Neural and Behavioral Responses to Tryptophan Depletion in UnmedicatedPatients With Remitted Major Depressive Disorder and Controls

Context  An instructive paradigm for investigating the relationship between brain serotonin function and major depressive disorder (MDD) is the response to tryptophan depletion (TD) induced by oral loading with all essential amino acids except the serotonin precursor tryptophan.

Objective  To determine whether serotonin dysfunction represents a trait abnormality in MDD in the context of specific neural circuitry abnormalities involved in the pathogenesis of MDD.

Design  Randomized double-blind crossover study.

Setting  Outpatient clinic.

Participants  Twenty-seven medication-free patients with remitted MDD (18 women and 9 men; mean ± SD age, 39.8 ± 12.7 years) and 19 controls (10 women and 9 men; mean ± SD age, 34.4 ± 11.5 years).

Interventions  We induced TD by administering capsules containing an amino acid mixture without tryptophan. Sham depletion used identical capsules containing hydrous lactose. Fluorodeoxyglucose F 18 positron emission tomography studies were performed 6 hours after TD. Magnetic resonance images were obtained for all participants.

Main Outcome Measures  Quantitative positron emission tomography of regional cerebral glucose utilization to study the neural effects of sham depletion and TD. Behavioral assessments used a modified (24-item) version of the Hamilton Depression Rating Scale.

Results  Tryptophan depletion induced a transient return of depressive symptoms in patients with remitted MDD but not in controls (P<.001). Compared with sham depletion, TD was associated with an increase in regional cerebral glucose utilization in the orbitofrontal cortex, medial thalamus, anterior and posterior cingulate cortices, and ventral striatum in patients with remitted MDD but not in controls.

Conclusion  The pattern of TD-induced regional cerebral glucose utilization changes in patients with remitted MDD suggests that TD unmasks a disease-specific, serotonin system–related trait dysfunction and identifies a circuit that probably plays a key role in the pathogenesis of MDD.

Major depressive disorder (MDD) has been associated with abnormallyreduced function of central serotonergic systems by a variety of in vivo andpostmortem findings. Relative to controls, patients with MDD have lower plasmatryptophan levels,^1,2 reducedcerebrospinal fluid 5-hydroxyindolacetic acid levels,^3,4 decreasedplatelet serotonin uptake,⁵ and blunted neuroendocrineresponses in challenge studies of different serotonin receptors.^6-12 Findingsfrom brain imaging studies^13-16 andautoradiographic studies in tissue sections obtained post mortem^17,18 suggestwidespread impairment of serotonergic function in depression with and withoutsuicidality.

Tryptophan depletion (TD) has been used to test the hypothesis thatdecreased serotonin function is associated with MDD (for a review see Booijet al¹⁹). Administration of a tryptophan-freeamino acid mixture of essential amino acids produces a rapid and substantialdecrease in plasma tryptophan levels and a decrease in brain tryptophan, brainserotonin, and 5-hydroxyindolacetic acid levels in rats.^20,21 Studiesin humans show profound decreases in plasma and cerebrospinal fluid tryptophanlevels²² and in cerebrospinal fluid levelsof 5-hydroxyindolacetic acid^23,24 afteroral administration of an amino acid mixture without tryptophan.

Findings that TD induces depressive symptoms in controls at high familialrisk of depression^25-28 andin medication-free, remitted patients with a history of depression^29,30 but not in euthymic control subjectswithout familial risk of depression^{24,25,29,31-36} suggestthat either dysfunction within serotonergic systems or heightened sensitivityto reductions in serotonin transmission reflects a trait abnormality in depression.

The neural circuits associated with the core symptoms of depressionand increased vulnerability for depression have not been precisely defined.Previous neuroimaging studies had design limitations, including small samplesizes, medication confounds, and illness heterogeneity. Such previous positronemission tomography (PET) studies³⁷ of TD-induceddepressive symptoms in medicated, remitted individuals with depression showedthat glucose metabolism decreased in the orbitofrontal cortex, ventrolateralprefrontal cortex (PFC), frontal polar cortex, pregenual cingulate cortex,and thalamus in patients who relapsed during TD but not in individuals whodid not relapse. In addition, at baseline (before TD), patients who relapsedhad increased metabolism in the amygdala relative to those who did not relapse.These data were largely consistent with subsequent studies^38,39 usingPET with water labeled with oxygen 15 in patients with remitted MDD (rMDD)(most of whom were taking medication) that found that decreased tryptophanlevels and transient return of depressive symptoms were associated with diminishedcerebral blood flow in the ventral anterior cingulate cortex, orbitofrontalcortex, and caudate nucleus. These areas have also been implicated in depressivepatients during an acute episode of depression⁴⁰ andthus may represent an important circuitry in MDD.

A major limitation of these investigations pertaining to identificationof the primary serotonin-related neural circuits related to high risk of depressionwas that patients were studied while taking antidepressant medications. Inthis context, the purpose of the present study is to examine the behavioraland neural effects of TD in unmedicated patients with rMDD and controls. Thepresent study extends previous research performed in MDD by revealing a traitabnormality of serotonin function in MDD in the context of specific neuralcircuitry abnormalities that may be involved in the pathogenesis of MDD.

Participants were entered into the study after they received a fullexplanation of the purpose of the study and the study procedures and afterthey provided written consent as approved by the National Institute of MentalHealth institutional review board. Twenty-seven remitted, unmedicated patientswith a diagnosis of at least 2 past major depressive episodes on the basisof the Structured Clinical Interview for DSM-IV–Nonpatient Version⁴¹ and19 controls were included in the study (Table 1). Duration of depressive illness and number of episodeswere estimated from the Past History of MDD addendum to the Structured Clinical Interview for DSM-IV–NonpatientVersion. Remission was defined as at least 3 months during which thepatient did not take an antidepressant agent and had 24-item Hamilton DepressionRating Scale (HDRS) scores in the nondepressed range (<8).⁴² Informationabout family history of mental illness (Axis I diagnoses) was obtained fromthe study participants for all first-degree relatives using the Family Interviewfor Genetic Studies.⁴³ Controls had no personalor family (first-degree relatives) history of psychiatric disorders. No useof psychotropic medications was allowed during the study. Participants werefree of medical illness on the basis of history and results of physical examination,electrocardiography, and laboratory tests, including liver and kidney functiontests, hematologic profile, thyroid function tests, urinanalysis, and toxicology.Pregnant and nursing women were excluded. Premenopausal women were studiedduring the follicular phase of the menstrual cycle. The menstrual phase wasdetermined using plasma estradiol and progesterone concentrations, time sinceonset of last menses, and home urine ovulation kits to detect the mid-cycleluteinizing hormone surge (Clear Plan Easy; Whitehall Laboratories, Madison,NJ) to identify the time of ovulation within the index menstrual cycle.

Participants were enrolled into a double-blind, placebo-controlled crossoverTD study and were randomly assigned to undergo either TD first and sham depletion(SD) second or SD first and TD second. To avoid carryover effects, depletionprocedures were separated by at least 8 days. On the TD day, participantsingested 70 white capsules containing an amino acid mixture consisting ofisoleucine (4.2 g), leucine (6.6 g), lysine (4.8 g), methionine (1.5 g), phenylalanine(6.6 g), threonine (3.0 g), and valine (4.8 g) at 7 AM. DuringSD, participants received 70 white capsules containing a total of 31.5 g oflactose at 7 AM. Patients were restricted from eating on day1 of the study until completion of PET at about 4 PM, at whichtime they returned to unrestricted food intake. Study raters (M.G. and T.W.)were blind to whether the individual was a patient or a control. The effectsof TD and SD were evaluated using measures of depression and measures of plasmatotal and free tryptophan concentrations.

Baseline clinical ratings were obtained at 7 AM and then7 and 24 hours later using a modified (24-item) version of the HDRS. The itemsassessing insomnia, weight change, and diurnal variation were removed becausethey could not be meaningfully assessed on the days of the study.

Plasma total and free tryptophan levels were measured on each studyday at baseline and then 5, 7, and 24 hours after intake of the capsules.Collected blood samples were immediately centrifuged for 15 minutes at 4°Cand 3000 rpm. Plasma was frozen at −70°C until analyzed. Immediatelyafter thawing, plasma proteins were precipitated by adding 20 µL of70% perchloric acid to 400 µL of plasma, followed by centrifugationfor 30 minutes at 20 000g at 4°C. For detectionof total tryptophan, 100 µL of the supernatant was injected into thehigh-performance liquid chromatography system, leaving another 100 µLfor a second injection. For the detection of free tryptophan, samples werefiltered through a 10-kDa centrifugal filter device (Amicon Ultrafree-MC;Millipore Corp, Bedford, Mass) before injection. The system was calibratedusing an external standard solution of tryptophan dissolved in a phosphate-bufferedsaline solution containing bovine serum albumin, 0.5 mg/mL. The standard solutionscontained tryptophan concentrations ranging from 0.31 to 20.0 µg/mLfor total tryptophan and 0.125 to 10.0 µg/mL for free tryptophan. (Toconvert tryptophan to micromoles per liter, multiply by 48.97.) For calibration,10 µL of 70% perchloric acid was added to 200 µL of standard solution,thawed immediately before use, and handled in the same way as the plasma samples.The high-performance liquid chromatography system consisted of the Waters2690 Separations Module (Waters Corp, Milford, Mass). The operational isocraticchromatographic conditions for this high-performance liquid chromatographysystem were set as follows: column temperature, 25.0°C; and flow rate,1.0 mL/min. The mobile phase consisted of 2.5 g of sodium acetate, 100 mgof disodium EDTA, and 50 mg of sodium octyl sulfonate, which were dissolvedin 2500 mL of deionized water and 150 mL of acetonitrile. A pH of 4.50 wasreached by the addition of acetic acid to the buffer before acetonitrile wasadded. This solution was filtered through a 0.47-µm membrane filterand degassed before use. The analytical column was a 250 × 4–mmSupersphere 60 RP-select B, packed with C8 (LiChroCART 250-4; MERCK KGaA,Darmstadt, Germany).

Approximate run time after injection until detection of tryptophan was10 minutes. A scanning fluorescence detector (Waters 474; Waters Corp) (λexcitation wavelength = 300 nm and λ emission wavelength = 350 nm)was used to detect tryptophan. The amount of a substance was obtained by theratio of the peak height to the peak height of the calibration curve of theexternal standards. The tryptophan recovery evaluated by the amount of spikedvs nonspiked plasma after extraction was 90% to 100%. Intra-assay and interassayvariations were less than 5%. The signal-to-noise ratio of the lowest standard(0.125 µg/mL) was greater than 30:1; the R² of the calibration curve was greater than 0.9992.

Magnetic resonance (MR) images were obtained for each participant usinga 3.0-T scanner (Signa; GE Medical Systems, Waukesha, Wis) and a 3-dimensionalmagnetization-prepared rapid acquisition gradient-echo sequence (echo time,2.982 milliseconds; repetition time, 7.5 milliseconds; inversion time, 725milliseconds; voxel size, 0.9 × 0.9 × 1.2 mm) to provide an anatomicframework for analysis, partial volume correction of the PET images, and morphologiccharacterization so that individuals with anatomic abnormalities could beexcluded.

Because the biochemical and behavioral effects of TD peak 5 to 7 hoursafter administration of the amino acid mixture, fluorodeoxyglucose F 18 (FDG)was infused approximately 6 hours after administration of the capsules containingeither lactose or amino acids. Regional cerebral glucose utilization (rCMRGlu)was measured noninvasively by combining left ventricular chamber time-activitycurve data with venous blood sample values to give the input function neededto calculate the metabolic rate. The left ventricular input function was obtainedfrom dynamic PET imaging of the heart, with venous blood samples obtainedconcurrently with imaging after injection of 4.5 mCi (166.5 MBq) of FDG. Imagedata from the heart were acquired using a whole-body PET scanner (GE Advance;GE Medical Systems) in 2-dimensional mode for 35 minutes (ten 30-second framesand ten 3-minute frames). This was followed by a 10-minute emission and an8-minute transmission brain scan 45 minutes after tracer injection. Cardiacslices were reconstructed, and 5 left ventricular slices were identified forregion of interest (ROI) placement. The 0- to 5-minute frames were averagedto allow location of the left ventricular blood pool, whereas the 25- to 35-minuteframes allowed identification of myocardial FDG uptake. Circular, 2-cm-diameterROIs over the left ventricular chamber were positioned on each of the differenceimages (left ventricular image blood pool minus myocardial FDG uptake) suchthat spillover from the myocardium was minimized. An average left ventriculartime-activity curve was obtained from the time-activity curves derived fromthe ROI in each of the 5 slices. The time-activity curve was extended in timeto include the period of brain imaging by using venous blood sample values.The average values of the venous blood samples taken at approximately 25,30, 35, and 50 minutes and the average left ventricular concentration duringthe 25- to 35-minute period were divided. This ratio was then used to scalethe 50-minute venous sample concentration, which was then appended to theleft ventricular curve, completing the input function used to generate parametricimages of rCMRGlu.⁴⁴

The effects of TD on rCMRGlu were assessed using whole-brain and MRimage–based ROI analysis using image processing and analysis software(MEDx; Medical Numerics Inc, Sterling, Va). Whole-brain FDG uptake was measuredusing an MR image–based template. Primary ROIs were selected based onresults of previous monoamine depletion studies^37,45 inmedicated patients with rMDD, which revealed abnormalities in the orbitofrontalcortex, posterior cingulate cortex, medial thalamus, and dorsolateral PFC.Additional secondary ROIs were selected based on these regions showing alterationsin functional imaging studies in untreated, symptomatic depressed patients⁴⁰ and consisted of the ventral striatum, pregenualcingulate cortex, subgenual cingulate cortex, ventrolateral PFC, and amygdala.All ROIs were defined a priori on an MR imaging template. These regions wereplaced on each patient's registered MR image. A binary mask of the gray matterwas then used to ensure that only gray matter pixels were included in theanalysis. Regions were then transferred to the coregistered PET images, andthe mean metabolic activity was obtained for each ROI. The whole-brain measurewas used to normalize the regional measures to factor out nonspecific globaleffects.

To explore TD-induced changes in regions outside the primary ROIs, theimages were analyzed post hoc by voxel-by-voxel analysis using a statisticalparametric mapping software package (SPM99; Wellcome Department of ImagingNeuroscience, London, England). The FDG images were coregistered to the MRimages and spatially normalized to the standardized space. Each image wasthen smoothed using a 12-mm gaussian kernel to compensate for errors in coregistrationand normalization.

Treatment (TD vs SD) × group (rMDD vs controls) repeated-measuresanalyses of variance were performed for each of the ROIs for whole-brain–normalizeddata. Greenhouse-Geisser P values were used to dealwith concerns about the sphericity of the repeated-measures factors. Bonferroni-adjustedsimple effects tests were used to evaluate the locations of differences. Fourregions were considered a priori primary ROIs, and 5 regions were considereda priori secondary ROIs, so the α level was Bonferroni adjusted forthe number of comparisons within each of these groups. All P values are 2-tailed and are reported before correction for multiplecomparisons.

The same analysis of variance model and simple-effects tests were used,with the addition of a time factor for the total HDRS score and the HDRS sadnessitem score and for the plasma total and free tryptophan levels. The Fisherexact test was used to compare the relapse rates of patients and controlsafter TD. Return of symptoms was defined as a 10-point increase on the HDRSafter baseline. Exploratory analysis of sex effects on rCMRGlu in the primaryand secondary ROIs included adding sex as a factor in the analysis of variancemodel where treatment and group were factors. Follow-up tests were conductedin the same manner as those for the initial ROI analysis.

To study differences in rCMRGlu, the SPM analysis to visualize regionalchanges on a voxelwise basis between patients with rMDD and controls acrossconditions used a fixed-effects multigroup design incorporating HDRS scoresas a covariate. The threshold for statistical significance was set at P<.05 corrected. Clusters with corrected values of P<.05 are reported. Voxels with uncorrected significanceof P<.001 are reported if the voxels fell withinthe a priori, hypothesized primary and secondary ROIs.

An SPM analysis was also carried out to examine the relationship betweenmood state and rCMRGlu. Difference images obtained by subtracting normalizedrCMRGlu images for the TD sessions from those for the SD sessions were compared,using the change in HDRS score as a regressor.

Consistent with the literature, TD reduced plasma free and total tryptophanlevels, whereas levels remained unaffected during SD (Figure 1). Plasma free tryptophan levels were reduced by 78% inpatients with rMDD and by 76% in controls. Plasma total tryptophan levelswere reduced by 74% in patients with rMDD and by 71% in controls. Order oftest sessions, age, and sex did not affect the outcome.

Tryptophan depletion, but not SD, was associated with a significantlygreater increase in depressive symptoms, as reflected by an increase in HDRStotal scores in patients with rMDD relative to controls (treatment ×time × group interaction: F_1.2,51.0 = 23.54; P<.001) (Figure 2). Peakeffects of TD on mood were found approximately 7 hours after administrationof the amino acid mixture. The analysis for the sadness item (item 1) on theHDRS showed a robust increase during TD in patients with rMDD but not in controls,with no effects on mood in either group during SD (3-way interaction: F_1.2,53.5 = 9.51; P = .002). Sixteen (59%) of27 patients with rMDD experienced a transient return of depressive symptomsduring TD, whereas none met similar criteria during SD (Fisher exact test, P<.001). No control subject had depressive symptomsduring TD or SD. Each patient with rMDD who experienced a transient returnof depressive symptoms during TD reported feeling back to baseline on assessmentat the follow-up (day 2) interview. Again, order of test sessions, age, andsex did not affect the outcome.

No statistically significant between-group differences in whole-brainabsolute rCMRGlu were found on the 2 test days. Thus, only normalized dataare reported. Statistically significant between-group differences in responseto TD were found in the orbitofrontal cortex and the posterior cingulate cortex(Figure 3). We found a statisticallysignificant group × treatment interaction for the lateral orbitofrontalcortex (F_1,40 = 7.36; P = .01), with anincrease in rCMRGlu during TD relative to SD in patients with rMDD (P = .004) but not in controls (P =.32). A significant group × treatment interaction also was found inthe posterior cingulate cortex (F_1,40 = 9.69; P = .003). Relative to SD, TD induced an increase in rCMRGlu in patientswith rMDD (P = .03), whereas rCMRGlu decreased duringTD relative to SD in controls (P = .04). In contrast,no significant group × treatment interactions were found in the medialthalamus and the dorsolateral PFC (Table2). No statistically significant differences in rCMRGlu were foundbetween patients with rMDD who had a return of depressive symptoms and thosewho did not experience a return of depressive symptoms during TD. Also, nodifferences in rCMRGlu were found at baseline (during SD) between patientswith rMDD and controls.

Analyses of the secondary ROIs, implicated by functional imaging studiesusing PET in symptomatic depressed patients and postmortem studies, showeda significant group × treatment interaction in the ventral striatum(F_1,42 = 4.73; P = .04). After Bonferronicorrection, this interaction was not significant (P =.20). Patients with rMDD had a statistically significant increase in rCMRGluduring TD relative to SD (P = .009), whereas rCMRGludid not differ in controls between conditions. No significant group ×treatment interactions were found for the pregenual cingulate cortex, subgenualcingulate cortex, ventrolateral PFC, or amygdala (Table 2).

Except for the posterior cingulate cortex, we did not see an effectof age or sex on rCMRGlu in any of the primary or secondary ROIs. Men withrMDD largely accounted for differences in the posterior cingulate cortex.

The results of the voxel-based analysis of rCMRGlu comparing patientswith rMDD and controls across treatments (TD vs SD) extended findings formthe ROI analysis and showed significant between-group differences for themedial thalamus (z = 4.18), anterior cingulate cortex(Brodmann area 32; z = 3.97), and right putamen (z = 3.81; P<.001 for all).The analysis of rCMRGlu for regions that showed a significant group ×treatment interaction comparing patients with rMDD who showed a return ofdepressive symptoms relative to controls revealed significant (cluster-levelcorrected) between-group differences in the posterior cingulate cortex (Brodmannarea 31; z = 2.80; P = .003),left superior temporal cortex (z = 2.75; P = .003), anterior cingulate cortex (Brodmann area 32; z = 3.03; P<.001), and anterolateral PFC(z = 2.99; P = .003). TherCMRGlu of patients with rMDD who showed a return of depressive symptoms differedsignificantly from that of patients with rMDD without symptom recurrence inthe posterior cingulate cortex (Brodmann area 31; z =2.60; P = .005). We used a regression analysis toassess the effect of mood change on the TD and SD days in the rMDD group.Increasing depression scores were significantly (cluster-level corrected)correlated with increased rCMRGlu in the right precuneus (z = 3.10) and the right putamen (z = 3.04; P<.001 for both).

Medication-free, euthymic patients with a history of MDD differ statisticallysignificantly from controls in their response to TD. Most individuals witha history of MDD experienced a transient return of depressive symptoms, whereascontrols remained unaffected by TD. Neither group showed a depressive reactionduring SD. These findings extend previous TD studies in patients with rMDDtaking^37,46-49 andnot taking^29,30 antidepressantmedications by demonstrating the specificity of the depressiogenic effectsof TD for patients with MDD and suggests that MDD is associated with a traitlikeabnormality involving central serotonergic systems.

Our data further indicate that TD unmasks functional changes in theneural circuits implicated in the pathogenesis of MDD. In patients with rMDDrelative to controls, TD was associated with increased rCMRGlu in the orbitofrontalcortex, anterior and posterior cingulate cortices, medial thalamus, and ventralstriatum. Previous studies in unmedicated patients with MDD scanned duringa spontaneous episode of MDD have consistently reported elevations of cerebralblood flow and glucose metabolism in patients with MDD relative to controlsin these same regions (for a review see Drevets⁴⁰).Previous imaging studies comparing patients with MDD during a spontaneousepisode of MDD with controls were interpreted such that the elevated activityin these regions was mood dependent, partly because it was found also duringexperimentally induced sadness and anxiety in healthy individuals⁵⁰ and in symptomatic depressed patients in responseto negative stimuli.⁵¹ Antidepressant drugtreatments result in a decrease in cerebral blood flow and metabolism towardnormative levels in these regions.^51-54 Thekey finding that distinguishes the present study from previous similar workis that we did not find statistically significant differences in rCMRGlu betweenpatients who had a return of depressive symptoms and those who did not duringTD. This leads to the question of whether results from other functional neuroimagingstudies really reflect current mood states or truly reflect an underlying"trait" abnormality in brain metabolism or blood flow associated with MDD.Our data suggest that TD unmasks a circuitry that truly represents a traitabnormality in MDD. Several researchers^55-57 haveshown that the anterior and posterior cingulate cortices and the anteromedialPFC are involved in processing the affective salience of sensory stimuli andare involved in tasks that elicit emotional responses or require emotionalevaluations. It can be hypothesized that these areas activate acutely duringmood challenges, whereas dysfunctions in the limbic-cortical-striatal-pallidal-thalamiccircuits represent a traitlike abnormality in MDD.

The magnitudes of the regional metabolic changes observed between theSD and TD conditions in the rMDD sample ranged from approximately 2% to 3%.Similarly, the normalized regional metabolic values measured in patients withMDD during TD-induced depressive relapse were only 1% to 3% greater than thecorresponding values in controls under either TD or SD conditions (Table 2). The magnitudes of these metabolicdifferences were thus modestly smaller than those of the metabolic differencesfound between currently depressed MDD cases and controls in the same regions,which have typically ranged from 3% to 6% when measured using ROIs placedusing PET-MR image collocation.⁵⁴ The relativelysmaller metabolic differences observed during TD-induced return of depressivesymptoms in the present study may simply reflect the corresponding differencesin depression severity between such studies. The severity of the depressivesyndrome achieved during TD-induced return of depressive symptoms in the presentstudy was mild, whereas previous studies comparing depressed MDD samples andcontrols imaged the depressed patients when they were in the moderately toseverely depressed range. The relatively subtle changes in rCMRGlu we observedduring TD-induced return of symptoms is nevertheless in the range expectedfor physiologic activation of tissue when taking into account the dilutionaleffects of our relatively large ROIs. During physiologic activation of cerebraltissue, the focal increases in metabolism are small (10%-40%) when measureddirectly over the area of maximal change.⁵⁸ Inthe present study, because we were unsure where the focal area of increasewould occur within relatively broad areas of interest in individuals, metabolismwas measured using low-resolution image analysis approaches to reduce typeII error. In the MR image–based ROI analysis, the ROI size was substantiallylarger than the amount of cortex typically involved during physiologic activation,so the signal that we were trying to detect was substantially diluted. Similarly,in the SPM analysis, the images were blurred to a lower resolution to minimizethe effects of misalignment errors that occur during the process of spatiallynormalizing the variable brain anatomy across individuals. As a result, themagnitudes of the metabolic changes detected by the SPM analysis were alsodiluted.

Tryptophan depletion did not induce significant metabolic differencesbetween patients with rMDD and controls in the amygdala. Increased left amygdalametabolism has been reported during spontaneous episodes of MDD, and a positivecorrelation between increased metabolism and the severity of the depressivesyndrome has been reported previously,⁵⁹ whereaslower right amygdala activity seems to predict a favorable treatment outcome.⁵² Evidence suggests that normal serotonin functionis important for proper amygdala function. Serotonin inhibits glutamate-evokedneuronal activity in the amygdala and modulates transmission of emotionallysalient sensory information from the sensory cortices to the amygdala.⁶⁰ Reduced serotonin activity in MDD may disinhibitexcitatory activity by reducing the stimulation of serotonin 1A receptorslocated on pyramidal cells, where they inhibit action potential formation,and of serotonin 3 receptors located on γ-aminobutyric acidergic interneurons,where they stimulate γ-aminobutyric acid release.^60-62 Thefailure of TD to induce significant changes in amygdala rCMRGlu may conceivablybe related to the short-lived effects of TD, resulting in an only transientdisruption of serotonin metabolism. Dysfunction in serotonin function sustainedover a longer period and more pronounced depressive symptoms might be necessaryto evoke changes in amygdala metabolism, as reported in the literature duringspontaneous depressive episodes.

Our results differ somewhat from those of previous studies in medicatedpatients with rMDD who underwent mood challenges during scanning. Studiesof cerebral blood flow³⁸ and rCMRGlu³⁷ during TD or during mood provocation with autobiographicalmemory scripts⁶³ reported diminished cerebralblood flow in the ventral anterior cingulate cortex, orbitofrontal cortex,and caudate nucleus^38,63 and decreasedmetabolism in the dorsolateral PFC, orbitofrontal cortex, and thalamus.³⁷ In the latter study,³⁷ relapse-pronepatients with rMDD had higher amygdala and PFC metabolism on the SD day thanthose who did not show depressive symptoms during TD. The present study inunmedicated patients with rMDD demonstrates, in contrast, areas with increasedmetabolism that agree with the literature in unmedicated symptomatic patientswith MDD. These discrepancies in the direction of change and the circuitryinvolved may be explained by the interaction of the medications patients hadbeen taking at the time of their studies with the TD and mood challenges vsthe drug-free status in our patients. Another difference is the smaller doseof amino acids used in the present study to deplete tryptophan and the useof lactose during SD. We found reductions in plasma total and free tryptophanlevels similar to the decrements reported by previous investigators.^22-27,29-39

Our results add to functional imaging studies in untreated patientswith MDD, lesion analyses, and postmortem studies that suggest that a well-characterizedcircuit seems to play a key role in the pathogenesis of MDD and probably representsa trait marker for MDD. Tryptophan depletion seems to be capable of unmaskingthis trait abnormality in MDD. As a potential phenotypic trait marker, TDseems to be a useful tool to study the genetic basis of MDD. An obvious nextstep in this type of investigation is to determine whether serotonin-relatedgenes account for the robust differential neural and behavioral responsesto TD in patients with rMDD and controls.

Correspondence: Alexander Neumeister, MD, Section on ExperimentalTherapeutics and Pathophysiology, Mood and Anxiety Disorders Program, NationalInstitute of Mental Health, North Drive, Bldg 15K, Room 200, Bethesda, MD20892-2670 (neumeisa@intra.nimh.nih.gov).

Submitted for publication January 6, 2004; final revision received February24, 2004; accepted March 15, 2004.

This study was presented in part in abstract form at the 42nd AnnualMeeting of the American College of Neuropsychopharmacology; December 9, 2003;San Juan, Puerto Rico.