Table S2: TPM tables for high resolution experiments (DONE)
Table S3: DNA binding genes within each cluster (DONE)
Table S4: adjusted p-values for light sensitivity comparisons (DONE)
Not necessary, but might add if reviewers insist: Table S5: lists of genes within each interacting pair

The dawn gene expression waves are delayed in both phyAphyBcry1cry2 (red and blue light signalling) and prr5prr7prr9 (circadian) mutants

DNA binding proteins appear to have multiple coordinated bursts of expression at dawn and these bursts are either light or temperature dependent or both.  To characterise the gene regulatory mechanisms that underly these changes, we performed RNA-seq on a series of mutant strains that perturbed red- and blue-light sensitivity (phyA phyB cry1 cry2, Ler background), the circadian clock (prr5 prr7 prr9, Col-0 background) and temperature sensitivity (hsfA1QK, Col-0 background), all of which have all been previously characterised \cite{Yanovsky_2000,Liu_2013,Yoshida_2011}.  These were sampled at the most informative time points according to NITPicker (CITE).   We observe the observe the same patterns of expression in this subset of  time points as we were able to see in the original time series in Col-0 (Figure 2C).  We also find that the genes within each of these clusters have perturbed patterns of expression in the mutant strains (Figure 2C).  In particular, the early expressed genes were up-regulated in both the prr5 prr7 prr9 and phyA phyB cry1 cry2 mutants and were expressed for a larger portion of the time series, clusters 5-6 (which include many light-responsive genes) are expressed later in the time series (at 105-120 minutes instead of 45-60 minutes) in the prr5 prr7 prr9 and phyA phyB cry1 cry2 mutants and at higher expression levels. Clusters 7-8 (clusters that we previously found to have higher expression at elevated temperatures) were found to have higher expression levels in hsf1aQKprr5 prr7 prr9, and phyA phyB cry1 cry2 compared to Col-0 and Ler.  Finally, the gene expression within clusters 9-10 was less coordinated in the mutants than the WT and there wasn't as consistent a pattern of up- or down-regulation.    

An early morning gene regulatory network

Next, we wished to identify possible regulatory links by inferring a gene regulatory network.  Because genes change their expression rapidly at dawn, it is likely that there is a time delay between the time the gene is expressed and the time the protein is expressed.  For this reason, we applied a time delay network inference method called dynGenie3 (CITE).  Because all large-scale network inference algorithms have large false-positive rates, we wished to confirm as many edges as possible using available DNA binding data (via DAP-seq from \citealt{O_Malley_2016}). 
Among TFs in our network with available binding data, we found that 34.3% of our high confidence edges (top 1% highest scoring edges) were consistent with DAP-seq.  In contrast, we only observed a 12.6% overlap with DAP-seq data among edges in our network that received a score of '0' using dynGenie3 among TF with available DAP-seq data.  This suggests that dynGenie3 is successfully enriching for biologically plausible edges.    
From now on, we only consider the subset of the dynGenie3 network that is consistent with DAP-seq data.  This network is shown in Figure 4A: the network naturally forms a U-shape, with early expressed genes primarily on the left side of the U and late expressed genes primarily on the right side.  These genes are enriched for photoperiodism-related GO terms, including GO terms associated with light sensing, heat response and circadian rhythms (Fig. 3B) .
A number of transcription factors that are known to play a role in light signalling and photomorphogenesis sit at the base of the U, linking the early and late expressed genes;  these include  ELONGATED HYPOCOTYL 5 (HY5), B-BOX DOMAIN PROTEIN 31 (BBX31),CYCLING DOF FACTOR 5 (CDF5) and REVEILLE 1 (RVE1) (Figure 4A).  These transcription factors are all involved in light-controlled hypocotyl elongation \cite{Chattopadhyay1998,Rawat2009,Martín2018,Heng2019}. While RVE1 and CDF5 expression is under strong circadian control and peaks at dawn \cite{Rawat2009,Henriques2017}, HY5 and BBX31 are light-induced genes \cite{Oyama1997,Heng2019} and their transcript levels display a strong peak post dawn in our time course data (Fig. X?). Based on the timing of their expression and their position within our network, RVE1 and CDF5 may promote expression of growth-related genes before dawn, while HY5 and BBX31 appear to coordinate the down-regulation of early dawn genes as well as the induction of subsequent transcriptional waves.