Increased Activation of Anterior Paralimbic and Executive Cortex FromWaking to Rapid Eye Movement Sleep in Depression FREE

Depression has been consistently linked with sleep dysregulation,^1- 8 includingalterations in electroencephalographic (EEG) measures of rapid eye movement(REM) sleep. In addition, changes in REM sleep have been linked with the clinicalcourse of the disorder,^9- 12 andmost effective antidepressant medications suppress REM sleep. Therefore, clarificationof the neurobiology of REM sleep in depression may provide clinically relevantinsights into the pathophysiology of the disorder.

Rapid eye movement sleep is generated by brainstem mechanisms.¹³ Recent functional neuroimaging studies demonstratethat REM sleep preferentially activates anterior limbic and paralimbic structures,such as the amygdala and the anterior cingulate cortex, in the absence ofprefrontal activation.^14- 18Activation of limbic and paralimbic cortex in these studiesrefers to a relatively greater level of blood flow or metabolism during REMsleep in relation to a waking baseline condition. Given the lower functionalactivity of these structures in the preceding nonrapid eye movement (NREM)period,^17,19- 20 onecould conceptualize this as a reactivation of limbic and paralimbic cortexwithin REM sleep. The persistence of this waking vs REM sleep pattern acrossblood flow and metabolic functional neuroimaging studies led us to proposethat such a comparison may serve as a naturalistic probe of limbic and paralimbicfunction in patients with mental disorders, such as depression.^18,21 Giventhe involvement of these structures in emotional regulation and motivatedbehavior, we hypothesize that alterations in REM sleep in depressed patientsmay reflect a functional alteration in these structures that may be centralto the neurobiology of depression. Since depressed patients have increasedREM sleep in relation to healthy subjects, we hypothesized that depressedpatients would exhibit a greater activation of limbic and paralimbic cortexfrom waking to REM sleep relative to healthy subjects. To test this hypothesis,we compared waking to REM sleep changes in regional cerebral metabolism between28 depressed patients and 14 healthy control subjects using the [¹⁸F]fluoro-2-deoxy-D-glucose([¹⁸F]-FDG) positron emission tomography (PET) method.²²

We studied 24 depressed subjects (15 women and 9 men; mean ±SD age, 41 ± 10 years; right-handed only) and 14 healthy subjects (11women and 3 men; mean ± SD age, 37 ± 10 years; right-handedonly). Preliminary analyses of 6 of the depressed patients and 8 of the healthycontrols have been reported earlier.²¹ Theresearch study was reviewed and approved by the University of Pittsburgh InstitutionalReview Board. All subjects provided written informed consent after the procedureswere fully explained, and they were compensated for participation in the study.Depressed subjects met Research Diagnostic Criteria²³ formajor depression on the basis of an interview with either the Schedule forAffective Disorders and Schizophrenia or the StructuredClinical Interview for DSM-III-R.²⁴ Depressedsubjects were required to have a minimum score of 15 on the first 17 itemsof the Hamilton Rating Scale for Depression²⁵ ora score of 17 or greater on the Beck Depression Inventory.²⁶ Theywere excluded if they met Research Diagnostic Criteria for schizophrenia,lifetime history of substance abuse or alcoholism, borderline or antisocialpersonality disorder, organic affective disorder, schizoaffective disorder,or psychotic subtype of major depression or bipolar depression. Healthy subjectswere required to have a score of 6 or lower on the first 17 items of the HamiltonRating Scale for Depression and to be free of any lifetime history of a mentaldisorder as previously described.²¹ All subjectswere required to be free of medications that could affect mood or sleep forat least a 2-week period of time (8 weeks for fluoxetine) prior to EEG sleepand PET studies. Subjects who could not remain drug- or alcohol-free duringthe study, as verified by nightly drug screens, were excluded. Medical exclusioncriteria for all subjects were met, as previously described.²¹ Subjectivesleep quality was assessed using the Pittsburgh Sleep Quality Index. Psychologicaldistress was assessed using the Symptom Check List-90-Revised. Any subjectwith an Apnea-Hypopnea Index of 10 or higher on night 1 screening was excludedfrom further study. All subjects underwent magnetic resonance scanning priorto their EEG sleep and PET studies using a 1.5-T scanner (Signa; GE MedicalSystems, Milwaukee, Wis), as previously described.¹⁹ Todetermine whole-brain metabolism, a whole-brain mask was created by applyinga brain/nonbrain segmentation to the magnetic resonance data that minimizedthe dilution of whole-brain metabolic values by the individually variablecontribution of cerebrospinal fluid spaces.²⁷

Electroencephalographic sleep studies were performed at the Universityof Pittsburgh General Clinical Research Center. Electroencephalographic sleepwas monitored on nights 1, 2, and 3. Night 1 was an adaptation night and sleepdisorders screening night. Night 2 data were used for the collection of baselineEEG sleep data. Bedtime was determined by the mean bedtime during the 7 dayspreceding sleep studies, as determined by review of a 7-day sleep diary. Onnights 1 and 2, subjects had sham intravenous tubing taped over their forearms.This was inserted through a cannula portal to a monitoring room for the purposeof accommodation to an indwelling intravenous tube used on night 3 for injectionof the radioisotope. The EEG sleep montage consisted of a C4/A1-A2 EEG channel,2 electro-oculography channels (right and left eyes) referenced to linkedmastoids, and a submental electromyography channel. All electrode impedanceswere determined to be greater than 5000 Ω. The EEG signal was collectedusing Grass 7P511 amplifiers (Grass-Telefactor, West Warwick, RI). Filtersettings for the EEG were 0.3 to 100 Hz. The electromyograph was bipolar,with a filter setting of 10 to 90 Hz. Electroencephalographic sleep was scoredvisually by raters blind to clinical information, according to the Rechtschaffenand Kales criteria.²⁸ In addition to sleepcontinuity and sleep architecture measures, the primary REM sleep dependentvariables included REM sleep percentage, REM latency (time between sleep onsetand first REM period minus any wakefulness occurring during the interval),and REM density in the first REM period (average automated REM counts perminute of REM sleep in the first REM period²⁹).Interrater sleep scoring reliability for major sleep variables were checkedperiodically with κ values ranging from 0.76 to 0.85. Definitions forvisually scored sleep variables have been provided elsewhere.³⁰

Regional cerebral glucose metabolism was assessed during both wakingand REM sleep using the [¹⁸F]-FDG PET method.²² Thewaking PET study occurred on the morning following the second night of sleep.The REM sleep PET study occurred on the third night of study. All PET studiesused a 4- to 6-mCi (148- to 222-mBq) dose of [¹⁸F]-FDG injectedvia the cannulas portal method in order to minimally disturb subjects. Thetime of [¹⁸F]-FDG injection for the waking study was approximately2 to 4 hours following awakening from the second night of sleep. The timeof [¹⁸F]-FDG injection for the REM sleep study immediately followedthe identification of the first REM of the second REM period of the thirdnight of sleep in the laboratory. While depressed patients often exhibit changesin REM sleep in the first REM period, the first REM period is often too brieffor an [¹⁸F]-FDG PET imaging study. In both the waking and REMsleep periods, subjects were monitored via polysomnography while lying ona bed. They were left undisturbed for a 20-minute period following injectionof the radioisotope. Subjects were allowed to leave the laboratory environmentfollowing their morning waking PET scan. For the waking study, they were giveninstructions to remain awake but with eyes closed in a dimly lit room. Twentyminutes after injection, subjects were transported to the PET imaging room.Scanning included a 30-minute emission scan (6 summed sequential 5-minutePET emission scans beginning 60 minutes after injection of the [¹⁸F]-FDG),followed by a 15-minute rod-windowed transmission scan. A modified simplifiedkinetic method³¹ was used as an indirect measureof absolute glucose metabolism (MRDglc), as previously described.³² In this approach, the plasma integral was estimatedfrom 6 nonarteriolized venous plasma samples collected every 8 minutes from45 to 95 minutes after [¹⁸F]-FDG injection. The acquisition protocolincluded 3-dimensional (septa retracted) mode in an ECAT HR+ PET scanner (CTIMolecular Imaging Inc, Knoxville, Tenn). The head was positioned such thatthe lowest scanning plane was parallel to and 1 cm below the canthomeatalline. All PET images were reconstructed using standard commercial softwareas 63 transaxial slices (each 2.4 mm thick), as previously described.¹⁹ All subsequent alignments and coregistrations wereperformed using a modification of Roger Woods' automated algorithms for PETto PET alignment and PET to magnetic resonance cross modality registration,as previously described.^21,33- 36 Themethods for translating the PET images into a common Talairach space for usein the grouped Statistical Parametric Mapping program, 1999 version,^37- 38 analyses have been previously described.²¹

χ² Tests and t tests were usedto test group differences in categorical measures and continuous clinicaland demographic measures. Group differences in EEG sleep measures were determinedusing a multivariate analysis of variance (MANOVA) to minimize errors relatedto performing multiple comparisons. Variables entered into the MANOVA includedtotal recording period; sleep efficiency; percentages of stage 1, 2, delta,and REM sleep; REM latency; and the density of REM sleep in the first REMperiod. Post hoc analyses of variance were performed on individual variablesafter detecting a significant group effect in the MANOVA. For the measure,MRDglc, a repeated-measures analysis of variance (groups = control and depressed;repeated measure = wake and REM sleep MRDglu), was used to test for group× time interactions and group and time effects. The above analyses wereperformed using SPSS software (SPSS Inc, Chicago, Ill). To determine differencesin relative regional metabolism between waking and REM sleep for each group,as well as group (control vs depressed) by time (wake vs REM sleep) interactions,we used the Statistical Parametric Mapping program. This program was alsoused to test post hoc group differences in waking relative regional metabolismand in REM sleep relative regional metabolism. The control and depressed wakingand REM sleep PET images were entered into an analysis of covariance usingglobal metabolism and age as covariates. Age in years was used as a covariategiven known variations in regional cerebral metabolism with age.³⁹ Statisticimages (t scores converted to z scores) were created for each analysis. Local statistical maximain these images were identified by their Talairach atlas (x-, y-, and z-axis)coordinates (see Talairach and Tournoux⁴⁰).Regions of interest were defined from preliminary analyses reported earlier.²¹ Results were corrected for multiple comparisons basedon number of voxels in the whole brain and within regions of interest.

Depressed and healthy groups did not differ in age or sex (Table 1). Depressed subjects had mild tomoderate severity of depression, global distress, and subjective sleep disturbance.Twenty-one depressed subjects had recurrent major depression (average ageat onset, 28 ± 5 years [mean ± SD]; average duration, 49 ±52 weeks [mean ± SD]), while 3 had single-episode major depression(average age at onset, 28 ± 5.5 years [mean ± SD]; average duration,51 ± 29 weeks [mean ± SD]). According to the insomnia questionson the Hamilton Rating Scale for Depression,²⁵ sleepdisturbances were moderately severe and distributed evenly across the night(initial, middle, and delayed).

The MANOVA revealed that the EEG sleep (from the second undisturbedbaseline night of sleep) of depressed patients differed significantly fromthat of the healthy control group (F₂₉ = 2.8, P = .02). Secondary analyses showed differences in measures of sleepcontinuity, NREM sleep, and REM sleep (Table 2). No significant group differences were found in the EEGsleep distribution of waking, REM, and NREM sleep during the initial [¹⁸F]-FDG uptake period of the REM sleep study. The mean ± SDnumber of 20-second epochs of REM, wake, and NREM sleep, respectively, inthe 20 minutes following injection of [¹⁸F]-FDG were 49 ±9, 2 ± 2, and 8 ± 8 for the controls and 50 ± 11, 3 ±4, and 8 ± 8 for the depressed group.

We predicted that the decline in whole-brain MRDglc from waking to REMsleep would be less in depressed patients than in healthy subjects. A trendtowards this group (depressed vs control) × state (wake vs REM sleep)interaction was noted (1-tailed P = .08). We predictedthat depressed patients would show greater MRDglc in REM sleep. A trend wasnoted (1-tailed P = .09). No significant differencein waking MRDglc was found (depressed = 8.14 ± 2.18 µmol/(100mL · min) and control = 8.66 ± 1.51 µmol/(100 mL ·min) t₃₁ = .71, 2-tailed t test, P = .48). A main effect of state wasnoted (REM MRDglc < wake MRDglc, F_1,30 = 9.2, P = .005).

In healthy subjects (Figure 1),relative metabolism increased from waking to REM sleep in a broad collectionof anterior limbic and paralimbic structures with some tendency towards righthemispheric increases (cluster level P = .002 correctedfor multiple comparisons; voxels in cluster = 1765; voxel of maximum significancewithin cluster at Talairach x, y, and z coordinates 6, 44, and 16). One largeconfluent region extended posteriorly and superiorly from the supplementarymotor area, then arched anteriorly and inferiorly in the dorsal anterior cingulateand medial prefrontal cortex. This region continued into pregenual and subgenualanterior cingulate cortex, the nucleus accumbens, anterior ventral pallidum,anterior ventral caudate, and lateral hypothalamus. Inferiorly and laterallyin the right hemisphere, this region continued into the amygdala and uncus,then arched laterally and posteriorly along the right hippocampal gyrus. Finally,this region extended into the right medial temporal cortex and into the rightinsular cortex.

In the depressed subjects (Figure 1), relative metabolism increased from waking to REM sleep in a broadcollection of anterior paralimbic structures (cluster level P<.001 corrected for multiple comparisons; voxels in cluster = 9883;voxel of maximum significance with cluster at Talairach x, y, and z coordinates8, 4, and 32). The spatial extent of activation in depressed patients wasmuch broader (9883 voxels in anterior paralimbic cortex cluster in depressedsubjects vs 1765 voxels in this region in healthy control subjects) and morebilateral in nature, and some regional variations seemed apparent (see resultsof interaction analysis below). The posterior limit of the anterior paralimbicregion bilaterally was in the superior parietal cortex (Brodmann area [BA]7), extending anteriorly including primary sensorimotor cortex bilaterally(BA 1-5), cingulate cortex bilaterally (BA 23, 24), the supplementary motorarea, and the premotor area (BA 6). Laterally, this region continued bilaterallyinto dorsolateral prefrontal cortex. Medially, beginning at the dorsal anteriorcingulate (BA 24, 32), this region continued predominantly on the right hemisphereinto pregenual anterior cingulate and medial prefrontal cortex. This regiondid not extend prominently into the subgenual anterior cingulate cortex (BA25). More posteriorly increased relative metabolism was seen in ventral pallidum/basalforebrain. Increased relative metabolism of basal ganglia was present onlyon the left side. There was increased relative metabolism of bilateral insularcortex. Significant increases in relative metabolism were seen in the rightbut not left hippocampus. In the brainstem, the midbrain reticular formationshowed increased relative metabolism in REM sleep (voxel level P = .002; voxels in cluster = 266; voxel of maximum significance atTalairach coordinates −2, −30, −4). On the left side, thisarea was confluent with a region of relatively increased metabolism in theanterior lobe of the cerebellum. A similar increase in relative metabolismwas noted in the anterior lobe of the cerebellum on the right side.

After correcting for multiple comparisons across all brain voxels, depressedsubjects showed greater increases in relative metabolism from waking to REMsleep than healthy subjects in a broadly distributed region of predominantlyleft hemispheric dorsolateral prefrontal, parietal, and temporal cortex (Figure 1; cluster level P = .04 corrected; voxels in cluster = 1005; voxel of maximum significancewithin cluster at Talairach x, y, and z coordinates −22, 16, 36). Thelargest region was in the left dorsolateral prefrontal and parietal cortexextending anteriorly from the frontal cortex at Talairach x, y, and z coordinates−25, 43, 15, then arching dorsally along the cortical mantle to a posteriorborder in parietal cortex at x, y, and z coordinates −34, −58,47. This area included the dorsolateral prefrontal cortex (BA 46), the FrontalEye Field in the superior precentral sulcus (x, y, and z coordinates −25,−12, and 53), and the Parietal Eye Fields in the superior and inferiorparietal lobule (x, y, and z coordinates −25, −57, and 52 to −31,−41, and 40).

After correcting for multiple comparisons within a priori regions ofinterest, 3 additional areas reached significance. One area included lefthemispheric insular and superior temporal cortex (BA 22, 38, 41, and 42) rangingfrom y = −20 to 0 and from x = −39 to −53 (cluster level P = .023 uncorrected; voxels in cluster = 451; voxel ofmaximum significance within cluster at Talairach x, y, and z coordinates −42,−12, and 4). A second area included left hemispheric primary sensorimotorcortex (BA 1, 2, 3, 4) (cluster level P = .04 uncorrected;voxels in cluster = 378; voxel of maximum significance within cluster at Talairachx, y, and z coordinates −48, −18, and 32). A third area was inthe left-sided midbrain reticular formation including the pretectal area (voxellevel P = .007 uncorrected; voxels in cluster = 92;voxel of maximum significance at Talairach x, y, and z coordinates −4,−28, and 0).

To control for the possibility that overall lower waking metabolismin depressed patients may be driving the results, we first created a binaryimage mask that included all voxels showing a significant group × stateinteraction. We then used this mask in a small-volume correction analysisin a group comparison (depressed vs control) of waking metabolism. Structuresshowing both the interaction and waking hypometabolism in depressed patientsincluded bilateral superior temporal gyrus (right hemisphere 164 voxels, P<.001 at x, y, and z coordinates −56, −16,and 16; left hemisphere 57 voxels, P = .03 at x,y, and z coordinates 58, −14, and 12) and a trend towards significancein the left superior frontal gyrus (BA 10) (37 voxels, P = .07 at x, y, and z coordinates −22, 52, and −4). Tosee if the observed interactions were associated with relative REM sleep hypermetabolismin the depressed patients, we used the same binary image mask in a small-volumecorrection analysis in a group comparison (depressed vs control) of REM sleepmetabolism. Structures showing both the interaction and REM sleep hypermetabolismin depressed patients included a large cluster (316 voxels, P<.001) inclusive of left primary sensorimotor cortex, inferiorparietal cortex, and left temporal cortex; a second large cluster (381 voxels, P< .001) inclusive of the left superior temporal gyrusand the left frontal eye fields, at Talairach x, y, and z coordinates −28,20, 36 (90 voxels, P = .01).

After we corrected for multiple comparisons across all brain voxels,healthy subjects did not show greater increases in relative metabolism inany area from waking to REM sleep than depressed subjects.

We performed a contrast that corrected for multiple comparisons withina priori regions of interest. Control subjects showed greater increases fromwaking to REM sleep in both the subgenual anterior cingulate cortex (Talairachx, y, and z coordinates 10, 26, and −4) and the pregenual anterior cingulatecortex (x, y, and z coordinates 6, 44, and 16). In depressed patients, wakinghypermetabolism in this region of the anterior cingulate was noted (302 voxels; P<.05 corrected; maximum significance at Talairach x,y, and z coordinates −8, 32, and 20).

Depressed patients showed increases in relative metabolism from wakingto REM sleep in the midbrain reticular formation and in a larger region ofanterior paralimbic cortex than did healthy control subjects. In addition,relative to controls, depressed patients showed greater activation of executivecortex, including cortical eye fields and bilateral dorsolateral prefrontalcortex, from waking to REM sleep. These findings support our hypothesis thatdepressed patients would demonstrate a greater activation of limbic and anteriorparalimbic structures in a waking to REM sleep functional neuroimaging probe.

While the EEG sleep profile of the depressed group was similar to priorreports, some features did not differ from our control sample. Depressed patientsdisplayed poorer sleep efficiency and longer sleep latencies, and they hada greater percentage of stage 1 sleep, although no reductions in slow-wavesleep were observed. They also had a greater REM sleep percentage. The differencesfor REM density and REM latency were in expected directions but did not reachstatistical significance.

The first primary finding in this study is the increased activationof the brainstem reticular formation from waking to REM sleep in depressedpatients. This is consistent with the model of an altered balance in brainstemmonoaminergic (norepinephrine and serotonin) systems and brainstem acetylcholineneuronal systems in depressed patients as proposed by McCarley.⁴¹ Rapideye movement sleep is generated by cholinergic nuclei in the brainstem interactingreciprocally with monoaminergic cell groups.¹³ Asinhibitory monoaminergic input declines from waking to NREM sleep and ceasesin REM sleep,⁴² cholinergic REM–generatingcells are disinhibited. This altered balance may be related to a supersensitivecholinergic system in depressed patients, as evidenced by supersensitive responsesto cholinergic REM–generating agents in depression.^43- 45 Alternatively,or in addition, a fundamental alteration in monoaminergic tone could leadto the observed findings. For example, application of 5-hydroxytryptamine1A (5HT1A) agonists into the raphe decreases postsynaptic serotonin release,leading to an increase in REM sleep.^46- 47 Boutrelet al⁴⁸ described increased amounts of REMsleep in 5-HT1A knockout mice. In light of other work,^49- 51 thetonic inhibitory role of serotonin on REM sleep in the 5-HT1A knockout studymay be mediated by postsynaptic 5-HT1A receptors in mesopontine cholinergicREM-on neurons.

A second primary finding in this study is the increased activation oflimbic and anterior paralimbic (hippocampus, basal forebrain/ventral pallidum,anterior cingulate, and medial prefrontal) cortex from waking to REM sleepin the depressed patients. Activation of these areas in REM sleep has previouslybeen noted in cats¹⁵ and healthy human subjects.^16- 18 Mesulam et al⁵² have shown that the highest density of cholinergicaxons is in core limbic structures such as the hippocampus and amygdala, followedby nonisocortical then isocortical sectors of paralimbic cortex, and lastlyin primary sensory, unimodal, and heteromodal association areas. Limbic andanterior paralimbic cortices also have high densities of inhibitory 5-HT1Apostsynaptic receptors in relation to other areas of the cortex.^53- 55 Increasedactivation of limbic and anterior paralimbic structures from waking to REMsleep in depressed patients, therefore, may also reflect a monoaminergic/cholinergicimbalance in the forebrain in addition to that seen in the brainstem reticularformation.

A third primary finding in this study is the relatively greater activationof executive cortex from waking to REM sleep in depressed patients. Studieshave shown reduced executive function in depression.⁵⁶ Afundamental difference between waking brain function and REM-sleep brain functionis the degree to which cortical activation is influenced by monoaminergicvs cholinergic ascending projections.⁵⁷ Inwaking, depression-associated reductions in monoaminergic function may accountin part for reductions in executive cortex function. In REM sleep, a cholinergicallydriven state, the depression-associated increased cholinergic tone and reducedmonoaminergic tone may not only activate cholinergically rich limbic and paralimbiccortex but also provide significant activation of other cortical areas notactivated in healthy subjects during REM sleep. This activation may occurvia brainstem cholinergic projections to the cortex or via basal forebraincholinergic projections that are also under modulatory influence of monoaminergicsystems.⁵⁸

Findings in affective neuroscience suggest that our PET findings maybe fundamentally related to the behavioral features of depression. Functionalneuroimaging studies have found that hippocampus, amygdala, and anterior cingulatecortex, structures showing supersensitive activation in REM sleep in depressedpatients, also activate in response to negatively valenced stimuli or increasedaffective states.^59- 64 Giventhe negative affect of depressed patients during waking, we speculate thatthe increased activation of these structures in depressed patients may reflecta susceptibility of depressed patients to experience stimuli in a more affectivelyintense, negative context. Some support for this comes from EEG sleep studiesin depressed patients, which showed an association of increased REM densitywith greater intensity of affect.⁶⁵ Both REMdensity and intensity of affect declined following resolution of depressivesymptoms.

Observations from cognitive neuroscience suggest that our findings inthe dorsolateral prefrontal cortex may reflect a greater involvement of executivefunction during REM sleep in depressed patients, perhaps in response to theincreased affective state produced by the abnormal activation of limbic andparalimbic cortex during REM sleep in depressed patients. Studies of cognitivecontrol suggest that the anterior cingulate cortex plays a role in the monitoringof cognitive performance, while the dorsolateral prefrontal cortex plays arole in the implementation of cognitive control.^66- 67 Activationof the anterior cingulate cortex within REM sleep^16- 18 mayreflect an internal monitoring process assessing the presence of affectivelyarousing stimuli. In depressed subjects, the additional bilateral activationof the dorsolateral prefrontal cortex is consistent with the recruitment ofhigher-order cognitive processes that may be recruited by depressed patientsto process the negative affective state. The increased activation of the frontaland parietal eye fields during REM sleep in depressed patients, which areknown to underlie the awake control of eye movements,⁶⁸ furthersupport an increased cognitive involvement within REM sleep in the depressedsubjects. If they were solely related to the eye movements of REM sleep, wewould have expected them to be right- as opposed to left-lateralized, giventhat eye movements are strongly right lateralized.⁶⁹

In relation to prior EEG sleep studies on depression, therefore, alterationsin REM sleep may reflect an increased sensitivity of the brainstem cholinergicREM– generating system, an increased stimulation of limbic and paralimbiccortex, and an additional involvement of executive cortex in REM sleep. Thesealterations in forebrain function in depression may be attributed to an imbalancein monoaminergic/cholinergic function in depression that affects not onlythe brainstem generation of REM sleep but also the manner in which the forebrainresponds to the stimuli of REM sleep. In the context of affective and cognitiveneuroscience studies of the behavioral roles of these structures, the increasedactivation of limbic and paralimbic cortex may reflect an increase in affectiveresponsivity in depression. Involvement of executive cortex during REM sleepin depression may reflect the recruitment of the cortex to develop strategiesfor managing the heightened affective arousal in depressed patients. Futurestudies are needed to clarify the relationships between these forebrain patternsand an imbalance in monoaminergic/cholinergic systems that may underlie depressiveneurobiology. Future studies are also needed to characterize the relationshipbetween affective and cognitive processes that may be abnormal in depressionand that may produce these forebrain patterns observed in REM sleep.

Corresponding author: Eric A. Nofzinger, MD, Western PsychiatricInstitute and Clinic, 3811 O'Hara St, Pittsburgh, PA 15213-2593 (nofzingerea@upmc.edu).

Submitted for publication July 28, 2003; final revision received November10, 2003; accepted February 3, 2004.

This research was supported in part by grants MH61566, MH66227, MH37869,MH01414, MH30915, RR00056, MH24652, and MH52247 from the National Institutesof Health, Bethesda, Md and Glaxo Wellcome Inc, Research Triangle Park, NC.

We thank the technical staffs of the Sleep Imaging Research Program,the Clinical Neuroscience Research Center, the General Clinical Research Center,the PET Center, and the Depression Treatment and Research Program at the Universityof Pittsburgh Medical Center for their help in conducting this work. We alsothank the anonymous reviewers of this work whose comments have been integratedinto the report.