In Vivo Brain Concentrations of N-Acetyl Compounds, Creatine, and Choline in Alzheimer Disease FREE

ALZHEIMER DISEASE (AD) results in substantial loss of brain tissue.^1- 15 This decline is attributable to deterioration of cell processes, shrinkage of neurons, and perhaps neuronal death.^10,16- 20 Healthy elderly individuals show an age-related gray matter volume decline in the neocortex^21- 24 (see Guttman et al²⁵) but their volume loss may not be accompanied by neuronal death.^26- 29 At issue is the functionality of the tissue remaining in the brains of healthy elderly patients and patients with AD.

In vivo proton magnetic resonance spectroscopic imaging (¹H MRSI) allows examination of brain tissue integrity. TheN-acetyl (NAc) peak is composed of several NAc compounds, including NAc aspartate, that are present only in living, mature neurons and not glia.³⁰ Many studies have reported lower NAc signals in patients with AD compared with healthy elderly individuals ^31- 33; some,^34- 37 but not others,^38- 43 have reported age-related decline in brain NAc in healthy people. Most studies have employed single-voxel spectroscopy³¹ and have expressed NAc as a ratio^44- 47 of 1 or 2 other metabolites: creatine plus phosphocreatine (Cr), which reflects high-energy phosphate metabolism, and choline (Cho), which increases in signal intensity with membrane synthesis and degradation.^48- 49 Many proton spectroscopy studies assume that neither age nor AD seriously affects Cr or Cho levels and thus express data as ratios.

We developed an MRSI method to estimate absolute proton metabolite signal intensities in gray matter and white matter separately.³⁸ Concentration, expressed as institutional signal intensity units per tissue volume, revealed more NAc in gray matter than white matter, consistent with many, ^34,50- 56 but not all,^57- 60 studies. Despite significant gray matter volume deficits in elderly healthy individuals, the young and elderly groups had equivalent concentrations of NAc in gray matter and white matter. By contrast, Cr and Cho concentrations demonstrated significant age effects. Cho concentrations were greater in gray matter in older controls; Cr concentrations were greater in gray matter and white matter in older subjects. These observations draw into question the use of Cr and Cho as appropriate referents for determining NAc concentration.

We applied our MRSI method to patients with AD and compared their results with those from our study of normal aging.³⁸ We expected the patients with AD, unlike the healthy elderly controls, to have abnormally low NAc concentrations in gray matter and white matter. We also anticipated that Cho levels would be even higher in patients with AD than they were in healthy elderly controls.

The 3 study groups included 15 young controls (all men; mean [±SD] age, 25.3±2.9 years), 19 healthy elderly controls (9 men, 10 women; mean age, 73.3±4.1 years), and 16 patients with AD (6 men, 10 women; mean age, 73.4±6.0 years). All analyses were performed blind to subject identity.

The patients with AD were recruited from the Geriatric Psychiatry Rehabilitation Unit and the National Institute of Mental Health Aging Clinical Research Center at the Veterans Affairs Palo Alto Health Care System, Palo Alto, Calif (Table 1). These centers have a 90.2% autopsy-confirmed AD success rate of patients who in life had a diagnosis of probable AD. Patients met National Institute of Neurological and Communicative Diseases and Stroke–Alzheimer's Disease and Related Disorders Association criteria for probable Alzheimer disease.^61- 62 Controls were recruited from the local community. Subjects were excluded if they had any significant history of psychiatric or neurological disorder unrelated to their diagnosis (stroke, closed head injury), current alcohol or other drug abuse or dependence, or a life-threatening medical condition. Screening included a psychiatric interview and medical examination; informed consent was obtained from all participants (Table 1).

The elderly control and AD groups were matched in age and sex, were comparable in years of education (elderly control mean [±SD], 16.3±2.2 years; AD mean, 15.4±2.0 years), and achieved similar scores on the National Adult Reading Test, an estimate of premorbid intelligence (elderly control mean [±SD], 115.6±5.8; AD mean, 110.6±9.5). Patients with AD (mean [±SD] score, 19.8±4.6) had significantly lower scores than the elderly controls (mean score, 28.5±1.2) on the Mini-Mental State Examination.⁶³

Cognitive data were collected within 3 months (0-84 days) of magnetic resonance imaging. Brain-behavior relationships were based on tests assessing verbal and nonverbal recognition and word finding. The Warrington Recognition Test⁶⁴ assessed memory for words and faces, and the modified Boston Naming Test⁶⁵ assessed confrontation naming.

In vivo proton MRSI and structural magnetic resonance imaging scans were acquired using a quadrature head coil on a 1.5-T magnetic resonance imaging scanner (5.6 System software; GE Signa, Milwaukee, Wis) with echo-speed gradient hardware (GE Medical Systems; Milwaukee, Wis) (2.2-G/cm maximum gradient amplitude, and 185-µs minimum rise time). Data were obtained with oblique anatomic prescriptions parallel to the anteroposterior commissure (AC-PC) line identified from midsagittal structural images. The image acquisition protocol and analysis are given in other studies.³⁸

A modified version of a 3-dimensional MRSI sequence using a time-varying readout gradient in the slice selection direction was used to image multiple contiguous slices.⁶⁶ Excitation was accomplished with a pair of spin-echo, spectral-spatial pulses preceded by an adiabatic inversion as described in other studies.³⁸ Collection parameters were repetition time (TR), 2 seconds; inversion time (TI), 170 milliseconds; echo time (TE), 144 milliseconds. The nominal voxel size was 1.1 cm³ and the total MRSI time was 17 minutes. Data reconstruction and metabolite estimation were performed as described elsewhere.⁶⁶ Corrections for receiver gain and coil loading were made when images were reconstructed to allow comparability of metabolite signal levels between subjects.⁴¹ The final metabolite images were 16 images of 32×32 pixels each.

An automated shimming procedure^67- 68 with 3 linear terms and 6 nonlinear terms was used to minimize main field variations. A final 3-dimensional fieldmap was collected at a resolution of 64×64×32 voxels (TR, 40 milliseconds; and TE, 10 milliseconds; flip angle, 20°; effective slice thickness, 6.4 mm; and field of view, 24 cm) to measure residual field inhomogeneities.

A midsagittal, gradient-recalled echo image was used to compute slice positions with 0.5-mm accuracy for all 3 scans in this protocol (anatomical, fieldmap, and MRSI). Anatomic images were acquired with an axial fast-spin echo protocol (TR, 3000 milliseconds; and TE, 20/80 milliseconds; echo train length, 8; 3-mm skip, 0.2 mm; 256 × 256 pixels matrix; field of view, 24 cm; number of excitations, 1; and time, 3 minutes 18 seconds). Sixteen slices were collected, the most inferior slice beginning at the anteroposterior commissure line, corresponding to the 8 middle spectroscopic slices and providing 2 anatomical slices for each MRSI slice. Twelve of these high-resolution images, corresponding to 6 slices of metabolite data, were used in the structure/metabolite analysis.

An average imaging session with the above-described protocol took about 1 hour.

Six MRSI slices were used, beginning with the slice 12.8 mm above the anteroposterior commissure line and extending superiorly (Figure 1). These slices were chosen because they had the least amount of signal loss and artifacts due to field inhomogeneity. Within these 6 slices, only pixels with good homogeneity (B₀ shifts within the range of ±5 Hz) were included for analysis. These slices were also manually edited to remove regions, usually outside of the brain, of obvious lipid and/or water artifacts. To further guard against the possibility that incompletely suppressed water signal contaminated the MRSI data, especially for Cho and Cr concentrations in the medial frontal region, an exclusion region roughly corresponding to the cingulate gyrus was constructed for each slice by proportionate geometric positioning. The metabolite signals were calculated as magnitude values, so the noise in the metabolite maps had a nongaussian (Rician) noise distribution in the low signal-to-noise ratio range. To account for the effects of this nongaussian noise distribution, a bias correction was applied to the metabolite signal intensity values.⁶⁹

Nonbrain (ie, dura, skull, and scalp) tissue was removed. The images were converted into segmentation maps, with each pixel designated as gray matter, white matter, or cerebrospinal fluid, and volumes were derived for each compartment. The data from each of the 2 segmented, high-resolution, structural slices corresponding to a metabolite slice were combined to provide 128 (8×8×2) segmented voxels underlying each metabolite voxel. Brain tissue composition of each subject was computed, with and without the voxels excluded in the metabolite analysis (Figure 1).

The segmented structural data were filtered to yield the same spatial frequency characteristics as the MRSI data. The optimum gray matter and white matter NAc, Cr, and Cho levels were found using a robust, least absolute deviation–regression algorithm.³⁸ The metabolite spectrum in each voxel was assumed to be a combination of signals from tissue compartments of gray matter and white matter. The gray matter and white matter tissue composition of each spectroscopic voxel was calculated from the segmented structural data. Given this information, the signal intensity of both gray matter and white matter was estimated for each subject and each metabolite.

Group differences were tested with repeated-measures analysis of variance and Student t tests. We performed exploratory correlations between metabolite measures and cognitive test scores with Pearson product-moment correlations. We report correlations that reached P<.05 or less (1 tailed), with predictions that better performance would be associated with higher concentrations of NAc and lower concentrations of Cr and Cho; family-wise Bonferonni adjustment for 3 comparisons (3 tests for each metabolite) requiredP<.03. Correlations were done for the AD and elderly control groups independently to avoid the possibility that the results would merely reflect group differences.

The patients with AD had smaller gray matter volumes than the elderly controls, who had smaller volumes than the young controls (F_2,49=115.904,P=.001). Cerebrospinal fluid volumes showed a complementary step-wise group effect (F_2,49=70.966, P=.001). The overall group difference was significant for white matter (F_2,49=3.476,P=.04), but only the AD group exhibited a volume deficit relative to the young and the elderly control groups (Figure 2). The pattern of group differences was virtually identical when examining only those structural voxels that corresponded to the spectroscopic voxels meeting the inclusion criteria (Table 2).

Despite the significant tissue volume deficit in patients with AD, the number of spectroscopic voxels meeting criteria for analysis did not differ significantly among the 3 groups (F_2,49=2.399,P=.10). The concentration estimation model provided separate metabolite concentration estimates for gray matter and white matter (Figure 3).

A repeated-measures analysis of variance (3 groups by 2 tissue types) for NAc concentration yielded significant effects of group (F_2,47=6.913,P=.002) and tissue type (F_2,47=234.108, P=.001) but no interaction (F_2,47=1.708,P=.19). The overall group difference was significant in gray matter (F_2,49=5.283, P=.008); the AD group had significantly lower gray matter NAc levels than the young (P=.03) and elderly (P=.009) control groups, which did not differ from each other. The ratio of gray to white matter NAc concentration was similar for the 3 groups (young control ratio=1.31; elderly control ratio=1.38; and patients with AD ratio=1.30) (F_2,49=1.024,P=.37).

For Cr concentration, a repeated-measures analysis of variance (3 groups by 2 tissue types) yielded significant group (F_2,47=21.25,P=.001) and tissue-type effects (F_1,47=884.214, P=.001) and interaction (F_2,47=3.934,P=.03). For both tissue types, the elderly controls and patients with AD had higher Cr concentrations than the young controls (elderly vs young controls for gray matter: t₃₂=4.678,P=.001, and for white matter: t₃₂=3.935,P=.001; patients with AD vs young controls for gray matter: t₂₉=4.619, P=.001, and for white matter:t₂₉=2.572, P=.02). By contrast, the elderly control and AD groups did not differ significantly from each other in either gray matter (t₃₃=1.065,P=.29) or white matter (t₃₃=1.386, P=.18) Cr concentrations. Again, the ratio of gray matter to white matter Cr concentration was similar for the 3 groups (young control group ratio=1.98; elderly control group ratio=1.95; and AD group ratio=2.12) (F_2,49=1.503,P=.23).

For Cho concentrations, a repeated-measures analysis of variance (3 groups by 2 tissue types) yielded a significant group effect (F_2,47=10.964,P=.001) and interaction (F_2,47=12.073, P=.001) and trend toward a tissue-type effect (F_1,47=3.691,P=.06). Gray matter Cho concentrations were lowest in the young controls and highest in the patients with AD, and all group-paired comparisons were significant (young vs elderly: t₃₂ =3.925,P=.001; elderly vs patients with AD: t₃₃=2.506,P=.02; young vs patients with AD: t₂₉=4.993,P=.001). As indicated by the interaction, white matter Cho showed a different pattern from gray matter; only the comparison between the elderly controls and patients with AD reached significance; the elderly group had higher Cho concentrations than the AD group (t₃₃=3.074, P=.004). Unlike NAc and Cr, the ratio of gray matter to white matter for Cho concentration showed a group difference, with the value for the AD group higher (ratio=1.20) than values for the young (ratio=.79,t₂₉=4.061, P=.001) and elderly control (ratio=.91,t₃₃=3.362, P=.002) groups; the young and elderly controls did not differ significantly from each other (t₃₂=1.351,P=.19).

The 3 group-by-metabolite analyses of variance were recalculated excluding women, because the younger group was composed of men only. The results were the same without the women, except that the group-by-Cr interaction was no longer significant in the men-only analysis.

As expected, the AD group was impaired on the 3 cognitive measures compared with the elderly controls (Table 3). There were several brain-behavior correlations in the predicted direction. In elderly controls, face recognition scores were positively correlated with gray matter NAc concentration (r=0.80,P=.001, n=13). In patients with AD, higher Cr gray matter concentrations were related to lower word-recognition scores (r=−0.67, P=.03, n=11), and higher Cho gray matter concentrations were related to lower face recognition scores (r=−0.70,P=.02, n=11).

Independent estimation of the concentrations of NAc, Cr, and Cho revealed different patterns across the groups: NAc showed a disease effect, Cr showed an age effect, and Cho showed disease and age effects. N-Acetyl compounds concentration was significantly reduced only in AD in gray matter but not white matter, even though both the AD and elderly control groups had substantial gray matter volume deficits relative to the young controls. Both the elderly healthy and AD groups had an excess of Cr in gray matter. Choline concentration in gray matter was notably higher in the elderly than in the young control group, and the AD values were even higher than those measured in the elderly controls.

While neuronal soma and processes may shrink, the current consensus is that little if any loss of neuronal cell numbers occurs in normal aging.^26- 29 Gray matter volume deficits without NAc concentration deficit suggests normal cell integrity in the healthy elderly. Like postmortem studies,^{10,16- 20,28} the current in vivo study revealed severe gray matter volume deficits and an accompanying deficit in gray matter NAc concentration in AD.

For literature comparison, we calculated whole brain raw signal intensity NAc/Cr ratio values of 1.382, 1.274, and 1.225 for the young, elderly, and AD groups, respectively. Concentration-corrected ratios of NAc/Cr for gray matter only presented a similar pattern of 1.38, 1.229, and 1.098 for the young, elderly, and AD groups, whereas for white matter only the pattern was 2.073, 1.734, 1.743. Thus, the NAc/Cr ratio differences seen in whole-brain mixed gray-white tissue represent a decrease in the NAc/Cr ratio with aging but no additional AD effect in white matter and a further decrease in this ratio in the AD group in gray matter.

Consistent with other in vivo^{31,44- 47,70- 72} and ex vivo or postmortem^{33,54,73- 74} reports, NAc concentration, whether expressed as a ratio of Cr or Cho or in absolute terms, was substantially lower in patients with AD than in age-matched controls. Kwo-On-Yeun et al⁵⁴ also noted substantially more reduction in NAc in gray matter than in white matter in comparing patients with AD to controls. As in other studies, NAc signal intensity was greater in gray matter than white matter for all 3 groups in our study.

Across all 3 groups, the calculated Cr concentration in gray matter was almost twice that in white matter, whereas the Cho concentration was more uniformly distributed between gray and white matter. Given these age- and disease-related differences in metabolite concentrations for gray and white matter, it is clear that neither Cr nor Cho concentration is constant, and, thus, they should not be used without regard to age and disease effects as referents in ratio expression of the NAc signal. The 3 group comparisons—young controls, elderly controls, and elderly patients with AD—help to disentangle disease effects from effects of normal aging. The former should be superimposed on the latter. Patients with AD did not significantly differ from the elderly controls in the volume of Cr in either gray or white matter, but both groups differed from the young controls. This pattern indicates that increased Cr concentration in patients with AD is attributable to advanced age rather than to the disease. The elevated gray matter Cho concentration in AD, however, seems to be the result of the additive effects of advanced age plus disease. Chang et al³⁹ noted decreased brain water content (noncerebrospinal fluid) with age, which could explain a relative decrease in cortical volume with the same number of cells, leading to increased metabolite concentration. Miller et al⁴⁹ reported that increased choline reflects degree of cellular density in brain tumors. Adding gliosis to this aging effect could explain the additional increase in Cho level we observed in patients with AD. White matter Cho concentration showed the opposite pattern, ie, no increase with age but a decrease with AD. Moats et al⁷¹ observed age-related increases in Cho levels in normal elderly but no further differences between normal elderly and patients with AD. Our observed increased Cho concentration in gray matter may be the result of cell membrane turnover.

Pettegrew et al,⁷⁵ using in vitro phosphorus 31 (³¹P) MRS, found elevated phosphomonoester levels in patients with AD. Brown et al⁷⁶ showed that phosphomonoester, phosphomonoester/phosphodiesther, and inorganic phosphate levels were elevated in patients with AD compared with controls, whereas no significant differences in any ³¹P indexes were found by Bottomley et al.⁷⁷ Similar findings were reported by Murphy et al.⁷⁸ More recently, Gonzalez et al⁷⁹ reported a 50% increase in phosphomonoester/phosphoiesther levels in patients with AD, but unchanged β-nucleoside 5‘-triphosphate, phosphocreatine, and inorganic phosphate levels in AD, speculating that the phosphodiesther difference reflected changes in the biophysical state of membrane phospholipids. Using quantitative ³¹P and ¹H perchloric acid extracts, Klunk et al⁸⁰ found increased myoinositol, aspartate, L-glutamate, alanine, phosphocholine, and phosphodiester levels, and decreased phosphoetanolamine and NAc-L-aspartate levels. They concluded that compounds related to membrane degradation and excitatory neurotransmission increased in patients with AD, while those related to neuronal integrity and inhibitory neurotransmission decreased. Pettegrew et al⁸¹ reported increased phosphomonoesther and phosphocreatine levels that preceded dementia for 1 subject.

The proton Cr signal is a combination of phosphocreatine and creatine.⁸² Thus, one cannot separate the contribution of creatine and phosphocreatine to the total Cr signal or relate findings directly to phosphorus spectroscopy studies, which, for instance, report a decreased ratio of phosphocreatine to inorganic orthophosphate in AD.⁸³ Similarly, several Cho-containing compounds contribute to the Cho peak in proton magnetic resonance spectroscopy. As reviewed by Michaelis et al⁸² Cho plasmogen (0.6 mmol/kg), glycerophosphorylcholine (0.4 mmol/kg), phosphorylcholine (0.4 mmol/kg), cytedinediphosphate-choline (0.05 mmol/kg), acetylcholine (0.03 mmol/kg), and choline (0.02 mmol/kg) contribute to the ¹MRS choline signal. Contributions from 15- to 18-mmol/L lipid-soluble phosphatidylcholine are minor.^49,82 In vivo ³¹P MRS could clearly play an important role in identifying the sources of the observed Cho as well as Cr signals in¹H MRSI.

A limitation of our method is imposed by the inhomogeneity of the main magnetic field. While we applied second-order, nonlinear shims to improve the homogeneity, the remaining field inhomogeneities, particularly in the more inferior and frontal brain regions, limit the useful extent of the observed volume. Because we apply a late-echo acquisition, metabolites with short T₂ relaxation times, such as myoinositol, are not observed.

The level of brain metabolite concentration seen with ¹H MRSI appears to have functional significance, given the correlations found with cognitive measures. The metabolite concentration–memory test correlations are consistent with our previous work, which showed a selective relationship between these tests and degree of hippocampal volume loss in patients with AD.^3,6 Together, these results lend support to the hypothesis that the prominent disease-related increase in gray matter Cho concentration marks cellular degeneration resulting in reduced NAc concentration and poorer cognitive function in AD.

Accepted for publication October 2, 1998.

This work was supported by grants AG11427, AA05965, and MH30854 (Dr Pfefferbaum), AA10723 and MH40041 (Dr Sullivan), and RR09784 and CA48269 (Dr Spielman) from the National Institutes of Health, Bethesda, Md.

We thank the the State of California Alzheimer's Disease Research Center, Veterans Affairs Palo Alto Health Care System, Palo Alto, Calif, for providing subjects for the study.

Corresponding author: Adolf Pfefferbaum, MD, Neuropsychiatry Program (BN166), SRI International, 333 Ravenswood Ave, Menlo Park, CA 94025 (e-mail: dolf@synapse.sri.com).