The influence of the circadian clock on diurnal fluctuations of both proteins and transcripts is limited.
To determine if the significant changes we measured in the diurnal proteome could be controlled by the circadian clock, we next compared our data to a quantitative proteomics dataset acquired under free-running (continuous light) conditions (Krahmer et al., 2019). Our dataset of quantified proteins contains 1800 of the 2038 proteins (88%) reported in this study, suggesting a comparable depth of proteome analysis (Supplemental Data 7). To directly compare our data set with the one reported in Krahmer et al. (2019), we performed the same JTK cycle analysis to identify proteins cycling with 22 or 24 h period (Hughes et al., 2010). When corrected for multiple testing, we found a total of 21 proteins that exhibit a significant fluctuation in abundance, of which 3 demonstrated a similar pattern under continuous light conditions (Figure 3A). A less stringent test using “shuffling of protein levels” (Hughes et al., 2010) identified a total of 147 fluctuating proteins (211 were found to fluctuate under continuous light conditions using this method; Krahmer et al., 2019). Between our study and Krahmer et al. (2019), 3 proteins fluctuate in both studies, 1 only in L-D conditions and 7 only in continuous light. The fact that of these 11 proteins 10 have significant JTK-cycle fluctuations in continuous light (i.e., free-running condition) suggests that they are under circadian control, although additional proteome analysis of normal photoperiods prior to free-running conditions is needed to substantiate this possibility. We did find alpha-crystallin domain 32.1 (ACD32.1; AT1G06460) to fluctuate at the protein-level independent of the circadian clock. ACD32.1 was previously shown to be regulated diurnally at the transcript level in continuous light (Covington et al., 2008), but it did not fluctuate in the proteome data of Krahmer et al., 2019. ACD32.1 is a peroxisome-targeted chaperone protein (Pan et al., 2018) implicated in the suppression of protein aggregation (Ma, Haslbeck, Babujee, Jahn, & Reumann, 2006). The protein peaks in abundance immediately after dark, suggesting a need for peroxisomal protein stability in the dark to maintain peroxisome functions required for plant growth, including fatty acid oxidation (Pan et al., 2018). To determine if a relationship exists between protein and transcript levels among the proteins identified in the JTK-cycle analysis, we queried the Diurnal Database v2.0 (http://diurnal.mocklerlab.org/) and found many gene sets with thousands of genes fluctuating in their transcript levels. We searched the Diurnal Database for the genes encoding the 21 proteins showing a significant change in protein abundance in our dataset (Figure 3A, magenta and blue) and found 18 of the genes, of which 3 did not have fluctuating transcript levels. Thus, only few protein–transcript pairs have both fluctuating transcripts and proteins. For the few pairs that do, neither protein nor the corresponding transcript levels were peaking at a specific Zeitgeber time (Figure 3B). But when comparing the patterns of individual pairs, there was typically a median delay of 5.5 h between the transcript peak and protein peak, and the expression patterns are correlated (Figures 3C and D). Since such a shifted dependency of transcript and protein expression pattern is rare in our proteome dataset, its biological significance needs to be investigated further.
The identification of only a single highly significant JTK-cycling protein in our dataset, together with limited fluctuations of proteins reported for measured proteomes of Arabidopsis wild-type and circadian clock mutants growing in free-running cycles of continuous light (Choudhary et al., 2015; Krahmer et al., 2019) is unexpected at first sight. However, previous studies of growing Arabidopsis leaves during the diurnal cycle (Baerenfaller et al., 2012) have revealed that the abundance of most measured proteins is not affected by corresponding changes in their transcript levels. This disconnect between protein and transcript abundances might be explained by one or more of the following factors. First, to some extent, absence of protein abundance fluctuations can be accounted for by JTK_Cycle stringency, which only tests for periodical protein changes following a sinus function. Second, the temporal separation of the fluctuating circadian transcriptome and the mostly stable proteome may provide a currently unknown adaptive advantage. This possibility is consistent with previous results reported from proteome analyses of circadian clock mutants (Graf et al., 2017). Third, turnover rates of proteins may increase or decrease with constant translation of transcripts whose levels fluctuate, resulting in stable diurnal abundances of most proteins on a proteome-scale basis (Martin-Perez & Villen, 2017). Finally, regular changing diurnal conditions (light, temperature, humidity, etc.) may not significantly impact the levels of most proteins. Instead of regulating proteins by adjusting their abundance, plants might react to expected diurnal changes in conditions by regulating the activity of proteins. This would be supported by the fact that most circadian regulation in Arabidopsis involves post-translational modifications of proteins such as protein phosphorylation (Choudhary et al., 2015; Krahmer et al., 2019; Uhrig et al., 2019) independent of changes in mRNA and/or protein levels. Further studies will be necessary to understand which factors explain these findings.