Intrinsically photosensitive ganglion cells contribute toplasticity in retinal wave circuitsLowry A. Kirkbya and Marla B. Fellerb,1

aBiophysics Graduate Group and bDepartment of Molecular and Cell Biology and the Helen Wills Neuroscience Institute, University of California, Berkeley,CA 94720

Edited by Lynn T. Landmesser, Case Western Reserve University, Cleveland, OH, and approved June 14, 2013 (received for review December 18, 2012)

Correlated spontaneous activity in the developing nervous systemis robust to perturbations in the circuits that generate it, suggest-ing that mechanisms exist to ensure its maintenance. We examinethis phenomenon in the developing retina, where blockade ofcholinergic circuits that mediate retinal waves during the firstpostnatal week leads to the generation of “recovered” wavesthrough a distinct, gap junction–mediated circuit. Unlike choliner-gic waves, these recovered waves were modulated by dopaminer-gic and glutamatergic signaling, and required the presence of thegap junction protein connexin 36. Moreover, in contrast to cholin-ergic waves, recovered waves were stimulated by ambient lightvia activation of melanopsin-expressing intrinsically photosensi-tive retinal ganglion cells. The involvement of intrinsically photo-sensitive retinal ganglion cells in this reconfiguration of wave-generating circuits offers an avenue of retinal circuit plasticity dur-ing development that was previously unknown.

retinal development | dopamine | degenerate circuit

The computations performed by neural circuits are not de-termined by hard-wired anatomy but rather can be altered by

experience or different neuromodulatory states (1–3). Thisplasticity is particularly important during development, whenneural circuits show remarkable robustness against perturbationsthat disrupt the patterned, spontaneous activity required for nor-mal development (4). For example, giant depolarizing potentialsin the developing hippocampus are maintained against decrea-ses in gamma-aminobutyric acid (GABAergic) transmission byincreasing the strength of glutamatergic transmission (5). Simi-larly, spontaneous network activity in the developing spinal cordis maintained against alterations in GABAergic transmission bychanges in both the intrinsic excitability of individual neuronsand changes in synaptic strength of glutamatergic synapses (6).The developing retina also shows robustness against perturba-tions in circuits that generate spontaneous retinal waves (4). Forexample, disruption of normal cholinergic transmission duringthe first postnatal week leads to the generation of waves via adistinct gap junction coupled network (7–9). These observationsindicate that degenerate circuit mechanisms exist in the de-veloping retina to maintain spontaneous activity.Here we explore the hypothesis that intrinsically photosensi-

tive retinal ganglion cells (ipRGCs) contribute to this wave cir-cuit plasticity. ipRGCs are a recently discovered class ofphotoreceptors that express the photopigment melanopsin (10)and are light sensitive in mice from birth, unlike rod and conephotoreceptors, which become photosensitive after 2 postnatalweeks of development (11). Although ipRGCs are typically in-volved in non-image-forming functions, such as entrainment ofcircadian rhythms (12), they have been shown to support intra-retinal signaling via gap junction coupling and by signaling todopaminergic amacrine cells (DACs) (13–15). Indeed, lightstimulation of ipRGCs can modulate cholinergic retinal circuitsduring development (16).We use multielectrode array (MEA) recordings to compare

the spatial and temporal properties of firing patterns recorded inthe dark versus the light from wild-type (WT) mice and knockout

mice lacking normal cholinergic waves (β2KO), in addition tolacking the gap junction protein connexin 36 (β2-cx36 dKO) orthe photopigment melanopsin (β2-Opn4 dKO). Our data sup-port the hypothesis that early light responses from ipRGCscontribute to the circuit that mediates the recovery of correlatedspontaneous firing patterns in the absence of cholinergic waves.

ResultsRecovered Waves in Mice Lacking Cholinergic Waves Are Modulatedby Light. To investigate firing patterns across retinal ganglioncells (RGCs) in both the dark and light, we performed MEArecordings on retinas acutely isolated from WT mice and in micelacking the β2 subunit of nicotinic acetylcholine receptors (nAChRs)(β2KO) at postnatal days 4–7 (P4–P7), when cholinergic wavesnormally occur (17, 18). At these ages, β2KO mice exhibit gapjunction waves in place of cholinergic ones (8, 9). In addition,rod and cone photoreceptors are not yet photosensitive anddo not contribute to ganglion cell light responses (19). WT re-cordings confirmed previous observations (7, 9, 20, 21), whereRGCs fired periodic bursts of action potentials that swept acrossthe retina as a wave (Fig. 1A). Immediately following lightonset, a subset of cells fired sustained bursts of action poten-tials, consistent with previously described light responses ofipRGCs in early postnatal development (22, 23).Recordings of β2KO mice also exhibited retinal waves with

distinct propagation properties from WT mice. In particular,waves in β2KO mice were less frequent and many RGCs firedaction potentials that were not associated with retinal waves andtherefore were not significantly correlated with one another, inagreement with previous studies (8, 9, 21, 24) (Fig. 1B). Sur-prisingly, we also observed that light stimulation led to an almosttwofold increase in the frequency of waves in β2KO retinas, incontrast to WT retinas, for which light stimulation had no effecton wave frequency (Fig. 1C, ***P < 0.001). This light-inducedincrease was observed over a range of starting frequencies,suggesting that the light-induced effect is independent of initialwave frequency. Light stimulation also increased the burst du-ration during a wave in both WT and β2KO retinas (Fig. 1D), aspreviously reported for WT retinas (16).Because ipRGCs are the only functional photoreceptors at

these ages and because the light-evoked increase in burst dura-tion of cholinergic WT waves is eliminated in melanopsinknockout mice (Opn4 KO) (16), we presumed that the observedlight-evoked effects on waves in β2KO retinas were mediatedby ipRGCs. To test this directly, we performed MEA recordingson a β2-melanopsin double KO mouse (β2-Opn4 dKO, kind-ly provided by David Copenhagen, University of California,

Author contributions: L.A.K. and M.B.F. designed research; L.A.K. performed research;L.A.K. analyzed data; and L.A.K. and M.B.F. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: mfeller@berkeley.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1222150110/-/DCSupplemental.

12090–12095 | PNAS | July 16, 2013 | vol. 110 | no. 29 www.pnas.org/cgi/doi/10.1073/pnas.1222150110

Dow

nloa

ded

by g

uest

on

July

10,

202

0

San Francisco). These mice exhibited retinal waves but did notshow an increase in wave frequency or burst duration duringa wave in the light (Fig. 1 C and D). These observations confirmthat the photic effects on waves in β2KO mice were mediatedby ipRGCs.

Waves in β2KO Mice Require cx36. Previous studies have shown thatthe recovered waves in β2KO mice are blocked by gap junctionantagonists but not fast neurotransmitter antagonists (8). Weassessed whether the presence of the neuronal gap junctionprotein cx36 was necessary for waves in β2KO mice by per-forming MEA recordings on a β2-cx36 dKO mouse (24). Cx36 isthe most abundant retinal connexin and couples most ganglioncells to other ganglion or amacrine cells (25). Although cx36KOmice have an increase in asynchronous firing in between retinal

waves, they exhibit cholinergic retinal waves with propagationpatterns that are indistinguishable from WT during the firstpostnatal week (24, 26). We found that β2-cx36 dKO mice didnot exhibit recovered retinal waves in either the dark or light(Fig. 2A). Rather, many RGCs fired asynchronous actionpotentials. We characterized the correlation properties of spik-ing neurons by computing correlation indices as a function ofintercellular distance for all cell pairs. This gives a measure ofthe likelihood relative to chance that two cells fire togetherwithin a given time window, where retinal waves are character-ized by a correlation index that is high for nearest neighbors andthat falls off with increasing intercellular distance (17). In con-trast, correlation index curves of β2-cx36 dKO mice were flat,confirming the absence of retinal waves (Fig. 2B). In some reti-nas, we observed synchronous bursting among subsets of cells inthe light, however this activity did not propagate in a wave-likemanner (Fig. S1A, Top). These observations confirm that re-covered waves in β2KO mice are gap junction mediated andshow that cx36 is required for their propagation.

ipRGCs Do Not Function as “Hub” Neurons for Recovered Waves.Oneway in which ipRGCs may increase β2KO wave frequency in thelight is by functioning as hub neurons, or highly connected nodes,that link together many cells and thereby impart synchronywithin the network. To test this possibility, we first identifiedneurons whose firing was highly correlated with a large numberof other cells and then determined whether these highly con-nected cells corresponded to ipRGCs.We first computed correlation indices as a function of in-

tercellular distance for all cell pairs. WT correlation indices werehigh for nearest neighbors and fell off with increasing in-tercellular distance, characteristic of retinal waves (Fig. 3A).β2KO correlation index curves were flatter than those for WT(Fig. 3A), consistent with larger waves and a faster propagationspeed, as described in previous studies (8, 9). Correlation indexcurves showed no difference in the dark and light for both WTand β2KO mice. We next used these correlation indices toconstruct connectivity maps, which show connections betweenneurons with the highest correlation indices (26, 27). In partic-ular, this strategy has been used in the developing hippocampusto identify highly connected hub neurons, which repeatedly ini-tiate activity (28). We defined two cells as being connected iftheir correlation index fell in the top 5% of all correlation indexvalues and extracted cells that were connected to at least 15% ofother cells (Fig. 3 B and C; “highly connected cells” correspondto red units and their connections are shown by blue lines).For WT retinas, highly connected cells fell in a cluster of ad-

jacent units with a median connection length of 145 μm (Fig. 3 Band D). Highly connected cells in β2KO mice were more dis-persed than in WT, with a median connection length of 260 μm(Fig. 3 C and D). Median connection lengths in the dark were

2 min

2 H

z

2 min

1 H

zA

B

Dark

WT β2KO β2-Opn4dKO

WT β2KO β2-Opn4dKO

Light

Dark Light

WT, P5

β2KO, P5

******

*

0

1

2

3

4

5

Bur

st d

urat

ion

durin

g w

ave

[sec

]

DC

0

0.5

1

1.5

Wav

e fre

quen

cy[p

er m

in]

darklight dark

light

Fig. 1. Recovered waves in mice lacking cholinergic waves are modulatedby light. (A and B) MEA recordings from a WT P5 (A) and β2KO P5 (B) retina.(Top) Activity pattern corresponding to boxed region in Middle. Each dotrepresents an electrode site, and the radius of the dot is proportional to thesingle unit firing rate recorded at that site. Frames correspond to 2 s.(Middle) Raster plot of spike trains of all single units. (Bottom) Averagefiring rate of all units. (C) Summary data of wave frequency in dark and lightfor WT, β2KO, and β2-Opn4 dKO mice. Open circles correspond to individualretinas and recordings from the same retinas are connected by dotted lines.(D) Summary data of burst duration during a wave in dark and light for WT,β2KO, and β2-Opn4 dKO mice. Error bars in C and D correspond to SEM.*P < 0.05; ***P < 0.001, paired t test.

B

00 500

25

Cor

rela

tion

inde

x

Distance [μm]

β2-cx36 dKO in lightA β2-cx36 dKO, P6

2 min

0.5

Hz

Dark Light

Fig. 2. β2KO waves require cx36. (A) MEA recording of spontaneous activityin β2-cx36 dKO mouse (as in Fig. 1A). (B) Correlation index versus in-terelectrode distance for pairs of spike trains in the light for β2-cx36 dKOmice. Data points correspond to averages of median values from individualretinas and error bars correspond to SEM.

Kirkby and Feller PNAS | July 16, 2013 | vol. 110 | no. 29 | 12091

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

July

10,

202

0

similar to those in the light for both WT and β2KO mice (Fig.3D), indicating that light does not alter the underlying spatialarchitecture of waves.We determined the locations of ipRGCs on the MEA by

identifying units that showed at least 45 additional spikes in the60-s window following light onset compared with the 60 s pre-ceding light onset (0.75 Hz increase; Fig. S2). This threshold wasdetermined using targeted cell-attached recordings of ipRGCs inOpn4–EGFP mice, in which GFP is expressed under the mela-nopsin promoter (29). We next determined whether identifiedipRGCs (turquoise units in Fig. 3 B and C) function as hub neuronsfor retinal waves by computing the percentage of ipRGCs thatwere highly connected and comparing this to the percentage wewould expect by chance. In the dark, the percentage of ipRGCsthat were highly connected did not differ from chance for bothWT and β2KO mice (Fig. 3E). Surprisingly, in the light thepercentage of ipRGCs that were highly connected was signifi-cantly less than chance (Fig. 3E, **P < 0.01). These observationsshow that ipRGCs do not function as hub neurons, as activationof ipRGCs in the light does not directly activate many otherRGCs. Rather, they suggest that the action of light on wavefrequencies in β2KO mice is an indirect effect of ipRGCs on thenetwork. Below we explore whether this modulation occurs viadopaminergic signaling.

β2KO Waves Are Modulated by Dopaminergic and GlutamatergicSignaling. ipRGCs are thought to influence retinal networks viaglutamatergic stimulation of dopamine release from DACs [14,15), but see ref. 30]. Dopamine is produced in mice as early as P4by DACs (31) and is a major player in regulation of gap junctioncoupling via cAMP-dependent posttranslational phosphorylation(25, 32). In general, activation of D1-like receptors decreases gapjunction coupling while activation of D2-like receptors increases

coupling, although the magnitude of the effects are highly cell-typespecific (33–35). Thus, one way that ipRGCs could exert an in-direct effect on network synchrony is via a dopaminergic pathway.We first tested whether dopamine signaling regulates wave

frequencies in WT and β2KO mice by performing MEArecordings in dopamine receptor antagonists. We found thatneither a D1 receptor antagonist (SCH23390, 10 μM) nor a D2receptor antagonist (raclopride, 8 μM) had an effect on wavefrequency in WT retinas (Fig. 4A). However, raclopride stronglyreduced wave frequency in β2KO mice in the light, whereasSCH23390 increased their frequency (Fig. 4B). Neither antago-nist significantly influenced mean baseline firing rates relative tothe mean firing rate in the light (Fig. 4C), suggesting that theobserved effects are likely due to modulation of gap junctioncoupling required for waves and not modulation of a cell’soverall firing properties. Furthermore, application of SCH23390to β2-cx36 dKO mice did not induce the generation of waves, nordid it change mean baseline firing rates relative to the meanfiring rate in the light (Fig. S1B), indicating that cx36 is a possibletarget of D1 receptor activation. These observations show thatdopamine signaling strongly modulates the frequency of re-covered waves, where waves are stimulated by D2 receptor ac-tivation but suppressed by D1 receptor activation.Recent experiments indicate that ipRGCs form excitatory

glutamatergic synapses onto DACs (14, 15). Thus, if the ob-served light-evoked increase in wave frequency in β2KO micewere mediated by ipRGC feedback onto DACs, we would expectthat blocking glutamate receptors would block the effect. To testthis hypothesis, we performed MEA recordings of β2KO mice inglutamate receptor antagonists in the light [D-AP5, 50 μM and6,7-dinitroquinoxaline-2,3-dione (DNQX), 20 μM]. In four outof five retinas tested, wave frequencies returned to dark levels in

B WT light, P5 C β2KO light, P5

WTlight

WTdark

β2KOlight

β2KO chanceWT chance

β2KOdark

D E

Length of connectionsfrom highly connected cells [μm]

ytilibaborpevitalu

muC

secnatsidfo

400 800

WT lightWT darkβ2KO lightβ2KO dark

0

1

.5

β2KO lightWT light

Distance [μm]

xedninoitalerro

C

00 500

25A

0

20

4054 71

eratahts

CG

Rp ifo tnecr eP

]%[

d etcenn ocyl hg i h

****

Fig. 3. ipRGCs do not function as hub neurons for retinal waves. (A) Cor-relation index versus interelectrode distance for pairs of spike trains for WT(black) and β2KO (blue) mice, in the light. Data points correspond to aver-ages of median values from individual retinas and error bars correspond toSEM. (B and C) Connectivity maps of WT P5 (B) and β2KO P5 (C) in the light.Circles correspond to location of single units with diameter scaled by themagnitude of the normalized correlation index. Red circles correspond tounits that were connected to at least 15% of other units, and blue lines showtheir connections. Turquoise circles correspond to units identified as ipRGCs.(D) Cumulative probability distributions of the distances of connections fromhighly connected cells for WT (black) and β2KO (blue) retinas in the dark andlight. (E) Percent of ipRGCs that are highly connected hub neurons for WT(gray) and β2KO (blue) activity, in dark and light. Box plots range from lowerto upper quartiles (25% and 75%) with median values indicated by centralblack line; whiskers (dotted lines) range from 5% to 95%. Faded box plotscorrespond to percentage of ipRGCs that are highly connected that wewould expect from chance. **P < 0.01, Wilcoxon rank sum test.

B

dark light rac SCH0

1

2

Wav

e fre

quen

cy [p

er m

in]

β2KO wave frequencyA

dark

WT wave frequency

light rac SCH0

1

Wav

e fre

quen

cy [p

er m

in]

dark light rac SCH0

1

Mea

n fir

ing

rate

[Hz]

C β2KO mean firing rate

Mea

n fir

ing

rate

[Hz]

β2KO mean firing rateE

0

1

Wav

e fre

quen

cy [p

er m

in] β2-Opn4 dKO

wave frequencyF

0

0.5

Wav

e fre

quen

cy [p

er m

in]

β2KO wave frequency

dark light DNQX+ D-AP5

dark light DNQX+ D-AP5

dark light DNQX+ D-AP5

D

0

1

*** * ns

** ** nsns***

Fig. 4. β2KO waves are modulated by dopaminergic and glutamatergicsignaling. (A and B) Wave frequency in dark, light, and in presence of do-pamine receptor antagonists in light for WT (A) and β2KO (B) mice. (C) Meanfiring rate in dark, light, and in presence of dopamine receptor antagonistsin light for β2KO mouse. (D) Wave frequency in dark, light, and in presenceof glutamate receptor antagonists in light for β2KO mice. (E) Mean firingrate in dark, light, and in presence of glutamate receptor antagonists in lightfor β2KO mice. (F) Wave frequency in dark, light, and in presence of glu-tamate receptor antagonists in light for β2-Opn4 dKO mice. For all plots,open circles correspond to individual retinas, and recordings from the sameretinas are connected by dotted lines. Error bars correspond to SEM. D1Rantagonist, SCH23390 (SCH), 10 μM; D2R antagonist, raclopride (rac), 8 μM.AMPAR and NMDAR antagonists, DNQX and D-AP5, 20 and 50 μM. *P < 0.05;**P < 0.01, repeated measures ANOVA with Holm-Sidak posthoc test.

12092 | www.pnas.org/cgi/doi/10.1073/pnas.1222150110 Kirkby and Feller

Dow

nloa

ded

by g

uest

on

July

10,

202

0

the presence of glutamate receptor antagonists with no effect onbaseline firing rates (Fig. 4 D and E), suggesting that the antag-onists influence wave properties but not a cell’s overall firingproperties. To test whether these glutamatergic effects were de-pendent on light-evoked firing of ipRGCs, we repeated themeasurements in β2-Opn4 dKO mice. We found that glutamatereceptor antagonists had no effect on wave frequencies in thesemice (Fig. 3F), indicating that the glutamatergic reduction of wavefrequency in β2KO mice in the light was melanopsin dependent.Together, these observations are consistent with the model thatipRGCs influence retinal networks via glutamatergic signaling,and likely via a dopaminergic pathway.

Light-Sensitive, Noncholinergic Wave Circuit Is Present but Latent inWT Mice. The differential modulation by light and dopamine re-ceptor antagonists between cholinergic and recovered gap junc-tion waves indicates that they are mediated by distinct circuits.Does a noncholinergic wave circuit exist as a “latent” circuit inWT retinas, or does it require an extended perturbation providedby the β2KO mouse? To address this question, we tested whetheracute block of cholinergic waves in WT mice unmasked a light-and dopamine-sensitive wave-generating circuit. We monitoredlight-evoked activity after a 20-min nAChR blockade using di-hydroß-erythroidine (DHβE, 8μM). Approximately 2 min afterlight onset, we observed rhythmic bursting with a periodicity of afew seconds (Fig. 5A), similar to that observed in another studyafter prolonged (10 h) nAChR blockade (7), but we did notdetect propagating waves.We next tested whether recovery of waves was being sup-

pressed by D1 receptor signaling by applying the D1 receptorantagonist SCH23390 for 20 min in the dark to DHβE-treatedretinas, and monitoring activity in the dark and light. Lightstimulation led to the generation of propagating, correlatedwave-like events, with similar spatial-temporal properties towaves in β2KO mice (Fig. 5 B–K). Specifically, connectivity mapsof recovered waves matched those observed in β2KO mice,where highly connected cells had a dispersed, gap junction sig-nature rather than a clustered, cholinergic one (Fig. 5 I and J). Inaddition, the coincidence of highly connected cells with ipRGCswas lower than chance in the light, as observed in β2KO mice(Fig. 5K, **P < 0.01). Finally, although some recovered waveswere present in the dark in some retinas, the frequency of wavesincreased significantly in the light (Fig. 5C, ***P < 0.001). To-gether, these observations show that an auxiliary wave-generat-ing circuit is latent in WT mice, where it is normally suppressedby a combination of nAChR activation together with D1R sig-naling. Light stimulation of ipRGCs facilitates the activation ofthis auxiliary circuit, indicating that ipRGCs contribute to therecovery of correlated spontaneous firing patterns in the absenceof cholinergic waves.

DiscussionIn this study we demonstrate that ipRGCs are used in de-generate circuit mechanisms to maintain correlated spontaneousactivity in the developing retina. These observations show thatretinal wiring diagrams are dynamic and malleable during de-velopment and suggest a unique function for ipRGCs in medi-ating retinal wave plasticity that underlies the maintenance ofcorrelated activity.We found that in the absence of cholinergic waves during

retinal development, a distinct light-modulated wave circuit wasactivated (Figs. 1 and 5). Cx36 was necessary for waves in β2KOmice (Fig. 2), and because neither light nor dopamine antago-nists induced waves in β2-cx36 dKO mice, we postulate that gapjunctions are the likely target of the observed light and dopaminemodulation of noncholinergic waves. Similar modulation of gapjunctions has been observed in many retinal circuits, implicatinggap junctions as sites of plasticity in both developing and adult

retinal circuits. For example, light and dopamine modulation ofcoupling has been extensively described for horizontal cells (36),AII amacrine cells (34, 37), and alpha-ganglion cells (33, 35).Based on the observation that ipRGCs likely signal to the

retina via dopaminergic signaling [(14, 15) but see ref. 30], wepropose the following model. During cholinergic waves, thedominant circuit is starburst amacrine cell release of acetylcho-line onto ganglion cells and other starburst amacrine cells (Fig.6A). Upon cholinergic block, ipRGCs increase their contributionto network dynamics and the dominant circuit becomes a gapjunction coupled network regulated by ipRGCs acting through

A

F

I J K

G H

zH

1

zH

1

10 min20 min

Dark LightDHβE (Dβ)

1 min 1 min

B

Dark LightDHβE+SCH (Dβ+S)

WT in DHβE (Dβ), P6 WT in DHβE+SCH (Dβ+S), P6

D

noit ar udtsruB

]ces [eva

wgni r ud

0

1

2

3

4

5 * nsns

Distance [μm]0 500 Distance [μm]0 500 Distance [μm]0 500

WT in Dβ in light WT in Dβ+S in light

WT in Dβ+S in light, P6

β2KO in lightWT in light

WT in Dβ+Sin light

xedninoitalerro

C

0

150 xedninoitalerro

C

xedninoitalerro

C

0

20

0

25

Length of connectionsfrom highly connected cells [μm]

400 800

ytilibaborpevitalu

muC

secnatsidfo

0

1

10 min20 min

ycneuqerfeva

W]ni

mrep[

WT (ligh

t)

β2KO (li

ght)

WT in Dβ (

light)

WT in Dβ+

S (dark

)

WT in Dβ+

S (ligh

t)

WT (ligh

t)

β2KO (li

ght)

WT in Dβ+

S (ligh

t)

WT (ligh

t)

β2KO (li

ght)

WT in Dβ+

S (ligh

t)(0)0

1

*****

** nsC

0

20

40

eratahts

CG

Rpifotnec reP

]% [

de tce nn ocylhgih WT in

Dβ+S(light)

WT inDβ+S

(chance)

**

E

g ni r u detar

gn iri F] z

H[eva

w

5

0

10

15 * nsns

β2KO in lightWT in light

WT in Dβ+Sin light

Fig. 5. Light-sensitive, noncholinergic wave circuit is present but latent inWT mice. (A and B) MEA recordings from a WT P6 retina after 20 min ofDHβE (Dβ) application in dark (A) and after 20 min of DHβE+SCH (Dβ+S)application in dark (B) (as in Fig. 1A). (C–E) Frequencies (C), burst duration(D), and firing rates (E) of waves in WT (gray), β2KO (blue), and WT in DHβE+SCH (red). Open circles correspond to individual retinas. (F and G) Correlationindices versus interelectrode distance for WT in DHβE alone (F) and DHβE+SCH (G) in the light. (H) Data from Figs. 3A and 5G for WT (gray), β2KO(blue), and WT in DHβE+SCH (red). (I) Connectivity map for WT in DHβE+SCHin the light (as in Fig. 3B). (J) Cumulative probability distributions of thelengths of connections from highly connected cells (WT and β2KO curvescorrespond to those from Fig. 3D). (K) Percent of ipRGCs that are highlyconnected hub neurons for WT in DHβE+SCH in the light compared tochance (box plots as in Fig. 3E). C–E, *P < 0.05, **P < 0.01, one-way ANOVA;C, ***P < 0.001, paired t test; K, **P < 0.01 Wilcoxon rank sum test. nAChRantagonist, DHβE (Dβ), 8 μM; D1R antagonist, SCH23390 (SCH or S), 10 μM.

Kirkby and Feller PNAS | July 16, 2013 | vol. 110 | no. 29 | 12093

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

July

10,

202

0

modulation of dopamine release (Fig. 6B). In this model, we de-pict the ipRGC–DAC synapse as a reciprocal connection, asipRGCs themselves express dopamine receptors (38). How a gapjunction coupled network is activated in the absence of cholinergicwaves remains to be delineated. However, one possibility is thatRGCs experience a change in gap junction coupling in response toreduced cholinergic input. This has previously been shown to oc-cur in rat adrenal medulla, where acute pharmacological block ofnAChRs leads to an increase in dye coupling of adrenal chro-maffin cells together with an increase in junctional currents (39).Further, we found that reduced D1 receptor signaling was

required to recover wave-like events in the absence of cholin-ergic waves in WT retinas (Fig. 5), suggesting that dopaminesignaling may normally suppress the auxiliary wave circuit duringcholinergic waves. In addition, we found that recovered waves inβ2KO mice were suppressed by activation of D1 receptors butstimulated by activation of D2 receptors (Fig. 4B). Because D2receptors are approximately 10-fold more sensitive to dopaminethan D1 receptors (34), a balance of the opposing effects of D1and D2 receptor activation can likely be achieved in vivo witha low concentration of dopamine. Interestingly, although blockadeof D2 receptors blocked waves, it did not decrease the mean firingrate of individual RGCs (Fig. 4C). Hence, we postulate that lightstimulation of ipRGCs does not increase wave frequency by ageneral increase in network excitability, but rather via further in-creases in gap junction coupling, thus lowering the threshold forwave events. Together, these observations suggest that the absenceof cholinergic waves might result in changes in dopamine sig-naling, which facilitate activation of recovered waves and sculpttheir dynamics.Our data suggest that the circuits that mediate both cholin-

ergic and recovered gap junction waves exist in WT retina asopposed to emerging after the prolonged activity blockade thataccompanies genetic deletions. Specifically, a short-term cho-linergic block resulted in periodic activity and correlated wave-like activity emerged when D1 receptor signaling was low (Fig.5). Hence, the wiring diagram of the developing retina manyinclude several “overconnected circuits,” in which some circuitsare closed and others activated depending on the internal stateof the system (2), such as classically described in the stomato-gastric ganglia (1) and recently described in Caenorhabditis ele-gans (40). The data presented here indicate that the developingretina may use a similar overconnection strategy as a means ofmaintaining spontaneous firing patterns. Such a strategy has theadvantage of allowing the network to rapidly change its prop-erties without having to construct new circuits. It is interesting topostulate whether this is a general mechanism used by otherdeveloping networks.

Materials and MethodsAnimals. Recordings were performed on mice aged postnatal day P4–P7 fromC57BL/6 WT (Harlan Laboratories, Indianapolis), β2KO (A. Beaudet, BaylorUniversity, Waco, TX) (41), β2-Opn4 dKO (David Copenhagen, University ofCalifornia, San Francisco), Opn4-GFP (P. Kofuji, Minnesota University, Min-neapolis) (29), β2KO/Opn4-GFP, and β2-cx36 dKO (24). Animal procedureswere approved by the University of California, Berkeley Institutional AnimalCare and Use Committees and conformed to the National Institutes ofHealth Guide for the Care and Use of Laboratory Animals, the Public HealthService Policy, and the Society for Neuroscience Policy on the Use of Animalsin Neuroscience Research. Animals were anesthetized with isofluorane, de-capitated, and the eyes were enucleated in a dark room with dim red am-bient light. Retinas were removed from eyecups in 95% O2–5% (vol/vol) CO2

bicarbonate buffered Ames’ solution (purchased from Sigma-Aldrich) underinfrared optics.

MEA Recordings. Isolated pieces of retina were placed RGC side down ontoa 60-electrode commercial MEA arranged in an 8 × 8 grid excluding the fourcorners, with 10 μm diameter electrodes at 100 μm interelectrode spacing(Multi Channel Systems). The retina was held in place using a dialysis mem-brane weighted with a ring of platinum wire. The recording chamber wassuperfused with Ames’ solution bubbled with 95% O2 and 5% CO2 andmaintained between 33 and 35 °C, pH 7.4. Each preparation was allowed toequilibrate for 20 min in the dark before starting data acquisition. Sponta-neous firing patterns were recorded for 30 min in the dark followed by 30min of unfiltered broad-band full-field light, delivered by a tungsten-halo-gen lamp with irradiance (in photons s−1·cm−2) of 2.4 × 1012 at 480 nm and2.9 × 1013 at 600 nm. This corresponds to a photon flux comparable to thatexperienced by newborn pups through closed eyelids (16). A second series ofdark-light recording conditions was repeated to ensure that any changes infiring patterns were not due to a change in recording conditions over ex-tended periods of time. Raw data were filtered between 120 and 2,000 Hz,and spikes sorted offline to identify single units using Plexon Offline Sortersoftware. The mean firing rate of all units over the duration of the recordingwas calculated and units with a mean firing rate less than 10% of the overallmean firing rate were excluded from further analysis. Spike-sorted datawere analyzed in MATLAB (MathWorks).

To identify wave events, we used a modified Poisson Surprise algorithm,outlined below (9, 42). The recording was divided into 1-s bins and the firingrate of each single unit in each 1-s bin was determined. From this, theprobability of chance occurrence of the firing rate in each bin given a unit’smean firing rate was determined using the Poisson distribution, where theprobability of c spikes occurring in a time bin, t = 1 s, for a mean firing rate r is

Pc =e−rtðrtÞc

c!:

A cell was considered to be bursting if Pc < 10−4 in any given bin. We thenidentified the time bins in which more than 5% of all cells in the recordingwere bursting with Pc < 10−4, and hence computed a pair-wise correlationindex, CI, as a function of distance between two cells for all spikes in thesebins, where

CI=NABðΔtÞ ·T

NAðTÞ ·NBðTÞ · ð2ΔtÞ:

NAB(Δt) corresponds to the number of spike pairs for which unit B fireswithin a time window ±Δt from unit A; NA(T) corresponds to the totalnumber of spikes fired by unit A during the total recording time, T (andsimilarly for NB(T)) (17). We used a correlation time window of Δt = 100 ms.Thus, only spikes in bins that displayed a decreasing nearest neighbor cor-relation index were accepted as waves and considered for analysis of waveproperties. Waves detected using this algorithm agreed with those de-termined by eye. Upon identification of waves, the wave frequency, burstduration during a wave, and firing rate during a wave were computed andaveraged for each unit.

The correlation index was calculated for all cell pairs in each retina. Thedistance between cells was approximated as the distance between theelectrodes of cell pairs. Pairs were then grouped according to intercellulardistance and themedians computed over all cell pairs. Themedian correlationindex was then plotted as a function of increasing intercellular distance. Toestablish connectivity maps, we defined two cells as being connected if theircorrelation index fell in the top 5% of all correlation index values (26). Wethen extracted cells that were connected to at least 15% of other cells (redunits in Fig. 3 B and C). We mapped these units back onto the electrodes on

RGC ipRGCRGC ipRGC

ASAC

nAChR

ACDAC

D1R

B SAC

gap junction

DAC

D2R

AC

DR

Cholinergic waves Gap junction waves

Fig. 6. Model of overlapping wave circuits. (A) During cholinergic waves,the dominant circuit is nAChR activation by acetylcholine, which is sponta-neously released from starburst amacrine cells. Activation of D1 receptorsinhibits the gap junction circuit. (B) Following cholinergic block, ipRGCs maymodulate dopamine release from DACs, which, via activation of D2 recep-tors, modulates the strength of gap junction coupling between ganglioncells (not pictured) or among a network of amacrine and ganglion cells. SAC,starburst amacrine cell; AC, amacrine cell; DAC, dopaminergic amacrine cell;RGC, retinal ganglion cell; ipRGC, intrinsically photosensitive retinal gan-glion cell.

12094 | www.pnas.org/cgi/doi/10.1073/pnas.1222150110 Kirkby and Feller

Dow

nloa

ded

by g

uest

on

July

10,

202

0

which they were recorded, and computed the distances to their connectedunits (blue lines in Fig. 3 B and C).

To identify ipRGCs, we computed the difference in a unit’s mean firing ratein the 60 s following light onset and its mean firing rate in the 60 s precedinglight onset. Most cells followed a narrow normal distribution centered about0 Hz difference (Gaussian fit parameters, μ = 0.07 Hz; σ = 0.35 Hz), which wereclassified as nonipRGCs. Units that showed an increase in mean firing rate of atleast 0.75 Hz (45 additional spikes in a 60-s window) in the light were classifiedas ipRGCs. This classification came from targeted cell-attached recordings froma transgenic mouse line in which GFP is expressed under the melanopsinpromoter (Opn4–EGFP mouse) (29). Cell-attached recordings showed that GFP+ RGCs exhibited an increase in mean firing rate following light onset of atleast 0.75Hz, whereas the difference in mean firing rates for non-GFP+ cellsfell into a cluster centered around 0 Hz (refer to Fig. S2). Each MEA unit wasinspected manually to verify that ipRGCs classified in this manner showeda light response. These units were mapped back on to the electrodes on whichthey were recorded (turquoise units in Fig. 3 B and C).

Pharmacology. DHβE (8 μM), D-AP5 (50 μM), DNQX (20 μM), SCH23390 hy-drochloride (10 μM), and raclopride (8 μM) were added to Ames’ media asstock solutions prepared in either distilled water (DHβE, D-AP5, DNQX, andSCH23390) or DMSO (raclopride). Antagonists were purchased from Tocris.

Electrophysiology. Isolated retinas were mounted RGC side up on filter paperover a small viewing hole. Retinas were superfused with Ames’ solutionbubbled with 95% O2 and 5% CO2 and maintained between 33 and 35 °C,pH. 7.4. Retinas were visualized with differential interference contrastoptics on an Olympus BX51WI microscope under a LUMPlanFL 60× water-immersion objective. ipRGCs were identified by GFP signal under epifluor-escent illumination at 488 nm. A hole was pierced in the inner limitingmembrane of the retina using a glass recording pipette to access the RGClayer. RGCs were targeted under control of a micromanipulator (MP-225,Sutter Instruments). Recording pipettes were pulled with a tip resistance of4–5 MΩ (Sutter Instruments) and filled with filtered NaCl (150 mM). Datawere acquired using pCLAMP 10.2 recording software and a Multiclamp700B amplifier (Molecular Devices), sampled at 6 kHz and filtered between120 and 2,000 Hz.

ACKNOWLEDGMENTS. We thank David Copenhagen for sharing his β2-Opn4 dKO mouse line with us. We also thank the anonymous reviewersfor their constructive feedback on an earlier version of this paper. Thiswork was supported by the National Institutes of Health R01 GrantEY013528, National Science Foundation Graduate Research FellowshipProgram (to L.A.K.), and Boehringer Ingelheim Fonds PhD Fellowship(to L.A.K.).

1. Nusbaum MP, Blitz DM, Swensen AM, Wood D, Marder E (2001) The roles of co-transmission in neural network modulation. Trends Neurosci 24(3):146–154.

2. Bargmann CI (2012) Beyond the connectome: How neuromodulators shape neuralcircuits. Bioessays 34(6):458–465.

3. Turrigiano G (2012) Homeostatic synaptic plasticity: Local and global mechanisms forstabilizing neuronal function. Cold Spring Harb Perspect Biol 4(1):a005736.

4. Blankenship AG, Feller MB (2010) Mechanisms underlying spontaneous patternedactivity in developing neural circuits. Nat Rev Neurosci 11(1):18–29.

5. Sipilӓ ST, et al. (2009) Compensatory enhancement of intrinsic spiking upon NKCC1disruption in neonatal hippocampus. J Neurosci 29(21):6982–6988.

6. Wenner P (2011) Mechanisms of GABAergic homeostatic plasticity. Neural Plast 2011:489470.

7. Stacy RC, Demas J, Burgess RW, Sanes JR, Wong ROL (2005) Disruption and recovery ofpatterned retinal activity in the absence of acetylcholine. J Neurosci 25(41):9347–9357.

8. Sun C, Warland DK, Ballesteros JM, van der List D, Chalupa LM (2008) Retinal waves inmice lacking the beta2 subunit of the nicotinic acetylcholine receptor. Proc Natl AcadSci USA 105(36):13638–13643.

9. Stafford BK, Sher A, Litke AM, Feldheim DA (2009) Spatial-temporal patterns ofretinal waves underlying activity-dependent refinement of retinofugal projections.Neuron 64(2):200–212.

10. Provencio I, Jiang G, De Grip WJ, Hayes WP, Rollag MD (1998) Melanopsin: An opsin inmelanophores, brain, and eye. Proc Natl Acad Sci USA 95(1):340–345.

11. Sernagor E (2005) Retinal development: Second sight comes first. Curr Biol 15(14):R556–R559.

12. Berson DM (2003) Strange vision: Ganglion cells as circadian photoreceptors. TrendsNeurosci 26(6):314–320.

13. Müller LPDS, Do MTH, Yau K-W, He S, Baldridge WH (2010) Tracer coupling of in-trinsically photosensitive retinal ganglion cells to amacrine cells in the mouse retina. JComp Neurol 518(23):4813–4824.

14. Zhang D-Q, et al. (2008) Intraretinal signaling by ganglion cell photoreceptors todopaminergic amacrine neurons. Proc Natl Acad Sci USA 105(37):14181–14186.

15. Zhang D-Q, Belenky MA, Sollars PJ, Pickard GE, McMahon DG (2012) Melanopsinmediates retrograde visual signaling in the retina. PLoS ONE 7(8):e42647.

16. Renna JM, Weng S, Berson DM (2011) Light acts through melanopsin to alter retinalwaves and segregation of retinogeniculate afferents. Nat Neurosci 14(7):827–829.

17. Wong RO, Meister M, Shatz CJ (1993) Transient period of correlated bursting activityduring development of the mammalian retina. Neuron 11(5):923–938.

18. Feller MB, Butts DA, Aaron HL, Rokhsar DS, Shatz CJ (1997) Dynamic processes shapespatiotemporal properties of retinal waves. Neuron 19(2):293–306.

19. Sernagor E, Eglen SJ, Wong RO (2001) Development of retinal ganglion cell structureand function. Prog Retin Eye Res 20(2):139–174.

20. Demas J, Eglen SJ, Wong ROL (2003) Developmental loss of synchronous spontaneousactivity in the mouse retina is independent of visual experience. J Neurosci 23(7):2851–2860.

21. McLaughlin T, Torborg CL, Feller MB, O’Leary DDM (2003) Retinotopic map re-finement requires spontaneous retinal waves during a brief critical period of de-velopment. Neuron 40(6):1147–1160.

22. Tu DC, et al. (2005) Physiologic diversity and development of intrinsically photosen-sitive retinal ganglion cells. Neuron 48(6):987–999.

23. Neumann T, Ziegler C, Blau A (2008) Multielectrode array recordings reveal physio-logical diversity of intrinsically photosensitive retinal ganglion cells in the chick em-bryo. Brain Res 1207:120–127.

24. Torborg CL, Hansen KA, Feller MB (2004) High frequency, synchronized burstingdrives eye-specific segregation of retinogeniculate projections. Nat Neurosci 8(1):72–78.

25. Bloomfield SA, Völgyi B (2009) The diverse functional roles and regulation of neu-ronal gap junctions in the retina. Nat Rev Neurosci 10(7):495–506.

26. Blankenship AG, et al. (2011) The role of neuronal connexins 36 and 45 in shapingspontaneous firing patterns in the developing retina. J Neurosci 31(27):9998–10008.

27. Feldt S, Bonifazi P, Cossart R (2011) Dissecting functional connectivity of neuronalmicrocircuits: Experimental and theoretical insights. Trends Neurosci 34(5):225–236.

28. Bonifazi P, et al. (2009) GABAergic hub neurons orchestrate synchrony in developinghippocampal networks. Science 326(5958):1419–1424.

29. Schmidt TM, Taniguchi K, Kofuji P (2008) Intrinsic and extrinsic light responses inmelanopsin-expressing ganglion cells during mouse development. J Neurophysiol100(1):371–384.

30. Cameron MA, et al. (2009) Light regulation of retinal dopamine that is independentof melanopsin phototransduction. Eur J Neurosci 29(4):761–767.

31. Yoshida M, et al. (2011) Overexpression of neurotrophin-3 stimulates a second waveof dopaminergic amacrine cell genesis after birth in the mouse retina. J Neurosci31(35):12663–12673.

32. Witkovsky P (2004) Dopamine and retinal function. Doc Ophthalmol 108(1):17–40.33. Mills SL, et al. (2007) Dopaminergic modulation of tracer coupling in a ganglion-

amacrine cell network. Vis Neurosci 24(4):593–608.34. KothmannWW, Massey SC, O’Brien J (2009) Dopamine-stimulated dephosphorylation

of connexin 36 mediates AII amacrine cell uncoupling. J Neurosci 29(47):14903–14911.35. Hu EH, Pan F, Völgyi B, Bloomfield SA (2010) Light increases the gap junctional

coupling of retinal ganglion cells. J Physiol 588(Pt 21):4145–4163.36. Bloomfield SA, Xin D, Persky SE (1995) A comparison of receptive field and tracer

coupling size of horizontal cells in the rabbit retina. Vis Neurosci 12(5):985–999.37. Mills SL, Massey SC (1995) Differential properties of two gap junctional pathways

made by AII amacrine cells. Nature 377(6551):734–737.38. Van Hook MJ, Wong KY, Berson DM (2012) Dopaminergic modulation of ganglion-

cell photoreceptors in rat. Eur J Neurosci 35(4):507–518.39. Martin AO, Mathieu M-N, Guérineau NC (2003) Evidence for long-lasting cholinergic

control of gap junctional communication between adrenal chromaffin cells. J Neu-rosci 23(9):3669–3678.

40. Macosko EZ, et al. (2009) A hub-and-spoke circuit drives pheromone attraction andsocial behaviour in C. elegans. Nature 458(7242):1171–1175.

41. Xu W, et al. (1999) Multiorgan autonomic dysfunction in mice lacking the beta2 andthe beta4 subunits of neuronal nicotinic acetylcholine receptors. J Neurosci 19(21):9298–9305.

42. Kerschensteiner D, Wong ROL (2008) A precisely timed asynchronous pattern of ONand OFF retinal ganglion cell activity during propagation of retinal waves. Neuron58(6):851–858.

Kirkby and Feller PNAS | July 16, 2013 | vol. 110 | no. 29 | 12095

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

July

10,

202

0