Endocrine-Related Cancer (2008) 15 439–449
Patterns and changes in gene expression following neo-adjuvant anti-estrogen treatment in estrogen receptor-positive breast cancer Vera Cappelletti1*, Manuela Gariboldi1,2*, Loris De Cecco1,2, Sara Toffanin1, James F Reid1,2, Lara Lusa1,2, Emilio Bajetta3, Luigi Celio3, Marco Greco4, Alessandra Fabbri5, Marco A Pierotti 2,6 and Maria Grazia Daidone1 1
Department of Experimental Oncology, Fondazione IRCCS Istituto Nazionale dei Tumori, Via Venezian 1, 20133 Milano, Italy Molecular Cancer Genetics, Fondazione Istituto FIRC di Oncologia Molecolare (IFOM), Milan, Italy Departments of Medical Oncology 4Surgery, 5Pathology and 6Scientific Direction, Fondazione IRCCS Istituto Nazionale dei Tumori, Via Venezian 1, 20133 Milano, Italy 2 3
(Correspondence should be addressed to M G Daidone, Department of Experimental Oncology, Fondazione IRCCS Istituto Nazionale dei Tumori, Via Venezian 1, 20133 Milano, Italy; Email:
[email protected]) *V Cappelletti and M Gariboldi contributed equally to this work
Abstract This study aimed to define a gene expression profile associated with response to anti-estrogen treatment in estrogen receptor a (ERa)-positive breast cancer from elderly patients and to identify possible candidate genes associated with resistance by detecting those modulated by treatment. Using cDNA microarrays containing 16 702 unique clones, 21 pre-treatment and 11 paired posttreatment samples collected in a neo-adjuvant toremifene trial on elderly patients with operable and locally advanced ERa-positive breast cancer were profiled. Gene expression profiles generated from pre-treatment samples were correlated with treatment-induced tumor shrinkage and compared with those obtained from post-treatment paired samples to define genes differentially modulated following anti-estrogen treatment. Correlation analysis on 21 pre-treatment samples highlighted 53 genes significantly related to treatment response (P!0.001). Genes involved in cell cycle and proliferation were more frequently upregulated in responders compared with nonresponders. Class comparison analysis identified 101 genes significantly modulated independently of treatment response; 82 genes were modulated in non-responders, whereas only 8 genes were differently expressed after treatment in responders. Gene expression profiles appear to be more frequently modulated by anti-estrogen treatment in non-responding patients and may harbor interesting genes possibly involved in anti-estrogen resistance, including clusterin, MAPK6, and MMP2. This concept was corroborated by in vitro studies showing that silencing of CLU restored toremifene sensitivity in the ER anti-estrogen-resistant breast cancer cell line T47D. Integration between neo-adjuvant therapy and transcriptional profiling has therefore the potential to identify therapeutic targets to be challenged for overcoming treatment resistance. Endocrine-Related Cancer (2008) 15 439–449
Introduction The use of hormonal therapy has significantly contributed to the general reduction in breast cancer mortality achieved in the last decade (Peto et al. 2000). Classical anti-estrogen therapies act by blocking or downregulating the estrogen receptor (ER) and represent the first example of biologically targeted
therapy in this tumor type. However, ER transcriptional effects are not simply determined by the ligand, and ER activity is regulated through complex interactions with multiple cell signaling pathways (Razandi et al. 2003, Stoica et al. 2003). Neo-adjuvant trials represent a tool for investigating biological markers associated with treatment response
Endocrine-Related Cancer (2008) 15 439–449 1351–0088/08/015–439 q 2008 Society for Endocrinology Printed in Great Britain
DOI: 10.1677/ERC-07-0274 Online version via http://www.endocrinology-journals.org
V Cappelletti, M Gariboldi et al.: Gene expression and anti-estrogen treatment as well as those modulated by the treatment, thus giving clues for understanding mechanisms involved in clinical resistance. The combination of this approach with genomic expression profiling could provide an opportunity to identify molecular signatures associated and/or predictive of treatment response. Preliminary data eliciting molecular signatures associated with clinical outcome following endocrine treatment are already available in the metastatic and adjuvant settings (Ma et al. 2004, Paik et al. 2004, Jansen et al. 2005). We exploited a neo-adjuvant trial recently concluded in our institution, originally planned to evaluate the predictive role of ER-b on response to the anti-estrogen toremifene (Cappelletti et al. 2004), to investigate the molecular signatures associated with tumor shrinkage on pre-treatment samples and to identify changes in gene expression pattern in post-treatment samples compared with pretreatment one. We report here a gene signature associated with response to treatment and a distinct set of genes modulated by treatment, which might be involved in anti-estrogen resistance mechanisms. The latter include genes already known to be involved in mechanisms of resistance to anti-estrogens, such as MPK6 and MMP2, and a new candidate CLU (clusterin). A validation in an experimental model is reported for one of the candidate genes, CLU. We demonstrated that toremifene treatment upregulated CLU expression and that silencing of CLU restored sensitivity to the anti-estrogen in an ERCantiestrogen-resistant breast cancer cell line. This suggests that the integration between neoadjuvant therapy and transcriptional profiling has the potential to identify therapeutic targets to be challenged to overcome treatment resistance.
Patients and methods Patients, tissue samples, and study design
This study employed leftover samples obtained from a prospective study (Cappelletti et al. 2004) in which women aged 65 years or older with operable or locally advanced ER-positive (ERC) breast cancer but previously untreated were subjected at Istituto Nazionale Tumori of Milan to pre-operative toremifene (Fareston, Shering Plough, Bulkham, Australia) at a dose of 60 mg once daily for 3 months and then underwent surgery. Pre-treatment tumor material was obtained by three to four core biopsies with a 14 gauge 440
needle. Post-treatment tissue was sampled by the pathologist within the residual tumor at surgery. For molecular analyses, core biopsies and surgical specimens were snap-frozen in liquid nitrogen within 30 min of removal, and stored at K80 8C; gene profile was carried out on samples containing more than 70% neoplastic cells, as assessed by hematoxylin and eosinstained sections from frozen tissue specimens. Clinical response was evaluated as percentage of residual tumor size from baseline after 3 months of treatment (Cappelletti et al. 2004). Patients were assigned to responder or non-responder groups according to WHO criteria except that patients with residual tumor !75 and R50% were classified as achieving minor changes and considered as responders (Geisler et al. 2001), according to previously reported criteria to evaluate clinical response to neoadjuvant endocrine treatment. We also investigated changes of gene expression pattern in the absence of any intercurrent systemic treatment on a series of seven paired bioptic and surgical specimens from patients subjected to diagnostic biopsy followed within 3–4 weeks by radical surgery without intervening therapies. The study has been approved by the IRB and Ethics Committee, and patients gave written informed consent to donate the leftover tissue after diagnosis to the Istituto Nazionale Tumori of Milan for the present and future research. RNA isolation and expression profiling
Total RNA extraction from core biopsies and surgical samples, probe labeling and hybridization were performed as described previously (De Cecco et al. 2004). Two types of slides were used: type 7 star slides (Amersham Biosciences) for the toremifene-treated cases and UltraGAPS (Corning, Lowell, MA, USA) for the control dataset. The samples and a reference RNA (Universal Human Reference RNA, Stratagene, La Jolla, CA, USA) were labeled directly with Cy3-dCTP (reference RNA) or Cy5-dCTP (sample RNA; Amersham Biosciences) and indirectly with 3DNA Submicro Expression Array Detection kit (Genisphere, Montvala, NJ, USA). Hybridizations were carried out in a hybridization station (Genomic Solutions, Ann Arbor, MI, USA), slides were scanned using the GenePix 4000B microarray scanner and quantified using GenePix Pro 5.0.1.24 (Axon Instruments, Molecular Devices, Sunnyvale, CA, USA). The RNAs were hybridized on two different cDNA microarrays containing a total of 16 702 unique clones selected from the Human sequence verified IMAGE www.endocrinology-journals.org
Endocrine-Related Cancer (2008) 15 439–449 clone collection (Research Genetics/Invitrogen) and spotted in triplicate. Quantitative real-time PCR
Total RNA isolated for the microarray analysis was used to verify the quantity of specific messengers by real-time PCR for 11 biologically or clinically relevant differentially expressed genes. RNA was reverse transcribed using the High-Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA, USA). The samples were amplified in multiplex PCRs using one of the assays of interest labeled with FAM and the housekeeping gene GADPH labeled with VIC. Data analysis was done using the Sequence Detector version 1.9 software (Applied Biosystems, Foster City, CA, USA) and statistical analyses were performed using the R-statistical computing programing language (R development Core Team 2006). The same statistical tests employed in the gene expression experiment were used to confirm the results with quantitative real-time PCR. Relative log expression of the genes (KDCt) was obtained subtracting the number of cycle threshold observed for the GADPH gene from that observed for the gene of interest. siRNA transfection
T47D cells were pre-incubated with Lipofectamine 2000 (Invitrogen-Life Technologies Inc.) in serum-free Opti-MEM for 20 min before adding siRNA oligonucleotides (25 nM), diluted in the same medium (3 mg/ml lipofectamine final concentration). siRNAs, CLU-V-siRNA (selective for the cytoplasmic form of CLU), 5 0 -AUG AUG AAG ACU CUG CUG C-3 0 and 3 0 -UAC UAC UUC UGA GAC GAC G-5 0 ; control siRNA (scrambled of BIRC6 gene), 5 0 -GCA GUA CAU GGU AUG AUU AdTdT-3 0 , and 3 0 -dTdTCGU CAU GUA CCA UAC UAA U-5 0 were custommade (Dharmacon, Research Inc., Lafayette, CO, USA). After 24 h, medium containing the RNA duplex and lipofectamine was replaced with Dulbecco’s modified Eagle’s medium/Ham’s F12 (DMEM/F12) supplemented with 5% stripped FBS and cells were treated with 10K7 M 4-OH–TOR. The cell number was determined after 3 days by direct counting of viable cells in a Burker chamber. Experiments were repeated thrice. Western blotting
Cytosolic proteins (30 mg) were separated by SDSPAGE and transferred onto nitrocellulose membranes that were blocked for 1 h in Dulbecco’s phosphate www.endocrinology-journals.org
buffered saline (DPBS) 1!, 0.1% Tween-20 with 5% non-fat dry milk and incubated with primary antibody against CLU (1:500, sc-6419; Santa Cruz Biotechnology, Santa Cruz, CA, USA) and against b-actin (1:1000, 20– 33; Sigma–Aldrich). The membranes were washed and incubated for 1 h at room temperature with peroxidaseconjugated anti-goat (Santa Cruz Biotechnology) and anti-rabbit (Amersham Biosciences Europe) secondary antibodies (1:2000). Signals were detected by chemiluminescence (ECL, Amersham Biosciences Europe) according to the manufacturer’s instructions. The relative amounts of CLU protein were quantified by densitometric analysis and normalized with respect to b-actin signal. Data analysis and statistics
Raw microarray images and quantifications were stored and processed in BioArray Software Environment (BASE, Lund, Sweden; Saal et al. 2002). Poor signal quality of background-corrected Cy3 and Cy5 intensities were flagged and a lowess normalization (Yang et al. 2002) applied to each slide. Replicated spots were averaged and their log (base2) expression ratios (tumor/reference) were downloaded from BASE and imported into BRB-ArrayTools version 3.2.2 (Bethesda, MD, USA) for further analyses (Simon et al. 2007). All IMAGE clone annotations were updated with the latest release of NCBI Unigene (build no. 194) using SOURCE (Diehn et al. 2003). In the present study, we analyzed two different datasets. The first one contained 32 hybridizations and was used to evaluate: 1) gene expression patterns of pre-treatment core biopsies correlated with tumor shrinkage (as a categorical dichotomous or a continuous variable) after toremifene treatment (21 pre-treatment samples) and 2) gene expression changes in residual tumor specimens to identify molecular patterns after toremifene exposure (11 paired pre- and post-treatment samples). The second one containing 14 hybridizations, from 7 paired biopsy–surgical specimens, was used to evaluate gene expression changes in residual tumor specimens without any intercurrent local or systemic treatment. Starting from both chips containing 8303 and 8399 clones spotted in triplicate, low-quality spots (filtering criteria are described in the web Supplementary Material, which can be viewed online at http://erc. endocrinology-journals.org/supplemental/) from both datasets were flagged (set to missing value) and subsequently filtered to contain no more than 20% missing values for each clone, which reduced the first dataset to 12 784 clones and the second one to 4205 clones. 441
V Cappelletti, M Gariboldi et al.: Gene expression and anti-estrogen treatment Genes whose expression was significantly correlated with treatment response where identified by computing a statistical significance level for each gene by testing the hypothesis that Spearman’s correlation between gene expression and tumor shrinkage was zero. Class comparisons were performed using random-variance t-tests (Wright & Simon 2003; paired when applicable). In all analyses, genes were considered statistically significant if their P value was !0.001. The probability of finding a given number of genes below this threshold by chance was tested by repeated analysis on permutated class labels (Simon et al. 2003). Prior to the functional annotation of the lists using SOURCE (Diehn et al. 2003), DAVID (Dennis et al. 2003), and data from PubMed, clones with no gene symbols assigned to them by means of Unigene were mapped to the UCSC Human Genome using BLAT (Kent 2002, Karolchik et al. 2003). Hierarchical clustering of samples and genes was done using the 1-Pearson correlation as distance and average linkage method. All protocols, expression data, sample description, the list of annotated genes, and the microarray data in a format that conforms to the Minimum Information About a Microarray Gene Experiment (MIAME) guidelines of the Microarray Gene Expression Data Society (MGED) are available at http://pierotti.group. ifom-ieo-campus.it/suppl/br_tor.html. Microarray data have been deposited to the EBI ArrayExpress.
Results Pre-treatment gene profiles
Individual clinicopathological and biological tumor features of the 21 patients included in this study are reported in Table 1. Tumors were clinically classified as T2 (90%), node-negative (62%), and PgR-positive (71%), and the clinical response (i.e., tumor shrinkage R25%) was achieved in 67% of patients. Patients were comparable for clinicopathological and biological findings to the original subset of 38 toremifene-treated patients entered a prospective study adequately statistically sized to evaluate the predictive role of ER-b expression (Cappelletti et al. 2004) whose leftover biological samples were used for the present study. Initial class comparison results of clinical response considered as a binary variable (i.e., responders versus non-responders) identified only 15 genes at the 0.001 level with a high probability of finding such a number by chance (PZ0.153). Thereafter, when clinical response was considered as a continuous variable defined as percent tumor shrinkage (Table 1), 53 genes (Fig. 1), either positively (32) or negatively (21) correlated with tumor reduction (P!0.001), were identified (PZ0.0024 of finding 53 genes at 0.001 level by chance). Functional analysis (excluding unknown reporters) showed that they were mainly involved in cell cycle and proliferation (19.4%), signal transduction (16.7%), nucleic acid
Table 1 Clinicopathological and biological tumor characteristics
Patient ID 09 10 13 18 20 21 22 24 25 26 29 30 32 33 34 37 38 45 51 57 79
442
Initial (pre-treatment) tumor size (cm) 3!3 4!4 2.5!2.5 6!5 2.5!2 3!3 4.5!4 3.5!3.5 3.5!2.5 3!3 5!3.5 3.5!4 3!3 4!4 3!2 2.5!2 2.5!2.5 4!3.5 3!2.5 2.1!2.1 2.5!2.5
Clinical axillary positive nodes
PgR
% Tumor variation
Post-treatment specimen availability
No Yes Yes Yes No No Yes No No Yes Yes Yes No No No Yes No No No No No
C C C C K C C C C K K K C C C C C C K K C
K100 K75 K100 K80 K25 K30 K32 K60 K74 K17 K14 K14 K17 0 C50 K55 K64 K25 0 K49 K52
No Yes Yes No No No Yes No No No Yes Yes Yes Yes Yes Yes No No Yes Yes No
www.endocrinology-journals.org
Endocrine-Related Cancer (2008) 15 439–449 Genes involved in cell proliferation were more frequently upregulated in responding patients than in resistant ones (26.3% vs 11.8%) and included MOBK1B, DYRK2, RUNX3, BMI1, and GSPT1. So were also genes involved in membrane solute transport (including members of the solute carrier family as SLC11A2, SLCO3A1, SLC27A1, and ABCA5, belonging to ATP-binding cassette transporter family). Conversely, genes involved in nucleic acid processing, including zinc finger proteins (LOC90379, ZNF558, and PIAIS1) and helicases (SBNO1), were more frequently upregulated in non-responders. Treatment-modulated gene profiles
Before comparing the gene expression pattern in matched pre- and post-treatment specimens, we investigated expression profile modulation in serial paired biopsy and surgical specimens in the absence of intercurrent treatments. Unsupervised hierarchical clustering demonstrated that five out of the seven pairs had closely related transcriptional profiles. The class comparison based on paired t-test identified, at the 0.001 significance level, only two differentially expressed genes: an expressed sequence tag and TFDP1. Unsupervised analysis of paired pre- and post-treatment specimens
Figure 1 Heatmap of genes correlated with response to toremifene (percentage of tumor reduction). For each gene, expression levels above the mean are shown in red and those below the mean are shown in blue. The color scale ranges from K1 (darkest blue) to C1 (darkest red). IMAGE clone identifiers and gene symbols are reported in the rows whereas the samples and their percentage of tumor reduction are reported in the columns. The gene expression matrix was mean centered and the gene expression profiles were clustered using the average linkage hierarchical clustering method and 1-Pearson correlation as the distance method. All samples (14 responders C7 non-responders) were ordered according to the tumor shrinkage. *Annotated using BLAT.
processing/transcription (16.7%), and membrane solute transport (13.9%). We also identified some genes involved in protein transport/metabolism (11.1%), immune response (5.5%), and cell adhesion/ cytoskeleton (2.8%), and 17 genes (32.1%) had unknown function. In Table 2, the list of known genes associated with response is reported as subdividing genes according to biological function. www.endocrinology-journals.org
For 11 patients (five responders and six non-responders), paired samples before and after toremifene treatment were available. Unsupervised hierarchical cluster analysis using all available genes showed that pre- and post-treatment specimens clustered together in four out of five responding patients and in only two out of the six matched specimens from non-responding patients (Fig. 2). Supervised analysis of paired pre- and post-treatment specimens
Class comparison results considering all paired pre- and post-treatment samples identified 101 differentially expressed genes (P!0.001) following toremifene treatment (PZ0.009 of getting at least 101 genes significant by chance), with a negligible overlap with the 53 genes associated with tumor shrinkage. Only four genes were common to the two lists: MNT, SYAP1, and FLJ90757 and an expressed sequence tag, which were all downregulated by the treatment and negatively associated with tumor reduction. Since hierarchical clustering result suggested that gene expression before and after treatment seemed more similar in responders than in non-responders, 443
V Cappelletti, M Gariboldi et al.: Gene expression and anti-estrogen treatment Table 2 Genes associated with treatment response Biological function Genes positively associated with response Adhesion/cytoskeleton Protein transport/metabolism Signal transduction
Nucleic acid processing and transcription
Cell cycle/proliferation Immunology Membrane solute transport Other
Genes negatively associated with response Protein transport/metabolism Signal transduction
Nucleic acid processing and transcription Cell cycle/proliferation
Steroid metabolism Immunology Membrane solute transport
Other
we speculated that many of the differentially expressed genes after treatment were in part due to the heterogeneity observed in the latter subset. As expected, the class comparison of treatment-induced expression among the responders (five patients) identified 8 differentially expressed genes (P!0.001, PZ0.3125 of observing such a number), while 82 differentially expressed genes (P!0.001, PZ0.03125 of observing such a number) were identified among non-responders (six patients). Within genes modulated independently of treatment outcome protein transport/metabolism, cell signaling, and nucleic acid processing/transcription were more represented, while the cell adhesion class was less represented (Table 3). Genes coding for proteins related to cell stress response and apoptosis pathways, 444
Gene symbol
P value
C1QTNF7 VPS39 UBA52 FRS2 CLEC2D INPP4B ZNF558 PIAS1 SBNO1 LOC90379 MNT MAPK8 BPHL SYAP1 CLPTM1 ADCK4 TCTE3
0.0006799 0.0003019 0.0004614 0.0002194 0.0002404 0.0008168 0.0008585 0.0008585 0.0008853 0.0009408 0.0001893 0.0007346 1.29!10K5 0.0004961 0.0003625 0.000575 0.0007346
GALNT13 EXOC1 PIP5K1B CAB39 GRBP SNAPC3 INTS12 MOBK1B BMI1 GSPT1 DYRK2 RUNX3 OSBPL8 COL4A3BP SLCO3A1 ABCA5 SLC27A1 SLC11A2 C1QC
0.0009408 0.0007878 0.0004431 0.0005004 0.0008775 0.0003496 0.0009488 !1!10K7 0.0002498 0.0005487 0.0007878 0.0001817 9.5!10K6 !1!10K7 0.0005407 0.0005487 0.0005842 0.0007395 0.0003251
which were absent among the 53 genes associated with response, were also represented (8.3%). Among genes modulated in non-responding patients, cell adhesion/cytoskeleton, nucleic acid processing/ transcription, and signal transduction emerged as the most represented functional categories Among genes modulated in all patients upregulation predominated in the cell cycle proliferation category (JUNB, S100A10, and PPARG) and slightly in the cell adhesion and cytoskeleton category (CTTN, KAL1, and MXRA8). On the contrary, in the category of protein transport/metabolism genes, downregulation (18.9%) slightly overcame upregulation (14.3%). The former included RAB27A involved in protein transport and also in small GTPase-mediated signal transduction, SYT11, RPE, TTC7L1, and ANKIB1 involved in www.endocrinology-journals.org
Endocrine-Related Cancer (2008) 15 439–449 Among genes exclusively modulated in nonresponders upregulation of cell stress/apoptosis was more frequent and included both antiapoptotic (NCL and CLU) and pro-apoptotic (BCL2L13) genes (Table 4). Genes related to signal transduction and cell cycle/proliferation were more frequently downregulated and included only two upregulated genes, DKK3, a negative regulator of WNT signaling and MAPK6. Furthermore, a striking upregulation (41.2%) of genes associated with cell adhesion/cytoskeleton (ITGB1, SMOC2, FBLN2, FMOD, SERPIN1, TIMP3, and MMP2) was observed.
Figure 2 Hierarchical clustering of the gene expression matrix containing no missing values of 11 paired samples (five responders and six non-responders) before and after treatment. The solid lines indicate paired samples with similar gene expression profiles of responding patients and the dashed lines indicate pairing of non-responding patients. The percentage of tumor reduction is reported under each pre-treatment sample.
Table 4 Genes modulated by treatment in non-responder patients onlya Biological function Upregulated genes Cell stress/apoptosis
Adhesion/cytoskeleton
ubiquitin-dependent protein catabolism, the mitochondrial carrier protein MCART6 and CRLS1, a synthase involved in phospholipid biosynthesis, while upregulated genes included PDZRN3 involved in protein ubiquitination, PPIG, a peptidyl-prolyl isomerase accelerating protein folding and also involved in regulation of pre-mRNA splicing, and the translation initiation factor EIF3S2, besides a mannosidase (MAN1C1) involved in glycosylation. Table 3 Classification into functional categories of genesa modulated by the treatment independently from treatment response Functional categories of genes modulated by treatment Stress/apoptosis Cell adhesion/cytoskeleton Protein transport/ metabolism Signal transduction Nucleic acid processing Cell cycle proliferation ER signaling Steroid metabolism Immunology Membrane solute transport Other classes a
In all patientsb
In non-responder patientsb
8.3 6.9 16.7
9.1 20.0 10.9
13.9 13.9 8.3 – 2.8 1.4 9.7 18.1
14.5 20.0 7.3 1.8 – – 7.3 9.1
The percentage of unknown reporters was 28.7% on the overall series and 49.1% exclusively in specimens from nonresponder patients. b Relative percentages on known genes.
www.endocrinology-journals.org
Protein transport/ metabolism Signal transduction Nucleic acid processing and transcription Cell cycle/proliferation Membrane solute transport Other Downregulated genes Adhesion/cytoskeleton
Gene symbol
P value
CLU BCL2L13 NCL FMOD TIMP3 SERPIN1 FBLN2 SMOC2 MMP2 ITGB1 TOMM40
0.0001582 0.0006041 0.0008061 0.000246 0.0004362 0.0005444 0.0005989 0.0008047 0.000805 0.0008549 0.0009673
DKK3 ZNF71
0.0009238 0.0004078
GABPB2 MAPK6 SLC39A13
0.0005375 8.56!10K5 0.0006409
PPT1R11
0.0006333
C1QTNF7 CLDN18 ARPC4 Signal transduction ADRBK2 DKFZP686A10121 PTCH TXK Nucleic acid processing ZNF677 and transcription SBNO1 TAF1B ZNF331 SRPK2 Cell cycle/proliferation ORC1L HSMPP8 ER signalling NCOA5 Other BACE DLD
0.708 0.0002255 0.0004617 0.0004755 0.0005211 0.0006665 0.0009083 0.0002018 0.0004757 0.0006169 0.0007066 0.0007598 0.0001482 0.0007817 0.0008897 0.0004043 0.0005991
a Only those genes that are modulated in non-responders are reported, excluding genes common to other gene lists.
445
V Cappelletti, M Gariboldi et al.: Gene expression and anti-estrogen treatment Confirmation of gene expression data by real-time PCR
We validated cDNA microarray data by real-time PCR for eight genes differentially expressed in pre- and posttreated specimens from non-responding patients (TIMP3, MMP2, FBLN2, NLC, BCL2L13, CLU, MAPK6, and DKK3) and three genes whose expression correlated with tumor shrinkage (RUNX3, GSPT1, and DYRK2). Paired t-test and Spearman’s rank correlation analysis confirmed all but one (DYRK2) of the tested genes (see Supplementary Table 1). Biological validation of CLU
CLU, one of the genes upregulated following treatment exclusively in the subset of non-responders (Supplementary data Table 1), was identified as possible candidate for attempting to modulate the sensitivity to anti-estrogens in a toremifene-resistant cell line: T47D. CLU was chosen as candidate gene due to its association with survival and apoptosis (Shannon et al. 2006). Treatment with 10K7 M 4-OH toremifene or with the similar drug tamoxifen slightly upregulated CLU expression (Fig. 3A). CLU expression was
Figure 3 (A) Western blotting of CLU active forms in T47D cells treated for 6 days with 10K7 M 4-OH toremifene and 10K7 M 4-OH tamoxifene. (B) Western blotting of CLU precursor and active forms, in controls and in siRNA-treated samples. T47D cells were treated with 25 nM control siRNA and siRNA against the secreted form of CLU for 24 h in serum-free medium. The medium was replaced and after 3 days cells were harvested and lysed. The figure reports representative blots of three separate experiments.
446
silenced treating the cells with an siRNA specific for cytoplasmic CLU, the form responsible for the cytoprotective effects. CLU appeared on blots as the 60 kDa precursor and an about 40 kDa protein smear comprising the two a and b dimers and their glycosylated forms. Treatment with siRNA effectively suppressed the expression of the precursor and the active forms (K75 and K95% respectively; Fig. 3B), while no effects on protein expression were observed with the control siRNA. Treatment with 10K7 M toremifene for 3 days did not affect cell proliferation (percentage of untreated control: 102G4% vs 100G 5%). Addition of unrelated control siRNA, which did not affect CLU expression, resulted in a slight but not a statistically significant reduction (83G14%), while treatment with the siRNA CLU-V significantly inhibited cell proliferation down to 56G0.6% (PZ0.032 with respect to siRNA ctrl).
Discussion The combination of primary endocrine therapy with global gene profiling provides the opportunity to collect information to predict response to SERM and to identify genes involved in anti-estrogen resistance. The clinical response of the tumor can in fact be regarded as a surrogate of response to systemic treatment, and genes modulated in non-responding patients might help to identify pathways associated with resistance. In the present study, few genes were differentially expressed between responders and non-responders confirming the difficulties in finding a reliable molecular predictor of response using a top-down approach. Limited duration of clinical treatment, which resulted in few patients achieving pCR, scanty cases available for microarray analysis, and their heterogeneity could account for this failure. We did, however, identify a gene set significantly correlated with toremifene-induced tumor shrinkage evaluated on a continuous scale in pre-operative samples, without overlapping with previously identified signatures of endocrine treatment response in the adjuvant setting (Ma et al. 2004) or in women with advanced disease (Paik et al. 2004, Jansen et al. 2005). Conversely, a novel contribution of our study to the relation between gene expression profile and treatment effect resides in the analysis of transcriptome changes after antiestrogen treatment. To exclude the possibility that genes differentially expressed before and after treatment could be due to surgical procedure, elapsed time or simply to some random effects, rather that to treatment itself, we verified in advance that a negligible fraction of genes (2/4205) were differentially www.endocrinology-journals.org
Endocrine-Related Cancer (2008) 15 439–449 expressed in matched biopsies and surgical specimens collected within a time lapse of 3–4 weeks and not submitted to intercurrent treatment. Investigations on gene modulation in the absence of treatment are generally underrepresented in the literature, and even studies dealing with post-treatment changes in gene expression are scanty, with only one paper reporting differentially expressing genes following anastrozole in locally advance breast cancer (Nedelcheva et al. 2005), and few other papers investigating transcriptional profiling following docetaxel alone (Chang et al. 2003, 2005) or in combination with doxorubicin and cyclophosphamide (Hannemann et al. 2005), or erlotinib (Yang et al. 2005). In our study, treatment-modulated genes appear to differ from treatment outcome-associated genes, a finding also reported with a slightly different type of analysis in the only other study investigating posttreatment changes as a function of clinical response to anastrozole (Nedelcheva et al. 2005). These authors observed that most pre- and post-treatment tumor specimens clustered together, indirectly suggesting the existence of a set of treatment outcome-associated genes unchanged after treatment. This finding indirectly confirms and provides support to our observation of a negligible overlap between treatment response-associated and response-modulated genes. In addition, in our study treatment outcome appears to affect gene modulation, since transcriptome variations were minimal in responding patients in agreement with Chang et al. (2005) but in contrast with Hannemann et al. (2005). However, disagreements on gene modulations according to clinical response might be due to differences in treatment type and criteria used to assess clinical/pathological response, and to the possibility that in non-responding patients a differential gene expression reflects biological differences between epithelial and stromal cells (the latter frequently increased after primary treatment). Despite such problems and potential bias due to investigating treatment-induced changes mainly in partially or nonresponding patients, we reasoned that focusing on genes exclusively modulated in non-responders could provide information on anti-estrogen resistance. Indeed, many among the genes modulated only in non-responders are involved in the balance between pro-apoptotic and anti-apoptotic signals as well as in cell signaling from membrane receptors, which are known to play a role in the establishment of hormone resistance. This observation represents a proof of our hypothesis that genes upregulated after treatment in resistant patients may include potential therapeutic targets as well as genes representative of pathways www.endocrinology-journals.org
linked to resistance. This latter concept was further supported by the fact that our data indirectly confirm in a clinical setting some experimental studies aimed at understanding the action mechanism of SERM stimulation and the role of membrane ER (Shou et al. 2004), which associates with the scaffolding protein caveolin and activates proximal signaling molecules like G-proteins (Razandi et al. 2003). Such a signaling results into stimulation of phospholipase C and ErbB2/ EGFR heterodimers, which lead to activation of PIrK and AKT pathway (Stoica et al. 2003), and finally of MAPK. Specific G-proteins allow the activation of MMP able to transactivate heparin-bound EGF (Razandi et al. 2003). In our study, toremifene affected different genes involved in G-protein-mediated signaling, including GEM, which was upregulated independently of treatment response and could play a role as a regulatory protein in receptor-mediated signal transduction. Exclusively in the non-responder subset, toremifene upregulated MMP2 and MAPK6, which were shown to be activated by estradiol following binding to the membrane receptor, a mechanism involved in anti-estrogen resistance. A validation of gene signatures obtained from this study on independent patients set was not possible due to the lack of patients treated with pre-operative anti-estrogens. We, however, attempted a biological validation of the concept that genes modulated by the treatment exclusively in the non-responding patients, may include genes involved in hormone resistance. We chose an ERC anti-estrogenresistant cell line, T47D. Treatment with anti-estrogens upregulated CLU expression in agreement with that of Warri et al. (1993), although to a lesser extent because the anti-estrogen-resistant cell line expressed much more CLU compared with the sensitive MCF7 cells used by Warri et al. Silencing of clusterin using an siRNA specific for the cytoplasmic variant restored the sensitivity of T47D cells toward growth inhibitory effects of toremifene. Such results do not demonstrate a direct involvement of CLU in anti-estrogen-resistant mechanisms, but strongly support the concept that CLU may modulate response to anti-estrogens and represents therefore an interesting pharmaceutical target. Overall, this study suggests that knowledge of molecular networks modulated by treatment is particularly important in view of the abundance of promising molecularly targeted inhibitors currently available as treatment options.
Acknowledgements This work was supported by the Italian Association for Cancer Research (AIRC) and Special Grant ‘Ricerca 447
V Cappelletti, M Gariboldi et al.: Gene expression and anti-estrogen treatment Finalizzata’ from the Italian Health Ministry. The authors declare that there is no conflict of interest that would prejudice the impartiality of this scientific work.
References Cappelletti V, Celio L, Bajetta E, Allevi A, Longarini R, Miodini P, Villa R, Fabbri A, Mariani L, Giovanazzi R et al. 2004 Prospective evaluation of estrogen receptor-b in predicting response to neoadjuvant antiestrogen therapy in elderly breast cancer patients. EndocrineRelated Cancer 11 761–770. De Cecco L, Marchionni L, Gariboldi M, Reid JF, Lagonigro MS, Caramuta S, Ferrario C, Bussani E, Mezzanzanica D, Turatti F et al. 2004 Gene expression profiling of advanced ovarian cancer: characterization of a molecular signature involving fibroblast growth factor 2. Oncogene 23 8171–8183. Chang JC, Wooten EC, Tsimelzon A, Hilsenbeck SG, Gutierrez MC, Elledge R, Mohsin S, Osborne CK, Chamness GC, Allred DC et al. 2003 Gene profiling for the prediction of therapeutic response to docetaxel in patients with breast cancer. Lancet 362 362–369. Chang JC, Wooten EC, Tsimelzon A, Hilsenbeck SG, Gutierrez MC, Tham YL, Kalidas M, Elledge R, Mohsin S, Osborne CK et al. 2005 Patterns of resistance and incomplete response to docetaxel by gene expression profiling in breast cancer patients. Journal of Clinical Oncology 23 1169–1177. Dennis G Jr, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC & Lempicki RA 2003 DAVID: database for annotation, visualization, and integrated discovery. Genome Biology 4 P3. Diehn M, Sherlock G, Binkley G, Jin H, Metese JC, Hernandez-Boussard T, Reees CA, Cherry JM, Botstein D, Brown PO et al. 2003 SOURCE: a unified genomic resource of functional annotations, ontologies, and gene expression data. Nucleic Acids Research 31 219–223. Geisler J, Detre S, Berntsen H, Ottestad L, Lindtjorn B, Dowsett M & Eistein Lonning P 2001 Influence of neoadjuvant anastrazole (Arimidex) on intratumoral estrogen levels and proliferation markers in patients with locally advanced breast cancer. Clinical Cancer Research 7 1230–1236. Hannemann J, Oosterkamp HM, Bosch CA, Velds A, Wessels LF, Loo C, Rutgers EJ, Rodenhuis S & van de Vijver MJ 2005 Changes in gene expression associated with response to neoadjuvant chemotherapy in breast cancer. Journal of Clinical Oncology 23 3331–3342. Jansen MP, Foekens JA, van Staveren IL, Dirkzwager-Kiel MM, Ritstier K, Look MP, Meijervan Gelder ME, Sieuwerts AM, Portengen H, Dorssers LC et al. 2005 Molecular classification of tamoxifen-resistant breast carcinomas by gene expression profiling. Journal of Clinical Oncology 23 732–740.
448
Karolchik D, Baertsch R, Diekhans M, Furey TS, Hinrichs A, Lu YT, Roskin KM, Schwartz M, Sugnet CW, Thomas DJ et al. 2003 The UCSC genome browser database. Nucleic Acids Research 31 51–54. Kent WJ 2002 BLAT – the BLAST-Like alignment tool. Genome Research 12 656–664. Ma XJ, Wang Z, Ryan PD, Isakoff SJ, Barmettler A, Fuller A, Muir B, Mohapatra G, Salunga R, Tuggle JT et al. 2004 A two-gene expression ratio predicts clinical outcome in breast cancer patients treated with tamoxifen. Cancer Cell 5 607–616. Nedelcheva Kristensen V, Sorlie T, Geisler J, Yoshimura N, Linegjaerde OC, Glad I, Frigessi A, Harada N, Lonning PE & Borresen-Dale AL 2005 Effects of anastrazole on the intratumoral gene expression in locally advanced breast cancer. Journal of Steroid Biochemistry and Molecular Biology 95 105–111. Paik S, Shak S, Tang G, Kim C, Baker J, Cronin M, Baehner FL, Walker MG, Watson D, Park T et al. 2004 A multigene assay to predict recurrence of tamoxifentreated, node-negative breast cancer patients. New England Journal of Medicine 351 2817–2826. Peto R, Boreham J, Clarke M, Davies C & Beral V 2000 UK and USA breast cancer deaths down 25% in year 2000 at ages 20–69 years. Lancet 355 1822. Razandi M, Pedram A, Park ST & Levin ER 2003 Proximal events in signalling by plasma membrane estrogen receptors. Journal of Biological Chemistry 278 2701–2712. R Development Core Team 2006 R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-90005107-0, URL http://www.R-project.org. Saal LH, Troein C, Vallon-Christersson J, Gruvberger S, Borg A & Peterson C 2002 BioArray software environment (BASE): a platform for comprehensive management and analysis of microarray data. Genome Biology 3 software0003.1–software0003.6. Shannon B, Seifert M, Leskov K, Willis J, Boothman D, Tilgen W & Reichrath J 2006 Challenge and promise: roles for CLU in pathogenesis, progression and therapy of cancer. Cell Death and Differentiation 13 12–19. Shou J, Massarweh S, Osborne CK, Wakeling Wakeling AE, Ali S, Weiss H & Schiff R 2004 Mechanisms of tamoxifen resistance: increased estrogen receptor-HER2/neu crosstalk in ER/HER2-positive breast cancer. Journal of the National Cancer Institute 96 926–935. Simon R, Korn E, McShane L, Radmacher MD, Wright GW & Zhao Y 2003 In Design and Analysis of DNA Microarray Investigations, ch 7. New York: Springer-Verlag. Simon R, Lam A, Li MC, Ngan M, Menenzes S & Zhao Y 2007 Analysis of gene expression data using BRB-Array tools. Cancer Informatics 2 11–17. Stoica GE, Franke TF, Moroni M, Mueller S, Morgan E, Iann MC, Winder AD, Reiter R, Wellstein A & Martin MB
www.endocrinology-journals.org
Endocrine-Related Cancer (2008) 15 439–449 2003 Effect of estradiol on estrogen receptor-a gene expression and activity can be modulated by the ErbB2/PI 3-K/Akt pathway. Oncogene 22 7998–8011. Wa¨rri AM, Huovinen RL, Laine AM, Martikainen PM & Harkonen PL 1993 Apoptosis in toremifene-induced growth inhibition of human breast cancer cells in vivo and in vitro. Journal of the National Cancer Institute 85 1412–1418. Wright GW & Simon R 2003 A random variance model for detection of differential gene expression in small microarray experiments. Bioinformatics 19 2448–2455.
www.endocrinology-journals.org
Yang YH, Dudoit S, Luu P, Lin DM, Peng V, Ngai J & Speed TP 2002 Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variation. Nucleic Acids Research 3 e15. Yang SX, Simon RM, Tan AR, Nguyen D & Swain SM 2005 Gene expression patterns and profile changes pre-and post-erlotinib treatment in patients with metastatic breast cancer. Clinical Cancer Research 11 6226–6232.
449
