Imaging the pathophysiology of major depressive disorder

14 Pages·2014·0.92 MB·English

Imaging the pathophysiology of major depressive
disorder - from localist models to circuit-based
analysis
Citation
Treadway, Michael T., and Diego A Pizzagalli. 2014. “Imaging the pathophysiology of major
depressive disorder - from localist models to circuit-based analysis.” Biology of Mood & Anxiety
Disorders 4 (1): 5. doi:10.1186/2045-5380-4-5. http://dx.doi.org/10.1186/2045-5380-4-5.
Published Version
doi:10.1186/2045-5380-4-5
Permanent link
http://nrs.harvard.edu/urn-3:HUL.InstRepos:12152827
Terms of Use
This article was downloaded from Harvard University’s DASH repository, and is made available
under the terms and conditions applicable to Other Posted Material, as set forth at http://
nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of-use#LAA
Share Your Story
The Harvard community has made this article openly available.
Please share how this access benefits you. Submit a story .
Accessibility
TreadwayandPizzagalliBiologyofMood&AnxietyDisorders2014,4:5
Biology of
http://www.biolmoodanxietydisord.com/content/4/1/5
Mood & Anxiety Disorders
REVIEW Open Access
Imaging the pathophysiology of major depressive
disorder - from localist models to circuit-based
analysis
Michael T Treadway* and Diego A Pizzagalli*
Abstract
The neuroimaging literature of Major Depressive Disorder (MDD)has grown substantially overthelast several
decades, facilitating great advances inthe identification of specific brain regions,neurotransmitter systems and
networks associated withdepressiveillness.Despite this progress, fundamental questions remain about the
pathophysiology and etiology ofMDD. More importantly, this body of work has yet to directly influence clinical
practice. It has long been a goal for thefields of clinical psychology and psychiatryto have a means of making
objective diagnosesof mental disorders. Frustratingly littlemovementhas been achieved onthis front, however,
and the ‘gold-standard’of diagnostic validity and reliability remains expert consensus. Inlight of this challenge, the
focus of thecurrent review is to provide acritical summary of key findingsfrom different neuroimaging approaches
inMDD research, including structural, functional and neurochemicalimaging studies. Following this summary, we
discuss some of the current conceptual obstacles to better understanding the pathophysiology of depression, and
conclude withrecommendations for future neuroimaging research.
Keywords: Major Depression, Neuroimaging, PET, MRI, Serotonin, Dopamine, MRS, Glutamate, GABA, Inflammation
Introduction Even more importantly, this body of work has yet to
The neuroimaging literature of Major Depressive Dis- influence clinical practice in any substantive way. A
order (MDD) has exploded in recent years, with the longstanding goal of clinical psychology and biological
current pace of research including over 250 new articles psychiatry research has been the development of object-
listed each year in PubMed alone. A substantial majority ive tests for diagnosing mental disorders. Frustratingly
of these studies have been focused on identifying pu- little progress has been made on this front, however, and
tative biological and neural variables that differentiate the ‘gold-standard’ of diagnostic validity and reliability
individuals with MDD from psychiatrically healthy con- remains expert consensus, a practice that is essentially
trols. This program of research has been successful in unchanged from Meehl’s day [1]. Despite our capacity to
demonstrating a large number of abnormalities in MDD measure an astonishing array of biological signals in
samples, including alterations across measures of brain MDD patients, we have yet to find a single measure - or
structure and function; endocrine, immune and neuro- a combination of variables - that tracks symptom ex-
transmittersystems;andlarge-scalenetworkorganization. pression with the requisite specificity and sensitivity to
Despite this progress, however, fundamental questions bereliably meaningfulinthe clinic.
remain about the pathophysiology and etiology of MDD It is against this backdrop that we present the current
aswellasthestrengthsandpitfallsofneuroimagingmeth- review. First, we provide a (non-exhaustive) summary of
odologiesinattemptingtoanswerthem. the major findings that have emerged from different
neuroimagingapproaches.Thisincludesareviewofstruc-
tural,functional,neurochemical,neuroendocrineandneu-
roimmune imaging studies in MDD. Following this
*Correspondence:[email protected];[email protected].
edu summary, we discuss some of the current conceptual ob-
CenterforDepressionAnxietyandStressResearch,McLeanHospital/Harvard stacles to better understanding the pathophysiology of
MedicalSchool,115MillStreet,Belmont,MA02478,USA
©2014TreadwayandPizzagalli;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsofthe
CreativeCommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,
distribution,andreproductioninanymedium,providedtheoriginalworkisproperlycredited.TheCreativeCommonsPublic
DomainDedicationwaiver(http://creativecommons.org/publicdomain/zero/1.0/)appliestothedatamadeavailableinthis
article,unlessotherwisestated.
TreadwayandPizzagalliBiologyofMood&AnxietyDisorders2014,4:5 Page2of13
http://www.biolmoodanxietydisord.com/content/4/1/5
depression,andpresenttheuseofcircuit-basedanalysisas studies provide the strongest evidence linking these
amethodologicalpathforward. changestotheonsetandremissionofadepressivestate,
suggesting that they are either causally involved in
Review MDD or that they at least meaningfully fluctuate with
NeuroimagingandthepathophysiologyofMDD illness progression. Future longitudinal work, especially
Morphometricneuroimagingstudies with prospective-cohort designs, will help further eluci-
A large number of studies in MDD patients to date have date the role of these morphometric alterations in the
identified structural alterations across multiple tissue etiopathophysiologyofMDD.
classes. These findings have been summarized using me
ta-analytic approaches reporting on structural alter- Functionalneuroimagingstudies
ations observed using regions-of-interest (ROI) tracing- Functional imaging studies of MDD have also grown
based methods [2,3], voxel-based methods (VBM) [4], substantially over the last two decades, with a wide var-
post-mortem tissue analysis [5], and diffusion tensor im- iety of functional domains probed using a large number
aging of white-matter integrity [6]. Tracer-based methods of tasks. We will therefore limit our focus to two general
have provided especially strong evidence for reduced domainsthathavebeenmostfrequentlyexaminedinde-
hippocampal volume and enlarged ventricles in MDD pression studies: (1) emotion processing and regulation,
[3]. These results have been recapitulated by voxel-based and (2) reward processing (Figure 1). For a more com-
methods, which additionally implicate a more distributed prehensive discussion of other aspects of the functional
network of structural alterations associated with MDD, neuroimaging literature in major depression, we refer
includingtheanteriorcingulatecortex(ACC),medialpre- readers to several excellent recent reviews and meta-
frontal cortex (mPFC), orbitofrontal cortex (OFC), dorso- analyses:[18-20].
lateral prefrontal cortex (dlPFC), the striatum, and the
amygdala. A limitation of these findings is that they are Functionalneuroimagingofemotionprocessing Argu-
mostlydrawnfromcross-sectionaldesigns.Therefore,itis ably the most common domain assessed by functional
difficult to know whether such structural differences rep- imaging studies of depression is responses to emotional
resent a biological diathesis, a compensatory adaptation, stimuli. Examples include studies of responses to both
oraconsequenceoftheillness. explicit and implicit presentations of affect-laden stimuli
A handful of longitudinal studies have addressed this [21-23],recruitmentofcognitive controlmechanismsre-
significant limitation by investigating structural changes quired to gate out affective ‘distractors’ during simple
as a function of depressive state and treatment outcome. working memory and attention tasks [24,25], and deli-
\Forsomeregions,includingthehippocampusandmedial berate top-down control of affective responses to positive
prefrontal areas, several studies have suggested that grey- and negative stimuli [26-29]. The most replicated result
matter volume may decline monotonically over multiple observed duringpassive presentation of emotionalstimuli
depressive episodes [7,8]. In contrast, the amygdala may is a heightened responsivity in limbic regions - especially
become enlarged prior to a first depressive episode [9]. theamygdala-tonegativelyvalencedstimuliindepressed
Structuralintegrityoftheseregionshasalsobeenfoundto individuals. For tasks that require subjects to efficiently
partially predict symptom remission. Specifically, hippo- ‘gate-out’ affective content in order to better attend to
campal volumes have been found to prospectively correl- non-emotionalaspectsofataskorstimulus, elevatedlim-
ate with treatment outcome at both one- and three-year bic activity is often accompanied by hypo-activation in
follow-ups [10,11], and longitudinal studies have found prefrontal areas, including aspects of ventromedial PFC,
thatdecreasedhippocampalvolumeswerepartiallyrestored ventrolateral prefrontal cortex (vlPFC), ACC, and dlPFC.
following successful treatment or spontaneous remission Itisnoteworthythatthesesameregionsfrequentlyexhibit
[11-13]. Importantly, similar morphometric changes in volumetricabnormalities.
these regions have also been associated with high levels While prefrontal hypo-activations are commonly inter-
oftraitnegativeaffectinnon-depressedindividualswho preted as evidence of a top-down control ‘deficit’, it is
have an elevated polygenic risk profile for developing unclear whether they reflect a local deficit in network re-
MDD. This has been observed using both genome-wide cruitment or simply a failure to engage in the task as ef-
analysis [14] and examinations of non-depressed indi- fectively as controls.Interestingly,when task performance
viduals withafamily historyofMDD[15,16],consistent is matched across depressed and non-depressed individ-
withthehypothesisthatthesestructuraldecreaseslikely uals, there is evidence for hyper-response in prefrontal
reflectanendophenotypemarker[17]. areas [30,31], possibly indicating cortical inefficiency. In
Taken together, structural imaging studies have found addition, the specificity of alterations in amygdalar and
robust evidence for group-level differences in grey-matter prefrontal networks to depression is unclear, as similar
volumeacrosscorticalandsub-corticalareas.Longitudinal patternsarefrequentlyobserved instudiesofanxiety,and
TreadwayandPizzagalliBiologyofMood&AnxietyDisorders2014,4:5 Page3of13
http://www.biolmoodanxietydisord.com/content/4/1/5
Figure1Regions,transmittersandcircuitsimplicatedinthepathologyofmajordepressivedisorder(MDD)byhumanneuroimaging
studies.PaststudieshaveidentifiedalterationsinmonoaminelevelsandreceptoravailabilityaswellasalterationsinglutamateandGABA.These
neurotransmittersystemsparticipateinlargercircuitsinvolvedintheexperienceandregulationofemotion,responsestostress,andprocessingof
rewards.Note:placementofstructurelabelsisapproximate.Amyg=amygdala;Caud=Caudate;GABA=GABAergicprojections;Glu=glutamatergic
projections;Hipp=hippocampus;NAcc=nucleusaccumbens;Put=Putamen;SN=substantianigra;VP=ventralpallidum;VTA=ventraltegmental
area.RepublishedwithpermissionfromTreadwayandZald[49].
onlyafewdirectcomparisonstudieshavebeenconducted while others have found no differences [27-29]. These
todate[32,33].Futureresearchisneededtofurtherisolate studies have also generally failed to observe impaired
thespecificalterationsincortico-limbicresponsestoemo- cortico-amygdala interactions during explicit emotion
tion in MDD, and to determine the extent to which these regulation in MDD. Consequently, these data suggest
effects are specific to a depressed mood or rather repre- that emotion regulation deficits in MDD do not reflect
sentacommonmechanismassociatedwithotherformsof a true inability to regulate emotion when explicitly di-
internalizingpsychopathology. rected to do so, at least not in the context of typical
In contrast to experimental paradigms that require laboratory-basedaffectivestimuli.
either passive emotional processing or implicit emotion
regulationintheform ofattentional control, findingsof Functional neuroimaging of reward processing An-
studies of directed emotion regulation in MDD are other primary area of functional neuroimaging research
highly variable. In healthy controls, down-regulation of in MDD involves responses to rewarding stimuli. While
negative emotion has been consistently associated with early functional magnetic resonance imaging (fMRI)
increased activation in medial and dlPFC areas and re- (and non-imaging) studies frequently operationalized re-
ducedactivityintheamygdala[34].Theseobservations, ward in terms of the passive viewing or consumption of
combined with observations of impaired functional positively valenced stimuli (for example [36-38]), more
coupling between mPFC and amygdala during passive recent work has increasingly emphasized constructs
viewing of affective stimuli [35], led investigators to of reward anticipation [39-42], reinforcement learning
hypothesize that depressed patients would be less suc- [43,44] and motivation [45-47], which are psychologi-
cessful in reducing amygdala reactivity - and associated cally and neurobiologically distinct. This shift has been
negative emotions - when explicitly regulating emo- motivated largely by the enhanced understanding of
tional responses to negative stimuli. Empirical support functional segregation of dopaminergic cortico-striatal
for this hypothesis, however, has been mixed. Only one systems in reward processing, which have been found to
study has reported that depressed patients experience underlie anticipation, learning, and salience of rewards,
moredifficultyindecreasingsadnessthancontrols[26], rather than affective responses to them [48]. Indeed,
TreadwayandPizzagalliBiologyofMood&AnxietyDisorders2014,4:5 Page4of13
http://www.biolmoodanxietydisord.com/content/4/1/5
reward-related symptoms are especially amenable to a Neurochemical imaging of serotonin systems in MDD
translational neuroscience approach, given how well Interest in serotonin (5-HT) has been central to depres-
characterized reward-related pathways are by both pre- sion research over the last three decades, owing primarily
clinical and human neuroeconomic studies (see [49] for a to reported success of antidepressant pharmacotherapies
longer discussion). The most common observation from that selectively target the serotonergic system in both
thisbodyofworkishypo-recruitmentinMDDpatientsof humans and animal models. Evidence from preclinical
striatal regions associated with reward salience, anticipa- studies further supports a role for serotonin in MDD
tion and learning, possibly reflecting alterations in the symptoms, particularly those related to the processing of
availabilityofpre-synapticpoolsindopaminergicafferents stress.Undernormalconditionsofwakefulness, serotonin
tostriatalsub-regions[50-52](seealsodiscussionofdopa- neuronsaretonicallyactive[55]andthedistributionofse-
mine imaging studies below). In addition, altered cross- rotonergictoneisrelativelyevenacrossmostbrainregions
talk between cortical and ventral striatal regions has been [56] - a pattern that has been found to support normal
associated with rapid habituation to rewarding stimuli, network functioning for a variety of cognitive and goal-
whichisalsoconsistentwithanhedonicpresentation[53]. directedbehaviors.Incontrast,exposuretostresscanpro-
In sum, these studies highlight cortico-striatal path- duce a surge in 5-HT signaling, which has been found
ways as critically involved in specific symptom domains to disrupt the emotion-regulatory functions of cortico-
of MDD. Of note, there is arguably greater consistency amygdalar networks [57]. Further, evidence suggests that
in studies of reward processing in MDD than of other medial prefrontal projections to serotonin-releasing neu-
cognitiveprocesses.This may reflectthe fact thatreward ronsinthedorsalrapheplayacrucialroleindetermining
processing studies have focused on a more homoge- adaptive versus non-adaptive responses to stress [58,59].
neous symptom domain and that the neurobiology of Consequently,impairedserotoninsignalingmaybeasub-
normativerewardfunctioning isbetterunderstood. strate involved in stress vulnerability and a key risk factor
inthedevelopmentofMDD[60-62].
For these reasons, serotonin is among the most widely
NeurochemicalimagingstudiesinMDD imaged neurochemical systems in MDD, with over 35
The hypothesis that specific neurotransmitter systems studies exploring group differences in the expression of
represent a core pathology of mood disorders is among serotonin receptor sub-types as well as the serotonin
the oldest in biological psychiatry (see, for example, transporter (for recent reviews, see [63,64]). To date,
Schildkraut [54]). For most of modern psychiatric history, however, results have been mixed, with investigators fre-
this line of work has emphasized alterations in mono- quently reporting higher or lower serotonin receptor or
amines, given early observations that administration of transporter expression inMDDparticipantsthanincon-
various monoamine-reducing drugs or pharmacological trols [63]. For example, of the 15 studies investigating
manipulations could induce depressive symptoms. It was expression of 5-HT receptor in depressed patients
1A
onlydecadeslater,however,whentheinvivovisualization relative to healthy controls, nine reported decreased ex-
ofthesesignalingpathwayscouldbeachieved. pression in MDD, four reported increased expression,
Currently, the two most widely used approaches to and two observed no change. Similar discrepancies have
neurochemical imaging in psychiatric populations are been observed for other proteins involved in 5-HT sig-
Positron Emission Tomography (PET) and Magnetic naling pathways, including the 5-HT receptor, 5-HT
2A 1B
Resonance Spectroscopy (MRS). A less commonly used receptor,andtheserotonin transporter (SERT).
technique is Single Photon Emission Computed Tom- It is important to note that most of these studies are
ography (SPECT). Both PET and SPECT rely on the relatively small in size (between 9and 22 MDD patients)
measurement of radioactive decay from an injected iso- and are, therefore, underpowered to explore within-
topeasthebasisoftargetingthespatialdistributionofa sample relations between serotonin function and specific
particular receptor or protein. In contrast, MRS takes symptomdimensions. Thisisapotentiallycritical limita-
advantage of the different magnetic resonance signa- tion, as the substantial heterogeneity of MDD is likely to
tures associated with distinct molecular compounds, be associated with divergent effects on neurotransmitter
andcanbeuseful forquantifying theavailabilityofrela- systems. In addition, most of these studies have not in-
tively abundant neurotransmitters such as glutamate vestigated the function of serotonin signaling systems,
(Glu) and γ-aminobutyric acid (GABA). Both of these compared to baseline expression. Moreover, no longitu-
methods have contributed to the study of pathophysi- dinal studies have been performed. Therefore, as with
ology in MDD, and are notable for both their positive cross-sectionalstudiesofstructureorfunction,itisdiffi-
and null findings. In this section,we reviewsome of the cult to know whether 5HT abnormalities should be
primary neurotransmitter systems that have been inves- interpreted as a primary deficit, a downstream conse-
tigatedinMDDusingthesetechniques. quence, a risk factor, or a compensatory mechanism.
TreadwayandPizzagalliBiologyofMood&AnxietyDisorders2014,4:5 Page5of13
http://www.biolmoodanxietydisord.com/content/4/1/5
What is clear, however, is that to the extent that the is roughly equivalent. One study reported reduced D1
5-HT system is involved in the etiopathophysiology, its availability in left middle caudate [84], but this finding
effect size is modest, and likely dependent on interac- has not yet been replicated. Taken together, these studies
tionswith numerous othersystems. suggestapossibleroleofD2-likereceptorsindownstream
effects of antidepressant treatment, although the precise
Neurochemical imaging of catecholamine systems in nature of the effect and how alterations in D2-like recep-
MDD Other monoamines that have long been associ- toravailabilitymayberelatedtoDAfunctionareunclear.
ated with MDD are the catecholamines dopamine (DA) As with other conflicting reports in neuroimaging
and norepinephrine (NE) [65,66]. DA is well established studies of MDD, part of the discrepancy across studies
as being necessary for motivation, reward-based learn- likely reflects the heterogeneity of the disorder. Support-
ing, and goal-directed behavior [48,67,68] and, therefore, ingthisassertionistheobservationofslightlymore con-
is believed to be a substrate of reward-related symptoms sistent effects when MDD samples are selected on the
such as anhedonia, fatigue, and anergia in psychiatric basis of a particular symptom profile. For example, one
disorders [69,70]. Unlike 5-HT, which is relatively uni- study that restricted its MDD patient sample to individ-
form in its distribution across the brain, DA expression uals with anhedonic symptoms reported decreased DAT
is densest in the striatum, a key structure involved in binding [85]. In addition, L-DOPA alterations in the stri-
valuation,decision-making andaction. atum are present in depressed individuals with flat affect
NeuroimagingevidenceforalteredDAsystemsinMDD or psychomotor slowing, but not in depressed individuals
comes primarily from PET, SPECT and pharmacological withoutthesesymptoms[86,87].DecreasesinDA synthe-
challenge studies. This research has found that MDD sis have also been observed in patients who develop de-
is associated with changes in DA synthesis capacity as pressive symptoms after undergoing IFN-α therapy [50].
indexed by L-3,4-dihydroxyphenylalanine (L-DOPA) up- This therapy stimulates inflammation signaling cascades,
take [71], as well as changes in the regional distribution which have been found to disrupt DA synthesis, and may
and availability of DA receptors, and the DA transporter provide a link between elevated inflammation in MDD
(DAT).Aswiththe5-HTstudiessummarizedabove,how- and specific symptoms related to perturbances of DA
ever, imaging studies of DA systems have produced con- signaling, such as motivation and anhedonia [50,88].
flicting results. In PETand SPECTstudies of DAT, MDD Overall, these studies provide mixed evidence for gen-
has been associated with both lower [72] and higher eral DA alterations in MDD, with additional evidence
[73-75] binding potential in the striatum. Interestingly, all highlightingtheimportanceof examining links between
studies reporting DAT increases have used SPECT, which DAsystemsandspecificsymptomsinMDD,ratherthan
has much lower sensitivity than PET [76] and often inthedisorderasawhole.
employed tracers that have equal affinity for the SERT In contrast to DA, molecular imaging methods of NE
andDAT(forexample,β-CIT)andthusdonotallowcon- signaling pathways have been much slower to develop.
clusive interpretations. Moreover, post-mortem studies Currently, only studies of the NE transporter (NET) have
supporttheobservationofreducedDATexpression[77]. been performed in MDD [89], with no studies examining
Studies of DA receptor availability in MDD have also NE receptors in MDD due to a lack of available ligands.
yielded mixed results. In some cases, increased striatal Pharmacologicalfunctionalimagingstudieshavealsobeen
D2/D3receptorbindinghasbeenshowntooccurinhet- used to indirectly explore effects of NE-increasing agents,
erogeneous depressed samples [78,79]. This increase in though many of these studies have used drugs such as
D2/D3 receptor availability appears to contradict animal duloxetine, which simultaneously target both 5-HT and
data in which antidepressant responses are associated NE transporters. Duloxetine reduced connectivity within
with increased D2-like binding in the striatum [80]. resting-stateandtask-positivenetworks,[90],andboosted
Other studies using medication-naïve or medication-free ventral striatal responses during a reward task [91], while
patients have failed to find group differences in striatal the NET-selective agent reboxetine increased thalamic
receptor binding [81,82], while one additional small dorsolateral prefrontal responsivity to emotional pic-
study reported variable changes in D2-like binding fol- tures [92,93]. While these studies provide promising
lowing treatment with selective serotonin reuptake in- leads,insufficientfunctionalormolecular imagingwork
hibitors (SSRIs) with patients who showed increased of NE function in the context of MDD is available, des-
binding exhibiting more clinical improvement than pitesignificantevidenceforitsroleinthedisorder[94].
thosewhodidnot[83].WithrespecttotheD1receptor,
fewer studies have examined this system given the lack NeurochemicalimagingofglutamatergicandGABAergic
of available ligands that reliably distinguish between the systems in MDD In recent years there has been sub-
D1 and serotonin 5-HT receptor, especially in extra- stantial interest in the contribution of non-monoamine
2A
striatalareaswherethereceptordensityofD1and5HT neurotransmitters to the pathophysiology of MDD,
2A
TreadwayandPizzagalliBiologyofMood&AnxietyDisorders2014,4:5 Page6of13
http://www.biolmoodanxietydisord.com/content/4/1/5
particularlytheexcitatoryandinhibitoryaminoacidtrans- beensuggestedtoplayacriticalroleinmediatingnegative
mitters of glutamate (Glu) and GABA, respectively. At an feedback of hypothalamic-pituitary-adrenal (HPA)-axis
intuitive level, the hypothesis that these systems would be activity [115,116]. Thus, decreased GABAergic tone
implicated in depression holds significant appeal; the in- may foment excess glucocorticoid exposure in mPFC,
nervationofGlu-andGABA-releasingneuronsvastlyout- as reviewed above. The combination of increased glu-
numbers all other neurotransmitter systems in the brain, cocorticoid exposure and elevated GABA has been
makingthesetwoneurochemicalsresponsibleforthebulk hypothesized to be a combination that may lead to in-
of information processing related to learning, cognition, creased excitotoxicity in these regions, thereby partially
memory,anddecision-making[95].Whenconsideringthe explaining the structural alterations in these areas sum-
scope of this diverse functional anatomy, it is difficult to marized in the preceding section. To date, however, the
imagine that Glu and GABA would not be directly, or at number of studies focused on GABA is relatively small,
leastindirectly,involved. makingitdifficulttodrawfirmconclusions.
Evidence for alterations of Glu transmission in MDD
have long been reported, but findings have been mixed, Imaging neuroendocrine and neuroimmune systems
with increased Glu levels observed in plasma samples in MDD Lastly, there has been growinginterestin using
and post-mortem tissue as compared to decreased levels neuroimagingtostudythefunctionalandstructural con-
found in neuroimaging studies [96-98]. These discrepan- sequence of other neurochemicals, such as hormones,
cies may be due in part to the multiple roles that Glu factorsandotherendogenoussignalingmolecules.While
plays in the brain (for a more extended discussion, see technical limitations generally prevent the imaging of
[99]). A recent meta-analysis or MRS imaging studies such molecules directly, their effects on structure and
found that MDD was associated with a substantial de- metabolism can nevertheless be observed using MRI. In
creaseinGlulevelswithin themPFC/ACC[100],though the case of MDD, this work has focused most heavily on
it should be noted that not all studies were able to dis- pro-inflammatory factors, including families of cytokines
tinguish between Glu and glutamine, a common metab- such as interleukins and interferons, hormones such as
olite of astrocyte reuptake processes. Studies published glucocorticoids (cortisol), thyroid stimulating hormone
after this meta-analysis provided additional evidence of (TSH), andghrelin [117,118].
reduced Glu concentration in the mPFC of MDD sub- Dysregulation of stress hormones has been widely re-
jects [101-103], and similar alterations have also been ported in MDD (albeit with significant variance), and is
detected in children with depressive symptoms [104] as believed to partially mediate some of the observed struc-
well as remitted MDD subjects [102], raising the possi- tural alterations associated with the disorder, especially
bility that they constitute a trait-like vulnerability factor within the hippocampus and mPFC [119,120]. These re-
for MDD. Highlighting the clinical significance of these gions are well known for playing a critical role in the
findings, among MDD subjects, increased pre-treatment regulation of stress hormones via direct and indirect
Glu levels predicted better electroconvulsive therapy projections to the hypothalamus, and have been shown
(ECT)response[103]. to be structurally vulnerable to glucocorticoid-mediated
PET imaging studies of metabotropic Glu receptors excitotoxicity. Animal studies using either chronic stress
have also revealed changes in Glu signaling pathways in protocols or local corticosteroid injections have repeat-
MDD[105]andinrelationtoMDDsymptoms[106].The edly demonstrated tissue damage following excessive
rapid antidepressant effects of ketamine, an N-methyl-D- glucocorticoid exposure, including loss of dendritic
aspartate (NMDA) partialagonist[107,108] furtherimpli- spines and de-arborization [121-124]. In human studies,
catetheGlupathway.Finally,aberrationsinGlusignaling comparable relationships have been observed between
and Glu neurotoxicity have been associated with mPFC daily cortisol levels and grey-matter volumeindepressed
volumetric reductions discussed above[95].Insum,while patients [125]. Given that elevated stress is a major pre-
investigationofGludysfunctioninMDDisrelativelynew, cipitant of first-time depressive episodes [60], the associ-
given the near ubiquitous distribution of Glu signaling ation between stress and regional microdamage is highly
throughout the brain, it is likely that many of the alter- relevant.
ations in neural circuit function observed using fMRI In addition to stress hormones, depressive states have
studiespartiallyreflectGlu-relatedpathology. been strongly associated with an elevated inflammatory
IncontrasttoGlu,studiesofGABAarelessfrequentin load[126,127],andtherehasbeengrowinginterestinsig-
MDD. GABA alterations have been documented in MDD naling pathways related to the metabolic syndrome and
[17,109], including reports of reduced GABA levels in excess adipose tissue as potential mediators of chronic
plasma and cerebrospinal fluid [110-112], as well as spe- low-grade inflammation [128,129]. Neuroimaging has
cificGABAreductionsinthemPFCasassessedwithMRS therefore been employed as an aid to understanding
[113,114]. Moreover, GABA function in this region has the possible consequences of altered inflammatory and
TreadwayandPizzagalliBiologyofMood&AnxietyDisorders2014,4:5 Page7of13
http://www.biolmoodanxietydisord.com/content/4/1/5
metabolic factors on neural systems [130,131]. Animal based studies (for example, small-volume correction).
models suggest that elevations of peripheral cytokines Even when group differences emerge, they are often
and subsequent CNS microglia activation can disrupt present onlyat thelevelofgroup average, withcompar-
the synthesis of both 5-HTand DA [132,133], and in- able ranges for both groups (for example, [142]). As a
duce symptoms of fatigue and motivational anhedonia result, the field has been unable to identify any neural
[134,135]. While direct evidence of increased microglia signature that may serve as useful biomarker in the
activity in MDD has not been detected using available diagnosisofMDD,andguidetreatmentselection.
PET ligands [136], functional neuroimaging studies The lack of stable pathophysiological markers of MDD
have sought to better understand the downstream con- after so many neuroimaging studies raises the possibility
sequences of increased cytokine activity by examining that the biological origins of depressive symptoms are
correlations between peripheral cytokine levels and simply too dynamic to produce consistent patterns using
resting or task-induced fluctuations in Blood-Oxygen- case–control designs. For example, many receptor sub-
Level Dependent (BOLD) signals. Of note, healthy con- types, including those of monoamine, Glu and GABA
trols receiving an endotoxin challenge exhibited blunted pathways show rapid, activity-dependent changes in ex-
neuralresponsestorewardanticipationintheventralstri- pression and ligand affinity [143]. This plasticity may be
atum during a Monetary Incentive Delay (MID) task affected by the time of day, the amount of sleep some-
[137], similar to what is observed in MDD [42]. Further onereceivedtwonightsbefore,andhowmuchtheyhave
evidencecomesfromimagingstudiesofpatientsreceiving been taxing their working memory in recent weeks just
IFN-α therapy. Following IFN-α treatment - which ro- as much as it is affected by MDD ([105,143]). Against all
bustly increases inflammation - subjects exhibited de- these additional sources of variance, it is perhaps unsur-
creased DA turnover within the striatum, as measured prising that cross-sectional designs have been unable to
using a pre/post PET imaging of DA uptake with [18F] identifyreliablebiomarkers.
Dopa (F-DOPA) [50]. By beginning to localize the func- As a result, an increasing number of researchers have
tional architecture of brain regions that are affected by called for a better understanding of specific circuits
stress hormones and inflammation and by linking such that may mediate transdiagnostic symptom expression
abnormalities to specific cardinal symptoms of MDD (for [144,145]. While much of this discussion has rightly em-
example, anhedonia), these studies are playing a critical phasized the importance of animal models [146,147], the
roleinadvancingamorepreciseunderstandingofdepres- final section presents several conceptual and methodo-
sionsymptomatology. logical approaches to clinical imaging studies that we
feel may aid in the identification of circuits rather than
Summary of neuroimaging studies In review of the regions.
findings above, several themes emerge. The most prom-
isingresultisthatregardlessofimagingmodality,neuro- Futuredirectionsandcircuit-basedanalysis
imaging studies repeatedly isolate a similar network of A clear limitation of current neuroimaging studies in
regions in which MDD patients differ from controls. In- MDD has been the tendency to assess particular mea-
deed, the greatest success of neuroimaging studies in sures of brain function independently, despite clear evi-
MDD has been to identify core nodes involved in the ex- dence that these measures are highly inter-related. In
pression of depressive symptoms. Neural responses in contrasttomeasures ofspecific brainregions,chemicals,
cortico-striato-limbiccircuitshavebeenshowntodiscrim- or tissue classes, circuit-based analysis provides a con-
inatebetweenresponderstodifferenttreatmentmodalities ceptual framework that is organized around a specific
[20,138] and have been the empirical foundation for new behavioral process. Circuits are defined by the combined
treatment techniques, such as deep brain stimulation structural and functional properties that enable a spe-
(DBS) [139,140] and transcranial magnetic stimulation cific set of computations to be performed on a discrete
(TMS)[141]. set of inputs. As such, circuit-analysis integrates across
But this is perhaps where the good news ends; for many different levels and measures of brain function,
while the same regions are often implicated, the direc- but eschew the complexity of simultaneous whole-brain
tion of the effects is often contradictory (for example, analysis. By focusing on discrete computations with a
greater or lesser BOLD signal, depending on the task). circumscribed array of possible inputs and outputs, cir-
Moreover, some of this consistency is undoubtedly due cuit analysis meets the basic scientific requirement of
to confirmatory bias in ROI selection; that is, reports of simplification, but does so without neglecting biological
group differences in a given region increase the prob- complexity.
ability that future studies will focus on the region, either Oneofthemostpowerfulandreadilyavailablemethods
with targeted measurement(for example,volumetric tra- for circuit-analysis in patient populations is functional
cing) or with more liberal statistical thresholds in voxel- connectivity. Whereas much of the first two decades of
TreadwayandPizzagalliBiologyofMood&AnxietyDisorders2014,4:5 Page8of13
http://www.biolmoodanxietydisord.com/content/4/1/5
neuroimaging research emphasized localization of func- the scope of multi-modal imaging studies has been ex-
tion to a particular brain region while individuals were pandedtoincludeneurotransmitterssuchasmonoamines,
either at rest or performing a particular task, growing ap- Glu and GABA. In an important study by Northoff and
preciationfortheroleoffunctionalnetworkshasemerged colleagues, researchers identified shifts in neurotransmit-
inrecentyears.Thisshiftreflectsabroaderrecognitionin ter availability of Glu and GABA as major drivers of al-
the field of neuroimaging that the brain is comprised of tered functional responses to emotional stimuli in MDD
discrete networks, which show local properties (for ex- [152]. Similarly, decreases in pre-synaptic striatal DA as
ample, ‘small-worldness’) and can rapidly re-configure measured using F-DOPA were associated with blunted
themselves to adapt to current environmental demands ventral striatal BOLD signalduringrewardanticipationin
[148,149]. Consequently, focusing on task-driven changes depressed subjects, demonstrating a clear link between
within a single brain area provides only limited insight DAbioavailabilityandstriatalresponsestoreward[50].
into the specific computations being performed. Indeed, Another essential benefit of circuit-based analysis is the
significantprogresshasbeenmadeinthe characterization bridgeitcreatestobehavioralandmolecularneuroscience.
of several stable networks that support domain-general Animalmodelsplayapivotalroleintestingcausalhypoth-
functions,includingattentionalcontrol,noveltyandthreat esesaboutbrainfunction[146].Suchmodelsarenot only
detection, default-mode and social cognition, and rein- useful in aiding the interpretation of correlative neuroim-
forcementlearninganddecision-making,allofwhichhave aging data, they can also help refine the mapping of psy-
beenimplicatedinMDD[150](Figure2). chiatric symptom definitions to discrete and dissociable
Complementing functional connectivity as a path to- circuits. For example, the symptom of anhedonia - which
wards understanding MDD from a systems-level ap- wasonceperceivedasaunitaryconstructtappingintothe
proach is the growing use of multi-modal imaging. The experience of pleasure - has gradually come to be recog-
most common example is the combined use of struc- nized as involving multiple sub-components, including
tural and function neuroimaging data. Such data are motivation and hedonic response, each of which involves
often collected together, as standard preprocessing pipe- distinct neural substrates [69,70,153]. However, because
lines for functional neuroimaging data typically require these sub-components in humans are often heavily corre-
high-resolution structural scans [151]. More recently, lated with each other at the level of symptom expression,
Figure2Schematicdepictionofcommonlyidentifiedfunctionalnetworksandtheirassociatedcognitiveandsymptomdomains.
RepublishedwithpermissionfromBuckholtzandMeyer-Lindenberg[150].
TreadwayandPizzagalliBiologyofMood&AnxietyDisorders2014,4:5 Page9of13
http://www.biolmoodanxietydisord.com/content/4/1/5
animal models were required to demonstrate that motiv- the past week, but happens to have a brief lifting during
ational and hedonic aspectsofreinforcement wereindeed the two- to three-hour window in which the lab experi-
neurobiologicallydissociable[48,67]. ment occurs, it may be more important to consider the
Pharmacological manipulations and other interven- ‘in the moment’ affect rather than feelings aggregated
tional techniques also provide tremendous advantages over the past weeks when trying to interpret associated
over imaging measures alone; these studies can identify imaging data; this stands in contrast with the traditional
plasticity dynamics that can help unpack some of the assumption that individuals with shared symptom seve-
cross-sectional observations. For example, behavioral rity over a one-week period will have more variance in
pharmacology studies of DA in MDD suggest that pa- common than individuals with shared experience of a
tients experience a much stronger euphoria response to particular experimental task. One approach to address-
amphetamine than controls [154,155], possibly indicat- ing this challenge is the development of measures that
ing an up-regulation of post-synaptic DA receptors sites seek to tease apart ‘believing self’and ‘experiencing self’,
and/or DAT. The cross-sectional imaging studies dis- with the aim of identifying separate biological correlates.
cussed above, however, suggest down-regulation of both It is likely that both are implicated in the maintenance
D1andD2aswellasDATproteins.Sincetheexpression of depressive symptoms [163], but current symptom as-
of these proteins is dynamic, imaging studies done in sessment inventories are poorly suited to distinguish be-
conjunction with pharmacological challenges (within- tweenthesedistinct modesoftypesofsubjective report.
subject) hold the promise for stronger mechanistic infer- In sum, clinical imaging studies can contribute to
ences regarding how neurotransmitter systems are able circuit-based analysis through a focus on network-based
to adapt to changing conditions in MDD. Such pharma- analytical techniques, such as functional connectivity,
cological challenges can therefore greatly aid in the un- multi-modal imaging methods, the use of within-subject
derstanding of plasticity within circuits, and help shift pharmacological challenge designs, and greater sensitivity
away from a focus on the identification of a stable, per- to potential discrepancies between ‘believing self’ and
sistent pathological marker, which appears unlikely to ‘experiencingself’thatmaymaskimportantdistinctions
exist. in the relationships between subjective report and neu-
Finally, circuit-based analysis may help facilitate a shift roimagingdata.
in the conceptualization and measurement of psychiatric
symptom inventories that are often used to define
groups and regress against imaging data. Psychiatric Conclusion
measurement largely relies on subjective self-reporting The neuroimaging literature of depression has grown
of distress as they sine qua non of symptom diagnosis. tremendously over the last several decades. The primary
Whenitcomestoreportinghowonefeels,however,stud- fruit of these efforts has been the identification of brain
ies in healthy populations have increasingly observed dis- regionsandstructuresthataremostcriticaltotheexpres-
sociationbetweenthe‘believingself’andthe‘experiencing sion of depressive symptomatology, while also increasing
self’ [156-158]. Accordingly, while the former describes ourknowledgeofhowtheseregionsinteractwithparticu-
how an individual summarizes their experience over a lar neurotransmitter systems, neurochemicals, hormones,
period of time, the latter refers to experiential reports and other signaling proteins. Despite a wealth of positive
made ‘in the moment’. Whereas these constructs would findings, translationstotreatmentremain elusive.Moving
theoretically be expected to correlate highly, growing evi- forward,theintegrationofthese variousmethodsthrough
dencesuggeststhattheyareonlymoderatelycorrelatedat the use of circuit-based analysis willbe critical for the de-
best[157,159].Thisisinpartduetothepresenceofwell- velopment of a biologically-based nosology and personal-
known retrospective biases that reflect a heightened izedmedicineinpsychiatry.
sensitivity to recency or maximum intensity of emo-
tional experiences (so-called ‘peak and end’ effects)
Abbreviations
[160]. In disorders like schizophrenia, such retrospect- 5-HT:Serotonin;ACC:anteriorcingulatecortex;BOLD:blood-oxygen
ive biases can result in almost completely uncorrelated level-dependent;CNS:centralnervoussystem;DA:dopamine;
DAT:dopaminetransporter;DBS:deep-brainstimulation;dlPFC:dorsolateral
findings of emotional response across retrospective and
prefrontalcortex;Glu:glutamate;GABA:gamma-aminobutyricacid;HPA
in-the-momentreports[161]. axis:hypothalamic-pituitary-adrenalaxis;IFN:interferon;MDD:major
This renders significantly liable the common practice depressivedisorder;MID:monetaryincentivedelay;mPFC:medialprefrontal
cortex;MRI:magneticresonanceimaging;MRS:magneticresonance
of regressing symptom severity measures (for example,
spectroscopy;NE:norepinephrine;NET:norepinephrinetransporter;
the BDI-II) against imaging data [162], as both the inde- OFC:orbitofrontalcortex;PET:positronemissiontomography;PFC:prefrontal
pendent and dependent measures likely reflect a com- cortex;ROI:regionofinterest;SPECT:singlephotonemissioncomputed
plex mix of ‘believing self’ and ‘experiencing self’. For tomography;SERT:serotonintransporter;TMS:transcranialmagnetic
stimulation;TSH:thyroidstimulatinghormone;VBM:voxel-based
example, ifsomeone hasreported severe depression over morphometry;vlPFC:ventrolateralprefrontalcortex.