Professional Documents
Culture Documents
https://doi.org/10.1038/s41586-023-06638-9 Joshua Hahn1,16, Aboozar Monavarfeshani2,16, Mu Qiao3,15, Allison H. Kao2, Yvonne Kölsch4,
Ayush Kumar1, Vincent P. Kunze5, Ashley M. Rasys6, Rose Richardson7,
Received: 30 March 2023
Joseph B. Wekselblatt8, Herwig Baier4, Robert J. Lucas7, Wei Li5, Markus Meister3,
Accepted: 13 September 2023 Joshua T. Trachtenberg9, Wenjun Yan2, Yi-Rong Peng10, Joshua R. Sanes2 ✉ &
Karthik Shekhar1,11,12,13,14 ✉
Published online: 13 December 2023
Open access
The basic plan of the retina is conserved across vertebrates, yet species differ profoundly
Check for updates
in their visual needs1. Retinal cell types may have evolved to accommodate these
varied needs, but this has not been systematically studied. Here we generated and
integrated single-cell transcriptomic atlases of the retina from 17 species: humans,
two non-human primates, four rodents, three ungulates, opossum, ferret, tree shrew,
a bird, a reptile, a teleost fish and a lamprey. We found high molecular conservation
of the six retinal cell classes (photoreceptors, horizontal cells, bipolar cells, amacrine
cells, retinal ganglion cells (RGCs) and Müller glia), with transcriptomic variation
across species related to evolutionary distance. Major subclasses were also conserved,
whereas variation among cell types within classes or subclasses was more pronounced.
However, an integrative analysis revealed that numerous cell types are shared across
species, based on conserved gene expression programmes that are likely to trace back
to an early ancestral vertebrate. The degree of variation among cell types increased
from the outer retina (photoreceptors) to the inner retina (RGCs), suggesting that
evolution acts preferentially to shape the retinal output. Finally, we identified rodent
orthologues of midget RGCs, which comprise more than 80% of RGCs in the human
retina, subserve high-acuity vision, and were previously believed to be restricted
to primates2. By contrast, the mouse orthologues have large receptive fields and
comprise around 2% of mouse RGCs. Projections of both primate and mouse
orthologous types are overrepresented in the thalamus, which supplies the primary
visual cortex. We suggest that midget RGCs are not primate innovations, but are
descendants of evolutionarily ancient types that decreased in size and increased in
number as primates evolved, thereby facilitating high visual acuity and increased
cortical processing of visual information.
The ability to assess gene conservation among species has been of enabled related activity focused on determining the extent to which cell
great value in multiple ways. It has revealed the evolutionary history types, the functional units of complex tissues6,7, are conserved among
of specific genes, highlighted crucial developmental and functional species. Analysing patterns of cell-type conservation across phylogeny
pathways, informed strategies for rational in vivo manipulations and can serve as a conceptual foundation for reconstructing the evolution
helped guide choices of animal models that mimic human diseases3,4. of cell types and identifying conserved developmental programmes8–10.
Comparative genomics was enabled by advances in DNA sequencing, as The neural retina, the portion of the brain that resides in the back of
well as statistical methodologies for sequence alignment and phyloge- the eye, is well-suited for this type of analysis. It is arguably as complex
netic inference5. Advances in high-throughput single-cell RNA sequenc- as any other part of the brain, but its compactness and accessibility
ing (scRNA-seq) and single-nucleus RNA sequencing (snRNA-seq) have facilitate detailed investigations of structure and function11. Moreover,
Department of Chemical and Biomolecular Engineering, University of California, Berkeley, Berkeley, CA, USA. 2Department of Cellular and Molecular Biology, Center for Brain Science, Harvard
1
University, Cambridge, MA, USA. 3Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, CA, USA. 4Max Planck Institute for Biological Intelligence,
Martinsried, Germany. 5Retinal Neurophysiology Section, National Eye Institute, National Institutes of Health, Bethesda, MD, USA. 6Department of Cellular Biology, University of Georgia, Athens,
GA, USA. 7Division of Neuroscience and Centre for Biological Timing, Faculty of Biology Medicine and Health, University of Manchester, Manchester, UK. 8Division of Chemistry and Chemical
Engineering, California Institute of Technology, Pasadena, CA, USA. 9Department of Neurobiology, David Geffen School of Medicine at UCLA, Los Angeles, CA, USA. 10Department of
Ophthalmology, Stein Eye Institute, UCLA David Geffen School of Medicine, Los Angeles, CA, USA. 11Helen Wills Neuroscience Institute,Vision Science Graduate Group, University of California,
Berkeley, Berkeley, CA, USA. 12Biological Systems and Engineering Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA. 13Center for Computational Biology, Biophysics Graduate
Group, University of California, Berkeley, Berkeley, CA, USA. 14California Institute of Quantitative Biosciences (QB3), University of California, Berkeley, Berkeley, CA, USA. 15Present address:
LinkedIn, Mountain View, CA, USA. 16These authors contributed equally: Joshua Hahn, Aboozar Monavarfeshani. ✉e-mail: sanesj@mcb.harvard.edu; kshekhar@berkeley.edu
400
ONL
300
OPL
200
AC HC
MG BC INL
100
IPL 0
et l t
an ue w se ys us irre eep ow Pi
g re m rd ck sh rey
AC GCL um aq os hre ou om ysc C r
ss
u
iza hi afi mp
c m s q u Sh Fe L C r
H a ar e M bd
o m S po Ze
b
La
M M Tre a r O
RGC Rh Pe
c CHX10 AP2A RBPMS DAPI
Squirrel Mouse Peromyscus Tree shrew Human Sheep Cow Pig
ONL
INL
GCL
Fig. 1 | Conserved retinal structure across vertebrates. a, Cartoon of a inner plexiform layer (IPL). b, Phylogeny of the 17 vertebrate species analysed
section through a vertebrate retina showing the arrangement of its six major in this work. The scale bar on the right indicates estimated divergence time.
cell classes: photoreceptors (including rods (r) and cones (c)), horizontal cells c, Sections from retinas of eight species immunostained for RBPMS (a pan-RGC
(HC), bipolar cells (BC), amacrine cells (AC), retinal ganglion cells (RGC) and marker), CHX10 (also known as VSX2) (a pan-bipolar cell marker) and AP2A
Müller glia (MG). The outer segments of rods and cones (OS), outer nuclear (also known as TFAP2A) (a pan-amacrine cell marker) and stained with the nuclear
layer (ONL), inner nuclear layer (INL) and ganglion cell layer (GCL)—which stain DAPI. Scale bars, 25 µm. Figures are representative of images from three
contain cell somata—are indicated, as are the outer (synaptic) layer (OPL) and retinas.
unlike other brain regions (for example, the cerebral cortex), the basic similarities in gene expression. This principle extends to identified
structural blueprint of the retina is highly conserved among verte- subclasses of photoreceptors, bipolar cells and amacrine cells. Tran-
brates1. The retina contains five neuronal classes—photoreceptors, scription factors implicated in cell and subclass specification are also
horizontal cells, bipolar cells, amacrine cells and retinal ganglion cells evolutionarily conserved, pointing to common programmes of retinal
(RGCs)—and a resident glial class called Müller glia12. The cell somata are development. Within each cell class, the transcriptomic variation across
arranged in three nuclear layers separated by two plexiform (synaptic) species increases with evolutionary time in a manner incompatible with
layers (Fig. 1a) with information flowing through them in a defined purely ‘neutral’ evolution18. Second, we assessed the extent of evolu-
direction: photoreceptors in the outer nuclear layer sense light and tionary variation among cell types within photoreceptors, horizontal
transmit visually evoked signals to interneurons in the inner nuclear cells, bipolar cells and RGCs, which have been comprehensively classi-
layer; the interneurons (horizontal cells, bipolar cells and amacrine fied in mice19–21 and primates22–24. We identify numerous evolutionarily
cells) process the information and supply it to RGCs in the innermost conserved types but find that variation is more extensive in RGCs than
layer; and the RGCs send axons through the optic nerve to visual cen- in other classes, suggesting that natural selection acts preferentially to
tres in the brain. Several of the neuronal classes can be subdivided shape the retinal output. Finally, we identify non-primate orthologues
into subclasses, and all classes comprise multiple types that differ in of midget RGCs, which account for more than 80% of RGCs in humans
morphology, physiology, connectivity and molecular composition6,11–14. and are primarily responsible for high-acuity vision. To our knowledge,
The specificity of connections between interneuronal and RGC types no counterparts of these cell types have previously been identified in
endows many RGC types with selective responsiveness to small subsets non-primates, precluding mechanistic analysis of blinding diseases
of visual features such as edges, directional motion and chromatic- involving RGC loss, such as glaucoma. This orthology suggests that
ity14,15. As a result of neural computations in the retina, the optic nerve rather than appearing de novo in primates, midget RGCs evolved from
transmits a set of parallel representations of the visual scene to the rest cell types that were present in the common mammalian ancestor.
of the brain for further processing16,17.
Despite these conserved features, vertebrate species differ greatly
in their visual needs1. Some species are diurnal, others are nocturnal; Retinal cell atlases of 17 species
some are terrestrial, others are aquatic; and some mainly hunt, whereas Previously, we used scRNA-seq and snRNA-seq to study retinal cell
others forage for colourful fruits. It is likely that variations in retinal types in five species: Mus musculus19,20,25,26 (hereafter referred to as
cell types across species emerged during the course of evolution to ‘mouse’), cynomolgus macaque22 (Macaca fascicularis), human23 (Homo
serve these diverse needs. However, the evolutionary relationships sapiens), chick27 (Gallus gallus) and zebrafish28 (Danio rerio). For the
among retinal cell types have not been mapped systematically. Here we present study, we generated atlases from 12 additional species: ferret
address this gap by using single-cell transcriptomics to compare retinal (Mustela putoriusfuro), brown anole lizard (Anolis sagrei), deer mouse
cell classes, subclasses and types in 17 vertebrate species (Fig. 1b,c). (Peromyscus maniculatus bairdii), tree shrew (Tupaia belangeri chinensis),
First, we show that the conserved functional and morphological pig (Sus domesticus), sheep (Ovis aries), cow (Bos taurus), opossum
character of the six cell classes is mirrored by marked cross-species (Monodelphis domestica), marmoset (Callithrix jacchus), 4-striped
0.25
0.50
0.75
1.00
0.25
0.50
0.75
0.25
0.50
0.75
1.00
1.00
0
Normalized expression Spearman correlation Spearman correlation
MG
MG
PR
HC
AC
PR
BC
RGC
HC
Peromyscus
Rhabdomys
Tree shrew
BC
Marmoset
C
AC
G
PR
C
Opossum
Zebrafish
Macaque
H
M
G
Squirrel
Human
R
Mouse
Sheep
Lizard
Ferret
Chick
Cow
Pig
AC
d Cones Glycinergic ACs OFF BCs
10 RGC
BC AC
5
PR ARR3 SLC6A9 GRIK1
UMAP2
0
Rods GABAergic ACs ON BCs
RGC HC
−5
MG
−10 BC
RBPMS
RBPMS2
SLC17A6
THY1
NEFL
NEFM
POU4F1
VSX1
VSX2
OTX2
GRM6
TRPM1
CABP5
GRIK1
TFAP2A
TFAP2B
TFAP2C
GAD1
GAD2
PAX6
ROBO1
SLC6A9
ONECUT1
ONECUT2
LHX1
CALB1
VIM
SLC4A3
WDR72
SAG
PDC
GNGT1
GNGT2
NRL
SLC1A3
RLBP1
GLUL
SFRP2
RCVRN
APOE
0.4
divergence
0.2
0
0 0.4 0.8 1.2 0 0.4 0.8 1.2 0 0.4 0.8 1.2 0 0.4 0.8 1.2 0 0.4 0.8 1.2 0 0.4 0.8 1.2
Evolutionary divergence (substitutions per 100 bp)
Fig. 2 | Class- and subclass-specific transcriptomic signatures. a, Heat map indicating class identity (left) or expression levels of subclass-specific markers
showing average expression of marker genes (columns) within each major (right). GAD1, a marker for GABAergic amacrine cells, is also expressed by some
cell class in 17 species (rows). Rows are grouped by cell class (left). Within each horizontal cells, and ISL1, a marker for ON bipolar cells, is also expressed by
class, species are ordered as in Fig. 1b. Grey tiles indicate data that are missing some RGCs, horizontal cells, and amacrine cells. Details of gene expression
owing to the absence of the corresponding orthologue in the species annotation. by species are shown in Extended Data Fig. 8d. e, Pairwise mean squared
Colours indicating cell class are uniform in a–e. PR, photoreceptor. b, Cross- divergence of class-specific pseudobulk gene expression profiles between
correlation matrix (Spearman) of pseudobulk transcriptomic profiles for species (y axis) increases with evolutionary distance, as estimated by
the 16 jawed vertebrates. Rows and columns are grouped by class, and then substitutions per 100 bp (x axis). Data from mammals, chicken and lizard are
ordered by phylogeny within a class. A total of 4,560 1:1 gene orthologues were included. Data including zebrafish are presented in Extended Data Fig. 7e.
used to calculate the correlation values. c, As in b, with rows and columns Solid lines represent power law (y = ax b) regression fits. Across the graphs,
grouped by species instead of class. Matrices including lamprey are shown a ∈ [0.33, 0.47] and b ∈ [0.23, 0.35]. The coefficient of determination (R 2) values
in Extended Data Fig. 7c,d. d, Left, uniform manifold approximation and range from 0.75 to 0.92.
projection (UMAP) embedding of integrated cross-species data, with points
grass mouse (Rhabdomys pumilio), 13-lined ground squirrel (Ictidomys cells) were not analysed further. Biological replicates within each collec-
tridecemlineatus) and sea lamprey (Petromyzon marinus) (Fig. 1b,c). We tion exhibited a high degree of concordance (Extended Data Figs. 3–6).
also profiled around 185,000 nuclei from 18 human donors, thereby The numbers of cells in each class for each species are summarized in
allowing us to identify over 30 more cell types than had been detected Supplementary Table 1.
in the dataset analysed previously23, including 10 additional RGC types
(Extended Data Fig. 1). To obtain sufficient numbers of bipolar cells and
RGCs for comprehensive analysis, we enriched these classes in some Molecular conservation of neuronal classes
collections (Extended Data Figs. 2–6 and Methods). We also collected We analysed the expression of class markers that have been validated
cells without enrichment to ensure representation of all classes. in mice and primates; that is, genes that are co-expressed within a
We used a standardized computational pipeline to normalize, correct retinal cell class but exhibit little or no expression in other retinal
batch effects, reduce dimensionality and cluster the data from each cell classes19,20,22–26. Many showed similar expression patterns in
species separately29 (Methods). Cells that did not belong to the six other vertebrates (Fig. 2a). Using these markers, we assigned cells
canonical classes named above (for example, microglia or endothelial within each species to one of the six classes. We then assessed the
Expression
Marmoset
Mouse 4
Opossum 2
Peromyscus
0
Pig
Rhabdomys
Sheep
Squirrel
Tree shrew
g
ow
k
rd
p
an
se
el
ue
sh
um
t
ys
re
e
c
Pi
cu
ee
rr
re
os
za
hi
ou
afi
um
aq
r
C
ui
ss
Fe
Sh
ys
sh
C
do
Li
m
br
M
Sq
ac
po
H
m
ar
ee
Ze
ab
M
ro
O
M
Tr
Rh
Pe
e Orthotype differentially expressed genes f Orthotypes corresponding to rod BCs and BC1B
oBC1A No. of species
oBC1A
oBC1B 0 oBC1B
oBC2 5 oBC2
oBC3A 9 oBC3A Mapping
oBC3B oBC3B (%)
13
BC orthotype
BC orthotype
oBC4 oBC4 0
oBC5A Average oBC5A 20
oBC5B expression oBC5B 40
oBC5C 3 oBC5C >60
oBC5D 2 oBC5D
oBC6 oBC6
oBC7 1 oBC7
oBC8/9 0 oBC8/9
oRBC –1 oRBC
Pe 6
Tr 6
Sq e_8
Sh 13
Rh e_5
P 5
M po _8
M _BC 3
H c_O B
_O Fx
C ow X
P− 2
Tr 11
Sq 15
Sh _9
Rh e_1
Pi 2
Pe 1
M Opo 1
_R 1
M BC
M _RB
um B
Fe B
C _1
_1
LR CA
4
5D
PT 5
KI RM
PL L3
U 2
PT A
AD 1A
ER 2
N 4
PC H2
N 10
3
M K
1
r_
_B 1
1
g_
r_
ou _
C F
1
H _R
_R
M
Y
BB
IN
2
PR
O ig
r
ow
um F
z_
u_
a_
ou _
hi _
e_
a_
F
E
DH
SO
C
SH
AN
C
XD
XP
K
RT
KA
P
RR
ar
ac
Li
N
PR
g
Species PR types to orthotypes Mapping Mapping
o_Rod (%) h Species HC types to orthotypes (%)
o_sCone 0 30
25 oHC1 40
o_mlCone oHC2 50
50
75 60
M _R d
ar od
M e_R d
Rh _R d
a od
Sq r_R d
Sh _R d
C e_R d
d
g d
r d
um o_ d
ar Co d
Tr _sC ne
ou C e
Rh _sCone
Sq _s one
Sh _sCone
ow C e
Pi _sC ne
ou lC ne
a_ lC e
lC e
u lC e
l e
Fe _sC ne
H po Co e
um _s ne
ac lC e
ar lC e
l e
ow l ne
g_ lC e
O r_m Cone
po l e
lC ne
e
ac 1
ar 1
Tr _H1
Sq_H1
Sh u_H
ow 1
Pi _H1
po 1
M _H1
Rh _H
Pe _H
Pe _Ha
ac b
ar 2
2
O w_ 2
O _H 2
2B
_H A
M e_s on
C e_s on
O _s on
M _m Con
M _m on
Tr _m on
Rh _m on
Pe m on
Sq r_m on
Sh _m on
Pi _m on
Fe ml on
on
M _H
M _H
C e_H
O _H
M _H
Sh _H
C e_H
po H
ac o
Tr _Ro
ou o
Pe _Ro
u o
o
ow o
Pi _Ro
Fe _Ro
H Op _Ro
M _s Ro
M r_H
po 2
100 70
o
M e_m Co
_m Co
g o
C e_m Co
M _R
ou
a
u C
r
um
g
um
o
H
a
r
H
Fig. 3 | Multispecies integration of bipolar cells. a, UMAP of mammalian orthotype. Mapping that includes non-mammalian orthotype is shown in
bipolar cells computed with the raw (left) and integrated (right) gene Extended Data Fig. 9e. e, Dot plot showing differentially expressed genes
expression matrices. Cells are coloured by species of origin. b, Feature plots within each bipolar cell orthotype. The size of the dot represents the number
showing expression within the integrated space of the rod bipolar cell marker of mammalian species (out of 13 mammalian species in total) that express the
PRKCA, the ON bipolar cell marker ISL1, and the OFF bipolar cell marker GRIK1. gene in at least 30% of cells in the corresponding orthotype, and the colour
c, As in a, but with cells coloured by orthotype identity. d, Left, confusion represents normalized expression level. f, Confusion matrix showing the
matrix showing the percentage of cells from each mouse bipolar cell-type species bipolar cell clusters (columns) that map specifically to the orthotype
mapping to each mammalian bipolar cell orthotype. Each column sums to oRBC and oBC1B. Bipolar cell types are named on the basis of their species of
100%. See Extended Data Fig. 9a for a higher magnification view. Centre, origin and within-species bipolar cell cluster ID (Extended Data Figs. 1 and 3–6);
confusion matrix showing specific mapping between mammalian bipolar cell for example, Peromyscus bipolar cell cluster 1 is called ‘Per_1’. g,h, Confusion
orthotypes and bipolar cell clusters within each mammalian species (Extended matrix showing mapping of mammalian photoreceptor (g) and horizontal cell
Data Figs. 1 and 3–6). Right, confusion matrices showing the mapping of bipolar (h) types to orthotype. Format as in d, centre.
cell clusters in lizard, chick and zebrafish to the mammalian bipolar cell
We repeated the orthotype analysis for photoreceptors and horizon- only a single horizontal cell type. Again, orthotype analysis separated
tal cells, which are less diverse classes than bipolar cells. As noted above, horizontal cells into two groups (Fig. 3h). Many non-mammalian ver-
photoreceptors are divided into two subclasses, rods and cones. Most tebrates are more complex in these respects, with 4 or 6 photorecep-
mammals have a single rod type and two cone types, tuned to respond tor types and 4 horizontal cell types in birds (including chicken) and
best to short wavelengths (S cones, also known as blue cones) and fish27,34,35 (including zebrafish); these species mapped less well onto
medium wavelengths (M cones, also known as green cones), respec- the mammalian orthotypes.
tively. However, many primates have a third cone type (L cones, also
known as red cones) that is sensitive to longer wavelengths34. Orthotype
analysis separated mammalian M and L cones from S cones effectively, Retinal ganglion cell orthotypes
with the few exceptions probably being due to insufficient cell num- We next performed orthotype analysis on RGCs, the only output neu-
bers (Fig. 3g). Similarly, most mammals have two horizontal cell types, rons in the retina. We identified 21 RGC orthotypes in mammals and
called H1 and H2—although mice and perhaps other rodents—have found differentially expressed genes that distinguished them (Fig. 4a–c
Orthotypes
oRGC5 oRGC10
Mouse oRGC11
Opossum Average
oRGC7oRGC16 oRGC12 expression
Peromyscus oRGC13
Pig 3
oRGC1 oRGC14
Rhabdomys oRGC2 oRGC15 2
Sheep oRGC4 oRGC16
Squirrel oRGC17 1
Tree shrew oRGC9 oRGC18
oRGC8 oRGC19 0
oRGC20 −1
oRGC21
G A2
G L2
13 9
TB K3
0
M X6
N IC
P T4
AP 1
1
RX 3
RG
2C
C 6
R 1
DH 3
N 7
RU IB
SG 1
PL CD
C 6
EM ST
DH
SO H1
PC CS
1
X2
1R X
LP
X
T
DF
DS
SY
IR
IL RO
IR
XN
RI
U
TM CH
D
d
oRGC1
oRGC2
oRGC3
oRGC4
oRGC5
oRGC6
oRGC7
Mapping
oRGC8
(%)
oRGC9
RGC orthotype
0
oRGC10 25
oRGC11 50
oRGC12 75
oRGC13 100
oRGC14
oRGC15
oRGC16
oRGC17
oRGC18
oRGC19
oRGC20
oRGC21
F P F P F P s l t
w se ys cu rre ee
p
Pi
g
rre um rd ck sh
re ou m ui ss za hi afi
an e et sh do ys Sh Fe Li C br
u e M m Sq po
um aq os re ha
b ro O Ze
H ac ar
m T R Pe
M M
oRGC12
oRGC7 Average
oRGC3 expression
oRGC5 3
oRGC6 2
oRGC16 1
oRGC1 0
oRGC11 −1
oRGC2
oRGC4
oRGC9
oRGC8
8 a 2 7 4 ) 8 7 7 2 1 l 5 6 9 2 3 4 l 9 0 9 5 5 4 6 0 4 1 1 1 2
C _MX M1 1b _M C C4 NS C1 C2 C3 NT _S _S FFS GC ove C1 C2 C2 3D 3D C3 FFT OFF OFF ove iON iON 3B C1 C2 C3 C3 C2 C_N C1 C3 _DV C1 C2 3L 3D C1 3L
ES
TB 3
M R1
FO B
N SA 2
RO 2
AP 2
BN D
M O C C 2
X
XP
EU TB
TF D
R n i i C
d n W
AF
22 3_ 0_ 31 O aO in id C_n mi mi 6_ C
ha G G haO 5_J C_ W W _W
IR
_W 0_W SG
M
C C3 C4 C a h G 1_ _ 13
EO
h l p R R 3 p m m F F C D S C C2
lp A _T _T lp C RG 2 3 l F F G _ _ o D
(A
4 1_ 21 17 2_A _ T C C
5 _A C4_ 28_ _FR 38 C3 _o oo
C
3 C C C C4 9 4 C 32 C 1 2 _
C
4 C C C C 16
C
Fig. 4 | Multispecies integration of retinal ganglion cells. a, Integrated primates. Right, confusion matrices showing the mapping of RGC clusters
UMAP of RGCs from 12 mammals (cow was excluded owing to the paucity of (columns) in lizard, chick and zebrafish to the 21 mammalian RGC orthotypes.
RGC data.). Cells are labelled by species of origin. For primates, cells from fovea Mapping to the single non-mammalian RGC orthotype is shown in Extended
and periphery are plotted separately. b, As in a, with RGCs labelled by orthotype. Data Fig. 10d. e, Left, confusion matrix showing that mouse RGC types (rows;
c, Dot plot showing differentially expressed genes within each RGC orthotype. naming as in ref. 20) belonging to transcription factor-based subsets39
Representation as in Fig. 3e. d, Left, confusion matrices showing that species- (colours) map to the same orthotypes (columns). Right, dot plot showing
specific RGC clusters (Extended Data Figs. 1 and 3–6) map to mammalian RGC specific expression patterns of subclass-specific transcription factor-
orthotypes in a specific fashion. Representation as in Fig. 3d, centre, except encoding genes39 in orthotypes. Representation as in Fig. 3e.
that clusters from fovea (F) and periphery (P) are mapped separately for
and Extended Data Fig. 10a). Eighty-one per cent of mammalian RGC by non-mammalian species (Extended Data Fig. 10b–d and Supple-
clusters (329 out of 408) mapped predominantly to a single orthotype mentary Table 3).
(Fig. 4d). In species that contain more RGC types than orthotypes, tran- To test the reliability of orthotype analysis for RGCs, we searched
scriptomically similar RGC clusters mapped to the same orthotype. As for orthologues of an evolutionarily ancient set of RGC types called
was the case for bipolar cells, RGC orthotypes remained stable when intrinsically photosensitive RGCs (ipRGCs). ipRGCs contain the pho-
lizard, chick and zebrafish were included in the integration (Fig. 4d, topigment melanopsin (encoded by OPN4), which enables them to
right), but were supplemented by an additional orthotype dominated generate visually evoked signals without input from photoreceptors36.
oRGC11 20
oRGC12 40
>60
oRGC13
oRGC14
oRGC15
oRGC16
oRGC17
oRGC18
oRGC19
oRGC20
oRGC21
Mac_MGC_OFF
Mar_MGC_OFF
MarFov_fRGC12
Mou_C17_TRGC_S1
Mou_C42_AlphaOFFS
Fer_14
Fer_20
Opo_2
Per_18
Per_23
Pig_1
Pig_17
Rha_26
She_2
Squ_27
Tre_1
Tre_28
Hum_MGC_ON
Mac_MGC_ON
Mar_MGC_ON
Mou_C43_AlphaONS_M4
Fer_19
Fer_9
Opo_28
Per_34
Pig_4
Tre_3
Hum_PGC_OFF
HumFov_RGC4
HumPer_RGC4
Mac_PGC_OFF
Mar_PGC_OFF
Mou_C34
Mou_C45_AlphaOFFT
Opo_22
Opo_27
Per_36
Per_38
Rha_30
Squ_30
Tre_34
Hum_PGC_ON
HumFov_RGC10
HumFov_RGC21
HumPer_RGC10
HumPer_RGC20
HumPer_RGC21
Mac_PGC_ON
MacFov_fRGC7
MacPer_pRGC9
Mar_PGC_ON
MarFov_fRGC7
Mou_C18
Mou_C27
Mou_C37
Mou_C41_AlphaONT
Fer_22
Per_14
Pig_14
Pig_22
Pig_25
Pig_34
Rha_24
Rha_25
Rha_33
She_9
Squ_31
Tre_17
Tre_21
Tre_32
Tre_33
Hum_MGC_OFF
500
250
GCL 0 Sustained
–250
OFF C42 ON C43 OFF C45 ON C41 –600
RGCs –400 –500
MGC αOFF MGC αON PGC αOFF PGC αON –200 –750
OFF 0
OTs oRGC1 oRGC4 oRGC5 oRGC2 200 –1,000
400 –1,250 Transient
Polarity OFF ON OFF ON 600
ON
Kinetics Sustained Transient
d e
0.5
No. of mouse combinations
20 Mouse
0.4
Macaque
Proportion
0.3
10 0.2
Orthotype
0.1
oRGC1 (MGC OFF)
oRGC4 (MGC ON)
0 0
0.45 0.50 0.55 0.60 0.65
ys
Sq e
m el
po t
ee m
m ig
tP
a F
aq P
H eF
um P
F
O rre
cu
s
ee
ro irr
re
P
Tr ssu
M set
an
M que
an
m
ou
M ose
sh
u
Sh
do
um
M
o
m
ab
ac
ac
H
ar
ar
Rh
Pe
Fig. 5 | Mammalian orthologues of midget and parasol RGCs. a, Confusion species, polarity and kinetics (see Supplementary Note 2). d, Matching MGCs
matrix showing RGC clusters from different species that map specifically to and PGCs to mouse types by GAGE. Inset, given sets of mouse and primate
oRGC1, oRGC4, oRGC5 and oRGC2, which contain OFF and ON midget RGCs RGC types, the model fits the arrangement of their cluster centroids in gene
(MGCs) and OFF and ON parasol RGCs (PGCs). Representation as in Fig. 3f. expression space by assuming a shape that is simply shifted to the other species
Column names corresponding to primate midget and parasol types are shown via a linear translation. Symbols mark the four response types: circle, sustained;
in red, and mouse α-RGC types are shown in blue. b, Schematic delineating square, transient; open, ON; filled, OFF. The graph is a histogram of the fraction
morphological and physiological similarities between primate and midget of explained variance showing for each proposed combination of four mouse
RGCs and their α-RGC orthologues. Orthotypes (OTs) of each pair as well as cell types how well the resulting shape fits the macaque RGC geometry. The red
the orthology among bipolar cell types that innervate them are also shown. bar shows the set of four α-RGC types. Green bars show combinations containing
Morphologies of neuronal types were created on the basis of published data three α-RGC types. Grey bars, remaining sets of four mouse cell types as shown
(Supplementary Note 1). Within each pair, the left column corresponds to in Supplementary Table 4. e, Relative proportion of OFF and ON midget RGC
primate types and the right column corresponds to mouse types. c, FLDA orthologues in mammalian species based on frequencies of cells in oRGC1 and
projection of the scRNA-seq data for primate midget and parasol types and oRGC4.
mouse α-RGC types onto the corresponding 3D space, with axes representing
matches the macaque shape (Methods). The four α-RGC types pro- transcriptomically defined mouse RGC types20. The four α-RGC types
duced the strongest match by a large margin, followed by several com- with the correct matching of polarity and kinetics with the MGCs and
binations containing three α-RGC types (Fig. 5d). Finally, we considered PGCs scored in second place out of all such combinations. The top
matches for all 3,575,880 possible combinations of 4 drawn from the 45 match was biologically implausible (see Extended Data Fig. 12e).
−1
25
50
75
0
1
2
0.0
0.1
0.2
0.3
0
c f
MG
Cone
Rod
HC
GlyAC
GabaAC
MGC_OFF
DB5 0.8
RB
DB4
MGC_ON
BBGB
IMB
DB6
d RGC10
DB3b
DB2
RGC21
RGC20 DB3a
FMB
RGC15
OFFX
PGC_ON
RGC13 DB1
RGC18 0.3
RGC6 RGC8
RGC14
RGC19 RGC7
RGC11 RGC10
RGC5
RGC9
RGC8
RGC4
PGC_OFF RGC4 RGC17
RGC16 RGC12 RGC7 PGC_OFF
RGC9 PGC_ON
e
MGC_OFF
MGC_ON
DB3a
RGC14
RGC13
DB1
DB2 RGC21
OFFX
DB6 RGC18
RGC15
RB RGC6
FMB DB3b
RGC5
RGC17
DB5 IMB BBGB RGC16
RGC20
DB4 RGC12
RGC11
RGC19
mac_neun+_563
mac_neun+_8691
mac_all_218
peri_all_218
mac_all_220&235
peri_all_220
mac_neun+_235
mac_neun+_355_s2
peri_all_235
mac_neun+_887
mac_neun+_591
mac_neun+_218
mac_neun+_355_s1
peri_neun+_1059&564
peri_neun+_563
mac_neun+_220
peri_neun+_887&869
mac_all_822
peri_neun+_326&355
mac_neun+_326
mac_neun+_056
mac_neun+_978&194&150
mac_neun+_1005&1053&1017
0.6
T H A6
S A 1
O LC D2
U 9
L T1
AL 1
T P B1
SA 3
P DG
R C
P D HO
C H
C 3
LB 3
AP P 1
S L BP 2
G X2
VS 6
VSY1
O X2
AB 1
T F R IK5
TF P2 1
T FAP 2A
AP B
G 2C
A RX
E
C MS
G D
EC A
R 1A
M
C HX
SL RR
R MS
C X
G P
O
E6
N 6
17
A
R
A
BP
R
Features
NEUN
FSC
BCs
10µm
DAPI CHX10
b Human: Macula
d
BCs
10µm
DAPI CHX10
e
NEUN CHX10 DAPI
ONL
INL
GCL
Extended Data Fig. 2 | Nuclear enrichment strategies for retinal ganglion in ~90% yield for RGCs; BCs were not analyzed in this experiment. c. Brightfield
cells (RGCs) and bipolar cells (BCs). a. Examples of gating strategy in image showing the morphology and integrity of FACS-purified nuclei. d.
fluorescent activated cell sorting (FACS) experiments for collecting single Confocal image of DAPI stained FACS-purified nuclei. e. Retinal sections from
nuclei labeled with either PE-conjugated NEUN, which enriches RGCs, or six species show that PE-conjugated NEUN (red) and APC-conjugated CHX10/
APC-conjugated CHX10 (also known as VSX2), which enriches BCs. Data shown VSX2 labels RGCs and BCs, respectively. Retinal sections were co-stained for
are representative from experiments in the pig retina. NEUN and CHX10-based DAPI (blue) to visualize nuclei. Scale bar, 50 μm. Images in panels a–e
enrichment resulted in ~90% yield for RGCs and ~95% yield for BCs. b. Same as representative of n ≥ 3 experiments.
panel a, for human macular retina samples. NEUN-based enrichment resulted
Article
Pct Exp Avg Exp Proportion
Pct Exp Avg Exp Proportion
a b
0.6
0.4
0.2
−1
0.0
25
50
75
Tree shrew
100
0
1
2
Sheep
25
50
75
0.0
0.2
0.4
0.6
0.8
−1
0
0
1
2
MG
Cone MG
Rod
1 Cone
HC
GlyAC Rod
GabaAC
BC_6 0.8
HC
BC_3 GlyAC
BC_7
BC_10 GabaAC
BC_11 BC_10 0.9
BC_9
BC_13 BC_6
BC_8 BC_14
BC_4
BC_2 BC_2
BC_12
BC_1
BC_5
BC_1 BC_12
RGC_29 0.3
BC_5
RGC_3
RGC_20 BC_8
RGC_19
BC_9
RGC_15
RGC_6 BC_3
RGC_16
BC_4
RGC_39
RGC_37 BC_11
RGC_13
BC_7
RGC_1
Identity
RGC_35 BC_13
RGC_38 0.8
RGC_24 RGC_8
RGC_30 RGC_12
RGC_27
RGC_28 RGC_11
RGC_32 RGC_19
RGC_7
RGC_26 RGC_5
RGC_18 RGC_4
RGC_2
RGC_5 RGC_15
RGC_36 RGC_16
RGC_9
RGC_33 RGC_9
RGC_34 RGC_14
RGC_17
RGC_21 RGC_3
RGC_31 RGC_13
RGC_12
RGC_14 RGC_6
RGC_23 RGC_10
RGC_10
RGC_22 RGC_7
RGC_4 RGC_17
RGC_8
RGC_11 RGC_1
RGC_25
RGC_18
chx10+_s1
chx10+_s2
neun-_chx10-_s1
neun-_chx10-_s2
chx10+_s3
neun+_s2
neun+_s3
neun+_s1
0.1
RGC_2
1
S L −1
PO S2
Y1
X1
TF X2
T F 2A
2B
2− 1
N A9
EC 1
T2
PD G
H
R 3
P1
BP S
X
AD AD
4F
O UT
R
M
E6
R
SA
U
AS
M
TH
LB
VS
T
AP
AP
AR
C
BP
C
O
EC
G
0.2
R
chx10+_s1
chx10+_s2
neun+_s2
neun+_s5
neun+_s6
chx10+_s3
neun+_s3
neun+_s4
N
R
T HA6
S L BP S 2
N Y1
N FL
V SF M
O X2
G X2
V M6
A 1
G BP 5
T FAP 2 1
T F AP A
AP 2B
G 2C
S A 1
O LC D2
U 9
L T1
AL 1
T P B1
SA 3
P DG
C
E 6O
C H
C 3
LB 3
AP P 1
C MS
ARR X
E
O
T F R IK
G D
EC A
R 1A
C SX
C HX
SL R
O
P DR H
G
R M
17
N 6
T
A
R
E
Features
BP
R Features
0.25
0.50
0.75
100
Cow Pig
25
50
75
−1
100
0
0
1
2
0.0
0.2
0.4
0.6
0.8
25
50
75
−1
0
0
1
2
MG MG
Cone
Rod
Cone
HC 0.8
GlyAC
Rod GabaAC
BC_18 0.8
BC_7
HC BC_6
BC_3
GlyAC BC_11
BC_16
BC_10
GabaAC BC_5
0.9 BC_4
BC_13
BC_13
BC_12
BC_9
BC_15 BC_14
BC_17
BC_8
BC_14 BC_15
BC_2
BC_6 BC_1
RGC_36 0.6
RGC_5
BC_3 RGC_11
RGC_10
RGC_29
BC_8
RGC_16
RGC_6
BC_5 RGC_21
RGC_7
RGC_13
BC_16 RGC_31
RGC_23
BC_7 RGC_15
RGC_18
RGC_3
BC_9 RGC_26
RGC_37
RGC_14
BC_4
RGC_4
RGC_1
BC_11 RGC_17
RGC_2
RGC_22
BC_2 RGC_12
RGC_8
BC_1 RGC_9
RGC_20
RGC_30
BC_10 RGC_24
RGC_19
RGC_28
BC_12
RGC_27
0.9 RGC_32
RGC RGC_35
RGC_34
RGC_33
neun+_s1
neun+_s6
RGC_25
chx10+_s1
chx10+_s2
all_s2
neun+_s2
neun+_s3
neun+_s4
chx10_s3
neun+_s5
T HA6
S A 1
O LC D2
U 9
L T1
AL 1
T P B1
SA 3
P DG
C
E 6O
C H
C 3
LB 3
AP P 1
S L BP S 2
N Y1
N FL
V SF M
O X2
G X2
V M6
A 1
G BP 5
T FAP 2 1
T F AP A
AP 2B
G 2C
C MS
ARR X
E
T F R IK
G D
EC A
R 1A
C SX
C HX
SL R
O
PD H
R M
17
N 6
T
A
R
E
neun-chx10-_s2
neun-chx10+_s3
neun+chx10-_s3
neun-chx10+_s2
neun+chx10-_s2
neun-chx10+_s4
neun+chx10-_s4
cd73-_s1a
cd73-_s1b
cd90+_s1a
cd90+_s1b
0.2
BP
T F 2A
2B
1
SL 2
EC 9
N T1
T2
R 3
P1
S2
Y1
C 2
G 5
T F IK1
BP S
E
R
AD
AD
O C6A
X
R
M
O
PD
SA
H
O U
U
AB
LB
M
TH
VS
AP
AP
AR
AP
Features
BP
R
EC
G
G
R
N
R
Features
100
Peromyscus Ferret
100
25
50
75
0.0
0.2
0.4
0.6
−1
0.0
0.2
0.4
0.6
25
50
75
0
−1
0
1
2
0
1
2
MG
MG
Cone
Rod Cone
HC 0.6 Rod
GlyAC 0.9
HC
GabaAC
GlyAC
BC_9 0.7
BC_11 GabaAC
BC_5 BC_7 0.7
BC_8 BC_4
BC_7
BC_3
BC_3
BC_2 BC_2
BC_6 BC_12
BC_10 BC_1
BC_1
BC_10
BC_4
RGC_17 0.5 BC_11
RGC_24 BC_8
RGC_27 BC_5
RGC_13
BC_9
RGC_9
RGC_8 BC_6
RGC_26 0.8
RGC_2
RGC_5 RGC_13
RGC_6
RGC_23
RGC_32
RGC_21 RGC_24
RGC_22 RGC_19
RGC_19 RGC_17
RGC_18
RGC_14
RGC_20
RGC_16 RGC_10
RGC_7 RGC_12
RGC_14 RGC_8
RGC_37
RGC_16
RGC_12
RGC_34 RGC_18
RGC_36 RGC_1
RGC_38 RGC_15
RGC_23 RGC_21
RGC_29
RGC_33 RGC_3
RGC_30 RGC_6
RGC_10 RGC_25
RGC_35 RGC_22
RGC_11
RGC_31 RGC_9
RGC_26 RGC_20
RGC_28 RGC_2
RGC_4 RGC_11
RGC_15
RGC_3 RGC_4
RGC_1 RGC_7
RGC_25 RGC_5
0.1 0.1
all_s1
all-s2a
all-s2b
cd73-_s2a
cd73-_s2b
cd90+_s2a
cd90+_s2b
cd90+_s1
SL 2
S2
Y1
X2
G 2
T F IK1
T F 2A
2B
EC 9
N T1
T2
PD O
H
R 3
P1
BP S
cd73-_s1a
cd73-_s1b
cd73-_s1c
cd73-_s2a
cd73-_s2b
cd73-_s3
cd90+_s1a
cd90+_s1b
TF P5
T F 2A
G B
G 1
SL 2
S2
Y1
X2
C 2
N A9
N T1
T2
G
O
C
P1
AD
AD
O C6A
BP S
R X
E
TX
R
M
E6
R
SA
AD
AD
TX
R
O U
O
M
TH
LB
PD
R
SA
VS
AP
AP
H
AR
C
BP
O CU
U
AB
M
TH
O C6
LB
VS
AP
AP
O
AR
EC
C
G
AP
BP
R
O
EC
R
E
N
R
R
Features Features
Pct Exp Avg Exp Proportion Pct Exp Avg Exp Proportion
0.0
0.2
0.4
0.6
100
0.0
0.1
0.2
0.3
c d
−1
25
50
75
Opossum Lizard
25
50
75
−1
0
1
2
0
0
0
1
2
MG MG
Cone Cone
Rod
Rod
HC
GlyAC HC
GabaAC GlyAC
BC_13 0.8 GabaAC
BC_5 BC_16 0.9
BC_8
BC_5
BC_14
BC_8
BC_7
BC_10 BC_6
BC_12 BC_15
BC_11 BC_4
BC_2 BC_10
BC_15
BC_7
BC_17
BC_9 BC_11
BC_6 BC_12
BC_3 BC_1
BC_1 BC_14
BC_4
BC_13
BC_16
BC_3
RGC_30 0.5
RGC_28 BC_9
RGC_31 BC_2
Identity
RGC_29 BC_0
RGC_23 RGC_23 0.2
RGC_20
RGC_19
RGC_12
RGC_5 RGC_22
RGC_15 RGC_21
RGC_13 RGC_11
RGC_17 RGC_20
RGC_18 RGC_15
RGC_6
RGC_4
RGC_10
RGC_3 RGC_5
RGC_26 RGC_16
RGC_24 RGC_18
RGC_19 RGC_14
RGC_11
RGC_6
RGC_27
RGC_14 RGC_7
RGC_8 RGC_9
RGC_2 RGC_1
RGC_25 RGC_17
RGC_7 RGC_12
RGC_1
RGC_13
RGC_16
RGC_21 RGC_10
RGC_4 RGC_2
RGC_22 RGC_8
RGC_9 RGC_3
neun-_s1
neun-_s3a
neun-_s3b
neun-_s4a
neun-_s4b
neun+_s1
neun+_s5a
neun-_s2
neun+_s5b
neun+_s3
neun+_s4
0.1 0.2
TH 6
VS 1
X2
G 2
VS 6
G 1
T F IK1
T F P 2A
T F P 2B
G C
G 1
S L D2
EC 9
C T1
B1
C
P D HO
AR H
SL R3
R A3
AP 1
E
A
AD
N A
Y
TX
U 6
VS 2
C X2
G 5
O 1
PR X2
T F CA
G A
S L D2
LH 9
C X1
TP 1
3
G
C
P D HO
AR H
SL R3
R A3
AP 1
O
C S
E
2
PD
E6
neun-_s1
neun-_s2
neun+_s1
neun+_s2a
neun+_s2b
all_s3
U
P O 7A
IK
6A
B
M
17
O C6
AL
1
LB
6F
P
A
SL PM
O
AP
PD
E6
SA
R
C
O
AB
AL
1
LB
A
A
AP
A
K
R
R
C
C
B
SL
Features Features
Extended Data Fig. 4 | Summary of cell type atlases for Peromyscus, ferret, 49,972 cells) (b), opossum (n = 5 animals; 76,763 nuclei) (c), and brown anole
opossum, and brown anole lizard. Panels a-d depict the atlases (as in Extended lizard (n = 3 animals; 42,848 nuclei) (d).
Data Fig. 3) for peromyscus (n = 3 animals; 44,223 cells) (a), ferret (n = 2 animals;
Pct Exp Avg Exp Proportion Pct Exp Avg Exp Proportion
a Rhabdomys
0.00
0.25
0.50
0.75
1.00
100
100
0.0
0.1
0.2
0.3
0.4
0.5
b
25
50
75
25
50
75
−1
−1
Squirrel
0
0
0
1
2
0
1
2
MG MG
Cone Cone
Rod
Rod
HC
HC
GlyAC
GlyAC
GabaAC
GabaAC
BC_15 0.9
BC_3 BC_8 0.9
BC_1 BC_6
BC_11 BC_13
BC_7 BC_14
BC_17 BC_11
BC_13 BC_15
BC_14 BC_4
BC_16 BC_3
BC_6 BC_12
BC_9 BC_10
BC_5 BC_7
BC_12 BC_2
BC_10 BC_1
BC_4 BC_5
BC_2
BC_9
BC_8
RGC_30 0.2
BC_18
RGC_29
RGC_32 0.2
RGC_32
RGC_30
RGC_31 RGC_24
RGC_33 RGC_25
RGC_17 RGC_20
RGC_26 RGC_NA
RGC_22 RGC_9
RGC_25 RGC_28
RGC_29 RGC_11
RGC_24 RGC_2
RGC_16 RGC_21
RGC_3 RGC_19
RGC_1 RGC_26
RGC_28 RGC_16
RGC_19 RGC_31
RGC_20 RGC_18
RGC_4
RGC_3
RGC_7
RGC_13
RGC_11
RGC_15
RGC_21
RGC_12
RGC_10
RGC_9 RGC_8
RGC_23 RGC_22
RGC_6 RGC_27
RGC_12 RGC_17
RGC_18 RGC_7
RGC_15 RGC_4
RGC_14 RGC_14
RGC_8 RGC_23
RGC_13 RGC_5
RGC_27 RGC_1
RGC_2 RGC_10
RGC_5 RGC_6
0.2 0.2
all_s2a
all_s2b
neun-_s2
neun-_s1
neun+_s2
neun+_s1a
neun+_s1b
T HA6
V 6
P DG
R C
P D HO
A 6H
C 3
LB 3
AP P 1
S L BP 2
N Y1
VS FL
O X2
G X2
AB 1
T F R IK5
TF P2 1
A
AP B
G 2C
S A 1
O LC D2
U 9
L T1
AL 1
T P B1
SA 3
E
C MS
chx10+_s1
neun+_s1
TH 6
S L BP M 2
VS 1
O 2
G 2
VS 6
C X1
G P5
A 1
A A
AP B
G 2C
G D1
O LC D2
EC 9
LH 1
C X1
B1
G
R C
P D HO
H
R A3
P1
C S
C X
M
AD
R 1A
EC A
SL RR
R MS
C SX
G P
C HX
T FAP 2
T F IK
N 6A
R MS
Y
X
TX
T
TF P2
TF P2
PD
E6
S L CR
17
N 6
SA
T
E
R
U
17
AB
AL
1
LB
A
S A
A
R
G
R
BP
BP
R
Features Features
Pct Exp Avg Exp Proportion Pct Exp Avg Exp Proportion
c d Sea-lamprey
100
0.00
0.25
0.50
0.75
1.00
100
0.2
0.4
0.6
0.8
Marmoset
25
50
75
25
50
75
−1
−1
0
0
0
1
2
0
1
2
MG MG
Cone
Rod Cone
HC
Rod
GlyAC
GabaAC HC
BC_DB1 0.9
BC_FMB GlyAC
BC_DB3a GabaAC
BC_BB/GB*
BC_IMB BC_8 0.6
BC_DB6
BC_3
BC_RB
BC_DB4_2 BC_7
BC_DB4
BC_DB5* BC_5
BC_DB3b BC_2
BC_DB2
pRGC_pRGC10 0.2 BC_1
pRGC_pRGC8
BC_4
pRGC_pRGC7_b
pRGC_pRGC7_a BC_6
pRGC_pRGC4/9
pRGC_OFF_MGC RGC_15 0.3
pRGC_OFF_PGC_a RGC_18
pRGC_pRGC12/13
pRGC_OFF_PGC_b RGC_17
pRGC_pRGC3
RGC_12
pRGC_pRGC14
pRGC_pRGC5 RGC_9
pRGC_pRGC11
pRGC_ON_PGC RGC_3
pRGC_pRGC6 RGC_13
pRGC_ON_MGC
pRGC_pRGC1 RGC_8
fRGC_fRGC10
RGC_7
fRGC_ON_PGC
fRGC_OFF_PGC_a RGC_5
fRGC_fRGC2
fRGC_fRGC12 RGC_16
fRGC_fRGC11
RGC_10
fRGC_fRGC6
fRGC_fRGC3 RGC_11
fRGC_fRGC5
RGC_2
fRGC_fRGC4
fRGC_fRGC9 RGC_1
fRGC_OFF_PGC_b
fRGC_fRGC8 RGC_14
fRGC_OFF_MGC
RGC_6
fRGC_fRGC1
fRGC_ON_MGC RGC_4
fRGC_fRGC7
peri_cd73-_s1a
peri_cd73-_s1b
peri_cd90+_s1
fov_all_s1a
fov_all_s2a
fov_all_s2b
fov_all_s1b
fov_all_s2c
all_s1
all_s2
0.3
2
X2
T1
B
X1
S
1
2
0.2
H
O
T H A6
TF RI 5
G 1
N 6 2
U 9
L T1
AL 1
T P B1
SA 3
P DG
C
E 6O
C H
C R3
LB 3
S L BP S 2
VSY1
O X2
G TX2
C SX6
A 1
T FAP 1
T F AP A
AP 2B
G 2C
AP P 1
S LAR X
17 S
6F
AD
6A
TX
PN
O
M
_X
_X
VM
AD
O LCAD
EC A
R 1A
G BP
C HX
C M
O
R
E6
P DR H
RH
U
LH
VS
R M
AP
BP
U
R
C
2B
2B
EC
G
O
PD
BP
PO
SL
R
AP
AP
S
N
R
O
TF
TF
Features
Features
Extended Data Fig. 5 | Summary of cell type atlases for Rhabdomys, squirrel, 22,821 cells) (b), marmoset (n = 2 animals; 52,559 cells) (c), and Sea-lamprey
marmoset and sea-lamprey. Panels a-d depict atlases (as in Extended Data (n = 2 animals; 18,928 cells) (d).
Fig. 3) for Rhabdomys (n = 2 animals; 65,338 nuclei) (a), squirrel (n = 1 animal;
Article
Pct Exp Avg Exp Proportion Pct Exp Avg Exp Proportion
a b
100
Macaque
0.0
0.1
0.2
0.3
0.4
0.5
0.6
Mouse
100
0.0
0.1
0.2
0.3
0.4
25
50
75
−1
25
50
75
−1
0
1
2
0
0
1
2
MG
MG Cone
Cone Rod
Rod HC
HC GlyAC
GabaAC
GlyAC BC7 0.8
GabaAC BC5A
DB5* 0.8 BC6
BC5B
DB4
BC5C
DB6 BC9
BB/GB* BC8
OFFx BC5D
BC1A
DB1 BC1B
FMB BC2
DB2 BC3B
DB3b BC4
BC3A
DB3a RBC
RB C42_AlphaOFFS 0.4
IMB C33_M1a
pRGC_pRGC17 0.3 C41_AlphaONT
C45_AlphaOFFT
pRGC_pRGC11 C44
pRGC_pRGC7 C22_MX
pRGC_pRGC16 C31_M2
C20
pRGC_pRGC9
C39
pRGC_pRGC13 C14
pRGC_pRGC10 C35
pRGC_pRGC6 C28_FmidiOFF
C27
pRGC_pRGC15 C37
pRGC_pRGC18 C38_FmidiON
pRGC_pRGC12 C23_W3D2
pRGC_PGC_ON C11
C43_AlphaONS_M4
pRGC_pRGC5 C10
pRGC_pRGC14 C32_FRGC_novel
pRGC_PGC_OFF C19
pRGC_MGC_ON C36
C25
pRGC_pRGC8 C17_TRGC_S1
pRGC_MGC_OFF C7
fRGC_fRGC14 C13_W3L2
C5_JRGC
fRGC_fRGC16
C34
fRGC_fRGC15 C1_W3L1
fRGC_fRGC12 C6_W3B
fRGC_fRGC13 C3_FminiON
C16_ooDSGC_DV
fRGC_fRGC7 C21_TRGC_S2
fRGC_fRGC11 C9_TRGC_novel
fRGC_fRGC10 C40_M1b
fRGC_fRGC9 C2_W3D1
C30_W3D3
fRGC_fRGC6 C18
fRGC_fRGC8 C8
fRGC_fRGC5 C24
C12_ooDSGC_N
fRGC_MGC_ON
C15
fRGC_PGC_ON C29
fRGC_PGC_OFF C4_FminiOFF
fRGC_MGC_OFF C26
all_s1
all_s2
all-s3
all-s4
cd90+_s1
cd90+_s2
cd90+_s3
cd90+_s4
cd90+_s5
cd90+_s6
cd90+_s7
cd90+_s8
cd90+_s9
cd90+_s10
0.1
peri_cd73-_s1a
peri_cd73-_s1b
peri_cd73-_s2a
peri_cd73-_s2b
peri_cd73-_s2c
peri_cd90+_s1a
peri_cd90+_s1b
peri_cd90+_s1c
peri_cd90+_s1d
peri_cd90+_s1e
peri_cd90+_s1f
peri_cd90+_s1g
peri_cd90+_s1h
peri_cd90+_s1i
fov_all_s3a
fov_all_s1a
fov_all_s4a
fov_all_s4b
fov_all_s1b
fov_all_s1c
fov_all_s4c
fov_all_s1d
fov_all_s3b
fov_all_s1e
fov_all_s1f
fov_all_s1g
fov_all_s2a
fov_all_s2b
fov_all_s2c
fov_all_s2d
fov_all_s2e
fov_all_s2f
fov_all_s3c
fov_all_s2g
fov_all_s2h
0.4
T HA6
T FAP A
AP 2B
G 2C
G 1
N 6 2
U 9
C HX 1
AL 1
T P B1
SA 3
P DG
P DR H C
E 6O
C H
R 1A3
LB 3
AP P 1
S L BP S 2
VSY1
O X2
G TX2
C SX6
A 1
TF RI 5
T FAP 1
S LAR X
17 S
E
AD
O LC D
VM
EC A
M
G BP
L T
CR
2
T HA6
C M
O
C S2
VSY1
O X2
G X2
VS 6
G X1
TF P2 1
A A
AP B
G 2C
G D1
O LC D2
U 9
L T1
C HX 1
SA 1
P DG
R C
P D HO
C H
C 3
R 1A3
AP P 1
R
A RX
E
M
T F IK
R M
EC A
A
SL RR
TF P2
O
E6
R
SL PM
17
N 6
AL
LB
T
A
S A
R
BP
R
A
S
B
R
R
Features Features
Pct Exp Avg Exp Proportion Pct Exp Avg Exp Proportion
c Chicken d Zebrafish
100
0.0
0.1
0.2
0.3
100
25
50
75
0.0
0.1
0.2
0.3
0.4
0.5
−1
25
50
75
0
−1
0
1
2
0
0
1
2
MG MG
Cone Cone
Rod Rod
HC 0.9 HC
GlyAC GlyAC
GabaAC GabaAC
BP−5 0.7 BC_9 0.8
BP−8 BC_16
BP−2 BC_6
BP−12 BC_8
BP−18
BC_18
BP−11
BP−15 BC_11
BP−16 BC_15
BP−17 BC_14
BP−13 BC_7
BP−19 BC_23
BP−21 BC_22
BP−10 BC_21
BP−1 BC_17
BP−3 BC_10
BP−14 BC_12
BP−4
BP−6 BC_13
BP−7 BC_4
BP−22 BC_2
BP−9 BC_3
BP−20 BC_19
RGC−40 0.2 BC_5
RGC−13 BC_20
RGC−37 BC_1
RGC−41 RGC_24 0.2
RGC−23 RGC_30
RGC−31
RGC_31
RGC−22
RGC−29 RGC_25
RGC−18 RGC_29
RGC−27 RGC_20
RGC−28 RGC_4
RGC−14 RGC_18
RGC−8 RGC_27
RGC−35 RGC_6
RGC−33 RGC_21
RGC−34 RGC_9
RGC−30 RGC_7
RGC−4
RGC−9 RGC_22
RGC−36 RGC_3
RGC−16 RGC_13
RGC−38 RGC_5
RGC−32 RGC_10
RGC−10 RGC_32
RGC−19 RGC_23
RGC−1 RGC_28
RGC−39 RGC_11
RGC−11 RGC_14
RGC−20
RGC_12
RGC−6
RGC−26 RGC_16
RGC−3 RGC_19
RGC−21 RGC_2
RGC−15 RGC_8
RGC−17 RGC_17
RGC−5 RGC_15
RGC−7 RGC_26
RGC−12 RGC_1
RGC−2 0.3
RGC−25
P O S 2B
TH 2
VS 1
S L D2
PR X2
O A
VS 2
TF X1
G A
EC 9
C T1
TP 1
3
AR HO
P D 3A
G 6H
AP LA
EB
N A
B
M
6F
TX
RGC−24
KC
U
O C6
AL
LU
AP
R
E
O
R
cd73-_s1
cd90+_s15
U
cd73-_s2
cd73-_s3
cd73-_s4
cd73-_s5
cd73-_s6
cd73-_s7
cd73-_s8
cd90+_s1
cd90+_s2
cd90+_s3
cd90+_s4
cd90+_s5
cd90+_s6
cd90+_s7
cd90+_s8
cd90+_s9
cd90+_s10
cd90+_s11
cd90+_s12
cd90+_s13
cd90+_s14
M
0.1
BP
all_s1
all_s2
all_s3
all_s4
cd90+_s1
cd90+_s2
cd90+_s3
cd90+_s4
G B
G 1
S L D2
U 2
TH 1
VS 1
VS 1
G 2
O 1
TF TX2
T F P 2A
E 9
EC 1
AR 2
3
G 1
T2
PD O
B
R C
P1
BP S
R
IK
AD
N A
P O MS
4F
Y
X
X
N T
T
R
S
R PM
E6
PD
H
Features
O CU
G
O C6
LB
AP
R
R
R
A
N
B
R
Features
Extended Data Fig. 6 | Summary of cell type atlases for macaque, mouse, replicates; 60657 cells) (d). Cluster labels are consistent with published
chick and zebrafish. Panels a-d depict atlases (as in Extended Data Fig. 3) for annotations19,20,22,27,28. Each biological replicate in zebrafish involved a pooling
macaque (n = 4 animals; 146,054 cells) (a), mouse (n = 10 animals; 51,162 cells) of eyes from multiple (5-8) fish.
(b), chick (n = 4 animals; 34,788 cells) (c), and zebrafish (n = 15 biological
a b
s_AC
Tree shrew_AC
C
Chicken_AC
Mac oset_AC
Pero
um_A
S heep_ AC
Rhabdomy
C
Lizard_
HC
e_ A
mys se_ A
Ferr _ AC
Pig_AC
Cow_A C
HC
w_
Tre an_ A
Mo rel_ A C
Oposs
aqu
Sq
cus
Marm
hre
et_ A
rre e_ H s_
u
uir p_ H C
AC
Ze mo an_
S h fish BP
M r om C
Fe ous yscu
s
Hum
M Hu
Pe w_ H
br
e
C
e
ar
e
Rh
a et_ BP
10
C
ab
HC
Co
C
d
s
_
Ch om C _ HC C
t_
m
A
Tre ick ys_ H
e g_ en _ H
Op
sh
r
en BP
_ P i hick sum Chicken
os ew_ BP C os HC _ HC
su
m_ P
B Op ard_ mys Cow
P ig BP Liz abdo HC
Co _ BP R h irrel_ HC 5 Ferret
u
w_ S q ue_ Human
S he
ep_ P
B caq HC
Ma
Ferr BP an_ Lamprey
et_ B Hum oset_HC
Lizard P Marm G Lizard
_BP ish_M
Peromysc Zebraf
us_BP
Zebrafish_
BP Macaque
UMAP_2
Mouse_BP
Zebrafish_HC 0 Marmoset
Squirrel_BP
Macaque_BP Lizard_ PR Mouse
Zebrafish_R
GC Squirrel_P
R Opossum
C
Cow_ RG
Chicken
_P R Peromyscus
_R GC Tree
S heep GC P ig_
shrew
_P R −5 Pig
et_ R C Pero R
P
Ferr RG Rhabdomys
rd_ C R h myscu
Liza _ R G a
Op bdom s_ P R Sheep
n GC
ma Mo ossu ys_ P
Hu ue_ R GC R Squirrel
q
ca ys_ R GC Ze use m_ P
Ma m R C M bra _ P R R Tree shrew
do P ig_ R G C M arm fish
ab _ ac o _ −10 Zebrafish
Rh m RG aq set P R
su s_
S q use R G C
Hu ee P R
ue _ P
Mo ken _ R G
os cu
rel_ R G C
S h w_ P R
_P R
m p_ P
p
ys
Pero hrew R G C
Co rret_
an
O R
ic et
RG C
Fe _ MG
m
_P
_
Ch os
ro
Co MG
Marm cus_ C
ose MG
Pe
P ig p_ MG
_
R
m
Hum t_ MG
S he
w
ar
R
G
S qu
G
Lizard
_MG
_
M
Rhabdo
Tree shrew_M
Mouse_ MG
Ferret_MG
Opossum_MG
an_ M
_
ue_M
uir
e
irrel_
s
Chicken
−10 −5 0 5 10
_MG
my
es
Macaq
MG
mys_ MG
UMAP_1
Tre
c d
5
0
5
0
5
0
5
0
0.2
0.5
0.7
1.0
0.2
0.5
0.7
1.0
MG
PR
HC
AC
BC
RGC
e RGC BC AC HC PR MG
0.8
expression distance
0.6
Mean squared
0.4
0.2
y
0.0
0.0 0.8 1.6 2.4 0.0 0.8 1.6 2.4 0.0 0.8 1.6 2.4 0.0 0.8 1.6 2.4 0.0 0.8 1.6 2.4 0.0 0.8 1.6 2.4
Evolutionary Divergence (substitutions per 100 bp)
Extended Data Fig. 7 | Evolutionary conservation of retinal classes. a class. d. Same as panel c, but rows and columns grouped based on species
a. Dendrogram showing transcriptional relationships among pseudobulk instead of class (compare with Fig. 2c). e. Pairwise mean-squared distance of
expression vectors following integration. Each node is a cell class within a class-specific cell-averaged gene expression profiles between all 16 jawed
particular species. Dendrograms were computed via hierarchical clustering vertebrate species (y-axis) increases with evolutionary divergence, as estimated
analysis (correlation distance, average linkage). b. Same as Fig. 2d, with cells by substitutions per 100 bp (x-axis) (compare with Fig. 2e). Gray shaded regions
colored by species of origin. Inset shows a magnified region containing samples demarcate species pairs involving zebrafish. Solid lines represent power law
from all species. c. Cross-correlation matrix (spearman) of class- and species- (y = ax b) regression fits. Across the panels, a ∈ [0.34, 0.47] and b ∈ [0.29, 0.45].
specific cell-averaged profiles for all 17 vertebrates (compare with Fig. 2b). The coefficient of determination (R 2) values range from 0.79-0.93.
Rows and columns are grouped by class, and then ordered by phylogeny within
Article
a b MEIS2 TCF4
THRB LHX4
10 10
GabaAC
5 5
Cone GlyAC
NRL NR2E3
UMAP_2
UMAP_2
0 0
Rod
−5 −5
−10 −10
−10 −5 0 5 10 −10 −5 0 5 10
UMAP_1 UMAP_1
c e f
FEZF2 LHX3
5
0
5
0
0.2
0.5
0.7
1.0
5
0
5
0
0.2
0.5
0.7
1.0
10
Cone
5
Rod
ISL1 ST18
GlyAC
UMAP_2
GabaAC
−5
OFF BC
−10 OFF BC ON BC
ON BC
−10 −5 0 5 10
UMAP_1
Cone
Norm. Exp.
Rod 0.00
0.25
0.50
0.75
1.00
GlyAC
GabaAC
OFF BC
ON BC
B
3
F4
M 9
2
3
L1
8
6A
AD
IK
T1
X
2E
IS
ZF
X
R
IS
TC
LH
LH
N
AR
TH
R
E
LC
S
R
FE
G
G
N
Cone BC OT distribution
oBC1A
BC1B
BC2 oBC1B
BC OrthoType (OT)
BC5D oBC5B
0.25
BC6 oBC5C
BC7 oBC5D
BC8/9
oBC6
RBC 0.00
oBC7
ep
ig
ea
t
ry
ew
se
el
ow
a
ys
um
ry
rre
ve
cu
r
ve
P
irr
he
he
he
ov
om
he
ou
hr
Fe
ss
M _ B 1A
M u_ 1B
M _ B C2
M _ B 3A
M u_ 3B
M _ B C4
M u_ B 5A
M _ B 5B
M _ B 5C
M _ B5D
M _B 6
M _B 7
M u_ B 8
_R 9
BC
ys
Fo
Fo
qu
r ip
r ip
r ip
ou C
ou C
o C
ou C
F
oBC8/9
M
S
S
bd
po
m
ou C
o C
ou C
o C
o C
ou C
ou B
ou B
ou C
ou C
S
an
et
Pe
Pe
e
Pe
ee
ro
M _B
ha
O
qu
os
um
Pe
Tr
an
et
R
ue
ou
m
ac
os
H
um
aq
ar
M
M
Mouse Type (Shekhar et al., 2016)
ac
H
ar
M
M
c d f
RBC DE genes across mammals
Chicken Opossum Avg Exp
Cow oRBC
Ferret 2.5
Ferret 2.0
NM_ON Pig 1.5
Human Cow 1.0
Lizard oBC3A oBC7 Sheep 0.5
Macaque
oBC6 0.0
Squirrel −0.5
Marmoset oBC8/9 oBC5C Peromyscus
Mouse Rhabdomys Pct Exp
oBC5B
Opossum Mouse 0
Peromyscus oBC5D Tree_shrew 25
oBC2
Pig oBC1B oBC5A Marmoset 50
NM_OFF
Rhabdomys Macaque 75
Sheep Human 100
oBC1A
Squirrel
C IK
P 3
01 W 2
AR P C MD
G 9
C A7
XC 1
T S 14
AK HB
N P1
N D2
R 1
U B
C E2
C T9
PP B
S A A2
16 L
G 28
T L1
SD N
S H T H 7B
O 4
SL TP 4
C R2
IP
oBC3B
Tree shrew
SH
H DH
C RK
G G
8L C2
RO S D
M
30 V KP
S C IN2
PA G
T H C2
EM
R
oBC4
16
L
AT AP
12
A
B
S
O
R
N
H
IT
Zebrafish
I
SL
A8
e
oBC1A
oBC1B
oBC2
oBC3A
oBC3B
Mapping %
oBC4
0
oBC5A 25
oBC5B 50
75
oBC5C
100
oBC5D
oBC6
oBC7
oBC8/9
oRBC
NM_ON
NM_OFF
ys
s
an
ue
et
se
ow
rd
ck
h
rre
re
cu
ee
Pi
is
re
om
os
za
hi
ou
um
aq
r
C
af
ss
ui
ys
Fe
Sh
Sh
C
m
Li
br
M
bd
ac
Sq
po
H
m
ar
Ze
ee
M
ha
ro
ow
M
Tr
Pe
R
Br
Extended Data Fig. 9 | Bipolar Cell OrthoType analysis including non- enriched for non-mammalian BCs from chick, lizard and zebrafish. The two
mammals. a. Confusion matrix showing the rationale behind naming OTs, named NM_OFF and NM_ON, are enriched for OFF and ON BCs from
mammalian BC OTs (rows) based on the mapping patterns of mouse BC types non-mammals (also see panel e). e. Confusion matrices showing the mapping of
(columns)19. Representation as in Fig. 3d, with each column summing to 100%. species-specific BC clusters (columns) to BC OTs (rows) identified by integrating
OT BC8/9 contains mappings from both mouse BC8 and BC9, which are BCs from all jawed vertebrates (panel c). Representation as in Fig. 3d’. Mammalian
transcriptionally proximal. b. Barplot showing within-species relative BC clusters predominantly map to the mammalian OTs (rows 1-14), and the
frequencies (y-axis) of the 13 cone BC OTs within each mammalian species pattern of mapping is similar to Fig. 3d. Chick, Lizard and Zebrafish BCs largely
(x-axis). The foveal and peripheral data from primates are plotted separately. map to the non-mammalian OTs NM_OFF and NM_ON (rows 15-16). f. Dotplot
c. Integrated UMAP of BCs from all 16 jawed vertebrates. Cells are colored by showing species-specific genes (columns) expressed in RBC orthologs in
species of origin. Lamprey, a jawless vertebrate, was excluded from the analysis mammals (rows). The size and color of each dot represent the percentage of
due to the paucity of shared orthologous genes. d. Same as c, with cells colored cells within the species cluster expressing the gene and the average expression
by OT identity. The integration of all jawed vertebrates recovers all the level, respectively.
mammalian BC OTs listed in Fig. 3c, but additionally identifies two OTs
a OT
b c
1.00
oRGC1 oRGC12 Chicken oRGC13
oRGC2 oRGC13 Ferret
HumanFovea
0.75 oRGC3 oRGC14 HumanPeriphery oRGC21
oRGC12 oRGC14
Lizard
oRGC4 oRGC15 oRGC10
MacaqueFovea oRGC15
Proportion
M ac eri ea
os eF y
m ip a
ry
sh a
Pe bdo e
m s
S us
S el
ep
F g
po t
um
Zebrafish
w
O erre
ro my
m qu er
M etP ove
ee ve
ha us
i
irr
P
a c m he
os he
re
c
M P v
he
ss
ar a ph
ys
ue Fo
Tr tFo
R Mo
qu
M Hu rip
ar er
Pe
e
an
um
H
d
oRGC1
oRGC2
oRGC3
oRGC4
oRGC5
oRGC6
RGC OrthoType (OT)
oRGC7
oRGC8
oRGC9
oRGC10
oRGC11
oRGC12
oRGC13
oRGC14
oRGC15
oRGC16
oRGC17
oRGC18
oRGC19
oRGC20
oRGC21
oRGC_NM
P
F
oRGC1
oRGC2
oRGC3
oRGC4
oRGC5
oRGC6
RGC OrthoType (OT)
oRGC7
oRGC8
oRGC9
oRGC10
oRGC11
oRGC12
oRGC13
oRGC14
oRGC15
oRGC16
oRGC17
oRGC18
oRGC19
oRGC20
oRGC21
oRGC_NM
oRGC8 0
oRGC9
oRGC10 20
oRGC11 40
oRGC12
oRGC13 >60
oRGC14
oRGC15
oRGC16
oRGC17
oRGC18
oRGC19
oRGC20
oRGC21
OPN4+
EOMES+
Liz 11
3
H Fo Fer 6
M Pe R G 7
M Fo _ R 6
M Per fR G 6
F pR 5
ou fR C8
22 C1
ou X
Pe 8
P r_ 7
P 15
ha 8
S 6
S _3
Tr _ 6
H Fo Fe 29
M Pe R G 26
M Fo R G 13
M Per fR G 13
G 4
M u_ GC 5
M _ C 31_ 8
_ C 3_ 2
M _M a
_ b
ou 4
Pe C7
r 6
P _ 37
ha 1
S _ 27
Ch Tre 4
G 0
_2
C
C
ar _ C
_C
R _1
_1
Pe pR 1
M r_ p C1
ou C 1
ou 3 M
40 M1
ou 1
M C4
Pe _ 1
R _3
_2
i_R _ 2
r_
um v_ _
M _M
C-
ac r_ C
ac v_ C
ac _ C
_
he
he
e_
_ _
ac v_ G
M ov_ G
_C G
_
Fe
ig
ig
ig
um v r
qu
o R
ac r
um
um
H
b
R M _M
Proportion a G
EOMES
TLE1
PROX1
MAFB
ZEB2
GRM5
IRX3
TBR1
PLXNA2
TFAP2D
MEIS2
ha M c_ M C_
bd M M ar_ GC OF
0.00
0.03
0.06
0.09
0.12
M ou_ arF OF _ O F
ou C ov F _ F
om
_ C 17 _ f M F
M ys 42 _ T R G GC
o _ A R G C1
lp C 2
ha _ S
O 1
Pe S q use
Fe F F S
ro uir
m re r
Fe _ 14
r
ys l
c O _ 20
po
S us Pe _ 2
r
he
Pe _ 18
r_
ep
O Fer P i 23
p r P i g_ 1
M Tr os et R g_ 1
ha 7
_
ar ee su
m S h 26
S q e_
sh m
u_ 2
os re
M et w Tr 27
Tr e_ 1
e_
M ar m P e P
H 28
ac o ri ig
p um
M
aq se h
ou M _M
u t e _C a c_ GC
M e P Fo ry M
43 a MG _
_ A r_ O C ON
lp N _ O
H aca eri vea
um qu ph ha _ N
O MG
N C
S_
an e F er
y F e M4
H Pe ove r_
F 19
um ri a
p O er_
e
an he po 9
Pe _ 28
r_
Fo ry
P i 34
g
ve
a Tr _ 4
e_
3
H
um
H _P
um G
OT
H F o C_
u O
M mP v_ R F F
a e G
M c_ P r_ R C4
M ar G G
_O C C
ou
F _O 4
specific clusters. The size and color of each dot represent the percentage of
45 MF _ P F F
_A o G
lp u_ C C
cells within the species cluster expressing the gene and the average expression
ha 3
Extended Data Fig. 11 | Midget and Parasol OTs. a. Dotplot showing examples
O 4
O FF
p T
O o_ 2
po 2
oRGC2 (PGC ON)
Pe _ 27
r
oRGC5 (PGC OFF)
Pe _ 36
R r_ 3
ha 8
Sq _3
u 0
Tr _ 30
e_
34
H
H um_
um P
H F G
um ov C
H Fo _ R _ ON
u v G
H mP _ R C1
−1
50
40
30
20
10
Pct Exp
Avg Exp
genes
select candidate
combination
multicollinearity removal
Primate MGC ON
Primate PGC ON
Mouse αOFFs
Mouse αOFFt
Mouse αONs
Mouse αONt 1
2
3
4