ABA Regulates Apoplastic Sugar Transport and is a Potential Signal for Cold-Induced Pollen Sterility in Rice

Abstract

Cold temperatures cause pollen sterility and large reductions in grain yield in temperate rice growing regions of the world. Induction of pollen sterility by cold involves a disruption of sugar transport in anthers, caused by the cold-induced repression of the apoplastic sugar transport pathway in the tapetum. Here we demonstrate that the phytohormone ABA is a potential signal for cold-induced pollen sterility (CIPS). Cold treatment of the cold-sensitive cultivar Doongara resulted in increased anther ABA levels. Exogenous ABA treatment at the young microspore stage induced pollen sterility and affected cell wall invertase and monosaccharide transporter gene expression in a way similar to cold treatment. In the cold-tolerant cultivar R31, ABA levels were significantly lower under normal circumstances and remained low after cold treatment. The differences in endogenous ABA levels in Doongara and R31 correlated with differences in expression of the ABA biosynthetic genes encoding zeaxanthin epoxidase (OSZEP1) and 9-cis-epoxycarotenoid dioxygenase (OSNCED2, OSNCED3) in anthers. The expression of three ABA-8-hydroxylase genes (ABA8OX1, 2 and 3) in R31 anthers was higher under control conditions and was regulated differently by cold compared with Doongara. Our results indicate that the cold tolerance phenotype of R31 is correlated with lower endogenous ABA levels and a different regulation of ABA metabolism.

Abscisic acid — Cold — Gene expression — Pollen sterility — Rice — Tapetum

Introduction

Cool temperatures adversely affect rice crops in temperate regions of the world. At the beginning of the season, chilling affects germination and seedling vigor, delaying heading and grain development (Andaya and Mackill 2003). Cold spells during the reproductive stage at the end of the growing season are especially damaging, causing pollen sterility and massive losses in grain yield. In Australia, cold snaps during the reproductive stage can lead to 20–40% yield losses (Angus and Lewin 1991, Jacobs and Pearson 1994, Jacobs and Pearson 1999).

Reproductive development in plants is very sensitive to a variety of abiotic stresses. Morphological investigations and cross-pollination experiments demonstrated that for stresses such as drought and cold, the most vulnerable stage is the young microspore (YM) stage of male gametophyte development (Bingham 1966, Hayase et al. 1969, Satake and Hayase 1970, Nishiyama 1984, Saini et al. 1984, Patterson et al. 1987). Stress sensitivity of the ovule occurs later in reproductive development during anthesis and the early stages of grain filling (Westgate and Boyer 1985). In the cold-sensitive Australian commercial rice cultivar Doongara, a high level of sterility is induced by cold temperatures applied specifically at the YM stage or the transition of the tetrad to the early uninucleate stage of pollen development. At this stage, the tapetum is functioning at maximal capacity; the pollen cell wall is deposited and the locular fluid is synthesized (Oliver et al. 2005). Our previous studies have indicated that cold treatment of Doongara at the YM stage leads to increased anther sucrose and hexose levels, followed by lack of starch in the pollen grains at anthesis. When the Chinese cold-tolerant cultivar R31 is subjected to the same temperature regime, sucrose accumulation does not occur and the pollen remains fertile (Oliver et al. 2005). The activity of cell wall-bound acid invertase, an enzyme that plays a critical role in the transport pathway of sucrose to sink tissues (Roitsch 1999, Sturm 1999, Sturm and Tang 1999), is repressed in cold-stressed Doongara anthers, but was maintained at high levels in R31 anthers. Similarly, expression of the cell wall invertase gene OSINV4 (Os04g33720), which is expressed specifically in the tapetum at the YM stage and in the pollen grains from the early bicellular (EB) stage onwards, is repressed by cold in Doongara anthers, but high levels of OSINV4 expression are maintained in R31 anthers throughout the cold treatment (Oliver et al. 2005). These studies indicate that CIPS is associated with the down-regulation of a key component of apoplastic sugar transport in anthers.

To extend our investigation of the effect of cold treatment on other components of the apoplastic sugar transport pathway in anthers, we have identified two rice anther monosaccharide transporter genes, OSMST7 (Os01g38680) and OSMST8 (Os01g38670; Oliver et al. 2007). These tandem duplicated genes on chromosome 1 have 92% identical DNA sequences but show different temporal and spatial expression features. OSMST7 is expressed predominantly later in pollen development (starting at the YM stage and peaking between the vacuolated uninucleate stage and the EB stage). OSMST7 may be expressed in the pollen or the anther wall, but not the tapetum, which starts to degrade from the vacuolated uninucleate stage onwards. OSMST8 expression is similar to that of OSINV4, with expression in the tapetum at the YM stage and high expression from the vacuolated microspore stage onwards until maturity (Oliver et al. 2007). Here, we describe the effect of cold on the expression of these genes, to determine whether cold could act to down-regulate the entire apoplastic sugar transport pathway in anthers.

We also investigate the role of ABA as a signal for cold-induced pollen sterility (CIPS). ABA plays an important regulatory role in a variety of abiotic stress responses in plants, including drought and cold stress (Tomashow 1999, Shinozaki and Yamaguchi-Shinozaki 2000, Shinozaki et al. 2003, Chinnusamy et al. 2004). ABA accumulates in drought-stressed wheat spikelets (Morgan 1980, Saini and Aspinall 1982, Morgan and King 1984, Westgate et al. 1996). ABA is also part of a regulatory network that integrates sugar and light signaling (Finkelstein and Gibson 2002, Brocard-Gifford et al. 2003). Our results indicate that ABA levels increase in cold-stressed rice anthers and that ABA treatment mimics the effect of cold. In addition, the levels of ABA biosynthesis and degradation differ between cold-tolerant and cold-sensitive rice, suggesting that there is a strong association between ABA metabolism and the cold tolerance phenotype.

Results

Monosaccharide transporter genes are expressed differently in Doongara and R31

The effect of cold on the expression of the monosaccharide transporter genes OSMST8 and OSMST7 was analyzed at two developmental stages: the cold-sensitive YM stage and the EB stage, which is relatively insensitive to cold. In Doongara, OSMST8 expression at the YM stage was low and was repressed on average 5-fold by cold treatment (P < 0.05, n = 5; Fig. 1a). Doongara OSMST8 expression at the EB stage was much higher than at the YM stage, and was repressed 2-fold by cold (P < 0.05, n = 4; Fig. 1a). OSMST8 expression at the YM stage was higher in R31 than in Doongara, and the gene was not as strongly affected by cold (Fig. 1a). At the EB stage, OSMST8 expression in R31 was about 2-fold lower than in Doongara, and there was no reduction by cold (Fig. 1a).

Fig. 1

The effect of cold (CD) on the expression of the OSMST8 (a) and OSMST7 (b) genes in anthers from the cold-sensitive cultivar Doongara and the cold-tolerant cultivar R31. α-Tubulin was used as a loading control. Gene expression results are shown for young microspore stage anthers of control (YM C) and cold-treated plants (YM CD), and early bicellular stage anthers of control (EB C) and cold-treated plants (EB CD). Southern blot detection of the RT–PCR gel was used to detect OSMST8 expression at the young microspore stage due to the low expression of this gene at this stage of pollen development (middle panel in a).

OSMST7 expression in Doongara was induced by cold, and induction was stronger at the YM stage (3.8-fold; P < 0.05; n = 4; Fig. 1b) than at the EB stage (2.6-fold; P < 0.05; n = 5; Fig. 1b). In R31, expression of OSMST7 at the YM stage was 20-fold higher than in Doongara, and there was no significant effect of cold treatment (Fig. 1b; n = 4). At the EB stage, OSMST7 expression in R31 was 4.5-fold higher than in Doongara and the gene was repressed 2-fold by cold (Fig. 1b; n = 3).

These results further establish that anther sugar metabolism and sugar transport to the tapetum are affected by cold. In Doongara, cold represses apoplastic sugar transport pathway genes that are active in the tapetum (OSINV4 and OSMST8) and induces OSMST7 which may be present in the pollen or anther wall. All three genes show a different response to cold treatment in the cold-tolerant cultivar R31.

Cold treatment causes ABA accumulation in rice anthers

In Doongara panicles, a 3 d cold treatment at the YM stage caused a 34% increase in ABA levels (P < 0.05, Fig. 2a). In YM anthers, ABA levels increased by 36% after 5 h of chilling (P < 0.05) and remained 33% higher than untreated controls after 24 and 72 h of cold treatment (P < 0.05, Fig. 2b). ABA levels did not change significantly in cold-treated EB anthers (Fig. 2b). The ABA concentration in untreated YM anthers (average: 1,035 ± 20 ng g⁻¹ DW) was 2.4 times greater than at the EB stage (average: 438 ± 51 ng g⁻¹ DW).

Fig. 2

Measurement of ABA levels in cold-treated rice plants. (a) ABA concentration in young microspore stage panicles of control (YM C) and cold-treated (3 d at 12°C; YM CD) Doongara plants. The graph shows the average of two biological replicate experiments. Within each experiment, between two and four panicles per treatment were measured. Error bars represent standard errors. dw = dry weight. (b) ABA concentration in anthers of Doongara plants cold treated at the young microspore (YM) and early bicellular (EB) stage for 0 h (C), 5, 24 and 72 h (12°C). The results shown are the average of two biological repeat experiments. Within each experiment, the means of each treatment were calculated as the average of three or four samples. Each sample consisted of pooled anthers from at least two plants. Error bars represent standard errors. (c) ABA concentration in leaves of Doongara plants with anthers at the young microspore (YM) and early bicellular (EB) stage. Cold treatment was for 0 (C), 5, 24 and 72 h at 12°C. The graph shows the average leaf ABA concentrations measured using two different batches of plants for the YM treatments, one of which was also used for the EB treatments. From each batch, three or four plants per treatment were measured. Error bars represent standard errors. (d) Comparison of ABA levels in anthers (D1) and leaves of Doongara and R31 plants. ABA levels were determined at the young microspore stage of pollen development under control conditions (YM C) and cold conditions (YM CD). The cold treatment (12°C) was for 3 d in anthers (D1) and 24 h in leaves (D2). The anther data show the average of two biological repeat experiments. Within each experiment, three or four samples per treatment were measured, where each sample consisted of pooled anthers from at least two plants, and the leaf data show the average of four plants per treatment. Error bars represent standard errors.

In untreated leaves, ABA levels were about half those in untreated anthers (607 ng g⁻¹ DW; Fig. 2c) and were not significantly altered after 5 h of chilling. In leaves of plants at YM and EB stages, ABA levels did not rise until 24 h of cold treatment and the increase was greater in leaves (3.9-fold; P < 0.01) than in anthers (1.4-fold). Leaf ABA levels remained high in plants chilled for 72 h (P < 0.05; Fig. 2c). ABA increased in leaves regardless of pollen developmental stage, but an increase in anthers was only observed at the YM stage, i.e. the stage of greatest cold sensitivity. This observation, together with the fact that ABA accumulates earlier in anthers than in leaves, suggests that ABA is not transported to the anthers from the leaves or other plant tissues, but rather is synthesized in the anther.

In YM anthers of the cold-tolerant cultivar R31, ABA levels under normal conditions were on average 30% lower than in Doongara anthers (P < 0.01), and increased after 3 d of cold treatment to approximately the same level as that found in untreated Doongara anthers (Fig. 2d1). Similar results were obtained after a 24 h cold treatment, with 644 ± 30 ng g⁻¹ FW under control conditions and 758 ± 89 ng g⁻¹ FW after cold treatment. Thus, in R31, anther ABA levels are significantly lower than in Doongara anthers and remain low after cold treatment. R31 leaves also showed a different response to cold: leaf ABA levels were significantly lower under both control and cold conditions compared with Doongara leaves (P < 0.05), and cold treatment even led to a small but significant decrease in R31 leaves (P < 0.05, Fig. 2d2).

ABA treatment mimics the effect of cold

The highest level of spikelet sterility in Doongara plants (as measured by the number of unfilled grains) occurs when cold is applied at the YM stage (Fig. 3a). Considerable amounts of spikelet sterility are also observed before and after the YM stage; this is due to variability in the auricle distance (AD) measurement system and the fact that rice panicles flower from top to bottom (the AD scale was calibrated to score the top florets). Application of ABA (10⁻⁴ M) also caused spikelet sterility (Fig. 3a). Spikelet sterility levels following ABA treatment were much higher than following cold treatment, and 10⁻⁵ M ABA treatments resulted in only marginally lower sterility levels (data not shown). The highest sterility occurs following application at the YM stage, but the difference between treatments before and after the YM stage is not as large as with cold treatment (Fig. 3a). One potential explanation for this difference with the cold treatments is that it is impossible to control the length of the ABA treatment, and the hormone may remain present in the flower tissues for longer periods of time. In control plants, the majority of pollen (95.9 ± 1.3%, n = 3) developed normally and filled with starch. Doongara anthers cold treated at the YM stage only contained 11.4 ± 1.4% (n = 3) fertile pollen, and pollen from plants injected with ABA at the YM stage also showed strongly reduced fertility (17.1 ± 4.7%, n = 3; Fig. 3c).

Fig. 3

Effect of cold and ABA treatment on spikelet sterility of Doongara and R31 panicles. The percentage of spikelet sterility was scored for panicles treated with cold or ABA before (<YM: AD −4 to 0), after (>YM: AD +2.5 to +7) and at the young microspore stage (YM; AD 0 to +2.5). (a) Effect of cold and ABA treatment on Doongara spikelet sterility. Sterility levels for both cold and ABA treatments peak when plants are treated at the young microspore stage of pollen development. The following numbers of Doongara panicles (n) were scored for this data set: control (D C; n = 112), D CD <YM (n = 18), D CD YM (n = 30), D CD >YM (n = 35), mock-injected control (D Ci; n = 36), D ABA <YM (n = 37), D ABA YM (n = 41), D ABA >YM (n = 45). (b) Effect of cold and ABA treatment on R31 spikelet sterility. The following numbers of R31 panicles (n) were scored for this data set: control (R31 C; n = 104), R31 CD <YM (n = 28), R31 CD YM (n = 37), R31 CD >YM (n = 78), mock-injected control (R31 Ci; n = 35), R31 ABA <YM (n = 22), R31 ABA YM (n = 33), R31 ABA >YM (n = 32). (c) Starch staining of pollen from control, cold- and ABA-treated Doongara panicles. Average fertility levels were 95.9 ± 1.3, 11.4 ± 1.4 and 17.1 ± 4.7% for control, cold- and ABA-treated plants, respectively (n = 3 plants). (d) Starch staining of pollen grains from control, cold- and ABA-treated R31 plants. Average fertility levels were 97.4 ± 1.3, 96.6 ± 0.6 and 6.4 ± 1.4% for control, cold- and ABA-treated plants, respectively. Plants were cold treated or injected with ABA at the YM stage, and pollen was collected from 3–4 florets per plant just. Pictures were taken at a magnification of 40× and scale bars represent 50 μm. Both cold and ABA treatment result in a reduction in the number of pollen grains, and the pollen grains are slightly smaller and often irregular in shape.

Interestingly, although R31 is highly cold tolerant, ABA treatment caused high levels of spikelet sterility (Fig. 3b). Similar to Doongara, sterility was highest when ABA was applied at the YM stage, but a high proportion of sterility also occurred when ABA was applied before and after the YM stage (Fig. 3b). Average sterility following ABA treatment was 1.6× higher in R31 than in Doongara irrespective of the developmental stage (P < 0.01; n = 123 plants). R31 anthers contain similar amounts of fertile pollen to Doongara anthers under control conditions (97.4 ± 1.3%, n = 3), but following cold treatment at the YM stage most pollen remains fertile (96.6 ± 0.6%, n = 3; Fig. 3d). ABA treatment strongly reduced the number of fertile pollen grains (6.4 ± 1.4%, n = 3; Fig. 3d).

Both cold and ABA treatment resulted in less pollen being produced, and pollen grains were slightly smaller and deformed. This could be due to premature abortion of pollen development and incomplete deposition of the pollen cell wall. Premature abortion may also explain why starch is not deposited in cold- and ABA-treated pollen, since engorgement of rice pollen grains only occurs a few days before anthesis (Raghavan 1988).

Our data show that R31 is significantly more sensitive to ABA treatment than Doongara (Fig. 3.B). ABA treatment of R31 anthers resulted in spikelet sterility and absence of starch accumulation towards anthesis (Fig. 3d).

ABA regulates OSINV4, OSMST7 and OSMST8 expression in anthers

Doongara panicles were treated in vivo with ABA, and anthers were collected after 2 d. Reverse transcription–PCR (RT–PCR) analysis showed that following ABA treatment OSINV4 expression on average was reduced to 0.61- and 0.53-fold at the YM and EB stage, respectively (P < 0.01, n = 4; Fig. 4a, b). The expression of OSMST8 was also repressed by ABA and was 0.35- and 0.64-fold the untreated levels at the YM and EB stages, respectively (P < 0.01, n = 4; Fig. 4a, b). Expression of OSMST7 was induced 8.12- and 1.28-fold at the YM and EB stage, respectively (P < 0.01, n = 4; Fig. 4a, b). ABA, which induces pollen sterility in a way similar to cold, also regulates expression of OSINV4, OSMST7 and OSMST8 in the same way as cold.

Fig. 4

RT–PCR (a) and quantification results (b) showing the effect of exogenous ABA application on the expression of the sugar metabolic genes OSINV4, OSMST8 and OSMST7 in Doongara anthers. Gene expression results are shown for untreated control young microspore (YM C) and early bicellular stage (EB C) anthers, and ABA-treated young microspore (YM ABA) and early bicellular stage (EB ABA) anthers. Results are expressed as relative expression levels compared with untreated controls (standard error bars are shown). α-Tubulin was used for normalization of the RT–PCR results. YM C expression for OSINV4 in (a) was detected using Southern blot hybridization.

Cold induces ABA biosynthesis in rice anthers

Rice has three NCED genes (OSNCED1, Os03g44380; OSNCED2, Os12g42280; and OSNCED3: Os07g05940; Tan et al. 2003), and the annotation of the rice genome shows a single ZEP gene (OSZEP1, Os04g37619).

All three rice NCED genes are expressed in anthers, and OSNCED1 expression levels were slightly lower than those of OSNCED2 and OSNCED3 (data not shown). Expression of OSNCED1 was induced on average 2.2-fold (P < 0.05, n = 14) and 1.9-fold (P < 0.05, n = 7) in cold-stressed Doongara and R31 YM anthers, respectively; expression levels were not significantly different in R31 and Doongara YM anthers under both control and cold conditions (Fig. 5). OSNCED2 expression was not significantly affected by cold in Doongara YM anthers (P > 0.05; n = 10; Fig. 5). Under control conditions, expression of OSNCED2 in R31 anthers was significantly lower than in Doongara anthers (0.3-fold; P < 0.01; n = 6), and the gene was induced 2.9-fold by cold to final expression levels that are not significantly different from those of Doongara following control and cold treatments (P > 0.05; n = 5; Fig. 5). Although there was no difference in OSNCED3 expression between Doongara and R31 under normal conditions (P > 0.05, n = 18; Fig. 5), expression of the gene was significantly different between Doongara and R31 under cold conditions: OSNCED3 was induced 2-fold by cold in Doongara (P <0.01; n = 24), but there was no induction by cold in R31 (P > 0.05; n = 11; Fig. 5).

Fig. 5

RT–PCR expression studies for the rice OSNCED1, OSNCED3 and OSZEP1 genes in anthers of Doongara (D) and R31 plants. Gene expression levels are shown for young microspore stage anthers of control (YM C) and cold-treated (YM CD) plants. The data were normalized to the average expression levels observed in control Doongara young microspore stage anthers (D YM C), and error bars represent standard errors.

The expression levels of OSZEP1 were similar in Doongara and R31 anthers under control conditions. OSZEP1 was induced 2.4-fold by cold in Doongara YM anthers (P < 0.01; n = 6), but not in R31 YM anthers (P > 0.05; n = 12; Fig. 5). The expression pattern of OSZEP1 is similar to the expression pattern of OSNCED3: both genes are induced by cold in Doongara YM anthers, but expression levels are unaffected by cold in R31 YM anthers.

The results show that OSZEP1 and two of the OSNCED ABA biosynthetic genes (OSNCED2 and OSNCED3) are expressed differently in cold-tolerant rice compared with cold-sensitive rice.

Cold affects ABA catabolism differently in Doongara and R31 anthers

The C-8′ hydroxylation pathway is thought to be the predominant ABA catabolic pathway in plants (Nambara and Marion-Poll 2005). The rice genome has three ABA 8′-hydroxylase (CYP707A) genes (Saika et al. 2007): Os02g47470 (OSABA8OX1), Os08g36860 (OSABA8OX2) and Os09g28390 (OSABA8OX3).

In Doongara anthers, OSABA8OX1 expression was induced 2.9-fold by cold treatment (P < 0.01, n = 16), and 1.5-fold by ABA treatment (P < 0.05, n = 8; Fig. 6a). In R31 anthers, OSABA8OX1 expression was on average 1.7-fold higher than in Doongara under control conditions (P < 0.05, n = 12; Fig. 6a). The gene was also induced by cold treatment (P < 0.05, n = 8; Fig. 6a) and the overall expression level following cold treatment was not significantly different from the expression levels in cold-treated Doongara anthers (Fig. 6a). Treatment with ABA did not lead to a significant induction of OSABA8OX1 in R31 anthers.

Fig. 6

Effect of cold and ABA treatment on the expression of the ABA-8′-hydroxylase genes OSABA8OX1 (a), OSABA8OX2 (b) and OSABA8OX3 (c) in Doongara and R31 anthers. Error bars represent standard errors.

The OSABA8OX2 gene was induced 3.3-fold by cold (P < 0.01; n = 12) and 3.7-fold by ABA (P < 0.01; n = 3; Fig. 6b) in Doongara anthers. Under control conditions, overall expression levels of OSABA8OX2 were 2.4-fold higher in R31 compared with Doongara (P < 0.01; n = 10). In contrast to OSABA8OX1, OSABA8OX2 expression in R31 was repressed by cold (P < 0.05; n = 5) and expression levels in R31 were only 1.3-fold higher than in Doongara under control conditions (P < 0.05, n = 5), but significantly lower than Doongara cold expression levels (P < 0.05; n = 5). OSABA8OX2 was induced by ABA in R31 (P < 0.01; n = 5), and expression levels following ABA treatment were similar in R31 compared with Doongara (3.6-fold Doongara control levels; Fig. 6b).

OSABA8OX3 is induced 2-fold (P < 0.05; n = 7) and 2.2-fold (P < 0.05; n = 3) by cold and ABA, respectively, in Doongara (Fig. 6c). Control expression levels of OSABA8OX3 were 2.6-fold higher in R31 compared with Doongara (P < 0.01; n = 4). OSABA8OX3 expression in R31 was reduced but not significantly lower under cold conditions compared with R31 control anthers (P > 0.05; n = 3). R31 cold expression levels were only 1.1-fold higher than in Doongara under control conditions but not significantly lower than Doongara cold expression levels (P > 0.05; n = 3). OSABA8OX3 was induced 3.6-fold above Doongara control levels by ABA in R31 (P < 0.01; n = 5), and expression levels following ABA treatment were similar in R31 and Doongara (Fig. 6c).

These results indicate that Doongara and R31 show differences not only in ABA biosynthesis but also in its catabolism in response to cold stress. The expression of rice ABA 8′-hydroxylase genes is induced by ABA, and this capacity is also different in Doongara and R31anthers.

Discussion

The early stages of pollen development (pollen mother cell stage to VM stage) are characterized by active growth and high metabolic activity in the anther. Anthers have the highest sink strength in the flower, and large amounts of sugars are mobilized to the anthers to support early development (Clément et al. 1996, Castro and Clément 2006). Our previous results have shown that cold stress blocks the mobilization of sugar to the YMs, culminating in abortion of pollen development and sterility. The tapetum is the interface between the sporophyte and gametophyte, and as such plays a critical role in pollen development. OSINV4 expression at the YM stage is tapetum specific and therefore its repression by cold is at the basis of the blockage of sugar supply to the tapetum and pollen grains (Oliver et al. 2005). Sugar accumulates in cold-stressed anthers, and later starch accumulates in the anther wall (Satake 1976).

Glucose and fructose, the reaction products of cell wall invertase, also accumulate in cold-treated anthers, suggesting that the monosaccharide transporter step in the apoplastic sugar transport pathway is also repressed by cold (Oliver et al. 2005). Synchronous regulation of these genes has previously been reported (Ehness and Roitsch 1997). The rice OSMST8 gene has a spatial and temporal expression pattern that strongly resembles OSINV4 expression (Oliver et al. 2007), and herein we show that this gene is also repressed by cold, indicating that OSMST8 may function in the same pathway as OSINV4. However, the substrate specificity of OSMST8 remains unknown and the size and complexity of the monosaccharide transporter family in plants suggest that other rice anther MST genes may also be involved. In Arabidopsis, only one (AtSTP2) of the 14 MST genes was expressed at the YM stage (Truernit et al. 1999). OSMST7 is induced by cold in anthers but its temporal expression pattern suggests that it is not expressed in the tapetum (Oliver et al. 2007). OSMST7 may be expressed in the pollen grains or in the anther wall. Both glucose and sucrose have been shown to induce MST gene expression (Atanassova et al. 2003), suggesting that the induction of this gene could be a consequence of the sugar accumulation in cold-stressed anthers.

The observation that sugar metabolic genes additional to OSINV4 are affected by cold is indicative of the severity of the disturbance cold causes in anther sink–source relationships. Sugars are not transported to the tapetum and pollen but rather are redirected to the anther wall where they accumulate as starch (Satake 1976). This interruption of sugar transport to the tapetum and pollen is likely to be an important contributor to pollen sterility, because the provision of sugars to the pollen for starch synthesis is essential for pollen fertility (Pacini 1996, Clément and Audran 1999, Castro and Clément 2006). In addition, the fact that OSINV4, OSMST8 and OSMST7 are not affected by cold in anthers of the cold-tolerant cultivar R31 suggests that there is a strong relationship between the expression of these genes and the cold tolerance phenotype.

We were interested in investigating the role of ABA as a signal for CIPS because of the similarity in the phenotype between drought- and cold-induced pollen sterility. Both cold temperatures and water stress applied specifically at the YM stage cause sucrose accumulation in anthers, repression of cell wall invertase activity and gene expression, and depletion of starch in mature pollen (Saini 1997, Koonjul et al. 2005, Oliver et al. 2005). This similarity suggests that both stresses use a common regulatory mechanism to induce abortion of pollen development. ABA is suggested to play a causative role in drought-induced pollen sterility in wheat spikelets (Morgan 1980, Saini and Aspinall 1982, Morgan and King 1984, Westgate et al. 1996). In addition, the promoters of OSINV4, OSMST7 and OSMST8 contain ABA, cold and drought response elements (Oliver et al. 2007), suggesting that ABA could regulate their expression during CIPS. ABA also interacts with sugar signaling (Finkelstein and Gibson 2002, Brocard-Gifford et al. 2003), and attempts to isolate sugar response mutants resulted in the identification of genes that were previously known as ABA mutants (Arenas-Huertero et al. 2000, Laby et al. 2000, Rook et al. 2001, Arroyo et al. 2003).

Our results provide correlative evidence that ABA acts as a regulatory signal for the induction of CIPS. Cold treatment causes endogenous ABA levels to increase in Doongara, and exogenous application of ABA results in pollen sterility, showing higher sensitivity at the YM stage. ABA treatment and cold also have the same effect on the expression of OSINV4, OSMST7 and OSMST8 in Doongara. The likelihood of ABA being a signal in CIPS is further strengthened by the comparison of the cold-sensitive cultivar Doongara with the cold-tolerant cultivar R31. Compared with Doongara, R31 has low anther ABA levels which remain low following cold treatment.

The genes encoding the first two enzymes of the ABA biosynthetic pathway, zeaxanthin epoxidase (ZEP) and 9-cis-epoxycarotenoid dioxygenase (NCED), are induced by osmotic stresses (Taylor et al. 2000, Mulholland et al. 2003), and stress resistance in plants can be manipulated by changing their expression (Xiong and Zhu 2003). The differences in ABA levels between Doongara and R31 are paralleled by differences in expression of genes encoding enzymes of ABA biosynthesis, OSNCED2, OSNCED3 and OSZEP1. OSNCED2 expression is much lower in R31 YM anthers. While the gene is not induced by cold in Doongara anthers, cold induction in R31 anthers does not increase OSNCED2 expression above Doongara expression levels. In contrast, while expression of OSNCED3 and OSZEP1 is induced by cold in Doongara anthers, their expression is not affected in R31. Additionally, R31 anthers maintain higher expression levels of the OSABA8OX1, OSABA8OX2 and OSABA8OX3 genes encoding ABA 8′-hydroxylase under normal conditions, and these genes are regulated differently following cold treatment in R31 compared with Doongara anthers. Although all three OSABA8OX genes are induced by cold in Doongara, only OSABA8OX1 shows some degree of induction in R31, while OSABA8OX2 and OSABA8OX3 are down-regulated by cold treatment. Our anther ABA measurements indicate that the net result of these changes in ABA biosynthetic and catabolic gene expression ultimately is significantly lower ABA levels in R31 compared with Doongara anthers. The maintenance of lower ABA levels, or, perhaps more importantly, maintaining a fine balance between ABA biosynthesis and degradation, appears to be important for determining cold tolerance in rice anthers. The fact that exogenous ABA application causes pollen sterility in R31 suggests that it is the lower level of ABA that is responsible for cold tolerance and that a regulatory mechanism controlling ABA biosynthesis and catabolism functions differently in R31 compared with Doongara.

It is possible that this signal for the regulation of ABA metabolism is a long-distance signal that originates in other plant parts. The observation that leaves of R31 plants also have less ABA than Doongara under normal and cold conditions (Fig. 2d) suggests that this regulatory mechanism acts plant-wide and is not confined to the anther. Interestingly, R31 anthers appear to be more sensitive to exogenous ABA application than Doongara anthers. It is possible that the lower endogenous ABA levels of R31 are the consequence of a feedback regulatory system of ABA biosynthesis that is more sensitive to endogenous ABA levels than in Doongara. Alternatively, ABA levels could be controlled through interactions with other plant hormones. Arabidopsis mutants with altered ABA sensitivity were shown to be ethylene response mutants (Beaudoin et al. 2000, Ghassemian et al. 2000). Recently it was shown that ethylene promotes rice ABA 8′-hydroxylase gene expression under submergence conditions (Saika et al. 2007), suggesting that ethylene plays an important role in regulating ABA levels.

Our results may point to a central role for ABA in regulating pollen development under abiotic stress conditions such as cold and drought. In predominantly self-fertilizing cereals such as rice and wheat, regulation of anther sugar supply for pollen development by ABA offers a powerful mechanism to control fertility and ultimately grain number under adverse environmental conditions. ABA is unlikely to be the sole regulator of pollen development, as ABA is known to act antagonistically with the growth-stimulating hormone gibberellic acid (Jacobsen 1973, White and Rivin 2000). Gibberellic acid biosynthetic genes are also expressed in the tapetum layer of rice anthers (Weiss et al. 1995, Kaneko et al. 2003, Murray et al. 2003, Kaneko et al. 2004). A correct balance of ABA/gibberellic acid is required for normal pollen development (Wassom et al. 2001), and abnormally high levels of ABA in tomato anthers are associated with pollen sterility (Singh and Sawhney 1998). ABA also interacts with ethylene, which is involved in senescence and cell death responses (Beaudoin et al. 2000, Ghassemian et al. 2000, Sharp 2002, Benschop et al. 2005). Increased ABA levels were shown to induce premature cell death and senescence of daylily petals (Panavas et al. 1998). ABA plays an essential role in controlling the programmed cell death event that leads to natural degradation of the tapetum (Wang et al. 1999, Wu and Cheun 2000, Ku et al. 2003). The timing of this event is absolutely crucial, and various metabolic disturbances that lead to premature or delayed tapetal cell death cause male sterility (Mariani et al. 1990, van der Meer et al. 1992, Worrall et al. 1992, Tsuchiya et al. 1995, Hernould et al. 1998, Ku et al. 2003, Yui et al. 2003).

Investigations of the Chinese cold-tolerant cultivar R31 have revealed that the cold tolerance phenotype may be due to differences in a regulatory mechanism that fine-tunes the balance of ABA biosynthesis and catabolism. This regulatory mechanism may involve interactions with other plant hormones. The mapping of cold tolerance quantitative trait loci of R31 is currently in progress and the availability of the rice genome sequence will enable us to map the gene(s) that is (are) responsible for the cold tolerance phenotype. The availability of molecular markers for cold tolerance genes will greatly facilitate the breeding process for cold-tolerant rice.

Materials and Methods

Plant growth and treatment conditions

The cold-sensitive Australian japonica cultivar Doongara and the tolerant Chinese japonica cultivar R31 were used throughout these experiments (Oliver et al. 2005). Seeds were sown in 15 cm pots containing a mix of 25% compost, 75% vermiculite and approximately 5 g of osmocote fertilizer. Plants were grown in the greenhouse (light, 28°C; dark, 24°C) with the pots and root system submerged in water. Determination of anther developmental stages was carried out using the AD method (AD = the distance in cm between the nodes of the flag leaf and the penultimate leaf; Oliver et al. 2005). For this study we used the following pollen developmental stages: before the YM stage (<YM; AD −5 to 0 cm), the YM stage (AD 0 to +2.5 cm), the EB stage (AD +6 to +8 cm) and after the YM stage (>YM; AD +2.5 to +9 cm). Cold treatment was carried out in a controlled growth chamber (light, 13 h at 400 μmol m⁻² s⁻¹; dark, 11 h) for 3 d at a constant temperature of 12°C. For gene expression studies, the AD was anticipated and plants started the 3 d cold treatment about 1–2 cm ahead of the AD where the YM and EB stages are reached. For large-scale cold treatments, the AD refers to the AD at the start of the cold treatment. Anthers were collected using a dissecting microscope, and were always harvested in the afternoon to minimize diurnal variation effects. Upon harvesting, anthers were immediately frozen in liquid nitrogen, and stored at −80°C.

Plants were injected with a 10⁻⁴ M ABA (+/− 2-cis, 4-trans-abscisic acid; Sigma, St Louis, MO, USA) solution (in 50 mM sodium phosphate buffer, pH 7.2, and one drop of Tween-20 per 50 ml; Zeng et al. 1985). Control mock-injected plants were treated with the same solution lacking ABA. Plants were injected at a fixed time in the afternoon to minimize diurnal variation effects. The solution was injected with a syringe and a fine 25 G 5/8″ needle into the space between the leaf sheaths surrounding the developing panicle. For gene expression studies, anthers were collected for RNA extraction 2 d after ABA injection. Control anthers were collected from mock-injected plants.

For scoring spikelet sterility, plants were returned to normal growth conditions (cold treatment) and allowed to develop until grain setting. The percentage of spikelet sterility is expressed as the number of unfilled grains at maturity [number of mature sized unfilled florets relative to the total number of mature sized florets (empty + filled)]. In rice, cold treatment mainly affects pollen formation (Hayase et al. 1969) and therefore a spikelet is due to absence of fertilization and pollen sterility. The two-sample t-test was used to test for significant differences between control and ABA- or cold-treated sample means. Monitoring starch content in pollen using iodine staining was as described previously (Oliver et al. 2005). Starch staining of pollen grains is a good viability staining method for cold-stressed pollen and can be used to determine pollen fertility (Gunawardena et al. 2003).

ABA measurements

The ABA concentration in anthers, panicles and leaves was measured using the Phytodetek competitive enzyme-linked immunosorbent assay (ELISA) kit (Agdia, Elkhardt, IN, USA). Plants were cold treated for 5, 24 or 72 h, and then anthers, panicles and leaves were collected at a fixed time of the day (to minimize diurnal effects) and frozen in liquid nitrogen; material was also collected at the same time from control plants. Anther material was freeze-dried and the dry weight determined; panicle and leaf material was first ground in liquid nitrogen, and then the powder was freeze-dried and weighed. The powder was extracted overnight at 4°C in cold 80% methanol. The mixture was then centrifuged at 5,000 r.p.m. for 5 min and the supernatant was collected. The overnight extraction was repeated on the pellet, and the supernatants were pooled. The pellet was washed a further three times in cold 80% methanol by vortexing and spinning, and the supernatant for each sample was pooled and dried down in a SpeedyVac until approximately 50 μl of liquid remained. TBS buffer (25 mM Tris–HCl pH 7.5, 100 mM NaCl, 1 mM MgCl₂, 3 mM NaN₃) was added to a final volume of 250 μl (for anthers), 1 ml (for panicles) or 500 μl (for leaves). These extracts were diluted 10-fold in TBS and used in the ELISA according to the Phytodetek protocol. A standard curve of different ABA dilutions was constructed to calculate the sample ABA concentrations. ABA concentrations were calculated as ng per g DW. Each measurement was replicated 3–4 times using different pooled anther, leaf or panicle samples (biological repeats) and in replicate experiments using different batches of plants as indicated in the figure legends. The two-sample t-test was used to test for significant differences between control and cold-treated sample means.

RNA expression studies using semi-quantitative RT–PCR

Anther RNA was isolated using the RNeasy Plant Kit (Qiagen, Hilden, Germany). Total RNA was treated with RQ1 DNase (Promega, Madison, WI, USA) and precipitated using 2 M LiCl. Each RNA sample was extracted from the anthers harvested at the same developmental stage from 1–2 plants. RNA concentrations were measured spectrophotometrically and an RNA gel was run from each batch of RNA samples in order to check the quality of the RNA and the accuracy of the concentration. Semi-quantitative RT–PCRs were carried out by reverse transcription of 0.5 μg of total anther RNA using Thermoscript reverse transcriptase (Invitrogen, Carlsbad, CA, USA). The PCR step was carried out using an Advantage GC genomic PCR kit (BD Biosciences, Palo Alto, CA, USA) with gene-specific primers (see Table 1). PCR products were all cloned and sequenced to ascertain that the correct template was amplified. α-Tubulin was used as a control in the RT–PCRs, and the α-tubulin results were used to normalize the gene expression results (Oliver et al. 2005). RT–PCR products were analyzed on 1.5% agarose gels and quantified by image scanning software (Multigauge software; Fujifilm, Tokyo, Japan). For some RT–PCRs, reaction products were detected by Southern blot hybridization using radiolabeled probes of the cloned PCR fragments. Probes were prepared using a random primer extension kit (Perkin Elmer, Boston, MA, USA). Hybridized filters were analyzed and quantified using a Fujifilm FLA-5000 phosphorimager and Multigauge software (Fujifilm, Tokyo, Japan). The two-sample t-test was used for statistical analysis of all the gene expression data. All rice gene locus numbers are according to release 4.0 of the TIGR Rice database (www.tigr.org/tdb/e2k1/osa1/index.shtml). Gene expression data for each sample were normalized relative to α-tubulin expression levels, and average expression levels were calculated for several biological repeats. The two-sample t-test was used for significance testing, and the P-value as well as the number of repeat RNA samples (n) is indicated in the text.

Table 1

List of RT–PCR primers used in this study

Gene	Primer sequences
OSINV4	5′-GCTTCGTCGACGTCGACATCG
	3′-CTAGTACTGCTACAACCACC
α-Tubulin	5′-GGCTTGTGTCTCAGGTTATCTCATC
	3′-CATGGAGGATGGCTCGAAGG
OSMST7	5′-ATGTTCGCCACCTTCGTCTCCATC
	3′-ACCGCACTAGAGCTGTAAGATAGA
OSMST8	5′-CCTCGGCCGCCGCAAGCTCCTCC
	3′-GCAATGGATCGATGTAGCCAGCAG
OSNCED1	5′-GCGCGCCATCCTGCCGCCGTCGAGC
	3′-AACGTGGAGGTGTTCGATCGATGGATC
OSNCED2	5′-CCGACCACCTTCACGCCCAATT
	3′-TTCTTCTTGCCACACCTCCTCGCC
OSNCED3	5′-ATGAAGCAATTTAGTCATGCGT
	3′-CTCGCGATCGATCGATTTGGCG
OSZEP1	5′-TATGTGACTGATAATGGAAGTG
	3′-ATGGCTGACTGAAGTCTCTCGT
OSABA8OX1	5′-GCCGTGGAGGACGTGGAATACCAAG
	3′-TCACTCCTGCTCGGTGTTCTTGCG
OSABA8OX2	5′-GTTCGGGAGCGGCGTGCACGCGTGC
	3′-TCGATCGATCGACCATTGCATCCACAC
OSABA8OX3	5′-AGTACAGCCCATTCCCTGTG
	3′-ACGCCTAATCAAACCATTGC

Acknowledgments

The authors wish to thank Ms. J. Edlington for excellent technical assistance. S.N.O. was supported by a PhD fellowship from the Cooperative Research Centre for Sustainable Rice Production (#3209). This research is supported by the Rural Industries Research and Development Corporation (#US-143A).z

Abbreviations:

Abbreviations:
• ABA8OX

  ABA-8′-hydroxylase

• AD

  auricle distance

• CIPS

  cold-induced pollen sterility

• EB

  early bicellular

• ELISA

  enzyme-linked immunosorbent assay

• NCED

  9-cis-epoxycarotenoid dioxygenase

• RT–PCR

  reverse transcription–PCR

• YM

  young microspore

• ZEP

  zeaxanthin epoxidase.

Author notes

³Present address: Max Planck Institute of Molecular Plant Physiology, Am Muehlenberg 1, D-14476 Golm, Germany.