White Matter Microstructure Associated with the Antidepressant Effects of Deep Brain Stimulation in Treatment-Resistant Depression: A Review of Diffusion Tensor Imaging Studies

Treatment-resistant depression (TRD) is a severe disorder characterized by high relapse rates and decreased quality of life. An effective strategy in the management of TRD is deep brain stimulation (DBS), a technique consisting of the implantation of electrodes that receive a stimulation via a pacemaker-like stimulator into specific brain areas, detected through neuroimaging investigations, which include the subgenual cingulate cortex (sgCC), basal ganglia, and forebrain bundles. In this context, to improve our understanding of the mechanism underlying the antidepressant effects of DBS in TRD, we collected the results of diffusion tensor imaging (DTI) studies exploring how WM microstructure is associated with the therapeutic effects of DBS in TRD. A search on PubMed, Web of Science, and Scopus identified 11 investigations assessing WM microstructure in responders and non-responders to DBS. Altered WM microstructure, particularly in the sgCC, medial forebrain bundle, cingulum bundle, forceps minor, and uncinate fasciculus, was associated with the antidepressant effect of DBS in TRD. Overall, the results show that DBS targeting selective brain regions, including the sgCC, forebrain bundle, cingulum bundle, rectus gyrus, anterior limb of the internal capsule, forceps minor, and uncinate fasciculus, seem to be effective for the treatment of TRD.


Introduction
Major depressive disorder (MDD) is a disabling psychiatric disorder characterized by low mood, anhedonia, guilt, and neurovegetative symptoms, such as fatigue, loss of appetite or weight, and insomnia (American Psychiatric Association, DSM-5-TR). The 1-year prevalence of MDD is approximately 6%, although it varies considerably across countries [1], whereas its lifetime prevalence ranges from 10 to 20%; moreover, MDD is twice as common in women as in men [2,3]. Although for most patients MDD is episodic, the course of the illness is highly variable, and the duration, the number, and the pattern of episodes changes considerably between subjects [4]. Nonetheless, antidepressants represent the first line of treatment for MDD, with psychotherapy used as add-on treatment [5]. Notably, approximately 30% of patients with MDD fail to achieve remission with two or more adequate dose-duration antidepressants from different classes, and are Int. J. Mol. Sci. 2022, 23, 15379 2 of 13 referred to as having "treatment-resistant depression" (TRD) [6], which is characterized by lower acute remission rates and increased probability of relapse [7]. In addition, compared with treatment-responsive MDD, patients with TRD present more frequently a history of childhood adversities, negative life events, comorbid personality disorders, higher medical comorbidities, and increased risk of suicide [8,9]. Therefore, the identification of effective therapies is paramount for this disabling disorder. In this regard, several therapeutic strategies have been proposed for TRD, spanning from pharmacological treatments, including antipsychotics [10], mood stabilizers [11], and esketamine [12], to neurostimulation treatments, such as vagus nerve stimulation (VNS), electroconvulsive therapy (ECT), transcranial magnetic stimulation (TMS), and deep brain stimulation (DBS) [13]. Among these, in the last few years, DBS has evolved to become a promising and effective strategy in the management of TRD, considering its relatively favorable side effect profile and the vast array of potential stimulation parameters [14]. In DBS surgery, the electrodes are stereotactically implanted into specific brain areas, where stimulation is provided via a pacemaker-like stimulator that delivers an electrical stimulation [15]. Neuroimaging studies have provided evidence that MDD is characterized by structural and functional alterations in numerous brain areas, such as the cingulate cortex, dorsolateral and dorsomedial prefrontal cortex, striatum, thalamus, and insula [16][17][18]. These investigations have guided the choice of targets for DBS, which include the subgenual cingulate cortex (sgCC), ventral capsule/ventral striatum, nucleus accumbens, lateral habenula, inferior thalamic peduncle, medial forebrain bundles, and the bed nucleus of the stria terminalis [19]. The exact mechanism of DBS is still unknown, although it has been hypothesized that the suppression of gamma oscillations and facilitation of theta-gamma coupling by DBS is mediated by both sgCC activation of inhibitory circuits and an increase in plasticity in the frontal cortex [20], ultimately leading to the normalization of abnormal network connectivity [21]. Furthermore, in recent years, the results from diffusion tensor imaging (DTI) provided a useful tool not only in the definition of surgical targets for DBS [22,23], but also in the study of WM tracts in relation to treatment response [24] This is because DTI provides useful information on the orientation and integrity of white matter (WM) tracts in TRD [25], and allows us to investigate the changes in WM and connectivity between frontal and subcortical structures following DBS [26,27]. It is a relatively sensitive neuroimaging technique that assesses water molecule diffusion in extracellular space within the tissue, and has the potential to provide useful information regarding alterations in the microstructure of the WM [28]. Fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) are four indices that are derived based on the DTI approach for the purpose of quantifying the tensors that are included within each voxel, and can be indicative of microstructural markers [29,30]. FA is indicative of the degree of anisotropic movement of water. MD quantifies the average movement of water in three-dimensional directions. Moreover, RD and AD measure the movement of water molecules perpendicular to and alongside the nerve fibers, respectively. Lower microstructural integrity is associated with decreased FA and increased MD, RD, or AD. In TRD, reduced FA in the right anterior limb of the internal capsule, the body of the corpus callosum, bilateral external capsule, in the left limbic lobe uncus, left middle frontal gyrus and right cerebellum posterior lobe, bilateral inferior fronto-occipital fasciculus, bilateral inferior longitudinal fasciculus, bilateral superior longitudinal fasciculus, forceps major and forceps minor, and bilateral cingulum have been described in patients with TRD compared with healthy controls (HC) [31]. In this context, with this review, we aim at providing a comprehensive overview of WM microstructure associated with the therapeutic effects of DBS in TRD, with the final goal of clarifying the potential applications and advantages of the use of this technique in patients with TRD.

Demographic, Clinical, and Stimulation Characteristics of Included Studies
A total of 11 studies conducted between 2012 and 2020 investigating the association between presurgical WM measurements and response to DBS treatment were included in this review. The sample size ranged from 2 [32] to 24 [33] patients. Seven studies [22,24,26,27,[34][35][36] included only patients with MDD, while one study [37] assessed both patients with MDD and bipolar disorder (BD). Three studies [32,33,38] did not specify the type of underlying disorder. Patients were middle-aged men and women in most studies. When reported, the average age of onset and duration of current major depression episode ranged from 15.2 to 25.8 years old and 5.7 to 64.1 months. Treatment response was assessed using the Hamilton Depression Rating Scale (HDRS) [39], Montgomery-Åsberg Depression Rating Scale (MADRS) [40], Beck's Depression Inventory (BDI) [41], or the Global Assessment of Functioning (GAF) [42]. The majority of included studies applied DBS to the sgCC [32,[34][35][36][37] or medial forebrain bundle [22,27,33]. The remaining investigations applied DBS to the ventral capsule/ventral striatum [38], posterior rectus gyrus [26], and ventral anterior limb of the internal capsule [24]. The methods and key findings of the studies are detailed in Table 1. The most commonly reported WM tracts are depicted in Figure 1.

Investigating the Role of WM Alterations in DBS Therapeutic Effect Using Tractography
Except for one study [24], the others used 3T MRI machines with a b-value of 1000. They all applied tractography (Table 1). Overall, the investigations used four major approaches to conduct tractography: (a) seed-to-target probabilistic tractography between pre-defined target regions and patient-specific volume of activated tissue (VAT) [32,34,36], defined as the spatial extent of the electrical field delivered by the lead to the anatomical structures near the selected target [43,44]; (b) whole-brain probabilistic tractography with VAT as the seed [22,27,35,37,38]; (c) bundle-specific tractography, considering some specific WM tracts for the analysis [24,33,35]; (d) whole-brain unconstrained probabilistic tractography [26].

WM Tracts Mediating the Effects of Subgenual Cingulate Cortex DBS on TRD
Of the 11 included studies, 5 applied DBS to the sgCC [32,[34][35][36][37]. Overall, the results of these studies demonstrated that three primary fibers might mediate the effects of sgCC DBS on TRD, including cingulum bundle, forceps minor, and uncinate fasciculus. In particular, a pivotal study by Riva-Posse et al. (2014) reported that, compared to nonresponders, responders to DBS showed robust connectivity in bilateral and medial aspect of the uncinate fasciculus connecting VAT to the medial frontal cortex, the cingulum bundle connecting the VAT to the rostral and dorsal anterior cingulate cortex and middle cingulate cortex, as well as short descending midline fibers connecting the VAT to subcortical nuclei, including the nucleus accumbens, caudate, putamen, and anterior thalamus. Moreover, the patients that converted from non-responders at six-month follow-up to responders at 2-year follow-up gained connectivity from VAT to the above-mentioned regions through these fibers [37]. In accordance, Tsolaki et al. (2017) used seed-to-target probabilistic tractography to investigate the connectivity of VAT to bilateral medial prefrontal cortex via forceps minor and uncinate fasciculus, ipsilateral ventral striatum, and anterior cingulate cortex in two subjects with TRD. They found that VAT showed structural connectivity to all four target regions in the responder patient, while there was no such connectivity in the non-responder [32]. Conversely, the findings of Clark et al. (2020) did not agree with these two studies. Specifically, these authors reported that the whole-brain probabilistic tractography, with VAT as the seed, showed more limited connectivity in non-responders compared to responders from VAT to the medial frontal and to the temporal lobe. In addition, projections to the striatum were limited in both groups, but were more prominent in the non-responders. The authors created a common tract map for responders and non-responders, in which all voxels were shared by all group members. The cingulum bundle was not identified in the common tracts map of either responders or non-responder groups at either time point. However, the bundle-specific analysis showed different results at 6-month and 1-year follow-ups. In particular, responders had significantly more VAT overlap with the cingulum bundle than non-responders at 6 months, but significantly less VAT overlap with the forceps minor tract at 1 year. The profile of other tracts, including uncinate fasciculus and frontostriatal projections, did not differ between responders and non-responders [35]. In contrast to previous studies, Howell et al. (2019) investigated the WM pathways of interest surrounding the sgCC, including the forceps minor, cingulum bundle, uncinate fasciculus, and frontal pole, with a different approach. In a sample of six patients, they assessed whether axonal activation could predict time to response. Their findings showed that cingulum bundle and forceps minor were activated in most patients. Using regression analysis, they demonstrated that four regressors (forceps minor, left cingulum bundle, right cingulum bundle, and right frontal pole (FP)) explained 99% of the variation in the patient's time to stable response. However, activation of the right cingulum bundle alone explained 84% of the variance in time to stable response, and the addition of forceps minor, left cingulum bundle, and right FP did not contribute significant information to the linear model. Therefore, they concluded that increased predictive power from the inclusion of forceps minor, left cingulum bundle, and right FP in the linear model was likely because of overfitting [36]. Lastly, the study by Riva-Posse et al. (2019) assessed the role of WM connections in heart rate alterations. A significant relationship was found between the estimated structural connectivity of the left sgCC VAT to the middle cingulate cortex and the change in heart rate. Notably, it was found that the greater the structural connectivity between the sgCC VAT and the middle cingulate cortex (MCC), the more the heart rate increased for specific stimulation trials [34].

WM Tracts Mediating the Effects of Medial Forebrain Bundle DBS on TRD
Three studies applied DBS to the medial forebrain bundle [22,27,33]. Two studies by Fenoy et al. were conducted on an overlapping group of patients [22,27]. In the first study, contrary to non-responder patients, the three responder patients presented strong connectivity between the target location of the active electrode contact points and the medial prefrontal cortex [22]. Similar results were observed at a one-year follow-up [27]. In contrast,  found no significant relationships between microstructural measures of superolateral medial forebrain bundle and treatment response. Nonetheless, a positive relationship was reported between treatment response and enlargement of WM in left frontopolar superolateral medial forebrain bundle [33].

WM Tracts Mediating the Effects of DBS on Other Regions in Patients with TRD
The study by Lujan et al. (2012) assessed the axonal activation after DBS on the ventral anterior internal capsule and ventral striatum. They analyzed axonal activation by determining the extracellular voltages along each axon model (228,960 axon models for the 14 patient-specific DBS electric fields) by interpolating the patient-specific 3D electric fields onto each axon model compartment. Their results demonstrated that five pathways passing through the ventral anterior internal capsule and coursing lateral and medial to the ventral striatum, or dorsal and lateral to the nucleus accumbens, were common among 75% of responder patients, while one pathway adjacent to the ventromedial surface of the dorsal striatum was common among 75% of non-responders [38]. Conversely, Acolla et al. (2016) applied whole-brain unconstrained probabilistic tractography after DBS on the posterior rectus gyrus. They demonstrated that the probability of projections reaching the medial prefrontal cortex through the forceps minor and nucleus accumbens, or the anterior caudate through the uncinate fasciculus, was higher in the responder patient than in the average non-responders. Conversely, lower connectivity probability was found for tracts reaching the anterior cingulate cortex and middle cingulate cortex through the cingulum bundle in the responder patient [26]. Lastly, Liebrand et al. (2020) employed DBS on the ventral anterior limb of the internal capsule and reconstructed the superolateral medial forebrain bundle and anterior thalamic radiations for all patients using probabilistic tractography. There was a significant relationship between the average distance between tracts of interest and VAT and percentage change in HDRS [24].

Discussion
In this study, we reviewed all the available evidence to investigate whether patterns of WM microstructural integrity are associated with treatment response after DBS in TRD patients. In general, despite some inconsistencies in the results, it seems that DBS applied to different anatomical targets located in cortical, cingulate, and subcortical areas might stimulate different nodes of an interconnected network. This hypothesis is based on overall findings showing that treatment response after DBS with different targets, i.e., the cingulum bundle, forceps minor, uncinate fasciculus, medial forebrain bundle, and sgCC, is associated with the microstructural connectivity of similar WM tracts. However, future studies with graph theoretical analysis could better elucidate the network connectivity after DBS for TRD.
More in detail, the main targets of DBS in the included studies were the sgCC and the superolateral medial forebrain bundle. This is not surprising, as these regions play a key role in the pathophysiology of depression. Specifically, the sgCC is a crucial node with widespread connections to cortical and subcortical regions. Evidence from studies on primates has demonstrated that it is in reciprocal connection with the prefrontal, orbitofrontal, and dorsal cingulate cortices [45][46][47], as well as subcortical areas, such as the amygdala, thalamus, ventral striatum, and brainstem [46,48]. In addition, human studies using PET and fMRI have demonstrated the involvement of the sgCC in emotional processing in depressed individuals [49,50]. Conversely, the medial forebrain bundle is an important tract of the mesolimbic reward system, interconnecting the ventral tegmental area, hypothalamus, nucleus accumbens, and limbic lobe [51,52]. Notably, the ventral tegmental area is a critical hub within the dopaminergic reward system [53,54], and appears to modulate depressive symptoms [54]. Interestingly, it has been hypothesized that DBS effects on the medial forebrain bundle might lead to the recruitment of glutamatergic fibers from the medial prefrontal cortex to the ventral tegmental area, and it might therefore regulate ventral tegmental area dopaminergic activity [55]. In agreement, overall, the reviewed studies showed that stimulating various anatomical targets with DBS, including superolateral medial forebrain bundle, sgCC, and also the rectus gyrus, led to clinical improvement in patients with TRD. This suggests that aberrant electrical activity in these regions contributes to the development of TRD, and thus, acts as crucial hubs for the treatment of depression in patients with TRD.
In addition, from the reviewed studies emerged that other WM tracts seem to mediate the effects of DBS on TRD, including the cingulum bundle, forceps minor, and uncinate fasciculus, all structures that connect key neuroanatomical structures in cortical (prefrontal and temporal cortex), cingular, and subcortical (striatum) areas, and have been implicated in the pathophysiology of MDD. Specifically, the cingulum bundles are the main intra-hemispheric association pathways connecting intrahemispheric structures with the middle cingulate cortex [56]. Notably, simultaneous intraoperative behavioral and tractography assessments demonstrated that the activation of the cingulum bundle leads to changes in interoception [57], which leads to the modulation of the rostral anterior cingulate cortex and anterior insular cortex [58], two key regions communicating through the subgenual cingulum bundle [59].
Furthermore, the subgenual cingulum bundle is also associated with the dorsal anterior cingulate cortex [60,61], which is an important target for the treatment of MDD [62]. In this regard, previous studies have shown that lesion severity of the cingulum bundle is a prognostic predictor of poor treatment response in depression [63], and that the microstructural integrity of the cingulum bundle is associated with better treatment outcomes after DBS in TRD [36]. Noteworthy, also, is that the forceps minor, which is the biggest WM tract connecting the medial forebrain to other brain regions, has been implicated in the pathophysiology of affective disorders [64], with previous evidence also showing that targeting this region has proven to be effective in reducing symptoms in TRD subjects [37]. Similarly, the uncinate fasciculus, which is the key fiber connecting the orbitofrontal to the temporal cortex [65], is also engaged in the processing and regulation of emotion [66,67]. Interestingly, DTI studies have described an impaired microstructural integrity of uncinate fasciculus in patients with MDD [68], indicating the role of this structure in the development of depression. However, in contrast to the cingulum bundle and forceps minor, the evidence regarding the role of the uncinate fasciculus in mediating the effects of DBS on depression is still limited.
Finally, the results of DBS-DTI studies suggest that treatment resistance was generally associated with less WM connectivity among crucial brain regions, while responder patients showed a more interconnected network. This evidence seems to indicate that a connected WM architecture might be needed for an optimal response after DBS for TRD.
Importantly, the results of the reviewed studies should be considered in light of several limitations, which may influence the generalizability of our conclusions. First, in 7 studies, the sample size was <10, and in the remaining 4, it ranged from 14 to 24. Sample size is a limitation in most DBS studies in psychiatric patients, therefore, care must be taken when interpretating the results. Second, there was a significant heterogeneity in terms of methodologies (e.g., target regions of DBS and DBS protocols) and populations (e.g., diagnostic criteria for TRD, mixed diagnosis, heterogeneous medications regimens), which in turn might decrease the generalizability of the findings. Indeed, specifically for medications, previous studies have shown that the use of antidepressants is associated with altered WM microstructural integrity [69]. Third, the retrospective nature of several of the included studies represents an intrinsic limitation. Fourth, since DTI is an indirect measure of the structural integrity of WM, to date, it is not entirely clear how perturbations in DTI parameters reflect in the axonal structure of WM fiber bundles. Therefore, DTI results should be interpreted with caution.

Article Selection
A systematic search on PubMed, Scopus, and Web of Science was conducted to identify all relevant studies published before January 2022 with no language restrictions. The keywords included the following terms: "treatment resistant depression" OR "resistant depression" OR "refractory depression" AND "Deep Brain Stimulation" OR DBS AND "Diffusion Tensor MRI" OR "Diffusion Tensor Magnetic Resonance Imaging" OR "DTI MRI" OR "Tractography" OR "white matter" OR "microstructural damage" OR "fractional anisotropy" OR "Diffusion Tensor Imaging". We also traced the references of the relevant articles to find additional eligible studies. We included articles if they (1) explored WM using DTI or diffusion tensor MRI; (2) were conducted on individuals with TRD undergoing DBS. We excluded articles that (1) were conducted on subjects with disorders other than TRD (e.g., Parkinson's disorder, obsessive compulsive disorder); (2) did not perform DBS.
We also excluded letters to the editor, commentaries, and meeting abstracts. Data selection was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [70]. Two authors (GC and HSM) independently performed the eligibility assessment. Our search resulted in 295 articles, 240 of which remained after removing duplicates. After title/abstract screening, 224 papers did not meet our inclusion criteria, and were excluded. After full-text reading, 3 articles were excluded, thus resulting in 11 eligible DTI studies ( Figure 2). Details of the characteristics of the selected studies are shown in Table 2.
relevant articles to find additional eligible studies. We included articles if they (1) explored WM using DTI or diffusion tensor MRI; (2) were conducted on individuals with TRD undergoing DBS. We excluded articles that (1) were conducted on subjects with disorders other than TRD (e.g., Parkinson's disorder, obsessive compulsive disorder); (2) did not perform DBS. We also excluded letters to the editor, commentaries, and meeting abstracts. Data selection was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [70]. Two authors (GC and HSM) independently performed the eligibility assessment. Our search resulted in 295 articles, 240 of which remained after removing duplicates. After title/abstract screening, 224 papers did not meet our inclusion criteria, and were excluded. After full-text reading, 3 articles were excluded, thus resulting in 11 eligible DTI studies ( Figure 2). Details of the characteristics of the selected studies are shown in Table 2.

Data Extraction
One author (HSM) extracted the following data, which were checked by a second author (GC): (a) characteristics of the sample (age, gender, age of onset, duration of current depressive symptoms, medication profile, duration of follow-up); (b) diagnostic and psychopathological assessments; (c) technical specification of image acquisition; (d) DBS location; (e) key WM findings; (f) clinical outcomes.

Conclusions
In conclusion, the evidence that emerges from the reviewed studies demonstrates that DBS targeting the sgCC, superolateral medial forebrain bundle, rectus gyrus, and anterior limb of the internal capsule is associated with significant improvement in depressive symptoms. More importantly, the evidence suggests that the cingulum bundle, forceps minor, uncinate fasciculus, medial forebrain bundle, and sgCC seem to be the main WM structures mediating the beneficial effects of DBS on depressive symptoms in TRD. Nonetheless, future studies are indispensable to achieve enough clinical detail and statistical power to further our understanding on the association between WM structure and response to DBS.