Genome-Wide Transcript and Small RNA Profiling Reveals Transcriptomic Responses to Heat Stress^,

This comprehensive and integrated spatiotemporal expression profile of heat-responsive mRNAs and small RNAs reveals complex and dynamic transcriptomic responses to heat stress in maize.

Abstract

Because of climate change, crops will experience increasing heat stress. However, the ways in which heat stress affects crop growth and yield at the molecular level remain poorly understood. We generated spatiotemporal mRNA and small RNA transcriptome data, spanning seven tissues at three time points, to investigate the effects of heat stress on vegetative and reproductive development in maize (Zea mays). Among the small RNAs significantly induced by heat stress was a plastid-derived 19-nucleotide small RNA, which is possibly the residual footprint of a pentatricopeptide repeat protein. This suggests that heat stress induces the turnover of certain plastid transcripts. Consistently, genes responsible for photosynthesis in chloroplasts were repressed after heat stress. Analysis also revealed that the abundance of 24-nucletide small interfering RNAs from transposable elements was conspicuously reduced by heat stress in tassels and roots; nearby genes showed a similar expression trend. Finally, specific microRNA and passenger microRNA species were identified, which in other plant species have not before been reported as responsive to heat stress. This study generated an atlas of genome-wide transcriptomic responses to heat stress, revealing several key regulators as potential targets for thermotolerance improvement in maize.

Global climate change has immense consequences for food production as extreme temperature can severely hinder and damage the growth and development of plants (Long and Ort, 2010; Kellermeier et al., 2014; Fang and Xiong, 2015). According to the global climate report from the National Oceanic and Atmospheric Administration (https://climate.nasa.gov/vital-signs/global-temperature/), the yield of staple crops, including rice (Oryza sativa), wheat (Triticum aestivum), and maize (Zea mays), decreased along with the increase in global average surface temperature, which was 0.8°C higher in 2018 than in the 20th century (Lobell et al., 2011; Bita and Gerats, 2013). Extreme temperature and heat stress are governed by the large-scale circulation of the Earth’s atmosphere and are expected to continue in the future (Horton et al., 2015). Accordingly, the improvement of high temperature tolerance in crops represents an urgent need, and whole-genome analyses of plant heat stress responses will help identify key regulators and pathways as potential targets for thermotolerance improvement.

At the cellular level, heat stress alters membrane fluidity and disrupts protein homeostasis, thereby affecting cellular processes critical for plant growth and development (Bokszczanin et al., 2013). To adapt to abnormally high temperatures, plants have developed multiple thermotolerance strategies. Heat stress responses are conserved in plants (Mittler et al., 2012) and have been shown to be induced by the following thermosensors: a calcium-permeable channel on the plasma membrane (Saidi et al., 2009), a nuclear-localized histone sensor (Kumar and Wigge, 2010), and two unfolded protein sensors localized to the endoplasmic reticulum (Deng et al., 2011) and cytosol (Sugio et al., 2009). In addition, phytochrome B functions as a thermosensor to signal through the temperature-responsive transcription factor PHYTOCHROME-INTERACTING FACTOR4 (Jung et al., 2016; Qiu et al., 2019). Signals recognized by these thermosensors ultimately activate regulatory proteins, such as heat stress transcription factors (HSFs; Kotak et al., 2007), MULTIPROTEIN BRIDGING FACTOR1C (Suzuki et al., 2008), Basic leucine zipper (Zhang et al., 2017), and WRKYGQK motif–containing proteins (WRKY; Li et al., 2010), which control the generation of heat shock proteins (HSPs) that are important components of plant heat stress responses (Kotak et al., 2007). However, the regulators important for inducing the expression of those transcription factor genes have yet to be identified, and the relationship between different heat stress response pathways remains unclear.

Small RNAs (sRNAs), a population of 20- to 24-nucleotide (nt) noncoding RNA molecules, participate in plant abiotic stress responses (Zhao et al., 2016). sRNAs are produced from the helical regions of larger RNA precursors by RNase III Dicer-like (DCL) proteins into a double-stranded duplex with 2-nt overhangs (Axtell, 2013). One strand of the duplex is loaded into an Argonaute protein and regulates the expression of its target genes through sequence complementarity with target transcripts. In plants, sRNAs are classified into two classes based on their precursors: microRNAs (miRNAs) derived from hairpin RNAs and small interfering RNAs (siRNAs) derived from double-stranded RNAs (Axtell, 2013).

miRNAs are universal regulators of gene expression in eukaryotes (Lee et al., 1993; Rogers and Chen, 2013; Ha and Kim, 2014; Iwakawa and Tomari, 2015; Tang and Chu, 2017). In plants, miRNAs impact a multitude of developmental and physiological processes (Voinnet, 2009; Sun, 2012) as well as responses to various environmental stresses (Ruiz-Ferrer and Voinnet, 2009; Shriram et al., 2016; Islam et al., 2018). The passenger strand (miRNA*) of the miRNA/miRNA* duplex is usually degraded after the formation of mature miRNA-Argonaute1 complexes (Rogers and Chen, 2013). However, studies in both animals and plants have shown that many miRNA* species can accumulate considerably, and their function may be as important as, or sometimes even more crucial than, the function of the corresponding guide miRNA (Meijer et al., 2014; Liu et al., 2017). miRNAs involved in heat stress tolerance have been identified in several plant species. Guan et al. (2013) reported that in Arabidopsis (Arabidopsis thaliana), miRNA398 is rapidly induced by heat stress and enhances plant thermotolerance by promoting the accumulation of HSF proteins through repression of three target genes involved in oxidative homeostasis: Copper/zinc Superoxide Dismutase1 (CSD1), CSD2, and a gene encoding a copper chaperone for both CSD1 and CSD2 (CCS). miR156 is an important regulator in juvenile-to-adult phase transition in plants (Chuck et al., 2007). Arabidopsis miR156 is also induced by heat stress and mediates the maintenance of acquired thermotolerance by down-regulating SQUAMOSA promoter binding protein-like genes (SPLs; Stief et al., 2014). In sunflower (Helianthus annuus), miR396 was found to regulate HaWRKY6 during early responses to high temperature, and miR396-resistant transgenic plants exhibited enhanced tolerance to heat stress (Giacomelli et al., 2012). A very recent study revealed that miR160 overexpression in Arabidopsis enhanced thermotolerance by repressing the expression of target genes AUXIN RESPONSE FACTOR10 (ARF10), ARF16, and ARF17, thereby activating a series of downstream HSP genes (Lin et al., 2018). In wheat, miR159 is down-regulated after heat stress and the miR159 overexpression rice lines were more sensitive to heat stress, indicating the negative regulation of miR159 in heat stress–related signaling pathway (Wang et al., 2012; Li et al., 2016b). While these results demonstrate that certain miRNAs play vital roles in plant heat stress tolerance, much remains unknown about the global impacts of miRNAs and other sRNAs in heat responses.

In plants, siRNAs can be further classified into phased siRNAs (phasiRNAs), heterochromatic siRNAs, and natural antisense transcript siRNAs based on their distinct biogenesis pathways and modes of action (Axtell, 2013). To date, only a few studies have described the involvement of siRNAs in heat stress responses. In Arabidopsis, the Copia-type retrotransposon ONSEN was found to be activated upon heat stress in siRNA biogenesis mutants, suggesting a potential role of siRNAs in regulating plant responses to heat stress (Ito et al., 2011). The phasiRNAs are sRNAs generated in a head-to-tail arrangement, and their production is triggered by miRNAs from phasiRNA-producing loci (PHAS loci), which include both protein-coding genes and noncoding genes. phasiRNA biogenesis requires RNA-dependent RNA polymerase 6 to synthesize double-stranded RNAs that are subsequently processed into 21- or 24-nt phasiRNAs (Fei et al., 2013). Transacting siRNAs are a special class of phasiRNAs generated from transacting siRNA (TAS) loci (Vazquez et al., 2004). HEAT-INDUCED TAS1 TARGET1 (HTT1) and HTT2 are highly induced by heat stress and are targeted by TAS1-derived siRNAs in Arabidopsis (Li et al., 2014). Overexpression of TAS1a, which enhanced TAS1-siRNA accumulation, caused increased sensitivity of plants to heat stress (Li et al., 2014). In dicots, PHAS loci mainly comprise disease resistance genes (Zhai et al., 2011), while in monocots, phasiRNAs largely originate from long noncoding RNAs and are abundant in inflorescence tissues (Johnson et al., 2009; Zhai et al., 2015). In maize, miR2118 triggers 21-nt phasiRNA accumulation preferentially during premeiosis, and miR2275 triggers 24-nt phasiRNA accumulation mainly during meiosis in anther cells (Zhai et al., 2015). In rice, the long noncoding RNA Photoperiod-sensitive Male Sterility 1 Transcript, which is highly expressed in young panicles and gives rise to 21-nt phasiRNAs triggered by miR2118, regulates photoperiod-sensitive male sterility under long-day conditions (Fan et al., 2016). Taken together, these studies link phasiRNAs with reproductive growth in monocots, but the roles of these sRNAs in heat stress tolerance in reproductive tissues remain unclear.

Maize (Z. mays ssp. mays) is the second most abundant crop worldwide and is expected to be the top cereal in the future due to its high yield potential as a C4 plant (Jones, 2009; Ort and Long, 2014). Most maize genotypes have been adapted to warm temperatures (Hufford et al., 2012). Nevertheless, heat stress can cause thin leaf morphology and lead to a reduction in the CO₂ assimilation rate in maize seedlings (Karim et al., 2000), which contributes to pollination failure and kernel abortion (Gong et al., 2015). It is crucial to uncover the molecular mechanism of maize thermotolerance, especially of the vital regulatory factors, to pave the way for genetic breeding of heat-resistant maize.

In this study, maize plants were subjected to 2-h (short-term) or 48-h (long-term) heat stress, and different vegetative- and reproductive-stage tissues were subsequently collected for mRNA sequencing (mRNA-seq) and sRNA-seq. By comparing the mRNA transcriptome profiles, we identified differentially expressed genes (DEGs) in specific tissues. In particular, genes responsible for photosynthesis in chloroplasts were repressed after short-term heat stress. From the sRNA transcriptome analysis, we discovered that a plastid-derived 19-nt sRNA, presumably the footprint of a pentatricopeptide repeat (PPR) protein, was significantly induced after heat stress. In contrast, we found that the abundance of a population of 24-nt siRNAs from transposable elements (TEs) in roots and tassels was conspicuously reduced after heat stress. Moreover, nearby genes also showed a similar decreased trend, indicating that the expression of siRNAs and their nearby genes may be coregulated. Finally, we identified specific miRNAs, as well as miRNA* species, that have not been reported in other plant species to be responsive to heat stress. This study provides a comprehensive expression profiling of heat-responsive mRNAs and sRNAs and elicidates the molecular mechanisms that regulate heat stress in maize at transcriptional and posttranscriptional levels.

RESULTS

Transcriptomic Responses to Heat Stress Differ between Vegetative and Reproductive Tissues

Maize inbred line B73 plants grown at normal temperature were exposed to heat stress (38°C) for a short (2-h) or long (48-h) period at the V3 stage (collar of third leaf visible at the vegetative stage) and at the R1 stage (silk just visible outside the husks at the reproductive stage; Abendroth et al., 2011). After heat stress, the newly emerged leaves of maize at the V3 stage became curlier than the control leaves (Supplemental Fig. S1A), and the leaf chlorophyll content reduced slightly as well (Supplemental Fig. S1, B and C). At the R1 stage, sticky and sweet drops on stalks of maize were observed after a long period of heat stress (Supplemental Fig. S2A), which might be osmolytes secreted to regulate osmotic pressure in response to heat stress (Kaplan et al., 2004). In addition, the vigor of pollen stained by I₂-KI solution decreased from 24 h after heat stress (Supplemental Fig. S2, B and C). These results indicated that heat stress affected maize growth and development.

At 0, 2, and 48 h after heat stress (HAH), leaves, stalks, and roots were collected from V3-stage plants, and tassels, ears, silks, and leaves were collected from R1-stage plants. RNA was isolated from the samples and processed for mRNA-seq, which yielded ∼20 million mapped reads per sample and ∼15 million uniquely mapped reads per sample (Supplemental Data Set 1). At least two biological replicates were conducted for each sample, and the correlation coefficients between biological replicates were high (r > 0.94; Fig. 1A; Supplemental Fig. S3A; Supplemental Data Set 2). For each analyzed tissue, genes were randomly selected for reverse transcription quantitative PCR (RT-qPCR) verification, and the results were in agreement with those from mRNA-seq (Supplemental Fig. S4). Heat stress treatment did not affect the global patterns of gene expression in the analyzed tissues except for R1-tassels, in which the average gene expression level decreased at 48 HAH (Supplemental Fig. S3B). This change was probably due to the high sensitivity of pollen to heat stress (Müller and Rieu, 2016). Of the 43,943 annotated maize genes (Jiao et al., 2017), more DEGs were identified in tissues at the R1 stage (2,901/3,245 up-/down-regulated genes at 2 HAH; 6,201/6,923 up-/down-regulated genes at 48 HAH) than at the V3 stage (1,701/1,096 up-/down-regulated genes at 2 HAH; 2,149/2,301 up-/down-regulated genes at 48 HAH) compared to the untreated control (Supplemental Fig. S3C; Supplemental Data Set 3). The only common DEGs between the V3 stage and the R1 stage were 40 up-regulated genes at 2 HAH (Fig. 1B). Based on gene ontology (GO) enrichment analysis, these common DEGs were mainly associated with heat acclimation, response to heat, and protein folding (Fig. 1C). For tissue-specific DEGs (Fig. 1B), GO analysis showed that photosynthesis-associated genes were highly up-regulated in R1-tassels at 2 HAH but down-regulated at 48 HAH (Supplemental Fig. S5). Genes with the GO terms reproductive process, noncoding RNA process, and noncoding RNA metabolic process were down-regulated at 48 HAH in R1-tassels, suggesting that heat stress for 48 h may have affected pollen production or vigor. Among the down-regulated genes enriched in reproductive process, five genes (Zm00001d034701, Zm00001d007388, Zm00001d053886, Zm00001d053286, and Zm00001d053895) were homogenous to male-sterility genes reported in other plant species (Supplemental Data Set 3; Wan et al., 2019). In contrast to the pattern in R1-tassels, genes with the GO terms noncoding RNA metabolic process and RNA process were up-regulated at 48 HAH in V3-leaves, including some sRNA biosynthesis genes (Supplemental Fig. S5).

Figure 1.

RNA-seq analysis of transcriptomic responses to heat stress in different maize tissues at the vegetative and reproductive stages. A, Hierarchical clustering dendrogram using Pearson correlation coefficients for seven tissues at the vegetative and reproductive stages. B, Modified Venn diagrams showing the common and specific heat-responsive genes across different maize tissues. C, GO enrichment analysis of genes up-regulated by heat in all maize tissues at 2 HAH.

In general, heat stress could have larger effects on the reproductive stage than the vegetative stage. We compared the molecular expression changes of leaves at both V3 and R1 stages in response to heat stress. As expected, we observed more DEGs in R1-leaf at the reproductive stage, suggesting that the stress at the reproductive stage has larger effects than the vegetative stage (Supplemental Fig. S6). In addition, the DEGs were quite different in leaves at vegetative and reproductive stages, which provides insight into the different regulatory mechanisms induced by heat stress between vegetative and reproductive tissues (Supplemental Fig. S6). Specifically, when exposed to heat stress for a short period (2 h), in leaves at the vegetative stage (V3-leaf), transferase- and terpene synthase–related genes were up-regulated, while phenylpropanoid metabolic process copper ion binding–related genes were down-regulated. In leaves at the reproductive stage (R1-leaf), RNA processing and nuclease-related genes were up-regulated, while membrane- or transporter-related genes were down-regulated. When exposed to heat stress for a long period (48 h), in V3-leaf, nuclear import and nucleoside binding-related genes were up-regulated, while photosynthesis-related genes were down-regulated. In R1-leaf, chloroplast- or plastid-related genes were up-regulated, while transmembrane transporter–related genes were down-regulated (Supplemental Fig. S6).

Photosynthesis-Related Genes Are Repressed in Maize Leaves under Heat Stress

It has been reported that net photosynthesis in maize leaves is inhibited at temperatures above 38°C (Crafts-Brandner and Salvucci, 2002). Genes in chloroplasts are transcribed by two types of RNA polymerases: nuclear-encoded RNA polymerase (NEP) and plastid-encoded RNA polymerase (PEP; Kremnev and Strand, 2014; Yu et al., 2014). NEP is a T3-T7 bacteriophage-type enzyme with a single subunit that predominantly mediates the transcription of housekeeping genes including PEP. PEP is a bacterial-type multisubunit enzyme, whose four core subunits are encoded by the genes RNA polymerase A (rpoA) , rpoB, rpoC1, and rpoC2, and it dominates the transcription of photosynthesis genes. There is a single copy of the NEP gene RNA polymerase T phage-like1 in the maize genome (Schnable and Freeling, 2011), and its expression levels in the leaves of both vegetative-stage and reproductive-stage plants were induced by heat stress (Supplemental Fig. S7A; Supplemental Data Set 4). The expression levels of rpoA, rpoB, rpoC1, and rpoC2 were all dramatically reduced at 2 HAH in R1-leaves but markedly increased at 24 and 48 HAH according to the RT-qPCR results (Fig. 2A), exhibiting rapid adaptation to heat stress. Because the chloroplast PEP core enzyme requires sigma-like transcription factors encoded by the nuclear genome for promoter recognition and gene transcription initiation (Chi et al., 2015), the expression of sigma-like transcription factors was also examined. Among the seven sigma factor genes (Sigs) in the maize genome, only two exhibited significant changes in expression after heat stress. ZmSig3, the ortholog of AtSig3, which specifically initiates the transcription of subunit N of PSII (psbN) in PSII (Zghidi et al., 2007), had increased transcript levels at 24 HAH in R1-leaves (Supplemental Fig. S7B; Supplemental Data Set 4), while ZmSig8, whose Arabidopsis homolog is induced by low temperatures (Nagashima et al., 2004), showed a sharp reduction at 2 HAH in R1-leaves (Supplemental Fig. S7B; Supplemental Data Set 4). Next, we examined the expression of genes involved in photosynthesis in the chloroplast. The PSII psbB operon (psbB-psbT-psbH-photosynthetic electron transport B [petB]-petD) has a promoter recognized by PEP and is highly conserved in vascular plants. psbN is another gene on the opposite strand of the psbB gene cluster between psbT and psbH (Stoppel and Meurer, 2013). According to RT-qPCR, the expression levels of psbN, psbC, and all genes in the psbB operon except for psbT were significantly reduced (Fig. 2, B and C). Moreover, the expression levels of major genes in the photosynthesis light reaction, such as gene encoding subunit C in PSI genes (psaC), genes encoding the ATP synthase subunits a and b (atpa and atpb), and the cytochrome b₆-f complex subunit gene cybt6, were all coordinately down-regulated at specific time points after exposure to heat stress (Fig. 2, D and E). The large-chain gene of Rubisco (rbcl) gene encodes a subunit of Rubisco, the first key enzyme in carbon fixation reactions, and its expression levels were also significantly repressed after heat stress (Fig. 2E). In summary, the 2-h heat stress treatment dramatically inhibited photosynthesis-related genes encoded by the maize chloroplast genome in leaves at both the vegetative and reproductive stages, which may result in reduction of chlorophyll content in vegetative leaves after heat stress (Supplemental Fig. S1). Photosynthesis-related genes encoded by the nuclear genome showed only a slight reduction in expression at 48 HAH in leaves, except for the significant changes observed for NEP and the two sigma factors ZmSig3 and ZmSig8 (Supplemental Fig. S7C; Supplemental Data Set 4).

Figure 2.

Most photosynthesis-related genes in chloroplasts were dramatically repressed by heat stress in maize leaves. A, Expression patterns of four PEP core subunit genes in chloroplasts. B, Expression of two PSII genes were repressed after heat stress. C, Expression of genes in the psbB operon were repressed after heat stress. D, Expression of the iron-sulfur center protein gene psaC in PSI was decreased after heat stress. E, ATP synthase subunit a and b genes (atpa and atpb), the cytochrome b₆-f complex subunit gene cybt6 and the Rubisco large subunit gene rbcl were all down-regulated after heat stress. All genes in the chloroplast were examined by qPCR after RT using random primers in R1-leaves after heat stress. Asterisks indicate significant differences compared to 0 HAH (Student’s t test: *P < 0.05, **P < 0.01, ***P < 0.001). Each analysis includes three independent replicates. Error bars indicate sd. Sample size n ≥ 3.

Genes Encoding PPR Proteins Are Induced by Heat Stress in Maize Leaves

To more fully understand the maize transcriptomic responses to heat stress, sRNA libraries were constructed and sequenced, and more than 5 million mapped reads were obtained for each sample (Supplemental Data Set 5). The sRNA populations in the V3-stalk, V3-root, R1-tassel, R1-ear, and R1-silk tissues were characterized by two prominent size classes, 22 and 24 nt (Supplemental Fig. S8), as previously observed in maize plants (Nobuta et al., 2008; Zhai et al., 2013; Lunardon et al., 2016). Strikingly, there was a distinct enrichment of 19-nt sRNAs in leaves at both the V3 and R1 stages (Fig. 3A; Supplemental Fig. S8). In particular, in R1-leaves at 48 HAH, the proportion of 19-nt sRNAs increased dramatically compared to that of the untreated control (Fig. 3A). We searched the corresponding reads and found that the 19-nt peak was due to a single sRNA generated from the 3′ untranslated region of the psbH gene (Fig. 3B), whose product is a PSII subunit in chloroplasts. The psbH gene is located within the pentacistronic psbB operon along with psbB, psbT, petB (cytochrome b₆), and petD (subunit IV of the cytochrome b₆f complex; (Stoppel and Meurer, 2013). The significant increase in the abundance of the 19-nt read (GGT​AGT​TCG​ACC​GTG​GAT​T) from psbH at 48 HAH detected in the sRNA-seq analysis (Fig. 3C) was further verified by northern blot (Fig. 3D). In fact, the 19-nt sequence corresponded precisely to the binding site of high-chlorophyll-fluorescence 152 (HCF152) protein, a PPR protein that protects psbH mRNA from degradation in the chloroplast. Thus, the accumulation of this sRNA fragment represented a heat-induced footprint of HCF152 (Zhelyazkova et al., 2012). The mRNA-seq data showed that the expression of the HCF152 gene was induced at 2 HAH (Fig. 3E), and the change was verified by RT-qPCR (Fig. 3F). Meanwhile, the expression of psbH was reduced at 2 HAH (Fig. 2C). These data suggest that heat stress caused the degradation of psbH RNA, and the induction of HCF152 partially counteracted this effect.

Figure 3.

PPR proteins in response to heat stress in maize. A, Size distribution (in nucleotides) of sRNAs from maize leaves after heat stress at the R1 stage. R1, silks just visible outside the husks at the reproductive stage. B, The HCF152 protein affects the accumulation of specific psbB operon transcripts encoded by the chloroplast genome in maize R1-leaves in response to heat stress. Chr. Pt, chromosome of chloroplast. C and D, sRNA-seq data (C) and northern blotting (D) both indicated a dramatic induction of the 19-nt sRNA shown in (B) after heat stress in maize R1-leaves. The probe labeled by biotin at both the 3′- and 5′-terminal ends was reverse complementary to the 19-nt sRNA. Northern blotting for miR827 was used to mark the 21-nt sRNA position. E and F, HCF152 expression was induced by heat stress in maize R1-leaves according to RNA-seq (E) and RT-qPCR (F). Error bars indicate sd. Sample size n ≥ 3. G, Heatmap showing the expression patterns of PPR genes differentially expressed in R1-leaves at 48 HAH. The cutoff for DEGs was as follows: fold-change > 2 and reads per kilobase per million mapped reads (RPKM) > 10 for at least one time point. Asterisks indicate significant difference compared to 0 HAH (Student’s t test: *P < 0.05, **P < 0.01, ***P < 0.001).

The maize B73 genome encodes at least 495 PPR-related proteins (Supplemental Data Set 6; Cheng et al., 2016), which are characterized by several canonical repeats of a 35-amino acid motif and are involved in RNA binding in chloroplasts and mitochondria (Barkan and Small, 2014). The PPR proteins in plants consist of two major subfamilies: the P-class proteins, which contain only tandem arrays of P motifs and are implicated in RNA stabilization, processing, splicing, and translation, and the PLS-class proteins, which possess characteristic triplets of P, L, and S motifs and function mainly in RNA editing (Fujii and Small, 2011; Barkan and Small, 2014). Among the 495 PPR-related genes in maize, 58 and 4 PPR genes were up-regulated and down-regulated, respectively, in R1-leaves at 48 HAH (Fig. 3G; Supplemental Data Set 6). Thus, heat stress may influence the regulation of PPR proteins in chloroplasts and mitochondria. According to the classification in Cheng et al. (2016), more than half of the 58 up-regulated PPR genes (32) belong to the P subfamily, and the others belong to five subgroups of the PLS subfamily (PLS, E1, E2, E+, and DYW; Fig. 3G).

TE-Derived 24-nt siRNAs Are Dramatically Decreased in Response to Heat Stress in Maize Tassels and Roots

As shown by the sRNA length distribution, the abundance of 24-nt sRNAs was reduced dramatically by more than half at 48 HAH compared to 0 HAH in maize V3-roots and R1-tassels (Fig. 4A), but did not change significantly in other five tissues (Fig. 3A; Supplemental Fig. S8). By mapping the sRNAs to different genomic components, statistical analysis showed that 24-nt TE-derived sRNAs decreased dramatically at 48 HAH in V3-roots and R1-tassels (Fig. 4C), and the reduced 24-nt sRNAs were mainly derived from TEs (Fig. 4D). Moreover, sRNAs derived from the promoter and intergenic region were also found to be statistically reduced after heat stress (Fig. 4E). In plants, most 24-nt sRNAs are siRNAs generated by plant-specific RNA polymerase IV and correlate with RNA-dependent DNA methylation and transcriptional gene silencing (Zhang et al., 2007; Mosher et al., 2008; Erdmann and Barciszewski, 2011). To identify sRNAs differentially expressed in response to heat stress, the maize genome was divided into consecutive nonoverlapping 500-bp windows, and the normalized sRNA read counts in each window were compared between the heat-treated and control samples. Specifically, we compared the abundance of sRNAs at 48 and 0 HAH in each 500-bp window for each sRNA size class (21, 22, 23, and 24 nt) for V3-roots and R1-tassels. In both tissues, the analysis revealed more hypo- (48 HAH < 0 HAH) differential sRNA regions (DSRs) than hyper- (48 HAH > 0 HAH) DSRs (Fig. 4F; Supplemental Data Set 7). The maize genome possesses the highest level of TEs compared to any other plants (Baucom et al., 2009; Schnable et al., 2009). In maize, ∼90% of TEs are located in pericentromeric regions and heterochromatic maize knobs (Baucom et al., 2009), and the majority of genes (85.5%) have TEs within 1 kb (Li et al., 2015). Interestingly, the distribution of 24-nt hypo-DSRs was consistent with the gene distribution pattern on maize chromosomes (Fig. 4F), indicating the possible functions of gene-proximal TEs in gene expression regulation.

Figure 4.

TE-derived 24-nt sRNAs were dramatically decreased after heat stress in maize tassels and roots. A, Size distribution (in nucleotides) of sRNAs from V3-roots and R1-tassel after heat stress. B, Composition of sRNAs in roots and tassels according to their genomic location. C, Abundance of 24-nt sRNAs generated from TEs were significantly reduced at 48 HAH compared to 0 HAH in both V3-root and R1-tassel. Error bars indicate sd. Sample size n ≥ 3. D, Composition of reduced 24-nt sRNAs in V3-root and R1-tassel at 48 HAH according to their genomic location. E, Proportion of sRNAs from different genomic components. Numbers above the columns are P-values for χ² test. F, Distribution of differentially expressed sRNAs along the maize genome after heat stress in roots and tassels, respectively. The genome was divided into 500-bp windows for analysis. The arrowheads show the location of centromeres. Chr, Chromosome; hyper, up-regulated; hypo, down-regulated. Asterisks indicate significant difference (Student’s t test: *P < 0.05, **P < 0.005).

To analyze the possible relationship between changes in TE-derived 24-nt sRNAs and changes in the expression of their nearby genes, we analyzed the maize genome to identify TEs within 1 kb of nearby genes (Supplemental Data Set 8). Next, we identified TEs that generated 24-nt sRNAs within 1 kb of genes in R1-tassels and V3-roots at 0, 2, and 48 HAH. The analysis indicated that ∼20%–30% of TEs that generated 24-nt sRNAs had nearby genes within 1 kb, and this proportion was not influenced by heat stress in either R1-tassels or V3-roots (Fig. 5A). Interestingly, we found that TEs producing 24-nt hypo-DSRs under heat stress were significantly enriched in the vicinity of genes within 1 kb (Fig. 5A). Generally, TEs are classified into two classes based on their transposition mechanism. Class I TEs are also known as retrotransposons, and their transposition strategy involves a copy-and-paste mechanism via RNA intermediates; class II TEs are known as DNA transposons and transpose using a cut-and-paste mechanism via DNA (Wicker et al., 2007). The two types of TEs can be further classified into superfamilies defined by their sequence conservation and structural relationships (Fig. 5B; Wicker et al., 2007). Retrotransposons from the long terminal repeat (LTR)/Copia and LTR/Gypsy superfamilies and DNA transposons from the Helitron superfamily are the most abundant transposons throughout the maize genome. The genomic analysis showed that all DNA transposon families and three retrotransposon families exhibited preference to locate nearby genes within 1 kb (Fig. 5B). But among the TEs producing down-regulated 24-nt sRNA in R1-tassel at 48 HAH, only the Helitron family, which belongs to DNA transposon subclass 2, showed enrichment nearby genes within 1 kb (Fig. 5B). These results indicated that maize TEs that responded to heat stress were enriched in the vicinity of genes and most of them belonged to Helitron family of DNA transposons.

Figure 5.

Genes near TEs producing 24-nt hypo-DSRs tended to be down-regulated under heat stress. A, TEs producing down-regulated 24-nt sRNAs are enriched near genes. B, Classification of TEs generating down-regulated 24-nt sRNAs in R1-tassels at 48 HAH. DR, Down-regulated. Numbers above the columns are P-value for χ² test. Numbers in blue indicated that the TE family was enriched nearby genes, while no numbers or numbers in black were not enriched. hAT, the name derives from three well-described TE families: hobo from Drosophila, Ac-Ds from maize, and Tam3 from snapdragons; PIF,P instability factor; RTE, retrotransposable element; SINE, short interspersed nuclear element; TIR, terminal inverted repeat. C, Hyper-geometric tests showing that the reduced accumulation of 24-nt sRNAs at TEs was correlated with the down-regulation of nearby genes within 1 kb under heat stress at 48 HAH. P-values were calculated according to the hyper-geometric test. D, Graphs showing the location of TEs relative to nearby genes within 1 kb.

To further analyze the relationship between TEs generating 24-nt hypo-DSRs and the expression of their proximal genes, we examined the mRNA-seq data and found that in R1-tassels, 169 genes that were down-regulated were within 1 kb of TEs producing 24-nt hypo-DSRs at 48 HAH compared to 0 HAH (Supplemental Data Set 9). Moreover, according to hyper-geometric test, these genes were significantly enriched among the total down-regulated genes in response to heat stress (Fig. 5C). Among these 169 down-regulated genes, ∼60% have TEs inserted in the promoter region and 31% have TEs inserted in the gene body, which is higher than the normal distribution from the whole genome (Fig. 5D). These results suggested that under heat stress, the reduced accumulation of 24-nt sRNAs at TEs was correlated with the down-regulation of nearby genes.

miRNA Expression Profiles in Different Tissues in Response to Heat Stress

Using data from miRbase (http://www.mirbase.org/) and miRNEST (http://rhesus.amu.edu.pl/mirnest/copy/), we identified 73 miRNA families in maize, and 52 of these families had reads per million mapped reads (RPM) values > 0 in at least one analyzed tissue (Supplemental Data Set 10). Comparing the expression levels between heat-stressed and control samples, we identified differentially expressed miRNAs at 2 and 48 HAH in different maize tissues at the vegetative and reproductive stages, respectively (Fig. 6, A and B; Supplemental Fig. S9). Many more miRNAs responded to heat stress in V3 tissues than those in R1 tissues (Supplemental Fig. S10). In V3-leaves, almost all differentially expressed miRNAs at both 2 and 48 HAH were up-regulated, while the differentially expressed miRNAs in V3-roots and R1-tassels were mostly down-regulated at 48 HAH (Fig. 6, A and B; Supplemental Fig. S9). Moreover, the total abundance of 21-nt miRNAs in these analyzed tissues exhibited the same trends (Supplemental Fig. S11). Among the 33 differentially expressed miRNAs in the analyzed tissues, 18 were mature miRNA species and the other 15 were miRNA* species (Supplemental Fig. S10). miRNAs were randomly selected for validation by northern blot, and the analysis showed expression patterns similar to those from the sRNA-seq analysis (Fig. 6, C and D).

Figure 6.

miRNAs in response to heat stress in different maize tissues at 48 HAH. A and B, Scatter plots comparing the abundance of miRNAs between 0 and 48 HAH in different maize tissues at the vegetative stage (A) and reproductive stage (B). sRNA-seq was performed using at least two biological replicates for each maize tissue under heat stress. Purple and blue dots indicate miRNAs up-regulated and down-regulated by heat stress, respectively. miRNAs mentioned in the text were highlight in dark purple (up-regulated) or blue (down-regulated). C and D, Northern blotting for sRNA-seq data verification at the vegetative stage (C) and reproductive stage (D). At least two biological replicates were conducted for each blot and one is shown. U6 served as a loading control. Numbers below the blots indicate the relative abundance of the sRNAs to the 0 HAH control samples.

The four most abundant miRNAs, miR159, miR166, miR396, and miR408, were differentially expressed in response to heat stress (Supplemental Fig. S12). miR159 targets Repeat 2 and Repeat 3 MYB domain transcription factor genes and regulates the timing of vegetative development in Arabidopsis (Guo et al., 2017). A number of studies have also reported changes in miR159 abundance in response to various abiotic and biotic stresses (Liu et al., 2012; Wang et al., 2012; Du et al., 2014; Yang et al., 2014), and silencing miR159 caused curly leaves in Arabidopsis (Allen et al., 2007). miR159 was slightly down-regulated in V3-leaf after heat stress (Supplemental Fig. S12), which might account for the aforementioned curly leaf phenotype in the newly emerged leaf (Supplemental Fig. S1A). In addition, miR159 was up-regulated at 48 HAH in V3-stalks but down-regulated in V3-roots and R1-leaves from 2 HAH (Supplemental Fig. S12). miR166 is highly conserved in plants and involved in various developmental processes, such as root development, vascular patterning of the shoot, and nutrient ion uptake (Kim et al., 2005; Boualem et al., 2008; Iwamoto and Tagiri, 2016). Heat stress induced different expression pattern changes in different miR166 family members (Supplemental Fig. S12). Mature miR166c/d/g/h/i and miR166l/m were up-regulated in V3-leaves at 48 HAH, while miR166a/n/q/r, miR166k/j/o, and miR166p were up-regulated in V3-stalks. At the R1 stage, most of the miR166 members were down-regulated in the analyzed maize tissues. miR396 family members, which play important roles in grain size determination (Gao et al., 2015; Li et al., 2016a), were dramatically induced at 2 HAH in V3-leaves and at 48 HAH in R1-silk tissue but repressed in V3-stalks and V3-roots (Supplemental Fig. S12). miR408, which accumulates to high levels in senescing leaves of Arabidopsis (Thatcher et al., 2015) and is responsive to various abiotic stresses (Abdel-Ghany and Pilon, 2008; Zhang et al., 2014; Ma et al., 2015), was dramatically reduced after long-term heat stress in V3-roots and R1-tassels (Supplemental Fig. S12).

Several monocot-specific miRNAs such as miR1432, miR444, and miR528 were also differentially expressed upon heat stress. miR1432 has been implicated in drought stress (Ferdous et al., 2017) and grain yield determination (Zhao et al., 2018). In our study, miR1432 was highly expressed in R1-leaves and was dramatically induced at 48 HAH, while its expression levels were low in other maize tissues (Fig. 6D; Supplemental Fig. S10). miR528 is involved in repressing maize lignin biosynthesis and lodging resistance under nitrogen-luxury conditions, and it also plays a role in antiviral defense by regulating reactive oxygen species accumulation in rice (Wu et al., 2017; Sun et al., 2018). In our data, miR528 was markedly reduced in V3-roots after heat stress (Fig. 6C), while the passenger strand miR528* was dramatically induced by heat stress in V3-leaves and V3-stalks (Fig. 7A; Supplemental Fig. S10). miR444, which functions in the rice nitrate signaling pathway (Yan et al., 2014) and antiviral pathway (Wang et al., 2016), was induced in V3-leaves at 48 HAH but dramatically repressed in R1-tassels (Fig. 6; Supplemental Fig. S10).

Figure 7.

miRNA* species in response to heat stress in different maize tissues. A and B, Northern blot for miR528* (A) and miR168* (B) in maize tissues at the vegetative stage (V3) and reproductive stage (R1). At least two biological replicates were conducted for each blot and one is shown. U6 served as a loading control. Numbers below the blots indicate the relative abundance of the sRNAs to the 0 HAH control samples. C and D, Validated targets of miR528* (C) and miR168* (D) using degradome-seq. The target plots (T-plots) show the abundances for degradome tags along the full-length of the target mRNA sequence. The frequencies of degradome tags with 5′ ends at the indicated positions are shown in black, with the frequency at the miRNA cleavage site highlighted with the red dot. The alignments show the miRNA with a portion of its target sequence (bottom). Red arrow indicates signatures consistent with miRNA-directed cleavage. The underlined nucleotide in the target transcript indicates the cleavage site detected in the degradome analysis. NAC, NAM, ATAF1/2 and CUC2 domain. E and F, Expression patterns of miR528* (E), miR168* (F), and their target genes. The y axis indicates the log₂ transformation of RPM for miRNAs and reads per kilobase per million mapped reads (RPKM) for genes. R represents the negative correlation coefficient between the expression patterns of miRNAs and their predicted target genes.

Three miRNAs previously reported to be associated with abiotic stress responses, miR398, miR399, and miR827, had varied expression patterns in response to heat stress in the different maize tissues. miR398 participates in many abiotic and biotic stress responses (Zhu et al., 2011). The expression of miR398 was much higher in leaves at both the V3 and R1 stages and remarkably low in V3-roots and R1-tassels, and at 48 HAH, its levels increased in R1-silks (Fig. 6; Supplemental Fig. S10). miR399 and miR827 were shown to be dramatically induced upon phosphate starvation (Liu et al., 2014). In our sRNA-seq data, miR827 was preferentially expressed in R1-stage tissues, and miR399e/i/j were highly expressed in R1-leaves at 0 HAH (Supplemental Fig. S10). miR827 was significantly induced in V3-leaves and R1-ear, while miR399e/i/j were slightly repressed in R1-silks from 2 HAH (Fig. 6; Supplemental Fig. S10).

Another notable miRNA in terms of heat stress response was miR162, whose target gene is DCL1 (Xie et al., 2003). miR162 expression was repressed by heat stress in V3-roots and R1-tassels at 48 HAH (Fig. 6), and the mRNA-seq data indicated that DCL1 expression increased slightly after heat stress in V3-roots (Supplemental Data Set 11).

In summary, overall miRNAs tend to be more actively responsive to heat stress at vegetative tissues than reproductive tissues, and long-term heat stress has greater impacts on the expression of miRNAs than that of short-term heat stress, indicating the complex and dynamic miRNA responses to heat stress in maize.

Heat Stress Alters the Expression of Specific miRNA* Species in Various Tissues

In heat-stressed maize leaves at the vegetative stage (V3), several miRNA* species accumulated to high levels, and in most cases, the corresponding mature miRNAs did not exhibit any obvious changes (Supplemental Fig. S10). These included the miRNA* strands for miR164c/h family members, miR167g, miR168a/b family members, miR396a/b/i family members, miR398c, miR528a/b family members, and miR529 (Supplemental Fig. S13A). Only two miRNAs, miR166p and miR827a, had increased expression trends for both the mature and passenger strands in heat-stressed maize leaves (Supplemental Fig. S13B). The most striking miRNA* response to heat stress was that of miR168* (miR168a/b-3p; Fig. 7B). Its expression was high at the R1 stage, although the mature miR168 was only minimally expressed in all analyzed maize tissues (as visualized by Integrative Genomics Viewer; Supplemental Fig. S14; Supplemental Data Set 10), which is notably different from its abundance in Arabidopsis (Li et al., 2012; Iki et al., 2018) and rice (Du et al., 2011). The expression of miR168* was largely induced by heat stress as early as 2 HAH in V3-leaves and V3-roots but significantly down-regulated in R1-stage tissues (Supplemental Fig. S10). These expression pattern changes were validated by northern blot (Fig. 7B), indicating an important role of miR168* in the vegetative-to-reproductive stage transition during maize development as well as different responses to heat stress during the two developmental stages. Another miRNA* whose expression changed significantly under heat stress was miR528*. In V3-leaves, miR528* was highly induced by heat stress at 48 HAH, whereas in V3-stalks, miR528* was first down-regulated at 2 HAH and then up-regulated at 48 HAH (Fig. 7A), indicating the tissue-specific and dynamic responses of miR528* to facilitate plant adaptation to heat stress. To determine whether the passenger strand of miRNAs have targets, we performed degradome-seq for R1-silk under normal condition. This analysis revealed Zm00001d011589 and Zm00001d021605 as potential target genes for miR168* and miR528*, respectively (Fig. 7, C and D). Zm0001d011589 is annotated to encode a NAC domain-containing protein, and Zm00001d021605 is an unknown function gene. As expected, these two target genes showed opposite expression patterns compared to their corresponding miRNA*s (Fig. 7, E and F; Supplemental Data Set 12), indicating that miR168* and miR528* are functional and may play roles in maize heat stress responses through regulating their targets.

To further investigate the potential roles of the differentially expressed miRNA and miRNA* species in regulating heat tolerance in maize, we predicted their target genes using the psRNATarget program (Dai et al., 2018) and analyzed the expression correlation between miRNAs and their target genes. Based on the scores and the negative correlation coefficients of the expression between miRNAs and their target genes (Fig. 8; Supplemental Data Set 13), 43 predicted candidate target genes were selected for further GO enrichment analysis (gene IDs in red in Supplemental Data Set 13). Interestingly, 5 of the 43 genes had terms associated with thylakoid membrane organization (Supplemental Fig. S15), which is related to functions in chloroplasts. All five genes had down-regulated expression levels in leaves at the V3 and R1 stages (Supplemental Data Set 13), consistent with the our results indicating that photosynthesis was repressed by heat stress in maize leaves (Fig. 2). We also validated the target genes for some of the heat stress–responsive miRNAs and miRNA*s by degradome-seq (Supplemental Data Set 14).

Figure 8.

Predicted target genes showing opposite expression patterns to those of miRNAs. The y axis indicates the log₂ transformation of RPM for miRNAs and reads per kilobase per million mapped reads (RPKM) for genes. R represents the negative correlation coefficient between the expression patterns of miRNAs and their predicted target genes.

DISCUSSION AND CONCLUSION

In addition to cereals such as maize, numerous crop species are greatly affected by heat stress at both the vegetative and reproductive stages, ultimately resulting in extensive yield loss (Barnabás et al., 2008; Lesk et al., 2016). Studies on the physiological effects of heat stress have mainly emphasized increased membrane fluidity, reactive oxygen species accumulation (Sangwan et al., 2002; Atkin and Tjoelker, 2003), reductions in shoot biomass, leaf extension rate and the CO₂ assimilation rate at young stages (Karim et al., 2000; Kumar et al., 2012; Mathur et al., 2014) as well as impaired pollen formation and fertilization at the reproductive stage (Müller and Rieu, 2016). Research over the past several decades has elucidated molecular mechanisms of the plant heat stress response (Liu et al., 2015; Ohama et al., 2017), with an emphasis on HSPs and heat shock transcription factors. Nevertheless, many upstream regulators and their regulatory mechanisms have yet to be revealed, along with the differences in the mechanisms between vegetative and reproductive tissues. The current study used next-generation sequencing to obtain a comprehensive and integrated spatiotemporal transcriptome and sRNAome of maize in response to heat stress. The analysis identified candidate heat response regulators in both vegetative and reproductive tissues. There were more DEGs specifically expressed in reproductive tissues than in vegetative tissues in response to heat stress. Components of the photosynthesis network in chloroplasts were hindered in leaves after heat stress, and a chloroplast-derived 19-nt sRNA accumulated to high levels in heat-stressed leaves. Moreover, we found that many TE-derived 24-nt siRNAs were dramatically reduced after exposure to high temperature in tassel, and the nearby genes tended to be down-regulated, indicating a possible coexpression between them (Figs. 4 and 5). The presently identified heat-responsive miRNAs and miRNA* species included some sequences that have not been reported to be involved in heat stress in either vegetative or reproductive tissues. Our analysis identified a collection of potential key regulators and provides insight into the molecular response mechanisms induced by heat stress in different maize tissues at both the vegetative and reproductive stages.

DEG Responses to Heat Stress at Different Developmental Stages Are Tissue Specific

Compared to the seedling stage, reproductive tissues are more sensitive to high temperature (Barnabás et al., 2008). However, only a few studies have examined the heat stress–induced molecular mechanisms in maize reproductive tissues. In this study, maize plants at both the vegetative and reproductive stages were treated with high temperature, and different tissues were collected for transcriptome analysis. Under the 2-h heat stress treatment, there were 40 heat-responsive genes detected in all of the analyzed tissues from both stages (Fig. 1C). By contrast, there were many more tissue-specific DEGs, and they were associated with diverse GO terms (Fig. 1; Supplemental Fig. S5). In tassels, for example, 1,782 and 2,341 genes were found to be specifically up- and down-regulated, respectively, in response to heat stress at 48 HAH (Fig. 1B). In summary, the mRNA-seq results underscore the distinct expression response profiles of different tissues at different developmental stages, indicating that the whole-plant heat response likely comprises tissue-specific molecular mechanisms.

Previous studies have investigated the genome-wide responses of maize seedlings to heat stress (Frey et al., 2015; Makarevitch et al., 2015; Li et al., 2017; Qian et al., 2019). The findings of Li et al. (2017) indicated the existence of universal response mechanisms in maize induced by distinct abiotic stresses; specifically, the report identified 167 genes that were differentially expressed in four separate analyses for salinity, drought, heat, and cold stress compared to the respective controls. Among the 2,346 identified DEGs (1,481 up- and 865 down-regulated), more than half were heat stress specific (957 up- and 436 down-regulated). Similar results were obtained by analyzing data from Makarevitch et al. (2015); Supplemental Fig. S16). The expression patterns of genes under heat stress were quite different from those induced by other stresses (Makarevitch et al., 2015). By comparing the DEGs from our seedling leaf data to those in the heat-stressed sample from Makarevitch et al. (2015), we identified 228 up-regulated genes present in both data sets (Supplemental Fig. S16A). Furthermore, GO analysis of these genes indicated enrichment of the following GO terms: response to heat, protein folding, response to H₂O₂ and response to temperature stimulus (Supplemental Fig. S16B). Differences in the identified DEGs between our data and the published data may be attributable to different growth conditions, sampling stages, heat treatment conditions, sequencing depth, and other relevant biological factors.

PPR Genes Induced by Heat Stress May Potentially Relieve the Adverse Effects of Stress on Chloroplasts

In Chinese cabbage (Brassica campestris ssp. pekinensis), the production of some chloroplast sRNAs is suppressed by high temperature (Wang et al., 2011). In this study, we identified a 19-nt chloroplast-derived sRNA that was dramatically induced (from ∼0.2 million reads to ∼0.8 million reads) by heat stress in maize leaves. The 19-nt sRNA is generated from the 3′ untranslated region of the psbH gene, which encodes a PSII subunit in chloroplasts (Fig. 3B). Moreover, the 19-nt sRNA sequence corresponds exactly to the binding site of PPR protein HCF152, which protects psbH mRNA from degradation (Zhelyazkova et al., 2012). As the psbH RNA is degraded, the HCF152 binding site is protected and accumulates. Thus, the accumulation of the 19-nt HCF152 footprint during heat stress implies that heat stress induces the degradation of psbH RNA. PPR genes compose a large gene family involved in mRNA metabolism in chloroplasts and mitochondria, and they play essential roles in maintaining organelle stability and plant development (Barkan and Small, 2014). In the current study, heat stress broadly suppressed the expression of photosynthetic genes in chloroplasts, while a large number of PPR genes were activated, and the accumulation of the aforementioned 19-nt sRNA were significantly enhanced (Figs. 2 and 3). One possibility is that PPRs protect and stabilize photosynthesis-related genes from degradation, as a mechanism to counter or relieve the adverse effects of heat stress. Accordingly, increased accumulation of the HCF152 PPR protein may have led to the over-accumulation of the 19-nt sRNAs, which may act as retrograde signals from chloroplast to nucleus. A proposed model is offered in Figure 9A to illustrate the transcriptomic responses in chloroplast after heat stress. Evolutionary analysis of the sequence of the 19-nt sRNA showed that it is highly conserved across angiosperms (Ruwe and Schmitz-Linneweber, 2012). It would be interesting to know whether heat stress–induced accumulation of the 19-nt sRNA is widespread in plants. It would also be interesting to know whether the complex of HCF152 and the 19-nt sRNA has any regulatory role or is simply a by-product of heat stress–induced psbH RNA degradation.

Figure 9.

Proposed models on maize transcriptomic responses to heat stress. A, Responses of photosynthesis-related genes in chloroplast after heat stress. Upon heat stress, photosynthesis-related genes are repressed in chloroplasts. PPR transcriptions are activated and they may participate in protecting photosynthesis-related genes from degradation, resulting in the production of elevated 19-nt sRNAs that might function as retrograde signals from chloroplast to nucleus. Bold lines represented up-regulated transcription after heat stress, and thin lines indicated down-regulated transcription after heat stress. SIG, sigma factor. B, Potential regulating networks of heat stress–responsive sRNAs in maize. Maize sRNAs showed different responses to heat stress in different tissues and stages. Diagram (left) showed that heat stress repressed the production of 24-nt siRNAs and the transcription of nearby genes. Among the reduced 24-nt siRNAs, phasiRNAs may lead to pollen developmental defects and male sterility. The orange lines indicated repression. Diagram (right) exhibited miRNA-mediated responses to heat stress. The stress-responsive miRNAs regulate their target genes at the posttranscriptional level and lead to morphological changes and physiological adaptations of maize plants under heat stress. S1, V3-leaf; S2, V3-stalk; S3, V3-root; S4, R1-tassel; S5, R1-leaf; S6, R1-silk; S7, R1-ear.

Heat-Induced miRNA and miRNA* Accumulation May Be Species Specific

miRNAs and miRNA* species have been found to play intricate roles in plant heat stress responses (Zhao et al., 2016). In Arabidopsis, several miRNAs are known to be specifically involved in plant thermotolerance. miR398 is rapidly induced upon heat stress to repress the transcripts of its target genes CSD1, CSD2, and CCS to promote the accumulation of HSF proteins (Guan et al., 2013). miR156 is also induced upon heat stress and helps maintain acquired thermotolerance by down-regulating SPL genes (Stief et al., 2014). Another study revealed that miR160 overexpression enhanced thermotolerance by repressing the expression of its target genes ARF10, ARF16, and ARF17, thereby activating a series of downstream HSP genes (Lin et al., 2018). In sunflower, miR396 regulates the expression of HaWRKY6 during early responses to high temperature, and transgenic miR396-resistant plants were found to have enhanced tolerance to heat stress (Giacomelli et al., 2012). In contrast, studies investigating the functions of miRNA* species in plant thermotolerance have been limited.

To compare the presently identified heat stress–responsive miRNAs in maize to those reported in other plant species, we analyzed previously published genome-wide miRNA profiling data for heat-stressed rice (Mangrauthia et al., 2017), tomato (Zhou et al., 2016), and Arabidopsis (Barciszewska-Pacak et al., 2015). All of these studies used vegetative-stage seedlings. To make the data more comparable, we used the differentially expressed miRNAs identified in maize vegetative-stage leaves for the comparison. In the four plant species, only miR166 was commonly up-regulated in response to heat stress. Four miRNA species (guide or passenger strand) were up-regulated in both maize and tomato (miR166, miR166*, miR168*, and miR396*), and five sequences, including the monocot-specific miR444 and miR528*, were up-regulated in both maize and rice after heat stress (Supplemental Fig. S17). One interesting phenomenon was that almost all of the differentially expressed heat stress–induced miRNAs in maize V3-leaves were up-regulated (Fig. 6A; Supplemental Fig. S10), which is distinct from the patterns in other plant species (Supplemental Fig. S17). Additionally, there were fewer down-regulated miRNAs in rice than in tomato and Arabidopsis, suggesting that monocots and dicots might use different regulatory mechanisms to endure heat stress. Compared to previous findings, we also identified a few tissue- and stage-specific heat stress–responsive miRNAs, including the following: miR1432 (specifically up-regulated in R1-leaves), miR168* (specifically up-regulated in V3-stage tissues and down-regulated in R1-stage tissues), miR162 (down-regulated only in V3-roots and R1-tassels), miR159 (specifically up-regulated in V3-stalks and down-regulated in V3-roots and R1-leaves), miR408 (specifically up-regulated in V3-stalks and down-regulated in V3-roots and R1-tassels), and miR827 (specifically up-regulated in R1-ears and R1-silks). Most of these expression changes were validated by northern blot (Fig. 6C). However, it is important to draw attention to the myriad sources of variation among different studies in different plant species, including different growth conditions, sampling methods, and heat stress treatment. As such, the identification of species-specific, stress-responsive miRNAs must always take into consideration the influence of experimental variation.

siRNA-Triggered TE Silencing May Be Largely Relieved after Heat Stress

Extensive analyses in plants suggest that most transposons are inactive and targeted by siRNAs (Lister et al., 2008; Lisch, 2009). However, several studies have shown that abiotic stresses can reverse transposon silencing (Grandbastien et al., 2005; Ito et al., 2011; Yasuda et al., 2013). For example, the LTR/Copia type retrotransposon ONSEN in Arabidopsis was found to be activated by heat stress (Ito et al., 2011). Generally, TE silencing can be triggered by 24-nt siRNAs derived from the TE loci themselves (Nosaka et al., 2012). In this work, the abundance of 24-nt sRNAs was dramatically reduced by more than half after long-term (48-h) heat stress in tassels and roots, and most of the reduction occurred at TE loci (Fig. 4A). Although these results suggest that heat stress may greatly reduce the abundance of TE-derived 24-nt siRNAs and reactivate TE transcription, the transcription level of these TEs in response to heat stress requires confirmation. In this study, the TE loci that generated 24-nt siRNAs reduced by heat stress were enriched near genes within 1 kb (Fig. 5A), and in some tissues, the down-regulated genes were also enriched near these particular TE loci (Fig. 5C). These results suggest that the production of 24-nt siRNAs at TE loci may be correlated with the expression of nearby genes. However, the putative mechanism that controls the relationship among TEs, TE-derived 24-nt siRNAs, and nearby genes in response to heat stress in maize requires further in-depth study.

On the other hand, the heat-responsive 24-nt sRNAs also included 24-nt phasiRNAs, that is, phased secondary siRNAs produced from PHAS loci. Zhai et al. (2015) reported 176 24-nt PHAS loci in the maize genome and an enrichment of 24-nt phasiRNAs in the meiotic stage of tassels that were important for pollen development. In the current study, ∼8.85% of 24-nt sRNAs were produced from 24-nt PHAS loci in the control tassel sample; after long-term heat stress, the percentage was significantly reduced to 4.58% (Supplemental Fig. S18). These results provide experimental evidence that the production of 24-nt phasiRNAs in tassels was severely repressed by high temperature stress, possibly indicating disrupted pollen development.

Based on present results, we hypothesize that maize orchestrated sRNAs including siRNAs and miRNAs to acclimatize to hot environment (Fig. 9B). Heat stress repressed the production of 24-nt siRNAs and the transcription of nearby genes, maybe resulting in the activation of a plethora of genes participating in stress tolerance. Decrease of phasiRNAs may account for pollen developmental defects and male sterility. Moreover, the heat stress–responsive miRNAs may lead to morphological changes and physiological adaptations of maize plants through regulating their target genes at the posttranscriptional level under heat stress. The exact molecular mechanisms of sRNAs functioning in heat stress tolerance of maize plants remain elusive and require further study.

MATERIALS AND METHODS

Plant Materials and Growth Conditions

Maize (Zea mays ssp. mays) ‘B73’ plants (Jiao et al., 2017) were cultivated in a phytotron at 25°C with 65% relative humidity under a 14-h/10-h light/dark cycle. Heat stress was conducted at the V3 stage and R1 stage.

Plants were subjected to 38°C heat treatment with comparable humidity for 2, 24, and 48 h. After heat stress, tissues at the V3 stage (leaves, stalks, and roots) and R1 stage (leaves, tassels, ears, and silks) were collected and immediately frozen in liquid nitrogen. Samples from at least five plants were pooled for each biological replicate, and at least two biological replicates were performed.

Chlorophyll Exaction and Determination

Maize leaves were ground in 96% (v/v) ethanol (0.1 g of fresh weight leaves in 20 mL of 96% [v/v] ethanol). Ground tissue was centrifuged at 8,000 rpm for 5 min to pellet any insoluble material after being soaked overnight in darkness (to prevent chlorophyll degradation) at 4°C. The contents of chlorophyll a and b were determined by measuring at 665 and 649 nm on a spectrophotometer (Sunrise absorbance microplate reader, Tecan) and computed according to Wintermans and de Mots (1965).

Analysis of Pollen Vigor

Pollens from at least three individual plants were pooled as one biological replicate and stained with 1% I₂-KI solution to examine the vigor of pollen grains. Three replicates were performed for the analysis.

RNA Extraction and Library Preparation

Total RNA was extracted from the maize tissues using TRIzol reagent (Molecular Research Center) according to the manufacturer’s instructions. Agilent Bioanalyzer 2100 system was used to check the quality of the extracted RNA, and only RNA of high integrity was used for the analysis.

For the mRNA-seq library, TruSeq RNA Sample Prep Kit v2 (RS-122-2001, Illumina) or NEBNext Ultra RNA Library Prep Kit for Illumina (New England Biolabs) was used according to the manufacturer’s recommendations. The samples were pooled and sequenced (150-bp paired-end reads) on the Illumina HiSeq 2500 platform or on the Illumina HiSeq 4000 platform at Berry Genomics (Beijing, China).

For the sRNA-seq library, 20 µg of total RNA for each sample was run in a 15% polyacrylamide/urea gel after denaturation, and gel slices corresponding to 15- to 40-nt RNA fragments were separated. The RNA was precipitated by C₂H₃NaO₂ and ethanol and used for sRNA library construction with the NEBNext Multiplex Small RNA Library Prep Set for Illumina (New England Biolabs, E7300S). The pooled sRNA libraries were sequenced on the Illumina HiSeq 2500 platform (Berry Genomics) to produce 50-bp single-end reads.

For degradome-seq library construction, RNA samples were extracted from R1-silk. The degradome-seq library was constructed based on a previously described method (Zhai et al., 2014). Degradome-seq was performed on the Illumina HiSeq 2500 platform (Berry Genomics) to produce 50-bp single-end reads.

mRNA-seq Data Processing

All raw mRNA-seq reads that passed the FastQC quality control steps were trimmed to remove the adapter sequences and then mapped to the B73 RefGen_v4 genome (Jiao et al., 2017) using Hisat2 software (Kim et al., 2015) with default settings. Only reads that mapped to unique positions were used for further analyses.

sRNA-seq Data Processing

The 3′ adapter sequences (AGATCGGAAGAGC) were removed from the Illumina sRNA-seq reads using cutadapt v1.9.1 (Martin, 2011), and only 18- to 26-nt reads were used for subsequent analysis. Adapter-free reads mapped to ribosomal RNA, tRNA, small nucleolar RNA, and small nuclear RNA sequences were filtered using Bowtie 2 (Langmead and Salzberg, 2012), and the remaining reads were mapped to the B73 RefGen_v4 genome (Jiao et al., 2017) using ShortStack (Johnson et al., 2016). The genome was divided into 500-bp windows to calculate the number of reads whose 5′ end nucleotides mapped to a given window. To quantify miRNA expression, adapter-free reads were counted and assigned to each miRNA (miRbase v21 and miRNest 2.0), with a 1-nt shift allowed on the two ends. Reads mapped to each of the 24-nt PHAS loci (Zhai et al., 2015) in the maize genome were also counted. RPM values were calculated for normalization for each window. sRNA abundance was compared between different treatments using the R package DESeq2 (Love et al., 2014).

Degradome-Seq Data Processing

The raw sequences were filtered to remove low-quality reads. The adapter sequences (TGGAATTCTCGGG) were removed using in-house perl script. The remaining sequences were then analyzed using CleaveLand4 (Addo-Quaye et al., 2009). Degradome sites were selected with P-value < 0.05.

Differential Gene Expression and Enrichment Analysis

EdgeR (Robinson et al., 2010) was used to normalize gene expression levels and identify statistically significant DEGs (FDR < 0.05 and fold-change > 2).

GO terms enriched in the up-/down-regulated gene sets from individual time point comparisons of heat stress treatment were identified using Fisher test as implemented in agriGO v2.0 (Tian et al., 2017). Each gene set was compared against the full set of genes in the B73 RefGen_v4 genome (Jiao et al., 2017) as background. GO terms with FDR < 0.05 were identified as significantly enriched.

RT-qPCR Verification

cDNA was synthesized using reverse transcriptase Moloney murine leukemia virus (PrimeScript RT Reagent Kit with gDNA Eraser, TaKaRa), and qPCR experiments were conducted on the StepOne Plus Real-Time PCR System (Applied Biosystems). Random primer was used to synthesize cDNA for all genes. The reaction mixture contained 10 µL of 2× SYBR Green premix, 5 µL of template (∼20 ng), 0.8 µL of PCR forward primers (10 µm), 0.8 µL of PCR reverse primers (10 µm; Supplemental Data Set 13), and 4.2 µL of nuclease-free water. Ubiquitin was used as the reference gene (Zm00001d045049), and three biological replicates were performed for each sample.

sRNA Northern Hybridization

Total RNA (10 µg for each sample) was separated in a 15% polyacrylamide/urea gel after denaturation at 70°C for 10 min. After the RNA was transferred onto a neutral nylon membrane (Hybond-NX, GE Healthcare), crosslinking to the membrane was performed using 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride (Sigma). The probes (Supplemental Data Set 13) labeled with biotin were hybridized with sRNAs on the membrane, and stabilized streptavidin-horseradish peroxidase was used to detect the biotin signal. Three biological replicates were performed for each sample.

miRNA Target Prediction

miRNA target genes were predicted using the psRNATarget program (Dai et al., 2018; http://plantgrn.noble.org/psRNATarget/) with a maximum expectation of 3.

Statistical Analysis

R software was used to perform the hyper-geometric tests to analyze the relationship between TEs and their proximal genes. The equation is as follows (Barash et al., 2001):

C represents Combination. N is the number of expressed genes and M indicates the number of down-regulated genes. n and k represent the number of expressed genes and down-regulated genes within 1 kb of TEs producing 24-nt down-regulated sRNAs, respectively.

Primers and Probes

All primers used for real-time PCR and probes used for northern blotting are listed in Supplemental Data Set 15.

Accession Numbers

The mRNA-seq and sRNA-seq Illumina reads for all samples have been deposited into the Sequence Read Archive at the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/sra) under accession number PRJNA520822, and degradome-seq data were deposited under the accession number PRJNA542014.

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. Morphological appearance of maize at V3 stage under heat stress.

• Supplemental Figure S2. Phenotype of different maize tissues at R1 stage under heat stress.

• Supplemental Figure S3. General analysis of mRNA-seq data for different maize tissues under heat stress.

• Supplemental Figure S4. Verification for mRNA-seq data by RT-qPCR.

• Supplemental Figure S5. GO enrichment analysis of differentially expressed genes in response to heat stress in maize.

• Supplemental Figure S6. Comparative analysis of heat stress-responsive genes in leaves between V3 and R1 stages.

• Supplemental Figure S7. Expression patterns of photosynthesis-related genes encoded by the nuclear genome in maize leaves.

• Supplemental Figure S8. Length distribution of sRNAs from different maize tissues after heat stress.

• Supplemental Figure S9. miRNAs in response to heat stress in different maize tissues at 2 HAH.

• Supplemental Figure S10. Expression patterns of miRNAs in response to heat stress in different tissues.

• Supplemental Figure S11. Total abundance of 21-nt miRNAs in different maize tissues at the vegetative (V3) and reproductive (R1) stages under heat stress.

• Supplemental Figure S12. Most abundant miRNAs were differentially expressed to heat stress.

• Supplemental Figure S13. Expression patterns of miRNAs and miRNA*s in response to heat stress.

• Supplemental Figure S14. Read mapping of miR168 (miR168-5p) and miR168* (miR168-3p) visualized by Integrative Genomics Viewer (IGV).

• Supplemental Figure S15. Hierarchical tree graph of GO terms of differentially expressed miRNA target genes in response to heat stress.

• Supplemental Figure S16. Comparative analysis of heat stress-responsive genes.

• Supplemental Figure S17. Comparative analysis of heat stress-responsive miRNAs in different plant species.

• Supplemental Figure S18. Heat stress dramatically repressed the production of 24-nt phasiRNAs in tassels.

• Supplemental Data Set S1. Summary of mRNA-seq libraries of different maize tissues under heat stress.

• Supplemental Data Set S2. Correlation coefficients of mRNA-seq biological replicates.

• Supplemental Data Set S3. List of differentially expressed genes in different maize tissues under heat stress (FDR < 0.05, fold-change > 2).

• Supplemental Data Set S4. Heat stress-responsive genes associated with photosynthesis-related genes encoded in the maize nuclear genome.

• Supplemental Data Set S5. Summary of sRNA-seq data for different maize tissues under heat stress.

• Supplemental Data Set S6. Expression of PPR genes in maize under heat stress.

• Supplemental Data Set S7. DSRs in response to heat stress in V3-root and R1-tassel of maize.

• Supplemental Data Set S8. Genes within 1 kb of TEs in the maize genome.

• Supplemental Data Set S9. Expression of 169 down-regulated genes within 1 kb of TEs producing down-regulated 24-nt sRNAs at 48 HAH in R1-tassels.

• Supplemental Data Set S10. Expression of miRNAs in different maize tissues under heat stress.

• Supplemental Data Set S11. Expression patterns of maize sRNA pathway genes in the RNA-seq data.

• Supplemental Data Set S12. Validated targets of miR168* and miR528* by degradome sequencing.

• Supplemental Data Set S13. Target gene prediction of differentially expressed miRNAs in response to heat stress in maize.

• Supplemental Data Set S14. Validated targets of heat stress responsive miRNAs by degradome sequencing.

• Supplemental Data Set S15. Primers and probes used in this work.