Representative magnetic resonance spectroscopy spectra across varying degrees of neurocognitive impairment. T1-weighted images and magnetic resonance spectroscopy spectra from a representative control subject (A), a neurocognitively normal human immunodeficiency virus–positive patient (B), a mildly impaired human immunodeficiency virus–positive patient (C), and a moderately to severely impaired human immunodeficiency virus–positive patient (D). A multivoxel chemical shift imaging grid (outlined in white) is centered over subcortical gray matter for each of the subjects. Spectra are taken from voxels overlapping the lenticular nuclei (blue squares). The x-axis for all of the spectra ranged from 0.2 to 4.3 ppm. Spectra display the resonances of choline (Cho) at 3.20 ppm, creatine (Cr) at 3.02 ppm, N-acetyl aspartate (NAA) at 2.02 ppm, and lactate (Lac) and lipid (Lip) at 1.33 ppm.
Markers of inflammation and oxidative stress are elevated in human immunodeficiency virus–positive patients with varying degrees of neurocognitive impairment compared with controls. The data are from lenticular nuclei voxels, showing the ratio of lactate (Lac) to creatine (Cr) at an echo time of 135 milliseconds (A) and the ratio of Lac and lipid (Lip) to Cr at an echo time of 30 milliseconds (B). Error bars indicate SEM. *Results from the t test are statistically significant (P < .05).
Roc AC, Ances BM, Chawla S, Korczykowski M, Wolf RL, Kolson DL, Detre JA, Poptani H. Detection of Human Immunodeficiency Virus–Induced Inflammation and Oxidative Stress in Lenticular Nuclei With Magnetic Resonance Spectroscopy Despite Antiretroviral Therapy. Arch Neurol. 2007;64(9):1249–1257. doi:10.1001/archneur.64.9.noc60125
Single-voxel magnetic resonance spectroscopy measurements of N-acetyl aspartate, choline, and creatine (Cr) are affected in patients with human immunodeficiency virus (HIV) and neurocognitive impairment. However, these metabolic markers are often normalized in affected central nervous system regions, such as the lenticular nuclei, after initiation of highly active antiretroviral therapy (HAART).
To examine whether lactate (Lac), a marker of inflammation and anaerobic glycolysis, and lipid, an indicator of cell membrane turnover resulting from oxidative stress, could serve as surrogate biomarkers within the lenticular nuclei of HIV-positive patients with different degrees of neurocognitive impairment.
Three-tesla 2-dimensional–chemical shift imaging magnetic resonance spectroscopy at echo times of 30 milliseconds and 135 milliseconds was performed in voxels overlapping the lenticular nuclei of seronegative controls and a spectrum of HIV-positive patients (neurocognitively normal, mildly impaired, or moderately to severely impaired).
University of Pennsylvania, Philadelphia.
Ten seronegative controls and 45 HIV-positive patients with different degrees of neurocognitive impairment (15 neurocognitively normal patients, 12 mildly impaired patients, and 18 moderately to severely impaired patients).
Main Outcome Measures
In vivo 2-dimensional–chemical shift imaging magnetic resonance spectroscopy analysis of N-acetyl aspartate:Cr, choline:Cr, Lac:Cr, and (lipid + Lac):Cr ratios among the various groups. In addition, the effect of the degree of HAART central nervous system penetration (high vs low) on these ratios was studied.
No significant lenticular nuclei atrophy was detected with volumes similar across all of the groups. Both N-acetyl aspartate:Cr and choline:Cr ratios were similar across all of the groups at either echo time. In contrast, the Lac:Cr ratio was significantly greater in HIV-positive patients with moderate to severe impairment compared with seronegative controls. The (lipid + Lac):Cr ratio was significantly elevated within each HIV-positive subgroup compared with seronegative controls. Within HIV-positive patients receiving HAART, the degree of central nervous system penetration (high vs low) did not affect metabolic ratios.
As seen with 2-dimensional–chemical shift imaging magnetic resonance spectroscopy, HIV induces inflammation and oxidative stress in HIV-positive patients despite HAART. Lipid and Lac are more sensitive inflammatory biomarkers that may be used to differentiate HIV-positive subgroups. However, no significant difference in efficacy, as measured by metabolic ratios, exists for high– vs low–central nervous system–penetrating HAART.Published online July 9, 2007 (doi:10.1001/archneur.64.9.noc60125).
Human immunodeficiency virus (HIV) induces both structural and functional cerebral metabolic changes1- 4 that cause HIV-associated neurocognitive impairments (HNCIs). The HNCIs develop from host and viral factors produced by macrophages and glial cells rather than from secondary infection or malignancy.4 The lenticular nuclei (LN) are often affected, causing psychomotor retardation.5 Previous immunopathological studies of the LN have demonstrated a strong correlation between host and viral factors and severity of HNCI.6,7 A greater understanding of HIV-induced functional metabolic changes may be gleaned from in vivo neuroimaging.
Proton magnetic resonance spectroscopy (1H-MRS) provides a reliable, noninvasive, in vivo approach to assess HNCI-induced biochemical changes. Magnetic resonance spectroscopy has greater sensitivity than typical structural magnetic resonance imaging (MRI) to monitor disease severity.8,9 Frequently studied metabolites include N-acetyl aspartate (NAA), a marker of mature neurons and axons; choline-containing compounds (Cho), a marker of membrane turnover; and creatine (Cr), a marker of high-energy metabolism. Reduced levels of NAA along with increased levels of Cho occur in the frontal white matter and LN of neurocognitively normal HIV-positive patients and patients with HNCI.5,8- 16 Decreased NAA levels may reflect neuronal injury or loss, whereas elevated levels of Cho most likely indicate gliosis or membrane remodeling. The introduction of highly active antiretroviral therapy (HAART) normalizes both NAA:Cr and Cho:Cr ratios,17- 19 limiting their predictive abilities.
In addition to neurodegeneration and gliosis, inflammation and oxidative stress resulting from increased cell turnover are common in patients infected with HIV.20 Activated monocytes (ie, microglia and macrophages), whether infected or uninfected, invade the brain and induce cellular factors (eg, cytokines, Tat, gp120, chemokines, and nitric oxide), promoting a cascade of immune activators.1 Lactate (Lac) levels can be elevated in HIV-positive patients compared with seronegative controls; this most likely reflects increased anaerobic glycolytic demands of macrophages due to inflammation.12 Human immunodeficiency virus also causes oxidative stress and cell membrane turnover leading to increased lipid (Lip) peroxidation,21,22 which induces cellular dysfunction and neuronal cell death.23 Both Lip and Lac can be measured by MRS12 but have not been extensively studied in HIV-positive patients prescribed HAART.
Human immunodeficiency virus–induced biochemical changes have typically been examined at 1.5 T using single-voxel MRS. Single-voxel MRS provides a high signal to noise ratio within a large voxel. This may cause partial volume effects and decrease the specificity of detecting focal metabolite differences.24 The recent development of 2-dimensional (2-D)–chemical shift imaging (CSI) MRS offers the unique ability to acquire multiple contiguous higher-resolution voxels simultaneously. In this study, we used 2-D–CSI MRS at 3 T to evaluate subgroups of HIV-positive patients with different degrees of cognitive impairment (neurocognitively normal, mildly impaired, and moderately to severely impaired).25,26 The presence of Lip and Lac as biomarkers of inflammation and oxidative stress was compared with the presence of commonly measured metabolites. The impact of high- vs low-penetrating regimens on these biochemical markers was assessed in HIV-positive patients receiving HAART.
Forty-five HIV-seropositive patients and 10 healthy seronegative controls were recruited for neuroimaging studies using 1H-MRS. The HIV serological status was confirmed by positive Western blot results and/or plasma HIV. Baseline laboratory screening evaluations were performed within 2 weeks of imaging and included a basic metabolic panel, complete blood cell count, thyroid panel, syphilis testing, CD4 cell count, and plasma viral load.
Patients who had been infected with HIV for more than 2 years were recruited using the following inclusion criteria: history of CD4 cell count less than 500/μL; negative urine toxicology screen results (ie, for cocaine, amphetamine, marijuana, benzodiazepine, barbiturates, and opiates); no other chronic medical or psychiatric illnesses; MRI results without major structural abnormalities (eg, cortical infarcts, arteriovenous malformations, or tumors); absence of head trauma with loss of consciousness for more than 30 minutes; lack of seizure disorders, hypertension, and diabetes; and no recent opportunistic infections or inflammation within the LN. All of the patients receiving HAART had been adhering to particular regimens for at least 2 months prior to their scan. Healthy age- and education-matched controls were also recruited. Controls were not receiving medications and did not have a history of substance abuse. Prior to enrollment, all of the subjects were verbally informed of the study protocol and informed written consent was obtained. The study protocol was approved by the Human Subjects Institutional Review Board at the Hospital of the University of Pennsylvania, Philadelphia.
The HIV-positive patients were classified into 1 of 3 groups of neurocognitive impairment (neurocognitively normal, mildly impaired, or moderately to severely impaired) according to recently proposed classification schemes27 that use neurobehavioral performance assessment scores.28,29 According to this scheme, a classification of mildly impaired would include those subjects who have been recently referred to as having asymptomatic neurocognitive impairment or mild neurocognitive impairment. A classification of moderately to severely impaired would include those subjects often referred to as having HIV-associated dementia.
Participants completed neuropsychological tests that assessed verbal memory and recall, psychomotor skills, motor skills and praxis, and executive functioning. Raw scores were converted to T scores with corrections applied for age, sex, and education. The T scores from each test were averaged to derive an overall global deficit score.28,29 Patients were classified as neurocognitively normal (global deficit score < 0.500), mildly impaired (0.500 ≤ global deficit score < 1.125), or moderately to severely impaired (global deficit score ≥ 1.125).30 A board-certified neurologist (B.M.A.) performed a detailed standardized neurological examination of cranial nerve function, motor strength and coordination, reflexes, gait, and sensation.27
Patients and controls underwent conventional 1H-MRS using a 3-T Trio scanner (Siemens, Erlangen, Germany) equipped with a standard quadrature head coil provided by the manufacturer. The imaging protocol consisted of the following: (1) a localizer fast low-angle shot sequence (TurboFLASH; repetition time, 20 milliseconds; echo time [TE], 5 milliseconds; matrix size, 144 × 192; 3 slices; 9.6-mm slice thickness; 9-second acquisition time); (2) an axial 3-dimensional T1-weighted sequence (MPRAGE) to identify anatomical structures (repetition time, 1620 milliseconds; TE, 3.9 milliseconds; inversion time, 950 milliseconds; matrix size, 192 × 256; 160 slices; 1-mm slice thickness; 5-minute acquisition time); and (3) two spin-echo 2-D–phase-encoding–CSI MRS sequences using a TE of 135 milliseconds and another with a TE of 30 milliseconds (repetition time, 2000 milliseconds; matrix size, 16 × 16; field of view, 20 × 20 cm; 20-mm slice thickness; 8-minute approximate acquisition time for each sequence).
For 2-D–CSI MRS, the axial slice was acquired from the deep gray matter that included the LN. Nominal voxel size using the CSI grid was 1.25 × 1.25 × 2.00 cm, which is smaller than conventional voxel-of-interest analysis using single-voxel MRS. Eight outer-volume saturation pulses (50-mm thickness) were applied for lipid suppression outside and immediately adjacent to the voxel of interest. A long TE of 135 milliseconds was used to detect the presence of Lac at 1.3 ppm, as it appears as an inverted doublet owing to J modulation, whereas a TE of 30 milliseconds was used to observe metabolites with short T2 relaxation times, including Lip and Lac.31 For MRS scans, automatic prescanning was conducted to achieve water suppression and optimal full-width half-maximum values of water resonance.
Additional imaging sequences were acquired after the T1-MPRAGE sequence and prior to MRS sequences with the total scan time per patient of approximately 45 to 50 minutes. Unfortunately, some HIV-positive patients could not complete all of the MRS scans. In 2 patients, one classified as mildly impaired and the other as moderately to severely impaired, MRS at a TE of 135 milliseconds could not be obtained owing to withdrawal during the scan. Additionally, 5 neurocognitively normal, 3 mildly impaired, and 6 moderately to severely impaired HIV-positive patients withdrew during the scan with a TE of 30 milliseconds.
The LN, which includes the putamen and globus pallidus, was chosen because it is often affected by HIV.32 The caudate was not included owing to its proximity to the ventricles. Other regions such as the frontal lobe were not chosen as a result of slice prescription. Atrophy has previously been observed with advancing HNCI.18 To determine whether HIV-induced LN atrophy was present, a manually segmented region of interest was drawn in each hemisphere using high-resolution T1-weighted images. A neuroradiologist (R.L.W.) who was blinded to the status of the subject verified the accuracy of segmentation and ensured a lack of signal abnormalities within the LN. The total number of voxels within the LN was expressed in millimeters.3
Multivoxel CSI spectra were analyzed on a Leonardo workstation using the Syngo software (Siemens). A prominent NAA resonance centered at 2.02 ppm was used as an internal chemical shift reference. For each spectrum in the CSI field of view, the acquired MRS signal (free induction decay) was zero filled (2048 data points), smoothed (Hanning filter; width, 200 milliseconds), and Fourier transformed. This was followed by phase (zero- and first-order polynomial) and baseline correction for optimal linear frequency dependence. The area under the peak for each resonance was measured for relative quantification by fitting each peak with a standard gaussian function. At a TE of 30 milliseconds, a signal at 1.33 ppm was assigned to Lip and Lac (Lip + Lac) (Figure 1B). At a TE of 135 milliseconds, an inverted peak at 1.33 ppm was defined as Lac (Figure 1D). The presence of Lip + Lac and Lac peaks were considered significant if a peak greater than 1 SD above the noise was observed.31 Noise levels were measured in the clear region of the spectrum at 0.0 to 0.5 and 8.0 to 8.5 ppm. Metabolic ratios for NAA:Cr, Cho:Cr, (Lip + Lac):Cr, and Lac:Cr were computed.
The HAART central nervous system (CNS) penetration scores were characterized using a hierarchical approach based on the best available evidence. Available data on chemical characteristics, cerebral spinal fluid (CSF) pharmacology, and neuroeffectiveness in the CNS were reviewed for Food and Drug Administration–approved antiretroviral drugs (ARVs) using package inserts, drug references, published articles, and conference abstracts. Using this approach, ARVs were classified into 3 categories.33 An ARV was considered to be in the lowest-penetration category if it met the following criteria: (1) its chemical properties supported relatively poor penetration into the CNS; (2) its CSF concentrations either were unmeasurable in human or animal studies or were measurable but consistently did not exceed the population median inhibitory concentration (IC50) for wild-type HIV as defined by Monogram Biosciences (South San Francisco, California); or (3) clinical studies demonstrated its ineffectiveness in reducing CSF viral load or improving cognition. An ARV was considered to be in the high-penetration category if it met the following criteria: (1) its molecular and pharmacological properties supported relatively high penetration into the CNS; (2) its concentrations in CSF were measurable in human or animal studies and consistently exceeded the population IC50 for wild-type HIV; or (3) clinical studies demonstrated its effectiveness in reducing CSF viral load or improving cognition. A third, intermediate category was assigned if the following criteria were met: (1) clinical studies did not consistently detect the drug in CSF; (2) its concentrations were measurable but did not consistently exceed the population IC50; or (3) the chemical properties did not clearly support penetration. These rules were applied in a hierarchical manner: clinical effectiveness studies were considered the strongest evidence, pharmacokinetic studies were considered the next strongest evidence, and chemical data were considered in the absence of the other 2 types of evidence.34 Using such empirical data, Table 1 summarizes categorization for individual ARVs. For the purposes of categorization, protease inhibitors that were boosted with ritonavir were considered a single drug distinct from the unboosted parent protease inhibitor. Individual ARVs were assigned a score based on penetration category (0 = lowest penetration, 0.5 = intermediate penetration, 1 = highest penetration). The CNS penetration effectiveness (CPE) score was then determined by summing the individual penetration scores for each ARV in a HAART regimen. Patients receiving HAART were then classified into either high– or low–CNS-penetration groups, with the high-penetration group having a CPE score greater than 1.5 and the low-penetration group having a CPE of 1.5 or less.
Four voxels from the CSI grid that completely overlapped the LN (2 voxels per hemisphere) were used in the evaluation. Metabolic ratios were averaged from each hemisphere as well as the combination from both hemispheres for each subject. For some cases (10% [1 of 10] of the controls, 20% [3 of 15] of the neurocognitively normal HIV-positive patients, 17% [2 of 12] of the mildly impaired HIV-positive patients, and 22% [4 of 18] of the moderately to severely impaired HIV-positive patients), data were pooled from only 2 or 3 voxels as a result of poor spectral resolution (ie, full width at half maximum > 10 Hz) or a low signal to noise ratio such that the signal could not be distinguished from the noise.
Metabolite ratios and LN volumes were compared across groups (controls, neurocognitively normal patients, mildly impaired patients, and moderately to severely impaired patients) by 1-way analysis of variance (ANOVA) (P < .05). If results from an ANOVA were significant, a post hoc unpaired t test (P < .05) was performed. If normality and equal variance tests failed, then an ANOVA on ranks was performed followed by multiple comparisons using a Mann-Whitney rank sum test (P < .05). Additionally, correlations were determined between commonly measured laboratory values and various metabolic ratios across all of the groups. To assess the impact of possible differences in HAART CNS penetration, t tests were performed comparing various metabolic ratios for HIV-positive patients on either high– or low–CNS-penetrating regimens.
Subject demographics and laboratory values are provided for controls (n = 10) and for neurocognitively normal (n = 15), mildly impaired (n = 12), and moderately to severely impaired (n = 18) HIV-positive patients (Table 2). There were no significant differences in age, sex, or years of formal education for the various classifications. Neither CD4 cell counts nor logarithmic plasma viral loads significantly differed among the HIV-positive subgroups. All of the HIV-positive patients were not naïve to HAART, with many (60% [9 of 15] of the neurocognitively normal patients, 75% [9 of 12] of the mildly impaired patients, and 78% [14 of 18] of the moderately to severely impaired patients) adhering to stable regimens at the time of the scan.
To determine whether HIV-induced atrophy was present, LN volumes were obtained for each subject. Within bilateral LN, mean ± SEM volumes were 8400 ± 305 mm3 for healthy controls (n = 10), 8291 ± 351 mm3 for neurocognitively normal patients (n = 15), 7931 ± 346 mm3 for mildly impaired patients (n = 12), and 7824 ± 365 mm3 for moderately to severely impaired patients (n = 18). Although the mean LN volume became progressively smaller for increasing degrees of impairment, these differences were not significant across groups (ANOVA, P = .42), with subsequent multiple comparisons of means showing similar results. Because no significant differences were observed, atrophy corrections were not performed.
Group sizes at a TE of 135 milliseconds included 10 healthy controls, 15 neurocognitively normal patients, 11 mildly impaired patients, and 17 moderately to severely impaired patients. At a TE of 135 milliseconds, no significant differences were observed across various subgroups for NAA:Cr ratios (Table 3) (ANOVA, P = .98). Similarly, Cho:Cr ratios at a TE of 135 milliseconds were comparable across all of the groups, with no significant differences observed (Table 3) (ANOVA, P = .24).
Group sizes at a TE of 30 milliseconds included 10 healthy controls, 10 neurocognitively normal patients, 9 mildly impaired patients, and 12 moderately to severely impaired patients. At a TE of 30 milliseconds, no significant differences were observed across the various subgroups for NAA:Cr ratios (Table 3) (ANOVA, P = .18). The Cho:Cr ratios at a TE of 30 milliseconds were similar across all of the 3 groups, with no significant differences observed (Table 3) (ANOVA, P = .54).
At a TE of 135 milliseconds, 10% (1 of 10) of the controls, 40% (6 of 15) of the neurocognitively normal patients, 36% (4 of 11) of the mildly impaired patients, and 59% (10 of 17) of the moderately to severely impaired patients had a Lac peak present at 1.33 ppm. At a TE of 30 milliseconds, 10% (1 of 10) of the controls, 50% (5 of 10) of the neurocognitively normal patients, 67% (6 of 9) of the mildly impaired patients, and 75% (8 of 12) of the moderately to severely impaired patients had a peak at 1.33 ppm indicating Lip and Lac. At a TE of 135 milliseconds, the magnitude of the Lac:Cr ratio was significantly greater within moderately to severely impaired HIV-positive patients compared with seronegative controls (t test, P = .03) (Figure 2A). At a TE of 30 milliseconds, the magnitude of the (Lip + Lac):Cr ratio was significantly greater in all of the HIV-positive subgroups compared with seronegative controls (t test, P = .02 for neurocognitively normal patients, P = .01 for mildly impaired patients, and P = .004 for moderately to severely impaired patients all compared with seronegative controls) (Figure 2B). However, no significant differences were seen among the various HIV-positive subgroups (P = .24).
We assessed whether standard markers of HIV disease progression (CD4 cell count and logarithmic plasma viral load near the time of the scan) correlated with ratios of Lac:Cr at a TE of 135 milliseconds or (Lip + Lac):Cr at a TE of 30 milliseconds. No significant correlation was evident between each metabolic marker and measured laboratory values (data not shown). In addition, because the LN is proposed to be involved in results on individual neuropsychological performance tests such as the Grooved Pegboard Test and Trail-Making Test B, we assessed whether T scores for these test results correlated with the Lac:Cr ratio at a TE of 135 milliseconds or the (Lip + Lac):Cr ratio at a TE of 30 milliseconds.35 Results on both of these neuropsychological performance tests were not significantly correlated with the levels of metabolic markers (data not shown).
To further assess the effects of HAART on neurometabolic markers, we analyzed the subset of HIV-positive patients on stable regimens within each of the subgroups. Among the 9 neurocognitively normal HIV-positive patients receiving HAART, 5 were receiving highly penetrating regimens, whereas 4 were receiving low-penetrating regimens. Among the 9 mildly impaired HIV-positive patients receiving HAART, 4 were receiving regimens with high CPE scores, whereas 5 were receiving regimens with low CPE scores. Finally, among the 14 moderately to severely impaired HIV-positive patients receiving HAART, 8 were receiving highly penetrating regimens and 6 were receiving low-penetrating regimens. Usable spectra could not be obtained for all of the patients on either high– or low–CNS-penetrating regimens because some HIV-positive subgroups comprised only 2 or 3 patients. The calculated power to detect a significant difference between high- and low-penetrating drugs across the HIV-positive subgroups ranged from a low of 0.06 to a high of only 0.40, confirming that the power would indeed be too low for 2-way ANOVAs. Therefore, we combined HIV-positive patients across the various subgroups to determine whether differences in the degree of HAART CNS penetration impacted metabolic ratios. No significant differences were observed for each of the metabolic ratios at either TE between the HIV subgroups on high- vs low-penetrating regimens (Table 4).
Using 2-D–CSI MRS, we demonstrated the presence of Lip and Lac, markers of oxidative stress and anaerobic glycolysis, within the LN of HIV-positive patients. We observed significant anaerobic glycolysis and inflammation only in moderately to severely impaired HIV-positive patients as noted by the presence of an elevated Lac:Cr ratio at a TE of 135 milliseconds. The presence of an elevated (Lip + Lac):Cr ratio at a TE of 30 milliseconds but no increase in the Lac:Cr ratio within neurocognitively normal and mildly impaired HIV-positive patients suggest a primary increase in lipid peroxidation attributable to oxidative stress.
Although moderately to severely impaired HIV-positive patients had smaller LN volumes, no significant differences were observed across the groups. In the past, subcortical atrophy has primarily been reported in the caudate30,32,36,37 but not in LN.32 Metabolic markers obtained by MRS, rather than structural MRI changes, may provide an additional method for differentiating the degree of HNCI.
Prior studies at 1.5 T have reported reduced NAA:Cr ratios and increased Cho:Cr ratios in subcortical gray matter of patients with HNCI,15,18,38- 40 consistent with both neuronal loss and increased inflammation. However, our study revealed relatively similar NAA:Cr and Cho:Cr ratios in the LN for all of the HIV-positive subgroups compared with controls. These results might reflect the positive contribution of HAART on MRS metabolites. All of the patients in our study either were receiving stable HAART regimens at the time of the scan or had been previously exposed to medications. Reversals in metabolic changes have in fact been observed after introduction of ARVs.15,18,41,42 Although NAA and Cho may be used as biomarkers prior to starting medications, the predictive value of these metabolites in differentiating HIV-positive subgroups may be significantly diminished once patients initiate HAART.
As markers of active inflammation and oxidative stress, Lip and Lac peaks have been reported in HIV-positive patients with opportunistic infections such as toxoplasmosis, progressive multifocal leukoencephalopathy,14,43 epilepsy,44 or stroke.45 Previous studies have demonstrated that HIV can induce rapid sphingolipid cell membrane turnover as well as increase in de novo synthesis and hydrolysis of lipids into secondary messengers such as ceramide and 4-hydroxynonenal.21,22 The presence of elevated (Lip + Lac):Cr ratios in all of the HIV-positive subgroups suggests that oxidative stress may occur early and persist throughout as HNCI progresses.46,47 Continued lipid peroxidation may also suggest the presence of latent HIV reservoirs inducing the irreversible initiation of neuronal damage despite overt changes in neurocognitive impairment as determined by neuropsychological assessments. Overall, our results suggest that HIV may be similar to Alzheimer and Parkinson diseases48,49 in that extended periods of oxidative stress may lead to continued neuronal damage and eventual cognitive decline.21,22
The presence of Lac may reflect anaerobic glycolysis and cellular toxicity due to activated macrophages.14 In contrast to combined Lip and Lac results for a TE of 30 milliseconds, Lac levels were only elevated in moderately to severely impaired HIV-positive patients. The elevated Lac levels may possibly result from lipid peroxidation, as disturbances in the integrity of cellular membranes can lead to leakage of cytoplasmic enzymes such as lactate dehydrogenase.50 These results suggest an eventual “burned-out” state with irreversible initiation of apoptotic signaling resulting in increases in anaerobic metabolism seen at the end degrees of HNCI.
Highly active antiretroviral therapy may limit the progression and/or severity of blood-brain barrier dysfunction by inhibiting the pathogenic pathways required for the continued presence and production of HIV and HIV proteins.51- 53 In fact, HAART can induce systemic lipodystrophy54 and could theoretically reduce Lip and Lac levels. However, the observance of these metabolites argues for the continued presence of a viral sanctuary that is inducing inflammation.35 Despite a reduction in viral replication by HAART, impaired reuptake of neurotransmitters (such as glutamate) may stimulate the release of cytokines, ceramide, 4-hydroxynonenal, interleukins, nitric oxide, and chemokines, promoting a cascade of inflammation.20 Neuroprotective and anti-inflammatory approaches that inhibit sphingolipid breakdown and subsequent second messenger signaling may provide novel therapeutic targets.21
Attainment of virological suppression in the CNS by HAART could theoretically be important for 2 reasons. First, viral inhibition in the CNS may protect against additional neural injury due to increased viral burden and thereby may ameliorate neurocognitive impairments. Second, HAART-induced viral suppression in the CNS may forestall compartmentalized evolution of medication resistance. Similar to previously published evidence that the penetration of ARVs into brain tissue may not be sufficient to control viral replication in infected CNS target cells,55,56 we observed that MRS markers of oxidative stress and inflammation remained elevated in HIV-positive patients on either high– or low–CNS-targeted HAART regimens. The results are similar to recent findings showing that no significant differences in brain viral load reduction were observed when comparing subjects receiving 2 or more CNS-penetrating drugs with those receiving only 1 or none.33 Our findings imply that CNS-targeted HAART does not substantially inhibit oxidative stress and inflammation. Additional neuroprotective medications should be considered in the treatment of HIV-positive patients.
A potential limitation of this study is that Lac and Lip cannot be reliably distinguished from each other at short TEs (30 milliseconds). Additional sequences that can selectively isolate Lac from Lip resonances exist.57 Availability of these sequences could be extremely beneficial in assessing the chronic inflammation and oxidative stress that are present in HIV-positive patients and may allow for the assessment of additional neuroprotective agents. However, the application of these pulse sequences on existing scanners is not trivial. On the other hand, the MRS techniques used in this study have general applicability, as they are included within standard packages accompanying clinical MRI units. Another potential limitation of the current study is the use of metabolite ratios rather than absolute concentrations for the data analysis. Absolute quantification of spectroscopic data is not insignificant, as knowledge of T1 and T2 relaxation times leads to prohibitive scan times. Consequently, metabolite ratios were used in this study to evaluate spectroscopic data.10- 13,15,16,40,58 Furthermore, myoinositol could not be consistently evaluated in our cohort as a result of suboptimal water suppression in some cases, which yielded a residual water peak overlapping the myoinositol resonance. In previous studies of HIV-positive patients, myoinositol has been reported as a nonspecific marker of inflammatory processes or gliosis in the frontal white matter49 and basal ganglia.59 In addition to Lip and Lac as markers of inflammatory processes,14,47,49 it is possible that myoinositol levels may also be elevated in our cohort of patients, reflecting glial cell turnover or osmoregulation in glial cells.59 Finally, it should be noted that our study was cross-sectional in nature and the observed group differences may be caused by unknown factors specific to the patient populations. Variations in the efficacy of ARVs to cross the blood-brain barrier may explain the wide variability in CD4 cell, plasma, and viral load levels observed in our cohort of HIV-positive patients. Additional longitudinal studies that follow Lip and Lac levels prior to and after the initiation of HAART are required.
We have demonstrated that Lip and Lac are present more often in HIV-positive patients than in controls. These results suggest that HIV-induced oxidative stress and inflammation occur even after initiation of HAART. No significant differences were seen between high– and low–CNS-penetrating HAART regimens. Compared with typical markers such as NAA and Cho, Lip and Lac may provide additional measures for differentiating HIV-positive subgroups with varying degrees of neurocognitive impairment.
Correspondence: Harish Poptani, PhD, Department of Radiology, University of Pennsylvania, B6 Blockley Hall, 423 Guardian Dr, Philadelphia, PA 19104 (email@example.com).
Accepted for Publication: September 26, 2006.
Published Online: July 9, 2007 (doi:10.1001/archneur.64.9.noc60125).
Author Contributions: Drs Roc and Ances contributed equally to this work. Study concept and design: Roc, Ances, Detre, and Poptani. Acquisition of data: Roc, Ances, and Kolson. Analysis and interpretation of data: Roc, Ances, Chawla, Korczykowski, Wolf, and Poptani. Drafting of the manuscript: Roc, Ances, and Chawla. Critical revision of the manuscript for important intellectual content: Roc, Ances, Korczykowski, Wolf, Kolson, Detre, and Poptani. Statistical analysis: Roc and Ances. Obtained funding: Ances and Detre. Administrative, technical, and material support: Roc, Ances, Korczykowski, and Kolson. Study supervision: Ances, Kolson, Detre, and Poptani.
Financial Disclosure: None reported.
Funding/Support: This work was supported by grants AI045008 from the Center for AIDS Research (A.C.R., B.M.A., D.L.K., and J.A.D.), NS045839 from the Center for Functional Neuroimaging at the University of Pennsylvania (A.C.R. and M.K.), NIH AI 32783 from the University of Pennsylvania AIDS Clinical Trials Unit (A.C.R., B.M.A., D.L.K., and J.A.D.), CF05-SD-301 from the Universitywide AIDS Research Program (B.M.A.), and 106729-40-RFRL from the American Foundation for AIDS Research (B.M.A.).
Role of the Sponsors: No funding sources had a role in study design; collection, analysis, and interpretation of data; writing of the report; or the decision to submit the manuscript for publication.
Additional Information: This study was a collaborative effort between the Center for Functional Neuroimaging and the Departments of Neurology and Radiology at the Hospital of the University of Pennsylvania.
Additional Contributions: Norman Butler, BA, Doris Cain, BA, and Tanya Kurtz, BA, provided assistance. Ron Ellis, MD, PhD, Brian Schweinsberg, PhD, Mariana Cherner, PhD, and Christopher Ake, PhD, provided helpful comments on the manuscript. The reviewers provided insightful suggestions.