Abstract
The middle temporal (MT) area of the extrastriate visual cortex has long been studied in adulthood for its distinctive physiological properties and function as a part of the dorsal stream, yet interestingly it possesses a similar maturation profile as the primary visual cortex (V1). Here, we examined whether an early-life lesion in MT of marmoset monkeys (six female, two male) altered the dorsal stream development and the behavioral precision of reaching-to-grasp sequences. We observed permanent changes in the anatomy of cortices associated with both reaching (parietal and medial intraparietal areas) and grasping (anterior intraparietal area), as well as in reaching-and-grasping behaviors. In addition, we observed a significant impact on the anatomy of V1 and the direction sensitivity of V1 neurons in the lesion projection zone. These findings indicate that area MT is a crucial node in the development of primate vision, affecting both V1 and areas in the dorsal visual pathway known to mediate visually guided manual behaviors.
SIGNIFICANCE STATEMENT Previous studies have identified a role for the MT area of the visual cortex in perceiving motion, yet none have examined its central role in the development of the visual cortex and in the establishment of visuomotor behaviors. To address this, we used a unilateral MT lesion model in neonatal marmosets before examining the anatomic, physiological, and behavioral consequences. In adulthood, we observed perturbations in goal-orientated reach-and-grasp behavior, altered direction selectivity of V1 neurons, and changes in the cytoarchitecture throughout dorsal stream areas. This study highlights the importance of MT as a central node in visual system development and consequential visuomotor activity.
Introduction
Vision relies on a multiplicity of areas within the brain to communicate synergistically to produce an accurate percept. The seminal works of Goodale and Milner (1992) and Ungerleider and Mishkin (1982) proposed the existence of two neocortical pathways that have a dichotomous visual functional specialization. More specifically, the dorsal stream pathway is involved in perceiving motion and vision for action, and the ventral stream pathway is associated with shape/object perception and vision for perception. Central to the dorsal stream is the middle temporal (MT) area (Fig. 1A,B). Based on established anatomic markers of neural circuit maturation, area MT matures early and in parallel with primary sensory areas, including the primary visual cortex (V1; Bourne and Rosa, 2006). Further, because the monosynaptic V1-MT connection is yet to be fully established during this process, the early maturation of MT is thought to be independent of V1 input (Warner et al., 2012). Together, these findings and the presence of disynaptic retinal inputs to MT via the thalamus (Sincich et al., 2004; Warner et al., 2010) support the notion that it is a primary-like area and serves as an anchor early in life to support the establishment of the dorsal stream (Rosa, 2002).
Marmosets used in the current study
In adulthood, both the marmoset and macaque MTs receive significant input directly and indirectly from V1 (Ungerleider and Mishkin, 1979; Mundinano et al., 2019b) and are interconnected with multiple dorsal stream extrastriate areas, including V3 and the dorsomedial (DM; area V6 in Old World monkeys and humans area, hosts many direction-selective neurons like MT and plays a central role in motion recognition and perception) area, the anterior intraparietal (AIP) and lateral intraparietal (LIP) areas (Maunsell and van Essen, 1983; Krubitzer and Kaas, 1990; Colby et al., 1993; Lyon and Kaas, 2001), and the frontal eye fields (FEF; Maioli et al., 1983; Majka et al., 2020; Fig. 1A). Area MT also has abundant connections with subcortical structures, including (1) reciprocal connectivity with the medial portion of the inferior pulvinar (Lin and Kaas, 1980; Warner et al., 2010, 2012), (2) input from the koniocellular layers of the lateral geniculate nucleus (LGN; Sincich et al., 2004), and (3) output to the superior colliculus (Maunsell and van Essen, 1983; Ungerleider et al., 1984; Lyon et al., 2010).
A substantial proportion of neurons within MT are highly tuned for direction (Dubner and Zeki, 1971; Albright, 1984), which is a clear distinction from V1, suggesting that MT plays a central role in motion perception and integration of visual information. However, few studies have interrogated how injury to MT impairs visual perception and global visual behavior. Newsome et al. (1985) and Dürsteler et al. (1987) provided behavioral evidence of how MT contributes to motion perception following mechanical ablation of the MT, while motion blindness/akinetopsia has been observed clinically in a patient with a bilateral MT injury (Zihl et al., 1983; Zeki, 1991). Further, others have demonstrated that V2 direction-tuning properties are dependent on MT (Jansen-Amorim et al., 2011).
In the phenomenon of blindsight, the extensive multiplexed circuitry of the dorsal stream, with a central component being area MT, has been long implicated as the neural substrate that provides the ability to perform specific visually driven tasks in the absence of V1(Weiskrantz et al., 1974; Mundinano et al., 2019a). A more specific investigation revealed that MT remains active when visual stimuli are presented within the scotoma (Leh et al., 2010; Tran et al., 2019).
Few studies have taken into account the role of MT in the development of the dorsal stream and establishment of specific visuomotor behavior, but developmental studies have inferred that the MT acts as the primary node in the establishment of the dorsal stream (Bourne and Rosa, 2006; Mundinano et al., 2015). Further, early-life lesions of the geniculostriate pathway have highlighted the conserved integrity of MT in both humans and monkeys (Warner et al., 2015). Therefore, we hypothesized that an early-life lesion in MT will permanently perturb reaching-to-grasp behaviors and result in dysfunction of areas directly connected with MT.
Materials and Methods
Subjects
Five New World marmoset monkeys (Callithrix jacchus) received a mechanical ablation of the MT area at postnatal day (P)14 (Table 1). Following surgical ablation, the animals were allowed 12 months of recovery before undergoing training for visual behavior experimentation. Three aged-matched animals were used as controls for behavioral experiments (Table 1). All experiments were conducted in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes. All procedures were approved by the Monash University Animal Ethics Committee, which also monitored the health and well-being of the animals throughout the experiments.
Mechanical ablation of area MT
Neonatal marmosets (P14, n = 5) were anesthetized with isoflurane (1–2% in medical oxygen at 0.5 L · min−1) and placed into a custom-built nonferromagnetic stereotaxic frame to facilitate T2-weighted MRI to allow for demarcation (dorsal shift in layer 4; Fig. 1C) and targeted ablation of area MT. Full details of the preparation of the animals and how to localize areas of interest with MRI guidance have been described previously (Mundinano et al., 2016). Scans were exported as Digital Imaging and Communications in Medicine (DICOM) files, and the brain images were visualized using Horos version 3.3.5 software (https://horosproject.org). Once MT was visualized, ablations were performed using a biopsy punch (diameter = 3.0 mm). Prior calculation of the depth of the cortical sheet at the site of tissue ablation assisted with limiting damage to the underlying white matter. T2-weighted MRI scans at 6 weeks confirmed intact white matter, which was observed to be degenerating by 36 weeks (Fig. 1C) and histologically (Fig. 1D). Following ablation of cerebral tissue, the cavity was filled with Gelfoam, the cranium was reconstructed, and the scalp was sutured. Animals received postsurgical antibiotic and analgesic medication and were then allowed a minimum of 12 months for recovery to allow for plastic reorganization of the brain.
Longitudinal MR imaging
During the animal's recovery, structural T2 images were acquired on a 9.4T small-bore animal scanner (Agilent) at 6 weeks and 36 weeks postlesion to identify the extent of the lesion (Fig. 1C). The parameters for the T2 scan were the following: MRI structural comprised of T2-weighted images [9.4 Tesla 18 cm bore MRI scanner (Bruker)], 2D Rapid Acquisition with Relaxation Enhancement (RARE) sequence, repetition time (TR) = 12,000 ms, echo time (TE) = 48 ms, matrix size = 200 × 180, field of view (FOV) = 40× 36 mm2, number of slices = 80, slice thickness = 0.4 mm, RARE factor = 8, axial plane, four averages, scan time 30 min. Additionally, at the same time, diffusion tensor imaging (DTI) was acquired to analyze changes in microstructure in the visual cortical tissue following injury. DTI was acquired in the axial orientation using a six-shot spin-echo echo-planar imaging (EPI)-based DTI sequence. Scan parameters included the following: TR/TE of 3000/31 ms, FOV of 38.4 × 38.4 mm2, data matrix of 96 × 96, and 50 slices with a thickness of 0.4 mm. Diffusion encoding gradients were applied in 30 directions, with a b value of 800, gradient duration of 5 ms, and separation of 18 ms. A reference scan with inverted phase and readout gradients was acquired for phase correction to reduce EPI and DTI-specific artifacts. The average of the reference and the original scan also increased the signal-to-noise ratio (SNR). The DTI sequence was repeated four times and saved separately with the aim of further increasing the SNR and generating backup data in case any unexpected motion artifacts occurred during the relatively long scan. The whole DTI scan took ∼80 min.
Diffusion-weighted data were preprocessed and analyzed using tools from different software packages that work well with the small-size marmoset brain. MRIcron dcm2nii script (http://people.cas.sc.edu/rorden/mricron/dcm2nii.html) was used for conversion of DICOM to NIfTI files; averaging the orientation of different DWI repetitions and corrections for movement distortions was scripted using Functional MRI of the Brain Software Library (FSL version 5.0.9; https://fsl.fmrib.ox.ac.uk/fsl). Binary brain masks were generated using BrainSuite version 15c (https://www.brainsuite.org). Diffusion tensors and derived fractional anisotropy (FA) maps were calculated using MRtrix3 (https://www.mrtrix.org). To identify and delineate the boundaries of cortical areas showing FA changes, we used the Brain/MINDS 3D digital marmoset brain atlas (Woodward et al., 2018). The images containing the significant changes in FA values were coregistered to the space of the marmoset atlas using the registration tool ITK-SNAP (https://www.itksnap.org/; Yushkevich et al., 2006) and the Advanced Normalization Tools package (Avants et al., 2009).
Visual behavior training
Visual behavior experiments commenced following a minimum of 12 months of recovery from the MT ablative surgery. Detailed stepwise training and habituation protocols to implement reach-and-grasp behavioral experiments in the marmoset have been previously described (Fox et al., 2019).
Animals were first trained to enter a custom fabricated behavior training box. Food rewards were given to encourage participation in desired behaviors positively. Following habituation with the behavior box, the animal was transferred to an adjacent room where the behavioral experimentation was conducted. Food rewards were once again provided to habituate the animal to the room.
Following complete habituation, the ability of the animal in the goal-oriented reach and grasp of static objects was examined, whereby the animal was presented with a target object in the form of a food reward (piece of fruit). The food reward was placed in one of two positions on a pedestal, both of which were equidistant to an aperture in the behavior box (Fig. 2A,B). These two positions could favor either the left or right hand. To determine the hand preshaping behavior of the animal and whether the size of the target object would have a significant effect on performance, the food reward was cut into cuboids with the largest face ∼5 × 5 mm or ∼10 × 10 mm. The precise size of the food reward was qualified during the video analysis of each trial with built-in tools in Tracker software. Before commencement of a static task trial, a blind was placed in front of the transport box and its apertures so the animal could not prime the hand to the position of the reward in anticipation. Each static object retrieval session consisted of 15–20 trials. The proportion of trials of small- versus large-object presentation was equal.
To assess the reaching and grasping capabilies of moving objects within our subjects, a single-size food reward (cuboid face ∼10 × 10 mm) was placed on the edge of a custom-built rotating turntable (ø = 14 cm; Fig. 3A). The center of the turntable was aligned with the center of the transport box, equidistant to both right and left apertures. Objects were consistently placed in the same spot of the turntable. Stops were used to occlude the apertures to control the use of a specific hand (i.e., the animal would use the right hand when the right aperture was unblocked and vice versa). The stop was removed, and the trial commenced once the food reward had undergone one revolution to ensure that the turntable can accelerate to the desired speed for the task. Animals were given a 10 s response time for each trial. Animals were allowed a maximum of three attempts within each trial for it to be considered successful. Background white noise was played throughout the sessions to minimize any impact the noise produced by the turntable would have in the performance by the animal.
The following independent variables were incorporated into the turntable task: (1) six different revolution speeds ranging from 10 to 60 revolutions per minute (RPM), (2) the direction of the revolution (clockwise/anticlockwise), and (3) the hand used to retrieve the object (Table 2). This resulted in 24 different conditions that would be presented during a training session. The order of the trial conditions was randomized across sessions, and a single condition (e.g., object moving at a speed of 30 RPM in a clockwise direction with the aperture opened for the left hand) was not repeated in the same session.
Independent variables used in moving reach-and-grasp task
Analysis of visual behavior performance
Trials for the static object task were binned into three main categories; a successful trial was defined as a single coordinated motor action to retrieve the object. Corrected trials occurred when animals required fewer than three corrections in either their reach or grasp to retrieve the object successfully. Failure to meet the conditions of the two previous bins, or not completing the action within the given time frame, was assigned to the failed trial category. Trials for the moving object task were binned into two main categories; a successful trial was when the animal retrieved the rotating object within three attempts, a failed trial was when more than three attempts were required for retrieval or the object was dislocated from the initial position. Trials with no response were considered invalid and were discarded.
All trials were recorded using a GoPro Hero4 Black camera (1280 × 790 pixel resolution, 120 frames/s, narrow view) from the transverse plane perpendicular to the plane of motion (i.e., top-down view), and videos captured were analyzed off-line. The same investigator (C-K.C.) performed and analyzed all videos off-line manually in a blinded manner to determine the performance and kinematics of each trial. The performance as expressed by error rate was determined by dividing the number of failed trials by the total number of trials within a given session. The overall performance of a given task in each animal was determined with permutation testing of the performance of all sessions.
Video files were transferred to an open-source software, Tracker version 4.93 (https://physlets.org/tracker/), designed to perform a 2D kinematic analysis. A reference scale of known size remained in the field of view and was positioned at the height of the reaching movements to eliminate any perspective errors. The edges of the reaching platform served as reference points for x- and y-axes to track movement trajectories across each trial. Three metrics were measured in the video analysis: the maximum grip aperture (MGA) between the animal's thumb and index finger when the animal preshapes the hand in acquiring an object, the velocity at which the animal would move to grasp an object, and the acceleration. Additionally, the junctions between the nail and digit of the index finger and thumb were used as reference points to measure the Euclidean distance over time. Movement duration was defined as time elapsed from the moment the thumb reference point was visible outside the reaching aperture in the box to the moment when either the index finger or thumb made contact with the piece of fruit (identified for at least three consecutive frames). The length of the food reward in each trial was measured against the reference scale and plotted against the MGA for the trial using Prism software version 8.4.2, to determine whether the animals exhibited any hand preshaping behavior. All failed trials were further sorted to determine whether the animals exhibited a premature or delayed reach while attempting to retrieve the food reward. The angle between the final hand position over the turntable and the food reward was then measured on Tracker version 4.93 to determine the accuracy of targeting the object (Fig. 3C). Mean angulation was calculated in each session for each animal to reveal the overall angulation of failed trials across all sessions (p ≤ 0.05 was considered to be statistically significant; all data are presented as mean ± SEM).
Electrophysiological recordings
Animals M1695 and F1696 underwent a single recording session under deep anesthesia. The detailed methodology of how the animals were prepared for electrophysiology is described in Bourne and Rosa (2003). In brief, animals underwent a tracheotomy and two craniotomy procedures with anesthesia maintained with a continuous intravenous infusion of a mixture of pancuronium bromide (0.1 mg · kg−1 · h−1), sufentanil (6 µg · kg−1 · h−1), and dexamethasone (0.4 mg · kg−1 · h−1), in a saline/glucose solution. During the recordings, animals were also ventilated with nitrous oxide and oxygen (7:3).
The electrophysiological experiments consisted of two phases. First, mapping of receptive fields in the MT complex was used to define the perimeter of the V1 scotoma. Following identification of the scotoma, recordings were conducted in V1 within and outside the lesion projection zone (LPZ). Recordings were performed with tungsten microelectrodes (∼1 MΩ). Insertion of the electrodes to the eccentricity of interest was guided by stereotaxic and topographical data described in previous studies (Fritsches and Rosa, 1996; Mundinano et al., 2019b). Amplification and filtering were achieved via an A-M Systems Model 1800 Microelectrode Alternating Current Amplifier. Stimuli presentation and methods for quantification of tuning for orientation and direction and spatial and temporal frequency within V1 were performed as per previous studies (Yu et al., 2010, 2013).
Histology and immunohistochemical tissue processing
Animals were killed with an overdose of sodium pentobarbitone (>100 mg/kg) then transcardially perfused with 10 mm PBS that had been supplemented with heparin (50 IU/ml of PBS), followed by 4% paraformaldehyde in 10 mm PBS. Brains were postfixed in 4% paraformaldehyde then dehydrated in serial solutions of PBS-sucrose before being snap frozen in 2-methylbutane chilled to −50°C and stored in a −80°C freezer until cryosectioning.
Brain samples were cryosectioned in the coronal plane at 50 um, divided into five series, and stored free-floating in a cryoprotective solution as described previously (Bourne and Rosa, 2006).
Half a series was stained for myelin using the silver impregnation (Gallyas, 1979) to demarcate area MT and to qualify the location and size of the ablative lesion (Fig. 1C). Free-floating sections were labeled with GABAergic interneuronal marker parvalbumin (PV; 1:3000; PV27, Swant).
Microscopy and digital image processing
Brain sections were imaged with an Axio Imager Z1 microscope (Zeiss). Images were obtained with a Zeiss Axiocam HRm digital camera using AxioVision software (version 4.8.1.0) at a resolution of 1024 × 1024 or 2048 × 2048 pixels and saved in Zeiss Vision Image and exported to TIFF format. The objectives used were Zeiss EC Plan-Neofluar 5× 0.16 (catalog #420330-9901) and EC Plan-Neofluar 10× 0.3 (catalog #420340–9901). Filter sets used for visualizing immunolabeled cells were Zeiss 38 HE eGFP (catalog #489038-9901-000).
Images used for cell density quantification were taken with the 10× objective. Stitching of images and adjustments to contrast and brightness were performed using Adobe Photoshop 2020 version 21.0.1 or Zeiss MosaiX software. The contours, boundaries, and line art of all figures were drawn using Adobe Illustrator 2020 version 24.0.1.
Experimental design and statistical analysis
Behavioral data were collected from two nonlesioned control animals and two MT-lesioned animals. Each animal undertook goal-orientated reach-and-grasp tasks to permit interrogation of visuomotor capabilities. Assessment of the behavioral performance of an animal was determined through the error rate in each session. The error rate and the frequency of successful retrieval on the first attempt were examined using a nonparametric permutation test (Microsoft Excel version 16.62). The kinematic analysis of reach-and-grasp behaviors, including MGA, velocity, acceleration, and angulation, was also examined using the same nonparametric permutation test. Data were repeatedly resampled within the permutation test (5000 times) to determine any significance between lesion and control cohorts (p ≤ 0.05 was considered statistically significant).
To determine whether the food reward size influenced performance, an ANCOVA was performed to compare the error rate of large and small sizes across sessions (Prism version 8.4.2). Linear regression was also performed to determine the hand preshaping behavior against food lengths (Prism version 8.4.2; p ≤ 0.05 was considered statistically significant).
For voxel-based morphometry statistical analysis of FA maps, a general linear model following a two-sample t test was used comparing left (lesioned) and right (contralesional) hemispheres (SPM12, Wellcome Trust Center for Neuroimaging) as follows. FA maps from all MT-lesion animals were registered to a marmoset brain template (http://brainatlas.brain.riken.jp/marmoset/; BSI-Neuroinformatics, Riken) using the normalized mutual information function (7-degree B-spline). The resulting images were smoothed with 1 mm isotropic Gaussian kernel. Subsequently, images were flipped and registered again to original unflipped images. The significance level was p < 0.05, corrected for the familywise error rate with an extended threshold of 1 voxel; z-scores converted to p values were displayed on the Brain/MINDS 3D digital marmoset atlas (Woodward et al., 2018) using Mango 4.1 software (Research Imaging Institute; ric.uthscsa.edu/mango).
Quantification of interneuronal density was conducted as described previously (Mundinano et al., 2015). PV cell density within layers 2 and 3, 4, or layers 5 and 6 was determined within each area of interest. For each animal, six immunolabeled sections were randomly selected across the anterior–posterior axis across our areas of interest. The areas sampled were V1; intraparietal areas AIP, LIP, and MIP; and posterior parietal (PE) area (Rosa et al., 2009; Paxinos et al., 2012).
Thresholding and density quantification in photomontages of PV immunoreactive cells was conducted using Fiji image software (Schindelin et al., 2012). A nonparametric Mann–Whitney U test examined comparison between the ipsilesional and contralesional V1 (p ≤ 0.05 was considered statistically significant; Prism version 8.4.2 software). The same investigator (W.C.K.) performed all quantifications in a blinded manner.
Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Results
We used early postnatal ablations to study the role MT serves in the development of the visual cortex and visually guided behaviors. At 6 weeks postsurgery, MRI T2-weighted images were acquired to validate unilateral biopsy punch ablation of MT in the five neonatal (P14) marmosets (Fig. 1C). Before lesion, the location of area MT was visualized in MRI images (T2 weighted) by a dorsal shift in layer 4 from adjacent areas (Fig. 1C, inset). During the excision of MT tissue, particular care was taken to remove all six cortical layers while leaving the underlying white matter tract intact. When the animals subsequently reached young adulthood (>36 weeks) and underwent DTI scans for evaluation of affected cortices, another T2 sequence was obtained to map the extent of the lesion (Fig. 1B). In these scans, we often observed that the lesion core had scar tissue, and subtle degeneration of the underlying white matter had occurred by 36 weeks, which was not present at the 6 week point. Following perfusion of the cerebral tissues, histologic confirmation of the lesion was achieved with a Gallyas (myelin) silver stain (Fig. 1D), and the lesion was entirely reconstructed to establish its topographical extent for each animal (Fig. 1E).
Monosynaptic connectivity of the MT area of the visual cortex and unilateral early-life ablative lesion approach. A, Wiring diagram detailing monosynaptic reciprocal connectivity of area MT and proposed behavioral function. B, Lateral view of the left hemisphere of the marmoset cortical surface highlighting anatomic positioning of areas in A. C, T2-weighted MRI of a marmoset brain at the level of the parafoveal MT, which was targeted for ablation at P14. MT was demarcated based on the dorsal shift in layer 4 (left, inset, red asterisk). In the anterior/posterior plane, the core of MT was denoted as when the most caudal aspect of the corpus callosum becomes visible (arrow). Further T2-weighted scans were conducted at 6 weeks and 1 year postinjury to confirm lesion location and also DTI for fractional anisoptropy analysis (Fig. 5, DTI results). Scale bar, 5 mm. D, Histologic analysis with myelin silver staining for ex vivo demarcation of MT and localization of the lesion in all five cases. Anterior/posterior positioning of the lesion is referenced against the coordinates outlined in the Paxinos marmoset brain atlas (Paxinos et al., 2012). Scale bar, 2 mm. E, Summary of the extent and retinotopic relationship of MT lesion in each lesioned lesion animal. Squares represent the horizontal meridian. Circles represent the vertical meridian. Dashed lines and associated numbers represent the eccentricity from the fovea in degrees. Solid lines represent area boundaries. Each 2D representation was created by reconstructing area MT and the lesion from histologic sections in D. Scale bar, 2 mm. Details of the experimentation each animal underwent are shown in Table 1. FEF, frontal eye fields; PE, area PE; MIP, medial intraparietal area; AIP, anterior intraparietal area; V1, primary visual cortex.
The absence of MT during development perturbs reach-and-grasp kinematics as well as interaction with static and moving objects
As MT neurons are highly tuned for direction and the perception of motion (Rosa and Elston, 1998), we wanted to determine whether accurate reward-driven visuomotor behavior was still achievable following the early-life loss of MT. To ensure adequate compliance with the task, we first trained the animals to enter a removable transport box on the home cage, which was then taken into the behavior laboratory (Fig. 2A; Fox et al., 2019). We used naturalistic, goal-oriented, reach-and-grasp tasks once neonatal MT lesion animals (P14) had reached adulthood (>18 months). These were undertaken both statically (Fig. 2B) or on a variable speed/direction turntable (Fig. 3A,B). Each food retrieval trial was 2D video captured (Figs. 2B, 3B), which allowed us to study the animal's proficiency in executing the task and the kinematics of the reach-and-grasp phases.
Early-life MT lesions perturbs goal-orientated reach-and-grasp performance in static objects. A, Experimental setup of reach-and-grasp tasks. The animals are presented with a static or moving object retrieval task. B, Example of 720 pixel footage of an animal successfully grasping a static reward. C–D, A small (∼125 mm3) and large (∼1000 mm3) food reward was first used to determine whether the object size affected the ability of the animal to retrieve the static object. Object size had no significant effect on the ability of the control cohort to retrieve the reward (ANCOVA, F(1,52) = 0.13, p = 0.71; C). The size of the reward object also had no significant effect on the lesion cohort in object retrieval (ANCOVA, F(1,52) = 0.68, p = 0.41; D). E, Performance (error rate) of control versus lesioned animals in reaching and grasping static objects. Lesioned animals performed significantly worse (permutation test, *p = 0.04). F–H, Kinematic analysis of the animal's reward retrieval abilities and the frequency distribution across all static object trials for control and lesioned animals was examined. Lesioned animals exhibited a larger MGA in their grasp (F) and reach with greater velocity (G) and acceleration (H).
Animals were first presented with a static task, which involved presenting a food reward that was placed on a pedestal in front of the animal (Fig. 2B). Food rewards were cut into two cuboid sizes (small face ∼5 × 5 mm, large face ∼10 × 10 mm) and were first used to determine if the object size would affect overall performance. No significant difference in retrieval performance was observed between the two sizes of the fruit reward in both control (ANCOVA, F(1,52) = 0.13, p = 0.71) and lesioned cohorts (ANCOVA, F(1,52) = 0.68, p = 0.4; Fig. 2C,D).
When examining the collective retrieval attempts of both sizes of the statically presented food reward, we found the overall performance within the MT-lesioned group was significantly impaired when compared with controls (permutation test, p = 0.04*; Fig. 2E). Further analysis of the reach-and-grasp kinematics revealed significant differences in hand preshaping and reaching behavior. Specifically, during the preshaping phase, neonatal MT-lesioned animals exhibited a significantly larger maximum grip aperture than controls (permutation test, p = 0.0037; Fig. 2F). Further characterization of the behavior revealed neonatal MT-lesioned animals exhibit a feedforward movement with greater velocity (permutation test, p = 0.04; Fig. 2G) and acceleration (permutation test, p = 0.02; Fig. 2H) when approaching the target object.
To interrogate the ability of the animal to accurately engage a moving object in a naturalistic setting, animals were again required to reach and grasp a food reward, but in this instance the reward was placed on a variable speed/direction turntable in front of the transport box (Fig. 3A–C; Movie 1). A difference in the object retrieval error rate was observed between the two cohorts, with the neonatal MT-lesioned group having an increased error rate (permutation test, p = 0.042; Fig. 3D). Further analysis of reach-and-grasp behavior within the lesion cohort revealed they were less coordinated and consequently had less success in retrieval of the reward on their first attempt (permutation test, p = 0.023; Fig. 3E). When lesioned subjects were unsuccessful in retrieving their food reward, this was largely because of the premature execution of their reaching action (Fig. 3F). When lesioned animals failed in retrieval of moving objects, the magnitude of their failure (angle between the moving object and the grasp attempts made over the rotating table) was greater compared with controls (permutation test, p= 0.003; Fig. 3G). The MGA, peak velocity, and acceleration observed between both cohorts for the moving task were similar (Fig. 3H–J).
Visuomotor performance in reaching to grasp moving objects is impaired following early-life loss of MT. A, Experimental setup for the moving object reach-and-grasp task. B, B', Example of 720 pixel footage of an animal successfully reaching for and grasping a moving reward. C, An example trial where the animal was late in executing its reach for the food reward and failed in the moving object reach-and-grasp trial, the magnitude of which was calculated based on the angle between hand (touch of plate) and object. D, Performance of control versus lesioned animals in reaching and grasping moving objects. Lesioned animals performed significantly worse (permutation test, *p = 0.042). E, Control animals were also observed to more often succeed in retrieving a moving object on their first attempt (permutation test, *p = 0.023). F, Failed trials within the lesioned cohort were largely because of the animal reaching prematurely. Control animals revealed close to a 1:1 proportion of premature and late reaching actions in failed trials, whereas MT-lesioned animals exhibited a tendency to reach prematurely. G, The average magnitude of error as denoted by angulation in the failed trials. When lesioned animals failed in reaching and grasping moving objects, the magnitude of the missed grasp was greater than controls (permutation test, **p = 0.003). H–J, Kinematic analysis of the animal's retrieval abilities and the frequency distribution across all moving object trials for control and lesioned animals was examined. Control and lesioned animals exhibited comparable MGA in their grasp (H), reach velocity (I), and acceleration (J).
Moving object task: successful and failed reach-and-grasp attempts. An example showcasing successful and failed attempts of an animal when undertaking the moving object task.
V1 responses are altered following an early-life lesion in MT
As there is a significant reciprocal monosynaptic connectivity between MT and V1, we wanted to observe whether an early-life lesion of MT had an impact on the response properties of neurons in the LPZ of the ipsilesional V1 in adulthood. Visually evoked single-unit activity was recorded in the LPZ of V1 from two animals (M1695 and F1696; Fig. 4A), with a total of 122 units recorded. Visual stimuli were presented to determine the orientation and direction selectivity (DS) as well as optimal spatial and temporal frequencies for each of the units.
Tuning properties of V1 neurons following unilateral MT lesion. A, MT receptive fields were mapped by recording on the periphery of the lesion core. Following receptive field mapping, the scotoma is delineated, and single unit recordings of V1 neurons were performed within the MT-V1 LPZ (see above, Materials and Methods). B, The orientation selectivity of all recorded V1 neurons as determined by the HWHH bandwidth (median HWHH = 29.5° from 122 units). C, Direction index of the 122 V1 units recorded within the LPZ. A direction index of >0.5 suggests the neuron is directionally selective. D, E, Only 10 of 122 (∼8%) of neurons were tuned for direction. V1 neurons within the LPZ responded optimally to a spatial frequency (SF) of 0.7 cycles/° (D) and temporal frequency (TF) of 2.3 Hz (E).
Our first observation was that the basic retinotopic organization of the V1 was unaffected by the neonatal V1 lesion (Fig. 4A), which was expected. Our single-unit recordings revealed that 86.1% (105 of 122) of V1 neurons within the LPZ were found to be orientation selective. The half-width-half-height (HWHH) bandwidth, which is a traditional measure for orientation selectivity (Henry et al., 1973; Rose and Blakemore, 1974) within the MT-lesion-affected V1, was observed to be 29.5° (Fig. 4B). This is comparable to the median HWHH bandwidth of 29.0°, which has been reported previously for normal V1 neurons throughout the entire visual field (Yu and Rosa, 2014). The DS of a cell was determined by calculating a direction selectivity index (Ringach et al., 2002; Yu and Rosa, 2014), whereby an index of >0.5 suggests the cell is tuned for direction. Previous reports within the marmoset V1 demonstrate that ∼20% of V1 neurons are direction selective. Our findings were lower than this, with <10% (10 of 122) classified to be direction selective (Fig. 4C). Although the proportion of DS can vary with eccentricity, that is, central vision, 18% DS; near periphery, 26% DS; and far periphery, 20% DS (Yu and Rosa, 2014), our observations of ∼10% fell below this previously reported proportion.
Neurons in V1 within the two subjects responded optimally to a spatial frequency of 0.7 cycles/° (Fig. 4D) and a temporal frequency of 2.3 Hz (Fig. 4E). Previously determined values for optimal spatial frequency was 1.08 cycles/° centrally and 0.45 cycles/° in the near periphery, with an optimal temporal frequency between 3.0 and 4.0 Hz across the entire visual field. Therefore, the spatiotemporal sensitivities appeared unaffected by an early-life MT ablation.
Together, the basic retinotopic layout and physiological properties of V1 neurons appeared unaffected by the early-life MT lesion, with one exception. Namely, across the recorded areas, the proportion of directionally selective neurons was lower than previously reported, suggesting that MT may play a role in shaping these responses.
Early-life lesions in MT lead to widespread changes in cortical architecture
An advantage of our early-life lesion model is that it allows us to perform longitudinal MR imaging following the lesion. We performed whole-brain DTI scans on our lesioned animals at 36 weeks to determine the extent of chronic disruption to the cortical architecture. Voxel-based morphometric analysis of the DTI maps between the lesioned and nonlesioned hemisphere revealed a reduction in FA in the intraparietal cortical areas AIP and MIP, posterior parietal area PE (dorsal stream areas), the FEFs, and V1 (p value <0.05; Figs. 1A, 5A,B) areas that MT has strong monosynaptic connections with (Abe et al., 2018). This result suggests an altered state in the neuroanatomy of these areas, where a reduction in FA has been correlated with both axonal and gray matter damage (Huisman et al., 2004; Shereen et al., 2011).
DTI highlights the extent of microstructural changes to the cortex following a unilateral MT lesion. A, B, Voxel-based morphometry of FA maps. Yellow-red clusters represent significant hemispheric differences in FA, which were lower and observed in the ipsilesional primary visual cortex, intraparietal cortex, posterior parietal cortex, and frontal eye fields (left vs right, n = 5). Vertical lines (α, β, γ, δ) represent coronal sections in B.
As we previously demonstrated, there is a correlation between changes in FA and remodeling of the local circuitry (Mundinano et al., 2018). Therefore, we examined more closely the cellular neuroanatomy in the areas affected by the lesion in MT. Specifically, quantification of the calcium-binding protein PV-positive interneurons was undertaken as PV+ neurons have garnered significant interest as a component in the maturation of the neocortex and are functionally capable of amplifying circuit function (Hendrickson et al., 1991; Cardin et al., 2009; Sohal et al., 2009). When compared with the nonlesioned hemisphere, PV+ cell density within the ipsilesional hemisphere was reduced in V1 (Mann–Whitney U test, layers 2/3, p = 0.0008; layers 5/6, p = 0.046) and PE (Mann–Whitney U test, layers 5/6, p = 0.0029; Fig. 6A,B). Although there appears to be a reduction in PV+ cells in the other layers of V1 and PE, as well as in the IPC in the lesion animals, these were not significant.
Altered cellular maturation of visuomotor areas as revealed by PV fluorescent immunohistochemistry. A, Immunofluorescent labeling of PV+ interneurons in regions implicated to have undergone microstructural perturbations, as revealed by DTI (Fig. 5). Scale bar, 200 µm. B, Scatter plots reveal significant reduction in PV neuronal density in infragranular layers of PE (p = 0.0029), supragranulayer layers V1 (***p = 0.0008) and infragranular layers of V1 (*p = 0.046) as well as infragranular layers of PE (**p = 0.0029).
Discussion
This study was conceived with the objective to determine how MT contributes to the development of the dorsal stream. Further, an early-life unilateral lesion of MT would enable us to determine the lifelong behavioral deficits associated with loss of this central area of the dorsal stream and the impact this has on upstream and downstream cortices. Clinical studies that examined the consequence of losing V1 in early life versus adulthood demonstrated a marked difference in behavioral performance (Weiskrantz et al., 1974; Leh et al., 2010). Patient BI, who suffered a bilateral V1 injury at 9 d of age, retains the ability to perceive color and possesses significant conscious vision. When comparing the visual abilities of BI to an extensively studied blindsight patient, such as GY, who sustained an injury to V1 at the age of 8 years, BI possesses more remarkable preservation of vision. Because there is a strong correlation between the age when an injury is acquired and the preservation of abilities that neuroplasticity offers, we wanted to explore the effects an MT lesion would have on visuomotor behavior in a developing visual system. Additionally, we wanted to determine whether there are behaviors and brain areas that critically depend on MT. From our results, it is clear that an early-life lesion in MT significantly affects visual system behavior and architecture and provides evidence that area MT is crucial in the normal establishment of the dorsal stream network.
Seminal works studying the effect of lesions in MT on visual behavior have unfortunately only looked at adult-acquired loss. Newsome and Paré (1988) provided awake behavioral evidence in the macaque that sensitivity to motion coherence is significantly disrupted following a lesion in MT. Furthermore, MT is required to track moving stimuli with smooth pursuit eye saccades. Visuomotor behavior following a bilateral MT lesion in adulthood has been examined in the macaque (Gattass et al., 2020). The subjects were presented with a battery of tests to probe reach-and-grasp behavior in static and moving objects. Most notably, MT-lesioned animals showed impairment in their ability to engage in a static reach-and-grasp task in which the food reward is placed in a narrow slot so that the animal must only use the index and thumb for successful retrieval of the food reward. Closer examination of the retrieval mechanism used by the adult MT-lesioned animals revealed they had difficulty orientating the hand and typically had to grasp the food reward with the whole hand and could not exhibit finer grasps with the fingertips. The authors were unable to determine the impact that the adult-acquired bilateral MT loss would have on visuomotor behavior in moving objects. The moving object retrieval task in their study consisted of a spinning bowl that rotated at 240 RPMs. A banana pellet was dropped in the spinning bowl and bounced around. All animals had difficulty with this task, and no significant difference was observed between control and lesioned cohorts. In our study, our animals also showed perturbations in hand posture when engaging in the static task as seen with MT-lesioned animals, exhibiting a larger MGA when reaching for a food reward. This suggests that MT is required for prehension throughout life. However, it is possible that the deficits observed in this adult lesion model are because of the lesion extending beyond MT, as removal of underlying white matter was a component of the ablation, and degeneration of the LGN was observed, suggesting damage to the optic radiation.
Clinically, in a few cases of individuals with motion blindness, damage to both the lateral temporo-occipital cortex, which includes area MT, and the parieto-occipital cortex, which includes V6, has been observed (Blanke et al., 2003). In one example, patient LM suffered a bilateral vascular insult to the temporal portions of her brain, which included MT, in adulthood with no injury evident in V1(Zihl et al., 1983). Visual assessment of patient LM demonstrated that she had perturbations in reach-and-grasp behavior, oculomotor scanning patterns, and significant impairment in perceiving moving stimulus, although her ability to discriminate the color and form of objects remained normal (Zeki, 1991; Zihl and Heywood, 2015). Although LM clearly had akinetopsia, she demonstrated an ability to reach and grasp static objects and moving objects up to speeds of 0.5 m/s (Schenk et al., 2000). Her performance greatly diminished at speeds >0.5 m/s, and she often executed her reach with greater velocity to moving objects.
A notable difference between our study and LM's capabilities is that LM performed remarkably well when reaching for static or slowly moving objects, and her ability to intercept objects at low speeds was on par with that of controls. Early-life MT-lesioned marmosets performed poorly when intercepting moving objects and displayed greater peak velocity and acceleration in their executed reach compared with controls during the static task. Clinically, there is evidence that even with static object retrieval, transcranial magnetic stimulation of MT interferes with fluid goal-directed reaching (Whitney et al., 2007). Our early-life-lesioned marmosets, as well as the macaques with adulthood lesions in MT in the Gattass et al. (2020) study, were unable to replicate the success LM had with static object retrieval. We believe LM's ability to reach out to stationary and slow-moving stimuli could be a testament to her remarkable adaptive strategies in using the static visual cues available for her to estimate trajectories of slow-moving items and her hand relative to the stationary target object. However, it could also reflect the incomplete lesion in MT and the satellites. As the marmosets in our study displayed impairment in retrieval of moving objects akin to that of patient LM, we postulate that the plasticity of the developing brain does not permit recovery of visuomotor function in moving targets. As such, MT serves as a critical node for visuomotor behaviors that require information from moving stimuli throughout life. However, it must be noted that area DM/V6 may be responsible for the residual capacity in reach and grasp, as this area, through the V6–V6A–parietal area PEc circuit, has previously been demonstrated in macaques to be capable of eliciting reach-to-grasp behaviors (Fattori et al., 2017).
Further dissection of the failed behavioral trials by each marmoset in the moving object task revealed that the lesioned cohort initiated the reach prematurely. This observed behavior could be because of the inability of the animal to accurately detect the speed of the moving stimulus (akinetopsia), a deficit that was also observed in LM. It should be noted that motion is encoded in multiple cortical areas and not just in MT (Tolias et al., 2005; Pitzalis et al., 2010). Therefore, the intrinsic ability of the brain to translate motion perception into a meaningful and precise goal-directed action relies on more than just sensitivity to motion coherence and further reinforces how integral MT is for object interaction and the integral role it plays in the establishment of the associated networks during development.
The existing literature has highlighted the importance of the MIP and PE areas in reaching actions and limb coordination (Snyder et al., 1997; Filimon et al., 2009). Considering that we observed anatomic alterations specifically in these areas following an early-life lesion in MT, we suggest they are complicit in the perturbed reaching behavior observed in both the static and moving object tasks. However, this does not explain why the lesioned marmosets demonstrate a similar kinematic profile as their control counterparts when intercepting moving targets. As marmosets do not have a refined hand use akin to that of macaques or chimpanzees (Chiang, 1967; Boesch and Boesch, 1983), it is most likely that the tendency to execute reach-and-grasp tasks on moving targets with maximum velocity is a natural teleological behavior of marmosets, especially for fast arboreal maneuvers. Despite the comparable kinematics in the moving object task, the lesioned animals ultimately tended to reach and fail to collect moving objects prematurely.
To the best of our knowledge, this is the first study examining the tuning properties of V1 neurons following an early-life injury to an extrastriate area. Previously, studies have examined the physiological properties of neurons in MT following a lesion in V1 (Rodman et al., 1989; Kaas and Krubitzer, 1992), revealing that receptive field size and tuning properties were dramatically affected. However, MT responses remained robust and largely unaffected following V1 lesions sustained in early life (Yu et al., 2013). The only noticeable deficit was a reduced proportion of DS neurons. This supports the notion that the visual brain is able primarily to execute contingency mechanisms during development to provide a higher level of functional recovery following the early-life loss of V1. This concept has also been observed clinically in subject BI, who, following a bilateral early-life injury to V1 has significant preservation of vision, including sensitivity to moving stimuli. We wanted to observe the converse; whether the tuning properties of V1 are vulnerable during development and the extent of codependency between V1 and MT to develop DS or any other tuning property. In this study, we observed no dramatic perturbations in the tuning properties of V1 neurons. We did observe a lower proportion of DS neurons, which parallels previous findings for MT following an early-life lesion in V1 (Yu et al., 2013). Modifications of V1 tuning properties, such as DS or even orientation selectivity, have been reported previously (Tretter et al., 1975; Pasternak and Leinen, 1986; Kohn and Movshon, 2004) but none specifically concerning V1–MT circuitry. Therefore, further investigation is warranted to determine whether the MT-V1 projection serves a role in early life to tune DS within V1 neurons.
As PV interneurons form most GABAergic interneurons in V1(Brederode et al., 1990), we sought to determine whether there were disruptions to the local circuits by quantifying PV+ interneurons. Further, it is well documented that the local interneuron circuitry is more susceptible during development than in adulthood (Gomes et al., 2019). MT has extensive reciprocal connectivity with layers 3C and 6 of V1 (Mundinano et al., 2019b), and we revealed a reduced PV+ cell density within both supra and infragranular layers. PV+ neurons in V1 have a diversity of feature-specific visual responses that include sharp orientation and DS, small receptive fields, and bandpass spatial frequency tuning (Lee et al., 2012). These results suggest that subsets of PV interneurons are components of specific cortical networks and that perisomatic inhibition contributes to the generation of precise response properties. Ablation of PV in V1 interneurons has previously been observed to decrease DS properties within V1(Runyan et al., 2010). Therefore, it is conceivable that the reduced PV population in V1 following the early-life lesion in MT is the underlying basis for the observed lower proportion of DS neurons in V1.
The implications of this current study extend our understanding of the central role area MT occupies in the early establishment of the dorsal stream and associated behaviors and the role as an anchor in the developing visual cortex. Further, our observation of a lower proportion of DS neurons within V1 could suggest a level of dependency on MT in early life in the tuning and appropriate maturation of V1 neurons. Although marmosets lack fine motor skills, they have demonstrated their ability to use tools (Yamazaki et al., 2011) and have the circuitry for accurate reach-and-grasp behaviors. Teleologically, the early establishment of the dorsal stream network provides primates with the capacity to process vision for action, including accurate reaching-and-grasping behavior, which is integral to their survival.
Footnotes
This work was supported by the National Health and Medical Research Council (NHMRC) Project Grant APP1042893. W.C.K was supported by the NHMRC Dora Lush Postgraduate Scholarship APP1190007. J.A.B was supported by the NHMRC Senior Research Fellowship APP1077677. The Australian Regenerative Medicine Institute was supported by grants from the State Government of Victoria and the Australian Government. The authors thank David A. Leopold, who read through earlier versions of the manuscript; Mitchell J. de Souza for technical assistance; and Qizhu Wu for MRI acquisition and diffusion tensor imaging preprocessing.
The authors declare no competing financial interests.
- Correspondence should be addressed to James A. Bourne at James.Bourne{at}monash.edu
