Decreased Number of Parvalbumin and Cholinergic Interneurons in the Striatum of Individuals with Tourette Syndrome

Introduction

Tourette syndrome (TS) is a childhood-onset neuropsychiatric illness characterized by motor and vocal tics and a high incidence of co-morbid obsessive-compulsive disorder (OCD) and attention-deficit hyperactivity disorder (ADHD) (Graybiel and Canales, 2001; Leckman, 2002). Patients with TS also show deficits in procedural learning (Marsh et al., 2004). Insights from human lesions and degenerative disorders as well as animal models suggest that the basal ganglia are key for the initiation and coordination of sequential motor actions and may play a larger role in procedural learning (Albin et al., 1995; Graybiel et al., 1994). The cerebral cortex projects to the striatum, which comprises the caudate (Cd) and the putamen (Pt). Striatal medium spiny neurons (MSNs) send inhibitory connections, directly or indirectly through the external segment of the globus pallidus (GPe), to the output nuclei of the basal ganglia, which are the internal segment of the globus pallidus (GPi) and the substantia nigra reticulata (SNr). In turn, the GPi and SNr project inhibitory fibers to motor and intralaminary nuclei of the thalamus, which send fibers back to the striatum and to the cortex. This ensemble forms the cortico-striatal-thalamo-cortical (CSTC) circuitry. This basal ganglia network has been suggested to be crucial for the integration of automatic sequences into goal-directed behavior (Graybiel and Canales, 2001).

A large-scale structural imaging study (Peterson et al., 2003) found a small (5%) but significant decrease in Cd volume in both children and adults with TS, suggesting that this phenotype may represent a useful biomarker for this syndrome. Striatal MSNs projecting to the GP and SNr represent the vast majority of neuronal elements in this region. The MSNs receive cortical inputs; dopaminergic (DA) inputs from the midbrain, and modulatory inputs from four distinct classes of local circuit interneurons, namely, parvalbumin (PV), calretinin (CR), somatostatin/nitric oxide synthase/neuropeptide Y (SOM/NOS/NPY) and cholinergic interneurons. The PV, CR, SOM neurons represent inhibitory GABAergic cells that are involved in various forms of feed-forward inhibition within the striatum (Gurney et al., 2004). The fast-spiking PV+ interneurons are interconnected by gap junctions, forming a widespread, non-habituating inhibitory network (Kawaguchi, 1993; Kita et al., 1990). The CR+ medium-sized aspiny interneurons (10-20 μm) are the most abundant interneurons in the primate striatum. CR is also expressed by cholinergic interneurons (Cicchetti et al., 1998), which can be distinguished by their large size (24-42 μm) and by their expression of the acetylcholine synthetic enzyme choline acetyltransferase (ChAT) (DiFiglia, 1987).

A possible biological reason for the reduction in Cd volume in TS is a deficit in DA innervation of the striatum, as abnormal striatal DA function has been reported in TS patients (Albin et al., 2003; Singer et al., 2002). Another possibility is a deficiency in striatal interneurons or in a subset of MSNs. In a recent unbiased stereological study using postmortem basal ganglia tissue from individuals with TS and normal controls (NC), we reported a deficiency in PV+ interneurons in the Cd of three patients with severe TS (Kalanithi et al., 2005). To investigate possible alterations in other striatal cell populations, we undertook a follow-up study in which we assessed the regional density of PV, CR, ChAT-containing interneurons as well as MSNs in five TS patients and five NC. The results suggest that cholinergic, as well as PV+ interneurons, are markedly reduced in the striatum of severely affected TS individuals in a region-specific manner.

Methods

Results

The density of the three major striatal interneurons types (PV+, CR+ and ChAT+ interneurons), of DARPP-32+ medium spiny neurons (MSNs) and of cresyl violet-stained total neurons was estimated by stereological methods in the striatum (Cd and Pt) of NC and TS individuals. Using the anterior edge of the GPe as a landmark, density estimates were separately obtained for anterior and posterior portions of the Cd and Pt (denominated, respectively, Cd head and Pt rostral and Cd body and Pt caudal), excluding the NA (see Figure 1). Most of the anterior striatum defined in this way corresponds to the associative region, whereas the posterior mostly corresponds to the sensorimotor region (for a more precise delimitation of these functional regions, see below). Consistent with our previous data that included three of the five subjects used in the present study (Kalanithi et al., 2005), we found a 55.7% decrease in the density of PV-immunoreactive cells in the striatum of TS individuals (Figure 2 A, B; Figure 3; Table 3). The density of PV+ cells in the NC and TS striatum was, respectively, 331.4 ± 32.7 and 137.8 ± 32.7 cells/mm³ (mean ± SEM) and the difference between NC and TS was highly significant (F (1, 8) = 14.085, p = 0.006, ANOVA). There was no interaction between diagnosis (NC, TS) and area (Cd and Pt) or subarea (anterior and posterior), indicating that the decrease in PV+ interneurons in TS was equally significant in all regions of the striatum. This was confirmed with a non-parametric test, the Mann-Whitney U Test (p < 0.0005).

To determine if additional populations of neurons other than PV+ neurons are also altered in TS brains, we examined other classes of interneurons. The most abundant population of interneurons in the striatum is a class of medium sized aspiny interneurons (10-20 μm) expressing CR (Figure 4 A-D, arrowheads). There was no difference in the number of medium-sized CR+ interneurons between NC and TS individuals (Figure 5 C; F (1, 5) = 0.056, p = 0.822, ANOVA). Cell size and morphology did not appear to differ between NC and TS patients (Figure 4 A-D). However, approximately 10% of all CR+ cells are a distinct population of neurons with a much larger soma size (24-42 μm) and a more complex, multipolar pattern of dendritic branching. These latter cells are morphologically identical to ChAT+ cells (Figure 4 C, D arrows). This is consistent with previous studies showing that about 80% of the large-sized CR+ neurons co-localize ChAT, the acetylcholine synthetic enzyme (Cicchetti et al., 1998). The density of large-size CR+ interneurons, as determined by stereological analyses, was 236.0 ± 14.5 and 111.1 ± 14.5 cells/mm³ (mean ± SEM) in the striatum of NC and TS individuals, respectively (Figure 4 A-D; Figure 5 A, B; Table 3). A statistically significant 52.9% decrease of large-size CR+ interneurons was present in TS (F (1, 8) = 36.994, p = 0.001, ANOVA) and was confirmed with the Mann-Whitney U Test (p < 0.0005). Although the average decrease in large-size CR+ neurons was larger in the anterior portion than in the posterior portion of the striatum (62.9% versus 41.5%, respectively), neither the area (Cd/Pt) nor the subarea (anterior/posterior) showed any statistically significant interaction with the diagnosis. Thus, as for PV+ neurons, the decrease in large-size CR+ interneurons in TS was equally significant in all regions of the striatum.

To confirm the deficit of cholinergic interneurons, we examined ChAT immunoreactive neurons in the striatum. The density (cells/mm³, mean ± SEM) of ChAT+ cells in the striatum of NC and TS individuals was 213.6 ± 22.9 and 108.1 ± 22.9, respectively (Figure 6 A-D; Figure 7 A, B; Table 3). The data revealed a highly significant 49.4% decrease in ChAT+ neuron density in the TS striatum (F (1, 8) = 10.641, p = 0.011, ANOVA). Again, area (Cd/Pt) and subarea (anterior/posterior) did not interact with diagnosis. The overall difference between NC and TS was again confirmed with the Mann-Whitney U Test (p < 0.0005). Qualitatively, no obvious differences in the soma size of cholinergic neurons nor in the arborization of individual cells were detected. The overall level of ChAT+ immunoreactivity within the neuropil appeared to be decreased, likely due to the decreased cell number (Figure 6 A-D). Together with the previous results, these data suggest that there is a combined decrease in PV+ and cholinergic interneurons in the striatum of TS patients, as compared to NC.

To understand whether these changes extended to more ventral, limbic portions of the basal ganglia, we assessed neuron density in the NA. PV neurons were too sparse in this region to be reliably counted. There was no significant change in large-size CR+ interneurons in the NA (Table 3; p = 0.86, Mann Whitney U test). Similarly, no difference in ChAT+ neurons was detected between NC and TS individuals in this region (Table 3; p = 0.114, Mann Whitney U test).

To understand whether these changes were specific to PV+ and cholinergic neurons or were attributable to a more general loss of total neurons within the striatum, we stained adjacent serial sections with cresyl violet. Neurons were distinguished from glia based on somatic size and nuclear morphology. The density of cresyl violet stained neurons (cells/mm³, mean ± SEM) in the striatum of NC and TS individuals was 45,709 ± 2,083 and 38,352 ± 1,980, respectively, indicating no significant difference between NC and TS (F (1, 8) = 4.357, p = 0.070). Furthermore, the density of cresyl violet stained neurons in the Cd head, the area with the strongest differences with respect to the PV and cholinergic interneurons, was almost identical in TS and NC (Table 3; p = 0.841, Mann Whitney U test). However, the TS samples showed a trend to a decrease in cresyl violet stained neurons in the Cd body and in the Pt (Table 3, Supplementary Material 2). Although we cannot exclude that these trends may become significant with the addition of more brain samples to the study, the decrease in density of PV+ and cholinergic interneurons is clearly not attributable to a generalized neuronal loss, particularly in the Cd head.

To obtain an independent confirmation of the specificity of changes in PV+ and cholinergic neuron density, we immunostained the sections for DARPP-32, a protein expressed by the majority of MSNs. The density of DARPP-32+ neurons in the head of the Cd was 21,843 ± 3,038 cells/mm³ and 17,337 ± 1,075 cells/mm³ in NC and TS brains, respectively (p = 0.4, Mann Whitney U test). Hence, TS patients demonstrated no significant change in MSNs in the head of the Cd. Collectively, the data suggest that cell losses were specific for PV and cholinergic neurons in the Cd head in this group of TS individuals.

To further evaluate physiological implications of these changes for cholinergic circuitry, we investigated the distribution of large-sized CR+ interneurons (i.e., cholinergic interneurons) in the three main functional subdivisions of the striatum (receiving associative, sensorimotor and limbic afferents). For this analysis we used the StereoInvestigator program to accurately delineate in our serial tissue sections the associative, sensorimotor and limbic territories according to previous studies (Figure 8A) (Bernácer et al., 2007; Morel et al., 2002; Parent and Hazrati, 1995) and analyzed large-sized CR immmunoreactive cell density in both control and TS brains. A statistical comparison of diagnosis versus regional effects by repeated measures ANOVA revealed significant main effects of diagnosis (F (1, 8) = 25.852; p= 0.001) and region (F (2, 16) = 5.335; p= 0.017) and a significant interaction of diagnosis versus region (F (2, 16) = 5.631; p= 0.014). Indeed, in NC there was a gradient in the density of cholinergic neurons, with highest values in the associative and lowest in the limbic regions (Figure 8B). Differences between associative and limbic and associative and sensorimotor cholinergic neuron densities were both statistically significant in NC (Sidak post-hoc test, p= 0.016 and p= 0.031, respectively); in contrast, no significant difference in cholinergic neuron density was present among subregions in the TS striatum. A comparison of TS with NC region by region by the Sidak post-hoc test indicated that the associative region showed the most pronounced decreases in cholinergic neuron density (Figure 8B; 60.0% decrease, p < 0.0005) followed by the sensorimotor region (Figure 8B; 44.7% decrease, p = 0.008) and that the limbic region was not significantly different between TS and NC individuals (Figure 8B; p = 0.402). The last finding is also in accordance with the lack of observable changes in cholinergic markers in the NA described above.

Since there was a statistically significant difference in age between NC and TS using the Mann-Whitney U Test, we re-ran the ANOVA using age as a covariate for each of the individual main variables. In no case the effect of age was statistically significant (PV, p = 0.236; CR-large, p = 0.747; ChAT, p = 0.543; CR-large in functional subdivisions, p = 0.485), suggesting that age does not account for the variance between NC and TS. To further understand whether there was a significant correlation between age and interneuron density, we then performed regression analyses. In both NC and TS, we found no significant correlation between age and the density of any cell type in the striatum (PV-NC, p = 0.655; PV-TS, p = 0.703; CR-large-NC, p = 0.204; CR-large-TS, p = 0.593; ChAT-NC, p = 0.858; ChAT-TS, p = 0.29).

Discussion

Our data indicate that cholinergic interneurons may be reduced in the striatum of patients with severe and persistent TS, as indicated by two cellular markers, CR and ChAT. The largest decrease was observed in the more rostral associative territory, an area comprising the head of Cd and the adjacent rostral Pt. In more posterior sensorimotor portions of the Cd and Pt, differences in cholinergic cells were smaller and more variable, but still statistically significant. In contrast, the limbic territory, which includes the NA and adjacent anterior/ventral striatum, as defined by its connections with allocortical areas and the medial prefrontal cortex (PFC) and its continuity with the NA and shell of the amygdala (Haber and McFarland, 1999; Heimer, 2000; Holt et al., 1997), failed to show significant changes. We cannot exclude that with the analysis of more brain samples, a statistically significant difference in cholinergic cells in the NA might emerge. Nevertheless, one of the most remarkable and robust findings of the current study is that the gradient in cholinergic neuron density found in normal subjects (highest in associative regions, intermediate in sensorimotor and lowest in limbic regions) is completely abolished in TS. These data strongly implicate associative and sensorimotor regions of the basal ganglia in the pathophysiology of TS or in long-term adaptive compensatory changes to this disorder.

We also report a strong deficit in PV+ interneurons in the Cd and Pt of TS patients, which confirms and extends previous data (Kalanithi et al., 2005). The average decrease in PV+ neuron number was strongest in the rostral Pt, although no statistically significant interaction between diagnosis and regions was found, suggesting that PV+ cells are equally depleted in the whole striatum. We cannot exclude that the addition of more patients may reveal a statistically significant difference between subregions. PV+ cell bodies and fibers are enriched in the sensorimotor territory, extend into the more rostral associative territory, but are very sparse in the anterior/ventral limbic portions of the basal ganglia (Morel et al., 2002). Indeed, PV+ cells were too sparse in the NA to reliably assess their differences.

In contrast to PV and cholinergic interneurons, CR+ medium-sized interneurons, which represent the largest population of interneurons within the striatum, were not affected in this group of TS individuals. A count of total neuron densities in the striatum also revealed no statistically significant difference between TS and NC. PV and cholinergic neurons each represents less than 1% of the total neuronal population, and thus losses in these cells are not expected to have a significant impact on total neuron number. Together, the data suggest that there is a selective decrease in PV+ and cholinergic interneurons, rather than a generalized decrease in total neuron number in the striatum of TS individuals.

The main limitation of this study is the small sample size, and consequently the limited power to detect additional differences between the control and the patient groups and to control for sample variables. One of these variables is age, as TS are significantly younger in our sample (possibly because of frequent accidental deaths in this group). However, detailed statistical analyses suggested that age could not account for the significant differences between patients and controls.

Another variable is that most TS patients have been treated with dopamine D2 receptor blocking drugs. The possibility that the cellular changes may be the consequence of this treatment cannot be completely ruled out at present, nevertheless, it must be noted that long-term neuroleptic treatment in rats (4-12 months) did not decrease striatal levels of the GABA-synthesizing enzymes, neuropeptides and calcium binding proteins, including PV or CR, despite clear behavioral evidence of a movement disorder (Jolkkonen et al., 1994; Mithani et al., 1987).

Inconsistent results have been reported in the literature regarding the effect of chronic antipsychotic treatment on the cholinergic system. While some studies have reported that short-term neuroleptic treatment in rodents (40-90 days) leads to decreases in ChAT activity and ChAT-immunostained neurons in several brain regions (Mahadik et al., 1988; Terry et al., 2003), long-term treatment (6-12 months) produces more variable results, including increases, decreases, or no significant effect on ChAT activity, ChAT-immunoreactive neurons, acetylcholine levels and other indices of cholinergic function (Grimm et al., 2001; Lohr et al., 2000; Mithani et al., 1987; Murugaiah et al., 1982; Rupniak et al., 1986; Terry et al., 2007). These discrepancies may be explained by region-specific effects (the decrease in ChAT+ cells being more prominent in ventral striatum and NA) (Grimm et al., 2001), length of administration and types of drugs used (typical neuroleptics being more likely to cause changes in cholinergic function than the atypical risperidone and clozapine) (Friedman et al., 1983; Terry et al., 2003). Recent studies have reported that the density of cholinergic neurons in the limbic striatum is decreased in patients with schizophrenia (Holt et al., 1999) and that the lowest density of ChAT+ neurons was found in two schizophrenic individuals that had not been treated with neuroleptics. This is in agreement with pharmacological data suggesting that dopamine D2 receptor antagonists increase cholinergic neuron activity, choline utilization and acetylcholine levels in the striatum (Pedata et al., 1980; Rupniak et al., 1986). In aggregate, the above findings are not consistent with the idea that the decrease in ChAT+ cells observed in the present study is attributable to neuroleptic treatment.

Chronic exposure of macaque monkeys to typical and atypical antipsychotic drugs has been shown to be associated with decreased brain volume (Dorph-Petersen et al., 2005), which may be caused by a reduction in number of glial cells (Konopaske et al., 2008). However, our changes in density values cannot be explained by such an effect, as this would have produced a generalized increase in density across all neuron types examined. Thus, changes in volumes due to chronic neuroleptic treatment are unlikely to be the explanation for the selective changes in PV+ and ChAT+ neuron density that we have found in TS individuals.

A similar decrease in cholinergic neuron markers has been described in the Cd of schizophrenic patients, albeit the NA was also affected in this population (Holt et al., 2005; Holt et al., 1999). As the Cd nucleus receives direct inputs from the dorsolateral PFC, the data are in line with the PFC hypometabolism and functional deficit in this disorder. Disruptions in the PFC-Cd circuitry may also be pivotal for long-term outcome in TS, since a favorable future outcome in children with TS is associated with increased Cd size (Bloch et al., 2005) and tic suppression is concomitant with activation of the PFC (Peterson et al., 1998). Thus, the cholinergic neuron alterations described here may be a manifestation of hypofunctional frontostriatal circuitry, a characteristic in common between severe TS and schizophrenia (Yoon et al., 2007).

The electrically coupled PV interneuron network is responsible for downregulating electrical activity in the striatum (Kawaguchi, 1993; Kita et al., 1990). The PV neurons receive a tonic input from the spontaneously active cholinergic interneurons (Koos and Tepper, 2002). Thus the cholinergic neurons, through their connection to the PV neurons, can exert widespread inhibitory influence over MSNs (Aosaki et al., 1994; Graybiel et al., 1994; Pakhotin and Bracci, 2007). The cholinergic neurons receive excitatory projections from the thalamus, and particularly the ventralis anterior (VA)/ventralis lateralis (VL) and centromedian-parafascicular nuclear complex (CM-Pf) (McFarland and Haber, 2000; McFarland and Haber, 2001). These thalamic nuclei in turn receive inhibitory projections from GPi neurons, a region exhibiting high-frequency tonic and oscillatory activity (Difiglia and Rafols, 1988). Collectively, these observations suggest that self-reinforcing, feedback inhibitory circuits involving the PFC and sensorimotor thalamus impinge on the ChAT-PV interneuron network in the striatum. The possible importance of this inhibitory network for tic-related motor activity is reinforced by neurosurgical experiments, in which patients with intractable TS found partial relief from their symptoms when thalamic nuclei of the CM-Pf complex were implanted with deep brain stimulation electrodes (Bajwa et al., 2007; Servello et al., 2008; Temel and Visser-Vandewalle, 2004). In TS, the combined loss or dysfunction of the cholinergic and PV+ cells in the associative and sensorimotor striatum may severely impair the ability of the PFC and thalamus to carry out the inhibitory modulation of MSNs, which may result in tics and deficits in procedural learning. The finding that deficits in procedural learning in both children and adults with TS are correlated with tic severity (Marsh et al., 2004) argues that a dysfunctional PV/cholinergic neuron network may well be a core feature of TS anatomy and not an epiphenomenon.

GABAergic and cholinergic neurotransmission powerfully modulate the activity of oscillating networks. The firing of striatal PV+ neurons precedes that of the other cell types in the circuit, suggesting that these cells pace the oscillations (Berke et al., 2004). Consequently, a dysfunction in the PV-cholinergic circuit may lead these networks to become dysrhythmic, producing a loss of control of sensory information and motor action (Leckman et al., 2006; Llinás et al., 2005; Llinás et al., 1999). This work suggests the intriguing notion that a dysfunction of PV-cholinergic neurons in the associative and sensorimotor regions of the basal ganglia may underlie the emergence of tics and other forms of disinhibited behavior characterizing TS symptomatology.

Supplementary Material