Arabidopsis thaliana accession Col-0 was used as WT. The following mutants were used: ahk2-5 ahk3-7 (Riefler et al., 2006), cca1-1 lhy-20 (Nitschke et al., 2016). Arabidopsis plants were grown on soil for 5 weeks under short day (SD) conditions (8 h light/16 h darkness) in a growth chamber with light intensities of 100–150 μmol m^–2 s^–1, using a combination of Philips Son-T Agros 400 W and Philips Master HPI-T Plus, 400 W/645 lamps, at 22°C and 60% relative humidity.

For stress treatments, 5-week-old SD-grown plants were used. Photoperiod stress was induced by a 24 h prolongation of the light period (prolonged light period, PLP) followed by a normal 16 h night (Figure 1A). Control plants remained under SD conditions. Stress parameters were analyzed in leaves 7–10. The harvest during the dark period was performed in green light.

Leaves 11–13 of 4-week-old Arabidopsis plants were used for inoculation with Pseudomonas syringae pv. tomato DC3000 (Pst), the method was carried out as described by Griebel and Zeier (2008). For pathogen infection we used an initial inoculum at OD₆₀₀ of 0.002.

Total RNA was extracted from leaf material (leaves 7–10). Only the distal parts of leaves 7–10 were harvested which is the most affected part of the leaves. Leaves were flash-frozen in liquid nitrogen and homogenized with a Retsch Mixer Mill MM2000 (Retsch, Haan, Germany). Total RNA was extracted using the NucleoSpin^® RNA plant kit (Machery and Nagel, Düren, Germany) as described in the user’s manual and in Frank et al. (2020). For RNA-seq analysis, RNA was isolated from three biological samples at four different time points. The isolated RNA was send to BGI (Hongkong, China) and processed as described in Cortleven et al. (2019). In brief, a NanoDrop NA-1000 and a Bioanalyzer Agilent 2100 (Agilent Technologies, Santa Clara, CA, United States) were used to check RNA concentration, integrity and rRNA contamination. After DNase I treatment, mRNA was enriched by using oligo (dT) magnetic beads and fragmented into shorter fragments. First-strand cDNA was synthesized by using random hexamer primers, followed by second strand synthesis. After purification, end repair, and 3′ end single-nucleotide A (adenine) addition, sequence adaptors were ligated. Following PCR amplification and quality control by the Agilent 2100 Bioanalyzer and ABI StepOnePlus Real-Time PCR System (Thermo Fischer Scientific, Waltham, MA, United States), the library products were sequenced on the BGI SEQ-500 platform. More than 22 million raw sequencing reads were obtained for each sample. After the removal of adaptors and low-quality reads, the obtained clean reads (approximately 21 million) were stored in FASTQ format (Cock et al., 2010). Sequences were aligned to the TAIR11 Arabidopsis reference genome using Bowtie2 (Langmead and Salzberg, 2012). Gene expression levels were quantified using RSEM (Li and Dewey, 2011) and DEGs were identified using the DESeq method (Love et al., 2014) considering three different parameters (time, genotype and treatment) with the following default criteria: fold change ≥ 2 and Bonferroni correction (p-value ≤ 0.05). RNA-seq data are deposited in NCBI’s Gene Expression Omnibus and are accessible through GEO Series accession number GSE173899.

Gene Ontology (GO) annotation was performed using PANTHER (Mi et al., 2013, 2019). Heatmaps were created using MEV (MultiExperiment Viewer; Saeed et al., 2003; Howe et al., 2011). For cluster analysis, MEV was used. Quality Threshold (QT) clustering was done using the following parameters: diameter: 0.7; minimum cluster size: 50; Euclidean distance or with a Pearson’s correlation: diameter: 0.3, minimum cluster size: 10. For hierarchical clustering, Euclidean distance and average linkage was used. PCA analysis was performed using the PCAplot function in R (version 3.6.2).

For quantitative real-time PCR (qRT-PCR), leaf material was collected at the same time points as for RNA-seq analysis. Quantitative real-time PCR analysis was performed as described in Cortleven et al. (2019). Sequences of primers used for gene expression analysis are listed in Supplementary Table 1. Gene expression data were normalized against three or four different nuclear-encoded reference genes (UBC21, TAFII15, PP2A, and/or MCP2A) according to Vandesompele et al. (2002) and expressed relative to the control treatment at time point 0 h which was set to 1.

The transcriptomic profile of photoperiod stress was compared with the transcriptomic profiles of different specific biotic and abiotic stresses. The percent overlap of changes caused by photoperiod stress with those caused by other stresses was calculated as follows:

JA, SA and camalexin were extracted and their levels determined by UPLC-MS/MS (Q-ToF-ESI; Synapt G2-S HDMS; Waters^®, Milford, Massachusetts, United States) as described in Valsamakis et al. (2020).

Statistical analyzes were performed using SAS v.9.2 (SAS Institute GmbH, Heidelberg, Germany) or R (version 3.6.2). Data were analyzed by a Welch t-test or ANOVA followed by Tukey’s post-hoc test. Normality and variance homogeneity of datasets were tested using the Shapiro-Wilk and Levene tests (Neter et al., 1996). To meet the assumptions, datasets were transformed using logarithmic or square root transformations.

In order to obtain genome-wide information of the transcriptional response to photoperiod stress, we analyzed changes of transcript abundance during the night following a light period that was prolonged by 24 h in SD-grown plants (WT, ahk2 ahk3, cca1lhy). This treatment causes a strong stress response (Nitschke et al., 2016), but shorter prolongations of the photoperiod in the range of few hours also cause a significant although weaker stress response (Abuelsoud et al., 2020). Samples for RNA-seq analysis were harvested at different time points (0, 4, 6, and 12 h) following the prolonged light treatment (Figure 1A). These time points reflected the timing of stress responses occurring during the night with photoperiod stress marker gene activation (BAP1 and ZAT12) starting around 5 h in WT and the photoperiod stress-sensitive mutants. Visible phenotypical consequences (flabby leaves in photoperiod stress-sensitive genotypes) appear about 8 h after the start of the night and coincide with a stronger induction of the photoperiod stress marker genes and the formation of ROS (Nitschke et al., 2016; Abuelsoud et al., 2020). Thus, the time points for sampling were chosen to allow monitoring of early transcriptional changes occurring before the onset of visible stress symptoms as well as to detect later transcriptional changes during the night when physiological and phenotypical consequences start to appear. A scheme of the experimental setup is shown in Figure 1A. Principal component analysis (PCA) of the DEGs showed that control samples cluster together (red circle; Supplementary Figure 1A) indicating that changes in gene expression due to the genotype are much smaller than those caused by the treatment. Samples harvested at the end of the PLP (0 h time point) cluster together with the control samples suggesting that control and photoperiod stress-treated samples do not differ significantly at timepoint 0 h. The 4 h and 6 h time points cluster separately from control samples and a division among the different genotypes is visible. Especially at the 12 h time point, a strong separation of the genotypes (blue circles; Supplementary Figure 1A) is evident. At this time point, the ahk2 ahk3 and cca1 lhy samples are clearly divergent from WT, which is consistent with the stronger photoperiod stress phenotype of these mutants (Nitschke et al., 2016).

We analyzed how many genes were regulated dependent on the genotype, time and treatment component (Supplementary Figure 1B). In total 10,278 genes were differentially expressed in a genotype-dependent manner, 17,465 genes in a time-dependent manner and 16,612 genes in a treatment-dependent manner (across genotypes and time points). The genotype-dependent DEGs were analyzed in detail by clustering and GO term overrepresentation analysis. Quality Treshold (QT) clustering revealed 21 different clusters (Supplementary Figure 2). Fifty two percent of all DEGs are found in cluster 1 and 2 which showed a downregulation (cluster 1) or an upregulation (cluster 2) during the night following the photoperiod stress treatment. This regulation has a higher amplitude in the photoperiod stress-sensitive mutants (Supplementary Figures 1C,D). According to GO term analysis, genes involved in photosynthesis- or chloroplast-related processes are overrepresented in cluster 1, whereas genes involved in autophagy, responses to endoplasmatic reticulum stress and Golgi vesicle transport are overrepresented in cluster 2 (Supplementary Figures 1E,F).

To get more insight in the DEGs following photoperiod stress, we made pairwise comparisons between photoperiod stress-treated and control samples for each genotype and time point. In all genotypes, the number of DEGs increased over time (Figures 1B–D and Supplementary Datas 1, 2). For instance, in WT there are 388 DEGs at time point 0 h, 1,226 DEGs at time point 4 h, 3,419 DEGs at time point 6 h and 4,912 DEGs at time point 12 h (Figure 1B and Supplementary Datas 1, 2). A large number of the early-regulated genes showed also an increased steady state level at later time points. The number of DEGs increased over time in all different genotypes, but the number of regulated genes being considerable higher in the mutants than in WT, which reflects their increased photoperiod stress sensitivity. The total number of DEGs at time point 12 h is 10,453 in ahk2 ahk3 and 7,556 in cca1 lhy reflecting their higher sensitivity to PLP.

GO enrichment analysis of the DEGs identified by comparing photoperiod stress-treated and control samples, revealed that cellular responses to oxygen levels, response to chitin and plant-type hypersensitive responses and positive regulation of defense responses were among the top five significantly enriched GO terms at 4 and 6 h in all genotypes (Figures 1E–G and Supplementary Table 2). This indicates that the early changes on the transcriptomic landscape of photoperiod stress can be associated with responses to oxygen and biotic stress responses, which have in common the occurrence of oxidative stress (Wojtaszek, 1997; Bolwell et al., 2002; Fukao and Bailey-Serres, 2004; Schmidt et al., 2018).

There are no indications for a photoperiod stress response at time point 0 h as neither photoperiod stress marker genes nor stress response genes are upregulated at this time point (Nitschke et al., 2016; Frank et al., 2020). Therefore, we considered the changes at the 0 h time point as genotype-dependent effect caused by the prolongation of the light period that do not belong to the specific photoperiod stress response occurring during the night. However, DEGs at time point 0 h might be relevant for the perception of photoperiod stress and the development of the initial response. All genotypes had 90 genes in common that were significantly affected by PLP at time point 0 h (Table 1 and Supplementary Table 3). GO term analysis revealed that this core-set of DEGs is related to the circadian clock.

The DEGs that are shared between WT, ahk2 ahk3 and cca1 lhy at time points 4, 6, and 12 h resulted in a core-set of 388 photoperiod stress-regulated genes, of which 382 genes were upregulated and 6 downregulated (Table 1 and Supplementary Table 3). GO overrepresentation analysis (Supplementary Table 4) revealed that the core-set of photoperiod stress-regulated genes belong mostly to cellular responses to hypoxia or to oxygen levels and defense responses to pathogens. Hence, the results support that photoperiod and oxidative stress are similar. We further evaluated these similarities below.

The transcriptional changes in the two stress-sensitive genotypes showed, beside a larger number of DEGs, a number of peculiarities which might be functionally relevant for the enhanced phenotype and which will be explored in more detail elsewhere. About two third of the responsive genes of WT were also found in the two photoperiod stress-sensitive genotypes. However, during the course of the night, there was an increasing number of DEGs characteristic for the stress-sensitive genotypes ahk2 ahk3 and cca1 lhy (Figure 2 and Supplementary Table 5), which do not occur in WT. One obvious difference to WT was the downregulation of numerous SMALL AUXIN UP-RNA (SAUR) genes. Photoperiod stress is characterized by an oxidative burst (Abuelsoud et al., 2020) which is more pronounced in the photoperiod stress-sensitive genotypes. Oxidative stress is known to affect auxin signaling (Blomster et al., 2011) which might be a cause for the downregulation of the numerous SAUR genes in these genotypes. Investigation of the functional relevance of these SAUR genes is an interesting direction for future research on the photoperiod stress syndrome.

A previous study (Abuelsoud et al., 2020) demonstrated that photoperiod stress is associated with a nightly increase in peroxide content resulting in an oxidative burst due to increased peroxidase and decreased catalase activity. Therefore, we focused especially on changed transcript levels of genes related to oxidative stress. Consistent with the increased oxidative stress after photoperiod stress (Abuelsoud et al., 2020), genes related to oxidative stress were found among the top 20 strongest up- and downregulated DEGs for the different genotypes (Supplementary Tables 6–11). Among them are OXIDATIVE SIGNAL-INDUCIBLE1 (OXI1), RESPIRATORY BURST OXIDASE HOMOLOG C (RBOHC), PEROXIDASE4 (PRX4), PRX37, ZAT11, CALMODULIN LIKE 37 (CML37), CML38, and ETHYLENE RESPONSE FACTOR 71 (ERF71). OXI1 encodes a protein kinase necessary for oxidative burst-mediated signaling in Arabidopsis (Rentel et al., 2004). RBOHC is required for the production of ROS in response to an extracellular ATP stimulus (Kaya et al., 2019) and both PRX4 and PRX37 encode apoplastic oxidative burst peroxidases (Valerio et al., 2004; Daudi et al., 2012; O’Brien et al., 2012). The TFs, ZAT11, and ERF71, are involved in nickel ion tolerance (Liu et al., 2014) or hypoxia (Hess et al., 2011), respectively, and can be induced by H₂O₂ (Hieno et al., 2019). The calmodulin-like proteins CML37 and CML38 are involved in drought stress and herbivore tolerance (Scholz et al., 2014, 2015) or in hypoxia stress (Lokdarshi et al., 2016), respectively. The induction of these genes was confirmed for all three genotypes by qRT-PCR analysis, which also showed that the induction was stronger in the mutants (Figure 3 and Supplementary Table 12).

To investigate the response profile of the oxidative stress-regulated genes during photoperiod stress, we used datasets of several studies identifying core-sets of ROS-responsive genes for comparisons to our dataset. Hieno et al. (2019) unraveled a core-set of 60 H₂O₂-responsive transcription factors (TFs) after H₂O₂ treatment; Zandalinas et al. (2019) identified in response to short high light treatment 82 H₂O₂- and RBOHD-dependent genes and Lai et al. (2012) investigated the relation between the circadian clock and ROS-responsive genes and identified a core-set of 167 genes of which 140 were clock-regulated. In addition, transcriptional profiles specific for the response to H₂O₂, superoxide and singlet oxygen were identified by Gadjev et al. (2006). Overall, these four different core-sets of ROS-responsive genes showed only a small overlap (Supplementary Figure 3A), probably due to the different experimental setups used to increase ROS production. Therefore, these sets of genes were pooled to form a new master core-set of 283 ROS-responsive genes (Supplementary Data 3). QT clustering of transcript levels after photoperiod stress of this master core-set of ROS-responsive genes indicated that there is a strong regulation of these genes starting at 4 h during the dark relaxation (Figure 4A). Four different clusters were identified: Cluster I shows time-dependent upregulation of genes starting at the 4 h time point; Cluster II shows time-dependent upregulation of genes starting at the 6 h time point; Cluster III shows first upregulation of genes after 4 h and 6 h and then a decrease in expression at the 12 h time point; cluster IV shows time-dependent downregulation of genes. In all clusters, the response of the photoperiod stress-sensitive mutants ahk2 ahk3 and cca1 lhy is stronger. Transcript levels of representative genes (ZAT12, ERF1, PACLOBUTRAZOL RESISTANCE1 (PRE1) and C-REPEAT/DRE BINDING FACTOR2 (CBF2) of the different clusters measured by qRT-PCR confirm the transcriptional regulation (Figures 4B–E). In addition, we analyzed the proportion of genes of the ROS core-set that are regulated at different time points in the different genotypes after photoperiod stress. Results showed that in all genotypes this proportion increased over time with the highest co-regulation at time points 6 and 12 h (Supplementary Figure 3B).

Comparison of the photoperiod stress-responsive transcriptomic profile with the distinct superoxide-, singlet oxygen- and H₂O₂-induced gene profiles identified by Gadjev et al. (2006) (Supplementary Figure 3C) revealed a strong overlap with the singlet oxygen-induced transcript profile (Supplementary Figure 3B and Supplementary Data 4). Similarly, a strong overlap with singlet oxygen-UV-B early, RBOHF-related and oxidative stress (ROS)-related transcript profiles (Figure 4F and Supplementary Data 5) can be recognized when comparing the core-set of DEGs after photoperiod stress with the different ROS footprints (Willems et al., 2016). Together, these results indicate the involvement of ROS signaling on the transcriptomic response to photoperiod stress.

Because photoperiod stress causes an oxidative burst (Abuelsoud et al., 2020), we investigated the transcriptional regulation of 221 genes encoding enzymes involved in the scavenging of ROS. QT cluster analysis revealed three major clusters (Supplementary Figure 4 and Supplementary Data 6). Cluster I showed an upregulation of genes over time in all genotypes, which is stronger in the photoperiod stress-sensitive genotypes, including the photoperiod stress-responsive PRX34 (Abuelsoud et al., 2020), Cluster II shows a downregulation of genes over time, which is even stronger in the stress-sensitive mutants. CAT2 (CATALASE2) is one of the genes of cluster 2. This downregulation of CAT2 is in agreement with Abuelsoud et al. (2020) who found a decreased catalase activity after photoperiod stress. Cluster III consists of genes with a particular high expression at 4 h and 6 h time points for ahk2 ahk3 mutants. PRX4, which was identified by Abuelsoud et al. (2020) as one of the strongly regulated genes upon photoperiod stress in relation to oxidative stress, belongs to this cluster. Among the top 20 significantly regulated genes are also a number of genes encoding glutathione-S-transferases (GSTs) (Supplementary Tables 6–11). GSTs are involved in the metabolization of toxic reactive compounds such as ROS during an oxidative burst (Gallé et al., 2019).

Taken together, these results indicate that photoperiod stress affects the expression of genes encoding enzymes involved in the scavenging of oxidative stress.

Photoperiod stress is a relatively new form of abiotic stress and not much is known about similarities with other stresses. We therefore compared the transcriptomic profile of plants in response to photoperiod stress with those in response to other biotic (Truman et al., 2007; Winkelmüller et al., 2021) and abiotic stresses, including a shift to blue light (BL), drought stress, heat stress, cold stress, salt stress, ozone treatment, fluctuating light and high light stress (Lee et al., 2005; Tosti et al., 2006; Kleine et al., 2007; Truman et al., 2007; Huang et al., 2008, 2019; Larkindale and Vierling, 2008; Ding et al., 2014; Schneider et al., 2019; Figure 5 and Supplementary Data 7). In addition, we compared our dataset with the transcriptomic profile of circadian clock-regulated genes (Covington et al., 2008) as a previous study found a link between photoperiod stress and the circadian clock (Nitschke et al., 2016). At the 0 h time point, the number of commonly regulated genes was relatively low in all genotypes, which indicate that the photoperiod stress response starts later during the night following the PLP, similar to the ROS-responsive transcript profile overlaps (Figure 4). Ozone treatment, flg22-induced PTI and pathogen attack showed the highest number of commonly regulated genes with photoperiod stress treatment (Figure 5). Already after 4 h, approximately 37% of photoperiod stress-regulated genes were identical to those regulated by ozone stress. The percentage of commonly regulated genes even increased to almost 70% at time point 12 h. 50% of the genes responding to pathogen attack and approximately 70% of the genes responding to flg22-induced PTI were also regulated by photoperiod stress at time point 12 h. In the photoperiod stress-sensitive mutants, the overlap between the transcriptional response to photoperiod stress and other stresses was stronger, however, again the highest overlap was detected between photoperiod and ozone stress, pathogen attack and flg22-induced PTI. Interestingly, the stresses that show the highest overlap of their DEGs with photoperiod stress cause an oxidative burst as does photoperiod stress (Abuelsoud et al., 2020).

Conspicuously, the transcriptomic response to photoperiod stress shows overlap with the response to high light (HL) stress (Figure 5). However, the photoperiod stress response differs from a HL stress response as plants display an effect directly after the HL stress, e.g., reduced photosynthetic capacity, while after photoperiod stress treatment plants only show a stress phenotype during the following night (Cortleven et al., 2014). Moreover, a prolongation of the light period alone is not causative for the photoperiod stress response as a night of at least 7.5 h is necessary to induce a cell death phenotype specific for the photoperiod stress response (Nitschke et al., 2016).

The similarity of the transcriptional responses to photoperiod stress, pathogen attack and ozone stress (Figure 5) motivated us to explore possible links between these response pathways. Besides the oxidative burst evoked by these stresses, JA and SA biosynthesis and signaling are common signaling pathways affected by these stresses. Both, JA and SA, are involved in downstream signaling of PTI and ETI (Zhang et al., 2018). During ozone stress, SA accumulates (Yalpani et al., 1994; Sharma et al., 1996) and the ozone-induced hypersensitive cell death is modulated by JA signaling (Rao et al., 2000). Therefore, we explored the expression pattern of SA (Figure 6A and Supplementary Data 8) and JA (Supplementary Data 9 and Supplementary Figure 5) biosynthesis/signaling genes. The analysis revealed that numerous SA-related genes were strongly upregulated by photoperiod stress (Figure 6). Transcriptional regulation of selected SA biosynthesis and signaling genes has been confirmed by qRT-PCR for EDS1, PAD4, ICS1 and PATHOGENESIS-RELATED GENE 1 (PR1) in WT and the stress-sensitive genotypes ahk2 ahk3 and cca1 lhy (Figures 6B–E). Transcript levels of JA biosynthesis and signaling genes such as LIPOXYGENASE3 (LOX3), LOX4 and JASMONATE-ZIM-DOMAIN PROTEIN 1 (JAZ1), were strongly upregulated as well after photoperiod stress in WT and the photoperiod stress-sensitive mutants (Supplementary Figure 5 and Supplementary Data 9). Besides the increased transcriptional regulation of SA and JA biosynthesis genes, also the levels of SA and JA-Ile were strongly increased after photoperiod stress in WT and the photoperiod stress-sensitive mutants (Figure 7).

During pathogen defense responses, SA is an essential signaling molecule for both basal defense mechanisms and SAR. In addition, SA induces the formation of camalexin, which is one of the major phytoalexins in plant defense responses decreasing bacterial growth after an infection (Glawischnig, 2007). After photoperiod stress, camalexin concentrations increase strongly and the transcript levels of PHYTOALEXIN DEFICIENT3 (PAD3), a key enzyme of camalexin biosynthesis increased upon photoperiod stress treatment (Figures 8A,B). In addition to these SA-dependent defense responses, we also observed that transcript levels of FMO1, which encodes flavin-dependent mono-oxygenase that is an essential component of the biologically induced SAR independent of SA (Mishina and Zeier, 2006), is strongly upregulated during the night at 6 and 12 h after photoperiod stress treatment (Figure 8C). Together, these results clearly demonstrate that photoperiod stress induces responses similar as pathogen infection.

Because plants respond to photoperiod stress similar as to pathogen infection we asked whether photoperiod stress enhanced the plant resistance against (hemi-) biotrophic pathogens.

To answer this question, we inoculated previously photoperiod-stressed and none-stressed plants with Pseudomonas syringae pv. tomato DC3000 (Pst). In the first experiment, we inoculated not yet fully developed leaves (11–13) with Pst in the morning the day after the photoperiod stress treatment (24 h light prolongation). These leaves were chosen as they do not become flabby as mature leaves do after a 24 h prolonged light period (Nitschke et al., 2016), which might interfere with the pathogen infection. Colony forming units were counted 3 days post infection. In plants pretreated with photoperiod stress, bacterial growth was strongly reduced (Figure 8D).

In a second experiment, we inoculated mature leaves but decreased the duration of the photoperiod stress treatment to 8 h light prolongation (16 h PLP) thus avoiding the flabby phenotype in mature leaves. This pretreatment also decreased the bacterial growth in WT (Figure 8E). Similar trends were observed in the photoperiod stress-sensitive genotypes (Figure 8F).

The results of both experiments demonstrate that photoperiod stress results in an enhanced immunity in the absence of a pathogen.