Measurement of Brain Regional α-[11C]Methyl-L-Tryptophan Trapping as a Measure of Serotonin Synthesis in Medication-FreePatients With Major Depression FREE

The serotonin hypothesis of major depression (MD) postulates that anabsolute or relative deficit in serotonergic neurotransmission in the brainpathways regulating mood may serve as a vulnerability diathesis for MD.^1- 4 Withthe development of selective radioactive probes for various components ofthe serotonin system, functional neuroimaging studies of specific neurochemicalprocesses across disease stages are now deemed possible.

Reduced serotonin neurotransmission could result from any of the followingmechanisms, alone or in combination: reduced availability of the precursor L-tryptophan (Trp) for serotonin synthesis; reduced metabolism, release,or both; altered neuronal reuptake; impaired receptor(s) function(s); andtransduction(s). Evidence supporting a reduction of serotonin₂ andserotonin_1A receptor binding potential^5- 8 andserotonin transporter^9- 10 availabilityin the brain of medication-free patients with MD has been recently reported;whether such alterations are primary or secondary, compensating for dysfunctionsin other neural systems, is unknown.

The synthesis of serotonin from the essential amino acid Trp is a 2-stepenzymatic process catalyzed by tryptophan hydroxylase,¹¹ whichis considered to be the rate-limiting enzyme for serotonin synthesis; in thebrain, it is found exclusively in serotonin neurons.¹² Theproduct of the Trp hydroxylation reaction, 5-hydroxy-L-Trp,undergoes decarboxylation catalyzed by aromatic amino acid decarboxylase.Serotonin is metabolized into 5-hydroxyindolacetic acid, which is removedfrom the brain into the cerebrospinal fluid and eventually into the bloodby a probenecid-sensitive system.¹³

There are at least 3 lines of evidence suggesting that impairment ofserotonin neurotransmission in patients with MD could result from alteredTrp uptake or metabolism, leading to reduced conversion of plasma Trp intobrain serotonin: (1) depressive symptoms can be induced in vulnerable individualsby a procedure known to acutely reduce serotonin content in the brain, thatis, Trp depletion^14- 15; (2) inhibitionof tryptophan hydroxylase by p-chlorophenylalanine,¹⁶ in patients' remission of depressive symptoms, cantrigger a rapid relapse of the depressive state; and (3) likewise, vulnerablepatients remitted from an episode of MD were reported to experience a relapseof depression after acute Trp depletion.^17- 18

Positron emission tomographic (PET) studies measuring serotonin synthesisrates directly in patients with MD initially relied on the use of 5-hydroxy-3-[¹¹C]tryptophan¹⁹ as a tracer. However,besides its rapid conversion into 5-hydroxy-3-[¹¹C]tryptamine,this tracer has the main disadvantage of bypassing the rate-limiting enzymetryptophan hydroxylase for serotonin synthesis, hence being a substrate forall brain monoamine aromatic amino acid decarboxylase.^20- 21

A more recent method for estimating brain regional serotonin synthesisrelies on the use of α-[¹¹C]methyl-L-tryptophan(α-[¹¹C]MTrp),^22- 23 whichis a synthetic analog of the serotonin precursor Trp. Findings from autoradiographicstudies indicate that α-[³H]MTrp is taken up by serotoninneurons,²⁴ where it is trapped in the serotoninsynthesis precursor pool or metabolized into α-methylserotonin.^22,25 It has been proposed that the rateof this irreversible trapping of α-[¹¹C]MTrp (K*) providesan index of regional rates of serotonin synthesis²³ becauseit can be related to the metabolic conversion of Trp into serotonin.^22,26

In the present study, measurements of brain regional K* were obtainedand compared in medication-free patients with current MD (n = 17) and controls(n = 17) matched for age and sex. Brain regional K* was measured bilaterallyin 5 areas (10 volumes of interest [VOIs]) selected a priori based on thecurrent literature on the functional brain correlates of mood, cognitive,and autonomic dysregulations that characterize MD: the dorsolateral prefrontalcortex (DLPFC), cingulate gyrus, mesial temporal lobe, caudate nucleus, andthalamus.^27- 30 ThePET studies of cerebral blood flow or glucose metabolism in patients withprimary depression identified a pattern of corticolimbic regional dysfunctionsbelieved to underlie the pathogenesis or expression of MD: reduced frontallobe function (dorsolateral and ventrolateral PFC) is perhaps the most consistentlyreported abnormality, together with changes in cingulate cortex (CC), limbicand paralimbic areas (anterior temporal, amygdala, and insula), and subcorticalnuclei (basal ganglia and thalamus).³⁰ Particularlyrelevant to the a priori selection of the brain areas to be examined prospectivelyin this study is the recent observation that serotonin_1A and serotonin₂ in vivo binding potentials and the metabolic response to serotoninagonists are reportedly reduced in patients with MD in some of the same brainregions (DLPFC, medial prefrontal cortex, and anterior CC [ACC]).^{7- 8,31- 32}

Primary inclusion criteria were as follows: (1) current MD, as per the Structured Clinical Interview for the DSM-IV Axis I Disorders³³; (2) Hamilton Depression Rating Scale17³⁴ score of 18 or higher; (3) medicationfree for 2 weeks or for more than 5 elimination half-lives of the drug, whicheverwas more; (4) no current substance abuse; and (5) never having used the putativeserotonin neurotoxins 3,4-methylenedioxymethamphetamine and 3,4-methylenedioxyamphetamine.Twenty-nine consecutive patients with a provisional diagnosis of MD referredby Montreal-area psychiatrists and general practitioners were assessed forstudy participation; 17 met the entry criteria. Reasons for exclusion includedhypothyroidism (n = 2), hyperthyroidism (n = 1), a history of severe braintrauma (n = 1), severe diabetes mellitus (n = 1), current substance abuse(n = 1), dropout (n = 1), not meeting the diagnostic criteria for currentMD (n = 2), and technical breakdown (n = 3).

All controls, recruited via newspaper advertisements, were interviewedusing the Structured Clinical Interview for the DSM-IV AxisI Disorders.³³ All subjects were physicallyhealthy, as determined by a physical examination and standard laboratory tests.Exclusion criteria included a personal history of past or current DSM-IV Axis I psychiatric disorder, a DSM-IV AxisI psychiatric disorder in a first-degree relative, a Beck Depression Inventory³⁵ score greater than 10, and past use of 3,4-methylenedioxymethamphetamineor 3,4-methylenedioxyamphetamine. Seventeen controls were selected and matchedto the patient group for age and sex.

On the day of the PET study, all participants had negative findingson a urine drug screen sensitive to cocaine, opiates, phenylcyclohexyl piperidine,tetrahydrocannabinol, barbiturates, benzodiazepines, and amphetamines (TriagePanel for Drugs of Abuse, Biosite Diagnostics Inc, San Diego, Calif). Allwomen of fertile age were scanned during their follicular phase. Written informedconsent was obtained from all participants. The study was carried out in accordancewith the Declaration of Helsinki,³⁶ and itwas approved by the research ethics committee of the Montreal NeurologicalInstitute and by the institutional review board of McGill University.

The α-[¹¹C]MTrp was prepared as described in a previousarticle.³⁷ All patients observed an overnightfast (water allowed ad libitum), preceded by a low-protein diet, to reduceinterindividual variability in plasma amino acid concentrations.³⁸ AllPET studies were performed between 11 AM and 2 PM usinga whole-body scanner (ECAT HR+; CTI Molecular Imaging, Inc/Siemens, Knoxville,Tenn). All images were collected and reconstructed using a 3-dimensional modewith an intrinsic resolution of 5 × 5 × 5 mm full width at halfmaximum. Transmission scans for attenuation correction were performed usinga ⁶⁸Ge/Ga source. After the intravenous injection of 10 to 20 mCi(185-740 MBq) of α-[¹¹C]MTrp, dose not scaled to body weight,administered as a 2-minute slow infusion, 60-minute dynamic PET data wereacquired. During each PET scan, 13 blood samples were drawn from the antecubitalvein to compute the α-[¹¹C]MTrp input function. The inputfunction was derived from sinus radioactivity (0-20 minutes) and venous plasma(20-60 minutes) as described in previous publications.^39- 40 Three2-mL venous blood samples were collected from each participant during thePET study. All samples were centrifuged, and ultrafiltrates were stored at−80°C for measurement of free plasma Trp concentrations using high-performanceliquid chromatography.^25,38 Twoadditional plasma samples were treated with trichloroacetic acid (2:1) fordetermination by high-performance liquid chromatography of total plasma Trpconcentrations.

All participants underwent high-resolution magnetic resonance imagingusing a 1.5-T superconducting magnet system (Philips Gyroscan; Philips MedicalSystems, Eindhoven, the Netherlands). Images were collected using 3-dimensionalvolume acquisition, T1 weighted (3-dimensional fast-field echo scan: repetitiontime, 18 milliseconds; echo time, 10 milliseconds; and flip angle, 30°),over the whole brain. Magnetic resonance imaging data were stored as a 256× 256 × 160–mm matrix with 1-mm³ isotropic voxels.

The Patlak graphic method⁴¹ was usedto calculate K* (in milliliters per gram per minute) using 40 minutes of dynamicPET data collected 20 to 60 minutes after tracer injection.^38,40 Comparisonsof regional normalized K* values between controls and patients with MD werecarried out using a magnetic resonance imaging–based VOI method. Individualmagnetic resonance imaging data were corrected for field inhomogeneities⁴² and were resampled in a standard stereotaxic space.⁴³ Tissue classification into gray and white matterand cerebrospinal fluid was performed using INSECT (Intensity Normalized StereotaxicEnvironment for the Classification of Tissue).⁴⁴ Thesedata were subsequently submitted to an ANIMAL (Automatic Nonlinear ImagingMatching and Anatomical Labelling)⁴⁵ for segmentationinto 48 anatomical volumes (5 structures for a priori–selected analysisand 19 for exploratory analysis, all bilaterally). All volumes were visuallyinspected, and errors were corrected manually when necessary.

Bilateral DLPFC, cingulate complex, mesial temporal cortex, thalamus,and caudate constituted the a priori–selected VOIs. An additional 38exploratory VOIs were placed throughout the brain. These VOIs were convolvedwith a 7-mm full width at half-maximum gaussian kernel filter and then wereresampled into PET native space. Volumes of interest were then applied todynamic native PET space to extract time-activity curves. To minimize theeffect of individual global differences in regional values, all regional K*values were normalized by the mean global K* to 100. Volumetric measurementswere performed in brain volumes in Talairach space using the software DISPLAY,⁴⁶ which allows for simultaneous labeling of a targetstructure in multiple planes.

All statistical procedures were performed using STATISTICA (version4.1; Statsoft Inc, Tulsa, Okla), with the main comparisons based on the useof repeated-measures analyses of variance, with hemisphere (left vs right)as a within-subject factor (repeated-measure factor) and sex and group (controlsvs patients with MD) as between-subject factors. When the group × sex× hemisphere interaction was significant (P<.05),post hoc analysis using the Student-Newman-Keuls test was performed. All comparisonswere carried out only between patients and controls of the same sex. Significancewas assessed at P<.05. Correlations between regionalnormalized K* values in VOIs and demographic and clinical variables, includingage, duration of illness, number of relatives with affective illness, freeplasma Trp concentration, and Hamilton Depression Rating Scale and Beck DepressionInventory scores, were examined using the Pearson product moment correlation.In the exploratory analysis, all probabilities were corrected for multiplecomparisons (n = 38), a procedure that, in this instance, might be deemedconservative given the high degree of regional correlation among variables.⁴⁷

A total of 17 patients (9 women: mean ± SD age, 41 ± 11years; and 8 men: mean ± SD age, 41 ± 11 years) and 17 controls(9 women: mean ± SD age, 37 ± 15 years; and 8 men: mean ±SD age, 32.5 ± 9.9 years) participated in the study (Table 1). Three patients were diagnosed as having single-episodeMD, 12 (5 women and 7 men) had major depressive disorder, and 2 met the criteriafor bipolar disorder type II. Mean ± SD Hamilton Depression RatingScale scores were 27.3 ± 6.6 for women with MD and 25.8 ± 5.2for men with MD. There was no statistically significant change in the normalizedglobal K* with age.

All but 4 patients (2 men and 2 women) had a documented positive familyhistory of mood disorder in first-degree relatives, and 2 patients were adopted.Only 1 patient reported a history of suicidal attempt. All patients with MD,except 1, and all controls were right handed. There were no correlations betweenthe severity of the index episode and duration of the medication-free period,number of previous episodes, or number of affected relatives.

There were no group differences between unmedicated patients with MDand age- and sex-matched controls in plasma concentrations of total or freeTrp or in global K* values. The repeated-measures analyses of variance forthe regional normalized K* values, focusing on a priori–defined VOIs,found for the CC main effects of group (F_1,30 = 4.75; P = .04) and sex (F_1,30 = 7.73; P<.01)and a group × sex × hemisphere interaction (F_1,30 =7.52; P<.01). The post hoc Student-Newman-Keulsanalysis confirmed that the K* was reduced bilaterally in the CC in womenwith MD and on the left side in men with MD (P<.001for both) (Table 2). As illustratedby a group × hemisphere interaction (F_1,30 = 7.50; P<.02), normalized K* was also significantly reducedin the left mesial temporal cortex in patients with MD relative to controls(P<.001). The clinical severity of depressiondid not correlate with normalized K* values in any of the a priori–definedVOIs.

Exploratory analyses were performed in 38 additional VOIs. The comparisonsbetween patients with MD and controls suggested significantly decreased normalizedK* values in some brain structures, although most of those group differences(except in the superior parietal lobule bilaterally in women [uncorrected P<.001]) disappeared after correction for multiple comparisons(critical value of P = .001).

Volumetric analysis of the CC was performed to determine whether thereduction in normalized K* was secondary to volume reduction in the CC inpatients with MD, as suggested by Drevets et al.⁴⁸ Becauseautomatic segmentation methods only provide total volume of the cingulateformation, further segmentation was performed, distinguishing between posterior,anterodorsal, subgenual, and subcallosal cingulate gyri. Gray and white matterboundaries were determined using an automatic tissue classification procedure.⁴⁴ All measurements were performed by the same rater(N.G.) masked to age, sex, and diagnosis. Repeated-measures analysis of variancewas performed to examine group differences, taking into account sex and hemisphere.Patients with MD did not differ statistically from controls on morphometricmeasurements of the segmented cingulate formation.

A refined VOI analysis for normalized K* measurements performed foreach individual subdivision of the CC indicated that patients with MD, relativeto controls, exhibited statistically significantly lower normalized K* valuesin the anterodorsal cingulate, bilaterally in women and on the left in men.Normalized K* values in the ACC did not correlate with Beck Depression Inventoryor Hamilton Depression Rating Scale scores.

In this study, medication-free depressed patients exhibited a statisticallysignificant reduction of the regional normalized trapping constant (K*) of α-[¹¹C]MTrp, a proxy for serotonin synthesis, in the CC. The differencein the CC was statistically significantly more robust in women, and, aftersegmentation, the differences between patients with MD and controls in normalizedK* values are primarily in the ACC (Figure1).

The serotonin hypothesis of depression predicts that a reduced availabilityof serotonin in certain brain systems involved in the regulation of mood mediatesaspects of vulnerability to mood disorders. The present finding of reducednormalized K* values in parts of the paralimbic cortex but not in the DLPFC,orbital frontal cortex (OFC), or caudate provides some preliminary indicationabout the potential serotonin subsystems that may be relevant in the modulationof various facets of the depression phenotype. Germane to the finding of reducednormalized K* values in the ACC is the relatively consistent observation thatthis paralimbic structure is activated during induced negative mood statesin controls and deactivated in patients with MD or alexithymia.^49- 50 Whereasother interconnected limbic structures, such as the amygdala and the hippocampus,⁵¹ are known to be involved in emotional responses tosimple perceptual stimuli, the ACC is thought to play a larger role in thecognitive evaluation of more complex affectively relevant events and moodregulation.^52- 53 It receives abundantascending serotonin neuron projections from the median and dorsal raphe⁵⁴ and projects to the OFC and limbic caudate, forminga circuit hypothesized to be functionally deficient in patients with MD.^55- 56 Measurements of cerebral blood flowduring acute Trp depletion in patients recovering from an episode of MD showthat the severity of depressive symptoms, elicited by the reduced serotoninneurotransmission, correlates with the inhibition of neuronal activity inseveral cortical, subcortical, and limbic brain regions, including the subgenualACC.^54- 55 The possibility thata reduction in serotonin synthesis in serotonergic fibers innervating theACC accounts, in part, for the reduced positive emotional salience of externaland internal stimuli (anhedonia) often experienced by patients with MD, althoughspeculative, is tantalizing.

Patients with MD did not statistically significantly differ from controlsin normalized K* in the DLPFC and OFC. This is somewhat surprising given therecent study, although not confirmed by others,⁵⁷ ofa reduced glucose metabolic response to the serotonin-releasing agent fenfluraminehydrochloride in the PFC of medication-free patients with MD.⁵⁸ TheDLPFC is known to be part of circuit-mediating attentional processes and workingmemory. Depressed patients with psychomotor retardation and impaired executivefunction often demonstrate reduced activity in the DLPFC.⁵⁹ Becausewe did not collect measurements of executive function in these patients, wecannot comment on the significance of this apparent negative finding or ruleout the possibility of a false-negative finding. This notwithstanding, anassociation was recently reported in schizophrenic patients between reducedPFC function and enhanced dopaminergic striatal neurotransmission.⁶⁰ These findings are consistent with the hypothesisthat reduced executive functions and DLPFC activity in patients with MD could,in part, be preferentially mediated, directly or indirectly, by dopamine-related,rather than serotonergic, mechanisms.

Reduced metabolic activity and metabolic responsivity to serotonergicprobes in the OFC has recently been associated with impulsive aggression andsuicide.⁶¹ The absence of a statistically significantdifference in normalized K* values in the OFC in patients with MD vs controlsmay, in part, be accounted for by the relatively low incidence of suicidalbehavior in this sample of depressed patients (1 of 17 individuals). Consistentwith this hypothesis, reduced α-[¹¹C]MTrp trapping in themedial OFC was reported recently in patients with cluster B personality disordersendowed with high impulsivity and suicidal morbidity⁶² andin patients studied in the immediate aftermath of a severe suicide attempt.⁶³

Normalized K* values were also statistically significantly reduced inthe mesial temporal cortex of depressed patients, which included the amygdalaand the anterior part of the hippocampus, structures densely innervated byserotonin terminals. The amygdala is a pivotal structure involved in the appraisalof fear stimuli and in the acquisition and expression of anxiety-related responses⁶⁴; the hippocampus, in comparison, plays a role inmediating contextual memory functions. Increased anxiety levels and subjectivememory impairments are common expressions of the phenomenology of clinicaldepression. The amygdala and the hippocampus form important connections withthe CC and with the OFC and the DLPFC.^65- 66 Morphometricand metabolic alterations in both areas have been described in patients withMD.⁵³

The validity of the finding of reduced normalized K* values in patientswith MD rests on the following methodological considerations: (1) Is the α-[¹¹C]MTrp/PET method a valid technique for estimating serotonin synthesis?(2) Is the VOI approach valid? Why not use statistical parametric mapping?(3) Is the depression sample representative? (4) Could the reduction in normalizedK* values reflect a partial volume effect due to anatomic differences betweendepressed patients and controls? (5) Could the reduction in normalized K*values be an artifact of altered brain hemodynamics, that is, reduced cerebralblood flow, a common biological observation in patients with MD? These considerationsare discussed in the following paragraphs.

Some reservations have been expressed about the significance of the α-[¹¹C]MTrp/PET method on the basis of a variety of technical points, includingthe suggestion that the method might measure blood-brain barrier transportof tryptophan rather than synthesis of serotonin.⁶⁷ Thesereservations have been addressed in detail in several related method studiesand reviews from our group^22,38- 40 andothers.^23,68- 69 Medication-freepatients with MD did not differ statistically significantly from controlsin total and free plasma tryptophan concentrations or in free plasma tryptophanfraction, thereby, it is unlikely that reduced normalized K* values reflectgroup differences in circulating tryptophan concentrations. This is importantbecause changes in plasma tryptophan levels have been related to changes inbrain serotonin synthesis.^{15,18,27- 28,38,56} Inthis study, we elected to report all data as normalized K* values rather thanas calculated rates of serotonin synthesis R because the lumped constant requiredfor conversion of K* to R is not known for the human brain.

In this study, we chose a VOI-based analysis in PET native space forbetween-group comparisons. This choice was based on the following considerations:(1) the literature on cerebral blood flow and glucose metabolic disturbancesin MD permitted the a priori selection of regions of interest, (2) the analysiscould follow established neuroanatomy, (3) normalized K* values could be calculatedfor the purpose of brain-behavior correlations, and (4) the analysis relieson a nonbiased method of brain segmentation and minimizes the potential impacton brain functional measurements of group-related anatomic differences inshape and volume, which are not completely controlled for by the current spatialnormalization procedures used in VOI-based analysis.^70- 71 Thesetechniques are well validated and have been successfully used by other researchers.^65,72 Automatic labeling methods, however,have some limitations; in particular, manual corrections may occasionallybe necessary, specifically for small VOIs. Statistical parametric maps comparingnormalized K* values in depressed patients and controls were also obtainedfor the purpose of cross validation and confirmed the finding of a reducednormalized K* value in the ACC of patients with MD (t =3.38, z = 3.10; P<.001; x = 10, y = 44, z = 24)(Figure 2).

The sample size is modest but well within the range of similar studiesin the field. Much attention was focused on preventing contamination of thebiological measure of interest by nonspecific factors: although all patientswith MD were medication free, 13 of 17 were drug naive at the time of scan,all had negative findings on a comprehensive toxicologic screen the day ofscan, and all participants were studied within a 3- to 4-hour period (11 AM to 3 PM); matching for age and sex was rigorous,and women were all tested during their follicular phase. Yet, depression researchis often compounded by clinical and biological heterogeneity, and this studyis no exception. Until the reported observation is replicated in independentpatient samples, caution suggests that the conclusion drawn from this studybe applicable only to depressed patients whose clinical characteristics aresimilar to those of the reported sample: MD, outpatients with moderate tosevere symptoms, nonmelancholic type, with no history of previous suicidality,of whom 65% reported a positive family history of mood disorder in 1 or morefirst-degree relatives.

It has been proposed that reduced tissue volume in the subgenual regionof the ACC might confound the results of functional brain imaging studiesin patients with MD. The analyses on segmented cingulate did not identifystatistically significant differences in tissue volume between patients andmatched controls. The absence of statistically significant morphometric differencesbetween patients and controls suggests that the observed difference in thenormalized K* values in parts of the ACC is primarily attributable to functionalchanges, rather than partial volume effects, secondary to anatomic differencesbetween patients and controls.⁷⁰ We assessedthis further by analyzing normalized K* values after corrections for individualdifferences in tissue volume. After segmentation of the CC, normalized K*values remained lower in the ACC bilaterally in women with MD relative tocontrols but only on the left in men with MD. Whether this reflects a partialvolume effect in men or a true sex-related difference in normalized K* valuesis unclear.

It is unlikely that the observed group differences in regional K* valuesare attributable to changes in cerebral blood flow because tracers with alow plasma-brain rate constant, such as α-[¹¹C]MTrp, areinsensitive to variations in cerebral blood flow.⁷³

In conclusion, this study reports reduced normalized K* values in theACC and the left mesial temporal cortex in medication-free patients with MDin the absence of gross anatomic defects. Although these findings are highlysuggestive, they do not formally prove altered serotonin synthesis or alteredserotonergic neurotransmission in MD. More generally, this finding adds tothe cumulative evidence supporting serotonergic dysfunction in MD and pointsto a critical role played by serotonergic innervation to parts of the limbicand paralimbic cortex, such as the ACC, the amygdala, and the hippocampus,in the dysregulation of mood. Finally, these findings raise important questionsfor future studies: Are the reported alterations of K*, an index of serotoninsynthesis, trait or state related or genetically or environmentally determined?What, if any, is their developmental time course? Longitudinal measurementsin unaffected individuals at risk of major depressive disorder should representa promising avenue of research in the field.

Corresponding author: Mirko Diksic, PhD, Department of Neurologyand Neurosurgery, McGill University, 3801 University St, Montreal, Quebec,Canada H3A 1A1 (mirko.diksic@mcgill.ca).

Submitted for publication June 13, 2003; final revision received January18, 2004; accepted January 21, 2004.

This study was supported by grant MOP-42438 from the Canadian Institutesof Health Research, Ottawa, Ontario.

This study was presented in part at the American College of Neuropsychopharmacologymeeting; December 14-18, 1998, San Juan, Puerto Rico; and at the 46th annualmeeting of the Society of Nuclear Medicine; June 6-10, 1999; Los Angeles,Calif.

We thank the staff of the Medical Cyclotron-Radiochemistry and PositronEmission Tomography Units, Montreal Neurological Institute.

Dr Diksic is a Killam Scholar at the Montreal Neurological Institute,McGill University; Drs Debonnel and Benkelfat each hold a Chercheur NationalAward; and Dr Leyton holds a Chercheur Boursier from Le Fonds de la rechercheen santé du Québec.