Evidence of separate influence of moon and sun on light synchronization of mussel’s daily rhythm during the polar night

Summary Marine organisms living at high latitudes are faced with a light climate that undergoes drastic annual changes, especially during the polar night (PN) when the sun remains below the horizon for months. This raises the question of a possible synchronization and entrainment of biological rhythms under the governance of light at very low intensities. We analyzed the rhythms of the mussel Mytilus sp. during PN. We show that (1) mussels expressed a rhythmic behavior during PN; (2) a monthly moonlight rhythm was expressed; (3) a daily rhythm was expressed and influenced by both sunlight and moonlight; and (4) depending on the different times of PN and moon cycle characteristics, we were able to discriminate whether the moon or the sun synchronize the daily rhythm. Our findings fuel the idea that the capability of moonlight to synchronize daily rhythms when sunlight is not sufficient would be a crucial advantage during PN.


INTRODUCTION
Recently, a habitual paradigm in Arctic marine biology stating that life processes are drastically reduced during the polar night (PN) have been challenged. 1 Despite low temperature, limited food availability, and a reduction in illumination, PN dormancy does not appear to be a general feature among arctic marine organisms. Recent reviews have listed many active biological processes during the PN, 2,3 among them being light-mediated biological rhythms expressed by zooplankton [4][5][6] and an active daily rhythmic behavior in arctic scallops. 7,8 Daily rhythms are usually synchronized by a 24-h solar-day cycle. 9 Moreover, moonlight could also be involved in the lunar-month rhythm that is synchronized by a 29.5-day synodic moon cycle related to the moon surface illumination. 10 It has been shown that Arctic zooplankton migrations can be mediated by lunar light during parts of the PN, 11 but no studies have hitherto been able to document the absolute and relative influences of sun-and moonlight in relation to synchronizing biological rhythms during the PN. In the case of polar regions, and especially during the PN, the effect of sun and moon illumination to synchronize the rhythms are highly different from temperate areas where sun and moon cycles at the daily and monthly scales are mostly unchanged throughout the year. During PN, with the sun remaining below the horizon for a full 24-h cycle, light variation during the day still occurs, but in a very low range of intensity. 12 At midday, the quantitative difference between the polar day and the polar night is in the order of 10 8 measured in absolute quanta. 12 Measured at 79 N within a diel cycle, the ratio of illumination between midday and midnight is approximately similar in the polar day and PN and six orders of magnitude lower than during the spring and autumn equinox. 12,13 In polar regions, the moon cycles are also different from those observed at middle latitudes, due to the elliptic orbit of the earth around the sun. During a lunar month, we observe successively lunar-day cycles either always above or always below the horizon with between, as in temperate areas, cycles with the moon below and above the horizon each day. Moreover, the moon illumination is still changing at a 29.5-day cycle. At 79 N the moon light during a full moon, in the darkest part of the PN, is two orders of magnitude higher than the solar illumination (irradiance = $10 À3 mmol m À2 s À1 and $10 À5 mmol m À2 s À1 , respectively) and hence the dominating factor regulating the lighting conditions. 12,13 The blue mussel is a reemerged resident in the high Arctic archipelago of Svalbard after a 1000-year absence in Svalbard. 14  iScience Article synchronization of biological rhythms under the governance of light at very low intensities and subsequent advantages. [17][18][19][20][21][22][23][24] Thus, we have investigated if the light intensity from the sun and the moon during PN were sufficient to synchronize the behavioral rhythms of mussels and then, if it was possible to quantitatively separate the influences of the two.

RESULTS AND DISCUSSION
Synodic lunar-month cycles synchronize valve behavior of mussels during the polar night To determine mussel's rhythmic valve behavior during PN, we monitored valve opening amplitude (VOA) using a high-frequency non-invasive (HFNI) valvometer biosensor 25 at Ny-Å lesund, in Kongsfjorden, Svalbard (78 56 0 N, 11 56 0 E) ( Figure 1A). This biosensor has continuously recorded the behavior of 15 mussels equipped with valvometric electrodes ( Figure 1B) during two consecutive PN (timetable in Table S1). To determine whether a moonlight rhythm was expressed, we analyzed PN 2016-2017 lasting 116 days.  Figure 1D) were determined by Lomb-Scargle periodogram spectral analysis 26 to estimate the period and by the cosinor model 27,28 to determine the characteristic and the strength of the rhythm. A significant period of 29.30 G 0.81 days (mean G ES) was found in the range of the synodic lunar-month cycle ($29.53 days).. In Figure 1E, significant VOA differences were showed according to moon phase during PN (p = 0.012). VOA was minimal during new moon, intermediate during the first quarter of the moon, and maximal during full moon and the third quarter of the moon. A behavioral moonlight rhythm of a mollusk bivalve was already shown in the oyster Crassostrea gigas living in temperate areas. In winter, the daily rhythm of the C. gigas showed a maximal VOA during the new moon, the darkest time of the night. 25,29 In contrast, we show here that during PN the mussels increase their VOA during the full moon, when moon light intensity is maximal. At an annual scale, we observed a higher VOA when the photoperiod and light intensity increased, 30 suggesting a photophilic behavioral pattern. The influence of the lunar cycle during PN was previously reported for the vertical migration of pelagic zooplankton in the high Arctic. 11 Thus, in this study, we show that eye-less benthic organisms such as Mytilus sp. have sensitivity and synchronicity toward the lunar cycle during PN.

Mussels exhibit daily behavioral rhythms during the polar night
To investigate whether light coming from lunar-day or solar-day cycles could synchronize valve behavior during PN, individual rhythmic activity of VOA on a daily scale was determined during two subsequent polar nights (details Table S1). Figure S1 shows the range of the daily rhythms measured during the two PN. These rhythms are calculated according to the different twilights, the moon phases, and the lunar cycle characteristics (31 periods studied in total, details Table S2). We showed that a significant rhythm was always expressed during PN regardless of the tested intervals. The percentage of rhythmic mussels ranges from 25% to 93.8% (see details Table S2). Moreover, the rhythms analyzed here showed periods ranging between 22.1 G 1.1 h and 27.4 G 0.5 h. It suggests a variable and quite weak synchronization of the sun or the moon at a daily scale during PN. That might be explained by the very low light intensity coming from the sun or the moon. To reveal a possible relationship between the light intensity of the sun reflected by the skyglow and the percentage of rhythmic mussels, we quantified percentages according to the different twilight intervals [31][32][33] (Figure 2A). We show that the difference of light intensity in each twilight cannot explain the differences of percentage of rhythmic mussels alone. The results expressed as quartiles showed that around 73%-75% of mussels behaved rhythmic during civil and nautical twilights and $ 60% during the astronomical twilight (exact values in Table S3A). No significant difference (p = 0.372) was found between the three twilight conditions. These results suggested that a light cue coming from the sun is not able solely to entrain behavioral daily rhythms in mussels during PN. Then, we analyzed the percentages of rhythmic mussels ( Figure 2B) and moon surface illumination ( Figure 2C) according to the different lunar-day cycle characteristics. The results expressed a significant difference (p = 0.002) for rhythmic mussels. Indeed, during a lunar-day cycle with the moon above and below the horizon, the percentage of rhythmic mussels  Figure S1 and Table S1).
Results are shown as quartiles in green (25% and 75% quartiles are defined by the box edge, 50 % median value by the line inside the box). In red solid line, the mean.
(A) Percentage of rhythmic mussels according to the different twilights of PNs: civil twilight (maximum sun height : 0 to À6 , n = 12 tested periods); nautical twilight (maximum sun height : À6 to À12 , n = 15 tested periods); astronomical twilight (maximum sun height : À12 to À18 , n = 4 tested periods iScience Article (62.8% G 4.1%) was significantly lower compared with a lunar day with the moon always above the horizon (77.9% G 2.6%) or always below the horizon (83.6% G 3.8%). The mean illumination of the moon showed significant variation (p < 0.0001) between different moon cycles. The maximum percentage of illumination corresponds to different lunar day cycles. The mean illumination was 83.5% G 5.2% with the moon always above the horizon, 54.7% G 5.6% with the moon above and below the horizon, and 23.4% G 6.0% with the moon always below the horizon. Considering both results ( Figures 2B and 2C), we show that daily rhythm synchronization cannot be explained solely by lunar illumination or lunar cycle characteristics. Indeed, the percentage of rhythmic mussels was the same for lunar-day cycles, both always below and above the horizon, despite the variation of lunar disc illumination (exact values in Table S3B). Consequently, we conclude that the light cue coming from the moon is not sufficient by itself to explain behavioral rhythms at the daily scale. In conclusion, we show that daily rhythms of mussels are largely expressed during the polar night, but their synchronization cannot be explained separately by sunlight or moonlight cues.
Both lunar day and solar day synchronize mussel's daily rhythm during the polar night By classic spectral analysis or rhythmic models, it is not possible to statistically discriminate the contribution of either sun cycle (24 h) or lunar cycle (24.8 h) to the mussel's synchronization. Thus, to address the issue of the origin of the daily mussel rhythm during PN, we analyzed the evolution of the significant periods found in accordance to the interaction of light sources coming from the sun and the moon. First, to get an overview, in the Figure 3A, we plotted measured solar irradiance, expressed in energy of photosynthetic active radiation (E PAR ), recorded in Ny-Alesund during PN 34,35 according to its angle below the horizon. The maximal moonlight irradiance during full moon was measured at $1.5.10 À3 mmol m À2 s À1 , which also corresponded to a sunlight illuminance at an altitude of $À8 below the horizon. 12, 32 We were able to determine three different conditions that could be the zeitgeber of the daily rhythms. First, from 0 to À8 of sun elevation, only the sun could have a light influence. Then, in the range of À8 and À12 of sun elevation, sunlight and moonlight could have a mixed influence. And finally, below À12 of sun elevation (corresponding to the beginning of astronomical twilight), only the moon could have a light influence. To test the hypothesis that the evolution of the daily rhythms could reveal the different source of light synchronization, we plotted the significant periods found in the daily range during the two PN studied (Table S2) according to the sun elevation and the moon cycle positions ( Figures 3B-3D). In the condition where the lunar-day cycle is always above the horizon ( Figure 3B) and with a maximal moon surface illumination ( Figure 2C), we show that the daily periods found were near 24 h in civil twilight, revealing a clear solely sun cycle influence and then moved toward periods close to a lunidian day at 24.8 h with the decrease of sun altitude to À12 . In the condition with the moon above and below the horizon during the lunar-day cycle ( Figure 3C) and a moon surface with a mid-illumination ( Figure 2C), we observe a higher variability of periods found in the daily range due to the mixed influence of the sun and the moon. But this variability decreased with the decrease of sun elevation to reach, at the end of the nautical twilight and the beginning of the astronomical one, daily periods closer to the lunidian cycle at 24.8 h. Finally, in the condition with the moon always below the horizon during the lunar-day cycle ( Figure 3D) and a moon surface with a low illumination (Figure 2C), we clearly show that the periods found in the civil twilight up to the middle of the nautical twilight were close to 24 h under the influence of the sun. Then, at decreased sun elevation and lower lunar illumination, the periods found seem less synchronized by the sun or the moon, although a shift seem to occur to a period closer to 24.8 h rather than to 24 h.
The strength of the moon entrainment to synchronize daily rhythm during the polar night depends on lunar-day cycle characterization, moon illumination, and the sun elevation To clarify and to order the respective influences of the moonlight and the sunlight in PN as the source entraining daily rhythms, we quantified the difference (d) between the periods measured, with either the exact daily period of the sun cycle (24 h) or the daily period of the moon cycle (24.8 h). This parameter was used to evaluate the strength of entrainment of the moon and the sun, i.e. the lower the d, the stronger the moon or sun influence. Figure 4 shows d between a 24-h sun cycle (d 24h ) or 24.8-h moon cycle (d 24.8h ) and the actual periods found according to the moon cycle characteristics and the sun elevation. We can show a significant  iScience Article correlation between d 24.8h and sun elevation, for conditions when the moon cycle was always above the horizon (p = 0.009, r 2 = 0.78) and moon cycle above and below the horizon (p < 0.001, r 2 = 0.66) (Figures 4A-4B). In these lunar conditions, the lower the sun elevation, the more d 24.8h became closer to zero. On contrary, under these lunar conditions, we did not find a significant correlation between d 24h and sun elevation ( Figures 4D-4E). These results highlight that moonlight was the main driver of the daily rhythm under these conditions. Then, we showed ( Figure 4F) a significant correlation between d 24h and sun elevation, when the moon cycle was always below the horizon (p = 0.007, r 2 = 0.73). In this condition, the higher the sun elevation, the more d 24h approached zero. On contrary, we did not observe a significant correlation between d 24.8h and sun elevation ( Figure 4C), showing an absence of moon influence to synchronize the daily rhythm.
During the polar night, moonlight and sunlight share the timing of daily rhythm We have focused on the mollusk bivalve Mytilus sp. to investigate the role of the moonlight during PN in the synchronization of daily behavior. We have shown that moonlight, besides its role as a driver of monthly rhythm, also shares the role of timing daily behavior together with the sunlight. Usually, it is assumed that the daily rhythm is the consequence of the circadian clock synchronization by sunlight. However, during PN conditions, moonlight intensity is comparable to sunlight and could set the mussel's daily timing system.
Our finding does not allow to discriminate if the observed daily activity is due to a direct reaction to moonlight/sunlight perception or is due to endogenous clockwork mechanism entrainment. However, a previous study done on the arctic scallop Chlamys islandica in Svalbard 8 showed the first evidence of the persistence of clock gene expression oscillations during PN, suggesting that functional clockwork could entrain rhythmic behaviors. Moreover, earlier studies 36,37 on bivalves such as the oyster C. gigas have highlighted the plasticity of the so-called circadian clock that was able to run at tidal periodicity ($12.4 h) under field conditions and under constant darkness in controlled lab conditions, whereas this clock ran at $24 h under light/dark conditions. Although our finding strongly suggests a photophilic behavior of the mussels in PN, it does not prove it stricto sensu. Indeed, there may exist an indirect effect through the interaction with the food sources or predators that would be under the synchronization of the lunar cycles. However, evidence that the circadian clock could be sensitive to low light intensity in the range of moonlight have already been shown in different terrestrial animals 38,39 and plants. 40 This plasticity of the circadian clock has also been suggested for other marine phyla. Molecular work on the bristleworm Platynereis dumerilii revealed that moonlight is perfectly able to synchronize swarming daily rhythm thanks to the interplay of two light sensors, a melanopsin ortholog (R-opsin1) and a cryptochrome (L-cry), known to be involved in the circadian clockwork. 41 Finally, if we accept the idea that maintaining daily rhythm during PN is crucial for animals to temporally prioritize internal physiological processes, the ability to synchronize the clock system by moonlight when sunlight is not sufficient would be a key advantage for animals in polar regions.

Limitations of the study
Although we measured an apparent light-mediated rhythm, we are not able to disentangle a direct response to light of the mussels from a response controlled by an endogenous underpinning clockwork mechanism. Moreover, in this field experiment, we did not strictly prove the photophilic behavior of the mussels. Thus, only the expression of clock genes in the same conditions would allow to give an answer about the photophilic behavior and the involvement of the clock in the generation of the behavioral rhythm in such low light intensity and also the implication of the circadian clock on the light perceived from the moon. Thus, a molecular study remains to be achieved to complete the answer of this issue.  34,35 Vertical solid green line corresponds to the full moon irradiance ($ 1.5⸱10 À3 mmol⸱m À2 ⸱s À1 ). Horizontal orange and dark orange dashed lines correspond to the sun elevation of À8 and À12 , respectively, which separated the sun and moon lights respective predominate influences.
(B-D) Significant daily rhythms of VOA (mean G ES, n = 15 mussels) during PN according to the maximal sun angle below the horizon and the moon cycle positions, determined by Lomb-Scargle periodogram and validated by cosinor analyses.
(B) Lunar-day cycles with the moon always above the horizon (n = 7 tested periods); (C) lunar-day cycles with the moon above and below the horizon (n = 16 tested periods); (D) lunar-day cycles with the moon always below the horizon (n = 8 tested periods). For the details of the tested intervals, see Table S2. 24 Figure 4. Entrainment strengths of the solar-day and lunar-day cycles on the behavioral rhythm accordingly to the interaction between the sun elevation and the moon cycle characteristics (A-F) Correlation between the sun angle position below the horizon and the difference in time duration (deltaä) between lunar day (d 24.8h , (A-C)) or solar day (d 24h , (D-F)) and the actual periods found. Each correlation tested were done according to the moon cycle positions. (A, D) Lunar-day cycles with the moon always above the horizon (n = 7 tested periods); (B, E) lunar-day cycles with the moon cycles above and below the horizon (n = 16 tested periods), (C, F) lunar-day cycles with the moon always below the horizon (n = 8 tested periods). Lines solid black for significant and dotted gray for nonsignificant regressions. For the details of the tested intervals, see Table S2. In bolt, the significant p value of the linear regression and its corresponding r 2 . Solar irradiance was expressed in energy of photosynthetic active radiation (E PAR , mmol⸱m À2 ⸱s À1 ). We plotted E PAR data measured at midday (12h UTC). To avoid light coming from the moon, for data with sun angle below À8 (when moon light becomes influent), only data with moon below the horizon and/ or with the percentage of lunar disc illumination <10% were considered.

Recording of mussel's valve activity behavior in the field study
Valve behavior was recorded in situ using a high-frequency non-invasive (HFNI) valvometer. 25 Briefly, a pair of lightweight electrodes, designed to minimize disturbance to mussel's behavior, were glued on each shell. These valvometer electrodes were connected to the HFNI valvometer by flexible wires, which allowed the mussels to move their valves without constraints. The measurement is magnetic principle based. The electrodes are made with small self-inductance solenoids (material: ferrite; size: 3.2 mm 3 2.5 mm x 2 mm; weight: $0.06 g), whose specifications are: inductance: 470 mH; rated current: 45 mA; self-resonance frequency: 5 mHz. Thanks to these electrodes specificity, a very low electromagnetic current (1-2 nT) was generated between the electrodes by the biosensor, which allowed measurements of the amount of valve opening. The signal was recorded at 10 Hz using custom acquisition cards (Nanog manufacturer, Pessac, France), and the data were automatically transmitted daily to a data processing center at the Arcachon Marine Biological Station (France) using internet network.

Valve behavior quantification
Field valve activity data were analyzed using LabView 8.0 software (National Instruments). The valve behavior endpoints were expressed as the hourly valve opening amplitude (VOA, %) of each individual (54,000 data acquisition/day/mussel). Mean hourly VOA was reported as a percentage, with 100% indicating that the valves were opened at their maximum amplitude and 0% indicating that the valves were closed, during the entire time studied, respectively.

Chronobiological analysis Determination of tested periods for chronobiological analysis
To determine the existence of individual mussel's lunar month rhythm based on the synodic lunar cycles lasting 29.53 days, chronobiological analysis were applied on the 2017-2018 PN-period, comprising 4 entire synodic lunar cycles. Such chronobiological analysis couldn't be done in the first PN studied (2016-2017) at the monthly scale due to the absence of data during 15 days during this time due to electrical failures. Lunar month periodicity of behavioral rhythm was defined significant for a period of 29.53days G5 days. To calculate VOA according to the moon phases (full moon, first quarter of the moon, new moon and third quarter of the moon), we measured for each mussel the mean of VOA during 3 days around the exact date of each moon phase, and so for each of the 4 lunar-month cycles comprised in PN.
Then, we determined daily rhythms. In this study, we used the term daily for all periods found in the range 18-30h, entrained by a 24h sun daily cycle (solar day) or a 24.8h lunar daily cycle (lunidian day). The extent of the classic circadian range (20-28h) previously defined for terrestrial animals was set to consider possible looser periodicity in unsynchronized conditions or weakly synchronized for aquatic organisms in PN.
To determine rhythm at the daily scale, the two PN were analyzed. These two PN were divided in several intervals according to the lunar day cycles (see details Tables S1 and S2).

Methods to quantify significant rhythms
Chronobiological analyses were performed using TSA Serial Cosinor 8.0 software. Several steps were required to validate a significant rhythm. 42,43 Four steps must be validated. First, the quality of the dataset was assessed by controlling for the absence of randomness using the autocorrelation diagram. Second, the absence of a stationary phenomenon was checked by using a partial autocorrelation function (PACF) calculation. 44 Third, the recorded data were tested for periodicities by the spectral method of the Lomb and ll OPEN ACCESS iScience 26, 106168, March 17, 2023 iScience Article Scargle periodogram, which combines the principle of a regression analysis and Fourier transformations. 26 This method gives a threshold of probability (p = 0.95) defining the limit below which the signal can be regarded as ''noise''. Fourth, the rhythmicity was validated and modeled with the cosinor model, which uses a cosine function calculated by regression. 27,28 For a given period, the model is written as Y (t) = Acos (pt/t + 4) +M + ε (t) where Y (t) is an observation of the mean VOA at time t, A is the amplitude, 4 is the acrophase, t is the period, M is the mesor and ε is the relative error. Two key tests validated the calculated model and the existence of a rhythm: the elliptic test had to be rejected, and the probability for the null amplitude hypothesis had to be <0.05. For a set of data, several significant periodicities could occur. To identify significant secondary periodicities, we reinjected the previously calculated residues of the Cosinor model to remove the trend related to the first statistical period and then repeated the entire procedure (1-4 steps). This entire procedure was necessary to validate secondary periodicities. In this study, the procedure was repeated up to four times to reveal significant rhythmicity.

Statistical analysis
Differences between conditions were investigated using one-way ANOVA for repeated measures, after checking assumptions (normality of data and equal variance tests). When assumptions were not validated, non-parametric tests were performed. A Kruskal-Wallis One-Way ANOVA on rank for repeated measures was applied to compare distributions, followed by Dunn's Method for all pairwise multiple comparisons. For all statistical results, a probability of p < 0.05 was considered significant. Analyses were performed with SigmaPlot (Version 13, Systat, Chicago, USA).