Targeted Destruction of Photosensitive Retinal Ganglion Cells with a Saporin Conjugate Alters the Effects of Light on Mouse Circadian Rhythms

In order to show specific ablation of melanopsin expressing cells, RGC-5 cells stably expressing melanopsin were exposed to the UF008/SAP conjugate. After 4 days of exposure, the cells were killed in a dose-dependent manner. The highest concentration of UF008/SAP (1000 pg UF008/SAP /µl) caused maximum cell death whereas the same concentration of the non-immunized IgG/SAP conjugate did not cause any cell death (Figure 1). Also, RGC-5 cells that do not express melanopsin were not affected by the UF008/SAP conjugate, even at the highest dose tested.

Injection of the UF008/SAP conjugate into the vitreous of adult C57BL/6J mouse eyes killed ipRGCs in a dose-dependent manner (Figure 2A and B), the curve becoming asymptotic at approximately 400 ng/eye. All control retinas from various dose groups had similar melanopsin cell densities and were combined. Relative to the average control retina, about 57% of melanopsin cells were killed by 400 ng UF008/SAP per eye which was not significantly different than the cell death achieved with the highest dose group (800 ng/eye). Representative images from each dose group (Figure 2A) show the decreasing number of immunopositive cells in retinas exposed to increasing UF008/SAP amounts. A time course study showed that maximal melanopsin cell death occurs about 2 weeks after 400 ng UF008-SAP injection (Figure 2C). Each group was significantly different from controls. Retinas of the 4 week group that received 400 ng/eye UF008/SAP were analyzed for regional differences in the efficacy of the conjugate. Analysis of peripheral to central regional variation in UF008/SAP-dependent melanopsin cell death in the mouse retina revealed that the density of remaining cells was uniform indicating no regional differences (Figure 3).

There are various ways of assessing retinal health, but relative thickness of the outer nuclear layer (ONL) has been used previously as a direct measurement of the degree of photoreceptor cell death [11], [12]. Figure 4A and B shows that neither relative ONL nor relative inner nuclear layer (INL) thickness of retinas from mice receiving 800 ng/eye UF008/SAP differed from corresponding measurements from PBS-injected retinas. Because there was no significant effect of the highest dose, no effects were expected with the lower doses.

Exposure of the retina to UF008/SAP does not affect its tissue morphology, as non-melanopsin cells remain intact. Cellular analysis in Figure 4C confirms that UF008/SAP (400 ng/eye) causes the loss of intrinsically photoreceptive ganglion cells (ipRGCs), but does not appear to alter other structural characteristics of the mouse retina. A normal distribution of glial fibrillary acid protein (GFAP; labels Muller cells), calbindin- (CALB; labels horizontal cells) and choline acetyltransferase (ChAT; labels cholinergic amacrine cells) immunoreactivity was observed in both control and UF008/SAP injected eyes.

Anatomical evaluation of the RHT terminal plexus in the SCN was performed in mice that had received unilateral UF008/SAP injections. The lesioned eye was given an intra-vitreous injection of cholera toxin subunit B (CT-B) conjugated to Alexa488 and the intact eye was injected with CT-B/Alexa594 conjugate. Ipsi- and contralateral retinas contribute equally to mouse SCN innervation [13], allowing the contributions of lesioned and intact retinas to be compared. Ablation of ipRGCs yielded an unexpected topography of remaining SCN retinal innervation. Fluorophore-conjugated CT-B labeled a dense corona of terminals along the lateral and ventral SCN border, with very sparse innervation centrally and medially. The normal retinal input occupied the central/medial area, as well as overlapping the more lateral terminal corona (Figure 5).

Having revealed there are no significant morphological differences between retinas dissected from UF008/SAP injected and control eyes, we subjected mice to a visual cliff task, in order to test general visual function. Figure 6 shows that bilaterally sighted and injected mice differ slightly in their responses to a visual cliff test, but both groups perform significantly above 50% chance levels. Mice were also given a light-dark preference test that revealed no difference between UF008/SAP-injected and control mice with respect to the time spent in the dark or the number of times the mice entered the lighted chamber (65.2±3.6 vs 71.6±5.5% time in the dark and (279±53 vs 185±40 entries into the light for UF008/SAP and controls, respectively; the corresponding values for 9 blind controls were 49.7±4.3 and 316±45, both values significantly different from the other groups, p<.02). When the comparison with controls was limited to the 6 mice that were free-running under LD, there was also no difference between groups.

Mice in the rhythm regulation studies were unilaterally injected with 400 ng UF008/SAP or IgG/SAP and contralaterally enucleated (to reduce overall variability in ipRGC ablation) under LD12∶12 conditions before being subjected to various lighting regimes. Figure 7A shows actograms representative of behavior across four different lighting conditions. None of the mice lost entrainment in LD12∶12 as an immediate consequence of UF008/SAP treatment. When switched to a graded photoperiod in which the irradiance gradually declined to zero (LD15G∶9; Figure 7B), all animals entrained, but with markedly different phase angles between groups (Figure 7C). Large differences in phase angle of entrainment were related to the correspondingly large differences in irradiance at the time of activity onset (Figure 7B). In constant darkness (DD), the circadian periods of the two groups did not differ. In constant light (LL), the circadian period of UF008/SAP mice did not change, whereas for control mice the period lengthened, as expected (Figure 7D). When the mice were returned from LL to the original LD12∶12 (without regard to the phase of the individual mice), 9 of 10 UF008/SAP mice required 16 or more days to re-entrain (median = 24 days), whereas every control animal required 14 or fewer days (median = 3 days). Subsequent histology showed that, on the average, UF008/SAP treatment yielded approximately 80% loss of melanopsin cells. Among the UF008/SAP-treated mice, the 3 individuals with the highest densities of remaining ipRGCs required 0, 16 and 16 days to re-entrain (corresponding to 81, 27 and 21 cells/mm², respectively). Four of the UF008/SAP-injected mice never re-entrained. The density of melanopsin expressing RGCs averaged 15.2 cells/mm² in these four animals.

Figure 7E shows that constant light exposure induces a suppression of total wheel-running activity (masking) in control (IgG/SAP) mice, but not in UF008/SAP injected mice. Entrained, UF008/SAP-injected mice had relatively normal masking in response to a 1 hr light pulse, but recovered more slowly than expected at the end of the pulse (Figure 8). In contrast, UF008/SAP mice that fail to entrain also do not show masking to a 1 hr light pulse administered about 30 minutes after activity onset.

Three separate investigations using different methods have now demonstrated that intact melanopsin cells are necessary in order for the mouse non-image forming visual system to function properly ([22], [23] and this report). Cell ablation greatly reduces the size of the RHT terminal field. In all likelihood, it is this specific change that accounts for the absence of light-induced phase shifts and masking. However, an alternative explanation is possible. Melanopsin cells have dendro-dendritic synaptic contacts with amacrine cells [24], [25] and may transmit photic information to them [26]. Therefore, loss of melanopsin cells might be interrupting necessary transmission of information to the SCN that would pass through amacrine cells in contact with non-melanopsin-containing retinal ganglion cells that also project to the SCN.

The present data (Figure 7C) suggest the possibility that there may be a sharply delineated minimum melanopsin cell density that is required for normal non-image forming visual responses. Potential melanopsin cell photoreceptive network properties may be lost when cell density falls below approximately 25 cells/mm². A similar, sharp threshold effect on rat working memory has been observed following saporin-induced destruction of basal forebrain cholinergic cells [27]. In that instance, memory was impaired only if an approximate 75% cell destruction threshold was exceeded.

The mouse RHT provides bilaterally symmetrical innervation to the SCN [28]. This characteristic has enabled us to compare, in individual mice, the terminal distributions from a normal retina with those from an ipRGC-ablated retina. The results indicate a specialized topography of the RHT terminal field that remains after melanopsin cell cell death. Whether the particular distribution of terminals observed is of functional importance remains to be seen. However, it is not associated with any previously described sector of the mouse SCN based on distributions of peptidergic cell types, clock gene expression or afferent terminal fields.

The thalamic intergeniculate leaflet (IGL) is a component of the circadian system particularly noteworthy because of the robust projection it provides to the SCN. The IGL also receives direct photic input from a predominantly contralateral retinal projection [29]. In mice sustaining UF008/SAP-induced ablation of melanopsin cells, retinal projections to the contralateral IGL are virtually eliminated. This suggests that all or nearly all retinal projections to the contralateral IGL arise from melanopsin cells. A functional implication concerns the role of the IGL as a mediator of circadian system response to tonic light exposure. In the hamster, loss of the IGL reduces the period lengthening effect of LL by approximately the same amount as loss of melanopsin photopigment does in the mouse [20], [30]. Classical photoreceptors account for the remaining effect of LL on period length and may exert their action directly on the SCN via the RHT.

Our results are consistent with those from two earlier studies (Table 1) employing different methods to reach similar conclusions about the essential role played by melanopsin-expressing retinal ganglion cells in the regulation of certain non-image forming visual responses [22], [23]. All three clearly indicate that the absence of melanopsin cells causes a major deficiency of entrainment to LD cycles and a lack of period lengthening in LL. Collectively, the three sets of data demonstrate that melanopsin cells are not necessary for at least rudimentary image forming vision as indicated by response to a wide variety of visual tasks. Unexpectedly, our light-dark preference test, which is unlikely to involve either image formation or optomotor reflexes, was performed equally well by both sighted and melanopsin cell deficient mice, implying that this is behavior relies on different neural mechanisms than used for simple irradiance detection (cf., [31].

With respect to circadian rhythm regulation, our results are most similar with those of Hatori and colleagues [22] who also used a procedure that acutely killed melanopsin cells in adult mice. In contrast, Guler and colleagues [23] observed effects (longer circadian period in DD, gradual loss of melanopsin cell projections) that are likely the consequence of altered circadian system development. The inducible lesion procedure applicable to adult mice developed by Hatori et al. appears to be the most consistently successful of the three methods used to kill melanopsin cells. In particular, all mice with diphtheria toxin activated melanopsin cell death rapidly lost entrainment and exhibited free-running rhythms. Nevertheless, as is true for the other procedures, this method also fails to kill all melanopsin cells.

The advantages of the UF008/SAP procedure are two-fold. First, we have shown that complete ablation of retinal ganglion cells is not required to observe massive deficits in non-image forming visual behavior of mice, such as circadian rhythm entrainment and masking. Second, this technique has broad potential applicability across species, limited only by the availability of appropriate antibodies that can be conjugated to saporin, creating an melanopsin cell-specific immunotoxin. Finally, this is the first report of a functional immunotoxin made with an antibody to a G protein-coupled receptor. Previous toxins to G protein-coupled receptors have relied on the ligand for targeting and internalization of a toxin [32] –[34]. Antibody targeting of cell-surface proteins has become an effective tool in cancer therapy [35] –[37], and these data indicate that targeting G protein-coupled receptors with antibodies may likewise be useful.

UF008, an anti-melanopsin polyclonal antibody raised in rabbit against the 15 N-terminal extracellular amino acids of mouse melanopsin, was covalently conjugated to saporin (Advanced Targeting Systems, San Diego, California). Synthesis was as described previously [38].

The RGC-5-mWT-2 cell line was established by stably transfecting RGC-5 cells using calcium phosphate precipitation [39] with a ScaI linearized vector containing the mOpn4 open reading frame (gi∶14349304) in parental vector pcDNA3.1/Zeo(+) (Invitrogen, Carlsbad, CA). Melanopsin expression was confirmed by immunocytochemistry. The RGC-5 cell line was a generous gift of Dr. Neeraj Agarwal [40]. It was derived by transformation of neonatal retinal cells with psi2 E1a, a retroviral vector containing the 12S E1A adenoviral oncogene. RGC-5 cells express mRNA markers characteristic of retinal ganglion cells. Cells were maintained in DMEM (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (Atlanta Biologicals, Norcross, GA). Triplicate cultures were exposed to increasing concentrations of the UF008/SAP immunotoxin ranging from 0 to 1000 pg/µl. Controls were exposed to 1000 pg/µl of the nonsense immunoglobulin conjugate IgG/SAP which consists of a general non-specific anti-rabbit IgG covalently conjugated to saporin. In addition, RGC-5 cells not expressing mOpn4 were exposed to 0 or 1000 ng/µl of UF008/SAP or 1000 ng/µl of IgG/SAP. After 4 days in standard culture conditions, representative images of the triplicate cultures were visualized and captured under Hoffman modulation contrast on an Olympus CK2 inverted microscope fitted with a camera mount.

C57BL/6J male mice (Jackson Laboratories, Bar Harbor, ME) were used in this study. All the experimental procedures were carried out in accordance with Association for Assessment of Laboratory Animal Care policies and approved by the University of Virginia Animal Care and Use Committee or by the Stony Brook University Institutional Animal Care and Use Committee.

Injections (2 µl/eye) were performed in animals anesthetized with isoflurane and eyes topically anesthetized with one drop of 0.5% proparacaine (Akorn Inc, Buffalo Grove, IL). Eyelids were gently retracted with fingers, allowing the eyeball to protrude. A Hamilton syringe with a 30 gauge needle was used to make the injection at the level of the ora serrata into the posterior chamber of the eye and needle was left in place for about 2 minutes after the injections. The lids were then slowly released allowing the eyeball to retract back into the orbit. Animals were systemically administered a mouse ketoprofen analgesic mixture after the injections.

Three month old C57BL/6J male mice were divided into six dose groups (n = 5) The right eye of each animal was injected with 2 µl phosphate buffered saline (PBS) vehicle and the left eye was injected with 25, 50, 100, 200, 400 or 800 ng of UF008/SAP suspended in 2 µl PBS vehicle. All animals were sacrificed by CO₂ asphyxiation 4 weeks post-injection. Eyes from 1 animal per group were sectioned whereas eyes from the remaining 4 animals were used for retinal flat-mounts.

Eyes were excised and hemisected. Posterior eyecups containing the lens were immersed into freshly prepared 4% paraformaldehyde (Electron Microscopy Sciences, Fort Washington, PA) in PBS and kept at 4°C for 24 hours. For flat-mounts, lenses were removed and retinas dissected from eyecups after which they were spread on a filter paper. All immunohistochemical procedures on flatmounts were carried out in 1.5 ml microfuge tubes. For retinal sections, eyecups were cryo-protected in 30% sucrose in 0.1 M PB overnight at 4°C, embedded and frozen in a prepared (3% gelatin, 30% egg albumin in distilled H₂O) or commercially available (Tissue-Tek OCT 4583 Compound, Sakura Finetek, Torrance, CA) mounting medium, and sectioned in a cryostat at 16 µm thickness. Sections were thaw-mounted onto gelatin/chromium-coated glass microscope slides, air dried, and stored frozen (−20°C or −80°C) until further processing.

Flat-mounted retinas and sections were washed 3 times (10 min each) in tris-buffered saline (TBS) (Quality Biological, Gaithesburg, MD) and blocked for 30 minutes in 6% normal goat serum (Vector Laboratories, Burlingame, CA) in TBS. Blocking was followed by 3 washes with TBS followed by incubation at 4°C for 24 hrs in a 1∶2500 dilution of the primary anti-mouse melanopsin antiserum (UF006) in a TBS buffer containing 1% BSA, 0.25% carrageenan lambda, and 0.3% Triton X-100. Tissues were then washed 3 times with TBS and incubated for 1 hr at room temperature in Cy3-conjugated anti-rabbit IgG secondary antibody (Jackson Immunoresearch, West Grove, PA) diluted 1∶500 in incubating buffer, followed by three TBS washes. Flat mounts were removed from the filter paper and transferred to glass slides. Sections and flat-mounts were mounted in DAPI-containing Vectashield (Vector Laboratories, Burlingame, CA), coverslipped and sealed.

Grayscale images were captured on a Zeiss epifluorescence microscope equipped with a SPOT charge-coupled device camera. Photoshop 6.0 (Adobe Systems, San Jose, CA) was used to enhance image files for brightness and contrast. In each quadrant of the flat-mount, three images corresponding to an area of 0.61 mm² were captured sequentially, from the periphery to the center (optic nerve), for a total of 12 images/retina. These three images were termed "peripheral", "middle" and "central". The cell count in each of the 12 frames was converted to cell number/mm². Selected images were pseudocolored with Photoshop to reflect the long wavelength emission fluorescence of the Cy3 fluorophore.

Three month old C57BL/6J male mice were divided into five groups (n = 4). Four groups received bilateral ocular injections of UF008/SAP (400 ng/eye) and were killed 1, 2, 3 or 7.5 weeks post-injection. The fifth group was bilaterally injected with the nonsense immunoglobulin conjugate IgG/SAP as a control and killed at 7.5 weeks. All animals were sacrificed by CO₂ asphyxiation 4 weeks post-injection. Eyes from 1 animal per group were sectioned whereas eyes from the remaining animals were used for retinal flat-mounts. Tissues were processed as described previously for the dose/response studies.