Chemokines in cerebrospinal fluid correlate with cerebral metabolite patterns in HIV-infected individuals.

Abstract

Chemokines influence HIV neuropathogenesis by affecting the HIV life cycle, trafficking of macrophages into the nervous system, glial activation, and neuronal signaling and repair processes; however, knowledge of their relationship to in vivo measures of cerebral injury is limited. The primary objective of this study was to determine the relationship between a panel of chemokines in cerebrospinal fluid (CSF) and cerebral metabolites measured by proton magnetic resonance spectroscopy (MRS) in a cohort of HIV-infected individuals. One hundred seventy-one stored CSF specimens were assayed from HIV-infected individuals who were enrolled in two ACTG studies that evaluated the relationship between neuropsychological performance and cerebral metabolites. Concentrations of six chemokines (fractalkine, IL-8, IP-10, MCP-1, MIP-1β, and SDF-1) were measured and compared with cerebral metabolites individually and as composite neuronal, basal ganglia, and inflammatory patterns. IP-10 and MCP-1 were the chemokines most strongly associated with individual cerebral metabolites. Specifically, (1) higher IP-10 levels correlated with lower N-acetyl aspartate (NAA)/creatine (Cr) ratios in the frontal white matter and higher MI/Cr ratios in all three brain regions considered and (2) higher MCP-1 levels correlated with lower NAA/Cr ratios in frontal white matter and the parietal cortex. IP-10, MCP-1, and IL-8 had the strongest associations with patterns of cerebral metabolites. In particular, higher levels of IP-10 correlated with lower neuronal pattern scores and higher basal ganglia and inflammatory pattern scores, the same pattern which has been associated with HIV-associated neurocognitive disorders (HAND). Subgroup analysis indicated that the effects of IP-10 and IL-8 were influenced by effective antiretroviral therapy and that memantine treatment may mitigate the neuronal effects of IP-10. This study supports the role of chemokines in HAND and the validity of MRS as an assessment tool. In particular, the findings identify relationships between the immune response-particularly an interferon-inducible chemokine, IP-10-and cerebral metabolites and suggest that antiretroviral therapy and memantine modify the impact of the immune response on neurons.

Full text

Page 1

Chemokines in cerebrospinal fluid correlate with cerebral
metabolite patterns in HIV-infected individuals
Scott L. Letendre & Jialin C. Zheng & Marcus Kaul &
Constantin T. Yiannoutsos & Ronald J. Ellis &
Michael J. Taylor & Jennifer Marquie-Beck &
Bradford Navia & for the HIV Neuroimaging Consortium
Received: 20 September 2010 /Revised: 14 November 2010 /Accepted: 24 November 2010 /Published online: 19 January 2011
# The Author(s) 2011. This article is published with open access at Springerlink.com
Abstract Chemokines influence HIV neuropathogenesis
by affecting the HIV life cycle, trafficking of macrophages
into the nervous system, glial activation, and neuronal
signaling and repair processes; however, knowledge of their
relationship to in vivo measures of cerebral injury is
limited. The primary objective of this study was to
determine the relationship between a panel of chemokines
in cerebrospinal fluid (CSF) and cerebral metabolites
measured by proton magnetic resonance spectroscopy
(MRS) in a cohort of HIV-infected individuals. One
hundred seventy-one stored CSF specimens were assayed
from HIV-infected individuals who were enrolled in two
ACTG studies that evaluated the relationship between
neuropsychological performance and cerebral metabolites.
Concentrations of six chemokines (fractalkine, IL-8, IP-10,
MCP-1, MIP-1β, and SDF-1) were measured and com-
pared with cerebral metabolites individually and as com-
posite neuronal, basal ganglia, and inflammatory patterns.
IP-10 and MCP-1 were the chemokines most strongly
associated with individual cerebral metabolites. Specifically,
(1) higher IP-10 levels correlated with lower N-acetyl
aspartate (NAA)/creatine (Cr) ratios in the frontal white
matter and higher MI/Cr ratios in all three brain regions
considered and (2) higher MCP-1 levels correlated with
lower NAA/Cr ratios in frontal white matter and the parietal
cortex. IP-10, MCP-1, and IL-8 had the strongest associa-
tions with patterns of cerebral metabolites. In particular,
higher levels of IP-10 correlated with lower neuronal pattern
scores and higher basal ganglia and inflammatory pattern
scores, the same pattern which has been associated with
HIV-associated neurocognitive disorders (HAND). Subgroup
analysis indicated that the effects of IP-10 and IL-8 were
influenced by effective antiretroviral therapy and that
memantine treatment may mitigate the neuronal effects of
IP-10. This study supports the role of chemokines in HAND
and the validity of MRS as an assessment tool. In particular,
the findings identify relationships between the immune
response—particularly an interferon-inducible chemokine,
IP-10—and cerebral metabolites and suggest that antiretro-
viral therapy and memantine modify the impact of the
immune response on neurons.
Keywords CSF.Chemokines.Magnetic resonance
spectroscopy.HIV.Brain
Background
Chemokines are multifunctional, immunomodulatory pro-
teins that influence HIV neuropathogenesis by multiple
mechanisms. First, chemokines can modify HIV replication
S. L. Letendre (*):R. J. Ellis:M. J. Taylor:J. Marquie-Beck
University of California, San Diego,
220 Dickinson Street, Suite A,
San Diego, CA 92103, USA
e-mail: sletendre@ucsd.edu
J. C. Zheng
University of Nebraska Medical Center,
Omaha, NE, USA
M. Kaul
The Sanford-Burnham Institute,
La Jolla, CA, USA
C. T. Yiannoutsos
Indiana University School of Medicine,
Indianapolis, IN, USA
B. Navia
Tufts Medical School,
Boston, MA, USA
J. Neurovirol. (2011) 17:63–69
DOI 10.1007/s13365-010-0013-2

Page 2

(Lane et al. 2003) and disease progression (Gonzalez et al.
2005) because HIV uses their receptors to enter cells
(Alkhatib et al. 1996; Deng et al. 1996; Dragic et al. 1996;
Feng et al. 1996). Second, chemokines can affect macro-
phage activation (Kaul et al. 2005) and trafficking across
the blood–brain barrier (Weiss et al. 1999). Once in the
nervous system, HIV-infected macrophages can produce
intact virions and inflammatory neurotoxins (Gonzalez-
Scarano and Martin-Garcia 2005). Third, since most neural
cells express chemokine receptors (Bajetto et al. 2001;
Gabuzda et al. 1998), chemokines can affect neuronal
signaling and repair in the brain, leading to aberrations in
glial and neuronal functions. The evidence supporting these
functions derives from multiple sources, including cell
culture experiments (Eugenin et al. 2003; Zheng et al.
2001), human and animal studies of chemokine expression
in body fluids (Kelder et al. 1998; Zink et al. 2001;
Andersson et al. 1998; Letendre 2005; Sevigny et al. 2004)
and brain tissue (Sanders et al. 1998; Sui et al. 2003), and
population studies of chemokine-encoding genes (Gonzalez
et al. 2002; Letendre et al. 2004).
Most of the studies that have evaluated chemokines in
vivo have compared their concentrations to either clinical
dementia staging or performance on neuropsychological
(NP) testing. These studies have provided important
insights but have limitations since they may be affected
by factors such as the effort of the subject, the composition
of the testing battery, or the expertise of the assessor.
Subjects who have abnormal clinical staging or NP
performance seem to have elevated levels of several
chemokines, but this finding does not demonstrate whether
each protein is functioning in a predominantly neurotoxic,
neuroprotective, or bystander role. Imaging methodologies,
such as proton magnetic resonance spectroscopy (MRS),
may address these limitations.
MRS is a sensitive, reliable in vivo method that detects
changes in specific cerebral metabolites, which reflect
disturbances in cellular function. MRS can measure several
cerebral metabolites, including (1) N-acetyl aspartate (NAA),
which reflects neuronal injury or loss, (2) choline (Cho),
which reflects membrane remodeling, (3) myoinositol (MI),
which reflects osmolar shifts and astroglial proliferation, and
(4) creatine (Cr), which reflects cellular metabolism. HIV-
associated brain injury is characterized by decreased NAA/
Cr ratios in white matter and increased Cho/Cr and MI/Cr
ratios in white matter and basal ganglia (Lee et al. 2003).
Using MRS, our group identified three cerebral metabolite
patterns that were associated with the risk of HIV-associated
neurocognitive disorders (HAND): a neuronal pattern that
was associated with a threefold increased risk, a basal
ganglia pattern that was associated with a twofold increased
risk, and an inflammatory pattern that was weakly associated
with HAND (Yiannoutsos et al. 2004).
The primary objective of this study was to evaluate the
relationships between chemokine concentrations in CSF
and cerebral metabolites measured by MRS. We hypothe-
sized that chemokine concentrations would be associated
with the same pattern of cerebral metabolites that is
associated with HAND (i.e., lower neuronal pattern scores,
higher basal ganglia pattern scores, and higher inflamma-
tion pattern scores). We predicted that chemokines would
differ in their associations with cerebral metabolite patterns
and postulated that those which most closely reflected the
pattern associated with HAND may most strongly influence
HIV neuropathogenesis. To address these objectives, we
selected six chemokines [monocyte chemotactic protein
(MCP)-1, interferon-inducible protein (IP)-10, fractalkine
(FKN), macrophage inflammatory protein (MIP)-1β, stro-
mal derived factor (SDF)-1, and interleukin (IL)-8] that
have been implicated in HAND but differ by the types of
cells that produce them, the types of receptors they bind,
and the types of cells they attract.
Methods
Subject cohort
Demographic and disease-related data were obtained from
AIDS Clinical Trials Group (ACTG) protocol 700, which
evaluated the relationship between neurocognitive impair-
ment and cerebral metabolites, as measured by MRS. This
protocol provided a single MRS assessment from a group
of HIV-infected, neurologically unimpaired subjects, and
two repeated MRS evaluations from a group of neurolog-
ically impaired individuals enrolled in another ACTG
study, protocol 301, a randomized, placebo-controlled
16-week clinical trial of memantine for HAND. This latter
subject group had repeated MRS evaluations at visit week 0
(prior to initiation of study drug or placebo) and after
16 weeks of study drug or placebo. In all, there were 129
patients providing data for this study: 30 HIV-infected,
neurologically asymptomatic controls and 99 ACTG 301
co-enrolled subjects (48 randomized to the placebo group
and 51 to the memantine group; Table 1). The local
Institutional Review Boards approved the studies at all
participating sites. All subjects or their legal guardians gave
informed consent prior to screening and enrollment.
Participant assessments
HAND was assessed using the AIDS dementia complex
(ADC) staging approach according to the guidelines
published by Price and Brew (Price et al. 1991). Magnetic
resonance spectroscopy measured four cerebral metabolites
(NAA, Cho, MI, and Cr) in three regions of interest
64J. Neurovirol. (2011) 17:63–69

Page 3

(parietal cortex, white matter, and basal ganglia). For the
purposes of analysis, NAA, Cho, and MI levels were
normalized to creatine levels (i.e., NAA/Cr, Cho/Cr, and
MI/Cr ratios were calculated). Composite neuronal, basal
ganglia, and inflammatory pattern variables were calculated
according to the method of Yiannoutsos et al. (Yiannoutsos
et al. 2004):
Neuronal ¼ NAA=CrWMþ NAA=CrPC
BasalGanglia ¼ NAA=CrBGþ Cho=CrBG
Inflammatory ¼ MI=CrBGþ Cho=CrWMþ MI=CrWM
þ Cho=CrPCþ MI=CrPC
Participants who were eligible for this analysis had
consented to and successfully undergone lumbar puncture.
Cerebrospinal fluid (CSF) specimens were stored at −70°C
until they were assayed for six chemokines (FKN, IL-8, IP-
10, MCP-1, MIP-1β, and SDF-1) by enzyme-linked
immunosorbent assays. Concentrations were adjusted for
dilution and for the sensitivities of the assays, which were
30 (FKN), 10 (IL-8), 7.8 (IP-10), 62 (MCP-1), 94 (MIP-
1β), and 18 (SDF-1), all expressed in picograms per
milliliter. The volume of the 171 stored CSF samples
varied from 200 μL to more than 1 mL so the number of
chemokines assayed in each sample was determined by a
priori prioritization. As a result, the number of concentra-
tion values ranged from 53 to 171 across the different
chemokines (summarized in Table 2). Assay operators were
blinded to other study results.
After assay completion, data were combined with
demographic data, clinical staging, other disease-related
markers, cerebral metabolites, and treatment arm (when
applicable). The other disease-related markers included
HIV RNA levels, which were measured by RT-PCR in
plasma and CSF (Roche, Amplicor, nominal limit of
detection 50 copies/mL), and CD4+ lymphocyte counts,
which were measured by flow cytometry. The distribution
of MIP-1β concentrations was undetectable in 92% of
specimens, so it was eliminated from all analyses.
Statistical analyses
All results were considered statistically significant at the 5%
alpha level. No adjustment was performed to account for
multiple comparisons when, for example, up to 45 correlation
coefficients were calculated (three cerebral metabolite ratios
multiplied by three regions of interest multiplied by five
chemokines) so these results should be interpreted as
hypothesis generating. When estimating correlations between
measures which were collected repeatedly (at baseline and
week 16), we accounted for the within-subject correlation by
usingtheper-subjectmeanofeachmeasureandestimatingthe
correlation between the means of the measures rather than the
measures themselves. As some subjects had two and some
only had one observation, we followed the recommendation
byBlandandAltman(BlandandAltman1995) and weighted
the correlation analysis by the number of available measures
on each subject. A repeated-measures analysis of variance
was used to compare chemokine levels between subjects
with early versus late ADC stage while accounting for the
correlation of measurements obtained repeatedly within the
same subject.
Table 2 Descriptive statistics of chemokine data
Sample size Median IQR% detectable
FKN
IL-8
IP-10
MCP-1
MIP-1β
SDF-1
53
53
99
30.0
24.7
134.9
640.4
96.0
459.0
0.0, 60.0
10.0, 39.3
153.3, 386.6
438.5, 900.3
96.0, 96.0
266.0, 700.0
49
96
100
100171
171
123
8
97
Concentrations are in picograms per milliliter
IQR interquartile range
Table 1 Subject demographics and disease characteristics among the
N = 129 subjects involved in this study
Subject subgroup
HIV-infected controls
HIV-infected impaired subjects
Placebo
Memantine
Age at baseline (years)
Median (IQR)
HIV RNA, plasma (log10copies/mL)
Median (IQR)
HIV RNA, CSF (log10copies/mL)
Median (IQR)
CD4+ cell count (cells/μL)
Median (IQR)
Gender (male)
N (%)
Ethnicity (% white)
Antiretroviral use
N (%)
ADC stage
Normal (stage 0)
Subclinical (stage 1)
Stage 2
Stage 3
30
99
48
51
(23%)
(77%)
(37%)
(40%)
42(36–46)
2.2 (<50–4.2)
<50(<50–2.5)
274(168–428)
115
88
(89%)
(68%)
118 (91%)
30
83
14
(23%)
(64%)
(11%)
(2%)2
J. Neurovirol. (2011) 17:63–6965

Page 4

Results
Subject characteristics
Table 1 summarizes the demographic and disease character-
istics of the 129 subjects included in the analysis. Subjects
were mostly middle-aged, white men. Antiretroviral use
was common in the cohort, resulting in low HIV RNA
levels in most subjects. Most subjects exhibited at least a
mild degree of neurological impairment (77%).
Chemokine concentrations
Table 2 summarizes chemokine concentrations in CSF.
Almost all the measured chemokines were within the
detectable range in the majority of specimens. The principal
exception was MIP-1β, which was undetectable in nearly
all (92%) specimens. As a result, MIP-1β concentrations
were not included in subsequent analyses.
Relationships between chemokines and either clinical ADC
stage or individual cerebral metabolites
Higher levels of FKN were associated with worse clinical
ADC stage after accounting for the repeated nature of the
data at weeks 0 and 16 (p < 0.001). Using a Box–Cox
(logarithmic) transformation on the original FKN levels and
after the repeated-measures adjustment, the estimated mean
CSF FKN level was108 pg/mL (ADC stage ≥2) versus
34 pg/mL (ADC stage ≤1; Fig. 1). No other chemokine was
significantly different between mild and more severe ADC
patients.
Comparing chemokine concentrations to cerebral metabo-
lites, higher IL-8 levels correlated with higher MI/Cr levels in
theparietalcortex(Spearmanr = 0.43; p = 0.019). IP-10 levels
correlated with lower NAA/Cr ratios in the frontal white
matter (r = −0.43, p = 0.003) and higher MI/Cr ratios in the
basal ganglia (r = 0.36; p = 0.038) and the parietal cortex
(r = 0.44, p < 0.003). Higher MCP-1 levels were associated
with lower NAA/Cr ratios in frontal white matter (r = −0.27,
p = 0.023) and in the parietal cortex (r = −0.32, p = 0.006).
These results are summarized in Table 3. Levels of chemo-
kines, which were associated with cerebral metabolites at a
significance level between 5% and 10%, are also shown.
Relationships between chemokines and patterns of cerebral
metabolites
Figure 2 summarizes the results of the analyses of chemo-
kines and cerebral metabolite patterns. This analysis
demonstrated that higher levels of either IP-10 (r = −0.41,
p = 0.006) or MCP-1 (r = −0.31, p = 0.008) were inversely
correlated with the neuronal metabolite pattern and IL-
8 levels were positively correlated with the inflammatory
factor pattern (r = 0.42, p = 0.045).
To account for the use of medications that might
confound the relationships between chemokines and cere-
bral metabolites, we performed two subgroup analyses.
First, to account for the potentially confounding effect of
antiretroviral therapy (ART), we limited the analysis to only
those subjects who were taking no or failing ART, as
evidenced by having detectable HIV RNA levels in plasma.
In this subgroup, IL-8 no longer correlated with the
inflammatory pattern scores, MCP-1 no longer correlated
ADC stage <=1 ADC stage >1
0
25
50
75
100
125
150
Mean within-subject CSF FKN levels (pk/mL)
ADC status (Early versus Late stage)
Fig. 1 Box plot of mean within-subject CSF FKN levels between
HIV-infected subjects who were neurologically normal or had slight
neurological impairment (ADC stage ≤1) and patients with advanced
neurological impairment (ADC stage >1)
Table 3 Correlation table for chemokines and MRS metabolite ratios
FKNIL-8 IP-10MCP-1
Basal ganglia
NAA/Cr
Cho/Cr
MI/Cr
Frontal white matter
NAA/Cr
Cho/Cr
MI/Cr
Parietal cortex
NAA/Cr
Cho/Cr
MI/Cr
–
–
0.48
–
–
0.40
–
0.31
0.36
−0.24
–
–
–
–
–
–
–
−0.43
–
–
−0.27
–
–
–
–
0.42
–
–
0.43
−0.28
0.28
0.44
−0.32
–
–
Numbers are values of Spearman’s correlation coefficient between the
within-patient means at weeks 0 and 16
Statistically significant P values ≤0.05 are italicized. P values between
0.05 and 0.10 are shown in regular typeface. Cells with dashes
indicate P values >0.10. Levels of SDF-1 were not significantly
associated with any of the metabolite ratios and are not shown here.
All p values are unadjusted for multiple comparisons
66 J. Neurovirol. (2011) 17:63–69

Page 5

with the neuronal pattern score, and the relationship
between IP-10 levels and the neuronal pattern score was
attenuated (r = −0.29; p = 0.13). Together, these weakened
associations in individuals with detectable HIV RNA levels
suggest that ART influences the relationships between
chemokines and cerebral metabolites.
Next, to account for the potentially confounding effects
of memantine, we repeated our analysis in the subgroup of
subjects who were not taking memantine. In this subgroup
analysis, the correlation between IL-8 and the inflammatory
factor score appeared to be maintained (r = 0.50, p = 0.11)
while the correlation between IP-10 and the neuronal factor
score appeared to strengthen (r = −0.66; p = 0.001),
indicating that memantine use may mitigate IP-10-
associated neuronal injury. The correlation between MCP-
1 and the neuronal pattern score did not meet statistical
significance in this subgroup or the complementary
subgroup of individuals who took memantine.
Discussion
Our observations support several hypotheses about the
predominant in vivo effects of the study’s targeted chemo-
kines. Specifically, our analyses support that IP-10 and
MCP-1 are predominantly neurotoxic because they were
associated with evidence of neuronal loss (evidenced by
their negative correlations with NAA/Cr ratios and IP-10’s
correlation with neuronal factor scores). On the other hand,
the role of FKN is less clear as FKN levels were associated
with more severe ADC and with MRS evidence of
inflammation, but not with MRS evidence of neuronal
injury or loss. IL-8 levels were also associated with
inflammation, as evidenced by positive correlations with
inflammatory factor scores and MI/Cr ratios, but also had
no evidence of association with factors of neuronal loss.
SDF-1 was not associated with individual metabolites or
patterns of cerebral metabolites, and MIP-1β was below
detection in too great a proportion of subjects in this
heavily treated cohort to reliably assess its association with
cerebral metabolites.
Particularly noteworthy was the finding that IP-10 was
associated with the same pattern of cerebral metabolites
found in individuals with HAND, implicating interferon
pathways in HIV neuropathogenesis. In the brain, IP-10, or
CXCL10, is expressed by astrocytes and microglia and can
attract T lymphocytes and monocytes/macrophages (Taub
et al. 1993). Its expression can be induced by type I or type
II interferons (Neville et al. 1997), gp120 (Asensio et al.
2001), or tat (Kutsch et al. 2000). Its receptor, CXCR3, is
expressed on neurons but is not a primary co-receptor for
HIV cell entry. While some evidence supports that IP-10 may
be neuroprotective (Narumi et al. 2002), the preponderance of
published data support its neurotoxic properties. For example,
IP-10 can stimulate HIV replication (Lane et al. 2003), is
associated with accumulation of activated T cells in the CNS
(Kolb et al. 1999), is markedly upregulated in brains of
r=-0.41
p=0.006
r=-0.31
p=0.008
r=0.42
p=0.045
Fig. 2 Graphical depictions of the relationships between Box–Cox-
transformed MCP-1 (left panel) and IP-10 levels (middle panel) with
neuronal pattern scores and IL-8 levels with inflammatory pattern
scores (right)
J. Neurovirol. (2011) 17:63–69 67