Maternally inherited defects in the formation of male flower organs leading to cytoplasmic male sterility (CMS) indicate an involvement of mitochondrial genes in the control of flower formation. In the ‘carpeloid’ CMS type of carrot, stamens are replaced by carpels. The florets thus resemble well-investigated homeotic flower mutants of Arabidopsis and Antirrhinum, in which organ identity is impaired because of the mutation of specific nuclear MADS box genes. We have isolated five cDNAs encoding MADS box proteins (DcMADS1–5) from a flower-specific library of carrot. Structural features deduced from their sequence and transcript patterns in unmodified carrot flowers determined by in situ hybridisation relate them to known MADS box transcription factors involved in specification of flower organs. In ‘carpeloid’ CMS flowers, we detected a distinctly reduced expression of DcMADS2 and DcMADS3, homologues of the Antirrhinum genes GLOBOSA and DEFICIENS. Our data strongly suggest that the ‘carpeloid’ CMS phenotype is caused by a cytoplasmic (mitochondrial) effect on the expression of two MADS box factors specifying organ development at whorls 2 and 3 of carrot flowers.
Flower development in plants is mainly controlled by the nuclear genome. The well-known phenomenon of cytoplasmic male sterility (CMS; Kaul, 1988), however, indicates a participation of extra-nuclear genes in flower formation. CMS plants are unable to produce fertile pollen and hence are male sterile. The maternal inheritance of the CMS trait clearly indicates that the genetic information of either the mitochondria or the plastids participates in certain steps of flower formation. Mitochondrial genes have been identified as being causally involved in all well-investigated CMS systems. In many cases of CMS, rearrangements in the mitochondrial genome were found to create unique chimeric genes, and often function of genes encoding subunits of the mitochondrial ATP synthase is impaired, suggesting that reduced mitochondrial respiration and/or ATP production lead to male sterility (reviewed by Budar and Pelletier, 2001; Mackenzie and McIntosh, 1999; Schnable and Wise, 1998). The mitochondrial dysfunction of CMS plants can be either maintained or suppressed by specific nuclear gene functions leading, in the latter case, to restored male fertility. Genetic and molecular mechanisms of restoration vary among the different CMS systems (reviewed by Schnable and Wise, 1998). Elimination of CMS-related mitochondrial DNA segments (Janska et al., 1998), restoration of effective editing of specific mitochondrial mRNAs (Howad and Kempken, 1997) as well as post-transcriptional decrease of aberrant mitochondrial gene products (e.g. Gagliardi and Leaver, 1999; Menassa et al., 1999) and compensatory effects at the metabolic level (Cui et al., 1996; Liu et al., 2001) have been described. Tissue-specific differences in mitochondrial gene expression and in energy demands are suggested explanations for the specific effects of mitochondrial dysfunction on male flower organs and pollen formation, respectively (e.g. Bergman et al., 2000; Conley and Hanson, 1995; Ducos et al., 2001). Consequently, the mitochondria have to interfere with the function of certain, yet unidentified nuclear genes during flower formation.
Similar homeotic organ alterations affecting two neighbouring flower whorls are known from well-investigated nuclear mutants of Antirrhinum and Arabidopsis (Bowman et al., 1989; Schwarz-Sommer et al., 1990). Their detailed investigation led to the detection of MADS box transcription factors with major functions in flower development and to a model explaining floral patterning (Weigel and Meyerowitz, 1994). The model proposes three gene classes, A–C, that specify identity of the four flower organs: sepals, petals, stamens and carpels. It has recently been extended by inclusion of class D- and E-genes (Theissen, 2001). The A-function determines sepal development, whereas concerted expressions of A, B and E, and of B, C and E specify petals and stamens, respectively. The C- and E-functions specify development of carpels; class D-genes are needed for ovule formation. Most of these genes encode transcription factors of the MADS box gene family sharing a conserved 58 amino acid DNA-binding domain (Schwarz-Sommer et al., 1990).
The aim of the present study was to find out if MADS box genes are involved in the expression of a certain CMS phenotype and to identify nuclear genes that are affected by disturbed mitochondrial–nuclear interactions that lead to male sterility. We report here on the identification of five carrot MADS box genes, DcMADS1–5, involved in flower formation. To understand their roles in organ development, we characterised temporal and spatial transcript patterns in unmodified and CMS flowers of carrots. In the ‘carpeloid’ CMS type, we observed altered transcript patterns of DcMADS2 and DcMADS3, the carrot homologues of genes with B-function. Our findings indicate a specific cytoplasmic (mitochondrial) influence on the expression of MADS box genes for B-activity during early differentiation of petals and stamens leading to the ‘carpeloid’ CMS phenotype.
Phenotype and histology of ‘carpeloid’ and ‘petaloid’ CMS flowers
Stamens of unmodified carrot flowers (Figure 1a) are replaced by carpels in the ‘carpeloid’ CMS type (Figure 1g) or by petals in ‘petaloid’ CMS florets (Figure 1m). Additionally, the petals of both CMS types have sepaloid features because at least midribs or whole organs are greenish. To identify the stage at which organ development of homeotic CMS flowers starts to deviate, we compared stained longitudinal sections by light microscopy in unmodified carrot (Figure 1a–f) as well as in ‘carpeloid’ (Figure 1g–l) and in ‘petaloid’ CMS types (Figure 1m–r). Floral organogenesis of carrot was subdivided into seven stages as described previously (Linke et al., 1999). They include formation of single floret primordia (stages s0–s3), establishment of organ primordia until stage s5 and beginning organ differentiation during stages s6 and s7. Separation of single florets (Figure 1b; arrows), out-forming of whole floret primordia (stages s2–s3) and establishment of organ primordia (stage s5) were identical in all three flower types. Development of ‘carpeloid’ (Figure 1h,i) and ‘petaloid’ florets (Figure 1n,o) remained unchanged until organ primordia in whorl 3 emerged (late s5, Figure 1c,i,o). At early stage s6, the ‘carpeloid’ floret base was found to be more extended than that in unmodified flowers (Figure 1d; st). The primordia of whorl 3 were still globular but slightly broadened (Figure 1j; w3c).
At the same stage s6, the third-whorl primordia of ‘petaloid’ CMS florets are narrower, but erected, and slightly elongated. Indentations at the primordia bases normally introducing filament formation (Figure 1d; arrow) are missing. At stage s7, when the female organ is formed in the centre (Figure 1e,c), unmodified stamens become distinguishable into filaments and anthers (Figure 1e). At this stage, ‘carpeloid’ flowers exhibit bicarpellary structures at this position (Figure 1k; w3c) while ‘petaloid’ florets show no morphological difference between genuine petals and elongating third-whorl organs (Figure 1q,r). Unmodified stamens of mature florets are completely differentiated into filaments and the globular pollen sacs (Figure 1f). Mature ‘carpeloid’ flowers show a reduced size of sepaloid petals, and styles of homeotic carpels are elongated (Figure 1l). Thus, ‘carpeloid’ and ‘petaloid’ CMS florets indicate an aberrant development of third-whorl organs from early s6 onwards.
Cloning of DcMADS1–5 and expression in unmodified flowers
A cDNA library of unmodified, male fertile carrot flowers was established and screened with gene probes from the Antirrhinum MADS factors GLOBOSA, DEFICIENS and PLENA. Five MADS box homologous carrot clones, designated as DcMADS1–5 (Dc, D. carota), were identified (EMBL accession numbers AJ271147–AJ271151). They encoded putative proteins of 207–255 amino acids containing the highly conserved MADS domain (Figure 2) with a potential calmodulin-dependent phosphorylation site (Schwarz-Sommer et al., 1990). Multiple alignments of the MADS domains of DcMADS1–5 with putative equivalents of other plant species were performed (Figure 2). Analysis of conserved amino acid positions enabled assignment of the carrot clones to five subgroups of the MADS box family as defined by Theissen and Saedler (1995). Thereby, DcMADS1 was assigned to the SQUAMOSA (SQUA) group, DcMADS2 to the GLOBOSA (GLO) group, DcMADS3 to the DEFICIENS (DEF) group, DcMADS4 to the AGAMOUS (AG) group and DcMADS5 to the SEPALLATA group (former AGL2-subfamily, Ma et al., 1991). As mediators of organ specification, the SEPALLATA genes belong to class E-factors of the advanced ABCDE model (Theissen, 2001). Sequence comparison of subdomains of the deduced proteins (M-I-K-C-regions) between DcMADS5 and SEPALLATA1–3 of Arabidopsis (former AGL2, AGL4 and AGL9) indicates high similarities to SEPALLATA1 (see Tables S1–S5).
Table S1. Percentage of nucleotide identity in the MADS box encoding DNA region of Daucus compared to Antirrhinum (and Arabidopsis)
Table S2. Percentage of amino acid identity between deduced MADS box proteins of Daucus and of Antirrhinum (and Arabidopsis)
Table S3. Percentage of amino acid identity/similarity between MADS domain, I-region and K-box (M-I-K) of Daucus and Antirrhinum (values for Arabidopsis are shown in parentheses)
Table S4. Nucleotide identity between DcMADS5 and the SEPALLATA- and other relative AGL-like genes (AGL3, AGL6)
Table S5. Percentage of amino acid identity/similarity between DcMADS5 and SEPALLATA1–3 (former AGL2, AGL4 and AGL9) of Arabidopsis (values for further relatives AGL3 and AGL6 are included)
Northern analysis and RT-PCR of DcMADS1–5 transcripts revealed expression in unmodified male fertile carrot and in the CMS-flower phenotypes, but not in leaf or root tissue (not shown). For detailed analysis of expression in different tissues of early and advanced floret stages, in situ hybridisation studies were performed. Figure 3 shows the temporal and spatial transcript patterns from initiation until early differentiation of organ primordia, i.e. from stages s3 to s7. At stage s1, transcripts of a carrot gene homologous to LEAFY of Arabidopsis and FLORICAULA of Antirrhinum were observed (B. Linke and M. Szklarczyk, unpublished data); however, none of the DcMADS genes showed detectable transcription. Among the MADS genes, DcMADS5 and DcMADS1 were the ones showing transcript accumulation earliest in flower formation (stage s2) followed by DcMADS3 and somewhat later by DcMADS2 and DcMADS4. At stage s2, transcripts of DcMADS1 were distributed across the whole floral primordium. At stage s3, expression of DcMADS1 became restricted to the shoulder region of the flower primordium (Figure 3a) where subsequently sepals (Figure 3a; arrow) and petals (Figure 3b) arise. During further differentiation, the mRNA remained detectable in sepals, in petals and in the pedicel of the floret (Figure 3c). Expression of DcMADS1 was found to be coincident to transcription of SQUAMOSA in Antirrhinum (Huijser et al., 1992) and AP1 in Arabidopsis (Mandel et al., 1992). The mRNA of DcMADS2 (Figure 3d–f) appeared at early stage s3 when shoulder regions of the floret primordium start to extend (corresponding floret shapes are indicated in Figure 1). It was restricted to the area of whorls 2 and 3 (Figure 3d). Transcripts of DcMADS3 (Figure 3g–i) were already detected at stage s2 when floret shape was still globular and displayed expression throughout the whole floret (Figure 3g; arrow). At s3, mRNA of DcMADS3 became limited to those regions where subsequently petals and stamens develop (Figure 3g; right primordium). During advanced stages, the transcripts of DcMADS2 and DcMADS3 were uniformly distributed in developing petals and stamens. Very low expression levels were also detectable in the fourth whorl (Figure 3f,i). Expression patterns of DcMADS2 and DcMADS3 corresponded to those described for GLOBOSA and DEFICIENS of Antirrhinum (Schwarz-Sommer et al., 1992; Tröbner et al., 1992). The transcripts of DcMADS4 were observed at late stage s3, first in the floral centre (Figure 3j) and subsequently in stamens and carpels (Figure 3k,l), revealing expression patterns as reported for PLENA in Antirrhinum (Bradley et al., 1993) and AGAMOUS in Arabidopsis (Yanofsky et al., 1990). Until organ primordia start to develop, transcripts of DcMADS5 were found distributed over the whole floret (not shown).
The expression patterns indicate early roles for DcMADS1 and DcMADS5 before organ primordia arise. They could at least partially be involved in specification of meristem identity or might act as mediators between the meristem and the organ identity network. Expression from stage s3 onwards and spatial distribution of transcripts indicate a function of DcMADS1–4 as organ identity genes. As has already been suggested from their homology to the respective genes of Arabidopsis and Antirrhinum, DcMADS1 may participate in A-function (as AP1 of Arabidopsis) and DcMADS2 and DcMADS3 act as B-genes, whereas DcMADS4 may be a prominent member of C-function. Hence, we focused on our further analyses on DcMADS1–4.
Transcript patterns of DcMADS1–4 in ‘petaloid’ and ‘carpeloid’ CMS flowers
Carrot CMS plants with ‘petaloidy’ of third-whorl organs resemble homeotic mutants with impaired C-function to a certain degree. According to the ABC-model, an impaired C-function promotes A-function in organs of the inner whorls. Therefore, we first analysed whether third-whorl organs adopt a petal identity because of an ectopic expression of DcMADS1 in ‘petaloid’ CMS flowers. However, as in unmodified flowers, transcripts of DcMADS1 were found in the perianth organs (Figure 4a,e), but not in third-whorl petals (Figure 4e; w3p). On the other hand, expression of DcMADS4 was found in ‘petaloid’ third-whorl organs (w3p) as well as in the genuine carpels (c) (Figure 4d,g), revealing the same expression area and signal strengths as found in unmodified flowers. Transcript patterns of DcMADS2 (Figure 4b) and DcMADS3 (Figure 4c,f) were also as in unmodified flowers. Thus, replacement of stamens by homeotic petals was not associated with altered transcript patterns of DcMADS1–4.
Until organ primordia were established (late s5), we found no altered expression patterns of DcMADS1–4 in ‘carpeloid’ florets (Figure 5a–d and not shown). During stage s6 (Figure 5e–g) when organ development deviates (cf. Figure 1j), we analysed transcript patterns of DcMADS1–4 in serial sections. The mRNA accumulation of DcMADS1 and DcMADS4 was found to be as in unmodified flowers: expression of DcMADS1 limited to perianth (Figure 5e) and of DcMADS4 to reproductive organs (Figure 5g) facilitated proper assignment of whorl identity enabling immediate comparison of expression states in organs of whorls 2 and 3 (Figure 5e–g).
A distinct difference from unmodified flowers was observed in case of transcript accumulation of the B-genes DcMADS2 and DcMADS3. Comparison of consecutive sections revealed a clearly diminished mRNA abundance of DcMADS2 (not shown) and DcMADS3 (Figure 5f) compared to DcMADS1 or DcMADS4 (Figure 5e,g) and to expression in unmodified flowers (Figure 3h). The transcript levels of DcMADS3, and also of DcMADS2 (not shown), were only slightly diminished in sepaloid petals (Figure 5f; sp) and predominantly found in distal regions and along the midrib regions (Figure 5l). The faint expression the distal parts of the homeotic carpel styles is shown in Figure 5(m). Low but significant transcript amounts were also seen in the floral centre of ‘carpeloid’ (Figure 5f) and unmodified florets. Transcript abundance of DcMADS2 and DcMADS3 remained decreased during subsequent differentiation (Figure 5j,k) while mRNA levels of DcMADS1 and DcMADS4 were as in unmodified flowers (Figure 5h,i). Obviously, the DcMADS2 and DcMADS3 transcript accumulation decreases only after establishment of the primordia of third-whorl organs. The low but detectable transcript level in the third-whorl primordia is kept from stage s6 onwards. Similarly, the expression of DEFICIENS in homeotic third-whorl organs of Antirrhinum B-mutants was reduced, but not completely absent (Schwarz-Sommer et al., 1992; Zachgo et al., 1995).
We have identified five carrot cDNAs, DcMADS1–5, potentially encoding MADS box factors involved in organ identity during flower development. Based on their expression pattern during flower development as well as on their sequence similarity to specific MADS box transcription factors from Arabidopsis and Antirrhinum, we conclude that DcMADS1 is involved in A-function, DcMADS2 and DcMADS3 in B-function and DcMADS4 in C-function, whereas DcMADS5 has significant similarities to the SEPALLATA group (former AGL2-subfamily, Ma et al., 1991) essential for proper B- and C-function in Arabidopsis (Honma and Goto, 2001; Pelaz et al., 2000). Reduction of transcript levels of the two B-activity genes, DcMADS2 and DcMADS3, is correlated with abnormal flower development in the ‘carpeloid’ CMS type.
The ‘carpeloid’ flowers are developed by a variant of the ‘petaloid’ CMS type of carrots. They strongly resemble homeotic Arabidopsis and Antirrhinum mutants with impaired B-function by possessing sepal-like structures in whorl 2 and homeotic carpels in whorl 3 (Figure 1). The initiation of their petal and stamen primordia starts exactly as in unmodified flowers, while first morphological deviations are seen at stage s6. Transcript patterns of the putative carrot B-factors DcMADS2 and DcMADS3 were unaltered during early formation of organ primordia (Figure 5; s3–s5), indicating proper initial transcription during these stages. In Antirrhinum, initial expression of B-genes is controlled by factors specifying meristem identity (e.g. Hantke et al., 1995) and by products directing spatial boundaries of B-expression (Simon et al., 1994). As initial expression of both DcMADS2 and DcMADS3 was neither delayed nor reduced, a proper function of these earlier acting co-regulators can be assumed. When organs start to differentiate and aberrant development of third-whorl primordia was conspicuous, however, we observed a drastically reduced transcript abundance of DcMADS2 and DcMADS3 in ‘carpeloid’ flowers (Figure 5). This is a specific response of DcMADS2 and DcMADS3, as transcript accumulation of the other investigated genes remained unchanged throughout ‘carpeloid’ flower formation.
‘Carpeloid’ flowers represent a genetically well-studied CMS type inherited in a non-Mendelian maternal manner (Börner et al., 1995; Straub, 1971), characterised by a specific combination of certain nuclear genes with certain mitochondrial and plastid genes. As CMS was found to be caused by interactions of nuclear and mitochondrial gene products in other well-studied systems (Budar and Pelletier, 2001; Mackenzie and McIntosh, 1999; Schnable and Wise, 1998), this is most likely also the case for the ‘petaloid’ CMS type and its ‘carpeloid’ variety (Nakajima et al., 2001; Scheike et al., 1992; Szklarczyk et al., 2000). Therefore, a likely scenario for ‘carpeloid’ flower formation is that a specific interaction of nuclear and mitochondrial gene products leads, probably via a mitochondrial dysfunction and a hitherto unknown signalling pathway, to the lowered transcript accumulation of DcMADS2 and DcMADS3. We cannot rule out the possibility that the mitochondrial effect on DcMADS2 and DcMADS3 is an indirect one, i.e. the disturbed nuclear–mitochondrial interaction may act primarily on another gene/gene product. Yet the striking similarity of the ‘carpeloid’ flower phenotype with flowers of the homeotic Antirrhinum and Arabidopsis mutants with impaired B-function and the fact that B-genes are master regulators of stamen and petal identity support the idea of a causal relationship between the low level of DcMADS2 and DcMADS3 transcripts and flower malformation in this CMS type. If it is not a B-activity gene itself that causes the defect in flower development of ‘carpeloid’ carrots (that would best explain the observed phenotype), it should be a gene involved in controlling B-activity genes. With DcMADS5, a carrot MADS gene with homology to the E-class factor SEPALLATA1 (AGL2) of Arabidopsis, we have analysed a likely candidate for such a gene. In support of a direct effect of the mitochondria on B-activity genes, DcMADS5 did not exhibit an altered transcript accumulation during ‘carpeloid’ flower formation. However, further research at the DcMADS5 protein level and inclusion of more genes would be needed to precisely elucidate the molecular basis of the ‘carpeloid’ phenotype.
The simultaneously altered expression patterns of DcMADS2 and DcMADS3 in ‘carpeloid’ flowers point to their interdependence as described for their counterparts in Antirrhinum and Arabidopsis: these B-factors are known to cooperate with each other by a mechanism of auto- and cross-regulation during advanced floral stages (Goto and Meyerowitz, 1994; Jack et al., 1992; Schwarz-Sommer et al., 1992; Tröbner et al., 1992). As expression of both DcMADS2 and DcMADS3 is properly induced at s3 but diminished from s6 onwards, it is obvious that only ‘late’ B-regulation is disturbed in the ‘carpeloid’ flower phenotype. We found a different extent of transcriptional decrease with minor expression of DcMADS2 and DcMADS3 in organs of whorl 3 than whorl 2 (Figure 5f). This is in agreement with the co-regulated organ-specific expression as described for B-genes of Arabidopsis and Antirrhinum (Samach et al., 1997; Zachgo et al., 1995) and can explain the minor phenotypic defects of second-whorl organs where sepal-like features are often limited to greenish basal streaks (whereas from third-whorl primordia a completely novel organ type develops). Weak hybridisation signals of DcMADS3 in the tips of homeotic carpels indicate that B-function was not completely lacking in advanced stages of ‘carpeloid’ flower development. As shown for Antirrhinum, intact stamen specification depends on a proper induction as well as on a continuous expression of the B-genes throughout flower formation (Zachgo et al., 1995). Therefore, the reduced expression during late B-activity in ‘carpeloid’ florets implies that the ‘threshold’ of remaining B-function required for ultimate stamen formation is insufficient. Instead, the intact C-function gives rise to carpeloid organ formation in the third-whorl.
Interestingly, homeotic flower modifications of Nicotiana (+Hyoscyamus) cybrids revealed a significant down-regulation of a transcript hybridising with a GLOBOSA probe in Northern blot analysis (Zubko et al., 2001), also suggesting a cytoplasmic influence on the expression of a putative B-activity gene. The identity of the hybridising transcript remains to be elucidated, also whether this gene exclusively is functionally affected in the cybrids, and at which developmental stage its expression decreases. In contrast, homeotic flowers of alloplasmic tobacco do not seem to be associated with altered expression levels of B- and C-genes at any stage of flower development. Here, flower malformation was suggested to depend on impaired regulation of floral genes affecting early cell division and/or boundary-specification between appropriate organ types (Bereterbide et al., 2002; Farbos et al., 2001).
Also ‘petaloid’ flowers, though genetically related to the ‘carpeloid’ CMS phenotype, showed no altered expression of B-genes in our studies. This finding was not unexpected because the ‘petaloid’ CMS phenotype differs from flowers of B-mutants. To a certain extent, ‘petaloid’ flowers resemble homeotic C-mutants of Antirrhinum and Arabidopsis (Bradley et al., 1993; Yanofsky et al., 1990), but C-mutants are additionally modified in whorl 4 by revealing new flowers inside the original florets. ‘Petaloid’ florets retain determinacy and always have true carpels, and, in accordance with that phenotype, we also found no altered transcript accumulation of the A-gene DcMADS1 and the C-gene DcMADS4 in homeotic third-whorl petals of ‘petaloid’ flowers (Figure 4).
In conclusion, our data suggest that a disturbed interaction between nuclear and extranuclear (mitochondrial) gene products leads to an impaired expression of MADS box genes with B-function that in turn results in flower malformation and male sterility of ‘carpeloid’ carrot plants. Virtually no information exists about signals between mitochondria and nucleus that regulate transcription in these organelles in plant cells (Lohrmann et al., 2001; Mackenzie and McIntosh, 1999). Regulatory feedback mechanisms from plastids directed to the nucleus include redox-regulated phosphorylation steps as well as metabolites like porphyrins, sugars or reactive oxygen species (Rodermel, 2001). Similar mechanisms have been reported to be involved in the communication of mitochondria with the nucleus in yeast and animals (e.g. Biswas et al., 1999; Duchen, 2000; Sekito et al., 2000) and might also operate in plants.
Male fertile plants of cultivated carrot (Daucus carota sativus) with unmodified flowers as well as CMS plants with ‘petaloid’ and ‘carpeloid’ flowers were obtained from the collection of the Institute of Horticultural Crops (BAZ), Quedlinburg. ‘Petaloid’ CMS lines were derived from combinations of CMS-inducing cytoplasm of line W33 (Banga et al., 1964; Thompson, 1961) with nuclear backgrounds of several European carrot cultivars of the BAZ collection. Plants of the ‘carpeloid’ CMS type were descended from the Institute for Applied Genetics, University Hannover and were obtained by crosses of a petaloid CMS line (W33) with different varieties of the cultivated carrot according to Straub (1971) at the Institute of Breeding Methods in Vegetable (BAZ), Quedlinburg. The maternal inheritance of the male sterility trait in association with ‘petaloid’ and ‘carpeloid’ flower features was certified by genetic investigation (Börner et al., 1995; Scheike et al., 1992; Szklarczyk et al., 2000). Plants were cultivated under glasshouse conditions after vernalisation for 6 weeks at 5°C. Flower traits were evaluated from umbels of first to fifth order.
Construction and screening of a flower-specific cDNA library
Umbels that provide floral buds of early and advanced stages (3–10 mm) were collected from single plants. Poly(A)+ RNA was prepared (QIAGEN, Hilden, Germany) and 5 µg was used for cDNA synthesis followed by directed ligation into the λgt22A phage vector (Gibco/BRL, Eggenstein, Germany). Ligation products were packaged with the Gigapack III Gold System (STRATAGENE, Heidelberg, Germany). About 250 000 independent recombinant clones were obtained. Totally, 750 000 clones of the amplified library were screened with a mixture of probes containing MADS box encoding regions from the Antirrhinum genes GLOBOSA, DEFICIENS and PLENA. Hybridisation was carried out at 53°C overnight in sodium phosphate buffer (0.25 m Na2HPO4, pH 7.2, 7% SDS). Filters were washed in 2× SSC, 0.1% SDS for 15 min first at room temperature and then at 53°C followed by two washing steps for 15 min at 53°C in 1× SSC, 0.1% SDS. Putative MADS box homologous clones were amplified by PCR from single plaques, using primers deduced from vector regions. Inserts were classified by their different sizes and signal strengths under hybridisation conditions at 56°C and exceeding washing procedures up to 30 min. Sequencing was done on an automated sequencer (ABI 373 A, Applied Biosystems, Weiterstadt, Germany). Database research and alignments were performed using the blast (Altschul et al., 1990) and fasta (Pearson and Lipman, 1988) programs, respectively.
Umbels were fixed, wax-embedded and sectioned as previously described (Linke et al., 1999). Sections of 10 µm were mounted onto positively charged Superfrost + slides (Fisher Scientific, Berlin, Germany). Staining with haematoxylin and eosin was performed according to standard procedures.
In situ hybridisation
For preparation of RNA probes of DcMADS1–4, the 3′-DNA regions downstream of the MADS domain were amplified by PCR and subcloned into pCRTM II (Invitrogen, San Diego, USA). Digoxigenin-labelled antisense RNA probes were generated according to the instructions of the supplier (Boehringer, Mannheim, Germany). Labelled sense riboprobes were used as a negative control (data not shown).
Pre-treatment before hybridisation was performed as described by Huijser et al. (1992). Sections were covered with 100 ng of denatured RNA probe in 60 µl hybridisation solution (AMERSHAM, Buckinghamshire, UK). Hybridisation was done in a humidified chamber for 16 h at 53°C. Washing steps were performed in 0.2× SSC, 0.1% SDS, 2× for 30 min at 50°C and 2× for 20 min in 0.1× SSC, 0.1% SDS. Immunological detection using antibodies against digoxigenin coupled with alkaline phosphatase and final staining procedure with NBT/BCIP were done according to the suppliers manual (BOEHRINGER). Digital analysis was performed using an Olympus BH2 microscope (Olympus Optical Co., Japan) combined with a Panasonic F15HS video camera (Matushita Electric Industrial Co. Osaka, Japan). Image processing was done using analysis 2.1 (Soft-Imaging Software (SIS), Münster, Germany).
We thank Barb Sears for critically reading the manuscript and gratefully acknowledge H. Saedler (Max Planck Institute for Plant Breeding, Cologne, Germany) for kindly providing gene probes of SQUAMOSA, GLOBOSA, and PLENA from snapdragon. We are obliged to Marek Szklarczyk for providing the partial sequence of DcFLO and grateful to three anonymous referees for helpful suggestions. We also thank J. Müller and B. Liebe for technical help. This work was supported by grants from the Deutsche Forschungsgemeinschaft and the Fond der Chemischen Industrie to T.B. and by a HSPIII fund of the HU, Berlin to B.L.
Tables S1–S3 show sequence comparison between DcMADS1–4 and appropriate orthologs of Antirrhinum and Arabidopsis at nucleotide and amino acid levels, and Tables S4 and S5 show sequence comparison between DcMADS5 and the SEPALLATA group SEP1–3 (former AGL2, AGL4, AGL9) of Arabidopsis thaliana.