The balance of inhibitory to excitatory (I/E) synaptic inputs is thought to control information processing and behavioral output of the central nervous system. We sought to test the effects of the decreased or increased I/E ratio on visual circuit function and visually guided behavior in Xenopus tadpoles. We selectively decreased inhibitory synaptic transmission in optic tectal neurons by knocking down the γ2 subunit of the GABAA receptors (GABAAR) using antisense morpholino oligonucleotides or by expressing a peptide corresponding to an intracellular loop of the γ2 subunit, called ICL, which interferes with anchoring GABAAR at synapses. Recordings of miniature inhibitory postsynaptic currents (mIPSCs) and miniature excitatory PSCs (mEPSCs) showed that these treatments decreased the frequency of mIPSCs compared with control tectal neurons without affecting mEPSC frequency, resulting in an ∼50% decrease in the ratio of I/E synaptic input. ICL expression and γ2-subunit knockdown also decreased the ratio of optic nerve-evoked synaptic I/E responses. We recorded visually evoked responses from optic tectal neurons, in which the synaptic I/E ratio was decreased. Decreasing the synaptic I/E ratio in tectal neurons increased the variance of first spike latency in response to full-field visual stimulation, increased recurrent activity in the tectal circuit, enlarged spatial receptive fields, and lengthened the temporal integration window. We used the benzodiazepine, diazepam (DZ), to increase inhibitory synaptic activity. DZ increased optic nerve-evoked inhibitory transmission but did not affect evoked excitatory currents, resulting in an increase in the I/E ratio of ∼30%. Increasing the I/E ratio with DZ decreased the variance of first spike latency, decreased spatial receptive field size, and lengthened temporal receptive fields. Sequential recordings of spikes and excitatory and inhibitory synaptic inputs to the same visual stimuli demonstrated that decreasing or increasing the I/E ratio disrupted input/output relations. We assessed the effect of an altered I/E ratio on a visually guided behavior that requires the optic tectum. Increasing and decreasing I/E in tectal neurons blocked the tectally mediated visual avoidance behavior. Because ICL expression, γ2-subunit knockdown, and DZ did not directly affect excitatory synaptic transmission, we interpret the results of our study as evidence that partially decreasing or increasing the ratio of I/E disrupts several measures of visual system information processing and visually guided behavior in an intact vertebrate.
- temporal precision
- receptive field
- γ-aminobutyric acid
- GABAA receptor
- synaptic transmission
- avoidance behavior
- first spike latency
the central nervous system (CNS) is composed of networks of excitatory and inhibitory neurons that process information and control behavior. The strengths of excitatory and inhibitory synaptic inputs to neurons are balanced, such that the distribution and strengths of excitatory inputs appear to be matched by inhibitory inputs (Ascoli et al. 2008; Burkhalter 2008). A balance of excitatory and inhibitory inputs is postulated to allow neurons to operate over a greater dynamic range in response to inputs (Tiesinga et al. 2008), and imbalances in excitatory and inhibitory inputs are thought to disrupt information flow within and between brain networks (Rubenstein and Merzenich 2003).
We used the visual system of Xenopus tadpoles to test the consequences of manipulating the ratio of inhibitory to excitatory inputs (I/E) on visual response properties and behavior. The visual system is particularly useful for probing the postulated function of a balance of inhibition to excitation on circuit function. Neuronal responses to visual inputs, visual information processing, and visually controlled behaviors have been well characterized in a variety of experimental systems, and perturbations in these response properties can be readily recognized. For instance, information about the sensory world is represented in the timing of stimulus-induced spikes (Daw et al. 2006; Haider et al. 2006; Hasenstaub et al. 2005; Priebe and Ferster 2008; Tiesinga et al. 2008), spatial and temporal receptive fields of neurons (Liu et al. 2009; Nowak et al. 2009; Richards et al. 2010; Tao and Poo 2005), and input/output relations (Carvalho and Buonomano 2009; Mittmann et al. 2005; Pouille and Scanziani 2001; Tiesinga et al. 2008). In addition, visually guided behaviors provide a readout of visual system function (Dong et al. 2009; Gahtan et al. 2005; Orger et al. 2008; Prusky et al. 2004).
In the visual system of Xenopus tadpoles, excitatory inputs from retinal ganglion cells activate a highly interconnected circuit of inhibitory and excitatory tectal neurons (Akerman and Cline 2006; Dong et al. 2009; Pratt et al. 2008; Tao and Poo 2005), some of which project to the hindbrain and spinal cord to control swimming behavior (Gahtan et al. 2005; Orger et al. 2008). The majority of synaptic GABAA receptors (GABAAR) in the CNS contains the γ2 subunit (Fritschy et al. 2003; Sieghart and Sperk 2002), which is required for anchoring GABAAR at postsynaptic sites (Alldred et al. 2005; Essrich et al. 1998; Meier and Grantyn 2004; Schweizer et al. 2003). Furthermore, a peptide, called the intracellular loop (ICL), which corresponds to 95 amino acids of the ICL between transmembrane domains 3 and 4 of the γ2 subunit, prevents anchoring of γ2-subunit-containing GABAAR to synaptic sites (Alldred et al. 2005; Essrich et al. 1998; Meier and Grantyn 2004; Schweizer et al. 2003). Our strategy was to selectively decrease GABAAR-mediated synaptic transmission by expression of ICL or knocking down the γ2 subunit with antisense morpholino (MO) oligonucleotides. We increased GABAAR-mediated synaptic transmission by exposure to the GABAAR-positive allosteric modulator, diazepam (DZ). Because these manipulations did not directly affect excitatory synaptic transmission, they provided us with the opportunity to assess the effect of decreased and increased ratios of I/E synaptic input on visual-response properties of tectal neurons and visually guided behavior. The data indicate that modulating I/E in tectal neurons interferes with visual circuit function and visually guided behavior.
MATERIALS AND METHODS
All experimental protocols were approved by the Institutional Animal Care and Use Committee at The Scripps Research Institute.
Animals and transfection.
Xenopus laevis tadpoles were generated from the lab colony or purchased from Nasco (Fort Atkinson, WI) and reared from stage 23 in an incubator at 16°C with a 12-h dark/12-h light cycle. Albino tadpoles of either sex were used for experiments. For tectal neuron transfections, stage 47/48 tadpoles were anesthetized in 0.02% MS-222 (Tricane methanesulfonate; Sigma-Aldrich, St. Louis, MO) and electroporated with DNA plasmids (∼0.6 μM) or MOs (200 μM) (Bestman et al. 2006). Bulk electroporation with cytomegalovirus (CMV) promotor constructs and MOs transfects both inhibitory and excitatory neurons in the optic tectum. Electroporation of the left or right half of the brain is accomplished by using unipolar current pulses, whereas bilateral electroporation is accomplished by switching the polarity of current pulses, as described (Haas et al. 2002).
Plasmids and MOs.
A dual CMV promoter plasmid expressing enhanced green fluorescent protein (EGFP) and ICL, EGFP and mutant ICL (mICL), or EGFP alone was used for transfections. ICL was modified from the murine sequence of ICL (kindly provided by Dr. Simon Rumpel, The Institute for Molecular Pathology, Vienna, Austria) (Shen et al. 2009). The mICL is generated by exchanging Ser (S) to Ala (A) of ICL (HYFVSNRKPSKDKDKKKKNPLLRMFSFKAPTIDIRPRSATIQMNNATHLQERDEEYGYECLDGKDCASFFCCFEDCRTGAWRHGRIHIRIAKMDS) (Shen et al. 2009). The X. laevis cDNA of the GABAAR γ2 subunit was cloned from a cDNA library made with the SMART rapid amplification of cDNA ends amplification kit (Clontech Laboratories, Mountain View, CA), according to the manufacturer's protocol, and sequenced. To knock down the GABAAR γ2 subunit, we transfected neurons with a translation-blocking MO against the Xenopus γ2 subunit (γ2-MO; GATCGCATCTCTCGTCGCGTTTCGA) or a five-pair nucleotide mismatch control MO (Ctrl-MO; GATgGCATgTCTCcTCGCcTTTgGA) from Gene Tools (Philomath, OR), both of which were tagged with carboxyfluorescein.
Stage 47 tadpole midbrains were dissected and homogenized immediately in ice-cold radioimmunoprecipitation assay buffer containing a protease inhibitor cocktail (Roche Applied Science, Indianapolis, IN) and 1 mM sodium orthovanadate. Equal amounts of protein homogenates for each experimental group were separated by SDS-PAGE (12%) and transferred to nitrocellulose membranes. Blots were blocked in 5% nonfat milk with 0.1% Tris-buffered saline and Triton X-100 (Sigma-Aldrich) and incubated with rabbit polyclonal GABAAR γ2 antibody (1:200, ab4073, Abcam, Cambridge, MA) diluted in blocking solution. Blots were rinsed and incubated with horseradish peroxidase-conjugated secondary antibodies (Amersham Biosciences, Piscataway, NJ). Bands were visualized using enhanced chemiluminescence reagent (Amersham Biosciences). Blots were stripped and reprobed with β-tubulin antibody (1:200, sc-9104, Santa Cruz Biotechnology, Santa Cruz, CA).
Anesthetized stage 47 tadpoles were dissected to remove the skin and meninges over the brain. Animals were transferred to 4% paraformaldehyde (Electron Microscopy Sciences, Fort Washington, PA) in phosphate buffer (PB; pH 7.4), exposed to a brief microwave pulse, and left to fix for an additional 2 h at room temperature. Animals were cryoprotected overnight in 30% sucrose (Sigma-Aldrich), and brains were removed and cut into 20-μm cryostat sections, which were permeabilized with 1% Triton X-100 in PB and blocked with 5% goat serum (Gibco, Grand Island, NY) in 1% Triton X-100 and incubated in rabbit polyclonal GABAAR γ2 antibody (diluted 1:200 in 2% goat serum with 0.1% Triton X-100 in PB) overnight at 4°C. Sections were rinsed with PB and incubated in secondary antibodies (Alexa Fluor 568 goat antirabbit, Invitrogen, Eugene, OR). To quantify the immunoreactivity, background-subtracted fluorescence intensity was measured per unit area in the neuropil with ImageJ 1.36 (Wayne Rasband, National Institutes of Health, Bethesda, MD). We compared the ratio of intensity between the MO-transfected and contralateral control sides of tectum by measuring intensity in the same size and location of the neuropil area. For each tectum, the immuofluorescence intensity from three different locations was averaged.
Experiments were performed at room temperature (20–22°C). For all recordings, the brains were perfused with extracellular saline containing (in mM): 115 NaCl, 2 KCl, 3 CaCl2, 1.5 MgCl2, 5 HEPES, 10 glucose, and 0.01 glycine, pH 7.2, osmolality 255 mOsm, throughout the recording. For recordings from neurons in isolated brains, tadpoles were anesthetized with 0.02% MS-222 (Sigma-Aldrich), brains were dissected, and neurons were recorded in whole-cell mode as described (Shen et al. 2009). When required, TTX (1 μM), DZ (10 μM), 2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f]quinoxaline-2,3-dione (NBQX; 20 μM), and dl-2-amino-5-phosphonopentanoic acid (D,L-APV) (50 μM, all Sigma-Aldrich) were included in the extracellular saline. Miniature inhibitory and excitatory postsynaptic currents (mIPSCs and mEPSCs, respectively) were recorded from neurons in isolated brains using a cesium (Cs)+-based pipette solution (in mM: 114 Cs-methane sulfonate, 1.5 MgCl2, 3 tetraethylammonium-chloride, 20 HEPES, 10 EGTA, 2 ATP, and 0.3 GTP, pH 7.2, osmolality 255 mOsm) in the presence of 1 μM TTX. The first 30 miniature events per neuron were used for analysis to ensure similar sampling from each cell. For electrical stimulation experiments, a bipolar-stimulating electrode (FHC, Bowdoin, ME) was placed in the optic chiasm or tectum to stimulate axons with a Master-8 stimulator (A.M.P.I., Israel).
For in vivo recordings, we used a regenerating ATP internal solution (Wang et al. 2003): (in mM) K-gluconate 95, MgCl2-6H2O 1.5, Na2-phosphocreatine 10, KCl 13, HEPES 20, EGTA 10, ATP 4, GTP 0.3, and creatine phosphokinase 10 units/ml. A cell-attached recording was started at least 5 min after gigaseal was obtained. After the cell-attached recording, the membrane in the patch was ruptured, and membrane potential was voltage clamped at −60 mV to record α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR)-mediated currents or at 0 mV to record GABAAR-mediated currents.
Recording micropipettes were pulled from borosilicate glass capillaries (Warner Instruments, Hamden, CT) with a pipette puller (P-97, Sutter Instrument, Novato, CA) and had resistances in the range of 7–9 MΩ when filled with Cs+-based pipette solution. All cells were recorded from the middle of the rostrocaudal axis of the tectum. Liquid junction potential was adjusted during recording. Signals were filtered at 2 kHz with a Multiclamp 700B amplifier (Molecular Devices, Palo Alto, CA) and a Digidata 1440A data acquisition system (Molecular Devices). Data were sampled at 10 kHz with Clampex 10 (Molecular Devices) and analyzed using a Clampfit 10 (Molecular Devices), Mini Analysis Program (Synaptosoft, Fort Lee, NJ), or MATLAB (The MathWorks, Natick, MA). IPSC decay kinetics were measured by a weighted decay time constant (τdecay) by fitting with an exponential function with Clampfit 10.
The full-field visual stimuli with three luminance steps (10, 20, and 250 cd/m2) were generated by MATLAB and presented from lowest to highest luminance from a projector [Samsung SP-P310ME light-emitting diode (LED)]. Luminance values were measured by a photometer (model IL1400BL, International Light Technologies, Peabody, MA). Each stimulus was presented 10 times at a frequency of 0.1 Hz with an interval of 0.05 Hz. Because the majority of tectal neurons at these stages shows predominant light-off responses, we analyzed the responses, including spikes, excitatory compound synaptic currents (eCSCs), and inhibitory CSCs (iCSCs) to light-off stimuli. As controls, eCSCs and iCSCs were recorded from cells transfected with Ctrl-MO and mICL. The ratios of iCSCs/eCSCs were not significantly different for Ctrl-MO and mICL neurons and were therefore combined. The iCSC/eCSC ratios for the three luminance values tested for the Ctrl-MO and mICL neurons were luminance 10 cd/m2: 1.89 ± 0.71 (Ctrl-MO), 2.06 ± 1.32 (mICL); luminance 20 cd/m2: 1.57 ± 0.41(Ctrl-MO), 2.04 ± 1.01 (mICL); and luminance 250 cd/m2: 1.32 ± 0.47 (Ctrl-MO), 0.97 ± 0.28 (mICL); n = 4/3; P > 0.05 for all comparisons.
Receptive field mapping.
For temporal receptive field mapping, full-field Gaussian white noise flicker stimuli at 50 Hz were presented to the eye from a fiber optic, and neurons from the contralateral tectum were recorded. The stimulus luminance was γ corrected. Spikes were recorded from tectal neurons in cell-attached mode for 20 min. Temporal receptive fields were measured by spike-triggered average for cell-attached recording or reverse correlation for whole-cell recording (Vislay-Meltzer et al. 2006). Many tectal cells in tadpoles have monophasic temporal receptive fields with negative deflections, consistent with observations that responses to light-off stimuli are stronger than light-on stimuli in the majority of tectal cells. The temporal receptive fields were normalized to the peak response over a window of 600 ms when the average response returns to baseline. The full width at half maximum (FWHM) of the normalized receptive field was used as the measure for the length of the temporal receptive fields (Butts et al. 2007; Vislay-Meltzer et al. 2006).
Spatial receptive fields were mapped using methods modified from previous studies (Butts et al. 2007; Dong et al. 2009; Engert et al. 2002; Vislay-Meltzer et al. 2006). A projector (Samsung SP-P310ME LED) was used to project computer-generated stimuli to the retina through a back-projector screen (see Fig. 5A). Tadpoles were anesthetized and the skin peeled back to expose the tectum. Tadpoles were stabilized on a sylgard cushion with pins. The sylgard cushion was attached to a bar in the recording chamber, which can be rotated so that the tadpole was mounted with one eye facing the projector screen. Brightness was controlled by Kodak neutral density filters (luminance range: 3–55 cd/m2), and image size from the projector was controlled by a convex lens. The receptive field was measured with an 8 × 8 grid of 0.5 × 0.5 cm2 nonoverlapping squares covering a 4.5 × 4.5-cm2 area in the visual field. White squares on a dark background or dark squares on a white background were presented for 1.5 s in random order, 64 times with 5-s intervals, until the entire visual field is mapped. The mapping sequence was repeated two or three times to eliminate possible contamination by spontaneous activity. The response at each grid position was determined as the average number of total spikes per stimulus and presented as a gray-scale map. After cell-attached recording, negative pressure was applied to the electrode to break through for whole-cell voltage clamp recording. The same 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 0.6 s from the onset of stimulus was normalized and computed by MATLAB to show spatial receptive field size. All values that are larger than three times the standard deviation of spontaneous activity were included in the measurements of spatial receptive fields. Only neurons from which both excitatory receptive fields (eRFs) and inhibitory receptive fields (iRFs) were successfully recorded were included in the correlation analysis. For cell-attached recording, the value at each grid position was normalized to the total number of evoked spikes within 0.6 s after the onset of the off-stimulus. Spatial receptive fields for synaptic currents are represented on a map with each square of the grid coded in a gray scale, according to the normalized total charge transfer in response to the visual stimulus. Spontaneous activity was measured during the last 1 s after each off stimulus.
Visual avoidance assay.
Animals were screened using an optomotor response, which does not require the optic tectum (Dong et al. 2009; Pronych et al. 1996) and serves as an indicator of nontectal visual function and animal health. Animals were placed in a 16 × 10-cm tank filled with ∼1-cm rearing solution and exposed to a stimulus of alternating black and white bars drifting in one direction. After 30 s, we counted the number of animals located in one-half of the chamber, toward which the gratings drifted, as described previously (Dong et al. 2009). Animals that do not respond to the optomotor stimulus are not used for further behavior tests. We then screened animals a second time with a visual stimulus of randomized moving spots. Only tadpoles that show an avoidance response to the moving spot stimulus with a diameter of 0.4 cm are used for further analysis. Animals in which the optic tectum was electroporated with EGFP, mICL, ICL, Ctrl-MO, or γ2-MO exhibit a normal optomotor response, indicating that electroporation itself or the specific reagents do not affect the ability of tadpoles to swim or respond to visual stimuli. Animals treated with DZ (10 μM or 30 μM) show normal swimming behavior (see Fig. 9F). In some experiments, we injected 1∼2 μl DZ (1 mM) or vehicle (1% DMSO in 0.9% NaCl) into the tectal ventricle with a micropipette and let the tadpoles recover for 2 h before testing behavior.
For the visual avoidance assay, single tadpoles are placed in a 16 × 10-cm tank filled ∼1 cm with rearing solution. The custom-made tank with a back-projection screen is mounted on top of a clear Plexiglas. Each animal is given 2 min to distribute within the tank, and then visual stimuli are presented for 30 s/stimulus using a microprojector (3M MPro110, luminance range: 3–90 cd/m2) positioned below the tank. The positions of the tadpoles are illuminated by four arrays of infrared LEDs. Videos of tadpoles and the stimuli captured on the screen are recorded with a Hamamatsu ORCA-ER digital camera at 15 frames/s. The entire system is enclosed in a light-tight compartment. Visual stimuli are generated and presented by MATLAB 2009b (The MathWorks, Psychophysics Toolbox extensions). Tadpoles are tested once for a series of different stimuli, in which spot size, luminance, velocity, and contrast are varied and presented in random order. An avoidance event is counted when a tadpole displays a sharp turn as a spot approaches the head at an angle between 75° and 105°. Avoidance behavior is counted from the first 10 encounters with moving spots and plotted as percent avoidance. Wild-type and albino tadpoles show similar avoidance behavior when either of them had prescreening with optomotor test and spot test (data not shown). The Wilcoxon signed-rank test is used to test whether behavior was significantly different from random. A Mann-Whitney test is used to test for significant differences between experimental groups.
To estimate the size (s) of the spots used as visual stimuli, we used the equation V = 2 arctan (s/2D) to calculate the visual angle (V) corresponding to each spot size. D represents the distance from tadpole's eye to the spot. Because the height of the solution in the tank is ∼1 cm, we estimate that the distance D from the eye to the spot is at most 1 cm. Visual angles of 75°, 55°, 30°, 16°, and 6° correspond to spots with estimated diameters of 0.6, 0.4, 0.2, 0.1, and 0.04 cm, respectively.
Student's t-tests were used for paired data. An ANOVA test followed by post hoc Tukey's test was used for multiple group data. Where noted, Wilcoxon signed-rank tests or Mann-Whitney tests were used. Data are presented as individual data points or as mean ± SE unless otherwise noted. Correlations were evaluated by Pearson's correlation coefficient. Experiments and analysis were performed blind to the experimental condition, unless noted.
Knockdown of the γ2 subunit of GABAAR.
Our previous studies showed that expression of ICL decreases the frequency of GABAergic mIPSCs without affecting mEPSCs and decreases the I/E ratio of optic nerve-evoked retinotectal synaptic responses in Xenopus tectal neurons (Shen et al. 2009). To establish an independent method to decrease inhibitory synaptic transmission, we transfected tectal neurons with γ2-MO to knock down the endogenous γ2-subunit expression. γ2-MO transfection decreased γ2-subunit immunolabeling in tissue sections of the optic tectum (Fig. 1, A–C) and in Western blots by about 36%, 2 days after transfection (Fig. 1D), suggesting that γ2-MO-transfected cells have lower expression of the γ2 subunit than controls. The partial loss of the γ2 subunit from transfected tecta could arise from partial transfection of the population of tectal cells or partial knockdown of protein in MO-transfected neurons.
γ2-Subunit knockdown selectively decreases GABAergic synaptic transmission.
To assess the effect of γ2-subunit knockdown on GABAergic and glutamatergic synaptic transmission, we transfected neurons with γ2-MO or Ctrl-MO. We recorded GABAAR-mediated mIPSCs and glutamate-mediated mEPSCs 2 days later. The interevent interval of mIPSCs in γ2-MO neurons is significantly greater than in Ctrl-MO neurons (Fig. 2, A and C), whereas the amplitudes of mIPSCs were not statistically different (Fig. 2C). In contrast, the interevent interval and the amplitudes of mEPSCs in γ2-MO neurons are not significantly different from Ctrl-MO neurons (Fig. 2D). These data indicate that γ2-MO transfection selectively decreases GABAergic synaptic inputs, similar to the effect of ICL expression (Shen et al. 2009).
The γ2 subunit is required for benzodiazepine binding to GABAAR (Sigel 2002). We therefore used the response to DZ to test whether γ2-MOs decrease γ2-subunit-containing GABAARs at postsynaptic sites. We used a stimulating and recording configuration, which allowed us to record direct inhibitory synaptic activity from optic tectal neurons (Fig. 2E): we placed a stimulating electrode in the tectum to excite local axons and collected whole-cell patch-clamp recordings at 0 mV holding potential from nearby neurons in the presence of NBQX to block excitatory transmission. Current injection through the stimulating electrode evoked direct inhibitory synaptic inputs in the recorded neurons (Fig. 2F). To test whether inhibitory synaptic transmission onto γ2-MO neurons can be enhanced with DZ, we recorded evoked inhibitory synaptic currents from γ2-MO neurons or Ctrl-MO neurons before and after perfusion of DZ (10 μM). DZ doubled the amplitude of evoked, GABAAR-mediated currents in Ctrl-MO neurons and lengthened the decay time of evoked inhibitory responses by ∼50%, consistent with the presence of the γ2 subunit containing synaptic GABAAR, but DZ did not affect the amplitude or time course of evoked GABAAR-mediated currents in γ2-MO cells (Fig. 2, F–H). Together, these results indicate that γ2-MO selectively decreases inhibitory synaptic inputs without affecting AMPAR-mediated mEPSCs and that this results from a selective loss of γ2-subunit-containing synaptic GABAAR. The experiments with DZ also show that DZ significantly enhances monosynaptic-evoked GABAergic responses in control tectal neurons but that the inhibitory synaptic responses remaining in γ2-MO cells are mediated by GABAAR, which lack γ2 subunits. We reported a similar specificity of ICL expression in the decreasing synaptic γ2 subunit containing synaptic GABAAR responses (Shen et al. 2009).
γ2-MO decreases, and DZ increases the I/E ratio.
Glutamatergic retinal input synapse on both excitatory and inhibitory tectal neurons and the inhibitory neurons forms feed-forward synapses on tectal neurons, which are driven by visual input (Akerman and Cline 2006). Consequently, optic nerve stimulation evokes both excitatory and inhibitory synaptic inputs in the majority of tectal neurons. We tested whether γ2-MOs can be used to decrease the ratio of evoked retinotectal I/E synaptic inputs in tectal neurons and whether DZ can be used to increase the ratio of evoked retinotectal I/E synaptic inputs. We placed a stimulation electrode in the optic chiasm of the isolated brain preparation and recorded optic nerve stimulation-evoked synaptic responses sequentially at holding potentials of 0 mV and −60 mV from the same neurons (Fig. 3A). Since the amplitudes of the evoked synaptic responses depend on the numbers of retinal axons stimulated, we normalized the peak amplitudes of evoked IPSCs to the peak amplitudes of evoked EPSCs for each neuron. The ratio of the amplitudes of evoked IPSCs to EPSCs (or the I/E ratio) in γ2-MO cells is less than one-half of the I/E ratio in Ctrl-MO cells, whereas the ratio of the amplitudes of evoked retinotectal IPSCs to EPSCs was increased significantly in DZ (Fig. 3, B and C). These results indicate that γ2-MO transfection can be used to decrease the ratio of evoked I/E synaptic responses in tectal neurons, similar to our previous finding with ICL expression (Shen et al. 2009), and that DZ can be used to increase the ratio of evoked I/E synaptic responses.
Extrasynaptic GABAergic responses are unaffected by ICL or γ2-subunit knockdown.
Activation of extrasynaptic GABAAR has been shown to affect neuronal excitability, sensory gating, and responses to synaptic inputs (Farrant and Nusser 2005; Mody and Pearce 2004; Tang et al. 2011). Most extrasynaptic receptors are comprised of α, β, and δ subunits and lack γ2 subunits (Jacob et al. 2008; Mody and Pearce 2004). Extrasynaptic GABAAR have ∼10-fold higher affinity for GABA than synaptic GABAAR, consistent with their activation by spillover of GABA from synaptic sites (Farrant and Nusser 2005; Mody and Pearce 2004; Sanna et al. 2009). Although tectal neurons have extrasynaptic GABAAR (Akerman and Cline 2006), we predicted that neither ICL nor γ2-MO would affect responses of extrasynaptic GABAergic responses, because they lack γ2 subunits. To test this directly, we recorded tonic GABAergic currents in whole-cell mode (holding at −60 mV) from control, γ2-MO, and ICL-expressing tectal neurons from isolated brains, while blocking glutamatergic currents with kynurenic acid (2 mM) (Fig. 4). Traces from representative cells are shown in Fig. 4, A, C, and E. We collected a 5-min period of baseline recording, during which spontaneous IPSCs (sIPSCs) can be seen in control neurons (Fig. 4A). We applied 1 μM muscimol to activate high-affinity, extrasynaptic GABAAR, as described (Sanna et al. 2009). Application of muscimol induced a hyperpolarizing shift in the holding current and increased the noise variance in all neurons recorded, consistent with the presence of extrasynaptic GABAergic receptors (Farrant and Nusser 2005; Mody and Pearce 2004; Sanna et al. 2009). Application of the GABAAR antagonist, bicuculine (100 μM), blocked the tonic current. Note that exposure to bicuculine does not result in a significant shift from the baseline (Fig. 4, A, C, and E), suggesting that there is little ambient GABA available to activate extrasynaptic receptors in the isolated brain preparation used for these experiments. Although γ2-MO- and ICL-expressing neurons have few sIPSCs during the baseline period, their tonic response to muscimol is comparable with control cells (Fig. 4, C and E). We generated all-point histograms of the current amplitudes for 2 min each during the baseline-recording period in the presence of muscimol and in the presence of bicuculine for Ctrl-MO cells, γ2-MO-transfected neurons, and ICL-expressing neurons (Fig. 4, B, D, and F). Muscimol application produced a comparable increase in tonic current in all groups of cells, which was blocked by bicuculine. There is no significant difference in tonic responses between Ctrl-MO cells and γ2-MO-transfected neurons or ICL-expressing neurons (Fig. 4G). These data indicate that ICL expression and γ2-subunit knockdown do not affect extrasynaptic GABAAR.
Evoked visual responses show shifts in I/E.
Results from experiments in Figs. 2 and 3 indicate that the ICL and γ2-MO decrease the ratio of optic nerve stimulation-evoked I/E synaptic inputs to tectal neurons. Here, we determined the effect of ICL and γ2-MO transfection and of DZ exposure on CSCs recorded from tectal neurons following full-field visual stimulation. Visual stimulation evokes CSCs in tectal neurons, which are the result of direct excitatory retinal input and polysynaptic excitatory and inhibitory tectal activity (Dong et al. 2009; Pratt et al. 2008; Zhang et al. 2000). To determine whether manipulation of inhibitory synaptic inputs affects visual responses in intact animals, we recorded CSCs from tectal neurons following whole-field visual stimulation of increasing luminance steps (10, 20, and 250 cd/m2). We recorded the evoked eCSCs at a holding membrane potential of −60 mV and evoked iCSCs at a holding potential of 0 mV. Tectal neurons were transfected with γ2-MO or Ctrl-MO and recorded 2 days later, or they were transfected with a dual promoter plasmid expressing EGFP and either ICL or mICL, and EGFP-expressing neurons were recorded 2 days later. In addition, we recorded CSCs in the presence of DZ (10 μM) to test the effect of increasing inhibitory synaptic transmission on visually evoked responses (Fig. 5, A and B). Data from Ctrl-MO and mICL-expressing neurons were not significantly different and were combined as controls (see materials and methods, Electrophysiology).
Visual stimuli of increasing luminance values from 10 to 250 cd/m2 evoked iCSCs and eCSCs with corresponding increases in integrated current in control neurons (Fig. 5B). Visually evoked eCSCS have an initial component, which is the response to direct retinal inputs, and a later component of prolonged synaptic currents from recurrent inputs. Visually evoked iCSCs in control neurons are larger and longer lasting than eCSCs, as reported previously (Akerman and Cline 2006; Richards et al. 2010; Tao and Poo 2005; Zhang et al. 2003). Visually evoked CSCs recorded from ICL and γ2-MO neurons and in the presence of DZ are different from controls (Fig. 5B and Table 1). To test whether ICL expression, γ2-MO, or DZ changes the ratio of I/E in evoked visual responses, we determined iCSC/eCSC ratios of total integrated charge transfer (Fig. 5C). For control neurons, the average iCSC/eCSC ratio was greater for lower luminance values, and responses were more variable between cells than for responses to higher luminance stimuli. For ICL and γ2-MO neurons, the iCSC/eCSC ratios were significantly lower than control neurons for stimuli of 10 and 20 cd/m2 luminance, but the iCSC/eCSC ratios were not different between control and ICL-expressing or γ2-MO-transfected neurons for stimuli with stronger luminance intensity values (250 cd/m2). iCSC/eCSC ratios were significantly greater in DZ-treated neurons compared with controls for stimuli of 20 and 250 cd/m2 luminance (Fig. 5C). These data suggest that luminance intensity of the stimulus differentially recruits excitatory and inhibitory inputs to tectal neurons in the control optic tectum, as seen in visual cortex, hippocampus, and pyriform cortex (Carvalho and Buonomano 2009; Stokes and Isaacson 2010; Tucker and Fitzpatrick 2006). It is interesting to note that despite the variability in individual CSCs, the ratios of I/E CSCs are less variable, consistent with the idea that inhibition and excitation are coordinately recruited within the circuit. The data also indicate that increasing or decreasing the ratio of I/E synaptic transmission onto individual tectal neurons within the tectal circuit has ramifying effects on circuit responses to visual input.
Regulation of spontaneous activity by I/E.
We analyzed spontaneous spiking in cell-attached recordings and spontaneous synaptic currents with whole-cell recordings from control EGFP- or ICL-expressing tectal neurons and DZ-treated neurons in intact animals. TTX was not present during these recordings, so the synaptic currents are a report of spiking activity in tectal neurons presynaptic to the recorded neuron. The frequency of spontaneous action potential spiking in ICL-expressing neurons is significantly greater than control neurons (Fig. 6A). These data indicate that ICL increases spontaneous spiking activity in ICL-expressing neurons, which could have an effect on neurons throughout the tectal circuit. Furthermore, the frequency of sIPSCs is lower in ICL-expressing neurons compared with controls, whereas the frequency of spontaneous EPSCs (sEPSCs) is not significantly different (Fig. 6, B and C). Although DZ did not affect spontaneous action potential activity in tectal neurons, it increased the frequency of sIPSCs and decreased the frequency of sEPSCs (Fig. 6, B and C).
Temporal fidelity of visual responses.
The precision of spike timing is important for coding information and for accurate information transfer between neurons (Tiesinga et al. 2008). We tested whether increasing or decreasing I/E affects the temporal fidelity of first spike latency of visual responses in tectal neurons. We recorded in cell-attached mode from neurons transfected with γ2-MO or ICL or with Ctrl-MO or mICL as controls and from neurons exposed to DZ. Individual responses and superimposed traces show that the first spike latency is more variable in neurons expressing γ2-MO or ICL and less variable in neurons exposed to DZ compared with controls (Fig. 7A). A plot of the variation of first spike latencies relative to the average latency per cell shows that the temporal fidelity of the first spike latency is lower in ICL and γ2-MO neurons and higher in DZ-treated neurons (Fig. 7B).
We analyzed the effect of luminance on the temporal fidelity of the first spike latency by determining the coefficient of variance (CV) of the first spike latencies relative to the median latency of each cell at each luminance. First spike latency is consistently reliable in control neurons for visual stimuli across luminance values from 10 to 250 cd/m2. When inhibition is decreased with ICL or γ2-MO, the variation of first spike latency is significantly greater for stimuli of 10 cd/m2 luminance but is comparable with controls for stimuli of 250 cd/m2 (Fig. 7C and Table 2). DZ decreases the CV for responses to 20 cd/m2 luminance compared with controls but not for other luminances tested, suggesting increased I/E can improve temporal fidelity of responses for some stimuli.
The synaptic I/E ratio has been shown to affect the latency to fire action potentials (Carvalho and Buonomano 2009; Marder and Buonomano 2004; Tiesinga et al. 2008). First spike latencies are significantly longer in response to a stimulus of 10 cd/m2 than for a stimulus of 250 cd/m2 for all cell groups tested. Furthermore, first spike latencies in response to stimuli of 10 cd/m2 are significantly longer in ICL, γ2-MO, and DZ-treated neurons compared with controls (Fig. 7D and Table 2). Analysis of interspike intervals (ISIs) over the first 50 ms of the visually evoked response shows that ISIs decrease with increasing luminance values in control neurons, but not in ICL, γ2-MO, or DZ-treated neurons (Fig. 7E and Table 2), consistent with a previous study indicating that I/E affects ISIs within a burst (Berg et al. 2007).
These data indicate that decreasing inhibitory synaptic inputs with ICL expression or γ2-subunit knockdown increases the variance in first spike latency of visual responses and can increase the latency of visual responses, whereas increasing inhibitory synaptic transmission with DZ increases the fidelity of visual responses. The data suggest that increasing or decreasing the I/E ratio of synaptic inputs to tectal neurons significantly affects spike timing in response to visual inputs. In the following experiments, we tested the effect of decreasing or increasing I/E on visual information processing in the optic tectum and visually guided behavioral output, using ICL expression, γ2-subunit knockdown, or DZ.
I/E ratio shapes spatial receptive fields.
Spatial receptive fields for visual input represent the regions of the visual world from which information is detected and processed in the CNS. Because spatial receptive fields are an emergent property of a circuit generated from organized sensory afferents, local circuit activity, and modulatory inputs, they serve as a readout of circuit function in the CNS (Tiesinga et al. 2008). Previous work has shown that inhibitory inputs shape spatial receptive fields (Fagiolini et al. 2004; Gabernet et al. 2005; Hensch 2005; Hicks and Dykes 1983; Liu et al. 2009; Morales et al. 2002; Sillito 1975; Tao and Poo 2005), possibly by affecting a balance of I/E synaptic inputs (Farrant and Nusser 2005; Hensch 2005; Zhang et al. 2008). Here, we tested directly whether increasing or decreasing I/E affects spatial receptive fields. With the use of a chamber that allows us to map receptive fields without physically removing the lens from the retina, we recorded spatial receptive fields sequentially in cell-attached mode and in whole-cell mode at −60 mV and 0 mV in response to light-off visual stimuli presented in a pseudorandom pattern across the 64 positions of an 8 × 8 grid (Fig. 8, A–C). Gray-scale spatial receptive field maps of the action potential activity and synaptic currents over a 600-ms window following stimulus onset are generated from an average of two to three repetitions of the stimuli (Fig. 8D). In the example in Fig. 8, cell-attached recordings reveal a highly focused spiking receptive field (sRF), in which visual stimulation in one square of the visual field produces strong action potential activity in the cell, whereas spiking activity in response to the same stimulus in immediately adjacent areas of the visual field is much lower. By contrast, whole-cell recordings taken at −60 mV show that the neuron receives excitatory synaptic inputs in response to light-off stimulation over a much larger area of the visual field (Fig. 8D). We call this the eRF. Similarly, the iRF, recorded in whole-cell mode at 0 mV, is larger than the sRF (Fig. 8D). The recordings show that tectal cells receive excitatory and inhibitory synaptic inputs over a 600-ms window in response to stimuli presented in every portion of the visual field, but the output of the cell reports visual input over a subset of the visual field.
To test whether decreasing the ratio of I/E synaptic transmission onto tectal neurons affects the cell's spatial receptive fields, we recorded spikes, first in cell-attached mode, followed by whole-cell recordings of excitatory and inhibitory synaptic inputs from neurons transfected with ICL and EGFP or EGFP alone. We also tested whether an increasing ratio of I/E synaptic transmission onto tectal neurons affects spatial receptive fields by recording from neurons in animals exposed to DZ. Two examples each are shown for EGFP-expressing control neurons, ICL-expressing neurons, and neurons from DZ-treated animals (Fig. 9A). sRFs in ICL-expressing neurons are significantly larger than sRFs in control neurons, whereas sRFs are significantly smaller in DZ-treated neurons compared with controls (Fig. 9, A–C). The cumulative distribution of sRF sizes in ICL-expressing neurons is shifted significantly toward larger receptive fields compared with EGFP-expressing neurons (Fig. 9B). Conversely, the cumulative distribution of sRF sizes in DZ-treated neurons is shifted toward smaller values (Fig. 9B). Similarly, the mean sRF size of control neurons is greater than the mean sRF size of DZ-treated neurons and smaller than the mean sRF size of ICL-expressing neurons (Fig. 9C).
Considering the synaptic inputs, we find that eRFs are significantly larger in ICL-expressing cells than in controls (Fig. 9D). The relative size of the iRF is not different among ICL-expressing neurons, DZ-treated neurons, and controls (Fig. 9E), although iRFs are smaller in ICL-expressing neurons and larger in DZ-treated neurons, because RF measurements are generated from responses normalized to the maximum response. Since the iRF size indicates the spatial extent of feed-forward inhibitory inputs (Akerman and Cline 2006), these data suggest that the convergence of inhibitory inputs onto the recorded neurons is unaffected by ICL expression or DZ, although the strength of synaptic inputs is affected by the treatment.
Plots of the sizes of the eRFs vs. sRFs show that even though receptive fields of control neurons are smaller than ICL neurons, the sizes of sRFs and eRFs are similarly correlated in control and ICL neurons (Fig. 9F). Although the sizes of sRFs and iRFs are correlated in control neurons, they are not well correlated in ICL neurons (Fig. 9G). Furthermore, the sizes of eRFs and iRFs are correlated in control neurons, but they are not well correlated in either ICL-expressing or DZ-treated neurons (Fig. 9H). These data indicate that partially decreasing or increasing the ratio of I/E synaptic inputs in tectal neurons is sufficient to disrupt the information processing in the tectal circuit that is represented as spatial receptive fields. To determine the input-output correlations from the receptive field mapping, we plotted the maximum spikes from the 8 × 8 grid from each neuron, determined by cell-attached recording, and the corresponding eCSCs (Fig. 9I) or iCSCs (Fig. 9J), determined by whole-cell recordings. ICL-expressing neurons fire more action potentials relative to synaptic inputs compared with control neurons, consistent with data presented in Fig. 6A.
To determine the degree to which excitatory and inhibitory inputs are correlated in response to visual stimuli, we plotted the charge transfer of the iCSCs and eCSCs over the 8 × 8 grid from each neuron (Figs. 9K). The currents of iCSCs and eCSCs are well correlated in EGFP- and ICL-expressing and DZ-treated neurons. The regression line between iCSCs and eCSCs in ICL-expressing neurons is lower than controls, consistent with a lower I/E, as shown in Fig. 5C. In DZ-treated neurons, the linear trend line is shifted toward larger values of iCSCs than controls, consistent with a higher I/E in these neurons. This analysis indicates that ICL expression and DZ treatment shift the balance of excitation to inhibition in the CSC inputs recorded with receptive field mapping.
These data demonstrate that spatial receptive fields are sensitive to increases and decreases in inhibition and suggest that both increases and decreases in I/E affect spatial receptive fields. The larger subthreshold iRFs and eRFs relative to the sRF in control neurons and the change in receptive field size in response to changes in I/E may reflect the capacity for receptive field plasticity seen with sensory training protocols or learning paradigms (Dong et al. 2009; Engert et al. 2002; Gandhi et al. 2008; Moore et al. 1999; Mu and Poo 2006; Richards et al. 2010; Vislay-Meltzer et al. 2006).
I/E affects temporal response properties of spatial receptive fields.
We recorded spatial receptive field responses over a 600-ms window after the onset of the stimulus at each location in the visual field. This allowed us to determine the temporal features of responses at each of the 64 stimulus locations. To test whether ICL expression or DZ affects temporal parameters of spatial receptive field responses, we generated maps of the sRFs over intervals of 100 ms for the 600-ms recording period, shown for three representative neurons in Fig. 10A. Control neurons generate spikes encoding spatial receptive field information within a 100-ms window following stimulus onset, whereas sRFs in ICL-expressing and DZ-treated neurons can be recorded over larger proportions of the 600-ms period (Fig. 10A). Plotting the temporal distribution of normalized spike numbers for the representative spatial receptive fields in Fig. 10A shows that the receptive field in the control is rapid and temporally concise (Fig. 10B). The ICL-expressing example fires more spikes than the control neuron. The onset of the spatial receptive field response in the ICL-expressing neuron shown is delayed relative to the control and persists over a 600-ms window. In DZ, the onset of the response is also delayed, and the cell fires fewer action potentials, but they extend over a larger temporal window than controls. We determined the FWHM as a measure of response duration (Lesica et al. 2007). The mean FWHM of responses in ICL-expressing neurons is significantly greater than control neurons (Fig. 10C), consistent with the idea that the prolongation of spatial receptive field information is due to recurrent excitatory activity in the tectal circuit. The mean FWHM of responses in DZ was also significantly greater than control neurons (Fig. 10C), suggesting that increased GABAergic transmission and an increased I/E ratio disrupt the temporal response properties of visual information processing. We explored this idea further by plotting mean spike latency for the groups of cells recorded (Fig. 10D). Consistent with the spike latencies of the individual representative cells shown in Fig. 10, A and B, the mean spike latencies of ICL and DZ cells are greater than control neurons. ICL expression increases the total number of spikes fired over 600 ms in response to visual stimulation (Fig. 10E). Poststimulus time histograms show that the temporal distribution of action potentials in ICL-expressing neurons is longer than control neurons, whereas DZ-treated neurons fire less than control neurons, but their activity continues longer (Fig. 10F). These data indicate that I/E controls spatial and temporal properties of tectal cell visual responses by regulating the recruitment of local tectal circuitry following visual stimulation.
ICL increases the synaptic integration window.
Temporal receptive fields are the period over which synaptic inputs are integrated to generate neuronal output. Regulation of temporal receptive fields therefore controls synaptic integration related to information processing and learning rules (Pouille and Scanziani 2001). As an independent measure of the function of inhibition in controlling temporal response properties, we calculated temporal receptive fields of visual responses using a spike-triggered average or reverse correlation to a whole-field Gaussian white noise flicker of luminance steps. The FWHM of temporal receptive fields of ICL-expressing cells determined from cell-attached or whole-cell recording is significantly longer than control cells (Fig. 11, A–D). Furthermore, comparing the FWHM values from cell-attached recordings with whole-cell recordings indicates that the temporal integration window of synaptic inputs closely matches the temporal receptive field of the cells' output spiking activity. Finally, the τdecay of the excitatory synaptic temporal receptive fields of ICL-expressing neurons is significantly greater than in control neurons (EGFP: 40.8 ± 5.0 ms, n = 11; ICL: 66.0 ± 9.6 ms, n = 15; P < 0.05). Together, these results indicate that ICL expression lengthens the synaptic integration window.
Visually guided avoidance behavior.
The optic tectum receives and processes visual information and transforms sensory inputs to outputs, some of which, in turn, regulate behavior mediated by circuits in the hindbrain and spinal cord (Orger et al. 2008). We have found that the temporal precision of visually evoked tectal cell spiking activity (Fig. 7), temporal receptive fields (Fig. 11), spatial receptive fields (Figs. 9 and 10), and tectal cell input/output relations is disrupted by experimental manipulations that either increase or decrease inhibition and alter the ratio of inhibition to excitation by 30–50% of control I/E values. These data suggest that multiple aspects of visual responses in the optic tectal neurons are disrupted when I/E is modified. We next tested whether a partial decrease or increase in I/E affects tectally mediated, visually guided behavior using a visual avoidance behavior assay modified from Aizenman and colleagues (see Dong et al. 2009).
The behavior assay tests whether animals avoid visual stimuli by determining their swimming behavior in response to visual stimuli. Animals turn away from moving spots of certain sizes that approach their eyes at approximately a right angle (Fig. 12A). Data are presented as an avoidance index or the fraction of times that individual animals respond to the stimulus out of 10 encounters with the moving spot stimuli. Control (untransfected) and EGFP-transfected tadpoles selectively avoid moving spots corresponding to diameters larger than 0.4 cm (about as big as their head; Fig. 12B). Avoidance of stimuli of 0.4 and 0.04 cm shows the maximal and baseline responses in control animals, respectively, which are significantly different from one another (Mann-Whitney test, P = 0.014). Animals electroporated with ICL or γ2-MO show deficits in visual avoidance behavior, whereas the animals electroporated with mICL or Ctrl-MO are comparable with controls (Fig. 12, C and D).
To test whether DZ affects visual avoidance behavior, 12 tadpoles were tested sequentially in normal Steinberg's solution, after 1 and 2 h exposure to 10 μM DZ, after 1 h exposure to 30 μM DZ, and finally, after 1 h in Steinberg's solution (Fig. 12E). DZ significantly reduces the avoidance response to stimuli of 0.4 and 0.6 cm. The longer 2-h exposure to 10 μM DZ did not cause a further reduction in the visual avoidance response; however, 30 μM DZ completely blocked the avoidance response, which recovered after 1 h in Steinberg's solution. DZ did not affect the time animals spent swimming (Fig. 12F) and optomotor responses (data not shown), suggesting it does not affect responses to visual input or locomotor activity. In an effort to limit the potential effect of DZ on the retina or motor systems, we injected 1 mM DZ into the tectal ventricle and tested for visual avoidance behavior 2 h later. Intraventricular DZ blocked the visual avoidance behavior (Fig. 12G), whereas behavior after intraventricular injection of the vehicle, 1% DMSO, was comparable with control (Fig. 12H). These data indicate that manipulations that decrease or increase I/E in tectal neurons interfere with visually guided behavior.
Deficits in information processing, cognition, and behavior are thought to arise from an imbalance of inhibitory and excitatory synaptic inputs (Berg et al. 2007; Rubenstein and Merzenich 2003). We therefore sought to test the functional significance of modulating I/E on tectal cell visual response properties and visually guided behavior. We used molecular genetic reagents to decrease inhibitory synaptic transmission and used DZ, a positive GABAAR allosteric modulator, to increase inhibitory synaptic transmission. Because ICL expression, γ2-MOs, and DZ do not directly affect excitatory synaptic transmission, we interpret our results as evidence that modulating I/E disrupts several measures of circuit function, notably the precision and latency of visually evoked spike timing, control of recurrent activity within the tectal circuit, spatial and temporal receptive fields, input/output relations of tectal cells, and visually guided behavior.
Spike timing precision.
Temporal information coding conveyed by spike timing is largely attributed to the precision of first spike latency (Daw et al. 2006; Johansson and Birznieks 2004; Mainen and Sejnowski 1995; Tiesinga et al. 2008). Temporal precision of spike timing is regulated by feed-forward or feed-back inhibition (Burkhalter 2008; Gabernet et al. 2005; Pouille and Scanziani 2001; Tiesinga et al. 2008), suggesting that inhibition regulates information transfer between neurons. In the tadpole visual system, tectal neurons receive direct excitatory inputs from convergent excitatory retinal ganglion cells and feed-forward inhibition from GABAergic tectal neurons (Akerman and Cline 2006), similar to the configuration of excitatory and inhibitory inputs in other systems (Burkhalter 2008). The onset of feed-forward inhibition curtails excitatory inputs and has three related consequences: 1) the window over which convergent excitatory inputs sum to generate action potentials is narrowed; 2) the window over which action potentials are evoked is shortened; and 3) the temporal precision of spikes in response to repeated stimuli is increased (Akerman and Cline 2006; Kremkow et al. 2010; Mittmann et al. 2005; Pouille and Scanziani 2001; Wehr and Zador 2003).
Our results show that an ∼50% decrease in the I/E ratio in tectal neurons causes more jitter in the first spike latency, increases the synaptic integration window, and increases the length of temporal receptive fields. Conversely, an ∼30% increase in the I/E ratio generates less jitter and increases first spike latency. Together, these observations show that spike timing in response to direct sensory input is sensitive to changes in I/E. First spike latency carries information about stimulus features, including intensity and location of sensory inputs (Heil 2004; Petersen et al. 2002; Tiesinga et al. 2008). Interestingly, we found that the effect of modulating I/E on first spike jitter depends on luminance intensity. The variance in first spike latencies is significantly greater for visual stimuli of low luminance values, whereas responses to higher luminance stimuli have greater temporal fidelity. Similarly, I/E of visually evoked synaptic currents is larger and more variable for low luminance stimuli, suggesting that I/E in tectal neurons can convey information about stimulus intensity. Higher luminance stimuli recruit more inhibitory inputs within the tectal circuit, similar to “veto” responses in visual cortex (Tucker and Fitzpatrick 2006), which in turn, increases the precision and reliability of spikes in visual cortex (Haider et al. 2010). Evidence for selective recruitment of inhibitory inputs with stronger stimuli has been reported in slice preparations of the pyriform cortex and hippocampus (Carvalho and Buonomano 2009; Stokes and Isaacson 2010), suggesting that dynamic changes in I/E in response to input strength may be a general property of neuronal circuits. These results demonstrate that the extent to which excitatory and inhibitory components of tectal circuits are engaged in visual responses reflects the strength of the sensory inputs. They further suggest that the dynamic range of changes in I/E may be essential for information processing in vivo and that mechanisms that limit the I/E dynamic range may impair CNS function.
I/E ratios control input/output relations.
A fundamental challenge of neuroscience is to determine the mechanisms that govern the input/output transformation. In many brain regions, including optic tectum, the magnitudes of excitatory and inhibitory synaptic conductance are balanced and temporally offset (Akerman and Cline 2006; Haider et al. 2010; Liu et al. 2007; Nowak et al. 2009; Richards et al. 2010; Tiesinga et al. 2008; Zhang et al. 2003). Modeling studies and experiments in reduced preparations suggest that the ratio of I/E conductances governs input/output relations of neurons and circuits by affecting, among other parameters, the reliability with which inputs drive spikes in downstream neurons (Carvalho and Buonomano 2009; Kremkow et al. 2010; Marder and Buonomano 2004; Pouille and Scanziani 2001; Tiesinga et al. 2008). Our data provide experimental evidence that modulating I/E disrupts input/output relations of individual neurons and the functional circuit in intact animals. The increased variance in latency to fire spikes and increased recurrent activity when I/E is decreased likely contribute to functional changes in the input/output relation of individual tectal neurons, similar to effects seen with changing the timing of inhibition in cultured hippocampal neurons (Marder and Buonomano 2004; Pouille and Scanziani 2001). It would be interesting to test directly how decreasing or increasing I/E in tectal neurons affects activity in hindbrain and spinal cord neurons that have been shown to trigger swimming in response to visual stimulation (Orger et al. 2008).
Lack of homeostatic control of the I/E ratio.
The balance of I/E is actively established and maintained during circuit development by unknown mechanisms (Davis 2006; Farrant and Nusser 2005; Lee et al. 2010; Liu 2004; Sillito 1975). Large-scale modifications in sensory input, such as sensory deprivation, result in homeostatic modifications in both excitatory and inhibitory synaptic transmission (Farrant and Nusser 2005; Morales et al. 2002; Sillito 1975). Although the strength of excitatory and inhibitory inputs can be modified independently by activity-dependent mechanisms, homeostatic mechanisms may scale the magnitude of changes in synaptic strength to maintain a balance of inhibitory and excitatory inputs (Davis 2006; Liu et al. 2007; Maffei et al. 2006). Despite homeostatic mechanisms that operate to maintain I/E and neuronal firing rate, the relatively moderate and cell-autonomous decrease in inhibition, accomplished by ICL expression or γ2-subunit knockdown, does not trigger homeostatic mechanisms that reset I/E, at least within the time course of our experiments.
Tectal I/E regulates visually guided swimming behavior.
Our data indicate that modulating I/E in the tectum blocks visually guided behavior. They also demonstrate that the tectum is a critical visual processing station for visually guided behavior. Failure to generate an avoidance response may arise from deficits in information processing and distorted spatiotemporal responses in tectal neurons that interfere with generating output from the tectum to the hindbrain and spinal cord. The deficits observed with ICL expression or γ2-subunit knockdown may be a response to decreased inhibitory input over the 2 days prior to our recordings. Time-lapse imaging showed that the tangential extent of tectal cell dendrites was larger than controls 3 days after transfection with ICL but were not different from controls 2 days after ICL transfection (Shen et al. 2009). In addition, ICL-expressing neurons fail to show the visual experience-dependent, structural plasticity seen in control neurons (Shen et al. 2009). The decreased I/E and resultant increased spiking in transfected cells could affect their development and connections within the visual circuit. Several studies have shown that visual experience and activity-dependent synaptic plasticity refine the developing retinotectal circuit and alter visual response properties, indicating that a variety of mechanisms, including spike timing-dependent plasticity, could translate changes in I/E and activity into changes in circuit connectivity and function (Ruthazer and Aizenman 2010).
Both increasing and decreasing I/E impairs visually guided behavior. These experiments are the first to test experimentally the consequences of opposite changes in I/E on neuronal response properties and behavior. The data indicate that increasing and decreasing I/E does not necessarily have opposing effects on neuronal function and visual response properties. The degraded visually guided behavior arises because processing of visual information and input/output relations are disrupted in both experimental conditions and suggests that diverse deficits in circuit function can disrupt behaviors controlled by the visual circuit.
Our transfections result in ICL expression in 10–20% of tectal neurons and MO transfection in ∼50% of tectal cells, suggesting that decreasing I/E in a subset of tectal cells within the retina-tectum-spinal cord circuit is sufficient to degrade behavioral output. Increases in spontaneous and recurrent activity in transfected neurons indicate that activity in other neurons in the tectal circuit is increased and suggest that the changes in visually guided behavior may result from these circuit-wide changes in activity. Several studies indicate that altering the function of a minority of cells in the circuit can compromise circuit function. In rat fear conditioning, decreasing glutamatergic transmission in ∼20% of amygdala neurons blocked the conditioned response (Rumpel et al. 2005), whereas recent reports suggest that increasing activity in subpopulations of neurons, for instance, by expression and activation of channelrhodopsin-2, affects circuit function and behavior (Douglass et al. 2008; Houweling and Brecht 2008; Li et al. 2009; Pagliardini et al. 2011; Tye et al. 2011; Witten et al. 2010).
Approximately 30% of tectal neurons are GABA immunoreactive. We estimate that transfection of 10–20% of tectal neurons results in transfection of a corresponding minority of GABAergic neurons in the tectum. Although decreasing inhibitory input to inhibitory neurons might be expected to increase net inhibition in the tectal circuit, the visually evoked iCSCs in ICL and MO-transfected neurons (Fig. 5) indicate that decreasing the ratio of I/E synaptic inputs to tectal neurons decreases net inhibitory currents following visual stimulation. Therefore, we cannot distinguish a potential contribution of modifying inhibitory input to inhibitory neurons.
I/E is thought to be disrupted in several neurodevelopmental diseases, including autism spectrum disorders (ASD) (Rubenstein and Merzenich 2003). Visual processing is also disrupted in some people with ASD (Annaz et al. 2010; Marco et al. 2011). Our study shows that disrupting I/E in a minority of neurons within a circuit is sufficient to cause deficits in information processing and behavior, suggesting that diseases characterized by errors in information processing could arise from deficits in a subset of neurons within a circuit.
This work was supported by the National Eye Institute (EY011261), the Nancy Lurie Marks Foundation, an endowment from the Hahn Foundation, and Dart Neuroscience, LLC to H. T. Cline.
No conflicts of interest, financial or otherwise, are declared by the author(s).
We thank Massimo Scanziani, Carlos Aizenman, and members of the H. T. Cline lab for helpful discussions.
Present address of J. A. Demas: St. Olaf College, Northfield, MN 55057.
- Copyright © 2011 the American Physiological Society