MADS goes genomic in conifers: towards determining the ancestral set of MADS-box genes in seed plants

Abstract

Background and Aims

MADS-box genes comprise a gene family coding for transcription factors. This gene family expanded greatly during land plant evolution such that the number of MADS-box genes ranges from one or two in green algae to around 100 in angiosperms. Given the crucial functions of MADS-box genes for nearly all aspects of plant development, the expansion of this gene family probably contributed to the increasing complexity of plants. However, the expansion of MADS-box genes during one important step of land plant evolution, namely the origin of seed plants, remains poorly understood due to the previous lack of whole-genome data for gymnosperms.

Methods

The newly available genome sequences of Picea abies, Picea glauca and Pinus taeda were used to identify the complete set of MADS-box genes in these conifers. In addition, MADS-box genes were identified in the growing number of transcriptomes available for gymnosperms. With these datasets, phylogenies were constructed to determine the ancestral set of MADS-box genes of seed plants and to infer the ancestral functions of these genes.

Key Results

Type I MADS-box genes are under-represented in gymnosperms and only a minimum of two Type I MADS-box genes have been present in the most recent common ancestor (MRCA) of seed plants. In contrast, a large number of Type II MADS-box genes were found in gymnosperms. The MRCA of extant seed plants probably possessed at least 11–14 Type II MADS-box genes. In gymnosperms two duplications of Type II MADS-box genes were found, such that the MRCA of extant gymnosperms had at least 14–16 Type II MADS-box genes.

Conclusions

The implied ancestral set of MADS-box genes for seed plants shows simplicity for Type I MADS-box genes and remarkable complexity for Type II MADS-box genes in terms of phylogeny and putative functions. The analysis of transcriptome data reveals that gymnosperm MADS-box genes are expressed in a great variety of tissues, indicating diverse roles of MADS-box genes for the development of gymnosperms. This study is the first that provides a comprehensive overview of MADS-box genes in conifers and thus will provide a framework for future work on MADS-box genes in seed plants.

INTRODUCTION

MADS-box genes comprise a large gene family coding for transcriptions factors (Gramzow et al., 2010). They are characterized by the presence of a MADS-box that encodes the DNA-binding domain of the corresponding MADS-domain proteins. Two types of MADS-box genes are distinguished, Type I or SRF-like and Type II or MEF2-like, which had probably already been established in the most recent common ancestor (MRCA) of extant eukaryotes (Alvarez-Buylla et al., 2000; Gramzow et al., 2010). In line with this, both types of MADS-box genes have been identified in most eukaryotes studied so far, even though taxa exist in which the one or other type of MADS-box genes has been lost (Gramzow et al., 2010). In plants, the proteins encoded by Type II MADS-box genes have a conserved, characteristic domain structure, with the MADS (M) domain followed by an Intervening (I) domain, a Keratin-like (K) domain and a C-terminal domain. The genes encoding these types of proteins have therefore been termed MIKC-type genes (Ma et al., 1991). In plants, the total number of MADS-box genes increased greatly to about 100 in flowering plants (angiosperms), while it remained low in all other eukaryotic groups (Alvarez-Buylla et al., 2000; Gramzow et al., 2010). MADS-box genes in land plants have been further subdivided into groups and clades based on their phylogeny and structural features (Gramzow and Theissen, 2010). The Type I genes have been subdivided into the three groups, Mα, Mβ and Mγ based solely on phylogenetic criteria, while, in the case of Type II genes, MIKC^C- and MIKC*-group genes are distinguished by different lengths of their encoded K-domains and also on phylogenetic criteria (Henschel et al., 2002; Parenicova et al., 2003; Kwantes et al., 2012). Finally, about a dozen ancient clades of MIKC^C-group genes have been recognized in angiosperms (Becker and Theissen, 2003), and a few other clades in mosses and ferns (Münster et al., 2002).

Our knowledge of the functional importance of the different types of MADS-box genes differs greatly in plants. Only a few Type I genes have been functionally characterized and it has been shown that they are mainly involved in female gametophyte, embryo and seed development (Yoo et al., 2006; Bemer et al., 2008; Kang et al., 2008; Steffen et al., 2008; Walia et al., 2009). Large fractions of their function may be hidden by redundancy, and many of these genes may have a function that is only weak, at best (Bemer et al., 2010). In contrast, the crucial functions of MIKC-type genes have long been recognized based on informative mutant phenotypes and are well studied (Schwarz-Sommer et al., 1990; Yanofsky et al., 1990; Trobner et al., 1992; Mandel et al., 1992; Pelaz et al., 2001). These genes are involved in controlling nearly all aspects of sporophyte and male gametophyte development (for recent reviews see Gramzow and Theissen, 2010; Smaczniak et al., 2012). Most prominent are their roles in flower and fruit development of angiosperms.

Land plants evolved from unicellular green algae (Cronk, 2001). The transition to land was accompanied by the evolution of structures that allow the regulation of water loss, such as cuticles and stomata (Peterson et al., 2010). Land plants comprise liverworts, mosses, hornworts (collectively called bryophytes) and tracheophytes. The tracheophytes evolved roots and vascular tissue for the transport of water and nutrients, with lycophytes being the most basal group that has these structures. The next clade that branches off from the tracheophyte tree comprises ferns and their allies such as horsetails. These ‘euphyllophytes’ represent the most basal group of land plants having true leaves. Thereafter, seeds evolved facilitating the dispersal of the corresponding plants. Seed plants comprise the ancestral gymnosperms and angiosperms. According to most molecular analyses, extant gymnosperms, comprising conifers, gnetophytes, cycads and Ginkgo, are monophyletic (Bowe and Coat, 2000; Chaw et al., 2000; Xi et al., 2013). However, the phylogenetic relationships between the different gymnosperm groups remain equivocal; nevertheless, the most recent comprehensive analyses suggest that cycads plus Ginkgo form a clade that is sister to all remaining extant gymnosperms (Wu et al., 2011; Xi et al., 2013). Angiosperms evolved having flowers that develop ovules enclosed in carpels, and seeds protected and distributed by fruits as new structural features. Hence, the evolution of some land plant lineages is characterized by the addition of new structures leading to more complex body plans.

Whole genome sequences are available for several green algae species, the moss Physcomitrella patens, the spikemoss (lycophyte) Selaginella moellendorffii and a number of angiosperms (The Arabidopsis Genome Initiative, 2000; Goff et al., 2002; Derelle et al., 2006; Tuskan et al., 2006; Jaillon et al., 2007; Merchant et al., 2007; Rensing et al., 2008; Banks et al., 2011). Hence, the complete set of MADS-box genes in the genomes of green algae, moss, spikemoss and angiosperms can be evaluated. While only one or two MADS-box genes have been identified in green algae, moss and spikemoss genomes encode around 20 MADS-box genes (Gramzow et al., 2012; Barker and Ashton, 2013). This number further increases in angiosperms, which have roughly about 100 MADS-box genes (Parenicova et al., 2003; Leseberg et al., 2006; Arora et al., 2007).

With these data, it is also possible to infer the minimal ancestral sets of MADS-box genes in the MRCA of plants, land plants, vascular plants and angiosperms. The MRCA of plants probably encoded only few MADS-box genes, but at least one Type I and one Type II gene (Gramzow and Theissen, 2010). In contrast, the MRCA of mosses and vascular plants and the MRCA of vascular plants both probably encoded at least two Type I, one MIKC^C- and one MIKC*-group gene. The MRCA of extant angiosperms probably already possessed at least three Type I, 11 MIKC^C- and two MIKC*-group genes, a great increase as compared with the ancestor of vascular plants. Looking at these numbers it appears likely that the increase in the number of MADS-box genes is correlated with the increasing complexity of land plants and hence, given the function of these genes in developmental control of extant organisms, that MADS-box genes are probably involved in the phenotypic evolution of plants (Theissen et al., 2000).

Due to the previous lack of whole genome data for ferns and allies and gymnosperms, the set of MADS-box genes in the MRCA of euphyllophytes and seed plants, respectively, could not be reliably inferred. From studies aiming at the isolation of MADS-box genes from gymnosperm transcriptomes, it is known that gymnosperms have several MIKC^C-group genes that are orthologous to those known from angiosperms (Becker et al., 2000; Futamura et al., 2008; Carlsbecker et al., 2013). Hence, the MRCA of seed plants also probably contained at least ten MIKC^C-group genes. However, whether the MRCA of extant seed plants contained even more MIKC^C-group genes and what the ancestral number of Type I and MIKC*-group genes is in seed plants is not known.

Recently, the genomes of Picea abies, Picea glauca and Pinus taeda have been sequenced (Birol et al., 2013; Nystedt et al., 2013). Together with the increasing amount of transcriptome data (Wegrzyn et al., 2008; Lorenz et al., 2012), it is now possible to get a much more detailed picture about the ancestral set of MADS-box genes in seed plants and correlate MADS-box gene evolution with the evolution of seeds and flowers. Here we use this treasure trove of data to get a more detailed picture of the dynamics of MADS-box gene evolution during the origin of seed plants and the diversification of conifers.

MATERIALS AND METHODS

Identification of MADS-box genes

The whole genome assembly of Picea abies (Nystedt et al., 2013) was downloaded from Umea University in July 2012 (now available at http://congenie.org/). The whole genome assembly of Picea glauca (Birol et al., 2013) was downloaded from SMarTForests (http://www.smartforests.ca/) in January 2013 and the 0.8 version of the assembly of Pinus taeda was downloaded from PINEREFSEQ (http://pinegenome.org/pinerefseq/). Similarly, transcriptome data were downloaded from Dendrome (http://dendrome.ucdavis.edu/; Wegrzyn et al., 2008). Further-more, transcriptome data described by Lorenz et al. (2012) were downloaded from the NCBI Short Read Archive (Sayers et al., 2012).

The scaffolds of the whole genome assemblies as well as the assembled transcriptomes were translated in all six possible reading frames to create amino acid sequences using a customized perl script. These amino acid sequences were searched for MADS domains using hmmsearch of the HMMer package (Eddy, 1996) with a customized Hidden Markov Model for plant MADS domains (Gramzow and Theißen, 2013). For the whole genome data, all results with a length of at least 30 amino acids were kept and used for phylogeny reconstruction. For transcriptome data, the complete transcript sequences were obtained for each of the HMM results with a length of at least 30 amino acids. The identified transcript sequences were assembled separately for each gymnosperm species using Sequencher v 5.1 (Gene Codes Corp., Ann Arbor, MI, USA) with a minimum match percentage of 95 and a minimum overlap of 20. For the assembled transcript sequences, the open reading frames (ORFs) were then determined using Batch ORF Finder at greengene.uml.edu or ORF Finder at NCBI (Sayers et al., 2012). The ORFs were translated into protein sequences.

To determine which transcript sequences of P. abies, P. glauca and P. taeda correspond to which MADS-domain sequences identified from these genomes, we conducted local BLAST searches (Altschul et al., 1990) using the transcript sequences as query sequences and the whole genome assemblies of the P. abies, P. glauca and P. taeda genomes as database. If a transcript sequence had a match with more than 97 % sequence identity over a length of approx. 180 nucleotides to a genomic region that was identified to have a MADS-domain by hmmsearch, only the transcript sequence was kept and the MADS-domain sequence identified from the genome was removed from the dataset. As the number of transcript sequences for which no genomic region was identified was quite high using these stringent criteria, we later also considered shorter BLAST results.

The remaining MADS-domain-containing sequences identified from gymnosperm genomes and transcriptomes were combined and identical sequences were kept only once using the function ‘Remove Redundancy’ with a threshold of 100 of the program Jalview (Waterhouse et al., 2009). The MADS-domain-containing sequences remaining after this step were named using a two- or three-letter code for the species from which the sequences were identified followed by the keyword ‘MADS’ and incrementing numbers.

Phylogeny reconstruction

The reduced dataset of MADS-domain sequences identified from gymnosperm genomes and transcriptomes were aligned with all MADS-domain proteins from Arabidopsis thaliana, Oryza sativa, Populus trichocarpa, Vitis vinifera, Amborella trichopoda, Selaginella moellendorffii and Physcomitrella patens and with MADS-domain proteins annotated from gymnosperms and ferns as far as known using hmmalign (Eddy, 1996) with a hidden Markov model for plant MADS domains (Gramzow and Theißen, 2013) and the –trim option to remove non-homologous residues from the alignment. Furthermore, amino acids corresponding to insert states were removed from the alignment. The alignment was used to reconstruct a phylogeny using the RAxML program (Stamatakis, 2006) with the -f option to conduct a rapid bootstrap analysis with 1000 replicates and search for the best-scoring maximum-likelihood (ML) tree in one program run at the CIPRES Science gateway (Miller et al., 2010). Based on this ML phylogeny, MADS-domain sequences from gymnosperms were classified into Type I and Type II MADS domains depending on their grouping relative to the Type I and Type II MADS-domain sequences of the other species which had been classified previously (Parenicova et al., 2003; Leseberg et al., 2006; Arora et al., 2007; Diaz-Riquelme et al., 2009; Gramzow et al., 2012; Barker and Ashton, 2013).

We then aligned Type I and Type II MADS-domain sequences separately using sequences from the same species as described above and the program Probalign (Roshan and Livesay, 2006). For Type I MADS-domain proteins, the alignment was cropped to contain only the MADS domain, and for Type II proteins, the alignment was cropped to contain only the MADS and K domains. Phylogenies for these two datasets were reconstructed using RAxML as described above.

Based on the ML phylogeny for Type II proteins, these proteins were separated into 11 clades. To test the stability of these clades, separate datasets were compiled containing all the MADS sequences belonging to a specific clade according to the Type II ML phylogeny plus MADS sequences from other clades of A. thaliana, V. vinifera and O. sativa. Separate phylogenies for each of the 12 clades were constructed as described above for Type I and Type II phylogenies. Finally, datasets for each clade were compiled containing the MADS proteins belonging to the corresponding clade in the Type II as well as in the separate phylogeny. SEP3 from A. thaliana was used as a representative of the outgroup for all clades except the clade containing AGL2-, AGL6-, FLC- and SQUA-like genes, in which AGL12 of A. thaliana was used as an outgroup representative, and the clade containing AG-like genes in which AGL15 of A. thaliana was used. Alignments and phylogenies were constructed as described above. For each clade, two phylogenies were reconstructed in this way, one containing all MADS proteins belonging to this clade and another only containing MADS proteins with transcript support. To test the stability of the phylogenies for the different clades of Type II proteins, we also reconstructed phylogenies with MrBayes (Ronquist and Huelsenbeck, 2003) where we only included MADS proteins with transcript support. We used the WAG model of amino acid substitutions (Whelan and Goldman, 2001), generated 6 million generations, sampled every 1000th phylogeny and excluded the first 25 % from further analysis.

Gene expression data

Information about gene expression as revealed by transcriptome data was obtained from the NCBI expressed sequence tag database and the NCBI short read archive (Sayers et al., 2012).

RESULTS

Number of MADS sequences in gymnosperms

Our search of genome and transcriptome data for Gnetum gnemon and 18 conifer species (Fig. 1) identified 1064 MADS-domain sequences (Table 1, Supplementary Data Table S1). In species for which whole-genome information is available, Picea abies, Picea glauca and Pinus taeda, 261, 121 and 367 MADS sequences were found, respectively. However, there is evidence of transcription only for 49, 23 and 60 of these MADS sequences, respectively. We found a total of 43 sequences in transcriptome data of P. abies, P. glauca and P. taeda for which we did not find a corresponding sequence in the genome. This number decreases to 19 transcriptome sequences that were not found in the genomes when less stringent criteria were used. The fact that we did not find some transcript sequences in the genomes may be due to the incomplete knowledge of gymnosperm genomes. For P. abies and P. glauca, only about 61 % of the whole genome has been sequenced (Birol et al., 2013; Michael and Jackson, 2013; Nystedt et al., 2013). While most of the missing sequence is supposed to represent repetitive elements, some genes may have also escaped sequencing so far. We believe, however, with our combination of genome and transcriptome data, that we have made a major leap towards a complete overview of the MADS-box genes in gymnosperm genomes.

Fig. 1.

Relationships of conifer genera considered here, modified after Chaw et al. (1997) and Gugerli et al. (2001). Gnetum, angiosperms, ferns, lycophytes, mosses and liverworts are shown as outgroup representatives; Gnetum is shown on a dashed line due to the unresolved position. The different colours of the branches are used in the following figures to indicate the host taxa of the corresponding genes.

Table 1.

Number of MADS sequences identified from gymnosperm genome and transcriptome data

Order	Family		Abbreviation	Total	Type I	Type II
Gnetales	Gnetaceae	Gnetum gnemon	GgMADS	41	0	41
Coniferales	Pinaceae	Cedrus atlantica	CaMADS	13	0	13
		Picea abies	PaMADS	253 (41) + 8	12	249
		Picea glauca	PgMADS	107 (9) + 14	3	118
		Picea sitchensis	PsMADS	17	1	16
		Pinus banksiana	PbMADS	2	0	2
		Pinus contorta	PcMADS	14	0	14
		Pinus lambertiana	PlMADS	41	0	41
		Pinus palustris	PpaMADS	21	0	21
		Pinus pinaster	PpiMADS	10	0	10
		Pinus taeda	PtaMADS	346 (39) + 21	17	350
		Pseudotsuga menziesii	PmeMADS	40	0	40
	Podocarpaceae	Podocarpus macrophyllus	PmaMADS	16	1	15
	Araucariaceae	Wollemia nobilis	WnMADS	11	0	11
	Sciadopityaceae	Sciadopitys verticillata	SvMADS	22	1	21
	Taxaceae	Taxus baccata	TbMADS	3	0	3
	Cephalotaxaceae	Cephalotaxus harringtonia	ChMADS	35	0	35
		Cryptomerica japonica	CjMADS	10	0	10
	Cupressaceae	Sequoia sempervirens	SsMADS	19	0	19

For species for which whole-genome information is available the numbers are given as follows: number of MADS sequences identified from genome data (number of MADS sequences identified from genome data and supported by transcriptome data) + number of MADS sequences identified from transcriptome data for which the genomic locus could not be identified.

Based on our RAxML phylogeny (Supplementary Data Fig. S1), we classified the identified MADS sequences into Type I and Type II. Bootstrap values are quite low. However, the grouping of the known MADS-box genes is correct and thus the classification of the gymnosperm sequences into Type I and Type II should also be accurate in most cases. Only 35 gymnosperm MADS sequences were classified as Type I. Even for the species for which whole-genome information is available, the percentage of Type I sequences is always less than 5 % of all MADS sequences. In contrast, 1029 gymnosperm MADS sequences were classified as Type II. In the species for which whole genome information is available, P. abies, P. glauca and P. taeda, the number of Type II sequences is 249, 118 and 350, respectively.

Type I genes

In our phylogenies of Type I genes there are two major branches that may represent large clades (Fig. 2, Supplementary Data Fig. S2). One putative clade contains genes classified as Mα genes in A. thaliana, P. trichocarpa and O. sativa (Parenicova et al., 2003; Leseberg et al., 2006; Arora et al., 2007) and MADS-box genes from V. vinifera, gymnosperms, S. moellendorffii and P. patens. The other clade comprises genes classified as Mβ and Mγ genes in A. thaliana, P. trichocarpa and O. sativa and MADS-box genes from V. vinifera, gymnosperms, S. moellendorffii and P. patens. This suggests that there are two ancient clades of Type I MADS-box genes in land plants. Again, the bootstrap values supporting the two groups are quite low but our results concur with previous studies (Gramzow et al., 2012).

Fig. 2.

Phylogeny of Type I MADS-box genes. The two major putative clades of Type I genes, Mα and Mβ/γ (Gramzow et al., 2012), are marked by labelled lines on the right. The colours of the branches correspond to the colours of the gymnosperm genera in Fig. 1. Abbreviations of gymnosperm gene names are described in Table 1. Lycophyte sequences are from Selaginella moellendorffii and moss sequences are from Physcomitrella patens. Clades of angiosperm genes are collapsed into triangles and their classification according to (Parenicova et al., 2003; Leseberg et al., 2006; Arora et al., 2007) is shown on the right. Numbers at nodes denote bootstrap values. The fully resolved phylogeny is available as Supplementary Data Fig. S2.

As mentioned above, the number of Type I genes identified in gymnosperms is very low. The 35 Type I genes of gymnosperms are distributed approximately evenly between the two clades with 14 genes belonging to putative clade I (Mα) and 21 genes belonging to putative clade II (Mβ/Mγ). Evidence of transcription exists only for few of these genes. Our analysis of transcriptome data revealed that two genes of clade I are expressed in mixed shoot tissues. For clade II, four genes were found in transcriptome data derived from bud, male cone and embryo tissues.

Clades of Type II genes

Based on our phylogeny of Type II sequences (Supplementary Data Fig. S3), we defined 11 branches (putative clades, hereafter termed ‘clades’ for simplicity) for separate analyses (MIKC*, AGL2/AGL6/FLC/SQUA, DEF/GLO/OsMADS32/GGM13, AGAMOUS, AGL12, AGL15, AGL17, GpMADS4, StMADS11, TM3 and TM8). All of these clades contain gymnosperm sequences (Table 2).

Table 2.

Number of gymnosperm genes in different clades of Type II MADS-box genes

Species	MIKC*	AGL2/AGL6/FLC/SQUA	DEF/GLO/OsMADS32/GGM13	AGAMOUS	AGL12	AGL15	AGL17	GpMADS4	StMADS11	TM3	TM8
S. sempervirens	0	0	0	0	0	0	0	0	8	4	4
C. japonica	0	1	6	1	0	0	0	0	0	0	1
T. baccata	0	0	0	0	0	0	0	0	1	1	1
C. harringtonia	0	3	0	0	0	0	0	0	6	10	13
S. verticillata	0	1	0	0	0	0	0	0	0	6	14
W. nobilis	1	0	0	0	0	0	0	0	4	0	6
P. macrophyllus	0	1	1	0	0	0	0	0	6	1	6
P. taeda	1 (41)	11 (38)	2 (64)	3 (24)	4 (15)	0 (0)	0 (4)	0 (2)	7 (20)	25 (77)	6 (58)
P. pinaster	0	1	0	0	0	0	0	0	2	4	3
P. palustris	0	1	0	0	0	0	0	0	0	15	5
P. lambertiana	0	6	1	0	1	0	0	2	6	19	6
P. contorta	0	1	0	0	0	0	0	1	2	8	2
P. banksiana	0	0	0	0	0	0	0	0	0	1	1
P. abies	4 (5)	2 (15)	8 (20)	2 (15)	0 (5)	0 (1)	0 (13)	2 (9)	14 (53)	14 (85)	4 (27)
P. glauca	0 (1)	3 (10)	1 (8)	2 (5)	1 (6)	0 (0)	0 (6)	0 (4)	7 (26)	8 (32)	1 (17)
P. sitchensis	0	1	0	1	0	0	0	0	5	6	3
P. menziesii	0	1	2	2	0	0	0	0	1	31	3
C. atlantica	0	2	0	0	0	0	0	0	3	8	0
G. gnemon	2	4	3	4	0	0	0	0	4	2	18
Sum	8 (50)	39 (86)	24 (105)	15 (52)	6 (27)	0 (1)	0 (23)	5 (18)	76 (147)	163 (310)	97 (188)

Generally, the number of expressed genes is given. For P. taeda, P. abies, P. glauca and sum, the total number of genes including those solely identified from genome data without evidence of transcription is given in parentheses.

MIKC*-group genes

In our type II phylogeny, 50 gymnosperm sequences grouped together with MIKC*-group genes from mosses, ferns and angiosperms (Supplementary Data Fig. S3). Only eight of these are supported by expression data. The bryophyte sequences branch off before the differentiation of the S- and P-clade, each containing sequences from ferns, gymnosperms and angiosperms (Supplementary Data Fig. S4). In the phylogeny containing only sequences with expression data (Fig. 3), PaMADS17 is the only gymnosperm member of the S-clade. Interestingly, also 39 genomic sequences from P. taeda without evidence of transcription belong to the S-clade in our phylogeny. Ten gymnosperm sequences (seven with evidence of transcription) form the P-clade together with sequences from ferns and angiosperms. The association of the gymosperm sequences to the P- and S-clades in our MrBayes phylogeny is consistent with that in our RAxML phylogenies (Supplementary Data Fig. S4). The identified MIKC*-group genes of the P-clade, PaMADS29, PaMADS31 and PaMADS26 from P. abies, show expression in stem, wood, vegetative shoots and female cones. PaMADS26 was also identified from male cones. cDNA of GgMADS1 from G. gnemon was isolated from female cones. In contrast, the expression of the single S-clade gene PaMADS17 was detected only in male cones.

Fig. 3.

Phylogeny of MIKC*-group genes. Colours of the branches, gene names, triangles and numbers at nodes are as described in Fig. 2. The subclades S and P (Nam et al., 2003) are marked by labelled lines on the right. The liverwort sequence is from Marchantia polymorpha, fern sequences are from Ceratopteris richardii, lycophyte sequences are from Selaginella moellendorffii, moss sequences are from Physcomitrella patens and the separate angiosperm sequences are from Vitis vinifera and from Populus trichocarpa. The fully resolved phylogeny is available as Supplementary Data Fig. S4b.

AGL2/AGL6/FLC/SQUA-like genes

Recently, it was established that FLC-like genes belong to the superclade of AGL2-, AGL6- and SQUA-like genes (Ruelens et al., 2013). Hence, we here analysed these genes together. In our phylogenies, two large subclades can be defined, one consisting of AGL2- and AGL6-like genes and the other consisting of FLC- and SQUA-like genes (Supplementary Data Fig. S5a). Of the gymnosperm sequences identified here, 19 belong to the clade of AGL2/AGL6-like genes and 67 belong to the clade of FLC/SQUA-like genes. When we restrict our phylogeny to sequences for which we have evidence of transcription, the AGL2- and the AGL6-like genes of angiosperms form sister clades as well as the FLC- and SQUA-like genes of angiosperms, with the exception of FLC-like genes from rice, which cluster within a clade of gymnosperm sequences (Fig. 4). This suggests that the duplications giving rise to AGL2- and AGL6-like genes and to FLC- and SQUA-like genes occurred in angiosperms and that the MRCA of extant seed plants possessed at least one AGL2/AGL6-like gene and at least one FLC/SQUA-like gene. As the bootstrap values for our RAxML phylogenies were quite low we also reconstructed a phylogeny using MrBayes (Supplementary Data Fig. 5c). The sister-group relationships of AGL2- and AGL6-like genes and of FLC- and SQUA-like genes of angiosperms was confirmed in our MrBayes phylogeny with stronger support (posterior probability of 0·51 and 0·90, respectively). However, all gymnosperm sequences appear more closely related to AGL2/AGL6-like genes than to FLC/SQUA-like genes in our MrBayes phylogeny.

Fig. 4.

Phylogeny of AGL2/AGL6/FLC/SQUA-like genes. Colours of the branches, gene names, triangles and numbers at nodes are as described in Fig. 2. The subclades AGL2/AGL6 and FLC/SQUA are marked by labelled lines on the right. The fully resolved phylogeny is available as Supplementary Data Fig. S5b.

Analysing transcriptome data, we found expression of gymnosperm genes belonging to the AGL2/AGL6 clade in shoots, needles and reproductive tissues. FLC/SQUA-like genes from gymnosperms were identified in transcriptome data from a wide variety of tissues (roots, shoots, stems, bark and female cones).

DEF/GLO/OsMADS32/GGM13-like genes

In our phylogenies, OsMADS32-like genes cluster with DEF-like, GLO-like and GGM13-like genes (Supplementary Data Fig. S3). Conse-quently, we analysed these clades together. Based on our Type II phylogeny including MADS sequences identified from genome and transcriptome projects, 105 sequences belong to this clade (Supplementary Data Fig. S3a). In our separate phylogenies for this superclade, DEF- and GLO-like genes from angiosperms form sister clades that in turn are sister clades to OsMADS32-like genes of angiosperms (Supplementary Data Fig. S6a). This suggests that there have been two duplications near the base of extant angiosperms giving rise first to OsMADS32- and DEF/GLO-like genes and then to DEF- and GLO-like genes. A clade of 60 MADS sequences from gymnosperms is sister to the DEF/GLO/OsMADS32-superclade. GGM13-like genes from angiosperms form a clade in our phylogeny to which 45 MADS sequences from gymnosperms are related.

When we exclude gymnosperm MADS sequences solely predicted on genomic sequences (Fig. 5), 24 MADS sequences from eight gymnosperm species cluster with DEF/GLO/OsMADS32-like genes from angiosperms and nine MADS sequences from five gymnosperm species cluster with GGM13-like genes from angiosperms. The phylogeny and distribution of the sequences suggests that there has been at least one DEF/GLO/OsMADS32-like gene and one GGM13-like gene in the MRCA of extant seed plants. In our MrBayes phylogeny, some gymnosperm sequences are sister to the OsMADS32- and to the DEF-like genes from angiosperms, respectively, rather than being ancestral to a superclade of DEF-, GLO- and OsMADS32-like genes from angiosperms (Supplementary Data Fig. 6c).

Fig. 5.

Phylogeny of DEF/GLO/OsMADS32/GGM13-like genes. Colours of the branches, gene names, triangles and numbers at nodes are as described in Fig. 2. The subclades DEF/GLO/OsMADS32 and GGM13 are marked by labelled lines on the right. The fully resolved phylogeny is available as Supplementary Data Fig. S6b.

Gymnosperm DEF/GLO/OsMADS32-like sequences are mainly derived from transcriptome data of male reproductive tissues. GGM13-like genes were mainly identified from mixed tissues. Only for GGM13-like genes of P. abies are specific tissues, namely female reproductive organs, given.

AGAMOUS- and AGL12-like genes

In our phylogenies, 52 of all identified gymnosperm MADS sequences belong to the AGAMOUS clade (Supplementary Data Fig. S7a). When we exclude sequences for which no evidence of transcription exists, 15 sequences remain. In this reduced phylogeny, all AG-like genes of gymnosperms are sister to the AG-like genes of angiosperms, suggesting that the MRCA of extant seed plants had at least one but may not have had more than one AG-like gene (Fig. 6). Furthermore, the AG-like genes of gymnosperms form three subclades in our RAxML phylogeny and two subclades in our MrBayes phylogeny (Supplementary Data Fig. 7c), indicating that the MRCA of extant gymnosperms may have possessed at least two to three AG-like genes. Interestingly, AG-like sequences derived from transcripts were identified from a variety of tissues, including roots, shoots, stems, bark, leaves and reproductive organs.

Fig. 6.

Phylogeny of AG-like genes. Colours of the branches, gene names, triangles and numbers at nodes are as described in Fig. 2. Three putative subclades of AG-like genes in gymnosperms, GBM5, GGM3 and DAL2, are marked by labelled lines on the right. The fully resolved phylogeny is available as Supplementary Data Fig. S7b.

The sister group of AG-like genes are the AGL12-like genes (Becker and Theissen, 2003). We identified 27 AGL12-like sequences from gymnosperms, where there is evidence of transcription for six of them (Fig. 7, Supplementary Data Fig. S8). Expression was mostly found in mixed tissues, twice in root and once in shoot and/or needles. Our phylogeny suggests that one AGL12-like gene was present in the MRCA of extant seed plants.

Fig. 7.

Phylogeny of AGL12-like genes. Colours of the branches, gene names, triangles and numbers at nodes are as described in Fig. 2. The fully resolved phylogeny is available as Supplementary Data Fig. S8b.

AGL15- and AGL17-like genes

In our Type II phylogeny, one sequence of P. abies clusters with AGL15-like genes from angiosperms and 23 sequences from P. taeda, P. abies and P. glauca cluster with AGL17-like genes of angiosperms (Supplementary Data Figs S3, S9 and S10). However, all of these gymnosperm sequences have been identified solely from genomic sequences and no evidence of transcription exists for these genes. Alignments reveal strong similarity of at least some of these gymnosperm sequences to the AGL15- and AGL17-like MADS sequences of angiosperms, respectively (Figs 8 and 9). Nevertheless, the alignments also reveal three positions each where the gymnosperm sequences are different from all of the angiosperm MADS sequences of the respective clades. Hence, further analyses are needed to confirm or deny the association of the gymnosperm sequences to the clades of AGL15- and AGL17-like genes.

Fig. 8.

Phylogeny of AGL15-like genes (A) and alignment of the MADS domains of AGL15-like proteins (B). For the phylogeny, colours of the branches, gene names, triangle and number at nodes are as described in Fig. 2. The alignment is coloured according to conservation and biochemical properties of amino acids (Thompson et al., 1997) and shows only non-conserved residues. Names of angiosperm genes are shaded grey. Positions at which the gymnosperm sequence differs from all angiosperm sequences are boxed. The consensus sequence is given below the alignment.

Fig. 9.

Phylogeny of AGL17-like genes (A) and alignment of the MADS domains of AGL17-like proteins (B). For the phylogeny, colours of the branches, gene names, triangle and numbers at nodes are as described in Fig. 2. Colouring and labelling of the alignment is described in Fig. 8. Positions at which the gymnosperm sequences differ from all angiosperm sequences are boxed. The consensus sequence is given below the alignment.

GpMADS4-like genes

We found 18 sequences belonging to the gymnosperm-specific clade of GpMADS4-like genes (Supplementary Data Fig. S3). The known genes DAL10 of P. abies and GpMADS4 of G. parvifolium constitute the GpMADS4-subclade together with 13 other gymnosperm sequences (Supplementary Data Fig. S11). DAL21 of P. abies forms another subclade together with three other gymnosperm sequences, termed the DAL21-subclade here. These clades are also stable in our MrBayes phylogeny (Supplementary Data Fig. S11b). For both subclades, there are three genes supported by expression data (Fig. 10), where expression was found in bark, shoots, stems, needles, buds and cones.

Fig. 10.

Phylogeny of GpMADS4-like genes. Colours of the branches, gene names and numbers at nodes are as described in Fig. 2. The two putative subclades GpMADS4 and DAL21 are marked by labelled lines on the right.

StMADS11-like genes

In our phylogenies, 147 gymnosperm sequences cluster together with StMADS11-like genes from angiosperms, of which 76 gymnosperm sequences have expression data support (Supplementary Data Fig. S12a). There are two subclades of StMADS11-like genes from conifers in our phylogeny (Fig. 11), suggesting that there was a duplication of an ancestral StMADS11-like gene near the base of extant conifers. We termed the two resulting subclades PaMADS19-like and PaMADS20-like genes after the P. abies gene with the lowest number in the corresponding clade. These two subclades also appear in our MrBayes phylogeny (Supplementary Data Fig. S12c). For the PaMADS19 subclade, sequences of all the conifer species studied here were found with the exception of those of S. verticillata. The subclade PaMADS20 contains only sequences of P. macrophyllus and the family Pinaceae. In both subclades several species-specific duplications occurred, for example in S. sempervirens, W. nobilis, C. harringtonia, P. macrophyllus and species of the family Pinaceae. Expression of gymnosperm StMADS11-like genes was found mainly in shoots, but also in other tissues such as roots, stems, wood, bark and needles.

Fig. 11.

Phylogeny of StMADS11-like genes. Colours of the branches, gene names, triangle and numbers at nodes are as described in Fig. 2. The two putative subclades of StMADS11-like genes in gymnosperms, PaMADS19 and PaMADS20, are as marked by labelled lines on the right. The fully resolved phylogeny is available as Supplementary Data Fig. S12b.

TM3-like genes

The clade of TM3-like genes includes 310 gymnosperm sequences from the species studied here, 163 of which have expression support, and four sequences from Pinus radiata (Supplementary Data Fig. S13). In our reduced phylogeny there are two main clades of gymnosperm sequences, one comprising genes from the conifer species S. verticillata, S. sempervirens, T. baccata and C. harringtonia (Fig. 12), and the other containing sequences of G. gnemon, P. macrophyllus and Pinaceae species. The clade of Pinaceae TM3-like genes can be further subdivided into three subclades, termed DAL19-, DAL3- and PrMADS4-like genes here. There are a number of species-specific expansions of TM3-like genes, for example in Pinaceae, S. verticillata, S. sempervirens and C. harringtonia. The MrBayes phylogeny shows different subclades to those observed in our RAxML phylogenies (Supplementary Data Fig. S13c). Hence, even though the sheer number of TM3-like genes in conifers indicates a number of duplications, the timing and extent of duplications cannot be determined based on our data. Transcripts of TM3-like genes were isolated from various tissues, such as shoots, stems, needles, buds, and female and male cones.

Fig. 12.

Phylogeny of TM3-like genes. Colours of the branches, gene names, triangle and numbers at nodes are as described in Fig. 2. Three putative subclades of TM3-like genes in Pinaceae, DAL19, DAL3 and PrMADS4, are marked by labelled lines on the right. The fully resolved phylogeny is available as Supplementary Data Fig. S13b.

TM8-like genes

The angiosperm TM8-like genes group with 188 MADS-sequences from gymnosperms (Supplementary Data Fig. S14a). For 97 sequences expression data are available. In the reduced phylogeny including only genes with expression data, two subclades are evident, termed as GgMADS2-like and GgMADS25-like genes (Fig. 13). The clade of GgMADS2 contains numerous sequences of the family Pinaceae and 16 sequences of G. gnemon. The Pinaceae sequences can be further subdivided into the subclades PaMADS15 and PaMADS24. Transcript-based sequences belonging to these two subclades were isolated from buds, shoots, needles and stems and only one sequence of P. taeda PtaMADS27, belonging to the PaMADS15 subclade, was found to be expressed in roots. The clade of GgMADS25-like genes contains sequences of non-Pinaceae conifers and two sequences from G. gnemon. The clades of GgMADS25-like genes also appears in our MrBayes phylogeny (Supplementary Data Fig. S14c). However, the genes belonging to the GgMADS2-like clade in our RAxML phylogeny are split into two clades that are not sister clades in our MrBayes phylogeny. Expression of genes from this clade was also found in a variety of tissues, such as shoots, needles and male strobili.

Fig. 13.

Phylogeny of TM8-like genes. Colours of the branches, gene names, triangle and numbers at nodes are as described in Fig. 2. Two putative subclades of TM8-like genes in gymnosperms, GgMADS2 and GgMADS25, are marked by labelled lines on the right. Two putative subclades of the GgMADS2-subclade in Pinaceae, PaMADS15 and PaMADS24, are likewise marked by labelled lines. The fully resolved phylogeny is available as Supplementary Data Fig. S14b.

DISCUSSION

Large total number of MADS-box genes but low percentage of Type I genes in conifer genomes

The total number of MADS-box sequences identified from the three sequenced conifer genomes ranges from 121 to 367 with an average of 249. In sequenced angiosperm genomes, this number ranges from 60 to 265 where the average number is 116 (Gramzow and Theißen, 2013). Hence, the number of putative MADS-box genes (including pseudogenes) in conifers is on average much higher than in angiosperms. This is consistent with the large genome size of gymnosperms. The genomes studied here of Picea abies, Picea glauca and Pinus taeda have genome sizes of more than 20 Gbp (Birol et al., 2013; Nystedt et al., 2013). Despite the large genome sizes, there is no evidence for genome duplications in the gymnosperm lineage (Nystedt et al., 2013). The total number of genes also seems to be higher in gymnosperms than in most angiosperms. In P. abies, 70 968 protein-coding loci were predicted, of which 28 354 were considered to be of high confidence (Nystedt et al., 2013). Hence, the large number of MADS-box genes we found in gymnosperm genomes correlates with the large genome size and gene number in gymnosperms. However, many of the MADS-box genes we found are probably pseudogenes. We identified premature stop codons in the MADS boxes of 17 sequences from P. abies, eight sequences from P. glauca and 22 sequences from P. taeda. A complete analysis of the number of pseudogenes in the MADS-box gene family is not possible yet due to the short length of some scaffolds on which MADS-boxes have been identified. A large fraction of pseudogenes has also been identified for other gene families in gymnosperms, such as phytochrome, cdc2 and WUSCHEL-type homeobox genes (Kinlaw and Neale, 1997; Kvarnheden et al., 1998; Garcia-Gil, 2008; Hedman et al., 2013). How many of the identified MADS-box genes in conifers are functional remains to be determined. The number of genes supported by transcripts identified for P. abies, P. glauca and P. taeda ranges only from 23 to 60. This low number of MADS-box genes for which transcriptome data were found may have several reasons. First, the number of different tissues sampled was low to moderate. For P. abies, 22 transcriptome samples of different tissues were determined (Nystedt et al., 2013) while for P. taeda and P. glauca six and eight transcriptome datasets, respectively, were available (Wegrzyn et al., 2008; Lorenz et al., 2012). Furthermore, many MADS-box genes are known to be expressed at low level or only at specific stages of development (Becker and Theissen, 2003; De Bodt et al., 2003; Nam et al., 2004; Bemer et al., 2010). Hence, more transcriptome data may help to identify more MADS-box genes as expressed genes. However, as mentioned above, a number of the MADS-box genes we identified may also be pseudogenes.

In contrast to the large overall number of identified MADS-box genes in conifer genomes, the number of Type I MADS-box genes ranges from only three to 17 in the sequenced conifer genomes. Only two angiosperm genomes examined so far have fewer than 17 Type I MADS-box genes, namely Cucumis sativus (11) and Zea mays (14) (Gramzow and Theißen, 2013). Percentage-wise, the difference between the amount of Type I genes in conifer and angiosperm genomes is even clearer: in all angiosperm genomes that have been examined, the percentage of Type I genes is greater than 20 % of all MADS-box genes, while in examined conifers, the percentage is always lower than 5 %. For angiosperms, it has been shown that Type I genes have higher birth-and-death rates than Type II genes (Nam et al., 2004). Our phylogeny of Type I genes shows lineage-specific expansions of Type I genes for conifers, similar to what is observed for Type I genes in angiosperms (Fig. 2). However, there are fewer clades of conifer Type I genes, which generally also have fewer genes than the clades of angiosperm Type I genes (Supplementary Data Fig. S2). Hence, either the birth rate of Type I genes is lower or the death rate of Type I genes is higher in conifers than in angiosperms. Further studies are needed to clarify the evolutionary patterns of Type I genes in conifers.

Strengths and weaknesses of MADS-box gene phylogenies

In general, the support values in our phylogenies are quite low. This may be due to large datasets including quite diverse sequences and is common for MADS-box gene phylogenies in plants (Gramzow and Theißen, 2013). Therefore, we used two independent phylogenetic reconstructions using ML and Bayesian methods to test the stability of clades. Furthermore, we often observe paraphyletic relationships between the different gymnosperm sequences (Figs 2–6, 9, 11 and 13). Long branch attraction may, at least partially, explain the apparent paraphyletic relationship even though the methods we used here are less prone to long branch attraction than other phylogenetic methods (Bergsten, 2005). Also, some of the gymnosperm sequences are very short and may not provide enough information for the phylogenetic reconstruction algorithms to correctly place them in the phylogeny. Hence, the paraphyletic pattern of gymnosperm genes may often be an artefact and we largely ignored paraphyletic relationships and rather assumed monophyly of gymnosperm genes when estimating the number of MADS-box genes in the MRCAs of seed plants and gymnosperms.

Number of MADS-box genes in the MRCA of extant seed plants

According to our phylogenies, the MRCA of extant seed plants contained at least 13–16 different MADS-box genes (Fig. 14). Of these, at least two were Type I genes [one Type I gene of clade I (Mα) and one Type I of clade II (Mβ/Mγ)], at least two were MIKC*-group genes [one MIKC*-group gene of the S clade and one MIKC*-group gene of the P clade (Nam et al., 2004)] and at least 9–12 were MIKC^C-group genes (one each of the following clades: AGL2/AGL6, FLC/SQUA, DEF/GLO/OsMADS32, GGM13, AG, AGL12, StMADS11, TM3 and TM8; and possibly AGL15, AGL17 and GpMADS4). For MIKC^C-group genes, the presence of AGL15- and AGL17-like genes needs to be verified as thus far gymnosperm genes with high similarity to the angiosperm AGL15- and AGL17-like genes have been identified solely based on genomic data and no evidence of transcription exists. Furthermore, support values for the corresponding clades are quite low and alignments of the MADS domains do not give a clear signal of whether the gymnosperm sequences belong to the AGL15- and AGL17-clades or not. The clade of GpMADS4-like genes is specific to gymnosperms. The deep branching of this clade in relation to the other ancient clades of Type II genes was not clear in our phylogenies. As it does not form a clear clade with another ancient clade of Type II genes, we hypothesize that this clade was present in the MRCA of seed plants and was subsequently lost in angiosperms. Given that phylogenies from previous studies also did not reveal any clear association of this clade with another major clade of MADS-box genes (Melzer et al., 2010; Carlsbecker et al., 2013), this hypothesis seems reasonable. However, it remains possible that phylogenic reconstructions have failed to place GpMADS4-like genes correctly thus far. Hence, more transcript sequences from gymnosperms may help to clarify the origin of AGL15-, AGL17- and GpMADS4-like genes.

Fig. 14.

Clades of MADS-box genes plotted on a simplified phylogeny of vascular plants. The crown groups of gymnosperms and angiosperms are shown as open triangles. At each node, the clades of MADS-box genes that had been established at this node according to our phylogenies are listed. The clades are sorted into three columns at each node where the first column lists Type I genes, the second column lists MIKC*-group genes and the third column lists MIKC^C-group genes. Clade names generally refer to the first member that had been published, except for Mα and Mβ/γ, which denote different groups of Type I MADS-box genes, MIKC^c and MIKC*, which refer to the two groups of Type II MIKC-type genes, and MIKC* S and MIKC* P, which are subclades of MIKC*-group genes. Clade names composed of clade names separated by slashes (‘/’) refer to clades that have not undergone the corresponding split. For example, the AGL2/AGL6-like genes in the stem group of seed plants gave rise to the clades of AGL2-like, AGL6-like and AGL9-like genes in the MRCA of extant angiosperms. MIKC^c-group clades belonging to the same ancestral seed plant clade are shown in the same colour. A question mark behind a clade name indicates that the presence of this clade in the corresponding MRCA is less strongly supported than that of the other clades because putative clade members were only found in genome but not transcriptome data (AGL15 and AGL17) or because of ambiguities in tree topology (GpMADS4). The putative loss or absence of GpMADS4-like genes early in the evolution of angiosperms is indicated by minus signs. For lycophytes, the clades given at the terminal branch indicate that at least one clade member has been identified in at least one extant species of lycophytes. For angiosperms and gymnosperms, the clades given have been established in the stem group except for the two clades of TM8-like genes and the two clades of GpMADS4-like genes in gymnosperms (indicated by ^x). In these cases, our taxon sampling indicates presence of these clades in the MRCA of gnetophytes and conifers but as data from cycads and Ginkgo are missing we cannot infer presence in the MRCA of all extant gymnosperms.

Our work extends previous studies and increases the minimum number of MADS-box genes identified in the MRCA of extant seed plants by one to four. Earlier studies recognized the presence of two ancient clades of Type I genes that trace back close to the MRCA of land plants (Gramzow et al., 2012). At this stage, however, no gymnosperm Type I genes of clade I (Mα) could be identified but only a few gymnosperm Type I genes of clade II (Mβ/γ). Here we demonstrate that Type I genes of clade I are also present in gymnosperms. Similarly, the presence of two ancient clades of MIKC*-group genes that originated in the MRCA of euphyllophytes, the P- and the S-clade, has been described (Kwantes et al., 2012; Gramzow et al., 2012). In these previous studies, however, only gymnosperm members of the P-clade had been found whereas here we identified gymnosperm genes of both clades. At least eight MIKC^C-group genes were predicted to be present in the MRCA of seed plants (Becker et al., 2000; Melzer et al., 2010). Our analyses reveal that this number should be increased to 9–12. In contrast to previous studies (Becker et al., 2000; Melzer et al., 2010; Kim et al., 2013), our phylogenies indicate that the duplication giving rise to AGL2- and AGL6-like genes may be specific to angiosperms and that the genes previously thought to be orthologues of AGL6-like genes in gymnosperms are actually orthologues to an ancestral AGL2/AGL6-like gene (Fig. 4). According to our scenario, only one AGL2/AGL6-like gene, instead of one AGL2- and one AGL6-like gene, was present in the MRCA of extant seed plants. Consequently, we do not have to assume a loss of AGL2-like genes in gymnosperms but two consecutive duplications at the base of angiosperms giving rise first to AGL2- and AGL6-like genes and then to SEP1- (AGL2-) and SEP3- (AGL9-) like genes. In line with a recent study (Ruelens et al., 2013), FLC- and SQUA-like genes of angiosperms form a clade in our phylogenies with the exception of FLC-like genes from rice. Neighbouring this clade are a number of MADS-box genes from gymnosperms. This suggests that FLC/SQUA-like genes are present in gymnosperms and that the MRCA of extant seed plants possessed at least one FLC/SQUA-like gene. Previous studies could not identify SQUA-like genes in gymnosperms (Becker et al., 2000; Melzer et al., 2010). Furthermore, our study indicates that AGL15-, AGL17- and GpMADS4-like genes were also present in the MRCA of extant seed plants. However, this finding needs to be tested with additional data, as described above. Finally, our study clarifies the origin of OsMADS32-like genes, which had previously been thought to be specific to monocots (Sang et al., 2012). In our phylogeny, we identify an OsMADS32-like gene in the basal angiosperm species Amborella trichopoda, Am.tr.OsM32 (Supplementary Data Fig. S6) as confirmed by other studies (The Amborella Genome Project, 2013). The clade of OsMADS32-like genes is sister to a clade of DEF- and GLO-like genes (Fig. 5), suggesting that OsMADS32-like genes originated by a duplication of an ancestral DEF/GLO/OsMADS32-like gene near the base of extant angiosperms.

Number of MADS-box genes in ancestral gymnosperms

From our phylogenies, we can infer some duplications in the lineage leading to the MRCA of gnetophytes (represented by Gnetum) and conifers after the lineage that led to angiosperms split off (Fig. 14). One duplication each happened in the TM8-clade, leading to GgMADS2- and GgMADS25-like genes, and in the GpMADS4-clade, generating GpMADS4- and DAL21-like genes. Hence, the number of MIKC^C-group MADS-box genes is nearly as high in the MRCA of gnetophytes and conifers (14) as in the MRCA of angiosperms (17). If recent phylogenetic reconstructions are correct (Wu et al., 2011; Xi et al., 2013), data from Ginkgo and cycads will be required to determine whether the two gene duplications discussed above pre-date even earlier diversifications of extant gymnosperms than the split between gnetophytes and conifers.

Inferred functions of MADS-box genes in the MRCA of seed plants

Type I MADS-box genes

Only a few studies of Type I MADS-box gene functions have been published so far (Portereiko et al., 2006; Yoo et al., 2006; Bemer et al., 2008; Colombo et al., 2008; Kang et al., 2008; Steffen et al., 2008). These studies suggest as commonality a role for Type I MADS-box genes in female gametophyte, embryo and seed development in angiosperms (Gramzow and Theissen, 2010). Expression of one gymnosperm Type I gene was also found in embryo. Hence, a role in embryo development may be an ancestral function of Type I genes in seed plants. The expression of other gymnosperm Type I genes in a wide range of tissues (e.g. shoots, needles and cones) may represent transcriptional noise, indicate additional functions of gymnosperm Type I genes or restriction of Type I gene functions in angiosperms. Unfortunately, there are currently no data available, such as detailed gene expression pattern or mutant phenotypes, that would clarify the function of Type I genes in gymnosperms.

MIKC*-group genes

In ferns, these genes are expressed in male and hermaphroditic gametophytic tissue and in sporophytic tissue such as roots and stipes (Kwantes et al., 2012). In contrast, their expression is usually restricted to male gametophytic tissue (pollen) in basal and in derived angiosperms (Kwantes et al., 2012; Liu et al., 2013). The identified gymnosperm MIKC*-group genes are expressed in gametophytic and sporophytic tissues as well, indicating that no restriction to male gametophytic tissue occurred before the split of angiosperms and gymnosperms. Our analyses of transcriptome data reveal that P-clade genes are expressed in vegetative shoots, wood, and female and male reproductive organs, while the S-clade gene was found to be expressed in male reproductive organs. The broad expression pattern of genes of the P-clade suggests diverse functions during plant development, whereas S-clade genes seem to have a more restricted function in specifying male organs in gymnosperms. However, further investigations are needed to clarify the functions of MIKC*-group genes of extant gymnosperms. The data suggest that MIKC*-group genes may have had a role in the development of both male and female reproductive organs and in the development of vegetative tissues in the MRCA of extant seed plants. After the divergence of the lineages that led to extant gymnosperms and angiosperms, MIKC*-group genes became functionally restricted to the male gametophyte in angiosperms while they may have kept a broader role in gymnosperms.

AGL2/AGL6- and FLC/SQUA-like genes

The gymnosperm AGL2/AGL6-like genes identified here are expressed in shoots, needles and reproductive tissues. DAL1 and DAL14 from P. abies, which have been described as AGL6-like genes previously, are expressed in male and female cones (Carlsbecker et al., 2004, 2013). DAL1 is additionally expressed in vegetative tissues and has therefore been proposed to be involved in vegetative development and to regulate phase change from juvenile to adult (Carlsbecker et al., 2004). Angiosperm AGL2- and AGL6-like genes have roles in the transition to flowering and in lateral organ and flower development (Pelaz et al., 2000; Koo et al., 2010; Yoo et al., 2011). It has also been shown that some AGL6-like genes can act redundantly with AGL2-like genes in flower organ formation (Rijpkema et al., 2009). Combining this information suggests that ancestral AGL2/AGL6-like genes may have had roles in the transition to reproductive development and in the development of vegetative and reproductive organs. Angiosperm FLC-like genes have roles in developmental phase changes (Michaels and Amasino, 2001; Deng et al., 2011). The expression of gymnosperm FLC/SQUA-like genes identified here in a number of vegetative tissues may point to a similar role for these genes. Furthermore, gymnosperm FLC/SQUA-like genes are expressed in female cones. The function of SQUA-like genes in angiosperms in the development of the sterile organs of the flower and in meristem identity specification (Mandel et al., 1992; Ferrándiz et al., 2000) may represent a new function for these genes in angiosperms.

DEF/GLO/OsMADS32-like and GGM13-like genes

The gymnosperm genes that are sister to the clades of DEF-, GLO- and OsMADS32-like genes of angiosperms were found to be expressed in male reproductive tissues. This is consistent with previous reports about the expression of GGM2 of G. gnemon in male cones, of DAL11, DAL12 and DAL13 of P. abies in male bud meristems, and of PrDGL of P. radiata as well as of CjMADS1 and CjMADS2 of C. japonica in male strobili (Mouradov et al., 1999; Winter et al., 1999; Fukui et al., 2001; Sundstrom and Engstrom, 2002). Similarly, DEF- and GLO-like genes in angiosperms are expressed in male reproductive organs and in petals and specify stamen and petal identity (Schwarz-Sommer et al., 1990; Goto and Meyerowitz, 1994). These expression patterns point to an ancestral function of DEF/GLO/OsMADS32-like genes in the development of male reproductive organs. In contrast, gymnosperm GGM13-like genes are mainly expressed in female reproductive organs (Becker et al., 2002; Carlsbecker et al., 2013; Lovisetto et al., 2013), which is again consistent with what is known about the expression of GGM13-like genes in angiosperms (Becker et al., 2002; Yang et al., 2012). Hence, these genes may have an ancestral function in the development of female reproductive organs.

AG-like and AGL12-like genes

In previous publications, expression of gymnosperm AG-like genes has been observed mainly in reproductive organs, such as DAL2 of P. abies in female cones, JcMADS2 of Juniperus communis, TdMADS3 of Thujopsis dolabrata and CjMADS4 of C. japonica in seed cones and pollen cones, and SAG1 of P. mariana and GGM3 of G. gnemon in male and female cones (Rutledge et al., 1998; Tandre et al., 1998; Winter et al., 1999; Englund et al., 2011; Groth et al., 2011). However, GBM5 of G. biloba was reported to be expressed not only in reproductive organs, but also in vegetative leaves (Jager et al., 2003). Interestingly, our analysis of transcriptome data reveals expression of gymnosperm AG-like genes in many different tissues ranging from roots via shoots and leaves to reproductive organs. The possibly wide expression pattern of gymnosperm AG-like genes as observed in G. biloba and in transcriptome data does not completely comply with the function of AG-like genes in angiosperms in the development of reproductive organs and roots (Yanofsky et al., 1990; Liljegren et al., 2000; Pinyopich et al., 2003; Moreno-Risueno et al., 2010). The expression of gymnosperm AG-like genes in other tissues may represent just transcriptional noise, or the ancestral function of AG-like genes in the MRCA of seed plants might have been broader, and AG-like genes in angiosperms lost functions outside of roots and reproductive organs.

The expression of gymnosperm AGL12-like genes in roots and shoots/needles was found in trancriptome data here and had previously been described for DAL5 of P. abies (Carlsbecker et al., 2013). This fits with the expression of AGL12-like genes from angiosperms in roots, leaves and floral meristems and their function in root development and transition to reproductive development (Tapia-López et al., 2008). Hence, the ancestral function of AGL12-like genes in the MRCA of extant seed plants probably involved root development and phase change to reproductive development.

AGL12-like genes form the sister clade of AG-like genes (Becker and Theissen, 2003). The probably broad functions of the ancestral AG-like gene as described above and the probable involvement of the ancestral AGL12-like gene in root development and transition to reproductive development suggest that the ancestral AG/AGL12-like gene had broad functions as well, and that the functions of AGL12-like genes were restricted to root development and phase change to reproductive development early after the duplication leading to AG- and AGL12-like genes before the divergence of angiosperms and gymnosperms.

AGL15- and AGL17-like genes

The presence of AGL15- and AGL17-like genes in the MRCA of extant seed plants is not clear but is suggested by our phylogeny. For possible gymnosperm AGL15- and AGL17-like genes, no expression data are available. From angiosperms, some functions of AGL15- and AGL17-like genes are known, such as root and leaf development and transition to reproductive development (Zhang and Forde, 1998; Adamczyk et al., 2007; Kutter et al., 2007; Han et al., 2008). It is possible that MADS-box genes are involved in these processes in gymnosperms as well and that putative AGL15- and AGL17-like genes carry out these functions also in gymnosperms.

GpMADS4-like genes

Expression of GpMADS4 was detected in female reproductive organs (Shindo et al., 1999). However, this was the only tissue studied. The GpMADS4 orthologues GGM7 and DAL10 were found to be expressed during the development of female and male cones (Becker and Theissen, 2003; Carlsbecker et al., 2003, 2013). DAL10 has been hypothesized to be involved in meristem determination of reproductive buds establishing reproductive identity (Carlsbecker et al., 2013). In contrast to genes of the GpMADS4-subclade, DAL21 of the DAL21-subclade is specifically expressed in female cones (Carlsbecker et al., 2013). Hence, the ancestral GpMADS4-like gene in gymnosperms may have had a function in determining both types of reproductive structures. After the duplication event during gymnosperm evolution, the function of the DAL21-subclade may have been restricted to the development of female cones. According to our phylogeny, GpMADS4-like genes were present in the MRCA of seed plants but lost in angiosperms. Speculation about the function of GpMADS4-like genes in the MRCA of seed plants is difficult, but their expression patterns in extant gymnosperms suggest a role in reproductive organ development also in the MRCA of extant seed plants. This role may have been taken over by other genes or may have become dispensable in angiosperms leading to a loss of GpMADS4-like genes in angiosperms.

Expansions of the clades of StMADS11-, TM3- and TM8-like genes in gymnosperms

By far the highest numbers of MADS-box genes in gymnosperms were found for the StMADS11, TM3 and TM8 clades. MADS-box genes belonging to these clades were identified in nearly all gymnosperm species studied.

We identified two large subclades, PaMADS19- and PaMADS20-like genes of conifer StMADS11-like genes. Similarly, four subclades, SVP-, ZMM17-, MPF1- and MPF2-like genes, have been defined for angiosperm StMADS11-like genes (Khan and Ali, 2013). Angiosperm StMADS11-like genes are mainly expressed in vegetative tissues such as roots, leaves and shoots (Borner et al., 2000; Lee et al., 2000; Michaels et al., 2003; Fornara et al., 2008; Wingen et al., 2012). Different functions have been described for these genes. AGL24 and SVP of A. thaliana act as promoter and repressor of floral transition, respectively (Hartmann et al., 2000; Michaels et al., 2003). The three StMADS11-like genes of rice, OsMADS22, OsMADS47 and OsMADS55, are involved in the negative regulation of brassinosteroid responses (Fornara et al., 2008; Lee et al., 2008) and MPF2 of Physalis floridana is important for male fertility and the development of the ‘inflated calyx syndrome’ (He and Saedler, 2005). The identification of conifer StMADS11-like genes from transcriptome data of mixed shoots, developing buds and roots indicates that these genes may have diverse functions in conifers as well.

We observed two basal duplications of TM3-like genes from Pinaceae resulting in the three subclades DAL19-, DAL3- and PrMADS4-like genes. In P. radiata an expansion of the clade of TM3-like genes has been previously suggested (Walden et al., 1999). The four sequences studied by Walden et al. (1999) are distributed over the three subclades of our phylogeny. All PrMADS genes are expressed in male cones, roots, needles and shoots with the exception of PrMADS6, which is not expressed in roots. Furthermore, four TM3-like genes have been identified from P. abies previously (Tandre et al., 1995; Carlsbecker et al., 2013; Uddenberg et al., 2013) of which two – DAL3 and DAL19 – are expressed in seedlings, cambium, vegetative shoots, and female and male cones at different developmental stages. This indicates that conifer TM3-like genes may be involved in the development of different tissues. DAL19 is hypothesized to be involved in phase change from vegetative to reproductive development (Uddenberg et al., 2013). The angiosperm genes of the TM3-clade AGL20/SOC1 from A. thaliana, its paralogues AGL19, AGL42, AGL71 and AGL72, and the rice genes OsMADS50 and OsMADS56 act as key regulators of flowering time (Lee et al., 2000; Tadege et al., 2003; Schonrock et al., 2006; Ryu et al., 2009; Dorca-Fornell et al., 2011). Hence, a function in the transition to reproductive growth may have already been performed by the ancestral TM3-like gene in the MRCA of extant seed plants. Subsequently, some TM3-like genes evolved other functions, such as AGL14/XAL2 from A. thaliana, which is involved in root development (Garay-Arroyo et al., 2013) and AGL42/FYF, which is involved in flower senescence/abscission (Chen et al., 2011). Similarly, diversification of the TM3-like genes of conifers could have led to different functions similar to the evolution of TM3-like genes in angiosperms.

TM8 was first described in tomato where its transcripts were detected in gynoecia, stamens and petals (Pnueli et al., 1991). The TM8-like gene ERAF17 of cucumber is expressed in sepals, petals and ovaries of female flowers but not in male flowers (Ando et al., 2001). TM8-like genes have been lost in a number of angiosperm lineages such as Brassicaceae and Poaceae (Parenicova et al., 2003; Arora et al., 2007). In gymnosperms three TM8-like genes from G. biloba and one TM8-like gene from T. baccata were described and their expression patterns were examined recently (Lovisetto et al., 2012). GbMADS11 and GbMADS6 from G. biloba are expressed in leaves, male strobili and ovules, whereas GbMADS7 is expressed only weakly in these tissues. The TbTM8 gene from T. baccata is expressed strongly in ovules and developing arils. Two transcript sequences from C. japonica were identified from male strobili. The sequences from P. abies identified here were detected only in wood tissues and buds. Other conifer sequences were isolated from needles, shoot tips and wood tissues as well. Given the diverse expression patterns of TM8-like genes in gymnosperms and angiosperms, it is difficult to infer a function for the ancestral TM8-like gene in the MRCA of seed plants. Judged by the number of gene duplications and gene losses observed for this clade, it seems that TM8-like genes are a clade of fast evolving genes, which have acquired different functions in different species of conifers as well as of angiosperms.

CONCLUSIONS

We show that the minimal number of Type I genes was much lower than the minimal number of Type II MADS-box genes in the MRCA of seed plants. Our analysis of transcriptome data reveals that gymnosperm MADS-box genes are expressed in a great variety of tissues, indicating diverse roles of MADS-box genes for the development of gymnosperms. Our study is the first that provides a comprehensive overview about MADS-box genes in conifers and thus will provide a framework for future work on MADS-box genes in seed plants.

SUPPLEMENTARY DATA

Supplementary data are available online at www.aob.oxfordjournals.org and consist of the following. Table S1: MADS-box genes identified from gymnosperm genomes and transcriptomes. Fig. S1: RAxML phylogeny of MADS-box genes in plants. Fig. S2: RAxML phylogeny of Type I MADS-box genes. Fig. S3: RAxML phylogenies of Type II MADS-box genes. Fig. S4: phylogenies of MIKC* MADS-box genes. Fig. S5: phylogenies of AGL2/AGL6/FLC/SQUA-like MADS-box genes. Fig. S6: phylogenies of DEF/GLO/OsMADS32/GGM13-like MADS-box genes. Fig. S7: phylogenies of AG-like MADS-box genes. Fig. S8: phylogenies of AGL12-like MADS-box genes. Figs S9 and S10: RAxML phylogenies of AGL15- and AGL17-like genes. Fig. S11: phylogenies of GpMADS4-like MADS-box genes. Fig. S12: phylogenies of StMADS11-like MADS-box genes. Fig. S13: phylogenies of TM3-like MADS-box genes. Fig. S14: phylogenies of TM8-like MADS-box genes.