Major depressive disorder (MDD) is among the most prevalent and debilitating of all psychiatric illnesses, affecting nearly 20% of the US population, or more than 30 million adults, at some point in their lives 1. MDD is characterized most commonly by sadness and/or by a marked decrease in the ability to experience reward or pleasure (that is, anhedonia). Individuals diagnosed with MDD also often exhibit psychomotor retardation and deficits in executive function, typically reflected by difficulties concentrating, making decisions, and terminating the processing of negative material 2. Neuroanatomical models of depression, however, have yet to fully account for the range of symptom domains that characterize this disorder.

Investigators examining neural functioning in depressed individuals have frequently reported anomalies in the structure and function of the frontal lobes. For example, altered regional blood flow 3 and glucose metabolism 4 in the prefrontal cortex (PFC), a region frequently implicated in affect regulation and in a number of cognitive and motor processes, have been linked to both psychomotor retardation and anhedonia in depressed individuals. Interestingly, these studies also implicate anomalous functioning of the striatum in MDD, a revealing co-occurrence given the neuroanatomical linkages between the PFC and striatum. The striatum, composed within each hemisphere of the caudate nucleus, putamen, and core of the nucleus accumbens, comprises the primary input structure to the basal ganglia circuitry, receiving extensive projections from the frontal cortex. Output neurons from the basal ganglia send projections to midbrain structures and the thalamus, which completes a cortico-striatal-pallidal-thalamic (CSPT) circuit by returning projections to the frontal cortices. Whereas the PFC is generally thought to serve central control or goal representation functions, investigators have argued that CSPT circuitry aids in the filtering and focusing of cortical input, selecting among potential cognitive and behavioral representations 56. This formulation has provided a foundation for models of behavior selection 6 and working memory updating 7, and supports the hypothesis that abnormal functioning of frontostriatal pathways contributes to the anhedonic, ruminative, and psychomotor dysfunctions characteristic of MDD.

Importantly, topographical organization within CSPT circuitry is largely preserved, such that corticostriatal neurons from the motor and premotor cortices project most densely to the caudal putamen ('motor' division), whereas input from dorsolateral PFC (DLPFC) innervates the dorsal caudate ('executive' division), and projections from the orbitofrontal cortex (OFC) and anterior cingulate cortex (ACC) tend to populate the ventral striatum ('limbic' or 'affective' division). In general, these divisions along the ventral-dorsal axis of the striatum are posited to be relatively segregated, thereby differentially supporting affective, cognitive, and motor processing 8. Given that the difficulties experienced in MDD span these domains, it is possible that dysfunction in this system is attributable to global perturbations in microstructure or macrostructure, or in neurotransmission, rather than to abnormalities restricted to a particular CSPT division. Alternatively, the ascending flow of information through the parallel CSPT circuits 9 may enable a circumscribed abnormality in CSPT function to propagate to other subdivisions (for example, disrupting the translation of reward motivation into goal-directed motor behavior). Nonetheless, despite hypotheses that CSPT circuitry, and the frontostriatal connections in particular, may be compromised in MDD 101112, little empirical work has explicitly examined CSPT circuit function in this disorder.

Over the last decade, there has been a surge in the development and use of methods to investigate network-level brain function 13. These methods have recently been used to map the correlated activity, or functional connectivity, of normally and abnormally functioning corticostriatal networks in human subjects at rest 1415. Using one such method, investigators have documented attenuated functional connectivity between the ACC and a region of interest (ROI) spanning parts of the putamen and globus pallidus in depressed individuals 16. This finding suggests anomalous frontostriatal connectivity in MDD, but does not permit us to draw conclusions about the subdivisions involved. In the present study, we examined patterns of frontostriatal connectivity in adults diagnosed with MDD, focusing on the dorsal and ventral striatal subdivisions posited to differentially support motor (dorsal putamen), executive (dorsal caudate), and affective (ventral caudate and putamen) functions, all of which have been found to be compromised in MDD. Given previous findings that mood 17 and reward-related behaviors 18 vary with the extent of connectivity between the ventral striatum (VS) and the subgenual anterior cingulate cortex (sACC), we hypothesized that, compared with non-depressed individuals, depressed participants would exhibit abnormal connectivity between the frontal cortex and the striatum, most markedly in this ventral, limbic circuit. Further, given the frequent presentation of cognitive and psychomotor dysfunction in MDD, we hypothesized that depressed persons would also exhibit attenuated frontostriatal connectivity within the executive and motor networks.

The present study was designed to examine alterations in the functional connectivity of frontostriatal networks in MDD. Patterns of connectivity in both control and depressed participants were consistent with the results of previous studies 14; importantly, however, striking differences between the depressed and non-depressed participants were detected in networks that have been implicated in affective and motivational processing 28. In particular, we found that, compared with healthy controls, depressed individuals exhibited attenuated connectivity between seed regions approximating the ventral striatum/nucleus accumbens and ventral rostral putamen, and vmPFC and sACC targets.

It is important to note that these prefrontal target regions fall within the bounds of regions known to have anatomical connectivity with the VS 89. As noted earlier, the pattern of projections from the cortex to the striatum maintains much of its anatomical distinction, such that corticostriatal neurons from the motor and premotor cortices project most densely to the caudal putamen, whereas input from the OFC and vmPFC tends to populate the anteroventral caudate and ventromedial putamen. Indeed, this ventral corticostriatal network is posited to support reward-related processing and affect 8. Consistent with this framework, several of the ventromedial PFC structures found in this study to correlate differentially with striatal activity in participants with MDD and CTL subjects have previously been implicated in depression and in the processes affected by this disorder 2930.

Because of its unique interconnectivity with subcortical and limbic structures, the sACC has been hypothesized to be involved in both affective processes and visceromotor integration 3132, and has specifically been associated with the induction of negative mood 33. Functional 29, metabolic 34, and volumetric 3435 anomalies involving the sACC are often reported in studies of MDD. Interestingly, Harrison and colleagues 17 reported a relation between cytokine-induced decrements in mood and the degree of connectivity between the nucleus accumbens and the sACC. Considered in light of our current findings, this body of work suggests that multiple forms of brain pathology, including aberrant structure or functionality of either striatal or frontal regions, ultimately produce a similar, deleterious effect on network-level function. Thus, future research should examine the extent to which volumetric and metabolic anomalies contribute to the pathophysiology of MDD by altering patterns of mesolimbic or corticostriatal connectivity.

Situated just rostral to the sACC, the vmPFC is associated with the production 36 and flexible shifting of emotional states 37 and behavior, as well as with reward processing 38 and reinforcement learning 39. Interestingly, the integrity of fibers linking the vmPFC, as well as the medial OFC, with the VS appears to mediate individual differences in reward-related behavior. For example, using diffusion tractography, Cohen and colleagues found a positive correlation between the integrity of this tract and reward dependence, a personality trait characterized by learning from reward signals and persistence in repeating actions associated with rewards 18. Importantly, impaired reinforcement or reward learning is characteristic of MDD, and has been linked to anhedonic symptoms in depressed individuals 30; the current finding of diminished functional connectivity between the vmPFC and VS in MDD suggests a potential neural correlate of this deficiency. Future work should examine the relation between differences in frontostriatal connectivity and compromised motivational and reward-learning processes in anhedonic depression.

We also observed attenuated connectivity between the dorsal striatum and medial PFC in depressed, relative to control, participants. In contrast to the ventral corticostriatal 'limbic' networks, those networks comprising more dorsal (and caudal) regions of the striatum, along with their respective inputs from the frontal cortices, are thought to support executive function and motor-planning processes, including the initiation of behavior 40. In addition to detailing associations between MDD and compromised executive and motor function, studies have linked depression to difficulties in the initiation of motor movements 41. For example, Sabbe et al. report that the highest degree of motor slowing in MDD is observed during the early stages of a movement, suggesting a particular difficulty in initiating behaviors 42. Subsequent work has demonstrated an increased reliance on external cues during the performance of sequential movements in melancholic depressed persons, reminiscent of the deficits of initiative seen in Parkinson's disease (PD) 43. While preliminary analysis of the relation between psychomotor retardation and dorsal striatal connectivity suggests a plausible mechanism by which psychomotor function is compromised in MDD, future research with larger sample sizes and more detailed assessments of symptom severity will help clarify the extent and specificity of this relation.

We found an unanticipated increase in connectivity between the dorsal caudate and the middle frontal gyrus in MDD. Given the association between dorsal corticostriatal macrocircuitry and both the filtering and manipulation of content in working memory 4445, it is possible that this result reflects the inability to update the contents of working memory 46 and the maintenance of negative information in MDD 47. Preliminary support for this formulation comes from our observation that, for depressed participants, increasing tendency to ruminate predicted increasing DC-DLPFC connectivity during rest, though this correlation did not reach statistical significance in the current sample. It is possible that the observed increase in functional connectivity reflects a compensatory mechanism; if output from the basal ganglia circuitry to the frontal cortex (via the thalamus) is compromised in individuals with depression, cortical input to the basal ganglia may be ramped up in an attempt to drive an underactive basal ganglia-thalamus complex.

In addition to differences between depressed and non-depressed participants in the degree or extent of connectivity within corticostriatal networks, alterations observed in this study in the relation between these networks and other known brain systems are noteworthy. Relative to healthy individuals, participants with MDD exhibited enhanced connectivity between the VS and the PCC, a cortical-midline region central to the default mode network (DMN) 48. The DMN, composed of medial frontal, lateral and medial parietal, as well as temporal lobe regions, has been implicated broadly in self-referential processing 49. Activity across the various nodes of this network is reduced during periods of non-self-referential processing, such as those that occur during the performance of demanding cognitive tasks 4849. Recently, Kelly et al. demonstrated that administering L-3,4-dihydroxyphenylalanine (L-DOPA), a precursor to dopamine (DA), to healthy participants diminishes functional connectivity between the PCC and caudate regions 24; these authors propose that the striatum plays a role in the suppression of the DMN under conditions requiring a shift from self-directed, to externally-directed, processing. In contrast, our finding indicates a reduction in the anticorrelation of these two networks in MDD, potentially reflecting the difficulty exhibited by depressed persons shifting attention away from self-referential processing and reducing DMN activity in the face of external demands 50.

Although we did not quantify dopaminergic factors in the present study, there is reason to believe that DA plays an important role in the modulation of these corticostriatal connections. Whereas augmentation of DA levels has been shown to increase functional connectivity between the VS and vmPFC 24, DA depletion has been found to attenuate the functional coupling between VS and both dorsal and ventral PFC 51. Similarly, connectivity analyses in PD patients have demonstrated attenuated connectivity between the ventromedial caudate (approximating our VS seed) and the vmPFC 52. The similarity between the connectivity patterns identified in both temporary and disease-related DA-depleted states and those identified in our depressed sample is intriguing, and argues for a more systematic investigation of the relations among DA tone, corticostriatal connectivity, and symptom presentation in MDD.

In closing, we note a number of limitations of this study. First, our sample was composed entirely of females, thereby limiting our ability to generalize the current results to male subjects. Second, our depressed sample was heterogeneous with respect to comorbid DSM-IV diagnoses and psychoactive medication status. Although statistical tests indicated that medication status [see Additional file 1] did not mediate the aberrant patterns of frontostriatal connectivity we found to characterize depressed individuals, we cannot rule out the possibility that certain drug classes might selectively affect these circuits. While previous studies have reported no change in resting state basal ganglia-ACC functional connectivity following selective serotonin reuptake inhibitor (SSRI) treatment 53 and only decreases in DC-frontal lobe connectivity following administration of L-DOPA to healthy controls 24, future research might examine directly the effects of dopaminergic agents on corticostriatal connectivity, particularly in the context of symptom remediation. Third, we may be underpowered in the current study to detect and precisely characterize the associations between clinical variables and frontostriatal abnormalities. It is also possible that certain of the documented abnormalities in frontostriatal connectivity reflect risk factors for the development of depression. In this case, we might not expect to identify correlations with disease or symptom severity. These possibilities highlight the importance of examining corticostriatal connectivity in a larger sample of MDD subjects, in those at elevated risk for the development of MDD, and in individuals who have recovered from the disorder. Finally, the present findings of altered frontostriatal connectivity in MDD may be attributable, in part, to structural changes within the striatum itself 545556. Although we assume that the seeds used in this study represent discrete functional divisions of the striatum, it is possible that the boundaries between such divisions are shifted in the brains of depressed persons as a result of structural alteration. Future work directly examining the consequences of gross anatomical, metabolic and/or functional changes on frontostriatal circuitry could advance our understanding of how regional changes in brain function alter system-level function in psychiatric disorders. Despite these limitations, however, the results of the present study provide important empirical support for the hypothesis that MDD is characterized, and even maintained, by compromised frontostriatal function.

The authors declare that they have no competing interests.

DJF and IHG designed the study, and DJF and JPH conducted the statistical analyses. DJF wrote the first draft of the manuscript. All authors contributed to, and approved, the final manuscript.