Abstract
Humans adjust their movement to changing environments effortlessly via multisensory integration of the effector’s state, motor commands, and sensory feedback. It is postulated that frontoparietal (FP) networks are involved in the control of prehension, with dorsomedial (DM) and dorsolateral (DL) regions processing the reach and the grasp, respectively. This study tested (five female and five male participants) the differential involvement of FP nodes [ventral premotor cortex (PMv), dorsal premotor cortex (PMd), anterior intraparietal sulcus (aIPS), and anterior superior parieto-occipital cortex (aSPOC)] in online adjustments of reach-to-grasp coordination to mechanical perturbations (MP) that disrupted arm transport. We used event-related transcranial magnetic stimulation (TMS) to test whether the nodes of these pathways causally contribute to the processing of proprioceptive information when reaching for a virtual visual target at two different perturbation latencies. TMS over aSPOC selectively altered the correction magnitude of arm transport during late perturbations, demonstrating that aSPOC processes proprioceptive inputs related to mechanical perturbations in a movement phase-dependent manner.
Significance Statement
Detailed knowledge regarding specific brain regions, the timing of their involvement, and roles in the online updating of sensory input during the control of the reach-to-grasp movement is critical for understanding the deficits resulting from various diseases (e.g., stroke) and developing effective intervention strategies. Our results provide evidence for the involvement of the anterior superior parieto-occipital cortex in the modulation of the response to a mechanical perturbation of arm transport during the later stage of the movement, suggesting that this area participates in estimating the effector state based on proprioceptive input. Our results can provide key information for identifying brain targets of engagement in therapeutic applications of noninvasive brain stimulation to ameliorate motor deficits stemming from parietal damage.
Introduction
The dynamic nature of our environment often requires fast and accurate corrections of our actions for unexpected perturbations. For example, when reaching to grasp an object, a sudden bump on the arm (perturbing the internal state of the body) rarely gives us pause because we are so deft at rapid online adjustments. This flexible, coordinated control of reach-to-grasp movements depends on the integration of information about the motor command and sensory feedback arising from visual and proprioceptive feedback about the movement (Desmurget and Grafton, 2000). Based on convergent evidence from human and nonhuman primate research, two frontoparietal (FP) pathways linking the posterior parietal cortex (PPC) to the premotor cortex (PM) have been identified as subserving sensorimotor integration and control of the reach-to-grasp movement (Galletti et al., 2001; Buneo and Andersen, 2006, 2012; Filimon et al., 2009). A dorsomedial (DM) “reach” pathway consisting of the dorsal premotor cortex (PMd, macaque homolog area F2) and the superior parieto-occipital cortex [anterior superior parieto-occipital cortex (aSPOC) including area hV6A, macaque homolog area V6A] supports the visuomotor transformations necessary for reaching, and the dorsolateral (DL) “grasp” pathway consisting of the ventral premotor cortex (PMv, macaque homolog area F5) and the anterior intraparietal cortex [anterior intraparietal sulcus (aIPS)/phAIP, macaque homolog area AIP] supports the visuomotor transformations necessary for grasping (Jeannerod, 1984, 1999; Tanné-Gariépy et al., 2002; Culham et al., 2003, 2006; Frey et al., 2005; Pitzalis et al., 2013, 2015; Orban, 2016). These pathways contain neurons that respond to both visual and proprioceptive inputs (Davare et al., 2011; Breveglieri et al., 2023c, 2023a) as well as during prehension movements (Franklin et al., 2016).
To evaluate the causal contributions of FP brain areas to online updating of reach-to-grasp movements in humans, transcranial magnetic stimulation (TMS) “virtual lesions” (Pascual-Leone, 1999; Silvanto and Muggleton, 2008a) have been employed during tasks requiring rapid corrections to unexpected perturbations. To date, nearly all studies of this design have utilized visual perturbations of the target object (Tunik et al., 2005; Buch et al., 2010; Vesia et al., 2010; Reichenbach et al., 2011; Verhagen et al., 2013; Breveglieri et al., 2023c) providing important information about the specialization of FP nodes for specific reach-to-grasp functions.
Less well understood is the specialization of FP brain areas for online corrections in response to physical perturbations of the limb. In a previous study (Schettino et al., 2015), we used a mechanical perturbation of the index finger during a reach-to-grasp to a visual target, while single-pulse TMS was applied to one of the two nodes of the DL pathway in a random order. Our results showed a modulation of the rapid response to the finger perturbation following TMS at ∼50 ms after movement onset but only for PMv and not to aIPS. Our conclusions were that PMv, but not aIPS, may play a role in the updating of the response to the perturbation of the effector state. While numerous other studies have strongly suggested that FP pathways (Fattori et al., 2010, 2017; Breveglieri et al., 2016, 2023b, 2023a; Galletti and Fattori, 2018) are involved in proprioceptive information processing for online correction, often through the removal of visual feedback, the generalization of these findings to physical (proprioceptive) perturbations of limb state remains speculative. Furthermore, studies from our lab and others have provided evidence that FP networks rely on continuous state monitoring for updating and error correction (Desmurget et al., 1999; Mulliken et al., 2008a, 2008b; Archambault et al., 2009) and the role of FP nodes in processing a perturbation may be different given the state of the movement. However, this information is also primarily derived from studies using visual perturbations, and therefore generalization of findings to physical perturbations is limited.
Therefore, to directly test the role of FP nodes in the online processing of physical perturbations of the limb, we applied a mechanical (proprioceptive) perturbation of the arm (limb state) during reach-to-grasp in a virtual reality (VR) environment and delivered TMS to each of the nodes in the two FP pathways at two different timings (early, 100 ms; late, 300 ms) during the movement. We hypothesized that while both DM and DL may be involved in proprioceptive processing, the specific action (reach or grasp) and timing (early or late in the movement) of their involvement may be different, with the medial pathway involved in arm transport during the early stage of the reach-to-grasp and the lateral pathway involved in the control of hand aperture at latter stages (Elliott et al., 2010).
Materials and Methods
Participants
All protocols were conducted in conformance with the Declaration of Helsinki and were approved by the Institutional Review Board of Northeastern University. Ten right-handed (Oldfield, 1971; Steenhuis and Bryden, 1989) participants (mean ± 1SD age = 24 ± 7.3 years; five females) with no reported muscular, orthopedic, or neurological health concerns, voluntarily participated in this study after providing verbal and written informed consent.
Experimental setup
Each participant reached to grasp virtual objects in an immersive virtual environment (VE) developed in UNITY (ver. 5.6.1f1, 64-bit, Unity Technologies) and delivered via an HTC Vive Pro head-mounted display (HMD; HTC; Fig. 1). The interpupillary distance in the HMD was individually adjusted to each participant. Motion tracking of the head was achieved by streaming data from an inertial measurement unit and laser-based photodiodes embedded in the HMD. An eight-camera motion tracking system (PPT Studio NTM, WorldViz; sampling rate, 90 Hz) recorded the 3D motion of infrared emitting diode markers attached to the participant's wrist (at the center of the segment running between the ulnar and radial styloid process) and the thumb and index fingertips. Participants viewed their thumb and index fingertips as two 3D spheres (green in color, 0.8 cm diameter) in VE, corresponding to the 3D position of the respective IRED markers (Furmanek et al., 2021). A Phantom Premium haptic device (3D Systems) attached to each participant's wrist was used to apply mechanical perturbations (MP) of the reach (Fig. 1). The schedule of trials, virtual renderings of the target object, and timing/triggering of the perturbation were controlled using custom software developed in C#. Our previous work has shown that the coordination pattern of reach-to-grasp movements is preserved irrespective of whether individuals reach to grasp physical objects or their exact virtual renderings in this immersive haptic-free VE (Furmanek et al., 2019, 2021; Mangalam et al., 2021).
Experimental setup and perturbation conditions. Top panel, Seated participants in an immersive virtual environment. Middle panel, TMS-induced cortical perturbation was applied to PMv, PMd, aIPS, and aSPOC, verified by finite-element modeling of the electric field (E-field) induced by TMS. Example E-fields (V/m) calculated for actual stimulations in our pilot data are shown. Bottom panel, 100 ms (early, E) or 300 ms (late, L) after movement onset, a haptic device perturbed the internal state of hand transport by applying a 5 N force resisting the hand motion along the axis of the reach. Mechanical perturbation (MP) of hand transport was time-locked with TMS. No perturbation was applied in the control (unperturbed) conditions.
Experimental procedure
Each participant sat on a chair in front of a table with their right hand resting on the tabletop. At the start position, the thumb and index finger straddled a 1.5-cm-wide wooden peg located 12 cm in front and 24 cm to the right of the sternum, with the thumb depressing a switch (Fig. 1, top). Lifting the thumb off the switch marked movement onset. Upon an auditory cue (timing jittered on each trial to reduce anticipation) that indicated the start of a trial, participants reached to grasp the virtual object presented in the HMD, lifted it, and held it until it disappeared (3.5 s from movement onset, i.e., the moment the switch was released) and returned their hand to the starting position (Fig. 1, bottom). A custom collision detection algorithm was used to determine when the virtual object was grasped (Furmanek et al., 2019, 2021). The virtual rectangular object, 5.4 × 2.5 × 8.0 cm in dimension, was located 30 cm away from the hand position and oriented 75° relative to the coronal axis, allowing for a more naturalistic grasp that did not require excessive wrist extension (Fig. 1, top). Participants were instructed to grasp the object using a pincer grasp directed at the center of the object's lateral surfaces. Successful grasp was indicated by turning the object's color to red and allowing the object to be lifted by the virtual grasp. Participants were familiarized with the VE and the reach-to-grasp task prior to data collection by allowing up to 120 practice trials (e.g., until subjects could reach and grasp the object at a comfortable speed). Our previous work has shown that this familiarization period with the VE and the reach-to-grasp task allows for stable performance (Furmanek et al., 2019).
TMS-induced cortical perturbations and perturbation timing
Each participant completed one session per brain site (aSPOC, PMd, aIPS, PMv). Mechanical perturbations (MP) were applied (early, 100 ms, or late, 300 ms, after movement onset) using a robotic manipulandum (Phantom Premium haptic device) attached to the wrist, which exerted a force of 5 N resisting the motion of the hand along the axis of the reach. The perturbation lasted for the remainder of the trial. Each session consisted of 512 trials, lasted ∼2 h, and was divided into four 30 min miniblocks punctuated by 2 min breaks to prevent task fatigue. Each miniblock comprised 128 trials. TMS was delivered concurrently to each perturbation in 50% of all trials. MP comprised 25% of trials in each miniblock (32 trials). Early (100 ms after movement onset) and late (300 ms after movement onset) perturbations were equally divided among the 128 perturbed trials. Therefore, each combination of mechanical and cortical (TMS) perturbation (two perturbation timings multiplied by two TMS conditions) consisted of eight trials. The randomized trial order within each miniblock was repeated four times. Based on our prior work (Schettino et al., 2015; Furmanek et al., 2019), the early and late perturbations were timed to occur prior to and shortly after, respectively, the time when peak transport velocity (PTV) would occur in unrestrained movements. Control conditions included 48 unperturbed without TMS (control) trials as well as 24 early and 24 late unperturbed with TMS trials (U-TMS-E, U-TMS-L). The control conditions are meant to test the results of experimental trials against a natural, unimpeded reach-to-grasp movement. The method for calculating the outcome measures following perturbation trials (P-TMS-E, P-TMS-L) relative to noTMS conditions (P-noTMS-E, P-noTMS-L) is presented in Figure 3.
Neuronavigated transcranial magnetic stimulation
Each participant's anatomical high-resolution MRI was acquired (Siemens MAGNETOM Prisma 3T; T1-weighted 3D MPRAGE scan; Inversion Time, 1,100 ms; Echo Time, 2.63 ms; Repetition Time, 2,000 ms; Field of View, 256 × 192 mm; 256 × 192 × 160 acquisition matrix; 1 × 1 × 1 mm voxels). Fiducial locations on the MRI were coregistered with the participant’s head to allow frameless neuronavigation to ensure consistency of stimulation location (Brainsight, Rogue Research; Fig. 2). All TMS procedures utilized a Magstim BiStim2 with a D702 figure-of-eight coil (Magstim). The location of the cortical hotspot for the first dorsal interosseous muscle was identified by performing a course mapping of the left precentral gyrus hand knob area using standard procedures (Weiss et al., 2013). The hotspot was selected as the stimulation site producing the largest motor-evoked potential, quantified as the peak-to-peak amplitude of the electromyography signal in a 10–40 ms window following the TMS pulse. The resting motor threshold (RMT) was then defined as the stimulator intensity, in percent of maximum stimulator output, required to elicit motor-evoked potentials >50 µV on 5 out of 10 consecutive trials (Rossini et al., 2015). RMT was found to be 41.1 ± 6.6% across subjects.
Reconstructed three-dimensional images of each participant's brain (n = 10) and the target sites over which the TMS coil was positioned. PMd, dorsal premotor cortex; PMv, ventral premotor cortex; aSPOC, anterior superior parieto-occipital cortex; aIPS, anterior intraparietal sulcus.
Localization of brain sites and TMS
Each of the four brain sites (aSPOC, PMd, aIPS, PMv) was marked based on anatomical landmarks identified on each participant's reconstructed volumetric MR image (Fig. 2), guided by established localizers from prior published fMRI and TMS studies on related tasks: (1) anterior superior parieto-occipital cortex (aSPOC), along the medial surface of the parietal lobe and just anterior/rostral to the parietal occipital (PO) sulcus, commonly referred to as the precuneate region; it includes the parietal convexity adjacent to and just off the midline precuneate region but medial to the intraparietal sulcus (Cavina-Pratesi et al., 2010; Vesia et al., 2010, 2013); (2) dorsal premotor cortex (PMd), superior portion of the precentral gyrus, as delimited by the superior frontal sulcus (Davare et al., 2006); (3) anterior intraparietal sulcus (aIPS), the junction between the anterior extent of the intraparietal sulcus and the postcentral sulcus (Tunik et al., 2005); and (4) ventral premotor cortex (PMv), caudal part of pars opercularis of the inferior frontal gyrus, between the vertical ramus of the lateral fissure and the precentral gyrus (Davare et al., 2006, 2010). During the experiment, TMS was applied as two pulses spaced 50 ms apart to one of the four brain sites at 120% of RMT. A double pulse was used to extend the virtual lesion effect in order to ensure disruption of processing is coincidental with afferent feedback about the perturbation. The use of double pulses in virtual lesion experiments is common, and prior research has shown that double pulses (40–50 ms interstimulus interval) can be used to maintain temporal resolution while maximizing behavioral effects due to summation of pulse effects on the cortex (O’Shea et al., 2004; Ellison et al., 2007; Kalla et al., 2008; Sliwinska et al., 2014; Luber et al., 2020). In our experiment, the double pulse ensures the duration of the virtual lesion coincides with afferent feedback about the perturbation. On all trials, TMS was automatically triggered by a TTL signal (transistor-transistor logic, set to 5V)at the latencies (early, 100 ms, and late, 300 ms, after movement onset) described above.
Finite-element modeling of the electric field (E-field)
We performed finite-element (FE) modeling of the TMS-induced electric field to validate that stimulation at each desired node would indeed target the intended brain regions (Fig. 1, middle). For this, we constructed subject-specific FE head models and computed the TMS-induced E-fields in the brain, per established procedures (Htet et al., 2019; Saturnino et al., 2019a), to maximize effects and reduce intersubject variability. Head models were created by tissue segmentation of high-resolution structural MRIs using the SPM SimNIBS toolbox (Saturnino et al., 2019b), generating a detailed tetrahedral mesh (4–6 million elements) with conductivity values assigned to each element depending on tissue type. Using the SCIRun problem-solving environment and its Brain Stimulator toolkit (Parker and Johnson, 1995; Dannhauer et al., 2017), the vector potential of the TMS coil was approximated by small magnetic dipoles (Thielscher and Kammer, 2004; Windhoff et al., 2013), and the E-field induced by these dipoles in the head model was calculated (Janssen et al., 2013; Rampersad et al., 2019).
Kinematics data processing
Data were analyzed offline using custom Matlab routines (The MathWorks). Kinematic data were low-pass filtered at 6 Hz with a fourth-order Butterworth filter. Trials were cropped from movement onset (start switch) to moment offset (contact of both thumb and index finger with the object). In VR, the offset was defined as the timestamp when the virtual object was successfully grasped (thumb and index finger markers met the collision detection criteria; Furmanek et al., 2021).
Outcome measures
The traditional view of TMS as a “virtual lesion” postulates that TMS to a given cortical area that is presumed to be involved in behavior will result in delayed or deteriorated performance of that behavior. However, recent findings suggest that the effects of TMS on a given cortical area may affect behavior in a manner that is dependent on the state of the neural plant; that is, as a function of whether neurons involved in the task are inhibited, excited, or not modulated at the time of stimulation (Pascual-Leone et al., 2000; Silvanto and Muggleton, 2008b; Siebner et al., 2009; Ziemann, 2010). Based on this view, we selected two outcome measures describing unique aspects of the corrective response for transport and aperture components (Fig. 3).
Method for calculating outcome measures. Left, Transport position and its derivatives. Right, Aperture and its derivatives. Vertical lines mark “late” perturbation onsets (P), 300 ms after switch release. The gray areas indicate the perturbation activity. Correction magnitude: difference between the local minimum and maximum in the velocity profiles. Correction latency: delay between perturbation onset and the first zero crossing in the acceleration profiles; see text for details.
Correction magnitude
This refers to the change in velocity at the time of the first two zero crossings in the acceleration profile (corresponding to a local minimum and local maximum in the velocity profile, respectively), occurring in the immediate postperturbation period (100–300 ms after perturbation, i.e., 100–300 ms for perturbation applied at 100 ms and 300–500 ms for perturbation applied at 300 ms). The change in peak-to-peak correction magnitude indicates the extent to which TMS led to a suppressed or exaggerated corrective action. We interpret TMS-induced changes in correction magnitude in terms of modulation of the signal-to-noise ratio within a cortical node, which consequently modifies the information output from that node (Silvanto and Cattaneo, 2017), or disruption of state estimation that results in a corrective response formed from old or noisy information about the current or future limb state (Miall et al., 2007).
Correction latency
This is quantified as the delay between perturbation onset and the first zero crossing in the acceleration profile (corresponding to a local minimum in the velocity profile), occurring in the immediate postperturbation period (100–300 ms for perturbation applied at 100 ms and 300–500 ms for perturbation applied at 300 ms). Prolongation of the correction latency would indicate the extent to which the TMS delayed online updating of the reach-to-grasp movement. We interpret TMS-induced increases in correction latency in terms of delays in the serial processing of the perturbation, which may reflect either a delay in recognition of the perturbation, a delay in information flow, or a delay in output to motor cortices (Veerman et al., 2008; Oostwoud Wijdenes et al., 2014; Holmes and Dakwar, 2015).
Statistical analysis
We applied separate 4 × 2 × 2 rm-ANOVAs to examine the effects of “brain site” (aSPOC, PMd, aIPS, PMv), “cortical perturbation” (noTMS, TMS), and “timing” (early, late) on dependent measures of correction magnitude and latency for both transport and hand aperture. Given that we were interested in the effects of TMS at specific cortical sites and at specific movement phases, we followed up only the third-order significant interaction with 4 × 2 rm-ANOVAs to examine the effects of “brain site” (aSPOC, PMd, aIPS, PMv) and “cortical perturbation” (noTMS, TMS) at each “timing.” Significant main effects of “brain site” were followed by Holm's sequential Bonferroni’s procedure (Holm, 1979). The Shapiro–Wilk test was used to verify the normality of data distribution; all variables met this assumption. The Mauchly's sphericity test was used to validate assumptions for repeated-measures ANOVAs. Each test was performed in Statistica ver. 13 (TIBCO Software), and each test statistic was considered significant at the two-tailed α level of 0.05. Significant effect sizes are reported as partial eta-squared (η2) in Table 1.
Outcomes of 4 × 2 × 2 rm-ANOVAs examining the effects of “brain site” (aSPOC, PMd, aIPS, PMv) on which TMS was applied, “cortical perturbation” (noTMS, TMS), and “timing” (early, late) on online updating of reach-to-grasp movements to unpredictable mechanical perturbation of hand transport
Results
The current study used mechanical perturbations of the reach to disrupt the effector state and elicit a correction via proprioceptive input. We also used cortical perturbations by means of TMS to test for causal links between the processing of sensorimotor input and the behavioral correction to the mechanical perturbation. The mean MNI coordinates for all participants were as follows: aSPOC (x = −17.5 ± 2.5, y = −82.4 ± 3.8, z = 48 ± 5), PMd (x = −25.6 ± 3.1, y = −0.3.4 ± 4.5, z = 67.5 ± 2.5), aIPS (x = −50.2 ± 3.7, y = −44.4 ± 5.7, z = 55.8 ± 2.1), and PMv (x = −57.3 ± 2.2, y = 15.2 ± 3.5, z = 21.1 ± 4.6). We tested the involvement of the individual nodes of the DM and DL reach-to-grasp pathways at two different timings following movement initiation (early) or movement shaping (late). All participants tolerated stimulation well, and no adverse events occurred. Figure 4 presents the mean kinematic profiles of the transport and aperture components of the reach-to-grasp movement with (TMS) and without TMS (noTMS) for the “unperturbed” (Fig. 4A,D), “early perturbation” (Fig. 4B,E), and “late perturbation” (Fig. 4C,F) conditions in a representative participant. The “control” condition is depicted in black in perturbation plots.
Effects of TMS on the reach (left column) and grasp (right column) components of the reach-to-grasp movements, averaged across trials for a representative participant. Transport velocity profiles for (A) unperturbed, (B) early perturbation (100 ms), and (C) late perturbation (300 ms). Aperture profiles for (D) unperturbed, (E) early perturbation (100 ms), and (F) late perturbation (300 ms). Shaded areas indicate ± 1SD. Insets in B and C highlight the effect of TMS on transport velocity.
Analysis of no perturbation conditions
To test whether TMS alone, in the absence of the mechanical perturbation, may have resulted in any observable effects, we performed a 4 × 3 rm-ANOVA using all control conditions (control, U-TMS-E, U-TMS-L) for the following parameters: movement time, peak transport velocity, and peak aperture (Fig. 4A,D). None of these analyses showed significant effects (main effects and interactions, p > 0.05), indicating that it is unlikely that TMS had any appreciable effect on kinematics when the movements were performed without a need to make online updates.
Analysis of the effects of TMS on corrections to the transport component
As expected, the control condition is characterized by a single velocity peak. Mechanical perturbations resulted in a notable double peak in transport velocity as participants produced a second velocity peak to counteract the pull of the perturbation. In the early condition, application of the mechanical perturbation during the initiation phase of reach-to-grasp resulted in a lower peak transport velocity (PTV) relative to control, followed by a secondary peak (Fig. 4B). Conversely, in the late condition, mechanical perturbations applied in the shaping phase increased peak deceleration without affecting PTV (relative to control; Fig. 4C). The transport correction magnitude was generally observed to be larger for the late than the early condition (Fig. 4, compare B, C).
To test the involvement of each cortical site in the online updating of the reach to compensate for the mechanical perturbation, two 4 × 2 × 2 rm-ANOVAs: “brain site” (aSPOC, PMd, aIPS, PMv), “timing of perturbation” (early, late), and “cortical perturbation” (noTMS, TMS) were run for each of the transport variables (transport correction magnitude and transport correction latency). A significant interaction of “brain site,” “timing,” and “cortical perturbation” was observed for transport correction magnitude (F(3,27) = 3.07, p = 0.044, Table 1). We followed the three-way interaction with a Bonferroni–Holm post hoc test for each “brain site” and “timing” (4 × 2) to test for the effect of TMS. This analysis involved eight comparisons (each brain site early/late/noTMS vs each brain site early/late/TMS; Table 2). For this purpose, we had to adjust the α to 0.00625 (0.05/8 conditions). Only the aSPOC-late comparison survived this analysis, exhibiting a significant effect of TMS (p = 0.001), which was observed behaviorally as a decreased correction magnitude of the reach component (Fig. 5A,B). The mean transport velocity trajectories for late perturbation [noTMS (blue) vs TMS (red)] and control conditions for each participant are shown in Figure 6.
Effects of TMS to aSPOC on transport responses to an unpredictable mechanical perturbation of hand transport—resisting the hand motion along the axis of the reach—applied at 300 ms after movement onset (“late perturbation”) during reach-to-grasp movements. A, Mean transport velocity trajectories when TMS was applied to aSPOC. The vertical lines denoted by “P” indicate perturbation onset (300 ms). TMS to aSPOC resulted in a smaller correction magnitude (red) than the noTMS condition (blue). B, Mean (±1SE) correction magnitude for noTMS and TMS conditions across all participants (n = 10), for each of the four brain sites. Thin gray lines indicate values for individual participants. **p = 0.001.
Effects of TMS to aSPOC on transport velocity responses to an unpredictable mechanical perturbation of hand—resisting the hand motion along the axis of the reach—applied at 300 ms after movement onset (“late perturbation”) during reach-to-grasp movements. Transport velocity trajectories for control (black), late perturbation without TMS (noTMS, blue), and late perturbation with TMS (TMS, red) conditions for each participant. Note decreased correction magnitude of transport velocity with TMS. The vertical lines denoted by “P” indicate late perturbation onset at 300 ms.
Bonferroni–Holm post hoc paired t test of transport correction magnitude for each “brain site” (aSPOC, PMd, aIPS, PMv) for “early” and “late” unpredictable mechanical perturbation of hand transport
Analysis of the effects of TMS on corrections to the aperture component
The effects of the mechanical perturbation to transport on the aperture were best characterized by a delay in the closure of the fingers in both the early and late conditions (Fig. 4E,F). To test the involvement of each cortical site in online updating of the grasp to compensate for the mechanical perturbation, two 4 × 2 × 2 rm-ANOVAs: “brain site” (aSPOC, PMd, aIPS, PMv), “timing of perturbation” (early, late), and “cortical perturbation” (noTMS, TMS) were run for each of the aperture variables (aperture correction magnitude and aperture correction latency). A significant interaction of brain site and cortical perturbation for grasp aperture correction latency (F(3,27) = 3.57, p = 0.026, Table 1) was detected. We followed the two-way interaction with a Bonferroni–Holm post hoc test for each brain site to test for the effect of TMS. This analysis involved four comparisons (each brain site/noTMS vs each brain site/TMS; Table 3). For this purpose, we had to adjust the α to 0.00125 (0.05/4 conditions). No tests survived the correction, and aIPS resulted in p = 0.036.
Bonferroni–Holm post hoc paired t test of aperture correction latency for each “brain site” (aSPOC, PMd, aIPS, PMv) for unpredictable mechanical perturbation of hand transport
Discussion
This study investigated the causal involvement of frontoparietal nodes in the online updating of reach-to-grasp movements to unpredictable mechanical perturbations of hand transport. Very few studies have employed mechanical perturbations of the reach (Haggard and Wing, 1995; Rand et al., 2004; Reichenbach et al., 2014) or the grasp (Schettino et al., 2015) during reach-to-grasp movements, and only two (Reichenbach et al., 2014; Schettino et al., 2015), to our knowledge, have employed TMS to determine the causal role of specific brain sites in humans. Our results showed a clear effect on the transport correction magnitude following a concurrent mechanical perturbation and TMS stimulation of aSPOC in the late condition (shaping phase of the grasp), pointing to a causal role for that region in the correction to a disruption of the effector state during the movement. The involvement of aSPOC/V6A in proprioceptive processing has been suggested by a number of human (Filimon et al., 2009; Vesia et al., 2010; Breveglieri et al., 2023a, 2023c) and nonhuman primate studies (Breveglieri et al., 2002). For example, Breveglieri et al. (2002) found somatosensory responses to passive manipulation and joint rotation limited to the upper limbs in region V6A in monkeys. Subsequent work by the same group (Fattori et al., 2010, 2017; Breveglieri et al., 2018) reported neuronal activity modulation in V6A during grasping movements in the absence of visual feedback. The authors concluded that through the integration of visual and motor inputs (from PMd), aSPOC might play a role in the monitoring of bodily state during reaching and grasping. In our experiment, the effect observed occurred during the shaping phase of the grasp, after peak transport velocity (PTV), when sensory input is thought to play a considerable role in the control of the movement (Elliott et al., 2017; Gallivan et al., 2018; Furmanek et al., 2019). In related work, Vesia and colleagues used repetitive TMS to disrupt the activity of SPOC, the angular gyrus (AG), and the medial intraparietal sulcus (mIPS) during a reaching from memory task and found that the endpoints of the participants’ reaching motions became more variable (Vesia et al., 2010). Critically, they found that providing visual feedback of the moving hand allowed their participants to correct their movement endpoints, but only if the stimulation was over mIPS or AG and not over SPOC. These results suggest that aSPOC processes a proprioceptive signal that cannot be recalibrated using a visual input. Given the above, it is reasonable to suggest that the DM pathway is involved in the correction of late mechanical perturbations of the reach via the monitoring of proprioceptive information. Fattori and collaborators (Fattori et al., 2017; Gallivan et al., 2018) have proposed that aSPOC/V6A may act as a state estimator receiving a corollary discharge from PMd that allows for the comparison of the ongoing motor command to incoming sensory (visual, proprioceptive) feedback. Our data confirm a role for aSPOC in the dynamic correction to proprioceptive perturbations and demonstrate that it is a necessary node in the visuomotor control of the reach-to-grasp movement.
In terms of the lack of effects observed for the other nodes in the FP networks included in this study, we note that in our early condition, the mechanical perturbation, and TMS pulses occurred 100 ms after the onset of the movement, during the initiation of the motion (before PTV). Furthermore, while information about the target object was processed visually, the perturbation was processed proprioceptively. In our previous work (Schettino et al., 2015), we tested the involvement of the DL nodes (PMv and aIPS) in a mechanical perturbation of the grasp. Our results showed a role only for PMv 50 ms following the perturbation, but no effects were observed on either node for TMS stimulation when it was concurrent with the perturbation or 100 ms after the perturbation. This suggests that during movement initiation, the precise timing at which the premotor cortices may receive and process proprioceptive information is within a relatively small window. It is possible that a TMS pulse concurrent with the perturbation during early control may have missed that window despite the use of a double pulse “virtual lesion.”
The paucity of effects at the late stage of the movement for the DL nodes (aIPS, PMv) is more surprising. Nonetheless, there is evidence that the DL nodes are involved in the processing of visual information for the control of the grasp relative to the control of the reach (Galletti and Fattori, 2018). While we did see a double peak following the mechanical perturbation of the reach, the effects of TMS on aperture velocity or magnitude did not reach significance (Table 3). It is, therefore, possible that the DL pathway, in spite of receiving proprioceptive information, may preferentially employ it to control the grasp component, which was not perturbed in this study. The work of Reichenbach et al. (2014) supports this notion. In that work, the authors looked at the role of aIPS and mIPS in correcting for proprioceptively sensed perturbations during a reaching task. Participants held a robot arm that presented random force perturbations to their reaching arm as they transported their hand toward visually defined targets with or without visual feedback of hand position. Their results showed that TMS to aIPS increased the deceleration time of the movement while hand feedback remained available. However, when visual feedback was removed, TMS to mIPS but not aIPS disrupted endpoint accuracy. The authors suggested that while mIPS appeared to process proprioceptive information during the reach, aIPS processed more general sensory information. Similarly, our previous study using a grasping perturbation found that aIPS-TMS did not result in modifications of the grasp. The evidence supports the notion that area aIPS may be more related to the determination of object properties through visual input, relying on more dorsal and medial regions of the PPC for the online control of the reach and the grasp through proprioceptive means.
Given the clear effect of TMS on aSPOC during the response to the mechanical perturbation, it was surprising to note a lack of effect for PMd. Current thinking regarding the role played by the DM pathway based on nonhuman primate studies (Fattori et al., 2017; Galletti et al., 2022) suggests that during an unexpected change in effector state, aSPOC and PMd interact to calculate the magnitude of the motor error to implement a modified response. Further studies testing the specific contributions of the DM nodes may clarify their roles.
Conclusion
Mechanical perturbations of bodily state are relatively common during natural behavior, underscoring the importance of describing the specific roles of the frontoparietal networks for the understanding of normal and disordered behavior. Our study is the first to test the causal role of FP pathways in the correction of mechanical perturbations of arms transport during a reach-to-grasp movement in humans. We found evidence for the involvement of aSPOC in the modulation of the response to the perturbation during the later stage of the movement (during hand preshaping), suggesting that this area participates in the estimation of the effector state based on proprioceptive input. This evidence supports previous findings regarding proprioceptive processing in aSPOC/V6A (Fattori et al., 2017; Breveglieri et al., 2018, 2023a; Galletti and Fattori, 2018) and extends them by showing that this area is necessary for the compensatory response observed following a proprioceptively defined perturbation of the reach.
Footnotes
This work was supported by NIH Grants R01NS085122 (E.T. and S.V.A) and R01HD058301 (S.V.A. and E.T.), and National Institue on Daily, Independent Living, and Rehabilitation Research–funded Rehabilitation Engineering Research Center Grant 90RE5021 (S.V.A.). We thank Alex Huntoon and Samuel Berin for their work on the virtual reality platform that enabled this work. We also thank Sambina Anthony for helping with data collection and preparing Figure 1. We thank Dr. Sumientra Rampersad for helping with fine element head models and computing the transcranial magnetic stimulation–induced electric fields in the brain.
↵*M.P.F. and L.F.S. contributed equally to this work.
The authors declare no competing financial interests.
- Correspondence should be addressed to Mariusz P. Furmanek at mariusz.furmanek{at}uri.edu.
