Altered immune pathway activity under exercise challenge in Gulf War Illness:
An exploratory analysis
Gordon Brodericka,⇑,1, Rotem Ben-Hamob,2, Saurabh Vashishthaa,3, Sol Efronib,4, Lubov Nathansonc,5,
Zachary Barnesd,6, Mary Ann Fletcherd,7, Nancy Klimasc,e,8
aDepartment of Medicine, University of Alberta, Edmonton, Canada
bGoodman Faculty of Life Science, Bar Ilan University, Ramat Gan, Israel
cMiami Veterans Affairs Medical Center, Miami, FL, USA
dDepartment of Medicine, University of Miami, Miami, FL, USA
eInstitute for Neuro-Immune Medicine, Nova Southeastern University, Ft. Lauderdale, FL, USA
a r t i c l ei n f o
Received 5 July 2012
Received in revised form 28 October 2012
Accepted 13 November 2012
Available online 29 November 2012
Gulf War Illness
a b s t r a c t
Though potentially linked to the basic physiology of stress response we still have no clear understanding
of Gulf War Illness (GWI), a debilitating illness presenting with a complex constellation of immune, endo-
crine and neurological symptoms. Here we compared male GWI (n = 20) with healthy veterans (n = 22)
and subjects with chronic fatigue syndrome/myalgic encephalomyelitis (CFS/ME) (n = 7). Blood was
drawn during a Graded eXercise Test (GXT) prior to exercise, at peak effort (VO2 max) and 4-h post exer-
cise. Affymetrix HG U133 plus 2.0 microarray gene expression profiling in peripheral blood mononuclear
cells (PBMCs) was used to estimate activation of over 500 documented pathways. This was cast against
ELISA-based measurement of 16 cytokines in plasma and flow cytometric assessment of lymphocyte pop-
ulations and cytotoxicity. A 2-way ANOVA corrected for multiple comparisons (q statistic <0.05) indi-
cated significant increases in neuroendocrine-immune signaling and inflammatory activity in GWI,
with decreased apoptotic signaling. Conversely, cell cycle progression and immune signaling were
broadly subdued in CFS. Partial correlation networks linking pathways with symptom severity via
changes in immune cell abundance, function and signaling were constructed. Central to these were
changes in IL-10 and CD2+ cell abundance and their link to two pathway clusters. The first consisted
of pathways supporting neuronal development and migration whereas the second was related to
androgen-mediated activation of NF-jB. These exploratory results suggest an over-expression of known
exercise response mechanisms as well as illness-specific changes that may involve an overlapping stress-
potentiated neuro-inflammatory response.
? 2012 Elsevier Inc. All rights reserved.
Within months of returning from Operation Desert Storm an
alarming number of Gulf War veterans began reporting fatigue,
musculoskeletal discomfort, skin rashes, and cognitive dysfunction
(Haley, 1997; Fukuda et al., 1998; Wolfe et al., 1998). We still have
no clear understanding of Gulf War Illness (GWI) although the basic
physiology of response to stress whetherpsychological, chemical or
other provides a starting point. Indeed clinical presentation of GWI
overlaps strongly with that of another stress-mediated illness:
Chronic fatigue syndrome/myalgic encephalomyelitis (CFS/ME)
(Kang et al., 2003; Eisen et al., 2005). Though brain imaging has
shown promise (Liu et al., 2011; Haley et al., 2009), single biomark-
ers for these illnesses in the lymphocyte transcriptome remain
elusive (Byrnes et al., 2009). Early RT-PCR amplification of cytokine
associated transcripts by Zhang et al. (1999) indicated that Gulf
War veterans diagnosed with chronic fatigue syndrome (CFS/ME)
0889-1591/$ - see front matter ? 2012 Elsevier Inc. All rights reserved.
⇑Corresponding author. Address: Div. of Pulmonary Medicine, Department of
Medicine, University of Alberta, WMC 2E4.41 WC Mackenzie Health Sciences, 8440-
112 Street, Edmonton, AB, Canada T6G 2R7. Tel.: +780 492 1633; fax: +780 407
systemsbiomed.org (R.Ben Hamo), email@example.com (S. Vashishtha), sefroni@
miami.edu (Z. Barnes), MFletche@med.miami.edu (M.A. Fletcher), nklimas@nova.
edu (N. Klimas).
1Tel.: +1 780 492 1633.
2Tel.: +1 780 566 9406.
3Tel.: +972 3 7384519.
4Tel.: +972 3 7384518.
5Tel.: +1 305 607 8227.
6Tel.: +1 305 243 6218.
7Tel.: +1 305 243 6288.
8Tel.: +1 305 575 3267.
Brain, Behavior, and Immunity 28 (2013) 159–169
Contents lists available at SciVerse ScienceDirect
Brain, Behavior, and Immunity
journal homepage: www.elsevier.com/locate/ybrbi
had significantly higher levels of IL-2, IL-10, IFN-c, and TNF-a tran-
script than healthy controls while non-veteran CFS/ME subjects did
not differ. More recent work by our group (Whistler et al., 2009) fo-
cused on the expression in Gulf War veterans of transcripts associ-
ated with natural killer (NK) cell function before, during and after a
maximal graded exercise challenge (VO2 max). In this specific
group of 108 microarray probe sets, 49 were increased in expres-
sion 2-fold and none were decreased in GWI. Specifically, mRNA
for perforin (PRF1) and granzyme (GZMB), along with killer cell lec-
tin-like receptors (KLR) and death receptor induction (FASLG) indi-
cated a depressed NK cell cytotoxic response to exercise in GWI.
Additional surveys have been done using animal models and
have focused on exposure to chemical agents used in theatre;
namely organophosphates and acetylcholine esterase (AChE)
inhibitors such as pyridostigmine bromide (PB) (Golomb, 2008;
Amourette et al., 2009) and sarin (Shewale et al., 2012; Mach
et al., 2008). Though much of this has focused on proteomic and
phospho-proteomic profiling (Torres-Altoro et al., 2011; Zhu
et al., 2010), recent work by Barbier et al. (2009) surveyed the
expression of genes associated with stress response, learning and
memory in the hippocampus and hypothalamus of mice. Their re-
sults indicated that stressed animals exposed to PB showed in-
creases in hippocampal expression of three genes implicated in
memory development: brain-derived neurotrophic factor (BDNF),
tropomyosin-related kinase B (TrkB) and calcium/calmodulin-
protein kinase II alpha (CamKII). This type of survey is not easily
performed in human subjects however circulating lymphocytes
have beenshown to reflect changes inthe brain(Mutezet al., 2011).
As cholinergic signaling is a known modulator of immune re-
sponse (Kawashima and Fujii, 2003) we wanted to explore if NF-
jB activity might be altered in the circulating lymphocytes of
GWI subjects perhaps as a lasting result of exposure in theatre to
the AChE inhibitors described above. We used a maximal exercise
challenge to improve detection as NF-jB (Kim et al., 2009) and
many other components of immune function (Walsh et al., 2011)
are responsive to exercise as are neuro-endocrine modulators
thereof, such as cortisol and neuropeptide Y (NPY) (Jonsdottir,
2000). Consistent with this, our group found previously that IL-1,
an NF-jB modulator, exerted a strong influence on exercise-in-
duced immune cell signaling in GWI (Broderick et al., 2011). Based
on this, we hypothesized that cholinergic modulation of endocrine-
immune function during exercise would differ in GWI subjects.
This would manifest not only through changes in a variety of
inflammatory pathways but also in associated metabolic co-regu-
lators such as mTOR (Waickman and Powell, 2012; Chi, 2012) as
well as in neurotransmission pathways responsive to exercise
To evaluate these effects we performed an exploratory analysis
of gene expression using a novel methodology (Efroni et al. (2007,
2008)) for estimating the activity level of over 400 pathways de-
scribed in the National Cancer Institute (NCI)/Nature Pathway
Interaction Database (PID) (Schaefer et al., 2008) and the Kyoto
Encyclopedia of Genes and Genomes (KEGG) database (Kanehisa
et al., 2010). Subjects were then compared in terms of the activity
level of these pathways. This is a significant departure from con-
ventional analysis, which is based only on those pathways sup-
ported by genes expressed significantly across groups (Kerr et al.
2008a,b; Whistler et al., 2009) and where no estimate of pathway
activation is produced. Using this alternative approach we esti-
mated pathway activation from gene expression in peripheral
blood mononuclear cells (PBMC) collected before exercise, at peak
effort and 4 h after exercise in veterans with GWI and healthy vet-
erans. We also used a smaller pilot cohort of CFS/ME subjects as a
disease control group. Using a 2-way ANOVA and correcting for
false discovery we found 19 pathways uniquely expressed in
GWI compared to CFS/ME and healthy control subjects. To further
enforce biological relevance we constructed association networks
describing the cascade of pathway activation to lymphocyte popu-
lation, cytotoxic function, cytokine signaling and symptom severity
in GWI. These suggested a broad support of pathways involved in
neuronaldevelopmentand migration as wellas hormone-mediated
NF-jB activation and changes in apoptotic signaling. While
elements of these signatures overlapped in part with expected
exercise response mechanisms, we found that their activity was
disproportionate in GWI and that components not normally
responsive to exercise were being recruited.
2. Material and methods
2.1. Sample collection and processing
2.1.1. Cohort recruitment
As part of a larger ongoing study a subset of CFS/ME (n = 7), GWI
subjects (n = 20) and healthy but sedentary Gulf War era veterans
(n = 22) were recruited from the Miami Veterans Administration
Medical Center. All subjects were comparable in age, body mass in-
dex (BMI), ethnicity and duration of illness. Subjects were male
and ranged in age between 30 and 55. Inclusion criteria was de-
rived from Fukuda et al. (1998), and consisted in identifying veter-
ans deployed to the theater of operations between August 8, 1990
and July 31, 1991, with one or more symptoms present after
6 months from at least 2 of the following: fatigue; mood and cog-
nitive complaints; and musculoskeletal complaints. Subjects were
in good health prior to 1990, and had no current exclusionary diag-
noses (Reeves et al., 2003). Medications that could have impacted
immune function were excluded. Use of the Fukuda definition in
GWI is supported by Collins et al. (2002). Summary results of sub-
ject demographics and exercise performance are listed in Table 1.
Ethics statement. All subjects signed an informed consent ap-
proved by the Institutional Review Board of the University of Mia-
mi. Ethics review and approval for data analysis was obtained by
the IRB of the University of Alberta.
2.1.2. Subject assessment
All subjects received a physical examination and medical his-
tory including the GWI symptom checklist as per the case defini-
tion. Psychometric questionnaires included the Multidimensional
Fatigue Inventory (MFI) (Smets et al., 1995), a 20-item self-report
instrument designed to measure fatigue, and the Medical Out-
comes Study 36-item short-form survey (SF-36) (Ware and Sher-
bourne, 1992) assessing health-related quality of life. The Krupp
Fatigue Severity Inventory (Krupp FSI) was also used to measure
perceptions of fatigue severity (Krupp et al., 1989) while the im-
pact of symptoms on the activities of daily life was measured with
the Sickness Impact Profile (SIP) (Bergner et al., 1981). Subjects
Summary of mean and standard error values () for demographic variables and
exercise performance in GWI, CFS/ME and healthy control groups.
Number of subjects
Body–mass index (BMI) (kg/m2)
Peak VO2 max (mL/kg/min)
% Predicted peak VO2 max (%)
Time to VO2 max (min)
aTwo-tailed t test significant at 0.05 level versus HC.
G. Broderick et al./Brain, Behavior, and Immunity 28 (2013) 159–169
were screened for quantity and quality of sleep, and evaluated for
the likelihood of primary sleep disorders using the Pittsburgh Sleep
Quality Index (PSQI) (Buysse et al., 1989). Designed to assess symp-
toms of post-traumatic stress disorder (PTSD), the Davidson Trau-
ma Scale (DTS) (Davidson et al., 1997) was applied to those
subjects who reported a traumatic experience (death of loved
one, assault, injury, etc....). This instrument is divided into three
components: intrusion, avoidance, and hyper-arousal. While 66%
of the GWI cohort presented DTS scores consistent with PTSD, this
data is available only for a small fraction of CFS/ME subjects and
healthy control subjects. Finally, aspects of cognitive impairment
were assessed using the Paced Auditory Serial Addition Task (PA-
SAT). This serial-addition task is used to assess the rate of informa-
tion processing,sustained attention,
(Gronwall, 1977). A summary of symptom severity measures and
results obtained in each group is presented in Supplemental table
Immune response was stimulated with a standard maximal
Graded eXercise Test (GXT) using a Vmax Spectra 29c Cardiopul-
monary Exercise Testing Instrument, Sensor-Medics Ergoline 800
fully automated cycle ergometer, and SensorMedics Marquette
MAX 1 Sress ECG. According to the McArdle protocol (McArdle
et al., 2007) subjects pedaled at an initial output of 60 W for
2 min, followed by an increase of 30 W every 2 min until the sub-
ject reached: (1) a plateau in maximal oxygen consumption (VO2);
(2) a respiratory exchange ratio >1.15; or (3) the subject stopped
the test. A first blood draw was conducted prior to exercise follow-
ing a 30-min rest. Second and third blood draws were conducted
upon reaching peak effort (VO2 max) and at 4-h post exercise
respectively. Summary exercise performance results are presented
in Table 1. All control subjects were screened as sedentary based
on their response to a questionnaire upon recruitment. This was
subsequently confirmed by the aerobic capacities obtained (VO2
max) (Hurwitz et al., 2010). Adjusted maximum VO2 levels were
comparable between GWI and control groups but were signifi-
cantly lower for the CFS disease control group (p < 0.05). Trends ex-
isted towards shorter exercise bouts in both illness groups but
these did not achieve statistical significance.
2.1.3. Gene expression
At each blood draw three 8-mL tubes of blood were collected in
CPT vacutainers (B-D-Biosciences, San Jose, CA). The peripheral
blood mononuclear cells (PBMC) were isolated and stored in liquid
nitrogen under conditions designed to maintain viability. Specifi-
cally, whole blood was added to Ficoll-Paque, centrifuged at
1000g for 25 min. PBMC’s were isolated from the PBMC ring atop
the Ficoll layer into a separate tube, centrifuged at 300g for
10 min, then re-suspended in PBS. Cells were then counted using
a Beckman Coulter viCell, and cryopreserved in freezing media
(temperature lowered 1 ?C/min until @ ?80 ?C).
Total RNA was extracted using TRI Reagent (Molecular Research
Center, Cincinnati, OH) following the manufacturer’s protocol. The
quality and quantity of RNA was assessed using the Agilent Bioan-
alyzer 2100 RNA 6000 Nano Kit (Agilent Technologies, CA). From
each sample, 300 ng of total RNA was converted into cDNA by re-
verse transcription using a T7-oligo(dT) primer and the Affymetrix
30IVT Express Kit (Affymetrix, Santa Clara, CA) according to stan-
dard manufacturer protocol. The generated cDNA was purified
using the GeneChip Sample Cleaning Module (Affymetrix) and la-
beled cRNA was generated by in vitro transcription using the bio-
tinylated nucleotide mix. This was then purified with the Cleaning
Module and quantified using the Nanodrop ND-1000 spectropho-
tometer (NanoDrop Technologies, Inc., Wilmington, DE USA). In
each preparation 11 lg cRNA was fragmented in Fragmentation
Buffer (Affymetrix) in a final reaction volume of 25 ll.
Hybridization, washing, staining and scanning were done using
Affymetrix GeneChip instruments (Hybridization Oven 640, Fluid-
ics Station 450Dx, Scanner GCS3000Dx) and Affymetrix Human
U133 2.0 arrays (Affymetrix) as per manufacturer’s standards.
Microarray image files (.cel data) were generated using the
Affymetrix GCOS software tool with default microarray analysis
parameters to provide overall within chip normalization of the im-
age intensity distribution. The quality parameters that were mon-
itored besides cRNA total yield and cRNA A260/A280 ratio
included: (i) background noise (Q value), (ii) percentage of present
called probe sets, (iii) scaling factor, (iv) information about exoge-
nous Bacillus subtilis control transcripts from the Affymetrix Poly-A
control kit (lys, phe, thr, and dap), and (v) the ratio of intensities of
30probes to 50probes for a housekeeping gene (GAPDH).
In a quality control step a custom panel of 10 gene transcripts,
plus GAPDH as a housekeeping gene, were surveyed using the
NanoString nCounter system (http://www.nanostring.com/prod-
ucts/assays/). This code set contained a 30biotinylated capture
probe and a 50reporter probe tagged with a fluorescent barcode,
two sequence-specific probes for each of 10 transcripts. Probes
were hybridized in triplicate for 19 h at 65 ?C to 100 ng of total
RNA collected from 15 patients (5 patients with CFS, 5 patients
with GWI and 5 healthy control patients) for all three time points.
Excess capture and reporter probes were removed and transcript-
specific ternary complexes were immobilized on a streptavidin-
coated cartridge using the NanoString preparation station robotic
fluids handling platform as per manufacturer’s standard protocol.
The nCounter Digital Analyzer was used to count individual fluo-
rescent barcodes and quantify target RNA molecules present in
each sample. Background hybridization signal was determined
using spike in negative controls, and all mRNAs with expression
below mean background + 2 standard deviations were considered
below the detection limit. Raw code count was then normalized
based on the relative number of positive control counts. This back-
ground was then subtracted from the raw transcript count. A com-
parison of nCounter and mRNA-seq data is available in Sun et al.
(2011). Analysis was conducted at the Oncogenomics Core Facility
at the University of Miami (Miami, USA) Results of this partial val-
idation step are presented in Table S2.
2.1.4. Cell cytometry and cytotoxicity
Flow cytometry was performed on each sample to determine
lymphocyte subset abundance using a Beckman/Coulter FC500.
Whole blood samples were stained in 5 color combinations, with
the appropriate concentrations of antibodies, erythrocytes lysed
and the cells fixed with the Optilyse C reagent (Beckman-Coulter
Corp., Hialeah, FL). Lymphocyte, monocyte and granulocyte popu-
lations were determined using light scatter and back gating on
fluorescence for the CD45 bright and CD14 negative population.
The isotype control served as reference for negative events. Spec-
tral compensation was established daily. Quality control included
optimization for lymphocyte recovery, purity of analysis gate, lym-
phosum, and replicate determined according to CDC guidelines. In
addition, measurement of intracellular cytotoxic protein concen-
trations was performed using quantitative fluorescence. Levels of
intracellular perforin, granzyme A or granzyme B, conjugated to
phycoerythrin within lymphocytes subsets, both NK cells and
CD8+ T cells, were simultaneously assessed with a 5-color system
using a maximum-yield protocol (Maher et al., 2002; Broderick
et al., 2011). Summary statistics for these markers are listed at each
phase of exercise for each subject group in Table S3.
2.1.5. Cytokine profiles
Plasma was also separated from each sample within 2 h of col-
lection and stored at ?80 ?C until assayed. We measured 16 cyto-
kines in plasma using Quansys reagents and instrument (Quansys
G. Broderick et al./Brain, Behavior, and Immunity 28 (2013) 159–169
Biosciences, Logan, Utah). The Quansys Imager, driven by an 8.4
megapixel Canon 20D digital SLR camera, supports 96 well plate
based chemiluminescent imaging. The Q-Plex™ Human Cytokine-
Screen (16-plex) is a quantitative enzyme-linked immunoabsor-
bent assay (ELISA)-based test where sixteen distinct capture
antibodies have been absorbed to each well of a 96-well plate in
a defined array. The range of the standard curves and exposure
time were adjusted previously to provide reliable comparisons
between subject groups in this illness population at both low
and high cytokine concentrations in plasma. Second order polyno-
mial regression models were used as standard calibration curves.
Quadruplicate determinations were made, i.e., each sample was
run in duplicate in two separate assays. Statistics reported previ-
ously in Broderick et al. (2010) indicated an average coefficient
of variability (CV) of 0.20 for inter-assay and 0.09 for intra-assay
repeatability. Summary statistics for cytokine concentrations at
each phase for each subject group are listed in Table S3.
2.2. Numerical analysis
2.2.1. Comparative analysis of gene expression
Prior to analysis raw microarray data was corrected for chip-to-
chip variability. Removal of background signal intensity was per-
formed using GeneChip RMA (GC-RMA), an extension of classical
Robust Multichip Average (RMA) algorithm (Wu et al., 2004; Wu
and Irizarry, 2005; Katz et al., 2006). The range of intensity values
were adjusted using quantile normalization (Bolstad et al., 2003).
Expression values were transformed logarithmically (log2) to im-
prove normality. The significance of changes in gene expression
across subject groups and across all 3 time-points was evaluated
using a 2-way analysis of variance (ANOVA-2). In all cases, null
probability values were adjusted for multiple comparisons using
the method proposed by Storey (2002) for estimating the false dis-
covery rate (FDR) and probability of false discovery q. Gene and
pathway associations of Affymetrix probe sets were obtained using
the Protein Analysis THrough Evolutionary Relationships (PAN-
THER) software system (Mi et al., 2010).
2.2.2. Estimating pathway activity
Individual gene products are expressed in a highly coordinated
manner that supports the function of biochemical reaction path-
ways in the cell. In accordance with this we have used a novel
algorithm developed by Efroni et al. (2007, 2008) to estimate
the activity of known pathway segments from the expression of
the genes that encode their components. The first step consists
of converting a continuous measure of gene expression into a dis-
crete gene status, namely up-expressed or down-expressed. In
brief, for every individual gene the distribution of expression val-
ues across all samples are numerically fit to two separate gamma
probability distribution functions: one describing the distribution
of expression values supporting the up state and one describing
the distribution of values supporting the down state. The expecta-
tion maximization (EM) algorithm is applied to provide a set of
maximum-likelihood estimates for the parameters ai and bi as
well as the values of the mixture coefficients githat weigh the
contribution of the gamma distribution (Eqs. 1–3) for each state
i = 1,2. The vector of these parameters is estimated iteratively
from H0to H such that the function Q (H, H0) is maximized
xt;iðlogðgiÞ ? logðcðxt;ai;biÞÞð1Þ
The probability of a gene being up-expressed given its expres-
sion level x is the probability of occurrence of the ‘‘up’’ state overall
pUP= NUP/N multiplied by the probability of expression level x cor-
responding to an up-expressed state or c(x;aUP, bUP); where NUPis
the number of genes that are up-expressed among all N genes. It is
important to note that this manipulation does not assign a discrete
up or down state to a gene but instead provides a continuous scale
of expression that includes being neither up nor down-expressed.
Every pathway consists of a collection of reaction steps or inter-
actions. These can be modeled as logic functions with genes serv-
ing as inputs and outputs. In the current protocol the activation
level of such a logical function was computed based on the joint
conditional probability that the input genes k 2 set I are in an
up-expressed state based on measured gene expression (Eq. 4).
As we have estimated the corresponding probability that the set
of output genes k 2 O are also in an up-expressed state (Eq. 5),
we can use the agreement between input and output as a measure
of consistency Csfor the activation of reaction step s (Eq. 6).
Cs¼ pðs ¼ activeÞ ? pðO ¼ \up"Þ þ ð1 ? pðs
¼ activeÞÞ ? ð1 ? pðO ¼ \up"ÞÞ
Using the approach outlined in Eqs. (1)–(6), activation levels
and consistency were calculated in each patient sample for all
reaction steps described in 582 candidate pathways aggregated
from the National Cancer Institute (NCI)–Nature Pathway Interac-
tion Database (PID) (Schaefer et al., 2008) and the Kyoto Encyclo-
pedia of Genes and Genomes (KEGG) database (Kanehisa et al.,
2010). The NCI-Nature PID database is itself an aggregation of
135 pathways curated by the NCI-Nature team with an additional
322 pathways imported from the BioCarta (www.biocarta.com)
and Reactome (Croft et al., 2011; Matthews et al., 2009) databases.
In each individual sample the activity of each pathway was calcu-
lated as the average activation level across all component interac-
tions. Activation scores in each subject group were log transformed
to improve normality and compared for each pathway at each time
point using both parametric (t test) and non-parametric (Wilcoxon
rank sum) tests. Once again, two-way analysis of variance (ANO-
VA-2) was used to assess the significance of group, time and
group ? time interactions with the false discovery rate (FDR) esti-
mated using Storey (2002). The potential of each pathway activa-
tion score as a diagnostic marker was described in terms of
receiver-operating characteristics (ROC) by the increase above ran-
dom in the area under the curve (AUC > 0.50). All the computations
were conducted with the MATLAB software environment (The
MathWorks Inc., Natick, MA).
pðs ¼ activeÞ ¼
pðO ¼ \up"Þ ¼
2.2.3. Identifying patterns of pathway-symptom interaction
To describe in statistical terms the correlation of pathway activ-
ity with symptom severity via changes in immune cell population,
function and signaling we constructed multi-layer empirical net-
works using partial linear correlation as a measure of association
(Emmert-Streib, 2007; Magwene and Kim, 2004). This measure ad-
justs the pair-wise correlation of marker xiwith marker xjfor the
indirect correlation contributed by a confounding marker xk.
G. Broderick et al./Brain, Behavior, and Immunity 28 (2013) 159–169
rðxi;xjÞ ? rðxi;xkÞrðxj;xkÞ
In order to cast each pair-wise association in the proper context
these were adjusted for age and BMI within the GWI subject group.
We summarized the contribution of these confounding effects by
computing the first principal component for these two markers
and using this as the context variable xk(Eq. 7). Null probability
p was computed by transforming the correlation to create a t
statistic having n?2 degrees of freedom for n observations. Confi-
dence bounds were based on an asymptotic normal distribution
of 0.5?log((1 + r(xixj|xk))/(1?r(xixj|xk))), with an approximate vari-
ance equal to 1/(n?3) when variables have a multivariate normal
distribution. Once again the probability of false discovery q was
estimated according to Storey (2002) with q < 0.05 indicating sig-
Based on this measure of association, multi-level hierarchical
networks were constructed in GWI linking pathway activation to
symptom severity through changes in immune cell abundance,
function and signaling. Minimum paths pi jlinking each pathway
xito each symptom yjwere computed and subjected to a logical
consistency constraint (Eq. 8). This stipulated that the direction
of change in the symptom score in GWI, sign(Dyj), must be consis-
tent with the direction of change in pathway activation, sign(Dxi),
when propagated along the sequence of correlation coefficients
rn massociated with each segment [n,m] of the shortest path pi j.
All calculations related to network identification and rationali-
zation, including minimum path calculations, were conducted with
the MATLAB software environment (The MathWorks Inc., Natick,
MA). The graphical rendering of correlation networks as well as
the analysis of network attributes was done using the Cytoscape
platform (Smoot et al., 2011) with the NetworkAnalyzer compo-
nent (Assenov et al., 2008).
ð1 ? r2ðxi;xkÞÞ
ð1 ? r2ðxj;xkÞÞ
signðDxiÞ ¼ signðDyjÞ
3.1. Differential expression of transcription-associated markers
The number of microarray probe sets changing in expression
across time, subject group, or showing a group ? time interaction
with a probability of false discovery q < 0.05 are shown in Table 2.
Over 1000 probes sets (1080) differed significantly in expression
across time in GWI. Using the PANTHER database this corre-
sponded to 644 genes associated with 79 pathways. For the CFS/
ME disease control group we found close to 300 probe sets (280)
changing across time. These mapped to 180 genes that were in turn
associated with 30 pathways. These changes across time were
dwarfed by the very broad differences in expression observed
across subject groups. Close to 9000 probe sets (8819) were differ-
entially expressed in GWI across all time points compared to the
healthy control group. These mapped to 4404 genes associated
with 126 pathways in the PANTHER database. In CFS/ME over
10,000 probe sets (10,222) changed significantly in expression
compared with the healthy control group. These probe sets corre-
sponded to 4963 genes that in turn mapped onto 132 pathways.
While significant effects were found for both time and group no
probe sets presented significant time ? group interactions with
q < 0.05.
3.2. Preferential activation of known pathways
Using the methodology proposed by Efroni et al. (2007, 2008)
we obtained estimates of pathway activation levels for each sam-
ple. As with the previous probe-level analysis, a 2-way ANOVA of
pathway activation found no significant time–group interactions.
However we also found that by enforcing pathway regulatory logic
and aggregating across multiple step reactions, changes across
time were no longer discernable. As shown in Table 2, only group
effects were significant at q < 0.05. In GWI, 127 pathways differed
significantly in activation level compared to the healthy control
group. This is comparable with the number found by relying on
the gene and pathway annotation of the probe sets. The changes
in the predicted activation level of 19 of these pathways were
found only in GWI and exceeded by 10% the activation level found
in the healthy control group for that same pathway (Table 3). Re-
sults indicated decreased activation in several key elements of
apoptotic signaling. For example, we found reduced pathway activ-
ity for tumor necrosis factor receptor 1 (TNFr1) and FAS/CD95
receptor signaling as well as tyrosine kinase mediated activation
of T cell receptor signaling and thrombospondin-1 (TSP-1) medi-
ated induction of cell death. This also included decreased kinase
A-anchoring protein AKAP95 activity and its involvement via cas-
pase-3 in the apoptotic caspase cascade. This was accompanied
by reduced activity in several transcriptional regulatory pathways.
In particular we found reduced activation of nuclear receptors het-
erodimers retinoic acid receptor (RAR) and retinoid X receptor
(RXR). In the absence of ligand, DNA-bound RXR/RARA are known
to represses transcription via co-repressors NCOR1, SMRT (NCOR2)
and histone deacetylase (HDAC). Also disrupted was activation of
RAs-related Nuclear protein (RAN), an androgen receptor (AR) co-
activator, via the RAN binding protein, as well as activation of
RNA polymerase III, a decisional regulator of cell survival during
In contrast, we observed increased activation of pathways
involving broad-acting immune modulator NF-jB (nuclear factor
kappa-light-chain-enhancer of activated B cells) (Figure S1). We
also found significantly elevated visual signal transduction and
activation of a KEGG super-pathway describing broad-scale recep-
tor–ligand interactions in neurotransmission. This was accompa-
nied by increased biosynthesis of phenylpropanoid compounds, a
process promoted by the dopamine precursor phenylalanine. Final-
ly, with possible ties to environmental exposure, we also observe
chronic activation of the detoxification pathway supporting degra-
dation of the androgen-disrupting hydrocarbon bisphenol.
While CFS/ME and GWI are commonly considered sister ill-
nesses, we found little overlap at the level of pathway activation.
Specifically, we found 25 pathways that were uniquely expressed
in CFS/ME (q < 0.05) and exceeded by 10% the corresponding acti-
vation level found in the healthy control group (Table S4). These
typically involved suppression of transcriptional regulation and
cell cycle. Changes in energy metabolism included activation of
antioxidant lipoic acid metabolism accompanied by decreases in
caffeine-induced lipolysis and leptin-mediated control of insulin
resistance. These changes coincided with decreased B cell receptor
signaling. Only 8 pathways were shared by both CFS/ME and GWI
(q < 0.05) with activation ranging from 93% to 110% of healthy
control (Table S5). Control of skeletal myogenesis calmodulin-
dependent kinase (CaMK) and granzyme A mediated apoptosis
were the most up and down-expressed respectively.
In a first partial validation, data collected beforehand for quality
control were examined for concordance with these results. A sur-
vey of 10 genes selected a priori for their involvement in immune
signaling was conducted using the NanoString nCounter system.
Results of a 2-way ANOVA indicated significant group effects for
the expression of beta actin (ACTB), fibroblast growth factor-bind-
ing protein 2 (FGFBP2), acidic ribosomal protein P0 (RPLP0) as well
as zinc finger and BTB domain-containing protein 16 (ZBTB16).
This was especially noticeable in GWI subjects at peak effort (T1)
for RPLP0 and ZBTB16 expression. Pathway annotations listed in
G. Broderick et al./Brain, Behavior, and Immunity 28 (2013) 159–169
Table S2B for ACTB and ZBTB16 show an overlap with the results in
Tables 3 and S5, in particular with regard to ACTB’s role in nAChR
signaling and its subsequent up-regulation of NF-jB activity
(Chernyavsky et al., 2010) via CaMK. Likewise changes in ZBTB16
aligned with predicted changes in co-repressor SMRT activity and
3.3. Pathway associations with symptom severity in GWI
The pathways described above are not expressed independently
of one another but instead belong to a well-integrated biochemical
network that both drives and responds to changes across multiple
levels of biology. To explore how these changes in pathway activa-
tion might be relevant to GWI we examined their correlation with
changes in 24 indicators of symptom severity measured at rest,
corrected for age and body–mass index (BMI) (Table S1). Rather
than examine direct association between pathway and symptom,
we introduced intermediate layers describing changes 12 lympho-
cyte sub-populations, 7 measures of cell cytotoxicity, and the
expression of 16 cytokines, NPY and cortisol. A change in pathway
activation was considered relevant to illness severity if it first cor-
related significantly changes in immune cell demographics and
function, then cell–cell signaling. We also imposed a consistency
constraint whereby the direction of correlation along this path
had to align with differences in pathway and symptom expression
observed across subject groups. This analysis of cell and protein
profiling was used to provide additional validation, limit spurious
associations and improve biological interpretation.
Using these constraints multi-level networks were constructed
from 518 candidate intracellular, cellular, intercellular and symp-
tom-level nodes. As some features of GWI might correlate best
with exercise capacity we constructed separate networks linking
baseline symptom severity with these markers at rest (T0), peak ef-
fort (T1) and at recovery (4 h after peak effort) (T2). Only associa-
tions with a partial correlation coefficient | r | > 0.10 and a
significance of q < 0.05 were conserved. From these networks we
extracted the shortest paths linking each symptom measure to
the 19 pathways significantly expressed in GWI (Table 3). When
enforcing hierarchical and significance constraints, no sub-net-
works were found linking symptom expression to GWI pathways
at rest. However at peak effort and in the recovery phase all 19
pathways were linked to symptom severity by between 4 and 12
edges (Tables S6 and 7), with correlation strength typically exceed-
ing | r | = 0.50 (Figure S2).
At peak effort this network consisted of 23 symptom nodes sep-
arated from the 19 GWI pathway nodes by a layer of 9 soluble
Summary of microarray probe sets with significant difference in expression level in GWS and CFS/ME groups compared to the healthy control group. Significance was based on
p-values for time, group and time ? group effects estimated from a 2-way ANOVA and set as p < 0.05. Adjustment for multiple comparisons was based on Storey’s q-statistic
(q < 0.05).
p Time < 0.05
q < 0.05
p Group < 0.05
q < 0.05p Time ? group < 0.05
q < 0.05
Number of probe sets
GWI vs HC all times
CFS vs HC all times
Number of genes (Unigene IDs mapped)
GWI vs HC all times
CFS vs HC all times
Number of pathways (from mapped IDs)
GWI vs HC all times
CFS vs HC all times
Number of pathways (PID path activation)
GWI vs HC all times
CFS vs HC all times
Significant differences (ANOVA-2) in inferred pathway activity unique to GWI (q < 0.05) with a mean absolute fold change (FC) in pathway activity in excess of 10% compared to
healthy control subjects (HC) (1.10 < pathway activity FC < 0.90). Performance as potential diagnostic markers is shown as area under receiver–operator characteristic (ROC)
curve (AUC) in excess of 0.50, the value expected from a random assignment.
Node # Pathway # Pathway name
q ValueROC AUC > 0.50Pathway activity FC (GWI:HC)
AKAP95 role in mitosis and chromosome dynamics (BioCarta)
TSP-1 induced apoptosis in microvascular endothelial cell (BioCarta)
RNA polymerase III transcription (BioCarta)
LCK and FYN tyrosine kinases in initiation of TCR activation (BioCarta)
Signaling events mediated by HDAC class III (NCI/Nature)
Lissencephaly gene (LIS1) in neuronal migration and development
FAS signaling pathway (CD95) (BioCarta)
Caspase cascade in apoptosis (BioCarta)
TNFr1 signaling pathway (BioCarta)
FAS signaling pathway (CD95)(NCI/Nature)
MAP kinase inactivation of SMRT corepressor (BioCarta)
RXR and RAR heterodimerization with other nuclear receptor
Bisphenol a degradation (KEGG)
Neuroactive ligand–receptor interaction (KEGG)
Stress induction of HSP regulation (BioCarta)
Phenylpropanoid biosynthesis (KEGG)
Atypical NF-jB pathway (NCI/Nature)
Visual signal transduction: cones (NCI/Nature)
NF-jB signaling pathway (BioCarta)
G. Broderick et al./Brain, Behavior, and Immunity 28 (2013) 159–169
mediators (cytokines, NPY and cortisol), 6 cell population–function
nodes and 10 intermediate pathway nodes. A similar network was
produced at recovery. Enforcing logical consistency propagated
downward from symptom severity resulted in rationalization of
this network leaving 10 of the original 19 GWI pathways (Fig. 1)
at peak effort only. No pathways satisfied this constraint at recov-
ery (T2). Scores for DTS, PASAT as well as SF36 social function and
emotional limit clustered together as a sub-network linked to IL-10
levels. Interestingly though IL-10 was significantly lower in GWI as
a group, within-group increases aligned with increased illness
severity (Figure S3). Similarly SF36 measures for physical function,
physical limit, pain, and vitality cluster together and were linked to
IL-1a, IL-2 and IL-5 levels. MFI indicators were more diffuse in their
association with reduced motivation, for example, standing apart
and correlating positively with changes in IL-4, 12 and IL-10.
SF36 general score also stood apart and correlated negatively with
levels of IL-10. Of the 6 cytokines that remained in the sub-
network, IL-10 had the broadest effects on severity with direct
connections to 6 symptom constructs including SF36 social
function which itself supported a cluster of 10 symptoms.
Within-group changes in IL-10 and IL-1a levels, along with IL-4
and IL-12, aligned with a shift in the relative abundance of CD2+ T
and NK lymphocytes. Similarly changes in IL-2 and IL-5 aligned
with changes in intracellular perforin in cytotoxic T cells
(CD3+CD8+). Both branches correlated with expression of the axon
guidance pathway (Fig. 1), a hub linking 8 of the original 19 GWI
pathways under the broad theme of neuronal development and
migration. Aligning with results from the nString panel, two of
these 8 satellite pathways involved regulation of transcriptional
co-repressor SMRT, and its associated activation for RAR and RXR
heterodimers. The latter were linked to axon guidance via the reg-
ulation of Cdk5 and LPA4 signaling respectively, two pathways
associated with neuronal maturation and migration. Also consis-
tent with this theme were pathway nodes for regulation of lissen-
cephalygene LIS-1and a-synuclein
neuroactive ligand–receptor signaling and visual signal transduc-
tion, the second most over-expressed GWI pathway. This cluster
signaling as wellas
centered about the axon guidance pathway was tethered to cyto-
toxic T cell perforine levels via apoptosis-inducing factor (AIF)
activity, which in turn was linked to SMRT co-repressor activity
and biosynthesis of phenypropanoid, an inhibitor of apoptosis.
The latter was also linked to modulation of PI3K/AKT/mTOR activ-
ity. In parallel with this major hub surrounding axon guidance, a
second minor branch from the CD2+ cell population node consisted
of 2 pathways associated with androgen-mediated regulation of
NF-jB activation. Finally, changes in the relative abundance of
Th17 cells (CD4+CD26+) aligned with increased induction of heart
shock protein (hsp) and the expression of IL-12, an NK cell primer.
In this study we use a standard exercise challenge to stimulate
immune and endocrine signaling in a group of veterans with Gulf
War Illness (GWI), a disease control group of non-veterans with
chronic fatigue syndrome (CFS/ME) and healthy control subjects.
Circulating lymphocytes collected at initial rest, peak effort (VO2
max) and 4 h post-effort were analyzed for gene expression using
both a conventional single-gene approach and a more novel ap-
proach whereby transcript levels were projected onto the regula-
tory circuitry of known pathways. This projection resulted in a
loss of resolution with respect to time however strong group ef-
fects were found with 19 pathways being uniquely expressed in
GWI compared to both healthy controls and CFS/ME, an illness
with similar clinical presentation. Activity was suppressed in
roughly two thirds of these pathways including TNF receptor (or
death receptor) signaling and elements of apoptotic signal trans-
duction. Although signatures obtained in CFS and GWI shared
some commonalities, CFS presented with an almost opposite
suppression of cell cycle, actin remodeling and metabolism. Cal-
cium-mediated T cell activation, IFN signaling and B cell receptor
activation were also suppressed in CFS. Indeed, calcium induced
activation of calcineurin and subsequent calmodulin binding was
a recurring theme among pathways expressed in CFS. Among the
Fig. 1. GWI pathway sub-network interface with symptom severity at peak effort. Association sub-network in GWI at peak effort (T1) during exercise showing only
uninterrupted paths linking symptom nodes with pathway nodes, through layers containing cytokine/hormone and cell abundance/toxicity nodes. Links or edges indicate
pair-wise partial correlations with a probability of false discovery q 6 0.05; positive correlations are shown as green edges, negative correlations are shown as red edges. Bold
green (+) and bold red (?) indicate measured statistically significant increases or decreases in expression respectively, while the same symbols in black font indicate
estimated direction of change based on the consistency constraint of Eq. (8). (For interpretation of the references to colour in this figure legend, the reader is referred to the
web version of this article.)
G. Broderick et al./Brain, Behavior, and Immunity 28 (2013) 159–169
latter we found control of skeletal myogenesis by CaMK, a process
involving activation of P2X5 receptor (Ryten et al., 2002). Calmod-
ulin and calcineurin-mediated signaling has also been implicated
in desensitization of pain receptor TRPV1 (Numazaki et al., 2003).
Involvement of purinergic P2X5, TRPV1 as well as IL-10 immune
signaling has been substantiated in recent work by Light et al.
(2012). However, as this group was very small (n = 7) a more ade-
quate interpretation would be that the CFS subjects studied here
could not be easily reconciled with GWI in terms of pathway
General suppression of apoptotic signaling in GWI was accom-
panied by a significant increase in broad-scale ligand–receptor
interactions supporting neurotransmission as well as increased
activation of master transcriptional regulator NF-jB. Indeed, when
these pathways were cast in the context of symptom severity and
changes in immune cell population, function and signaling, the
pathway mediating axon guidance arose as a major hub aggregat-
ing these effects with a number of signal transduction pathways
including visual signal transduction and neuronal migration. Inter-
estingly retinal axon branching and similar processes are promoted
by neurotrophic factor (BDNF) and its receptor tropomyosin-
related kinase B (Marler et al., 2008). In recent work Barbier
et al. (2009) found alterations in gene expression for both BDNF
and TrkB in stressed animals exposed to PB. Other pathways
include LPA4 mediation of PI3K/AKT pathway (Lee et al., 2008;
Yanagida et al., 2007) and signaling via a-synuclein, a key protein
in Parkinson’s disease (Tani et al., 2010). All support either directly
or indirectly processes related neuronal migration. Although these
measures of pathway activation were expressed in peripheral
blood, similar profiles have been found in PBMC gene expression
of other neuro-immune and neurodegenerative disorders. For
example, a recent analysis of PBMC gene expression in subjects
with Parkinson’s disease reported similar disruption of pathways
associated with axonal guidance, calcium signaling, death signal-
ing and apoptosis (Mutez et al., 2011). Also found were altered
activation of MAP kinase pathway and actin-mediated cytoskele-
ton formation. These results are consistent with our observations
of altered FAS signaling, MAP kinase inactivation of SMRT and their
link with the axon guidance hub pathway. Axon guidance was also
one of the most significant pathways found in a functional enrich-
ment analysis of the top 100 disease-related hub genes expressed
in the PBMC of amyotrophic lateral sclerosis (ALS) patients (Saris
et al., 2009). Differences in FAS-mediated apoptotic signaling were
also found relevant to ALS pathogenesis (Raoul et al., 2006). Of
note, results obtained in blood by Mutez et al. (2011) showed sub-
stantial alignment with gene expression in biopsies collected from
the substantia nigra pars compacta of PD subjects, supporting the
relevance of PBMC profiling in these studies. Similar evidence has
been reported in the study of other neurological conditions includ-
ing Huntington’s and Alzheimer’s disease (Runne et al., 2007; Maes
et al., 2007).
In parallel with neuronal migration pathways we found activa-
tion of master transcriptional regulator nuclear factor-jB (NF-jB)
in relation to prolactin (PRL) signaling. Up-regulation of the NF-
jB/Rel family has been observed in monocytes of men (O’Donovan
et al., 2011) and PBMC of women with PTSD (Pace et al., 2012). In
animal models, selective inhibition of this transcription factor sig-
nificantly reduced the prevalence of extreme PTSD-like response to
stress (Cohen et al., 2011). Prolactin, a stress-responsive pituitary
hormone, has been shown to increase the binding activity of NF-
jB and interferon regulatory factor-1 (IRF-1), promoting TNF-a
and IL-12 secretion (Brand et al., 2004). Secretion of IL-10 was also
induced but only at high levels of PRL. A T-cell cytokine, prolactin is
important in maintaining immune system homeostasis under
stress and preventing glucocorticoid-induced lymphocyte cell
death (Yu-Lee, 2002; Buckley, 2001). This role would be consistent
with the down-regulation of death signaling found in our current
analysis as well as recent evidence pointing to HPA axis dysfunc-
tion in these subjects, potentially involving altered pituitary func-
tion (Golier et al., 2007, 2009). Finally the role of PRL in mediating
neurogenesis (Walker et al., 2012), astrocyte proliferation and
associated IL-1a production (DeVito et al., 1995) may explain at
least in part the elevated PRL levels found in autoimmune illnesses
like systemic lupus erythematosus (SLE) and multiple sclerosis
(MS) (Shelly et al., 2012). More importantly, this role might offer
a tentative mechanistic basis supporting the association of GWI
severity with PRL/NF-jB and neurogenic pathway activation, via
IL-10 and IL-1a in particular.
Also consistent with known actions of prolactin, we found PRL-
mediated NF-jB activity to be associated with GWI severity
through changes in T and NK cell abundance (CD2+) as well as
through IL-10 and IL-12 expression. As mentioned previously, IL-
10 emerged as a hub cytokine that supported changes in virtually
all GWI symptom measures along with IL-1a. Elevated levels of IL-
10 have been associated with increased severity of depressive
symptoms in MS (Heesen et al., 2005). Indeed, a recent gene set
analysis performed on tissue samples from the prefrontal cortex
of patients with major depressive disorder (MDD) suggested up-
regulation of a variety of pro- and anti-inflammatory cytokines
including those identified here as correlates of GWI severity
(Fig. 1), namely IL-1a, IL-10, and IL-12 as well as IL-2 and IL-5
(Shelton et al., 2011; Schmidt et al., 2003). Not surprisingly, NF-
jB signaling plays a key role in mediating the actions of IL-1 and
stress in the onset of depression-like behavior (Koo et al., 2010).
Changes in IL-1R signal transduction, one of only 2 pathways rep-
resenting IL-1 signaling in this dataset, were statistically significant
(q = 0.045) but the magnitude of changes in pathway activation
were less than 10% of levels found in healthy control. In a previous
analysis we found a strong influence of IL-1a on neuroendocrine-
immune signaling in GWI even though levels in plasma did not dif-
fer significantly from control (Broderick et al., 2010). This same
work also pointed to a muted cortisol response in GWI, which is
consistent with over-expression of NF-jB activity (Wolf et al.,
Though several of the markers implicated here are also respon-
sive to exercise, e.g. T and NK cell abundance (Walsh et al., 2011),
IL-1a (Kimura et al., 2001), IL10, prolactin release (Rojas Vega et al.,
2012) as well as neurogenic (Walker et al., 2012) and NF-jB acti-
vation (Kim et al., 2009), it is important to remember that they
were selected on the basis of changes in symptom severity and
pathway activation across subject groups. As a result we found
some aspects of normal exercise response to be over-expressed
or expressed earlier than normal in GWI. For example, in GWI
we found an excessive activation of normal acute IL-1a response
(Kimura et al., 2001) with involvement of IL-10 (Walsh et al.,
2011) and IL-4 (Rosa Neto et al., 2011) at peak effort when these
are typically expressed much later. In addition, there were notable
deviations from normal exercise response. Activity of cyclin-
dependent kinase 5 (Cdk5) (Ghiani et al., 2007) is typically up-reg-
ulated during exercise while a-synuclein activity is down-regu-
lated (Dimatelis et al., 2012). While we recovered the expected
trend in Cdk5 activation, we predicted an opposite trend towards
up-regulation of synuclein signaling in GWI. Similarly, we inferred
an inverse trend in PI3K/AKT pathway activity, which is normally
increased during exercise (Walsh et al., 2011). Though our analysis
suggested decreasing CD2+ relative abundance consistent with
exercise-induced lymphocytosis, we observed a concurrent de-
crease in several facets of apoptotic signaling as well as positive
correlation of IL-2 with increasing severity (Gleeson and Bishop,
This work points to chronic NF-jB activation as a potentially
key component of GWI. Though recent work in animal models sup-
G. Broderick et al./Brain, Behavior, and Immunity 28 (2013) 159–169
ports the involvement of altered cholinergic-immune signaling in
GWI as a result of exposure to acetylcholine (ACh) mediators (PB,
organophosphates, sarin, etc....), we did not observe direct evi-
dence of this here. However it must be emphasized that our anal-
ysis was constrained to a specific set of documented pathways and
was by no means comprehensive. For example, ACh signaling is
known to be a significant modulator of immune function
(Kawashima and Fujii, 2003) and activation of nicotinic acetylcho-
line receptors (nAChR) by SLURP-1 is known to induce of NF-jB
expression (Chernyavsky et al., 2010). Blocking nAChR function
significantly increased basal ACTH and corticosterone in animals
and completely inhibited PRL response to stress (Weidenfeld
et al., 1983). This is consistent with bi-directional ACh signaling
in lymphocytes (Kawashima et al., 2012).
In addition to the limited scope of the pathways examined, we
found the method of Efroni et al. (2007, 2008) to be conservative
since significant up-expression or down-expression of a pathway
requires a corresponding change in the majority of individual
genes supporting that pathway. As a result, it almost certainly
leads to the loss of more subtle signaling mechanisms. Indeed in
this instance, there was a loss of temporal trends across the phases
of exercise. This being said, we might also expect that those path-
ways identified in this analysis were well supported at the level of
gene expression. The method also provides a much-needed frame-
work for the functional interpretation of gene expression results.
Many findings from this work would not have emerged from a con-
ventional comparative analysis, emphasizing the importance of
casting marker expression in the confirmatory context of multiple
layers of biology. This is especially relevant in the study of complex
disorders like GWI. Although the classification performance of
pathway activation levels (AUC < 0.70) did not encourage their
use single biomarkers, this was still sufficiently elevated to moti-
vate future investigation of their use in combination as a bio-
marker panel for diagnosis and possible sub-typing of GWI
(Haley et al., 1997).
In summary, we found statistical association of baseline GWI
symptom burden with increased activation at peak effort in
pathways supporting neuronal development along with down-
regulation of apoptotic signaling. This was accompanied by prolac-
tin-mediated increases in NF-jB activation, a shift in T and NK cell
population and the expression of IL-10 and IL-1a principally. These
changes not only exceeded normal exercise response but also dif-
fered in nature. NF-jB proteins have been detected in many
chronic inflammatory and autoimmune conditions including can-
cer (Ben-Neriah and Karin, 2011). Though the link to ACh signal
disruption was not substantiated directly, disruption of prolactin-
mediated NF-jB activation in concert with altered signaling sup-
porting neurogenesis and apoptosis merit further scrutiny. As this
was an exploratory analysis, confirmatory profiling of gene expres-
sion along candidate pathways identified here, NF-jB in particular,
should be undertaken using a high-resolution method such as PCR.
In addition, pharmacological blockage of NF-jB activity is possible
using anti-TNF, anti-IL-1 as well as IjB kinase (IKK) inhibitors
(Ben-Neriah and Karin, 2011). These might constitute important
avenues for therapy in GWI as well as providing a treatment-based
approach to verifying the illness processes proposed in this
Conceived and designed the experiments: M.A.F., N.G.K., G.B.
Performed the experiments: L.N, M.A.F., N.G.K., Z.B. Analyzed the
data: R.B.H., G.B., S.V., S.E. Contributed reagents/materials/analysis
tools: S.E., L.N., M.A.F. Wrote the paper: G.B., R.B.H., S.E., N.G.K.,
This analysis was funded by grants from the US Department of
Defense CMDRP program (N. G. Klimas, M. A. Fletcher, G. Broder-
ick), Merit Awards from the US Department of Veterans Affairs
(N. G. Klimas, M. A. Fletcher) and from the CFIDS Association of
America (G. Broderick, N. G. Klimas, M. A. Fletcher).
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bbi.2012.11.007.
Amourette, C., Lamproglou, I., Barbier, L., Fauquette, W., Zoppe, A., Viret, R., Diserbo,
M., 2009. Gulf War illness: effects of repeated stress and pyridostigmine
treatment on blood-brain barrier permeability and cholinesterase activity in rat
brain. Behav. Brain Res. 203 (2), 207–214.
Assenov, Y., Ramirez, F., Schelhorn, S.E., Lengauer, T., Albrecht, M., 2008. Computing
topological parameters of biological networks. Bioinformatics 24 (2), 282–284.
Barbier, L., Diserbo, M., Lamproglou, I., Amourette, C., Peinnequin, A., Fauquette, W.,
2009. Repeated stress in combination with pyridostigmine. Part II: changes in
cerebral gene expression. Behav. Brain Res. 197 (2), 292–300.
Ben-Neriah, Y., Karin, M., 2011. Inflammation meets cancer, with NF-jB as the
matchmaker. Nat. Immunol. 12 (8), 715–723.
Bergner, M., Bobbitt, R.A., Carter, W.B., Gilson, B.S., 1981. The Sickness Impact
Profile: development and final revision of a health status measure. Med. Care 19
Bolstad, B.M., Irizarry, R.A., Astrand, M., Speed, T.P., 2003. A comparison of
normalization methods for high density oligonucleotide array data based on
variance and bias. Bioinformatics 19 (2), 185–193.
Brand, J.M., Frohn, C., Cziupka, K., Brockmann, C., Kirchner, H., Luhm, J., 2004.
Prolactin triggers pro-inflammatory immune responses in peripheral immune
cells. Eur. Cytokine Netw. 15 (2), 99–104.
Broderick, G., Fuite, J., Kreitz, A., Vernon, S.D., Klimas, N., Fletcher, M.A., 2010. A
formal analysis of cytokine networks in chronic fatigue syndrome. Brain Behav.
Immun. 24 (7), 1209–1217.
Broderick, G., Kreitz, A., Fuite, J., Fletcher, M.A., Vernon, S.D., Klimas, N., 2011. A pilot
study of immune network remodeling under challenge in Gulf War Illness.
Brain Behav. Immun. 25 (2), 302–313.
Buckley, A.R., 2001. Prolactin, a lymphocyte growth and survival factor. Lupus 10,
Buysse, D.J., Reynolds 3rd, C.F., Monk, T.H., Berman, S.R., Kupfer, D.J., 1989. The
Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and
research. Psychiatry Res. 28 (2), 193–213.
Byrnes, A., Jacks, A., Dahlman-Wright, K., Evengard, B., Wright, F.A., Pedersen, N.L.,
Sullivan, P.F., 2009. Gene expression in peripheral blood leukocytes in
biomarker. PLoS ONE 4 (6), e5805.
Chernyavsky, A.I., Arredondo, J., Galitovskiy, V., Qian, J., Grando, S.A., 2010.
Upregulation of nuclear factor-kappaB expression by SLURP-1 is mediated by
alpha7-nicotinic acetylcholine receptor and involves both ionic events and
activation of protein kinases. Am. J. Physiol. Cell Physiol. 299 (5), C903–C911.
Chi, H., 2012. Regulation and function of mTOR signalling in T cell fate decisions.
Nat. Rev. Immunol. 12 (5), 325–338, Review.
Cohen, H., Kozlovsky, N., Matar, M.A., Zohar, J., Kaplan, Z., 2011. The characteristic
long-term upregulation of hippocampal NF-jB complex in PTSD-like behavioral
stress response is normalized by high-dose corticosterone and pyrrolidine
dithiocarbamate administered immediately after exposure. Neuropsychophar-
macology 36 (11), 2286–2302.
Collins, J.F., Donta, S.T., Engel, C.C., Baseman, J.B., Dever, L.L., Taylor, T., Boardman,
K.D., Martin, S.E., Wiseman, A.L., Feussner, J.R., 2002. The antibiotic treatment
trial of Gulf War Veterans’ Illnesses: issues, design, screening, and baseline
characteristics. Control. Clin. Trials 23 (3), 333–353.
Croft, D., O’Kelly, G., Wu, G., Haw, R., Gillespie, M., Matthews, L., Caudy, M., Garapati,
P., Gopinath, G., Jassal, B., Jupe, S., Kalatskaya, I., Mahajan, S., May, B., Ndegwa,
N., Schmidt, E., Shamovsky, V., Yung, C., Birney, E., Hermjakob, H., D’Eustachio,
P., Stein, L., 2011. Reactome: a database of reactions, pathways and biological
processes. Nucleic Acids Res. 39 (Database issue), D691–D697.
Davidson, J.R., Book, S.W., Colket, J.T., Tupler, L.A., Roth, S., David, D., Hertzberg, M.,
Mellman, T., Beckham, J.C., Smith, R.D., Davison, R.M., Katz, R., Feldman, M.E.,
1997. Assessment of a new self-rating scale for post-traumatic stress disorder.
Psychol. Med. 27 (1), 153–160.
DeVito, W.J., Avakian, C., Stone, S., Okulicz, W.C., Tang, K.T., Shamgochian, M., 1995.
Prolactin induced expression of interleukin-1 alpha, tumor necrosis factor-
alpha, and transforming growth factor-alpha in cultured astrocytes. J. Cell
Biochem. 57 (2), 290–298.
Dimatelis, J.J., Hendricks, S., Hsieh, J.H., Vlok, M.N., Bugarith, K., Daniels, W.M.,
Russell, V.A., 2012. Exercise partially reverses the effect of maternal separation
chronic fatigue:noevidence ofa
G. Broderick et al./Brain, Behavior, and Immunity 28 (2013) 159–169
on hippocampal proteins in 6-hydroxydopamine lesioned rat brain. Exp.
Physiol. [Epub ahead of print].
Efroni, S., Schaefer, C.F., Buetow, K.H., 2007. Identification of key processes
underlying cancer phenotypes using biologic pathway analysis. PLoS ONE 2
Efroni, S., Carmel, L., Schaefer, C.G., Buetow, K.H., 2008. Superposition of
transcriptional behaviors determines gene state. PLoS ONE 3 (8), e2901.
Eisen, S.A., Kang, H.K., Murphy, F.M., Blanchard, M.S., Reda, D.J., Henderson, W.G.,
Toomey, R., Jackson, L.W., Alpern, R., Parks, B.J., Klimas, N., Hall, C., Pak, H.S.,
Hunter, J., Karlinsky, J., Battistone, M.J., Lyons, M.J.Gulf War Study Participating
Investigators, 2005. Gulf War veterans’ health: medical evaluation of a US
cohort. Ann. Int. Med. 142, 881–890.
Emmert-Streib, F., 2007. The chronic fatigue syndrome: a comparative pathway
analysis. J. Comput. Biol. 14 (7), 961–972.
Fukuda, K., Nisenbaum, R., Stewart, G., Thompson, W.W., Robin, L., Washko, R.M.,
Noah, D.L., Barrett, D.H., Randall, B., Herwaldt, B.L., Mawle, A.C., Reeves, W.C.,
1998. Chronic multisymptom illness affecting Air Force veterans of the Gulf
War. JAMA 280, 981–998.
Ghiani, C.A., Ying, Z., de Vellis, J., Gomez-Pinilla, F., 2007. Exercise decreases myelin-
associated glycoprotein expression in the spinal cord and positively modulates
neuronal growth. Glia 55 (9), 966–975.
Gleeson, M., Bishop, N.C., 2005. The T cell and NK cell immune response to exercise.
Ann. Transplant. 10 (4), 43–48.
Golier, J.A., Schmeidler, J., Legge, J., Yehuda, R., 2007. Twenty-four hour plasma
cortisol and adrenocorticotropic hormone in Gulf War Veterans: relationships
to posttraumatic stress disorder and health symptoms. Biol. Psychiatry 62 (10),
Golier, J.A., Schmeidler, J., Yehuda, R., 2009. Pituitary response to metyrapone in
Gulf War veterans: relationship to deployment. PTSD and unexplained health
symptoms. Psychoneuroendocrinology 34 (9), 1338–1345.
Golomb, B.A., 2008. Acetylcholinesterase inhibitors and Gulf War illnesses. Proc.
Natl. Acad. Sci. USA 105 (11), 4295–4300.
Gronwall, D.M., 1977. Paced auditory serial-addition task: a measure of recovery
from concussion. Percept. Mot. Skills 44 (2), 367–373.
Haley, R.W., 1997. Is Gulf War syndrome due to stress? The evidence reexamined.
Am. J. Epidemiol. 146 (9), 695–703.
Haley, R.W., Kurt, T.L., Hom, J., 1997. Is there a Gulf War Syndrome? Searching for
syndromes by factor analysis of symptoms. JAMA 277 (3), 215–222, Erratum in
JAMA 1997, 278(5), 388.
Haley, R.W., Spence, J.S., Carmack, P.S., Gunst, R.F., Schucany, W.R., Petty, F., Devous
Sr., M.D., Bonte, F.J., Trivedi, M.H., 2009. Abnormal brain response to cholinergic
challenge in chronic encephalopathy from the 1991 Gulf War. Psychiatry Res.
171 (3), 207–220.
Heesen, C., Koehler, G., Gross, R., Tessmer, W., Schulz, K.H., Gold, S.M., 2005. Altered
cytokine responses to cognitive stress in multiple sclerosis patients with
fatigue. Mult. Scler. 11 (1), 51–57.
Hurwitz, B.E., Coryell, V.T., Parker, M., Martin, P., Laperriere, A., Klimas, N.G.,
Sfakianakis, G.N., Bilsker, M.S., 2010. Chronic fatigue syndrome: illness severity,
sedentary lifestyle, blood volume and evidence of diminished cardiac function.
Clin. Sci. (Lond). 118 (2), 125–135.
Jonsdottir, I.H., 2000. Special feature for the Olympics: effects of exercise on the
immune system: neuropeptides and their interaction with exercise and
immune function. Immunol. Cell Biol. 78 (5), 562–570, Review.
Kanehisa, M., Goto, S., Furumichi, M., Tanabe, M., Hirakawa, M., 2010. KEGG for
representation and analysis of molecular networks involving diseases and
drugs. Nucleic Acids Res. 38, D355–D360.
Kang, H.K., Natelson, B.H., Mahan, C.M., Lee, K.Y., Murphy, F.M., 2003. Post-
traumatic stress disorder and chronic fatigue syndrome-like illness among Gulf
War veterans: a population-based survey of 30,000 veterans. Am. J. Epidemiol.
Katz, S., Irizarry, R.A., Lin, X., Tripputi, M., Porter, M.W., 2006. A summarization
approach for Affymetrix GeneChip data using a reference training set from a
large, biologically diverse database. BMC Bioinformatics 7, 464.
Kawashima, K., Fujii, T., 2003. The lymphocytic cholinergic system and its
contribution to the regulation of immune activity. Life Sci. 74 (6), 675–696,
Kawashima, K., Fujii, T., Moriwaki, Y., Misawa, H., 2012. Critical roles of
acetylcholine and the muscarinic and nicotinic acetylcholine receptors in the
regulation of immune function. Life Sci. 91 (21–22), 1027–1032.
Kerr, J.R., Petty, R., Burke, B., Gough, J., Fear, D., Sinclair, L.I., Mattey, D.L., Richards,
S.C., Montgomery, J., Baldwin, D.A., Kellam, P., Harrison, T.J., Griffin, G.E., Main, J.,
Enlander, D., Nutt, D.J., Holgate, S.T., 2008a. Gene expression subtypes in
patients with chronic fatigue syndrome/myalgic encephalomyelitis. J. Infect.
Dis. 197 (8), 1171–1184.
Kerr, J.R., Burke, B., Petty, R., Gough, J., Fear, D., Mattey, D.L., Axford, J.S., Dalgleish,
A.G., Nutt, D.J., 2008b. Seven genomic subtypes of chronic fatigue syndrome/
myalgic encephalomyelitis: a detailed analysis of gene networks and clinical
phenotypes. J. Clin. Pathol. 61 (6), 730–739.
Kim, S.Y., Jun, T.W., Lee, Y.S., Na, H.K., Surh, Y.J., Song, W., 2009. Effects of exercise on
cyclooxygenase-2 expression and nuclear factor-kappaB DNA binding in human
peripheral blood mononuclear cells. Ann. NY Acad. Sci. 1171, 464–471.
Kimura,H., Suzui, M.,Nagao,F., Matsumoto,
determination of plasma cytokines by time-resolved fluoroimmunoassay;
effect of bicycle exercise on plasma level of interleukin-1 alpha (IL-1 alpha),
tumor necrosis factor alpha (TNF alpha), and interferon gamma (IFN gamma).
Anal. Sci. 17 (5), 593–597.
K., 2001.Highly sensitive
Koo, J.W., Russo, S.J., Ferguson, D., Nestler, E.J., Duman, R.S., 2010. Nuclear factor-
kappaB is a critical mediator of stress-impaired neurogenesis and depressive
behavior. Proc. Natl. Acad. Sci. USA 107 (6), 2669–2674.
Krupp, L.B., LaRocca, N.G., Muir-Nash, J., Steinberg, A.D., 1989. The fatigue severity
scale. Application to patients with multiple sclerosis and systemic lupus
erythematosus. Arch. Neurol. 46, 1121–1123.
Lee, Z., Cheng, C.T., Zhang, H., Subler, M.A., Wu, J., Mukherjee, A., Windle, J.J., Chen,
C.K., Fang, X., 2008. Role of LPA4/p2y9/GPR23 in negative regulation of cell
motility. Mol. Biol. Cell. 19 (12), 5435–5445.
Light, A.R., Bateman, L., Jo, D., Hughen, R.W., Vanhaitsma, T.A., White, A.T., Light,
K.C., 2012. Gene expression alterations at baseline and following moderate
exercise in patients with Chronic Fatigue Syndrome and Fibromyalgia
Syndrome. J. Int. Med. 271 (1), 64–81.
Liu, P., Aslan, S., Li, X., Buhner, D.M., Spence, J.S., Briggs, R.W., Haley, R.W., Lu, H.,
2011. Perfusion deficit to cholinergic challenge in veterans with Gulf War
Illness. Neurotoxicology 32 (2), 242–246.
Mach, M., Grubbs, R.D., Price, W.A., Nagaoka, M., Dubovicky ´, M., Lucot, J.B., 2008.
Delayed behavioral and endocrine effects of sarin and stress exposure in mice. J.
Appl. Toxicol. 28 (2), 132–139.
Maes, O.C., Xu, S., Yu, B., Chertkow, H.M., Wang, E., Schipper, H.M., 2007.
Transcriptional profiling of Alzheimer blood mononuclear cells by microarray.
Neurobiol. Aging 28 (12), 1795–1809.
Magwene, P.M., Kim, J., 2004. Estimating genomic coexpression networks using
first-order conditional independence. Genome Biol. 5 (12), R100.
Maher, K.J., Klimas, N.G., Hurwitz, B., Schiff, R., Fletcher, M.A., 2002. Quantitative
fluorescence measures for determination of intracellular perforin content. Clin.
Diagn. Lab. Immunol. 9 (6), 1248–1252.
Marler,K.J., Becker-Barroso,E., Martínez,
Poopalasundaram, S., Hindges, R., Soriano, E., Comella, J., Drescher, U., 2008. A
TrkB/EphrinA interaction controls retinal axon branching and synaptogenesis. J.
Neurosci. 28 (48), 12700–12712.
Matthews, L., Gopinath, G., Gillespie, M., Caudy, M., Croft, D., de Bono, B., Garapati,
P., Hemish, J., Hermjakob, H., Jassal, B., Kanapin, A., Lewis, S., Mahajan, S., May,
B., Schmidt, E., Vastrik, I., Wu, G., Birney, E., Stein, L., D’Eustachio, P., 2009.
Reactome knowledgebase of human biological pathways and processes. Nucleic
Acids Res. 37, D619–D622.
McArdle, W.D., Katch, F.I., Katch, V.L., 2007. Exercise Physiology: Energy, Nutrition,
and Human Performance. Lippincott Williams & Wilkins, London, p. 954.
Mi, H., Dong, Q., Muruganujan, A., Gaudet, P., Lewis, S., Thomas, P.D., 2010.
PANTHER version 7: improved phylogenetic trees, orthologs and collaboration
with the Gene Ontology Consortium. Nucleic Acids Res. 38 (Database issue),
Mutez, E., Larvor, L., Leprêtre, F., Mouroux, V., Hamalek, D., Kerckaert, J.P., Pérez-Tur,
J., Waucquier, N., Vanbesien-Mailliot, C., Duflot, A., Devos, D., Defebvre, L.,
Kreisler, A., Frigard, B., Destée, A., Chartier-Harlin, M.C., 2011. Transcriptional
profile of Parkinson blood mononuclear cells with LRRK2 mutation. Neurobiol.
Aging 32 (10), 1839–1848.
Numazaki, M., Tominaga, T., Takeuchi, K., Murayama, N., Toyooka, H., Tominaga, M.,
calmodulin. Proc. Natl. Acad. Sci. U.S.A. (United States) 100 (13), 8002–
O’Donovan, A., Sun, B., Cole, S., Rempel, H., Lenoci, M., Pulliam, L., Neylan, T., 2011.
Transcriptional control of monocyte gene expression in post-traumatic stress
disorder. Dis. Markers 30 (2–3), 123–132.
Pace, T.W., Wingenfeld, K., Schmidt, I., Meinlschmidt, G., Hellhammer, D.H., Heim,
C.M., 2012. Increased peripheral NF-jB pathway activity in women with
childhood abuse-related posttraumatic stress disorder. Brain Behav Immun 26
Raoul, C., Buhler, E., Sadeghi, C., Jacquier, A., Aebischer, P., Pettmann, B., Henderson,
C.E., Haase, G., 2006. Chronic activation in presymptomatic amyotrophic lateral
sclerosis (ALS) mice of a feedback loop involving Fas, Daxx, and FasL. Proc. Natl.
Acad. Sci. USA 103 (15), 6007–6012.
Reeves, W.C., Lloyd, A., Vernon, S.D., Klimas, N., Jason, L.A., Bleijenberg, G., Evengard,
B., White, P.D., Nisenbaum, R., Unger, E.R.International Chronic Fatigue
Syndrome Study Group, 2003. Identification of ambiguities in the 1994
chronic fatigue syndrome research case definition and recommendations for
resolution. BMC Health Serv. Res. 3 (1), 25.
Rojas Vega, S., Hollmann, W., Strüder, H.K., 2012. Influences of exercise and training
on the circulating concentration of prolactin in humans. J. Neuroendocrinol. 24
Rosa Neto, J.C., Lira, F.S., Zanchi, N.E., Oyama, L.M., Pimentel, G.D., Santos, R.V.,
Seelaender, M., Oller do Nascimento, C.M., 2011. Acute exhaustive exercise
regulates IL-2, IL-4 and MyoD in skeletal muscle but not adipose tissue in rats.
Lipids Health Dis. 10, 97.
Runne, H., Kuhn, A., Wild, E.J., Pratyaksha, W., Kristiansen, M., Isaacs, J.D., Régulier,
E., Delorenzi, M., Tabrizi, S.J., Luthi-Carter, R., 2007. Analysis of potential
transcriptomic biomarkers for Huntington’s disease in peripheral blood. Proc.
Natl. Acad. Sci. USA 104 (36), 14424–14429.
Ryten, M., Dunn, P.M., Neary, J.T., Burnstock, G., 2002. ATP regulates the
differentiation of mammalian skeletal muscle by activation of a P2X5
receptor on satellite cells. J. Cell Biol. 158 (2), 345–355.
Saris, C.G., Horvath, S., van Vught, P.W., van Es, M.A., Blauw, H.M., Fuller, T.F.,
Langfelder, P., DeYoung, J., Wokke, J.H., Veldink, J.H., van den Berg, L.H.,
Ophoff, R.A., 2009. Weighted gene co-expression network analysis of the
peripheral blood from Amyotrophic Lateral Sclerosis patients. BMC Genomics
G. Broderick et al./Brain, Behavior, and Immunity 28 (2013) 159–169
Schaefer, C.F., Anthony, K., Krupa, S., Buchoff, J., Day, M., Hannay, T., Buetow, K.H., Download full-text
2008. PID: the Pathway Interaction Database. Nucleic Acids Res. 37 (Database
Schmidt, E.D., Aguilera, G., Binnekade, R., Tilders, F.J., 2003. Single administration of
interleukin-1 increased corticotropin releasing hormone and corticotropin
releasing hormone-receptor mRNA in the hypothalamic paraventricular
nucleus which paralleled long-lasting (weeks) sensitization to emotional
stressors. Neuroscience 116 (1), 275–283.
Shelly, S., Boaz, M., Orbach, H., 2012. Prolactin and autoimmunity. Autoimmun. Rev.
11 (6–7), A465–A470.
Shelton, R.C., Claiborne, J., Sidoryk-Wegrzynowicz, M., Reddy, R., Aschner, M., Lewis,
D.A., Mirnics, K., 2011. Altered expression of genes involved in inflammation and
apoptosis in frontal cortex in major depression. Mol. Psychiatry 16 (7), 751–762.
Shewale, S.V., Anstadt, M.P., Horenziak, M., Izu, B., Morgan, E.E., Lucot, J.B., Morris,
M., 2012. Sarin causes autonomic imbalance and cardiomyopathy: an important
issue for military and civilian health. J. Cardiovasc. Pharmacol. 60 (1), 76–87.
Smets, E.M., Garssen, B., Bonke, B., De Haes, J.C., 1995. The Multidimensional Fatigue
Inventory (MFI) psychometric qualities of an instrument to assess fatigue. J.
Psychosom. Res. 39 (3), 315–325.
Smoot, M., Ono, K., Ruscheinski, J., Wang, P.-L., Ideker, T., 2011. Cytoscape 2.8: new
features for data integration and network visualization. Bioinformatics 27 (3),
Storey, J.D., 2002. A direct approach to false discovery rates. J. Royal Stat. Soc. 64 (3),
Sun, Z., Asmann, Y.W., Kalari, K.R., Bot, B., Eckel-Passow, J.E., Baker, T.R., Carr, J.M.,
Khrebtukova, I., Luo, S., Zhang, L., Schroth, G.P., Perez, E.A., Thompson, E.A.,
2011. Integrated analysis of gene expression, CpG island methylation, and gene
copy number in breast cancer cells by deep sequencing. PLoS ONE 6 (2), e17490.
Tani, M., Hayakawa, H., Yasuda, T., Nihira, T., Hattori, N., Mizuno, Y., Mochizuki, H.,
2010. Ectopic expression of a-synuclein affects the migration of neural stem
cells in mouse subventricular zone. J. Neurochem. 115 (4), 854–863.
Torres-Altoro, M.I., Mathur, B.N., Drerup, J.M., Thomas, R., Lovinger, D.M.,
O’Callaghan, J.P., Bibb, J.A., 2011. Organophosphates dysregulate dopamine
signaling, glutamatergic neurotransmission, and induce neuronal injury
markers in striatum. J. Neurochem. 119 (2), 303–313.
Waickman, A.T., Powell, J.D., 2012. mTOR, metabolism, and the regulation of T-cell
differentiation and function. Immunol. Rev. 249 (1), 43–58.
Walker, T.L., Vukovic, J., Koudijs, M.M., Blackmore, D.G., Mackay, E.W., Sykes, A.M.,
Overall, R.W., Hamlin, A.S., Bartlett, P.F., 2012. Prolactin stimulates precursor
cells in the adult mouse hippocampus. PLoS ONE 7 (9), e44371.
Walsh, N.P., Gleeson, M., Shephard, R.J., Gleeson, M., Woods, J.A., Bishop, N.C.,
Fleshner, M., Green, C., Pedersen, B.K., Hoffman-Goetz, L., Rogers, C.J., Northoff,
H., Abbasi, A., Simon, P., 2011. Position statement. Part one: immune function
and exercise. Exerc. Immunol. Rev. 17, 6–63, Review.
Ware Jr., J.E., Sherbourne, C.D., 1992. The MOS 36-item short-form health survey
(SF-36). I. Conceptual framework and item selection. Med. Care. 30 (6), 473–
Weidenfeld, J., Siegel, R., Conforti, N., Mizrachi, R., Brenner, T., 1983. Effect of
intracerebroventricular injection of nicotinic acetylcholine receptor antibodies
on ACTH, corticosterone and prolactin secretion in the male rat. Brain Res. 265
Whistler, T., Fletcher, M.A., Lonergan, W., Zeng, X.R., Lin, J.M., Laperriere, A., Vernon,
S.D., Klimas, N.G., 2009. Impaired immune function in Gulf War Illness. BMC
Med. Genomics 2, 12.
Wolf, J.M., Rohleder, N., Bierhaus, A., Nawroth, P.P., Kirschbaum, C., 2009.
Determinants of the NF-kappaB response to acute psychosocial stress in
humans. Brain Behav. Immun. 23 (6), 742–749.
Wolfe, J., Proctor, S.P., Davis, J.D., Borgos, M.S., Friedman, M.J., 1998. Health
symptoms reported by Persian Gulf War veterans two years after return. Am. J.
Ind. Med. 33 (2), 104–113.
Wu, Z., Irizarry, R.A., 2005. Stochastic models inspired by hybridization theory for
short oligonucleotide arrays. Proc. RECOMB 2004. J. Comput. Biol. 12 (6), 882–
Wu, Z., Irizarry, R.A., Gentleman, R., Murillo, F.M., Spencer, F., 2004. A Model Based
Background Adjustment for Oligonucleotide Expression Arrays. J. Am. Stat.
Assoc. 99 (468), 909–917.
Yanagida, K., Ishii, S., Hamano, F., Noguchi, K., Shimizu, T., 2007. LPA4/p2y9/GPR23
mediates rho-dependent morphological changes in a rat neuronal cell line. J.
Biol. Chem. 282 (8), 5814–5824.
Yu-Lee, L.Y., 2002. Prolactin modulation of immune and inflammatory responses.
Recent Prog. Horm. Res. 57, 435–455.
Zhang, Q., Zhou, X.D., Denny, T., Ottenweller, J.E., Lange, G., LaManca, J.J., Lavietes,
M.H., Pollet, C., Gause, W.C., Natelson, B.H., 1999. Changes in immune
parameters seen in Gulf War veterans but not in civilians with chronic
fatigue syndrome. Clin. Diagn. Lab. Immunol. 6 (1), 6–13.
Zhu, H., O’Brien, J.J., O’Callaghan, J.P., Miller, D.B., Zhang, Q., Rana, M., Tsui, T., Peng,
Y., Tomesch, J., Hendrick, J.P., Wennogle, L.P., Snyder, G.L., 2010. Nerve agent
exposure elicits site-specific changes in protein phosphorylation in mouse
brain. Brain Res. 1342, 11–23.
G. Broderick et al./Brain, Behavior, and Immunity 28 (2013) 159–169