In A. thaliana the PPI network of floral homeotic proteins shows a characteristic scale-free structure where most proteins display a highly limited set of interaction partners and only few members possess promiscuous interactions, most notably SEP3 (de Folter et al., 2005; Immink et al., 2009; Al Hindi et al., 2017). PPI networks of floral homeotic proteins have also been investigated for several other core eudicot species such as Antirrhinum majus (Causier et al., 2003), Petunia x hybrida (Immink et al., 2003), Solanum lycopersicum (Leseberg et al., 2008; Al Hindi et al., 2017) and Gerbera hybrida (Ruokolainen et al., 2010) as well as for the more distantly related early diverging eudicot species Euptelea pleiospermum, Akebia trifoliata and
Pachysandra terminalis (Liu et al., 2010) and for the monocot Oryza sativa (Cooper et al., 2003).
Comparison of the PPI networks reveals that the protein-protein interactions that are required for the formation of the different floral quartets are highly conserved throughout all investigated
Discussion
99
networks pointing towards an essential role of these interactions for the development of the primary flower architecture (Liu et al., 2010). Beside the conserved interactions also several variable interactions were found that are presumed to account for deviant flower morphologies of the examined species (Liu et al., 2010). In addition and no less remarkable also the absence of numerous interactions appeared to be ‘conserved’. For example, in none of the investigated eudicot and monocot species a direct interaction of AP3-like proteins with AP1-, AG- or STK-like proteins was detected (Liu et al., 2010; Ruokolainen et al., 2010). Furthermore, also no direct interaction between AG- and STK-like proteins was found (Liu et al., 2010). The absence of certain interaction patterns suggests that these interactions probably bring about detrimental changes in gene regulation that may cause severe malfunctions during floral organ identity determination.
In our study on the evolution of the obligate heterodimerization of AP3 and PI my colleagues and I investigated protein-protein interaction capabilities of floral homeotic proteins from the early diverging angiosperm species Amborella trichopoda, Nuphar advena and from the magnoliid
Liriodendron tulipifera (Manuscript I: Melzer et al., 2014). We could show that even floral
homeotic proteins of such distantly related species possess all protein-protein interactions that are necessary for the formation of the different floral quartets (Manuscript I: Melzer et al., 2014). However, beside these highly conserved interaction patterns we observed additional interactions among floral homeotic proteins of early diverging angiosperms that are not found among their orthologs from monocots and eudicots (Manuscript I: Melzer et al., 2014). For example AP3-like proteins from A. trichopoda and N. advena do not only interact with PI-like proteins but also with AG-like and AGL6-like proteins. This finding is consistent with other studies on the interactions of floral homeotic proteins from A. trichopoda and N. pumila that detected direct interactions of AP3- and PI-like proteins with AP1-, AGL6- and STK-like proteins, respectively (Amborella Genome Project, 2013; Li et al., 2015). The more promiscuous interactions of floral homeotic proteins from early diverging angiosperms suggest that ancestral precursors of floral homeotic proteins being present at the base of angiosperm evolution may also possessed a wide interaction spectrum, pointing towards a shift from promiscuity to specificity in the PPI network.
To better comprehend changes within the PPI network of floral homeotic proteins during angiosperm evolution two different approaches could be applied. The method of ancestral
Discussion
100
character state reconstruction uses experimentally determined protein-protein interaction data of proteins from extant species, maps the interaction data onto the phylogeny of the examined proteins and eventually reconstructs the ancestral states at internal nodes (Manuscript II: Rümpler et al., 2015a). By contrast the technique of ancestral sequence reconstruction calculates the most likely amino acid sequence of ancestral proteins at certain time points during evolution. The reconstructed amino acid sequences can be used to synthesize the encoding nucleotide sequences and to subsequently examine the protein-protein interaction behavior of the corresponding proteins experimentally (Gumulya and Gillam, 2017). In recent years both approaches were applied to reconstruct the state of the PPI network of floral homeotic proteins in the MRCA of extant angiosperms (Li et al., 2015; Ruelens et al., 2017). The results consistently substantiate the assumption that ancestral floral homeotic proteins indeed possessed more promiscuous interaction capabilities compared to their orthologs of extant eudicots and monocots (Li et al., 2015; Ruelens et al., 2017).
In the context of this thesis protein-protein interaction data of floral homeotic proteins from phylogenetically informative angiosperms were examined to infer how the PPI network controlling flower development changed during early angiosperm evolution. However, to conceive the evolutionary trajectories that shaped the structure of the PPI network at the base of angiosperm evolution it is necessary to also consider the interaction capabilities of orthologs of floral homeotic proteins from gymnosperms, angiosperms closest extant relatives. The seven subfamilies of floral homeotic genes that are present in angiosperms (AP1-, AP3-, PI-, AG-,
STK-, SEP1- and SEP3-like genes) most likely trace back to four ancestral gene families
(ancestral AP1/FLC-, ancestral AP3/PI-, ancestral AG/STK- and ancestral SEP-like genes) that were present in the MRCA of extant seed plants i.e. angiosperms and gymnosperms (Gramzow and Theißen, 2010; Ruelens et al., 2013; Gramzow et al., 2014; Chen et al., 2017). In angiosperms ancestral AP1/FLC-like genes diverged into AP1- and FLOWERING LOCUS C (FLC)-like genes of which latter function as central repressors of flowering in A. thaliana. Ancestral AP3/PI-like genes diverged into AP3- and PI-like genes, a duplication of an ancestral
AG/STK-like gene gave rise to AG- and STK-like genes and ancestral SEP-like genes diverged
into SEP1- and SEP3-like genes (Ruelens et al., 2013; Gramzow et al., 2014; Chen et al., 2017). Phylogeny reconstructions of orthologs of floral homeotic genes from the gymnosperm Gnetum
Discussion
101
and an ancestral AG/STK-like gene, respectively (Becker et al., 2003; Wang et al., 2010; Gramzow et al., 2014). Two further genes of G. gnemon GGM9 and GGM11 were found to be phylogenetically closely related to AP1- and SEP-like genes of angiosperms, although their exact phylogenetic relationship is still discussed controversially in the literature (Kim et al., 2013; Ruelens et al., 2013; Gramzow et al., 2014). Studies on the protein-protein interactions of the encoded proteins revealed direct interaction of GGM2 with GGM3, GGM9 and GGM11 as well as direct interaction of GGM3 with GGM9 and GGM11 (Winter et al., 2002b; Wang et al., 2010). Thus similar to the direct interactions of AP3-, PI-, AG- and STK-like proteins observed in early diverging angiosperms (Amborella Genome Project, 2013; Melzer et al., 2014; Li et al., 2015) also the AP3/PI- and AG/STK-like proteins of the gymnosperm G. gnemon are capable to directly interact. Thus it appears most likely that also the ancestral AP1/FLC-, AP3/PI-, AG/STK- and SEP-like proteins being present at the base of seed plant evolution possessed promiscuous interaction capabilities. The angiosperm specific duplications of floral homeotic genes increased the number of floral homeotic proteins which probably still possessed promiscuous interaction capabilities resulting in a complex PPI network with highly connected nodes (Li et al., 2015; Ruelens et al., 2017). During early angiosperm evolution numerous protein-protein interactions were lost leading to a shift from promiscuous to more specific interactions within the PPI network of floral homeotic proteins (Ruelens et al., 2017). Remarkably the loss of interactions thereby was not random but rather predominantly involved interactions of AP3- and PI-like proteins and to a certain extent also AG- and STK-like proteins (Li et al., 2015; Ruelens et al., 2017). In contrast SEP-like proteins retained their promiscuous interaction behavior eventually leading towards the scale-free network structure found in extant eudicots and monocots with SEP-like proteins serving as hubs that mediate interaction of most other floral homeotic proteins (Ruelens et al., 2017).