In this longitudinal study, we present the first investigation of microstructural changes in the hippocampus following ECT in severely depressed patients using neurite imaging. Our aim was to test the hypothesis and identify in vivo biomarkers of ECT-induced hippocampal neurogenesis. In addition to noting a volumetric increase, we observed an increase in neurite density and dispersion, coupled with a reduction in the isotropic fraction (i.e., free water), suggesting enhanced synaptic branching in the hippocampus.
Our findings of hippocampus volume increase after ECT are in line with recent meta-analyses (10,31–33) and mega-analyses from the Global ECT-MRI Research Collaboration (GEMRIC) (11,34) reporting a significant increase in hippocampus volume within one or two weeks after ECT.
In addition to macroscopic volume changes, we report microstructural changes following ECT, which include a reduction in generalized fractional anisotropy, mean diffusivity, and axial diffusivity, as well as changes in the free water compartment. We also observed an increase in neurite density and orientation dispersion. Previous DTI studies have reported a decrease in MD in the bilateral hippocampi following ECT (35–37). This finding supports neuroplastic changes rather than postictal edema, which typically presents with an increase in MD due to extracellular fluid augmentation. In this study, we observed a significant bilateral decrease in AD and left MD, consistent with the previously reported DTI findings (35–37). We also detected an early ECT-induced generalized FA decrease bilaterally, in line with Jorgensen et al. (35), who found a transient FA decrease after ECT. This generalized FA decrease could be related to the expansion of axonal and dendritic processes of pyramidal neurons, as described in the postmortem ferret brain (38). Of note, we did not detect an association between hippocampal MD and FA decreases and a favorable response to ECT , as observed in earlier studies (35–37).
We used the NODDI model for the first time to analyze hippocampal ECT-induced changes at the sub-voxel level. Our specific aim was to determine whether the changes observed in previous DTI studies could be attributed to hippocampal neurogenesis rather than cellular swelling, such as intracellular edema observed in postictal states or acute ischemic stroke (39). Here, we observed an increase in neurite density in the right hippocampus from baseline to post-ECT, an initial bilateral decrease in Fiso, and a consistent increase in ODI across all three time points. In cases of cellular swelling, an increase in neurite density and a decrease in Fiso are also expected due to a shift of water from the extracellular to the intracellular space. However, no change in dendritic orientation is anticipated. The hypothesis of cellular swelling after ECT appears unlikely given the ODI increase we observed in both hippocampi, instead supporting enhanced dendritic complexity and synaptogenesis, as described in neonates' grey matter (40), aligning with our initial hypothesis.
Remarkably, the analyses of the hippocampal subfields revealed that some changes were specific to the dentate gyrus. Indeed, the NDI increase and Fiso decrease were observed in the dentate gyrus but not in the hippocampal tail, consistent with neurogenesis occurring in the dentate gyrus (13). The FA decrease and ODI increase in the entire hippocampus, including the tail, could reflect synaptogenesis, potentially occurring across all hippocampal subfields (41).
The time-course of these microstructural changes after ECT raises the question: are we truly observing active new cell formation? In our study, changes in Fiso and ODI were observed at V2 (two weeks after the start of ECT), preceding the changes in neurite density, which occurred at V3 (on average 70 days after baseline).Spontaneous hippocampal neurogenesis takes place over a period of several months, during which neural stem cells mature into progenitor cells, immature granule cells that migrate in the dentate granule cell layer, to become mature and finally establish functional connections (42). The recent findings of a post-mortem study (13) contribute to the discussion. While they observed an increase in doublecortin, a dendritic growth marker (43) in subjects treated with ECT, they found no differences in Ki-67, a marker of cell proliferation. If the short duration of Ki-67 expression, compared to the delay between ECT and the time of death, could be an explanation, the authors also proposed that it may indicate that neuroplasticity, rather than active new cell formation, increased after ECT in their sample. The literature presents two potential theories that could explain the pathophysiology of such hippocampal neuroplasticity. Firstly, ECT might enable existing neurons in the dentate gyrus to undergo ‘dematuration’, i.e. mature cells could become activated and make new dendritic connections (44). An alternative hypothesis could be that ECT promotes the dendritic maturation of immature granule cells stored in the dentate granule cell layer, bypassing the previous maturation steps and thereby accelerating the spontaneous neurogenesis process (45). Interestingly, in animal studies, electroconvulsive stimulation has been shown to induce dendritic spine maturation of new granular cells, increase the dendritic spine density of mature granular cells (46) and enhance synaptogenesis and dendritic branching in the Cornu Ammonis 1 region of the hippocampus (7,47) in rats.
Additional analyses could be conducted and correlated to these microstructural changes, including assessments of structural (48) and functional (49) connectivity of the hippocampus during ECT. Such analyses could help determine whether the observed regional alterations are associated with connectivity changes, providing additional support for the hypothesis of hippocampal neuroplasticity.
Importantly, we observed a significantly larger volume increase in the right hippocampus after 5 ECT sessions among ECT-responders compared to ECT-nonresponders, a finding not previously reported in the literature (34,50). These findings may enable the early prediction of treatment efficacy and expedite discontinuation of ECT in non-responsive patients. These findings were possibly facilitated by the fixed V2 time-point, scheduled after 5 ECT sessions, enabling a more reliable comparison between groups. However, unlike a study using a 7T MR scanner (51), volumetric variations of the dentate gyrus were not correlated with the clinical response in our study, possibly attributable to the thinness of this region and the lower spatial resolution of our 3T MR acquisitions. Microstructural metrics also failed to correlate to the clinical response. One explanation could be that diffusion exhibit a poorer spatial resolution than 3D T1-w images (2 mm versus 1 mm isotropic voxels) and is more prone to partial volume effects. While achieving high-resolution brain diffusion MRI has historically been challenging due to prolonged acquisition times to avoid signal loss, recent advancements in diffusion MRI suggest that achieving submillimeter resolution may become clinically feasible in the near future. This potential advancement could overcome this limitation in future studies.
Our study has several limitations. First, only 27 out of 43 patients completed all three MRI evaluations, while 7 patients only had a baseline MRI. To address this issue, we used linear mixed-effects models, which well adapted to missing data in longitudinal designs. Time was modeled as a categorical variable with three time points; however, a more effective approach could have been to utilize trajectory techniques to account for the variability in the intervals between baseline and the V3 time point (52,53). Third, our data only capture changes occurring during the first three months after initiating ECT. Several studies report a normalization of the hippocampus volume between 6 to 12 months after ECT (35,54–56), suggesting a transient effect of ECT. Due to the study design, we cannot provide evidence of long-term persistence of the observed changes. However, this aspect was not critical in relation to our primary goal, which was to elucidate the pathophysiology of the hippocampal volume increase observed in the weeks following ECT.