Stress and breast cancer: from epidemiology to molecular biology

Stress exposure has been proposed to contribute to the etiology of breast cancer. However, the validity of this assertion and the possible mechanisms involved are not well established. Epidemiologic studies differ in their assessment of the relative contribution of stress to breast cancer risk, while physiological studies propose a clear connection but lack the knowledge of intracellular pathways involved. The present review aims to consolidate the findings from different fields of research (including epidemiology, physiology, and molecular biology) in order to present a comprehensive picture of what we know to date about the role of stress in breast cancer development.


Introduction
Studies have estimated that approximately 50% of breast cancer incidence can be attributed to known genetic, physiologic, or behavioral risk factors [1], with genetic risk factors accounting for 5 to 10% of breast cancer cases [2]. Established physiological and behavioral risk factors for breast cancer include having a fi rst-degree relative with breast cancer, early menarche, late menopause, nulli parity or bearing of fi rst child at a later age, overweight after menopause, certain types of benign breast diseases, alcohol consumption, and long-term use of meno pausal estrogen replacement therapy [3]. In addition to these well-characterized contributors, other factors, whose eff ects have been more diffi cult to evaluate, are suspected of conferring increased breast cancer risk. Th ese factors include smoking, certain aspects of nutrition (meat and fat consumption), physical activity, and psychological stress [4,5].
Th e possible contribution of psychological stress to breast cancer development has been extensively studied. Literature on the topic is not only abundant, spanning several decades, but is scattered between the fi elds of epidemiology, physiology, and molecular biology. Whereas the largest amount of literature focuses on using epidemiologic methods to look for a connection between exposure to stress and subsequent breast cancer diagnosis, this portion of the literature is also the most diffi cult to assess. Studies in this area diff er greatly in their fi ndings, probably due to diff erences in factors such as study design, the eff ects of confounding, type of stress exposure, and timing of stress exposure or stress exposure measurement. On the other hand, animal literature on the topic is simpler to interpret, for the most part pointing to a connection between physiological stress signaling and breast cancer development. Th is literature is limited, however, by a lack of variability in stress exposure parameters and the diff erences in breast physiology and development between mice and humans. Finally, molecular studies demonstrate a clear overlap between intracellular stress signaling and protumorigenic pathways within breast cells, but need to be integrated with other stress-breast cancer research in order to obtain an unambiguous assessment of the potential cause-eff ect relationship.
In the present review, we provide a comprehensive picture of the current state of knowledge in the stressbreast cancer fi eld. We shall fi rst begin with epidemiology (which looks at the stress-breast cancer association in human populations) as the most directly relevant study of the human condition, will then continue with the physiological rationale behind a stress-breast cancer asso ciation, will present physiological research in this fi eld (which examines the direct correlation between experimentally-induced stress and tumor development in animals), and shall end with molecular biology (providing the cellular mechanisms through which stress may contribute to breast cancer development). We will identify the sources of study discrepancy (in the case of epidemiology) or the gaps in knowledge (in the case of physiology and molecular biology) that should be addressed in order to further advance scientifi c

Epidemiologic evidence for a stress-breast cancer association
Epidemiologic research on this topic concentrates on testing for an association between exposure to a stressful stimulus and an outcome of breast tumor diagnosis. Historically, this was attempted primarily with the use of retrospective case-control study designs where incident breast cancer cases are usually matched with populationbased or hospital-based controls. Recollection of expo sure to stress was recorded through questionnaires and assessed for its association with breast cancer. Inter pretation of these studies, however, was complicated by a number of limitations, the most important of which is potential recall bias -subjects are more likely to remember stress exposure if they have been diagnosed with breast cancer.
More recently, a type of prospective study design has been employed in which exposure information is obtained prior to knowledge of breast cancer diagnosis (limited prospective design). In such studies, subjects who have undergone a biopsy for suspected breast disease, but have not yet received a diagnosis, are asked about their previous exposure to stress. Although the goal of this type of study design is to reduce the recall bias associated with case-control studies, it has been observed that in many cases women are able to correctly predict their diagnosis after biopsy [6]. Recall of stressful events may thus still be compromised. Consequently, there has been a recent shift in the fi eld towards the employment of truly prospective designs. Th ese designs aim to determine whether exposure information obtained at the beginning of the study is associated with increased risk of breast cancer determined after a long period of follow-up. Results from some large prospective cohorts have recently become available.
In this section we aim to present an overview of the current state of knowledge by outlining the trends demon strated by the two main meta-analyses on the topic of stress and breast cancer in the fi eld of epidemiology. In addition, with the help of more recently published fi ndings (Table 1), we aim to identify the factors that need to be taken into consideration in order to defi nitively establish whether stress plays a role in breast cancer etiology.
A meta-analytical review of epidemiologic studies in the area of stress and breast cancer incidence was carried out by Petticrew and colleagues, with specifi c testing for an association between bereavement and breast cancer, and other life events (including total number of life events, major life changes, separation, war experiences, serious illness, fi nancial problems, and work problems) and breast cancer [7]. Owing to the scope of the literature at the time, the analysis was limited to case-control and limited prospective designs: 11 studies for bereavement and breast cancer (six case-control studies and fi ve limited prospective studies); 15 studies for other life events and breast cancer (eight case-control studies and seven limited prospective studies). Th e studies spanned the time period between 1966 and 1997. Th is metaanalysis demonstrated no association between bereavement and breast cancer (odds ratio (OR) = 1.06, 95% confi dence interval (CI) = 0.95 to 1.18), but a more than twofold increase in breast cancer risk was associated with other life events (OR = 2.63, 95% CI = 2.34 to 2.96). Limiting the analysis to only the highest quality studies (fi ve for other life events) resulted in no apparent association (OR = 0.8, 95% CI = 0.61 to 1.06).
A more recent meta-analysis was performed by Duijts and colleagues [8]. Twenty-seven studies published between 1966 and December 2002 examining the relation ship between stressful life events and breast cancer risk were analyzed. Th e studies encompassed a wider variety of research designs and included ten retrospective case-control studies, four prospective casecontrol studies, nine limited prospective cohort studies, and four prospective cohort studies. Th e fi ndings concluded that variables signifi cantly associated with breast cancer risk are an increased number of stressful life events (OR = 1.77, 95% CI = 1.31 to 2.40), death of a signifi cant other (OR = 1.37, 95% CI = 1.10 to 1.71), and death of a relative or friend (OR = 1.35, 95% CI = 1.09 to 1.68). Interestingly, a disparity in results was found on the basis of study design. Prospective studies showed a higher summary OR associated with stressful life events than retrospective design studies (OR = 2.46, 95% CI = 0.98 to 6.18 and OR = 1.93, 95% CI = 1.13 to 3.31, respectively). Th is was attributed to the possible presence of recall bias in retrospective studies. In addition, studies that took into account well-established breast cancer risk Th e diff erences between the above-described metaanalytical results refl ect the wide diversity in fi ndings between studies. Th ey also, however, highlight the factors that may be sources of discrepancies and that need to be addressed in future research in order to move this research fi eld forward. Th ese factors and some other possible causes of discordance are the type of study design, control for confounding, the timing of stress exposure, and the type of stress exposure. Further support for the importance of consideration of these factors in study design or analysis is lent by investigations published following the release of the discussed metaanalytical reviews.

Study design and control for confounding
Th e infl uence of study design and control for confounding on results is quantitatively demonstrated by the diff erences in ORs observed in the analysis by Duijts and colleagues for retrospective studies and prospective studies and for studies accounting for confounding versus those that do not [8]. Study design and confounding probably contributed to the diff erence in results between the two meta-analyses presented. At the time of the earlier analysis, fewer prospective studies had been published. In addition, the majority of earlier research does not account for the eff ect of confounding factors. Several factors -such as diet, exercise, and hormone exposure -are associated with both the exposure (stress) and the outcome (breast cancer) of interest, and therefore need to be considered in this context as potential confounders [8][9][10].
Another important variable to take into account is the eff ect of socioeconomic status on the association of stress with breast cancer risk. At this point in time, it is unclear whether stress acts as a mediating factor between socioeconomic status and breast cancer, or whether socioeconomic status and stress are independently correlated with breast cancer risk. Evidence has been presented for both possibilities [11,12]. Th e analysis of stress-breast cancer associations should therefore be carried out in both the presence of and the absence of socioeconomic status in order to rule out a confounding eff ect.

Type of stress
Stress can be acute (short-lived) or chronic (repetitive or occurring over an extended period of time) [13,14]. Th ere appears to be a tendency towards stronger associations being observed in studies looking at certain types of stressful life events and breast cancer incidence, compared with those studies examining chronic stress such as work-related stress, caregiving stress, or everyday/total stress.
Th e eff ect of work-related stress on breast cancer risk was assessed in some recent prospective cohort studies. In a study by Kuper and colleagues, the eff ect of job strain on breast cancer incidence was examined among 36,332 Swedish women participating in the Women's Lifestyle and Health Cohort Study, followed for an average of 13 years [15]. Th e results showed that both low job control and high job demands were associated with breast cancer risk among women working full-time, whereas an even stronger association was observed among women exposed to both of these variables (Table 1). Th ese results were contradicted, however, by results from nurse cohort studies. In a prospective cohort study in which 18,932 women in the Danish Nurse Cohort were assessed for the eff ect of work-related stressors on breast cancer [16], high work pressure, infl uence on job organization and long working hours were not associated with breast cancer risk. High work tempo/speed showed an association with increased risk but no dose-response eff ect. Similarly, the Nurses' Health Study showed no association between high-strain jobs and breast cancer risk [17]. Since the above studies employ a similar design, the disparity in fi ndings regarding work stress suggests that the relationship of this type of stress to breast cancer is diffi cult to assess accurately, and may be aff ected by the specifi c measurement method of work stress or by some of the other sources of study discrepancy discussed.
Caregiving stress was examined in the Nurses' Health Study in a cohort of 32,826 women followed for 8 years.
No association between caregiving stress and breast cancer was observed [18].
Th e existence of a relationship between everyday/total stress and breast cancer risk has also proven diffi cult to establish. Participants in the Copenhagen City Heart Study who had reported high levels of stress at baseline were found to have a lower risk of developing breast cancer [19]. A study examining lifetime social adversity within 11,467 women in the European Prospective Investi gation into Cancer -Norfolk cohort found no association for specifi c measures of social adversity, such as diffi culties reported in childhood, and perceived stress within the previous 10 years [20]. Th ese results are opposed by the fi ndings that moderate and high daily stress showed a mild correlation with higher breast cancer risk in the West of Scotland Collaborative Study [21], and that self-reported stress in the 5 years preceding the study baseline is associated with breast cancer risk [11].
In contrast to the above stress categories, stress induced by life events shows a more consistent association with breast cancer risk in prospective studies. For example, a life event-breast cancer association was observed in the Finnish Twin Cohort prospective study [22]. An associa tion for life events with breast cancer risk was also observed in a recent registry study [23], in a limited prospective design study [24], and in the Women's Health Initiative cohort study [10]. Th e European Prospective Investigation into Cancer -Norfolk study, however, found no association for life events within the 5 years previous to study baseline [20], demonstrating that factors other than the type of stress also contribute to study discrepancy.

Timing of stress exposure
Variability between studies may also stem from the distinctive time windows of relevant exposure to stress examined. Th e latency period between stress exposure and breast cancer initiation is unknown. In general, the process of breast cancer development is estimated to occur over 10 to 20 years [25,26]. In agreement with this, a study by Lillberg and colleagues found that breast cancer risk is most strongly correlated with life events that have occurred within 11 years prior to diagnosis [27]. Since the time periods of exposure examined in diff erent studies vary widely, some of the assessed exposures may fall outside the biologically relevant time, thus leading to diff erent study results.
Another time-associated factor that needs to be taken into consideration is the possibility that exposure to stress at certain periods of development may have a stronger impact on breast physiology. Women diagnosed with breast cancer were found to be signifi cantly more likely to have undergone stress due to losses or social defi cits in childhood and adolescence, for example [28]. Similarly, death of a mother during a girl's childhood was found to be signifi cantly associated with increased breast cancer risk in a prospective cohort in the Baltimore Epidemiologic Catchment Area Study [29]. Finally, in a registry study and a prospective cohort study looking at association of stress caused by exposure to the Holocaust or to the siege of Leningrad with breast cancer risk or mortality, respectively, stress among women younger than 18 years old was found to be associated with the highest risk and mortality of breast cancer (although the association in the Leningrad study did not reach statistical signifi cance) [30,31]. Th ese studies also suggest that the cancer-causing eff ects of stress may be fairly specifi c to the breast. In both cases, all types of cancer were examined, and breast cancer showed the strongest association with stress.

Conclusions from the epidemiologic literature
Epidemiologic evidence in the area of stress and breast cancer outlines several trends. Firstly, the eff ect of stress appears to be relatively breast specifi c, possibly due to the biological roles of stress signaling in the mammary gland (described in the next section).
Secondly, the type of stress examined aff ects the study results in this area. Th e stress most strongly associated with increased breast cancer risk appears to be stress induced by major life events, whereas fi ndings regarding work-related, caregiving, or everyday stress vary considerably. Th e dependence of results on the stress type may stem from variable biological eff ects associated with diff erent types of stress, an area that needs to be further investigated with the help of physiological methods.
Th e timing of stress exposure also has an eff ect on breast cancer risk, with early life stress exhibiting the strongest association with breast cancer.
Finally, a greater number of prospective studies is necessary for each subcategory of stress or timing of stress exposure in order to more defi nitively understand the stress-breast cancer relationship.

Biological plausibility of a stress-breast cancer association
Th e plausibility of a stress-breast cancer association stems from two important physiological roles of the stress hormone cortisol. Cortisol plays an essential part in mammary gland development and function, which may sensitize mammary tissues to modulations in cortisol signaling in the presence of stress. It also has an impact on certain aspects of estrogen activity in the mammary gland, which may initiate protumorigenic changes during periods of stress.

The physiological stress response
Stress is defi ned as 'an alteration in the body's hormonal and neuronal secretions caused by the central nervous system in response to a perceived threat' [32,33]. Th e long-term response to psychosocial stressors in humans consists of activating the HPA axis of hormonal signaling [34]. Cortico tropin-releasing hormone produced in the hypothalamus stimulates the release of adrenocorticotropic hormone from the anterior pituitary. Adrenocorticotropin, in turn, signals the adrenal cortex to produce the 'stress hormone' , cortisol. Cortisol generates a physical response to the stress signal by binding to its cytoplasmic receptor, GR, and promoting protein, lipid, and carbo hydrate catabolism [35,36].
In its ordinary physiological role, cortisol has protective eff ects on the organism by regulating immune function, promoting memory of dangerous events, increasing blood pressure and heart rate to meet the physical demands of a fi ght or fl ight response, and making fuel available for sustaining increased physical activity [37]. Prolonged stress-response conditions similar to those stimulated by stressful life events, however, have been shown to predispose for illnesses such as hypertension, atherosclerosis, osteoporosis, immune dysfunction, and cancer in laboratory studies [37][38][39]. Breast cancer development has been proposed to be correlated with such prolonged stress exposure [39]. As pointed out above, this may be a function of the natural role of cortisol in the development and activity of the mammary gland.

Role of stress signaling pathways in the mammary gland
Th e mammary gland begins to form early during embryogenesis and continues to develop in defi ned stages that are correlated with sexual development and reproduction [40][41][42]. Laboratory studies have demonstrated that cortisol plays a physiological role in the mammary gland mainly in the latter part of pregnancy and during lactation (Figure 1). At that stage, placental lactogens stimulate DNA synthesis in the mammary cells, and cortisol induces the formation of the rough endoplasmic reticulum, where milk proteins will be synthesized [42]. Prolactin release upon birth causes lobular diff erentiation, and the secretion of early milk proteins such as βcasein. Cortisol, on the other hand, predominantly regulates the expression of late milk proteins, although it has also been shown to play an important role in the regulation of β-casein expression [43].
Th ese breast-specifi c functions of cortisol signaling have been confi rmed by studies in mice defective for the intracellular mediator of cortisol, the GR. Th ese studies have demonstrated that GR activity is involved in the ductal development of the virgin mammary gland [44], in lobuloalveolar mammary gland development during pregnancy (via the induction of proliferation) [45], and in stimulating milk production during lactation. Owing to the natural functions of cortisol in regulating mammary gland biology, its misregulation and prolonged presence during periods of stress is likely to expose breast cells to the activation of downstream biological pathways outside their normal context. Studies indicate that stressful life events, as well as job stress, produce elevated cortisol levels years after the exposure [46][47][48]. Th is elevation would signifi cantly prolong the eff ect of cortisol on the mammary gland. In addition, stress causes permanent changes in HPA-axis responsiveness [49,50], which may compromise the body's ability to buff er any negative eff ects of subse quent stress exposure.
In concordance with directly aff ecting the molecular signaling pathways of mammary cells, cortisol may indirectly contribute to breast tumorigenesis by altering the generation or activity of estrogen. A large body of evidence suggests that estrogen plays a key role in breast cancer etiology [51,52]. One important source of estrogen in the breast is the conversion of adrogens to estrone in stromal cells by the enzyme aromatase [53,54]. Research has shown that aromatase activity [55] and mRNA levels [56] are increased in stromal areas surrounding mammary tumors and that aromatase is able to stimulate the proliferation of breast cancer cells in vitro [57]. Cortisol has been characterized as an inducer of aromatase activity when in the presence of serum [58] or plateletderived growth factor [59]. Treatment of stromal breast cells with the glucocorticoid inhibitor RU486 results in loss of aromatase activation by the aromatase inducer dibutyryl-cAMP [60], demonstrating a necessity for cortisol participation in this process. Stimulation of breast cancer development by prolonged or elevated presence of cortisol during periods of stress may there fore partially occur as a result of augmented estrogen production.
Research in our laboratory has also demonstrated a functional interaction between cortisol and estrogen signaling. We had previously found that cortisol treatment results in downregulation of expression of the gene encoding the tumor-suppressor breast cancer susceptibility gene 1 (BRCA1). Th e BRCA1 promoter is known to contain an estrogen receptor response element and to be activated by estrogen activity [61]. Since BRCA1 is capable of blocking estrogen-initiated cell proliferation and estrogen signaling pathways [62], the upregulation of BRCA1 expression by estrogen is thought to represent a feedback mechanism by which rapidly proliferating cells control their growth [61]. Factors that obstruct this mechanism are expected to produce unchecked cell proliferation. We have determined that cortisol may act as one such factor. Cortisol treatment of breast cells resulted in loss of estrogen-induced stimulation of BRCA1 expression [63].
Th e above-described functions of cortisol may contribute to breast cancer development in response to stress. Th eir eff ect, however, may be dependent on individual genetic stress susceptibility. Specifi c polymorphisms in the GR gene have been shown to play a role in an individual's response to stress. For example, increased GR responsiveness to glucocorticoid signaling has been indicated for an aspartic acid to serine change in codon 363 of the GR (N363S), found in 3 to 7% of Caucasians [64]. Th is heightened sensitivity to stress has been correlated with a number of cortisol-specifi c metabolic changes [64][65][66][67][68], and is likely to amplify the eff ect of cortisol signaling on the mammary gland. Importantly, such a genetic predisposition to stress-induced breast cancer development is supported by the observation that a highly polymorphic dinucleotide CA repeat (D5S207), suspected to occur in linkage disequilibrium with a polymorphism within the GR locus, is associated with an increase in breast cancer risk [69].

Physiological studies on stress and breast tumor growth
As discussed in the previous section, multiple variables may impact the magnitude or measurement of an association between exposure to a stressful stimulus and breast cancer development. Th is complexity makes an unequivocal relationship between stress and breast cancer diffi cult to demonstrate using epidemiologic methods. Laboratory investigations can therefore comple ment epidemiologic fi ndings by allowing for a measurement of the impact of stress exposure on mammary gland physiology in a controlled environment. Such studies (summarized in Table 2) have been carried out in animals by employing the method of stress induc tion via social isolation. Social isolation and a change in social environment are methods previously correlated to an increase in the release of adrenocorticotropic hormone and corticosterone [70], and to enhanced tumor size [71,72]. Th e use of these techniques by the studies outlined below has allowed for the establishment of a contributing eff ect of stress on breast tumor growth and incidence.
In a study using male mice transplanted with an androgen-responsive mammary tumor, Weinberg and Emerman showed that socially isolated animals, additionally exposed to acute daily novelty stress (changes in housing conditions), exhibit a marked increase in tumor growth as compared with group-housed animals [73]. Th ese results were confi rmed in a follow-up study, in which mammary-tumor-transplanted mice switched from group housing to individual housing were found to have faster tumor growth rates as compared with mice switched from individual housing to group housing [74]. Other hormones involved in the diff erent developmental stages are also listed. Estrogen and progesterone promote ductal system proliferation during puberty. However, the DNA binding function of the glucocorticoid receptor also appears to be required. During pregnancy, cortisol contributes to lobuloalveolar development of the mammary gland, in conjunction with estrogen and progesterone. Prolactin and cortisol prepare the mammary cells for lactation and stimulate milk protein production following parturition. In addition, cortisol contributes to the maintenance of lactation by suppressing involution. ER, endoplasmic reticulum. In addition to tumor growth, social isolation was shown by Hermes and colleagues to have an eff ect on lifetime risk of mammary tumor incidence [75]. Socially isolated female Norway rats followed for spontaneous mammary tumor development were found to exhibit a relative risk of 3.3 for developing at least one malignant tumor as compared with group-housed animals, although there was no signifi cant diff erence in the risk of developing a palpable tumor mass in general. Th e tumor burden in isolated rats was also 84 times higher. Similar characteristics of tumor development were observed in work by Williams and colleagues in socially isolated mice [76]. Th e study demonstrated that tumor incidence and tumor size are both increased in isolated mice when compared with group-housed controls, and that this is correlated with a higher number of poorly diff erentiated adenocarcinomas.
Interestingly, animals exposed to social isolation early in life in the study by Hermes and colleagues showed a hyperactive response to psychological stressors later in life (this was also observed by Williams and colleagues), as well as sustained changes in HPA-axis signaling [75]. Hyperactivity of the stress response was, in turn, correlated with increased tumor burden. It is tempting to speculate that this model presents a situation analogous to the eff ect of stressful life events on the risk of developing breast cancer later in life. In fact, human studies indicate that HPA-axis signaling in response to a stressor is permanently altered following adverse life events, such as childhood abuse (although the direction  [104] lymphocytes in response to psychiatric patients distressed individuals X-ray irradiation Glaser and DNA damage DNA repair capacity during Medical students during Increase in DNA repair capacity colleagues [105] period of stress examination during period of stress, possibly as initial response to increased DNA damage Cohen and DNA damage Levels of O6-methylguanine-Stressed vs. control rats Levels of DNA repair enzyme colleagues [106] methyltransferase following are reduced in stressed rats stress exposure Fischman and DNA damage Rate of sister chromatid exchanges Stressed vs. control rats Increase in sister chromatid colleagues [107] in response to γ-irradiation, exchanges in stressed rats; mitomycin-C in the presence of increased susceptibility to environmental stressors mutagenesis of change diff ers between studies) [49,50,77], and that basal cortisol levels may be aff ected by early and late life events [78,79]. One mechanism for these long-term HPA-axis signaling changes has been revealed through recent investigations employing lack of grooming as a stressor. In these studies, rat pups prevented from being groomed exhibited a heightened stress response when exposed to stressors as adults [80,81]. Th is augmented response was associated with methylation of a far-upstream element in the promoter of the GR gene in cells of the hippocampus, subsequent low levels of GR gene expression, and longterm alteration of HPA-axis signaling due to reduced sensitivity to negative feedback control by glucocorticoids [82,83].
In addition to its eff ects on GR expression, social isolation was demonstrated to result in global changes in gene expression, and to cause the activation of certain glycolytic and lipogenic metabolic pathways previously correlated with tumor development [76]. Th ese fi ndings provide some of the possible pathways aff ected by stress, linking physiological and molecular literature on the topic of stress and breast cancer.

Future directions in physiological research
Th e animal studies discussed above for the most part demonstrate a connection between stress exposure and breast tumor incidence and growth, corroborating the fi ndings of a subset of the epidemiologic literature. Th is fi eld suff ers from a limited number of studies, however, and needs to be expanded into research directions that would strengthen its applicability to human biology. Th e most important step forward would be to extend the use of physiological measures of stress to human studies. Although, a retrospective epidemiologic design does not allow for a measurement of stress signaling parameters at the time of stress exposure, certain prospective designs may allow for incorporation of physiological stressmeasurement techniques. For example, physio logical measures can be employed in prospective studies of longterm stress (such as work strain or caregiving) where the exposure is ongoing at the time of study initiation, or to test for increased cortisol levels in the presence of a history of life events. Validated methods for stress quantifi cation that can be applied include the measurement of morning and evening salivary or blood cortisol levels [46,84]. Th e detection of anti-Epstein-Barr virus antibodies in the blood has also been implicated as an indirect biological marker of stress [85,86].
An important parallel that emerges between animal and human physiology is the specifi c role of early life stress on breast cancer development. Epidemiologic studies, however, demonstrate that in addition to the timing of stress exposure, the stress-breast cancer asso ciation depends on the type of stress. A limitation of the physiological literature is therefore the lack of variability in stress measures. Th e primary type of stress employed in animal studies when studying tumor development is social isolation. Th is type of stress has been proposed to be representative of chronic stress exposures in humans [87]. Th e correlation between species-specifi c stress signaling responses, however, requires validation. Social stress has also been shown to aff ect the development of the mammary gland during puberty, and may therefore not be appropriate for usage in the study of early life events [88]. In addition, there is a need for animal studies repre sentative of other human stressors, such as life events. Th ese studies would allow for the draw of more direct parallels between animal and human fi ndings.

Molecular studies
Th e contribution of molecular biology to the stressbreast cancer fi eld of research is twofold. Firstly, molecular biology contributes to the understanding of stress by allowing the elucidation of the intracellular molecular changes which occur in response to stress signaling. Secondly, by examin ing whether stress impacts tumorigenic pathways within the mammary gland, molecular biology tests the bio logical plausibility of the stress-breast cancer association observed in some of the epidemiologic literature. Molecular evidence currently points to apoptosis modulation [89], regulation of immune function, and changes in DNA repair mechanisms as the main biological mechanisms through which stress contributes to cancer development.

Stress and apoptosis in the mammary gland
Consistent with its role in stimulating lactation, cortisol has been shown to suppress apoptosis and involution in the mammary gland [90] and to stimulate prosurvival pathways in nonmalignant mammary epithelial cells in vitro [91]. Th ese observations indicate that prolonged exposure to stress may facilitate tumor progression by suppressing the removal of genetically altered cells. In accordance with these genetic instability-inducing eff ects, glucocorticoids have been found to inhibit chemotherapy-induced apop tosis [92] and to stimulate invasiveness [93] and adhesion [94] of human breast tumor cells. Furthermore, the GR was found to be overexpressed in 94.4% of metaplastic carcinomas and in 92.3% of malignant phyllode tumors [95], implying a link between increased intracellular stress signaling and the development of those types of breast malignancies. Interestingly, some studies have pointed to an inhibition of the growth of certain mammary tumor cell lines in the presence of cortisol [96,97], which may indicate tumortype specifi city of the cortisol eff ect.
Few breast tissue-specifi c molecular mechanisms have been proposed for the antiapoptotic eff ect of stress signaling in such cells. Th e GR, however, has been shown to be involved in several apoptosis-related pathways in general, which have been reviewed previously [98,99]. In the mammary gland in particular, glucocorticoids have been found to induce both the mRNA levels of the activator protein-1 family members c-fos, jun B, and jun D, and activator protein-1 DNA binding activity [90]. Th is induction, in turn, leads to the inhibition of activator protein-1 target genes and the suppression of apoptosis and of mammary gland involution. In malignant mammary cells, the synthetic glucocorticoid dexamethasone has been found to repress mitogen-activated protein kinase-induced apoptosis by upregulating the expression of mitogen-activated protein kinase phosphatase-1 [92]. Th e GR has also been shown to signal through the serum and glucocorticoid-regulated kinase-1 to decrease the proapoptotic activity of Fork head transcription factor 3a in the breast cancer cell line SKBR-3 [100]. Finally, the synthetic glucocorticoid dexa metha sone has been found to inhibit TNFα-induced apoptosis by suppressing the degradation of several inhibitors of apoptosis proteins, including cIAP1, cIAP2, and XIAP [101].

Stress and immune detection of transformed cells
Stress may also contribute to mammary tumor development by aff ecting immune system function and the elimination of transformed mammary cells. Th e eff ect of stress signaling on immunity has been extensively studied and was recently reviewed by Webster Marketon and Glaser [102]. With respect to cancer development, cortisol release in response to psychological stress has been suggested to produce a shift in the levels of Th 1 and Th 2 cytokines towards a Th 2 response, a decline in natural killer cell activity, and a decrease in IFNγ production, all of which would aff ect the ability of the immune system to detect and respond to the presence of tumor cells [103]. Holden and colleagues have also suggested a causal immunological model in which a stress-related increase in TNFα results in decreased activity of tyrosine phos phatase and in a consequently diminished expression of the class-I MHC antigen on the surface of malignant cells, allowing transformed cells to escape detection by the immune system [38].

Stress and DNA repair capacity
Alternatively, psychological stress has been demonstrated to modulate DNA repair capacity and to promote mutagenesis (Table 2). For example, highly distressed psychiatric patients show signifi cantly lower DNA repair capacity following X-ray irradiation then less distressed individuals [104]. Also, in peripheral blood lymphocytes and in spleens from stressed rats, levels of an important DNA repair enzyme, O6-methylguanine-methyltransferase, are reduced as compared with those obtained from control rats [105]. Interestingly, a study of medical students during examination stress demonstrated an increase in DNA repair capacity during the period of stress, which the authors interpreted as an initial cellular response to an increase in DNA damage [106]. Such promotion of mutagenesis in response to stress has been observed in other studies. Rats exposed to environmental stressors exhibit a marked increase in the rate of sister chromatid exchanges (an event correlated with both mutagenesis and an increase in cancer risk) [107], and are more susceptible to mutagenesis induced by γ-irradiation and mitomycin-C [108]. Similarly, a higher rate of DNA mutation occurrence due to oxidative damage is found in rats undergoing psychological distress [109].
Some molecular in vitro results have recently become available with regard to the involvement of stress signaling in DNA repair. In murine 3T3 cells, cortisol treatment was found to increase intracellular DNA damage by approximately fi vefold and to interfere with the repair of ultraviolet-induced DNA damage [110]. In addition, a gene array experiment demonstrated that cortisol modulates the expression of genes involved in DNA damage signaling, including the proto-oncogene Cdc25A (involved in cell cycle delay following DNA damage) and the genes coding for the DNA damage sensors Chk1, Rad53 and Rad9 [110].
Our own studies have shown in mouse mammary cell lines that cortisol downregulates the expression of the BRCA1 gene [63]. As an important regulator of both DNA repair and apoptosis, the expression of this gene is critical in the development of breast cancer -demonstrated by its frequent mutation in familial breast cancers [111,112] and its downregulation in many sporadic breast tumors [113]. Th e loss of 50% of BRCA1 function in BRCA1 mutation carriers leads to an altered profi le of gene expression, which resembles the genetic profi le of BRCA1-associated hereditary carcinomas [114]. Th e approximately 50% decrease in BRCA1 levels we observed in response to elevated cortisol levels is therefore biologically signifi cant. To date, BRCA1 regulation is the only breast-specifi c DNA repair pathway that has been linked to cortisol signaling.

Future directions in molecular research
Many of the molecular systems described need to be tested in the context of mammary gland biology in order to validate their specifi c contribution to breast cancer development in response to stress. Another future direction in this fi eld should be the application of molecular tools to other scientifi c disciplines studying stress and breast cancer. Th is expansion would allow for a demonstration of the biological signifi cance of diff erent types of stress exposure (that is, the molecular pathways aff ected) or for obtaining a quantitative measure of the eff ect of stress on mammary cells. For example, as previously discussed, the seemingly diff erent epidemiologic eff ects of work strain or life events on breast cancer risk may stem from discor dance in the molecular mechanisms involved. Th is can be tested by looking at changes in relevant gene expression (for example, expression of apoptotic or DNA repair genes) in response to each type of stress. Th e observed eff ect of stress timing on breast cancer risk can also be explained by studying the eff ect of stress on gene expres sion at diff erent stages of  Alternatively, the eff ect of stress exposure on known GR targets or changes in GR gene methylation patterns can be used to confi rm stress exposure occurrence in both animal and human studies.

Summary and conclusion
Elucidation of the factors contributing to the incidence of breast cancer is of crucial importance for the development of therapeutic or preventative strategies targeting the disease. Exposure to psychological stress and stressrelated cortisol release have been proposed to be associated with increased breast cancer risk. In the present review we have provided an overview of the current knowledge on stress signaling and breast cancer, incorporating fi ndings from the several disciplines involved (a summary model is presented in Figure 2). Epidemiologic evidence has been diffi cult to interpret due to use of diff erent measures of stress and other study methods that lead to inconsistent fi ndings. Recent reviews have helped to pinpoint the areas of strongest association and have demonstrated that fi ndings depend on the type of study design and the method of exposure measurement. Th ese reviews, however, have also demonstrated a need for more well-devised studies -ideally, prospective cohorts that take into account confounding factors and focus on specifi c types and timing of stress exposure. Th e large population-based cohorts currently underway are likely to soon generate more defi nitive answers.
Consolidation of knowledge in the area of physiological research demonstrates that laboratory studies, in the most part, support a stress-breast cancer association. Th is is consistent with fi ndings from epidemiologic studies looking at the eff ect of life events on breast cancer risk. Since epidemiologic evidence is strongest for a specifi c type of stress exposure, a need exists for a variation in the types of stress methods used in physiologic studies.
Molecular work on the eff ects of stress in breast cells in particular is still in its early stages. Some initial eff orts have been applied towards identifying the intracellular pathways aff ected by cortisol. Th e suppressive eff ect of cortisol on the apoptotic ability and DNA repair capacity of cells, as well as its negative eff ect on immunity, suggests that a connection between stress signaling and tumor development is biologically plausible. Additional research is needed to identify the particular DNA repair and apoptosis genes aff ected by cortisol and to verify whether pathways aff ected in other cell types are also targeted by cortisol in breast cells.
Th e study of stress and mammary gland malignancy may be greatly aided by the employment of an interdisciplinary approach to research in this fi eld. As knowledge of stress signaling in the breast increases, these new fi ndings can be applied to other research areas. For example, assessment of the activity of GR-related genes or binding partners could be included in physio logical studies in order to establish a clear path between external stress exposure and intracellular eff ects, or a combination of physiological measures of stress, such as salivary cortisol levels, with epidemiologic methods could be included to provide a more accurate assessment of stress exposure at various relevant periods of life. Th ere is a need for under standing the diff ering physiological eff ects of types or times of stress exposure. Th is need can be addressed with the use of both physiological methods to measure hor monal eff ects and molecular methods to identify diff er ences in gene regulation. Finally, the use of molecular measures of stress susceptibility -such as the identifi cation of GR polymorphisms in epidemiologic studies -may help to eliminate a potentially important source of study discrepancy in that fi eld.