Subtype-Specific Alterations of γ-Aminobutyric Acid and Glutamatein Patients With Major Depression FREE

Several lines of evidence from animal and human studies suggest thatthe γ-aminobutyric acid (GABA) system contributes to the neurobiologicalfeatures underlying major depressive disorder (MDD). This subject has beenexamined by several researchers.^1- 5 Collectively,the studies suggest MDD is associated with deficits in GABAergic function.This hypothesis has been clearly supported by consistent findings of reducedGABA content in the plasma and cerebrospinal fluid of depressed individualsrelative to comparison subjects.^6- 12

Consistent with reports of widespread reductions in GABA content, invivo proton magnetic resonance spectroscopy (hydrogen 1 [¹H]–MRS)has more recently been used to demonstrate reduced occipital cortex GABA concentrationsin a sample of severely depressed subjects.¹³ Otherrecent studies^14- 15 demonstratingthat treatment of depression with electroconvulsive therapy or selective serotoninreuptake inhibitor agents increases occipital cortex GABA concentrations inMDD patients suggest that GABA abnormalities may be normalized following treatment.If reproducible, these findings significantly strengthen the association betweenimpaired GABAergic function and MDD, and may provide a novel perspective intothe pathophysiological processes related to MDD. The primary objectives ofthis study were as follows: (1) to replicate the original findings of reducedGABA levels in the occipital cortex of MDD subjects in a larger sample, witha broader range of depression severity, controlling for cortical volume andtissue composition; (2) to expand the analysis to include glutamate and othermetabolites detectable by ¹H-MRS; and (3) to explore potentialclinical correlates to the differences in GABA and metabolite concentrations.

All subjects provided written informed consent using forms and proceduresapproved by the Yale Human Investigations Committee. Forty-four depressedsubjects meeting DSM-IV criteria for MDD based onthe Structured Clinical Interview for DSM-IV AxisI Disorders: Patient Edition¹⁶ were recruitedthrough newspaper advertisement and referrals from community physicians. Anactive substance abuse disorder within the 6 months preceding study enrollmentwas considered an exclusion criterion. Subjects who were taking medicationfor the treatment of their depression completed a minimum 2-week medicationwashout before spectroscopy, with only diphenhydramine hydrochloride, 25 to50 mg, being administered for insomnia.

The 38 healthy control subjects had no personal, or first-degree family,history of MDD or other DSM-IV diagnoses, per interview.

Studies were performed with a 2.1-T magnet (Oxford Magnet Technology,Oxford, England) with a 1-m bore and a spectrometer (Avance; Bruker Instruments,Billerica, Mass). Subjects lay supine with the occipital surface of the headagainst a 7-cm distributed capacitance surface coil. Gradient-echo scout imageswere acquired from 5-mm-thick slices, with a 1-cm center-to-center slice separationand a field of view of 240 mm divided into 128 × 128 pixels. Noniterativeshimming was performed using an in vivo automatic method (FASTERMAP)¹⁷ in a 3.4-cm-diameter spherical volume. A 3.0 ×3.0 × 1.5-cm volume was selected across the midline of the brain, centered2 cm from the dura. Localization was achieved with selective excitation; 3-dimensional,image-selected, in vivo spectroscopy; outer volume suppression; and a surfacespoiler. After adjusting the radiofrequency power levels, a J-editing sequence^18- 19 yields pairs of subspectra that aresubtracted to obtain the edited GABA signal. Briefly, one subspectrum wasacquired with an inversion pulse applied to the GABA C4 proton and one subspectrumwas acquired without the inversion pulse. The phase of the coupled GABA C3proton resonance was inverted in one subspectrum relative to the other, butthe other, uncoupled, resonances in the region (creatine and choline) werethe same in both spectra. When the 2 subspectra were subtracted, creatineand choline vanished, leaving GABA. The data were acquired in interleavedfashion, toggling between 64-second sets of acquisitions, a set with the inversionpulse and a set without the inversion pulse. The detection was run for 20minutes to yield 9 pairs of subspectra. Subspectra were acquired with 8000data points in a 510-millisecond acquisition, with a 2-second repetition timeand an echo time of 68 milliseconds.

By using technical computing software (MATLAB; The Mathworks, Inc, Natick,Mass), each sub-free induction decay was line broadened by 3 Hz, zero filledto 32 000 points, and Fourier transformed. The subspectra were phasebased on the spectrum with the inversion pulse applied, and the area of thecreatine resonance was measured. The 2 subspectra were subtracted, and thearea of the GABA resonance was measured. If patients moved during any of the9 sets of GABA editing scans, the much larger and sharper creatine and cholineresonances caused well-defined subtraction errors that prevented the measurementof GABA, and any pairs of subspectra with such patient movement were not processedfurther. The level of GABA was determined by adding together the scans thatshowed no signs of patient motion. The concentration of GABA was determinedfrom the ratio of GABA–total creatine resonances according to thefollowing¹⁸:

[GABA-] = (Area_GABA/Area_Creatine − 0.07) × (0.93 × 1.01) × (3/2) × [Creatine]

where 0.07 was subtracted to account for the contribution ofcoedited macromolecules^20- 23 tothe GABA resonance, 0.93 is the integral correction factor for the differencein integration limits due to the 0.03–part per million difference inthe chemical shift of creatine and GABA, 1.01 corrects for the editing efficiency,and 3/2 adjusts for the difference of 2 protons detected in GABA vs 3 in creatine.The concentration of creatine was 9 mmol/kg.²⁴

In 55 subjects, spectroscopic measurements of N-acetylaspartate(NAA) and other metabolites were obtained in the same voxel. Some subjectswere not able to tolerate the additional time in the scanner. The sequencewas applied in pairs to yield subspectra that contain either macromoleculesalone or macromolecules combined with the metabolites of interest.²⁵ Briefly, one subspectrum was acquired with a hyperbolicsecant inversion pulse applied across the spectral bandwidth, followed byan inversion-recovery delay to null the metabolites, and one subspectrum wasacquired without the inversion pulse. The macromolecule subspectrum was subtractedfrom the other subspectrum to obtain a spectrum of metabolites alone. Thedata were acquired in interleaved fashion, toggling between individual 48-secondsets of inverted and uninverted acquisitions. Each block was stored, and thesequence run for 20 minutes to yield 16 pairs of subspectra. The subspectrawere acquired with 8000 data points in a 510-millisecond acquisition, witha 5-second repetition time and an echo time of 12 milliseconds. A scan ofunsuppressed tissue water was also acquired to evaluate the absolute levelof creatine and for eddy current correction of the short-echo sequence. Byusing technical computing software (MATLAB), each sub-free induction decaywas processed using a lorentzian-to-gaussian conversion of −1 and 6Hz, Fourier transformation, spectral phasing, and subtraction. The peak amplitudesat 3.23, 3.03, 2.75, 2.60, 2.45, 2.29, and 2.02 parts per million were measuredand deconvolved using model spectra obtained in solutions of GABA, creatine,glutamate, glutamine, aspartate, and NAA to determine the metabolite levelsrelative to the resonance of total creatine, deconvolving GABA using its concentrationfrom the subject's J-edited GABA measurement. The concentrations were determinedassuming a concentration of 9 mmol/kg for creatine.²⁴

To account for potential changes in tissue composition, a series of3-mm-thick contiguous images of T1 were used to quantify the amount of graymatter, white matter, and cerebrospinal fluid in the voxel of interest.²⁶ The images of T1 were measured using a series ofinversion-recovery images that required images of the spatial distributionof the radiofrequency power to overcome the problems associated with radiofrequencyinhomogeneity.^27- 28 The concentrationsof GABA, glutamate, and other metabolites were compared for the 2 groups.

The primary objective of this study was to replicate the finding ofreduced occipital cortex GABA concentrations associated with MDD.¹³ The data for the 71 subjects were analyzed usingan analysis of covariance (ANCOVA), with GABA levels as the response variableand diagnosis (depressed vs healthy) as the main predictor variable. Percentagetissue, percentage white matter, and NAA level were individually consideredas covariates, then dropped from the model because they were nonsignificantat the P = .05 level. Analysis of covariance wasalso used to evaluate the effect of diagnosis on the levels of glutamine,glutamate, NAA, aspartate, and choline, the GABA-glutamate ratio, and thepercentages of white matter and tissue. The percentages of white matter andtissue and the NAA level were considered as covariates in the models for glutamate,glutamine, and GABA-glutamate ratio. Age and sex were considered as covariatesin all models, but when not significant, they were dropped from the models.

Correlations among metabolite levels and measures of tissue compositionwere computed for depressed subjects and healthy controls (N = 71). Correlationsof metabolites and measures of tissue composition with Hamilton DepressionRating Scale scores were computed only for depressed subjects (n = 33). Bonferronicorrection was applied.

Most subjects in this study were outpatients with disease in the mildto moderate range of severity, recruited through advertisement, while mostsubjects in the original study by Sanacora et al¹³ wererecruited from an inpatient research facility and had much more severe levelsof depression. We combined the data sets to broaden the range of subjectsin an effort to study potential differences in cortical GABA concentrationbetween various subtypes and levels of depression severity (total sample size,105: 71 from study 2 and 34 from study 1). Analysis of covariance was usedto evaluate the GABA measurements, using DSM-IV defined(Structured Clinical Interview for DSM-IV Axis IDisorders: Patient Edition–confirmed) subtype as a cofactor, where thesubtypes were melancholic depressed, atypical depressed, depressed with nosubtype, and healthy control. Age and sex were included as covariates. Theeffect of the study (ie, previously published data vs new data) was also includedas a cofactor because data from the old study could differ from data fromthe new study, because of population differences or the technical upgradesin the magnetic resonance system. Multiple comparisons among subtypes wereperformed using the Tukey-Kramer method.

All statistical tests were performed using α = .05, with the Bonferronicorrection when appropriate, and SAS statistical software, version 8.2 (SASInstitute Inc, Cary, NC). Data are given as mean ± SD unless otherwiseindicated.

Of the 44 depressed subjects enrolled in the study, 33 successfullycompleted the baseline GABA ¹H-MRS study, resulting in spectraof acceptable quality for evaluation. Two subjects were dropped from the studyafter being given medications by another physician between the time of enrollmentand the ¹H-MRS study. Three subjects aborted the study while inthe magnet because of anxiety and claustrophobia, and data were lost for 6subjects due to either movement artifact or system malfunction (primarilysoftware errors). Twenty-nine subjects completed the short-echo ¹H-MRSstudy, providing quality spectra allowing for measures of other metabolites,including glutamate. Thirty-eight healthy controls completed the GABA studyand 28 completed the short-echo study.

The mean age of the depressed subjects was 41.87 ± 9.88 years,and was significantly higher than that of the controls (35.74 ± 11.40years; t₆₉ = −2.40, P = .02). The median (range) ages were 41.13 (19.08-57.13) and 31.98(19.51-62.60) years, respectively. There was also a trend for men to composemore of the depressed group (22 [67%] of 33 subjects) than the control group(15 [39%] of 38 subjects) (χ²₁ = 5.23, P = .02). Therefore, age and sex were considered as covariates in allfurther analyses.

Depression ranged from mild to severe, as reflected by the 25-item HamiltonDepression Rating Scale scores. A complete summary of the patients' clinicalhistories is provided in Table 1.

Magnetic resonance spectroscopic measures of GABA levels (Figure 1) were analyzed for an effect of diagnosis. As shown in Figure 2, the magnetic resonance spectrashowed that the concentration of cortical GABA was significantly lower fordepressed subjects (1.20 ± 0.42 mmol/kg) than for controls (1.42 ±0.28 mmol/kg) (F_1,69 = 6.79, P = .01).The percentages of tissue (P = .26) and white matter(P = .36), the NAA level (P =.80), age (P = .83), and sex (P = .97) did not have a significant effect on GABA levels. Of specificinterest was the fact that only 1 of the 38 healthy controls had GABA concentrationsbelow 1 mmol/kg of brain tissue, while 15 of the 33 depressed subjects hadGABA concentrations in this range. There was no evidence to suggest a medicationwashout effect is related to the decreased GABA content because the GABA concentrationfor the 8 subjects who required a washout was actually higher than for theMDD group as a whole (1.40 ± 0.50 mmol/kg of tissue).

Levels of glutamate, glutamine, NAA, aspartate, and choline, the GABA-glutamateratio, and tissue composition were evaluated for an effect of diagnosis. Glutamateconcentrations were significantly higher in the patients (9.20 ± 1.26mmol/kg) than in the control subjects (7.99 ± 1.13 mmol/kg) (ANCOVAF_1,54 = 29.8, P<.001) (Figure 3), with no effect of age or sex but a significant effectof NAA (t₅₄ = 4.72, P<.001). The ratio of GABA-glutamate was also significantly higherin patients than in control subjects (ANCOVA F_1,55 = 10.3, P = .002). Chemicals other than glutamate and GABA didnot show significant differences at P = .05. AfterBonferroni correction (P = .05/8), the glutamatelevel and the GABA-glutamate ratio were still significantly (P = .001) higher for patients than for controls.

The NAA concentration was not significantly different between the depressedand healthy groups (P = .16). Age was significantlyassociated with NAA concentrations (ANCOVA F_1,55 = 4.69, P = .03). Sex had a significant effect on choline level(ANCOVA F_1,55 = 6.21, P = .01), with womenand men having choline levels of 1.48 ± 0.15 and 1.39 ± 0.12mmol/kg, respectively.

The ratio of creatine–tissue water (×100 000) was 0.27± 0.03 and 0.26 ± 0.06, respectively, in the 27 control subjectsand 28 patients from whom the water measure was obtained. A 2-sample t test showed no significant difference between groups(P = .66). Pairwise t testsof the subtypes showed no significant difference in the ratio of creatine–tissuewater among control subjects and subtypes (lowest value, P = .67).

Depressed subjects had significantly less solid tissue in the voxel(93.9% ± 4.1%) than did controls (96.2% ± 2.2%) (ANCOVA F_1,65 = 7.22, P = .009), with neither age norsex reaching a significance level of .05. Voxels from depressed subjects alsocontained a significantly smaller proportion of white matter (36.3% ±5.4%) than from the healthy controls (39.7% ± 7.1%) (ANCOVA F_1,64 = 7.32, P = .009), even after controllingfor age (ANCOVA F_1,64 = 4.44, P = .04),and a significantly larger portion of gray matter (ANCOVA F_1,64 =6.25, P = .02) controlling for age (ANCOVA F_1,64 = 7.29, P = .009) (Table 2).

After Bonferroni correction for multiple tests (P = .05/45), only 3 significant correlations remained between the variablesin this study. These were the negative correlation between GABA and glutamate(r = −0.60, P<.001)(Figure 4), the positive correlationbetween percentage tissue and percentage white matter in the MRS voxel (r = 0.50, P<.001), and thepositive correlation between NAA and aspartate (r =0.60, P<.001). The latter relationship was mostlikely accounted for by spectral overlap of the metabolites, which, therefore,confounded any biological meaning of the correlation. Although not meetingthe conservative Bonferroni correction test for significance, there were alsocorrelations at the .01 significance level between GABA and glutamine (r = 0.37, P = .005), between NAAand glutamate (r = 0.36, P =.006), and between NAA and percentage tissue (r =0.40, P = .003).

By combining the data from this study with the data from the initialpreviously published study,¹³ we were ableto increase our sample size sufficiently to allow us to better explore clinicalcorrelates, including the difference between the various DSM-IV–defined subtypes of MDD.²⁹ Multiplelinear regression analysis was used to evaluate the GABA measurements in theold study and the present data, looking at melancholic depression, atypicaldepression, depression with no subtype, and control subjects, controllingfor age, sex, and a study effect. By using this model, we found a significantsubtype effect (ANCOVA F_3,100 = 11.09, P<.001),and a multiple comparison by Tukey-Kramer showed that the GABA levels weresignificantly lower in patients with melancholic depression compared withcontrols (P<.001) and in patients with no subtypecompared with controls (P = .002), but not in patientswith atypical depression (Figure 5).Interestingly, the melancholic subjects with psychotic features all had GABAconcentrations below 0.8 mmol/kg of tissue. There was also a significant studyeffect (ANCOVA F_1,100 = 5.08, P = .03),but this was largely accounted for by differences in the depressed group becausethere was no difference in GABA concentrations between the control subjectsfrom study 1 (1.45 ± 0.39 mmol/kg) and from study 2 (1.42 ±0.28 mmol/kg). Pairwise t tests of the subtypes didnot show any significant differences in the ratio of creatine–tissuewater among controls and subtypes (lowest value, P =.67).

There was no correlation between cortical GABA concentrations and theHamilton Depression Rating Scale score based only on the present study (n= 33) (Pearson product moment correlation coefficient r = −0.15, P = .41). There was a trendbased only on the subjects in the original study (n = 14) (Pearson productmoment correlation coefficient r = −0.51, P = .07). Based on the combined data set, there is a significantcorrelation (r = −0.38, P = .01). However, controlling for the study (Pearson product momentpartial correlation coefficient r = −0.22, P = .15) reduces the finding back to a trend level.

By using the short-echo data collected primarily in the second study,glutamate differed significantly among the 4 groups (ANCOVA F_3,59 =8.59, P<.001). The multiple comparisons by Tukey-Kramershowed significant differences in glutamate between patients with melancholicdepression and controls (P<.001) and between patientswith no subtype and controls (P = .03) (Figure 6). The comparison between the melancholicdepressed group and the no subtype group was a trend (P = .05). Patients with atypical depression did not have significantlydifferent GABA (P = .10) and glutamate (P = .11) levels when compared with controls and patients not meetingeither subtype criteria. However, this is likely because of the limited samplesize.

The present study yielded 3 principal findings. (1) A reduction in occipitalcortical GABA levels was found in a new and less severely ill group of patientswith MDD, controlling for differences in cortical tissue composition. (2)Increased cortical glutamate levels and GABA-glutamate ratios were observedin the occipital cortex of MDD subjects. (3) The reductions in cortical GABAlevels and increased glutamate levels were particularly associated with melancholicand psychotic features in the MDD subjects.

The finding of significantly reduced occipital cortex GABA concentrationsin the depressed subjects relative to the control subjects is consistent withthat of the previous ¹H-MRS study. The difference in the magnitudeof the mean reduction between the 2 studies is likely related to differencesin the clinical characteristics of the MDD subjects recruited into the 2 studies.The first study examined a cohort of more severely depressed subjects recruitedmostly from an inpatient psychiatric unit. The GABA levels in these patientswere reduced by 52%, with almost no overlap between the MDD and control subjects'levels. Subjects in the present study, recruited primarily through newspaperadvertisements, included a broader range of illness severity, and showed asmaller group difference in GABA levels. However, the many depressed subjectscompared with control subjects with GABA concentrations below 1 mmol/kg ofbrain tissue suggests the existence of a subgroup of depressed subjects withmarkedly decreased levels of GABA in the occipital cortex, similar to thosein the first study. Thus, we believe the greater heterogeneity among the clinicalcharacteristics of the MDD subjects in the second sample likely explains thelarger range and higher mean GABA concentrations that were observed in thestudy.

Analyzing the combined data sets from the original and present studiessupports this hypothesis. The post hoc analysis of the combined data set providesstrong evidence that differences exist in cortical GABA concentrations betweensubsets of depressed subjects. Depressed subjects with melancholic featuresseem to have the largest and most consistent GABA reductions. This seems tobe especially clear in the subset of melancholic subjects who also have psychoticfeatures. In contrast, normal or near-normal GABA concentrations were foundin most atypically depressed subjects. This finding is not completely unprecedented.Roy et al¹¹ reported only finding a significantdifference in cerebrospinal fluid GABA concentrations between depressed andcontrol subjects when they limited the comparison to the 13 melancholic patientsin the study and the controls. Petty et al⁶ alsoreported that the lowest plasma GABA concentrations appeared in subjects likelyto meet the current diagnosis of melancholic depression.

The addition of a short-echo scan to the protocol allowed us to collectother measures in addition to GABA level in this study. Of all the other compoundsexamined, only glutamate was significantly different between the groups aftercontrolling for multiple comparisons. Because of the increased noise and significantoverlap with glutamine, the measurement of glutamate is associated with increasedvariability in the concentrations obtained. However, there again seems tobe a maximum value that is rarely breached by control subjects but frequentlyviolated in a subgroup of depressed subjects. Again, most markedly outlyingconcentrations were from the subjects with the melancholic subtype of MDD.This finding is of interest in light of the fact that glutamatergic abnormalitieshave previously been reported in several studies^30- 35 ofdepressed subjects. Although these studies have failed to show a consistentdirection to the glutamate abnormalities in the varied brain regions examined,the growing series of studies^36- 41 demonstratingantidepressant effects in several agents with antiglutamatergic activity furthersuggest that the abnormal glutamate concentrations may have clinical relevance.

In sum, these results suggest the existence of a subgroup of MDD subjectswith coexistent abnormalities in the major excitatory and inhibitory neurotransmittersystems. These findings add further to the mounting evidence suggesting thatGABA and glutamate contribute to the pathophysiological features and treatmentof mood disorders.⁴² The large increase inthe ratio of GABA-glutamate in this subgroup of MDD subjects may have significantimplications regarding neurophysiological processes, such as cortical excitability,and potential excitotoxicity that may be directly related to the pathophysiologicalfeatures and pathogenesis of the disorder.

We are left to speculate on the mechanisms that could account for thecombined abnormalities in the amino acid neurotransmitter systems. The metabolicpathways regulating the synthesis and cycling of GABA and glutamate are tightlycoupled,^43- 45 suggestingthat a single alteration in a shared pathway may account for the elevatedglutamate and reduced GABA concentrations. The growing number of postmortemstudies demonstrating decreased glial cell number and density associated withMDD^46- 54 providesone interesting possibility to account for the observed amino acid abnormalities.Astrocytes play a critical role in the function and regulation of the aminoacid neurotransmitter systems. They provide the primary source of energy toneurons,⁵⁵ and furnish the major pathway forneuronal glutamate and GABA synthesis.^45,56 Reports^57- 59 of abnormally elevatedlevels of S100B, a calcium-binding peptide produced by astrocytes, in patientswith melancholic depression further support a potential relationship amongastrocyte pathological features, altered amino acid neurotransmitter function,and MDD.

Decreased astrocytic function would be expected to result in decreasedflux through the glutamate-glutamine cycle and, because glutamine generatedthrough this cycle is the primary precursor of GABA synthesis,⁴⁵ decreasedGABA synthesis as well. This could account for the consistent findings ofreduced GABA concentrations in the brain, cerebrospinal fluid, and plasmaspecimens of depressed subjects. What effect this would have on static glutamateand glutamine levels is much more difficult to predict because release mayalso be significantly reduced in this situation, resulting in increased intracellularstores. Unfortunately, the ¹H-MRS measures used in this study provideonly a static picture of metabolites measured, and do not allow us to determinerates of metabolism or cycling that may help to further isolate a specificsite or sites in the metabolic pathways that may be disrupted. However, preliminaryresults using carbon 13 (¹³C)–MRS measures of glucose incorporationinto the carbon skeleton of the metabolites suggest that the rates of GABAsynthesis and glutamate-glutamine cycling are reduced in depressed subjects,in a manner consistent with this hypothesis.⁶⁰ Futureexplorations using this approach are likely to provide additional informationregarding the specific mechanisms related to the observed abnormalities associatedwith MDD.

A recent preliminary report⁶¹ indicatingthat the glial reductions in the amygdala of depressed subjects are due, atleast in part, to decreases in oligodendrocytic density suggests that morethan one glial cell type may be altered in patients with depression. The potentialloss of oligodendrocytic material is consistent with the significantly lowercontent of white matter that was observed in the depressed subjects, and maysuggest more diffuse glial cell pathological features. Interestingly, we alsoobserved an increased percent gray matter in the depressed subjects. Whileunanticipated, this finding is consistent with a recent report by Ballmaieret al,⁶² and may be related to complex structuralchanges resulting in altered T₁ relaxation properties of the tissue.

The reciprocal relationship between glutamate and GABA that is seencould also be the result of other causes of reduced glutamatergic function.In the context of reports of reduced glucose use^63- 64 andreduced cortical activity⁶⁵ associated withMDD, it may be that the low GABA level is related to long-term decreased glutamaterelease. A similar phenomenon has previously been observed following lightdeprivation⁶⁶ and deafferation.⁶⁷ Itis possible that lower glutamatergic activity leads to a reduction in corticalGABA level by a regulatory metabolic process, such as changes in the saturationof enzymes involved in neurotransmitter cycling or control mechanisms imposedby the brain in reaction to reduced glutamate release.^68- 69

These data must be interpreted in light of several limitations. Somedata were lost because of poor spectra quality associated with movement artifact.This loss of data may have introduced a systematic error into the analysesif subjects who moved more tended to have some consistent level of corticalGABA. It is also possible that differences in the macromolecular contributionsto the spectra between the depressed and control subjects might account fordifferences in GABA and glutamate. However, it seems unlikely given that theonly reports of macromolecular changes are in diseases with severe structuraldamage, like multiple sclerosis and stroke.^70- 72 Wealso limited the region of interest to the occipital cortex because of technicallimitations in the method. This has not been a region thought to be closelyassociated with the symptoms most commonly characterizing MDD. However, recentstudies have demonstrated abnormalities in serotonin_1A receptorbinding^73- 74 and signal transductionpathways,⁷⁵ along with reduced metabolic activity⁷⁶ in the occipital cortex of depressed subjects. Nevertheless,the limitation of the region-specific analysis further limits our abilityto draw any causative relationships between the observed changes and the clinicalmanifestations of the disorder. Last, the ¹H-MRS technique onlyprovides static measures of metabolite concentrations and does not providespecific information about the synthesis or degradation of the metabolites.Future studies exploring additional brain regions, such as the prefrontalcortex, should provide information about the regional generalizability ofthe findings, and studies using ¹³C-MRS methods allow us to specificallyprobe the metabolic pathways regulating GABA and glutamate synthesis and degradationin the brain.

In conclusion, the findings of this study are consistent with thoseof a previous ¹H-MRS study demonstrating lower GABA concentrationsin the occipital cortex. The new observation of elevated glutamate levelsin the same region is consistent with emerging evidence that suggests bothamino acid neurotransmitter systems contribute to the pathophysiological featuresof MDD. It also suggests that a metabolic pathway common to both systems maybe a primary site of pathological features in MDD. Continued advances in ourunderstanding of the physiological features associated with these systemsare likely to provide new targets for investigation and may help guide futuredrug development. Last, the finding that GABA and glutamate abnormalitiesseem limited to a subset of MDD patients with identifiable clinical correlatessuggests that the ¹H-MRS measures may have potential value as diagnostictools.

Correspondence: Gerard Sanacora, MD, PhD, Clinical Neuroscience ResearchUnit, Abraham Ribicoff Research Facilities, Connecticut Mental Health Center,34 Park St, New Haven, CT 06519 (gerard.sanacora@yale.edu).

Submitted for publication November 12, 2003; final revision receivedFebruary 6, 2004; accepted February 13, 2004.

This study was supported by grant DF99-067 from The Patrick and CatherineWeldon Donaghue Medical Research Foundation (Dr Sanacora); grants K02 AA00261-01(Dr Krystal) and K02 AA13430-01 (Dr Mason) from the National Institute onAlcohol Abuse and Alcoholism, Bethesda, Md; the Veterans Affairs ResearchEnhancement Award Program (Drs Sanacora and Krystal); National Alliance forResearch on Schizophrenia and Depression, Great Neck, NY (Drs Sanacora andMason); grant KO8MH01715-01 from the National Institute of Mental Health,Rockville, Md (Dr Sanacora); the Dana Foundation, New York, NY (Drs Sanacoraand Epperson); grant P50AA12870 from the National Institute on Alcohol Abuseand Alcoholism, Bethesda (Dr Krystal), grant K23 MH01830 from the NationalInstitute of Mental Health, Rockville (Dr Epperson); grant MH30929-21 fromthe Mental Health Clinical Research Center at Yale (Drs Krystal and Mason);and The Stanley Foundation, Muscatine, Iowa (Dr Mason).

This study was presented in part at the Annual Meeting of the Societyfor Biological Psychiatry; May 17, 2003; San Francisco, Calif.