Architecture and Activity-Mediated Refinement
of Axonal Projections from a Mosaic
of Genetically Identified Retinal Ganglion Cells
Andrew D. Huberman,1,* Mihai Manu,1Selina M. Koch,2Michael W. Susman,1Amanda Brosius Lutz,1Erik M. Ullian,2
Stephen A. Baccus,1,3and Ben A. Barres1,3
1Department of Neurobiology, Fairchild Science Building D235, Stanford University School of Medicine, Palo Alto, CA 94305, USA
2Department of Ophthalmology and Physiology, University of California, San Francisco, San Francisco, CA 94117, USA
3These authors contributed equally to this work
Our understanding of how mammalian sensory cir-
cuits are organized and develop has long been
hindered by the lack of genetic markers of neurons
with discrete functions. Here, we report a transgenic
mouse selectively expressing GFP in a complete
mosaic of transient OFF-a retinal ganglion cells
(tOFF-aRGCs). This enabled us to relate the mosaic
spacing, dendritic anatomy, and electrophysiology
of these RGCs to their complete map of projections
in the brain. We find that tOFF-aRGCs project exclu-
geniculate nucleus and are restricted to a specific
of tOFF-aRGC are also organized into columns in the
SC. Both laminar and columnar specificity develop
through axon refinement. Disruption of cholinergic
retinal waves prevents the emergence of columnar-
but not laminar-specific tOFF-aRGC connections.
Our findings reveal that in a genetically identified
sensory map, spontaneous activity promotes synap-
tic specificity by segregating axons arising from
RGCs of the same subtype.
The transmission of topographically organized sensory informa-
tion poses a substantial problem for central nervous system
(CNS) development. Sensory axons must travel long distances
to reach their targets yet still maintain their local relationships
guidance cues steer axons through a series of decision points
leading to their targets and then spontaneous neural activity
and guidance cues refine those projections into an orderly, con-
tinuous topographic map (McLaughlin and O’Leary, 2005; Luo
and Flanagan, 2007). Within this general model, however, sev-
eral aspects of the construction of mammalian sensory circuits
remain poorly understood. For instance, a hallmark feature of
sensory pathways is their parallel organization: different neurons
respond to and convey qualitatively distinct information about
a given sensory modality (Callaway, 2005). How axons carrying
distinct typesof sensoryinformation areorganizedwithincontin-
uous topographic maps and how those projections develop is
unclear. A major obstacle for studying this aspect of circuit
wiring—often called ‘‘synaptic specificity’’ (Benson et al., 2001)
is a lack of genetic markers that ensure stable, selective, and
complete labeling of functionally identified classes of neurons
in mammals. Several basic and important questions related to
development of sensory circuits in the mammalian CNS there-
fore remain unresolved. For example, it is unknown whether
emergence of synaptic specificity involves refinement of initially
inaccurate projections or whether functionally specific circuits
are accurately formed from the outset, as they are in the inverte-
brate nervous system (Meinertzhagen and Hanson, 1993). A
related question is whether neural activity contributes to the
emergence of synaptic specificity.
In principle, the connections associated with different types of
gence of synaptic specificity in the mammalian CNS. Conven-
physiologically distinct subtypes of RGCs (reviewed in Wa ¨ssle,
2004; Berson, 2008). Each RGC subtype receives qualitatively
of its laminar-specific dendritic connections with bipolar and
amacrine cells in the inner plexiform layer (IPL) (Roska and Wer-
blin, 2001) and in turn, conveys that information to the brain by
son, 2008). Despite extensive study of the physiology and con-
nectivity associated with different RGCs, the total number of
functionally distinct RGC subtypes and the complete wiring dia-
gram corresponding to any one individual RGC subtype still re-
mains unclear (Sun et al., 2002; Kong et al., 2005; Berson, 2008).
In order to study the organization and development of neural
circuitry corresponding to distinct RGC subtypes, we screened
a library of BAC transgenic mice with GFP expressed under
the control of different promoters (Gong et al., 2003). Our goal
saic of RGCs throughout the retina- because mosaic spacing
suggests that cell population is comprised of a functionally ho-
mogeneous RGC subtype (Wa ¨ssle, 2004) and (2) there are few
Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc. 425
or no cells expressing GFP in retinorecipient brain areas, so the
organization of axonal projections from the GFP-expressing
RGCs can be visualized during development and adulthood.
A Transgenic Mouse with a Complete Mosaic
of One RGC Subtype Expressing GFP
Calretinin-EGFP mice (CB2-GFP mice) met both criteria of our
screen. In the retinas of these mice, GFP is expressed by a pop-
Figure 1. A Mosaic of GFP+OFF-aRGCs in
(A) Whole-mount CB2-GFP retina, showing GFP+
RGCs. Along the relieving cuts, GFP+amacrine
cells are also evident (see [E]). D/V/N/T: dorsal,
ventral, nasal, temporal axes. Scale is 500 mm.
(B) GFP+RGCs (arrows). Inset shows a single
GFP+RGC (arrow) with its axon (arrowhead). (C
and D) Every GFP+RGC is immunopositive for
in (B) inset is 50 mm.
(E) Vertically-sectioned retina. Blue, DAPI staining
indicates cellular layers. GFP+cells are amacrine
cells in the inner nuclear layer (INL), and regularly
(F) Density profile of GFP+RGC cells as a function
of distance from other GFP+RGCs (Rodieck,
(G) Four neighboring GFP+RGCs filled with DiI to
reveal their entire dendritic arbors.
(H) Dendritic field boundaries of the same four
GFP+cells. (G and H) Scale is 75 mm.
(I) Retinal section stained for DAPI, GFP, and the
vesicle-acetylcholine transporter (VAChT) which
labels the middle of the ON and OFF sublaminae
in the IPL (horizontal lines denote boundaries of
IPL). GFP+RGC dendrites ramify within the OFF
sublaminae at the ?30%–35% depth of the IPL.
Arrow indicates the GFP+RGC’s axon. Arrow-
heads indicate axons from other GFP+RGCs
whose somata lie outside the field of view. Scale
is 75 mm.
ulation of regularly spaced cells through-
out the ganglion cell layer (GCL) (Figures
1A, 1B, and 1E; mean = 368 ± 14 GFP+
RGCs per retina; n = 6 mice). The pres-
ence of an axon emanating from each
GFP+cell confirmed they were RGCs
(Figure 1B, inset). In addition, as ex-
pected for RGCs, all the GFP+cells in
the RGC layer degenerated in response
to optic nerve transection (data not
shown). Every GFP+RGC stained with
the antibody SMI-32 (Figures 1B–1D),
which recognizes a neurofilament antigen
tion, the relatively large (?20–25 mm) and spherical GFP+somata
resembled mouse aRGCs (Pang et al., 2003; Schubert et al.,
2005; Volgyi et al., 2005). In CB2-GFP mice, GFP is also ex-
pressed by a population of amacrine cells in the inner nuclear
layer (INL) (Figures 1A and 1E). Based on their morphology and
calretinin expression (Lee et al., 2004) we assume those to be
AII amacrines, although they may include other amacrine types.
In vertical retinal sections (Figure 1E), the GFP+amacrines in the
INL and the regularly spaced GFP+RGCs (Figure 1E, arrows)
were both clearly visible. The somas of GFP+RGCs formed
Mapping a Mosaic of Genetically Identified RGCs
426 Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc.
a highly regular mosaic across the entire retina. This distribution
was not random, as cell bodies obeyed a minimum spacing rule,
avoiding each other to a distance of approximately 100 mm (Fig-
ure 1F). To visualize clearly the dendrites of single cells, we tar-
geted individual GFP+RGCs with electrodes and filled them
with DiI (n = 22 RGCs from 5 mice). The dendrites of neighboring
GFP+RGCs tiled the retina with the amount of overlap typical of
mouse OFF-aRGCs (Figure 1G and see Figure S1 available
online; Schubert et al., 2005; Volgyi et al., 2005). The DiI fills
also revealed that the GFP+RGCs possessed dendritic features
characteristic of mouse OFF-aRGCs (Figures 1G, 1H, and S1;
Pang et al., 2003; Schubert et al., 2005; Volgyi et al., 2005). We
observed that, in every case, the dendrites of the GFP+RGCs
were monostratified at the inner 30%–35% depth of the IPL
(Figure 1G), within the ‘‘OFF’’ sublaminae (Famiglietti and Kolb,
1976). When the retina was sectioned in the vertical plane and
stained for the vesicle acetylcholine transporter (VAChT) (which
delineates ON and OFF sublaminae in the IPL), it was apparent
that every GFP+RGC projected its dendrites exclusively into
the OFF sublaminae (Figure 1I; >600 GFP+RGCs examined
from 12 CB2-GFP mice). Collectively, these observations indi-
cate that a mosaic of OFF-aRGCs selectively express GFP in
this mouse line.
We next recorded intracellularly from the GFP+RGCs by visu-
microscope. In response to a flashing visual stimulus, these cells
showed transient OFF-type responses, depolarizing and spiking
briefly at the offset of a flash (Figure 2A). Visual responses of
different GFP+RGCs were very homogeneous, in that the aver-
age of each cell’s flash response was highly correlated with the
population average (correlation coefficient r = 0.93 ± 0.02, n = 7;
Figure 2B). Further characterization of these cells using a ran-
domly flickering visual stimulus confirmed the transient OFF-
type nature of these cells, which showed a biphasic temporal
response that peaked at 105–121 ms relative to the visual stimu-
lus (n = 3; Figures 2C and 2D). Taken with our anatomical obser-
vations, we conclude that a retinotopic mosaic of transient OFF-
aRGCs (tOFF-aRGCs) expresses GFP in this mouse line.
A Highly Organized Map of the tOFF-aRGC
Mosaic in the Brain
To determine the connections made by tOFF-aRGCs in the brain
we injected cholera toxin beta conjugated to Alexa 594 (CTb-
594; red label in Figures 3–8), which labels the axons of all
RGCs, into both eyes of CB2-GFP mice and then examined
the brain for axons that were both CTb-594+and GFP+. Of the
approximately two dozen brain regions targeted by RGC axons,
tOFF-aRGCsproject toonlytwoofthese:(1)thesuperior collicu-
lus (SC), a midbrain structure involved in sensory-motor integra-
tion and (2) the dorsal lateral geniculate nucleus (dLGN), a fore-
The mouse SC receives projections from many different types
entire contralateral retina. The dorsal-ventral axis of the retina is
mapped along the lateral-medial axis of the SC and the tempo-
ral-nasal retinal axis is mapped along the rostral-caudal axis of
the SC (Dra ¨ger and Hubel, 1976; May, 2005). When the SC is
viewed in the coronal plane (Figure 3A) the axons of RGCs can
be seen entering the SC deep to the pial surface, in the stratum
opticum (SO), and their dense terminal arborizations can be
observed in the overlying stratum griseum superficialis (SGS)
(Figure 3A). As the main retinorecipient division of the SC, the
SGS is divided into lower (lSGS) and upper (uSGS) lamina based
on the types of afferent input and postsynaptic cells residing
along its depth. A limited number of RGC axons also project to
the stratum zonale (SZ), a thin layer located just below the pial
margin (May, 2005). We observed that the GFP+axons of tOFF-
aRGCs terminate in a highly specific and stereotyped manner
within the SC. Their axons entered the SC through the SO, along
with all other RGC axons, and terminated in the overlying SGS.
Within the SGS, however, the GFP+arbors of tOFF-aRGCs
were faithfully restricted to the lSGS (Figures 3B and 3C and Fig-
ures 3B0and 3C0), avoiding the uSGS and SZ entirely. Viewed in
the sagittal plane, laminar-specific targeting of tOFF-aRGC pro-
GFP+tOFF-aRGC projects to the SC (data not shown). Taken
the full lateral-medial (Figures 3B and 3C) and rostral-caudal
(Figure 3D) extent of the SC, we can conclude that every tOFF-
aRGC obeys strict laminar specificity of its axonal projection to
this target, regardless of its retinotopic position.
Figure 2. Visual Responses of GFP+tOFF-aRGCs
(A) Membrane potential response of a GFP+tOFF-aRGC in response to a 1 s
periodic flash. Action potentials are truncated.
(B) Average subthreshold flash response of seven cells (gray trace), compared
to the average flash response of two example cells (colored traces). Spikes
were digitally removed prior to averaging.
(C and D) Spatiotemporal receptive field of a GFP+tOFF-aRGC computed
from the membrane potential response of the cell to a randomly flickering
checkerboard. (C) Temporal filter averaged across the spatial receptive field
center. (D) Two spatial slices of the receptive field taken at the times indicated
by the colored lines in (C), indicating the biphasic nature of the receptive field
Mapping a Mosaic of Genetically Identified RGCs
Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc. 427
Another salient feature of the tOFF-aRGC projections to the
SC was that, within the lSGS, their arbors were regularly spaced
to form patches of innervated and noninnervated territory (Fig-
ures 3A–3D). Based on their regular increased density of projec-
tions within the topographic map, we refer to these structures as
columns. Retinocollicular projections are known to be organized
into columns of alternating right-eye and left-eye RGC input in
cats (May, 2005). The GFP+columns observed in the SC of
CB2-GFP mice do not relate to eye specificity, however, be-
cause all the GFP+axons in the SC originate from the contralat-
eral retina (Figure S2).
We counted ?90 GFP+columns per SC hemisphere (88–96
columns; n = 6 mice). Given that an entire mosaic of ?370
tOFF-aRGCs expresses GFP and projects to the contralateral
SC, we could not determine how many tOFF-aRGCs project to
each column. In the caudal SC, however, RGC axons were
Figure 3. Axonal Projections of GFP+tOFF-aRGCs
(Left margin) Schematics of RGC projections in the mouse.
(Top left) Dorsal view. The retinas and optic pathways are in
red. Superior colliculus (SC), dorsal lateral geniculate nucleus
(dLGN),and olivary pretectal nucleus (OPN). Here and in (A–G)
R: rostral, C: caudal, L: lateral, M: medial. Schematics of cor-
onal and sagittal view of the brain at the level of the SC (middle
panels) and dLGN (bottom panel).
(A–D) Laminar- and columnar-specific tOFF-aRGC axonal
projections to the SC. (A) CTb-594 labeling (red) of all RGC
axons. CTb-594 weakly labels RGC axon shafts but densely
labels RGC axon terminals and thereby completely labels all
retinorecipient nuclei. SO, stratum opticum; lSGS, lower stra-
tum griseum superficialis; uSGS, upper stratum griseum
superficialis; SZ, stratum zonale. (A0) High-magnification
view of box in (A). (B) GFP+tOFF-aRGC axons in the SC.
The axons enter the SC through the SO, and selectively arbor-
ize in the lSGS. The GFP+arbors are also organized into col-
umns. Small arrow indicates a GFP+cell in the deeper SC;
such cells do not contribute to the GFP+axons in the retinor-
ecipient SC (see Figure S2). (B0) High-magnification view of
box in (B). (C) Merged view of (A) and (B). (C0) High-magnifica-
tion view of box in (C). (A–C) Scale is 250 mm. Coronal plane is
shown; D: dorsal, L: lateral. Scale in (A0)–(C0) is 125 mm. (D)
Sagittal view of CTb-594+and GFP+tOFF-aRGC axons in
the SC. Scale is 250 mm.
(E–G) Retino-dLGN projections viewed in the coronal plane.
(E) CTb-594 labeled RGC axons fill the entire dLGN, IGL,
the vLGN and IGL, turn medially into the dLGN, and project
through the lateral third of the dLGN (asterisks) to selectively
arborize in the inner dLGN, forming a ‘‘layer.’’ (G) Merged
view of (E) and (F). Scale is 250 mm.
sparse enough that we could observe cases where
individual GFP+axons bifurcated to terminate in
two neighboring, separate columns (Figure S3).
Because the degree of retinocollicular conver-
gence varies according to position within the reti-
notopic map (Dra ¨ger and Hubel, 1976; Berson,
1988), it is unclear if such bifurcation is a general
rule for all tOFF-aRGCs or is limited to the caudal
SC. We observed regularly spaced GFP+columns
across all axes of the SC (Figures 3A–3D), indicat-
ing that there is a highly organized columnar map corresponding
to the mosaic of tOFF-aRGCs within this target.
Unlike the dLGN of carnivores and primates (Callaway, 2005),
the rodent dLGN lacks overt laminar cytoarchitecture corre-
sponding to eye-specific inputs or to functionally specialized
visual processing streams (Reese, 1988). We were therefore
jections to the dLGN. GFP+axons projected through the lateral
third of the dLGN to selectively terminate within the inner/medial
portion of the nucleus, where they form a well-demarcated
‘‘layer’’ (Figures 3E–3G). Eye removal experiments indicated that
the GFP+axons projecting to the dLGN arose exclusively from
the contralateral retina (Figure S2). Given that every GFP+RGC
projects to the SC, the layer-specific tOFF-aRGC projections
to the dLGN must be the collaterals of retinocollicular axons.
Together, the presence of laminar- and columnar-specific
Mapping a Mosaic of Genetically Identified RGCs
428 Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc.
projections to the SC and the laminar-specific projections to
dLGN indicates there is a highly precise brain map associated
with the tOFF-aRGC mosaic.
Emergence of Laminar and Columnar Specificity
Involves Axonal Refinement
How do the highly specific patterns of axonal connections made
by tOFF-aRGCs arise during development? This could occur by
cisely to its proper layer and column(s), as has been described
for retinal projections in the fly and in the chick (Meinertzhagen
and Hanson, 1993; Inoue and Sanes, 1997). Alternatively,
tOFF-aRGC axons could at first project broadly within their tar-
gets, and subsequently refine to achieve laminar and columnar
specificity, as is the case for retinotopic and eye-specific map-
ping in mammals (McLaughlin and O’Leary, 2005; Torborg and
Feller, 2005). Because tOFF-aRGC projections to the SC
exhibit both laminar and columnar specificity (Figure 3) and
because the SC develops relatively early (Edwards et al.,
1986), we examined the development of these features in the
Studies in mice and hamsters have shown that RGC axons
reach the SC slightly before birth and initially overshoot their cor-
rect retinotopic termination zone. Between postnatal day 0 (P0)
retinotopic map (McLaughlin and O’Leary, 2005). Subsequently,
RGC axons elaborate terminal arbors in the overlying SGS
(Sachs et al., 1986). We observed that on P4, the axons of
tOFF-aRGCs do not exhibit their mature pattern of laminar or
columnar specificity (Figures 4A–4C0). While tOFF-aRGC axons
tended to arborize in the lSGS and rudimentary columns were
evident (Figures 4A–4C), it was readily apparent that many
GFP+tOFF-aRGC axons extend across the entire retinorecipient
depth of the SC (Figures 4A–4C0), a pattern never observed in
adulthood (Figure 3). Indeed, on P4, manyGFP+axons projected
all the way up to the pial margin of the SC (arrows in Figures 4B0
and 4C0). On P10, by contrast, laminar and columnar specificity
of tOFF-aRGC retinocollicular projections resembled the adult
pattern (Figures 4D–4F0). Quantitative analysis confirmed that a
significantly greater number of GFP+axons project to the
uSGS and SZ (the upper half of the retinorecipient SC; 0%–
50% depth from the pial margin) in P4 mice as compared to
P10 or adult mice (Figure 4G). There was also significantly less
columnar specificity of tOFF-aRGC retinocollicular projections
in P4 mice compared to P10 or adult mice (Figure 4H).
It is unlikely that the lack of laminar and columnar specificity
observed in P4 CB2-GFP mice is related to immaturity of the
SC layers or to increased numbers of RGCs expressing GFP
at this age because development of the retinorecipient SC
layers and RGC apoptosis are completed by P4 (Edwards
et al., 1986; Farah and Easter, 2005). Indeed, we found that
both the thickness of the retinorecipient SC (Figure 4I) and
the number of GFP+RGCs per retina were indistinguishable
between P4, P10 and adult CB2-GFP mice (Figure 4J). Also,
as in the adult mouse (Figure S2), the GFP+axons observed
in the SC of developing CB2-GFP mice arise solely from
tOFF-aRGCs in the contralateral retina and not from other
sources (Figure S4). We therefore conclude that the develop-
ment of laminar- and columnar-specific retinocollicular projec-
tions involves axonal retraction and refinement. Thus, in addi-
tion to any potential mechanisms that direct tOFF-aRGC
axons to their correct laminar and columnar positions, there
must also be mechanisms in place to remove the incorrect
projections made by these RGCs during early postnatal devel-
Cholinergic Spontaneous Retinal Activity Regulates
Emergence of Columnar Specificity
Laminar and columnar specificity of tOFF-aRGC projections to
the SC emerges between P4 and P10, which is prior to the onset
of vision (Demas et al., 2003). During this period, however,
‘‘waves’’ of neural activity sweep across the retina, inducing cor-
P10 coincides with the presence of cholinergic-mediated retinal
waves (Bansal et al., 2000; McLaughlin et al., 2003), we tested
the role of that activity on the emergence of laminar- and colum-
nar-specific tOFF-aRGC projections to the SC and the dLGN.
In the first set of experiments, we mated CB2-GFP mice to
mice lacking the b2- subunit of the nicotinic acetylcholine recep-
tor (b2nAChR?/?) (Picciotto et al., 1995; Xu et al., 1999) to gener-
ate CB2-GFP:b2nAChR?/?mice. b2nAChR?/?mice exhibit sig-
nificantly altered patterns of RGC bursting during the first 8 days
of postnatal life, after which spontaneous and visually evoked
RGC activity returns to normal (Bansal et al., 2000; McLaughlin
et al., 2003). At P4/5, the degree of laminar- and columnar-
specificity in the SC was qualitatively and quantitatively indistin-
guishable between CB2-GFP:b2nAChR+/+
CB2-GFP:b2nAChR+/?mice, and CB2-GFP:b2nAChR?/?mice
(Figure 5). On P10, however, CB2-GFP:b2nAChR?/?mice ex-
hibited a severely abnormal pattern of tOFF-aRGC projections
to the SC (Figures 6A–6F). Whereas overall target selection
and laminar specificity were normal in the SC (Figures 6A–6E)
and in the dLGN (Figure 7), therewas anobvious reduction in co-
lumnar specificity among tOFF-aRGC axons in the SC of CB2-
GFP:b2nAChR?/?mice (Figures 6C, 6D, and 6F).
Remarkably, even though spontaneous retinal activity in
b2nAChR?/?mice returns to normal by P8 (Bansal et al., 2000;
McLaughlin et al., 2003; Torborg and Feller, 2005), columnar
GFP:b2nAChR?/?mice (Figures 6F and 6I). This shows there is
a critical period for columnar segregation of tOFF-aRGC projec-
tions that is mediated by b2nAChR-mediated spontaneous
retinal activity during the first postnatal week. Importantly, the
number of GFP+tOFF-aRGCs and the retinorecipient thickness
of the SC were both normal in CB2-GFP:b2nAChR?/?mice
(Figures 6G and 6H). We thus attribute the permanent loss of
columnar specificity in CB2-GFP:b2nAChR?/?mice to a lack of
or indirect effects of the mutation on SC cells.
b2nAChR?/?mice lack b2nAChRs throughout the brain, in-
cluding in the SC where they are normally expressed (Picciotto
et al., 1995; Xu et al., 1999). So even though several aspects of
RGC axon targeting are normal in CB2-GFP:b2nAChR?/?mice
(Figures 6 and 7), the above experiments do not unequivocally
determine whether cholinergic spontaneous retinal activity in-
structs columnar-specific refinement of tOFF-aRGC projections
abnormal in adultCB2-
Mapping a Mosaic of Genetically Identified RGCs
Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc. 429
or whether there is a more general requirement for b2nAChRs for
columns to develop. To address this, we injected epibatidine
(epi), a cholinergic agonist that perturbs spontaneous RGC
bursting in the early postnatal retina (Penn et al., 1998; Cang
et al., 2005) into one eye, and we injected saline into the other
eye of CB2-GFP:b2nAChR+/+(WT) mice every 48 hr from P4 to
Figure 4. Development of Laminar- and Columnar-Specific tOFF-aRGC Projections Involves Axonal Refinement
(A–C0) Targeting of tOFF-aRGC axons to the SC is imprecise early in development. (A and B) Coronal view of the SC in a P4 CB2-GFP mouse. tOFF-aRGC axons
exhibit rudimentary columnar and laminar specificity, but unlike in adult CB2-GFP mice (Figure 3), many GFP+axons project into the uSGS and SZ,, and up to the
pial margin (arrows). (B0) Higher-magnification view of the boxed region in (A) and (B). Arrows indicate GFP+axons projecting to the uppermost SC. (C and C0)
Sagittal view of the SC on P4. GFP+axons are distributed across the retinorecipient depth of the SC. Arrows indicate GFP+axons projecting to the pial margin
(arrows). Asterisks in (A) indicate GFP+glia at the pial margin and a few GFP+neurons in the deeper, non-retinorecipient SC (see Figure 3). GFP expression in
these glia disappears by ?P8. Neither the GFP+glia nor GFP+SC neurons contribute to the GFP+axons in the young or adult SC (Figures S2 and S3).
(D–F0) Targeting of tOFF-aRGC axons to the SC is adult-like by P10; GFP+axons are restricted to the lSGS and segregated into distinct columns. (A–F) Scale is
250 mm. (B0, C0, E0, and F0) Scale is 125 mm.
(G) Quantification of laminar specificity as a function of age (see Experimental Procedures). The number of GFP+axons/mm projecting to the 25%–50% (upper
third) and 0%–25% (upper quarter) of the SC depth is significantly greater on P4 compared to P10 or adult (p < 0.0001). P10 versus adult were not significantly
different (25%–50% depth, p = 0.74; 0%–25% depth, p = 0.23; t test, ±SEM) (n = 4 adult mice, n = 6 P10 mice, n = 7 P4 mice).
(H) Columnar segregation index (see Experimental Procedures). There is significantly less columnar specificity on P4, as compared to P10 or adult (p < 0.0001),
whereas columnar specificity is indistinguishable between P10 and adult (p = 0.53) (n = 3 mice per age).
(I)The depth(retinorecipient thickness) oftheSCisnotdifferent between P4,P10,oradult CB2-GFP mice(P4 versusP10,p=0.3;P4 versusadult,p=0.48;ttest;
n = 5 P4 mice, n = 4 P10 mice, n = 4 adult mice).
(J) The number of GFP+RGCs per retina is not different on P4 compared to P10 and adult (P10+) (p = 0.96; t test; ± SEM; n = 5 P4 mice, n = 4 P10 mice, and n = 4
Mapping a Mosaic of Genetically Identified RGCs
430 Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc.
P10. We then examined the pattern of tOFF-aRGCs projections
to the SC on P10. Since all the GFP+tOFF-aRGCs in each eye
project to the contralateral SC (Figures S2 and S4), we could
directly compare the retinocollicular projections from the epi-
treated retina to the retinocollicular projections from the saline-
treated retina, in the two SC hemispheres of the same mouse.
In the SC hemisphere contralateral to the saline-injected eye,
tOFF-aRGC axons were restricted normally to the lSGS and
organized into distinct columns (Figures 8A and 8B). In the SC
ity was normal but columnar specificity was perturbed (Figures
8A and 8B). The defects in columnar specificity induced by epi
treatment were nearly as severe as those observed in CB2-
GFP:b2nAChR?/?mice (Figure 8C). We therefore conclude
that b2nAChR-mediated spontaneous activity that occurs in
the retina from P4–P10 mediates columnar-specific refinement
of tOFF-aRGC projections to the SC.
Elucidation of the Map of Projections
from a Mosaic of RGCs
We provide the first description of a complete mosaic of RGCs
(tOFF-aRGCs) selectively labeled with GFP. This enabled us to
relate the mosaic spacing, dendritic anatomy, and electro-
physiology of these RGCs with their complete map of connec-
tions in the brain. Mouse tOFF-aRGCs are thought to be
homologous to OFF-brisk-transient-Y RGCs in cats and mon-
keys (Pang et al., 2003; Schubert et al., 2005; Volgyi et al.,
2005; Berson, 2008). However, the complete pattern of projec-
tions arising from tOFF-a/Y RGCs has not previously been elu-
cidated. We find here that tOFF-aRGCs are remarkably spe-
cific with respect to which brain nuclei they project to, and
the organization of their projections within the retinotopic
maps of those nuclei. In the mouse, tOFF-aRGCs project
Figure 5. Genetic Removal of b2nAChRs Does Not Alter the Initial Targeting of tOFF-aRGC Axons
(A–D) On P4, targeting of tOFF-aRGC axons to the SC is normal in CB2-GFP:b2nAChR+/?mice and CB2-GFP:b2nAChR?/?mice. Coronal view of the SC in a P4
CB2-GFP:b2nAChR+/?mouse (A and B) and a P4 CB2-GFP:b2nAChR?/?mouse (C and D). As in WT P4 mice (Figure 4) laminar and columnar specificity are not
mature at this age. Arrows indicate GFP+axons that project up near the pial margin. Scale is 250 mm.
(E) Quantification of laminar specificity for each genotype on P4/5; laminar specificity is indistinguishable between WT and b2nAChR+/?mice (25%–50% depth,
p = 0.74; 0%–25% depth, p = 0.54), between b2nAChR+/?and b2nAChR?/?mice (25%–50% depth, p = 0.82; 0%–25% depth, p = 0.45), and between WT and
b2nAChR?/?mice (25%–50% depth, p = 0.73; 0%–25% depth, p = 0.60), indicating that all genotypes start out with similar patterns of axon targeting on P4/5.
(F) Quantification of columnar specificity across genotypes on P4/5 (WT versus b2nAChR+/?, p = 0.67; b2nAChR+/?versus b2nAChR?/?, p = 0.61; WT versus
b2nAChR?/?, p = 0.97).
Error bars are ± SEM.
Mapping a Mosaic of Genetically Identified RGCs
Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc. 431
only to the dLGN and the SC, and within both these targets,
they form precisely targeted laminar-specific connections.
Laminar-specificity of retinocollicular projections from function-
ally distinct RGCs has been observed previously by recording
from and filling individual RGC axons in the optic tract of cats,
monkeys, and hamsters (Bowling and Michael, 1980; Berson,
1988; Mooney and Rhoades, 1990; Tamamaki et al., 1995)
and was suggested from retrograde filling experiments in
mice (Hofbauer and Dra ¨ger, 1985). Those studies agree with
our findings that a/Y RGCs project to the lSGS. But because
previous studies relied on filling of randomly-targeted individ-
ual RGC axons, our results are the first to delineate the path-
way corresponding to a complete mosaic of one subtype of
Figure 6. Cholinergic Spontaneous Activity Mediates Columnar- but Not Laminar-Specific Refinement of tOFF-aRGC Projections to the SC
(A and B)Coronalview of tOFF-aRGC projections totheSCinaP10CB2-GFP:b2nAChR+/?mouse. Laminarspecificityand columnar specificityaresimilartoWT
P10 mice (see Figure 4).
(C and D) Coronal view of tOFF-aRGC projections to the SC in aP10 CB2-GFP:b2nAChR?/?mouse. Laminar specificityis normal for this age but columnar spec-
ificity is severely perturbed. (A–D) Scale is 250 mm.
(E)Quantificationoflaminarspecificityaccording togenotype.Nosignificantdifferences werepresentbetweenWT,b2nAChR+/?,orb2nAChR?/?CB2-GFPmice
on P10 (p = 0.3–0.7; ±SEM; n = 3 mice per condition).
(F) Quantification of columnar specificity in the P10 and adult SC, according to genotype. At P10 and at adulthood, CB2-GFP:b2nAChR?/?mice exhibit signif-
icantly less columnar specificity than b2nAChR+/?mice or WT (b2nAChR+/+) mice (**p < 0.001; n = 3 mice per genotype; ±SEM; Student’s t test).
(G) The thickness of the retinorecipient SC depth is not different between WT, b2nAChR+/?or b2nAChR?/?mice at P10 or in adulthood (p = 0.2–0.5, t test, n = 3
mice per genotype).
(H) The number of GFP+RGCs per retina is not different between WT and b2nAChR?/?CB2-GFP mice (p = 0.75; t test; ±SEM; n = 4 mice per genotype).
Mapping a Mosaic of Genetically Identified RGCs
432 Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc.
Whereas laminar organization of retinogeniculate projections
from functionally distinct RGCs is a hallmark feature of the carni-
vore and primate visual pathway (Callaway, 2005), this feature
was thought to be absent in rodents (Reese, 1988). We find,
however, that tOFF-aRGCs project to stereotyped lamina in
the dLGN of the mouse, revealing an unexpectedly high degree
of synaptic specificity in the rodent visual system. Our discovery
that the axons from the mosaic of tOFF-aRGCs form a columnar
map in the SC reveals the precision with which single RGC sub-
types arrange their projections within a topographic map. These
results and those in recent reports (Yonehara et al., 2008; Kim
et al., 2008) highlight the usefulness of genetic methods to iden-
tify functionally distinct RGC subtypes and to delineate their pro-
jections within multiple brain targets, even when those targets
consist of compact neuropil with no obvious cytoarchitectural
Emergence of Synaptic Specificity in the Mammalian
CNS Involves Axonal Refinement
Our results show that mammalian RGC projections achieve
laminar and columnar specificity by first transitioning through a
period of imprecision (Figure 4). The emergence of synaptic
specificity in the mammalian visual system thus differs from
that of the invertebrate CNS where laminar and columnar spec-
ificity of photoreceptor projections emergebydirected growth of
appropriately targeted synaptic connections (Meinertzhagen
and Hanson, 1993; Clandinin and Zipursky, 2002). It is unlikely
that the imprecision we observed simply reflects tOFF-aRGCs
axons contacting the proper target SC cells that have not yet
migrated to their final positions, because cell migration and lam-
ination in the SC is completed by P5 (Edwards et al., 1986). Also,
the SC follows an inside-out gradient of neurogenesis (Edwards
et al., 1986), meaning that lSGS neurons develop before the
uSGS.Therefore, tOFF-aRGCsproject through andbeyond their
correct target lamina before retracting to form laminar-specific
connections in the lSGS.
Spontaneous Retinal Activity Segregates Axons
Arising from RGCs of the Same Subtype
Cholinergic retinal waves are thought to segregate RGC axons
from the two eyes and glutamatergic waves are thought to seg-
Feller, 2005; Kerschensteiner and Wong, 2008). Here, we pro-
vide evidence that cholinergic spontaneous retinal activity also
segregates axons arising from RGCs in the same eye and of
the same subtype. In previous studies, the consequences of al-
labeling individual RGC axons (Sretavan et al., 1988), (2) labeling
all the RGC axons arising each eye (Penn et al., 1998; Huberman
et al., 2002; Torborg and Feller, 2005), or (3) by labeling groups
(?500)of axons arisingfrom retinotopically closebut functionally
diverse RGCs (Grubb etal.,2003; McLaughlin etal.,2003; Chan-
drasekaran et al., 2005). Thus, it was unclear precisely which
features of axon targeting and refinement were impacted by ac-
tivity. Indeed, some studies have argued that activity is merely
permissive for axon growth and targeting (Crowley and Katz,
2000; Huberman et al., 2003). Our observation that retinocollic-
ular projections are similar between b2nAChR?/?mice and con-
trol mice before refinement on P4 (Figure 5) and that laminar
specificity emerges despite manipulations that severely perturb
columnar specificity (Figures 6–8) argues against a merely per-
missive influence of spontaneous retinal activity on axon growth.
The data presented here thus identify cholinergic retinal waves
as a key influence on the emergence of synaptic specificity by
clustering groups of axons arising from RGCs of the same
Our finding that cholinergic spontaneous retinal activity is not
necessary to refine tOFF-aRGCs projections into specific lamina
within their targets differs from eye-specific laminar refinement,
which clearly requires cholinergic transmission (Penn et al.,
1998; Rossi et al., 2001; Muir-Robinson et al., 2002; Huberman
et al., 2002; Torborg and Feller, 2005). It is worth noting that
glutamatergic waves appear in b2nAChR?/?mice by P8 (Bansal
et al., 2000) whereas we evaluated laminar specificity on P10
because it is the earliest age when laminar and columnar refine-
ment are adult-like. It is therefore possible that glutamatergic
waves rescue laminar specificity of tOFF-aRGC projections
between P8and P10.Nevertheless, allowinganextended period
of survival past P10 does not rescue columnar specificity (Fig-
ure 6), indicating a lasting, critical role for cholinergic spontane-
ous retinal activity in columnar but not laminar refinement.
The fact that laminar specificity can emerge in the absence of
cholinergic retinal activity is consistent with previous studies on
the role of activity in lamina-specific targeting of retinal projec-
tions in the fly and chick (Inoue and Sanes, 1997; Hiesinger
et al., 2006). We also observed normal laminar-specific targeting
Figure 7. Disruption of Cholinergic Sponta-
neous Retinal Activity Does Not Prevent
Laminar-Specific Targeting of tOFF-aRGC
Axons to the dLGN
(A–C) Retino-dLGN projections in a P10 CB2-
GFP:b2nAChR?/?mouse. (A) CTb-594 (red) la-
beled RGC retino-dLGN projections. (B) GFP+
tOFF-aRGC axons avoid the outer third (asterisks)
ner portion of the dLGN, forming a distinct layer
just as in WT CB2-GFP mice (see Figure 3). Note
also that the GFP+tOFF-aRGC axons avoid the
vLGN and IGL, just as in WT CB2-GFP mice. (C)
Merged view of (A) and (B). Scale is 150 mm.
Mapping a Mosaic of Genetically Identified RGCs
Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc. 433
of tOFF-aRGCs dendrites to the OFF sublamina of the IPL of
CB2-GFP:b2nAChR?/?mice (data not shown), which was ex-
pected based on previous results (Bansal et al., 2000). Although
we do not rule out the possibility that noncholinergic forms of
retinal spontaneous activity could mediate laminar-specific
targeting of tOFF-aRGC axons or dendrites, we speculate that
RGCs select their laminar-specific connections primarily on the
basis of cell-type specific expression of adhesion molecules,
1997; Clandinin and Zipursky, 2002; Chen and Clandinin, 2008;
Nern et al., 2008; Yamagata and Sanes, 2008). In CB2-GFP
mice, tOFF-aRGCs can be identified by virtue of their GFP ex-
pression early in development (Figure S5), making these mice
potentially useful for studying development of RGC morphology,
mosaic spacing and tiling, as well as for identifying candidate
genes that control lamina-specific targeting of tOFF-aRGC
axons and dendrites.
Activity-Dependent Segregation of tOFF-aRGC
Projections into Columns: Relevance to Models of
Ocular Dominance Column Development
Our findings that emergence of columnar-specific projections
from tOFF-aRGCs involves axonal refinement and requires
of cortical ocular dominance column (ODC) formation (LeVay
et al., 1978; Stryker and Harris, 1986; Huberman et al., 2006)
and with experiments on ODCs in the tecta of three-eyed frogs
(Reh and Constantine-Paton, 1985). Our results contrast, how-
toreceptor axons is precise from the outset and does not require
synaptic transmission (Hiesinger et al., 2006). This raises the
question of whether, in mammals, the segregation of visual pro-
jections into columns reflects generic mechanisms of activity-
dependent axon sorting that occur during development (Adams
related to the types of neural information they carry.
Along these lines, we considered what aspects of retinocollic-
ular circuitry tOFF-aRGC columns might correspond to and why
they require cholinergic spontaneous retinal activity in order to
develop. Based on the mapping of retinal coordinates onto
the SC in the mouse (Dra ¨ger and Hubel, 1976), we determined
that the size of each GFP+column in the SC corresponds to
about 240 to 300 mm across the retina. Cholinergic waves cor-
relate the firing of RGCs whose cell bodies are located within
0 to 300 mm of each other (McLaughlin et al., 2003). Given the
average spacing of tOFF-aRGCs, cholinergic waves should cor-
relate nearby groups of 3 to 4 GFP+RGCs. These measure-
ments support the idea that the columns described here arise
from the axon terminals of neighboring GFP+RGCs that experi-
ence correlated activity in the retina. In this way, the columns we
describe can be thought of as anchoring axons arising from
closely positioned RGCs of the same subtype, through Hebbian
Why does columnar specificity fail to develop in b2nAChR?/?
mice? In these mutants, correlated RGC firing is significantly
lower than normal for RGCs situated up to 300 mm apart
(McLaughlin et al., 2003). This should reduce Hebbian strength-
ening of axonal inputs from groups of GFP+RGCs that project to
the same SC cells and prevent weakening and elimination of
RGC axons misprojecting to intercolumn zones or to incorrect
columns. This model is consistent with our observation that
qualitatively and quantitatively, some columnar architecture is
present in both WT and b2nAChR?/?mice on P4/5 (Figure 5)
and that columnar specificity is severely reduced but not entirely
absent in P10 and adult b2nAChR?/?mice (Figure 6). This model
is also consistent with studies that perturbed retinal waves or
blocked NMDA receptors in the SC during development and
observed increased receptive field sizes of SC neurons (Chan-
drasekaran et al., 2005; Huang and Pallas, 2001).
One consideration is that mouse tOFF-aRGCs are coupled to
one another through connexin36 gap junctions (Schubert et al.,
2005). Genetic removal of connexin36 abolishes tOFF-aRGC
coupling (Schubert et al., 2005), and yet the same manipulation
does not perturb early spontaneous retinal activity or eye-spe-
cific refinement in the dLGN (Torborg and Feller, 2005). Thus, it
is unlikely that genetic removal of b2nAChRs or epibatidine
Figure 8. Columnar-Specific Refinement Is Mediated by Cholinergic Spontaneous Activity in the Retina
(AandB)RGCprojections totheSCofaP10WTCB2-GFPmousethatreceivedepibatidine (epi)injectionsintotherighteyetodisruptspontaneousretinalactivity
and saline (control) injections into the left eye, from P4–P10. Normal laminar-specific targeting of GFP+tOFF-aRGC axons to the lSGS is evident in both SC hemi-
spheres. Columnar segregation is normal in the saline hemisphere but is perturbed in the SC hemisphere contralateral to the epi-injected eye. Scale is 250 mm.
(C) Quantification of columnar specificity in the SC hemisphere receiving input from control, saline- or epi-injected retina (epi versus saline **p < 0.001; saline
versus control p = 0.67; ±SEM; t test; n = 2 mice).
Mapping a Mosaic of Genetically Identified RGCs
434 Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc.
prevents columnar segregation of tOFF-aRGC axons by reduc-
ing gap junction coupling of these RGCs. We thus conclude
that altered correlated RGC firing (Bansal et al., 2000; McLaugh-
lin et al., 2003; Cang et al., 2005; Torborg and Feller, 2005) is the
factor primarily responsible for the defective columnar refine-
ment of tOFF-aRGC projections in CB2-GFP:b2nAChR?/?and
At this time, we do not know what the columns arising from
tOFF-aRGCs contribute to visual processing or behavior. Inter-
estingly, othershave described columns of somatosensory input
to the deeper nonretinorecipient SC layers that are in register
with the retinotopic map (Dra ¨ger and Hubel, 1976). Whether so-
but it is tempting to speculate about their possible relationship in
multimodal sensory integration. Ultimately, tools that allow ma-
nipulation of neural activity in particular RGC subtypes will allow
us to determine their contribution to visual perception and be-
havior. In the meantime, our findings show that transgenic
mice with genetically identified subtypes of retinal neurons ex-
pressing GFP can serve as useful tools for elucidating previously
uncharted structure-function relationships in the visual system
and for studying the mechanisms underlying development of
synaptic specificity in the mammalian CNS.
Mice were obtained from the Mouse Regional Resource Center (detailed infor-
at: http://www.mmrrc.org/strains/283/0283.html). Mice were generated on
Swiss-webster background and were crossed onto C57/Bl6 background for
multiple generations in our laboratory. In the text, postnatal day 0 corresponds
to the day of birth. GFP expression in the retina of CB2-GFP was reflective of
mRNA insitufor calretinin inthe retina, buttherewas less GFP inthebrain than
would be expected on the basis of calretinin mRNA expression (data not
shown). This reflects positional effects of the CB2-GFP transgene insertion
(Gong et al., 2003).
These mice were generated by breeding b2nAChR?/?mice (Xu et al., 1999;
obtained from D. Feldheim, UC Santa Cruz) to CB2-GFP mice. The CB2-
GFP:b2nAChR+/?offspring were backcrossed to b2nAChR?/?mice on
a C57/Bl6 background, and the mixed background CB2-GFP:b2nAChR?/?
out as previously described (Xu et al., 1999).
Mice were perfused with saline followed by 4% parformaldehyde (PFA), the
eyes removed and postfixed 4 hr in 4% PFA at 4?C, then the retina was
extracted. For whole mount, relieving cuts were made to allow the retina to
lay flat. Retinas were incubated in blocking solution (10% goat serum, 0.25%
Triton-X in PBS) for 2 hr, then transferred to primary antibodies (1:1000 rabbit
anti-GFP from Molecular Probes; 1:5000 mouse SMI-32, Sternberger mono-
clonals) for 48 hr at 4?C, washed in PBS (5 3 1 hr), transferred to secondary
antibody (1:1000 goat anti-rabbit Alexa 488, 1:1000 goat anti-mouse Alexa
594;Molecularprobes) for2hr,washed inPBS(53 1hr),mounted,and cover-
slipped with Vectashield (Vectorlabs). Retinal sections were as follows: the
retina was immersed in 30% sucrose, then cryosectioned at 12 mm. Block =
1 hr, primary incubation = 24 hr at 4?C, secondary incubation = 1 hr. VAChT
staining (1:1000 goat anti-VAChT, Chemicon), required the use of donkey
serum rather than goat serum.
DiI Labeling of Single RGCs
Methods wereasdescribedpreviously(KimandJeon, 2006).Wholeeyeswere
fixed for 8 hr in 4% PFA, the retina extracted and stored in PBS. GFP+cells
were filled iontophoretically using an Axoprobe-1A amplifier and the lipophilic
dye,DiI.Electrodeshadresistancesofabout50MOhmand wereabout 0.5mm
in size. A positive current of between 5 and 20 nA was applied for 20–30 min to
fill each cell.
Quantification of GFP+RGCs
The number of GFP+RGCs per retina was counted from 53 or 103 montaged
parisons here and unless otherwise noted, for all quantitative measurements
listed below. SEMs are shown.
Quantification of Mosaic Regularity
The density recovery profile (Rodieck, 1991) describes the average density of
cells at a given distance from the soma, and was calculated as
where niðsÞ is a histogram of the number of cells in an image at distance s
from cell i, and AiðsÞ is the area contained in that image at distance s from
cell i. rðsÞ was calculated from all GFP+RGCs in 27 images of central and
peripheral retina from 6 different mice.
The dark-adapted mouse retina was isolated under an infrared microscope
into oxygenated Ringer’s solution (124 mM NaCl, 2.5 mM KCl, 2 mM CaCl2,
2 mM MgCl2,1.25 mM NaH2PO4, 26 mM NaHCO2, and 22.2 mM glucose),
and the pH was maintained at 7.4 by bubbling of 95% O2, 5% CO2at 36?C.
A flattened retina-eyecup preparation (Hu et al., 2000) was placed with gan-
glion cells facing upwards in a superfusion chamber on the stage of an upright
fluorescence microscope and held in place with a transparent dialysis mem-
brane containing several ?200 mm holes. For intracellular recordings, sharp
microelectrodes were filled with 2 M potassium acetate, having a final imped-
ance of 250–350 MU. Fluorescent retinal ganglion cells were detected at 340
magnification by excitation with a blue LED. To minimize effects of adaptation
and photodamage, single fluorescent images were taken at a very dim excita-
tion intensity (?10 mW/m2), within the range of intensities used for visual
cell soma by the intracellular electrode, the inner limiting membrane was
microdissected using a patch pipette. After a fluorescent cell was identified,
it was impaled under oblique infrared illumination.
Stimulation and analysis was as follows: light stimuli were projected from
a computer video monitor through a custom-made lens system and focused
from above onto the photoreceptor layer through a 310 water immersion
objective a mean photopic intensity of ?8 mW/m2. Flashes (0.5 Hz) were
presented using the maximum and minimum intensities of the monitor. The
spatiotemporal receptive field of all neurons was measured by the standard
method of reverse correlation of the membrane potential to a flickering check-
erboard stimulus (110 mm regions changing every 30 ms; Baccus et al., 2008).
To compute correlation coefficients between the average flash responsesof
neurons, spikes were first removed by setting a threshold for the derivative of
the membrane potential, and then the average was computed from 10–15
flashes. The correlation coefficient r between two average flash responses
xðtÞ and yðtÞ was calculated as
where T is the duration of the recording, and h.i denotes the time average.
ðxðtÞ ? hxiÞðyðtÞ ? hyiÞdt
ðxðtÞ ? hxiÞ2dt
ðyðtÞ ? hyiÞ2dt
Mapping a Mosaic of Genetically Identified RGCs
Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc. 435
Mice were perfused, brains removed, postfixed for 8 hr in 4% PFA, immersed
was carried out as above except that primary antibody incubation was over-
night at 4?C, and secondary incubation was 90 min at RT.
CTb-594 Labeling of RGC Axons
Two to five microliters of CTb-594 (0.5% diluted in saline; Molecular Probes)
was injected into the vitreous of each eye (2 ml/eye on P4/5, 3.5 ml/eye at
P10; 5 ml/eye in adults). The mouse was perfused 24 hr later (ages in text refer
to age of animal at perfusion). CTb-594 was visualized without any immunos-
One microliter of 1 mM epi was injected into the vitreous of one eye every 48 hr
on P4, P6, P8 (Rossi et al., 2001). An equivalent volume of saline vehicle was
injected into the opposite eye and both eyes were labeled with CTb-594 on
P9, and brain tissue harvested on P10.
Quantification of Retinorecipient SC Depth
The distance from the base of the SO to the pial surface of the SC was
measured from CTb-594 labeling at four locations across each SC section.
Measures were taken from 2–4 SC sections per animal and averaged before
Quantification of Laminar Specificity
The depth of the retinorecipient SC was divided into quarters along its depth
such that the pial margin was 0% and the base of the SO was 100%. Zero
to fifty percent depth corresponds well to the combined thickness of the
uSGSandSZ.Oneand one-half totwomillimeter linescansweremade across
the rostral-caudal (sagittal sections) and medial lateral (coronal sections) SC
and the number of GFP+axons intersected per mm was calculated in the
0%–25% and 25%–50% depths. Two to four sections per animal were
analyzed to generate an average per animal. The density of GFP labeling in
the 50%–75% and 75%–100% depths of the SC (the lSGS and SO) was too
high to count individual axons. We therefore quantified the degree of mistar-
geting to the 0%–50% depth (the uSGS and SZ).
Quantification of Columnar Specificity
To measure the degree of columnar specificity in the SC, we calculated a
column segregation index (CSI) (Crowley and Katz, 2000) according to
CSI=1 ?m ? b
c ? b
where c is the image intensity in the center of a fluorescent patch, m is
the intensity at the midpoint between patches, and b is the intensity of the
background tissue, measured in square regions 100 mm on a side. CSI ranged
between 0 and 1 and was averaged over all adjacent patch/interpatch pairs in
age or condition was the CSI zero, indicating that some columnar structure
was present even on P4 and after activity manipulations. The near-complete
lack of columnar segregation in some CB2-GFP:b2nAChR?/?mice made it
difficult to identify interpatch zones in those animals. We therefore made
line scans across the extent of the lSGS and took the troughs as interpatch
The Supplemental Data include five figures and can be found with this article
online at http://www.neuron.org/cgi/content/full/59/3/425/DC1/.
We are grateful to D. Feldheim, T. Clandinin, T. Cutforth, and J. Krey for helpful
comments and suggestions. We thank B. Chapman for use of her microscope
and J. Yamada and M. Martin for genotyping. This work was supported by NIH
RO1 EY011310-13, NIH R21 EY018320 (B.A.B.), NIH R01 EY016842 (S.A.B.),
a March of Dimes Basil O’Connor Award (E.M.U), Research to Prevent Blind-
ness Young Investigator Award (E.M.U.), NIH EY02162 (E.M.U.), NIH
R21MH083090 (E.M.U.), and T32 EY07120 (S.K.). A.D.H. was supported by
a Helen Hay Whitney Postdoctoral Fellowship.
Accepted: July 8, 2008
Published: August 13, 2008
Adams, D.L., and Horton, J.C. (2003). Capricious expression of cortical
columns in the primate brain. Nat. Neurosci. 6, 113–114.
Baccus, S.A., Olveczky, B.O., Manu, M., and Meister, M. (2008). A retinal
circuit that computes object motion. J. Neurosci. 28, 6807–6817.
Bansal, A., Singer, J.H., Hwang, B.J., Xu, W., Beaudet, A., and Feller, M.B.
(2000). Mice lacking specific nicotinic acetylcholine receptor subunits exhibit
dramatically altered spontaneous activity patterns and reveal a limited role
for retinal wavesin forming ON and OFF circuits in theinner retina. J.Neurosci.
Benson, D.L., Colman, D.R., and Huntley, G.W. (2001). Molecules, maps and
synaptic specificity. Nat. Rev. Neurosci. 2, 899–909.
Berson, D.M. (1988). Retinal and cortical inputs to cat superior colliculus:
composition, convergence and laminar specificity. Prog. Brain Res. 75,
The Senses: A Comprehensive Reference, Volume 1, R. Masland and T.D.
Albright, eds. (San Diego, CA: Academic Press), pp. 491–520.
Bowling, D.B., and Michael, C.R. (1980). Projection patterns of single physio-
logically characterized optic fibres in cat. Nature 286, 899–902.
Callaway, E.M. (2005). Structure and function of parallel pathways in the pri-
mate early visual system. J. Physiol. 566, 13–19.
Cang, J., Renteria, R.C., Kaneko, M., Liu, X., Copenhagen, D.R., and Stryker,
M.P. (2005). Development of precise maps in visual cortex requires patterned
spontaneous activity in the retina. Neuron 48, 797–809.
Chandrasekaran, A.R., Plas, D.T., Gonzalez, E., and Crair, M.C. (2005).
Evidence for an instructive role of retinal activity in retinotopic map refinement
in the superior colliculus of the mouse. J. Neurosci. 25, 6929–6938.
Chen, P.L., and Clandinin, T.R. (2008). The cadherin Flamingo mediates level-
dependent interactions that guide photoreceptor target choice in Drosophila.
Neuron 58, 26–33.
Clandinin, T.R., and Zipursky, S.L. (2002). Making connections in the fly visual
system. Neuron 35, 827–841.
Crowley, J.C., and Katz, L.C. (2000). Early development of ocular dominance
columns. Science 290, 1321–1324.
nous activity inthe mouse retina is independent of visual experience. J.Neuro-
sci. 25, 9347–9357.
Dra ¨ger, U.C., and Hubel, D.H. (1976). Topography of visual and somatosen-
sory projections to the mouse superior colliculus. J. Neurophysiol. 39, 91–101.
Edwards, M.A., Caviness, V.S., and Schneider, G.E. (1986). Development of
cell and fiber lamination in the mouse superior colliculus. J. Comp. Neurol.
Famiglietti, E.V., and Kolb, H. (1976). Structural basis for ON- and OFF-center
responses in retinal ganglion cells. Science 194, 193–195.
Farah, M.H., and Easter, S.S. (2005). Cell birth and death in the mouse retinal
ganglion cell layer. J. Comp. Neurol. 489, 120–134.
Nowak, N.J., Joyner, A., Leblanc, G., Hatten, M.E., and Heintz, N. (2003). A
gene atlas of the central nervous system based on bacterial artificial chromo-
somes. Nature 425, 917–925.
Mapping a Mosaic of Genetically Identified RGCs
436 Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc.
Grubb, M.S., Rossi, F.M., Changeux, J.P., and Thompson, I.D. (2003). Abnor-
mal functional organization in the dorsal latera; geniculate nucleus of mice
lacking the beta 2 subunit of the nicotinic acetylcholine receptor. Neuron 30,
Hiesinger, P.R., Zhai, R.G., Zhou, Y., Koh, T.W., Mehta, S.Q., Schulze, K.L.,
Cao, Y., Verstreken, P., Clandinin, T.R., Fischbach, K.F., et al. (2006). Activ-
ity-independent prespecification of synaptic partners in the visual map of Dro-
sophila. Curr. Biol. 16, 1835–1843.
Hofbauer, A., and Dra ¨ger, U.C. (1985). Depth segregation of retinal ganglion
cells projecting to the mouse superior colliculus. J. Comp. Neurol. 234, 465–
preparation suitable for electrophysiological studies of neurons visualized with
trans-scleral infrared illumination. J. Neurosci. Methods 103, 209–216.
Huang, L., and Pallas, S.L. (2001). NMDA antagonists in the superior colliculus
prevent developmental plasticity but not visual transmission. J. Neurophysiol.
Huberman, A.D., Stellwagen, D., and Chapman, B. (2002). Decoupling eye-
specific segregation from lamination in the lateral geniculate nucleus.
J. Neurosci. 22, 9419–9429.
Huberman, A.D., Wang, G.Y., Liets, L.C., Collins, O.A., Chapman, B., and
Chalupa, L.M. (2003). Eye-specific segregation independent of normal neuro-
nal activity. Science 300, 994–998.
Huberman, A.D., Speer, C.M., and Chapman, B. (2006). Spontaneous retinal
activity mediates development of ocular dominance columns and binocular
receptive fields in v1. Neuron 52, 247–254.
Inoue, A., and Sanes, J.R. (1997). Lamina-specific connectivity in the brain:
regulation by N-cadherin, neurotrophins, and glycoconjugates. Science 276,
Kim, T.J., and Jeon, C.J. (2006). Morphological classification of parvalbumin-
containing retinal ganglion cells in mouse: single cell injection after immunocy-
tochemistry. Invest. Ophthalmol. Vis. Sci. 47, 2757–2764.
ular identification of a retinal cell type that responds to upward motion. Nature
Kerschensteiner, D., and Wong, R.O. (2008). A precisely timed asynchronous
pattern of ON and OFF retinal ganglion cell activity during propagation of
retinal waves. Neuron 58, 851–858.
Kong, J.H., Fish, D.R., Rockhill, R.L., and Masland, R.H. (2005). Diversity of
ganglion cells in the mouse retina: unsupervised morphological classification
and its limits. J. Comp. Neurol. 489, 293–310.
Lee, E.J., Kim, H.J., Lim, E.J., Kim, I.B., Kang, W.S., Oh, S.J., Rickman,
D.W., Chung, J.W., and Chun, M.H. (2004). AII amacrine cells in the mamma-
lian retina show disabled-1 immunoreactivity. J. Comp. Neurol. 470, 372–
LeVay, S., Stryker, M.P., and Shatz, C.J. (1978). Ocular dominance columns
J. Comp. Neurol. 179, 223–244.
Luo, L., and Flanagan, J.G. (2007). Development of continuous and discrete
neural maps. Neuron 56, 284–300.
May, P.J. (2005). The mammalian superior colliculus: laminar structure and
connections. Prog. Brain Res. 151, 321–378.
McLaughlin, T., and O’Leary, D.D. (2005). Molecular gradients and develop-
ment of reitnotopic maps. Annu. Rev. Neurosci. 28, 327–355.
McLaughlin, T., Torborg, C.L., Feller, M.B., and O’Leary, D.D. (2003). Retino-
topic map refinement requires spontaneous retinal waves during a brief critical
period of development. Neuron 40, 1147–1160.
Meinertzhagen, I.A., and Hanson, T.E. (1993). The development of the optic
lobe. In The Development of Drosophila Melanogaster, M. Bate and A. Marti-
nez-Arias, eds. (New York: CSHL Press), pp. 1363–1491.
Mooney, R.D., and Rhoades, R.W. (1990). Relationships between physiologi-
cal and morphological propoeties of retinocollicular axons in the hamster.
J. Neurosci. 10, 3164–3177.
Muir-Robinson, G., Hwang, B.J., and Feller, M.B. (2002). Retinogeniculate
axons undergo eye-specific segregation in the absence of eye-specific layers.
J. Neurosci. 22, 5259–5264.
Nern, A., Zhu, Y., and Zipursky, S.L. (2008). Local N-cadherin interactions
mediate distinct steps in the targeting of lamina neurons. Neuron 58,
Pang, J.J., Gao, F., and Wu, S.M. (2003). Light-evoked excitatory and inhibi-
tory synaptic inputs to ON and OFF alpha ganglion cells in the mouse.
J. Neurosci. 23, 6063–6073.
Peichl, L. (1991). Alpha ganglion cells in mammalian retinae: common proper-
ties, species differences, and some comments on other ganglion cells. Vis.
Neurosci. 7, 155–169.
Penn, A.A., Riquelme, P.A., Feller, M.B., and Shatz, C.J. (1998). Competition in
retinogeniculate patterning driven by spontaneous activity. Science 279,
Picciotto, M.R., Zoli, M., Lena, C., Bessis, A., Lallemand, Y., Le Novere, N.,
Vincent, P., Pich, E.M., Brulet, P., and Changuex, J.P. (1995). Abnormal avoid-
ance learning in mice lacking functional high-affinity nicotinic in the brain.
Nature 374, 65–67.
Reese, B.E. (1988). Hidden lamination in the dorsal lateral geniculate nucleus:
the functional organization of this thalamic region in the rat. Brain Res. 472,
Reh, T.A., and Constantine-Paton, M. (1985). Eye-specific segregation
requires neural activity in three-eyed Rana pipiens. J. Neurosci. 5, 1132–
Rodieck, R.W. (1991). The density recovery profile: a method for the analy-
sis of points in the plane applicable to retinal studies. Vis. Neurosci. 6, 95–
Roska, B., and Werblin, F. (2001). Vertical interactions across ten parallel,
stacked representations in the mammalian retina. Nature 410, 583–587.
Rossi, F.M., Pizzorusso, T., Porciatti, V., Marubio, L.M., Maffei, L., and Chan-
guex, J.P. (2001). Requirement of the nicotinic acetylcholine receptor b2
subunit for the anatomical and functional development of the visual system.
Proc. Natl. Acad. Sci. USA 98, 6453–6458.
Sachs, G.M., Jacobson, M., and Caviness, V.S. (1986). Postnatal changes in
arborization patterns of murine retinocollicular axons. J. Comp. Neurol. 246,
Schubert, T., Degen, J., Willecke, K., Hormuzdi, S.G., Monyer, H., and Weiler,
R. (2005). Connexin36 mediates gap junctional coupling of alpha-ganglion
cells in mouse retina. J. Comp. Neurol. 485, 191–201.
Sretavan, D.W., Shatz, C.J., and Stryker, M.P. (1988). Modification of reitnal
ganglion cell axon morphology by prenatal infusion with tetrodotoxin. Nature
Stryker, M.P., and Harris, W.A. (1986). Binocular impulse blockade prevents
the formation of ocular dominance columns in cat visual cortex. J. Neurosci.
Sun, W., Li, N., and He, S. (2002). Large-scale morphological survey of mouse
retinal ganglion cells. J. Comp. Neurol. 451, 115–126.
Tamamaki, N., Uhlrich, D.J., and Sherman, S.M. (1995). Morphology of phys-
iologically identified retinal X and Y axons in the cat’s thalamus and midbrain
as revealed by intraaxonal injection of biocytin. J. Comp. Neurol. 354,
Torborg, C.L., and Feller, M.B. (2005). Spontaneous patterned retinal activity
and the refinement of retinal projections. Prog. Neurobiol. 76, 213–235.
Volgyi, B.,Abrams,J.,Paul,D.L., and Bloomfield, S.A.(2005). Morphology and
tracer coupling pattern of alpha ganglion cells in the mouse retina. J. Comp.
Neurol. 492, 66–77.
Wa ¨ssle, H. (2004). Parallel processing in the mammalian retina. Nat. Rev.
Neurosci. 5, 747–757.
Mapping a Mosaic of Genetically Identified RGCs
Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc. 437
Wong, R.O. (1999). Retinal waves visual system development. Annu. Rev. Download full-text
Neurosci. 22, 29–47.
Xu, W., Orr-Utreger, A., Nigro, F., Gelber, S., Sutcliffe, C.B., Armstrong, D.,
Patrick, J.W., Role, L.W., Beaudet, A.L., and De Biasi, M. (1999). Multiorgan
autonomic dysfunction in mice lacking the beta2 and the beta4 subunits of
neuronal nicotinic acetylcholine receptors. J. Neurosci. 19, 9298–9305.
ina-specific synaptic connections in vertebrate retina. Nature 451, 465–469.
Yonehara, K., Shintani, T., Suzuki, R., Sakuta, H., Takeuchi, Y., Nakamura-
Yonehara, K., and Noda, M. (2008). Expression of SPIG1 reveals development
of a retinal ganglion cell subtype projecting to the medial terminal nucleus in
the mouse. PloS ONE. 3, e1533. 10.1371/journal.pone.0001533.
Mapping a Mosaic of Genetically Identified RGCs
438 Neuron 59, 425–438, August 14, 2008 ª2008 Elsevier Inc.