Age-related changes in neural volume and microstructure associated with interleukin-6 are ameliorated by a calorie-restricted diet in old rhesus monkeys.

Abstract

Systemic levels of proinflammatory cytokines such as interleukin-6 (IL-6) increase in old age and may contribute to neural atrophy in humans. We investigated IL-6 associations with age in T1-weighted segments and microstructural diffusion indices using MRI in aged rhesus monkeys (Macaca mulatta). Further, we determined if long-term 30% calorie restriction (CR) reduced IL-6 and attenuated its association with lower tissue volume and density. Voxel-based morphometry (VBM) and diffusion-weighted voxelwise analyses were conducted. IL-6 was associated with less global gray and white matter (GM and WM), as well as smaller parietal and temporal GM volumes. Lower fractional anisotropy (FA) was associated with higher IL-6 levels along the corpus callosum and various cortical and subcortical tracts. Higher IL-6 concentrations across subjects were also associated with increased mean diffusivity (MD) throughout many brain regions, particularly in corpus callosum, cingulum, and parietal, frontal, and prefrontal areas. CR monkeys had significantly lower IL-6 and less associated atrophy. An IL-6xCR interaction across modalities also indicated that CR mitigated IL-6 related changes in several brain regions compared to controls. Peripheral IL-6 levels were correlated with atrophy in regions sensitive to aging, and this relationship was decreased by CR.

Full text

Page 1

Age-related changes in neural volume and microstructure associated with
interleukin-6 are ameliorated by a calorie-restricted diet in old rhesus monkeys
A.A. Willetteb, B.B. Bendlina,d, D.G. McLarena,c,d, E. Canua, E.K. Kastmana,d, K.J. Kosmatkad, G. Xua,
A.S. Fieldf, A.L. Alexandere, R.J. Colmang, R.H. Weindrucha,d,f, C.L. Coeb, S.C. Johnsona,d,g,⁎
aGeriatric Research Education and Clinical Center, Wm. S. Middleton Memorial Veterans Hospital, Madison, WI 53705, USA
bHarlow Primate Laboratory, Department of Psychology, Madison, WI 53715, USA
cNeuroscience Training Program, University of Wisconsin, Madison, WI 53706, USA
dDepartment of Medicine, University of Wisconsin, Madison, WI 53705, USA
eWaisman Imaging Center, University of Wisconsin, Madison, WI 53705, USA
fDepartment of Radiology, University of Wisconsin, Madison, WI 53792, USA
gWisconsin National Primate Research Center, Madison, WI 53715, USA
a b s t r a c t a r t i c l ei n f o
Article history:
Received 28 October 2009
Revised 2 March 2010
Accepted 4 March 2010
Available online 15 March 2010
Keywords:
Monkey
Voxel-based morphometry
Calorie restriction
Aging
Interleukin-6
Atrophy
Systemic levels of proinflammatory cytokines such as interleukin-6 (IL-6) increase in old age and may
contribute to neural atrophy in humans. We investigated IL-6 associations with age in T1-weighted segments
and microstructural diffusion indices using MRI in aged rhesus monkeys (Macaca mulatta). Further, we
determined if long-term 30% calorie restriction (CR) reduced IL-6 and attenuated its association with lower
tissue volume and density. Voxel-based morphometry (VBM) and diffusion-weighted voxelwise analyses
were conducted. IL-6 was associated with less global gray and white matter (GM and WM), as well as smaller
parietal and temporal GM volumes. Lower fractional anisotropy (FA) was associated with higher IL-6 levels
along the corpus callosum and various cortical and subcortical tracts. Higher IL-6 concentrations across
subjects were also associated with increased mean diffusivity (MD) throughout many brain regions,
particularly in corpus callosum, cingulum, and parietal, frontal, and prefrontal areas. CR monkeys had
significantly lower IL-6 and less associated atrophy. An IL-6×CR interaction across modalities also indicated
that CR mitigated IL-6 related changes in several brain regions compared to controls. Peripheral IL-6 levels
were correlated with atrophy in regions sensitive to aging, and this relationship was decreased by CR.
Published by Elsevier Inc.
Introduction
Peripheral and central production of proinflammatory cytokines
(PICs), including interleukin-6 (IL-6), upregulate oxidative stress in
brain and increases age-related neural atrophy and risk for neurode-
generative pathologies (Ershler, 1993; Kagiwada et al., 2004;
Baranowska-Bik et al., 2008; Qin et al., 2008; Chen et al., 2008;
Baune et al., 2009). IL-6 concentrations in sera and spleen typically
increase with age (Ferrucci et al., 1999; Ershler and Keller, 2000;
Bruunsgaard et al., 2001) and age-related chronic inflammatory states
(Brouqui et al. 1994; Devaux et al. 1997; De Keyser et al. 1998), which
may prime microglia to overproduce PICs and contribute to neuronal
degradation (Godbout et al., 2005; Deng et al., 2006; Cunningham
et al., 2007; Kumagai et al., 2007; Perry et al., 2007; Teeling and Perry,
2009). In humans, peripheral IL-6 is related to lower global and
regional gray and white matter volume (GM, WM; Hauss-Wegrzyniak
et al., 2000; Jefferson et al., 2007; Marsland et al., 2008). Aging is
similarly related to diffuse regional atrophy (Sowell et al., 2003), and
therefore, it was of interest to further investigate the relationship
between neural indices and systemic IL-6 using a whole-brain voxel-
based approach in an aged animal model, the rhesus monkey (Macaca
mulatta), that undergoes some of the age-related brain atrophy seen
in humans (Roth et al., 2004).
IL-6 is uniquely suited for conducting these analyses and
translating results to humans. The protein structure of IL-6 shares
greater than 98% homology between monkeys and humans, and there
is a strong relationship between peripheral and central IL-6 levels in
rhesus monkeys (Reyes and Coe, 1996). In addition, because IL-6 is
produced by many types of cells in the body, it is a stable biomarker
of many physiological systems (Rao et al., 1994; Willette et al., 2007).
By contrast, other PICs such as interleukin-1 and tumor necrosis
factor-alpha are quite low in the blood stream in the absence of
infection. IL-6 is also more readily measured in the healthy individual
and is widely used as a biomarker for chronic, low-grade systemic
inflammation typically associated with aging (Bruunsgaard, 2002)
and many disease states (Fonseca et al., 2009).
Neural atrophy related to IL-6 levels may be reduced by a
calorically restricted diet (CR), which downregulates mRNA and
NeuroImage 51 (2010) 987–994
⁎ Corresponding author. Geriatric Research Education and Clinical Center, D-4225
Veterans Administration Hospital, 2500 Overlook Terrace, Madison, WI 53705, USA.
E-mail address: scj@medicine.wisc.edu (S.C. Johnson).
1053-8119/$ – see front matter. Published by Elsevier Inc.
doi:10.1016/j.neuroimage.2010.03.015
Contents lists available at ScienceDirect
NeuroImage
journal homepage: www.elsevier.com/locate/ynimg

Page 2

protein expression of PICs (Spaulding et al., 1997; Arvidsson et al.,
2004; You et al., 2007). CR also decreases Alzheimer's disease-like
neural sequelae in animal models (Qin et al., 2006; Halagappa et al.,
2007), and can lower proinflammatory or oxidative stress factors in
aged murine neocortex (Lee et al., 2000) and hippocampus (Wu et al.,
2008). CR-induced reductions in visceral adiposity, body weight,
and inflammatory pathologies may mediate decreases in peripheral
IL-6 and were determined as possible mediating health factors
(Bruunsgaard, 2002; Bulcao et al., 2006; Colman et al., 2009). Finally,
to confirm that IL-6 is a stable biomarker over time (Rao et al., 1994),
this protein was assessed twice: when the monkey reached 20 years
of age and later at the time of the scan.
This cross-sectional study used animals from a longitudinal CR
project initiated in 1989 (Ramsey et al., 2000). Voxelwise analyses
of segmented T1-weighted and diffusion tensor imaging (DTI) data
were performed. We hypothesized that higher IL-6 levels acquired
proximal to the time of the MRI scan would be associated with
(1) decreased global GM and WM (Pang et al., 2003; Jefferson et al.,
2007) and (2) decreased regional volumes, as well as lower fractional
anisotropy (FA) and increased mean diffusivity (MD), in diffuse areas
and tracts sensitive to age-related atrophy including cortical and
intrahemispheric parietal areas, superior temporal sulcus, corpus
callosum, thalamus, the dorsal convexity post central sulcus, cingulate
cortex, insula, and dorsal or medial portions of prefrontal cortex (PFC;
Hauss-Wegrzyniak et al., 2000; Good et al., 2001; Sowell et al., 2003;
Pfefferbaum et al., 2005; Sullivan et al., 2006; Kochunov et al., 2007;
Pagani et al., 2008; Alexander et al., 2008; Marsland et al., 2008). FA is
typically used to examine intravoxel tract coherence sensitive to
changes in myelin and axons, while MD can broadly reflect changes
in WM and GM tissue density. CR was expected to lower IL-6 and
reduce associated neural atrophy. Further, Colman et al. (2009)
recently showed that CR interacts with age to protect against atrophy.
A complementary set of interaction analyses was thus conducted to
assess the extent to which CR monkeys showed less atrophy per
picogram-per-milliliter (pg/mL) increase in IL-6.
Materials and methods
Subjects
Forty-five rhesus monkeys between 19 and 31 years of age were
utilized from a longitudinal CR project at the Wisconsin National
Primate Research Center (WNPRC). Eighteen animals were fed a
normal diet (mean±SD age=23.84±2.79 years; 12 females and 6
males), while 27 CR subjects had been on a moderately restricted diet
(30%reductionofintake)forapproximately12to17 years(mean±SD
age=24.32±2.77 years; 15 females and 12 males). Details of the CR
manipulation have been described previously (Kemnitz et al., 1993;
Ramsey et al., 2000). Animals were maintained at 21 °C, humidity of
50–65%, at 12 h/12 h light–dark cycle. Food was present for 6–8 hours
per day; water was available ad libitum. Animals were individually
housed and had continuous visual and auditory contact with other
monkeys in the room. Research and care staff regularly inspected and
interacted with the cohort approximately three times per day.
Manipulanda and other environmental enrichment were provided to
alleviate boredom. Scan acquisition, image processing, and IL-6 assays
wereconductedbypersonnelblindtodietarycondition.Thestudywas
approved by the Institutional Animal Care and Use Committee.
Interleukin-6 assessment
Blood was collected at two time points: (1) within 6 months of the
neuroimaging scan date to determine basal levels of circulating IL-6
and (2) when each animal reached 20 years of age. IL-6 was assayed
using a commercial enzyme-linked immunosorbent assay (ELISA;
R&D Systems, Minneapolis, MN) using methods similar to those
described previously (Willette et al., 2007).
Body mass, visceral adiposity, and morbidity
Body weight (kg) was recorded within 48 hours prior to the MRI
scan. Central adipose tissue in the peritoneum was assessed by MRI
using the following procedure. First, a region of interest (ROI) was
traced around the viscera of the thoracic cavity subjacent to
subcutaneous adipose tissue. Subject-specific intensity thresholding
was then used to isolate adipose tissue, followed by summing all
voxels to obtain total central adiposity. It has been reported that body
mass index (BMI) does not appear to influence the relationship of IL-6
and neural atrophy (Marsland et al., 2008). Thus, for the purposes of
this report, visceral adiposity was evaluated as a correlate of IL-6
values. Finally, the presence of veterinarian-diagnosed inflammatory
or illness conditions near the time of scan was considered because
such pathologies might markedly raise PICs (Brouqui et al. 1994;
Devaux et al. 1997; De Keyser et al. 1998). The small sample size for
glucoregulatory pathologies such as diabetes precluded statistical
analysis (controls, n=4; CR, n=0). Inflammatory ailments included
arthritis (n=3), peritonitis (n=1), endometriosis (n=3), adenocar-
cinoma (n=4), gastric bloat (n=1), and bloat comorbid with
endometriosis (n=1). No acute or severe clinical illnesses due to
infectious pathogens were noted near the time of sampling or scan.
Monkeys were thus grouped into a binary variable as clinically
healthy or having an inflammatory condition that could potentially
affect IL-6 levels.
Image acquisition
Animalsweresedatedandimagesacquiredaspreviouslydescribed
on a General Electric 3.0 T Signa MR unit (GE Medical Systems,
Milwaukee, WI, USA) using a quadrature Tx/Rx volume coil 18 cm in
diameter (McLaren et al., 2009). Briefly, monkeys were maintained
immobile in an optimal plane of anesthesia. A three-dimensional
head-only coronal T1-weighted inversion recovery-prepped spoiled
gradient echo (IR-prepped SPGR) sequence was collected with the
following parameters: repetition time (TR)=8.772 ms, echo time
(TE)=1.876 ms, inversion time (TI)=600 ms, flip angle=10°,
number of excitations (NEX)=2, matrix=256×224, field of
view=160 mm. One hundred twenty-four coronal slices were
acquired with a thickness of 0.7 mm and no gap resulting in near
isotropic 0.625×0.625×0.7 mm voxels. Head-only DTI data were
acquired using a single-shot, spin-echo, diffusion-weighted echo-
planar imaging (DW-EPI) sequence with diffusion gradients in 12
optimaldirectionsandwasemployedusingminimumenergycriterion
(Hasan et al., 2001). A diffusion weighting of b=816 s/mm2was
used. Other imaging parameters for this sequence were axial
acquisition, TR/TE=10,000/77.2 ms, flip angle=90°, NEX=6, field
of view=160 mm, in-plane matrix=120×120, slice thick-
ness=2.5 mm, slice gap=2.5 mm, and 124 slices. Image distortion in
the DW-EPI data was minimized by a higher-order shimming protocol
that was run before the DTI scan. Voxel size was resampled to
0.5×0.5×0.5 mm for both scan sequences. Finally, for the torso, a
three-dimensional coronal T1-weighted fast gradient echo (FGRE)
sequence was acquired during the same scan session. Parameters
were TR/TE=15.7/1.404 ms, 27 coronal sections, flip angle=20°, slice
thickness=4 mm, slice gap=6 mm, NEX=2, matrix=256×128, field
of view=200 mm, and 0.781×0.781×6 mm voxels.
Head-only MRI preprocessing
T1-weighted images were manually deskulled, rotated, normal-
ized to the 112RM-SL monkey atlas, and segmented into GM and WM
probability maps in SPM5 as previously described (McLaren et al.,
988
A.A. Willette et al. / NeuroImage 51 (2010) 987–994

Page 3

2009, 2010). Subject data were normalized to a common space for
VBM analyses through Diffeomorphic Anatomical Registration using
Exponentiated Lie Algebra (DARTEL; Ashburner, 2007; McLaren et al.,
2010). These warped and modulated GM and WM images were thus
moved into the same space, coordinates, and orientation as the D99-
SL brain (Saleem et al., 2002). A 4mm FWHM Gaussian kernel was
applied to smooth the images for analysis. An absolute threshold of
0.1wasappliedfor GMorWM volumeto restrictanalysestothetissue
class of interest.
The DTI data were processed as follows: image distortions caused
by eddy currents were corrected using FSL's FDT Diffusion Toolbox
(http://www.fmrib.ox.ac.uk/fsl/fdt/index.htm). Three-dimensional
maps of the diffusion tensor and derived measures were calculated
using DTIFIT in FSL, including FA, MD, axial diffusivity (λ1), and radial
diffusivity ([λ2+λ3]/2]). To bring DTI maps into the 112RM-SL atlas
for voxelwise analysis, each subject's non–diffusion-weighted image
(B0map) was coregistered and normalized to a T2-weighted template
through the combination of two transforms, where T2 acquisition has
been described elsewhere (McLaren et al., 2009). The normalized
images were smoothed using a 4mm FWHM isotropic Gaussian
kernel. FA analyses were restricted to WM by using a binary prior
probability map to discern changes in major fiber tracts; an absolute
threshold of 0.1 was applied for each monkey to reduce the influence
ofmorphologyonresults.MDmapsweremaskedusingabinaryversion
of the 112RM-SL rhesus atlas because MD is not overtly biased toward
GMorWMtissueclassesrelativetoFA.MDresultsdidnotchangewhen
binary GM and WM prior probability maps were separately used as
masks and the resulting MD result maps were combined.
Imaging artifacts
Images were inspected by experts (A.S.F. and G.X.) to locate
scanner artifacts that could influence volumetric or microstructural
results. One T1-weighted scan and nine DTI scans were excluded from
analyses due to motion, acquisition, respiration, and other distortion
artifacts.
Statistical analyses and atlases
Due to wide variation in IL-6 levels for each group, a two-tailed
Mann–Whitney ranks test using the Z-statistic was used to assess
group differences in peripheral IL-6 at time of scan, which is reported
as the median and interquartile range (IQR). This nonparametric test
is less sensitive to the influence of very high or low values than a
comparison of means. Two-tailed Pearson's r correlations were used
to evaluate if IL-6 levels were related to body weight, visceral
adiposity, and health status, as well as to TBV, global GM and WM. IL-6
levels at 20 years of age andat timeof scan were also compared. α was
set at .05 for these analyses.
For neural association analyses, covariates included age, gender,
and dietary condition; TBV was additionally incorporated when
testing for regional volume associations. GM or WM probability maps
were entered into regional analyses of volume as the dependent
variable (Ashburner and Friston, 2000). The independent variable for
association analyses was IL-6 collected near the time of scan—
although the average of IL-6 at both time points was used to confirm
the stability of result maps. Of particular interest was if CR monkeys
showed incrementally less atrophy per picogram-per-milliliter-unit
increase in IL-6 relative to controls, suggesting a protective effect.
Thus, an IL-6×CR interaction term was entered into separate general
linear models to test for this relationship; IL-6 and dietary condition
vectors, in addition to standard covariates, were included to remove
their main effects. The t-statistic threshold at the voxel level was set at
Pb0.005 (uncorrected). Analyses were corrected for multiple com-
parisons at the cluster level using Monte Carlo simulations, in the
Analysis of Functional NeuroImages (AFNI) package AlphaSim, to
derive a cluster-corrected P value less than 0.05 (Forman et al., 1995).
This correction required GM volume clusters to contain at least 875
edge-connectedvoxelsto besignificant. ForWMvolumeareas,as well
as FA maps, the necessary cluster size after correction was 854 voxels.
MD analyses examining both GM and WM required a cluster size of at
least 1,106 voxels. Although violation of nonstationarity is a concern
for cluster correction of structural imaging data, the use of a 4mm
FWHM kernel does not appreciably bias the true α when using this
permutational estimate of cluster size (Hayasaka et al., 2004).
To determine if alterations in myelin or axons predominantly
drove FA results (Song et al., 2002), the IL-6 association FA map was
used as an explicit mask for post hoc voxelwise analyses of radial and
axial diffusivity thresholded at a voxel α of 0.05 (uncorrected) over 50
contiguous voxels. Coordinates of significant clusters correspond to
the space of the Saleem–Logothetis atlas and are displayed on the
112RM-SL underlays (Saleem et al., 2002). Major fiber tracts were
designated based on autoradiography conducted in rhesus monkeys
(Schmahmann and Pandya, 2006).
Results
Demographics, pathology, body composition, and peripheral interleukin-6
The subject composition of the two dietary conditions did not
differ by age or sex. CR monkeys had lower basal IL-6 at the scan time
point than did the age- and sex-matched controls [Z(45)=−2.201,
Pb0.05]. There was a median difference of 1.12 pg/mL between
the CR and control groups (median=1.43 pg/mL, IQR=2.05, versus
2.55 pg/mL, IQR=2.43, respectively; Fig. 1). For the healthy monkeys
without pathologies at the time of scan, IL-6 was correlated with early
values obtained at 20 years of age [r(29)=0.47, Pb0.05]. By contrast,
monkeys with some type of inflammatory pathology near time of
scan did not show this association, which also negated the stability
of the IL-6 values over time for all subjects [r(45)=−0.01, NS]. The
two diet conditions had statistically similar IL-6 values at 20 years
of age (Z(45)=−0.754, NS). Volumetric and microstructural result
maps related to IL-6 were comparable whether cytokine data were
used from 20 years of age and at time of scan independently or
averaged together (data not shown).
Fig. 1. The main effect of a calorie-restricted diet on systemic IL-6 levels collected near
the time of scan. The black bar in each column indicates the median of each group. High
or low IL-6 values did not affect the significance of the result. For display purposes, a
scalar transform is used employing the natural log.
989
A.A. Willette et al. / NeuroImage 51 (2010) 987–994

Page 4

The presence of inflammation-related conditions was similar in
both CR (n=8) and controls (n=7) in this subsample from the larger
project. A binary variable representing this marker of health was
correlated with higher IL-6 concentrations near the time of scan
[r(45)=0.35, Pb0.05]butnot at 20 years ofage[r(45)=0.13, NS].The
two monkeys with the highest IL-6 levels at the time of the scan had
been diagnosed with chronic endometriosis. This gynecological
condition is common in older female monkeys; therefore, they were
included in all neural analyses as representative of normative aging in
this species. Central adiposity and total body mass were not correlated
with IL-6 sampled near the time of scan.
Global and regional morphometry
Interleukin-6 and T1-weighted volumetric associations
The TBV of the control and CR monkeys did not differ significantly.
Higher peripheral IL-6 near scanning was moderately correlated with
lower TBV [r=−0.41, Pb0.01]. This relationship appeared stronger for
global GM [r=−0.52, Pb0.001] as compared to WM [r=−0.34,
Pb0.05]. IL-6 was associated with reduced GM volume in parietal and
temporal regions sensitive to age-related atrophy. Several clusters were
foundinassociationcorticesalongtheintraparietalsulcus,medialparietal
cortex, and superior temporal sulcus in lateral, medial, and superior
temporal areas (Figs. 2A–C; Table 1). GM regions typically not associated
with age-related atrophy included primary visual cortex and extrastriate
areas. There was no significant relationship between IL-6 levels and
regional WM volume after correction for multiple comparisons.
Volumetrics: interleukin-6×dietary condition interactions
AnIL-6×CRstatusmodelexaminedifCRmonkeyshadmorevolume
than controls across subjects per picogram-per-milliliter increase in IL-
6, suggesting a protective effect against IL-6 related atrophy. For GM
volume, no clusters were significant after type 1 error correction. For
WM volumes (Figs. 2D–G), CR monkeys with higher IL-6 showed
mitigated atrophy relative to controls within intraparietal areas
along the inferior longitudinal fasciculus [−22, 2, 14; t(38)=3.41;
voxel cluster size=1,636] and a local maxima in medial longitudinal
fasciculus[−17, 4,22; t(38)=3.07];cingulum,uncinatefasciculus, and
other subcortical bundles along the frontal forceps [2, 38, 16; peak
t value=3.36; voxel cluster size=2,525], as well as a local maxima in
uncinate fasciculus in deep WM [−8, 37, 14; peak t value=3.04]; and
extreme capsule and striatal bundle proximal to orbital PFC [14, 32, 14;
peak t value=3.34; voxel cluster size=1,108].
Fractional anisotropy
Interleukin-6 and FA associations
Progressively higherIL-6 across all animalswas related to lowerFA
values along the genu, body, and splenium of the corpus callosum, as
well as posterior cingulum (Fig. 3A; Table 2). Clusters were also found
within thalamic bundles at the levels of the ventral–posterior and
pulvinar nuclei, as well as WM proximal to the somatosensory areas
consistent with the superior longitudinal fasciculus (Figs. 3B and C).
Other associations were noted in arcuate fasciculus along an
intraparietal extent of lateral sulcus, as well as anterior commissure
and uncinate fasciculus. A small anterior portion of this latter tract
showed overlap with the WM volume result map generated by the IL-
6×CR contrast. To determine whether WM volume influenced the FA
result, a peak voxel (2, 44, 16) in this area of overlap was taken. A full
correlation indicated a mild-moderate negative relationship between
FA and IL-6 [r(32)=−0.392, P=0.013], which was not influenced
when partially out WM voxel values [r(29)=−0.388, P=0.016].
FA: interleukin-6 dietary condition interactions
An IL-6×CR Status model was used to test where CR monkeys had
higher FA relative to controls across subjects per picogram-per-
milliliter increase in IL-6, suggesting a protective effect. After
correction, the CR monkeys showed a mitigated relationship with
IL-6 associated atrophy in superior longitudinal fasciculus along the
dorsal convexity anterior to the central sulcus (Figs. 3D and E;
Table 2). Removal of individual IL-6 values at the high or low end of
the range did not undermine the associations or group differences
beyond an expected loss of statistical power. Split-level partial
correlations indicated that the positive association of IL-6 and FA for
CR monkeys was nonsignificant (data not shown).
FA: radial and axial diffusivity post hoc analyses
To determine whether FA results were driven primarily by myelin
or axonal changes, post hoc analyses of radial and axial diffusivity
Fig. 2. Association of IL-6 and IL-6×CR status interaction in GM and WM volume in the
axial plane in arbitrary units (AU). IL-6 was associated with decreased GM primarily
in (A) intraparietal sulcus, (B) temporo-occipital and extrastriate regions, and
(C) superior temporal sulcus. For WM, the interaction indicated that CR monkeys
showed less IL-6 related atrophy per picogram-per-milliliter increase relative to
controls along intraparietal sulcus (D), the frontal forceps (E), and subjacent to orbital
prefrontal cortex (F). A peak voxel proximal to uncinate fasciculus depicts the IL-6×CR
interaction (G). A scalar employing the natural log of IL-6 is used for display purposes.
The ‘warm’ and ‘cool’ color schemes for association and interaction analyses represent t
values. Brains are oriented in neurological space.
990
A.A. Willette et al. / NeuroImage 51 (2010) 987–994

Page 5

maps were conducted within the space of the FA results map.
Increased peripheral IL-6 was related to greater radial diffusion along
the corpus callosum, thalamic bundles, and other areas highly
consistent in cluster size and significance with the FA map
(Supplementary Figure 1). By contrast, higher axial diffusivity was
restricted to part of the splenium and corticocortical fibers adjacent to
the somatosensory areas.
Mean diffusivity
Interleukin-6 and MD associations
Comparable to FA, higher levels of IL-6 at scan were associated
with increased MD of WM throughout an extensive cluster encom-
passing the corpus callosum and cingulum (Fig. 4; Table 3). MD in GM
tissue was also higher in several brain areas. The main cluster ranged
from the global maxima adjoining right hemisphere insula and
somatosensory areas out to local maxima caudally in the splenium
and rostrally to ventral PFC. The caudal extent in part included
occipitosplenial fibers, intraparietal and parietal cortex areas, tempo-
ral regions generally along the superior temporal sulcus, posterior
cingulatecortex,andlongitudinal fasciculiin parietallobe. Association
regions extended anteriorly to dorsal premotor cortex, supplemen-
tary motor area (SMA), pre-SMA, frontal eye field, and then dorsal
PFC. This rostral portion of the cluster consisted of anterior cingulate
cortex; dorsal, infralimbic, orbitofrontal, and ventral PFC, as well as
gyrus rectus; precentral operculum; optic tract; agranular insula; and
the basal forebrain. Medial and lateral extents incorporated striatum;
dysgranular anterior insula and claustrum; uncinate, arcuate, and
longitudinal fasciculi; superior temporal sulcus in medial temporal
lobe; and medial temporal lobe structures including parahippocam-
pus, hippocampus, and amygdala. Another separate cluster included
occipitosplenial fibers and extrastriate cortex.
MD: interleukin-6 dietary condition interactions
An IL-6×CR Status model tested where CR monkeys had lower
MD relative to controls for every picogram-per-milliliter increase in
IL-6 across all subjects. A striking cluster extended approximately
25 mm along the frontal convexity in WM and GM, spanning from
dorsal premotor cortex to dorsal PFC and additionally incorporating
superior longitudinal fasciculus, SMA, pre-SMA, frontal eye field, and
anterior cingulate cortex (Fig. 4; Table 3). This region bilaterally
overlapped with part of the IL-6 main effect. Clusters were also found
Fig. 3. Association of IL-6and IL-6×CRstatus interaction inFA of WM along sagittal slices.
IL-6 was negatively associated with FA in areas such as the corpus callosum (A), thalamic
bundles (B), and superior longitudinal fasciculus (C). CR diet contributed to a protective
effect against IL-6 related atrophy per picogram-per-milliliter increase in superior
longitudinal fasciculus in the dorsal convexity (D). A peak voxel in superior longitudinal
fasciculusillustratestheIL-6×CRinteraction(E).Ascalaremployingthenaturallogisused
for display purposes. The ‘warm’ and ‘cool’ color schemes for association and interaction
analyses represent t values. Brains are oriented in neurological space.
Table 2
Fractional anisotropy of myelinated white matter and IL-6.
LocationCoordinates Peak
t value
Cluster size
(voxels)
xyz
IL-6 association
Splenium of CC
Splenium of CC
Splenium of CC
TP
TP
SLF III
Genu of CC
Genu of CC
Body of CC
AC; UF
UF
UF
8022
22
21
12
14
21
24
18
25
9
14
18
6.22
5.31
3.98
5.33
5.20
4.72
4.31
4.28
3.39
3.77
3.46
2.92
6939
−8
−2
0
4
4
162708
2372
2423
3995
11
22
3
15
26
34
18
21
44
46
7
7
8
14865
12102
−6
IL-6×CR status effect
SLF I
SLF I
−5
−6
29
20
32
36
4.23
3.89
854
Local maxima of larger clusters are noted below the highest maxima. Voxel and cluster
thresholds were Pb0.005 uncorrected and Pb0.05 corrected, respectively. AC = anterior
commissure,CC=corpuscallosum,SLF = superiorlongitudinal fasciculus,UF= uncinate
fasciculus, TP = thalamic peduncle. Positive x-coordinates correspond to the right
hemisphere.
Table 1
Regional gray matter and interleukin-6.
LocationCoordinates Peak
t value
Cluster size
(voxels)
xyz
Ventral intraparietal area (IPS)
Somatosensory areas (IPS)
Medial parietal cortex
Extrastriate cortex (LS)
Medial temporal area (STS)
Extrastriate cortex (IOS)
Primary visual cortex
Lateral temporal lobe (STS)
Ventral intraparietal area (IPS)
144
8
26
34
34
32
24
14
16
4.83
3.15
2.85
4.06
3.85
3.75
4.01
3.79
3.71
4905
6
1
−2
−12
−4
−4
−22
18
15
19
28
13
23
4847
3087
1537
1606
4
−106 30
Local maxima of larger clusters are noted below its highest maxima. Voxel and cluster
thresholds were Pb0.005 uncorrected and Pb0.05 corrected, respectively. IOS =
inferior occipital sulcus, IPS = intraparietal sulcus, STS = superior temporal sulcus,
LS = lunate sulcus. Positive x-coordinates correspond to the right hemisphere.
991
A.A. Willette et al. / NeuroImage 51 (2010) 987–994