Profiling molecular and behavioral circadian rhythms in the non-symbiotic sea anemone Nematostella vectensis

The behavioral rhythms ofwere studied by monitoring the locomotor activity of individuals using a tracking system, which was equipped with an infrared (IR) camera and time-controlled white LED illumination that can be set to different intensities (). In parallel with the locomotor activity tracking, some of the experiments were recorded with a video camera in order to enable visualization of the different movement patterns. Three major movement types were recorded (): head movement from side to side, body banding and constant peristaltic movement along the body axis. Behavioral rhythms were initially characterized over 3 days under 12 h light: 12 h dark (LD) conditions at 23 °C. Automated infrared tracking showed thatexhibits greater locomotor activity during the subjective night. Under LD conditions, during the night (ZT12–ZT24), the tested animals (n = 35) moved a total of 188.6 cm on average (standard error (SE) = 34.3), compared to only 84.5 cm on average (SE = 4.1) during light hours (ZT0–ZT12;). The averagedlocomotor activity peaked between four and nine hours after dark onset (ZT16–ZT21). Within this time period, the animals moved on average 101.9 cm (SE = 20.8) compared to only 16.3 (SE = 3.9) in the equivalent time period during light hours (ZT4–ZT9). Fourier analysis of the locomotor activity average ratios during the three days in LD conditions resulted in a single significant periodogram peak at 23.99 h (n = 35), indicating circadian frequency (in red). To normalize the differences in the absolute distance covered betweenindividuals that may originate from differences in size or metabolic rate, we have calculated the relative locomotor activity as a percentage of the maximum locomotor activity recorded for each animal (in LD the average relative locomotor activity ranged between a minimum of 11.2% during light hours and a maximum of 46% during dark hours,). Similar to the LD results, a single significant frequency peak at 23.98 h was identified in constant dark free-running conditions (DD; n = 20;in blue) with average relative locomotor activity ranging from 19.9% to 53.4% and following the same oscillation pattern as in LD (). These results support the existence of endogenous clock oscillator; however, in contrast with previous observation ¹⁶, our results didn’t show any significant circadian oscillation frequency during the constant light free-run (LL; n = 30;in green). Under LL conditions, the average relative locomotor activity ranged from 22.2% to 32.1% with no significant dominant frequency (). Nematostella Nematostella Nematostella Nematostella Fig. 1 Supplementary Video S1 Fig 2A Fig. 2B Fig. 3A Fig. 2B Fig 3B Fig. 2B Fig. 3C

As a first approach to study whetheroscillator exhibits temperature compensation, we inquired if the rhythms observed in LD conditions were also maintained at a lower temperature, we monitored the locomotor activity during three days under LD conditions at 18 °C (5 °C below the temperature of all other experiments). We observed a similar locomotor activity oscillation pattern as in the LD and DD experiments with locomotor activity ratios averages of 10.4% to 55.7% (). The tested approach may have some limitations associated with potential masking effects of light. Nevertheless, the obtained results suggest potential temperature compensation ofcircadian system. Nematostella Nematostella Fig. 3D

We tested the effect of 1 h dark and 1 h light pulses on the locomotor activity oscillation. The pulses were performed during normal LD conditions, while the effect was tested under DD free-run in order to prevent entrainment by a light cue. When a 1 h dark pulse was applied between ZT9 and ZT10 (2 h before the entrained dark onset), the oscillation phase was advanced and changes were observed in the cycle length. The observed locomotor activity cycle length (based on peak locomotor activity) was advanced by 2 h on the next day (ZT14), by 8 h on the second day after the pulse (ZT8) and by 6 h on the third day after the pulse (ZT10) (n = 11;). In contrast, a 1 h light pulse, between ZT21 and ZT22 (2 h before the entrained light onset) caused a complete disruption of the locomotor activity cycle during the following dark free-run for the rest of the experiment (). Due to the nature of the experimental system, prolonged behavioral monitoring was not possible, so it is not possible to determine whether the observed disruption is transient or permanent. Fig. 4A Fig. 4B

The casein kinase I (CK1) family consists of serine/threonine protein kinases, some of which are key regulators of circadian timing in bilaterian animals, fungi and green algae. CK1-like genes have previously been identified in bothandand were suggested as components of circadian gene network in these organisms. Reciprocal BLASTx searches of human andCK1 sequences against predicted proteins in theJGI genomic database revealed six CK1 family members inThree of theCK1 sequences grouped into a clade withas well as humanand(NvCK1_12115, NvCK1_12051, NvCK1_88486) Two others (NvCK1_159193 and NvCK1_161273) grouped withand humangenes, and the finalgene (NvCK1_192152) grouped with humanand(). ^{22 23} Supplementary Fig. S1 Acropora Nematostella Drosophila Nematostella Nematostella. Nematostella Drosophila Doubletime CK1δ CK1ε Drosophila CK1 CK1α Nematostella CK1γ1 CK1γ3

To investigate a potential role for CK1 activity in circadian function inwe characterized the effects of two specific pharmacological inhibitors of vertebrate CK1 activity on circadian behavioral rhythms in. One of these inhibitors (PF-4800567) specifically targets CK1δ. The second (PF-670462) inhibits both CK1δ and CK1ε and has been shown to disrupt behavioral rhythms in distantly related organisms, such as the green alga. The two inhibitors were tested at concentrations that have been shown to specifically inhibit circadian function in zebrafish. Nematostella, Nematostella Ostreococcus tauri ^{22 24}

Twelve hours prior to the initiation of the locomotor tracking,individuals were incubated in 1 μM of the pan-CK1δ/ε inhibitor or CK1δ-selective inhibitor. Over the next two days (48 h), locomotor activity tracking was performed under DD free-running conditions followed by inhibitor-free recovery of 1.5 days (36 h) under LD conditions. CK1δ/ε inhibitor-treatedlost their locomotor activity oscillation (n = 12,), while CK1δ inhibitor-treatedmaintained their original oscillation. (n = 12,). The locomotor activity oscillation of CK1δ/ε inhibitor-treatedwas successfully recovered after replacing the water with inhibitor-free water and changing the light conditions back to LD (). This suggests that one or more CK1 family members may be involved in the regulation of circadian behavior in Nematostella Nematostella Nematostella Nematostella Nematostella. Fig. 5A Fig. 5B Fig. 5A

To better understand the molecular forces that regulate the circadian locomotor activity rhythm inwe conducted transcriptional profiling using the Illumina HiSeq platform with samples collected every four hours over two days under LD conditions identical to those in the behavioral assay (BioProject accession number: PRJNA246707). Using Fourier analysis, the possible diel rhythmicity (i.e., 24-h periodicity) of all the genes was quantified, and the 180 transcripts exhibiting a g-factor >0.5 were further analyzed. Through K-means clustering, these transcripts were divided into 5 groups, each with a characteristic peak expression time. The 50 transcripts exhibiting the strongest diel rhythm are shown in, and expression data for all 180 genes are listed in; we subsequently refer to these as diel cycle genes (DCGs). Because these genes were identified based on their oscillations under LD conditions they were characterized as “diel control genes” (DCGs) rather than as “clock-controlled genes” (CCGs), which have been specifically demonstrated to maintain a cycle under constant conditions. Nematostella, Fig. 6 Supplementary Table S1

We annotated 143 of the DCGs through BLASTp-based searches of the SwissProt database. In addition, we identified putative homologues for 59% (22/37) of the unannotated genes through BLAST searches of thegenomic database. These may represent taxonomically restricted genes. GO terms were associated with 135 of the DCGs; however, none of these GO terms were statistically enriched in comparison with thetranscriptome. Acropora millepora Nematostella

In the present study,,andexhibited diel periodicity with similar timing of peak expression to that reported by Reitzel(). Specifically,expression peaked late in the day (ZT9-13),andpeaked during mid-day (ZT4-11 and ZT5-9, respectively), andpeaked during early morning or late night (ZT0-4 inand ZT21 in the present study). Bothand cryptochromes play central roles in regulating circadian cycles in bilaterians. Some cryptochromes are light sensitive and act to directly coupling the circadian clock with exogenous light cues. NvClock NvCry1a, NvCry1b NvCry2 et al. NvClock Cry1a Cry1b Cry2 Clock ¹⁵ Table 1 ^{15 25}

and the scleractinian coralare both anthozoan cnidarians, but they differ profoundly in terms of habitat and symbiont composition. Individualpolyps lack algal symbionts and colonize salt marsh environments, whileforms calcified colonies on tropical reefs through an obligate symbiosis for dinoflagellates. Genes with circadian expression patterns in both taxa are likely to serve fundamental roles in circadian physiology of cnidarians. Nematostella Acropora millepora Nematostella A. millepora

We first compared the set of 180DCGs with a set of CCGs that were identified from a previous microarray-based study of. Of note, thegenes exhibited daily oscillations both under LD conditions and under DD free-run. We mapped the differentially expressed microarray probes to 99 uniquetranscripts, 9 of which are putative homologs ofDCGs (). Among the shared genes (DCGs andCCGs) were two cryptochromes.andeach contain two Type I cryptochromes and one Type II cryptochrome. In each species, the Type II cryptochrome and one of the Type I cryptochromes exhibited diel oscillations (i.e., were identified as DCGs inand as CCGs in). TheType I and Type II cryptochromes have a similar oscillation pattern, which generally overlaps with aType I cryptochrome (), peaking at 12 pm (ZT18) but not withType II cryptochrome (), which peaks at 4 pm (ZT10;). Nematostella Acropora millepora A. millepora A. millepora Nematostella Nematostella A. millepora Nematostella A. millepora Nematostella A. millepora, Acropora Nematostella NvCry1a Nematostella NvCry2 ²⁰ Table 2 Fig. 7a Fig. 7a

Additional genes that exhibited diel oscillations in both species were two heat shock proteins (members of the Hsp70 and Hsp90 families) and, all of which act as chaperones to maintain correct protein folding. In, expression of these three genes peaked at 4 pm (ZT10), which was hypothesized to correspond to diel patterns of stress. In contrast, in, all three genes exhibited peak expression during subjective night (12 am, ZT17,). protein disulfide isomerase A. millepora Nematostella ²⁰ Fig. 7b

Four additional genes exhibited diel oscillations in bothand:, a heme-binding protein in the SOUL family, a high mobility group B protein (HMGB), and a transcript with no similarity to genes of known function. The unannotated gene did exhibit significant similarity (40–50% amino acid identity, e-values around 1 × 10) to predicted proteins of unknown function from diverse metazoans.exhibited a strong diel expression pattern in both species, with peak expression at noon forand 8 am for. Shoguchishowed thatfalls within a clade of basic-helix-loop-helix transcription factors that contain an orange domain (bHLH-O). The SOUL family member exhibited similar expression in both species (daytime maxima), but the HGMB and unannotated gene did not. Nematostella A. millepora Hes/Hey-like Hes/Hey-like Nematostella A. millepora et al. Nematostella Hes/Hey-like −40 ²⁶

Thetranscripts that exhibited diel oscillations in expression were also compared with an Illumina-based study of gene expression during the day and night inlarvae. Of the 180genes with diel oscillations in expression, we identified putative homologues of 108 genes within thedata set. Six of thesetranscripts exhibited ≥3-fold higher expression during the night, and eight exhibited ≥3-fold higher expression during the day (). Three of these genes (also exhibited circadian expression patterns in themicroarray. In larvae, bothandwere expressed at higher levels during the day, as they were in the microarray study of adult corals.was also expressed most highly during the day in coral larvae. Comparison of the larval dataset with theDCGs revealed additional shared genes that were not identified in the microarray study. For example,expression in larvae was about four times higher during the day compared with the night. A putative homolog of theexhibited greatly elevated expression during the night inlarvae and also exhibited peak expression during the night inCIPC is a mammalian protein that regulates period length by forming complexes with CLOCK, leading to enhanced phosphorylation and degradation. Althoughwas initially described as absent from invertebrates, similar predicted protein sequences are present in urchins and molluscs (e.g., XP005109657and XP800566). It is unknown whether the CIPC-like protein fromor other invertebrates forms complexes with CLOCK and/or performs a circadian function. Interestingly, in contrast to adultthe larvae did not show significant (≥3-fold) transcription change between day and night in any chaperone homologues, this may be related to the fact that in the larvae was sampled only once during the daytime and once during the nighttime (ZT10 and ZT22, respectively). Nematostella Acropora millepora Nematostella A. millepora A. millepora AmCry1, AmCry2, Hes/Hey-like) A. millepora AmCry1 AmCry2 Hes/Hey-like Nematostella Clock clock-interacting circadian pacemaker (CIPC) A. millepora Nematostella. CIPC Aplysia californica Strongylocentrotus purpuratus Nematostella A. millepora, ¹⁹ Table 3 ^{27 28}

Two transcripts belonging to the superfamily of cytochrome P450 mono-oxygenases (CYPs) exhibited diel periodicity in expression, but with a 4–8 hour phase-shift from one another (). The difference in timing might indicate that the two enzymes catalyze different steps within a metabolic pathway, producing metabolites that cycle out of phase with one another. In a phylogenetic analysis of animal CYPs, these twoCYPs fell into a clade that included the vertebrate steroidogenicandgenes. Synthesis of vertebrate-type steroids requires side-chain cleavage of cholesterol by the vertebrate-specific CYP11; CYP17 and CYP21 then act catalyze downstream steps in the synthesis of sex steroids and corticosteroids. Several mammalian CYPs, including, exhibit circadian oscillations in expression, which result in daily cycles in cholesterol homeostasis and hormone concentrations. While the substrate of the-like genes is unknown, mammalian CYP17 is able to metabolize a variety of substrates including the steroid precursor squalling. Fig. 7c ^{29 30 31 32 33} Nematostella Cyp17 Cyp21 Cyp17 Nematostella Cyp17

Severaltranscripts exhibited strong diel oscillations that had not previously been implicated in cnidarian circadian signaling. Among these, a transcript (NV_200090) similar to() exhibited strong cycling with peak expression at night (,). In vertebrates QN1 helps to regulate the cell cycle during retinal development and serves a motor protein during mitosis. Of the 50 transcripts oscillating with the strongest diel periodicity (,), four were collagen family members. Collagen transcripts undergo circadian cycles in mammalian cartilage, but rhythms in collagen expression have not been previously identified in cnidarians. Also of note, many of these strongly oscillating genes (7 of 50), exhibited no significant similarity to annotated genes, or could only be weakly annotated as possessing a conserved domain (e.g, LEM superfamily member, GIY-YIG superfamily member, ARID domain-containing protein). Clearly a great deal remains to be learned regarding the function of these cyclic genes. Nematostella QN1 Centrosomal protein quail neuroretina 1 Supplementary Table S1 Fig 6 ³⁴ Supplementary Table S1 Fig. 6 ³⁵

K-means clustering demonstrated that distinct groups of DCGs exhibit peak expression throughout the day and night. As previously mentioned, three chaperone proteins exhibited peak expression during subjective night (, top cluster;, cluster 1). Beyond this grouping, genes with similar apparent functions did not necessarily cluster together. For example the genes identified as likely circadian regulators (Clock, CIPC, Cryptochromes, Hes/Hey-like) are distributed broadly among clusters. Because circadian regulation is characterized by feedback from intersecting transcriptional/translational loops, it makes sense that expression patterns of regulatory components will be offset. The four collagen-like DCGs were distributed among three expression clusters. While the reason for this offset is unknown, it’s possible that serial expression of different collagen forms helps to stabilize total collagen levels or that the different forms are necessary for specific components that are produced during on a daily cycle. Fig. 6 Supplementary Table S1

Laboratory-bredwere maintained in plastic containers with one-third strength artificial sea water (33% ASW, Reef crystals) at 18 °C under a 12 : 12 h (7 am–7 pm/7 pm–7 am) LD cycle. Animals were fed five times per week with freshly-hatched brine shrimp, and water was renewed weekly. Animals were gradually acclimated to 23 °C and starved for two days prior to behavioral experiments and transcriptional profiling. Nematostella

Locomotor activity of individual Nematostella were monitored using two Noldus DanioVision XT tracking devices, each equipped with an IR camera and white LED illumination that can be set to different intensities and LD cycles (). The data collection and analysis were carried out by EthoVision XT8 video tracking software (Noldus information technology, Wageningen, Netherlands). Animals were isolated in wells of six-well plates, each of which was manually defined as a tracking ‘arena’ in the EthoVision software. Center-point detection with gray scaling (detection range of 25–77, contour erosion of 1 pixel, high pixel smoothing) was used to monitor movements, which were calculated according to the change in position of the average center pixel each second (). Fig. 1 Fig. 1

Illumination was provided within the DanioVision tracking device by the integral white LED light with an intensity of 200 (+/−10) lux (25% of its maximum intensity) and did not significantly affect the experimental temperature (23 °C). When needed (as for the 18 °C experiment), a chiller pump was used to keep the water temperature fixed during the duration of the experiment. The illumination cycles were the same as used for culturing (12 : 12 h LD). Since this is the first application of this tracking system to measurement of sea anemone movements, we tested the system background noise using measurements of six immobilized (paralyzed with MgCl)individuals for 1 h. The recorded movement in this test was less than 1 cm, and was considered as insignificant background noise (compared with the average movement of the non-paralyzed animals). Parameters were optimized to ensure that organisms were detected throughout the entire observation period. 2 Nematostella

The total distance moved was summed in hourly bins and expressed as a percentage of the maximum hourly distance measured for each individual. The average and standard errors were calculated for all tested animals based on the normalized values of each hour. The oscillation frequencies were evaluated based on the average values of each experiment using Fourier analysis, as previously described. ⁵⁷

individuals were monitored in 6-well plates containing one-third strength ASW with one of two casein kinase inhibitors; the pan-CK1δ/ε inhibitor PF-670462 or the CK1δ-selective inhibitor PF-4800567 (Pfizer Global Research and Development, TOCRIS Bioscience) dissolved in DMSO. In order to determine the effective concentrations, we performed an initial toxicity assay based on the range tested by Smadja Storz. We tested the viability of the animals based on response to mechanical touch 1, 3 and 10 days after adding the inhibitors to the water in the six-well plates to final concentrations of 0.1, 1 and 10 μM (n = 6). Allindividuals survived up to 10 days after incubation in 0.1 and 1 μM of both inhibitors, but all died 10 days after incubation in 10 μM concentration of either inhibitor. Based on these results, all inhibition experiments were conducted in final concentrations of 1 μM. All controls were treated with identical concentrations of DMSO (0.05%). Nematostella et al. Nematostella ²⁴

We used RNA-seq technology to identify diel cycle genes (DCGs) infollowing an experimental design previously used in circadian studies of the coral. Anemones were acclimated and maintained during the experiment inside the Noldus DanioVision XT tracking device under identical light (LD) and temperature conditions as in the behavioral assay. Five anemones were sampled every 4 h over two consecutive days, starting at 8 am. Total RNA was extracted from pools of five individuals using the Qiagen RNeasy Mini Kit. The Illumina TruSeq protocol was used to prepare libraries from the RNA samples. We performed one biological replicate by constructing and sequencing two Illumina libraries from different samples of five animals collected at same time point (the second time point, 12pm). The libraries were multiplexed on 2 lanes of an Illumina HiSeq2000. On average, ~15 million 50 base-pair paired-end reads were obtained for each library. The data was deposited as an SRA BioProject (accession number: PRJNA246707). Reads were aligned to thegenomeusing TopHat. Only reads that uniquely aligned to protein coding regions with up to two mismatches were retained. Thegene information was downloaded from Joint Genome Institute database (). A custom Perl script was used to parse the output from TopHat (Sequence Alignment/Map (SAM) format) and to convert it into raw number of reads aligned to each position in eachgene. The dataset was de-duplicated to remove multiple reads with identical start positions in the genome, as these might represent PCR artifacts. Library quality was assessed in comparison with a benchmark library described by Levin and colleagues. All library quality parameters met the benchmark standards, including mapping of reads to unique genome start sites and evenness in expressed gene coverage. Nematostella Acropora millepora Nematostella Nematostella Nematostella ^{20 11 58 59 59} http://genome.jgi-psf.org/Nemve1/Nemve1.info.html↗

We tested the effect of biological variation by comparing two libraries derived from different anemones collected at the same time and light condition. The differences between the samples are close to the expected technical noise (96% of the genes are within the expected 99%-region of Poisson noise), as described recently for miRNA-seq multiplexing. ⁶⁰

The logarithmically-transformed gene expression values were normalized using a modification of the TMM method, in which the mRNA profiles were scaled such that the log-fold changes of all the mRNAs are distributed around zero (after trimming the higher and lower quartiles of the log-fold changes). The scaling factor was thus set so that the trimmed mean of log-folds vanish. The mean was weighted using the inverse standard deviation, as estimated from Poisson distribution of counts. ^{61 60}

The time-dependent signal was converted into a frequency-dependent signal using the Fast Fourier Transform (FFT). We used in-house scripts that were previously found to be accurate in detecting circadian genes, as attested by ~90% true positive rate in independent validation experiments (). The extent to which the original signal contains a 24-hr rhythm was quantified by the ratio (‘g-factor’) of the power (squared amplitude) of the frequency which corresponds to a 24-h period, to the sum of powers of all frequencies. The higher the g-factor, the higher is the confidence that the transcript exhibits a diel rhythm. Changing the definition of the g-factor by adding the powers of higher harmonics of the 24-h period to the numerator, gave similar results compared to the use of the definition above. The genes with the highest g-factor (g-factor greater than 0.5 was used as a cutoff) were sorted into five clusters with similar temporal expression patterns using a K-means clustering, implemented in Matlab as described by Levy. ^{20 62 20} et al.

Functional annotation oftranscripts, including predicted homologs within the Swissprot database and from the transcriptome of the coralwere downloaded from the Joint Genome Institute database. Annotations were manually curated for genes exhibiting strong diel periodicity in their expression patterns (50 genes with highest g-factor) and those identified through our comparative analysis (see below). Manual curation was based on BLASTp searches of the Swissprot and NR databases and, in a few cases, published phylogenetic analyses (cryptochromes,/). Nematostella Acropora millepora Hes Hey-like

We compared the set of DCGs identified in our study with genes exhibiting circadian expression patterns or strong day/night differences in two published studies of the coralBradyused Illumina-based transcriptional profiling to compare gene expression between coral larvae collected during day and night (12 : 12 h LD cycle, samples collected 10 hours after lights on (ZT10) and 10 hours after lights off (ZT22)). They reported the number of counts and fold change, but did not provide any further statistical analysis. From the 47,666 transcripts that they identified, we selected the 10,294 genes that exhibited a three-fold difference in expression between the day and night and identified potential homologs of the putative DCGs fromLevyused an experimental design similar to the present study:colonies were sampled every 4 hours over 2 days under LD and DD conditions. They conducted expression profiling using a cDNA microarray. We selected 200 genes exhibiting the strongest circadian expression patterns (g-factor > 0.6468), identified the associated probe sequences in the NCBI Gene Expression Omnibus (GEO) database (Platform), and annotated them using BLAST searches of alarval transcriptome database hosted on SymBioSys (http: sequoia.ucmerced.edu/SymBioSys). We then identified potential homologs among the putative DCGs from. Acropora millepora. et al. Nematostella. et al. Acropora millepora Acropora millepora Nematostella ^{19 20} GPL6941