The Roles of Rods, Cones, and Melanopsin in Photoresponses of M4 Intrinsically Photosensitive Retinal Ganglion Cells (ipRGCs) and Optokinetic Visual Behavior

All procedures were approved by the Institutional Animal Care and Use Committee at the University of Michigan. Four mouse strains were used: (1)mice (“melanopsin-knockout mice”) in which both copies of the melanopsin open reading frame are replaced by the gene encoding Cre recombinase (Ecker et al.,); (2)mice with non-photosensitive rods (Calvert et al.,); (3)mice with non-photosensitive cones (Chang et al.,); and (4) B6129SF2/J wild-type mice made by crossing C57BL/6J with 129S1/SvImJ mice (Jackson Laboratory stock # 101045). Animals were 4–8 months old, included both sexes, and were housed in a 12 h light 12 h dark cycle with all experiments done during the light phase. Opn4 Gnat1 Gnat2 / Cre Cre −/− 3 cpfl ^{2010 2000 2006}

A mouse was dark-adapted overnight before the day of an experiment. Under dim red light, the mouse was euthanized using COfollowed by cervical dislocation, and its eyes were enucleated and put in Ames' medium gassed with 95% Oand 5% CO. Under infrared-based night vision viewers, the retinas were isolated from the retinal pigment epithelium (RPE), the vitreous was removed using forceps, and each retina was cut into quadrants. A quadrant was flattened on the bottom of a superfusion chamber with the ganglion cell side up and superfused by bubbled Ames' medium at ~3 mL min, with temperature maintained at 33°C. The ganglion cell layer was viewed under infrared transillumination, and whole-cell recordings were made from the largest somas using an internal solution containing (in mM) 120 K-gluconate; 5 NaCl; 4 KCl; 10 HEPES; 2 EGTA; 4 Mg-ATP; 0.3 Na-GTP; 7 Tris-phosphocreatine; ~0.1% Lucifer Yellow for visualizing cellular morphologies; and KOH to set pH at 7.3. 2 2 2 −1

The retina was kept in darkness except when presented with test stimuli, which were all full-field light. In the experiment measuring responses to a sum-of-sinusoids stimulus lasting 23.0 s, the temporal pattern of light intensity modulation was generated by summing 21 equal-amplitude sinusoids of different frequencies (in Hz): 0.3, 0.4, 0.5, 0.7, 0.95, 1.3, 1.9, 2.45, 3.1, 4.1, 5.3, 6.3, 7.2, 8.9, 10.7, 13.9, 16.7, 19.6, 23.1, 27.5, and 31.7. This stimulus was identical to the “high contrast” sum-of-sinusoids stimulus described in (Howlett et al.,) with the highest-intensity point being ~3.4 log units above the lowest-intensity point (Figure), except that: (1) our light source was a 470 nm light-emitting diode rather than white light; and (2) the intensity range was 9.1 log−12.5 log photons cms, with a mean intensity of 10.8 log photons cmswhich is in the high mesopic range (Dacey et al.,). Light intensity was varied by a microcontroller board (Arduino; Scarmagno, Italy) using pulse width modulation, and the light was delivered to the retina via the objective lens. In all other experiments, the stimuli were 480 nm light presented from below the transparent bottom of the superfusion chamber. Stimulus presentation orders and interstimulus intervals were standardized across all cells. Each cell was presented with just one of the three sets of stimuli, i.e., light steps, flickers, or the sum-of-sinusoids stimulus. Each cell's identity as an M4 ipRGC was verified first by its tonic depolarizing response lasting throughout a 10 s light step (10.5 log−11.5 log photons cms) and then by post-recording visualization of its morphology (Ecker et al.,; Estevez et al.,). All cells with other light-step responses and/or morphologies were discarded. M4 cells in all genotypes were morphologically similar (Figure), confirming that the lack of rod, cone, or melanopsin activity did not significantly alter their cellular morphologies. To produce the photos in Figure, each cell's dye fill was imaged at 3–8 focal planes using epifluorescence microscopy. After all out-of-focus cellular structures in each image had been masked manually, all focal planes were z-projected using ImageJ software (National Institutes of Health, Bethesda MD). ²⁰¹⁷ 8A ^{2005 2010 2012} 1A 1A −2 −1 −2 −1 −2 −1

All experiments employed the virtual optokinetic system (Prusky et al.,) sold by CerebralMechanics (Lethbride AB, Canada). Briefly, a mouse was placed on a platform at the center of an arena formed by four computer monitors, and the lid of the apparatus was closed prior to testing. In each trial, the monitors presented a grayscale vertical sine wave grating that drifted either clockwise or counterclockwise at 12 deg/s, and the mouse's tracking behavior was assessed by an experimenter blind to the animal's genotype. Tracking was determined by identifying tracking episodes lasting 1–1.5 s. The two parameters we tested were spatial frequency, and Michelson contrast defined as (max luminance – min luminance) / (max luminance + min luminance). Contrast sensitivity was the reciprocal of the Michelson contrast threshold. For each mouse, acuity testing lasted 7–15 min while contrast sensitivity testing lasted 15–30 min. Light intensity at the mouse's cornea was about 12 μW cm, corresponding very approximately to 13.5 log photons cms. When measuring this intensity, wavelengths exceeding 700 nm were blocked using a low-pass filter as they are irrelevant to mouse retinal photoreception. ²⁰⁰⁴ −2 −2 −1

For analyzing graded membrane potentials (), action potentials in the whole-cell recordings were first filtered out using a 20 Hz low-pass filter (for the light-step and flicker responses) or a 40 Hz low-pass filter (for the 21-frequency responses). Six parameters were analyzed: V m

When constructing spike histograms, action potentials in the unfiltered whole-cell recordings were counted and grouped into 1-s bins, with spike counts during the 5-s prestimulus baseline normalized to 0. To quantify the persistence of light-step responses, spike counts relative to the prestimulus baseline were summed 10–50 s after stimulus offset.

Statistical comparisons of the four genotypes were made using one-way ANOVA followed by post hoc Tukey tests.-values smaller than 0.05 indicate significant differences. P

In the first whole-cell recording experiment, we presented each cell with an intensity series of 10-s light steps ranging from 8.5 log to 15.5 log photons cms. As exemplified by the representative responses in Figure, M4 cells give excitatory spiking responses lasting the duration of the light steps, and at sufficiently high stimulus intensities, the excitation persists for many seconds after a transient hyperpolarization at light offset. To assess how melanopsin, rod input, and cone input contribute to these light-step responses, we tested wild-type, melanopsin-knockout, rod-functionless (), and cone-functionless () M4 cells, and population averages of the recordings are shown in Figure. Even the lowest intensity of 8.5 log photons cmsevoked a clear response in the rod-functionless cells, suggesting it is above the cone threshold. To estimate the melanopsin activation threshold, we recorded from wild-type,andM4 cells in the presence of the rod/cone signaling blockers L-(+)-2-amino-4-phosphonobutyric acid (50 μM), 6,7-dinitroquinoxaline-2,3-dione (40 μM), and D-(-)-2-amino-5-phosphonopentanoic acid (25 μM) (Slaughter and Miller,; Hensley et al.,), and found 12.5 log photons cmsto be the lowest intensity needed to stimulate melanopsin (Figure), the slight hyperpolarizing response at 11.5 log photons cmswas previously shown to be likely a photoelectric artifact (Zhao et al.,). −2 −1 −/− 3 cpfl −2 −1 −/− 3 cpfl −2 −1 −2 −1 1B 2 ^{1981 1993} 2 ²⁰¹⁴ Gnat1 Gnat2 Gnat1 Gnat2 rightmost column

Statistical comparisons between the wild type's and the other genotypes' light-step responses are shown in Figure. The wild-type cells' peak response amplitudes (measured near light onset) were nearly “all-or-none,” with the lowest intensity inducing only slightly smaller responses than the higher intensities (Figure). This could be due to Caspikes at the beginning of the responses (Hu et al.,) which were not eliminated by the 20 Hz low-pass filter. Knocking out melanopsin did not significantly affect any of the peak responses (Figure), whereas abolishing rod or cone function significantly reduced the peak response amplitude at several light intensities (Figureand). 3 3A ²⁰¹³ 3A 3A 2+ left middle right

In the second analysis, we calculated final-to-peak amplitude ratios to quantify the sustainability of the light-step responses—the more sustained a response is, the higher its final-to-peak ratio would be. Eliminating melanopsin did not significantly alter the final-to-peak ratio for any of the responses (Figure), while disrupting rod or cone input significantly altered this ratio at one and two intensities, respectively (Figureand). 3B 3B left middle right

In the third analysis, we quantified the poststimulus persistence of the light-step responses by averagingrelative to the baseline 10–50 s after light offset; we excluded the first 10 poststimulus seconds from this analysis to avoid the transient hyperpolarization seen in many responses. The wild-type cells had large persisting depolarizations at the four highest light intensities (Figure). Abolishing rod input dramatically attenuated these poststimulus depolarizations at three of the four intensities (Figure), but neither the melanopsin-knockout cells (Figure) nor the cone-functionless cells (Figure) had significantly reduced poststimulus depolarizations. V middle left right m 1B 3C 3C 3C

Contrary to our finding, a previous study reported a near-abolition of the poststimulus spiking increase in melanopsin-knockout M4 cells (Schmidt et al.,). Thus, we repeated the Figureanalysis by counting spikes instead of measuring. Figureshows averaged spike histograms of the four genotypes' light-step responses at the four highest intensities, with prestimulus spiking normalized to zero, and Figureplots the spike counts measured 10–50 s after light offset. This spike analysis confirmed that significant reduction in the poststimulus persistence was observed only in the M4 cells lacking rod input (Figure), not in those without melanopsin (Figure) or cone input (Figure). ²⁰¹⁴ 3C 4A 4B 4B 4B 4B V middle left right m

The remaining whole-cell recording experiments assessed M4 cells' responses to temporally varying light. In the first experiment we presented 20-s flickers containing square light pulses that alternated with darkness, at 3 frequencies (0.5, 2, and 5 Hz) and 2 intensities (11.5 log and 14.5 log photons cms). Representative recordings are shown in Figure, and population averages in Figure. In the presence of normal Ames' medium, cells of the four genotypes seemed to show varying abilities to track the individual light pulses. Responses to the individual pulses gradually increased over the course of most flickers, probably indicating adaptation (Figurefirst four). By contrast, melanopsin-only responses recorded during rod/cone block did not show any tracking of the pulses (Figure), indicating that synaptic input is necessary for pulse tracking. −2 −1 5 6 6 6 columns rightmost column

Statistical comparisons between the wild type's and the other genotypes' flicker responses are shown in Figure. To quantify a cell's ability to track the light pulses, we measured the amplitude of its response to the final pulse in each flicker. At the higher light intensity, wild-type cells' tracking ability tended to improve as the flicker frequency increased, and the ability to track the 0.5 Hz flicker increased significantly in the rod-functionless cells (Figure). At the lower intensity, wild-type cells tracked all three flicker frequencies more or less equally well, and tracking ability was improved in rod-functionless cells for all three frequencies (Figure). 7 7A 7A top bottom

The various genotypes' averaged flicker responses seemed to show different degrees of sustainability (Figurefirst four). To quantify the sustainedness of the flicker responses, we computed ratios of the peak amplitude in the final-pulse response to the peak amplitude in the first-pulse response. The only significant change in flicker response sustainedness was seen when melanopsin was knocked out, making the cells' responses less sustained for the 0.5 Hz flicker at the higher intensity (Figure). 6 7B columns top left

The previous experiment tested only three frequencies and all pulses were delivered against a dark background, but the visual scene typically contains far more temporal frequencies and much of vision occurs in light-adapted conditions. Thus, in the next experiment we presented a sum-of-sinusoids stimulus with 21 frequencies summated around a 10.8 log photons cmsmean intensity (Figure) to rapidly assay M4 cells' responses to a wider range of temporal frequencies under a somewhat more light-adapted state. The range of this stimulus (9.1 log−12.5 log photons cms) was chosen because it is just high enough to stimulate melanopsin to some degree (Figure) while probably low enough to avoid excessive bleaching of rods and cones. Fast Fourier transform was used to calculate each cell's response amplitude at each of the 21 frequencies found in the stimulus. As shown in Figure, knocking out melanopsin had no significant effect at any of the frequencies, whereas eliminating rod or cone input significantly increased responses at 13 and 6 frequencies, respectively. −2 −1 −2 −1 8A 2 8B rightmost column

As mentioned in the Introduction, ipRGCs mediate not only non-image-forming behavioral responses but also pattern vision, as previously demonstrated using an optokinetic tracking assay (Schmidt et al.,). Thus, we concluded this study by examining how the optokinetic tracking behavior of mice was impacted by the elimination of melanopsin, rod, or cone function. In the first assay, we determined the various genotypes' acuity, i.e., the highest grating spatial frequency evoking a tracking response. Abolishing cone function reduced acuity most substantially, although disrupting rod function also caused a significant reduction. In contrast, knocking out melanopsin had no effect (Figure). ²⁰¹⁴ 9A

In the second assay we determined the animals' contrast sensitivities for four different grating spatial frequencies. Contrast sensitivity was significantly reduced for the second highest spatial frequency in melanopsin-knockout, rod-functionless and cone-functionless mice, by comparable amounts. For the cone-functionless mice, contrast sensitivity was also significantly reduced at the highest spatial frequency (Figure). Incidentally, the contrast sensitivities we measured were generally lower than those reported in Schmidt et al. (), which could be because we detected tracking responses by identifying longer (1–1.5 s) tracking episodes. 9B ²⁰¹⁴

M4 cells' responses to light-step onset were reduced at several intensities in the rod-functionless and cone-functionless cells but not the melanopsin-knockout cells, suggesting that only rod and cone inputs participate in this short-latency response, as expected from the sluggish onset of melanopsin-based photoresponses (Berson et al.,). However, the intensities at which light-on responses were reduced were somewhat lower in the rod-functionless cells than in the cone-functionless cells, consistent with rods being more photosensitive than cones (Fain and Dowling,). ^{2002 1973}

The observation that negative masking responses were shortened in melanopsin-knockout mice lent support to the notion that melanopsin is crucial for tonic ipRGC photoresponses (Mrosovsky and Hattar,). Similarly, knocking out melanopsin rendered light responses in the suprachiasmatic nucleus (which receives ipRGC input) more transient (Mure et al.,), and light-evoked pupil constriction also became less tonic (Zhu et al.,; Keenan et al.,). But there were also reports of impressively long-lasting rod/cone-driven light responses in both the suprachiasmatic nucleus (van Oosterhout et al.,) and ipRGCs (Wong,), so the relative importance of melanopsin vs. rod/cone signaling in photoresponse sustainability remained unresolved. Here we found that only rod and cone inputs contribute significantly to the sustainedness of M4 cells' responses to 10-s light steps at certain intensities. For the flickering stimuli, however, eliminating rod or cone activity does not impact the sustainedness of responses to any of the flickers tested, while knocking out melanopsin makes responses to one of the three higher-intensity flickers less sustained. ^{2003 2007 2007 2016 2012 2012}

Taken together, the present observations suggest that some of the seemingly contradictory results in previous papers regarding response sustainability could be partly due to differences in photostimulation conditions such as intensities, durations, and temporal characteristics. Functional differences among the ipRGC types and post-retinal processing (Keenan et al.,) could also add to the discrepancies between ipRGC recordings and behavioral data. ²⁰¹⁶

We found that the rod input to M4 cells is responsible for the majority of their poststimulus depolarization; in fact, melanopsin does not seem to contribute at all. Early recordings showed that rod/cone-driven ipRGC responses to light steps had rapid offset (Dacey et al.,; Perez-Leon et al.,; Wong et al.,; Schmidt and Kofuji,), leading to the hypothesis that the poststimulus persistence of the ipRGC photoresponse is mostly if not entirely due to melanopsin. Indeed, the post-illumination sustained pupil responses in primates were reported to have melanopsin-like wavelength sensitivity (Gamlin et al.,), forming the basis for the now-widespread use of this persistent response to assess melanopsin function in humans (Kankipati et al.,; Park et al.,). However, the subjects in Gamlin et al. () were partially light-adapted, potentially limiting rods' contribution to the pupil responses. Considering our present data, investigators using the post-illumination pupil response to measure melanopsin function should minimize rod contribution to the response, either by using a rod-saturating background or by limiting the duration of pre-testing dark adaptation. ^{2005 2006 2007 2010 2007 2010 2011 2007}

Contrary to our results, an earlier paper reported that knocking out melanopsin dramatically reduced the poststimulus persistence of M4 cells' light-step responses (Schmidt et al.,). Both that study and ours used retinas that had been isolated from the RPE so that bleached rhodopsin could not be replenished through the retinoid cycle (Wang and Kefalov,), but our light steps were one third in duration and we tested fewer stimuli at high intensities (11.5 log−15.5 log photons cms), so rods were likely less desensitized in our recordings. By contrast, because melanopsin has a much higher threshold than rhodopsin and is comparatively resistant to photobleaching (Sexton et al.,; Emanuel and Do,; Zhao et al.,), it was probably well preserved in both studies. Consequently, Schmidt et al. probably overestimated the importance of melanopsin relative to the rod input. ^{2014 2011 2012 2015 2016} −2 −1

Using melanopsin alone, ipRGCs can track flickering light up to only ~0.2 Hz (Walch et al.,). In the present study, all the frequencies tested in the flicker and sum-of-sinusoids experiments were above 0.2 Hz and thus should not be trackable by melanopsin, and indeed knocking out melanopsin did not significantly affect any of the frequencies examined. ²⁰¹⁵

For both flickering and sum-of-sinusoids stimuli, tracking was enhanced at many frequencies in M4 cells lacking rod input, which is consistent with a reduction in the poststimulus response persistence, thereby allowing the cell to recover from the previous light pulse more rapidly. In the sum-of-sinusoids experiment, silencing rods affected M4 cells' responses at up to 16.7 Hz, which may appear to contradict classic studies reporting the temporal acuity of rod-mediated vision to be at most 15 Hz (Hecht and Shlaer,), but more recent studies showed that rods can track stimuli at 28 Hz or even higher frequencies (Conner and MacLeod,; Nusinowitz et al.,). ^{1936 1977 2007}

For M4 cells lacking cone input, the flicker and sum-of-sinusoids experiments yielded seemingly contradictory results: responses to the flickers were not significantly affected at 0.5, 2, or 5 Hz, but responses to the sum-of-sinusoids increased at all frequencies between 0.5 and 2.45 Hz. This can be explained by cones influencing rods via gap-junction coupling (Raviola and Gilula,). Unlike the flickers which were delivered in darkness, the sum-of-sinusoids stimulus was presented against a background light, and in wild-type retina this background hyperpolarizes rods and cones, thereby compressing their responses to the sinusoids. Rod responses to the sinusoids are presumably further compressed as cones transmit their background-evoked hyperpolarization to rods via gap junctions. Although cones are known to adapt rapidly to prolonged illumination, they maintain a steady-state hyperpolarization about half the size of the peak hyperpolarization (Normann and Perlman,; Burkhardt,), enabling them to continuously suppress the coupled rods. Butcones lack the G-protein that mediates phototransduction (Chang et al.,), so they are constitutively depolarized and do not add to the background-induced compression in the coupled rods, ultimately allowing larger rod-driven photoresponses in downstream ipRGCs. Consistent with this explanation, cone-driven electroretinograms measured in the presence of a background light are substantially larger inmice than in wild-type mice, probably because rods in themice remain depolarized and thus no longer suppress cones through gap junctions (Cameron and Lucas,). The lack of wild-typedifferences outside the 0.5–2.45 Hz range (Figure) could reflect the bandpass properties of rod-cone coupling. ^{1973 1979 1994 2006 2009} 8B Gnat2 Gnat1 Gnat1 vs. Gnat2 right 3 cpfl −/− −/− 3 cpfl

Using an optokinetic tracking assay identical to ours, a previous study detected a role for melanopsin in contrast sensitivity (Schmidt et al.,). Our results showed that at least for some spatial frequencies, melanopsin's role in contrast sensitivity may be as substantial as rods' and cones'. The promotion of contrast sensitivity by melanopsin could theoretically be mediated by not only ipRGCs' projections to image-forming visual centers, but also their intraretinal signaling to amacrine cells which subsequently regulate image-forming retinal circuits (Hankins and Lucas,; Allen et al.,). Two intraretinal ipRGC signaling pathways have been identified: glutamatergic transmission from M1 cells to a subset of dopaminergic amacrine cells (Zhang et al.,; Prigge et al.,), and gap junctional transmission from multiple types of ipRGCs to some non-dopaminergic amacrine cells (Muller et al.,; Reifler et al.,; Chervenak et al.,). Although Schmidt and colleagues ruled out the involvement of M1 cells and hence the glutamatergic pathway in the modulation of contrast sensitivity (Schmidt et al.,), the involvement of the gap junctional pathway remains possible. ^{2014 2002 2014 2008 2016 2010} [2015a] ^{2016 2014}

On the other hand, we found spatial acuity to be determined by rods and cones alone, consistent with the absence of optokinetic tracking in rod/cone-functionless, melanopsin-only mice (Ecker et al.,). For acuity, cones were found to be more important than rods as silencing cones reduced acuity by a greater amount than silencing rods. This is expected since the stimulus intensity at the cornea, >13 log photons cms, was well within the photopic range (Dacey et al.,). Nonetheless, rods did contribute significantly to acuity, which may seem surprising given the photopic conditions. A previous study found that mice possessing functional rods but lacking functional cones and melanopsin continued to exhibit optokinetic reflexes in photopic conditions as long as the pupils were allowed to constrict (Cahill and Nathans,). More recent work has further shown that even though daylight conditions initially saturate rods, these cells gradually escape saturation and regain contrast sensitivity (Tikidji-Hamburyan et al.,). ^{2010 2005 2008 2017} −2 −1

In both the flicker and sum-of-sinusoids experiments, responses at many frequencies were increased rather than decreased in mice lacking rod or cone function, but these mice never showed enhanced acuity or contrast sensitivity in the optokinetic assay—the apparent contrast sensitivity increase at the lowest spatial frequency inmice (Figure) was statistically insignificant. But a direct comparison is impossible because optokinetic tracking is driven by multiple types of ganglion cells (Berson,), and testing conditions (e.g., adaptational states, and stimulus durations, waveforms and intensities) also differed between the single-cell and behavioral experiments. Gnat1 middle −/− 9B ²⁰⁰⁸

At least two more roles of melanopsin have been suggested in the literature. An early study observed reduced pupil constriction at high irradiances in melanopsin-knockout mice (Lucas et al.,). But in our light-step experiment, we did not detect any reduction in the peak response amplitude or final-to-peak amplitude ratio when melanopsin was knocked out in M4 cells. While Schmidt and colleagues did detect reduced light-evoked spiking in melanopsin-knockout M4 cells at one irradiance, responses at higher irradiances were wild-type-like (Schmidt et al.,). This apparent discrepancy between single-cell recordings and a behavioral output could reflect differences between M4 and M1, the main ipRGC type driving pupillary constriction (Ecker et al.,). ^{2003 2014 2010}

Schmidt and colleagues showed that while wild-type M4 cells' spike rates could follow the intensity changes in a stimulus that first became brighter over 6 min and then dimmer over 6 min, melanopsin-knockout M4 cells could only signal the upward ramp and stopped spiking shortly after the stimulus began to dim, suggesting melanopsin is needed to faithfully encode irradiance (Schmidt et al.,). But as discussed above in the context of post-illumination persistence, a caveat is that these recordings were made under conditions where rhodopsin was irreversibly bleached by light. Thus, by the time the downward ramp started, most rhodopsin molecules could have already been bleached by the preceding 6 min of illumination, leaving melanopsin as the only photopigment capable of signaling the intensity decrement. It remains to be tested how well melanopsin-knockout M4 cells can signal intensity reduction under conditions permitting rhodopsin regeneration. ²⁰¹⁴

Our central strategy was to assess whether eliminating each photoreceptive system would significantly alter certain parameters of the ipRGC photoresponse or spatial visual behavior. Because the disruption of photosensitivity was unconditional, it could have induced developmental alterations of retinal circuits (Rao et al.,) and perhaps even mild degeneration in theretinas (Chang et al.,). Despite this potential caveat, our results are largely consistent with known properties of the various photoreceptors, as discussed above. We had tried to use a small-molecule antagonist of melanopsin phototransduction, AA92593, to acutely block melanopsin-based light responses (Jones et al.,), but in our hands this drug failed to significantly reduce these responses. Lucas et al. have studied the roles of melanopsin and cones by using melanopsin- and cone-selective light stimuli (Lall et al.,; Allen et al.,), thereby avoiding any developmental complications in knockout mice, but this approach cannot be used to achieve our present goal, i.e., eliminating melanopsin, cone or rod activity to evaluate the functional necessity of each photoreceptor class. ^{2013 2006 2013 2010 2014} Gnat2 3 cpfl

Another limitation is that all light-step and flicker responses were measured under scotopic conditions. Though a steady background was embedded in the sum-of-sinusoids stimulus, this background was in the mesopic range and so the M4 cells likely remained fairly well dark-adapted. We performed all whole-cell recordings under scotopic/mesopic conditions and limited stimulus durations to 10 s (for the light steps), 20 s (for the flickers), and 23 s (for the sum-of-sinusoids stimulus) to minimize irreversible bleaching of rod and cone photopigments, as the isolated retinas used in this study did not have access to the RPE's retinoid cycle.

Finally, we examined only M4 cells and the findings may not be applicable to all the other ipRGC types. For instance, melanopsin could conceivably contribute more prominently to the light responses of M1–M3 ipRGCs, which express this photopigment at higher levels than M4 cells (Berson et al.,; Estevez et al.,). Moreover, M2 cells have been found to exhibit larger cone-driven light responses than M1 cells (Schmidt and Kofuji,), suggesting that cone signaling may differentially impact the various ipRGC types. ^{2010 2012 2010}

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer ES and handling Editor declared their shared affiliation.