Professional Documents
Culture Documents
Article
Neuron 90, 1243–1256, June 15, 2016 ª 2016 Elsevier Inc. 1243
Figure 1. Cholinergic Inputs Play a Central
Role in Initiating Responses in DSGCs
during Natural/Low-Contrast Stimulation
Conditions
(A) A schematic of the standard model for direction
selectivity in which DSGCs compute direction
by comparing glutamatergic excitation mediated
by bipolar cells (Glu; red) with GABAergic inhibition
from null-side SACs (null SAC shown in green).
Arrow indicates the preferred direction of the
DSGC.
(B) Diagram of the proposed model in which the
SAC network generates direction selectivity. SAC
dendrites are optimally stimulated by centrifugal
motion (soma-to-dendrite; Euler et al., 2002)
and therefore preferred-side SACs (blue) drive
preferred-direction cholinergic excitation, and null-
side SACs (green) provide null-direction GABAergic
inhibition (and also ACh). As only null-side SACs
make strong synaptic connections with DSGCs
(Briggman et al., 2011), ACh transmission may
occur through paracrine mechanisms (dotted blue
arrows), while GABAergic inhibition acts via con-
ventional synaptic mechanisms (solid green arrow).
The secondary contribution of bipolar cells medi-
ated by NMDARs has not been shown for simplicity.
(C) Spike raster plot in response to the repeated
presentation of a 5 s natural movie clip (100 trials,
presented at 0.1 Hz; a uniform gray background
was presented during the intervals). The spiking
in response to natural stimuli were DS (see Fig-
ure S1). The gray trace (top) shows the changes in
light intensity over the DSGC receptive field. Trials
in the presence of an nAChR antagonist (100 mM
HEX) are indicated in red. The instantaneous
spike rates (estimated through convolution with a
Gaussian kernel, s = 25 ms) averaged over seven
trials are shown below for control (black) and in
nAChR antagonists (red).
(D) The average total number of spikes generated/
trial under different conditions (n = 7; mean ± SEM is
plotted; *p < 0. 001) demonstrates that the potent
block by nAChR antagonists is completely reversible.
(E) The polar plot illustrates the peak ON spike rate evoked in eight directions (three trials) in control (black) and in the presence of an nAChR antagonist (red). The
DSI is indicated by the solid radial line. Example spike trains for both preferred (0 ) and null (180 ) motion in control and in the presence of nAChR antagonist are
illustrated around the polar plot.
(F) Similar to (E) but using low-contrast stimuli (C50).
(G) Plot of the average peak ON spike rate for responses evoked in the preferred direction under conditions shown in (E) and (F) (n = 8; mean ± SEM is plotted;
*p < 0. 001). For extended analysis of DS properties and quantification of the OFF spike rates, see Figure S2.
(Figures 1C and 1D). Natural movies were obtained using a CCD 1982; Fried et al., 2005; Grzywacz et al., 1998; Kittila and Mas-
camera mounted on a mouse’s head while it freely roamed in its sey, 1997; Lee et al., 2010; Lipin et al., 2015; Park et al., 2014;
cage (Froudarakis et al., 2014). Spike recordings were made Weng et al., 2005), possibly due to the differences in the visual
from ON-OFF DSGCs in an isolated whole-mount retina, onto stimuli used. One major difference was that the equivalent
which this movie was projected. DSGCs responded to certain contrast of the movie used here (as estimated by Tadmor and
moving features in the movie (Figure 1C), and rotating the Tolhurst, 2000) is significantly lower than that of high-contrast
movie revealed that indeed these responses were DS (Figure S1, moving bars, spots, or gratings that have been typically used
available online). Interestingly, the application of an nAChR in previous studies. Hence, we next tested the role of nAChRs
antagonist (100 mM hexamethonium; HEX; or 50 mM curare) in mediating responses to moving spot stimuli of varying
strongly and reversibly suppressed DSGC responses, indicating contrast.
a critical role for cholinergic excitation in generating responses Indeed, saturating ‘‘high’’ contrast spots (C100) moving on
to motion in the movie (Figures 1C and 1D). This strong suppres- a featureless background evoked stronger spiking responses
sion of DSGC responses contrasts with the weaker effect of in DSGCs compared to natural stimulation (the maximum peak
nAChR antagonists reported in the literature (Ariel and Daw, firing rate was 206 ± 14 Hz, n = 8, compared to 95 ± 10 Hz,
n = 7 under natural stimulation conditions). Consistent with decreased responses to stimuli below C50, but only reduced
the previous literature, we found that the application of maximal responses by 49% ± 9% (n = 7, p < 0.005; Figure 2B).
nAChR antagonists reduced the DSGC peak spike rate to In addition, the C50 was significantly larger in the presence of an-
the leading edge of full contrast moving spots (250 mm in diam- tagonists (C50 = 24% ± 3% in control; C50 = 41% ± 3% in the
eter, 1 mm/s; Figures 1E and 1G) but did not compromise its presence of nAChR antagonist; n = 7; p < 0.005; Figure 2B). This
ability to code direction (Figures 1E and S2). Similar effects suggested that the direct AMPAR-mediated inputs from bipolar
were observed on the trailing edge response (OFF response), cells contribute less to responses evoked by low- compared to
and measuring spike rate or total spike number did not signifi- high-contrast stimuli. Conversely, when the contrast response
cantly impact our results (Figure S2). function of the cholinergic input to DSGCs was estimated by
When ‘‘low’’ contrast spots that produced 50% of the maximal subtracting the responses measured in nAChR antagonists
response (C50; 110 ± 12 Hz, n = 8) were used, application of from those measured under control conditions (Figure 2A), it
nAChR antagonists completely blocked the spiking response became apparent that cholinergic inputs dominate in the low-
(Figures 1F and 1G). This suggested that cholinergic inputs play contrast regime, accounting for >80% of the total excitatory in-
a central role in mediating DSGC responses in this regime. puts for contrast levels below C50 (Figure 2C).
Thus, it appears that simple low-contrast stimuli are similar to Previous studies have shown that voltage-dependent
natural scenes in their sensitivity to cholinergic antagonists. For NMDARs play a major role in mediating the DSGC’s light re-
simplicity, in the rest of this study we utilized low-contrast moving sponses (Kittila and Massey, 1997; Lee et al., 2010; Poleg-Pol-
stimuli to probe the properties of the DS circuit. Importantly, low- sky and Diamond, 2016), raising the possibility that glutamate
contrast stimuli drive a reliable DS response in DSGCs, indicating signals were mediated by these receptors in the low-contrast
that cholinergic/GABAergic inputs are balanced in such a way regime. To test this hypothesis, synaptic responses evoked
that DS responses are maintained, despite being weakly stimu- by low-contrast stimuli (C50) were measured at different holding
lated (Figures 1F and S2). potentials in the presence of an nAChR antagonist (Figure 3A).
Indeed, synaptic responses measured under these conditions
The Role of nAChRs, AMPARs, and NMDARs in Contrast were highly non-linear, as depicted in the current-voltage (I-V)
Encoding plot (Figure 3C), indicating the involvement of NMDARs. Subse-
To better understand the organization of excitatory inputs to quent addition of an NMDAR antagonist decreased the current
DSGCs, we used voltage-clamp methods to assess the roles amplitudes and linearized the I-V, shifting the reversal potential
of AMPARs, nAChRs (Figure 2), and NMDARs (Figure 3). Surpris- close to 40 mV (Figures 3B and 3D). This indicated that
ingly, we found that these receptors were not activated in parallel the residual component was made up of weak GABAA recep-
when stimulus contrast was varied, giving rise to several unex- tor (GABAR)- and AMPAR-mediated conductances (GABAR
pected consequences, as described below. 2.1 nS; AMPAR 0.6 nS). In contrast, the I-V of the currents
First, we recorded contrast response functions for AMPAR/ blocked by D-AP5 was ‘‘J’’ shaped, characteristic of NMDARs.
nAChR-mediated excitatory postsynaptic currents (EPSCs) in The maximal conductance (12.5 nS at +40 mV; Figure 3E) was
DSGCs (Figures 2A and 2B; VHOLD 60 mV, near ECl; large compared to other ganglion cell types (Buldyrev et al.,
D-AP5 was included in the bath to block NMDARs), which 2012; Manookin et al., 2010). Despite the large NMDA conduc-
were fit using the Naka-Rushton equation (Naka and Rushton, tance, the absence of any spiking activity in the same condi-
1966; Figure 2B). Applying an nAChR antagonist substantially tions in which NMDAR currents were measured indicates that
To validate the linear regression analysis, we compared the that linear regression provides a reliable estimate of the relative
performance of three different models (ACh + AMPA + GABA, contribution of different synaptic receptors underlying DSGC re-
ACh + GABA, or AMPA + GABA) under different experimental sponses to preferred direction motion.
conditions that systematically biased the contribution of the indi-
vidual receptor components. For low-contrast responses, the Spatiotemporal Profiles of ACh and GABA Underlying
errors associated with the fitting procedure (root mean square Low-Contrast Direction Selectivity
error; RMS error) are similar for the three-component model Previous studies have demonstrated SAC-mediated inhibition,
and the ACh + GABA model. The AMPA + GABA model, how- but not excitation, to be spatially asymmetric (Lee et al., 2010;
ever, performed relatively poorly. Together, these results confirm Yonehara et al., 2011). However, these studies relied on artificial
the lack of detectable AMPAR-mediated glutamatergic input optical or electrical methods to stimulate SACs, not capturing
in this regime (Figures 6A–6E). On the other hand, responses their dynamic behavior in the context of the network. Further-
evoked by high-contrast stimuli were best fit with the three- more, these experiments were performed in conditions in which
component model, consistent with the idea that nAChRs, network activity was pharmacologically blocked, casting doubt
AMPARs, and GABARs contribute to the synaptic responses on the physiological relevance of ACh (Taylor and Smith,
(Figures 6F–6J). Both of the two-receptor component models 2012). Thus, the temporal dynamics of ACh and GABA evoked
performed poorly in the high-contrast regime. Furthermore, by moving stimuli remain to be investigated. As we found basis
AMPARs and nAChRs appear to contribute equally to the synap- functions for the nAChRs, GABARs, and AMPARs did not signif-
tic response (Figure 6G), consistent with the pharmacology (Fig- icantly change over the duration of the synaptic response (Fig-
ure 2B). Finally, when we measured responses to high-contrast ure S4), we could reiterate the linear regression procedure
stimuli in the presence of nAChR antagonists, we found throughout the entire light response at 1 ms intervals (Figure 7B).
AMPA + GABA were sufficient to model the synaptic responses This provided a direct estimate of the temporal dynamics of
(Figures 6K–6O). The ACh component usually observed in con- nAChR-, GABAR-, and AMPAR-mediated synaptic responses
trol conditions was completely abolished by nAChR antagonists evoked by moving stimuli (Figure 7A; see Figure S5 for compar-
(n = 6; p < 0. 001, t test). Under these conditions, adding an ACh isons of data to model predictions).
component did not improve the fit, and the ACh + GABA model As noted previously, the contribution of AMPARs under these
provided a poor estimate of the response. Thus, we conclude conditions was minimal (Figures 7B and 7C). Importantly, under
low-contrast stimulation conditions, ACh and GABA signals ity of 1 mm/s). Interestingly, as inhibition increased, cholinergic
differed in several respects during motion in the preferred or excitation ceased, consistent with the idea that cholinergic
null directions. First, during preferred motion, excitation arose release was strongly controlled by GABAergic inhibition (Fried
a few milliseconds before inhibition (40 ± 13 ms delay between et al., 2005). The temporal delay between excitation and inhibi-
peak gACh and gGABA; n = 6 DSGCs; Figures 7B and 7F), which tion results in a net excitation in the preferred direction in a similar
corresponds to a spatial lag of 40 mm (given the stimulus veloc- time window in which spiking was observed in DSGCs (Figures
7B and 7E). Second, during null motion, GABAergic inhibition terns of activity. However, convincing evidence that SACs alone
was significantly stronger than that observed during preferred could generate direction selectivity through mixed transmission
motion (Taylor and Vaney, 2002; Figures 7B and 7D), while was obtained when we measured responses to moving stimuli
cholinergic excitation was approximately equal in both direc- evoked by either photoreceptors or ChR2-expressing SACs in
tions (Figures 7B and 7C), indicating a differential transmission the same DSGC (Figures 8A–8D). In these experiments, after
of ACh and GABA signals to DSGCs. In the null direction, both characterizing the tuning properties of the DSGCs under normal
GABAergic inhibition and cholinergic excitation had a similar stimulation conditions (Figure 8B), photoreceptor output was
time course (Figures 7B and 7G), but since the GABA conduc- blocked using a standard cocktail of glutamate receptor antag-
tance was significantly larger than the cholinergic conductance, onists/agonist (Figure 8C). The stimulus intensity was then
the net activity was inhibitory (Figure 7B; bottom trace). While increased by 5 log units in order to stimulate ChR2 (Figure 8D).
directional inhibition and spatiotemporal offset inhibition/excita- Remarkably, we found that stimulation of SAC-ChR2 rein-
tion have been previously demonstrated to be key factors in stated the DS spiking responses after photoreceptor blockade
shaping direction selectivity (Taylor and Vaney, 2002), what is (Figures 8B–8D). Importantly, each cell’s preferred direction
surprisingly revealed here is that these features can arise from measured in control conditions remained unchanged when
the SAC network alone under physiological conditions, without measured under SAC-ChR2 stimulation (Figures 8E and 8F),
direct bipolar cell input to DSGCs. and SAC-ChR2-mediated direction selectivity was observed in
DSGCs coding a variety of directions (Figure 8F), indicating
Optogenetic Stimulation of SACs Evokes DS Responses that this mechanism was not specific for a particular DSGC
In general, results from optogenetic experiments are difficult to subtype. The observed direction-selective index (DSI; see
interpret because these stimulation techniques are not usually Experimental Procedures) during SAC-ChR2 stimulation was
designed to drive physiologically relevant spatiotemporal pat- increased compared to control (Figure 8G), likely due to a
All procedures were performed in accordance with the Canadian Council on AUTHOR CONTRIBUTIONS
Animal Care and approved by the University of Victoria’s Animal Care Commit-
tee. Experiments were performed using adult Hb9eGFP (RRID: MGI_109160),
S.S. designed and performed most experiments, analyzed results, and
or ChatCre (RRID: MGI_5475195) crossed with reporter mice, Ai9 (RRID:
contributed to writing the paper. A.J.M. and A.H. collected data that contrib-
MGI_3809523) or Ai32 (RRID: MGI_5013789). Retinae were extracted from
uted to Figures 3F and 3G and Figures 1E and 1F, respectively. G.D. and
mouse eyes under dim red light and flat mounted on a piece of filter paper
D.J.S. helped design the conductance analysis/natural stimuli. G.B.A. de-
with a pre-cut window, through which visual stimuli were presented. Superior
signed experiments and wrote the paper.
coding GFP+ DSGCs were identified in the Hb9eGFP retina using two-photon
laser-scanning microscopy techniques (Trenholm et al., 2013a), while other
ACKNOWLEDGMENTS
types of GFP DSGCs were identified by their medium-sized elliptical somata
and by their DS responses. Light stimuli, produced using a digital light projec-
We thank Kerry Delaney, Bob Chow, and Craig Brown for their useful discus-
tor (Hitachi Cpx1, refresh rate 75 Hz), were focused onto the outer segments of
sions, and Laura Hanson and Rishi Vasandani for providing helpful technical
the photoreceptors using the sub-stage condenser. The background lumi-
support. This work was supported by operating grants from the Foundation
nance, measured with a calibrated spectrophotometer (Ocean Optics), was
Fighting Blindness (Canada), Brain Canada, and Canadian Institutes for Health
set to 10 photoisomerisations/s (R*/sec). Visual stimuli created in the Matlab
Research (CIHR MOP 142301) awarded to G.B.A.
environment (Psychtoolbox) were of positive contrasts, ranging between 15%
and 1,000% (Weber contrast). In some experiments, 50–100 ms light flashes
Received: November 2, 2015
using a blue LED (470 nm) were used to drive ChR2 (10 mW/cm2 or 108
Revised: March 9, 2016
R*/sec). All experiments were performed at 35 C–37 C.
Accepted: April 18, 2016
Light-evoked synaptic currents were recorded at membrane potentials be-
Published: May 26, 2016
tween 70 mV and +20 mV, stepped in 5–10 mV increments, and were fit to
the following equation using Matlab’s built-in Levenberg-Marquardt algorithm:
REFERENCES
IðVÞ = gACh fACh ðVÞ ðV ECat Þ + gGABA fGABA ðVÞ ðV ECl Þ + gAMPA
fAMPA ðV ECat Þ; Ariel, M., and Daw, N.W. (1982). Pharmacological analysis of directionally
sensitive rabbit retinal ganglion cells. J. Physiol. 324, 161–185.
where I is the instantaneous total current at the membrane voltage V,
Borst, A., and Euler, T. (2011). Seeing things in motion: models, circuits, and
Ecat the cationic reversal potential = 0 mV, and ECl the chloride reversal
mechanisms. Neuron 71, 974–994.
potential = 56mV. fACh, fGABA, and fAMPA are basis functions for these recep-
tors, normalized to 1 at 60 mV (described below). gACh, gAMPA, and gGABA Brecha, N., Johnson, D., Peichl, L., and Wässle, H. (1988). Cholinergic ama-
denote the conductance mediated by nAChR, AMPAR, and GABARs, respec- crine cells of the rabbit retina contain glutamate decarboxylase and gamma-
tively, estimated by the fitting procedure. The fitting was performed at a time aminobutyrate immunoreactivity. Proc. Natl. Acad. Sci. USA 85, 6187–6191.
interval of 1 ms (Figure 7) or over a 30–50 ms period (Figures 5 and 6). Briggman, K.L., Helmstaedter, M., and Denk, W. (2011). Wiring specificity in
The Woodhull equation, which well describes the voltage-dependent inward the direction-selectivity circuit of the retina. Nature 471, 183–188.
rectification of nAChRs (Haghighi and Cooper, 1998), was used to define its Buldyrev, I., Puthussery, T., and Taylor, W.R. (2012). Synaptic pathways that
basis function (fACh), shape the excitatory drive in an OFF retinal ganglion cell. J. Neurophysiol.
Gmax 107, 1795–1807.
fACh ðVÞ =
Cafaro, J., and Rieke, F. (2010). Noise correlations improve response fidelity
1 + ½S
kd
and stimulus encoding. Nature 468, 964–967.
Carandini, M., and Ferster, D. (2000). Membrane potential and firing rate in cat
VzdF
kd = kdð0Þ e RT ; primary visual cortex. J. Neurosci. 20, 470–484.
Cohen, E.D., and Miller, R.F. (1995). Quinoxalines block the mechanism of
where Gmax is the maximum conductance, [S] is the internal concentration of
directional selectivity in ganglion cells of the rabbit retina. Proc. Natl. Acad.
spermine (100 mM), kd is the dissociation constant at a given voltage V, kd (0)
Sci. USA 92, 1127–1131.
is the dissociation constant at 0 mV estimated as 10.66 mM, z is the valence
of spermine assumed to be 3.8 at physiological pH (Haghighi and Cooper, Demb, J.B. (2007). Cellular mechanisms for direction selectivity in the retina.
1998), d is the fraction of the membrane electrical field sensed by spermine Neuron 55, 179–186.
from the intracellular side and was estimated to be 48.4%, F is the Faraday Diamond, J.S., and Copenhagen, D.R. (1993). The contribution of NMDA and
constant, and R and T are the universal gas constant and absolute tempera- non-NMDA receptors to the light-evoked input-output characteristics of retinal
ture, respectively. The GABAR basis function (fGABA) was described by an ganglion cells. Neuron 11, 725–738.
Froudarakis, E., Berens, P., Ecker, A.S., Cotton, R.J., Sinz, F.H., Yatsenko, D., Nichols, C.W., and Koelle, G.B. (1968). Comparison of the localization of
Saggau, P., Bethge, M., and Tolias, A.S. (2014). Population code in mouse V1 acetylcholinesterase and non-specific cholinesterase activities in mammalian
facilitates readout of natural scenes through increased sparseness. Nat. and avian retinas. J. Comp. Neurol. 133, 1–16.
Neurosci. 17, 851–857. Park, S.J., Kim, I.J., Looger, L.L., Demb, J.B., and Borghuis, B.G. (2014).
Gillespie, D.C., Kim, G., and Kandler, K. (2005). Inhibitory synapses in the Excitatory synaptic inputs to mouse on-off direction-selective retinal ganglion
developing auditory system are glutamatergic. Nat. Neurosci. 8, 332–338. cells lack direction tuning. J. Neurosci. 34, 3976–3981.
Greene, M.J., Kim, J.S., and Seung, H.S.; EyeWirers (2016). Analogous Patneau, D.K., and Mayer, M.L. (1990). Structure-activity relationships for
convergence of sustained and transient inputs in parallel on and off pathways amino acid transmitter candidates acting at N-methyl-D-aspartate and quis-
for retinal motion computation. Cell Rep. 14, 1892–1900. qualate receptors. J. Neurosci. 10, 2385–2399.
Grzywacz, N.M., and Amthor, F.R. (1993). Facilitation in ON-OFF directionally Pei, Z., Chen, Q., Koren, D., Giammarinaro, B., Acaron Ledesma, H., and Wei,
selective ganglion cells of the rabbit retina. J. Neurophysiol. 69, 2188–2199. W. (2015). Conditional knock-out of vesicular GABA transporter gene from
starburst amacrine cells reveals the contributions of multiple synaptic mech-
Grzywacz, N.M., Amthor, F.R., and Merwine, D.K. (1998). Necessity of acetyl- anisms underlying direction selectivity in the retina. J. Neurosci. 35, 13219–
choline for retinal directionally selective responses to drifting gratings in rabbit. 13232.
J. Physiol. 512, 575–581.
Poleg-Polsky, A., and Diamond, J.S. (2011). Imperfect space clamp permits
Gutiérrez, R., Romo-Parra, H., Maqueda, J., Vivar, C., Ramı̀rez, M., Morales, electrotonic interactions between inhibitory and excitatory synaptic conduc-
M.A., and Lamas, M. (2003). Plasticity of the GABAergic phenotype of the ‘‘glu- tances, distorting voltage clamp recordings. PLoS ONE 6, e19463.
tamatergic’’ granule cells of the rat dentate gyrus. J. Neurosci. 23, 5594–5598.
Poleg-Polsky, A., and Diamond, J.S. (2016). NMDA receptors multiplicatively
Haghighi, A.P., and Cooper, E. (1998). Neuronal nicotinic acetylcholine recep- scale visual signals and enhance directional motion discrimination in retinal
tors are blocked by intracellular spermine in a voltage-dependent manner. ganglion cells. Neuron 89, 1277–1290.
J. Neurosci. 18, 4050–4062.
Root, D.H., Mejias-Aponte, C.A., Qi, J., and Morales, M. (2014). Role of gluta-
Keeley, P.W., Whitney, I.E., Raven, M.A., and Reese, B.E. (2007). Dendritic matergic projections from ventral tegmental area to lateral habenula in aver-
spread and functional coverage of starburst amacrine cells. J. Comp. sive conditioning. J. Neurosci. 34, 13906–13910.
Neurol. 505, 539–546.
Sagdullaev, B.T., McCall, M.A., and Lukasiewicz, P.D. (2006). Presynaptic
Kerchner, G.A., and Nicoll, R.A. (2008). Silent synapses and the emergence of inhibition modulates spillover, creating distinct dynamic response ranges of
a postsynaptic mechanism for LTP. Nat. Rev. Neurosci. 9, 813–825. sensory output. Neuron 50, 923–935.
Kim, J.S., Greene, M.J., Zlateski, A., Lee, K., Richardson, M., Turaga, S.C., Saunders, A., Granger, A.J., and Sabatini, B.L. (2015). Corelease of acetylcho-
Purcaro, M., Balkam, M., Robinson, A., Behabadi, B.F., et al.; EyeWirers line and GABA from cholinergic forebrain neurons. eLife 4, http://dx.doi.org/
(2014). Space-time wiring specificity supports direction selectivity in the retina. 10.7554/eLife.06412.
Nature 509, 331–336.
Schmidt, M., Humphrey, M.F., and Wässle, H. (1987). Action and localization of
Kim, T., Soto, F., and Kerschensteiner, D. (2015). An excitatory amacrine cell acetylcholine in the cat retina. J. Neurophysiol. 58, 997–1015.
detects object motion and provides feature-selective input to ganglion cells
Sethuramanujam, S., and Slaughter, M.M. (2015). Properties of a glutamater-
in the mouse retina. eLife 4, http://dx.doi.org/10.7554/eLife.08025.
gic synapse controlling information output from retinal bipolar cells. PLoS ONE
Kittila, C.A., and Massey, S.C. (1997). Pharmacology of directionally selective 10, e0129133.
ganglion cells in the rabbit retina. J. Neurophysiol. 77, 675–689. Shabel, S.J., Proulx, C.D., Piriz, J., and Malinow, R. (2014). Mood regulation.
Kolb, H., and Famiglietti, E.V. (1974). Rod and cone pathways in the inner GABA/glutamate co-release controls habenula output and is modified by an-
plexiform layer of cat retina. Science 186, 47–49. tidepressant treatment. Science 345, 1494–1498.
Kostadinov, D., and Sanes, J.R. (2015). Protocadherin-dependent dendritic Stafford, B.K., Manookin, M.B., Singer, J.H., and Demb, J.B. (2014). NMDA
self-avoidance regulates neural connectivity and circuit function. eLife 4, and AMPA receptors contribute similarly to temporal processing in mamma-
http://dx.doi.org/10.7554/eLife.08964. lian retinal ganglion cells. J. Physiol. 592, 4877–4889.
Krishnaswamy, A., Yamagata, M., Duan, X., Hong, Y.K., and Sanes, J.R. Strang, C.E., Long, Y., Gavrikov, K.E., Amthor, F.R., and Keyser, K.T. (2015).
(2015). Sidekick 2 directs formation of a retinal circuit that detects differential Nicotinic and muscarinic acetylcholine receptors shape ganglion cell response
motion. Nature 524, 466–470. properties. J. Neurophysiol. 113, 203–217.