Thalamic, Amygdalar, and Hippocampal Nuclei Morphology and their Trajectories in First Episode Psychosis: A Preliminary Longitudinal Study

Abstract

The thalamus, amygdala, and hippocampus play important pathophysiologic roles in psychosis. Few studies have prospectively examined subcortical nuclei in relation to predicting clinical outcomes after a first-episode of psychosis (FEP). Here, we examined volumetric differences and trajectories among subcortical nuclei in FEP patients and their associations with illness severity. Clinical and brain volume measures were collected using a 1.5T MRI scanner and processed using FreeSurfer 6.0 from a prospective study of antipsychotic-naïve FEP patients of FEP-schizophrenia (FEP-SZ) (baseline, n=38; follow-up, n=17), FEP non-schizophrenia (FEP-NSZ) (baseline, n=23; follow-up, n=13), and healthy controls (HCs) (baseline, n=47; follow-up, n=29). Compared to FEP-NSZ and HCs, FEP-SZ had significantly smaller thalamic anterior nuclei volume at baseline. Longitudinally, FEP-SZ showed a positive rate of change in the amygdala compared to controls or FEP-NSZ, as well as in the basal, central and accessory basal nuclei compared to FEP-NSZ. Enlargement in the thalamic anterior nuclei predicted a worsening in overall psychosis symptoms. Baseline thalamic anterior nuclei alterations further specify key subcortical regions associated with FEP-SZ pathophysiology. Longitudinally, anterior nuclei volume enlargement may signal symptomatic worsening. The amygdala and thalamus structures may show diagnostic differences between schizophrenia and non-schizophrenia psychoses, while the thalamus changes may reflect disease or treatment related changes in clinical outcome.

1. Introduction

Psychosis spectrum disorders have a median lifetime prevalence of 7.5 per 1000 people (Moreno-Kustner et al., 2018), and its chronic form, schizophrenia (SZ), is ranked among the top 10 leading causes of disability worldwide. First-episode psychosis (FEP) studies are particularly important in understanding the pathophysiology of SZ as they limit the potential confounds of illness chronicity and antipsychotic treatment effects (Akudjedu et al., 2020; Keshavan et al., 1998). Furthermore, there are neurobiological differences between FEP-Non schizophrenia (FEP-NSZ) and FEP-SZ groups, with the latter group showing greater brain alterations (Ohtani et al., 2018; Velakoulis et al., 2006). Thalamic, amygdalar, and hippocampal structures have been consistently implicated in SZ (Hegde et al., 2020; Hill et al., 2017; Vita et al., 2006). For example, in two prospective meta-analyses, bilaterally reduced thalamic, amygdalar, and hippocampal volumes were observed in SZ subjects, while hippocampal and amygdalar volume loss was observed in bipolar disorder (BD) compared to healthy controls (HCs) (Hibar et al., 2016; van Erp et al., 2016). These structures were examined in these studies as a whole, ignoring their highly heterogeneous structure, comprising of multiple nuclei with different structural connectivity and primary functions relevant to the neurobiology of psychosis (Rhindress et al., 2017). Advances in semi-automated surface-based neuroimaging techniques now allow for the parcellation of these brain regions into their nuclei, which improve the accuracy of neuroanatomical analyses.

1.1. Thalamic alterations in psychosis

The thalamus is a major relay center for the cortex and plays an active role in cognitive processes, including attention, speed of information processing, learning and memory, flexible adaptation, sleep spindles, and visual processing (Aggleton, 2014; London et al., 2013; Manoach et al., 2020; Pergola and Suchan, 2013; Schmahmann and Pandya, 2008; Van der Werf et al., 2003; Wolff and Vann, 2019). Thalamic nuclei alterations have been of great interest to researchers trying to understand the pathophysiology of SZ. The anterior and mediodorsal nuclei (Figure 1a) contribute to declarative memory, where the former influences the archival, while the latter coordinates and selects retrieval processes (Child and Benarroch, 2013; Van der Werf et al., 2003). The pulvinar has strong connections to the visual cortex and therefore play a role in visual attention, social cognition, and modulation of behavioral response (Benarroch, 2015; Sherman and Guillery, 2006).

Figure 1:

Thalamic nuclei volume differences between groups. (a) Representative image of thalamic nuclei; (b) Heatmap showing adjusted Cohen’s d effect size for thalamic volumetric analysis between FEP-NSZ (First-Episode Psychosis Non-Schizophrenia), FEP-SZ (FEP-Schizophrenia) and HCs (Healthy Controls) groups at baseline (c) Heatmap showing adjusted Cohen’s d effect size for thalamic volumetric analysis between FEP-NSZ, FEP-SZ, and HCs at follow-up. Data were adjusted for age, sex, race, socioeconomic status, handedness, and intracranial volumes at baseline and sex, handedness, and intracranial volumes at follow-up. * denotes q<0.1.

Several studies have demonstrated that thalamic abnormalities are present in subjects with psychosis. For example, a meta-analysis consisting of thirteen structural MRI studies identified significant bilateral thalamic volume reductions in FEP and chronic SZ subjects when compared to HCs (Adriano et al., 2010). Within antipsychotic-naïve populations, FEP-SZ subjects show greater thalamic volume reduction compared to chronic SZ subjects (Shepherd et al., 2012). In a 5-year longitudinal study, time and time by sex interactions were found to be related to greater thalamic volume loss in SZ (van Haren et al., 2016). Significantly smaller mediodorsal and pulvinar nuclei volumes have also been identified in SZ (Kemether et al., 2003; Shepherd et al., 2012), while FEP subjects were found to have smaller left mediodorsal and anterior nuclei compared to HCs (Gilbert et al., 2001). A recent review of 26 post-mortem studies found consistent evidence for reduced volume and cell number in the pulvinar of people with SZ (Dorph-Petersen and Lewis, 2017).

1.2. Amygdalar alterations in psychosis

The amygdala, a structure located deep in the temporal lobe of the brain, is considered to be a key structure in regulating emotional information and its dysfunction is associated with SZ (Olucha-Bordonau et al., 2015). Neuroimaging studies examining the relationship between amygdala volume and clinical features have been mixed, which may be due to treatment confounds (Ho et al., 2019). The amygdala receives inputs from sensory systems via the lateral nuclei (Figure 2a), while the central nucleus is the main output region to other subcortical sites as well as being responsible for mediating both fear and anxiety-related responses through its connections to the basal forebrain, hypothalamus, and brainstem (Kalin et al., 2004; Kelly and Stefanacci, 2009; LeDoux, 2007). The basal and accessory basal nuclei receive projections from the cornu ammonis 1 (CA1) and subiculum subfields of the hippocampus (Canteras and Swanson, 1992; Phelps and LeDoux, 2005), and are thought to influence emotional memories before acting as the main source of amygdala output to cortical sites (Kelly and Stefanacci, 2009).

Figure 2:

Amygdala nuclei volume differences between groups. (a) Representative image of amygdala nuclei; (b) Heatmap showing adjusted Cohen’s d effect size for amygdala volumetric analysis between FEP-NSZ (First-Episode Psychosis Non-Schizophrenia), FEP-SZ (FEP-Schizophrenia) and HCs (Healthy Controls) groups at baseline (c) Heatmap showing adjusted Cohen’s d effect size for amygdala volumetric analysis between FEP-NSZ, FEP-SZ, and HCs at follow-up. Data were adjusted for age, sex, race, socioeconomic status, handedness, and intracranial volumes at baseline and sex, handedness, and intracranial volumes at follow-up. * denotes q<0.1.

Currently, there is no consensus in the literature regarding amygdala nuclei volumes in FEP. In a meta-analysis of FEP subjects, researchers failed to show amygdala volume differences compared to HCs (Vita et al., 2006). While Joyal et al (2003) (Joyal et al., 2003) and Witthaus (2009) (Witthaus et al., 2009) reported a reduction in amygdala volume in FEP subjects, Velakoulis et al. (2006) (Velakoulis et al., 2006) reported greater amygdala volume in FEP non-SZ (FEP-NSZ) subjects, but not in people at ultra-high-risk for psychosis, FEP-SZ (FEP-SZ) or chronic SZ. Cross-sectionally, researchers found reduced left lateral and bilateral basal nuclei in SZ (Zheng et al., 2019), and only lateral nuclei in FEP (Armio et al., 2018). Volume loss of the lateral nuclei can distinguish non-affective from affective psychosis subjects (Mahon et al., 2015). A postmortem study of the amygdala in SZ identified a decrease in the total number of neurons in the lateral nuclei, which may have been due to the effect of antipsychotic treatment (Berretta et al., 2007). Changes in amygdala volume over time are less certain with one 5-year study reporting that SZ subjects had a greater age-related volume loss compared to control subjects (van Haren et al., 2016).

1.3. Hippocampal alterations in psychosis

The hippocampus is a limbic structure involved in the formation of episodic and declarative memories (Richter-Levin, 2004). Greater anterior hippocampus volume has been associated with better accuracy/performance speed and cognitive performance in SZ and FEP, respectively (Bilder et al., 1995). The presubiculum, subiculum, CA1, CA3, and CA4 (Figure 3a) are the most investigated hippocampal nuclei in psychosis. The presubiculum receives inputs from the subiculum, while the subiculum receives large projections from CA1, which acts to amplify hippocampal output (O’Mara, 2006). CA1, CA3, and CA4 are thought to contribute to episodic memory processing, which has been demonstrated by training memory ability in rats and tracking neuronal loss in these regions with reductions in declarative memory capabilities (Coras et al., 2014; Hunsaker et al., 2008).

Figure 3:

Hippocampal nuclei volume differences between groups. (a) Representative image of hippocampal nuclei; (b) Heatmap showing adjusted Cohen’s d effect size for hippocampal volumetric analysis between FEP-NSZ (First-Episode Psychosis Non-Schizophrenia), FEP-SZ (FEP-Schizophrenia) and HCs (Healthy Controls) groups at baseline (c) Heatmap showing adjusted Cohen’s d effect size for hippocampal volumetric analysis between FEP-NSZ, FEP-SZ, and HCs at follow-up. Data were adjusted for age, sex, race, socioeconomic status, handedness, and intracranial volumes at baseline and sex, handedness, and intracranial volumes at follow-up.

While there is a greater consensus on hippocampal volume loss in psychosis, few studies have examined longitudinal changes. From cross-sectional studies, hippocampal volume is reported to be reduced bilaterally in chronic SZ, unilaterally in FEP, with more subtle differences in the ultra-high-risk group (Velakoulis et al., 2006). Meta-analyses of cross-sectional and longitudinal studies have confirmed the reduction of hippocampal volume in FEP subjects compared to HCs (Steen et al., 2006; Vita et al., 2006). Longitudinally, both HCs and SZ subjects are known to experience a loss in hippocampal volume with the former following a curvilinear trajectory and the latter following a linear pattern (van Haren et al., 2016). Researchers studying hippocampal nuclei have demonstrated volumetric reductions in bilateral presubiculum, subiculum, CA1, CA3, and CA4 in psychotic disorders, most notably in schizophrenia (Baglivo et al., 2018; Haukvik et al., 2015; Mathew et al., 2014; Zheng et al., 2019). Lastly, in a 12-week follow-up study of antipsychotic treatment, researchers found that FEP subjects demonstrated significant reductions in CA4 volume and increased subiculum volume over time (Rhindress et al., 2017), while two other studies did not identify any hippocampal changes over time (Lieberman et al., 2001; Wood et al., 2001).

It is clear that subcortical nuclei abnormalities exist in psychosis, however, changes in these structures over time and their association with clinical outcomes remain to be fully elucidated. Thus, using a longitudinal sample of antipsychotic-naïve FEP subjects, we sought to examine 1) thalamic, amygdalar, and hippocampal nuclei volume differences at baseline and 2) at ~1 year follow up using an annualized rate of change (ARCH) score, as well as determining 3) their association with clinical measures. We hypothesized that FEP-SZ would have smaller thalamic, amygdalar, and hippocampal nuclei at baseline and greater age-related volume loss compared to FEP-NSZ and HCs. Further, we hypothesized that lower structural volumes and greater volume loss over time would be associated with a worse clinical outcome.

2. Methods

2.1. Participants

This research was approved by the University of Pittsburgh Institutional Review Board. All participants provided written informed consent. For subjects below the age of 18 years we obtained consent from a parent or guardian, and informed assent from the participants. Subjects were recruited from the inpatient and outpatient services of the Western Psychiatric Institute and Clinic from 1996 to 2004. Inclusion criteria included; age 15 to 45 years, IQ >75, no active substance abuse or dependence, no significant medical or neurological illness, no history of head injury, and all psychosis participants met the DSM-IV criteria for a psychotic disorder with <2 weeks of lifetime antipsychotic treatment (Keshavan et al., 2003). Subjects were evaluated by trained clinicians using the Structured Clinical Interview for DSM-IV, Patient Edition (SCID-P). Subjects were naturalistically followed up with antipsychotic treatment. All diagnoses were determined using DSM IV and confirmed using consensus meetings including all clinical and SCID-P data after at least 6 months of follow-up with a final diagnosis determined at ~1-year. HCs were recruited from the same geographic location and they did not have a current or past axis I disorder based on the Structured Clinical Interview for DSM-IV–Non-Patient Edition, neurologic or chronic medical condition, family history of psychosis in first-degree relatives or of psychotropic medications within six months of baseline assessment (Keshavan et al., 2003).

All participants were assessed for handedness and parental socio-economic scale (SES) using the Hollingshead Four-Factor Index (Hollingshead, 1975). For FEP participants, the Clinical Global Impression (CGI) (Haro et al., 2003), Global Assessment of Functioning (GAF) (Endicott et al., 1976), and the Scale for the Assessment of Positive Symptoms (SAPS) and Negative Symptoms (SANS) (Andreasen, 1990) were assessed to measure illness severity, function and symptoms of psychosis.

The baseline sample size included 108 participants diagnosed at ~1 year: 38 FEP-SZ (SZ n= 11, schizoaffective n= 5, schizophreniform n=1, or residual/undifferentiated SZ diagnosis n= 21), 23 FEP-NSZ (Bipolar 1 disorder n=2, major depressive affective disorder n=9, delusional disorder n=1, or reactive psychosis not otherwise specified or psychosis not otherwise specified n=11) and 47 HCs at baseline. The longitudinal sample included 59 participants: 17 FEP-SZ (SZ n=4, schizoaffective n=3, or residual/undifferentiated SZ diagnosis n=10), 13 FEP-NSZ (Bipolar 1 disorder n=1, major depressive affective disorder n=4, or delusional disorder n=1, reactive psychosis not otherwise specified or psychosis not otherwise specified n=7) and 29 HCs. In the proband group, only four subjects were taking antipsychotics at a short duration at baseline which we sufficiently categorize this group as antipsychotic-naïve. Participants either had a week 4 follow-up MRI scan (5 FEP-SZ, 3 FEP-NSZ, and 7 HCs) or a year 1 follow-up MRI scan (12 FEP-SZ, 10 FEP-NSZ, and 22 HCs).

2.2. Image acquisition and processing

All participants underwent magnetic resonance imaging (MRI) on a 1.5-T Signa whole-body GE Scanner (General Electric Medical Systems, Milwaukee, Wisconsin). Regional brain volumes were measured with three-dimensional spoiled gradient recall acquisition in the steady-state pulse sequence, which obtained 1.5-mm thick contiguous and parameters as described in (Keshavan et al., 2003). Automated brain segmentation was performed using FreeSurfer 6.0 and manually edited to remove sinuses, dura and correct other segmentation errors in cortical and subcortical regions. Volume measurements for the whole hippocampus, amygdala, and thalamus nuclei were extracted from the additional FreeSurfer modules which parcellate the respective nuclei. These subcortical nuclei were parcellated based on a probabilistic atlas that was built upon either high-resolution ex-vivo MRI data or histological data (Iglesias et al., 2018, 2015; Saygin et al., 2017). From correspondence with Dr. Iglesias (Iglesias et al., 2018, 2015; Saygin et al., 2017), it was determined that the use of a 1.5T scanner to perform subcortical subfield segmentations is adequate as long as post segmentation visual inspection is performed. FreeSurfer atlases underwent rigorous visual inspection (quality control), first rating overall T1 image quality for motion artifacts, and then for proper nuclei segmentation rating them between 0 (low), 1 (intermediate), and 2 (good quality). Out of 108 MRI scans at baseline, 10 scans were rated as 1 for the thalamus, 1 scan was rated 1 for the amygdala, and 13 scans were rated as 1 for the hippocampus. The remaining subcortical scans were rated as 2. Out of 59 longitudinal scans, 6 scans were rated as 1 for thalamus, 2 scans were rated 1 for the amygdala, and 14 scans were rated as 1 for the hippocampus. The remaining subcortical scans were rated as 2. There were no low ratings, hence we did not remove any scans from the analysis. We did not perform additional manual editing of subcortical segmentations. Total intracranial volume (ICV) was also extracted for covarying purposes. Follow up MRI scans were processed through the FreeSurfer longitudinal stream, which used an unbiased within-subject template space and image that was created using robust, inverse consistent registration (Reuter et al., 2012; Reuter and Fischl, 2011).

The averaged right and left nuclei were included for the thalamus (anterior, mediodorsal, ventral lateral, and pulvinar, Figure 1a), amygdala (lateral, basal, accessory basal, and central, Figure 2a), and hippocampus (CA1, CA2/CA3, CA4, presubiculum; and subiculum, Figure 3a). We included these nuclei since they are the most studied regions, have high intraclass correlation coefficients, and can be reliably obtained from 1.5T MRI data (Iglesias et al., 2018, 2015; Saygin et al., 2017).

2.3. Statistical analysis

All statistical analyses were performed using the R statistical analysis software (version 3.5.0, https://www.r-project.org/). Demographic and clinical variables were assessed using chi-squared tests or analysis of variance (ANOVA). We used univariate ANOVAs for testing moderator effects of age, sex, race, handedness, socioeconomic status, duration of illness (since first psychotic symptoms), quality control for whole amygdala, hippocampus, and thalamus, as well as scan interval, antipsychotic status, and chlorpromazine (CPZ) equivalent dose days (CPZ equivalents * dose days * antipsychotic dose) on average nuclei measures (Supplementary table 1). Based on these sensitivity analyses, we covaried for age, sex, race, socioeconomic status, handedness, and intracranial volume (ICV) at baseline and sex, handedness, and ICV at follow-up. We performed a post hoc analysis using the amygdala ARCH values while covarying for the whole amygdala quality control measure (Supplementary Figure 1). At baseline, there were no outliers (defined as above 4 standard deviation (SD)); those between 3 to 4 SD were winsorized to the 3rd SD (n=4). Group contrasts (FEP-NSZ to HCs, FEP-SZ to HCs, FEP-NSZ to FEP-SZ) at baseline and follow-up were compared using a general linear model while adjusting for sex, handedness, and ICV. Adjusted Cohen’s d effects sizes (denoted by “d”) are reported. Longitudinal neuroimaging measures were transformed to annualized rates of change (ARCH) for each region of interest (ROI), [ROI_ARCH= (ROI_FU-ROI_BL/ ROI_BL)/ Interval]*100%, where Interval is the time between baseline (BL) and follow-up (FU) scans in years (Cannon et al., 2015). ARCH measures greater than 6 SDs were removed from the analysis (n=0) and those between 4-6 SD were winsorized to the 4th SD (n=7). A higher threshold for outlier removal was used to preserve the size of the longitudinal sample. Prospective clinical percent change measures were calculated using FU–BL/BL*100 for each clinical measure of interest and was adjusted for age, sex and race. Partial Spearman correlations and Fisher’s rto-z transformations were performed to test the relationship between subcortical nuclei and clinical measures (GAF, CGI, SAPS/SANS scores, and CPZ dose days) within groups (probands, FEP-SZ, and FEP-NSZ) and between groups (FEP-SZ and FEP-NSZ). Adjusted Cohen’s D effect sizes are reported. Multiple comparison correction was performed using the Benjamini Hochberg method and the significance threshold (denoted as “q”) was set at <0.1 (Benjamini and Hochberg, 1995). Multiple comparison correction was performed for the thalamic (n=5), amygdalar (n=5) and hippocampal (n=6) regions separately. Post-hoc analyses were performed for q<0.1 nuclei findings. Partial correlations were corrected for multiple comparisons within groups (FEP-SZ or FEP-NSZ) for nuclei (n=16) and clinical measures (n=4).

3. Results

3.1. Demographics

Sociodemographic and clinical information for each of the study groups at the baseline and follow-up assessment is presented in Table 1. At baseline, the three groups were not significantly different in age, race, and handedness, but they were significantly different in sex or SES. FEP-SZ and FEP-NSZ were not significantly different in the duration of illness, antipsychotics use, and SANS scores, but FEP-SZ had significantly worse SAPS, SAPS/SANS, and CGI scores compared to FEP-NSZ. At follow-up, there were no significant differences in age, sex, race, handedness, SES, and CPZ dose days across all groups. FEP-SZ had significantly greater psychotic use in addition to worse SAPS/SANS and CGI score compared to the FEP-NSZ group.

Table 1:

Sociodemographic and Clinical Information for all Groups by Baseline and Longitudinal Assessment

	Baseline Assessment					Longitudinal Assessment
	FEP- NSZ (n=23)	FEP-SZ (n=38)	HC (n=47)	χ2,F	p- value	FEP- NSZ (n=13)	FEP-SZ (n=17)	HC (n=29)	χ2,F	p- value
Age (mean, sd)	24.0 (8.7)	26.9 (8.4)	25.1 (6.7)	1.1	0.348	23.0 (8.5)	25.0 (5.9)	25.2 (6.3)	0.5	0.61
Sex (Male/Female)	20/3	30/8	25/22	10.8	0.004	12/1	11/6	18/11	4.1	0.13
Race (AA/CA/OT)	5/18/0	10/24/4	10/34/3	3.2	0.52	2/11/0	4/11/2	6/21/2	2.2	0.71
Handedness (Right/Left/Mixed)	18/4/1	31/2/4	41/1/3	6.6	0.16	10/0/3	15/1/1	26/2/1	5.2	0.27
Socioeconomic Status (mean, sd)	41.4 (12.7)	38.6 (13.2)	45.6 (10.1)	3.8	0.027	40.9 (10.8)	38.5 (14.8)	46.0 (10.1)	2.4	0.10
Follow up time in years (mean, sd)	-	-	-	-	-	0.9 (0.4)	0.8 (0.5)	1 (0.4)	0.8	0.45
Antipsychotics (Yes/No)	2/21	2/36	-	2.8	2.46	5/8	15/2	-	37.4	<0.001
CPZ dose days (mean, sd)	-	-	-	-	-	46042 (50135)	41015 (47884)	-	0.04	0.84
Duration of illness in years (mean, sd)	2.1 (3.8)	4.6 (6.4)	-	2.9	0.096	2.0 (15)	2.9 (2.1)	-	1.6	0.21
SAPS and SANS score (mean, sd)	1.2 (0.3)	1.3 (0.3)	-	5.1	0.028	0.8 (0.2)	1.1 (0.3)	-	13	0.001
SANS score (mean, sd)	1.8 (0.5)	2.0 (0.5)	-	0.9	0.337	1.5 (0.3)	1.9 (0.4)	-	8.1	0.01
SAPS score (mean, sd)	0.5 (0.4)	0.7 (0.3)	-	7.8	0.007	0.1 (0.2)	0.4 (0.3)	-	7.9	0.01
CGI score (mean, sd)	3.6 (0.9)	4.2 (0.7)	-	6.5	0.014	2.7 (0.8)	3.7 (0.7)	-	14.8	<0.001

Note: FEP-NSZ = First-Episode Psychosis Non-Schizophrenia; FEP-SZ = FEP-Schizophrenia; HC = Healthy Controls; sd = standard deviation; AA = African American; CA = Caucasian; OT = Other; CPZ = Chlorpromazine; SAPS = Scale for Assessment of Positive Symptoms; SANS = Scale for Assessment of Negative Symptoms; CGI = Clinical Global Impression. SAPS and SANS score = average combined score of SAPS and SANS; CPZ dose days = CPZ equivalent * dose * days on medication. Bolded text denotes p<0.05.

3.2. Thalamus results

At baseline, lower thalamus and its nuclei volumes were observed in the FEP-SZ group compared to HCs and FEP-NSZ groups (Figure 1b). Specifically, FEP-SZ subjects had smaller volumes in the anterior nuclei (combined d=−0.59, q=0.095) and this effect is mainly driven by the right hemisphere (d=−0.70, p=0.002) compared to HCs. Significantly lower volumes in the whole thalamus (combined d=−0.66, q=0.095; left d=0.65, p=0.015; right d=0.60, p=0.020) were found in FEP-SZ when compared to FEP-NSZ individuals (Figure 1b). There were no significant differences between all groups at follow-up (Figure 1c). See Supplementary Table 2 for the subcortical nuclei mean and standard deviations.

3.3. Amygdala results

At baseline, subjects with FEP-SZ generally showed lower volumes in the amygdala and its nuclei compared to the HCs and FEP-NSZ groups, while FEP-NSZ individuals demonstrated few volumetric changes compared to HCs (Figure 2b). There were no significant group differences at baseline assessment between groups. Longitudinally, subjects with FEP-SZ generally had volume enlargement in the amygdala and its nuclei over time compared to HCs and FEP-NSZ groups (Figure 2c), while FEP-NSZ individuals showed a greater rate of volume decline compared to HCs. In particular, FEP-SZ subjects had significantly greater volume enlargement in the whole amygdala (combined d=0.72, q=0.098) compared to HCs and this effect was mainly driven by the left amygdala (combined d=0.74, p=0.036). Compared to FEP-NSZ, FEP-SZ showed an enlargement in the whole amygdala (combined d=0.88, q=0.089; left d=0.78, p=0.035; right d=0.76, p=0.048), basal (combined d=0.74, q=0.098), central (combined d=0.85, q=0.089; right d=0.93, p=0.009), and accessory basal (combined d=0.86, q=0.089; right side d=0.86, p=0.017). There were no significant volumetric differences between FEP-NSZ and HCs at follow up.

3.4. Hippocampus results

At baseline, hippocampal volumes in FEP-SZ individuals did not differ from HCs or FEP-NSZ groups (Figure 3b). There were no significant group differences at baseline and follow-up assessment between groups. Longitudinally, FEP-SZ individuals demonstrated greater rates of volume reduction over time, except for the presubiculum region when compared to HCs, and except for CA3 and CA4 when comparing to FEP-NSZ (Figure 3c).

3.5. Correlations results

We investigated whether average subcortical nuclei volume measures (baseline or ARCH measures) were associated with clinical symptoms or CPZ dose days at baseline or follow up. In individuals with FEP-SZ, improvements in overall psychosis symptoms (Δ SAPS/SANS scores) were associated with volumetric reductions in the thalamic anterior nuclei (r=0.79, q=0.03), but not in FEP-NSZ (Figure 4a, Supplementary Table 3b). The correlation between Δ SAPS/SANS scores and the Δ anterior nuclei in FEP-SZ was significantly different from the same correlation observed in FEP-NSZ (Fisher’s Z test, z= 2.47, p=0.014, Supplementary Table 3b). At baseline, smaller anterior nuclei volume was associated with worse CGI scores in the proband group (Supplementary Table 3a). At follow-up, improvements in global functioning (Δ CGI score) were associated with volumetric reductions in the thalamic anterior nuclei (r=0.51, p=0.046), but not in FEP-NSZ (Figure 4b, Supplementary Table 3b). The correlation between Δ CGI score and the Δ anterior nuclei in FEP-SZ was significantly different from the same correlation observed in FEP-NSZ (Fisher’s Z test, z=2.41, p=0.016, Supplementary Table 3b). See supplementary table 3a, b for additional correlations between nuclei and clinical measures. There were no significant relationships between CPZ dose days and any of the nuclei volumes.

Figure 4.

a: Anterior nuclei volume reduction is associated with improved overall psychosis symptoms. Scatter plot demonstrating the correlation between the average change in anterior volume and change in SAP and SANS scores, which is the average combined score of SAPS and SANS. Volume data as adjusted for sex, handedness, and intracranial volumes. Symptom score was adjusted for age, sex, and race, ^#signifies q=0.03.

b: Anterior nuclei volume reduction is associated with improved clinical global impression (CGI). Scatter plot demonstrating the correlation between the average change in anterior volume and change in CGI scores. Volume data as adjusted for sex, handedness, and intracranial volumes. Symptom score was adjusted for age, sex, and race.

4. Discussion

In this study, we comprehensively analyzed baseline and longitudinal subcortical nuclei differences among FEP-SZ, FEP-NSZ, and HCs and investigated the association between these structures and clinical outcomes. Consistent with the literature, we showed that at baseline the FEP-SZ group had lower overall volume for the thalamus (Adriano et al., 2010; Akudjedu et al., 2020; Shepherd et al., 2012) compared to HCs, and this effect was statistically significant for the anterior nuclei. Anterior thalamic nuclei expansion over time was associated with overall worsening of psychosis symptoms. Additionally, in FEP-SZ there was a significant increase in the whole amygdala compared to HCs and FEP-NSZ, as well as an increase in the basal, central and accessory basal nuclei compared to FEP-NSZ. Lastly, we did not find an effect of antipsychotic burden on any of the nuclei volumetric changes over time.

4.1. Thalamus findings

In this study, we demonstrated that smaller whole thalamic volume may be a useful biomarker to distinguish FEP-SZ from FEP-NSZ subjects at baseline, while lower anterior volume distinguishes FEP-SZ from HCs. Overall, we found that healthy controls had decreased thalamic volume over time and this is consistent with a previous longitudinal study (Rapoport et al., 1997). Additionally, increases in anterior nuclei volume over time may be predictive of symptomatic worsening in FEP-SZ. These findings can be understood in the context of the function of the anterior nuclei, which is connected to mammillary bodies, anterior cingulate, subiculum, retrosplenial cortex, and the hippocampus, which together modulate episodic memory (Dorph-Petersen and Lewis, 2017). Lesions to the anterior nuclei of the thalamus are responsible for the episodic memory deficits observed in Wernicke-Korsakoff syndrome and thalamic strokes (Child and Benarroch, 2013). Reduced anterior nuclei volume has also been proposed as a potent endophenotype for psychosis (Steullet, 2019). Additionally, our group has previously identified lower anterior nuclei volume in FEP-SZ (Gilbert et al., 2001), which we replicated in this larger sample of FEP-SZ individuals, and extended this finding by demonstrating the effects of anterior nuclei enlargement on psychosis symptom worsening over time. From a pathophysiological standpoint, two post-mortem studies reported fewer neurons in the anterior nuclei of individuals with SZ (Young et al., 2000), while one study found reduced oligodendrocyte numbers (Byne et al., 2006). Lower levels of N-Methyl-D-aspartate (NMDA) Receptor 2C mRNA and binding to polyamine and glycine sites of the NMDA receptor binding have been identified in the anterior nuclei of individuals with schizophrenia suggesting diminished glutamatergic activity (Ibrahim et al., 2000). Dopaminergic alterations have also been associated with the anterior nuclei in SZ, including higher levels of D1, D3 and D5 mRNA expression, greater vesicular monoamine transporter (VMAT2, presynaptic dopaminergic innervation) activity, and higher calcyon mRNA expression (Clinton et al., 2005). Lastly, enlargement of the right thalamus and its nuclei have been observed under the effects of atypical antipsychotics and neuroleptic treatments (Cho et al., 2018). Though we did not observe relationships between antipsychotic dose and these structural changes, we cannot rule out the possibility that dopaminergic and glutamatergic upregulation in treatment nonresponsive patients may account for some thalamic structural changes. Thus, anterior nuclei abnormalities in FEP-SZ may be due to disease as well as treatment related alterations in glutamatergic and/or dopaminergic tone resulting in neuroanatomical alterations of the anterior nuclei.

4.2. Amygdala findings

In this study, FEP-SZ subjects demonstrated a significant volumetric enlargement of the whole amygdala compared to HCs and a significant volumetric increase of the whole amygdala, basal, central, and accessory basal nuclei volume compared to the FEP-NSZ group, suggesting that FEP-SZ and FEP-NSZ have orthogonal trajectories. We observed amygdalar volume loss in healthy controls, which has been shown by another longitudinal study (Bois et al., 2016). Enlargement of the overall amygdala and its subcortical structures is consistent with prior studies investigating subjects at high-risk for childhood or early-onset SZ (Levitt et al., 2001; Velakoulis et al., 2006; Welch et al., 2010). This enlargement may be attributed to an increase in oligodendrocytes density as a compensatory mechanism (Williams et al., 2013) or due to the effects of antipsychotic medications (Tebartz van Elst et al., 2004), although the latter was not the case in our study. Pathophysiological support also comes from a postmortem study in individuals with SZ and HCs that examined the amygdala transcriptome and found a downregulation of 'synaptic transmission' and 'behavior' genes, as well as an upregulation of 'immune response' and 'blood vessel development' genes in schizophrenia ((Chang et al., 2017). Lastly, hyperconnectivity of the amygdala is linked to visual hallucination, an example of positive symptoms of SZ (Ford et al, 2015). Thus, enlargement of the amygdala, combined with its strong connectivity, may play an important role in the pathophysiology of FEP-SZ and more studies are needed to further understand this relationship.

4.3. Hippocampus findings

In this study we did not identify any significant hippocampal alterations at baseline or follow up, and there weren’t any correlations with clinical outcomes that survived multiple comparison correction. These findings were surprising, especially since significant correlations between hippocampal regions and positive symptoms severity have been previously reported in the literature (Kühn et al., 2012; Mathew et al., 2014). However, the relatively small sample size in our study may have led to reduced power to detect subtle differences. While we did not observe significant hippocampal alterations, volume loss in FEP-SZ may be due to a decrease in interneuron populations, which has been observed in post-mortem studies (Heckers and Konradi, 2015; Zhang and Reynolds, 2002). We found that healthy controls had increased hippocampal volume over time and this is consistent with a previous longitudinal study (Bois et al., 2016). Postmortem studies showing immature dentate granule cells (Walton et al., 2012) and lower dentate gyrus neural stem cell proliferation (Reif et al., 2006) support the hypothesis of hippocampal pathophysiological abnormalities in SZ. Only a few studies have examined hippocampal alterations in FEP, but there is emerging literature suggesting that the degree of hippocampal hyperactivity or the amount of glutamatergic dysfunction predicts the severity of hippocampal volume loss over time in subjects with SZ (Kraguljac et al., 2013).

4.4. Strengths and Limitations

This report has several strengths, including the minimally anti-psychotic treated FEP population, longitudinal assessment at one year, within-subject design, and the use of advanced semi-automated software for determining subcortical nuclei volume followed by rigorous quality control. While this study focused on three subcortical structures (amygdala, hippocampus, thalamus) and their nuclei, we did not examine other subcortical structures such as the caudate, putamen, globus pallidus, and nucleus accumbens. Other limitations include a relatively small sample size, no data on mood stabilizers or lithium status, and the male preponderance of the probands, which was addressed by adjusting for sex as a covariate. The findings from this study should be replicated in future studies. Furthermore, it is important to acknowledge that the distinct findings in this study compared to others are most likely accounted for by variations in the population studied, methodologies, confounding pathologies and potential moderating factors such as genetic heterogeneity or environmental risks (e.g. stress, smoking, body mass index, medical comorbidities) (Heckers and Konradi, 2015).

Subcortical Nuclei Abnormalities in FEP Highlights.

• At baseline, FEP-Schizophrenia has lower thalamic anterior nuclei volume than HC

• At baseline, FEP-Schizophrenia has smaller whole thalamic volume than FEP-Non Schizophrenia

• Longitudinally, anterior nuclei volume enlargement may signal symptomatic worsening