Circatidal control of gene expression in the deep-sea hot vent shrimp Rimicaris leurokolos

Abstract

Biological clocks are a ubiquitous feature of all life, enabling the use of natural environmental cycles to track time. Although studies on circadian rhythms have contributed greatly to the knowledge of chronobiology, biological rhythms in dark biospheres such as the deep sea remain poorly understood. Here, based on a free-running experiment in the laboratory, we reveal potentially endogenous rhythms in the gene expression of the deep-sea hydrothermal vent shrimp Rimicaris leurokolos. Oscillations with approximately 12 h periods, probably reflecting tidal influence, greatly prevail over others in the temporal transcriptome, indicating R. leurokolos probably depends on a circatidal clock consisting of at least some components independent from the circadian clocks. The tidal transcripts exhibit an antiphased expression pattern divided into two internally synchronized clusters, correlated with wide-ranging biological processes that occur in the nucleus and cytoplasm, respectively. In addition, the tidal transcripts showed great similarities with genes in fruit flies and mice exhibiting approximately 12 h ultradian rhythms, indicating that the tide probably had a broad impact on the evolution of approximately 12 h oscillations found across the Metazoa. These findings not only provide new insights into the temporal adaptations in deep-sea organisms but also highlight hydrothermal vent organisms as intriguing models for chronobiological studies, particularly those linked to approximately 12 h ultradian rhythms.

1. Introduction

As an adaptation to natural periodic environmental changes, biological clocks have been evolving in all kingdoms of life to keep track of time [1–3]. Numerous organisms exhibit endogenous rhythms dominated by circadian periodicity of approximately 24 h at behavioural, physiological, and molecular levels, which help them to anticipate daily changes such as light, temperature, and food availability [4–6]. To achieve this, animals rely on their internal circadian clock systems, like transcription–translation feedback loops (TTFLs), that can be entrained to match the environmental cycles [7]. In dark biospheres such as underground and deep-sea habitats, however, the absence of sunlight veils organisms there from the direct effects of everlasting day–night transitions. How organisms living in the dark schedule their activities through time, as an adaptation to darkness, remains obscure [8,9].

The deep ocean, the largest but the least studied habitat on Earth [10], is of great interest for chronobiological research owing to its distinct ecological environment [11]. With increasing depth, the sunlight intensity dramatically decreases and the visible light spectrum rapidly narrows to result in zero sunlight penetration below 1000 m even in the clearest oceanic waters [12]. From the ecological perspective, it is more appropriate for deep-sea organisms to synchronize their biological rhythms with other zeitgebers (e.g. tidal cycles) instead of the light cycle [11,13]. Indeed, various environmental fluctuations related to tidal cycles have been observed in the deep, notably in hydrothermal vent ecosystems [14–17]. Vent faunas rely strongly on their local environment and survive on organic compounds produced by chemosynthetic microbes utilizing dissolved chemicals in the vent fluids [18]. Therefore, vent animals are likely to be affected by spatial and temporal variations in environmental factors modulated by the tide. Several previous studies have reported potential tide-related rhythms of behaviour and community dynamics in deep-sea organisms from chemosynthetic ecosystems [19–21] and a recent study revealed that tidal cycles dominated the gene expression of the deep-sea vent mussel Bathymodiolus azoricus through periodic in situ sampling across 24.8 h [22]. Although various evidence suggests that vent organisms exhibit biological rhythms in the wild, whether they need endogenous rhythms or clocks to accommodate and anticipate local environmental changes to maximize their fitness is still poorly understood.

In the Iheya North hydrothermal vent field of the Okinawa Trough (figure 1a), tidal modulation on the deep-sea current dynamics was observed previously [23]. According to a published long-term temperature measurement (mean ± s.d. = 4.4 ± 0.1°C) recorded at the periphery of the vent field where bathymodioline mussels typically occur, a strong semidiurnal tidal cycle (12.4 h periodicity) was found in the environmental fluctuations as well (figure 1b). Unlike mussels at the periphery, however, the alvinocaridid shrimp Rimicaris leurokolos (formerly Shinkaicaris leurokolos [24]) lives close to the vigorously venting orifice (0.2−0.8 m and occasionally less than 0.2 m from the vent [25]) where the hot fluid and cold surrounding seawater mix spontaneously and dramatically ([26]; figure 1c). In the R. leurokolos colony, environmental factors fluctuate frequently and dramatically on the seconds-minute scale, which adds unpredictable noises to the tidal cycles. In fact, the temperature could rapidly fluctuate between 7.63 and 17.74°C (mean ± s.d. = 10.16 ± 1.71°C) within approximately 2.5 min (figure 1d). As such, R. leurokolos is probably unable to directly use the unsteady, highly variable environmental factors as zeitgebers to entrain its biological rhythm. However, patterns of venting activity across a longer time span are known to be linked to tidal cycles [27]. Therefore, we hypothesized that an internal clock may integrate all periodic cues, whether they are clear (e.g. pressure) or noisy (e.g. water chemistry) and thus help hot vent shrimps adapt and keep track of periodic environmental changes.

Figure 1.

Deep-sea chronobiology study on the hydrothermal vent shrimp Rimicaris leurokolos. (a) This work was focused on the Iheya North hydrothermal vent field, Okinawa Trough. (b) Time series of published temperature data covering 351 days recorded at the peripheral area of the active NBC mound vent. Lomb-Scargle (LS) periodogram analysis reveals a strong semidiurnal tidal cycle dominated by the M2 tidal constituent. M2, principal lunar semidiurnal constituent; S2, principal solar semidiurnal constituent; N2, larger lunar elliptic semidiurnal constituent; K1, luni-solar declinational diurnal constituent; O1, lunar declinational diurnal constituent. (c) Rimicaris leurokolos shrimps (in the yellow circle) live in a hot zone near the vent orifice where hydrothermal fluids mix with cold surrounding seawater spontaneously. (d) Time series of in situ temperature and dissolved oxygen records during 162 s when the sensor was placed in the R. leurokolos colony showing frequent environmental fluctuations with high variability.

Although circadian patterns are dominant in terrestrial and coastal animals, recent studies revealed robust ultradian rhythms (e.g. approx. 12 h cycling gene expressions) in model organisms such as mice and fruit flies [28,29]. These rhythms were found to be cell-autonomous, driven by a distinct approximately 12 h pacemaker largely independent from the circadian clock. As many homologous genes were found to be rhythmic with a period of approximately 12 h in both terrestrial and marine animals, a hypothesis was then proposed that the circatidal rhythms in marine animals and approximately 12 h rhythms in terrestrial animals are evolutionarily conserved [30]. However, given the dominant effects of the approximately 24 h circadian system, it is challenging to distinguish ultradian rhythms from circadian rhythms in terrestrial or coastal animals. The hot vent shrimp studied herein lives in the deep-sea environment lacking day–night cycles and thus provide an opportunity to examine this hypothesis.

Here, through a laboratory-based free-running experiment across 72 h, we examined the biological rhythms of gene expression in the vent shrimp R. leurokolos and tested if it can maintain rhythmicity under constant conditions. Our work reveals free-running rhythms dominated by approximately 12 h tidal patterns in the transcriptome of R. leurokolos, which are likely driven by an unidentified endogenous clock, including but not limited to circatidal components. The tidal transcripts exhibit a bimodal gene expression pattern and influence broad cellular processes such as DNA replication, DNA repair, regulation of transcription, transport, cell cycle and responses to environmental stresses. In addition, we found R. leurokolos shares many common genes and biological processes in the tidal rhythms with approximately 12 h rhythms in sea anemones, fruit flies, and mammals, suggesting the evolutionary conservation of gene expression rhythms at approximately 12 h across a broad range of phyla. These findings highlight the prospect for using deep-sea organisms as models in chronobiology, especially in the evolution and molecular mechanisms of approximately 12 h ultradian rhythms.

2. Material and methods

(a). Environmental temperature monitoring

The local fluctuations in the temperature of an R. leurokolos colony on the North Big Chimney (NBC; 126°53.80′ E, 27°47.45′ N; 982 m deep) of the Original Site, Iheya North vent field, Okinawa Trough were measured using a RINKO I dissolved oxygen and temperature sensor (JFE Advantech Co., Ltd.) during dive no. 1610 of the remotely operated vehicle (ROV) Hyper-Dolphin on board R/V KAIYO cruise KY14−01 for 162 s and data were recorded every second (electronic supplementary material, data S1). Published temperature data in a peripheral area of the Original Site, Iheya North vent field, far away from active chimneys, which were recorded continuously every 10 min from 4 September 2010 to 20 August 2011, were re-analysed [31]. The Lomb–Scargle periodogram was conducted to detect periodicity in the temperature time series.

(b). Sampling procedure and experiment design

A total of 184 adult individuals of R. leurokolos were collected by a suction sampler from the top of the NBC chimney at the Original Site, Iheya North vent field, during dive no. 1613 of ROV Hyper-Dolphin on board R/V KAIYO cruise KY14-01. Specimens of R. leurokolos were kept in rotary canister bottles on the ROV prior to recovery on board the research vessel. After sampling, it took about 2 h (15.17−17.07 JST, Japan Standard Time) for the ROV to be recovered. Once on-board the ship, under natural light, 154 live shrimps, excluding 25 dead individuals and five ovigerous females, were immediately transferred into a lightproof seawater container. This container was placed in a dark, cold room at 4°C by 19.30 JST for the subsequent free-running experiment. The free-running experiment began at 22.00 JST, aiming to investigate the gene expression of R. leurokolos in a constant environment over time. At each sampling timepoint, shrimps were collected by a meshed scoop, dissected with scissors without anaesthesia into cephalothorax and abdomen and fixed every 4 h for a total duration of 72 h. The cephalothoraxes of each individual were preserved in RNAlater (Invitrogen) and used for RNA-Seq in this study. Details of the free-running experiment procedure are available in the electronic supplementary material, methods S1.

(c). RNA extraction, sequencing and assembly

Total RNA was extracted from individuals using TRIzol reagent (ThermoFisher) following the manufacturer’s protocol. The messenger RNA from each shrimp cephalothorax was enriched by Oligo-dT probes and further reverse transcribed to first-strand complementary DNA (cDNA) for library construction. All cDNA libraries were sequenced on an Illumina NovaSeq 6000 platform with a 150 bp paired-end mode. A total of 57 RNA-Seq samples were generated from R. leurokolos during 72 h of sampling at 4 h intervals with three replicates per time point. Raw reads were firstly trimmed using Trimmomatic v. 0.39 [32] and then potentially associated bacteria contamination reads were removed with Kraken2 v. 2.0.8-beta [33]. De novo assembly was performed with Trinity v. 2.11.0 [34] (set Jaccard clip option). Salmon v. 1.4.0 [35] was used to quantify the expression of each assembled transcript.

(d). Gene quantification and annotation

The trimmed mean of M values (TMM) cross-sample normalization method [36] was applied to remove the sequencing variations and potential ‘batch effects’ among samples using the abundance_estimates_to_matrix.pl script in the Trinity toolkit. In addition, lowly expressed transcripts with 0 counts in at least 38 samples were filtered out, leaving 23 905 transcripts for subsequent analyses. The predicted proteins were annotated by searching the candidate protein sequences against the NCBI non-redundant database using the proteic basic local alignment search tool (BLASTP) with Diamond v. 2.0.13 [37] with ultra-sensitive mode and an E-value threshold of 1 × 10^–5, the Kyoto Encyclopedia of Genes and Genomes (KEGG) database via KEGG Automatic Annotation Server (KAAS) and the Gene Ontology (GO) via BLAST2GO version 6.0 [38]. The gene enrichment analyses were performed with ClusterProfiler v. 4.2.2 [39].

(e). Gene expression and detrend analyses

Principal components analysis (PCA) was performed using the plot PCA function in the R package DESeq2 v. 1.34.0 [40]. Permutational multivariate analysis of variance (PERMANOVA) and analysis of similarities (ANOSIM) were applied to test the statistical significance of the PCA results. The potential monotonic trends in the raw gene expression of all transcripts were determined using the Mann–Kendall trend test, and we pinpointed two overall clusters of transcripts that had either an upward trend (S score > 0 and p‐value < 0.05) or a downward trend (S score < 0 and p‐value < 0.05). To remove the trend and produce a steady baseline for subsequent rhythm analyses, we used CEEMDAN [41], an adaptive decomposition method designed for non-stationary and nonlinear time series signals, which minimizes the impact of distorting the period, phase or amplitude change of the oscillations. To verify the effectiveness of detrend analysis, we compared the rhythm detection results using gene expression data before and after detrending. The results suggested that the detrend analysis increases the power of rhythmicity detection methods and thus helps us to get a clearer picture of the periodic patterns within the gene expression data. Details of detrend analysis are available in the electronic supplementary material, methods S1, figures S1–S3.

(f). Rhythm detection analyses

Overall, we chose four rhythm detection algorithms in this study, the mathematical method eigenvalue/pencil [28] (carried out with customized scripts), the extended circadian harmonic oscillator application v. 4.0.1 [42] (ECHO) and two non-parametric statistical methods, RAIN v. 1.28.0 [43] and empirical JTK_CYCLE v. 1.0 [44] (eJTK), which were designed to detect asymmetric waveforms. Tidal and daily transcripts were defined by a significant period of 12 and 24 h (RAIN and eJTK) or a period range of 12 ± 2 and 24 ± 2 h (eigenvalue/pencil and ECHO). Multiple testing was carried out to adjust p-values using the Benjamini–Hochberg method for RAIN and ECHO and an empirical method based on the gamma distribution for combining dependent p-values for eJTK. Transcripts with an false discovery rate (FDR) ≤ 0.05 were considered statistically rhythmic. We also used ECHO to fit data with an extended fixed amplitude model and categorize the oscillation type using the amplitude change coefficient (γ) as forced (−0.15 < γ < −0.03; amplified amplitude), harmonic (−0.03 ≤ γ ≤ 0.03; steady amplitude), damped (0.03 < γ < 0.15; reduced amplitude), overexpressed (γ ≤ −0.15) and repressed (γ > 0.15), of which the latter two are considered arrhythmic [42]. Details of rhythm detection analysis are available in the electronic supplementary material, methods S1, table S1 and data S2.

(g). Identification of putative circadian clock genes

We searched for putative circadian clock genes in our de novo assembled transcriptome using BLASTP carried out by Diamond with a series of known core clock and clock-associated genes as queries (electronic supplementary material, data S3). For candidate circadian clock transcripts, sequences were scanned against the coding regions and conserved functional domains using the InterProScan tools [45]. To determine the sequence homology of the candidate transcripts, we further conducted phylogenetic analyses using IQ-Tree v. 2.1.3 [46] with the ultrafast bootstrap approximation (UFBoot) and ModelFinder for examining the best-fit model. Details of phylogenetic analyses are available in the electronic supplementary material, methods S1.

(h). Conserved rhythmic orthologs analysis

The approximately 12 h rhythmic gene expression data of mice liver and Drosophila S2 cells used in this study were published previously [28,29]. Together with the de novo transcriptome of R. leurokolos assembled in this study, the protein sequences from mouse genome assembly GCF_000001635 and fruit fly genome assembly GCF_000001215 were used to construct orthogroups using Orthofinder v. 2.5.4 [47] with default settings. The conserved rhythmic analysis was performed with the single-copy orthologous genes. Pathway network enrichment analysis of conserved single-copy orthologous genes was performed with aPEAR v. 1.0.0 [48]. Lists of descriptions and GO enrichment results of approximately 12 h rhythmic single-copy orthologous genes are accessible in the electronic supplementary material data S4.

3. Results and discussion

(a). Transcriptome analysis of deep-sea vent shrimp Rimicaris leurokolos

To study the potential endogenous rhythms in R. leurokolos, we carried out a 72 h free-running experiment with 4 h sampling intervals under constant conditions. According to previous studies on crustaceans, we worked on the whole cephalothorax that contained the nervous and visual system where the central clock probably lies, as its exact location in crustaceans remains unclear [49]. The reference de novo transcriptome contained 90 755 predicted open reading frames, which showed a 92% complete BUSCO score. Candidate protein-coding transcripts with low expression were filtered out, leaving 23 905 transcripts with a 60.4% annotation rate for rhythmic analyses. A PCA of normalized raw gene expression showed a rough trend from the start of sampling (0 h) to the end (72 h; electronic electronic supplementary material, figure S4). Performing the Mann–Kendall trend test on the raw dataset, we found that 10 273 transcripts (i.e. 42.97% of the transcriptome) had a significant monotonic trend (p‐value < 0.05). The trend was then removed using CEEMDAN [41] to produce a steady baseline and enhance the strength of rhythm detection methods (electronic electronic supplementary material, figure S5). In the PCA results plotted using detrended gene expression, the trend disappeared and sample points from different sampling times clustered together (figure 2a), indicating a minimal external effect on the experiment as the shrimps were seemingly experiencing the same conditions.

Figure 2.

The Rimicaris leurokolos temporal transcriptome was dominated by tidal oscillations under constant conditions. (a) Principal components analysis (PCA) plot showing the detrended transcriptomic data from the free-running experiment samples (n = 57). The significance of inter-group differences was examined by the PERMANOVA (p‐value = 0.376) and ANOSIM (p‐value = 0.958) analyses both on Bray–Curtis dissimilarities. (b) Distribution of the periods of oscillations identified by the eigenvalue/pencil method and the ECHO application. Oscillations with a period range of 10−14 h and 22−26 h are regarded as tidal and daily cycling, respectively. (c,d) Venn diagrams detailing the number of rhythmic transcripts (FDR < 0.05) identified by ECHO, RAIN and eJTK in the (c) tidal range (d) and daily range. (e) Raincloud plot of the root mean square amplitude for approximately 12 h tidal transcripts and approximately 24 h daily transcripts. The stars indicated the level of significance: ****p‐value < 0.0001.

(b). General biological rhythms in the temporal transcriptome of Rimicaris leurokolos

We first used the eigenvalue/pencil [28] and ECHO [42] algorithms to carry out unbiased identification of any potential periodic oscillations in the gene expression of R. leurokolos. The results uncovered similar rhythmicity, with most cycling transcripts identified as ultradian rhythms, which oscillated in the tidal range of 10−14 h, while much fewer cycling transcripts oscillated in the daily range of 22−26 h (figure 2b). The eigenvalue/pencil algorithm was able to identify superimposed oscillations from one cycling transcript. For cycling transcripts with multiple components composed of tidal and daily oscillations, the amplitude of the tidal component is significantly larger, suggesting a stronger rhythmicity (electronic electronic supplementary material, figure S6). Despite the high sensitivity of detecting weak ultradian rhythms, the eigenvalue/pencil method could have high positive false rates [50]. To validate the results, we then detected cycling transcripts with two robust nonparametric algorithms, RAIN [43] and eJTK [44]. Searching for cycling transcripts (FDR < 0.05) with periods in the range of 8−28 h at regular intervals of 4 h, we found the number of rhythmic transcripts identified by RAIN and eJTK both reached a peak at the 12 h period that corresponded to a tidal cycle (electronic supplementary material, figure S7). Intriguingly, the results from ECHO shared many common cycling transcripts with RAIN (60.8% of transcripts) and eJTK (65.8% of transcripts) but detected numerous extra tidal transcripts, probably owing to the inclusiveness of identifying different harmonics of oscillating elements [51]. Merging transcripts that were identified as rhythmic by at least one method, we identified a total of 8122 rhythmic transcripts (33.98% of the transcriptome) in the tidal range (figure 2c) versus 1230 rhythmic transcripts (5.15% of the transcriptome) in the daily range (figure 2d). Then, we calculated and compared the root mean square amplitude between tidal and daily transcripts and found the tidal transcripts had a significantly larger amplitude than daily ones, indicating tidal cycles contributed more biologically relevant effects (figure 2e).

These results demonstrate sustainable endogenous rhythmic gene expression in R. leurokolos under constant conditions. The temporal transcriptome is dominated by the approximately 12 h tidal rhythms, which greatly prevail over other oscillations with different periods, especially the 24 h daily transcripts that are probably unrelated to the circadian rhythms.

(c). Patterns of tidal gene expression in Rimicaris leurokolos

To uncover the biological rhythms ticking in the deep, we focused on the dominant tidal rhythms in the temporal transcriptome of R. leurokolos. Our 72 h duration of sampling covered approximately six tidal cycles, allowing us to test the stability and robustness of rhythmicity in the tidal gene expression [52]. To capture the amplitude changes over time, we fitted the gene expression to an extended fixed-amplitude model using ECHO [42] and categorized the results into five types of oscillation groups (for details, see §2). Nearly half of the tidal transcripts (n = 3 294) were identified as damped type, showing a reduced amplitude (figure 3a).

Figure 3.

Free-running tidal transcriptomic patterns of Rimicaris leurokolos. (a) Density distribution based on amplitude change coefficient (γ) of tidal transcripts detected by ECHO. Oscillation types are categorized as forced (−0.15 < γ < −0.03; amplified amplitude), harmonic (−0.03 ≤ γ ≤ 0.03; steady amplitude), damped (0.03 < γ < 0.15; reduced amplitude), overexpressed (γ ≤ −0.15), and repressed (γ > 0.15) with the latter two considered arrhythmic. (b) Rose chart showing phase distribution comparison of tidal transcripts in clusters 1 and 2. For each oscillation, the phase is divided by period length and then multiplied by 360 (degrees) to normalize phases of transcripts with different period lengths. (c) Heatmap showing highly synchronized gene expression of tidal transcripts (ECHO FDR < 0.05). Each row represents a transcript with three individuals per time point. Transcripts are clustered and ordered by phases, oscillation type, and monotonic trend. The expression of each gene is Z-score normalized. (d) Bubble diagram showing significantly enriched gene ontology (GO) terms related to biological processes of tidal subgroups categorized by three classification schemes (from left to right: cluster, oscillation type, trend). Cross symbols represent a lack of enriched terms in the corresponding subgroup.

Although there was a significantly amplified or reduced amplitude in at least 75% of tidal rhythmic transcripts, the period and phase of most tidal cycles were seemingly fixed. We divided the tidal transcriptome into two clusters by gene expression using K-means clustering (electronic supplementary material, figure S8a,b). The two clusters showed a largely antithetical gene expression pattern over time and transcripts within each cluster were peaking at the same time, resulting in an almost complementary phase distribution in one tidal cycle for the two clusters (figure 3b). This bimodal expression pattern is widely observed in the gene expression of endogenous rhythms. For example, terrestrial organisms such as mice and baboons experience transcriptional ‘rush hours’ preceding dawn and dusk, which is considered to assist in preparing for the impending change in lighting [5,6]. In the hydrothermal vent, tidal cycles could strongly affect variations in environmental factors such as temperature, water chemistry, and dissolved oxygen. Vent fauna would thus experience two opposite environmental conditions in a tidal cycle: one characterized by high temperature, high sulfide concentrations, low oxygen concentrations, etc., while the other has comparatively low temperature, low sulfide concentrations, high oxygen concentrations, etc. In this case, the bimodal tidal expression pattern could help R. leurokolos save energy during the transition periods between acrophases and troughs in the tidal cycle to make preparations for upcoming significant environmental changes [53–55].

The rhythmicity of 3837 transcripts in cluster 1 was relatively stable with almost equivalent amounts of modest forced, harmonic and damped oscillations, whereas the 2524 transcripts in cluster 2 were nearly occupied by damped oscillations that decreased obviously in amplitude, particularly during the second half of the experiment (figure 3c). Indeed, there was a seemingly strong correlation between trend directions and oscillation types (electronic supplementary material, figure S8c). In other words, a tidal transcript with a decreasing expression trend was more likely to have a reduced amplitude, and vice versa. This likelihood was extremely high in cluster 2, where over 94% of tidal transcripts with a downward trend were also damped. As we isolated R. leurokolos shrimps from the hydrothermal vent and cultivated them in the laboratory under constant conditions for 72 h, there were unsurprisingly considerable changes in the gene expression over time corresponding to transcriptional responses to potential stresses such as decompression. Therefore, the amplitude probably decreases as a byproduct of declining gene expression levels, especially for most transcripts in cluster 2. Although desynchronization between different tissues of R. leurokolos could lead to arrhythmic gene expression in our free-running study, we still detected a considerable number of rhythmic transcripts with highly synchronized expression, suggesting the effect of desynchronization was probably minor.

(d). Wide range of biological processes related to tidal cycles in Rimicaris leurokolos

The tidal rhythmic transcripts could be categorized by gene expression, oscillation types and original trends, all of which included several subgroups that probably reflected similar or different biological relevance. To search for potential biological functions modulated by tidal rhythms, we performed functional enrichment analyses over the tidal transcriptome based on these three classification schemes using ClusterProfiler [39]. GO and KEGG pathway analyses revealed that tidal rhythms generally affected various biological processes such as DNA repair, regulation of transcription, translation, transport, organelle organization, and citrate cycle (figure 3d; electronic supplementary material, figure S9). There was an expected high correlation between the biological functions of classified subgroups. For example, terms enriched in cluster 2 were more or less the same as those in the damped or downward subgroup. Notably, rhythmic processes related to the nucleus, such as chromatin organization, DNA biosynthetic process and transcription, were uniquely enriched in cluster 1, whereas rhythmic processes supported by the cytoplasm, such as translation, protein folding, vesicle-mediated transport and oxidative phosphorylation, were highly enriched in cluster 2. Thus, it is likely that R. leurokolos shrimps can organize various rhythmic biological processes in terms of spatial and temporal expression patterns to coordinate their internal physiology, regardless of the presence of clear zeitgebers.

In hydrothermal vents, particularly near the fluid orifice where R. leurokolos inhabits, hot, anoxic fluids with abundant potentially toxic chemical substances (e.g. hydrogen sulfide) mix dramatically with the cold surrounding seawater, resulting in an extreme and variable environment. Vent organisms have evolved various adaptive strategies at morphological, physiological, and biochemical levels [18]. We hypothesized that intrinsic rhythms could help vent animals schedule their activities to cope with the harsh environment. As expected, we found many tidally oscillating transcripts that were related to various cellular stress response processes such as sulfur metabolism, reactive oxygen species homeostasis, cellular detoxification, heat and hypoxia response, and innate immune response (electronic supplementary material, figures S10a–d). The representative tidal genes related to stress responses include Sqor, Cat, Sod, Gst, Gpx, Egln1, Dnaja1, Hsp90, and Ctsl, among others. Most of these transcripts showed damped rhythmic expression over time under constant conditions, suggesting that in situ environmental cues were necessary for the maintenance and synchronization of rhythmic processes related to stress responses. The results reveal broad tidal regulation of gene expression in the cellular response to various environmental stresses, suggesting an adaptive advantage for R. leurokolos living in harsh, fluctuating environments.

Mounting evidence suggests the presence of cross-talking between circadian clocks and the cell cycle [56,57]. Here, we found the annotation of many tidal transcripts was related to the cell cycle (electronic supplementary material, figure S11a–c). The transition between the G1 and S phase is controlled and regulated by key genes involved in the checkpoint pathways [57–59], of which several transcripts were tidally oscillating in the transcriptome, including Ccnd2, TP53, Mycbp2, Rbl1, E2f2, Tipin, Mdm2, among others. In addition, genes that encoded major components of the RAD9–RAD1–HUS1 clamp complex (Rad9a, Rad1, Hus1), which play pivotal roles in DNA repair, also oscillated in the tidal range [60]. As crucial elements of the spindle assembly checkpoint, genes such as Mad2, Bub1, and Bub3 that regulate chromosome separation were also found to be tidally cycling [61]. Taken together, tidal rhythms appear to generally regulate the progression of the cell cycle in R. leurokolos to cope with DNA damage derived from periodic intrinsic and environmental stresses.

(e). Possible molecular basis and zeitgebers of biological rhythms in Rimicaris leurokolos

We searched for and identified in R. leurokolos homologues of the canonical circadian clock genes that form the molecular oscillator in the fruit fly Drosophila melanogaster, including Clock (Clk), Period (Per), Timeless (Tim), Cryptochrome 2 (Cry2), Clockwork orange (Cwo), Doubletime (Dbt), Vrille (Vri), and Pdp1ε (electronic supplementary material, material S12a–d). Indeed, all identified R. leurokolos clock gene homologues lacked 24 h rhythmicity (electronic supplementary material, figure S13). This may provide intrinsic fitness for these hydrothermal vent shrimps that are probably completely isolated from the surface as adults [62]. Moreover, most clock genes showed a lack of rhythmicity in the tidal range as well. Although the transcriptional repressor Period and Doubletime, which encodes a kinase that phosphorylates the PERIOD protein [63], were statistically oscillated in the tidal range, the Period transcript showed an opposite expression with the Timeless transcript in the second half of the free-running experiment, which is abnormal for the circadian clock regulated by the TTFL mechanism. As a robust endogenous clock is required to be free-running with a self-sustained period for maintaining the rhythmicity of tidal oscillations in R. leurokolos, we hypothesize that there is a circatidal clock with at least part of the molecular components being independent from the circadian clock. Separate circadian and circatidal clocks were also reported in the intertidal isopod Eurydice pulchra [64] and the mangrove cricket Apteronemobius asahinai [65]. Nevertheless, the core clock genes still encode functional proteins with conserved domains (electronic supplementary material, figure S12e) and we cannot exclude the possibility that some core circadian clock genes are necessary for tidal rhythms [66,67], even though they have lost the free-running transcription rhythmicity. These circadian clock genes may also be useful during the dispersal of planktotrophic larvae, which possibly take place in the photic zone [68].

To precisely synchronize with the tidal cycle, we consider R. leurokolos shrimps to depend on an endogenous circatidal clock to integrate all potential periodic cues. In the microhabitat where R. leurokolos lives, crucial environmental factors such as temperature, water chemistry and dissolved oxygen can vary irregularly and dramatically in a very short time, generating noisy signals. Therefore, although these environmental factors are crucial for the survival of hot vent shrimps, they are probably poor periodic cues. Hydrostatic pressure and tidal currents, however, are largely in line with the standard of a synchronization cue [69,70]. In addition, they show significant coherence with temperature, water chemistry, and dissolved oxygen [71]. The closer the distance from hot vents, the stronger the modulation of tidal pressure on environmental factors [17]. Therefore, we propose hydrostatic pressure as the most likely synchronization cue for R. leurokolos. Over longer time scales, the overall levels of venting activity are known to be modulated by the tidal cycle. As such, by synchronizing to pressure, it is possible for hot vent shrimps to predict major environmental fluctuations. In addition to pressure and currents, light emissions from vents [72], food sources [73], or even soundscape cues from hot vents [74,75] and the surface [76] could be potentially useful periodic cues for R. leurokolos. Nevertheless, how exactly hot vent shrimps synchronize their endogenous rhythms with environmental cycles remains a question for the future.

(f). Rhythmicity comparison of two vent species living in different microenvironments

To compare the tidal rhythms in two deep-sea hydrothermal vent organisms, we applied our transcriptome analysis and rhythm detection pipeline to the published in situ gene expression dataset of the Atlantic deep-sea vent mussel Bathymodiolus azoricus [22]. Despite the different geographical localities of the two species, B. azoricus and R. leurokolos both show similar biological rhythms predominated by a tidal pattern that is related to the responses to environmental stresses such as DNA damage and thermal, immune, and oxidative stress, suggesting a generally temporal adaptation to the ‘extreme’ deep-sea hydrothermal environments. Compared with R. leurokolos, however, the proportion of tidal transcripts in the transcriptome was lower in B. azoricus regardless of the detection algorithm used (electronic supplementary material, figure S14a,b). Indeed, considering that rhythmic gene expression was organ-specific [5], the result was disputable as we detected overall rhythmicity within mixed tissues of R. leurokolos, whereas the study on B. azoricus only used the gills. We further examined and compared the strength of tidal rhythms in the two species by calculating the rhythmicity index using the autocorrelation analysis and the results suggested the general rhythmicity of tidal gene expression was greater in R. leurokolos (electronic supplementary material, figure S14c). This is unexpected because it should be easier for B. azoricus, living in a relatively distant area from the vent orifice with stable diffuse flow, to detect environmental factors influenced by tidal cycles. A possible explanation is that R. leurokolos needs to depend on a strong pacemaker to integrate all the periodic cues for predicting environmental fluctuations and preparing for transcriptional regulation, whereas B. azoricus could schedule the transcription by either relying on an endogenous clock or responding directly to the tidal-modulated environmental cues [22]. Why R. leurokolos inhabiting a more dynamic environment shows more robust rhythmicity compared with B. azoricus is an intriguing topic warranting future studies, although this could also be related to intrinsic differences between mussels and shrimps, as well as the influence of chemosymbiosis on the host [77].

(g). The potential evolutionary conservation between marine tidal rhythms and terrestrial approximately 12 h rhythms

We show that the deep-sea vent shrimp R. leurokolos exhibits robust endogenous rhythms related to tidal effects instead of daily cycles, making it a useful model for testing the hypothesis that marine tidal rhythms and approximately 12 h rhythms of terrestrial animals share conserved features. To investigate this, we compared the tidal gene expression of R. leurokolos with published, approximately 12 h, ultradian rhythms in the Bmal1 knockout mice liver of Mus musculus [28] and Drosophila S2 cells from the fruit fly D. melanogaster [29]—two systems without canonical circadian clock gene expressions. A total of 406 single-copy orthologous genes were identified as showing approximately 12 h rhythms in these three species. The vent shrimp Rimicaris shared more genes with the fruit fly compared with the mouse (54 genes versus 36 genes), agreeing with their closer phylogenetic relationship. We also found 21 genes with a period of approximately 12 h rhythms in all three species (figure 4a). These genes were enriched in the processes of DNA replication, regulation of transcription, ribosome biogenesis, protein folding and trans-Golgi network, which are also highly enriched in approximately 12 h rhythms of human beings ([78]; figure 4b). The heat shock protein 70 cochaperone Dnaja1, which was also reported in approximately 12 h rhythms of mammals [78] and tidal rhythms of the sea anemone Aiptasia diaphana [77], also showed a robust free-running tidal rhythm in R. leurokolos. These results support the evolutionary conservation of rhythmic gene expression at approximately 12 h related to the processes of central dogma information flow in both terrestrial and marine animals across a broad range of phyla, indicating that such approximately 12 h rhythms in animals probably initially evolved for adjusting to the tidal cycle.

Figure 4.

Homologous genes with a period of approximately 12 h rhythms between Rimicaris leurokolos, Mus musculus, and Drosophila melanogaster. (a) Venn diagram showing the comparison of approximately 12 h rhythms of single-copy orthologous genes between three species. (b) Pathway enrichment network showing biological functions of 21 conserved approximately 12 h genes in three species.

4. Conclusion

Overall, our analyses provide evidence for endogenous biological rhythms in the deep-sea hydrothermal vent shrimp R. leurokolos through a 72 h free-running experiment and constrain the possible zeitgeber for biological clocks ticking in the deep, shedding light on biological rhythms in the light-absent biosphere (figure 5). We propose that rhythms with a period of 24 h driven by the conventional circadian clock largely decrease or even disappear in adult deep-sea hydrothermal vent fauna. Instead, rhythmic gene expression coordinated with tidal cycles at approximately 12 h, which are probably controlled by an independent circatidal clock, evolved as a temporal adaptation to environmental fluctuations in hydrothermal ecosystems. Independent of solar energy but being influenced by tidal cycles, organisms in the deep-sea hydrothermal vent are probably powerful models for chronobiologists to decipher the evolution and molecular mechanisms of circatidal clocks and biological rhythms at approximately 12 h, important across a wide range of living organisms.

Figure 5.

A new perspective of biological rhythms in the dark biosphere. Terrestrial and shallow-water organisms largely rely on circadian clocks for anticipating periodic environmental changes such as the day–night transition. Terrestrial organisms such as human beings exhibit robust approximately 24 h rhythms driven by the circadian clock [2] and potential approximately 12 h ultradian rhythms (with an asterisk) that are assumed to be homologous with circatidal rhythms [79], though the biological functions and synchronization cues are still unclear. The intertidal mussel Mytilus californianus exhibits both circatidal and circadian rhythms, while the latter is dominant. The copepod Calanus finmarchicus also shows behavioural rhythms in diel vertical migration [80]. However, a different scenario can be found in the dark biosphere. Living away from direct sunlight, subterranean mole rats scarcely depend on circadian rhythms for survival and the endogenous circadian system is understood to be in the process of regression, even though still present [81]. Here, we found that in addition to the vent mussel Bathymodiolus azoricus [22], the vent shrimp Rimicaris leurokolos also shows strong endogenous rhythms in gene expression around 12 h, probably reflecting key impacts of tidal cycles on deep-sea hot vent organisms.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

Raw RNA-Seq data are available on NCBI as BioProject PRJNA993859.

Supplementary material is available online [82,83].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

H.Z.: data curation, formal analysis, methodology, visualization, writing—original draft; T.Y.: conceptualization, data curation, investigation, methodology, project administration, supervision; N.M.: conceptualization, methodology; C.C.: conceptualization, investigation, methodology, writing—review and editing; Q.J.: visualization; P.-Y.Q.: funding acquisition, project administration; J.S.: conceptualization, data curation, formal analysis, funding acquisition, project administration, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was supported by: Science and Technology Innovation Project of Laoshan Laboratory (LSKJ202203104); Fundamental Research Funds for the Central Universities (202172002 and 202241002); Young Taishan Scholars Program of Shandong Province (tsqn202103036); Natural Science Foundation of Shandong Province (ZR2023JQ014); Principal investigator projects of the Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou) (2021HJ01); Major Basic and Applied Research Projects of Guangdong Province (2019B030302004-04); Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou) (SMSEGL20SC01); Hong Kong Special Administrative Region government (16101822, C2013-22GF). The participation of Chong Chen in the R/V KAIYO research cruise KY14-01 was supported by an InterRidge Cruise Bursary.