Distinct Volume Alterations of Thalamic Nuclei Across the Schizophrenia Spectrum

Abstract Background and Hypothesis Abnormal thalamic nuclei volumes and their link to cognitive impairments have been observed in schizophrenia. However, whether and how this finding extends to the schizophrenia spectrum is unknown. We hypothesized a distinct pattern of aberrant thalamic nuclei volume across the spectrum and examined its potential associations with cognitive symptoms. Study Design We performed a FreeSurfer-based volumetry of T1-weighted brain MRIs from 137 healthy controls, 66 at-risk mental state (ARMS) subjects, 89 first-episode psychosis (FEP) individuals, and 126 patients with schizophrenia to estimate thalamic nuclei volumes of six nuclei groups (anterior, lateral, ventral, intralaminar, medial, and pulvinar). We used linear regression models, controlling for sex, age, and estimated total intracranial volume, both to compare thalamic nuclei volumes across groups and to investigate their associations with positive, negative, and cognitive symptoms. Study Results We observed significant volume alterations in medial and lateral thalamic nuclei. Medial nuclei displayed consistently reduced volumes across the spectrum compared to controls, while lower lateral nuclei volumes were only observed in schizophrenia. Whereas positive and negative symptoms were not associated with reduced nuclei volumes across all groups, higher cognitive scores were linked to lower volumes of medial nuclei in ARMS. In FEP, cognition was not linked to nuclei volumes. In schizophrenia, lower cognitive performance was associated with lower medial volumes. Conclusions Results demonstrate distinct thalamic nuclei volume reductions across the schizophrenia spectrum, with lower medial nuclei volumes linked to cognitive deficits in ARMS and schizophrenia. Data suggest a distinctive trajectory of thalamic nuclei abnormalities along the course of schizophrenia.


Introduction
Schizophrenia is a severe mental disorder that is heritable, polygenic, and heterogeneous in symptoms as well as in their course. 1,2Symptoms cluster into positive (eg, psychotic), negative (eg, anhedonia or social withdrawal), and cognitive symptoms (eg, attention or executive function deficits).While negative and cognitive symptoms show a rather stable course from their usual onset in early adolescence, psychotic symptoms typically fluctuate between psychotic episodes and psychotic remission.Thus, "pre-stages" exist in which psychotic risks or episodes occur without the presence of a manifest disorder.4][5] Individuals at risk for psychosis were found to have a 36% transition rate at 3 years, 6 which is tremendously higher than a 1% lifetime risk in the general population. 2 When psychotic symptoms appear for the first time, the occurrence is called "first episode psychosis" (FEP). 7ARMS, FEP, and established schizophrenia (SCZ) together are referred to as the "schizophrenia spectrum." 3,8,91][32] This has been facilitated by the recent development of a pipeline for segmenting thalamic nuclei on T1-weighted MR images. 335][36][37] Particularly, lower mediodorsal thalamic nuclei volumes 38 as well as aberrant prefrontalmediodorsal thalamic connectivity 10,[39][40][41] have previously been linked to cognitive impairments in schizophrenia.Moreover, cortico-thalamic hypoconnectivity has been reported in ARMS. 42However, it remains to be elucidated whether thalamic nuclei volume reductions occur prior to disease manifestation since previous studies have focused mainly on one stage of the disorder and have not covered the entire spectrum of schizophrenia.In addition, there is scarce evidence for nuclei-based volumetric studies in the pre-stages of the disorders.
Thus, our study of thalamic nuclei volumes across the schizophrenia spectrum pursued the following objectives: First, based on previous findings, particularly in established schizophrenia, we hypothesized that only specific nuclei, primarily the medial and pulvinar subregions, exhibit lower volumes than the control group.Due to conflicting and incomplete results derived from previous literature investigating pre-chronic phases, we did not have a clear hypothesis regarding these stages.To examine our hypothesis, the volumes of thalamic nuclei were determined in three stages of the schizophrenia spectrum as well as in age-and sex-matched healthy controls using a pipeline commonly used in and validated by the neuroimaging community. 33Second, we sought to study a potential link between volumetric aberrations and distinct symptom dimensions, namely positive, negative, and cognitive, in the patient groups.Given prior, albeit inconsistent, evidence regarding the possible involvement of medial and pulvinar nuclei groups in cognitive impairments, 10,[38][39][40][41] we expected a significant association between altered medial subregion volume and cognitive deficits in schizophrenia.

Participants
Structural T 1 -weighted brain images and clinical data of 522 subjects were collected from four sites, comprising control participants without psychopathology (HC) as well as three stages of the schizophrenia spectrum: ARMS, FEP, and established SCZ.These data have been used previously to investigate orbitofrontal-striatal abnormalities in relation to negative symptoms 43 as well as to assess basal-forebrain cholinergic nuclei alterations in relation to cognitive symptoms 44 across the schizophrenia spectrum.See table 1 and Supplementary Methods for an overview and more detailed demographics of each group per scanning site.
Munich Dataset.][47] Patients were in psychotic remission according to DSM-IV-TR criteria. 48,49Exclusion criteria for healthy controls were a history of axis I disorder, substance abuse, and first-degree relatives with a history of psychosis.Substance abuse was eliminated via urine screening or clinical interview.Patients' antipsychotic medication was kept stable for at least 2 weeks prior to the scan.Psychotic and negative symptoms were evaluated with the Positive and Negative Syndrome Scale (PANSS). 50

COBRE
Dataset.An open-source dataset provided by the Center for Biomedical Research Excellence (COBRE) database (http://fcon_1000.projects.nitrc.org/indi/retro/cobre.html)comprising 72 schizophrenic patients meeting DSM-IV criteria 49 and 73 healthy controls were incorporated into the analysis.Antipsychotic medication was kept stable for a minimum of 4 weeks before the study.HC had no history of DSM-IV axis I disorders or psychosis in any first-degree relative.Urine screening was Applied to detect substance abuse.Psychotic and negative symptoms were measured using PANSS. 50rich Dataset.Data of 48 patients with schizophrenia, 26 FEP subjects, as well as 28 healthy controls were acquired and previously analyzed. 43,51,52Clinical diagnosis was based on the structured Mini-International Neuropsychiatric Interview for DSM-IV (M.I.N.I). 53The inclusion criterion for SCZ was a clinical diagnosis of schizophrenia.Subjects with FEP were recruited during their first psychiatric admission in outpatient (n = 6) and inpatient (n = 20) units of the Psychiatric Hospital of the University of Zurich. 52FEP was defined as having a first clinical diagnosis of brief psychotic disorder, schizophreniform disorder, or first-episode schizophrenia using M.I.N.I. 53 FEP subjects with a positive subscale item higher than 5 on PANSS were excluded because the initial aim of the experiment was to study the link between striatal connectivity aberrations with negative symptoms. 43,51,52For all patients, antipsychotic medication was kept stable for a minimum of 2 weeks before the study.Patients were excluded if another DSM-IV axis I disorder was diagnosed, benzodiazepines (> 1 mg/day Lorazepam-equivalent) were taken, or if extrapyramidal side effects were evident.Exclusion criteria for HC were psychiatric disorders, history of psychiatric disorders, and substance abuse.
Basel Dataset.Data from 77 FEP individuals, 73 ARMS subjects, and 44 healthy controls were retrieved from previous studies. 54,55All individuals potentially at risk for an early stage of schizophrenia were intensively assessed based on variables previously shown to be predictors of psychosis such as social decline or genetic risk. 56,57Each individual identified to be at-risk was followed up every 1-3 months during the first 3 years, then yearly, for a thorough examination of potential predictors of schizophrenia.In particular, ARMS for psychosis was defined based on the Basel Screening Instrument for Psychosis 58 and the Brief Psychiatric Rating Scale (BPRS) 4,59 as showing "attenuated" psychotic symptoms, brief limited intermittent psychotic symptoms, or having a firstor second-degree relative with psychotic disorder and a marked decline in social functioning.More precisely, the decline in social functioning was defined as a "marked deterioration of psychosocial functioning with serious consequences for work, education, relationships (occurrence during the last 5 years and persisting up to now)." 57 After a clinical follow-up of 33.3 months, 15 ARMS individuals had transitioned to psychosis.First-episode psychosis included subjects who fulfilled the criteria for acute psychotic disorder according to the International Statistical Classification of Diseases and Related Health Problems, 10th Revision, 60 or DSM-IV criteria. 49In particular, a BPRS 4,59 score of ≥4 on the hallucination item, or ≥5 on the unusual thought content, suspiciousness item, or conceptual disorganization item were set as the threshold. 61,62Symptoms needed to be present at least several times a week and the change in mental state lasting >1 week. 5727 FEP patients were medication-free, while 45 were prescribed antipsychotics.Exclusion criteria for HC were psychiatric disorders or history of psychiatric disorders, head trauma, neurological illness, serious medical illness, substance abuse, or psychiatric disorders in family history.The Scale for the Assessment of Negative Symptoms (SANS) was used to assess the severity of negative symptoms. 48n summary, 73 ARMS participants, 104 FEP participants, 146 patients with schizophrenia, as well as 146 control subjects were examined.Following quality control of MRI cortical and thalamic nuclei segmentation, the final cohort consisted of 66 ARMS participants, 89 FEP individuals, 126 patients with schizophrenia, and 137  S1.
All studies were approved by the local ethics committees and all participants provided written informed consent.

Symptom Ratings: Positive, Negative, and Cognitive Symptoms
For schizophrenia patients and some FEP subjects, psychotic, and negative symptoms were evaluated with the Positive and Negative Syndrome Scale (PANSS). 50For ARMS and some FEP individuals, the Brief Psychiatric Rating Scale (BPRS) 61 and Scale for the Assessment of Negative Symptoms (SANS) 48 were used alternatively to assess positive and negative, or negative symptomatology, respectively.Refer to Supplementary table S1 for an overview.
As the four acquisition sites used different neuropsychological batteries for cognitive testing, several test scores were assessed to investigate patients' cognitive abilities (Supplementary table S1).The Verbal Fluency Test was available for all patient groups as well as HC.It captures processing speed and semantic memory by asking participants to name as many words as possible in 60 s.Semantic fluency tasks involve words that belong to a specific category (here: animals; "animal fluency"), while phenological fluency tasks ask participants to name as many words as possible that begin with a specific letter (here: letter S). 63,64 In addition, the Trail Making Test Part A (TMT-A) and the Symbol Coding Task of the Brief Assessment of Cognition in Schizophrenia (SCT) were available for SCZ and HC subjects only.In the TMT-A, processing speed and visual attention are assessed by measuring the time to join numbers with a line in seconds. 65Similarly, the SCT determines processing speed and attention by asking subjects to decode symbols into numbers with a given decoding key. 66For an overview of relevant cognitive processes as well as linked thalamic outcomes associated with the cognitive tasks conducted, see Supplementary table S2.

Assessment of Clinical Characteristics
The assessment of several clinical characteristics was conducted by psychiatrists at each site.Current antipsychotic medication in terms of chlorpromazine (CPZ) equivalents was determined in milligrams per day (mg/ day), 67 whereas the cumulative effect of anticholinergics was determined in terms of anticholinergic burden of medication (ACB) defined in arbitrary units (a.u., see https://www.acbcalc.com/). 68

Image Acquisition, Processing, Volume Outcomes, and Harmonization
T 1 -weighted structural brain MRI scans were acquired at each site.MRI data acquisition parameters are described in detail in the Supplementary Methods section for each site.
Next, the scans were processed using FreeSurfer's (v 7.1.1;https://surfer.nmr.mgh.harvard.edu/)cortical reconstruction process recon-all.In brief, artifact correction, skull stripping, normalization into standard space as well as cortical segmentation and parcellation were performed.After assessing reconstruction quality visually, following FreeSurfer recommendations (https://surfer.nmr.mgh.harvard.edu/fswiki/QATools),25 individuals were excluded due to processing inaccuracies (see a detailed description of exclusion criteria in Supplementary Methods).
Afterwards, a built-in thalamus segmentation pipeline 33 was applied to segment the thalamus into 26 nuclei per hemisphere.Quality inspection following recent standards of the community 69,70 led to the exclusion of a further 50 participants due to unsatisfactory segmentation results.Detailed exclusion criteria and quality assessment procedures are described in the Supplementary material.Volumes in mm 3 were extracted for each subject and each nucleus directly from FreeSurfer.
In accordance with prior literature, 33,69 we grouped these nuclei into different subregions: anterior, lateral, ventral, intralaminar, medial, and pulvinar (see table 2 and figure 1).Note: According to suggestions in previous literature, 33,69 thalamic subregions were defined incorporating the nuclei distinguished with the Thalamic Nuclei pipeline 33 listed in the left column.
Since the naïve combination of neuroimaging data across different sites introduces variance mostly attributed to varying scanning protocols and hardware, volumetric outcomes were retrospectively harmonized for scanner effects using NeuroCombat (v0.12.2, in Python). 71Adopted from genomics, 72 Combat models site-specific scaling factors by using an empirical Bayes method to adjust the data for batch effects while preserving a priori known sources of biological variance. 71The method has repeatedly shown great potential to remove scanner variation in diffusion tensor imaging, 73 structural, 71,74 and functional 75 MRI data.

Thalamic Nuclei Across the Schizophrenia Spectrum
Estimated total intracranial volume (eTIV) was corrected while preserving biological variation related to diagnosis, sex, and age, whereas eTIV variance was additionally retained when harmonizing subregional thalamic measures.Harmonization successfully removed variance attributed to the hardware (see Supplementary figure S1).However, since we are aware that there are limitations to NeuroCombat if sites acquired data from different groups, 76,77 we compared subregion volumes within the Basel cohort before harmonization (Supplementary Results).

Statistical analysis
Group Difference in Thalamic Subregion Volumes.To estimate group differences, analyses of covariance (ANCOVA) were run on the harmonized data in IBM SPSS Statistics (IBM Corp. Released 2021.IBM SPSS Statistics for Windows, Version 28.0.Armonk, NY: IBM Corp).Mean volumes of the six thalamic subregions served as the within-subject variables, whereas group was the between-subject variable.Biological sex, age, and scanner-corrected eTIV were included as covariates of no interest.In particular, eTIV was included to control for individual head size differences among subjects.Fisher's Least Significant Difference (LSD) test was applied for post hoc analyses separately for each test.Subsequently, Benjamini-Hochberg correction for multiple comparisons 78 was applied to correct for the use of six different tests in total.Partial eta square (ηp 2 ) was used to estimate effect sizes.
Control Analyses for Group Difference in Thalamic Subregion Volumes.Multiple linear regression models, as described in the following paragraph, were used to link reduced thalamic nuclei volumes with other confounding variables, namely medication, age, and illness duration.See Supplementary Methods for a detailed description.

Associations Between Thalamic Subregion Volumes and
Cognitive Performance.Differences in cognitive scores between groups were determined using an ANCOVA with the respective test as a dependent variable, group as a factor, and age and sex as covariates.
Testing the associations between cognitive performance and thalamic subregion volumes was conducted in SPSS using multiple linear regression analyses.To illustrate the effect of each covariate on the association between symptoms and subregional volume, we performed various regression analyses in a hierarchical framework, that is, we extended the model stepwise by one covariate.Thus, we came up with four different models to test the interaction between volume and positive, negative, and cognitive symptom scores (sym pos , sym neg , and sym cog , respectively):

Results
An overview of the demographic and clinical characteristics of the final pooled samples for each clinical group is shown in table 1 and Supplementary table S1.For comparisons of these characteristics across patients and healthy controls, see Supplementary Results.For example, age and sex are comparable between patient groups and corresponding healthy control groups for each site.

Consistently Lower Medial Nuclei Volumes Across All Stages of the Schizophrenia Spectrum, Lower Lateral Nuclei Volumes Only in Established Schizophrenia
To evaluate volumetric changes in grouped thalamic nuclei subregions across the schizophrenia spectrum, we applied one-way ANCOVA adjusting for age, sex, and eTIV followed by post hoc tests.We found group differences only in lateral and medial nuclei groups (figure 2, Supplementary table S3).More specifically, a significant group difference was detected in medial nuclei (F 3,414 = 4.115, P unc = .007,P FDR = .042,ηp 2 = 0.029).Lower volumes were observed in ARMS (P = .046),FEP (P = .005),and established schizophrenia (P = .003)compared to the control group (figure 2a).Furthermore, we found a significant difference in the lateral nuclei subregion (F 3,414 = 3.588, P unc = .014,P FDR = .042,ηp 2 = 0.026).
Post hoc analysis demonstrated that volumes of lateral nuclei were significantly lower in schizophrenia patients compared to healthy controls (P = .005)as well as between schizophrenia patients and FEP (P = .010;figure 2b).These findings point to region-specific volume reductions in the medial and lateral thalamic nuclei, with the two nuclei showing a different pattern of reduction across the clinical groups.

Control Analyses
Group-based Volumetric Analysis of the Thalamic Subregions per Hemisphere To assess group differences in thalamic subregions per hemisphere, we conducted an ANCOVA adjusting for age, sex, and eTIV for each hemisphere separately.In line with our main analysis, we found that only lateral and medial subregions were significantly different between patients and controls (Supplementary table S4).

Association of Thalamic Nuclei Volumes With Medication, Age, and Illness Duration
Regarding medication status, approximately 70% of participants with FEP and almost all patients with schizophrenia were receiving antipsychotic medication (measured as CPZ equivalents) and/or medication with anticholinergic effect or burden (measured as ACB, see Supplementary table S1).To assess the link between medication status and thalamic nuclei volumes, multiple linear regression analyses were performed.Neither antipsychotic medication nor the anticholinergic burden of medication revealed a significant association with lateral or medial subregional volumes in any patient group investigated (Supplementary table S6).

Impact of Age or Illness Duration in the SCZ Group
Furthermore, we investigated whether age or illness duration might impact thalamus nuclei volumes using multiple linear regression analyses in the schizophrenia group.Regarding lateral subregion volumes, no significant associations with age were found.In contrast, regarding medial subregion volumes, there was a significant link to age in the healthy control group (β-coefficient = −.451,P < .001) as well as the schizophrenia group (β-coefficient = −.373,P < .001;see Supplementary table S7).Since the groups are age-matched and age is related to medial volume in both groups, but medial volume is smaller in patients than in controls (even when controlling for age), we conclude that the aberration in medial volume is driven by one or more other factors, such as pathology or factors related to chronicity (eg, hospitalizations, years of antipsychotic medication intake, etc.).Moreover, there was no significant interaction between age and disease stage (lateral: β-coefficient = −.248,P = .131;medial: β-coefficient = −.138,P = .830),suggesting that the disorder effect on medial thalamus volume did not further increase with age (Supplementary table S8).Neither lateral nor medial subregion volumes were associated with illness duration in the schizophrenia group (Supplementary table S7.2), implying that there is no significant relationship between longer illness duration and greater reduction in thalamic nucleus volumes in schizophrenia.Thalamic Nuclei Across the Schizophrenia Spectrum

Medial Nuclei Volumes are Specifically Linked to Cognitive Impairments
Next, we analyzed the relationship between altered medial and lateral thalamic nuclei subregions and different symptom clusters (ie, positive, negative, and cognitive) for each stage of the schizophrenia spectrum separately.We again want to stress here that clinical and cognitive data were only available for a subset of the participants and that neuropsychological test batteries differed between acquisition sites (see Supplementary table 1 for an overview).For each clinical group, we first analyzed cognitive impairments by examining differences in cognitive test performance compared to healthy controls using one-way ANCOVAs adjusting for age and sex.Second, multiple linear regression analyses were performed in a hierarchical framework to examine the associations between smaller volumes in the lateral or medial subregions and symptoms.For this purpose, we used four different models (see "Methods" section), each one adding a covariate to the model to track the influence of this covariate on the association between volume and symptoms.
In subjects with FEP, phonemic fluency was significantly lower compared to healthy controls (F 1,91 = 14.351,
Finally, cognitive impairments of patients with schizophrenia were not linked to lower lateral nuclei volumes (eg, TMT-A, model 3: β-coefficient = −.182,P = .153;see Supplementary tables S10.6-S10.8),implying that only medial nuclei volume decrease is implicated in cognitive impairments in the schizophrenia spectrum.Specificity.Of note, neither positive (assessed with PANSS positive or BPRS positive) nor negative symptoms (assessed with PANSS negative or SANS) were found to be significantly related to medial or lateral volume in any patient group (Supplementary table S10), suggesting that associations of medial nuclei volumes were specific for cognitive symptoms across stages of the spectrum.
In addition, there was no relation between medial subregional volume and any cognitive score detected in the healthy sample (Supplementary table S11).Yet, by the use of interaction analysis, we did not detect a specific link between medial nuclei volume and cognitive performance in the SCZ group (see Supplementary Results).

Discussion
We examined thalamic nuclei volumes across the schizophrenia spectrum and associated nuclei volume changes with symptom scores.Medial nuclei volumes were specifically lower across the spectrum and associated with cognitive symptoms.In contrast, lateral nuclei volumes were lower only in established schizophrenia and were not linked to symptoms.To the best of our knowledge, this study is the first to demonstrate distinct volume reductions of thalamic nuclei across the entire schizophrenia spectrum.Regarding the potential course of schizophrenia, our findings might support a model of ongoing involvement of medial nuclei across developmental stages, impacting cognitive function, while thalamic impairment may encompass lateral nuclei at later stages.

Thalamic Nuclei Show Different Patterns of Volume Reduction
Thalamic Volume Reductions are Specific to Medial and Lateral Nuclei.The principal finding of the present study is that only medial and lateral nuclei in the schizophrenia spectrum had significantly lower volumes, while the volumes of the remaining nuclei did not differ significantly between groups.This implies a nuclei-specific impact of the disorder.
While postmortem findings regarding smaller thalamic nuclei in patients with schizophrenia are less consistent for the mediodorsal nucleus than the pulvinar, 23,24,36,37,74,80 we note that neuroimaging literature has often reported alterations in both regions, 30,32 but there are exceptions. 35,81The inconsistency between postmortem and imaging findings, particularly regarding the pulvinar, may be related to methodological differences and inherent limitations of each approach (eg, indirect measure of brain characteristics for MRI, small sample sizes for postmortem studies, etc.).Nevertheless, this discrepancy merits further investigation.
In contrast to the considerable literature in established schizophrenia, studies on pre-stages are sparse.However, there is evidence that the thalamus volume as a whole is smaller in FEP individuals, 82,83 particularly the mediodorsal region, 84 and that its volume is reduced longitudinally. 85Findings in ARMS have been controversial, with both reduced 86,87 and increased 88 volumes detected.So far, there is solely one investigation of thalamic nuclei volumes in ARMS, reporting no significant differences from healthy controls. 35However, only 4 of the 38 participants examined developed psychosis later, suggesting that these results may underestimate true differences.

Medial and Lateral Nuclei Reductions Exhibit Different
Patterns Across the Schizophrenia Spectrum.We noted Thalamic Nuclei Across the Schizophrenia Spectrum distinct volume reduction patterns across disease Specifically, lateral nuclei exhibited reductions solely in schizophrenia patients, whereas medial nuclei consistently showed lower volumes throughout the entire spectrum.As expected, medial volume reductions were less pronounced in ARMS and FEP subjects than in schizophrenia.Noteworthy, there was no significant difference of any subregional volume when assessing the Basel cohort only (see Supplementary Results, control analysis), before data harmonization.We argue that analysis in only one cohort is underpowered and that pooling data across sites is necessary to draw conclusions based on larger samples.
One interpretation of this pattern regarding schizophrenia development might be that medial nuclei impairments manifest early in the disease, possibly due to genetic predisposition, and worsen over disease stages.In contrast, lateral nuclei impairments may arise later during the disorder and result from different secondary mechanisms.
Indeed, there is evidence from genetic 89 and animal 90,91 studies that thalamic nuclei show different gene expression profiles and develop in distinct trajectories.We speculate that the alterations in the medial nuclei are due to differences in neurodevelopment, whereas this is less or not at all the case for the lateral nuclei.For instance, medial nuclei volumes were associated with a critical neurodevelopmental transcription factor, 89 whose mutations have been linked to schizophrenia. 92,93This idea is supported by findings from neuroimaging studies.For instance, Xi et al 41 compared cortico-thalamic functional connectivity within schizophrenia patients, their siblings, and healthy controls.Whereas increased sensorimotorthalamic (ie, lateral) functional connectivity was specifically found in schizophrenia patients, decreased prefrontal-mediodorsal connectivity was shared between patients and their healthy siblings.Reduced prefrontalmediodorsal connectivity was also observed in clinical high-risk individuals and in adolescents with early-onset schizophrenia, 15,40,42,94 with subjects that later transitioned to schizophrenia showing the most apparent reductions. 42ogether, the decrease of prefrontal-mediodorsal functional connectivity might to be at least partly genetically imprinted.Conversely, lower cell count might also lead to hypoconnectivity, although longitudinal studies are lacking.Our findings seem compatible with a developmental model of early impaired mediodorsal nuclei (ie, in ARMS) but later changes in lateral nuclei (ie, in an established disorder) in the course of schizophrenia.However, longitudinal studies are necessary to test this hypothesis.

Volumetric Aberrations of the Medial Nuclear Group are Linked to Cognitive Impairments Across the Schizophrenia Spectrum
We detected significantly reduced cognitive performance in all clinical groups in at least one test of cognitive functioning.Furthermore, significant volume reductions in the medial nuclei group were specifically linked to cognitive symptoms in ARMS and SCZ patients.Thereby, specificity was apparent in two dimensions: (1) regionspecific as medial not lateral volume reductions were associated with cognitive symptoms, and (2) symptomspecific as no link was detected between volume reductions and positive or negative symptoms.
In more detail, we found consistently lower cognitive scores in schizophrenia patients compared to controls for all cognitive tests, while we observed mixed findings in ARMS and FEP subjects (Supplementary table S9).6][97] Our results imply a worsening of symptoms across the spectrum and heterogeneity between subjects.
In the schizophrenia group, lower performances in TMT-A and SCT were significantly linked to lower medial nuclei volumes in several models, suggesting a key role of the medial thalamus in cognitive symptomatology.Medial volume is a significant predictor of cognition that is not explained by differences in gender or head size.However, when also accounting for age, cognition is not significantly explained by medial volume.As known from previous studies, [98][99][100] age is strongly associated with volume and cognition.If variables in the model are highly correlated with each other (ie, multicollinearity), this leads to instability in estimates.Alternatively, the relationship between cognitive symptoms and volume might be non-linear, and age might capture these nonlinear effects.
The implication of medial nuclei as a key modulator in cognitive functioning is well documented 101,102(pp. 312-317) and in line with its extensive reciprocal connections with medial and lateral prefrontal cortices (PFC), critically controlling cognitive functioning. ,101,102(p. 307)103Moreover, they may modulate the connectivity between the hippocampus and the mPFC, being also involved in the control of cognitive functions. 104,105Accordingly, lesion studies in patients with focal lesions demonstrated that the mediodorsal thalamus is necessary for cognitive control. 106,1070][41] However, this association between medial thalamus volume and cognitive impairment was not detected when age was additionally included as a covariate, pointing out that the association could be driven by age-related performance differences.
In contrast to the schizophrenia lesion model, where lower medial volumes are associated with reduced cognitive performance, we observed the opposite trend in ARMS.Specifically, we found a significant link between lower medial volume and higher cognitive scores.In general, this non-intuitive pattern might be attributed to a non-linear course the relationship between medial thalamus volume and cognition over the course of schizophrenia.Given that prefrontal-mediodorsal connectivity seems to be neurodevelopmentally decreased in schizophrenia 41 and that dysconnectivity might entail 108,109 or is at least related to 110 gray matter loss, prefrontalmediodorsal connectivity going along with cognitive decline might precede substantial volume reductions in the medial thalamus across the whole at-risk population.The anti-correlation observed in ARMS may suggest that compensatory mechanisms are effective early in the disease course leading to a relatively better performance in cognitive performance despite thalamic structural differences.However, as the disease progresses to FEP and SCZ stages, these compensatory mechanisms might become less efficient, resulting in no significant correlation and, ultimately, in a different pattern of correlation.
In healthy controls, we did not find a significant correlation between cognition and medial thalamus volume.This is in contrast to a previous study, 30 but might be explained by the employed cognitive testing, mainly focused on processing speed-related cognitive functions.Nevertheless, our results show a significant link between cognitive symptomatology and medial thalamic volume reductions in some spectrum groups, following the assumption that especially medial nuclei are relevant for various domains of cognition. 19,101,106

Strengths and Limitations
This study extends prior research in several ways: first, the pipeline utilized to segment thalamic nuclei is well evaluated 33 and has been successfully applied to a variety of disorders. 30,31,69,111Therefore, a well-justified and highresolution analysis of thalamic subregions is provided.Second, intensive quality control as well as adjustment for multi-site batch effects attest to the methodological precision of this investigation.Third, the presentation of findings based on a large sample encompassing multiple stages of the schizophrenia spectrum allows the generation of valuable hypotheses regarding developmental trajectories.
Despite these strengths, several limitations require consideration.Segmentation of thalamic nuclei suffers from poor contrast of thalamic margins on T1w images.Although we intensively scrutinized segmentation quality according to the latest recommendations 69,70 and have excluded some nuclei due to their small size and poor contrast in 1 mm 3 T1w images, we are aware that criticism exists regarding the anatomical relevance of this pipeline. 112Furthermore, despite correcting for batch effects of MRI data using gold standard procedures, 71 some effects resulting from site and hardware differences may remain.Especially, since the results for the Basel site alone were not significant for schizophrenia pre-stages, we cannot rule out that effects were affected by the harmonization procedure.By demonstrating that these differences are not significant, large effects are not to be anticipated.Additionally, due to the lack of respective tools, clinical and cognitive data were not harmonized across sites, possibly leading to non-biological sources of variance in the data.Although we emphasize the uniqueness and importance of the spectrum design of our study for hypotheses regarding disease progression, we caution that a cross-sectional design does not permit temporal inference of thalamic nuclei volume alterations.In addition, comparisons between the data sets regarding clinical and cognitive characteristics were only possible to a limited extent due to the different batteries employed, which were limited in scope (ie, mainly assessing processing speed) and were only completed by a subset of the participants.

Fig. 1 .
Fig. 1.Thalamic nuclei grouping based on segmentation in FreeSurfer.The thalami are segmented into 26 nuclei for each hemisphere, as listed in table 2. (a) Based on literature recommendations, 33,69 thalamic nuclei were then grouped into six nuclei clusters per hemisphere (see table 2): anterior (red), ventral (blue), lateral (pink), medial (dark green), intralaminar (orange), and pulvinar (light green).Excluded nuclei due to previous recommendations or overall poor segmentation quality are summarized and shown in white.(b) Depicts an actual segmentation performed on a 30-year-old healthy male of the COBRE dataset.In (c), thalamic subregions are overlaid on a bias-field corrected T1w image of the MNI152 template as obtained from FSL standards.MNI, Montreal Neurological Institute.Panel (a) was created with BioRender.com.

Fig. 3 .
Fig. 3. Association of medial thalamic nuclei volume with cognitive impairments across the schizophrenia spectrum.Volumes not adjusted for age, sex, and eTIV are shown for visualization purposes.(a) Significant association with animal fluency in at-risk mental state patients.(b) Significant association with TMT-A in established schizophrenia.(c) Significant association with SCT in established schizophrenia.

Table 1 .
Demographics Per Acquisition Site

Thalamic Nuclei Across the Schizophrenia Spectrum healthy
controls (see below).For detailed information about data-based exclusion criteria, see Supplementary Methods.A summary of all available measures as well as patient characteristics is given in Supplementary table