Decreased density of cholinergic interneurons in striatal territories in Williams syndrome

Abstract

Williams syndrome (WS) is a rare neurodevelopmental disorder caused by the hemideletion of approximately 25–28 genes at 7q11.23. Its unusual social and cognitive phenotype is most strikingly characterized by the disinhibition of social behavior, in addition to reduced global IQ, with a relative sparing of language ability. Hypersociality and increased social approach behavior in WS may represent a unique inability to inhibit responses to specific social stimuli, which is likely associated with abnormalities of frontostriatal circuitry. The striatum is characterized by a diversity of interneuron subtypes, including inhibitory parvalbumin-positive interneurons (PV+) and excitatory cholinergic interneurons (Ch+). Animal model research has identified an important role for these specialized cells in regulating social approach behavior. Previous research in humans identified a depletion of interneuron subtypes associated with neuropsychiatric disorders. Here, we examined the density of PV+ and Ch+ interneurons in the striatum of 13 WS and neurotypical (NT) subjects. We found a significant reduction in the density of Ch+ interneurons in the medial caudate nucleus and nucleus accumbens, important regions receiving cortical afferents from the orbitofrontal and ventromedial prefrontal cortex, and circuitry involved in language and reward systems. No significant difference in the distribution of PV+ interneurons was found. The pattern of decreased Ch+ interneuron densities in WS differs from patterns of interneuron depletion found in other disorders.

Introduction

Although traditionally best understood with reference to motor function, a growing body of literature has highlighted the importance of basal ganglia structures in cognitive and emotional processes (Grahn et al. 2008, 2009). The striatum comprises the principle input nuclei for the basal ganglia, and receives important overlapping projections from cortical areas that topographically organize its territories into functional loops (Nakano et al. 2000; Haber and Knutson 2010; Choi et al. 2012; Haber 2016). Together, these functional territories form an important site of convergence for integrating information from diverse cortical sources underlying cognitive processes (Morris et al. 2016). The striatum is characterized by the distribution of a variety of inhibitory, GABAergic interneurons, and excitatory, cholinergic interneurons that modulate the function of inputs from regions sharing connectivity with the striatal nuclei and output from striatal medium spiny neurons. These diverse populations of inhibitory interneurons express calbindin, calretinin, parvalbumin, somatostatin, neuropeptide Y, and NADPH-diaphorase, and combined, constitute around 20% of total striatal neuron populations in the neurotypical human brain (Wu and Parent 2000; Bernácer et al. 2012; Lecumberri et al. 2018). Among these, fast-spiking interneurons express the calcium-binding protein parvalbumin (PV+) and forms multiple synapses with single cortical afferents in the striatum (Koós and Tepper 1999; Tepper et al. 2010; Berke 2011).

Striatal cholinergic interneurons (Ch+) are excitatory neurons that comprise less than 1% of total neuron populations in the human striatum (Bernacer et al. 2007, 2012; Stephenson et al. 2017). Despite their sparse distribution, these extremely large, heavily branched neurons share extensive connections with cortical and subcortical projections, and modulate inputs to local inhibitory interneurons and medium spiny projection neurons within the striatum (Bolam et al. 1984; Contant et al. 1996; Pakhotin and Bracci 2007). Data produced from rodent models support the hypothesis that Ch+ interneurons contribute to the modulation of social behavior. Martos and colleagues (2017) showed that selective ablation of Ch+ interneurons led to a significant increase in compulsive, socially directed (not object-focused) exploratory behavior in mice. Rapanelli and colleagues (2017) additionally demonstrated that the ablation of both Ch+ and fast-spiking PV+ in the caudate and putamen may result in ‘autistic’ behaviors, including repetitive and stereotyped behavior and reduced social interaction, particularly in male (but not female) mice. Importantly, the behavioral pattern was observed only when both types of neurons were targeted. These deficits may relate to the relationship between interneurons and medium spiny projecting neurons in the striatum. Together, these findings suggest that Ch+ and PV+ interneurons contribute to a mechanism that may selectively regulate socially directed behavior.

Williams syndrome (WS) is a rare genetic disorder with an unusual cognitive phenotype characterized most saliently by the profound disinhibition of social behavior (Järvinen-Pasley et al. 2008; Järvinen et al. 2013). Individuals with WS are known to have an unusually high drive to engage in social interactions, which may be attributed to a deficit in the inhibitory control of social behavior (Little et al. 2013) and dysfunction of behavioral response inhibition (Mobbs et al. 2007). Further evidence for pathology of striatal morphology and its contribution to functional deficits come from structural imaging, where decreases in gray matter volume in the caudate, as well as decreased volume of the putamen and nucleus accumbens have been reported (Reiss et al. 2000, 2004; Fan et al. 2017). Taken together, these findings suggest that the unique cognitive profile in WS may be associated with abnormalities of frontostriatal circuitry.

Imbalances of inhibitory and excitatory neuronal activity, both in the cerebral cortex as well as in subcortical structures, are thought to contribute to dysfunction in a variety of psychiatric and neurodevelopmental disorders (Yizhar et al. 2011; Marín 2012), including autism (Gogolla et al. 2009; Zikopoulos and Barbas 2013; Nelson and Valakh 2015; Hashemi et al. 2017). Our preliminary findings in WS (Hanson et al. 2018a) suggested a percentwise decrease in the density of Ch+ interneurons in the caudate. Here, we examined the density of Ch+, as well as PV+, interneurons in the striatum in an expanded sample of neurotypical subjects (NT) and subjects with WS. Our aim was to determine if there are differences in the density of Ch+ and PV+ interneurons between NT subjects and subjects with WS within striatal territories with shared connectivity with the prefrontal cortex (PFC), and how the pattern of their distribution may relate to differences observed in other neuropsychiatric disorders. Given the unusual social phenotype observed in WS, it was expected that altered Ch+ and PV+ interneuron densities may contribute to deficits in behavioral inhibition observed in WS, due to these cell types’ putative role in modulating social approach behavior.

Methods

Materials

Table 1 lists information on all subjects from whom postmortem materials were used in the current study. All brain tissue samples were collected from donors for whom an informed consent for brain autopsy and use of clinical data and demographic information for research was obtained. The brains of seven subjects with behaviorally confirmed diagnoses of WS (four males and three females, average age 38.5 ± 7.3 years) in the Ursula Bellugi Williams Syndrome Brain Collection were included in the sample. Genetic confirmation of WS, based on FISH for elastin, was confirmed for five out of seven subjects. Medical history for two subjects, for whom genetic deletion data were not available (WS1 and WS13) indicated key phenotypic traits, including distinctive facies and significant history of supravalvular aortic stenosis (SVAS), consistent with the genetic disorder’s biomedical profile (Morris and Mervis 2000; Pober 2010). Six NT subjects (four males and two females, average age 39.0 ± 6.3 years) were provided in collaboration with the University of Maryland Brain and Tissue Bank, who granted ethical permission for use of postmortem materials in this research. All brain samples were cryosectioned at 40 μm and a 1-in-10 series stained with 0.25% thionine solution for Nissl. Independent sample t tests revealed no significant difference in age (t₁₁ = 0.05, p = 0.96), postmortem interval (t₁₁ = 0.16, p = 0.87), or fixation duration (t₁₁ = 0.04, p = 0.97) between WS and NT subjects.

Table 1.

Subject background information

Subject	Age at death (years)	Sex	Diagnosis	PMI (hours)	Cause of death
WS 10	18	M	WS	24	WS/cardiac complications
4916	19	M	NT	5	Drowning
WS 15	25	F	WS	24	WS/cardiac complications
5350	25	F	NT	26	Sepsis
WS 1	31	M	WS	26	WS/cardiac complications
5614	31	M	NT	24	Acute drug intoxication
WS 14	42	F	WS	12	WS/cardiac complications
WS 9	43	F	WS	12	WS/cardiac complications
5445	42	F	NT	10	Pulmonary thromboembolism
WS 12	45	M	WS	24	WS/cardiac complications
5875	45	M	NT	19	Aortic dissection
WS 13	69	M	WS	8	WS/cardiac complications
5943	69	M	NT	23	Acute coronary artery thrombosis

PMI postmortem interval

Immunohistochemical staining

Two 1-in-20 series of sections were selected from the interval spanning the rostral striatum and head of the caudate nucleus rostral to the emergence of the anterior commissure in the coronal plane for each individual for immunohistochemistry. Prior to immunohistochemical processing, sections were stored at −20 °C in cryoprotective solution. Sections were stained for choline acetyltransferase (ChAT) and parvalbumin using an avidin–biotin peroxidase method previously described (Raghanti et al. 2008) and developed for optimal use in archival fixed tissues. Briefly, floating tissue sections were rinsed in PBS for 50 min, and pre-treated for antigen retrieval utilizing a 0.05% citraconic acid solution at 86 °C for 30 min. Endogenous peroxidases were quenched by incubation in a methanol and hydrogen peroxide solution. Sections were pre-blocked in a solution of normal sera and Triton-X detergent prior to incubation primary antibodies for ChAT and PV.

Cholinergic interneurons

Free-floating tissue sections were incubated in an anti-ChAT polyclonal antibody raised in goat (Millipore AB-144P) at a 1:500 dilution concentration, continuously shaken for 24 h at room temperature, followed by 24 h at 4 °C. The primary antibody selected labels a band at approximately 68 kDa in humans and nonhuman primates, and has been validated for immunohistochemical use in both nonhuman primate (Raghanti et al. 2008; Stephenson et al. 2017) and human (Bernácer et al. 2007; Raghanti et al. 2008; Stephenson et al. 2017) tissue.

Parvalbumin

Adjacent 1-in-20 series of free-floating tissue sections from each subject were incubated in a monoclonal antibody raised against parvalbumin (Swant #235) at a 1:5000 dilution concentration, and continuously shaken for 24 h at room temperature followed by 24 h at 4 °C. The primary antibody specifically stains the ⁴⁵Ca-binding site of parvalbumin, and has been extensively validated for use in human and nonhuman primate fixed tissue (Prensa et al. 1999; Sherwood et al. 2009).

Following incubation in primary antibody, all sections were rinsed and incubated in a biotinylated secondary antibody (dilution 1:200), followed by an avidin–peroxidase complex (ABC kit; PK-6100, Vector Laboratories, Burlingame, CA). A 3,3′-diaminobenzidine chromogen with nickel enhancement (SK-4100, Vector Laboratories) was utilized to visualize immunoreactive cells and fibers with robust staining of cell bodies and their processes (Figs. 1 and 2). Stained sections were mounted on gelatin-coated slides that were blinded to the investigator for diagnosis prior to quantification. No qualitative differences in staining or in the distribution of positively stained cells were observable between NT and WS subjects that would distinguish them during counting procedures.

Fig. 1.

Examples of neurons and fibers positively stained for ChAT+ in NT (left) and WS (right) subjects in five striatal regions of interest including the dorsal (a, b) and medial (c, d) caudate, dorsal (e, f) and medial (g, h) putamen, and nucleus accumbens (i, j) regions. Arrows denote incomplete cell bodies of neurons that did not meet inclusion criteria. Scale bar = 50 um

Fig. 2.

Examples of neurons and fibers positively stained for PV in NT (left) and WS (right) subjects in four striatal regions of interest including the dorsal (a, b) and medial (c, d) caudate, dorsal (e, f) and medial (g, h) putamen regions. Arrows denote incomplete cell bodies of neurons that did not meet inclusion criteria. Scale bar = 50 um

Regions of interest

Boundaries for five territories, including the dorsal and medial territories of the caudate nucleus, the nucleus accumbens region, and the medial associative and dorsal associative-sensorimotor regions of the putamen were established from adjacent series of Nissl-stained sections as previously described (Hanson et al. 2018a, b; Fig. 3). Care was taken to sample within these territories consistently, avoiding transitional areas and the fibers of the internal capsule. Due to their highly sparse distribution and consistent with previous observations (Kataoka et al. 2010), PV+ interneuron density could not be reliably quantified in the nucleus accumbens region, and was, therefore, not measured.

Fig. 3.

Regions of interest sampled corresponding to the dorsal (dC) and medial (mC) caudate nucleus, medial putamen (mP), dorsal putamen (dP), and nucleus accumbens (NA) regions. Dorsal regions of the putamen and dorsomedial regions of the caudate receive overlapping sensorimotor and associative projections. Adapted from Bernácer et al. (2007); Haber and Knutson (2010); Averbeck et al. (2014)

Stereological quantification

Quantitative analyses were performed by computer-assisted stereology using a Dell workstation receiving live video feed from an Optronics MicroFire color video camera (East Muskogee, OK) attached to a Nikon 80i light microscope, equipped with a Ludl MAC5000 motorized stage and a Heidenhain z-axis encoder. The optical dissector probe was used in conjunction with fractionator sampling using StereoInvestigator software (v. 10, MBF Bioscience, Williston, VT). Regions of interest were traced at 4× magnification (N.A. 0.75), and a square grid no larger than 1,000 μm² was superimposed in random orientation over the region of interest. Six sections per subject spanning the rostrocaudal extent of the head of the caudate nucleus, putamen, and nucleus accumbens regions (800 μm apart) were sampled for density of Ch+ and PV+ interneurons. Interneurons were counted at 20× magnification (N.A. 0.75) in a 350 μm² counting frame with a dissector height of 9 μm, with a coefficient of error (Gundersen m = 1) less than or equal to 0.1. Interneurons were counted only if they had a well-defined, darkly stained soma that came into view within the counting frame and initial segments of at least one visibly stained process. Incomplete neurons were excluded and comprised only a small proportion of stained neurons (typically less than 2 in approximately 20–25 sampling sites per section). Soma area for interneurons was additionally measured in one in every five interneurons counted using the Nucleator probe with a 4-point array in StereoInvestigator. Total density of interneurons was calculated as the estimated population divided by the planimetric volume of the region sampled. We additionally compared the density of Ch+ and PV+ interneurons relative to the total density of neurons for all regions of interest, as previously determined from Nissl-stained sections and quantified using previously published parameters (Hanson et al. 2018a, b) in the newly sampled subjects and territories. Additional stereological parameters, including average number of sampling sites and coefficient of error by region of interest, are listed in Tables 2 and 3.

Table 2.

Stereological sampling parameters for Ch+ interneurons

Region of interest	Diagnosis	Mean no. sampling sites	Mean no. neurons counted	Mean CE value (Gundersen m = 1)
dC	WS	256	117	0.10
	NT	238	108	0.09
mC	WS	430	125	0.10
	NT	324	154	0.09
dP	WS	274	100	0.10
	NT	215	95	0.10
mP	WS	321	116	0.09
	NT	286	124	0.10
NAcc	WS	211	86	0.10
	NT	227	146	0.09

Table 3.

Stereological sampling parameters for PV+ interneurons

Region of interest	Diagnosis	Mean no. sampling sites	Mean no. Neurons counted	Mean CE value (Gundersen m = 1)
dC	WS	223	139	0.09
	NT	293	153	0.09
mC	WS	225	107	0.09
	NT	330	167	0.10
mP	WS	279	141	0.09
	NT	235	121	0.09
dP	WS	219	100	0.10
	NT	249	81	0.10

Statistical analyses

All statistical analyses were performed using GraphPad Prism (Version 7, La Jolla, CA). Two-way ANOVAs were calculated (α < 0.05) to compare the density of Ch+ and PV+ interneurons with region of interest and condition (WS v. NT) as independent variables. All data were subject to Grubb’s test (GraphPad test for outliers) to determine if any single value represented a significant outlier from others by diagnostic category (WS or NT), but no outliers were detected. The percent of Ch+ and PV+ interneurons relative to the total density of neurons was additionally compared across all regions. Mann–Whitney U tests were utilized as post hoc tests to compare values for interneuron densities, percentages, and soma area in each region of interest between WS and NT subjects.

Results

Results for the density of Ch+ and PV+ interneurons in all five areas are summarized in Figs. 4 and 5. Table 4 summarizes average interneuron densities and the percentage of total neurons these account for in all areas. For Ch+ interneurons, ANOVAs revealed significant effects of region of interest (F_4,53 = 6.75, p = 0.0002) and diagnosis (F_1,53 = 12.89, p = 0.0007), though no interaction between diagnosis and region of interest was observed (F_4,53 = 0.69, p = 0.60). A significant effect of region of interest was observed for PV+ interneuron density (F_3,37 = 3.36, p = 0.03), but no significant effect of diagnosis (F_1,37 = 1.50, p = 0.23) or interaction between diagnosis and region of interest (F_3,37 = 1.43, p = 0.25) was found. The percent of Ch+ neurons as a proportion of total neurons showed no significant interaction between region of interest or diagnosis (F_4,53 = 0.64, p = 0.63), but significant effects of region of interest (F_4,53 = 4.94, p = 0.002) and diagnosis (F_1,53 = 5.05, p = 0.03) were observed. The percent of PV+ interneurons as a proportion of total neurons showed no significant interaction between region of interest and diagnosis (F_3,43 = 0.33, p = 0.84), and no significant effect of diagnosis (F_1,43 = 0.19, p = 0.66), but a significant effect of region of interest (F_3,43 = 3.61, p = 0.02) was observed.

Fig. 4.

Results of stereological analyses for Ch+ interneuron density. A significantly lower density of Ch+ interneurons was found in the medial caudate (mC) and nucleus accumbens (NAcc) regions, but not in the putamen

Fig. 5.

Results of stereological analyses for PV+ interneuron density. No significant difference in the density of PV+ interneurons was observed between WS and NT subjects

Table 4.

Mean density of interneurons sampled and percent of total neurons by region of interest

	Dorsal caudate	Medial caudate	Dorsal putamen	Medial putamen	Nucleus accumbens
NT Ch+ density % total neuron density	194 ± 22.5 0.82%	176 ± 21.0 0.67%	262 ± 37.8 1.04%	158 ± 20.6 0.51%	230 ± 22.2 0.72%
WS Ch+ density % total neuron density	175 ± 22.2 0.78%	108 ± 10.3 0.41%	204 ± 24.7 0.66%	136 ± 16.2 0.53%	167 ± 12.6 0.53%
% Difference in Ch+	−11.64%	−51.58%*	−0.22%	+ 0.01%	−40.13%*
NT PV density % total neuron density	294 ± 47.3 1.06%	190 ± 36.1 0.78%	159 ± 19.7 0.66%	195 ± 30.6 0.68%	Not measured
WS PV+ density % total neuron density	209 ± 20.1 1.02%	153 ± 20.3 0.65%	169 ± 21.4 0.67%	146 ± 5.3 0.62%	Not measured
% Difference in PV+	−33.74%	−31.8%	+ 0.07%	−27.3%	Not measured

^*indicates a significant decrease in density

Dorsal caudate

The highest average values for Ch+ density in the caudate were found in the dorsal portion. No difference in the density of Ch+ interneurons (mean ± SEM) was observed between WS (175 ± 22.2/mm³) and NT (194 ± 22.5/mm³) subjects (Mann–Whitney U test, p = 0.53). Soma area for Ch+ interneurons also did not differ significantly (837 ± 58.5 μm² WS vs. 801 ± 52.8 μm² NT, p = 0.55). No significant difference was observed in PV+ interneuron density (209 ± 20.1/mm³ WS vs. 294 ± 47.3/mm³ NT, p = 0.23) or soma area (283 ± 13.5/mm³ WS vs. 286 ± 12.7/mm³ NT, p = 0.85). No difference in the percent of total neurons comprised by PV+ (1.02% WS vs. 1.06% NT, p = 0.95) or Ch+ (0.78% WS vs. 0.82% NT, p = 0.84) neurons was observed in this region.

Medial caudate

A significant decrease in the density of Ch+ interneurons was observed in the medial caudate nucleus of WS individuals compared to NT (108 ± 10.3/mm³ in WS vs. 176 ± 21.0/mm³ NT; p = 0.0082). Average Ch+ soma area did not differ significantly between NT (731 ± 78.2 μm²) and WS (768 ± 57.0 μm², p = 0.84). No significant difference in the density (153 ± 20.3/mm³ WS vs 190 ± 36.1/mm³, p = 0.63) or soma area (276 ± 11.1 μm² in WS vs. 287 ± 18.9 μm² NT; p = 0.66) of PV+ interneurons was observed. A significant decrease in the percent of total neurons comprised by Ch+ neurons was observed (0.41% WS vs. 0.67% NT, p = 0.01), but no significant difference was observed in percent of PV+ interneurons (0.65% WS vs. 0.78% NT, p = 0.63).

Dorsal putamen

Greater average density of Ch+ interneurons was observed in the dorsal territory of the putamen as compared to its medial territories. No significant difference in the distribution of PV+ (159 ± 19.7 NT vs. 169.4 ± 21.4 WS, p = 0.63) or Ch+ (262 ± 37.8 NT vs. 204 ± 24.7, p = 0.24) interneurons or soma area was observed between NT and WS. No difference was observed in total soma area for Ch+ (722 ± 32.0 WS vs. 754 ± 21.7 NT, p = 0.59) or PV+ (338 ± 7.1 WS vs. 312 ± 8.2 NT, p = 0.23) neurons. Percent of total neurons represented by Ch+ (0.66% WS vs. 1.04% NT, p = 0.72) and PV+ (0.67% WS vs. 0.66% NT, p = 0.63) was not significantly different.

Medial putamen

The lowest density of Ch+ interneurons in NT subjects was observed in the associative region of the medial putamen, consistent with previously reported average values for its precommissural territories (Bernácer et al. 2007). Average values for Ch+ interneuron density did not differ significantly between NT and WS subjects (158 ± 20.6/mm³ NT vs. 136 ± 16.2/mm³ WS, p = 0.54). Though average density for PV+ interneurons was 27.3% lower in WS as compared to NT subjects, this difference was not statistically significant (146 ± 5.3/mm³ WS vs. 195 ± 30.6/mm³, p = 0.0513), nor did they comprise a smaller percentage of total neurons (0.62% WS vs. 0.68% NT, p = 0.95). No significant difference in soma area was observed for PV+ (302 ± 11.0 μm² WS vs. 320 ± 2.4 μm², p = 0.40) or Ch+ neurons (746 ± 76.3 μm² NT vs. 751 ± 61.4 μm² WS; p = 0.90) in this region.

Nucleus accumbens

Average density of Ch+ interneurons was significantly lower in WS subjects (167 ± 12.6/mm³ WS vs. 230 ± 22.2 NT, p = 0.04), as was the percent of total neurons staining positive for Ch+ (0.53% WS vs. 0.72% NT, p = 0.01). Ch+ soma area did not differ significantly in WS (757 ± 57.1 μm²) as compared to NT subjects (686 ± 51.6 μm², p = 0.3859). PV+ interneurons were not quantified in the NA region due to their sparse distribution.

Discussion

In the present work, our aim was to analyze the density of Ch+ and PV+ interneurons in the rostral territories of the striatum in WS as compared to NT subjects. The regions of interest examined here comprise a significant portion of the striatum rostral to the anterior commissure, and were targeted for their connectivity with several areas of the PFC: orbitofrontal (BAs 11, 13, 14, and 12), anterior cingulate and ventromedial (BAs 24b, 25, and 32) and dorsal (BA 9/46) areas of the PFC (Haber 2003; Haber et al. 2006; Haber and Knutson 2010). Ch+ interneuron densities in NT subjects for all regions of interest reported here lie within the range of variation previously reported for neurotypical subjects for the regions of interest (Bernácer et al. 2007; Kataoka et al. 2010). Values for PV+ interneuron density in NT subjects also fell within the range of variation previously reported (Kataoka et al. 2010; Bernácer et al. 2012; Lecumberri et al. 2018). In WS, we found a significant reduction in the density of Ch+ interneurons in the medial caudate nucleus and nucleus accumbens, and no difference compared to NT for PV+.

These findings add to a growing body of literature that point to significant changes in WS in structures guiding social responses and inhibitory behavioral control, suggesting that the circuits critically involved in these behaviors may be compromised. Our previous analyses (Hanson et al. 2018b) have indicated that the caudate nucleus in WS is characterized by an excess of glia, and particularly oligodendrocytes, which may relate to local differences in connectivity with specific areas of the PFC. These findings are particularly interesting given previous reports (Lew et al. 2017) of a decrease in neuron density in the orbital PFC of WS individuals (specifically, BA 11, which projects to the mC) that was not observed in primary somatosensory, motor, or visual cortex (BAs 3, 4, and 18). This area, as well as BA 10, also displayed an altered pattern of dendritic complexity in pyramidal neurons as compared to NT subjects (Hrvoj-Mihic et al. 2017). Additionally, BA 25 of the ventromedial PFC, which projects to the most ventral territories of the medial caudate and nucleus accumbens, showed decreased neuronal density in infragranular layers, and a substantially greater density of glia in WS (Wilder et al. 2018). These key regions of the PFC form important connections with the striatum at the specific sites where we observed decreased Ch+ neuron density, supporting the assertion that frontostriatal systems are particularly affected in the WS neuroanatomical phenotype.

Our observations in the striatum of individuals with WS can be considered in the context of other neuropsychiatric disorders where striatal interneurons are found to be decreased in density (Table 5). A decrease in Ch+ interneuron density has been observed in schizophrenia, which is characterized by disruptions in emotional regulation, deficits in reward prediction, and generalized anhedonia, attributed in part to abnormalities of the midbrain dopaminergic system affecting innervation of the basal ganglia (Perez-Costas et al. 2010). The decrease is most notable in the ventral striatum, inclusive of the most ventral territories of the caudate and nucleus accumbens regions (Holt et al. 1997, 2005). These areas are also affected in our WS sample. Depletion of Ch+ is found in Tourette syndrome (TS), a disorder characterized by the persistence of multiple involuntary motor and vocal tics. In TS, this depletion is found primarily in associative and sensorimotor territories of the dorsal striatum (Kataoka et al. 2010). By contrast, we found a decrease of Ch+ interneurons specific to the medial caudate nucleus in WS, with no significant differences observed in other territories. Specific deficits of Ch+ interneurons were noted in early immunohistochemical studies (Selden et al. 1994a) of Alzheimer’s disease (AD) in the ventral striatum, where a dramatic (> 70%) decrease in numbers was observed in a two-dimensional study of sections stained for acetylcholinesterase. Nevertheless, given recent advances in the study and complexities of the disease, the significance of these findings to contemporary insights of AD is complex (see Craig et al. 2011; Nava-Mesa et al. 2014; Sanabria-Castro et al. 2017 for review). Ch+ interneurons seem to be differentially affected in the limbic territories in the striatum in AD, which may relate to the high density of neurofibrillary tangles observed in the nucleus accumbens (Selden et al. 1994b) but not in the head of the caudate nucleus.

Table 5.

Summary of evidence for aberrant striatal interneuron distribution in selected neurological disorders in humans

Disorder	Key finding	Citation
Williams syndrome (WS)	Decreased density of ChAT+ neurons in medial caudate and accumbens (IHC; postmortem stereological study)	Present study
Autism spectrum disorders (ASD)	Decreased density of small CR+ neurons in caudate	Adorjan et al. (2017)
Schizophrenia	Reduced density of Ch+ interneurons in medial caudate, ventral striatum	Holt et al. (1997, 2005)
Tourette’s syndrome (TS)	Reduced density of striatal PV+ and Ch+ interneurons in associative/sensorimotor territories of dorsal striatum	Kalanithi et al. (2005), Kataoka et al. (2010)
Alzheimer’s disease (AD)	Decreased density of Ch+ neurons in ventral striatum (by area)	Lehéricy et al. (1989), Selden et al. (1994a)

Data for striatal PV+ interneuron densities do not exist in the literature for autism spectrum disorder (ASD), but a reduced density of small, calretinin-positive (CR) interneurons was found in the caudate nucleus in ASD as compared to neurotypical subjects (Adorjan et al. 2017). CR+ neurons comprise the most predominant cell type in the striatum in primates, and display a range of morphotypes with diverse properties (Wu and Parent 2000). Approximately 90% of the large CR+ interneurons in the striatum are cholinergic interneurons, but these large neurons did not show a similar pattern of reduced density in ASD subjects examined. Reduction of small GABAergic CR+ interneurons may relate to variants observed in the gene CNTNAP2 in ASD, which was associated with decreased CR+ neurons in the cerebral cortex in an animal model (Peñagarikano et al. 2011). One postmortem study has additionally confirmed that CR+ neurons are reduced in number in BA 9 in ASD (Hashemi et al. 2017). CNTNAP2 is not a part of the deletion observed in WS, and its variation in ASD may identify a subset of cases within this heterogeneous disorder with diverse genetic etiology. Nevertheless, calretinin-positive neurons may represent a further site of altered neuronal distribution in WS and provide a future direction for ongoing research into inhibitory systems in the disorder.

The medial caudate corresponds to territories that form networks important for decision-making, motivation, and reward (Haber and Behrens 2014). This region is additionally important for language function and vocalizations, and corresponds to territories where humans with dysfunctional FOXP2 genes display hyperactivation relative to NT participants in fMRI and PET language tasks (Vargha-Khadem et al. 1998; Liégeois et al. 2003). Although language function in WS is typically referred to as preserved relative to other cognitive faculties, such as visuospatial cognition (Nakamura et al. 2001; Farran and Jarrold 2003; Atkinson et al. 2003, 2007; Meyer-Lindenberg et al. 2004), global IQ (Searcy et al. 2004), and working memory (Morris and Mervis 2000; Robinson et al. 2003; Sampaio et al. 2008, 2010), select abnormalities of language use and function are observed in the disorder, including deficiencies in syntax, phonology, and lexical semantics (Brock 2007) as well as spatial and temporal elements of language (Karmiloff-Smith et al. 1997; Phillips et al. 2004). Additional study of neurotransmitter innervation may be of particular interest for understanding these language deficits in WS, given the role the putamen may play (Booth et al. 2007; Balsters et al. 2017; Viñas-Guasch and Wu 2017) in language processing and production. Though no significant difference was observed in the total density of PV+ interneurons, previous research (Fan et al. 2017) has suggested that gross differences in putamen volume may contribute to the disorder’s distinctive phenotype.

In addition to language functions, the medial caudate may be particularly important for social interaction and language, and has been the site of important neuroanatomical changes during human evolution (Raghanti et al. 2008). While the whole striatum has undergone a substantial reduction in relative size in the human brain in recent primate evolution (Barger et al. 2014), significant increases in dopaminergic innervation of the medial caudate have been observed in humans as compared to anthropoid primates, including apes (Raghanti et al. 2016). While cholinergic innervation did not differ across taxa in the medial caudate, humans and apes display a significant predominance of multipolar Ch+ as compared to unipolar and bipolar morphotypes, which displayed a higher frequency in monkeys (Stephenson et al. 2017). Though basal ganglia components are frequently regarded as being highly conserved, these differences may indicate enhanced synaptic plasticity and neurotransmission, which may relate to the evolution of language and reward systems.

Patterns of morphological differences in ASD serve as an interesting counterpoint to those observed in WS, particularly in their contrasting behavioral phenotypes and genetic profiles. Duplications in the deleted region in WS have been shown to result in a canonically autistic behavioral phenotype characterized by language delays and reduced social behavior (Berg et al. 2007; Velleman and Mervis 2011; Sanders et al. 2012), suggesting disparate effects of the gene deletion or duplication in WS and ASD respectively, with implications for the underlying neuroanatomy. Structural imaging studies have revealed increased caudate volume (Sears et al. 1999; Langen et al. 2007) and early overgrowth of the caudate in ASD (Langen et al. 2012), which is strongly associated with repetitive behaviors in the disorder (Estes et al. 2011) and not commonly observed in WS (Rodgers et al. 2012). Decreased frontostriatal connectivity in the caudate nucleus has also been observed (Di Martino et al. 2011) in ASD, and functional imaging studies have noted reduced frontostriatal activation (Shafritz et al. 2008), as well as aberrant processing of reward and reduced connectivity in the nucleus accumbens (Delmonte et al. 2013), particularly related to social stimuli (Kohls et al. 2013). Whereas reduced neuron density is observed in the caudate and nucleus accumbens in ASD (Wegiel et al. 2014), no differences in total neuron density were observed in these structures in WS, but an excess of glia, and particularly oligodendrocytes, was found (Hanson et al. 2018b) in the caudate, suggesting abnormalities of cellular distribution that may contribute to frontostriatal dysfunction. Additionally, an increase in the number of neurons in the lateral nucleus of the amygdala has been observed in WS (Lew et al. 2018), whereas this region showed evidence for a decrease in neuron number in ASD (Schumann and Amaral 2006). The amygdala also demonstrates a disparate pattern of serotonergic innervation in WS and ASD, including an increased density of serotonergic axons in ASD and a reduced density in WS as compared to NT individuals (Lew et al. 2020). In light of these contrasting patterns in microstructure, additional perturbations of neural circuitry may be seen in the striatum in ASD cases.

Conclusion

The present study targeted PV+ and Ch+ interneurons in territories of the striatum in WS due to their putative role in social approach behavior. In WS, we found a significant decrease in cholinergic interneuron density in the medial caudate nucleus that likely contributes to the reduced inhibitory control of social behavior, in a specific territory of the striatum targeted for its role in social cognition. Multiple neurological disorders involve perturbations of interneuronal circuitry, which is likely important for social behavior, emotional regulation, and behavioral control. Decreases in interneuron density in neurological disorders follow distinctive patterns that seem to relate to shared connectivity with cortical regions and modulation of cortical afferents. The topographical organization of the striatum, and particularly the caudate nucleus, by projections from regions of the PFC and other areas indicates that a more nuanced parcellation of its territories and variation in the distribution of neurons in these territories may help to elucidate variation in specific frontostriatal circuits contributing to the distinctive pathophysiology of select neurological disorders. For closely related disorders like WS and ASD, understanding variation in neuronal distribution in discrete functional territories may help to elucidate perturbations of neural circuitry in specific circuits and connectivity pathways, which will help to clarify the genetic and developmental mechanisms underlying these disorders.