REVIEW

Retinal ganglion cell interactions shape the developingmammalian visual systemShane D’Souza1,2,3,* and Richard A. Lang1,2,4,5

ABSTRACTRetinal ganglion cells (RGCs) serve as a crucial communicationchannel from the retina to the brain. In the adult, these cells receiveinput from defined sets of presynaptic partners and communicate withpostsynaptic brain regions to convey features of the visual scene.However, in the developing visual system, RGC interactions extendbeyond their synaptic partners such that they guide developmentbefore the onset of vision. In this Review, we summarize our currentunderstanding of how interactions between RGCs and theirenvironment influence cellular targeting, migration and circuitmaturation during visual system development. We describe the rolesofRGCsubclasses in shaping unique developmental responseswithinthe retina and at central targets. Finally, we highlight the utility of RNAsequencing and genetic tools in uncovering RGC type-specific rolesduring the development of the visual system.

KEY WORDS: Retinal ganglion cells, Visual system,Non-autonomous, IpRGCs, Cell-cell interactions, Retina

IntroductionThe retina serves as the visual gateway to the world around us. Thislight-sensitive neural tissue is located in the posterior of the eye(Fig. 1A) and has evolved to detect photons, extract visual features andconvey this information to the rest of the brain (Baden et al., 2020).Together, the eye and all brain regions involved in processing retinalinput form the visual system (Seabrook et al., 2017). Through thecomplex and integrated function of multiple cell classes in the retina(Fig. 1A), we are capable of sensing, experiencing and responding tovisual scenes that range from subtle and static to vivid and dynamic.Photoreceptors (rods and cones) express photopigment proteins and

detect light. These photoreceptors signal to bipolar interneurons,which then provide excitatory input to retinal ganglion cells (RGCs)(Fig. 1A). Horizontal cells provide inhibitory input to photoreceptors,whereas amacrine cells inhibit RGCs and bipolar cells (Masland,2012). RGCs have axons that extend through the optic nerve andconvey information from the retina to numerous targets in the brain –shaping not only visual perception, but also our physiology andcircadian behavior (Schmidt et al., 2011; Baden et al., 2020).

Unraveling how visual processes take place requires an understandingof the cell diversity, mapping their connections, decipheringcommunication patterns and determining how each cell is generated.In addition, it is vital to determine how transient developmentalinteractions between cells shape the visual system. By systematicallyaddressing these various aspects of the visual system, we can slowlybegin to tease apart the complexities of visual mechanisms.

RGCs, the sole projection neurons of the retina, have been well-studied for their diversity, connections, signaling and development.Between electrophysiological profiling and single-cell transcriptomics,this class can be further divided into 30-46 distinct types inmice (Badenet al., 2016; Tran et al., 2019). Each RGC type optimally encodesunique features of the visual scene (contrast, motion, color, etc.), andprojects to diverse ‘central targets’ in the brain that decode each feature(Dhande et al., 2015). Some RGC types express the photopigmentmelanopsin (Opn4), conferring them with direct photosensitivity(Berson et al., 2002; Hattar et al., 2002). Collectively, these multiplephotosensitive RGCs are referred to as intrinsically photosensitiveRGCs (ipRGCs) and make up a subclass representing six unique RGCtypes (M1-M6; reviewed by Sondereker et al., 2020).

Interestingly, RGCs also represent one of the first cell classesgenerated during retinal development (Fig. 1B) (Young, 1985) andtheir cellular diversity is established by postnatal day (P) 5 (Rheaumeet al., 2018; Tran et al., 2019), before eye opening in mice. Theirnumerous interaction partners in the brain and retina, in additionto their early generation, indicate that RGCs could regulatedevelopmental events along the visual pathway. Over the last twodecades, evidence has emerged indicating that RGCs function asmorethan visual feature detectors and information conduits. Althoughsparse (∼2.7% of all retinal cells) (Young, 1985), RGCs serve as adevelopmental nexus for the visual system. By interacting withmigrating retinal neurons, retinal and vascular progenitors, astrocytesand central targets, they influence the course of visual systemdevelopment, well before the visual experience.

In this Review, we provide an overview of the evidence suggestingthat RGCs play a developmental role in mammalian visual systemdevelopment. When applicable, we describe key concepts gleanedfrom non-mammalian vertebrate models. We describe how earlyRGC activity, structure and secreted factors establish structural,molecular and physiological properties within the visual system.When available, we highlight data that implicate particular RGCsubclasses (e.g. ipRGCs) as modulators of retinal and centraldevelopment, and how this influences behavior. Finally, wedescribe how next-generation tools (bioinformatics and genetics)can be applied towards understanding how RGCs contribute todevelopment of the retina, brain and mammalian behavior.

The roles of RGCs in retinal developmentThe developmental influence of RGCs can be categorized into twotypes: activity-dependent input (in which RGCs depolarize andrelease neurotransmitters) and activity-independent input (secreted

1The Visual Systems Group, Cincinnati Children’s Hospital, Cincinnati, OH 45229,USA. 2Center for Chronobiology, Abrahamson Pediatric Eye Institute, Division ofPediatric Ophthalmology, Cincinnati Children’s Hospital, Cincinnati, OH 45229,USA. 3Molecular and Developmental Biology Graduate Program, University ofCincinnati, College of Medicine, Cincinnati, OH 45229, USA. 4Division ofDevelopmental Biology, Cincinnati Children’s Hospital, Cincinnati, OH 45229, USA.5Department of Ophthalmology, University of Cincinnati, College of Medicine,Cincinnati, OH 45229, USA.

*Author for correspondence (shane.dsouza@cchmc.org)

S.D., 0000-0001-6344-1434

This is an Open Access article distributed under the terms of the Creative Commons AttributionLicense (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use,distribution and reproduction in any medium provided that the original work is properly attributed.

1

© 2020. Published by The Company of Biologists Ltd | Development (2020) 147, dev196535. doi:10.1242/dev.196535

DEVELO

PM

ENT

factors or structural aspects of the cell). In this section, we describehow these forms of early ‘RGC input’ influence the developingretina, optic tract and optic chiasm (Fig. 1C).

RGC-derived factors modulate progenitor proliferation anddifferentiationThe neural retina (Fig. 1A) is derived from a pool of multipotentretinal progenitor cells (RPCs). These cells give rise to all retinalneurons and Müller glia, largely relying on intrinsic regulation (suchas transcriptional regulators and chromatin state) and extrinsic cues(paracrine factors) to guide specification into one of seven post-mitotic cell classes (Cepko, 2014). Two models of intrinsic regulationhave been proposed to describe how progenitors generate each of theseclasses in a temporally defined sequence (Fig. 1B). The deterministicmodel suggests that intrinsically different RPCs are competent togenerate specific groups of classes (Turner et al., 1990; Trimarchiet al., 2008). By contrast, the stochastic model suggests that RPCs areequipotent and, through a probabilistic process, generate cell classesin the quantities and order observed during retinal development(Cayouette et al., 2003; Gomes et al., 2011; He et al., 2012). Ourcurrent understanding, however, suggests that both modalities ofdevelopment likely occur in the retina (Cepko, 2014). In addition todifferentiation of these classes, RPCs must also self-renew to generatecells in appropriate quantities. Although intrinsic factors appear to be amajor driver of clone size (and thus number of RPC-derived retinalcells) (He et al., 2012), there are documented circumstances in whichRGC-derived signals (extrinsic cues) shape the generation of specificcell classes and cell quantity during development.Early experimental approaches using dissociated or explanted

rodent retina cultures, heterochronic mixing (combining cells ofdifferent developmental ages) and media supplementation suggestthat post-mitotic cells have the capacity to influence RPCdifferentiation (Watanabe and Raff, 1990, 1992; Altshuler andCepko, 1992; Waid and McLoon, 1998). The presence of RGCs (inculture) prevents progenitors from adopting the RGC fate (Waid andMcLoon, 1998), and the responsible RGC factor(s) appears to bediffusible as opposed to juxtacrine. The morphogen sonic hedgehog

(Shh) was found to be expressed by developing RGCs, while itscognate receptor (patched) is expressed by RPCs (Jensen andWallace, 1997; Zhang and Yang, 2001). These discoveries positionthis axis as a feedback pathway that titrates RGC production in theretina. More recent studies have firmly established a developmentalrole for RGC-derived Shh. Murine Shh is initially expressed byRGCs of the central retina at embryonic day (E) 12 (Wang et al.,2005). In fish and mice, Shh expression expands in a central-to-peripheral wave that closely mirrors, but lags behind, the wave ofRGC generation (Neumann and Nuesslein-Volhard, 2000; Wanget al., 2005). RGC-derived Shh serves two major roles during retinaldevelopment: promoting proliferation of RPCs, and simultaneouslysuppressing RGC generation. Loss of Shh in the peripheral retinaleads to reduced RPC proliferation, increased RGC production, andfailure of later-born cells to be generated (Wang et al., 2005).

Another diffusible factor, vascular endothelial growth factor(Vegf), is expressed by RGCs, and its receptor (Flk1; also knownas Kdr) is expressed by endothelial cells (Yamaguchi et al., 1993) andRPCs (Yang and Cepko, 1996). This expression pattern is conservedin chicken, in which secreted Vegf suppresses RGC-genesis andpromotes RPC proliferation, reminiscent of Shh function in thedeveloping retina (Hashimoto et al., 2006). RGC-specific ablationexperiments (Mu et al., 2005), and mice with RGC loss (Gan et al.,1996; Brown et al., 2001), also complement the claim that RGCs are asource of proliferative cues. By contrast, not all RGC-released factorsinfluence RPC proliferation. RGC-derived growth differentiationfactor 11 (Gdf11) inhibits progenitors from adopting the RGC fatewithout altering cell cycle dynamics. Instead, it is thought that Gdf11temporally restricts the expression of key RGC-competence factors(Kim et al., 2005). The gene encoding one such transcription factor,Atoh7, has been extensively studied for its requirement in RGC-genesis through the induction of an RGC-fate transcriptional program(Liu et al., 2001; Wu et al., 2015). In several vertebrate classesassessed (Brown et al., 2001; Kay et al., 2001; Liu et al., 2001),including humans (Prasov et al., 2012), mutations in Atoh7 or itsregulatory elements lead to an almost complete loss of RGCs, withadditional pathologies associated with the eye (Ghiasvand et al.,

OS/IS

ONL

OPL

INL

IPLGCL

A Retina organization B Retinal neuron glia generation C Intra-retinal RGC signaling

Cell

prop

orti

on

Developmental ageE10 P10

ShhGdf11Vegf

ShhPdgfa

Key

Rod Cone Horizontal cell Bipolar cell Amacrine cell RGC Müller glia RPC Astrocyte Vasculature

Fig. 1. Retinal architecture, cell generation and intra-retinal signaling during retinal development. (A) The mammalian neural retina highlightingphotoreceptors (rods, cones), interneurons (horizontal cells, bipolar cells and amacrine cells), projection neurons (RGCs), Muller glia and retinal progenitor cells(RPC) structured into three nuclear/cellular layers (ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer) and two neuropils (OPL, outerplexiform layer; IPL, inner plexiform layer). (B) Developmental generation of mouse retinal cell types (based on Young, 1985) depicting both the sequence ofgeneration and relative proportion of each cell type. (C) Interactions between immature RGCs and other developing cell types in the context of retinaldevelopment. Arrows depict RGC-derived factors described in the text. OS/IS, outer segment/inner segment of photoreceptor cells.

2

REVIEW Development (2020) 147, dev196535. doi:10.1242/dev.196535

DEVELO

PM

ENT

2011; Prasov et al., 2012; Kondo et al., 2016; Atac et al., 2019). Inaddition, recent data suggest that Atoh7+ lineage RGCs non-autonomously promote the survival or generation of Atoh7−

lineage RGC types (Brodie-Kommit et al., 2020 preprint). Thus,RGCs appear to regulate their own production by feeding into RPCtranscriptional programs via paracrine signaling.Although it is clear that RGCs enhance progenitor proliferation

through secreted factors, it is debated whether the RGC-to-RPC axisinfluences fate determination. Studies involving gene deletions of keyfactors in RGC development and production [Shh, Atoh7, Brn3b(Pou4f2), etc.] argue that the RGC-to-RPC signaling axis, in additionto regulating proliferation, differentially influences the productionand survival of certain cell classes (Wang et al., 2002, 2005; Moshiriet al., 2008; Bai et al., 2014). Studies using intersectional genetics tospecifically ablate RGCs (Six3-cre; Brn3bz-DTA) suggest that all non-RGC cell classes are produced in correct proportions but, overall,retinal cell number is greatly reduced (Mu et al., 2005). However,both forms of manipulation have limitations that make it difficult tointerpret their meaning for fate-determination. For example, Shhexpression is not restricted to RGCs during development – murineamacrine cells and cones also express Shh at later developmentalstages (Jadhav et al., 2006). Thus, alterations in germline Shhknockouts may reflect alternative sources of Shh. In addition, thekinetics of RGC death in the Six3-cre; Brn3bz-DTA line is not fullyunderstood, making it possible that transient RGCs are sufficient toshape fate determination. Further work is required to establish RGCsas bona fide modulators of fate determination in the retina.

Non-autonomous regulation of lamination via RGCsDuring development, retinal neurons must be generated in theappropriate numbers, but must also locate their synaptic partners inorder to effectively encode the visual scene. In the first few postnatalweeks, the processes of synaptic formation, pruning and circuitestablishment occur through cellular activity (Demarque et al., 2002;Goldberg et al., 2002), expression of attractive or repulsive cell-surface proteins (Sanes and Zipursky, 2010; Missaire and Hindges,2015; Sanes and Zipursky, 2020), and release of neurotransmitters(Morgan et al., 2008; Kerschensteiner et al., 2009; Blankenship andFeller, 2010). In particular, these events take place in the innerplexiform layer (IPL) (Fig. 1A), a synaptic ‘highway’ in whichbipolar axons meet RGC and amacrine dendrites in distinct laminarsublayers. Here, the visual field is preprocessed by the retina andseparated into parallel channels based on information content, beforebeing conveyed to the brain (Sanes and Zipursky, 2010, 2020).IPL laminar structure is generated via attractive interactions

between presynaptic and postsynaptic cells (Yamagata et al., 2002;Yamagata and Sanes, 2008). Given the early generation, migrationand activity of RGCs (Young, 1985) (Fig. 2A), one might expectRGCs to guide axonal and dendritic lamination of the later-bornbipolar and amacrine classes (Fig. 1B). There is some evidence forthis in the zebrafish retina, in which RGCs are transiently requiredfor appropriate amacrine dendrite lamination in the IPL (Kay et al.,2004). In mice, as we have discussed, studies using genetic ablationor loss of RGCs reveal that ganglion cells may not be required forgeneral targeting or lamination of neuronal processes in themammalian IPL (Mu et al., 2005; Moshiri et al., 2008). It ispossible that, within a circuit, interacting inner retinal neuronsmutually restrict lamination patterns. For example, the ramificationof ON-OFF direction-selective RGC dendrites is dependent on thehomophilic interaction between cadherins expressed by these cellsand their synaptic partners, the starburst amacrine cells (SACs)(Duan et al., 2018). Although it is clear that lamination requires the

interaction between multiple cell types, RGCs do not appear to be amajor influence during construction of the IPL. Further researchinto specific circuits and RGC types will help determine the role ofRGCs in synaptic lamination.

Interestingly, one type of RGC, the M1 ipRGC, extends itsdendrites beyond the IPL during development, and is poised toregulate more than just its synaptic partners. ipRGC types aredefined by their dendritic structure and lamination pattern, somasize, melanopsin expression and projections to central retinal targets(Sondereker et al., 2020). M1 ipRGCs terminate their dendrites inthe outer IPL (Provencio et al., 2002), but during postnataldevelopment, they extend their dendrites towards the outer retina(Fig. 2B,C) making them ‘biplexiform cells’ (Renna et al., 2015).These atypical projections, termed outer retinal dendrites (ORDs),are transient and their distribution is limited to the dorsal retina(Sondereker et al., 2017). The close proximity of these dendrites tocone terminals suggests a potential interaction between these twoclasses of photoreceptor. Indeed, early activity of murine ipRGCs isnecessary to restrict cone photoreceptors to the outer retina (Tuffordet al., 2018). This occurs through an indirect mechanism: ipRGCORDs contact dendrites of dopaminergic amacrine cells (DACs)and through potential regulation of dopamine release, cones arerestricted to their appropriate layers (Fig. 2D). In addition, cone cellbodies are further restricted to the apical aspect of the outer nuclearlayer (ONL) via light- and activity-independent input of ipRGCs(Fig. 1A; Tufford et al., 2018). Taken together, these studies suggestthat, as a class, RGCs do not shape their local synaptic plexus (i.e.the IPL), but signaling of an RGC subclass (ipRGCs) restrictsnon-synaptic partners to their appropriate location.

Developing astrocyte and vascular networks require immatureRGC inputRGC-mediated regulation of visual development is not limited toneuronal cells of the retina (Fig. 2A). As RGCs are generated, theiraxons project towards the optic disc (Fig. 2B,E), fasciculate to formthe optic nerve, and travel great distances to contact central targets. Atthe optic disc, RGCs are poised to interact with cells of the optic stalk,astrocytes and the hyaloid vasculature – a transient structure thatserves as the primary nutrient source for the developing eye (Fig. 2E).Cues originating from developing RGC axons appear to stimulateproliferation of astrocyte precursors (Burne and Raff, 1997).Surprisingly, in addition to its local role in the neural retina, RGC-derived Shh is transported down the axon and promotes astrocyteprecursor and optic stalk progenitor proliferation in the optic disc(Fig. 2E,F; Wallace and Raff, 1999; Dakubo et al., 2003). Overall,RGC-derived Shh serves similar roles across the developing visualsystem, promoting proliferation of retinal, astrocyte and optic stalkprogenitor cells.

Astrocytes enter the retina through the optic disc, migrate radiallyand provide a framework for endothelial cell growth and vasculardevelopment (Fruttiger, 2007; Sun and Smith, 2018). Radialmigration of astrocytes depends on paracrine signaling andstructural components of RGCs: platelet derived growth factoralpha (Pdgfa) and the RGC axon itself. RGC-derived Pdgfa binds toits receptor (Pdgfra) on astrocytes and serves as a mitogen andchemoattractant (Fruttiger et al., 1996; Tao and Zhang, 2016),facilitating migration towards the peripheral retina. While on theirmigratory route, astrocytes closely associate with RGC axons and arebelieved to use them as a pre-existing scaffold for motility and spatialpatterning. In the retina of mice that lack RGCs (Atoh7 knockouts) orhave RGC axon defects (RGC-specific Robo1/2 mutants), retinalastrocytes display polarization defects, inappropriate migratory

3

REVIEW Development (2020) 147, dev196535. doi:10.1242/dev.196535

DEVELO

PM

ENT

patterns and delayed vascular plexus formation (O’Sullivan et al.,2017). Notably, physical RGC-astrocyte interactions are requiredspecifically for directional movement towards the peripheral retina,whereas astrocyte-inner limiting membrane (ILM) interactions are

required for any form of astrocyte movement (Gnanaguru et al., 2013;Tao and Zhang, 2016). Thus, RGCs influence the astrocyte lineage –from generation to migration (Fig. 2E,F), and contribute to thedevelopment of the retinal vascular landscape.

G1

SG2

M

G1

SG2

M

A

B C D

E F

~E12 P0 (birth) P7 P12 (eye opening) P30

GenerationActivity

Inner plexiform

Astrocyte migration

Superficial

IntermediateDeep

Hyaloid

RGC

Vasc

ular

C, D

E, F

OPL

INL

IPLGCL

Cone terminals

DAC M1

Cone Drd4 &lamination

DAC(dopamine)

M1(Glutamate?)

Light (through eyelid)

RGC axon

Neu

ral r

etin

aO

ptic

dis

c

Axon

Pdgfa

Shh

Astrocytes

Immature astrocytes

Astrocyte precursors

Astrocytes

RGCs

Endothelial cells(hyaloid artery)

Optic disc

RGCs

AstrocytesRGCs

Optic disc

Fig. 2. RGCs interact with non-synaptic partners to guide lamination and vascular development. (A) Timeline of retinal ganglion cell (RGC) generation andactivity relative to development of the inner plexiform, migration of astrocytes and vascular growth/regression in the retina. (B) Whole-mount schematicrepresentation of a developing retina depicting astrocytemigration from the optic disc. (C) Dendrites extending fromM1 intrinsically photosensitive retinal ganglioncells (M1) and dopaminergic amacrine cells (DACs) extend beyond the appropriate synaptic location in the inner plexiform layer (IPL) towards the outerplexiform layer (OPL), where cone photoreceptor synaptic terminals are located. (D) Communication between the intrinsically photosensitive retinal ganglion cell(ipRGC) and DAC leads to dopamine release, activation of D4-type dopamine receptor (Drd4) on cones to restrict lamination of the cells to the outer nuclear layer(ONL). (E) Immature RGCs interact with migrating astrocytes and the hyaloid artery through RGC structure (axons) and RGC activity (photoreception),respectively. (F) Timeline of the influence of RGC-derived factors and structure on the glial/astrocyte lineage in the optic disc and to neural retina. GCL, ganglioncell layer; INL, inner nuclear layer.

4

REVIEW Development (2020) 147, dev196535. doi:10.1242/dev.196535

DEVELO

PM

ENT

Following astrocyte migration, endothelial cells are recruitedfrom the optic disc and migrate radially to form a superficialvascular plexus. From this initial plexus, endothelial cells reorienttheir migration and expand towards the outer retina, forming thedeep and intermediate plexus (Fig. 2A) (Sun and Smith, 2018).Beyond RGC-astrocyte interactions, immature retinal neuronactivity has also been associated with the development of thesevascular plexuses (reviewed by Biswas et al., 2020). Spontaneousactivity from retinal neurons, in the form of retinal waves(Huberman et al., 2008), coincides with development of thevascular networks (Weiner et al., 2019). Ablation or chemogeneticinhibition of SACs prevents development of the deep plexus(Weiner et al., 2019). On the other hand, ablating amacrine andhorizontal cells severely reduces development of the intermediateplexus (Usui et al., 2015). Although RGCs participate in retinalwaves, it has yet to be tested whether their wave-associated activityhas a direct effect on vascular development. However, experimentsfocused on RGC subclasses have hinted towards influence of RGCactivity on vascular development.ipRGCs express melanopsin as early as E15 in the mouse

(McNeill et al., 2011), and are one of the earliest establishedphotoreceptors in the neonatal retina (Sekaran et al., 2005; Kirkbyand Feller, 2013; Caval-Holme et al., 2019). During vasculardevelopment, ipRGCs detect light and suppress angiogenesis, whilesimultaneously promoting the timely regression of the hyaloidartery (Rao et al., 2013). Visual deprivation (particularly during thefetal period in mice) or deleting melanopsin (Opn4 knockout mice)leads to persistence of the hyaloid artery and overgrowth of theretinal vascular plexus. Elevated retinal neuron numbers in dark-reared or Opn4 knockout mice leads to a more hypoxic retina,increased retinal Vegfa production and feedback to the vascularcompartment that expresses VEGF receptors (Vegfr2) (Rao et al.,2013; Cristante et al., 2018). In addition, RGCs that express anotheropsin, Opn5, indirectly regulate vascular development (Nguyenet al., 2019). From P3-P8, Opn5-RGCs promote light-mediatedreuptake of the neurotransmitter dopamine in the neural retina. Thisactivity stabilizes the hyaloid artery by preventing the accumulationof dopamine in the vitreous, where it would act on D2 receptors(Drd2) of endothelial cells to suppress the activity of Vegfr2(Fig. 2E). Thus, the activity of opsin-expressing RGCs convergeson hyaloid artery regression, but acts through distinct mechanisms –one via a Vegfa-Vegfr2 pathway, and the other through a dopamine-D2-Vegfr2 pathway. In support of RGCs as regulators of hyaloidregression, loss of Atoh7 in mice and humans leads to persistence ofthe hyaloid artery and failure of the retinal vasculature to effectivelydevelop (Ghiasvand et al., 2011; Edwards et al., 2012; Prasov et al.,2012). Together, the RGC population is a vital signaling andstructural hub that establishes the framework for the development ofretinal astrocytes and vascular architecture.

Inter-RGC interactions in the developing visual systemThus far, we have discussed the various interactions between RGCsand other cell classes in the developing retina. However, RGCs alsointeract with each other to guide their own development. RGC-to-RGC interactions take place within the retina and at the optic chiasm(where eye-specific axons cross) and appear to be a component ofbinocular vision development.During development, spontaneous bursts of synchronized

spatiotemporal neuron activity sweep across the retina (termedretinal waves) (Fig. 3A). These correlated bursts of activity aregenerated through distinct mechanisms over developmental time,and are thought to refine RGC synaptic connections with image-

forming brain regions (Fig. 3B; Wong, 1999; Firth et al., 2005;Huberman et al., 2008; Blankenship and Feller, 2010). As a retinalwave sweeps across the ganglion cell layer (GCL), neighboringRGCs fire action potentials followed by a brief refractory period, inwhich a wave can no longer propagate. This information is receivedby the visual thalamus and midbrain (McLaughlin et al., 2003;Ackman et al., 2012; Burbridge et al., 2014) and the competition ofeye-specific wave activity drives refinement of contralateral andipsilateral projections (Fig. 3C,D).

RGCs do not passively participate in waves, but rather activelyinfluence eye-specific segregation through diverse mechanisms. Theearliest forms of retinal waves (phase I waves) are thought to bepropagated by connexin (Cx) gap-junction networks betweendendrites of neighboring RGCs (Bansal et al., 2000; Syed et al.,2004) (Fig. 3B). The mechanisms that generate phase I waves are stillpoorly understood; however, gap-junction propagation plays a role inaxon refinement at the dorsal lateral geniculate nucleus (dLGN) of thevisual thalamus. Mice that lack gap-junction proteins Cx36 and Cx45(Cx36/45 double knockout) do not display alterations in phase II or IIIretinal waves, but have altered inter-wave firing patterns (Blankenshipet al., 2011). As a consequence, eye-specific projections in the dLGNare less segregated than wild-type or Cx45 knockout mice. Still, theseresults warrant further investigation for two reasons. First, Cx36 andCx45 are both expressed in dLGN neurons, which are recipients ofRGC input. Therefore, it is possible that defects observed in the Cx36/45 double knockout arise as a consequence of intrinsic changes at thetarget rather than within the retina (Belluardo et al., 2000; Maxeineret al., 2003; Lee et al., 2010; Zlomuzica et al., 2010). Second,amacrine and bipolar cells express Cx36/45 and may influence retinalwaves independently of the RGC network (Mills et al., 2001; Han andMassey, 2005; Hansen et al., 2005). Targeted deletion of all detectedconnexins from RGCs (Cx36, Cx45 and Cx30.2) will provide betterinsight into the role of inter-RGC gap-junctions in shaping retinalmapping (Völgyi et al., 2009; Müller et al., 2010; Akopian et al.,2014).

During development, ipRGCs also form gap-junction networkswith neighboring ipRGCs and conventional RGCs (Kirkby andFeller, 2013; Caval-Holme et al., 2019). Through these intra-retinalgap-junction circuits, ipRGCs increase their own sensitivity to lightand simultaneously confer light-sensitivity to conventional RGCs(Caval-Holme et al., 2019). In the absence of phase II retinal waves(β2nAChR knockout and ChAT knockout mice), the retinagenerates compensatory light-dependent waves that are driven bythese ipRGC circuits (Kirkby and Feller, 2013; Stacy et al., 2005).Initially, experiments implicated melanopsin-driven light responsesas modifiers of retinal waves and retinal mapping to the dLGN(Renna et al., 2011). More recent analysis, by contrast, suggests that,although ipRGCs participate in phase II waves, their light-dependent activity does not modulate wave properties (Kirkbyand Feller, 2013; Chew et al., 2017). M1 ipRGCs, however, appearto be required physically for normal wave properties and maturationof the visual system. Ablation of M1 ipRGCs (Opn4DTA line) altersproperties of phase II retinal waves, leading to failure of eye-specificsegregation of retinal-dLGN projections (Chew et al., 2017). As M1ipRGCs sparsely innervate the dLGN (Hattar et al., 2006; Li andSchmidt, 2018) their interactions likely reflect intra-retinal signalingas opposed to target maturation (ipRGC-to-dLGN axis).Interestingly, M1 ipRGCs project axon collaterals into the innerretina (Joo et al., 2013; Chew et al., 2017) and participate in bothdendritic (Tufford et al., 2018) and axon collateral signaling toamacrine cells (Zhang et al., 2008, 2012; Atkinson et al., 2013;Newkirk et al., 2013; Prigge et al., 2016) – thus, these two signaling

5

REVIEW Development (2020) 147, dev196535. doi:10.1242/dev.196535

DEVELO

PM

ENT

modalities may explain how ipRGCs shape retinal wave dynamics.Given that the photopigment melanopsin is unnecessary for retinalwave initiation or wave properties, it is still unclear whether activity-dependent or independent mechanisms are employed by M1ipRGCs to regulate intra-RGC and visual circuit development.

RGC crosstalk at the optic chiasm shapes binocular visionAlthough ipRGCs appear to regulate signaling within the localenvironment of the retina, some RGCs project their axons to the opticchiasm and into the retina of the other eye (Fig. 3C). These structures,termed retino-retinal or ‘R-R projections’, have been described inamphibians (Bohn and Stelzner, 1981; Humphrey and Beazley, 1985;Tóth and Straznicky, 1989; Tennant et al., 1993), birds (McLoon andLund, 1982; Thanos, 1999) and mammals (Bunt and Lund, 1981;Braekevelt et al., 1986; Müller and Holländer, 1988; Nadal-Nicoláset al., 2015; Tang et al., 2016). Initially, they were considered non-functional or an experimental artifact, but recent work suggests thatR-R projections likely serve a role during development of the visualsystem. R-R projections have been traced to the GCL of the opposingretina, in which they closely associate with immature SACs(pacemakers of phase II retinal waves) (Fig. 3B; Murcia-Belmonteet al., 2019). These structures likely represent bona fide retinal

projections, because the Unc5 axon guidance molecule they expressis both necessary and sufficient to generate R-R projections.Although it is difficult to manipulate or record activity of theseprocesses, mathematical modeling suggests that R-R projections mayprovide inter-retinal synchrony to maximize bilateral eye-specificsegregation (Murcia-Belmonte et al., 2019). This is especially true inorganisms that have disparate size relationships between the retinaand central targets (Murcia-Belmonte et al., 2019). Organisms withsimilarly sized retina and targets (zebrafish) appear to lack thesetransient RGC projections and retinal Unc5 expression, primarilyrelying on molecule-guided retinal mapping and less on retinalwaves. Conversely, in mice, in vivo recording of retinal waves at thesuperior colliculus reveals an unexpected level of correlation betweenwaves emerging from both retinae (Ackman et al., 2012), leadingauthors to speculate that R-R projections functionally synchronizeactivity in each eye.

R-R projections reflect a small fraction of RGC projections, as themajority direct their axons to central targets. Along their journeyfrom the retina to the brain, axons from both eyes converge at theoptic chiasm (Fig. 3C), at which a portion cross over (contralateralor c-RGC) and the remainder maintain their trajectory (ipsilateral ori-RGC). These contra- and ipsilateral-projecting pathways are the

?A Interneuron-to-RGC signaling B Timeline and mechanism of retinal wave generation

SAC BC

RGCD

irec

tion

of

wav

e ac

tivi

ty

Phase I

Phase II

Phase III

Gap-junction

Cholinergic

Glutamatergic

Starburst amacrine cell

Bipolar cell

Age P0 P10

C D

Centraltargets

Opticchiasm

RGC axons: Contralateral Ipsilateral Overlap

dLGN

vLGN

dLGN

vLGN

Inter-RGC interactions

RGC gap-junctions

ipRGCs

R-R projections

GCL

Fig. 3. Inter-RGC interactions shape early retinal activity. (A) Retinal waves are mediated through release of acetylcholine from starburst amacrine cells(SACs), glutamate via bipolar cells (BC), or through gap-junction coupling between retinal ganglion cells (RGCs). The activity of these cells driveswaves of depolarization across the ganglion cell layer (GCL, right) to central targets involved in binocular vision (C,D) and refines synapses between RGCs andtheir targets. (B) Over developmental time, retinal waves are generated via distinct mechanisms (cell within circle) with partial overlap. Phase I (∼E16-P0), gap-junction-mediated waves; phase II (∼P0-P10), cholinergic waves; and phase III (∼P10-P14), glutamatergic waves. (C) RGCs from each retina extendaxons towards the brain and cross at the optic chiasm or continue along the same side (pink versus purple axons). (D) At the visual thalamus (dLGN, dorsal lateralgeniculate nucleus; vLGN, ventral lateral geniculate nucleus), retinal waves drive refinement of retinal projections from the eye on the same side (ipsilateral; pink)or opposite side (contralateral; purple). Inter-RGC interactions (bottom) are considered to drive a portion of this refinement process, promoting segregationof projection and thus less overlap (black regions in dorsal LGN). ipRGCs, intrinsically photosensitive RGCs; R-R projections, retino-retinal projections.

6

REVIEW Development (2020) 147, dev196535. doi:10.1242/dev.196535

DEVELO

PM

ENT

basis of binocular vision (Rasband et al., 2003) in vertebrates. Theoptic chiasm, where RGC axons cross, is a location where axonguidance molecules are interpreted to be repulsive or attractive(reviewed by Erskine and Herrera, 2007; Rasband et al., 2003). Thisis largely accomplished by interactions between the niche and theintrinsic type of RGC (c-RGC or i-RGC) responding to these cues.As previously discussed, Shh is expressed by RGCs and acts as a

potent mitogen both locally (RPC proliferation) and at the optic disc(astrocyte precursor proliferation). This suggests that it is traffickeddown the axon and released at distal locations (Wallace and Raff,1999; Dakubo et al., 2003). Later in development, Shh is expressedby the earlier-born c-RGC population (Sánchez-Camacho andBovolenta, 2008), whereas the corresponding Shh receptor Boc isexpressed by later-born i-RGCs (Fabre et al., 2010). Given thetemporal sequence of birth and axon targeting, it is speculated thatc-RGCs influence the targeting of i-RGCs. Consistent with theseobservations, c-RGCs produce, traffic and deposit Shh in the opticchiasm before the generation of i-RGCs (Peng et al., 2018). As retinaldevelopment progresses, i-RGCs project their axon to the chiasm, arerepelled by Shh signaling and fail to cross the midline, and thusmaintain their ipsilateral trajectories (Peng et al., 2018; Ferent et al.,2019). i-RGC competency to respond to Shh is mediated byexpression of the receptor Boc, which is both necessary and sufficientfor axon repulsion at the chiasm (Fabre et al., 2010). Taken together,RGCs use multiple mechanisms to shape the development of thebinocular visual system – modulation of retinal waves, inter-retinalsignaling and regulation of axon guidance allow RGCs to shapecommunication with central targets before the initiation of vision.

The roles of RGCs in central target development andbehaviorThe central portion of the visual system comprises brain regionsdedicated to receiving, decoding and dispatching informationarriving from the retina. In mice, RGCs project their axons to morethan 40 unique regions of the brain (Morin and Studholme, 2014;Martersteck et al., 2017), which we collectively refer to as ‘centraltargets’. Retinal axons terminate in the hypothalamus, thalamus andmidbrain, and thus regulate a variety of behaviors that can be broadlycategorized as image forming and non-image forming. In this section,we describe our current understanding of howRGCs shape features ofcentral target cells, influence the innervation of non-retinal inputs andguide the development of behaviors (Fig. 4).

RGCs in the development of image-forming central targetsAs presented in previous sections, RGCs are capable of regulating thedevelopment of multiple cell types within a tissue, through diversemechanisms (activity, secreted factors and structure). This degree ofregulation is true for thalamic targets such as the dLGN, in whichRGCs synapse onto thalamocortical (TC) relay neurons and inhibitoryinterneurons. Spontaneous retinal activity, in the form of retinal waves,is largely responsible for the maturation of dLGN circuitry (Failoret al., 2015), and developmental remodeling of TC relay neurons(El-Danaf et al., 2015) and interneurons (Charalambakis et al., 2019).In addition, layer-specific recruitment of dLGN inhibitory interneuronsis accomplished by spontaneous activity via RGCs (Golding et al.,2014). RGC axons also provide a physical scaffold and secreted cuesto guide development of the dLGN (activity-independent input).Although multiple mechanisms may explain how RGCs recruitinterneurons to the appropriate central target, recent transcriptionalprofiling of the LGN in the absence of retinal input provides someinsight into these complex interactions. Within the developing LGN,RGC axons stimulate local astrocytes to produce fibroblast growth

factor 15 (Fgf15), which serves as a chemoattractant for newlygenerated, migrating inhibitory interneurons (Su et al., 2020).

A large portion of input to the visual thalamus consists of axonsoriginating from the cortex (corticothalamic axons), thalamicreticular nucleus and various brainstem nuclei (Guido, 2018).These non-retinal inputs modulate thalamic signaling behavior andare integral to transmission of retinal input to multiple dLGN targets.Given that both retinal and non-retinal axons terminate in the LGN, itis not surprising that both timing and targeting are regulated viaaxonal interactions. Retinal axons, which arrive at the dLGN early,limit dLGN innervation by corticothalamic axons (Brooks et al.,2013; Seabrook et al., 2013) by inhibiting the degradation of therepulsive chondroitin sulfate proteoglycan (CSPG) aggrecan (Brookset al., 2013). This axon innervation pathway involves RGC regulationof dLGN interneurons that release aggrecanases (enzymes that breakdown aggrecan) and serves as a developmental switch to regulatetimely innervation of the thalamus. Conversely, projections from thebrainstem (ascending cholinergic input) rely on RGCs to promotetheir timely arrival and axon arborization in the dLGN (Sokhadzeet al., 2018). As such, RGCs regulate local events (such as astrocytesignaling and interneuron recruitment) and long-range events (i.e.timely recruitment of corticothalamic and brainstem projections)during thalamic development.

Beyond the visual thalamus, RGCs project axons to other nucleiinvolved in image forming vision, such as the superior colliculus ofthe midbrain (SC), as well as nuclei involved in image stabilization,including the medial terminal nucleus (MTN), dorsal terminalnucleus (DTN) and nucleus of the optic tract (NOT). As in thedLGN, complementary ligand-receptor interactions recruit RGC-subclass-specific axons to the MTN and SC (Osterhout et al., 2015;Sun et al., 2015; Ito and Feldheim, 2018). However, whether RGCsinfluence the development of circuits, non-retinal input or cellswithin these targets remains to be explored. Given that a single RGCinnervates multiple targets (Ellis et al., 2016), and known RGCtypes that contact the dLGN simultaneously innervate the SC, NOT,DTN and MTN (Kay et al., 2011), it is likely that these centraltargets all receive similar developmental input from RGCs.

ipRGCs regulate maturation of the circadian clock and circadianentrainmentThe unique ability of ipRGCs to detect photons is matched by theirunique projections to central targets. In addition to the visualthalamus, ipRGC axons terminate in multiple nuclei, most notablythe suprachiasmatic nuclei (SCN), the intergeniculate leaflet (IGL)and the olivary pretectal nucleus (OPN) (Hattar et al., 2006). Thesenuclei are collectively termed ‘non-image forming’ because they donot participate in conscious vision; rather, they are involved inbehavioral and circadian processes (Fu et al., 2005). As ipRGCs arethought to exclusively express melanopsin, these cells are readilyaccessed genetically and, therefore, much is known about theirdevelopmental influence on non-image forming central targets andtheir behavioral output.

Throughout the day, rods, cones and melanopsin detect ambientlight, which is conveyed via ipRGC axons to the SCN (the circadianpacemaker). Innervation of the SCN via ipRGCs takes place duringpostnatal life in the mouse (McNeill et al., 2011; Chew et al., 2017),and ipRGCs serve as the primary retinal input to this structure(Hatori et al., 2008; Ecker et al., 2010; Kofuji et al., 2016). ThisipRGC-SCN signaling axis allows the organism to synchronizebehavior with the light-dark cycle (called photoentrainment). In theabsence of an external stimulus or cue, the circadian clock has anendogenous free-running period which is close to – but not exactly –

7

REVIEW Development (2020) 147, dev196535. doi:10.1242/dev.196535

DEVELO

PM

ENT

24 h (Herzog et al., 2017; Hastings et al., 2018). ipRGC activity isnecessary for acute entrainment of the circadian clock with the light-dark cycle, but has been shown to also regulate the free-runningperiod of the circadian pacemaker in mice. Removal of the eyes(enucleation) of mice at birth lengthens the circadian period,whereas enucleation into adulthood (P60) has no effect on settingthe period, suggesting a crucial window for RGC input to shapeSCN output and circadian behavior (Chew et al., 2017). Mice raisedin the absence of light or lacking M1 ipRGCs (Opn4DTA mutants)also have lengthened circadian periods, further implicating light-and activity-dependent ipRGC signaling as a potential mechanismfor period setting. Thus, retinal input from ipRGCs is requireddevelopmentally, to shape the period length of the clock, andacutely, to synchronize the clock with the light-dark cycle.M1 ipRGC interactions with the SCN not only set the pace of the

circadian clock, but also impact development of non-retinal SCNcircuitry. The IGL provides input to the SCN and is involved innonphotic entrainment of the circadian clock. In the absence of alight-dark cycle, an animal can behaviorally entrain to events thatoccur at a predictable time across the circadian day, such as time-restricted feeding (Janik and Mrosovsky, 1994; Edelstein and Amir,1999). ipRGC innervation of the SCN is necessary for the expression

of neuropeptide Y (NPY) in IGL neurons projecting to the SCN(Fernandez et al., 2020). Without ipRGC innervation and NPYrelease at the SCN, the ability to anticipate time-restricted feedingcues is greatly diminished (Fernandez et al., 2020). Together withretinal input-driven thalamic development, these data lead us toreason that RGC input is required to sculpt and shape the intrinsicproperties, non-retinal inputs and behavioral output that emerge fromcentral targets in the brain, independently of their adult roles.

Leveraging transcriptomic and genetic tools to unravel RGCtype-specific regulation of visual system developmentThis Review highlights the developmental roles of RGCs, roles thatshape and tune the visual system. As discussed, developmental RGCinteractions are diverse and not limited to their adult synaptic partners.As we uncover the diversity of RGCs as a class, a few questionsbecome apparent: what are the contributions of other RGC subclassesand types, if any, to visual system development? Is there amechanistic bias for certain RGC types within their regulatory rolesduring development (synaptic, structural or peptide release)?

In the last decade, genetic approaches using Cre recombinasemouse lines that target entire RGC subclasses have provided insightinto these questions. This is particularly true for the ipRGC subclass

A

B

Central targets of RGCs – 40+ regions (mouse) Targets

Hypothalamus Thalamus Midbrain

Retina

RGCs

Activity

Releasedfactors

Neuronrecruitment

Circuitmaturation

Non-retinalinnervation

Central targets during development

Behavioral output

Retina

RGCs

Activity

Releasedfactors

Neuronrecruitment

Circuitmaturation

Non-retinalinnervation

Central targets during development

SCN

dLGNvLGNIGL

SCOPN

AstrocyteImmatureinterneuron

Maturinginterneuron

Non-RGCprojection

RGCaxon

Key

Fig. 4. RGC-regulated central targetdevelopment. (A) Projections from theretina terminate in unique central targetlocations –more than 40 in themouse brain(Martersteck et al., 2017) – that correspondto regions of the hypothalamus (blue),thalamus (red) and midbrain (yellow).Coronal brain sections highlight majortargets of retinal ganglion cells (RGCs) inthe hypothalamus (SCN, suprachiasmaticnucleus), thalamus (dLGN, dorsal lateralgeniculate nucleus; IGL, intergeniculateleaflet; vLGN, ventral lateral geniculatenucleus) and midbrain (OPN, olivarypretectal nucleus; SC, superior colliculus).(B) RGC projections from the retinainfluence the recruitment of interneuronsvia astrocytes, maturation of thalamiccircuitry and non-retinal innervation ofcentral targets.

8

REVIEW Development (2020) 147, dev196535. doi:10.1242/dev.196535

DEVELO

PM

ENT

that express Opn4 and can be accessed through the Opn4cre mouseline (Ecker et al., 2010). The remarkably diverse influence of ipRGCson visual system development (Rao et al., 2013; Johnson et al., 2010;Tufford et al., 2018), circadian behavior (Güler et al., 2008; Chewet al., 2017; Fernandez et al., 2020; Hatori et al., 2008), bodytemperature (Rupp et al., 2019), mood (Fernandez et al., 2018; Huanget al., 2019), learning (Fernandez et al., 2018) and sleep (Rupp et al.,2019) have been assessed using the specificity ofOpn4 expression incombination with Cre-dependent ablation (Rosa26iDTR) and neuronalsilencing alleles (Rosa26TeNT). It is also clear that, within the ipRGCsubclass, distinct RGC types are integral for the development ofparticular behaviors. These discoveries were made possible throughintersectional genetics, with knowledge of RGC type-specific genes(e.g. non SCN-projecting M1 ipRGCs express Opn4 but not Brn3b)and intersectional alleles that allow for recombinase-mediatedmanipulation of the cell (Opn4cre; Pou4f2z-DTA). It is unlikely thatipRGC types are alone in their roles as developmental regulators.Given that the entire RGC repertoire is established rather early, it islikely that many RGC types and subclasses contribute todevelopment of the retinal and central landscapes.Recent RGC single-cell RNA-sequencing (scRNA-seq) and

physiological profiling studies provide us with a foundation toaddress RGC type-specific developmental questions (Baden et al.,2016; Rheaume et al., 2018; Tran et al., 2019). Sequencing hasrevealed candidate marker genes that may exist as recombinase lines(Cre or Flp), which can offer us genetic access to specific RGC types.Although these studies suggest that RGC types can be described by

combinations of at least two genes (Tran et al., 2019), dual-recombinase systems are currently implemented to study retinalneuron types in the adult (Jo et al., 2018) and have the potential to beapplied to visual system development. Combining recombinase lines(single or dual) with recombinase-dependent ablation/toxin alleles(Mu et al., 2005) or signaling modifying alleles (Huang and Zeng,2013; Madisen et al., 2015) would allow us to explore the importanceof each RGC type and potential activity-dependent or -independentmechanisms on development of the visual system. In addition,temporal control of ablation (Rosa26iDTR), silencing (hM4Di) oractivation (hM3Dq) can be used to assess when RGC type-specificsignaling influences the developmental landscape of the visualsystem. Considering that the influence of RGCs is not limited to asingle cell type within the visual system, it will be important todetermine the broader influence of these signaling hubs – particularlythrough unbiased techniques, such as scRNA-seq, proteomics andbehavioral analysis (Fig. 5). Together, scRNA-seq data and geneticsmay shed light on the emerging developmental roles of RGCs.

ConclusionTo summarize, we have reviewed the emerging roles of RGCs in thedevelopment of the visual system. Along the visual pathway, fromretina to targets, we have discussed how RGC activity-dependentand activity-independent inputs regulate the development of diversecell types within a given niche (neuronal and non-neuronal). Manyquestions remain to be addressed, such as whether all RGC types areequally poised to regulate development, or which components of

RGC diversityRGC type

genetic accessRecombinase-dependent

RGC manipulation

Organism behavior:Central development:Retinal development:

Molecular profiling

Physiologicalprofiling

Marker 1Marker 2Marker 3Marker 4Marker 5 ...

Type #1

Type #2

Cre

Flp

Cre

Flp

Marker 1

Marker 2

Marker 3

Marker 4

Pou4f2DTA

Rosa26DTR

Ablation alleles

Silencing alleles

Rosa26TeNT

Rosa26hM4Di

Single-cell transcriptomics

Cell number and diversity

Projection mapping

Physiology and circuit maturation

Visual tasks

Circadian behavior

Fig. 5. Deciphering the roles ofunique RGC types on thedevelopment of the visual system.To address the fundamental roles ofeach of the retinal ganglion cell (RGC)types requires a comprehensiveunderstanding of the typology of thesecells (RGC diversity). The diversity ofRGCs can be defined by their uniquetranscriptomes (Tran et al., 2019) orelectrophysiological responses tovisual stimuli (Baden et al., 2016), or acombination of the two. From this, wecan define unique combinations ofgenes that define types of RGCs (e.g.marker 1 and 2 define type #1, andmarker 3 and 4 define type #2, etc.)and gain genetic access to these cellsby using a combination ofrecombinase types (e.g. Cre, Flp)driven by subclass-specific genes.Furthermore, genetic access to RGCtypes would allow us to test whetherthe cell or its activity is required.Intersectional recombinase-dependent toxin lines (DTA, diphtheriatoxin A subunit; DTR, diphtheria toxinreceptor) or silencing lines (hM4Di,human muscarinic receptor 4 coupledto Gi; TeNT, tetanus toxin) can beapplied with RGC type-specificrecombinase(s). Finally, as RGCsregulate intra- and extra-retinaldevelopment, it will be vital to assessthe changes that occur within andoutside the retina to elucidate thedevelopmental roles of RGC types.

9

REVIEW Development (2020) 147, dev196535. doi:10.1242/dev.196535

DEVELO

PM

ENT

vision are a consequence of developmental or adult activity ofRGCs. With the derivation of new tools and ‘-omics’ datasets forRGCs, we can begin to determine the RGC-autonomous, non-autonomous and type-specific roles in shaping and establishing thevisual system – from molecules and synapses, to circuitry andbehavior. Clinically, these forms of integrative multi-systemanalyses will undoubtedly influence our understanding of howearly human RGC loss in conditions such as anophthalmia, primarycongenital glaucoma (Wilson and Di Polo, 2012) and Atoh7 loss-of-function (Prasov et al., 2012; Atac et al., 2019) influence visualsystem development and behavior beyond conscious vision.

AcknowledgementsWe would like to thank our reviewers, Allie Pendery, Brian A. Upton and Dr JoshuaR. Sanes for their helpful insight while preparing this Review.

Competing interestsThe authors declare no competing or financial interests.

FundingThis work was supported by the National Institutes of Health (R01 EY027077 andR01 EY027711 to R.A.L.), in addition to funds provided by the Goldman Chair of theAbrahamson Pediatric Eye Institute at Cincinnati Children’s Hospital Medical Center.Deposited in PMC for release after 12 months.

ReferencesAckman, J. B., Burbridge, T. J. and Crair, M. C. (2012). Retinal waves coordinatepatterned activity throughout the developing visual system. Nature 490, 219-225.doi:10.1038/nature11529

Akopian, A., Atlasz, T., Pan, F., Wong, S., Zhang, Y., Volgyi, B., Paul, D. L. andBloomfield, S. A. (2014). Gap junction-mediated death of retinal neurons isconnexin and insult specific: a potential target for neuroprotection. J. Neurosci. 34,10582-10591. doi:10.1523/JNEUROSCI.1912-14.2014

Altshuler, D. and Cepko, C. (1992). A temporally regulated, diffusible activity isrequired for rod photoreceptor development in vitro. Development 114, 947-957.

Atac, D. G., Koller, S., Hanson, J. V. M., Feil, S., Tiwari, A., Bahr, A., Baehr, L.,Magyar, I., Kottke, R., Gerth-Kahlert, C. et al. (2019). Atonal homolog 7(ATOH7) loss-of-function mutations in predominant bilateral optic nervehypoplasia. Hum. Mol. Genet. 29, 132-148. doi:10.1093/hmg/ddz268

Atkinson, C. L., Feng, J. and Zhang, D.-Q. (2013). Functional integrity andmodification of retinal dopaminergic neurons in the rd1 mutant mouse: roles ofmelanopsin and GABA. J. Neurophysiol. 109, 1589-1599. doi:10.1152/jn.00786.2012

Baden, T., Berens, P., Franke, K., Roman Roson, M., Bethge, M. and Euler, T.(2016). The functional diversity of retinal ganglion cells in the mouse. Nature 529,345-350. doi:10.1038/nature16468

Baden, T., Euler, T. andBerens, P. (2020). Understanding the retinal basis of visionacross species. Nature Rev. Neurosci. 21, 5-20. doi:10.1038/s41583-019-0242-1

Bai, L., Kiyama, T., Li, H. andWang, S.W. (2014). Birth of cone bipolar cells, but notrod bipolar cells, is associated with existing RGCs. PLoS ONE 9, e83686. doi:10.1371/journal.pone.0083686

Bansal, A., Singer, J. H., Hwang, B. J., Xu, W., Beaudet, A. and Feller, M. B.(2000). Mice lacking specific nicotinic acetylcholine receptor subunits exhibitdramatically altered spontaneous activity patterns and reveal a limited role forretinal waves in forming ON and OFF circuits in the inner retina. J. Neurosci. 20,7672-7681. doi:10.1523/JNEUROSCI.20-20-07672.2000

Belluardo, N., Mudo, G., Trovato-Salinaro, A., Le Gurun, S., Charollais, A.,Serre-Beinier, V., Amato, G., Haefliger, J.-A., Meda, P. and Condorelli, D. F.(2000). Expression of connexin36 in the adult and developing rat brain. Brain Res.865, 121-138. doi:10.1016/S0006-8993(00)02300-3

Berson, D. M., Dunn, F. A. and Takao, M. (2002). Phototransduction by retinalganglion cells that set the circadian clock. Science 295, 1070-1073. doi:10.1126/science.1067262

Biswas, S., Cottarelli, A. and Agalliu, D. (2020). Neuronal and glial regulation ofCNS angiogenesis and barriergenesis. Development 147, dev182279. doi:10.1242/dev.182279

Blankenship, A. G. and Feller, M. B. (2010). Mechanisms underlying spontaneouspatterned activity in developing neural circuits. Nature Rev. Neurosci. 11, 18-29.doi:10.1038/nrn2759

Blankenship, A. G., Hamby, A. M., Firl, A., Vyas, S., Maxeiner, S., Willecke, K.and Feller, M. B. (2011). The role of neuronal connexins 36 and 45 in shapingspontaneous firing patterns in the developing retina. J. Neurosci. 31, 9998-10008.doi:10.1523/JNEUROSCI.5640-10.2011

Bohn, R. C. and Stelzner, D. J. (1981). The aberrant retino-retinal projection duringoptic nerve regeneration in the frog. II. Anterograde labeling with horseradishperoxidase. J. Comp. Neurol. 196, 621-632. doi:10.1002/cne.901960408

Braekevelt, C. R., Beazley, L. D., Dunlop, S. A. and Darby, J. E. (1986). Numbersof axons in the optic nerve and of retinal ganglion cells during development in themarsupial Setonix brachyurus. Dev. Brain Res. 25, 117-125. doi:10.1016/0165-3806(86)90158-6

Brodie-Kommit, J., Clark, B. S., Shi, Q., Shiau, F., Kim, D.W., Langel, J., Sheely,C., Schmidt, T., Badea, T., Glaser, T. et al. (2020). Atoh7-independentspecification of retinal ganglion cell identity. bioRxiv. doi:10.1101/2020.05.27.116954

Brooks, J. M., Su, J., Levy, C., Wang, J. S., Seabrook, T. A., Guido, W. and Fox,M. A. (2013). A molecular mechanism regulating the timing of corticogeniculateinnervation. Cell Rep. 5, 573-581. doi:10.1016/j.celrep.2013.09.041

Brown, N. L., Patel, S., Brzezinski, J. and Glaser, T. (2001). Math5 is required forretinal ganglion cell and optic nerve formation. Development 128, 2497-2508.

Bunt, S. M. and Lund, R. D. (1981). Development of a transient retino-retinalpathway in hooded and albino rats. Brain Res. 211, 399-404. doi:10.1016/0006-8993(81)90712-5

Burbridge, T. J., Xu, H.-P., Ackman, J. B., Ge, X., Zhang, Y., Ye,M.-J., Zhou, Z. J.,Xu, J., Contractor, A. and Crair, M. C. (2014). Visual circuit developmentrequires patterned activity mediated by retinal acetylcholine receptors.Neuron 84,1049-1064. doi:10.1016/j.neuron.2014.10.051

Burne, J. F. and Raff, M. C. (1997). Retinal ganglion cell axons drive theproliferation of astrocytes in the developing rodent optic nerve. Neuron 18,223-230. doi:10.1016/S0896-6273(00)80263-9

Caval-Holme, F., Zhang, Y. and Feller, M. B. (2019). Gap junction coupling shapesthe encoding of light in the developing retina.Curr. Biol. 29, 4024-4035.e5. doi:10.1016/j.cub.2019.10.025

Cayouette, M., Barres, B. A. and Raff, M. (2003). Importance of intrinsicmechanisms in cell fate decisions in the developing rat retina. Neuron 40,897-904. doi:10.1016/S0896-6273(03)00756-6

Cepko, C. (2014). Intrinsically different retinal progenitor cells produce specific typesof progeny. Nature Rev. Neurosci. 15, 615-627. doi:10.1038/nrn3767

Charalambakis, N. E., Govindaiah, G., Campbell, P. W. and Guido, W. (2019).Developmental remodeling of thalamic interneurons requires retinal signaling.J. Neurosci. 39, 3856-3866. doi:10.1523/JNEUROSCI.2224-18.2019

Chew, K. S., Renna, J. M., McNeill, D. S., Fernandez, D. C., Keenan, W. T.,Thomsen, M. B., Ecker, J. L., Loevinsohn, G. S., VanDunk, C., Vicarel, D. C.et al. (2017). A subset of ipRGCs regulates both maturation of the circadian clockand segregation of retinogeniculate projections in mice. eLife 6, e22861. doi:10.7554/eLife.22861

Cristante, E., Liyanage, S. E., Sampson, R. D., Kalargyrou, A., De Rossi, G.,Rizzi, M., Hoke, J., Ribeiro, J., Maswood, R. N., Duran, Y. et al. (2018). Lateneuroprogenitors contribute to normal retinal vascular development in a Hif2a-dependent manner. Development 145, dev157511. doi:10.1242/dev.157511

Dakubo, G. D., Wang, Y. P., Mazerolle, C., Campsall, K., McMahon, A. P. andWallace, V. A. (2003). Retinal ganglion cell-derived sonic hedgehog signaling isrequired for optic disc and stalk neuroepithelial cell development. Development130, 2967-2980. doi:10.1242/dev.00515

Demarque, M., Represa, A., Becq, H., Khalilov, I., Ben-Ari, Y. and Aniksztejn, L.(2002). Paracrine intercellular communication by a Ca2+- and SNARE-independent release of GABA and glutamate prior to synapse formation.Neuron 36, 1051-1061. doi:10.1016/S0896-6273(02)01053-X

Dhande, O. S., Stafford, B. K., Lim, J.-H. A. and Huberman, A. D. (2015).Contributions of retinal ganglion cells to subcortical visual processing andbehaviors. Annu. Rev. Vis. Sci. 1, 291-328. doi:10.1146/annurev-vision-082114-035502

Duan, X., Krishnaswamy, A., Laboulaye, M. A., Liu, J., Peng, Y.-R., Yamagata,M., Toma, K. and Sanes, J. R. (2018). Cadherin combinations recruit dendrites ofdistinct retinal neurons to a shared interneuronal scaffold. Neuron 99,1145-1154.e6. doi:10.1016/j.neuron.2018.08.019

Ecker, J. L., Dumitrescu, O. N., Wong, K. Y., Alam, N. M., Chen, S.-K., LeGates,T., Renna, J. M., Prusky,G. T., Berson, D.M. andHattar, S. (2010). Melanopsin-expressing retinal ganglion-cell photoreceptors: cellular diversity and role inpattern vision. Neuron 67, 49-60. doi:10.1016/j.neuron.2010.05.023

Edelstein, K. and Amir, S. (1999). The role of the intergeniculate leaflet inentrainment of circadian rhythms to a skeleton photoperiod. J. Neurosci. 19,372-380. doi:10.1523/JNEUROSCI.19-01-00372.1999

Edwards, M. M., McLeod, D. S., Li, R., Grebe, R., Bhutto, I., Mu, X. and Lutty,G. A. (2012). The deletion of Math5 disrupts retinal blood vessel and glialdevelopment in mice.Exp. Eye Res. 96, 147-156. doi:10.1016/j.exer.2011.12.005

El-Danaf, R. N., Krahe, T. E., Dilger, E. K., Bickford, M. E., Fox, M. A. and Guido,W. (2015). Developmental remodeling of relay cells in the dorsal lateral geniculatenucleus in the absence of retinal input. Neural Dev. 10, 19. doi:10.1186/s13064-015-0046-6

Ellis, E. M., Gauvain, G., Sivyer, B. and Murphy, G. J. (2016). Shared and distinctretinal input to themouse superior colliculus and dorsal lateral geniculate nucleus.J. Neurophysiol. 116, 602-610. doi:10.1152/jn.00227.2016

10

REVIEW Development (2020) 147, dev196535. doi:10.1242/dev.196535

DEVELO

PM

ENT

Erskine, L. and Herrera, E. (2007). The retinal ganglion cell axon’s journey: insightsinto molecular mechanisms of axon guidance. Dev. Biol. 308, 1-14. doi:10.1016/j.ydbio.2007.05.013

Fabre, P. J., Shimogori, T. andCharron, F. (2010). Segregation of ipsilateral retinalganglion cell axons at the optic chiasm requires the Shh receptor Boc. J. Neurosci.30, 266-275. doi:10.1523/JNEUROSCI.3778-09.2010

Failor, S., Chapman, B. and Cheng, H.-J. (2015). Retinal waves regulate afferentterminal targeting in the early visual pathway. Proc. Natl. Acad. Sci. USA 112,E2957-E2966. doi:10.1073/pnas.1506458112

Ferent, J., Giguere, F., Jolicoeur, C., Morin, S., Michaud, J.-F., Makihara, S.,Yam, P. T., Cayouette, M. and Charron, F. (2019). Boc Acts via Numb as a Shh-dependent endocytic platform for Ptch1 internalization and Shh-mediated axonguidance. Neuron 102, 1157-1171.e5. doi:10.1016/j.neuron.2019.04.003

Fernandez, D. C., Fogerson, P. M., Lazzerini Ospri, L., Thomsen, M. B., Layne,R. M., Severin, D., Zhan, J., Singer, J. H., Kirkwood, A., Zhao, H. et al. (2018).Light affects mood and learning through distinct retina-brain pathways. Cell 175,71-84.e18. doi:10.1016/j.cell.2018.08.004

Fernandez, D. C., Komal, R., Langel, J., Ma, J., Duy, P. Q., Penzo, M. A., Zhao, H.and Hattar, S. (2020). Retinal innervation tunes circuits that drive nonphoticentrainment to food. Nature 581, 194-198. doi:10.1038/s41586-020-2204-1

Firth, S. I., Wang, C.-T. and Feller, M. B. (2005). Retinal waves: mechanisms andfunction in visual system development. Cell Calcium 37, 425-432. doi:10.1016/j.ceca.2005.01.010

Fruttiger, M. (2007). Development of the retinal vasculature. Angiogenesis 10,77-88. doi:10.1007/s10456-007-9065-1

Fruttiger, M., Calver, A. R., Kruger, W. H., Mudhar, H. S., Michalovich, D.,Takakura, N., Nishikawa, S. I. and Richardson, W. D. (1996). PDGFmediates aneuron-astrocyte interaction in the developing retina. Neuron 17, 1117-1131.doi:10.1016/S0896-6273(00)80244-5

Fu, Y., Liao, H.-W., Do, M. T. H. and Yau, K.-W. (2005). Non-image-forming ocularphotoreception in vertebrates. Curr. Opin. Neurobiol. 15, 415-422. doi:10.1016/j.conb.2005.06.011

Gan, L., Xiang, M., Zhou, L., Wagner, D. S., Klein, W. H. and Nathans, J. (1996).POU domain factor Brn-3b is required for the development of a large set of retinalganglion cells. Proc. Natl. Acad. Sci. USA 93, 3920-3925. doi:10.1073/pnas.93.9.3920

Ghiasvand, N. M., Rudolph, D. D., Mashayekhi, M., Brzezinski, J. A., IV,Goldman, D. and Glaser, T. (2011). Deletion of a remote enhancer near ATOH7disrupts retinal neurogenesis, causing NCRNA disease. Nature Neurosci. 14,578-586. doi:10.1038/nn.2798

Gnanaguru, G., Bachay, G., Biswas, S., Pinzon-Duarte, G., Hunter, D. D. andBrunken, W. J. (2013). Laminins containing the β2 and γ3 chains regulateastrocyte migration and angiogenesis in the retina.Development 140, 2050-2060.doi:10.1242/dev.087817

Goldberg, J. L., Espinosa, J. S., Xu, Y., Davidson, N., Kovacs, G. T. A. andBarres, B. A. (2002). Retinal ganglion cells do not extend axons by default.Neuron 33, 689-702. doi:10.1016/S0896-6273(02)00602-5

Golding, B., Pouchelon, G., Bellone, C., Murthy, S., Di Nardo, A. A., Govindan,S., Ogawa, M., Shimogori, T., Luscher, C., Dayer, A. et al. (2014). Retinal inputdirects the recruitment of inhibitory interneurons into thalamic visual circuits.Neuron 81, 1057-1069. doi:10.1016/j.neuron.2014.01.032

Gomes, F. L. A. F., Zhang, G., Carbonell, F., Correa, J. A., Harris, W. A., Simons,B. D. and Cayouette, M. (2011). Reconstruction of rat retinal progenitor celllineages in vitro reveals a surprising degree of stochasticity in cell fate decisions.Development 138, 227-235. doi:10.1242/dev.059683

Guido, W. (2018). Development, form, and function of the mouse visual thalamus.J. Neurophysiol. 120, 211-225. doi:10.1152/jn.00651.2017

Guler, A. D., Ecker, J. L., Lall, G. S., Haq, S., Altimus, C. M., Liao, H.-W., Barnard,A. R., Cahill, H., Badea, T. C., Zhao, H. et al. (2008). Melanopsin cells are theprincipal conduits for rod-cone input to non-image-forming vision. Nature 453,102-105. doi:10.1038/nature06829

Han, Y. and Massey, S. C. (2005). Electrical synapses in retinal ON cone bipolarcells: subtype-specific expression of connexins. Proc. Natl. Acad. Sci. USA 102,13313-13318. doi:10.1073/pnas.0505067102

Hansen, K. A., Torborg, C. L., Elstrott, J. and Feller, M. B. (2005). Expression andfunction of the neuronal gap junction protein connexin 36 in developingmammalian retina. J. Comp. Neurol. 493, 309-320. doi:10.1002/cne.20759

Hashimoto, T., Zhang, X.-M., Chen, B. Y. and Yang, X.-J. (2006). VEGF activatesdivergent intracellular signaling components to regulate retinal progenitor cellproliferation and neuronal differentiation. Development 133, 2201-2210. doi:10.1242/dev.02385

Hastings, M. H., Maywood, E. S. and Brancaccio, M. (2018). Generation ofcircadian rhythms in the suprachiasmatic nucleus. Nature Rev. Neurosci. 19,453-469. doi:10.1038/s41583-018-0026-z

Hatori, M., Le, H., Vollmers, C., Keding, S. R., Tanaka, N., Schmedt, C., Jegla, T.and Panda, S. (2008). Inducible ablation of melanopsin-expressing retinalganglion cells reveals their central role in non-image forming visual responses.PLoS ONE 3, e2451. doi:10.1371/journal.pone.0002451

Hattar, S., Liao, H.-W., Takao, M., Berson, D. M. and Yau, K.-W. (2002).Melanopsin-containing retinal ganglion cells: architecture, projections, andintrinsic photosensitivity. Science 295, 1065-1070. doi:10.1126/science.1069609

Hattar, S., Kumar, M., Park, A., Tong, P., Tung, J., Yau, K.-W. and Berson, D. M.(2006). Central projections of melanopsin-expressing retinal ganglion cells in themouse. J. Comp. Neurol. 497, 326-349. doi:10.1002/cne.20970

He, J., Zhang, G., Almeida, A. D., Cayouette, M., Simons, B. D. and Harris, W. A.(2012). How variable clones build an invariant retina. Neuron 75, 786-798. doi:10.1016/j.neuron.2012.06.033

Herzog, E. D., Hermanstyne, T., Smyllie, N. J. and Hastings, M. H. (2017).Regulating the suprachiasmatic nucleus (scn) circadian clockwork: interplaybetween cell-autonomous and circuit-level mechanisms. Cold Spring Harb.Perspect. Biol. 9, a027706. doi:10.1101/cshperspect.a027706

Huang, Z. J. and Zeng, H. (2013). Genetic approaches to neural circuits in themouse. Annu. Rev. Neurosci. 36, 183-215. doi:10.1146/annurev-neuro-062012-170307

Huang, L., Xi, Y., Peng, Y., Yang, Y., Huang, X., Fu, Y., Tao, Q., Xiao, J., Yuan, T.,An, K. et al. (2019). A visual circuit related to habenula underlies theantidepressive effects of light therapy. Neuron 102, 128-142.e8. doi:10.1016/j.neuron.2019.01.037

Huberman, A. D., Feller, M. B. and Chapman, B. (2008). Mechanisms underlyingdevelopment of visual maps and receptive fields. Annu. Rev. Neurosci. 31,479-509. doi:10.1146/annurev.neuro.31.060407.125533

Humphrey, M. F. and Beazley, L. D. (1985). Retinal ganglion cell death during opticnerve regeneration in the frog Hylamoorei. J. Comp. Neurol. 236, 382-402. doi:10.1002/cne.902360307

Ito, S. and Feldheim, D. A. (2018). The mouse superior colliculus: an emergingmodel for studying circuit formation and function. Front. Neural Circuits 12, 10.doi:10.3389/fncir.2018.00010

Jadhav, A. P., Cho, S.-H. and Cepko, C. L. (2006). Notch activity permits retinalcells to progress through multiple progenitor states and acquire a stem cellproperty. Proc. Natl Acad. Sci. USA 103, 18998-19003. doi:10.1073/pnas.0608155103

Janik, D. and Mrosovsky, N. (1994). Intergeniculate leaflet lesions andbehaviorally-induced shifts of circadian rhythms. Brain Res. 651, 174-182.doi:10.1016/0006-8993(94)90695-5

Jensen, A. M. and Wallace, V. A. (1997). Expression of Sonic hedgehog and itsputative role as a precursor cell mitogen in the developing mouse retina.Development 124, 363-371.

Jo, A., Xu, J., Deniz, S., Cherian, S., DeVries, S. H. and Zhu, Y. (2018).Intersectional strategies for targeting amacrine and ganglion cell types in themouse retina. Front. Neural Circuits 12, 66. doi:10.3389/fncir.2018.00066

Johnson, J., Wu, V., Donovan, M., Majumdar, S., Renterıa, R. C., Porco, T., VanGelder, R. N. and Copenhagen, D. R. (2010). Melanopsin-dependent lightavoidance in neonatal mice. Proc. Natl Acad. Sci. USA 107, 17374-17378. doi:10.1073/pnas.1008533107

Joo, H. R., Peterson, B. B., Dacey, D. M., Hattar, S. and Chen, S. K. (2013).Recurrent axon collaterals of intrinsically photosensitive retinal ganglion cells. Vis.Neurosci. 30, 175-182. doi:10.1017/S0952523813000199

Kay, J. N., Finger-Baier, K. C., Roeser, T., Staub,W. and Baier, H. (2001). Retinalganglion cell genesis requires lakritz, a Zebrafish atonal Homolog. Neuron 30,725-736. doi:10.1016/S0896-6273(01)00312-9

Kay, J. N., Roeser, T., Mumm, J. S., Godinho, L., Mrejeru, A., Wong, R. O. L. andBaier, H. (2004). Transient requirement for ganglion cells during assembly ofretinal synaptic layers. Development 131, 1331-1342. doi:10.1242/dev.01040

Kay, J. N., De la Huerta, I., Kim, I.-J., Zhang, Y., Yamagata, M., Chu, M. W.,Meister, M. and Sanes, J. R. (2011). Retinal ganglion cells with distinctdirectional preferences differ in molecular identity, structure, and centralprojections. J. Neurosci. 31, 7753-7762. doi:10.1523/JNEUROSCI.0907-11.2011

Kerschensteiner, D., Morgan, J. L., Parker, E. D., Lewis, R. M. and Wong,R. O. L. (2009). Neurotransmission selectively regulates synapse formation inparallel circuits in vivo. Nature 460, 1016-1020. doi:10.1038/nature08236

Kim, J., Wu, H.-H., Lander, A. D., Lyons, K. M., Matzuk, M. M. and Calof, A. L.(2005). GDF11 controls the timing of progenitor cell competence in developingretina. Science 308, 1927-1930. doi:10.1126/science.1110175

Kirkby, L. A. and Feller, M. B. (2013). Intrinsically photosensitive ganglion cellscontribute to plasticity in retinal wave circuits. Proc. Natl Acad. Sci. USA 110,12090-12095. doi:10.1073/pnas.1222150110

Kofuji, P., Mure, L. S., Massman, L. J., Purrier, N., Panda, S. and Engeland,W. C. (2016). Intrinsically Photosensitive Retinal Ganglion Cells (ipRGCs) arenecessary for light entrainment of peripheral clocks. PLoS ONE 11, e0168651.doi:10.1371/journal.pone.0168651

Kondo, H., Matsushita, I., Tahira, T., Uchio, E. and Kusaka, S. (2016). Mutationsin ATOH7 gene in patients with nonsyndromic congenital retinal nonattachmentand familial exudative vitreoretinopathy. Ophthalmic Genet. 37, 462-464. doi:10.3109/13816810.2015.1120316

Lee, S.-C., Cruikshank, S. J. and Connors, B. W. (2010). Electrical and chemicalsynapses between relay neurons in developing thalamus. J. Physiol. 588,2403-2415. doi:10.1113/jphysiol.2010.187096

11

REVIEW Development (2020) 147, dev196535. doi:10.1242/dev.196535

DEVELO

PM

ENT

Li, J. Y. and Schmidt, T. M. (2018). Divergent projection patterns of M1 ipRGCsubtypes. J. Comp. Neurol. 526, 2010-2018. doi:10.1002/cne.24469

Liu, W., Mo, Z. and Xiang, M. (2001). The Ath5 proneural genes function upstreamof Brn3 POU domain transcription factor genes to promote retinal ganglion celldevelopment. Proc. Natl. Acad. Sci USA 98, 1649-1654. doi:10.1073/pnas.98.4.1649

Madisen, L., Garner, A. R., Shimaoka, D., Chuong, A. S., Klapoetke, N. C., Li, L.,van der Bourg, A., Niino, Y., Egolf, L., Monetti, C. et al. (2015). Transgenic micefor intersectional targeting of neural sensors and effectors with high specificity andperformance. Neuron 85, 942-958. doi:10.1016/j.neuron.2015.02.022

Martersteck, E. M., Hirokawa, K. E., Evarts, M., Bernard, A., Duan, X., Li, Y., Ng,L., Oh, S. W., Ouellette, B., Royall, J. J. et al. (2017). Diverse central projectionpatterns of retinal ganglion cells. Cell Rep. 18, 2058-2072. doi:10.1016/j.celrep.2017.01.075

Masland, R. H. (2012). The neuronal organization of the retina.Neuron 76, 266-280.doi:10.1016/j.neuron.2012.10.002

Maxeiner, S., Kruger, O., Schilling, K., Traub, O., Urschel, S. and Willecke, K.(2003). Spatiotemporal transcription of connexin45 during brain developmentresults in neuronal expression in adult mice. Neuroscience 119, 689-700. doi:10.1016/S0306-4522(03)00077-0

McLaughlin, T., Torborg, C. L., Feller, M. B. and O’Leary, D. D. M. (2003).Retinotopic map refinement requires spontaneous retinal waves during a briefcritical period of development. Neuron 40, 1147-1160. doi:10.1016/S0896-6273(03)00790-6

McLoon, S. C. and Lund, R. D. (1982). Transient retinofugal pathways in thedeveloping chick. Exp. Brain Res. 45, 277-284. doi:10.1007/BF00235788

McNeill, D. S., Sheely, C. J., Ecker, J. L., Badea, T. C., Morhardt, D., Guido, W.and Hattar, S. (2011). Development of melanopsin-based irradiance detectingcircuitry. Neural Dev. 6, 8. doi:10.1186/1749-8104-6-8

Mills, S. L., O’Brien, J. J., Li, W., O’Brien, J. and Massey, S. C. (2001). Rodpathways in the mammalian retina use connexin 36. J. Comp. Neurol. 436,336-350. doi:10.1002/cne.1071

Missaire, M. and Hindges, R. (2015). The role of cell adhesion molecules in visualcircuit formation: from neurite outgrowth to maps and synaptic specificity. Dev.Neurobiol. 75, 569-583. doi:10.1002/dneu.22267

Morgan, J. L., Schubert, T. and Wong, R. O. L. (2008). Developmental patterningof glutamatergic synapses onto retinal ganglion cells. Neural Dev. 3, 8. doi:10.1186/1749-8104-3-8

Morin, L. P. and Studholme, K. M. (2014). Retinofugal projections in the mouse.J.Comp.Neurol. 522, 3733-3753. doi:10.1002/cne.23635

Moshiri, A., Gonzalez, E., Tagawa, K., Maeda, H., Wang, M., Frishman, L. J. andWang, S. W. (2008). Near complete loss of retinal ganglion cells in the math5/brn3b double knockout elicits severe reductions of other cell types during retinaldevelopment. Dev. Biol. 316, 214-227. doi:10.1016/j.ydbio.2008.01.015

Mu, X., Fu, X., Sun, H., Liang, S., Maeda, H., Frishman, L. J. and Klein, W. H.(2005). Ganglion cells are required for normal progenitor- cell proliferation but notcell-fate determination or patterning in the developingmouse retina.Curr. Biol. 15,525-530. doi:10.1016/j.cub.2005.01.043

Muller, M. and Hollander, H. (1988). A small population of retinal ganglion cellsprojecting to the retina of the other eye. Exp. Brain Res. 71, 611-617. doi:10.1007/BF00248754

Muller, L. P. D. S., Dedek, K., Janssen-Bienhold, U., Meyer, A., Kreuzberg,M. M., Lorenz, S., Willecke, K. and Weiler, R. (2010). Expression andmodulation of connexin 30.2, a novel gap junction protein in the mouse retina.Vis. Neurosci. 27, 91-101. doi:10.1017/S0952523810000131

Murcia-Belmonte, V., Coca, Y., Vegar, C., Negueruela, S., de Juan Romero, C.,Valin o, A. J., Sala, S., DaSilva, R., Kania, A., Borrell, V. et al. (2019). A retino-retinal projection guided by unc5c emerged in species with retinal waves. Curr.Biol. 29, 1149-1160.e4. doi:10.1016/j.cub.2019.02.052

Nadal-Nicolas, F. M., Valiente-Soriano, F. J., Salinas-Navarro, M., Jimenez-Lopez, M., Vidal-Sanz, M. and Agudo-Barriuso, M. (2015). Retino-retinalprojection in juvenile and young adult rats and mice. Exp. Eye Res. 134, 47-52.doi:10.1016/j.exer.2015.03.015

Neumann, C. J. and Nuesslein-Volhard, C. (2000). Patterning of the zebrafishretina by a wave of sonic hedgehog activity. Science 289, 2137-2139. doi:10.1126/science.289.5487.2137

Newkirk, G. S., Hoon, M., Wong, R. O. and Detwiler, P. B. (2013). Inhibitory inputstune the light response properties of dopaminergic amacrine cells in mouse retina.J. Neurophysiol. 110, 536-552. doi:10.1152/jn.00118.2013

Nguyen, M.-T. T., Vemaraju, S., Nayak, G., Odaka, Y., Buhr, E. D., Alonzo, N.,Tran, U., Batie, M., Upton, B. A., Darvas, M. et al. (2019). An opsin 5-dopaminepathway mediates light-dependent vascular development in the eye. Nat. CellBiol. 21, 420-429. doi:10.1038/s41556-019-0301-x

Osterhout, J. A., Stafford, B. K., Nguyen, P. L., Yoshihara, Y. and Huberman,A. D. (2015). Contactin-4 mediates axon-target specificity and functionaldevelopment of the accessory optic system. Neuron 86, 985-999. doi:10.1016/j.neuron.2015.04.005

O’Sullivan, M. L., Pun al, V. M., Kerstein, P. C., Brzezinski, J. A.,IV, Glaser, T.,Wright, K. M. and Kay, J. N. (2017). Astrocytes follow ganglion cell axons to

establish an angiogenic template during retinal development.Glia 65, 1697-1716.doi:10.1002/glia.23189

Peng, J., Fabre, P. J., Dolique, T., Swikert, S. M., Kermasson, L., Shimogori, T.and Charron, F. (2018). Sonic hedgehog is a remotely produced cue that controlsaxon guidance trans-axonally at a midline choice point. Neuron 97, 326-340.e4.doi:10.1016/j.neuron.2017.12.028

Prasov, L., Masud, T., Khaliq, S., Mehdi, S. Q., Abid, A., Oliver, E. R., Silva, E. D.,Lewanda, A., Brodsky, M. C., Borchert, M. et al. (2012). ATOH7 mutationscause autosomal recessive persistent hyperplasia of the primary vitreous. Hum.Mol. Genet. 21, 3681-3694. doi:10.1093/hmg/dds197

Prigge, C. L., Yeh, P.-T., Liou, N.-F., Lee, C.-C., You, S.-F., Liu, L.-L., McNeill,D. S., Chew, K. S., Hattar, S., Chen, S.-K. et al. (2016). M1 ipRGCs influencevisual function through retrograde signaling in the retina. J. Neurosci. 36,7184-7197. doi:10.1523/JNEUROSCI.3500-15.2016

Provencio, I., Rollag, M. D. and Castrucci, A. M. (2002). Photoreceptive net in themammalian retina. This mesh of cells may explain how some blind mice can stilltell day from night. Nature 415, 493. doi:10.1038/415493a

Rao, S., Chun, C., Fan, J., Kofron, J. M., Yang, M. B., Hegde, R. S., Ferrara, N.,Copenhagen, D. R. and Lang, R. A. (2013). A direct and melanopsin-dependentfetal light response regulates mouse eye development. Nature 494, 243-246.doi:10.1038/nature11823

Rasband, K., Hardy, M. and Chien, C.-B. (2003). Generating X. Neuron 39,885-888. doi:10.1016/S0896-6273(03)00563-4

Renna, J. M., Weng, S. and Berson, D. M. (2011). Light acts throughmelanopsin toalter retinal waves and segregation of retinogeniculate afferents. Nat. Neurosci.14, 827-829. doi:10.1038/nn.2845

Renna, J. M., Chellappa, D. K., Ross, C. L., Stabio, M. E. and Berson, D. M.(2015). Melanopsin ganglion cells extend dendrites into the outer retina duringearly postnatal development. Dev. Neurobiol. 75, 935-946. doi:10.1002/dneu.22260

Rheaume, B. A., Jereen, A., Bolisetty, M., Sajid, M. S., Yang, Y., Renna, K., Sun,L., Robson, P. and Trakhtenberg, E. F. (2018). Single cell transcriptome profilingof retinal ganglion cells identifies cellular subtypes.Nat. Commun. 9, 2759. doi:10.1038/s41467-018-05134-3

Rupp, A. C., Ren,M., Altimus, C.M., Fernandez, D. C., Richardson, M., Turek, F.,Hattar, S. and Schmidt, T. M. (2019). Distinct ipRGC subpopulations mediatelight’s acute and circadian effects on body temperature and sleep. eLife 8,e44358. doi:10.7554/eLife.44358

Sanchez-Camacho, C. and Bovolenta, P. (2008). Autonomous and non-autonomous Shh signalling mediate the in vivo growth and guidance of mouseretinal ganglion cell axons. Development 135, 3531-3541. doi:10.1242/dev.023663

Sanes, J. R. and Zipursky, S. L. (2010). Design principles of insect and vertebratevisual systems. Neuron 66, 15-36. doi:10.1016/j.neuron.2010.01.018

Sanes, J. R. and Zipursky, S. L. (2020). Synaptic specificity, recognitionmolecules, and assembly of neural circuits. Cell 181, 536-556. doi:10.1016/j.cell.2020.04.008

Schmidt, T. M., Chen, S.-K. and Hattar, S. (2011). Intrinsically photosensitiveretinal ganglion cells: many subtypes, diverse functions. Trends Neurosci. 34,572-580. doi:10.1016/j.tins.2011.07.001

Seabrook, T. A., El-Danaf, R. N., Krahe, T. E., Fox, M. A. and Guido, W. (2013).Retinal input regulates the timing of corticogeniculate innervation. J. Neurosci. 33,10085-10097. doi:10.1523/JNEUROSCI.5271-12.2013

Seabrook, T. A., Burbridge, T. J., Crair, M. C. and Huberman, A. D. (2017).Architecture, function, and assembly of the mouse visual system. Annu. Rev.Neurosci. 40, 499-538. doi:10.1146/annurev-neuro-071714-033842

Sekaran, S., Lupi, D., Jones, S. L., Sheely, C. J., Hattar, S., Yau, K.-W., Lucas,R. J., Foster, R. G. and Hankins, M. W. (2005). Melanopsin-dependentphotoreception provides earliest light detection in the mammalian retina. Curr.Biol. 15, 1099-1107. doi:10.1016/j.cub.2005.05.053

Sokhadze, G., Seabrook, T. A. and Guido, W. (2018). The absence of retinal inputdisrupts the development of cholinergic brainstem projections in themouse dorsallateral geniculate nucleus. Neural Dev. 13, 27. doi:10.1186/s13064-018-0124-7

Sondereker, K. B., Onyak, J. R., Islam, S. W., Ross, C. L. and Renna, J. M.(2017). Melanopsin ganglion cell outer retinal dendrites: Morphologically distinctand asymmetrically distributed in the mouse retina. J. Comp. Neurol. 525,3653-3665. doi:10.1002/cne.24293

Sondereker, K. B., Stabio, M. E. and Renna, J. M. (2020). Crosstalk: The diversityof melanopsin ganglion cell types has begun to challenge the canonical dividebetween image-forming and non-image-forming vision. J. Comp. Neurol. 528,2044-2067. doi:10.1002/cne.24873

Stacy, R. C., Demas, J., Burgess, R. W., Sanes, J. R. and Wong, R. O. L. (2005).Disruption and recovery of patterned retinal activity in the absence ofacetylcholine. J. Neurosci. 25, 9347-9357. doi:10.1523/JNEUROSCI.1800-05.2005

Su, J., Charalambakis, N. E., Sabbagh, U., Somaiya, R. D., Monavarfeshani, A.,Guido, W. and Fox, M. A. (2020). Retinal inputs signal astrocytes to recruitinterneurons into visual thalamus. Proc. Natl. Acad. Sci. USA 117, 2671-2682.doi:10.1073/pnas.1913053117

12

REVIEW Development (2020) 147, dev196535. doi:10.1242/dev.196535

DEVELO

PM

ENT

Sun, Y. and Smith, L. E. H. (2018). Retinal vasculature in development anddiseases. Annu. Rev. Vis. Sci. 4, 101-122. doi:10.1146/annurev-vision-091517-034018

Sun, L. O., Brady, C. M., Cahill, H., Al-Khindi, T., Sakuta, H., Dhande, O. S.,Noda, M., Huberman, A. D., Nathans, J. and Kolodkin, A. L. (2015). Functionalassembly of accessory optic system circuitry critical for compensatory eyemovements. Neuron 86, 971-984. doi:10.1016/j.neuron.2015.03.064

Syed, M. M., Lee, S., Zheng, J. and Zhou, Z. J. (2004). Stage-dependent dynamicsand modulation of spontaneous waves in the developing rabbit retina. J. Physiol.560, 533-549. doi:10.1113/jphysiol.2004.066597

Tang, X., Tzekov, R. and Passaglia, C. L. (2016). Retinal cross talk in themammalian visual system. J. Neurophysiol. 115, 3018-3029. doi:10.1152/jn.01137.2015

Tao, C. and Zhang, X. (2016). Retinal proteoglycans act as cellular receptors forbasement membrane assembly to control astrocyte migration and angiogenesis.Cell Rep. 17, 1832-1844. doi:10.1016/j.celrep.2016.10.035

Tennant, M., Bruce, S. R. and Beazley, L. D. (1993). Survival of ganglion cellswhich form the retino-retinal projection during optic nerve regeneration in the frog.Vis. Neurosci. 10, 681-686. doi:10.1017/S095252380000537X

Thanos, S. (1999). Genesis, neurotrophin responsiveness, and apoptosis of apronounced direct connection between the two eyes of the chick embryo: a naturalerror or a meaningful developmental event? J. Neurosci. 19, 3900-3917. doi:10.1523/JNEUROSCI.19-10-03900.1999

Toth, P. and Strznicky, C. (1989). Retino-retinal projections in three anuranspecies. Neurosci. Lett. 104, 43-47. doi:10.1016/0304-3940(89)90326-1

Tran, N. M., Shekhar, K., Whitney, I. E., Jacobi, A., Benhar, I., Hong, G., Yan, W.,Adiconis, X., Arnold, M. E., Lee, J. M. et al. (2019). Single-cell profiles of retinalganglion cells differing in resilience to injury reveal neuroprotective genes.Neuron104, 1039-1055.e12. doi:10.1016/j.neuron.2019.11.006

Trimarchi, J. M., Stadler, M. B. and Cepko, C. L. (2008). Individual retinalprogenitor cells display extensive heterogeneity of gene expression.PLoSONE 3,e1588. doi:10.1371/journal.pone.0001588

Tufford, A. R., Onyak, J. R., Sondereker, K. B., Lucas, J. A., Earley, A. M., Mattar,P., Hattar, S., Schmidt, T. M., Renna, J. M. and Cayouette, M. (2018).Melanopsin retinal ganglion cells regulate cone photoreceptor lamination in themouse retina. Cell Rep. 23, 2416-2428. doi:10.1016/j.celrep.2018.04.086

Turner, D. L., Snyder, E. Y. and Cepko, C. L. (1990). Lineage-independentdetermination of cell type in the embryonic mouse retina. Neuron 4, 833-845.doi:10.1016/0896-6273(90)90136-4

Usui, Y., Westenskow, P. D., Kurihara, T., Aguilar, E., Sakimoto, S., Paris, L. P.,Wittgrove, C., Feitelberg, D., Friedlander, M. S. H., Moreno, S. K. et al. (2015).Neurovascular crosstalk between interneurons and capillaries is required forvision. J. Clin. Invest. 125, 2335-2346. doi:10.1172/JCI80297

Volgyi, B., Chheda, S. and Bloomfield, S. A. (2009). Tracer coupling patterns ofthe ganglion cell subtypes in the mouse retina. J. Comp. Neurol. 512, 664-687.doi:10.1002/cne.21912

Waid, D. K. and McLoon, S. C. (1998). Ganglion cells influence the fate of dividingretinal cells in culture. Development 125, 1059-1066.

Wallace, V. A. and Raff, M. C. (1999). A role for Sonic hedgehog in axon-to-astrocyte signalling in the rodent optic nerve. Development 126, 2901-2909.

Wang, Y. P., Dakubo, G., Howley, P., Campsall, K. D., Mazarolle, C. J., Shiga,S. A., Lewis, P. M., McMahon, A. P. and Wallace, V. A. (2002). Development of

normal retinal organization depends on Sonic hedgehog signaling from ganglioncells. Nat. Neurosci. 5, 831-832. doi:10.1038/nn911

Wang, Y., Dakubo, G. D., Thurig, S., Mazerolle, C. J. and Wallace, V. A. (2005).Retinal ganglion cell-derived sonic hedgehog locally controls proliferation and thetiming of RGC development in the embryonic mouse retina. Development 132,5103-5113. doi:10.1242/dev.02096

Watanabe, T. and Raff, M. C. (1990). Rod photoreceptor development in vitro:intrinsic properties of proliferating neuroepithelial cells change as developmentproceeds in the rat retina. Neuron 4, 461-467. doi:10.1016/0896-6273(90)90058-N

Watanabe, T. and Raff, M. C. (1992). Diffusible rod-promoting signals in thedeveloping rat retina. Development 114, 899-906.

Weiner, G. A., Shah, S. H., Angelopoulos, C. M., Bartakova, A. B., Pulido, R. S.,Murphy, A., Nudleman, E., Daneman, R. and Goldberg, J. L. (2019).Cholinergic neural activity directs retinal layer-specific angiogenesis and bloodretinal barrier formation. Nat. Commun. 10, 2477. doi:10.1038/s41467-019-10219-8

Wilson, A. M. and Di Polo, A. (2012). Gene therapy for retinal ganglion cellneuroprotection in glaucoma. Gene Ther. 19, 127-136. doi:10.1038/gt.2011.142

Wong, R. O. L. (1999). Retinal waves and visual system development. Annu. Rev.Neurosci. 22, 29-47. doi:10.1146/annurev.neuro.22.1.29

Wu, F., Kaczynski, T. J., Sethuramanujam, S., Li, R., Jain, V., Slaughter, M. andMu, X. (2015). Two transcription factors, Pou4f2 and Isl1, are sufficient to specifythe retinal ganglion cell fate. Proc. Natl. Acad. Sci. 112, E1559-E1568. doi:10.1073/pnas.1421535112

Yamagata, M. and Sanes, J. R. (2008). Dscam and Sidekick proteins direct lamina-specific synaptic connections in vertebrate retina. Nature 451, 465-469. doi:10.1038/nature06469

Yamagata, M.,Weiner, J. A. and Sanes, J. R. (2002). Sidekicks: synaptic adhesionmolecules that promote lamina-specific connectivity in the retina. Cell 110,649-660. doi:10.1016/S0092-8674(02)00910-8

Yamaguchi, T. P., Dumont, D. J., Conlon, R. A., Breitman, M. L. and Rossant, J.(1993). flk-1, an flt-related receptor tyrosine kinase is an early marker forendothelial cell precursors. Development 118, 489-498.

Yang, X. and Cepko, C. L. (1996). Flk-1, a receptor for vascular endothelial growthfactor (VEGF), is expressed by retinal progenitor cells. J. Neurosci. 16,6089-6099. doi:10.1523/JNEUROSCI.16-19-06089.1996

Young, R. W. (1985). Cell differentiation in the retina of the mouse. Anat. Rec. 212,199-205. doi:10.1002/ar.1092120215

Zhang, X. M. and Yang, X. J. (2001). Regulation of retinal ganglion cell productionby Sonic hedgehog. Development 128, 943-957.

Zhang, D.-Q., Wong, K. Y., Sollars, P. J., Berson, D. M., Pickard, G. E. andMcMahon, D. G. (2008). Intraretinal signaling by ganglion cell photoreceptors todopaminergic amacrine neurons. Proc. Natl. Acad. Sci. USA 105, 14181-14186.doi:10.1073/pnas.0803893105

Zhang, D.-Q., Belenky, M. A., Sollars, P. J., Pickard, G. E. and McMahon, D. G.(2012). Melanopsin mediates retrograde visual signaling in the retina. PLoS ONE7, e42647. doi:10.1371/journal.pone.0042647

Zlomuzica, A., Reichinnek, S., Maxeiner, S., Both, M., May, E., Worsdorfer, P.,Draguhn, A., Willecke, K. and Dere, E. (2010). Deletion of connexin45 in mouseneurons disrupts one-trial object recognition and alters kainate-induced γ-oscillations in the hippocampus. Physiol. Behav. 101, 245-253. doi:10.1016/j.physbeh.2010.05.007

13

REVIEW Development (2020) 147, dev196535. doi:10.1242/dev.196535

DEVELO

PM

ENT