Disrupted-in-Schizophrenia (DISC1) Functions Presynaptically at Glutamatergic Synapses

The pathophysiology of schizophrenia is believed to involve defects in synaptic transmission, and the function of many schizophrenia-associated genes, including DISC1, have been linked to synaptic function at glutamatergic synapses. Here we develop a rodent model via in utero electroporation to assay the presynaptic function of DISC1 at glutamatergic synapses. We used a combination of mosaic transgene expression, RNAi knockdown and optogenetics to restrict both genetic manipulation and synaptic stimulation of glutamatergic neurons presynaptic to other layer 2/3 neocortical pyramidal neurons that were then targeted for whole-cell patch-clamp recording. We show that expression of the DISC1 c-terminal truncation variant that is associated with Schizophrenia alters the frequency of mEPSCs and the kinetics of evoked glutamate release. In addition, we show that expression level of DISC1 is correlated with the probability of glutamate release such that increased DISC1 expression results in paired-pulse depression and RNAi knockdown of DISC1 produces paired-pulse facilitation. Overall, our results support a direct presynaptic function for the schizophrenia-associated gene, DISC1.


Introduction
DISC1 was identified as a schizophrenia susceptibility gene because a chromosomal translocation that results in a c-terminal truncation of the DISC1 gene was found to co-segregate with major mental illness in an extended Scottish pedigree [1,2]. Studies have shown that DISC1 is a scaffolding protein with a variety of functions during all aspects of neural development [3,4]. RNAi knockdown of DISC1 during early neocortical and hippocampal development resulted in several phenotypes including the disruption of neurogenesis, migration, altered dendritic arborization, and the density of spines [5,6,7,8,9]. Transgenic animals expressing a truncated version of DISC1 (DISC1DC) under control of the CaMKII promoter were shown to display abnormal behavioral phenotypes, enlarged ventricles, decreased levels of cortical dopamine, fewer parvalbumin-positive neurons and altered spine density [10,11]. These studies provide strong evidence for DISC1 having important roles in postsynaptic physiology and structure, however evidence also exists that suggest DISC1 has important presynaptic functions. DISC1 immunoreactivity is observed at the ultrastructural level in presynaptic terminals [12,13]. Both acute RNAi knockdown and a knockin mouse that creates a truncating lesion in DISC1 resulted in altered axonal targeting of mossy fibers [14,15]. The knockin mouse also produced changes in short-term plasticity at the mossy fiber/CA3 synapse [15]. Moreover, RNAi knockdown of DISC1 disrupts the transport of synaptic vesicles and mitochondria [16,17], two cellular organelles important for synaptic transmission [18]. Here we use optogenetics and wholecell electrophysiology to specifically test for a presynaptic function of DISC1 in cortical layer 2/3 pyramidal neurons. We show expression of DISC1DC enhances mEPSC frequency and alters the kinetics of the evoked glutamate transient. In addition, we show the expression level of DISC1 in presynaptic neurons regulates the probability of glutamate release. Overall, our data provide several lines of evidence that suggest DISC1 has direct functions in presynaptic transmission.

Ethics Statement
All studies were conducted in accordance with protocols that were approved by the University of Connecticut Institutional Animal Care and Use Committee (IACUC; Assurance No. A09-025, 2/2011). The facilities at the University of Connecticut are accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care (AAALAC).

In utero electroporation and inducible plasmid expression
In utero electroporation was performed on Wistar rats as previously described [19]. For inducible expression we used a 4-OHT-activatible form of Cre recombinase (pCAG-ER T2 -CreER T2 , 1.5 ug/mL) and Cre-dependent inducible expression vectors (pCALNL-dsRed, pCALNL-GFP, pCALNL-wtDISC1, and pCALNL-DISC1DC, 1.5 mg/mL; a gift from T. Matsuda and C. Cepko, Harvard Medical School, Boston). Both DISC1 and DISC1DC constructs were gifts from A. Kamiya and A. Sawa, John Hopkins School of Medicine, Balitmore, MD and were subcloned into the pCALNL plasmid for this study. The D1 shRNA (pUEG-D1) was a gift from H. Song, John Hopkins School of Medicine, Baltimore, MD. Channelrhodophsin plasmid (pcDNA3.1hChR2-EYFP) was a gift from K Diesseroth, Stanford University, Stanford, CA, and was subcloned into the pCAG plasmid for this study.
The pCALNL-DISC1DC-GFP and pCAG-wtDISC1-GFP was made by fusing eGFP to the C-terminus. DISC1DC-GFP was only used in Figure 1A-B to demonstrate protein localization and wtDISC1-GFP was only used for RNAi rescue in Figure 3 D-F, all other experiments lacked the GFP fusion.

4-OHT administration
4-OHT administration was performed as described by Manent et al., 2009. Briefly, 4-hydroxytamoxifen (4-OHT; Sigma) was dissolved in 95% ethanol at a concentration of 20 mg/ml and diluted in 9 volumes of corn oil. Diluted 4-OHT (2 mg/100 g body weight) was administered to the animals via i.p. injection on P5 and P7. Vehicle-treated animals were injected with the same solution without 4-OHT.

Histological procedures and microscopy
Animals (p28) were transcardially perfused under deep anesthesia with 4% paraformaldehyde in PBS. Brains were removed and postfixed 24 hours in the same fixative solution, prior to coronal sectioning with a vibratome (Leica, Nussloch, Germany) Brain sections were processed for immunohistochemistry as floating sections. Primary antibody was goat-anti-GFP (1:1000, Molecular Probes) and secondary antibody was rabbit-anti-goat conjugated with Alexa 488 (1:200, Molecular Probes). Photomicrographs were taken with a Leica TCS SP2 confocal microscope (Nussloch, Germany) and Zeiss Axio Imager 2 with Zeiss ApoTome module. For spine counting, the primary basal dendrites of layer 2/3 pyramidal cells from at least three animals in each condition were imaged. All counting was done blind using Neurolucida (MBF Bioscience, Williston, VT) for analysis. For axon arborization measurements were performed as previously described [21].
Except, to compensate for variability in the efficacy of labeling callosal axons with GFP, the densiometric line scans were normalized by the average GFP intensity measured just above the white matter tract below the area of interest. To analyze the number of presynaptic active zones per length of axon we co-transfected layer 2/3 neurons with synpatophysin-RFP (pCAG-syp-RFP), pCAG-ER T2 -CreER T2 , pCALNL-GFP (control) or pCALNL-GFP plus pCALNL-DISC1DC. pCAG-syp-RFP was subcloned from pTRE-Bi-SG-T (Addgene plasmid 26084 [22]), fused with RFP and placed behind the pCAG promoter. NIH Image J Software was used to measure the length of contralateral axons and syp-RFP positive puncta were manually counted. All counts were done blind and at least three animals per condition were used.

Light-activated synaptic transmission
ChR2-venus was activated by 2 ms pulses of blue light (473 nM; ,1 mW) from a 20 mW laser (Dragon Lasers, China) attached to a fiber optic cable. The end of the fiber optic cable was attached to a ceramic patch pipette holder and manipulator. The tip of the fiber optic cable was submerged into the bath above the brain slice. Light-evoked EPSC amplitudes were monitored as the fiber optic was moved until the maximum evoked amplitude was achieved. Gabazine (5 mM) was then washed into the bath to block feedforward inhibition and the underlying EPSC was revealed.

Data Analysis and Statistics
We used Axograph on a Macintosh computer for analysis. For analysis of EPSC kinetics the rise time equals the duration of time between 10% and 90% of the maximum peak amplitude. EPSC peak location equals the duration of time between the EPSC onset (5% of the peak amplitude) and the maximum peak amplitude. Weighted decay equals the total charge from the peak of the response back to baseline divided by the peak amplitude. EPSC charge was measured for 50 ms following the EPSC onset. EPSC rise time and peak location was measured from normalized EPSCs  that were the average of 10 consecutive sweeps. For all experiments, statistical significance was determined using standard t-tests, 1-way ANOVA with Student Newman-Keuls post-hoc test. All statistical significance is indicated on the figures with asterisks.
Averaged data values are reported as mean 6 SEM.

Results
In order to determine whether DISC1 regulates synaptic transmission we used in utero electroporation to alter DISC1 expression in approximately 20% of neocortical layer 2/3 pyramidal neurons by conditionally expressing full-length DISC1 (wtDISC1), DISC1DC, or constitutive expression of an shRNA previously shown to create effective RNAi knockdown of DISC1 (D1 RNAi [6,14,20]; see methods). In utero electroporation produces high fidelity co-transfection of multiple plasmids and reliable inducible expression with no leaky expression in the absence of tamoxifen ( Figure 1A [19]). Conditional expression of DISC1DC on P5 is after neurogenesis and neuronal migration is complete and therefore does not result in early developmental disruptions as previously reported for embryonic expression [5]. To identify the cellular localization of our inducible DISC1DC construct we fused it with GFP (pCAG-DISC1DC-GFP) and observed expression in the soma, dendrites, axons ( Figure 1B), and axon terminals of layer 2/3 pyramidal neurons ( Figure 1C). This distribution throughout the neuron is consistent with previous studies showing nearly ubiquitous distribution of DISC1 [12].
We next used whole-cell patch clamp recording to characterize effects of DISC1DC expression on the electrophysiology of neocortical pyramidal neurons. We found no significant differences in the intrinsic membrane properties, including resting membrane potential, input resistance and spike firing rates in neurons expressing DISC1DC, wtDISC1, D1 RNAi or GFP ( Figure S1). However, the frequency of miniature excitatory synaptic currents (mEPSCs) mediated by glutamatergic synaptic activity was nearly doubled by DISC1DC expression (Figure 1C,D; control transfected 1.5860.31 (n = 9), control untransfected 1.7960.32 (n = 5), DISC1DC transfected 3.9060.96 (n = 10), DISC1DC untransfected 4.4960.7 (n = 6), wtDISC1 transfected 1.3760.29 (n = 7); ANOVA p,0.002). This increase in synaptic activity was present to the same extent in both transfected cells expressing DISC1DC and neighboring non-transfected cells and indicates presynaptic expression of DISC1DC is sufficient to explain the increase in mEPSC frequency. DISC1DC expression had no significant on mEPSC amplitudes, rise times, or decays, compared to wtDISC1 or GFP controls (but see D1 RNAi; Figure  S2A-C). Furthermore, we observed no significant difference in the density of spines, complexity of contralateral axonal projections, or the density of presynaptic terminals as visualized with a synaptophysin-RFP (syp-RFP) fusion protein ( Figure S2D,E). Together, these results suggest the DISC1DC-dependent enhancement of mEPSC frequency is not due to increases in synapse number, but rather reflects an alteration in presynaptic function.
To directly test the presynaptic function of DISC1 we combined our genetic manipulations with co-expression of channelrhopsin (ChR2). We then stimulated this transfected population of neurons with a laser pulse and recorded evoked excitatory postsynaptic currents (EPSCs) from untransfected layer 2/3 neurons ( Figure  S4B). Any altered synaptic transmission recorded in untransfected neurons would be due to manipulation of DISC1 in the photoactivated presynaptic neurons. In all conditions, laser stimulation reliably evoked single EPSCs that were completely blocked by TTX or NBQX ( Figure S4B1), confirming that the light-stimulated population of neurons was indeed presynaptic to the untransfected recorded cells.
We next assayed how expression of our DISC1 constructs would affect the light-evoked glutamate release by monitoring EPSC kinetics. Presynaptic stimulation of neurons expressing DISC1DC consistently produced unusual multiphasic EPSCs that exhibited significantly slower rise-times and significantly delayed EPSC peaks compared to GFP controls, wtDISC1 and D1 RNAi (Figure 2A We did not observe a significant difference in the total EPSC charge between conditions ( Figure 2B; ANOVA p = 0.24) suggesting that the total number of vesicles released is unchanged by DISC1DC expression.
DISC1DC-dependent slowing of glutamate release was not due to an effect on action potential generation. Using cell attached recordings from either DISC1DC or GFP transfected neurons, we did not observe a difference in the trial-to-trial temporal jitter of action potentials that were generated by light activation of ChR2. Recordings from GFP (n = 9) or DISC1DC (n = 8) transfected neurons showed less than a 50 ms variation in the to time to action potential peak from trial-to-trial. This temporal variation is significantly less than what is observed for the DISC1DCdependent slowing of EPSC kinetics (approx. 2 ms).
One possibility for slowed kinetics observed in averaged synaptic responses, as measured above, is that presynaptic expression of DISC1DC may increase trial-to-trial variability in the EPSC waveform. To determine and temporally map the possible change in variability in response, we normalized 10 consecutive responses to their peak amplitude for each recording from control and DISC1DC transfection conditions and computed the average trial-to-trial variance over the duration of the EPSC waveform ( Figure 2C). The averaged time-resolved variances for responses for all recordings show a significant increase in the trialto-trial variance across time for DISC1DC (n = 34) compared to control (Figure 2d; (n = 24)). Moreover, the difference in variance is maximally different during the rise-time and decay of the EPSC ( Figure 2E) suggesting further a desynchronization in vesicle release across the entire EPSC waveform. These results together indicate that presynaptic expression of DISC1DC inhibits the synchronous nature of vesicle release, an effect distinct from the effects of DISC1 knockdown or DISC1 overexpression and confirms previous reports that DISC1DC acts as a dominant negative construct [5].
The specific presynaptic mechanisms that underlie the effects of DISC1DC are unknown at this time, however they are not related to measurable changes in intrinsic membrane properties, presynaptic excitability ( Figure S1) or axonal structure ( Figure S3). Many central synapses display asynchronous release or delayed release that is observed during periods of high frequency stimulation, and this property of asynchronous release is believed to be separate from synchronous release [24]. DISC1DC-dependent slowing of EPSC kinetics observed here does not appear to be associated with changes in this type of asynchronous release, as we did not observe a slow build-up of charge during high frequency stimulation nor did we observe an increase in spontaneous EPSC (sEPSC) frequency following high frequency stimulation ( Figure S5). This suggests that the DISC1DC effect on desynchronizing release is something distinct from asynchronous release.
We next asked if the expression level of DISC1 could regulate the probability of release by measuring paired-pulse ratios (PPR) and the coefficient of variation (CV). Light-evoked release from control neurons expressing GFP at 50 ms interval on average resulted in a slight paired-pulse facilitation (PPF; control, Figure  . The abnormal EPSC kinetics and increased trial-to-trial variability observed when DISC1DC was expressed precluded our ability to use peak amplitudes to analyze the probability of release from this condition. Together, these results indicate the level of DISC1 expression in presynaptic neurons regulates the probability of transmitter release such that overexpression of DISC1 enhances release probability and decreased DISC1 expression lowers the release probability.

Discussion
We provide several lines of evidence showing DISC1 regulates glutamate release from presynaptic terminals. We show expression of DISC1DC enhances the frequency of mEPSCs and disrupts the synchronous nature of evoked glutamate release. Furthermore, we show the expression level of DISC1 in presynaptic neurons is correlated with the probability of glutamate release. Our results suggest RNAi knockdown of DISC1 produces different effects from those seen with expression of DISC1DC, and indicates DISC1DC acts as a dominant negative as suggested by others [5,10].
Understanding the function of DISC1DC is relevant to schizophrenia not only because of the chromosomal translocation segregates with mental illness in the Scottish pedigree [1,2], but also because several alternative splice variants of DISC1 were found to have higher expression in patients with schizophrenia [25]. Expression of DISC1DC under the control of the CaMKII promoter in a transgenic mouse line resulted in several phenotypes related to schizophrenia including enlarged lateral ventricles, reduction in parvalbumin immunoreactivity, reduced cortical dopamine levels and behavioral abnormalities [10]. Another mouse model that more closely models the human translocation by introducing a truncating lesion in the endogenous murine Disc1 ortholog showed several presynaptic phenotypes including, abnormal axonal targeting in hippocampus, altered short-term synaptic plasticity, decreased volume of synaptic vesicles, and elevated cAMP levels [15]. DISC1DC may alter DISC1 function through its interaction with full-length DISC1. In cell models, truncated DISC1 was shown to form dimers with wild-type DISC1 that resulted in abnormal microtubule dynamics and defects in neuronal migration [5]. We show that postnatal expression of DISC1DC results in an enhancement of mEPSC frequency in both transfected and neighboring untransfected neurons, suggesting either an enhancement in structural synaptic connectivity or an alteration in spontaneous vesicle fusion, or both. We therefore assessed whether DISC1DC changed morphological measures of connectivity in cortex. We compared dendritic spine densitities, axonal arborization, and the density of presynaptic active zones labelled by synaptophysin-mRFP fusion (syp-RFP) between control and DISC1DC expressing neurons, and found no significant evidence for DISC1DC altering any of these morphological measures of connectivity ( Figure S5). This lack of effect on connectivity mirrors those obtained from a transgenic mouse model in which expression of truncated DISC1 was induced postnatally and no effect on spine density was observed [11]. Therefore these results suggest that the enhancement of mEPSC frequency by DISC1DC, are due to functional changes in transmitter release that are largely independent of changes in axonal sprouting or changes in spine number.
To further investigate the presynaptic function of DISC1 we utilized optogenetics to specifically stimulate presynaptic neurons expressing our DISC1 constructs. This analysis produced several distinct presynaptic phenotypes including effects on the kinetics of evoked glutamate release and the probability of glutamate release. Similar to the effects on mEPSC frequency, the kinetics of glutamate release was only altered by expression of DISC1DC. We observed that expression of DISC1DC increased the trial-to-trial variance over the duration of the EPSC waveform. One explanation for this result is that DISC1DC disrupts synchronous vesicle fusion normally observed at these synapses. The molecular mechanism responsible for this effect is currently unknown. However, an intriguing candidate mechanism involves the major Ca 2+ -sensor for vesicle fusion, synpatotagmin.
DISC1 interacts indirectly with synaptotagmin through an interaction with FEZ-1, and expression of DISC1DC was shown to attenuate synaptic vesicle transport in primary cortical neuronal cultures [16]. Genetic deletion of synaptotagmin results in a complete loss of synchronous release, dramatically enhances spontaneous vesicle fusion, and has very little effect on asynchronous release [24,26,27,28]. These synaptotagmin-dependent effects on synchronous release are strikingly similar to our results observed by overexpression of DISC1DC, whereby DISC1DC appears to inhibit synchronous release in a dominant negative manner while also enhancing the frequency of spontaneous vesicle fusion. Future experiments directed at FEZ-1 expression may provide insight into this potential mechanism.
Our results also indicate DISC1 expression can regulate the probability of glutamate release, whereby overexpression of DISC1 results in paired-pulse depression and an increase in CV. Conversely, RNAi knockdown of DISC1 produces paired-pulse facilitation and a decrease in CV. These results further indicate a presynaptic function for DISC1, however the mechanism associated with this phenotype is not apparent. One potential mechanism involves DISC1 regulation of synaptic vesicle trafficking [16]. Alternatively, the regulation of mitrochondria trafficking by DISC1 may be important, as defects in the trafficking of mitochondria are known to alter several aspects of synaptic transmission including short-term plasticity [17,18,29,30].
Overall, our results provide the strong evidence for presynaptic effects of DISC1 at glutamatergic synapses in the neocortex. The level of DISC1 expression appears to regulate the probability of release and therefore may function to control the reliability of glutamate release. In contrast, expression of DISC1DC appears to inhibit synchronous glutamate release and may consequently affect the timing of synaptic transmission through neocortical circuits.  Figure S4 Activation of ChR2 with a 2 ms pulse of blue light evokes glutamatergic synaptic transmission. In utero electroporation on E15-E16 results in transfection of approximately 20% of layer 2/3 neurons in the targeted cortical area. Measuring light-activated synaptic transmission is amendable in this circuit because of the high level of recurrent connections between layer 2/3 neurons. A) Schematic depicting a recording in a layer 2/3 neuron transfected with ChR2 (green cell). (A1) A 2 ms pulse of blue light (473 nM; ,1 mW) generates a single action potential in a control neuron that is blocked by TTX application. B) Schematic depicting a recording from an untransfected neuron (open cell) surrounded by layer 2/3 neurons transfected with ChR2. B1) A 2 ms pulse of blue light stimulates surrounding ChR2-positive neurons to fire action potentials and results in an EPSC in the untransfected neuron that is completely blocked by TTX (1 mM, 94.464.8% block of control response, n = 13, p,0.005) or the AMPA receptor antagonist NBQX (10 mM, 96.560.8% block of control response, n = 11, p,0.0005). All EPSC recordings performed in the presence of the GABAa antagonist (gabazine, 5 mM). (TIF) Figure S5 Presynaptic expression of DISC1DC does not alter asynchronous release. A) A representative train of EPSCs recorded from an untransfected neuron and evoked from presynaptic neurons expressing ChR2 and dsRed (control). Ten consecutive traces are overlaid. EPSCs were evoked with a train of 15 light pulses at 50 Hz. sEPSCs were collected 400 ms before and after the stimulus train (inset). B) A representative train of EPSCs recorded from an untransfected neuron and evoked from presynaptic neurons expressing ChR2 and DISC1DC. C) Summary data showing this stimulation protocol was effective in producing a significant asynchronous release for control recordings, as the frequency of sEPSCs is significantly enhanced following the stimulus train in control but not DISC1DC condition (control frequency before train 1.260.2 Hz vs. after train 2.660.6 Hz (n = 12); p,0.02 paired t-test; DISC1DC frequency before train 2.160.4 Hz vs. after train 2.660.4 Hz (n = 13); p = 0.2). However, the post stimulation/prior stimulation ratio of sEPSC frequency was not statistically different between control and DISC1DC terminals (control 2.1260.36 (n = 12) vs. DISC1DC 1.4860.19 (n = 13); p = 0.13), suggesting the DISC1DC-dependent slowing of EPSC kinetics is separate from asynchronous release. (TIF)