High-Temperature-Induced Defects in Tomato (Solanum lycopersicum) Anther and Pollen Development Are Associated with Reduced Expression of B-Class Floral Patterning Genes

Sexual reproduction is a critical process in the life-cycle of plants and very sensitive to environmental perturbations. To better understand the effect of high temperature on plant reproduction, we cultivated tomato (Solanum lycopersicum) plants in continuous mild heat. Under this condition we observed a simultaneous reduction in pollen viability and appearance of anthers with pistil-like structures, while in a more thermotolerant genotype, both traits were improved. Ectopic expression of two pistil-specific genes, TRANSMITTING TISSUE SPECIFIC and TOMATO AGAMOUS LIKE11, in the anthers confirmed that the anthers had gained partial pistil identity. Concomitantly, expression of the B-class genes TOMATO APETALA3, TOMATO MADS BOX GENE6 (TM6) and LePISTILLATA was reduced in anthers under continuous mild heat. Plants in which TM6 was partially silenced reacted hypersensitively to temperature elevation with regard to the frequency of pistilloid anthers, pollen viability and pollen quantity. Taken together, these results suggest that high-temperature-induced down-regulation of tomato B-class genes contributes to anther deformations and reduced male fertility. Improving our understanding of how temperature perturbs the molecular mechanisms of anther and pollen development will be important in the view of maintaining agricultural output under current climate changes.


Introduction
Sexual reproduction is a key process in the life-cycle of plants, both in natural and in agricultural settings, with fitness and yield, respectively, depending directly on its success. The efficiency of the process, however, is strongly influenced by environmental conditions, such as water availability and ambient temperature [1,2]. With the current, rapid changes in global climate, the average and maximum temperatures are expected to increase in many parts of the world as well as the frequency of heat waves [3]. Studies in various plant species, monocot and dicot, have shown that development and functioning of especially the male gametophyte is highly sensitive to both continuous mild increases and short extremes in temperature [4,5]. A number of studies have addressed the problem of heat-induced male infertility in plants and identified deviations in various tissues [6][7][8] and metabolic processes [9][10][11][12][13][14][15][16]. During their development, the male gametophytes are actively supported by the surrounding sporophytic cell layer, called the tapetum [17]. Because developmental defects in the pollen under high temperature are often accompanied by developmental defects in the tapetum [7,8,[18][19][20][21][22][23] it has been suggested that the the latter constitutes the primary defect [1,24]. The sensitivity of the tapetum to high temperature has been confirmed by a transcriptome analysis in rice, showing that heat-repressed genes in anthers preferentially locate to this tissue [16].
Patterning of floral organs is primarily under control of a set of MADS-box genes. These genes encode transcription factors with a conserved MADS-box DNA binding domain and are thought to exist in all eukaryotes. In plants, MADS-box genes are involved in controlling major aspects of development, including male and female gametophyte development, embryo and seed development, as well as root, flower and fruit development [25].The MADSbox genes involved in floral organ identity are also referred to as the ABC genes. The ABC model was proposed as a result of research on mutants with homeotic deformations [26]. The model divides the floral meristem into four whorls, each characterized by the expression of specific (combinations) of ABC genes. This, in turn leads to the formation of the four kinds of flower organs: activity of A-class genes alone in the first whorl will result in the formation of sepals, the combination of A-and B-class genes leads to petals in the second whorl, B-and Cclass genes produce stamens in the third whorl and C-class genes alone direct pistil development in the central fourth whorl [25,27].
When examining tomato flowers grown under continuous mild heat conditions, we observed not only a reduction in pollen number and viability, but also the appearance of anther deformations resembling a partial conversion to pistil identity. The aim of this study was to identify the molecular changes underlying these deformations and determine the relation between anther deformation and impaired development of the male gametophyte under mildly elevated temperatures.

Plant material and treatments
S. lycopersicum cultivar Red Setter was obtained from Metapontum Agrobios (Metaponto, Italy). The cultivar Microtom (TOMJP00001) was provided by University of Tsukuba, Gene Research Center, through the National Bio-Resource Project (NBRP) of the MEXT, Japan. The TM6-RNAi line (TM6i6) in the Microtom background was kindly provided by Dr Vivian Irish (Yale University, New Haven, USA; De Martino et al., 2006). The lines AtGRXS17-3 and AtGRXS17-9 were in the Rubicon background [28].
Plants were raised in climate chambers (14 hours per day light,~200 μmol/m 2 s at plant level,~70% humidity) under control conditions (25˚C day, 19˚C night). Upon appearance of the first inflorescence, buds and flowers were removed and the plants were kept under control conditions, continuous mild heat (CMH32: 32˚C day, 26˚C night) or higher continuous mild heat (CMH34: 34˚C day, 28˚C night). Flowers that developed fully under these conditions were used to determine the frequency of anther deformations, light microscopy, gene expression studies and pollen germination rate. Pollen germination was analysed by in-vitro pollen germination as described by [29] with minor changes (sucrose and PEG400 were adjusted to 5% and 25%, respectively). To determine the frequency of anther deformations in Red Setter at least eight plants were used per treatment and all flowers produced over a period of two weeks were harvested (24 flowers under control conditions, 91 flowers under CMH32 and, due to reduced flower set, 10 flowers under CMH34). To analyse the lines AtGRXS17-3 and AtGRXS17-9 three plants were used per treatment and ten flowers per plant, that did fully develop under these conditions, were harvested. Flowers that did showed spacing between anthers, bent anther tips, green or shortened stamen were counted as deformed.
For gene expression studies seven and four plants of the cultivar Red Setter were grown under control and CMH32, respectively. Anthers and pistils of the flowers were harvested separately and sampled randomly into pools (5 mature flowers or 10 flower buds with an anther length of 2-3 mm per pool).
To study the Microtom TM6-RNAi line, Microtom wild type and transgenic plants were grown in climate cabinets (MC1600, Snijders Labs, The Netherlands) under control conditions (constant 22˚C, 12 hours LED-light from Philips Green Power LED DR/B/FR 120,~250 μmol/ m 2 s, 60% humidity) and moved into milder continuous mild heat (CMH30: 30.5˚C day, 25.5˚C night,) upon flowering. To determine the rate of flower deformations newly opened flowers were harvested daily on day 8 to 15 after moving the plants into CMH30, flowers were scored as described above. Pollen germination was determined from newly opened flowers on day 8 after moving into CMH30 and pollen number on the days 13-15 after moving into CMH30. Flowers and pollen from control conditions were harvested and analysed on the same days. Pollen germination was analysed as described above and pollen number was determined with a Bürker-Türk counting chamber.

Light microscopy
Anthers were fixed in FAA for 24 hours at 4˚C, dehydrated in a graded ethanol series and embedded in Technovit 7100. Sections of 5-7μm were cut with a rotary microtome and stained with toluidine blue 0.1% in 1% borax. Sections were viewed and photographed with a Leitz Orthoplan microscope, equipped with a Leica camera DFC 420C.

qRT-PCR
RNA was extracted from the flower organs using Trizol (Invitrogen) and treated with DNase I. cDNA was synthesized using the IScript kit (Bio-Rad) and the PCR (end volume of 25 μl, containing cDNA equivalent to 10 ng RNA) was performed using the SybrGreen iQ mix (Bio-Rad). The primer pairs (Table 1) used for the analysis were designed using Beacon Designer (Premier Biosoft). Cq values were determined by CFX Manager 3.0 (Bio-Rad) and the efficiency was calculated using LinReg PCR. The normalization factor was calculated using GeNorm (Biogazelle), using SAND, CAC, EF1α and RPL8 as reference genes. The qRT-PCR was performed with four and five biological replicates from flower buds and mature flowers, respectively.

Statistics
Frequencies of anther deformation under three temperatures and in three genotypes were averaged on plant level and compared using a non-parametric Kruskal-Wallis one-way ANOVA followed by Mann-Whitney U test to compare group means pairwise. As the data on frequencies of anther deformation in the TM6 experiment did not comply with assumptions of ANOVA, they were tested using a generalized linear model on count data. Pollen germination data and pollen number data were logit-transformed and log-transformed, respectively, and analysed by a one-way ANOVA with Tuckey or two-way ANOVA with Tuckey followed by a one-way ANOVA to test for differences between individual groups, as appropriate. For analysis of the qRT-PCR data, log-transformed normalized relative quantities were compared using a student's t-test, according to [30]. All analyses were performed using IBM SPSS statistics version 21 (IBM).

Co-occurrence of floral deformations and low pollen viability under continuous mild heat conditions
As has been reported before [31], tomato pollen viability was compromised when plants were grown under contiunuous mild heat (CMH), with clear effects being visible at~32˚C/26˚C day-and night-time temperatures (CMH32; Fig 1A). In addition, we observed that under these conditions part of the anthers developed deformations, consisting of spacing between anthers, twisting and greening of the tips ("antheridial cone splitting"; Fig 1B). The frequency of these deformations increased with higher temperatures (CMH34), while at the same time pollen viability was further reduced (Fig 1A). To test whether the negative correlation between the occurrence of anther deformation and pollen viability remained when using a different perturbation than temperature, we compared wild-type plants with plants with increased heat tolerance, due to overexpression of the glutaredoxin GRXS17 [28]. When grown under control conditions, transgenic lines were indistinguishable from the wild type. In contrast, when grown at CMH32, the AtGRXS17-overexpressing lines showed significantly fewer flowers with anther deformations than the wild type, while the pollen viability was higher than in the wild type ( Fig 1C). Thus, in both experimental systems, pollen viability and the presence of anther deformation were negatively correlated within the same temperature range.

CMH-induced anther deformations are homeotic
Closer inspection revealed that in some cases aberrant tissue was present on the adaxial side of the anthers grown under CMH32 (Fig 1B). Cytological analysis showed the presence of clusters of relatively small cells, along the length of the anthers (compare Fig 2A and Fig 2B). This abnormal tissue bore a similarity to the transmitting tissue found in the centre of the style ( Fig  2C). Furthermore, ovule-like protrusions were found at the base of the anthers (Fig 2D). Thus, tissues and organs resembing those normally found in the inner-most whorl of tomato flowers  were present on the anthers of flowers grown under CMH32, suggesting that CMH32 induced a partial homeotic conversion from anthers to pistil.
To independently confirm that the anthers gained pistil-like identity under CMH32, we harvested anthers and pistils from plants grown under control and CMH32 conditions and analyzed the expression of two pistil-specific genes. In tobacco, TRANSMITTING TISSUE SPE-CIFIC (TTS) is known to be expressed in the transmitting tissue of the style [32]. The tomato ortholog (Solyc02g078100), which we named SlTTS, was identified on the basis of sequence similarity. Gene expression analysis showed that its transcript indeed accumulated specifically in pistils, with strongest expression in mature flowers (Fig 3A). Similarly, we confirmed that Frequency of anther deformations and pollen germination rate of wild type (WT) tomato plants (cv. Rubicon) and transgenic lines AtGRXS17-3 and AtGRXS17-9 grown under CMH32. Under control conditions, no anther deformations were observed, and percentage pollen viability was 60 ± 12 (SD), 60 ± 13 and 56 ± 9 for WT, AtGRXS17-3 and AtGRXS17-9, respectively (no significant differences between genotypes). Values represent the mean ± SD, different letters indicate statistically significant differences (per trait, P<0.05).
doi:10.1371/journal.pone.0167614.g001 the expression of the MADS-box gene TOMATO AGAMOUS LIKE11 (TAGL11), involved in ovule development [33], was pistil specific (Fig 3B). Subsequent analysis showed that the anthers of plants grown under CMH32 had significantly higher expression of SlTTS and TAGL11 then those from plants grown under control conditions (Fig 3C and 3D). This result supports the hypothesis that the observed anther deformations are homeotic conversions.

Expression of B-and C-class genes under CMH32
Anther identity depends on activity of B-and C-class genes and a decreased ratio of B-to Cclass gene expression may lead to pistil-like identity in the anthers [34]. Therefore, we analyzed expression of these classes of genes in anthers grown under control and CMH32 conditions. In core eudicots, the APETALA3 (AP3) B-class gene lineage has diverged into two paralogous lineages: euAP3 and TOMATO MADS BOX GENE6 (TM6) [35]. While in Arabidopsis and Antirrhinum only the euAP3 lineage has been retained [36], tomato has both lineages, i.e. TOMATO AP3 (TAP3) and TM6 [34]. A loss-of-function mutation in TAP3 results in a homeotic conversion of stamen to pistil-like tissue, as well as conversion of petals to sepal-like structures [37]. Silencing of TM6, on the other hand, only leads to homeotic conversions of the stamen [34,37]. Similarly, tomato contains two PISTILLATA-type B-class genes, i.e. Lycopersicum esculentum PISTILLATA (LePI) and TOMATO PISTILLATA (TPI) [34,38]. Fig 4 shows that the relative expression of the B-class gene TAP3 was already reduced in young anthers grown under CMH32, while in mature anthers also TM6 and LePI were expressed significantly lower. The B-class gene TPI was not differentially expressed, and also expression of the C-class genes TOMATO AGAMOUS1 (TAG1) and TAG-LIKE1 (TAGL1) was not significantly affected by CMH32. Interestingly, the more heat tolerant AtGRXS17-overexpression lines (Fig 1C) not only showed markedly reduced ectopic expression of the pistil marker TAGL11 in young anthers under CMH32 than the wild type, but also significantly higher AP3 expression (Fig 5).

Flower phenotype of a TM6-RNAi line under CMH30 and control conditions
To determine whether the observed down-regulation of B-class genes in anthers under CMH32 acted up-or downstream of the change in identity, we analysed a weak TM6-RNAi line [34]. As was previously described, this line showed only very few anther-to-pistil conversions when grown under control conditions (Table 2). However, when grown under CMH30, a temperature regime not high enough to cause homeotic conversions in the wild-type background, the frequency of anther-to-pistil related deformations in this line increased significantly (Table 2). At this temperature regime, most of the TM6-RNAi flowers had strong anther deformations, even stronger than those of wild-type plants under CMH32. The enhanced effect at CMH30 might be due to combined mild reductions in TM6 and AP3 expression, although activities of these genes are not dependent on each other [34].

Effect of reduced B-Class gene activity on pollen viability
To determine whether reduction of B-Class gene activity also affected pollen development, we analysed pollen number and pollen germination potential of the wild type and TM6-RNAi lines grown under control and CMH30. As expected, pollen number and viability was somewhat reduced by CMH30 in the wild type. Already under control conditions, TM6-RNAi plants produced significantly fewer pollen than the wild type, and the remaining pollen had lower viability (Table 2). Under CMH30, these characteristics were further affected, especially pollen viability. There was a significant interaction between genotype and treatment for both traits, in that the TM6-RNAi line responded singnificantly more strongly to the treatment than the wild type (Table 2). Taken together, our results show that an artificial decrease in TM6 expression mimics/enhances high temperature-induced phenotypic defects in both, anther and pollen development.

Discussion
In the current era of global warming, high temperature damage to plants is a increasingly frequent phenomenon, particularly around the reproductive period [39,40]. Heat-induced male infertility has been noted in a wide variety of angiosperms and a link between defects in pollen development and the function of sporophytic anther tissue under high temperature has been hypothesized. Data presented here suggest a mechanistic basis for this relation. Our results fit a model in which continuous mild heat represses B-class gene activity, which normally acts to direct stamen identity and thereby supports pollen development. Interestingly, gross morphological changes in stamen size and morphology have been noted to accompany reduction of pollen viability not only in the various tomato accessions in this study, but also in a number of other species [41][42][43][44]. Furthermore, closer inspection of a data set from [8] on transcriptomic changes in developing flower organs of barley under CMH revealed a similar down-regulation of B-class genes (data not shown). Together, these findings suggest that the model proposed may be more broadly applicable to flowering plants. Support for our results comes from the tomato stamenless type of mutants mutant. The sl mutant has complete homeotic conversion of anthers to carpels (fused to the central carpel) and was recently found to be a TAP3 mutant [37,45]. sl-2 is likely to be allelic to sl [46], but has a weaker homeotic phenotype, analogous to the TM6-RNAi line used in our studies [37,47]. Interestingly, the flower phenotypes of these mutants are temperature dependend, as flower morphology and fertility of both can be rescued by growth at reduced temperatures and the anther and pollen defects of the sl-2 mutant are strongly enhanced when the plants are grown at elevated temperature [45,47]. These findings are in accordance with the fact that TAP3/TM6 gene expression is reduced by high temperature (Fig 4) and induced by low temperature treatment [48]. Thus, B-class gene activity seems to be negatively regulated by ambient temperature over a broad range of temperatures.
Studies in Arabidopsis have shown that expression of B-class genes is regulated by different pathways over the course of flower development. During the floral meristem patterning stage, their expression is induced by LEAFY and APETALA1 [49,50]. At later stages of anther and pollen development, however, B-class gene expression is maintained by gibberellin (GA) via degradation of DELLA proteins, which act as repressors of the B-class MADS-box genes [50,51]. Notably, a GA-hyposensitive mutant in wheat has recently been shown to be hypersensitive to high temperature regarding seed set [52] and among rice cultivars, the ability to maintain pollen viability under high temperatures was shown to correlate with higher contents of GA [53].
Phenotypically, male infertility as caused by high growth temperature is very similar to that caused by drought during flower development, including abberent development of the tapetum [1,24]. In Arabidopsis, it has been shown that AP3 gene expression is also reduced under drought [54], which may mean that the similarity extends towards the underlying molecular defect. Interestingly, homeotic anther conversion has also been described for a cytoplasmic male sterile (CMS) Brassica napus line and expression of the B class genes was strongly reduced in the CMS line [55]. The reason for the anther phenotype is unknown, but as the CMS is caused by mitochondrial dysfunction, it has been suggested to be related to higher reactive oxygen species (ROS) production [56]. Because high temperature is known to affect ROS levels, it is tempting to hypothesize that this represents the initial defect in this condition as well. This would agree to our finding that overexpression of a glutaredoxin gene improves anther and pollen development and AP3 gene expression under high temperature. Future studies combining manipulations of B-class gene activity, GA signaling strength and ROS levels should shed light on this.