CYCLIC NUCLEOTIDE-GATED ION CHANNELs 14 and 16 Promote Tolerance to Heat and Chilling in Rice.

Calcium signaling has been postulated to be critical for both heat and chilling tolerance in plants, but its molecular mechanisms are not fully understood. Here, we investigated the function of two closely related cyclic nucleotide-gated ion channel (CNGC) proteins, OsCNGC14 and OsCNGC16, in temperature-stress tolerance in rice (Oryza sativa) by examining their loss-of-function mutants generated by genome editing. Under both heat and chilling stress, both the cngc14 and cngc16 mutants displayed reduced survival rates, higher accumulation levels of hydrogen peroxide, and increased cell death. In the cngc16 mutant, the extent to which some genes were induced and repressed in response to heat stress was altered and some Heat Shock factor (HSF) and Heat Shock Protein (HSP) genes were slightly more induced compared to the wild type. Furthermore, the loss of either OsCNGC14 or OsCNGC16 reduced or abolished cytosolic calcium signals induced by either heat or chilling stress. Therefore, OsCNGC14 and OsCNGC16 are required for heat and chilling tolerance and are modulators of calcium signals in response to temperature stress. In addition, loss of their homologs AtCNGC2 and AtCNGC4 in Arabidopsis (Arabidopsis thaliana) also led to compromised tolerance of low temperature. Thus, this study indicates a critical role of CNGC genes in both chilling and heat tolerance in plants, suggesting a potential overlap in calcium signaling in response to high- and low-temperature stress.

Plant growth and development is highly regulated by environmental factors including temperature. Extreme temperatures cause an array of biochemical, physiological, and morphological changes in plants and often negatively impact plant productivity (Hatfield and Prueger, 2015;Liu et al., 2016;Leeggangers et al., 2017). Developing plant varieties that are tolerant to high and low temperatures is a challenge for researchers in plant biology and agriculture (Atkinson and Porter, 1996). Both high and low temperatures induce complex responses in plants, some of which may provide protection against stresses and damage (Wada et al., 1990;Murakami et al., 2000). Primary temperature sensing is thought to occur on membranes potentially through membrane fluidity and membrane-localized channels or receptors (Murata and Los, 1997). Sensing is followed by the generation of secondary messengers, including Ca 21 and reactive oxygen species (ROS; Yan et al., 2006). Both heat and cold induce transient Ca 21 influx into the cell cytoplasm (Finka et al., 2012;Ma et al., 2015), and repeated cold treatment can induce repetitive Ca 21 transients (Krebs et al., 2012). ROS, at moderate amounts, may induce ROS scavengers and other protective mechanisms for stress tolerance (Prasad et al., 1994). Excess ROS cause oxidative damage to cells, leading to growth defects or initiation of programmed cell death (Mittler, 2002). The secondary Ca 21 signals and ROS can initiate another cascade of signaling events, which may have different signaling and physiological consequences depending on the primary stimuli . In heat signaling, a transient increase of Ca 21 is thought to promote expression of the calmodulin (CaM) genes (Liu et al., 2003), and CaM can regulate the activities of CaM-binding protein kinase or heat-activated mitogenactivated protein kinase (Nguyen and Shiozaki, 1999;Liu et al., 2008). The activated kinases phosphorylate heat shock factors (HSFs) and promote the expression of Heat Shock Proteins (HSPs), leading to the heat shock response (HSR) and the onset of thermotolerance (Saidi et al., 2011). HSPs are chaperone proteins important for controlling damage from denatured and aggregated proteins caused by heat (Ohama et al., 2017). HSPs can be induced by both the HSR and an active unfolded protein response in the endoplasmic reticulum (ER); the unfolded protein response can stabilize, renature, and degrade unfolded proteins to enhance survival under heat (Fragkostefanakis et al., 2016;Rieu et al., 2017). In cold signaling, it is thought that cold sensors located on membranes trigger Ca 21 influx, leading to activation of kinases that ultimately modulate the activity and abundance of Inducer of C-repeat Binding Factor (CBF) Expression1 (ICE1). In Arabidopsis (Arabdopsis thaliana), AtICE1 regulates the expression of CBF genes as well as other cold-responsive transcription factors, resulting in increased freezing tolerance (Stockinger et al., 1997;Liu et al., 1998;Zhang et al., 2019). In rice (Oryza sativa), OsICE1 is phosphorylated by mitogen-activated protein kinase3, resulting in a higher protein stability, which leads to the accumulation of trehalose and improved chilling tolerance .
Interestingly, deficiency in AtCNGC2 resulted in more heat susceptibility at the reproductive stage in contrast to the enhanced heat tolerance at seedling stage (Katano et al., 2018), suggesting a developmental stagedependent function of CNGCs in stress tolerance. AtCNGC6 was shown to be activated by cytosolic cAMP and to promote expression of HSP genes and acquired thermotolerance (Gao et al., 2012). AtCNGC16 plays an essential function in pollen fertility under heat stress (Tunc-Ozdemir et al., 2013). The diverse function of CNGCs is thought to result from their functions in modulating calcium influx. AtCNGC2 is shown to affect Ca 21 influx into leaf cells to prevent Ca 21 -induced hydrogen peroxide (H 2 O 2 ) accumulation, cell death, and leaf senescence . In rice, 16 CNGC members have been identified (Nawaz et al., 2014), but only two are functionally characterized. The OsCNGC13 gene is a maternal sporophytic factor that functions in regulating rice seed setting and the OsCNGC9 gene was recently shown to mediate pathogen-associated molecular pattern-induced Ca 21 influx in plant immunity Wang et al., 2019).
In this study, we found that the rice OsCNGC14 and OsCNGC16 genes play critical roles in heat tolerance as well as low-temperature tolerance. In addition, they affect Ca 21 influx to the cytosol in response to heat or cold treatments. Our study thus provides evidence for the role of CNGC proteins in calcium regulation and uncovers a further function of CNGC proteins in chilling and freezing tolerance in addition to their known roles in heat tolerance in plants.

Generation of the cngc14 and cngc16 Mutants
Given that AtCNGC2, AtCNGC4, and PpCNGCb play key roles in thermotolerance in Arabidopsis and moss (Finka et al., 2012), we investigated the roles of their rice homologs in temperature-stress tolerance. The complementary DNA (cDNA) sequences of these three genes were used to Blast search the rice genome sequences, and their most similar genes in rice were identified as OsCNGC14, OsCNGC15, and OsCNGC16 (Supplemental Fig. S1). Because we were not able to generate mutants in OsCNGC15 as in OsCNGC14 and OsCNGC16, we focused our study on the latter two genes. We first determined the subcellular localization of these two rice CNGC proteins by fusing them with GFP. Transient expression of these fusion proteins in Nicotiana benthamiana revealed that OsCNGC14 and OsCNGC16 are both localized to the plasma membrane (Supplemental Fig. S2). A search of the Rice Expression Profile Database showed that the OsCNGC14 and OsCNGC16 genes are expressed in most organs, including leaves, stems, inflorescence, and reproductive organs, and OsCNGC14 has a much higher expression level than OsCNGC16 in general (Supplemental Fig. S3, A and B). We assessed the expression of these genes in response to heat and cold treatments by reverse transcription quantitative PCR (RT-qPCR). The OsCNGC14 gene was induced by 42°C treatment to a maximum of 5.6-fold at 3 h, dropped at 6 h, and came down to almost the basal level at 9 h of heat treatment. The OsCNGC16 gene was induced to a maximum of 3.4-fold at 3 h and decreased to almost the basal level at 6 h of heat treatment (Supplemental Fig. S4A). In response to the 6°C chilling treatment, the expression of OsCNGC14 was increased by 1.6-fold at 1 h and dropped to the basal level at 3 h, whereas expression of the OsCNGC16 gene reached a maximum of 3.1-fold at 6 h compared to the basal level and dropped to basal level at 9 h (Supplemental Fig. S4B).
To assess the function of OsCNGC14 and OsCNGC16 genes in rice, we generated their mutants in the cv Nipponbare background via CRISPR/Cas9 technology (Xing et al., 2014). Homozygous mutants were identified among the T 0 population by sequencing the region around the CRISPR target sites. Three OsCNGC14 mutants (cngc14-1, cngc14-2, and cngc14-3) and three OsCNGC16 mutants (cngc16-1, cngc16-2, and cngc16-3) were obtained. The cngc14-1 mutant contained an 8-bp deletion (193-200 bp) and a 1-bp deletion at 224 bp relative to the translation start site, leading to a predicted 3-amino acid deletion (amino acids 65-67) in its predicted coding protein (Fig. 1, A and B). The cngc14-2 mutant contained a 25-bp deletion (176-200 bp) and a 1bp insertion at 225 bp, leading to a predicted 8-amino acid deletion (amino acids 59-66; Fig. 1, A and B). The cngc14-3 mutant contained a 28-bp deletion (197-224 bp), which is predicted to produce an altered protein sequence starting at amino acid 66 and a stop codon at amino acid 83 (Fig. 1, A and B). The cngc16-1 mutant contained 1-bp insertions at 585 and 718 bp, which are predicted to produce an altered protein sequence starting at amino acid 195 and a stop codon at amino acid 262 (Fig. 1, C and D). The cngc16-2 mutant contained a 3-bp deletion (586-588 bp) and a 1-bp insertion at 718 bp, causing a predicted 1-amino acid deletion at position 196 and a reading frame shift after position 240 ( Fig. 1, C and D). The cngc16-3 contained a 7-bp deletion (585-591 bp), a 1-bp insertion at 718 bp, and a 1-bp substitution at 719 bp, which is predicted to produce an altered protein sequence starting at amino acid 196 and a stop codon at amino acid 204 (Fig. 1, C and D).
We assayed the mRNA expression of OsCNGC14 and OsCNGC16 in their respective mutants by semiquantitative RT-PCR using primers spanning the CRISPR target sites. The transcripts of OsCNGC14 were found to be greatly reduced in all three cngc14 mutant alleles compared to the wild type, with cngc14-3 having the lowest expression ( Fig. 1B; Supplemental Fig. S5A). The transcripts of OsCNGC16 were also greatly reduced in all three cngc16 mutant alleles compared to the wild type with cngc16-3 having the lowest expression ( Fig. 1D; Supplemental Fig. S5B). The drastically reduced expression of OsCNGC14 or OsCNGC16 associated with some alleles is likely due to nonsense-mediated mRNA degradation. Taken together, cngc14-1 and cngc14-2 are likely reduction-of-function mutants, because lower levels of mRNA are produced and small deletions are present in the predicted proteins. The cngc14-3 allele and all three cngc16 alleles are most likely loss-offunction mutants, because the amounts of mRNA present are lower than in the wild type and the predicted proteins have large truncations and therefore are likely nonfunctional (Fig. 1, B and D).
OsCNGC14 and OsCNGC16 Promote Thermotolerance in Rice at Seedling Stage To determine whether OsCNGC14 and OsCNGC16 play roles in heat tolerance, the rice cngc14 and cngc16 mutants were subjected to a heat treatment. None of these cngc14 and cngc16 mutant plants had any overt abnormalities or defects when grown under normal conditions. Two-week-old seedlings were treated at 48°C for 48 h, and the loss of green color and survival rates were analyzed after a 2-week recovery growth at 26°C. The cngc14 and cngc16 mutants had more withered and yellower leaves compared to the wild-type cv Nipponbare plants ( Fig. 2A). The mutants also had lower survival rates than the wild type. The cngc14-1, cngc14-2, cngc14-3, cngc16-1, cngc16-2, and cngc16-3 had survival rates of 66%, 51%, 24%, 23%, 23%, and 30%, respectively, each of which was significantly lower than the rate of 70% for the wild type except that of cngc14-1 (Fig. 2B). The difference in survival rates among cngc14-1, cngc14-2, and cngc14-3 is consistent with the prediction that cngc14-3 is a loss-of-function allele and the other two are reduction-of-function alleles. We further tested seedling tolerance by longer exposure (72 h) to a moderate heat of 43°C. After the heat treatment, the cngc14 and cngc16 mutant plants had severely wrinkled and rough leaves and the cngc16 mutant developed necrotic spots on leaves, whereas the wild type was only slightly wrinkled at the leaf tip (Supplemental Fig. S6A). In addition, ion leakage in leaves, an indicator of membrane damage, was dramatically increased in the cngc14-2 mutant and three cngc16 mutants compared with wild-type plants after heat treatment but not in plants without heat treatment (Fig. 2C). Furthermore, whereas before heat treatment the mutants had low accumulation of H 2 O 2 , similar to that of the wild type, a higher accumulation relative to the wild type was observed after heat treatment, as assayed by staining with 3,39-diaminobenzidine (DAB; Fig. 2D). Heat also induced moderate cell death in the wild type, as indicated by trypan blue staining, and the cngc14 and cngc16 mutants exhibited more cell death compared with the wild-type plants after heat treatment but not without heat treatment (Fig. 2E). These results indicate that OsCNGC14 and OsCNGC16 positively contribute to heat tolerance in rice and that loss or reduction of their function leads to higher heat susceptibility associated with higher accumulation of H 2 O 2 and cell damage.
OsCNGC14 and OsCNGC16 Impact Cytosolic Calcium Increase in Response to Heat Treatment Because heat shock induces a transient increase in cytosolic Ca 21 (Saidi et al., 2009;Zheng et al., 2012), we investigated whether OsCNGC14 and OsCNGC16 affect the production of such a calcium signal. The cytosolic calcium concentration was detected by a nuclear export signal (NES) version of the calcium sensor Yellow Cameleon 3.6 (YC3.6) based on fluorescence resonance energy transfer (FRET; Krebs et al., 2012). This NES-YC3.6 reporter was introduced into the cngc14-2 and cngc16-1 mutants by crossing, and isogenic lines of the wild type and the cngc mutants with the same reporter gene were generated (Supplemental Fig. S7, A-C). Calcium concentrations, measured by the FRET signals (cpVenus/cyan fluorescent protein), were analyzed in stomatal cells where YC3.6 has a strong expression (Supplemental Fig. S7, D-G). Under normal growth conditions, no significant difference was observed Figure 1. Mutants of OsCNGC14 and OsCNGC16 generated by CRISPR/Cas9 technology. Diagram of the genomic sequences (A and C) and predicted protein structures (B and D) of the wild type (WT) and mutants of OsCNGC14 (A and B) and OsCNGC16 (C and D). A and C, Diagrams of the genomic DNA, with exons, introns, and untranslated regions indicated by black boxes, lines between boxes, and white boxes, respectively. Below the diagrams are sequence alignments between the mutants and wild type, with genomic sequences on the left and protein sequence on the right. Minus (2) and plus (1) signs and the letter S indicate the number of nucleotides deleted, inserted, and substituted, respectively, within the CRISPR/Cas9 target sequences T1 and T2. The asterisk indicates a stop codon generated. F, Forward primers; R, reverse primers. B and D, RNA expression and predicted protein structures of OsCNGC14 (B) and OsCNGC16 (D) in the wild type and the cngc mutants. On the left are products of semiquantitative RT-PCR of OsCNGC14 and OsCNGC16 in their respective mutants and the wild type. Actin1 was used as a control gene. On the right are diagrams of predicted protein structure in the wild type and different mutant alleles. For each structure, transmembrane helices are indicated by cylinders, the cyclic nucleotide binding domain by a rectangle, and the calmodulinbinding domain by a solid circle. Locations of amino acid alterations are indicated by arrows. Red lines indicate altered sequences from frameshift. between wild-type and cngc14 or cngc16 mutants in cytosolic Ca 21 levels (Supplemental Fig. S7H). Drastic differences in Ca 21 levels were observed when heat shock was applied to leaves. In response to heat treatment, the cytosolic Ca 21 concentration rose quickly in the wild type, reached the maximum at 15 s (seconds), and declined sharply (Fig. 3, A and B; Supplemental Movie S1). This "spike and shoulder"-shaped cytosolic Ca 21 was consistent with earlier reports in Arabidopsis and rice (Behera et al., 2015). Each of the subsequent heat treatments at the frequency of once every minute induced a calcium spike with a similar shape to the first spike (at least up to the seventh treatment; Fig. 3, A and B; Supplemental Movie S1). The cngc14-2 mutant displayed a significant decrease in cytosolic calcium spike after the first heat treatment, and it showed weaker spikes with subsequent treatments compared to the wild type (Fig. 3, A and C; Supplemental Movie S2). This indicates that Ca 21 kinetics was milder in the cngc14-2 mutant compared to the wild type. The cngc16-1 mutant exhibited a much stronger defect in calcium signal compared to the cngc14-2 mutant. The heat-induced transient increase of cytosolic Ca 21 was largely abolished in the cngc16-1 mutant (Fig. 3, A and D; Supplemental Movie S3). The maximal changes in FRET signals (apparent FRET efficiency/apparent rest FRET efficiency [DE app /E apprest ]) in cngc14-2 and cngc16-1 were 0.10 6 0.09 and 0.02 6 0.07, respectively, which were significantly lower than the signal 0.33 6 0.15 seen in the wild type for the first peak (Fig. 3, E and F). In conclusion, OsCNGC14 and OsCNGC16 are positive modulators of heat-induced increase of cytosolic calcium. It remains to be determined whether the lesser severity of defects in the cngc14-2 mutant compared to the cngc16-1 mutant is due to the reduction-of-function nature of the cngc14-2 allele.

Transcriptome Analysis of the cngc16 Mutant
To further investigate how CNGCs regulate heat tolerance in rice, we carried out RNA sequencing (RNA-Seq) to profile transcriptomes of the wild type and the cngc16-3 mutant. Rice plants grown at 26°C for 2 weeks were heat treated at 42°C, and RNAs were extracted from leaf tissues at 0, 3, and 24 h after , cngc14, and cngc16 seedlings before and after heat treatment. Seedlings were hydroponically grown at 26°C for 2 weeks and then treated at 48°C for 48 h followed by a recovery at 26°C for 2 weeks. Plants were photographed after the recovery. Scale bars 5 2 cm. B, Survival rates for the cngc14 and cngc16 mutants shown in A. Shown are the means 6 SD for three biological replicates (n 5 48 for each replicate). At least three independent experiments were carried out with similar results. C, Cellular ion leakage with or without 48°C treatment for 48 h in the third leaves of 2-week-old seedlings grown hydroponically at 26°C. Shown are means 6 SD from four biological replicates, with 12 plants for each line. Asterisks indicate significant difference by Students t test (*P , 0.05 and **P , 0.01). D and E, H 2 O 2 accumulation detected by DAB staining (D) and cell death detected by trypan blue staining (E) in the third leaves of 3-week-old seedlings grown in the soil before and after treatment at 48°C for 72 and 76 h, respectively (n 5 15). Scale bars 5 1 mm. heat treatment. Differentially expressed genes (DEGs) among samples were defined by a fold change of jlog 2 j $ 1 and a false discovery rate (FDR) #0.01. Overall, high correlation values were observed for three biological repeats for each genotype and time point combination (r $ 0.92; Supplemental Fig. S8A), indicating a good reproducibility of the samples. Principal component analysis (PCA) also revealed that the biological repeats are clustered together (Supplemental Fig. S8B). Interestingly, the 0-h samples of the wild type and the cngc16-3 mutant were closely clustered, indicating that without heat treatment, the wild type and the mutant are similar in RNA expression. PC1 divided the 0-h samples from the 3-and 24-h samples, indicating that temperature, and not genotype, has the largest impact on transcriptomes. PC2 separated the 0-, 3-, and 24-h samples and also showed that RNA expression at 3 h is more different from that at 0 and 24 h. The genotype difference was less pronounced than the heat and time differences, indicating that the wild type and the mutant are in general similar in transcriptome responses before and after heat treatment. In addition, PC1 distribution revealed that the cngc16-3 mutant at 3 and 24 h is more distant from the wild type or cngc16-3 at 0 h compared to the wild type at the corresponding 3-and 24-h time points (Supplemental Fig. S8B), suggesting that there are more changes in the cngc16-3 mutant than in the wild type in response to heat.
Comparisons of DEGs between the wild type and the cngc16-3 mutant, as well as between heat treatment and non-heat treatment, also indicated that the transcriptome responses to heat were similar in the wild type and the mutant. At 0, 3, and 24 h, a total of 150, 404, and 510 genes, respectively, had increased expression, whereas a total of 84, 44, and 576 genes had decreased expression, in the cngc16-3 mutant compared to the wild type (Supplemental Table S1). No significant gene ontology (GO) enrichment was observed for these DEGs between the cngc16-3 mutant and the wild type at 0, 3, or 24 h of heat treatment. When comparing heat treatment (3 or 24 h) to non-heat treatment (0 h), a larger number of DEGs (in thousands) was found for both the Figure 3. Heat-induced cytosolic Ca 21 changes in guard cells of cngc14 and cngc16 mutants. A, Emission images of cytosolic Ca 21 in the stomatal cells from leaves of wild-type (WT), cngc14-2, and cngc16-1 10-d-old seedlings expressing NES-YC3.6. Shown are representative images before heat shock treatment (60 s) and after heat shock buffer administration (75 and 120 s). The representative image of relative cytosolic Ca 21 is represented as the emission fluorescence ratio and scaled by a pseudo-color bar (bottom). Scale bars 5 50 mm. B to D, Normalized cytosolic Ca 21 FRET ratio in stomatal cells. The luminescence was recorded at 1-s intervals. Shown are representative images of calcium signals in stomatal cells of the wild type, cngc14-2, and cngc16-1. Lowercase letters refer to the time points at which these ratio images were taken in A. A total of 30 stomatal cells (one leaf/plant, 1-2 stomatal cells/leaf) in the third leaves from at least 15 different seedlings were measured. The two numbers underneath each genotype indicate the number of cells with the representative calcium pattern versus the number of total cells analyzed. E, Representative traces of calcium signals from the wild type and mutants are shown together. F, Scatter plots of the heat-induced changes in cytosolic calcium (DE app /E apprest ) in wild type, cngc14-2, and cngc16-1, as shown in B to D. The D cytosolic Ca 21 indicates the changes between the [Ca 21 ] peak and [Ca 21 ] rest . Data are from all 30 cells that were monitored for each genotype. Asterisks indicate significant difference by Student's t test (**P , 0.01).
wild type and the mutant (Fig. 4, A and B; Supplemental Tables S2 and S3). A Venn diagram showed significant overlap between the wild type and the cngc16-3 mutant in both up-and down-regulated DEGs at 3 and 24 h (Fig. 4, A and B). This suggests that the mutant is not drastically different from the wild type before or after heat treatment. GO-term analysis also revealed that the wild type and the mutant had almost identically enriched GO terms in DEGs at 3 h (or 24 h) versus 0 h (Supplemental Fig. S9, A-D). Heat-induced DEGs in the wild type and the cngc16-3 mutant were both enriched in GO terms including "response to salt stress," "response to reactive oxygen species," "cellular response to heat," "response to cold," and "chaperone-mediated protein folding" (Supplemental Fig. S9, A and B). The heatrepressed DEGs in the wild type and the cngc16-3 mutant were enriched in GO terms related to growth and development, such as "plant-type cell wall biogenesis," "sterol biosynthetic process," "metal ion transport," and "cellular amino acid biosynthetic process" (Supplemental Fig. S9, C and D).
We further examined DEGs between the wild type and cngc16-3 mutant at 3 or 24 h to determine whether they represented any unique pathways. GO-term analysis did not reveal any enriched terms; however, these DEGs have significant overlaps with DEGs of heat versus nonheat in the wild type. Among the 404 upregulated DEGs in the cngc16-3 mutant compared to the wild type at 3 h, 105 were heat induced and 94 were heat repressed in the wild type (Supplemental Fig.  S10A). This suggests that heat responses (induction or repression) are larger for some and smaller for other genes in the cngc16-3 mutant compared to the wild type (Supplemental Fig. S10A). This was similarly observed for DEGs between the wild type and the mutant at 24 h, that is, they have significant overlaps with heat-induced or heat-repressed genes in the wild type, and they are more responsive or less responsive to heat in the mutant compared to the wild type (Supplemental Fig. S10, C and D). Taken together, there was no qualitative difference between the wild type and the cngc16-3 mutant in terms of heat response at the transcriptome level, but the extent of responses for some genes differs between the mutant and the wild type.
Because the Arabidopsis cngc2 mutant had increased expression of HSP genes compared to the wild type in response to heat (Finka et al., 2012), we examined the expression of HSF and HSP genes in the cngc16 mutant. Among the 100 HSF and HSP genes in rice, 79 were identified as DEGs (log 2 fold change $1 or #21 and FDR , 0.01) using the normalized expression data in this study, which is almost exclusively due to their induction by heat (Supplemental Table S4). Among all of these HSF and HSP DEGs, only two displayed a significant expression difference between the mutant and the wild type at 24 h, and none were DEGs at 0 or 3 h (Supplemental Table S4). This indicates that expression levels of HSF and HSP genes in general are similar between the mutant and the wild type (,2-fold difference) with or without heat treatment. Cluster analyses of these 79 DEGs of HSF and HSP genes by their expression revealed five coexpression groups (Fig. 4C). Heatmap analysis showed that genes in groups 1 and 3 had lower relative expression levels in the cngc16-3 mutant compared to the wild type at 0 and 24 h, respectively ( Fig. 4C; Supplemental Table S4). By contrast, genes in groups 2, 4, and 5 had higher induction by heat in the cngc16-3 mutant compared to the wild type at 3 and/or 24 h ( Fig. 4C; Supplemental Table  S4). These results indicate that the rice cngc16-3 mutant has slightly increased induction of primarily HSF and HSP genes after heat treatment at 3 or 24 h. cngc16-3 does not have an attenuated induction of HSFs and HSPs during heat shock, which is in contrast to the enhanced induction observed in the . RNA-Seq analysis of the cngc16 mutant. A and B, Venn diagrams of heat-induced (A) and heatrepressed (B) genes in the wild type (WT) and cngc16-3. Two-week-old seedlings were treated at 42°C for 0, 3, and 24 h for RNA-Seq analysis. Expression at 3 or 24 h was compared to that at 0 h, and DEGs were defined by log 2 fold change $1 or #21, and FDR , 0.01. The DEGs were used to draw the Venn diagram via tools accessed at http://bioinfogp.cnb.csic. es/tools/venny/. Statistical significance (P , 0.01) of the overlap is indicated by an asterisk, calculated using the online program http://nemates.org/MA/progs/ overlap_stats.html. C, Heatmap representation of expression levels of DEGs of HSF and HSP genes at 0, 3, and 24 h after 42°C heat treatment in wild type and the rice cngc16-3 mutant.
OsCNGC14 and OsCNGC16 Enhance Chilling Tolerance in Rice at Seedling Stage CNGCs have not been associated with cold responses in plants. The drastic calcium signal defect of the rice cngc14 and cngc16 mutants in response to heat promoted us to analyze their function in cold tolerance, where calcium signal is also thought to be a secondary messenger. The cngc14 and cngc16 mutants, as well as the wild type, were subjected to a chilling (cold but above freezing) treatment of 6°C for 72 h performed on 2-week-old seedlings followed by a 2-week recovery growth period at 26°C. Each of the mutants analyzed, namely cngc14-1, cngc14-2, cngc14-3, cngc16-1, cngc16-2, and cngc16-3, exhibited reduced growth and yielded yellower leaves after chilling treatment compared to the wild-type plants ( Fig. 5A; Supplemental Fig. S6B). In addition, all mutants had lower survival rates from chilling treatment compared to the wild type (Fig. 5B). cngc14-1, cngc14-2, cngc14-3, cngc16-1, cngc16-2, and cngc16-3 exhibited survival rates of 43%, 36%, 22%, 28%, 17%, and 27%, respectively, significantly lower than the rate of 65% for the wild type (Fig. 5B).
We subsequently examined the response to chilling stress in the cngc14 and cngc16 mutants at the cellular level. Chilling induced increased ion leakage in leaf cells of the wild-type plants due to damage to cellular membranes (Fig. 5C). The cngc14 and cngc16 mutant plants all had significantly increased ion leakage in leaves compared to the wild-type plants at 72 h after chilling, whereas they did not differ from the wild type before chilling treatment (Fig. 5C). Chilling induced accumulation of ROS, including H 2 O 2 , as assayed by DAB staining (Fig. 5D). Both the cngc14 and cngc16 mutants accumulated more H 2 O 2 than the wild-type plants after 96 h of chilling treatment, whereas both the mutants and the wild type had low amounts of H 2 O 2 before chilling treatment (Fig. 5D). Chilling also Figure 5. Mutants of OsCNGC14 and OsCNGC16 are chilling susceptible at seedling stage. A, Morphological phenotypes of wild-type (WT), cngc14, and cngc16 seedlings before and after chilling treatment. Seedlings were hydroponically grown at 26°C for 2 weeks and then treated at 6°C for 72 h followed by recovery at 26°C for 2 weeks. Plants were photographed after the recovery. Scale bars 5 2 cm. B, Survival rates of the cngc14 and cngc16 mutants after chilling treatment as in A. The data are the means 6 SD of three biological replicates (n 5 48 for each replicate). Three independent experiments were carried out with similar results. C, Cellular ion leakage with or without 6°C treatment for 72 h in the third leaves of 2-week-old seedlings grown hydroponically at 26°C. Data represent means 6 SD from four biological replicates with 12 plants for each line. Asterisks indicate significant difference by Student's t test (*P , 0.05 and **P , 0.01). D and E, H 2 O 2 accumulation detected by DAB staining (D) and cell death detected by trypan blue staining (E) in the third leaves of 3-week-old seedlings grown in soil before and after a treatment of 6°C for 96 h (n 5 15). Scale bars 5 1 mm.
induced cell death in the wild-type plants, as indicated by trypan blue staining (Fig. 5E). The cngc14 and cngc16 mutants exhibited more cell death compared to the wild type after 96 h of chilling treatment, whereas they did not show increased cell death before chilling (Fig. 5E). Taken together, these results indicate that OsCNGC14 and OsCNGC16 are important for chilling tolerance in rice and that the loss or reduction of their function leads to reduced chilling tolerance, which is associated with more H 2 O 2 accumulation and cell death.
OsCNGC14 and OsCNGC16 Impact Cytosolic Calcium Increase in Response to Chilling Treatment To investigate whether OsCNGC14 and OsCNGC16 might affect chilling-induced changes in cytosolic Ca 21 , we monitored calcium levels in stomatal cells of the wild type and the cngc mutants before and after chilling treatment. Chilling triggered a large increase in cytosolic Ca 21 in the wild type in 5 s, and subsequent chilling treatments induced very mild Ca 21 spikes, with no calcium change observed at the third treatment (Fig. 6, A and B; Supplemental Movie S4). In 17 of the 24 cells observed, the cngc14-2 mutant displayed a significant decrease in cytosolic Ca 21 spike following the first cold treatment, and there was almost no calcium change in any of the 24 cells observed following the second and subsequent chilling treatments ( Fig. 6A and C; Supplemental Movie S5). The cngc16-1 mutant had a stronger defect compared to cngc14-2 after the chilling treatment. No change in cytosolic Ca 21 was observed in the cngc16-1 mutant after chilling in 19 of 21 cells recorded (Fig. 6, A and D; Supplemental Movie S6). The cytosolic Ca 21 levels, measured by maximal changes in FRET signals (DE app /E apprest ) of YC3.6, were 0.22 6 0.14 in the wild type, 0.12 6 0.09 in the cngc14-2 mutant, and 0.05 6 0.04 in the cngc16-1 mutant (Fig. 6, E and F). These results thus indicate that OsCNGC14 and Figure 6. Cold-induced cytosolic Ca 21 changes in guard cells of cngc14 and cngc16 mutants. A, Emission images of cytosolic Ca 21 in the stomatal cells from leaves of wild-type (WT), cngc14-2, and cngc16-1 10-d-old seedlings expressing NES-YC3.6. Shown are representative images before chilling treatment (60 s) and after chilling treatment (65 and 120 s). The representative image of relative cytosolic Ca 21 is represented as the emission fluorescence ratio and scaled by a pseudo-color bar (bottom). Scale bars 5 50 mm. B to D, Normalized cytosolic Ca 21 FRET ratio in stomatal cells. The luminescence was recorded at 2-s intervals. Shown is the representative image of calcium signals in stomatal cells of the wild type, cngc14-2, and cngc16-1. Lowercase letters refer to the time points at which these ratio images were taken in A. At least 21 stomatal cells (one leaf/plant, 1-2 stomatal cells/ leaf) in the third leaves from at least 10 different seedlings were measured. The two numbers underneath each genotype indicate the number of cells with the representative calcium pattern versus the number of cells analyzed. E, Representative traces of calcium signals from the wild type and the mutants are shown together. F, Scatter plots of the cold-induced changes in cytosolic calcium (Dcytosolic Ca 21 ) in wild type, cngc14-2, and cngc16-1 as shown in B to D. The Dcytosolic Ca 21 indicates the changes between the [Ca 21 ] peak and [Ca 21 ] rest . Data are from all cells that were monitored (30, 24, and 21 cells for the three genotypes respectively). Asterisks indicate significant difference by Student's t test (**P , 0.01).
OsCNGC16 are positive modulators of chilling-induced increases in cytosolic Ca 21 .
Arabidopsis AtCNGC2 and AtCNGC4 Promote Chilling Growth and Freezing Tolerance Given that the OsCNGC14 and OsCNGC16 contribute positively to both heat and chilling tolerance in rice, we asked whether Arabidopsis AtCNGC2 and AtCNGC4 affect low-temperature responses in addition to heat tolerance. The Arabidopsis cngc2 and cngc4 mutants were assayed for hypocotyl elongation and rosette growth under chilling conditions. The wild-type seedlings displayed shorter hypocotyls when grown at 6°C compared to at 22°C in darkness. The cngc2 and cngc4 mutants had shorter hypocotyls but these differences were not statistically significant compared to the wildtype plants at 22°C (Fig. 7, A and B). However, the cngc2 and cngc4 mutants displayed significantly shorter hypocotyls compared to wild-type plants at 6°C (Fig. 7, A  and B). Chilling (4°C) also inhibited rosette growth in Figure  7. Arabidopsis mutants of AtCNGC2 and AtCNGC4 are chilling and freezing susceptible. A and B, Phenotypes (A) and quantification (B) of hypocotyl elongation of cngc2 and cngc4 in the dark at 22°C (for 36 h) and 6°C (for 4 d). Data represent means 6 SD from three biological replicates, each of which had at least 15 seedlings. C and D, Phenotypes (C) and quantification (D) of rosette growth of the wild type, cngc2, and cngc4 grown on soil at 22°C for 3 weeks and at 6°C for 45 d. Data represent means 6 SD from three biological replicates with 10 plants per replica. Double factor variance analysis shows interaction between genotype and temperature. E to G, Phenotypes (E) and survival rate quantifications (F and G) of the wild type and mutants after 28°C treatment for 3 h followed by a 1-week recovery. Data represent means 6 SD from nine repeats (one pot per repeat) for each genotype. Asterisks indicate significant difference by Student's t test (**P , 0.01). Scale bars 5 0.5 cm (A), 1 cm (C and D), and 2 cm (E). the light in the wild type. The fresh weight after 45 d of growth at 4°C was only 30% of that after 21 d of growth at 22°C (Fig. 7, C and D). Both cngc2 and cngc4 mutants exhibited more inhibited growth by chilling in the light, with smaller rosettes and reduced fresh weight compared to the wild type when grown at 22°C (Fig. 7C). Compared to the wild type, both mutants had greater reduction in fresh weight (not rosette outgrowth) at 4°C than at 22°C, as indicated by double variant analysis (G 3 T; Fig. 7D). In addition, both the cngc2 and cngc4 mutants exhibited yellow or chlorotic leaves when grown at 4°C, which was not observed in the wild type (Fig. 7C).
We further examined responses to freezing treatments in the Arabidopsis cngc2 and cngc4 mutants. When treated at 28°C for 3 h followed by a 1-week recovery period, the cngc2 mutants displayed a 32% survival rate, significantly lower than the 73% survival rate for the wild-type plants grown side-by-side in the same pot (Fig. 7, E and F). Similarly, the cngc4 mutants had a 28% survival rate, significantly lower than the 76% rate for the wild type grown in the same pot (Fig. 7, E and G). Taken together, these results indicate that the Arabidopsis cngc2 and cngc4 mutants are both impaired in chilling growth and freezing tolerance.

DISCUSSION
Here we report that the rice genes OsCNGC14 and OsCNGC16 have critical functions in heat and chilling tolerance. The loss of either OsCNGC14 or OsCNGC16 leads to a higher susceptibility to heat and chilling compared to the wild type. In addition, these two rice CNGCs are important for generating cytosolic calcium signals in response to heat and cold, which might further influence transcriptional reprogramming under these stress conditions. Heat tolerance was shown to involve the functions of four Arabidopsis CNGC genes (AtCNGC2, AtCNGC4, AtCNGC6, and AtCNGC 16) and a moss gene, PpCNGCb (Finka et al., 2012;Gao et al., 2012;Tunc-Ozdemir et al., 2013). This study demonstrates that rice CNGCs function in heat-stress tolerance. Moreover, these data provide evidence for a function of plant CNGCs in chilling/freezing tolerance. The Arabidopsis AtCNGC2 and AtCNGC4 genes, similar to rice OsCNGC14 and OsCNGC16 genes, are required for growth at chilling temperatures and for freezing tolerance. Although the interpretation of this defect could be complicated by the growth defect in the Arabidopsis cngc2 and cngc4 mutants, it corroborates a positive role of CNGCs in low-temperature responses observed in rice. Therefore, these findings suggest that the role of CNGCs encompasses a further stress tolerance: chilling and freezing tolerance.
Early studies found a negative impact on heat-stress tolerance from AtCNGC2 and AtCNGC4 in Arabidopsis as well as PpCNGCb in moss (Finka et al., 2012). Here, we find that two of their closest homologs in rice, OsCNGC14 and OsCNGC16, contribute positively to heat tolerance. The first scenario explaining this apparent contradiction is that basal tolerance is assayed in rice and acquired tolerance is assayed in Arabidopsis. A more comprehensive analysis of basal and acquired heat tolerance under different treatment conditions would be needed to compare the roles of Arabidopsis and rice CNGCs. Second, the function of CNGC in heat tolerance could be tissue-or developmental-stage dependent. Consistent with this hypothesis, the AtCNGC2 gene was found to be a negative regulator of heat tolerance at the seedling stage but a positive regulator of heat tolerance at the reproductive stage (Katano et al., 2018). Third, the loss of CNGC function might generate stress conditions that trigger acclimation when the stress level is above a threshold. It is possible that the extent of stress induced by the loss of AtCNGC2 or AtCNGC4 in Arabidopsis might be larger than that induced by the loss of OsCNGC14 or OsCNGC16 in rice, and therefore mutations of AtCNGC2 or AtCNGC4 in Arabidopsis induce stress acclimation and enhanced heat tolerance. Consistent with this hypothesis, the rice cngc16 mutant and the wild type had low numbers of DEGs and no enriched GO terms under nonstress conditions, suggesting that the mutant does not significantly up-or downregulate a pathway or process. Nevertheless, in contrast to AtCNGC2 and AtCNGC4, a distantly related gene in Arabidopsis, AtCNGC6, was shown to be a positive regulator of acquired heat tolerance (Gao et al., 2012). This indicates that different CNGCs might have different or even opposite impacts on heat tolerance in the same plant.
OsCNGC14 and OsCNGC16 likely are involved in the early signaling events of both heat and cold responses. Without OsCNGC14 or OsCNGC16, calcium spikes, whether from the first treatment or subsequent treatments of heat or chilling, were reduced in amplitude or totally abolished (Figs. 3 and 6). These results indicate that OsCNGC14 and OsCNGC16 are directly or indirectly required for Ca 21 influx to the plant cells in response to heat or cold. Similar calcium dynamics in response to heat or chilling have been observed between Arabidopsis and rice Kiegle et al., 2000;Mori et al., 2018;Ma et al., 2015). In Arabidopsis, two Ca 21 -permeable mechanosensitive channels, MCA1 and MCA2, were found to mediate chilling-induced calcium influx (Mori et al., 2018). The cold-induced cytosolic Ca 21 increase was reduced, but not abolished, in the mca1 mca2 double mutant, suggesting additional modulators or channels for cold sensing. In rice, chilling-tolerance divergence1 (COLD1), a G-protein signaling regulator, can modulate the activity of the channel protein for calcium influx (Ma et al., 2015). In Arabidopsis, AtCNGC6, which mediates the acquisition of thermotolerance, was reported to be required for heat-induced Ca 21 influx using the wholecell patch-clamp technique (Gao et al., 2012). In moss, PpCNGCb is also required for calcium influx under heat, although the calcium dynamics after heat is different from that in rice (Finka et al., 2012). The impact of OsCNGC14 and OsCNGC16 on calcium signals and tolerance to heat and chilling makes them candidates for channels responsible for the chilling response. Consistent with a function of OsCNGC14 or OsCNGC16 in early temperature-stress signaling, their loss of function or reduction of function led to more signs of stress and damage, such as ROS accumulation, and more cell death in leaves under heat or chilling (Figs. 2,D and E,and 5,D and E). The slightly higher induction of some HSF and HSP genes in the cngc16 mutant ( Fig. 4C; Supplemental Table S4) also suggests the perception of more severe stress in the mutant after heat treatment.
The study here reveals that the homologous genes OsCNGC14 and OsCNGC16 have similar functions in temperature-stress tolerance, but their functions are not redundant or totally overlapping. This indicates that either they function in parallel for maximum responses or they interact to form functional complexes. Complexes of heterodimers or heterotetramers have been observed for CNGCs. In Arabidopsis, a homotetramer of the pollen tube-specific AtCNGC18 is an active Ca 21 channel, but the heterotetramer of AtCNGC18 with AtCNGC7 (or AtCNGC8) is inactive . In rice, OsCNGC15 is very closely related to OsCNGC14 and OsCNGC16 (Supplemental Fig. S1), but a failure to generate its knockout mutant prevented investigation of its function. It will be interesting to examine whether these proteins form complexes among themselves and how complex formation affects their activities in calcium signal generation.
In sum, this study reveals a critical role of CNGCs in low-temperature tolerance as well as in heat tolerance in plants and suggests a shared molecular mechanism in calcium signaling in response to both temperaturestress conditions. Future study should reveal whether and how CNGCs form channels for calcium influx and how they respond to temperature changes and impact temperature-stress signaling.

Plant Materials and Growth Conditions
The rice genotype used in this study is Oryza sativa 'Nipponbare'. Freshly harvested seeds were treated for 2 weeks or more at 42°C to break dormancy, soaked in water at room temperature for 3 to 5 d, and then germinated for 1 d at 37°C. For hydroponic growth, seeds at a similar germination stage were transferred to a 96-well plate (with its bottom cut) sitting in water, as described previously (Huang et al., 2009). Seedlings were grown in chambers with 16 h light at 26°C. For soil growth, germinated seeds were transferred to a mix of vermiculite:special substrate:organic substrate (1:1:1) and cultured at 26°C in the greenhouse.
The Arabidopsis (Arabidopsis thaliana) mutants cngc2 (stock no. CS6523) and cngc4 (stock no. CS6523) were obtained from the Arabidopsis Biological Resource Center. For Arabidopsis growth, seeds were sown in a mixture of vermiculite:soil (1:1), cold treated at 4°C for 3 d, and germinated and grown in controlled chambers.

Plasmid Constructions and Rice Transformation
For generating CRISPR/Cas9 constructs, specific targets of guide RNA spacer sequences were designed using the CRISPR-PLANT web site (http:// www.genome.arizona.edu/crispr/CRISPRsearch.html). Two guide RNA targets were cloned into the PHUE411 construct (Xing et al., 2014) and expressed in cv Nipponbare by Agrobacterium tumefaciens-mediated transformation (Nishimura et al., 2006). For subcellular localization, the coding sequences from the genomic DNA were cloned first into the entry vector pDONR221 (Magnani et al., 2006) by BP Clonase II Enzyme mix (Invitrogen) and then into the destination vector phpt-pSATN1-GW (Tzfira et al., 2005) using LR Clonase II Enzyme mix (Invitrogen).
All primer sequences are listed in Supplemental Table S5.

Protein Subcellular Location
Nicotiana benthamiana leaves expressing GFP proteins were analyzed with a ZEISS LSM 780 confocal microscope (ZEISS Microsystems). The excitation lines for imaging GFP and mCherry fusions were 488 and 580 nm, respectively. Images were analyzed using ZEISS Microsystems software.

Heat and Cold Tolerance Assays
Heat and cold tolerance assays were performed using 2-week-old rice plants. Genotypes to be compared were planted side by side each occupying half a plate in the same 96-well plate and treated with 48°C for 48 h (heat) or 6°C for 72 h (chilling). Three plates were treated at the same time and at least three independent treatments were conducted.
Hypocotyl elongation assays were done on seeds sown on one-half strength Murashige and Skoog plates, stratified at 4°C for 3 d, and allowed to germinate at 22°C for 24 h. Seeds at a similar germination stage were then grown at 22°C for 36 h or at 6°C for 4 d in the dark. For measurement of rosette growth under chilling, seedlings were grown first at 22°C for 1 week followed by 45 d at 4°C. Freezing tolerance assays were done on 2-week-old seedlings grown at 22°C in chambers. Plants were moved to a 28°C chamber for 3 h before moving back to the 22°C chamber for 1 week of recovery. Plants of two different genotypes were planted side by side in the same pot for comparison. Nine replicate pots were freezing-stress treated at the same time.

DAB and Trypan Blue Staining
Accumulation of H 2 O 2 and cell death were detected by DAB and trypan blue staining, respectively, as described previously (Shen et al., 2015;Thordal-Christensen et al.,1997).

Ion Leakage Analysis
Ion leakage was measured as previously described, with some modifications (Charng et al., 2007). For each sample, three leaves were combined and cut into 0.5-cm pieces and incubated in 0.4 M mannitol at room temperature with gentle shaking for 3 h before the initial conductivity of the solution was measured with a conductivity meter (DDS-307A, Ningbo Hinotek Technology). Total conductivity of the solution was measured after incubation at 85°C for 20 min. The ion leakage rate was expressed as the percentage of initial conductivity divided by total conductivity.

RNA Extraction, RT-PCR, and RT-qPCR
Total RNAs were isolated from seedling tissues using TRIpure Reagent (Bioteke) according to the manufacturer's protocols. One microgram total RNA was reverse transcribed into cDNA using HiScript II RT SuperMix (Vazyme). The cDNAs were then used as templates for qPCR with gene-specific primers. Relative quantitative expression was normalized to the reference gene OsActin1 (LOC_Os03g61970) and calculated as described previously (Gao et al., 2012). The RT-qPCR was performed using AceQ qPCR SYBR Green Master Mix (Vazyme) and the Bio-Rad CFX96 real-time PCR detection system (Bio-Rad Laboratories). Primer sequences are listed in Supplemental Table S5.

RNA-Seq and Data Analysis
Two-week-old rice seedlings grown in soil at 26°C were heat treated (42°C), and leaf tissues were collected at 0, 3, and 24 h after treatment. Total RNAs were extracted using TRIpure Reagent (Bioteke) according to the manufacturer's protocol. The transcriptomes were sequenced using Illumina HiSeq X Ten (PG150, Illumina). The adaptor sequences and low-quality sequences were removed using Hisat2 (Kim et al., 2015). Approximately 4.7 GB of clean reads were generated from each sample. The clean reads were mapped to the rice reference genome (IRGSP-1.0). Gene expression normalization among samples was performed using DESeq2 (Love et al., 2014).
PCA analysis was carried out via the web tool https://www.omicshare. com/tools/Home/Soft/pca. The spearman correlation was calculated by the Python Pandas package. Venn diagram analysis was done via the web tool http://bioinfogp.cnb.csic.es/tools/venny/, and the statistical test for overlapping significance was based on the online tool http://nemates.org/MA/ progs/overlap_stats.html.

Ca 21 Imaging
FRET-based Ca 21 imaging was performed as described previously . For heat and cold treatment, images were acquired every 1 and 2 s, respectively.

Statistical Analysis
Data were processed in Excel version 2010 (Microsoft Corporation). Comparisons between group means were performed with Student's t test.

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. Phylogenetic tree of CNGC proteins in Arabidopsis, moss, and rice.
Supplemental Figure S3. Global expression profiles of OsCNGC14 and OsCNGC16 at various stages of development.
Supplemental Figure S4. Expression pattern of OsCNGC14 and OsCNGC16 genes in response to heat-and cold-stress conditions.
Supplemental Figure S5. RNA expression of OsCNGC14 and OsCNGC16 in the wild type and the cngc mutants.
Supplemental Figure S6. Mutants of OsCNGC14 and OsCNGC16 are heat and cold susceptible at seedling stage.
Supplemental Figure S7. Signals of NES-YC3.6 in the wild type and the cngc14 and cngc16 mutants.
Supplemental Figure S8. Overall analyses of transcriptomes of the wild type and the cngc16-3 mutant.
Supplemental Figure S9. GO enrichment analysis of DEGs between the cngc16-3 mutant and the wild type at 3 h and 24 h.
Supplemental Figure S10. Venn diagram showing overlaps of high temperature-responsive genes and DEGs of cngc16-3 mutant.
Supplemental Table S2. Heat-induced and heat-repressed genes in wildtype plants at 3 and 24 h.
Supplemental Table S3. Heat-induced and heat-repressed genes in the cngc16-3 mutant at 3 and 24 h.
Supplemental Table S4. Expression levels of HSFs and HSPs in the wild type and the cngc16-3 mutant at 0, 3, and 24 h.