Cell-autonomous regulation of structural and functional plasticity in inhibitory neurons by excitatory synaptic inputs

Functional circuit assembly is thought to require coordinated development of excitation and inhibition, but whether they are co-regulated cell-autonomously remains unclear. We investigated effects of decreased glutamatergic synaptic input on inhibitory synapses by expressing AMPAR subunit, GluA1 and GluA2, C-terminal peptides (GluA1CTP and GluA2CTP) in developing Xenopus tectal neurons. GluACTP decreased excitatory synaptic inputs and cell-autonomously decreased inhibitory synaptic inputs in excitatory and inhibitory neurons. Visually-evoked excitatory and inhibitory currents decreased proportionately, maintaining excitation/inhibition. GluACTP affected dendrite structure and visual experience-dependent structural plasticity differently in excitatory and inhibitory neurons. Deficits in excitatory and inhibitory synaptic transmission and experience-dependent plasticity manifested in altered visual receptive field properties. Both visual avoidance behavior and learning-induced behavioral plasticity were impaired, suggesting that maintaining excitation/inhibition alone is insufficient to preserve circuit function. We demonstrate that excitatory synaptic dysfunction in individual neurons cell-autonomously decreases inhibitory inputs and disrupts neuronal and circuit plasticity, information processing and learning.

Activity plays a critical role in the refinement and maintenance of functional neural circuits 1 , which is thought to require coordinated development of its two principle components: excitatory and inhibitory neurons 2 . Although proportional co-regulation of excitation and inhibition and a constant excitation/inhibition ratio have been widely observed during circuit development 2 our understanding of how glutamatergic excitatory inputs affect the development of inhibition at synaptic and neuronal levels remains incomplete. Mounting evidence from different brain regions and species suggests that perturbing activity or sensory experience decreases development of inhibition and disrupts the maturation and specification of inhibitory neurons and circuits 3-7 , however most of these studies perturbed activity broadly and were unable to resolve cellautonomous and circuit-based outcomes. Direct evidence that glutamatergic synaptic inputs drive the cell autonomous development of inhibitory input in individual neurons is unclear.
As the predominant mediator of fast excitatory synaptic transmission, AMPARs provide the initial depolarization that is essential for the activation of NMDARs and subsequent secondary signal transduction and synaptic plasticity mechanisms. Four types of AMPAR subunit (GluA1-4) form different hetero-and homo-dimers of AMARs, with GluA1 and GluA2 being the major AMPAR subunits. Regulation of the trafficking of postsynaptic AMPAR underlies activitydependent plasticity of synaptic strength [8][9][10] . Regulatory sites within the C-terminal region of GluA1 and GluA2 subunits are required for synaptic trafficking of AMPARs 8,11 . Expression of peptides corresponding to the GluA C-terminal peptides (GluA1CTP or GluA2CTP) impairs AMPAR trafficking, decreases excitatory synaptic transmission and disrupts experiencedependent synaptic plasticity [12][13][14] . GluACTPs are therefore effective tools to disrupt AMPARmediated excitatory transmission in individual neurons, permitting study of outstanding questions concerning the role of excitatory synaptic inputs in structural and functional development of neurons and circuits.
Here, we expressed GluA1CTP or GluA2CTP, referred to collectively as GluACTPs, in individual tectal neurons to assess the effects of impaired excitatory synaptic transmission on inhibitory synaptic inputs and the development of structural and functional properties in excitatory and inhibitory neurons in vivo. We show that GluACTP expression proportionally decreased excitatory and inhibitory synaptic inputs, resulting in a constant balance of excitation to inhibition in both inhibitory and excitatory neurons. In vivo time-lapse imaging demonstrated that deficits in excitatory synaptic inputs had distinct effects on dendritic arbor development and experience-dependent structural plasticity in excitatory and inhibitory neurons. GluACTPmediated decreases in excitatory and inhibitory transmission also manifested in deficits in visual 4 information processing, recorded as impaired spatial and temporal receptive field properties, and visuomotor behavior. Finally, GluACTP expression blocked learning-induced behavioral plasticity. Our results demonstrated that excitatory synaptic dysfunction led to cell-autonomous inhibitory synaptic dysfunction, which then ramified to impair neuronal and circuit properties and degrade behavioral performance.

GluACTP expression reduced both excitatory and inhibitory synaptic transmission in tectal neurons.
To test whether decreasing glutamatergic synaptic inputs in individual neurons affects GABAergic synaptic transmission, we sparsely transfected tectal neurons with constructs coexpressing GFP and GluA1CTP or GluA2CTP and recorded mEPSCs and mIPSCs from GFP+ neurons 5-8 days later (Fig. 1a). mEPSC frequency was significantly reduced in both GluA1CTP and GluA2CTP expressing neurons, with no significant change in mEPSC amplitudes (Fig. 1bc). The decrease in mEPSC frequency likely reflects loss of synapses over several days of GluACTP expression. Interestingly, both the frequency and amplitude of mIPSCs were significantly reduced in GluACTP-expressing neurons (Fig. 1d-e), suggesting that excitatory synaptic inputs govern the development of inhibitory synaptic inputs in a cell-autonomous manner. By contrast, disrupting inhibitory synaptic inputs by interfering with GABA A R trafficking does not affect excitatory input onto the same neurons 15 . Paired pulse ratios of excitatory synaptic currents were comparable in neurons expressing GluA1CTP (n=5), GluA2CTP (n=5) or controls (n=7) (Fig. 1f-g), consistent with a deficit in AMPAR trafficking into postsynaptic sites [12][13][14]16 .
Tectal neurons receive direct excitatory retinal inputs as well as feed forward and feedback inhibitory inputs within tectal circuits 10,17,18 . To determine if the decreased excitatory synaptic transmission affected the E/I balance of evoked synaptic responses, we recorded excitatory and inhibitory compound synaptic currents (eCSCs and iCSCs) from transfected tectal neurons evoked by full field visual stimulation in intact animals (Fig. 1h). Visually-evoked eCSCs and iCSCs recorded from either GluA1CTP or GluA2CTP-expressing neurons were significantly smaller than controls (Fig. 1i-j), however, the ratio of total integrated charge transfer between iCSCs and eCSCs remained unchanged between GluACTP-expressing and control neurons (Fig.   1k). These data further demonstrate that interfering with GluA1-and GluA2-containing AMPAR 5 trafficking not only decreases excitatory synaptic transmission onto the transfected cells, but also induces a proportional decrease in the inhibitory synaptic transmission onto the same neurons.

Decreased excitatory inputs induced cell-autonomous decreases in inhibitory synaptic inputs in both excitatory and inhibitory neurons.
Excitatory and inhibitory tectal neurons demonstrate different visual experience-dependent structural and functional plasticity 19 . To test whether disrupting excitatory synaptic transmission affects excitatory and inhibitory neurons differentially, we combined immunohistochemical labeling of the excitatory and inhibitory synaptic markers PSD95 and gephyrin, with GABA immunolabeling to examine the density of excitatory and inhibitory postsynaptic puncta in dendrites of GluACTP-expressing neurons.
In the optic tectum, individual tectal neurons express a mixture of AMPARs with different subunit compositions (GluA2-lacking or GluA2-containing). The proportion of GluA2-lacking to GluA2-containing AMPARs varies among individual tectal neurons such that more immature neurons show higher content of calcium permeable GluA2-lacking-AMPARs 20 . Both excitatory and inhibitory neurons express GluA1 and GluA2 in the developing optic tectum, as shown by double immunolabeling with GABA and GluA1 or GluA2 antibodies ( Fig.2a-b).
We examined the effects of GluACTP expression on the density of PSD95 and gephyrin puncta in dendrites of sparsely transfected excitatory or inhibitory tectal neurons. Excitatory and inhibitory tectal neurons receive excitatory and inhibitory synaptic inputs (Fig. 2c). Both PSD95 and gephyrin immunolabeling are highly punctate, with higher puncta density in the neuropil than the somatic region (Fig. 2d). GFP+ dendritic segments were identified as inhibitory or excitatory by GABA immunolabeling. In both excitatory and inhibitory neurons, GluA2CTP expression reduced the density of both PSD95 and gephyrin puncta ( Fig. 2e- indicating a decrease in the number of both excitatory and inhibitory synapses onto transfected neurons. GluA1CTP expression decreased both PSD95 and gephyrin puncta in excitatory neurons but not in inhibitory neurons. Given that over 70% of tectal neurons are excitatory 19 , these results are consistent with decreased mEPSC and mIPSC frequency seen in electrophysiological recordings from randomly recorded neurons, and demonstrate cellautonomous loss of inhibitory synaptic inputs induced by decreased excitatory input in both excitatory and inhibitory neurons. 6 GluACTP expression differentially affects dendritic arbor growth and experiencedependent structural plasticity in excitatory and inhibitory neurons. To assess effects of disrupted excitatory synaptic transmission on dendritic arbor development and experience-dependent structural plasticity in excitatory and inhibitory neurons, we performed in vivo time-lapse imaging of GluACTP-and GFP-co-expressing individual neurons. Excitatory and inhibitory neurons were identified by post-hoc GABA immunolabeling ( Fig. 3a-b). Total dendritic branch length (TDBL) and total branch tip number (TBTN) from 3-D reconstructions of the imaged neurons indicated that GluA1CTP and GluA2CTP expression in excitatory neurons significantly decreased TBTN but not TDBL, resulting in decreased branch density, without affecting dendritic arbor branching pattern (Fig. 3c, e). In inhibitory neurons, GluA2CTP expression significantly increased TDBL, without affecting TBTN or branch density (Fig. 3d).
Interestingly, GluA2CTP expression also changed the branching pattern of inhibitory neurons, causing neurons to branch significantly farther from the soma (Fig. 3f), possibly reflecting a compensatory response to decreased excitatory inputs, consistent with observations of activitydependent redistribution of synapse in the absence of normal activity 21 .
Previous studies showed that 4hr of short-term enhanced visual experience (STVE) increased dendritic arbor growth rate in tectal neurons compared to 4hr in dark, and that GluACTP expression blocks this visual experience-dependent dendritic arbor structural plasticity 13 . These studies imaged randomly sampled tectal neurons, therefore the results likely reflect plasticity in excitatory neurons, the majority of tectal neurons. Inhibitory tectal neurons, on the other hand, demonstrate a bimodal experience-dependent plasticity, with an inverse correlation between the valence of plasticity in response to dark and STVE in individual neurons. Furthermore, inhibitory neurons cluster into two functional groups, one group increases dendritic arbor growth rate in response to STVE and decreases arbor growth rate in the dark, while the other decreases arbor growth rate in STVE and increases it in dark.
To test if the bimodal experience-dependent structural plasticity in inhibitory neurons is affected by disrupting excitatory synaptic inputs, we collected time-lapse images of individual tectal neurons co-expressing GFP and GluA1CTP or GluA2CTP in animals exposed to 4hr of dark followed by 4hr of STVE (Fig. 4a) Comparing dendritic arbor growth rates over 4h in STVE and 4h in dark for individual neurons demonstrated that GluACTP selectively blocked the structural plasticity of excitatory neurons in 7 response to STVE but not in dark (Fig. 4b- of structural responses to either dark or STVE was not different between GluACTP-expressing and control neurons (Fig. 4d). By contrast, plotting dendritic arbor growth rates over 4h in STVE versus dark for individual neurons demonstrates that GluACTP expression disrupted the inverse correlation between the valence of structural plasticity in dark and STVE (Fig. 4e).
Application of unsupervised cluster analysis based on ΔTBTN in response to STVE versus dark clustered control inhibitory neurons into two evenly-sized subpopulations, called Group I and Group II neurons (Fig. 5a, Group I: n=14; Group II: n=16). Group I and Group II neurons displayed experience-dependent structural plasticity with opposite valence, accounting for the lack of experience-dependent plasticity in the pooled population (Fig 4c). The plasticity profile of Group I inhibitory neurons was similar to excitatory neurons, retracting dendrites in the dark and extending them in STVE. The plasticity profile of Group II inhibitory neurons was the opposite, extending dendrites in the dark and retracting them in STVE (Fig. 5aiii,). We applied the same cluster analysis to GluACTP-expressing inhibitory neurons and assigned the clustered subgroups by the difference between average ΔTBTN in dark and STVE as in control neurons (Group I: In addition, comparison of the structural plasticity in response to dark or STVE between GluACTP-expressing and control inhibitory neurons showed that within both Group I and Group II neurons, GluA2CTP expression significantly altered the plasticity in response to dark, not STVE ( Fig. 5d-g). By contrast, the plasticity in response to either dark or STVE in GluA1CTP-8 expressing inhibitory neurons was not different from controls in either group. These results indicate that GluA2 is of particular importance for the bi-modal plasticity in inhibitory neurons.
Considering that only GluA2CTP-expression significantly decreased PSD95 and gephyrin puncta in inhibitory neurons ( Fig. 2g-h), these results again suggest that, unlike excitatory neurons, inhibitory tectal neurons are more sensitive to the disruption of GluA2-mediated AMPAR trafficking. The loss of the bi-modal plasticity response indicates that disrupting excitatory synaptic inputs to inhibitory neurons changed their circuit connectivity 19 .
Using unsupervised cluster analysis, control excitatory neurons cluster into two groups distinguished by their plasticity in the dark ( showed that GluA1CTP and GluA2-CTP significantly affected the plasticity in response to STVE but not dark ( Fig. 6d-g). The observation that decreased excitatory inputs significantly affected the STVE response in excitatory neurons and the dark response in inhibitory neurons provides further evidence that the plasticity of inhibitory tectal neurons is actively regulated in dark 19,22 .

Decreased excitatory and inhibitory inputs disrupt receptive field properties.
Interaction of excitatory and inhibitory inputs is thought to be essential for the developmental refinement of visual receptive field (RF) properties [23][24][25] . Here, we showed that interfering with glutamatergic inputs compromises both excitatory and inhibitory inputs yet E/I remains constant.
To test whether RF properties are affected by GluACTP expression, we measured the spatial and temporal RF properties in transfected neurons. We recorded spatial receptive fields in both cellattached mode, to measure the spiking receptive fields (sRF), and whole-cell mode, to measure excitatory receptive fields (eRF) and inhibitory receptive fields (iRF) respectively in response to 9 light off visual stimuli (Fig. 7a). sRFs, eRFs and iRFs were all significantly smaller in GluA1CTP-and GluA2CTP-expressing neurons than controls (Fig. 7b-d). GluACTP expression disrupted the convergence of eRF and iRF as shown by significantly greater distance between the eRFs and iRFs centers (Fig. 7e). We generated temporal receptive field (tRF) maps by binning the number of spikes in 100 ms intervals over the 700 ms recording period following the OFF stimulus (Fig. 7f). The tRFs in control GFP+ neurons was confined to 200 ms following the stimulus (Fig. 7g). By contrast, the tRF in GluA1CTP-and GluA2CTP-expressing neurons extended up to 600 ms after the stimulus. Consequently, the average spike latency and the full width at the half maximal response (FWHM) of the tRFs in GluA1CTP-and GluA2CTPexpressing neurons were significantly greater than controls (Fig. 7h-i). The decreased convergence of eRF and iRF and the increased temporal span of the visually evoked spikes are consistent with a decreased inhibition 15,18,24,25 . Together, these data indicate that decreasing excitatory synaptic inputs and the subsequent cell autonomous decrease in inhibitory synaptic inputs in tectal neurons disrupted the tectal circuits underlying visual information processing and impaired both spatial and temporal RF properties.

GluACTP expression impairs visual avoidance behavior and learning-induced behavioral plasticity.
We next addressed whether GluACTP expression affects visual avoidance behavior and behavioral plasticity. The results described above predict two outcomes of GluACTP expression: On one hand, if constant E/I is sufficient for circuit function underlying behavior, then the coregulation of inhibitory inputs in response to GluACTP expression and the resultant constant E/I, predicts that visual avoidance behavior would be intact. By contrast, GluACTP expression impaired visual information processing and RF properties, predicting that circuit function and plasticity would be impaired. We bulk electroporated tectal neurons with GluA1CTP, GluA2CTP, or GFP expression constructs and evaluated visual avoidance behavior, and avoidance behavior plasticity. The visual avoidance behavior is a tectal-mediated visually-guided behavior in which an animal changes swim trajectory in response to an approaching visual stimulus 15 . Animals improve their avoidance behavior following a visual-conditioning training protocol 26 . Control tadpoles avoided dots with diameters ranging from 0.2 to 0.6 cm, with the peak avoidance response to dots of 0.4 cm diameters ( Fig. 8a-b). GluA1CTP-and GluA2CTP-expressing tadpoles had significantly lower avoidance indices to stimuli of 0.4 cm and 0.2 cm, and GluA2CTPexpressing animals also showed a significantly decreased response to 0.6 cm dots (Fig. 8b), indicating impaired visual behavior in these animals despite of balanced E/I. GluA1CTP and GluA2CTP expression also blocked visual-conditioning mediated plasticity of the behavior (Fig.   8c). This learning deficit is consistent with the compromised experience-dependent structural plasticity and visual information processing observed in individual neurons expressing GluACTPs.

Discussion
Genetic variants of proteins associated with glutamatergic synaptic function, such as CTNAP2 and SHANK3, have been implicated in the etiology of neuropsychiatric disorders, placing excitatory synapse dysfunction in the spotlight as a candidate mechanism underlying pathogenesis of these disorders 27,28 . Animal models with these genetic deficits also show reduced inhibitory tone, recapitulating human patient studies 29 . Some of these neurological disorders are thought to have neurodevelopmental origins, such as ASD and schizophrenia, raising the intriguing questions: Is there a causal link between defective excitatory synaptic function during development and reduced inhibition? And how do deficits in excitatory synaptic functions lead to deficits in inhibitory function?
Here we examined the role of glutamatergic excitatory synaptic transmission in the coordinated development of inhibition and excitation at synaptic, neuronal, circuit and behavioral levels, focusing on inhibitory synapses and neurons. We report a coordinated cell-autonomous reduction in synaptic inhibition in response to decreasing glutamatergic transmission by GluACTP expression in individual neurons. The decreased inhibitory input was not only observed in spontaneous activity but also in visually evoked inhibitory synaptic responses, which drastically altered the spatial and temporal visual receptive field properties in transfected tectal neurons.
Disrupting excitatory transmission also blocked experience-dependent structural plasticity in both inhibitory and excitatory neurons. Interestingly, the structural plasticity deficit in inhibitory neurons occurred in the response to dark, whereas in excitatory neurons the deficit occurred in response to visual stimulation. These synaptic and cellular defects translated into behavioral deficits when GluACTPs were expressed more extensively in tectal neurons.
The C-terminals of AMPARs include core regulatory sites for AMPAR trafficking. CTPs compete with endogenous AMPARs for binding partners, thereby interfering with AMPAR trafficking into synapses and decreasing excitatory synaptic transmission. Different mechanisms may underlie the disruption of excitatory synaptic inputs by expression of GluA1CTP and GluA2CTP due to different synaptic delivery mechanisms of GluA1-and GluA2-containing AMPARs 11,12 . GluA1CTP does not affect basal AMPAR-mediated currents but abolishes activity-dependent synaptic potentiation. GluA2CTP significantly decreases basal AMPARmediated synaptic transmission, which produces greater synaptic potentiation in response to LTPinducing protocols, but due to impaired synaptic delivery of GluA2-containing receptors, this increased synaptic strength is not maintained 9 . AMPAR trafficking also affects homeostatic plasticity where GluACTP blocks inactivation-induced synaptic scaling 30  preferentially co-localizes with GluA1 and knocking it down in cultured neurons led to abnormal cytoplasmic aggregation of GluA1, suggesting a role in AMPAR trafficking 34 . The dysfunction of inhibitory synaptic transmission observed with CNTNAP2 knockdown may be a secondary consequence of disrupted excitatory transmission due to defective AMPAR trafficking, as we observed here. Another consequence of the coordinated decrease in inhibition induced by decreased excitation is that E/I remained relatively stable. E/I is thought to be critical for neural circuit stability and normal brain function 2 . Disrupted E/I is associated with several neurological diseases, including epilepsy, schizophrenia and autism spectrum disorders [35][36][37] . Here we showed that even though E/I was resilient to disruption of excitatory inputs, nervous system function was still significantly compromised at both neuronal and circuit levels, shown by abnormal dendritic morphology, altered experience-dependent plasticity and defective receptive field properties, suggesting that balanced E/I is not sufficient to maintain normal brain function. Mice with MECP2 knockout in forebrain excitatory neurons provide another example of simultaneous reduction in excitation and inhibition resulting in functional deficits despite relatively balanced E/I 38 .
Classical studies have shown that excitatory sensory inputs modify inhibition, however studies specifically investigating the direct cell autonomous involvement of AMPAR-mediated excitatory inputs in the development of inhibitory neurons have been sparse 39  Specific cellular mechanisms underlying the cell-autonomous down regulation of inhibition by decreased excitatory inputs are still unclear. One potential mechanism is retrograde signaling through BDNF, which has been shown to regulate formation and maintenance of inhibitory inputs in response to excitatory inputs 48 . NO has also been proposed as a retrograde messenger that mediates heterosynaptic potentiation of GABAergic synapses 49 . Previous studies indicate that cell-autonomous regulation of inhibitory synaptic inputs is independent of postsynaptic spiking 4 , suggesting that excitatory postsynaptic activity alone is sufficient to control the formation and maintenance of inhibitory synapses. When GABAergic currents are hyperpolarizing and AMPARs are major source of synaptic depolarization, as is the case in our experiments, disrupting AMPAR-mediated excitatory synaptic inputs reduces the depolarization that is crucial for the activation of NMDARs, which has been shown to be obligatory for the development of 13 inhibitory synaptic transmission 49,50 . Our data provide direct evidence for an essential role for glutamatergic excitatory transmission in the cell autonomous development of GABAergic inhibition and distinguished effects of excitatory inputs on excitatory and inhibitory neurons.

Animals and Transfection
Albino Xenopus laevis were reared as previously described 19   For receptive field mapping, white squares on a dark background were presented in an 8×8 grid of 0.5×0.5 cm 2 non-overlapping squares covering a 4×4 cm 2 area in the visual field. The entire visual field was mapped by randomly presenting white squares for 1.5 s with 5 s intervals.
For spike receptive field mapping, using cell attached recordings, the response within 700 ms after the onset of the off stimulus at each grid position was determined as the average number of total spikes per stimulus from two or three repeats. After cell-attached recording, whole cell voltage clamp recording was accomplished by applying negative pressure. The same visual stimuli were used to measure inhibitory spatial receptive fields and excitatory spatial receptive fields, by holding membrane potential at 0 mV and -60 mV respectively. Total synaptic charge transfer over 700 ms from the onset of stimulus was normalized to the peak response and computed by Matlab to show spatial receptive field size. All values larger than 3 times the standard deviation of spontaneous activity were included in the measurements of spatial receptive fields. The center of the spatial receptive field was defined as the center of the square that elicited the highest responses (maximum number of spikes).

In Vivo Time Lapse Imaging of Dendritic Arbor Structure and Data Analysis
Animals were electroporated with DNA constructs at stage of 45-46 and screened for those with sparsely transfected and well-isolated cells. For imaging, animals were anesthetized with 0.01% MS-222 (Sigma) and were placed in a Sylgard chamber covered by a glass coverslip. Images were collected every 4hr before and after each visual experience session (dark or STVE). Twophoton z-series were collected at 1 µm steps with a 20x water immersion objective (Olympus XLUMPlanFL 0.95NA) at 3-4x scan zoom using a custom-built microscope modified from an Olympus FV300 system 19 .
Complete dendritic arbors of each neuron were reconstructed using a semi-manual function in the Filament module of Imaris (Bitplane, US). Total dendritic length and branch tip number were automatically calculated by the software. 3D Sholl analysis calculated the number of branches that intersect concentric circles at increasing distances from the cell soma, using a customized Matlab program with reconstructed filament data exported from Imaris.

Visual Avoidance Assay and Visual Conditioning
The visual avoidance assay was conducted as previously described 15 . Tadpoles were placed in an 8Χ3 cm tank filled to a depth of ~1 cm with Steinberg's rearing solution. Visual stimuli were presented to a back-projection screen on the bottom of the chamber using a microprojector (3M, MPro110). Videos of tadpoles illuminated by IR LEDs were recorded with a Hamamatsu ORCA-ER digital camera. Visual stimuli were generated and presented by MATLAB. Randomly positioned moving spots of 0.04, 0.2, 0.4 and 0.6 cm diameter were presented in random order for 60 seconds. Visual avoidance behavior was scored as a change in the swimming trajectory in the first ten encounters between each tadpole and moving spots (the percentage of avoidance responses out of 10 encounters, plotted as avoidance index). For visual conditioning, animals were exposed to moving bars (1 cm width; 0.3 Hz; Luminance: 25 cd/m 2 ) continuously for 2 or 4 hours, or for 3 repeats of 5 minutes of moving bars interleaved by 5-minute blank, for a total of 30 minutes, as described 26 .

Cluster Analysis and Statistical Tests
Custer analysis was performed based on ΔTBTN over STVE versus darl of individual neurons using an unsupervised agglomerative hierarchical tree method in MATLAB (linkage.m) based upon their pair-wise vectorial distance in the constructed 2D space (pdist.m) 19 .