To eliminate reciprocal GABAergic inhibition among neighboring SACs, we used a conditional knockout (KO) mouse line in which Slc32a1 is selectively removed from SACs (Slc32a1 KO) to block GABA release from SACs. This line contains homozygous floxed Slc32a1 alleles (acronym: Slc32a1^flox/flox) to replace the endogenous Slc32a1 gene, and a choline acetyltransferase (Chat)-IRES-Cre knockin allele for SAC-specific Cre expression (acronym: C). A floxed tdTomato transgene (acronym: T) was also included when SACs were targeted for electrophysiological recordings (Figure 1C). We have previously shown that these KO mice display disrupted GABAergic inhibition from SACs to DSGCs and no detectable developmental compensation (Pei et al., 2015). To confirm that GABAergic synapses between neighboring SACs are also lost in these mice, we performed paired recordings from neighboring SACs in the ganglion cell layer with intersoma distance of 50~100 um. Evoked GABAergic inhibitory postsynaptic currents (IPSCs) were measured from one SAC at holding potential of −80 mV while the other SAC was depolarized to 0 mV for 20 ms in voltage clamp configuration. We detected reciprocal inhibition between SACs in 9 out of 9 pairs in the control group (Figure 1D, mean peak amplitude 16.4 ± 2.8 pA, five mice), consistent with previous studies (Kostadinov and Sanes, 2015; Lee and Zhou, 2006). We did not detect any evoked IPSCs in the KO group (Figure 1E, n = 9 pairs, three mice), indicating that reciprocal SAC-SAC inhibition is abolished in Slc32a1 KO mice.

Next, we examined centrifugal direction selectivity of SAC dendrites in these KO mice using two-photon imaging of the calcium indicator GCaMP6. We imaged calcium transients in varicosities located in the furthest distal 20 μm of GCaMP6-expressing SAC dendrites while a bright bar moved against a homogeneous dark background (Figure 2A). We term this stimulus "simple moving bar stimulus”. As expected, the leading edge of the bright bar evoked calcium transients in On SACs and the trailing edge evoked responses in Off SACs (Figure 2B, and Videos 1 and 2). We used the peak amplitude of the calcium transients in varicosities as a measure of the strength of dendritic activation. We quantified direction selectivity of responses using the direction selectivity index (DSI, see Experimental methods) and the vector sum. In control mice, dendritic calcium signals in both On and Off SACs show strong directional tuning to motion in the centrifugal direction (Figure 2B and C). In Slc32a1 KO mice, centrifugal direction selectivity (both DSI and vector sum) of On SACs is similar to that of control mice (Figure 2D). However, Off SACs exhibit a small, but statistically significant, increase in DSI and vector sum values (Figure 2F) due to decreased centripetal direction response (Figure 2—figure supplement 1). Adding the GABA_A receptor antagonist SR95531 to Slc32a1 KO retinas reduces these direction selectivity parameters for both On and Off SACs (Figure 2D–2G), indicating a role for GABAergic inhibition of SACs by non-SAC amacrine cells. Our results demonstrate that, for a simple moving bar stimulus, losing reciprocal SAC-SAC inhibition in Slc32a1 KO mice does not affect the direction selectivity of On SACs, and enhances the direction selectivity of Off SACs.

Figure 2 with 1 supplement see all

Download asset Open asset

Video 1

Download asset

Video 2

Download asset

The GABAergic inhibition required for generating the centrifugal direction selectivity of On and Off SACs could come from two sources besides other SACs. The first is the GABAergic synapse onto presynaptic bipolar cells from other amacrine cell types (Hoggarth et al., 2015; Lee and Zhou, 2006). The second is direct GABAergic input onto SACs from non-SAC amacrine cells (Ding et al., 2016). To distinguish between these two possibilities, we examined the direction selectivity of SACs when GABA_A receptors are removed from SACs but not from bipolar cells. If presynaptic inhibition from non-SAC amacrine cells onto bipolar cells is sufficient for direction selectivity, we would expect normal direction selectivity under this condition. In comparison, if non-SAC-mediated GABAergic inputs directly onto SACs are important, we would expect impaired direction selectivity.

We eliminated all direct GABAergic inhibition onto SACs by generating a conditional KO mouse line in which the α2 subunit of GABA_A receptors, Gabra2, is selectively knocked out from SACs (Gabra2 KO mice, Figure 3A). Gabra2 expression colocalizes with On and Off SAC dendrites in the IPL (Brandstätter et al., 1995) and has been shown to be important for direction selectivity (Auferkorte et al., 2012). For targeted recording from these KOs, we also labeled SACs and a subtype of On-Off DSGCs that prefer motion in the posterior direction (pDSGCs) (Huberman et al., 2009) with tdTomato and GFP, respectively (Figure 3A). To confirm loss of functional GABA_A receptors in SACs of Gabra2 KO mice, we recorded their spontaneous IPSCs and found that they are eliminated (Figure 3B and C). We also performed paired recordings from neighboring SACs, and found that none of the pairs showed evoked IPSCs (Figure 3D, n = 10 pairs, three mice). Therefore, GABAergic inputs onto SACs are abolished in Gabra2 KO mice.

Figure 3

Download asset Open asset

Although multiple studies have demonstrated that the emergence of retinal direction selectivity occurs independent of neural activity (Elstrott et al., 2008; Hamby et al., 2015; Sun et al., 2011; Wei et al., 2011), we needed to rule out potential developmental changes in this circuit in Gabra2 KO mice. Therefore, we examined other critical synapses involved in direction selectivity: the GABAergic and cholinergic synapses between SACs and DSGCs. Using paired voltage-clamp recording, we depolarized SACs and measured evoked GABAergic IPSCs and cholinergic excitatory postsynaptic currents (EPSCs) in pDSGCs. We targeted SAC-pDSGC pairs with overlapping dendritic arbors and intersoma distances of 60–80 µm. In control mice, SACs from the null side provide strong GABAergic inputs onto pDSGCs while SACs from the preferred side provide weak GABAergic inputs. In Gabra2 KOs, this asymmetric wiring pattern remains unchanged (Figure 3E and F). Similarly, cholinergic transmission in the KO is unaffected (Figure 3E and F). Thus our results show that GABAergic and cholinergic synapses from SACs to DSGCs are not altered in the Gabra2 KO mice. In addition, the amplitude of EPSCs in pDSGCs evoked by a bright flashing spot is similar between control and Gabra2 KO groups (Figure 3G), indicating that both cholinergic and glutamatergic synapses onto pDSGCs are unaltered.

How do SACs respond to visual motion when their direct ionotropic GABAergic inhibitory inputs are lost? To address this question, we measured centrifugal direction selectivity by monitoring GCaMP6 fluorescence in On and Off SACs of Gabra2 KO mice during the simple moving bar stimulus described above. We found that Off SACs in these KOs display a significant increase in centripetal-direction response (Figure 4A and B, Video 3), resulting in reduced direction selectivity (Figure 4C–4F). Therefore, direct GABAergic inputs onto Off SACs from non-SAC amacrine cells are important for centrifugal direction selectivity. In contrast, we found that On SACs in the KOs exhibit centrifugal preference similar to those of control mice (Figure 4, Video 4), suggesting that direct GABAergic inputs are not required for direction selectivity in On SACs during the simple moving bar stimulus. Bath application of SR95531 reduced direction selectivity of On SACs in KOs (Figure 4—figure supplement 1), suggesting a role for presynaptic inhibition onto bipolar cells in generating direction selectivity.

Figure 4 with 1 supplement see all

Download asset Open asset

Video 3

Download asset

Video 4

Download asset

Since centrifugal preference of Off SAC dendrites is selectively impaired in Gabra2 KO mice, we would expect that the Off component of inhibitory inputs onto DSGCs would also display reduced direction selectivity. We tested this hypothesis by performing whole-cell voltage clamp recordings in pDSGCs to measure light-evoked IPSCs during the simple moving bar stimulus. To quantify direction selectivity of the inhibitory inputs onto DSGCs, we calculated DSI and vector sum of the peak amplitude and total charge transfer of IPSCs in response to the simple moving bar stimulus. In control mice, both On and Off components of IPSCs are direction selective (Figure 5A–5D). In Gabra2 KO mice, the Off component of IPSCs shows reduced selectivity (Figure 5C and D) but the On component does not (Figure 5A and B). Therefore, in the absence of GABAergic inputs onto SACs, the Off component of inhibitory inputs onto DSGCs displays impaired directional tuning.

Figure 5

Download asset Open asset

To test if loss of lateral inhibition onto Off SACs impacts the ultimate output of the retinal direction selective circuit, we determined its effect on the spiking activity of DSGCs. We recorded the spiking activity of pDSGCs in loose-patch configuration, and compared the direction selectivity of the leading edge-evoked On response and the trailing edge-evoked Off response in control and Gabra2 KO mice during the simple moving bar stimulus. Consistent with the reduced directional tuning of IPSCs observed in the Off component, we found that the Off response of pDSGC spiking activity in Gabra2 KO mice shows less direction selectivity than that in control mice (Figure 6A and C). In contrast, the On responses in KOs display direction selectivity similar to that of controls (Figure 6A and B). To rule out the possibility that the Off response is influenced by prior activation of the On pathway, we reversed the contrast of the moving bar stimulus by presenting a dark bar against a bright background. We found that direction selectivity of the leading edge-evoked Off component in Gabra2 KOs was again less than that in controls (Figure 6D and F), while the trailing edge-evoked On response of KOs shows direction selectivity similar to controls (Figure 6D and E). Therefore, the Off component of pDSGC spiking activity becomes less directionally tuned in Gabra2 KO mice.

Figure 6

Download asset Open asset

In the experiments above, we used a simple moving bar stimulus and found that direction selectivity of the Off pathway requires lateral inhibition onto Off SACs, but direction selectivity of the On pathway does not require direct GABAergic input onto On SACs. What, then, is the role of lateral inhibition in the On pathway? We hypothesized that our stimulus conditions did not optimally activate this inhibitory pathway. In particular, a noisy background may more effectively engage this inhibition and reveal its functional importance in direction selectivity. To test this idea, we measured the direction selectivity of pDSGCs in Gabra2 KO mice using a bright bar moving against a flickering checkerboard background. This flickering checkerboard provides "white noise" that is commonly used for mapping receptive fields in the spike-triggered average method (Chichilnisky, 2001). We set the brightness of this moving bar equal to that of our simple moving bar stimulus, and varied the brightness of the checker pattern from 0% to 100% of the moving bar intensity. In control mice, the pDSGC spiking response maintains direction selectivity over a wide range of checker intensities (Figure 7A–7C). Consistent with the earlier finding that lateral inhibition is crucial for direction selectivity of the Off component (Figure 6C and E), the direction selectivity of the Off response in Gabra2 KO mice is reduced regardless of checker intensity (Figure 7A and C). Interestingly, however, replacing the gray background with a flickering checkerboard also significantly reduces the On direction selectivity in KO mice (Figure 7A and B), indicating a functional contribution of lateral inhibition during more complex stimuli.

Figure 7

Download asset Open asset

The flickering checkerboard not only adds noise to the visual stimuli, but also changes the average background illuminance and therefore the effective contrast of the moving bar. To test if changing either ambient illuminance or contrast alone can mimic the effect of introducing the flickering checkerboard, we measured the direction selectivity of pDSGCs in Gabra2 KO mice using the simple moving bar stimulus with different contrast and background illuminance levels that lie in the upper and lower range of the flickering checkerboard stimuli. We found that, similar to the results with our original simple bar stimulus (Figure 6), the On component of the pDSGC's spiking response displays normal direction selectivity, while the Off component shows reduced direction selectivity at all illuminance and contrast levels tested (Figure 7D). Therefore, the impaired direction selectivity seen in the On pathway in Gabra2 KO mice with the flickering checkboard background is not due to changes in stimulus contrast or ambient illuminance. Instead, the noise introduced by the flickering checkerboard is important for the functional recruitment of lateral inhibition in the On pathway.

Although using conditional KO mice offers the unique advantage of achieving the cell type and synapse specificity, a major concern of this genetic approach is that neural circuits may potentially be altered due to developmental compensations. In our study, we selectively remove from SACs the two genes involved in GABAergic transmission, Slc32a1 and Gabra2, during the early postnatal period (Xu et al., 2016). Therefore, we must verify that aside from loss of SAC-mediated GABA release in Slc32a1 KO mice and loss of GABA_A receptors in SACs in Gabra2 KO mice, the rest of the direction selectivity circuit is not altered. To address this issue, we examined the other known synapse types that are involved in retinal direction selectivity, including the cholinergic and GABAergic synapses between SACs and DSGCs, as well as the glutamatergic inputs onto DSGCs from bipolar cells. We found that these synapses showed no detectable changes (Pei et al., 2015) and Figure 3), indicating that removing GABA release or GABA receptors from SACs does not lead to compensatory developmental changes at other synapses in the direction selective circuit. We cannot discount the possibility that developmental compensation has occurred at sites not examined in this study. However, our findings are consistent with multiple previous studies that demonstrate that the early development of retinal direction selectivity is highly resistant to changes in neural activity (Chan and Chiao, 2008; Elstrott et al., 2008; Hamby et al., 2015; Sun et al., 2011; Wei et al., 2011).

After direction selectivity is established, visual experience has been shown to play a role in refining the clustered distribution of preferred directions of On-Off DSGCs (Bos et al., 2016). In both Slc32a1 KO and Gabra2 KO mice, visually evoked responses are present in all retinal neurons including SACs and pDSGCs. Furthermore, pDSGCs in both KO lines remain tuned to the posterior direction (Pei et al., 2015); data not shown). Given that all known synaptic connections in the direction selective circuit and their strengths are preserved in our conditional KO mice, we consider these mouse lines valuable tools for circuit analysis and for testing existing models.

Lateral inhibition plays an important role in feature selection in multiple sensory systems. One elaborated form, the reciprocal inhibition between inhibitory neurons, has been described in multiple brain areas such as retina, thalamus, midbrain and cortex (Lee and Zhou, 2006; Machens et al., 2005; Miller and Wang, 2006; Papadopoulou et al., 2011). In sensory and cognitive systems, lateral inhibition motifs have been implicated in selecting the target feature in an environment when other 'competitor' features are present (Mysore and Knudsen, 2012; Sharpee, 2012). In the visual system, moving objects in natural scenes have backgrounds that are rarely homogeneous and stationary, but rather are spatially and temporally noisy. These can serve as 'competitor' stimuli that pose challenges to the direction selective circuit for reliably detecting motion direction. We postulate that lateral inhibition onto On SACs functions to safeguard direction selectivity against a noisy environment. A noisy background increases activation of bipolar cells, which in turn provides stronger glutamatergic drive to the downstream amacrine cells including SACs. Under this condition, the lateral inhibitory network may be in a more 'primed' state to be efficiently recruited by the motion stimuli and to exert its function. In contrast, amacrine cells involved in lateral inhibition are less activated when bipolar cells adapt to the gray background, and thus the moving bar alone may not be sufficient to significantly recruit lateral inhibition in the On pathway. Consistent with this hypothesis, direction selectivity of the On response of DSGCs in Gabra2 KO mice is normal when the background is homogeneous but deteriorates as soon as white noise is introduced into the background. These findings represent an intriguing example in which additional neural mechanisms are recruited when the visual stimulus more closely resembles natural viewing conditions, but simpler visual stimuli involving only the feature of interest may not reveal the functional significance of these mechanisms (David et al., 2004; Felsen and Dan, 2005; Felsen et al., 2005; Turner and Rieke, 2016). A notable analogy has been reported in the visual cortex: a prominent inhibitory component in the receptive field of visual cortical neurons is uniquely revealed by more complex natural stimuli but is not observed with synthetic sinusoidal gratings (David et al., 2004).

The role of lateral inhibition onto SACs in direction selectivity has been exclusively studied in On SACs using bath application of the GABA_A receptor antagonist SR95531 with variable results (Ding et al., 2016; Euler et al., 2002; Gavrikov et al., 2006; Hausselt et al., 2007; Lee and Zhou, 2006; Oesch and Taylor, 2010; Vlasits et al., 2016). This variability may be due to a number of factors, such as species differences, conditions of visual stimulation, somatic recordings versus calcium imaging at the dendrites, and locations of varicosities along SAC dendrites. Furthermore, the effect of the antagonist cannot be attributed to specific type(s) of GABAergic synapses since GABA_A receptors are extensively expressed in amacrine cells and bipolar cells in the IPL. Here, we combined pharmacology with the genetic dissection of the GABAergic circuitry to determine the synaptic loci that participate in direction selectivity. This approach allowed us to assign functional roles to different inhibitory synaptic components, and to identify a new component in the functional wiring diagram of the direction selective circuit: inhibitory inputs from non-SAC amacrine cells onto Off SACs.

Pharmacological experiments cannot establish the ultimate effect that lateral inhibition onto SACs has on the spiking output of DSGCs because SR95531 disrupts SAC-DSGC inhibition. However, by genetically eliminating GABA_A receptors but not neurotransmitter release from SACs in Gabra2 KO mice, we preserved the critical synaptic connections between SACs and DSGCs, and determined the role that GABAergic inputs onto SACs play in the spiking output of DSGC’s. We found that the direction selectivity of On-Off DSGC spiking decreases when SAC centrifugal preference is reduced. Therefore, the inhibitory mechanisms contributing to the centrifugal preference of SACs that we have studied here are functionally relevant for determining the signals that are relayed from the retina to the brain.

The Gabra2^flox/flox mouse line was a generous gift from Dr. Uwe Rudolph at Harvard Medical School. Slc32a1^flox/flox mice (Slc32a1<tm1Lowl>/J), Chat-IRES-Cre mice (129S6-Chat^tm2(cre)Lowl/J) and floxed tdTomato mice (129S6-Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze}/J) were acquired from the Jackson Laboratory. Drd4–GFP mice were originally developed by MMRRC (http://www.mmrrc.org/strains/231/0231.html) in the Swiss Webster background, and were subsequently backcrossed to C57BL/6 background. All strains were backcrossed to the C57BL/6 background in our laboratory, and crossed to each other to create the lines used in this study. Mice of both sexes between postnatal days 18–35 were used for paired recording experiments, and those between postnatal days 24–35 were used for light response experiments. All procedures to maintain and use mice were in accordance with the University of Chicago Institutional Animal Care and Use Committee (Protocol number ACUP 72247) and in conformance with the NIH Guide for the Care and Use of Laboratory Animals and the Public Health Service Policy.

Mice were anaesthetized with isoflurane and decapitated after dark adaptation. Under infrared illumination, retinas were isolated from the pigment epithelium at room temperature in oxygenated Ames’ medium (Sigma-Aldrich, St. Louis, MO) for visual stimulation experiments or in artificial cerebrospinal fluid (ACSF) containing 119.0 mM NaCl, 26.2 mM NaHCO₃, 11 mM D-glucose, 2.5 mM KCl, 1.0 mM K₂HPO₄, 2.5 mM CaCl₂, and 1.3 mM MgCl₂ for dual patch clamp recording. Isolated retinas were then cut into dorsal or ventral halves and mounted ganglion-cell-layer-up on top of a 1 mm² hole in a small piece of filter paper (Millipore, Billerica, MA). The orientation of the preferred direction (posterior) of Drd4-GFP positive neurons was noted for each piece. Retinas were kept in darkness at room temperature in Ames’ medium or ACSF bubbled with 95% O2/5% CO² until use (0–7 hr).

Recording electrodes of 3–5 MΩ were filled with a cesium-based internal solution containing 110 mM CsMeSO₄, 2.8 mM NaCl, 4 mM EGTA, 5 mM TEA-Cl, 4 mM adenosine 5′-triphosphate (magnesium salt), 0.3 mM guanosine 5′-triphosphate (trisodium salt), 20 mM HEPES, 10 mM phosphocreatine (disodium salt), 5 mM N-Ethyllidocaine chloride (QX314), 0.025 mM Alexa 488 (for SACs) and 0.025 mM Alexa 594 (for pDSGCs), pH 7.25. tdTomato-labeled SACs and GFP-labeled pDSGCs were identified with epifluorescence imaging (X-Cite) or two-photon microscopy (Bruker Nano Surfaces Division, Middleton, WI) under a water immersion objective (60x, Olympus LUMPlanFl/IR). Data were acquired using PCLAMP 10 recording software and a Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA), low-pass filtered at 4 kHz and digitized at a sampling rate of 10 kHz. Light-evoked responses were recorded at a bath temperature of 32–33°C. Spontaneous IPSCs in SACs and paired SAC-SAC and SAC-pDSGC recordings were obtained at room temperature.

During paired SAC-pDSGC recordings, the evoked IPSCs and EPSCs in pDSGCs were isolated by holding the cells at reversal potentials for these conductances (0 mV for GABAergic, −60 mV for cholinergic) in the presence of 0.05 mM D-AP5 and 0.05 mM DNQX disodium salt. The reversal potentials were calculated and then experimentally confirmed by obtaining current-voltage (I-V) relationships of pharmacologically isolated GABAergic and cholinergic currents in pDSGCs. To measure spontaneous IPSCs in SACs and evoked IPSCs in SAC-SAC paired recordings, 0.008 mM DHβE, 0.05 mM D-AP5 and 0.05 mM DNQX disodium salt were bath applied to the retinas to block NMDA-, AMPA/kainite- and nicotinic receptors. Since depolarizing SACs to 0 mV caused poorly controlled voltage-gated conductances at the beginning of the depolarizing step, we clamped the postsynaptic SACs at −80 mV to measure inward chloride currents while all excitatory channels were pharmacologically blocked. Reported potentials have been corrected for the liquid junction potential (~9 mV).

Light-evoked IPSCs in pDSGCs were isolated by holding cells at 0 mV. Three repetitions of raw synaptic traces were recorded and averaged to obtain the mean response for each stimulus condition. The peak amplitude and total charge transfer of IPSCs evoked by the leading and trailing edges of the moving bar were used to calculate the direction selective index (DSI) and vector sum. DSI is defined as ${\displaystyle \left|\frac{P-N}{P+N}\right|}$, where P is the peak amplitude or charge transfer of IPSCs in the preferred direction, and N is that in the null direction.

A white organic light-emitting display (OLEDXL, eMagin, Bellevue, WA; 800 × 600 pixel resolution, 60 Hz refresh rate) was controlled by an Intel Core Duo computer with a Windows seven operating system and was presented to the retina at a resolution of 1.1 μm/pixel. Moving bar stimuli were generated by MATLAB and the Psychophysics Toolbox (Brainard, 1997), and projected through the condenser lens of the two-photon microscope onto the photoreceptor layer. For the flashing spot stimulus in Figure 3G, we used 10 repetitions of a 110 µm radius white spot (2 s black, 2 s white, 2 s black) to test On and Off responses. For the simple moving bar stimulus, a positive-contrast bar (110 µm wide, 385 µm long) moved along the long axis in 8 or 12 pseudo-randomly chosen directions at a speed of 440 µm/sec over a 660µm-diameter field on the retina; and three to five trials were recorded for each direction. The percent stimulus contrast was calculated as (L_stimulus – L_background)/ (L_stimulus + L_background). Unless otherwise noted (i.e., Figure 7D), the intensity of the moving bar was ~6.3×10⁴ isomerizations (R*)/rod/s, lying in the photopic range, and the background intensity was ~1800 R*/rod/s, lying at the lower end of the photopic range. Individual checker size was 55×55 µm with each checker’s intensity equal to either the background (0) or a white value (expressed as a percentage of the moving bar intensity) drawn from a binomial distribution. Checker pattern was updated at 15 Hz.

Genetically encoded calcium indicator CCaMP6m was expressed in sparse populations of On and Off SACs by intravitreal injection of an AAV vector carrying floxed GCaMP6m (University of Pennsylvania Vector Core) into control (C) and KO mice (Gabra2^flox/flox C and Slc32a1 ^flox/flox C). GCaMP6 fluorescence of isolated retinas in oxygenated Ames at 32–33°C was imaged in a customized two-photon laser scanning fluorescence microscope (Bruker Nano Surfaces Division). GCaMP6 was excited by a Ti:sapphire laser (Coherent, Chameleon Ultra II, Santa Clara, CA) tuned to 920 nm, and the laser power was adjusted to avoid saturation of the fluorescent signal. Onset of laser scanning induces a transient On response in On SACs that adapts to the baseline in ~3 s. Therefore, to ensure the complete adaptation of this laser-induced response and a stable baseline, visual stimuli were given after 10 s of continuous laser scanning. To separate the visual stimulus from GCaMP6 fluorescence, a band-pass filter (Semrock, Rochester, MA) was placed on the OLED to pass blue light peaked at 470 nm, while two notched filters (Bruker Nano Surfaces Division) were placed before the photomultiplier tubes to block light of the same wavelength. Imaging was performed in the regions of the retina that contain sparsely labeled SAC varicosities so that individual varicosities could be resolved and the orientation of the dendrites relative to the soma could be determined. The objective was a water immersion objective (60x, Olympus LUMPlanFl/IR). Time series of each imaging window (~20×20 µm) were collected at 30–40 Hz.

Analysis was performed using ImageJ and MATLAB. Circular regions of interest (ROIs) corresponding to individual varicosities and a background region with no GCaMP6 expression were manually selected for each imaging window in ImageJ. The fluorescent time course of each ROI was determined by averaging all pixels within the ROI. The fluorescence of the background region was subtracted from the raw fluorescent signals of the ROIs in the same imaging window at each time frame. The resulting fluorescence measurements were then used to calculate the visually evoked responses in varicosities. The relative fluorescence change in each varicosity was taken as ${\displaystyle \mathrm{\Delta }\mathrm{F}=\frac{F-F_0}{F_0}}$, where F is the peak amplitude of fluorescence and $F_0$ is the baseline fluorescence level. Mean ΔF was obtained by averaging 3–5 trials for each direction. To assess direction selectivity of SAC distal dendrites, we used the vector sum or direction selectivity index (DSI), defined as ${\displaystyle \left|\frac{{\mathrm{\Delta }F}_{cf}-{\mathrm{\Delta }F}_{cp}}{{\mathrm{\Delta }F}_{cf}+{\mathrm{\Delta }F}_{cp}}\right|}$, where ${\displaystyle {\mathrm{\Delta }F}_{cf}}$ is the relative fluorescence change in centrifugal motion and ${\displaystyle {\mathrm{\Delta }F}_{cp}}$ is that in centripetal motion. For statistical analysis, we compared variance across varicosities on the dendrites of the same SAC versus variance across varicosities of different SACs. Since the latter was larger, we averaged DSI and vector sum values of all the varicosities belonging to one SAC to get a single data point, and took N in the statistical analysis to be the number of cells, which approximately equals the number of imaging windows.

The two-photon targeted recording of light responses was performed as in (Pei et al., 2015). Briefly, the retinas were perfused with oxygenated Ames at 32–33°C. Cells were visualized with infrared light (>900 nm) and an IR-sensitive video camera (Watec). Drd4-GFP-positive cells were identified using a two-photon microscope (Bruker Nano Surface Division) and a Ti:sapphire laser (Coherent Chameleon Ultra II) tuned to 920 nm. An electrode of 3–5 MΩ was filled with Ames’ medium for loose patch recordings of spikes. Then the electrode was carefully removed, and a new electrode filled with cesium-based internal solution and 25 μm Alexa 594 was used to fill the recorded cell. An image stack of the Alexa-488 filled pDSGC was acquired with the two-photon microscope at z intervals of 1.5 μm and was resampled three times for each z-plane using a 60x objective (Olympus LUMPlanFl/IR 60x/0.90W). Images were acquired to cover the entire dendritic field of the cell to verify the bistratified dendritic arbor and the cofasciculation with tdTomato-expressing SACs. Rarely, we encountered Drd4-GFP positive neurons that did not show the above two characteristics in both control and knockout groups. These mislabeled cells are not included in the analysis.

Data were analyzed using custom protocols in MATLAB. The number of spikes evoked by the leading (On) and trailing (Off) edges of the bar were counted using MATLAB and averaged across trials in each direction. Spiking DSI is defined as (N_pref− N_null)/ (N_pref + N_null) where N is the number of spikes evoked by the leading or trailing edge of the moving bar.

Grouped data with error bars are presented as mean ± SEM and tested for normality. For Figures 2, 3, 5 and 7, statistical differences were examined using one-way analysis of variance and post hoc comparisons using Student’s t-test with Bonferroni correction. For Figures 4 and 6, a two-sample Kolmogorov–Smirnov test was used and multiple comparisons were corrected by Bonferroni correction.

The funders provide financial support to this manuscript in study design, data collection and interpretation, and the decision to submit the work for publication.

We thank Chen Zhang for managing a mouse colony and intravitreal injections, Dr. Uwe Rudolph from Harvard Medical School for the generous gift of the Gabra2^flox/flox mouse line, and the Genetically-Encoded Neuronal Indicator and Effector (GENIE) Project and the Janelia Research Campus of the Howard Hughes Medical Institute GENIE Program and the Janelia Research Campus, specifically Vivek Jayaraman, Ph.D., Douglas S Kim, Ph.D., Loren L Looger, Ph.D., Karel Svoboda, Ph.D. for the AAV-GCaMP6 vectors. This work was supported by NIH R01 EY024016, E. Matilda Ziegler Foundation Grant, Whitehall Grant, Sloan Research Fellowship, Karl Kirchgessner Foundation Grant and Brinson Foundation Award. The authors declare no competing financial interests.

Animal experimentation: All procedures to maintain and use mice were in accordance with the University of Chicago Institutional Animal Care and Use Committee (Protocol number ACUP 72247) and in conformance with the NIH Guide for the Care and Use of Laboratory Animals and the Public Health Service Policy.

© 2016, Chen et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.