Temporally diverse excitation generates direction selective responses in on and off type retinal starburst amacrine cells

In OFF SACs, type 2 cone bipolar cells prefer-entially contact proximal dendrites, whereas type 3a cone bipolar cells preferentially contact intermediate or distal dendrites further Broa

Temporally Diverse Excitation Generates Direction-Selective Responses in ON- and OFF-Type Retinal Starburst Amacrine Cells

Graphical Abstract

Highlights

d Direct recordings of the motion-evoked responses of OFF

starburst amacrine cells

d ON and OFF starburst amacrine cells receive

spatiotemporally patterned excitation

d Spatiotemporally patterned excitation generates

direction-selective responses

d Direction selectivity is an excitation-driven emergent

receptive field property

Authors

James W Fransen, Bart G Borghuis

Correspondence

bart.borghuis@louisville.edu

In Brief

Fransen and Borghuis use whole-cell electrophysiology to demonstrate that ON- and OFF-type starburst amacrine cells, key players in retinal direction selectivity, receive temporally diverse excitation across their dendritic arbors Integration of this input in model simulations generates direction selective responses as an emergent receptive field property.

Fransen & Borghuis, 2017, Cell Reports18, 1356–1365

February 7, 2017ª 2017 The Author(s)

http://dx.doi.org/10.1016/j.celrep.2017.01.026

Cell Reports Report

Temporally Diverse Excitation Generates

Direction-Selective Responses in ON- and OFF-Type Retinal Starburst Amacrine Cells

James W Fransen1and Bart G Borghuis1 , 2 ,*

1Department of Anatomical Sciences and Neurobiology, University of Louisville School of Medicine, Louisville, KY 40202, USA

2Lead Contact

*Correspondence:bart.borghuis@louisville.edu

http://dx.doi.org/10.1016/j.celrep.2017.01.026

SUMMARY

The complexity of sensory receptive fields increases

from one synaptic stage to the next In many cases,

increased complexity is achieved through

spatio-temporal interactions between convergent

excit-atory and inhibitory inputs Here, we present

evidence that direction selectivity (DS), a complex

emergent receptive field property of retinal starburst

amacrine cells (SACs), is generated by

spatiotem-poral interactions between functionally diverse

excitatory inputs Electrophysiological whole-cell

re-cordings from ON and OFF SACs show distinct

tem-poral differences in excitation following proximal

compared with distal stimulation of their receptive

fields Distal excitation is both faster and more

tran-sient, ruling out passive filtering by the dendrites

and indicating a task-specific specialization Model

simulations demonstrate that this specific

organiza-tion of excitaorganiza-tion generates robust DS responses in

SACs, consistent with elementary motion detector

models These results indicate that selective

integra-tion of spatiotemporally patterned excitaintegra-tion is a

computational mechanism for motion detection in

the mammalian retina.

INTRODUCTION

Direction-selective ganglion cells (DSGCs) in the mammalian

retina were discovered more than 50 years ago (Barlow and

Hill, 1963) Despite an apparently compact circuit architecture

(DSGCs are located just two [excitation] or three [inhibition]

syn-apses from the photoreceptor input), our understanding of the

neural mechanisms underlying direction selectivity (DS) remains

incomplete DS originates in the dendrites of starburst amacrine

cells (SACs) (Euler et al., 2002; Yoshida et al., 2001) and is

trans-mitted to DSGCs through inhibitory synapses onto the DSGC

dendrites (Briggman et al., 2011) Although several mechanisms

have been shown to contribute to direction selective responses

in the SAC dendrites, none has proven essential Thus, the

fundamental computation of visual motion direction remains unresolved

SACs comprise ON- and OFF-center subtypes with distinct anatomy and sensitivity to bright and dark moving stimuli, respectively (Figure 1A;Famiglietti, 1983; Peters and Masland, 1996; Taylor and Wa¨ssle, 1995) ON SAC somas are located in the ganglion cell layer They are readily accessible for electro-physiological whole-cell recording, and their response proper-ties have been studied extensively in rabbit (Dmitriev et al., 2012; Euler et al., 2002; Fried et al., 2002; Gavrikov et al., 2006; Hausselt et al., 2007; Lee et al., 2010; Lee and Zhou, 2006; Taylor and Wa¨ssle, 1995; Zhou and Lee, 2008) and mouse (Ozaita et al., 2004; Pei et al., 2015; Vlasits et al., 2014, 2016; Wei

et al., 2011) ON SACs preferentially respond to radial motion away from their soma (centrifugal) versus toward their soma (centripetal) (Euler et al., 2002; Hausselt et al., 2007; Lee and Zhou, 2006; Oesch and Taylor, 2010) This direction-dependent response asymmetry forms the basis for direction selective release of GABA from synaptic sites located in their distal den-drites Few studies have targeted OFF SACs for whole-cell recording (Vlasits et al., 2014) due to the less accessible location

of their somas in the retina’s inner nuclear layer (Famiglietti, 1983) While one study showed motion-evoked Ca2+responses

in OFF SACs (Ding et al., 2016), no electrophysiological record-ings of motion-evoked responses have been reported, and how OFF and ON SAC response properties compare is not known Several mechanisms have been demonstrated to contribute to

ON SAC DS, and several models exist (Taylor and Smith, 2012) Reciprocal SAC-SAC GABAergic inhibition enhances DS (Lee and Zhou, 2006; Zhou and Lee, 2008), but SAC DS persists with GABA receptors blocked (Hausselt et al., 2007; Oesch and Taylor, 2010) and in the absence of surround stimulation required to evoke GABAergic surround inhibition (Hausselt

et al., 2007); thus, inhibition is not strictly required An excita-tion-driven, dendrite autonomous model in which excitatory input interacts with cell-intrinsic properties recapitulates ON SAC DS (Hausselt et al., 2007) This model critically depends

on a dendritic voltage gradient that is maintained by low membrane resistance at the soma combined with tonic excita-tion caused by tonic glutamate release from presynaptic ON bipolar cells While ON bipolar cells presynaptic to ON SACs tonically release glutamate (Borghuis et al., 2013), ON SAC DS

in rabbit persist in the absence of tonic excitation following

1356 Cell Reports 18, 1356–1365, February 7, 2017ª 2017 The Author(s)

This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

pharmacological perturbation (Oesch and Taylor, 2010)

More-over, OFF-type bipolar cells presynaptic to OFF SACs do not

release glutamate tonically (Borghuis et al., 2013), indicating

that OFF SACs may use a mechanism for DS that is different

from that proposed for ON SACs (Hausselt et al., 2007)

Recently, an alternative mechanism for DS, ‘‘space-time

wir-ing,’’ was proposed based on connectomic reconstruction of the

OFF and ON SAC presynaptic circuits (Greene et al., 2016; Kim

et al., 2014) Both OFF and ON SACs showed cell-type-specific

spatial organization of presynaptic bipolar cell contacts across

the dendritic arbor In OFF SACs, type 2 cone bipolar cells

prefer-entially contact proximal dendrites, whereas type 3a cone bipolar

cells preferentially contact intermediate or distal dendrites (further

Broad Temporal Tuning and Persists in the Absence of GABAergic Inhibition

(A) Diagram showing the spatial arrangement of synaptic connections between OFF and ON star-burst amacrine cells (SACs) and a direction-se-lective ganglion cell (DSGC).

(B) Left: two-photon fluorescence images of tdTomato-expressing OFF SACs (magenta) in a whole-mount ChAT-Cre::Ai9 transgenic mouse retina before (top) and after whole-cell recording (bottom) Dye-fill (green) identifies the recorded cell (arrow) Right: illustration of the radial motion stimulus; a schematic of SAC morphology (green) was added for reference.

(C) Electrophysiological whole-cell recordings of the membrane voltage, excitatory, and inhibitory current responses of OFF and ON SACs during outward (red) and inward (black) motion stimula-tion (250 mm/s).

(D) Top: average slope of the response onset for outward (red) and inward motion (black) for OFF (solid symbols) and ON SACs (open symbols) 1 Hz modulation; sine wave slope was added for refer-ence (dashed, gray; *p % 0.031) Bottom: average response amplitude (peak-trough) for all recorded SACs (n = 11 ON, n = 6 OFF; *p % 0.013) Error bars indicate ±1 SEM.

(E) Example traces of the membrane voltage response during outward (red) and inward motion stimulation (black) at different stimulus velocities (x axis scaled to one period) Responses to out-ward and inout-ward motion were normalized to have equal amplitude to emphasize the difference in response slope Dashed lines indicate linear fits to quantify slope of the response onset.

(F) Quantification of responses to outward and inward motion stimulation across cells (includes data shown in E; amplitude and slope were calculated prior to normalization) Shaded area represents ±1 SEM.

See also Figure S2

referred to as ‘‘distal’’;Vlasits et al., 2016)

In ON SACs, type 7 bipolar cells predomi-nantly synapse proximally, and each of the three type 5 bipolar cells synapse distally

A subsequent connectomic study with superior methods (Ding et al., 2016) extended these results by demonstrating excitatory input from one additional OFF bipolar cell type (type 1) onto OFF SACs but agreed with the earlier claim that predominant bipolar cell type input differs between proximal versus distal regions of the SAC dendritic arbor

The temporal kinetics of bipolar cell calcium responses (Baden

et al., 2013) and glutamate release (Borghuis et al., 2013) at the anatomical levels where proximal and distal-connecting bipolar cell types stratify their axonal arbors suggests that they express sluggish and sustained and fast and transient responses to light stimulation, respectively Elementary motion detector models (Barlow and Levick, 1965; Hassenstein and Reichardt, 1956) predict that this difference in proximal versus distal excitation

is sufficient for generating DS in ON and OFF SACs However,

the space-time wiring model for DS has remained controversial,

and one electrophysiological study of mouse ON SACs found no

support for it (Stincic et al., 2016) A direct test in OFF SACs is

lacking To address this, we developed methods for efficient

whole-cell recording of both ON and OFF SACs in the intact

(whole-mount) mouse retina in vitro and applied these methods

to study their visually evoked excitatory response properties

RESULTS

Radial Motion Stimuli Evoke DS Excitatory Responses in

Both OFF and ON SACs

We first established that mouse OFF SACs, like ON SACs,

generate robust DS responses We made loose-patch spike

re-cordings from postsynaptic, fluorescently labeled ON-OFF

DSGCs (ooDSGCs; Figures S1A and S1B) and isolated the

OFF SAC contribution to DS by eliminating light-evoked

re-sponses in the ON SAC presynaptic pathway (L-AP4, 20mM)

ON-pathway block eliminated ooDSGC responses to light

incre-ments, as expected, but left intact DS responses to light

decre-ments (Figure S1B) This is consistent with recordings in rabbit

(Kittila and Massey, 1995) and indicates robust DS responses

in OFF SACs

For direct measurements of motion-evoked responses in ON

and OFF SACs, we obtained whole-cell recordings of

fluores-cently labeled SACs in whole-mount retinas of transgenic

ChAT-Cre::Ai9 mice (Figure 1B) Recorded in current-clamp

mode, the membrane voltage response of both cell types was

direction selective, characterized by an increased response

amplitude during outward compared with inward radial motion

(Figure 1C) Responses to outward and inward radial motion

also showed direction-selective harmonic distortion (skew;

Fig-ure 1C), consistent with harmonic frequency components

(Figures S2A and S2B), as previously reported for ON SACs

(Hausselt et al., 2007)

Slope of response onset and peak-to-trough amplitude

directly impact the activation of voltage-gated ion channels

that drive DS synaptic release, particularly in the presence of

leak currents Thus, we use these two physiologically relevant

parameters, rather than Fourier amplitude and phase, to quantify

the motion-evoked responses of ON and OFF SACs

Next, we tested if the DS membrane voltage response

re-flected DS synaptic input, either excitatory or inhibitory, by

recording synaptic currents during voltage clamp near the

reversal potential for chloride (ECl, 67 mV) and cations (Ecation,

0 mV), respectively We found significant DS in the slope and

amplitude of excitatory postsynaptic potentials (EPSPs;Figures

1C and 1D; paired t test ON SACs, slope: p < 0.0001, amplitude:

p = 0.001, n = 9; OFF SACs, slope: p = 0.009, amplitude: p =

0.035, n = 8) Inhibitory postsynaptic potentials (IPSPs;Figures

1C and 1D) were generally larger in OFF SACs than in ON

SACs, but in neither cell type was the slope or amplitude of

inhi-bition significantly different for outward versus inward motion

(Figure 1D; ON SACs, slope: p = 0.58, amplitude: p = 0.21, n =

8; OFF SACs, slope: p = 0.84, amplitude: p = 0.84, n = 5) These

data show that direction-selective excitation contributes to both

ON and OFF SAC DS

OFF and ON SACs Express Broad Velocity Tuning ooDSGCs and ON SACs show broad DS velocity tuning (He and Levick, 2000; Sivyer et al., 2010) To test if OFF and ON SACs exhibit similarly broad velocity tuning, we compared membrane voltage responses of OFF and ON SACs during motion stimulation across an1 decade velocity range (62.5–500 mm/s) OFF and ON SAC responses were distinctly DS, both in slope and amplitude of the response at all but the lowest velocities (Figures 1E, 1F, and S2D), demonstrating broad velocity tuning of the mechanism un-derlying direction-selective responses in both cell types

ON SAC DS is enhanced by reciprocal GABAa receptor-medi-ated (Zhou and Fain, 1995) inhibition between neighboring ON SACs (Lee and Zhou, 2006) but persists without it (Hausselt

et al., 2007; Oesch and Taylor, 2010) To test if excitation alone

is sufficient for DS in OFF SACs, we compared responses to in-ward versus outin-ward radial motion under control conditions and with GABAareceptors blocked (gabazine; SR95531, 10mM) Ga-bazine eliminated stimulus-evoked inhibitory currents and caused tonic depolarization as expected, but it did not eliminate direction-selective differences in slope and amplitude of OFF and ON SAC responses to inward versus outward motion (Figures S2E and S2F) Thus, neither OFF nor ON SAC DS requires GABAergic inhi-bition Because glycinergic inhibition does not contribute signifi-cantly to ON and OFF SAC DS responses, as evidenced by near-complete loss of inhibition following GABAareceptor block,

we conclude that DS is a SAC emergent receptive field property that can be generated through excitatory synaptic input alone SAC Excitation Is Faster and More Transient following Distal Compared with Proximal Receptive Field Stimulation

If proximal and distal SAC dendrites receive excitatory input from temporally distinct bipolar cell populations, as indicated by con-nectomic reconstruction (Ding et al., 2016; Greene et al., 2016; Kim et al., 2014), then the time course of excitation at different distances from the SAC soma should differ To test this, we pre-sented circular white noise (concentric rings of increasing diam-eter and fixed width, centered on the recorded cell’s soma) and recorded the excitatory response during binary white-noise lumi-nance-contrast modulation of each ring (38 Hz; 100% binary contrast;Figure 2A)

Reverse correlation of the excitatory current response recorded

at the soma and the luminance history of each ring gave a linear approximation (impulse response, or ‘‘filter’’) of the time course

of excitation at each ring’s eccentricity (Figure 2B) Simulations showed that a linear-nonlinear (LN) model (Carandini et al., 2005; Chichilnisky, 2001) constructed with the measured linear filters and a single static nonlinearity after spatial summation captured most of the response variance (Figures S3A–S3C) We found only modest differences between the static nonlinearities for each ec-centricity when computed separately, indicating that the linear fil-ters were representative of the time course of excitation at their respective eccentricities (Figures S3D and S3E) Since the static nonlinearity scales the linear response without influencing the shape of the temporal filter, it is not considered further here Excitatory receptive field centers extended100 mm from the soma (Figure 2C), consistent with previous reports (Ding et al., 2016; Greene et al., 2016; Kim et al., 2014; Vlasits et al., 2016)

1358 Cell Reports 18, 1356–1365, February 7, 2017

The most eccentric filters (e.g., 135–150mm) revealed an

inhibi-tory surround consistent with presynaptic inhibition of the bipolar

cell axon terminals Most significant in the context of our study,

which aimed to test for spatiotemporal differences in SAC

exci-tation, time to peak of SAC excitatory filters showed a distinct

temporal gradient with eccentricity in OFF SACs; this was also

observed in ON SACs but was less pronounced (OFF SAC, 26

± 1.1 ms/100 mm; ON SAC, 8.0 ± 1.2 ms/100 mm;Figure 2D)

In-dependent measurements using conventional annulus stimuli

gave similar results (Figures S4A and S4B) Here, too,

differ-ences in proximal and distal excitation were less pronounced

in ON SACs than in OFF SACs

In both ON and OFF SACs, filters representing distal

exci-tation were more biphasic than filters representing proximal

50 μm

A

B

−200

−100 0

0.5 s

Stimulus - ring #5

white black EPSC, V hold = -69 mV

OFF SAC

Time (ms)

ON SAC

Time (ms)

0 - 15

15 - 30

Eccentricity (μm)

30 - 45

45 - 60

60 - 75

75 - 90

90 - 105

105 - 120

120 - 135

135 - 150

Vhold = -69 mV

100 f.u.

−50 0 50 100 150

ON SAC

0 50 100 150

0 50 100

Eccentricity (μm)

OFF SAC C

D

Eccentricity (μm)

120 110 100 90 80

E

0 1

Eccentricity (μm)

0.5

300 μm

15 μm

EPSC impulse response

Faster and More Transient following Distal Compared with Proximal Receptive Field Stimulation

(A) Left: example frame of the radial white noise stimulus All rings are equal width (15 mm) The stimulus shown represents one time point, with some rings black and others white according to each ring’s unique binary white noise stimulus sequence Schematic SAC morphology is shown for scale (magenta) Top right: example luminance time course for one stimulus ring Bottom right: excitatory current response of an OFF SAC re-corded during radial white noise stimulation (B) Impulse responses (filters) of the excitatory synaptic input evoked by white noise stimulation at each eccentricity (V hold = 69 mV) Traces repre-sent averages of 9 OFF SACs and 18 ON SACs ±1 SEM shown in gray Vertical dashed line illustrates longer time to peak of proximal OFF SAC filters Arrows indicate time to peak of proximal (solid arrows) and distal input (open arrows) Arrow-heads indicate monophasic response profile of proximal (solid) versus biphasic response profile of distal input (open) Gray lines represent ±1 SEM (C) Amplitude of excitatory filters of OFF and ON SACs.

(D) Time to peak of excitatory filters of OFF and ON SACs Dashed lines represent linear fits Arrows indicate time to peak of proximal (solid arrows) and distal input (open arrows).

(E) Biphasic index (BI), defined as the relative amplitude of the excitatory peak and trough

(BI = A t / A t + A p) of excitatory filters for OFF and

ON SACs Arrowheads indicate monophasic response profile of proximal (solid) versus biphasic response profile of distal input (open).

In (C)–(E), the shaded area represents ±1 SEM See also Figures S3 and S4

excitation (25mm versus 85 mm: biphasic index OFF SAC, 0.05 ± 0.05 versus 0.40 ± 0.05 [n = 9, p < 0.0001]; ON SAC, 0.16 ± 0.04 versus 0.55 ± 0.06 [n = 17, p < 0.0001, paired t test]) (Figure 2E) This is consistent with the apparent transient and sustained response kinetics of bipolar cell glutamate release in different synaptic layers (Borghuis et al., 2013)

Model Simulations Demonstrate Sufficiency of an Excitatory Mechanism for SAC DS

To test if the observed spatiotemporal pattern of excitation in ON and OFF SACs is sufficient for DS, we constructed a linear model based on the cells’ average recorded filters (Figures 2B and3A) Convolving these model ON and OFF SACs with a radial motion stimulus gave DS responses (Figures 3A and 3B) similar those re-corded in the SACs (Figures 1C and 1D) An alternative model with spatially uniform temporal kinetics did not give DS re-sponses (amplitude asymmetry in OFF SACs: 0.9% of control,

ON SACs: 3.4% of control;Figure 3C) Thus, the observed DS

A B D

C

Figure 3 Spatiotemporal Excitatory Interactions Generate an Outward-Motion-Preferring Response Asymmetry in OFF and ON SACs

(A) Top: model spatiotemporal receptive field constructed from measured filters ( Figure 2 B) at their respective eccentricities (blue through red; subset of filters shown on right) Bottom left: space-time plot of the simulated receptive field (y = 0; blue, increasing excitatory current; red, decreasing excitatory current) Bottom right: space-time plot of the stimulus (y = 0).

(B) Convolving the excitatory receptive field model with the motion stimulus generated asymmetric responses with increased amplitude and faster onset for outward compared with inward motion.

(C) Top: assigning the same filter (ring #5) to all eccentricities generated symmetrical motion-evoked responses Reversing filter sequence (surround becomes center, and vice versa) reversed the asymmetry of motion responses, causing a preference for inward motion.

(D) Simulated responses following specific manipulations of the model, as indicated (see text for details).

(E) Summary of the amplitude and slope of simulated responses to outward (red circles) and inward (black circles) radial motion in a model OFF (left) and ON SAC (right) for the configurations shown in (B)–(D).

(F) Direction selectivity indices (DSIs) calculated from the responses shown in (D).

(G) Elementary motion detector model A correlator integrates excitatory input from two receptive fields (RF1 and RF2) separated by distanceDx In the original

model ( Hassenstein and Reichardt, 1956 ), low-pass filtering by RF1 generates a delayDt in the transmitted signal that results in direction selectivity at the level of

the correlator The detector shown would prefer motion in the RF1/RF2 direction, and responds maximally to stimulus velocity v = Dx /Dt.

(H) Illustration of hypothetical responses of RF1 and RF2 and their sum at the level of the correlator Asterisk indicates a motion response through synaptic release when the response threshold is crossed (red dashed line) In this example, the integrator uses summation; other integration modes (e.g., multiplication) may enhance detector performance Right: variation on the elementary motion detector model shown on the left, with motion detection based on integration of

(legend continued on next page)

1360 Cell Reports 18, 1356–1365, February 7, 2017

(Figure 3B) resulted from interactions between spatially

sepa-rated excitatory inputs with different temporal kinetics Indeed,

reversing filter order (assigning center filters to the surround

and surround filters to the center) changed the response from

outward preferring to inward preferring (Figure 3C),

demon-strating that the specific spatial organization of temporally

diverse excitatory input onto the SAC dendrites generates

out-ward-preferring DS Next, we performed three specific model

tests to determine how different aspects of the SAC

spatiotem-poral receptive field contribute to DS

To test the contribution of the inhibitory surround, we assigned

filter values of zero to the outermost three rings (105–150mm

radius) This manipulation resulted in responses that were nearly

identical to control conditions (amplitude asymmetry: 94.7% of

control in OFF SACs and 87.3% of control in ON SACs;Figures

3D–3F) Thus, receptive field surround interactions from

presyn-aptic inhibition onto the bipolar cell contribute negligibly to

SAC DS

To test the contribution of temporal interactions between

proximal excitation (which has longer time-to-peak and

sus-tained temporal kinetics) and distal excitation (which has shorter

time-to-peak and transient temporal kinetics;Figures 2D, 2E,

and 3A), we assigned filter values of zero to the centermost

four rings, identified by their monophasic kinetics Losing center

input reduced asymmetry in the motion-evoked response

ampli-tude by 50.5% in OFF SACs and 76.8% in ON SACs (Figures 3D–

3F), demonstrating that central excitation contributes strongly

to DS

To test the contribution of response transience of distal

exci-tation to DS, we eliminated transience by setting all positive filter

values to zero This manipulation rendered all filters monophasic

(i.e., sustained), while preserving differences in time to peak of

proximal versus distal input The rationale is that temporal

inter-actions between sustained and transient input may drive

corre-lation detection even in the absence of a response latency

differ-ence (Figures 3G and 3H) Loss of transient excitation reduced

asymmetry of the response amplitude by 50.1% in OFF SACs

and 83.0% in ON SACs (Figures 3D–3F), demonstrating that

distal transient excitation contributed strongly to DS

Collec-tively, these model simulations show that spatiotemporal

inter-actions of measured excitatory responses in both OFF and ON

SACs are sufficient for generating DS with a magnitude that

ex-ceeds a commonly used threshold for direction selectivity

(direc-tion selectivity index [DSI] > 0.3;Figures 3E and 3F)

Our data show that the SAC excitatory response to distal

versus proximal visual stimulation differs In both ON and OFF

SACs, proximal responses are sustained whereas distal

sponses are transient In addition, in OFF SACs, proximal

re-sponses are significantly more delayed than distal rere-sponses

One interpretation is that the response properties of the

presyn-aptic bipolar cells differ, which is consistent with connectomic

reconstructions that show that contacting bipolar cell types

segregate across the SAC arbors (Ding et al., 2016; Greene

et al., 2016; Kim et al., 2014) However, our measurements cannot rule out that the observed differences in excitation arise postsynaptically, e.g., through differences in glutamate receptor composition of proximal versus distal synapses Selective expression of NMDA receptors in particular, which pass cations

in a membrane voltage-dependent manner, could influence the time course of the postsynaptic excitatory response during spatial stimulation

To test if NMDA receptors differentially shape the excitatory response at proximal versus distal synapses on the SAC den-drites, we recorded visually evoked SAC responses in the pres-ence and abspres-ence of the NMDA receptor selective pharmaco-logical blocker D-AP5 Amplitude and slope of the membrane voltage response remained direction-tuned with NMDA recep-tors blocked (Figure 4A), demonstrating that NMDA receprecep-tors are not required for ON and OFF SAC DS Current recordings further showed that the time course of excitation during circular white noise stimulation remained monophasic (sustained) prox-imal and biphasic (transient) distal in both ON and OFF SACs (Figures 4B–4D), demonstrating that these properties are NMDA-receptor independent

NMDA-receptor block had a negligible effect on the temporal response in ON SACs but significantly shortened the duration of the sustained, proximal excitation in OFF SACs (Figure 4B) This faster shutoff of excitation with NMDA receptors blocked pushed the time to peak of the excitatory response forward in time, thus reducing the temporal difference between proximal and distal time to peak (Figure 4C) NMDA receptor block did not change the biphasic index for excitation at proximal or distal locations (Figure 4D)

Bath-applied D-AP5 blocks NMDA receptors throughout the retina Thus, its effect on proximal OFF SAC excitation may be presynaptic (e.g., through specific NMDA-receptor-dependent inhibitory circuitry impinging on the proximal-connecting bipolar cell types) and/or postsynaptic (through preferential NMDA-re-ceptor expression in proximal synapses) We resolved this ambi-guity from current voltage (I-V) curves obtained by recording the current response to proximal visual stimulation at different hold-ing potentials (Figure 4E) If synapses on the proximal OFF SAC dendrites express NMDA receptors, then subtraction of the I-V curves recorded in the absence and presence of D-AP5 should reveal a significant J-shaped residual current, characteristic of the NMDA receptor response (Dingledine et al., 1999) Indeed, OFF SACs showed a J-shaped difference current in control minus D-AP5 conditions, indicating that OFF SACs express NMDA receptors at proximal synapses (Figure 4F) ON SACs,

on the other hand, did not show a J-shaped residual current dur-ing proximal stimulation, consistent with the observation that D-AP5 does not affect the time course of excitation in these cells (Figures 4B and 4C) D-AP5 broadly increased the current ampli-tude in ON SACs, indicating that NMDA receptors control the presynaptic bipolar cell response Thus, NMDA receptors influ-ence the OFF SAC response by slowing its proximal excitatory

spatially offset transient and sustained RF inputs Because the transient input may combine with the sustained input to cross threshold at a range of time points following onset of the sustained response, this detector would exhibit broad temporal tuning (see Discussion for details) To better illustrate the key features of the proposed model, in this schematic, the magnitude of differences in temporal response kinetic (transient versus sustained) has been increased relative to what was measured.

Figure 4 NMDA Receptors Slow Proximal Excitation in OFF SACs

(A) Membrane voltage response (whole-cell current clamp) of OFF and ON SACs during inward (black) and outward (blue) radial motion in the absence (top) and presence (bottom) of the NMDA receptor blocker D-AP5 (50 mM) Panel shows single-cell examples representative of the recorded population (OFF n = 5;

ON n = 6) Response slope and amplitude did not change significantly following NMDA receptor block (all p > 0.10).

(B) Impulse responses (filters) of the excitatory synaptic input evoked by proximal, distal, and surround circular white noise stimulation (V hold = 69 mV) in the absence (black) and presence (red) of D-AP5 (50 mM) Traces represent averages of 4 OFF SACs and 5 ON SACs ±1 SEM shown in gray.

(C) Response time-to-peak of the excitatory filters measured with circular white noise under control conditions (black) and with NMDA receptors blocked (red; data partially shown in B) Asterisks indicate significant differences (p < 0.05).

(D) Biphasic index of the excitatory filters measured with circular white noise under control conditions (black) and with NMDA receptors blocked (red; data partially shown in B) Biphasic indexes were always greater at distal compared with proximal locations and did not change significantly following NMDA receptor block (n.s.) Error bars represent ±1 SEM.

(E) Current response of an OFF SAC at different holding potentials (bottom) during visual stimulation of the proximal receptive field (schematic in inset, top left; time course shown below traces).

(legend continued on next page)

1362 Cell Reports 18, 1356–1365, February 7, 2017

input, but they do not mediate the sustained versus transient

dif-ference in temporal kinetics of excitation at proximal versus

distal synaptic sites

DISCUSSION

Targeted whole-cell recordings from ON and OFF SACs show

that excitation in ON and OFF SAC distal versus proximal

den-drites is temporally diverse; excitation following distal stimulation

is both faster and more transient than that following proximal

stimulation Model simulations show that this specific functional

architecture generates direction-selective responses with a

pref-erence for outward motion—a hallmark property of ON and OFF

SACs Because this mechanism acts at the level of the excitatory

input, it appears to generate the first instance of direction

selec-tivity in the SAC We conclude that selective integration of

tran-sient and sustained excitation is a fundamental mechanism for

SAC direction selectivity

Pharmacological perturbation showed that NMDA receptors

prolong the proximal excitatory response and increase the

tem-poral delay of proximal, but not distal, excitation in OFF SACs

Thus, selective NMDA receptor expression is a cell-intrinsic

mechanism for diversifying proximal versus distal temporal

re-sponses in OFF SACs Since temporal delay of excitation

im-pacts DS, this mechanism likely contributes to these cells’

veloc-ity tuning Importantly, excitation in ON and OFF SACs remained

sustained proximal and transient distal following NMDA receptor

block A parsimonious explanation, consistent with connectomic

data (Ding et al., 2016; Greene et al., 2016; Kim et al., 2014) is

that this difference in excitation reflects synaptic input from

func-tionally distinct bipolar cell types across the dendritic arbor The

circuit model supported by our results bears a striking

resem-blance to the visual motion detection circuits in Drosophila

(Beh-nia et al., 2014; Leonhardt et al., 2016; Serbe et al., 2016; Tuthill

and Borghuis, 2016) A similar synaptic architecture,

character-ized by dendritic domains with functionally distinct synaptic

input, has been reported for mouse hippocampal pyramidal

cells, where it serves as a mechanism for coincidence detection

in synaptic plasticity and learning (Stuart and Spruston, 2015)

Our work provides functional evidence for an excitation-based

mechanism for SAC DS but does not rule out additional

contribu-tions from previously described mechanisms, including dendrite

autonomous computation in ON SACs (Hausselt et al., 2007) and

reciprocal SAC-SAC inhibition (Lee and Zhou, 2006) Our results

apparently contradict those of a recent study that found no

dif-ference in the time course of excitation in ON SACs measured

with flashed annuli (Stincic et al., 2016) One possible

explana-tion is that the concentric white noise stimulus used in our study

provides a more sensitive measure of the time-course of

excita-tion during dynamic visual stimulaexcita-tion; another is that flashed

annuli may evoke nonlinear postsynaptic response properties

not activated by spatially dense white noise, which stimulates

at a net-zero mean level across the SAC dendritic arbor Consis-tent with this explanation, ON SACs in our study showed substantially less temporal diversity in proximal versus distal excitatory conductance when stimulated with annuli compared with circular white noise (Figure S4B)

We used linear systems analysis to compute the time course

of excitation at different eccentricities The filter characteristics are linear approximations of the full excitatory response, which likely includes non-linearities The use of spatiotemporal white noise stimuli helped resolve response dynamics with high tem-poral resolution while minimizing nonlinearities from contrast and luminance gain control Contributions from additional non-linear mechanisms such as spatiotemporal interactions (motion detection) were excluded by definition due to the anal-ysis method While spatiotemporally correlated stimuli (e.g., texture motion) would evoke the specific nonlinear response properties that drive DS synaptic release, our method was opti-mized for quantitative assessment of the time course of local excitation to test for predicted temporal differences in the excit-atory response selective input from functionally diverse bipolar cell pathways impinging on proximal and distal dendrites (Greene et al., 2016; Kim et al., 2014)

The magnitude of the asymmetry of ON and OFF SAC re-sponses to outward compared with inward motion is small (Fig-ure 1) compared with the robust directional asymmetry of GABA release evident in postsynaptic DSGCs (e.g.,Park et al., 2014)

In the current working model, the relatively small, initial asymmetry

in the SAC membrane voltage is amplified nonlinearly by various mechanisms, including voltage-gated channels and reciprocal in-hibition, and thresholded to generate strongly DS synaptic release

DS of ooDSGC responses extends over a wide range of stim-ulus velocities This broad temporal tuning is present already at the level of the GABAergic inhibitory input from SACs shown here (Figure 1F) and in rabbit (Sivyer et al., 2010) Broad temporal tuning is inconsistent with a strictly spatiotemporal delay-based mechanism for motion detection, which would give narrow tem-poral tuning with peak sensitivity defined by the ratio of spatial extent of the excitatory input (Dx) and relative delay between proximal and distal input (Dt; Figures 3G and 3H) Our data show that two independent components of OFF and ON SAC excitatory responses contribute to coincidence detection under-lying DS (Figure 3H): (1) a relative delay (Dt) between proximal and distal excitation, which was substantial in OFF SACs (26 ms/100mm) but minor in ON SACs (8 ms/100 mm;Figure 2D); and (2) a difference in temporal kinetics (transient versus sustained) between proximal and distal excitation, which was observed in both OFF and ON SACs (Figures 2B and 2E) Because the transient response can trigger SAC synaptic release as long as the sustained response persists (Figure 3H), transient-sustained integration provides a mechanistic basis for the well-established but incompletely understood broad tem-poral tuning of ooDSGCs (Sivyer et al., 2010)

(F) Current-voltage relation of the response to proximal visual stimulation Left: amplitude of the initial, transient response component (‘‘T’’ in E, average of latter three stimulus cycles) Right: amplitude of the sustained response component (‘‘S’’ in E) Magenta curve shows the difference between the current response under control (black) and D-AP5 conditions (red) Asterisks indicate significant differences (p < 0.05) The significant J-shaped difference curve of the sustained response component in OFF SACs (top right, arrowhead) indicates that in OFF SACs, NMDA receptors contribute to the excitatory current at proximal synapses.

In (B), (C), and (F), the shaded area represents ±1 SEM.

EXPERIMENTAL PROCEDURES

Animals

All animal procedures were approved by the Institutional Animal Care and Use

Committee at the University of Louisville and were in compliance with National

Institutes of Health guidelines Mice of either sex, maintained on C57BL/6J

backgrounds, were studied between 1.5 and 3 months of age Data were

obtained from transgenic ChAT-IRES-Cre mice (Jackson Laboratory stock

#006410) crossed with the Ai9 tdTomato ROSA26 reporter line (Jackson

Lab-oratory stock #007905) ooDSGC recordings of Figure S1 used an

Thy1-iGluSnFR line (Looger lab, Janelia Research Campus) with low-level

expres-sion of iGluSnFR in various cell types, including ooDSCGs.

Electrophysiological Recording

Whole-cell electrophysiological recordings were obtained from the ventral,

whole-mount mouse retina in vitro as described previously ( Borghuis et al.,

2013 ) See Supplemental Experimental Procedures for details.

Visual Stimulation

Visual stimuli were generated with an Apple G4 computer and custom

C-lan-guage software Stimuli were displayed using a DLP video projector (HP

AX325AA; Hewlett-Packard), with the image projected onto the photoreceptor

layer using the microscope condenser Stimuli comprised inward and outward

drifting radial square waves ( Figure 1 B), flashed annuli, and concentric binary

white noise ( Figure 2 A).

Data Analysis and Model Simulations

Data are presented as mean ± SEM unless stated otherwise Statistical

signif-icance was assessed using Student’s t test as indicated Model analysis was

performed with numerical simulations using custom algorithms in MATLAB

(MathWorks).

SUPPLEMENTAL INFORMATION

Supplemental Information includes Supplemental Experimental Procedures

and four figures and can be found with this article online at http://dx.doi.org/

10.1016/j.celrep.2017.01.026

AUTHOR CONTRIBUTIONS

J.W.F and B.G.B conceived of the study, designed the experiments, and

collected and analyzed the data B.G.B developed the model analysis and

simulations B.G.B wrote the manuscript Both authors read and approved

the final version of the manuscript.

ACKNOWLEDGMENTS

We thank Dr J Demb for helpful comments on the model analysis and

simu-lations This work was supported in part by grants from the E Matilda Ziegler

Foundation for the Blind and the University of Louisville School of Medicine.

Received: April 25, 2016

Revised: December 5, 2016

Accepted: January 11, 2017

Published: February 7, 2017

REFERENCES

Baden, T., Berens, P., Bethge, M., and Euler, T (2013) Spikes in mammalian

bipolar cells support temporal layering of the inner retina Curr Biol 23, 48–52.

Barlow, H.B., and Hill, R.M (1963) Selective sensitivity to direction of

move-ment in ganglion cells of the rabbit retina Science 139, 412–414.

Barlow, H.B., and Levick, W.R (1965) The mechanism of directionally

selec-tive units in rabbit’s retina J Physiol 178, 477–504.

Processing properties of ON and OFF pathways for Drosophila motion

detec-tion Nature 512, 427–430 Borghuis, B.G., Marvin, J.S., Looger, L.L., and Demb, J.B (2013) Two-photon imaging of nonlinear glutamate release dynamics at bipolar cell synapses in

the mouse retina J Neurosci 33, 10972–10985 Briggman, K.L., Helmstaedter, M., and Denk, W (2011) Wiring specificity in

the direction-selectivity circuit of the retina Nature 471, 183–188 Carandini, M., Demb, J.B., Mante, V., Tolhurst, D.J., Dan, Y., Olshausen, B.A., Gallant, J.L., and Rust, N.C (2005) Do we know what the early visual system

does? J Neurosci 25, 10577–10597 Chichilnisky, E.J (2001) A simple white noise analysis of neuronal light

re-sponses Network 12, 199–213 Ding, H., Smith, R.G., Poleg-Polsky, A., Diamond, J.S., and Briggman, K.L (2016) Species-specific wiring for direction selectivity in the mammalian

retina Nature 535, 105–110 Dingledine, R., Borges, K., Bowie, D., and Traynelis, S.F (1999) The glutamate

receptor ion channels Pharmacol Rev 51, 7–61 Dmitriev, A.V., Gavrikov, K.E., and Mangel, S.C (2012) GABA-mediated spatial and temporal asymmetries that contribute to the directionally selective

light responses of starburst amacrine cells in retina J Physiol 590, 1699–

1720 Euler, T., Detwiler, P.B., and Denk, W (2002) Directionally selective calcium

signals in dendrites of starburst amacrine cells Nature 418, 845–852 Famiglietti, E.V., Jr (1983) ‘Starburst’ amacrine cells and cholinergic neurons:

mirror-symmetric on and off amacrine cells of rabbit retina Brain Res 261,

138–144 Fried, S.I., M€unch, T.A., and Werblin, F.S (2002) Mechanisms and circuitry

underlying directional selectivity in the retina Nature 420, 411–414 Gavrikov, K.E., Nilson, J.E., Dmitriev, A.V., Zucker, C.L., and Mangel, S.C (2006) Dendritic compartmentalization of chloride cotransporters underlies directional responses of starburst amacrine cells in retina Proc Natl Acad.

Sci USA 103, 18793–18798 Greene, M.J., Kim, J.S., and Seung, H.S.; EyeWirers (2016) Analogous convergence of sustained and transient inputs in parallel on and off pathways

for retinal motion computation Cell Rep 14, 1892–1900 Hassenstein, B.v., and Reichardt, W (1956) Systemtheoretische Analyse der Zeit-, Reihenfolgen- und Vorzeichenauswertung bei der Bewegungsperzep-tion des R€usselka¨fers Chlorophanus Z Naturforschg 11, 513–524 Hausselt, S.E., Euler, T., Detwiler, P.B., and Denk, W (2007) A dendrite-auton-omous mechanism for direction selectivity in retinal starburst amacrine cells.

PLoS Biol 5, e185.

He, S., and Levick, W.R (2000) Spatial-temporal response characteristics of the ON-OFF direction selective ganglion cells in the rabbit retina Neurosci.

Lett 285, 25–28 Kim, J.S., Greene, M.J., Zlateski, A., Lee, K., Richardson, M., Turaga, S.C., Purcaro, M., Balkam, M., Robinson, A., Behabadi, B.F., et al.; EyeWirers (2014) Space-time wiring specificity supports direction selectivity in the retina.

Nature 509, 331–336 Kittila, C.A., and Massey, S.C (1995) Effect of ON pathway blockade on

direc-tional selectivity in the rabbit retina J Neurophysiol 73, 703–712 Lee, S., and Zhou, Z.J (2006) The synaptic mechanism of direction selectivity

in distal processes of starburst amacrine cells Neuron 51, 787–799 Lee, S., Kim, K., and Zhou, Z.J (2010) Role of ACh-GABA cotransmission in

detecting image motion and motion direction Neuron 68, 1159–1172 Leonhardt, A., Ammer, G., Meier, M., Serbe, E., Bahl, A., and Borst, A (2016) Asymmetry of Drosophila ON and OFF motion detectors enhances real-world

velocity estimation Nat Neurosci 19, 706–715 Oesch, N.W., and Taylor, W.R (2010) Tetrodotoxin-resistant sodium channels

contribute to directional responses in starburst amacrine cells PLoS ONE 5,

e12447

1364 Cell Reports 18, 1356–1365, February 7, 2017