Basal-Forebrain Cholinergic Nuclei Alterations are Associated With Medication and Cognitive Deficits Across the Schizophrenia Spectrum

Abstract Background and Hypothesis The cholinergic system is altered in schizophrenia. Particularly, patients’ volumes of basal-forebrain cholinergic nuclei (BFCN) are lower and correlated with attentional deficits. It is unclear, however, if and how BFCN changes and their link to cognitive symptoms extend across the schizophrenia spectrum, including individuals with at-risk mental state for psychosis (ARMS) or during first psychotic episode (FEP). Study Design To address this question, we assessed voxel-based morphometry (VBM) of structural magnetic resonance imaging data of anterior and posterior BFCN subclusters as well as symptom ratings, including cognitive, positive, and negative symptoms, in a large multi-site dataset (n = 4) comprising 68 ARMS subjects, 98 FEP patients (27 unmedicated and 71 medicated), 140 patients with established schizophrenia (SCZ; medicated), and 169 healthy controls. Results In SCZ, we found lower VBM measures for the anterior BFCN, which were associated with the anticholinergic burden of medication and correlated with patients’ cognitive deficits. In contrast, we found larger VBM measures for the posterior BFCN in FEP, which were driven by unmedicated patients and correlated at-trend with cognitive deficits. We found no BFCN changes in ARMS. Altered VBM measures were not correlated with positive or negative symptoms. Conclusions Results demonstrate complex (posterior vs. anterior BFCN) and non-linear (larger vs. lower VBM) differences in BFCN across the schizophrenia spectrum, which are specifically associated both with medication, including its anticholinergic burden, and cognitive symptoms. Data suggest an altered trajectory of BFCN integrity in schizophrenia, influenced by medication and relevant for cognitive symptoms.

negative symptoms were evaluated with the Positive and Negative Syndrome Scale (PANSS). 6This study was approved by the Ethics Review Board of our institution and all participants gave written informed consent.

MRI data acquisition in the Munich cohort
All participants were scanned in an MR-scanner at the Klinikum Rechts der Isar, Munich, Germany.Data acquisition was performed on a hybrid whole-body mMR Biograph PET/MRI scanner (Siemens-Healthineers, Erlangen, Germany), with a vendor-supplied 12-channel phase-array coil.Anatomical data of the whole brain were collected using a T1-weighted MPRAGE sequence with TR/TE/flip angle: 2300 ms/2.98 ms/9°; 160 slices (gap 0.5 mm) covering the whole brain; FoV: 256 mm; matrix size: 256×256; voxel-size: 1×1×1mm3.

The COBRE cohort
The dataset of 72 SCZ meeting DSM-IV-TR criteria 5 (age range: 18-65 years; mean: 38.16±13.89years) and 73 HC (age range: 18-65 years; mean: 35.60±11.50years) was derived from the COBRE dataset.Two subjects had to be excluded.Antipsychotic medication was kept stable for minimum four weeks before the study.
HC had no history of DSM-IV Axis I disorders, or psychosis in any first-degree relatives.Substance abuse was eliminated via urine-screening.Psychotic and negative symptoms were measured by PANSS. 6Participants completed their written informed consent.

The Zurich cohort
26 FEP (age range: 18-48 years; mean: 24.35±6.89years), 48 SCZ (age range: 19-49 years; mean: 32.96±7.92years), and 28 HC (age range: 18-55 years; mean: 32.54±9.158][9] Two subjects had to be excluded.FEP were defined as patients with a clinical diagnosis of brief psychotic disorder, schizophreniform disorder or first-episode schizophrenia using Mini-International Neuropsychiatric Interview for DSM-IV (M.I.N.I). 10 FEP with a positive subscale item higher than 5 on PANSS were excluded.SCZ had a clinical diagnosis of schizophrenia.Exclusion criteria for patients were other current DSM-IV axis I disorder, benzodiazepines (>1mg/d lorazepam-equivalent), or extrapyramidal side effects.All FEP and SCZ were treated with antipsychotic medication, the dose was kept stable for minimum two weeks before the study.Exclusion criteria for HC were psychiatric disorders, history of psychiatric disorders, and substance abuse.In schizophrenia, psychotic and negative symptoms were measured by PANSS. 6

MRI data acquisition in the Zurich cohort
Data acquisition was performed on a Philips Achieva 3.0 T magnetic resonance scanner with a 32-channel SENSE head coil at the MR-Zentrum of the Psychiatric Hospital, University of Zurich.Anatomical data of the whole brain were collected using an ultra-fast gradient echo T1-weighted sequence with TR/TE/flip angle: 8.4ms/3.8ms/8°;160 slices covering the whole brain; matrix size: 240×240; voxelsize: 1×1×1mm3.
ARMS were defined as individuals meeting the following inclusion criteria based on the Basel Screening Instrument for Psychosis: 13 "attenuated" psychotic symptoms, brief limited intermittent psychotic symptoms, or a first-degree relative with psychotic disorder and a marked decline in social functioning.After a clinical follow up of 33.3 month, 15 individuals had transited to psychosis.FEP was diagnosed based on the operational criteria for first-episode psychosis according to the ICD-10 or DSM-IV, 14 but not yet for schizophrenia. 15Specifically, patients' scores on the BPRS were 4 or above on the hallucination item or 5 or above on the unusual thought content, suspiciousness or conceptual disorganization items. 1527 FEP were unmedicated, while 45 were treated with antipsychotics.Exclusion criteria for HC were psychiatric disorders or history of psychiatric disorders, head trauma, neurological illness, serious medical illness, substance abuse, and psychiatric disorder in family history.
The Scale for the Assessment of Negative Symptoms (SANS) was used to assess the severity of negative symptoms. 4

MRI data acquisition in the Basel cohort
Data acquisition was performed on a 3T magnetic resonance imaging scanner (Magnetom Verio, Siemens Healthcare, Erlangen, Germany) at the Basel University Hospital.Anatomical data of the whole brain were collected using a T1-weighted MPRAGE sequence with TR/TE/flip angle: 2000 ms/3.4 ms/8°; 160 slices covering the whole brain; FoV: 176 mm; matrix size: 256×256; voxel-size: 1×1×1mm3.

Assessment of cognitive deficits
For control and specificity analyses the Symbol-coding Task (SCT) of the Brief Assessment of Cognition in Schizophrenia, and Multiple-choice vocabulary intelligence test Part B (MWT-B) were investigated.Both are paper-pencil-based metrics.SCT evaluates attention and processing speed, 16 whereas MWT-B evaluates verbal intelligence. 17

Multi-site harmonization across scanners
As the present study combined data from four distinct acquisition sites, we controlled for the scanner effect using the open-source toolbox neuroCombat developed by Fortin et al. 18 Total intracranial volume (TIV), anterior, and posterior BFCN-VBM were harmonized using Empirical Bayes.Sex, age, group, and TIV (for BFCN-VBM) were used as covariates of no interest.For more information, see: https://github.com/Jfortin1/neuroCombat.

TIV normalization
To correct for individual differences in head size, BFCN-and gray matter (GM)-VBM were scaled to TIV by dividing each ROI multiplied by 10 6 (and 10 3 for global GM, respectively) by TIV, as previously described. 19

Statistical analysis
Specificity analysis for BFCN-VBM.To test whether data were normally distributed D'Agostino's K-squared test was used.Pearson and Spearman correlation analyses were used to investigate the putative association between BFCN-VBM and anticholinergic-burden of medication and antipsychotic medication in the patient groups showing alterations.Anticholinergic-burden of medication was measured using the anticholinergic-burden score (ACB; see: http://www.acbcalc.com), 20tipsychotic medication was calculated based on chlorpromazine equivalents (CPZ). 21Only medicated patients were included in the medication association analysis.The specificity of altered BFCN-VBM relative to changes in global GM was examined using ANCOVA across the spectrum, while controlling for the influence of global GM.To test whether age influences and moderates changes in BFCN-VBM, correlation analysis was performed between BFCN-VBM and age in both HC and patients, followed by a moderation analysis using the PROCESS package in SPSS https://www.processmacro.org/download.html.The effect of illness duration on BFCN-VBM in SCZ was investigated using Spearman correlation, and multiple regression analysis with interaction with anterior BFCN-VBM as dependent variable and age and illness duration as independent variables.Specificity analysis for altered BFCN-VBM and cognitive deficits.To control for the possible influence of ACB on the association between BFCN-VBM and cognitive deficits, we conducted partial correlation analysis with ACB scores as covariates of no interest.The specificity of cognitive symptoms associated with BFCN-VBM was investigated using correlation analysis with positive and negative PANSS and SANS scales, respectively, instead of cognitive scores, in patients.

Control and specificity analyses for lower anterior BFCN-VBM in schizophrenia
First, the influence of medication was tested by correlating the lower anterior BFCN-VBM with both anticholinergic-burden of medication, measured by ACB score, and with antipsychotic medication, measured by CPZ, respectively, in SCZ.The ACB score was significantly negatively correlated with lower anterior BFCN-VBM (rho=-0.22,p=0.009), suggesting that a higher level of anticholinergic-burden of medication is relevant for lower anterior BFCN-VBM in SCZ (Figure S1A).To evaluate the influence of anticholinergic burden of medication, ACB was used as an additional covariate in the ANCOVA model of anterior BFCN changes across the SCZ spectrum.Anterior BFCN-VBM remained at trend significant when controlling for ACB (F4,469=2.50,p=0.06).However, Dunnett's post hoc did not reveal significant differences between patient groups and HC.No significant correlation was found between anterior BFCN-VBM and CPZ (rho=-0.09,p=0.29), indicating that it is unlikely that current antipsychotic medication influences anterior BFCN-VBM of SCZ.
Next, changes of anterior BFCN-VBM were compared with changes in the general amount of global GM in SCZ, which might contribute to BFCN changes.Differences in TIV-normalized global GM-VBM between the four groups were tested by the use of ANCOVA.We found significant differences in TIV-normalized global GM across the schizophrenia spectrum (F3,470=5.87,p=0.01).Specifically, using Dunnett's' post hoc, global GM-VBM was lower in SCZ compared to HC (Figure S1B).Then, global GM-VBM was used as an additional covariate in the above-mentioned ANCOVA model of anterior BFCN changes across the spectrum to examine the potential relevance of global GM variations on the group differences observed in anterior BFCN-VBM.Global GM-VBM changes affected our ANCOVA result, which remained only at-trend significant when controlling for global GM-VBM (F3,469=2.07,p=0.10), suggesting that lower anterior BFCN-VBM in SCZ is not independent from global GM-VBM changes.To disentangle the effect of global GM and anticholinergic burden of medication, a partial correlation analysis was conducted between ACB and anterior BFCN-VBM in SCZ with GM-VBM as covariate-of-no-interest.When controlling for global GM-VBM, the correlation between ACB and anterior BFCN-VBM did not remain significant (rho=-0.12,p=0.15), indicating that the association between anterior BFCN-VBM and ACB is not independent of global GM-VBM.Furthermore, we found that ACB was also linked to global GM-VBM in SCZ (rho=-0.23,p=0.007), suggesting an influence of anticholinergic-burden of medication on global GM differences.
Next, we tested whether the lower anterior BFCN-VBM in SCZ is influenced by age.
Previous studies demonstrated that age influences BFCN volumes. 19,22It is unclear whether this is also the case in SCZ and whether age might modify the disorder effect on BFCN-VBM.First, by using correlation analysis, we verified a negative correlation between age and anterior BFCN-VBM in both HC (rho=-0.41,p<0.001) and SCZ (rho=-0.44,p<0.001), respectively.This result suggests that age influences anterior BFCN-VBM not only in HC but also in SCZ (Figure S1C).Then, moderation analysis was used to test whether age moderates the disorder effect (factor disorder with levels HC and SCZ) on anterior BFCN-VBM.The analysis showed no moderation of age on the relationship between disorder and anterior BFCN-VBM (ΔR 2 =1.6%, F1,306=0.60,p=0.44), indicating that lower anterior BFCN-VBM in SCZ do not interact with age.
Finally, we asked whether illness duration might modify the disorder effect on anterior BFCN-VBM.To test this, one should remember that age and illness duration are highly correlated (in our case: rho=0.74,p<0.001).Therefore, any test of illness duration effects has to account for age effects too.First, correlation analysis revealed a significant association between anterior BFCN-VBM and illness duration in SCZ (rho=-0.28,p<0.001; Figure S1D).Restricted to SCZ, we performed a multiple regression analysis with interaction with anterior BFCN-VBM as dependent variable and age and illness duration as independent variables.This analysis demonstrated that age and illness duration affect anterior BFCN-VBM in SCZ (F3,136=14.01,p<0.001,R 2 =0.24).We found again (as above) a main effect of age on anterior BFCN-VBM (p=0.003) but no significant effects for both illness duration (p=0.08) and its interaction with age (p=0.10).This result indicates that illness duration is at least at-trend independent from changes in global GM-VBM beyond age in SCZ.Since age does not appear to significantly modify the effect of the disorder on anterior BFCN-VBM, our result provides indirect support for the model that illness duration is unlikely to modify the disorder effect on BFCN-VBM in SCZ.

Control and specificity analyses for larger posterior BFCN-VBM in firstepisode psychosis patients
First, the potential effect of medication on larger posterior BFCN-VBM in FEP was assessed by the use of correlation analysis (Figure S2A).ACB, was not associated with larger posterior BFCN-VBM (rho=-0.07,p=0.56), suggesting that anticholinergic-burden of medication do not affect posterior BFCN-VBM.Similarly, antipsychotic mediation, CPZ, did not correlate with posterior BFCN-VBM in FEP (rho=0.11,p=0.35), suggesting no influence of current antipsychotic medication on BFCN-VBM changes.Furthermore, posterior BFCN-VBM differs between the groups (F2,262=7.15,p=0.001) and is larger in FEP-unmedicated compared to HC (p<0.001).This result indicates on the one hand, that it is unlikely that larger posterior BFCN-VBM in FEP is driven by antipsychotic medication, and on the other hand, that first episode psychosis might be associated with BFCN-VBM increases.Interestingly, no significant difference in posterior BFCN-VBM was found between HC and FEPmedicated, suggesting that treatment with antipsychotic medication might contribute to normalize larger posterior BFCN-VBM in FEP.
Next, we asked whether larger posterior BFCN-VBM in FEP was influenced by global GM changes.When controlling for global GM-VBM, posterior BFCN-VBM differed at-trend across the schizophrenia spectrum (F3,469=2.18,p=0.09; Figure S2B).Additionally, there were no differences in global GM-VBM between HC and FEP, suggesting that larger posterior BFCN-VBM in FEP was not influenced by global GM-VBM changes.To analyze this aspect further, we performed ANCOVA for the two groups (controlling for age, sex, and global GM-VBM).We found a significant difference between the groups (F1,262=8.59,p=0.04), with a larger volume in FEP compared to HC (p=0.004), using Dunnett's post hoc test.
To test whether age affects the group difference in posterior BFCN-VBM correlation analysis was used in the HC and FEP groups (Figure S2C).Significant negative correlations between posterior BFCN-VBM and age were found in HC (rho=-0.45,p<0.001), but not in FEP (rho=-0.15,p=0.14).Identical to the approach in anterior BFCN-VBM, moderation analysis for HC and FEP was performed.We found an attrend moderation effect of age on the effect of group on posterior BFCN-VBM (ΔR²=8.9%,F1,263=3.10,p=0.08), suggesting that that larger posterior BFCN-VBM in FEP might be driven by interactions between age and psychosis.

Control and specificity analyses for the association between anterior BFCN-VBM and cognitive impairment
First, the effect of anticholinergic-burden of medication on the association between anterior BFCN-VBM and TMT-A was studied using nonparametric partial correlation analysis.The correlation remained significant after controlling for ACB (rho=-0.22,p=0.04), suggesting that anticholinergic-burden of medication do not influence the link between anterior BFCN-VBM and TMT-A.
To investigate whether additional variables such as ACB, sex, and age influenced the relationship between lower anterior BFCN-VBM and TMT-A scores, we computed multiple regression analysis.The overall regression model was significant (R 2 =0.13,F81,4=4.19,p=0.004).We found that age significantly predicted TMT-A (beta=0.49,p=0.015).ACB, sex, and anterior BFCN-VBM had no significant effect on TMT-A.However, as the link between TMT-A and age is well-known, 23 and age was significantly correlated both with TMT-A (rho =0.42, p<0.001) and anterior BFCN-VBM (rho=-0.44,p<0.001), the latter indicating high collinearity between two predictors.Therefore, multiple regression is not the ideal method to investigate the predictions of TMT-A.We used hierarchical multiple regression with 4 blocks to investigate the effects of each variable on TMT-A (Table S1).The hierarchical multiple regression demonstrated that anterior BFCN-VBM predicts TMT-A values, as long as age is not included in the model.However, this is not surprising as age is highly collinear with anterior BFCN.
To ensure that the link between anterior BFCN-VBM and cognitive deficit does not depend on the test used for assessing cognitive performance, we investigated the relation between anterior BFCN-VBM and SCT as an alternative measure of cognitive performance with focus on attention and processing speed.ANOVA demonstrated that SCZ had significantly lower SCT scores compared to HC (F178,1=38.67,p<0.001).Correlation analysis demonstrated a positive correlation with SCT (rho=0.36,p<0.001) indicating that lower anterior BFCN-VBM in SCZ is associated with lower cognitive test performance, independent of the applied cognitive testing procedure (Figure S3A).
Next, the specificity of the association of lower anterior BFCN-VBM with cognitive deficit in SCZ was examined by correlating the cognitive scores with the anterior BFCN-VBM of HC (Figure S3B).No association was detected between anterior BFCN-VBM and TMT-A (r=-0.11,p=0.30).These findings suggest that the association between lower anterior BFCN-VBM and cognitive functioning is specific to SCZ.
To test for specificity of the association between BFCN-VBM and cognitive performance with respect to other symptom dimensions, anterior BFCN-VBM was correlated with positive and negative PANSS scales, respectively, representing psychotic and negative symptoms of SCZ (Figure S3C).No associations were found between lower anterior BFCN-VBM and PANSS positive (rho=-0.08,p=0.37) or negative (rho=-0.02, p=0.80), respectively, indicating that these symptoms do not link with the lower anterior BFCN-VBM in SCZ.In other words, lower anterior BFCN-VBM appear to link specifically only with cognitive deficits in SCZ.

Control and specificity analyses for the association between posterior BFCN-VBM and cognitive impairment
First, to test whether the association between posterior BFCN-VBM and Phonemic Fluency was influenced by anticholinergic-burden of medication, partial correlation analysis between posterior BFCN-VBM and Phonemic Fluency scores was computed.After controlling for ACB, no association between posterior BFCN-VBM and Phonemic Fluency was detectable (rho=-0.07,p=0.74), suggesting an influence of anticholinergic-burden of medication on this association (Figure S4A).
Next, to ensure that the link between posterior BFCN-VBM and cognitive deficit does not depend on the test used for assessing cognitive performance, we investigated the relation between posterior BFCN-VBM and MWT-B score as an alternative measure of cognitive performance.ANOVA demonstrated significantly reduced performance in MWT-B in FEP compared to HC (F111,1=4.39,p=0.038).Pearson's correlation analysis demonstrated that posterior BFCN-VBM correlated with MWT-B score at-trend (r=-0.20,p=0.10), suggesting that the relevance of posterior BFCN-VBM increases for impaired cognitive performance does not depend -at least attrend -on the used cognitive test.
Furthermore to test for specificity with respect to group, associations with cognitive impairment measured by Phonemic Fluency and posterior BFCN-VBM were also tested in the other groups of the spectrum and HC using correlation analysis (Figure S4B).The posterior BFCN-VBM showed no correlation in HC (r=0.02p=0.82) nor in ARMS (r=-0.29,p=0.16).However, a link was found between posterior BFCN-VBM and Phonemic Fluency in SCZ, which was at-trend to significance (r=0.17,p=0.09).This finding suggests that posterior BFCN-VBM of clinically more affected groups of the spectrum including FEP might be relevant for cognitive performance.
Finally, to test for specificity of the association between posterior BFCN-VBM and cognitive performance with respect to other symptom dimensions, the correlation of posterior BFCN-VBM with negative and psychotic symptoms, respectively, was investigated in FEP (Figure S4C).No associations were found between posterior BFCN and psychotic symptoms (r=-0.28,p=0.17), measured by PANSS positive, and negative symptoms (rho=-0.01,p=0.95), measured by SANS scale.These results suggest that neither psychotic nor negative symptoms link with posterior BFCN-VBM in FEP.

Limitations
The present study has several limitations.First, regarding the measurement, socalled volumetric MRI does not directly measure brain volume , making clear-cut and consistent interpretations of MRI-based changes difficult. 24,25Second, the integrity of the cholinergic system is assessed indirectly by MRI-based techniques.The volumetric measurements of the BFCN were determined by stereotactic mapping of the BFCN, and thus non-cholinergic neuronal populations interacting with cholinergic neurons in the BF are included within the ROI. 26,27A recent study reported that stereotactic mapping of nucleus basalis Meynert generates deviating results from post-mortem findings. 28While our results indicate that structural changes in the basal forebrain are relevant for patients' cognitive deficits, we cannot demonstrate that the functional outcome solely relies on changes in cholinergic neurons.Functional imaging studies might disentangle the functional relevance of BFCN alterations along the course of schizophrenia to a greater extent.Third, the present study examined cross-sectional data, thus, we cannot derive individual disorder progression trajectories.Longitudinal studies might be employed to provide better insights into the effect of psychosis, long-term treatment, and chronicity on structural changes in the cholinergic system.Fourth, distinct clinical and cognitive variables were acquired across the different sites, which engraves comparisons of associations across the whole dataset.Therefore, findings should be evaluated carefully.Finally, the sample size of specific subgroups, namely both at-risk individuals that transited to psychosis and unmedicated first-episode patients, was relatively modest, calling for larger sample size studies in the future.Table S1.Hierarchical multiple regression with four blocks investigating the effect of anterior BFCN-VBM, ACB, sex, and age on TMT-A in schizophrenia.ACB anticholinergic-burden of medication, BFCN basal-forebrain cholinergic nuclei.

Figure S3 .
Figure S3.Control and specificity analyses for the association between anterior BFCN-VBM and cognitive impairment.(A) Associations between anterior BFCN-VBM and symptoms were investigated with Spearman correlation.(i) Plot shows significant correlation between anterior BFCN-VBM and TMT-A score when controlling for ACB in SCZ (rho=-0.22,p=0.04).(ii) Plot shows a significant correlation between anterior BFCN-VBM and SCT in SCZ (rho=0.36,p<0.001).(B)

Figure S4 :
Figure S4: Control and specificity analyses for the association between posterior BFCN-VBM and cognitive impairment.Associations between posterior BFCN-VBM and symptoms were investigated using correlation analysis.(Ai) No significant correlation was found between posterior BFCN-VBM and PF when controlling for ACB in FEP (rho=-0.07,p=0.74).(ii) Plot shows a correlation at-trend to significance between posterior BFCN-VBM and MWT-B in FEP (r=-0.20,p=0.10).
FEP-medicated regarding posterior BFCN-VBM was investigated using ANCOVA controlling for age and sex.Dunnett's was used for post-hoc analysis.*<0.05,