Cerebrospinal fluid concentration of complement component 4A is increased in first-episode schizophrenia

Excessive synapse loss is a core feature of schizophrenia and is linked to the complement component 4A gene (C4A). In two independent cohorts, we show that cerebrospinal fluid (CSF) C4A concentration is elevated in first-episode psychosis patients who develop schizophrenia and correlates with CSF measurements of synapse density. Using patient-derived cellular modeling, we find that disease-associated cytokines increase neuronal C4A expression and that IL-1beta associates with C4A in patient-derived CSF.

expression has been confirmed in SCZ postmortem brain tissue 1 . During brain development, 48 microglia utilize complement signaling for selective removal of supernumerary synapses by 49 complement receptor 3 (C3R)-dependent phagocytosis 2,3 . In line with the observed decrease 50 in synapse density in SCZ 4,5 , excessive microglial synapse elimination has been observed in 51 SCZ-derived in vitro models 6 . This suggests that the increased SCZ risk in microglial synapse 52 removal in early stages of the disease. C4A CNs also correlate with complement deposition in 53 patient-derived models, as well as microglial engulfment of synaptic structures 6 . Notably, 54 C4B CNs, not linked to SCZ risk 1 , do not influence neuronal complement deposition or 55 synapse elimination in these models 6 . Recently, these in vitro findings were confirmed in vivo 56 using mouse models overexpressing human C4A or C4B 7 . 57 Despite the accumulating evidence from experimental models linking C4A to 58 excessive synapse removal in SCZ, data showing increased in vivo protein levels in patients 59 has been lacking. One reason is that these measurements have been limited by difficulties in 60 distinguishing C4A and C4B as the peptide sequence only differ by a few amino acids. In the 61 present study, we developed a targeted mass spectrometry method capable of detecting unique 62 peptide sequences in C4A and C4B protein. We applied this method to human cerebrospinal 63 fluid (CSF) collected from two independent cohorts of first-episode psychosis (FEP) patients 64 and healthy controls (HCs) (Figure 1a). 65 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 24, 2021. ; https://doi.org/10.1101/2021.08.17.21262131 doi: medRxiv preprint In the first cohort (KaSP; discovery sample) we included 44 FEP patients and 20 age-66 and sex-matched HCs. FEP patients were stratified depending on if they subsequently 67 developed SCZ (FEP-SCZ; 29 subjects) or not (FEP-nSCZ; 15 subjects). The groups 68 displayed no significant differences in terms of demographics and clinical characteristics 69 Table 1). We observed significantly higher CSF C4A concentrations in 70 FEP-SCZ patients as compared to HCs or FEP-nSCZ patients (Figure 1b). On the contrary, 71 CSF C4B concentrations were similar across the three groups (Figure 1c ). Fourteen FEP-72 SCZ patients and 8 FEP-nSCZ patients had been prescribed an antipsychotic, although in no 73 instance for more than one month before CSF collection. At the group level, concentrations of 74 both C4A and C4B were similar between patients who had been exposed to an antipsychotic 75 and those being antipsychotic-naïve, and FEP-SCZ patients displayed higher C4A 76 concentrations also after adjusting for the use of antipsychotic medication (Supplementary 77

Figure 1). 78
Results were validated in an independent FEP cohort (GRIP; replication sample), 79 comprising of 31 FEP patients (FEP-SCZ; n=17, FEP-nSCZ; n=14) together with 21 HCs. 80 FEP-nSCZ patients were more commonly smokers than HCs but otherwise the three groups 81 did not differ in terms of demographics and clinical characteristics (Supplementary Table 2). 82 However, smoking was unrelated to CSF C4A or C4B concentration (rpearson(p)=-0.16; 83 P=0.253, and r(p)=-0.23; P=0.102, respectively). In accordance with our results from the 84 discovery cohort, we observed significantly higher CSF C4A concentrations in the FEP-SCZ 85 group as compared to HCs or the FEP-nSCZ group (Figure 1d), while CSF C4B 86 concentrations were similar across groups (Figure 1e). Twenty-two patients (9 FEP-SCZ and 87 13 FEP-nSCZ) were prescribed an antipsychotic, with FEP-SCZ patients still displaying 88 higher CSF C4A concentrations compared to FEP-nSCZ patients after adjusting for use of 89 antipsychotic medication (Supplementary Figure 2). 90 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint  is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 24, 2021. ; https://doi.org/10.1101/2021.08.17.21262131 doi: medRxiv preprint variable effect on C4A versus C4B expression 12,13 . The pro-inflammatory cytokines, 116 interleukin(IL)-1 and IL-6 both have repeatedly been shown to be elevated in CSF of SCZ 117 patients 14,15 , and we therefore explored their influence on neuronal C4A expression. To this 118 end, we generated stably inducible neurogenin 2 (NGN2) neural progenitor cells from an SCZ 119 patient-derived induced pluripotent stem cell line, with equal CNs of C4A and C4B, and 120 differentiated these cells to cortical excitatory neurons. These cultures were stimulated for 24 121 hours with either IL-1 (1ng/ml), IL-6 (1ng/ml), or a combination of IL-1 and IL-6 ( Figure  122 2a). We observed significantly increased expression of C4A as measured by qPCR when IL-123 1 and IL-6 were combined (Figure 2b), while no such effect was evident for expression of 124 C4B (Figure 2c), thus suggesting that the induction is predominately influencing C4A 125 expression. To corroborate this observation in a clinical context and to determine the relative 126 importance of each cytokine in clinically relevant concentrations, we measured CSF IL-1 127 and IL-6 in 25 FEP patients with available CSF (KaSP cohort). Controlling for genetically 128 predicted C4A RNA expression using available whole genome sequencing data 1 , we observed 129 a significant positive correlation between IL-1 and CSF C4A concentration (Figure 2e), 130 while the correlation between IL-6 and CSF C4A concentration was less pronounced and did 131 not reach significance (Figure 2f). In contrast, there was no correlation between these two 132 cytokines and CSF C4B concentrations (Figure 2g and 2h). Collectively, this suggests that in 133 FEP patients primarily elevated IL-1 levels could contribute to excessive synaptic pruning 134 by specifically inducing C4A expression per C4A CN. 135 Lastly, as C4A CNs predict synaptic complement deposition and synaptic pruning in 136 experimental models 6 , we explored associations between CSF C4A protein concentrations 137 and an in vivo proxy of synapse density. CSF levels of neuronal pentraxins (1, 2, and 138 receptor) were recently shown to inversely correlate with prefrontal cortical thickness 139 determined with MRI, as well as to predict synapse loss in dementia and Alzheimer's 140 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 24, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 24, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 24, 2021. ; https://doi.org/10.1101/2021.08.17.21262131 doi: medRxiv preprint at the tertiary unit (i.e., study baseline) was 2 years. This time period, as well as time to SCZ 208 diagnosis within the study period, was unrelated to CSF C4A or C4B levels, as well as to a 209 SCZ-vs. a non-SCZ-related diagnosis (data not shown). Patients who did not receive a SCZ 210 diagnosis at the diagnosis evaluation around 2 years after baseline received a diagnosis of 211 bipolar disorder, brief psychotic disorder, substance-induced psychotic disorder (not known at 212 the initial evaluation), or a psychotic disorder not otherwise specified. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 24, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 24, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 24, 2021. ; https://doi.org/10.1101/2021.08.17.21262131 doi: medRxiv preprint peptides) of each trypsin digested CSF sample, was injected and run in triplicates. Pooled CSF 283 samples were run as quality control samples throughout all the samples. 284 The linear range of each heavy peptide was measured in pooled CSF digest. A mixture 285 of heavy AQUA peptides was spiked at five different concentrations into pooled CSF 286 samples. The concentration range was adjusted for each individual peptide according to its 287 expected endogenous signal. Each sample was analyzed three times, and the peak areas of 288 heavy peptides were plotted against the spiked heavy peptides concentrations 289 (Supplementary Figure 4). 290 All the raw data generated on the Fusion MS were imported to Skyline v4.1 (MacCoss 291 Lab Software, USA) for data analysis. Peak integration was done automatically by the software 292 and was manually inspected to confirm correct peak detection. Peak identities were confirmed 293 by measured transitions (dotp) and also between endogenous and corresponding heavy peptides 294 (rdotp). The best 3 to 5 transitions were selected for the quantification and was performed by 295 matching light and heavy peak area ratios. All calculations were performed using Microsoft 296

Excel. 297
For quality control, we also measured a peptide shared for C4A and C4B (tC4; GRIP). 298 The sum of C4A + C4B peptide levels in the combined sample strongly correlated to tC4 299 (rs=0.91; P<1x10 -15 ). One subject with C4A CNs also displayed non-detectable levels of the 300 C4A peptide. We decided to exclude all subjects not displaying detectable peptide levels in the 301 main analyses. However, analyses were also performed including zero values, as well as 302 imputing zero values with half of the limit of detection, but with very similar results (data not 303 shown). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 24, 2021. ; https://doi.org/10.1101/2021.08.17.21262131 doi: medRxiv preprint

Molecular analysis of C4 structural elements (ddPCR) 308
CNs of C4 structural elements (C4A, C4B, C4-HERV CNs) in the GRIP cohort were measured 309 using ddPCR as previously described 6  is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint  is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 24, 2021. ; https://doi.org/10.1101/2021.08.17.21262131 doi: medRxiv preprint after stimulus onset. The difference of the most positive peak and most negative in a 20-150 358 ms window after pulse onset was used to calculate the peak response amplitude. The 359 following startle measures were examined; 1) reactivity, or the magnitude of response (startle 360 amplitude (SA), 2) PPI or the percentage of change in startle magnitude to prepulse + pulse 361 versus pulse-alone trials (((pulse − prepulse + pulse) / pulse) * 100) 19  Reprogramming, Inc. (www.cellular-reprogramming.com), as previously described 372 previously 6 . Briefly, induced pluripotent stem cell (iPSC) colonies were obtained using 373 mRNA reprogramming in a feeder-free culture system. Stable iPSCs were expanded in 374 NutriStem XF medium (Biological Industries) and on biolaminin 521 LN-coated (BioLamina) 375 plates to at least passage 3. iPSCs were then purified using MACS with anti-TRA-1-60 376 MicroBeads (Miltenyi Biotec) on LS columns according to the manufacturer's instructions. 377 All fibroblasts and iPSCs were screened and found negative for Mycoplasma; they stained 378 positive for octamer-binding transcription factor 4 (POU domain, class 5, transcription factor 379 1) and TRA-1-60 (Figure 2a). 380 We then generated NGN2 expressing stable NPC lines using TALEN-based plasmids 381 as previously described 6 . Briefly, the doxycycline-inducible NGN2 AAVS1 knock-in plasmid, 382 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 24, 2021. ; https://doi.org/10.1101/2021.08.17.21262131 doi: medRxiv preprint 9.
Hedberg, M., Imbeault, S., Erhardt, S. & Schwieler, L. Disrupted sensorimotor gating 523 in first-episode psychosis patients is not affected by short-term antipsychotic treatment. 524 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 24, 2021. ; https://doi.org/10.1101/2021.08.17.21262131 doi: medRxiv preprint