Basal Ganglia Shape Features Differentiate Schizoaffective Disorder from Schizophrenia

Abstract

There is growing evidence that schizophrenia and schizoaffective disorder represent closely related syndromes that vary in severity along a neurobiological continuum. In the present study, volume and shape of the basal ganglia was examined in people with schizophrenia and schizoaffective disorder relative to healthy controls and hypothesized that unique neuroanatomical differences would be observed in each patient group. Magnetic resonance 1.5T images were obtained from schizophrenia (n=47), schizoaffective disorder (n=15), and from healthy control (n=42) participants, matched for age, gender, parental socioeconomic status, and race. The caudate, putamen, and globus pallidus were characterized using high-dimensional brain mapping procedures (Csernansky et al., 2004b). Results revealed significant shape deformations between schizophrenia and schizoaffective disorder that also differed from control subjects. Relative to schizophrenia, schizoaffective subjects showed exaggerated inward deformations indicative of localized volume loss in subregions of the caudate, putamen, and globus pallidus (all p<0.001). These shape features correlated with mental flexibility and negative symptoms in schizophrenia (all p<0.05), but not schizoaffective disorder. To the extent that differences in important basal ganglia substructures reflect biological heterogeneity among these two psychotic illnesses, this data could prove useful in improving diagnostic precision, as well as informing the affective component of mental illness.

1. Introduction

Recent scientific efforts are aimed at refining categorization of psychiatric disorders based on converging evidence from genetic and neurobiological studies suggesting their relatedness (Quinlan et al., 2020). For example, early family studies provided the initial support for the idea of a “schizophrenia spectrum” of illnesses (Kendler and Diehl, 1993). More recently, genome-wide association studies of large populations of patients with schizophrenia and affective psychoses suggest that there is substantial overlap in the alleles associated with their risk (Lichtenstein et al., 2009) despite clear biological distinctions (Wolfers et al., 2021, 2018). In addition, studies of cognition in schizoaffective disorder and schizophrenia demonstrate similar impairment profiles relative to healthy individuals (Barch, 2009). However, some findings suggest the presence of distinguishable neurobiological abnormalities that support the conceptual and categorical separation of the syndromes (Smith et al., 2011).

The basal ganglia are a set of deep-brain nuclei topographically connected to the cortex via the thalamus (Alexander and Crutcher, 1990) and participate in the regulation of motor functions, cognition, and emotional responses (DeLong and Wichmann, 2007). Initial reports indicated basal ganglia volumes are reduced in schizophrenia (Lawrie and Abukmeil, 1998; Shenton et al., 2001); however, recent meta-analyses suggest enlargement of these structures is characteristic of the disease (Erp et al., 2016; Okada et al., 2016). Furthermore, a small number of recent studies have observed changes in the surface shape of the basal ganglia in schizophrenia, which highlight localized abnormalities in these structures (Mamah et al., 2016).

In contrast, relatively little is known about basal ganglia integrity in schizoaffective disorder. Most studies of psychosis tend to assign individuals with schizophrenia and schizoaffective disorder into a single, unitary clinical group for analysis, thereby obscuring distinguishable features between the groups. However, recent efforts have been made to characterize the features of schizoaffective disorder as a distinct condition (e.g., the Bipolar-Schizophrenia Network on Intermediate Phenotypes; (Tamminga et al., 2013). A thorough understanding of the unique neurobiological features in schizoaffective disorder relative to other forms of psychosis is especially relevant given the affective dysregulation associated with the disorder.

The overarching aim of this study was to characterize both the volume and shape of three major basal ganglia nuclei—the caudate nucleus, globus pallidus and putamen—in three groups of individuals: schizophrenia, schizoaffective disorder, and healthy control subjects. Prior research on the comparison of clinical and cognitive features in these groups suggests individuals with schizoaffective disorder may experience attenuated deficits and symptom severity relative to schizophrenia (Heinrichs et al., 2008; Peralta and Cuesta, 2008; Smith et al., 2009). Based on this work, it was hypothesized that alterations in the degree of neuroanatomical change in schizophrenia and schizoaffective subjects would also follow a stepwise pattern (i.e., schizophrenia>schizoaffective disorder>controls). Similar findings have also been observed in the corpus callosum of first-episode affective and non-affective psychosis samples (Walterfang et al., 2009). However, because of the involvement of specific basal ganglia substructures in affective neural circuits, such as the caudate and globus pallidus (Bonelli and Cummings, 2007) it was expected that some regions would be preferentially affected in individuals with schizoaffective disorder, but not those with schizophrenia. Given these deep-brain regions are implicated in various behavioral processes, shape features were also examined relative to cognitive and clinical dimensions of the psychosis groups, with the prediction that greater abnormal shape deformation of the caudate and globus pallidus would relate to increased impairment in behavioral measures.

2. Methods

2.1. Sample

Participants included 47 individuals with schizophrenia (SCZ), 15 individuals with schizoaffective disorder (SA), and 42 control subjects (CON) from an existing NIH-funded dataset on neuromorphometry in schizophrenia. The recruitment process is previously described (Delawalla et al., 2006; Smith et al., 2009) with SA subjects recruited specifically for this study and diagnosed using the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID; First et al., 1996). A subset of SCZ and CON subjects were selected from a larger sample of available participants (Csernansky et al., 2004a) in order to specifically match against SA subjects with respect to age, gender, parental socioeconomic status, and race. Participants completed a battery of neuropsychological tests that evaluated cognitive domains relevant to the psychosis profile (Nuechterlein et al., 2004). All cognitive and symptom ratings were transformed to z-scores using the mean and standard deviation of the current sample and averaged within clusters. The project was approved by the IRB at Washington University, and informed consent was obtained from each subject after a complete description of the study was given. See online Supplementary Material for more detailed demographic information, including inclusion/exclusion criteria, cognitive/clinical measures, and medication history.

2.2. Surface mapping

Using magnetic resonance images (see Supplemental Materials for acquisition parameters), three specific regions of the basal ganglia were selected for examination in this study: caudate nucleus, putamen, and globus pallidus. Surfaces of these regions were generated by applying Large-Deformation High-Dimensional Brain Mapping (Csernansky et al., 2004b; Wang et al., 2001), which is an atlas-based procedure that utilizes diffeomorphic transformations to accurately align a template image of the structures to a target (i.e., subject) image that allows independent matching of individual surface points while preserving the unique morphological features of each subject (Beg et al., 2005; Miller et al., 2002). Reliability and validity of HDBM-LD for mapping the basal ganglia has been previously established (Mamah et al., 2008, 2007).

Localized shape differences were characterized by compiling deformation values, computed as a contrast from the sample mean, from triangulated surface points of each structure for all subjects into a principal components analysis (one per structure) to diminish the high dimensionality of the data, thus generating an orthonormal set of eigenvectors to represent variation in the shape of the left and right basal ganglia structures (Beg et al., 2005; Mamah et al., 2007). The bulk of variation in these surfaces was defined by the first 10 eigenvectors; caudate = 87%; globus pallidus = 92%; putamen = 88%.

2.3. Statistical analysis

First, X² and one-way ANOVA models were used to examine group differences on demographic variables. A multiple analysis of variance (MANOVA) model was used to assess group differences on cognition, while a standard one-way analysis of variance (ANOVA) model was used to examine differences in psychopathology between SCZ and SA. For analysis of basal ganglia volume, a repeated measures multivariate analysis of variance (RM-ANOVA) with group status as the between-subjects effect and hemisphere (left/right) as the within-subject effects was used for each region. Given patient group differences in anti-depressant use, the effect of this variable on basal ganglia volumes was examined using ANCOVA models.

To test for global surface shape differences across groups, RM-MANOVA models were used with group status as the between-subjects effect, with hemisphere (left/right) and eigenvector (top 10 from the principal components analysis) as the within-subject effects. Separate models were run for each region of the basal ganglia. If an interaction effect that included group was significant for a region, similar follow-up RM-ANOVA models were then conducted to identify significant differences between each group (i.e., SA vs. SCZ, SA vs. CON, SCZ vs. CON).

Visualization of shape deformation patterns were constructed with maps of the composite surface of each structure at every graphical vertex for the basal ganglia. Shape displacements were estimated at each surface point as the difference between the means of the group vectors in magnitude. Inward and outward displacements, or deformations, are estimated as representations of localized volume loss or exaggeration at the neurobiological level (Hanko et al., 2019). A human brain atlas was consulted to associate any observed deformation patterns to each basal ganglia region (Mai et al., 2008).

To characterize shape features for use in statistical models with cognitive and clinical variables, a maximum likelihood estimate of the linear predictor (i.e. xBeta) was generated from a logistical regression that predicted psychosis group status (i.e., SCZ and SA) per hemisphere for each structure in the basal ganglia. xBeta is a single score representing shape differences between 2 groups where low scores reflect SA shape and high scores represent SCZ shape. These values were entered into Pearson partial correlation models (controlling for the effects of the generalized deficit in schizophrenia) to examine their relationship with cognition and clinical symptoms in the patient groups only. Results from these analyses were then used to guide specific linear regression models to determine whether brain shape (xBeta as the independent variable) could predict cognition or clinical symptomatology in SCZ and SA.

2.4. Sensitivity power analysis

Sample sizes for the current study are fixed given data was derived from an archival longitudinal study of schizophrenia-spectrum disorders (Wang et al., 2008). Sensitivity power analyses using Cohen’s f were conducted for each ANOVA model above using G*Power (Faul et al., 2007) to determine the minimum effect that could be detected at 80% power and a type I error rate of 0.05 (Cohen, 2013). With a combined sample of 104 subjects in 3 groups, results from the sensitivity analysis revealed there was power to detect the following minimal Cohen’s f effect sizes for: cognition MANOVA = 0.09; psychopathology one-way ANOVA = 0.36; basal ganglia volume RM-ANOVA for interaction effects = 0.08; basal ganglia surface shape for interaction effects RM-MANOVA = 0.39. It appears the sample is adequate for addressing the proposed research questions with the ability to detect at least small to moderate effect sizes if present (Grove and Cipher, 2016). Cohen’s f values were calculated for each comparison above using criteria from Cohen (Cohen, 2013) and Lenhard & Lenhard (Lenhard and Lenhard, 2016).

3. Results

3.1. Participant characteristics

In terms of demographics (Table 1), the three groups did not differ on age, socioeconomic status, race, and gender (all p>0.40). In addition, SCZ and SA groups did not differ in duration of illness (p=0.97), as well as the number of individuals with previous substance use disorders (all p>0.65). Both 1^st-generation- and 2^nd-generation antipsychotic usage between SCZ and SA groups did not differ (all p>0.60). With respect to other medication use, the groups did not differ on mood-stabilizers (p=0.29), but there was a significant difference on antidepressant use in the SA group (80%) compared to SCZ (29.8%). There was a significant group effect for cigarette use (F_3,102=7.4, p<0.001), with SCZ and SA demonstrating higher rates than CON, but not differing from each other.

Table 1.

Demographic Characteristics of Study Sample

	scz (n = 47)		SA (n =15)		CON (n = 42)		Statistic
	Mean	(SD)	Mean	(SD)	Mean	(SD)	F-test	df	p
Age (years)	41.2	(12.3)	42.6	(9.8)	42.6	(12.1)	0.18	103	0.84
Duration of Illness (years)	18.0	(14.2)	17.9	(12.6)	--		0.002	61	0.97
Mean SES	3.6	(0.9)	3.4	(0.9)	3.3	(1.0)	0.80	95	0.43
Cigarette use (cigarettes per year)*	4908	(4811)	3750	(3971)	1595	(2968)	7.4	102	<0.001
Medication Use
1st-Generation (Dose Years)a	6.1	(7.9)	4.9	(2.1)	--		0.04	13	0.84
2nd-Generation (Dose Years)b	3.7	(3.5)	4.3	(4.1)	--		0.25	50	0.62
	N	(%)	N	(%)	N	(%)	X2	df	p
Gender, No. (% male)	26	(55.3%)	8	(53.3%)	23	(54.8%)	0.01	2	0.99
Race (% White)	23	(48.9%)	7	(46.7%)	23	(54.8%)	0.87	4	0.93
Substance Use Disorder
Alcohol	19	(40.4%)	7	(46.7%)	--		0.18	1	0.67
Cannabis	17	(36.2%)	5	(33.3%)	--		0.04	1	0.84
Cocaine	8	(17%)	4	(26.7%)	--		0.39	1	0.41
Stimulants	2	(4.3%)	1	(6.7%)	--		0.14	1	0.71
Hallucinogens	1	(2.1%)	1	(6.7%)	--		0.002	1	0.39
Sedatives	2	(4.3%)	--		--		0.66	1	0.42
Opioids	3	(6.4%)	--		--		1.01	1	0.32
Medication Use
Mood Stabilizer	7	(14.9%)	4	(26.7%)	--		1.08	1	0.299
Anti-Depressant	14	(29.8%)	12	(80.0%)	--		11.77	1	<0.001

^*SCZ > CON (p < 0.001); SA > CON (p = 0.081)

^aSCZ (n = 12), SA (n = 2)

^bSCZ (n = 38), SA (n = 13)

3.2. Cognitive performance and clinical symptoms

Regarding cognition (Table 2), there was an overall main effect for group (Λ_(10,178)=5.02, p<0.001). Among individual groups, significant Tukey mean differences were observed between SCZ and CON on Digit Span (DS; p<0.003), Letter-Number Sequencing (LNS; p<0.001), Spatial Span (SS; p<0.002), Trail Making Test-B (TMT-B; p<0.001), and Wisconsin Card Sorting Test (WCST; p<0.003). Differences were also observed between SA and CON on DS (p<0.04), LNS (p<0.01), and WCST (p<0.01). Between SA and SCZ groups, only TMT-B performance notably differed (p<0.002). Effects sizes for these comparisons were moderate to large. Regarding clinical psychopathology, SA and SCZ did not significantly differ on any aspect of symptomatology, although the effect size difference for negative symptoms was moderate (f=0.42).

Table 2.

Cognition and Psychopathology Standardized Scores, MANOVA Results and Effect Sizes

	scz ( n = 47 )	SA ( n = 15 )	CON ( n = 42 )	3-Group MANOVA
	Mean (SD)			F	df	p	Cohen’s f
Cognition				Λ = 5.02	10	<0.0001
WAIS-III Vocabulary	−0.38 (0.96)*	−0.004 (0.84)	0.49 (0.93)	9.54	2	<0.0001	0.45
WMS-III Digit Span	−0.35 (0.89)*	−0.35 (0.52)*	0.31 (0.97)	6.54	2	0.01	0.37
WMS-III Letter-Number	−0.62 (0.95)*	−0.19 (0.67)*	0.53 (0.76)	19.46	2	<0.0001	0.62
Sequencing
WMS-III Spatial Span	−0.36 (0.84)*	−0.14 (0.83)	0.32 (0.88)	6.53	2	0.003	0.37
Trail Making Test, Part B	−0.73 (1.14)* †	0.20 (0.45)	0.35 (0.61)	17.04	2	<0.0001	0.68
Wisconsin Card Sorting Test – Perseverative Errors	−0.51 (1.21)*	−0.63 (1.32)*	0.27 (0.61)	7.66	2	0.001	0.41
					One-Way ANOVA
Psychopathology
Positive Symptoms	0.12 (0.78)	0.26 (1.12)	--	0.32	1	0.57	0.12
Negative Symptoms	0.25 (0.69)	−0.10 (0.58)	--	3.28	1	0.07	0.42
Disorganized Symptoms	0.16 (0.61)	0.16 (0.64)	--	0.00	1	0.99	0.00

^*= differs from CON at p < 0.05

^†= differs from SA at p < 0.05; based on Tukey post-hoc analyses

3.3. Volume analysis

Right and left basal ganglia volumes (Supp. Table 1) in the RM-ANOVA models revealed no main effect for group on the volume of the caudate nucleus (F_2,101=0.1, p=0.93), putamen (F_2,101=0.3, p=0.77), or globus pallidus (F_2,101=0.6, p=0.52). Effect of brain hemisphere was significant for the caudate nucleus (Left>Right: F_1,101=5.1, p=0.03) and putamen (Left>Right: F_1,101=16.7, p<.001) but not the globus pallidus (F_1,101=0.02, p=0.96). No significant antidepressant medication-by-group interactions were present using ANCOVA models.

3.4. Shape analyses

Caudate.

RM-MANOVA results (Table 3) revealed a significant group-by-eigenvector interaction (λ=0.54, F_18,186=3.7, p<0.001). Follow-up comparison models revealed significant group-by-eigenvector interactions between SA and SCZ (F_9,51=5.8, p<0.001), SA and CON (F_9,47=6.7, p<0.001), but not SCZ and CON (F_9,79=1.3, p=0.26). The SA and SCZ comparison was covaried for use of antidepressant medication, which did not significantly influence the model (F_1,59=1.2, p=0.29).

Table 3.

RM-ANOVA Models of Basal Ganglia Shape Characteristics

	3-Group MANOVA
	Group x Eigenvector			Group x Hemisphere			Group x Hemisphere x Eigenvector
Structure	F	df	p	F	df	p	F	df	p	Cohen’s f* LH/RH
Caudate Nucleus	Λ = 0.541	18	<0.001	Λ = 0.934	2	0.032	Λ = 0.672	18	0.003
SCZ vs SA	5.88	18	<0.001	5.13	2	0.027	3.82	18	0.001	0.34/0.19
SA vs CON	6.73	18	<0.001	6.14	2	0.016	1.22	18	0.301	0.18/0.03
SCZ vs CON	1.28	18	0.257	0.00	2	0.965	1.99	18	0.051	0.18/0.18
Putamen	Λ = 0.295	18	<0.001	Λ = 0.929	2	0.024	Λ = 0.604	18	<0.001
SCZ vs SA	12.55	18	<0.001	0.34	2	0.558	1.94	18	0.067	0.06/0.04
SA vs CON	25.56	18	<0.001	2.18	2	0.145	3.49	18	0.002	0.05/0.06
SCZ vs CON	1.91	18	0.062	6.95	2	0.010	2.97	18	0.004	0.10/0.01
Globus Pallidus	Λ = 0.476	18	<0.001	Λ = 0.954	2	0.095	Λ = 0.592	18	<0.001
SCZ vs SA	6.77	18	<0.001	2.42	2	0.125	2.85	18	0.008	0.74/0.39
SA vs CON	7.92	18	<0.001	3.78	2	0.057	4.14	18	0.001	0.65/0.20
SCZ vs CON	1.38	18	0.208	1.82	2	0.180	1.45	18	0.178	0.17/0.23

^*= calculated from weighted averages of all ten eigenvectors for each hemisphere

Visual representations of caudate shape (Figure 1) revealed SA was characterized by inward deformation along dorsal aspects of the caudate and outward deformation in ventral regions when compared with CON. Inward deformations were also observed along the dorsal, but also anterior ventral surface when compared with SCZ. While not significant in the ANOVA models, SCZ shape was noted as having modest inward deformations along the dorsal surface and outward deformation in anterior ventral regions of the caudate.

Figure 1.

Caudate nucleus surface shape displacement maps between: schizophrenia (SCZ) patients and control (CON) participants; schizoaffective (SA) patients and CON participants; SA and SCZ patients. Cooler colors indicate significant regions of inward shape differences and warmer colors indicate significant regions of outward shape differences.

Putamen.

RM-MANOVA across all groups revealed a significant group-by-eigenvector interaction (λ=0.516, F_18,186=8.69, p<0.001). Follow-up comparisons found significant differences between SA and SCZ (F_9,51=12.6, p<0.001), as well as SA and CON (F_9,47=25.6, p<0.001), but a non-significant difference between SCZ and CON (F_9,79=1.9, p=0.06). Antidepressant medication did significantly affect the SA/SCZ comparison (F_1,59=18.2, p<0.001) and was included in the model.

Visualization of putamen shape differences (Figure 2) revealed prominent inward deformation in ventral-lateral, ventral-posterior and dorsal aspects in SA relative to SCZ and CON. In addition, greater outward deformation in lateral posterior areas were also observed. In SCZ, some anterior-dorsal inward deformations were noted in the right putamen relative to CON with some medial posterior involvement.

Figure 2.

Putamen surface shape displacement maps between: schizophrenia (SCZ) patients and control (CON) participants; schizoaffective (SA) patients and CON participants; SA and SCZ patients. Cooler colors indicate significant regions of inward shape differences and warmer colors indicate significant regions of outward shape differences.

Globus Pallidus.

RM-MANOVA results for all groups revealed a significant group-by-eigenvector interaction (λ=0.48, F_18,186=4.6, p<0.001). Follow-up comparisons found significant differences between SA and SCZ (F_9,51=5.55, p<0.001) and SA and CON (F_9,47=7.92, p<0.001) groups, but none between SCZ and CON (F_9,79=1.4, p=0.21). Covarying for antidepressant medication use in the SA and SCZ comparisons did not significantly affect the model (F_1,59=0.96, p=0.33).

Visual inspection of the globus pallidus (Figure 3) revealed prominent SA-related inward shape deformations in anterior, dorsal and ventral posterior regions relative to both CON and SCZ. While no significant shape differences were noted in the ANOVA models for SCZ, some mild inward deformation was observed in the anterior aspects of the globus pallidus.

Figure 3.

Globus pallidus surface shape displacement maps between: schizophrenia (SCZ) patients and control (CON) participants; schizoaffective (SA) patients and CON participants; SA and SCZ patients. Cooler colors indicate significant regions of inward shape differences and warmer colors indicate significant regions of outward shape differences.

3.5. Shape relationships with cognition and clinical symptoms

Descriptive statistics of xBeta values revealed mean residual values in the negative range for SA, and in the positive range for SCZ in all structures (Supp. Table 2). Regarding cognition, there were no significant relationships between basal ganglia shape features and DS, SS, LNS, or WCST in patients (all p=0.07 to 0.94). However, negative correlations were noted between TMT-B and shape of all structures bilaterally, including the caudate (L: r=−0.25, p=0.055; R: r=−0.25, p=0.050), putamen (L: r=−0.47, p<0.001; R: r=−0.38, p=0.003), and globus pallidus (L: r=−0.31, p=0.013; R: r=−0.47, p<0.001). The direction of these inverse correlations suggested SCZ-related basal ganglia shape related with poorer TMT-B performance relative to SA (see Figure 4A). These relationships were maintained even after controlling for the global cognitive deficit in the correlation models.

Figure 4.

Basal ganglia shape progressing from schizoaffective (SA) to schizophrenia (SCZ) correlated with A) poorer TRAILS B performance including the caudate (L: r = −0.25, p = 0.055; R: r = −0.25, p = 0.050), putamen (L: r = −0.47, p < 0.001; R: r = −0.38, p = 0.003), and globus pallidus (L: r = −0.31, p = 0.013; R: r = −0.47, p < 0.001); and B) increased negative symptoms including caudate (L: r = 0.26, p = 0.046; R: r = 0.29, p = 0.025), globus pallidus (L only: r = 0.31, p = 0.016) and globus pallidus (L only: r = 0.34, p = 0.007). CD = caudate nucleus, PU = putamen, PL = globus pallidus

Regarding clinical symptoms, Pearson correlations revealed no relationships between basal ganglia shape and positive or disorganized symptoms (all p=0.19 to 0.90). There was, however, consistent correlations with negative symptoms observed with caudate (L: r=0.26, p=0.046; R: r=0.29, p=0.025), globus pallidus (L only: r=0.31, p=0.016) and putamen (L only: r=0.34, p=0.007). The direction of these correlations indicate that negative symptoms worsen as shape of the structures becomes more SCZ-related (see Figure 4B).

Follow-up regression models were used to further explore the predictive relationship of shape for TMT-B and negative symptoms. Using significant basal ganglia regions from the correlation models above, results of a linear regression with shape as a predictor of TMT-B performance revealed a significant model (F_(6,55)=3.72, p=0.004, R²=0.288). However, a similar model examining the predictive ability of shape on negative symptoms was non-significant (F(4,57)=2.24, p=0.076, R²=0.136).

4. Discussion

The aim of this study was to investigate basal ganglia shape abnormalities in individuals with schizoaffective disorder, and whether they followed a pattern similar to, or different from that observed in schizophrenia. Based on previous studies comparing these groups on other types of neurobiological and cognitive variables (Barch, 2009; Smith et al., 2011), we expected basal ganglia abnormalities in schizoaffective disorder would be similar but attenuated relative to schizophrenia. This would be consistent with findings from studies on genetic influences in schizophrenia and schizoaffective disorder (Gershon et al., 2011; Kendler and Diehl, 1993). Overall, we found lack of support for this general hypothesis, and noted that while there were some commonalities, basal ganglia shape in schizoaffective disorder on average differed from schizophrenia and in some cases the abnormalities were exaggerated.

While no group differences in global volume of the caudate, globus pallidus and putamen were observed, a number of differences existed between groups in localized volume loss characterized by inward deformation of surface shape. Individuals with schizoaffective disorder significantly differed from both control and schizophrenia groups in the shape of all structures, while schizophrenia subjects differed from the control group only in shape of the putamen. Given the basal ganglia has been associated with regulation of emotion and cognition (Bonelli and Cummings, 2007), we investigated the relationship of shape abnormalities with working memory, executive functioning and clinical symptoms in the patient groups. A unique pattern emerged where schizophrenia-related shape abnormalities in all structures correlated with a measure of mental flexibility load and motor speed (TMT-B) and negative symptoms. Follow-up regression models revealed that shape abnormality predicted TMT-B performance, but not negative symptoms; indicating a possible substrate for deficits in adaptive and flexible thinking for motor speed in psychosis. These findings are revealing but also consistent with the growing body of literature on the use of computational methods to improve detection of subtle neurobiological changes in studies of neuropsychiatric disorders (Nguyen et al., 2016).

In the caudate, schizoaffective shape-related abnormalities were noted in dorsal and ventral anterior regions including the tail. While inward deformation on the dorsal body of the caudate were observed similarly in both psychosis groups, inward deformation on the anterior ventral surface was unique to schizoaffective subjects and may indicate a structural basis for affective disturbance. The ventral caudate, which intersects the caudate and ventral striatum, is densely interconnected with the limbic system (Draganski et al., 2008) and is associated with multiple affective functions such as reward processing (Benningfield et al., 2014), pain perception (Martikainen et al., 2015) and fatigue (Miller et al., 2014). In addition, stimulation of the ventral caudate is a potential source of treatment for affective conditions such as OCD and major depression (Aouizerate et al., 2004). Also interesting is shape deformations of the caudate similar to those found in the schizoaffective group were reported in a study of never-medicated individuals with bipolar disorder (Hwang et al., 2006). In addition, Mamah and colleagues (Mamah et al., 2016) found shape abnormalities in subjects with bipolar disorder that also mimicked those in our schizoaffective group. Our findings highlight the ventral caudate as an affective substrate unique to schizoaffective disorder and not schizophrenia.

Regarding the putamen, exaggerated inward deformations were noted bilaterally along anterior lateral, dorsal and posterior medial regions relative to controls; however, outward deformations were also observed in posterior lateral aspects. These findings however were observed to be exaggerated in schizoaffective disorder compared to schizophrenia. This finding is particularly interesting given the putamen’s role as part of densely connected frontal cerebral circuitry. While the majority of the putamen’s anatomy is well known for receiving excitatory inputs from motor regions (DeLong and Wichmann, 2007) it is also connected with prefrontal cortex. In particular, anterior putamen demonstrates connectivity with medial, ventral and dorsolateral prefrontal regions that play a major role in emotional/motivational states (Lehéricy et al., 2004; Postuma and Dagher, 2006). There is also converging evidence that the strength of anterior putamen connectivity with frontal regions is associated with aspects of affective regulation such as inhibitory control (Franken and Buitelaar, 2018; Zandbelt and Vink, 2010). The significant anterior ventral abnormalities in the putamen observed only in the schizoaffective group suggests alterations in this region may contribute to their affective disturbance. This interpretation is reinforced by recent reports describing prominent ventral putamen shape changes in individuals with untreated major depressive disorder (Lu et al., 2016) and bipolar disorder (Mamah et al., 2016). As was observed in the caudate, these overlapping findings implicate disruption of this particular region as a distinctive part of the pathophysiological process in affective disorders.

Results from the globus pallidus analyses showed no significant abnormalities in schizophrenia, while deformation in schizoaffective disorder was significantly different from that of either schizophrenia and control groups. This schizoaffective pattern consisted of inward deformations along the dorsal, ventral anterior, and posterior regions with some outward deformation in the posterior extremities. Again, the unique shape features in the schizoaffective group suggest features that are distinct to this group. As the main output relay in the basal ganglia, the globus pallidus is implicated in multiple motor and behavioral processes (Albin et al., 1989; DeLong and Wichmann, 2007). While previous work has found significant schizophrenia-related shape differences in the putamen and globus pallidus (Mamah et al., 2008), this study only found attenuated changes. This may reflect an effect due to group composition where all schizophrenia spectrum subjects were combined into a single group. Overall, the above findings in basal ganglia shape analysis demonstrate prominent shape abnormalities in schizoaffective disorder that are distinct from those observed in schizophrenia. In particular, inward deformation in specific regions on the ventral caudate, anterior ventral putamen, and dorsal regions of the globus pallidus set it apart from schizophrenia. These regions have connections with frontal and limbic systems known to subserve aspects of cognition and emotion implicated in affective regulation, and are also disrupted in other affective states such as depression and bipolar disorder.

Another interesting finding from this study pertains to the relationship between shape features and various aspects of behavior in schizophrenia. The only cognitive measure that consistently related to basal ganglia shape was TMT-B, which was also the only task where performance significantly differed between psychosis groups. Correlation models revealed TMT-B performance systematically decreased as bilateral shape in the caudate, globus pallidus and putamen resembled that of schizophrenia. A follow-up regression model reinforced this finding by demonstrating shape was also predictive of performance, suggesting these structures aid in the regulation of mental flexibility load and motor speed. Many lines of research indicate the importance of basal ganglia function in mental flexibility and set-shifting for both animal models (Crofts et al., 2001) and humans (Monchi et al., 2006; Ravizza and Ciranni, 2002). Given the basal ganglia has reciprocal connections to prefrontal and limbic cortices (DeLong and Wichmann, 2007; Lehéricy et al., 2004), the observed abnormalities suggest a structural basis for cognitive deficits represented in these psychotic disorders. Our findings indicate that despite the presence of unique basal ganglia shape abnormalities in schizoaffective disorder, these alterations do not relate to the (attenuated) cognitive impairment experienced by the group. Rather, cognition appears more susceptible to alterations in these regions in schizophrenia.

A similar pattern was observed regarding negative symptoms in the patient groups, where schizophrenia-related basal ganglia shape changes correlated with negative symptomatology. Although the observed correlation models were significant, the regressions results failed to demonstrate any predictive value of these relationships. Previous studies have observed relationships between basal ganglia integrity and the expression of negative symptoms (Spinks et al., 2005; Strub, 1989), supporting the modest evidence from our study that this is a feature primarily observed in schizophrenia.

Several conclusions may be drawn from these findings, particularly that unique shape features in schizoaffective disorder are not strongly supportive of a spectrum approach to severity in psychotic disorders. While some similarities were observed between the patient groups, indicating a shared component, prominent basal ganglia shape differences were noted in regions of the putamen and globus pallidus. This may be a function of separable disease processes or could also reflect the influence of antipsychotic medication effects. Previous work suggests typical antipsychotics (as a function of dosage, treatment time and individual sensitivity) may contribute to enlargement of basal ganglia structures, thus minimizing potential shape differences (Erp et al., 2016; Mamah et al., 2016). While our patient groups did not statistically differ on 1^st generation antipsychotic use, there was generally higher use in schizophrenia, which may explain some of their lack of inward deformation. Current approaches include combining schizophrenia and schizoaffective disorder into a single diagnostic category when conducting psychosis-spectrum research; however, the results of our study suggest exercising caution when doing so, particularly when examining subcortical features. This also raises the question of whether schizophrenia and schizoaffective disorder should be uniformly treated as distinct entities when attempting to characterize their pathophysiology or identify treatments targets.

One primary limitation of the study was the relatively small sample sizes of the groups. However, the sensitivity analysis revealed there was adequate power to detect even small-to-moderate effects in the proposed comparisons. Given many of the Cohen’s f effect sizes for the significant shape ANOVA models were moderate-to-large, the findings were not likely underpowered or spurious. As previously mentioned, the effect of antipsychotic medication on basal ganglia morphology may also play a role in our findings. However, some of this concern may be mitigated given both groups were relatively similar on 1^st and 2^nd generation usage, as well as the highly significant findings in the schizoaffective group. Furthermore, antidepressant use between the psychosis group did differ, and there is some indication that antidepressants can increase the volume of deep-brain structures, particularly the globus pallidus and putamen (Bykowsky et al., 2019). However, including antidepressant use as a covariate in our models revealed a significant influence only on shape of the putamen, suggesting a selective impact on that structure alone. Thus it appears that surface metrics may demonstrate some relative insensitivity to the effects of certain medications, which has been noted in prior work (Womer et al., 2014). Another potential limitation is that the MR images used were from a relatively low-field scanner (1.5-Tesla), which may have limited our ability to detect differences due to a weaker contrast-to-noise ratio. Finally, white matter abnormalities are known to exist in both of these populations in regions surrounding the basal ganglia (Szeszko et al., 2005), thus it is possible the abnormalities we observed may be a function of white matter pathology not captured by our approach.

The absence of differences among groups in global volume of the basal ganglia was somewhat surprising, particularly given recent meta-analytic studies that have observed enlargement of these structures in schizophrenia (Erp et al., 2016; Okada et al., 2016). The disparate finding is likely a function of our sample composition, as well as the increasingly recognized heterogeneity that exists in schizophrenia which has identified unique neurobiological subtypes (Chand et al., 2020). It is important to note that roughly half of the participants were African-American, representing an uncommonly diverse sample. The racial distribution of subjects in many larger or meta-analytic studies is unclear as this data is generally not reported, thus direct comparison of findings can be difficult given this complicating factor. It is likely our findings may be incorporating elements unique to minority populations, thus not completely aligning with studies where the majority of subjects are Caucasian.

In summary, our findings suggest that there is a component of structural change in the basal ganglia, more unique than common, in schizoaffective disorder relative to schizophrenia, but that may form a structural basis for cognitive deficits that is more prominent in schizophrenia. Specific differences may also suggest a basis for the affective dysregulation used to clinically distinguish the two disorders. The role of antipsychotic effects on shape changes in our study is unclear, and future studies in medication naïve patients would provide insights on this. The significance of our findings would be greatly extended should these changes correlate with other biological markers, such as genetic findings, neurochemical differences, or functional changes. Should such associations be present, they would shed light on the relationship between brain structure and clinical presentations and may possibly provide biological underpinnings for symptomatic heterogeneity among psychotic illnesses. Thus, these findings may someday allow for more accurate prediction of illness progression or risk and may even allow for the design of diagnosis-specific treatments.

Highlights.

• Schizoaffective disorder appears neurobiologically distinct from schizophrenia

• Surface-mapping reveals basal ganglia abnormalities in schizoaffective disorder

• Shape features relate to mental flexibility and negative symptoms in schizophrenia