TOXICOLOGICAL SCIENCES 85, 963–975 (2005)
doi:10.1093/toxsci/kfi153
Advance Access publication March 23, 2005
The Effects of Early Lead Exposure on the Brains of Adult Rhesus
Monkeys: A Volumetric MRI Study
Robert E. Lasky,*,1 Melissa L. Luck,† Nehal A. Parikh,‡ and Nellie K. Laughlin†
*Center for Clinical Research and Evidence Based Medicine, The University of Texas-Houston Medical School, 6431 Fannin Street, MSB 2.104, Houston,
Texas 77030; †Harlow Center for Biological Psychology, 22 North Charter Street, Madison, Wisconsin 53715; ‡The University of Texas-Houston Medical
School, Department of Pediatrics, 6431 Fannin, Houston, Texas 77030
Received October 22, 2004; accepted March 18, 2005
Little is known about direct effects of exposure to lead on
central nervous system development. We conducted volumetric
MRI studies in three groups of 17-year-old rhesus monkeys: (1)
a group exposed to lead throughout gestation (n ¼ 3), (2) a group
exposed to lead through breast milk from birth to weaning (n ¼ 4),
and (3) a group not exposed to lead (n ¼ 8). All fifteen monkeys
were treated essentially identically since birth with the exception
of lead exposure. The three-dimensional MRI images were
segmented on a computer workstation using pre-tested manual
and semi-automated algorithms to generate brain volumes for
white matter, gray matter, cerebrospinal fluid, and component
brain structures. The three groups differed significantly in the
adjusted (for total brain size) volumes of the right cerebral white
matter and the lateral ventricles. A significant reduction was
noted in right cerebral white matter in prenatally exposed
monkeys as compared to controls ( p ¼ 0.045). A similar reduction
was detected in the white matter of the contralateral hemisphere;
however, this difference did not achieve statistical significance
( p ¼ 0.143). Prenatally exposed monkeys also had larger right
( p ¼ 0.027) and left ( p ¼ 0.040) lateral ventricles. Depending on
the timing of exposure during development, lead may exhibit
differential effects with resultant life-long alterations in brain
architecture.
Key Words: lead; rhesus monkey; segmentation; volumetric
MRI.
Since the early 1990s there has been a steady decline in body
lead burdens in children in the U.S. (Stephenson, 2003). This
encouraging news has been tempered by two recent independent reports that blood lead levels below those previously
considered to have no adverse consequences are associated
with reduced IQ at three and five years of age (Bellinger and
Needleman, 2003; Canfield et al., 2003). It is now hypothesized
that there is no safe body lead burden (Needleman and
Landrigan, 2004); rather, lead produces a continuum of effects,
1
To whom correspondence should be addressed. Fax: (713) 500-0519.
E-mail: [email protected].
from subtle impairments in cognitive function and behavioral
abnormalities at low burdens to encephalopathy and death at
high burdens.
Our understanding of what structures in the brain are
affected by lead is rudimentary. Technological innovations in
imaging have dramatically improved our ability to noninvasively assess brain structure and function (Ashburner and
Friston, 2000; Fischl et al., 2002; Iosifescu et al., 1997; Smith
et al., 2002; Wei et al., 2002). The few neuroimaging reports of
lead exposed adults suggest several vulnerable regions in the
brain. High level exposure is associated with lesions in the
cerebellum, thalamus, putamen, basal ganglia, cerebral cortex,
periventricular white matter, and pons on CT and anatomic
magnetic resonance imaging (MRI). All of the cases reported
in these studies also had significant neurobehavioral symptoms
(Mani et al., 1998; Schroter et al., 1991; Teo et al., 1997; Tuzun
et al., 2002).
The only two published studies concerning the effects of
lead toxicity on the brains of children assessed by anatomic
MR imaging reported no abnormalities despite blood lead
levels ranging from 23 to 65 lg/dl (Trope et al., 1998, 2001).
However, there was reduced N-acetylaspartate (a marker of
neuronal loss) measured by magnetic resonance spectroscopy
(MRS) in the frontal gray matter of these children. Weisskopf
et al. reported similar metabolite abnormalities in the frontal
lobe as well as the hippocampus and midbrain in a set of adult
twins with chronic lead exposure. These patients also had
lesions indicative of microinfarcts on MRI (Weisskopf et al.,
2004). Subtle central nervous system effects of lead may not be
detected by standard neuroimaging evaluation. MRS and other
methods such as quantitative volumetric MRI may prove to be
more informative in identifying vulnerable brain regions and
correlating structural deficits with the neurobehavioral abnormalities associated with lead toxicity.
Animal studies offer unique advantages over human studies
in evaluating lead toxicity including random allocation to
treatment groups and known lead exposure histories of the
study participants. The problems of generalizing the results of
animal studies to humans are lessened when the animal model is
Toxicological Sciences vol. 85 no. 2 Ó The Author 2005. Published by Oxford University Press on behalf of the Society of Toxicology. All rights reserved.
For Permissions, please email: [email protected]
964
LASKY ET AL.
a non-human primate. This is especially true for studies addressing the effects of lead on the brain. There are no comprehensive
methodologic or normative 3-D MRI brain volumetric studies on
non-human primates. Normative MRI volumetric data for the
majority of brain structures in the non-human primate are lacking,
and for the rhesus monkey only volumes for the whole brain,
amygdala, corpus callosum, thalamus, putamen, and caudate
nucleus have been reported (Franklin et al., 2000; Matochik et al.,
2000; Schindler et al., 2002).
Neuropathology in laboratory animals has documented the
adverse effects of lead on microstructural and molecular
pathways in the visual area of the occipital lobe (Reuhl et al.,
1989), cerebral cortex, hippocampus, and cerebellum among
other brain structures (Ma et al., 1997; Reddy et al., 2003;
Slomianka et al., 1989; Struzynska et al., 2001). To date, there
are no published reports of CT, MRI, or volumetric MRI in lead
exposed animals.
The present study is an exploratory analysis of the effects of
lead exposure in utero and during the immediate postnatal
period on regional brain volumes of 17-year-old rhesus
monkeys. Little is known about the persistent effects of
antenatal and postnatal lead exposure on the mature adult
brain. We also describe the reliability of the methods we used
to manually and semi-automatically segment various brain
structures because of the promise of this methodology in
studying the effects of neurotoxins on the brain.
Lead induced effects on the anatomy of the brain are of
concern if there are associated functional deficits. This study
does not address that important association. Rather, it presents
a methodology and preliminary data that may prove useful in
understanding the structural underpinnings of the functional
deficits induced by exposure to lead and other toxins. By
identifying affected brain structures, the associated functional
deficits can be more efficiently targeted and understood.
MATERIALS AND METHODS
Subjects. The study cohort consisted of 15 rhesus monkeys (Macaca
mulatta) born to adult female monkeys in the spring and summer of 1981 at the
University of Wisconsin Harlow Center for Biological Psychology. These
monkeys were part of a larger cohort of monkeys participating in a study of the
effects of early lead exposure on the auditory system. Details concerning these
monkeys, their lead exposure, and lead effects on the auditory system are
described elsewhere (Lasky et al., 1995).
Eight of these monkeys served as controls. Four of the control monkeys were
males, and the other four were females. The other seven monkeys in the study
cohort had been exposed to lead either in utero or postnatally until they were
weaned. Four of the lead exposed monkeys were males (two prenatally and two
postnatally exposed monkeys), and the other three were females (one prenatally
and two postnatally exposed monkeys). An eighth lead exposed monkey had
a MRI scan of compromised quality due to artifact and was not included in this
study because of the inadequate scan.
Lead exposed breeding females were administered lead acetate daily in
their drinking water beginning three months prior to their time mated
pregnancy (8.6 mg/kg body weight per day) and continuing until 5 months
postpartum (9.1 mg/kg body weight per day). Details of the lead exposure
histories of these monkeys are given in Schantz et al. (1986). All newborn
monkeys were cross fostered shortly after birth to recently parturient mothers
to create three groups: a prenatal group exposed to lead from conception to birth
(n ¼ 3), a postnatal group exposed to lead through breast milk from birth to
weaning (n ¼ 4), and a control group whose birth and rearing mothers were
never exposed to lead (n ¼ 8). The study monkeys were placed in separate
cages with their rearing mothers until weaning at six months postpartum.
Weaned monkeys were housed in groups of five until 2 ½ years postpartum
after which they were housed separately. The monkeys were treated identically
with the exception of their lead exposures. The study monkeys were fed a set
diet according to the Harlow Center protocol intended to promote optimal
health. With some individual adjustments, males received 12 pieces of monkey
chow and females 10 pieces daily.
Blood samples from the 15 monkeys in the study sample were obtained by
venipuncture biweekly from birth to one year of age, monthly during year two,
quarterly until 3½ years postpartum, and periodically thereafter. Four biological maternal blood samples were collected at regular intervals from
conception until the birth of their study offspring. Four nursing maternal blood
samples were collected at regular intervals from a few days after birth until the
weaning their study offspring. All blood samples were analyzed for lead by the
Wisconsin State Laboratory of Hygiene by atomic absorption spectroscopy
(Model 503, Perkin-Elmer, Boston, MA) using a modified delves cup technique
(Ediger and Coleman, 1972). Absorbance was read in duplicate at 283.3 nm
with deuterium arc background correction. Average blood lead levels were
calculated for each study monkey from birth to weaning and for their biological
and nursing mothers during the times they were administered lead.
The median maternal blood lead level for the prenatal group during
pregnancy was 62.0 lg/dl (minimum ¼ 42.0, maximum ¼ 81.5 lg/dl). The
median maternal blood lead level for the postnatal group during nursing was
97.8 lg/dl (minimum ¼ 63.0, maximum ¼ 209.5 lg/dl). The median prenatal
offspring blood lead level during nursing was 26.5 lg/dl (minimum ¼ 23.2,
maximum ¼ 40.1 lg/dl). The elevated blood lead levels in the prenatal
offspring during nursing reflect exposure in utero because there was no further
exposure to lead postnatally in these monkeys (i.e., their ‘‘postnatal’’ nursing
mothers were never exposed to lead). The median postnatal offspring blood
lead level during nursing was 55.1 lg/dl (minimum ¼ 45.5, maximum ¼
70.8 lg/dl). The median control offspring blood lead level during nursing was
4.5 lg/dl (minimum ¼ 3.4, maximum ¼ 6.3 lg/dl). Blood lead levels for all
lead exposed infant monkeys declined after weaning and were <10 lg/dl by 2 ½
years postpartum and <5 lg/dl by 4 ½ years of age. All monkeys (mothers and
offspring) were without overt signs of lead toxicity at all times.
The monkeys had been in good health since birth and were in good health at
the time of the MRI scans for this study. The scans were obtained over a sixmonth interval from 4 January through 18 June 1999. The median age of the leadexposed monkeys at the time of the scans was 17.53 years (the range was from
17.48 to 18.14 years). The median age of the control monkeys at the time of the
scans was 17.73 years (the range was from 17.61 to 18.16 years). All care and
testing of the study cohort was approved by the University of WisconsinMadison Animal Care and Use Committee and conformed to the guidelines
established by the National Institutes of Health (NIH publication #86-23, 1985).
Procedures. The monkeys were housed and studied at the University of
Wisconsin-Madison Harlow Center for Biological Psychology. The MRI
scanner was located at the Waisman Center (about 2 km from the Harlow
Center) on the University of Wisconsin-Madison campus. The monkeys were
food deprived the night before they were scanned. Ketamine hydrochloride
(Ketaject Phoenix Scientific, Inc., St. Joseph, MO) was given im (10–15 mg/kg)
to anesthetize the monkeys for transport by the University Primate Center van.
Three members of the research team (R.E.L., M.L.L., and a research assistant)
attended to the monkeys in the van and during the entire procedure. At the
Waisman Center, the condition of the monkeys was evaluated. An iv line was
placed in the monkeys’ saphenous vein and a propofol (Diprivan, Stuart
Pharmaceuticals, Wilmington, DE) drip was initiated with a 1 ml bolus and
a maintenance dose set to effect (no movement of the monkey with stable vital
signs). Heart rate, respiration, and SpO2 were continuously monitored before,
EFFECTS OF LEAD ON THE RHESUS BRAIN
TABLE 1
MRI Acquisition Details
Parameters
TR
TE
Inversion time
Pulse angle
Slice thickness
Slice orientation
# of slices
Voxel
dimensions
Field of view
MATRIX
NEX
Values used
11.4 ms
2.2 ms
400 ms
20°
1.2 mm (Gap ¼ 0)
Coronal
124
0.886 mm3
(width and height ¼ 0.859 mm,
depth ¼ 1.200 mm)
22 3 16.5 cm
IS: 256 mm AP: 256 mm:
LR: (covered head with 124 slices)
2
during, and after the scan (Model 4402, Sensor Devices, Inc., Waukesha, WI).
Pre- and post-scan rectal body temperatures were also monitored. The monkeys
were placed supine in the scanner. Their heads were positioned upright by
a molded foam constraint developed to scan the heads of adult rhesus monkeys.
From start to finish the scans took about 45 min. The monkeys were transported
back to the Harlow Center and placed in an observation cage overnight in
a room adjacent to the larger room housing their home cages. They were
returned to their home cages the morning after their scans. None of the
monkeys experienced any difficulties.
Apparatus. The monkeys were scanned with a GE 1.5 Tesla Oxford Style
Magnet with version 5.7 software (General Electric, Milwaukee, WI). A
standard GE Quad Head Coil was used. A brief Sagittal T1 localizer scan
preceded the Coronal 3D T1 Fast Spoiled Gradient Echo Recall (FSPGR)
sequence that was used for the volumetric measurements. The scanning
parameters of the 3D T1 FSPGR sequence are listed in Table 1. No postprocessing adjustments to reduce movement artifacts and non-uniform intensity
levels were employed.
Volumetric measurements. The volumetric measurements were performed by R.E.L. using ANALYZE software (version 4.0, Biomedical Imaging
Resource, Mayo Clinic, Rochester, MN). R.E.L. was masked to lead exposure
of the study monkeys. Coronal slices were scored sequentially in a posterior to
anterior direction. Each scan was scored two times with at least two months
separating scoring the same scan. A lengthy training procedure preceded
scoring the scans. Primary anatomical references included Paxinos et al.
(2000), Martin and Bowden (1996), and Internet accessible monkey atlases
(BrainInfo, Comparative Mammalian Brain Collections, and Laboratory of
Neuroimaging, UCLA). Local experts were also enlisted to verify the structures
measured. It took about 6 h to completely score a single scan.
The ANALYZE software provides automated and manual (i.e., outlining
structure boundaries by hand) segmentation software tools that were used to
segment anatomically based structures (Regions Of Interest or ROIs). Gray
scale differences alone did not distinguish all the anatomical structures of
interest. Therefore, anatomical landmarks (i.e., the spatial relationships among
anatomical structures) in addition to gray scale differences were used to
manually segment structures that could not be reliably segmented (automatically) by gray scale thresholding alone. The manually identified structures may
combine and blur anatomical distinctions evident with greater resolution and
additional information (e.g., different MR protocols, histological, and biochemical data). Consequently, some structures that were scored reflect common
anatomical classifications, while others do not. An example of the latter would
be structures medial to the cerebrum and superior to the midbrain (we labeled
these structures ‘‘medial gray matter’’). The hippocampus, amygdala, lenticular
965
nucleus, and caudate nucleus could be differentiated by gray scale differences
and anatomical landmarks. Medial to those structures, tissues imaged as
predominantly gray matter included thalamic structures but also basal ganglia
and cerebral structures as well. We could not reliably make those additional
distinctions and did not attempt to do so for this study.
Table 2 identifies the brain structures scored for this study. It also specifies
how those structures were scored. As a result of pre-testing, an invariant
procedure was used to score each T1 scan. Figure 1 exemplifies the scoring at
several steps in the procedure. The ANALYZE software was used to orient the
coronal sections scored so that they were orthogonal to the plane defined by the
axis passing through the anterior and posterior commissures and the interaural
axis. The structures that could not be reliably scored automatically were
manually outlined (using the ANALYZE ‘‘Manual Trace’’ software tool). The
left and right hemispheres and the cerebellum were then manually isolated
(using the ANALYZE ‘‘Auto Trace Limit’’ software tool) for subsequent
automated scoring. Tissues of the head external to the skull were also isolated
so that automated scoring did not include those tissues as brain parenchyma or
CSF. Prior to scoring, the intensity limits distinguishing gray and white matter
and CSF and other opaque tissues were identified from histograms of the voxel
intensities of a randomly selected monkey. Cutoffs to distinguish gray matter
(56–125), white matter (126), and CSF and other tissues (0–55) were selected
from the generated histograms and verified by visual inspection of the images
(i.e., the cutoffs appeared to correctly distinguish the intended structures on the
randomly selected monkey as well as the other monkeys in the dataset). Those
intensity limits were then used to segment (using the ANALYZE ‘‘Auto Trace’’
software tool) the cerebral hemispheres and the cerebellum (both gray and
white matter) as well as the CSF and skull (scored as ‘‘extra parenchymal
volume’’).
Analyses. The study results are presented in four sections. The first
concerns the reliability of making two independent volumetric measurements
by the same person (R.E.L.) on the same scan from the same monkey for the
brain structures identified in this study. We followed Bland and Altman’s (1996)
recommendations and used Intra-Class Correlation (ICC) coefficients, within
standard deviations (wSDs), and repeatability to characterize measurement
reliabilities. The wSDs are the average difference between the two measurements for the entire sample (i.e., the SDs of the repeated measurements).
Repeatability is defined as the difference between two measurements as large as
95% of the differences between repeat measurements. The intra-scorer
reliabilities for all the structures scored are presented. These reliabilities were
calculated for the eight control monkeys.
There may be gender differences in the effects of lead on the brain. In the
second section gender differences were explored by a 2(gender) 3 2(grouplead exposed and controls) analysis of variance.
The third section considers the relationships among the brain structures
measured. A principal components analysis was not possible given the small
sample size. Instead, we calculated Pearson correlation coefficients and partial
correlation coefficients (adjusting for total brain volume) to identify brain
structures correlated with each other. These analyses also determined whether it
was necessary to adjust brain structures by total brain volume because brain
structure volumes reflect overall brain size as well as the relative sizes of each
individual structure.
The final section evaluates lead effects on brain structures. A multivariate
analysis of variance (MANOVA) was not possible because we had more
dependent variables than monkeys, some of those dependent variables (right
and left analogs of the same structure) were highly correlated raising concerns
of multicollinearity among the dependent variables, and the sparseness of the
data made it difficult to test assumptions of multivariate normality and
homogeneity of the covariance matrices. Therefore, we adopted a univariate
analytic approach to identify possible lead effects on the structures we
measured. We calculated one way analyses of variance (ANOVA) to evaluate
the effects of lead exposure (prenatally, postnatally, or no exposure) on total
brain volume. We calculated one way analyses of covariance (ANCOVAs) for
each of the individual brain structures (ROIs) we measured. The independent
variable in these analyses was lead exposure (prenatal, postnatal, or no
966
LASKY ET AL.
TABLE 2
Definitions of the Brain Structures Measured
TABLE 2—Continued
Left lenticular nucleus
Brain structure
Definition
Right amygdala
Total brain
Cerebrum
Right cerebrum
Left cerebrum
Right cerebral white matter
Left cerebral white matter
Right cerebral gray matter
Left cerebral gray matter
Cerebellum
Cerebellum, white matter
Cerebellum, gray matter
Basal ganglia
Right basal ganglia
Left basal ganglia
Right caudate
Left caudate
Right lenticular nucleus
Sum of all structures listed in this Table
except the ventricles and the extra
parenchymal volume.
Sum of the right and left cerebrums.
Sum of the right cerebral gray and white
matter.
Sum of the left cerebral gray and white
matter.
All structures except the cerebrum and extra
parenchymal volume were identified first.
The right and left cerebral hemispheres
were then segregated manually by
bisecting the longitudinal fissure. Cerebral
white matter was then automatically
identified by the intensity limits for white
matter determined from pretesting (the
limits that visually separated white from
gray from cerebral spinal fluid and other
structures).
The same definition as for right cerebral
gray matter.
After identifying the right cerebral gray
matter, the same procedure was used to
identify the right cerebral white matter
except the intensity limits identifying
white matter were used.
The same definition as for right cerebral
white matter.
Sum of the cerebellar gray and white matter.
The cerebellum was manually segregated
from the rest of the brain. Cerebellar
white matter was then automatically
identified by the intensity limits for white
matter determined from pretesting.
After identifying the right cerebellar white
matter the intensity limits identifying
gray matter were used to automatically
identify the cerebellar gray matter.
Sum of the right and left basal ganglia.
Sum of the right caudate, lenticular nucleus,
amygdala, and acumbens.
Sum of the left caudate, lenticular nucleus,
amygdala, and acumbens.
Gray matter surrounding and extending lateral
from the anterior horn of the right lateral
ventricle and inferior to the corpus
callosum.
Gray matter surrounding and extending
lateral from the anterior horn of the left
lateral ventricle and inferior to the corpus
callosum.
Gray matter in the right cerebral hemisphere
medial to the insular cortex and the
adjacent extreme capsule, inferior and
lateral to the right caudate nucleus,
lateral to the internal capsule, and
superior to the hippocampus, auditory
and optic radiations, and the anterior
commissure.
Left amygdala
Right acumbens
Left acumbens
Right hippocampus
Left hippocampus
Diencephalon
Right diencephalon
Left diencephalon
Right medial gray matter
Left medial gray matter
Right hypothalamus
Left hypothalamus
Brainstem
Midbrain
The same definition as for the right
lenticular nucleus.
Proceeding from posterior to anterior
coronal slices, the right amygdala was
initially defined as the gray matter
superior to the right hippocampus and
separated from it by the inferior horn of
the right lateral ventricle. It was inferior
to the internal capsule, the auditory and
optic radiations, the anterior
commissure, and the right lenticular
nucleus. Proceeding anterior, the right
hippocampus and lateral ventricles
diminish being replaced entirely by the
amygdala (it was the only structure
occupying the temporal pole other than
the cerebrum).
The same definition as for the right
amygdale.
Gray matter that connected the inferior
right caudate laterally to the superior
right lenticular nucleus.
The same definition as for the right
acumbens.
First defined in posterior coronal sections
as gray matter medial to the posterior
horn of the right lateral ventricle. More
anterior, it became localized inferior to
the posterior and inferior horns of the
right lateral ventricle.
The same definition as for the right
hippocampus.
Sum of the right and left diencephalons.
Sum of the right medial gray matter and
the right hypothalamus.
Sum of the left medial gray matter and the
left hypothalamus.
After segregating the right hemisphere,
gray matter lateral to midline and
bounded by the internal capsule, the
auditory and optic radiations, and the
anterior commissure.
The same definition as for the right medial
gray matter.
Gray matter in the right hemisphere,
anterior to the midbrain, lateral to the
midline, and inferior to the thalamus and
anterior commissure.
The same definition as for the right
hypothalamus.
Sum of the midbrain, pons, and medulla.
First defined in posterior coronal sections
by the colliculi, then as the gray matter
superior to the pons connected to the
thalamus and adjacent hemispheres.
The superior extent was defined at
midline by the inferior terminus of the
third ventricle and by an arc extending
from that point to the superior extent of
the cerebral spinal fluid separating the
midbrain and the right cerebral
hemisphere.
967
EFFECTS OF LEAD ON THE RHESUS BRAIN
TABLE 2—Continued
Brain structure
Pons
Medulla
Ventricles
Right lateral
Left lateral
Third
Cerebral aqueduct
Fourth
Corpus callosum
Septum
Fornix
Optic Chiasm
Pineal
Pituitary
Extra parenchymal volume
Definition
Gray matter between the midbrain and
medulla oblongata. The bulging middle
cerebellar peduncles define the lateral
extent of the pons and the superior and
inferior borders defined by the peduncles
were used to differentiate the pons from
the midbrain and the medulla.
Trapezoidal gray mater extending inferior
from the pons and connecting to the
narrower diameter spinal cord. The
inferior boundary was defined manually at
the point below which the spinal cord was
of constant diameter.
Sum of right and left lateral ventricles, the
third and fourth ventricles, and the
cerebral aqueduct.
The structure in the intensity range of
cerebral spinal fluid (defined visually in
pretesting) localized laterally within the
parenchyma of the right hemisphere.
The same definition as for the left lateral
ventricle.
CSF at midline inferior to the lateral
ventricles and superior to the midbrain.
CSF at midline within the midbrain.
CSF within the pons and medulla oblongata.
The structure in the intensity range of white
matter (defined visually in pretesting)
superior to the lateral ventricles and
caudate nuclei and crossing the midline.
Narrow sliver of medial white matter inferior
and orthogonal to the corpus callosum and
bounded laterally by the right and left
lateral ventricles.
First defined in posterior coronal sections as
white matter extending inferior from the
medial corpus callosum and laterally to
the posterior horns of the lateral
ventricles. More anteriorally it is defined
as the enlarged inferior base extending
from the septum.
The juncture of the right and left optic
nerves as they join at midline anterior to
the midbrain.
Small, spherical midline structure surrounded
by CSF, superior to the superior colliculi
and inferior to the corpus callosum.
Spherical midline structure surrounded by
CSF and inferior to the optic chiasm.
The head was first segregated manually
from the surrounding space. After defining
all the other structures listed in this
Table, the extra parenchymal tissue was
automatically identified by intensity limits
determined from pretesting (the limits that
visually separated gray and white
parenchymal matter from cerebral spinal
fluid and other structures constituting the
structures defining the extra parenchymal
volume).
exposure). The covariate in these analyses was total brain volume. These
analyses indicate whether volume differences in specific brain structures were
associated with lead exposure (in utero or the early postnatal period) adjusting
for total brain volume.
Our small sample size dictated that we could reliably detect only very large
effect sizes. Accepting a type I error of 0.05, the control monkeys would have to
have a brain structure volume 2.85 standard deviations greater than that of
the prenatal and the postnatal monkeys to be detected with 80% power by
a one-way ANOVA. Because of our limited power to detect even sizeable
effects, trends in the data were of interest to avoid missing biologically important effects.
Whether to adjust the nominal significance level for multiple comparisons is
an important consideration. We scored 36 independent structures (we also
combined those structures to form composite structures). The probability that
one of the 36 univariate ANCOVAs we calculated would be ‘‘significant’’ at the
0.05 level is 0.84. Nevertheless, we did not adjust for multiple comparisons.
Following Rothman’s (1990) arguments, we chose to identify patterns in the
results and avoid type II errors rather than adjust the type I error rate for
multiple comparisons. In exploratory research, identifying real effects is more
important than avoiding false positives. Our results must be interpreted with
these considerations in mind.
The results of the formal tests and diagnostics we conducted indicated that
the assumptions made by our parametric testing were not violated, however the
power of those formal tests was limited by our small sample size. Because of
the small sample size and uncertainty about the distributions of the measurements, non-parametric analyses were conducted to confirm our parametric
tests. Both fixed effect parametric and Kruskal-Wallis non-parametric one way
ANOVAs were calculated to compare total brain volume of the prenatally lead
exposed group, the postnatally lead exposed group, and the controls. We used
the residual of the linear regressions of total brain volume on each of the
individual component brain structures as the dependent variable in KruskalWallis non-parametric one way ANOVAs. Those ANOVAs are non-parametric
analogs of the ANCOVAs we calculated for those same structures. The results
of the non-parametric analyses confirmed the parametric analyses and are not
reported.
Dose response relationships were explored by correlating the blood lead
levels of the study monkeys from birth to weaning and from the mothers (during
their pregnancies for the prenatal monkeys and while nursing for the postnatal
monkeys) with brain structure volumes identified by the ANCOVAs as affected
by lead exposure. Both Pearson and Spearman partial correlation coefficients
were calculated (partialling out the effect of total brain volume). The calculated
parametric and non-parametric partial correlations were similar. We also
calculated linear regressions (adjusting for total brain volume) in order to
quantify the change in brain structure volumes associated with measured blood
lead levels.
The analyses were conducted using SPSS (version 10.0.7; June 2000; SPSS,
Inc., Chicago, IL. 60606) and NCSS (version NCSS 2004; March 2004; NCSS,
Kaysville, UT) statistical software packages.
RESULTS
Reliability
The ICCs and their 95% confidence intervals (C.I.) are
presented in Table 3 for each of the brain structures measured.
Some of the structures listed in Table 3 are combinations of
component structures (e.g., the basal ganglia included the
caudate nucleus, the lenticular nucleus, the amygdala, and the
acumbens). Table 3 also presents the within standard deviations
(wSDs) and repeatability for the brain structures measured. The
reliabilities of our measurements were very high; all would be
968
LASKY ET AL.
FIG. 1. An example of scoring an MRI slice. (a) An example of an unscored 3D T1 FSPGR MRI slice, one of the 124 slices acquired on each monkey in this
study. (b) The manually identified structures listed in Table 2 were scored using the Analyze ‘‘Manual Trace’’ software tool. (c) After manually identifying
structures, predetermined intensity limits were used to identify right and left cerebral white matter and cerebellar white matter automatically using the Analyze
‘‘Auto Trace’’ software tool. The Analyze ‘‘Auto Trace Limit’’ software tool was used to isolate the right and left cerebral hemispheres, the cerebellum, and the
brain from the spinal cord. (d) After identifying the cerebral white matter, right and left cerebral gray matter was identified using the predetermined intensity limits
and the Analyze ‘‘Auto Trace’’ software tool. Not shown was the automatic identification of ‘‘extraparenchymal’’ tissue (including CSF and skull). The Analyze
‘‘Auto Trace Limit’’ software tool was used to restrict the cerebral gray matter and the extraparenchymal to the brain and skull respectively. After scoring the first
slice, the remaining slices were scored in the same manner as described above. These coronal slices were scored in a posterior to anterior direction. When all 124
slices were scored, the ANALYZE software generated volume measurements as well as other statistics for each of the structures identified and listed in Table 2.
Those measurements were then compared for monkeys with different exposures to lead.
characterized as having excellent reliability. Automatically
scored structures (e.g., cerebral gray and white matter) had
among the highest reliabilities as expected; however, even
manually scored structures with the poorest reliabilities had
ICCs greater than 0.77. As scored, the volumetric measurements
have little intra-scorer measurement error associated with them.
Gender Effects
Female rhesus monkeys have smaller brains than males
(Cupp and Uemura, 1981; Franklin et al., 2000). Gender
differences in total brain volume in the study sample were
explored by a 2(gender) 3 2 (group) analysis of variance.
Because of the small sample size, the two lead exposed groups
were combined for this analysis. There was a significant gender 3
group interaction (F(1,11) ¼ 5.22; p ¼ 0.043) for total brain
volume, the only significant (p < 0.05) effect. (Male total brain
volumes were more variable than female brain volumes
violating the homoscedasticity assumption.) Control female
monkeys (mean ¼ 100,384 mm3, SD ¼ 4499 mm3) control
male monkeys (mean ¼ 97,546 mm3, SD ¼ 12,278 mm3), and
lead-exposed male monkeys (mean ¼ 100,906 mm3, SD ¼
7844 mm3) had similar sized brains. In contrast, lead exposed
females had smaller brains (mean ¼ 84,534 mm3, SD ¼
3,220 mm3) than the other monkeys. These results suggest that
the control males had unusually small brains. There is some
support for this conjecture because the control males also
weighed less than expected. They were similar in body weight
to the control females (median body weight for the control
males was 7.18 kg, range ¼ 6.75 to 11.10 kg; for the control
females it was 8.20 kg, range ¼ 6.80 to 8.50 kg) and
significantly lighter than the lead exposed males (median body
weight ¼ 12.22 kg, range ¼ 9.95 to 12.90 kg). As expected the
lead exposed males were significantly heavier than lead
exposed females (median body weight ¼ 6.65 kg, range ¼
6.40 to 7.50 kg). One implication of these analyses is the need to
adjust for differences in total brain volume because of sampling
variation in the sizes of the study monkeys.
EFFECTS OF LEAD ON THE RHESUS BRAIN
TABLE 3
Reliabilities of the Measured Brain Volumes
Brain
structure
Total brain
Cerebrum
Right cerebrum
Left cerebrum
Right cerebral gray matter
Left cerebral gray matter
Right cerebral white matter
Left cerebral white matter
Cerebellum
Cerebellum, gray matter
Cerebellum, white matter
Basal ganglia
Right basal ganglia
Left basal ganglia
Right caudate
Left caudate
Right lenticular nucleus
Left lenticular nucleus
Right amygdala
Left amygdala
Right acumbens
Left acumbens
Right hippocampus
Left hippocampus
Diencephalon
Right diencephalon
Left diencephalon
Right medial gray matter
Left medial gray matter
Right hypothalamus
Left hypothalamus
Brainstem
Midbrain
Pons
Medulla
Ventricles
Right lateral
Left lateral
Third
Cerebral aqueduct
Fourth
Corpus callosum
Septum
Fornix
Optic Chiasm
Pineal
Pituitary
Extra parenchymal volume
ICC
(95% C.I.)
.9997
.9982
.9980
.9919
.9959
.9970
.9962
.9913
.9981
.9997
.9997
.9598
.9542
.9628
.9868
.8990
.8550
.9447
.9808
.9771
.8386
.9331
.9540
.9628
.9308
.8531
.8199
.8776
.7788
.8355
.9008
.9492
.9431
.9718
.9470
.9837
.9281
.9865
.8593
.9518
.8244
.9972
.9224
.9390
.7773
.9941
.9787
.9966
(.9987, .9999)
(.9905, .9996)
(.9906, .9996)
(.9635, .9984)
(.8576, .9994)
(.9792, .9994)
(.9805, .9993)
(.9578, .9983)
(.9908, .9996)
(.9987, .9999)
(.9985, .9999)
(.8229, .9917)
(.7772, .9908)
(.8420, .9923)
(.9163, .9975)
(.6036, .9786)
(.4789, .9686)
(.7545, .9886)
(.9069, .9961)
(.8793, .9955)
(.4255, .9649)
(.5801, .9873)
(.7915, .9906)
(.8309, .9924)
(.7196, .9855)
(.4444, .9686)
(.1779, .9638)
(.5043, .9743)
(.0236, .9577)
(.4123, .9642)
(.4193, .9810)
(.7842, .9895)
(.7498, .9883)
(.8782, .9942)
(.7591, .9892)
(.9240, .9967)
(.7082, .9849)
(.9323, .9973)
(.4750, .9698)
(.7801, .9902)
(.3904, .9615)
(.9873, .9994)
(.6746, .9839)
(.7056, .9877)
(.2387, .9509)
(.9735, .9988)
(.8981, .9957)
(.9840, .9993)
Within subject SD in mm3
(repeatability in mm3)
128.08
221.11
120.24
223.96
121.47
126.85
149.34
163.04
55.10
28.69
35.37
131.13
69.49
66.77
16.69
24.80
37.11
33.17
17.79
24.00
10.61
11.48
17.26
19.49
48.93
50.47
38.49
46.04
43.45
10.42
11.16
74.22
42.31
25.17
45.72
19.53
18.98
10.03
2.61
1.72
9.39
16.20
3.29
14.90
17.91
0.66
5.15
347.17
(354.77)
(612.47)
(333.05)
(620.38)
(336.47)
(351.38)
(413.67)
(451.62)
(152.63)
(79.46)
(97.97)
(363.22)
(192.47)
(184.95)
(46.23)
(68.71)
(102.80)
(91.87)
(49.27)
(66.48)
(29.39)
(31.79)
(47.81)
(53.99)
(135.54)
(139.81)
(106.63)
(127.52)
(120.35)
(28.86)
(30.91)
(205.60)
(117.21)
(69.73)
(126.65)
(54.09)
(52.58)
(27.78)
(7.24)
(4.75)
(26.01)
(44.86)
(9.12)
(41.28)
(49.61)
(1.84)
(14.27)
(961.65)
Correlations among Structures
Table 4 presents correlations among major brain structures.
Total brain volume correlates most strongly with the cerebrum
followed by the cerebellum, basal ganglia, brainstem, diencephalon, the ventricles, and is negatively but non-significantly
correlated with extra parenchymal volume. Not surprisingly,
969
monkeys with larger brains tended to have larger component
structures. Extra parenchymal volume was the only variable not
included in total brain volume. The negative correlation with
extra parenchymal volume suggests that the larger the brain the
less the extra parenchymal volume.
We calculated partial correlations to determine the correlations among brain structures adjusting for total brain volume
(i.e., standardizing the sizes of the brain structures scored by
the overall size of the brain). Those partial correlations are
presented below the diagonal in Table 4. The partial correlation
between the cerebrum and cerebellum became negative after
adjusting for total brain volume. The partial correlations among
the other structures were nonsignificant with a few exceptions
(e.g., a significant positive partial correlation between extra
parenchymal volume and the basal ganglia). For the cerebrum
the white and gray matter was negatively (but nonsignificantly)
correlated, while the negative correlation between cerebellar
gray and white matter was significant (r ¼ .786, p ¼ 0.021).
Not surprisingly, right and left analogs of the same component
structure were also positively correlated for the brain structures
scored, significantly so for gray (r ¼ .754, p ¼ 0.031) and white
(r ¼ .734, p ¼ 0.038) cerebral matter.
Lead Effects
Table 5 presents group differences for the brain structures
measured. Except for total brain volume, all of the other measurements were adjusted by total brain volume. There were no
significant ( p < 0.05) differences in total brain volume as a function of lead exposure group by a one way ANOVA. ANCOVAs
with total brain volume as the covariate were calculated to evaluate
lead effects on individual brain structures adjusted for variation in
total brain volumes. Only the lateral ventricles (F(2,11) ¼ 5.105;
p ¼ 0.027 for the right lateral ventricle and F(2,11) ¼ 4.389; p ¼
0.040 for the left lateral ventricle) and cerebral white matter
(F(2,11) ¼ 4.158; p ¼ 0.045 for the right cerebral white matter
and a similar but non-significant difference for the left cerebral
white mater, F(2,11) ¼ 2.331; p ¼ 0.143) significantly differed
among the lead exposure groups of monkeys. Monkeys exposed to
lead prenatally had the largest lateral ventricles. Control monkeys
had more cerebral white matter than lead exposed monkeys. The
prenatally exposed monkeys had the least cerebral white matter.
Differences in cerebral white matter were similar for both
hemispheres, but more consistent for the right cerebral hemisphere. There also tended to be more cerebral gray matter in the
brains of lead-exposed monkeys although that trend was not
statistically significant. The results and trends in Table 5 are
summarized and presented graphically in Figure 2.
Dose response relationships were explored by correlating
blood lead levels with the increased cerebral white matter and
decreased lateral ventricle volumes. Partial correlations were
calculated in order to adjust for total brain size. The partial
correlations between blood lead levels in the study monkeys
over the first six months of postnatal life were negatively but
970
LASKY ET AL.
TABLE 4
Pearson Correlation Coefficients among Major Brain Structures above the Diagonal and Partial Correlation Coefficients Adjusting
for Total Brain Volume Below the Diagonal
Brain
structure
Total brain
Cerebrum
Cerebellum
Basal ganglia
Diencephalon
Brainstem
Ventricles
Extra parenchymal
volume
Cerebrum
Cerebellum
.99*
.89*
.83*
.79*
.27
.24
.25
.30
.19
.23
.01
.35
.35
.15
Basal
ganglia
.87*
.84*
.73*
.40
.60
.45
.76*
Diencephalon
.67
.64
.68
.44
Brainstem
.78*
.75*
.70
.60
.68
.33
.45
.74
.42
.18
Ventricles
.36
.40
.18
.11
.55
.71*
Extra
parenchymal
.48
.52
.49
.09
.60
.28
.25
.09
*Indicates p < 0.05, two-tailed test.
non-significantly correlated with white matter in the right (r ¼
0.34, p ¼ 0.228) and the left (r ¼ 0.38, p ¼ 0.180) cerebral
hemispheres. Those partial correlations were significant when
maternal blood lead levels (during pregnancy for the prenatal
monkeys and during nursing for the postnatal monkeys) rather
than study monkey blood lead levels were correlated with
cerebral white matter volumes (r ¼ 0.62, p ¼ 0.019 for the
right cerebrum and r ¼ 0.62, p ¼ 0.018 for the left cerebrum).
None of the partial correlations between blood levels and
lateral ventricle volumes were significant although all four
correlations were positive as expected.
The changes in cerebral white matter volumes associated
with measured blood lead levels were quantified by the slopes
of linear regressions. There was a loss of 54.4 mm3 (95% C.I. ¼
38.9, 147.7 mm3) and 49.3 mm3 (95% C.I. ¼ 26.1, 124.8
mm3) of right and left cerebral white matter respectively for
every lg/dl increase in blood lead levels of the study monkeys.
There were losses of 39.7 mm3 (95% C.I. ¼ 7.8, 71.5 mm3) and
32.7 mm3 (95% C.I. ¼ 6.7, 58.8 mm3) of right and left cerebral
white matter respectively for every lg/dl increase in maternal
blood lead levels.
DISCUSSION
Franklin et al. (2000) report that male whole brain volumes
are approximately 20% larger than female whole brain
volumes. As expected, lead exposed males had larger brains
(by 16%) than lead exposed females in the study sample.
However, male controls had similar sized brains as female
controls. They were also similar in body weight suggesting by
chance the sample included unusually small male control
rhesus monkeys. Therefore, it would be misleading to interpret
the gender 3 lead interaction in the study sample as a lead
effect. Understanding gender differences in brain volumes is
important. They must be addressed definitely with a larger and
more representative sample.
We did not detect reliable differences between lead
exposed and control monkeys in total brain volume. Our
ability to detect differences in total brain volume may have
been compromised by unrepresentative sampling as discussed
above. We evaluated lead effects on brain structures adjusting
for total brain volume to reduce concerns about unrepresentative sampling. Monkeys exposed to lead early in life had
less cerebral white matter in their brains than the control
monkeys. The deficits were larger in monkeys exposed to
lead prenatally. The blood lead levels of the mothers and the
study monkeys were negatively correlated with cerebral
white matter suggesting a dose response relationship. The
relationships between blood lead levels and cerebral white
matter may have been stronger with maternal blood lead
levels than blood lead levels of the study monkeys because
the latter only indirectly reflected lead exposure for the
monkeys exposed prenatally. The exploratory nature of the
analyses cautions against over-interpreting these results.
A study by Deng and colleagues may explain why lead
exposure was associated with a reduction in cerebral white
matter in this study. Deng et al. studied the effects of lead
exposure on rat oligodendrocyte progenitor cells and myelin
production. Chronic lead exposure resulted in interference of
the timely developmental maturation of oligodendrocyte
progenitors resulting in hypo- and demyelination of axons
(Deng et al., 2001).
Because the overall size of the cerebrum was not affected by
lead exposure in this study, cerebral tissue scored as gray
matter on T1 MRIs may have been unmyelinated white matter.
The differentiation of gray and white matter was done
automatically on the basis of intensity level differences.
Therefore, the trend to more cerebral gray matter in lead
exposed monkeys may be explained by including unmyelinated
white matter as ‘‘gray matter’’. Alternatively, the fluid content
in some cerebral tissue may be increased in lead exposed
monkeys accounting for the reduced ‘‘white matter.’’ The
increased size of the lateral ventricles also suggests that
971
EFFECTS OF LEAD ON THE RHESUS BRAIN
TABLE 5
Differences in Brain Structure Volumes (Mean and 95% C.I. in mm3) Adjusted by Total Brain Volume Among Prenatally Lead
Exposed Monkeys, Postnatally Lead Exposed Monkeys, and Monkeys Never Exposed to Lead
Brain structure
Total braina
Cerebrum
Right cerebrum
Left cerebrum
Right cerebral gray matter
Left cerebral gray matter
Right cerebral white matter
Left cerebral white matter
Cerebellum
Cerebellum, gray matter
Cerebellum, white matter
Basal ganglia
Right basal ganglia
Left basal ganglia
Right caudate
Left caudate
Right lenticular nucleus
Left lenticular nucleus
Right amygdala
Left amygdala
Right acumbens
Left acumbens
Right hippocampus
Left hippocampus
Diencephalon
Right diencephalon
Left diencephalon
Right medial gray matter
Left medial gray matter
Right hypothalamus
Left hypothalamus
Brainstem
Midbrain
Pons
Medulla
Ventricles
Right lateral
Left lateral
Third
Cerebral aqueduct
Fourth
Corpus callosum
Septum
Fornix
Optic Chiasm
Pineal
Pituitary
Extra parenchymal volumea
Prenatal monkeys
(n ¼ 3)
98707
70415
35276
35139
24498
24229
10778
10910
9766
6614
3152
4185
2075
2109
650
709
979
949
433
430
14
21
530
533
2913
1364
1549
1250
1441
114
108
4838
1828
1937
1073
1288
556
564
42
30
97
1514
30
289
170
5
120
41048
(64051, 133363)
(69126, 71705)
(34327, 36225)
(34039, 36240)
(20305, 28690)
(20255, 28203)
(6842, 14715)
(7323, 14498)
(8689, 10843)
(4464, 8764)
(131, 6172)
(3681, 4689)
(1784, 2367)
(1875, 2344)
(494, 807)
(596, 822)
(889, 1069)
(867, 1031)
(272, 593)
(274, 586)
(18, 45)
(20, 63)
(433, 627)
(432, 635)
(2583, 3243)
(1198, 1531)
(1346, 1752)
(1058, 1443)
(1227, 1654)
(77, 151)
(61, 155)
(4473, 5202)
(1604, 2052)
(1648, 2226)
(821, 1324)
(1060, 1515)
(455, 657)
(445, 682)
(25, 59)
(20, 39)
(65, 128)
(1233, 1795)
(14, 45)
(192, 386)
(118, 223)
(7, 16)
(75, 166)
(11355, 70741)
Postnatal monkeys
(n ¼ 4)
90276 (78885, 101667)
70666 (69476, 71856)
36326 (35451, 37202)
34340 (33325, 35355)
21503 (17634, 25371)
22401 (18735, 26067)
14824 (11192, 18456)
11939 (8629, 15248)
9946 (8953, 10940)
7281 (5297, 9264)
2666 (121, 5452)
4645 (4180, 5109)
2305 (2036, 2574)
2340 (2123, 2556)
695 (550, 839)
643 (538, 747)
1012 (929, 1095)
1067 (991, 1142)
574 (426, 722)
594 (450, 738)
25 (4, 53)
36 (2, 74)
487 (398, 577)
502 (409, 595)
2368 (2063, 2672)
1162 (1009, 1316)
1206 (1018, 1393)
1062 (884, 1240)
1089 (892, 1286)
101 (67, 134)
116 (73, 160)
4740 (4404, 5077)
1812 (1605, 2019)
1771 (1504, 2037)
1158 (926, 1389)
968 (758, 1178)
430 (337, 523)
400 (291, 510)
42 (26, 58)
20 (12, 30)
76 (47, 104)
1638 (1378, 1897)
31 (17, 45)
278 (188, 367)
158 (110, 206)
15 (5, 25)
155 (113, 197)
42291 (41321, 43261)
Control monkeys
(n ¼ 8)
98965
71680
36003
35677
19221
21056
16782
14621
9252
5892
3361
4247
2130
2118
690
695
990
970
431
433
18
22
479
481
2695
1354
1341
1244
1243
110
98
4906
2039
1750
1117
897
385
380
34
23
76
1348
36
225
200
8
141
42044
(91697, 106233)
(70879, 72481)
(35414, 36593)
(34994, 36361)
(16617, 21826)
(18588, 23525)
(14336, 19227)
(12392, 16849)
(8584, 9921)
(4556, 7228)
(1484, 5237)
(3934, 4560)
(1949, 2311)
(1972, 2263)
(593, 787)
(624, 765)
(935, 1046)
(917, 1019)
(331, 531)
(336, 530)
(1, 38)
(4, 48)
(419, 539)
(418, 544)
(2490, 2901)
(1251, 1458)
(1215, 1467)
(1124, 1364)
(1110, 1375)
(88, 133)
(69, 127)
(4680, 5133)
(1900, 2178)
(1571, 1930)
(960, 1273)
(756, 1039)
(322, 448)
(306, 454)
(23, 44)
(17, 29)
(56, 95)
(1174, 1523)
(26, 45)
(164, 285)
(167, 232)
(1, 15)
(113, 170)
(36264, 47825)
Note. Structures that significantly ( p < 0.05) differed among the three lead exposure groups by a 3(group) ANCOVA are in italics. Total brain volume was the
covariate in these analyses.
a
Unadjusted volumes.
972
LASKY ET AL.
FIG. 2. A summary of adjusted volumetric brain differences among monkeys with different lead exposures. Each pie represents the adjusted brain volume for
monkeys with prenatal lead exposure (n ¼ 3), postnatal lead exposure (n ¼ 4), or no exposure to lead (n ¼ 8). Pie slices represent the percentage of the total brain
volume that was cerebral white matter (wm), cerebral gray matter (gm), other wm, other gm, or the ventricles. Offset slices include structures (right cerebral wm
and the right and left lateral ventricles) that differed significantly ( p < 0.05) by univariate ANCOVAs among the three lead exposure groups.
proportions of the tissues constituting the brain may be affected
by exposure to lead early in development.
The existing literature is too diverse at this point to confirm
or refute these exploratory results. The few reports on the
effects of early lead exposure on brain anatomy differ in lead
exposures, species, measurement methodologies, and other
factors. However, results of a previous study may explain in
part why cerebral tissue may be differentially affected (Cremin
et al., 1999). Adult rhesus monkeys were administered lead
orally for five weeks to reach and maintain target blood lead
levels of 35–40 lg/dl. Cremin et al. measured lead in the blood
and the prefrontal cortex, frontal lobe, hippocampus, and
striatum of the brain. Lead concentrations were greatest in
the prefrontal cortex, followed by the frontal lobe, the
hippocampus, and then the striatum. Although the Cremin
et al. study concerned lead exposures as adults, a limited
number of brain structures, and did not distinguish white and
gray matter, it does suggest that lead may concentrate in the
frontal cerebral cortex relative to the hippocampus and
striatum. Thus, in the present study, cerebral white matter
may have been differentially affected because of higher
concentrations of lead in the cerebrum and because myelination is rapid during the period of lead exposure in this study. It
is of note in the Cremin et al. study, that prefrontal cortex lead
concentrations were significantly correlated with integrated
blood lead levels over the entire lead exposure period but not
with blood lead levels collected concurrently with the prefrontal cortex biopsy, emphasizing the necessity of animal
models because lead exposure histories are not known
in humans.
In addition to reporting lead effects on the brain, our study
employed a promising measurement methodology for neurotoxicological studies. MRIs are relatively non-invasive, increasing their potential applications. They avoid some of the
distortions that are consequences of preserving (fixture effects)
and evaluating the brain using laboratory methods. They are
also efficient in that they can produce quantitative information
concerning the brain with a relatively small investment in
analysis time and effort. Non-invasive MRI and invasive
laboratory approaches are complementary, MRIs providing
a broader perspective that can be further refined by detailed
laboratory analyses. Other MRI methodologies promise additional insights. For example, our results suggest myelination
may be disrupted in monkeys exposed to lead early in life.
Diffusion Tensor Imaging (DTI) is a more direct and sensitive
approach to evaluating myelination. Follow-up studies to
identify brain insults due to early lead exposure should include
DTI. In addition, MRS has been used to identify neuronal loss
in the forebrains of lead exposed children and is another
imaging methodology to consider in lead research (Trope et al.,
EFFECTS OF LEAD ON THE RHESUS BRAIN
1998, 2001). Finally, functional MRI directly relates functional
deficits to specific brain structures significantly enhancing our
understanding of the effects of lead on the brain. An advantage
of all these approaches is they can be conducted efficiently and
non-invasively in humans and laboratory animals facilitating
cross-species comparisons. In short, MRI technologies have
much to offer research evaluating the effects of lead and other
environmental toxins on the brains of humans and laboratory animals.
The scoring procedures adopted for this study were demonstrated to be highly reliable. There are several factors that
contributed to the consistency of scoring. Some of the
procedures were semi-automated identifying all voxels within
a defined intensity range as belonging to the defined structures.
It is not surprising that semi-automated approaches to identifying structures are highly reliable. It was surprising that very
high reliabilities were achieved for structures defined manually.
The reliability of scoring the MRI structures benefited from
the way volumetric data are scored and generated. Minor
differences between independent measurements of the same
brain are likely to be randomly distributed. By summing
measurements from successive slices to calculate volumetric
measurements, those random measurement errors tend to
cancel each other resulting in highly reliable measurements.
For an individual slice, the reliabilities are more modest than
the reliabilities presented in Table 3.
It should be noted that only intra-scorer reliability was
assessed in this study. The measurement errors associated with
different scorers (inter-scorer reliability) and repeating an MRI
on the same subject (test-retest reliability) were not assessed.
The former is critical in estimating the generalizability of the
results by other investigators, the latter in estimating how well
an individual MRI characterizes the brain of the monkey. A
complete characterization of the measurement errors associated with quantitative MRI methods requires evaluating all
sources of measurement variability. Nevertheless, our reliability results indicate that the differences reported for the MRIs on
the study sample are robust. Whether we would replicate our
results with a different scorer or if we obtained different scans
on the study monkeys are important questions we did not
address in this study.
For this study, we chose to employ manual and semiautomated methods to analyze total and component brain
volumes. Fully automated methods for segmentation offer
greater objectivity, replicability, generalizability, and efficiency
(Ashburner and Friston, 2000; Caviness et al., 1999; Fischl
et al., 2002). Furthermore, they presuppose no working
knowledge of brain anatomy. At present we believe that
semi-automated methods involving a scorer with knowledge
of rhesus brain anatomy will produce more accurate results for
some structures. Statistical pattern recognition methods based
on a finite mixture model that partition the brain into gray
matter, white matter, and CSF have been reported by Anderson
et al. (2002). Automated methods are beginning to incorporate
973
anatomical landmarks into the identification of brain structures
because relying on gray scale differences alone is inadequate to
distinguish all brain structures. Multispectral approaches that
combine the information from different scanning methodologies (e.g., T1 and T2) to make discriminations that cannot be
differentiated by the individual scanning methodologies afford
improved identification capabilities. Because of the advantages
of automated methods, conducting and interpreting studies of
the brain can be greatly facilitated.
There have also been considerable gains in the analysis and
interpretation of MRI scans. Notable is the work of Friston,
Worsley, Ashburner, and colleagues. They have developed
Statistical Parametric Mapping (SPM) to test hypotheses about
imaging data, primarily fMRI and PET. Their approach is also
suitable for segmentation studies and includes Voxel-Based
Morphometry or VBM (Ashburner and Friston, 2000). Their
analytic software is available at no cost to the neuro-imaging
community (www.fil.ion.ucl.ac.uk/spm/). Their analytic approach evaluates the relationships between design effects
(experimental groups or stimulus effects) and voxel intensity
adjusting for identified covariates using a general linear model
(GLM) framework. The results of their analyses are statistical
maps that identify the likelihood of group (or stimulus in the
case of functional imaging) differences at an individual voxel
level. They adjust for multiple comparisons using Gaussian
Random Field (GRF) theory. GRF rather than a Bonferroni
type correction is used because voxel intensities are not
independent but are correlated. Their approach lends itself to
hierarchical mixed modeling in which voxels could be
identified as belonging to higher level factors such as brain
structures and those to even higher levels factors such as
experimental groups. Prior information can be incorporated
from previous scans, other measurement modalities, or from
the literature by adopting a Bayesian framework.
Despite the appeal of Friston et al.’s approach, we did not
adopt it for this study. Their approach is voxel based; ours is
anatomical structure based. The distinctions they make depend
on intensity level differences. For the reasons we have alluded
to, we have felt it necessary to also include anatomical
landmarks to identify many brain structures of interest. An
SPM identifies voxels that are likely to differ in intensity
between lead exposed and control monkeys. To do so the
images must be co-registered so that ‘‘lead exposed monkey’’
voxels can be compared to the corresponding ‘‘control
monkey’’ voxels. Co-registering is not a trivial task and best
achieved when the referent is well-characterized (artifact free
and representative of the referent population) and the comparison group is not markedly discrepant from the referent. The
more discrepant the two groups the more smoothing of the
voxels required for co-registration, compromising the resolution of the analysis. Small sample sizes such as the present
study exacerbate these requirements for valid co-registration.
In contrast, the approach we adopted does not require coregistering of the scans but does require the identification of the
974
LASKY ET AL.
structures of interest by a trained scorer. By using hierarchical
mixed modeling or planned rather post hoc comparisons, it
would be possible to define ‘‘anatomical contrasts’’ within
a VBM approach similar to those we have made; however, it
would require a priori identification of the voxels belonging to
those anatomical structures and, therefore, the same anatomical
classification our approach depends on.
Our approach also differs from Friston’s VBM approach in
that we did not adjust our results for multiple comparisons.
Specifically, our concern was not false positives but missing
true differences if they exist. Because of the sheer number of
voxels analyzed, adjusting for dependences among those
voxels is necessary for VBM. The number of anatomical
structures we analyzed, although sizable, was more manageable. Furthermore, all structures we identified were not equal
(although we did not feel planned comparisons were appropriate given a sparse literature). Specifically, we identified differences in cerebral white matter. Cerebral white matter was
automatically identified, reliably scored, and among the largest
structures scored.
We measured many structures because the specifics of lead
effects on the brain are largely unknown. Therefore, our results
may be explained by chance. Furthermore, we could only
expect to identify lead effects that were sizeable because of our
small sample size. Therefore, our study is vulnerable to missing
more subtle lead effects. We did demonstrate that highly
reliable volumetric measurements of brain structures can be
recorded from adult rhesus monkey MRIs. Consequently, the
differences we reported characterize the study monkeys on the
MRIs collected for this study. However, it is less clear whether
these results would generalize to other samples of monkeys (or
humans) or even to the same monkeys with different MRIs.
This study needs replication on a larger sample. Nevertheless,
our results suggest that early lead exposure may disrupt
cerebral myelination in rhesus monkeys.
ACKNOWLEDGMENTS
The authors would like to thank Nicole Giles, Heather Gottfried, and Kyle
Meyer for their technical assistance with this project. The data for this study
were collected under a grant from NIEHS awarded to N.K.L. (NIH R01
ES04860). The analyses were supported by a grant awarded to R.E.L. from the
Center for Clinical Research and Evidence-Based Medicine, the University of
Texas-Houston Medical School.
REFERENCES
Anderson, A. H., Zhang, Z., Avison, M. J., and Gash, D. M. (2002). Automated
segmentation of multispectral brain MR images. J. Neurosci. Methods 122,
13–23.
Ashburner, J., and Friston, K. J. (2000). Voxel-based morphometry-the
methods. Neuroimage 11, 805–821.
Bellinger, D. C., and Needleman, H. L. (2003). Intellectual impairment and
blood lead levels. N. Engl. J. Med. 349, 500–502.
Bland, J. M., and Altman, D. G. (1996). Statistics notes: Measurement error and
correlation coefficients. BMJ 313, 41–42.
BrainInfo. View the Template Atlas page. Available at: http://braininfo.rprc.
washington.edu. Accessed 10 February 2004 and 30 June 2004.
Canfield, R. L., Henderson, C. R., Jr., Cory-Slechta, D. A., Cox, C., Jusko,
T. A., and Lanphear, B. P. (2003). Intellectual impairment in children with
blood lead concentrations below 10 microg per deciliter. N. Engl. J. Med.
348, 1517–1526.
Caviness, V. S., Jr., Lange, N. T., Makris, N., Herbert, M. R., and Kennedy,
D. N. (1999). MRI-based brain volumetrics: Emergence of a developmental
brain science. Brain Dev. 21, 289–295.
Comparative Mammalian Brain Collections. The Rhesus Monkey page.
Available at: www.brainmuseum.org. Accessed 10 February 2004 and 30
June 2004.
Cremin, J. D., Jr., Luck, M. L., Laughlin, N. K., and Smith, D. R. (1999).
Efficacy of succimer chelation for reducing brain lead in a primate model of
human lead exposure. Toxicol. Appl. Pharmacol. 161, 283–293.
Cupp, C. J., and Uemura, E. (1981). Body and organ weights in relation to age
and sex in Macaca mulatta. J. Med. Primatol. 10, 110–123.
Deng, W., McKinnon, K. D., and Poretz, R. D. (2001). Lead exposure delays
the differentiation of oligodendroglial progenitors in vitro. Toxicol. Appl.
Pharmacol. 174, 235–244.
Ediger, R. D., and Coleman, R. L. (1972). A modified Delves cup atomic
absorption procedure for the determination of lead in blood. At. Absopt.
Newsl. 11, 33–36.
Fischl, B., Salat, D. H., Busa, E., Albert, M., Dieterich, M., Haselgrove, C., van
der Kouwe, A., Killiany, R., Kennedy, D., Klaveness, S., Montillo, A.,
Makris, N., Rosen, B., and Dale, A. M. (2002). Whole brain segmentation:
Automated labeling of neuroanatomical structures in the human brain.
Neuron 33, 341–355.
Franklin, M. S., Kraemer, G. W., Shelton, S. E., Baker, E., Kalin, N. H., and
Uno, H. (2000). Gender differences in brain volume and size of corpus
callosum and amygdala of rhesus monkey measured from MRI images.
Brain Res. 852, 263–267.
Institute of Laboratory Animal Resources, Committee on the Care and Use of
Laboratory Animals. (1985). Guide for the care and use of laboratory
animals. NIH publication #86-23. U.S. Department of Health and Human
Services, Washington, DC.
Iosifescu, D. V., Shenton, M. E., Warfield, S. K., Kikinis, R., Dengler, J., Jolesz,
F. A., and McCarley, R. W. (1997). An automated registration algorithm for
measuring MRI subcortical brain structures. Neuroimage 6, 13–25.
Lasky, R. E., Maier, M., Snodgrass, E., Hecox, K. E., and Laughlin, N. K.
(1995). The effects of lead on otoacoustic emissions and auditory evoked
potentials in monkeys. Neurotoxicol. Teratol. 17, 633–644.
Laboratory of Neuroimaging, UCLA. The Monkey Atlas page. Available at:
www.loni.ucla.edu/Research/Atlases/MonkeyAtlas.html. Accessed 10 February 2004 and 30 June 2004.
Ma, T., Chen, H. H., Lim, D. K., Hume, A. S., and Ho, I. K. (1997). Effects of
subacute lead exposure on [3H]MK-801 binding in hippocampus and
cerebral cortex in the adult rat. Brain Res. 760, 187–192.
Mani, J., Chaudhary, N., Kanjalkar, M., and Shah, P. U. (1998). Cerebellar
ataxia due to lead encephalopathy in an adult. J. Neurol. Neurosurg.
Psychiatry 65, 797.
Martin, R. F., and Bowden, D. M. (1996). A stereotaxic template atlas of the
macaque brain for digital imaging and quantitative neuroanatomy. Neuroimage 4, 119–150.
Matochik, J. A., Chefer, S. I., Lane, M. A., Woolf, R. I., Morris, E. D., Ingram,
D. K., Roth, G. S., and London, E. D. (2000). Age-related decline in striatal
volume in monkeys as measured by magnetic resonance imaging. Neurobiol.
Aging 21, 591–598.
EFFECTS OF LEAD ON THE RHESUS BRAIN
Needleman, H. L., and Landrigan, P. J. (2004). What level of lead in blood is
toxic for a child? Am. J. Public Health 94, 8.
975
Smith, S. M., Zhang, Y., Jenkinson, M., Chen, J., Matthews, P. M., Federico, A.,
and De Stefano, N. (2002). Accurate, robust, and automated longitudinal and
cross-sectional brain change analysis. Neuroimage 17, 479–489.
Paxinos, G., Huang, X. F., and Toga, A. W. (2000). The Rhesus Monkey Brain in
Stereotaxic Coordinates. Academic Press, San Diego, CA.
Reddy, G. R., Basha, M. R., Devi, C. B., Suresh, A., Baker, J. L., Shafeek, A.,
Heinz, J., and Chetty, C. S. (2003). Lead induced effects on acetylcholinesterase activity in cerebellum and hippocampus of developing rat. Int. J.
Dev. Neurosci. 21, 347–352.
Struzynska, L., Bubko, I., Walski, M., and Rafalowska, U. (2001). Astroglial
reaction during the early phase of acute lead toxicity in the adult rat brain.
Toxicology 165, 121–131.
Reuhl, K. R., Rice, D. C., Gilbert, S. G., and Mallett, J. (1989). Effects of
chronic developmental lead exposure on monkey neuroanatomy: Visual
system. Toxicol. Appl. Pharmacol. 99, 501–509.
Teo, J. G., Goh, K. Y., Ahuja, A., Ng, H. K., and Poon, W. S. (1997).
Intracranial vascular calcifications, glioblastoma multiforme, and lead
poisoning. AJNR Am. J. Neuroradiol. 18, 576–579.
Rothman, K. J. (1990). No adjustments are needed for multiple comparisons.
Epidemiology 1, 43–46.
Schantz, S. L., Laughlin, N. K., Van Valkenberg, H. C., and Bowman, R. E.
(1986). Maternal care by rhesus monkeys of infant monkeys exposed
to either lead or 2,3,7,8-tetrachlorodibenzo-p-dioxin. Neurotoxicology 7,
637–650.
Trope, I., Lopez-Villegas, D., Cecil, K. M., and Lenkinski, R. E. (2001).
Exposure to lead appears to selectively alter metabolism of cortical gray
matter. Pediatrics 107, 1437–1442.
Schindler, M. K., Wang, L., Selemon, L. D., Goldman-Rakic, P. S., Rakic, P.,
and Csernansky, J. G. (2002). Abnormalities of thalamic volume and shape
detected in fetally irradiated rhesus monkeys with high dimensional brain
mapping. Biol. Psychiatry 51, 827–837.
Stephenson, J. (2003). CDC report on environmental toxins: Some progress,
some concerns. JAMA 289, 1230–1233.
Trope, I., Lopez-Villegas, D., and Lenkinski, R. E. (1998). Magnetic resonance
imaging and spectroscopy of regional brain structure in a 10-year-old boy
with elevated blood lead levels. Pediatrics 101, E7.
Tuzun, M., Tuzun, D., Salan, A., and Hekimoglu, B. (2002). Lead encephalopathy: CT and MR findings. J. Comput. Assist. Tomogr. 26, 479–481.
Schroter, C., Schroter, H., and Huffmann, G. (1991). Neurologic and
psychiatric manifestations of lead poisoning in adults (case report and
review of the literature). Fortschr. Neurol. Psychiatr. 59, 413–424.
Wei, X., Warfield, S. K., Zou, K. H., Wu, Y., Li, X., Guimond, A., Mugler, J. P.
3rd, Benson, R. R., Wolfson, L., Weiner, H. L., and Guttmann, C. R. (2002).
Quantitative analysis of MRI signal abnormalities of brain white matter with
high reproducibility and accuracy. J. Magn. Reson. Imaging 15, 203–209.
Slomianka, L., Rungby, J., West, M. J., Danscher, G., and Andersen, A. H.
(1989). Dose-dependent bimodal effect of low-level lead exposure on the
developing hippocampal region of the rat: A volumetric study. Neurotoxicology 10, 177–190.
Weisskopf, M. G., Hu, H., Mulkern, R. V., White, R., Aro, A., Oliveira, S., and
Wright, R. O. (2004). Cognitive deficits and magnetic resonance spectroscopy in adult monozygotic twins with lead poisoning. Environ. Health
Perspect. 112, 620–625.