To gain new insights into the mechanism underlying cytoplasmic male sterility (CMS), we compared the nuclear gene expression profiles of flowers of a Brassica napus CMS line with that of the fertile B. napus maintainer line using Arabidopsis thaliana flower-specific cDNA microarrays. The CMS line used has a B. napus nuclear genome, but has a rearranged mitochondrial (mt) genome consisting of both B. napus and A. thaliana DNA. Gene expression profiling revealed that a large number of genes differed in expression between the two lines. For example, nuclear genes coding for proteins that are involved in protein import into organelles, genes expressed in stamens and pollen, as well as genes implicated in either cell-wall remodeling or architecture, were repressed in the CMS line compared with B. napus. These results show that the mt genome of the CMS line strongly influences nuclear gene expression, and thus reveal the importance of retrograde signalling between the mitochondria and the nucleus. Furthermore, flowers of the CMS line are characterized by a replacement of stamens with carpelloid organs, and thus partially resemble the APETALA3 (AP3) and PISTILLATA (PI) mutants. In accordance with this phenotype, AP3 expression was downregulated in the stamens, shortly before these organs developed carpelloid characteristics, even though it was initiated correctly. Repression of PI succeeded that of AP3 and might be a consequence of a loss of AP3 activity. These results suggest that AP3 expression in stamens depends on proper mt function and a correct nuclear–mt interaction, and that mt alterations cause the male sterility phenotype of the CMS line.
Cytoplasmic male sterility (CMS) is a maternally inherited trait where the flower fails to produce functional pollen. Besides naturally occurring CMS in wild populations of plants, the trait can be artificially obtained. Alloplasmic lines can be produced either by sexual crossings or by protoplast fusions followed by backcrosses, where the nuclear genome from one species is combined with the cytoplasm of another (Kaul, 1988). The novel nuclear-cytoplasmic combinations often result in aberrant expression of the mitochondrial (mt) genes (de la Canal et al., 2001; Håkansson and Glimelius, 1991; Kitagawa et al., 2003; Leino et al., 2005), and different CMS phenotypes have been associated with certain open reading frames (orfs) composed of novel sequences of unknown origin combined with sequences of known mt genes. The expression of the CMS-associated mt orfs can often be reverted and the male sterile trait restored to fertility by nuclear restorer gene(s) (reviewed by Hanson and Bentolila, 2004). This demonstrates that interactions between nuclear and mt genomes are involved in the regulation of male fertility and flower development. As CMS clearly includes a mt influence on the nuclear gene expression, CMS systems could be used to study retrograde signalling in plants. The mt retrograde signalling pathway has mostly been studied in yeast and mammals.
Besides inhibited pollen production, the CMS lines often display additional alterations in flower development such as abnormal floral organs (reviewed by Hanson and Bentolila, 2004; Kofer et al., 1991). For example, in the Brassica napus CMS line analysed in this study, the morphology of flower buds of the CMS line is similar to that of the wild type during early flower development, but during later stages the stamens are replaced by carpelloid organs with ovule-like structures found at the internal margins of the unfused carpelloid structures (Leino et al., 2003; Teixeira et al., 2005a). Similar homeotic modifications have also been found in other CMS systems such as Nicotiana tabacum (Farbos et al., 2001; Kofer et al., 1991; Zubko et al., 1996), Daucus carota (Linke et al., 1999, 2003), and Triticum aestivum (Murai et al., 2002; Ogihara et al., 1997). Notably the homeotic conversions of the third-whorl organs observed in these CMS systems show striking similarities with the third-whorl organ phenotypes of Arabidopsis thaliana mutants affected in the B-class genes APETALA3 (AP3) and PISTILLATA (PI) (Bowman et al., 1989, 1991), which are involved in the specification of petals and stamens. This suggests that the regulation of the B-class genes, or of other components in the same pathway, is disturbed in many CMS systems. Studies in CMS lines of N. tabacum, D. carota, T. aestivum and B. napus have indeed found correlations between a downregulation of B-class genes and the CMS phenotype (Linke et al., 2003; Murai et al., 2002; Teixeira et al., 2005a; Zubko et al., 2001). In spite of this progress, our understanding of the function of nuclear genes in developing CMS flowers is currently limited.
To gain new insights into the cause of aberrant flower development in CMS lines, we analysed nuclear gene expression during flower formation in a B. napus CMS system on a global scale using A. thaliana flower-specific cDNA microarrays. The B. napus CMS line we used for this analysis contains a B. napus nuclear genome, and a rearranged mt genome mainly consisting of A. thaliana DNA (Leino et al., 2003). The fertile maintainer line B. napus cv. Hanna was used as a reference. As flower development in wild-type B. napus is similar to that of A. thaliana, we defined floral stages in accordance with the staging system described for A. thaliana flowers by Smyth et al. (1990). The genera of Brassica and Arabidopsis are closely related, and A. thaliana and B. napus share about 85% exon sequence similarity (Cavell et al., 1998). Because previous studies have shown that A. thaliana microarrays can be successfully used for the analysis of gene expression in Brassica species (Girke et al., 2000; Lee et al., 2004), we assumed that the genes detected by the microarrays represent the B. napus homolog of the A. thaliana genes. For simplicity, we use A. thaliana gene annotations throughout this study.
Results and discussion
Early flower development in B. napus and its corresponding CMS line
For this study, we aimed to compare the nuclear transcriptome of B. napus flowers with that of a B. napus CMS line to detect differences in gene expression resulting from the rearranged mt genome of the CMS line. To this end, we followed and characterized flower development of both lines in detail by scanning electron microscopy (SEM) (Figure S1). No phenotypic differences were observed between the CMS line and fertile B. napus in the early stages (stages 0–5), whereas homeotically transformed third-whorl organs were readily recognized in stage 8 CMS flowers in accordance with Teixeira et al. (2005a,b).
The altered nuclear transcriptome of the CMS line is concordant with the CMS phenotype
As young flower buds of the CMS line are morphologically similar to those of B. napus (Figure S1 and Teixeira et al., 2005a), we reasoned that an analysis of the nuclear gene expression profiles of early stage flowers would allow us to trace the initiation of gene expression alterations, which lead to the CMS phenotype and are affected by the rearranged mt genome of the CMS line, and subsequently affect flower development. We therefore collected tissue samples from the CMS line, as well as from the B. napus maintainer line, representing young floral buds of different developmental stages. The first tissue sample contained flower buds of up to stage 5, as well as the inflorescence meristem (hereafter referred to as stages 0–5). For the second sample we collected flower buds of stage 6 and 7, and for the third sample buds of stage 8 were pooled. We then compared the gene expression profiles of these samples by microarray analysis using the experimental design depicted in Figure S1. Gene expression was compared between lines at different stages of development, as well as between developmental stages within lines (Figure S1). Dye-labelled RNA populations from the individual tissue samples were co-hybridized to cDNA microarrays, the elements of which were derived from transcripts that are enriched in A. thaliana flowers. These microarrays contain 10 816 elements corresponding to about 5000–6000 unique genes (Wellmer et al., 2004).
Our experiments identified a large number of genes that were differentially expressed in the CMS line in comparison to B. napus, which shows that the presence of A. thaliana recombined mtDNA in the CMS line strongly influences nuclear gene expression during flower development. Of the 10 816 microarray elements, 1350 displayed significant expression differences between the CMS line and B. napus at either one or more floral stages (false discovery rate, FDR-corrected P < 0.001; Table S1). These elements corresponded to 244 unique genes (see Experimental procedures and Table S2).
The number of differently expressed genes increased with progressing developmental stages. At stages 0–5, we detected 29 genes as differentially expressed (five with higher expression and 24 with lower expression in the CMS line compared with B. napus). At stages 6–7, 93 genes showed significant expression changes (53 with higher expression and 40 with lower expression). Finally, at stage 8, when phenotypic differences between the flowers of B. napus and the CMS line were clearly visible (see above), 208 genes were differentially expressed (104 with higher expression and 104 with lower expression).
Cluster analysis of the microarray data for the differentially expressed genes revealed five groups of genes with correlated expression profiles (Figure 1c). Genes in clusters I, II and IV displayed a significantly reduced expression in the CMS line compared with B. napus at one or more floral stages. Genes in cluster I were predominantly repressed during stages 0–5, whereas clusters II and IV comprised genes the expression of which showed the strongest reduction at later stages of flower development (Table S2). In contrast to clusters I, II and IV, cluster III contained genes with a higher expression in the CMS line during stages 6–7 and at stage 8. Cluster V was composed of only two sequences, which showed a higher expression in flowers of the CMS line compared with that of B. napus in all samples analysed.
Within the groups of co-expressed genes we found several with either known or presumed functions during the development of stamens and carpels, respectively. For example, 11 of the 109 genes in cluster III (Table S2) are either specifically or predominantly expressed in carpels according to a previous study (Wellmer et al., 2004). These genes include the floral regulators SHATTERPROOF 1 and 2 (Ma et al., 1991), SEEDSTICK (Rounsley et al., 1995) and FRUITFULL (Mandel and Yanofsky, 1995). Upregulation of these genes in flowers of the CMS line is a likely consequence of the transformation of stamens into carpelloid organs.
In contrast, genes expressed in stamens were strongly downregulated in the CMS line. Clusters I, II and IV comprised a total of 133 genes, of which 33 have been previously predicted (Wellmer et al., 2004) as being predominantly expressed during stamen and/or pollen development (Table S2). However, we also found genes in these clusters that are more broadly expressed during flower development. For example, NOZZLE/SPOROCYTLESS (NZZ/SPL), a key regulator of sporogenesis (Schiefthaler et al., 1999; Yang et al., 1999) (cluster IV), is expressed in stamens and carpels, whereas the floral organ identity genes AP3 and PI (cluster II) are expressed in the petals and stamens of wild-type flowers (see above). Taken together the results of this analysis suggest that the majority of the differentially expressed genes identified in the experiment are likely to be involved in stamen and carpel development, and that most of the observed expression changes are a consequence of the homeotic transformations in the CMS line.
Genes implicated in cell-wall remodelling display a reduced expression in the CMS line when stamen primordia emerge
Genes in cluster I displayed a significantly lower expression in the CMS line compared with B. napus at stages 0–5 (see above). However, only a few of them also showed significant expression differences at later floral stages. Most of the genes in this cluster have been predicted as predominantly expressed in stamens and/or pollen by previous studies (Honys and Twell, 2003; Wellmer et al., 2004; Zimmermann et al., 2004). Among these genes we found several that encode members of the families of pectinesterases, multicopper oxidases (of type 1) and glycoside hydrolases (of family 28). Limited functional data are available for individual members of these protein families, but they are generally thought to be involved in cell-wall modifications (Micheli, 2001; Torki et al., 1999). Interestingly, the differentially expressed members of these large gene families generally showed a close evolutionary relationship (Figures S2, S3 and S4), which may indicate that they are active during similar cellular processes. In the pectinesterase and multicopper oxidase gene families we also found individual genes that showed a close phylogenetic relationship and had up to 80% sequence identity with differentially expressed genes, but themselves did not differ significantly between the CMS line and fertile B. napus. Hence, in our microarray experiments, we could discriminate between genes with up to 80% sequence identity. As A. thaliana and B. napus have an estimated 85% sequence identity in the ORF regions (Cavell et al., 1998), the risk of wrongly assigning genes as being differentially expressed because of cross-hybridization is very limited.
In A. thaliana, four archesporial cells appear soon after stamen primordia emerge at stage 5 of flower development (Sanders et al., 1999), and these cells undergo divisions to form microsporangia (Scott et al., 2004). At flower stage 7, the archesporial cells have divided to form parietal and sporogenous cell lineages. The reduced expression of several genes implicated in either cell-wall remodelling or architecture suggests that early division events of archesporial cells may be disturbed in the CMS line. In our study, no obvious morphological differences between stage-5 buds from the CMS line and the B. napus cultivar were observed by SEM analysis. However, cell patterning defects in areas where the initiation of anther primordia takes place have recently been described for CMS-line flowers based on the results of detailed histological studies (Teixeira et al., 2005a).
Genes implicated in either energy production or protein metabolism are downregulated in the CMS line
To further characterize the differentially expressed genes identified in our experiment we analysed their assignment to functional categories using Gene Ontology annotations (Table S2). We found that none of the functional categories analysed were significantly enriched in the dataset compared with their distribution among all genes represented on the microarray. However, when we compared the up- and downregulated genes of the groups, we detected an over-representation of genes that were assigned to one of five categories (Figure 1a, b): genes involved in Cellular Transport and Transport Mechanisms, or in Metabolism were enriched among genes upregulated in the CMS line, whereas downregulated genes showed an overrepresentation for the categories Energy, Protein Fate and Protein Synthesis, respectively.
The thirteen genes assigned to the category Energy encode proteins involved in glycolysis, in the citric acid cycle or are components of electron transport chains. Ten of these genes had a significantly lower expression in the CMS line at stage 8 of flower development compared with B. napus. Downregulation of genes involved in the supply of energy is in agreement with the results of previous studies that reported reduced ATP levels in the CMS lines of B. napus (Teixeira et al., 2005b) and N. tabacum (Bergman et al., 2000), respectively. In fact, one hypothesis that has been put forward to explain the floral phenotypes of CMS lines is based on the idea that an increased demand for respiratory function and energy equivalents during flower development cannot be provided for by the abnormal mitochondria of a CMS line (reviewed by Hanson and Bentolila, 2004; reviewed by Linke and Börner, 2005; Tadege and Kuhlemeier, 1997). The downregulation of genes involved in energy supply that we detected at around the developmental stage when phenotypic alterations in flowers of the CMS line occur is in agreement with this hypothesis. Taken together, our data suggest a reduced energy supply caused by a repression of genes involved in the production of energy equivalents.
Expression of nuclear genes encoding proteins targeted to the mitochondria is reduced in the CMS line
Among the genes we identified in the experiment as differentially expressed, and that are involved in the supply of energy, we found several that were predicted as targeted to the mitochondria. This result suggested an influence of the abnormal mitochondria of the CMS line on the expression of nuclear genes that encode mt proteins. To further investigate this, we searched the dataset for genes that are likely to encode mt proteins and identified a total of 25 among the 244 differentially expressed genes (i.e. ∼10%). For most of the corresponding proteins (19 of 25), mt localization has been demonstrated experimentally (Table S2). Twenty-two of the identified genes displayed a lower expression in the CMS line compared with B. napus, indicating that the expression of nuclear genes encoding mt proteins is overall repressed in the CMS line.
It has been proposed that the developmental aberrations observed in CMS lines are a consequence of a disrupted cross-talk, or retrograde signalling, between the abnormal mitochondria and the nucleus, and that this defect may result in altered mt and nuclear gene expression (Geddy et al., 2005; reviewed by Mackenzie and McIntosh, 1999). In fact, such transcriptional changes have been described for several genes in different CMS systems (reviewed by Hanson and Bentolila, 2004). In addition, it has been shown that in leaves treated with the electron transport inhibitor anti-mycin A, nuclear gene expression is strongly affected (Yu et al., 2001). The results of our transcriptome analysis also show a marked influence on nuclear gene expression, and thus are in agreement with the idea of a disrupted communication between mitochondria and the nucleus in CMS systems. Furthermore, they suggest that a primary effect of the presence of a recombined mt genome in the CMS line may be a repression of nuclear genes that encode mt proteins.
Among these repressed genes we found several that encode proteins involved in protein import into organelles. These included the mitochondrial processing peptidases (α-MPP) (At1g51980) and β-MPP (At3g02090), which remove N-terminal mt targeting signals (reviewed by Glaser and Dessi, 1999; Lister et al., 2004), as well as the translocase of the outer mt membrane, TOM40 (At3g20000; Lister et al., 2004). Downregulation of these essential genes suggests that either the import or the processing of nuclear encoded mt proteins might be affected in the CMS line. These defects, combined with the changes in nuclear gene expression we have detected in the experiment, may result in a severely altered mt proteome. This, in turn, may lead to a reduced production of ATP (see above) and to the abnormal ultrastructure that has been previously reported for the mitochondria of the CMS line (Teixeira et al., 2005a).
Two mitochondrial transcripts are highly expressed in the CMS line
Cluster I is composed of two microarray elements that showed a higher expression in the CMS line in comparison with B. napus throughout flower development (Figure 1c). These elements are most likely hybridized to transcripts encoded by the mt, and not the nuclear genome, as the ATPase subunit 9 (atp9) is a gene known to be encoded by the mitochondria, and element GP001E3 has been shown to be expressed in the mitochondria (Holec et al., 2006; Leino et al., 2005). Because our labelling method was based on poly-dT priming, detection of these mt transcripts in the microarray experiments suggests that they might have been tagged with a poly-A tail by the mt transcription machinery in order to be degraded by polynucleotide phosphorylase (PNPase) (Holec et al., 2006). The transcripts we identified were AtMg01080 (atp9) and a sequence located upstream in the 5′-UTR of orf139 (bases 47 130–47 518 and 181 295–181 683 in the mt genome, matching a 500-nucleotide sequence expressed in A. thaliana (Holec et al., 2006)), here named GP001E3 after the designation of the corresponding microarray element (Lin et al., 1999; Unseld et al., 1997). GP001E3 is specific to A. thaliana and is not present in either the mt or the nuclear genome of B. napus (Handa, 2003; Leino et al., 2005). Thus, GP001E3 expression in the CMS line originates from A. thaliana mtDNA. Using circular RT-PCR, it has been previously shown that GP001E3 is co-transcribed with orf139 (Leino et al., 2005). To test whether GP001E3 is expressed throughout developing CMS line flowers or in a specific pattern, we performed in situ hybridization experiments with a GP001E3 anti-sense probe and detected expression in developing ovules and in the marginal regions of carpelloid stamens (Figure 2a). Notably, it is from these marginal tissues that ovules are sometimes formed in the carpelloid third-whorl organs of the CMS line. For orf139 we found an expression pattern similar to that of GP001E3 (Figure 2b), in agreement with the reported co-transcription of these genes (see above). Thus, both GP001E3 and orf139 are expressed in a specific pattern during flower development. Because floral organ-specific patterns of mt transcripts as a result of high mt number have been previously reported (de la Canal et al., 2001; Geddy et al., 2005; Huang et al., 1994; Smart et al., 1994), we tested whether the expression of GP001E3 and of orf139 correlates with that of nad6 (AtMg00270), an essential mt gene that encodes NADH dehydrogenase subunit 6 (Unseld et al., 1997). As for GP001E3 and orf139, we detected nad6 expression in ovules and marginal regions of carpelloid stamens (Figure 2c). This result suggests that the specific expression of these mt transcripts in flowers of the CMS line is indeed the result of a high mt activity in rapidly dividing tissues.
AP3 and PI expression is reduced in the CMS line and is restricted to whorl 2 at later floral stages
Flowers of the CMS line show a phenotype that partly resembles that of the A. thaliana B-class mutants AP3 and PI. In fact, a reduction of AP3 and PI expression in flowers of the CMS line has been previously demonstrated (Teixeira et al., 2005a). The results of our microarray experiments suggested that differences in AP3 and PI expression between the CMS line and B. napus are more pronounced in the later stages of floral development. To verify these results, we determined the AP3 and PI transcript levels in B. napus and CMS-line flowers by quantitative real-time RT-PCR (qRT-PCR). We found reduced expression levels for these genes in the CMS line compared with B. napus during stages 6–7 and 8, but not at stages 0–5 (Figure 3), in agreement with our microarray results.
To determine whether the downregulation of AP3 and PI expression in flowers of the CMS line is a consequence of either reduced expression levels or reduced expression domains, or both, we performed in situ hybridization experiments for AP3 and PI using sections of B. napus and CMS-line inflorescences. In B. napus, we detected AP3 expression in stage-3 flowers in cells that will give rise to petal and stamen primordia. Expression in these organs remained high throughout flower development (Figure 4a–b), in agreement with the previously reported expression pattern for AP3 in both A. thaliana (Jack et al., 1992) and B. napus (Teixeira et al., 2005a). In stage-3 flowers of the CMS line, AP3 transcripts were detected, as in B. napus, in whorls 2 and 3 (Figure 3d; see also Teixeira et al., 2005a). However, at stage 5, AP3 expression was excluded from third-whorl-organ primordia (Figure 4e), whereas expression in petals appeared to be unaffected (Figure 4b). In later stage flowers, AP3 expression remained restricted to petals and did not reappear in third-whorl organs (Figure 4f). Notably, downregulation of AP3 in the third-whorl coincided with the aberrant cell division found in flowers of the CMS line (Teixeira et al., 2005a).
In corresponding experiments with a PI anti-sense probe, we detected expression in second- and third-whorl organs of stage-3 B. napus and CMS-line flowers (Figure 4g,j; see also Teixeira et al., 2005a), in a pattern that is similar to that of AP3. PI expression was restricted to second-whorl organs in older flowers of the CMS line. However, compared with AP3, downregulation of PI expression appeared to be delayed, as PI transcripts were readily detectable in third-whorl primordia of stage-5 CMS flowers (Figure 4k). The PI hybridization signal in third-whorl organs of CMS flowers diminished after stage 7 and could not be detected after stage 10 (Figure 4l).
In A. thaliana, the regulation of AP3 and PI expression can be subdivided into two distinct phases: an activation phase and a maintenance phase (Jenik and Irish, 2001). During the activation phase, which occurs during floral stages 3–6, AP3 and PI expression is induced by the combined activity of the floral meristem identity genes APETALA1 (AP1) and LEAFY (LFY) (Busch et al., 1999; Ng and Yanofsky, 2001; Sessions et al., 2000). During the maintenance phase, from floral stage 6 onward, AP3 and PI expression is regulated by an auto-regulatory feedback loop. Thus, expression of both genes is dependent on their own gene products. Provided that A. thaliana and B. napus are similar with respect to the regulation of AP3 and PI expression, the results of our in situ hybridization experiments suggest a downregulation of AP3 in third-whorl organs of CMS-line flowers during the initiation phase, and a subsequent repression of PI during the maintenance phase. Downregulation of PI is therefore a likely consequence of reduced AP3 activity.
It has been suggested based on the results of qRT-PCR analyses that apart from AP3 and PI, two upstream regulators of AP3 and PI expression, namely LFY (Schultz and Haughn, 1991) and UNUSUAL FLOWER ORGANS (UFO; Wilkinson and Haughn, 1995), are upregulated in the CMS line as compared with B. napus (Teixeira et al., 2005a). We did not detect significant expression changes for these genes in our study, which may be a result of differences in sensitivity between the two methods used. In any case, the reported upregulation of the activators LFY and UFO in CMS-line flowers cannot be readily reconciled with a reduction in AP3 transcript levels. Thus, our data strongly suggest that AP3 activity in stamens is regulated by an as yet unknown whorl-specific factor that is dependent, either directly or indirectly, on proper mt function.
The CMS-line 4:19 used for this study was derived from protoplast fusions between B. napus cv. Hanna and A. thaliana accession Landsberg erecta (Forsberg et al., 1998). The original somatic hybrids were backcrossed to the B. napus cultivar. Several recent publications provide a detailed characterization of the CMS line (Leino et al., 2003, 2005; Teixeira et al., 2005a,b). We used the parental and fertile B. napus cultivar as a reference for our experiments. The fertile line and the CMS line are isogenic with respect to the nuclear genome, but have different mt genomes. The mt genome of the fertile line is pure B. napus, whereas the CMS mitochondria consists of a rearranged genome from both B. napus and A. thaliana (Leino et al., 2003, 2005). Plants were grown in a soil:perlite mixture under the following conditions: 16-h photoperiod (light intensity 400 μmol m−2sec); 85% air humidity; 22°C day/18°C night temperatures.
Inflorescences containing flower buds up to stage 12 (Smyth et al., 1990) from B. napus and the CMS line were fixed with 2.5% glutaraldehyde in 0.05 m phosphate buffer, pH 6.8 (buffer A) at room temperature for 2 h, rinsed in buffer A, and then post-fixed in 2% osmium tetroxide (in buffer A) for 2 h at room temperature. After three rinses with buffer A, the material was dehydrated in a graded ethanol series (20%, 40%, 60%, 80%, 95% and 100%). The material was then subjected to critical-point drying using CO2. The inflorescences were mounted on stubs and dissected using glass needles. After mounting the different floral organs, the material was coated with gold and analysed with a JSM-6320F scanning microscope (JEOL, Tokyo, Japan).
Isolation of flower buds
Flower buds were isolated from B. napus, as well as from the CMS line, and sorted according to their developmental stage. The individual tissue samples contained flower buds of stages 0–5 (including the inflorescence meristem), 6–7 and 8, respectively. Stages were defined in accordance with the staging system described for A. thaliana flower development (Smyth et al., 1990). Flower buds were collected from at least two different biological repetitions within each line, from between two and four inflorescences on each plant, and from between two and four different plants from each line. Different flower bud samples were collected at approximately the same time of the day to avoid diurnal effects on gene expression.
Total RNA from flower-bud samples was isolated using the Qiagen RNeasy Plant Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. mRNA populations from total RNA samples were amplified by in vitro transcription using T7 RNA polymerase (Ambion, Austin, TX, USA), as previously described (Wellmer et al., 2004).
The CMS line was compared with B. napus at three different stages of development: 0–5, 6–7 and 8 (Figure S1). For each stage, two biological repetitions were used, each with two experimental repetitions, resulting in four hybridizations per developmental stage. The developmental stages were also compared between B. napus and the CMS line (Figure S1). For each comparison, we performed two hybridizations using RNA from independent biological samples. In all replicate experiments, the dyes used for RNA-labelling were switched to reduce dye-related artefacts.
The microarrays used in this study contained 10 816 elements, which were derived from A. thaliana flower-specific cDNA libraries. These elements represent approximately 5000–6000 unique genes. For further information about the flower-specific cDNA microarray, see Wellmer et al. (2004).
We sequenced several (in total 89) of the microarray elements that had not been previously annotated and that reported differential expression in our experiment. We found that the majority (68 of 89) of these elements represent genes, for which probes had been previously identified on the cDNA microarray. The remaining 21 elements corresponded to 11 unique genes. Thus, we estimate that the array elements that were reported as being differentially expressed and remained unsequenced represent ∼110 additional genes.
Probe labelling and microarray hybridization
The protocols used for probe labelling and hybridization are listed in Appendix S1. In short, first- and second-strand cDNA was synthesized from 5 μg of total RNA using a poly(dT)-primer containing a T7 promoter sequence. Subsequently, in vitro transcription was performed using the Megascript T7 kit (Ambion, Austin, TX, USA). Dye-labelled cDNA was generated from 5 μg of amplified mRNA using aminoallyl-modified nucleotides and chemical coupling of cyanine 3 (Cy3) and cyanine 5 (Cy5) fluorescent dyes (Amersham Biosciences, Little Chalfont, UK), respectively. Labelled cDNA was hybridized to microarrays at 50°C for ca. 16 h.
Data acquisition, normalization and statistical analysis
Hybridized arrays were scanned with an Axon 4000B scanner using the GenePix 3.1 analysis software (Axon Instruments, Foster City, CA, USA). The GenePix 3.1 software was also used to quantify spot intensities, as well as the local background intensities. All subsequent data analyses were performed in R (http://www.r-project.org) using marray and limma in the open-source Bioconductor project (http://bioconductor.org). Data points were removed either if a spot was flagged during data acquisition (e.g. as a ‘bad spot’, ‘absent spot’, etc.) or if the spot intensity was below a threshold determined by a model-based statistical method, which is based on the standard deviation of background differences between a spot and the neighbouring spots (Yang et al., 2001). Spot intensities were then corrected for the background using the normexp option (offset = 50) in limma. The data were normalized using within-printtip-group intensity dependent normalization with the marray package. A linear model was fitted for each gene using the limma package (Smyth, 2004), and contrasts for differential expression between CMS and B. napus at stages 0–5, 6–7 and 8 were estimated. Differentially expressed genes were identified using the moderated t-statistic in limma and correcting for multiple testing using FDR (Benjamini and Hochberg, 1995). The test was performed in a hierarchical fashion, first at the overall gene level and then at the contrast level. We first used an FDR-adjusted P-value of 0.001 at the overall gene level. The largest P-value among the selected genes (0.00019) was then used as the FDR-adjusted cut-off for significance at each of the three floral stages analysed.
Quantitative real-time RT-PCR
RNA extracted from flower buds at different stages (0–5, 6–7 and 8) from two biological samples were used for qRT-PCR of AP3 and PI. cDNA was synthesized using Superscript III RNase-Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions. qRT-PCR reactions were carried out using an ABI Prism 7000 Sequence Detector in MicroAmp Optical 96-well reaction plates with optical covers, according to the manufacturer's instructions (Applied Biosystems, Foster City, CA, USA). PCR reactions (final volume of 25 μl) were performed containing gene-specific primers (Teixeira et al., 2005a) and the passive reference dye ROX (a glycine conjugate of 5-carboxy-x-rhodamine with succinimidyl ester) (Applied Biosystems), in order to normalize fluorescence across the plate. Reaction conditions were as follows: 50°C for 2 min, 94°C for 10 min, followed by 40 cycles of 94°C for 15 sec, and 60°C for 1 min. To avoid the amplification of genomic DNA contaminations, primers were designed so that they would flank introns using the software Primer Express (Applied Biosystems). Relative quantification values and standard deviations were calculated using the comparitive CT method (ΔΔ - CT) according to the manufacturer's instructions (User Bulletin #2; ABI Prism 7700 Sequence Detection system, updated 10/2001, Applied Biosystems). Values were normalized to the expression of the reference actin, as well as the calibrator B. napus. Differences in expression were identified using the Student's t-test.
In situ hybridization
All tissues were fixed in 4% paraformaldehyde and embedded in 100% Paraplast plus (Sigma, St Louis, MO, USA). Sections (8-μm thick) were fixed to Probe-on Plus slides (Fischer Scientific, Pittsburgh, PA, USA) at 42°C and hybridized with digoxygenin-UTP-labelled (Boehringer Mannheim Corp., Indianapolis, IN, USA) probes, which were generated by in vitro transcription using either SP6 or T7 RNA polymerase. In situ pre-hybridization, hybridization and detection were performed as previously described (Carr and Irish, 1997).
To sort the differentially expressed genes in our dataset into functional categories, and to identify genes encoding proteins targeted for the mitochondria, we used the ‘Arabidopsis Mitochondrial Protein Database’ (AMPDB; http://www.ampdb.bcs.uwa.edu.au/; Heazlewood and Millar, 2005; Heazlewood et al., 2004). Default settings were used to retrieve the functional category for each gene. Using the settings PAGE confirmed [results generated from a blue native (BN-PAGE) study on mitochondria], LCMS confirmed [results from a liquid chromatography-tandem mass spectrometry (LC-MS/MS) study on mitochondria] and default settings, we identified genes that were experimentally validated with PAGE and/or LCMS to encode mt proteins. Genes encoding proteins putatively targeted to mitochondria were identified using targetp (Emanuelsson et al., 2000), mitoprot (Claros and Vincens, 1996), ipsort (Bannai et al., 2002) and Predotar (Small et al., 2004) (four different prediction programs for mitochondrially targeted proteins), and genes identified by three or more of the predicting programs were selected. Differences between functional categories were identified using the chi-squared test.
We thank I. Eriksson, G. Rönnqvist and S. Thyselius for their excellent technical assistance and Dr M. Leino for valuable discussions. This work was supported by the strategic research programme Agricultural Functional Genomics (AgriFunGen) at the Swedish University of Agricultural Sciences, as well as The Swedish Research Council (VR), The Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS) and The Swedish Foundation for International Cooperation in Research and Higher Education (STINT).