Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery

Brain, Behavior, and Immunity xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Brain, Behavior, and Immunity journal homepage: www.elsevier.com/locate/ybrbi

Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery E. Cash a,b,c,1, S.E. Sephton b,c,⇑,1, A.B. Chagpar d, D. Spiegel e,f, W.N. Rebholz b, L.A. Zimmaro b, J.M. Tillie e, F.S. Dhabhar e,f,g,⇑,1 a

Department of Surgery, Division of Otolaryngology–HNS, University of Louisville School of Medicine, Louisville, KY, United States Department of Psychological & Brain Sciences, University of Louisville, Louisville, KY, United States James Graham Brown Cancer Center, University of Louisville, Louisville, KY, United States d The Breast Center – Smilow Cancer Hospital at Yale-New Haven, Department of Surgery, Yale University School of Medicine, New Haven, CT, United States e Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, Stanford, CA, United States f Stanford Cancer Institute, Stanford University School of Medicine, Stanford, CA, United States g Institute for Immunity, Transplantation, and Infection, Stanford Cancer Institute, Stanford University School of Medicine, Stanford, CA, United States b c

a r t i c l e

i n f o

Article history: Received 20 May 2014 Received in revised form 9 February 2015 Accepted 20 February 2015 Available online xxxx Keywords: Breast cancer Rest–activity Cortisol Rhythms Biomarker Angiogenesis Metastasis Invasion Immunosuppression Inflammation

a b s t r a c t Psychological distress, which can begin with cancer diagnosis and continue with treatment, is linked with circadian and endocrine disruption. In turn, circadian/endocrine factors are potent modulators of cancer progression. We hypothesized that circadian rest–activity rhythm disruption, distress, and diurnal cortisol rhythms would be associated with biomarkers of tumor progression in the peripheral blood of women awaiting breast cancer surgery. Breast cancer patients (n = 43) provided actigraphic data on rest–activity rhythm, cancer-specific distress (IES, POMS), saliva samples for assessment of diurnal cortisol rhythm, cortisol awakening response (CAR), and diurnal mean. Ten potential markers of tumor progression were quantified in serum samples and grouped by exploratory factor analysis. Analyses yielded three factors, which appear to include biomarkers reflecting different aspects of tumor progression. Elevated factor scores indicate both high levels and strong clustering among serum signals. Factor 1 included VEGF, MMP-9, and TGF-b; suggesting tumor invasion/immunosuppression. Factor 2 included IL-1b, TNF-a, IL-6R, MCP-1; suggesting inflammation/chemotaxis. Factor 3 included IL-6, IL-12, IFN-c; suggesting inflammation/TH1-type immunity. Hierarchical regressions adjusting age, stage and socioeconomic status examined associations of circadian, distress, and endocrine variables with these three factor scores. Patients with poor circadian coordination as measured by rest–activity rhythms had higher Factor 1 scores (R2 = .160, p = .038). Patients with elevated CAR also had higher Factor 1 scores (R2 = .293, p = .020). These relationships appeared to be driven largely by VEGF concentrations. Distress was not related to tumor-relevant biomarkers, and no other significant relationships emerged. Women with strong circadian activity rhythms showed less evidence of tumor promotion and/or progression as indicated by peripheral blood biomarkers. The study was not equipped to discern the cause of these associations. Circadian/endocrine aberrations may be a manifestation of systemic effects of aggressive tumors. Alternatively, these results raise the possibility that, among patients with active breast tumors, disruption of circadian activity rhythms and elevated CAR may facilitate tumor promotion and progression. Ó 2015 Elsevier Inc. All rights reserved.

Abbreviations: CAR, cortisol awakening response; EFA, exploratory factor analysis; ER, estrogen receptor; HER2/neu, human epidermal growth factor 2; HPA, hypothalamic–pituitary–adrenal; IES, Impact of Events Scale; IFN, interferon; IL, interleukin; MCP, monocyte chemotactic protein; MEMS, Medication Event Monitoring System; MMP, matrix metallopeptidase; POMS, Profile of Mood States; PR, progesterone receptor; SCN, suprachiasmatic nucleus; TGF, transforming growth factor; TH, Thelper; TNF, tumor necrosis factor; VEGF, vascular endothelial growth factor; WASO, wake after sleep onset. ⇑ Corresponding authors at: 2301 South Third Street, 317 Life Sciences, Louisville, KY 40292, United States. Tel.: +1 (502) 852 1166 (S.E. Sephton). 300 Pasteur Drive, MC 5135, Stanford, CA 94305, United States. Tel./fax: +1 (650) 736 8565 (F.S. Dhabhar). E-mail addresses: [email protected] (S.E. Sephton), [email protected] (F.S. Dhabhar). 1 Equal authorship contribution. http://dx.doi.org/10.1016/j.bbi.2015.02.017 0889-1591/Ó 2015 Elsevier Inc. All rights reserved.

Please cite this article in press as: Cash, E., et al. Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery. Brain Behav. Immun. (2015), http://dx.doi.org/10.1016/j.bbi.2015.02.017

2

E. Cash et al. / Brain, Behavior, and Immunity xxx (2015) xxx–xxx

1. Introduction Disruption of circadian rhythms can increase breast cancer risk and hasten disease progression. Compelling new animal, cell, and molecular research is revealing the biology that links circadian disruption with tumor incidence and growth (Escobar et al., 2012, 2010; Fu and Lee, 2003). An accumulation of evidence, including convincing epidemiologic data, led the World Health Organization to conclude in 2007 that shift work is probably carcinogenic (Class 2A; Straif et al., 2007). In turn, animal research has shown conclusively that circadian disruption accelerates tumor progression (Fu and Lee, 2003). Several independent clinical studies have demonstrated that circadian rhythms have prognostic value with regard to mortality rates among patients with breast cancer (Sephton et al., 2000), colorectal cancer (Mormont et al., 2000; Innominato et al., 2009; Lévi et al., 2014), renal cell carcinoma (Cohen et al., 2012), and lung cancer (Sephton et al., 2013). Circadian disruption in cancer patients may have multiple etiologies: a portion of the increased variance in rhythms may be attributed to physiological effects of the tumor (Mormont and Levi, 1997). However, psychological distress – a symptom commonly experienced by cancer patients – can also markedly disrupt sleep and circadian rhythms (Sephton and Spiegel, 2003; Van Reeth et al., 2000). Studies suggest that distress and poor sleep may have serious repercussions for breast cancer patients (Eismann et al., 2010; Palesh et al., 2007). Psychological distress appears to increase breast cancer risk and shorten survival time (Chida et al., 2008; Spiegel and Giese-Davis, 2003), and poor sleep efficiency is prognostic for early breast cancer mortality (Palesh et al., 2014). The effects of stress and circadian disruption on cancer progression

may be interconnected, and may be mediated by endocrine and immune factors (Antoni et al., 2006; Eismann et al., 2010). The central circadian clock in the hypothalamic suprachiasmatic nucleus (SCN) generates behavioral rhythms and coordinates circadian fluctuations in peripheral cell growth, protein synthetic machinery, activation/secretion, and apoptosis (Mohawk et al., 2012; Lowrey and Takahashi, 2011). Because each of these activities is relevant to tumor growth, factors that entrain or disrupt SCN rhythms have potential to affect cancer progression (Filipski et al., 2002). The means by which the SCN coordinates peripheral cell rhythms is under investigation, but a strong candidate mechanism is via fluctuation in hypothalamic–pituitary–adrenal (HPA) axis activity. Other signal cascades may arise from circadian rhythms in sympathetic nervous activity and melatonin release. Circadian disruption may amplify the detrimental, cancer-promoting aspects of certain psychoneuroendocrine and immune pathways. Poorly coordinated behavioral and endocrine circadian rhythms increase daily tumor growth rates (Savvidis and Koutsilieris, 2012) and impact anti-tumor immunity (Cermakian et al., 2013; Eismann et al., 2010; Hanahan and Weinberg, 2011). Our group has developed a theoretical model for use in the cancer context, positing pathways by which circadian, psychological distress, and endocrine factors influence one another and may in turn affect tumor progression (Sephton and Spiegel, 2003; Eismann et al., 2010). This theory is based upon data from a wide range of research areas, and suggests three pathways by which cancer-relevant immunity may be affected via circadian rest–activity rhythms, psychosocial factors, and endocrine activation. Each of these pathways has documented influence on aspects of immunity relevant to tumor defense (Fig. 1; Eismann et al., 2010).

Fig. 1. Hypotheses were based on a model that posits associations of circadian rhythms (Arrow F), distress (Arrow G) and endocrine (Arrow H) function with tumor promoting serum biomarkers (black arrows; modified from Eismann et al., 2010). Tests of Arrows A, C, and D were already conducted within the current sample and reported in Dedert et al., 2012.

Please cite this article in press as: Cash, E., et al. Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery. Brain Behav. Immun. (2015), http://dx.doi.org/10.1016/j.bbi.2015.02.017

E. Cash et al. / Brain, Behavior, and Immunity xxx (2015) xxx–xxx

1.1. Circadian effectors of tumor promoting biology (Arrow F in Fig. 1) Clinical data show cancer patients and cancer survivors with clear circadian disruption have related alteration of inflammatory mediators (Rich et al., 2005; Irwin et al., 2013; Bower et al., 2005; Miller et al., 2008). Chronic inflammatory processes affect all stages of tumor development including initiation, promotion, and progression (Elinav et al., 2013). Thus, cancer patients with marked circadian disruption may have suppressed functional cellular immunity and overactive inflammatory responses, which could promote tumor growth, angiogenesis, and metastasis (Eismann et al., 2010). 1.2. Psychological effectors of tumor promoting biology (Arrow G in Fig. 1) Many breast cancer patients experience clinically significant distress at the time of diagnosis (Hegel et al., 2006), and may continue to experience emotional difficulty throughout treatment (Lethborg et al., 2000). Psychological distress is known to alter endocrine activation relevant to tumor progression, which may, in turn, reduce anti-tumor immune capabilities. While distress may be a normal response to such challenges, cancer patients and cancer survivors with greater anxiety, depression, fatigue, and cognitive difficulty also have elevated inflammatory responses (Bower et al., 2011; Miller et al., 2008; Rich et al., 2005; Dantzer et al., 2014; Seruga et al., 2008). 1.3. Endocrine effectors of tumor promoting biology (Arrow H in Fig. 1) Among cancer patients with cortisol rhythm aberrations, researchers have also observed suppressed functional cellular immunity and overactive inflammatory responses (Desantis et al., 2011; Mravec et al., 2008). Circadian cortisol rhythms may be considered a code of hypothalamic-immune communication (Arjona and Sarkar, 2008). Endocrine mediation of tumor growth has been shown to occur via products of the HPA and sympathetic stress response systems (Volden and Conzen, 2013; Lutgendorf and Sood, 2011), as well as glucocorticoid effects on tumor suppressor genes (Antonova and Mueller, 2008). New data suggest that glucocorticoids may also dysregulate immunity through epigenetic mechanisms (Krukowski et al., 2011). Our model and additional data (Lyon et al., 2008) suggest comprehensive assessment of cytokine profiles is needed to illuminate circadian–neuroendocrine–immune–tumor relationships. Such data may reveal mediating mechanisms and inform treatment strategies in cancer care. For a tumor to progress, it must acquire distinctive ‘‘Hallmark’’ capabilities that enable it’s growth and dissemination (Hanahan and Weinberg, 2011). It must sustain cell proliferation; evade growth suppression, immune destruction, and apoptosis; enable replicative immortality, invasion, and metastasis; and stimulate tumor-promoting inflammation (hereafter referred to as hallmarks of tumor progression; Hanahan and Weinberg, 2011). Epigenetic alterations that occur in tumor cells or tumor-associated stromal cells can result in altered secretory activity. By measuring secretory products, or their downstream effects, we might learn whether tumors are attaining hallmark capabilities. For example, secretion of vascular endothelial growth factor (VEGF) signals an increase in angiogenic capability, which aids tumor growth and metastasis (Veikkola and Alitalo, 1999; Claffey et al., 1996). Studies suggest high peripheral VEGF is prognostic for early mortality in several cancer types (Jin-no et al., 1998; Karayiannakis et al., 2002). Matrix metallopeptidase (MMP)-9, an enzyme that facilitates the process of tissue remodeling, is an important player in tumor invasiveness and metastasis, especially in response to stress hormone signaling

3

(Sood et al., 2006). Transforming growth factor (TGF)-b is involved in cancer cell growth and spread, induction of epithelial–mesenchymal transition (Kalluri and Weinberg, 2009), and plays differing roles at varying stages of disease (Zheng et al., 2008). Other measurable evidence of tumor progression may include systemic inflammatory markers, which also predict shortened cancer survival time (Salgado et al., 2003). Proinflammatory markers such as interleukin (IL)-1b, tumor necrosis factor (TNF)-a, IL-6, IL-12, and interferon (IFN)-c measured in the tumor microenvironment or in the periphery are likely indicative of tumors acquiring ‘‘Hallmark’’ capabilities (Mantovani et al., 2008). IL-6 levels are significantly elevated in lung and breast cancer patients and are associated with poor prognosis (Hodge et al., 2005). IL-6 signals via a receptor complex (IL-6R) and triggers downstream effectors of tumor growth (Grivennikov and Karin, 2008). Monocyte chemotactic protein (MCP)-1 perpetuates inflammatory processes and promotes invasion of certain breast cancer cell lines (Fujimoto et al., 2009). Tumor-associated MCP-1 levels correlate with tumor stage (Valkovic´ et al., 1998), suggesting a role in invasion. Circadian endocrine dysregulation may also shift the balance of T-helper (TH)1 versus TH2-dependent responses in a way that favors tumor growth. Human clinical and population-based studies show circadian endocrine disruption, including flattening of cortisol rhythms, co-occurs with suppressed functional cellular immunity and overactive inflammatory responses (Desantis et al., 2011; Miller et al., 2008). These changes in host biology could promote tumor cell proliferation, adhesion, migration, invasion, angiogenesis, and/or immune evasion (Antoni et al., 2006; Armaiz-Pena et al., 2009; Thaker and Sood, 2008). Additionally, cortisol and the HPA axis appear to play a vital role in the diurnal nature and coordination of immune surveillance (Trifonova et al., 2013). Research on circadian and psychosocial effects in cancer would benefit from inclusion of such biomarkers of successful tumors. Presurgical cancer patients present a unique opportunity to explore in the context of an active tumor the associations of circadian, psychosocial, and endocrine factors; markers indicative of tumor growth; and a tumor’s attainment of hallmark capabilities. Most prior work in this research area has relied on postsurgical and metastatic patients, or breast cancer survivors. Research with the intent of exploring associations between circadian, psychoneuroendocrine, and tumor promoting factors should also focus on subjects with active tumor, such as presurgical patients. Such studies would offer high relevance with regard to biology, research design, and clinical information. Given that bidirectional associations between biology and behavioral comorbidity likely take time to develop, it is unclear whether these associations exist in patients suffering the recent stressor of a breast cancer diagnosis, and to what extent a common group of symptoms (e.g., fatigue, poor mood, cognitive difficulty) is driven by cancer-related distress versus physiological factors. Although there is clear evidence that depression likely impacts both cancer biology and mortality (Howren et al., 2009; Pinquart and Duberstein, 2010; Spiegel and Giese-Davis, 2003), the impact of normal reactions to cancer diagnosis, such as distress, needs further exploration. Some early evidence suggests the potential relevance of persistent distress early in the cancer trajectory. One study showed that depression measured at the initiation of treatment for head and neck cancer is prognostic for early recurrence/mortality (Lazure et al., 2009). Another study that invites further inquiry is a randomized trial that demonstrated a 10-year survival benefit of psychiatric care during hospitalization for gastrointestinal cancer surgery (Küchler et al., 2007). The period of cancer diagnosis, surgery, and the initiation of post-surgical treatment may be acutely stressful. Among people suffering from stress of a relatively acute nature, the cortisol awakening response (CAR) may be elevated (Chida and Steptoe, 2009), but few studies have

Please cite this article in press as: Cash, E., et al. Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery. Brain Behav. Immun. (2015), http://dx.doi.org/10.1016/j.bbi.2015.02.017

4

E. Cash et al. / Brain, Behavior, and Immunity xxx (2015) xxx–xxx

explored the CAR as a potential effector of stress-related outcomes in cancer. Psychological distress after receipt of a cancer diagnosis has potentially significant implications for disease progression and mortality. Within a presurgical cancer patient sample, our research group has begun to test our model of circadian effects in cancer-relevant psychoneuroimmunology (Eismann et al., 2010): Among 57 breast cancer patients, we tested associations between circadian rest– activity rhythms, psychosocial factors, and endocrine activation (arrows A, C, and D in Fig. 1; Dedert et al., 2012) and found significant associations between actigraphic and cortisol-based measures of circadian disruption. Patients with high distress or poor coping evidenced greater rest–activity rhythm disruption (Dedert et al., 2012). The current analysis tests a different portion of our model: Associations of tumor promoting immune factors with rest– activity rhythms, psychosocial distress, and cortisol profiles. This analysis uses data from the same patient sample (Dedert et al., 2012), and uniquely includes data on peripheral blood biomarkers that potentially indicate a breast tumor’s attainment of hallmark capabilities that allow for growth and dissemination (Hanahan and Weinberg, 2011). Hypothesized associations are shown as arrows F, G, and H in Fig. 1.

Table 1 Sample characteristics (N = 43).

2. Methods 2.1. Participants English-proficient women awaiting surgical treatment for breast cancer were eligible for this study. Our recruitment site was a breast surgical oncology clinic. Most women referred to the clinic were patients with primary/localized disease, but a small number of women with recurrent or metastatic disease were also referred. Patients underwent biopsy to confirm diagnosis, and were introduced to the study by the breast surgeon (A.B.C.). Because data were all collected prior to treatment and definitive staging (from surgical pathology), the existence and site of any metastatic disease was unknown at time of data collection. Research assistants provided informed consent. Participants collected data concurrently over 4 days prior to neoadjuvant or surgical treatment. The study was approved by Institutional Review Boards at the University of Louisville and Stanford University. Ninety-one patients were referred, of which 60 were eligible and enrolled in the study. Reasons for refusal included response burden and/or time constraints. Fifty-seven participants provided enough actigraphic, salivary, and psychometric data to allow for calculation of summary statistics for at least two of these variables (reported in Dedert et al., 2012). Of those, seven participants declined to provide socioeconomic data, and circadian data were incomplete for two. We were unable to collect blood data on five patients. Reasons for missing blood samples include participant declining blood draw, problems with the sample (e.g., hemolyzation), or insufficient time to obtain sample prior to treatment initiation. Available data from 43 participants comprise the sample in current analyses. Definitive surgical diagnoses included 41 primary and two recurrent cases that had not received treatment within the past year. Most (67.5%) had Stage I or II disease. Pathology reports revealed DCIS (Stage 0) in three participants. Four patients elected lumpectomy, 35 chose mastectomy, and two underwent neoadjuvant treatment. Surgical data was unavailable for two patients. Frequently reported medical comorbidities included hypertension (n = 8) and diabetes (n = 7). Sample characteristics are presented in Table 1. 2.2. Assessments Participants provided data on demographics, medical history, and medications via questionnaire. Surgical pathology reports

a

Variable

Frequency

Percentage

Ethnicity White/caucasian African–American Native American Asian

27 13 2 1

62.8% 30.2% 4.7% 2.3%

Marital status Single Married/partnered Divorced Widowed

6 16 13 8

14.0% 37.2% 30.2% 18.6%

Years of education Middle School (8 years) High School (12 years) AA/Technical School (14 years) College Degree (16 years) Master’s Degree (18 years) Doctoral Training (20 years)

3 29 4 4 2 1

7.0% 67.4% 9.3% 9.3% 4.7% 2.3%

Household income Less than $20,000 20,000–39,999 40,000–59,999 60,000–79,999 80,000–99,999 100,000 and greater

18 10 3 3 4 5

41.9% 23.3% 7.0% 7.0% 9.3% 11.6%

Currently employed Yes No

22 21

51.2% 48.8%

Stage 0 (DCIS) I IIA IIB IIIA IIIB IIIC IV

3 19 5 2 7 0 3 4

7.0% 44.2% 11.6% 4.7% 16.3% 0% 7.0% 9.3%

Karnofsky ratinga 100 90 80 70 60

15 18 5 0 2

34.9% 41.9% 11.6% 0% 4.7%

Days between diagnosis and enrollment Range = 1–121

M = 19.5

SD = 22.88

Days between diagnosis and surgery Range = 4–200

M = 55.61

SD = 43.00

Age at diagnosis Range = 21–79

M = 52.49

SD = 13.35

Karnofsky scores were not obtained for three participants.

were the source of definitive staging and receptor status, including estrogen (ER), progesterone (PR), and human epidermal growth factor 2 (HER2/neu).

2.2.1. Circadian disruption Wrist-worn actigraphy devices (Micro Mini-Motionlogger; Ambulatory Monitoring Catalog No. 24.000BEF; Tamura et al., 2009), recorded movement consecutively for 3 days and four nights in 1-min epochs. Previous research suggests this recording interval is sufficient to calculate circadian disruption indices (Mormont et al., 2000; Innominato et al., 2009; Lévi et al., 2014). Each morning, patients provided logs of the time of sleep onset for the previous night and the time of awakening. Actigraphic estimates of time of sleep onset and waking were also calculated. Sleep intervals were set prior to calculation of summary variables by integrating self-reported sleep times with actigraphy output,

Please cite this article in press as: Cash, E., et al. Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery. Brain Behav. Immun. (2015), http://dx.doi.org/10.1016/j.bbi.2015.02.017

5

E. Cash et al. / Brain, Behavior, and Immunity xxx (2015) xxx–xxx

which was chosen when the two were discrepant (n = 27 time points, 3.5%). We did not query about time in bed spent reading or watching TV, because these activities may be better distinguished from sleep by actigraphy scoring algorithms than self-reports. Commercial software (Action 4; Ambulatory Monitoring) was used to calculate rest/activity rhythm (autocorrelation coefficient) and two dichotomy indices based on motion while in (I) compared to out (O) of bed: nighttime restfulness and daytime sedentariness. The autocorrelation coefficient correlated epochs of activity on 1 day to those at the same time on other days. This is considered a measure of circadian consistency or similarity of rest/activity patterns across days. Greater nighttime restfulness (I < O dichotomy index) was calculated as the percentage of activity when the patient is in bed that falls below the median level of activity when out of bed. Conversely, daytime sedentariness (O < I dichotomy index) indicates the percentage of activity when out of bed that falls below the median level of activity while in bed. In both cases, a lower percentage score indicates less frequent movement. Greater nighttime restfulness (higher I < O) is consistent with strong circadian rhythmicity, while greater daytime sedentariness (higher O < I) is consistent with circadian disruption (Mormont et al., 2000). Data were examined for differences in weekday versus weekend data collection, which was noted only for daytime sedentariness (t(41) = 2.177, p = .035). Calculation of sleep parameters was based on the same epochs. Variables included sleep latency, total sleep time, minutes spent awake after sleep onset (WASO), and sleep efficiency, the proportion of time in bed spent asleep (previously described; Dedert et al., 2012). Subjective sleep quality was also reported each morning with respect to the past night using a single item measure. 2.2.2. Psychological distress Participants completed the Impact of Event Scale (IES; Horowitz et al., 1979), a 15-item measure assessing symptoms of intrusion and avoidance during the previous week, which was keyed to a potentially traumatic experience: diagnosis of breast cancer. This measure has demonstrated validity and reliability in cancer populations (Cordova et al., 1995). They also provided reports on the Profile of Mood States (POMS), a 65-item measure of affective states during the past week (McNair et al., 1971). The POMS has been used frequently in cancer patients (Stanton et al., 2000), and the Tension-Anxiety subscale has previously been related to cytokine expression among ovarian cancer patients (Lutgendorf et al., 2008a). The IES total score, POMS-Total Mood Disturbance, and POMS-Tension-Anxiety subscales were used separately in hypothesis tests. Mean total cancer-specific distress score on the IES was 29.69 (SD = 15.14). Mean score for POMS-Total Mood Disturbance was 44.79 (SD = 38.65), and 15.85 (SD = 8.35) for Tension-Anxiety. 2.2.3. Endocrine disruption Participants collected saliva samples for assessment of cortisol at waking, 30 min after waking (+30 min), 4 p.m., and immediately prior to going to bed over 3 days to yield 12 samples. Medication Event Monitoring System (MEMS; Aardex, Union City, CA) bottles were used to store cotton swabs for saliva collection. Electronic bottle caps recorded the time and date of sample collection. Participants were asked to not brush their teeth, eat, drink, or smoke for 30 min prior to each sample collection. A research assistant centrifuged, aliquoted, and stored all saliva samples at 80 °C. Assays were conducted in Dr. Sephton’s lab using a salivary EIA (Salimetrics, Inc.) sensitive to 0.007 lg/dL. Inter-assay CVs were 10.45% for low and 4.03% for high controls, respectively. Detailed information about data preparation procedures has been previously reported (Dedert et al., 2012). Briefly, MEMS and self-reported saliva collection times were very highly correlated

(rs = .962), suggesting reliable self-reports, consistent with prior research (Kraemer et al., 2006). For discrepant reports, MEMS and actigraphy data were reviewed. When there was >10 min discrepancy between self-reported collection times and MEMS recorded times, MEMS collection times were utilized (n = 6 samples, 1.2%). Collection time (n = 2 samples, 0.4%) and cortisol values (n = 6 samples, 1.2%) that were >4 SD from the mean were considered outliers and deleted. Data from one participant were excluded based on the potential effects on circadian regulation due to her shift work schedule. After examining for non-adherence and statistical outliers, a total of 11.8% of the individual cortisol values (n = 61 samples) were excluded due to statistical outlier status (noted above; n = 8), nonadherence to +30 min sample timing (<15 or >45 min after waking sample; n = 6), and determination based on corroborating actigraphy data that samples were not collected at actual time of waking or bedtime (n = 47 samples). Cortisol values were logtransformed. The diurnal cortisol slope was calculated excluding +30 min samples using unstandardized beta of log-transformed cortisol regressed on collection time. The cortisol awakening response (CAR), or slope of change in cortisol after awakening, was similarly calculated using waking and +30 min samples. Diurnal mean, an estimate of the overall exposure of cortisol, was calculated using all 12 log-transformed cortisol values. Mean and 95% confidence intervals of raw cortisol levels averaged over each saliva collection time are presented in Fig. 2. 2.2.4. Serum biomarkers A trained phlebotomist drew a single blood sample using Vacutainer™ tubes. Serum was collected as close to study entry as possible, and always prior to surgical treatment. Since biomarkers of interest are known to exhibit circadian patterns of release in systemic circulation (Kronfol et al., 1997), the timing of blood draw was restricted as much as possible. Our clinic scheduled all new patients at 11 a.m. and 1 p.m. Mean blood collection time was 1:15 p.m. (SD = 119 min; median = 1:32 p.m.); none were postprandial. Serum aliquots were frozen at 80 °C within 2 h of blood draw. Parameters were analyzed in duplicate using electrochemiluminescent quantitative sandwich immunoassays from MesoScale Discovery (Gaithersburg, MD). A multiplex assay was used to quantify IL-1b, IL-6, TNF-a, IL-12, and IFN-c. Single-plex assays were used to quantify the other factors. Assay plates were imaged on a SECTOR Imager 2400A, and data analyzed using Discovery Workbench software (MesoScale). Eight values (1.2%) fell below detection limits of the assays, and these values were replaced with

0.500

0.400

0.300

0.200

0.100

0

Wake Wake+30

16:00

Bedtime

Fig. 2. Mean raw diurnal salivary cortisol levels over the 3-day collection period among pre-surgical breast cancer patients (n = 43).

Please cite this article in press as: Cash, E., et al. Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery. Brain Behav. Immun. (2015), http://dx.doi.org/10.1016/j.bbi.2015.02.017

6

E. Cash et al. / Brain, Behavior, and Immunity xxx (2015) xxx–xxx

0.001 for the purpose of analysis. Respectively, the detection limits, intra-assay, and inter-assay CVs were: VEGF (9.9, 2.1%, 2.4%), MMP9 (0.07, 4.3%, 0.2%), TGF-b (0.003, 5.6%, 1.3%), IL-1b (0.06, 5.9%, 2.9%), TNF-a (0.29, 7.1%, 4.2%), sIL-6R (0.06, 9.7%, 3.1%), MCP-1 (0.25, 6.1%, 2.3%), IL-6 (0.09, 5.2%, 3.1%), IL-12p70 (0.50, 9.4%, 7.8%), and IFN-c (0.34, 6.1%, 0.1%). Values that fell >4 SD from the mean and displayed as extreme values on box and whisker plots (n = 6) were considered outliers and deleted. Summaries of actigraphy, cortisol, and serum biomarkers are presented in Table 2. 2.2.5. Control variables All analyses included three covariates, determined by their strong relationships to cancer outcomes in general, and as noted in a prior study (Dedert et al., 2012): Two theoretical control variables were entered, age at diagnosis and cancer stage, as both are associated with variance in speed of tumor progression and level of associated psychological distress. Third, recruitment took place in two different clinics under the same surgical oncologist, where average income varied significantly. Additionally, we previously observed that income level was related to outcomes of interest, prompting inclusion as an empirical control (Dedert et al., 2012). No covariates demonstrated significant relationships with factor scores. 2.3. Statistical analysis 2.3.1. Data reduction The precise nature of potential cancer-related distress and tumor effects on systemic tumor progression biomarkers is unknown. Thus, we were not able to hypothesize exactly which serum factors would covary in this unique sample of patients. As such, we tested parameters based on a data-driven approach, rather than grouping parameters based on theoretically determined function. Exploratory factor analysis (EFA) was used to explore how serum parameters covary in the presence of active disease.

Table 2 Summary statistics of all physiological variables collected as part of this study (N = 43, except CAR N = 35). Variable Actigraphy Rest–activity rhythm, coefficient Nighttime restfulness, % Daytime sedentariness, % Mean sleep time, min Sleep efficiency, % WASO, min Salivary cortisol Diurnal slope, log(lg/dL)/h CAR, log(lg/dL)/h Diurnal mean, log(lg/dL) Mean waking, log(lg/dL) Mean bedtime, log(lg/dL) Serum biomarkers VEGF, pg/mL MMP-9, pg/mL TGF-b, pg/mL IL-1b, pg/mL TNF-a, pg/mL IL-6R, ng/mL MCP-1, pg/mL IL-6, pg/mL IL-12, ng/mL IFN-c, pg/mL

Mean (SD)

Median

1st–3rd quartile range

.27 (.16)

.28

.16–.41

97.21 (3.60) 6.59 (5.88) 386.94 (76.55) .89 (.08) 13.00 (6.30)

98.46 4.80 405.67 .91 11.75

97.57–99.08 2.56–8.84 350.00–448.00 .85–.94 9.75–18.00

.071 (.090) .082 (.515) 1.213 (.390) 1.330 (.540) 2.791 (.993)

.071 .002 1.265 1.307 2.777

377.62 (225.36) 28.35 (16.13) 2.67 (1.01) 0.25 (0.43) 8.10 (3.54) 6199.11 (1570.55) 150.10 (62.28) 2.23 (2.13) 11.60 (17.34) 1.26 (1.12)

344.93 24.12 2.63 0.18 7.63 6250.59

210.26–474.71 18.15–35.75 2.04–3.07 0.12–0.25 6.26–8.72 4936.66–7189.08

139.24 1.54 4.50 0.94

104.31–188.32 1.14–2.55 1.49–12.34 0.62–1.39

.127 to .012 .105 to .223 1.406 to .952 1.607 to .928 3.384 to 1.908

EFA determined the degree to which serum biomarkers varied together. The Kaiser–Meyer–Olkin (KMO) statistic, a measure of sampling adequacy, should be above .5 to help ensure the data meet the assumptions of the analysis (Field, 2005). For this model, KMO = .552. Anti-image statistics for each molecule entered into the model were examined to ensure that the amount of variance explained by each was sufficient to warrant inclusion in the final model. Further assurance that the parameters varied together was sought by examining the Factor Loading, which reflects both the concentration of the molecule and its tendency to covary with other molecules that loaded onto that factor. Factor Loadings <0.4 were not interpreted, as they do not explain sufficient variance within the overall model (Field, 2005). The analysis also returns a corresponding factor score for each participant, which is a linear combination of the log-transformed values of the serum parameters weighted by the corresponding Factor Loading. A factor score was calculated for each of the three clusters that emerged from the EFA, and represents an individual’s score on each Factor. The factor scores were then used separately as the dependent variables for tests of hypotheses. 2.3.2. Preliminary analyses To ensure that there were no significant patterns of disruption that may confound results, independent samples t-tests were used to compare circadian, distress, and endocrine variables among the 43 participants that make up the current sample versus the remaining 14 patients from our full sample of 57. 2.3.3. Hypothesis tests After ensuring the data met statistical assumptions, hierarchical linear regressions controlling for age at diagnosis, cancer stage, and income examined the hypothesized relationships. Circadian (rest– activity rhythm, nighttime restfulness, daytime sedentariness), distress (IES, POMS-Total Mood Disturbance, POMS-TensionAnxiety) and endocrine (diurnal cortisol slope, CAR, diurnal mean) variables were entered individually as predictors, and the three factor scores separately as outcome variables. When significant relationships emerged, each biomarker was then individually examined as a dependent variable. This allowed us to determine if one analyte in particular could possibly account for the significant association with factor scores. The purpose of this approach was to glean information on the unique signaling molecule clusters that may exist among patients with an active tumor, and to inform future studies. 2.3.4. Secondary analyses Secondary analyses were conducted with the intent to identify potential confounding variables that might explain the observed significant associations. As a check, we adjusted for contributors of variance in circadian and endocrine parameters and tumor receptor status by re-running tests of hypotheses after adjustment for weekday versus weekend days of data collection, hormonal receptor status of the tumor, and ethnicity. Tests of hypotheses were repeated after removal of recurrent and metastatic breast cancer cases. Behaviorally measured rest–activity rhythms provide a good approximation of centrally mediated circadian rhythms, which regulate the timing of periods of sleepiness and wakefulness throughout the day. However, other factors may also drive sleep, and few studies have explored the relative value of rest–activity versus sleep parameters in the cancer context. After again adjusting for age, stage, and income, we explored the association of three additional measures with potential relevance to serum biomarkers (factor scores): actigraphically-measured total sleep time and sleep efficiency, and subjective sleep quality. To confirm which aspects of cortisol related to serum biomarkers, we also explored

Please cite this article in press as: Cash, E., et al. Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery. Brain Behav. Immun. (2015), http://dx.doi.org/10.1016/j.bbi.2015.02.017

7

E. Cash et al. / Brain, Behavior, and Immunity xxx (2015) xxx–xxx

associations of mean waking and evening cortisol levels with factor scores. 2.3.5. Power analysis For EFA models, general convention suggests five participants per variable should be sufficient (Field, 2005). Since we planned to examine ten serum analytes in this sample of 43 participants, caution was used when interpreting Factor Loadings and model fit, and measurements of sampling adequacy were checked carefully. Based on previously reported effect sizes of predictors and control variables (Rich et al., 2005), a medium effect size (f2 = .15; Cohen, 1988) was proposed for relationships between distress and serum parameters, with smaller effect sizes (f2 = .11) for tests of circadian and endocrine relationships with serum parameters. All models were predicted to have 81% power or greater (Faul et al., 2007). 3. Results

Table 4 Hierarchical linear regressions tested relationships between actigraphic measures of circadian disruption and factor scores derived from an exploratory factor analysis of a panel of serum biomarkers. Control variables (age at diagnosis, cancer stage, income) were entered on the first step for each regression model. No significant associations with factor scores were observed (p’s > .05; not shown). Step two for all analyses is shown (N = 43). Variable

The EFA demonstrated good fit (Bartlett’s test of sphericity p < .001) and good sampling adequacy (KMO > .5). EFA resulted in a model of resting serum biomarker levels yielding three Factors. Factor 1 consisted of three molecules generally associated with tumor progression, specifically tumor invasion and immunosuppression: VEGF, MMP-9, and TGF-b. Factor 2 included both proinflammatory and chemotactic markers. Factor 3 was comprised of factors generally associated with proinflammatory and TH-1 type processes. These statistically derived clusters and their corresponding Factor Loadings are presented in Table 3. No significant differences emerged when comparing the circadian, distress, and endocrine variables of the sample of 43 with the 14 women whose data were not included in the current analyses. 3.2. Circadian disruption related to Factor 1 and VEGF (tests of Arrow F in Fig. 1) Rest/activity rhythm was significantly related to Factor 1, such that women with more robust circadian rest/activity cycles had lower levels and a weaker correlation among biomarkers of tumor invasion and immunosuppression (VEGF, TGF-b, MMP-9; Table 4; b = .364, partial r = .338, p = .038). For descriptive purposes, participants were split into two equal groups based on their rest/activity rhythm scores. A comparison of the two groups is illustrated in Fig. 3, Panel A. Models testing relationships between nighttime restfulness, daytime sedentariness, and factor scores were non-significant. In tests using individual serum parameters from the significant model as outcomes, women with more robust rest/ activity rhythms had significantly lower serum VEGF levels Table 3 Factors emerging from an exploratory factor analysis of tumor-promoting serum biomarkers are presented. The Factor Loading is also presented, which indicates of the amount of variance each biomarker contributes to the Factor on which it loads. Factor

Biomarker

Loading

1: Tumor invasion/immunosuppression

VEGF MMP-9 TGF-b

.829 .805 .720

2: Inflammation/chemotaxis

IL-1b TNF-a IL-6R MCP-1

.625 .743 .734 .578

IL-6 IL-12 IFN-c

.627 .561 .671

3: Inflammation/TH1-type immunity

SE B

Factor 1 (VEGF, MMP-9, TGF-b) Rest–activity rhythm 1.395 Nighttime restfulness .023 Daytime sedentariness .007 Factor 2 (IL-1b, TNF-a, IL-6R, Rest–activity rhythm Nighttime restfulness Daytime sedentariness

.647 .036 .024

MCP-1) .967 1.166 .086 .047 .010 .016

Factor 3 (IL-6, IL-12, IFN-c) Rest–activity rhythm 1.794 Nighttime restfulness .010 Daytime sedentariness .031 ⁄

3.1. Preliminary data analysis

B

.970 .043 .028

R2

DR2

p of DR2

.364⁄ .101 .050

.160 .090 .082

.108 .010 .002

.038 .533 .776

.146 .288 .106

.049 .109 .071

.017 .077 .009

.412 .078 .552

.291 .035 .182

.239 .171 .197

.069 .001 .026

.072 .821 .270

b

p < .05

(b = .409, partial r = .374, p = .017). Rest/activity rhythm was not significantly related to MMP-9 or TGF-b (all p’s > .124). For descriptive purposes and to better understand the observed relationships, levels of each serum biomarker were plotted separately for participants above versus below the median rest/activity rhythm score (Fig. 3, Panel B). While the effect size of the relationship between rest/activity rhythm and Factor 1 was moderate (f2 = .13), a slightly larger effect size was observed between rest/activity rhythm and VEGF (f2 = .16). 3.3. Psychological distress did not relate to serum biomarker profiles (tests of Arrow G in Fig. 1) Hierarchical regressions entered IES scores, POMS-Total Mood Disturbance, and POMS-Tension-Anxiety scores separately as predictors of factor scores. No models emerged as significant. 3.4. Endocrine disruption related to Factor 1 and VEGF (tests of Arrow H in Fig. 1) The CAR was significantly associated with serum Factor 1 such that women with greater cortisol responses to awakening had higher levels and greater cohesion among biomarkers of tumor invasion and immunosuppression (VEGF, TGF-b, MMP-9; Table 5; b = .386, partial r = .428, p = .020). For descriptive purposes, Factor 1 scores are shown for participants split at the median CAR (Fig. 4, Panel A). Models testing diurnal cortisol slope and mean were non-significant. In tests using individual serum parameters from the significant model as outcomes, women with a higher CAR had significantly higher serum VEGF levels (b = .534, partial r = .539, p = .001), but CAR was not significantly related to MMP9 or TGF-b (all p’s > .620). To further describe these differences, levels of each serum molecule were plotted for groups split at the median CAR level (Fig. 4, Panel B). The effect size of the relationship between CAR and Factor 1 was moderate (f2 = .24), while a large effect size was observed for the relationship between CAR and VEGF (f2 = .41). 3.5. Secondary analyses In secondary analyses conducted with the intent to identify potential confounding variables that might explain the observed associations, separate models adjusted for weekday versus weekend collection, tumor receptor status, and ethnicity. Significant

Please cite this article in press as: Cash, E., et al. Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery. Brain Behav. Immun. (2015), http://dx.doi.org/10.1016/j.bbi.2015.02.017

8

E. Cash et al. / Brain, Behavior, and Immunity xxx (2015) xxx–xxx

Rest/Activity Rhythm

* p=.038

* p=.017

0.50

0

-0.50

35 500 30 400

25 20

300

15 200 10 100

5

-1.00

A

B

Mean Raw Serum Level (pg/mL)

40

600

1.00 Mean Raw Serum Level (pg/mL)

Mean Serum Factor 1 Score (VEGF, MMP-9, TGF-β)

Below Median (n=21) Above Median (n=22)

0

0 VEGF

MMP-9

TGF-β

Fig. 3. Mean exploratory factor analysis Factor 1 scores (Panel A) and raw serum level of each cytokine that loaded into Factor 1 (Panel B) are displayed. Levels are displayed in two equal groups based on rest/activity rhythm scores, derived for descriptive purposes only.

Table 5 Hierarchical linear regressions tested relationships between endocrine (salivary cortisol) disruption and factor scores derived from an exploratory factor analysis of a panel of serum biomarkers. Control variables (age at diagnosis, cancer stage, income) were entered on the first step for each regression model. No significant associations with factor scores were observed (p’s > .05; not shown). Step two for all analyses is shown (diurnal slope and diurnal log mean N = 43; CAR N = 35). Variable

B

SE B

b

R2

DR 2

p of DR2

Factor 1 (VEGF, MMP-9, TGF-b) Diurnal slope 1.016 1.468 CAR .606 .247 Diurnal log mean .099 .345

.109 .386⁄ .046

.120 .293 .111

.011 .142 .002

.493 .020 .775

Factor 2 (IL-1b, IL-6R, TNF-a, MCP-1) Diurnal slope .682 .845 CAR .196 .161 Diurnal log mean .273 .444

.133 .211 .101

.089 .141 .042

.016 .042 .009

.425 .233 .542

Factor 3 (IL-6, IL-12, IFN-c) Diurnal slope .806 CAR .381 Diurnal log mean .284

.072 .188 .109

.159 .232 .164

.005 .034 .011

.641 .261 .484

⁄

1.717 .332 .401

p < .05

results persisted after controlling for each of these factors. We also ran separate analyses that removed recurrent and metastatic breast cancer cases, which produced no substantive changes in the significance of the results. No significant results were obtained from exploratory analyses examining associations of Factor 1 scores with actigraphic sleep variables, self-reported sleep quality, mean log waking, and evening cortisol. 4. Discussion We found support for two of the relationships hypothesized in Fig. 1. Supporting arrow F, presurgical breast cancer patients with uncoordinated rest–activity rhythms (i.e., poor inter-daily activity rhythm stability) had significantly elevated levels and a strong clustering of serum signals related to tumor progression (Factor 1). These included biomarkers of angiogenesis (VEGF), immunosuppression and epithelial–mesenchymal transition (TGF-b), tumor invasion (MMP-9) and metastasis (VEGF & MMP-9).

Supporting arrow H, higher cortisol responses after awakening were associated with the same signal cluster. Secondary analyses suggested that both findings were driven largely by VEGF concentrations. We did not find an association between distress and immunosuppression markers or systemic inflammation. These findings support a growing body of literature linking circadian disruption with tumor promotion and accelerated cancer progression (Fu and Lee, 2003; Filipski et al., 2002, 2004). We used two different indicators of circadian regulation, both of which have demonstrated prognostic value in cancer: rest–activity rhythms and cortisol salivary cortisol rhythms. This strategy was intended to elucidate potential biological links that might explain the associations between circadian disruption and cancer prognosis, by using a novel presurgical sample with active tumor. Our data represent a group of women whose circadian coordination, distress, and HPA rhythms lie generally between that of normal controls and patients with more advanced cancer. Rest– activity rhythms observed in this sample suggest circadian coordination poorer than that of healthy controls (Fernandes et al., 2006) and similar to that of patients with metastatic colorectal cancer (Lévi et al., 2014). However, our patients demonstrated more homogeneous restfulness at night (a more restricted range of values indicating inactivity while in bed) as compared with this group of patients with more advanced disease. A large proportion (47%) of the sample reported cancer-specific distress above the clinical cutoff for post-traumatic stress disorder (IES score P 35; Horowitz et al., 1979). In comparison with breast cancer patients within 9 months of diagnosis (Goldsmith et al., 2010), our sample was more distressed. However, their distress was similar to that of patients awaiting breast biopsy (Flory and Lang, 2011). Up to 90% of healthy controls demonstrate robust diurnal cortisol rhythms (Stone et al., 2001). In contrast, aggregate data from cancer samples often demonstrate flattening of the cortisol rhythm, with data from individual patients showing idiosyncratic changes in secretion including phase shifting, aberrant peaks and troughs, or consistently flattened levels. We expected to see some degree of diurnal cortisol disruption in our sample based simply on the presence of cancer (Abercrombie et al., 2004). Among our sample, 49% demonstrated continually descending diurnal cortisol after the

Please cite this article in press as: Cash, E., et al. Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery. Brain Behav. Immun. (2015), http://dx.doi.org/10.1016/j.bbi.2015.02.017

9

E. Cash et al. / Brain, Behavior, and Immunity xxx (2015) xxx–xxx

A

Below Median (n=17) Above Median (n=18)

0.50

0

-0.50

* p=.020

600

* p=.001

40 35

500

30 400

25 20

300

15 200

10

Mean Raw Serum Level (pg/mL)

1.00

Mean Raw Serum Level (pg/mL)

Mean Serum Factor 1 Score (VEGF, MMP-9, TGF-β)

Cortisol Awakening Response

100

5 -1.00

0

B

0 VEGF

MMP-9

TGF-β

Fig. 4. Mean exploratory factor analysis Factor 1 scores (Panel A) and raw serum level of each cytokine that loaded into Factor 1 (Panel B) are displayed. Levels are displayed in two equal groups based on cortisol awakening response (CAR) levels, derived for descriptive purposes only.

morning peak, indicating normal HPA rhythmicity. Aberrant rhythms in the remaining half of our sample can be characterized, in decreasing order of prevalence, as: evening cortisol peaks, flattened profiles, and afternoon peaks. Cortisol rhythm disruption occurred among fewer of our presurgical patients (51%) than previously observed among women with advanced breast cancer (73%; Sephton et al., 2000). Viewed in aggregate, diurnal cortisol decline in our sample was steeper (i.e., appeared more healthy) than that of fatigued breast cancer survivors (Bower et al., 2005), but flatter than that observed in a sample of newly diagnosed, postsurgical breast cancer patients in the Netherlands (Vedhara et al., 2006). Our data support the notion that circadian and endocrine dysregulation, and concomitant loss of HPA flexibility and responsivity, are present to some degree early in the cancer trajectory, and that they become more pronounced over time with advancing cancer (Eismann et al., 2010; Mormont et al., 2000). Exploratory Factor Analysis grouped the ten serum biomarkers into three Factors using a data-driven technique. This strategy allowed the complex and unique data – multiple biomarkers from peripheral blood of patients with active tumor – to describe relationships between biomarkers involved in in vivo tumorassociated physiology. We purposefully chose a statistical technique that would not impose biological assumptions to allow us to learn about relationships between biomarkers in the cancer context. The significant pattern of covariance, or correlation, that emerged showed cohesion among biomarkers related to tumor angiogenesis, invasion, metastasis, and immunosuppression (Factor 1: VEGF, MMP-9 and TGF-b). Factors 2 and 3 revealed associations between proinflammatory, chemotactic, and TH-1 type molecules (Table 3). These biomarkers are not typically grouped together based on biological theory or previous research, but are all known to facilitate tumor progression. The strong cohesion between levels of these molecules may indicate an active influence by the tumor on secretion of these particular agents, or weaknesses in host tumor-defenses that permitted the production of these tumor-promoting biomarkers. The Factors suggest a way of thinking about the biology and provide biological outcomes that

uniquely reflect tumor–host physiology in women with active breast tumor. However, they do not provide any mechanistic information and cautious interpretation is appropriate. We hypothesized (arrow F) that patients with uncoordinated rest–activity rhythms, marked by poor inter-daily activity rhythm stability, would show elevation of biomarkers of tumor progression. In primary analyses, Factor 1 was associated with uncoordinated rhythms. Although all three biomarkers loaded with similar strength on Factor 1 (Table 3), differences were seen in tests of their individual associations with rest–activity rhythm: significant associations persisted only for VEGF, not for MMP-9 or TGF-b. While the source of serum VEGF from the host versus tumor cells was not identified, its potential biological relevance to tumor growth is clear (Kimura et al., 2007). Elevations in serum VEGF are a potential indicator of a tumor gaining the ability to stimulate angiogenesis, one of the identified hallmarks of progression (Hanahan and Weinberg, 2011). VEGF is a primary stimulus for angiogenesis, but also contributes to other key aspects of tumorigenesis such as enhancing tumor invasion, survival (Hormbrey et al., 2002), and the function of cancer stem cells (Goel and Mercurio, 2013). Tumors may scavenge VEGF from the circulating blood, trapping it using specific receptors, and then re-releasing it in paracrine fashion to stimulate local tumor angiogenesis (Hashiguchi et al., 2000). The circadian rest/activity rhythm is a strong indicator of brainmediated, central circadian rhythms (Ancoli-Israel et al., 2003). In turn, the central circadian clock regulates the timing of peripheral tissue production of VEGF through periodic activation of peripheral clock genes Per1 and Per2 (Koyanagi et al., 2003). Centrally mediated circadian rhythm disruption could deregulate this action of peripheral clocks in tumor or host defense cells, altering peripheral Per1 and Per2 activity and leading to elevated tissue levels of VEGF that may mediate faster tumor growth (Jensen et al., 2012; Jensen, 2014). In support of this notion, down-regulation of Per1 expression increases cancer cell growth in vitro by accelerating tumor growth at certain times of day (Yang et al., 2009). Further, the maintenance of regular circadian cycles appears to be crucial

Please cite this article in press as: Cash, E., et al. Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery. Brain Behav. Immun. (2015), http://dx.doi.org/10.1016/j.bbi.2015.02.017

10

E. Cash et al. / Brain, Behavior, and Immunity xxx (2015) xxx–xxx

for avoiding excessive oxidative stress, which has also been linked with elevated VEGF production. Hypoxia-induced expression of VEGF is known to increase tumor angiogenesis (Koyanagi et al., 2003). Several other molecular regulators of the circadian clock have been found to interact with cell cycle signals, suggesting circadian disruption may alter tumor growth via multiple other pathways (Sahar and Sassone-Corsi, 2009). Thus, the associations between rest–activity rhythm and Factor 1/VEGF may be a marker of another effector, rather than a mediator (Yasuniwa et al., 2010; Sahar and Sassone-Corsi, 2009). In contrast with the rest–activity rhythm findings, we observed no association of sleep with serum biomarkers, perhaps highlighting a biologically and/or clinically meaningful distinction between sleep and circadian parameters. Neither the actigraphic, nor the self-report estimates of sleep were associated with serum biomarkers. Sleep disturbances are common in cancer patients receiving treatment, and among some survivors (Savard and Morin, 2001; Penttinen et al., 2011). Despite their moderate to severe rest– activity rhythm disruption, this sample displayed relatively good sleep with most women having sleep efficiency scores above the generally accepted cutoff of .85 for clinically significant insomnia (Morin et al., 1993). Thus, there may have been too little variation in sleep parameters to allow relationships with biomarkers to emerge statistically. Our findings highlight the need for future research to consider both sleep and circadian rhythms in exploring psychoneuroimmune relationships. Sleep and circadian rhythms are regulated by somewhat different mechanisms (Hennig et al., 1998; Zeitzer, 2013), and differences may extend to their respective associations with biological factors relevant to tumor growth. Interestingly, we saw no association between distress and immunosuppression markers, or between distress and systemic inflammation. Previous research has shown that psychological distress can influence immune function, particularly in individuals with advanced (i.e., metastatic) cancer or cancer survivors (Bower et al., 2011; Miller et al., 2008; Rich et al., 2005; Dantzer et al., 2014; Seruga et al., 2008). Women awaiting breast biopsy are perhaps more similar to our presurgical sample, because their data present a view of the biology of an active tumor for those who will be so diagnosed. Among a pre-biopsy sample, distress has been associated with immunosuppression (suppressed natural killer cell activity and increased proinflammatory cytokines; Witek-Janusek et al., 2007). Similarly, among postsurgical breast cancer patients, distress has been linked with suppressed natural killer cell activity and lymphocyte proliferation (Andersen et al., 1998). Notably, our presurgical sample was assessed at a different time during the cancer trajectory. Our immune assessments were also different. We measured peripheral tumor-relevant biomarkers including some inflammatory mediators, but these were not functional assays. Further there may be differences in the type and magnitude of interactions between distress and biomarker activity when comparing pre-biopsy patients with those who have a confirmed diagnosis and are awaiting breast cancer surgery. It is generally accepted that reliable estimates of the diurnal salivary cortisol slope, gathered using multiple days of morning and evening collection, may be indicative of circadian disruption and/or chronic exposure to stress. In contrast, the CAR more likely reflects awakening-induced activation of the HPA axis by the SCN (Clow et al., 2010). The CAR is known to increase with stress, and specifically stress perceived upon awakening (Chida and Steptoe, 2009). The notion that the CAR may be closely related to sympathetic, rather than strictly HPA, activation invites further research on CAR associations with circadian, sympathetic, and tumor progression biomarker responses in the context of cancer (Chida and Steptoe, 2009; Clow et al., 2010; Edwards et al., 2001). Our observed associations of the CAR with Factor 1 scores and VEGF might be understood in the context of other research that has

shown alterations in cortisol secretion can affect VEGF and MMP9 production. Elevated glucocorticoid levels indicative of greater chronic stress burden have been linked with higher tumor-associated tissue levels of VEGF and greater tumor burden (Dhabhar et al., 2012). In one study, cortisol enhanced production of VEGF by two ovarian tumor cell lines, though effects varied by cell line and the dose of glucocorticoids (Lutgendorf et al., 2003). Further, another study showed glucocorticoids have been shown to cause a twofold increase in MMP-9 levels in ovarian tumor cells (Lutgendorf et al., 2008b). It is possible that women in our sample with elevated CAR also tended toward greater HPA reactivity to stress, and that the observed association with Factor 1 scores and VEGF was driven by cortisol reactivity. However, no implication regarding the direction or causation of the finding can be gleaned from our data. It is also possible these results were driven by a third, unmeasured confounding factor associated both with CAR and Factor 1/VEGF. Breast tumors are a heterogeneous group of diseases that are distinguished into subtypes by the hormone receptors they carry (Sotiriou and Pusztai, 2009; Reis-Filho and Pusztai, 2011). The presence or absence of ER, PR, and/or HER2/neu receptors determines whether certain targeted drug therapies can be of potential benefit in specific cases. Effects of glucocorticoids on treatment outcomes also vary by tumor receptor status (Volden and Conzen, 2013). In a recent study, analysis by ER status produced different results with regard to effects of Supportive-Expressive Group Therapy on metastatic breast cancer survival, with survival benefits observed only among ER negative patients (Spiegel et al., 2007). The current findings remained significant after adjustment for tumor receptor status in the analyses. Given that our presurgical sample had not yet received any differential cancer treatment based on their tumor receptor subtype, this finding is not surprising. Future work should aim to clarify potential effects of circadian and HPA aberrations among heterogeneous tumor types. Lending support to our findings, results remained significant after adjusting the relevant models for weekday versus weekend data collection (tests of F and H; Fig. 1), effects of ethnicity, and the potential influence on results from our few recurrent/metastatic cancer cases (tests of F, G, and H). We also explored wake and evening cortisol levels as predictors in models examining arrow H, and these were not significant. This suggests that the diurnal rhythm of cortisol secretion, rather than cortisol levels per se, is more strongly linked with peripheral tumor-associated biomarkers in this sample. This sample included a high proportion of African American patients. Understanding psychoneuroimmune responses is especially relevant among these patients, who have a lower breast cancer incidence but higher mortality rate compared to non-Hispanic whites (Siegel et al., 2014). However, the study is limited in several ways. Without normative data for comparison, definitive conclusions are not possible. We examined a relatively narrow window of 3 to 4 days of cortisol and actigraphy assessment. It may be preferable to collect fewer cortisol samples over more days (i.e., waking and evening samples over 5–8 days) to reliably characterize diurnal cortisol slope (Segerstrom et al., 2014). We also had only one blood draw for biomarker assessment. It could be more informative to include multiple blood samples. Given that physiological rhythm dysregulation may take time to develop, it may also be beneficial to consider the effects of chronic life stress when examining relationships between distress, circadian rhythms, and tumor-promoting biomarkers. Inconsistencies between our data and the results of prior research may be ascribed to differences in the timing of assessment along the cancer treatment trajectory and/or alternate forms of measurement and quantification. The associations of rest–activity rhythm and CAR with Factor 1 and VEGF may reflect a biology of rapidly growing tumor among

Please cite this article in press as: Cash, E., et al. Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery. Brain Behav. Immun. (2015), http://dx.doi.org/10.1016/j.bbi.2015.02.017

E. Cash et al. / Brain, Behavior, and Immunity xxx (2015) xxx–xxx

presurgical patients who also have poor circadian coordination. Observed associations between the circadian and endocrine factors and biomarkers may simply be a manifestation of the systemic effects of aggressive tumors. Variation in both the circadian variables and the biomarkers might also have been driven by a third confounding factor that we did not measure. Alternatively, given strong extant evidence that cancer progression is accelerated by circadian disruption, it is reasonable to posit that circadian abnormalities are mediators, driving alterations of biomarkers that favor tumor growth. Indeed, disruption of circadian activity and endocrine rhythms may facilitate the secretion of biomarkers that promote tumor growth and progression. More research is needed to explore this possibility, as the current findings provide no information to the direction of the associations or their causation, nor do they discern the biology of the tumor from that of the host. Nevertheless, we believe this is the first in vivo cancer study to demonstrate relationships between rest–activity rhythm, CAR, and the serum biomarkers that formed Factor 1. These novel data suggest biological mechanisms that should be explored in future studies aimed at understanding the prognostic significance of circadian disruption in cancer survival. Support Developmental Cancer Research Award, awarded to authors F.S.D., S.E.S., and D.S. from the Stanford Cancer Institute, Stanford University School of Medicine, Stanford, CA; University of Louisville Intramural Research Incentive Grant for Research on Women awarded to S.E.S. from the University of Louisville Office of the Vice President for Research; Carl & Elizabeth Naumann Startup Fund awarded to F.S.D. Conflict of interest statement All authors declare that there are no conflicts of interest. References Abercrombie, H.C., Giese-Davis, J., Sephton, S., Epel, E.S., Turner-Cobb, J.M., Spiegel, D., 2004. Flattened cortisol rhythms in metastatic breast cancer patients. Psychoneuroendocrinology 29, 1082–1092. Ancoli-Israel, S., Cole, R., Alessi, C., Chambers, M., Moorcroft, W., Pollak, C., 2003. The role of actigraphy in the study of sleep and circadian rhythms. American Academy of Sleep Medicine review paper. Sleep 26, 342–392. Andersen, B.L., Farrar, W.B., Golden-Kreutz, D., Kutz, L.A., MacCallum, R., Courtney, M.E., Glaser, R., 1998. Stress and immune responses after surgical treatment for regional breast cancer. J. Natl. Cancer Inst. 90, 30–36. Antoni, M.H., Lutgendorf, S.K., Cole, S.W., Dhabhar, F.S., Sephton, S.E., McDonald, P.G., Stefanek, M., Sood, A.K., 2006. The influence of bio-behavioural factors on tumour biology: pathways and mechanisms. Nat. Rev. Cancer 6, 240–248. Antonova, L., Mueller, C.R., 2008. Hydrocortisone down-regulates the tumor suppressor gene BRCA1 in mammary cells: a possible molecular link between stress and breast cancer. Genes Chromosomes Cancer 47, 341–352. Arjona, A., Sarkar, D.K., 2008. Are circadian rhythms the code of hypothalamicimmune communication? Insights from natural killer cells. Neurochem. Res. 33, 708–718. Armaiz-Pena, G.N., Lutgendorf, S.K., Cole, S.W., Sood, A.K., 2009. Neuroendocrine modulation of cancer progression. Brain Behav. Immun. 23, 10–15. Bower, J.E., Ganz, P.A., Dickerson, S.S., Petersen, L., Aziz, N., Fahey, J.L., 2005. Diurnal cortisol rhythm and fatigue in breast cancer survivors. Psychoneuroendocrinology 30, 92–100. Bower, J.E., Ganz, P.A., Irwin, M.R., Arevalo, J.M., Cole, S.W., 2011. Fatigue and gene expression in human leukocytes: increased NF-jB and decreased glucocorticoid signaling in breast cancer survivors with persistent fatigue. Brain Behav. Immun. 25, 147–150. Cermakian, N., Lange, T., Golombek, D., Sarkar, D., Nakao, A., Shibata, S., Mazzoccoli, G., 2013. Crosstalk between the circadian clock circuitry and the immune system. Chronobiol. Int. 30, 870–888. Chida, Y., Hamer, M., Wardle, J., Steptoe, A., 2008. Do stress-related psychosocial factors contribute to cancer incidence and survival? Nat. Clin. Pract. Oncol. 5, 466–475. Chida, Y., Steptoe, A., 2009. Cortisol awakening response and psychosocial factors: a systematic review and meta-analysis. Biol. Psychol. 80, 265–278.

11

Claffey, K.P., Brown, L.F., del Aguila, L.F., Tognazzi, K., Yeo, K.T., Manseau, E.J., Dvorak, H.F., 1996. Expression of vascular permeability factor/vascular endothelial growth factor by melanoma cells increases tumor growth, angiogenesis, and experimental metastasis. Cancer Res. 56, 172–181. Clow, A., Hucklebridge, F., Thorn, L., 2010. The cortisol awakening response in context. Int. Rev. Neurobiol. 93, 153–175. Cohen, J., 1988. Statistical Power Analysis for the Behavioral Sciences, second ed. Lawrence Erlbaum Associates, Hillsdale, NJ. Cohen, L., Cole, S.W., Sood, A.K., Prinsloo, S., Kirschbaum, C., Arevalo, J.M., Pisters, L., 2012. Depressive symptoms and cortisol rhythmicity predict survival in patients with renal cell carcinoma: role of inflammatory signaling. PLoS One 7, e42324. Cordova, M.J., Andrykowski, M.A., Kenady, D.E., McGrath, P.C., Sloan, D.A., Redd, W.H., 1995. Frequency and correlates of posttraumatic-stress-disorder-like symptoms after treatment for breast cancer. J. Consult. Clin. Psychol. 63, 981–986. Dantzer, R., Heijnen, C.J., Kavelaars, A., Laye, S., Capuron, L., 2014. The neuroimmune basis of fatigue. Trends Neurosci. 37, 39–46. Dedert, E.A., Lush, E.D., Chagpar, A., Dhabhar, F., Segerstrom, S.C., Spiegel, D., Dayyat, E., Daup, M., McMasters, K., Sephton, S.E., 2012. Stress, coping and circadian disruption among women awaiting breast cancer surgery. Ann. Behav. Med. 44, 10–20. DeSantis, A.S., DiezRoux, A.V., Hajat, A., Golden, S.H., Jenny, N.S., Sanchez, B.N., Shea, S., Seeman, T.E., 2011. Associations of salivary cortisol levels with metabolic syndrome and its components: the multi-ethnic study of atherosclerosis. J. Clin. Endocrinol. Metab. 96, 3483–3492. Dhabhar, F.S., Saul, A.N., Holmes, T.H., Daugherty, C., Neri, E., Tillie, J.M., Oberyszyn, T.M., 2012. High-anxious individuals show increased chronic stress burden, decreased protective immunity, and increased cancer progression in a mouse model of squamous cell carcinoma. PLoS One 7, e33069. Edwards, S., Clow, A., Evans, P., Hucklebridge, F., 2001. Exploration of the awakening cortisol response in relation to diurnal cortisol secretory activity. Life Sci. 68, 2093–2103. Eismann, E.A., Lush, E., Sephton, S.E., 2010. Circadian effects in cancer-relevant psychoneuroendocrine and immune pathways. Psychoneuroendocrinology 35, 963–976. Elinav, E., Nowarski, R., Thaiss, C.A., Hu, B., Jin, C., Flavell, R.A., 2013. Inflammationinduced cancer: crosstalk between tumours, immune cells and microorganisms. Nat. Rev. Cancer 13, 759–771. Escobar, C., Salgado-Delgado, R., Gonzalez-Guerra, E., Tapia Osorio, A., AngelesCastellanos, M., Buijs, R.M., 2012. Circadian disruption leads to loss of homeostasis and disease. Sleep Disord. 2011, e964510. Faul, F., Erdfelder, E., Lang, A.-G., Buchner, A., 2007. G⁄Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behav. Res. Methods 39, 175–191. Fernandes, R., Stone, P., Andrews, P., Morgan, R., Sharma, S., 2006. Comparison between fatigue, sleep disturbance, and circadian rhythm in cancer inpatients and healthy volunteers: evaluation of diagnostic criteria for cancer-related fatigue. J. Pain Symptom Manage. 32, 245–254. Field, A.P., 2005. Discovering Statistics Using SPSS. Sage Publications Ltd., London, England. Filipski, E., Delaunay, F., King, V.M., Wu, M.W., Claustrat, B., Gréchez-Cassiau, A., Francis, L., 2004. Effects of chronic jet lag on tumor progression in mice. Cancer Res. 64, 7879–7885. Filipski, E., King, V.M., Li, X., Granda, T.G., Mormont, M.C., Liu, X., Claustrat, B., Hastings, M.H., Levi, F., 2002. Host circadian clock as a control point in tumor progression. J. Natl. Cancer Inst. 94, 690–697. Flory, N., Lang, E.V., 2011. Distress in the radiology waiting room. Radiology 260, 166–173. Fujimoto, H., Sangai, T., Ishii, G., Ikehara, A., Nagashima, T., Miyazaki, M., Ochiai, A., 2009. Stromal MCP-1 in mammary tumors induces tumor-associated macrophage infiltration and contributes to tumor progression. Int. J. Cancer 125, 1276–1284. Fu, L., Lee, C.C., 2003. The circadian clock: pacemaker and tumour suppressor. Nat. Rev. Cancer 3, 350–361. Goel, H.L., Mercurio, A.M., 2013. VEGF targets the tumour cell. Nat. Rev. Cancer 13, 871–882. Goldsmith, R.E., Jandorf, L., Valdimarsdottir, H., Amend, K.L., Stoudt, B.G., Rini, C., Hershman, D., Neugut, A., Reilly, J.J., Tartter, P.I., Feldman, S.M., Ambrosone, C.B., Bovbjerg, D.H., 2010. Traumatic stress symptoms and breast cancer: the role of childhood abuse. Child Abuse Negl. 34, 465–470. Grivennikov, S., Karin, M., 2008. Autocrine IL-6 signaling: a key event in tumorigenesis? Cancer Cell 13, 7–9. Hanahan, D., Weinberg, R.A., 2011. Hallmarks of cancer: the next generation. Cell 144, 646–674. Hashiguchi, T., Arimura, K., Matsumuro, K., Otsuka, R., Watanbe, O., Jonosono, M., Maruyama, Y., Maruyama, O., Osame, M., 2000. Highly concentrated vascular endothelial growth factor in platelets in Crow–Fukase syndrome. Muscle Nerve 23, 1051–1056. Hegel, M.T., Moore, C.P., Collins, E.D., Kearing, S., Gillock, K.L., Riggs, R.L., Ahles, T.A., 2006. Distress, psychiatric syndromes, and impairment of function in women with newly diagnosed breast cancer. Cancer 107, 2924–2931. Hennig, J., Kieferdorf, P., Moritz, C., Huwe, S., Netter, P., 1998. Changes in cortisol secretion during shiftwork: implications for tolerance to shiftwork? Ergonomics 41, 610–621. Hodge, D.R., Hurt, E.M., Farrar, W.L., 2005. The role of IL-6 and STAT3 in inflammation and cancer. Eur. J. Cancer 41, 2502–2512.

Please cite this article in press as: Cash, E., et al. Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery. Brain Behav. Immun. (2015), http://dx.doi.org/10.1016/j.bbi.2015.02.017

12

E. Cash et al. / Brain, Behavior, and Immunity xxx (2015) xxx–xxx

Hormbrey, E., Gillespie, P., Turner, K., Han, C., Roberts, A., McGrouther, D., Harris, A.L., 2002. A critical review of vascular endothelial growth factor (VEGF) analysis in peripheral blood: is the current literature meaningful? Clin. Exp. Metastasis 19, 651–663. Horowitz, M., Wilner, N., Alvarez, W., 1979. Impact of event scale: a measure of subjective stress. Psychosom. Med. 41, 209–218. Howren, M.B., Lamkin, D.M., Suls, J., 2009. Associations of depression with Creactive protein, IL-1, and IL-6: a meta-analysis. Psychosom. Med. 71, 171–186. Innominato, P.F., Focan, C., Gorlia, T., Moreau, T., Garufi, C., Waterhouse, J., Giacchetti, S., Coudert, B., Iacobelli, S., Genet, D., Tampellini, M., Chollet, P., Lentz, M., Mormont, M., Lévi, F., Bjarnason, G.A., 2009. Circadian rhythm in rest and activity: a biological correlate of quality of life and a predictor of survival in patients with metastatic colorectal cancer. Cancer Res. 69, 4700–4707. Irwin, M.R., Olmstead, R.E., Ganz, P.A., Haque, R., 2013. Sleep disturbance, inflammation and depression risk in cancer survivors. Brain Behav. Immun. 30, S58–S67. Jensen, L.D., 2014. The circadian clock and hypoxia in tumor cell de-differentiation and metastasis. Biochim. Biophys. Acta. Jensen, L.D., Cao, Z., Nakamura, M., Yang, Y., Bräutigam, L., Andersson, P., Cao, Y., 2012. Opposing effects of circadian clock genes Bmal1 and period2 in regulation of VEGF-dependent angiogenesis in developing zebrafish. Cell Rep. 2, 231–241. Jin-no, K., Tanimizu, M., Hyodo, I., Nishikawa, Y., Hosokawa, Y., Doi, T., Okada, Y., 1998. Circulating vascular endothelial growth factor (VEGF) is a possible tumor marker for metastasis in human hepatocellular carcinoma. J. Gastroenterol. 33, 376–382. Kalluri, R., Weinberg, R.A., 2009. The basics of epithelial–mesenchymal transition. J. Clin. Invest. 119, 1420–1428. Karayiannakis, A.J., Syrigos, K.N., Polychronidis, A., Zbar, A., Kouraklis, G., Simopoulos, C., Karatzas, G., 2002. Circulating VEGF levels in the serum of gastric cancer patients: correlation with pathological variables, patient survival, and tumor surgery. Ann. Surg. 236, 37. Kimura, K., Hasiguchi, T., Deguchi, T., Horinouchi, S., Uto, T., Oku, H., Setoyama, S., Maruyama, I., Osame, M., Arimura, K., 2007. Serum VEGF – as a prognostic factor of atherosclerosis. Atherosclerosis 194, 182–188. Koyanagi, S., Kuramoto, Y., Nakagawa, H., Aramaki, H., Ohdo, S., Soeda, S., Shimeno, H., 2003. A molecular mechanism regulating circadian expression of vascular endothelial growth factor in tumor cells. Cancer Res. 63, 7277–7283. Kraemer, H.C., Giese-Davis, J., Yutsis, M., Neri, E., Gallagher-Thompson, D., Taylor, C.B., Spiegel, D., 2006. Design decisions to optimize reliability of daytime cortisol slopes in an older population. Am. J. Geriatr. Psychiatry 14, 325–333. Kronfol, Z., Nair, M., Zhang, Q., Hill, E.E., Brown, M.B., 1997. Circadian immune measures in healthy volunteers: relationship to hypothalamic–pituitary–adrenal axis hormones and sympathetic neurotransmitters. Psychosom. Med. 59, 42–50. Krukowski, K., Eddy, J., Kosik, K.L., Konley, T., Janusek, L.W., Mathews, H.L., 2011. Glucocorticoid dysregulation of natural killer cell function through epigenetic modification. Brain Behav. Immun. 25, 239–249. Küchler, T., Bestmann, B., Rappat, S., Henne-Bruns, D., Wood-Dauphinee, S., 2007. Impact of psychotherapeutic support for patients with gastrointestinal cancer undergoing surgery: 10-year survival results of a randomized trial. J. Clin. Oncol. 25, 2702–2708. Lazure, K.E., Lydiatt, W.M., Denman, D., Burke, W.J., 2009. Association between depression and survival or disease recurrence in patients with head and neck cancer enrolled in a depression prevention trial. Head Neck 31, 888–892. Lethborg, C.E., Kissane, D., Burns, W.I., Snyder, R., 2000. ‘‘Cast adrift’’ the experiencing of completing treatment among women with early stage breast cancer. J. Psychosoc. Oncol. 18, 73–90. Lévi, F., Dugué, P.A., Innominato, P., Karaboué, A., Dispersyn, G., Parganiha, A., Giacchetti, S., Moreau, T., Focan, C., Waterhouse, J., Spiegel, D., 2014. Wrist actimetry circadian rhythm as a robust predictor of colorectal cancer patients survival. Chronobiol. Int., 1–10 Lowrey, P.L., Takahashi, J.S., 2011. Genetics of circadian rhythms in mammalian model organisms. Adv. Genet. 74, 175–230. Lutgendorf, S.K., Cole, S., Costanzo, E., Bradley, S., Coffin, J., Jabbari, S., Rainwater, K., Ritchie, J.M., Yang, M., Sood, A.K., 2003. Stress-related mediators stimulate vascular endothelial growth factor secretion by two ovarian cancer cell lines. Clin. Cancer Res. 9, 4514–4521. Lutgendorf, S.K., Lamkin, D.M., DeGeest, K., Anderson, B., Dao, M., McGinn, S., Zimmerman, B., Maiseri, H., Sood, A.K., Lubaroff, D.M., 2008a. Depressed and anxious mood and T-cell cytokine expressing populations in ovarian cancer patients. Brain Behav. Immun. 22, 890–900. Lutgendorf, S.K., Lamkin, D.M., Jennings, N.B., Arevalo, J.M., Penedo, F., DeGeest, K., Langley, R.R., Lucci 3rd., J.A., Cole, S.W., Lubaroff, D.M., Sood, A.K., 2008b. Biobehavioral influences on matrix metalloproteinase expression in ovarian carcinoma. Clin. Cancer Res. 14, 6839–6846. Lutgendorf, S.K., Sood, A.K., 2011. Biobehavioral factors and cancer progression: physiological pathways and mechanisms. Psychosom. Med. 73, 724–730. Lyon, D.E., McCain, N.L., Walter, J., Schubert, C., 2008. Cytokine comparisons between women with breast cancer and women with a negative breast biopsy. Nurs. Res. 57, 51–58. Mantovani, A., Allavena, P., Sica, A., Balkwill, F., 2008. Cancer-related inflammation. Nature 454, 436–444. McNair, D.M., Lorr, M., Droppleman, L.F., 1971. Manual for the Profile of Mood States. Educational & Industrial Testing Service, San Diego. Miller, A.H., Ancoli-Israel, S., Bower, J.E., Capuron, L., Irwin, M.R., 2008. Neuroendocrine–immune mechanisms of behavioral comorbidities in patients with cancer. J. Clin. Oncol. 26, 971–982.

Mohawk, J.A., Green, C.B., Takahashi, J.S., 2012. Central and peripheral circadian clocks in mammals. Ann. Rev. Neurosci., 445–462 Morin, C.M., Stone, J., Trinkle, D., Mercer, J., Remsberg, S., 1993. Dysfunctional beliefs and attitudes about sleep among older adults with and without insomnia complaints. Psychol. Aging 8, 463. Mormont, M.C., Levi, F., 1997. Circadian-system alterations during cancer processes: a review. Int. J. Cancer 70, 241–247. Mormont, M.C., Waterhouse, J., Bleuzen, P., Giacchetti, S., Jami, A., Bogdan, A., Lellouch, J., Misset, J.L., Touitou, Y., Levi, F., 2000. Marked 24-h rest/activity rhythms are associated with better quality of life, better response, and longer survival in patients with metastatic colorectal cancer and good performance status. Clin. Cancer Res. 6, 3038–3045. Mravec, B., Gidron, Y., Hulin, I., 2008. Neurobiology of cancer: Interactions between nervous, endocrine and immune systems as a base for monitoring and modulating the tumorigenesis by the brain. Semin. Cancer Biol. 18, 150– 163. Palesh, O.G., Collie, K., Batiuchok, D., Tilston, J., Koopman, C., Perlis, M.L., Butler, L.D., Carlson, R., Spiegel, D., 2007. A longitudinal study of depression, pain, and stress as predictors of sleep disturbance among women with metastatic breast cancer. Biol. Psychol. 75, 37–44. Palesh, O., Aldridge-Gerry, A., Zeitzer, J.M., Koopman, C., Neri, E., Giese-Davis, J., Jo, B., Kraemer, H., Nouriani, B., Spiegel, D., 2014. Actigraphy-measured sleep disruption as a predictor of survival among women with advanced breast cancer. Sleep 37, 837–842. Penttinen, H.M., Saarto, T., Kellokumpu-Lehtinen, P., Blomqvist, C., Huovinen, R., Kautiainen, H., Hakamies-Blomqvist, L., 2011. Quality of life and physical performance and activity of breast cancer patients after adjuvant treatments. Psychooncology 20, 1211–1220. Pinquart, M., Duberstein, P.R., 2010. Depression and cancer mortality: a metaanalysis. Psychol. Med. 40, 1797–1810. Reis-Filho, J., Pusztai, L., 2011. Gene expression profiling in breast cancer: classification, prognostication, and prediction. Lancet 378 (9805), 1812–1823. Rich, T., Innominato, P.F., Boerner, J., Mormont, M.C., Iacobelli, S., Baron, B., Jasmin, C., Levi, F., 2005. Elevated serum cytokines correlated with altered behavior, serum cortisol rhythm, and dampened 24-hour rest–activity patterns in patients with metastatic colorectal cancer. Clin. Cancer Res. 11, 1757–1764. Sahar, S., Sassone-Corsi, P., 2009. Metabolism and cancer: the circadian clock connection. Nat. Rev. Cancer 9, 886–896. Salgado, R., Junius, S., Benoy, I., Van Dam, P., Vermeulen, P., Van Marck, E., Dirix, L.Y., 2003. Circulating interleukin-6 predicts survival in patients with metastatic breast cancer. Int. J. Cancer 103, 642–646. Savard, J., Morin, C.M., 2001. Insomnia in the context of cancer: a review of a neglected problem. J. Clin. Oncol. 19, 895–908. Savvidis, C., Koutsilieris, M., 2012. Circadian rhythm disruption in cancer biology. Mol. Med. 18, 1249. Segerstrom, S.C., Boggero, I.A., Smith, G.T., Sephton, S.E., 2014. Variability and reliability of diurnal cortisol in younger and older adults: implications for design decisions. Psychoneuroendocrinology 49, 299–309. Sephton, S.E., Lush, E., Dedert, E.A., Floyd, A.R., Rebholz, W.N., Dhabhar, F.S., Salmon, P., 2013. Diurnal cortisol rhythm as a predictor of lung cancer survival. Brain Behav. Immun. 30, S163–S170. Sephton, S., Sapolsky, R.M., Kraemer, H.C., Spiegel, D., 2000. Diurnal cortisol rhythm as a predictor of breast cancer survival. J. Natl. Cancer Inst. 92, 994–1000. Sephton, S., Spiegel, D., 2003. Circadian disruption in cancer: a neuroendocrine immune pathway from stress to disease? Brain Behav. Immun. 17, 321–328. Seruga, B., Zhang, H., Bernstein, L.J., Tannock, I.F., 2008. Cytokines and their relationship to the symptoms and outcome of cancer. Nat. Rev. Cancer 8, 887– 899. Siegel, R., Ma, J., Zou, Z., Jemal, A., 2014. Cancer statistics, 2014. CA Cancer J. Clin. 64, 9–29. Sood, A.K., Bhatty, R., Kamat, A.A., Landen, C.N., Han, L., Thaker, P.H., Li, Y., Gershenson, D.M., Lutgendorf, S., Cole, S.W., 2006. Stress hormone-mediated invasion of ovarian cancer cells. Clin. Cancer Res. 12, 369–375. Sotiriou, C., Pusztai, L., 2009. Gene-expression signatures in breast cancer. N. Engl. J. Med. 360, 790–800. Spiegel, D., Butler, L., Giese-Davis, J., Koopman, C., Miller, E., DiMiceli, S., Classen, C., Fobair, P., Carlson, R., Kraemer, H., 2007. Effects of supportive-expressive group therapy on survival of patients with metastatic breast cancer: a randomized prospective trial. Cancer 110, 1130–1138. Spiegel, D., Giese-Davis, J., 2003. Depression and cancer: mechanisms and disease progression. Biol. Psychiatry 54, 269–282. Stanton, A.L., Danoff-Burg, S., Cameron, C.L., Bishop, M., Collins, C.A., Kirk, S.B., Sworowski, L.A., Twillman, R., 2000. Emotionally expressive coping predicts psychological and physical adjustment to breast cancer. J. Consult. Clin. Psychol. 68, 875–882. Straif, K., Baan, R., Grosse, Y., Secretan, B., El Ghissassi, F., Bouvard, V., Altieri, A., Benbrahim-Tallaa, L., Cogliano, V., 2007. Carcinogenicity of shift-work, painting, and fire-fighting. Lancet Oncol. 8, 1065–1066. Stone, A.A., Schwartz, J.E., Smyth, J., Kirschbaum, C., Cohen, S., Hellhammer, D., Grossman, S., 2001. Individual differences in the diurnal cycle of salivary free cortisol: a replication of flattened cycles for some individuals. Psychoneuroendocrinology 26, 295–306. Tamura, Y., Matsuda, M., Chiba, S., 2009. Actigraphy in sleep medicine and sleep research. Nihon Rinsho, Jpn. J. Clin. Med. 67, 1576–1580. Thaker, P.H., Sood, A.K., 2008. Neuroendocrine influences on cancer biology. Semin. Cancer Biol. 18, 164–170.

Please cite this article in press as: Cash, E., et al. Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery. Brain Behav. Immun. (2015), http://dx.doi.org/10.1016/j.bbi.2015.02.017

E. Cash et al. / Brain, Behavior, and Immunity xxx (2015) xxx–xxx Trifonova, S.T., Zimmer, J., Turner, J.D., Muller, C.P., 2013. Diurnal redistribution of human lymphocytes and their temporal associations with salivary cortisol. Chronobiol. Int. 30, 669–681. Valkovic´, T., Lucˇin, K., Krstulja, M., Dobi-Babic´, R., Jonjic´, N., 1998. Expression of monocyte chemotactic protein-1 in human invasive ductal breast cancer. Pathol. Res. Pract. 194, 335–340. Van Reeth, O., Weibel, L., Spiegel, K., Leproult, R., Dugovic, C., Maccari, S., 2000. Physiology of sleep (review) – interactions between stress and sleep: from basic research to clinical situations. Sleep Med. Rev. 4, 201–219. Vedhara, K., Tuinstra, J., Miles, J.N., Sanderman, R., Ranchor, A.V., 2006. Psychosocial factors associated with indices of cortisol production in women with breast cancer and controls. Psychoneuroendocrinology 31, 299–311. Veikkola, T., Alitalo, K., 1999. VEGFs, receptors and angiogenesis. Semin. Cancer Biol. 9, 211–220. Volden, P.A., Conzen, S.D., 2013. The influence of glucocorticoid signaling on tumor progression. Brain Behav. Immun. Suppl., S26–S31.

13

Witek-Janusek, L., Gabram, S., Mathews, H.L., 2007. Psychologic stress, reduced NK cell activity, and cytokine dysregulation in women experiencing diagnostic breast biopsy. Psychoneuroendocrinology 32, 22–35. Yang, X., Wood, P.A., Ansell, C.M., Quiton, D.F.T., Oh, E.Y., Du-Quiton, J., Hrushesky, W.J., 2009. The circadian clock gene Per1 suppresses cancer cell proliferation and tumor growth at specific times of day. Chronobiol. Int. 26, 1323–1339. Yasuniwa, Y., Izumi, H., Wang, K.Y., Shimajiri, S., Sasaguri, Y., Kawai, K., Kohno, K., 2010. Circadian disruption accelerates tumor growth and angio/stromagenesis through a Wnt signaling pathway. PLoS One 5, e15330. Zheng, Q., Safina, A., Bakin, A.V., 2008. Role of high-molecular weight tropomyosins in TGF-beta-mediated control of cell motility. Int. J. Cancer 122, 78–90. Zeitzer, J., 2013. Control of sleep and wakefulness in health and disease. Prog. Mol. Biol. Transl. Sci. 119, 137–154.

Please cite this article in press as: Cash, E., et al. Circadian disruption and biomarkers of tumor progression in breast cancer patients awaiting surgery. Brain Behav. Immun. (2015), http://dx.doi.org/10.1016/j.bbi.2015.02.017