Disrupted-in-schizophrenia-1 is required for normal pyramidal cell–interneuron communication and assembly dynamics in the prefrontal cortex

Article activity feed

1. 

   Version published to 10.7554/elife.79471 on eLife

   Oct 14, 2022
2. 

   eLife

   Aug 12, 2022

   Author Response

   Reviewer #1 (Public Review):

   The authors of the paper provide new evidence of how prefrontal cortex of mutant mice used as a disease model of schizophrenia differs from wild type littermates. By analyzing local network dynamics at the level of specific cell type, authors shed new light on the circuit mechanisms that underlie changes in network dynamics in these mice.

   The claims in the submitted manuscript are supported by the data. I have a few comments and questions that need to be clarified.

   We thank the reviewer for highlighting the novelty of our work and its relevance (…shed new light on the circuit mechanisms that underlie changes in network dynamics in these mice…) for the field and the validity of our data (….claims in the submitted manuscript are supported by the data).

   1. Average firing rates

   Authors …

   Author Response

   Reviewer #1 (Public Review):

   The authors of the paper provide new evidence of how prefrontal cortex of mutant mice used as a disease model of schizophrenia differs from wild type littermates. By analyzing local network dynamics at the level of specific cell type, authors shed new light on the circuit mechanisms that underlie changes in network dynamics in these mice.

   The claims in the submitted manuscript are supported by the data. I have a few comments and questions that need to be clarified.

   We thank the reviewer for highlighting the novelty of our work and its relevance (…shed new light on the circuit mechanisms that underlie changes in network dynamics in these mice…) for the field and the validity of our data (….claims in the submitted manuscript are supported by the data).

   1. Average firing rates

   Authors claim that they saw a significant reduction in interneuron firing rates in Disc1 mutant mice compared to control mice Fig.1c. However, the difference could be general and not interneuron specific. Due to the high firing rates of interneurons, the statistical test will work better on interneurons than on pyramidal cells as pyramidal cells average firing rates are lower. What I suggest to do is to take interneuron cells that fire at a lower rate (lower 33% for example ) and compare the control and Disc1 groups. Also I would suggest to take pyramidal cells that have higher firing rates (upper 33% for example) and compare firing rates across the same groups. One would like to see if these differences are not due to changes in firing rates per se.

   We thank the reviewer for pointing out this important aspect. In our original analysis, we did not take into account that additional differences in the PYR population might be present but ‘masked’ by the overall lower firing rate of that neuronal population. As suggested by the reviewer, we separately considered the firing rate of the ‘top 33%” of the PYR population, which did not significantly differ between genotypes (p=0.958, n=209 control and 245 Disc1 PYRs, Welch’s test). As suggested, we moreover considered the ‘bottom 33%’ of INT firing rates, for which the significantly lower rates of Disc1-mutant INTs remained (control: 4.2 ± 0.6 Hz vs. Disc1: 1.8 ± 0.2, n=26 and 34 neurons, p=0.013, Mann-Whitney U-test). Since only few INTs were recorded per session in some cases (ranges: Disc1: 2-12/session; control: 2-19/session), we performed this analysis on the basis of individual cells (see also our reassessment of the main statistical comparisons in response to #1 by reviewer 2 and #4 by reviewer 3). These data are now reported in the new Fig. 1 – figure supplement 3 and referred to in line 76 ff. (line 72 ff. without tracked changes) of the revised manuscript.

   1. Optogenetic tagging

   Authors indicate that light triggered and spontaneous spike waveform are similar Fig.1d. This is nice, but would be better to see all the tagged neurons. I would suggest showing all optically tagged neurons spike features. Authors can impose with a different color spike features of tagged neurons in Fig.1a. I suspect that since all PVI are narrow spiking and they must fall into the area of blue colored cells in Fig.1a.

   Following the reviewers suggestions, we included the average waveforms with and without light for all opto-tagged PVIs in the revised Fig. 1f. Moreover, we included the kinetic features of opto-tagged PVIs in Fig. 1a (red dots), and separately for control and Disc1-mutant mice in the new Figure 1-figure supplement 2. As predicted by the reviewer, the PVIs indeed cluster with the other putative INTs. We would moreover like to point to our new analysis in response to #2 of reviewer 2 addressing the spike kinetics of optotagged PVIs versus untagged putative INTs, which are similar in their trough-to-peak duration and asymmetry index. These data are shown in the novel Fig. 1 – figure supplement 2.

   1. It was not clear why authors assessed only firing rates in last 25ms (line 348-349). If they have a clear justification for this they should provide it. But why not use the latency of the first spike also as an additional metric. A well tagged cell will respond to light pulse with short latency (within 5 ms). My concern is that non PVI cells may increase firing rate after 25ms of stimulation of PVI cells due to disinhibition.

   Despite the latency to the first spike being frequently used as a method to detect ChR2-positive neurons, the laser stimulation produced significant photoartefacts in our hands. We were therefore concerned that spikes that happen shortly after the onset of the light pulse might be missed, and hence the latency to the first spike might be misinterpreted. Selecting a later time point in the stimulation interval allowed us to assess the firing rate during light application without the interference by artefacts. Nevertheless, we fully agree with the reviewer’s concern that ChR2-negative non-PVIs might increase their rate due to disinhibition, and that these neurons might thus be falsely classified as PVIs. However, we are confident that that is not the case. First, optotagged PVIs cluster well within the population of electrophysiologically identified INTs (see our response to your first remark on ‘optogenetic tagging’) and were indistinguishable from this population in terms of spike kinetics (see our response to #2 of reviewer 2 and the new Fig. 1 – figure supplement 2), suggesting that no disinhibited PYRs were included in the optotagged sample of cells. Second, we performed an additional analysis to address the time course of firing rate changes in optotagged PVIs. We computed smoothed spike trains (convolved with a 5 ms SD Gaussian kernel), and extracted the average firing rate of each optogenetically identified PVI centered on the onset of the light pulses. This analysis revealed a rapid increase in firing rate upon light delivery, arguing against disinhibitory network effects. These new data are now shown in the new Fig. 1 – figure supplement 5 and reported in line 89 (85 without tracked changes) of the revised manuscript.

   1. Spike cross-correlations

   The authors show that spike transmission probability from PYR to PVI is reduced in Disc1 mice compared to the controls Fig.2d and Fig.2e, but what happens to PVI to PYR spike transmission probability? Is it different in those groups? Answering this question is important since the authors discuss this topic in line 185-193.

   Inhibitory synaptic interactions are indeed detectable by spike-train cross-correlation. However, we find these to be harder to quantitatively interpret than excitatory connections. Those interactions are not visible as spike transmission but rather as a reduction in spike transmission. Reliable estimates of the reduction in spike rate of postsynaptic PYRs require very large spike numbers of postsynaptic neurons that need to be sampled. For instance, Senzai et al., 2019 (Neuron 101: 500-513.e5) identified inhibitory interactions in continuous recordings lasting up to 68 h. Since we did not explicitly design our experiments to investigate inhibitory interactions, our recordings were substantially shorter than the required length. Using the method of Senzai et al., 2019 to identify inhibitory interactions, we detected only 5 INT-INT interactions (in the pooled Disc1-mutant and control data set). This low number does not allow the quantification of potentially reduced spike transmission. Thus, attempts to quantify inhibitory interactions properly would require a substantial amount of additional long-duration recordings. While the point raised by the reviewer is highly relevant and should be investigated in future, we think that given the extensive amount of experimentation needed to address this question, it is beyond the scope of the current manuscript.

   1. Authors could try to link oscillations with spike transmission probabilities. On line 180 authors discuss that lower synchrony between PVI might be responsible for observed reduction in gamma power in Disc1 mutant mice. With the available data authors could test this hypothesis. They can look at spike cross correlations in their pool of INT and PVI (if they have pairs of PVI recorded in the same session) population.

   We thank the reviewer for this excellent suggestion! We computed the cross-correlations for all simultaneously recorded putative INTs and quantified the baseline-subtracted mean cross-correlation within 10 ms around zero time lag. This analysis revealed weaker cross-correlation in Disc1-mutant mice (p=0.026, Mann-Whitney U test, tested on averages from n=7 control and Disc1 mice with at least 2 INTs recorded simultaneously), suggestive of reduced synchronization of putative INTs at short time lags. These new data are now included in the new Fig. 4 and reported in line 201 ff. (185 ff. without tracked changes) of the revised manuscript.

   1. An alternative way to link oscillations with lower spike transmission probabilities in PYR-PVI pairs is to use synchrony triggered LFP analysis. One could take all time points when PVI and PYR cells fired acausal spikes within 2ms window and look at the LFP around this time point. Than take the average of the synchrony-triggered LFP and look at the power spectrum.

   The proposal to link spike transmission with LFP power is indeed intriguing. As suggested by the reviewer, we extracted the 60-90 Hz-filtered LFPs triggered by INT spikes that followed a spike in a presynaptic PYR by <2 ms and measured the average gamma amplitude in a window of 20 ms around the INT spike. This analysis revealed comparable gamma amplitudes in Disc1 compared to control pairs. This finding suggests that local PYR-INT loops are still capable to produce gamma oscillations, and that the gamma oscillation defect of Disc1 mice is likely not caused by such a local defect. To investigate the relationship between INT spike timing and gamma oscillations more generally, we further extracted gamma amplitudes of spike-triggered LFPs using all available spikes of the INTs. Moreover, we compared the data to gamma amplitudes measured at randomly selected time points. ANOVA analysis followed by Tukey tests performed on the level of mouse averages indicated that while INT spiking-associated gamma amplitudes were significantly larger than those depicted from random time points in wild type mice (p=0.001). However, the same was not true for Disc1-mutant mice (p=0.591). Furthermore, this analysis revealed significantly reduced spike-triggered high gamma amplitudes in Disc1-mutant compared to control mice (p=0.011). While these results argue against a driving role of local connection alterations in gamma defects, they generally confirm the impaired synchrony of INT spiking relative to gamma oscillation that we observed in our analysis of phase coupling. These data are now shown in the new Fig. 4, which summarizes all new analyses regarding gamma oscillations and phase-coupling, and in figure 4 – figure supplement 2. The new results are described in the main text of the revised manuscript in line 188 ff. (172 ff. without tracked changes).

   Considering the reduced short time scale synchronization of INTs (see our new results towards the reviewer’s #5) and reduced gamma amplitude of INT spike-triggered LFPs, it is possible that impaired synchronization among prefrontal INTs might contribute to the observed reduction in gamma power of Disc1-mutant mice (thereby, essentially, reflecting impaired INT gamma (ING)). Additionally, reduced long-range excitatory drive maintaining local gamma oscillations might be a contributing factor. For example, recent work showed that high gamma oscillations in the mPFC occur synchronized with gamma oscillations in the olfactory bulb (Karalis & Sirota, 2022, Nat Commun 13:467). It remains to be investigated whether local INTs are rhythmically driven by input from the olfactory bulb (in a multi-synaptic pathway including olfactory cortex) and to what extent that drive maintaining afferent gamma might be altered in Disc1-mutant mice. While the current data set does not allow a systematic evaluation of these possibilities, they should be further explored in future experiments.

   1. Cell assembly analysis

   The authors used 10ms for testing synchronization among pairs of PYR neurons in Fig.4a but 25ms for analysis of assembly dynamics. I think the authors justified why they used 25ms bin size, but it was not clear why they used 10ms? Could the authors clarify the reasons behind this decision?

   The synchronization analysis was originally applied to PYRs converging on a common postsynaptic INT. English et al. (Neuron 95:505-520, 2017) systematically tested the effect of presynaptic cooperativity on spike transmission in the hippocampus (their Fig. 5). Their analysis revealed a maximum in cooperativity at ~10 ms. To maximize the sensitivity of our approach, we thus focused on 10 ms for this analysis. However, we agree that using the same time window as for assembly extraction is a reasonable proposal, in particular since we find no difference in the synchronization of identified presynaptic PYRs (Fig. 3e of the revised manuscript). Thus, we have recomputed cross-correlations using a 25 ms bin size. To further improve the analysis, we restricted it to neurons with at least 1000 spikes and simplified the quantification of excess spiking by using the ‘coinicident_spikes’ function of the Python package neuronpy.utils.spiketrain. Excess synchrony is now estimated by quantifying the number of coincident spikes between a reference and a comparison spike train detected in a 25 ms time window normalized by the firing rate expected by chance (2*frequency of comparison train * synchrony window * number of the reference train).

   By using this improved analysis with a 25 ms time window, we could replicate our original finding of enhanced synchronization of PYR spiking. However, when we averaged the data on the basis of individual mice as suggested in #1 of reviewer 2 and #4 of reviewer 3, we could not observe this effect (irrespective of whether we used the new, coincident spikes-based analysis or the original excess synchrony analysis at either 10 or 25 ms synchrony window). This result is now stated in line 215 ff. (199 ff. without tracked changes) of the revised manuscript.

   Reviewer #2 (Public Review):

   This is an interesting paper, in which the authors assessed spiking and network deficits in a well-established mouse model of schizophrenia. This mouse model carries a genetic deletion of the Disrupted-in-schizophrenia-1 (Disc1) gene, which is highly penetrant in the human condition. The authors combined behavioral analyses with state-of-the-art electrophysiological recordings in vivo, coupled to optogenetic tagging, to study a subnetwork formed by a major inhibitory neuron subclass (the parvalbumin (PV)-expressing interneuron) and principal excitatory pyramidal neurons in the medial prefrontal cortex. This work indicates reduced firing rates of PV cells in Disc1-KO mice, likely due to reduced coupling with pyramidal neurons, leading to alterations in local network activity. Indeed, the authors found that Disc-KO mice exhibited reduced levels of gamma oscillations and somewhat hypersynchronous networks.

   Taking advantage of novel techniques and analytical strategies, the manuscript provides rich, novel insight into the neurobiology of a mouse model of this severe psychiatric condition. The data is of high quality, the findings interesting and the manuscript is well written.

   Overall, the results support the authors' conclusions, although some additional analyses are necessary to corroborate their interpretations.

   Although the paper does not give information on how PV cell dysfunctions are engaged during cognitive tasks, this study can be considered as an important first step in advancing our knowledge on the basic dysfunctions of cortical networks in this model of schizophrenia

   We thank the reviewer for praising the ‘high quality’ of our work, and the ‘rich, novel insights’ on the neurobiology of a mouse model of a psychiatric disorder.

   1. The major findings stem from the analysis of the spiking activity of individual neurons recorded using either silicon probes or arrays of tetrodes. Both techniques allow simultaneous recording of many neurons from a single animal; therefore, from a statistical point of view neurons recorded from one animal are pseudo replicas and cannot be considered as independent measurements. Throughout the manuscript, the authors perform two-sample tests on the pooled data from all recorded neurons to compare differences between genotypes; therefore, artifactually increasing the power of statistical tests. Comparisons between genotypes should be performed using each mouse as an independent measurement.

   To be able to compare the data on the basis of mouse averages, we performed additional recordings, which resulted in a final data set of 9 Disc1 and 7 control mice. We recomputed the main results of this study based on mouse averages. First, consistent with our original cell-by-cell analysis, we found significantly reduced firing rates of putative INTs but not of PYRs (line 72 (69 without tracked changes)). Moreover, we confirmed our results on decreased spike transmission probability at PYR-INT connections (line 121 (107 without tracked changes)), decreased spike transmission in the resonance window (line 163 (147 without tracked changes)), reduced high gamma power (line 173 ff. (157 ff. without tracked changes)), lower phase-coupling of INT spikes to high gamma oscillations (line 178 (162 without tracked changes)), and reduced strength of assembly activations in Disc1 compared to control mice (line 229 ff. (211 ff. without tracked changes)). Similarly, we performed new analysis on INT-INT synchronization and INT spike-triggered gamma amplitudes (as requested by reviewer 1 #5 & 6), which showed significant effects on the level of mouse averages (line 188 ff. (line 172 without tracked changes)). Second, our original finding on significant differences in the synchronization of individual PYR-PYR pairs could not be reproduced on the level of individual mice. This is reported in line 215 (199 without tracked changes) of the revised manuscript. Finally, the analyses based on optogentically identified PVIs did not allow comparison by mouse averages due to the low number of experiments (n=3 mice each). Given that the vast majority of our conclusions is based on electrophysiologically identified INTs, with optogenetic identification experiments being only confirmatory in nature, and that performing additional experiments for optogentic identification of PVIs would be very laborious, we report the results of these analyses as comparisons between neurons or connected pairs. This is clearly stated at the respective sections throughout the revised manuscript. We hope that the reviewer can agree with our decision.

   1. The superficial layers of the mPFC are difficult to reach with a vertical approach of the probes due to the presence of a large blood vessel located medially in the frontal dura. Therefore, the authors are most likely reaching mPFC deep layers where PYR neurons produce fast spikes at high rates. If this is the case, this would make it difficult to sort the spiking of PYR from that of INs based on the spike kinetics and rate. The authors used opto-tagging of PVIs in a set of experiments. It would be reassuring to confirm that the spike waveform and kinetics that they extracted from PVIs are similar to those they assigned as INTs in their experiments with no opto-tagging. Identified PVIs should be statistically different from putative PYRs (not responding to light). Although opto-tagging of PVIs can solve this issue, the amount of cells isolated remains low and the number of animals is not stated. Opto-tagged cells are subsequently used for further analyses but the statistical value of those remain unclear. Since the entire interpretation of the rest of the results depend on this result, this must be clarified.

   As correctly pointed out by the reviewer, we indeed targeted deep layers of the mPFC (~0.4 mm lateral of the midline; see also the histological information about the recordings sites that is now included in Figure 1 – figure supplement 1), where higher spike rates are expected compared to superficial layers. To assess whether this might have influenced the identification of putative INTs, we separately plotted the duration and asymmetry index used to classify the neurons in PYRs and putative INTs for Disc1 and control mice. This analysis yielded well separated clusters in both cases. In addition, as suggested by the reviewer, we compared the kinetic properties (spike duration and asymmetry index) and rates of PYRs, putative INTs, and optotagged PVIs. In both genotypes, ANOVA analysis followed by Tukey post-hoc testing revealed significant differences between the PYRs and both groups of INTs, both for rate (smaller in PYRs) and kinetic properties (longer spikes of PYRs) while we found no difference between putative INTs and PVIs. These results thus suggest that the method used to identify INTs works reliably. These new data are now shown in the revised Fig. 1a and the new Figure 1 – figure supplement 2 and mentioned in line 89 ff. (85 without tracked changes) of the revised manuscript.

   We agree that the number of experiments using PVI opto-tagging is low (n=3 mice per genotype, this information is now included in the main text in line 93 ff. (88 ff. without tracked changes)). However, our analysis of spike transmission probability using the population of untagged putative fast-spiking INTs revealed similar results as for the sample of optogenetically identified PVIs. We view the PVI optotagging experiment as an additional confirmation that the difference in firing rate and spike transmission did likely not arise from sampling from different INT types in Disc1 and control mice, as pointed out in line 80 (76 without tracked changes) of the revised manuscript. The limitation of the low number of PVIs in our study is critically reflected in the revised discussion in line 249 ff. (229 without tracked changes).

   1. Proportion of gamma coupled neurons. The authors mention the use of pairwise phase consistency (PPC). PPC is a good method to measure phase coupling independent of differences in firing rates. However, it is not entirely clear how PPC is used to measure the extent of phase locking. In the methods, the authors mention that they ran the PPC analysis after determining significant phase locking with Rayleigh's test. Moreover, they provide PPC values for high gamma oscillations but not for other frequency ranges. Perhaps, it would be better to test significant coupling of all units by nonrandom spike-phase distributions crossing a confidence interval, estimated by Monte Carlo methods from independent surrogate data set. These can be obtained upon randomly jittering each spike times. Indeed, PPC values estimated by the authors for high gamma are higher for PYR than INT (Fig. 1- Fig. Suppl 4 b). This is at odds with previously published observations in V1 (e.g. Perrenoud et al., PLoS Biol. 2016 PMID: 26890123). Given the existing reports of reduced excitatory transmission in DISC-1 mice, phase locking of PYR to other frequency bands might be affected.

   Following the reviewer’s suggestion we have revised our phase-coupling analysis. First, Perrenoud et al (2016) show that gamma oscillations occur in short bursts of high power. To better reflect the coupling of putative INTs to those transient gamma events, we restricted the phase-coupling analysis to epochs within the largest quintile of gamma amplitude (assessed by the envelope of the gamma-filtered signal obtained by Hilbert transformation). Second, instead of the Rayleigh test, we obtained for each unit randomized spike trains by shuffling the inter-spike intervals (500 iterations). Significant phase locking was then obtained by testing whether two consecutive bins of the phase histogram exceeded the 95th percentile of the random distribution. This analysis was performed separately for the low (20-40 Hz) and high gamma bands (60-90 Hz) for both putative INTs and PYRs. Third, the depth of phase coupling was assessed by PPC for all significantly phase-coupled neurons. While this metric is more robust against changes in spike rates than traditional measures, it is still not completely independent of it. Perrenoud et al, for instance, showed using spike sub-sampling that the reliability in estimating PPC depends on spike rate (with >1000 spikes being optimal). However, our data set of PYRs contained fewer than 1000 spikes during high gamma events (mean Disc1: 657 ± 32, mean control: 840 ± 43). To better account for the effect of rate dependence, we restricted the analysis to neurons with >250 spikes. To further limit the potential impact of different spike counts across neurons, we used random subsampling with a fixed spike number of 250 (100 iterations per cell), computed PPC in each iteration, and averaged over the PPC estimates per cell. Finally, in response to the reviewers point 1, the results of all neurons (PYR and INT separately) were then averaged for each mouse.

   Consistent with our original analysis, we found a significantly reduced proportion of phase-coupled INTs but unaltered PPC of significantly coupled INTs to the high gamma band. Moreover, we observed no significant effects for low gamma oscillations or for the phase-coupling of PYRs to either low or high gamma bands. These results are now shown in the new Fig. 4 and the new Figure 4 – figure supplement 1, and are described in line 170 ff. (154 without tracked changes) of the revised manuscript. In addition, we provide a detailed explanation of the revised phase coupling analysis, including a formal description how PPC is computed, in the Methods section of the revised manuscript in line 524 ff. (486 without tracked changes).

   Using the revised phase-coupling analysis, we observed comparable PPC values of significantly coupled PYRs (0.013) and INTs (0.014) to high gamma in control mice. While the improved analysis thus resolved the paradoxical finding of lower PPC in INTs, we did not observe weaker phase-coupling of PYRs as reported in Perrenoud et al. (2016). A possible explanation for this discrepancy might be genuine differences in gamma coupling of the PYR population between visual cortex (Perrenoud et al., 2016) and the prefrontal cortex (our study), which will require further investigation in future.

   Reviewer #3 (Public Review):

   In the present study, the authors aim to assess network activity alterations in the prefrontal cortex of mice with a deletion variant in the schizophrenia susceptibility gene DISC1 ("DISC1 mutants"). Using silicon probe in vivo recordings from the prefrontal cortex, they find that mutant mice show reduced firing rates of fast-spiking interneurons, reduced spike transmission efficacy from pyramidal cells to interneurons, and enhanced synchronization and activation of cell assemblies. The authors conclude that "interneuron pathology is linked with the abnormal coordination of pyramidal cells, which might relate to impaired cognition in schizophrenia."

   The cellular and circuit basis of psychiatric disorders has received strong interest in the recent past. In particular, alterations of the "excitation-inhibition balance" in cortical circuits has been the focus of extensive scrutiny (reviewed in pmid 22251963). Specifically, in both human samples as well as in mouse models, disruption of interneuron development and function have been implicated in the pathogenesis of schizophrenia. In the DISC1 mouse model, studies have reported disrupted interneuron development (e.g. pmid 23631734, 27244370), reduced numbers of GABAergic neurons (e.g. pmid 18945897), reduced inhibition from GABAergic neurons ex vivo (e.g. pmid 32029441), and reduced firing rates of fast-spiking neurons in vivo in the basal forebrain (pmid 34143365).

   The present manuscript makes a potentially important contribution to this question by probing the microcircuitry of the prefrontal cortex in vivo in the DISC1 mouse model of schizophrenia. It goes beyond previous work in assessing circuit dynamics in vivo in more detail, albeit with indirect methods. The experiments and analysis have generally carefully been performed, though the statistical analysis raises some questions. The advances made by the present work compared to previous studies could be delineated more clearly.

   We thank the reviewer for praising the analysis of our data ‘…have generally carefully been performed..’ and the ‘important contribution’ of our work to the field.

   Read the original source
3. 

   eLife

   Jul 5, 2022

   Evaluation Summary:

   This manuscript investigates the consequences of Disrupted-in-schizophrenia-1 (Disc1) gene knock out in the medial prefrontal cortex (mPFC) of mice. This work marks a potentially significant contribution to elucidate cortical circuits alterations in this genetic model of schizophrenia. The main message is that communication between cortical pyramidal neurons and fast spiking interneurons is altered with consequence on cortical network activities. The data generally support the conclusions made but analyses of electrophysiological data should improve.

   (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)

   Read the original source
4. 

   eLife

   Jul 5, 2022

   Reviewer #1 (Public Review):

   The authors of the paper provide new evidence of how prefrontal cortex of mutant mice used as a disease model of schizophrenia differs from wild type littermates. By analyzing local network dynamics at the level of specific cell type, authors shed new light on the circuit mechanisms that underlie changes in network dynamics in these mice.

   The claims in the submitted manuscript are supported by the data. I have a few comments and questions that need to be clarified.

   Average firing rates

   Authors claim that they saw a significant reduction in interneuron firing rates in Disc1 mutant mice compared to control mice Fig.1c. However, the difference could be general and not interneuron specific. Due to the high firing rates of interneurons, the statistical test will work better on interneurons than on pyramidal cells …

   Reviewer #1 (Public Review):

   The authors of the paper provide new evidence of how prefrontal cortex of mutant mice used as a disease model of schizophrenia differs from wild type littermates. By analyzing local network dynamics at the level of specific cell type, authors shed new light on the circuit mechanisms that underlie changes in network dynamics in these mice.

   The claims in the submitted manuscript are supported by the data. I have a few comments and questions that need to be clarified.

   Average firing rates

   Authors claim that they saw a significant reduction in interneuron firing rates in Disc1 mutant mice compared to control mice Fig.1c. However, the difference could be general and not interneuron specific. Due to the high firing rates of interneurons, the statistical test will work better on interneurons than on pyramidal cells as pyramidal cells average firing rates are lower. What I suggest to do is to take interneuron cells that fire at a lower rate (lower 33% for example ) and compare the control and Disc1 groups. Also I would suggest to take pyramidal cells that have higher firing rates (upper 33% for example) and compare firing rates across the same groups. One would like to see if these differences are not due to changes in firing rates per se.

   Optogenetic tagging

   Authors indicate that light triggered and spontaneous spike waveform are similar Fig.1d. This is nice, but would be better to see all the tagged neurons. I would suggest showing all optically tagged neurons spike features. Authors can impose with a different color spike features of tagged neurons in Fig.1a. I suspect that since all PVI are narrow spiking and they must fall into the area of blue colored cells in Fig.1a.

   It was not clear why authors assessed only firing rates in last 25ms (line 348-349). If they have a clear justification for this they should provide it. But why not use the latency of the first spike also as an additional metric. A well tagged cell will respond to light pulse with short latency (within 5 ms). My concern is that non PVI cells may increase firing rate after 25ms of stimulation of PVI cells due to disinhibition.

   Spike cross-correlations

   The authors show that spike transmission probability from PYR to PVI is reduced in Disc1 mice compared to the controls Fig.2d and Fig.2e, but what happens to PVI to PYR spike transmission probability? Is it different in those groups? Answering this question is important since the authors discuss this topic in line 185-193.

   Authors could try to link oscillations with spike transmission probabilities. On line 180 authors discuss that lower synchrony between PVI might be responsible for observed reduction in gamma power in Disc1 mutant mice. With the available data authors could test this hypothesis. They can look at spike cross correlations in their pool of INT and PVI (if they have pairs of PVI recorded in the same session) population.

   An alternative way to link oscillations with lower spike transmission probabilities in PYR-PVI pairs is to use synchrony triggered LFP analysis. One could take all time points when PVI and PYR cells fired acausal spikes within 2ms window and look at the LFP around this time point. Than take the average of the synchrony-triggered LFP and look at the power spectrum.

   Cell assembly analysis

   The authors used 10ms for testing synchronization among pairs of PYR neurons in Fig.4a but 25ms for analysis of assembly dynamics. I think the authors justified why they used 25ms bin size, but it was not clear why they used 10ms? Could the authors clarify the reasons behind this decision?

   Read the original source
5. 

   eLife

   Jul 5, 2022

   Reviewer #2 (Public Review):

   This is an interesting paper, in which the authors assessed spiking and network deficits in a well-established mouse model of schizophrenia. This mouse model carries a genetic deletion of the Disrupted-in-schizophrenia-1 (Disc1) gene, which is highly penetrant in the human condition. The authors combined behavioral analyses with state-of-the-art electrophysiological recordings in vivo, coupled to optogenetic tagging, to study a subnetwork formed by a major inhibitory neuron subclass (the parvalbumin (PV)-expressing interneuron) and principal excitatory pyramidal neurons in the medial prefrontal cortex. This work indicates reduced firing rates of PV cells in Disc1-KO mice, likely due to reduced coupling with pyramidal neurons, leading to alterations in local network activity. Indeed, the authors found that …

   Reviewer #2 (Public Review):

   This is an interesting paper, in which the authors assessed spiking and network deficits in a well-established mouse model of schizophrenia. This mouse model carries a genetic deletion of the Disrupted-in-schizophrenia-1 (Disc1) gene, which is highly penetrant in the human condition. The authors combined behavioral analyses with state-of-the-art electrophysiological recordings in vivo, coupled to optogenetic tagging, to study a subnetwork formed by a major inhibitory neuron subclass (the parvalbumin (PV)-expressing interneuron) and principal excitatory pyramidal neurons in the medial prefrontal cortex. This work indicates reduced firing rates of PV cells in Disc1-KO mice, likely due to reduced coupling with pyramidal neurons, leading to alterations in local network activity. Indeed, the authors found that Disc-KO mice exhibited reduced levels of gamma oscillations and somewhat hypersynchronous networks.

   Taking advantage of novel techniques and analytical strategies, the manuscript provides rich, novel insight into the neurobiology of a mouse model of this severe psychiatric condition. The data is of high quality, the findings interesting and the manuscript is well written.

   Overall, the results support the authors' conclusions, although some additional analyses are necessary to corroborate their interpretations.

   Although the paper does not give information on how PV cell dysfunctions are engaged during cognitive tasks, this study can be considered as an important first step in advancing our knowledge on the basic dysfunctions of cortical networks in this model of schizophrenia

   1. The major findings stem from the analysis of the spiking activity of individual neurons recorded using either silicon probes or arrays of tetrodes. Both techniques allow simultaneous recording of many neurons from a single animal; therefore, from a statistical point of view neurons recorded from one animal are pseudo replicas and cannot be considered as independent measurements.
   Throughout the manuscript, the authors perform two-sample tests on the pooled data from all recorded neurons to compare differences between genotypes; therefore, artifactually increasing the power of statistical tests. Comparisons between genotypes should be performed using each mouse as an independent measurement.

   2. The superficial layers of the mPFC are difficult to reach with a vertical approach of the probes due to the presence of a large blood vessel located medially in the frontal dura. Therefore, the authors are most likely reaching mPFC deep layers where PYR neurons produce fast spikes at high rates. If this is the case, this would make it difficult to sort the spiking of PYR from that of INs based on the spike kinetics and rate. The authors used opto-tagging of PVIs in a set of experiments. It would be reassuring to confirm that the spike waveform and kinetics that they extracted from PVIs are similar to those they assigned as INTs in their experiments with no opto-tagging. Identified PVIs should be statistically different from putative PYRs (not responding to light).
   Although opto-tagging of PVIs can solve this issue, the amount of cells isolated remains low and the number of animals is not stated. Opto-tagged cells are subsequently used for further analyses but the statistical value of those remain unclear. Since the entire interpretation of the rest of the results depend on this result, this must be clarified.

   3. Proportion of gamma coupled neurons. The authors mention the use of pairwise phase consistency (PPC). PPC is a good method to measure phase coupling independent of differences in firing rates. However, it is not entirely clear how PPC is used to measure the extent of phase locking. In the methods, the authors mention that they ran the PPC analysis *after* determining significant phase locking with Rayleigh's test. Moreover, they provide PPC values for high gamma oscillations but not for other frequency ranges. Perhaps, it would be better to test significant coupling of all units by nonrandom spike-phase distributions crossing a confidence interval, estimated by Monte Carlo methods from independent surrogate data set. These can be obtained upon randomly jittering each spike times. Indeed, PPC values estimated by the authors for high gamma are higher for PYR than INT (Fig. 1- Fig. Suppl 4 b). This is at odds with previously published observations in V1 (e.g. Perrenoud et al., PLoS Biol. 2016 PMID: 26890123). Given the existing reports of reduced excitatory transmission in DISC-1 mice, phase locking of PYR to other frequency bands might be affected.

   Read the original source
6. 

   eLife

   Jul 5, 2022

   Reviewer #3 (Public Review):

   In the present study, the authors aim to assess network activity alterations in the prefrontal cortex of mice with a deletion variant in the schizophrenia susceptibility gene DISC1 ("DISC1 mutants"). Using silicon probe in vivo recordings from the prefrontal cortex, they find that mutant mice show reduced firing rates of fast-spiking interneurons, reduced spike transmission efficacy from pyramidal cells to interneurons, and enhanced synchronization and activation of cell assemblies. The authors conclude that "interneuron pathology is linked with the abnormal coordination of pyramidal cells, which might relate to impaired cognition in schizophrenia."

   The cellular and circuit basis of psychiatric disorders has received strong interest in the recent past. In particular, alterations of the "excitation-inhibition …

   Reviewer #3 (Public Review):

   In the present study, the authors aim to assess network activity alterations in the prefrontal cortex of mice with a deletion variant in the schizophrenia susceptibility gene DISC1 ("DISC1 mutants"). Using silicon probe in vivo recordings from the prefrontal cortex, they find that mutant mice show reduced firing rates of fast-spiking interneurons, reduced spike transmission efficacy from pyramidal cells to interneurons, and enhanced synchronization and activation of cell assemblies. The authors conclude that "interneuron pathology is linked with the abnormal coordination of pyramidal cells, which might relate to impaired cognition in schizophrenia."

   The cellular and circuit basis of psychiatric disorders has received strong interest in the recent past. In particular, alterations of the "excitation-inhibition balance" in cortical circuits has been the focus of extensive scrutiny (reviewed in pmid 22251963). Specifically, in both human samples as well as in mouse models, disruption of interneuron development and function have been implicated in the pathogenesis of schizophrenia. In the DISC1 mouse model, studies have reported disrupted interneuron development (e.g. pmid 23631734, 27244370), reduced numbers of GABAergic neurons (e.g. pmid 18945897), reduced inhibition from GABAergic neurons ex vivo (e.g. pmid 32029441), and reduced firing rates of fast-spiking neurons in vivo in the basal forebrain (pmid 34143365).

   The present manuscript makes a potentially important contribution to this question by probing the microcircuitry of the prefrontal cortex in vivo in the DISC1 mouse model of schizophrenia. It goes beyond previous work in assessing circuit dynamics in vivo in more detail, albeit with indirect methods. The experiments and analysis have generally carefully been performed, though the statistical analysis raises some questions. The advances made by the present work compared to previous studies could be delineated more clearly.

   Read the original source
7. 

   Version published to 10.1101/2021.12.10.472099v1 on bioRxiv

   Dec 11, 2021