The propensity of human immunodeficiency virus (HIV) to primarily affect the CNS is well documented (1-5). However, imaging studies have only been able to demonstrate the characteristic magnetic (MR) findings of atrophy and white matter signal abnormalities relatively late in the course of the disease (6,7). Because there is evidence that HIV does enter the CNS early in the course of infection (5,8,9), there may be metabolic changes that occur on the cellular level that are too subtle to detect with conventional imaging. Because MR spectroscopy (MRS) may be able to detect alterations in the biochemistry of tissues, we hoped that it would prove to be more sensitive at detecting early CNS involvement by HIV than conventional magnetic resonance imaging (MRI).
Previously, we and others have shown that proton MRS is able to detect abnormalities in the brains of HIV-infected patients attributable to their primary HIV infection (10-13). Investigators have also demonstrated metabolic abnormalities in symptomatic patients who have either normal MRI scans or in regions of normal MRI imaging appearance (14,15). Several studies have been unable to demonstrate MRS abnormalities in asymptomatic HIV-infected patients (10,11,16). Although their inability may be due to an inherent limitation of MRS, it may also reflect limitations that are technical in nature. The goal of our study was to determine whether, by slightly modifying MRS, it would be capable of detecting early CNS involvement by HIV.
MATERIALS AND METHODS
We recruited nine patients with known HIV infection from a clinic that specialized in the treatment of drug abuse. These patients were all undergoing neuropsychological testing as part of another study. We required our cases to have laboratory evidence of HIV infection as well as absence of neuropsychological symptoms. Because all nine of these patients had the confounding history of drug abuse, we also studied 10 HIV-infected patients without a history of drug abuse. These additional patients were randomly chosen from a larger cohort of neurologically asymptomatic HIV-infected patients who were enrolled in a longitudinal study. Ten controls who were presumed to be HIV seronegative were recruited from the Radiology Department. None of the controls had risk factors for HIV or demonstrated clinical evidence of disease either at the time of examination or after 2 years of follow-up.
Potential subjects were excluded for the following reasons: (a) Medically unstable or hematological, renal, or hepatic dysfunction; (b) cardiac pacemarker; (c) intracranial clips, metal implants, or external clips within 10 mm of the head; (d) intraocular metal; (e) pregnant or nursing; (f) claustrophobia; (g) abnormal MR scan at initial evaluation; (h) neuropsychological symptoms at initial evaluation.
Clinical and Neuropsychological Evaluation
All patients received imaging, spectroscopy, and neurological examination on the same day. Either a board-certified or boardeligible neurologist evaluated the patient's mental status, cranial nerves, motor and sensory function, coordination, and deep tendon reflexes.
The neuropsychological testing was conducted as part of an ongoing multidisciplinary longitudinal study that included cognitive assessments at 6-month intervals. A comprehensive neuropsychological battery including measures of abstraction and mental flexibility, verbal cognition, spatial-constructional ability, verbal and visual memory and learning, expressive and receptive language, attention and speeded visual processing, and fine motor speed was administered to all subjects who participated in the MRS study. The test battery was based on the used in prior studies of HIV-infected individuals, and detailed methods and test references are available in earlier reports (17,18). This evaluation coincided with the 1-year-interval follow-up. Neuropsychological testing was usually within a week of the study.
Imaging and Spectroscopy Protocols
Before spectroscopy, each patient underwent an imaging protocol consisting of the following: (a) sagittal repetition time (TR) 600 ms/echo time (TE) 12 ms, 5 mm thick, matrix 192, NEX 1, field of view (FOV) 22 cm; (b) axial fast spin-echo TR 2,500 ms, Te 18/90 ms, 3 mm thick, matrix 192, 1 repetition, field of view 22 cm.
Subsequently, solvent-suppressed proton spectra were acquired using the point resolved spectroscopy (PRESS) sequence. An attempt was made to obtain bilateral voxels of interest (VOIs) centered in the middle of the centrum semiovale in each subject; however, not all subjects were cooperative enough to complete both voxels. The parameters for the spectra were as follows: TE 16, 1,000-Hz sweepwidth, 2K points, 2-s repetition time, eight-step phase cycle, 128 averages per spectrum, and voxel size 1.5 × 1.5 × 1.5 cm. Acquisition time for the VOIs was ≈8.5 min, with the total duration of the MR spectroscopic examination being ≈45 min.
Each spectrum was processed individually with ProNMR (Softpulse Software, Guelph, Ontario, Canada). Spectroscopic data were phased by one of the authors (R.E.L.), who was blind to the serostatus of the subject. Quantitative analysis of the spectra was then performed using an automated frequency domain software program that first fitted peaks from the spectrum and then calculated the area under the curves for all peaks. The solvent suppression of each case was between 500 and 1,000, which was sufficient to make baseline correction unnecessary. The following peaks were readily identified on all spectra (Fig. 1A): N-acetyl aspartate (NAA) at 2.0 ppm, creatine (Cr) at 3.0 ppm, and choline (Cho) at 3.2 ppm. Additionally, we calculated the area of the “marker” peak, a region that we had previously identified between 2.1 and 2.6 ppm (13). This conglomerate peak consists of the resonances for glutamate, glutamine, and γ-aminobutyric acid, which are not resolvable at 1.5 T.
Laboratory tests obtained on the patients included an HIV serum antibody test, and CD4 and CD8 cell counts. Only patients, and not controls, had these blood tests performed. The study protocol was approved by the institutional review board and all of the patients were enrolled after obtaining informed consent.
We used the Mann-Whitney U-Wilcoxon rank sum W test to calculate the significance of the difference between patients and controls for various metabolite ratios as well as age. We used the same test to examine CD4 counts with respect to gender. We used stepwise multiple linear regression to develop equations relating CD4 counts and the mean neuropsychological test scores to metabolite ratios (SPSS version 6.0 for Macintosh, 1995).
Our 19 HIV-infected subjects ranged in age from 34 to 48 years, with a mean age and SD of 39.1 years ± 7.2 (Note: All errors are SDs.) There was no statistically significant difference in the mean age of patients with and without a history of intravenous drug use (IDU) (39.9 ± 4.7 versus 38.4 ± 9.9, respectively). Our 10 control subjects had a mean age of 34.2 ± 6.2 years, with a range of 25-47 years. This difference in mean ages between HIV-infected subjects and controls approached statistical significance (p = 0.05). Fifteen of the patients (79%) and six of the controls (60%) were male. All of the patients and controls had normal MR imaging. Specifically, there was no evidence of atrophy or white matter abnormalities.
We obtained 33 spectra from the 19 HIV-infected patients, 14 patients having had bilateral spectra, and 10 spectra from 10 control subjects. When HIV-infected subjects had bilateral voxels obtained, they were combined by using the mean values of metabolite ratios. This combination assumes that HIV is primarily a diffuse process. This is a conservative assumption, because if the focal effects of HIV are more important, this analysis would tend to minimize the differences between the patients and controls. We did a second comparison using the individual voxels as the unit of analysis, which did not differ significantly from the primary analysis. Table 1 lists the metabolite ratios for HIV-infected subjects and controls as well as mean neuropsychology test scores and CD4 counts in HIV-infected subjects.
Figure 2 compares the metabolite ratios of HIV-infected subjects with controls. The marker/Cr ratio was significantly different between patients and controls. This difference in marker peak area is illustrated in Fig. 1B, the proton spectrum of a patient, when compared with Figure 1a, the proton spectrum of a control subject. If one uses a cutoff of 1 SD above the mean of controls as the upper limit of normal for metabolite ratios, then the marker/Cr ratio has a sensitivity of 16/19 = 84%, a specificity of 90%.
To account for the possibility of drug abuse being responsible for the difference of marker peak/Cr between HIV-infected subjects and controls, we compared the metabolite ratios of those HIV-infected subjects with a history of drug abuse to those without such a history. There were no significant differences in the metabolite ratios between these two groups. Specifically, the marker/Cr ratio was nearly identical (1.8 ± 0.85 versus 1.6 ± 0.36, p = 0.96). In contrast, the marker/Cr ratio was significantly different between the HIV-infected patients without a history of drug abuse and controls (1.6 ± 0.35 versus 0.49 ± 0.51, p = 0.0005).
We examined the relationship between the two continuous measures, CD4 count, and mean neuropsychological score, and the metabolite ratios by building linear regression models. The CD4 count for one of our patients was an outlier (with a value of 1,276, more than 2 SD above the mean), and our first model, which contained this patient, showed NAA/Cr and age to be significant predictors of CD4 count. When the outlier was removed and the regression repeated, gender was the only significant predictor. The mean CD4 count for males was 204, whereas for females it was 442. This difference was significant (p = 0.02), indicating that the men in our cohort seemed to have more severe disease than the women.
We obtained similar results when building a linear regression model for the mean neuropsychological test score. None of the metabolite ratios were significant predictors of this variable.
Two important conclusions emerge from our study. First, despite our small sample size, we demonstrated a significant difference between the metabolite ratios of asymptomatic HIV-infected patients and controls. The best discriminator between cases and controls was the marker peak to creatine ratio. The NAA/Cr ratio did not differ significantly between patients and controls. The use of short echo times is crucial for the detection of the marker peak. Prior studies of asymptomatic HIV-infected patients used long echo times, which failed to detect the marker peak resonances (10,11,16). This technical difference may account for the discrepancy between our results and other studies. Our results are highly encouraging regarding the ability of MRS to detect involvement of the CNS by HIV early in the course of disease. Second, we were able to exclude IDU as a confounder for the metabolite changes that are present in HIV-infected patients. The metabolite changes that we observed in our HIV-infected cohort were present regardless of whether or not the subject had used intravenous drugs.
Even though the mean value of the marker peak/Cr ratio differs between patients and controls, we could not demonstrate its ability to predict either CD4 count or neuropsychological score. Although such a significant relationship could still exist, but because of our small sample size, our study lacked the power to demonstrate all but the largest relationships. Another factor is possibly related to the colinearity that existed in our sample between gender, marker/Cr and CD4 count. If variables are closely associated in a linear fashion, then it is difficult to demonstrate the significance of the individual regression coefficients (19).
The lack of difference in the NAA/Cr ratio between patients and controls is not surprising. Similar results were recently reported by Chong et al., who demonstrated lower NAA levels in neurologically symptomatic HIV-infected patients compared with neurologically normal patients, but failed to demonstrate a significant difference between neurologically healthy patients and controls (10). Because NAA is thought of as a marker of neuronal injury, it may be that the preservation of this ratio reflects lack of neuronal damage early in the course of infection (20).
We should emphasize that all our patients had normal MR imaging. One conclusion from the discrepancy between imaging and spectroscopy is that the amount of demyelination may be too small to detect by imaging, yet is still detectable by proton spectroscopy. An alternative possibility is that rather than demyelination, spectroscopy is detecting evidence of macrophage infiltration.
In prior work, we interpreted the elevation of marker/Cr in patients as indicating myelin damage secondary to an immunologically mediated process (21). Early in the course of disease, when patients are still asymptomatic, myelin pallor with perivascular inflammation is the prominent pathological finding. Therefore, we expected to uncover evidence of myelin damage in these patients. The fact that we see spectroscopic abnormalities may in fact be such evidence of damage, indicating that spectroscopy may be more sensitive than imaging.
An alternative, and more likely explanation, is that an elevation of the marker/Cr reflects immune activation and/or an inflammatory process, without frank demyelination. Previously, we had examined a cohort of HIV-infected patients with magnetization transfer imaging, a technique that can distinguish demyelination from edema (22). This revealed no significant evidence of demyelination early in the course of disease, when imaging was normal (23). Other recent work by our group suggests that in patients with multiple sclerosis, gadolinium enhancement on MR had a positive linear correlation with marker/Cr (24). Moreover, it has been shown that gadolinium enhancement showed a better correlation with macrophage infiltration and astroglial response than with perivascular lymphocyte infiltration (25). Thus, the marker/Cr may represent amino acids, or small-molecular-weight peptides or proteins found in macrophages. Based on comparisons between NMR spectra obtained from the cytosolic fraction of brain tissue and purified high-performance liquid chromatography separations of extracts, Kauppinen et al. (26-30) have suggested that resonances from Thymosin B4 can make significant contributions to the in vivo NMR spectrum of brain tissue. Thymosin B4 is a polypeptide (43 amino acids) found in significant concentrations in macrophages. If macrophages are present and activated, it is possible that this polypeptide, which contains 11 gluamate or glutamine residues, could contribute to the resonances in the region of the marker peaks (31).
A source of bias in our study is the difference in ages between patients and controls. Although the mean age of controls is only 5 years less than for patients, this is a statistically significant difference. It remains unclear as to whether this is a clinically significant difference. In future studies, either larger numbers or better age matching will be necessary.
Our results must be regarded as pilot data, and as such, we must be circumspect in drawing firm conclusions. Larger, prospective, multicenter trials need to be conducted to verify the improved sensitivity of spectroscopy over imaging, as well as to examine what role MRS may play in predicting the course of neurological function. We are currently collecting logitudinal data from a larger cohort of asymptomatic HIV-infected patients. By performing serial proton spectroscopy on these patients and following their clinical course over several years, we hope to better understand the alterations in the spectra that these patients manifest. Although presently the drug armamentarium for treating HIV is quite limited, in the future, as better chemotherapeutic agents are developed, the need for monitoring the CNS efficacy of such drugs will expand. MRS may be able to accomplish this task.
HIV-infected patients are just one group that may benefit from proton MRS. The ability to detect changes early in the course of disease likely extends to other processes, such as multiple sclerosis. As our ability to treat these CNS ailments improves, the importance of diagnostic modalities in general, and MRS in particular, will increase.
Acknowledgment: This work was supported by the Robert Wood Johnson Clinical Scholars Program (J.G.J.), the GE-Radiology Research Academic Fellowship (J.G.J.), and NIH HIV: NS 31464 (R.E.L.).
1. Carne C, Smith A, Elkington S, et al. Acute encephalopathy coincident with seroconversion for anti-HTLV-III. Lancet
2. Grant I, Atkinson J, Hesselink J. Evidence for early central nervous system involvement in the acquired immunodeficiency syndrome (AIDS) and other human immunodeficiency virus (HIV) infections. Ann Intern Med
3. Koralink I, Beaumanoir A, Hausler R, et al. A controlled study of early neurologic abnormalities in men with asymptomatic human immunodeficiency virus infection. N Engl J Med
4. Resnick L, Berger J, Shapsak P, et al. Early penetration of the blood-brain barrier by HIV. Neurology
5. Resnick L, DiMarzo-Veronese F, Schupbach J, et al. Intra-blood-brain-barrier synthesis of HTLV-III-specific IgG in patients with neurologic symptoms associated with AIDS or AIDS-related complex. N Engl J Med
6. Post M, Berger J, Quencer R. Asymptomatic and neurologically symptomatic HIV-seropositive individuals: prospective evaluation with cranial MR imaging. Radiology
7. Post JD, Levin BE, Berger JR, Duncan R, Quencer RM, Calabro G. Sequential cranial MR findings of asymptomatic and neurologically symptomatic subjects. Am J Neuroradiol
8. Rolfs A. Early findings in the cerebrospinal fluid of patients with HIV-1 infection of the central nervous system. N Engl J Med
9. Ho D, Rota T, Schooley R, et al. Isolation of HTLV-III from cerebrospinal fluid and neural tissues of patients with neurologic syndromes related to the acquired immunodeficiency syndrome. N Engl J Med
10. Chong W, Paley M, Wilkinson I, et al. Localized cerebral proton MR spectroscopy in HIV infection and AIDS. Am J Neuroradiol
11. Chong W, Sweeney B, Wilkinson I, et al. Proton spectroscopy of the brain in HIV infection: correlation with clinical, immunologic, and MR imaging findings. Radiology
12. Menon D, Baudouin C, Tomlinson D, Hoyle C. Proton MR spectroscopy and imaging of the brain in AIDS: evidence of neuronal loss in regions that appear normal with imaging. J Comput Assist Tomogr
13. Jarvik JG, Lenkinski R, Grossman RI, Gomori JG, Schnall MD, Frank I. Proton MR spectroscopy of HIV-infected patients: characterization of abnormalities with imaging and clinical correlation. Radiology
14. Barker P, Lee R, McArthur J. AIDS dementia complex: evaluation with proton MR spectroscopic imaging. Radiology
15. Meyerhoff D, MacKay S, Bachman L, Poole N, Dillon W, Weiner M, Fein G. Reduced brain N-acetylaspartate suggests neuronal loss in cognitively impaired human immunodeficiency virus-seropositive individuals: in vivo 1H magnetic resonance spectroscopic imaging. Neurology
16. Menon DK, Ainsworth JG, Cox IJ, et al. Proton spectroscopy of the brain in AIDS dementia complex. J Comput Axial Tomogr
17. Saykin A, Janssen R, Sprehn G, Kaplan J, Spira T, Weller P. Neuropsychological dysfunction in HIV infection: characterization in a lymphadenopathy cohort. Int J Clin Neuropsychol
18. Saykin A, Janssen R, Sprehn G, Kaplan J, Spira T, O'Connor B. Longitudinal evaluation of neuropsychological function in homosexual men with HIV infection: 18-month follow-up. J Neuropsychol
19. Fisher L, van Belle G. Biostatistics: a methodology for the health sciences
. New York: John Wiley, 1993.
20. Nadler J, Cooper J. N-acetyl-L-aspartic acid content of human neural tumours and bovine peripheral nervous tissues. J Neurochem
21. Grossman R, Lenkinski R, Ramer K, Gonzalez-Scarano F, Cohen J. MR proton spectroscopy in multiple sclerosis. Am J Neuroradiol
22. Dousset V, Grossman R, Ramer K, et al. Experimental allergic encephalomyelitis and multiple sclerosis: lesion characterization with magnetization transfer imaging. Radiology
23. Jarvik J, Grossman R. Magnetization transfer imaging in HIV infected patients: a preliminary characterization of brain findings. Radiological Society of North America, 1992.
24. Hirsch J, Lenkinski R, Grossman R. Magnetic resonance spectroscopy
in the evaluation of enhancing lesions in the brain of multiple sclerosis patients. AJNR Am J Neuroradiol
25. Nesbit GM, Forbes GS, Scheithauer BW, Okazaki H, Rodriguez M. Multiple sclerosis: histopathologic and MR and/or CT correlation in 37 cases at biopsy and three cases at autopsy. Radiology
26. Kauppinen RA, Palvimo J. Contribution of cytoplasmic polypeptides to the 1H NMR spectrum of developing rat cerebral cortex. Magn Reson Med
27. Kauppinen RA, Nissinen T, Karkkainen AM, Pirttila TR, Palvimo J, Kokko H, Williams SR. Detection of thymosin beta 4 in situ in a guinea pig cerebral cortex preparation using 1H NMR spectroscopy. J Biol Chem
28. Kauppinen RA, Kokko H, Williams SR. Detection of mobile proteins by proton nuclear magnetic resonance spectroscopy
in the guinea pig brain ex vivo and their partial purificaition. J Neurochem
29. Kauppinen RA, Niskanen T, Hakumaki J, Williams SR. Quantitative analysis of 1H NMR detected proteins in the rat cerebral cortex in vivo and in vitro. NMR Biomed
30. Kauppinen RA, Williams SR. Nuclear magnetic resonance spectroscopy
studies of the brain. Prog Neurobiol
31. Davie C, Hawkins C, Barker G, Brennan A, Tofts P, Miller D, McDonald W. Serial proton magnetic resonance spectroscopy
in acute multiple sclerosis. Brain