A Cryptochrome adopts distinct moon- and sunlight states and functions as sun- versus moonlight interpreter in monthly oscillator entrainment

The moon’s monthly cycle synchronizes reproduction in countless marine organisms. The mass-spawning bristle worm Platynereis dumerilii uses an endogenous monthly oscillator to phase reproduction to specific days. Classical work showed that this oscillator is set by full moon. But how do organisms recognize such a specific moon phase? We uncover that the light receptor L-Cryptochrome (L-Cry) is able to discriminate between different moonlight durations, as well as between sun- and moonlight. Consistent with L-Cry’s function as light valence interpreter, its genetic loss leads to a faster re-entrainment under artificially strong nocturnal light. This suggests that L-Cry blocks “wrong” light from impacting on the monthly oscillator. A biochemical characterization of purified L-Cry protein, exposed to naturalistic sun- or moonlight, reveals the formation of distinct sun- and moonlight states characterized by different photoreduction- and recovery kinetics of L-Cry’s co-factor Flavin Adenine Dinucleotide. In vivo, L-Cry’s sun-versus moonlight states correlate with distinct sub-cellular localizations, indicating different signalling. In contrast, r-Opsin1, the most abundant ocular opsin, is not required for monthly oscillator entrainment. Our work reveals a new concept for correct moonlight interpretation involving a “valence interpreter” that provides entraining photoreceptor(s) with light source and moon phase information. These findings advance our mechanistic understanding of a fundamental biological phenomenon: moon-controlled monthly timing. Such level of understanding is also an essential prerequisite to tackle anthropogenic threats on marine ecology.


22
The moon's monthly cycle synchronizes reproduction in countless marine organisms. The mass-23 spawning bristle worm Platynereis dumerilii uses an endogenous monthly oscillator to phase 24 reproduction to specific days. Classical work showed that this oscillator is set by full moon. But how 25 do organisms recognize such a specific moon phase? 26 We uncover that the light receptor L-Cryptochrome (L-Cry) is able to discriminate between different 27 moonlight durations, as well as between sun-and moonlight. Consistent with L-Cry's function as light 28 valence interpreter, its genetic loss leads to a faster re-entrainment under artificially strong nocturnal 29 light. This suggests that L-Cry blocks "wrong" light from impacting on the monthly oscillator. A 30 biochemical characterization of purified L-Cry protein, exposed to naturalistic sun-or moonlight, 31 reveals the formation of distinct sun-and moonlight states characterized by different 32 photoreduction-and recovery kinetics of L-Cry's co-factor Flavin Adenine Dinucleotide. In vivo, L-Cry's 33 sun-versus moonlight states correlate with distinct sub-cellular localizations, indicating different 34 signalling. In contrast, r-Opsin1, the most abundant ocular opsin, is not required for monthly 35 oscillator entrainment. Our work reveals a new concept for correct moonlight interpretation 36 involving a "valence interpreter" that provides entraining photoreceptor(s) with light source and 37 moon phase information. These findings advance our mechanistic understanding of a fundamental 38 biological phenomenon: moon-controlled monthly timing. Such level of understanding is also an 39 essential prerequisite to tackle anthropogenic threats on marine ecology. In order to test for a functional involvement of L-Cry in monthly oscillator function, we generated 79 two l-cry mutant alleles (Δ34 and Δ11bp) (Fig.1a) using TALENs 23 . In parallel, we generated a 80 monoclonal antibody against Platynereis L-Cry. By testing mutant versus wildtype worms with the 81 anti-L-Cry antibody in Western blots (Fig.1b) and immunohistochemistry (Fig. 1e-j), we verified the 82 absence of L-Cry protein in mutants. Furthermore, we confirmed that the staining of the antibody in 83 wildtype worms (Fig. 1e-h) matches the regions where l-cry mRNA is expressed (Fig. 1d). These tests 84 confirmed that the engineered l-cry mutations result in loss-of-function alleles. In turn, they validate 85 the specificity of the raised anti-L-Cry antibody. 86  Given that recent, non-inbred isolates from the same habitat as our lab inbred strains (which is the 135 same habitat as the data collected in ref. 29 ) exhibit a broad spawning distribution under standard 136 worm culture light conditions (which includes the bright artificial moon light) 30 , we hypothesized 137 that the difference in spawning synchrony between wildtype laboratory cultures and populations in 138 their natural habitat is caused by the rather bright nocturnal light stimulus typically used for the 139 standard laboratory culture (Extended Data Figure 1a vs. b). 140 Figure 2: L-Cry shields the circalunar clock from light that is not naturalistic moonlight (a-d,j) Spawning of l-cry +/+ (a), l-cry +/-(Δ34) (b) and l-cry -/-( Δ34/ Δ34) (c) animals over the lunar month in the lab with 8 nights of artificial moonlight (a-c), under natural conditions in the sea (d, replotted from ref. 29,31 ) and in the lab using naturalistic sunand moonlight (j, 8 nights moonlight). (eh,k) Data as in (a-d,j) as circular plot. 360° correspond to 30 days of the lunar month. The arrow represents the mean vector, characterized by the direction angle µ and r (length of µ). r indicates phase coherence (measure of population synchrony). pvalues inside the plots: result of Rayleigh Tests. Significance indicates non-random distribution of data points. The inner circle represents the Rayleigh critical value (p=0.05). (i,l) Results of multisample statistics on spawning data shown in (ah,j,k). The phase differences in days can be calculated from the angle between the two mean vectors (i.e. 12°= 1 day).

142
Lunar spawning precision of wild-type animals depends on 143 naturalistic moonlight conditions 144 We next tested the resulting prediction that naturalistic moonlight should increase the spawning 145 precision of the wildtype population, using naturalistic sun-and moonlight devices we specifically 146 designed based on light measurements at the natural habitat of Platynereis dumerilii 26 (Extended 147 Data Extended Data Fig. 1a,c). We assessed the impact of the naturalistic sun-and moonlight 148 (Extended Data Fig. 1a,c) on wildtype animals, maintaining the temporal aspects of the lab light 149 regime (i.e. 8 nights of "full moon"). Indeed, merely adjusting the light intensity to naturalistic 150 conditions increased the precision and phase coherence of population-wide reproduction: After 151 several months under naturalistic sun-and moonlight, wildtype worms spawned with a major peak 152 highly comparable to the wildtype precision reported at its natural habitat (Fig. 2d,h vs. j,k), and also 153 exhibited an increased population synchrony (r=0.398 compared to r=0.295 under standard worm 154 room light conditions). This increased similarity to the spawning distribution at the natural habitat 155 ("Sea") is confirmed by statistical analyses (Fig. 2l): The phase difference (angle between the two 156 mean vectors) is only one day (corresponding to 12°). In contrast, the spawning distribution of 157 wildtypes under standard worm room light versus naturalistic light conditions is highly significantly 158 different in linear and circular statistical tests and has a phase difference of 7.7 days (Fig. 2l). 159 These findings show that it is the naturalistic light that is critical for a highly precise entrainment of 160 the monthly clock of wildtype worms. Given that l-cry -/animals reach this high precision with the 161 artificial light (i.e. standard lab light) implies that in wildtype L-Cry blocks artificial, but not naturalistic 162 full moon light from efficiently synchronizing the circalunar clock. This block is removed in l-cry -/-163 animals, leading to a better synchronization of the l-cry -/population. This finding suggests that L-164 Cry's major role could be that of a gate-keeper controlling which ambient light is interpreted as full 165 moon light stimulus for circalunar clock entrainment. 166 l-cry functions as a light signal gate-keeper for circalunar clock 167 entrainment 168 A prediction of this hypothesis is that mutants should entrain better to an artificial full moon light 169 stimulus provided out-of-phase than their wildtypes counterparts (in which L-Cry should block the 170 "wrong" moonlight at least partially from re-entraining the circalunar oscillator). 171 We thus compared the spawning rhythms of l-cry +/+ and l-cry -/worms under a re-entrainment 172 paradigm, where we provided our bright artificial culture full moon light at the time of the subjective 173 new moon phase (scheme Fig.3a). In order to compare the spawning data distribution relative to the 174 initial full moon (FM) stimulus, as well as to the new full moon stimulus (i.e. new FM), we used two 175 nomenclatures for the months: months with numbers are analyzed relative to the initial nocturnal 176 light stimulus (i.e. FM), whereas months with letters are analyzed relative to the new (phase-shifted) 177 nocturnal light stimulus (i.e. new FM, Fig.3a). When the nocturnal light stimulus is omitted (to test 178 for the oscillator function) we then refer to 'free-running FM' (FR-FM) or 'new free-running FM' (new 179 FR-FM), respectively (Fig.3a). Using these definitions, the efficiency of circalunar clock re-180 entrainment will be reflected in the similarity of spawning data distributions between month 1 and 181 month D, i.e. the more similar the distribution, the more the population has shifted to the new 182 phase. 183 When using the artificial nocturnal light conditions, the re-entrainment of l-cry -/animals was both 184 faster and more complete than for their wildtype relatives, as predicted from our gate keeper 185 hypothesis. This is evident from the linear data analysis and Kolmogorov-Smirnov tests when 186 comparing the month before the entrainment (month 1) with two months that should be shifted 187 after the entrainment (months C,D, Fig. 3b,c,f,g). 188 Most notably, while l-cry-/worms were fully shifted in month D ( Wheeler test (see Methods section for details), which shows less/no significance in the comparison 195 of mutants before and after entrainment, but very highly significant differences in the distribution of 196 the wildtype spawning data in the same comparison. Consistently, the phase differences in days 197 calculated from the angle between the two mean vectors from the circular analysis is smaller in the 198 mutants than in the wildtypes when comparing the phase of the month before the entrainment 199 (month 1) with two months after the entrainment (months C,D) ( Fig. 3d-g). The fact that there are 200 still differences in the mutant population before and after entrainment is likely due to the fact that 201 even the mutants are not fully re-entrained. However, they have significantly shifted stronger in 202 response to an artificial nocturnal light stimulus than the wildtypes. This provides further evidence 203 that in wildtype worms L-Cry indeed blocks the "wrong" light from entering into the circalunar clock 204 and thus functions as a light gate-keeper.

217
L-Cry functions mainly as light interpreter, while its contribution as 218 direct moonlight entraining photoreceptor is minor 219 We next tested to which extent L-Cry function itself as a sensor for re-entrainment signal under 220 naturalistic light conditions. Based on the finding that l-cry-/worms can still re-entrain the circalunar 221 oscillator (see above), it is clear that even if L-Cry also directly contributed to the entrainment, it 222 cannot be the only moonlight receptor mediating entrainment. With the experiments below, we 223 aimed to test if L-Cry has any role as an entraining photoreceptor to the monthly oscillator. 224 Thus, we tested how the circalunar clock is shifted in response to a re-entrainment with naturalistic 225 moonlight in Platynereis wt versus l-cry-/worms. For this, animals initially raised and entrained 226 under standard worm room light conditions of artificial sun-and moonlight (Extended Data Figure  227 1b,e) were challenged by a deviating FM stimulus of 8 nights of naturalistic moonlight (Fig. 4a, 228 Extended Data Figure 1c,e). This re-entraining stimulus was repeated for three consecutive months 229

238
Significance indicates non-random distribution of data points. The inner circle represents the Rayleigh critical value 239 (p=0.05). (f,g) Results of multisample statistics on spawning data shown in (a-e). Phase differences in days can be calculated 240 from the angle between the two mean vectors (i.e. 12°= 1 day).

241
The resulting spawning distribution was analysed for the efficacy of the naturalistic moonlight to 242 phase-shift the circalunar oscillator. In order to test if the animals had shifted their spawning to the 243 new phase, we again compared the spawning pattern before the exposure to the new fullmoon 244 stimulus (months with numbers: data distribution analyzed relative to the initial/old FM, see scheme 245 Consistently, the direction angle (µ) of the mean vectors before and after the shift was highly similar, 254 resulting in a phase difference of only 0.2 days between months 1 and C and 0.5 days between 255 month 2 and month C (Fig. 4f, for details see methods). 256 Of note, wildtype worms would eventually reach the high spawning precision found under 257 naturalistic moonlight only after several more months based on independent experiments (Fig. 2j,k). 258 When we analyzed the spawning distribution of l-cry mutants in the same way as the wildtypes, we 259 found that the data distribution exhibited significant differences in the linear Kolmogorov-Smirnov 260 test when comparing months 1 and 2 before the shift to the months C and D after the shift ( Based on the statistical significant difference in the re-entrainment of l-cry -/-, but not wild-type 266 populations under a naturalistic sun-and moonlight regime, we conclude that L-Cry also contributes 267 to circalunar entrainment as a photoreceptor. However, as these differences are rather minor, 268 compared to the much stronger differences seen under artificial light regime, we conclude that its 269 major role is the light gate keeping function. 270 In an independent study that focused on the impact of moonlight on daily timing, we identified r-271 Opsin1 as a lunar light receptor that mediates moonlight effects on the worms' ~24hr clock 32 . We 272 tested if r-opsin1 is similarly important for mediating the moonlight effects on the monthly oscillator 273 of the worm, analyzed here. This is not the case. r-opsin1 -/animals re-entrain as well as wildtype 274 worms under naturalistic light conditions (Extended Data Figure 4). This adds to and is also consistent 275 with our above observation that the spawning distribution is un-altered between r-opsin1 -/and 276 wildtype animals under artificial light conditions (Extended Data Fig.3a,b,e,f). This finding also further 277 enforces the notion that monthly and daily oscillators use distinct mechanisms, but both require L-278 Cry as light interpreter. 279 L-Cry discriminates between naturalistic sun-and moonlight by  We then investigated the response of L-Cry to ecologically relevant light, i.e. sun-and moonlight 301 using naturalistic sun-and moonlight devices we designed based on light measurements at the 302 natural habitat of Platynereis dumerilii 26 (Extended Data Fig.1a,c,e). Upon naturalistic sunlight 303 illumination, FAD was photoreduced to FAD°-, but with slower kinetics than under the blue light 304 source, likely due to the intensity differences between the two lights (Extended Data Fig.1c-e). 305

317
While blue-light illumination led to a complete photoreduction within 110s (Extended Data Fig.5f), 318 sunlight-induced photoreduction to FAD°-was completed after 20 min (Fig. 5b) and did not further 319 increase upon continued illumation for up to 2h (Extended Data Fig.6a). Dark recovery kinetics had 320 time constants of 3.2min (18°C) and 5min (ice) (Fig.1c, Extended Data Fig.6b,c). 321 As the absorbance spectrum of L-Cry overlaps with that of moonlight at the Platynereis natural 322 habitat (Extended Data Fig. 1a), L-Cry has the principle spectral prerequisite to sense moonlight. 323 However, the most striking characteristic of moonlight is its very low intensity (1.79 x 10 10 324 photons/cm 2 /s at -5m, Extended Data Fig. 1a,e). To test if Pdu-L-Cry is sensitive enough for 325 moonlight, we illuminated purified L-Cry with our custom-built naturalistic moonlight, closely 326 resembling full moon light intensity and spectrum at the Platynereis natural habitat (Extended Data 327 photoreduction, but with different kinetics and different final FAD°-product/FADox adduct ratios, we 340 wondered how purified L-Cry would react to transitions between naturalistic sun-and moonlight (i.e. 341 during "sunrise" and "sunset"). 342 Mimicking the sunrise scenario, L-Cry was first illuminated with naturalistic moonlight for 6 h 343 followed by 20 min of sunlight exposure. This resulted in an immediate enrichment of the FAD°-state 344 (Fig. 5g). Hence, naturalistic sunlight immediately photoreduces remaining oxidized flavin molecules, 345 that are characteristic of moonlight activated L-Cry, to FAD°-, to reach a distinct fully reduced sunlight 346 state. 347 In contrast, when we next mimicked the day-night transition ("sunset") by first photoreducing with 348 naturalistic sunlight (or strong blue light) and subsequently exposed L-Cry to moonlight, L-Cry first 349 returned to its full dark state within about 30 min (naturalistic sunlight: τ=7min (ice): Fig.5h, 350 Extended Data Fig.6e; blue light: τ=9 min (ice): Extended Data Fig.6f-h), despite the continuous 351 naturalistic moonlight illumination. Prolonged moonlight illumination then led to the conversion of 352 dark-state L-Cry to the "moonlight state" (Fig. 5i, Extended Data Fig. 6f). Hence, fully photoreduced 353 "sunlight-state" L-Cry first has to return into the dark state before entering the "moonlight-state" 354 characterized by the stable presence of the partial FAD°-product/FADox adduct. In contrast to 355 "sunlight-state" L-Cry, "moonlight-state" L-Cry does not return to the oxidized ("dark") state under 356 naturalistic moonlight, i.e. moonlight maintains the "moonlight-state", but not the "sunlight-state". 357 Taken together, these results indicate the existance of kinetically and structurally distinct "sunlight" 358 and "moonlight" states of L-Cry (Fig.5j, Extended Data Fig.6i). 359 Naturalistic sun-and moonlight differently affect L-Cry subcellular

383
In contrast to sunlight, exposure to an equal length (8hrs) of naturalistic moonlight did not cause a 384 reduction in L-Cry levels compared to an equivalent time (8hrs) in darkness (FM-1, zt0-10min versus 385 FM7, zt0-10min: Figs.6b,c, Extended Data Fig.7). Thus, any potential moonlight signalling via L-Cry 386 occurs via a mechanism independent of L-Cry degradation. 387 We next examined the spatial distribution of L-Cry in worm heads (scheme Fig. 6d) at lunar and diel 388 timepoints ( Fig. 6a-a''). After 8hrs of a dark night (i.e. NM, zt0-10min), L-Cry is found predominantly 389 in the nucleus of individual cells, (Fig. 6e-e''', quantification as numerical data, i.e. 390 nuclear/cytoplasmic ratio: Fig. 6h, for quantification as categorical data 34 : Extended Data Fig. 8a'-c'', 391 d-f). Given that an equivalent time of 8hrs of sunlight exposure results in strong degradation of L-Cry 392 and hence loss of staining signal (see Western blots above), we analyzed L-Cry's localization after a 393 short exposure. Already after 10mins of exposure to naturalistic sunlight (NM zt0+10mins, Fig. 6a,a'), 394 the L-Cry nuclear localization strongly diminished, becoming predominantly cytoplasmic (Fig. 6f- Given the degradation of L-Cry by naturalistic sunlight, we next asked the question if L-Cry is present 399 at night timepoints, allowing for sufficient exposure to naturalistic moonlight to reach the moonlight 400 state. We tested two diel timepoints of the first night lit by the naturalistic moonlight for circalunar 401 entrainment (FM1): at zt16 (just after the naturalistic sunlight is off and moonlight is on) and at zt20 402 (after 4hrs of naturalistic moonlight exposure) (Extended Data Fig. 9a,a'). We observe that low levels 403 of L-Cry can already be detected at FM1 zt16 (Extended Data Fig. 9b-b'''), and increase within the 404 next hours (see FM1 zt20, Extended Data Fig. 9c-c'''), with a predominantly nuclear L-Cry localization. 405 At this timepoint still 4hours of moonlight illumination remain for the protein to biochemically reach 406 the full moonlight state (ZT20 to ZT24). Based on these data we conclude that within the organism 407 and under natural conditions (with the moon illuminating at least 8h of the night under full moon 408 conditions even during summer photoperiods), L-Cry has sufficient time to reach its moonlight state 409 (by changing from sunlight to dark to moonlight state and/or by de novo synthesis of dark adapted L-410 Cry that reaches the moonlight state within 4hrs-see biochemical kinetics, Fig.5d-j, Extended Data 411 Figures 6f,g). 412 Upon further naturalistic moonlight exposure for seven continuous nights (FM7, zt0-10min) L-Cry 413 remained clearly nuclear (Fig. 6g-g''', numerical quantification Fig. 6h, categorical quantification: 414 Extended Data Fig. 8f). Thus, the sunlight and moonlight-states of L-Cry correlate with distinct 415 subcellular distribution patterns. In fact, we observed that L-Cry at FM7, zt0-10min is even more 416 nuclear restricted than at zt0-10min under NM, both in the numerical analysis of the 417 nuclear/cytoplasmic ratio (Fig. 3h), as well as in the blind categorical scoring (Extended Data Fig 8f). 418 This suggests that also the dark and moonlight states of L-Cry have distinct subcellular distribution 419 patterns. 420 Complementing the spawning analyses on genetically mutated animals, these findings show that 421 moonlight and sunlight impact differentially on L-Cry quantity and localization. 422 This allows us to put forward a model, in which L-Cry directly via its biochemical states and 423 connected cellular signalling properties is able to discriminate between (naturalistic) sun-and 424 moonlight and to function as a gate-keeper for potentially entraining light stimuli for the circalunar 425 oscillator (Fig. 7b). But why would it be required to do this in nature? As we expand in more detail in 426 the discussion, we speculate that this is necessary to entrain to a specific moon phase, which is the 427 full moon phase for Platynereis. This moon phase is specifically characterized by the long duration of 428 detectable moonlight, i.e. moonlight during the entire night 21 (Fig.7a). Interestingly, this matches the 429 biochemical kinetics of at least 6hours of light exposure to acquire L-Cry's biochemical moonlight 430 state. However in nature, where the setting of the full and waning moons is immediately followed by 431 sunrise (i.e. no darkness window, Fig.7a, 21 ), measuring the duration of light exposure alone would 432 not allow the worms to detect a specific moonphase (Fig.7a). Thus, under the natural conditions of 433 waning/waxing moonphases and sunrise/sunsets, being able to detect the switch from moonlight to 434 sunlight is essential to determine the end of the moonlight phase and thus to discriminate between 435 full moon and waning moon phases (Fig.7a). 436 Furthermore, L-Cry's gate-keeping mechanism likely also makes the entrainment system more stable 437 against irregular illumination as it could arise from thunderstorms. 438

450
Our work delivers the first molecular entry point into the mechanisms underlying a moonlight-451 entrained monthly oscillator. It also provides the new concept that a light receptor does not just 452 sense light, but by its intensity and duration can give light a valence that is relevant to discriminate 453 between different naturally existing light sources (Fig. 7a,b). While we see the most apparent 454 behavioral/physiological phenotype of the l-cry-/worms under artificial lab light conditions, these 455 conditions are nevertheless highly informative about l-cry's role as light valence detector. As briefly 456 mentioned above, we interpret that this valence detection is under natural conditions necessary in 457 order for the worms to synchronize to a specific moonphase: full moon. Full moon has the specific 458 property that it is the moonphase during which the moon illuminates the entire night from sunset to 459 sunrise (Fig.7a). In order for the organism to 'know' this specific moonphase under natural conditions 460 with waning and waxing moons as well as sunrise/sunsets (Fig.7a), it needs to determine the 461 duration not just of illumination, but specifically of the dim light illumination. This discrimination is 462 made by L-Cry. Under the lab artificial light conditions, the moonlight stimulus is much more intense 463 and misses signals of the waning/waxing moon phases. We hypothesize that under these artificial 464 situations, the circalunar clock is still somewhat entrained, because there is no other entrainment 465 stimulus it otherwise can entrain to. However, L-Cry signals that it is not really the "right" nocturnal 466 light, which results in the observable, rather low population synchronization. If L-Cry is not present 467 (as in the l-cry-/worms), the nocturnal artificial light signal of the lab condition fully impacts on the 468 circalunar clock. As it is (artificially) highly precise without possibly confusing waning and waxing 469 moon signals, the entrainment results in the observed higher synchrony of the l-cry-/population. If In its role as a light-signal gate keeper, only the accumulation of L-Cry molecules in the nuclear 501 moonlight signalling state following prolonged moonlight exposure during full-moon, would enable 502 lunar entrainment via an additional photoreceptor "X", which by itself is not able to discriminate 503 between the correct full-moon signal and other "wrong" signals, such as sunlight or (in our lab 504 experiment) the artificial/non-naturalistic nocturnal light source (Fig. 7b). 505 When L-Cry is photoreduced by light other than (naturalistic) moonlight, the light signalling of 506 photoreceptor X towards the circalunar oscillator is inhibited (Fig.7b). L-cry mutant worms lack this 507 inhibitory mechanism, resulting in the observed (unnatural) synchronisation to artificial moon light. 508 On the other hand, the somewhat better response of wildtype worms to naturalistic moonlight under 509 re-entraining conditions indicates that the accumulation of moon-light state L-Cry not only releases 510 the inhibition, but might enhance the activity of the yet to be identified photoreceptor X or provide 511 additional light signalling by itself to the monthly oscillator. 512 Connected to the question of the transmission of the moonlight signal to the circalunar oscillator is 513 also a better understanding of L-Cry's "moonlight state". Is this state "just" a partial photoreduction 514 of the state reached upon artificial sunlight exposure or perhaps also a conformationally different 515 state with distinct formation and decay kinetics? And what is the role of the L-Cry dimers? An 516 intriguing observation is, that in presence of moonlight the moonlight state can be stably maintained 517 over several hours, whereas the sunlight state completely reverts to the fully-oxidized dark-state 518 within minutes without accumulating the moonlight state while transitioning through partial 519 photoreduction (Fig.5j). These different responses to moonlight illumination suggest that the 520 moonlight-and sunlight states are conformationally and kinetically not equivalent (Extended Data 521 Figure 6i). Based on its sequence homology and the similarity of its FAD photoreaction to Drosophila 522 CRY (dCry), it is conceivable that L-Cry also displaces the regulatory C-terminal tail in the 523 photoreduced state as observed for dCry 33,36 . However, as dCry is monomeric, L-Cry homodimer 524 formation may impact these conformational changes, and these may further vary depending on 525 whether moonlight or sunlight operates on the initial dark-state L-Cry homodimer. We propose, that 526 partial FAD photoreduction in the moonlight state could be related to the formation of asymmetric L-527 Cry dimers, where one monomer retains oxidized FAD, while in the second monomer FAD is 528 photoreduced to FAD°- (Fig. 5j). This requires, that the flavins in the two L-Cry monomers have 529 different redox potentials, likely resulting from different chemical environments due to 530 conformational differences between the monomers (Extended Data Figure 6i). Hence different 531 amounts of energy (photon numbers) would be needed to photoreduce the flavins in the two L-CRY 532 monomers. Moonlight, due to its very low intensity can only induce the lower energy transition, 533 resulting in the partially photoreduced moonlight state. In presence of intense sunlight, however, the 534 larger energy barrier to photoreduce the second flavin can also be overcome. Certainly, more 535 extensive mechanistic studies are required to further support our model. However, this model is 536 consistent with all our current in vitro data, and moreover, it plausibly illustrates how the very 537 different intensities of moon-and sunlight can lead to the formation of conformationally distinct dark 538 state (new moon), moonlight state (full moon) and sunlight state L-Cry proteins. Thereby L-Cry could 539 translate different light qualities into different cellular signaling events, e.g. by changing L-Cry's 540 subcellular localizations and cellular degradation rates (Fig. 6), to ultimately affect moonlight 541 dependent physiology (Fig. 2-4). 542 Finally, an evolutionary consideration: Monthly synchronization by the moon has been documented 543 for a wide range of organisms-including brown and green algae, corals, crustaceans, worms, but also 544 vertebrates (reviewed in 6 ). Furthermore, recent reports also provide increasing evidence that the 545 lunar cycle influences human behaviour (reviewed in 21,37 ). Are the lunar effects mediated by 546 conserved or different mechanisms? 547 When considering monthly oscillators with period lengths in the range of weeks, our implication of L-548 Cry as a light receptor in the circalunar entrainment pathway at first glance rather suggests that such 549 monthly oscillator might not be conserved, given that direct L-Cry orthologs are not present in all the 550 groups that are affected by the lunar cycle 38 . However, taking further aspects into account, such a 551 conclusion might be too quick. Could other members of the Cry/photolyase family take over similar 552 functions? Furthermore, our entrainment data suggest the presence of additional moonlight 553 entrainment photoreceptors, which might be conserved. Last, but not least the molecular 554 mechanisms underlying the circalunar oscillator also await identification, and it is possible that 555 conservation exists on this level. Examples are known from circadian biology and it will now require 556 further work to reach a similar level of understanding for moon-controlled monthly rhythms and 557 clocks. 558

559
Natural light measurements 560 Under water measurements of natural light at the habitat of Platynereis dumerilii were acquired 561 using a RAMSES-ACC-VIS hyperspectral radiometer (TriOS GmbH) for UV to IR spectral range. In 562 coastal waters of the Island of Ischia, in the Gulf of Naples, the two radiometers were placed on sand 563 flat at 5m depth near to Posidonia oceanica meadows, which are a natural habitat for P. dumerilii. 564 Measurements were recorded automatically every 15min across several weeks in the winter 565 2011/2012. To obtain a fullmoon spectrum, measurements taken from 10pm to 1am on a clear 566 fullmoon night on the 10.11.2011 were averaged. To subtract baseline noise from this measurement, 567 a NM spectrum was obtained by averaging measurements between 7:15pm to 5am on a NM night on 568 24.11.2011, and subtracted from the FM spectrum. Resulting spectrum: Extended Data Fig. 1a. To 569 benchmark these moonlight spectra measured under water with moonlight measured on land, we 570 compared the underwater spectra to a publicly available full moon spectrum measured on land on 571 14.04.2014 in the Netherlands (Extended Data Fig. 1g, spectrum available at 572 http://www.olino.org/blog/us/articles/2015/10/05/spectrum-of-moon-light). As expected, light with 573 longer wavelengths was strongly reduced in the underwater measurements compared to the surface 574 spectrum, since longer wavelengths penetrate water less efficiently. For the sunlight spectrum, 575 measurements taken from 8am to 4pm on a sunny day on 9.11.2011 were averaged. 576

659
To assess if sunlight can further increase FAD photoreduction starting from the moonlight activated 660 state, L-CRY was first illuminated with continuous moonlight for 6 hours, followed by 20 min of 661 sunlight illumination (on ice). Complete UV-VIS spectra from 300 -700 nm were measured in each 662 case. 663

664
Platynereis dumerilii were grown as previously described 15,39 . All animal work was conducted 665 according to Austrian and European guidelines for animal research. Generation and Genotyping of l-cry KO worms 671 Design and construction of TALENs targeting l-cry is described in 23 . For genotyping, DNA extraction of 672 immature and premature worms was conducted by cutting 5-10 tail segments with a scalpel and 673 incubating them in 20µl 50mM NaOH at 95° for 20min. After adding 5µl of Tris/HCl pH 7.5, the 674 supernatant was used as template for the PCR reaction. Mature worms were frozen as whole at -675 20°C and DNA was later extracted using NucleoSpin Tissue Mini kit for DNA from cells and tissue 676 Immunohistochemistry, microscopy and L-Cry localization determination 690 Worm heads were dissected with jaws and fixed in 4% PFA for 24h at 4°C. Samples were 691 subsequently permeabilized using methanol, digested for 5 min with Proteinase K at room 692 temperature without shaking and post-fixed with 4% PFA for 20 min at room temperature. Next, 693 samples were washed 5 times for 5 min with 1x PTW and incubated in hybridization mixture 40 used in 694 in situ hybridization protocol, at 65° C overnight. Worm heads were washed with 50% formamide/2X 695 SSCT -standard saline citrate containing 0.1% Tween 20® (Sigma Aldrich) (2x, 20 min), then with 2X 696 SSCT (2x, 10 min) and with 0.2X SSCT (2x, 20 min); all washing steps at 65° C. After blocking for 90 697 min with 5% sheep serum (Sigma-Aldrich) at room temperature, samples were incubated in L-Cry 698 antibodies 5E3-3E6-E8 (1:100) and 4D4-3E12-E7 (1:50) in 5% sheep serum (Sigma-Aldrich). Secondary 699 the appropriate primary antibody diluted in 2.5% milk/PTW at 4°C overnight. [anti-L-Cry 5E3-3E6-E8 734 (1:100) and anti-L-Cry 4D4-3E12-E7 (1:100); anti-beta-Actin (Sigma, A-2066, 1: 20.000)]. After 3 rinses 735 with 1xPTW the membrane was incubated with the species specific secondary antibody [anti-Mouse 736 IgG-Peroxidase antibody, (Sigma, A4416, 1:7500); Anti-rabbit IgG-HRP-linked antibody (Cell Signaling 737 Technology, #7074, 1:7.500] diluted in 1xPTW/1% slim milk powder for 1 hour, RT. After washing, 738 SuperSignal™ West Femto Maximum Sensitivity Substrate kit (Thermo Fisher Scientific) was used for 739 HRP-signal detection and finally signals were visualized by ChemiDoc Imaging System (BIORAD). 740 Specific protein bands were quantified in "Image J" and L-Cry was normalized to beta-Actin. 741 Collection and analysis of spawning data 742 Worm boxes were checked daily for mature worms. Worms which had metamorphosed into their 743 sexually mature male or female form and had left their tube to perform their nuptial dance were 744 scored as mature animals. 745 The recordings of mature animals in nature (collected from June 1929 to June 1930 in Naples 29,31 ) 746 were digitalized and all months were aligned to relative to the same moonphase and combined. For 747 comparisons of these data with our spawning data from the lab, we aligned the first day after full 748 moon in nature with the last day of full moon stimulus in the lab, since Platynereis dumerilii 749 synchronizes its circalunar clock to the end of the full moon stimulus 5 . 750 For analysis, each day of the lunar month was assigned a number from 1 to 30. For linear plots, the 751 percentage of mature worms per lunar day was then plotted as a histogram. The spawning 752 distributions of two conditions were compared using the Kolmogorov-Smirnov Test. For the circular 753 analysis 43-45 of spawning data, the lunar day of spawning was multiplied by 12 for each worm, so that 754 the 30 lunar days regularly distributed on the 360° circle. Each dot represents one mature worm 755 unless stated otherwise. Circular data can be described using the mean vector (displayed as an 756 arrow), which is defined by its direction angle (µ) and its length (r). The direction angle µ is given 757 relative to 0° (moon off). The value of length r (also called phase coherence) ranges from 0 to 1, 758 where higher values indicate higher phase coherence (i.e synchrony). In order to test, if the observed 759 data distribution is significantly different from random, we performed the Rayleigh's Uniformity Test 760 and used p<0.05 as cutoff for significance. Non-uniform distribution is consistent with lunar 761 rhythmicity. For comparing two circular datasets (e.g. of different genotypes or different months in 762 the phase shift experiments), we used the non-parametric Mardia-Watson-Wheeler test. Circular 763 analysis of these data was performed using Oriana (Version 4.02, Kovach Computing Services). 764