Structural connectivity of low-frequency subthalamic stimulation for improving stride length in Parkinson’s disease

Highlights

• •Low-frequency stimulation (LFS) may improve gait impairments, but the underlying structural network is still not well understood.
• •We retrospectively assessed objective gait performance of 19 STN-DBS patients with 85 Hz and 130 Hz stimulation. A normative connectome was used to estimate structural connectivity.
• •Clinical improvements were related to the associative parts of the STN with connections to the supplementary motor area, prefrontal cortex, and pedunculopontine nucleus.
• •Non-beneficial effects were linked to stimulation of dorsolateral fibers near the STN.
• •Targeting the non-motor subregion of the STN with LFS may pose a potential therapeutic approach for gait disorders.

Abstract

Background

A reduction in stride length is considered a key characteristic of gait kinematics in Parkinson’s disease (PD) and has been identified as a predictor of falls. Although low-frequency stimulation (LFS) has been suggested as a method to improve gait characteristics, the underlying structural network is not well understood.

Objective

This study aims to investigate the structural correlates of changes in stride length during LFS (85 Hz).

Methods

Objective gait performance was retrospectively evaluated in 19 PD patients who underwent deep brain stimulation (DBS) at 85 Hz and 130 Hz. Individual DBS contacts and volumes of activated tissue (VAT) were computed using preoperative magnetic resonance imaging (MRI) and postoperative computed tomography (CT) scans. Structural connectivity profiles to predetermined cortical and mesencephalic areas were estimated using a normative connectome.

Results

LFS led to a significant improvement in stride length compared to 130 Hz stimulation. The intersection between VAT and the associative subregion of the subthalamic nucleus (STN) was associated with an improvement in stride length and had structural connections to the supplementary motor area, prefrontal cortex, and pedunculopontine nucleus. Conversely, we found that a lack of improvement was linked to stimulation volumes connected to cortico-diencephalic fibers bypassing the STN dorsolaterally. The robustness of the connectivity model was verified through leave-one-patient-out, 5-, and 10-fold cross cross-validation paradigms.

Conclusion

These findings offer new insights into the structural connectivity that underlies gait changes following LFS. Targeting the non-motor subregion of the STN with LFS on an individual level may present a potential therapeutic approach for PD patients with gait disorders.

1. Introduction

Gait impairments are frequently observed in patients with Parkinson’s disease (PD) and contribute significantly to disability and loss of independence (Okuma, 2014). Reduced gait velocity, step height, and step duration represent the hallmarks of gait kinematics in PD (Mirelman et al., 2019, Agharazi et al., 2023). An insufficient stride length could thus be a fundamental factor (Hausdorff, 2009) and has been recognized as a substantial indicator for predicting falls (Hoskovcová et al., 2015).

The role of deep brain stimulation (DBS) in modulating the ability of PD patients to ambulate has been controversial since its inception. Although electrical stimulation in the subthalamic nucleus (STN) effectively alleviates cardinal motor symptoms in PD (Pötter-Nerger and Volkmann, 2013, Horn et al., 2017), its efficacy for improving gait impairments and other axial signs is not as predictable (Barbe et al., 2020). These diverging results may be due to differences in stimulation parameters (Fan et al., 2023, Moreau et al., 2008, Ramdhani et al., 2015), which have led to interest in low-frequency stimulation (LFS, at 60–––85 Hz) as an effective approach to immediately address these challenges (Mügge et al., 2023). In addition to pulse configuration, pinpointing the exact location of stimulation could also be a method to facilitate gait kinematics of PD patients after surgery. Nevertheless, there is an ongoing debate regarding the identification of the most beneficial target for achieving optimal therapeutic outcomes (Weiss et al., 2013, Fasano et al., 2015).

The concept that subcircuits within the STN account for distinct symptoms in PD is widely accepted (Eisenstein et al., 2014). A subparcellation into associative, limbic, and motor parts of the STN has fundamentally shaped the rationale for the dorsolateral lead placement (Haynes and Haber, 2013). Recent evidence suggests that varying volumes of activated tissue (VAT) can lead to comparable outcomes for motor symptoms (Waldthaler et al., 2021), thus reinforcing the importance of incorporating additional anatomical information such as fiber tracts into clinical practice for both surgical planning and parameter programming. In addition, stimulation site network mapping has allowed the identification of treatment targets at group levels, which could enable tailored stimulation protocols linked to specific symptoms (Fan et al., 2023, Siddiqi et al., 2023, Strelow et al., 2022, Temiz et al., 2022). Cortico-subthalamic fibers originating from the prefrontal cortex (PFC) with connections to the mesencephalic locomotor region (MLR) have been associated with gait improvements, while stimulation at the lenticular fasciculus (Strelow et al., 2022) and fibers bypassing the pedunculopontine nucleus (PPN) (Moreau et al., 2008) may exert detrimental effects on gait kinematics. These findings underscore the critical role of accurate lead placement (Pötter-Nerger and Volkmann, 2013) and lend credence to a multinetwork model as a structural correlate of gait impairment in PD (Mirelman et al., 2019).

This retrospective analysis of a randomized controlled trial aimed at exploring the underlying network associated with changes in gait under different DBS frequencies. We hypothesized that the network includes structural connectivity to frontal cortical areas and the MLR.

2. Methods

2.1. Patients

Twenty-three PD patients (nine females, 57.0 ± 8.1 years) with chronic STN-DBS (Boston Scientific Neuromodulation Corporation, Valencia, CA 91355, USA) were recruited from the outpatient clinic of the Department of Neurology at the Philipps-University Marburg, Germany. All patients had undergone DBS surgery more than three months previously and had provided written informed consent. The study was approved by the local ethics committee (reference 33/20) and was conducted in accordance with the latest version of the Declaration of Helsinki. Four participants had to be excluded due to insufficient imaging data quality, resulting in 19 patients for the final analysis. For descriptive analyses see Table 1.

Table 1.

Clinical details. Abbreviations: DD: Disease duration, LEDD: Levodopa equivalent daily dose, MDS-UPDRS: Movement Disorder Society-Unified Parkinson’s Disease Rating Scale, NA: Not applicable, SD: Standard deviation.

ID	Age [years]	DD [years]	MDS-UDPRS III (Med ON, Stim ON)	Subtype	LEDD [mg]	Side prevalence	Time since surgery [months]
1	65	18	14	Bradykinetic-rigid	700	Right	10
2	58	20	13	Bradykinetic-rigid	450	Right	17
3	62	11	15	Equivalent	550	Right	12
4	72	8	25	Tremor-dominant	450	Left	12
5	56	9	9	Bradykinetic-rigid	500	Right	18
6	56	7	9	Bradykinetic-rigid	500	Right	8
7	43	11	14	Equivalent	475	Left	14
8	60	10	4	Equivalent	750	Left	6
9	56	6	10	Equivalent	650	Right	8
10	52	10	13	Bradykinetic-rigid	750	Right	14
11	46	5	4	Bradykinetic-rigid	NA	Right	12
12	63	25	7	Tremor-dominant	600	Right	24
13	58	11	4	Bradykinetic-rigid	125	Right	24
14	56	6	35	Bradykinetic-rigid	650	Left	26
15	56	3	23	Bradykinetic-rigid	500	Left	4
16	64	16	9	Bradykinetic-rigid	525	Left	8
17	68	12	19	Equivalent	450	Right	23
18	41	7	7	Bradykinetic-rigid	450	Left	27
19	51	10	21	Bradykinetic-rigid	500	Right	40
Mean ± SD	57.0 ± 8.1	10.8 ± 5.5	13.4 ± 8.2	NA	531.9 ± 144.7	NA	16.2 ± 9.2

2.2. Assessment of gait and statistical analyses

This study analyzed data from a randomized controlled trial conducted by Mügge and Kleinholdermann et al. (Mügge et al., 2023). Commercially available mobile inertial measurement units (Portabiles HealthCare Technologies GmbH, version 1.2.0) were utilized to measure gait kinematics (Jakob et al., 2021). The system employed inertial measurement units to record linear acceleration and angular rate data at a frequency of 102.4 Hz. It also utilized stride detection and parametrization, enabling objective gait assessments (Hannink et al., 2017, Barth et al., 2015). Stride length was selected as the main variable due to its recognized role as a predictor of falls (Hausdorff, 2009, Hoskovcová et al., 2015) and its correlation with other features of gait (Mügge et al., 2023). Furthermore, we considered a two-minute free walking task to be the most representative measure for analyzing the relationship between different DBS settings and gait performance. To this end, subjects completed a randomized sequence of free walking under DBS frequencies of 85 and 130 Hz. The order was thereby chosen randomly and the person responsible for DBS adaptations did not disclose the adjustments to the participants. Changes of motor performance were evaluated based on individual DBS settings and with respect to stimulation with 85 Hz. Medication remained unaltered during the entire study. Linear mixed effects models were utilized with the stimulation condition as a fixed effect and the different gait conditions as covariates. A linear contrast was used for post-hoc testing.

2.3. Localization of electrodes and individual stimulation volumes

The Lead-DBS software (version 2.6) was used for image processing according to the proposed steps (https://www.lead-dbs.org) (Horn and Kühn, 2015, Horn et al., 2019). Briefly, after coregistration of preoperative T1- and T2-weighted magnetic resonance imaging (MRI) and postoperative computed tomography (CT) scans, all images were spatially normalized into the standard MNI space (Fonov et al., 2011). Leads were localized with the PaCER (Precise and Convenient Electrode Reconstruction for Deep Brain Stimulation) algorithm, including their rotation (Husch et al., 2018). Visualization of electrodes and active contacts was performed using Lead-Group (Treu et al., 2020). Individual VAT, according to a finite element method with four compartments (gray and white matter, electrode contacts, and insulation), were subjected to group analyses for the relationship between the intersection of the VAT and the different STN subregions (motor, associative, and limbic) with respect to changes in objective gait performance. For this purpose, the individual percentage change in stride length between 85 Hz and 130 Hz (Formula 1) of the best clinical setting was included as a covariate (Treu et al., 2020). Spearman’s rank correlations were performed between gait outcomes and the VAT-STN intersections. Atlas segmentations of the STN were defined by the DISTAL Minimal Atlas (Ewert et al., 2018). A table containing the clinical outcome parameters and individual stimulation settings is included as supplementary material.

$$\frac{\mathit{Stride}\mathit{length}\mathit{at}85\mathit{Hz}\left[\mathit{cm}\right]-\mathit{Stride}\mathit{length}\mathit{at}130\mathit{Hz}\left[\mathit{cm}\right]}{\mathit{Stride}\mathit{length}\mathit{at}130\mathit{Hz}\left[\mathit{cm}\right]}x100$$ (1)

2.4. Region of interest analysis

The PFC, the supplementary motor area (SMA), and the precentral gyrus (PG) were selected as a priori defined regions of interest (ROI) based on their well-established connection to gait control (Mirelman et al., 2019, Takakusaki, 2013, Bohnen and Jahn, 2013, Snijders et al., 2016). Structural connectivity analysis was performed utilizing an openly available group connectome derived from diffusion-weighted MRI of 85 participants of the Parkinson’s Progression Markers Initiative database (Marek et al., 2011). Based on the normative connectome, the numbers of streamlines passing through each VAT were isolated and fiber counts to predefined ROI were correlated to clinical scores using Spearman’s rank-correlation (Treu et al., 2020). Cortical areas were derived from the brain parcellation using the Automated Anatomical Labelling Atlas 3 (AAL) (Rolls et al., 2020).

In addition to cortical areas, we investigated the structural connectivity to mesencephalic motor regions, i.e., the PPN and the substantia nigra (SN), according to their implications in gait changes following DBS targeting (Weiss et al., 2013, Strelow et al., 2022, Lin et al., 2020). Individual location of the PPN resulted from the Harvard Ascending Arousal Network Atlas, which has previously been used for structural connectivity estimates (Strelow et al., 2022, Edlow et al., 2012). Moreover, we correlated the overlap between the VAT and the SN (defined by the Nigral Organization Atlas (Zhang et al., 2017)with changes in clinical scores to determine whether stimulation sites in this structure were associated with clinical improvement. Multiple testing was corrected using the Benjamin-Hochberg method to control type-I error at a significance level of p < 0.05.

2.5. Whole-brain structural connectivity

To characterize the structural correlates of stride length improvement under LFS, we employed patient-specific VAT as seed regions for all the fiber tracts of the normative connectome. To this end, individual stimulation volumes were dichotomized, indicating whether or not they touched the fiber. With a two-sample t-test, we compared the clinical outcome parameter between the connected and unconnected stimulation volumes. As described by Li and colleagues, this approach resulted in t-scores for each fiber, with high t-values reflecting higher explanatory power for LFS induced changes in stride length (Li et al., 2020). Only high explanatory fibers were selected for further statistical analysis (top 20 % based on fiber t-scores) (Lofredi et al., 2020). In addition, we conducted Spearman’s rank correlation between the fiber density of connected cortical voxels and the LFS induced effects on stride length, visually represented as R-maps. To validate the obtained connectivity profiles, we conducted a leave-one-out cross-validation (Horn et al., 2019), resulting in a spatial correlation coefficient that was consequently subjected to a linear regression model to predict the effects of LFS on the clinical outcome of the removed patient. In addition, the robustness of the model was verified using 5- and 10-fold cross-validation (Hacker et al., 2023).

3. Results

3.1. Behavioural results

The coefficient estimate for stride length of an entire gait stance in the LFS condition (85 Hz) obtained from the linear mixed effects model was 118.8 cm, which was significantly higher than the estimate for the 130 Hz stimulation condition, which was 115.7 cm (z = 9.882, p < 0.001, Supplementary Fig. 1).

3.2. Localization of DBS leads and VAT and relationship between intersected VAT and STN subregions

The location of all DBS leads in standard space is illustrated in Fig. 1. In the right hemisphere, normalized to the total volume of the respective VAT, 24.0 ± 16.0 % [range (13.5, 31.5)] of the VAT were localized in the motor subregion, while 12.0 ± 11.5 % [range (2.8, 16.5)] and 5.1 ± 5.4 % [range (0.7, 7.5)] were found in the associative and limbic subregions, respectively. A slightly different distribution was observed in the left hemisphere. While 29.1 ± 15.5 % [range (17.2, 37.8)] of the VAT were identified within the motor subregion, 7.8 ± 11.3 % [range (1.6, 10.6)] and 3.6 ± 2.7 % [range (2.0, 5.0)] were found in the associative and limbic segments, respectively.

Fig. 1.

Lead positions in standard space in the right (A) and left (B) subthalamic nucleus (STN) in lateral view: The motor subregion of the STN is illustrated in brown, the associative subregion in red, and the limbic subregion in white, parcellation according to the DISTAL minimal atlas (Ewert et al., 2018). Active contacts are illustrated in red.

The VAT intersection with the associative subregion of the STN showed a positive correlation with change scores of stride length in the right hemisphere (left: rho = 0.01, p = 0.47; right: rho = 0.56, p = 0.03). No such correlation was observed for the motor (left: rho = 0.07, p = 0.47; right: rho = 0.21, p = 0.28) or the limbic (left: rho = 0.27, p = 0.28; right: rho = 0.33, p = 0.23) parts of the STN (Table 2). This implies that a greater proportion of the VAT localized in the associative subregion of the STN corresponded to a greater improvement in stride length with LFS (Fig. 2 A).

Table 2.

Spearman's rank correlations of changes in stride length with VAT-STN intersections. Abbreviations: VAT: volume of activated tissue, STN: subthalamic nucleus. Correlation p-values were corrected for type-I error.

	Intersection VAT - right motor STN		Intersection VAT - right associative STN		Intersection VAT - right limbic STN		Intersection VAT - left motor STN		Intersection VAT - left associative STN		Intersection VAT - left limbic STN
	rho	p	rho	p	rho	p	rho	p	rho	p	rho	p
change in stride length (130 Hz – 85 Hz)	0.21	0.28	0.53	0.03	0.33	0.23	0.07	0.47	0.01	0.47	0.27	0.28

Fig. 2.

Discriminative fibers connected with the volumes of activated tissue (VAT) and associated with changes in stride length under low-frequency stimulation (LFS). (A) Spearman’s rank correlations of VAT intersections with the associative subregion of the subthalamic nucleus (STN) and the LFS induced changes in stride length. (B-C) Close-ups of the right STN in lateral and medial view, respectively. Positively correlated fibers (red tractography) appear to enter the non-motor subregions of the STN, whereas negatively correlated tracts bypass the STN dorsolaterally (blue tractography).

3.3. Regions of interest based structural connectivity

Analysis of the cortical gait network, previously defined as ROI, revealed a significant positive correlation in both hemispheres for the SMA (left: rho = 0.50, p = 0.03; right: rho = 0.54, p = 0.03) and the superior frontal gyrus (left: rho = 0.50, p = 0.03; right: rho = 0.52, p = 0.03). Furthermore, significant structural connectivity was found for the middle frontal gyrus (left: rho = 0.28, p = 0.14; right: rho = 0.54, p = 0.03) and for the inferior frontal gyrus (left: rho = 0.32, p = 0.11; right: rho = 0.48, p = 0.03) in the right hemisphere (Fig. 3 A-C), as well as the PG in the left hemisphere (left: rho = 0.58, p = 0.03; right: rho = 0.12, p = 0.33). This suggests that higher fiber counts connecting VAT to these ROIs were associated with greater LFS induced improvement in stride length (Table 3).

Fig. 3.

Region of interest (ROI) based analysis and whole-brain structural connectivity maps. (A-D) Significant Spearman’s rank correlations of fiber counts connecting volume of activated tissue (VAT) to the cortical gait network and mesencephalic locomotor regions (MLR). (E) Fiber density of connected cortical voxels and the low-frequency stimulation (LFS) induced effects on stride length. Correlation coefficients are color coded for each voxel of the normative connectome. SMA, supplementary motor area; MFG, middle frontal gyrus; IFG, inferior frontal gyrus; PPN, pedunculopontine nucleus.

Table 3.

Spearman's rank correlations of changes in stride length with the volumes of activated tissue (VAT) connecting fiber counts to regions of interest (ROI). Abbreviations: PG: precentral gyrus, SMA: supplementary motor area, SFG: superior frontal gyrus, MFG: middle frontal gyrus, IFG: inferior frontal gyrus, PPN: pedunculopontine nucleus. Correlation p-values were corrected for type-I error.

	right PG		left PG		right SMA		left SMA		right SFG		left SFG		right MFG		left MFG		right IFG		left IFG		PPN
	rho	p	rho	p	rho	p	rho	p	rho	p	rho	p	rho	p	rho	p	rho	p	rho	p	rho	p
change in stride length (130 Hz – 85 Hz)	0.12	0.33	0.58	0.03	0.50	0.03	0.54	0.03	0.52	0.03	0.50	0.03	0.54	0.03	0.28	0.14	0.48	0.03	0.32	0.11	0.48	0.03

Subcortically, we calculated the overlap from the VAT to the SN and the fiber connection to the PPN. We found that an improvement in stride length under LFS was positively correlated with the structural connectivity of VAT to the PPN in both hemispheres (rho = 0.48, p = 0.03) (Fig. 3 D), whereas no significant correlation was observed between the overlap of VAT and the SN (rho = 0.30, p = 0.11).

3.4. Whole-brain structural connectivity

The whole-brain structural connectivity maps derived from voxel-wise Spearman correlations are depicted in Fig. 3 E. In the leave-one-out validation, the LFS effects of single individuals on stride length were predictable on the respective discriminative fiber profiles of the remaining 18 participants (R = 0.42, p = 0.03) (Supplementary Fig. 2 A). This implies that higher similarity of the individual set of discriminative fibers to the group fiber profile correlated with a larger DBS induced effect on stride length. Repeating the same analysis in 5- or 10-fold cross-validation also led to significant correlations (R = 0.44, p = 0.02 and R = 0.44, p = 0.03, respectively) (Supplementary Fig. 2 B-C). Strikingly, the connectivity profile demonstrated a lateralization that favored discriminative fibers in the right hemisphere. Fibers linked to an improvement in stride length appeared to enter the non-motor subregions of the STN (Fig. 2 C-D) and projected to frontal and prefrontal areas. Strong correlations were also observed between voxels of the PFC and changes in stride length in the whole-brain structural connectivity maps (Fig. 3 E). Conversely, fiber tracts associated with a lack of improvement passed by the STN dorsolaterally and traversed to the right PG and paracentral lobule, providing downstream connections to the cerebellum (Fig. 4 A-B). In a post-hoc analysis, the volumetric size of the VAT localized outside the STN was not predictive of changes in stride length (p > 0.05). This suggests that the observed effect was likely due to the bypassing fiber bundles as opposed to unspecific stimulation of tissue outside the STN.

Fig. 4.

Discriminative fiber tracts seeding from bilateral volume of activated tissue (VAT). (A) Lateral view of significant fibers connected to the right VAT. Fibers linked to an improvement in stride length (red tractography) appear to originate from premotor areas and the prefrontal cortex. Subcortically, they traverse the internal capsule and project to the non-motor subregions of the subthalamic nucleus (STN) before descending to spinal motor neurons. Conversely, negatively correlated tracts (blue tractography) seem mainly to derive from the primary motor region, bypassing the STN lateral to its border and putatively encompassing the cerebellum. (B) Coronal view of fibers seeding from bilateral VAT. Activation profiles exhibit a lateralization favoring discriminative fibers in the right hemisphere.

4. Discussion

In this study, we aimed to investigate the structural network that underlies the effective use of low-frequency stimulation for gait in PD patients. We could thereby show beneficial effects when stimulation was delivered to the associative subregion of the STN, whilst no such association was found for the motor part. In addition, improvements in stride length were correlated with volumes of activated tissue that were connected to the supplementary motor area and the prefrontal cortex, with the PPN also being part of this facilitatory network. Conversely, stimulation of fibers dorsolaterally adjacent to the STN and originating in the primary motor cortex was not found to be beneficial for stride length. These findings provide a new perspective by offering a differentiated view of the subcircuits that may be involved in measurable gait disturbances and their modulation.

Models of the basal ganglia attribute a pivotal role to the STN as a key structure for motor control (Gulberti et al., 2023, Kravitz et al., 2010). Our results suggest that the location of the VAT within the STN can affect stride length, particularly when located in its associative subregion. Strikingly, there was no correlation between stimulation volumes intersecting the dorsolateral motor part, which is consistent with previous studies showing that LFS is most effective when distributed ventrally (Ramdhani et al., 2015, Xie et al., 2017, Zibetti et al., 2016, Khoo et al., 2014). Hence, for gait disorders, optimizing active contacts towards ventral stimulation sites may improve its clinical efficacy while potentially maintaining effective symptom management (Khoo et al., 2014).

In addition to local effects, a growing body of evidence points to a network-dependent progression of symptoms in PD (McGregor and Nelson, 2019, DeLong and Wichmann, 2015, Holtbernd and Eidelberg, 2012). In particular, this study builds upon a demonstrated association between aberrant hypersynchronization within striatocortical and mesocorticolimbic pathways and gait disturbances (Steidel et al., 2021). Our results indicate that a physiological connection between the non-motor subregion of the STN and the PFC and SMA is a prerequisite for achieving beneficial effects on gait with LFS. Progressive dysfunction of circuits involving these structural connections has been implicated in the loss of ability to sequence and coordinate postural adjustments and stepping movements (Fling et al., 2014), along with impairments in cognitive decision-making processes (Ranchet et al., 2020). Furthermore, the integration of the STN within a broader network that includes associative cortico-subthalamic fiber recruitment (Rodriguez-Rojas et al., 2022) highlights its significance as a stimulation target to modulate different neural circuits involved in motor functions. This perspective reinforces the crucial role of the STN, not only as a central hub in motor control, but also in wider-ranging circuits that can be therapeutically harnessed for gait improvement in PD.

Building on this understanding of the integrative role of the STN, our findings also shed light on subcortical mechanisms, particularly the impact of LFS on the outflow tracts of the PPN, an integral part of the MLR (Lin et al., 2020). The MLR plays a crucial role in the regulation of gait, embedded within a network of dense interconnections to brainstem circuits and cortical areas (Nutt et al., 2011). It has been proposed that decoupling between the SMA and the motor cortex can result in an increased recruitment of the hyperdirect pathway, leading to enhanced GABAergic inhibition of the MLR and ultimately causing gait disturbances (Bardakan et al., 2022, Lewis and Barker, 2009, Lewis and Shine, 2016). Interestingly, PPN-DBS has shown potential benefits in improving gait characteristics that do not respond to levodopa treatment (Ramdhani et al., 2015, Ferraye et al., 2010). Further credence to this argument may be provided by the effects of targeting non-beneficial fiber tracts that bypass the STN dorsolateral to its border. These fiber bundles are closely located to the zona incerta and the lenticular fasciculus, both of which have been associated with stimulation induced gait deterioration (Strelow et al., 2022, Fleury et al., 2016). The lenticular fasciculus originates from the globus pallidus internus, passes through the posterior limb of the internal capsule and merges with the ansa lenticularis to form the thalamic fasciculus in Forel’s field H1, situated dorsal to the zona incerta (Fleury et al., 2016). From a network perspective, the effects of STN-DBS on gait may be partially mediated by cerebellar-thalamo-cortical circuits (Hill et al., 2013), consistent with our observations that these fibers appear to originate in the right primary motor cortex and provide downstream connections to the cerebellum.

We corroborate previous literature indicating differences in structural connectivity between hemispheres, with activation profiles exhibiting a right lateralization (Fan et al., 2023, Temiz et al., 2022). This observed asymmetry may be attributed to volumetric variances of STN subregions between hemispheres (Temiz et al., 2022) and the distinct influence of the right STN on locomotion, alongside its cortical projections (Marchal et al., 2019, Aron et al., 2014). Another explanation may be reflected in the utilization of asymmetric stimulation volumes for individual patients, addressing the lateralized manifestation of parkinsonian motor symptoms (Postuma et al., 2015). Ultimately, we cannot draw any definitive conclusions about laterality since the clinical assessment was only conducted with bilateral stimulation and confirmatory data should be collected to elucidate the mechanisms underlying this phenomenon.

4.1. Limitations

There are some limitations to consider. First, due to the retrospective design, DBS parameters were not refined using the results of our connectivity analysis. Therefore, it is speculative whether directing current to non-motor STN subregions would have been tolerated in a daily routine considering the lower energy levels and the potential for increased motor disability. This may be particularly important in patients with tremor-dominant PD, as LFS has been found to affect tremor control (Phibbs et al., 2014); emphasizing the importance of careful patient selection. Additionally, one could argue that the beneficial effects might diminish over time (Xie et al., 2017). One possible solution could be the use of a multicurrent approach instead of a single LFS setting, as most modern systems allow for adjusting stimulation parameters accordingly (Pollo et al., 2014). Another limitation of our VAT modeling approach is that it does not take into account frequency-dependent changes in tissue conductivity, which could potentially affect our results. Moreover, normative connectome analyses do not account for individual variations in structural connectivity estimates. Finally, the relatively small sample size may hamper the generalizability of the results.

5. Conclusions

Our results suggest that the stride length benefits of LFS may be due to current application to the associative subregion of the STN and recruitment of premotor and prefrontal subthalamic pathways with structural connections to the PPN. Moreover, we observed the presence of fibers originating from the primary motor cortex that were associated with a lack of improvement. Taken together, this suggests a complex and widespread network involved in the modulation of gait in PD. In a clinical setting, individual VAT targeting the non-motor subregion of the STN may pose a therapeutic strategy for PD patients with gait disorders.

Funding

AC was supported by the travel grant of the Thiemann Foundation.

CRediT authorship contribution statement

Alexander Calvano: Writing – original draft, Formal analysis, Data curation, Conceptualization. Urs Kleinholdermann: Writing – review & editing, Visualization, Formal analysis, Data curation. Amelie-Sophie Heun: Investigation, Data curation. Miriam H.A. Bopp: Writing – review & editing, Validation. Christopher Nimsky: Investigation, Data curation. Lars Timmermann: Investigation, Data curation. David J. Pedrosa: Writing – review & editing, Visualization, Supervision, Methodology, Formal analysis, Data curation, Conceptualization.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.