Quantitative analysis and bioinformatics
Total proteome - Label‑free precursor (MS1) intensity based quantification was performed using Progenesis QI for Proteomics (version 2.1, www.nonlinear.com) to quantify total proteome changes. Briefly, for each individual fraction, automatic alignment was reviewed and manually adjusted before normalization. From each Progenesis peptide ion (default sensitivity in peak picking) a maximum of the top five tandem mass spectra per peptide ion were exported as a Mascot generic file (*.mgf) using charge deconvolution and deisotoping option and a maximum number of 200 peaks per MS/MS. Searches were done in Mascot 2.4.1 (Matrix Science) against a decoyed (reversed) Arabidopsis protein database from TAIR (release TAIR10) concatenated with a collection of 261 known mass spectrometry contaminants. Precursor ion mass tolerance was set to 10 ppm and the fragment ion mass tolerance was set to 0.6 Da. The following search parameters were used: trypsin digestion (1 missed cleavage allowed), fixed modifications of carbamidomethyl modified cysteine and variable modifications of oxidation of methionine, deamidation of asparagine and glutamine, and acetylation of protein N terminal peptides. Mascot searches were imported into Scaffold 4.2.1 (Proteome Software). The following thresholds were applied: peptide FDR ≤ 5, protein FDR ≤ 10, 1 minimum peptide. Spectrum reports were imported again into Progenesis. After this, individual fraction analyses were combined into the full quantitative Progenesis experiment. From this, quantitative peptide values were exported for further processing. Only peptides that could be unambiguously assigned to a single protein (gene model annotation) were kept for quantification. A Hi‑4 strategy (Grossmann et al., 2010) was applied to obtain protein quantitative values. Proteins with 2 or more peptides assigned were considered as quantifiable. Following these criteria, the final protein level FDR was estimated at 0.013.
Phosphoproteome - Quantification of changes in identified phosphopeptides was performed using MaxQuant (version 1.3.0.5) with default settings and the following modifications: fixed peptide modification by carbamidomethylation of cysteines and variable peptide modifications by phosphorylation of serine, threonine and tyrosine, and oxidation of methionine, and false discovery rate (FDR) tolerances of ≤ 0.05 (protein) and ≤ 0.01 (peptide). MaxQuant outputs were subsequently filtered for phosphopeptides with a phosphorylation site probability score ≥ 0.8 and presence in at least 2 of 4 biological replicates and 2 of 3 time‑points for each light transition.
Data Analysis - Significant fluctuations in protein abundance and phosphopeptides were determined using an ANOVA analysis: total proteome (P value ≤ 0.05 and Fold-change (FC) ≥ 1.5) and phosphoproteome (P value ≤ 0.05). The significantly changing proteome was subjected to cluster analysis using GProX (Rigbolt, Vanselow, & Blagoev, 2011). Six clusters were generated in an unsupervised clustering manner based on the fuzzy c-means algorithm. Significantly changing proteins and phosphoproteins were subjected to gene set enrichment analysis (GSEA) using the SetRank algorithm relative to the identified proteome and phosphoproteome, respectively (Simillion, Liechti, Lischer, Ioannidis, & Bruggmann, 2017). Enrichment was calculated for all the available databases included in the SetRank R package. Only terms with a size ≥ 2 were considered (gene set size ≥ 2). For each protein cluster, a SetRank corrected P value ≤ 0.01 was applied as threshold. For phosphoproteins changing at the L-D or D-L transition, a SetRank corrected P value ≤ 0.01 and an FDR ≤ 0.05 were applied. To test for significantly non-changing proteins at the transitions to light, (i.e., at dawn, ZT23 to ZT1, and dusk, ZT11 to ZT13), a TOST equivalence test (equivalence R package) was applied with an ε = 0.4. Significance threshold was P value ≤ 0.05. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository. Data are available via ProteomeXchange with identifier PXD007600.
Additional Analyses - To compare protein and mRNA profiles, mRNA data generated by the Alison Smith laboratory was obtained from the Diurnal database (http://diurnal.mocklerlab.org; Mockler et al., 2007). Data was standardized to plot both protein and mRNA data in the same graph. Predicted subcellular localization of all changing proteins and phosphoproteins was performed using the consensus subcellular localization predictor SUBAcon (suba3.plantenergy.uwa.edu.au) (Tanz et al., 2013). String DB network analyses were undertaken using both proteome and phosphoproteome data. String DB analyses were performed in Cytoscape using the String DB plugin stringApp (Szklarczyk et al., 2017). A minimum correlation coefficient of 0.5 was used along with a second layer of 5 additional nodes to infer network connectedness.
JTK Analyses – To compare diurnal protein fluctuations to free running circadian clock fluctuations published by Krahmer et al. (2019) we performed an equivalent analysis using the JTK cycle to identify proteins cycling with 22 or 24 h period (Hughes, Hogenesch, & Kornacker, 2010). The exact loading script JTK_analysis.zip is available upon request. The data was then used to produce Figure 3B, C and D. Proteins identified to fluctuate were normalized such that they fluctuate around a median of 0 with maximal amplitudes of 2. Transcriptome data from Diurnal DB (http://diurnal.mocklerlab.org; Mockler et al., 2007) was used to determine if the associated transcripts were also fluctuating, and if so, when. To estimate a confidence interval for the relative expression or protein level errors, their relative levels were compared to the theoretical cosine function at the same timepoint. Based on all errors, irrespective of the exact timepoint, a 99% confidence interval was computed.