Circadian clock-controlled gene expression in co-cultured, mat-forming cyanobacteria

Natural coastal microbial mat communities are multi-species assemblages that experience fluctuating environmental conditions and are shaped by resource competition as well as by cooperation. Laboratory studies rarely address the natural complexity of microbial communities but are usually limited to homogeneous mono-cultures of key species grown in liquid media. The mat-forming filamentous cyanobacteria Lyngbya aestuarii and Coleofasciculus chthonoplastes were cultured under different conditions to investigate the expression of circadian clock genes and genes that are under their control. The cyanobacteria were grown in liquid medium or on a solid substrate (glass beads) as mono- or as co-cultures under a light–dark regime and subsequently transferred to continuous light. TaqMan-probe based qPCR assays were used to quantify the expression of the circadian clock genes kaiA, kaiB, and kaiC, and of four genes that are under control of the circadian clock: psbA, nifH, ftsZ, and prx. Expression of kaiABC was influenced by co-culturing the cyanobacteria and whether grown in liquid media or on a solid substrate. Free-running (i.e. under continuous light) expression cycle of the circadian clock genes was observed in L. aestuarii but not in C. chthonoplastes. In the former organism, maximum expression of psbA and nifH occurred temporally separated and independent of the light regime, although the peak shifted in time when the culture was transferred to continuous illumination. Although functionally similar, both species of cyanobacteria displayed different 24-h transcriptional patterns in response to the experimental treatments, suggesting that their circadian clocks have adapted to different life strategies adopted by these mat-forming cyanobacteria.


Results
total RnA and qpcR statistics. The extracted total RNA was of good quality (average RIN: 7.7 ± 1) and yielded higher concentrations from C. chthonoplastes than from L. aestuarii (Table 1). Co-cultures yielded less total RNA than mono-cultures. Similarly, biofilm-grown cultures yielded less total RNA when compared to liquid cultures ( Table 1).
The qPCR-derived Ct values of the biological replicates of the targeted circadian clock gene cluster kaiABC, nitrogenase gene nifH, cell division protein gene ftsZ, photosynthesis D1 protein gene psbA, peroxiredoxin gene prx and two commonly used cyanobacterial housekeeping genes, rnpA and ppc ranged from 23 to 39 in L. aestuarii and from 21 to 38 in C. chthonoplastes. The average r 2 and amplification efficiencies of the qPCR assays were 0.95 and 100% in L. aestuarii and 0.92 and 92% in C. chthonoplastes (Table 2). Regardless the taxon and treatment, the highest absolute transcription was observed for the photosystem II protein D1 coding gene psbA with on average 153,425 gene-copies/µl in L. aestuarii and 78,592 gene-copies/µl in C. chthonoplastes.
Validation of housekeeping genes. Each qPCR run was accompanied by the quantitation of housekeeping genes for normalization. The genes rnpA (protein component of ribonuclease P) and ppc (phosphoenolpyru- Table 1. Average quantity (ng/µl) and standard deviations (SD) and the average quality (RIN) of total RNA extracted from L. aestuarii (LA) and C. chthonoplastes (CC). Average quantities and qualities are given for each treatment (L = liquid, B = immobilized, M = mono-culture, C = co-culture).  Table 2. RT-qPCR primer (forward (F), reverse (R)) and TaqMan-probe (P) information for C. chthonoplastes and L. aestuarii. Amplification efficiency and r 2 depict average values. All qPCR assays were run using an annealing temperature of 64 °C. *Housekeeping gene.    Table 4) and explained ~ 81% of the observed variance in their expression. Unexpectedly, only kaiA was significantly influenced by the light regime (Χ 2 = 0.017, p = 0.02) ( Table 4). Potential rhythmic expression was deduced from comparing and fitting the expression patterns with a sine curve using the circadian cycle predictor program MetaCycle 29 in accordance with a previous study in natural microbial mats 30 Figure  S2). Peak expression of kaiB varied with whether the organisms were co-cultured or not and ranged from ZT9 (ML, CL and MB) to ZT4.5 (CB). The highest peak expression measured for kaiB was observed in biofilmgrown co-cultures (    Table 4). Cultures growing under a light-dark regime revealed a sinusoidal pattern of psbA gene expression (p = 0.007-0.02) (Fig. 3a, b, Supplementary Figure  S2) with the exception of CB cultures. Highest psbA expression was found during the beginning of the light period (ZT4.5-ZT9) followed by a sharp drop in expression in liquid cultures at ZT14.5 ( Fig. 3a, b, Supplementary Figure S1A). In biofilm-grown cultures the overall expression level of psbA was lower than in the liquid  Figure S2). The sinusoidal pattern was maintained under continuous illumination for CL and MB conditions biofilm-grown (p = 0.01-0.04) (Fig. 3a, b, Supplementary Figure S2).  Figure S1A). Moreover, nifH expression was significantly higher in liquid cultures compared to the biofilm-grown cultures. A significant sinusoidal curve under LD illumination was observed under all conditions (p = 0.005-0.01) except for the cocultured biofilm-grown cultures (CB) (Fig. 3a, Supplementary Figure S2). The imposed light regime significantly affected the daily transcription pattern of nifH (Χ 2 = 0.296, p = 0.001) (Fig. 1, Table 4). Under continuous light, the pattern of expression in the liquid cultures peaked more often than in biofilm-grown cultures with high expression at CT0, CT4.5 and CT21. (Fig. 3b, Supplementary Figure S1A). In biofilm-grown mono-cultures nifH expression peaked at CT14.5 and CT21 and in co-cultures at CT17. 45. No significant sinusoidal nifH gene expression was found under continuous illumination.
ftsZ expression in L. aestuarii. The variation in expression of ftsZ in L. aestuarii during a 24-h period was large and only in co-cultured biofilm-grown cultures at LD illumination a sinusoidal like curve was observed with a maximum at ZT0 (p = 0.03) (Fig. 3a, Supplementary Figure S2). In contrast, under continuous illumination an opposite pattern is found with all the three other growth conditions providing significant sinusoidal ftsZ expression (p = 0.008-0.01) (Fig. 3b, Supplementary Figure S1A). Due to other more variable patterns, each of  (Fig. 1, Table 4).  Table 4). Furthermore, the expression of kaiA and kaiC differed when the organism was grown as a mono-culture or co-cultured with L. aestuarii (mono-vs. co-culture: Χ 2 = 0.053-0.077, p = 0.001), while the way of culturing significantly impacted on the expression of kaiB and kaiC during a 24-h period (liquid-vs. biofilm-grown-culture: Χ 2 = 0.040-0.833, p = 0.001-0.009) (Fig. 2, Table 4). The expression of kaiA and kaiB did not seem to follow a particular pattern and only in liquid co-cultures a potential sinusoidal curve was discerned for kaiA expression (p = 0.005) and in biofilm-grown mono-cultures for kaiB (p = 0.03) (Fig. 4a, Supplementary Figure S1). Peak expression for kaiA in liquid co-cultures and biofilmgrown cultures was at the beginning of the light period (ZT0) and at the end of the dark period (ZT21) and peaked at ZT0 and ZT9 in liquid mono-cultures (Fig. 4a). A sinusoidal distribution was also observed for kaiC expression of cells grown in ML and CB conditions with respective highest expression at ZT0 and ZT4.5 (p = 0.01) (Fig. 4a, Supplementary Figure S2).
Under continuous illumination, kaiA expression was sinusoidal at LL-MB (p = 0.01) and kaiB at LL-CB (p = 0.003) with peak expression at CT17.5 and CT14.5 (Fig. 4b, Supplementary Figure S2). Potential rhythmic expression of kaiC under LL conditions was not observed. www.nature.com/scientificreports/ prx expression in c. chthonoplastes. The expression pattern of prx in C. chthonoplastes differed significantly between a mono-culture and co-culture with L. aestuarii (Χ 2 = 0.263, p = 0.001), and between a liquid medium and biofilm-grown culture (Χ 2 = 0.053, p = 0.002), as well as upon transfer to continuous light (Χ 2 = 0.022, p = 0.007) (Fig. 2, Table 4). Overall, the expression of prx was 5-(LL) to tenfold (LD) higher in biofilm-grown than in liquid culture (Fig. 4a, b, Supplementary Figure S1B). Expression of prx under standard conditions (ML-LD) gradually increased from the beginning of light to the dark period while an almost opposite trend was observed in MB cultures with highest expression at time points ZT0 and ZT21. A sinusoidal pattern was observed in the co-cultures and in biofilm-grown mono-cultures independent of the light regimes tested (p = 0.006-0.04) (Fig. 4a, b, Supplementary Figure S2).
psbA expression in c. chthonoplastes. Expression of psbA in C. chthonoplastes was significantly influenced by two of the growth conditions (mono-/co-culture: Χ 2 = 0.029, p = 0.029, liquid vs. biofilm-grown: Χ 2 = 0.028, p = 0.031) (Fig. 2, Table 4). A sinusoidal pattern was observed for psbA under CB conditions at LD illumination (p = 0.01) and under CL and MB conditions under continuous illumination (p = 0.003-0.01) (Fig. 4a, b, Supplementary Figure S2). The expression of psbA in LD-grown cultures peaked during the early dark to early light period (ZT17.45-ZT0) (Fig. 4a, Supplementary Figure S1B). Under continuous illumination, co-cultures and MB cultures revealed peaks in the expression of psbA between CT14.5 and CT21, while the ML culture was characterized by high psbA expression at CT0 (Fig. 4b, Supplementary Figure S1B).  Table 4). Relative nifH expression in biofilm-grown, co-culturegrown and LL-grown cultures displayed less fluctuation than displayed in liquid-grown, mono-culture-grown and LD-grown cultures. Expression of nifH in C. chthonoplastes was overall low with a single peak at the late dark/early light period in liquid cultures (ZT21 (ML), ZT0 (CL)) and during the light period in biofilm-grown cultures (ZT9 (MB), ZT4.5 (CB)) ( Fig. 4a, Supplementary Figure S1B). In the liquid mono-culture at LL a single peak in the expression of nifH was observed at CT0 (Fig. 4b, Supplementary Figure S1B). No significant sinusoidal patterns were found for C. chthonoplastes nifH expression under any of the tested conditions. ftsZ expression in c. chthonoplastes. Expression of the cell division protein coding gene ftsZ was significantly affected whether co-cultured or not (Χ 2 = 0.122, p = 0.001) and liquid-or biofilm-grown culture (Χ 2 = 0.025, p = 0.016) (Fig. 2, Table 4). The patterns of expression of ftsZ during a 24-h period varied. A sinusoidal curve was only observed in mono-cultures (p = 0.006-0.01) and at LL-CL and LL-MB (p = 0.008-0.04) (Fig. 4a, b, Supplementary Figure S2) with peak expression during the late dark and early light period in LDgrown cultures and at same time points in LL-grown cultures. (Fig. 4a, b, Supplementary Figure S1B).

Discussion the circadian clock of L. aestuarii and C. chthonoplastes. Rhythmic gene expression in cyanobac-
teria involves the following three steps: (1) the external cues, mostly light, temperature and redox potential that (2) entrain the "clock" proteins that control the rhythmic expression of (3) the functional protein coding genes involved in processes that are separated in time.
The only controlled rhythmic cue in our experimental set-up was the illumination period while temperature was kept the same in all experiments. Given the large number of variables tested within one cycle and the need for at least 6 time points per cycle, we were not able to assay over a 48 h or longer period that would be needed to determine truly circadian control. However, the observed significant differences in expression patterns are sufficient to infer that circadian control may also be affected by the greater complexity of natural conditions. Both the way of culturing (liquid or biofilm-grown) as well as whether the cyanobacteria were co-or monocultured affected the expression of the circadian clock genes kaiA, kaiB, and kaiC in both species, albeit in different ways.
The maximum expression of the circadian clock genes kaiABC of L. aestuarii shifted 4-8 h between liquid and biofilm-grown cultures. Shifting the cultures to continuous illumination for 2 days, from which samples were taken on the second day, had little influence on the variation of expression of the three kai genes over a 24-h period, which is indicative of the free-running cyanobacterial clock as has been previously shown for unicellular cyanobacteria 31,32 . In contrast, the cultures of C. chthonoplastes did not display a free-running clock. Under continuous light, maximum gene expression of kaiABC in this organism shifted in regard to what has been found under a dark-light regime when grown in biofilm-grown cultures and co-culture.
Based on the aforementioned results we suggest that, despite being both members of the order of Oscillatoriales, the two species display major differences in how stringent the control over their circadian clocks is. These differences may be attributed to the life strategies of the two cyanobacteria regarding their response to high light. For example, C. chthonoplastes avoids light stress by migration 33,34 , which may trigger several transcriptional responses in line with a metabolism that is not optimally adapted to continuous high light intensities. In contrast, Lyngbya only moves when differentiating into hormogonia: short, sheet-less motile trichomes 35 , and therefore cannot rapidly avoid the imposed light stress. However, Lyngbya species evolved strategies to endure high light, desiccation and UV-stress. To avoid the latter, Lyngbya species synthesize UV-screening compounds such as mycosporine amino acid-like substances (MAAs) and carotenoids, in addition to coiling of their trichomes 36  www.nature.com/scientificreports/ to an 'hourglass' clock, which has been described for kaiA-lacking species such as Prochlorococcus marinus 37 . However, since C. chthonoplastes does have, and expresses the kaiA gene, the differences in expression with L. aestuarii may reflect specific adaptation to their natural habitat. In support of this assumption is the low identity between kaiA nucleotide and protein sequences (< 60% identity) of L. aestuarii and C. chthonoplastes (and also compared to the kaiA sequence of S. elongatus), which is also the lowest in comparison to kaiB and kaiC (> 75% identity), suggesting that kaiA is the least conserved gene and may experience the lowest selective pressure among the kai genes (Supplementary Table S1, Supplementary Figure S3). Even though the two species co-occur in many marine microbial mats, L. aestuarii and C. chthonoplastes may occupy different ecological niches that allow them to share light and micronutrients using differently tuned circadian clocks.
A kaiABC-independent circadian oscillator has been linked to the periodicity of peroxiredoxin which follows the generation of stress-induced reactive oxygen species 25,38 . Overall, the expression patterns of prx reveal a significant sinusoidal curve over the 24-h sampling period for most conditions except the standard liquid mono-cultures in C. chthonoplastes and under some conditions in L. aestuarii. However, the expression levels of prx shifted significantly between mono-and co-cultures of liquid (L. aestuarii) and biofilm-grown (C. chthonoplastes) cultures. Highest prx expression was found in mono-cultures in the dark to early light period. This was also found in Crocosphaera watsonii strain WH8501 and it was attributed to the accumulation of reactive oxygen species (ROS) towards the end of the light period 39 . In contrast to mono-cultures, prx expression in co-cultures of L. aestuarii and C. chthonoplastes displayed maxima towards the late light period. This would suggest that the accumulation of ROS occurs faster in co-cultures as the result of a higher raise of oxygen concentration due to the combined photosynthetic activity. Robust free-running prx expression 25 without a peak shift was only found in co-cultures of liquid (C. chthonoplastes) and biofilm-grown (L. aestuarii) cultures under continuous illumination. Regardless of the illumination regime and treatment, average prx levels of C. chthonoplastes were more than a 100-fold higher than in cultures of L. aestuarii and showed a tenfold increase in biofilm-grown cultures in comparison to liquid cultures. The large transcriptional discrepancy between the species' prx levels and between C. chthonoplastes' liquid and biofilm-grown cultures suggests a higher sensitivity of C. chthonoplastes and especially of its biofilm-grown cultures, to oxidative stress and explains its migratory behaviour to avoid high light intensities. the expression of circadian clock regulated genes in mono-and co-cultures grown in free-living or biofilm-grown mode. Depending on growth conditions, expression of psbA in C. chthonoplastes and L. aestuarii, and nifH in L. aestuarii followed a sinusoidal pattern during the 24-h period. In the unicellular cyanobacteria Cyanothece sp., expression of psbA and nifH is under control of the kaiABC regulatory network 40 similar as for psbA expression in S. elongatus 41 . The peak in expression of psbA in the early light period and of nifH in the dark is typical for non-heterocystous, circadian clock controlled cyanobacteria in which nitrogenase activity is confined to the low light periods where photosynthetically produced oxygen is low 42,43 . Remarkably, we did not find such distinction between the peaks of psbA and nifH expression in L. aestuarii grown as a biofilm-grown co-cultured with C. chthonoplastes. Instead, both psbA and nifH peaked during the dark. It is unclear why the expression of psbA shifted to the dark in biofilm-grown co-cultures. Moreover, under LD conditions psbA and nifH peak transcription levels in L. aestuarii were up to 11 times lower in biofilm-grown cultures compared to liquid cultures. Albeit less distinct, the same was seen in C. chthonoplastes. The lower expression of psbA in biofilm-grown cultures may have been caused by the lower exposure to light of the biofilm-grown trichomes or by a decreased rate of growth as has been observed in aggregated cells in biofilms of Escherichia coli 44 . However, the latter explanation was not supported by lower transcription levels of the cell division gene ftsZ in biofilm-grown cultures. Alternatively, as is the case in S. elongatus, reactive oxygen species may target the de novo synthesis of psbA mRNA 45,46 decreasing the rate of photosynthesis 47 . This explanation is supported by the increase of prx expression in biofilm-grown cultures that is possibly a response to an increase in ROS.
When L. aestuarii was transferred to continuous illumination the expression peak of psbA shifted 4-8 h later during the 24-h cycle. In contrast, C. chthonoplastes, which does not display a free-running circadian rhythm, the peak of the expression of psbA did not shift when the culture was transferred to constant illumination. These results challenge the degree of control of the circadian clock on the expression of psbA in these cyanobacteria and call for another controlling mechanism.
The measurement of expression of nifH in C. chthonoplastes was hindered by technical difficulties in the qPCR reaction yielding in some runs false positive reactions above the threshold level in the non-template controls, which was attributed to a known reagent contamination problem 48 . However, this problem is mostly neglectable when large amounts of species specific nifH containing DNA is added as template. With some caution we conclude that nifH expression in C. chthonoplastes revealed low transcript numbers and lack of temporal separation from psbA expression. This is unusual for non-heterocystous cyanobacteria. Moreover, it is still uncertain whether C. chthonoplastes actually fixes atmospheric dinitrogen since it has not yet been possible to grow the species diazotrophically in culture 49 . Although this cyanobacterium possesses the nifHDK gene cluster 50 , it lacks the accessory genes nifOTWXZ, which are present in, for instance, L. aestuarii 51 . The function of these accessory nif genes is not well understood and it is therefore not certain whether they are indispensable for N 2 fixation. Daily patterns of nifH expression in biofilm-grown C. chthonoplastes cultures are similar to those observed in a coastal microbial mat 52 as well as metatranscriptomes from the same mat 30 . These studies report that nifH expression is high in the beginning and/ or middle of the light period.
In conclusion, comparison of gene expression patterns between laboratory cultures and field samples are rare and may be contradictory. The daily variations of gene expression obtained from the metatranscriptome of oceanic surface water containing abundant picocyanobacteria (Synechococcus) were similar to those obtained from liquid cultures 53  www.nature.com/scientificreports/ these planktonic species than would be the case for benthic microorganisms. The transcriptomes of Salinibacter ruber and Haloquadratum walsbyi reveal large differences in relative expression levels in a subset of genes while relative expression levels of other genes are well conserved in the metatranscriptomes from natural samples in which these organisms thrive (Bolhuis, unpublished data).
Our study shows that gene expression in two filamentous non-heterocystous cyanobacteria respond differently to external stimuli and potential zeitgebers. Phase shifts as well as changes in expression levels occurred as a function of the mode of growth (liquid medium versus biofilm-grown culture) and to the presence of the other competitor species. The observed species-specific expression patterns most likely reflect different life strategies, hinted at by differences in expression of prx, psbA, and nifH. In this study, neither co-or mono-cultivation nor liquid or biofilm-grown growth could be held as main responsible effector for the observed differences in gene expression patterns. However, in combination these effectors are able to alter gene expression not only in the laboratory but also in nature. As a consequence, most laboratory derived cyanobacterial circadian clock (controlled) expression patterns are insufficient predictors for expression patterns in the field. This emphasizes the need to mimic the natural environment when testing bacteria in the laboratory.
Material and methods experimental setup. Stock cultures of the filamentous, non-heterocystous cyanobacteria Lyngbya aestuarii PCC8106 (synonym = Lyngbya sp. CCY9616) and Coleofasciculus (Microcoleus) chthonoplastes PCC7420 (CCY9604) served as inoculum for the experiments. These cyanobacteria were isolated from similar coastal microbial mats but at different times and at a large geographical distance (Mellum, Germany (L. aestuarii) and Woods Hole, USA (C. chthonoplastes)). Although the two species, both members of the order Oscillatoriales, are naturally co-occurring in coastal microbial mats and in many other similar habitats 54-57 , the isolates tested here did not have any previous interaction other than the three weeks of co-culturing for the experiments described here. The cyanobacteria were grown at 23 °C under continuous fluorescent light (L: photon density: 40 µmol m −2 s −1 ) in 40 ml culture flasks (TPP, Switzerland) containing 25 ml of BA + medium (1:1 mix of BG11 and ASN 3) (www.dsmz.de) with added nitrate ( +) (13.29 mM). The culturing was done in triplicate. Liquid cultures (L) were grown in the 40 ml TP culture flasks. Biofilm-grown cultures (B) were grown in 6-well culture plates equipped with filter trays (0.4 µm pore size) (NUNC™, Denmark), which were filled with sterile glass beads (3.5 g/well) (ø 0.1 mm) and 4 ml/well of BA + medium (which did not submerse the beads). Sterile tweezers and disposable sterile Pasteur pipettes (VWR, USA) were used to distribute equally sized pieces of cyanobacterial trichomes on the glass beads in order to obtain an even distribution. Co-culturing (C) was done by inoculating similar sized filaments of both species (1:1 ratio) in culture flasks or by placing their trichomes on the glass bead surface. The cultures were grown for three weeks under a 16 h light and 8 h dark cycle (LD) to entrain the circadian clock. Two days prior to sampling, half of the LD cultures of each treatment was exposed to constant light (LL) in order to test the free-running characteristics of the circadian clock. During the three-week culturing period, the cyanobacteria did not show signs of degradation or bleaching. Evaporation of medium during culturing was compensated by refilling wells and culture flasks with autoclaved MilliQ water. Cultures transferred to continuous light (LL) were sampled at the corresponding time points of the LD series but were labelled as Circadian Time (CT). Samples were taken in biological triplicates and immediately submerged in liquid nitrogen in a sterile mortar and ground with a sterile pestle. From the liquid cultures an equal number of subsamples were taken and from the biofilm-grown samples a complete well with cell material was used for RNA extraction. After grinding, the material was placed in bead tubes provided by the ZR-Fungal/ Bacterial RNA MiniPrep isolation kit (ZYMO research, USA) and quickly frozen in liquid nitrogen prior to short-term storage at − 80 °C. RNA extraction was done according to the manufacturers' protocol. Quantity, quality, and efficiency of DNA removal was measured after DNAse treatment of each RNA sample (TurboDNAse, LifeTechnology, USA) by using a Bioanalyzer 2,100 (Agilent Technologies, USA). Three µl of each of the DNAse treated samples containing 2 ng/µl RNA (RIN > 5) were reverse transcribed into cDNA using SuperScript III reverse transcriptase (200U/µl) and random hexamer primers (100 ng/μl) followed by an RNase treatment to remove residual RNA according to the manufacturers' protocol (Life Technologies, USA).
Gene quantification. Based on the genome sequences of L. aestuarii and C. chthonoplastes, primers and TaqMan probes of the targeted circadian clock gene cluster kaiABC, the circadian clock reporter gene cikA, nitrogenase gene nifH, cell division protein gene ftsZ, photosynthesis D1 protein gene psbA, peroxiredoxin gene prx and two commonly used cyanobacterial housekeeping genes, rnpA and ppc 26 , were designed using the genetic analysis software Geneious R 8.1.7 58 (Table 2). The proper annealing temperatures of the primers were established by gradient PCR on a thermocycler (BioMetra) and specificity was confirmed by Sanger sequencing of the amplicons (BaseClear, Leiden, The Netherlands). Standard curves for RT-qPCR were prepared by using dilution series of PCR products and the primers were checked for species specificity by means of crossover PCRs and gel electrophoreses. Samples, standard curves and non-template controls were run in technical triplicates on a Rotor-Gene 6000™ (Qiagen, USA).

Statistics.
To evaluate the validity of the housekeeping genes across treatments, Ct values of genes of interest and housekeeping genes were used in the program BestKeeper 27 . T-tests were applied to examine the sample-wise expression within each genes' expression profile and the influence of treatments on transcription was assessed using R-based CCA scripts and ANOVA (Χ 2 , F-and p values (< 0.05)). In these tests normalized transcript abundances (( GOI geomean(HKGs) )) of each gene served as input data. CCAs were plotted and supplemented with confidence ellipses (95%) of the standard expression of the different genes. Potential significant rhythmic gene expression following a sinusoidal curve was predicted by the R based script Metacycle 29 using the transcription levels at different time points as input matrix. The cycle analysis was performed with the meta2D function that integrates multiple cycle prediction methods and combines their p-values. Training of the cycling algorhythm was performed by combining the datasets, thereby mimicking a 72 h sampling pattern and a p-value cut-off of 0.05 was used to determine potential sinusoidal patterns.