In flowering plants, class-B floral homeotic genes encode MADS-domain transcription factors, which are key in the specification of petal and stamen identity, and have two ancient clades: DEF-like and GLO-like genes. Many species have one gene of each clade, but orchids have typically four DEF-like genes, representing ancient gene clades 1, 2, 3 and 4. We tested the ‘orchid code’, a combinatorial genetic model suggesting that differences between the organs of the orchid perianth (outer tepals, inner lateral tepals and labellum) are generated by the combinatorial differential expression of four DEF-like genes. Our experimental test involves highly sensitive and specific measurements, with qRT-PCR of the expression of DEF- and GLO-like genes from the distantly related Vanilla planifolia and Phragmipedium longifolium, as well as from wild-type and peloric Phalaenopsis hybrid flowers. Our findings support the first ‘orchid code’ hypothesis, in that absence of clade-3 and -4 gene expression distinguishes the outer tepals from the inner tepals. In contrast to the original hypothesis, however, mRNA of both clade-3 and -4 genes accumulates in wild-type inner lateral tepals and the labellum, and in labellum-like inner lateral tepals of peloric flowers, albeit in different quantities. Our data suggest a revised hypothesis where high levels of clade-1 and -2, and low levels of clade-3 and -4, gene expression specify inner lateral tepals, whereas labellum development requires low levels of clade-1 and -2 expression and high levels of clade-3 and -4 expression.
The developmental program determining floral organ identity is a relatively well-studied aspect of angiosperm development, and is essential for understanding the morphological evolution of the flower. The ABCDE genetic model of floral organ identity determination (Theißen, 2001) is a unifying paradigm and the core of comparative analysis of flower development and evolution (Ferrario et al., 2004; Theißen and Melzer, 2007). This model is an extension of the ABC model that emerged from the comparative study of homeotic mutants in the eudicotyledoneous plants Arabidopsis thaliana and Antirrhinum majus (Coen and Meyerowitz, 1991; Weigel and Meyerowitz, 1994). The ABCDE model comprises five major classes of homeotic selector genes: A, B, C, D and E, most of which are MADS-box genes (Theißen, 2001). Each of these gene classes is involved in the combinatorial determination of the identity of different floral organs: A and E genes specify sepals (Ditta et al., 2004); genes A, B and E determine petals; the combination of genes B, C and E specify stamens (male reproductive organs); genes C and E determine carpels (female reproductive organs); and genes D and E determine ovules (reviewed in: Theißen, 2001; Krizek and Fletcher, 2005; Theißen and Melzer, 2007). The conservation of this system across major plant groups indicates that floral homeotic genes constitute an ancient regulatory network for most angiosperms (reviewed in: Rijpkema et al., 2010; Litt and Kramer, 2010). As the different functional classes of MADS-box genes are members of distinct clades (reviewed in: Becker and Theißen, 2003), studying the expression and function of the genes in the comparative context of their phylogeny has been important for testing hypotheses on the evolution of flower morphology.
Many monocot species have several copies of DEF- and GLO-like genes (putative class-B genes), which are expressed slightly differently from their single-copy orthologs in Antirrhinum majus and Arabidopsis thaliana (Kim et al., 2004; Zahn et al., 2005). For instance, the heterotopic expression of DEF- and GLO-like genes probably determines the development of petaloid tepals instead of sepals in the outer whorl of the flowers of Tulipa gesneriana and other petaloid monocots (van Tunen et al., 1993; Kanno et al., 2003; Nakamura et al., 2005; Nakada et al., 2006). Although this research documents alternative ways by which class-B proteins may regulate their own transcription and may be associated with novel morphologies, it is still not clear how MADS-box gene duplication and transcriptional divergence has driven the evolutionary diversification of floral morphology.
An exciting opportunity to address this question comes from recent studies on putative class-B genes of Orchidaceae (Asparagales), which indicate that the unprecedented perianth diversity of this family might be associated with the duplication of MADS-box genes (Hsu and Yang, 2002; Tsai et al., 2004, 2005; Xu et al., 2006; Kim et al., 2007; Chang et al., 2010). With more than 25 000 species, orchids are the second largest plant family, and their flowers have a unique zygomorphic floral structure, including three types of perianth organs: three outer tepals (often also termed ‘sepals’) in the first floral whorl, and two inner lateral tepals (‘petals’) as well as a frequently highly modified inner median tepal called labellum (or lip) in the second floral whorl (Rudall and Bateman, 2002) (Figure 1a). The labellum is, with few exceptions, always different from the other perianth organs, and is elaborately adorned with calli, spurs, glands and a distinctive pattern of coloration (Figure 1). The labellum is probably homologous to the adaxial (dorsal) tepal of other monocot flowers, but its position in the perianth is often the lowest because of resupination, a 180° developmental torsion of the pedicel and/or the ovary that changes floral orientation (Arditti, 2002; Bateman and Rudall, 2006). The abaxial orientation of the resupinate labellum and its location in direct opposition to the fertile anther strongly suggest that its morphological elaboration is the result of adaptation to specific pollinators.
The orchid gynostemium (or column) is a compound structure formed by the complete or partial union of male and female organs (Dressler, 1993; Rudall and Bateman, 2002) (Figure 1). The gynostemium shows a remarkable level of species-specific morphological variation in size and shape, presence of appendages, position and number of the anthers, as well as the shape of the pollinia and the degree of coherence between pollen grains (Dressler, 1993).
In contrast to other monocot flowers with two whorls of fully developed and functional stamens, in all orchid subfamilies the three adaxial stamens are suppressed. In the subfamily Apostasioidea, either all three of the abaxial stamens develop (Neuwiedia) or just the two inner abaxial stamens develop (Apostasia) (Bateman and Rudall, 2006). This configuration also characterizes Cypripedioideae. The flowers of the members of the subfamilies Vanilloideae, Orchidoideae and Epidendroideae develop a single, outer abaxial stamen (Figures 2–5 in Bateman and Rudall, 2006).
As in the monocot families Hypoxidaceae and Amaryllidaceae, the orchid ovary is inferior with respect to the bases of the perianth organs (epigyny), and is formed by three carpels. In most orchids there are no divisions between carpels, but in genera from the subfamilies Apostasioideae and Cypripedioideae, like Apostasia, Selenpedium and Phragmipedium, the ovary has three chambers or locules (Dressler, 1993).
The flowers of orchids are of considerable morphological diversity, and their reproduction and development have intrigued scientists for centuries (Brown, 1810; Brown, 1831; Darwin, 1862; Rolfe, 1909; Garay, 1960; Chen, 1982; Cozzolino and Widmer, 2005). For instance, Darwin and others interested in the morphological evolution of the orchid flower employed information on the vascularization pattern of the flower organs to infer the relationships of homology with those of flowers of other petaloid monocots (Brown, 1810; Brown, 1831; Darwin, 1862; Lindley, 1840; Nelson, 1967). These analyses aimed at determining the developmental transitions behind the suppression of most stamens and the origin of the labellum. These researchers hypothesized that the labellum originated from the union of the inner median tepal with two adaxial stamens (Brown, 1831; Lindley, 1840; Darwin, 1862; Nelson, 1967). This hypothesis was eventually rejected by analysis of the pattern of vascularization and microscopical observations of the organ ontogeny, demonstrating the morphological and developmental independence of perianth and stamens (Brown, 1810; Crüger, 1864; Swamy, 1948; Kurzweil, 1987, 1998). Specifically, the morphological analysis of Kurzweil showed that in monandrous orchids the labellum and the anther are initiated at different time points, and from distinct primordia (Kurzweil, 1987, 1998).
Previously, we linked the pattern of expression of DEF-like genes, documented by analyses of wild-type and perianth-mutant (peloric) orchids (Hsu and Yang, 2002; Tsai et al., 2004; Xu et al., 2006; Kim et al., 2007), with data that we had generated on the molecular phylogeny of these genes (Mondragón-Palomino and Theißen, 2008, 2009). We found that DEF-like genes of orchids are grouped into four well-supported ancient clades (Figure 1b), and that genes from a specific clade are expressed in the same organs of the perianth. Specifically, genes from clade 1 (PeMADS2-like genes) and clade 2 (OMADS3-like genes) are expressed in all tepals (Hsu and Yang, 2002; Tsai et al., 2004; Xu et al., 2006), whereas genes from clade 3 (PeMADS3-like genes) are only expressed in inner tepals (Tsai et al., 2004; Xu et al., 2006; Kim et al., 2007), and genes from clade 4 (PeMADS4-like genes) are exclusively expressed in the labellum (Tsai et al., 2004). This suggested a simple combinatorial model for organ identity specification that we termed the ‘orchid code’ (Mondragón-Palomino and Theißen, 2008, 2009). Assuming that class-A, -C and -E genes are expressed as described in the ABCDE model, according to the ‘orchid code’, expression of clade-1 plus clade-2 genes alone leads to the development of organs with outer tepal identity, expression of clade-1 plus clade-2 plus clade-3 genes leads to the development of inner lateral tepals, and expression of clade-1 plus clade-2 plus clade-3 plus clade-4 genes specifies labellum development. According to our hypothesis, the difference between outer and inner tepals depends on the differential expression of clade-3 genes. Similarly, the distinction between inner lateral tepals and the labellum is determined by differential expression of clade-4 genes (Mondragón-Palomino and Theißen, 2008, 2009).
So far the ‘orchid code’ has been supported by a phylogenetically very limited set of gene expression data, covering only a few genes in just two (Orchidoideae and Epidendroideae) of the five extant subfamilies of orchids (Figure 1c). To test it more extensively we present here a comparative analysis of the expression of DEF- and GLO-like genes in Vanilla planifolia (Vanilloideae) and Phragmipedium longifolium (Cypripedioideae), as well as in wild-type and peloric flowers from the Phalaenopsis hybrid ‘Athens’ (Epidendroideae) (Chase et al., 2006) (Figure 1d,e). In this analysis we employed quantitative real-time PCR (qRT-PCR) with a well-validated normalization approach based on the stable expression of three independent reference genes (Vandesompele et al., 2002; Hellemans et al., 2007). This set-up enables a sensitive cross-species characterization and statistical comparison of the levels of mRNA of each of four highly similar DEF-like paralogs, which, like others encoding transcription factors, are expressed at a relatively low level.
With this approach we evaluated two essential aspects of the ‘orchid code’: the key involvement of clade-4 DEF-like genes in the development of the labellum and the applicability of the ‘orchid code’ to a wider spectrum of orchid subfamilies.
The pattern of expression of paralogous orchid DEF-like and GLO-like genes is highly conserved and different in the organs of the outer and inner perianth
Our comparative analysis showed a high degree of conservation of the pattern of expression of DEF- and GLO-like genes in the orchids Vanilla planifolia (Vanilloideae), Phragmipedium longifolium (Cypripedioideae) and the Phalaenopsis hybrid ‘Athens’ (Epidendroideae), which belong to three orchid subfamilies (Figure 1c) that diverged from each other between 60 and 70 Ma (Ramírez et al., 2007). Specifically, DEF-like genes from clades 1 and 2 have a relatively high level of expression in the outer tepals, inner lateral tepals, gynostemium and ovary (Figure 2). The lowest level of expression of these two gene types was consistently observed in the labellum of Phragmipedium longifolium and Phalaenopsis (Figure 2). In the case of Vanilla planifolia, expression of VaplaDEF1 (clade 1) has its lowest level in the labellum, but in the outer tepals, inner lateral tepals and ovary, expression is also low in comparison with the other two species (Figure 2). In several cases, the relative levels differ between the two bud sizes analyzed (P ≥ 0.05, d.f. = 9, Wilcoxon matched pairs signed ranks test; Table S1); however, the patterns of expression persist across developmental stages and species (Figures 4a and S1).
In contrast, the genes from clades 3 and 4 of these three species have a significantly higher level of relative mRNA accumulation in the inner lateral tepals, labellum and gynostemium than genes from clades 1 and 2. In particular, the expression of these two genes is higher in the labellum of all species analyzed than in all other organs (Figure 3). On the other hand, the activity of genes of clades 3 and 4 in the outer tepals and ovary is for most cases the lowest measured in all species (Figure 3). In the case of genes of clades 1 and 2, some of the differential patterns of expression previously described in the earlier stages of flower development are statistically different from those we documented at later stages in each species (P ≥ 0.05, d.f. = 9, Wilcoxon matched pairs signed ranks test; Table S1), but the relatively high level of activity of genes from clades 3 and 4 in the inner perianth and gynostemium persisted (Figures 4a and S1).
In agreement with previous observations, the levels of expression of genes in clades 3 and 4 within a species are highly correlated with each other. Namely, pairwise comparison of the levels of activity of VaplaDEF2 and VaplaDEF3 in Vanilla planifolia, PhlonDEF3 and PhlonDEF4 in Phragmipedium longifolium, and PeMADS3 and PeMADS4 in Phalaenopsis, showed a Spearman’s correlation coefficient of 0.77, 0.96 and 0.84 in the earliest developmental stages investigated, respectively (Table S2). Despite these correlations, the level of activity of genes from clade 3 is always higher than that of the corresponding genes from clade 4 (Figures 3, 4a and S1).
We also determined a high correlation coefficient between the activities of genes in clades 1 and 2 in the earliest developmental stages (Table S2). Specifically, the levels of expression of PhlonDEF1 and PhlonDEF2 of Phragmipedium longifolium showed a correlation coefficient of 0.67, whereas in Phalaenopsis the levels of mRNA of PeMADS2 and PeMADS5 had a correlation coefficient of 0.73.
In agreement with the high level of positive correlation between the activity of the genes in sister clades 1 and 2, as well as the genes in clades 3 and 4 in Phalaenopsis, we observed a negative correlation of −0.75 and −0.79 in the expression of PeMADS5 (clade 2) and PeMADS4 (clade 4) in buds of 0.5 and 1 cm length, respectively (Table S2).
The different levels of expression of DEF-like genes in the perianth and the reproductive organs of the species investigated seem to follow a regular pattern, with the clade-3 genes being the most highly expressed ones, followed by clade-4 genes. In contrast, genes in clades 1 and 2 and the single GLO-like gene have relatively low levels of expression (Figures 2, 3, 4a, 5 and S1).
The relative expression of DEF-like genes contrasts with the generally uniform and lower level of expression of GLO-like genes in all of the investigated organs (Figures 4a, 5, S1 and S3).
The differential expression of DEF-like genes from clades 3 and 4 are associated with the development of the orchid labellum
The previous results suggest that the combined, higher differential expression of clade-3 and -4 DEF-like genes is associated with the development of the orchid labellum from the three species analyzed. We obtained additional evidence for this relationship from the expression of class-B genes in flowers of the Phalaenopsis hybrid ‘Athens’, where the inner lateral tepals are transformed into labellum-like structures (Figures 1e and 4).
In the labellum-like inner lateral tepals of the peloric Phalaenopsis hybrid ‘Athens’, both PeMADS3 (clade 3) and PeMADS4 (clade 4) are expressed at levels 3 and 2.8 times higher than in the inner lateral tepals of the wild-type flower, respectively (Figure 4). Whereas the level of expression of these genes is higher in the inner lateral tepals of the peloric form than in the wild-type inner lateral tepals, this relative level is still lower or at the same level than that in the labellum of both wild-type and peloric flowers (Figure 4). The activity of PeMADS2 (clade 1) increased twofold in the gynostemium, and the levels of PeMADS3 (clade 3) and PeMADS4 (clade 4) increased 1.4 and 5 times, respectively, in the outer tepals of peloric orchids. However, because of the low level of expression of these genes in wild-type orchids, these changes do not lead to high levels of gene expression, and obviously do not cause visible morphological changes in these structures (Figure 4).
The higher expression of PeMADS3 and PeMADS4 in the labellum-like tepals of the peloric Phalaenopsis hybrid ‘Athens’ were independently documented in the peloric flowers produced by three individual plants with the same phenotype, suggesting, but not conclusively demonstrating, that the ectopic expression of PeMADS3 and PeMADS4 causes the transition from organs with inner lateral tepal identity into labellum-like tepals (Figure S2).
Combined differential activity of class-B genes during the development of the orchid gynostemium and ovary
We consistently observed that in the column of wild-type flowers, clade-3 and -4 DEF-like genes have a higher level of expression than in the ovary. This observation is consistent with the fact that the gynostemium holds the male reproductive structures, and that in angiosperms class-B genes (together with class-C genes) determine the development of stamens (Figures 3 and 4).
The expression of clade-1 and -2 DEF-like genes in reproductive organs tends to follow a species-specific pattern, and might be associated with the different developmental stages analyzed (Figure 2). The differences in the expression of these genes in the gynostemium and ovary of wild-type and peloric orchids are relatively low (with 37–200% more mRNA in peloric orchids than in wild-type orchids), particularly considering that the overall level of expression of these genes in the reproductive organs is lower than that of clade-3 and -4 genes (Figure 4). However, the distinct patterns observed suggest that these genes might contribute differently to the development of reproductive organs. These variations might be associated with the frequent lack of stamens (Figure 1e) and fruit abortion in plants producing peloric flowers (M. Mondragón-Palomino, unpublished observations).
As observed in the perianth organs, the expression of GLO-like genes in the gynostemium and ovary of wild-type flowers is comparatively low, but it increases by about 50% in the gynostemium and decreases by about 25% in the ovary of peloric flowers (Figure 4).
Refining the ‘orchid code’
The ‘orchid code’ model was based on species from the most derived subfamilies Epidendroideae and Orchidoidea (Figure 1c), which comprise most of the Orchidaceae species (Chase et al., 2003). However, the association between the expression of clade-4 genes and the development of the labellum, a crucial aspect of the model, was based solely on a single analysis of wild-type and mutant flowers of Phalaenopsis equestris (Tsai et al., 2004). Therefore, we investigated the wider applicability of the ‘orchid code’ by measuring and comparing the expression of the unique GLO-like gene and the four DEF-like genes in the outer tepals, inner lateral tepals, labellum, gynostemium and ovary of Vanilla planifolia and Phragmipedium longifolium, as well as in wild-type and peloric flowers of the horticultural variety Phalaenopsis hybrid ‘Athens’. Vanilla planifolia and Phragmipedium longifolium represent the Vanilloideae and Cypripedioideae, two groups of orchids with an intermediate phylogenetic position between the diverse Orchidoideae and Epidendroideae and the basalmost Apostasioideae (Figure 1c; Cameron, 2006). Unfortunately, we were not able to include species from the Apostasioideae in our analyses because they are not available in cultivation beyond their range in Southern Asia.
DEF-like genes in clades 1 and 2
Despite the fact that the species analyzed belong to subfamilies that diverged between 60 and 70 Ma (Ramírez et al., 2007), and that their flowers appear to be quite different, we found that class-B genes from specific phylogenetic clades have a highly similar pattern of expression in the perianth, and to a certain extent also in reproductive organs. This documents a strong conservation of the gene expression patterns and suggests an equally strong conservation of gene functions. In agreement with previous non-quantitative expression analyses our results show that both genes from clades 1 and 2 are expressed throughout the perianth organs (Hsu and Yang, 2002; Tsai et al., 2004; Xu et al., 2006). However, our quantitative analysis demonstrates that in the labellum of all species analyzed the mRNA levels of both clade-1 and -2 DEF-like genes are up to ten times lower than in the other perianth organs (Figure 2). These results are comparable, and in general agreement, with the expression of OMADS5 (clade 1) and OMADS3 (clade 2) in the organs of wild-type and peloric flowers of Oncidium Gower-Ramsey (Epidendroideae) (Chang et al., 2010). However, these authors argue that OMADS5 (clade 1) is absent from the labellum, and has a reduced level of expression in the ectopic labellum-like organs of the mutants analyzed. They interpret this lack of expression as a condition necessary for the development of the labellum or labellum-like structures. Whereas our analysis supports the significantly reduced expression of clade-1 and -2 genes in the labellum of wild-type and peloric flowers, we did not observe a significant reduction in the expression of PeMADS2 (clade 1) in the labellum-like inner lateral tepals of peloric Phalaenopsis hybrid ‘Athens’, the ortholog of OMADS5, and only a 22% reduction in the mRNA level of PeMADS5 (clade 2) (Figure 4). This suggests that the observed reduction in the expression of clade-1 genes in the inner lateral tepals is not required for the establishment of labellum-like organ identity.
Although both our analysis and that of Chang et al. (2010) detected expression of clade-2 genes in peloric flowers of Phalaenopsis hybrid ‘Athens’ or Oncidium Gower Ramsey, respectively, an earlier expression analysis employing northern-blot hybridization reported that expression of the clade-2 gene ortholog, PeMADS5, of Phalaenopsis equestris is absent from the perianth and gynostemium of a similar three-labellum mutant (Tsai et al., 2004). As both qRT-PCR and northern-blot hybridization are highly sensitive techniques, a feasible explanation for these contrasting results is that the mutants analyzed have similar phenotypes but a different genetic or epigenetic basis, as has already been suggested for other cases of orchid pelorism (Rudall and Bateman, 2003). Considering that clade-1 and -2 genes have very similar patterns of expression in the perianth, it is possible that the lack of activity of PeMADS5 in the mutant of Phalaenopsis equestris is somehow compensated by the activity of PeMADS2, and thus might not directly cause the ectopic development of labellum-like structures.
DEF-like genes in clades 3 and clade 4
In contrast to the patterns of expression of clade-1 and -2 genes, the qRT-PCR analysis presented here indicates that both genes from clades 3 and 4 have a high relative activity in the inner lateral tepals, labellum and gynostemium. This suggests that their combined action and differential levels of expression are involved in the development of these distinct structures in the Orchidaceae. Specifically, in the two flower bud sizes of the orchid species analyzed, the expression of clade-3 and -4 genes is particularly high in the labellum of both wild-type and peloric orchids, as well as in the labellum-like inner lateral tepals of peloric Phalaenopsis hybrid ‘Athens’. This suggests that the specific development of labellum identity involves the combined and significantly higher level of expression of both clade-3 and -4 genes. In this context, the specific pattern of activity of class-B genes in orchids might control target genes that realize the identity of distinct perianth organs as well as downstream factors regulating the development of species-specific features.
The pattern of expression of orchid clade-3 DEF-like genes documented here, and the ensuing notion that this gene plays a role in the differentiation of the inner from the outer perianth, is in agreement with other expression studies in orchids (Tsai et al., 2004; Xu et al., 2006; Kim et al., 2007). Similarly, Chang et al. (2010) documented the expression of OMADS9 (clade 3) in the inner lateral tepals and labellum of wild-type flowers, as well as in the lips and labellum-like ectopic tepals in two mutants of Oncidium Gower-Ramsey. Despite major similarities between our results and those of Chang et al. (2010), Chang et al. found that the expression of OMADS9 was higher in the inner lateral tepals than in the labellum or labellum-like organs, whereas we found the opposite. This difference might be caused by assaying the expression of OMADS9 without regarding the activity of a clade-4 gene of Oncidium Gower Ramsey. Given the high level of similarity between these genes this could feasibly interfere with the process of quantification. Thus, the differences between our results and those of Chang et al. (2010) may be clarified when the expression of the clade-4 gene of Oncidium Gower Ramsey is characterized.
The only previous study on a clade-4 gene reported that expression of PeMADS4 in Phalaenopsis equestris is restricted to the labellum and the ectopic labellum-like structures of a peloric mutant with three lips. This is not consistent with our finding that clade-4 genes from Vanilla planifolia, Phragmipedium longifolium and Phalaenopsis are also highly expressed in the inner lateral tepals, labellum and gynostemium, albeit having an overall relatively lower mRNA level than clade-3 genes. Only specific information about the experiments of Tsai et al. (2004) would help to explain the differences between their results and ours. Concerning the experiments presented here, the specificity of the primers for all qRT-PCR experiments was confirmed by sequencing the resulting products, and by the similarity of the patterns of expression observed in different species.
Previously, the orchid code hypothesis considered only the presence or absence of the expression of class-B genes in specific perianth organs (Figure 6b; Mondragón-Palomino and Theißen, 2008, 2009). However, the experimental analyses outlined here (summarized in Figure 6a) instead suggest a model, the ‘refined orchid code’, in which the relative levels of expression are key to the development of specific structures (Figure 6c). Specifically, the expression of clade-1 and -2 genes and the absence of clade-3 and -4 gene expression leads to the development of outer tepals. The higher expression levels of clade-1 and -2 genes, compared with the levels of expression of clade-3 and -4 genes, leads to the development of lateral inner tepals. Labellum identity is determined by the lower/absent expression of clade-1 and -2 genes, and the expression of both clade-3 and -4 genes at significantly higher level than in inner lateral tepals (Figure 6).
In addition to refining the ‘orchid code’, the analyses presented here showed that in the reproductive organs DEF-like genes of all four clades have a higher level of expression in the gynostemium than in the ovary (Figure 6), except that clade-1 and -2 genes follow an opposite pattern in wild-type flowers of Phalaenopsis (Figures 2 and 4). For a specific understanding of the development of the identity of gynostemium and ovary, the expression of genes encoding classes C and D of floral homeotic transcription factors must be analyzed. In all species analyzed, GLO-like genes are expressed in all organs at a relatively low level.
Evolution of orchid DEF-like genes
Previous phylogenetic analyses showed that orchid DEF-like genes of clades 1 and 2, on the one hand, and clades 3 and 4, on the other hand, form two pairs of sister clades (Figure 1b; Mondragón-Palomino et al., 2009). Our analysis showed that paralogs within these sister groups have positively correlated patterns of expression. This correlated expression of paralogs within sister clades suggests that after duplication they retained similar regulatory regions, and are still under the control of similar upstream factors. Additional evidence of this is the parallel increase in activity of both PeMADS3 and PeMADS4 in the labellum-like inner lateral tepals of the peloric Phalaenopsis hybrid ‘Athens’ (Figure 4b).
Moreover, the correlated patterns of expression documented here may suggest that these paralogs are redundant. Although testing this requires experiments in planta that are currently not feasible for orchids, we do not consider this likely because these genes have been preserved for tens of millions of years under purifying selection (Mondragón-Palomino et al., 2009). In addition, despite the correlated patterns of expression, the specific levels of mRNA are very different (Figures 2, 3, 5 and S1), and might then be associated with different developmental roles. Furthermore, the corresponding proteins have distinctive motifs in their K and C domains, suggesting they might have different interaction partners (Figure 5 in Mondragón-Palomino et al., 2009).
An evolutionary interpretation of the data presented here suggests that genes in clades 1 and 2 follow an ancestral pattern of expression, because these genes are expressed in all perianth organs like DEF-like genes from other petaloid monocots (e.g. Tulipa gesneriana and Crocus sativus, reviewed in Kanno et al., 2007). Conversely, the pattern of expression of clade-3 and -4 genes represents a derived state associated with the development of labellum and inner lateral tepals.
Gene duplication and developmental modularization of the perianth
Previously, we proposed that gene duplication and transcriptional divergence created modules in the orchid perianth that develop into distinct structures under the control of a specific combination of class-B genes. Mutations and natural selection on specific genes or their targets would affect individual modules without changing the rest of the flower, and thus facilitated the spectacular morphological diversification of the orchid labellum (Mondragón-Palomino and Theißen, 2009). This also holds in the revised model. The only difference to the old model is that organ identities of the modules represented by outer tepals, inner lateral tepals and labellum are determined by the combined differential expression of distinct levels of mRNA from four DEF-like genes, rather than the simple ‘on’ versus ‘off’ patterns of differential gene expression (compare Figure 6b with c).
Although the single orchid GLO-like gene is expressed in all flower organs, this does not mean that in angiosperm evolution only duplication and transcriptional divergence of DEF-like genes is associated with morphological diversification. Recently, the duplication of GLO-like genes in species of Zingiberales was associated with the development of a staminodial labellum, which was, in a way, analogous to the ‘orchid code’ (Bartlett and Specht, 2010), suggesting that contingency might associate either lineage with morphological diversification.
Pre-anthesis flower buds of the orchids Vanilla planifolia (Vanilloideae) and Phragmipedium longifolium (Cypripedioideae) were collected in the botanical gardens of Halle (Saale), Wilhelma (Stuttgart) and Heidelberg. Plants of Phalaenopsis hybrid ‘Athens’ (Epidendroideae), with wild-type or peloric flowers, were purchased from Valerius Orchideen (http://www.orchideen-valerius.de) and maintained under glasshouse conditions of 21°C for 14 h and 15°C at night. These species were chosen based on their subfamily and the availability of blooming individuals. Flower buds from these species were collected by size. Most buds collected from Vanilla planifolia were 1.5 or 2 cm in length, whereas the buds collected from Phragmipedium longifolium were 1 and 3 cm in length. We collected buds of 0.5 and 1 cm in length from Phalaenopsis hybrid ‘Athens’ with wild-type flowers, and of 1 cm in length from three specific plants where the inner lateral tepals were replaced by labellum-like organs (Figure 1). The plants of Phalaenopsis that we employed produced either wild-type only or peloric flowers. Flower buds of a specific size were dissected into individual tubes, immediately shock-frozen with liquid nitrogen and then stored at −80°C.
Dissected flower organs were individually ground in liquid nitrogen with mortars and pestles that had been treated with HCl and were baked. Total RNA was extracted with Biomol’s reagent (Biomol, http://www.biomol.de), following the manufacturer’s protocol.
Quantification and quality assessment of each sample
The total RNA concentration was measured with a Nano Vue device (General Electric, http://www.gelifesciences.com), and its integrity was assessed on 1.2% non-degrading agarose gels. In further procedures we only employed samples where the bands of 16S and 18S rRNA were sharp and visible in the samples of all flower organs from a specific batch of buds of a given size.
Total RNA was treated with Turbo DNase from Ambion (2U μl; Ambion, http://www.invitrogen.com/ambion) following the manufacturer’s protocol. Inactivation and elimination of DNase was performed with an AquaPhenol/Chloroform extraction (Carl Roth GmbH, http://www.carlroth.com). RNA was precipitated overnight at −20°C on a half volume of 7.8 m NH4AcOH, pH 6.0, and two volumes of EtOH 100%. To isolate the RNA precipitate we centrifuged the product at 16 000 g, held at 4°C, for 30 min. The resulting pellet was washed twice in 75% EtOH and dissolved in nuclease-free H2O (Ambion). The concentration and integrity was re-assessed as described above.
cDNA was synthesized with 500 ng of total RNA from each sample with 1 μl of oligo(dT)14 reverse transcriptase Super Script III (Invitrogen, http://www.invitrogen.com) following the instructions of the manufacturer.
Primer design and standard curves
All primer pairs were designed with primer 3 v0.4.0 (http://frodo.wi.mit.edu/primer3), based on cDNA sequences previously obtained for all DEF- and GLO-like class-B genes from Vanilla planifolia, Phragmipedium longifolium (Mondragón-Palomino et al., 2009) and Phalaenopsis equestris (Tsai et al., 2004, 2005). As reference genes for normalization we designed species-specific primers against the sequences of Actin, Ubiquitin and EF1α. These genes were chosen based on experimental and computational assessment with genorm v3.5 (Vandesompele et al., 2002), indicating that they were stably expressed in the tissues of interest. Their sequences were obtained by 3′ rapid amplification of cDNA ends (RACE), as previously described (Mondragón-Palomino et al., 2009), and the sequences were deposited in GenBank (HQ693760–HQ693767). All primers were ordered from IDT (http://eu.idtdna.com), and their specificity was verified by the size of their amplicons (Table S3) and in most cases, also by cloning, sequencing and alignment. The amplification efficiency (E) of all primer pairs was calculated with a standard curve with six dilution points (Table S3). Primer sequences will be made available by the authors upon request.
Assays were performed with Absolute QPCR SYBR green mix and ROX, as a passive reference dye (Thermo Fisher, http://www.thermofisher.com), in a Mx3005P system (Stratagene, now Agilent, http://www.genomics.agilent.com). The components of each reaction are as follows: 7.5 μl of 2x SYBR green mix; 0.5 μl ROX (1:500 dilution of 1 mm); 2 μl of primer mix (2.5 nm); 1 μl of a 1:5 dilution of cDNA from a specific flower organ and 4 μl of H2O. Triplicate reactions for each target gene and organ were arranged according to the principle of sample maximization in a 96-well plate (Derveaux et al., 2010). In each plate we included a positive control (cDNA from all floral organs), and a negative control (NTC). In the analyses of the Phalaenopsis hybrid ‘Athens’ we also included an inter-run calibrator (IRC). For all reactions the amplification program was as follows: 15 min at 95°C; 15 s at 95°C, 30 s at 60°C and 30 s at 72°C, for 40 cycles; followed by a melting curve from 1 min at 95°C, 30 s at 60°C and 30 s at 95°C.
Normalization and data analysis
The cycle threshold (Ct) values of individual qRT-PCR runs were exported to qbase plus v1.5 (Biogazelle, http://www.biogazelle.com). As a measure of quality control this program only takes replicate Ct values that do not differ from each other by more than 0.5 cycles. qbase implements a modified ΔΔCt method (Hellemans et al., 2007) that takes the gene-specific amplification efficiencies calculated for each primer pair with standard curves (Table S3). In this case, qbase employed the Ct values of EF1α, Ubiquitin and Actin to generate a normalization factor for samples from individual species. The normalized quantities were rescaled relative to the sample with the lowest relative quantity and, with the corresponding standard errors, were exported to excel v12.2.7 for Mac (Microsoft, http://www.microsoft.com) and to r v2.3.1 (R Foundation for Statistical Computing, http://www.r-project.org/foundation) to perform the two-sided, paired Wilcoxon rank test as well as Spearman’s correlation test on log-transformed data.
MMP was funded by a postdoctoral fellowship from the Volkswagen Foundation (I/81 901). We thank Jo Vandesompele, Jan Hellemans and Stefaan Derveaux from Biogazelle for advice on the design and analysis of qRT-PCR assays, Dajana Lobbes and Dominik Schmidt for technical advice, to the Botanical Gardens of Halle, Stuttgart and Heidelberg (Germany) for access to their living collections as well as Domenica Schnabelrauch from the MPI of Chemical Ecology for sequencing. Special thanks to an anonymous reviewer whose comments helped us to improve our manuscript, as well as to all of the members of the Theißen Laboratory and the Friedrich Schiller University, Jena, for general support.