Regulation of molecular clock oscillations and phagocytic activity via muscarinic Ca²⁺ signaling in human retinal pigment epithelial cells

Abstract

Vertebrate eyes are known to contain circadian clocks, however, the intracellular mechanisms regulating the retinal clockwork remain largely unknown. To address this, we generated a cell line (hRPE-YC) from human retinal pigmental epithelium, which stably co-expressed reporters for molecular clock oscillations (Bmal1-luciferase) and intracellular Ca²⁺ concentrations (YC3.6). The hRPE-YC cells demonstrated circadian rhythms in Bmal1 transcription. Also, these cells represented circadian rhythms in Ca²⁺-spiking frequencies, which were canceled by dominant-negative Bmal1 transfections. The muscarinic agonist carbachol, but not photic stimulation, phase-shifted Bmal1 transcriptional rhythms with a type-1 phase response curve. This is consistent with significant M3 muscarinic receptor expression and little photo-sensor (Cry2 and Opn4) expression in these cells. Moreover, forskolin phase-shifted Bmal1 transcriptional rhythm with a type-0 phase response curve, in accordance with long-lasting CREB phosphorylation levels after forskolin exposure. Interestingly, the hRPE-YC cells demonstrated apparent circadian rhythms in phagocytic activities, which were abolished by carbachol or dominant-negative Bmal1 transfection. Because phagocytosis in RPE cells determines photoreceptor disc shedding, molecular clock oscillations and cytosolic Ca²⁺ signaling may be the driving forces for disc-shedding rhythms known in various vertebrates. In conclusion, the present study provides a cellular model to understand molecular and intracellular signaling mechanisms underlying human retinal circadian clocks.

Introduction

Daily behavioral and physiological rhythms are governed by the circadian clock system, which is composed of multiple oscillators in the body. The master circadian clock is located in the hypothalamic suprachiasmatic nucleus (SCN) in mammals^1,2, which organizes rest of oscillators and ultimately coordinates the system’s circadian rhythms³. In addition, as in the lower vertebrate clock⁴, the mammalian eye contains a complete circadian clock⁵. For example, photoreceptor disc shedding^6,7,8, dopamine synthesis⁹, and retinal electrical responses to light¹⁰ are all under the control of the circadian clock. Notably, melatonin release from cultured retina represented temperature-compensated circadian rhythms and could entrain to the light-dark cycles^11,12,13, proving that the mammalian retina contains a self-sustaining and functional circadian clock. Consistently, clock gene expressions have been identified in the inner layer of the mammalian retina^14,15,16, retinal ganglion cells¹⁷, Müller cells¹⁸ and retinal pigmental epithelium (RPE) cells^19,20. Within various peripheral (i.e., non-SCN) circadian clocks, the importance of the clock in the eye should be emphasized as its exceptional role for the photic input (i.e., resetting) system to the central SCN clock.

Microarray assays have demonstrated that nearly 300 genes display circadian transcriptional activities within the eye²¹. Of the many molecular oscillators, the clock gene Bmal1 may play a pivotal role in the retina, because a conditional knockout of Bmal1 in the retina using CHX10-Cre resulted in a loss of circadian rhythm of inner retinal electrical activity in response to light²¹. Conversely, CHX10-Cre might not knockout Bmal1 in RPE cells, because CHX10 is a transcriptional factor localized to the inner nuclear layer, particularly in bipolar cells^22,23. Thus, it is still unknown how clock gene oscillations in RPE cells^19,20 contribute to physiological rhythm generations in the eye. Because disc shedding of photoreceptor outer segments (OS) is mediated largely by phagocytic activities of RPE cells^24,25,26, and OS binding to RPE cells evokes cytosolic Ca²⁺ spikes in RPE cells²⁷, it is reasonable to hypothesize that molecular clock oscillations and intracellular Ca²⁺ signaling in RPE cells are involved in the generation of intrinsic disc-shedding rhythms. However, substantial evidence is lacking to prove this process.

Our group focused on interactions between clock gene transcriptional rhythms and cellular physiological rhythms using long-term Ca²⁺ measurements with yellow cameleon (YC) Ca²⁺ sensor proteins^28,29,30,31. Here, to address molecular and cellular activity rhythms in the RPE, we established a human RPE cell line (hRPE-YC) that stably co-expressed Bmal1-luciferase¹⁹ and the YC3.6 Ca²⁺ sensor³⁰. Using hRPE-YC cells, we visualized interactive rhythms in Bmal1 transcriptions, cytosolic Ca²⁺, and phagocytic activities in these cells. In addition, because we observed consistent cytosolic Ca²⁺ mobilizations via M3 muscarinic acetylcholine receptors in hRPE-YC cells, the effect of a muscarinic agonist (carbamylcholine, carbachol) on phase responsiveness in Bmal1-luciferase rhythms was analyzed in detail.

Results

Functional expression of M3 muscarinic acetylcholine receptors

Live hRPE-YC cells were stimulated with various receptor agonists and the responses were screened by Ca²⁺ imaging. Of these, cholinergic reagents, acetylcholine and carbachol, increased cytosolic Ca²⁺ in nearly all hRPE-YC cells examined (number of cells = 272 in seven separate experiments; Fig. 1A). The acetylcholine and carbachol-induced Ca²⁺ elevations were both concentration-dependent with EC₅₀ values of 1.0–3.3 μM and 9.4–22.9 μM, respectively (Fig. 1B). The magnitude of the Ca²⁺ responses was also analyzed as a function of circadian time (CT), which is defined by the average Bmal1-luciferase rhythms in a culture dish. The magnitude of the Ca²⁺ response was greater at CT20 (the time point with peak chemiluminescence in the Bmal1-luciferase rhythms) than at CT2 or CT14 (F_2,89 = 23.61; P < 0.001 by two-way ANOVA), whereas the EC₅₀ values were not significantly changed across the circadian cycle (Fig. 1B). Conversely, nicotine (100 μM), dopamine (100 μM), serotonin (100 μM), melatonin (100 μM), and high potassium (80 mM) failed to produce apparent Ca²⁺ mobilizations (number of cells = 286 in seven experiments; Fig. 1A,C). The carbachol-induced Ca²⁺ mobilization was completely abolished by the general muscarinic antagonist pirenzepine (10 μM; number of cells = 115 in three experiments; Fig. 1D) or by the M3 selective antagonist darifenacin (10 μM; number of cells = 119 in three experiments; Fig. 1E). The decay time constant in carbachol-induced Ca²⁺ mobilization was not modified by continuous (20 min) exposure to carbachol (number of cells = 124 in three experiments; Fig. 1F), indicating rapid desensitization of receptors following carbachol stimulations.

Figure 1: Cytosolic calcium mobilizations in hRPE-YC cells.

(A) Acetylcholine (ACh, 100 μM) and the muscarinic receptor agonist carbachol (CCh, 100 μM), but not nicotine (Nic, 100 μM), evokes Ca²⁺ transients in hRPE-YC cells. Three representative cell responses are shown. Black bars under the traces denote the timing of perfusion bulb switching, prior to actual drug delivery for 1 min. Virtual color Ca²⁺-concentration images at three timings specified in CCh session (a–c) are shown on the top. (B) The concentration-response curves for ACh and CCh were analyzed at three different CTs. The magnitudes of Ca²⁺ responses were largest at CT20 both for ACh and CCh stimulations. ***P < 0.001 by two-way ANOVA. (C) Apparent Ca²⁺ mobilizations were not produced by 100 μM dopamine (DA), 100 μM serotonin (5-HT), 100 μM melatonin (Mel), and 80 mM high potassium (High K⁺). (D) The CCh-induced Ca²⁺ mobilizations were inhibited by 10 μM pirenzepine (PZP). (E) The CCh-induced Ca²⁺ mobilizations were also inhibited by 10 μM darifenacin (DFC). (F) The CCh (50 μM)-induced Ca²⁺ mobilizations depended on the onset of CCh stimulations, but were not enhanced by continuous stimulations. All above experiments were reproducible for at least three independent trials in separate culture dishes.

Gene expression profiles of G_q-coupled muscarinic receptors (M1, M3, and M5) were also analyzed in hRPE-YC cells using real-time RT-PCR. The highest gene expression was observed for the M3 subtypes (F_2,21 = 232.5; P < 0.001 by Duncan’s multiple range test following one-way ANOVA; Fig. 2A). The expression of M3 receptor genes tended to be greater at CT20, whereas the difference was not statistically significant (F_2,18 = 0.04; n.s. by one-way ANOVA; Fig. 2B).

Figure 2: Transcriptional profiles of muscarinic receptor subtypes.

(A) Within G_q-coupled muscarinic receptor subtypes, M3 is the dominant subtype in hRPE-YC cells. Relative mRNA abundance was evaluated by real-time RT-PCR. ***P < 0.001 by one-way ANOVA. A housekeeping gene (β-actin) was evaluated to estimate relative expression levels. Cells were sampled during subjective nighttime (CT12–24). (B) The gene expression levels of M3 receptors at three different CTs. Although the levels tend to be greater at CT20, this difference was not statistically significant.

Circadian rhythms in spontaneous Ca²⁺ spiking frequencies

Physiological activity rhythms in hRPE-YC cells were estimated by long-term Ca²⁺ imaging techniques. The continuous fluorescent ratio monitoring at a low sampling rate (image pair acquisition per 10 min) demonstrated that there were no circadian rhythms at baseline Ca²⁺ concentrations (number of cells = 53 in three experiments; Fig. 3A, Supplementary Movie 1). Interestingly, a transient decrease in cytosolic Ca²⁺ concentration was observed immediately before and after cell divisions (Fig. 3A; Supplementary Movie 1). In addition, intrinsic Ca²⁺ spiking activities were observed in hRPE-YC cells, whereas the sampling rate in long-term Ca²⁺ imaging could not precisely report the amplitude and frequency of Ca²⁺ spikes. Therefore, we further analyzed the population average of Ca²⁺ spiking frequencies at a higher sampling rate with minimized photo-toxicity using two-photon microscopy (Fig. 3B). The results demonstrated that the Ca²⁺ spiking frequencies in hRPE-YC cells were smaller during early subjective day (CT2) than during other circadian time points (F_3,893 = 12.98; P < 0.001 by Duncan’s multiple range test following one-way ANOVA; Fig. 3C). The circadian variations in the Ca²⁺ spiking frequencies were abolished (Fig. 3C; n.s. by one-way ANOVA) by transfection of dominant-negative Bmal1 (DN-Bmal1), which reduced the amplitude of Bmal1 transcription rhythms by nearly 70% (Supplementary Fig. 1). The spontaneous Ca²⁺ spikes in hRPE-YC cells were caused by the release of Ca²⁺ from internal Ca²⁺ stores, because switching the regular extracellular buffer to Ca²⁺-free buffer did not block Ca²⁺ spikes for 10 min (number of cells = 42 in three experiments; Fig. 3D). Thus, these results indicate the presence of Ca²⁺ spiking rhythms driven by clock gene transcriptions and Ca²⁺ release from internal Ca²⁺ stores.

Figure 3: Long-term Ca²⁺ imaging.

(A) Conventional 3-day imaging of cytosolic Ca²⁺ demonstrated an absence of circadian rhythms in baseline cytosolic Ca²⁺ levels in hRPE-YC cells. Red, green, blue, and black traces correspond to continuous cytosolic Ca²⁺ changes in four cells pointed by red, green, blue, and white arrows in virtual color images on the top (a–c). Interestingly, temporal reduction in cytosolic Ca²⁺ levels immediately before and after cell divisions was observed at approximately 24-h cycles (asterisks). Additionally, intrinsic Ca²⁺ spiking activities were occasionally monitored, whereas it was hard to visualize their circadian rhythmicity by this analysis. (B) Ca²⁺ spiking frequencies were further analyzed for 30 min at CT2 or CT14 using a two-photon confocal microscope at higher sampling rate. Each colored trace represents the Ca²⁺ spiking profile from single cell. (C) Analysis at four different CTs demonstrated the lowest Ca²⁺ spiking frequency at CT2. ***P < 0.001 by one-way ANOVA. Transfection of DN-Bmal1 canceled circadian variations in Ca²⁺ spiking frequency. (D) Temporal replacement of extracellular buffer to Ca²⁺-free buffer failed to inhibit Ca²⁺ spikes, indicating store-driven Ca²⁺ spikes in these cells.

Carbachol and forskolin, but not light-pulse, phase-shift the Bmal1-luciferase rhythm

To analyze molecular clock behaviors, Bmal1-luciferase activities were visualized using a chemiluminescent imager. Following 1 μM dexamethasone treatment, hRPE-YC cells displayed synchronous induction of chemiluminescence that oscillated with circadian period (Fig. 4A). Thus, to analyze the field intensity changes during circadian cycles, we used an eight-channel luminometer for the following analyses. On the second day after monitoring, carbachol was applied to dishes by collection of 10% (v/v) of the culture medium, which was returned to the dishes with carbachol (the final diluted concentration was 50 μM). Compared with the Bmal1-luciferase rhythms without carbachol supplementation, dishes with carbachol displayed phase-shifted rhythms in the subsequent circadian cycle depending on the timing of applications; application at CT14 produced phase delays and application at CT20 produced phase advances (Fig. 4B). By quantifying phase gaps as a function of timing of carbachol applications, a phase response curve (PRC) was fitted (Fig. 4C). The fitting curve obeyed a typical type-1 PRC for carbachol stimulations (Fig. 4C).

Figure 4: Bmal1 transcriptional rhythms.

(A) The Bmal1-luciferase chemiluminescent images of hRPE-YC cells at four different time points after a 1 μM dexamethasone pre-treatment. Note that synchronous and cyclic intensity changes were observed. (B) The Bmal1-luciferase intensity (average in 35-mm dish) was quantified using a multi-channel chemiluminescent analyzer. Arrows indicate onset of 50 μM carbachol (CCh) exposures. Subsequent troughs or peaks of circadian waves were compared with groups with non-treated controls. (C) Based on the CCh-induced phase shifts at various time points, a type-1 PRC was successfully fitted (red dotted line). The sham-treated group (10% v/v medium exchange) failed to induce significant phase shifts (green circles). (D) The identical analysis, but with 5 μM forskolin (Fsk), produced larger phase shifts and formed the type-0 PRC (blue dotted lines).

Using the same experimental paradigm, effects of forskolin (5 μM) were also analyzed. Forskolin shifted Bmal1-luciferase rhythms to the same direction of carbachol stimulation with larger magnitudes. Eventually, the type-0 PRC was fitted to the forskolin-induced phase shifts (Fig. 4D). A 10% v/v culture medium exchange occasionally produced small circadian phase shifts, with no apparent direction. The non-specific phase shifts were not due to temporal dim lighting (<5 lux) during medium exchanges, because bright light exposure (2,000 lux for 5 min) at corresponding timing failed to produce apparent phase shifts (Fig. 5A). Little photosensitivity in hRPE-YC cells may be due to limited expression of photo-sensor molecules, such as Cryptochrome 2 (Cry2) and melanopsin (Opn4), in these cells (Fig. 5B).

Figure 5: Lack of light pulse-induced phase shifts in hRPE-YC cells.

(A) Bmal1-luciferase recordings as in Fig. 4B, but with a 5 min light pulse (2,000 lux, arrows). The light pulse exposures failed to induce phase shifts (red traces) compared with cells without light exposures (blue traces). (B) Gene expression profiles of Cry1, Cry2, and Opn4 were analyzed using real-time RT-PCR in total mouse retina (left) and hRPE-YC cells (right). These gene expression levels did not depend on the time of day of tissue sampling (ZT10, 2 h before dark onset; ZT22, 2 h before light onset) or CTs in hRPE-YC cell cultures. Note that Cry2 levels were significantly smaller and Opn4 levels were negligible in hRPE-YC cells.

To address the mechanisms underlying the differential PRCs induced by carbachol and forskolin, the phosphorylation levels of cAMP response element-binding protein (CREB) were analyzed by immunocytochemistry in hRPE-YC cells (Fig. 6A). Immediate (<10 min) nuclear elevation in phosphorylated CREB (pCREB) was observed after both the carbachol and forskolin applications, with a slightly larger elevation after carbachol stimulation (+26% fluorescent intensity; P < 0.001 by two-tailed Student’s t-test; Fig. 6B). However, the carbachol-induced nuclear pCREB elevation was transient and the level was even lower (−24.5% fluorescent intensity) than that of unstimulated hRPE-YC cells following a 30-min exposure to carbachol (P < 0.001 by two-tailed Student’s t-test; Fig. 6B). Conversely, the forskolin-induced nuclear pCREB elevation was enhanced following 30 min of exposure, and the level was 2.3-fold higher than that after carbachol exposure (P < 0.001 by two-tailed Student’s t-test; Fig. 6B). Thus, these results indicate that the duration of the nuclear pCREB induction, which depends on the stimulant, determines the type of PRCs in hRPE-YC cells.

Figure 6: Differential CREB phosphorylation levels following carbachol (CCh) and forskolin (Fsk) exposure.

(A) Immunofluorescent staining of pCREB (red color in merged picture) following 10 or 30 min exposure to CCh or Fsk. Counter-staining using DAPI (blue color in merged picture) demonstrates the nuclear localization of pCREB signals in hRPE-YC cells. (B) Nuclear pCREB levels were quantified as a function of nuclear fluorescent intensity. Dashed line denotes the average intensity of unstimulated cells. Each average was calculated from 300–500 cells in at least three separate experiments. Significant differences were found between the CCh and Fsk groups. ***P < 0.001 by two-tailed Student’s t-test.

Regulation of phagocytic activities by circadian clock and carbachol

Phagocytic activities of hRPE-YC cells were analyzed by culturing cells with red fluorescent latex beads for 3 h, followed by removal of extracellular fluorescent signals by rinsing and quenching immediately before optical quantification. With confocal microscopy, red spots were visualized at the tip of cellular projections (Fig. 7A). The number of red fluorescent beads that internalized from late subjective night (CT23) to early subjective day (CT2) was significantly larger than that at other CTs (analysis based on seven to eight imaging fields in three dishes; F_3,26 = 5.96, P < 0.01 by Duncan’s multiple range test following one-way ANOVA; Fig. 7A,B). Interestingly, when cells were stimulated with 50 μM carbachol 30 min prior to image acquisition, the large fluorescent signal at CT2 significantly reduced to the level observed at other CTs, resulting in loss of circadian variations (n.s. by one-way ANOVA). In addition, a similar analysis in hRPE-YC cells that underwent transfection of DN-Bmal1 failed to display apparent circadian variations in phagocytic activity levels (n.s. by one-way ANOVA).

Figure 7: Phagocytic activity rhythms in hRPE-YC cells.

(A) Upper. Three-hour phagocytic activities of the fluorescent latex beads (red spots) were compared between early subjective day (CT2) and early subjective night (CT14). Green fluorescence of hRPE-YC cells was used to estimate the field cell density. Because phagocytosis generally occurred at fine cellular projections, red spots were visualized outside of green fluorescence. Middle. Similar analysis as above, but with 50 μM carbachol (CCh) pre-treatment. Lower. Similar analysis of controls, but with cells transfected with DN-Bmal1. (B) The relative intensity of red fluorescence was analyzed as an index of phagocytic activity at four different CTs. The largest phagocytic activity was found at CT2. Circadian variations in phagocytic activities were abolished either by CCh pretreatment or DN-Bmal1 transfection. Means and standard errors were calculated from 4 independent trials for each group. **P < 0.01 by one-way ANOVA.

Discussion

In the earlier work for a human RPE cell line¹⁹, Bmal1 transcriptional rhythms, whose periods were lengthened by high concentrations of lithium, were characterized, whereas the mechanisms by which temporal cues directly regulate transcriptional and cellular activity rhythms were not described. By generating hRPE-YC cells, we were able to visualize interactive rhythms in Bmal1 transcriptions, cytosolic Ca²⁺, and phagocytic activities in these cells. It has been described that phagocytosis of RPE cells determines disc shedding of photoreceptor OS. Therefore, we suggest that molecular clock oscillations and cytosolic Ca²⁺ rhythms in RPE cells may be involved in disc-shedding rhythms. In addition, the present results suggest that the cholinergic system in the eyes is involved in the regulation of the circadian rhythm in RPE cells.

Daily photoreceptor disc shedding underlies circadian rhythms in photic sensitivities, yet the mechanism remains unclear. Phagocytosis of photoreceptor OS in RPE cells is triggered by light. Large phagosomes have been observed at the light-onset time in daily light-dark cycles^6,32,33. In addition, circadian rhythms of disc shedding and phagocytosis persist in constant darkness^6,7,8, suggesting the presence of intrinsic oscillatory mechanisms for the control of disc-shedding rhythms. Photic signal to control phagocytic rhythms could be processed within the eye and not via central circadian clock outputs because a SCN lesion failed to modulate phagocytic rhythms in RPE cells⁷. As discussed in a recent review paper by McMahon⁵, it is still poorly understood whether the disc-shedding rhythms are produced by retinal photoreceptor cells or with interaction to RPE cells. The present results demonstrated (i) the presence of Bmal1-dependent rhythms in Ca²⁺ spiking frequencies and phagocytic activities and (ii) the absence of photic regulation in Bmal1 transcriptional rhythms in hRPE-YC cells. Although DN-Bmal1 may affect not only the clockwork, but also diverse intracellular events including metabolic controls³⁴, and the present data were derived from a single cell line, it is reasonable to hypothesize that (i) intrinsic Ca²⁺-spiking rhythms in RPE cells may be involved in disc-shedding rhythms and (ii) photic regulation of phagocytic activities in RPE cells may be conducted via photoreception outside the RPE cells and subsequent signal transduction from photoreceptors to RPE cells. Although latex bead phagocytosis does not depend on a specific binding process to photoreceptor OS³⁵, the intrinsic rhythms in the RPE cells suggested in this study could be a strong driving force for circadian disc-shedding rhythms.

In addition to the intrinsic (i.e., Bmal1-dependent) phagocytic activity rhythms, the present results demonstrated carbachol-induced inhibition of phagocytic activities in hRPE-YC cells. It has been recently shown that gene knockout of L-type Ca²⁺ channels (Ca_v1.3−/−) partially decreased light-onset phagocytic activities, but increased midday phagocytic activities in in vivo RPE cells³⁶. Notably, carbachol action found in the present study was far beyond the effect of Ca_v1.3−/−, which resulted in the cancellation of intrinsic phagocytic rhythms. The presence of choline acetyltransferase and high-affinity choline transporter in photoreceptor OS has been shown³⁷, and thus such inter-retinal networks may underlie the regulatory pathway. M3 muscarinic receptor couples phospholipase C (PLC) to mobilize intracellular Ca^2+ 38,39. Importantly, it has shown that photoreceptor disc shedding and the phagocytic process in RPE cells are also associated with phosphoinositide signaling⁴⁰: the process begins with shedding of photoreceptor OS discs, which expose phosphotidylserine at the tips of photoreceptors⁴¹ to facilitate binding to αvβ5 integrin^42,43 and CD36⁴⁴ on the apical surface of RPE cells. Subsequently, the Mer tyrosine kinase is used for internalization⁴⁵, and this process may coincide with PLC activation, which promotes inositol triphosphate/diacylglycerol production and intracellular Ca²⁺ mobilization⁴⁰. Indeed, metabotropic Ca²⁺ spiking follows OS binding to RPE cells²⁷. Involvement of Ca²⁺ release from internal stores was also suggested as a knockdown of bestrophin-1, a Ca²⁺/Cl⁻ co-transporter at the endoplasmic reticulum, partially increased phagocytic activities of RPE cells^36,46. Because the present results demonstrated carbachol-induced inhibition of phagocytic activities, large cytosolic Ca²⁺ mobilizations may be involved in the termination of phagocytic processes. This interpretation is consistent with dual roles of Ca²⁺ for the regulation of phagocytic activities in RPE cells, which were hypothesized using Ca_v1.3−/− mice³⁶.

The present results also demonstrated muscarinic regulation of Bmal1 transcriptional rhythms in hRPE-YC cells. Magnitudes of Ca²⁺ mobilizations via acetylcholine and carbachol were not significantly different between subjective day and night. Thus, the refractory period in the PRC during the subjective day is apparently not a matter of the size of Ca²⁺ mobilizations, but may be more downstream events linking to the gene transcriptions. It has been shown that Per2-luciferase rhythms were phase-shifted in type-1 and type-0 PRCs, depending on the duration of light exposure (0.5–12 h) to NIH3T3 mouse fibroblasts overexpressing G_q-coupled photo-sensor melanopsin⁴⁷. Although how light exposure impacts net cytosolic Ca²⁺ levels was not described in that study, it has been suggested that the duration of cytosolic Ca²⁺ mobilizations could be the determinant for types of PRCs. In the present study, carbachol and forskolin were continuously bath-applied to cells after specific time points, whereas carbachol-induced phase-shifting profiles were similar to PRCs via short-term light exposures in the NIH3T3 fibroblast models⁴⁷. This may be due to rapid desensitization of muscarinic receptors as continuous carbachol stimulation, which evoked only the onset Ca²⁺ rise in hRPE-YC cells.

Forskolin, which is lipophilic and a direct activator for adenylyl cyclase, produced further larger phase shifts in the same experimental paradigm and formed type-0 PRC, similar to the results with longer light exposures to melanopsin-expressing fibroblasts⁴⁷. Exposure of melanopsin-expressing fibroblasts to light induces phosphorylation of CREB⁴⁸, which is known to take place in the SCN of hamsters following nocturnal light exposure⁴⁹. Taken together, it is reasonable to assume that intracellular signaling underlying phase responses in hRPE-YC cells may also take place via phosphorylation of CREB, which is induced either by forskolin (i.e., via cAMP-dependent protein kinase) or by carbachol (i.e., via Ca²⁺/calmodulin-dependent protein kinase). Indeed, we observed (i) a transient Ca²⁺ elevation and pCREB induction upon carbachol exposure, and (ii) a long-lasting pCREB induction following forskolin exposure. The difference in the PRC shape upon carbachol and forskolin stimulation can thus be explained by the duration and magnitude of CREB phosphorylation. It has been shown that the human circadian clock is phase shifted by single or repeated light exposure and formed type-1 or type-0 PRCs, depending on the intensity of stimulus or light exposure paradigm^50,51,52. Thus, the present results suggest that such human clockwork could be modeled by intracellular signaling levels using hRPE-YC cells.

Day-night and/or circadian rhythms in photoreceptor OS phagocytosis have been extensively explored in many species. Compared with cone phagocytosis, rod phagocytosis more regularly occurs shortly after light onset, regardless of diurnal and nocturnal species, as in mice⁵³, rats⁶, ground squirrels⁵⁴, and rhesus monkeys⁵⁵. Thus, it is possible that intrinsic phagocytic rhythms in hRPE-YC cells are involved in general rod disc-shedding rhythms. Mutant mice lacking functional rods, but retaining both cone and melanopsin phototransduction pathways, exhibit impaired photoentrainment at a wide range of light levels⁵⁶. Thus, rods play crucial roles for circadian photoentrainment. Maintenance of rod functions by RPE cells, therefore, could be important for photic input pathways to the central circadian clock, as well as clock works within the eye. The generation of a conditional knockout of Bmal1 in RPE cells may provide further insights into the total contribution of RPE clocks for the systems circadian clock, although the study is beyond the scope of current study.

The mechanisms involved in disturbance of the human biological clock and resultant diseases, such as circadian rhythm sleep disorders, remains poorly understood. Thus, the human circadian clock system has been extensively analyzed from clinical aspects and by physiological studies on healthy volunteers, yet studies on the cellular and molecular machineries remain limited. For example, the transcriptional rhythm of the clock gene Per3 in leukocytes may be coupled to the donor’s (system’s) clock movements, because they are phase advanced by exposure of the donor to light at appropriate timing⁵⁷. In addition, recent studies have demonstrated a particular correlation between clock gene transcriptional rhythms in human fibroblasts and sleep-wake profiles in fibroblast donors^58,59,60. However, fundamental information is lacking on human cellular clockwork, because substantial regulatory mechanisms producing phase responses have not yet been demonstrated in cell culture models. In this regard, the present results provide a powerful model for understanding the intracellular signaling mechanisms underlying or regulating human cellular clockwork.

In conclusion, the present results provide a cellular model for understanding the molecular and intracellular signaling mechanisms underlying the human retinal circadian clock and propose a de novo function of the cholinergic system in human eyes.

Methods

Generation of hRPE-YC clones

Human immortalized RPE cells stably expressing Bmal1-luciferase (hRPE cells) were generated from the hTERT-immortalized human retinal pigment epithelial cell line (hTERT RPE-1; purchased from the American Type Culture Collection, Manassas, VA)¹⁹. The hRPE cells were cultured with Dulbecco’s-Modified Eagle Medium/F12 (DMEM/F12) supplemented with 10% FBS (Invitrogen, Carlsbad, CA), sodium bicarbonate (1.2 g/L), and 1% penicillin/streptomycin antibiotics (Invitrogen) under constant temperature (37 °C) and 5% CO₂.

Because hRPE cells¹⁹ are resistant to neomycin, the YC3.6 gene was ligated to the multiple cloning site of a zeocin-resistant vector (pcDNA3.1/zeo; Invitrogen) and transfected into hRPE cells using Lipofectamine-2000 (Invitrogen). Subsequently, the cells were cultured in medium containing zeocin (400–800 μg/mL) for cell selection. Four colonies grown from single cells were picked up by a cloning ring for further subcloning using 96-well plates. Finally, one clone steadily expressing YC3.6 was used in the present study.

Ca²⁺ imaging

For the receptor screening assay, the cells were seeded onto 35-mm glass-bottom dishes and cultured as described above in a CO₂ incubator (cell density: 2–4 × 10⁵ cells/dish). The culture medium was gently rinsed from the dishes using buffered salt solution (BSS) consisting of (in mM) 128 NaCl, 5 KCl, 2.7 CaCl₂, 1.2 MgCl₂, 1 Na₂HPO₄, 10 glucose, and 10 HEPES/NaOH (pH 7.3). hRPE-YC cells were placed on an inverted microscope stage (TE300; Nikon, Tokyo, Japan) and continuously perfused with BSS at a flow rate of 2 mL/min through an in-line heater (SF-28; Warner Instruments, Hamden, CT) set at 36 °C. Fluorescent image pairs (535 ± 15 nm and 480 ± 15 nm) were produced by a 440 ± 5 nm light pulse (300 msec pulses generated by Lambda-LS 300 W Xenon lamp house, Sutter Instrument, Novato, CA), which was conducted to the microscope through a liquid light guide and reflected using a dichroic mirror (FT 445 nm). These images were acquired using a cooled CCD camera (CoolSnap Fx, Photometrics, Tucson, AZ) through a 20× objective lens (Plan Fluor 20×/0.50, Nikon) and a filter wheel (Lambda 10-3, Sutter Instrument) attached in front of the camera. Timings of shutter gating and image acquisitions at 6-s intervals were regulated by digital imaging software (MetaFluor ver. 6.0; Japan Molecular Devices, Tokyo, Japan). The background fluorescence was also subtracted using the software. Dopamine hydrochloride (Sigma-Aldrich, St. Louis, MO), serotonin hydrochloride (Sigma), melatonin (Sigma), L-glutamate monohydrate (Wako Pure Chemical Industries, Ltd., Osaka, Japan), acetylcholine chloride (Sigma), carbamylcholine chloride (Sigma), nicotine hemisulfate salt (Sigma), pirenzepine dihydrochloride (Abcam, Cambridge, UK), and darifenacin hydrobromide (LKT Laboratories, Inc., St. Paul, MN) were delivered to the cells by switching the perfusate. For analysis of CT dependencies in cholinergic Ca²⁺ responses, Bmal1-luciferase rhythms were monitored as below by the luminometer system. Following online estimation of chemiluminescent rhythms, culture dishes were replaced to the Ca²⁺ imaging system and stimulated with acetylcholine or carbachol at different concentrations to obtain concentration response curves.

The following two approaches were used to analyze circadian rhythms in cytosolic Ca²⁺ concentrations. First, hRPE-YC cells plated as above on glass-bottom dishes were treated for 1 h with 1 μM dexamethasone (Sigma) before recordings. Subsequently, cells were rinsed and cultured with fresh culture medium. The culture dish was kept in a temperature- and CO₂-controled (36 ± 0.5 °C and 5 ± 0.2%) custom-built chamber, which enclosed an entire inverted microscope (DM-IRB; Leica Microsystems, Tokyo, Japan). An optical fiber light source (EXFO Photonic Solutions Inc., Ontario, Canada) set outside the recording chamber was used to supply excitation light. Intensity of excitation light was reduced by 70% using a neutral density filter. The fluorescent images were viewed using a 20× objective lens (HC PL-Fluotar 20×/0.50, Leica) and the same optical filter sets for the receptor screening assay. Fluorescent image pairs were acquired continuously for 3 days at 10-min intervals using a cooled CCD camera (Cascade 1k; Photometrics) and digital imaging software (Image Pro-Plus; Media Cybernetics, Bethesda, MD). The fluorescent intensity was enhanced by 4 × 4 pixel binning and electron-multiplier in the CCD.

Although the above method efficiently analyzed slow baseline Ca²⁺ fluctuations, it was not applicable for the analysis of events faster than the sampling rate. To analyze circadian rhythms in spontaneous Ca²⁺ spike frequencies, therefore, the present study additionally used a multiphoton-confocal laser-scanning microscope (A1MP plus, Nikon). For this analysis, a fluorescent image was acquired by two-photon excitation (880 nm) with a 16× water-immersion objective lens (LWD 16×/0.80, Nikon) and digital imaging software (NIS-Elements AR4.10, Nikon). The fluorescent image pairs (535 ± 15 nm and 480 ± 15 nm) were acquired over 30 min at 2-s intervals separately at four different CTs, normalizing the time after dexamethasone treatment (see below in Bmal1-luciferase assay). In this analysis, Ca²⁺ spike frequencies observed in six dishes were compared.

Bmal1-luciferase assay

To monitor the clock gene transcriptional rhythms in hRPE-YC cells, the cells were plated on 35 mm plastic dishes (2 × 10⁵ cells/dish and seven dishes/experiment). On the second day after plating, cells grew at a 90% confluent condition and were treated for 1 h with 1 μM dexamethasone. Subsequently, the culture medium was rinsed with standard culture medium and supplemented with 50 μM beetle luciferin (Promega, Madison, WI). As a blank control, an empty dish filled with the same luciferin-supplemented medium (1 mL) was also prepared. These dishes were light-sealed by aluminum foil and further incubated for 1 h in a CO₂ incubator prior to recording. Recording chamber of Kronos-Dio luminometer system (Model AB-2550, ATTO Co. Ltd., Tokyo, Japan) was set at 37 °C throughout the experimental period. Duration of photon-counting was set for 1 min per sampling with sampling intervals at 10 min. Background intensity was subtracted using the blank control. The time point with peak chemiluminescent level in the Bmal1-luciferase rhythms was regarded as CT20 according to the data based on an ex vivo model⁶¹. To analyze the influence of CTs in the Ca²⁺ imaging and phagocytosis assays, analyses at different CTs were conducted by normalizing the time elapsed after dexamethasone treatment. For example, the data at CT2 were obtained equivalently from two groups at approximately 26 h and 50 h after dexamethasone treatment. Thereafter, the data at CT2 were compared with the data at CT14, which is located between the two groups (approximately 38 h after dexamethasone treatment). The chemiluminescence imager equipped with a deep-cooled electron-multiplier CCD camera (LV200; Olympus, Tokyo, Japan) was used to monitor spatiotemporal distribution of luminescent signals. For this assay, hRPE-YC cells were cultured with 500 μM beetle luciferin and chemiluminescent signals were exposed to the camera for 10 min at 20-min intervals.

To analyze PRCs against pharmacological stimulations, Kronos recordings were paused for 5 min and all dishes were drawn to the on-site clean bench. Ten percent of culture medium (100 μL) was collected from each dish. Carbachol (500 μM) or forskolin (50 μM; Sigma) was added to the collected culture medium and gently returned to the culture dish (final diluted concentration of 50 μM carbachol and 5 μM forskolin). Dishes were kept under dim red light (<5 lux) during initial dish installations and the above medium exchange processes. As controls, dishes underwent a medium exchange without carbachol supplementation or exposing the dishes to bright light (2,000 lux for 5 min). These pharmacological stimulations were examined at the second circadian cycle of Bmal1-luciferase rhythms, because the first cycle was strongly influenced by luciferin uptake. Eventual phase-shifts of Bmal1-luciferase rhythms at the third to forth circadian cycle were quantified by referring unstimulated controls to the same Kronos recording chamber.

To analyze the effects of DN-Bmal1, hRPE-YC cells on 35-mm glass-bottom dishes were transfected with vectors (pcDNA3.1) carrying DN-Bmal1³¹ or blank vectors using Lipofectamine-2000 1day prior to the Kronos recording. General transfection rates were 60–75% using Lipofectamine-2000 in hRPE-YC cells and consistent reductions (~70%) in the amplitude of Bmal1-luciferase rhythms were observed on the second day or later after DN-Bmal1 transfections (Supplementary Fig. 1). Using the residual circadian rhythms in Bmal1-luciferase, CTs were determined as above and used for the further assay for Ca²⁺ imaging or for phagocytosis.

Immunofluorescent confocal imaging

To examine the effects of carbachol and forskolin on the CREB phosphorylation levels, hRPE-YC cells plated on 35-mm glass-bottom dishes were stimulated with carbachol (50 μM) or forskolin (5 μM) for 10 or 30 min during subjective nighttime. Immediately after the stimulations, hRPE-YC cells were fixed in 4% phosphate-buffered paraformaldehyde for 15 min and washed three times with phosphate-buffered saline (PBS) consisting of (mM) 137 NaCl, 2.7 KCl, 1.5 KH₂PO₄ and 8.1 Na₂HPO₄ (pH 7.4). The fixed samples were then incubated for 2 h at room temperature in 10% donkey serum (Jackson Immuno Research Laboratories, West Grove, PA) dissolved in 0.3% Triton-X (Sigma) PBS. Next, samples were incubated with 1:100 affinity-purified rabbit anti-P-CREB (pSer¹³³) (Sigma) dissolved in 5% donkey serum PBS for 24 h at 4 °C. After three 20 min PBS rinses, samples were incubated in 1:400 Cy3-conjugated donkey anti-rabbit IgG (Jackson Immuno Research Laboratories) for 2 h at room temperature. Finally, samples were rinsed with PBS (four 15 min rinses on an orbital shaker) and mounted using Vectorshield (Vector Laboratories, Burlingame, CA) containing 4′,6-diamidino-2-phenylindole (DAPI). Images were acquired using a confocal laser-scanning microscope (FV1000, Olympus) with a laser diode (405 nm), helium neon laser (534 nm) and a 20× objective lens (UPL-SAPO 20×/0.75, Olympus). All staining experiments were repeated at least three times. The nuclear immunofluorescent intensity (8-bit depth) was analyzed using Photoshop CS 6 software (Adobe Systems, San Jose, CA).

Phagocytosis assay

Phagocytic activities were analyzed using a phagocytosis assay kit (Cayman Chemical, Ann Arbor, MI) according to the manufacturer’s instruction. Cells at CT23, CT5, CT11 and CT17 were treated with latex bead-rabbit IgG-Phycoerythrin conjugate (5 μL/1 mL medium/dish) for 3 h. Following rinsing out the beads with standard culture medium, extracellular fluorescence was quenched by 50 μM trypan blue in standard medium. Images were acquired using a FV1000 confocal laser-scanning microscope with argon (488 nm) and helium neon (534 nm) lasers and 60× oil-immersion objective lens (UPL-SAPO 60×/1.35, Olympus). Cytosolic YFP fluorescence from hRPE-YC cells was used to estimate cell shapes and numbers. To analyze the effect of carbachol stimulations, cells were stimulated with 50 μM carbachol 30 min prior to image acquisition according to methods for the Bmal1-luciferase assay.

Real-time RT-PCR assay

The hRPE-YC cells in 35 mm dishes at 90% confluency were rinsed, suspended in PBS, transferred to 1.5 mL RNase-free tubes, and centrifuged for 5 min at 1,000 rpm at room temperature. The cell pellets were transferred to 350 μL of RLT buffer (RNeasy Kit; Qiagen, Chatsworth, CA) and homogenized using a bio-masher (Funakoshi, Tokyo, Japan) at 2,500 rpm for 30 s. The resultant cell lysates were diluted with an equivalent volume of 70% ethanol and stored at −80 °C until RNA extraction. The present study also analyzed transcriptional levels of Cry1, Cry2, and Opn4 in whole mice retina. For this analysis, three 2-month-old male C57BL/6 J mice maintained on a 12 h light/dark cycle at a constant ambient temperature (23 ± 1 °C) were deeply anesthetized with an intraperitoneal injection of sodium pentobarbital (50 mg/kg, body weight). Bilateral eyes were then removed and directly frozen on dry ice. The crystalline lens and vitreous body were carefully removed in ice-cold PBS by stereoscopic surgery, and the remaining bodies, including the sclera, was homogenized as above using a bio-masher. Total RNA (4 μg/sample) was extracted from tissue homogenates using the RNeasy Kit according to the manufacturer’s instructions. The PCR primers used were described in Supplementary information.

Each primer (100 μM) was used in Rotor-Gene SYBR Green RT-PCR Master Mix (Qiagen) according to standard methods. Finally, the PCR amplification was monitored in a strip tube (25 μL reaction volume) set in the 72-well rotor of a real-time PCR system (Rotor Gene 3000 A; Corbett Research, Sydney, NSW, Australia) with the following temperature steps: reverse transcription at 55 °C for 10 min (Hold 1): initial PCR activation at 95 °C for 5 min (Hold 2); and 60 thermal cycles of 95 °C for 5 s and 60 °C for 10 s. The reactions in four separate tubes were averaged for each sample.

Statistical analysis

Data are presented as means ± standard error. Two-way ANOVA or one-way ANOVA followed by Duncan’s multiple range test were used for statistical comparisons across multiple means. A two-tailed Student’s t-test was used for pairwise comparisons. A 95% confidence level was considered to indicate statistical significance. To estimate the concentration-response curve for acetylcholine and carbachol, culture dishes were stimulated twice by low (≤10⁻⁶ M) and high concentration (>10⁻⁶ M) drugs with a 20 min gap. Responses in different dishes were averaged and the concentration-response curve was fitted by a four-parameter Hill function. For the chemiluminescent rhythm assay, background intensity was subtracted and general decaying tendency was de-trended using standard Kronos software (ATTO). The PRCs were eye-fitted by three experienced investigators.