Integrate Small RNA and Degradome Sequencing to Reveal Drought Memory Response in Wheat (Triticum aestivum L.)

Drought has gradually become one of the most severe abiotic stresses on plants. Plants that experience stress training can exhibit enhanced stress tolerance. According to MicroRNA (miRNA) sequencing data, this study identified 195 candidate drought memory-related miRNAs in wheat, and targets of 64 (32.8%) candidate miRNAs were validated by degradome sequencing. Several drought memory-related miRNAs such as tae-miR9676-5p, tae-MIR9676-p3_1ss21GA, tae-miR171a, tae-miR531_L-2, tae-miR408_L-1, PC-3p-5049_3565, tae-miR396c-5p, tae-miR9778, tae-miR164a-5p, and tae-miR9662a-3p were validated as having a strong response to drought memory by regulating the expression of their target genes. In addition, overexpression of drought memory-related miRNA, tae-miR531_L-2, can remarkably improve the drought tolerance of transgenic Arabidopsis thaliana. Drought memory can regulate plant cellular signal transduction, plant biosynthetic processes, and other biological processes to cope with drought via transcriptional memory. In addition, drought memory-related miRNAs can promote starch and sucrose catabolism and soluble sugar accumulation and regulate proline homeostasis to improve plant drought resistance. Our results could contribute to an understanding of drought memory in wheat seedlings and may provide a new strategy for drought-resistant breeding.


Introduction
Drought is one of the most critical threats to agriculture because of the increased water resource scarcity [1]. As the global climate changes, the frequency, severity, and extent of drought tend to further aggravate agriculture [2]. Wheat (Triticum aestivum L.) is the most widely cultivated crop, grown on 210 million hectares, and provides 30% of the food consumed worldwide [3]. It is estimated that by 2050, global wheat yields must increase by 60% to meet the needs of a rapidly growing population [4]. However, wheat production has been falling significantly due to various abiotic stresses, especially drought [5].
Because the environment is constantly changing, plants may experience recurrent environmental stresses throughout their entire life cycle [6]. The molecular mechanisms of plant response to similar abiotic stresses are different from its first encounter [7]. It was reported that many plants could enhance their tolerance to the same pressure [8][9][10]. These quicker and more aggressive responses include structural, genetic, and biochemical modifications that may be defined as stress memory [8,11]. A growing body of research has focused on stress memory in plants [12]. In Arabidopsis thaliana (L.) Heynh, stress memory develops through a phenomenon similar to genomic imprinting when exposed to stress [13], and the epigenetic regulation of the proline biosynthetic genes in developing stress memory represents a new research area [14]. Stress memory has a genetic effect on plants. The cold memory was inherited maternally but not paternally in the zygote and early embryos [15]. The H3K27me3 demethylase-HSFA2 regulatory loop can orchestrate

The Influence of Drought Memory to Plant Drought Resistance
The drought resistance of wheat seedlings that experienced drought training previously could be improved ( Figure 1A). The proline content and Phi2 of the samples following three treatments at 0, 24, 48, and 72 h were measured ( Figure 1C,D). The proline accumulation in the DM group was significantly higher than CG after 24 h and 72 h drought treatment and lower than in the DD group at all time points. In addition, the photosynthetic efficiency of the DM group was significantly higher than the DD group. This observation means that drought memory pre-treated wheat under severe drought conditions is less damaged than non-drought-trained wheat. Therefore, drought memory wheat needs less proline to maintain cellular homeostasis and preserve higher photosynthetic efficiency than non-drought pre-treated wheat.
needs less proline to maintain cellular homeostasis and preserve higher photosynthetic efficiency than non-drought pre-treated wheat. Figure 1. Experimental design and physiological index measurement. (A) Drought resistance detection of seedlings in three groups. Trileaf stage wheat seedlings in DD and DM groups were treated with half-strength Hoagland's liquid medium containing 19.2% PEG for six days, and seedlings grown in normal conditions were used as control. (B) Trileaf stage seedlings were divided into three groups, including the control (CG), direct drought (DD), and drought memory (DM). The DM group was pre-treated with 10% (M/V) PEG. Seedlings of the DD and DM groups were then treated with 19.2% PEG, while CG was grown under normal conditions. Whole seedlings of three groups were collected at 0, 1, 6, and 12 h with three biological repetitions and immediately frozen in liquid nitrogen. (C) Proline content comparison amongst the three groups after treatment (19.2% (M/V) PEG) for 0, 24, 48, and 72 h. * p < 0.05; ** p < 0.01. (D) Comparison of the efficiency of photosystem II (Phi2) amongst the three groups after treatment (19.2% (M/V) PEG) for 0, 24, 48, and 72 h. Error bars represent the SD. Points on the polyline followed by different letters are statistically different according to the analysis of variance followed by Duncan's Multiple Range Test (Comparison of different treatments at the same time).

High-Throughput Sequencing of miRNAs
Raw reads were first stripped of the 3′ connector sequence and the low-quality reads (Table S1). After removing the PCR duplications and data filtering, 68.9 million unique raw reads were obtained. After filtering by comparing with the mRNA, Rfam, and Repbase databases, 30.0 million unique high-quality valid reads were retained for further analysis (Table S2). Twenty-four nucleotides miRNA represented the highest abundance and accounted for 49.4% of all detected miRNAs. Different length miRNAs belonging to the same miRNA family may have diverse functions [35]. The miRNA length proportion differed amongst the three groups and four time points, suggesting that varying length miRNAs may play diverse roles in drought tolerance and drought memory ( Figure S1A).

Identification of Conserved, Known and Novel miRNAs
Unique sequences of 18-25 nucleotides (nt) were mapped to wheat precursors of miRBase 22.0 using a BLAST search to identify miRNAs (Table 1, Figure S1D). Our results Figure 1. Experimental design and physiological index measurement. (A) Drought resistance detection of seedlings in three groups. Trileaf stage wheat seedlings in DD and DM groups were treated with half-strength Hoagland's liquid medium containing 19.2% PEG for six days, and seedlings grown in normal conditions were used as control. (B) Trileaf stage seedlings were divided into three groups, including the control (CG), direct drought (DD), and drought memory (DM). The DM group was pre-treated with 10% (M/V) PEG. Seedlings of the DD and DM groups were then treated with 19.2% PEG, while CG was grown under normal conditions. Whole seedlings of three groups were collected at 0, 1, 6, and 12 h with three biological repetitions and immediately frozen in liquid nitrogen. (C) Proline content comparison amongst the three groups after treatment (19.2% (M/V) PEG) for 0, 24, 48, and 72 h. * p < 0.05; ** p < 0.01. (D) Comparison of the efficiency of photosystem II (Phi2) amongst the three groups after treatment (19.2% (M/V) PEG) for 0, 24, 48, and 72 h. Error bars represent the SD. Points on the polyline followed by different letters are statistically different according to the analysis of variance followed by Duncan's Multiple Range Test (Comparison of different treatments at the same time).

High-Throughput Sequencing of miRNAs
Raw reads were first stripped of the 3 connector sequence and the low-quality reads (Table S1). After removing the PCR duplications and data filtering, 68.9 million unique raw reads were obtained. After filtering by comparing with the mRNA, Rfam, and Repbase databases, 30.0 million unique high-quality valid reads were retained for further analysis (Table S2). Twenty-four nucleotides miRNA represented the highest abundance and accounted for 49.4% of all detected miRNAs. Different length miRNAs belonging to the same miRNA family may have diverse functions [35]. The miRNA length proportion differed amongst the three groups and four time points, suggesting that varying length miRNAs may play diverse roles in drought tolerance and drought memory ( Figure S1A).

Identification of Conserved, Known and Novel miRNAs
Unique sequences of 18-25 nucleotides (nt) were mapped to wheat precursors of miRBase 22.0 using a BLAST search to identify miRNAs (Table 1, Figure S1D). Our results identified 644 known miRNAs (gp1a) and 837 conserved miRNAs (gp1b, gp2a, gp2b, and gp3). Amongst these, 141 known miRNAs and 222 conserved miRNAs exist in all libraries (Table S3). After eliminating low expressed miRNAs, 146 novel miRNAs were predicted, of which 61 miRNAs were common to all of the 33 miRNA libraries (Table S4). Except for miRNAs PC-3p-1639_10290 and PC-5p-576_26392, which were expressed at high levels, most of the novel miRNAs exhibited moderate expression level categories. The base distribution for each position of known, conserved, and novel miRNAs was analyzed to investigate base preference. We also found that the first base preference is largely relevant to the length of the miRNA ( Figure S1B,C,E,F).

Identification of Drought Memory-Related miRNAs
After drought pretreatment, a total of 198 significantly (p-value < 0.05) differentially expressed miRNAs were found (DM_0 h vs. CG_0 h). Of these, 75 miRNAs were downregulated, and the remainder were up-regulated expressed. Furthermore, 18 differentially expressed miRNAs exhibited a high expression level ( Figure S2A). Simultaneously, 228 direct drought response-related miRNAs (union set of DD_1 h vs. CG_1 h, DD_6 h vs. CG_6 h and DD_12 h vs. CG_12 h) were identified ( Figure S3).
Direct drought resistance and drought memory are two kinds of drought response mechanisms, and there are both similarities and differences between them. After pretreatment with a periodic repetition of low-level drought (10% PEG), wheat may respond to drought more quickly or aggressively through structural, genetic, and biochemical modifications. When the pre-treated wheat seedlings suffered a higher level of drought stress (19.2% PEG), 243 miRNAs (union set of DM_1 h vs. CG_1 h, DM_6 h vs. CG_6 h, and DM_12 h vs. CG_12 h) were significantly differentially expressed, and they included 128 known and conserved miRNAs and 115 novel miRNAs (Figures 2A and S2B-D). However, when drought pretreatment plants encounter a persistent severe drought, miRNAs related to drought memory and direct drought response are both expressed. Consequently, 57 miRNAs that were common to direct drought and drought memory were removed from the 243 drought memory-related miRNAs. Furthermore, nine miRNAs, included in 57 shared miRNAs but significantly differentially expressed between DM and DD groups, were also considered drought memory-related miRNAs. Finally, 195 miRNAs, including 186 drought memory-specific and nine significantly differentially expressed shared miRNAs, were identified as candidate drought memory-related miRNAs (Figure 2A

Validation of Target Genes of Drought Memory-Related miRNAs in Wheat Using Degradome Sequencing
Degradome sequencing analysis was conducted to narrow down the prediction range of miRNA-target pairs and avoid false positives caused by the prediction software The unique transcript-mapped reads were obtained from the CG, DD, and DM libraries respectively (Table S5). Cleavage classifies 'target plots' (T-plots) peaks into four catego ries (0-4), with peaks of categories 0-3 having >1 read per peak, and examples of T-plo for the targets of wheat miRNAs are shown in Figure S4.
Target genes of 195 candidate drought memory-related miRNAs were also validated in the degradome results, with categories <3 as a threshold. Sixty-four (32.8%) candidat drought memory-related miRNAs with 445 target transcripts were detected (Table S6) Among them, 12 miRNAs were expressed at a high level, 37 at a middle level, and 15 at low level. It is worth mentioning that eight novel miRNAs were also detected, and th pre-miRNA stem-loop secondary structure of these putative drought memory-related miRNAs is shown in Figure S5A. All the mature sequences of the novel wheat miRNA are observed in the stem regions of the stem-loop structures. Furthermore, 19 miRNA significantly differentially expressed in pretreatment drought (DM_0 h vs. CG_0 h) can also be detected in these 64 validated drought memory-related miRNAs ( Figure S5B). Enrichment analysis revealed that regulation of transcription, DNA-templated (GO:0006355), DNA binding (GO:0003677), and nucleus (GO:0005634) were the top thre

Validation of Target Genes of Drought Memory-Related miRNAs in Wheat Using Degradome Sequencing
Degradome sequencing analysis was conducted to narrow down the prediction range of miRNA-target pairs and avoid false positives caused by the prediction software. The unique transcript-mapped reads were obtained from the CG, DD, and DM libraries, respectively (Table S5). Cleavage classifies 'target plots' (T-plots) peaks into four categories (0-4), with peaks of categories 0-3 having >1 read per peak, and examples of T-plot for the targets of wheat miRNAs are shown in Figure S4.
Target genes of 195 candidate drought memory-related miRNAs were also validated in the degradome results, with categories <3 as a threshold. Sixty-four (32.8%) candidate drought memory-related miRNAs with 445 target transcripts were detected (Table S6). Among them, 12 miRNAs were expressed at a high level, 37 at a middle level, and 15 at a low level. It is worth mentioning that eight novel miRNAs were also detected, and the pre-miRNA stem-loop secondary structure of these putative drought memory-related miRNAs is shown in Figure S5A. All the mature sequences of the novel wheat miRNAs are observed in the stem regions of the stem-loop structures. Furthermore, 19 miRNAs significantly differentially expressed in pretreatment drought (DM_0 h vs. CG_0 h) can also be detected in these 64 validated drought memory-related miRNAs ( Figure S5B). Enrichment analysis revealed that regulation of transcription, DNA-templated (GO:0006355), DNA binding (GO:0003677), and nucleus (GO:0005634) were the top three GO terms that were significantly enriched in the biological process, molecular function, and cellular component, respectively. Targets of drought memory-related miR-NAs were annotated to respond to the hormone, especially the auxin-activated signaling pathway (GO:0009734). Furthermore, drought memory-related transcripts were annotated to ion transport GO terms, such as calcium ion transport (GO:0006816), cadmium ion transport (GO:0015691), iron ion transport (GO:0006811), and response to zinc ion (GO:0010043). Furthermore, the GO terms of histone modification (GO:0016570), seed development (GO:0048316), and leaf development (GO:0048366) were also found in our results ( Figure S5C). Phenylalanine metabolism (ko00360), pyruvate metabolism (ko00620), phenylpropanoid biosynthesis (ko00940), and fructose and mannose metabolism (ko00051) are four significantly enriched KEGG pathways. Target transcript of the drought memory-related miRNA tae-miR164a-5p annotated to the environmental information processing-mTOR signaling pathway (ko04150) and targets of the tae-MIR1862d-p3_1ss20TC are taking part in the glutathione metabolism pathway (ko00480). In addition, the glucose metabolism and amino acid metabolism pathways were also enriched in our results ( Figure S5D).

Validation of Drought Memory-Related miRNAs and Their Target Genes by RT-qPCR
Ten drought memory-related miRNAs, including nine known and conserved miRNAs and one novel miRNA, and their target genes were selected for validation by RT-qPCR based on the abundance and predicted functions ( Figure 3). GO terms that were significantly enriched in the biological process, molecular function, and cellular component, respectively. Targets of drought memory-related miRNAs were annotated to respond to the hormone, especially the auxin-activated signaling pathway (GO:0009734). Furthermore, drought memory-related transcripts were annotated to ion transport GO terms, such as calcium ion transport (GO:0006816), cadmium ion transport (GO:0015691), iron ion transport (GO:0006811), and response to zinc ion (GO:0010043). Furthermore, the GO terms of histone modification (GO:0016570), seed development (GO:0048316), and leaf development (GO:0048366) were also found in our results ( Figure  S5C). Phenylalanine metabolism (ko00360), pyruvate metabolism (ko00620), phenylpropanoid biosynthesis (ko00940), and fructose and mannose metabolism (ko00051) are four significantly enriched KEGG pathways. Target transcript of the drought memory-related miRNA tae-miR164a-5p annotated to the environmental information processing-mTOR signaling pathway (ko04150) and targets of the tae-MIR1862d-p3_1ss20TC are taking part in the glutathione metabolism pathway (ko00480). In addition, the glucose metabolism and amino acid metabolism pathways were also enriched in our results ( Figure S5D).

Validation of Drought Memory-Related miRNAs and Their Target Genes by RT-qPCR
Ten drought memory-related miRNAs, including nine known and conserved miR-NAs and one novel miRNA, and their target genes were selected for validation by RT-qPCR based on the abundance and predicted functions ( Figure 3). (A-J) The relative expression level of ten candidate drought memory-related miRNAs and their target genes. The relative expression level of each miRNA and their target genes was compared between DD and DM groups at 0, 1, 6, and 12 h, with the CG as control. The expression levels of the miRNAs and target genes were normalized using U6 (GenBank: X63066.1) and Elongation Factor 1-Alpha genes, respectively. All reactions were performed in triplicate for each sample, and (A-J) The relative expression level of ten candidate drought memory-related miRNAs and their target genes. The relative expression level of each miRNA and their target genes was compared between DD and DM groups at 0, 1, 6, and 12 h, with the CG as control. The expression levels of the miRNAs and target genes were normalized using U6 (GenBank: X63066.1) and Elongation Factor 1-Alpha genes, respectively. All reactions were performed in triplicate for each sample, and expression levels were calculated according to the 2 −∆∆CT method. Error bars represent the SD, * p < 0.05; ** p < 0.01.
Our results indicated that the expression patterns of five drought memory-related miRNAs, including tae-miR9676-5p, tae-miR408_L-1, PC-3p-5049_3565, tae-miR9778, and tae-miR164a-5p, were similar between the DD and DM groups; however, their expression levels were significantly different between the two groups, and the remaining five miRNAs showed opposite expression patterns between groups DD and DM. Furthermore, the expression patterns of almost all miRNA-target pairs were contrasting in the DM group, except for tae-miR396c-5p and tae-miR9778. These results confirm the reliability of our sequencing results and indicate that selected candidate miRNAs and their target genes may play crucial roles in drought memory. Moreover, the expression level of drought memory-related miRNAs revealed that drought memory-induced drought response and direct drought response were two different but interrelated processes.

Overexpression of the tae-miR531_L-2 Precursor in Arabidopsis thaliana
The precursor of tae-miR531_L-2 was overexpressed in Arabidopsis thaliana to evaluate the roles of drought memory-related miRNAs in plant drought resistance.
Our results found that tae-miR531_L-2 was significantly overexpressed in transgenic Arabidopsis thaliana, and the expression level was significantly higher than WT under drought stress. In addition, tae-miR531_L-2 can also strongly respond to cold and salt stresses, but there was no significant response to heat stress ( Figure 4A,B). Our phenotypic experiment demonstrated that the overexpression of drought memory-related miRNA tae-miR531_L-2 could significantly improve the drought resistance of Arabidopsis thaliana in many ways. It can also considerably enhance the seed germination and seedling survival rates under drought stress ( Figure 4C). On the other hand, the overexpression of tae-miR531_L-2 could not only remarkably improve the survival rate of Arabidopsis thaliana under severe soil drought but also considerably improve the seedling biomass and seedling growth rate under prolonged mild drought conditions ( Figure 4D). Our results indicated that the expression patterns of five drought memory-related miRNAs, including tae-miR9676-5p, tae-miR408_L-1, PC-3p-5049_3565, tae-miR9778, and tae-miR164a-5p, were similar between the DD and DM groups; however, their expression levels were significantly different between the two groups, and the remaining five miR-NAs showed opposite expression patterns between groups DD and DM. Furthermore, the expression patterns of almost all miRNA-target pairs were contrasting in the DM group, except for tae-miR396c-5p and tae-miR9778. These results confirm the reliability of our sequencing results and indicate that selected candidate miRNAs and their target genes may play crucial roles in drought memory. Moreover, the expression level of drought memoryrelated miRNAs revealed that drought memory-induced drought response and direct drought response were two different but interrelated processes.

Overexpression of the tae-miR531_L-2 Precursor in Arabidopsis thaliana
The precursor of tae-miR531_L-2 was overexpressed in Arabidopsis thaliana to evaluate the roles of drought memory-related miRNAs in plant drought resistance.
Our results found that tae-miR531_L-2 was significantly overexpressed in transgenic Arabidopsis thaliana, and the expression level was significantly higher than WT under drought stress. In addition, tae-miR531_L-2 can also strongly respond to cold and salt stresses, but there was no significant response to heat stress ( Figure 4A,B). Our phenotypic experiment demonstrated that the overexpression of drought memory-related miRNA tae-miR531_L-2 could significantly improve the drought resistance of Arabidopsis thaliana in many ways. It can also considerably enhance the seed germination and seedling survival rates under drought stress ( Figure 4C). On the other hand, the overexpression of tae-miR531_L-2 could not only remarkably improve the survival rate of Arabidopsis thaliana under severe soil drought but also considerably improve the seedling biomass and seedling growth rate under prolonged mild drought conditions ( Figure 4D).  (D) Drought resistance detection in the seedling stage. Two-week-old soil-grown seedlings were air-dried and then rehydrated to detect extreme drought resistance. The seven-day-old soil-grown seedlings were watered with 10% PEG solution for two weeks for drought resistance detection under prolonged mild drought.

Identification of Drought Memory-Related miRNAs and Their Target Genes in Wheat
Drought memory is a complex biological process regulated by the expression of many genes. Previous studies have reported that miR156, miR391, and miR831 are heat stress memory-related miRNAs in Arabidopsis thaliana, and miR398 and miR408 were associated with drought memory in coffee plants [36,37]. Our results suggest that some of the above stress memory-related miRNAs are equally important in drought memory-induced drought resistance. In this study, the tae-miR408_L-1 may be involved in drought memory by regulating the expression of genes associated with the electron transport chain and metal ion binding. Some miRNAs involved in direct drought may also play a role in drought memory. Accumulation studies have shown that many miRNAs, including miRNA408, 171, 164, and 396, play essential roles in the drought stress resistance of plants [38,39]. Our results found that the drought memory-related miRNA tae-miR408_L-1 is indispensable in drought memory and direct drought response. In addition, the drought memory-related miRNA tae-miR164a-5p could target MAP kinase activity (GO:0004707) and mTOR signaling pathway (ko04150) related transcripts, which might improve the drought tolerance of wheat [40,41]. Target transcript of tae-miR396c-5p was associated not only with abscisic acid (GO:0009737) and mannitol (GO:0010555) but were also positively regulated in response to water deprivation (GO:1902584). ABA is a crucial drought memory-related plant hormone of rice [21]. The miR171 in rice might improve salt tolerance by modulating physiological changes, stomatal development, ABA-dependent pathways, and expression of stressrelated genes [42]. Our results showed that the drought memory-related miRNA, tae-miR171a, could respond to both direct drought and drought memory, and the expression patterns were different from each other ( Figure 3C).
Our study also identified some new drought memory-related miRNAs. A novel drought memory-related miRNA, PC-3p-5049_3565, was detected, and it targeted fructose bisphosphate aldolase (FBA). It has been reported that FBA is a salt stress-responsive marker protein in chloroplasts that catalyzes the cleavage of fructose-1-6-bisphosphate into D-glyceraldehyde-3-phosphate, thus producing dihydroxyacetone phosphate and adenosine triphosphate [43]. Moreover, FBA may be one of the target proteins responsible for Al stress in wheat roots [44]. The function of FBA in drought stress memory in wheat deserves to be studied further.

miRNA-Gene-GO Association Analysis of Drought Memory-Related miRNAs
Enrichment analysis was conducted amongst target transcripts of 64 drought memoryrelated miRNAs and used 42 significantly (p < 0.05) enriched GO terms for association analysis ( Figure 5). The results revealed that drought memory displayed a transcriptional stress memory (GO:0003677, GO:0006355, GO:0006351, and GO:0005634). Transcriptional stress memory can respond to drought by regulating other specific functions. The present study found that drought memory wheat may respond to severe drought by regulating plant cellular signal transduction. Firstly, drought memory might regulate the multiple signaling pathways by drought memory-related miRNAs (tae-miR160a_L+1R-1, tae-MIR167a-p5_2ss6TG19AC, tae-MIR160f-p5_2ss6TG19AC, and tae-MIR160b-p5_2ss6TG19AC) that target transcripts that significantly enriched the auxin-activated signaling pathway (GO:0009734). What is more, drought memory-related miRNA tae-MIR1432-p5_2ss6TG19AC_1 could target calcium-transporting ATPase genes. The exogenous calcium was reported to modify long-term memory and further control regeneration in plant cells [45], and the regulation of the calcium-sensing receptor in both stomatal movement and photosynthetic electron transport is crucial for water use efficiency and drought tolerance in Arabidopsis thaliana [46]. exogenous calcium was reported to modify long-term memory and further control regeneration in plant cells [45], and the regulation of the calcium-sensing receptor in both stomatal movement and photosynthetic electron transport is crucial for water use efficiency and drought tolerance in Arabidopsis thaliana [46]. Figure 5. The miRNA-gene-GO association analysis of drought memory-related miRNAs. Candidate drought memory-related miRNAs, their target transcripts, and annotated GO terms were used to investigate the relationship between them. The miRNAs are denoted in green, target transcripts are denoted in cyan, and the GO terms are red. The annotated functions of target transcripts are labeled in red.
Drought memory may also regulate plant biosynthetic processes through transcriptional memory. Drought memory-related miRNA tae-MIR1862d-p3_1ss20TC could target glutathione biosynthetic transcripts. Glutathione and its related enzymes have great significance in the response of plants to drought [47]. In addition, miR167 family members, tae-miR167a-3p_L-1_1ss23CT and other drought memory-associated miRNAs, are also involved in drought response through the CCAAT-binding factor complex, NAC transcription factors and heat shock protein 90, protein modification, electron transport, energy metabolism, amino acid metabolism, plant growth and development, and P1B-ATPase 2 for heavy metal transport.

Proposed miRNA-Dependent Regulatory Pathways That Participate in Drought Memory
Drought memory is a complex process that results from the synergistic effects of a series of regulatory pathways, including biosynthesis and metabolism of drought-resistant substances and signal transduction. This study found several pathways that may Drought memory may also regulate plant biosynthetic processes through transcriptional memory. Drought memory-related miRNA tae-MIR1862d-p3_1ss20TC could target glutathione biosynthetic transcripts. Glutathione and its related enzymes have great significance in the response of plants to drought [47]. In addition, miR167 family members, tae-miR167a-3p_L-1_1ss23CT and other drought memory-associated miRNAs, are also involved in drought response through the CCAAT-binding factor complex, NAC transcription factors and heat shock protein 90, protein modification, electron transport, energy metabolism, amino acid metabolism, plant growth and development, and P1B-ATPase 2 for heavy metal transport.

Proposed miRNA-Dependent Regulatory Pathways That Participate in Drought Memory
Drought memory is a complex process that results from the synergistic effects of a series of regulatory pathways, including biosynthesis and metabolism of drought-resistant substances and signal transduction. This study found several pathways that may be important for drought memory (Figure 6). A novel drought memory-related miRNA PC-3p-5049_3565 targeted two transcripts involved in the transition between glyceraldehyde triphosphate and D-fructose 1,6 diphosphate. Target transcripts of tae-MIR1862d-p3_1ss20TC are involved in the starch and sucrose metabolism (ko00500) pathway and are linked to carbon fixation in the photosynthetic organisms pathway via D-fructose-6-phosphate. In addition, PC-3p-325338_28 and tae-MIR1862d-p3_1ss20TC participate in the phenylalanine metabolism (ko00360) and phenylpropanoid biosynthesis (ko00940) pathways. On the one hand, phenylalanine, tyrosine, and tryptophan metabolisms can synthesize proteins, other secondary metabolites, and plant hormones to promote plant metabolism [48]. On the other hand, the down-regulate expression of genes in these pathways could increase carbohydrate metabolism by glycolysis and produce more energy in response to drought [49]. Otherwise, the target transcript of tae-MIR1862d-p3_1ss20TC is also involved in glutathione, cysteine, and methionine metabolism pathways to respond to drought memory.
be important for drought memory (Figure 6). A novel drought memory-related miRNA PC-3p-5049_3565 targeted two transcripts involved in the transition between glyceraldehyde triphosphate and D-fructose 1,6 diphosphate. Target transcripts of tae-MIR1862d-p3_1ss20TC are involved in the starch and sucrose metabolism (ko00500) pathway and are linked to carbon fixation in the photosynthetic organisms pathway via D-fructose-6-phosphate. In addition, PC-3p-325338_28 and tae-MIR1862d-p3_1ss20TC participate in the phenylalanine metabolism (ko00360) and phenylpropanoid biosynthesis (ko00940) pathways. On the one hand, phenylalanine, tyrosine, and tryptophan metabolisms can synthesize proteins, other secondary metabolites, and plant hormones to promote plant metabolism [48]. On the other hand, the down-regulate expression of genes in these pathways could increase carbohydrate metabolism by glycolysis and produce more energy in response to drought [49]. Otherwise, the target transcript of tae-MIR1862d-p3_1ss20TC is also involved in glutathione, cysteine, and methionine metabolism pathways to respond to drought memory. Figure 6. A proposed regulatory mechanism involving the differentially expressed miRNAs and their target genes in wheat drought memory. Targets of eight miRNAs were enriched to fifteen pathways that might be crucial to drought memory.
What is more, target transcripts of the tae-MIR9676-p3_1ss21GA (TraesCS1A02G246600.1) participate in two closely related pathways, including alanine, aspartate, and glutamate metabolism (ko00250) and arginine biosynthesis (ko00220). As the arginine biosynthesis pathway product, arginine is further involved in the arginine and proline metabolism pathways. The proline accumulation under osmotic stress was accompanied by increased soluble sugar concentration [50,51]. Sucrose could inhibit ABAinduced proline accumulation, and ABI4 is involved in ABA and sugar signaling [52]. Our results revealed that proline accumulation in the DM group was significantly slower than that of the DD group ( Figure 1C). Therefore, we propose a drought memory-induced drought response model in plants: drought memory-related miRNAs can improve plant drought resistance by promoting starch and sucrose catabolism, soluble sugar accumulation, and regulating proline homeostasis. Figure 6. A proposed regulatory mechanism involving the differentially expressed miRNAs and their target genes in wheat drought memory. Targets of eight miRNAs were enriched to fifteen pathways that might be crucial to drought memory.
What is more, target transcripts of the tae-MIR9676-p3_1ss21GA (TraesCS1A02G246600.1) participate in two closely related pathways, including alanine, aspartate, and glutamate metabolism (ko00250) and arginine biosynthesis (ko00220). As the arginine biosynthesis pathway product, arginine is further involved in the arginine and proline metabolism pathways. The proline accumulation under osmotic stress was accompanied by increased soluble sugar concentration [50,51]. Sucrose could inhibit ABA-induced proline accumulation, and ABI4 is involved in ABA and sugar signaling [52]. Our results revealed that proline accumulation in the DM group was significantly slower than that of the DD group ( Figure 1C). Therefore, we propose a drought memory-induced drought response model in plants: drought memory-related miRNAs can improve plant drought resistance by promoting starch and sucrose catabolism, soluble sugar accumulation, and regulating proline homeostasis.

Plant Materials and Treatments
'Chinese Spring' is a local wheat variety from China that is sensitive to drought. Seeds of the 'Chinese Spring' wheat, preserved in our laboratory, were selected as research material for this study. It was germinated on moist filter paper in Petri dishes and then transferred to a half-strength Hoagland's liquid medium [53]. Seedlings were grown in a growth chamber under controlled conditions (23 ± 1 • C, 16 h light/8 h dark cycle). Trileaf stage seedlings were divided into three groups: control group (CG), direct drought (DD), and drought memory (DM) groups. The DM group was pre-treated with 10% (M/V) PEG for 24 h in Hoagland's nutrient solution. They were then allowed to grow in Hoagland's nutrient solution for 24 h, and this treatment was repeated three times [23,54]. At the same time, the CG and DD groups were cultured in Hoagland's nutrient solution. Subsequently, all seedlings from the DD and DM groups were treated with 19.2% (M/V) PEG, whereas the CG group was grown under normal conditions. Whole seedlings of 33 samples from the three groups at drought treated 0, 1, 6, and 12 h (with three biological repetitions) were collected and immediately frozen in liquid nitrogen for the subsequent test ( Figure 1B). Furthermore, the leaves of the three groups (CG, DD, and DM) after treatment (19.2% (M/V) PEG) for 0, 24, 48, and 72 h were collected to measure proline content using a proline assay kit (Nanjing Jiancheng Bioengineering Institute, Nanjing, China) with three biological repetitions, and the photosystem II (Phi2) was measured at 0 to 72 h intervals on leaves using a Photosynq device [55].

Small RNA and Degradome Library Construction and Sequencing
Total RNA was isolated and purified using the TRIzol reagent (Invitrogen, Carlsbad, CA, USA) following the manufacturer's instructions. The RNA amount and purity of each sample were quantified using a Nano Drop ND-1000 instrument (Nano Drop, Wilmington, DE, USA). Formaldehyde denatured gel electrophoresis was used to assess RNA integrity. UMI labeling technology was then used to construct miRNA libraries to eliminate the effects of polymerase chain reaction (PCR) duplications according to the TruSeq Small RNA Sample Preparation Guide [56]. Finally, the library was sequenced on an IlluminaHiSeq 2500 device with PE150 mode at the LC-BIO (Hangzhou, China).
Every 12 samples representing the four time points (0, 1, 6, and 12 h) in the same group were pooled to construct a degradome library, and three libraries (CG, DD, and DM) were finally constructed. The degradome cDNA library was prepared following the method previously described by Axtell, with some modifications [57]. A 3 -adapter random primer was used to make the first strand of cDNA from mRNA, and size selection was performed using AMPure XP beads. Following the manufacturer's recommended protocol, the cDNA libraries were 50 base pairs (bp) single-end sequenced on an IlluminaHiSeq 2500 (LC Bio, Hangzhou, China).

Identification of Known and Potential Novel miRNAs
After the Illumina sequencing, the 3 connector sequence and low-quality data of the raw reads were first stripped, and the 30-80 bp data was retained, then the UMI sequence was extracted to remove PCR duplication and the PCR duplication ratio was calculated. The remaining sequences were subjected to an in-house program, ACGT101-miR_v4.2 (LC Sciences, Houston, TX, USA), to remove adapter dimers, junk, low complexity, common RNA families (rRNA, tRNA, snRNA, snoRNA), and repeats. After this, unique sequences of 18-25 nt were mapped to wheat precursors in miRBase 22.0 (ftp://mirbase.org/pub/ mirbase/CURRENT/ (accessed on 1 June 2021)) by BLAST search using Bowtie2 to detect known miRNAs [58]. The unmapped reads were then mapped to multiple databases, including mRNA (ftp://ftp.ensemblgenomes.org/pub/plants/release-41/fasta/triticum_ aestivum (accessed on 1 June 2021)), Repbase (http://www.girinst.org/repbase (accessed on 1 June 2021)), Rfam14 (http://rfam.janelia.org (accessed on 1 June 2021)), and genomic DNA (ftp://ftp.ensemblgenomes.org/pub/plants/release-41/fasta/triticum_aestivum (accessed on 1 June 2021)) using Bowtie2. The genome-mapped reads extended on the flanking region and the predicted secondary structure tended to form hairpins that could be considered the novel miRNAs. A modified global normalization was used to correct copy numbers among different samples, and the method corresponded to the previous study [59]. Based on bioinformatics analysis, these miRNA sequences were classified into four groups (gp1-gp4) based on the classification method [60]. gp1a are known miRNAs that have been reported in wheat; gp1b, gp2, and gp3 represent conserved miRNAs; and gp4 contains novel miRNAs.

Analysis of Differentially Expressed miRNAs
The T-test (http://en.wikipedia.org/wiki/Student\T1\textquoterights_t-test (accessed on 1 June 2021)) and ANOVA test (http://en.wikipedia.org/wiki/Oneway_analysis_of_ variance (accessed on 1 June 2021)) were chosen for pairwise-group and multi-group comparisons, respectively, to identify differential expression of miRNAs based on normalized deep-sequencing counts. The significance threshold was set at a p-value < 0.05 in each test. The expression level of the miRNAs was divided into three levels: high, medium, and low. High expression is the number of reads for the miRNAs that were higher than the average copy of the dataset. The medium expression is the number of reads for the miRNAs that were higher than ten reads and less than the average copy of the dataset, and the low expression was the number of reads reported as less than ten.

Degradome Validation and Annotation of Target Genes of miRNAs
ACGT101-DEG software (LC-bio Sciences, Houston, TX, USA) was used to inspect the targets and evaluate the correctness of the identified target, and the CleaveLand (v3.0) program was used to identify potentially cleaved targets [61]. The degradome reads were then mapped to the wheat transcriptome data. According to five categories (0-4), all targets were categorized according to the T-plot peaks as described in previous studies [62,63]. Furthermore, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) annotation were performed to predict the functions of target genes of the miRNAs according to the information collected from BioMart in Ensembl (http://plants.ensembl.org/Triticum_aestivum/Info/Index (accessed on 1 June 2021)). A hypergeometric test was then used to find the functions or pathways that were significantly enriched in the selected miRNA-mRNA pairs.

Validation of Differentially Expressed miRNAs and Their Target Genes
Reverse transcription-quantitative real-time PCR (RT-qPCR) was conducted to detect the expression level of drought memory-related miRNAs and their target genes in wheat. The sample collection, total RNA extraction, purification, quantification, and integrity detection methods were consistent with those described above. The first-strand cDNAs of miRNAs were synthesized using an miRNA 1st Strand cDNA Synthesis Kit (by stem-loop) (MR101-01, Vazyme, Nanjing, China) with the mixed reference gene reverse transcription (RT) primers and miRNA-specific stem-loop RT-primers. In addition, the first-strand cDNAs were also synthesized using Oligo dT (18T) primer and random 6 mers primer with the Evo M-MLV Mix Kit with gDNA Clean for qPCR (AG11728, Accurate Biotechnology (Hunan) Co., Ltd., Hunan, China). The RT-qPCR analysis was conducted using the QuantStudio™ 7 Flex System (Thermo Fisher Scientific, Waltham, MA, USA) with SYBR ® Green Premix Pro Taq HS qPCR Kit (Rox Plus) (AG11718, Accurate Biotechnology (Hunan) Co., Ltd., Hunan, China). The expression levels of the miRNAs and target genes were normalized using U6 (GenBank: X63066.1) and Elongation Factor 1-Alpha genes, respectively. All reactions were performed in triplicate for each sample. The primers used in this study are listed in Table S7, and expression levels were calculated according to the 2 −∆∆CT method [64].

Overexpression of tae-miR531_L-2 in Arabidopsis thaliana
The precursor DNA sequence (70 bp) of the tae-miR531_L-2 was obtained from the miRNA sequence result, the precursor sequence contained 200 bp upstream and downstream, and the stem-loop sequence was cloned from the wheat genome with the primers F: 5 -GTCAACGTCAGTACCTCATGC-3 and R: 5 -AACTGCAGCGGAGCATCAACG-3 . The cloned sequence was validated with Sanger sequencing and then inserted into the plant binary expression vector pBI121 by XbaI and KpnI restriction sites to construct an overex-pression vector. This construct was introduced into Agrobacterium tumefaciens GV3101 and then transformed into Arabidopsis thaliana (Columbia) using the floral dip method [65].
All plants were grown in a growth chamber with 16 h light/8 h dark cycles and 22 • C/19 • C day and night temperature. Soil-grown two-week-old T 3 homozygous transgenic line and wild type (WT) seedlings treated with salt (300 mM NaCl, 6 h), cold (4 • C, 6 h), heat (37 • C, 6 h), and drought (20% PEG, 0 h, 1 h, 6 h and 12 h) were collected for RT-qPCR analysis. The WT seedlings under the same treatment were used as a control to detect the relative expression level of tae-miR531_L-2 in transgenic Arabidopsis thaliana. Methods of total RNA extraction, first-strand cDNA synthesis, RT-qPCR, and relative expression level calculation were the same as those described above. For the drought resistance experiment of transgenic lines and WT, (1) seeds were surface-sterilized and germinated on half-strength Murashige and Skoog (MS) agar plates containing 0 mM and 350 mM D-mannitol for two weeks; (2) two-week-old soil-grown seedlings were air-dried and then rehydrated; (3) seven-day-old soil-grown seedlings were watered with 10% PEG solution for two weeks.

Conclusions
Repeated drought training can improve drought resistance in wheat through miRNA regulatory networks. Drought memory can regulate plant cellular signal transduction, plant biosynthetic processes, and other biological processes to cope with drought via miRNAregulated transcriptional memory. This study proposes a model of drought memoryinduced drought response in plants: drought memory-related miRNAs can improve plant drought resistance by promoting starch and sucrose catabolism and soluble sugar accumulation as well as by regulating proline homeostasis. Furthermore, overexpression of the drought memory-related miRNA tae-miR531_L-2 in plants significantly improved the drought resistance of transgenic Arabidopsis thaliana. The results presented here might provide novel insight into resolving wheat drought tolerance mechanisms and droughtresistant breeding. Institutional Review Board Statement: Not applicable.

Informed Consent Statement: Not applicable.
Data Availability Statement: The raw reads, including small RNA and degradome sequencing, in this study are publicly available at the NCBI Sequence Read Archive (SRA) under accession codes PRJNA799639 and PRJNA799641, respectively.