Cerebrospinal fluid, antineuronal autoantibody, EEG, and MRI findings from 992 patients with schizophreniform and affective psychosis

The central role played by cerebrospinal-fluid (CSF) examinations including antineuronal autoantibody (Ab) testing is increasingly recognized in psychiatry. The rationale of this study was to present a multimodally investigated group of patients. In total, 992 patients were analyzed for CSF alterations: 456 patients with schizophreniform and 536 with affective syndromes. Ab measurement included testing for established antineuronal IgG-Abs against intracellular antigens in serum (Yo/Hu/Ri/cv2[CRMP5]/Ma1/Ma2/SOX1/TR[DNER]/Zic4/amphiphysin/GAD65) and for cell surface antigens in the CSF (NMDAR/AMPA-1/2-R/GABA-B-R/LGI1/CASPR2/DPPX). In 30 patients with “red flags” for autoimmune psychosis, “tissue tests” were performed. Additional diagnostics included MRI and EEG analyses. CSF white-blood-cell counts were increased in 4% and IgG indices in 2%; CSF-specific oligoclonal bands were detected in 4%; overall, 8% displayed signs of neuroinflammation. In addition, 18% revealed increased albumin quotients. Antineuronal Abs against intracellular antigens were detected in serum in 0.6%. Antineuronal Abs against established cell surface antigens were detected in serum of 1% and in the CSF of 0.3% (CSF samples were only questionably positive). Abnormal IgG binding in “tissue tests” was detected in serum of 23% and in CSF of 27%. In total, 92% of the Ab-positive patients demonstrated at least one sign of brain involvement in additional diagnostics using CSF, MRI, EEG, and FDG-PET. In summary, CSF basic analyses revealed signs of blood–brain-barrier dysfunction and neuroinflammation in relevant subgroups of patients. Established antineuronal IgG-Abs were rare in serum and even rarer in the CSF. “Tissue tests” revealed frequent occurrences of Ab-binding; therefore, novel antineuronal Abs could play a relevant role in psychiatry.


Introduction
In the last decade, the study of autoimmune encephalitis (AE) and autoimmune psychosis (AP) has rapidly developed 1,2 , largely due to the discovery of anti-N-methyl-D-aspartate receptor (NMDAR) encephalitis in 2007. Prior to this discovery, patients with suspected AE were mainly tested for antineuronal autoantibodies (Abs) against intracellular antigens in the context of paraneoplastic processes 3 . Since 2007, however, the importance of AE has increased in the field of psychiatry with the recognition that anti-NMDAR encephalitis often manifests with psychotic symptoms and that these patients are usually seen initially by psychiatrists [4][5][6][7] . At the same time, several other antineuronal Abs against cell surface antigens (e.g., LGI1) have been discovered that are associated with psychiatric symptoms. These Abs seem to play a direct pathophysiological role and can occur non-paraneoplastically 8 . The discovery of new Abs is expected in the future; therefore, in this respect, screening examinations using unfixed rodent tissue sections can be helpful for Ab detection 9,10 . Nevertheless, at present, large investigations of the prevalence of antineuronal Abs in patients with psychoses have been limited to unimodal studies using serum [11][12][13] . Smaller investigations of cerebrospinal fluid (CSF) have revealed antineuronal Abs at significantly lower numbers when compared to serum analyses 14,15 . Multimodal studies that include electroencephalography (EEG), magnetic resonance imaging (MRI), and especially CSF basic analyses with quantification of antineuronal Abs are lacking at present. CSF diagnostics play a central role in the context of antineuronal Abs, since low antigen-specific immunoglobulin (Ig) G Ab titers can be detected even rarely in the serum of healthy individuals [11][12][13] . This finding emphasizes the necessity of evaluating the pathophysiological relevance of these Abs by CSF analyses 2,7,16 . In addition, extended diagnostics using EEG, MRI, or [ 18 F] fluorodeoxyglucose positron emission tomography (FDG-PET) would allow the detection or exclusion of brain involvement in sero-positive patients. According to recently published international consensus criteria for AP, a diagnosis of "probable AP" requires typical CSF, MRI, or EEG findings, while confirmation of a diagnosis of "definite AP" requires the detection of IgG antineuronal Abs in the CSF 2 .

Rationale
At the Department of Psychiatry and Psychotherapy of the University Hospital in Freiburg, patients with schizophreniform syndromes have routinely been offered lumbar punctures (LPs) since approximately June 2009, based on new developments and personal experiences with AEs/ APs 14,17,18 . The aim of the present study was to conduct a retrospective evaluation of a large cohort of patients with schizophreniform and affective syndromes who underwent multimodal examinations consisting of CSF analyses, antineuronal Ab testing in serum/CSF, EEG, and MRI.

Patients and methods
The study received approval from the local ethics committee of the University of Freiburg (EK Fr 396/18). All patients gave written informed consent before LP. Between January 2006 and November 2019, 992 patients were included in the present study.

Patient cohort
All inpatients with schizophreniform syndromes (according to the International Statistical Classification of Diseases and Related Health Problems criteria, version 10 [ICD-10]: F20.X-F29.X, F06.0-2, F10.5-F19.5) and affective syndromes (unipolar depression following ICD-10: F32.X, F33.X, F06.3 and bipolar disorder following ICD-10: F30.X, F31X, F06.3) who underwent an LP at the Department of Psychiatry and Psychotherapy were included. Patients who were transferred to the Department of Neurology for further investigations were not included. Only the first LP results were analyzed for each patient. The patients were clinically diagnosed by experienced senior psychiatrists according to the ICD-10 criteria. For statistical analyses, patients were classified according to their predominant psychiatric syndromes. Patients with schizophreniform or affective syndromes who were also diagnosed with dementia were excluded (ICD-10: F00.X-F04.X). Other preexisting (e.g., earlier stroke) or newly described (e.g., migraine) neurological comorbidities were recorded but not considered as an exclusion criteria if the LP was performed within the diagnostic process of the psychiatric disorder. Since approximately June 2009, patients with schizophreniform syndromes have been offered CSF analysis routinely at our institution, whereas patients with affective syndromes have not been examined routinely. Clinical data were extracted from the patient discharge letters. Some parameters were also taken from the basic clinical documentation, such as the Clinical Global Impression (CGI; ref. 19 ), Global Assessment of Functioning (GAF; ref. 20 ), and psychopathological scores following the German Association for Methodology and Documentation in Psychiatry ("AMDP-scores"; ref. 21 ).

Antineuronal antibodies against cell surface antigens
The analysis with fixed biochip assays has been established since 2011 (Euroimmun-kits®, Lübeck, Germany). This initially involved the testing of IgG Abs against five antigens (NMDAR, AMPA-1/2-R, GABA-B-R, LGI1, CASPR2). In 2018, testing for Abs against DPPX was added ("mosaic 6" from Euroimmun®; Lübeck, Germany). The tests were initially performed exclusively in the CSF, but since approximately January 2016, both CSF and serum samples have been routinely analyzed. Prior to that date, combined CSF and serum samples were only conducted in particular cases. The Ab findings were divided into questionably positive (+), slightly positive (+ +), and clearly positive (+++). From 2006 to 2011 (and later in special cases), material was sent to the reference laboratory at John Radcliffe Hospital (Prof. Vincent, Oxford, United Kingdom) for anti-NMDAR IgG Abs testing using live cell assays and for anti-VGKC IgG Abs testing using RIAs. The results of these tests are already published 14,24 . For reasons of consistency, the results of the previously published work are listed in the results section; additional unsystematic investigations at Oxford for individual cases are not analyzed here. Testing for Abs associated with demyelinating diseases (AQP4 and MOG) has been established since 2018 on our ward for schizophreniform psychosis using Euroimmun® biochip kits (Lübeck, Germany). Since the end of 2018, tissue-based assays using indirect immunofluorescence on unfixed murine brain tissue were established in patients with "red flags" for AP (e.g., catatonia or CSF specific OCBs) 27,28 (Prof. Prüss, Charité and DZNE, Berlin, Germany; see exemplary in 9 ). Only positive (+++) IgG antibody binding patterns were included in the analysis.

Instrument-based diagnostics EEG
All patients were offered an EEG examination on admission. The EEGs included a resting-state EEG for approximately ten minutes and (if possible) a hyperventilation (HV) phase for~3 min. The EEGs were evaluated by the responsible physicians. In addition, an automated detection of intermittent generalized rhythmic delta/theta activity (IRDAs/IRTAs) was performed. The methodology has been described in the previous papers 29,30 , and the findings were divided into pre-HV-, post HV-, HVdifference (post-HV-pre-HV), and overall-IRDAs/IRTAS.

MRI
The MRI protocol included at least T1-weighted (axial 5 mm thick fast spin echo slices on a 1.5 Tesla, MPRAGE sequence with isotropic 1 mm 3 voxels on a 3 Tesla scanner), DWI (axial 5 mm thick slices), and FLAIR sequences (coronal 3 mm thick fast spin echo slices on a 1.5 Tesla, 3D SPACE sequence with isotropic 1 mm 3 voxels on a 3 Tesla scanner). The evaluation was performed by experienced senior physicians in neuroradiology.

Available datasets
Due to the retrospective approach, not all parameters were available for all patients; moreover, the procedures have been continuously optimized and adapted over the past years. The available datasets are presented in Table 1.

Statistical analyses
Statistical analysis was performed using the Statistical Package for the Social Sciences, version 24 (IBM Corp., Armonk, NY, USA). The results are largely presented in a descriptive manner. Independent sample t-tests were used for the comparison of dimensional variables between the subgroups of patients without age difference. ANCOVA analyses with age correction were used to compare all other dimensional variables (e.g., CSF protein concentration between patients with schizophreniform and affective syndromes) between the subgroups with age difference. Categorical variables (e.g., sex) were compared using Chi 2 tests. A binary logistic regression was performed for age-dependent categorical variables (e.g., number of positive OCBs) between different aged groups. Correlations between CSF basic parameters (WBC count, protein, AQ, and IgG index) with EEG-IRDA/IRTA rates, laboratory results (T3/T4, TSH), clinical findings (number of suicide attempts and number of earlier inpatient stays), and psychometric scores (GAF, CGI, AMDP-scores) were analyzed using Spearman correlation. For correlation analyses, all patients were analyzed together. A p-value of <0.05 was defined as statistically significant for group comparisons and correlation analyses. Due to the  exploratory approach of statistical analyses, no correction was performed for multiple testing.

Description of the study population
A total of 992 patients were analyzed. Overall, 456 patients presented with schizophreniform syndromes (46%) and 536 with affective syndromes (54%; 455 with unipolar depression and 81 with bipolar disorder). The two subgroups differed significantly in age (p < 0.001). The detailed findings are summarized in Tables 2 and 3. The increase in LPs during the study period is summarized in Fig. 1.
Patients with pronounced signs of a BBB dysfunction were further analyzed. Thirty percent (N = 7/23) of the patients with a greatly increased protein concentration of >1000 mg/L (5% or N = 23/448 of the patients with increased protein concentrations and 2% or N = 23/991 of all patients) suffered from schizophreniform and 70% (N = 16/23) from affective syndromes. Comparing the age of patients with protein concentrations >1000 mg/L (N = 23; M = 50.48±16.21 years) and all patients with protein levels <1000 mg/L (N = 968; M = 42.59 ± 17.93 years) highlighted significant differences (F = 1.012, p < 0.037). When comparing the patients with protein concentrations >1000 mg/L (N = 23) and all patients with protein levels <1000 mg/L (N = 968), no significant differences in the rate of EEG (Wald = 1.497, p = 0.221) or MRI alterations (Wald = 0.196, p = 0.658) were found. In terms of inflammatory CSF alterations (WBC count, IgG Indices, CSF specific OCBs), significant differences (Wald = 9.187, p = 0.002) were detected with higher rates in patients with protein concentrations >1000 mg/L (26%; vs. 7% in patients with protein concentration <1000 mg/L). There were differences in the number of earlier suicide attempts with higher rates in patients with protein concentrations >1000 mg/L (64%) versus patients with protein concentration <1000 mg/L (35%; Wald = 4.307, p = 0.038), but no differences in the number of earlier patient instays, and different psychopathological scores (GAF/ AMDP scores). Among the patients with elevated agerelated AQs (N = 174/989), 49% (N = 85/174) suffered from schizophreniform and 51% (N = 89/174) from affective syndromes. Their average age was 42.88 (±16.43) years. When comparing the patients with elevated AQs and those with normal AQs (N = 815/989), no differences in the rate of EEG/MRI and inflammatory CSF abnormalities were found. Patients with elevated age-related AQs had higher rates of earlier suicide attempts (in 47%) compared with patients with normal AQs (34%; Chi 2 = 4.143, p = 0.042).
The group of patients with first-episode schizophreniform syndromes (N = 188) showed the following altera-

Autoantibody testing
The detailed Ab findings are summarized in Table 5. Anti-thyroid Abs were detected in 17% of all patients.
Established antineuronal Abs against cell surface antigens were detected in the serum of six patients (out of 475 patients tested overall [1%]) and in the CSF of two patients (out of 741 patients tested overall [0.3%]). The two conspicuous CSF samples were only questionably positive. Positive antineuronal Abs against intracellular antigens in serum were detected in six patients (out of 826 patients tested overall [0.7%]). In two of these patients, Abs were also detected in the CSF (not systematically analyzed). Overall, 31 of 826 patients showed slight Ab reactivity against intracellular antigens (4%). One of 102 tested patients was positive for anti-MOG Abs in serum (1%).
Overall, positive established antineuronal Abs were detected in the serum of 12 patients (1% of 826 tested patients), and in the CSF of four patients (0.5% of 741 tested patients). However, not every single Ab was tested in all patients. A tendency was noted for a more frequent presence of serum Abs against cell surface antigens in patients with schizophreniform syndromes (p = 0.061). ). Seven patients (23%) had positive results in their serum, and eight patients were positive in their CSF (27%), mostly with predominant IgG binding to cerebellar and/or hippocampal granule cells (for details see Table 5).
Initially, selected samples (N = 39) were also examined in the reference laboratory at Oxford. The results have already been published for five positive cases, with low titer anti-VGKC Abs in four patients, and clearly positive anti-NMDAR Abs in one female patient 14,24 .

Instrument-based diagnostics
The EEG showed abnormalities in 25% of the patients, most frequently as alterations in the form of IRDAs/ IRTAs (in 17%). They were significantly more frequent in patients with schizophreniform psychosis (p > 0.01). In the automated IRDA/IRTA detection, tendencies for different IRDA/IRTA rates after HV (p = 0.075), and for the IRDA/IRTA difference (p = 0.066), with higher rates in patients with schizophreniform syndromes, were found. MRIs revealed overall changes in 72% of patients, with the most frequent being non-specific white matter changes (in 42%, including each individual non-specific lesion), in 9% the MRI findings were compatible with (post-) inflammatory changes. The findings are shown in detail in Table 6.

Description of antineuronal antibody-positive patients
A total of 24 patients were positive for antineuronal Abs (this includes an anti-NMDAR Ab-positive, older case tested at Oxford; however, weakly positive anti-VGKC Ab titers or weak reactivities in the ravo blot® for Abs against intracellular antigens were not considered as positive). This group included significantly more patients with schizophreniform psychoses (N = 18; Chi 2 = 6.577, p = 0.010). Overall, 58% of the Abpositive patients had CSF alterations (signs of inflammation in 22%; increased AQs in 21%), 54% had MRI signs, and 33% had EEG abnormalities. In addition, 60% of the patients examined with FDG-PET (N = 9/1 5) displayed abnormalities. In summary, signs of brain involvement were detected in 92% of the clearly Abpositive cases (one alteration in 33%, two alterations in 29%, three alterations in 25%, and four alterations in 4%). The findings in the clearly Ab-positive patients are presented in detail in Table 7.

Discussion
This study describes the multimodal diagnostic assessment of a large group of patients with schizophreniform and affective psychoses in a naturalistic inpatient setting in a tertiary care hospital. The main CSF results were signs of BBB dysfunction with increased AQs in 18% and inflammatory CSF alterations in 8% of all patients. Positive antineuronal IgG Abs against established intracellular antigens were detected in serum in 0.6% of the patients. Antineuronal IgG Abs against established cell surface antigens were detected in serum of 1% of the patients and in the CSF of 0.3% (CSF samples were only questionably positive). However, patterns of novel antineuronal Abs using tissue tests were detected in the serum and/or CSF of 30% of patients with schizophreniform or affective syndromes and red flags for AP.
The frequent signs of BBB dysfunction in 18% of all patients are consistent with results from a recent metaanalysis of patients with schizophrenia and affective disorders 31 . The BBB dysfunction significantly correlated with more severe symptoms (as measured by suicide attempts and CGI score) in the present sample. Current knowledge does not clarify whether these changes are primarily involved in the pathophysiology of mental illness or whether they are triggered secondarily by psychotropic drugs [32][33][34] . However, our data rather suggest that the findings are not caused by medication, since the results did not differ significantly between patients with and without psychotropic drug administration. Irrespective of the cause, a disturbance in BBB function can induce a harmful interaction between the innate brain and adaptive peripheral immunity 35 . This, in turn, allows the transfer of antineuronal Abs (e.g., against NMDAR) from the serum to the CSF, thereby leading to anti-brain effects 36 . Inflammatory changes that included mild pleocytosis, elevated IgG indices, or CSF specific OCBs were also detected in a relevant, subgroup (8%) of all patients. CSF pleocytosis was usually only subtly pronounced (in 87% from 5 to 30/µL); therefore, higher cell counts can be assumed to lead to a fulminant disorder and these patients are not treated on psychiatric wards. With regard to the 8% of patients who showed schizophreniform syndromes with inflammatory changes, the lower prevalence figure compared to some preliminary studies 14,15,23,24,37 may reflect the screening approach, which has led to an increasing number of patients undergoing LPs over the last few years (see Fig. 1). These inflammatory changes are compatible with pathogen-related pathologies, but they would also be typical for an AE/AP. Both AEs and APs are associated with slightly increased WBC counts or increased IgG indices/CSF specific OCBs 1,2,6,16 .
Antineuronal Ab-associated AEs have recently been described mostly in association with schizophreniform symptomatology 2 . In fact, in the present study samples as  well, the detection of serum IgG Abs tended to be more frequent in patients with schizophreniform syndromes.
The finding of only a few patients showing questionable CSF Ab positivity is consistent with another study in which 124 patients with schizophreniform psychosis, examined using the same methodology, displayed only negative CSF samples 15 . However, notably, in most of our cases, the serum Ab-positive patients also showed signs of brain involvement in further investigations; indeed, 92% of the Ab-positive patients showed at least one alteration in CSF, MRI, EEG, or FDG-PET findings. Therefore, our assessment is that all patients with antineuronal IgG serum Abs should undergo a careful diagnostic workup to verify possible brain involvement. This holds even if the     patients are CSF Ab negative, since isolated serum Abs can still have therapeutic consequences 38 . According to current international consensus criteria, the detection of antineuronal Abs in the serum, in combination with typical EEG or CSF alterations, is indicative of a "probable AP" 2 .
The role of tissue tests in selected clinical cases is also worth noting and was clinically relevant in several cases. Tissue tests analyzing serum and CSF were frequently conspicuous in patients with red flags for AP. Experiences with some patients have already been published 9,39,40 . Besides granule cell patterns, Abs against vascular structures were found, most likely directed against endothelial cells. The significance and specificity of these findings is not yet clear, as data on the prevalence of these findings in healthy controls is lacking. Similar findings were recently described in association with neuromyelitis opticum (NMO) spectrum diseases 41 . Pathophysiologically, Abs directed against endothelial cells might lead to a BBB dysfunction. In this context, different pathophysiological mechanisms could contribute secondarily to the development of psychiatric syndromes 35 . In addition, Abs against myelin structures were found. These findings are interesting in light of the constantly expanding range of NMO spectrum diseases 42 and their previously described association with psychotic and affective symptoms 43,44 . In Table 5 continued *One patient had two +++ antibody findings: HuD and SOX1. **One patient had multiple + antibody findings: Zic4, Sox1, Yo, CV2, and HuD. ***One patient had multiple + antibody findings: Zic4, Sox1, CV2(CRMP5). ****One patient with two slightly positive antibody findings: anti-SoX1 and Zic4. *****One patient had multiple + antibody findings: anti-Zic4, Sox1, and Ma2. ******One patient had two + antibody findings: Sox1 and Zic4. The overall alterations are marked in bold. summary, our opinion is that novel Abs against so far unknown antigens could play a decisive role in a subgroup of patients with severe mental disorders. A major limitation of the present study is its retrospective, open, and uncontrolled design, which meant that many patients did not receive all tests. For example, the measurement of established antineuronal Abs against cell surface antigens was not introduced until 2011 in our department. Initially, these Abs were only tested in the CSF, as CSF testing was considered more sensitive for the detection of anti-NMDAR encephalitis 45 . Our own observations revealed that anti-NMDAR encephalitis can probably also occur in patients with isolated positive serum results 38 , so serum analyses were introduced later, in addition to CSF analysis. The retrospective approach of the study also meant that confirmatory test results for the positive Ab findings from other investigators or with other methods or in other laboratories were not routinely performed. In some cases, the questionable Ab-positive cases could not be confirmed externally (e.g., for patient 1 in Table 7), while other findings (e.g., for patient 9 in Table 7) were only slightly positive. However, this is precisely the situation encountered by clinicians in their everyday lives. In the psychiatric setting, weakly positive Ab findings or Ab titers below the current detection threshold of the standard assays could also be relevant. For example, a possibly long-lasting but milder antibody effect on the brain, occurring via processes such as synaptic reconstruction, could lead to subtler psychiatric phenotypes. For this reason, we have openly described all findings, including questionable results and those from external laboratories/follow-up tests, and the additional findings for these patients are summarized in Table 7. Some patients showed constellations of an AP/AE (e.g., case 2 in Table 7), whereas several other cases had assessments that remained more nebulous (e.g., in case 5 in Table 7). In our department, the use of tissue tests was not fully established until the end of 2018. Nevertheless, even now, this very laborious examination remains reserved for selected cases with high suspicion of AP 9,39,40 . The open design, the broad inclusion criteria (e.g., not excluding patients with different comorbidities), and the fact that a tertiary referral center would obviously attract patients for organic differential diagnosis could have led to a distortion of the results. Similarly, the use of an uncontrolled design precluded estimation of the prevalence of CSF alterations and positive tissue tests in healthy individuals. However, comparative values are available from neurological control groups. For example, an increased AQ in patients with retrobulbar neuritis was    In the reference laboratory in Oxford (using live CBAs) positive, not tested using biochip-assays initially. Additionally slightly positive for anti-SOX1 antibodies in serum.
c In an additional measurement externally serum titer was elevated up to 1:160 using CBAs (reference <1:20 In the laboratory in Oxford (using radioimmunoassays) testing for anti-VGKC antibodies was negative.
g Serum testing using Ravo blot® was positive and CSF was negative, in the Euroimmun immunblot® anti-Yo reactivity was found in serum and CSF, in addition, a weak anti-Ma2-reactivity was found in the serum.
h The concentration was measured by radioimmunoassays.
i Confirmatory analysis externally using live CBAs positive with titer determination. External testing was performed unsystematically in Laboratory Krone (Bad Salzuflen; Germany) or Laboratory Stöcker (Lübeck, Germany).
found in only 3.8% 46 , whereas the authors detected significantly higher percentages (18%) in similarly aged patients with schizophreniform syndromes. Three subgroups of the patients presented here have already been described in the previous studies 14,23,24 . Multimodal prospective screening studies combining all available methods are desirable in future, especially since the sensitivity of different antineuronal Ab test methods differs significantly 47 . These variations in sensitivity may also explain the low Ab prevalence observed in the present study. Unlike some previous large studies analyzing the serum antineuronal Ab prevalence 11,12 , the present study was focused only on IgG Abs that are clearly associated with an AE 48 .

Conclusions
CSF findings often revealed a dysfunction of the BBB and, less frequently, signs of neuroinflammation. Established high-level antineuronal Abs in serum were rare, and they occurred even less frequently in the CSF. However, several serum-only Ab-positive patients showed evidence of brain involvement in instrument-based clinical studies. Surprisingly, the use of screening tissue tests frequently detected pathologies in pre-selected patients. Novel antineuronal Abs with so far unknown antigens could, therefore, play a decisive role in psychiatry. Further multimodal, prospective, and controlled studies are necessary.