OsMKKK70 Negatively Regulates Cold Tolerance at Booting Stage in Rice

Cold stress at the booting stage leads to a lower seed setting rate and seriously threatens the production of rice (Oryza sativa L.), which has become a major yield-limiting factor in higher-altitude and -latitude regions. Because cold tolerance at the booting stage (CTB) is a complex trait and is controlled by multiple loci, only a few genes have been reported so far. In this study, a function of OsMKKK70 (Mitogen Activated Protein Kinase Kinase Kinase 70) in response to CTB was characterized. OsMKKK70 expression was rapidly induced by cold stress at the booting stage. OsMKKK70 overexpression (OsMKKK70-OE) plants were more sensitive to cold stress at the booting stage with a lower seed setting and pollen fertility, but there was no significant difference between the osmkkk70 mutant and WT. Considering the effect of functional redundancy, we further tested the CTB response of osmkkk62/70 and osmkkk55/62/70, the double and triple mutants of OsMKKK70 with its closest homologs OsMKKK62 and OsMKKK55, and found that osmkkk62/70 and osmkkk55/62/70 displayed significantly increased CTB with a higher seed setting and pollen fertility, indicating that OsMKKK70 negatively regulates rice CTB. Moreover, under the low-temperature (LT) condition, the osmkkk62/70 mutant had slightly higher Gibberellin (GA) contents, increased expression of GA biosynthesis genes, and lower protein level of OsSLR1 in anthers than those in WT. By contrast, OsMKKK70-OE anther had a lower GA biosynthesis than that of WT. Together, these findings suggest that OsMKKK70 negatively regulates rice CTB by fine-tuning GA levels in anthers.


Introduction
Rice is an important cereal crop and widely cultivated around the world. Cold tolerance at the booting stage (CTB) is a major limiting factor for rice production and geographic distribution [1][2][3][4]. Cold stress at the booting stage severely injures microspore development and pollen fertility, which seriously threaten the yield and quality of rice [3]. Cold stress at the booting stage is more severe and frequent in high-latitude and highaltitude regions. In China, the annual loss of rice production due to low temperature is as high as 3-5 million tons [4]. Therefore, it is an urgent task to improve cold tolerance at the booting stage for rice breeding.
Cold tolerance is a complex quantitative trait and is controlled by multiple genes [5]. Cold stress at the booting stage is challenging in terms of phenotypic evaluation, which limits the identification of cold-related genes or the dissection of the underlying molecular mechanisms. Therefore, although some multiple QTLs controlling CTB have been mapped, only a few genes, including CTB1, CTB2, CTB4a, OsbZIP73, LTT1, OsWRKY53, OsLEA9, transcriptome analysis using the young rice panicles exposed to the low temperature (LT) of 15 • C at the booting stage. We identified approximately 1000 differentially expressed genes, and OsMKKK70 was one of LT-induced expressed genes. To confirm this result, we re-examined the expression of OsMKKK70 treated with LT at the booting stage, and found that OsMKKK70 was rapidly induced, peaked at 1 h of LT treatment, and showed a gradual descent thereafter ( Figure 1A). To study the biological function of OsMKKK70, the OsMKKK70 overexpression plants (OsMKKK70-OE) were generated [6]. Interestingly, OsMKKK70-OE plants exhibited obviously shriveled and pale anther morphology and lower seed setting rate at the normal condition (NC), which is similar to wild type (WT) subjected to LT injury, implying that OsMKKK70 might be involved in cold tolerance at the booting stage (CTB) [6]. We, therefore, selected OsMKKK70 for further characterization in this study.

OsMKKK70 Might Be a Novel Regulator of Cold Tolerance at Booting Stage
Cold stress at the booting stage is an important abiotic stress that severely affects the production and geographic distribution of rice [1][2][3][4]. To explore the cold-tolerance-related genes at the booting stage and the underlying regulatory pathways, we performed transcriptome analysis using the young rice panicles exposed to the low temperature (LT) of 15 °C at the booting stage. We identified approximately 1000 differentially expressed genes, and OsMKKK70 was one of LT-induced expressed genes. To confirm this result, we re-examined the expression of OsMKKK70 treated with LT at the booting stage, and found that OsMKKK70 was rapidly induced, peaked at 1 h of LT treatment, and showed a gradual descent thereafter ( Figure 1A). To study the biological function of OsMKKK70, the OsMKKK70 overexpression plants (OsMKKK70-OE) were generated [6]. Interestingly, OsMKKK70-OE plants exhibited obviously shriveled and pale anther morphology and lower seed setting rate at the normal condition (NC), which is similar to wild type (WT) subjected to LT injury, implying that OsMKKK70 might be involved in cold tolerance at the booting stage (CTB) [6]. We, therefore, selected Os-MKKK70 for further characterization in this study. (G) Quantification of relative stained pollen ratio in (E). Data are shown as means ± SE (n = 10). Each dot represents the result from one biological replicate, and error bars indicate means ± SE. p values were calculated by Student's t-test; ** p < 0.01. . Data are shown as means ± SE (n = 10). (G) Quantification of relative stained pollen ratio in (E). Data are shown as means ± SE (n = 10). Each dot represents the result from one biological replicate, and error bars indicate means ± SE. p values were calculated by Student's t-test; ** p < 0.01. First, we carefully examined the seed setting rate of OsMKKK70-OE and showed that it was significantly lower than those of the WT in both NC and LT conditions ( Figure 1B,C). After LT for 4 d at 15 • C, the relative seed setting rate of the OsMKKK70-OE plant was around 21.1%, while that of WT was around 38% ( Figure 1D). This result indicated that OsMKKK70-OE is hypersensitive to cold stress at the booting stage. Cold stress at the booting stage usually affects anther development and pollen fertility in rice, which results in a decreased seed setting rate [15,20]. Therefore, we also observed the pollen fertility and anther morphology of OsMKKK70-OE. It was shown that, under NC, the pollen fertility of OsMKKK70-OE was lower than that of WT ( Figure 1E,F). Upon LT treatment, the pollen fertility of the OsMKKK70-OE was even much lower than that of WT ( Figure 1E-G), which was consistent with the lower seed setting rate of the OsMKKK70-OE plant ( Figure 1B-D). In addition, under NC, the anther of OsMKKK70-OE was shriveled and pale, compared with the normal anther of WT ( Figure S1A). Upon LT treatment, the degree of shrunken and pale of OsMKKK70-OE anther was further enhanced ( Figure S1A). Collectively, these results indicated that overexpressing OsMKKK70 leads to more sensitivity to cold stress at the booting stage, suggesting that OsMKKK70 is involved in rice CTB.

OsMKKK70 Negatively Regulates Rice CTB
To confirm that OsMKKK70 is involved in CTB, we generated osmkkk70 mutants via CRISPR/Cas9-mediated genome editing [6]. The osmkkk70 mutant showed a comparable seed setting rate to WT under NC and LT treatment ( Figure S2A-C). Our previous work demonstrated that OsMKKK70 has two homologs (OsMKKK62 and OsMKKK55) with high sequence similarity and these three OsMKKK genes function redundantly [6]. Therefore, we also examined the cold tolerance of osmkkk55 and osmkkk62 single mutants, and found that osmkkk55 and osmkkk62 showed comparable cold tolerance to WT ( Figure S2D-I).
To further explore the biological function of OsMKKK70 in response to CTB, osmkkk62/70 double and osmkkk55/62/70 triple mutants were used for the cold stress assay [6]. osmkkk62/70 and osmkkk55/62/70 mutants had comparable seed setting rates to WT under NC (Figure 2A,B). However, under LT conditions, osmkkk62/70 and osmkkk55/62/70 mutants exhibited much higher seed setting rates and relative seed setting rates than those of the WT (Figure 2A-C). Indeed, the relative seed setting rate of the osmkkk62/70 was 69.8% and 59.2% after LT for 4 or 5 d at 15 • C, respectively; the osmkkk55/62/70 mutant was 72.2% (after LT for 4 d) and 60.1% (after LT for 5 d); while that of WT only reached 33.9% (after LT for 4 d) and 20.9% (after LT 5 d) ( Figure 2C). To confirm that osmkkk62/70 and osmkkk55/62/70 mutants increased the CTB, we performed two more independent replications in two consecutive years from 2019 to 2020. It was shown that osmkkk62/70 and osmkkk55/62/70 mutants reproducibly displayed a higher seed setting rate than WT in LT treatment ( Figure S3A-D). Combined with the OsMKKK70-OE being hypersensitive to LT and osmkkk62/70 and osmkkk55/62/70 mutants having increased CTB, these results suggest that OsMKKK70 plays the major role in negatively regulating rice CTB, and OsMKKK55 and OsMKKK62 might have minor and redundant functions in this process.

OsMKKK70 and OsMKKK62 Affect Pollen Fertility and Anther Development
It has been proposed that anther length and anther morphology are positively correlated with CTB and can be used as criteria for evaluating CTB in rice [9,13]. To study how OsMKKK70 negatively regulates rice CTB, we examined the pollen fertility and the anther morphology of osmkkk62/70 and osmkkk55/62/70. It was shown that, under NC, the pollen fertility of osmkkk62/70 and osmkkk55/62/70 mutants was comparable to WT plants ( Figure 3A,B). By contrast, under LT treatment, the fertile pollen ratio of osmkkk62/70 and osmkkk55/62/70 mutants was much higher than that of WT plants ( Figure 3A  and relative seed setting rate (C) of WT, osmkkk62/70, and osmkkk55/62/70 mutants grown under NC, and after LT treatment for 4 or 5 days. Data are shown as means ± SE (n = 10). Dotted lines were drawn to separate the NC and LT, and the comparison is among different genotypes in the same condition. Each dot represents the result from one biological replicate; error bars indicate means ± SE. Statistically significant differences are indicated by different lowercase letters (p < 0.05, one-way ANOVA with Tukey's significant difference test).

OsMKKK70 and OsMKKK62 Affect Pollen Fertility and Anther Development
It has been proposed that anther length and anther morphology are positively correlated with CTB and can be used as criteria for evaluating CTB in rice [9,13]. To study how OsMKKK70 negatively regulates rice CTB, we examined the pollen fertility and the anther morphology of osmkkk62/70 and osmkkk55/62/70. It was shown that, under NC, the pollen fertility of osmkkk62/70 and osmkkk55/62/70 mutants was comparable to WT plants ( Figure 3A,B). By contrast, under LT treatment, the fertile pollen ratio of osmkkk62/70 and osmkkk55/62/70 mutants was much higher than that of WT plants ( Figure 3A  (B,C) Seed setting rate (B) and relative seed setting rate (C) of WT, osmkkk62/70, and osmkkk55/62/70 mutants grown under NC, and after LT treatment for 4 or 5 days. Data are shown as means ± SE (n = 10). Dotted lines were drawn to separate the NC and LT, and the comparison is among different genotypes in the same condition. Each dot represents the result from one biological replicate; error bars indicate means ± SE. Statistically significant differences are indicated by different lowercase letters (p < 0.05, one-way ANOVA with Tukey's significant difference test).

OsMKKK70 and OsMKKK62 Negatively Regulate GA Contents in Anthers
Low-temperature treatment resulted in anther GA contents being markedly decreased and the application of exogenous GA can successfully rescue cold-injured pollen sterility [9,20]. Recently, it was reported that the oswrky53 mutant displays increased CTB by fine-tuning GA levels in anthers [9]. Moreover, it was also reported that OsMKKK70 functions upstream of OsWRKY53 via activating the OsMAPKK4-OsMAPK6-OsWRKY53 cascade in regulating rice grain size and leaf angle [6]. These results have motivated us to test whether OsMKKK70 also regulates rice CTB by fine-tuning GA contents in anthers. To this end, we quantified the endogenous GA levels in anthers of WT and mutant plants grown under both NC and LT. Given that osmkkk62/70 and osmkkk55/62/70 have a similar CTB phenotype, we only measured the GA level of osmkkk62/70 ( Figure 4). GA 1 , GA 3 , GA 4 , and GA 7 are bioactive GAs, and GA 4 and GA 7 are the predominant GAs in reproductive organs, while GA 1 and GA 3 are the main GAs in vegetative organs [6]. We found that the osmkkk62/70 mutant and WT accumulated similar levels of GA 4 and GA 7 under NC ( Figure 4A). Importantly, following 15 d at 18 • C LT treatment, the levels of GA 4 and GA 7 significantly decreased in WT relative to NC, whereas they obviously did not decrease in osmkkk62/70 ( Figure 4A). GA 1 and GA 3 contents were slightly decreased after LT treatment in both WT and osmkkk62/70, but the osmkkk62/70 mutant still had a higher level than WT under LT conditions ( Figure 4A). The levels of immediate precursors bioactive GA, including GA 9 , GA 12 , GA 24 , GA 5 , GA 19 , and GA 53 , were also higher in the osmkkk62/70 mutant than those in WT under both NC and LT ( Figure 4B,C). Additionally, the osmkkk62/70 mutant contained significantly more deactivated GAs (GA 8 , GA 34 , and GA 51 ) than in WT ( Figure 4D). Considering that GA 4 and GA 7 are predominant bioactive GAs in reproductive organs, and the osmkkk62/70 mutant had a higher level of GA 4 and GA 7 than in WT after LT treatment ( Figure 4A), these results demonstrate that the osmkkk62/70 mutant might accumulate more bioactive GAs than WT under LT treatment. Values are means ± SE (n = 10). Dotted lines were drawn to separate the NC and LT, and the comparison is among different genotypes in the same condition. Each dot represents the result from one biological replicate; error bars indicate means ± SE. Statistically significant differences are indicated by different lowercase letters (p < 0.05, one-way ANOVA with Tukey's significant difference test).

OsMKKK70 and OsMKKK62 Negatively Regulate GA Contents in Anthers
Low-temperature treatment resulted in anther GA contents being markedly decreased and the application of exogenous GA can successfully rescue cold-injured pollen sterility [9,20]. Recently, it was reported that the oswrky53 mutant displays increased CTB by fine-tuning GA levels in anthers [9]. Moreover, it was also reported that Os- Values are means ± SE (n = 10). Dotted lines were drawn to separate the NC and LT, and the comparison is among different genotypes in the same condition. Each dot represents the result from one biological replicate; error bars indicate means ± SE. Statistically significant differences are indicated by different lowercase letters (p < 0.05, one-way ANOVA with Tukey's significant difference test).
NC and LT ( Figure 4B,C). Additionally, the osmkkk62/70 mutant contained significan more deactivated GAs (GA8, GA34, and GA51) than in WT ( Figure 4D). Considering th GA4 and GA7 are predominant bioactive GAs in reproductive organs, and the mkkk62/70 mutant had a higher level of GA4 and GA7 than in WT after LT treatme ( Figure 4A), these results demonstrate that the osmkkk62/70 mutant might accumula more bioactive GAs than WT under LT treatment. To confirm this result, we examined the protein level of OsSLR1 (Slender rice 1, r DELLA protein) in anthers, as OsSLR1 degradation is promoted by GA and the OsSL level can be used as a marker of endogenous GA content [28]. Results showed that t osmkkk62/70 mutant and WT accumulated similar OsSLR1 protein levels under N however, after the LT treatment, the OsSLR1 protein level in the osmkkk62/70 mutant w markedly lower than that in WT ( Figure 5A). Simultaneously, we also examined t OsSLR1 protein level in OsMKKK70-OE anthers under NC. The result showed that t OsSLR1 protein level in OsMKKK70-OE was higher than that in WT ( Figure 5B). GAM (GA-regulated MYB domain transcription factor gene) and its target gene CYP703A3 a GA-induced anther specifically expressed genes [29]. We also examined the expressi level of these two marker genes and found that the expression levels of GAMYB a CYP703A3 in WT were comparable to those in osmkkk62/70 and osmkkk55/62/70 mutan under NC. However, at the LT condition, the expressions of GAMYB and CYP703A3 osmkkk62/70 and osmkkk55/62/70 mutants were significantly higher than those in W To confirm this result, we examined the protein level of OsSLR1 (Slender rice 1, rice DELLA protein) in anthers, as OsSLR1 degradation is promoted by GA and the OsSLR1 level can be used as a marker of endogenous GA content [28]. Results showed that the osmkkk62/70 mutant and WT accumulated similar OsSLR1 protein levels under NC; however, after the LT treatment, the OsSLR1 protein level in the osmkkk62/70 mutant was markedly lower than that in WT ( Figure 5A). Simultaneously, we also examined the OsSLR1 protein level in OsMKKK70-OE anthers under NC. The result showed that the OsSLR1 protein level in OsMKKK70-OE was higher than that in WT ( Figure 5B). GAMYB (GA-regulated MYB domain transcription factor gene) and its target gene CYP703A3 are GA-induced anther specifically expressed genes [29]. We also examined the expression level of these two marker genes and found that the expression levels of GAMYB and CYP703A3 in WT were comparable to those in osmkkk62/70 and osmkkk55/62/70 mutants under NC. However, at the LT condition, the expressions of GAMYB and CYP703A3 in osmkkk62/70 and osmkkk55/62/70 mutants were significantly higher than those in WT ( Figure 5C). Consistent with this result, the expressions of GAMYB and CYP703A3 in OsMKKK70-OE were significantly lower than those in WT ( Figure 5D). Collectively, these findings suggest that OsMKKK70 and its homologs might negatively regulate rice CTB by modulating the GA content in anther. ( Figure 5C). Consistent with this result, the expressions of GAMYB and CYP703A3 in OsMKKK70-OE were significantly lower than those in WT ( Figure 5D). Collectively, these findings suggest that OsMKKK70 and its homologs might negatively regulate rice CTB by modulating the GA content in anther.

OsMKKK70 Negatively Regulates GA Biosynthesis Gene Expression in Anther
In rice, OsGA20ox1, OsGA20ox3, and OsGA3ox1 are vital genes that catalyze to generate bioactive GAs and are highly expressed in reproductive organs [16,[30][31][32]. It was reported that LT markedly decreased the expression of OsGA20ox1, OsGA20ox3, and OsGA3ox1, and resulted in lower GA content [9]. To address whether OsMKKK70 functions in CTB via regulating GA biosynthesis genes expression, we first checked the expression of these GA biosynthesis genes in WT, osmkkk62/70, and osmkkk55/62/70 mutants. We found that, after LT treatment, the transcript levels of OsGA20ox1, OsGA20ox3, and OsGA3ox1 were significantly higher in osmkkk62/70 and osmkkk55/62/70 mutants than those in WT ( Figure 6A). Given the higher content of GA12 in the osmkkk62/70 mutant relative to WT under both NC and LT ( Figure 4B), we also examined the expression level of ent-kaurenoic acid oxidase (KAO), which plays key roles in the early steps of GA biosynthesis [16]. The result showed that osmkkk62/70 and osmkkk55/62/70 mutants had a higher expression level of KAO relative to WT under NC and LT treatment, which partially explained the greater amount of GA12 in the osmkkk62/70 mutant relative to WT

OsMKKK70 Negatively Regulates GA Biosynthesis Gene Expression in Anther
In rice, OsGA20ox1, OsGA20ox3, and OsGA3ox1 are vital genes that catalyze to generate bioactive GAs and are highly expressed in reproductive organs [16,[30][31][32]. It was reported that LT markedly decreased the expression of OsGA20ox1, OsGA20ox3, and OsGA3ox1, and resulted in lower GA content [9]. To address whether OsMKKK70 functions in CTB via regulating GA biosynthesis genes expression, we first checked the expression of these GA biosynthesis genes in WT, osmkkk62/70, and osmkkk55/62/70 mutants. We found that, after LT treatment, the transcript levels of OsGA20ox1, OsGA20ox3, and OsGA3ox1 were significantly higher in osmkkk62/70 and osmkkk55/62/70 mutants than those in WT ( Figure 6A). Given the higher content of GA 12 in the osmkkk62/70 mutant relative to WT under both NC and LT ( Figure 4B), we also examined the expression level of ent-kaurenoic acid oxidase (KAO), which plays key roles in the early steps of GA biosynthesis [16]. The result showed that osmkkk62/70 and osmkkk55/62/70 mutants had a higher expression level of KAO relative to WT under NC and LT treatment, which partially explained the greater amount of GA 12 in the osmkkk62/70 mutant relative to WT ( Figure S4). Deactivation of GA is mostly catalyzed by GA 2-oxidase [31]. Given that the osmkkk62/70 mutant accumulated more deactivated GAs (GA 8 , GA 34 , and GA 51 ) than WT ( Figure 4D), we also examined the expression of OsGA2ox genes. The results showed that osmkkk62/70 and osmkkk55/62/70 mutants had higher expression levels of OsGA2ox1 and OsGA2ox3 compared to WT under the LT condition ( Figure S4). To confirm that OsMKKK70 negatively regulates GA biosynthesis gene expression in anthers, we further examined the GA biosynthesis genes expression in OsMKKK70-OE under NC and LT treatment, and found that the transcript levels of OsGA20ox1, OsGA20ox3, and OsGA3ox1 in OsMKKK70-OE were lower than those in WT under both NC and LT treatment ( Figure 6B). Collectively, these results suggest that OsMKKK70 negatively regulates the GA content during cold stress at the booting stage, and consequently regulates the rice CTB.  Figure S4). Deactivation of GA is mostly catalyzed by GA 2-oxidase [31]. Given that the osmkkk62/70 mutant accumulated more deactivated GAs (GA8, GA34, and GA51) than WT ( Figure 4D), we also examined the expression of OsGA2ox genes. The results showed that osmkkk62/70 and osmkkk55/62/70 mutants had higher expression levels of OsGA2ox1 and OsGA2ox3 compared to WT under the LT condition ( Figure S4). To confirm that OsMKKK70 negatively regulates GA biosynthesis gene expression in anthers, we further examined the GA biosynthesis genes expression in OsMKKK70-OE under NC and LT treatment, and found that the transcript levels of OsGA20ox1, OsGA20ox3, and OsGA3ox1 in OsMKKK70-OE were lower than those in WT under both NC and LT treatment ( Figure 6B). Collectively, these results suggest that OsMKKK70 negatively regulates the GA content during cold stress at the booting stage, and consequently regulates the rice CTB.

OsMKKK70 Might Act Upstream of OsWRKY53 in Regulating Rice CTB
MAPK cascades have been shown to function upstream of several group Ia members of WRKY proteins. For example, OsWRKY53 is a phosphorylated substrate of Os-MAPK6, and functions downstream of the OsMKKK10-OsMKK4-OsMAPK6 and Os-MKKK70-OsMKK4-OsMAPK6 cascades to regulate seed size [6,33,34]. Recently, it has been found that OsWRKY53 negatively regulates rice CTB by fine-tuning GA levels in anther [9]. In this study, we found that osmkkk62/70 and osmkkk55/62/70 mutants showed increased CTB phenotypes, including higher seed setting rates and more GA contents in anthers at the LT condition, which is similar to the oswrky53 mutant as reported (Figures  2A-C, 4A-D and S3) [9]. In addition, OsMKKK70-OE lines showed a decreased seed setting rate, unhealthy anther morphology, and lower expression level of GA biosynthesis genes in anthers under NC, which is also similar to OsWRKY53-OE plants (Figures 1B-G

OsMKKK70 Might Act Upstream of OsWRKY53 in Regulating Rice CTB
MAPK cascades have been shown to function upstream of several group Ia members of WRKY proteins. For example, OsWRKY53 is a phosphorylated substrate of OsMAPK6, and functions downstream of the OsMKKK10-OsMKK4-OsMAPK6 and OsMKKK70-OsMKK4-OsMAPK6 cascades to regulate seed size [6,33,34]. Recently, it has been found that Os-WRKY53 negatively regulates rice CTB by fine-tuning GA levels in anther [9]. In this study, we found that osmkkk62/70 and osmkkk55/62/70 mutants showed increased CTB phenotypes, including higher seed setting rates and more GA contents in anthers at the LT condition, which is similar to the oswrky53 mutant as reported (Figures 2A-C, 4A-D and S3) [9]. In addition, OsMKKK70-OE lines showed a decreased seed setting rate, unhealthy anther morphology, and lower expression level of GA biosynthesis genes in anthers under NC, which is also similar to OsWRKY53-OE plants ( Figures 1B-G and S1) [6,9]. These results imply that the OsMKKK70-OsMKK4-OsMAPK6-OsWRKY53 cascade might also be involved in rice CTB. Notably, we found that smg1-1 (OsMKK4 mutant) and dsg1 (OsMAPK6 mutant) showed comparable relative seed setting rates to WT at the LT condition ( Figure S7A-F). This might be due to the decreased seed setting rates of smg1-1 and dsg1 under NC ( Figure S7A-F). In addition, we also overexpressed the constitutively active (CA) form of OsWRKY53 in the osmkkk62/70 background, and generated osmkkk62/70 CA-OsWRKY53 plants [6]. We found that, compared with WT and osmkkk62/70, osmkkk62/70 CA-OsWRKY53 plants showed a lower seed setting, decreased fertile pollen ratio, and shriveled and pale anther morphology under NC, which is very similar to OsWRKY53-OE plants ( Figure 7A-D) [6,9]. We further examined the GA biosynthesis genes expression in osmkkk62/70 CA-OsWRKY53 under NC. We found that the transcript levels of OsGA20ox1, OsGA20ox3, and OsGA3ox1 in CA-OsWRKY53 and osmkkk62/70 CA-OsWRKY53 were comparable, which were lower than those in the WT and osmkkk62/70 ( Figure S8). These results indicated that OsMKKK70 might function upstream of OsWRKY53 in regulating rice CTB. and S1) [6,9]. These results imply that the OsMKKK70-OsMKK4-OsMAPK6-OsWRKY53 cascade might also be involved in rice CTB. Notably, we found that smg1-1 (OsMKK4 mutant) and dsg1 (OsMAPK6 mutant) showed comparable relative seed setting rates to WT at the LT condition ( Figure S7A-F). This might be due to the decreased seed setting rates of smg1-1 and dsg1 under NC ( Figure S7A-F). In addition, we also overexpressed the constitutively active (CA) form of OsWRKY53 in the osmkkk62/70 background, and generated osmkkk62/70 CA-OsWRKY53 plants [6]. We found that, compared with WT and osmkkk62/70, osmkkk62/70 CA-OsWRKY53 plants showed a lower seed setting, decreased fertile pollen ratio, and shriveled and pale anther morphology under NC, which is very similar to OsWRKY53-OE plants ( Figure 7A-D) [6,9]. We further examined the GA biosynthesis genes expression in osmkkk62/70 CA-OsWRKY53 under NC. We found that the transcript levels of OsGA20ox1, OsGA20ox3, and OsGA3ox1 in CA-OsWRKY53 and osmkkk62/70 CA-OsWRKY53 were comparable, which were lower than those in the WT and osmkkk62/70 ( Figure S8). These results indicated that OsMKKK70 might function upstream of OsWRKY53 in regulating rice CTB.

Discussion
In this study, we demonstrated that OsMKKK70 and its homologs might be novel negative regulators of CTB by fine-tuning GA levels in anther. First, OsMKKK70 expression is rapidly induced by cold treatment at the booting stage ( Figure 1A), and the OsMKKK70-OE plant is more sensitive to cold stress at the booting stage ( Figures 1B-G and S1). Second, osmkkk62/70 and osmkkk55/62/70 mutants have increased CTB due to higher pollen fertility and healthier anther morphology (Figures 2, 3 and S3). Third, the osmkkk62/70 mutant displays higher GA contents relative to WT at the LT condition (Figures 4 and 5). Fourth, OsMKKK70 negatively regulates GA biosynthesis, and then fine-tunes the GA contents during cold stress at the booting stage ( Figure 6). Taken together, these results demonstrate that OsMKKK70 and its homologs might negatively regulate CTB by modulating the GA content in anthers.  [25][26][27]. Our previous study demonstrated that OsMKKK70 has two homologs (OsMKKK62 and OsMKKK55) with high sequence similarity and these three OsMKKK genes function redundantly to regulate seed size and leaf angle in rice [6]. In this study, we discovered that osmkkk55, osmkkk62, and osmkkk70 mutants show no observable phenotype compared with WT in CTB ( Figure S2), but osmkkk62/70 double and osmkkk55/62/70 triple mutants display increased CTB (Figure 2, Figure 3 and Figure S3). Notably, unlike OsMKKK70-OE, OsMKKK55-OE and OsMKKK62-OE have a comparable seed setting rate to WT, and the expression of OsMKKK55 and OsMKKK62 was not obviously induced during the booting stage under the LT condition ( Figure S5A-D). In addition, we also discovered that OsMKKK62 and OsMKKK55 have higher levels in anther ( Figure S6A,B). These results indicated that OsMKKK70 might function redundantly with its homologs in regulating rice CTB, and OsMKKK70 plays the dominant role. However, it cannot be ruled out that OsMKKK70 and two homologs (OsMKKK62 and OsMKKK55) function in different pathways for CTB.

OsMKKK70-OsMKK4-OsMAPK6-OsWRKY53 Cascade Might Mediate a Trade-Off between Grain Size and CTB
Grain size and CTB are critical determinants of final grain yield. In our previous work, we found that overexpressing OsMKKK70 caused plants to produce larger seeds, the osmkkk62/70 double mutant and the osmkkk55/62/70 triple mutant displayed significantly smaller seeds [6]. Similarly, OsWRKY53 overexpression led to enlarged increased grain size, in contrast to the smaller seeds in the oswrky53 mutant [34]. We also demonstrated that OsMKKK70 regulates rice grain size through the OsMKK4-OsMAPK6-OsWRKY53 cascade signaling pathway [6]. Moreover, we also reported that the oswrky53 mutant shows an increased CTB phenotype via tuning the GA level in anthers [9]. In this study, we found that osmkkk62/70 and osmkkk55/62/70 mutants show an increased CTB phenotype (Figures 2A-C, 4A-D and S3). In addition, overexpressing CA-OsWRKY53 can rescue the small seed size of osmkkk62/70 [6]. This study shows that the seed setting, fertile pollen ratio, anther morphology, and GA biosynthesis genes expression level of the osmkkk62/70 CA-OsWRKY53 plant are very similar to those of CA-OsWRKY53 plants under NC (Figures 7A-D and S8). These results indicated that the OsMKKK70-OsMKK4-OsMAPK6-OsWRKY53 cascade might mediate a trade-off between grain size and CTB.
To support OsMKKK70 negatively regulating rice CTB by modulating the GA content in anther, we examined the expression level of GA biosynthesis genes in OsMKKK70-OE and WT plants under NC and after LT treatment for 4 days. The result showed that the expression levels of GA biosynthesis genes in OsMKKK70-OE are lower than those in WT under NC ( Figure 6B). After LT treatment, the expression levels of GA biosynthesis genes in OsMKKK70-OE show a greater degree of reduction, which are even lower than those in WT ( Figure 6B). Simultaneously, the level of OsSLR1 protein, GA-induced genes, and GA biosynthesis genes is comparable between the osmkkk70 mutant and WT under NC and after LT treatment, which might result from functional redundance ( Figure S9).
Notably, unlike oswrky53 showing no accompanying yield penalty, osmkkk62/70 and osmkkk55/62/70 mutants exhibit a decreased grain yield under NC ( Figure S10A,B). The grain number per panicle and 1000-grain weight of the osmkkk62/70 and osmkkk55/62/70 mutants are significantly lower than those of WT ( Figure S10A,B). In addition, osmkkk62/70 and osmkkk55/62/70 mutants show reduced plant height and panicles length compared with WT, which are also limiting factors for direct application ( Figure S10A,B). Although the application potential of osmkkk62/70 and osmkkk55/62/70 mutants is not as good as the oswrky53 mutant, the function characterization of OsMKKK70 in CTB provides a novel target for cold tolerance breeding in rice.

Plant Materials and Growth Conditions
Rice (Oryza sativa ssp. japonica) cultivar Longjing 11 (LJ11) was used to generate the OsMKKK70 transgenic plants and as the WT control. OsMAPK6 mutant dsg1 [35] and OsMKK4 mutant smg1-1 [36] were used to perform cold tolerance at booting stage (CTB) assays. Rice cultivars, SF43 and ZH11 (Oryza sativa ssp. japonica), were used as the wildtype control to compare with the corresponding mutants in CTB assays. The plants were cultivated in an experimental field (Heilongjiang Province, China) under natural long-day conditions or in a growth chamber at 30 • C for 14 h (light) and 24 • C for 10 h (dark).

Cold Tolerance at Booting Stage (CTB) Assays
The CTB assay was performed in a greenhouse as described previously with [9, 10,20] with some modifications. We selected germinated seedlings with a similar vigor of wildtype and mutant plants, cultivated them in pots containing a mixture of soil and water with fertilizer, and grew them in the greenhouse for normal growth (30 • C for 14 h (day) and 24 • C for 10 h (night)). For the point of low-temperature treatment, we established an internal reference with the distance between the auricles of the flag leaf and the penultimate leaf of the main panicles. When the distance between the auricles of the flag leaf and the penultimate leaf of the main panicle was within −5~0 cm, the main panicles were labeled and then transferred to a low-temperature greenhouse (15 • C constant temperature, 14 h light and 10 h dark photoperiod). After the indicated number of days of cold treatment, treated plants were returned to the greenhouse for normal growth until maturity, at which point we harvested and measured traits on labeled main panicles. The seed setting rate of labeled panicles were investigated for evaluating cold tolerance. The relative seed setting rate is the percentage of seed setting rate under low-temperature treatment relative to normal conditions. More than 10 plants per line were used for treatment in each biological replicate.
For RT-qPCR analysis of cold-regulated gene expression, and detecting OsSLR1 protein level, NC sampling was taken on the beginning day of low-temperature treatment, and LT sampling was taken on 4 days after low-temperature treatment described as above.
For measuring GA contents in cold-treated anthers, when the distance between the auricles of the flag leaf and the penultimate leaf of the main panicles was −8~−6 cm, LT samples were the plants transferred to the low-temperature greenhouse (18 • C, 14 h light and 10 h dark photoperiod) for continuous growing until around 15 d. NC samples were the plants growing under the normal condition until 1 d before flowering (around 11 d).

Microscopy
Starch staining of pollen grains and observations of anther morphology were performed 1 d before flowering. Anther morphology was observed directly under a microscope (Olympus, Shanghai, China, SZX16). For measuring pollen fertility, anthers were fixed in a formalin-acetic acid-alcohol (FAA) solution. Fixed anthers were ground to release pollen grains, stained with 1% I 2 -KI solution, then observed under a light microscope (Olympus, Shanghai, China, BX53). Blue-stained pollen grains were counted to determine pollen fertility.

Quantification of GAs
For measuring GA contents in cold-treated anthers, when the distance between the auricles of the flag leaf and the penultimate leaf of the main panicles was −8~−6 cm, LT samples were the plants transferred to the low-temperature greenhouse (18 • C, 14 h light and 10 h dark photoperiod) for continuous growing until around 15 d. NC samples were the plants growing under the normal condition until 1 d before flowering (around 11 d). The anthers of osmkkk62/70 and wild-type plants were harvested 1 d before flowering and stored at −80 • C until further use. The contents of endogenous GAs in anthers were measured by Wuhan Greensword Creation Technology Company (Wuhan, China) (http://www.greenswordcreation.com), referring to the method previously described [39]. Each series of experiments was performed in biological triplicates.

Total RNA Isolation and RT-qPCR Analysis
Total RNA was extracted from the samples using Trizol (Invitrogen, Carlsbad, CA, USA). Reverse transcription was carried out with SuperscriptII Reverse Transcriptase (Invitrogen, USA) using 1000 ng of total RNA. Real-time PCR was conducted with a Lightcycler 480 using SYBR Green PCR master mix (Takara, Kyoto, Japan). The rice ubiquitin gene was used as an internal reference to normalize all gene expression data. Three biological repeats were performed for each analysis. Values are means ± SE of three biological repeats. The primers used are listed in Table S1.

Protein Gel Blot Analysis
For analysis of the protein level of OsSLR1, the total proteins were extracted using the protein extraction buffer (25 mM Tris-HCl (pH 7.0), 10 mM MgCl2, 10 mM NaCl, and 5 mM DTT). The protein samples were separated by 8% SDS-PAGE and detected by immunoblot analysis with anti-OsSLR1 antibody [40] and anti-ACTIN antibody (Abmart, Shanghai, China; M20009M).

Statistical Analysis
Statistically significant differences were determined by Student's t-test and one-way ANOVA. The means and SE were based on independent biological samples.