30 Nov 2001 107 AR AR146-04.tex AR146-04.SGM LaTeX2e(200105.docx

30 Nov 2001 10:7 AR AR146-04.tex AR146-04.SGM
LaTeX2e(2001/05/10) P1: GJC
Annu. Rev. Psychol. 2002. 53:83–107
Copyright c�2002 by Annual Reviews. All rights reserved
EMOTIONS, MORBIDITY, AND MORTALITY:
New Perspectives from Psychoneuroimmunology
Janice K. Kiecolt-Glaser1, Lynanne McGuire2,
Theodore F. Robles3, and Ronald Glaser4
1,2Department of Psychiatry and 4Department of Molecular
Virology, Immunology, and
Medical Genetics, The Ohio State University College of
Medicine, 1670 Upham Drive,
Columbus, Ohio 43210; e-mail: [email protected]
3Department of Psychology, The Ohio State University,
Columbus, Ohio 43210;
e-mail: [email protected]
Key Words depression, immune function, social support,
interleukin 6
■ Abstract Negativeemotionscan intensifyavarietyofhealth
threats.Weprovide
a broad framework relating negative emotions to a range of
diseases whose onset
and course may be influenced by the immune system;
inflammation has been linked
to a spectrum of conditions associated with aging, including
cardiovascular disease,
osteoporosis,arthritis,
type2diabetes,certaincancers,Alzheimer’sdisease, frailtyand

functional decline, and periodontal disease. Production of
proinflammatory cytokines
thatinfluencetheseandotherconditionscanbedirectlystimulatedby
negativeemotions
andstressfulexperiences.Additionally,negativeemotionsalsocontr
ibute toprolonged
infection and delayed wound healing, processes that fuel
sustained proinflammatory
cytokineproduction.Accordingly,wearguethatdistress-
relatedimmunedysregulation
may be one core mechanism behind a large and diverse set of
health risks associated
with negative emotions. Resources such as close personal
relationships that diminish
negativeemotionsenhancehealth inpart through their positive
impacton immuneand
endocrine regulation.
CONTENTS
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . 84
NEGATIVE EMOTIONS, MORBIDITY,
AND MORTALITY: THE EVIDENCE . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . 85
Depression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . 85
Anxiety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . 86
Hostility/Anger . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . 87
PATHWAYS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . 87
Morbidity, Mortality, and Aging: Central Immunological
Mechanisms . . . . . . . . . . 87

Emotions and Immune System Alterations . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . 89
Emotions and Neuroendocrine Alterations . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . 91
Health Behaviors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . 92
0084-6570/02/0201-0083$14.00 83
LaTeX2e(2001/05/10) P1: GJC
84 KIECOLT-GLASER ET AL.
Allostatic Load: Conceptual Similarities and Differences . . . . .
. . . . . . . . . . . . . . . 93
VULNERABILITY AND RESILIENCE FACTORS . . . . . . . . . .
. . . . . . . . . . . . . . . . 94
Sociodemographic Variables . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . 94
Personality Traits and Coping . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . 97
Social Relationships . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . 98
CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . 99
INTRODUCTION
The idea that emotions are linked with morbidity and mortality
has existed for
over two millennia (Sternberg 1997). Hippocrates (c. 500 B.C.)
theorized that

health was related to the balance of four bodily humors, which
contributed to
specific temperaments. Galen (A.D. 131–201) took this idea
further, proposing
that a balance of the “passions” was essential for physical
health. Indeed, severe
emotional reactions were considered causes of diseases such as
stroke, birth de-
fects, asthma, ulcers, and, ultimately, even death (Sternberg
1997). These beliefs
persisted through the medieval period and the early
Renaissance; in The Anatomy
of Melancholy, Robert Burton (1621/1893) wrote, “the mind
most effectually
works upon the body, producing by his passions and
perturbations miraculous
alterations . . . crueldiseasesandsometimesdeath
itself.”Although this ideadom-
inatedmedicalpracticeformuchofearlycivilization, in
themodernerathescience
of the biological bases of health and disease has far surpassed
the science of emo-
tions.Inthisreviewweconsidernewevidencethatsuggestshownegati
veemotions
may contribute to disease and death through immune
dysregulation.
We first address the evidence that negative emotions are related
to morbidity
andmortality.Wehighlight theconsequencesofdepression,
anxiety, andhostility,
threebroademotions thathavebeen linked
toverifiablehealthoutcomes;although
therearemanypotential commonpathsamong
thenegativeemotions, there isalso
evidence that different emotions may make unique contributions

to some disease
processes (Leventhal et al. 1998). Next we consider key
pathways, focusing on
a central immunological mechanism that serves as a gateway for
a range of age-
associated diseases, the dysregulation of proinflammatory
cytokine production.
In the final section we consider vulnerability and resilience
factors, including
sociodemographic variables, personality traits and coping,
social relationships,
and positive emotions.
Althoughit isclear thatnegativeemotionscan
intensifyawidevarietyofhealth
threats, positive emotions have received considerably less
attention, perhaps re-
lated to theprevailingviewofphysical andmentalhealthas
theabsenceofdisease
and negative emotions (Ryff & Singer 1998), as well as the fact
that positive emo-
tionsarefewerinnumberandlessdifferentiatedthannegativeemotion
s(Ellsworth
& Smith 1988). Indeed, although a substantial empirical
literature exists for “de-
pression” and objective measures of health, almost none exists
for “happiness”
and health, and thus we concentrate on the former.
This review concentrates on the pathways from negative
emotions to illness
anddeath; theeffectsofdiseaseonemotionaldistresswill
notbeaddressed inany

LaTeX2e(2001/05/10) P1: GJC
EMOTIONS AND PSYCHONEUROIMMUNOLOGY 85
detail, although the relationships are clearly bidirectional.
Indeed, cytokines have
substantial effects on the central nervous system, including
production and en-
hancement of negative moods, physical symptoms including
lethargy and fatigue,
and a range of sickness behaviors from shivering to loss of
appetite (Leventhal
et al. 1998, Watkins & Maier 2000); accordingly, negative
emotions may also
reflect a prodromal or active disease process (Leventhal et al.
1998). In fact, al-
though we focus on the impact of emotions on immune and
endocrine responses
and disease, there is plausible evidence that the immune system
has a role in the
neuroendocrine and behavioral features of both depressive and
anxiety disorders
(Miller 1998).
NEGATIVE EMOTIONS, MORBIDITY,
AND MORTALITY: THE EVIDENCE
Depression
Depression is themostcommonpsychiatric
illness,andbothmajordepressionand
subthresholddepressivesymptomscarrysubstantialhealthrisks.An
umberofwell-
controlled prospective studies have linked depressive symptoms
with coronary

heart disease (CHD), the leadingcauseofdeath in
theUnitedStates.Forexample,
a 13-year prospective study suggested that individuals with
major depression had
a 4.5 times greater risk of a heart attack compared with those
with no history of
depression(Prattetal.1996).Depressivesymptomsalsoplacepatient
sat jeopardy;
across a series of studies, healthy individuals who had elevated
depression scores
at baseline had a 1.5- to 2-fold increased risk for a first heart
attack (Glassman
& Shapiro 1998). Not surprisingly, patients who had preexisting
cardiovascular
disease also had poorer outcomes if they were depressed
(Glassman & Shapiro
1998); mortality among patients who had suffered a heart attack
was four times
higher among the depressed than the nondepressed (Frasure-
Smith et al. 1993).
Onerecentwell-
controlledstudyfoundthatchronicdepressedmoodwaslinked
to cancer risk; after adjusting for sociodemograhic variables and
risk factors, the
hazardratioacrossarangeofcancerswas1.88(Penninxetal.1998b).In
contrast to
thesefindings,otherresearchershavenotfoundevidenceforalinkbet
weendepres-
sion and malignant disease (Croyle 1998, Whooley & Browner
1998). However,
most prior literature has relied on a single assessment of
depressive symptoms;
whenPenninxetal.
(1998b)usedasimilarstrategywiththeirowndata, theydidnot
find the relationship between dysphoria and cancer that emerged

when depressive
symptomsexceededcutpoints atbaselineaswell
as3and6yearsbeforebaseline.
Thus, some of the inconsistencies among cancer studies may
reflect methodolog-
ical differences. Additionally, it should also be noted that many
related cancer
studies have assessed a wide range of malignancies with very
different etiologies,
geneticcontributions,behavioral influences(e.g.,smoking),etc.;
theheterogeneity
makes it difficult to assess evidence in this arena. Our
mechanistic discussion in
the next section suggests that some cancers may show stronger
relationships with
negative emotions than others (Ershler & Keller 2000).
LaTeX2e(2001/05/10) P1: GJC
Depression influencesoutcomes inavarietyofother
illnesses.Depressedmood
was an independent risk factor for all-cause mortality in
medical inpatients
(Herrmann et al. 1998). Among 1286 persons who were 71 or
older, baseline de-
pressive symptoms predicted greater physical decline over the
subsequent 4 years
(Penninx et al. 1998a). Depression heightens the risk for
osteoporosis; either past
or current depression in women was associated with lower bone
mineral density

(Michelson et al. 1996). Among older men, depressed mood at
baseline was asso-
ciated with an increased risk for declines in muscle strength
over a 3-year period;
important as an indication of current physical functioning, grip
strength is also a
powerful predictor of future functional limitations and
disability (Rantanen et al.
2000). Depression has also been associated with reduced
rehabilitation effective-
nessinaspectrumofdiseases(e.g.,stroke,fractures,andpulmonarydi
sease)(Katz
1996). Similarly, depressed diabetics are less likely to follow
recommendations
for dietary management and glycemic control (Katon 1998).
Pain, a pervasive medical problem, accounts for substantial
levels of disability
andcontributesgreatly to theoverall burdenof illness
(Turk&Melzack1992). In-
extricablylinkedtodepressionandothernegativemoods,paincanincr
easedisease
severity and mortality (Staats 1999, Wells et al. 1989). Pain can
provoke increases
in heart rate and blood pressure, enhance secretion of stress-
related hormones in-
cluding catecholamines and cortisol, and dysregulate a range of
immunological
activities (Kiecolt-Glaser et al. 1998, Liebeskind 1991).
Additionally, pain may
disrupt many aspects of physical, mental, and social functioning
(Leventhal et al.
1998). Accordingly, depression can amplify morbidity by
magnifying pain and
disability across a range of acute and chronic health problems.
How large are the effects? For mortality, the increased risk

among elderly
women in one large study was “. . . similar to that conferred by
other cardiovascu-
lar risk factors, such as hypertension, cigarette smoking,
hyperlipidemia, obesity,
anddiabetes”(Whooley&Browner1998,p.2132).
Inanotherstudy,depressionat
baselineincreasedtheriskthatparticipantswoulddevelopadisability
overthenext
6yearsby73%(Penninxetal. 1999).Data from11,242outpatients in
theMedical
Outcomes Study showed that patients with either a current
depressive disorder or
depressive symptoms in the absence of a syndromal disorder
had worse physical,
social, and role function, worse perceived current health, and
greater bodily pain
than patients with no chronic conditions (Wells et al. 1989).
The poorer func-
tioning that was uniquely associated with depressive symptoms
was comparable
to—orevenworse than—
thatuniquelyassociatedwitheightchronicmedical con-
ditions. Thus, the increased morbidity and mortality associated
with depression is
substantial.
Anxiety
Although depression has been the best-studied negative
emotion, anxiety also has
adverse effects, particularly in the cardiovascular realm, where
it plays a role in
the development of CHD and contributes to poorer prognosis
after acute coronary

events, including death and recurrent ischemic events. Phobic,
panic-like anxiety
LaTeX2e(2001/05/10) P1: GJC
predicted 3 times the risk of fatal CHD at a 7-year follow-up
compared with no
anxiety (Hainesetal.1987). Indata
fromtheNormativeAgingStudyhigher levels
ofanxietywereassociatedwithalmostdoubletheriskoffatalCHD(Ka
wachietal.
1994b). Similarly, men in the Health Professionals Follow-up
Study who reported
the highest levels of anxiety had more than double the risk for
fatal CHD and
nonfatal myocardial infarction (Kawachi et al. 1994a). Anxiety
symptoms were
associatedwithsignificantly increasedriskofmyocardial
infarctionandcoronary-
related death over a 20-year period in women who were
homemakers (Eaker et al.
1992).Anxietyalsohasnegativeconsequencesforrecoveryfromsurg
ery(Kiecolt-
Glaser et al. 1998).
Hostility/Anger
Chronic anger and hostility also negatively impact health. One
excellent 9-year
population-based study found that men high in hostility had
more than twice the

risk of all-cause and cardiovascular mortality compared with
men low in hostility
(Eversonet al. 1997).Similarly, a largeprospective
studyofemployees found that
hostilitypredicted the totalnumberof long-
termmedicallycertifiedabsencesover
a4-yearperiodamongmenbutnotwomen(Vahteraetal.1997).
Indeed,arigorous
meta-analysis concluded that hostilitywasa robust risk factor
forCHD,aswell as
for all-cause mortality (Miller et al. 1996).
Although the weight of the evidence clearly implicates negative
emotions, par-
ticularly depression, in all-cause mortality, the findings have
been inconsistent,
the discrepancies undoubtedly fueled by notable methodological
shortcomings in
a number of studies, including small samples, low mortality,
brief follow-up pe-
riods, incomplete follow-up, and absence of control for relevant
health behaviors
or premorbid status (Schulz et al. 2000). Successive
assessments that provide in-
formation on health problems, medications, smoking, and
alcohol use are crucial;
indeed, the absence of positive findings in some studies may
well be related to
failure to assess and control for smoking and alcohol use
(Wulsin et al. 1999),
key health behaviors that impact a spectrum of diseases
(Kiecolt-Glaser & Glaser
1988). The effects are clearly bidirectional, and illness can
enhance the risk for
the development of depression and anxiety symptoms and
disorders (Katz 1996).

Despite thesemethodologicalshortcomings, it isclear that
theburdensandstresses
that stimulate psychological morbidity also have clear and
notable consequences
for physical health.
PATHWAYS
Morbidity, Mortality, and Aging: Central
Immunological Mechanisms
Emotions can affect health through many pathways; these
influences may occur
indirectly, throughhealthbehaviorsor compliancewithmedical
regimens, anddi-
rectly,throughalterationsinthefunctioningofthecentralnervoussys
tem,immune,
LaTeX2e(2001/05/10) P1: GJC
endocrine, and cardiovascular systems. The primary focus of
our mechanistic dis-
cussion will be the immune and endocrine pathways to age-
related changes in
health; our choice is based on recent evidence that implicates
dysregulation of
proinflammatory cytokines, particularly interleukin 6 (IL-6), as
a central compo-
nentacrossa rangeofdiseases inolderadults.Wefirstprovideabrief
introduction
to cytokines, followed by a review of evidence relating cytokine

dysregulation to
a spectrum of health problems.
Cytokines are protein substances released by cells that serve as
intercellular
signals to regulate the immune response to injury and infection.
The relevance of
cytokines to the biobehavioral sciences is illustrated by the
appearance of reviews
ofcytokinebiology inpsychiatricandpsychological literature
(Kronfol&Remick
2000, Maier & Watkins 1998). The signaling properties of
cytokines are similar
to classic hormones of the endocrine system, and cytokines can
be differentiated
into two basic classes based on their effects on the immune
response, proinflam-
matory and antiinflammatory. The proinflammatory cytokines
include IL-1, IL-6,
and tumornecrosis factor (TNF); theypromote
inflammation,abeneficial reaction
in early immune responses to infection and injury (Glaser et al.
1999a). The pri-
mary actions of these cytokines are attracting immune cells to
the site of infection
or injury and causing them to become activated to respond.
Secondary actions
include changes in physiology that promote inflammation, such
as alterations in
metabolism and temperature regulation. Antiinflammatory
cytokines such as IL-
10 and IL-13 dampen the immune response, causing, for
instance, decreased cell
function and synthesis of other cytokines.
The immune system’s inflammatory response can be triggered in
a variety

of ways, including infection and trauma. The mechanisms
associated with in-
flammation are critical to resolving infections and repairing
tissue damage; how-
ever,chronicorrecurringinfectionscanprovokepathologicalchange
s(Hamerman
1999).Forexample, lowlevelsofpersistent
inflammationmayresultwhenchronic
infectious processes such as periodontal disease, urinary tract
infections, chronic
pulmonary disease, and chronic renal disease persistently
stimulate the immune
system, with the greatest repercussions among older adults who
already show
age-related increases in IL-6 production (Cohen 2000).
Indeed, inflammation has recently been linked to a spectrum of
conditions
associated with aging, including cardiovascular disease,
osteoporosis, arthritis,
type 2 diabetes, certain lymphoproliferative diseases or cancers
(including mul-
tiple myeloma, non-Hodgkin’s lymphoma, and chronic
lymphocytic leukemia),
Alzheimer’sdisease, andperiodontaldisease
(Ershler&Keller2000).Theassoci-
ation between cardiovascular disease and IL-6 is related in part
to the central role
that this cytokine plays in promoting the production of C-
reactive protein (CRP),
recently recognized as an important risk factor for myocardial
infarction (Papan-
icolaou et al. 1998). For example, high concentrations of CRP
predicted the risk
of future cardiovascular disease in apparently healthy men

(Ridker et al. 1997).
Further studies provided mechanistic links: chronic infections
amplified the risk
for development of atherosclerosis fourfold in subjects who
were free of carotid
LaTeX2e(2001/05/10) P1: GJC
atherosclerosis at baseline, conferring increased risk even in
subjects lacking con-
ventional vascular risk factors (Kiechl et al. 2001). Indeed, the
increased risk for
artery-clogging plaque was greater than that conferred by
elevated blood pressure
or cholesterol (Kiechl et al. 2001). Cardiovascular disease is the
leading cause of
death, and individuals with high levels of both IL-6 and CRP
were 2.6 times more
likely to die over a 4.6-year period than those who were low on
both (Harris et al.
1999).
More globally, chronic inflammation has been suggested as one
key biological
mechanism that may fuel declines in physical function leading
to frailty, disabil-
ity, and, ultimately, death (Hamerman 1999, Taaffe et al. 2000).
For example,
elevated serum IL-6 levels predicted future disability in older
adults, a finding
the authors suggest may reflect the effects of the cytokine on

muscle atrophy,
and/or to the pathophysiologic role played by the cytokine in
particular diseases
(Ferrucci et al. 1999). Proinflammatory cytokines including IL-
6 may slow mus-
cle repair following injury and accelerate muscle wasting
(Cannon 1995); indeed,
IL-6 and CRP also play a pathogenic role in a range of diseases
associated with
disability among the elderly (e.g., osteoporosis, arthritis, and
congestive heart
failure) (Ferrucci et al. 1999). In this context it is interesting
that IL-6 is also asso-
ciated with self-rated health (Cohen et al. 1997a), a robust
predictor of mortality
(Leventhaletal.1998).Thus, theclinical importanceof
immunologicaldysregula-
tion forolderadults ishighlightedby
increasedrisksacrossdiverseconditionsand
diseases.
Emotions and Immune System Alterations
There is excellent evidence that depression and anxiety enhance
the production of
proinflammatory cytokines, including IL-6 (Dentino et al. 1999;
Lutgendorf et al.
1999; Maes et al. 1995, 1999, 1998). Higher plasma IL-6 levels
were associated
with greater distress in a sample of community women
(Lutgendorf et al. 1999).
Women who were caregiving for a relative with Alzheimer’s
disease had higher
levels of plasma IL-6 than either women who were anticipating
a housing reloca-
tion or community controls (Lutgendorf et al. 1999); the finding

was particularly
noteworthy because caregivers were 6–9 years younger, on
average, than women
in the other two groups. Chronic fatigue patients showed
increases in IL-6 fol-
lowing a severe life stressor (Hurricane Andrew) (Costello et al.
1998). Following
successfulpharmacologic treatment, elevated IL-6
levelsdeclined inpatientswith
a major depression diagnosis (Sluzewska et al. 1995).
Both physical and psychological stressors can provoke transient
increases in
proinflammatorycytokines(DeRijketal.1997,Zhouetal.1993);
inanimalmodels
both stress and administration of epinephrine elevated plasma
IL-6, consistent
with evidence that IL-6 production is stimulated through �-
adrenergic receptors,
among other pathways (Papanicolaou et al. 1998). Thus,
production of IL-6 and
other proinflammatory cytokines can be directly stimulated by
negative emotions
and stressful experiences, providing one direct pathway.
LaTeX2e(2001/05/10) P1: GJC
Negative emotions also contribute indirectly to the immune
dysregulation evi-
denced by proinflammatory cytokine overproduction. Repeated,
chronic, or slow-

resolving infections or wounds enhance secretion of
proinflammatory cytokines,
a process that can serve to further inhibit certain aspects of
immune responses
(e.g., IL-2, an important defense against infection), and thus
may contribute to
the immunodepression of aging (Catania et al. 1997). Stress
impedes the immune
response to infectious challenges, amplifying risks for
contagion and prolonged
illness episodes (Glaser et al. 1999b, Kiecolt-Glaser et al.
1996a, Sheridan et al.
1991); distress also provokes substantial delays in wound
healing (Glaser et al.
1999a, Kiecolt-Glaser et al. 1995, Marucha et al. 1998) and
enhances the risk for
wound infection after injury (Rojas et al. 2001). Thus, negative
emotions such as
depressionor anxietycandirectly affect thecellsof the
immunesystemandeither
up-ordown-regulate thesecretionofproinflammatorycytokines;
inaddition,neg-
ative emotions may also contribute to prolonged or chronic
infections or delayed
wound healing, processes that indirectly fuel proinflammatory
cytokine produc-
tion. These changes are likely to be greatest, and to carry the
highest health risks,
among the elderly.
Although our focus thus far has been on the health
consequences associated
with secretion of proinflammatory cytokines, negative emotions
can also have
direct adverse effects on a variety of other immunological
mechanisms; both an-

imal and human studies have provided convincing evidence that
these immune
alterations are consequential for health. For example, to help
demonstrate causal
relationships between psychosocial stressors and the
development of infectious
illness, investigators have inoculated subjects with a variety of
vaccines (Glaser
et al. 1992, 2000, Kiecolt-Glaser et al. 1996a, Morag et al.
1999, Vedhara et al.
1999). Vaccine responses demonstrate clinically relevant
alterations in immuno-
logical responses to challenge under well-controlled conditions;
accordingly, they
serve as a proxy for response to an infectious agent. More
distressed and anxious
individuals produced immune responses to vaccines that were
delayed, substan-
tially weaker, and/or shorter lived; as a consequence, it is
reasonable to assume
these same individuals would also be slower to develop immune
responses to
other pathogens; thus, they could be at greater risk for more
severe illness. Con-
sistent with this argument, adults who show poorer responses to
vaccines also
experience higher rates of clinical illness, as well as longer-
lasting infectious
episodes (Burns & Goodwin 1990, Patriarca 1994). In addition,
other researchers
have shown that distress can alter susceptibility to cold viruses
(Cohen et al.
1998).
Increasedsusceptibilitytopathogensisaserioushealthproblemforol
deradults.

For example, although influenza is rarely fatal among healthy
younger adults,
together influenza and pneumonia, a common complication of
influenza virus
infection, constitute the fourth leading cause of death among
individuals who are
75orolder(Yoshikawa1983);distressedolderadultsdemonstratepo
orerresponses
to both influenza and pneumococcal vaccines (Glaser et al.
2000, Kiecolt-Glaser
etal.1996a,Vedharaetal.1999).Thus,datafromhumanstudiesnowpr
ovidesolid
LaTeX2e(2001/05/10) P1: GJC
evidence that negative emotions can increase susceptibility to
infectious disease
via alterations in the immune response.
Emotions and Neuroendocrine Alterations
The endocrine system serves as one prominent gateway across a
spectrum of dis-
easesbecauseemotionsprovoke the
releaseofpituitaryandadrenalhormones that
havemultipleeffects, includingalterations incardiovascularand
immunefunction
(Glaser&Kiecolt-
Glaser1994,Rozanskietal.1999).Bothanxiousanddepressed
moodscanactivate thesympathetic-pituitary-
adrenalmedullaryaxis,aswellas the

hypothalamic-pituitary-adrenocortical (HPA) axis (Miller
1998). Numerous stud-
ies have suggested that a variety of emotion-responsive
hormones including the
catecholamines (norepinephrine and epinephrine),
adrenocorticotropin hormone,
cortisol, growth hormone, and prolactin can impel quantitative
and qualitative
changes in immune function, and there is bi-directional
feedback between the
endocrine and immune systems (Rabin 1999). For example,
depression can sub-
stantially boost cortisol, and elevations in cortisol can provoke
multiple adverse
immunological changes including defects in vaccine responses
(Vedhara et al.
1999) and wound healing (Padgett et al. 1998). In contrast to
the generally neg-
ative effects of cortisol, growth hormone can enhance many
aspects of immune
function (Malarkey et al. 1996); growth hormone is lower in
depressed patients
(Dinan 1998), and growth hormone gene expression is altered in
mononuclear
cellsofchronicallydistressedcaregivers(Malarkeyetal.1996).Addi
tionally,both
anxious and depressive disorders and symptoms can elevate
catecholamines;
althoughbrief increases in response toacute
stressorsmaybeadvantageousunder
manycircumstances, longer-
termincreasesaregenerallyassociatedwith immuno-
logical down-regulation (Malarkey et al. 1996).
The hypercortisolemia associated with clinical depression is
well documented

(DeRijk et al. 1997); however, the endocrine system’s
involvement in the patho-
genesis of many stress-related disease processes is also likely
mediated in part
through frequent small daily excursions in hormonal levels
following stressful
events,and/or
throughdisturbanceofdiurnalrhythms(Dhabhar&McEwen1997).
The ability to “unwind” after stressful encounters, i.e., quicker
return to one’s
neuroendocrine baseline, influences the total burden that
stressors place on an in-
dividual (Frankenhaeuser 1986). Stressors that are resistant to
behavioral coping,
particularlystressorsperceivedasunpredictableanduncontrollable,
maycontinue
tobeassociatedwithelevatedstresshormonesevenafter
repeatedexposure(Baum
et al. 1993).
Our prior discussion focused on age-related immune
dysregulation; thus, it is
important to note that cytokines such as IL-6 also influence the
functioning of the
endocrine system, one of the many bi-directional relationships
between the two
systems. IL-6 is a potent stimulator of corticotropin-releasing
hormone produc-
tion, a mechanism that leads to heightened HPA activity,
including elevated lev-
els of plasma adrenocorticotropin hormone, followed by
increased cortisol levels

LaTeX2e(2001/05/10) P1: GJC
(Dentinoetal.1999).Thus,negativeemotions thatdysregulate IL-
6secretionmay
also promote neuroendocrine alterations that have immune
consequences.
Thecomplexityof thesepotential
interactionsisfurtherunderscoredbyoneline
of research that suggests that once cortisol levels rise, they can
initiate, perpetu-
ate, or aggravate syndromaldepression, depression-
likebehaviors, anddepressive
symptomssuchasanxiety,
insomnia,andpoormemory(Wolkowitz&Reus1999).
Such data are consistent with the conceptualization of major
depression as a dys-
function in the stress response (Sternberg et al. 1992), as well
as evidence that
both emotional distress and disease may be prompted by
common genetic and
constitutionalvariables (Leventhal et al. 1998).For
example,first-degree relatives
of depressed patients who have never been clinically depressed
have HPA axis
responses similar to their affected relatives and different from
controls (Holsboer
et al. 1995). Similarly, a 10-year follow-up of adolescents who
had served as part
of anormal control groupshowed that thebaselinepatternof sleep-
relatedgrowth
hormonesecretionwaspredictiveofsubsequentdepressiveepisodes
(Coplanetal.
2000).Accordingly,thehealthhazardsassociatedwithnegativeemot

ionsarelikely
to reflect multiple interacting risk factors, including important
genetic influences.
Although there are common genetic influences for depression
and neuroen-
docrine dysregulation, sufficiently stressful circumstances can
also produce clini-
callysignificant immuneandendocrinedysregulation in
individualswhoarenotat
risk. For example, the chronic strains of dementia spousal
caregiving were related
to the onset of syndromal depressive disorders in older adults
who had no prior
evidence of vulnerability through either personal or family
history (Dura et al.
1990). Moreover, although only a minority of caregivers
develop syndromal dis-
orders, men and women who provide long-term care for a
spouse or parent with
Alzheimer’s disease typically report high levels of distress as
they attempt to cope
with the familymember’sproblematicbehaviors; this
stressorhasbeenassociated
with prolonged endocrine and immune dysregulation, as well as
health changes,
including alterations in vaccine response and wound healing
(Castle et al. 1995;
Esterling et al. 1994, 1996; Glaser et al. 1998; Irwin et al. 1991;
Kiecolt-Glaser
et al. 1996a; Malarkey et al. 1996; Mills et al. 1999; Vedhara et
al. 1999; Wu et al.
1999).
Health Behaviors

In addition to the direct influences of psychological states on
physiological func-
tion, distressed individuals are more likely to have health habits
that put them
at greater risk, including poorer sleep, a greater propensity for
alcohol and drug
abuse, poorer nutrition, and less exercise, and these health
behaviors have cardio-
vascular, immunological, and endocrinological consequences
(Kiecolt-Glaser &
Glaser 1988). Psychosocial stressors that increase adverse
health behaviors also
provokemaladaptivephysiological
changes.Forexample,deepsleepprovides the
normal stimulus formuchof the
releaseofgrowthhormone,whichenhancesmul-
tiple aspects of immune function; thus, stressors that modify the
architecture of
LaTeX2e(2001/05/10) P1: GJC
sleep also lessen secretion of growth hormone (Veldhuis &
Iranmanesch 1996).
Moreover, even partial sleep loss one night results in elevated
cortisol levels the
nextevening(Leproultetal.1997).Adversehealthbehaviorscan
interactwithone
another; for example, heavy alcohol use is linked to poorer
sleep and nutrition.
Smoking makes substantial contributions to morbidity and
mortality; depressed

patients are more likely to smoke and less likely to quit than
nondepressed indi-
viduals (Wulsinet al. 1999).Depressedpatientsmaybe less likely
toseekmedical
care and take prescribed medications than those who are not
depressed (Penninx
et al. 1999, Whooley & Browner 1998).
Higher plasma IL-6 and CRP levels are associated with adverse
health habits:
Values for both are higher in smokers than nonsmokers, in
individuals who report
less physical activity, and in those with a higher body mass
index (Ferrucci et al.
1999, Taaffe et al. 2000). However, health habits including
smoking, physical
activity, and alcohol use have typically explained only a small
part of the excess
mortalityassociatedwithdepressionamongolderadults,e.g.,Pennin
xetal.(1999).
Similarly, IL-6 has robust relationships with morbidity and
mortality, even after
controlling for health behaviors (Ferrucci et al. 1999, Taaffe et
al. 2000); more
broadly, behavioral studies have demonstrated reliable
psychological influences
on immune function in populations selected in part on the basis
of health habits
(Kiecolt-
Glaseretal.1993).Thus,healthbehaviors,althoughobviouslyimport
ant,
are not sufficient to explain the relationship between emotions
and disease.
We have focused on the immune and endocrine systems, but
there are obvi-

ously many other physiological pathways through which
emotions can influence
health, including cardiovascular and neurobiological circuitry
(Davidson et al.
2001, Krantz & McCeney 2001, Leventhal et al. 1998).
However, many lines of
evidence now indicate that IL-6 may function as a “. . . global
marker of impend-
ing deterioration in health status in older adults” (Ferrucci et al.
1999, p. 645).
We have argued that negative emotions directly prompt immune
dysregulation,
and these processes may lead to subsequent maladaptive
immune and endocrine
changes. Thus, research that addresses the dysregulation of the
immune and en-
docrine systems associated with negative emotions could
substantially enhance
our understanding of psychological influences on health,
particularly among the
elderly.
Allostatic Load: Conceptual Similarities and Differences
Theimmunedysregulationwearediscussingisconsistentwith
thebroadallostatic
load formulation, the “. . . long-term effect of the physiologic
response to stress”
(McEwen 1998, p. 171); however, the breadth of disease
outcomes addressed and
theoperationalizationof
theconceptsandpathwaysaresomewhatdifferent. Inves-
tigators have used a broad battery of measures to gauge
allostatic load, including
blood pressure, overnight urinary cortisol and catecholamine
excretion, waist to

hip ratio, glycosylated hemoglobin, the ratio of serum high-
density lipoprotein in
the total serum cholesterol concentration, and
dehydroepiandrosterone (DHEA)
LaTeX2e(2001/05/10) P1: GJC
sulfate; individuals with higher scores on this broad battery
were more likely to
have incident cardiovascular disease as well as declines in
cognitive and physical
function when assessed at a 3-year follow-up (Seeman et al.
1997).
Inconcertwiththeunderlyingtenetsoftheallostaticloadformulation(
McEwen
1998),emotional influenceson theHPAandsympathetic-pituitary-
adrenalmedul-
lary axes—and potential long-term changes in each—are a
central focus of our
mechanistic discussion. Within both frameworks chronic stress
is highlighted,
with its capacity for inducing long-term decline via
overexposure to stress hor-
mones.However, ourmechanisticpath focusesmorenarrowlyon
the implications
ofadverseneuroendocrinechanges for
immunemodulation,aswellas thebidirec-
tional feedbackfromthe immunesystemtotheendocrinesystem—
thestimulation
of corticotropin-releasing hormone by IL-6—on the spectrum of

inflammation-
relatedhealthoutcomesdiscussedearlier.Clearly,
thebatteryofhealth indicesde-
scribedabove(Seemanetal.1997)haveimportantprognosticvalue;h
owever,even
after the point at which risk factors such as cholesterol,
hypertension, and obesity
predict health deterioration less successfully among the very
old, chronic inflam-
mationcontinuestobeanimportantmarker(Ferruccietal.1999).Final
ly,weplace
agreateremphasisonthetoll
thatdailystressplaysviaimmunedysregulation—the
extent to which negative emotions contribute to prolonged
infection and delayed
woundhealing,processesthatfuelsustainedproinflammatorycytoki
neproduction.
Thus, in thefinal sectionweaddress theenormousvariability in
stress responsive-
nessbyreviewingliteraturerelatedtoresilienceandvulnerabilityfact
ors identified
in psychoneuroimmunology research to date.
VULNERABILITY AND RESILIENCE FACTORS
Sociodemographic Variables
AGE Biologically, the largest deleterious or enhancing
consequences of negative
and positive emotions are likely to occur when biological
vulnerability is great-
est: early and late in life. Although our primary focus has been
on aging, intense
emotional experiences have the capacity to permanently alter
neuroendocrine and
autonomic responses, and thesemaybemost consequentialwhen

theyoccurearly
in life. For example, women with a history of childhood abuse
are at substantially
greater risk for depressive and anxiety disorders; they also show
larger pituitary-
adrenal andautonomic responses to laboratory stressors
thancontrols (Heimet al.
2000, Lemieux & Coe 1995) and possibly experience long-term
immunological
alterations (De Bellis et al. 1996). Data on maternal separation
in nonhuman pri-
matesprovidesstrongsupportiveevidencefromawell-
characterizedanimalmodel
(Coe 1993).
Changes in immune function associated with aging have already
been ad-
dressed. In addition, however, older adults appear to show
greater immunological
impairmentsassociatedwithdistressordepression
thanyoungeradults (Herbert&
LaTeX2e(2001/05/10) P1: GJC
Cohen1993,Kiecolt-
Glaseretal.1996a,Schleiferetal.1989).Further,olderadults
may be more vulnerable to negative emotions due to smaller
social support net-
works(Carstensen1992). Incontrast,however,
theintensityofemotionalreactions
may also decline with aging, providing some protection

(Leventhal et al. 1998).
Finally, the impactofagemayvarythroughitsassociationwithother
individual
differences, related to changes in social, psychological, and
biological resources
(Leventhal et al. 1998). For example, age of onset of depression
interacted with
gender such that onset after age 70 in women most strongly
predicted increased
morbidity and mortality among adults seeking treatment for
depression (Philibert
et al. 1997). Thus, aging can interact with distress and
depression to enhance risks
for morbidity and mortality among older adults.
GENDER There are established gender differences in well-
being, including dif-
ferences in major psychopathology (e.g., depression is more
common in women)
andnegativeandpositivemoods, thatmayderive
frombiological,personality, and
sociocultural influences (Nolen-Hoeksema & Rusting 1999).
Surprisingly, there
has been inconsistent attention paid to possible gender
differences in emotion and
health relationships, making it difficult to draw conclusions at
this time. Some
studieshaveusedonlymalesor females, andothershavenot
systematicallyexam-
inedgendereffects.Forexample,ameta-
analysisofhostilityandhealthconcluded
that too few studies have reported results by sex to draw
conclusions at this time
(Miller et al. 1996).
Estrogen and androgens can repress IL-6 expression, and thus

age-related in-
creases inIL-6geneexpressionandserumlevelsare thought
toberelated inpart to
theagingof theendocrinesystem(Ershler&Keller2000).These
linkagessuggest
that longitudinalcomparisonsofpostmenopausalwomenwhoare
takinghormone
replacement therapy with those who are not would be one
potentially profitable
avenue for exploration.
More broadly, gender effects have been demonstrated both in
emotional
experiences (e.g., cognitive, physiological responses), and in
health outcomes
(Frankenhaeuser 1991, Stoney et al. 1987, Verbrugge 1982).
Differential rates of
depression, anxiety, and hostility in men and women may lead
to different overall
associations between gender and health outcomes. Furthermore,
men and women
may experience similar emotions differently, perhaps in part
due to different con-
stellations of additional vulnerability and resilience factors
(e.g., age, social sup-
port), resulting in different associations with health outcomes
(Kiecolt-Glaser &
Newton2001,Tayloretal.2000b).Forexample,aninteractionbetwee
ndepression
severity,age,andgenderwasrecentlyfoundsuchthatamongtheelderl
y,severede-
pressionwasassociatedwith increasedmortality
inmenandwomen,whereasmild
depression predicted increased mortality solely in men
(Schoevers et al. 2000). In

another example, among community dwelling adults, chronic
strain, low sense of
mastery, and rumination were more common in women than in
men and mediated
the greater prevalence of depression in women (Nolen-
Hoeksema et al. 1999).
It is possible that differences in the qualitative experience of
emotions may be
LaTeX2e(2001/05/10) P1: GJC
associated with different health behavior and physiological
reactivity patterns,
leading to different health outcomes. An important focus for
future research is
the manner in which gender may act as a vulnerability or
resilience factor in its
interaction with emotion and health outcomes, and the
contribution of additional
contextual variables such as age and social support in these
relationships.
SOCIOECONOMIC STATUS Socioeconomicstatus(SES),
typicallymeasuredbyed-
ucation, income, and occupation, has inverse relationships with
major depression,
depressive symptoms, and hostility (Adler et al. 1994). SES also
shows strong
inverse relationships with most major causes of morbidity and
mortality across
populations(Tayloretal.1997).Therelationshipsaresostrongthatalt

houghlower
SES groups have higher rates of morbidity and mortality,
differences in social po-
sition relate to risk even at the upper levels of the hierarchy
(Adler et al. 1994).
The longer-termstressorsassociatedwith immunealterations
include“burnout”at
work (Lerman et al. 1999), job strain (Kawakami et al. 1997),
and unemployment
(Arnetzetal.1991).Taylorandcolleagues(1997)suggest
thatsocialclassandrace
provideacontextforunderstandingtheimpactofunhealthyenvironm
ents,withone
initial route to increased risk via obvious differential exposures
to chronic stress.
RACE Racial and ethnic disparities in morbidity and mortality
exist in a number
of health-related conditions, including cancer, cardiovascular
disease, diabetes,
HIV/AIDS, and preventable infectious illness (Williams 1997),
all of which in-
volve the immune system. These differences are due in part to
dispositional risk
factors, health behavioral risk factors (Myers et al. 1995), and
SES, which are not
exclusivetoparticularethnicgroups.Forexample,ahigherprevalenc
eofAIDSin-
dicator conditions (e.g., tuberculosis, pneumonia) has been
found in HIV-positive
racial and ethnic minorities compared with HIV-positive whites,
and is probably
influenced by differential exposure to etiologic agents,
diagnosis and reporting,
and access to treatment (Hu et al. 1995). Racial and ethnic
differences in health-

relatedoutcomesmaybeassociatedwithmentalhealthdisparities,
suchas ratesof
depression, thatmaybedrivenbySESandethnicdifferences in
seeking treatment
(US Dep. of Health and Human Services 1999). At the same
time, there appear
to be direct relationships between ethnicity and health, such as
poorer health out-
comes among African Americans across the socioeconomic
strata (Williams &
Collins 1995).
The immunological and genetics literatures generally suggest a
genetic con-
tribution to disease risk stratified by race/ethnicity, particularly
for autoimmune
disorders (Hess & Farhey 1994, Kalman & Lublin 1999), and
these differences
may be due to genetic factors such as cytokine polymorphisms.
Despite these ge-
neticstratifications, theconceptof“race”isnota
truebiologicalcharacteristic,and
the construct of “ethnicity” is atheoretical; both can lead to
simplistic interpreta-
tionsofintergroupdifferences(Meyerowitzetal.1998,Williams199
7).Moreover,
given that genetic factors generally determine susceptibility, but
not development
of disease, racial and ethnic influences on emotions, immunity,
and health may be
LaTeX2e(2001/05/10) P1: GJC

best understood along multiple dimensions, including culture,
ethnic identity, and
minority status (Phinney 1996).
Personality Traits and Coping
Personality and coping styles reflect individual differences in
appraisal and re-
sponse to stressful situations, and both have been associated
with the onset and
course of chronic and progressive health problems (Scheier &
Bridges 1995).
In fact, in longitudinal studies, personality and coping
characteristics have pre-
dictedphysical illness andmortality in initiallyhealthyadults
(Maruta et al. 2000,
Peterson et al. 1988), as well as in HIV-seropositive gay men
(Cole et al. 1997,
1996;Reedetal.1999)andadultsundergoingbonemarrowtransplant(
Molassiotis
et al. 1997). There is evidence that personality, coping, and
emotions may interact
to increaseordecrease individuals’
riskofnegativehealthoutcomes.Forexample,
the co-existence of the type “D” distressed personality style,
including depressive
and anxiety symptoms, and social inhibition, predicted cardiac
morbidity and
mortality over a 10-year period (Denollet & Brutsaert 1998,
Denollet et al. 1996),
whereas greater optimism, indicative of a positive emotion
personality style, pre-
dicted better health outcomes among cardiac patients (Scheier et
al. 1999).

A potent resilience factor for health outcomes may be the
induction and main-
tenance of positive emotion through personality and coping
styles. The broaden-
and-build model of positive emotions (Fredrickson 1998) posits
a broadening of
the individual’s scopeofattention, cognition, andaction,
andbuildingofphysical,
intellectual, and social resources. Positive emotion may include,
but is not limited
to,positive reappraisalof stressful lifeevents
(Folkman&Moskowitz2000),find-
ingmeaning(Taylor1983),developingpositive illusions
(Tayloretal.2000a), and
situational or dispositional optimism (Scheier & Carver 1992,
Taylor 1989). Fur-
thermore, positive emotions might “undo” the aftereffects of
negative emotions,
particularly in physiological recovery (Fredrickson 1998).
Positive emotion has
beenassociatedwithbetterhealthoutcomes,
forexampleamongmaleheart attack
survivors (Affleck et al. 1987) and HIV-seropositive men
experiencing bereave-
ment (Bower et al. 1998). The pathways through which positive
emotions impact
health outcomes are not well known at this point, but likely
occur through en-
docrine and immune mechanisms, as well as indirectly through
health behaviors
(Aspinwall & Brunhart 1996, Shepperd et al. 1996).
Personalityandcopingstylesmaypredisposeindividualstowardgrea
terrelative
negative or positive emotions, thereby maintaining

physiological alterations asso-
ciated with emotions. For example, personality and coping
styles, such as repres-
sion, rejection sensitivity, attributional style, and sociability,
have been associated
with altered immune cell counts in peripheral blood and
dysregulated cellular im-
mune function
(Segerstrom2000).Notably,givenourearlierdiscussion,oneposi-
tivecopingstrategy,attendanceatreligiousservices,hasbeenassocia
tedwithlower
levels of IL-6 in a large community sample of older adults
(Koenig et al. 1997).
Thus, the relationships among personality and coping styles and
health outcomes
LaTeX2e(2001/05/10) P1: GJC
may be mediated by their influences on negative and positive
emotions and im-
mune function, and these relationships are likely to be strongest
in the context of
relevant stressful events.
Social Relationships
Data from large, well-controlled epidemiological studies
suggest that social iso-
lation constitutes a major risk factor for morbidity and
mortality, with statistical
effect sizes comparable to those of such well-established health

risk factors as
smoking, blood pressure, blood lipids, obesity, and physical
activity (House et al.
1988).
Immunologicalalterationsprovideonepossiblephysiologicalpathw
ay: the
linkbetweenpersonal relationshipsand immunefunction isoneof
themost robust
findings in psychoneuroimmunology (Uchino et al. 1996). For
example, better re-
sponses on two immunological assays were associated with
higher social support
in women whose husbands were being treated for urologic
cancer (Baron et al.
1990). Medical students who reported better social support
mounted a stronger
immune response to a hepatitis B vaccine than those with less
support (Glaser
et al. 1992). Individuals with fewer social ties were more
susceptible to respira-
tory viruses (Cohen et al. 1997b). Spousal caregivers of
dementia sufferers who
reported lower levels of social support on entry into a
longitudinal study and who
were most distressed by dementia-related behaviors showed the
greatest and most
uniformlynegativechangesinimmunefunctiononeyearlater(Kiecol
t-Glaseretal.
1991). Several researchers reported immunological differences
between subjects
who disclosed traumatic or upsetting events, compared with
those in a nondisclo-
sure condition (Christensen et al. 1996, Esterling et al. 1990,
Pennebaker et al.
1988, Petrie et al. 1995). Loss of a spouse or partner through
bereavement or di-

vorce is associated with poorer immune function for a period of
time (Irwin et al.
1987;Kemenyet al. 1995;Kiecolt-Glaser et al.
1987,1988;Schleifer et al. 1983).
Marriage is obviously an important relationship, and marital
quality has been
associatedwith immuneandendocrine function (Kiecolt-
Glaser&Newton2001).
Forexample,womenwithrheumatoidarthritiswerefollowedfor12w
eeks(Zautra
et al. 1998); although both immune function and clinician’s
ratings changed dur-
ing a week of increased interpersonal stress, women who
reported more positive
spousal interaction patterns and less spousal criticism or
negativity did not show
as large an increase in clinical symptoms.
However, when close relationships are discordant, they can also
be associated
with depression and immune dysregulation. Both syndromal
depression and de-
pressive symptoms were strongly associated with marital
discord (Beach et al.
1998, Fincham & Beach 1999). In addition, pervasive
differences in endocrine
and immune function were reliably associated with hostile
behaviors during mar-
ital conflict among diverse samples that included newlyweds
selected on the basis
of stringent mental and physical health criteria, as well as
couples married an
average of 42 years (Kiecolt-Glaser et al. 1997, 1993, 1996b;
Malarkey et al.
1994). Thus, although supportive personal relationships are

associated with better
LaTeX2e(2001/05/10) P1: GJC
immune function (Kiecolt-Glaser & Newton 2001, Uchino et al.
1996), close
personal relationships that are chronically abrasive or stressful
may provoke de-
pression and other negative emotions as well as persistent
immune and endocrine
dysregulation.
CONCLUSIONS
Wesuggest thatresearchers interestedinpsychological
influencesonhealthshould
expand their consideration of the range of diseases whose onset
and course may
be influenced by the immune system; inflammation has recently
been linked to a
spectrum of conditions associated with aging, including
cardiovascular disease,
osteoporosis,arthritis,
type2diabetes,certainlymphoproliferativediseasesorcan-
cers, Alzheimer’s disease, frailty and functional decline, and
periodontal disease
(Ershler & Keller 2000). Production of IL-6 and other
proinflammatory cytokines
that influence these and other conditions can be directly
stimulated by negative
emotions and stressful experiences, providing one direct

pathway from emotions
to health. In addition, negative emotions may also contribute to
prolonged in-
fection or delayed wound healing, processes that fuel sustained
proinflammatory
cytokine production. Accordingly, we argue that distress-related
immune dys-
regulation may be one core mechanism behind the health risks
associated with
negative emotions. These direct and indirect processes pose the
greatest health
risks for older adults who already show age-related increases in
proinflamma-
torycytokineproduction.Thus, aging
interactswithnegativeemotions toenhance
risks formorbidityandmortality amongolder adults.Finally,
thepsychoneuroim-
munology literature provides evidence that resources such as
close personal re-
lationships or personality and coping styles that diminish
negative emotions may
enhance health in part through their positive impact on immune
and endocrine
regulation.
ACKNOWLEDGMENTS
Workon
thisarticlewassupportedbyNIHgrantsK02MH01467,R37MH4209
6,
K02 MH01467, PO1 AG16321, P50 DE17811, and T32
MH18831.
Visit the Annual Reviews home page at
www.AnnualReviews.org

LITERATURE CITED
AdlerN, BoyceT, ChesneyM, CohenS, Folk-
manS,etal.1994.Socioeconomicstatusand
health. Am. Psychol. 49:15–24
Affleck G, Tennen H, Croog S, Levine S.
1987. Causal attribution, perceived benefits,
and morbidity after a heart attack: an 8-
year study. J. Consult. Clin. Psychol. 55:29–
35
Arnetz BB, Brenner SO, Levi L, Hjelm R,
Petterson IL, et al. 1991. Neuroendocrine
Chronic stress promotes tumor growth and angiogenesis
in a mouse model of ovarian carcinoma
Premal H Thaker1,10, Liz Y Han1,10, Aparna A Kamat1,10,
Jesusa M Arevalo2, Rie Takahashi2, Chunhua Lu1,
Nicholas B Jennings1, Guillermo Armaiz-Pena1, James A
Bankson3, Murali Ravoori4, William M Merritt1,
Yvonne G Lin1, Lingegowda S Mangala1, Tae Jin Kim1, Robert
L Coleman1, Charles N Landen1, Yang Li1,
Edward Felix5, Angela M Sanguino6, Robert A Newman5, Mary
Lloyd7, David M Gershenson1,
Vikas Kundra4,8, Gabriel Lopez-Berestein6, Susan K
Lutgendorf 9, Steven W Cole2 & Anil K Sood1,7
Stress can alter immunological, neurochemical and
endocrinological functions, but its role in cancer progression

is not well understood. Here, we show that chronic behavioral
stress results in higher levels of tissue catecholamines, greater
tumor burden and more invasive growth of ovarian carcinoma
cells in an orthotopic mouse model. These effects are mediated
primarily through activation of the tumor cell cyclic AMP
(cAMP)–protein kinase A (PKA) signaling pathway by the b2
adrenergic receptor (encoded by ADRB2). Tumors in stressed
animals showed markedly increased vascularization and
enhanced expression of VEGF, MMP2 and MMP9, and we
found that angiogenic processes mediated the effects of stress
on tumor growth in vivo. These data identify b-adrenergic
activation of the cAMP–PKA signaling pathway as a major
mechanism by which behavioral stress can enhance tumor
angiogenesis in vivo and thereby promote malignant cell
growth. These data also suggest that blocking ADRB-mediated
angiogenesis could have therapeutic implications for the
management of ovarian cancer.
Epidemiologic and experimental animal studies have shown that
stress
may alter tumor growth1–5. However, the biological
mechanisms

underlying such effects are not well understood, and their
clinical
significance for human disease remains controversial as a result.
Data
from animal models suggest that stress can modulate the growth
of
certain tumors via neuroendocrine regulation of the immune
response
to tumor cells1. However, the uncertain role of the immune
system in
controlling solid tumors led us to consider an alternative
hypothesis:
stress mediators from the sympathetic nervous system might
directly
modulate the growth and malignant behavior of tumor cells,
inde-
pendent of effects on the immune system. Based on our prior
studies
linking behavioral factors to circulating VEGF levels in vivo
and
showing catecholamine regulation of tumor cell VEGF
production
in vitro6,7, we sought to determine whether sympathetic
nervous
system activity might mediate a causal effect of stress on the
growth
and metastasis of ovarian cancer in vivo. We also sought to
define the
role of angiogenic cytokines in mediating those effects.
These studies used an established orthotopic mouse model in
which
human ovarian carcinoma cells are inoculated into the
peritoneal
cavity of nude mice8. We then experimentally stressed animals

with a
new physical restraint system that we designed, in which
periodic
immobilization induces high levels of hypothalamic-pituitary-
adrenal
and sympathetic nervous system activity characteristic of
chronic
stress (Supplementary Fig. 1 online). We first examined the
effects
of stress on tumor growth and metastasis of HeyA8 ovarian
cancer
cells inoculated 7 d after the initiation of stress. In mice
receiving 0, 2
or 6 h of immobilization daily for 21 d, the number of tumor
nodules
increased by 259% in the 2-h stress group (P ¼ 0.005, two-
tailed
Student’s t-test) and 356% in the 6-h stress group (P o 0.001;
Fig. 1a). Mean tumor weight increased 242% in the 2-h stress
group (P ¼ 0.01) and 275% in the 6-h group (P ¼ 0.005). Tumor
growth was confined to the peritoneal cavity in all control mice,
but it
spread to the parenchyma of the liver or spleen in 50% of
stressed
mice (P ¼ 0.01). The number and weight of tumor nodules did
not
differ between the 2- and 6-h stress groups. Therefore, we used
the 2-h
restraint for all subsequent experiments. Additional experiments
with
the SKOV3ip1 ovarian cancer cell line and the MB-231
orthotopic
breast cancer model (Fig. 1b) showed that stress effects on
tumor
growth were evident in a wide range of tumor cell lines
(significant

differences in mean tumor weight and nodule number, all P o
0.01;
Fig. 1b). Subsequent studies using social isolation as an
alternative
stressor also showed a 187% increase in tumor weight (P o 0.01)
and
a 255% increase in nodule count (P o 0.001) compared with
group-
housed control animals (Fig. 1c). We also documented
longitudinal
Received 12 April; accepted 21 June; published online 23 July
2006; doi:10.1038/nm1447
1Department of Gynecologic Oncology, University of Texas
(U.T.) M.D. Anderson Cancer Center, 1155 Herman Pressler,
Unit 1362, Houston, Texas 77030, USA.
2Division of Hematology-Oncology, Department of Medicine,
11-934 Factor Building, University of California Los Angeles
(UCLA) School of Medicine, Los Angeles,
California 90095, USA. 3Department of Imaging Physics and
4Department of Experimental Diagnostic Imaging, U.T. M.D.
Anderson Cancer Center, 1515 Holcombe
Boulevard, Houston, Texas 77030, USA. 5Department of
Experimental Therapeutics, U.T. M.D. Anderson Cancer Center,
8000 El Rio Street, Houston, Texas 77054,
USA. 6Department of Experimental Therapeutics, 7Department
of Cancer Biology and 8Department of Diagnostic Radiology,
U.T. M.D. Anderson Cancer Center, 1515
Holcombe Boulevard, Houston, Texas 77030, USA.
9Department of Psychology, University of Iowa, E228 Seashore
Hall, Iowa City, Iowa 52241, USA. 10These authors
contributed equally to this work. Correspondence should be
addressed to A.K.S. ([email protected]).
NATURE MEDICINE VOLUME 12 [ NUMBER 8 [ AUGUST

2006 939
L E T T E R S
©
2
0
0
6
N
a
tu
re
P
u
b
lis
h
in
g
G
ro
u
p
h

tt
p
:/
/w
w
w
.n
a
tu
re
.c
o
m
/n
a
tu
re
m
e
d
ic
in
e

development of stress-induced differences in tumor mass using
a
subcutaneous ovarian cancer model (Supplementary Fig. 2
online).
To assess the effects of stress on sympatho-adrenal-medullary
activity, we measured the size of both adrenal glands9. In
animals
inoculated with HeyA8 cells, the left adrenal gland diameter
averaged
1.85 ± 0.25 mm for the control group versus 2.88 ± 0.56 mm for
the
stressed group (P ¼ 0.003, with similar differences noted for
right
adrenal; data not shown). Based on our previous in vitro studies
showing expression of b-adrenoreceptors on ovarian cancer
cells7, we
performed a series of experiments to evaluate their role in
mediating
stress effects on tumor growth. Beginning 4 d after tumor cell
injection, we treated daily mice bearing HeyA8 ovarian cancer
cells
with either (i) phosphate-buffered saline (PBS), (ii) 10 mg/kg
iso-
proterenol (a nonspecific b-receptor agonist (hereafter ‘b
agonist’)),
(iii) 5 mg/kg terbutaline (a b2 agonist), (iv) 1 mg/kg xamoterol
(a
b1 agonist), or (v) isoproterenol plus 2 mg/kg of the nonspecific
b antagonist propranolol. Isoproterenol increased the mean
number
of tumor nodules by 341% (P o 0.001) and the average tumor
weight
by 266% (P ¼ 0.001; Fig. 1d). We noted a similar increase in
tumor

burden with the b2 agonist terbutaline but not the b1 agonist
xamoterol. Propranolol blocked the isoproterenol-induced
increase
in tumor growth (Fig. 1d), showing that b-adrenergic signaling
is
essential for that effect. Propranolol also completely blocked
the effects
of immobilization stress on tumor growth, indicating a critical
role for
b-adrenergic signaling in behavioral stress effects on tumor
growth
(Fig. 1e). Among stressed animals, we observed invasive
disease such
as parenchymal metastasis to the liver, spleen, pancreas and
diaphragm
in 50% of the placebo-treated group, compared with 0% of the
propranolol-treated group (P ¼ 0.01).
To confirm the relevance of b-adrenergic receptors (encoded by
the
ADRB genes) for stress-induced tumor growth, we screened 17
additional epithelial ovarian cancer cell lines and identified two
(A2780 and RMG-II) that were negative for both ADRB1 and
ADRB2, as shown by RT-PCR, western blot and lack of
intracellular
cAMP response to norepinephrine or isoproterenol (data not
shown).
In contrast to other ovarian cancer cells, stress did not
significantly
affect tumor weight, number of nodules or the pattern of spread
in
mice injected with ADRB1- and ADRB2-negative A2780 or
RMG-II
cells (Fig. 1f). To further define the specific b-adrenergic
receptor

subtype responsible for stress effects and assess the relative
role of host
(mouse) versus tumor (human) adrenergic receptors in this
model, we
used short interfering RNA (siRNA) specific for human ADRB1
or
ADRB2 to selectively inhibit tumor cell adrenoreceptor
expression
(Fig. 1g). To facilitate in vivo distribution, we incorporated the
siRNA
into a neutral liposome, DOPC10. Both control and ADRB1
siRNA-
DOPC failed to block stress-induced enhancement of tumor
weight
and nodules (stress effects continued to be significant at P o
0.01),
but human ADRB2 siRNA-DOPC efficiently blocked stress-
mediated
enhancement of tumor cell growth (Fig. 1h). Together with data
from
b-adrenoreceptor–negative cell lines, these results identify
ADRB2
receptors on the human tumor cell as the critical mediator of
stress
effects in this in vivo system.
Based on our previous in vitro data showing that catecholamines
can promote VEGF production by ovarian cancer cells7, we
evaluated
the role of angiogenesis in stress effects on in vivo tumor
growth. In
the HeyA8 model, stress significantly enhanced mean vessel
density
(MVD) counts (P o 0.001), and that effect was blocked by the
b antagonist propranolol (Fig. 2a). We observed similar changes
in

MVD counts in tumors from non-stressed animals treated with
the
b agonists isoproterenol or terbutaline, which had significantly
higher
0 5 10
Number of nodules
2.521.510.50
Tumor weight (g)
**
**
*
*
Daily stress
Daily stress
Daily stress
Daily stress
Control
Control
Control
Control
Daily stress

Daily stress
Control
Control
Propranolol + daily stress
β1 + β2 siRNA
β2 siRNA
{
{
β1 siRNA
β2
s
iR
NA
β1
s
iR
NA
{
{
{

{
{
{
{
Control siRNA
Actin
ADRB1 ADRB2
96
h
48
h
96
h
48
h
Co
nt
ro
l s
iR
NA
Co
nt

ro
l s
iR
NA
128401.510.50
RMG-II
A2780
10501.510.50
* *
Propranolol
PBS + daily stress
PBS
10503210
Isoproterenol + propranolol
Xamoterol
Terbutaline
Isoproterenol
PBS
**
**
**
**

**
**
**
**
*
*
*
*
*
*
*
10503210
Social isolation
Control
Daily stress
Control
Daily stress
Control
Daily stress

Control
MB-231
SKOV3ip1
HeyA8
0 0.25 0.5
0 20 4010.50
10503210
10503210
Daily stress (6 h)
Daily stress (2 h)
Control
Number of nodulesTumor weight (g)a
b
c
d
e
f
g
h

Figure 1 Effect of chronic stress on in vivo ovarian cancer
growth.
(a) Quantification of tumor weights and tumor nodules in
control mice and
in mice stressed for 2 h or 6 h daily that were injected
intraperitoneally with
HeyA8 ovarian cancer cells. (b) Confirmation of the results in
mice injected
with two different b-adrenoreceptor–positive ovarian cancer cell
lines (HeyA8
and SKOV3ip1) intraperitoneally and MB-231 breast cancer
cells injected
into the mammary fat pad. (c) Effects of social isolation stress
on the HeyA8
ovarian cancer model. (d) The effects of immobilization are
mimicked by the
chemical stressor isoproterenol (a nonspecific b agonist) and
terbutaline
(b2 agonist) but not by xamoterol (b1 agonist). Mice treated
with both
isoproterenol plus propranolol (b blocker) have a tumor burden
similar to
non-stressed mice. (e) Propranolol counteracts the effects of
chronic stress,
confirming the importance of the b-adrenoreceptors. (f) Mice
injected with
b-adrenoreceptor–null ovarian cancer cell lines (A2780 and

RMG-II) did
not have accelerated tumor growth despite being chronically
stressed.
(g) Western blot of lysate from orthotopic HeyA8 tumors
collected 48 h and
96 h after a single dose of siRNA (either control siRNA or
siRNA targeted
to ADRB1 (b1) or ADRB2 (b2)) in DOPC liposomes. The
targeted receptors
were downregulated. (h) Effects of ADRB1- and/or ADRB2-
targeted siRNA
on stress-induced in vivo HeyA8 tumor growth. Results
represent the mean ±
s.e.m.; n ¼ 10 mice per group. *P r 0.01; **P r 0.001.
L E T T E R S
940 VOLUME 12 [ NUMBER 8 [ AUGUST 2006 NATURE
MEDICINE
©
2
0
0
6
N
a
tu

re
P
u
b
lis
h
in
g
G
ro
u
p
h
tt
p
:/
/w
w
w
.n
a
tu
re
.c

o
m
/n
a
tu
re
m
e
d
ic
in
e
MVD counts than did controls (P o 0.001); propranolol also
blocked
those effects (Fig. 2b). Stress-induced increases in angiogenesis
were
accompanied by significant elevation of VEGF protein (as
shown by
ELISA; P o 0.001) and mRNA (as shown by in situ
hybridization;
P o 0.001) within tumor tissue (Fig. 2a). We observed similar
increases in vessel density and VEGF levels in the ADRB-
positive
MB-231 breast cancer model (Supplementary Fig. 3 online) and
the
HeyA8 model with social isolation stress (Supplementary Fig. 3
online), but not in the ADRB-null RMG-II tumors

(Supplementary
Fig. 4 online). Immunohistochemistry in the HeyA8 restraint
model
also demonstrated stress-induced increases in levels of bFGF,
MMP-2
and MMP-9 proteins (Supplementary Fig. 5 online).
We carried out several additional studies to assess more
compre-
hensively the effects of stress on functional and anatomic
character-
istics of tumor vasculature. In an established Matrigel plug
assay
(Fig. 3a), exogenous VEGF (0.25 nM) enhanced angiogenesis as
expected (P o 0.01, Fig. 3b), and conditioned medium from
norepinephrine-treated tumor cells also significantly enhanced
angio-
genesis (P o 0.01) (Fig. 3a,b). Both catecholamines had modest
effects on in vitro migration and tube formation of mouse
mesenteric
endothelial cells11, but these effects were much smaller than
those
observed when endothelial cells were exposed to conditioned
medium
from norepinephrine-treated tumor cells (P o 0.001;
Supplementary
Fig. 6 online). Consistent with a key mechanistic role of ADRB
regulation of VEGF, the stimulatory effects of norepinephrine-
treated
tumor cell conditioned medium was efficiently blocked by
PTK787
(inhibits VEGFR2 and other tyrosine kinases) or by
pretreatment of
tumor cells with propranolol before norepinephrine treatment
for

collection of conditioned medium (Supplementary Fig. 6).
These
results paralleled confocal microscopy studies showing that
tumors
from stressed animals contained more tortuous and numerous
blood
vessels than controls (Fig. 3c). This was accompanied by a 24%
decrease (P o 0.01) in the proportion of blood vessels with
pericyte
coverage in tumors from stressed animals (Fig. 3d), suggesting
a more
immature vasculature. In addition, in vivo magnetic resonance
ima-
ging and kinetic analysis showed stress-induced increases in
tumor size
(Fig. 3e) and an increase in the volume transfer constant
between
plasma and the extravascular extracellular space (Ktrans, a
reflection of
flow and permeability; P o 0.05; Fig. 3f). Tumors from control
mice
showed a higher return rate constant between the extravascular
Isoproterenol +
propranolol
XamoterolTerbutalineIsoproterenolControl
V
E
G
F
I
S

H
V
E
G
F
C
D
3
1
V
E
G
F
I
S
H
Control Placebo +
daily stress
Propranolol Propranolol +
daily stress
V
E
G
F

C
D
3
1
*
*
*
* *
**
* *
0
0.5
1
O
D
m
e
a
su
re
m

e
n
t
0
10
20
30
40
50
Tu
m
o
r
V
E
G
F
(p
g
/m
g
)
0
10
20

30
40
M
V
D
0
0.5
1
O
D
m
e
a
su
re
m
e
n
t
0
10
20

30
40
Tu
m
o
r
V
E
G
F
(p
g
/m
g
)
0
10
20
30
40
M
V

D
a
b
Figure 2 Angiogenesis is increased in chronically stressed
animals. (a) HeyA8 tumor samples from control and stressed
animals treated with placebo or
propranolol were stained for VEGF and CD31 (for
quantification of mean vessel density (MVD) counts) by
immunohistochemistry. We quantified tumor
VEGF levels using an ELISA kit. We also stained tumor
samples for VEGF by in situ hybridization (ISH) and calculated
optical density (OD) measurements.
(b) HeyA8 tumor samples from mice treated with a placebo,
isoproterenol (nonspecific b agonist), terbutaline (b2 agonist),
xamoterol (b1 agonist), or
isoproterenol plus propranolol were stained for CD31 and VEGF
by immunohistochemistry and ISH. We quantified tumor VEGF
protein levels by ELISA. MVD
was higher in the isoproterenol and terbutaline groups than in
controls or in the combination therapy group. All photographs
were taken at 200�. The bars in
the graphs correspond sequentially to the labeled columns of
images at left. Error bars represent s.e.m. *P o 0.001.
L E T T E R S
2006 941
©

2
0
0
6
N
a
tu
re
P
u
b
lis
h
in
g
G
ro
u
p
h
tt
p
:/
/w

w
w
.n
a
tu
re
.c
o
m
/n
a
tu
re
m
e
d
ic
in
e
extracellular space and blood plasma (kep, the return rate of
contrast to
plasma; Fig. 3g). In comparison, the volume of the
extravascular

extracellular space per unit volume of tissue (ve) was greater in
the
stressed animals (Fig. 3h).
Given that behavioral stress induced significant anatomical and
functional alterations in tumor vasculature, we sought to
determine
whether VEGF activity had a key role in mediating stress-
induced
acceleration in tumor growth. Stress effects on the number and
weight
of HeyA8 tumor nodules were completely blocked by co-
administra-
tion of either the VEGF-R2 inhibitor PTK787 (50 mg/kg orally,
daily)
or the monoclonal VEGF-specific antibody bevacizumab (5
mg/kg
intraperitoneally twice per week) (Fig. 4a,b and Supplementary
Fig. 7 online).
To more clearly define the mechanism by which b-adrenergic
signaling regulates VEGF production, we examined
transcriptional
changes in activity of the VEGF gene after catecholamine
stimulation.
In SKOV3ip1 cells, VEGF mRNA expression increased by a
peak
842% 1.5 h after exposure to norepinephrine (Fig. 4c), and
promoter
activity increased by 1,240% (Fig. 4d). This effect was blocked
by
propranolol, was mimicked by isoproterenol and was not
observed in
ADRB-negative A2780 ovarian cancer cells, implicating
b-adrenoreceptors as a key mediator (Fig. 4c and data not

shown).
Subsequent studies showed that b-adrenergic stimulation
modulates
VEGF gene expression via the cAMP–protein kinase A (PKA)
signal-
ing pathway; the adenylyl cyclase activator forskolin (1 mM)
and the
PKA activator dibutyryl cAMP (db-cAMP; 1 mM) also
stimulated
3
0.6
210
kep (min
–1)
K trans (min–1)
0.40.20
ve
0.40.20
Control
Daily stress
Control
Daily stress
Control
Daily stress
Control

Daily stress
*
*
*
*
*
*
806040200
Percentage of vessels
with pericyte coverage
20100
SKOV3ip1 NE CM
VEGF
Hgb (mg/dl)
Control
SKOV3ip1 NE CMVEGFControl Daily stressControl
Daily stressControl Daily stressControl
a c
d e f

h
g
b
20 µm20 µm
Figure 3 Effects of chronic stress on angiogenesis. (a)
Representative images of Matrigel containing either serum-free
SKOV3ip1 conditioned medium (control), VEGF (0.25 nM), or
serum-free conditioned medium from norepinephrine-
treated SKOV3ip1 cells (SKOV3ip1 NE CM). Newly formed
blood vessels in the Matrigel are indicated by arrowheads.
(b) We excised the Matrigel plugs and used them for
quantification of angiogenesis by measuring the hemoglobin
(Hgb)
content in the Matrigel matrix. *P o 0.01 versus control. (c) We
visualized tumor vasculature from HeyA8 tumors in
control and stressed animals by confocal fluorescence
microscopy after perfusion with FITC-lectin. (d) We examined
vessel maturation by determining the extent of pericyte
coverage in tumors from control and chronic stress-exposed
animals using triple immunofluorescence (CD31 (red) for
endothelial cells, desmin (green) for pericytes and Hoechst
(blue) for nuclei). *P o 0.01. (e) Magnetic resonance imaging
(MRI) was used to evaluate the functional characteristics of the
tumor vasculature.
Representative in vivo axial images of intraperitoneal HeyA8
ovarian tumors (arrowheads) from control and stressed animals

were acquired in a 4.7-T scanner
using a T2-weighted fast spin echo sequence (FSE) and
respiratory gating (day 18 post-inoculation with HeyA8 cells).
Vascular parameters derived from
dynamic contrast–enhanced MRI included (f) Ktrans, the
volume transfer constant between blood plasma and
extravascular extracellular space, (g) kep, the
rate constant between extravascular extracellular space and
blood plasma and (h) ve, the volume of extravascular
extracellular (interstitial) space per unit
volume of tissue. We made measurements from the enhancing
periphery of each tumor (n ¼ 8 per group; *P o 0.05). Error bars
represent s.e.m.
**
**
**
**
**
**
**
151050
Luciferase activity
(multiple of control)
NE
Control

1050
Forskolin + KT5720
Forskolin
NE + propranolol
NE + KT5720
KT5720
Db-cAMP
VEGF mRNA (multiple of control)
Isoproterenol
NE
Control
1050210
Number of nodulesTumor weight (g)
Bevacizumab + daily stress
Bevacizumab
Control IgG + daily stress
Control IgG
1050
Number of nodules
1.510.50
Tumor weight (g)
PTK787 + daily stress

PTK787
PBS + daily stress
PBSa
b
c d
Figure 4 Effect of VEGF on stress-induced tumor growth. (a)
HeyA8-injected
mice were randomly assigned to one of the following four
groups (4 d after
tumor cell injection; n ¼ 10 per group): control, oral placebo
with stress,
50 mg/kg oral PTK787 (VEGF-R2 inhibitor) daily with no
stress, or PTK787
with stress. (b) HeyA8-injected mice were randomly assigned to
one of the
following four groups (4 d after tumor cell injection; n ¼ 10 per
group):
control, oral placebo with stress, bevacizumab (VEGF-specific
antibody;
5 mg/kg intraperitoneally, twice per week) with no stress, or
bevacizumab
with stress. Treatment with PTK787 or bevacizumab blocked
the stress-
induced increase in tumor weight and number of nodules
compared with

treatment with the oral placebo (PBS) plus stress. (c) VEGF
mRNA
increased significantly when SKOV3ip1 cells were stimulated
with
norepinephrine (1 mM), isoproterenol (1 mM), forskolin
(activator of cAMP)
(1 mM), or dibutyryl cAMP (db-cAMP) (1 mM). KT5720 (1
mM) is a selective
inhibitor of the cAMP-dependent protein kinase A and blocked
VEGF
induction by norepinephrine and forskolin. (d) In SKOV3ip1
cells, the VEGF
promoter activity was increased by 1,280% after norepinephrine
treatment
compared with vehicle control. *P r 0.01; **P r 0.001.
L E T T E R S
MEDICINE
©
2
0
0
6
N
a

tu
re
P
u
b
lis
h
in
g
G
ro
u
p
h
tt
p
:/
/w
w
w
.n
a
tu
re

.c
o
m
/n
a
tu
re
m
e
d
ic
in
e
VEGF mRNA (Fig. 4c), and the PKA inhibitor KT5720 could
efficiently block VEGF mRNA induction by norepinephrine and
forskolin (Fig. 4c).
Together, these results demonstrate a complete pathway by
which
behavioral stress can directly regulate the growth of ovarian
carcino-
mas in vivo. Stress-induced release of catecholamines can
activate
ADRB2 on ovarian carcinoma cells to trigger increased
expression of
the VEGF gene, which results in enhanced tumor

vascularization and
more aggressive growth and spread of malignant cells. In
addition to
providing the first experimental demonstration that behavioral
stres-
sors can enhance the pathogenesis of ovarian carcinoma in vivo,
these
data also suggest new approaches for adjuvant therapy in
metastatic
disease (for example, b-blockade).
The present study also expands our understanding of the general
pathways by which stress can regulate cancer pathogenesis.
Previous
research emphasized the role of the immune system in
mediating
stress effects on tumor growth and metastasis1,12,13, but the
present
results show that the neuroendocrine stress response can also
directly
affect the growth and activity of malignant tissue through
hormone
receptors expressed by tumor cells. Key to the demonstration of
this direct effect on tumor cell biology is the use of a mouse
host–human tumor hybrid model that provides opportunities for
species-specific molecular manipulations that focally affect the
tumor, such as the use of siRNA specific to human ADRB1 and
ADRB2 and the comparison of human tumors positive for
ADRB
expression with those lacking ADRB expression. In addition to
these tumor tissue-specific manipulations, we carried out
several
additional studies to ensure that generalized changes in host
physio-
logy (such as effects of the host cardiovascular or lymphatic
systems)

or direct effects on tumor cell proliferation are not responsible
for
the observed effects of stress on tumor growth (Supplementary
Fig. 8 online).
Pharmacologic and genetic manipulations identify b-adrenergic
signaling as a central mediator of stress effects in the model of
stress
used here, but these findings do not rule out the possibility that
other
neuroendocrine signaling pathways might modulate tumor cell
bio-
logy under other circumstances (for example, stress effects
involving
the hypothalamic-pituitary-adrenal/glucocorticoid axis14 or
inhibitory
effects of dopamine on VEGF activity15 and tumor growth16).
To the
extent that behavioral processes modulate the activity of
multiple
hormones via the central nervous system17–19 and those
hormones
influence tumor cell biology6,20, interventions targeting neuro-
endocrine function at the level of the central nervous system
could
represent new strategies for protecting individuals with cancer
from the detrimental effects of stress biology on the progression
of
malignant disease.
METHODS
Cell lines and culture conditions. We grew the HeyA8,
SKOV3ip1, A2780,
RMG-II (from N. Ueno and H. Itamochi, M.D. Anderson Cancer
Center) and

MB-231 (from J. Price, M.D. Anderson Cancer Center) cancer
cell lines as
previously described8.
Chronic stress model. We obtained 10- to 12-week-old female
athymic nude
mice from the US National Cancer Institute. All experiments
were approved by
the Institutional Animal Care and Use Committee of the M.D.
Anderson
Cancer Center. We used a restraint-stress procedure based on
previous
studies21, using a new physical restraint system that we
designed (Supplemen-
tary Fig. 1). For the social isolation model, we randomly
assigned mice to be
group housed (n ¼ 5 per cage) or individually housed (n ¼ 1 per
cage). There
was a wall and minimum distance of 24 inches between cages
for the isolated
mice. For both models, we injected the ovarian cancer cells
intraperitoneally
into mice 7 d after starting stress in all groups. We necropsied
animals 21 d after
tumor cell injection. For beta-blockade, we inserted the Alzet

osmotic mini-
pumps (DURECT) containing PBS or s-propranolol
hydrochloride (Sigma;
2 mg/kg/d)22,23 on the nape of neck 7 d before initiation of
restraint stress.
Short interfering RNA (siRNA) preparation. We purchased
siRNAs targeted
against ADRB1 (target sequence 5¢-
CCGATAGCAGGTGAACTCGAA-3¢) or
ADRB2 (target sequence: 5¢-CAGAGTGGATATCACGTGGAA-
3¢) from Qiagen
and incorporated them into a neutral liposome (1,2-dioleoyl-sn-
glycero-3-
phosphatidlycholine (DOPC)), as previously described10,24.
Inhibition of VEGF-R signaling. We blocked VEGF-R signaling
using PTK787
(Novartis) daily25. We also used the monoclonal VEGF-specific
antibody
bevacizumab (Genentech) in separate experiments.
Real-time RT-PCR. We extracted total RNA (Trizol reagent;
Invitrogen) and
performed reverse transcription using an oligo(dT) primer and
Moloney
murine leukemia virus (M-MLV) reverse transcriptase (Life
Technologies).

After PCR amplification, we obtained quantitative values (we
normalized each
sample on the basis of its I8S content). We obtained the VEGF
primers from
Applied Biosystems (number HS00173626-m1).
Promoter analysis. We transiently transfected ovarian cancer
cells
(Invitrogen Lipofectamine 2000) with the VEGF promoter-
reporter constructs
(obtained from L. Ellis, M.D. Anderson Cancer Center) and
determined
luciferase activity in triplicate as previously described26.
Immunohistochemistry. We stained paraffin sections for CD31
(1:800 dilu-
tion, Pharmingen), MMP-2 (1:500 dilution, Chemicon), MMP-9
(1:200 dilu-
tion, Calbiochem), VEGF (1:500 dilution, Santa Cruz
Biotechnology) and
bFGF (1:1,000 dilution, Sigma) at 4 1C (refs. 8,24). We
exposed control samples
to secondary antibody alone, and they did not show any
nonspecific staining.
Details for immunostaining fixed frozen sections for CD31 and

desmin have
been reported previously8. To quantify MVD, we examined ten
random
0.159 mm2 fields at 100� magnification for each tumor and
counted the
microvessels within those fields8. We evaluated the degree of
pericyte coverage
in five fields per tumor (200� magnification) as either absent or
present
(defined as covering 450% of the vessel). We then calculated,
for each tumor,
the average percentage of covered vessels relative to uncovered
vessels.
In vivo angiogenesis assay with Matrigel. We exposed
SKOV3ip1 cells
to serum-free medium with 1 mM of norepinephrine or vehicle
for 24 h.
We collected supernatants and centrifuged them to remove cells.
We infused
the conditioned medium with phenol-red free Matrigel (2:3
proportion,
total 500 ml; BD Biosciences)27. We performed subcutaneous
injections in
mice for the three treatment groups (n ¼ 3), and VEGF (0.25
nM)
served as positive control. We killed the mice after 8 d and
extracted the
Matrigel plug for hemoglobin content (QuantiChrom

hemoglobin assay;
BioAssay Systems).
In situ hybridization. We performed in situ hybridization for
VEGF and
analyzed results as described previously (Supplementary
Methods online)28.
Tumor vessel imaging. We anesthetized tumor-bearing mice and
injected
them intravenously with 100 ml of fluorescein Lycopersicon
esculentum lectin
(Vector Laboratories). Ten minutes later, we perfused mice
through the
ascending aorta with 4% paraformaldehyde for 2 min. Tumors
were extracted,
fixed, embedded in OCT, and sectioned for visualization using
confocal
fluorescence microscopy29.
Magnetic resonance imaging. We performed all magnetic
resonance imaging
experiments using a 4.7-T Biospec small animal imaging system
(Bruker
Biospin USA). For anatomic tumor imaging, we used axial T2-
weighted fast
spin echo sequence (TE/TR 60 ms/1,050 ms; in-plane resolution

120 mm; slice
thickness 1 mm; skip 0.25 mm; 3 nex) to delineate the tumor.
For functional
imaging, we acquired a series of axial, dynamic-contrast-
enhanced images
during and after contrast (diluted Magnevist) injection through
a tail-vein
catheter. We used an interleaved fast spoiled gradient echo
sequence (TE/TR
1.4 ms/40 ms; in-plane resolution 312 mm � 234 mm; 1 mm
slice thickness;
L E T T E R S
2006 943
©
2
0
0
6
N
a
tu
re
P

u
b
lis
h
in
g
G
ro
u
p
h
tt
p
:/
/w
w
w
.n
a
tu
re
.c
o
m

/n
a
tu
re
m
e
d
ic
in
e
update rate 5.5 s per slice package) to monitor contrast
accumulation in blood
vessels and tumor. We analyzed regions of interest
encompassing the enhancing
periphery of the tumor. We measured functional tissue
parameters by fitting
contrast levels to a two-compartment pharmacokinetic model30.
Assessment of tumor VEGF levels. We quantified
concentrations of VEGF in
tumor tissue homogenates using an ELISA kit (R&D Systems).
Statistical analysis. We compared continuous variables with
either Student’s

t-test or analysis of variance (ANOVA) and compared
categorical variables with
w2. We used nonparametric tests (Mann-Whitney test), if
appropriate,
to compare differences. We considered P o 0.05 to be
significant.
Note: Supplementary information is available on the Nature
Medicine website.
ACKNOWLEDGMENTS
The authors thank I.J. Fidler, J. Price, C. Bucana, G. Gallick
and A. Johnson for
their helpful input and discussions regarding this work. We
thank D. Reynolds
for assistance with immunohistochemistry. We also thank J.
Johnson and
E. Schrock for assistance with manuscript preparation. This
work was supported
by US National Institutes of Health (NIH) grants (CA-11079301
and CA-
10929801), the Donna Marie Cimitile-Fotheringham Award for
Ovarian Cancer
Research, the U.T. M.D. Anderson Ovarian Cancer SPORE
(2P50 CA083639-
06A1), and a Program Project Development Grant from the
Ovarian Cancer
Research Fund, Inc. to A.K.S., NIH grant CA-104825 to S.K.L.,
and NIH grant
A152737 to S.W.C.
AUTHOR CONTRIBUTIONS
P.H.T., L.Y.H. and A.A.K. designed and performed experiments
and wrote

portions of the manuscript. J.M.A. and R.T. designed and
performed VEGF
transcription and promoter assays and edited the manuscript.
C.L., N.B.J.,
G.A.-P., W.M.M., Y.G.L., M.L., L.S.M., T.J.K., C.N.L. and
Y.L. designed and
performed experiments and edited the manuscript. E.F. and
R.A.N. designed
and performed catecholamine analyses and edited the
manuscript. J.A.B.,
M.R. and V.K. designed and performed the imaging experiments
and edited
the manuscript. A.M.S. and G.L.-B. performed siRNA
incorporations and edited
the manuscript. R.L.C. and D.M.G. performed statistical
analyses and edited the
manuscript. S.K.L., S.W.C. and A.K.S. designed the overall
study, analyzed data
and edited the manuscript.
COMPETING INTERESTS STATEMENT
The authors declare that they have no competing financial
interests.
Published online at http://www.nature.com/naturemedicine/
Reprints and permissions information is available online at
http://npg.nature.com/
reprintsandpermissions/
1. Antoni, M.H. et al. The influence of bio-behavioural factors
on tumour biology:
pathways and mechanisms. Nat. Rev. Cancer 6, 240–248 (2006).
2. Justice, A. Review of the effects of stress on cancer in

laboratory animals: importance
of time of stress application and type of tumor. Psychol. Bull.
98, 108–138 (1985).
3. Ben-Eliyahu, S. The promotion of tumor metastasis by
surgery and stress: immuno-
logical basis and implications for psychoneuroimmunology.
Brain Behav. Immun. 17
(Suppl.), S27–S36 (2003).
4. Reiche, E.M., Nunes, S.O. & Morimoto, H.K. Stress,
depression, the immune system,
and cancer. Lancet Oncol. 5, 617–625 (2004).
5. McEwen, B. Seminars in Medicine of the Beth Israel
Deaconness Medical Center:
protective and damaging effects of stress mediators. N. Engl. J.
Med. 338, 171–179
(1998).
6. Lutgendorf, S.K. et al. Vascular endothelial growth factor
and social support in patients
with ovarian carcinoma. Cancer 95, 808–815 (2002).
7. Lutgendorf, S.K. et al. Stress-related mediators stimulate
vascular endothelial growth
factor secretion by two ovarian cancer cell lines. Clin. Cancer
Res. 9, 4514–4521
(2003).
8. Thaker, P.H. et al. Antivascular therapy for orthotopic human
ovarian carcinoma
through blockade of the vascular endothelial growth factor and
epidermal growth factor
receptors. Clin. Cancer Res. 11, 4923–4933 (2005).

9. Paredes, A. et al. Stress promotes development of ovarian
cysts in rats: the possible
role of sympathetic nerve activation. Endocrine 8, 309–315
(1998).
10. Landen, C.N. et al. Therapeutic EphA2 gene targeting in
vivo using neutral liposomal
small interfering RNA delivery. Cancer Res. 65, 6910–6918
(2005).
11. Langley, R.R. et al. Tissue-specific microvascular
endothelial cell lines from H-2K(b)-
tsA58 mice for studies of angiogenesis and metastasis. Cancer
Res. 63, 2971–2976
(2003).
12. Glaser, R. & Kiecolt-Glaser, J.K. Stress-induced immune
dysfunction: implications for
health. Nat. Rev. Immunol. 5, 243–251 (2005).
13. Lutgendorf, S. et al. Social support, psychological distress
and natural killer cell
activity in ovarian cancer. J. Clin. Oncol. 23, 7105–7113
(2005).
14. Chrousos, G.P. & Gold, P.W. The concepts of stress and
stress system disorders.
Overview of physical and behavioral homeostasis. J. Am. Med.
Assoc. 267,
1244–1252 (1992).
15. Basu, S. et al. The neurotransmitter dopamine inhibits
angiogenesis induced
by vascular permeability factor/vascular endothelial growth
factor. Nat. Med. 7,
569–574 (2001).

16. Teunis, M.A. et al. Reduced tumor growth, experimental
metastasis formation,
and angiogenesis in rats with a hyperreactive dopaminergic
system. FASEB J. 16,
1465–1467 (2002).
17. Seeman, T.E., Berkman, L.F., Blazer, D. & Rowe, J.W.
Social ties and support and
neuroendocrine function: the McArthur studies of successful
aging. Ann. Behav. Med.
16, 95–106 (1994).
18. Turner-Cobb, J.M., Sephton, S.E., Koopman, C., Blake-
Mortimer, J. & Spiegel, D.
Social support and salivary cortisol in women with metastatic
breast cancer.
Psychosom. Med. 62, 337–345 (2000).
19. Antoni, M.H. et al. Cognitive-behavioral stress management
reduces distress and
24-hour urinary free cortisol output among symptomatic HIV-
infected gay men.
Ann. Behav. Med. 22, 29–37 (2000).
20. Costanzo, E.S. et al. Psychosocial factors and interleukin-6
among women with
advanced ovarian cancer. Cancer 104, 305–313 (2005).
21. Sheridan, J.F. et al. Restraint stress differentially affects
anti-viral cellular and humoral
immune responses in mice. J. Neuroimmunol. 31, 245–255
(1991).
22. Aarons, R.D. & Molinoff, P.B. Changes in the density of
beta adrenergic receptors in rat

lymphocytes, heart and lung after chronic treatment with
propranolol. J. Pharmacol.
Exp. Ther. 221, 439–443 (1982).
23. Exton, M.S. et al. Behaviorally conditioned
immunosuppression in the rat is regulated
via noradrenaline and beta-adrenoceptors. J. Neuroimmunol.
131, 21–30 (2002).
24. Halder, J. et al. Focal adhesion kinase targeting for therapy
of ovarian carcinoma
using in vivo siRNA delivery in neutral liposomes. Clin. Cancer
Res. (in the
press).
25. Xu, L. et al. Inhibition of malignant ascites and growth of
human ovarian carcinoma by
oral administration of a potent inhibitor of the vascular
endothelial growth factor
receptor tyrosine kinases. Int. J. Oncol. 16, 445–454 (2000).
26. Reinmuth, N. et al. Induction of VEGF in perivascular cells
defines a potential
paracrine mechanism for endothelial cell survival. FASEB J. 15,
1239–1241 (2001).
27. Passaniti, A. et al. A simple, quantitative method for
assessing angiogenesis and
antiangiogenic agents using reconstituted basement membrane,
heparin, and fibro-
blast growth factor. Lab. Invest. 67, 519–528 (1992).
28. Radinsky, R. et al. A rapid colorimetric in situ messenger
RNA hybridization technique
for analysis of epidermal growth factor receptor in paraffin-
embedded surgical speci-

mens of human colon carcinomas. Cancer Res. 53, 937–943
(1993).
29. Hu, L. et al. Vascular endothelial growth factor trap
combined with paclitaxel strikingly
inhibits tumor and ascites, prolonging survival in a human
ovarian cancer model. Clin.
Cancer Res. 11, 6966–6971 (2005).
30. Daldrup, H. et al. Correlation of dynamic contrast-enhanced
MR imaging with
histologic tumor grade: comparison of macromolecular and
small-molecular contrast
media. AJR Am. J. Roentgenol. 171, 941–949 (1998).
L E T T E R S
MEDICINE
©
2
0
0
6
N
a
tu
re
P
u

b
lis
h
in
g
G
ro
u
p
h
tt
p
:/
/w
w
w
.n
a
tu
re
.c
o
m
/n

30 Nov 2001 107 AR AR146-04.tex AR146-04.SGM LaTeX2e(200105.docx

Recommended

Recommended

More Related Content

Similar to 30 Nov 2001 107 AR AR146-04.tex AR146-04.SGM LaTeX2e(200105.docx

Similar to 30 Nov 2001 107 AR AR146-04.tex AR146-04.SGM LaTeX2e(200105.docx (20)

More from rhetttrevannion

More from rhetttrevannion (20)

Recently uploaded

Recently uploaded (20)

30 Nov 2001 107 AR AR146-04.tex AR146-04.SGM LaTeX2e(200105.docx