The body’s redox cycle controls immune-triggered cell death differently from the genetic clock

The difficulty in elucidating the distinct circadian function of the redox rhythm stems from the fact that, in addition to the lack of the knowledge of its oscillatory mechanism, it is intricately intertwined with the genetic clock. Disruption of the redox rhythm can affect the period, the phase, or the amplitude of the genetic clocks depending on their molecular architectures and the nature of the disruption (12, 14, 15) and vice versa (9). Thus far, genetic clock-defective mutants have been used in an attempt to identify the specific transcriptional targets of the redox rhythm. However, the scant overlap between the oscillating transcripts in these mutants and the wild type (WT) (7, 8) raised the concern that the targets identified in the mutants might not be functionally relevant for the WT (17).

To test whether the genetic and redox rhythms can be disentangled by an alternative method, we used the Arabidopsis mutant of the clock genes PSEUDO-RESPONSE REGULATORs (PRRs) 7 and 9 (prr7 prr9) which has the intrinsic genetic clock periods of approximately 24 h at 12 °C, 32 h at 22 °C (typical temperature for growing Arabidopsis), and 36 h at 30 °C (18). Therefore, in contrast to the WT, which has the ability to maintain the ~24 h period across a physiological range of temperatures, the prr7 prr9 mutant is defective in temperature compensation. This temperature-dependent period variation in the prr7 prr9 mutant makes it a tractable system to separate the redox and genetic rhythms based on oscillation periods.

We used oxidized (GSSG) glutathione to measure the period of the redox rhythm for several reasons. First, GSSG provides a good approximation of the sum total of glutathione redox metabolism (i.e., a branch of the redox rhythm), because unlike GSH, which can be synthesized de novo, GSSG is produced solely during cellular redox reactions. Unsurprisingly, GSSG was previously shown to follow the same redox rhythm as peroxiredoxin oscillations which were detected in all lineages of life (9, 10). Second, relatively high glutathione abundance allows more robust measurements. Finally, glutathione plays more diverse roles in maintaining cellular redox homeostasis in plants (19) compared to other electron carriers, such as NADPH. While NADPH can be directly utilized for protein reduction by the thioredoxin-NADPH-dependent thioredoxin reductase (NTR) system, a significant portion of it is used to reduce GSSG to GSH via glutathione reductase (GR) to mitigate oxidative stress. GSH is not only used to reduce oxidized proteins but also plays a central role in direct scavenging of reactive oxygen species (ROS) (Fig. 1A). A more detailed overview of the redox regulation in plants can be found in the recent reviews (20, 21).

Prior to sample collection, we grew WT and prr7 prr9 plants in temperature- and humidity-controlled Percival chambers (22 °C, 65% RH) (22), under 12 h light/12 h dark (LD) and then released the plants to constant light (LL) conditions at either 22 °C or 30 °C to collect plant tissue for concurrent glutathione and transcriptomic analyses from 24 h to 92 h in LL (Fig. 1B). For transcriptomic measurements, we used the multiplex RNA annealing selection ligation-sequencing (RASL-seq) approach (23) to analyze the expression of a selected pool of ~700 genes (Dataset S1↗) as an alternative to RNA-seq to accommodate the large sample number for the time-course experiments (two genotypes, two temperatures, and 18 time points with 6 replicates for 22 °C and 3 replicates for 30 °C).

Harmonic regression analysis of the time-course data (12) revealed that in contrast to the core genetic clock gene LATE ELONGATED HYPOCOTYL (LHY) which showed longer periods in the prr7 prr9 mutant than in the WT (Fig. 1 C and D), GSSG exhibited close-to-WT periods in the prr7 prr9 mutant, albeit with different phases, at both 22 °C and 30 °C (Fig. 1 E and F). Thus, the redox rhythm, represented by GSSG, is distinguished from the genetic clock based on the period lengths in the prr7 prr9 mutant (Fig. 1 G and H). Notably, while the elevated temperature increased the period of LHY from 31.71 h to 37.34 h in the prr7 prr9 mutant, recapitulating the previously observed temperature overcompensation (18), the period of GSSG changed only slightly from 27.34 h to 26.57 h. Therefore, GSSG oscillation appears to exhibit temperature compensation, a known hallmark of circadian clocks (24), separately from the genetic clock.

In contrast to GSSG, the total glutathione (GSSG + GSH) oscillated with the genetic clock, likely through the regulation of the GLUTAMATE-CYSTEINE LIGASE (GSH1), also known as PHYTOALEXIN DEFICIENT 2 (PAD2) (25) (Fig. 1 A and I–N). This gene encodes the enzyme that catalyzes the rate-limiting step of glutathione biosynthesis in Arabidopsis (26) with its rhythmic pattern aligned with that of total glutathione. Since the total glutathione and GSSG levels were measured in the same samples, the separation of GSH+GSSG and GSSG oscillations further supports the independence of the redox rhythm from the genetic clock. We hypothesize that since GSSG is produced only through redox metabolism and is the much smaller and more dynamic fraction of the total glutathione pool (~2 to 5%; Fig. 1 E, F, I, and J) (27, 28), it may serve as an "indicator" of GSH oxidation reactions and hence the redox rhythm.

The regulation of the overall glutathione production by the genetic clock provides a possible mechanism for the genetic clock to entrain the redox rhythm. This interplay might also explain the altered phase of GSSG in prr7 prr9 compared to WT. To test this hypothesis, we modified a previously proposed simple mathematical model for the redox rhythm (13) to introduce periodic forcing by a long-period, genetic clock-driven glutathione oscillation (Fig. 2A), which resulted in a similar transient phase shift for the redox rhythm (Fig. 2B).

To verify our findings in planta, we used the glutaredoxin (GRX)-fused redox-sensitive GFP (roGFP2) reporter system. roGFP2 can be excited by two different wavelengths of light depending on its oxidation state, allowing real-time ratiometric monitoring of glutathione redox potential (Fig. 2C). Since chloroplasts are the primary source of reducing power and energy in plants, we transformed roGFP2 variant localized to chloroplastic stroma (roGFP2_chl) (29) into WT and prr7 prr9 mutant to monitor the redox rhythm in live plants. We used a custom-built fluorescence imaging chamber to perform time-lapse measurements of reduced and oxidized roGFP2_chl under LL at 30 °C, the temperature at which there is a large separation between the genetic clock and the redox rhythm periods (Fig. 1H). We used the normalized ratio of the roGFP2 fluorescence after excitation with 2 different wavelengths. We detected robust oscillations of the fluorescence ratio (Fig. 2D), indicating the viability of roGFP2 as an in vivo readout of the redox rhythm. The analysis using Biodare's FFT-NLLS algorithm (30) of individual traces confirmed that, in agreement with the GSSG data (Fig. 1H), the roGFP2 ratio in prr7 prr9 oscillated with a close-to-24 h period (Fig. 2E), instead of expected genetic clock period of ~36 h. A phase difference similar to the one in GSSG measurements was also observed (Figs. 1F and 2D), further indicating that these oscillations are endogenous and represent the redox rhythm as measured by GSSG.

The distinct oscillation periods for the redox rhythm and the genetic clock in the prr7 prr9 mutant provided an opportunity to identify their specific transcriptional targets via clustering. We designed a two-step strategy to identify oscillatory genes driven by the redox rhythm. Since we are interested in not only period length of the target genes, but also their coexpression in conferring a specific physiological function, we first clustered the transcripts according to their normalized temporal signature using k-means clustering of the circadian time course RASL-seq data in the prr7 prr9 mutant at 22 °C. After testing different k values, we chose k = 16 to accommodate all possible clusters for the genes oscillating with the genetic clock period of approximately 32 h (8 points for each of the possible phases given by the time resolution of 4 h, multiplied by 2 to account for positive and negative skew). For genes in each coexpression cluster defined using the prr7 prr9 at 22 °C data, the second step was to identify rhythmic transcripts in WT and the prr7 prr9 mutant at 22 °C and 30 °C using the CosinorPY package (31) with the significance cutoff of P < 0.05. While a close-to-24 h period was observed across all coexpression clusters in the WT at 22 °C, larger period length variations were found in the prr7 prr9 mutant with majority of the gene clusters oscillating with the expected genetic clock period of ~32 h (Fig. 3 A and B). Interestingly, clusters 2 and 6 (C2 and C6, Dataset S2↗) predominantly contained oscillatory genes (shown as dots) with periods close to that in WT. Since the output of k-means clustering can vary depending on the initial seed and the number of clusters, we also used a deterministic hierarchical algorithm to analyze the RASL-seq data, which largely recapitulated the results of the k-means clustering (SI Appendix, Fig. S1↗).

To rule out the possibility that the observed period-length separation of the circadian genes in prr7 prr9 was caused by tissue-specificity (32 –34), we performed further analysis on the RASL-seq data generated using homogenized leaf tissues by grouping genes with known tissue specificity versus those without (35) and found that C2 and C6 genes did not originate from a single tissue and were still separated from others with their distinctly shorter periods across all groups (SI Appendix, Fig. S2A↗), indicating that period separation of C2 and C6 is unlikely due to a tissue-specific effect. We also examined two genetic clock-driven genes CHLOROPHYLL A/B-BINDING PROTEIN 2 (CAB2) and CATALASE 3 (CAT3) which were reported to have a ~2 h-difference in period length in WT (36) and found that this difference persisted in prr7 prr9, but with both periods >32 h (SI Appendix, Fig. S2B↗), unlike those of genes in C2 and C6. Finally, we used another package, MetaCycle (37), to analyze our RASL-seq data by a combination of two additional algorithms, JTK_CYCLE and Lomb–Scargle periodogram, to independently verify our results from CosinorPY analysis. After analyzing statistically significant oscillations (P < 0.05), we again found C2 and C6 to oscillate with a shorter period in the prr7 prr9 mutant (SI Appendix, Fig. S2 C and D↗). Notably, oscillations for genes in both C2 and C6 also showed altered phase in prr7 prr9 compared to WT at 22 °C, resembling that of GSSG (Figs. 1E and 3 C and D), consistent with these genes being regulated by the redox rhythm.

Surprisingly, at 30 °C, virtually none of the genes in C2 and C6 kept the shorter period (Fig. 3 E–H and SI Appendix, Fig. S2 E and F↗). At this elevated temperature, these genes appeared to be driven by the genetic clock. Therefore, the C2 and C6 genes can be regulated by both the redox rhythm and the genetic clock, with the redox rhythm playing the primary role under the normal temperature and the genetic clock dominating at elevated temperatures (Fig. 3 A–H). Among the genes studied in our RASL-seq, a single gene from C13, the key flowering time regulator FLOWERING LOCUS T (FT) (38) reliably oscillated with a shorter period under both temperatures (SI Appendix, Fig. S3 A and B↗), in agreement with the near WT redox rhythm period observed in the prr7 prr9 mutant (Figs. 1 G and H and 2 D and E). These results imply that at 30 °C, while the C2 and C6 genes are delinked from the redox rhythm, FT, and perhaps other flowering time genes not present in our RASL-seq gene pool, is still under the circadian redox rhythm regulation, likely through a different signaling pathway.

Since time-course experiments may lead to inaccurate period estimation due to the low time resolution and plant sample variations, we next constructed bioluminescence reporter lines to verify our findings. Based on previous publications (12, 22), we chose the promoter of CATALASE 2 (CAT2), a morning-phased output gene of the genetic clock (SI Appendix, Fig. S3 C and D↗), to drive the luciferase reporter. To make a reporter for the redox rhythm output, we chose, from C6, the promoter of S-NITROSOGLUTATHIONE REDUCTASE 1 (GSNOR1), which is the key regulator of S-nitrosothiol (SNO) levels in plants (39) (Fig. 1A). Besides its robust oscillation detected in our RASL-seq data (SI Appendix, Fig. S3 E and F↗), the dependence of GSNOR1 activity on NADH and GSH (40) further connects the gene to the redox rhythm. CAT2p:LUC and GSNOR1p:LUC constructs were transformed into both WT and the prr7 prr9 backgrounds and all independent T1 lines were imaged to minimize insertion position effects. Since the leaf age may affect the period of the circadian clock (41) within the same plant, and to avoid reported tissue-specific effects (32 –34), we quantified the luminescence of the shoot apex, whose circadian rhythms are robust and highly synchronized (42). The period length was then compared for all rhythmic lines [relative amplitude error (RAE) less than or equal to 0.7] across three independent experiments. In agreement with the RASL-seq data, significant differences in period lengths of CAT2p:LUC and GSNOR1p:LUC oscillations were detected in prr7 prr9, but not in WT at 22 °C (Fig. 3 I–L). Interestingly, for the last oscillation cycle, GSNOR1p:LUC oscillation period appeared to lengthen to match the CAT2p:LUC oscillation. This is likely due to the entrainment of the redox rhythm by the genetic clock through rhythmic transcription of GSH1 (Fig. 1 L–N). In support of this hypothesis, we recapitulated similar results during the later timepoints of the mathematical model (Fig. 2B), indicating the transient nature of the period separation between the redox rhythm and the genetic clock in the prr7 prr9 mutant. To ascertain that the period separation is not limited to this pair of genes, we tested another pair of reporters. CATALASE 3 (CAT3) was chosen as an evening-phased genetic clock output (12) to ensure that both morning- and evening-phased genetic clock outputs were separated from the redox rhythm. For the redox rhythm, the GLUCOSE-6-PHOSPHATE DEHYDROGENASE 6 (G6PD6) gene encoding an enzyme from the oxidative Pentose Phosphate Pathway (PPP) (43) (Fig. 1A) was chosen because PPP is a hub for the interplay between the redox rhythm and the genetic clock in mammals (14, 15) and the G6PD6 gene represents a different redox-regulated gene cluster (C2 in Fig. 3B). As with the previous pair of reporters, we found that G6PD6p:LUC oscillated with a significantly shorter period than CAT3p:LUC (Fig. 3 M–P) in prr7 prr9, but not in WT. Therefore, we observed the coexistence of the redox and genetic circadian rhythms within the same tissue type (shoot apical meristem) based on the distinct periods of their output genes (Fig. 3 L and P), further confirming the results from the RASL-seq, glutathione, and roGFP2 measurements (Figs.1 G and H, 2E, and 3B).

To test whether genes with similar temporal patterns (i.e., clusters) have a common biological function, gene ontology (GO) enrichment analysis was performed, with the RASL-seq probe set as the background to eliminate the bias stemming from using a limited gene pool. Remarkably, even in this probe set mostly containing genes involved in different immune responses (last column in Fig. 4A), the C2 genes were found to be significantly enriched with those involved in the regulation of immunity and cell death (predominantly immune-induced PCD, known as hypersensitive response in plants), while C6 had no significant enrichment possibly due to the low number of genes in the cluster. Other clusters with significant GO term enrichments had photosynthesis (C3)-, growth (C4)-, and messenger RNA (mRNA) processing (C14)-related genes (Fig. 4A). When the redox rhythm target genes (C2 and C6) were combined for GO enrichment analysis, immune-related GO terms were found again, similar to C2 alone, though with the cell death becoming the top GO term (Fig. 4B). Further STRING (Search Tool for the Retrieval of Interacting Genes/Proteins) analysis uncovered a highly interconnected network, with known immune-induced cell death regulator genes such as SUPPRESSOR OF BIR1-1 (SOBIR1) (44), PHYTOALEXIN DEFICIENT 4 (PAD4) and ENHANCED DISEASE SUSCEPTIBILITY 1 (EDS1) (45) in hub positions (Fig. 4C). Moreover, in agreement with our clustering analysis based on temporal expression signatures, STRING detected a high degree of coexpression of genes in clusters 2 and 6 according to their publicly available data (black lines in Fig. 4C). Moreover, analysis of only those genes showing statistically significant oscillations (P < 0.05) (Dataset S3↗), produced similar results (SI Appendix, Fig. S4↗). Based on the identities of C2 and C6 genes, we hypothesized that the circadian redox rhythm regulates PCD during the effector-triggered immunity (ETI), which occurs when the presence of a pathogen effector is recognized by its cognate nucleotide-binding site, leucine-rich repeat (NB-LRR) receptor in the plant host. This immune response is known to be associated with ROS burst (46) and significant increase in ATP levels (47, 48). It is also known that glutathione depletion severely compromises PCD (25). Moreover, while the time-of-day circadian sensitivity to ETI-triggered PCD has been reported previously (49), no mechanism has been proposed for this regulation. Remarkably, consistent with the delinking of C2 and C6 genes from the redox rhythm at 30 °C (Fig. 3 F–H), ETI-mediated PCD is well known to be inhibited at elevated temperatures (50).

To test the circadian control of PCD by the redox rhythm, we used the bacterial pathogen Pseudomonas syringae pv. maculicola ES4326 carrying the effector gene avrRpt2 (Psm ES4326/avrRpt2) to infect the WT, genetic clock-defective CIRCADIAN CLOCK ASSOCIATED 1 overexpressing (CCA1ox) line (51) and the mutant lacking the NB-LRR receptor for avrRpt2, resistant to p. syringae 2 (rps2). Based on a previous study, which showed that, for two consecutive circadian cycles under constant light (LL), ETI-induced PCD was circadian-gated toward the morning in WT (49), we examined the presence or absence of this time-of-day regulation of PCD by performing ETI-induced cell death assays in the subjective morning and subjective evening after 2 d of growth under constant light (LL) to eliminate any transient oscillations in CCA1ox (SI Appendix, Fig. S5 A and B↗). In agreement with our hypothesis, both WT and CCA1ox showed stronger ETI-associated PCD in the subjective morning than in the subjective evening (Fig. 4 D and E), even though CCA1ox has been previously reported to be compromised in the time-of-day-sensitive basal resistance to the bacterial pathogen without the avrRpt2 effector (52). In contrast to the genetic clock-defective CCA1ox, the mutant of 6-PHOSPHOGLUCONOLACTONASE 3 (PGL3), encoding a key enzyme in the oxidative section of the plastidic PPP (Fig. 1A), which has markedly altered glutathione levels and constitutively enhanced resistance (53), had decreased PCD and lost the time-of-day sensitivity to ETI induction (Fig. 4 F and G). This result is in accordance with the finding that a functional PPP is required in the anucleate human red blood cells for maintaining the redox rhythm (10). Since the enhanced basal resistance phenotype of pgl3 is dependent on the constitutively activated NONEXPRESSOR OF PATHOGENESIS-RELATED GENES 1 (NPR1) (53), whose nuclear translocation is redox rhythm-sensitive (12), we also tested the npr1 single and the pgl3 npr1 double mutants for ETI-induced PCD. We found that the npr1 single mutant still had time-of-day sensitivity to immune-triggered PCD, but when combined with pgl3, it could not rescue the PCD phenotype of pgl3 (Fig. 4 F and G), indicating that the lack of circadian PCD in pgl3 is not due to constitutive activation of NPR1-mediated resistance. Finally, we used the toc1 lhy cca1 core clock gene triple mutant (54) as an alternative genetic clock-defective mutant and the pad2 mutant of GSH1 (25) (Fig. 1A) as an alternative redox mutant. We observed that toc1 lhy cca1 still exhibited significant time-of-day difference in PCD levels, in contrast to pad2 (SI Appendix, Fig. S5 C and D↗), further confirming that circadian PCD is driven by the redox rhythm, not the genetic clock.

NPR1 has previously been shown to be involved in salicylic acid (SA)-mediated inhibition of ETI-induced PCD (55) and gating of SA-target gene expression through an increase in the amplitude of the genetic clock (12). However, the circadian gating of PCD observed in the npr1 mutant suggests that this redox rhythm-mediated phenomenon is through an SA/NPR1-independent pathway. It is known that besides SA, other defense hormones, such as JA and ET, also contribute to ETI (56, 57). Among the target genes regulated by the redox rhythm, both EDS1 and PAD4 in C2 (Fig. 4C and Dataset S2↗) are critical regulators of SA and JA crosstalk during ETI (58), even though they are not required for the function of the coiled-coil (CC) class of NB-LRRs (45) used in this study. Furthermore, C2 and C6 also include regulators of immune responses mediated by JA and ET (Fig. 4C and Dataset S2↗). GSNOR1 in C6 is known to both regulate PCD (39, 59) and promote JA-induced defense against herbivores in tobacco (60). SOBIR1 and RECEPTOR LIKE PROTEINs (RLPs) in C2 are involved in PCD (61, 62) and JA/ET-mediated defense against necrotrophic fungi (63). These findings suggest that the circadian redox rhythm may regulate PCD through the JA and/or the ET defense pathways.

The JA defense pathway is normally suppressed by the JASMONATE-ZIM-DOMAIN PROTEIN (JAZ) proteins which are degraded during ETI (56). Downstream of JAZ, the JA pathway contains two mutually antagonistic branches: the JA-only branch, with MYC2 as the central transcription factor (TF) promoting defense against herbivores, and the JA/ET-shared branch, with ETHYLENE-INSENSITIVE3 (EIN3) and ETHYLENE-INSENSITIVE3-LIKE 1 (EIL1) as key TFs protecting against necrotrophic pathogens (64). To determine which of the JA signaling branches is involved in the circadian control of PCD, we used the noninducible jaz1Δjas mutant expressing a JAZ1 protein lacking the Jas degradation domain (65) and key TF mutants myc2 and ein3 eil1 to assess PCD after infiltration with Psm ES4326/avrRpt2. We found that while both the jaz1Δjas and ein3 eil1 mutants lost their time-of-day sensitivity to ETI-triggered PCD, myc2 retained it (Fig. 5 A and B), indicating that an intact JA/ET branch is necessary for circadian regulation of ETI-associated PCD. Similar results were obtained for WT and ein3 eil1 when comparing PCD levels after infection at subjective evening of day 2 and at subjective morning of day 3 under LL, indicating that the circadian redox rhythm, rather than prolonged exposure to constant light, is responsible for this regulation (SI Appendix, Fig. S5 E and F↗).

To further demonstrate the role of the redox rhythm in the circadian regulation of PCD, we used the cell permeable GSH derivative, glutathione monoethyl ester (GSHmee), to perturb the cellular redox rhythm. GSHmee treatment has previously been shown to disrupt the redox rhythm while reinforcing the genetic clock through the activity of NPR1 (12). To exclude a possible influence of GSHmee on the growth of the pathogen, we used the transgenic plants that express the bacterial effector avrRpt2 in response to dexamethasone (dex) treatment to induce PCD. We found that while the treatment in the subjective morning triggered a stronger PCD than in the subjective evening, cotreatment with GSHmee abolished this difference (Fig. 5 C and D). This further links cellular glutathione levels to the time-of-day sensitivity of ETI-induced PCD.

Finally, we used the roGFP2_chl with additional roGFP2 variants localized to the cytoplasm (roGFP2_cyt) and the mitochondrial matrix (roGFP2_mt) (29) to evaluate the subcellular compartment-specific redox contribution to ETI-mediated PCD. Whole-plant time-lapse imaging (Fig. 5 E and F) and subsequent confocal microscopy (SI Appendix, Fig. S6↗) showed that upon ETI-induction, chloroplasts and mitochondria first became strongly reduced at the "primed stage," while cytoplasmic redox remained unchanged compared to mock, even though ETI signaling is associated with ROS production through plasma membrane-associated NADPH oxidase activities (46). However, immediately before the onset of PCD ("onset stage"), we observed a switch to a more oxidized state in chloroplasts and the cytosol, but not in mitochondria. This observation aligns with the recently proposed hypothesis that during ETI, chloroplast stroma is first overreduced due to removal of RuBisCO through autophagy (66), which ultimately helps trigger the production of ROS, making chloroplasts the major source of cellular oxidation during ETI-induced PCD (67).

All Arabidopsis thaliana WT, mutants, and transgenic plants used in this study were in the Columbia-0 ecotype background. Psm ES4326/avrRpt2 was used for PCD assays and Agrobacterium tumefaciens strain GV3101 was used for plant transformation. The plasmids for the luciferase reporter lines were constructed using the Gateway Technology (Thermo Fisher Scientific).

RNA extraction was performed using TRIZOL (Ambion). Probe annealing, ligation, and library construction for RASL-seq was performed as described in ref. 81. For RT-qPCR, UBIQUITIN 5 (UBQ5) mRNA was used for normalization. Primer sequences are provided in SI Appendix, Table S1↗.

The traces for the RASL-seq time-course data were detrended and normalized to ensure that the dominant signature of clustering was the oscillatory pattern. Then, k-means clustering of the genes was performed based on their temporal signature using MATLAB (R2022a) while the hierarchical clustering was done using http://www.bioinformatics.com.cn/srplot↗. The period distribution for each cluster was obtained using either CosinorPY (v3.1) (31) or MetaCycle (37). For individual genes as well as total glutathione and GSSG, harmonic regression was performed using GraphPad Prism 9 as described previously (12).

Glutathione extraction using HCl and colorimetric measurements using the Victor X5 plate reader (PerkinElmer) were performed based on the method described in ref. 82.

The 6-ODE mathematical model by del Olmo et al. (13) was modified using Mathematica (v13.3) by introducing a term for oscillating reduced glutathione scavenging H₂O₂:-g×cos2πtT,

where g is the coupling constant (equal to 0.06 for Fig. 2A) and T is the period.

To induce PCD, either Psm ES4326/avrRpt2 (OD_{600 nm} = 0.02) or solution of 0.25 µM of dex and 0.01% Silwet L-77 (Plantmedia) with or without 30 µM GSHmee (Sigma) was used. Conductivity was measured using an Orion Star series meter (Thermo Scientific).

For luminescence, the plants were imaged using a custom-built chamber as described previously (22) using PIXIS 2048B camera. For fluorescence, a custom-built fluorescence chamber was used with excitation provided by CoolLED pE-4000 system and emission using the Andor iXon Ultra 888 EMCCD Camera. For microscopy, the imaging was done with the Zeiss 880 airyscan inverted microscope. All images were analyzed using ImageJ.

All statistical analyses and harmonic regression were done using GraphPad Prism 9. Data processing was done in MATLAB (R2022a) or R (v4.1.2). GO term analysis was performed using ShinyGO (v0.77) (83). In the graphs, asterisks indicate statistical significance reflecting the P values (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, and ns, not significant). Unless specified, experiments were repeated at least three times with similar results.