Computational modelling of EEG and fMRI paradigms reveals a consistent loss of pyramidal cell synaptic gain in schizophrenia

Diminished synaptic gain - the sensitivity of postsynaptic responses to neural inputs - may be a fundamental synaptic pathology in schizophrenia. Evidence for this is indirect, however. Furthermore, it is unclear whether pyramidal cells or interneurons (or both) are affected, or how these deficits relate to symptoms. Participants with schizophrenia (Scz, n=108), their relatives (n=57), and controls (n=107) underwent three electroencephalography paradigms - resting, mismatch negativity, and 40 Hz auditory steady-state response - and resting functional magnetic resonance imaging. Dynamic causal modelling was used to quantify synaptic connectivity in cortical microcircuits. Across all four paradigms, characteristic Scz data features were best explained by models with greater self-inhibition (decreased synaptic gain), in pyramidal cells. Furthermore, disinhibition in auditory areas predicted abnormal auditory perception (and positive symptoms) in Scz, in three paradigms. Thus, psychotic symptoms of Scz may result from a downregulation of inhibitory interneurons that may compensate for diminished postsynaptic gain in pyramidal cells.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.
. CC-BY 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is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint . CC-BY 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 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 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 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 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 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 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 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 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 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 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 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 Reduced excitatory synaptic gain (i.e. decreased slope of the presynaptic input-postsynaptic 19 response relationship) is believed to be a primary deficit in schizophrenia 1,2 . This reduction 20 may primarily affect pyramidal cells 1 or inhibitory interneurons 3 . Decreased interneuron 21 function in the disorder may thus be primary or a compensatory response to try to rebalance 22 excitatory and inhibitory transmission in cortical circuits 4 . These hypotheses are difficult to 23 assess in vivo, however. 24 25 Various mechanisms may reduce synaptic gain in schizophrenia: the most important is 26 probably hypofunction of N-methyl-D-aspartate receptors (NMDARs) and their postsynaptic 27 signalling cascade 1,2 . Evidence for this comes from psychiatric genetics 5 , magnetic resonance 28 spectroscopy (MRS) imaging 6 , neuropathological studies 7 , animal models 8 , etc, but of these, 29 only MRS is performed in humans in vivo, and its glutamatergic measures are difficult to 30 interpret. Other neuromodulatory dysfunctions in schizophrenia (e.g. reduced cortical 31 dopamine 9 or muscarinic receptors 10 ) can be assessed more directly using positron emission 32 tomography (PET)), but neither MRS nor PET assess synaptic gain directly. 33 . CC-BY 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 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint          This schematic illustrates the key steps in the preprocessing of the EEG (resting state, 2 mismatch negativity and 40 Hz auditory steady-state response) and resting state fMRI 3 paradigms, and their subsequent analysis using dynamic causal modelling (DCM) and 4 parametric empirical Bayes (PEB). Simplified depictions of the paradigms are shown in the 5 first column (see Online Methods for details), with group differences in EEG data features in 6 the second column (first three rows), and DCM in the third column. The EEG data Con vs 7 Scz group differences are (from first to third rows) in rsEEG θ, β, and γ frequency band 8 power (Figure 2A 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 January 8, 2021. ; https://doi.org/10.1101/2021.01.07.21249389 doi: medRxiv preprint

2
In what follows, we first describe conventional analyses of group differences in data features 3 for each paradigm. We then report the best explanation for these differences in terms of DCM 4 parameters. Figure 1 summarises the analysis (excluding results), and Table S1 describes the 5 participants. We used the DCM canonical microcircuit model (see below) to analyse the EEG 6 paradigms. For the MMN and 40 Hz ASSR paradigms, we analysed group differences using 7 conventional data features (event related potentials or power spectra), from which subject-8 specific DCM parameters were estimated. For rsfMRI, we modelled the network generating 9 the MMN (and 40 Hz ASSR, in part). We used PEB (see Online Methods) to analyse group 10 and individual differences in synaptic (model) parameters, with the exception of rsEEG, 11 where characteristic group responses were modelled. We interpret greater 'self-inhibition' of 12 pyramidal cells as an effective loss of pyramidal synaptic gain. Given known 13 pathophysiology in Scz, NMDAR hypofunction seems the most likely explanation for loss of 14 pyramidal gain, but other explanations are possible, for example an upregulation of (e.g., 15 parvalbumin positive) interneuron function (see Online Methods for further discussion). 16 17 Age, sex, smoking and chlorpromazine dose equivalent covariates did not significantly affect 18 the results, unless otherwise stated. All t-tests were two-tailed, ranksum tests were used if 19 distributions were skewed; none are Bonferroni-corrected unless stated. 20

21
In rsEEG, Scz have altered power in θ, β and γ frequency bands 22 We first examined rsEEG power spectra by subtracting the 1/f gradient, noting that gradients 23 did not differ between groups with eyes open or closed (P>0.2). The mean adjusted power 24 spectra within the Con (n=98) and Scz (n=95) groups are shown in Figure  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 January 8, 2021. All (except self-inhibition)

Figure 3 -MMN data and modelling analysis 1
A -Mismatch difference waves (i.e. deviant-standard, mean ±s.e.m.) for Con (n=94; blue), 2 Scz (n=96; red) and Rel (n=42; green) at electrode Fz. Group differences are computed using 3 t-tests (uncorrected) at each timepoint and are marked with red (Scz vs Con) and green (Rel 4 vs Con) bars above the difference waves. There were no significant Scz vs Rel differences. D -Microcircuit models were compared, differing only in which parameters were allowed to 14 change from their priors (estimated G connectivity parameters are shown, as in Figure 2C). 15 These models' free G parameters included various combinations of superficial (sp) and/or 16 deep (dp) pyramidal cell (blue) connections to or from inhibitory interneurons (ii, red), and 17 self-inhibition of sp and ii cells. Note that each parameter -within each microcircuit -could 18 differ between subjects but was constrained to be the same in every cortical area within 19 subjects, except for sp self-inhibition which could differ throughout. The final model also 20 estimated delay D and time constant T parameters (these were fixed in the other five models).  Figure 3C). Likewise, Scz had a reduced P100 1 response (peak at 82 ms, P(FWE)=0.003, t(376)=4.83), as did Rel (peak at 94 ms, P(unc)=0.001, 2 t(268)=3.02; Figure S2B). 3 4 DCM of MMN indicates increased frontal self-inhibition in Scz, and group-specific 5 relationships of self-inhibition with abnormal auditory percepts and cognitive 6 performance 7 We first used model comparison to establish whether it was best to fix or estimate various 8 microcircuit parameters in the MMN analysis (see Online Methods). We compared six 9 models ( Figure 3D): Model 6G estimates six connectivity (G) parameters, Models 4Ga-d 10 consider subsets of these six, and Model 6G,D,T also estimates delays and time constants. 11 Bayesian model selection preferred Model 6G (also in Con and Scz separately), with a 12 protected exceedance probability of P=0.89 ( Figure 3E, left). This model fitted most 13 participants' data accurately (e.g. Figure S3A): a histogram of R 2 values is shown in Figure  14 3E (right) -the group mean R 2 was 0.73. R 2 were slightly higher in Con (mean=0.76 15 ±std=0.13) than in Scz (0.70±0.14; ranksum Z=3.12, P=0.0018) and Rel (0.71±0.15; ranksum 16 Z=2.14, P=0.033) ( Figure S3C). 17

18
We then used PEB to ask which parameters best explained group differences in the MMN: 19 self-inhibition within areas or connections between areas. The reduced mismatch effect in 20 Scz was best explained by increased self-inhibition in deviant -relative to standard -trials in 21 L IFG (P>0.95) and R IFG (P>0.99; Figure 3F). Including chlorpromazine dose equivalent 22 covariates reduced the posterior probability to P>0.75, but age, sex and smoking had no 23 effect. Conversely, there was no overall group effect (across both standards and deviants) of 24 Scz on the microcircuit parameters (all P<0.95; Figure S4C We next considered induced responses during auditory steady-state stimulation. Group-7 averaged 40 Hz ASSR are shown in Figure 4A, and the distributions of participants' peak γ 8 (35-45 Hz) frequencies in Figure 4B. Scz had slightly reduced γ peak frequency: mean peak 9 frequencies (following subtraction of the 1/f gradient: Figure S2D priors for these parameters showed the Full model was superior (Figure 4E, left). The 40 Hz 27 thalamic drive was modelled using a Gaussian bump function of width w≤4 Hz (see Online 28 Methods): this width performed better than a narrower bump of 1 Hz (Model -w, Figure 4E). 29 Model fits for the winning model were reasonable ( Figure S3B; mean R 2 =0.53). Group 30 differences in R 2 were not detected ( Figure S3C self-inhibition (i.e. increased synaptic gain), and lower dp-ii connectivity. 30 All effects shown in F, G, H and I are also present without the addition of age, sex and 31 smoking covariates (P>0.95), and also with inclusion of chlorpromazine dose equivalents as 32 a covariate. 33 34 was because there were marked differences between Rel and Con parameters, only some of 1 which were shared by Scz ( Figure S5B). The 'genetic risk' effect was expressed as an 2 increased conduction delay in L A1 (P>0.95; Figure 4F), and reduced superficial pyramidal 3 (sp) to inhibitory interneuron (ii) connectivity (P>0.99; Figure 4G, left). The psychosis 4 'diagnosis' effect was associated with increased superficial pyramidal self-inhibition in 5 bilateral A1 in Scz (both P>0.99; Figure 4G, right). 6 7 40 Hz ASSR DCM links abnormal auditory percepts to A1 disinhibition in Scz and 8 cognitive performance to reduced A1 self-inhibition in all participants 9 In Scz, the auditory perceptual abnormalities 'trait' measure related to a disinhibited sp-ii-sp 10 circuit, i.e. increased sp-ii (P>0.99) and reduced ii-sp connectivity (P>0.99), as well as 11 greater self-inhibition in L A1 (P>0.99; Figure 4H). These associations were also seen in the 12 auditory 'state' measure but at lower posterior probability (P>0.95 for sp-ii, P>0.75 for ii-sp 13 and sp-sp, not shown). Across all participants, Digit Symbol score was associated with 14 reduced pyramidal self-inhibition in bilateral A1 (both P>0.95) and reduced deep pyramidal 15 to inhibitory interneuron (dp-ii) connectivity throughout (P>0.99; Figure 4I). Self-inhibition 16 and dp-ii effects were also present in both Con and Scz groups when analysed separately (the 17 former in L A1 in Con, and R A1 in Scz; both P>0.95, not shown). model used to explain fMRI data is simpler than the neural mass models used for EEG; 25 however, they retain inhibitory self-connections. Model fits were accurate: R 2 s were >0.7 in 26 all groups, with no group differences ( Figure S3C, ranksum tests: all P>0.05). 27 28 PEB analysis of subject-specific DCM parameters suggested that Scz is associated with 29 increased self-inhibition in L and R IFG (P>0.99 and P>0.95 respectively; Figure 5A). These 30 effects were robust to age, sex, and smoking covariates (and to the removal of the 10 31 participants with the lowest rsfMRI signal to noise ratio: 8 Scz and 2 Con; both P>0.95). 32 These effects did not survive addition of chlorpromazine dose equivalents (L IFG self-33 inhibition fell to P>0.75). However, Rel > Con showed the same increase in self-inhibition in A -For comparative purposes, the rsfMRI connectivity analysis was conducted on the same 2 network as the MMN analysis. Results for Con (n=85) and Scz (n=72) are shown in the same 3 format as Figure 3F. As in the MMN, Scz showed increased self-inhibition in bilateral IFG. 4 Inclusion of chlorpromazine equivalent dose as a covariate still showed increased self-5 inhibition in L IFG but only at P>0.75. 6 B -rsfMRI connectivity analysis without covariates for Con (n=85) and Rel (n=45) is shown. 7 Like Scz, Rel show increased self-inhibition in bilateral IFG, but this effect disappeared with 8 addition of the age covariate (P<0.75).   Figure 5B). This group difference did not survive addition of the 1 age covariate: Rel were older than Con (Rel mean age=45.4 ±16.6 years, Con mean age=39.4 2 ±14.3 years; t(162)=2.4, P=0.02). These differences were not detected using conventional 3 functional connectivity analyses (that cannot assess self-inhibition) or analyses of regional 4 variance (see Figures S6B to S6E and Online Methods for further discussion). 5 6 rsfMRI DCM reveals relationships of positive symptoms to cortical disinhibition in Scz 7 and of gain to cognition in Con 8 PEB analysis within Scz found that the auditory perceptual abnormalities 'trait' measure was 9 associated with increased self-inhibition in L and R IFG (both P>0.99, Figure 5C We repeated the rsfMRI analyses without global signal regression (Figures S7A-E). Many 30 results changed, with much greater between subject variability in parameters (the Scz > Con 31 contrast, Figure S7A) or the loss of significant effects ( Figures S7B, D and E). 32 33

Self-inhibition findings across EEG and rsfMRI paradigms are similar but individual 1 parameter fitting must be improved for use as a model-based biomarker 2
In summary, we found clear evidence for increased self-inhibition (evidence of reduced 3 synaptic gain) in Scz ( Figure 6A) in all data modalities and paradigms. However, 4 disinhibition within auditory areas was associated with auditory perceptual abnormalities 5 within Scz ( Figure 6B). Furthermore, reduced self-inhibition was associated with Digit 6 Symbol performance, especially in Con ( Figure 6C). Note that associations with independent 7 data speak to the predictive validity of the synaptic biomarkers furnished by DCM. A 8 sensitivity analysis (see Online Methods) confirmed that increased superficial pyramidal self-9 inhibition best reproduced the key data features of the MMN (i.e. decreased MMN amplitude 10 but unchanged latency; Figure S8A) and -along with loss of sp-ii connectivity -the 11 decreased 40 Hz ASSR ( Figure S8B). 12 13 We also assessed whether the synaptic differences in Scz might be driven by early or late 14 illness stages. We did not have access to illness duration data, but instead repeated the Scz > 15 Con group comparisons for each paradigm, after dividing the groups equally into younger 16 (≤36 years) and older (≥37 years) participants. The group differences in rsEEG features, 17 MMN parameters and 40 Hz ASSR parameters were driven by the older group ( Figure S9A-18 C); the rsfMRI parameter differences by the younger group ( Figure S9D). 19 20 Finally, in post hoc analyses, we asked whether self-inhibition parameters -that were altered 21 in Scz -had predictive validity across paradigms, thus licensing their use as 'model-based 22 biomarkers'. We used PEB to assess relationships in Scz between self-inhibition parameters 23 in IFG in the MMN, or in A1 in the 40 Hz ASSR, and their corresponding self-inhibition 24 parameters in the rsfMRI, in participants whose EEG and rsfMRI were <100 days apart 25 (MMN n=44, 40Hz ASSR n=40), as some participants underwent EEG and rsfMRI many 26 months or even years apart. Within Scz, there was evidence of associations between self-27 inhibition in R IFG in the MMN and rsfMRI (P>0.95; Figure S10A) and in R A1 in the 40 28 Hz ASSR and rsfMRI (P>0.95; Figure S10B). However, there were no such relationships in 29 the left hemisphere or in Con.  The microcircuit consists of superficial and deep pyramidal cells (sp and dp, blue), inhibitory 5 interneurons (ii, red), and spiny stellate cells (ss, green), interconnected with excitatory 6 (arrowheads) and inhibitory (beads) connections. replicated and all could be explained by increased self-inhibition in (superficial) pyramidal 8 cells. Likewise, Scz also showed an increase in prefrontal self-inhibition -as in the  in rsfMRI ( Figure 6A). Given the pathophysiology of Scz, this most likely reflects 10 diminished synaptic gain on pyramidal cells 28 (discussed below). 11 12 Second, abnormal auditory percepts in Scz was associated with decreased self-inhibition in 13 auditory areas selectively, across three paradigms ( Figure 6B). This is consistent with 40 Hz 14 Third, processing speed (i.e. Digit Symbol score) in Con was associated with disinhibition 24 across three paradigms ( Figure 6C). This plausibly reflects a relationship between cognitive 25 performance and synaptic gain, and complements a recent rsfMRI DCM study of healthy 26 adults (n=602) that found fluid intelligence was related to decreased self-inhibition in dorsal 27 attention and salience networks 32 . In Scz, self-inhibition and cognition were only associated -28 in the 40 Hz ASSR -in R A1. Interestingly, a rsEEG DCM study in Down syndrome (n=36) 29 found intelligence was associated with self-inhibition in V1 33 . Why is unclear: self-inhibition 30 may be estimated more efficiently in sensory regions. 31 Regarding the rsEEG, increased θ power in Scz rsEEG is a well-established finding 27,34 , as is 1 a 'U-shaped' change in spectral power (i.e. increased θ, decreased β, increased γ), although 2 this pattern has been seen across θ, α and β frequencies 27 . Surprisingly, older Scz drove the 3 increased γ effect: low γ (30-45 Hz) power is typically reduced in Scz with longstanding 4 diagnoses 35 . Age is an imperfect proxy for illness duration, however. which was not observed in our data. We assumed time constants did not differ in Scz in the 22 ASSR or MMN, and estimated connectivity parameters -and delays, in the ASSR -instead 23 (these can be regarded as synaptic rate constants). 24 25 A previous rsfMRI DCM analysis in Scz found disinhibition in anterior cingulate cortex 25 , 26 rather than increased self-inhibition in bilateral IFG ( Figure 5A). This combination speaks to 27 a pattern of altered intra-prefrontal functional connectivity in early Scz 40 : namely, increased 28 connectivity of medial areas and more modest decreases in connectivity in lateral areas. 29 Prefrontal hyperconnectivity correlated positively with positive symptoms 40 . We similarly 30 found positive symptoms were associated with disinhibition in bilateral IFG, and also A1 31 ( Figure 5D). This relationship echoes findings that increased functional connectivity of 32 primary sensory areas (to thalamus) correlates with PANSS scores 41 , and that increased A1 33 rsfMRI autocorrelation (a result of reduced self-inhibition) in Scz relates to auditory 34 hallucinations 42 (c.f. Figure 5C, right). R IFG self-inhibition's relationship to Digit Symbol 1 score (in Con; Figure 5E) mirrors the finding that the global functional connectivity of a 2 nearby region of left lateral prefrontal cortex relates to fluid intelligence 43 . Taken together, 3 these results may indicate that some functional connectivity relationships to either symptoms 4 or cognition may not depend on extrinsic connections between nodes, but on synaptic gain 5 within nodes. 6 7 Although global signal regression removes physiological 'noise', the global signal has greater 8 power in Scz, which may have a neuronal component 44 . Here, GSR strengthened existing but 9 weak effects in both Scz and Con ( Figures 5C-E and S7C-E), and converted reduced forward 10 connectivity into IFG (amidst a large amount of between-subject parameter variability, 11 Figure S7A) to reduced synaptic gain within IFG. Given this, and given GSR in Scz affords a 12 relatively uniform transform of the data 44 , the results with GSR are probably more reliable, 13 especially given their consilience with the EEG results ( Figure 6). 14 15 The Rel group showed mixed effects across paradigms, and more data are required to draw 16 firm conclusions. In the MMN, no effects exceeded P>0.95 despite Rel's similar data 17 features to Scz, perhaps because the Rel group was smaller. In the 40 Hz ASSR, pyramidal 18 self-inhibition was reduced in Rel ( Figure S5B), not increased (as in Scz). In the rsfMRI 19 however, Rel showed comparable IFG self-inhibition increases relative to Scz ( Figure 5B). 20 Interestingly, Rel's pattern of apparently normal (or decreased) self-inhibition in the EEG 21 paradigms -yet increased self-inhibition in rsfMRI -was recapitulated in the younger Scz 22 Regarding potential causes of reduced synaptic gain, some Scz data features imply NMDAR 33 hypofunction. In rsEEG, increased γ follows NMDAR antagonism 46 , e.g. using ketamine 34 (which also suppresses β) 47 or in NMDAR encephalitis (which also increases θ) 17,48 . In 1 contrast, LSD and psilocybin do not increase θ 49 , and dopamine 2 antagonists potentiate α 2 and β 50,51 . The 40 Hz ASSR is sensitive to NMDAR function 20 (but also cholinergic 52 , 3 dopaminergic 53 and serotonergic 54 manipulations): the latter do not affect the MMN, 4 however, which is quite specific to NMDAR function 11 . Ketamine also reduces rsfMRI 5 functional connectivity of IFG and auditory cortices 55 . Antipsychotic dose covariates 6 weakened the Scz MMN condition-specific effects ( Figure 3F) but strengthened the overall 7 MMN effects ( Figure S4C); they also weakened the Scz rsfMRI effects, but similar rsfMRI 8 effects emerged in unmedicated Rel ( Figure 5). Overall, these findings resemble NMDAR 9 hypofunction, and seem unlikely to be medication-induced. 10 11 Several limitations are addressable: given pathophysiology is dynamic in Scz 1 , and that 12 subgroups may exist 56 , larger datasets should be analysed, containing more early course (and 13 preferably unmedicated) Scz. Furthermore, we must understand why younger Scz drive 14 rsfMRI self-inhibition effects yet older Scz drive EEG self-inhibition effects. DCM models 15 with explicitly parameterised NMDA (and other) receptor conductances 17 can explore 'self-16 inhibition' in more detail. Our analyses were restricted to a few cortical regions, to maximise 17 cross-paradigm effects, but future work should broaden this focus, especially to other 18 frontotemporal regions. 19 20 Another key limitation involved the efficiency of EEG model inversion. We found 21 preliminary evidence for a relationship between self-inhibition parameters across EEG and 22 fMRI paradigms in Scz ( Figure S10), but the clustering of EEG estimates around their prior 23 values ( Figure S10C) must be finessed (by using more informative data or paradigm design) 24 to assess cross-paradigm self-inhibition reliability in all participants. If parameters show 25 cross-paradigm validity, hierarchical PEB can be used to estimate them from multiple 26 paradigms simultaneously 21 . 27 28 In conclusion, we found that dynamic causal modelling of multimodal neuroimaging data in 29 Scz produced some remarkably consistent results across several paradigms. These include 30 increased self-inhibition (i.e. diminished synaptic gain) in Scz, especially in frontal areas, but 31 disinhibition -in auditory areas in particular -correlating with psychotic symptoms and 32 auditory perceptual abnormalities specifically. Psychotic symptoms may therefore be caused 33 by the interneuron downregulation that is necessary to restore cortical 'excitation/inhibition 34 balance' in Scz. With appropriate model-fitting procedures, and larger datasets, model-based 1 biomarkers for psychotic disorders may be at hand. April 2016. The study participants comprised those with a diagnosis of schizophrenia (Scz, 10 n=107), controls who were approximately matched for age, gender and smoking status 11 (n=108), and first degree relatives of Scz (n=57): details of the groups are listed in Table S1. 12 Diagnoses were made using the Structured Clinical Interview for DSM Diagnoses (SCID) 1 . 13 Scz were recruited from outpatient clinics, relatives from Scz participants themselves and 14 also from media advertisements, and controls from media advertisements. Exclusion criteria 15 were: neurological illness, head injury, substance abuse or dependence (except for nicotine), 16 and daily use of benzodiazepine drugs. Controls and relatives were excluded if they met 17 criteria for an Axis I diagnosis using the SCID, and controls were required to have no family 18 history of psychosis within two generations. 19 20 All participants completed the Digit Symbol Substitution Task 2 -a symbol-copying task that 21 assesses learning, executive function and processing speed -and the Auditory Perceptual 22 Trait and State scale (APTS) 3 . The APTS is a self-rated scale that assesses the frequency of 23 abnormal auditory percepts, from altered characteristics of sounds to illusions, 24 pseudohallucinations and verbal and non-verbal hallucinations. The scale includes 'trait' and 25 'state' measures, defined as symptoms experienced over one's lifetime until two weeks ago, 26 and over the past week, respectively. The full scale is available at 27 http://www.mdbrain.org/APTS.pdf. Its test-retest reliability was assessed in 41 participants 28 about 4 months apart, which showed ICC = 0.81 for both the trait and state measures, 29 suggesting good reliability. 30 31 Use of psychotropic medication was recorded. 90 Scz were taking one or more antipsychotics 32 (including Clozapine; see Table S1), for 15 Scz no medication was recorded and 3 were unmedicated. Some participants were also taking antidepressants, hypnotics (as prescribed) 1 or mood stabilisers -the Scz group had higher proportions in each case. The 20 item Brief 2 Psychiatric Rating Scale (BPRS) was used to rate overall symptoms in the Scz group only. 3 4 Data acquisition and paradigms 5 6 For EEG recording, participants sat in a semi-reclining chair inside a sound-attenuated 7 chamber, wearing an electrode cap. 64 scalp electrode sites were recorded, according to the 8 10/20 International System, using Neuroscan Stim acquisition software and a Synamps2 or 9 Synamps2 RT amplifier. Recordings were grounded midway between FPZ and FZ using 10 silver/silver-chloride electrodes and referenced to the nose. Eye movements were monitored 11 by vertical and horizontal electrooculograms (EOGs). Data were continuously sampled at 12 1000 Hz with a DC/100 Hz band-pass filter (24 dB/octave roll-off). Impedances were kept 13 below 5 kΩ. 14 15 For the resting state EEG (rsEEG) paradigm, subjects were asked to remain awake whilst 16 EEG data were recorded. Two recordings were obtained on the same day, one with eyes 17 open, and one with eyes closed. Each recording lasted 5 minutes. 18

19
In the mismatch negativity (MMN) paradigm, participants were presented with 1000 auditory 20 stimuli through earphones, of which 800 (80%) were standard tones presented at 75 dB, for 21 60 msec, at 1000 Hz; 200 (20%) were duration-deviant tones at 75 dB, for 150 msec, at 1000 22 Hz. All tones had a 5 msec rise/fall time, with a stimulus onset asynchrony of 300 msec. 800 23 tones were presented in a single block. Participants were asked to ignore the tones while 24 viewing a silent movie. Some of the MMN data have been analysed elsewhere 4 , using 25 classical statistical tests and structural equation modeling to find correlations between MMN 26 responses and neurotransmitter levels. Here we obtained effective connectivity estimates 27 using DCM. 28

29
In the auditory steady state response (ASSR) paradigm, participants listened to click trains 30 delivered by headphones, at 2.5, 5, 10, 20, 40, and 80 Hz. 75 stimulus trains (trials) -each 31 consisting of 15 clicks, with each click at 72 dB and of 1 ms duration -were delivered at 32 each stimulus frequency. The duration ranged from 6 s per train for 2.5 Hz, to 0.1875 s per train for 80 Hz. The inter-train interval was 0.7 s. Therefore, the durations for 2.5, 5, 10, 20, 1 40, and 80 Hz were 8.38, 4.69, 2.82, 1.89, 1.42, and 1.19 min, respectively, presented in 6 2 separate blocks separated by 2 min. The order of the blocks was randomized. In this study, 3 only the data for the 40 Hz ASSR were analysed: data from this and the other frequency 4 bands have been analysed elsewhere 3 . 5 6 For the resting state fMRI (rsfMRI) paradigm, participants were asked to keep their eyes 7 closed and relax. fMRI data were collected at the University of Maryland Center for Brain MMN and 40 Hz ASSR data were imported into EEGLAB 5 (https://sccn.ucsd.edu/eeglab/) 23 and band-pass filtered between 1 and 70 Hz, with a notch filter at 59.5-60.5 Hz. rsEEG data 24 were band-pass filtered between 1 and 50 Hz as higher γ frequencies are more vulnerable to 25 artefacts in non-timelocked data. The data were then epoched: rsEEG into 6 s epochs, MMN 26 into epochs from -50 ms to 300 ms around the stimulus onset, and 40 Hz ASSR data into 27 epochs from -100 ms to 600 ms around the stimulus onset. The data then underwent an 28 automated artefact rejection pipeline within EEGLAB, also employing the Multiple Artefact 29 Rejection Algorithm (MARA) 6 . First, epochs with amplitudes of >±5 std were rejected, as 30 were epochs containing linear trends (using pop_rejtrend, max slope=5, min R2=0.7) and low 31 (0-2 Hz) or high (20-40 Hz) frequency bands (using pop_rejspec) within power thresholds of 32 -50 to 50 dB and -100 to 25 dB respectively (40 Hz ASSR data used a frequency band of 20-35 Hz given the click train was presented at 40 Hz). Channels were then rejected (using 1 pop_rejchan) on the basis of extreme values in their spectrum (-4 to 6 std), kurtosis (-7 to 15 2 std) or joint probability (-9 to 7 std). If >50% epochs were rejected, channel rejection was 3 performed first, and epoch rejection second. If >50% epochs were still rejected, or if >20% 4 channels were rejected, the data was discarded. Independent component analysis (ICA) was 5 then performed, and components with p>0.5 of being artefacts -determined using stored 6 artefact templates within MARA -were then rejected. MARA contains a supervised machine 7 learning algorithm that learns from expert ratings of 1290 components by extracting features 8 from the spectral, spatial and temporal domains, in order to identify artefacts of all kinds (e.g. 9 eye, muscle, loose electrodes, etc). After MARA, channel and epoch rejection were run a 10 second time using lower thresholds, as sometimes artefacts became apparent following 11 component rejection. Channels were rejected on the basis of the following values in their 12 spectrum (-6 to 5 std), kurtosis (-6 to 9 std) or joint probability (-7 to 7 std). Missing channels 13 were then interpolated and the data were rereferenced to the common average. The data were 14 then converted to SPM format and imported into SPM12 (v7219) 7  (http://surfer.nmr.mgh.harvard.edu/) was then used to segment gray and white matter in the 30 whole brain to produce individual cortical-subcortical anatomical segmentations 10 -31 specifically the pial and white matter cortical boundaries which were used to define a "cortical 32 ribbon", and a subcortical grey matter voxel mask. These were then used to define the hybrid surface/volume neural file for each subject in the Connectivity Informatics Technology 1 Initiative (CIFTI) "grayordinate" space 11 . 2 3 BOLD images were first slice time corrected using FSL slicetimer. Next, motion correction 4 was performed by aligning all BOLD data to the first frame of every run via McFLIRT. Motion 5 corrected BOLD images were then registered to T1w image using FreeSurfer BBR. Finally, all 6 the linear transformation matrices and nonlinear warp images were combined and the original 7 slice time corrected BOLD images were registered to MNI-152 brain template in a single step. 8 To exclude signal from non-brain tissue a brainmask was applied and cortical BOLD data were 9 converted to the CIFTI gray matter matrix by sampling from the anatomically-defined cortical 10 ribbon. These cortical data were then aligned to the HCP atlas using surface-based nonlinear 11 deformation 11 . Subcortical voxels in the BOLD data were extracted using the Freesurfer-12 defined segmentation to isolate the subcortical volume portion of the CIFTI space. 13

14
In addition to the HCP MPP, all BOLD images had to pass stringent quality assurance criteria 15 as previously reported 12 to ensure that all functional data were of comparable and high quality: 16 i) signal-to-noise ratios (SNR) >90, computed by obtaining the mean signal and standard 17 deviation (sd) for a given slice across the BOLD run, while excluding all non-brain voxels 18 across all frames 13 ; ii) movement scrubbing as recommended by Power et al 14,15 . 'Movement 19 scrubbing' refers to the practice of removing BOLD volumes that have been flagged for high 20 motion, in order to minimize movement artefacts, and is a widely used fMRI preprocessing 21 technique. Specifically, to further remove head motion artefacts, as accomplished previously 12 , 22 all image frames with possible movement-induced artefactual fluctuations in intensity were 23 identified via two criteria: First, frames in which sum of the displacement across all 6 rigid 24 body movement correction parameters exceeded 0.5 mm (assuming 50 mm cortical sphere 25 radius) were identified; second, root mean square (RMS) of differences in intensity between 26 the current and preceding frame was computed across all voxels divided by mean intensity and 27 normalized to time series median. Frames in which normalized RMS exceeded 1.6 times the 28 median across scans were identified. The frames flagged by either criterion were marked for 29 exclusion (logical or), as well as the one preceding and two frames following the flagged frame. 30 Movement scrubbing was performed for all reported analyses across all subjects. Subjects with 31 more than 50% of frames flagged were removed from subsequent analyses. Next, to remove 32 spurious signal in resting state data we completed additional preprocessing steps, as is standard 33 practice 16 : all BOLD time-series underwent high pass temporal filtering (>0.008 Hz), removal of nuisance signal extracted from anatomically-defined ventricles, white matter, and the 1 remaining brain voxels (i.e. global signal) (all identified via participant-specific FreeSurfer 2 segmentations 17 ), as well as 6 rigid-body motion correction parameters, and their first 3 derivatives using previously validated in-house Matlab tools 18 . Note that while the removal of 4 global signal is a debated topic in fMRI 19,20 , it remains the gold standard for removing spatially 5 pervasive artefact in the brain (although other techniques are emerging 21,22 ). BOLD signal 6 within the subject-specific cortical mask was spatially smoothed with a 6 mm full-width-at- population's membrane potential back into a firing rate using a sigmoid operator whose gain 32 or slope is controlled by parameter S. This firing rate becomes the input for another 33 population, determined by the network's intrinsic and extrinsic connectivity. Delays in transmission within the microcircuit (intrinsic) and from one brain area to another (extrinsic) 1 are parameterised by D. The microcircuit itself ( Figure S1A) contains four neural 2 populations: spiny stellate cells (ss), superficial pyramidal cells (sp), deep pyramidal cells 3 (dp) and inhibitory interneurons (ii). Extrinsic connections follow known anatomical 4 patterns 28 : 'forward' connections project from sp cells to ss and dp cells in the area above, 5 and 'backward' connections project from dp cells to sp and ii cells in the area below. Given 6 all of the above, the dynamic activity of all populations in the network can be computed for 7 any given input (for evoked responses like the MMN) or as a filtered spectral response to 8 endogenous neuronal fluctuations (a mixture of pink and white neuronal noise) as in the 9 rsEEG 25 . 10

11
Most of the analyses in this paper concern group differences in G parameters, i.e. 12 connectivity between neural microcircuit populations (e.g. sp to ii) or self-inhibition (e.g. of 13 sp to sp). Self-inhibitory connections may parameterise one or more of several physiological 14 effects: i) they control the responsiveness of a population to its inputs, as any mechanism 15 controlling synaptic gain does, e.g. NMDAR function, but also classical neuromodulatory 16 receptors such as cholinergic and dopaminergic receptors; ii) they may reflect the action of an 17 inhibitory interneuron population in a circuit, e.g. from sp cells projecting to parvalbumin+ 18 (PV + ) interneurons and back to sp cells 29 ; iii) in the case of interneurons, they may reflect 19 autapses (self-synapses: common on PV + cells 30 ). A model cannot distinguish between these 20 mechanisms unless they are explicitly encoded in the model itself; however, some 21 interpretations are more or less likely given what we know of pathology in schizophrenia. For 22 example, a finding of increased sp self-inhibition in Scz is more likely to be due to loss of 23 synaptic gain in those cells than a strengthening of the sp-PV + -sp circuit, which most 24 evidence implies is weakened in Scz 31 . The opposite is true for a finding of decreased sp self-25 inhibition in Scz: its most likely explanation would be a loss of ii inhibition of sp cells. 26

27
Similarly, the microcircuit model does not explicitly distinguish between PV + and 28 somatostatin+ (SST + ) interneurons. SST + interneurons are of interest as their cortical markers 29 are just as (if not more) reduced in Scz as PV +32 . That said, the dynamics of the sp self-30 inhibition connection in the model are faster than those of the sp-ii-sp circuit (see the 31 parameters in Figure S1A), hence the model can potentially distinguish between faster (likely 32 PV + ) and slower (e.g. SST + ) interactions between pyramidal cells and interneurons 29 . (Note however that the empirical priors used for the 40 Hz ASSR analysis accelerated the dynamics 1 of the sp-ii-sp circuit, in order to model the 40 Hz peak.) 2 3 The microcircuit model used was the same as that used by Shaw et al 33 : we termed this 4 spm_fx_cmc_2017. It makes a small adjustment to spm_fx_cmc_2014, in that it replaces a G 5 connection from sp to ss with a connection from sp to dp: see Figure 2C and Figure S1A. We 6 used this model because it is closer to known anatomy 28 than previous versions, but not too 7 complicated to fit to EEG data (e.g. more sophisticated models 34 contain separate interneuron 8 pools for sp and dp cells). One key aspect of the model is that it models cortical dynamics 9 only, and hence can model the γ and β peaks generated by superficial and deep cortical layers 10 (respectively) 35-37 , but it cannot reproduce an α peak without adding an additional (thalamic) 11 input 38,39 . Given the absence of group differences in α in the rsEEG, and for simplicity, we 12 did not model this frequency band (as in previous studies 33,40 ). 13

14
To simulate the power spectra in controls and Scz, we used the SPM12 (v7219) function 15 spm_induced_optimise, which computes transfer functions (i.e. representations of cortical 16 dynamics in the spectral domain) for parameters in the biophysical models in DCM. The 17 simulated spectra were normalised in exactly the same way as the empirical spectra, 18 subtracting the 1/f gradient using robustfit in Matlab. The standard priors in DCM were used 19 ( Figure S1A). 20

21
We did not try to fit the microcircuit model to the empirical rsEEG data (although this is 22 possible 40 ) because the choice of sources in rsEEG data can be problematic. We instead used 23 the microcircuit model to simulate the effects of various potential circuit pathologies in Scz, 24 and compared the results to the pattern in the rsEEG. The circuit pathologies were based on 25 reasonable hypotheses about microcircuit connectivity abnormalities in Scz ( Figure 2D): 26 Model 1 -a <30% loss in connectivity in all microcircuit connections, i.e. a global loss of 27 synaptic efficacy. 28 Model 2 -a <30% loss of connectivity to/from interneurons ('to' and 'from' are identical 29 from the modelling point of view -their effects are the same). 30 Model 3 -a <30% loss of interneuron self-inhibition, to model potential disinhibition of PV+ 31 interneurons by other PV+ interneurons 30 . 32 Model 4 -a <30% gain in interneuron self-inhibition, to model loss of synaptic gain (e.g. 33 hypofunction of NMDARs) on interneurons.
Model 5 -a <30% gain in superficial pyramidal self-inhibition, to model loss of synaptic 1 gain (e.g. hypofunction of NMDARs) on sp cells. 2 3 Data analysis and modelling: mismatch negativity 4 5 The MMN data were plotted (using electrode Fz, as is standard practice 41 ) as group-averaged 6 waveforms for the standard and deviant tones separately ( Figure S2A) and as the traditional 7 'difference' waves, i.e. deviant-standard waves, illustrating the mismatch effect ( Figure 3A). 8 Group differences in these waveforms were assessed using t-tests at each timepoint using an 9 α of p<0.05 (uncorrected for multiple comparisons), and are displayed on the figures. 10 11 To more robustly illustrate the mismatch effects and group differences, incorporating other 12 electrodes and cluster-based correction for multiple comparisons, the MMN sensor-level data 13 were analysed in SPM12, after smoothing using an 8x8x8 mm FWHM Gaussian kernel 14 ( Figures 3B, 3C and S2B). Each of these models used the same macroscopic structure ( Figure 2C, right), i.e. forward 26 and backward connections linking areas in adjacent hierarchical levels, and lateral 27 connections linking bilateral areas at the same level (not shown). Note that only forward, 28 backward and self-inhibitory connections could show condition-specific differences between 29 groups, i.e. differences between standard and oddball event related potentials. 30 31 MMN model fitting was performed using the 'spatial' or 'IMG' forward model in DCM for 32 evoked response potentials (ERPs). DCM fits the first up to eight modes of the prior 33 predicted covariance in sensor space 45 ( Figure S3A). In practice, the first 3-4 modes usually capture most of the variance (as in Figure S3A, right), so R 2 values generated for the MMN 1 data are based only on the first four modes. A boundary elements head model 46 was used to 2 approximate the brain, cerebrospinal fluid, skull and scalp surfaces. 3 4 Fitting non-linear neural mass models to EEG data is an ill-posed problem, and can lead to 5 difficulties in optimisation, such as models getting stuck in local optima. Empirical Bayes for 6 DCM 47 can circumvent this issue by performing model fitting iteratively, each time using the 7 group level posteriors over parameters as priors for each subject's parameters in the next 8 model inversion, yielding more robust and efficient parameter estimates 48 . In practice, it 9 substantially improved model fit: model inversions performed with and without it had R 2 10 values of ~0.7 and ~0.6 respectively. Local optima can also be avoided by fitting one fully 11 parameterised model and then pruning unnecessary parameters using Bayesian model 12 reduction (see below), rather than fitting many models with reduced numbers of parameters 13 which are more prone to local optima problems 49 . 14 15 Following model inversion, models were formally compared using Bayesian model 16 selection 50 ( Figure 3E). We used the protected exceedance probability 51 as the metric of 17 success: i.e. the probability that any one model is more frequent than the others, above and 18 beyond chance. The R 2 for the first four modes of the winning model was also computed 19 ( Figure 3E). Group differences between Con and Scz or Rel in their R 2 values were also 20 assessed using ranksum tests (as the distributions were skewed; Figure S3C). 21 22 Group differences in parameters and relationships between parameters and other measures 23 were analysed using Parametric Empirical Bayes (PEB; spm_dcm_peb) 49,52 . PEB is a 24 hierarchical Bayesian version of a general linear model that can assess which DCM 25 parameters differ between groups or covary with a continuous measure. One critical 26 advantage of PEB (over performing classical statistical tests on parameters) is that it takes 27 account of not just each parameter's expectation but also its covariance: therefore parameters 28 that cannot be estimated with high confidence (e.g. in models that fit less well) contribute less 29 to the inference. 30 31 A second key aspect of the PEB procedure is the use of Bayesian model reduction 53 and 32 Bayesian model averaging 49 (spm_dcm_peb_bmc). In essence, these steps prune away 33 unnecessary parameters and then obtain posterior estimates for the remaining parameters by averaging over the remaining models. The Bayesian model reduction procedure allows 1 hypotheses about model parameters to be formally tested. This is done by first defining a 2 model space of different parameter groupings that may best describe the modelled effect. In 3 the MMN analysis, two model spaces were used: the first asked which combinations of 4 extrinsic or self-inhibitory connections could best account for the mismatch effect ( Figure  5 S4A, left), and the second asked which combinations of intrinsic (microcircuit) connections 6 could best account for the overall group differences across conditions ( Figure S4B, left). 7 These model spaces were based on the assumption that microcircuit parameters would not 8 differ between conditions, but messages passed between areas (or pyramidal gain in specific 9 areas) may well do so. In Figure S4A  Note that if there is no evidence for some models, parameters unique to those models will be 21 redundant: these parameters are eliminated following Bayesian model reduction. Conversely,22 parameters that are present in all probable models are highly likely to explain the group 23 difference effect. The posterior probabilities of parameters differing between groups are 24 shown on the right ( Figure S4A): only six parameters have non-zero probabilities of 25 explaining the group difference effect, but only two of these are >0.95. These probabilities 26 generate the Bayesian confidence intervals for the group difference plots ( Figure 3F, in this 27 case). 28 29 Figure S4B (left) shows the model space for the group differences in microcircuit parameters. 30 The first six rows correspond to sp-sp self-inhibitory connections in areas from L A1 to R 31 IFG; the next five rows containing white boxes correspond to connections constrained to be 32 identical throughout the model. The The timecourses of the 40 Hz ASSR waveforms in electrode Fz are plotted in Figure 4A: ~40 6 Hz oscillations are superimposed on more sustained baseline changes (known to occur during 7 auditory click trains 54 ). Scz and Rel response baselines diverge (t-tests at each timepoint) 8 from Con around 250 ms, but we restricted our modelling to well-replicated group 9 differences in ~40 Hz power. Each subjects' EEG data were transformed into measurements 10 of phase and power in the frequency domain using a Morlet wavelet in spm_eeg_tf. For the 11 Fz sensor-level analyses, the spectra were normalised in order to assess the power at 40 Hz 12 relative to the background 1/f slope. For normalisation, we again subtracted the 1/f gradient 13 in log space (computed using robustfit in Matlab applied to the 10-30 Hz and 50-70 Hz 14 ranges, on a subject-specific basis): example gradients of group averaged data from electrode 15 Fz are shown in Figure S2D. γ peak frequency was the frequency in which the maximum 16 (normalised) power within the 35-45 Hz range occurred (plotted in Figure 4B). The 17 unnormalised group averaged time-frequency plots are shown in Figure S2E, and the 18 normalised plots in Figure 4C. Permutation testing assessed whether the number of observed 19 significant t-tests (P<0.05) for the Con v Scz, Con v Rel and Scz v Rel comparisons across 20 all timepoints in the 30-55 Hz range shown was likely due to chance. There were no group 21 differences in normalised power at other frequencies (t-tests at each timepoint and frequency, 22 not shown). 23

24
The 40 Hz ASSR paradigm is known to activate primary auditory cortex (Heschl's gyrus), 25 anterior to auditory evoked response sources 55,56 . To obtain good priors for source 26 localisation, we performed source localisation on all subjects in SPM12 using a minimum 27 norm approach and focussing purely on the 37-43 Hz frequency band. Across all subjects, 28 this generated three peaks of activation, either in or near right Heschl's gyrus and anterior to 29 the MMN sources. The closest to Heschl's gyrus (A1) was [50 -12 4] ( Figure 4D), which was 30 used (along with [-50 -12 4]) as the seed for reconstruction of virtual electrode data from a 31 broader band (35-50 Hz) for modelling in DCM. 32 Scz > Con analysis is shown in Figure S6A  but all likely models contained bilateral IFG self-inhibitory connections, hence these 20 connections are inferred to be highly likely to explain group differences (right). 21

22
For comparative purposes, in addition to the spectral DCM analysis, we performed a standard 23 functional connectivity analysis ( Figure S6B), i.e. computing the Pearson (zero lag) 24 correlation between two parcellations' timeseries. We also computed their (zero lag) 25 covariance ( Figure S6C), because comparing correlations in Con and Scz can be problematic 26 as their variances may differ 57 , and the corresponding zero lag covariance in the neuronal 27 states estimated by DCM ( Figure S6D). To see whether spectral DCM self-inhibition 28 parameters might be recapitulated in basic data features, we also assessed group differences 29 in the standard deviation of BOLD signal in each node ( Figure S6E), given that greater self-30 inhibition would be expected to relate to lower variance in a node 60 . 31 32 It is clear from Figures 5A and S6B-E that one cannot reliably associate effective 33 connectivity with functional connectivity or BOLD fluctuations. Reduced functional connectivity between A1 and STG in Scz ( Figure S6B and elsewhere 61 ) is less robust if 1 covariance is used instead (an issue highlighted previously 57 ). Conversely, increased IFG 2 self-inhibition in Scz ( Figure 5A) is not apparent in the BOLD standard deviations in IFG 3 ( Figure S6E), although reduced BOLD fluctuations in Scz have been reported in other 4 areas 60 . Measures of functional connectivity differ from effective connectivity in several 5 important ways: i) spectral DCM also estimates haemodynamic and neuronal/measurement 6 noise parameters -which may contribute to group differences in functional connectivity e.g. 7 if one group is older 62 or taking (antipsychotic) medication, or if one group moves more 8 (respectively); ii) spectral DCM connectivity includes non-zero lags, which indicate direction 9 of connectivity 58,63 ; iii) and -most importantly -spectral DCM also estimates self-inhibition 10 within cortical regions. 11 12 We also performed the same DCM and PEB analyses without global signal regression, to 13 ascertain the effects of this preprocessing step on the results ( Figure S7). 14 15 Data analysis and modelling: parameter sensitivity and cross-paradigm analysis

17
We performed parameter sensitivity analyses on the six estimated microcircuit (G) 18 parameters to address two potential weaknesses in the MMN and 40 Hz ASSR analyses. In 19 the MMN analysis, constraining condition-specific (mismatch) effects to only sp-sp self-20 inhibition amongst the G parameters -along with extrinsic connections -might miss such 21 effects in other G parameters. In the 40 Hz ASSR analysis, the data were afforded high 22 precision during model fitting in order to force the model to fit the unnatural 40 Hz peak: this 23 may lead to overfitting of other data features, however, and spurious results. We therefore 24 selected two subjects from each paradigm with EEG responses typical of the group average. 25 We simulated virtual electrode ('LFP') data from a single cortical area using 26 spm_induced_optimise and either spm_gen_erp (for the MMN) or spm_csd_mtf (for the 40 27 Hz ASSR). In each case we used these subjects' posterior G parameter estimates, varying 28 each G parameter by ±30% in turn; in the MMN, we also fixed delays (D) and time constants 29 (T) to the values used in Figure S1A. The simulated data are shown in Figure S8, with the 30 key data features that the parameters showing group differences ought to explain circled in 31 red. 32 We sought to investigate whether Scz with shorter or longer illness durations were driving 1 the key Scz > Con group difference effects in each paradigm. We thus repeated the Scz > 2 Con analyses from each paradigm, splitting each group by median age, as we did not have 3 illness duration data. Group differences in rsEEG power spectra in subjects ≤36 years old 4 (Con n=50, Scz n=49) and in subjects ≥37 years old (Con n=48, Scz n=46) are shown in 5 Figure S9A; group differences in MMN mismatch parameter effects in subjects ≤36 years old 6 (Con n=45, Scz n=47) and subjects ≥37 years old (Con n=48, Scz n=48) in Figure S9B; 7 group differences in 40 Hz ASSR parameters in subjects ≤36 years old (Con n=47, Scz n=48) 8 and in subjects ≥37 years old (Con n=46, Scz n=46) in Figure S9C; and group differences in 9 rsfMRI parameters in subjects ≤36 years old (Con n=44, Scz n=34) and in subjects ≥37 years 10 old (Con n=42, Scz n=37) in Figure S9D. 11 12 We used the PEB framework to assess whether the abnormal parameters in the Scz group 13 related to each other across modalities, i.e. across EEG and fMRI. If so, this would support 14 the use of multiple paradigms in combination to estimate these 'biomarker' parameters. We 15 selected the left and right IFG self-inhibition parameters from the MMN and left and right A1 16 self-inhibition parameters from the 40 Hz ASSR, and their counterparts from the rsfMRI. For 17 each of the four rsfMRI parameters separately, we used the PEB framework to reveal which 18 of all MMN or 40 Hz ASSR parameters related to it. Thus the posterior expectation and 19 covariance in the EEG parameters, but only the expected value of the rsfMRI parameter, was 20 used by the PEB model. (We reasoned that the rsfMRI parameters were likely estimated with 21 greater confidence, given their simpler models and more consistently high R 2 values 22 compared with the EEG models ( Figure S3C), hence it would be better to include the EEG 23 parameters' covariance in the PEB model.) Two of the four analyses yielded significant 24 results: R IFG self-inhibition across rsfMRI and MMN ( Figure S10A), and R A1 self-25 inhibition across rsfMRI and 40 Hz ASSR ( Figure S10B). 26 27 For comparative purposes, we also performed classical analyses of the relationships between 28 the same rsfMRI and EEG parameters. The Pearson correlations between the rsfMRI and 40 29 Hz ASSR or MMN self-inhibition parameters are shown in Figure S10C