European Journal of Neuroscience, Vol. 17, pp.

1793±1801, 2003 ß Federation of European Neuroscience Societies

Melanopsin retinal ganglion cells and the maintenance

Ma'ayan Semo,1 Stuart Peirson,1 Daniela Lupi,1 Robert J. Lucas,1 Glen Jeffery,2 and Russell G. Foster1
1
Department of Integrative & Molecular Neuroscience, Division of Neuroscience & Psychological Medicine, Faculty of Medicine,
Imperial College London, Charing Cross Hospital, Fulham Palace Road, London W6 8RF, UK
2
University College London, Institute of Ophthalmology, Bath Street, London EC1V 9EL, UK

Keywords: mouse, photoentrainment, pupil, retinal degeneration

Abstract
Melanopsin-expressing ganglion cells have been proposed as the photoreceptors mediating non-rod, non-cone ocular responses to
light. Here we use the aged (approximately 2 years) rodless and coneless (rd/rd cl) mouse to assess the impact of progressive inner
retinal cell loss on melanopsin expression, circadian entrainment and pupillary constriction. Aged rd/rd cl mice show substantial
transneuronal retinal degeneration leaving only the ganglion cell layer and little of the inner nuclear layer. Despite this loss, quantitative
reverse transcriptase-polymerase chain reaction showed normal levels of melanopsin expression, and immunocytochemistry
demonstrated both the presence and normal cellular appearance of these cells. Furthermore, the optic nerves of the two genotypes
(rd/rd cl and ‡/‡) were not obviously different in animals older than 2 years. However, this massive level of retinal degeneration left both
pupillary and circadian responses to light intact, even in rd/rd cl mice older than 2 years. Our data provide the ®rst positive correlation
between the persistence of melanopsin-expressing cells and the maintenance of both circadian and pupillary responses to light in
the absence of rods and cones. These ®ndings, together with recent studies on melanopsin knockout mice, are consistent with the
hypothesis that melanopsin-expressing ganglion cells are photosensitive and mediate a range of irradiance-detection tasks.

Introduction
Retinal rods and cones were thought to be the only ocular photo- and approximately 508 nm (green cone) (Sun et al., 1997), and do not
receptors. However, recently work in both teleosts (Soni et al., 1998) show any signi®cant ®t to the PLR action spectrum in rd/rd cl mice.
and mammals (Freedman et al., 1999; Lucas et al., 1999) has provided Further, an action spectrum for phase-shifting circadian rhythms of
overwhelming evidence that ocular photoreception is not limited to locomotor behaviour in rd/rd cl mice implicates OP479, and argues for
rods and cones. In mammals this was ®rst demonstrated in mice a single novel photopigment regulating both the PLR and circadian
lacking all functional rods and cones. Elimination of these photo- light responses (S. Thompson, J.M. Appleford, M.W. Hankins, R.J.
receptors in rd/rd cl mice does not block circadian responses to light, as Lucas and R.G. Foster, unpublished results).
demonstrated by entrainment of circadian locomotor rhythms and by A subset of type III retinal ganglion cells (RGCs) projecting to the
acute pineal melatonin suppression. Enucleation abolishes these light circadian pacemaker within the suprachiasmatic nuclei (SCN) (Moore
effects, showing that novel ocular photoreceptors are responsible et al., 1995; Provencio et al., 1998) have been proposed as the novel
(Freedman et al., 1999; Lucas et al., 1999). photoreceptors (Berson et al., 2002). These cells in the rat are intrinsi-
Recent studies on rd/rd cl mice have shown that non-classical ocular cally light responsive, as their light-evoked depolarizations persist
photoreceptors do more than regulate the circadian clock. Both after anatomical and pharmacological isolation. An action spectrum
pupillary light responses (PLR) and acute alterations in locomotor for these responses shows a best ®t to an opsin/vitamin A photopig-
behaviour (masking) persist in the absence of rods and cones (Lucas ment template with a lmax of approximately 484 nm (Berson et al.,
et al., 2001; Mrosovsky et al., 2001). Furthermore, an action spectrum 2002), and hence is strikingly similar to OP479 of rd/rd cl mice (Lucas
for PLR in rd/rd cl mice demonstrates the involvement of a single et al., 2001). Signi®cantly, these RGCs invariably express the candidate
opsin/vitamin A-based photopigment with a lmax of approximately photopigment melanopsin (Provencio et al., 2000), and also project to
479 nm (opsin photopigment/OP479) (Lucas et al., 2001). The known the olivary pretectal nucleus (OPN), a region important in the PLR
mouse photopigments peak at approximately 360 nm (UV cone) (Gooley et al., 2001; Hattar et al., 2002).
(Jacobs et al., 1991), approximately 498 nm (rod) (Bridges, 1959) Most recently, a series of studies on melanopsin null mice has shown
that light-dependent circadian entrainment and pupillary constriction
are attenuated (Panda et al., 2002; Ruby et al., 2002; Lucas et al.,
2003). Furthermore, the intrinsic photosensitivity of melanopsin RGCs
Correspondence: Dr R. G. Foster, as above. appears to be abolished in these knockout mice (Lucas et al., 2003).
E-mail: [email protected]
Collectively these results implicate melanopsin as an important com-
Received 10 January 2003, revised 26 February 2003, accepted 28 February 2003 ponent in these irradiance responses. Here we have taken a comple-

doi:10.1046/j.1460-9568.2003.02616.x
1794 M. Semo et al.

mentary approach to study the role of RGCs in the photic regulation of Pupillometry
irradiance responses to light. Our approach takes advantage of a highly Pupillometry was undertaken in unanaesthetized age-matched C3H/
degenerate retinal model in mice (rd/rd cl), which lacks rods, cones He wild-type (n ˆ 39, 90±851 days old) and rd/rd cl mice (n ˆ 40,
and most of the inner retina. We assess the impact of progressive inner 90±855 days old). A signi®cant diurnal rhythm in the murine PLR
retinal cell loss on melanopsin expression, circadian entrainment and has been noted previously (Lucas et al., 2001). Consequently, we
pupillary constriction. restricted our measurements to 6 and 2 h before lights off (12 h
light : 12 h dark cycle). For full details, see Lucas et al. (2001). Mice
Materials and methods were adapted to the dark for 1±1.5 h. One eye of each animal was
videotaped under infrared illumination using a CCD camera ®tted
Animals
with a 140-mm lens; mice were held such that the cornea and camera
As described previously (Lucas et al., 1999), we generated a line of lens were in a parallel plane. Light stimuli (1 min) were provided by
transgenic mice that lacks both rod and cone photoreceptors. This was a quartz-halogen light and transmitted along a quartz ®bre optic to
achieved by introducing a synthetic transgene (cl) consisting of an irradiate the whole eye. The intensity and wavelength of the light
attenuated diphtheria toxin coding sequence driven by a portion of the were controlled with a monochromatic (half-bandwidth ˆ 10 nm)
human red cone opsin promoter (Wang et al., 1992; Soucy et al., 1998) interference ®lter (505 nm). The irradiance used for each constriction
into mice that lack rods (rd/rd). All procedures were conducted (50±60 mW/cm2) was measured using an optical power meter
according to the Home Of®ce regulations, under the Animals (Macam Photometrics). Pupil area was measured from images cap-
(Scienti®c Procedures) Act of 1987. Wild-type and rd/rd cl C3H/He tured from the videotaped records using TINA 3.1 image analysis
mice were maintained at 22 8C, 50% humidity in a 12 : 12 h light : dark software. To correct for individual variations in the dark-adapted
cycle (unless otherwise described) with food and water available pupil area, data were normalized to the pupil area immediately
ad libitum. preceding light onset.
Histology Immunocytochemistry
Animals were deeply anaesthetized with sodium pentobarbitone Animals were deeply anaesthetized with sodium pentobarbitone
(60 mg/kg), perfused transcardially with phosphate-buffered saline (60 mg/kg), perfused transcardially with warm (32 8C) 0.9% NaCl,
(PBS, 0.1 M, pH 7.4) followed by 2% paraformaldehyde and 2% followed by 4% paraformaldehyde in PB (pH 7.4). Eyes were removed
glutaraldehyde in 0.1 M PBS (pH 7.4). The eyes were removed and and post®xed for 24 h. Subsequently, the cornea and lens were
post®xed in the above for a minimum of 24 h. The cornea and lens were removed and the eye-cup cryoprotected (30% sucrose in PBS), agar
removed, and the eye-cup dehydrated through a graded series of embedded and sectioned at 20±25 mm on a sledge microtome. Free-
alcohols and embedded in historesin (Leica, UK). The eyes were ¯oating sections were collected in PBS-A (0.01 M PB, 0.9% NaCl,
sectioned at 5 mm and a 1-in-5 series collected, mounted onto slides, 0.1% NaN3). Sections were incubated in 1% goat serum in PBST-A
and stained with Cresyl violet before cover slipping. Optic nerves were (0.1 M PB, 0.9% NaCl, 0.3% Triton-X100, 0.1% NaN3) for 60 min at
dissociated free from the orbit and placed in chilled 2% osmium 4 8C, followed by incubation in a polyclonal primary antibody raised
tetroxide for 2 h. They were then washed in phosphate buffer (PB, against murine melanopsin epitope at a dilution 1/5000 (generously
0.1 M, pH 7.4) and dehydrated and embedded in Araldite. Semi-thin donated by Dr Ignacio Provencio) (Provencio et al., 2002). Sections
sections were cut at 1 mm, and stained with Cresyl violet. Retinal and were incubated for 72 h at 4 8C in the primary antibody, rinsed in PBST
optic nerve sections were viewed on a Zeiss Axioplan and photo- and incubated in Alexa Red 546 nm conjugated to goat antirabbit IGG
graphed digitally. at a dilution of 1/200 (Molecular Probes). Sections were washed in
Circadian behaviour PBST, mounted on clean slides and cover-slipped with PBS, and
viewed with a Zeiss LSM 500 confocal microscope, and images were
Age-matched male mice C3H/He rd/rd cl (n ˆ 38, 133±723 days old) stored. These were viewed with Zeiss LSM 5 image examiner (Version
and wild-type mice (n ˆ 37, 77±792 days old) were individually 2.8 software).
housed in running wheel cages. The cages were placed into light-tight
boxes. Electronic switches were attached to the running wheels, and Quantitative polymerase chain reaction (PCR)
the activity was recorded using ClockLab data collection system
(Actimetrics, IL, USA). Mice were entrained to a 12 : 12 h light : dark Animals
cycle (¯uorescent light, irradiance approximately 300 mW/cm2), tem- Three age groups were used, termed young (94  12 days), middle-
perature 22 8C, 50% humidity for at least 7 days. The entraining light aged (355  13 days) and old (574  4 days). Animals were killed by
was then discontinued, and the free-running activity was monitored in cervical dislocation according to Schedule 1 of the Animal (Scienti®c
constant darkness (DD) for 7±12 days. On the day of the light pulse, Procedures) Act, UK. Four wild-type and four rd/rd cl animals of each
mice were moved to the pulse equipment under infrared illumination age group were killed at ZT (zeitgeber time) 14±16. Both eyes were
and exposed to a 15-min pulse of light of a de®ned irradiance and immediately enucleated using an infrared viewing system. The eyes
wavelength (lmax at 505 nm at an irradiance of 1.0 mW/cm2 measured were hemisected to expose the retina and placed in RNAlater solution
using an optical power meter, Macam Photometrics Scotland, UK). (Ambion, UK) for RNA extraction.
Light was administered 4 h after activity onset (circadian time 16,
CT16), as described previously (Foster et al., 1991). Handling control cDNA synthesis
mice were treated in a similar fashion but given no light. Mice were Total RNA was extracted from paired whole eyes from each individual
allowed to free-run for a further 7±12 days, and the magnitudes of the in 0.5 mL of Tri-Reagent (Sigma), following homogenization in a
phase shifts were calculated with actogram analysis software from FastPrep green tube (Bio 404, Anachem Ltd, UK; FP120 homogenizer,
ClockLab, using the difference between a regression line through Q-Biogene-Alexis Ltd, UK). Total RNA was quanti®ed using an
steady state activity onsets prior to and after the pulse (Foster et al., Eppendorf biophotometer (Eppendorf UK Ltd, UK). A portion
1991). (1 mg) of total RNA was treated with RNA-free DNase I (Sigma

ß 2003 Federation of European Neuroscience Societies, European Journal of Neuroscience, 17, 1793±1801
 Melanopsin ganglion cells and light responses 1795

Aldrich, UK) for 30 min at 37 8C to eliminate any genomic DNA assessed using a two-factor ANOVA (analysis of variance) followed by
contamination, before 1 mg of RNA was reverse-transcribed with post-hoc Dunnett's t-test.
random decamers using a RetroScript kit (Ambion, UK). A control,
omitting the reverse-transcriptase, was included for each sample to Results
con®rm the absence of any contaminating DNA.
Anatomy
Primers In the rd/rd cl, loss of the outer retina is followed by progressive and
Primers were designed using MacVector software (Accelrys, UK), substantial transneuronal degeneration resulting in a thinning and
using the guidelines suggested by Applied Biosystems (Foster City, regional loss of the inner nuclear layer (INL) (Fig. 1A±D). Beyond
CA, USA). Where the genomic structure was known, primers were 9 months old, we could detect little more than the ganglion cell layer
designed to span introns to prevent genomic DNA ampli®cation. (GCL) in many regions (Fig. 1D). This progressive loss of the neural
Optimization was performed using wild-type ocular cDNA to ensure retina was associated with regional movement of pigment into the
that a single band of the correct size was produced. All bands were retina (Fig. 1C and D). Whilst patches of the INL were present in some
sequenced to con®rm identity, and melting curves were also under- regions, these were abnormally thin and the cells within them appeared
taken prior to quanti®cation to determine the optimum recording larger. In the normal retina, the INL was approximately 6 cells deep
temperature. The primers used were as follows. (Fig. 1A). In spite of individual variation, the inner retina of rd/rd cl
18S ribosomal RNA: mice was never more than 3±4 cells deep after approximately 700 days
Forward, GTAACCCGTTGAACCCCAT, old (Fig. 1B). Only the GCL in these mice appeared to be unaffected by
Reverse, CCATCCAATCGGTAGTAGCG; outer retinal loss.
acidic ribosomal phosphoprotein (ARP): Recent studies have provided evidence that a subset of cells in the
Forward, CGACCTGGAAGTCCAACTAC, retinal ganglion cell layer (RGCL) are intrinsically photosensitive.
Reverse, ATCTGCTGCATCTGCTTG; These cells express melanopsin, and as a result melanopsin has been
TATA-binding protein (TBP): proposed as the photopigment mediating these responses to light
Forward, TGGGCTTCCCAGCTAAGTTC, (Berson et al., 2002). To determine whether melanopsin-expressing
Reverse, GGAAATAATTCTGGCTCATAGCTACTG; cells remain in the aged rd/rd cl retina, we ®rst used a melanopsin-
glial ®brillary acidic protein (GFAP): speci®c antibody to identify and localize these cells. Melanopsin-
Forward, TGAGGCTGGAGGCAGAGAACAAC, containing cells were present in both wild-type controls and in rd/rd cl
Reverse, CAGTTGGCGGCGATAGTCGTTAG; mice. In both cases they were distributed equally in each of the retinal
Thy1: quadrants, and restricted to the inner retina. No staining was found in
Forward, GTCGCTCTCCTGCTCTCAGTCTTG, the outer retina in wild-type animals. Although detailed counts of
Reverse, TCATCCTTGGTGAAGTTGGC; melanopsin-positive cells were not undertaken, the number and pattern
and melanopsin: of labelling in both appeared similar. The morphological variation in
Forward, TCACAGGGATGCTGGGCAATC, melanopsin-expressing cells of the inner retina in aged (>700 days) rd/
Reverse, TTCTTGTAGAGGCTGCTGGCAAAG. rd cl mice is shown in Fig. 2A±D. Despite the variability in morphol-
All primers were synthesized by Sigma-Genosys, Sigma Aldrich. ogy, the majority of melanopsin-expressing cells resembled a type III
RGC morphology. Figure 2A and B shows melanopsin-positive cells in
Real-time PCR the GCL with processes extending into the remnant portion of the INL.
All real-time PCR reactions were conducted using a Sybr Green real- Cells with these processes were relatively common. A similar example
time PCR assay run on an Applied Biosystems PRISMTM 7700 is shown in Fig. 2C, but in this case processes not only extend into the
Sequence Detector (Applied Biosystems, CA, USA). Each reaction remnant INL, but also appear, through a range of adjacent confocal
was run in triplicate and contained 1 mL of ocular cDNA template, sections, to project towards another melanopsin-expressing cell deeper
along with 300 nM of each primer. Cycling conditions consisted of an in the retina. Figure 2D shows a melanopsin-expressing cell located
initial Taq activation step of 95 8C for 10 min, followed by 40 cycles of deeper in the retina. It was not clear whether this was a displaced
95 8C denaturation for 15 s followed by a 60 8C annealing/extension ganglion cell, amacrine cell or a cell whose location had become
step for 1 min. A ®nal data collection step of 20 s was added to each distorted by transneuronal cell loss.
cycle, with ¯uorescence measured at 2 8C below the melting tempera-
ture of the desired product as determined by melting curve analysis. Quantitative reverse transcriptase-polymerase
This ensured that any primer±dimers did not contribute to measured chain reaction (RT-PCR)
¯uorescence. The retinae of aged rd/rd cl mice were fragile due to extreme cell loss.
As a result it was not possible to make whole-mount preparations.
Data analysis Further, the nature of the retinae also precluded the reliable collection
Real-time PCR data were collected using Sequence Detector Software of serial sections. Consequently, it was not possible to quantify the
(SDS version 1.7; Applied Biosystems). The slope of each ampli®ca- number of immunopositive melanopsin cells in rd/rd cl retinae and to
tion plot was used to calculate reaction ef®ciency (Liu & Saint, 2002), compare these with the wild-type. Hence, in an attempt to quantify the
and expression of each sample was then calculated from threshold melanopsin-expressing cells within the retina, we undertook quanti-
cycle values as ef®ciency-corrected relative expression (as described tative RT-PCR for melanopsin expression in both age-matched
by Pfaf¯, 2001). 18S rRNA, ARP and TBP were used as internal genotypes (Fig. 3A). A two-factor ANOVA revealed an overall effect
controls to correct for variations in RNA and cDNA quality and of genotype (F1,15 ˆ 4.743, P < 0.05) and age (F2,15 ˆ 19.241,
quantity. A normalization factor (NF3) was calculated from the P < 0.001), and also that there is a signi®cant interaction between
geometric mean of these three controls, as suggested by Vandesompele genotype and age (F2,15 ˆ 3.895, P < 0.05). Post-hoc Dunnett's t-tests
et al. (2002). The stability indices of all three genes were also found to con®rmed the signi®cant decline in melanopsin transcript with age
be similar (all below 0.3). Signi®cant changes in expression level were (young vs. old rd/rd cl, P < 0.01; young vs. old wild-type, P < 0.05).

ß 2003 Federation of European Neuroscience Societies, European Journal of Neuroscience, 17, 1793±1801
1796 M. Semo et al.

Fig. 1. Histological sections from retinae of wild-type and rd/rd cl mice. The rd/rd cl sections show the range of retinal degeneration observed in these mice. (A)
Central retina from a wild-type mouse (approximately 700 days old) showing all three cellular layers. (B) Isolated patch of rd/rd cl central retina (approximately
700 days old) showing limited preservation of the INL. INL thickness was 21.9  2.6 mm (mean  SD) in the central retina of wild-type mice, whilst the patch of INL
shown from the rd/rd cl retina was 17.3  4.1 mm (mean  SD). Mann±Whitney tests showed this to be a signi®cant difference (P < 0.01). Furthermore, cell size in the
INL varied markedly between the genotypes, being larger in the rd/rd cl mice. The average number of cells found across the depth of the INL was 5±6 in the wild-type,
but only 3±4 at best in the rd/rd cl where present. (C) Focal retinal loss in the rd/rd cl mouse associated with migration of pigment from the retinal pigment epithelium.
(D) Substantial retinal thinning in the rd/rd cl mouse. Showing only the preservation of the GCL. Scale bar, 50 mm. GCL, ganglion cell layer; INL, inner nuclear layer;
ONL, outer nuclear layer.

There is signi®cantly more melanopsin in the young rd/rd cl vs. young of Thy1 (Fig. 3D). With this normalization there was no statistical
wild-types (P < 0.01), but at other ages the transcript levels are difference in the levels of melanopsin expression within the retinae of
equivalent (Fig. 3A). rd/rd cl and wild-type mice at any age.
GFAP up-regulation in MuÈller cells has been associated with retinal Collectively, the quantitative RT-PCR analysis showed that despite
degeneration (de Raad et al., 1996; Humphrey & Moore, 1996). the loss of most of the outer retina, levels of melanopsin expression
Hence, GFAP transcript levels were also quanti®ed in the ageing remained similar in rd/rd cl and wild-type mice. This ®nding is
rd/rd cl and wild-type retina. A two-factor ANOVA revealed an overall consistent with our observations that the optic nerves of the two
effect of genotype (F1,18 ˆ 38.766, P < 0.001), and post-hoc Dun- genotypes do not grossly differ. Figure 4 A and B shows representative
nett's t-tests revealed that at all ages the transcript level of GFAP is cross-sections of the optic nerves from both groups. The nerves were of
approximately 3.5-fold higher in the rd/rd cl than wild-type (wild- a similar size, and the axons appeared normal in both. Further, there
type young vs. rd/rd cl young, P < 0.01; wild-type middle-aged vs. was no obvious increase in the area of the tissue occupied by glia in the
rd/rd cl middle-aged, P < 0.05; wild-type old vs. rd/rd cl old, rd/rd cl, which would normally be associated with axonal loss in the
P < 0.01), suggesting a chronic long-term pathology throughout optic nerve (Sandell & Peters, 2002). Given the above, it seems
the life of the rd/rd cl. Sites of potential gliosis are shown in Fig. 1C unlikely that there has been any signi®cant RGC loss in the aged
and D. rd/rd cl mouse.
Thy1 expression has been used as a general marker to identify rodent
RGCs (Barnstable & Drager, 1984). Thy1 expression was determined Pupillary and circadian responses to light
in rd/rd cl and wild-type mice in an attempt to assess any age- and Our previous studies have shown that both pupillary constriction and
genotype-related RGC loss. However, there is an inherent problem in circadian responses to light are preserved in young (approximately
comparing the transcript levels of wild-type and rd/rd cl retinae in that 100 days old) rd/rd cl mice (Freedman et al., 1999; Lucas et al., 2001).
the transcripts from ganglion cells will represent a greater proportion These assays of retinal function were used in the present study to
of the total cell number in the rd/rd cl mRNA pool than in the wild- examine whether a more complete lesion of the inner retina would
type. Consistent with this notion, Thy1 levels were signi®cantly higher attenuate or abolish these responses to light. Despite the near complete
in the retinae of young rd/rd cl compared with wild-type mice loss of the inner retina, both the phase-shifting effects of light on
(P < 0.05). As a result, we normalized melanopsin levels to those circadian locomotor rhythms and pupillary constriction were main-

ß 2003 Federation of European Neuroscience Societies, European Journal of Neuroscience, 17, 1793±1801
 Melanopsin ganglion cells and light responses 1797

Fig. 2. (A±D) Confocal images showing melanopsin-immunopositive cells from an rd/rd cl mouse aged 752 days. Single nonstacked false colour images of a 1-mm
optical slice through a 25-mm transverse retinal section. (A and B) Melanopsin-immunopositive cells in the RGCL showing processes extending into the remnant
portion of the INL. (C) Melanopsin-immunopositive cell in the RGCL with processes extending towards another melanopsin-expressing cell deeper in the retina. (D)
Melanopsin-expressing cell located deeper in the retina that may be a displaced ganglion cell, amacrine cell or a cell whose location had become distorted by
transneuronal cell loss. Scale bar, 50 mm.

tained without signi®cant age-related attenuation in mice over These ®ndings were con®rmed and extended in the present study for
700 days old (Figs 5A and B, and 6A and B). both young and aged individuals (Fig. 5). By contrast, pupillary
For both our circadian and pupillary responses to light we selected responses to light are attenuated in rd/rd cl mice by approximately
irradiance levels that were nonsaturating for rd/rd cl mice. We have 2±3 log units (Lucas et al., 2001). The irradiance levels used for
shown that the phase-shifting responses of rd/rd cl mice are not pupillary constriction in the current experiments would fully constrict
statistically different from wild-type mice (Thompson et al., 2003). the wild-type pupil but leave the rd/rd cl pupil only partially con-

ß 2003 Federation of European Neuroscience Societies, European Journal of Neuroscience, 17, 1793±1801
1798 M. Semo et al.

Fig. 3. Quantitative RT-PCR from young (94  12 days), middle-aged (355  13 days) and old (574  4 days) wild-type and rd/rd cl eyes. Filled bars rd/rd cl; un®lled
wild-type. All changes in expression level were assessed using a two-factor ANOVA followed by a post-hoc Dunnett's t-test. (A) Levels of melanopsin transcript show a
signi®cant age-related decline in both genotypes (rd/rd cl, P < 0.01; wild-type, P < 0.05) There is signi®cantly more melanopsin in the young rd/rd cl vs. young wild-
types (P < 0.01), but at other ages the transcript levels are equivalent. (B) GFAP expression was approximately three±®vefold greater in the rd/rd cl eye. Wild-type
young vs. rd/rd cl young, P < 0.01; wild-type middle-aged vs. rd/rd cl middle-aged, P < 0.05; old wild-type vs. old rd/rd cl, P < 0.01. (C) Thy1 expression in the
retinae of rd/rd cl and wild-type mice ANOVA revealed only an overall effect of genotype (F1,18 ˆ 10.308, P < 0.05), Post-hoc t-tests revealed that there is only
signi®cantly more Thy1 in the young rd/rd cl vs. the young wild-type, P < 0.05). (D) Melanopsin normalized to Thy1. No statistical difference in the levels of
melanopsin expression was observed between the eye of rd/rd cl and wild-type mice (see text for details).

Fig. 4. Representative transverse optic nerve sections (1 mm) taken from similar distal locations stained with Cresyl violet. (A) Wild-type 723 days old. (B) rd/rd cl
743 days old. There were no obvious qualitative differences in the size of the nerves or in the number of axons that they contained. Scale bar, 10 mm.

stricted. As a result, the levels of pupillary constriction recorded from INL appeared abnormal in cell size and distribution. By contrast, the
rd/rd cl mice were never as great as those for wild-type mice (Fig. 6A± RGC layer of rd/rd cl mice appeared largely unaffected by the loss of
C). Despite this fact, there was no change in the dark-adapted pupil most of the rest of the retina. Expression levels of Thy1 (a marker of
area with age or between genotype (Fig. 6C). rodent ganglion cells) did not differ between aged rd/rd cl and wild-
type mice, the optic nerves of the two genotypes were of a similar size
and the axon appeared normal. Previous reports on other retinally
Discussion
degenerate mouse models (e.g. rd/rd and rds) have reported INL
We have used the most advanced model of retinal degeneration yet degeneration as a consequence of substantial photoreceptor loss
described (rd/rd cl) to study non-rod, non-cone ocular photoreception. (Sorsby et al., 1954; Blanks et al., 1974; Sanyal et al., 1980), but
The retinae of rd/rd cl mice at their natural life span of approximately the extent of cell loss was far less extensive than that observed in rd/rd
2 years old show extreme transneuronal degeneration. No cells were cl mice.
present in the outer retina, and there was serious thinning of the INL, Despite the complete loss of the outer retina and most of the inner
and in many regions of the retina its complete absence. Even when retina in aged (approximately 2 years old) rd/rd cl mice, circadian
patches of the INL were present, the limited cellular population of the behaviour was indistinguishable from congenic wild-type animals.

ß 2003 Federation of European Neuroscience Societies, European Journal of Neuroscience, 17, 1793±1801
 Melanopsin ganglion cells and light responses 1799

Fig. 5. (A) Magnitude of phase delays in the onset of running wheel activity following a pulse of light given in the early part of the subjective night. Filled bars rd/rd
cl; un®lled wild-type. GLIMTM (Generalized Linear Interactive Modelling; version 3.77 ß 2003 1985 Royal Statistical Society London) software was used to
perform analysis of covariance. This showed that the age of the animal was insigni®cant in explaining the observed variance. Further, the average phase shifts of the
two genotypes (of all ages) were not signi®cantly different from each other: wild-type 115  8 min (mean  SEM), rd/rd cl 125  7 min (mean  SEM) (F on
deletion ˆ 0.36 < F1,77 from tables ˆ 4.0 at P ˆ 0.05). These data show that old rd/rd cl mice retain the ability to phase shift and that the magnitudes of phase shifts are
not altered with age and are not signi®cantly different from the wild-type mice. (B) Representative running wheel actograms from an aged rd/rd cl and wild-type
mouse. The vertical bar on the right indicates the lighting condition; each line of the actogram represents 24 h. Mice were ®rst entrained to a 12 : 12 h light : dark cycle.
The entraining light was then discontinued and the free-running activity was monitored in constant darkness (DD). On the 11th day in darkness at CT16 they were
exposed to a 15-min light pulse lmax ˆ 505 nm at a subsaturating irradiance of 1.0 mW/cm2. Mice were allowed to free-run for a further 11 days, and the magnitudes of
the phase shifts were calculated by actogram analysis software from ClockLab, using the difference between regression lines through steady state activity onsets prior
to and after the pulse (Foster et al., 1991). The wild-type shows a phase delay of 113 min, the rd/rd cl of 135 min.

Similarly there was no decline in the overall sensitivity of the pupillary If the melanopsin-expressing RGCs are indeed responsible for
light re¯ex in ageing rd/rd cl mice. Thus, the loss of inner retinal mediating the behavioural effects of light on both circadian and
neurons does not result in attenuated circadian or pupillary photo- pupillary responses, then the prediction would be that this class of
sensitivities. The maintenance of these responses does, however, RGCs would be largely unaffected in the aged rd/rd cl mice. Quanti-
correlate with the preservation of the RGC layer, suggesting that this tative RT-PCR analysis showed that levels of melanopsin expression
cell layer houses the non-rod, non-cone photoreceptors. Direct support within the retina were statistically indistinguishable between rd/rd cl
for this conclusion comes from recent studies showing that a subset of and wild-type mice. These ®ndings, together with the results from the
RGCs that express melanopsin are intrinsically photosensitive (Berson recent studies on melanopsin knockout mice (Panda et al., 2002; Ruby
et al., 2002). It should be noted that the biochemical role of melanopsin et al., 2002; Lucas et al., 2003), would support the hypothesis that
within the intrinsically photosensitive ganglion cells remains unclear. melanopsin RGCs do indeed regulate circadian and pupillary
This opsin-like protein has yet to be shown to form a functional responses to light.
photopigment, and both its amino acid identity and genomic structure Although the ®ndings in this paper would support the overall
differ markedly from the characterized photosensory opsins (Belling- conclusion of the melanopsin knockout studies, there is one important
ham & Foster, 2002; Foster & Bellingham, 2002). Despite the ambig- difference. The results in this paper suggest that melanopsin RGCs are
uous role of this protein, it remains, to date, the only reliable marker of suf®cient, in the absence of rods and cones, to maintain apparently
the intrinsically photosensitive non-rod, non-cone retinal neurons. normal circadian responses to light. By contrast, the melanopsin

ß 2003 Federation of European Neuroscience Societies, European Journal of Neuroscience, 17, 1793±1801
1800 M. Semo et al.

Fig. 6. (A) Pupillary constriction in ageing rd/rd cl and wild-type mice. (A) Average maximum pupil constriction attained during 1 min of light exposure
(lmax ˆ 505 nm, irradiance ˆ 50±60 mW/cm2). Filled bars rd/rd cl; un®lled wild-type. GLIMTM software was used for the analysis of covariance. This showed that the
age of the animal was insigni®cant in explaining the observed variance. But the maximum pupil constrictions attained between the two genotypes (of all ages) were
signi®cantly different from each other (see Results): rd/rd cl at 81.6% constriction is signi®cantly less than the wild-type constriction at 94.3% (see text for details). (F
on deletion ˆ 82 > F1,77 from tables ˆ 11.7 at P ˆ 0.001). (B) PLR in old (600±800 days old) and young (100±200 days old) wild-type and rd/rd cl mice. The area of
the pupil is normalized to its size immediately preceding lights on. Both reach maximum constriction within 15 s (n ˆ numbers of animals in each group). (C)
Maximum (dark-adapted) pupil area is not affected by genotype or age (analysis using a two-factor ANOVA). Minimum pupil area (attained during light
administration) is not affected by age in either genotype, but is signi®cantly larger in the rd/rd cl (two-factor ANOVA, followed by post-hoc Dunnett's t-test, P < 0.001),
n as in (B). The rd/rd cl is capable of equivalent constrictions at higher irradiances (Lucas et al., 2001).

knockout results show that circadian responses to light are merely presented in this paper. Finally, it is also possible that the melanopsin
attenuated (Panda et al., 2002; Ruby et al., 2002). The basis for this RGCs are not the only photoreceptor cells of the RGC layer. If this
discrepancy is unclear, but suggests that the precise role of melanopsin were the case, then rd/rd cl mice with an intact RGC layer would be
RGCs in normal circadian entrainment is likely to be complex. For expected to show normal circadian responses to light, whilst mela-
example, under normal circumstances it is possible that the rods and nopsin knockout mice would show attenuated circadian responses.
cones provide the primary light input to the SCN, with the melanopsin This possibility is also accommodated by the melanopsin knockout and
RGCs acting as a supplementary source of light information. This rd/rd cl results. In this context it is also worth noting that the teleost
would be consistent with the results from the melanopsin knockout retina contains two potential inner retinal vitamin A-based opsin
studies, but would not be predicted from the rd/rd cl results in this photopigments (OPs) (Soni et al., 1998; Bellingham et al., 2002).
paper. Alternatively, the melanopsin RGCs might provide the primary Furthermore, cryptochromes (CRYs) expressed in the INL and GCL of
source of light information to the SCN under normal circumstances, mice have been proposed as vitamin B2-based photopigments (Miya-
but in their absence, the rods and cones might be recruited to this task. moto & Sancar, 1998; Van Gelder et al., 2003). However, the CRYs
This possibility would accommodate both the melanopsin knockout have not yet been shown to form light-sensitive compounds, and the
results (Panda et al., 2002; Ruby et al., 2002) and the ®ndings absorption spectrum of the CRYs does not match the action spectra

ß 2003 Federation of European Neuroscience Societies, European Journal of Neuroscience, 17, 1793±1801
 Melanopsin ganglion cells and light responses 1801

derived from rd/rd cl mice (Lucas & Foster, 1999). In addition, it is Humphrey, M.F. & Moore, S.R. (1996) Microglial responses to focal lesions of
dif®cult to interpret the impact of CRY knockouts, as these proteins the rabbit retina: correlation with neural and macroglial reactions. Glia, 16,
325±341.
appear to have several functions, not least they form a critical Jacobs, G.H., Neitz, J. & Deegan, J.F. (1991) Retinal receptors in rodents
component of the molecular clock (Shearman et al., 2000). maximally sensitive to ultraviolet light. Nature, 353, 655±656.
In conclusion, this study provides the ®rst positive correlation Liu, W. & Saint, D.A. (2002) A new quantitative method of real time reverse
between the persistence of melanopsin-expressing ganglion cells transcription polymerase chain reaction assay based on simulation of poly-
merase chain reaction kinetics. Anal. Biochem., 302, 52±59.
and the maintenance of both circadian and pupillary responses to Lucas, R.J., Douglas, R.H. & Foster, R.G. (2001) Characterization of an ocular
light. Our results are consistent with the hypothesis that melanopsin- photopigment capable of driving pupillary constriction in mice. Nature
expressing ganglion cells are photosensitive and are likely to provide Neurosci., 4, 621±626.
the irradiance information that mediates non-rod, non-cone ocular Lucas, R.J. & Foster, R.G. (1999) Circadian rhythms: something to cry about?
Curr. Biol., 9, 214±217.
responses to light in the rd/rd cl retinal mutation. However, the precise
Lucas, R.J., Freedman, M.S., Munoz, M., Garcia-Fernandez, J.M. & Foster,
role of both melanopsin and the melanopsin RGCs in regulating non- R.G. (1999) Regulation of the mammalian pineal by non-rod, non-cone,
rod, non-cone photoreception remains unclear. Our results also show ocular photoreceptors. Science, 284, 505±507.
that the RGC layer is remarkably resistant to outer and inner retinal Lucas, R.J., Hattar, S., Takao, M., Berson, D.M., Foster, R.G. & Yau, K.-W.
neuronal loss, and that photoreception is still possible in a retina with (2003) Diminished pupillary light re¯ex at high irradiances in melanopsin-
knockout mice. Science, 299, 245±247.
extreme neuronal degeneration. Such ®ndings are relevant to the study Miyamoto, Y. & Sancar, A. (1998) Vitamin B2-based blue-light photoreceptors
of human retinal pathologies, and the appreciation that visual blindness in the retinohypothalamic tract as the photoactive pigments for setting the
need not result in the loss of all the photoreceptive capabilities of circadian clock in mammals. Proc. Natl. Acad. Sci. USA, 95, 6097±6102.
the eye. Moore, R.Y., Speh, J.C. & Card, J.P. (1995) The retinohypothalamic tract
originates from a distinct subset of retinal ganglion cells. J. Comp. Neurol.,
352, 351±366.
Acknowledgements Mrosovsky, N., Lucas, R.J. & Foster, R.G. (2001) Persistence of masking
responses to light in mice lacking rods and cones. J. Biol. Rhythms, 16,
This work was supported by The Wellcome Trust, and a Biotechnology and 585±587.
Biological Sciences Research Council (BBSRC) studentship to M. S. We thank Panda, S., Sato, T.K., Castrucci, A.M., Rollag, M.D., DeGrip, W.J., Hogenesch,
Dr Ignacio Provencio (Department of Anatomy, Physiology, & Genetics J.B., Provencio, I. & Kay, S.A. (2002) Melanopsin (Opn4) requirement for
Uniformed Services University, USA) for his generous donation of melanopsin normal light-induced circadian phase shifting. Science, 298, 2213±2216.
antibodies. Pfaf¯, M.W. (2001) A new mathematical model for relative quanti®cation in
real-time RT-PCR. Nucl. Acids Res., 29, E45±E45.
Provencio, I., Cooper, H.M. & Foster, R.G. (1998) Retinal projections in mice
Abbreviations with inherited retinal degeneration: implications for circadian photoentrain-
ment. J. Comp. Neurol., 395, 417±439.
ARP, acidic ribosomal phosphoprotein; CRY, cryptochrome; CT, circadian Provencio, I., Rodriguez, I.R., Jiang, G., Hayes, W.P., Moreira, E.F. & Rollag,
time; GCL, ganglion cell layer; GFAP, glial ®brillary acidic protein; INL, M.D. (2000) A novel human opsin in the inner retina. J. Neurosci., 20,
inner nuclear layer; OP, opsin photopigment; OPN, olivary pretectal nucleus; 600±605.
PB, phosphate buffer; PBS, phosphate-buffered saline; PLR, pupillary light Provencio, I., Rollag, M.D. & Castrucci, A.M. (2002) Photoreceptive net in the
responses; RGCs, retinal ganglion cells; RGCL, retinal ganglion cell layer; RT- mammalian retina. Nature, 415, 493.
PCR, reverse transcriptase-polymerase chain reaction; SCN, suprachiasmatic de Raad, S., Szczesny, P.J., Munz, K. & Reme, C.E. (1996) Light damage in the
nucleus; TBP, TATA-binding protein; ZT, zeitgeber time. rat retina: glial ®brillary acidic protein accumulates in Muller cells in
correlation with photoreceptor damage. Ophthalmic Res., 28, 99±107.
References Ruby, N.F., Brennan, T.J., Xie, X., Cao, V., Franken, P., Heller, H.C. & O'Hara,
B.F. (2002) Role of melanopsin in circadian responses to light. Science, 298,
Barnstable, C.J. & Drager, U.C. (1984) Thy-1 antigen: a ganglion cell speci®c 2211±2213.
marker in rodent retina. Neuroscience, 11, 847±855. Sandell, J.H. & Peters, A. (2002) Effects of age on the glial cells in the rhesus
Bellingham, J. & Foster, R.G. (2002) Opsins and mammalian photoentrain- monkey optic nerve. J. Comp. Neurol., 445, 13±28.
ment. Cell Tiss. Res., 309, 57±71. Sanyal, S., De Ruiter, A. & Hawkins, R.K. (1980) Development and degenera-
Bellingham, J., Whitmore, D., Philp, A.R., Wells, D.J. & Foster, R.G. (2002) tion of retina in rds mutant mice: light microscopy. J. Comp. Neurol., 194,
Zebra®sh melanopsin: isolation, tissue localisation and phylogenetic posi- 193±207.
tion. Mol. Brain Res., 107, 128±136. Shearman, L.P., Sriram, S., Weaver, D.R., Maywood, E.S., Chaves, I., Zheng,
Berson, D.M., Dunn, F.A. & Takao, M. (2002) Phototransduction by retinal B., Kume, K., Lee, C.C., van der Horst, G.T.J., Hastings, M.H. & Reppert,
ganglion cells that set the circadian clock. Science, 295, 1070±1073. S.M. (2000) Interacting molecular loops in the mammalian circadian clock.
Blanks, J.C., Adinol®, A.M. & Lolley, R.N. (1974) Photoreceptor degeneration Science, 288, 1013±1019.
and synaptogenesis in retinal-degenerative (rd) mice. J. Comp. Neurol., 156, Soni, B.G., Philp, A.R., Knox, B.E. & Foster, R.G. (1998) Novel retinal
81±94. photoreceptors. Nature, 394, 27±28.
Bridges, C. (1959) The visual pigments of some common laboratory animals. Sorsby, A.P., Koller, P.C., Att®eld, M., Davey, J.B. & Lucas, D.R. (1954)
Nature, 184, 727±728. Retinal dystrophy in the mouse: histological and genetic aspects. J. Exp.
Foster, R.G. & Bellingham, J. (2002) Opsins and melanopsins. Current Biol., Zool., 125, 171±197.
12, 543±544. Soucy, E., Wang, Y., Nirenberg, S., Nathans, J. & Meister, M. (1998) A novel
Foster, R.G., Provencio, I., Hudson, D., Fiske, S., DeGrip, W. & Menaker, M. signaling pathway from rod photoreceptors to ganglion cells in mammalian
(1991) Circadian photoreception in the retinally degenerate mouse (rd/rd). retina. Neuron, 21, 481±493.
J. Comp. Physiol. [A], 169, 39±50. Sun, H., Macke, J.P. & Nathans, J. (1997) Mechanisms of spectral tuning in the
Freedman, M.S., Lucas, R.J., Soni, B., von Schantz, M., Munoz, M., David- mouse green cone pigment. Proc. Natl Acad. Sci. USA, 94, 8860±8865.
Gray, Z.K. & Foster, R.G. (1999) Regulation of mammalian circadian Van Gelder, R.N., Wee, R., Lee, J.A. & Tu, D.C. (2003) Reduced pupillary light
behavior by non-rod, non-cone, ocular photoreceptors. Science, 284, responses in mice lacking cryptochromes. Science, 299, 222.
502±504. Vandesompele, J., De Preter, K., Pattyn, F., Poppe, B., Van Roy, N., De Paepe,
Gooley, J.J., Lu, J., Chou, T.C., Scammell, T.E. & Saper, C.B. (2001) Mela- A. & Speleman, F. (2002) Accurate normalization of real-time quantitative
nopsin in cells of origin of the retinohypothalamic tract. Nature Neurosci., RT-PCR data by geometric averaging of multiple internal control genes.
12, 1165. Genome Biol., 3, RESEARCH0034 1-0034. 11.
Hattar, S., Liao, H.W., Takao, M., Berson, D.M. & Yau, K.W. (2002) Mela- Wang, Y., Macke, J., Merbsl, S., Zack, D., Klaunberg, B., Bennett, J., Gearhart,
nopsin-containing retinal ganglion cells: architecture, projections, and intrin- J. & Nathans, J. (1992) A locus control region adjacent to the human red and
sic photosensitivity. Science, 295, 1065±1070. green visual pigment genes. Neuron, 9, 429±440.

ß 2003 Federation of European Neuroscience Societies, European Journal of Neuroscience, 17, 1793±1801