Artificial Light at Night Affects Microbiota and Growth in the Oyster Crassostrea gigas: Correlations with the Daily Rhythm Robustness

Abstract

Widespread in coastal environments, artificial light at night (ALAN) is suspected to disrupt organisms’ biological rhythms by altering natural light cycles and thus constitutes a growing threat to these ecosystems. This study evaluates the effects of ALAN exposure at low and realistic intensity (~1 lx) on a coastal keystone species, the oyster Crassostrea gigas. The results reveal that ALAN significantly impairs the expression of core circadian clock genes (CgClock and CgBmal1) as well as the valve opening behavior, affecting rhythmic characteristics such as its robustness and daily profile. At the same time, ALAN leads to a decrease in daily shell growth and to a disruption of the gill microbiota, associated with an obliterated day/night difference in microbial alpha diversity. A direct correlation between a decrease in daily rhythm robustness, limitation of shell growth, and some microbial strands is shown, suggesting that biological rhythm disruption caused by ALAN might have harmful physiological consequences in oysters.

1. Introduction

The alternation between daylight and the darkness of the night is of great importance in all ecosystems, since it is used by the endogenous molecular clock to synchronize and coordinate biological processes with the environment [1]. This synchronization allows physiological processes and behavior to be organized according to environmental periodic changes, conferring adaptive advantages to organisms [2,3]. artificial light at night (ALAN) disfigures the nocturnal landscape across almost the whole world and can alter the perception of natural light/dark cycles by organisms [4,5]. Thus, ALAN might disrupt the synchronization of the circadian clock with the light environment and alter the temporal organization of a large diversity of biological processes, from the individual to the community scale (sleep, foraging, growth, reproduction, microbiota, orientation, communication, prey/predator interactions…) [6,7,8,9,10]. The study of ALAN’s ecological and physiological effects has greatly expanded over the last decades and should be one of the main focuses of global change research in the 21st century, particularly given its rapid expansion (9.6% increase in sky brightness per year) [11,12]. This is especially true for coastal environments, in which the human population is predicted to grow, suggesting a faster increase in the ALAN that is already present in 1.6 million km² of coastal seas at a depth of 10 m [13,14]. However, despite the growing threat that ALAN poses to coastal ecosystems, studies of its impacts remain scarce.

As a keystone species of coastal environments likely exposed to ALAN, the oyster Crassostrea gigas is a relevant biological model to investigate its potential effects on benthic ecosystems. This oyster species is a sessile mollusk endemic to East Asia, introduced and cultivated in many places of the world, especially in coastal areas exposed to ALAN [4,15,16]. Although the oyster C. gigas is an eyeless organism, it can perceive natural or artificial light through photoreceptors nested in each cell, such as cryptochromes. These cryptochromes are directly involved in the oyster’s internal clock, whose putative functioning has been described [17,18]. Up to now, no central circadian pacemaker has been found in oysters. The clock mechanism is based on a negative feedback loop composed of genes coding for transcriptional factors Clock (CgClock) and Bmal1 (CgBmal1), which form, as proteins, a heterodimer activating the transcription of the cryptochrome 2 (CgCry2), period (CgPer), and timeless (CgTim) genes [18]. Then, the corresponding proteins form a trimer that represses the CLOCK/BMAL1 activity, and thus their own expression. This molecular clock is synchronized with light cycles thanks to the cryptochrome CRY1: when this photoreceptor captures two photons, it induces TIM degradation, preventing the formation of PER/TIM/CRY2 trimer and thus the inhibition of CLOCK/BMAL1 activity [18]. Therefore, ALAN can alter the functioning of the oyster molecular clock and its synchronization with environmental light cycles. Indeed, previous studies have shown that the oyster’s behavioral daily rhythm is disrupted by ALAN starting from a low intensity of 0.1 lx, as well as the expression of its circadian clock, light perception, and clock-associated genes, which have been largely studied [18,19,20,21,22,23]. These effects have also been shown to be wavelength-dependent and to worsen when applying the part-night lighting mitigation strategy in realistic conditions (part-night lighting combined with skyglow throughout nighttime) [23,24,25].

To go further into the understanding of ALAN’s effects on oysters, there remains at least one issue that is important to study, which is ALAN’s effects on oysters’ physiological functions. In particular, determining whether the circadian clockwork disruption by ALAN is involved in physiological impairment remains to be investigated. Among the multitude of physiological functions that ALAN could impair by acting through temporal disorganization, we chose to study two major physiological traits. The first one is the oysters’ shell growth, since bivalve shell increment has been shown to be driven by daily light–dark cycles or the annual photoperiod [2,20,26,27], and thus could be impacted by ALAN through its action on the clockwork. The second physiological trait studied is the microbiota (i.e., all microorganisms living in a given tissue), since it is known to be linked with its host’s daily rhythm [28,29,30,31] and to fulfill crucial functions (nutrition, immunity, development…) in organisms [30,31]. Here, in oysters, we chose to study the microbiota of their gills, an organ involved in respiration and nutrition, two functions directly dependent on the valve opening behavior and daily rhythm, already well-studied in C. gigas [20,32,33].

Thus, in this study, our working hypothesis is that ALAN weakens the functioning of the circadian clock mechanism of C. gigas and thus impairs the physiological outputs under its control, reducing the animal’s fitness. To test this hypothesis, we exposed oysters to a low ALAN intensity (~1 lx) and measured (1) ALAN’s impact on circadian clockwork and behavioral daily rhythm as a rhythmic output, (2) its consequences on two major physiological traits, i.e., the growth and the microbiota, and (3) the correlation between the daily rhythm disruption and physiological impairments.

2. Materials and Methods

2.1. Experimental Protocol

This experiment was conducted from February to March 2022 in the Arcachon Marine Station on 64 C. gigas oysters (86.9 ± 1.9 mm shell length and 40.0 ± 0.7 mm shell width; mean ± SE) from an oyster farm at the Arguin site (France), where oysters were not exposed to direct ALAN. The experiment occurred in an air-conditioned room where oysters were divided into two conditions (control and ALAN conditions), corresponding to two groups of 32 oysters, with each group placed in one tank (L × W × H: 74.8 × 54.8 × 40.8 cm). The oysters were randomly assigned to each group, between which their length and width were not significantly different (p = 0.836 and p = 0.234, respectively). To limit any vibration that may disrupt oysters’ behavior, the tanks were placed on an antivibrating bench. To avoid any light contamination, the tanks were also surrounded by black tarpaulins. The experimental units were continuously supplied with seawater from the Arcachon bay that went through the same treatment: it was filtered (<5 µm), oxygenized, and maintained at a stable temperature (16.1 ± 0.1 °C), pH (8.0 ± 0.1), and salinity (32.2 ± 0.1 ‰). In addition, the oysters were continuously fed during the whole experiment with a Chaetoceros calcitrans microalgae culture using peristaltic pumps. The final continuous flow rate (seawater and microalgae culture) was 100 mL/min for each tank. The microalgae concentration was measured each day in each tank and was 784 ± 54 and 846 ± 54 cell/mL in the control and ALAN conditions, respectively, throughout the experiment (mean ± SE), without a significant difference (p = 0.304).

Before the exposure to ALAN, both groups of 32 oysters were acclimated for 7 days to a light:dark 10:14 cycle mimicking the winter photoperiod, using programmable white LED light bars (MH3SP3 DSunY, 397–715 nm, peak at 553 nm; Supplementary Figure S1). The daytime was from 7:00 to 17:00 h (all times noted in local time, UTC+1). The daytime intensity varied gradually before and after the maximal illuminance intensity of 1423.89 ± 219.88 lx (corresponding to an irradiance of 28.99 µE/m²/s), which was reached between 11:30 h and 12:30 h. During the acclimation period, the behavior of the oysters’ valves was continuously monitored (Valvometry HFNI technique described below). To check that the oysters in both conditions were in the same physiological state before exposure to ALAN, the behavioral data on the last four days of the acclimation period was recorded. No significant differences in valve opening duration (VOD) (p = 0.925), strength (p = 0.084), amplitude (p = 0.229), and acrophase (p = 0.859) of the rhythm (see definitions of strength, amplitude, and acrophase below in Section 2.3) between the two conditions were observed, highlighting that the initial conditions were similar between control and ALAN groups before the exposure began (Supplementary Figure S2). Then, the oysters were exposed to ALAN for 29 days. In the control condition, oysters remained exposed to a light:dark 10:14 cycle. The light intensity measured during nighttime corresponded to an intensity below the detection limit of the spectroradiometer (0.05 lx). In the ALAN condition, during nighttime, oysters were exposed to ALAN from 17:00 h to 7:00 h at an illuminance intensity of 1.27 ± 0.06 lx, corresponding to an irradiance intensity of 0.04 µE/m²/s using LED strips (MiBoxer Mi-Light WL5, Supplementary Figure S1). ALAN exposure was performed using a white light spectrum (418–684 nm, peak at 551 nm). The choice of ALAN intensity of ~1 lx was motivated by previous studies in which this low and realistic intensity induced significant effects in oysters [23,25].

All underwater illuminances and irradiances were measured underwater at the oysters’ level using a handheld spectroradiometer (Blue-Wave UVN-100, StellarNet Inc., Tampa, FL, USA) and a handheld MPE-PAR (MICRO Class, PAR sensor, Biospherical Instrument Inc., San Diego, CA, USA).

2.2. Oysters’ Valve Behavior

Oysters’ valve behavior was continuously measured on 16 oysters per condition, for the first 28 days of the experiment (lasting 29 days), using High-Frequency Non-Invasive (HFNI) valvometry (EPOC Lab-made biosensor, Arcachon, France) [19,34]. For this, two lightweight electromagnets are glued in front of each other on each oyster’s valve. An electromagnetic current is generated between the electromagnets, which are linked to a valvometer device by flexible wires. For each oyster, these biosensors allow for the continuous measurement of the oyster’s valve activity, with the signal recorded every 4.8 s. Then, data are processed using Labview 8.0 (National Instrument, Austin, TX, USA). We focused on the oysters’ hourly valve opening duration (VOD), which is defined as the percentage of time that an oyster spends with its valve open each hour. When the oyster valves are closed for one hour, the VOD is 0%. On the contrary, when its valves are opened all the hour, the VOD is 100%. Individual hourly VOD data are available in Supplementary Data S1.

2.3. Chronobiological Analyses of Oysters’ Valve Behavior

VOD data were used for individual chronobiological analysis using the software Time Series Analysis Serie Cosinor 8.0 (Expert Soft Technologies, Esvres, France). First, the data quality was checked, controlling the absence of random repartition of data using the autocorrelation diagram, and determining the absence of stationary character using the Partial Autocorrelation Function (PACF) calculation [35]. Then, significant periodicities (p > 0.95) were searched in the dataset using the Lomb and Scargle periodogram, and the data rhythmicity was modeled using the Cosinor model. This model uses a cosine function calculated by regression [36,37]. For a given period, the Cosinor model is written as Y(t) = Acos(2πt/τ + ϕ) + M + ε(t), with A being the amplitude (the difference between the highest value of the rhythm and its average level), τ the given period (the interval between two identical events), ϕ the acrophase (peak value), M the mesor (the average level of the rhythm), and ε(t) the relative error. To validate the model and the existence of rhythmicity, two tests are used: the ellipse test, which must be rejected, and the probability for the null amplitude hypothesis, which has to be lower than 0.05. This model gives another chronobiological parameter, the Percent Rhythm (PR), representing the percentage of cyclic behavior explained by the model, used as a proxy of the strength (or robustness) of the rhythm.

2.4. Oysters’ Shell Growth

The HFNI valvometry biosensor allows the measurement of the daily shell growth of each oyster [20,38]. Shell growth in bivalves occurs by calcification over the shell’s internal surface in the extrapallial space, increasing the minimal distance between the two electromagnets when the oyster’s valves are closed. Therefore, for each individual, the daily shell growth was calculated first by isolating the minimum distance between electromagnets each day and by using the shell length measures made at the beginning and the end of the experiment. Finally, to calculate the daily shell growth, the following formula was used: GR = ([(e_dx − e_d1)/(e_d28 − e_d1)] × [L_d28 − L_d0])/L_d0, with GR being the growth, e the minimum distance between electromagnets, d the days, and L the shell growth length measured with a caliper. The shell growth measurements were conducted only on individuals for which the HFNI valvometry baseline signal was not noisy, i.e., on 15 oysters in the control condition and 13 oysters in the ALAN condition.

2.5. Oysters’ Gene Expression by Real Time qPCR

On the 29th day of experiment, in each condition, the gills of eight oysters were sampled during nighttime at 24:00 h, and stored in RNAlater (Invitrogen) at −80 °C for further molecular analyses. From these gill samples, total RNA was extracted using an SV Total RNA Isolation System kit (Promega, Madison, WI, USA). The total RNA quantity and quality were assessed by spectrophotometry (Synergy HTX model, BioTek, Shoreline, WA, USA) by measuring OD230, OD260, and OD280, and calculating the OD260/OD280 and OD260/OD230 ratios. Then, RNA reverse transcriptions were performed using GoscriptTM Reverse Transcription System kits (Promega, Madison, WI, USA). Real-time qPCR was realized using PowerUp™ SYBR™ Green Master Mix kit (Fisher Scientific, Illkirch, France). The primer sets of the two core clock genes CgClock and CgBmal1 and the three housekeeping genes (CgEf1, Cg28S, CgGadph) are listed in Supplementary Table S1. The determination of the relative transcript level of clock genes was performed by using the comparative Ct method 2^−ΔCt, where ΔCt = Ct (target gene) − Ct (housekeeping gene) [39]. After checking stability values, we choose to normalize the gene’s expression with the housekeeping gene CgEf1 since its expression was the most stable, even considering the geometric means of several housekeeping genes [40].

2.6. Oysters’ Gills and Water Bacterial Diversity Measured by Illumina MiSeq

On the 29th day of the experiment, in each condition, eight oysters were used to record valve behavior were randomly sampled for gill tissues during daytime at 12:00 h and during nighttime at 22:00 h. At the same sampling times, seawater samples were collected in triplicate in each tank using 1 L sterile jars and were filtered immediately after sampling on 0.22 μm sterile membranes using a sterile filtration unit linked to a manual pump. The 32 gill samples and the 12 seawater filter samples were frozen and kept at −80 °C until DNA extraction. Total DNA was extracted from each gill sample (around 250 mg) using the QIAmp^® PowerFecal^® Pro DNA Kit (Qiagen, Les Ulis, France) and from the water samples using the DNeasy^® PowerWater^® Kit (Qiagen, France) following the manufacturer’s instructions. DNA from each sample (gills and water), as well as negative controls (two extraction kit blanks and two PCR blanks) and a positive control (Mock community), were used for prokaryotic diversity amplification based on the 16S rRNA gene. The sequencing was performed by the Bordeaux Transcriptome Genome Platform (Cestas, France) on a 450 bp fragment of the 16S rRNA gene (V3–V4 variable region), which is frequently used to analyze the microbial diversity on the Illumina MiSeq platform, using 2 × 250 bp chemistry [41,42].

2.7. Bioinformatics Data Processing of Illumina MiSeq Results

The first step involves the removal of the primers from the paired-end reads of prokaryotic 16S rRNA with Cutadapt version 2.10 [43]. Then, a DADA2 version 1.22 pipeline was used to process all individual fastq files forward and reverse, where microbial communities are described using Amplicon Sequence Variants (ASVs) [44]. First, the filterAndtrim function was used to filter and trim reads based on sequence quality profiles: Q-scores had to remain above 30 (truncLen = c(230, 220), maxEE = c(2,2), maxN = 0, and truncQ at 2). Then, an error model was calculated for forward and reverse reads using the learnErrors function. Filtered reads were dereplicated using default parameters and merged with a minimum overlap of 20 nucleotides, which does not allow mismatches. The obtained amplicons were filtered by size (390–435 bp), and the removeBimeraDenovo function was used to delete the chimeras. The IdTaxa function was used to taxonomically assign ASVs using the Genome Taxonomy Database SilvaSSURef v138.1 [45]. Then, filtrations were performed on BLASTn taxonomic affiliation, for which only affiliations with a minimum coverage of 99% and a minimum identity of 97% between the sample sequence and the reference sequence were retained.

2.8. Microbiota Analysis

For oysters’ gills, microbiota analyses were conducted on 15 (seven daytime and eight nighttime) and fourteen (seven daytime and seven nighttime) individual samples for the control and ALAN conditions, respectively, while for water samples analyses were conducted on five (two daytime and three nighttime) and six (three daytime and three nighttime) individual samples for the control and ALAN conditions, respectively. The differences in taxon relative abundances associated with each treatment were studied using a model based on negative binomial distribution with the DESeq function in the DESeq2 R package v 1.34.0, in which the log2 fold change is the effect size estimate: log2 fold change > 0 indicates how much the relative abundance of a given ASV is higher in the ALAN condition compared to the control condition [46]. Venn diagrams, displaying the number of ASVs shared between the conditions, were generated using the ggvenn v0.1.10 R package, where ASVs’ relative abundance was compared with only ASVs with a relative abundance superior to 0.1% of total reads. The alpha diversity was computed using the Phyloseq v1.38.0 and Vegan v2.6-4 R packages [47,48]. Based on that, alpha diversity indexes (number of ASVs, Shannon index, Chao1 index, and Inverse Simpson index) were calculated [49]. Beta diversity analyses were performed on Weigh Unifrac distances on a rarefied dataset and then visualized using non-metric multi-dimensional scaling (NMDS) [50].

2.9. Statistical Analyses

The statistical analysis conducted on oysters’ valve activity, chronobiological parameters, gene expression level, daily shell growth, and variability indexes were performed using the SigmaPlot software (version 13.0; Systat Software, Palo Alto, CA, USA). t-tests were performed for the two groups’ comparisons after checking assumptions (normality of data and equal variance). If these assumptions were not validated, the non-parametric Mann–Whitney rank sum test was conducted. The one-way ANOVA teat was used for multiple comparisons after checking assumptions, and when they were not validated, the non-parametric Kruskal–Wallis one-way ANOVA on ranks test was performed. In the case of significant differences, a Student–Newman–Keuls test was performed for all pairwise multiple comparisons. Concerning tests based on data from repeated measures on oysters, i.e., behavioral, chronobiological, and growth analysis, paired t-tests or one-way repeated measures ANOVA were applied. Linear regressions were applied to chronobiological parameters to test the time effect. The Individual variability (Iv) of hourly VOD and daily shell growth was measured by using the formula Iv = σ/m, with σ being the standard deviation and m the mean of individual data for each hour or day. Statistics and the data visualizations of microbiota analysis were performed in R Studio (R version 4.1.0). In the alpha diversity indexes (number of ASVs, Shannon index, Chao1 index, and Inverse Simpson index), differences between conditions were tested using Kruskal–Wallis one-way ANOVA on ranks, followed by pairwise Wilcoxon tests in the case of significant differences. For beta diversity analyses, groups of samples were compared by a permutational multivariable analysis of variance (PERMANOVA, 999 permutations) using the Adonis function of the Vegan R package [51], after checking the homogeneity of dispersion, followed by pairwise Adonis tests in case of significant difference. For 24 oysters, we obtained all at once the valve behavioral rhythm, the shell growth, and the microbiota composition (n = 13 (six daytime and seven nighttime) and n = 11 (five daytime and six nighttime) individuals for control and ALAN conditions, respectively). Spearman correlation coefficients were used to establish whether significant correlations exist between four quantitative variables: ALAN exposure (i.e., 0.05 lx for the control condition and 1.27 lx for the ALAN condition), the strength of the behavioral rhythm (Percent Rhythm, %), the cumulated daily shell growth (%), and the relative abundance of each of the 11 ASVs significantly over- or under-abundant in the ALAN condition were used here as markers of gill microbiota dysbiosis. For all the statistical tests, a difference was considered significant for a p-value = 0.05.

3. Results

3.1. Oysters’ Behavior

The mean hourly valve opening duration (VOD) during the daytime and nighttime of each day of the experiment is shown in Figure 1a,b for the control and ALAN oysters, respectively. The mean daytime/nighttime VOD is shown in Figure 1c. Results reveal strong diurnal behavior, i.e., a significantly higher activity during daytime compared to nighttime (Figure 1c), throughout the experiment for both conditions (Figure 1a,b). However, this diurnal behavior is significantly reduced in the presence of ALAN, due to a significant VOD decrease in the daytime (−15%) and increase in the nighttime (+157%) compared to the control condition (Figure 1c). This lower daytime/nighttime difference with ALAN exposure appears to be maintained throughout the experiment (Figure 1b). Figure 1d,e focus on the oysters’ behavioral mean daily pattern at the hourly scale in the control and ALAN conditions, respectively. In both conditions, the valve behavior shows a strong diurnal pattern, characterized by a daily VOD peak at 10:00–11:00 h. However, in the ALAN condition, a second peak appeared 2 h after the beginning of the nighttime (18:00–19:00 h), highlighting that this peak is not a direct response to ALAN. Figure 1f shows the individual variability of the VOD data and reveals a significant increase in this individual variability (+35%) in the ALAN condition compared to the control one (p < 0.001).

3.2. Behavioral Daily Rhythm

In both conditions, all individuals showed a significant behavioral daily rhythm. Figure 2 shows the oysters’ behavioral daily rhythm parameters: strength (i.e., robustness) of the rhythm (Figure 2a,d), amplitude (Figure 2b,e), and acrophase (Figure 2c,f).

When oysters are exposed to ALAN, the strength of the daily rhythm is significantly lower on 12 of the 28 days of the experiment, compared to the control (Figure 2a). These 12 days are spread throughout the experiment duration. No significant trend of this daily rhythm strength is observed according to the time in both conditions (linear regressions with p > 0.05) (Figure 2a). The mean strength of the oysters’ daily rhythm is also significantly lower (−22.7%) when they are exposed to ALAN (p < 0.001; Figure 2d). Concerning the rhythm’s amplitude, it significantly decreases in the presence of ALAN on 9 of the 28 days (Figure 2b), leading to an overall significant decrease (−12.6%) in the amplitude when oysters are exposed to ALAN (p < 0.001; Figure 2e). Again, these 9 days were spread throughout the experiment duration, without a significant trend of this daily rhythm amplitude observed according to the time in both conditions (linear regressions with p > 0.05 for both conditions) (Figure 2b). Finally, the rhythm’s acrophase is significantly delayed in oysters exposed to ALAN only for the tenth day of the experiment (Figure 2c), but it is not significantly affected by ALAN over the overall 28 days of the experiment (p = 0.775; Figure 2f). No significant trend of this daily rhythm acrophase is observed according to the time in both conditions (linear regressions with p > 0.05) (Figure 2c).

3.3. Circadian Clock Gene Expression

Figure 3 shows the relative mRNA level of two core clock genes (CgClock and CgBmal1) in the control and ALAN conditions during nighttime (24:00 h). For both genes, the results reveal a significant difference in expression between conditions. During ALAN conditions, the relative expression of CgClock is significantly higher (+22.8%, p = 0.037), while the expression of CgBmal1 is significantly lower (−26.5%, p = 0.009).

3.4. Daily Shell Growth

Figure 4a shows the cumulated daily shell growth during the 28 days of the experiment. The results reveal an overall significantly lower shell growth in oysters exposed to ALAN compared to the control condition (p < 0.001). During the first 15 days of the experiment, the shell growth is almost null in the ALAN condition and increases in the second part of the experiment. On the contrary, in control oysters, the shell growth increases continuously during the whole experiment. In both conditions, a speed-up in shell growth is observed on the last day of the experiment. The daily shell growth is detailed in Figure S3. At the end of the experiment, the cumulated shell growth is 32% lower in the ALAN condition compared to the control condition. Figure 4b shows the individual variability of the oysters’ daily shell growth, which is significantly higher (+81%) when exposed to ALAN compared to the control (p < 0.001).

3.5. Bacterial Community Composition

Bacterial community analysis reveals a difference in phylum relative abundance between oysters’ gills, dominated by Spirochaetota, Proteobacteria, and Campylobacterota, and water samples, dominated by Proteobacteria, Bacteroidota, and Planctomycetota (Supplementary Figure S4). Figure 5a shows the bacterial community composition at the phylum level in the gill tissues of oysters during nighttime and daytime for each experimental condition (control and ALAN). In control oysters, the bacterial community composition in gills shows changes between daytime and nighttime, especially with a decrease in Spirochaetota from 72% to 51%, while in the ALAN condition, these relative abundances were similar between daytime and nighttime (65% and 64%). On the contrary, the relative abundance of the Proteobacteria phylum increases in the control condition from daytime (11%) to nighttime (23%), while in the ALAN condition, it decreases (from 21% to 16%). The third most abundant phylum, the Campylobacterota, increases from daytime (6%) to nighttime (8%) in the control condition. This increase also occurs in the ALAN condition, but to a more important matter (from 4% to 14%). Focusing on daytime, results reveal a slight decrease in Spirochaetota relative abundance from 72% to 65% in the presence of ALAN, as well as an increase in Proteobacteria relative abundance from 11% to 21%, and a decrease in Campylobacterota relative abundance from 6% to 4% (Supplementary Figure S5). On the other hand, during nighttime, the relative abundance of Spirochaetota increases from 51% to 64% in the presence of ALAN, while the relative abundance of Proteobacteria decreases from 23% to 16%, and the Campylobacterota relative abundance increases from 8% to 14% (Supplementary Figure S5). Figure 5b shows a Venn diagram displaying the number of Amplicon Sequence Variants (ASVs) shared among the four conditions. It reveals that among 246 observed ASVs (only ASVs > 0.1% relative abundance), combining all conditions, 39 ASVs (15.9%) are shared by all conditions (core microbiota), 20 ASVs (8.1%) are specific to the control condition, and 11 ASVs (4.5%) are specific to the ALAN condition. The taxonomic composition at the phylum, family, and genus levels of the core microbiota and the ASVs specific to the control and ALAN conditions are shown in Supplementary Table S2.

3.6. Bacterial Diversity

Figure 6(a1–a4) shows the bacterial alpha diversity for each condition during daytime and nighttime. Four alpha diversity indexes are shown: the Shannon index (accounting for the richness, i.e., number of ASVs, and evenness, i.e., the relative abundances of ASVs), the observed ASVs (row measure of richness), the Chao1 index (estimated richness index accounting for undetected species), and the Inverse Simpson index (accounting for dominance and evenness; reflects the probability that two sequences sampled at random come from the same species). The comparison of the Shannon index value (Figure 6(a1)) reveals a significantly higher alpha diversity during nighttime in the control condition (3.4 ± 0.3; mean ± SE) compared to all other conditions (daytime in control (2.3 ± 0.2), daytime in ALAN (2.3 ± 0.2), and nighttime in ALAN (2.1 ± 0.1)) (p = 0.021). Thus, the significant day/night difference observed in the control condition disappears in the ALAN condition. Although there is a tendency to an increase in control nighttime for all other indexes, no differences are significant for the observed ASVs (Figure 6(a2), p = 0.090), the Chao1 index (Figure 6(a3), p = 0.124), and the Inverse Simpson index (Figure 6(a4), p = 0.180). Concerning the water samples, no significant differences are found when comparing the four alpha diversity indexes between conditions: p = 0.571 for the observed ASVs, p = 0.662 for the Shannon index, p = 0.458 for the Chao1 index, and p = 0.154 for the Inverse Simpson index.

To better understand the observed bacterial communities, we investigated the beta diversity (Figure 6b), i.e., the measure of similarity or dissimilarity between bacterial communities, using the weighted UniFrac distance, plotted as non-metric multi-dimensional scaling (NMDS) ordination plots (stress = 0.071). The results show a main significant difference in beta diversity between conditions and daytime and nighttime (p = 0.039), but post hoc tests do not discriminate any further significant differences between conditions and/or time of the day.

3.7. Bacterial Differential Relative Abundance

A differential relative abundance analysis of gill microbiota between the control and ALAN conditions, including daytime and nighttime data, shows that 13 ASVs are significantly differentially abundant (Supplementary Table S3), of which 11 ASVs are identified at the family level and shown in Figure 7. In the ALAN condition, two ASVs belonging to the Nitrincolaceae family are under-abundant, while nine ASVs belonging to eight different families (Arcobacteraceae, Pirellulaceae, Flavobacteriaceae, DEV007, Coxiellaceae, Colwelliaceae, AB1, Burkholderiaceae) are over-abundant in gills compared to the control.

3.8. Individual Data-Based Correlations Between ALAN Exposure, Strength of the Daily Rhythm, Shell Growth, and Differentially Abundant ASVs

We performed Spearman correlations to investigate the correlation at the individual scale (n = 24 oysters) between the daily rhythm robustness, shell growth, gill microbiota dysbiosis, and ALAN exposure. In this analysis, we chose to use as markers of microbiota dysbiosis the relative abundance of the 13 ASVs that were differentially abundant with ALAN exposure compared to the control condition, shown in Figure 7 and Table S3. The abundances of four of these ASVs (ASVs written in bold in Figure 7) showed a significant correlation with the three other factors (ALAN exposure, rhythm robustness, and shell growth). For these four ASVs (ASV39, ASV59, ASV348, and ASV6), Figure 8(a1–a4)shows 3D-scatter plots of the distribution of control and ALAN individuals. It shows a clear segregation between the control and ALAN conditions for the parameters measured. In addition, Figure 8b shows the Spearman correlation coefficient table between these parameters and ALAN exposure. This table shows a significant negative relationship between ALAN exposure and the strength of the rhythm (r = −0.695) and the shell growth (r = −0.683). On the contrary, a positive correlation is shown between the shell growth and the strength of the rhythm (r = 0.723). Concerning the microbiota dysbiosis, a significant correlation between the four ASVs’ abundance and ALAN exposure is found, negative with ASV39 (r = −0.908) and ASV59 (r = −0.880), and positive with ASV348 (r = 0.836) and ASV6 (r = 0.803). Furthermore, the significant correlation with the strength of the rhythm is positive with all four ASVs (r = 0.616, r = 0.594, r = 0.599, and r = 0.469). Finally, the significant correlation with the shell growth is positive with ASV39 (r = 0.711), ASV59 (r = 0.611), and ASV348 (r = 0.536), and negative with ASV6 (r = −0.538). Figure S6 and Table S4 present the same results regarding the nine remaining ASVs (of the 13 ASVs significantly differentially abundant in the ALAN condition compared to the control). The abundances of these nine ASVs do not show a significant correlation with the three other factors (ALAN exposure, rhythm robustness, and shell growth), except for the ASV205, which had a positive correlation with ALAN exposure (r = 0.410). The nomenclature of ASVs is detailed in the Supplementary Table S3.

4. Discussion

The present study shows that when exposed to a low level of ALAN (~1 lx), C. gigas circadian core clock gene (CgClock and CgBmal1) expression was modified during nighttime, and its behavioral daily rhythm was modified in terms of robustness, daily profile, and rhythmic characteristics. In parallel, the oysters’ shell growth was significantly decreased by ALAN, with a 32% gap on the last day of the experiment, and a gill microbiota dysbiosis was observed. Finally, an individual data-based analysis revealed significant correlations between ALAN exposure, the behavioral daily rhythm robustness, the growth, and the relative abundance of four ASVs of the gill microbiota affected by ALAN. The more the robustness of behavioral daily rhythm was impaired by ALAN, the more the shell growth was limited and the more the relative abundance of these ASVs was disrupted.

Daily rhythms have a molecular origin, the circadian clock, based on several interlocked transcriptional and post-translational auto-regulatory feedback loops. A common feature among organisms was characterized by the formation of the heterodimer CLOCK:BMAL, which activated the transcription of genes under its control [52]. In the current study, with ALAN exposure, it was shown a difference in the expression level of CgClock and CgBmal1, the orthologue genes coding for CLOCK and BMAL1 in C. gigas [18]. These results suggest an impairment of the oyster’s circadian clock functioning in the presence of ALAN. To validate this assumption, the valve activity behavior was investigated as an output of the circadian clock, as previously shown in many studies on C. gigas [18,19,22,33]. The results show that the behavioral daily rhythmic profile is modified with ALAN, with a new peak of activity during early nighttime. Moreover, both amplitude and robustness of the behavioral daily rhythm decrease with ALAN. These results corroborate a previous one-week study performed in C. gigas with the same ALAN intensity (1 lx) [23]. Here, these impacts persist throughout the 28 days of exposure, without a time effect, suggesting no habituation or lag effect of ALAN exposure. The results also show that ALAN’s impact on the daily rhythm of oysters is illustrated by both the decrease in the diurnal opening duration and the increase in the nocturnal opening duration. This result once again supports the idea that ALAN’s effects on oysters’ behavior are not limited to a direct response during nighttime, as is often described in the literature [53], but are a consequence of a disruption of the circadian clock functioning [54]. Finally, our results show that ALAN induces an increase in the individual variability of valve activity, attesting to a loss of synchronicity between oysters. Alterations of behavioral daily rhythms by ALAN have been observed in both terrestrial (European blackbirds Turdus merula) and coastal species (sandy beach isopod Tylos spinulosus [55] and rockfish Girella laevifrons [56]). These impacts on activity patterns may be detrimental for the fitness of organisms, as they have evolved biological timing to limit the risks and optimize the opportunities of their cycling environment [2,54,57]. For oysters, valve activity is closely related to crucial functions such as respiration, nutrition, protection against predators, feces ejection, or spawning [32,34,58]. Thus, modifications of their behavioral timing along the days may have widespread consequences on their physiology, their biotic interactions (with predators and prey), and even on the biochemistry of their environment [59].

Then, the objective was to study whether the observed impacts due to ALAN on the circadian clock and daily behavior were associated with physiological impairments, focusing on shell growth and gill microbiota. Bivalves’ shell growth is the result of a biomineralization process, leading to the formation of growth increments. This process is not constant over time and has been shown to notably follow daily rhythms [23,24,60]. While the regulatory process is not fully characterized to date, the daily pattern of biomineralization would not be a direct response to light/dark cycles but rather would be under the control of the endogenous clock [26,27,60]. Here, we show that the cumulated daily shell growth is decreased by 32% on the last day of the experiment when oysters are exposed to ALAN. Specifically, under artificial nighttime lighting, we observed a lag effect on shell growth, with almost negligible shell growth until day 15. An increase was then observed during the second half of the experiment. In contrast, under the control condition, shell growth increased steadily throughout the 28 days of the experiment. At the end of these 28 days, shell growth in both the control and ALAN-exposed oysters appeared to accelerate similarly, and the results obtained in the two conditions seemed to converge. It would be interesting to conduct experiments over a longer period to determine whether the cumulative shell growth of ALAN-exposed oysters could catch up with, or even surpass, that of the control oysters. Nevertheless, the individual-based analysis reveals a significant correlation between the impairment of oysters ’daily rhythm robustness by ALAN and the deficit of shell growth. These results support the hypotheses that (1) the oysters’ shell growth is under the control of the endogenous clock, and (2) that ALAN exposure, through circadian clock disruption, impairs oysters’ shell growth. In several studies based on different species, the clock-controlled growth has already been revealed [61,62]. As an illustration, human growth hormone (GH) gene expression and protein levels oscillate over a 24 h cycle, regulated by the circadian machinery [62]. In plants, the circadian clock regulates a multitude of factors that affect growth, such as water and carbon availability, and hormone signaling pathways [61]. In oysters, the clock control of growth could, among other things, act through the circadian regulation of genes involved in the molecular pathways of the biomineralization process in the mantle tissue. Moreover, ALAN reducing growth has also been shown in other organisms, such as the freshwater gammarid Gammarus jazdzewskii or the juvenile toad Anaxyrus americanus, exposed to white LED at an intensity of 2 lx and 16.9 lx, respectively [63,64]. In the nocturnal amphipod Orchestroidea tuberculata, ALAN also impairs growth, in association with the disruption of daily activity rhythm and foraging behavior [65]. Finally, we show in this study that the presence of ALAN exacerbates the individual variability of the oysters’ shell growth, which might suggest some individual diversity in physiological responses to ALAN. Another factor that could affect physiological responses to ALAN is the season. Indeed, besides daily patterns, oysters’ shell growth clearly shows annual oscillations, hypothesized as the results of the endogenous clock entrainment at the annual scale by photoperiod [20]. Thus, we can hypothesize that ALAN could also impact the shell growth by disrupting its annual rhythm. Our experiment took place at the end of winter, when oysters have a very low shell growth compared to the spring, where they reach their maximum shell growth [20]. A future experiment covering the four seasons, preferably in field conditions, would be necessary to characterize how ALAN affects oysters’ shell growth, integrating the annual physiological rhythms, such as the reproduction cycle.

The role of microbiota in health and disease has been demonstrated by numerous studies [31]. Reciprocal interactions between a host’s circadian clock and its microbiota have been described in model species [28,29,66]. In our study, we reveal in C. gigas a day/night difference in the gill’s microbial phyla relative abundances and a day/night difference in alpha microbial diversity. Importantly, these day/night differences are altered (phyla abundances) or annihilated (microbial diversity) when oysters are exposed to ALAN. Finally, results of ALAN exposure suggest a dysbiosis in oysters’ gill microbiota, with 13 ASVs significantly under- or over-abundant compared to the control condition. Furthermore, despite the core microbiota [67], some ASVs are specifically observed in one or other conditions. Daily oscillations in microbiota abundance have already been observed many times in mice [28,29,66], but also in other invertebrate organisms such as the worm Macrostomum lignano and the cnidarian Nematostella vectensis [68,69]. In mice, it has been demonstrated that the microbiota rhythmicity is host dependent; daily oscillations in gut microbiota are damped or abolished in clock-disrupted hosts [28,29,66]. Here, we show that the disruption of the relative abundance of four ASVs by ALAN is directly correlated with the decrease in the behavioral daily rhythm robustness. These results are a first clue indicating that the observed impacts on oyster’s gill microbiota might be the consequence of the circadian clock disruption by ALAN. Our results also showed that the disruption of the relative abundance of these ASVs and of the daily rhythm induced by ALAN are correlated with the decrease in shell growth. In mice, the disruption of daily oscillations in microbiota abundances correlates with the disruption of daily feeding patterns in the clock-disrupted hosts [66]. By this analogy, the abrogation of day/night differences in microbiota diversity in oysters could suggest a disruption of nutritional rhythms and thus may impair the energy allocation for growth. This assumption is supported by the observed modification of valve behavior when oysters are exposed to ALAN, and considering that oyster nutrition is directly dependent on valve opening. Alternative routes, rather than feeding issues, might also explain how the host’s clock can regulate the microbiome. Indeed, as proposed in humans and mice [66], the disruption of the circadian regulation of endocrine factors like melatonin or corticosteroids may lead to microbiota impairment. In this sense, an annihilation of day/night difference in hiomt gene expression, a gene involved in melatonin synthesis, has been observed in oysters exposed to ALAN from 0.1 lx [23]. Both circadian and microbiota disruption have been shown to impair the growth hormone in mice [66]. Thus, considering the direct correlation between the disruption of the relative abundance of four ASVs affected by ALAN and the shell growth impairment observed in our study, we cannot exclude that microbiota dysbiosis may be involved in the observed deficit of shell growth in oysters exposed to ALAN. While microbiota studies in bivalves are still in their early stages [70,71,72,73], its role in the defense against pathogens, such as the herpesvirus OsVH-1, has been advanced several times [74,75,76,77,78]. Here, we highlight that besides the quality of the microbiota, the daily temporal dynamics of community composition of the microbiota may be decisive in health issues in bivalve species.

To conclude, our results highlight that even in eyeless organisms living in coastal areas, low nocturnal light intensity due to human activities might have strong and measurable consequences. Thus, this study shows that through the modification of marine organisms’ biological rhythms, ALAN could impact a plethora of key physiological functions and putatively modify the relationship between the individual and its biotope, leading to alterations in the coastal biodiversity.