located at +82 relative to the transcription start site (+1) of the ABCB1 promoter (Table 2).PCR was carried out as previously described (6,13).The expected size of the PCR product is 370 bp, designated as ut370.
Real-time PCR.Quantitative real-time PCR using SYBR Green dye was done to detect mRNA expression and to verify the presence or absence of genomic sequence in chromatin immunoprecipitation assays in MES-SA and its drug-selected cell lines.The primers were designed according to the published sequences of the BAC clone CTB-60P12 (accession no.AC002457), synthesized in the Stanford University Pan Facility, and summarized in Table 2. Glyceraldehyde-3-phosphate dehydrogenase was used as an internal control for the above assays.The final 20-AL reaction mixture included 300 nmol/L of each primer and 1Â SYBR Universal PCR master mix (Applied Biosystems, Foster City, CA).Thermal cycling was done with 10 minutes at 95jC followed by 40 cycles of PCR amplification: 95jC for 15 seconds and 62jC for 1 minute in the ABI Prism 7900 detection systems (Applied Biosystems).Amplification efficiencies were determined by serial dilutions.All reactions were done in triplicate.Dissociation curves were recorded after each run to confirm the primer specificity.
Chromatin immunoprecipitation.Chromatin immunoprecipitation analysis of the chromatin status of both ABCB1 far upstream and downstream regions was done using the Acetyl-Histone H3 Immunoprecipitation Assay Kit (Upstate Cell Signaling Solutions, Lake Placid, NY) according to the manufacturer's instructions.The detailed experimental procedures using the anti-acetyl-histone H3 antibody, a rabbit polyclonal IgG specific for acetyl-histone H3, are available at the web site of this company (http://www.upstate.com).Briefly, 2 Â 10 6 cells were used for each reaction in this assay.Histones were cross-linked to DNA and sonicated to 200 to 1,000 bp.The size of sonicated DNA was examined by gel electrophoresis after reversion of cross-links.Cell samples were precleared with 80 AL of salmon sperm DNA/protein A agarose/50% slurry for 30 minutes at 4jC before the addition of 10 Ag of a-acetylhistone H3 or normal rabbit IgGs (negative control, Sigma-Aldrich Chemical Co., St. Louis, MO).The protein A/antibody/histone/DNA complex was formed via overnight incubation at 4jC with rotation.Histone-DNA cross-links were reversed by adding 20 AL of 5 mol/L NaCl to 500 AL of the sample and incubated at 65jC for 4 hours.DNAs were recovered by phenol/chloroform extraction and ethanol precipitation and then dissolved in 250 AL of water.Nine microliters of purified DNAs were used for real-time PCR as described above.PCR primers used to amplify both the upstream ABCB1 and the ABCB1 P1 promoter regions were listed in Table 2.
Transient transfection experiments and chloramphenicol acetyltransferase assays.The plasmids used for transient transfection experiments were prepared according to the QIAGEN protocol (Qiagen,  (13).
CAT activity for each sample was normalized to h-galactosidase activity to account for differences in electroporation.

Results
The multidrug-resistant phenotype in doxorubicin singlestep-selected mutants.We previously established doxorubicin single-step-selected mutants (DSMs) via fluctuation analysis, with Abbreviations: ECW and ECP, the regions designated for the ABCB1 upstream primers as indicated in Fig. 5A; Exp., experiments listed in this study; F, forward primer; GSP, gene-specific primer; P1, the ABCB1 proximal promoter; ChIP, chromatin immunoprecipitation. *The cDNA sequences from the genes upstream of ABCB1 were based on the sequences with indicated accession numbers.The positive/minus (+/À) strands are indicated in the amplimers' sequences.cDNA amplimers were designated according to the H. sapiens BAC clone CTB-60P12 from chromosome 7 (accession no.AC002457).The underlined sequences represent restriction sites to facilitate PCR subcloning.bABCB1 coding region primers were based on the sequences from the accession no.M14758.
x GAPDH promoter region primers were obtained from Upstate Cell Signaling Solution.

Cancer Research
Cancer Res 2005; 65: (20).October 15, 2005 prevalent ABCB1 activation in MES-SA cells (6).We selected those DSMs that had stable ABCB1 expression to test their resistance to the selecting agent doxorubicin as well as to vinblastine and VP-16.
In general, the MDR phenotype in DSMs was similar to that which we previously reported (6).Functional expression of P-glycoprotein in the DSM was confirmed by flow cytometry using the UIC2 monoclonal antibody and Rh-123 (Fig. 1A, right).The multiply selected, highly MDR variant Dx5 cells were used as ABCB1positive controls in these studies.The partial reversion exhibited by Dx5 cells in this figure is evidence of multiple genetic steps in the evolution of those cells (Fig. 1A, middle).We also derived new DSM clones (e.g., MFS/DS-5) from an MES-SA subclone (MFS).These subclones displayed a typical MDR phenotype that can be completely reversed to the level of parental MES-SA cells by the P-glycoprotein inhibitor PSC833 (PSC), suggesting that activation of the ABCB1 gene is the predominant resistant mechanism in these drug-resistant MES-SA cells (Fig. 1B).We also isolated the doxorubicin-induced variants via a 6-day doxorubicin sublethal induction.A detailed description of both doxorubicin-selected and doxorubicin-induced variants is provided in Table 1 and in Supplementary Fig. 1.
Molecular characterization of ABCB1 activation in doxorubicin single-step-selected mutants.Molecular characterization of the DSMs was done and is depicted in Fig. 2. As shown in Fig. 2A, the ABCB1 gene was amplified 4-fold in the multistep-selected Dx5 cell line (lane 5), relative to its parental drug-sensitive MES-SA cells as determined by Southern blot.There is also a rearranged fragment on or around the ABCB1 promoter in Dx5 cells.However, neither ABCB1 gene amplification nor gene rearrangement was observed in the DSMs (lanes 2-4).In addition, we found no changes in gene copy numbers in genes upstream of ABCB1 (e.g., RPIB9 and MCFP) by real-time PCR (data not shown).These data suggest that either genomic alterations outside the ABCB1 promoter or transcriptional activation of ABCB1 might be important in these DSMs (see Supplementary notes for Fig. 2A).
We further mapped the transcription start site(s) and showed that the native ABCB1 P1 promoter protects 130-and 134-bp fragments in Dx5 cells (Fig. 2C, lane 5), but, surprisingly, not in those single-step mutants (2B-E3, 3B-B9, and 10B-E2) in which a f324-bp mRNA fragment was protected (Fig. 2C, lanes 2-4), indicating that the ABCB1 gene is predominantly initiated from upstream sites in these single-step mutants.To confirm the initiation site(s) of upstream ABCB1 transcripts in the DSMs (accession no.AF345625), we did 5V -RACE and isolated a 202-bp fragment (designated as RACE202) from a DSM, 10B-E2.Sequencing revealed that this band was homologous to two regions of the Homo sapiens BAC clone CTB-060P12 from 7q21.The mRNAs in the DSMs represent far upstream transcripts containing the previously described exon À1 of ABCB1 (17,18), which is transcribed from a locus that is 112 kb away from the ABCB1 P1 promoter (Fig. 2D).To verify the prevalence of the upstream transcripts in MDR cellular models, we designed specific primers to detect the transcripts containing exon À1.These primers generate a 370-bp RT-PCR product (ut370), which includes exon À1 (94 bp) and exon 1 (276 bp) as shown in Fig. 2D and in Supplementary Fig. 2. Transcripts originating at the upstream site were evident in 71% ( five of seven) of DSMs but not in Dx5 cells (Fig. 2E; Supplementary notes for Fig. 2E).
Identification of the spontaneous nature of ut370 in MES-SA cells.We have previously shown that either doxorubicin or vinblastine selects spontaneous ABCB1 mutants in MES-SA cells (6,9).To test whether activation of the ABCB1 upstream transcripts occur spontaneously in MES-SA sarcoma cells, we examined the presence or absence of the upstream transcripts in bulk MES-SA cells.Our data revealed that the ABCB1 upstream transcripts were detectable only after 40 cycles of PCR amplification in some batches (5 Â 10 6 cells per batch, obtained from different passages) of MES-SA cells with a frequency of two of nine (Fig. 3B, lanes 4-8, four of nine batches of MES-SA cells are shown).To show the spontaneous nature of this upstream transcript (ut370), we isolated a panel of MES-SA subclones by limiting dilution.One clone, termed MES-SA/ FS (MFS), displayed a karyotype similar to its parental MES-SA cells (14), and expressed the ABCB1 upstream transcripts (ut370) detectable by 40 cycles of PCR amplification (Fig. 3B, lane 4).Moreover, the ratio of the ABCB1 upstream transcripts between MFS cells and a vinblastine-resistant clone (Fig. 3B, lane 3, a positive control) was f1 to 32, implying that only a small portion of MFS cells acquired transcripts from upstream promoter during clonal expansion (Fig. 3A, open oval).Neither ABCB1 mRNA expression (determined by 50 cycles of RT-PCR; Fig. 3C) nor P-glycoprotein expression (data not shown) was observed in MFS cells, suggesting the acquisition of the upstream transcript (ut370) does not warrant a fully activated ABCB1 in these cells.
Complete activation of ABCB1 requires drug selection of the upstream ABCB1.To determine whether ABCB1 activation occurs preferentially in the preexisting spontaneous mutants (with ut370), we exposed MFS cells to the MDR substrates vinblastine and doxorubicin (Fig. 3A).The drug-resistant phenotype for vinblastine selection has been fully characterized (9).In general, both MFS/ vinblastine single-step mutants (VSM) and MFS/DSMs displayed a typical MDR phenotype related to ABCB1 activation (Fig. 1; Table 1; ref. 9).As we expected, the MDR phenotype in these cells was associated with expression of ut370 in both the MFS/VSM (eight of eight, Fig. 3D) and MFS/DSM variants ( four of four, with only three clones presented in Fig. 3E).Of note, VP-16, a weaker substrate for P-glycoprotein, failed to induce transcription from the upstream promoter and thus resulted in the selection of non-ABCB1-resistant mechanisms (Fig. 3E, lane 7).The ut370 mRNAs were also confirmed by sequencing analysis (accession nos.AF345624 and AF345623).Neither short-term (Supplementary Fig. 3) nor sublethal doxorubicin induction (Fig. 3F ) induces ABCB1 transcripts originated at the upstream promoter in MES-SA cells.Taken together, these data suggest ABCB1 activation in DSMs occurs as a result of the initiation of transcription at the ABCB1 far upstream promoter, and that additional mechanisms are induced during the course of single-step selection that promote ABCB1 transcription through the native, proximal ABCB1 promoter.
Activated RPIB9 mRNA expression correlated with ABCB1 expression.To verify whether the initiation of upstream transcripts was associated with transcriptional activity of the genes upstream of the ABCB1, we did real-time RT-PCR analysis of mRNA expression for both MCFP (H.sapiens mitochondrial carrier family protein, accession no.AF125531) and RPIB9 (H.sapiens Rap2binding protein 9, accession no.BC022520), which are located Abbreviations: GR, the ABCB1 gene rearrangement; WT, the wild-type ABCB1 genomic structure.B, structure of the PstI-Pst I probe (982 bp) of the ABCB1 P1 promoter used in RNase protection assays and the protected mRNA fragments from exon 1 (Ex1).C, RNase protection assay.The assay was done with parental MES-SA cells (negative controls), Dx5 (positive control), and the three DSMs (2B-E3, 3B-B9, and 10B-E2).Total RNAs were hybridized with the antisense RNA probe.Transcripts starting at the native ABCB1 P1 protect a 130-or 134-bp fragment.mRNAs with start sites 5V to these protect a 324-bp fragment as indicated.D, confirmation of far upstream sequences by 5V-RACE in the DSM.The 202-bp 5'-RACE fragment (RACE202) from the DSM (10B-E2) was isolated by 5V-RACE protocols, which is composed of 70 bp of exon 1a, 94 bp of exon À1, and an additional 38 bp upstream of exon À1 of the ABCB1 gene.Abbreviations: AUAP, abridged universal anchor primer; GSP, gene-specific primer.E, RT-PCR analysis of ut370 in DSMs.MES-SA cells (lanes 1-2) were negative controls.Lanes 4-11, DSMs.PCR cycle numbers (x ) are indicated.
upstream of exon À1 of ABCB1 (Fig. 4A).The MCFP gene is transcribed from a region that is f123 kb upstream of the upstream transcripts, whereas the RPIB9 gene is transcribed from the DNA strand opposite to the upstream transcripts and ABCB1.As shown in Fig. 4B, MCFP mRNA levels were increased (by 1.3-to 2.4-fold) in several drug-selected cell lines except 3B-B9.The expression pattern of MCFP was correlated with ABCB1 expression in the DSMs (10B-E2, 2B-E3, 3B-B9, and MFS/DS5) but not in multistep-selected Dx5 cells (Fig. 4B1 and B2).Furthermore, the expression pattern of RPIB9 was tightly correlated with ABCB1 expression, in that the mRNA ratios of RPIB9 expression in the four DSMs (10B-E2, 2B-E3, 3B-B9, and MFS/DS5) are 1:4.6:1.3:3,values that were consistent with those of ABCB1 (1:4.7:2.0:3.4;Fig. 4B).However, Dx5 cells did not express RPIB9, also consistent with the observation that Dx5 cells lack the upstream transcript (ut370).These data suggest that regional gene transcriptional activity may be associated with ABCB1 expression by altering regional chromatin structures that affect ABCB1 upstream transcription.
Chromatin remodeling by acetylated H3.To determine whether remodeling of chromatin structures by posttranslational modification of the structure of nucleosomes plays a role in activating the ABCB1 far upstream promoter region, we examined the acetylation status of the core histone protein H3 by chromatin immunoprecipitation assays using an anti-acetylated H3 antibody in MES-SA, two DSMs (2B-E3 and 10B-E2), and Dx5 cells.Acetylation of H3 histones within the 968 bp upstream of exon À1 was markedly increased by 18-and 25-fold, respectively, in 10B-E2 and 2B-E3, but only 1.7-fold in Dx5 cells compared with parental MES-SA cells (Fig. 5D3) as determined by at least five sets of primers that cover this area (described in Table 2 and partially listed in Fig. 5A).Modification of histone H3 was not evident in the 622-bp region upstream of the acetylated 968-bp region (Fig. 5D4).A reciprocal pattern of acetylation of histone H3 was observed in the ABCB1 P1 promoter region, in which Dx5 displayed a 35-fold increase in acetylated H3 as determined by real-time PCR using the five primer sets that cover the ABCB1 promoter region (Table 2,   5D).Thus, these data confirmed that ABCB1 activation in DSMs and Dx5 cells occurs at distinct sites in the 7q21 chromosomal region, although they may share some similar mechanisms such as acetylation of histone H3.However, initiation of ABCB1 transcription at the downstream promoter is controlled by an initiator sequence and transcriptional factors on the ABCB1 promoter ( 12), whereas the 320-bp region upstream of the exon À1 lacks typical promoter elements such as a TATA box and also lacks initiator sequences and DNA binding consensus for transcriptional factors (analysis not shown).Thus, how transcription initiation is mediated by RNA polymerase II in the ABCB1 upstream region remains obscure.
Functional analysis of the ABCB1 far upstream region and the ABCB1 P1 promoter.To determine whether the ABCB1 upstream transcripts represented a simple activation of an upstream promoter (P2), we used the putative promoter (P2)-CAT construct (ECP2-370) containing the 370-bp sequences upstream of exon À1 of ABCB1 to examine promoter activity in both MES-SA and its MDR variants (e.g., 10B-E2) that express the upstream transcripts (ut370).CAT activity was also compared with both the SV40 promoter and the ABCB1 P1 promoter (DC-1018).As shown in Fig. 6A, the electroporation experiment did not reveal significant basal promoter activity (f1.3-fold related to the vector control) with the ECP2-370 construct (Fig. 6A, columns 1-2) in MES-SA cells compared with both the SV40 and the ABCB1 promoter which displayed 6.3-and 8.4-fold CAT activity when compared with the vector control (Fig. 6A, columns 1, 3, and 4).There was only a 1.5-fold increase in CAT activity in the ECP2-370 vector transiently transfected into the DSM 10B-E2 (Fig. 6B, columns 2 and 4).Moreover, we also observed similar results in 2B-E3 and two MFS/VSM variants (MFS/VL20-3.1 and MFS/VL20-6.1;data not shown).These data indicate that the upstream region in MES-SA cells lacks basal promoter activity in contrast with the ABCB1 P1 promoter that has a strong basal transcriptional activity in MES-SA cells.
We further examined ABCB1 P1 promoter activity in 10B-E2 cells maintained in drug-free medium and only found a 1.5-fold increase in ABCB1 P1 promoter activity in 10B-E2 cells compared with MES-SA control (data not shown).Furthermore, we evaluated the ABCB1 basal promoter in parental MES-SA cells (Fig. 6C, lanes 1-2) and in drug-selected/induced sublines (Fig. 6C, lanes 3-10) by transiently introducing the ABCB1 reporter plasmid into these cells.Basal and drug-selected/induced ABCB1 promoter activity in the plasmid reflects the reactivity of transcriptional factors that act on the ABCB1 reporter construct and thus mimics the open chromatin of the ABCB1 promoter but not that of the chromatinembedded ABCB1 promoter.This basal promoter activity could be significantly modulated under drug selection conditions in sister clones (e.g., 10B-C5) as determined by quantitative analysis [ 14 C]labeled acetyl chloramphenicol species (Fig. 6C-D).This revealed 5-to 12-fold increases in ABCB1 P1 promoter activity in the DSMs, comparable with that in Dx5 (Fig. 6D), suggesting an interaction of either transcriptional factors or other epigenetic mechanisms with the ABCB1 P1 promoter during drug selection.

Discussion
The model for the activation of ABCB1 in MES-SA cells, based on our experimental data, is illustrated in Fig. 7.In this model, the single-step selection favors the isolation of clones with a spontaneous alteration that generates ABCB1 upstream transcripts (a partially activated ABCB1 status).It seems that an initial event contributes to an altered epigenetic mechanism (i.e., a major increase in acetylated H3 that modifies the chromatin structure of the ABCB1 far upstream region and thus initiates ABCB1 upstream transcripts).In MES-SA cells, MDR clones arise as a result of altered chromatin remodeling in the region of the upstream promoter.
Acquisition of spontaneous mutants as a discrete genetic step for ABCB1 activation.We previously hypothesized that a major genomic instability (with a mutation rate of 1.8 Â 10 À6 per cell per generation) leads to ABCB1 activation in DSM cells, based on fluctuation analyses (6).In this study, we were able to provide direct evidence supporting this hypothesis (Fig. 3).ABCB1 upstream transcripts are evident after 40 cycles of RT-PCR amplification in some bulk MES-SA cells and in the MFS clonal variant of the MES-SA cell line (Fig. 3B).MFS cells were derived by expanding one single cell to several million cells without any drug selection thus providing experimental evidence that the upstream ABCB1 transcripts are spontaneous in nature, as defined by a spontaneous mutation rate (6,9).However, the exact change that is responsible for the generation of the ABCB1 upstream transcript is unknown.It might be associated with genomic instability of ABCB1 upstream DNA sequences, which in turn affects the local chromatin structure around exon À1 or chromosomal alterations outside chromosome 7.These changes could result in expression of altered transcription factors that regulate the chromatinremodeling processes or modify the histone core of nucleosomes in the region around the ABCB1 upstream promoter (Fig. 7B).
A mutational event contributes to epigenetic modification of the histone core.We have shown that acetylated H3 was dramatically increased in the single-step mutants in a 968-bp region upstream of the promoter region, where ABCB1 upstream transcripts are generated.In eukaryotes, the basic unit of chromatin structures is assembled by wrapping f150 bp of DNA around histone octamers that are composed of four histone heterodimers (two each of H3-H4 and H2A-H2B).It has been well established in many cellular models that both deacetylation and acetylation of histones play a central role in gene regulation (12,(19)(20)(21).Our data suggest that an increase in acetylated H3 in the nucleosomes within the 968-bp upstream region proximal to the ABCB1 upstream promoter is one important step to open the chromatin structure, which in turn provides the accessibility for basal transcriptional machinery to act on exposed DNA and thus initiates gene transcription.The initiation of upstream transcripts may also depend on the acetylated region that interacts with local chromosomal/chromatin structures and distal enhancer elements.Whether a gain-of-function mutation of histone acetyltransferases, a loss-of-function mutation of histone deacetylases, or another alteration is the initial driving force to initiate transcription from the upstream promoter is unknown.
Upstream transcriptional activity correlates with ABCB1 mRNA expression.It has been shown that ABCB1 activation can be achieved by capturing a hybrid ABCB1 mRNA.In these cases, the hybrid partner usually originated from a region far upstream of exon À1 of the ABCB1 gene in chromosome 7 or from another chromosome following a rearrangement (15,22).We found an association of both MCFP and RPIB9 expression with ABCB1 expression in our MDR cellular models (Fig. 4B).The functions of both MCFP and RPIB9 are largely unknown (23).The tight correlation of RPIB9 and ABCB1 gene expression pattern in the DSMs, but not the multistep-selected Dx5 cells that lack the ut370 upstream transcripts (Fig. 4C), suggests that transcription of ABCB1 from the upstream promoter and transcription of the RPIB9 B, schema of the deduced chromatin structure around exon À1 of the ABCB1 gene.C, genomic DNA sequences of the 968-bp region upstream of exon À1 with partial primer sequences indicated for real-time PCR.D, chromatin immunoprecipitation and real-time PCR analysis of the precipitated genomic DNA sequences.D1, input control.Sonicated genomic DNAs without the chromatin immunoprecipitation steps and amplified with the glyceraldehyde-3-phosphate dehydrogenase (GAPDH ) promoter primer as listed in Table 2. D2-D5, chromatin immunoprecipitation assays with the anti-acetylated H3 antibody (aAH3) and real-time PCR analysis using ECP2, ECW, and ABCB1 P1 promoter primers (Table 2), respectively.D6, chromatin immunoprecipitation assays with rabbit IgG (control) and real-time PCR analysis using the ECP-2 primer.The result was normalized to the GADPH promoter control in each reaction.One of three independent experiments with triplicate is shown.
gene on the same chromosome 7 are initiated concurrently.The increased transcriptional activities at 7q21 might be result from alterations of regional chromatin structures within the 112-kb intronic sequences, upstream of the ABCB1 gene, that promote transcription through barriers in this region.It might also provide enhancer elements to facilitate ABCB1 transcription from the upstream promoter.Further study is needed to verify these possibilities.C, spontaneous mutations promote the acetylation of histone H3.D, the initiation of the upstream transcript containing exon À1 associated with regional transcriptional activity (e.g., RPIB9 activation).E, partially activated ABCB1 status with the transcript (containing ut370) terminated on the ABCB1 promoter, which is likely suppressed by the packed chromatin of the ABCB1 promoter.F, complete activation of ABCB1 may be achieved by either altered transcriptional factors (such as Sp1, AP1, NF-Y, YB-1, and C/EBPh; refs.12, 16) that interact with the ABCB1 promoter or epigenetic changes of the ABCB1 chromatin.These transcriptional and epigenetic changes, which may be acquired during the course of drug selection, permit transcriptional activation of the endogenous and chromatin-embedded ABCB1 promoter.
Interplay between genetic and epigenetic mechanisms at the ABCB1 promoter.Unselected MES-SA cells that contain transcripts originating at the upstream ABCB1 promoter lack ABCB1 mRNAs encompassing the coding regions, suggesting that the ABCB1 gene is not fully activated without drug selection (Fig. 3B-C and Fig. 7E).Thus, it seems that in some cases, the selection of preexisting mutations or epigenetic changes by chemotherapeutic agents (such as vinblastine, doxorubicin, or paclitaxel) is needed to give rise to a fully activated ABCB1 in these spontaneous mutants (Fig. 3D-E and Fig. 7F).Indeed, ABCB1 promoter activity was increased in the DSMs that are maintained in drug selection conditions, and this was significantly modulated in the presence of higher doxorubicin concentrations (e.g., 0.5 Amol/L doxorubicin; Fig. 6C-D).This suggests that additional trans-alterations (mediated by doxorubicin) act on the open chromatin structure of the ABCB1 promoter in the DSM cells to facilitate transcription from the ABCB1 promoter thus achieving complete activation of ABCB1.Trans-alterations that are involved in the above processes might include stable or transient expression or down-regulation of a panel of transcriptional factors.The stability of the MDR phenotype of the DSMs cultured in drug-free medium rules out transient expression (induction) of transcriptional as a major mechanism for ABCB1 expression in these cells (Table 1).The nature of the heritable changes that result in these stable MDR variants is not known but may include stable expression of transcription factors interacting with the ABCB1 promoter, drug-induced acetylation of other histones, and demethylation of the ABCB1 promoter.
In summary, our data suggest that ABCB1 expression in these single-step mutants arose from a discrete genetic step contributing to epigenetic modification of the histone core H3-initiating transcription from the ABCB1 upstream promoter.Transcriptional activity of the genes upstream of ABCB1 may be associated with ABCB1 mRNA expression.We conclude that interplay between genetic and epigenetic mechanisms at the putative ABCB1 upstream promoter and in the region proximal to this contributes to complete activation of ABCB1 in some cancer cells.

Figure 2 .
Figure 2. Molecular characterization of the DSMs.A, Southern blot.Fifty micrograms of genomic DNAs from MES-SA, its DSMs, and Dx5 cells were digested by HindIII, 1% agarose gel-fractionated, transferred onto a nylon membrane, and hybridized with the Pst I-Pst I fragment of the ABCB1 promoter labeled by [a-32 P]-dCTP as referenced in Materials and Methods.Abbreviations: GR, the ABCB1 gene rearrangement; WT, the wild-type ABCB1 genomic structure.B, structure of the PstI-Pst I probe (982 bp) of the ABCB1 P1 promoter used in RNase protection assays and the protected mRNA fragments from exon 1 (Ex1).C, RNase protection assay.The assay was done with parental MES-SA cells (negative controls), Dx5 (positive control), and the three DSMs (2B-E3, 3B-B9, and 10B-E2).Total RNAs were hybridized with the antisense RNA probe.Transcripts starting at the native ABCB1 P1 protect a 130-or 134-bp fragment.mRNAs with start sites 5V to these protect a 324-bp fragment as indicated.D, confirmation of far upstream sequences by 5V-RACE in the DSM.The 202-bp 5'-RACE fragment (RACE202) from the DSM (10B-E2) was isolated by 5V-RACE protocols, which is composed of 70 bp of exon 1a, 94 bp of exon À1, and an additional 38 bp upstream of exon À1 of the ABCB1 gene.Abbreviations: AUAP, abridged universal anchor primer; GSP, gene-specific primer.E, RT-PCR analysis of ut370 in DSMs.MES-SA cells (lanes 1-2) were negative controls.Lanes 4-11, DSMs.PCR cycle numbers (x ) are indicated.

Figure 3 .
Figure 3. Selection of spontaneous mutants with the ABCB1 upstream transcripts.A, schematic presentation of the isolation of the MES-SA subclone and single-step selection or sublethal selection by multiple anticancer drugs.The subclonal cell line MFS, generated from MES-SA cells by the limiting dilution method, was used for fluctuation analysis and single-step-selection by vinblastine(9), doxorubicin (in this study), and VP-16 (in this study).Open oval, spontaneous mutant cells with ut370 arisen from MFS cells.Open oval with bolded outlines, clones with both the ABCB1 upstream and downstream transcripts.Abbreviations: DOX, doxorubicin; VBL, vinblastine; VP, VP-16.B, RT-PCR analysis of ut370 in bulk MES-SA cells and in the MES-SA subclone (MFS).Different batches of MES-SA cells (four of nine, lanes 5-8) were included.C, analysis of expression of ABCB1 mRNA at coding regions was done by using primers spanning transmembrane 6 of the ABCB1 gene(9).The designated products were tm325.D, analysis of both tm325 (top ) and ut370 (middle ) in VSMs derived from MFS cells (lanes 1-11 ).Parental MES-SA cells (lanes 12-13 ) were used as negative controls.E, RT-PCR analysis of ut370 was done in the DSMs derived from MFS cells (lanes 3-5 ).MFS/VL20-4.1 (obtained by selection with 20 nmol/L vinblastine) and MFS/VP-S7 (obtained by selection with 0.5 Amol/L VP-16) were used as an ABCB1 -positive and ABCB1 -negative control, respectively, in this experiment.F, RT-PCR analysis of both ut370 and tm325 expression was carried out in the variant (MES-SA/DOX6D or DOX6D) derived from doxorubicin (80 nmol/L) sublethal exposure for 6 days.MES-SA and MFS/VL20-4.1 cells were used as a negative and positive control for both tm325 and ut370, respectively.rRNA was used as the control gene to normalize sample loading.The samples were analyzed by a 2% agarose gel and stained with ethidium bromide.

Fig.
Fig.5D).Thus, these data confirmed that ABCB1 activation in DSMs and Dx5 cells occurs at distinct sites in the 7q21 chromosomal region, although they may share some similar mechanisms such as acetylation of histone H3.However, initiation of ABCB1 transcription at the downstream promoter is controlled by an initiator sequence and transcriptional factors on the ABCB1 promoter (12), whereas the 320-bp region upstream of the exon À1 lacks typical promoter elements such as a TATA box and also lacks initiator sequences and DNA binding consensus for transcriptional factors (analysis not shown).Thus, how transcription initiation is mediated by RNA polymerase II in the ABCB1 upstream region remains obscure.Functional analysis of the ABCB1 far upstream region and the ABCB1 P1 promoter.To determine whether the ABCB1 upstream transcripts represented a simple activation of an upstream promoter (P2), we used the putative promoter (P2)-CAT construct (ECP2-370) containing the 370-bp sequences upstream of exon À1 of ABCB1 to examine promoter activity in both MES-SA and its MDR variants (e.g., 10B-E2) that express the upstream transcripts (ut370).CAT activity was also compared with both the SV40 promoter and the ABCB1 P1 promoter (DC-1018).As shown in Fig.6A, the electroporation experiment did not reveal significant basal promoter activity (f1.3-fold related to the vector control) with the ECP2-370 construct (Fig.6A, columns 1-2) in MES-SA cells compared with both the SV40 and the ABCB1 promoter which displayed 6.3-and 8.4-fold CAT activity when compared with the vector control (Fig.6A, columns 1, 3, and 4).There was only a 1.5-fold increase in CAT activity in the ECP2-370 vector transiently transfected into the DSM 10B-E2 (Fig.6B, columns 2 and 4).Moreover, we also observed similar results in 2B-E3 and two MFS/VSM variants (MFS/VL20-3.1 and MFS/VL20-6.1;data not shown).These data indicate that the upstream region in MES-SA cells lacks basal promoter activity in contrast with the ABCB1 P1 promoter that has a strong basal transcriptional activity in MES-SA cells.We further examined ABCB1 P1 promoter activity in 10B-E2 cells maintained in drug-free medium and only found a 1.5-fold increase in ABCB1 P1 promoter activity in 10B-E2 cells compared with MES-SA control (data not shown).Furthermore, we evaluated the

Figure 4 .
Figure 4. Real-time PCR analysis of mRNA expression of the genes (MCFP and RPIB9 ) upstream of ABCB1 .A, a physical map of ABCB1 and its upstream genes.RPIB9 is located on the opposite DNA strand of ABCB1 , in which RPIB9 partially overlaps with the far upstream region of ABCB1 .The MCFP gene is localized at a region that is f123 kb upstream of exon À1 of the ABCB1 gene.B, relative ABCB1, MFCP , and RPIB9 mRNA expression in MES-SA and its MDR derivatives.The mRNA expression was normalized to GAPDH expression.Columns (B1-B3 ), means of relative RT-PCR products from three independent amplifications; bars, SD.One of two independent experiments done in triplicate is shown.

Figure 5 .
Figure 5. Chromatin immunoprecipitation assays of chromatin status with the anti-acetylated H3 antibody.A, a physical map of both the ABCB1 upstream and downstream regions.The primer locations were labeled, with increased acetylated H3 in nucleosomes relative to MES-SA control ( * ) and unacetylated H3 ( y ) indicated.

Figure 6 .
Figure 6.Analysis of CAT activity of the upstream region of exon À1 and the ABCB1 P1 promoter in MES-SA and its drug-resistant derivatives.A, CAT ELISA assays of basal promoter activity of the upstream region (around exon À1, ECP2-370 or ECP2), the ABCB1 P1 promoter (DC-1018), and the SV40 promoter (pCAT/SV40, SV40).B, CAT ELISA assays of drug-selected or drug-induced promoter activity of the upstream region (around exon À1), the ABCB1 P1 promoter, and the SV40 promoter in the DSM 10B-E2.C, analysis of ABCB1 P1 promoter activity by transient transfection of pBLCAT3 (control = 1) and the ABCB1 -CAT vector (DC-1018 = 2) into the DSMs (DxS10 and 10B-C5), 0.5 Amol/L doxorubicin-induced 10B-C5 cells (D10B-C5), and Dx5.MES-SA cells were used as control in this experiment.The pSV-h-galactosidase plasmid was used an internal control for electroporation efficiency in these CAT assays.D, relative CAT activity of (C ) was determined by the [ 14 C] scintillation counts of acetylated chloramphenicols (A-CM ) and normalized to h-galactosidase expression levels.At least one of three independent experiments is shown.

Figure 7 .
Figure 7. Schematic model of the origins of MDR cells in cancer.The genetic and epigenetic steps (B-F ) that are required to develop MDR cells in a MES-SA sarcoma are diagrammed.A, ABCB1 genomic organization at 7q21 with ABCB1 far upstream (around exon À1) and the ABCB1 promoter (Ex1a and Ex1b), B, a nucleosomal representation of the ABCB1 genomic organization corresponding to (A ) with both the genomic instability region around exon À1 and the chromatin-embedded ABCB1 promoter indicated.C, spontaneous mutations promote the acetylation of histone H3.D, the initiation of the upstream transcript containing exon À1 associated with regional transcriptional activity (e.g., RPIB9 activation).E, partially activated ABCB1 status with the transcript (containing ut370) terminated on the ABCB1 promoter, which is likely suppressed by the packed chromatin of the ABCB1 promoter.F, complete activation of ABCB1 may be achieved by either altered transcriptional factors (such as Sp1, AP1, NF-Y, YB-1, and C/EBPh; refs.12, 16) that interact with the ABCB1 promoter or epigenetic changes of the ABCB1 chromatin.These transcriptional and epigenetic changes, which may be acquired during the course of drug selection, permit transcriptional activation of the endogenous and chromatin-embedded ABCB1 promoter.

Table 1 .
Drug sensitivity phenotype of MES-SA drugresistant variants (fold resistance) (6)14)clone of MES-SA cells was isolated by limiting dilution method(9,14).cDSMs of MES-SA cells were isolated via fluctuation analysis(6).Three representative clones of 13 populations are shown.bVSMswere derived from the MES-SA subclone cells (MFS; ref. 9).A representative clone of a population of eight is shown.xMFS/DSM (80 nmol/L) cells were generated in this study.One representative clone of each selection is shown.k MES-SA cells were undergone with sublethal induction/selection by doxorubicin (40 nmol/L).{ MES-SA cells were undergone with sublethal induction/selection by VP-16 (0.5 Amol/L).**Multistep doxorubicin-selected MDR cells from MES-SA cells

Table 2 .
Oligonucleotides used in PCR