Major depressive disorder (MDD) is a serious disease that poses major risks to human physical and mental health and will become the world's most common and economically burdensome illness by 2030 (Mathers and Loncar, 2006). Although there are some effective antidepressant treatments (Girardi et al., 2009; Murrough et al., 2013), the etiology and pathophysiology of MDD remain obscure. Depression is heritable (~35%) (Sullivan et al., 2000); however, the neural mechanisms of how risk genetic loci contribute to depression are still unclear. Accumulating evidence implicates the dopaminergic system in the pathophysiology of depression and antidepressant efficacy (Papakostas, 2006; Nelson et al., 2012; Dell'Osso et al., 2013). This suggests that depression is at least partly due to a decrease in extracellular dopamine (DA) levels (Lambert et al., 2000; Leggio et al., 2013). Thus, DA-related genetic variants that modulate endogenous dopaminergic neurotransmission might significantly contribute to depression pathophysiology. Although the results are not completely consistent, several studies have found that several DA-related genetic variants including catechol-O-methyltransferase (COMT), monoamine oxidase A (MAOA), DA D2 receptor (DRD2), and DA transporter gene (DAT) are closely associated with depression and antidepressant responses (Baune et al., 2008; Dong et al., 2009; Opmeer et al., 2010; Xu et al., 2011; Dannlowski et al., 2013). Specifically, Val alleles of the COMT gene polymorphism (rs4680) substantially increase COMT activity, improve DA metabolism and transmission, decrease prefrontal extracellular DA, and are associated with better processing of aversive stimuli (Stein et al., 2006; Winterer et al., 2006; Wu et al., 2012; Antypa et al., 2013). The G allele of the MAOA gene polymorphism (rs6323) encodes the higher activity form of the enzyme and thereby decreases dopaminergic tone, and is significantly associated with a poor antidepressant response in MDD patients (Shih et al., 1999; Ducci et al., 2006; Leuchter et al., 2009; Xu et al., 2011). Interestingly, the DRD2 and DRD3 genes are predominantly involved in the ventral striatum. For example, the C allele of the DRD2 gene polymorphism (rs6277) is associated with low striatal DRD2 availability and affects receptor affinity, with higher levels of depressive rumination in depressed individuals (Hirvonen et al., 2004, 2009; Whitmer and Gotlib, 2012). DRD3 is a DA receptor with distinct high distribution in the limbic system, especially the nucleus accumbens (NAc). Importantly, the Gly variant of the DRD3 gene polymorphism (rs6280) has 4–5 times greater affinity for DA compared to the Ser variant, and subjects with this form exhibit increased striatal reward-related DA release during a gambling task (Jeanneteau et al., 2006; Savitz et al., 2013). Collectively, these findings suggest that the multiple genetic effects of the DA pathway play a critical role in the pathophysiology of depression.

Depression is likely influenced by the joint influences of multiple genes operating within specific biological pathways (Woody and Gibb, 2015). Recently, the multilocus genetic profile score (MGPS) approach was employed to explore polygenic effects on psychiatric disorders (Bogdan et al., 2017). The MGPS approach summates several gene polymorphisms variants in a given functional neural system to obtain a composite of relative signaling (Nikolova et al., 2011; Bogdan et al., 2016). Researchers have begun to investigate the cumulative genetic effects of the DA pathway on depression (Stice et al., 2012; Pearson-Fuhrhop et al., 2014) by constructing MGPS from several gene variants involved in DA metabolism and signaling. These studies have revealed that lower MGPS reflecting DA signaling capacity predicts more severe depressive symptoms in healthy adults and depression patients (Pearson-Fuhrhop et al., 2014). On the other hand, neuroimaging genetics studies have demonstrated that the multilocus genetic composite of the DA pathway, but not a single single-nucleotide polymorphism (SNP) genotype, can predict reward circuitry activity during reward paradigm tasks (Nikolova et al., 2011; Stice et al., 2012). However, the cumulative effect of dopaminergic pathway genes on the intrinsic reward network in MDD patients is still unclear.

In the present study, we aimed to investigate the neural mechanism underlying the polygenic effects of the DA pathway on the reward network [constructed using the NAc functional connectivity (NAFC) network] in MDD patients. We hypothesized that the DA-MGPS would be significantly associated with the reward network and depressive symptoms and that dysfunctional connectivity within the reward network would mediate the association between DA-MGPS and depression in MDD patients.

Our findings demonstrate the cumulative effects of multiple gene polymorphisms in the dopaminergic pathway on both behavior and the intrinsic reward network in MDD patients. First, the DA-MGPS is associated with anxiety/somatization symptoms in MDD patients. Second, DA-MGPS widely influences the NAFC network and accounts for 7.5–19.7% of all variability of the NAFC network. Third, depression and DA-MGPS synergistically contribute to the dysfunctional reward network, especially in fronto-striatal circuits. Fourth, the relationship between the MGPS and HAMD anxiety/somatization factor is mediated by NAFC in MDD patients, suggesting that reward-related intrinsic connectivity could modulate the DA-MGPS effect on depression traits. These findings provide novel insight into the polygenic cumulative role of the dopaminergic pathway in the reward network in depression patients.

First, the DA-MGPS was significantly negatively correlated with the anxiety/somatization factor in the MDD group, suggesting that a lower score might predict higher levels of anxious depression. Individuals with a DA-MAGPS above seven in the HAMD anxiety/somatization factor would be considered as anxious depression patients (Fava et al., 2008). Moreover, 46–81% of MDD patients present anxious depression, subsequently leading to melancholic features, more severe symptoms, greater risk of suicide, and lower antidepressant response (Maurizio Fava et al., 2006, 2008; Lin et al., 2014). Although a previous study found an association between DA genetic risk scores and the HAMD total score in 1,267 depressed subjects (Pearson-Fuhrhop et al., 2014), it is not clear whether multiple genes in the dopaminergic pathway contribute to the underlying neuropathology of anxious depression. The present study provides indirect evidence that the DA genetic profile is involved in anxious depression.

Second, consistent with previous task-related fMRI studies, NAFC was decreased in MDD patients, and disruption was primarily located in the fronto-striatal pathway (Tremblay et al., 2005; Pizzagalli et al., 2009). These brain regions are closely linked to the constructs of reward valuation and anticipation, as well as motivation (Der-Avakian and Markou, 2012; Berridge and Kringelbach, 2015). In addition, the caudate, mOFC, and adjacent rACC are important hubs for reward circuits that contribute to processing reward valuation, subjective pleasure, and reward anticipation (Kringelbach et al., 2003; Georgiadis and Kringelbach, 2012). Moreover, the dlPFC, vlPFC, and dACC are involved in emotion regulation, evaluation of the self, and executive function (Northoff and Bermpohl, 2004; Smith et al., 2009). Therefore, for the important role of the fronto-striatal pathway in reward and cognitive control process, decreased NAFC in this pathway indicates that MDD patients might suffer from an attenuated reward experience and decreased cognitive control during the resting state.

Third, it is not surprising that the DA-MGPS was associated with a widespread effect on the intrinsic NAFC network including the frontal, temporal, parietal, occipital, and insular lobes. Specially, the DA-MGPS is positively correlated with functional connectivity between the NAc and hypothalamus. Both the hypothalamus and NAc play core roles in reward and motivation, especially for food “liking” and “wanting” (Aston-Jones et al., 2010), and their interaction may contribute to maintaining homeostasis and mediating affect and motivation (Castro et al., 2015). Though the sample is relatively small, we also found significant variability (7.5–19.6%) in the reward network between individuals that was explained by the DA-MGPS. These findings support the important and extensive role of DA in the reward network.

Fourth, although there was no difference in DA-MGPS between two groups, opposite patterns of MGPS effect on reward network were observed between groups. The different patterns between the CN and MDD groups might indicate that the DA-MGPS effect on reward network would be non-linear. Several studies reported an inverted U-shaped relationship between DA signaling and fronto-striatal brain activity and cognitive performance in healthy participants and subjects with Parkinson's disease (Bertolino et al., 2009; Cools and D'Esposito, 2011; Fallon et al., 2013). Our recent published work also revealed a non-linear modulation of the interaction between the COMT gene and depression on brain function (Gong et al., 2017a). Interestingly, the interaction regions identified in the present study included the mOFC, rACC, and putamen, which functionally encompass the fronto-striatal pathway, and these are implicated in reward anticipation, valuation, motivation, and attainment (Dillon et al., 2014). Our findings verify and expand the notion of a non-linear effect of DA signaling on the fronto-striatal pathway at a DA polygene level. According to the consistency of dysfunctional NAFC and the non-linear effect of DA-MGPS on the fronto-striatal pathway, we hypothesize that dysfunction in the reward network and a lower DA genetic profile make subjects vulnerable to depression.

Importantly, mediation analysis showed that the link between the DA-MGPS and HAMD-a score was mediated through functional connectivity between the NAc and putamen. This extends our observation of a link among the DA genetic profile, reward circuit activity, and depression traits in healthy adults and MDD patients (Felten et al., 2011; Nikolova et al., 2011; Stice et al., 2012; Pearson-Fuhrhop et al., 2014), and implies that NAc-putamen connectivity may play a substantial role in DA gene effects on anxious depression in MDD patients. The NAc and putamen are substructures of the striatum and play pivotal roles in regulating dopamine, thus mediating the responses induced by both stress and reward (Cabib and Puglisi-Allegra, 2012; Foti et al., 2014; Stringaris et al., 2015). A recent study reported that remitted MDD patients exhibited potentiated striatal reactivity during a psychological stress task (Admon et al., 2015). Another study found that the robust reward responsiveness in the NAc may protect against the depressogenic effects of stress (Nikolova et al., 2012). These previous findings leave open the possibility that the observed relationship between the DA genetic profile and anxious depression is driven by weaker striatal functional coupling due to chronic psychological stress in MDD patients. From this view, our findings suggest that local striatal functional connectivity could be a predictor of anxious depression in individuals with lower DA-MGPS.

The present study also has several limitations. First, because participant genetic information came from a large research sample that assayed only 14,398 SNPs in the exome region, several DA genetic variants of interest were not genotyped, such as DAT1 VNTR and DRD2 Taq1A (Brody et al., 2006). Additional DA pathway genes should be covered to generalize our findings. Second, we did not perform neuropsychological tests to measure reward or motivation. Future studies should include a detailed reward estimate test combining specific tasks to further clarify the relationship between DA genetic profile and reward processing in MDD patients. Third, the small sample size and mixed sample (both inpatients and outpatients, first onset and recurrent) of this study limit the generalizability of the current result. Lastly, the method of MGPS was often used and averaged each SNP equally and additively to influence DA signaling and brain function. A more effective approach is necessary to provide a polygenic risk score (Bogdan et al., 2017). However, we still believe that the current study provides a useful starting point for multilocus genetic imaging investigations for MDD.

In summary, our findings show that the DA-MGPS specifically contributes to the reward network in patients with MDD. In addition, NAc-putamen functional connectivity mediates the relationship between the DA genetic profile and anxious depression. These findings provide new insight to guide further exploration into the polygenic effects on depression risk variants and may help reveal the complex gene–brain–behavior relationship.

LG and CX: Conceived the study, designed and conducted the experiments, analyzed the data, and wrote the manuscript; CH, QY, YYY, and HW: Conducted the experiments, analyzed the data, and wrote the manuscript; HSZ, HXZ, and LL: Analyzed the data and wrote the manuscript; ZZ, YGY, and FB: Conceived the study, designed the experiments, analyzed the data, supervised the project, and wrote the manuscript.