Two light sensors decode moonlight versus sunlight to adjust a plastic circadian/circalunidian clock to moon phase

Significance The moon provides highly reliable time information to organisms. Whereas sunlight is known to set daily animal timing systems, mechanistic insight into the impact of moonlight on such systems remains scarce. We establish that the marine bristleworm Platynereis dumerilii times the precise hours of mass spawning by integrating lunar light information into a plastic daily timing system able to run with circadian (∼24 h) or circalunidian (∼24.8 h) periodicity. The correct interpretation of moonlight is mediated by the interplay of two light sensors: a cryptochrome and a melanopsin ortholog provide information on light valence and moonrise time, respectively. Besides its ecological relevance, our work provides a plausible explanation for long-standing observations of light intensity–dependent differences in circadian clock periods.

Platynereis dumerilii reproduces by nocturnal mass spawning, with sexually mature males and females synchronously rising from seagrass to the water surface (Fig. 1A) during the night (3). Whereas it is well established that this swarming is synchronized to specific nights of the month by a circalunar oscillator (1,4,5), we reasoned that it should further increase reproductive success if worms synchronized the onset of swarming behavior also to specific hours during those nights. In fact, such an interconnection of different timing systems is well established for polychaete relatives like the palolo worms (6) and fireworms (Odontosyllis) (7). Indeed, modulatory effects of the moon on behavior and physiology have been described for various animals (8,9).
This prompted us to investigate if P. dumerilii also exhibits preferred hours of swarming. We placed maturing, monthly (circalunar) entrained P. dumerilii adults (5,10) in individual wells of our automated behavioral recording device (11). As swarming is accompanied by a burst of swimming activity ("nuptial dance"), analysis by automated video tracking allowed us to systematically deduce the time of swarming onset with respect to the daylight/darkness ( light:dark [LD], 16:8 h) cycle (SI Appendix, Fig. S1 A and B and Movie S1). Analyses of 139 individuals revealed that swarming onset across the culture was indeed synchronized to an ∼1to 2-h window during the night (Fig. 1B). (Note that we selected about equal numbers of swarming worms per night. Therefore, the monthly swarming synchronization is invisible.) The precise time point depended on the time since the last artificial full moon (FM) night (Fig. 1B), which is provided to entrain the worms' monthly oscillator (4,5). In nights directly following the last FM night, animals started the characteristic swarming behavior directly following night onset. This onset of swarming gradually shifted by approximately 44 min per night within the first eight nights (Fig. 1B, days preceding the red arrow). For the remaining lunar month, the time of swarming onset remained unaltered at ∼5 h after night onset ( Fig. 1B and SI Appendix, Fig. S1B). To assess whether this synchronization was driven by an endogenous oscillator, we next monitored

Significance
The moon provides highly reliable time information to organisms. Whereas sunlight is known to set daily animal timing systems, mechanistic insight into the impact of moonlight on such systems remains scarce. We establish that the marine bristleworm Platynereis dumerilii times the precise hours of mass spawning by integrating lunar light information into a plastic daily timing system able to run with circadian (∼24 h) or circalunidian (∼24.8 h) periodicity. The correct interpretation of moonlight is mediated by the interplay of two light sensors: a cryptochrome and a melanopsin ortholog provide information on light valence and moonrise time, respectively. Besides its ecological relevance, our work provides a plausible explanation for long-standing observations of light intensity-dependent differences in circadian clock periods.
swarming onset in worms that were kept in constant darkness for several days. Under these dark-dark (DD) conditions, swarming was still synchronously initiated, with an average delay of ∼1 ± 0.1 h/d (Fig. 1C). This established that the specific hour of nocturnal swarming onset is controlled by an endogenous clock.
The time delay of about 44 min within the first eight nights after FM is reminiscent of the average delay of the rise of the waning moon (∼49 min per night; Fig. 1D). This apparent delay of moonrise time relative to sunset is caused by the period difference of the daily solar cycle (24 h) and the lunidian cycle (24.8 h; the average time span between two successive moonrises) (Fig. 1D). The latter matches the period length of the endogenous clock (∼25 h) controlling swarming onset under DD conditions (compare Fig. 1 C and D). The combination of these facts let us speculate that the worm's ∼24-h timing system could help to synchronize swarming onset to the darkest hours of the night but would require the moon for entrainment. Furthermore, the exact change of moonrise relative to sunset is not always exactly ∼49 min per night but varies under natural conditions ( Fig. 1D 0 ), making an additional adjustment by moonlight likely advantageous. We thus next studied if the endogenous clock was sensitive to moonlight for its exact entrainment. To mimic moonlight and sunlight under laboratory conditions, we complemented available surface measurements (12) by analyzing systematic light measurements at a natural habitat of Platynereis (SI Appendix, Fig. S2A), which guided the design of naturalistic sunlight and naturalistic moonlight illumination devices (SI Appendix, Fig. S2B, and see also refs. 1, 11). We next exposed animals (≥9 d after the end of the monthly nocturnal light stimulus; red arrow in Fig. 1B) to naturalistic moonlight (SI Appendix, Fig. S2B) provided during the second half of the night for five consecutive nights (Fig. 1E, LM1 to LM5). In response to this light regime mimicking waning moon, worms shifted their swarming onset gradually into the dark portion of these moonlit nights (Fig. 1E). The advanced swarming onset caused by the waning moonlight regime persisted when worms were subsequently released into constant darkness (Fig. 1E, DD1 and DD2), arguing that this shift was caused by an impact of moonlight on the endogenous clock, rather than being an acute masking effect (i.e., direct response to light). When the same naturalistic moonlight was provided during the first half of the night (mimicking times of waxing moon), no shift of swarming times was observed (Fig. 1F). Comparisons of a subsequent constant darkness period (Fig.  1F, DD1 and DD2) showed a significant difference of clockcontrolled swarming onset compared to the waning moonlight regime (Fig. 1G). Finally, under a constant naturalistic moonlight (MM) regime, swarming onset remained synchronized but occurred with a period length of ∼22.2 ± 0.3 h (Fig. 1H), a clear period shortening compared to DD conditions (Fig. 1C). Taken together, these results suggest the existence of a plastic oscillator system that regulates nocturnal swarming onset. In the absence of a moonlight stimulus it free-runs during the first nine nights after FM. Naturalistic moonlight mimicking the full or waning moon modulates the clock's period and/or resets the clock's phase. This results in a swarming preference during the dark portion of the night, consistent with natural observations. We refer to this clock as plastic circadian/circalunidian (PCC) clock.

L-Cry Is Required to Correctly Interpret Sunlight and Moonlight to Set the Swarming Hour
In order to understand how (naturalistic) sunlight and moonlight are sensed and distinguished by this system, we next sought to identify photoreceptor(s) relevant for the light impact on the PCC clock. One candidate receptor of particular interest was Platynereis L-Cryptochrome (L-Cry). Based on expression changes, a distant homolog (Cry2) in the coral Acropora has been speculated to mediate moonlight sensation (13). In a separate study, we uncovered that Pdu-L-Cry has the biochemical and cellular properties to discriminate between sunlight and moonlight and is required to correctly detect the FM phase for monthly oscillator entrainment (1).
To assess if Platynereis L-Cry is also relevant for the light input into an oscillator with the period length of ∼24 h, the PCC clock, we analyzed a Platynereis l-cry loss-of-function strain generated by Transcription activator-like effector nucleases (TALEN) technology (for details on mutants, see ref. 1). When exposed to constant darkness, l-cry À/À individuals still exhibited rhythmic initiation of swarming onset, with a period length (24.6 ± 0.3 h) indistinguishable from wild types ( Fig.  2A). This indicates that L-Cry is not required for the endogenous oscillation of the PCC clock.
To probe for roles of L-Cry in mediating light input into the PCC clock, we next investigated swarming rhythmicity in l-cry -/mutants exposed to constant naturalistic moonlight  Fig. 2A). In contrast, wild-type siblings shortened their period (MM) or became arrhythmic (LL) (Fig. 2 B and C). These clear differences between wild type and mutants let us conclude that L-Cry is relevant for the conveying of naturalistic sunlight and moonlight information to the PCC clock.
The absent adjustment of the PCC clock in l-cry À/À individuals to respond to light could be explained by a general reduction in light sensitivity. Alternatively, these findings are compatible with a role of L-Cry in distinguishing moonlight and sunlight, as L-Cry enables the PCC clock to respond differently to the two light conditions and as it also possesses this property in the context of the worm's monthly oscillator (for monthly oscillator, see ref. 1). To discriminate between the two possibilities, we exposed l-cry mutants to a combined naturalistic day/night light regime of 16:8 h, where they were exposed to naturalistic sunlight during the day and naturalistic moonlight during the entire night, mimicking the FM situation (light:moonlight [LM]) (Fig.  2D). If l-cry À/À animals were simply blind to light, they should continue to exhibit the swarming timing seen in Fig. 2 B and C. However, if l-cry À/À rather provided interpretation on the nature of the light stimulus to other photoreceptors, the prediction was that this would cause an increased behavioral variability between individual worms as critical light valence information is missing in a mixed naturalistic sunlight/moonlight regime. Indeed, unlike wild-type animals, which restricted swarming onset strictly to nocturnal hours ( Fig. 2 D and D 0 ), and different from the timing observed under constant moonlight (Fig. 2B) or constant sunlight (Fig. 2C), l-cry -/mutants exhibited aberrant, much more variable swarming onset timing under the complex naturalistic sunlight and moonlight regime (Fig. 2D), with a significant proportion (∼19%) of l-cry À/À worms initiating swarming during the day (Fig. 2D 0 ). Furthermore, and also consistent with L-Cry's valence function, all l-cry -/mutants restricted swarming onset to the night when no moonlight was present, albeit slightly less synchronized and earlier than wild type (Fig. 2 D [see LD before LM] and Fig. 2E). This further supports that the shifted timing into the day was caused by the naturalistic moonlight stimulus, which got misinterpreted by the l-cry À/À animals. Our findings indicate that the l-cry mutation does not simply render worms less sensitive to moonlight, but that L-Cry is required to provide the correct light valence information (valence is discrimination of sunlight versus moonlight based on intensity and/or spectrum) to the PCC clock. Furthermore, our data are consistent with the idea that constant moonlight during the night (i.e., FM) is used to reset the hour of swarming to the early night hours observed directly after FM, likely via modulating the period length of the PCC clock.
In an extended nontiled waning moonlight regime (as used in Fig. 1E), l-cry À/À mutants and wild-type individuals swarmed similarly (SI Appendix, Fig. S3A), except that l-cry À/À mutants initiated swarming again already slightly earlier during the LD cycle (SI Appendix, Fig. S3A and compare Fig. 2 D and E). By contrast, an abnormal (confused) swarming onset of l-cry À/À animals similar to the FM situation ( Fig. 2D) was observed in a light regime in which a tiled, waning moonlight stimulus (SI Appendix, Fig. S2C) was provided directly following the standard monthly culture FM stimulus (Fig. 2F). The tiled moonlight regime more closely mimics the natural timing of moonlight during nights in which swarming is observed ( Fig. 2F and compare Fig. 1 D and D 0 ). Under a corresponding LD regime lacking the moonlight stimulus, the abnormal swarming onset of the mutants was not observed (Fig. 2G).
We reason that the phenotypic difference between the nontiled and tiled waning moonlight regimes is likely due to the ability of worms to adjust their circadian output also to the solar photoperiod (a common phenomenon in various animals). Indeed, an equivalent long photoperiod has comparable effects (SI Appendix, Fig. S3B). Thus, even an unclear interpretation of light by l-cry -/mutant worms results in a similar behavioral output. Likely l-cry -/mutants misinterpreted the moonlight stimulus in a nontiled waning regime as a long photoperiod (SI Appendix, Fig. S3 A and B). By contrast, the more naturalistic tiled waning moonlight regime creates similarly abnormal swarming times in l-cry -/mutants, as does the naturalistic FM regime. This emphasizes the importance of naturalistic light regimes for functional light receptor studies. It further confirms the light discriminatory role of L-Cry, which is especially relevant under more naturalistic conditions.

Evidence for Distinct Signaling of L-Cry under Sunlight versus Moonlight
In the common view based on the work in Drosophila melanogaster, the fly 1:1 ortholog of L-Cry-dCRY (14)-undergoes light-dependent binding to Timeless, which leads to the degradation of both Timeless and dCRY, thus resetting the flies' circadian clock upon light input (reviewed in ref. 15). This binary signaling model is difficult to reconcile with our finding that in the adjustment of the Platynereis PCC clock, L-Cry is relevant for the distinction between moonlight versus sunlight, characterized by different irradiance levels and spectra.
Furthermore, we had observed that under conditions relevant for monthly oscillator entrainment, L-Cry's protein levels and subcellular localization markedly differ between naturalistic sunlight versus moonlight conditions (1). As the conditions relevant for the PCC clock entrainment are different from those relevant for the monthly oscillator entrainment, we therefore tested if L-Cry protein in the worm exhibited any differences when animals were exposed to naturalistic sunlight or moonlight under conditions relevant for the behavioral paradigms shown in Figs. 1 and 2 (and see Fig. 6) and SI Appendix, Figs. S3 and S7. For this, we made use of a Pdu-L-Cry-specific antibody (1). We first assessed L-Cry abundance in head extracts of animals sampled at the midpoint of the subjective night (at new moon), after 4 h of darkness or exposure to either naturalistic sunlight or moonlight ( Fig. 3 A, CT20, red arrows). As expected by the canonical Drosophila model, and consistent with our previous analyses in S2 cells (5), naturalistic sunlight led to a significant reduction of L-Cry compared to heads sampled from animals maintained in darkness ( Fig. 3 B and C). In contrast, the levels of L-Cry protein in the heads of naturalistic moonlight-exposed animals were indistinguishable from dark levels ( Fig. 3 B and C).
Immunohistochemical analyses at two distinct time points of the respective light regime (ZT/CT20; CT0, black arrows in J-K 00 ). The very low L-Cry levels we detect here under continuous naturalistic sunlight, together with our previous analyses (1,5), support the notion that a degradation pathway is triggered by sunlight but not moonlight. In combination with the behavioral phenotypes, this further strengthens the concept that L-Cry is required for the correct interpretation of sunlight versus moonlight for the PCC clock.

Pharmaceutical Disruption of Canonical Core Circadian Clock Oscillations Affects the PCC Clock
We next wondered whether the PCC clock required the activity of the conventional core circadian clock. We previously showed   Fig. S4. (J-K 00 ) Quantification of immunofluorescent images, with a threshold (gray) indicating the mean value of immunofluorescence in l-cry À/À mutants. *P < 0.05; **P < 0.01, ***P < 0.001; ****P < 0.0001 for unpaired t test; ns: not significant.
that an inhibitor of casein kinase 1δ/ε, PF670462, disrupts the worms' core circadian clock gene oscillations (5). The effect of this drug on the core circadian clock has also been shown in several other aquatic animals, as diverse as cnidarian, crustacean, and teleost fish species (16)(17)(18).
After validating that an incubation in 160 nM of PF670462 abolished molecular oscillations of core circadian clock transcripts (SI Appendix, Fig. S5A), we assessed the effects of the drug on the timing of swarming onset. In contrast to mocktreated controls, the swarming onset in constant darkness was disrupted upon drug treatment (SI Appendix, Fig. S5B). This finding is consistent with the notion that at least a subset of canonical circadian clock genes is required for the PCC clock, although we can at present not rule out that this effect could be caused by other targets of casein kinase 1δ/ε.

dCRY Prevents the Fly's Circadian Clock from Misinterpreting Moonlight
As a regular nocturnal stimulus, moonlight reaches aquatic and terrestrial habitats. The ability to properly discriminate between moonlight and sunlight is therefore likely important for any species that uses light-sensitive clocks. In many species, the conventional circadian clock should likely run with a constant period, irrespective of lunar phase. Thus, moonlight would need to be blocked from interfering with circadian rhythmicity in those organisms. Indeed, whereas fruit fly circadian behavior can be experimentally entrained to LD cycles with light below FM light intensity (19,20), and constant light at moonlight intensity can extend the period length of wild-type flies (21,22), moonlight does not cause major effects on the circadian clock when combined with an LD cycle in this species (23)(24)(25)(26).
Given our results about the importance of Platynereis L-Cry in discriminating between naturalistic sunlight versus moonlight and Drosophila dCRY being its direct 1:1 ortholog, we hypothesized that this principal functionality of the d/L-Cry family might also be present in D. melanogaster. Specifically, we wondered if nocturnal light mimicking moonlight would cause an increased shift of the circadian clock in cry mutant flies compared to controls. We monitored locomotor behavior of both cantonized cry 01 (27) and Canton-S wild-type flies under LM conditions, adapting an existing locomotor paradigm (28) and using an artificial moonlight source matching FM light intensities measured on land (SI Appendix, Fig. S2 D and E). In wild-type flies, moonlight delayed the evening peak to 2.2 ± 0.13 h (mean ± SEM) after night onset ( Fig. 4 A and C), in line with previous observations (24), whereas cry 01 mutants exhibited a significantly stronger delay, with the evening activity peak shifting to 4.4 ± 0.11 h (mean ± SEM) after night onset (Fig. 4 B and C). In comparison, no difference in evening peak activity was observed when wild-type versus cry 01 mutants were shifted from LD to DD (SI Appendix, Fig.  S6A). Consistent with our hypothesis, the difference was instead clearly visible between the same wild-type versus cry 01 mutant flies under MM (SI Appendix, Fig. S6 B and C).
The increased delay of the evening activity peak in cry 01 mutants versus wild type under moonlight could be caused either by acute effects of artificial moonlight on behavior or by a shift in the fly's circadian clock. In order to discriminate between these possibilities, we subjected flies to artificial LM conditions and used an established immunolabeling strategy to systematically assess, over 10 distinct time points, changes in the abundance of the core circadian clock protein Period (PER) in the lateral neurons harboring the fly's circadian pacemaker. Anatomical location and the presence or absence of immunoreactivity against the neuropeptide PDF allowed us to quantify Period abundance in l-LN v s and s-LN v s (below also referred to as morning [M] cells), as well as fifth s-LN v s and LN d s (clusters harboring the evening [E] cells) (Fig. 4 D-H).
Quantification across 132 Canton-S wild-type individuals exposed to LM conditions revealed that oscillations of Period protein levels in the different subclusters were in synchrony with each other (Fig. 4D). In contrast, the corresponding cry 01 mutants exhibited pronounced desynchronization of Period protein oscillations between cell groups, with E cells differing from M cells by ∼6 h (Fig. 4E). Similar analyses of cry 01mutant flies raised in various LD cycles have not revealed such desynchronization (29), indicating that the effects we observed were specifically caused by exposure to artificial moonlight. When comparing Period protein abundances for the different cell classes between cry 01 mutants and wild types, Period levels in E cells exhibited a stronger peak delay (∼8 h; Fig. 4 F and G) than M cells (∼2 h; Fig. 4H). This correlates with the fact that the peak of evening activity is significantly delayed in our behavioral analyses of cry 01 mutants compared to wild types under LM (Fig. 4 A and B). It should be noted that the time of fly evening activity is determined by the integrated action of all E cells including the dorsal E neurons (DN 1 -E) (30) that we have not assessed in our study. The DN 1 -E are closely coupled to s-LN v s in dark conditions and may have an earlier phase than the lateral E neurons (fifth s-LN v and LN d ) even in moonlight conditions. This may explain why the behavioral phase shift in evening activity does not completely match the large phase delay observed in the fifth s-LN v and LN d .
Taken together, these results indicate that the increased delay of the evening activity peak in cry 01 mutants under an LM light regime is the result of a desynchronization of the circadian clock rather than an acute light effect. This suggests that Drosophila dCRY is naturally required to reduce the effects of moonlight on circadian clock oscillations, in particular in the cell clusters harboring the evening oscillator.

L-Cry, but Not dCRY, Is Highly Sensitive to Moonlight
Given the genetic requirement of both L-Cry and dCRY to correctly interpret moonlight under a combined moonlight/sunlight regime, we next wondered if the biochemical light sensitivity of both orthologs was also comparable. For this, we purified both proteins in the presence of their cofactor flavine adenine dinucleotide (FAD) and tested for changes in absorbance after illumination. When light is sensed by dCRY (31) or L-Cry (1), it changes the oxidized FAD to the reduced anionic radical FAD°À form, visible in the proteins' absorbance spectrum (31). Extending our work on L-Cry's biochemical features, we find that Platynereis L-Cry does not only respond to naturalistic FM light (1) but does this even at intensities corresponding to 30% of FM intensity at 4 to 5 m seawater depths (Fig. 5A). In contrast, dCRY completely failed to respond to naturalistic moonlight levels equivalent to-and exceeding-those eliciting responses in Platynereis L-Cry (compare Fig. 5A with Fig. 5 B and C). However, dCRY was activated by naturalistic sunlight, reaching complete FAD reduction within 20 min (Fig. 5B) as observed for L-Cry (1), underscoring the integrity of the purified dCRY protein and the functionality of the assay.
Even though dCRY's sensitivity to dim light might be higher in its cellular context (32), this result clearly points at differences in the molecular mechanisms between dCRY and L-Cry functions. We provide further thoughts on this in Discussion. On the ecophysiological level, this might be connected to the different meanings that moonlight has as an environmental cue for the daily behavior of flies versus swarming worms: whereas fly circadian biology is likely optimized to buffer against the effect of moonlight, Platynereis worms, as shown in Fig. 1, use moonlight to precisely adjust their nocturnal swarming time to a favorable dark time window.

R-opsin1 Detects Moonrise to Optimize the Time of Swarming Onset
The retention of moonlight sensitivity in Platynereis l-cry -/mutants (as evidenced by the different mutant responses under the combined moonlight and sunlight regimes versus no-moonlight regimes; Fig. 2  sensation. We reasoned that the spectral sensitivity of these photoreceptors likely includes the blue-green range, given the relatively high levels of blue-green light in our moonlight measurements (SI Appendix, Fig. S2A). The gene encoding r-Opsin1 is expressed in the adult Platynereis eyes during both early development (33,34) and later stages (35). In a heterologous expression assay established for assessing photoreceptor action spectra (36), Platynereis r-Opsin1 exhibits an irradiance response peak in the blue range (λ max = approximately 470 nm) (37), similar to the peak of its human melanopsin homolog. When we assessed the respective sensitivities of both receptors in side-by-side comparisons, the halfmaximal effective irradiation (EI 50 ) of Platynereis r-Opsin1 (2.3 × 10 10 photons cm À2 s À 1) was ∼100 times lower than that of melanopsin (2.5 × 10 12 photons cm À2 s À1 ; Fig. 6A), indicating a remarkably high sensitivity of Pdu-r-Opsin1.
In the animal, this molecular sensitivity is combined with a high abundance of r-Opsin1: on the transcript level, a cellular profiling analysis revealed that r-opsin1 is one of the topmost expressed genes in Platynereis adult eye photoreceptors, outnumbering a distinct coexpressed opsin-r-opsin3-by nearly three orders of magnitude (37). Moreover, in the course of the metamorphic changes that occur during the days immediately prior to swarming, the outer segments of the eye photoreceptors-where Opsin molecules are concentrated in tightly packed membrane stacks-extend to around twice their length, suggesting an even increased sensitivity (38). All these facts imply that r-Opsin1 acts as a particularly high-sensitive light detector at the time of swarming.
To test whether r-Opsin1 was indeed required to mediate the impact of moonlight on the timing of swarming onset, we capitalized on an existing r-opsin1 À17/-17 loss-of-function allele (37). Under a constant darkness (DD) regime, r-opsin1 -/mutants were indistinguishable from wild type (SI Appendix, Fig. S7A), while they maintained synchronization under constant naturalistic sunlight, by this differing from wild type (SI Appendix, Fig. S7B). They also differ from l-cry -/mutants in that they advance their timing under LL (SI Appendix, Fig. S7B A E F G B C D Fig. 6. Pdu-r-Opsin1 functions as highly light-sensitive photoreceptor to adjust swarming onset to waning moon light. (A) Responses of Pdu-r-Opsin1 (violet) and human Melanopsin (black) to different blue light intensities (480 nm ± 10 nm), as quantified by a cell-based bioluminescent assay, reveal an ∼100-fold higher sensitivity of Pdu-r-Opsin1. (B) Swarming onset of r-opsin1 À/À and r-opsin1 +/+ worms entrained to 16:8 h LD cycles and then subjected to constant moonlight of FM light intensity during the night (LM). (C) None of the tested r-opsin1 À/À mutants initiated swarming onset during the day under this LM light regime (data LM2 to LM8 from B), whereas this number is significantly increased in l-cry À/À mutants (data LM2 to LM8 from Fig. 2D). (D-G) Swarming onset of r-opsin1 À/À and r-opsin1 +/+ worms entrained to 16:8 h LD cycles and then subjected to either moonlight during the second half of the night, mimicking a waning moonlight regime (D and E), or to constant moonlight (F and G), either with FM light intensity (dark green) (B, D, and F) or waning moon light intensity (light green indicates 20% of FM light intensity) (E and G). *P < 0.05; **P < 0.01; ***P < 0.001 for two-way ANOVA followed by Sidak's multiple comparison test. Black and violet lines in B, F, and G indicate linear regression lines of wild-type and r-opsin1 À/À mutants, respectively. The period length was calculated based on the slope of the regression line ± the 95% CI of the slope.
vs. Fig. 2C). As discussed in SI Appendix, SI Text, both the maintained synchronization and clock advancement might be connected to mechanisms underlying photoperiodic adjustments of the PCC clock, which also exist (SI Appendix, Fig.  S3B). When we exposed r-opsin1 -/mutants to the naturalistic FM regime they failed to reset swarming time to earlier hours ( Fig.  6B) but remained restricted to nocturnal hours. This contrasts with l-cry À/À mutants under the same paradigm (Fig. 6C). Our data are consistent with the notion that r-Opsin1 is critical for moonlight detection but not sunlight versus moonlight discrimination. In order to further test for r-Opsin1's requirement for moonlight detection, we subjected homozygous r-opsin1 17/-17 mutants and related wild-type individuals for 5 d to the established waning moon paradigm (Fig. 1E), providing naturalistic moonlight during the second half of the night (Fig. 6D). r-opsin1 À/À animals exhibited a significantly reduced ability to shift their swarming onset to the dark portion of the night compared to wild types (Fig. 6D). This difference became even stronger with naturalistic moonlight at lower intensities (as this would be the case for the natural waning moon) (Fig.  6E). Finally, we wondered if r-opsin1 -/mutants would also exhibit a reduced ability to reset the PCC clock under constant moonlight. Under constant moonlight at naturalistic FM (Fig. 6F) or waning moon (Fig. 6G) light intensities, r-opsin1 -/mutants were indistinguishable from wild type.
Taken together with the results under the more naturalistic mixed moonlight/sunlight regimes, this finding let us conclude that r-Opsin1 specifically enables the worms to detect moonrise to align the PCC clock accordingly. Under constant moonlight, however, the continuous activation of L-Cry appears to be sufficient for PCC clock period shortening. At present, we cannot fully explain the difference in r-opsin1 requirement for LM and MM regimes. Possibly, the LM regime does not provide a sufficiently long moonlight illumination for L-Cry to be fully activated, and/or the ability of L-Cry to advance the clock is restricted to certain phases of the clock. Future work will be required to understand the detailed mechanisms of how naturalistic light impacts on the oscillator under these conditions.
The distinct phenotypes of the l-cry -/versus r-opsin1 -/mutants argue for distinct roles of L-Cry and r-Opsin1 in decoding naturalistic moonlight and adjusting the PCC clock (Fig. 7): we hypothesize that the different subcellular levels of L-Cry during night and day (Fig. 7 A-C) provide key information on the level of irradiance. For the organism, this would harbor information on light valence, as the sources of lightsun versus moon-have significantly different temporal/ecological meanings.
Whereas r-opsin1 has no involvement in the entrainment of the monthly oscillator (1), our data indicate that in case of the PCC clock adjustment, it is relevant for mediating acute dim light information. The combination of L-Cry's and r-Opsin1's properties therefore allow the PCC clock to distinguish between not only sunlight, FM, and new moon but also the progressive phases of the waning moon, which are particularly relevant to set the right swarming hour (Fig. 7).

Discussion
Here we uncover an ∼24-h endogenous oscillator in marine broadcast-spawning worms that exhibits marked, moonlightdependent plasticity in its period length upon extended moonlight illumination, as it occurs during FM periods. The clock is also adjusted by moonrises during the waning moon phase. It is currently unclear if the latter adjustment also relies on a change in period length or by phase resetting. Furthermore, during nights when swarming time is fine-tuned by the light of the moon, the PCC is not reset by the following solar light (sunrise).
The modulation of the worms' ∼24-h oscillator by naturalistic moonlight provides a plausible model for how worms synchronize their nuptial dance, targeting a specific hour during the dark portion of moonlit nights. Restricting swarming behavior to the dark portion of the night might be advantageous to avoid predators that hunt during moonlight. On a mechanistic level, we suggest that this PCC clock shares elements with the conventional core circadian oscillator and reveal two highly sensitive light receptors, r-Opsin1 and L-Cry, that are critical to sense and interpret naturalistic moonlight.
Sensitivity to moonlight is directly relevant for a broad panel of marine broadcast spawners. The challenge of tagging nocturnal light information with the correct valence, however, likely extends beyond this specific ecological context. The classical categorization of organisms into nocturnal versus diurnal species (39,40) typically neglects the aspect of moonlight. Any animal entraining its ∼24-h clock to light will need to correctly interpret the occurrence of nocturnal light. Even though it has been shown that the circadian system of many species is sensitive to light levels as low as moonlight intensity, such as in flies (19,20) and mice (41), chronobiological studies have so far spent relatively little effort in dissecting how animal clocks prevent potential disturbance by moonlight and interpret naturalistic light regimes that combine both sunlight and moonlight.
The data presented here provide possible mechanistic explanations for the ability of the PCC clock to decode a combined sunlight and moonlight regime. A first tier is connected to the specific properties of cryptochrome: whereas under naturalistic moonlight, Platynereis L-Cry protein levels remain elevated, comparable to dark conditions, and are predominantly localized to the nucleus, the onset of sunlight causes a rapid degradation, with residual L-Cry protein found in the cytoplasm. On the biochemical level, L-Cry is highly sensitive to naturalistic moonlight. Moonlight evokes a different state in L-Cry than sunlight (see extensive comparison of sunlight vs. moonlight in ref. 1). Based on these detailed biochemical studies, including multiangle light scattering and size exclusion chromatography analyses, we hypothesize that Pdu-L-Cry functions as a dimer with different light sensitivities of each of the two monomers, potentially arising from different quantum yields for FAD photoreduction (1). This would allow for the differential response to high-intensity sunlight versus very low intensity moonlight, as moonlight can only photoreduce the flavin cofactor in the low-light sensitive L-Cry monomer, whereas sunlight can photoreduce both (1). Dimers as part of the photoresponse mechanism have been well documented for Cryptochromes in plants. These Cryptochromes form dimers or even multimers upon illumination, which are essential for the light response (42)(43)(44). Even though plant Cry2 members are frequently called lowlight receptors, their responses have only been analyzed under light conditions that still mimic sunlight (42,44), which are magnitudes away from moonlight (SI Appendix, Fig. S2, and ref. 45). Thus, the mechanism by which Pdu-L-Cry responds to the very dim moonlight has to be more efficient even compared to the low-light sensors of the plant Cry2 family, while L-Cry also still retains the ability to detect high-intensity sunlight. The dimer model of L-Cry fulfills this requirement (1).
In contrast to L-Cry, its direct Drosophila ortholog dCRY (14) does not form dimers (31). This difference is consistent with the lack of photoreduction of purified dCRY under naturalistic moonlight that we report here. Taken together, our data are consistent with the idea thatbesides the canonical strong-light induced degradation-based signaling pathway for cytoplasmic Cryptochrome-L-Cry, but not dCRY, possesses a second, dim-light-induced, nuclear mode of signaling. Its presence in the nucleus and absence of degradation under moonlight makes it tempting to speculate that L-Cry could function as a transcriptional repressor. Notably, Drosophila Cryptochrome has been proposed to function as transcriptional repressor in the flies' peripheral tissues under darkness (46). However, whereas the worms' vertebrate/animal-Cry2 ortholog tr-Cry (14) showed transcriptional repressor activity in a heterologous S2 cell assay, Pdu-L-Cry did not under the same conditions (5). Future work may determine if (worm) cell type-specific factors could be required for L-Cry to exert transcriptional activity or if it could rather regulate posttranscriptional responses.
A second lead on how moonlight can impact on ∼24-h timing is provided by our identification of r-Opsin1 as a second moonlight sensor. Whereas it remains to be uncovered how the r-Opsin1-dependent signals tie in with the different signaling states of L-Cry, the existence of two distinct sensors already opens up the possibility for a combinatorial setup, in which the (nuclear) presence of L-Cry allows an Opsin-dependent signal to be interpreted as moonlight, whereas the reduction of L-Cry upon sunlight illumination allows an incoming Opsin-dependent signal to be interpreted as sunlight (Fig. 7). Just the combination of the two signals alone would be sufficient for the light distinction. This combinatorial model could also explain the situation of Drosophila, where purified dCRY biochemically does not respond to naturalistic moonlight but is genetically required to shield the Drosophila circadian clock from moonlight impact. Opsin-based light input also plays a role for circadian entrainment, including the detection of light at moonlight intensity (47), and together with the information from dCRY, the natural source of light could be discriminated. For a diurnal/crepuscular species such as D. melanogaster, the main importance of the discrimination between sunlight and moonlight is to reduce the impact of moonlight on its ∼24-h clock and not to tune it to moonlight. Therefore, the moonlight sensitivity of the cryptochrome component itself is a dispensable feature for the fruitfly's dCRY. In contrast, the nocturnal bristle worm finetunes its plastic PCC with moonlight. A direct sensitivity of L-Cry to moonlight is likely beneficial for this moonlightresponding system and hence exists biochemically for L-Cry.
Evidence for plasticity of the conventional circadian clock has started to emerge from other marine systems: work on the circatidal oscillators of oysters maintained under controlled laboratory conditions revealed that core circadian clock genes exhibit ∼12.4-h cycles under constant darkness, whereas the transcripts of the same genes cycle with an ∼24-h oscillation under light/dark conditions (48). This provides evidence for the ability of the canonical clock to alternate between circadian (∼24 h) and (semi)circalunidian (∼12.4 h/∼24.8 h) periodicities. Of note, switches between circadian and circalunidian cycles might also occur in humans. For instance, mood switches of bipolar patients correlate with a period lengthening of their body temperature cycles that looks as if the circadian timing system can be intermittently entrained to a 24.8-h rhythm (49). Moreover, already classical chronobiological studies documented changes of the ∼24-h clock periodicity under dim light in various organisms, including birds, mice, hamsters, and humans (50,51), as well as the fruit fly D. melanogaster (22). Whereas the meaning of these results had remained enigmatic, they could well be explained by the conceptual framework of combined solar and lunar light cues that we present in our study. We anticipate that research on organisms for which lunar impact is of known biological relevance will be key to disentangle the interplay of solar and lunar timing cues.

Materials and Methods
Detailed methods on the following subjects are available in SI Appendix, SI Methods: natural light measurements, behavioral setup and analyses of swarming onset in worms, Western blots, immunohistochemistry, Period oscillations in Drosophila clock neurons, spectral sensitivity comparison of opsins, casein kinase inhibitor treatment and qPCR analyses, recombinant expression and purification of L-Cry and dCRY proteins, and UV/VIS spectroscopy of L-Cry and dCRY.
Worm Culture. Worms were grown as described previously (10). In short, worms were kept in plastic boxes filled with a 1:1 mixture of natural sea water and artificial sea water (30% Tropic Marine) and exposed to a 16:8 h light:dark light regime. To entrain their circalunar clock, worms received eight nights of continuous nocturnal light each month to mimic FM. Strains. l-cry À/À . Homozygous lcry Δ34/Δ34 worms were obtained by crosses between lcry Δ34/+ individuals generated and maintained in the VIO-strain background (see ref. 1). Wild-type worms used in the experiments for comparison to l-cry À/À worms were derived from the respective lcry +/+ relatives obtained in the crosses. r-opsin1 À/À . Homozygous r-opsin Δ17/Δ17 worms were obtained by crosses between r-opsin Δ17/+ individuals generated in the pMosfrops::egfpg vbci2 transgenic strain (37). Wild-type worms used for comparisons were derived from the pMosfrops::egfpg vbci2 transgenic strain in which the mutant was generated.
Recording of Locomotor Activity in D. melanogaster. Locomotor activity was recorded under constant temperature (20°C) from 0-to 1-d-old male Canton-S and cry 01 (Canton-S background) flies using the Drosophila Activity Monitors from Trikinetics Incorporation (28). Flies were recorded first for 5 d under 12 h light-12 h dark cycles (LD with ∼100 lx standard white light LED) and then for 7 d under 12 h light-12 h artificial moonlight cycles (LM cycles; forspectrum and intensity of artificial moonlight, see SI Appendix, Fig. S2C). The average actograms and the centers of maximal activity were calculated and plotted with ActogramJ (52). The phases of evening activity maxima under LM conditions were determined using the ActogramJ tool "acrophase." To test for differences in the acrophase of wild-type and cry 01 flies at LM4, an unpaired Student test was performed.
Imaging and Quantification of L-Cry Staining. Imaging of the worm heads was done on a Zeiss laser scanning confocal microscope (model LSM 700), using Plan-Apochromat 25X and Plan-Apochromat 40X objectives, a transmission photomultiplier tube (T-PMT) detection system and Zeiss ZEN 2012 software. Lasers used were at 405 nm and 555 nm. Image analysis was performed using the software Fiji/ImageJ (53).
Nuclei were segmented using Cellpose (54) on the DAPI channel images. Subsequently, the segmented nuclei were used as regions of interest (ROIs) in Fiji/ ImageJ (53) to quantify the signal intensity by calculating the corrected total cell fluorescence (CTCF) as follows: CTCF = Area ROI1 * Mean ROI1 À Area ROI1 * Mean backgroundROIs . CTCF was determined for the entire brain area as well and used for calculation of the signal intensity of nonnuclei, which was considered cytoplasmic (CTCF cytoplasm = CTCF total À CTCF nuclei ).
Statistical Analyses. We used one-way ANOVA followed by Dunnett's test to test if the timing of swarming onset during LD conditions differs compared to conditions where worms are subjected to moonlight conditions on top of an LD cycle. We used two-way ANOVA followed by Sidak's test to test if and during which days the timing of swarming onset differs between mutant and wild types across different days of a behavioral experiment. To compare if two sets of data had different variances, an F test as part of t test statistics was performed. Swarming onset data are shown as individual data points and additionally represented as box plots with whiskers reaching to the maximal and minimal values.
Western blot data, which were used to assess head L-Cry levels during sunlight, moonlight, and darkness conditions, were analyzed with one-way ANOVA followed by Tukey's multiple comparison test to test for significant differences in L-Cry abundance between the different light conditions.
To compare period oscillation in the different cell groups between cry01 mutants and wild-type flies over different ZTs we used two-way ANOVA followed by Sidak's test.
Data Availability. Light measurement data and data underlying the analyses of swarming times have been deposited in the Dryad Digital Repository (https:// datadryad.org) (DOI: 10.5061/dryad.2v6wwpzkr) (55). All other data are included in the manuscript and supporting information. Genetic animal strains used in the described work will be shared upon request with qualified researchers for their own use.