Content uploaded by Dennis Dacey
Author content
All content in this area was uploaded by Dennis Dacey on Mar 03, 2016
Content may be subject to copyright.
Proc.
Natl.
Acad.
Sci.
USA
Vol.
93,
pp.
582-588,
January
1996
Colloquium
Paper
This
paper
was
presented
at
a
colloquium
entitled
"Vision:
From
Photon
to
Perception,
"
organized
by
John
Dowling,
Lubert
Stryer
(chair),
and
Torsten
Wiesel,
held
May
20-22,
1995,
at
the
National
Academy
of
Sciences,
in
Irvine,
CA.
Circuitry
for
color
coding
in
the
primate
retina
(color
opponent/cone
photoreceptors/ganglion
cells/horizontal
cells/bipolar
cells)
DENNIS
M.
DACEY*
Department
of
Biological
Structure,
University
of
Washington,
Box
357420,
Seattle,
WA
98195-7420
ABSTRACT
Human
color
vision
starts
with
the
signals
from
three
cone
photoreceptor
types,
maximally
sensitive
to
long
(L-cone),
middle
(M-cone),
and
short
(S-cone)
wave-
lengths.
Within
the
retina
these
signals
combine
in
an
antag-
onistic
way
to
form
red-green
and
blue-yellow
spectral
op-
ponent
pathways.
In
the
classical
model
this
antagonism
is
thought
to
arise
from
the
convergence
of
cone
type-specific
excitatory
and
inhibitory
inputs
to
retinal
ganglion
cells.
The
circuitry
for
spectral
opponency
is
now
being
investigated
using
an
in
vitro
preparation
of
the
macaque
monkey
retina.
Intracellular
recording
and
staining
has
shown
that
blue-
ON/yellow-OFF
opponent
responses
arise
from
a
distinctive
bistratified
ganglion
cell
type.
Surprisingly,
this
cone
oppo-
nency
appears
to
arise
by
dual
excitatory
cone
bipolar
cell
inputs:
an
ON
bipolar
cell
that
contacts
only
S-cones
and
an
OFF
bipolar
cell
that
contacts
L-
and
M-cones.
Red-green
spectral
opponency
has
long
been
linked
to
the
midget
gan-
glion
cells,
but
an
underlying
mechanism
remains
unclear.
For
example,
receptive
field
mapping
argues
for
segregation
of
L-
and
M-cone
signals
to
the
midget
cell
center
and
surround,
but
horizontal
cell
interneurons,
believed
to
generate
the
inhibitory
surround,
lack
opponency
and
cannot
contribute
selective
L-
or
M-cone
input
to
the
midget
cell
surround.
The
solution
to
this
color
puzzle
no
doubt
lies
in
the
great
diversity
of
cell
types
in
the
primate
retina
that
still
await
discovery
and
analysis.
From
Cell
Types
to
Microcircuits
The
vertebrate
retina
is
that
part
of
the
central
nervous
system
where
multiple
parallel
representations
of
the
visual
world
first
emerge.
And
like
other
parts
of
the
brain,
the
retina
is
a
beautiful
and
complex
piece
of
neural
machinery,
although
it
has
taken
nearly
100
years
for
the
degree
and
nature
of
its
complexity
to
be
fully
appreciated.
Since
the
anatomical
renderings
of
Cajal
(1),
the
basic
framework
of
retinal
circuitry
has
been
known.
But
only
within
the
last
decade
has
it
become
clear
that
the
retina
contains
a
diversity
of
neural
cell
types
comparable,
in
fact,
to
that
of
the
cerebral
cortex.
Rods
and
two
or
three
types
of
cone
photoreceptors
relay
signals
to
at
least
10
types
of
bipolar
interneurons.
The
bipolar
cells
in
turn
contact
20-25
distinct
ganglion
cell
types,
which
give
rise
to
an
equal
number
of
parallel
pathways
to
the
visual
brain.
Addi-
tional
networks
of
interneurons
allow
lateral
interactions
to
modify
these
parallel
pathways:
2
horizontal
cell
types,
at
the
level
of
the
photoreceptor-bipolar
cell
synapse,
and
20-40
amacrine
cell
types
at
the
level
of
the
bipolar-ganglion
cell
synapse
(2-4).
Most
of
these
cell
types
have
not
yet
been
studied
in
detail,
but
their
existence
is
no
longer
disputed.
The
publication
costs
of
this
article
were
defrayed
in
part
by
page
charge
payment.
This
article
must
therefore
be
hereby
marked
"advertisement"
in
accordance
with
18
U.S.C.
§1734
solely
to
indicate
this
fact.
Without
doubt,
understanding
the
functional
architecture
of
the
retinal
circuitry
must
begin
by
first
characterizing
these
cell
types.
In
an
astounding
feat
of
neural
efficiency,
all
of
this
complex
circuitry
is
packaged
in
a
thin,
precisely
laminated
sheet
of
tissue.
Each
retinal
cell
type
shows
a
characteristic
set
of
physiological
properties
and
connections
with
other
cell
types
within
the
retinal
layers.
Each
type
also
shows
a
characteristic
density
and
spatial
arrangement
across
the
retina
and,
like
a
pattern
of
interlocking
tiles,
the
"mosaic"
of
cells
of
a
partic-
ular
type
forms
an
identifiable
unit
of
retinal
circuitry.
A
variety
of
techniques
have
revealed
these
distinctive
cell
mosaics
(5)
and
the
synaptic
links
among
them.
From
this
work
it
has
become
clear
that
the
diverse
retinal
cell
types
are
the
building
blocks
of
multiple
"microcircuits"
that
function
in
parallel
(2,
3).
The
diversity
of
retinal
cell
types
and
associated
microcir-
cuits
provides
a
new
framework
for
understanding
the
struc-
ture
and
function
of
the
primate
retina
and
its
role
in
the
visual
process.
One
important
function
of
the
primate
retina
is
to
transmit
color-related
signals,
and
in
this
paper
I
review
recent
attempts
to
identify
the
cell
types
and
microcircuits
that
are
responsible
for
the
complex,
color-coding
receptive
fields
of
primate
ganglion
cells.
This
work
has
been
done
in
macaque
monkeys,
a
group
that
shares
with
humans
(and
other
Old
World
species)
a
retina
that
contains
three-cone
photorecep-
tor
types,
each
maximally
sensitive
to
a
different
part
of
the
visible
spectrum.
The
cone
spectral
sensitivities
and
many
aspects
of
the
detailed
anatomy
of
the
retina
and
visual
pathways
in
macaque
are
virtually
identical
to
those
in
humans,
establishing
this
genus
as
an
excellent
model
for
the
neural
basis
of
human
color
vision.
Classical
Labeled
Line
Model
for
Color
Opponent
Circuitry
At
an
early
stage
in
visual
coding,
signals
from
the
three-cone
types
combine
in
an
antagonistic,
or
opponent,
fashion.
This
crucial
stage-
in
the
neural
representation
of
color
is
clearly
manifest
in
color
perception
(6).
Two
opponent
channels
exist:
in
the
red-green
opponent
pathway,
signals
from
long-
and
middle-wavelength-sensitive
cones
(L-
and
M-cones,
respec-
tively)
are
opposed;
and
in
the
blue-yellow
pathway,
signals
from
short-wavelength-sensitive
cones
(S-cones)
oppose
a
combined
signal
from
L-
and
M-cones.
A
neural
correlate
of
these
perceptual
opponent
channels
can
be
found
in
the
light
responses
of
certain
ganglion
cells
in
the
macaque
retina.
These
spectrally
opponent
neurons
are
excited
by
wavelengths
in
one
region
of
the
spectrum
and
inhibited
by
light
from
another
part
of
the
spectrum,
typically
showing,
at
some
intermediate
point,
a
null
response
where
Abbreviations:
L-cone,
M-cone,
and
S-cone,
cone
photoreceptors
maximally
sensitive
to
long,
middle,
and
short
wavelengths;
LED,
light-emitting
diode;
LGN,
lateral
geniculate
nucleus.
*To
whom
reprint
requests
should
be
addressed.
582
Proc.
Natl.
Acad.
Sci.
USA
93
(1996)
583
excitation
and
inhibition
cancel
(7).
Although
this
spectral
opponency
has
been
studied
for
more
than
30
years,
the
underlying
retinal
circuitry
remains
unclear.
Wiesel
and
Hubel
(8)
were
the
first
to
suggest
a
simple
circuitry
by
which
color
opponency
could
arise
in
macaque
ganglion
cells.
Recording
from
the
the
lateral
geniculate
nucleus
(LGN),
the
target
of
color-responsive
ganglion
cells,
they
reported
that
inputs
from
the
different
cone
types
ap-
peared
to
be
segregated
to
the
center
and
the
surround
of
the
classical
receptive
field.
Color
opponency
thus
could
arise
by
piggy-backing
on
the
antagonistic
center-surround
organiza-
tion
found
in
many
ganglion
cells.
For
example,
a
red-ON/
green-OFF
opponent
cell
would
receive
excitatory
L-cone
input
to
the
receptive-field
center
and
inhibitory
M-cone
input
to
the
receptive-field
surround.
A
consequence
of
this
com-
bined
spatial
and
cone
opponency
is
that
this
type of
cell
could
signal
achromatic
luminance
variation,
due
to
center-surround
spatial
antagonism,
and
also
signal
chromatic
change
that
engaged
both
the
excitatory
and
inhibitory
cone
pathways
(9).
This
type
of
spatially
and
chromatically
opponent
receptive
field
was
labeled
"Type
1"
(Fig.
1).
Wiesel
and
Hubel
(8)
described
a
second,
Type
2,
opponent
cell
class,
which
also
appeared
to
receive
excitatory
and
inhibitory
input
from
different
cone
types,
but
which
lacked
a
clear
center-surround
organization.
Instead,
opposing
cone
inputs
were
distributed
in
spatially
coextensive
ON
and
OFF
responding
fields
(Fig.
1).
As
recognized
by
Hubel
and
Wiesel
and
others
to
follow,
this
Type
2
receptive-field
organization
suggested
a
specialization
for
color
coding
independent
of
any
role
in
spatial
vision.
Cone
type-selective
circuitry
center-surround
Type
1
red-ON
Type
2
+be
blue-ON
FIG.
1.
Classical
cone-type-specific
circuitry
(labeled-line
model)
for
color
opponency
in
ganglion
cells.
In
the
Type
1
receptive
field,
inputs
from
different
cone
types
(L-
and
M-cones
in
this
example)
are
segregated
to
the
center
vs.
the
surround
of
the
receptive
field.
Type
1
cells
show
a
center-surround
antagonism
to
luminance
changes
and
a
spatially
uniform
response
to
full-field,
equiluminant
color
changes
(in
this
case
an
excitatory
response
to
a
shift
to
a
longer
wavelength).
In
Type
2
cells,
opposing
inputs
(S-cones
vs.
L-
and
M-cones)
form
two
spatially
coextensive
fields
and
thus
lack
the
center-surround
antag-
onism
to
luminance
changes.
Clearly,
although
they
both
display
opponency,
Type
1
and
Type
2
cells
must
be
linked
somewhat
differently
to
cones
and
interneurons.
In
Type
1
cells,
the
cone
inputs
must
be
segre-
gated
spatially,
while
in
Type
2
cells,
the
cone
inputs
are
coextensive
but
opposite
in
sign.
Nonetheless,
the
cornerstone
for
the
circuitries
of
both
Type
1
and
Type
2
cells
is
the
existence
of
labeled
lines,
that
is,
the
anatomical
segregation
of
the
different
cone
signals
from
the
receptors
through
the
connecting
interneurons
to
the
ganglion
cell.
This
labeled
line
model
predicts
a
retinal
circuitry
that
can
sort
out
the
L-
and
M-cone
signals
and
deliver
them
with
the
appropriate
sign
to
the
appropriate
part
of
the
receptive
field.
Identifying
the
Color
Opponent
Ganglion
Cell
Types
To
explore
the
labeled
line
model
and
determine
the
retinal
circuitry
giving
rise
to
red-green
and
blue-yellow
opponency
in
ganglion
cells,
the
ganglion
cell
types
that
transmit
these
signals
must
first
be
identified.
In
an
early
attempt,
DeMon-
asterio
(10),
using
intracellular
recording
and
staining
meth-
ods,
tentatively
suggested
that
a
morphologically
identified
group
of
ganglion
cells
with
large
cell
bodies,
called
parasol
cells,
were
the
blue-ON/yellow-OFF
opponent
cells
and
that
cells
with
small
cell
bodies
and
small
dendritic
trees,
called
midget
ganglion
cells,
probably
transmitted
red-green
oppo-
nent
signals.
Parasol
cells
have
since
been
shown
to
project
exclusively
to
the
magnocellular
LGN
layers
where
achromatic,
nonopponent
cells
are
recorded
and
so
play
no
part
in
color
coding.
However,
midget
ganglion
cells
provide
the
major
input
to
the
the
parvocellular
layers
of
the
LGN
where
both
red-green
and
blue-yellow
opponent
cells
are
found
(11,
12).
Thus,
the
midget
ganglion
cells
came
to
be
associated
with
the
overall
group
of
color
opponent
cells
despite
significant
dif-
ferences
in
the
receptive-field
properties
of
the
red-green
and
blue-yellow
opponent
ganglion
cells
(13,
14).
A
more
direct
link
between
anatomy
and
physiology
requires
the
direct
correlation
of
an
identified
ganglion
cell
type
with
a
color
opponent
receptive
field.
Studying
Color
Circuitry
with
an
in
Vitro
Preparation.
Recently
developed
techniques
have
enabled
breakthroughs
in
linking
structure
to
function
in
the
retina.
In
pioneering
studies
of
rabbit
retina
by
Masland
and
Vaney
and
their
colleagues
(15,
16),
an
isolated
retina
was
maintained
in
vitro,
and
fluorescent
markers
were
used
to
identify
cell
types
under
the
light
microscope.
Targeted
cells
could
then
be
intracellularly
filled
with
dyes
to
reveal
the
cell's
dendritic
morphology.
This
in
vitro
approach
was
later
applied
to
macaque
retina
(17,
18)
and
eventually
extended
to
combined
anatomical
and
physio-
logical
experiments
(19-22).
The
key
to
the
success
of
this
preparation
is
that
neuronal
light
responses
can
be
recorded
from
cell
types
that
have
been
visually
identified.
In
macaque,
the
L-,
M-,
and
S-cone
spectral
sensitivities
are
known,
so
the
method
of
silent
substitution
can
be
used
to
identify
cone
inputs
to
a
cell.
With
this
method,
two
lights
of
differing
spectral
composition
are
alternated
and
their
relative
radiances
adjusted
so
that
the
alternation
between
the
pair
of
lights
will
give
rise
to
a
modulated
response
in
one
but not
the
other
(the
silent)
cone
type
(23-25).
We
have
now
used
this
approach
in
macaque
retina
in
vitro
to
explore
circuits
that
underlie
opponency.
Circuitry
for
Blue-Yellow
Opponency
and
the
Role
of
the
Small
Bistratified
Cell.
The
first
cell
type
studied
with
this
approach
was
the
small
bistratified
ganglion
cell,
one
of a
number
of
ganglion
cell
types
that,
in
addition
to
the
midget
ganglion
cell,
projects
to
the
parvocellular
geniculate
layers
(26,
27).
The
cell's
distinctive
dendritic
tree
stratifies
in
two
separate
sublayers
within
the
inner
plexiform
layer
(Fig.
2).
The
innermost
tier
of
dendrites
costratifies
with
the
axon
terminals
of
a
cone
bipolar
cell
type
that
makes
exclusive
Colloquium
Paper:
Dacey
Proc.
Natl.
Acad.
Sci.
USA
93
(1996)
Blue-ON
bistratified
cell
a
blue/yellow
-
chromatic
red+green
blue
FIG.
2.
Wholemount
view
of
the
dendritic
morphology
of
the
blue-ON,
small
bistratified
ganglion
cell.
(a)
The
inner
dendritic
tree
costratifies
with
the
axon
terminals
of
the
"blue-cone"
bipolar
cell,
close
to
the
ganglion
cell
layer.
(b)
The
outer
dendritic
tree
is
more
sparsely
branching
than
the
inner
tree
and
stratifies
close
to
the
amacrine
cell
layer.
This
cell
was
injected
intracellularly
with
Neuro-
biotin,
and
the
morphology
was
demonstrated
by
horseradish
perox-
idase
(HRP)
histochemistry.
contact
with
S-cone
pedicles
(28),
suggesting
a
role
for
the
bistratified
ganglion
cell
in
an
S-cone
signal
pathway
(27).
Intracellular
recordings
from
small
bistratified
cells
in
vitro
confirmed
that
they
received
S-cone
signals
and
showed
that
they
corresponded
to
a
distinct
blue-ON/yellow-OFF
oppo-
nent
cell
type
(21)
(Fig.
3).
An
excitatory
input
from
S-cones
was
demonstrated
with
chromatic
and
S-cone-isolating
stimuli.
Surprisingly,
the
response
to
offset
of
a
yellow
light
was
also
excitatory-a
fast
depolarization
and
spike
discharge.
The
origin
of
the
opponent
OFF
component
thus
appears
to
arise
from
a
direct
excitatory
input
from
OFF-center
bipolar
cells
rather
than
from
inhibition
deriving
from
lateral
interactions.
Maps
of
the
spatial
structure
of
the
yellow-OFF
and
blue-ON
fields
revealed
a
Type
2
receptive
field,
with
coextensive
ON
and
OFF
regions
(Fig.
4a).
The
distinctive
morphology
of
the
small
bistratified
cell
suggests
a
simple
circuitry
that
could
account
for
Type
2
opponency
(Fig.
4b).
A
depolarizing
input
from
the
blue-cone
bipolar
cell
would
provide
the
excitatory
S-cone
ON
field;
similarly,
an
excitatory
input
from
a
second,
OFF-cone
bipolar
type
(summing
L-
and
M-cone
input)
to
the
outer
tier
of
dendrites
could
provide
the
coextensive
yellow-OFF
field.
Preliminary
analysis
of
the
bipolar
cell
inputs
to
the
small
bistratified
cell
strongly
supports
such
a
circuit
diagram
(29).
The
density
of
the
blue-ON
small
bistratified
cells
is
consistent
with
the
spatial
resolution
of
the
S-cone
pathway,
estimated
psychophysically
(30,
31),
suggesting
that
this
pathway
is
a
major
carrier
of
S-cone
signals.
Red-Green
Opponency
and
the
Role
of
the
Midget
Circuit.
The
proposed
circuitry
underlying
the
blue-ON/yellow-OFF
b
S-cone
isolating
green
red
blue
E
X1
N4
.......
0
200
400
600
800
1000
msec
FIG.
3.
Identification
of
strong
S-cone
input
to
the
blue-ON,
small
bistratified
ganglion
cell.
(a)
Light-emitting
diode
(LED)
stimulus
waveform
is
shown
at
the
top.
Red
and
green
LEDs
are
run
in
phase
and
set
equal
in
luminance
to
the
blue
LED,
run
in
counterphase
to
give
a
blue-yellow
chromatic
modulation.
Membrane
potential
is
shown
in
the
center.
The
cell
gives
a
strong
ON
response
in
phase
with
the
modulation
of
the
blue
LED.
A
poststimulus
time
histogram
of
the
spike
discharge
averaged
over
10
sec
is
shown
at
the
bottom.
(b)
Red
and
blue
LEDs
are
set
in
counterphase
to
the
green
LED,
and
relative
amplitudes
of
all
three
are
adjusted
to
selectively
modulate
the
S-cone
signal.
The
cell
response
follows
the
phase
of
the
S-cone
excitation
(solid
sine
wave).
cell
complies
with
the
labeled
line
model
for
Type
2
cells.
The
circuitry
of
the
red-green
opponent
pathway
remains
a
mys-
tery
for
two
reasons:
(i)
it
has
been
much
more
difficult
to
study,
and
(ii)
the
labeled
line
model
has
not
been
completely
successful
in
explaining
observations.
In
early
studies
of
red-green
spectral
opponency,
the
evi-
dence
for
cone-type-specific
labeled
lines
to
the
ganglion
cell
receptive
field
center
and
surround
was
indirect.
Monochro-
matic
adapting
lights
were
used
to
reduce
preferentially
the
sensitivity
of,
say,
the
L-cone
in
an
attempt
to
observe
a
cell's
response
to
the
M-cone
in
relative
isolation.
This
approach
could
not
provide
conclusive
evidence,
but
more
recent
ex-
periments
recording
from
centrally
located
midget
cells
and
using
the
silent-substitution
method
(32,
33)
support
Hubel
and
Wiesel's
original
vision
of
a
cone-type-specific
color-
coding
circuitry.
Strong
support
for
the
labeled
line
model
also
comes
from
the
anatomy
of
central
midget
ganglion
cells.
In
the
central
7-10
degrees
of
visual
field,
each
midget
cell
receives
its
sole
excitatory
connections
from
a
single
midget
bipolar
cell,
which
in
turn
connects
to
a
single
cone.
This
now
well-established
"private
line"
explains
why
a
given
midget
cell
responds
only
to
one
cone
type
(either
L,
or
M,
or
S)
in
its
receptive-field
center
(34,
35).
The
anatomy
of
th