Ketamine and Attentional Bias Toward Emotional Faces: Dynamic Causal Modeling of Magnetoencephalographic Connectivity in Treatment-Resistant Depression

CONCLUSION

This study used MEG recordings collected while participants completed a dot probe task with emotional faces in tandem with DCM to probe ketamine's effects in individuals with TRD and healthy volunteers. The goal was to measure changes in effective (causal) connectivity within and between the early visual cortex, fusiform cortex, amygdala, and inferior frontal gyrus, in addition to changes in AMPA, GABA, and NMDA receptor time constants, following ketamine administration. We were particularly interested in ketamine's effects in the amygdala, a key region implicated in the pathophysiology of depressiondemonstrating upregulation to positive faces and downregulation to negative faces during an attentional dot probe task following ketamine administration. Clinically, we found significantly reduced depressive symptoms in our TRD sample post-ketamine, consistent with previous findings. Controlling for the period-specific baseline and the participant-average baseline, ketamine was found to result in a 5.37-point reduction in MADRS score in the TRD sample. Behaviorally, no differences in reaction time bias scores were observed on the task. However, accuracy differences were observed between the two groups, with TRD participants significantly more accurate than healthy volunteers during the task. In addition, session effects were noted with regard to accuracy rates, with the best performance occurring during the baseline session, followed by the ketamine and then placebo sessions. Importantly, post-hoc tests found significant differences in accuracy between the baseline and placebo sessions only. These findings suggest that healthy volunteers were less engaged in the task and therefore did not perform as well as the TRD participants. In addition, task repetition led to poorer performance, especially following placebo saline infusion, where participants were perhaps least motivated to perform well-during the scan procedures. We modeled induced gamma-band activity during the dot probe task, identifying a network of brain regions involved in the task. We also modeled regions showing an effect of infusion (ketamine vs. placebo) and found increased gamma Parametric empirical Bayes was used to identify the mixing of parameters that contributed to the effect of group. Note that the timing of data collection (6-9 h post-ketamine administration) occurred past the half-life of ketamine. Meaningful parameters were defined as those with a posterior probability (Pp) >95%. Ten parameters were found to significantly contribute to group effects. These included the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) time constant within the amygdala (Amy), gamma aminobutyric acid (GABA) time constants within the fusiform gyrus (Fusi) and inferior frontal gyrus (IFG), and N-methyl-D-aspartate (NMDA) time constants within the early visual cortex (EV) and IFG. In addition, the inhibitory self-connections on spiny stellate cells (ss) within the EV, Fusi, and IFG differed between groups, as did the inhibitory self-connection on superficial pyramidal cells (sp) within the Fusi, and the excitatory connections between sp and deep pyramidal cells (dp) in the Amy. Finally, the inhibitory self-connection on inhibitory interneurons (ii) in the EV showed a group effect, though not at our threshold. *Pp > 0.95. Parametric empirical Bayes was used to identify the mixing of parameters that contributed to the effect of drug. Meaningful parameters were defined as those with a probability (Pp) >95%. Nine parameters were found to significantly contribute to drug effects. These included the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) time constant within the early visual cortex (EV), gamma aminobutyric acid (GABA) time constants within the amygdala (Amy) and inferior frontal gyrus (IFG), and N-methyl-D-aspartate (NMDA) time constants within the EV, Amy, and IFG. In addition, the excitatory connections between superficial pyramidal cells (sp) and deep pyramidal cells (dp), and the inhibitory connections between inhibitory interneurons (ii) and sp differed following ketamine in the EV, as did excitatory connections between spiny stellate cells (ss) and ii in the IFG. Finally, the inhibitory self-connection on ss in the IFG showed a drug effect, though not at our threshold. *Pp > 0.95. power in the amygdala post-ketamine vs. placebo for both TRD participants and healthy volunteers. These findings are in keeping with preclinical studies suggesting increased cortical excitation following ketamine administration, due to NMDA inhibition reducing the activity of putative GABA interneurons. At a delayed rate, this increases the firing rate of pyramidal neurons due to enhanced AMPA throughput (15) that, in turn, leads to increased cortical excitation. Given that gamma power in the amygdala showed a drug-specific effect, with increased cortical excitation post-ketamine, this suggests that increased cortical excitation in this key emotional face processing region may be related to previous reports of normalization of emotional processing following drug administration. Notably, normalization of amygdalar activity post-ketamine was previously described in an fMRI study that included an attentional dot probe task with emotional faces in TRD participants, though this was not specifically examined in the present study. Two plausible models of message passing between the early visual cortex and the inferior frontal gyrus were subsequently Parametric empirical Bayes was used to identify the mixing of parameters that contributed to group by drug interactions. Meaningful parameters were defined as those with a probability (Pp) >95%. Nine parameters were found to significantly contribute to group by drug effects. These included gamma aminobutyric acid (GABA) time constants within the early visual cortex (EV) and N-methyl-D-aspartate (NMDA) time constants within the EV, fusiform cortex (Fusi), and amygdala (Amy). In addition, the inhibitory self-connections on spiny stellate cells (ss) and inhibitory interneurons (ii), as well as inhibitory connections between ii and ss in the EV showed group by drug interactions. Excitatory connections between deep pyramidal cells (dp) and ii in the Amy, in addition to inhibitory self-connections on superficial pyramidal cells (sp) in the inferior frontal gyrus (IFG) also showed group by drug interactions. * Pp > 0.95. fit. A model that included traditional feedforward processing along the ventral stream to the amygdala in tandem with feedforward connections from the early visual cortex to the inferior frontal gyrus provided the best model fits, in line with ideas that top-down predictions serve to constrain bottom-up signal propagation. All fitted parameters were subsequently extracted, and a Bayesian modeling extension of DCM was used to test for meaningful parameters contributing to the group effect, drug effect, and group by drug interactions. Here, we focus on discussing group by drug interactions, as these are identified parameters where ketamine had differential effects between TRD participants and healthy volunteers. Four modeled receptor time constants showed group by drug interactions, including the GABA and NMDA time constants in the early visual cortex and the NMDA time constants in the fusiform cortex and amygdala. In the early visual cortex, ketamine administration led to faster GABA and NMDA transmission estimates for TRD participants, while GABA transmission slowed for healthy volunteers postketamine. In the fusiform cortex, faster NMDA transmission followed ketamine administration for TRD participants, though the rate of transmission slowed for healthy volunteers postketamine. Interestingly, a slowing of NMDA transmission was observed in the amygdala post-ketamine for both TRD and healthy volunteers, though healthy volunteers had significantly faster NMDA transmission at baseline/placebo than TRD participants. As the amygdala ROI was identified based on the effect of infusion (ketamine vs. placebo), slowing of NMDA transmission within this region is clearly related to drug effects. Although no association was noted between NMDA transmission in the amygdala and antidepressant response within our sample, future studies should examine whether these changes in NMDA time constants are related to other clinical measures of mood changes following drug administration. In addition to changes in receptor time constants, group by drug interactions were found for modeled intrinsic connectivity within the early visual cortex, amygdala, and inferior frontal gyrus. In the early visual cortex, three intrinsic connection parameters showed group by drug changes in inhibitory drive. First, decreased GABAergic inhibitory drive on self-connections were found for both inhibitory interneurons and spiny stellate cells following ketamine in the TRD participants, while healthy volunteers demonstrated increased GABAergic inhibitory drive post-ketamine. These self-connections reflect gain or precision of different cell types, suggesting reductions in self-gain on inhibitory interneurons and spiny stellate cells following ketamine administration in the TRD group. Second, reduced inhibitory drive was observed on the intrinsic connection from inhibitory interneurons to spiny stellate cells in the early visual cortex in our TRD and healthy volunteers. Third, ketamine increased the excitatory drive from deep pyramidal cells to inhibitory interneurons in the amygdala in TRD participants, while healthy volunteers showed reduced excitatory drive for this connection post-ketamine. Finally, ketamine also reduced the inhibitory self-gain on superficial pyramidal cells in the inferior frontal gyrus in our healthy volunteers only. Interestingly, these findings all reflect changes in intrinsic connectivity that regulate or modulate inhibition locally. Within the amygdala in particular, increased excitatory drive onto inhibitory interneurons for TRD participants seems at odds with an increased state of excitability within this region; however, similar accounts of increased pyramidal-to-inhibitory interneuron drive have previously been reportedand are thought to reflect a link between increased pyramidal cell excitability locally and downstream effects of increased gamma power. Separately, we tested whether any meaningful parameters identified in our analysis of group effects, drug effects, or group by drug interactions were associated with antidepressant response in our TRD participants. We specifically examined changes in parameter estimates from the baseline to ketamine sessions (baseline minus ketamine) and correlated them with change in MADRS score from baseline to 230 min post-ketamine (the time point closest to the MEG recording session). Two parameters were found to be associated with antidepressant response, both in the early visual cortex. The first was the AMPA time constant in the early visual cortex, where faster AMPA transmission post-ketamine was associated with better antidepressant response. The second was inhibitory selfgain on spiny stellate cells in the early visual cortex, where larger self-inhibition on spiny stellate cells post-ketamine was associated with better antidepressant response. The findings of an association between AMPA transmission and antidepressant response are particularly striking because AMPA receptor throughput following NMDA receptor blockade) is thought to result in delayed increases in synaptic potentiation and synaptogenesis, key mechanisms associated with ketamine's antidepressant effects. Similar associations between AMPA receptor connectivity and antidepressant response were also previously reported in a time window overlapping with our MEG recordings. One important limitation of this study is that MEG recordings were not collected during or immediately following infusions, but rather 6-9 h following ketamine administration in order to avoid side effects while measuring therapeutic drug effects. Thus, we cannot comment on acute changes in modeled parameter estimates. However, studies of ketamine's acute effects in healthy volunteers suggest robust changes in both gamma powerand AMPA and NMDA receptor drive (30) during ketamine infusion. Future studies should explore ketamine's acute effects in TRD participants to better understand the mechanisms via which ketamine reduces depressive symptoms. Another limitation is that we set a liberal criteria of p < 0.05 uncorrected for determining whether modeled parameters were associated with antidepressant response. Though this increases the likelihood of false positives, previous findings have demonstrated associations between AMPA parameters and antidepressant response in TRD. In addition, our study included secondary analyses of data collected during a clinical trial of ketamine's mechanisms of actions, and we limited our sample to participants having baseline, post-ketamine, and postplacebo scan data. Additional work should include a larger sample of study participants to model effective connectivity during a task probing attentional bias toward emotional faces, in order to better characterize effective connectivity changes in regons of the emotion processing network following ketamine administration.