Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Brain Behav Immun. Author manuscript; available in PMC 2010 February 1.
Published in final edited form as:
PMCID: PMC2677379

Depression, Social Support, and Beta-Adrenergic Transcription Control in Human Ovarian Cancer


Motivated by previous indications that beta-adrenergic signaling can regulate tumor cell gene expression in model systems, we sought to determine whether similar dynamics occur in primary human ovarian cancer. DNA microarray analyses of 10 ovarian carcinomas identified 266 human transcripts that were differentially expressed in tumors from patients with elevated biobehavioral risk factors (high depressive symptoms and low social support) relative to grade-and stage-matched tumors from low-risk patients. Promoter-based bioinformatic analyses confirmed increased activity of several beta-adrenergically-linked transcription control pathways, including CREB/ATF, NF-κB/Rel, STAT, and Ets family transcription factors. Consistent with increased beta-adrenergic signaling, high biobehavioral risk patients also showed increased intra-tumor concentrations of norepinephrine (but no difference in plasma norepinephrine). These data show that genome-wide transcriptional profiles are significantly altered in patients with high behavioral risk profiles, and they identify beta-adrenergic signal transduction as a likely mediator of those differences.

Keywords: Ovarian Cancer, Sympathetic Nervous System, Gene Expression, Transcription Factor, Beta-Adrenergic Receptor


Biobehavioral factors such as stress, depression, and social support have long been suspected to influence cancer onset and disease progression, but the molecular mechanisms of these effects are just beginning to be elucidated (Antoni, Lutgendorf et al. 2006). Recent findings from laboratory models suggest that biobehavioral processes can influence cancer biology via direct effects of the neuroendocrine system on the functional activity of tumor cells (Thaker, Han et al. 2006). Many solid epithelial tumors express receptors for neuroendocrine mediators from the sympathetic nervous system (SNS) (the catecholamines, epinephrine and norepinephrine) (Mitra and Carraway 1999; Lutgendorf, Cole et al. 2003; Nagmani, Pasco et al. 2003; Palm, Lang et al. 2006; Thaker, Han et al. 2006; Ramos-Jimenez, Soria-Jasso et al. 2007; Sastry, Karpova et al. 2007) or the hypothalamus-pituitary-adrenal (HPA) axis (the glucocorticoid, cortisol)(Wu, Chaudhuri et al. 2004). Analyses carried out in vitro and in mouse xenograft models have shown that neuroendocrine receptors regulate several biological processes involved in cancer metastasis and disease progression, including angiogenesis (Lutgendorf, Cole et al. 2003; Thaker, Han et al. 2006), tissue invasion (Sood, Bhatty et al. 2006; Yang, Sood et al. 2006), cell motility (Palm, Lang et al. 2006), and programmed cell death (Wu, Chaudhuri et al. 2004; Sastry, Karpova et al. 2007). Analyses of mouse xenograft models also show that experimentally imposed behavioral stress can increase the growth and metastatic activity of implanted human carcinoma cells via direct neuroendocrine regulation of tumor cell biology (Thaker, Han et al. 2006). However, it is unclear whether such dynamics occur in human clinical cancer because no studies have directly examined the biological activity of primary human tumors in relationship to biobehavioral characteristics of the host.

In model systems, neuroendocrine factors regulate cancer cell biology in large part by altering gene expression. For example, catecholamines can enhance the expression of several genes involved in angiogenesis (e.g., VEGF, IL6) and tissue invasion (MMP2, MMP9) in ovarian and asopharyngeal carcinoma cells (Lutgendorf, Cole et al. 2003; Sood, Bhatty et al. 2006; Yang, Sood et al. 2006; Nilsson, Armaiz-Pena et al. 2007). These effects are mediated by beta-adrenergic receptors, and subsequent activation of the cyclic 3′,5′-adenosine monophosphate / protein kinase A (cAMP/ PKA) signaling pathway. PKA regulates gene expression by phosphorylating multiple transcription factors, including members of the cAMP response element binding protein / activating transcription factor (CREB/ATF) family (Montminy 1997). Under certain circumstances, PKA can also cross-regulate activity of the pro-inflammatory NF-κB/Rel family of transcription factors (Shirakawa and Mizel 1989; Lahdenpohja, Henttinen et al. 1996; Zhong, Voll et al. 1998; Bierhaus, Wolf et al. 2003; Richlin, Arevalo et al. 2004), the pro-inflammatory STAT family of transcription factors (Landen, Lin et al. 2007), and the growth-promoting Ets transcription factors (Janknecht, Monte et al. 1996; Vossler, Yao et al. 1997; Luttrell, Ferguson et al. 1999; Wu and Janknecht 2002), providing multiple signaling pathways for functional genomic regulation by catecholamines.

In the present study, we sought to determine whether beta-adrenergic signaling might play a role in mediating the relationship between behavioral risk factors and gene expression in primary ovarian carcinomas. This hypothesis was motivated by previous observations that, 1.) the catecholamine ligands of beta-adrenergic receptors are increased in people with biobehavioral risk factors such as depression or low social support (Esler, Turbott et al. 1982; Seeman, Berkman et al. 1994; Veith, Lewis et al. 1994; Light, Kothandapani et al. 1998; Miller, Cohen et al. 1999; Hughes, Watkins et al. 2004; Hamer, Tanaka et al. 2007), 2.) ovarian cancer patients with those risk factors show elevated circulating levels of IL-6 and VEGF (Lutgendorf, Johnsen et al. 2002; Costanzo, Lutgendorf et al. 2005), and, 3.) beta-adrenergic signaling can enhance the expression of both IL6 and VEGFA genes in model systems (Lutgendorf, Cole et al. 2003; Thaker, Han et al. 2006; Nilsson, Armaiz-Pena et al. 2007). To determine whether beta-adrenergic regulation of gene expression also occurs in primary ovarian cancer, we profiled genome-wide transcriptional profiles in tumor tissues from patients with high biobehavioral risk profiles (high depressive symptoms and low social support) vs. low risk profiles (low depression and high social support), and used a hypothesis-testing bioinformatic strategy to assess CREB/ATF transcriptional signaling (Cole, Yan et al. 2005). We also tested secondary hypotheses that other transcription control pathways known to be activated by beta-adrenergic signaling, such as NF-κB/Rel, STAT, and Ets transcription factors, might also show increased activity in tumors from patients with high biobehavioral risk profiles.

Materials and methods


Tumor tissue and psychosocial data were collected from 20 patients undergoing primary surgical resection of ovarian carcinoma. These patients were sampled from a larger series of 60 adult women who were diagnosed with ovarian epithelial cancer, peritoneal cancer, or cancer of the fallopian tube, and were undergoing treatment at the University of Iowa (n = 53) or the University of Miami (n = 7). Patient samples were selected for analysis based on high vs. low biobehavioral risk profiles defined by depressive symptoms and social support (as detailed below). The only exclusion criteria were the presence of non-epithelial ovarian tumors, metastases to the ovaries from other organs, previous cancer diagnosis, regular use of systemic steroid medication in the last month, presence of a co-morbid condition with known effects on the immune system (e.g. autoimmune diseases), or inability to accurately answer questions (e.g. dementia).


Participants completed psychosocial and background measures one to seven days prior to surgery. Resected tumor tissues were acquired for microarray gene expression analysis as described below, and each tumor was classified by pathology according to grade, stage (FIGO classification), and histologic subtype (serous / mucinous, papillary / endometroid, carcinoma / adenocarcinoma). Individual tumor tissues from high- vs. low-risk patients were matched based on stage, grade, and histological subtype prior to genome-wide expression analysis. All procedures were approved by Institutional Review Boards at the University of Iowa, University of Miami, and University of California at Los Angeles.

Depression, social support, and background characteristics

Depressive symptoms were assessed by the Center for Epidemiologic Studies Depression Scale (CESD) (Radloff 1977). Social support was assessed by the Attachment subscale of the Social Provisions Scale (SPS-Attachment) (Cutrona and Russell 1987). Based on an a priori hypothesis derived from previous studies of biobehavioral risk factors in ovarian cancer (Costanzo, Lutgendorf et al. 2005; Lutgendorf, Sood et al. 2005), high psychosocial risk was defined by the presence of a CESD score ≥ 16 and SPS-Attachment score ≤ 15 (the median value). The Beck Depression Inventory (BDI) was also collected for comparison with other published studies, but the CESD served as the primary measure of depression. Background demographic characteristics (age, ethnicity, and socio-economic status) were assessed as potential confounders.

Tumor tissue gene expression

.1g tumor fragments were excised from tissue samples maintained at -70°C, homogenized in 600 μL of RNA-stabilizing lysis buffer (RNeasy RLT; Qiagen, Valencia CA), and supplemented with 400 μL of RNAse-free water (Qiagen) to produce a 1 ml nucleic acid lysate. 1 ml Qiazol reagent (Qiagen) and 300 μL chloroform were added, and lysates were centrifuged for 5 min at 1500 rpm in a 15 mL centrifuge tube maintained at 4°C. Aqueous phase products were mixed with 1 volume of 70% EtOH and applied to a RNeasy Mini spin column (Qiagen). Total RNA was extracted and treated with RNAse-free DNAse (Qiagen) following the manufacturer's protocol (supplemented by one additional wash in RW1 reagent and one additional wash in RPE reagent to remove excess salt). RNA purity and integrity were assessed using an Agilent 2100 bioanalyzer (Agilent Technologies, Santa Clara CA), and genome-wide transcription profiling was carried out using Affymetrix U133A high density oligonucleotide arrays (Affymetrix, Santa Clara CA) in the UCLA DNA Microarray Core, as previously described (Cole, Galic et al. 2003; Cole, Yan et al. 2005). Low level gene expression was quantified by Robust Multiarray Averaging (Bolstad, Irizarry et al. 2003), and differentially expressed genes were identified by a ≥ 2-fold difference in mean expression level (corresponding to a 5% False Discovery Rate) (Benjamini and Hochberg 1995; Cole, Galic et al. 2003) in paired comparisons of grade- and stage-matched pairs of tumors from high vs. low biobehavioral risk patients. Gene expression data are deposited in the National Center for Biotechnology Information Gene Expression Omnibus (GEO Series GSE9116).

RT-PCR verification

Differential gene expression was confirmed by quantitative real-time RT-PCR using commercial TaqMan Gene Expression Assays (Applied Biosystems, Foster City CA) and Quantitect Probe RT-PCR enzymes (Qiagen, Valencia CA), in 25 μL reactions carried out according to the manufacturer's specified 1-step thermal cycling protocol (60°C annealing) on an iCycler instrument (BioRad, Hercules CA). All reactions were carried out in triplicate, with analyte threshold cycle numbers normalized to GAPDH threshold cycle numbers prior to analysis.

Transcription control pathway analysis

The primary hypothesis regarding CREB/ATF transcription factors and secondary hypotheses regarding NF-κB/Rel and Ets family transcription factors were tested using a 2-sample variant of the Transcription Element Listening System (TELiS) (Cole 2005; Cole, Yan et al. 2005). TELiS contains data on the prevalence of 192 vertebrate transcription factor-binding motifs (TFBMs) from the TRANSFAC database (Wingender, Dietze et al. 1996) in the promoters of all genes assayed by the Affymetrix U133A high-density oligonucleotide array. TELiS differential expression analyses monitor the functional activity of the transcription factors binding to each TFBM by defining putative transcription factor-responsive genes (TFRGs) a priori based on promoter DNA sequence information, and then testing for empirical differential expression of those TFRGs using established statistical approaches (Cole, Yan et al. 2005). In the present study, activation of the CREB/ATF transcription control pathway was assessed by genes bearing the V$CREBP1_Q2 motif, activation of the NF-κB/Rel factors cRel and RelA were assessed by V$CREL_01 and V$NFKAPPAB65_01 motifs, activation of STAT family transcription factors was assessed by V$STAT_01, and activation of the prototypic Ets family transcription factor ELK1 was assessed by V$ELK1_01. Primary analyses utilized default motif detection parameters (Cole, Yan et al. 2005), and sensitivity analyses examined parametric variations of promoter length (-300 bp relative to RefSeq gene transcription start site, -600 bp, and -1000 bp to +200) and TFBM match stringency (MatSim = .80, .90, .95) (Cole, Yan et al. 2005).

Norepinephrine concentration

Norepinephrine levels in tumor and plasma extracts were determined by high performance liquid chromatography with electrochemical detection (HPLC-ED) as previously described (Hoffman, Sinkey et al. 2002). Pre-surgical blood samples were collected between 6:30 and 11:30 AM (average 8:20 AM, an average 3 hrs before tumor tissue capture), drawn into chilled EDTA Vacutainers (BD Biosciences, Franklin Lakes, NJ), and maintained on ice prior to centrifugation, separation of the plasma, and storage at -80°C until the time of extraction and assay. Plasma catecholamines were absorbed onto activated alumina at pH 8.6, washed, and eluted with dilute acid prior to injection onto a reverse-phase column to separate the individual catecholamines. Detection was accomplished by the use of a Coulochem II Dual Potentiostat Electrochemical Detector (ESA, Chelmsford, MA.). A calibration curve using “blank” human plasma (dialyzed to remove endogenous catecholamines) and linear regression analysis was used to determine sample plasma norepinephrine concentration. The calibration curve used the internal standard (DHBA, 3,4 dihyroxybenylamine) method which calculates an average response factor using peak areas and concentrations of added catecholamine in relation to the peak area of the internal standard. The interassay and intra-assay coefficients of variation were 3.4 and 3.1%, respectively, and the lower limit of detection was 25 pg / ml.

Dissected 2-3 mm3 tumor tissue samples were snap frozen in liquid nitrogen, pulverized using liquid nitrogen, and held on dry ice. The tissue powder was weighed and briefly homogenized (Fisher PowerGen 500 and sawtooth generator) in 5 ml 0.4 M perchloric acid containing 6 mM glutathione. Following centrifugation of the homogenate (1000 × g, 30 min), the supernatant was neutralized with 5 ml of TRIS buffer (1.5 M, pH 8.5) and catecholamines were absorbed onto acid washed alumina (100 mg). The alumina was washed twice with water and catecholamines were eluted with 1 ml 0.05 M perchloric acid containing 0.1 mM sodium metabisulfite. After microfiltration, the eluate was diluted in 4% acetic acid, and catecholamines were resolved using an Aquasil C18 4 μm (100 × 4.6 mm) HPLC column (Thermo Electron Corp., Bellefonte, PA), followed by electrochemical detection as described above. Chromatograph peak areas for norepinephrine and epinephrine were compared to the average peak areas determined from the injection of 100 pg pure standard, and corrected for extract dilutions and tissue wet weights.

Statistical analyses

Group differences in demographic and clinical parameters were tested using χ2 analysis (categorical variables) or paired t-tests (continuous variables) (Miller 1986). Analyses of promoter TFBM distributions tested for differential expression was using an independent sample t-test with Welch's correction for heteroscedasticity (Miller 1986). To confirm the statistical significance of differential TFBM distributions, we carried out randomization tests as previously described (Cole, Galic et al. 2003; Cole, Hawkley et al. 2007). Briefly, the total set of 10 samples was randomly partitioned into two groups of 5 samples, differentially expressed genes were identified based on ≥ 2-fold difference in mean expression level, and the resulting difference in the frequency of target TFBMs was quantified. This process was repeated 10,000 times to estimate the sampling distribution of differential TFBM distribution under the null hypothesis of random population sampling. The observed asymmetry in TFBM distributions in low vs. high biobehavioral risk patients was compared to this null hypothesis sampling distribution to estimate the tail probability of Type I (false-positive) errors (Bratley, Fox et al. 1983).


Tumor characteristics

Twenty ovarian carcinoma tissues were sampled from a series of 60 women undergoing primary surgical resection of Stage IA-IIIC ovarian cancer at two gynecologic oncology practices (Table 1). The majority of analyzed patients were diagnosed with advanced disease (55% stage III, 10% stage II, 35% stage I) and poorly differentiated tumor histology (55% grade 3, 15% grade 2, 25% grade 1). Participants ranged in age from 35 to 87 years (median = 61), 95% were white (1 African American), and they reported household incomes ranging from < $5,000 per year to > $80,000 per year (median = $30,000-$40,000). Depressive symptoms and social support were measured at pre-surgical clinic visits occurring 1-7 days prior to tumor resection. These variables showed a strong inverse correlation (r = -.43, p = .0015 in the total sample of 60; r = -.91, p < .0001 in the 20 individuals analyzed), and were therefore treated as two indicators of a common underlying biobehavioral risk cluster. Using established threshold values of CESD ≥ 16 and SPS-Attachment ≤ 15, 11 participants (42%) were identified by pre-surgical psychometric assessments as showing high levels of biobehavioral risk factors. Across these high- and low-risk strata, 5 pairs of tumors could be matched on common grade, stage, and histologic subtype, and were thus selected for microarray gene expression analysis (Table 1). These 10 tissue samples came from patients who were representative of the total sample on all assessed demographic and disease parameters (difference from sample-wide average, all p > .50). All analyzed tumors were grade 3 serous carcinomas / adenocarcinomas, recovered at stages ranging from IA to IIIC (80% IIIC). The selected high- and low-risk patients did not differ on any of the demographic dimensions assessed, or in tumor grade or stage (all p > .30). The only measured properties that differed significantly across groups were levels of depression (CESD, p = .0017; BDI, p = .0026) and social support (SPS-Attachment, p = .0129), as induced by the biobehavioral risk stratification process.

Table 1
Patient characteristics

Affymetrix U133A high-density oligonucleotide arrays were used to survey the expression of 22,283 human transcripts in ovarian tumor samples. Expression of 266 transcripts differed by more than 2-fold across high- vs. low-risk groups, including 220 transcripts up-regulated by at least 100% in tissues from high-risk patients, and 46 transcripts down-regulated by more than 50% in tissues from high-risk patients (Figure 1). Differentially expressed genes are listed in Supplementary Data (Table S1) and available on-line at 25 transcripts identified as differentially expressed in microarray analyses were confirmed by quantitative RT-PCR (Table S2, or on-line at Significant differential expression was confirmed for 23 of the 25 assayed transcripts (92%), with the magnitude of difference averaging 9.6-fold across the entire sample of 25, and p-values averaging .014. Because our primary interest focuses on testing the a priori hypothesis that upstream beta-adrenoreceptor-linked transcription control pathways drive the observed patterns of differential gene expression (rather than on any a priori hypothesis about the specific nature of differentially expressed genes), no descriptive analysis of gene functional characteristics was conducted.

Figure 1
Differentially expressed genes in ovarian cancers from patients with high vs. low biobehavioral risk profiles

Primary hypothesis: CREB/ATF-mediated transcription

TELiS bioinformatics analysis was applied to test this study's central hypothesis that genes up-regulated in ovarian cancers from women with high biobehavioral risk factors would show over-representation of targets of the CREB/ATF transcription control pathway, which mediates beta-adrenergic response to catecholamines through the cAMP/PKA signaling pathway. Results showed a > 3-fold elevation in the prevalence of CREB/ATF transcription factor-binding motifs (TFBMs) within promoters of genes up-regulated in tumors from patients with a high-risk biobehvioral profile (TRANSFAC V$CREBP1_Q2 motif: average .000 ± standard error .000 sites/promoter for genes over-expressed in tumors from low-risk individuals vs. .070 ± .026 in genes over-expressed in high-risk tumors; difference p = .0071 by t-test) (Figure 2a). Sensitivitiy analyses utilizing parametric variations of promoter length (-300 bp, -600 bp, -1000 to +200 bp) and TFBM match stringency (MatSim = .80, .90, .95) yielded similar results, with an average 2.35-fold (± .38) difference across all 9 parametric combinations (p = .0003).

Figure 2
Activity of beta-adrenoreceptor-linked transcription control pathways

Secondary hypotheses: NF-κB/Rel-, STAT-, and Ets-mediated transcription

Studies in in vitro model systems suggest that beta-adrenergic signaling can also activate NF-κB/Rel-family transcription factors, although these effects are variable across cell type and experimental conditions (Shirakawa and Mizel 1989; Lahdenpohja, Henttinen et al. 1996; Zhong, Voll et al. 1998; Bierhaus, Wolf et al. 2003; Richlin, Arevalo et al. 2004). In secondary analyses examining the distribution of cRel (cRel/p50 heterodimer) and RelA (p65/p50 heterodimer) promoter motifs, the prevalence of V$CREL_01 TFBMs was 3.7-fold higher in genes over-expressed in tumors from high-risk patients relative to low-risk patients (mean = .057 ± .040 in low risk vs. .210 ± .040 in high risk; difference, p = .0076) (Figure 2b). Similar results emerged across parametric variations of promoter length and scan stringency (mean = 2.80-fold ± .58 across all 9 combinations; p = .0014). For the RelA variant of NF-κB, TRANSFAC V$NFKAPPAB65_01 TFBMs were found at a 3.18-fold heigher prevalence in promoters of genes over-expressed in tumors from high-risk patients vs. low-risk patients (mean = .091 ± .026 in high risk vs. .029 ± .029 in low risk). This difference failed to reach statistical significance in primary analyses (p = .1103), but it was significant on average across all 9 parametric combinations of promoter length and TFBM match stringency (mean difference = 4.65-fold ± 1.73; p = .0275).

Analyses of ovarian carcinoma cells in vitro also indicate that beta-adrenergic receptors can activate the pro-inflammatory transcription factor STAT3 (Landen, Lin et al. 2007). TELiS analyses confirmed increased STAT transcripitonal activity in primary ovarian carcinoma cells by revealing a 4.7-fold higher prevalence of the V$STAT_01 TFBMs in promoters of genes up-regulated in tumors from high-risk patients (mean = .029 ± .029 in low risk vs. .133 ± .030 in high risk; difference, p = .0134) (Figure 2c). Parametric sensitivitiy analyses yielded consistent results, with an average 2.20-fold (± .38) increase in STAT activity across all 9 parametric combinations (p = .0141).

Data from in vitro models suggests that beta-adrenergic receptors can also activate the Ras / MAP kinase signaling pathway (Luttrell, Ferguson et al. 1999), and we therefore assessed activity of Ets family transcription factors that mediate Ras / MAP kinase-driven gene transcription. Analyses of the prototypic Ets factor, ELK1, revealed a significantly higher prevalence of the V$ELK1_01 TFBM in promoters of genes up-regulated in tumors from high-risk patients (mean = .000 ± .000 in low risk vs. .028 ± .014 in high risk; difference, p = .0451) (Figure 2d). Due to the high intrinsic stringency of the V$ELK1_01 nucleotide motif, this TFBM yielded non-zero prevalence frequencies for only 6 of the 9 total parametric combinations of promoter length and TFBM match stringency evaluated. However, among the 6 evaluable datasets, ELK1 TFBMs were approximately 3-fold more prevalent among the promoters of genes up-regulated in high-risk tumors relative to those in low-risk tumors (mean difference = .023 ± .007, p = .0081).

Joint relationships of beta-adrenergic transcription control pathways

To verify the accuracy of univariate statistical analyses and assess the combined significance of multiple beta-adrenergic transcription control pathways, we carried out randomization tests gauging the likelihood of the observed effects under the null hypothesis of random sampling (Bratley, Fox et al. 1983; Cole, Galic et al. 2003; Cole, Hawkley et al. 2007). The probability of observing 266 genes showing 2-fold differential expression was highly unlikely under random sampling (p = .0119), as were the probabilities of the observed differences in activity of CREB/ATF (p = .0287), NF-κB/Rel (p = .0092), and STAT family transcription factors (p = .0213). Differential activity of Ets family transcription factors approached statistical significance (p = .0821). Results indicated that simultaneous alteration in all 4 of these beta-adrenergically-linked transcription control pathways was exceedingly unlikely under random sampling (p < .0001).

Norepineprhine levels

To identify the basis for differential activation of beta-adrenregically responsive transcription control pathways in tumors from patients with high vs. low biobehavioral risk profiles, we quantified intra-tumor concentrations of the beta-adrenergic ligand, norepinephrine, in the 10 samples analyzed by DNA microarray. Tissues from patients with a low-risk profile showed minimal intra-tumor norepinephrine (all samples ≤ 0.1 pg / mg tissue assay lower limit of detection), whereas tissues from high-risk patients showed significantly greater norepinephrine concentrations (mean = 19.5 ± 6.9 pg / mg tissue, difference from low-risk, p = .0482) (Figure 3a). Parallel analyses of plasma collected an average 2 hrs prior to tumor resection failed to indicate any significant difference in circulating norepinephrine levels (low risk = 735.3 ± 60.5 pg / ml; high risk = 667.0 ± 13.5 pg / ml; difference, p = .1068) (Figure 3b). Plasma norepinephrine levels immediately prior to surgery were not significantly correlated with intra-tumor norepineprhine (r = -.18, p = .4362). Tumor norepinephrine concentrations did show moderate magnitudes of correlation with continuous scores on the CESD (r = .73, p = .1253) and SPS-Attachment scale (r = -.50, p = .2302), but those correlations failed to reach statistical significance due to the limited sample size. Plasma norepinephrine levels were not substantially correlated with either continuous measure of biobehavioral risk (CESD: r = -.32, p = .3814; SPS-Attachment: r = .22, p = .4233).

Figure 3
Norepinephrine levels in ovarian tumors and circulating blood


Based on previous indications that beta-adrenergic signaling can regulate tumor cell gene transcription in model systems (Lutgendorf, Cole et al. 2003; Sood, Bhatty et al. 2006; Thaker, Han et al. 2006; Landen, Lin et al. 2007; Nilsson, Armaiz-Pena et al. 2007), this study sought to determine whether similar dynamics occur in the context of primary human ovarian cancer. Results show that gene expression profiles in primary human tumor tissues are systematically altered in association with patient-level biobehavioral risk factors, and promoter-based bioinformatics analyses confirmed that beta-adrenergic transcription control pathways represent key mediators of those differences. TELiS analysis of 266 promoters found to be differentially active in ovarian carcinomas from patients with high depressive symptoms and low social support showed enrichment of regulatory elements for several beta-adrenergically regulated transcription factors, including CREB/ATF, NF-κB/Rel, STAT, and Ets factors. Consistent with genomic indications of increased beta-adrenergic signaling, tumors from patients with high biobehavioral risk profiles also showed elevated tissue levels of the sympathetic neuroeffector molecule norepinephrine. Thus, the present data are consistent with findings from laboratory animal and cellular models in suggesting increased activity of the sympathetic nervous system may constitute a primary mechanism by which biobehavioral factors impact gene expression in solid epithelial tumors.

One remarkable finding of this study involves the differential regulation of norepinephrine levels in tumor tissue vs. circulating blood. Patients with high biobehavioral risk profiles showed substantial elevations in intra-tumor norepinephrine concentrations, but no significant elevation (and, in fact, slightly lower levels) in plasma. The biological basis for this tissue-specific manifestation of noradrenergic signaling is not clear, but it could potentially stem from stress-induced up-regulation of tissue innervation by sympathetic neural fibers (e.g., as recently observed in lymphoid organs) (Sloan, Capitanio et al. 2007; Sloan, Capitanio et al. 2008). One potential mechanism for such effects in the context of ovarian cancer might involve chronic elevations in circulating catecholamines during tumor development (Weiner 1992), leading to increased tumor vascularization (Thaker, Han et al. 2006), and a consequent densification of vasculature-associated sympathetic fibers. The lack of correlation between blood and tumor norepinephrine levels is not surprising, given that circulating catecholamine levels are highly variable (Weiner 1992) and blood samples were collected an average of 3 hrs prior to tumor resection. However, volatile catecholamine levels cannot explain the emergence of systematically higher norepinephrine levels in tumors from patients with high biobehavioral risk profiles. The parallel expression of RNA transcriptional alterations, which lag receptor-mediated signal transduction by 2-4 hrs in ovarian carcinoma cells (Lutgendorf, Cole et al. 2003; Thaker, Han et al. 2006; Landen, Lin et al. 2007; Nilsson, Armaiz-Pena et al. 2007), also implies that the observed differences are not likely to stem from any transient alteration in tissue norepinephrine levels during surgery. The gene expression profiles observed here likely reflect the combined effects of basal individual differences in the biology of the tumor microenvironment (e.g., as driven by differential innervation or angiogenesis) and any transient effects occurring 2-4 hrs prior to surgery (i.e., well before any transient influences on tissue norepineprhine levels at tumor resection, and contemporaneous with the plasma catecholamine determinations). The most parsimonious explanation for the observed pattern of results involves stable alterations in sympathetic signaling within the tumor microenvironment. Regardless of the mechanism of these effects, the present data suggest that biobehavioral signaling dynamics can have marked effects in solid tissue that are not observable in circulating blood.

Although the specific nature of differentially regulated genes was not a target of this study, the simultaneous up-regulation of pro-inflammatory NF-κB/Rel and STAT family transcription factors suggests that tumors from high-risk patients may show heightened inflammatory signaling in the tumor microenvironment. Inflammation is believed to play a key role in the initiation and progression of may carcinomas (Coussens and Werb 2002; de Visser, Eichten et al. 2006), including those of the ovarian epithelium (Wang, Wang et al. 2005). To the extent that increased SNS activity promotes inflammation in the tumor microenvironment, inflammatory response genes may constitute a significant molecular mediator of biobehavioral influences in cancer. Given the known link between inflammatory gene expression and biology of malignant disease (Coussens and Werb 2002; de Visser, Eichten et al. 2006), the present data suggest that patients with high depressive symptoms and low social support might show accelerated disease progression in the aftermath of surgery (e.g., shorter times to disease recurrence or mortality). That speculation would be consistent with the existing epidemiologic literature in suggesting that biobehavioral factors exert their most pronounced effects on cancer by facilitating the progression of established tumors, rather than by spurring the development of new malignancies (Antoni, Lutgendorf et al. 2006).

Given the limited number of tissues available for this analysis, the present findings should be regarded as a preliminary indication that laboratory-identified biobehavioral mechanism extend into the human clinical setting. Nevertheless, it is notable that this study identified significant differences in tumor cell transcriptional dynamics that were both predicted by laboratory models a priori, and sufficiently large in magnitude to reach statistical significance despite the limited statistical power available in this small sample. If larger replication studies continue to confirm a significant role for beta-adrenergic control of multiple transcription factors in primary human cancers, this pleiotropic signaling pathway may constitute a novel molecular target for adjunctive therapies aimed at blocking catecholamine support for tumor cell biology (e.g., beta-blockade) (Antoni, Lutgendorf et al. 2006).

This study's analytic strategy rules out several potential alternative explanations for the observed differences in transcriptional activity in patients showing high vs. low biobehavioral risk profiles, including differences in tumor histological characteristics (which were controlled by matching high- vs. low-risk tissues for grade, stage, and histological subtype), and patient demographic characteristics (which were uncorrelated with depression or social support). However, several limitations of this study must be considered when interpreting the present results. These data come from a small sample of tissues that emerged from a stringent matching procedure that intentionally controlled for differences in tumor phenotype (e.g., grade, stage, histologic subtype) in high vs. low-risk patients. Larger studies will be required to establish the generality of these findings, and to explore any potential relationship between biobehavioral influences and the tumor phenotypes that were held constant by this study. Larger samples will also be required to distinguish any unique biological correlates of depressive symptoms from those associated with low social support. In the sample analyzed here, those two psychosocial risk factors were so highly correlated as to be statistically indistinguishable. Their strong correlation is consistent with previous indications that psychosocial risk factors often come in clusters which may reflect the propagation of adverse conditions throughout multiple aspects of an individual's life (Williams, Barefoot et al. 2003; Lehman, Taylor et al. 2005; Taylor, Lehman et al. 2006).

An additional limitation of this study involves its cross-sectional design. Previous experimental studies have shown that beta-adrenergic signaling can alter gene expression and cellular function in ovarian carcinoma cells (Lutgendorf, Cole et al. 2003; Antoni, Lutgendorf et al. 2006; Sood, Bhatty et al. 2006; Thaker, Han et al. 2006; Landen, Lin et al. 2007; Nilsson, Armaiz-Pena et al. 2007; Sood, Lutgendorf et al. 2007). However, it is conceivable that a reverse causal pathway might contribute to the cross-sectional association observed here, with spontaneous heterogeneity in tumor biology influencing behavioral characteristics (e.g., via effects of pro-inflammatory cytokines on the brain, resulting in decreased social motivation and increased depressive symptoms) (Dantzer, O'Connor et al. 2008). Although this hypothesis could explain an association between behavioral characteristics and pro-inflammatory transcription factors (e.g., NF-κB/Rel and STAT), it would not account for the increased CREB/ATF or Ets factor activity observed in patients with high-risk behavioral profiles. Moreover, any spontaneous mutagenic variations in the cancer genome that occur independently of biobehavioral influences would likely manifest as correlated differences in both transcriptional regimes and tumor histological phenotypes. This study's matching-based control for tumor phenotype prevents the misattribution of phenotype-related differences to biobehavioral influences, thus reducing the likelihood of reverse causality from tumor biology to behavior. Histological matching also “controls away” any biobehavioral influences that might affect tumor characteristics prior to detection (e.g., at the level of tumor initiation or promotion), and may thereby underestimate the total influence of biobehavioral factors. However, the appearance of differential gene expression profiles with a heightened inflammatory signature at the time of tumor diagnosis and resection suggests that subsequent progression of ovarian cancer may be accelerated in patients with high depressive symptoms and low social support.

The present findings are consistent with other recent data linking biobehavioral factors to gene expression differences via signaling pathways such as the inflammation-related glucocorticoid receptor, NF-κB/Rel, and JAK/STAT transcription factors, as well as the CREB/ATF and Ets factors (Irwin, Wang et al. 2006; Cole, Hawkley et al. 2007; Miller, Chen et al. 2008). This study is novel in identifying genome-wide transcriptional correlates of biobehavioral risk factors specifically within diseased tissue. The close concordance between the transcriptional mediators identified here in diseased tissue and those identified in leukocyte “reporter cell” studies underscores the common role of systemic neuroendocrine activity in shaping global gene expression profiles in both healthy and diseased tissue. As new therapeutic strategies are developed to block adverse effects of biobehavioral risk factors, the transcription control “fingerprints” identified in these studies may provide novel molecular targets for intervention, and functional genomic biomarkers for assessing the impact of those interventioins. As such, the gene expression fingerprint of depression and low social support in primary ovarian carcinomas provides new insights into the social epidemiology of cancer (Berkman and Kawachi 2000; Hernandez and Blazer 2006; Weinstein, Vaupel et al. 2007) and suggests novel strategies for protecting cancer patients from the adverse effects of stress on the progression of malignant disease (Antoni, Lutgendorf et al. 2006; Sood, Lutgendorf et al. 2007).

Supplementary Material




We thank Heena Maiseri, Stephanie McGinn, Vanessa Lehner, Cecilia Torres, and the UCLA DNA Microarray Core for technical assistance, and David Bender, Thomas Buekers, and Paige McDonald for their thoughtful discussions. Research supported by the National Cancer Institute (R21 CA88293, R01 CA104825, R01 CA109298, R01 CA110793 and R01 CA116778).


Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.


  • Antoni MH, Lutgendorf SK, et al. The influence of bio-behavioural factors on tumour biology: pathways and mechanisms. Nat Rev Cancer. 2006;6(3):240–8. [PMC free article] [PubMed]
  • Benjamini Y, Hochberg Y. Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society, Series B. 1995;57:289–300.
  • Berkman LF, Kawachi I. Social Epidemiology. New York: Oxford University Press; 2000.
  • Bierhaus A, Wolf J, et al. A mechanism converting psychosocial stress into mononuclear cell activation. Proceedings of the National Academy of Sciences of the USA. 2003;100(4):1920–1925. [PubMed]
  • Bolstad BM, Irizarry RA, et al. A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics. 2003;19(2):185–93. [PubMed]
  • Bratley P, Fox BL, et al. A guide to simulation. New York: Springer Verlag; 1983.
  • Cole S. Transcription Element Listening System (TELiS) 2005
  • Cole SW, Galic Z, et al. Controlling false-negative errors in microarray differential expression analysis: a PRIM approach. Bioinformatics. 2003;19(14):1808–16. [PubMed]
  • Cole SW, Hawkley LC, et al. Social regulation of gene expression in human leukocytes. Genome Biology. 2007;8(9):R189. [PMC free article] [PubMed]
  • Cole SW, Yan W, et al. Expression-based monitoring of transcription factor activity: The TELiS database. Bioinformatics. 2005;21(6):803–810. [PubMed]
  • Costanzo ES, Lutgendorf SK, et al. Psychosocial factors and interleukin-6 among women with advanced ovarian cancer. Cancer. 2005;104(2):305–13. [PubMed]
  • Coussens LM, Werb Z. Inflammation and cancer. Nature. 2002;420(6917):860–7. [PMC free article] [PubMed]
  • Cutrona CE, Russell DW. The provisions of social relationships and adaptation to stress. Advances in Personal Relationships. 1987;1:37–67.
  • Dantzer R, O'Connor JC, et al. From inflammation to sickness and depression: when the immune system subjugates the brain. Nat Rev Neurosci. 2008;9(1):46–56. [PMC free article] [PubMed]
  • de Visser KE, Eichten A, et al. Paradoxical roles of the immune system during cancer development. Nat Rev Cancer. 2006;6(1):24–37. [PubMed]
  • Esler M, Turbott J, et al. The peripheral kinetics of norepinephrine in depressive illness. Arch Gen Psychiatry. 1982;39(3):295–300. [PubMed]
  • Hamer M, Tanaka G, et al. The effects of depressive symptoms on cardiovascular and catecholamine responses to the induction of depressive mood. Biol Psychol. 2007;74(1):20–5. Epub 2006 Jul 24. [PubMed]
  • Hernandez LM, Blazer DG, editors. Genes, Behavior, and the Social Environment: Moving Beyond the Nature/Nurture Debate. Washington DC: National Academies Press; 2006. [PubMed]
  • Hoffman RP, Sinkey CA, et al. Systemic and local adrenergic regulation of muscle glucose utilization during hypoglycemia in healthy subjects. Diabetes. 2002;51(3):734–42. [PubMed]
  • Hughes JW, Watkins L, et al. Depression and anxiety symptoms are related to increased 24-hour urinary norepinephrine excretion among healthy middl-aged women. Journal of Psychosomatic Research. 2004;57(4):353–358. [PubMed]
  • Irwin MR, Wang M, et al. Sleep deprivation and activation of morning levels of cellular and genomic markers of inflammation. Arch Intern Med. 2006;166(16):1756–62. [PubMed]
  • Janknecht R, Monte D, et al. The ETS-related transcription factor ERM is a nuclear target of signaling cascades involving MAPK and PKA. Oncogene. 1996;13(8):1745–54. [PubMed]
  • Lahdenpohja N, Henttinen T, et al. Activation of the protein kinase A increases the DNA-binding and transcriptional activity of c-Rel in T cells. Scand J Immunol. 1996;43(6):640–5. [PubMed]
  • Landen CN, Jr, Lin YG, et al. Neuroendocrine modulation of signal transducer and activator of transcription-3 in ovarian cancer. Cancer Res. 2007;67(21):10389–96. [PubMed]
  • Lehman BJ, Taylor SE, et al. Relation of childhood socioeconomic status and family environment to adult metabolic functioning in the CARDIA study. Psychsomatic Medicine. 2005;67:846–854. [PubMed]
  • Light KC, Kothandapani RV, et al. Enhanced cardiovascular and catecholamine responses in women with depressive symptoms. Int J Psychophysiol. 1998;28(2):157–66. [PubMed]
  • Lutgendorf SK, Cole S, et al. Stress-related mediators stimulate vascular endothelial growth factor secretion by two ovarian cancer cell lines. Clin Cancer Res. 2003;9(12):4514–21. [PubMed]
  • Lutgendorf SK, Johnsen EL, et al. Vascular endothelial growth factor and social support in patients with ovarian carcinoma. Cancer. 2002;95(4):808–15. [PubMed]
  • Lutgendorf SK, Sood AK, et al. Social support, psychological distress, and natural killer cell activity in ovarian cancer. J Clin Oncol. 2005;23(28):7105–13. [PubMed]
  • Luttrell LM, Ferguson SS, et al. Beta-arrestin-dependent formation of beta2 adrenergic receptor-Src protein kinase complexes. Science. 1999;283(5402):655–61. [PubMed]
  • Miller GE, Chen E, et al. A genomic fingerprint of chronic stress in humans: Blunted glucocorticoid and increased NF-kappaB signaling. Biological Psychiatry. 2008 In press. [PMC free article] [PubMed]
  • Miller GE, Cohen S, et al. Personality and tonic cardiovascular, neuroendocrine, and immune parameters. Brain, Behavior, and Immunity. 1999;13:109–123. [PubMed]
  • Miller RG. Beyond ANOVA: Basics of applied statistics. New York: Wiley; 1986.
  • Mitra SP, Carraway RE. Synergistic effects of neurotensin and beta-adrenergic agonist on 3,5-cyclic adenosine monophosphate accumulation and DNA synthesis in prostate cancer PC3 cells. Biochem Pharmacol. 1999;57(12):1391–7. [PubMed]
  • Montminy M. Transcriptional regulation by cyclic AMP. Annu Rev Biochem. 1997;66:807–22. [PubMed]
  • Nagmani R, Pasco DS, et al. Evaluation of beta-adrenergic receptor subtypes in the human prostate cancer cell line-LNCaP. Biochem Pharmacol. 2003;65(9):1489–94. [PubMed]
  • Nilsson MB, Armaiz-Pena G, et al. Stress hormones regulate interleukin-6 expression by human ovarian carcinoma cells through a Src-dependent mechanism. J Biol Chem. 2007;282(41):29919–26. Epub 2007 Aug 23. [PubMed]
  • Palm D, Lang K, et al. The norepinephrine-driven metastasis development of PC-3 human prostate cancer cells in BALB/c nude mice is inhibited by beta-blockers. Int J Cancer. 2006;118(11):2744–9. [PubMed]
  • Radloff LS. The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement. 1977;1:386–401.
  • Ramos-Jimenez J, Soria-Jasso LE, et al. Histamine augments beta2-adrenoceptor-induced cyclic AMP accumulation in human prostate cancer cells DU-145 independently of known histamine receptors. Biochem Pharmacol. 2007;73(6):814–23. Epub 2006 Dec 1. [PubMed]
  • Richlin VA, Arevalo JM, et al. Stress-induced enhancement of NF-kappaB DNA-binding in the peripheral blood leukocyte pool: effects of lymphocyte redistribution. Brain Behav Immun. 2004;18(3):231–7. [PubMed]
  • Sastry KS, Karpova Y, et al. Epinephrine protects cancer cells from apoptosis via activation of PKA and BAD phosphorylation. J Biol Chem. 2007;12:12. [PubMed]
  • Seeman T, Berkman LF, et al. Social ties and support and neuroendocrine function: The MacArthur studies of successful aging. Annals of Behavioral Medicine. 1994;16:95–106.
  • Shirakawa F, Mizel SB. In vitro activation and nuclear translocation of NF-kappa B catalyzed by cyclic AMP-dependent protein kinase and protein kinase C. Mol Cell Biol. 1989;9(6):2424–30. [PMC free article] [PubMed]
  • Sloan EK, Capitanio JP, et al. Stress-induced remodeling of lymphoid innervation. Brain Behav Immun. 2008;22(1):15–21. Epub 2007 Aug 13. [PMC free article] [PubMed]
  • Sloan EK, Capitanio JP, et al. Beta adrenergic regulation of lymphoid innervation and SIV gene expression. Under review 2007
  • Sood AK, Bhatty R, et al. Stress hormone-mediated invasion of ovarian cancer cells. Clin Cancer Res. 2006;12(2):369–75. [PMC free article] [PubMed]
  • Sood AK, Lutgendorf SK, et al. Neurodocrine Regulation of Cancer Progression: I. Biological Mechanisms and Clinical Relevance. In: Ader R, editor. Psychoneuroimmunology. 4th. Burlington MA: Elsevier Academic Press; 2007.
  • Taylor SE, Lehman BJ, et al. Relationship of early life stress and psychological functioning to adult C-reactive protein in the Coronary Artery Risk Development in Young Adults Study. Biological Psychiatry. 2006;60:819–824. [PubMed]
  • Thaker PH, Han LY, et al. Chronic stress promotes tumor growth and angiogenesis in a mouse model of ovarian carcinoma. Nat Med. 2006;12(8):939–44. Epub 2006 Jul 23. [PubMed]
  • Veith RC, Lewis N, et al. Sympathetic nervous system activity in major depression. Basal and desipramine-induced alterations in plasma norepinephrine kinetics. Arch Gen Psychiatry. 1994;51(5):411–22. [PubMed]
  • Vossler MR, Yao H, et al. cAMP activates MAP kinase and Elk-1 through a B-Raf- and Rap1-dependent pathway. Cell. 1997;89(1):73–82. [PubMed]
  • Wang X, Wang E, et al. Ovarian cancer, the coagulation pathway, and inflammation. J Transl Med. 2005;3:25. [PMC free article] [PubMed]
  • Weiner H. Perturbing the Organism: The Biology of Stressful Experience. Chicago: University of Chicago Press; 1992.
  • Weinstein M, Vaupel JW, et al. Biosocial surveys. Washington D. C.: The National Academies Press; 2007.
  • Williams RB, Barefoot JC, et al. Psychosocial risk factors for cardiovascular disease: more than one culprit at work. JAMA. 2003;290(16):2190–2192. [PubMed]
  • Wingender E, Dietze P, et al. TRANSFAC: a database on transcription factors and their DNA binding sites. Nucleic Acids Res. 1996;24(1):238–41. [PMC free article] [PubMed]
  • Wu J, Janknecht R. Regulation of the ETS transcription factor ER81 by the 90-kDa ribosomal S6 kinase 1 and protein kinase A. J Biol Chem. 2002;277(45):42669–79. Epub 2002 Sep 3. [PubMed]
  • Wu W, Chaudhuri S, et al. Microarray analysis reveals glucocorticoid-regulated survival genes that are associated with inhibition of apoptosis in breast epithelial cells. Cancer Res. 2004;64(5):1757–64. [PubMed]
  • Yang EV, Sood AK, et al. Norepinephrine up-regulates the expression of vascular endothelial growth factor, matrix metalloproteinase (MMP)-2, and MMP-9 in nasopharyngeal carcinoma tumor cells. Cancer Res. 2006;66(21):10357–64. [PubMed]
  • Zhong H, Voll RE, et al. Phosphorylation of NF-kappa B p65 by PKA stimulates transcriptional activity by promoting a novel bivalent interaction with the coactivator CBP/p300. Mol Cell. 1998;1(5):661–71. [PubMed]