|Home | About | Journals | Submit | Contact Us | Français|
The subthalamic nucleus (STN) is one of the principal input nuclei of the basal ganglia. Using electrophysiological techniques in anesthetized rats, we show that the STN becomes responsive to visual stimuli at short-latencies when local disinhibitory injections are made into the midbrain superior colliculus (SC), an important subcortical visual structure. Significantly, only injections into the lateral, but not medial, deep layers of the SC were effective. Corresponding disinhibition of primary visual cortex also was ineffective. Complementary anatomical analyses revealed a strong, regionally-specific projection from the deep layers of the lateral SC to neurons in rostral and dorsal sectors of the STN. Given the retinocentric organization of the SC, these results suggest that lower-field stimuli represented in the lateral colliculus have a direct means of communicating with the basal ganglia via the STN that is not afforded to visual events occurring in the upper visual field.
The subthalamic nucleus (STN) occupies a central position in the functional architecture of the basal ganglia (Parent and Hazrati, 1995; Joel and Weiner, 1997), an interconnected group of subcortical nuclei that represent one of the brain's fundamental processing units (Gerfen and Wilson, 1996; Redgrave, 2007). Despite a prodigious volume of work, there is no agreed consensus concerning the function(s) of the basal ganglia; however, their dysfunction is accompanied by a well characterized range of symptoms which underlie such disorders as Huntington's disease and Parkinson's disease (Wichmann and DeLong, 1996; Obeso et al., 2000). In the latter case, the STN has achieved particular prominence in recent years as a therapeutic target for deep brain stimulation (Benabid, 2003). Nevertheless, our knowledge of the STN, certainly compared to other basal ganglia nuclei, is still limited, and this deficiency seriously impedes our understanding of basal ganglia function and the development of interventions to ameliorate their dysfunction. In particular, although the inter-relationship of the STN with other basal ganglia nuclei is well documented (Parent and Hazrati, 1995; Joel and Weiner, 1997), its regulation by external systems is not, even though it is increasingly considered to be a major entry point to the basal ganglia (e.g. Nambu et al., 2002).
Electrophysiological studies have revealed that the STN contains neurons whose activity is related to movement and others that respond to sensory stimuli (DeLong et al., 1985; Matsumura et al., 1992). This suggests the structure plays a role in transforming sensory signals into a behaviorally-relevant output. The transformation(s) performed will, in part, be determined by the informational content of the sensory signals coded by STN neurons. This in turn will be constrained by the perceptual properties of afferent sensory circuits. One strategy therefore to reveal the content of sensory signals encoded by STN neurons would be to determine the perceptual properties of afferent sensory structures. However, the origin of sensory inputs to the STN has remained largely unexplored.
Neurons in the STN exhibit short-latency (70-120 ms) responses to the onset of phasic visual stimuli (Matsumura et al., 1992). Retinal projections to either the geniculostriate system or the midbrain superior colliculus (SC) are the most likely sources of short-latency visual activation. However, the visual cortex provides no direct input to the STN; indeed, cortical afferents arise exclusively from prefrontal and motor areas (Afsharpour, 1985a; Nambu et al., 2002). Recently, we have shown that short-latency visual input to another important input station of the basal ganglia, the dopaminergic neurons of the ventral midbrain, is provided by the SC (Comoli et al., 2003; Dommett et al., 2005), a highly-conserved but evolutionarily ancient subcortical visual structure. This raises the intriguing possibility that the SC may also act as the primary source of short-latency visual input to the STN. Our present results confirm this supposition, firstly by establishing that the SC is critical for short-latency visual activation of STN neurons and then by demonstrating the existence of a direct excitatory projection from a specific sub-region of the SC to the STN.
All aspects of this study were performed with Home Office approval under section 5(4) of the Animals (Scientific Procedures) Act 1986, and experimental protocols received prior approval of the Institutional Ethics Committees.
Twenty-two male Hooded Lister rats (285-620 g) anesthetized with urethane (ethyl-carbamate, 1.25 g/kg, Sigma-Aldrich; Poole, UK) were used in the electrophysiology experiments. A contralateral approach using an angle of 35° (ML: 2.5-3.2, bregma) was used to position a parylene-C-insulated tungsten microelectrode (1-2 MΩ, A-M Systems Inc., Carlsborg, WA, USA) into the STN (AP: 3.6-4.1 mm, bregma; DV: 7.0-8.0 mm, dura). A second tungsten electrode coupled with a 30 gauge metal injector needle filled with the GABAA receptor antagonist bicuculline methiodide (40 ng per 400 nl saline, Sigma-Aldrich) was introduced vertically into the SC (AP: 5.8-6.3 mm, bregma; ML: 1.0-2.5 mm, bregma; DV: 3.5-5.0 mm, dura). Lateral separation between the electrode and the injector was 0.2-0.5 mm. Intracollicular injections of bicuculline (0.5 μl; 0.5 μl/min) were made with a 10 μl Hamilton syringe (Scientific Laboratory Supplies Ltd., Nottingham, UK) mounted on an infusion pump, connected to the needle by a length of plastic tubing. The experimental rationale in the present study was based on the important observation of Katsuta and Isa (2003) that local intra-collicular injections of bicuculline can reinstate visual sensitivity to the SC deep layers rendered insensitive by urethane anesthesia. (Note: Throughout the current article, we adhere to the convention of referring to all SC layers underlying the superficial layers as the ‘deep layers’). Because of the known functional organization of the rodent SC (Dean et al., 1989), particular attention was paid to the effects of local disinhibition within different regions of the retinocentric map (Wurtz and Albano, 1980; Stein and Meredith, 1993).
Multi-unit spike-related potentials from the STN and SC were amplified, band-pass filtered (300Hz-10Khz), digitized at 10Khz and recorded directly onto computer disc using a 1401+ hardware acquisition system (Cambridge Electronic Design: CED; Cambridge, UK) connected to a standard PC running CED Spike 2 software. Multi-unit electrophysiological responses in the SC were used to guide the position of the SC electrode. Recordings were made while the electrode/injector assembly was lowered into the SC in the presence of a whole field light flash (0.5 Hz, 10 ms duration) from a green LED (60 lux 570 nm) positioned 5 mm in front of the contralateral eye. Using the characteristically vigorous visual response of the SC superficial layers as a guide to position, the electrode was lowered into the subjacent intermediate layers of the medial or lateral sectors of the SC. When both the STN and SC electrodes were in place, sets of trials comprising 120 light flashes (parameters as above) were recorded. Responses were obtained during an initial baseline period before an SC injection of bicuculline. Sets of trials then were continued until the effect of the drug had worn off, typically 15-20 min.
In three cases, an insulated silver wire was inserted into the visual cortex (with the dura carefully retracted) to record visually-evoked field potentials. These were low-pass filtered (DC-1 kHz, 3dB points), digitized at 10 KHz and recorded directly to disk using the 1401+ acquisition system. Signal averaging over sets of 300 light flashes was performed by the Spike 2 software. After recording under baseline conditions, a Gelfoam swab soaked in a solution of bicuculline (100 μg/ml) was placed directly on the surface of the exposed cortex. Averaged field potentials were recorded for successive series of 300 flashes until the effect of the drug had worn off.
At the end of the experiments, the position of any subcortical recording sites was marked with a small lesion induced by passing a 10μA anodal current for 2.5 min through the tungsten recording electrode. Animals then were killed by an overdose of pentobarbitone and perfused with 0.9 % saline followed by 4 % paraformaldehyde. The brains were taken for histological analysis.
Following the experiments, peri-stimulus time interval histograms (PSTHs) were constructed based on SC and STN multi-unit data (bin width 1 ms). PSTHs were imported into an Excel program (Microsoft) (Peter Furness, Sheffield) which determined the following response characteristics before and after an injection of bicuculline: (i) Baseline activity: the mean number of multi-unit spikes occurring during the 500 1 ms bins prior to the light flash. (ii) Response latency: the time point at which post-stimulation spikes exceeded 3.0 S.D. of the baseline mean. (iii) Response duration: response offset was defined as the time point at which post-stimulation activity returned to 3.0 S.D of the baseline mean. (iv) Response amplitude: the mean number of multi-unit spikes between response onset and offset minus the baseline mean. The statistical reliability of differences between response latencies for the SC and STN, and between STN baseline firing rates before and after SC injections of bicuculline, were assessed by t-tests. Probabilities <0.05 (two tailed) were accepted as statistically reliable.
For the light microscopy experiments we used 15 male Hooded Lister rats (265-602g) bred in the Sheffield laboratory. Animals were anesthetised with an intraperitoneal injection of a mixture of Ketaset (0.765 ml/kg) and Rompun (1.1 ml/kg). Single injections of the anterograde tracers biotinylated dextran amine (BDA: Sigma-Aldrich) and Phaseolus vulgaris leucoagglutinin (PHA-L: Vector Laboratories, Peterborough, UK) were directed to different quadrants of the SC on opposite sides of the brain in the same subject (e.g., caudal-medial and rostro-lateral or caudo-lateral and rostro-medial). The stereotaxic co-ordinates were: AP: 6.0 to 7.3 mm, bregma; ML: 0.5-2.5 mm, bregma; DV: 3.5 to 5.0 mm, dura). Some injections of anterograde tracers were combined with an injection of the retrograde tracer Fluorogold (Fluorochrome LLC, Denver, CO, USA) into the substantia nigra pars reticulata (AP: 5.0 to 5.8 mm, bregma; ML: 1.0-3.0 mm, bregma; DV: 8.0 to 9.0 mm, dura) to help identify the STN. BDA (10 % in phosphate buffer; PB) and Fluorogold (4.0 % in distilled water) were pressure ejected in volumes of 30-90 nl, while PHA-L was ejected iontophoretically (5μA anodal current applied to a 2.5 % solution in PB, 7 s on/off for 15-20 min). To identify SC neurons projecting to the STN, small (6-10 nl) pressure ejections of the retrograde tracer cholera toxin subunit B (CTb; Quadratech Epsom, UK; 1% solution in PB) or larger injections of the fluorescent tracer True Blue (Sigma-Aldrich) (5% in distilled water) were made into the STN. After allowing 7 days for the transport of tracers, animals were reanesthetised with pentobarbitone and perfused transcardially with warm saline (40°C) followed by 4 % paraformaldehyde in PB (pH 7.4). Serial coronal or sagittal sections (30 μm) were cut on a freezing microtome and collected in PB, except for tissue containing True Blue, where the sections were collected directly onto slides, allowed to dry in a light protected box and coverslipped with DPX mountant.
For electron microscopy, an additional three rats received PHA-L injections as described above. Following 7 days survival, the animals were perfused with phosphate-buffered saline (0.01 M phosphate, PBS) followed by 3 % paraformaldehyde plus 0.1 % glutaraldehyde. Sagittal sections (70 μm) were cut on a vibratome and collected in PB.
To reveal the tracers (BDA, PHA-L, CTb or Fluorogold), free-floating sections were washed with 0.1 M PB followed by 0.1 M PB containing 0.3 % Triton X-100 (PB-TX) for 30 min. For animals injected with PHA-L, CTb or Fluorogold, the sections were incubated overnight in primary antibody solution (goat anti-PHA-L, 1:800-1,000 dilution, Vector, or goat anti-CTb, 1:4,000 dilution, Quadratech, or rabbit anti-Fluorogold, 1:50,000 dilution, Fluorochrome). The next day, sections were washed with PB-TX and incubated for 2 h in biotinylated rabbit anti-goat IgG (1:100, Vector, in PB-TX containing 2 % normal rabbit serum) for PHA-L and CTb or in biotinylated goat anti-rabbit IgG (1:100, Vector, in PB-TX containing 2 % normal goat serum) for Fluorogold. After 30 min washing, all the sections were incubated in Elite Vectastain ABC reagent (Vector, 1:100 in PB-TX) for 2 h. The peroxidase associated with the tracers was revealed by reacting tissue with H2O2 for approximately 1 min using nickel-enhanced diaminobenzidine (DAB) as the chromogen for BDA and CTb (black reaction product) while PHA-L and Fluorogold were revealed by incubation with VIP (Vector) chromogen (purple reaction product). Finally, sections were washed in PB, mounted on gelatine-coated slides, dehydrated in graded dilutions of alcohol, cleared in xylene and coverslipped in DPX.
Following injections of anterograde tracers into the SC, three coronal sections through the STN approximately equivalent to 3.6, 3.8 and 4.1 mm caudal to bregma in the atlas of Paxinos and Watson (1997), or three sagittal sections approximately equivalent to 1.9, 2.4 and 2.9 mm lateral to midline were selected for analysis. Photomicrographs of injection sites in the SC and of the STN on each section were taken with a Spot-2 digital camera mounted on a Nikon Eclipse 800.
The location of retrogradely-labeled neurons was plotted on four coronal sections through the SC separated by ~ 0.5 mm (equivalent to 5.8, 6.3, 6.8 and 7.3 mm caudal to bregma). A series of digital images (magnification 100 X) were taken and imported into a graphics program (Macromedia Freehand, Adobe, San Jose, CA, USA) where they were montaged. The borders and layers of the SC were drawn over the montage. The scheme used for identifying the layers of the rodent SC is the one we have used previously (Sahibzada et al., 1986; Dean et al., 1989; Redgrave et al., 1992) and is based on the morphometric study of Albers et al. (1988; i.e., SGS, stratum griseum superficiale; SO, stratum opticum; SGI, stratum griseum intermediate; SAI, stratum album intermediate; SP, strata profunda). Each layer was bisected into medial and lateral zones as described previously (Telford et al., 1996). For each section, retrogradely-labeled neurons appearing in each subdivision were plotted using a combination of the underlying photomontage and direct microscopic inspection of the tissue. A three-factor repeated measures ANOVA was used to analyze differences in the distribution of tecto-subthalamic neurons within different sectors of the SC. The factors were: i) “Rostral/caudal” with 4 levels (5.8, 6.3, 6.8 and 7.3); ii) “Medial/lateral” with two levels (medial or lateral); iii) “Layers” with four levels (SGS+SO, SGI, SAI and SP). Probabilities <0.05 (two tailed) were accepted as statistically reliable.
Tissue for electron microscopic analysis was treated in a similar manner, except that the sections were freeze-thawed using isopentane and liquid nitrogen, and Triton X-100 was not used. PHA-L was revealed using DAB as the chromogen for the peroxidase reactions (about 6 min). After post-fixation in osmium tetroxide (1 % in PB, 30 min), sections were embedded in resin and flat-mounted on microscope slides. Regions of the STN with anterograde labeling were re-embedded and re-sectioned for electron microscopic analysis. The sections were collected on Pioloform-coated, single-slot grids that were stained with lead citrate and examined in a Philips CM10 electron microscope.
As expected, neurons in the SC deep layers (Fig. 1Aa and 1Ba) and the STN (Fig. 1Ac and 1Bc) failed to respond to the light flash during a baseline period prior to the SC injection of bicuculline. However, after local injections of bicuculline, both medial and lateral parts of the SC deep layers exhibited short-latency, short-duration bursts of impulses in response to the light flash (Fig. 1Ab and 1Bb). By contrast, a corresponding burst of spikes was recorded from STN neurons only when bicuculline was injected in lateral parts of the SC deep layers (Fig. 1Ad). This effect was observed in all cases (13/13). Corresponding injections of bicuculline into medial SC were, however, consistently ineffective (6/6), leaving STN neurons unresponsive to the light flash (Fig. 1Bd). Systematic variations in the STN recording sites cannot account for these differences because they were similarly dispersed throughout the structure for cases receiving medial or lateral SC injections (supplementary Fig. 1 online). As the visual sensitivity of the SC was reinstated, several patterns of visually-evoked activity emerged in the STN: (i) a short-latency, short-duration phasic response (Fig. 2a); (ii) a short-latency response with a slower build up and a longer duration (Fig. 2b); (iii) a superimposition of response-types (i) and (ii) (Fig. 2c); and (iv) a long-latency, long-duration response (Fig. 2d). No reliable relationships were established between the various response patterns and either the loci of recording sites in the STN or the precise location of bicuculline injections into the SC (supplementary Fig. 1 online). However, in all cases the STN visual response latencies (mean latency: 86 ± 9 ms) were reliably longer than those recorded in the SC (mean latency: 53 ± 5 ms) (t = 3.679; df = 12; p < 0.01). Considered together, these results suggest that the lateral SC is an important source of short-latency visual input to the STN.
General disinhibition of neural activity within the SC also induced a significant increase of the tonic firing rate of STN neurons measured in the 500 ms before each light flash. Thus, the mean pre-stimulus spike rate of STN multi-unit activity increased from 26.62 ± 2.4 Hz to 45.17 ± 3.86 Hz following injections of bicuculline into the lateral SC (t = −4.19; df = 12; p < 0.01). By contrast, no comparable changes were found in the baseline firing rate of STN neurons following corresponding injections into the medial SC (mean number of multi-unit spikes recorded pre- and post-bicuculline were 27.24 ± 1.7 Hz and 35.00 ± 2.6 Hz respectively; t = −2.3; df = 5; p>0.05). These data provide further support for a regionally specific functional link between the lateral SC deep layers and the STN.
To determine whether enhanced activity in the visual cortex might also influence STN visual activity, bicuculline was applied to the primary visual cortex to increase its responsiveness to visual stimuli. Although the application of bicuculline significantly potentiated local cortical visually-evoked activity (Fig. 3a and b), in all cases (3/3), the population response of STN neurons remained insensitive to the light flash (Fig. 3c and d). Once again, the STN recording sites were among those from which clear visual responses had been recorded following disinhibition of the lateral SC (supplementary Fig. 1b online). Local disinhibition of the visual cortex also failed to influence the baseline firing rate of STN neurons (mean multi-unit spikes recorded pre- and post-bicuculline 26.16 ± 2.85 Hz and 30.77 ± 2.79 Hz respectively; t = − 0.925; df = 2; p>0.05). These data indicate that comparable disinhibition of cortical visual activity was unable to influence either the baseline activity or restore visual sensitivity to STN neurons in the anaesthetized preparation.
Taken together, the electrophysiological data suggest that the presence of short-latency visually-evoked activity in the STN is critically dependent upon input from the lateral SC deep layers but not from the medial SC or visual cortex. This degree of functional segregation within the SC was at odds with current descriptions of direct connections between the SC and the STN in rodents (Tokuno et al., 1994); a projection which seems, at best, modest with no evidence of regional specialization within the SC. Therefore, to better understand the anatomical basis of our electro-physiological data, we re-evaluated the SC-STN pathway in greater detail.
Injections of the anterograde tracers PHA-L (N=4) and/or BDA (N=8) into the SC revealed a more robust direct pathway to the STN than was previously reported (Tokuno et al., 1994) (Fig. 4). At least two different trajectories for SC efferent fibers could be identified. First, a bundle of fibers was seen leaving the rostral aspect of the SC, coursing vertically down through the parafascicular nucleus and then turning rostrally along the medial lemniscus to terminate in the STN. The second route appeared as a rostral extension of the pathway we have previously described (Comoli et al., 2003) from the SC to dopaminergic neurons in the vental midbrain, tecto-nigral pathway (Fig. 4).
The tecto-subthalamic projection originates almost exclusively in the lateral SC. Thus, injections of either PHA-L or BDA into lateral regions (N=6) were associated with high densities of labeled axons and terminals in ipsilateral STN (Figs. (Figs.44 and 5a-c). In contrast, the STN was virtually devoid of terminal labeling when injections (N=6) were directed to the medial SC (Fig. 5d-f). Moreover, tecto-subthalamic axons and terminals were differentially distributed within specific sub-regions of the STN. In the caudal aspects of the STN, anterogradely labeled axons and terminals were largely confined in a dorsal sheet (Fig. 4) that extended across the entire medio-lateral dimension of the nucleus (Fig. 5c). Rostrally, terminating fibers filled the entire medio-lateral and dorso-ventral area of the STN (Fig. 5b). There were many instances where anterogradely-labeled boutons were seen in close apposition to STN soma and dendrites (Fig. 5 g-j)
To determine whether the appositions between terminals labeled from the SC and neuronal elements in the STN established direct synaptic contacts, an electron microscopic analysis was carried out on tissue from three rats that received PHA-L injections into the lateral SC. A total of 85 labeled structures were examined, of which 27 were identified as synaptic boutons. These were found to form synaptic contacts with 32 postsynaptic elements within the STN. In each case, the synapse was of the asymmetrical type (Gray's Type 1). The labeled terminals made synaptic contact either with dendritic shafts (53 %, 17/32, Fig. 6A) or spines (47 %, 15/32, Fig. 6B) of STN neurons, with a small population of boutons making contact with both (19%, 5/27). These results demonstrate that there is a direct synaptic connection between the SC and the STN and the type of synapse formed suggests that the projection is predominantly excitatory.
To identify the regional distribution and morphology of tecto-subthalamic projection neurons, labeled neurons in the SC were plotted following injections of retrograde tracers CTb (N=3) or True Blue (N=3) into the STN. Retrogradely-labeled neurons were mainly located in lateral sectors of the SC deep layers (Fig. 7). This impression was confirmed by a 3-factor (Layers – 5 levels; Medial/lateral – 2 levels; Rostral/caudal – 4 levels) repeated measures ANOVA (interaction between the Layers and Medial/lateral factors: F = 5.254, df= 3, p < 0.05). Collicular neurons retrogradely-labeled by STN injections typically were small-to-medium sized multi-polar neurons (Saito and Isa, 1999) (mean cell-body area = 182.5 μm2 ± 12.7 SE; range 85.6 - 417.8μm2) (Fig. 8). Large tecto-reticulo-spinal output neurons (Bickford and Hall, 1989) were not labeled.
In summary, our tracing experiments demonstrated a substantial anatomical projection between the SC and the STN that originates from small-to-medium sized neurons concentrated in the far lateral aspects of the deep SC layers. Tecto-subthalamic fibers preferentially target dorsal and rostral parts of the STN where they make asymmetric, presumed excitatory, synaptic contacts.
The main findings of the present study are: (i) that local pharmacological disinhibition of the lateral, but not medial SC deep layers, nor comparable disinhibition of the visual cortex, leads to a short-latency visual activation of the STN; (ii) the direct projection from the SC to the STN is more substantial than previously proposed (Tokuno et al., 1994), arising predominantly from the lateral SC deep layers and terminating preferentially in the rostral and dorsal sectors of the STN; (iii) both electrophysiological and electron-microscopic data suggest that the pathway is predominantly excitatory. Together, these results identify the lateral SC as the principal, if not exclusive, source of short-latency visual input to the STN. These findings have broad implications for understanding the functional role(s) of the STN within the wider context of the basal ganglia (see below)
We have demonstrated that STN neurons respond at short latency to visual stimuli and that this response derives from the SC. A close association was observed between the location within the SC where local pharmacological disinhibition was required for visual stimuli to elicit short-latency activation of the STN and tecto-subthalamic cells-of-origin. Thus, the tecto-subthalamic projection arises mainly from small-to-medium sized neurons in the lateral deep layers of the SC, where pharmacological disinhibition was effective. Virtually no anterogradely-labeled fibers and/or terminals were observed in the STN following tracer injections into the medial SC deep layers and bicuculline injections into the medial SC left the STN visually unresponsive, despite the presence of robust visually-evoked population responses from medial deep SC layer neurons.
A second point of correspondence between the electrophysiological and anatomical data concerns the potential excitatory nature of the tecto-subthalamic projection. Unlike our previous electron-microscopic analysis of the projection from the SC to dopaminergic neurons in the vental midbrain (the tecto-nigral projection), where both asymmetric (presumed excitatory) and symmetric (presumed inhibitory) synaptic contacts were reported (Comoli et al., 2003), only labeled terminals forming asymmetric synapses were seen in the present study. Consistent with these anatomical findings, electrophysiological analysis of the tecto-nigral projection (Dommett et al., 2005) revealed both excitatory and inhibitory visually-evoked responses in nigral dopaminergic neurons. No inhibitory responses to visual stimulation were observed in the present study. However, some slight caution is required because inhibitory responses of individual STN neurons may have been masked by the dominating excitatory responses of others in the multi-unit record. Despite this qualification, the fact that we observed only excitatory visually-evoked multi-unit responses in the STN is consistent with our failure to detect symmetrical tecto-subthalamic synaptic contacts. To this extent, it appears that the tecto-nigral and tecto-subthalamic projections may be at least partly anatomically and functionally distinct.
An interesting observation in the presnt study was that, while visually-evoked responses could be recorded from locations throughout the STN, our anterograde tracing data showed that injections of tracer into the lateral SC consistently produced dense terminal labeling confined to a specific region of the STN (rostral and dorsal). There could be several reasons for this apparent inconsistency. First, in the rat, STN neurons typically have large dendritic arbors that can extend over half the length of the nucleus (Kita et al., 1983; Afsharpour, 1985b). In addition, many of these glutamatergic STN projection neurons (Smith et al., 1998) have intra-nuclear (local) axon collaterals making intrinsic, presumed excitatory connections within the STN (Chang et al., 1983). Neural modeling of this intrinsic architecture suggests that large regions of this highly interconnected nucleus respond to excitatory input with a widespread pulse of increased activity (Gillies and Willshaw, 1998). If these biologically-constrained simulations accurately reflect STN neural processing, excitatory visual input directed to one region of the STN (e.g., rostral and dorsal regions) could evoke responses throughout the structure. However, the extent to which intrinsic STN neurons are effectively interconnected has been questioned (Wilson et al., 2004), and following electrical stimulation of different regions of cerebral cortex, spatially-restricted zones of excitation were found within the STN (Kolomiets et al., 2001). A second possible explanation is that the observation of qualitatively different patterns of visually-evoked responses from the STN suggests the involvement of additional afferent pathways. In the light of similar findings following electrical stimulation of cerebral cortex, Kolomiets et al. (2001) concluded that different classes of electrophysiological response pattern in the STN probably reflected neural transmission along the different direct and indirect routes connecting the cerebral cortex with the STN (Maurice et al., 1998; Nambu et al., 2002). A similar explanation could account for the different classes of visually evoked responses recorded throughout the STN in the present study. Possible indirect routes might include relays in the parafascicular nucleus of the thalamus (Maurice et al., 1998; Krout et al., 2001), the pedunculopontine tegmental nucleus (Redgrave et al., 1987; Spann and Grofova, 1989), and dopaminergic neurons of substantia nigra (Tofighy et al., 2003; Dommett et al., 2005). The potential role of these indirect projections and the extent to which they may express divergent contacts remains to be determined.
Both the SC and the basal ganglia are phylogenetically ancient systems that have been highly conserved throughout vertebrate evolution (Medina and Reiner, 1995). Interactions between the SC and the basal ganglia would, therefore, have been established prior to the elaboration of cerebral cortex. Indeed, we have argued previously that, while closed-loop projections are an important component of cortico-basal ganglia-cortical architecture (Alexander et al., 1986; Parent and Hazrati, 1995), a similar, but phylogenetically-older, closed-loop architecture involving the striatum connects the basal ganglia with brainstem sensorimotor structures, including the SC (McHaffie et al., 2005). In recent years, it has been appreciated that direct connections between the cerebral cortex and the STN, the hyperdirect pathway (Nambu et al., 2002), represents a second important route providing input to the basal ganglia. The present demonstration of a significant projection to the STN from an evolutionarily primitive sensorimotor structure (i.e., the SC) reveals a further architectural correspondence between cortical and subcortical connections with the basal ganglia.
Clues to the function(s) of the STN have been gleaned by examining its impact on basal ganglia target structures such as the SC. The STN provides a major input to the basal ganglia output nuclei, including the substantia nigra pars reticulata, which contains the cells-of-origin of the nigro-tectal projection to the SC. Because the subthalamo-nigral projection is excitatory (Smith et al., 1998) and the nigro-tectal pathway is inhibitory (Chevalier and Deniau, 1990), a burst of excitatory STN activity transmitted to nigro-tectal neurons would deliver a burst of inhibitory activity to the SC: similar inhibitory signals would appear simultaneously at other projection targets of the basal ganglia output nuclei. Two potential, not necessarily exclusive, functions for such inhibition have been proposed: (i) Since the SC plays an important role in redirecting head and eye movements (i.e., gaze shifts) towards or away from sudden changes in the visual field (Wurtz and Albano, 1980), a short-latency burst of inhibitory activity could be part of a mechanism to interrupt current gaze-fixation prior to the redirection of gaze (Gillies and Willshaw, 1998; Redgrave et al., 1999). Inhibitory signals directed elsewhere could interrupt other components of behavioral output; (ii) A phasic, widely-broadcast excitatory signal from the STN to basal ganglia output nuclei could provide the prior ‘excitatory surround’ on which later focused inhibition from the striatum (representing the desired motor program for execution) is superimposed during action selection, in which the basal ganglia are believed to play a role (Mink, 1996; Gurney et al., 2001). The findings that visual responses in STN neurons have shorter latencies than those of gaze shifts required to fixate a target (i.e. STN latencies are generally <100ms compared with normal saccadic latencies of >150ms), and that short-latency input is provided by the SC, whose visual processing capabilities are rapid but relatively limited (Wurtz and Albano, 1980; Stein and Meredith, 1993), are consistent with both of the proposed putative functions, which require afferent sensory signals to the STN to be rapid but care less about the detailed content of those signals.
However, additional results in the present study raise problems for both suggestions. The ecology of the rodent is such that threatening stimuli (i.e. predators), presumably requiring the interruption of behavior, are most commonly detected in the upper visual field, which according to the retinocentric map is represented in the medial SC (Dean et al., 1989). Yet, the present study revealed virtually no contribution from the medial SC to the tecto-subthalamic projection, and hence no clear means for upper field stimuli to interrupt behaviour via this route. On the other hand, from an action selection point of view, it is not immediately apparent why selection mechanisms concerned with responses to visual events in the upper field require a radically different, non-STN associated architecture, compared with that used for the lower field. At the very least, the present results are inconsistent with the STN playing the global, ubiquitous role in action selection/behavioral interruption envisaged, although at a deeper level they may suggest that a reappraisal of the function of this nucleus is required.
The work was funded by grants from the BBSRC (to PO: 062742), Wellcome Trust (to PR: WT080943), Medical Research Council, UK (to JPB & JM) and NIH (to JGM: NS35008). The authors are grateful to Natalie Kennerley for her excellent histological assistance and to Peter Furness for the provision of software to analyze the electrophysiological data.