Increased expression of schizophrenia-associated gene C4 leads to hypoconnectivity of prefrontal cortex and reduced social interaction

Schizophrenia is a severe mental disorder with an unclear pathophysiology. Increased expression of the immune gene C4 has been linked to a greater risk of developing schizophrenia; however, it is not known whether C4 plays a causative role in this brain disorder. Using confocal imaging and whole-cell electrophysiology, we demonstrate that overexpression of C4 in mouse prefrontal cortex neurons leads to perturbations in dendritic spine development and hypoconnectivity, which mirror neuropathologies found in schizophrenia patients. We find evidence that microglia-mediated synaptic engulfment is enhanced with increased expression of C4. We also show that C4-dependent circuit dysfunction in the frontal cortex leads to decreased social interactions in juvenile and adult mice. These results demonstrate that increased expression of the schizophrenia-associated gene C4 causes aberrant circuit wiring in the developing prefrontal cortex and leads to deficits in juvenile and adult social behavior, suggesting that altered C4 expression contributes directly to schizophrenia pathogenesis.

Introduction Schizophrenia (SCZ) is a neurodevelopmental disorder characterized by disruptions in brain connectivity that lead to an array of symptoms including psychosis and deficits in cognition and social interactions [1,2]. Current treatment options are somewhat effective for treating psychosis but do not improve cognitive or social deficits, both of which contribute significantly to the long-term prognosis of people with SCZ [3][4][5]. Early social deficits are a feature of this disease and are often present even before the onset of psychosis [6]. Therefore, understanding the underlying mechanisms of early social deficits in SCZ could reveal a therapeutic window to alter disease progression in these individuals.
Among the many clinical traits of SCZ are deficits in social cognition such as social-cue perception and emotion regulation, which suggest dysfunction of the prefrontal cortex (PFC) [5]. Moreover, patients with SCZ exhibit abnormal activity in the PFC during affective face perception and cognitive reappraisal [7][8][9]. These results are not surprising given the well-established role of the PFC in social behaviors [10][11][12]. Although the mechanisms that cause social deficits in SCZ are unclear, evidence suggests that dysfunction in the PFC correlates with symptom onset and severity [13]. In fact, a neurological hallmark of SCZ is a reduction of gray matter in the PFC due to a loss of neuronal processes and synapses [14,15]. Therefore, it has been hypothesized that pathology in SCZ arises in part because of deficits in the pruning of cortical synapses, thus producing aberrant circuitry [14,15].
SCZ is highly heritable and genome-wide association studies have shown that the major histocompatibility complex (MHC) has the greatest genetic association with SCZ [16][17][18][19]. In particular, C4, which is located in the MHC locus, is associated with SCZ such that specific structural variants and mutations that cause increased expression of C4 confer greater risk for this brain disorder [16]. Human C4 (hC4) is encoded by two distinct genes, or isotypes, C4A and C4B, whereas mouse C4 (mC4), encoded by a single gene (C4b), contains an amino acid sequence that is approximately 80% similar to both hC4 isotypes [20][21][22]. C4 is a member of the innate immunity complement cascade that forms a proteolytic protein cascade that clears debris, enhances inflammation, and tags pathogens for engulfment or destruction [23]. It has been shown that microglia, the brain's resident macrophage, engulf neuronal and synaptic material dependent on the presence of complement proteins [24,25]. Microglia are the only cells in the brain that express the complement 3 receptor (CR3), which allows microglia to recognize and phagocytose material tagged by the complement cascade [24][25][26]. Microglia-mediated synaptic pruning is necessary for normal development of the brain but can also contribute to abnormal wiring of circuits and pathology when mis-regulated [16,[27][28][29][30].
Despite significant progress in this field, it is still not clear how reciprocal interactions between neurons and microglia contribute to the maturation and refinement of developing cortical circuits [31][32][33]. Mouse studies investigating the role of C4 and other complement proteins in synaptic refinement have relied on loss-of-function manipulations [16,24,25,34], whereas SCZ is linked to increased C4 expression [16]. Additionally, previous studies have focused on the role of complement proteins in the visual system [16,24,25] rather than interrogating their role in more SCZ-relevant brain regions, such as the PFC. Thus, overall, it is not known how increased C4 expression impacts the development of cortical microcircuits, which are commonly altered in neurodevelopmental disorders [35,36].
Herein, we investigated the effects of C4 overexpression on developing networks in layer (L) 2/3 of the mouse medial prefrontal cortex (mPFC), a cortical region where gray-matter loss is most prominent in SCZ [14,15]. We show that C4 overexpression induced transient structural and functional changes to L2/3 excitatory neurons in the mPFC. Additionally, we interrogated the role of microglia in complement-induced wiring deficits in the mPFC by measuring expressed (S3A and S3B Fig). We found C4-GFP localized to cell bodies and dendrites in L2/3 neurons (S3B Fig, yellow arrowhead). Overall, these results showed that mPFC L2/3 neurons express endogenous mC4, that a subpopulation of C4 is synaptically localized, and that IUE successfully increased mC4 levels at P21.

C4 overexpression causes dendritic spine alterations in apical arbors of L2/ 3 neurons
Since we found C4 in PSD fractions, we wondered whether increased expression of C4 in developing neurons led to structural alterations in dendritic spines. We used IUE to produce control (pCAG-GFP) or C4 overexpression conditions (pCAG-GFP with pCAG-mC4b or and quantification (bottom) of relative mC4 levels in total lysates and isolated PSD fractions from GFP or mC4 conditions from the electroporated region. Since C4 was expressed at relatively low levels, it could only be detected when brains were pooled. Lysates and PSDs from the electroporated region were prepared from the individual mice and pooled for the western blot analysis. N = 7 mice per group for P21 (left 4 lanes). N = 4 mice per group for P60 (right 4 lanes). Vinculin was used as loading control for lysates and PSD fractions. PSD-95 immunoblot served as a control for successful isolation of PSD fractions. For underlying data, see https://osf.io/7em3s/?view_only= 0e7ffde4ebd344dc83af83b5a605c451. E, embryonic day; GFP, green fluorescent protein; IUE, in utero electroporation; L2/3, layer 2/3; mC4, mouse C4; mPFC, medial prefrontal cortex; P, postnatal day; PSD, postsynaptic density; qPCR, quantitative PCR. pCAG-hC4A, for mC4 and hC4 conditions, respectively). Since early postnatal development is an important period for synaptic maturation and refinement of cortical circuitry [38], we measured dendritic spine density using confocal imaging of GFP in fixed brain sections collected at P7-9, P14-16, P21-23, and P55-60. Besides control and mC4 conditions, for comparison we also electroporated hC4A, which is highly associated with SCZ [16].
In control conditions, the density of apical tuft dendritic spines was developmentally regulated and increased nearly 8-fold in the 2 wk from P7 to P21 (Fig 2A, Tukey's test, ���� p < 0.0001). Whereas spine density in mC4-or hC4-overexpressing neurons was similar to that of control conditions at P7 and P14, apical tuft dendritic spine density was approximately 30% lower at P21 (Fig 2A and 2B). There were no differences in apical tuft protrusion density between any groups at P60 (Fig 2A). These results suggest that overexpression of C4 alters the developmental maturation of dendritic spines and that overexpression of either human or mouse homologues of C4 causes spine loss/dysgenesis in the mouse mPFC.
Closer inspection of protrusion morphology at P21 revealed a significant reduction of thin spines/filopodia and medium-sized spines in neurons overexpressing either mC4 or hC4, as compared to controls, whereas putative large mushroom spines remained intact (Fig 2C and  2D; spine types sorted by total integrated brightness [TIB], see Methods). Since dendritic spine volume positively correlates with PSD size and synaptic strength [39,40], this result suggests that increased expression of mC4 preferentially alters the development of weaker synaptic connections. Since overexpression of mC4 and hC4 led to similar defects in dendritic spine density at P21, in subsequent experiments we focused on the mouse homologue to investigate the role of C4 in cortical circuit function. In contrast to the structural alterations we observed in apical tufts, increased expression of mC4 did not significantly alter the density of basal dendritic spines, as compared to control conditions (Fig 2E and 2F). Taken together, these results demonstrate that increased expression of C4 in developing cortical neurons caused input-specific dendritic spine pathology and that less mature protrusions were preferentially affected, whereas larger mushroom spines remained intact.

C4 overexpression reduced functional connectivity in cortical neurons
To determine whether the morphological spine deficits observed in C4-overexpressing cells were accompanied by functional connectivity changes, we performed electrophysiological whole-cell voltage-clamp recordings in acute brain slices prepared from mPFC of P18-25 control and mC4 animals. Voltage-clamp recordings demonstrated that overexpression of mC4 caused an approximately 60% decrease in the frequency of miniature excitatory postsynaptic currents (mEPSCs) and an increase in the interevent interval (IEI) (Fig 3A-3C). Increased levels of mC4 also caused an approximately 18% reduction in the amplitude of mEPSCs (Fig 3D  and 3E), demonstrating that mC4 modifies not only the density of synaptic connections but also the postsynaptic responses of L2/3 neurons.
Further analysis of postsynaptic current kinetics revealed no differences in mEPSC Rise  or Decay tau between control and mC4-transfected neurons (Fig 3F and 3G). Although overexpression of mC4 did not alter the frequency or the amplitude of miniature inhibitory postsynaptic currents (mIPSCs) (Fig 3H-3L), it significantly slowed their kinetics (Fig 3M and 3N, 27% and 19% increase in Rise 10-90 and Decay tau , respectively), suggesting that there could be mC4-dependent changes in the location of inhibitory synapses. Taken together, our data suggest that increased expression of mC4 alters the developmental wiring of cortical neurons by decreasing their excitatory synaptic drive.

Overexpression of mC4 decreased membrane capacitance without altering overall excitability
We also monitored the electrophysiological properties of L2/3 mPFC neurons to determine whether overexpression of mC4 altered neuronal excitability. To do this, we injected hyperpolarizing and depolarizing current step injections of different amplitudes and recorded the membrane potential (V m ) in control and mC4-overexpressing neurons (Fig 4A and 4B). Overexpression of mC4 did not alter the number of action potentials (APs) in response to step current injections of incrementally increasing amplitudes (Fig 4C). In agreement with this, we did not see a change in the IEI of APs in response to step current injections (Fig 4D) or the minimal current injection that resulted in AP generation ( Fig 4E) in mC4-overexpressing neurons.
Together, these results suggest that overexpression of mC4 did not alter the intrinsic excitability of cortical neurons.
In contrast, increased levels of mC4 led to an approximately 29% decrease in the membrane capacitance relative to controls (Fig 4F). This effect on passive membrane properties was specific because increased expression of mC4 did not alter the neuron's resting membrane potential or membrane input resistance (Fig 4G and 4H). Since membrane capacitance is proportional to a cell's surface area, this result suggests that increased expression of mC4 could affect either the gross morphology of cortical neurons or plasma membrane properties.  Further investigation of neuronal morphology showed no difference between conditions in cell body area/diameter or in the width of primary dendrites (S4 Fig); there was also no change in dendritic complexity (S5 Fig), suggesting that gross neuronal morphology was not affected by mC4 overexpression. Besides the mC4-dependent changes in membrane capacitance, increased expression of mC4 did not have an effect on the morphology or intrinsic properties of transfected cells, thus suggesting that overall neuronal health was not affected by increased expression of mC4.

mC4 overexpression enhanced microglia engulfment of postsynaptic PSD-95
Several lines of evidence suggest that microglia contribute to circuit maturation and synaptic refinement through the secretion of effector molecules and direct interactions with neurons [41]. To determine whether increased expression of mC4 led to closer associations of microglia with neurons at P21, we measured the colocalization of transfected neuronal GFP with microglia (immunostained for ionized calcium binding adaptor molecule 1 [Iba1]) in single z-planes using confocal microscopy (S6A Fig). Using this approach, we found instances of microglia colocalized with GFP+ neuronal processes in both conditions (S6B Fig; white arrowheads), suggesting that microglia were in close proximity with the dendritic processes of transfected L2/3 neurons. Increased expression of mC4 led to an approximately 35% increase in the number of microglia that colocalized with GFP+ L2/3 neurons (S6C Fig) and an approximately 2-fold increase in the area of GFP signal that colocalized with the microglia cell body and proximal processes (S6D Fig). These results suggest that neuronal overexpression of mC4 led to an increase in the number of microglia in close proximity to neurons despite no change in the overall density of microglia in the transfected region (S7 Fig).
Since increased expression of mC4 led to a decrease in synaptic connectivity ( Next, we quantified the colocalization of PSD-95 (identified by PSD95-FingR immunofluorescence) with microglia (Iba1) or microglial lysosomes (CD68) (Fig 5). We observed clear instances of PSD-95 colocalized with Iba1 ( Fig 5A and 5C, white arrowheads) and CD68 signal (Fig 5B and 5D, white arrowheads), suggesting that synaptic material was engulfed by microglia and localized to microglia lysosomes. Furthermore, using orthogonal views, we confirmed that PSD-95 was localized within microglia soma and lysosomal compartments (Fig 5A-5D). We quantified the colocalization of fluorescent signals from PSD-95-FingR with both Iba1 and CD68 to measure postsynaptic puncta within the lysosomes of microglia in control and mC4 conditions. Overexpression of mC4 led to an approximately 2.3-fold increase in the percentage of microglia positive for engulfment ( Fig 5E). As a control, we analyzed the same images after pixel shifting (see Methods). In the pixel-shifted data, the percentage of microglia positive for triple-stained puncta dropped to under 3% in both groups (Fig 5E, red dotted line is mean shifted value for both groups), suggesting that the colocalization of signals in microglia did not occur by chance. Lastly, we observed no change in the area of PSD-95-FingR signal within CD68+ structures (Fig 5F). Taken together, these data show that C4 overexpression drives microglia engulfment of synaptic material.
Since increased expression of mC4 led to specific spine loss in L1 apical tufts with no changes in basal dendrites (Fig 2), we tested whether microglia engulfment of synaptic material was more prominent near the apical dendrites of cortical neurons. We measured the distance from the center of each microglia soma to the pia to identify its cortical depth, and this measurement was then compared to the area of microglia-engulfed PSD-95 signal. This allowed us to determine whether the connectivity deficits in apical tufts were due to altered microglia phagocytosis of synapses in a layer-specific manner. In control conditions, there was no correlation between the amount of PSD-95 colocalized with microglia and cortical depth, suggesting that, during normal development, microglia were not biased toward phagocytosis of PSD-95 in specific layers ( Fig 5G). However, we observed a significant correlation in the mC4 condition such that microglia closer to L1, where apical tufts are located, engulfed more PSD-95 material than their counterparts in deeper layers (Fig 5H). To test whether mC4-dependent enhancement of phagocytosis in L1 was due to an increased density of microglia, we quantified microglia density in L1 (cortical depth: 0-120 μm) and L2/3 (cortical depth: >120-300 μm) and found no difference between microglia density between layers in either condition (S7 Fig). There was also no difference in CD68 area within microglia between control and mC4 conditions (S7 Fig). In summary, our data support the hypothesis that C4 overexpression drives layer-specific circuit dysfunction through excessive microglia engulfment of postsynaptic material.
Additionally, we used expansion microscopy (ExM), a biological sample-preparation procedure that allows for nanoscale imaging with standard confocal microscopes [45] (S9 Fig), to confirm colocalization of fluorescent signals from PSD-95 and CD68 located in microglia. Using ExM, we increased the size of microglia soma by approximately 180% (equivalent to approximately 2.8× linear expansion) in control and mC4 conditions at P21 (S9D- S9F Fig). Microglia cell body size increased similarly along its long and short axis, supporting previous findings that have shown that ExM expands biological tissue linearly [45] (S9D- S9F Fig).
Other than increasing their size, expanded microglia had morphologies similar to cells in preexpansion conditions (S9 Fig), suggesting that ExM did not fundamentally alter the structure of these brain-resident macrophages.
Using ExM, we observed clear examples of PSD-95 colocalized with Iba1 ( Fig 6A and 6C, white arrowheads) and CD68 signal (Fig 6B and 6D), suggesting synaptic material was localized to microglia lysosomes. Confocal images obtained after ExM show examples of microglia lysosomes that were both negative and positive for PSD-95 in control and mC4 conditions (S10 Fig). Overexpression of mC4 in L2/3 neurons led to an approximately 3.3-fold increase in lysosomes positive for PSD-95 ( Fig 6E). Additionally, microglia in the mC4 condition contained significantly more lysosomes relative to control (Fig 6F, approximately 1.2-fold increase). In contrast, previously we found no change in CD68 reactivity using conventional confocal microscopy (S7B Fig); this could be due to the enhanced resolution made possible by ExM. For both conditions, lysosomes positive for PSD-95 were approximately 1.5-fold larger than those that were negative for postsynaptic material ( Fig 6G). Additionally, lysosome size did not differ between conditions for PSD-95(+) lysosomes (con[+] versus mC4[+]) or for PSD-95(−) lysosomes (con[−] vs mC4 [−]). This suggests that lysosomes positive for PSD-95 were morphologically different than lysosomes without synaptic material for both control and mC4 conditions. In summary, our data support the hypothesis that C4 overexpression drives layer-specific circuit dysfunction through excessive microglia engulfment of postsynaptic material.

Complement-dependent alterations in frontal cortical circuitry were sufficient to alter social interactions in juvenile and adult mice
To determine whether overexpression of mC4 in the frontal cortex is sufficient to cause deficits in early (P18) social behaviors, we administered a task that allowed us to measure sensorimotor abilities of mice and maternal-pup social interactions [46]. For these experiments, we used a modified IUE method (see Methods) to target large populations of L2/3 neurons in both hemispheres of the frontal cortex. Post hoc analysis of brains from these mice showed that most transfected cells were in prefrontal cortical regions, confirming that we were able to increase mC4 expression in large populations of PFC L2/3 neurons (S11 Fig).
To test sensorimotor abilities (maternal interaction 1 [MI1] task), we transferred P18 pups to an arena that contained fresh bedding in two neutral corners and nesting material from the animal's home cage in the corner opposite to the starting corner, and mice were free to explore the arena for 3 min (Fig 7A). Similar to controls, mC4 mice spent more time exploring the nest corner relative to the neutral corners, suggesting that mC4 mice had normal homing behavior ( Fig 7B). Although grooming occurrences of mC4 mice were more frequent and about 4.3-fold longer (S12 Fig) than in control mice, there was no interaction between experimental group and corner preference ( Fig 7B). These results indicate that although mC4 mice engaged in more repetitive behavior, they were able to explore the arena and interact with their nest bedding to the same extent as control animals. Additionally, these results suggest Complement component 4 and prefrontal cortical dysfunction that increased levels of mC4 in the frontal cortex did not impair overall sensory abilities or cause gross motor deficits in P18 pups.
To test maternal-pup social interactions (maternal interaction 2 [MI2] task), we placed P18 pups in an arena that contained nest bedding, an empty wire mesh cup, and a wire mesh cup containing the pup's dam, and the behavior of each pup was monitored for 5 min (Fig 7C). We found that mC4 mice showed an approximately 23% reduction in time spent exploring the cup containing the dam as compared to control mice ( Fig 7D). mC4 mice also exhibited an approximately 2.5-fold decrease in the latency to first approach the dam's cup ( Fig 7E). The proportion of pups that approached the dam's cup before the empty cup (at start of trial) was approximately 50% (Fig 7F), suggesting that pups in both conditions engaged in similar Complement component 4 and prefrontal cortical dysfunction exploratory behavior of the arena before seeking their mother. mC4 and control mice had similar latencies to first reach either the empty or dam's cup ( Fig 7G) and similar maximum speeds ( Fig 7H). Taken together, these results indicate that mC4 pups had intact motor and sensory abilities but were less motivated or interested in interacting with their mother.
Although we did not observe changes in dendritic spine density at P60 between conditions (Fig 2A), it is possible that increased C4 expression in adult mice resulted in changes in the function of frontal L2/3 neurons that could not be directly assessed by analyzing the mean dendritic spine density (e.g., miswiring of synaptic connections or intrinsic excitability). Therefore, we tested whether bilateral overexpression of mC4 in the frontal cortex led to changes in sensorimotor abilities or social behavior in adult (P60-70) mice (N = 22 control mice, N = 20 mC4 mice). We first tested adult mice in a novel-object task to assess whether interest in a novel object was affected by C4 overexpression (Fig 8A). Control and mC4 mice spent similar amounts of time exploring the novel object ( Fig 8B). Next, a novel-object recognition task was performed, in which mice were placed in an arena for 5 min with a novel object in one corner and a familiar object in the opposite corner ( Fig 8C). The discrimination index (DI) for object interaction (DI = [time with novel object − time with familiar object] / [time with novel object + time with familiar object]) was positive for both control and mC4 mice, indicating that mice in both conditions spent more time exploring the novel object relative to the familiar object ( Fig 8D). Additionally, the DI for control and mC4 mice was not different (Fig 8D), suggesting that overexpression of mC4 in the frontal cortex did not alter sensorimotor abilities or general motivation. In support of this, mC4 mice traveled at similar speeds and distances in an open field (OF) arena relative to the control group (Figs 8J and S13).
Next, we placed either control or mC4 adult mice in an arena that contained an empty wire mesh cup and a wire mesh cup containing a novel mouse (strain, sex, and age matched) in opposing corners, and the behavior of each mouse was monitored for 5 min (Fig 8E). We found that adult mC4 mice showed a significant reduction in social DI, a metric of sociability ([time spent with novel mouse − time spent with empty cup] / [time spent with novel mouse + time spent with empty cup]), as compared to control mice ( Fig 8F). This reduction in sociability was not due to an overall increase in vigilance or anxiety-like behavior in the mC4 condition, because both groups spent similar amounts of time in the open arms of the elevated-zero maze (EZM) (S13 Fig). We found no difference between control and mC4 mice in latency to approach either the novel mouse or empty cup, or in the proportion of mice choosing the novel mouse first (Fig 8G-8I). Overall, these results suggest that increased expression of mC4 in frontal L2/3 neurons led to long-lasting alterations in PFC circuitry that were sufficient to cause a reduction in social interactions in both juvenile and adult mice.

Discussion
We showed here that overexpression of C4 in the mPFC can perturb circuit development. We demonstrated that, during normal early postnatal development, excitatory neurons in the mouse mPFC express low levels of C4. Moreover, overexpression of C4 in these cells led to spine dysgenesis that was evident through a transient decrease in filopodia and medium-sized spine density during the third week of postnatal development. We also showed that C4-dependent spine abnormalities were accompanied by a decrease in excitatory synaptic drive onto pyramidal cells in the mPFC. C4 overexpression caused microglia to engulf more PSD-95 from transfected neuron's apical dendrites in L1, where spine loss was observed. Lastly, increased expression of C4 in the frontal cortex was sufficient to reduce social interactions in juvenile (P18) and adult (P60-70) mice. Our findings implicate C4 in shaping the developmental trajectory of cortical circuits and provide a causal link between increased C4 levels and PFC dysfunction. These data show for the first time, to our knowledge, that C4-driven circuit dysfunction in the PFC drives cellular and behavioral phenotypes observed in SCZ rather than merely being a consequence of disease and/or pharmacological treatment. Our morphological characterization of L2/3 mPFC neurons revealed that C4-mediated connectivity deficits are both circuit and spine-type specific. That is, overexpression of C4 caused a decrease in spine density of apical but not basal dendrites of L2/3 neurons. These data provide evidence that C4-driven synaptic alterations in the mPFC are cortical layer and/or input specific. Interestingly, previous data have shown that exposure to repeated stress in adolescent rats can reduce connectivity in superficial L1 of the mPFC [47], suggesting that this layer is susceptible to genetic and environmental perturbations during development. In SCZ, dendritic spines changes are most prominent in L2/3 of the mPFC [14]; therefore, we focused on how overexpression of mC4 causes anatomical and functional changes in this region. However, imaging studies have reported structural changes in a variety of brain regions in SCZ, including the frontal and temporal lobes and hippocampus [48]. It remains unknown whether mC4 overexpression in other brain regions leads to a decrease in spine density or other abnormalities. This question could be answered by targeting other brain regions using IUE or generating a global mC4-overexpressing transgenic mouse.
When we further investigated which spine types were affected by C4 overexpression, we found a specific loss of filopodia and medium-sized spines, whereas the density of mature mushroom spines were intact. This is in line with previous observations that complementmediated pruning is activity dependent [25]. Large mushroom and stubby spines are also associated with an accumulation of extracellular matrix proteins, which could protect them from synaptic elimination by microglia [49]. Future studies could shed light on the molecular mechanisms of input-specific plasticity by identifying "eat me"/"don't eat me" signals [50,51] or complement proteins that are differentially present in large mushroom versus filopodia spine types.
Major contributors to synaptogenesis are dendritic filopodia, which are thin, specialized postsynaptic structures that orchestrate synapse formation by dynamically sampling potential presynaptic partners, thus optimizing circuit wiring [52,53]. It is possible that C4 contributes to aberrant synaptic wiring by reducing the ability of developing circuits to form appropriate connections. At its simplest, increased levels of C4 may underlie PFC pathology and dysfunction by specifically altering filopodia-dependent synapse formation during early development. This interpretation would suggest that subsequent aberrant synaptic elimination in SCZ is a consequence of suboptimal circuit wiring. Transient deficits in dendritic spines and filopodia are seen in several mouse models of neurodevelopmental disorders [35]. Similar to these models, our data suggest that there are sensitive developmental periods during which cortical circuitry is especially susceptible to altered expression of C4.
In our experiments, we did not observe a significant difference in apical tuft protrusion density between adult (P60) control and C4 conditions. However, in human SCZ postmortem tissue, spine loss of PFC cortical L3 neurons persists into adulthood [14]. Notably, our genetic manipulation targeted a single gene, C4, whereas SCZ results from polygenetic mechanisms [54], implicating a multitude of genes. It is possible that the mechanisms that allow for the compensation of spine density in P60 mice are also impaired in SCZ. Therefore, our data suggest that the C4-induced connectivity changes observed during the third week of development could stimulate compensatory plasticity mechanisms not present in SCZ. Additionally, there may be differences in how human and mouse pyramidal neurons respond to increased levels of C4, which could explain the discrepancy in adult dendritic spine density between mice and humans. Importantly, our qPCR and western blot data show that C4 levels at P60 were not increased in the mC4 condition mice. Therefore, the fact that spine density in C4 conditions recovered to control levels by P60 could simply be due to the lack of C4 overexpression at P60. Although spine density in C4-overexpressing neurons returned to control levels by P60, our behavioral data in adult mice suggest that the underlying circuitry remains aberrant in mice overexpressing C4 in the frontal cortex. It remains to be determined whether increased expression of C4 can lead to other abnormal changes in circuitry (i.e., L2/3 output to deeper cortical layers or PFC network properties) that were not measured in this study that persist into adulthood.
Our electrophysiological data strengthen our morphological findings and provide additional insight into the functional abnormalities caused by C4 overexpression. Previous studies have shown that microglia are able to trogocytose-or "nibble" and phagocytose-small volumes of neurons and synapses [55]. It is possible that microglia trogocytose some of the PSD without fully eliminating synapses, thus causing a reduction in synaptic strength due to a loss of glutamate receptors. In our experiments, the decrease in excitatory synaptic drive caused by C4 overexpression could be the result of an increase in phagocytosis of synapses by microglia. Alternatively, microglia trogocytosis of synaptic material could trigger synaptic elimination mediated through conventional synaptic mechanisms such as long-term depression [56]. On the other hand, C4 did not induce a change in the frequency or amplitude of mIPSCs, which is in line with previous observations showing that complement does not target GABAergic synapses in mouse and cellular models of Alzheimer's disease [57]. The slowed kinetics of inhibitory transmission could be mediated through a loss of dendritic spines, which has been shown to alter inhibitory synapse location [58].
In addition to being the brain's resident macrophage, microglia can influence brain development by playing roles in synapse formation, elimination, and maintenance [28,41,59]. Our data show that increased C4 expression led to morphological changes in lysosomes and enhanced phagocytic activity. Since microglia are the only cells in the brain that express CR3 [59], they are equipped to recognize material that has been tagged by the complement cascade to initiate phagocytosis. Indeed, previous work has shown that in normal brain development, microglia phagocytose synaptic material, including PSD-95, and that this process of synapse elimination is necessary for normal brain wiring [27]. Our data in the PFC are consistent with and expand upon previous studies in the retinogeniculate pathway that show microglia engulf synapses in a complement-dependent manner [16,24,25], which suggests that complementdependent circuit wiring is necessary for normal development and contributes to pathology when mis-regulated.
Social deficits in individuals with SCZ are an early feature of this disease, and poorer social functioning is associated with worse functional outcome in adults with SCZ [60,6]. Therefore, investigating the causes of early social deficits could reveal potential targets for early therapeutic interventions in SCZ. In L1, the apical dendrites of pyramidal cells in multiple cortical layers interact with projections from higher-order areas [61]. It is thought that cortical "feedback" that inputs to L1 exerts "top-down" control that is important for higher-order cognitive processes [62]; therefore, the vulnerability of this layer to C4 overexpression could provide a cellular substrate for social dysfunction. The changes that we observed in connectivity and behavior parallel known SCZ phenotypes in humans [5,6,[13][14][15] and provide evidence that the PFC regulates social behavior. Future studies could interrogate the role of C4 in other SCZ-relevant phenotypes such as deficits in memory function [63].
Previous work has shown that mice deficient for various complement proteins, including C3R, C1q, C3, and C4, have a reduced pruning of synaptic terminals in the visual system during development [16,24,25]. Since it has been shown that other complement proteins play a role in developmental pruning, it is possible that C4 exerts its effects through the activation of the complement cascade. However, C4 might act on other pathways in neurons to alter connectivity, independent of the complement cascade. Future work could aim to elucidate noncanonical pathways that C4 could interact with by performing proteomics or transcriptomics in neurons.
Although polygenetic mechanisms underlie SCZ [17], we provide evidence that altering expression of a single gene, C4, in mPFC neurons is sufficient to cause cellular and behavioral phenotypes that resemble SCZ pathology. We propose that mPFC neurons express low levels of C4 during normal brain development, which makes them especially susceptible to genetic and environmental risk factors that increase C4 expression, which are associated with SCZ. In summary, we have identified a critical developmental window during which prefrontal cortical circuits are susceptible to alterations in C4 expression, thus opening the possibility for early therapeutic intervention to alter the developmental trajectory of SCZ.

Ethics statement
All experimental protocols were conducted according to the National Institutes of Health (NIH) guidelines for animal research and were approved by the Boston University Institutional Animal Care and Use Committee (IACUC; protocol #17-031).

Animals
All mice were group housed on a 12-hr light and dark cycle with the lights on at 7 AM and off at 7 PM and with food and water ad libitum. Offspring were housed with dams until weaning at P21. For spine density and morphology, microglia engulfment, electrophysiology, and behavior experiments, wild-type CD-1 mice of both sexes from age 7 to 70 d were used (CD-1 IGS; Charles River; strain code: 022). For in situ hybridization, C4b KO mice in a C57BL/6 genetic background (The Jackson Laboratory, stock number: 003643) and wild-type C57BL/6 mice of both sexes at age P30 were used (Charles River; strain code: 027). Males and females were used for all experiments, and no differences were found between sexes.

IUE
L2/3 progenitor cells in the mPFC were transfected via IUE [37]. pCAG-EGFP was electroporated in the control condition, and pCAG-EGFP plasmid was co-electroporated with either pCAG-mC4 or pCAG-hC4A plasmids for the mC4 or hC4 groups, respectively. Prior to surgery, all tools were sterilized by autoclaving. Aseptic techniques were maintained throughout the procedure, and a sterile field was prepared prior to surgery using sterile cloth drapes. Animals were weighed, and a combination of buprenorphine (3.25 mg/kg; SC) and meloxicam (1-5 mg/kg; SC) was administered as a preoperative analgesic. Timed-pregnant female CD-1 mice at E16 were anesthetized by inhalation of 4% isoflurane and maintained with 1%-1.5% isoflurane via mask inhalation. The abdomen was sterilized with 10% povidone-iodine and 70% isopropyl alcohol (repeated 3 times) before a vertical incision was made in the skin and then in the abdominal wall. The uterine horn was then exposed to allow injection of 0.5-1.0 μl of DNA solution (containing 1 μg/μl plasmid and 0.1% Fast Green) into the lateral ventricles using a pressure-injector (Picospritzer III, Parker Hannifin) with pulled-glass pipettes (Sutter Instrument, BF150-117-10). To target L2/3 progenitor cells in PFC for imaging and electrophysiological experiments, a custom-built triple electrode probe [37] was placed by the head of the embryo, with the negative electrodes placed near the lateral ventricles and the positive electrode placed just rostral of the developing PFC. For bilateral IUEs used in behavioral experiments, plasmid DNA (1 μg/μL) was injected into both lateral ventricles by positioning the glass pipette at a 90˚angle relative to the midline of the embryo's head and injecting 2-4 μl of DNA solution. These modifications ensured that the DNA solution would travel to the lateral ventricle contralateral to the injection site. Next, four square pulses (pulse duration: 50 ms, pulse amplitude: 36 V, interpulse interval: 500 ms) were delivered to the head of the embryo using a custom-built electroporator [66]. Embryos were regularly moistened with warmed sterile PBS during the surgical procedure. After electroporation, the embryos and uterine horn were gently placed back in the dam's abdominal cavity and the muscle and skin were sutured (using absorbable and nonabsorbable sutures, respectively). Finally, the dams were allowed to recover in a warm chamber for 1 hr and then returned to their cage.

Imaging
Fluorescence images were collected using an inverted laser scanning confocal microscope (Nikon Instruments, Nikon Eclipse Ti with C2Si + confocal) controlled by NisElements (Nikon Instruments, 4.51) including four laser lines (405, 488, 561, and 640 nm). For M-FISH and microglia experiments, confocal images were taken with a 60X Plan Apo objective (Nikon Instruments; Plan Apo, NA 1.4, WD: 0.14 mm, oil objective) using 1,024 × 1,024 pixel scans (pixel size = 0.27 × 0.27 μm). For dendritic spine imaging, images from P7-9, P14-16, P21-23, and P55-60 brain tissue were taken using a 40X Plan Apo λs objective (Nikon Instruments; Plan Apo, NA 1.3, WD: 0.2 mm, water objective) using 1,024 × 1,024 pixel scans (pixel resolution = 0.12). We imaged apical dendritic tufts in L1 and basal dendrites in L2/3. For dendritic spine imaging and neuronal reconstructions, we collected ROIs that consisted of stacks of images (approximately 20-40 optical sections, z-step = 0.3 μm and 1 μm, respectively). The area and diameter of neurons were analyzed in 40X images from the brightest z-plane of the soma. Brain sections from behavior mice were imaged using an upright wide-field microscope (Nikon Instruments, Nikon Eclipse Ni) controlled by NisElements (Nikon Instruments, 4.20) using a 10X objective (NA: 0.3, WD: 16 mm). DAPI and GFP-transfected cells in behavior brains were imaged using fluorescence filters for BFP (excitation: 370-401, dichroic: 420) and GFP HC (excitation: 470/40, dichroic: 495). Images in all conditions were collected using the same imaging conditions and exposure settings. Imaged stacks were imported in TIFF format into ImageJ (NIH). For confocal image analysis, we only measured neurons and glia in the anterior cingulate cortex and prelimbic, infralimbic, and medial orbital divisions of the mPFC. For behavior brain analysis, all GFP-positive cells were counted to quantify the extent of electroporation.

Multiplex fluorescence in situ hybridization
For M-FISH experiments, P21 and P30 brains were collected and immediately fresh-frozen on dry ice in O.C.T. compound (Fisher HealthCare, 23-730-571). Tissue was cut on a cryostat (Leica CM 1800) at 15-μm thickness, and M-FISH experiments were completed by using a commercial assay (RNAscope, Advanced Cell Diagnostics). In situ fluorescent probes were used to detect mC4 (#445161-C1), EGFP (#400281-C2), and CaMKIIα (#445231-C3). M-FISH assay and all reagents were obtained from Advanced Cell Diagnostics. All M-FISH experiments had N = 3 mice per condition. For in situ analysis, we quantified fluorescent signal from L2/3 CaMKIIα+ cell bodies of transfected GFP+ cells and their untransfected neighbors in mC4 conditions from single z-planes of 60X confocal images. This approach allowed us to control for variability of transcript expression between mice. We use the DAPI signal to identify the cell's nucleus, and all quantification of in situ signals was performed in the soma's brightest focal plane. CaMKIIα signal was used to delineate the perimeter of the cell body of L2/3 excitatory neurons in the mPFC and this ROI was used for analysis. Transfected cells were identified by presence of GFP mRNA, and nontransfected cells were GFP(−). Next, fluorescent signals from GFP and C4 transcripts were thresholded and binarized. The binarized signal was then used to calculate the percentage of soma area covered by GFP or C4 mRNA. We used the same binarization threshold and analysis procedure for both conditions. qPCR Animals at P21 or P60 were anesthetized with 4% isoflurane/oxygen (v/v). Brains were rapidly extracted following transcardiac perfusion using NMDG slicing solution: 92 mM NMDG, 2.5 mM KCl, 1.25 mM NaH 2 PO 4 , 30 mM NaHCO 3 , 20 mM HEPES, 25 mM glucose, 2 mM thiourea, 5 mM Na-ascorbate, 3 mM Na-pyruvate, 0.5 mM CaCl 2 �2H 2 O, and 10 mM MgSO 4 �7H 2 O. Sections (200 μm) were cut on a vibratome (LEICA VT1000 S) and kept in the NMDG slicing solution bubbled with 95% O 2 /5% CO 2 (295-305 mOsm). The sections were trimmed under a wide-field microscope to isolate the transfected region (containing GFP + cells) of the tissue. Each trimmed piece was flash frozen in dry ice. RNA was extracted from these tissue using Zymo Microprep kit (R1050), and cDNA was synthesized with Azuraquant cDNA synthesis kit (AZ-1997). RNA and cDNA were quantified using a Thermo Fisher Scientific NanoDrop One Spectrophotometer. Primer sets for qPCR were designed using NCBI primer blast (NIH) and were synthesized by Integrated DNA Technologies. Primer sets (Table 1) were verified to produce a single product following PCR amplification using Apex Taq RED Master Mix, 2.0X (Genesee Scientific 42-138) on a Bio-Rad T100 thermocycler. qPCR was carried out on an Applied Biosystems ABI 7900 Real Time PCR machine using AzuraView Greenfast qPCR Blue Mix HR (AZ-2401). All qPCR results were normalized to three biologically diverse internal reference genes (GAPDH, beta-Actin, and HPRT). These references were amplified in parallel with the target genes of interest and all samples were performed in triplicates. Dissociation curves were verified for every reaction, and only dissociation curves yielding single products were used in subsequent analysis.

Isolation of PSDs
PSD fractions were isolated as described with [57,67] minor modifications. Briefly, the electroporated region of the PFC was dissected and snap frozen and kept at −80˚C until further processing. Pieces of brain tissue were homogenized using a Teflon homogenizer. After 1,400g, 10-min centrifugation, the supernatant was transferred to a new tube and the pellet was rehomogenized and pelleted at 1,400g, 10 min. Supernatants were combined and centrifugated at 13,800g, 10 min. The pellet was resuspended in 0.32 M Tris-buffered sucrose and ultracentrifuged into 1.2, 1.0, 0.85 M sucrose gradient at 82,500g for 2 hr. The synaptosome fraction was collected, solubilized with 0.5% Triton X-100, and then centrifuged at 32,800g for 20 min to yield the PSD fraction. Samples from individual mice were isolated separately and pooled for immunoblot analysis.

Immunoblotting
Immunoblotting was performed as previously described [57] with minor modifications. Briefly, transfected HEK293 cells were directly lysed in the dish with 0.5% SDS in PBS, supplemented with complete protease inhibitor cocktail (Roche) and Benzonase Nuclease (Sigma). All protein samples were boiled in reducing SDS loading buffer and separated on Bolt Bis-Tris Plus gels (Thermo Fisher). After transfer to nitrocellulose membranes (Thermo Fisher), membranes were incubated with the following primary antibodies overnight: mouse anti-PSD-95 (clone K28/43, BioLegend), guinea pig anti-Synaptophysin 1 (Synaptic Systems), rabbit anti-C4 (in house, Biogen 931-946), Beta-Actin HRP-conjugate (clone AC-15, Sigma). HRP-conjugated secondary antibodies were from GE Healthcare. Immunoreactivity was detected on Che-miDoc XRS+ (Bio-Rad) and analyzed with Image Lab software. mC4 levels were quantified by normalizing to the loading control, vinculin.

Dendritic spine analysis
For the dendritic spine developmental time-course experiments in control, mC4, and hC4 conditions, we analyzed dendritic spines at multiple postnatal days (P7-9, P14-16, P21-23, P55-60). For all groups, we analyzed 20 dendrites from 7 mice (from multiple litters). Confocal image z-stacks (40X) were background subtracted and median filtered (radius 0.25 μm), and the presence or absence of a protrusion was determined by visually inspecting the entire z-stack of images. Dendrites were selected if the entire length to be analyzed was constricted within about 15 μm in the z-plane; this typically included from the tip of the branch to the bifurcation (for apical tufts). Brightest dendrites were selected from each image to ensure that dim protrusions, such as filopodia, were reliably identified. To reliably identify spines of a given dendrite, we analyzed GFP+ dendrites that were not occluded by other nearby cell processes. Dendrite selection and all following spine analysis were conducted blindly. For a dendritic protrusion to be counted, it had to clearly protrude out of the shaft by at least 3 pixels (approximately 0.36 μm). Dendritic protrusions were quantified from either apical dendritic tufts in L1 or secondary/tertiary basal dendrites in L2/3. Spine density was calculated by dividing the number of counted spines by the total dendritic length analyzed (50-80 μm long shafts). Since GFP brightness is monotonically related to the volume in each protrusion [68], we quantified the fluorescent intensity (TIB [a.u.]) of dendritic spines in apical tufts. To obtain a TIB value, the mean gray value of the dendritic spine was measured at the brightest focal plane and was divided by the mean fluorescence intensity of the adjacent dendritic shaft to normalize for varying imaging conditions. We sorted dendritic spine types into large mushroom/stubby (TIB > 75%), medium-size spines (25% � TIB � 75%), and thin-spine/filopodia (TIB < 25%) intensity groups based on percentile cutoffs determined from TIB distribution values of dendritic spines. We previously showed that this is a reliable, unbiased approach for classifying dendritic spine types [69].

Electrophysiological recordings and analysis
Mice (P18-25) were anesthetized with 4% isoflurane-oxygen mixture (v/v) and perfused intracardially with ice-cold external solution containing the following: 73 mM sucrose, 83 mM NaCl, 26.2 mM NaHCO 3 , 1 mM NaH 2 PO 4 , 22 mM glucose, 2.5 mM KCl, 3.3 mM MgSO 4 , 0.5 mM CaCl 2 and were bubbled with 95% O 2 /5% CO 2 (295-305 mOsm). Coronal slices (300-μm thickness) were cut on a VS1200 vibratome (Leica) in ice-cold external solution before being transferred to ACSF containing the following: 119 mM NaCl, 26 mM NaHCO 3 , 1.3 mM Shows forward and reverse primers used in qPCR experiments (Fig 1). . Slices were kept at 35˚C for 30 min before being allowed to recover for 30 min at room temperature. All recordings were performed at 30-32˚C. We only recorded from transfected neurons in the anterior cingulate cortex and prelimbic, infralimbic, and medial orbital divisions of the mPFC. Signals were recorded with a 5X gain, low-pass filtered at 6 kHz, and digitized at 10 kHz using a patch-clamp amplifier (Multiclamp 700B, Molecular Devices). Whole-cell voltage-clamp recordings were made using 3-5 MO pipettes filled with an internal solution that contained 125 mM Cs-gluconate, 3 mM NaCl, 8 mM CsCl, 4 mM EGTA, 4 mM MgATP, 0.3 mM NaGTP, and 10 mM HEPES (pH 7.3) with CsOH (280-290 mOsm). Series resistance (R s ) and input resistance (R in ) were monitored throughout the experiment by measuring the capacitive transient and steady-state deflection in response to a −5 mV test pulse, respectively. For mEPSC recordings, cells were voltage clamped at E rev GABA A (−70 mV) in the presence of 1 μM Tetrodotoxin (Tocris). For mIPSC recordings, cells were voltage clamped at E rev Glu (+5 mV) in the presence of 1 μM Tetrodotoxin (Tocris). For all groups, we analyzed 10-12 cells from 4 mice (from multiple litters). mPSCs were detected by fitting to mPSC amplitude template using pClamp10 analysis software (Molecular Devices). The peak current of each mPSC was calculated. Next, average mPSC amplitude and IEI was calculated for each cell and then averaged across each condition to determine the population mean and SEM. Bath application of 100 μM GABA A inhibitor picrotoxin at E rev Glu and 20 mM CNQX and 50 mM DL-AP5 at E rev GABA A eliminated mPSCs, thus confirming the identity of the recorded currents. Cells were excluded if R s varied by more than 20% during a recording. For all recordings, series resistance was close to 10 MO (control, 9.39 ± 0.53 MO, n = 12; mC4, 9.74 ± 0.44 MO, n = 10; p > 0.05) and was not compensated.
In current-clamp recordings, CNQX (20 mM), 50 mM DL-AP5, and picrotoxin (100 μm) were routinely added to the extracellular solution to block ionotropic synaptic transmission mediated by glutamate and GABA A receptors, respectively, to assess persistent firing. Neuronal excitability was assessed using the input-output curve measured from the changes in membrane potential (presence of APs) evoked by current steps (from V rest , start = −200 pA, step duration: 300 ms), increasing in increments of 10-15 pA in current-clamp mode. C m and R m were calculated in seal-test configuration from the decay and steady-state current of a transient generated in response to a −5 mV pulse test. Rheobase is the minimum current amplitude (300 ms) that resulted in an AP.

Neuronal morphology analysis
Confocal images of L2/3 GFP-positive neurons in mPFC (P21 and P60) were collected using a 40X objective. z-Stacks for images with a large field of view were taken at z-step of 1 μm through the entire section (about 200-μm thickness), taking care to include the entirety of the dendritic tree. Image stacks of dendritic trees and soma were imported in TIFF format into ImageJ (NIH). Cell somas and dendrites were traced manually to ensure accurate reconstruction. Dendritic arbors that could not be confidently and completely reconstructed were not used in the analysis. Dendritic reconstructions were confirmed by comparing independent reconstructions from two investigators (N = 10 neurons per condition). The dendritic parameters analyzed included the total dendritic length (of all dendrite branches), number of dendritic branches, number of branch points, number of dendritic end tips. Branch order for each dendritic branch was assigned starting at the cell body and increased after each branch point and the maximum branch order quantified. Sholl analysis was performed using Simple Neurite Tracer (SNT, ImageJ plugin, Longair MH, 2011). Each reconstructed neuron was thresholded and binarized, and the skeletons were imported to SNT. Sholl analysis was performed by calculating the number of dendrites that intersected concentric spheres that radiated from the soma in 10-μm radius increments. On some occasions, the dendritic arbor could not be reconstructed for several reasons: dendritic processes extended outside of the acquired field of view (or z-stack); the field of view became too dense with GFP-positive processes from neighboring cells, preventing unambiguous reconstruction; and some dendritic processes were deemed insufficiently bright to allow for unequivocal reconstruction. Thus, our reconstructions are an underestimate of the entire dendritic arbor. Neuronal soma area and diameter were measured from a single z-plane for each neuron. Soma diameter was measured perpendicular to the apical axis. Analysis was performed at P21 in the mPFC and 315 neurons were analyzed for control (6 ROIs from 3 mice) and 216 cells were analyzed for the mC4 condition (7 ROIs from 3 mice).

Microglia assays
Colocalization of GFP and Iba1. GFP colocalization with microglia (Iba1) was quantified at P21. Confocal images of mPFC (60X objective) were collected using the same exposure settings between conditions including channels for DAPI, GFP (transfected neurons), and Iba1 (microglia). All images were background subtracted and binarized using the same settings between experimental groups. Analysis was completed for single z-planes. To quantify microglia proximity to GFP+ neurons, we first defined the electroporated region (ROI) to be analyzed. We created an ROI for a single z-plane in the mPFC where transfected neurons were located. There were 26 ROIs for each condition. The ROIs of electroporated regions were, on average, 98,000 μm 2 (approximately 280 μm × 350 μm) and contained, on average, 13.6 microglia per ROI. We then identified microglia by tracing around the perimeter of the cell (using Iba1 signal), including the cell body and any proximal processes in the single analyzed z-plane. Microglia cell bodies were confirmed by the presence of DAPI. Colocalized GFP and Iba1 signal was isolated and quantified within the microglia by creating a thresholded mask. Colocalized signal within the microglia mask was only included in the analysis if it was equal to or larger than 1.23 μm (approximately 3 pixels). GFP-positive puncta within microglia were visually inspected in three-dimensional space by examining multiple z-planes to confirm the containment of the puncta to the microglia cytoplasm. The percentage of colocalized signal was quantified by dividing the area of the mask by the total area of the microglia (GFP/Iba1 colocalization [%] = microglia area colocalized with GFP / total microglia area). Using this analysis, we also calculated the percentage of microglia that were positive for GFP (GFP-positive microglia [%] = total number of GFP-positive microglia / total number of microglia), N = 26 ROIs (ROIs of electroporated regions from 3 mice per condition including 373 control and 334 mC4 microglia).
Synaptic engulfment and cortical depth correlation. To quantify engulfment of synaptic material in nonexpanded tissue, we electroporated (E16) a plasmid that labels endogenous PSD-95 (EF1a-PSD95.FingR-RFP) for control and mC4 conditions. Analysis was completed for single z-planes using DAPI, PSD-95-FingR-RFP, Iba1, and CD68 (a lysosomal marker). ROIs of electroporated regions were defined as described in the previous section (Colocalization of GFP and Iba1). Analysis was completed using the same method as in the GFP and Iba1 colocalization analysis, except synaptic material was only counted if it was within the lysosome of the microglia (PSD-95-postive puncta that colocalized with both Iba1 and CD68 signals). For each thresholded mask, the total colocalized area (microglia engulfment area [%] = microglia area colocalized with PSD-95 and CD68 / total microglia area) was quantified. We also quantified the percentage of microglia positive for engulfment (engulfment-positive microglia = [%] total number of positive microglia /total number of microglia). We also quantified the colocalization of fluorescent signals from PSD-95, Iba1, and CD68 in confocal images that were pixel shifted randomly by 12 μm in four possible directions for each channel, independently. The percent area of each microglia colocalized with CD68 signal was calculated as a measure of microglia reactivity (area of CD68 / microglia area). In a separate analysis, the cortical depth of each engulfment-positive microglia was determined by measuring the distance from the center of the cell body (using DAPI) to the adjacent pia mater. We restricted cortical depth analysis to L1 and L2/3 of the mPFC, where the dendritic spines of L2/3 cortical neurons are localized (depth � 300 μm). Microglia that did not contain triple-stained puncta (PSD-95/ Iba1/CD68) were excluded from correlation analysis. Control: N = 26 ROIs (from 5 mice; 345 microglia); mC4: N = 26 ROIs (from 5 mice; 319 microglia).
Microglia density. Microglia density analysis (P21) was completed using 50-μm z-stack images using DAPI and Iba1 signals. Microglia density was independently calculated for L1 (cortical depth: 0-120 μm) and L2/3 (cortical depth: >120-300 μm) of the mPFC. Density was calculated per ROI (electroporated region) by dividing the number of microglia in the ROI by the total volume of tissue for each ROI. Microglia density was measured in 19 control ROIs (from 5 mice, 2,146 microglia) and 17 mC4 ROIs (from 5 mice, 1,640 microglia) in the superficial layers of cortex, and in each ROI the density was calculated for L1 and L2/3. Microglia cell bodies were confirmed by the presence of DAPI.

ExM
Coronal mouse brain sections (50 μm) were permeabilized as previously described and stained at 4˚C for 24 hr with primary antibodies (rat anti-CD68, rabbit anti-RFP, guinea pig anti-Iba1) at a dilution of 1:500 for each. Tissue was subsequently washed three times for 10 min in 1X PBS with 0.025% Triton-100X. The tissue was then incubated with secondary antibodies (antirat 488, anti-rabbit Alexa 546, anti-rabbit 594, anti-guinea pig Atto 647, anti-guinea pig Alexa 647) at 4˚C for 24 hr. Tissue sections were again washed three times with PBS for 10 min. Sections were then anchored in 0.1 mg/ml Acryloyl-X SE in 1X PBS for 12 hr and polymerized within a polyacrylamide and sodium acrylate gel [70]. The gels were trimmed and subsequently digested with proteinase k (8 units/ml) for 8-10 hr. Following digestion, the sections were transferred into a #1.5 glass-bottom dish, washed three times with 1X PBS, and expanded with milliQ purified water for 15 min twice. Fluorescence images of the expanded tissue were acquired with a Nikon C2Si inverted laser scanning confocal microscope (Nikon Instruments, Nikon Eclipse Ti with C2Si+ confocal) with a 40X Plan Apo λs objective (Nikon Instruments; Plan Apo, NA 1.3, WD: 0.2 mm, water objective). Images of microglia near apical dendrites of L2/3 FingR-PSD95-RFP(+) neurons were acquired with a pixel sampling of (0.155 μm/pixel) and 1-μm z-axis step interval.
The scaling factor for each expanded microglia was determined by comparing pre-expansion measurements of the long and short cross sections of the microglial cell body in a maximum-intensity projection image, and these values were compared to expanded microglia measurements. The scaling factor for each microglia was calculated based on pre-expansion images. This scaling factor was applied to any measurements made within that expanded image (e,g., lysosomes area) to calculate the original size of the organelles. To determine the number and size of lysosomes, ROIs were manually drawn following the perimeter of the lysosome using the z-axis plane with the brightest CD68 signal. This z-plane was determined first qualitatively and confirmed by a line scan through the approximate center of the lysosome signal, which yielded the largest slope of intensity versus distance. The lysosome ROIs were counted and the cross-sectional areas measured and compared between experimental groups. CD68-positive signals were scored as a lysosome if they were 3 × 3 pixels, or 7-9 clustered pixels (0.216 μm 2 expanded), in size with an average signal intensity of at least two times above local background levels and were present in two consecutive z-planes.
Additionally, each lysosomal ROI was verified for FingR-PSD95-RFP signal using a z-axis profile plot for both CD68 and RFP signal. ROIs that show overlapping peaks between RFP and CD68 were verified for signal within the ROI in comparison to the surrounding local background. Local background was determined by calculating the mean background signal of RFP in a region surrounding each ROI. We considered lysosomes positive for synaptic engulfment if the ROI signal increased by two standard deviations over the local background signal. PSD95-FIngR-RFP-positive signal was scored as a PSD-95 internalized puncta if they were 2 × 2 pixels, or 4-5 clustered pixels (0.0961 μm 2 expanded), in size with an average signal intensity of at least two times above local background levels. The data were derived from 5 and 4 mice each for control and mC4 conditions, respectively, and 45 and 41 microglia were analyzed from control and mC4 conditions, respectively. We analyzed a total of 1,987 and 2,531 lysosomes from control and mC4 conditions, respectively.

MI task
For all juvenile behavior tasks, we used control (N = 15) and mC4 (N = 21) mice that were electroporated bilaterally at E16. Prior to the maternal homing test, dams were acclimated to the behavioral task by placing them in the mesh cup for 5-min periods, for 3 d consecutively. Mice were separated from their dams for 1 hr immediately before testing. The MI task consisted of two phases. In phase 1 (MI1), mice were placed in an OF with home bedding and fresh bedding, and in phase 2 (MI2) mice were placed in an OF with home bedding, and their dam was restrained in a small wire mesh cup and an empty wire mesh cup. Both phases of the behavioral task were run on P18.
For MI1, individual pups were transferred to a homemade acrylic arena (50 × 50 × 30 cm length-width-height) and placed in a "starting corner." The arena contained fresh bedding in two neutral corners and nest bedding in the corner opposite from the starting corner. Mice explored for 3 min, and the total time spent in the starting, nest, and fresh corners was measured using a homemade video tracking system written in MATLAB (MathWorks). Grooming occurrences and time spent grooming were annotated by a trained experimenter blind to experimental conditions. For MI2, two wire mesh cups were placed in opposite corners of a homemade acrylic arena (50 × 50 × 30 cm length-width-height), one containing the animal's dam and the other empty, and the start corner contained soiled home bedding. Mice spent 5 min exploring the MI2 environment, and we video-recorded behavior using a Logitech C270 Webcam at 30 fps. Time spent near each cup and other areas was measured using our homemade video tracking system that tracked the centroid of the mouse. Behaviors were recorded under a dim light (approximately 20 lux) positioned over the center zone of the arena. Between trials and mice, the arena was cleaned with 70% ethanol.

Adult behavior
For all adult behavior tasks, we used control (N = 22) and mC4 (N = 20) mice that were electroporated bilaterally at E16. The same cohort of adult mice were run in all adult behavioral tasks between the ages of P60-70. All behavior was recorded using a Logitech C270 Webcam at 20 fps. All mice were handled for 3 d consecutively prior to any behavioral testing to ensure familiarity to the experimenter. For all behavioral tasks, experiments were recorded under a dim light (approximately 20 lux) positioned over the center zone of the arena. For OF, mice were placed in a homemade acrylic arena (50 × 50 × 30 cm length-width-height) and were free to explore for 5 min. For EZM, mice were placed in the closed arm of a homemade EZM (track diameter = 50 cm, track width = 5 cm, wall height for closed arms = 40 cm, height of track = 61 cm) and were free to explore for 5 min. For OF and EZM, mice were not exposed to the arenas until testing day, since the initial response to the environment is measured. Sociability and object tasks were run in the OF arena. Mice were acclimated to the OF arena for 3 d prior to behavioral testing (5 min each day). For the novel-object interaction task, mice were placed in the center of the arena that contained a novel object (small plastic toy with smooth, cleanable surfaces) in one corner (corner alternated between mice). Mice were free to explore the arena for 5 min. For novel-object recognition, the arena contained a familiar object (object exposed in novel-object task) in one corner and a novel object in the opposite corner (corners alternated between mice), and mice were free to explore for 5 min. For the sociability task, the arena contained a novel mouse (age, sex, and strain matched) under a mesh wire cup in one corner and an empty mesh wire cup in the opposing corner, and mice were free to explore for 5 min. For all behavioral tasks, the arena was cleaned with 70% ethanol between all trials.
All adult behavior was analyzed using DeepLabCut [71], an open-source software package that uses deep neural networks to automatically track body parts from videos. We confirmed accurate tracking of mice based on this software by close inspection of videos once they had been annotated by DeepLabCut. For all adult behavior position tracking, we used the centroid of the mouse and compared amount of time spent in ROIs. For OF, we calculated the maximum and average velocity and the total distance traveled. For EZM, we compared the time spent in the open arms of the maze. For both novel-object and sociability tasks, we quantified amount of time spent in the corners of interest in the arena (time spent with novel object for novel-object interaction, time spent with novel versus familiar object for novel-object recognition, time spent with novel mouse versus empty cup for sociability).

Quantification of behavior brains
We counted the number of GFP-positive cells and assessed the extent of cell transfection in brains of animals that were tested in behavioral tasks. Experimental groups included mice from at least two litters (N = 15 mice for control and N = 21 for mC4 condition for juvenile behaviors; N = 22 mice for control and N = 20 for mC4 condition for adult behaviors). Somas were counted by multiple trained, independent experimenters. Coronal sections (50-μm thickness) from the entire brain were carefully inspected, and the DAPI signal was used to corroborate the presence of cell nuclei. We used a brain atlas [72] to confirm the location of GFP+ cell bodies and to delineate the boundaries of each brain region. Most transfected cells were located in frontal cortical regions, between Bregma +3.0 mm and 1.6 mm. We did not observe transfected cells in caudal cortical regions or subcortical areas.
To estimate the total amount of cells transfected in each brain, cells were counted from 50μm sections, including every other section. Since transfections were homogenous across neighboring brain sections, once we obtained the total number of cells per section, we plotted cell counts against their respective Bregma coordinates. We then used a linear interpolation to estimate the cell counts from missing brain sections. In a small number of brains (N = 3), we found no differences in total cell counts when we compared values obtained from counting GFP-positive cells in all transfected brain sections to estimated values obtained using the interpolation approach. In total, including juvenile and adult brains, we counted 174,041 neurons from 78 brains. All control and mC4 brains from mice used in behavioral tasks were GFP + and had similar rostro-caudal patterns of transfected neurons. We did not find any differences in total cell counts between groups for juvenile or adult conditions.

Statistical analysis
For confocal image analysis and electrophysiological recordings, we focused on neurons in the anterior cingulate cortex and prelimbic, infralimbic, and medial orbital divisions of the mPFC. All statistical analysis was completed in GraphPad Prism 8.0, and threshold for significance for all tests was set to 0.05 (α = 0.05). M-FISH and qPCR data sets were analyzed with an unpaired t test. Spine developmental data were analyzed using a one-or two-way ANOVA followed by Tukey's posttest. Dendritic spine fluorescent intensity values were sorted based off of percentile cutoffs determined from TIB distribution values of dendritic spines in control conditions, and differences in the density of spine types between groups were tested with a one-way ANOVA followed by Tukey's posttest. Electrophysiological data were analyzed with an unpaired t test, and cumulative distributions were analyzed with a Kolmogorov-Smirnov (KS) test. Microglia data were analyzed using either an unpaired t test or a one-way ANOVA followed by Tukey's posttest. Behavioral data were analyzed with a two-way ANOVA followed by Sidak's posttest and a t test with Welch's correction. Dendrite and soma morphology measurements were analyzed using a t test. Both male and female mice were used, and we did not observe any differences between the sexes among groups. Correlations were determined using Pearson's r correlation and linear regression. Analysis was performed blind to condition. Soma area/diameter and dendrite measurements were analyzed using GraphPad Prism 8.0 (GraphPad Software). Figures were prepared using CorelDRAW Graphics Suite X8 (Corel Corporation) and ImageJ (NIH). Custom-written routines for behavioral tracking and analysis are available upon request. Data are presented as the mean ± SEM, unless otherwise noted. staining with anti-C4 antibody (clone 931-946). Samples were from HEK293 cells transfected with GFP, hC4A, mC4, or mC4-GFP constructs. We detected C4 variants as approximately 250-kDa proteins, which is likely the unprocessed protein (predicted molecular weight is 193 kDa) or not fully reduced C4 protein or not fully reduced C4 proteint eurons? Also did you ever compare neuronal expression vs astrocytes vs microglia?ssion in WT a (C4 has disulfide bonds). (B) Fractionation scheme for the preparation of PSDs from mouse PFC region. Fractions that were used for immunoblot analysis are shown in bold. (C) Immunoblot of postsynaptic (PSD-95) and presynaptic (synaptophysin 1) marker proteins in PSD isolation steps. Synaptosome fraction contains both PSD-95 and synaptophysin 1. For underlying data, see https://osf.io/7em3s/?view_only=0e7ffde4ebd344dc83af83b5a605c451. GFP, green fluorescent protein; hC4A, human C4A; HEK, human epithelial kidney; mC4, mouse C4; PFC, prefrontal cortex; PSD, postsynaptic density. . Inset image of microglia shows the long (E) and short (F) diameter of soma (white line with arrows). The coefficient of variation between pre-expansion, post-expansion, and post-expansion scaled microglia sizes was 30.24%, 34.35%, and 29.22%, respectively. The variance between the pre-expansion and postexpansion scaled sizes was compared with an F-test, and no difference was found between these populations (one-way ANOVA with Tukey's test; p = 0.4212). Mean ± SEM. For underlying data, see https://osf.io/7em3s/?view_only=0e7ffde4ebd344dc83af83b5a605c451. ExM, expansion microscopy; Iba, ionized calcium binding adaptor molecule 1; PFC, prefrontal cortex. (TIF)