Two models for M4 phototransduction have been proposed. Our previous work suggests that melanopsin in M4 ipRGCs leads to potassium channel closure with a minor contribution of TRPC3/6/7 channel opening in bright light (Sonoda et al., 2018), while a separate study published concurrently proposed that HCN channels are opened by the melanopsin phototransduction pathway with no contribution from TRPC channels (Jiang et al., 2018). To begin to resolve these discrepancies, we first recorded the M4 photocurrent under conditions that matched those used in Jiang et al. as closely as possible (Table 1 and see ‘Materials and methods’). We stimulated M4 cells with brief, full-field, 50 ms flashes of high photopic (6.08 × 10¹⁵ photons · cm^–2 · s^–1) 480 nm light, the highest possible intensity for our LED light source. We identified large somata of putative M4 cells under infrared differential interference contrast (IR-DIC) and filled each recorded cell with Neurobiotin. We then confirmed the identity of each M4 cell post-recording using multiple, established criteria including morphology (verified ON stratification, large somata, highly branched dendritic arbors), physiology (ON-sustained responses to increments in light), and immunolabeling for SMI-32 (Schmidt et al., 2014; Lee and Schmidt, 2018; Sonoda et al., 2018; Sonoda et al., 2020; Figure 1A). We recorded the M4 melanopsin photocurrent in a cocktail of synaptic blockers (see ‘Materials and methods’) at –66 mV following exposure to a brief, 50 ms full-field flash. Control M4 cells consistently exhibited an inward photocurrent that was composed of both a relatively fast, transient component followed by a slower, larger inward current that persisted for more than 30 s after termination of the 50 ms light pulse (Figure 1B and C). This small, transient component was previously noted in a subset of M2 cells but has not been observed in M4 cells in photopic light, potentially due to the previous use of lower intensity photopic stimuli (Sonoda et al., 2018) or the use of bright, epifluorescent illumination to localize M4 cells that were subsequently recorded in high photopic light (Jiang et al., 2018; Ecker et al., 2010).

We next assessed the role of TRPC3/6/7 channels in M4 phototransduction. Our previous work described a relatively minor, but detectable, role for TRPC3/6/7 channels in M4 phototransduction in photopic light (10¹² photons · cm^–2 · s^–1) (Sonoda et al., 2018), while others have found no contribution of these channels at higher photopic light intensities (10¹⁸ photons · cm^–2 · s^–1) (Jiang et al., 2018). To determine the role of TRPC3/6/7 channels in M4 phototransduction, we compared the intrinsic photocurrent of control M4 cells to that of M4 cells in Trpc3^-/-; Trpc6^-/-; Trpc7^-/- (TRPC3/6/7 KO) retinas to 50 ms full-field flashes of the same bright, full-field, 480 nm light (6.08 × 10¹⁵ photons · cm^–2 · s^–1) (Figure 1C and D). We analyzed the photocurrent amplitude at Early, Intermediate, and Late timepoints in the recording, as well as the maximum photocurrent amplitude (Figure 1D and see ‘Materials and methods’). We found a significant decrease in the amplitude of the Early, transient, component in TRPC3/6/7 KO M4 cells, suggesting that this component is largely driven by TRPC3/6/7 channels. This was not due to changes in current density across genotypes because capacitance was not significantly different between control and TRPC3/6/7 M4 cells (Figure 1—figure supplement 1). Input resistance was also similar across genotypes (Figure 1—figure supplement 1). We also observed a slight, but not significant, decrease in all other quantified amplitudes. Collectively, these data are consistent with our previous findings that TRPC3/6/7 channels make a minor contribution to the M4 light response in bright light (Figure 1C and D; Sonoda et al., 2018).

We next examined the reported contribution of HCN channels to M4 phototransduction. HCN channels are a class of nonspecific cation channels opened primarily by membrane hyperpolarization whose activation voltage shifts in the presence of cyclic nucleotides (Biel et al., 2009). One compelling piece of evidence in support of melanopsin phototransduction opening HCN channels was a reduction of the M4 photocurrent following application of 50 μM ZD7288, an HCN antagonist (Jiang et al., 2018). We therefore first tested whether 50 μM ZD7288 blocked both the M4 photocurrent and HCN-mediated tail current using similar recording conditions to Jiang et al., combined with rigorous post hoc subtype identification of M4 cells. Because the incubation period for ZD7288 was not reported by Jiang et al., we first tested whether a period of 5–8 min incubation with ZD7288 was sufficient to abolish the HCN current in M4 ipRGCs. To do this, we recorded HCN tail currents from M4 ipRGCs first in the absence and then in the presence of 50 μM ZD7288 (Figure 2A; Van Hook and Berson, 2010; Chen and Yang, 2007). We found that treatment with 50 μM ZD7288 for 5–8 min was sufficient to eliminate the HCN tail current and achieve full HCN channel blockade in M4 ipRGCs (Figure 2A). After confirming tail current blockade by 5–8 min of 50 μM ZD7288, we then measured the photocurrent of these same M4 cells where HCN channels are fully blocked. If HCN channels are the sole target of melanopsin phototransduction in M4 cells, then this full blockade of HCN channels with 5–8 min of 50 μM ZD7288 should eliminate the melanopsin-dependent M4 photocurrent. However, we observed no reduction in the M4 photocurrent in 50 μM ZD7288 despite elimination of the HCN tail current under these conditions (Figure 2B and C). Thus, though M4 ipRGCs do express HCN channels, HCN channels are not required for melanopsin phototransduction.

As a second test of HCN involvement in M4 phototransduction, we investigated whether light exposure changes the amplitude of the HCN current or tail current in M4 ipRGCs. If melanopsin phototransduction opens HCN channels by shifting their voltage dependence, then in constant light at –66 mV (the starting potential in our HCN protocol) some proportion of HCN channels would already be open because we measure robust photocurrents at –66 mV (e.g. Figures 1C and 2B). It then follows that a hyperpolarizing step from –66 mV to –120 mV would evoke a smaller HCN current and tail current in light because there would be fewer HCN channels left to open compared to darkness when a larger pool of HCN channels are available to open with that same hyperpolarizing step. To test this, we compared the amplitude of the M4 HCN current and tail current in darkness and after 90 s of bright, background light (480 nm light at 6.08 × 10¹⁵ photons · cm^–2 · s^–1). This background light evoked a steady inward current as evidenced by the increased holding current required at –66 mV in light versus dark (Figure 2—figure supplement 1D and see ‘Materials and methods’). We found no change in either the M4 HCN current or tail current amplitudes (Figure 2—figure supplement 1A–C), indicating that light does not modulate the HCN current, further arguing against HCN involvement in M4 phototransduction.

The permeation properties of the M4 transduction channel(s) should be reflected in the photocurrent I–V relationship. Therefore, as a third test of HCN involvement in M4 phototransduction, we compared the reversal potential of the M4 HCN tail current versus that of the M4 photocurrent. We previously showed that the photocurrent I–V relationship of Control and TRPC3/6/7 KO M4 cells has a negative slope, reverses at the equilibrium potential for potassium channels (–90 mV), and drives an increase in input resistance and cellular excitability, indicating that potassium channels are the primary target of M4 phototransduction (Sonoda et al., 2018). However, the I–V relationship of the HCN tail current has not previously been reported for M4 ipRGCs. The reversal of the HCN tail current fit to a linear regression is used to estimate a reversal potential of HCN channels in other cell types and is usually between –25 mV and –50 mV (Van Hook and Berson, 2010). If melanopsin phototransduction does open HCN channels in M4 cells, then we would expect the extrapolated reversal potential of the HCN tail current to match the reported reversal potential of the M4 photocurrent. To test this, we measured the reversal potential of the M4 HCN tail current and compared it to that of the M4 photocurrent reported previously (Sonoda et al., 2018). To do this, we used a voltage-clamp protocol designed to open HCN channels and recorded the HCN tail current across multiple test potentials in the absence and then presence of 5–8 min of 50 μM ZD7288 (Van Hook and Berson, 2010; Chen and Yang, 2007; Figure 3A). We then calculated the HCN tail current amplitude by subtracting the tail current amplitude in ZD7288 from those generated in control solution at each test potential and plotted the ZD7288-sensitive current and used a linear fit to extrapolate the reversal potential (Van Hook and Berson, 2010; Chen and Yang, 2007; Figure 3A and B). Unlike the I–V relationship of the M4 photocurrent reversal at –90 mV, the linear fit of the M4 HCN tail current shows an extrapolated reversal at –26 mV. Though consistent with the properties of HCN channels (Biel et al., 2009), the extrapolated M4 HCN tail current I–V reversal is distinct from the reversal of the M4 photocurrent (M4 photocurrent replotted in Figure 3B from Sonoda et al., 2018, see also Figures 7D and S6 of Sonoda et al., 2018). These distinct differences further suggest that M4 phototransduction does not open HCN channels.

Though Jiang et al., 2018 reported complete blockade of the M4 photocurrent in 50 μM ZD7288, we saw no change in photocurrent amplitude in the same cells where we had confirmed complete blockade of HCN tail currents after 5–8 min incubation of ZD7288 (Figure 2; Jiang et al., 2018). We therefore sought to reconcile our experimental outcome with those previously described. ZD7288 has been reported to have off-target effects on other ion channels (Felix et al., 2003; Do and Bean, 2003; Sánchez-Alonso et al., 2008; Wu et al., 2012). If the reported prior blockade of the M4 photocurrent by 50 μM ZD7288 was due to off-target effects, then longer incubation with 50 μM ZD7288 could result in off-target effects that mimic the previously observed reduction in the M4 photocurrent (ZD7288 incubation time was not reported in the prior study) (Jiang et al., 2018). To test this, we treated M4 ipRGCs with 50 μM ZD7288 for 20 min. Importantly, this longer application period caused no further reduction in the HCN tail currents compared to cells incubated for 5–8 min, indicating that the shorter 5–8 incubation time is sufficient for full HCN blockade (Figure 2—figure supplement 2A and D). Despite identical HCN channel blockade, this longer 20 min incubation with 50 μM ZD7288 now essentially abolished the M4 photocurrent and increased M4 cell input resistance compared to control (Figure 2—figure supplement 2B–E). These results are consistent with potential nonspecific blockade of potassium channels by ZD7288 and provide a methodological explanation for previous conclusions of HCN involvement in M4 phototransduction (Do and Bean, 2003; Jiang et al., 2018). Collectively, our results argue against a role for HCN channels in M4 phototransduction and provide support for our previous conclusions that melanopsin phototransduction results in closure of potassium channels with a minor contribution of TRPC3/6/7 channels.

Previous work has reported that HCN and TRPC6/7 channels each contribute significantly to M2 phototransduction (Jiang et al., 2018). Considering the evidence against HCN involvement in M4 phototransduction, we next revisited the role of HCN channels in M2 phototransduction. We first sought to replicate previous findings that the HCN antagonist ZD7288 partially reduces the M2 photocurrent (Jiang et al., 2018). To do this, we performed whole-cell recordings of M2 ipRGCs identified under brief epifluorescent illumination in retinas of Opn4-GFP mice where only M1, M2, and M3 ipRGCs are labeled with EGFP (Schmidt et al., 2008). All cells were filled with Neurobiotin and identified as M2 ipRGCs post-recording using established criteria including ON stratification (M2 ipRGCs are the only ON type of ipRGC labeled in Opn4-GFP mice), large dendritic arbors, and lack of SMI-32 immunolabeling (Figure 4A; Schmidt and Kofuji, 2009; Lee and Schmidt, 2018; Lucas and Schmidt, 2019; see 'Materials and methods'). Notably, HCN tail currents of control M2 ipRGCs showed smaller, more variable HCN tail current amplitudes compared to those measured in M4 cells, consistent with previous reports (Jiang et al., 2018; Figures 2A and 4B). We then measured HCN tail currents in M2 cells before and after 5–8 min incubation with 50 μM ZD7288 and found that the HCN currents and tail currents were completely abolished under these conditions, identifying an effective concentration and incubation period of ZD7288 for full HCN channel blockade in M2 cells (Figure 4B). We next evaluated the role of HCN channels in M2 phototransduction. If HCN channels are involved in M2 phototransduction, then complete blockade of HCN channels via application of 50 μM ZD7288 for 5–8 min should significantly reduce the M2 melanopsin photocurrent (Jiang et al., 2018). To test this, we recorded the M2 photocurrent evoked by brief, full-field, 50 ms flashes of high photopic (6.08 × 10¹⁵ photons · cm^–2 · s^–1) 480 nm light in the presence or absence of 5–8 min of 50 μM ZD7288 (Figure 4C–E). M2 photocurrents in control cells consistently showed a transient peak in the Early phase of the response and then a smaller, slow component that persisted for more than 30 s following light offset (Figure 4C and D). As with M4 cells, HCN channel blockade with an effective concentration and incubation period of ZD7288 did not reduce the M2 photocurrent (Figure 4D and E), suggesting that HCN channels are not a target of melanopsin phototransduction in M2 cells. As a second test of HCN involvement in M2 phototransduction, we tested whether light altered the HCN current or tail current amplitude of M2 ipRGCs when stepped from –66 mV to –120 mV. Again, like M4 cells, we found no change in the amplitude of the M2 HCN tail current recorded in background light compared to darkness (Figure 4—figure supplement 1A–C), despite a steady, light-evoked inward current at the starting holding potential of –66 mV (Figure 4—figure supplement 1D and E). This indicates that the M2 HCN current is not modulated by light. This further supports the conclusion that HCN channels, though expressed in M2 cells, are not a target of M2 melanopsin phototransduction.

Our findings suggest that HCN channels are not opened by melanopsin phototransduction in M2 ipRGCs. However, previous work did report partial blockade of the melanopsin photocurrent in M2 ipRGCs following incubation with ZD7288 (Jiang et al., 2018). Given our findings that ZD7288 incubation can have off-target effects that decrease the M4 photocurrent (Figure 2—figure supplement 2), we postulated that prolonged exposure to high concentrations of ZD7288 may likewise reduce the M2 photocurrent via off-target effects (Felix et al., 2003; Do and Bean, 2003; Sánchez-Alonso et al., 2008; Wu et al., 2012). To test this, we measured the photocurrent of M2 ipRGCs following 20 min incubation with 50 μM ZD7288. Despite no additional reduction in HCN current, this longer incubation period resulted in an ~50% reduction in the maximum amplitude of the M2 photocurrent, consistent with the previously reported reduction (Figure 4—figure supplement 2A–D; Jiang et al., 2018). These findings suggest that this additional photocurrent blockade observed in this study and in previous work was due to off-target effects on non-HCN channels in M2 cells.

Given the lack of HCN involvement in M2 phototransduction, we next sought to identify the channels involved. TRPC6/7 have been reported to contribute to the M2 photocurrent (Jiang et al., 2018; Perez-Leighton et al., 2011). We therefore sought to determine the contribution of TRPC3/6/7 channels to M2 phototransduction. To do this, we recorded the M2 photocurrent in TRPC3/6/7KO; Opn4-GFP retinas. We found that the maximum amplitude was reduced by ~75% in TRPC3/6/7 KO M2 ipRGCs, and the Early component was essentially completely abolished (Figure 5A and B). This was not due to changes in intrinsic properties because capacitance and resistance were identical in TRPC3/6/7 KO and Control M2 Cells (Figure 5—figure supplement 1). Additionally, we observed no further reduction in the small remaining photocurrent of TRPC3/6/7 KO M2 following 5–8 min incubation with the HCN antagonist ZD7288 (50 μM) (Figure 5—figure supplement 2A–C). Furthermore, both the HCN tail current and inward current were not significantly altered by light exposure in TRPC3/6/7 KO M2 ipRGCs (Figure 5—figure supplement 3A and B). Thus, as with WT M2 phototransduction, we find no evidence of HCN involvement in TRPC3/6/7 KO M2 ipRGCs.

Our data show that TRPC channels are a major target of melanopsin phototransduction in M2 ipRGCs (Figure 5A and B), but the full I–V relationship of the M2 photocurrent has never been reported. We therefore sought to generate I–V curves for the Maximum, Early, Intermediate, and Late components of M2 phototransduction. To do this, we measured the photocurrent of M2 ipRGCs at multiple holding potentials from –106 mV to +34 mV following stimulation with brief, full-field, 50 ms flashes of high photopic (6.08 × 10¹⁵ photons · cm^–2 · s^–1) 480 nm light (Figure 6A and B, Figure 6—figure supplement 1). In control M2 ipRGCs, the inward current was the largest at –106 mV (Figure 6A, Figure 6—figure supplement 1A) and decreased at more depolarized holding potentials (Figure 6A, Figure 6—figure supplement 1A). The light-evoked I–V curves for each component had a positive slope, consistent with an increase in conductance via channels opening (Figure 6B). All photocurrent components in control M2 ipRGCs reversed between +10 mV and +25 mV, suggesting contributions from cation channels that are more permeable to sodium or calcium, such as TRPC channels (Figure 6B). In TRPC3/6/7 KO retinas, I–V relationships reversed between 0 mV and +10 mV, and the Maximum, Intermediate, and Late component amplitudes were reduced at more hyperpolarized, physiologically relevant, potentials while the Early component was reduced at all potentials (Figure 6, Figure 6—figure supplement 1B). Notably, HCN tail currents in M2 ipRGCs reversed at –32 mV, which is distinct from that of the photocurrent reversal at +19 mV for control and ~0 mV for TRPC3/6/7 KO M2 cells (Figure 7A and B), further arguing against HCN involvement in M2 phototransduction.

M1 and M2 ipRGCs heavily rely on TRPC channels for phototransduction, a commonality that would predict similar photocurrent I–V relationships for these subtypes. However, we observed that the I–V relationship of M1 and M2 ipRGCs differed in shape and reversal potential, with M1 cells having a more linear I–V relationship that reverses at 0 mV compared to a more complex I–V relationship that reverses +19 mV for M2 cells (Figure 8A; Warren et al., 2006; Hartwick et al., 2007; Graham et al., 2008; Xue et al., 2011; Perez-Leighton et al., 2011; Sonoda et al., 2018). This observation suggests that while both M1 and M2 cells require TRPC channels for melanopsin phototransduction, another type of channel may be interacting with TRPC channels in M2 ipRGCs to shift the reversal of the I–V relationship. Calcium entry through voltage-gated calcium channels (VGCCs) would shift the reversal potential of the photocurrent I–V to more positive voltages. Interestingly, VGCCs, like TRPC channels, can be modulated by Gq (Bloomquist et al., 1988; Scott et al., 1995; Niemeyer et al., 1996; Bertaso et al., 2003; Panda et al., 2005; Qiu et al., 2005; Hildebrand et al., 2007; Warren et al., 2006; Graham et al., 2008; Xue et al., 2011; Keum et al., 2014; Sonoda et al., 2018; Jiang et al., 2018). Moreover, in other cell types several TRPC subunits have been shown to interact with VGCCs (Soboloff et al., 2005; Onohara et al., 2006; Yan et al., 2009; Perissinotti et al., 2021). Therefore, we hypothesized that VGCCs may be a target of M2 phototransduction. If VGCCs are required for M2 phototransduction, then we would expect VGCC antagonists to reduce the M2 photocurrent. To test this, we recorded the photocurrent of M2 ipRGCs in control solution and in the presence of a cocktail of VGCC antagonists (see 'Materials and methods'; Figure 8B and C). Application of VGCC antagonists resulted in a significant reduction of the overall M2 photocurrent similar to that seen in TRPC3/6/7 KO M2 cells, indicating that VGCCs are required for M2 melanopsin phototransduction (Figure 8B and C). Importantly, VGCC blockade did not reduce the M1 photocurrent (Figure 8D and E). This indicates that VGCCs are not required for melanopsin phototransduction in M1 ipRGCs and that VGCCs may account for some of the observed differences in the M1 versus M2 photocurrent I–V relationships.

T-type VGCCS are one potential candidate target of the M2 phototransduction cascade. T-type VGCC mRNA expression has been reported in M2 ipRGCs (Figure 9—figure supplement 1A; Tran et al., 2019), and we detect T-type VGCC immunolabeling in ON-stratifying ipRGCs in retinal sections (Figure 9—figure supplement 1B). To test this functionally, we performed voltage clamp recordings of M2 cells stepping membrane voltage from –70 mV to more depolarized voltages (see 'Materials and methods' for detailed protocol) and compared current amplitudes in control solution and after application of the selective T-type VGCC antagonist TTA-P2 (10 µM) (Figure 9A, Figure 9—figure supplement 2; Wu et al., 2018; Randall and Tsien, 1997; Timic Stamenic et al., 2019; Zhang et al., 2023; Dreyfus et al., 2010; Choe et al., 2011). We found that TTA-P2 reduced M2 current amplitudes in this protocol, and that TTA-P2-sensitive currents showed I–V relationships consistent with T-type VGCC channels (Wu et al., 2018; Perez-Reyes, 2003; Randall and Tsien, 1997; Monteil et al., 2000; Fox et al., 1987; Dreyfus et al., 2010; Choe et al., 2011). Combined, these data provide evidence that M2 ipRGCs express functional T-type VGCC channels.

We next tested the hypothesis that T-type VGCCs are gated by the melanopsin phototransduction cascade in M2 cells. If T-type VGCCs are required for M2 melanopsin phototransduction, then blockade of T-type VGCCs should reduce or eliminate the M2 photocurrent. To test this, we recorded the M2 photocurrent under blockade of T-type VGCCs with the specific T-type antagonist TTA-P2 (10 µM). We found that T-type blockade nearly eliminates the M2 photocurrent without reducing the HCN tail current (Figure 9B and C, Figure 9—figure supplement 3). The M2 photocurrent was similarly reduced in the presence of a second T-type VGCC antagonist, mibefradil (10 µM) (Figure 10), suggesting that T-type VGCCs are required for melanopsin phototransduction. Importantly, mibefradil had no effect on the M2 photocurrent in TRPC3/6/7 KO M2 cells (Figure 10), suggesting that TRPC 3/6/7 channels and T-type channels act through a common pathway. We next recorded the M2 photocurrent in a cocktail of VGCC antagonists that did not contain T-type antagonists (see 'Materials and methods'). This VGCC cocktail lacking T-type antagonists did not significantly reduce the M2 photocurrent (Figure 10—figure supplement 1), further supporting a role for T-type VGCCs in M2 phototransduction. As with the full VGCC cocktail used above, TTA-P2 alone had no effect on the M1 photocurrent (Figure 9D and E), suggesting that VGCCs are dispensable for M1 phototransduction. Importantly, this complete lack of blockade of the M1 photocurrent by TTA-P2 or the VGCC antagonist cocktail also argues against reduction of the M2 photocurrent resulting from nonspecific effects of these drugs on TRPC channels. Collectively, these findings suggest a key, previously unappreciated, role for T-type VGCCs in M2, but not M1, melanopsin phototransduction.

Requests for reagents and resources should be directed to the lead contact, Tiffany Schmidt (tiffany.schmidt@northwestern.edu).

All procedures were approved by the Animal Care and Use Committee at Northwestern University.

Both male and female mice were used with a mixed B6/129 background. All mice were between 30 and 90 d of age. For M4 cell recordings, we used WT and Trpc3-/- (Hartmann et al., 2008; RRID:MGI:3810154); Trpc6-/- (Dietrich et al., 2005; RRID:MGI:3623137); Trpc7-/- (Perez-Leighton et al., 2011; RRID:MGI:5296035) mice. For M2 and M1 cell recordings, we used Opn4-GFP (Schmidt et al., 2008) and Opn4-GFP; Trpc3-/-; Trpc6-/-; Trpc7-/- mice. For slice immunohistochemistry, we used Opn4^cre/+ (Ecker et al., 2010; RRID:MGI:5285910) mice.

Mice were anesthetized with isoflurane delivered by Neurostar stereotaxic equipment (Robot Stereotaxic) and fixed to a nose cone under a surgical microscope. A 30-gauge needle was used to puncture a hole in the sclera, and each eye was injected with 1 µl of AAV2-hSyn-DIO-hM3D(Gq)-mCherry (6 × 10¹² vg/ml; Addgene, Cat#44361-AAV2; RRID:Addgene_44361) using a custom Hamilton syringe with a 33-gauge needle (Borghuis Instruments). Mice were euthanized and the eyes were removed 4 weeks after injection.

All mice were dark-adapted overnight and euthanized by CO₂ asphyxiation followed by cervical dislocation.

Eyes were enucleated, and retinas were dissected under dim red light in carbogenated (95% O₂-5% CO₂) Ames’ medium (Sigma-Aldrich). Retinas were then sliced in half and incubated in carbogenated Ames’ medium at 26°C for at least 30 min. Retinas were then mounted on a glass-bottom recording chamber and anchored using a platinum ring with nylon mesh (Warner Instruments). The retina was maintained at 30–32°C and perfused with carbogenated Ames’ medium at a 2–4 ml/min flow.

All recordings were made in Ames’ medium with 23 mM sodium bicarbonate. Synaptic transmission was blocked with 100 μM DNQX (Tocris), 20 μM L-AP4 (Tocris), 100 μM picrotoxin (Sigma-Aldrich), and 20 μM strychnine (Sigma-Aldrich) in Ames’ medium. Then, 500 nM tetrodotoxin (TTX) citrate (Tocris) was added to the synaptic blocker solution for voltage-clamp experiments. For whole-cell recordings, the internal solution (Jiang et al., 2018) used contained 120 mM K-gluconate, 5 mM NaCl, 4 mM KCl, 10 mM HEPES, 2 mM EGTA, 4 mM ATP-Mg, 0.3 mM GTP-Na₂ and 7-Phosphocreatine-Tris, with the pH adjusted to 7.3 with KOH. The internal solution was passed through a sterile filter with a 0.22 µm pore size (Sigma-Aldrich). Prior to recording, 0.3% Neurobiotin (Vector Laboratories) and 10 μM Alexa Fluor 594 (Thermo Fisher) were added to internal solution. The HCN antagonist, ZD7288 (Tocris), was dissolved in distilled water and added to the synaptic blockers for a final concentration of 50 μM. 50 μM ZD7288 was bath applied for 5–8 min (an incubation period we identified to fully block the HCN tail current, Figure 2A) with minimum off-target effects, or for 20 min, which led to additional non-specific effects on the photocurrent of M2 and M4 ipRGCs (Figure 2—figure supplement 2 and Figure 4—figure supplement 2). The cocktail of calcium (VGCC) blockers was added to the synaptic blockers and consisted of L-type blockers: 10 μM nifedipine and 5 µM nimodipine; P/Q-type blocker: 400 nM ω-agatoxin IVA; N-type blocker: 3 μM ω-conotoxin GVIA; R-type blocker: 3 nM SNX-482; and T-type blocker: 10 μM mibefradil dihydrochloride. Nifedipine (Tocris) and nimodipine (Tocris) were dissolved in DMSO and diluted in synaptic blockers. Mibefradil dihydrochloride (Tocris), SNX-482 (Tocris), ω-conotoxin GVIA (Tocris), and ω-agatoxin IVA (Tocris) were reconstituted in water and dilute in synaptic blockers to reach the final concentration. VGCC blockers were applied for 5 min to minimize off-target effects. To record calcium currents (Figure 9), a cesium based internal was used to block potassium and HCN channels. The cesium-based internal solution used contained 125 mM Cs-methanesulfonate, 10 mM CsCl, 1 mM MgCl₂, 5 mM EGTA, 10 mM Na-HEPES, 2 mM Na₂-ATP, 0.5 mM Na-GTP, and 10 mM Phosphocreatine. Cells were recorded in the presence of synaptic blockers with 2.5 mM TEA to block additional potassium channels.

The blue LED light (∼480 nm) was used to deliver light stimuli to the retina through a ×60 water-immersion objective. The photon flux was attenuated using neutral density filters (Thor Labs). Before recording, retinas were dark-adapted for at least 5 min. The photocurrent from ipRGCs was recorded following a 50 ms full-field flash of bright light with an intensity of 6.08 × 10¹⁵ photons · cm^–2 · s^–1.

The ganglion cell layer of retina was visualized using IR-DIC optics at 940 nm. M4 ipRGCs, synonymous with ON-sustained alpha RGCs (Schmidt et al., 2014), were identified in IR-DIC as cells with large somata (>20 μm) and characteristic ON-sustained responses to increments in light, as described in Sonoda et al., 2018. We opted to use IR-DIC localization to initially identify the large somata of putative M4 ipRGCs because it best minimizes bleaching compared to epifluorescence. After all cellular recordings, the identity of M4 ipRGCs was confirmed by verifying the dendrites stratified only in the ON-sublamina of the inner plexiform layer (IPL) and immunolabeled with SMI-32, an M4 ipRGC marker (Schmidt et al., 2014; Sonoda et al., 2018). M1 and M2 ipRGCs were targeted in the Opn4-GFP line (M1, M2, and M3 cells labeled, M4, M5, and M6 not labeled), based on their somatic GFP signals visualized under brief epifluorescent illumination. Following recording, the identities of M1 and M2 ipRGCs were confirmed by examining dendritic stratification (M1: OFF; M2: ON) in the IPL, achieved through intracellular dye (Alexa 594) immediately following recording and confirmed post-recording by Neurobiotin fill. M2 cell dendrites stratified only in the ON-sublamina of the IPL while M1 cell dendrites stratified in the OFF-sublamina of the IPL (described in Schmidt et al., 2008 and Schmidt and Kofuji, 2009). Both M1 and M2 were confirmed negative for SMI-32 immunolabeling as described in Lee and Schmidt, 2018. For all experiments, one cellular light response was recorded from each piece of retina to minimize light adaptation with the exception of TRPC3/6/7KO cells in mibefradil (Figure 10), which were stimulated twice approximately 3 min apart. Two stimulations do not affect photocurrent amplitude (data not shown).

Whole-cell recordings were performed using a Multiclamp 700B amplifier (Molecular Devices) and fire-polished borosilicate pipettes (Sutter Instruments, 3–5 MΩ for M4 cells, 5–8 MΩ for M2 and M1 cells). All voltage traces were sampled at 10 kHz, low-pass filtered at 2 kHz, and acquired using a Digidata 1550B and pClamp 10 software (Molecular Devices). All reported voltages are corrected for a –13 mV liquid junction potential calculated using Liquid Junction Potential Calculator in pClamp. We did not compensate for series resistance.

After recording, retina pieces were fixed in 4% paraformaldehyde (Electron Microscopy Sciences) in 1× PBS overnight at 4°C. Retinas were then washed with 1× PBS for 3 × 30 min at room temperature (RT) and then blocked overnight at 4°C in blocking solution (2% goat serum in 0.3% Triton PBS). Retinas were then placed in primary antibody solution containing mouse anti-SMI-32 (1:1000, BioLegend, Cat# 801701, RRID:AB_509997) in blocking solution for 2–4 d at 4°C. Retinas were washed in 1× PBS for 3 × 30 min at RT and transferred to secondary antibody solution containing Alexa 488 goat anti-mouse (1:1000, Thermo, Cat# A-21131, RRID:AB_2535771) and streptavidin conjugated with Alexa 546 (1:1000, Thermo, Cat# S-11225, RRID:AB_2532130) in blocking solution overnight at 4°C. Retinas were then washed in 1× PBS for 3 × 30 min at RT and mounted using Fluoromount aqueous mounting medium (Sigma).

All images were captured using a confocal laser scanning microscope (Leica DM5500 SPE, Leica Microsystems) with a ×20 objective. To include whole dendrites of ipRGCs, tiled image stacks spanning the ganglion cell layer to inner nuclear layer were collected. Images were processed using Fiji (Schindelin et al., 2012).

Mice were intracardially perfused with 4% paraformaldehyde (Electron Microscopy Sciences) in 1× PBS. The eyes were then removed and fixed in 4% paraformaldehyde (Electron Microscopy Sciences) in 1× PBS overnight at 4°C. The eyes were then washed in 1× PBS for 3 × 30 min at RT and then flash frozen in Optical Cutting Temperature Compound (Tissue Tek). The eyes were then sectioned at 20 µm on a cryostat (Leica CM1950, Leica Instruments).

The retinal slices were warmed to RT in 1× PBS for 30 min and then blocked for 2 hr at RT in blocking solution (5% donkey serum in 0.3% Triton-X in 1× PBS). Retinas were then placed in primary antibody solution containing chicken anti-mCherry (1:1000; Abcam, Cat# ab205402, RRID:AB_2722769) and rabbit anti-CaV3.1 (1:250; Alomone Labs, Cat# ACC-021, RRID:AB_2039779), rabbit anti-CaV3.2 (1:500; Alomone Labs, Cat# ACC-025, RRID:AB_2039781), or rabbit anti-CaV3.3 (1:250; Alomone Labs, Cat# ACC-009, RRID:AB_2039783) for 2 d at RT. The retinal slices were washed with 1× PBS for 5 × 15 min at RT and then incubated in secondary antibody solution containing Alexa 488 donkey anti-rabbit (1:1000; Thermo Fisher Scientific, Cat# A-21206, RRID:AB_2535792) and Alexa 594 donkey anti-Chicken (1:1000; Jackson ImmunoResearch Labs, Cat# 703-585-155, RRID:AB_2340377) for 2 hr at RT.

Images were captured using a confocal laser scanning microscope (Leica TCS SP8, Leica Microsystems) with a ×40 or ×63 objective in the Biological Imaging Facility at Northwestern (RRID:SCR_017767). Images were processed using Fiji (Schindelin et al., 2012).

All data were analyzed using custom scripts written in MATLAB (MathWorks; RRID:SCR_001622). The scripts are available on GitHub (https://github.com/schmidtlab-northwestern/Contreras-2023-M2-Phototransductioncopy archived at schmidtlab-northwestern, 2023).

For voltage-clamp experiments measuring the intrinsic melanopsin response to a 50 ms light stimulus (6.08 × 10¹⁵ photons · cm^–2 · s^–1), ipRGCs were voltage clamped at –66 mV, as in Jiang et al., 2018. Only a single cell was recorded per retina piece to ensure that the cell was not light adapted. We measured the maximum amplitude as well as the amplitude at three timepoints selected to represent different phases of the light response (highlighted in Figure 1B) of the photocurrent. The cells were smoothed using a 100 ms sliding average. The baseline current was quantified as the mean current in the first 3 s of the recording protocol. Maximum amplitude was calculated as the maximum current value of the smoothed trace that had the greatest change from baseline during the recording period. The Early, Intermediate, and Late timepoints represent the average change from baseline within the following timeframes after light onset: Early (141.7–440.4 ms), Intermediate (2857.7–6598.2 ms), and Late (9062.3–14062.3 ms). Values were reported as the absolute value of the current (pA), for each component.

The light-evoked I–V relationship for M2 cells was generated by recording the intrinsic melanopsin response for Control and TRPC3/6/7 KO M2 cells in response to a 50 ms light pulse. Cells were voltage clamped at potentials from –106 mV to +34 mV. Cells were allowed to stabilize at a given holding potentials for 1–3 min prior to light onset. Only a single cell was recorded per preparation to ensure dark adaptation state was as uniform as possible across cells/preparations. We plotted the I–V relationships for Maximum, Early, Intermediate, and Late amplitudes. The reversal potential was identified as where the I–V curve intersected the X-axis (i.e. I = 0).

For experiments measuring the HCN-mediated tail current, channel activation was evoked by hyperpolarizing the cell from –66 mV to –120 mV (Van Hook and Berson, 2010; Chen and Yang, 2007). Tail current amplitudes were defined as the maximum change from baseline upon return to –66 mV (after the voltage step to –120 mV) and were plotted as the absolute value of the current. Hyperpolarizing the membrane to –120 mV evoked an instantaneous current followed by the slowly activating inward current. The HCN inward current was defined as the time-dependent component at the end of the 4 s hyperpolarizing step. The HCN inward current amplitude was measured from the beginning of the slowly activating inward current (following the instantaneous current) to the end of the 4 s hyperpolarization step (Chen and Yang, 2007; Van Hook and Berson, 2010).

For experiments assessing blockade of HCN channels, cells were exposed to 50 μM ZD7288 and then subjected again to the same protocol. Blockade of HCN channels in ipRGCs was observed when the tail current amplitude was abolished (reviewed in Biel et al., 2009). Blockade of HCN channels, defined loss of tail currents, occurred between 5 and 8 min (5–8 min) of exposure to 50 μM ZD7288 for each individual cell. For experiments in Figures 2A and 4B, the HCN tail current was first measured (defined as control), and then the same cell was exposed to 50 μM ZD7288 for 5–8 min to block HCN channels. However, for Figures 2B and 4D, the photocurrent measurements were obtained from different cells: control and cells exposed to 50 μM ZD7288 for 5–8 min. The photocurrent was measured in different cells to avoid melanopsin photobleaching. To measure the I–V relationship of HCN channels in M2 and M4 ipRGCs (as described in Chen and Yang, 2007; Van Hook and Berson, 2010), cells were hyperpolarizing –66 mV to –120 mV to activate the majority of HCN channels. Cells were then depolarized to various test potentials from –106 mV to –66 mV. This protocol was repeated in the same cells following 5–8 min application of 50 μM ZD7288 to ensure blockade of HCN channels. The ZD7288 sensitive tail current was derived by subtracting 5–8 min ZD7288 recordings from Control. Tail current amplitudes were measured at each test potential. The tail current amplitudes during deactivation were plotted against each test potential to construct the I–V relationship. A linear fit was used to extrapolate the reversal potential of HCN tail currents.

For experiments assessing light modulation of HCN channels, the tail current was first measured in the dark. The same cell was then exposed to light for 90 s, which allowed the inward photocurrent to stabilize. We then measured the baseline current, HCN current, and HCN tail current as described above in dark versus light. As expected, all cells in background light exhibited a steady-state inward photocurrent at –66 mV.

For experiments assessing the effects due to prolonged application of 50 μM ZD7288, cells were exposed to 20 min of ZD7288 in whole-cell patch-clamp configuration and recordings were performed following this period. Recordings for Control cells were performed 20 min after whole cell was achieved to account for effects due to long recording time. Because of the long recording times and to avoid multiple light stimulations of the same cell, which would affect photocurrent amplitude, recordings with and without ZD7288 were performed in separate cells.

Input resistance was calculated as the slope of a linear fit of the steady-state voltage deflection evoked by a series of hyperpolarizing current injections from –300 pA to 0 pA and depolarizing current injections from 0 pA to 200 pA.

Before recording calcium currents, cells were voltage clamped at –70 mV and depolarized to –40 mV to block any transient channels. Cells were then stepped from –40 mV to voltages from –60 mV to +30 mV in 10 mV intervals (Randall and Tsien, 1997; Dhein et al., 2004; Cai et al., 2019; Zhang et al., 2023).

Graphing and statistical analysis were performed using GraphPad Prism 9 software (RRID:SCR_002798). For unpaired statistical comparisons, we used a non-parametric, two-tailed Mann–Whitney U test with a Bonferroni correction. For paired statistical comparisons, we used a non-parametric, two-tailed Wilcoxon matched-pairs signed-rank test. Significance was concluded when p<0.05.