Arabidopsis (Arabidopsis thaliana) Columbia-0 and Landsberg erecta (Ler) accessions were obtained from the Nottingham Arabidopsis Stock Centre. The GA-insensitive mutant gai was kindly shared by Patrick Achard. The DELLA double mutant rga-24 gai-t6 was previously described (Achard et al., 2006) and were kindly provided by Patrick Achard and Nicholas Harberd. The fluorescent tagged lines (FTLs) in the qrt1-2^-/- background (FTL1313 and FTL3332), used for genotyping unreduced gametes, were described earlier (Francis et al., 2007; Berchowitz and Copenhaver, 2008). FTL1313 marker is physically located at 11.2 cM on chromosome 1, and FTL3332 is positioned at 10.43 cM on chromosome 3¹ (Supplementary Figure S1A) (Francis et al., 2007; Singh et al., 2017). The pRGA::GFP-RGA transgenic plants were obtained from Nicholas Harberd. Primers used for mutant genotyping are listed in Supplementary Table S2.

Seeds were germinated on K1 medium for 6–8 days and seedlings were transferred to soil and cultivated in growth chambers at 12 h day/12 h night, 20°C, and less than 70% humidity. To stimulate flowering transition, the photoperiod was changed to a 16-h-day/8-h-night regime. For GA₃ and PAC treatment, flowering plants were sprayed by water (+0.02% Tween), 100 μM GA₃ (+0.02% Tween) and 1 mM PAC (+0.02% Tween), respectively. The chemical concentrations used in this study were chosen based on previous study (Liu et al., 2017b).

Young flowering Arabidopsis plants were treated with cold (4°C–5°C for 48 h) and the unicellular stage microspores were examined at 24–36 h after the cold treatment under microscope. The cold sensitivity of the rga-24 gai-t6 mutant was analyzed in the qrt1-2^-/- background. The microspores were released by squashing targeted stage buds on a microscope slide with a drop of orcein staining buffer. The assessment of unreduced microspores was done by comparing the size to the haploid microspores and/or by counting the number of nucleus. The cold sensitivity of sporogenesis was evaluated by quantifying the frequency of enlarged unicellular stage microspores among the haploid microspores in a same flower bud. For each plant individual, more than 500 meiotic products were counted for quantification; and if the total number could not reach 500, then all the meiotic products were counted. More than five biological replicates have been performed, and Student’s t-test was used for significance analysis at the 5% significance level (α = 0.05).

To determine the effect of cold on abundance of GFP-tagged DELLA RGA proteins, Image J software was used for the quantification of GFP intensity. First, Image J was set up by selecting ‘SET MEASUREMENTS’ in ANALYZE menu with AREA, INTEGRATED DENSITY and MEAN GRAY VALUE being selected. Second, using any of the drawing/selection tools in Image J, the entire anther area of a tetrad stage anther picture taken from a pRGA::GFP-RGA plant without cold treatment, was chosen for fluorescence quantification by clicking ‘MEASURE.’ The value was recorded as integrated density (ID1). Thirdly, a region in the same anther where developing meiocytes were located was chosen as background and the fluorescence intensity was quantified and recorded as mean fluorescence of background readings (FoBR1). Afterward, integrated density in pictures of another four anthers from four non-treated pRGA::GFP-RGA plants, and five anthers from five cold-treated (4°C–5°C for 24 h) pRGA::GFP-RGA plants was quantified (values are recorded as ID2-10, respectively). The mean background readings were recorded as FoBR2-10, respectively. When all samples were finished, the corrected total cell fluorescence (CTCF) for each sample was calculated using formula: CTCFn = IDn- (area of selected cell × FoBRn). Student’s t-test was used for significance analysis with significance level (alpha) = 0.05.

Callosic cell wall staining and the analysis of the male meiotic products (tetrad-stage analysis by aniline blue and orcein staining) were performed as described previously (Liu et al., 2017b). Flowering Arabidopsis qrt1-2^-/- plants segregating for the FTL1313 or FTL3332 pollen fluorescent marker were sprayed with 100 μM GA₃, and the mature pollen grains were observed at 7–9 days following GA treatment. FDR/SDR genotyping of mature pollen grains in the FTL lines (FTL1313 and FTL3332) was performed by releasing mature pollen grains in a drop of pollen extraction buffer (0.5 M EDTA) on a microscope slide, and visualized under fluorescence microscope. Five plant individuals were analyzed for either the FTL1313 or FTL3332 reporter. Meiotic spreads were prepared as described previously (Liu et al., 2017b).

The alpha-tubulin immunolocalization was performed according to the method of De Storme et al. (2012) with minor modifications. Briefly, the time of the first and second digestions by enzyme mix were adjusted to 3 and 1.5 h, respectively.

The young flowering Arabidopsis wild type Columbia-0 plants were treated with cold (4°C–5°C) for 0, 2, and 24 h, respectively, and the total RNA from young flower buds was isolated using the RNeasy Plant Mini Kit with additional on-column DNase I treatment (Qiagen). First-strand cDNA was synthesized using the GoScript^TM Reversion Transcription System Kit (Promega). Quantitative gene expression analysis was performed by qRT-PCR on a Stratagene MX3000 real-time PCR system using the GoTaq^® qPCR Master Mix (Promega). For each treatment group, young flower buds from ten plant individuals were collected, and mRNA was isolated and pooled for further use. Two more bulks of flower buds (also from ten plants for each bulk) were sampled, for a total of three biological replicates for each treatment. For each bulked mRNA sample, qPCR was performed twice, with three technical replicates for each gene being surveyed (six technical replicates in total). Data from the six assays were pooled and then analyzed. The Arabidopsis ACTIN2 gene (AT3G18780) was used as the reference gene. The data for each tested gene have been normalized to the value of internal control AtACTIN2, and the value of 2 and 24 h cold-treated samples were compared with non-cold treated samples (0 h cold treatment). The relative expression fold-change is presented and was calculated using 2ˆdelta-delta Ct method. Significance analysis was performed on the relative expression fold-change using Wilcoxon rank test. Significance level (alpha) was 0.05. Primers used for specific amplification of targeted gene transcripts are listed in Supplementary Table S3.

The microscopy performed in this study was according to the previous report (Liu et al., 2017b).

In Arabidopsis, recombination hotspots are distributed along entire chromatin except for centromeric regions (Salomé et al., 2011). To determine the type of GA-induced meiotic restitution in Arabidopsis, we analyzed segregation of the hemizygous centromere-linked FTL markers FTL1313 (dsRed – Chr. 1) and FTL3332 (YFP – Chr. 3) in the quartet1-2^-/- (qrt1-2^-/-) background. Because we have previously shown that GA treatment may induce around 5% meiotic restitution (Liu et al., 2017b), we here only quantified the number of meiotic restituted products and classified them into either FDR- or SDR-type (Supplementary Table S1). Under control conditions, plants hemizygous for either the FTL1313 or FTL3332 reporter construct produced tetrads in which two out the four pollen grains were GFP fluorescent reflecting regular segregation of these fluorescent markers (Supplementary Figure S1C). At 7–9 days post GA treatment, GA-sprayed plants hemizygous for FTL1313 produced abundant tetrads and, in addition, a small amount of triads with two regular-sized haploid pollen grains and one larger, unreduced pollen grain. When fluorescent expression in these pollen triads was exclusively confined to the unreduced pollen grain or to both haploid pollen grains, the triads most likely originate from SDR-type restitution (Supplementary Figures S1D,E). In contrast, when fluorescent expression in the triads occurred in both a haploid and an unreduced pollen grain, the triads most likely originated from FDR-type restitution (Supplementary Figure S1F). In addition, GA-treated plants also produced dyads in which homologous chromosomes harboring centromere-linked hemizygous FTL markers either segregate both to one single unreduced pollen (Supplementary Figure S1G) or each to one single unreduced pollen (Supplementary Figure S1H), indicating for either SDR- or FDR-type meiotic restitution, respectively. GA-induced dyads and triads in qrt1-2^-/- plants hemizygous for the FTL1313 reporter were found to contain 75.4% homozygous and 24.6% hemizygous for the locus containing the FTL1313 reporter. For GA-treated qrt1-2^-/- plants hemizygous for the FTL3332 marker, 92.1% appeared homozygous and 7.9% appeared heterozygous for the FTL3332 locus (Supplementary Figure S1B,I–L). The predominant homozygous status of unreduced microspores of both the FTL1313 and FTL3332 marker lines indicated that GA primarily induced SDR-type meiotic restitution. These results are similar to what has been reported for cold stress-induced male meiotic restitution (De Storme et al., 2012).

To test whether cold induces meiotic restitution by increasing endogenous GA level in anthers, flowering Arabidopsis plants were sprayed either with 100 μM GA₃ or 1 mM of the GA biosynthesis inhibitor PAC, after which they immediately were transferred to cold conditions (4–5°C) for 48 h. 24–36 h post cold treatment, unicellular stage microspores were examined for indirect quantification of male meiotic restitution (Figure 1). Under normal temperature conditions, both mock and PAC-treated plants produced 100% uniformly sized haploid microspores (Figure 1B, mock-treated; C, PAC-treated). In contrast, GA₃-sprayed qrt1-2^-/- plants produced around 3.8% enlarged microspores (Figures 1A,D,E). When an additional cold treatment was applied, mock, GA₃ or PAC sprayed plants displayed similar frequency of enlarged unicellular microspores (Figures 1A,F–K). These data indicate that cold-induced meiotic restitution is neither enhanced nor reduced by exogenous GA and PAC application, suggesting that GA homeostasis is not critical for evoking a cold response in male meiosis.

The impact of endogenous genetic alterations in GA signaling on the cold sensitivity of male sporogenesis was investigated by testing the cold response of both dominant and loss of function della mutant plants (Figure 2). The GA dominant insensitive gai mutant produces a non-degradable DELLA GAI protein and thus exhibits a constitutively repressed DELLA-dependent GA signaling (Peng et al., 1997). Under normal temperature conditions, the gai mutant produced normal tetrads and haploid microspores (Supplementary Figures S2C,D) as wild type plants (Supplementary Figures S2A,B). Upon exposure to cold (4–5°C for 48 h), gai plants exhibited male meiotic restitution with associated unreduced microspore formation at a similar level observed in wild type Ler plants (Figures 2A,E,F), indicating that impaired DELLA-dependent GA signaling does not block the cold sensitivity of Arabidopsis male meiotic cell division. The double rga-24 gai-t6 mutant plants exhibit constitutively activated GA signaling (Dill and Sun, 2001), and produced 4.8% unreduced male gametes under normal temperature conditions (Figure 2B). Cold was found to induce a significantly higher level of enlarged microspores compared to wild type plants (Figures 2B–D, wild type Ler; G and H, rga-24 gai-t6), suggesting an additive effect of cold stress on meiotic restitution in the della null mutation background. These data demonstrate that the cold sensitivity and response of Arabidopsis male sporogenesis does not rely on RGA- and GAI-dependent GA signaling.

We further examined male meiotic chromosome behavior and male meiotic cell wall formation in the GA-insensitive gai meiocytes. Cold stress did not disturb male meiotic chromosome segregation in both the gai mutant and wild type Ler plants (Supplementary Figures S3A–C). Under normal conditions, gai displayed regular callosic cell wall formation indicating normal meiotic cytokinesis (Supplementary Figures S3D,G). However, at 24 h after cold treatment (4–5°C), meiotic-restituted dyads and triads with incomplete callosic cell walls were observed in both Ler and the gai mutant (Supplementary Figure S3E,F,H,I) manifesting defective male meiotic cytokinesis.

Tubulin immunostaining was performed on the cold-shocked Ler and gai mutant male meiocytes (Figure 3). During the stages from prophase I to anaphase II, both the control and cold-stressed meiocytes in either Ler or gai plants displayed regular microtubule configurations (Figures 3A–E,G–K,M–Q,S–W). At prophase I, a network of microtubules surrounded the nucleus. Metaphase I showed formation of a single spindle, while metaphase II showed two perpendicular spindle formations. In contrast, the microtubule network; i.e., RMA, was clearly affected by cold at telophase II. Some of the nuclei in cold-stressed Ler and gai tetrad were adjacent to each other and were not separated by RMA (Figures 3R,X and Supplementary Figure S4). These data demonstrate that the cold-sensitivity of meiotic cytokinesis and RMA in male meiocytes does not rely on DELLA-dependent GA signaling.

To determine the effect of cold stress on abundance of DELLA proteins in developing Arabidopsis anthers, we used an Arabidopsis transgenic line harboring a recombinant pRGA::GFP-RGA reporter construct and monitored the in vivo GFP fluorescence signals in the developing anthers at 24 h under cold shock (4–5°C for 24 h). The abundance of RGA proteins was determined by quantifying the GFP intensity in the anthers (Figure 4A). If cold stress induces defective meiotic cytokinesis by promoting DELLA degradation, the fluorescence signals of GFP-RGA in tetrad stage anthers should be reduced. We observed that the pRGA::GFP-RGA was predominantly expressed in the tapetal cell layer and was not affected by 24 h cold stress (Figures 4B,C), suggesting that cold induces meiotic restitution in Arabidopsis anthers not by interfering with DELLA protein stability.

To determine the effect of cold stress on the expression of GA metabolic and signaling genes in young Arabidopsis flower buds, real-time quantitative PCR of the genes encoding for CBF1, GA biosynthesis GA3OX1, GA 2-oxidases GA2OX2 and GA2OX6, DELLA RGA and GAI proteins was performed (Figure 5). In 2 h cold-stressed young flower buds, the CBF1 and GA2OX6 transcript levels showed a significant increase and reduction compared to control plants without cold stress, respectively (Figures 5A,D), while for GA3OX1 and RGA, no significant alterations were detected (Figures 5B,E). The transcripts of GA2OX2 and GAI appeared stable throughout the cold treatment (Figures 5C,F). At 24 h under cold stress, the relative expression of CBF1 and GA3OX1 declined significantly (Figures 4A,B), contrary to RGA that showed an elevated expression level (Figure 5E). These data suggest that cold stress has an impact on the transcription of both GA synthesis and catabolic genes, and it modulates GA levels in Arabidopsis anthers in a complex manner.