Human activities change the Earth’s environment, increasingly more extreme environments result1. Among all abiotic stresses, low temperature is the most important factor limiting agricultural expansion and yield2. Low temperatures and late frosts harm crop production in both temperate and subtropical regions3. Therefore, understanding the key mechanisms of plant cold tolerance can help humans build more resilient food production systems. Cold tolerance in plants can be divided into three types: basic cold tolerance, acquired cold tolerance (plants acquire cold tolerance after transient cold induction) and maintenance of acquired cold tolerance (plants acquire cold tolerance after CS-ACC, and can maintain it even at normal temperature for a long time, also known as CS-memory). Previously, most studies focused on basic and acquired cold tolerance4, while there have been few studies on the maintenance of acquired cold tolerance5,6; this repeated cold stress is more consistent with the natural environment. Plants acquire CS-memory after CS-ACC because of the generation of transcriptional memory (TM)7–9. Currently all known cold acclimation and stress memory genes are not required for CS-memory, which is seriously limited CS-memory mechanism study10.
Our previous studies have found that inhibition of NADPH oxidase/respiratory burst oxidase homolog (RBOHs)-dependent signals through NADPH oxidase inhibitor (Diphenyleneiodonium, DPI) during recovery after CS-ACC reduced the maintenance of acquired cold tolerance11. We got two highly homologous and collinear RBOH genes CsaV3_5G002350 and CsaV3_4G006160 in chromosome 5 and chromosome 4 of cucumber, respectively and both of genes’ expression significantly increased during recovery after CS-ACC11. Moreover, CsaV3_5G002350 of cucumber had collinearity with Solyc08g081690 (RBOH1) of tomato and AT1G64060 (RBOHsF) of Arabidopsis. Both Solyc08g081690 and AT1G64060 have been shown to be involved in the regulation of abiotic stress12–14. Therefore, we speculated that CsaV3_5G002350 in cucumber may play a leading role in CS-ACC. Since chromosome 5 has two RBOH genes and CsaV3_5G002350 is the first location in the genome, we named it CsRBOH5.1.
We then knocked out CsRBOH5.1 using the clustered regularly interspaced short palindromic repeats/CRISPR-associated protein 9 (CRISPR-Cas9) (Extended Data Fig. 1a). Here about 6000 cucumber seeds were used and 5 positive transgenic plants were screened through continuous observation of GFP fluorescence in the differentiation and culture stage, which showed a conversion efficiency of about 0.8‰ (Extended Data Fig. 1b-c). Compared with WT, significant green GFP fluorescence were observed in tendrils, anthers, fruits (ovary) and seeds of transgenic cucumber (Extended Data Fig. 1c). A total of nine types of Csrboh5.1-edited lines were obtained (Extended Data Fig. 2a), of which we further screened 3 lines (-1 bp, Csrboh5.1-1; -31 bp, Csrboh5.1-2; -4 bp, Csrboh5.1-3; Extended Data Fig. 2b) by PCR amplification sequencing and GFP fluorescence that both undergo gene editing (multiple of non 3) and without the cas9 editor (GFP fluorescence) for the following breeding and tests. The transgenic seeds used here were T3 generation.
We then investigated whether knocking out CsRBOH5.1 affected CS-ACC of cucumber seedlings. The cucumber seedlings were subjected to a CS-ACC regime of 1 h at 10°C, 1.5 h at 25°C, and 3 h at 1°C. Following a long recovery (48 h, 25°C) after CS-ACC, a tester CS was applied for 24 h at 1°C (Fig. 1a). The cold tolerance of wild-type (WT) treated with CS-ACC was significantly increased with a lower chilling injury index (CII) and higher survival rate, than that of the nontreatment control (CK; unacclimated) (Fig. 1b-d), suggesting that CS-memory was preserved in cucumber during recovery after CS-ACC. However, knocking out CsRBOH5.1 made cucumber seedlings deacclimation. The CII of the three mutant lines were significantly higher than that of the WT and the survival rate were significantly lower than that of the WT and nontreatment control (Fig. 1b-d). We also took samples at 4 h and 48 h during the recovery after CS-ACC as well as at 48 h of nontreatment control and conducted RNA-seq (Extended Data Table 1a) to detect whether cucumber seedlings formed CS-TM. The biological replicates of these samples were good (Extended Data Fig. 3). We classify cold response genes into CS-memory and CS-induction genes (Extended Data Table 2). We referred to the up-up and up-no change groups as up-memory genes (um: 72), down-down and down-no change groups as down-memory genes (dm: 103), and the up-down and down-up groups as CS-induction genes (Fig. 2a and Extended Data Table 2). To further investigate the function of these CS-responsive genes, GO and KEGG enrichment analyses were performed (Fig. 2c-d and Extended Data Table 3). We selected the top 10 enriched GO terms and all enriched KEGG pathways for further analysis. The GO terms and KEGG pathways of the up-memory group were mainly related to ribosome or ribosome synthesis. The KEGG pathway of the down-memory group was related to endoplasmic reticulum protein processing. These memory genes related to protein synthesis may be the key to obtain CS-ACC and enhance cold tolerance15–18. While, as for CS-induction genes, the up-down group were mainly related to proteasome, peptidase, antioxidant substance (phenylalanine) and nucleotide and ATP binding and the down-up group were mainly related to photosynthesis.
When we knocked out CsRBOH5.1, about 70% of CS-memory genes unable to respond to CS-ACC and other 30% cannot maintained during recovery, completely destroyed the CS-TM (Fig. 2b). Except for CS-memory genes, transcription of CS-induction genes was always affected. For example, 50% of up-down genes unable to respond to CS-ACC, and 38% of that was down regulated during recovery. Moreover, about 20% of non-response genes in WT changed to cold responsive genes in Csrboh5.1 mutant (Fig. 2b). The above results indicated that CsRBOH5.1 controlled most of the expression of cold-responsive genes during recovery after CS-ACC and was essential for cucumber in responding CS-ACC.
We also studied some key cold-corresponding genes such as CsICEs, CsCBFs and CsCORs (Extended Data Fig. 4; Extended Data Table 4). Previous study has found that the expression of CsICEs-CsCBFs-CsCORs under CS significantly enhanced in cucumber19,20. However, compared with unacclimated seedlings, none of them enhanced during recovery at 4 h after CS-ACC (Extended Data Table 4b), indicated that CsICEs-CsCBFs-CsCORs mainly responded to the basal cold tolerance and cannot maintain response during short recovery even for 4 h. We then compared the expression of these genes between WT and Csrboh5.1 in CK, R4h and R48h, respectively, and found that CsRBOH5.1 knockout significantly decreased the expression of CsICE1, CsICE1-like and CsCBF2 (Extended Data Fig. 4). We further compared the changes of these genes between unacclimated and during recovery after CS-ACC in WT or Csrboh5.1 mutants. Knocking out CsRBOH5.1 resulted the expression of CsICE1 significantly decreased during recovery at 4 h after CS-ACC (Extended Data Table 4b), showed similar result in tomato ICE1 expression, when RBOH1 knockout21.
Since knocking out CsRBOH5.1 resulted in changes of almost global gene transcription, we hypothesized that CsRBOH5.1 must control some key regulatory factors that regulates gene expression. Because H3K4me3/H3K27me3 is widely deposited on genes that widely regulates gene expression, we asked whether CsRBOH5.1 can mediate H3K4me3/H3K27me3 affecting gene expression and then conducted chromatin immunoprecipitation followed by sequencing (ChIP-seq) for H3K4me3 or H3K27me3 at 4 h and 48 h during the recovery after CS-ACC, as well as at 48h of nontreatment control (Extended Data Table 1b). The biological replicates of these samples were good (Extended Data Fig. 5). We used biological replicates to obtain overlapping peaks for H3K4me3 and H3K27me3, respectively (Extended Data Table 5). The same to previous studies, H3K27me3 and H3K4me3 showed negative and positive correlations with the level of gene expression respectively22 (Extended Data Fig. 6). We then analyzed the presence of H3K4me3 or H3K27me3 peaks on genes. We found that 80%-95% of genes had H3K4me3 peaks, which was much higher than the results for H3K27me3 (Extended Data Fig. 7), suggesting that cucumber preferred H3K4me3 positively regulated gene transcription, which was consistent with previous study that H3K4me3 prevailed over H3K27me3 in gene regulation23,24. Moreover, there was no difference in the number of genes containing H3K4me3 among the different types of genes (Extended Data Fig. 7a). While, CsRBOH5.1 knockout did not change the number of genes containing H3K4me3 peaks. Thus, we further investigated the genetic distribution of H3K4me3 peak. We found that H3K4me3 peak frequencies in the regions from the TSS to 1500 bp on up-memory genes were significantly higher than other types of genes (Extended Data Fig. 8). Similarly, CsRBOH5.1 knockout did not change the H3K4me3 peak frequencies during recovery after CS-ACC on all types of genes, excepted for up-memory genes in unacclimated plants that significantly enhanced (Extended Data Fig. 8b-c).
We then asked whether CsRBOH5.1 knockout affected H3K4me3 peak signal. We noticed that H3K4me3 on up-memory gene mainly enriched in the promoter and the first 2/3 of the transcription start site (TSS) regions (Extended Data Fig. 9). Therefore, we divided the gene into two segments, including ‘promoter + 2/3TSS’ and the remaining regions (near ‘transcription end site’ (TES’)). As for ‘promoter + 2/3TSS’ region, we found that CsRBOH5.1 knockout decreased the H3K4me3 signal of up-down and down-memory genes, but increased the H3K4me3 signal of other types of genes in unacclimated plants (Fig. 3a). However, all types of genes were sharply decreased during recovery after CS-ACC, and the H3K4me3 peak signal was further weakened with the prolonged recovery time (Fig. 3b-c). For example, CsRBOH5.1 knockout reduced the H3K4me3 signal of all types of genes by 3–6% at 4 h recovery, and by 9–16% at 48 h recovery. We also detected the H3K4me3 on gene near TES region (Extended Data Fig. 10). We get almost similar result to ‘promoter + 2/3TSS’ region. These results strongly suggested that CsRBOH5.1 mainly decreased H3K4me3 deposition on genes during recovery after CS-ACC and this effect was more obvious with the extension of recovery time.
We then compared H3K4me3 deposition on genes between unacclimated and CS-ACC plants (Fig. 3d-h). We found that only the up-memory gene showed H3K4me3 epigenetic memory during recovery after CS-ACC (Fig. 3d). Then we investigated whether CsRBOH5.1 was essential for this epigenetic memory. We found that CsRBOH5.1 knockout did not reduce the H3K4me3 level of up-memory gene in unacclimated plants (Fig. 3a), but significantly reduced that during recovery in acclimated plants (Fig. 3b-d), completely destroying CS-TM (Fig. 2a-b). While as for heat stress memory, HsfA2 was the key to regulate H3K4me3 deposition, forming heat stress-TM25, which was different from CS-memory. And currently all known cold acclimation and stress memory genes are not required for CS-memory10. While, CsRBOH5.1 was a key gene for CS-memory through H3K4me3 regulation. Surprisingly, CsRBOH5.1 knockout also reduced H3K4me3 in CS-induction and even non-cold-responsive genes during recovery after CS-ACC (Fig. 3f-h), severely disrupting gene transcription (Fig. 2a-b).
We then analyzed the H3K4me3 signaling of each gene at the genome-wide level. We found that for unacclimated seedlings, CsRBOH5.1 knockout significantly increased H3K4me3 deposition. Compared with WT, 82.57% of genes were significantly up-regulated for H3K4me3 in Csrboh5.1 mutants (Fig. 3i). However, in the recovery stage after CS-ACC, 52%-68% of genes in Csrboh5.1 mutants decreased significantly for H3K4me3 compared with WT (Fig. 3j-k), and the number of decreased genes increased with the prolongation of recovery time after CS-ACC, which was consistent with the results in Fig. 3b-c. We then analyzed the H3K4me3 changes in WT and Csrboh5.1 mutants during recovery after CS-ACC, respectively. The results showed that after CS-ACC, H3K4me3 on 56.3% of genes in R4h (Fig. 3l) and on 63.5% of genes in R48h (Fig. 3m) of WT were significantly decreased compared with unacclimated plants (CK); and compared with R4h, 57.4% of the genes in R48h (Fig. 3n) had a significant decrease in H3K4me3. However, when Csrboh5.1 knockout, H3K4me3 on 94.2% of genes in R4h (Fig. 3o) and on 97.2% of genes in R48h (Fig. 3p) were significantly decreased compared with unacclimated plants. Moreover, compared with R4h, 77.1% of the genes in R48h (Fig. 3q) had a significant decrease in H3K4me3. Therefore, knocking out CsRBOH5.1 significantly reduced H3K4me3 levels of genes during recovery after CS-ACC, and the decrease was more obvious with the extension of recovery time.
We then detected the expression of genes related to H3K4me3 regulation during recovery after CS-ACC, including histone methyltransferase “writers”26–29, histone demethylase “erasers”30–33 and the histone modification “readers”9,34,35. We first compared the expression of these genes between WT and Csrboh5.1 in CK, R4h and R48h, respectively. We found that CsRBOH5.1 knockout significantly decreased the expression of most of these genes no matter for unacclimated seedlings or during recovery after CS-ACC compared with WT (Extended Data Fig. 11a-c and Extended Data Table 6a). We further compared the changes of these genes between unacclimated and during recovery after CS-ACC in WT or Csrboh5.1 mutants. We found that CsRBOH5.1 knockout significantly decreased the expression of CsATXR2 (histone methyltransferase) during recovery at 48 h after CS-ACC; however, the expression of some histone methyltransferase in WT including CsATXR6 and CsATX3 was significantly enhanced but not enhanced in Csrboh5.1 mutants during recovery at 48h after CS-ACC (Extended Data Table 6b). This was consistent with the result that H3K4me3 signal in WT increased at 48h during recovery after CS-ACC, but decreased in the Csrboh5.1 mutant (Fig. 3d-h). In general, CsRBOH5.1 was a key gene for genes to rebuild H3K4me3 deposition during recovery after CS-ACC, especially at 48h during recovery.
We further asked the difference between CS-memory and CS-induction gene in transcriptional regulation. We found that no matter unacclimated plants or plants during recovery after CS-ACC, H3K4me3 peak signals on up-memory genes were higher than that on down-memory genes; however, H3K4me3 peak signals on up-down genes were lower than that on down-up genes (Fig. 3a-c and Extended Data Fig. 9–10). Moreover, compared with other types of genes, H3K4me3 peak deposition on up-memory genes during recovery after CS-ACC were the highest (Fig. 3b-c and Extended Data Fig. 8), indicated that up-memory genes always maintained an easily transcribed H3K4me3 environment. Furthermore, although cucumber gene deposited little H3K27me3 peaks (Extended Data Fig. 7b), we also investigated H3K27me3 signal (total signal) of different types of genes in different samples. We found a converse results with the regulation of H3K4me3. H3K27me3 on up-memory genes were lower than that on down-memory genes; however, H3K27me3 on up-down genes were higher than that on down-up genes (Extended Data Fig. 12). Moreover, H3K27me3 deposition on up-memory genes and down-memory genes during recovery after CS-ACC were the lowest and highest, respectively, which further showed that CS-memory gene had a beneficial transcription environment in the genome. In general, deposition pattern of H3K4me3/H3K27me3 on CS-memory genes always contributed to CS-TM, while on CS-induction genes always went against their transcription pattern formation, showing a natural inherent property even after CS-ACC. While, CsRBOH5.1 knockout did not change this natural inherent property, even though CsRBOH5.1 knockout sharply decreased the H3K4me3 signal on genes, suggesting that this property was very conservative in cucumber.
In summary, we for the first time in plants, found that CsRBOH5.1 could regulate global genes’ H3K4me3 deposition and was essential for genes to rebuild H3K4me3 deposition during recovery after CS-ACC, which lays a foundation for human to regulate the whole genome H3K4me3 of plants in the future. Furthermore, we also found that the deposition pattern of H3K4me3/H3K27me3 on CS-memory genes always contributed to CS-TM, while on CS-induction genes always went against their transcription pattern formation, showing a natural inherent and conservative property.