Toward an indoor lighting solution for social jet lag

There is growing interest in developing artificial lighting that stimulates intrinsically photosensitive retinal ganglion cells (ipRGCs) to entrain circadian rhythms to improve mood, sleep, and health. Efforts have focused on stimulating the intrinsic photopigment, melanopsin; however, recently, specialized color vision circuits have been elucidated in the primate retina that transmit blue-yellow cone-opponent signals to ipRGCs. We designed a light that stimulates color-opponent inputs to ipRGCs by temporally alternating short and longer wavelength components that strongly modulate short-wavelength sensitive (S) cones. Two-hour exposure to this S-cone modulating light produced an average circadian phase advance of one hour and twenty minutes in 6 subjects (mean age = 30 years) compared to no phase advance for the subjects after exposure to a 500-lux white light equated for melanopsin effectiveness. These results are promising for developing artificial lighting that is highly effective in controlling circadian rhythms by invisibly modulating cone-opponent circuits.


26
People who spend most of their time under artificial light often suffer a phase delayed circadian 27 rhythm 1-3 . The discrepancy between an individual's delayed biological rhythm and the daily timing 28 determined by social constraints like school and work schedules causes "social jet lag" 4 which is 29 associated with disturbed sleep, daytime fatigue, reduced cognitive function, and a general feeling of 30  Khalsa (2003) 5 that is aligned with earth time so that the beginning of the internal biological night occurs at sunset and the end of the internal biological night occurs before wake time just after sunrise as indicated below the x-axis of the curve. B. (left) Illustration of the color vision circuitry for S-ON and S-OFF types of primate ipRGCs. (right) Illustration of the spectrally opponent response of an S-ON ipRGC with S -(L+M) cone inputs. C. Image of sunset in Seattle Washington illustrating how contrasting short and long wavelength light near the horizon produce a stimulus capable of driving spectrally opponent inputs to ipRGCs making them act as sunrise/sunset detectors. D. Spectral distributions of experimental light stimuli and their predicted effects on the color opponent inputs to ipRGCs. (Top left) Spectrum of the experimental white light with chromaticity coordinates 0.333, 0.333. (Top middle) Spectrum of the LED-derived experimental "blue" light with a spectral peak at 476 nm. (Bottom; left and middle) the product of wavelength-by-wavelength multiplication of the spectral distribution of the white light (Bottom left) times the spectrally opponent response of an ipRGC. Integration of the curve in across wavelength yields the predicted very small relative response of the ipRGC to the white light. (Bottom middle) The product of multiplication of the spectral distribution of the blue light times the spectrally opponent response of an ipRGC. Integration across wavelengths yields the predicted large relative response of the ipRGC to the blue light. (Right) The two spectra which are alternate to produce the S-cone modulating light. unwellness. A potential solution to social jet lag is to develop artificial lighting that is capable of 31 stimulating ipRGCs in the morning during times when such stimuli produce phase advances 5 ( Figure 1A). 32 With regard to circadian rhythms there has been an emphasis the effects of light on the intrinsic 33 photopigment, melanopsin, however ipRGCs can be activated by light absorption by cone 34 photoreceptors whose signals are carried by color opponent circuitry ( Figure 1B) in which short (S) and 35 long (L) plus middle (M) wavelength cones have opposite signs 6-8 . The color opponent input to ipRGCs 36 may have evolved so that changes in the color of sky at dawn and dusk ( Figure 1C) can contribute to 37 synchronization of the internal circadian clock such that the internal biological night begins at sunset 38 and ends before wake time just after sunrise. Previous experiments have provided evidence for a role 39 for color opponency in circadian phototransduction 9 and clear evidence for an S-cone contribution in 40 humans 10,11 . 41 Compared to melanopsin, cone-opponent circuits activate ipRGCs at much lower thresholds 12 . Thus, at 42 common indoor low illumination levels, lights optimized to stimulate the color-opponent circuits could 43 be much more effective in producing circadian phase advances than typical white artificial lighting. Color 44 opponent circuitry in humans is normalized through experience to null to white 13 . Thus, even though 45 artificial white light stimulates S-cones, because the excitatory and inhibitory cone components of the S 46 vs. (L+M) circuitry are balanced by white light it is predicted to have little net effect ( Figure 1D). 47 Narrowband lights that primarily stimulate one side of the opponent circuit are predicted to be much 48 more effective ( Figure 1D). Finally, the circuity carrying cone signals has relatively transient response 49 properties, so under laboratory conditions using narrow band lights that primarily stimulate S-cones, 50 their contributions decay upon extended light exposure 10,11 . Thus, the intensity, spectral and temporal 51 characteristics of the light must all be considered when developing indoor illumination capable of 52 combating social jet lag. 53 We designed a light that stimulates color-opponent inputs to ipRGCs by temporally alternating short and 54 longer wavelength components that strongly modulate short-wavelength sensitive (S) cones. We 55 determined the ability of a morning exposure of this light to produce a phase advance capable of 56 combatting social jet lag compared to a static white light and a static narrow band blue light. Our goal is 57 to evaluate the most effective dynamic lighting approach for circadian photoentrainment at the 58 comparatively low general lighting lux levels typical for homes, offices, schools, and health care facilities. 59 We hypothesize that practical lighting solutions that drive cone-based color-opponent inputs to ipRGCs 60 in the early morning can mediate circadian phase advances that will promote improved mood and 61 cognitive function, combat social jet lag and other circadian problems such as seasonal affective 62 disorder. 63

64
Participants circadian phase relative to solar time 65 When humans are exposed only to natural light, the internal circadian clock synchronizes to solar time 66 such that the internal biological night begins at sunset and ends before wake time just after sunrise 1 67 ( Figure 1A). We used dim light salivary melatonin onset (DLMO) as a measure of circadian phase. Figure  68 2A shows the rise in evening melatonin levels assayed from saliva samples for the six subjects who 69 participated in this study (each subject is represented by a different color). Compared to being 70 synchronized to solar time (shown by the dashed gray 71 curve; Figure 2A)  the excitatory and inhibitory sides of the color-105 opponent response, thus producing little net drive to 106 the ipRGCs from cones. In contrast, almost all 107 wavelengths in the blue light stimulate the S-cones on 108 the excitatory side of the response of the color-109 opponent system. Thus, the white light is expected to 110 produce a null response, and the blue light is predicted 111 Figure 2. Curves showing the nighttime dim rise in salivary melatonin levels under various conditions equated for melanopsin effectiveness. A. Rise in evening melatonin levels for the six subjects who participated in this study (each is shown in a different color). The dashed gray curve shows the predicted rise if the subjects were aligned to earth time where beginning of internal biological night occurs at sunset. On average, subjects were phase delayed 2.8 h. B. Average rise in evening melatonin after two-hour exposure to the static white light (gray curve) of Figure 1A compared to a baseline (dashed curve) measured on day one of the 3-day protocol. There was a slight, nonsignificant, phase delay associated with the white light exposure (n=3 subjects). C. Average rise in evening melatonin (blue curve) after a two-hour exposure to the 476nm blue light of Figure 1B compared to baseline (dashed curve) (n=6 subjects). The 476-nm light produced a phase advance of 40 minutes. D. The Rise in evening melatonin (orange curve) after two-hour exposure to 19 Hz S-cone modulated light compared to baseline (dashed curve) (n=6 subjects). This light produced a phase advance of 1 hour and 20 minutes.
to be many times more effective at driving the color-opponent pathways upstream of the ipRGCs ( Figure  112 1D). 113 To evaluate the ability of lights with different spectral and temporal characteristics to advance circadian 114 phase, we followed a 3-day protocol for each light condition. On the evening of the first day, subjects 115 collected saliva samples every hour starting at 6 PM ending at 2 AM. The following day, the samples 116 were analyzed to measure the rise in melatonin the evening before and the time of DLMO was 117 determined for each subject, defined as the time the melatonin levels reached 20% of maximum 14 . On 118 the morning of the third day of the protocol, each subject viewed a test light for two hours centered 119 10.5 hours after their individual DLMO. This corresponds to the time of circadian cycle expected to 120 produce the maximum light-induced phase advance ( Figure 1A) 5 . On the evening of the same day, 121 subjects again collected saliva samples that were used to evaluate whether the light exposure produced 122 a phase advance. 123 Figure 2B shows the results for the static white light. After exposure to the static white light, the average 124 rise in evening salivary melatonin levels did not differ significantly from the baseline, measured before 125 exposure). The slight phase delay after the exposure is within experimental error (p<0.05; paired t-test). 126 In contrast, the 470 nm blue light that was equated in melanopsin effectiveness to the static white light 127 produced a phase advance of 40 minutes ( Figure 2C). 128 Our goal is to develop lighting that can replace standard indoor white lighting and give people control of 129 their circadian phase. A static blue light (like Figure 1D; top left) is not an acceptable substitute for 130 standard lighting because it must be pure blue to drive the color vision circuitry. Any added long-131 wavelength components that make the light whiter, cancel the effectiveness. As an alternative, we 132 tested a temporally modulated light because, unlike the melanopsin drive to ipRGCs, which is quite 133 sustained, the cone inputs have transient responses. There are two types of color-opponent ipRGCs in 134 primates, S-ON and S-OFF, but both are ON-OFF cells responding both to the onset of one colored light 135 and the offset of the light of the opposing complementary color 6 . 136 Thus, theoretically, the best stimulus is a light that alternates between short and long-wavelength 137 components such that the color-opponent cells are being stimulated by the simultaneous offset of one 138 spectral component and the onset of the opposing component. It is possible to produce lights that, 139 when temporally alternated, appear white but strongly modulate S-cones. The S-cone inputs to ipRGCs 140 are tuned to respond to higher temporal frequencies than those serving hue perception making it 141 possible to modulate the S-cone input to ipRGCs strongly but minimize (and ideally eliminate) the 142 percept of flicker. The S-cone modulating light tested here consisted of a 19 Hz alternating pulse train 143 designed to modulate the quantal catch of S-cones with a differential of 100X between the two phases. 144 This was done by alternating the intensities of LEDs peaking at 427 nm vs. 545 nm, and the addition of 145 light from a 638 nm LED made the S-cone modulated pulse train appear nominally white. The intensity 146 of this light was adjusted to produce a time-averaged quantal catch in melanopsin matched to the 500-147 lux static white light of Figure 1D. As shown in Figure 2D, the S-cone modulated "white light" elicited a 148 striking 1 hr 20 min phase advance. 149 150

151
Blue lights are particularly effective in driving ipRGCs 15,16 , and it is often assumed this is mediated by 152 melanopsin. However, one novel aspect of the experiments here is that the blue and white lights were 153 equated for melanopsin effectiveness, thus, the large effect of blue compared to white cannot be 154 attributed to activation of melanopsin. Since the white condition nulls the color-opponent response 155 ( Figure 1D; left), it effectively isolated the melanopsin drive to the ipRGCs. We conclude that under the 156 relatively low light conditions and two-hour exposure duration used here, melanopsin activation is 157 insufficient to produce any significant circadian phase advance. Moreover, it follows that the substantial 158 phase advance produced by the blue light equated in melanopsin effectiveness to the white light is the 159 result of activation of the color-opponent circuitry, not melanopsin, as commonly assumed. The 160 implication of our result reported here is that since modest illumination level (ca. 500 lux) white lights 161 presented for relatively short duration exposures (≤ 2 hours) are ineffective in stimulating melanopsin 162 sufficiently to produce a phase advance, any practical indoor lighting solution to social jet lag and other 163 problems associated with a delayed circadian clock should focus on stimulating the color opponent 164 inputs to ipRGCs. 165 Previously, one hour of bright white (~10,000 lux) light produced a 40 minute advance in circadian 166 phase 17 . When white lights are sufficiently bright, they can produce a phase advance by activating the 167 much less sensitive melanopsin expressed in human ipRGCs compared to the 500-lux static white light 168 that was ineffective here ( Figure 2D). However, light that strongly modulates the S-cones for two hours 169 (500 lux X 2 hr vs. ~10,000 Lux X 1 hr) amounts to 10X fewer lux-hours but produced a circadian phase 170 advance per exposure hour that was twice as great. Thus, the S-cone modulating light is twice as 171 effective as very bright white light at 1/20 th the intensity. 172 As a different alternative to static illumination, Zeitzer et al. administered 60 2-msec pulses of 473 Lux 173 broad spectral band light over an hour and produced a phase change nearly half that of 1-hour 10,000 174 Lux static white light 18 . We assume that the increased effectiveness is due to the involvement of cone 175 circuits, as in the experiments reported here, since transient white flashes drive spectrally opponent 176 cone inputs to ipRGCs by virtue of differences in the temporal properties of their components. However, 177 because of the spectrally opponent nature of the cone inputs to ipRGCs, modulating S-vs. LM cones is 178 superior to non-spectrally selective cone modulation. The S-cone modulating light is 4 times more 179 effective and the exchange between long and short wavelength components can be invisible whereas 180 bright flashes every minute are not a practical alternative to traditional illumination. 181 Earlier, Spitschan and colleagues 19 measured melatonin suppression using two light stimuli which 182 differed exclusively in the amount of S-cone excitation by almost two orders of magnitude, but not in 183 the excitation L and M cones, rods, and melanopsin. Since the light with stronger S-cone excitation did 184 not differentially suppress melatonin, it might be interpreted to suggest a lack of support for a role for S-185 cone signals in circadian phototransduction. However, the Spitschan et al. experiment relies on the 186 assumption of additivity which doesn't apply to color opponent systems. Static white lights can produce 187 strong S-cone excitation but provide zero drive to ipRGCs because of the opponent nature of the cone 188 inputs. The "S cone isolating light" used by Spitschan was a pinkish color compared to "S-light" which 189 was orangish. This is because to equate the two lights for L and M effectiveness the S+ light had to 190 include about equal amounts of long and short wavelength light, nulling the color opponent response 191 like what occurs with the white light, as illustrated in Figure 1D. Thus show that color opponent circuitry is involved in circadian phototransduction 10,11 . 196 The color of the sky at sunrise and sunset ( Figure 1C) is the ideal cue for synchronizing one's internal 197 body clock to solar time. The intensity of light overhead can vary greatly for many reasons making it an 198 unreliable indicator of the time of day, but the orange color of the sky at the horizon always indicates 199 that it is sunrise or sunset. Retinal ganglion cells act as feature detectors. The color opponent inputs to 200 ipRGCs confer the ability to act as sunrise/sunset detectors. The orange color of the horizon that 201 characterizes the rising and setting sun produces a color contrast with the blue sky ( Figure 1C). The blue 202 and orange parts of the image on the retina produced by the sunset moving across the receptive field of 203 an ipRGC activates the transient color-opponent response very strongly. As shown in Figure 1A, when 204 our internal clock is aligned with solar time, sunrise occurs after the peak of the phase advance portion 205 of the phase response curve and sunset occurs before the peak of the delayed phase portion. When the 206 ipRGCs are strongly stimulated at both dawn and dusk the human phase response curve is perfectly 207 tuned to keep the phase of our internal pacemaker precisely aligned with solar time. 208 Color opponent mechanisms are associated with sensory systems that regulate circadian activity 209 throughout the animal kingdom including fish and reptiles 20,21 . Ancient single-celled organisms exhibit 210 color sensitivity that they use to their circadian activity 22 . It appears that the capacity to sense colors 211 originally evolved to serve circadian rhythms, not for hue perception 23 . The fact that primates have 212 evolved multiple independent circuits that provide color-opponent inputs to ipRGCs is a testament to 213 the importance of these sunrise and sunset detectors to our evolutionary survival. Thus, it makes 214 perfect sense to develop lighting to use these color vision circuits to take control of our circadian 215 wellbeing. 216 Our goal is to take control of our circadian rhythms by adding light exposures that strongly modulate S-217 cone opponency in the morning in the context of the light experience in people's regular daily lives. 218 Thus, here, each subject was exposed to the experimental lights on a background of their regular daily 219 lives as academics at the University of Washington. In this context, exposure to a 500-lux static white 220 produced no significant phase advance but a light with the same melanopsin effectiveness that 221 temporarily modulated S-cone color opponent circuitry produced phase advances, that if administered 222 in the context of a person's normal lighting routine, would be capable of offsetting the average 2.8-hour 223 delay, therefore eliminating social jet lag. 224 The discoveries of color vision circuitry inputs to primate ipRGCs 7,8 together with the evidence which has 225 accumulated showing the role that circuitry in circadian phototransduction, indicate a complete 226 paradigm shift in the strategy to develop healthy circadian lighting away from focusing on melanopsin to 227 emphasizing the cone inputs. Melanopsin might have been emphasized over the powerful effects of the 228 color-opponent inputs to ipRGCs because ideas about resetting of phase in humans have been 229 extrapolated from experiments on rodents that have emphasized melanopsin. While it has been 230 recognized that ipRGCs could be activated by classic photoreceptor input in the absence of melanopsin 231 in mice 24 , neither M1 or M2 ipRGCs in mice were reported to have inputs from the color-opponent 232 circuitry observed in primates; 25,26 however, more recently, differential input between S and M cones 233 were shown to produce responses in the suprachiasmatic nucleus of mice, recognizing the importance 234 of cone inputs for circadian entrainment, especially in cone dominated species 27 . Here, we demonstrate 235 that rather than focusing on melanopsin, under the constraints of making lights that appear white with 236 intensities like standard artificial lighting used indoors, stimulating ipRGCs by modulating S-cones has 237 promise to give people control of their circadian rhythms to improve mood, sleep, and health. 238

239
All methods were performed in accordance with the relevant guidelines and regulations. Data collected 240 and used in this study is available upon request. 241 Miniature, programmable, and portable ganzfeld design Minolta) positioned 1 meter behind each goggle. The two spectrums that were alternated temporally to 274 drive high S-cone modulation were calculated theoretically using retinal sensitivities for S-, melanopsin, 275 M-, and L-retinal sensitivities given by a photopigment template 28 with peaks set at 420 nm, 480 nm, 276 530 nm, and 559 nm, respectively, corrected for absorption by the lens 29 . For the S-cone modulating 277 light, the ratio of S-cone activation between the temporally alternated spectrums was 100:1, while L-278 and M-cone activations were held constant between the two temporal phases. The alternating 279 spectrums ( Figure 1D right; top and bottom) were programmed onto the goggles and modulated at 19 280 Hz presented as a square wave with 50% duty cycle. The radiance of these lights measured at the back 281 of the goggles was 150.5 μW/cm 2 . The alternation of the two spectrums produced approximately 500 lux 282 at the subject's pupil plane as measured with a lux meter (Digital Light Meter, LX 1330B). Melanopsin 283 activation was determined by integrating the measured time averaged spectrum with the corneal 284 sensitivity for melanopsin. The two other conditions, the static white light spectrum ( Figure 1A) which 285 produce a radiance measured at the back of the goggles of 72.9 μW/cm 2 and the static blue spectrum 286 from the 476 nm LED ( Figure 1B) which produce a radiance measured at the back of the goggles of 31.6 287 μW/cm 2 , were adjusted in intensity to produce the same time averaged melanopsin activation as the S-288 cone modulated light. 289

290
The Institutional Review Board at the University of Washington approved the human subject's research. 291 Research involving human subjects was performed in accordance with local and federal regulations. 292 Human subjects research adhered to the principles embodied in the Declaration of Helsinki. Informed 293 consent was obtained from all participants. The subjects were adult volunteers from the University of 294 Washington community in Seattle. 295 Six healthy adult (2 male and 4 female) subjects (mean age = 30; range 23-43) continued with their daily 296 academic lives during the winter months (December -February) in Seattle, WA over the course of the 297 experiments. The purpose of the experiments was to determine the effects on circadian phase of three 298 different lighting paradigms which were viewed for two hours centered 10.5 hours after their individual 299 DLMO. Lights administered at this time should produce the maximum circadian phase advance ( Figure  300 1A). Circadian phase was determined from the rise in evening melatonin levels assayed from saliva 301 samples. To measure phase accurately it was important to identify subjects with a robust, reliable 302 evening rise in salivary melatonin. In addition, it is important that our participants are stability entrained 303 to the 24-hour environmental cycle even though we expect most members of the University of 304 Washington university community to suffer from some amount of phase delay. New recruits collected 305 baseline evening salivary melatonin samples every hour from 6 PM until 2 PM. During this period, they 306 were instructed to generally keep illumination levels as measured by an illuminometer below 10 lux. 307 Short periods of higher illumination were allowed, when necessary, but were always kept below 30 lux. 308 Subjects also confirmed that they were keeping a regular sleep-wake schedule in the days surrounding 309 the experiment. After the first baseline salivary melatonin measurement, the only participants that 310 continued with the experiment were those that showed a robust rise in salivary melatonin between 6 311 pm to 2 am. Four of the original recruits did not meet this requirement. Failure may be because 312 subjects' internal clocks are free running, or they may be arrhythmic. This high number of failures may 313 be a consequence of the large number of gray and short winter days in Seattle. 314 Of the six subjects who met the inclusion criteria, all are graduate students, post-docs and one assistant 315 professor involved in studies related to circadian rhythms and five of them are co-authors on this 316 manuscript. As such, they were all very motivated to adhere to the somewhat grueling demands of the 317 protocol. These included adhering to the strict evening lighting regimen, collecting saliva on a strict 318 schedule, proper handling of the saliva samples and viewing the lights at the times and durations 319 specified. We believe that having motivated compliant, participants was a key to obtaining precise and 320 reliable results. Salivary melatonin measurements are objective so the fact that participants were not 321 naïve to the objectives of the experiment could not bias the results. 322 Experimental protocol for viewing light stimuli 323 The experiment was conducted during the COVID19 pandemic. Safety protocols prevented participants 324 from coming to the laboratory for experimental procedures, thus, all experiment procedures were 325 conducted in participants' homes. Saliva samples were collected by the subjects at one-hour intervals 326 starting at 6 PM PST and placed on dry ice immediately after collection. Two separate saliva samples 327 were collected at each time point, which were analyzed separately and averaged to minimize noise for 328 each individual timepoint. Since the experiments were done in the winter in Seattle, saliva collection 329 was done well after sunset so there was no possibility of exposure to sunlight during saliva collection 330 and subjects stayed in their homes with the illumination generally kept below 10 lux and always below 331 30 lux. Circadian timing was measured by the dim light salivary melatonin onset (DLMO, Salimetrics 332 melatonin ELISA). DLMO20% was calculated as the time point at which melatonin levels reached 20% of 333 the fitted peak-to-trough amplitude of each person's data. The data was fitted to an integrated Gaussian 334 (error function) by minimizing the sum of least squares. Maximum phase advances were assumed to 335 occur 10.5 hours after DLMO20%. Administrations of a 2-hour light pulse of the therapeutic lights were 336 therefore centered around 10.5 hours after DLMO20%. Lights were administered in the subjects homes 337 the morning after the baseline internal circadian timing was measured. To determine the phase advance 338 caused by each light, circadian timing was remeasured the evening of the day the light was 339 administered. Phase advances were calculated as the difference between DLMO20% after light 340 administration and baseline DLMO20%. Differences in phase produced by the light treatments were 341 evaluated using a paired t-test using each person DLMO measurement before and after treatment as a 342 pair. 343 344 Data Availability. Contact J.A.K. to request the data from this study. 345