Identification of a Genetic Signature of Activated Signal Transducer and Activator of Transcription 3 in Human Tumors (original) (raw)

Skip Nav Destination

Molecular Biology, Pathobiology, and Genetics| June 15 2005

James V. Alvarez;

1Department of Medical Oncology, Dana-Farber Cancer Institute, and Departments of

2Medicine and

Search for other works by this author on:

Phillip G. Febbo;

1Department of Medical Oncology, Dana-Farber Cancer Institute, and Departments of

2Medicine and

Search for other works by this author on:

Sridhar Ramaswamy;

2Medicine and

6Center for Cancer Research, Massachusetts General Hospital, Charlestown, Massachusetts

Search for other works by this author on:

Massimo Loda;

1Department of Medical Oncology, Dana-Farber Cancer Institute, and Departments of

3Pathology, Harvard Medical School; Departments of

Search for other works by this author on:

Andrea Richardson;

3Pathology, Harvard Medical School; Departments of

4Pathology and

Search for other works by this author on:

David A. Frank

1Department of Medical Oncology, Dana-Farber Cancer Institute, and Departments of

2Medicine and

5Medicine, Brigham and Women's Hospital, Boston, Massachusetts; and

Search for other works by this author on:

Requests for reprints: David A. Frank, Department of Medical Oncology, Dana-Farber Cancer Institute, 44 Binney Street, Mayer 522B, Boston, MA 02115. Phone: 617-632-4714; Fax: 617-632-6356; E-mail: david_frank@dfci.harvard.edu.

Received: November 30 2004

Revision Received: March 08 2005

Accepted: April 01 2005

Online ISSN: 1538-7445

Print ISSN: 0008-5472

2005

Cancer Res (2005) 65 (12): 5054–5062.

Article history

Received:

November 30 2004

Revision Received:

March 08 2005

Split-Screen
Views Icon Views
Open the PDF for in another window
Tools Icon Tools
Search Site
Article Versions Icon Versions
- Version of Record June 15 2005

Abstract

Signal transducer and activator of transcription 3 (STAT3) is a transcription factor that is activated in diverse human tumors and may play a direct role in malignant transformation. However, the full complement of target genes that STAT3 regulates to promote oncogenesis is not known. We created a system to express a constitutively active form of STAT3, STAT3-C, in mouse fibroblasts and used it to identify STAT3 targets. We showed that a subset of these targets, which include transcription factors regulating cell growth, survival, and differentiation, are coexpressed in a range of human tumors. Using immunohistochemical staining of tissue microarrays, we showed that these targets are enriched in breast and prostate tumors harboring activated STAT3. Finally, we showed that STAT3 is required for the expression of these genes in a breast cancer cell line. Taken together, these results identify a cohort of STAT3 targets that may mediate its role in oncogenesis.

Introduction

Several transcription factors are inappropriately activated in human tumors and capable of transforming cell lines in vitro. These oncogenic transcription factors function by activating or repressing target genes that collaborate to promote cell survival and proliferation. Accumulating evidence suggests that signal transducer and activator of transcription 3 (STAT3), a member of the STAT family of proteins, is such an oncogene. STAT3 is activated by phosphorylation on a tyrosine residue in its COOH terminus; this phosphorylation is followed by dimerization, nuclear translocation, DNA binding, and transcriptional activation of STAT3 target genes. Several human cancers, including breast cancer, prostate cancer, and head and neck cancer, and several hematologic malignancies, including acute myelogenous leukemia and multiple myeloma, contain persistently tyrosine-phosphorylated STAT3 (1). Furthermore, STAT3 is necessary for v-src transformation of fibroblasts, and a constitutively active mutant of STAT3 is sufficient to transform fibroblasts (2–4).

Several targets of STAT3 have been identified using human tumor cell lines and rodent fibroblasts. Proteins that regulate cell survival, including bcl-2, bcl-xL, mcl-1, and Fas, are direct targets of STAT3 (5–8). Cell cycle regulators cyclin D1, cyclin E1, and p21 can be activated by STAT3 in certain contexts (9). Finally, other transcription factors, including myc, jun, and fos, are themselves STAT3 targets (10–13). However, little is known about which of these targets are critical mediators of STAT3 in transformation. Furthermore, none of these targets is overexpressed in tumors in which STAT3 activation is observed. Thus, although many of these genes have independently been implicated in oncogenesis, their role in and contribution to human cancers in which STAT3 is activated is unclear.

In the present study, we undertake a genome-wide analysis of STAT3 target genes and analyze the expression of these genes in human tumor expression data sets. We find evidence for an expression signature for STAT3 and show that the expression of the genes composing this signature correlates with STAT3 activation. We thus identify several genes that may mediate the role of STAT3 in cancer. Further, this study suggests the benefit of combining cancer genomics with traditional cell biological methods to better understand the role of individual genes in the biology of human tumors.

Materials and Methods

Plasmids. To establish an inducible expression system, we used the GeneSwitch system from Invitrogen Life Technologies (Carlsbad, CA). pGeneB (Invitrogen) was modified to create pGene-stop by inserting a linker containing a stop codon upstream of the V5 and His tags. STAT3-C/pRcCMV (kindly provided by J. Bromberg, Memorial Sloan-Kettering Cancer Center, New York, NY) was cloned into the _Not_I-_Apa_I sites of pGene-stop. m67 pTATA TK-luc was kindly provided by J. Bromberg. pSwitch was from Invitrogen. The Renilla reporter phRL TK-luc was from Promega (Madison, WI).

Cell culture, transfection, and cytokines. To create an inducible STAT3-C cell line, STAT3-C/pGene-stop and pSwitch or pGene-stop and pSwitch were cotransfected into NIH3T3 cells using LipofectAMINE 2000 (Invitrogen). Cells were selected in 400 μg hygromycin (Roche Applied Science, Indianapolis, IN) and 700 μg zeocin (Invitrogen). Individual clones were picked using cloning rings (Fisher Scientific, Pittsburgh, PA), or transfected cells were pooled. Clones were screened for those that expressed no STAT3-C before induction and near-physiologic levels following induction. Mifepristone (Invitrogen) was used at 0.1 nmol/L for all experiments unless otherwise stated.

NIH3T3 cells, mouse embryonic fibroblasts, and MDA-MB-231 cells were grown in DMEM supplemented with 10% fetal bovine serum. Interleukin-6 (IL-6; R&D Systems, Minneapolis, MN) was used at 30 ng/mL, soluble IL-6 receptor (R&D Systems) was used at 50 ng/mL, and murine OSM (R&D Systems) was used at 25 ng/mL. Cells were pretreated with soluble IL-6 receptor for 1 hour before addition of IL-6.

Western blot analysis. Cells were lysed in radioimmunoprecipitation assay buffer [50 mmol/L Tris (pH 7.4), 150 mmol/L NaCl, 1% NP40, 0.5% sodium deoxycholate, 0.1% SDS] containing 1 mmol/L phenylmethylsulfonyl fluoride, 1 μg/mL pepstatin, and 1 mmol/L sodium vanadate on ice for 15 minutes. Protein (50 μg) was resolved on 8% SDS-polyacrylamide gels and transferred to nitrocellulose. Blots were probed with antibodies against the FLAG epitope (M2, Sigma, St. Louis, MO), STAT3 (Santa Cruz, Santa Cruz, CA), phospho-STAT3 (pSTAT3; Cell Signaling, Inc., Beverly, MA), or tubulin (Sigma).

Luciferase assays. 3T3/STAT3-C.CE, 3T3/STAT3-C.Pool, or 3T3/pGene cells were plated on 24-well plates and transfected with 1.8 μg m67-luc and 0.2 μg phRL TK-luc with LipofectAMINE 2000 for 16 hours. After transfection, cells were treated with the indicated concentration of mifepristone for 24 hours. Cells were lysed and luminescence was measured using the dual-luciferase reagents from Promega, according to the manufacturer's instructions, using a Luminoskan Ascent luminometer (ThermoLab Systems, Helsinki, Finland). STAT3-dependent luciferase production was normalized to control Renilla values.

RNA isolation, reverse transcription-PCR, and microarray analysis. RNA was isolated from cells using either Trizol reagent (Invitrogen) for microarray analysis or RNeasy kits (Qiagen, Valencia, CA) for reverse transcription-PCR (RT-PCR) according to each manufacturer's protocol. For semiquantitative RT-PCR, RNA (10 ng) was reverse transcribed and amplified using SuperScript One-Step RT-PCR kit (Invitrogen) for 30 cycles. For real-time RT-PCR, RNA (2 μg) was reverse transcribed with a polydeoxythymidylic acid primer using the SuperScript First-Strand Synthesis kit (Invitrogen). Real-time PCR was then done with a SYBR Green Master Mix (Stratagene, La Jolla, CA) on an ABI Prism 7000 instrument.

Primer pairs are as follows: mouse primers: SOCS3, forward GTTCCTGGATCAGTATGATGC and reverse CGCTTGTCAAAGGTATTGTCC; c-met, forward TCTCTCGAACAGCACACCTC and reverse TTGAGTCCATGTACCGCTGG; mcl-1, forward GACCGGCTCCAAGGACTC and reverse TGTCCAGTTTCCGGAGCAT; bcl-6, forward GAGCCCATAAGACAGTGCTCA and reverse GGTTGCATTTCAACTGGTCA; junB, forward GGTCAGGGATCAGACACA and reverse AAAGTACTGTCCCGGAGG; and actvr, forward GGGGACTGGTGTAACAGGAA and reverse TACTGCAAACACCACCGAGA; and human primers: bcl-6: forward TACCTGCAGATGGAGCAT and reverse ACTCTTCACGAGGAGGCT; mcl-1: forward GAGACCTTACGACGGGTT and reverse TTTGATGTCCAGTTTCCG; junB: forward AAATGGAACAGCCCTTCT and reverse TGTAGAGAGAGGCCACCA; egr1: forward AGCCCTACGAGCACCTGAC and reverse AGCGGCCAGTATAGGTGATG; calpain: forward GCAGGGATCTTTCACTTCCA and reverse GCTGAATGCACAAAGAGCAG; KLF4: forward TCCCATCTTTCTCCACGTTC and reverse AGTCGCTTCATGTGGGAGAG; and β-actin: forward TCCCTGGAGAAGAGCTACGA and reverse AGCACTGTGTTGGCGTACAG.

Microarray analysis, including the preparation of cRNA, oligonucleotide array hybridization to MG-U74Av2 GeneChip arrays (Affymetrix, Santa Clara, CA), and scanning of the arrays, was done by the Dana-Farber Microarray Core Facility. Briefly, RNA was harvested using Trizol reagent from untreated cells or cells treated with mifepristone for 4.5 hours. RNA was converted to cDNA and in vitro transcribed in the presence of biotinylated CTP and UTP. Labeled RNA was hybridized to an Affymetrix MG-U74Av2 chip containing 6,000 annotated genes and 6,000 expressed sequence tags, washed, and detected using phycoerythrin-conjugated streptavidin. The mean fluorescence intensity of each chip was normalized and the expression levels of all genes were compared. Data analysis was done using the DNA-Chip Analyzer software (14). Gene array data were normalized, and perfect match–only, model-based expression intensities were obtained in DNA-Chip Analyzer.

Hierarchical clustering and statistical analysis. Human tumor data sets (global cancer map, prostate, and leukemia) that we used have been described elsewhere and are available on the Broad Institute Web site.7

The breast tumor data set is unpublished data (A. Richardson). Hierarchical clustering was done using DNA-Chip Analyzer software. Kolmogorov-Smirnov analysis, used to measure coexpression of STAT3 targets and their association with the pSTAT3 phenotype, was done essentially as described (15, 16) using software designed for this purpose.8

A. Subramanian et al., manuscript in preparation.

Briefly, for Kolmogorov-Smirnov analysis, each gene present on the Affymetrix chip was queried with the set of STAT3 targets to generate a Kolmogorov-Smirnov score for that gene. All genes were ordered based on their Kolmogorov-Smirnov score, and the location of STAT3 targets on this ordered list was considered. The position of each STAT3 target on the list represents a nominal P for that gene; those targets in the top 5% of the list are coexpressed with the remaining STAT3 target genes to a statistically significant extent.

To test the enrichment of STAT3 targets in tumors with pSTAT3, all the genes on the chip are ranked based on their correlation with a given phenotype. We ranked the genes based on differential expression between tumors with and without pSTAT3 using the signal-to-noise score. For prostate tumors, we compared tumors with no pSTAT3 (score = 0) and those with high pSTAT3 (score = 3), and for breast tumors, we compared tumors with no pSTAT3 (score = 0) and those with intermediate and high pSTAT3 (score = 2 and 3). We then determined the location of STAT3 targets on this ordered list using the Kolmogorov-Smirnov score as described above; this Kolmogorov-Smirnov score measures the degree of enrichment and so is termed an enrichment score in Fig. 5. The statistical significance of this score was measured by randomly permuting the class labels (pSTAT3 status), recalculating signal-to-noise scores, and generating a Kolmogorov-Smirnov score for these random class distinctions. The nominal P represents how many times a Kolmogorov-Smirnov score from random class labels exceeds the test Kolmogorov-Smirnov score.

Immunohistochemistry. Breast tissue microarrays used in this study contained two representative 0.6 mm cores of each tumor and several cores of representative normal breast tissue, and prostate tissue microarrays contained three representative cores of each tumor. pSTAT3 protein expression was determined by immunohistochemistry using the pSTAT3 antibody (1:25 dilution). Immunohistochemistry was done using an autostainer (Optimax i6000, Biogenex, San Ramon, CA) with the VectaStain avidin-biotin complex kit (Vector Laboratories, Burlingame, CA) according to manufacturer's instructions. Briefly, sections were deparaffinized and rehydrated in water, and antigen retrieval was done by microwaving thrice in 10 mmol/L citrate buffer (pH 6.0) for 5 minutes each followed by 30 minutes of cooling. Sections were sequentially blocked with 3% hydrogen peroxide, avidin (Vector Laboratories), biotin (Vector Laboratories), and protein block (DAKO, Carpinteria, CA) and then incubated with pSTAT3 antibody (1:25 dilution) for 1 hour. Sections were washed thrice for 5 minutes in PBS and incubated with secondary antibody (MultiLink, Biogenex) for 20 minutes and avidin-biotin complex (horseradish peroxidase label, Biogenex) for 20 minutes. Sections were incubated with the VIP substrate (Vector Laboratories) for 10 minutes and counterstained with methyl green. Only nuclear reactivity was considered positive and was scored as 0 (no nuclear staining), 1+ (weakly positive, only slightly darker than negative nuclei), 2+ (moderately positive, nuclei were significantly darker than negative cells), or 3+ (strongly diffusely positive).

RNA interference. pRetroSuper (pRS) was generously provided by Anton Berns (Netherlands Cancer Institute, Amsterdam, the Netherlands). Oligonucleotides were designed, using guidelines described previously (17), to target the following sequence present in human STAT3: AACTTCAGACCCGTCAACAAA.

The oligonucleotides were designed with overhangs on each end to facilitate cloning into pRS. The sequence of the oligonucleotides were forward GATCCCCCTTCAGACCCGTCAACAAATTCAAGAGATTTGTTGACGGGTCTGAAGTTTTTGGAA and reverse AGCTTTTCCAAAAACTTCAGACCCGTCAACAAATCTCTTGAATTTGTTGACGGGTCTGAAGGGG. The targeting sequence and its complement are shown in bold.

pRS was transfected into 293 cells for packaging of virus (18). Viral supernatant (4 mL), supplemented with 4 mL growth medium, was used to infect MDA-MB-231 cells on a 10 cm plate for 16 hours in the presence of 4 μg/mL polybrene. Culture medium was then changed and puromycin was added to cells 24 hours later at 1 μg/mL. Puromycin-resistant cells were pooled and screened for STAT3 expression by Western blot.

Results

Development of a system to identify signal transducer and activator of transcription 3 transcriptional targets. To identify transcriptional targets of STAT3, we chose to express a constitutively active form of the protein in NIH3T3 cells. Expression of STAT3-C is sufficient for transformation of these cells and therefore provides a good system to identify targets of STAT3 involved in oncogenesis (4). We created several clones and one pool of 3T3 cells in which STAT3-C expression could be turned on with the small-molecule mifepristone (Fig. 1A). To test whether the induced protein was transcriptionally active, we transfected 3T3 cells with a reporter construct with four STAT binding sites in the promoter (19). Mifepristone treatment led to a 10- to 50-fold induction of luciferase activity, indicating that induced STAT3-C was able to activate expression from a STAT3-dependent promoter (Fig. 1B). The magnitude of STAT3-C-mediated target gene induction was similar to the magnitude of IL-6- and OSM-mediated gene induction, suggesting that STAT3-C was acting at physiologic levels. To determine whether STAT3-C could activate endogenous STAT3 targets, we measured the expression of a known STAT3 target, SOCS3. Following mifepristone treatment, SOCS3 mRNA was increased, further confirming that mifepristone led to induction of transcriptionally active STAT3-C (Fig. 1C). Recognizing that STAT3 can induce the expression of other transcriptional modulators, and because we wished to identify primary as opposed to secondary transcriptional effects of STAT3, we analyzed the time course of SOCS3 expression shortly after mifepristone treatment. SOCS3 mRNA was maximally induced 4.5 hours after treatment with mifepristone (Fig. 1D). We therefore chose this early time point for subsequent experiments.

Figure 1.

Inducible expression of STAT3-C in NIH3T3 cells. A, a system by which constitutively active STAT3-C could be inducibly expressed was introduced into NIH3T3 cells. A pool or two clones of NIH3T3/STAT3-C cells were treated with increasing concentrations of mifepristone (0.01-1 nmol/L) for 24 hours. A Western blot was done on whole cell lysate using an anti-FLAG antibody to detect STAT3-C. Arrow, concentration of mifepristone used for the remainder of experiments, 0.1 nmol/L. B, 3T3/STAT3-C or vector control cells were transfected with a STAT3-dependent luciferase construct, m67-luc, and a Renilla luciferase plasmid as an internal control. Cells were treated with varying concentrations of mifepristone, and luciferase expression was assayed 24 hours later. STAT3-dependent luciferase was normalized to Renilla luciferase. C, 3T3/STAT3-C or vector control cells were treated with 0.1 nmol/L mifepristone for 6 hours. Total RNA was harvested and SOCS3 expression was assayed by RT-PCR. D, 3T3/STAT3-C cells were treated with mifepristone for the times indicated, after which RNA was harvested and SOCS3 expression was determined by RT-PCR.

Figure 1.

Close modal

Genome-wide analysis of signal transducer and activator of transcription 3 targets. Equipped with a system to specifically activate STAT3, we sought to identify all the transcriptional targets of STAT3 using microarray expression analysis. A clone and a pool of 3T3/STAT3-C cells, as well as cells containing empty vector as a control, were starved of serum for 16 hours and then treated with mifepristone or vehicle for 4.5 hours. RNA was harvested from these cells and expression analysis was done using Affymetrix MG-U74Av2 oligonucleotide microarrays. One hundred eleven probe sets representing 66 known genes and 43 cDNAs or expressed sequence tags were increased ≥1.5-fold following mifepristone treatment in the clone, whereas 67 were increased by this magnitude in the pool (Fig. 2A; Supplementary Table S1). We chose this cutoff because a known STAT3 target, mcl-1, was increased by approximately this magnitude and because a previous study found known STAT3 targets to be induced by this magnitude (20). The clone and pool yielded nearly identical results (Fig. 2A): using a threshold of 1.5-fold, the correlation of probes sets between the clone and the pool was high (Pearson correlation = 0.944), so only data from the clone are discussed in detail here. Only one of these probe sets (TIEG, Gene ID 21847) was increased in the vector cells treated with mifepristone (data not shown), indicating that the genes identified are specific STAT3 targets and are not responding to mifepristone alone. Fourteen probes in total were changed by ≥1.5-fold in the pool, although none of these changed by >2-fold. Three probe sets, representing two genes (Ndr1, Gene ID 17990; Slc1a4, Gene ID 55963), were decreased ≥1.5-fold in response to mifepristone (data not shown). Several genes described previously as STAT3 targets were identified in the microarray screen, thus validating this approach for finding STAT3 targets. These include SOCS3, junB, CCAAT/enhancer binding protein β, mcl-1, and vascular endothelial growth factor (VEGF). A large number of the targets were themselves transcription factors, suggesting that STAT3 initiates a cascade of changes in the transcriptional profile of these cells. Notably, several of the genes suggested previously to mediate the role of STAT3 in oncogenesis, like c-myc, cyclin D1, and bcl-xL, were not activated by STAT3 at this time point. Although STAT3 has clearly been shown to be necessary for the expression of these genes in several contexts (9, 10), these findings reveal that STAT3 is not sufficient for their expression at this early time point. This raises the possibility that the protein products of other STAT3 target genes are required to collaborate with STAT3 to activate expression of these genes.

Figure 2.

Microarray analysis of STAT3 target genes. A, 3T3/STAT3-C cells or cells containing vector alone were starved overnight and then treated with mifepristone for 4.5 hours. Total RNA was harvested and the expression level of ∼12,000 genes was determined using Affymetrix MG-U74Av2 microarrays. Relative expression is shown by a color scale, with red representing higher expression and blue representing lower expression. The expression of all genes whose expression changed by ≥1.5-fold is shown. B, 3T3/STAT3-C cells were treated with mifepristone for 4.5 hours, and the expression of several STAT3 targets was determined by RT-PCR. C, mouse embryonic fibroblasts were treated with IL-6 or OSM for 2 hours. Total RNA was harvested and reverse transcribed, and the abundance of each transcript was determined by real-time PCR. Values are normalized to an internal control gene, hypoxanthine phosphoribosyltransferase, and expressed as fold induction over untreated cells. D, RNA was harvested from NIH3T3 cells and 3T3 cells transformed with v-src and was then reverse transcribed, and the abundance of each transcript was determined by real-time PCR. Values are normalized to an internal control gene, hypoxanthine phosphoribosyltransferase, and expressed as fold induction over NIH3T3 cells.

Figure 2.

Close modal

We validated the microarray results by analyzing the expression of several genes in response to mifepristone by semiquantitative RT-PCR. The expression of each gene we tested (c-met, mcl-1, bcl-6, and SOCS3) was increased in response to mifepristone (Fig. 2B). To further confirm that the genes we identified are true STAT3 targets, we tested their expression in response to other stimuli known to activate STAT3. The mRNA of several of the targets tested was increased in mouse embryonic fibroblasts after treatment with IL-6 or OSM for 2 hours (Fig. 2C). These genes were also induced when cells were pretreated with 10 μmol/L cycloheximide for 1 hour, suggesting that translation of an intermediate transcription factor is not required for their induction (Supplementary Fig. S1). Further, the expression of three of five targets tested was elevated in NIH3T3 cells transformed with v-src compared with parental NIH3T3 cells (Fig. 2D). Thus, these genes respond to physiologic and oncogenic activators of STAT3.

Signal transducer and activator of transcription 3 targets are coordinately expressed in human tumors. It is likely that only a few of these STAT3 targets are the key mediators of STAT3 in human cancers. We sought an alternative to in vitro systems to identify the important cancer-related STAT3 targets. Several data sets exist containing expression data for thousands of genes across hundreds of human tumors. We considered whether these data sets could reveal which of these STAT3 targets were likely to play a role in cancer.

We identified the human orthologues of the known mouse genes and analyzed their expression levels in various tumor data sets. The genes represented by probe sets varied slightly among the data sets (Supplementary Table S2). One such set, the global cancer map, contains expression data for 190 human tumors representing 14 distinct tumor types (21). We first did hierarchical clustering of the STAT3 target genes across these tumors (Supplementary Fig. S2). A subset of these genes was overexpressed in central nervous system tumors, leukemias, and prostate tumors; notably, STAT3 activation has been shown in each of these tumor types (22–24).

It is difficult to determine whether an apparent clustering of genes, as revealed by hierarchical clustering, is statistically significant. We reasoned that those STAT3 targets that subserve the critical oncogenic functions of STAT3 should be consistently highly expressed in tumors in which STAT3 is activated and expressed at low levels in tumors that lack STAT3 activation. These genes should thus be coordinately expressed in human tumor data sets, and their expression will serve as a genetic signature of STAT3 activation.

A strategy based on Kolmogorov-Smirnov analysis was used to determine which STAT3 targets were coexpressed in human tumors (Fig. 3A). Kolmogorov-Smirnov scanning has been used to discover genes whose expression correlates with the aggregate expression of a set of genes (15). We used Kolmogorov-Smirnov scanning to determine which of the STAT3 target genes were significantly coexpressed with the remaining targets using four independent data sets: the global cancer map; one data set comprising leukemias (25); a data set comprising 96 breast tumors;9

A. Richardson, unpublished data.

and a data set comprising prostate tumors (26). In this manner, we identified 12 genes that are coexpressed to a statistically significant extent in at least two of the four tumor data sets (Fig. 3B). These genes constitute an expression signature for STAT3 activation and, as suggested by their highly significant coexpression in human tumors, may represent the critical effectors of STAT3 activation in malignancy.

Figure 3.

STAT3 target genes are coordinately expressed in human tumors. A, strategy for determining coordinate expression of STAT3 targets. Every gene present on the Affymetrix chip was queried with the set of STAT3 target genes to generate a Kolmogorov-Smirnov (KS) score. Genes were ranked based on their Kolmogorov-Smirnov scores and the position of STAT3 targets within this ranked list was analyzed. Those targets in the top 5% of each list (P < 0.05) were considered as significantly coexpressed and constitute a STAT3 expression signature; these genes are listed in (B). B, genes constituting the STAT3 signature. For each gene, the Kolmogorov-Smirnov score and associated P in each tumor type are shown. The P is given by the position of each gene in a list of all genes ordered by Kolmogorov-Smirnov score divided by the total number of genes in the list.

Figure 3.

Close modal

Signal transducer and activator of transcription 3 signature genes are enriched in tumors with activated signal transducer and activator of transcription 3. We next determined whether this expression signature was associated with STAT3 activation in human tumors. Gene expression profiles of breast and prostate tumor cohorts have been obtained previously using Affymetrix gene arrays. The STAT3 activation status of the tumors was determined by immunohistochemical detection of activated STAT3 on tissue microarrays that had been prepared for these tumor cohorts. These tissue microarrays were stained with an antibody that detects the activated form of STAT3, which is phosphorylated on Tyr705 (pSTAT3). Samples were scored based on their nuclear pSTAT3 staining. We found that 66 of 96 (68%) breast tumors and 44 of 56 (78.6%) prostate tumors contain pSTAT3 (Fig. 4A and B). This is consistent with previous results showing STAT3 activation in 82% of prostate tumors (22) and 57% of breast tumors (27).

Figure 4.

$Figure 4. STAT3 is phosphorylated in sections from breast and prostate tumors. A, STAT3 activation was determined in sections from human breast tumors by immunohistochemical staining of tissue sections with an antibody against phosphorylated STAT3. The fraction of tumors in each class is shown. B, same as (A), except STAT3 activation was measured in human prostate tumors.$

STAT3 is phosphorylated in sections from breast and prostate tumors. A, STAT3 activation was determined in sections from human breast tumors by immunohistochemical staining of tissue sections with an antibody against phosphorylated STAT3. The fraction of tumors in each class is shown. B, same as (A), except STAT3 activation was measured in human prostate tumors.

Figure 4.

Close modal

Tumors were divided into two groups: those with no pSTAT3 immunoreactivity (score = 0) and those with high pSTAT3 immunoreactivity (2+ to 3+). Tumors with low pSTAT3 immunoreactivity (1+) were not used in this analysis. All genes represented on the Affymetrix microarray were then ranked based on differential expression between these groups using the signal-to-noise metric, such that genes expressed more highly in tumors with pSTAT3 were at the top of the list and genes expressed more highly in tumors lacking pSTAT3 were at the bottom. We then determined whether STAT3 targets are present near the top of the list of genes ranked by signal-to-noise using Kolmogorov-Smirnov analysis, which has been used as a sensitive technique to test whether a set of genes is associated with a given phenotype (16).

We tested whether the STAT3 signature genes, as well as the entire group of STAT3 targets we identified, were enriched in tumors with pSTAT3. The entire set of STAT3 targets we identified in our in vitro system showed a significant enrichment in prostate tumors harboring pSTAT3 (P = 0.024; Fig. 5A), and the genes composing the STAT3 signature also showed notable enrichment in tumors with pSTAT3 (P = 0.052; Fig. 5B). The entire set of STAT3 targets showed enrichment in breast tumors with pSTAT3, although this did not reach statistical significance (P = 0.079; Fig. 5C). In contrast, the refined list of STAT3 targets, which constitute the STAT3 signature (see Fig. 3B), did show statistically significant association with pSTAT3 in breast tumors (P = 0.043; Fig. 5D). As there is some tissue specificity to the STAT3 signature, we also tested whether the subset of STAT3 signature genes showing significant Kolmogorov-Smirnov scores in prostate were enriched in prostate tumors with pSTAT3 and whether those genes showing significant Kolmogorov-Smirnov scores in breast were enriched in breast tumors with pSTAT3. We found no greater enrichment with breast cancer–specific STAT3 signature genes but found greater enrichment with prostate cancer–specific STAT3 signature genes (P = 0.037).

Figure 5.

STAT3 targets are enriched in human tumors containing activated STAT3. A, correlation of all identified STAT3 target genes with activated STAT3 in prostate tumors was measured by a maximum enrichment score (ESmax), and the significance was assessed using permutation testing. B, correlation of genes composing the STAT3 signature with activated STAT3 in prostate tumors was measured by an maximum enrichment score, and the significance was assessed using permutation testing. C, correlation of all identified STAT3 target genes with activated STAT3 in breast tumors was measured by an maximum enrichment score, and the significance was assessed using permutation testing. D, correlation of genes composing the STAT3 signature with activated STAT3 in breast tumors was measured by an maximum enrichment score, and the significance was assessed using permutation testing.

Figure 5.

Close modal

These results indicate that the STAT3 targets we identified in NIH3T3 cells, and specifically the STAT3 signature we culled from this set using human tumor data sets, are expressed more highly in prostate and breast tumors with STAT3 activation. This provides further evidence that these genes may be critically involved in the contribution of STAT3 to human cancer.

Signal transducer and activator of transcription 3 is required for target gene expression in human cell lines with signal transducer and activator of transcription 3 activation. Although these results strongly suggest that STAT3 activation is responsible for the expression of these genes in tumors, we wished to directly address whether there is a causal relationship between STAT3 activation and expression of the STAT3 signature genes in human tumor cells. To do this, we used MDA-MB-231 cells, a human breast cancer cell line that has been shown previously to have phosphorylated STAT3. If activated STAT3 was driving expression of these target genes, then interrupting this signaling would lead to decreased expression of these genes.

We used virally delivered small interfering RNA (pRS) to suppress the levels of STAT3 protein. We screened pools stably expressing small interfering RNA targeting STAT3 and identified several pools with near complete suppression of STAT3 (Fig. 6A). As expected, these cells also had a reduction in the level of pSTAT3, suggesting that STAT3-dependent transcription would be abrogated. We isolated RNA from MDA-MB-231 cells stably expressing empty vector (pRS) or cells in which STAT3 was knocked down (pRS/STAT3) and analyzed the expression of STAT3 target genes by quantitative RT-PCR. The expression of all of the targets examined was decreased in cells lacking STAT3 (Fig. 6B). This suggests that STAT3 activation is required for the expression of these genes in human tumor cells that have acquired STAT3 activation during the course of transformation. Furthermore, cells in which STAT3 expression was knocked down grew more slowly than control cells and eventually lost knockdown of STAT3 (data not shown). This suggests that STAT3, and the expression of these STAT3 targets, is critical for the growth and survival of this breast cancer cell line.

Figure 6.

STAT3 inhibition decreases target gene expression in human breast cancer cells. A, MDA-MB-231 cells stably expressing empty vector (pRS) or a vector expressing small interfering RNA targeting STAT3 (pRS/STAT3). A Western blot was done on whole cell lysates using antibodies to STAT3 (top), pSTAT3 (middle), or tubulin (bottom). The experiment was done four times. A representative experiment is shown. B, total RNA was harvested and reverse transcribed, and the abundance of each transcript was determined by real-time PCR. Values are normalized to an internal control gene, β-actin, and expressed as relative expression compared with MDA-MB-231 pRS cells. The average of four independent experiments is shown.

Figure 6.

Close modal

Discussion

We have identified several targets of the transcription factor STAT3. By inducibly expressing a constitutively active form of STAT3, we were able to identify genes whose expression is regulated by STAT3 at very early time points. This excludes the identification of genes whose expression is indirectly affected by STAT3 activation, for instance through induction of intermediate transcription factors or through widespread changes in cellular conditions. Evaluating the role of these STAT3 targets in processes central to STAT3-directed transformation could be achieved by individually modulating the function of each STAT3 target and examining attendant changes in cellular phenotype. This approach is infeasible due both to the large number of targets identified and to the likelihood that several targets collaborate to produce a phenotype. In vitro models of transformation thus have limited utility in discerning which among hundreds of STAT3 targets are the crucial effectors of oncogenesis. We employed an alternate strategy based on the notion that human tumors contain in their gene expression profile information about the mechanism by which they became transformed. Specifically, we reasoned that if activated STAT3 contributes to a tumor's phenotype, there should exist in that tumor a genetic signature for STAT3 activation and that this signature should comprise those target genes that mediate the functions of STAT3 in oncogenesis.

We used two methods to identify a STAT3 expression signature, defined as a group of STAT3 targets that are highly coexpressed with each other in human tumors. We first did unsupervised hierarchical clustering across a range of diverse tumors. This led to the identification of a cluster of genes with high expression in central nervous system tumors, leukemias, and prostate tumors—all tumor types in which STAT3 activation has been observed. Next, to assess the statistical significance of this coexpression, we used a Kolmogorov-Smirnov test to measure the degree to which each STAT3 target was coexpressed with the remaining targets in several independent data sets. In this manner, we identified a signature for STAT3 activation. Most of the genes present in the signature are clustered together by hierarchical clustering, indicating that these two approaches yield similar results (see Supplementary Fig. S2). The STAT3 targets were also found to be more highly expressed in breast and prostate tumors with activated STAT3. This strongly suggests that the genes identified in vitro are regulated by STAT3 in vivo as well. The approach of combining a tractable in vitro system to identify direct transcriptional relationships, with human tumor samples and expression data sets that are most relevant to disease, is an especially informative method of studying the mechanism of oncogenic transcription factors.

Many of the STAT3 targets we identified, and specifically those present in the STAT3 signature, are themselves transcription factors. These include junB, egr1, KLF4, bcl-6, and NFIL3. Many of these have independently been implicated in oncogenesis likely through their regulation of proliferation (e.g., junB and egr1), survival (e.g., NFIL3), or differentiation (e.g., bcl-6; refs. 28–30). The large number of transcription factors also suggests that activated STAT3 initiates a widespread change in gene expression extending far beyond its direct, immediate targets, implying that the steady-state transcriptional profile of a cell in which STAT3 is chronically active might be profoundly different from a cell that lacks STAT3 activation. Another STAT3 target we identified, VEGF, has been well characterized as contributing to tumor progression through promoting angiogenesis (31). Further, STAT3 has been shown to activate VEGF expression in tumor cells (32). Our findings confirm that VEGF is a STAT3 target and, importantly, show association between STAT3 activation and VEGF expression in human tumors, providing further evidence for the role of STAT3 in promoting angiogenesis through VEGF. The protease calpain has been shown recently to be involved in cellular transformation and migration (33, 34), especially mediated by v-src. We show that calpain is a STAT3 target that is associated with STAT3 activation in human tumors, thus providing another potential mechanism by which STAT3 contributes to oncogenesis.

In conclusion, we have identified several transcriptional targets of STAT3, a subset of which are significantly coexpressed in human tumors and correlated with STAT3 activation in breast and prostate cancer. These target genes may mediate the role of STAT3 in these tumors and may provide novel targets for therapeutic intervention in tumors with activated STAT3.

Current address for P.G. Febbo: Departments of Medicine and Molecular Genetics and Microbiology, Duke Institute for Genome Sciences and Policy, Duke University, Durham, NC 27710.

Acknowledgments

Grant support: NIH grant CA79547, Albert J. Ryan Foundation, Friends of the Dana-Farber Cancer Institute, the family and friends of Warren R. Jacobson, and Harvard/DFCI SPORE in Breast 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 Aravind Subramanian for statistical and computational help with performing Kolmogorov-Smirnov analyses and Nandita Bhattacharjee for staining the tissue microarrays.

References

Frank DA. STAT signaling in the pathogenesis and treatment of cancer.

Mol Med

1999

;

432

–56.

Turkson J, Bowman T, Garcia R, Caldenhoven E, de Groot RP, Jove R. Stat3 activation by src induces specific gene regulation and is required for cell transformation.

Mol Cell Biol

1998

;

2545

–52.

Bromberg JF, Horvath CM, Besser D, Lathem WW, Darnell JE Jr. Stat3 activation is required for cellular transformation by v-src.

Mol Cell Biol

1998

;

2553

–8.

Bromberg JF, Wrzeszczynska MH, Devgan G, et al. Stat3 as an oncogene.

Cell

1999

;

295

–303.

Puthier D, Bataille R, Amiot M. IL-6 up-regulates mcl-1 in human myeloma cells through JAK/STAT rather than ras/MAP kinase pathway.

Eur J Immunol

1999

;

3945

–50.

Karni R, Jove R, Levitzki A. Inhibition of pp60c-Src reduces Bcl-XL expression and reverses the transformed phenotype of cells overexpressing EGF and HER-2 receptors.

Oncogene

1999

;

4654

–62.

Ivanov VN, Bhoumik A, Krasilnikov M, et al. Cooperation between STAT3 and c-jun suppresses Fas transcription.

Mol Cell

2001

;

517

–28.

Fukada T, Hibi M, Yamanaka Y, et al. Two signals are necessary for cell proliferation induced by a cytokine receptor gp130: involvement of STAT3 in anti-apoptosis.

Immunity

1996

;

449

–60.

Sinibaldi D, Wharton W, Turkson J, Bowman T, Pledger WJ, Jove R. Induction of p21WAF1/CIP1 and cyclin D1 expression by the Src oncoprotein in mouse fibroblasts: role of activated STAT3 signaling.

Oncogene

2000

;

5419

–27.

Bowman T, Broome MA, Sinibaldi D, et al. Stat3-mediated Myc expression is required for Src transformation and PDGF-induced mitogenesis.

Proc Natl Acad Sci U S A

2001

;

7319

–24.

Kojima H, Nakajima K, Hirano T. IL-6-inducible complexes on an IL-6 response element of the junB promoter contain Stat3 and 36 kDa CRE-like site binding protein(s).

Oncogene

1996

;

547

–54.

Yang E, Lerner L, Besser D, Darnell JE Jr. Independent and cooperative activation of chromosomal c-fos promoter by STAT3.

J Biol Chem

2003

;

278

15794

–9.

Kiuchi N, Nakajima K, Ichiba M, et al. STAT3 is required for the gp130-mediated full activation of the c-myc gene.

J Exp Med

1999

;

189

–73.

Li C, Wong WH. Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection.

Proc Natl Acad Sci U S A

2001

;

–6.

Lamb J, Ramaswamy S, Ford HL, et al. A mechanism of cyclin D1 action encoded in the patterns of gene expression in human cancer.

Cell

2003

;

114

323

–34.

Mootha VK, Lindgren CM, Eriksson KF, et al. PGC-1α-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes.

Nat Genet

2003

;

267

–73.

Brummelkamp TR, Bernards R, Agami R. Stable suppression of tumorigenicity by virus-mediated RNA interference.

Cancer Cell

2002

;

243

–7.

Ory DS, Neugeboren BA, Mulligan RC. A stable human-derived packaging cell line for production of high titer retrovirus/vesicular stomatitis virus G pseudotypes.

Proc Natl Acad Sci U S A

1996

;

11400

–6.

Besser D, Bromberg JF, Darnell JE Jr, Hanafusa H. A single amino acid substitution in the v-Eyk intracellular domain results in activation of Stat3 and enhances cellular transformation.

Mol Cell Biol

1999

;

1401

–9.

Brocke-Heidrich K, Kretzschmar AK, Pfeifer G, et al. Interleukin-6-dependent gene expression profiles in multiple myeloma INA-6 cells reveal a Bcl-2 family-independent survival pathway closely associated with Stat3 activation.

Blood

2004

;

103

242

–51.

Ramaswamy S, Tamayo P, Rifkin R, et al. Multiclass cancer diagnosis using tumor gene expression signatures.

Proc Natl Acad Sci U S A

2001

;

15149

–54.

Mora LB, Buettner R, Seigne J, et al. Constitutive activation of Stat3 in human prostate tumors and cell lines: direct inhibition of Stat3 signaling induces apoptosis of prostate cancer cells.

Cancer Res

2002

;

6659

–66.

Lin TS, Mahajan S, Frank DA. STAT signaling in the pathogenesis and treatment of leukemias.

Oncogene

2000

;

2496

–504.

Rahaman SO, Harbor PC, Chernova O, Barnett GH, Vogelbaum MA, Haque SJ. Inhibition of constitutively active Stat3 suppresses proliferation and induces apoptosis in glioblastoma multiforme cells.

Oncogene

2002

;

8404

–13.

Armstrong SA, Staunton JE, Silverman LB, et al. MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia.

Nat Genet

2002

;

–7.

Singh D, Febbo PG, Ross K, et al. Gene expression correlates of clinical prostate cancer behavior.

Cancer Cell

2002

;

203

–9.

Dechow TN, Pedranzini L, Leitch A, et al. Requirement of matrix metalloproteinase-9 for the transformation of human mammary epithelial cells by Stat3-C.

Proc Natl Acad Sci U S A

2004

;

101

10602

–7.

Virolle T, Krones-Herzig A, Baron V, De Gregorio G, Adamson ED, Mercola D. Egr1 promotes growth and survival of prostate cancer cells. Identification of novel Egr1 target genes.

J Biol Chem

2003

;

278

11802

–10.

Kuribara R, Kinoshita T, Miyajima A, et al. Two distinct interleukin-3-mediated signal pathways, Ras-NFIL3 (E4BP4) and Bcl-xL, regulate the survival of murine pro-B lymphocytes.

Mol Cell Biol

1999

;

2754

–62.

Albagli-Curiel O. Ambivalent role of BCL6 in cell survival and transformation.

Oncogene

2003

;

507

–16.

Dvorak HF. Vascular permeability factor/vascular endothelial growth factor: a critical cytokine in tumor angiogenesis and a potential target for diagnosis and therapy.

J Clin Oncol

2002

;

4368

–80.

Niu G, Wright KL, Huang M, et al. Constitutive Stat3 activity up-regulates VEGF expression and tumor angiogenesis.

Oncogene

2002

;

2000

–8.

Carragher NO, Frame MC. Calpain: a role in cell transformation and migration.

Int J Biochem Cell Biol

2002

;

1539

–43.

Carragher NO, Westhoff MA, Riley D, et al. v-Src-induced modulation of the calpain-calpastatin proteolytic system regulates transformation.

Mol Cell Biol

2002

;

257

–69.

2005

Supplementary data

612 Views

160 Web of Science

Identification of a Genetic Signature of Activated Signal Transducer and Activator of Transcription 3 in Human Tumors (original) (raw)

Abstract

Introduction

Materials and Methods

Results

Discussion

Acknowledgments

References

Supplementary data

Citing articles via

Email alerts