Circadian protein regulation in the green lineage I. A phospho-dawn of protein modification anticipates light onset in the picoeukaryote O. tauri

Diel regulation of protein levels and protein modification had been less studied than transcript rhythms. Here, we compare transcriptome data under light-dark cycles to partial proteome and phosphoproteome data, assayed using shotgun mass-spectrometry, from the alga Ostreococcus tauri, the smallest free-living eukaryote. 10% of quantified proteins but two-thirds of phosphoproteins were rhythmic. Mathematical modelling showed that light-stimulated protein synthesis can account for the observed clustering of protein peaks in the daytime. Prompted by night-peaking and apparently dark-stable proteins, we also tested cultures under prolonged darkness, where the proteome changed less than under the diel cycle. The dark-stable, prasinophyte-specific proteins were also reported to accumulate when O. tauri formed lipid droplets. In the phosphoproteome, 39% of rhythmic phospho-sites reached peak levels just before dawn. This anticipatory phosphorylation suggests that a clock-regulated phospho-dawn prepares green cells for daytime functions. Acid-directed and proline-directed protein phosphorylation sites were regulated in antiphase, implicating the clock-related, casein kinases 1 and 2 in phase-specific regulation, alternating with the CMGC protein kinase family. Understanding the dynamic phosphoprotein network should be facilitated by the minimal kinome and proteome of O. tauri. The data are available from ProteomeXchange, with identifiers PXD001734, PXD001735 and PXD002909. This submission updates a previous version, posted on bioRxiv on 4th April 2018, as https://www.biorxiv.org/content/10.1101/287862v1 Highlight The phosphorylation of most protein sites was rhythmic under light-dark cycles, and suggested circadian control by particular kinases. Day-peaking, rhythmic proteins likely reflect light-stimulated protein synthesis in this microalga.

Highlight (<30 words) 62 The phosphorylation of most protein sites was rhythmic under light-dark cycles, and 63 suggested circadian control by particular kinases. Day-peaking, rhythmic proteins likely 64 reflect light-stimulated protein synthesis in this microalga. 65 Abstract 66 Diel regulation of protein levels and protein modification had been less studied than transcript 67 rhythms. Here, we compare transcriptome data under light-dark cycles to partial proteome and 68 phosphoproteome data, assayed using shotgun mass-spectrometry, from the alga Ostreococcus 69 tauri, the smallest free-living eukaryote. 10% of quantified proteins but two-thirds of 70 phosphoproteins were rhythmic. Mathematical modelling showed that light-stimulated protein 71 synthesis can account for the observed clustering of protein peaks in the daytime. Prompted by 72 night-peaking and apparently dark-stable proteins, we also tested cultures under prolonged 73 darkness, where the proteome changed less than under the diel cycle. The dark-stable, 74 prasinophyte-specific proteins were also reported to accumulate when O. tauri formed lipid 75 droplets. In the phosphoproteome, 39% of rhythmic phospho-sites reached peak levels just 76 before dawn. This anticipatory phosphorylation suggests that a clock-regulated phospho-dawn 77 prepares green cells for daytime functions. Acid-directed and proline-directed protein 78 phosphorylation sites were regulated in antiphase, implicating the clock-related, casein kinases 79 1 and 2 in phase-specific regulation, alternating with the CMGC protein kinase family. 80 Understanding the dynamic phosphoprotein network should be facilitated by the minimal  Cells were grown (as described above) and independent, triplicate cultures were harvested at 264 the times indicated. Cultures were monitored using spectrophotometry at 600nm. Total   quantifying peptides were required to report protein abundance.

365
Equivalence testing 366 Using the R equivalence package, the statistical equivalence of mean abundance across time 367 was tested as the highest p-value from exhaustive pairwise Two one-sided test approach 368 (TOST) tests over all ZTs (Schuirmann, 1981;Westlake, 1981   with changing levels of RNA and/or PMs but not of protein ( Figure 1C).

480
Protein levels nonetheless changed smoothly, with distinct waveforms. Of the twenty most 481 highly-detected proteins, likely including the most abundant, 11 were significantly rhythmic 482 but with low amplitudes (Supplementary Figure S2A), such that only ostta10g03200 483 exceeded the 1.5-fold change threshold (Table S1). 15 of the twenty most highly-detected 484 PMs, in contrast, were rhythmic by both criteria (Supplementary Figure S2B). The more 485 stringent, "equivalence" test revealed 49 proteins with significantly non-changing protein

504
The regulation was strikingly reversed in the proteome (Fig. 1C, 1D), where the major   Tables S3-S5, Supplementary Figures S4, S5). We therefore tested whether 537 this light-regulated synthesis alone could explain the observed distribution of protein peaks.

539
We simulated protein dynamics (Fig. 2D-2F; Supplementary Figure S6 stronger daytime preference than in the data. ostta03g04520 is an example of an RNA that 544 peaks at ZT0 and its protein profile (Fig. 2G) was very similar to the predicted protein from 545 such an RNA (Fig. 2D). The overall distribution of protein profiles substantially reflects the 546 light-stimulated translation rate of this organism (see Discussion).

548
Unusual, night-time proteins suggest a 'dark state'

549
An intriguing pattern of protein regulation stood out from the common, daytime abundance.  Table S4). Four un-annotated, prasinophyte-specific proteins in cluster P6 553 not only peaked at night, but were also among the 11, highest-amplitude profiles of all the 554 rhythmic proteins (Fig. 3A) The proteomic landscape changed less during dark adaptation (DA) than under a standard LD  Table S6), suggesting that O. tauri selectively mobilised this protein pool in 581 darkness.

583
The night-abundant, prasinophyte proteins that accumulated in DA, and night-depleted 584 proteins that fell in DA (such as ostta06g02940, above; or PPDK ostta02g04360,

585
Supplementary Figure S7C), suggested that prolonged darkness preserved a night-like state.

586
An alternative explanation was that protein stability in general was altered in the putative 587 dark state. We sought to test that notion, using the protein degradation rates that were RNA level and slightly-increasing protein level in DA also had among the lowest protein 596 degradation rates in LD (Fig. 3C), and was among the most-detected proteins in these 597 conditions (Supplementary Figures S2A, S8B). The RNA data and protein degradation rates 598 suggested that the prasinophyte-specific proteins accumulated due to a focussed, regulatory 599 mechanism, rather than generalised refactoring of the proteome (see Discussion).  To test the phospho-dawn pattern by a different method, we estimated the bulk protein 617 phosphorylation across the diel cycle using protein gel staining (Supplementary Figures S9A-618 B). The proportion of phosphorylated proteins was lowest in the daytime and increased 619 during the night to peak at ZT0 (Supplementary Figures S9C). Total phosphorylation was 620 therefore broadly consistent with the distribution of PM profiles (Fig. 2C). Taken together, 621 these results indicate that a regulator other than light or protein abundance controls the O. 622 tauri phosphoproteome before dawn. Below, we report phosphosite sequences that suggested 623 its identity.

624
Functions of proteins with rhythmic phospho-motifs 625 The LD datasets confirmed that protein phosphorylation profiles often diverged from protein 626 abundance. Colour-coding in Fig. 1H shows that clustering of the phospho-motif (PM) 627 profiles aligned with the PC analysis more clearly than for the lower-amplitude, protein 628 profiles (Fig. 1F). The largest cluster PM1 reflected the profiles that peaked in the ZT0 629 timepoint, which PC analysis also highlighted (Supporting Figure S2D). Phosphopeptide  Table S5).

644
In contrast, the PM2, PM4, PM7 and PM8 clusters peaked at ZT16, with or without  Tables S4 and S5). We therefore analysed the phospho-regulators that might 648 control these profiles, including potential contributions to non-transcriptional timing.

650
Phase-specific target sites 651 We first analysed motifs of amino acids that were enriched in rhythmic PMs, compared with 652 all quantified phosphopeptides to avoid potential detection bias due to PM abundance. PMs   rhythmic PMs that peaked at ZT0 and the proline-directed motifs were depleted (Fig. 5C). Supplementary Figure S10C). The most-changing PM on a predicted protein phosphatase was 682 pT175 in ostta11g02830, related to human Dual-specificity phosphatase DUSP12 (Fig. 5D).

683
Among the clock-related protein kinases, we note the dusk-peaking PM of GSK3 (above).

684
CK2 subunits were not detected in our data and the PM on CK1 was not strongly rhythmic 685 (Fig. 4). 21 other protein kinases bore rhythmic PMs that are predicted targets of these clock-

709
In contrast, the largest number of PM profiles peaked in the pre-dawn, ZT0 timepoint. This

758
A preprint coincident with our first report showed that three of these night-expressed,  We are very grateful to K. Kis, L. Imrie and D. Kelly for expert technical help, to B. Kolody