Reduced glutamate in the anterior cingulate cortex in depression: an in vivo proton magnetic resonance spectroscopy study

ORIGINAL ARTICLES Reduced Glutamate in the Anterior Cingulate Cortex in Depression: An In Vivo Proton Magnetic Resonance Spectroscopy Study Dorothee P. Auer, Benno Pu¨tz, Eduard Kraft, Bernadette Lipinski, Julia Schill, and Florian Holsboer Background: Functional imaging studies suggest a specific role of the anterior brain regions in the pathogenesis of major depression. The aim of this study was to evaluate possible neurochemical alterations in the frontomesial cortex in patients with major depressive episode using in vivo proton magnetic resonance spectroscopy (1H-MRS). Methods: Single voxel 1H-MRS was performed in 19 patients with major depressive episodes and 18 agematched healthy controls within the anterior cingulate cortex and the parietal white matter. Absolute concentrations were estimated for N-acetyl-aspartate, choline-containing compounds, total creatine, myo-inositol, unresolved glutamate and glutamine (Glx) and glutamate alone (Glu). Voxel composition was analyzed by image segmentation into cerebrospinal fluid (CSF), grey and white matter. Results: MANOVA test for Glx and Glu using age, percent CSF and percent grey matter contribution as covariates yielded a significant group effect within the anterior cingulate due to decrease of Glx in patients (⫺10.4%, p ⫽ .013). Considering only severely depressed patients, both Glx and Glu (⫺14.3%, p ⫽ .03) showed a significant decrease. There was no significant group effect for the neuronal marker NAA, creatine, choline or myo-inositol in either localization. Conclusions: This study suggests a possible role of altered glutamatergic neurotransmission within the anterior cingulate in the pathogenesis of mood disorders. The otherwise unremarkable findings of major brain metabolites confirms lack of neurodegenerative or membrane metabolic changes in major depression. Biol Psychiatry 2000;47:305–313 © 2000 Society of Biological Psychiatry Key Words: Major depression, magnetic resonance proton spectroscopy, glutamate, cingulate cortex, regional differences

From the Max Planck Institute of Psychiatry, Munich, Germany Address reprint requests to Dr. Dorothee P. Auer, Max Planck Institute of Psychiatry, Kraepelinstr. 10, D-80804 Munich, Germany. Received March 11, 1998; revised November 12, 1998; accepted June 21, 1999.

© 2000 Society of Biological Psychiatry

Introduction

T

he majority of functional imaging studies in depressed patients show reductions in blood flow or neuronal energy consumption in the anterior regions of the brain. Most consistently, perfusion reductions have been found in the inferior frontal cortex, anterior cingulate, temporal cortex and basal nuclei. (Bench et al 1995; for review see Ebert and Ebmeier 1996 and Goodwin 1996). Hypotheses on the pathogenesis of depression include alterations of various neurotransmitter systems, e.g., defunct release, reuptake or receptor-mediated signaling of norepinephrine or serotonin (Meltzer et al 1987; Schildkraut 1965), or neuroendocrine hypotheses submitting that enhanced hypothalamic-pituitary-adrenocortical activity is a causal factor (Holsboer et al 1995). The experimental evidence for these hypotheses is derived mostly from indirect measurements such as plasma or cerebrospinal fluid (CSF) concentrations of neurotransmitters or neurohormones. The recent technological advances in brain imaging allow further testing of these hypotheses by using in vivo receptor imaging. For example, Larisch et al (1997) found dopamine type 2 (D2) receptor binding to be increased in the striatum or anterior cingulate gyrus of patients with depression treated with selective serotonin reuptake inhibitors. This finding does not imply, however, that dopaminergic neurotransmission is changed as, in untreated patients with schizophrenia and depression, D2 receptor binding was indistinguishable from that in controls. Apart from assessing receptor binding, the measurement of neurotransmitter concentrations in specific brain areas is a promising opportunity offered by proton magnetic resonance spectroscopy (1H-MRS). Charles et al (1994) reported an increased ratio between resonances of cholinecontaining compounds (Cho) and total (phospho-)creatine, creatinine (Cr) in the basal ganglia of depressed patients. Controversially, a more recent study found a reduced Cho/Cr in the basal ganglia in depressed patients, that was more pronounced in fluoxetine treatment responders (Renshaw et al 1997). An interesting temporary reduction in a number of neurobiochemicals including myo-inositol (mI) and glutamate (Glu) was noted in two patients with 0006-3223/00/$20.00 PII S0006-3223(99)00159-6

306

D.P. Auer et al

BIOL PSYCHIATRY 2000;47:305–313

Table 1. Clinical Data of Patients with Major Depressive Disorder Patient

Antidepressive treatment (mg)

Day of treatment

Age

ICD-10

HDS

1

F

31

36

Studya

4

2 3 4 5 6 7 8 9

F F F F M M F F

32 57 46 50 43 66 62 74

28 26 23 21 20 18 17 17

Doxepine 300 Trimipramine 200 Mirtazapine 15 Amitryptiline 150 None Paroxetine 20 None None

17 ? 2 7 — 5 — —

10 11

M M

46 48

16 14

None None

— —

Chloralhydrate 500 –1000 Lorazepam 1 Lorazepam 4 None None Chloralhydrate 1000 Lorazepam 3 None Flunitrazepam 1.5/ Lorazepam 2 None None

12 13 14 15 16 17 18

F F M F M F F

60 37 33 42 49 56 60

F32.11 F42.0 F31.5 F33.3 F33.2 F33.3 F33.2 F32.3 F32.2 F33.3 F13.1 F32.2 F32.2, F34.1 F32.2 F33.2 F33.2 F33.1 F32.11 F33.1 F32.3

14 12 ND ND ND ND ND

— 1 3 — 21 13 62/10/16

Cloralhydrate 500 None Lorazepam 2 None Lorazepam 2 Chloralhydrate 500 Lorazepam 0.5

3 — 3 — 15 16 16

19

F

62

F32.1

ND

None Amitriptiline 25 Amitryptiline 150 None Mirtazapine 30 Mirtazapine 30 Trimipramine 400/ Haloperidol 10/ Carbamazepine 300 Paroxetine 20

None

—

6

Sedatives (mg)

Day of treatment

Gender

9 1 6 — — 4 15 — 4/1 — —

ND, not done; HDS, Hamilton Rating Scale for Depression score. a Double-blind study testing paroxetine vs. tianeptine 37.5 mg.

secondary depression during Taxol/Neupogen chemotherapy for breast cancer (Cousins and Harper 1996). To further assess possible regional metabolic, especially glutamatergic, alterations in patients with major depression, we performed an 1H-MRS study in patients with major depression and age-matched controls. Two regions were chosen: the anterior cingulate gyrus, one of the target regions thought to be involved in mood regulation, and parietal white matter, a common reference region in spectroscopic studies that should not specifically be involved in major depression. To account for bias due to possible differences in anterior brain structures and variations in voxel positions between the groups, we used magnetic resonance imaging (MRI) to segment grey and white matter and CSF from individual voxels and calculated percent tissue composition in each voxel.

Methods and Materials Patients Nineteen patients with depression (mean age: 50.2 [12.2] years [SD]; 13 women and 6 men) who underwent routine MRI examination shortly after hospital admission were included in the study. All subjects were depressed at the time of examination. ICD-10 diagnoses were depressive episodes (9), recurrent depressive disorder (9, including 1 atypical depression) and bipolar affective disorder (1). The level of depression was scored

“severe” in 14 and “medium” in 5 patients according to ICD-10 severity criteria (compare Table 1) . Hamilton Rating Scale for depression scores were available in 13 patients with a mean of 22.5. At the time of examination, 7 patients were medication free or were receiving sedatives only, the remainder had had specific antidepressive therapy for 1 to 62 days (mean: 7.9 days). In 3 patients, psychiatric comorbidity was present: obsessive-compulsory behavior (ICD 10: F 42.0) once, dysthymic disorder (ICD 10: F 34.1) once and benzodiazepine abuse (ICD 10: F 13.1) once (Table 1).

Controls Thirty healthy volunteers were recruited via newspaper advertisements. Written informed consent was given according to the local ethical review board guidelines. An interview was performed to rule out any contraindications to MRI as well as any neurological or psychiatric abnormalities. In 18 volunteers (mean age 43.2 [14.6] years [SD], 10 women, 8 men) an anterior cingulate spectrum was acquired and in six patients a parietal white matter spectrum was also obtained. In the remaining 12 patients, only a parietal white matter spectrum was acquired, in the context of an unrelated study. The mean age of the 18 controls chosen for parietal white matter localization was 42.4 [13.1] (years [SD]); 12 were women and 6 men.

Magnetic Resonance Imaging and Spectroscopy The study was performed on a 1.5T GE whole body scanner (Signa Advantage, GE Medical Systems, Milwaukee) using a

1

H-MRS in Depression

BIOL PSYCHIATRY 2000;47:305–313

307

Figure 1. Voxel position and tissue segmentation in the anterior cingulate cortex. Upper left: Anatomic image with superimposed voxel borders. Upper right: Histogram analysis of signal intensities (arbitrary units) within total voxel. Bottom row: Segmented volume fractions of three pixel classes (left, cerebrospinal fluid, black; middle, gray matter, gray; right, white matter, white) as separated by interactively adjusted thresholds.

standard quadrature head coil. A PRESS sequence (TR/TE ⫽ 2000/35 ms) was used for 1H-MRS to minimize T1 and T2 relaxation effects within an acceptable examination time. The position and size of the voxels were visually adjusted on sagittal scouts to include mostly anterior cingulate cortex with minimum partial volume effect from corpus callosum (Figure 1) in all the patients and the 18 controls. Parietal white matter spectra were additionally acquired in 14 patients and the 18 controls as mentioned above, position and size of the voxel was graphically prescribed on axial and coronal T2 weighted scans within the deep parietal white matter to reduce partial volume effect from ventricle as well as cortex. Voxel size varied between 7.5 and 12 mls for the anterior cingulate and between 6.8 and 9.2 mls for the white matter spectra. The number of averages was standardized to 128 resulting in a total acquisition time of 6 min per spectrum. Shimming was performed by automated first order gradient shim with additional manual optimization. Water suppression was manually adjusted by controlling the transmit gain. Magnetic resonance imaging protocol for patients included T2 and proton density-weighted (T2w/PDw) fast spin echo (FSE: TR ⫽ 3000, TE ⫽ 30 and 80 ms) images in transaxial orientation, T2w FSE coronal scans and a T1-weighted 3Dspoiled gradient echo acquisition (3D-SPGR, TR ⫽ 10.3 ms, TE ⫽ 3.4 ms, flip ⫽ 20°, 0.9 ⫻ 0.9 ⫻ 1.1 mm pixel size) of the whole brain. In the controls, transaxial T2w FSE and a T1w 3D-SPGR or a shorter 3D gradient echo sequence (1 ⫻ 1 ⫻ 3 mm pixel size) were performed.

Data Analysis Postprocessing of spectra was standardized using commercially available software (SAGE/IDL) with zero filling, Fourier transformation, line broadening with 2 Hz and automated phase and eddy current correction. Metabolite concentrations were calculated using a time-domain fitting program (LC-model; Provencher 1993). Cr, mI,

N-acetyl-aspartate (NAA), Cho, Glu and unresolved glutamate and glutamine resonances (Glx) were analyzed and expressed in institutional units. The program also estimates the fitting error expressed in percent standard deviation. For robustness of final results, only metabolite information with an error ⬍20% SD was included in the final analysis. For visual qualitative analysis, spectra were averaged across the patient and the control group and difference spectra were calculated as follows: the baseline corrected fitted spectra as provided by the LC-model were scaled according to the estimated absolute metabolite concentrations, averaged for both groups over the relevant ppm range (0 to 4.2 ppm, covering all major metabolites) and then subtracted. This procedure allows only to show qualitative differences between the two groups because methodical limitations have to be considered when in vivo spectra are averaged concerning possible differences in frequency alignment, peak shape and baseline. To account for possible variation in voxel composition (CSF contamination, percentage of grey [GM] and white matter [WM] contribution) segmentation was done using an interactive program written in IDL that allows the visualization of the voxel borders on the simultaneously acquired 3D-SPGR and estimates the volume fraction of the different tissue types. Adaptation of thresholds for separation of CSF, GM and WM was supported by automatic histogram analysis and inspection of differently coloured tissue classes over 15 to 20 individual slices running through the prescribed voxel (Figure 1). Statistical analysis was performed with commercial SPSS software (SPSS 7.0 for Windows, SPSS software). MANOVA was used to test for significant group effects between controls and patients as well as for between controls and only severely depressed patients (n ⫽ 14) for the dependent variables Glx and Glu. To control for the remaining effects of age and tissue composition of the respective voxels, age, percent CSF and percent GM were used as covariables in the model. For possible group effects of the remainder metabolites, that can be consid-

308

D.P. Auer et al

BIOL PSYCHIATRY 2000;47:305–313

Results

Table 2. Tissue Composition

Anterior cingulate Controls (n ⫽ 18) All patients (n ⫽ 19) Severely depressed patients (n ⫽ 14) Parietal WM Controls (n ⫽ 18) All patients (n ⫽ 14) Severely depressed patients (n ⫽ 11)

GM

WM

CSF

57.5 (7.9) 58.1 (6.0) 58.6 (5.4)

27.7 (11.0) 24.8 (7.5) 24.4 (7.1)

14.8 (6.0) 17.1 (6.0) 17.0 (6.0)

14.2 (7.2) 14.3 (4.6) 13.0 (3.0)

83.2 (7.9) 83.2 (4.3) 84.3 (3.1)

2.6 (1.4) 2.4 (1.4) 2.7 (1.4)

Group mean values and standard deviations (SD) for grey matter (GM), white matter (WM) and cerebrospinal fluid (CSF) content of individual voxels expressed in volumetric % of total voxel.

ered as independent variables (NAA, Cho, Cr, mI), age, gender and tissue components (percent GM and WM and CSF voxel content) a two-tailed, nonparametric Mann–Whitney U test was applied.

Anterior Cingulate Spectra The patient and control groups did not differ significantly (at the p ⬍ .05 level, Mann–Whitney U test) in age, gender or respective tissue components (CSF, GM, WM, Table 2). The difference spectrum (Figure 2a) shows reduced signals in the range between 2.6 and 2.0 ppm, mainly attributable to Glx and only minor signal deviations for the other resonances. Quantitatively, no differences were seen between controls, depressed patients or severely depressed patients for mI, Cho, Cr or NAA resonances (Table 3) . Mean fitting errors varied between 3.7 (%SD, NAA) and 5.9 (%SD, mI) for patients and 3.9 (%SD, NAA) and 6.5 (%SD, mI) for controls. No single SD error of these resonances reached the 20% level, the highest error being 10% in 3 cases. An estimated SD of 20% or above was seen in one spectrum for Glx and Glu, and additionally, in one for Glu only, and these were discarded from further

Figure 2. Averaged fits from proton spectra within the (A) anterior cingulate and (B) parietal white matter of the control group (gray lines) and patient group (black lines). Difference spectra (control group ⫺ patient group) are displayed in the bottom rows. Cr, (phospho-)creatine; mI, myo-inositol; Cho, choline-containing compounds; Glx, unresolved glutamate and glutamine resonances; NAA, N-acetyl-aspartate.

1

H-MRS in Depression

BIOL PSYCHIATRY 2000;47:305–313

309

Table 3. Metabolite Concentrations in Anterior Cingulate Spectra

Controls (n ⫽ 18) All patients (n ⫽ 19) Severely depressed patients (n ⫽ 14)

Age (years)

mI (IU)

Cr (IU)

Cho (IU)

NAA (IU)

Glx (IU)

Glu (IU)

43.2 (14.6) 50.2 (12.2) 51.0 (12.6)

16.59 (2.08) 17.00 (1.40) 16.83 (1.28)

21.14 (2.40) 21.17 (2.52) 21.11 (2.50)

5.32 (0.64) 5.29 (0.72) 5.20 (0.51)

29.16 (3.01) 28.74 (4.22) 28.39 (3.85)

33.84 (5.1) 30.68 (4.04) 31.40 (3.48)

23.81 (4.08) 21.06 (3.51) 20.40 (2.90)

Group means (standard deviations). IU, institutional units. Abbreviation for metabolites as given in text.

analysis. Significant group effects (p ⫽ .043, Wilks test) could be shown for Glx and Glu between controls and depressed patients using MANOVA after correcting for covariable effects of age, CSF and GM (Table 2). The group effect could be attributed to the dependent variable Glx (univariate F test, F ⫽ .013) that showed a 10.4% group mean difference. There was a significant (Wilks test, p ⫽ .002) covariate effect for both variables Glx and Glu (univariate F test: p ⫽ .002 and 0.034 respectively). Individual univariate testing showed a significant effect for percent GM (t ⫽ .003) and a trend for percent CSF (t ⫽ .066) on Glx, whereas only percent GM showed a trend for an effect on the variable Glu (t ⫽ .093). The difference spectra would have suggested a somewhat larger effect size than 10% but for the quantitative analysis we have chosen a conservative exclusion criterion that led to omission of spectra with the lowest metabolite concentration thus possibly leading to an underestimation of the group difference. Considering only severely depressed patients in the same design using CSF, GM and age as covariables, a significant group effect was again found (p ⫽ .047, Wilks test), to which both Glx (F ⫽ .028, univariate F test) and Glu (F ⫽ .03, univariate F test) provided significant contributions. The respective mean difference was more pronounced for Glu (⫺14.3%) than Glx (⫺7.2%). No significant correlation was found between Hamilton Rating Scale for Depression scores and any of the analyzed metabolites using Pearson’s correlation coefficient.

Parietal White Matter Spectra 1

H-MRS spectra from parietal white matter were similar in the patient and control groups (Figure 2b) with no relevant difference signals of the major resonances. There was no statistical significant group difference (at the p ⬍ .05

level, Mann–Whitney U test) for age, gender and respective tissue components (CSF, GM, WM, Table 2) either. Metabolite data for NAA, mI, Cho and Cr (Table 4) could be quantified from all 32 parietal white matter spectra with mean fitting errors between 3.7 (%SD, NAA) and 8.1 (%SD, mI) in patients and 3.9 (%SD, NAA) and 8.7 (%SD, mI) for controls and showed no statistical significant difference between controls and patients. Severely depressed patients (n ⫽ 11) also did not display significant differences in either of the metabolites; however, 6/14 Glx and 8/14 Glu values in the patient group, as well as 7/18 Glx and 8/18 Glu values in the control group had to be discarded due to a high fitting error (20% SD or above). Again, no significant group effect for Glx and Glu was seen between patients and controls using MANOVA with and without covariate modelling for age, CSF and GM; however, a significant covariate effect could be seen for both Glx and Glu (p ⬍ .05). Individual univariate testing showed that CSF (t ⫽ .014 for Glx and t ⫽ .041 for Glu) and GM (t ⫽ .027 for Glx and t ⫽ .005 for Glu), but not age, had a significant effect on the variables Glx and Glu respectively.

Magnetic Resonance Imaging Results In the patient group (n ⫽ 19), 12 brain scans were considered normal. In the remaining seven, one scan showed minor white matter hyperintensities, four displayed increased CSF spaces (predominantly ventricular in two and predominantly cortical in two) and two showed an incidental vascular malformation. Both vascular malformations were located within the left frontal lobe and small (1–1.5 cm): radiological diagnosis was low-flow arterio-venous malformation (microangioma) in one and cavernoma in the other. In the control group there were minor age-related

Table 4. Metabolite Concentrations in Parietal White Matter Spectra

Controls (n ⫽ 18) All patients (n ⫽ 14) Severely depressed patients (n ⫽ 11)

Age (years)

mI (IU)

Cr (IU)

Cho (IU)

NAA (IU)

Glx (IU)

Glu (IU)

42.4 (13.1) 47.1 (11.3) 48.8 (11.6)

13.67 (1.83) 14.54 (1.58) 14.39 (1.62)

17.24 (2.24) 17.56 (2.19) 17.27 (2.18)

5.03 (0.65) 5.09 (0.91) 4.90 (0.63)

29.76 (2.49) 30.01 (2.66) 29.52 (2.71)

20.96 (3.05) 19.97 (3.80) 19.55 (3.99)

16.13 (1.03) 15.63 (1.59) 14.92 (0.91)

Group means (standard deviations). IU, institutional units. Abbreviation for metabolites as given in text.

310

BIOL PSYCHIATRY 2000;47:305–313

D.P. Auer et al

findings of enlargement of CSF spaces and small white matter hyperintensities, that were judged of no pathological significance. The presence of marked white matter hyperintensities and atrophy had been viewed as a contraindication in the recruitment phase. Only one volunteer with an incidental finding was not excluded from the control group in the study: a small asymptomatic occipital meningeoma without relevant space-occupying effect was present, but no morphological abnormality within the regions of interest for the spectroscopical studies could be detected.

Discussion 1

H-MRS allows regional analysis of brain metabolism, that can be helpful for differential diagnosis in focal and non-focal brain pathology. Furthermore, insight into underlying biochemical disorders may be offered, if the hypothetical pathophysiology involves one of the spectroscopically visible metabolites. In 1H-MRS at clinical field strength (mostly 1.5 T), visibility of metabolites is limited by several factors: low concentration (invisible below 1 mmol/l), relaxation enhancement (especially if bound to large macromolecular structures) and coupling effects in complex molecules, that all will reduce detectable signal amplitude. Additionally, chemical shift effects are field strength-dependent leading to considerable overlap of interesting metabolites at 1.5 T, that makes unequivocal assignment of resonance peaks difficult. For clinical studies only major metabolites may be analyzed: Glu, mostly unresolved from glutamine (Glx), both with overlapping contributions from GABA; NAA; mI, inseparable from glycine and betaine contributions as well as overlapping with glucose; Cho; and Cr.

Glutamate The major finding in this study was a significant decrease within the anterior cingulate of both the integral fit for Glx in depressed patients as well as Glx and Glu in severely depressed patients compared to age-matched healthy controls. Quantification and separation of glutamate and glutamine by use of 1H-MRS is technically difficult and no simple, routine method is available. Apart from general limitations in precise absolute spectroscopic quantification, these amino acids do not appear as single resonance peaks in the spectrum as NAA or Cho do, but are split into multiplets due to their chemical structures that, furthermore, overlap to a large extent in their chemical shift range (Figure 3) . Both factors lead to increased fitting errors for Glu, glutamine and GABA compared to other metabolites such as NAA, Cho and Cr despite high in vivo concentrations. Therefore, the combined Glx fitting and individual

Figure 3. Proton spectra of model solutions acquired at 1.5T (PRESS: TR/TE ⫽ 3000/35 ms). Largely overlapping multiplet signal structure can be seen. GABA, ␥-aminobutyric acid.

quantitation of Glu were performed, but not those for GABA or glutamine alone that would have resulted in insufficiently precise data. Undetected contributions from alterations in GABA levels can therefore not be ruled out in our findings. In fact, the larger difference seen in the group averaged spectra—that for methodical limitations represents only a qualitative result—may suggest additional differences in other resonances overlapping with Glx. Alterations of residual underlying resonances of macromolecules, however, are unlikely to account for the observed effect as about 90%–95% suppression can be expected at the chosen echo time of 35 msec as determined by T2 estimates provided in Behar et al (1994). The fact that both Glx and Glu showed significant reductions suggests that the effect is likely to be due to Glu, but not glutamine. The chosen threshold (⬍20% fitting error ⫽ a reasonable error limit compared to accepted reproducibility standards for individual metabolites with multiplet structure) does, however, not allow the analysis of the observed mean changes on an individual basis and therefore, only major group effects could be studied and allowed for. These were more distinct for the severely depressed patients, as defined by clinical assessment, compared to the overall group, but no significance emerged from the correlation of severity, as measured by the overall Hamilton Rating Scale for Depression score, and Glu or Glx levels. There was a marked regional difference in the control group with Glx being 38% and Glu 32% lower in white matter compared to anterior cingulate spectra, containing mainly grey matter. In keeping with this, a statistical significant covariate effect was found for percent grey

1

H-MRS in Depression

matter content of individual voxels. The finding of higher concentrations for Glu in grey matter compared to white matter confirms data from recent spectroscopic studies and autoptic data (Kno¨rle et al 1997; Pan et al 1996; Pouwels and Frahm 1998). The regional high Glu and Glx levels cannot, however, be explained by the high neuronal content alone, as comparison values for parietal cortex with comparable grey matter content show lower Glx values (unpublished data, n ⫽ 24 volunteers: mean Glx ⫽ 22.8 IU, mean age ⫽ 33.7 years). Even on the basis of receptor distribution, this finding cannot easily be explained. To our knowledge, data from anterior cingulate cortex for in vivo Glu levels are not yet available. Even though variations in anatomical positioning have to be expected between subjects, a systematic bias of the patient group could be ruled out by tissue segmentation showing similar mean values and possible residual individual effects were taken into account by treating individual grey matter values as covariates. Similarly, CSF contamination was controlled for and would have been expected to affect total metabolite content rather than a single metabolite. The fact that no difference was found in Glu content in parietal white matter suggests a regional disease effect. This tentative interpretation, however, needs to be consolidated, as higher fitting errors within the parietal white matter resulted in a rather high drop-out rate that may have masked disease-specific changes. The pathophysiological role of Glu and other excitatory amino acids in patients with major depressive episode has not been definitively explored. Altamura has proved increased plasma and decreased platelet levels for taurine as well as Glu in medication-free depressed patients (Altamura et al 1993). In a later study, they demonstrated a highly significant separation of depressed patients from controls using linear discriminant analysis with high loadings for increased taurine and decreased glycine as well as increased Glu (Altamura et al 1995). In contrast to these peripheral measures, surgical specimens taken from the frontal lobes of patients with depression did not show any alteration in Glu or other studied amino acids (Francis et al 1989). Postmortem receptor studies are also controversial: the group led by Skolnick has shown a reduced high-affinity binding to N-methyl-D-aspartate (NMDA) receptors in the frontal cortex of suicide victims compared to age- and postmortem interval-matched controls (Nowak et al 1995) and also found that some antidepressants produce alterations in the NMDA receptor complex, suggesting that the pathophysiology and treatment of depression involves glutamatergic pathways (Paul et al 1994). This is in contradiction, however, to Holemans et al (1993) who did not find any differences in the binding sites of NMDA receptors in brains of depressed patients who had committed suicide. The presented spectroscopic data further suggest a yet undefined role of altered glutamatergic neurotransmission in

BIOL PSYCHIATRY 2000;47:305–313

311

mood disorders. Specifically, reduced Glu together with changes in other neurotransmitters within the anterior cingulate may be relevant to the development of symptoms found among patients with depression. Medication effects would, of course, offer an alternative interpretation of the observed findings in this as well as in most of the referenced studies. Most antidepressants show adaptational effects on the NMDA receptor in rodent brains (Paul et al 1994; Reynolds and Miller 1988). Therefore, a drug-induced effect may be responsible for the finding of reduced Glu in patients with depression, although much lower dosages than those used in the rodents were used in the study reported here. We included acutely depressed patients shortly after their admission into hospital and only 7 out of 19 were nonmedicated at the time of MR examination, but overall, mean treatment time was low in the study group. In addition, the effects of benzodiazepines have to be considered, but studies on drug effects on in vivo proton spectra in humans are limited. A preliminary study using lorazepam indicated a significant drug effect on Cho, Cr and mI that showed a trend toward higher metabolite levels with lorazepam than without (Davanzo et al 1997). Another study with midazolam showed an increase in lactate and mI in normal volunteers (Burau et al 1997). Neither study reported on Glu effects and their results were largely dissimilar except for mI, so benzodiazepine treatment is unlikely to have caused the reported Glu decrease.

N-Acetyl-Aspartate The current concept of NAA is that of a neuronal marker due to its purely intraneural distribution in the adult brain (Urenjak et al 1992). Confirmatory results are available from a number of spectroscopic studies in focal, destructive as well as neurodegenerative diseases (see review Moats et al 1995). As there was no difference in NAA levels in either location, regional neuronal loss or changes in neuronal metabolism resulting in decreased NAA production can be ruled out in patients having an episode of major depressive disorder. Similarly, there was no difference in CSF content of voxels from patients and controls, that further evidence against regional brain atrophy and is in keeping with morphometric imaging studies (Coffey et al 1993). Recently reported findings, however, of hippocampal atrophy (Sheline 1996) warrants further studies on hippocampal NAA in patients suffering from major depressive episodes to clarify the origin of the observed changes in volume.

Myo-inositol Interest in mI has been raised by a possible link between the efficacy of lithium in mania and bipolar patients in reducing mI levels by blocking inositol monophosphatase

312

D.P. Auer et al

BIOL PSYCHIATRY 2000;47:305–313

and studies showing reduced mI levels in patients with depression (Barkai et al 1978). Furthermore, a number of treatment studies provide evidence for a therapeutic action of mI in major depressive episodes (Levine 1997). In two patients with depression related to Taxol treatment 1HMRS showed a 23–27% decrease in mI levels (Cousins et al 1996). To our knowledge, the current study is the first to report on localized cerebral mI levels in patients with a major depressive episode. Whereas regional mI levels within the parietal white matter and anterior cingulate were not dissimilar from control values, mI levels did vary according to region and showed a 21% mean increase in the anterior cingulate compared to levels in the white matter in the control group. Group differences in tissue composition of spectroscopic voxels could be ruled out by voxel segmentation and are thus highly unlikely to have masked a relevant mI decrease in the patient group. On the other hand, Levine et al (1996) who reported on CSF levels in non-medicated depressed patients and found no differences between a control group and the patients either, was also unable to show a predictive effect on clinical response to mI treatment. Therefore, within the limitations regarding drug treatment effects, a specific pathogenetic role of mI in mood disorders in general cannot be supported by the current study.

Choline As first proposed by Janowsky, depression may be causally linked to a central cholinergic predominance (Janowsky et al 1972). Among a large body of evidence supporting cholinergic mechanisms in mood disorders, induction of depressed mood has been shown by a number of cholinomimetics (Janowsky and Overstreet 1995). Previous spectroscopic studies have focused on basal ganglia and yielded opposing results: a state-dependent increase in the Cho/Cr ratio was described by Charles et al (1994) in drug-free patients with depression, whereas Renshaw et al (1997) found a lower Cho/Cr in a larger group of unmedicated depressed patients. The latter effect was more pronounced in treatment responders, but it remains to be studied whether drug treatment specifically affects Cho/Cr ratios. We could not substantiate Cho decrease or increase in our study either within parietal white matter or within the anterior cingulate, that may be related to regional differences and also to differences in data analysis. Absolute quantification schemes are robust against possible alterations in Cr levels and even though we did not attempt to estimate mmol/l concentrations, institutional units as used in this study offer absolute values and comparability between patients studied on the same system. Second, basal ganglia spectra are prone to broader line widths due to susceptibility effects, that further complicate adequate

quantification. Considerable overlap between Cho and Cr could be seen in the spectra displayed in both studies. In summary, this study provides evidence for reduced Glu concentrations in depressed patients within the anterior cingulate and thus supports the hypothesis of a glutamatergic role in the expression of mood disorders. Spectroscopic quantification of Glu, especially its separation from glutamine and GABA at clinical field strength is methodically still challenging. Therefore, the presented data need to be substantiated by further studies using improved techniques such as editing, J-refocused coherence transfer or C-PRESS methods (Hennig et al 1997; Hurd et al 1998; Keltner et al 1996; Lee et al 1995; Pan et al 1996; Petroff et al 1995). Also, the possible effects of regional distribution, medication, severity and duration of disease as well as potential reversibility need to be studied and will help to elucidate the pathophysiological role of glutamatergic systems in depression.

The authors would like to thank Dr. C. Heese for psychological assessment, A. Hofbauer, and Dr. T. Schirmer for technical assistance in spectra processing.

References Altamura CA, Mauri MC, Ferrara A, Moro AR, D’Andrea G, Zamberlan F (1993): Plasma and platelet excitatory amino acids in psychiatric disorders. Am J Psychiatry 150:1731– 1733. Altamura C, Maes M, Dai J, Meltzer HY (1995): Plasma concentrations of excitatory amino acids, serine, glycine, taurine and histidine in major depression. Eur Neuropsychopharmacol 5(suppl):71–75. Barkai AI, Dunner DL, Gross HA, Mayo P, Fieve RR (1978): Reduced myo-inositol levels in cerebrospinal fluid from patients with affective disorder. Biol Psychiatry 13:65–72. Behar KL, Rothman DL, Spencer DD, Petroff OAC (1994): Analysis of macromolecule resonances in 1H-NMR spectra of human brain. Magn Reson Med 32:294 –302. Bench CJ, Frackowiak RS, Dolan RJ (1995): Changes in regional cerebral blood flow on recovery from depression. Psychol Med 25:247–261. Burau T, Schilling AM, Seyfert S, Spies C, Wolf K (1997, April): Cerebral drug effects investigated by 1HMR-spectroscopy in volunteers and patients. Poster presented at the annual meeting of the International Society for Magnetic Resonance in Medicine, Vancouver, Canada. Charles HC, Lazeyras F, Krishnan KR, Boyko OB, Payne M, Moore D (1994): Brain choline in depression: in vivo detection of potential pharmacodynamic effects of antidepressant therapy using hydrogen localized spectroscopy. Prog Neuropsychopharmacol Biol Psychiatry 18:1121–1127. Coffey CE, Wilkinson WE, Weiner RD, Parashos IA, Djang WT, Webb MC, et al (1993): Quantitative cerebral anatomy in depression. A controlled magnetic resonance imaging study. Arch Gen Psychiatry 50:7–16.

1

H-MRS in Depression

Cousins JP, Harper G (1996): Neurobiochemical changes from Taxol/Neupogen chemotherapy for metastatic breast carcinoma corresponds with suicidal depression. Cancer Lett 110:163–167. Davanzo P, Oshiro T, Thomas MA, Shah B, Belin T, McCracken J, et al (1997, April): 1H-MR spectroscopy in human brain with and without lorazepam. Poster presented at the annual meeting of the International Society for Magnetic Resonance in Medicine, Vancouver, Canada. Ebert D, Ebmeier KP (1996): The role of the cingulate gyrus in depression–-from functional anatomy to neurochemistry. Biol Psychiatry 39:1044 –1050. Francis PT, Poynton A, Lowe SL, Najlerahim A, Bridges PK, Bartlett JR, et al (1989): Brain amino acid concentrations and Ca2⫹-dependent release in intractable depression assessed antemortem. Brain Res 494:315–244. Goodwin GM (1996): Functional imaging, affective disorder and dementia. Br Med Bull 52:495–512. Hennig J, Thiel T, Speck O (1997): Improved sensitivity to overlapping multiplet signals in in vivo proton spectroscopy using a multiecho volume selective (CPRESS) experiment. Magn Reson Med 37:816 – 820. Holemans S, De Paermentier F, Horton RW, Crompton MR, Katona CL, Maloteaux JM (1993): NMDA glutamatergic receptors, labeled with [3H]MK-801, in brain samples from drug-free depressed suicides. Brain Res 616:138 –143. Holsboer F, Lauer CJ, Schreiber W, Krieg JC (1995): Altered hypothalamic-pituitary-adrenocortical regulation in healthy subjects at high familial risk for affective disorders. Neuroendocrinology 62:340 –347. Hurd RE, Gurr D, Sailasuta N (1998): Proton spectroscopy without water suppression–-the oversampled J-resolved experiment. Magn Reson Med 40:343–347. Janowsky DS, Yousef MK, Davis JM, Sekerke JH (1972): A cholinergic-adrenergic hypothesis of mania and depression. Lancet 2:632– 635. Janowsky DS, Overstreet DH (1995): The role of acetylcholine mechanisms in mood disorders. In: Bloom FE, Kupffer DJ, editors. Psychopharmacology: The Fourth Generation of Progress, New York: Raven Press, pp 945–956. Keltner JR, Wald LL, Christensen JD, Maas LC, Moore CM, Cohen BM, et al (1996): A technique for detecting GABA in the human brain with PRESS localization and optimized refocusing spectral editing radiofrequency pulses. Magn Reson Med 36:458 – 461. Kno¨rle R, Assmann D, Landwehrmeyer B, Scheremet R, Mu¨ller K, Feuerstein J (1997): Aspartate, glutamate, glutamine, glycine and GABA in human bioptic neocortical areas: comparison to autoptic tissue. Neurosci Lett 221: 169 –172. Larisch R, Klimke A, Vosberg H, Loffler S, Gaebel W, Mullergartner HW (1997): In vivo evidence for the involvement of dopamine-D-2 receptors in striatum and anterior cingulate gyrus in major depression. Neuroimage 5:251–260. Lee HK, Yaman A, Nalcioglu O (1995): Homonuclear J-resolved spectral editing technique for quantification of glutamine and

BIOL PSYCHIATRY 2000;47:305–313

313

glutamate by 1H NMR spectroscopy. Magn Reson Med 34:253–259. Levine J, Kurtzman L, Rapoport A, Zimmerman J, Bersudsky Y, Shapiro J, et al (1996): CSF inositol does not predict antidepressant response to inositol. J Neural Transm 103: 1457–1462. Levine J (1997): Controlled trials of inositol in psychiatry. Eur Neuropsychopharmacol 7:147–155. Meltzer HY, Lowy MT (1987): The serotonin hypothesis of depression. In: Meltzer HY, editor. Psychopharmacology: The Third Generation of Progress, New York: Raven Press, pp 513–526. Moats RA, Watson L, Shonk T, Tokuyama S, Braslau D, Eto R, et al (1995): Added value of automated clinical proton MR spectroscopy of the brain. J Comput Assist Tomogr 19:480 – 491. Nowak G, Ordway GA, Paul IA (1995): Alterations in the N-methyl-D-aspartate (NMDA) receptor complex in the frontal cortex of suicide victims. Brain Res 675:157–164. Pan JW, Mason GF, Pohost GM, Hetherington HP (1996): Spectroscopic imaging of human brain glutamate by watersuppressed J-refocused coherence transfer at 4.1 T. Magn Reson Med 36:7–12. Paul IA, Nowak G, Layer RT, Popik P, Skolnick P (1994): Adaptation of N-methyl-D-aspartate receptor complex following chronic antidepressant treatments. J Pharmacol Exp Ther 269:95–102. Petroff OAC, Rothman DL, Behar KL, Mattson RH (1995): Initial observations on effect of vigabatrin on in vivo 1H spectroscopic measurements of gamma-aminobutyric acid, glutamate, and glutamine in human brain. Epilepsia 36:457– 464. Pouwels PJW, Frahm J (1998): Regional metabolite concentrations in human brain as determined by quantitative localized proton MRS. Magn Reson Med 39:53– 60. Provencher SW (1993): Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn Reson Med 30:672– 679. Renshaw PF, Lafer B, Babb SM, Fava M, Stoll AL, Christensen JD, et al (1997): Basal ganglia choline levels in depression and response to fluoxetine treatment–-an in vivo proton magnetic resonance spectroscopy study. Biol Psychiatry 41: 837– 843. Reynolds IJ, Miller RJ (1988): Tricyclic antidepressants block N-methyl-D-aspartate receptors: similarities to the action of zinc. Br J Pharmacol 95:95–102. Schildkraut JJ (1965): The catecholamine hypothesis of affective disorders: A review of supporting evidence. Am J Psychiatry 122:509 –522. Sheline YI (1996): Hippocampal atrophy in major depression: A result of depression-induced neurotoxicity? Mol Psychiatry 1:298 –299. Urenjak J, Williams SR, Gadian DG, Noble M (1992): Specific expression of N-acetylaspartate in neurons, oligodendrocyte type 2 astrocyte progenitors, and immature oligodendrocytes in vitro. J Neurochem 29:55– 62.