Immunological protein profiling of first-episode psychosis patients identifies CSF and blood biomarkers correlating with disease severity

A


Introduction
The pathophysiology of schizophrenia remains unknown, which has hampered the development of novel antipsychotic drugs.However, growing biochemical and genetic evidence suggests activation of the immune system is an underlying component in the pathology of the disease (Sekar et al., 2016).Numerous studies have shown that schizophrenia is associated with high levels of peripheral inflammatory proteins (Reale et al., 2021).Thus, compounds such as C-reactive protein (CRP), interleukin (IL)-6, IL-17, IL-18, interferon-gamma (IFN-γ), and several chemokines are correlated with symptom severity in various psychopathology domains (Miller and Goldsmith, 2019;Gonzalez-Blanco et al., 2018), including cognitive performance (Luo et al., 2019;Mondelli et al., 2015).Direct evidence of immune activation in the brain is still relatively sparse and involves higher levels of IL-1β, IL-6, and complement component 4A in the cerebrospinal fluid (CSF) of patients with first-episode psychosis, chronic schizophrenia, and, notably, also in patients with bipolar disorder with a history of psychosis (Söderlund et al., 2009;Söderlund et al., 2011;Sasayama et al., 2013;Gracias et al., 2022).
Given the complexity of the immune system, comprehensive investigations to detect markers related to schizophrenia pathophysiology, including validation of their functional roles, are critical for early diagnosis and development of potential pharmacotherapies.A challenge in this field is the dilemma of adequately analyzing immune-related molecules such as cytokines or chemokines.To a large extent, previous studies have used variants of ELISA to analyze these compounds but with variable success.The use of a broad range of analytical technologies, including diverse protein panels, is critical for biomarker discovery in schizophrenia, because an individual assay procedure may not allow the detection of all relevant compounds.Thus, in addition to proteomics and metabolomics, additional techniques may facilitate biological insight into the intricate mechanisms of immune activation in the disease.
In the present study, we used the Olink Proximity Extension Assay (PEA) to analyze the immune status of blood and CSF in a wellcharacterized cohort of patients with FEP and healthy controls.Furthermore, we investigated whether the immune profile differed between the patients who were later diagnosed with schizophrenia and those who were not.The present study also explored potential correlations between inflammatory markers, disease severity, and cognitive function.

Patients
This study was approved by the Stockholm Regional Ethics Committee (Dnr 2010/879-31/1).Before participation in the study, all subjects signed a written informed consent form in accordance with the Declaration of Helsinki.Healthy controls and FEP patients were recruited in the Karolinska Schizophrenia Project (KaSP) in coordination with four psychiatric facilities located in Stockholm, Sweden: PRIMA Vuxenpsykiatri, Södra Stockholms Psykiatri, Norra Stockholms Psykiatri, Psykiatri Nordväst.Subjects were included in the study between March 2011 and March 2019.Patients with neurologic illnesses or severe somatic illness, a history of illegal substance addiction, and the presence of co-existing neurodevelopmental abnormalities such as autism spectrum disorder were excluded from the study.Substance use was assessed using urine testing.Macroscopic brain abnormalities were excluded using magnetic resonance imaging (MRI).To assess the clinical characteristics of the patients, Global Assessment of Functioning (GAF; where symptom and functioning dimensions were assessed separately), Positive and Negative Syndrome Scale (PANSS), Clinical Global Impression (CGI), Alcohol Use Disorders Identification Tests, and Drug Use Disorders Identification Tests were used.A structured clinical interview following the Diagnostic and Statistical Manual of Mental Disorders IV (DSM-IV), or a consensus diagnostic process, was used to establish the diagnosis.Patients were invited for follow-up after approximately 18 months for re-assessment and the following DSM-IV diagnoses were finalized for 59 out 77 patients: paranoid schizophrenia (n = 23), unspecified schizophrenia (n = 11), other schizophrenia (n = 1), disorganized schizophrenia (n = 3), residual schizophrenia (n = 1), delusional disorders (n = 3), brief psychotic disorder (n = 1), psychosis not-otherwise-specified (n = 5), schizoaffective syndrome (n = 2), schizoaffective disorder bipolar type (n = 2), major depressive disorder (n = 1) and no diagnosis (n = 6).Patients were allowed to use tobacco products, and 22 of the patients (26%) used nicotine derivatives (smoking or snuff) [Information on nicotine is missing in 3 patients].Some patients required medications with sedatives and anxiolytics during their involvement in the study.22 out of 77 patients were treated with benzodiazepines (BZDs) during the time of plasma and CSF sampling [Information on 2 patients is missing].Approximately 50 % of the patients (n = 38 of 77) were treated with antipsychotics at the time of sampling (medication information was missing for one patient].The maximum number of days on antipsychotic treatment for these patients was 26 days, mean number of days was 9.6 ± 1.9 (mean ± s.e.m, n = 24).The remaining thirty-eight patients did not use antipsychotics prior to, or at the time of plasma and CSF sampling.Patients taking antipsychotics used the following medications: olanzapine, aripiprazole, risperidone, quetiapine, paroxetine, or haloperidol (see Supplementary Table S1).Close relatives of patients provided information on the duration of untreated psychosis (DUP).The MATRICS Consensus Cognitive Battery (MCCB) was used to evaluate cognition 11 .This battery captures 7 key cognitive domains relevant to schizophrenia.Plasma and CSF sampling, PANSS, GAF, and cognitive measurements were all performed within 10 days after admission (mean time (±s.e.m.): 5.9 ± 0.3 days) for most of the patients (n = 61).For sixteen of the patients these investigations were performed between 14 and 40 days (mean time (±s.e.m.): 16.4 ± 2.0 days).

Healthy controls
Fifty-six healthy subjects (27 males and 29 females) were enrolled in the study through advertisement.Several tests were used to examine the healthy control subjects, including a physical examination, MRI, as well as blood and urine sampling.Subjects were examined for previous psychiatric illnesses using The Mini International Neuropsychiatric Interview.Further exclusion criteria were previous or current use of illegal drugs and first-degree relatives with psychotic illnesses.All participants were free from medication and any form of substance abuse evaluated with Alcohol Use Disorders Identification Tests/Drug Use Disorders Identification Tests at the time of the study.None of the participants had any first-degree relatives with psychiatric diagnoses.MRI results were analyzed by an experienced neuroradiologist at the MR Center, Karolinska University Hospital, Solna, and one healthy control showed a structural abnormality.This subject showed early signs of demyelination; however, this was not sufficient for a multiple sclerosis diagnosis since no neurological symptoms had been present previously and a clinical neurological examination showed normal results.This individual exhibited oligoclonal bands in the CSF; however, there were no other abnormalities.Immune profiling of this subject was similar to those of other healthy controls, and therefore, it was decided to include the results from this subject in the current study.Plasma sampling, lumbar puncture and cognitive tests for healthy controls were performed within 36.8 ± 5.3 days (mean ± s.e.m.).

Plasma collection
Venous blood samples were collected using standard venipuncture techniques in 10-ml tubes containing EDTA (BD Vacutainer; BD Hemograd, K2 EDTA).Most subjects (n = 106; 58 FEP patients and 48 healthy controls) underwent a blood sampling procedure between 07:45 and 10:00 h after a night's sleep.Due to clinical routines, sampling of the remaining FEP patients (n = 19) was performed between 10:30 and 13:15 h.To control for this confounding factor, procedures for seven healthy controls were performed at the same time intervals.Participants were instructed not to engage in physical activity during the preceding eight hours.Samples were centrifuged at 2900 rpm for 15 min within one hour of collection and kept at − 70 • C storage until analysis.

CSF collection
Standard lumbar puncture protocols were followed for cerebrospinal fluid (CSF) sampling.For all individuals in the correct decubitus posture, a disposable atraumatic needle (22G Sprotte, Geisingen, Germany) was inserted at the L 4-5 level.CSF (18 ml) was allowed to drip into a lightprotected plastic test tube.Following centrifugation (Sigma 5810R, Eppendorf, Hamburg, Germany at 3500 r.p.m. (1438g) for 10 min) to separate cells and supernatant, CSF supernatant from all patients was divided into 10 aliquots and frozen at − 80 • C within 1 h of collection.Most subjects (n = 106; 58 FEP patients and 48 healthy controls) underwent lumbar puncture between 07:45 and 10:00 h after a night's sleep.Owing to clinical routines, morning sampling was not possible for F. Eren et al. the remaining FEP patients (n = 19).To control for this confounding factor, seven healthy controls underwent lumbar puncture during the same time interval (10:30 and 13:15 h).The participants of the study were instructed not to engage in physical activity during the preceding eight hours, yet it was not possible to monitor their activity.

Olink analysis
Inflammatory proteins in the plasma were analyzed by Olink Bioscience, Uppsala, Sweden using a multiplex proximity extension assay technology (PEA) inflammation panel.92 markers of the PEA assay were measured using a microtiter plate where 96 pairs of DNAlabeled antibody probes were present in each well.The PEA assay uses pairs of oligonucleotide-linked antibodies with affinities for each other.The binding of the target proteins and antibodies brings PEA probes in close proximity, leading to the initiation of DNA polymerization.The new DNA sequence, obtained by real-time PCR and then quantified, served as a marker for the targeted protein.After normalizing the PCR results for inter-and intra-run variation using an internal control and an interplate control, the processed data were presented in the form of an arbitrary unit called Normalized Protein eXpression (NPX).NPX is a log2 transformed value representing the relative quantification between samples.More information on the panel and protein-specific validation is accessible from https://www.olink.com/downloads.

Statistics
Data analysis was performed in RStudio (version 1.3.1093),using R (version 4.1.0).First, 2 samples that failed to pass the quality control procedure of the Olink NPX manager were excluded from analysis.Second, proteins with NPX values below the protein-specific LOD were excuded from analysis.In total, 19.6% (18 of 92) of the inflammatory proteins were excluded from analysis in the plasma inflammation panel.In CSF samples, almost 50 % (45 of 92) of the inflammatory proteins were excluded from the inflammation panel analysis.Third, outlier detection was performed using OlinkAnalyze package in R where the standard deviations from the mean interquartile range and mean sample median for every sample were checked.Outliers were defined as samples with an average NPX below 1.5 first quartile or above 1.5 third quartile.After that, one sample was identified as an outlier in plasma samples and removed from further analysis, although this did not change the statistical outcome.The remaining NPX values below the LOD were considered missing values.
Demographic information (Table 1) includes age, gender, body mass index (BMI), nicotine usage, medication information, symptom severity scores (PANSS scores, GAF assessment and CGI), cognitive test scores, and DUP.Comparisons of age and BMI differences were made using Mann-Whitney U tests between healthy controls and FEP patients and Kruskal-Wallis tests between healthy controls, SCZ, and NON-SCZ groups.Differences between gender and nicotine usage among groups were identified using chi-square tests.Comparisons between DUP, PANSS, GAF, and CGI scores among SCZ and NON-SCZ groups were calculated using Mann-Whitney U tests.
To identify differentially expressed proteins between groups, the limma package (Ritchie et al., 2015) was used to apply multiple linear models with plate and nicotine usage as confounders, when appropriate.As a secondary analysis we also analyzed the results by adding plate, age and gender as confounders.Limma is a powerful package (Kammers et al., 2015) for identifying differentially expressed proteins, and it employs an empirical Bayes method that provides an advantage in moderating the standard errors of residual variances.First, the differential expression was calculated for FEP compared to that in healthy controls.Second, SCZ and non-SCZ were separately compared to healthy controls.Adjusted p values of the proteins were calculated using Benjamini-Hochberg correction method (Benjamini and Hochberg, 1995) and the adjusted p value threshold (FDR) was chosen to be 0.05, which means that proteins with an FDR < 0.05 were identified as differentially abundant between groups.The number of tests corrected for plasma and CSF was 74 and 47, respectively, corresponding to the number of proteins that passed the LOD threshold.Correlations between significant proteins and disease severity scores were calculated using Spearman correlation analyses and adjusted using the Benjamini-Hochberg correction method 14 .Correlations were first performed for the entire FEP sample and subsequently for patients who later developed SCZ.
Additionally, a principal component analysis (PCA) was performed to identify samples deviating from the overall distribution pattern and potential confounders.PCA analysis were done with "OlinkAnalyze" package in R, using built-in "olink_pca_plot" function and samples were analyzed by their plate, age, gender, and nicotine usage to see if any of these variables were potential confounders.
To identify covariance between top proteins that were significantly different and disease severity we performed multivariate linear regression using "CCA" package R and cc function.Statistical tests for identified correlations between variables and the canonical variates were performed using "CCP" package in R with p.asym function.Test statistics were chosen as "Roy's Largest Root" to control for dependent measures of cognitive tests.

Participants
Table 1 and Table 2 shows demographics and clinical characteristics of the participants at inclusion admission.No significant differences were observed between groups in terms of age, gender, or BMI.Total PANSS score was 71.0 ± 2.2, and DUP was 5.2 ± 1.6 months for FEP patients (see Table 1).The PANSS score did not differ between patients later diagnosed with schizophrenia (72.7 ± 3.2) and those not receiving such diagnosis (69.1 ± 4.2), but DUP was significantly longer in the group of patients later diagnosed with schizophrenia (patients who received SCZ diagnosis have 7.0 ± 2.4 months compared to the group of patients not receiving a diagnosis 3.7 ± 3.6 months).With regard to all proteins detected in the serum or CSF, we observed no difference between drug-naïve subjects (n = 38) and those treated with antipsychotics (n = 38) using differential expression analysis.

CSF
Of the 92 markers in the inflammation panel, 47 proteins in the CSF were above LOD (cutoff > 0.30, except for CCL20 and TNFRSF9 which were included in the analysis as they are important inflammation markers).3 proteins were detected between > 40% and < 56%.11 proteins were detected between > 84% and < 99%.32 proteins were detected in all samples.IL-8 and IL-18 were reliably measured in CSF, yet no significant changes were observed between groups.We identified two proteins, IL-12B and CD5, with significantly different levels in FEP patients and healthy controls.A lower concentration of CD5 in FEP patients was observed and this protein was the only that remained significant after multiple hypothesis testing (logFC = -0.23,adj.p.val = 0.03, Table 3).Results from CSF proteins of each plate are reported in Supplementary Tables S7 and S8.

Plasma
Differential expression analysis between healthy controls and disease groups (SCZ and non-SCZ) showed that following correction for plate and nicotine multiple testing, no significantly abundant proteins were found in the non-SCZ group as compared with healthy controls.It is important to point out that, after stratification, there may be a loss of statistical power.However, 15 proteins were upregulated in patients diagnosed with schizophrenia compared with healthy controls.Here, we again identified AXIN1 as the most differentially expressed protein with   5. Table 6 includes data corrected for age, gender, nicotine, and plate.Analysis of the differential expression between disease groups (SCZ vs NON-SCZ) showed no significant differences.

CSF
There were no significant proteins detected after multiple hypothesis testing between healthy controls and patients later received schizophrenia diagnosis as well as between healthy controls and non-SCZ patients (Table 5).

Correlations between OLINK inflammation plasma proteins and symptom ratings in FEP patients
We further investigated correlations between differentially expressed proteins and symptom ratings.The NPX values for AXIN1 (r s = 0.36, p-adjusted = 0.0048, STAMBP (r s = 0.38, p-adjusted = 0.034, IL7 (r s = 0.50, p-adjusted = 0.018) showed a high positive correlation with PANSS positive score in FEP patients and IL-10RA (r s = -0.7,padjusted = 0.048) showed a negative correlation with PANSS positive score given it was differentially lower in FEP patients and these correlations remained significant after correction for multiple testing.IL10-RA showed a negative correlation with PANSS-total score and remained significant after correction for multiple testing (r s = -0.36,padjusted = 0.024).Other correlations did not remain significant after multiple testing (Fig. 2).Correlation coefficients of all significant proteins and disease severity scores are shown in Supplementary Table S4.To identify covariance between top proteins that were significant and disease severity we performed multivariate linear regression.Identified canonical dimensions between proteins and disease severity were not statistically significant (p value > 0.05) for all combined dimensions.

Correlations between OLINK inflammation proteins and cognitive test scores
In FEP patients, inflammatory proteins SIRT2 and CXCL1 correlated with performance of the Fluency cognitive test (r s = 0.36, p value = 0.03 and r s = 0.25, p value = 0.03, respectively).This test measures speed of processing.Furthermore, MSCEIT a test measuring social cognition showed a correlation with the chemokines CXCL1 and CXCL5 (r s = 0.3, p Fig. 2. Correlation between disease severity and significant plasma proteins in FEP patients.Heatmap of correlations of NPX levels between significant proteins and disease severity scores (PANSS, GAF and CGI).Correlation analysis was performed using Spearman correlation (r s ) and numbers in the boxes indicate r s values.Color intensity indicates the strength of the correlations, red refers to positive correlation blue refers to negative correlations.p values were adjusted for multiple comparisons with Benjamini-Hochberg, * adjusted p value < 0.05.Fig. 3. Correlation of disease severity with significant plasma proteins in subjects later diagnosed with schizophrenia.Heatmap of correlations of NPX levels of significant proteins and disease severity scores (PANSS, GAF and CGI).Correlation analysis is done with spearman correlation (r s ) and numbers in the boxes indicate r s value.Color intensity indicates the strength of the correlations, red refers to positive correlation and blue refers to negative correlations.p values did not remain significant after multiple testing with Benjamini-Hochberg, * adjusted p value < 0.05.value = 0.01 and r s = 0.33, p value = 0.005, respectively).Correlations were not significant after correction for multiple testing.In patients later diagnosed with schizophrenia, no inflammatory markers correlated with cognition.

Discussion
In this exploratory study, we analyzed 92 inflammatory markers in the plasma and CSF of FEP patients and healthy controls using proximity extension assay technology by Olink Biosciences.Here, we investigated concentration differences between FEP patients and healthy controls, as well as differences in diagnosis-based stratification of FEP patients (SCZ and NON-SCZ), and performed correlation analysis with disease severity scores and cognitive function.After correction for covariates (plate and multiple analyses), 12 inflammatory proteins in plasma and one in the CSF were significantly different between healthy controls and FEP groups.Differential expression analysis further showed that, compared to healthy controls, 15 proteins in plasma were higher in patients later receiving a diagnosis of schizophrenia, whereas no protein was changed in those not receiving a diagnosis.
The most significant difference compared to healthy controls, with the highest log-fold change, was plasma AXIN1 in both FEP patients and patients diagnosed with schizophrenia.Interestingly, AXIN1 modulates Wnt signalling and regulates catenin-β 1 (CTNNB1) by being a part of the β-catenin destruction complex (Goto et al., 2016).The Wnt signaling pathway is essential for hippocampal development, neuronal proliferation and migration, brain regionalization, and synapse formation.Hence, it is an important target for understanding dysregulated brain systems as observed in schizophrenia (Okerlund and Cheyette, 2011).Indeed, a disrupted Wnt signaling pathway (Nusse and Clevers, 2017;Karabicici et al., 2021;Ferrari, 2014;Vallée, 2022) has been proposed in diseases where neuroinflammation is suggested to be part of the pathophysiology, such as neurodegenerative diseases, depression, and autism spectrum disorder.Notably, increased mRNA expression of Wnt signaling pathway genes in the brain has also been found in schizophrenia and bipolar disorder (Hoseth et al., 2018).In addition, a study investigating KEGG pathways, protein-protein interaction networks and genome-wide association studies, along with a schizophrenia gene resource database, found that AXIN1 is a potential gene in the Wnt signaling pathway, suggesting an important role for this protein in schizophrenia (Jia and Zhao, 2010).Similar to STAMBP and IL-7, AXIN1 correlated with positive symptoms in FEP patients and was negatively correlated with the anti-inflammatory marker IL-10RA.Hence, the present results suggest that AXIN1 is a potential biomarker of psychosis.
Caspase-8/CASP-8, which was higher in both FEP patients as well as patients later diagnosed with schizophrenia, is another protein of great interest in schizophrenia research.Caspase families mediate important cellular processes, such as apoptosis, necrosis, proliferation, differentiation, and inflammation (Venero et al., 2013).In this regard, CASP-8 has been shown to promote the pro-inflammatory cytokine interleukin-1β (pro-IL-1β) via transcription factor nuclear factor (NF-κβ) activation, which induces pro-inflammatory cytokines and regulates aging and inflammatory pathways (Ketelut-Carneiro et al., 2018;Weng, 2014).Previous studies from our lab show that IL-1β activation, via CASP-8, induces Tryptophan 2,3-Dioxygenase (TDO2), thereby increasing endogenous production of the N-methyl-D-aspartate (NMDA) receptor antagonist kynurenic acid (KYNA) (Sellgren et al., 2016).Elevated brain levels of KYNA in schizophrenia have been shown to correlate with psychosis and cognitive deficits in humans, and to induce such behavioral alterations in rodents- (Jungbeck et al., 2009).
Sirtuin 2 (SIRT2), yet another significant protein, is abundantly expressed in the brain and suggested to influence critical biological processes, such as cell differentiation, cell cycle progression, and genomic stability (Chen et al., 2021;Kim et al., 2011).An intriguing relationship has been discovered between SIRT2 and NF-κβ where SIRT2 deacetylates p65 and regulates the expression of certain NF-κβ family genes (Rothgiesser et al., 2010).SIRT2 has also been implicated in various pathologies involving neuroinflammation such as Alzheimer's disease, Parkinson's disease, frontotemporal dementia, stroke, and brain injury (Chen et al., 2021).Like AXIN1, SIRT2 interacts with Wnt/ β-catenin signaling.Thus, depletion of SIRT2 results in activation of Wnt target genes (Nguyen et al., 2014); and recent studies suggest that SIRT2 contributes to the differentiation or proliferation of tumor cells in colorectal, breast, and lung cancer (Li et al., 2021;Zheng et al., 2012).The high plasma levels of SIRT2 in patients with schizophrenia could at least partly explain their increased cancer vulnerability and mortality (Wang et al., 2019;González-Rodríguez et al., 2020).The present results indicate that SIRT2 should be added to the list of biomarker candidates also for schizophrenia.
Additionally, peripheral levels of chemokines, such as CXCL1, CXCL5, and CXCL6, were higher in FEP patients and in patients diagnosed with schizophrenia.In the latter patient group also CXCL11, CCL8/MCP-2, and CCL11/Eotaxin-1 were elevated.The physiological importance of chemokines includes immune response function, mediating the migration of leukocytes and macrophages, as well as chemotaxis induction (Milenkovic et al., 2019).Their roles in different psychiatric disorders have been investigated earlier (Milenkovic et al., 2019;Hong et al., 2017;Teixeira et al., 2018;Misiak et al., 2020), as well as their influence on neurobiological pathways, such as neurogenesis, synaptic transmission, synaptic plasticity, and maintaining communication between neurons and glial cells, have previously been established (Stuart and Baune, 2014).CXCL1, CXCL5, CXCL6, and CXCL11 are related genes in the AKT pathway and are all involved in the development and functioning of the central nervous system.Their possible role in schizophrenia has been reviewed by Zheng et al. (Zheng et al., 2012;Kuo et al., 2012;Zhao et al., 2017;Liu et al., 2019).
In the present study, also TNFSF14 was found higher in the plasma of FEP patients who were later diagnosed with schizophrenia.This cytokine is primarily expressed in immature dendritic cells, monocytes, and activated T-cells (Murphy et al., 2006) and has been identified to play a role in the pathogenesis of various inflammatory diseases, such as COVID-19 (Arunachalam et al., 2020;Perlin et al., 2020), irritable bowel syndrome (Jungbeck et al., 2009), asthma, (Doherty et al., 2011) and non-alcoholic fatty liver disease (Herrero-Cervera et al., 2019).Furthermore, in a recent study, investigating osteogenesis, it was shown that treatment with recombinant TNFSF14 on human bone-marrowderived mesenchymal stem cells activates WNT/β-catenin signaling and p38 MAPK signaling pathways.Both pathways are suggested to be implicated in the schizophrenia pathogenesis (Heo et al., 2021;Funk et al., 2012;Vallée, 2022).
Generally, protein concentrations are lower in the CSF than in the plasma and the sensitivity of the OLINK immune panel allowed for the detection of only 47 CSF proteins.Surprisingly, CD5 was the only protein that was significantly changed in FEP patients.The biological role of CD5 is still obscure and apart from its use as an immune marker of T cells for decades, growing evidence suggests that the compound negatively regulates the B cell receptor and modulates autoimmune reactions (Burgueño-Bucio et al., 2019).Documentation of its role in the pathophysiology of schizophrenia is sparse, and, to our knowledge, the present study is the first to suggest that CD5 is downregulated in schizophrenia.One may argue that circulating immune mediators in the blood may not be essential for psychiatric diseases since they do not easily pass through the blood-brain barrier.However, the importance of blood protein levels in psychiatric disorders should not be ignored.Thus, circulating immune mediators may communicate with the brain in several ways, including activation of afferent nerves and endothelial cells in small blood vessels of the brain, or via active transport of immune-related molecules through the blood-brain barrier (Dantzer et al., 2008;Quan and Banks, 2007;Vichaya et al., 2020;Blomqvist and Engblom, 2018;Mapunda et al., 2022;D'Mello et al., 2016).Thus, peripheral low-grade inflammation may propagate by immune-related proteins knocking on the door to the brain, sending message of increased peripheral immune activity.
A recent study analyzed peripheral inflammatory markers in schizophrenia using Olink inflammatory panel (Klaus et al., 2021), similar to the one presently used (Cathomas et al., 2021).However, in that study, serum CCL20 and TNFSF10 were significantly higher in patients with schizophrenia than in healthy controls.This inconsistency with the present results could be related to differences in cohort characteristics.Thus, in Klaus et al., 2021, only patients with chronic schizophrenia treated with antipsychotics were investigated, whether present results were obtained from FEP patients involving drug-naive individuals.This discrepancy may emphasize that low-grade peripheral immune activation seen in schizophrenia may qualitatively vary along the course of the disease or may be affected by antipsychotic treatment.
The strengths of the present study include the relatively large number of proteins analyzed simultaneously, the unique longitudinal cohort of FEP patients with follow-up diagnoses, detailed knowledge of disease severity, and measurements of cognitive performance.However, comparison of abundance between proteins is not possible with current techniques, and a few inflammatory proteins were below proteinspecific LOD.Given the complexity of the immune system, the major limitation of this study is that the inflammatory panel used covers although many, but still a limited number of proteins.Thus, immunerelated proteins, like IL-1β and IL-6, as well as enzymes of the kynurenine metabolism (Erhardt et al., 2017) and some cytokines previously implicated in the pathophysiology of schizophrenia, have escaped our investigation in this regard.Important cytokines, such as IL-8 and IL-18 were reliably measured, however we did not observe any significant changes between groups in this regard.Another limitation of the present study is the somatic and mental status of the first-episode psychosis patients at admission.Thus, one should not ignore that conditions such as stress, anxiety, and sleep deprivation may have affected the levels of immune-related proteins.Importantly, proteins not found to be elevated in the CSF of patients with schizophrenia, e.g., due to the sensitivity of the presently used analysis techniques, may still be clinically relevant for the disease.
In conclusion, the results of the present study show that several immune markers in plasma are significantly higher in FEP patients.Some of these markers also correlated with illness severity and have been previously implicated in psychosis.Notably, at least three of the most prominent proteins found to be higher in the plasma of subjects with schizophrenia, i.e.AXIN1, SIRT2, and TNFSF14, directly interfere with WNT/β-catenin signaling, supporting a prominent role of this pathway in the pathophysiology of schizophrenia (Vallée, 2022).

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

F
.Eren et al.

Table 1
Demographics and clinical characteristics of healthy controls (HC) and FEP patients.

Table 2
Demographics and clinical characteristics of healthy controls (HC) and stratified patient groups.Differentially abundant proteins in plasma or CSF comparing healthy controls and first-episode psychosis patients (adjusted for plate).
HC: Healthy controls; SCZ: Patients who received schizophrenia diagnosis; NON-SCZ: Patients who did not receive schizophrenia diagnosis; BMI: Body Mass Index; DUP: Duration of untreated psychosis; PANSS: Positive and Negative Syndrome Scale; GAF: Global Assessment of Functioning; CGI: Clinical Global Impression.a: chi-square test.b:Mann-Whitney U test.c: Kruskal-Wallis test.Table 3 Logfold changes between groups and p values are calculated with limma package in R. Benjamini-Hochberg correction method is used for plasma and CSF to calculate adjusted p values.Significance level chosen for adjusted p value is < 0.05.logFC: Logfold change, CSF: Cerebrospinal fluid.*: Of all 92 proteins 74 proteins were above LOD in plasma.**: Of all 92 proteins 47 proteins were above LOD in CSF.a: Adjusted for plate.

Table 4
Differentially abundant proteins in plasma and CSF comparing healthy controls and first-episode psychosis patients (age, gender, and plate).
logFC: Logfold change, CSF: Cerebrospinal fluid.*: Of all 92 proteins 74 proteins were above LOD in plasma.**: Of all 92 proteins 47 proteins were above LOD in CSF.a: Adjusted for plate, age, and gender, corrected for 74 proteins in plasma and 47 proteins in CSF.Multiple hypothesis testing was conducted separately for plasma and CSF.

Table 5
Differentially abundant proteins in plasma and CSF comparing healthy controls and patient who later received a schizophrenia diagnosis (corrected for plate and nicotine).
**: Of all 92 proteins 47 proteins were above LOD in CSF.a: Adjusted for plate and nicotine.

Table 6
Differentially abundant proteins in plasma and CSF comparing healthy controls and patient who later received a schizophrenia diagnosis (corrected for age, gender, nicotine, and plate).Logfold changes between groups and p values are calculated with limma package in R. Benjamini-Hochberg correction method is used for plasma and CSF to calculate adjusted p values.Significance level chosen for adjusted p value is < 0.05.
logFC: Logfold change, CSF: Cerebrospinal fluid.*: Of all 92 proteins 74 proteins were above LOD in plasma.**: Of all 92 proteins 47 proteins were above LOD in CSF.a: Adjusted for plate, age, nicotine and gender.