Autophagy Mitigates High-Temperature Injury during Microsporogenesis 1 in Arabidopsis thaliana 2

13 Autophagy degrades cellular components during senescence, starvation, and stress. High14 temperature (HT) stress can inhibit microsporogenesis, but the involvement of autophagy in 15 HT injury is unknown. Here we show that Arabidopsis autophagy-defective (atg) mutants are 16 hypersensitive to HT stress during microsporogenesis but not during seedling growth. 17 Fertility was normal at 23 °C, but sporophytic male sterility occurred at 30 °C. At 30 °C, 18 wild-type developing anthers showed increased vacuolization in tapetum and lipophagy in 19 microspores. The atg5-1 mutant did not show these autophagic phenomena, but instead 20 showed irregularly enlarged vacuoles and subsequent shrinkage, and failure of the tapetum to 21 degenerate completely. HT specifically upregulated ATG8 in the developing anther, but not in 22 seedlings, and reduced MYB80 signaling in the anther, which is required for the regulation of 23 tapetal programmed cell death to promote microspore maturation. Interestingly, inhibition of 24 auxin activated the ATG8 signal in seedlings at both 23 and 30 °C. These results, combined 25 with our previous observation of anther-specific auxin depletion caused by HT, suggest that 26 autophagy mitigates HT injury to microsporogenesis by supporting tapetal degeneration and 27 micropore maturation in response to auxin reduction. 28


Introduction
Autophagy is a catabolic process that degrades cellular components within lysosomes and vacuoles, and is highly conserved among yeasts, plants, and mammals (Ohsumi 1999;Klionsky & Emr 2000;Yorimitsu & Klionsky 2005;Kim et al. 2007;Nakatogawa et al. 2009;Yoshimoto 2012). In plants, autophagy is activated during cell development, nutrient starvation, and senescence, and by environmental stressors such as pathogens, drought, salt, and oxidative damage (Bassham et al. 2006;Hofius et al. 2011;Yoshimoto 2012). In Arabidopsis and rice, a series of autophagy-related (ATG) genes have been identified, and many of the encoded essential amino acid residues are well conserved with those in yeasts (Doelling et al. 2002;Hanaoka et al. 2002;Chung et al. 2009;Xia et al. 2011;Yoshimoto 2012). Besides its role in cellular recycling, autophagy plays a role in cell death (Baehrecke 2005;Kroemer & Jäättelä 2005;Love et al. 2008;Parish & Li 2010). In plants, programmed cell death (PCD) occurs rapidly in the hypersensitivity response to pathogens, which requires prompt action to limit invasion, however autophagic cell death progresses relatively slowly and thus is observed in processes such as leaf senescence (Love et al. 2008;Parish & Li 2010;Wada et al. 2015).
Anther development, including specification of cell lineage and fate, follows well regulated programs. Among them, PCD is crucial in breaking down anther wall cells such as in the tapetum and middle layer during pollen grain maturation and anther dehiscence (Papini et al. 1999;Varnier et al. 2005). In Arabidopsis, rice, wheat, and cotton, the MYB80 transcription factor is required for the regulation of tapetal PCD through the expression of several genes, including UNDEAD (Phan et al. 2011(Phan et al. , 2012Xu et al. 2014). Autophagy plays a dispensable role in regular reproductive development, including tapetal PCD, in Arabidopsis, because almost all isolated autophagy-defective (atg) mutants can complete their life cycle under normal conditions (Yoshimoto 2012). Intriguingly, the occurrence of male sterility in a transgenic tobacco line overexpressing Arabidopsis AtATG6/BECLIN1 in the tapetum cells (Singh et al. 2010(Singh et al. , 2015 suggested that excessive autophagy might accelerate 3 tapetum cell death and lead to abortion of microsporogenesis in tobacco. Nevertheless, the rice autophagy null mutant OsATG7 showed a defect in tapetum cell degradation and sporophytic male sterility (Kurusu et al. 2014;Hanamata et al. 2014). These results strongly indicate that autophagy plays an important role in cell death during the degeneration of anther wall cells, but its biological significance and specific function in development are still unclear.
Early anther development is highly susceptible to several environmental stressors, in particular, high temperature (HT). We previously found that HT caused premature degradation of tapetum cells and abnormal vacuolization at 30 °C in barley and at >33 °C in Arabidopsis, and resulted in complete abortion of pollen development (Abiko et al. 2005;Oshino et al. 2007Oshino et al. , 2011Sakata et al. 2010). However, it is not clear whether elevated temperatures induce autophagy, and if so, how autophagy is involved in pollen and anther development under moderate HT stress. Here, we compared HT sensitivity during vegetative and reproductive stages between wild-type Columbia (Col-0; WT) plants and a series of atg mutants. We performed fine cytological analysis of anthers of WT and the atg5-1 mutant, which is a typical atg mutant used in research on ATG8-dependent autophagosome formation (Thompson et al. 2005;Nakamura et al. 2018). We also analyzed gene expression and protein production following moderate HT of 30 °C.
To test the effect of HT on plant growth from seedling to bolting stages, seeds were 4 directly sown into soil (1:1 mixture of Supermix A [Sakata] and vermiculite [Nittai]) and cultured at 23 or 30 °C in the growth chamber under the above conditions. To monitor the reproductive development and seed fertility of the primary inflorescence, 3-week-old plants (after initial anthesis) grown at 23 °C were transferred to 30 °C until the maturation of terminal siliques. To test recovery of pollen fertility, plants were held at 30 °C for 7 days, were then pollinated with pollen from plants of the same lines grown at 23 °C, and were kept at 30 °C.

Cytological analysis
To see HT-induced formation of YFP-ATG8 foci in anther wall cells and microspores, we observed developing anthers dissected from 3-week-old YFP-ATG8e-expressing plants grown at 23 °C and after 3 days' treatment at 30 °C under a laser scanning confocal microscope. The number of YFP-ATG8e foci per anther cell was quantified by IMARIS Microscopy Analysis Software (Bitplane).
For transmission electron microscopy, dissected stage 9 anthers of plants grown at 23 °C and exposed to 30 °C for 3 days were fixed in 2% paraformaldehyde and 2% glutaraldehyde in 0.05 M cacodylate buffer (pH 7.4) at 4 °C for 24 h. They were then washed 3 times in 0.05 M cacodylate buffer for 30 min each, and fixed in 2% osmium tetroxide at 4 °C for 4 h. They were dehydrated through a 70% to 100% alcohol series for 30 min each, rinsed twice for 1 h in propylene oxide, and embedded in resin (Quetol 651; Nisshin EM Co.) for 48 h at 60 °C.
Ultra-thin sections (80 nm) cut with a diamond knife on an ultramicrotome (Ultracut UCT, Leica) were mounted on copper grids, stained with 2% uranyl acetate at room temperature for 15 min, secondary-stained with lead stain solution (Sigma-Aldrich) at room temperature for 3 min, and then examined by transmission electron microscopy (JEM-1400 Plus, 80 kV; JEOL).
Transverse sections of stage 13 anthers and of pistils from 7-day HT-treated flowers at anthesis were prepared from embedded samples as above at a thickness of 1 µm and stained with 0.05% toluidine blue. Specimens were observed under a light microscope (BX51 Olympus) fitted with a CCD camera (DP73 Olympus). Dissected stage 13 anthers were also stained with iodine solution (Merck) to reveal pollen development and anther dehiscence.
After 7 h incubation in darkness with gentle shaking, the anthers were washed and observed under a microscope. The intensity of staining was quantified in ImageJ software with 4 different samples per treatment.

Semi-quantitative analyses of protein expression by MALDI-TOF MS
A total of 20 µg of crude extract from 50 stage 12 anthers was separated by SDS PAGE in a precast 5%-20% gradient gel (Oriental Inst HOG-0520). Three independent gels were stained with Coomassie Brilliant Blue (Sigma-Aldrich). The intensity of each band was calculated in ImageJ software, and the major proteins in each band were determined by matrix-assisted laser desorption/ionizationtime-of-flight mass spectrometry (MALDI-TOF MS; ABI5800, Applied Biosystems). Changes in relative levels of specific proteins by moderate HT treatment at 30 °C for 3 days was semi-quantified against specific peptide peak signals described in a previous protocol (Sasagawa et al. 2005): glutathione S-transferase / dehydroascorbate reductase 1 (DHAR1), L-ascorbate peroxidase 1 (APX1), and cytosolic triose-phosphate isomerase (CTIMC) were normalized to 60S ribosomal protein L7-2 (RPL7B) in MS spectrum data in the same gel slice from the 24-26-kDa protein band with the following peptides identified by MS/MS analysis: SHDGPFIAGER (theoretical molecular weight 1184.5) for DHAR1, ALLDDPVFRPLVEK (theoretical molecular weight 1610.9) for APX1, VIACVGETLEER (theoretical molecular weight 1374.6) for CTIMC, and ENFINELIR (theoretical molecular weight 1146.6) for RPL7B.

Statistical analyses
Each series of experiments was performed in triplicate. Statistics were calculated in MS Excel software. Statistical significance was assessed by unpaired Student's two-tailed t-test. Values were considered statistically significant at P < 0.05 or P < 0.01. One-way ANOVA followed by Tukey's post hoc test was performed for multi-sample study of silique length at P < 0.05.

Autophagy-defective mutants showed hypersensitivity to HT stress during reproductive stage and resulted in male sterility
In Arabidopsis thaliana, >33 °C HT causes male sterility owing to the abortion of pollen development (Sakata et al. 2010). To study the contribution of autophagy to HT injury, we measured silique length and seed set in 3-week-old plants (after initial anthesis) of WT and atg mutants moved from 23 °C to moderate HT of 30 °C. In WT, fully elongated siliques 7 from the 5th to the terminal blossoms were shortened to ½ to ⅓ of the length of those kept at 23 °C (Fig. 1A). In all atg mutants, they were reduced more severely to 1 ⁄6 of their length at 23 °C, which was normal (Fig. 1A). These shortened siliques in the atg mutants withered early. After the shift to 30 °C, the number of seeds per silique from the 5th to the terminal blossoms decreased to ¼ in WT and to almost nil in the atg mutants (Table 1). These results indicate that the 5th flower (flowering stage 10-11, Smyth et al. 1990;anther stage 9, Sanders et al. 1999) is a critical stage for susceptibility to elevated temperature, and that atg mutants became hypersensitive to HT stress during the reproductive stage.
We used crossing experiments to investigate whether the HT susceptibility of atg mutants was associated with the development of either pollen grains or ovules. When the pistils of atg plants that had developed at 30 °C (for 7 days) were pollinated with the pollen of the same lines that had developed at 23 °C, silique length was significantly recovered (Fig. 1B, C), indicating that HT affected pollen fertility but not ovules.
Since autophagy deficiency promotes the activation of SA signaling (Yoshimoto et al. 2009), we monitored the HT sensitivity of WT and atg5-1 plants carrying a bacterial SA hydroxylase, NahG. The results of Figure 2 and Table 1 show that NahG was unable to rescue the severe sterility of the atg5-1 mutant.
We also tested the effect of HT (at 30 and 35 °C) during the seedling stage on root elongation and plant growth. At 30 °C, root growth in both WT and atg mutants was retarded by about 20% (Supplementary Fig. 1). At 35 °C, it was retarded by ⅔ or more. Similarly, rosette leaf growth and expansion at 30 °C were suppressed in WT and atg5-1 ( Supplementary Fig. 2). In addition, bolting time was advanced by 1 week in both at 30 °C.
These results indicate that the degree of HT-induced obstacles did not change between WT and mutants from the seedling stage to the bolting stage. It suggests that autophagy suppresses HT injury to male reproductive development.

Deterioration of HT injury to anther and pollen development in atg mutants
At flowering stage, pistil morphology of the 5th to the terminal blossoms was normal in both WT and atg mutants, even after transfer to 30 °C (Fig. 3A). In contrast, anther size was reduced by approximately 20 to 40 % at 30 °C in all lines (Fig. 3B). The number of mature pollen grains was decreased in all anthers, severely so in atg anthers (Fig. 3B). Anther dehiscence still occurred in WT, even at 30 °C, and pollen grains were released, but anthers in all atg mutants barely dehisced (Fig. 3B).
These results indicate that autophagy is essential for septum and stomium breakage through degeneration of anther wall cells, including in the tapetum, and for pollen maturation at 30 °C.

Enhanced oxidative stress in anther of atg5-1 mutant
In general, reactive oxygen species (ROS) promote PCD, and spatiotemporal coordination of ROS production is essential for tapetal PCD progression and pollen development in rice and Arabidopsis (Hu et al. 2011;Xie et al. 2014). We monitored H2O2 levels in stage 12 anthers dissected from WT and atg5-1 plants held for 3 days at 30 °C. In WT, HT exposure significantly increased DAB staining in pollen grains and anther wall cells (Fig. 5).
Intriguingly, at 23 °C, staining was significantly higher in the atg5-1 anthers than in the WT, and at 30 °C, it further increased in the atg5-1 pollen grains (Fig. 5).
SDS-PAGE analysis showed that the intensity of the 24-26-kDa protein band normalized to 41-kDa actin significantly increased in atg5-1 stage 12 anthers held at 30 °C for 3 days (Fig. 6A, B). To identify which proteins increased, we analyzed the 24-26-kDa bands by MALDI-TOF MS. The major proteins detected were DHAR1, APX1, CTIMC, and 60S ribosomal protein L7-2 (RPL7B). Semi-quantitative MS peak analysis with normalization to RPL7B indicated that HT exposure significantly increased the level of DHAR1, which is involved in scavenging ROS (Dixon et al. 2002), in both WT and atg5-1 anthers, even though the basal level of DHAR1 at 23 °C was already higher in the mutant (Fig. 6C). Similarly, the basal level of CTIMC, a stress-inducible enzyme of the glycolysis and gluconeogenesis pathway (Sarry et al. 2006), was higher in the atg5-1 anthers and was induced by HT exposure, but to a lesser extent (Fig. 6C). In both WT and atg5-1, HT significantly induced APX1 (Fig. 6C). DHAR1 protein levels were consistent with DAB staining of H2O2 in WT and atg5-1 anthers in both control and HT groups.

Autophagy induced in anther wall cells and microspores by moderate HT
To study whether moderate HT induces autophagy in the developing anthers, we investigated the formation of ATG8 foci in YFP-ATG8e recombinant plants. GFP-ATG8 fusion proteins are used as reliable molecular markers of autophagy in yeast, mammalian, and Arabidopsis cells (Mizushima et al. 2004;Xie et al. 2008;Ishida et al. 2008;Nakayama et al. 2012;Merkulova et al. 2014). The number of YFP-ATG8e foci significantly increased in the anther wall cells and microspores of stage 9 anthers of plants grown at 30 °C for 3 days (Fig. 7). The foci measured 1-2 µm, comparable to the size of autophagic bodies marked with the GFP-ATG8 fusion protein in Arabidopsis Lin et al. 2015;Chung et al. 2010).
To further investigate how autophagy was promoted, we observed stage 9 anthers of WT and atg5-1 plants grown at 23 C or for 3 days at 30 °C by transmission electron microscopy.
At 23 °C, the tapetosomes and elaioplasts (components of the pollen coat) developed normally in tapetal walls of both WT and atg5-1 (Fig. 8). A few autophagosomes had the typical double membrane structure in the WT microspores at 23 °C (Fig. 8 inset). At 30 °C, WT tapetosomes developed abnormal vacuoles, and elaioplasts developed larger vacuoles with irregularly fused plastoglobuli (Fig. 8). In addition, the WT microspores showed an increase in autolysosomes, in which lipid bodies were trapped in the vacuoles and gained electron density (Fig. 8). Moreover, the number of autophagosomes increased in WT microspores and anther wall cells (Fig. 8). In contrast, larger vacuoles often appeared in the tapetum cells of atg5-1 grown at 23 °C (Fig. 8, blue arrows). At 30 °C, the microspores and tapetum cells were shrunken and had increased electron density, but neither vacuolization of elaioplasts and tapetosomes nor lipophage-like structures nor autophagosomes were observed in atg5-1 (Fig. 8).

HT repressed transcription of MYB80 and UNDEAD in developing anthers
The MYB80 transcription factor is required for the regulation of tapetal PCD; it upregulates UNDEAD, which encodes an A1 aspartic protease. It has been suggested that the AtMYB80/ UNDEAD system may regulate the timing of tapetal PCD, which is critical for viable pollen production (Phan et al. 2011(Phan et al. , 2012. To elucidate whether HT affects MYB80 transcriptional regulation in developing anther cells, we analyzed the expression of MYB80 and UNDEAD in stage 9 anthers of WT and atg5-1 plants. HT treatment at 30 °C for 3 days significantly reduced the expression of both MYB80 and UNDEAD in both WT and atg5-1 relative to 23 °C (Fig. 9). In particular, UNDEAD expression was much lower in atg5-1 than in WT at 30 °C, suggesting that normal tapetal PCD is considerably inhibited in atg5-1 under HT.

Discussion
Plant HT injury is one of the major obstacles to crop yield under warmer temperatures (Lobell and Field 2007). We analyzed the role of autophagy in the effect of HT stress on reproductive development of Arabidopsis. Although the roles of autophagy in senescence and in responses to nutrient starvation, pathogens, drought, salt, and oxidative stresses are well known (Doelling et al. 2002;Hanaoka et al. 2002;Bassham et al. 2006;Chung et al. 2009;Hofius et al. 2011;Yoshimoto 2012), there are few reports of its role in HT stress. Zhou et al. (2014) reported that silencing of ATG-related genes reduced the tolerance of tomato to severe heat stress at 45 °C. Although HT stress affects the whole plant, microsporogenesis is one of the most sensitive processes (Sakata & Higashitani 2008;Müller & Rieu 2016).
We previously showed that HT increases vacuolization and the development of autolysosome-or autophagosome-like structures in developing anther cells of barley (Oshino et al. 2007(Oshino et al. , 2011. Singh et al. (2010Singh et al. ( , 2015 showed that overexpression of the Arabidopsis tapetum-cell-specific autophagy-related ATG6/BECLIN1 in tobacco leads to premature degeneration of the developing tapetum cells. These reports suggest that disrupting autophagy would prevent the premature degradation of tapetum cells and improve HT tolerance. Instead, our results show that autophagy is essential for mitigating HT injury during pollen development in Arabidopsis: atg mutants became almost completely male-sterile at moderate HT of 30 °C, whereas WT plants still produced some fertile seeds (Figs. 1, 2; Table 1). In all atg mutants tested, 30 °C inhibited anther dehiscence and pollen development, leading to male sterility (Figs. 2, 3). Transverse section analysis of atg5-1 anthers clearly revealed the inhibition of anther septum and stomium breakage and shrunken abnormal tapetum (Fig. 4).
Anther development and differentiation, including specification of cell lineage and cell fate, are well regulated programs. The epidermis, endothecium, middle layer, and tapetum of anther wall cells are sequentially degraded during pollen maturation and anther dehiscence.
The proper timing of tapetal degradation is necessary for the production of viable pollen, and occurs via developmentally regulated PCD (Papini et al. 1999;Varnier et al. 2005;Parish & Li 2010). MYB80 is a key transcription factor that controls tapetal PCD by regulating at least 70 genes, including directly upregulating UNDEAD (Phan et al. 2011(Phan et al. , 2012Xu et al. 2014).
ATG8e foci increased in tapetum cells and microspores at 30 °C (Fig. 7). Size and number of vacuoles increased in the cytoplasm, elaioplasts, and tapetosomes, and autophagosomes, autolysosomes, and lipophage-like structures increased in microspores and anther wall cells (Fig. 8). In mammalian cells, the breakdown of lipid droplets by lipophagy contributes to energy generation under stress (Dong & Czaja 2011). In the atg5-1 mutant, these autophagic phenomena did not appear at 30 °C. Instead, abnormal large vacuolization and subsequent shrinkage (but not complete degeneration) of tapetum cells were observed, and the collapse of microspores resulted in complete male sterility (Figs. 4, 8). Together, these results indicate that HT stress alters the timing of tapetal PCD through repression of MYB80 transcriptional regulation, but it activates autophagy to compensate for the alteration of MYB80-regulated PCD by autophagic cell death for rescuing microspore development under HT.
A rice autophagy-defective mutant, the OsATG7 Tos 17 insertion line, is male-sterile owing to the inhibition of tapetum-cell degradation (Kurusu et al. 2014;Hanamata et al. 2014). Autophagy is involved in the breakdown of the lipid bodies and of lipids transferred from tapetum cells to the microspore surface in rice (Kurusu et al. 2014;Hanamata et al. 2014). However, Arabidopsis atg mutants completed their life cycle at normal temperature (Yoshimoto 2012;Figs. 1-4, Supplementary Figs. 1, 2), and thus rice and Arabidopsis differ in their need for autophagy (Kurusu et al. 2014;Hanamata et al. 2014). In brassicaceous species, including Arabidopsis, a distinctive organelle, the tapetosome, develops in the tapetum cells from the endoplasmic reticulum and stores triacylglycerols, flavonoids, alkanes, and pollen coat proteins (Wu et al. 1997;Suzuki et al. 2013;Hanamata et al. 2014). Another tapetal organelle, the elaioplast, stores a major component of the pollen coat: steryl esters inside numerous plastoglobuli (Wu et al. 1997;Suzuki et al. 2013;Hanamata et al. 2014).
Under HT, where tapetal PCD activity decreased, we found lots of vacuoles in tapetosomes and elaioplasts in WT anther wall cells, and lipophage-like bodies trapped in the autophagosomes and vacuoles in microspores (Fig. 8), similar to previous observations in rice (Kurusu et al. 2014;Hanamata et al. 2014), but not in atg5-1 cells (Fig. 8). These results strongly suggest that degradation of elaioplasts and tapetosomes promoted by autophagy leads to degeneration of tapetum cells and degrades lipid bodies in microspores to prevent abortion of microspores by inhibiting PCD at elevated temperatures.
In addition to the hypothesis that autophagic cell death can compensate for the decrease of tapetal PCD by HT, the following hypothesis is also conceivable. Here we found that HT induces ROS in developing anthers of both atg5-1 and WT plants (Fig. 5). Both APX1 and DHAR1 levels were significantly induced following HT exposure, DHAR1 greatly so (Fig.   6C). Recently, a similar pattern of APX1 and DHAR1 upregulation was reported in saltstressed Arabidopsis roots (Szymanska et al. 2019). Moreover, even under steady-state conditions, H2O2 and DHAR1 levels were higher in developing anthers of atg5-1 than in WT (Figs. 5,6). Likewise, higher ROS in atg2-1 and atg5-1 leaves was reported (Yoshimoto et al. 2009). In addition, it has recently been shown that excess ROS generated in a cytoplasmicmale-sterile wheat line delayed tapetal PCD initiation and led to pollen abortion (Liu et al. 12 2018), although the spatiotemporal regulation of ROS production is essential for tapetal PCD progression (Hu et al. 2011;Xie et al. 2014). Altogether, in addition to HT-induced ROS, over-accumulation of ROS in autophagy deficiency might strongly affect the spatiotemporal ROS signaling for tapetal PCD. In either hypothesis, the proper timing of pollen development, including tapetal PCD, is strictly controlled, and autophagy is important to minimize the effects of HT and to maintain this pollen-specific development.

Conclusion
In Arabidopsis, autophagy is not essential for completion of the life cycle under normal temperatures. Seedling growth of autophagy-deficient mutants was unaffected by HT.
However, pollen development was significantly damaged by HT: HT stress induced autophagy in developing anther wall cells and microspores. It caused oxidative damage and altered the tapetal PCD, but activated autophagy mitigated the HT injury.    and atg5-1 plants (lower panels) without (left) or with (right) ectopic NahG to suppress salicylic acid signaling. At 23 °C (blue borders), growth was normal. At 30 °C (orange borders), silique length was drastically reduced from the 5th flower. Scale bars, 1 mm.        atg10-1 24.9 ± 6.8 0.6 ± 1.9
b Seeds from 5 th to the terminal blossoms after temperature shift up.