Type I intrinsically photosensitive retinal ganglion cells of early post-natal development correspond to the M4 subtype

Intrinsically photosensitive retinal ganglion cells (ipRGCs) mediate circadian light entrainment and the pupillary light response in adult mice. In early development these cells mediate different processes, including negative phototaxis and the timing of retinal vascular development. To determine if ipRGC physiologic properties also change with development, we measured ipRGC cell density and light responses in wild-type mouse retinas at post-natal days 8, 15 and 30. Melanopsin-positive cell density decreases by 17 % between post-natal days 8 and 15 and by 25 % between days 8 and 30. This decrease is due specifically to a decrease in cells co-labeled with a SMI-32, a marker for alpha-on ganglion cells (corresponding to adult morphologic type M4 ipRGCs). On multi-electrode array recordings, post-natal day 8 (P8) ipRGC light responses show more robust firing, reduced adaptation and more rapid recovery from short and extended light pulses than do the light responses of P15 and P30 ipRGCs. Three ipRGC subtypes – Types I-III – have been defined in early development based on sensitivity and latency on multielectrode array recordings. We find that Type I cells largely account for the unique physiologic properties of P8 ipRGCs. Type I cells have previously been shown to have relatively short latencies and high sensitivity. We now show that Type I cells show have rapid and robust recovery from long and short bright light exposures compared with Type II and III cells, suggesting differential light adaptation mechanisms between cell types. By P15, Type I ipRGCs are no longer detectable. Loose patch recordings of P8 M4 ipRGCs demonstrate Type I physiology. Type I ipRGCs are found only in early development. In addition to their previously described high sensitivity and rapid kinetics, these cells are uniquely resistant to adaptation and recover quickly and fully to short and prolonged light exposure. Type I ipRGCs correspond to the SMI-32 positive, M4 subtype and largely lose melanopsin expression in development. These cells constitute a unique morphologic and physiologic class of ipRGCs functioning early in postnatal development.


Background
Rods, cones, and melanopsin-containing intrinsically photosensitive retinal ganglion cells (ipRGC) account for all behaviorally relevant visual and non-visual photoreception in the murine retina [1]. Mice are born visually blind, with rods and cones not forming active synapses until approximately post-natal day 10 (P10) and eyes not opening until P13-15 [2][3][4]. In contrast, ipRGCs produce light induced spikes from birth (P0) [5][6][7]. Recent work suggests ipRGCs have unique roles during this early developmental period that are separate from their adult roles in circadian entrainment and the pupillary light reflex (PLR). These include melanopsin-dependent influences on the early post-natal development of retinal vasculature [8] and the segregation of retinogeniculate projections [9]. The latter is thought to be a consequence of ipRGC modulation of intrinsic retinal waves [9,10]. Melanopsin also mediates negative phototactic behavior [11] and its accompanying vocalizations [12] in young pups.
Changes in ipRGC photosensitivity between early post-natal and adult ages have been reported in rd1/rd1 mice [5] and in transgenic mice expressing eGFP under the melanopsin promoter [13]. However, these differences have not been systematically studied in large numbers of cells from wild-type mice. It remains unclear whether changes in ipRGC photosensitivity mirror the changes in ipRGC functional roles from birth to adulthood in wildtype animals. Additionally, while several physiologically distinct ipRGC subtypes have been described in the early post-natal mouse retina (Types I-III), these types have not been associated directly with the morphologic classes of ipRGCs (M1-M5) described in adults [5,14]. Here we study wild-type ipRGC light responses over the course of post-natal development using multi-electrode array recording, and find a general reduction in photosensitivity with increasing age. This reduction in light sensitivity is largely restricted to one electrophysiologic subtype of ipRGC (the Type I cell). We also note a major reduction in the melanopsin expression in one anatomic subtype of ipRGC (the SMI-32+, M4 ipRGC) during post-natal development. Loose patch recordings confirm that these neonatal M4 cells possess Type I physiology. Mice thus possess a specific population of ipRGCs with heightened intrinsic photosensitivity in early development that is largely lost in adulthood.

Changes in ipRGC and melanopsin expression in early development
During the large-scale apoptotic events of early retinal development ipRGCs numbers drop dramatically. Their numbers then stabilize before eye-opening and into adulthood [7,15]. However, Tu et al. [5] showed a further decrease in the number of light active ipRGCs between P8 and adulthood. To study the change in ipRGC numbers during the post-apoptotic period, we measured ipRGC densities in wildtype P8, P15, P30 and P150 animals by melanopsin immunohistochemistry (see Fig. 1a). The density of total melanopsin-positive cells decreased by 17 % between P8 and P15 (from mean 173 mm −2 to 143 mm −2 , p = 0.025 by ANOVA, Additional file 1 for details), 10 % between P15 and P30 (from 143 mm −2 to 129 mm −2 , non-significant), and 25 % between P8 and P30 (173 mm −2 to 129 mm −2 , p = 0.001) ( Table 1, Fig. 1b). These densities are consistent with other recent ipRGC surveys [16,17].
To determine whether this reduction was specific to ipRGC subtypes, counted cells were subtyped by morphology of immunohistochemically stained cells. The subtypes M1, M2 and M3 were characterized by dendritic arborization in the IPL off-layer, on-layer, or both, respectively. Estevez et al. (2012) [20] previously demonstrated that the ipRGC M4 subtype is also an alpha-on ganglion cell. The neurofilament marker SMI-32 is a marker for alpha-on ganglion cells [18]. Therefore, M4 cells were characterized by melanopsin and SMI-32 co-labeling coupled with dendritic projections to the IPL on-sublayer. Previously, M4 cells have been described only in adults [19][20][21]. Here, we extend that nomenclature to include melanopsin+/SMI-32+ cells in the retina from P8 animals. For analysis, cells were grouped into 2 categories: 1) melanopsin+/SMI-32+ cells likely representing M4 cells, and 2) melanopsin+/SMI-32− cells representing M1, M2, and M3 subtypes ( Table 2).
We measured melanopsin expression levels in somata and dendrites of each subtype by calculating relative pixel intensities in confocal images. The average pixel intensities of somatic and dendritic melanopsin staining were measured from each of 6 M1, M2 and M4 cells within a single image. A total of 4 images from different retinas were analyzed at each age for a total of 24 cells per subtype, per age. Within each image all measurements were normalized to the intensity of the brightest cell measured (always an M1 cell) and termed relative intensities. The median relative intensities for M1 somata remained stable across ages regardless of absolute melanopsin levels. M1 dendritic intensities had a non-significant trend toward lower expression (Fig. 1c). Median relative intensities for M2 somata were also stable across development, remaining at~30 % of M1 values. However, median dendritic intensities for M2 cells decreased by 50 % between P8 and older retinas (p <0.05, Kruskal-Wallis (K-W), Additional file 1). Median relative intensities for M4 somata dropped by 30 % between P8 and older retinas (p < 0.005, K-W, Additional file 1), while dendritic values decreased 8 fold over the same time (p < 0.005, K-W, Additional file 1) (Fig. 1c).

Light response dynamics of ipRGCs in early development
The pronounced loss of melanopsin expression in M4 cells and the more moderate loss in M1 and M2 cells could result in altered intrinsic photosensitivity over the course of development. Changes in ipRGC photosensitivity between birth and adulthood have been noted previously in retinas of rd1/rd1 animals [5], and in transgenic animals  Spiking light responses were recorded following a 1-min 480 nm light stimulus at 3.98 × 10 13 photon cm −2 s −1 (IR 13.6) ( Table 3). Median on-latency increased from a mean of 6 s post lights-on at P8 to 9.5 s at P15 and 9 s at P30 (p < 0.05, K-W, Additional file 2). Median off-latency decreased significantly between the same groups, from 67 s at P8 to 17.5 s at P15 and 15.5 s at P30 (p < 0.05, K-W, Additional file 2). Peak firing rate decreased significantly between P8 and P30 only (p = 0.003, K-W, Additional file 2) (Table 3, Fig. 3). As a consequence of these changes, the total number of spikes elicited by a 1-min stimulus decreased significantly with age, from median 966 spikes at P8 to 410 spikes at P30 to the same stimulus (p < 0.05, K-W, Additional file 2). Overall these changes demonstrate an aggregate decrease in photosensitivity, consistent with other studies [5,13].

Recovery from 1-min bright light exposure
Adult ipRGCs are remarkably resistant to bleaching [22] and show slow adaptation, with some cells continuing to fire ex vivo after many hours of continuous light exposure [23]. The adaptation and bleaching kinetics of developing ipRGCs have not been previously studied. We first tested the ability of developing ipRGCs to recover from a 1-min, 480 nm light at IR 13.6. Exposed retinas were allowed to recover for 1 to 10-min in the dark, and then retested with the same light stimulus. Recovery in each test stimulus was calculated as ratio of spikes during the test stimulus compared to the first 1-min light. P8 ipRGCs recovered to 90 % after a 1-min recovery period and recovered completely by 6 min. Cells from P15 and P30 animals recovered to only 60−70 % after 1-min recovery and required more than 6 min to fully recover (Fig. 4b).
The differences in recovery times between P8 and P15 ipRGCs and P8 and P30 ipRGCS were statistically significant up to the 4-min recovery point. At 5-min and 6-min only the difference between P8 and P15 was significant (p < 0.05, Fig. 4b; LMM, Additional file 3 for details). Changes in peak firing rate and off-latency in recovery mirrored those of total spike activity. On-latencies for P8 cells remained near 100 % of pre-exposure values for all test intervals. On-latencies of P15 and P30 cells initially increased to a mean of 150 % pre-exposure values (longer latencies), but returned to 100 % by10-min recovery.
During 1-min recovery experiments, ipRGCs from P15 and P30 animals 'fatigued' following multiple 1-min light exposures, recovering to between only 40 and 60 % of baseline after 3 to 4 recovery trials, regardless of interval length. This was true even with a 15-min dark period between interval tests. P8 cells did not 'fatigue.' Thus, we utilized only data from cells exposed to 2 or fewer recovery cycles. This result does suggest that the recovery mechanisms of P8 ipRGCs differ substantially from those of older animals.

Recovery from 1-h bright light exposure
ipRGCs are noted for their ability to respond to long light exposures [22][23][24]. To determine how cell firing and recovery from long light exposures varies with development, we next exposed retinas of P8, P15, and P30 mice to a 1-h light (480 nm, IR 13.6), and followed recovery for 1-h with a 1-min test light (same as exposure light) every 10-min ( Fig. 5a-c). To test cell health, retinas were treated with KCl at the conclusion of each experiment. During the 1-h exposure, total spiking decreased  Average densities for melanopsin and SMI-32 stained cells over the first 20-min and stopped by 30-min in most cells. A small number of cells continued to fire throughout the 1-h. This pattern was seen at all ages. During recovery, age-specific differences were observed. Recovery was measured as a ratio of the total spikes during the 1 min post-exposure pulse compared with the first minute of the 1-h light exposure for each cell. Total spikes in P8 retinas recovered quickly, reaching 100 % by 20-min of dark recovery after the 1-h exposure. In comparison, P15 and P30 cells recovered much more slowly, reaching medians of 30 % and 20 %, respectively, over 1 h (Fig. 5a-c and Additional file 1). Between P8 and older ages there were also significant decreases in recovery of peak firing rate, on-latency, initial spikes (first 30-s of firing) and steady state spikes (second 30-s of firing) (all p < 0.05, LMM, Additional file 4: Figure; Additional file 5). Interestingly, off-latencies showed the opposite pattern. P15 and P30 cells recovered off-latencies near that of pre-exposure levels, while P8 cells never recovered to preexposure off-latencies (p = 0.039; LMM,; Additional file 4: Figure; Additional file 5).

Bleaching vs. adaptation
The observed age-dependent decreases in recovery after a 1-h light exposure could be caused by changes in either susceptibility to bleaching or to changes in light adaptation mechanisms (or both). To distinguish between these two possibilities, ipRGCs from P30 animals were exposed to either 9-cis-retinal to reverse bleaching, or to a ten-fold higher intensity light to test for adaptation. Using a protocol that allows retinaldehyde to penetrate the retina on an MEA and restore bleached ipRGC function [24], treatment with 9-cis-retinal during the recovery period did not increase recovery levels above vehicle control or untreated cells (Fig. 5d). Increasing light intensity at the 40, 50 and 60-min recovery time points increased percent recovery over control at all three times points (p = 0.01, Mann-Whitney, 60-min time point; Fig. 5f). Therefore the attenuated responses after long light exposure were not due to irreversible 'run down' in the tissue preparation or chromophore depletion. Taken together, these results suggest that changes in adaptation rather than melanopsin bleaching are responsible for age-dependent changes in recovery.

Recovery of ipRGC subtypes from bright light exposure
The age specific changes in adaptation can be further understood by looking at changes in ipRGC subtypes. Based on MEA recordings, Tu et al. (2005) defined three distinct ipRGC subtypesdenoted Type I, II, and IIIbased on photosensitivity and on-latency. Type I cells show relatively high sensitivity but slow on-latency; Type II cells show relatively low sensitivity and slow on-latency, and Type III cells show both relatively high sensitivity and relatively fast on-latency. We sought to determine if these subtypes showed differences in adaptation to long light exposure and how these differences might change in early development. Cells were subtyped as in Tu et al. (2005) with small modifications to reduce total light exposure before experiments (Table 4, Fig. 6a-c). Because of the relative paucity of total recordable Type II (n = 13) and III (n = 12) cells in older animals, P15 P30 subtypes were combined into a single 'post-eye-opening' group. This combination was justified by the lack of statistically significant differences in P15 and P30 responses ( Figure: Additional file 4). Total spikes and peak firing from the first minute of the 1-h light exposure showed a range of healthy responses in each subtype consistent with their relative photosensitivity (Fig. 6d-h). All subtypes had similar responses to KCl treatment at the end of each experiment and were assumed to be of similar health   (Table 5). Importantly, median KCl responses did not predict subtype recovery levels.
In P8 retinas the three ipRGC subtypes responded to and recovered from long, bright light exposure with significantly different dynamics. During the 1-h exposure, the vast majority of cells responded to light for the first 20-min. A subset of Type I cells continued firing throughout the 1-h light exposure either non-stop (4 cells, 9 %) or intermittently (5 cells, or 11 %). In contrast, 1 out of 7 Type III cells continued to fire continuously during the 1-h exposure, while none of the 13 Type II cells did (Fig. 7, Table 6). Following the 1-h exposure, Type I cells recovered more fully than Type III cells, and Type III cells recovered more fully than Type II cells. In Type I cells total spikes recovered to a median of just over 100 % preexposure activity. In Type III cells total spikes recovered to a median of 70 % of pre-exposure activity. In contrast, Type II cells showed minimal recovery after 1-h ( Fig. 7 and Additional file 6; p < 0.05 for differences between subtype LMM, Additional file 7). There were also  Method for categorizing ipRGC electrophysiologic subtypes. Irradiance (IR) is presented as the log 10 (photon cm −2 s −1 ). On-latency is the time from lights on until peak firing rate significant subtype differences in on-latency, peak-firing, and off-latency, initial spikes and steady-state spikes (all p < 0.05, LMM, Additional file 7, Additional file 6: Figure) Post-eye-opening Type II and Type III cells did not differ significantly from each other in any light response parameters measured. Within each subtype, however, differences in response and recovery were observed with age. Both post-eye-opening Type II and III cells showed a greater proportion of cells firing intermittently or continuously than the same class of cells at P8 during the 1-h exposure (Fig. 7, Table 6). This suggests that a subset of both cell types exhibited reduced adaptation to continuous exposure with increasing age. Type II cells showed an increase in recovery (decreased adaptation) with age in on-latency, total spikes, and initial spikes. In contrast, Type III cells showed a significant decrease in recovery (increased adaptation) with age in peak firing, total spikes, and initial spikes (LMM, Additional file 8, all p < 0.05; Additional file 6: Figure).

Patch clamp recordings of M4 ipRGCs show type I physiology exclusively
By P15, Type I cells were not detectable. Since M4/ alpha-on cells showed a decreased melanopsin expression over the same time course, we tested P8 M4 cells with loose patch recordings to determine their electrophysiologic subtype (Types 1-III). In wildtype P8 retina recordings were made from large cells with alpha-on  appearance, which were later confirmed to be M4/ alpha-on cells by SMI-32 and melanopsin co-staining (Fig. 8d). Using the MEA subtyping criteria (Table 4) 10/10 patched cells were identified as Type I cells (Fig. 8a).

Discussion
Recent work has demonstrated that ipRGCs subserve different aspects of light-dependent behavior and physiology in neonatal mice compared with adults [8,9,11,12].
Here we demonstrate that ipRGCs themselves undergo marked physiologic changes between P8 and P30. We find the previously described decrease in total ipRGC number after P8 [5,7] is specifically due to loss of melanopsin expression in SMI-32+, alpha-on ganglion cells, which in the adult correspond to the M4 subclass of ipRGCs [19,21]. As the density of M1 and M2 cells does not appear to substantially decline during the same time period, it is unlikely the reduced M4-type cell density is due to generalized retinal growth. As electrophysiologicallydefined Type I cells also disappear with the same time course, we hypothesized that these M4-type cells correspond to the Type I electrophysiologic class. Loose patch recording of M4/alpha-on cells confirmed they are  While M4 cells appear to be present at all ages, the low melanopsin expression in older M4 cells likely renders them undetectable in MEA recordings in the present study. A 2-log reduction in response threshold would place the adult M4 response beyond the intensity of the Xenon light source used here. Overall, this suggests that whereas the light response of P8 Type1/ M4 cells are driven entirely by melanopsin, in adults their light response is driven primarily by rod and cone input.
While 5 anatomically defined ipRGC subtypes have been described [26,27], we did not look specifically at M5 cells because they lack specific markers. Additionally, there is debate whether or not these cells are a distinct subtype or a variant of M4 cells (both appear to be alpha-on cells [18]).
Previous studies have shown that adult M4 ipRGCs do not express melanopsin levels highly enough to be detected reliably by standard immunochemistry [20,21,28]. However, we find that in young post-eye-opening animals at least a subset of alpha-on cells, as labeled with SMI-32, do stain for melanopsin using standard immunohistochemistry. The difference in results between previous studies and this study most likely stem from the age of animals examined. Previous studies used P42-P120 animals while the current study used primarily P8 to P35 animals. Our finding that only 3 % of P150 M4/ alpha-on cells express melanopsin suggests that as animals mature from post-eye-opening to adulthood (between P35 and P120), melanopsin levels become undetectable by standard techniques in all but a fraction of cells.
The decrease in Type I/M4 ipRGCs alone does not fully account for the decrease in recordable ipRGCs on multi-electrode recording seen with increasing age. While a decrease in total density from 173 mm −2 to 129 mm −2 was observed for melanopsin-positive cells between P8 and P30 (nearly all attributable to loss of melanopsin staining in SMI-32+ cells), the number of recordable cells per retina decreased from~20 cells/retina to 2 cells per retina. The reason(s) for this further decrease are not apparent, but could include possibilities of decreased overall cell viability in vitro with increased age, or of melanopsin-positive cells that are non-spiking.

Age-specific ipRGC light response recovery
As animals age, ipRGCs exhibit lower sensitivity to light. This is seen in the changes to cell on-latency, peak firing rate, off-latency, and total spikes. These changes are consistent with the results of Tu et al. (2005) and Schmidt et al. (2008). Additionally, the ability of ipRGCs to recovery from a 1-min light exposure decreases with age. Cells from older animals require nearly twice the time to fully recover than P8 cells, similar to recovery times for ipRGCs in older animals as reported by Do and Yau [29]. It is unlikely that the age-dependent decreases in light responses reflects melanopsin bleaching in the 1-min light exposures as mounting evidence indicates melanopsin is a bistable pigment [30] and highly resistant to photic bleaching [24].
Changes in recovery following a 1-h light exposure were more dramatic (Table 6). P8 Type I cells recovered to 100 % of initial activity after 20-min recovery while P15 and P30 cells recovered to only 20-40 % by the end of the 1-h recovery period. This raises the possibility that chromophore recycling and availability change with age, perhaps as a consequence of fully functioning rods and cones acting as sinks for available 11-cis-retinal in older animals. The absence of increased sensitivity after 9-cis-retinal supplementation argues against this. Instead, a difference in long-acting adaptation mechanisms is likely to explain these age-dependent changes, as suggested by increased light responses from older ipRGCs upon exposure to 10-fold brighter light. Such a prolonged recovery time is consistent with the work of Wong et al., [31] who showed, in older animals, that ipRGCs can take 3-h to dark adapt after even brief, strong light exposures. Interestingly, changes in firing rate and off-latency following long light exposure appears to be decoupled. In P8 ipRGCs, the steady-state spike number recovered to greater than pre-exposure levels while off-latency never recovered to pre-exposure levels. In contrast, older ipRGC steady-state spiking never fully recovered but off-latency did. This suggests that mechanisms governing adaptation of steady state firing and signal termination are independent and differentially regulated with development.
Previous work by Wong et al. [31] demonstrated ipRGC adaptation responses that are similar to those in visual photoreceptors. Here we extend that work to adaptation of the functional output of ipRGCs (i.e. cell spiking) instead of direct measurements of photocurrents. This is an important measurement, as the ultimate effect of light exposure history must be reflected in signaling to targets like the suprachiasmatic nuclei (controlling circadian rhythms) and the olivary pretectum (controlling pupillary light responses) in adults. Our data demonstrate that functional adaptation in most ipRGCs is profound and long lasting. However, those cells that continued to fire throughout the 1-h light exposure showed relatively low levels of adaptation and high rates of recovery to preexposure values. These cells could correspond to the rat ipRGCs found by Wong exhibiting continuous firing in response to a 10-h light stimulus [23]. The major difference between this study and Wong is the use of a retinal pigment epithelium (RPE)-attached preparation in the latter. It would be of interest to see if a similar RPE attached preparation in mouse could decrease the adaptation of cells seen here, even when direct retinoid supplementation could not. Alternatively, the difference may stem from the difference in animal pre-experiment dark adaptation time. We dark adapted for 1-h prior to experiments, while Wong dark adapted for 15-20 h. If this is the case, then the difference in results of the two studies reflects a profoundly long-lasting adaptation mechanism in ipRGCs. The identity of the cells showing minimal adaptation is unclear. Both P8 Type I and post-eye-opening Type II and III populations had adaptation-resistant cells. Whether these represent one end of a distribution of adaptation, or represent a distinct subpopulation of cells (such as the Brn3b-positive or -negative M1 population [15]) remains to be determined.

Conclusions
We find that there are general decreases in ipRGC density and photosensitivity with age and that these changes are subtype-specific, and primarily attributable to loss of Type I cells. Most importantly, we find that P8 Type I cells are M4/alpha-on cells and that loss of melanopsin in these cells likely accounts for the majority of changes observed. At P8 these cells behave differently from other subtypes, with increased sensitivity and reduced adaptation. This may reflect a simple maturation of Type1/M4 cells with no functional correlate. However, it is possible that the relative lack of adaptation in these cells could subserve developmental functions such as light modulation of spontaneous retinal waves in neonates and survival behaviors in post-natal animals like negative phototaxis in which sustained adaptation might be detrimental to the animal.

Mice
All experiments were performed in accordance with Association for Research in Vision and Ophthalmology guidelines for animal studies and under an approved animal study protocol at the University of Washington Institutional (Animal Care and Use Committee of University of Washington Protocol #4184-01). Mice were C57BL/6 (Jackson Laboratories, Bar Harbor, ME) studied at ages P8-P10 (P8), P15-20 (P15), and P30-P35 (P30). Animals were maintained in a 12:12-h light-dark cycle and provided food ad libitum.

Immunohistochemistry and analysis
Melanopsin immuno-positive cell densities were determined at P8, P15, and P30. Morphologic ipRGC subtypes were identified by a combination of dendritic projection patterns and immunohistochemical markers using confocal microscopy (Table 2) [28,32]. Melanopsin-immunopositive cells with dendritic projection to the off-sublayer of the inner-plexiform layer (IPL) were designated M1 cells, while those with dendritic projections to the on-sublayer of the IPL were designated M2/M4. Since M4 cells are also alphaon cells (Estevez et al. [20]) M4 cells were differentiated from M2 cells by co-staining with melanopsin and the alpha-on marker non-phosphorylated neurofilament h (SMI-32) [18,[33][34][35]. For melanopsin and SMI-32 immunohistochemistry, retinas were fixed for 1-h in 4 % paraformaldehyde and blocked overnight in 5 % donkey serum, 2 mg/ml BSA, 0.3 % Tx-100 in PBS. Retinas were then co-incubated with both primary antibodies for 3 nights at 4°C, using a custom polyclonal rabbit anti-mouse melanopsin antibody targeting the first 15 amino acids of melanopsin N-terminus ([1:5000]) and a mouse monoclonal antibody for SMI-32 ([1:1000], NE1023, Calbiochem, San Diego, CA). The melanopsin antibody used was previously characterized in [5], and does not stain retinas from melanopsin-deficient mice (data not shown). Secondary antibodies were incubated overnight at 4°C. These were donkey anti-rabbit Alexa 568 (A10042, Invitrogen, Grand Island, NY) and donkey anti-mouse Alexa 488 (A21020, Invitrogen) for melanopsin and SMI-32 respectively. Retinas were imaged by confocal microscopy. A region (average area = 1.29 mm 2 ) was sampled from each of the four retinal quadrants, with each region centered 1 mm from the optic disc. Four images per quadrant were combined into a single image. Cell counts were done by hand using the Cell Counter plug-in of Image J (NIH). Only cells with distinct somatic perimeter staining were counted as melanopsin positive. Counts were analyzed per group with a non-parametric Kruskal-Wallis test followed by a Mann-Whitney (2-tailed) test.
To determine the level of melanopsin staining in cell subtypes over time, relative average intensities were calculated for ipRGC somatic perimeters and dendrites of M1, M2 and M4 subtypes from P8, P15 and P30 mice. Average pixel intensity measurements were taken from structures in a single optical plane of confocal stacks with Image J (NIH). Measurements of soma perimeter average intensity were taken at each cell's widest extent before dendritic projections appeared. M1 dendritic average intensities were measured and averaged from 3 dendrites measured from the distance a dendrite remained in the optical plan after its first ramification in the IPL off-sublayer. Dendritic intensities for M2 and M4 cells were measured from their emergence from the cell soma to the length they could be followed reliably, (usually 1 to 2 cell diameter lengths). Intensities were averaged for 2 and 3 dendrites per cell for M2 and M4 cells, respectively. All measurements were backgroundsubtracted from an equivalent area adjacent to the structure measured. Six cells of each subtype were measured from a single image. All measured average intensity values for cell soma and dendrites were normalized to the value of the soma of the brightest cell measured per image (always an M1 cell). One image from each of 4 retinas was examined yielding a total of 24 cells per subtype per age. Intensity data was not normally distributed (Shapiro-Wilk test, p < 0.05) and did not have equal variance within cells types across age (Levene's test, p < 0.05). Each cell type was therefore compared over the three ages with Kruskal-Wallis tests. Significant Kruskal-Wallis tests were followed by Mann-Whitney (2-tailed) tests and Bonferroni corrected for 3 comparisons for specific age differences. Statistical analysis was performed with SPSS (IBM, Armonk NY).
MEAs consisted of a planar array of 60 electrodes (30 μm diameter, 200 μm inter-electrode spacing; Multi Channel Systems, Reutlingen, Germany). Raw electrical signals were amplified, filtered, and digitized through an A/D card (National Instruments, Austin, TX), written to disk and analyzed off-line, as described previously [5]. Retinas were stimulated with a Xenon light source (Sutter Instruments, Novato, CA) fed through a liquid light guide and diffusing filter (Thorlabs Inc., Newton, NJ). Light intensity and wavelength was adjusted with neutral density and narrow band-pass 480 nm interference filters, respectively (Thorlabs, Inc., Newton, NJ). Light intensity was measured with a radiometer (Advanced Photonics International, Fairfield, CT).

Light exposure protocols
For short-term bleaching/adaptation experiments, retinas were exposed to a 1-min, 480 nm light stimulus at an irradiance (IR) of 3.98 × 10 13 photon cm −2 s −1 (log 10 (3.98 × 10 13 photon cm −2 s −1 ) = IR 13.6) followed by dark recovery intervals of 1 to 10-min followed by a second 1-min test with the same light. Recovery during a test light was calculated as the percentage of spikes during the first 1-min exposure. Age-specific response dynamics were assessed from the first light exposure of each retina in the 1-min exposure experiment. For long-term bleaching/adaptation experiments, retinas were exposed to a continuous, 1-h, 480 nm light stimulus at IR 13.6 and allowed to recover in darkness for 1-h. Activity over the 1-h exposure was monitored continuously and analyzed in 1-min blocks taken every 5-min. To assess recovery, beginning 10-min into the recovery period and every 10-min thereafter, a 1-min test stimulus identical to the exposure light was administered.

ipRGC electrophysiologic subtypes
Categorizing ipRGC subtypes was done as previously described [5] with modifications to reduce light exposure prior to recovery experiments. Subtyping measurements were made at the beginning of each experiment. P8 cells were first divided into cells with on-latencies (time from light-on to time of peak firing) of less than 12-s and greater than 12-s in response to a 1-min, 480 nm light at an irradiance of IR 12.0. Cells with on-latencies less than 12-s were designated Type III cells. Cells with on-latencies greater than 12-s were further divided. Those with an IR 12.0 light response greater than 10 % of their IR 13.6 light responses were designated Type I cells. Those cells with an IR 12.0 light response of less than or equal to 10 % of their IR 13.6 response were categorized as Type II cells. In older animals, cells with on-latencies of less than 12-s in response to a 1-min, 480 nm light at IR 13.0 were designated Type III cells, while those with on-latencies greater than 12-s were Type II cells (Table 4).