Strigolactones and the control of plant development: lessons from shoot branching

Authors


Summary

Strigolactones (SLs) were originally identified through their activities as root exudates in the rhizosphere; however, it is now clear that they have many endogenous signalling roles in plants. In this review we discuss recent progress in understanding SL action in planta, particularly in the context of the regulation of shoot branching, one of the best-characterized endogenous roles for SLs. Rapid progress has been made in understanding SL biosynthesis, but many questions remain unanswered. There are hints of as yet unidentified sources of SL, as well as unknown SL–like molecules with important signalling functions. SL signalling is even more enigmatic. Although a likely receptor has been identified, along with some candidate immediate downstream targets, our understanding of how these targets mediate SL signalling is limited. There is still considerable uncertainty about whether the targets of SL signalling are primarily transcriptional or not. There is at least one non-transcriptional target, because a rapid primary response to SL is the removal of PIN1 auxin exporter proteins from the plasma membrane in vascular-associated cells of the stem. We discuss how the various early events in SL signalling could result in the observed changes in shoot branching.

Introduction

Strigolactones (SLs) are carotenoid-derived plant metabolites that function as signalling molecules, both endogenously as phytohormones and exogenously in the rhizosphere. SL was originally identified as a germination stimulant for Striga lutea (witchweed) (Cook et al., 1966). Striga lutea, like other members of the Striga genus and the related Orobanchaceae family, are parasitic weeds that use SLs exuded from the roots of host plants as a germination cue, thereby ensuring germination in close proximity to a suitable host. Since this discovery, SLs have been shown to function as rhizosphere signalling molecules in ancient and beneficial adaptive processes, making them attractive targets for the later evolution of the parasitic plant germination strategy. Most notably, exudation of SL from roots is important for the recruitment of arbuscular mycorrhizal (AM) fungi (Akiyama et al., 2005). Approximately 80% of land plants participate in symbioses with AM fungi (Parniske, 2008), a relationship in which the plant supplies photoassimilates to the fungus in exchange for nutrients, particularly phosphate. Consistent with this role, SL synthesis and secretion into the soil is upregulated in response to phosphate deficiency (Yoneyama et al., 2007a,b; Lopez-Raez et al., 2008; Umehara et al., 2008; Kohlen et al., 2011).

As endogenous signalling molecules, SLs have been implicated in a growing list of processes, including root growth, lateral root formation, root hair elongation, adventitious rooting, stem elongation, secondary growth, leaf expansion, leaf senescence, drought and salinity responses and, most prominently, shoot branching (Woo et al., 2001; Stirnberg et al., 2002; Snowden et al., 2005; Gomez-Roldan et al., 2008; Umehara et al., 2008; Agusti et al., 2011; Kapulnik et al., 2011; Ruyter-Spira et al., 2011; Rasmussen et al., 2012; de Saint Germain et al., 2013b; Ha et al., 2013). Analyses of the role of SL in shoot-branching control have been instrumental in characterizing the SL pathway, and genetic screens for shoot-branching mutants in several higher plant species led to the isolation of key genes required for SL synthesis and signalling. Several recent reviews have discussed SL signalling and its many roles in plant development (Janssen and Snowden, 2012; Brewer et al., 2013; de Saint Germain et al., 2013a; Foo and Reid, 2013; Ruyter-Spira et al., 2013). Here, we discuss recent advances in our understanding of SL biology with a specific emphasis on how it relates to our current understanding of shoot branching.

SL Biosynthesis and Transport

Biosynthesis

The naturally occurring SLs identified to date share a common tricyclic lactone structure composed of three rings (referred to as ‘ABC’), connected via an enol-ether bridge to a D–ring butenolide group (Figure 1). The D–ring and enol-ether bridge are an invariant feature of the known natural, active SLs (Mangnus et al., 1992; Zwanenburg et al., 2009, 2013). Several steps required for SL biosynthesis have now been assigned to enzymes identified in genetic screens for increased branching mutants in various species: more axillary growth (max) in Arabidopsis thaliana (Figure 2); dwarf (d) and high tillering dwarf (htd) in Oryza sativa (rice); ramosus (rms) in Pisum sativum (pea); and decreased apical dominance (dad) in Petunia hybrida (petunia) (Woo et al., 2001; Stirnberg et al., 2002; Sorefan et al., 2003; Booker et al., 2004, 2005; Ishikawa et al., 2005; Snowden et al., 2005; Johnson et al., 2006; Zou et al., 2006; Arite et al., 2007; Simons et al., 2007; Arite et al., 2009; Drummond et al., 2009; Lin et al., 2009; Drummond et al., 2012; Waters et al., 2012a,b). These mutants are characterized by their low levels of endogenous SLs, and the ability to rescue their shoot phenotypes by SL addition or, where tested, grafting to wild-type roots. Comparative studies have extended this list to other species, including Solanum lycopersicum (tomato), Zea mays (maize), Dendranthema grandiflorum (chrysanthemum) and Salix spp. (willow) (Vogel et al., 2010; Guan et al., 2012; Kohlen et al., 2012; Dong et al., 2013; Ward et al., 2013). Orthologues of SL biosynthetic genes and detectable SLs have been reported in more basal plants, including mosses, liverworts and green algae in the Charales (Proust et al., 2011; Delaux et al., 2012).

Figure 1.

The proposed pathway of strigolactone synthesis from the precursor all-trans-β-carotene to the simple strigolactone 5–DS. The molecular re-arrangements catalysed by D27, CCD7 and CCD8 are highlighted in blue, violet and green, respectively. The four rings of 5–DS are designated A–D. The synthetic strigolactone GR24 is shown to the left of 5–DS.

Figure 2.

Comparison of the shoot morphology of plants homozygous for wild type (left) and mutant (right) alleles of MAX1.

The first steps of SL synthesis involve isomerization and dioxygenase-mediated cleavage of a carotenoid precursor (or possibly precursors) via the action of the isomerase, D27, and CAROTENOID CLEAVAGE DIOXYGENASE 7 and CAROTENOID CLEAVAGE DIOXYGENASE 8 (CCD7 and CCD8), all of which are localized in the plastid (Sorefan et al., 2003; Booker et al., 2004; Auldridge et al., 2006; Arite et al., 2007; Lin et al., 2009; Waters et al., 2012a). SL was originally believed to be derived from all-trans-β-carotene by CCD7 and CCD8 (Schwartz et al., 2004; Alder et al., 2008), but recent work has demonstrated that an initial step in SL synthesis involves the isomerization of all-trans-β-carotene to 9-cis-β-carotene (Alder et al., 2012). In rice and Arabidopsis, this reaction is catalysed by an iron-containing enzyme, D27, the first all-trans-β-carotene/9-cis-β-carotene isomerase identified in plants (Lin et al., 2009; Waters et al., 2012a; Alder et al., 2012). It is possible that there is an alternative source of 9–cis-β-carotene, for example from spontaneous isomerization, because the branching phenotype of d27 mutants is not as severe as mutations in downstream enzymes, such as ccd8 (Lin et al., 2009; Waters et al., 2012a).

Following isomerization, 9–cis-β-carotene is converted into 9–cis-β-apo-10′-carotenal via a CCD7-mediated cleavage reaction at the C9′–C10′ position. From here, CCD8 performs a striking reorganization of the 9–cis-β-apo-10′-carotenal substrate, adding three oxygens and rearranging the backbone to form the A–ring, the characteristic D–ring and the enol-ether bridge, thereby producing the intermediate carlactone (CL) (Alder et al., 2012). CL has SL–like activity in rice and Arabidopsis, where it suppresses shoot-branching and other SL–regulated phenotypes in ccd8 mutants (Alder et al., 2012; Scaffidi et al., 2013). CL has recently been isolated from both Arabidopsis and rice, and feeding experiments with radiolabelled CL have verified that it is indeed an endogenous SL intermediate (Seto et al., 2014).

Downstream of D27, CCD7 and CCD8, the cytochrome P450 MAX1 is required for the synthesis of active SL (Booker et al., 2005; Kohlen et al., 2011; Scaffidi et al., 2013). This sequence of enzyme activity was originally inferred from reciprocal grafting experiments, where max1 rootstocks grafted to d27, max3 or max4 scions have wild-type levels of branching, whereas reciprocal grafts are highly branched (Booker et al., 2005; Waters et al., 2012a). These data support the existence of a mobile SL intermediate that is downstream of the plastid-localized enzymes described above. Consistent with this idea, MAX1 is predicted to be cytoplasmic, and its expression pattern does not match that of MAX3 and MAX4. MAX3 and MAX4 expression is highest in the root tips, cortex, and hypocotyls, a trend that is also observed for CCD7 and CCD8 orthologues in other dicotyledonous species (Booker et al., 2004; Bainbridge et al., 2005; Foo et al., 2005; Snowden et al., 2005; Johnson et al., 2006; Drummond et al., 2009). In contrast, MAX1 is expressed in the cambial region and xylem-associated parenchyma (Booker et al., 2005). The vascular localization of MAX1 hints that it may function during loading or unloading of the mobile SL precursor from the xylem. CL is the prime candidate for the mobile precursor acting between MAX4 and MAX1, as it over-accumulates in max1 and is unable to rescue max1 shoot phenotypes (Scaffidi et al., 2013; Seto et al., 2014).

It has been proposed that only a few steps, forming the B- and C–rings, are required to turn CL into a simple SL (Alder et al., 2012). Given the absence of additional SL biosynthetic candidates to date, it is possible that MAX1 catalyses all of these reactions. Still, unknown enzymes may perform intermediate steps required for SL synthesis and for further processing to yield the diverse SL structures identified so far (for reviews, see Xie and Yoneyama, 2010; Seto et al., 2012; Ruyter-Spira et al., 2013). The highly branched max1 mutant has so far only been reported in Arabidopsis, although MAX1 homologues from a wide range of other species can rescue the max1 mutant (Drummond et al., 2012; Challis et al., 2013; Cardoso et al., 2014). The lack of branching mutants affecting MAX1 family genes in other species may result from functional redundancy (Wang and Li, 2011; Challis et al., 2013). For example, multiple rice MAX1 orthologues are present in distinct clades, which are conserved among grasses (Umehara et al., 2010; Challis et al., 2013). A natural variation study in rice has recently described a quantitative trait locus containing two MAX1 orthologues, which is present in a low-tillering, high SL–containing cultivar, and is absent from a high-tillering, low SL–containing cultivar (Cardoso et al., 2014). It is tempting to speculate that the observed diversification in MAX1 could reflect the production of different SLs with only partially overlapping functions.

SL diversity

There is good evidence for functional diversity among SLs. The simplest SL isolated so far is 5–deoxystrigol (5–DS; Figure 1), detectable in monocots, dicots and the liverwort genus Marchantia (Awad et al., 2006; Yoneyama et al., 2008; Delaux et al., 2012). It contains no additional oxygen-containing groups on the A- and B–rings, and is proposed by Alder et al. (2012) as the simple SL made from CL. The addition of hydroxyl groups at different positions on the AB-ring produces several well-known SLs: strigol, orobanchol and sorgomol. It is not known whether decoration generally occurs before, after or concurrently with the formation of the ABC-ring. In Sorghum there is evidence that hydroxylation can occur after ABC-ring formation, because exogenously supplied 5–DS and ent-2′-epi-5-DS are taken up, hydroxylated at the C9 position, and exuded as sorgomol and ent-2′-epi-sorgomol, respectively (Motonami et al., 2013). A better understanding of where, when and how these different SLs are synthesized will be important in understanding their distinct roles. The presence of hydroxyl groups is generally associated with enhanced germination activity compared with acetyl-containing SLs (Sato et al., 2005; Xie et al., 2008). In contrast, hydrophobic or acetyl-containing SLs have stronger bud inhibition activity compared with hydroxyl-containing SLs in pea (Boyer et al., 2012).

Alternative routes to SL

Analysis of the max/rms/d biosynthetic mutants confirms that these genes are involved in the production of all major SLs; however, these mutations tend to reduce rather than eliminate detectable SLs. The roots of max1 and max4/ccd8 mutants have between three- and fivefold less orobanchol than the roots of the wild type (Kohlen et al., 2011). Orobanchol and orobanchyl acetate are not detected in rms1/ccd8 mutants, although detectable levels of fabacyl acetate are present in root exudates (Foo and Davies, 2011). The rice d27, d17/ccd7 and d10/ccd8 mutants have nearly undetectable levels of 2′–epi-5-DS, but at least one allele of d10 has detectable levels of 2′–epi-orobanchol or its isomer (Umehara et al., 2008; Lin et al., 2009). Overall, it seems that most SL biosynthetic mutants tested are capable of producing some SL species. It is unlikely that this is simply caused by the leakiness of the mutations. For example, moss mutants carrying a full deletion of the CCD8 gene still produce detectable levels of the SL strigol (Proust et al., 2011). Rather, these observations suggest that there is an alternative minor pathway for SL biosynthesis (Figure 3). In Arabidopsis, overexpression of the MAX2 SL signalling component (see below) partially reduces branching in max1, max3 and max4 (Stirnberg et al., 2007). This suggests either ligand-independent signalling or a MAX3/MAX4/MAX1-independent source of SL, consistent with the detection of orobanchol in some of these mutants (Kohlen et al., 2011). In this context it is interesting to note that a proposed endogenous butenolide signal, mimicked by smoke-derived karrikin compounds, and closely related to SLs, appears to be synthesized in a MAX3/MAX4/MAX1-independent manner (Nelson et al., 2011; Waters et al., 2012b; Scaffidi et al., 2013). There is good evidence for SL–related, non–SL compounds with germination stimulation activity. For example, non–SL sesquiterpene lactone compounds with Orobanche germination stimulation activity were recently reported in Helianthus annuus (sunflower; Raupp and Spring, 2013).

Figure 3.

An overview of strigolactone (SL) signalling and synthesis. Synthesis is depicted in the cell on the left and transduction in the cell in the middle, with some SL–regulated processes in plants indicated on the right. The major pathway of SL synthesis involves three plastid-localized enzymes D27, CCD7 and CCD8, which are required for the synthesis of the SL intermediate carlactone (CL). MAX1 and possibly other enzymes act downstream of CL to produce SL (5–DS is shown here). Other biosynthetic pathways may contribute to the SL pool. Another SL–like signal is probably also present in plants that signals through KAI2 and MAX2, as demonstrated for the SL–related molecule karrikin found in smoke. D14 and MAX2 are likely to function in the nucleus in SL perception and signal transduction. There may also be MAX2-independent perception of SL, possibly via KAI2. Several proteins that interact with D14 and/or MAX2 have been identified (see text), and these presumably mediate events downstream. These include the clathrin-dependent removal of PIN1 from the basal plasma membrane of cells in the stem, with consequent changes to auxin distribution. This could lead to transcriptional responses, for example BRC1 upregulation, or these transcriptional responses may occur independently of effects on PIN1. Arrows with dotted lines are used for more speculative interactions than arrows with solid lines.

Ultimately, the relative activity of different SLs will need to be determined by their relative affinities for their receptors, coupled to their ability to trigger immediate downstream events. What is known from structure–activity relationship studies is that although the ABC-ring is non-essential, the presence of the butenolide D–ring with a leaving group at the C2′ position is essential for SL activity in branching inhibition assays (Boyer et al., 2012; Fukui et al., 2013).

SL transport

Some advances have been made in uncovering how SLs are transported in plants, but this process is still largely unknown. Early grafting experiments in pea, Arabidopsis and Petunia demonstrated that SLs can move from the root to the shoot, but not vice versa (Beveridge et al., 1996; Napoli, 1996; Beveridge et al., 1997; Morris et al., 2001; Turnbull et al., 2002; Booker et al., 2005; Simons et al., 2007). Recently, SLs have been detected in the xylem sap of tomato and Arabidopsis, which is consistent with their unidirectional shootward movement (Kohlen et al., 2011).

In Petunia, SL exudation into the soil is known to require the PLEIOTROPIC DRUG RESISTANCE 1 (PDR1) protein (Kretzschmar et al., 2012). PDR1 is the only protein identified to date with SL transport function, supported by mutant and overexpression lines that possess altered SL exudation. PDR1 is part of the ATP BINDING CASSETTE (ABC) family of transporters, which have demonstrated roles in the transport of other hormones, such as ABA and auxin (Petrášek and Friml, 2009; Kang et al., 2010; Kuromori et al., 2010). Consistent with a role in SL exudation into the soil, PDR1 is highly expressed in the root tip, where SL biosynthesis gene expression is highest, and in passage cells further up the root, where AM hyphal entry usually occurs (Sharda and Koide, 2008). Interestingly, there is no PDR1 orthologue in Arabidopsis, with the closest homologue being the ABA transporter ABCG40 (Kang et al., 2010). The lack of a conserved SL transport protein in Arabidopsis, in contrast to the synthesis and signalling components, may result from the absence of mycorrhization. Consistent with this idea, Arabidopsis root exudates contain relatively low levels of SL compared with other species (Kohlen et al., 2011).

A role for active SL transport in the shoot is less clear. In Petunia, PDR1 is expressed in the nodes and stems, but not in dormant buds. The pdr1 mutant has increased bud outgrowth but only at basal nodes. This is in contrast to the dad mutants, which are also highly branched at more aerial nodes (Napoli, 1996; Napoli and Ruehle, 1996; Snowden and Napoli, 2003). It is not clear whether the pdr1 shoot phenotypes are dependent on root or shoot expression of PDR1. There could be reduced transport of SLs from the root to the shoot in these lines, or reduced movement of SLs locally in the shoot. Thus, the role of active SL transport in the shoot and the implications for branching control remain unclear.

SL Perception

D14

Although there are still gaps in our knowledge of SL synthesis and transport, the backbone of the major SL biosynthesis pathway is now established. The same cannot be said about the signalling events downstream of SL, despite rapid progress in this area. Of particular importance is the recent observation that the synthetic SL, GR24, binds to the α/β–hydrolase protein D14/DAD2. Mutations in rice OsD14, Petunia DAD2 and their Arabidopsis orthologue AtD14 all confer the typical highly branched SL–deficient phenotype, but these phenotypes cannot be rescued by SL addition, and indeed the rice d14 mutant significantly overproduces SLs (Arite et al., 2009; Gao et al., 2009; Liu et al., 2009; Hamiaux et al., 2012; Waters et al., 2012b). These data suggest that D14 is involved in SL perception, possibly in a manner similar to the α/β–hydrolase-like protein GID1 involved in gibberellin (GA) perception (Ueguchi-Tanaka et al., 2005); however, GID1 has lost its catalytic triad and does not exhibit hydrolase activity, whereas D14 possesses the conserved triad and is predicted to function accordingly. Indeed, it was recently shown that GR24 bound to the D14/DAD2 α/β–hydrolase undergoes hydrolytic cleavage at the enol–ether bond in the conserved Ser-His-Asp catalytic pocket (Hamiaux et al., 2012; Kagiyama et al., 2013; Zhao et al., 2013). Co-crystallization of GR24 and OsD14 demonstrates the formation of two hydrolysis products: the ABC-ring and the D–ring derivative 2,4,4–trihydroxy-3-methyl-3-butenal covalently bound to the Ser97 residue (Zhao et al., 2013). More recently, a second study co-crystallized OsD14 with GR24 and resolved a hydroxy D–ring product: D–OH, which participates in a hydrogen bond with the Trp205 residue (Nakamura et al., 2013). The authors propose that the D–OH product can be generated by in-crystal activity of D14 on the 2,4,4–trihydroxy-3-methyl-3-butenal intermediate. As a result of the formation of these hydrolysis products in crystallization studies, it is not strictly known how D14 initially binds GR24. Co-crystallisation with a non-hydrolysable version of GR24 would help to address this.

In light of these data, it is possible that D14 functions purely as an enzyme to produce a cleavage product with branch inhibition activity. There are conflicting reports about the bioactivity of GR24 hydrolysis products, but the data available suggest that the D–ring itself is unlikely to be the bioactive compound. In Petunia, the products of DAD2-hydrolysed GR24 do not inhibit bud outgrowth in Arabidopsis (Hamiaux et al., 2012). One study has reported that tiller outgrowth in d27 is inhibited by D–OH, but it should be noted that the effect observed is small and only seen with very high doses (Nakamura et al., 2013). Other studies using lower concentrations of D–OH compounds report no relevant bioactivity (Boyer et al., 2012; Fukui et al., 2013).

Regardless of the potential activity of D–OH, several lines of evidence support a receptor function for D14, perhaps with hydrolytic activity as a key part of the perception process. In particular, the cleavage reaction is very slow: only one molecule of GR24 is hydrolysed every 3 min, making it unlikely to be effective in producing an active signal. Furthermore, it seems likely that the mechanism of signalling is similar to that of karrikins, the germination stimulants found in plant-derived smoke, which signal through the closely related protein KAI2/D14-LIKE. Karrikins bind to KAI2 and may be hydrolysed through the action of the catalytic triad, but there is no leaving group, and thus the karrikin molecule is predicted to regenerate (Scaffidi et al., 2012; Waters et al., 2012c; Bythell-Douglas et al., 2013). Karrikin and SL signalling have been discussed in detail in recent reviews (Flematti et al., 2013; Waters et al., 2013, 2014).

Thermal stability and trypsin digestion assays demonstrate that GR24 addition destabilizes the DAD2 and D14 proteins, indicating that a conformational change may occur upon binding (Hamiaux et al., 2012; Nakamura et al., 2013), although another study has reported no change in D14 protein stability upon GR24 addition in a protease-digestion assay (Kagiyama et al., 2013). Nakamura et al. (2013) suggest that D–OH remains associated with D14 after cleavage and that the complex could function in signalling, with the D–OH forming part of the recognition surface for interacting proteins. This is reminiscent of the role of auxin as the molecular glue between members of the AUXIN BINDING F–BOX (AFB) family, and members of the AUX/IAA family of transcriptional repressors (Tan et al., 2007). Such activity could explain the low reported binding efficiency of D14 for GR24 (Hamiaux et al., 2012; Kagiyama et al., 2013), as strong binding may require the interacting partner, as is the case with auxin. Clearly, the elucidation of the precise mechanism of SL signalling through D14 and the role of hydrolysis will require the identification of the direct downstream effector(s). There have been multiple recent reports of proteins that function as downstream targets for SL signalling (see below). Although it is not yet certain how these target proteins function to control shoot branching, there is mounting evidence that D14 either promotes or participates an SL–sensitive receptor-degradation complex with MAX2 (Jiang et al., 2013; Wang et al., 2013; Zhou et al., 2013).

MAX2

Several ideas have been proposed about the downstream events activated by the binding and hydrolysis of SL by D14, and certainly the emerging story from the studies described above is that MAX2/D3/RMS4-mediated protein degradation and/or ubiquitination is integral to SL perception and signalling. Like D14, mutations in MAX2/D3/RMS4 result in a highly branched phenotype, which cannot be suppressed by exogenous SL addition or by grafting to wild-type roots (Beveridge et al., 1996; Stirnberg et al., 2002; Ishikawa et al., 2005; Johnson et al., 2006; Gomez-Roldan et al., 2008; Umehara et al., 2008; Drummond et al., 2012), and indeed clonal analysis in Arabidopsis suggests that MAX2 functions in a largely cell-autonomous manner (Stirnberg et al., 2007). Unlike d14, however, max2/d3/rms4 mutants have a range of additional phenotypes, such as an elongated hypocotyl and reduced germination, which have been attributed to its role in KAI2-dependent signalling (Stirnberg et al., 2002; Shen et al., 2007; Nelson et al., 2011; Waters et al., 2012b; Scaffidi et al., 2013). As mentioned above, KAI2 is thought to be the receptor for an SL–like compound and to have been co-opted as the karrikin receptor in fire-following species.

MAX2 encodes a member of the F–box protein family (Woo et al., 2001; Stirnberg et al., 2002). F–box proteins form the substrate selection subunit of SKP1-CULLIN-F-BOX PROTEIN (SCF)-type protein–ubiquitin ligase complexes. Protein ubiquitination often results in protein degradation by the 26S proteasome, but may also cause modification of protein activity or changes in protein localization (for a review, see Vierstra, 2012). MAX2 has been shown to participate in an SCF complex, and mutation of the F–box domain, required for its incorporation into the SCF complex, eliminates MAX2 function and instead imparts a weak dominant negative phenotype (Stirnberg et al., 2007). This could be caused by sequestration of the SCFMAX2 substrate, preventing its ubiquitination.

There are over 700 F–box proteins encoded by the Arabidopsis genome, suggesting that regulated ubiquitination is a common signalling mechanism in plants. Indeed, the signalling pathways of several other plant hormones involve SCF-mediated targeted protein degradation. For example, the aforementioned interaction of GA with its receptor GID1 triggers an interaction with members of the DELLA family of transcriptional regulators, which in turn leads to the interaction of this complex with SCFGID2/SLY1, resulting in the ubiquitination and degradation of the DELLAs and induction of GA-induced genes (Ueguchi-Tanaka et al., 2005). There has been considerable speculation that SL signalling occurs in a similar manner. There is now some evidence for this in both rice and Petunia, where it has been shown that GR24 promotes the interaction of D14/DAD2 with D3 and PhMAX2A (one of two MAX2 orthologues in Petunia) in two-hybrid and in vitro pull-down assays (Hamiaux et al., 2012; Jiang et al., 2013; Zhou et al., 2013). These two genes have overlapping expression patterns in rice, generally being highly expressed in xylem-associated cells in the leaf vasculature (Ishikawa et al., 2005). MAX2 is primarily nuclear-localized (Stirnberg et al., 2007; Dong et al., 2013), and D14-GFP reporter lines reveal both nuclear and cytoplasmic signals in Arabidopsis and rice protoplasts, making a D14–MAX2 interaction plausible in planta (Nakamura et al., 2013; Zhou et al., 2013); however, one study reports no such interaction in Arabidopsis between MAX2 and D14 (Wang et al., 2013).

Downstream Targets for Shoot-Branching Inhibition

A better understanding of the SL signalling pathway and the elucidation of the precise roles of D14 and MAX2 requires knowledge of the immediate and subsequent downstream targets. Most importantly, these SL targets must be causally linked to shoot-branching control and to other biological functions of SL. There are two ways to approach the question of what the targets of SL signalling might be. One is to start with the interactors of D14 and/or MAX2, and then work outwards to the phenotype. Alternatively, a detailed analysis of the phenotypes of SL–related mutants such as shoot branching can be used as a starting point, with the goal of working backwards to link them biochemically with the known components in the SL signalling pathway. These different approaches each have their own merits and have so far been used with differing degrees of success.

D14/MAX2 targets

Candidate gene approaches

An obvious way to identify downstream targets of SL is to isolate proteins that physically interact with either D14 or MAX2, or both, preferably in an SL–regulated manner. Research to identify such proteins in an untargeted way is continuing, and candidate-based approaches are also popular. Recently, the interaction between D14 and the rice DELLA protein SLR1 was tested in a yeast two-hybrid study, and an SL–dependent interaction was reported (Nakamura et al., 2013). Based on the observed D14–SLR1 interaction, it was suggested that D14 hydrolyses GR24 to induce an interaction with SLR1, and it was further proposed that this complex might be targeted for degradation by SCFMAX2/D3. Although della mutants may have altered branching patterns (Bassel et al., 2008), a careful analysis of the relationship between SLs and DELLAs in pea suggest that they act independently in the shoot. For example, multiple-mutant analysis demonstrates that SL deficiency suppresses internode elongation in the absence of DELLA proteins (de Saint Germain et al., 2013b). Furthermore, when a reporter line for the Arabidopsis DELLA protein RGA was used to assess responses to SL and GA, only GA was able to induce RGA degradation (de Saint Germain et al., 2013b).

A second MAX2/D14 target candidate is BRASSINOSTEROID INSENSITIVEI 1 EMS SUPPRESSOR (BES1), and its related proteins (Wang et al., 2013). These proteins are best known for their role in brassinosteroid signalling, where they function as brassinosteroid-activated transcription factors (Yin et al., 2002; Vert and Chory, 2006). BES1 interacts with MAX2 in various assays, but this interaction is SL–independent. Nonetheless, the synthetic SL GR24 can reduce the stability of the BES1 protein in a MAX2-dependent manner. Although no interaction with D14 has been detected, mutations in D14 appear to enhance the stability of BES1. These results suggest that SL signalling involves MAX2/D14-dependent degradation of BES1 and related proteins. Consistent with this idea, gain-of-function, hyperstable mutations in BES1 are associated with increased SL–resistant branching, whereas loss-of-function mutations are associated with reduced branching and can suppress branching in max2 mutants. All these data support a role for BES1 in the SL–mediated regulation of shoot branching; however, the phenotypes conferred by both classes of bes1 mutations are highly pleiotropic, making it difficult to determine whether these mutations affect bud activation directly.

These results highlight a problem with the candidate approach. Shoot branching, and indeed many of the other SL–related phenotypes, is regulated by multiple factors; hence the demonstration of a clear, causal chain of events linking SL to changes in shoot branching via a candidate intermediate is challenging. For example, as described for BES1 and DELLAs above, the proposal that the degradation of a specific candidate protein underlies the shoot-branching effects of SL is supported by the demonstration of an SL–dependent interaction with MAX2 and/or D14, a reduced branching phenotype in the loss-of-function mutant that is epistatic to the highly branched phenotype of SL mutants, and an SL–resistant highly branched phenotype of an overexpressor; however, any one of these observations can have several alternative explanations. A protein interaction observed in a yeast two-hybrid assay, an in vitro pull-down assay or even in planta does not necessarily infer functional significance. The mutation of a gene required for bud activation could result in a reduced branching phenotype epistatic to SL deficiency, even if this gene acts in an entirely parallel pathway, and overexpression of such a gene could similarly result in SL–resistant bud activation. Although these experimental approaches are informative, their unequivocal interpretation is risky. To overcome these problems, it is essential to understand better how SLs regulate bud activity, such that the gap between the immediate targets of D14/MAX2 and shoot branching can be narrowed sufficiently to allow for a better assessment of the specific functions of candidate interactors.

Second-site suppressors of d14/max2

A less targeted, but nonetheless potentially powerful, approach to identifying downstream events is to identify second-site suppressors of SL mutant phenotypes. The rationale here is that SL mutant phenotypes are probably caused by the over-accumulation of a protein or proteins normally targeted for degradation by SL via SCFMAX2. Loss-of-function mutations in such protein(s) would be expected to suppress SL mutant phenotype(s). Several such suppressors have been described for SL shoot-branching phenotypes (Rameau et al., 2002; Stirnberg et al., 2012a,b); however, in these cases the suppression is not direct and involves defects in early bud development or alterations in auxin sensitivity, again illustrating that many factors can affect shoot branching independently of SL.

A more promising suppressor of the max2 seedling phenotypes has recently been described, named SUPPRESSOR OF MAX2 1 (SMAX1; Stanga et al., 2013). The smax1 mutation rescues the elongated hypocotyl, small cotyledon size and reduced seed germination of max2. Although smax1 does not rescue the max2 shoot-branching phenotype, it may point to some other potential candidates worthy of investigation. SMAX1 encodes a chaperonin protein that is part of a small, eight-member clade of SMAX1-LIKE (SMAXL) genes in Arabidopsis (Stanga et al., 2013). These are most closely (albeit weakly) related to HEAT SHOCK PROTEIN 101 in the class–I Clp ATPase family. One of these, SMXL7, is relatively highly expressed in the shoot, consistent with a possible role downstream of MAX2 in SL–mediated shoot phenotypes.

Dominant max/d-like branching mutants

The complementary approach to second-site suppressor screens is to identify dominant max/d-like mutants. A major recent breakthrough in the field has come from one such mutant, d53, in rice (Jiang et al., 2013; Zhou et al., 2013). A dominant mutation in D53 results in the typical dwarf, high-tillering phenotype of SL–related mutants, and excitingly D53 was shown to encode a rice SMAXL family member. D14 and D53 interact, and this interaction is enhanced by SL in a concentration-dependent manner. D3 and D53 can also interact with or without D14 or SL. These interactions are observed in the nucleus, and the data suggest that D53 is polyubiquitinated and destabilized in the presence of D14, D3 and SL. Importantly, the dominant mutant d53 protein can still interact with D14 in an SL–dependent manner, but it is not destabilized (Jiang et al., 2013).

The close resemblance of d53 mutants to the previously isolated SL mutants, coupled with the biochemical evidence for SL–triggered degradation of D53, suggest this protein family is very likely to be central to SL–mediated branch regulation; however, as described above, until it is clear exactly how SL regulates shoot branching, the significance of D53 stability for this process will remain unresolved.

The mechanism(s) of SL action in the regulation of branching

The detailed characterization of SL branching mutants over many years has resulted in two, non-exclusive proposed mechanisms for SL action. The first is that SL signals through the transcriptional upregulation of particular genes in the TEOSINTE BRANCHED 1, CYCLOIDEA and PROLIFERATING CELL FACTORS 1 and 2 (TCP) family in buds (Braun et al., 2012; Dun et al., 2012). The second is that SL systemically triggers the removal of PIN1 auxin efflux proteins from the plasma membrane (Bennett et al., 2006; Crawford et al., 2010; Shinohara et al., 2013). These two targets of SL signalling form the basis of the two theories for how SL regulates bud outgrowth. Both can be used to explain an important feature of the SL branching phenotype, namely the resistance of SL mutant buds to inhibition by apical auxin (Beveridge et al., 2000; Sorefan et al., 2003; Bennett et al., 2006; Arite et al., 2007). Auxin is exported from the primary shoot apex, transported down the main stem in the polar auxin transport stream (PATS) and inhibits bud activity (Thimann and Skoog, 1933; Goldsmith, 1977). This auxin acts indirectly, however, as it does not enter the bud in appreciable quantities (Hall and Hillman, 1975; Morris, 1977). SLs are required in some way for this indirect auxin-mediated inhibition.

The transcriptional model

Related members of the TCP domain transcription factor family have been shown to play a major role in shoot branching in diverse species. Loss-of-function mutations in these genes result in increased shoot branching, exemplified by teosinte branched 1 (tb1) in maize, fine culm 1 (fc1) in rice and branched 1 (brc1) in Arabidopsis and pea (Doebley et al., 1997; Takeda et al., 2003; Aguilar-Martínez et al., 2007; Finlayson, 2007; Braun et al., 2012). These genes are expressed almost exclusively in buds and their enhanced expression is associated with bud inhibition (Doebley et al., 1997; Takeda et al., 2003; Lewis et al., 2008). Where tested, bud activation in the loss-of-function mutants is SL resistant (Brewer et al., 2009; Minakuchi et al., 2010; Braun et al., 2012). This has led to the proposal that these genes are a downstream target for the SL pathway in bud inhibition. This hypothesis has been investigated in several species and has particularly strong support in pea.

In pea, the expression of BRC1 is closely and negatively correlated with bud activity. SL mutant buds show reduced BRC1 expression, and the application of SLs to buds causes upregulation within 6 h in a cycloheximde-independent manner (Dun et al., 2012). This suggests that SLs can alter BRC1 transcription without the need for new protein synthesis, a hallmark of a primary response gene.

This mode of action of SL can explain the auxin resistance of SL mutant buds, with the additional observation that the transcription of SL biosynthetic genes is upregulated by auxin (Foo et al., 2005; Johnson et al., 2006; Zou et al., 2006; Arite et al., 2007; Hayward et al., 2009; Liang et al., 2010; Zhang et al., 2010). The BRC1 model is therefore essentially a linear pathway in which auxin positively regulates SL synthesis in the main stem, and SL moves up into the bud where it upregulates BRC1 expression, thereby suppressing branching (Brewer et al., 2009; Braun et al., 2012; Dun et al., 2012, 2013). SL activity downstream of auxin is consistent with the observed auxin resistance of rms/max mutant buds. It is also consistent with the observed SL suppression of bud activity in auxin-signalling mutants, including axr1 and tir1 afb1 afb2 afb3 in Arabidopsis (Brewer et al., 2009).

Although this model neatly fits the transcriptional and bud outgrowth responses in pea, there is still the question of whether it is the primary mode of SL action. BRC1 is expressed specifically and quite widely in buds, and this only partially overlaps with the much wider expression domains of MAX2 and D14, which are also expressed and presumably function systemically in vascular-associated tissues (Aguilar-Martínez et al., 2007; Stirnberg et al., 2007; Arite et al., 2009; Gao et al., 2009). Furthermore, in several species the correlation between the expression of BRC1 family members and bud activity is weak. For example, FC1 expression in rice d mutants is not reduced, overexpression of FC1 in d3 can only partially rescue the d3 branching phenotype and the addition of SL does not induce FC1 expression (Arite et al., 2007; Minakuchi et al., 2010). Similarly, in maize, where TB1 expression is constitutively high as a result of changes to the promoter selected during domestication (Doebley et al., 1997), SLs do not affect TB1 transcript levels. Mutations in ZmCCD8 result in increased branching, and GR24 treatment is not associated with altered TB1 expression but can reduce branch length (Guan et al., 2012). The branching phenotype of maize ccd8 mutants has been described as being relatively weak, but it is similar to the increased branching observed in Arabidopsis max4/ccd8 mutants. The phenotype only appears weak in comparison with the tb1 loss-of-function mutant, which is much more dramatic because of the dual role of maize TB1 in regulating both branch activity and branch identity. Because of this, tb1 mutants have long branches tipped by tassels, instead of short ear shoots (Hubbard et al., 2002). Thus, in both maize and rice, SL and TB1/FC1 are at least partially independent in their effects. It is possible that these differing responses reflect differing modes of regulation amongst species; however, even in pea the increased bud outgrowth observed at most nodes in the rms1/ccd8 mutant differs to the pronounced outgrowth that is restricted to basal nodes in the Psbrc1 mutant, suggesting SL–mediated branching inhibition, independent of BRC1 (Braun et al., 2012).

The auxin transport canalization model

An alternative explanation for the SL inhibition of bud outgrowth is based on the concept of auxin transport canalization (Bennett et al., 2006; Prusinkiewicz et al., 2009; Crawford et al., 2010; Shinohara et al., 2013). The auxin transport canalization hypothesis proposes that an initial passive flux of auxin between a source and a sink positively regulates and polarizes its own transport in the direction of the initial flux, and in doing so organizes its transport into files of cells, or canals, that efficiently move auxin from the source to the sink (Sachs, 1981, 2000). Under a canalization-based model for bud regulation, it is proposed that dormant buds must establish an efficient flow of auxin into the main stem PATS in order to activate (Li and Bangerth, 1999). There is good correlative evidence to support this (Morris, 1977; Balla et al., 2011). For example, in pea, the polarization of PIN1 auxin efflux carriers is observed between the bud and the main stem PATS following bud activation by decapitation of the apex, and this polarization is not observed when buds are inhibited by applying auxin to the decapitation site (Balla et al., 2011). One explanation as to why the establishment of a PATS out of the bud is necessary for bud activation is that auxin export away from incipient leaves on the flanks of the shoot apical meristem may be required for continued phyllotactic patterning and leaf initiation (Bayer et al., 2009).

According to the auxin transport canalization model for bud inhibition, the ability of a bud to canalize auxin flow into the main stem PATS would be competitively inhibited by other stronger sources of auxin, i.e. the primary apex and already active buds, supplying auxin into the main stem PATS and reducing its sink strength. In this way, auxin in the main stem PATS can inhibit bud activation indirectly by preventing auxin export from the bud.

Evidence for the role of SL in regulating buds through this mechanism comes from the analysis of their effects on the auxin export protein, PIN1. SL regulates the accumulation of PIN1 proteins on the plasma membrane, specifically by triggering their rapid depletion in a clathrin-dependent process that is likely to be endocytosis (Shinohara et al., 2013). The effect of SL on PIN1 depletion is independent of new protein synthesis but dependent on MAX2, and can be detected within 10 min of SL treatment in some cases. Consistent with this, the max mutants have increased stem auxin transport and stem PIN1 accumulation, accompanied by increased auxin levels in the PATS (Bennett et al., 2006). The vascular expression pattern of PIN1 (Gälweiler et al., 1998) matches well with that of D14 and MAX2 (Aguilar-Martínez et al., 2007; Stirnberg et al., 2007; Arite et al., 2009), but although MAX2 and D14 are nuclear, PIN1 depletion occurs at the plasma membrane. One hypothetical mechanism for SL action is that a protein that promotes PIN1 endocytosis is sequestered in the nucleus by a protein that is targeted for degradation in an SL-, D14- and MAX2-dependent manner. In line with this, there is some evidence for the release or exclusion of COP1 from the nucleus in response to SL treatment of seedlings (Tsuchiya et al., 2010).

An increased rate of removal of PIN1 proteins, conditioned by high SL, would dampen the positive feedback between auxin flux and PIN1 protein accumulation and polarization, thereby inhibiting the canalization of auxin transport out of buds. In this way, SL acts to enhance competition between buds and/or the primary apex for access into the main stem PATS (Ongaro et al., 2008; Crawford et al., 2010; Liang et al., 2010; Ward et al., 2013). According to this idea, SLs do not directly inhibit buds. They act systemically to dampen auxin transport canalization, thereby reducing the total number of buds that can be active. Consistent with this idea, solitary Arabidopsis buds on excised nodal stem segments can be inhibited by apical auxin supply, but not by basal SL supply; however, in excised stem segments bearing two buds, SL typically inhibits one bud but not the other. This result cannot be easily explained by a mode of action where SL locally regulates gene expression in buds, inhibiting their activity.

Further evidence that the effects of SL on auxin transport are causal in bud regulation includes the observation that the branching and auxin-resistant bud outgrowth phenotypes of the max mutants can be rescued by treatment with low levels of the auxin transport inhibitor 1–naphthylphthalamic acid (NPA) (Bennett et al., 2006). Similar effects have been reported in rice with d27 (Lin et al., 2009). This result is particularly striking because treatment of wild-type plants with the same levels of NPA results in increased branching, presumably because lower auxin in the main stem allows more buds to activate. Under this model, there are therefore two distinct ways to achieve a highly branched phenotype: the first with low PIN1 removal from the plasma membrane resulting in high PIN1 accumulation, high auxin transport and high auxin export from bud to stem; or the second with low PIN1 accumulation or activity, low auxin transport and lower auxin export from bud to stem. This leads to the prediction that the addition of SL to plants already compromised in their auxin transport ability would actually promote branching, a prediction that has been validated (Shinohara et al., 2013). This ability of SL to promote branching in some circumstances is also not easy to explain by the local regulation of gene expression in buds.

The strongest evidence against the idea that the primary mode of action of SL is enhancement of competition between buds via PIN1 depletion from the plasma membrane is the observation that all pea buds on a plant can be inhibited following decapitation and the simultaneous application of SL to buds at every node (Dun et al., 2013). Even so, some competition is observed, as the bud closest to the decapitation site still exhibits a small amount of outgrowth compared with intact plants, and this top bud remains fully inhibited if more basal branches are allowed to grow. Furthermore, under the canalization-based hypothesis, the application of sufficient quantities of SL directly to buds might be able to prevent canalization, even in the absence of a competing auxin source and hence strong stem sink strength. The ability to canalize depends on the relative rates of PIN1 insertion and PIN1 removal, and therefore particularly high levels of PIN1 removal may be sufficient to prevent canalization.

Further evidence against the central importance of SL effects on auxin transport in bud regulation comes from detailed comparisons of the application of NPA and GR24 on bud growth and auxin transport in recently activated buds. Here, GR24 and NPA were shown to have similar inhibitory effects on bud elongation, but NPA was much more effective than GR24 at inhibiting auxin transport, assessed by measuring the basipetal transport of co-applied radiolabelled auxin away from the GR24/NPA application site (Brewer et al., 2009). In interpreting these experiments it is important to remember that although both NPA and GR24 can reduce auxin transport, they act by completely different mechanisms and with different dynamics. SL is not strictly speaking an auxin transport inhibitor at all, but acts to reduce the level of PIN1 at the plasma membrane. In contrast, NPA is a strong pharmacological inhibitor of transport of both the PIN and ABCB family of auxin exporters (Yang and Murphy, 2009). In Arabidopsis stems treated with modest levels of NPA, only 10% of the auxin transport remains, whereas stems treated with extremely high concentrations of GR24 retain 70% of their auxin transport, compared with untreated controls (Crawford et al., 2010). Thus, co-application of NPA and radiolabelled auxin would prevent a large proportion of the auxin from entering the transport stream at the site of application, whereas GR24 would be predicted to have no such effect and would instead alter PIN accumulation dynamics at the plasma membrane. The inhibitory effect of these treatments on bud stem elongation could be caused by similarly different mechanisms: over-accumulation of auxin in the bud caused by auxin transport blockage with NPA treatment, or gradual depolarization of PINs in the bud with GR24 treatment.

It should be noted that the two models for SL action in bud inhibition are not mutually exclusive (Figure 3). Both could happen in parallel, or changes in BRC1 expression could result from SL–mediated changes in auxin transport. In this scenario, BRC1 could act as a stabilizer of bud repression, in which buds with low levels of auxin export, for example because of high SL levels, are kept in an off state by upregulation of BRC1. Interestingly, the majority of genes reported to be repressed following the treatment of seedlings with GR24 for 90 min are known or putative auxin-regulated genes (Mashiguchi et al., 2009), consistent with the idea that altered auxin transport is central to the subsequent transcriptional effects of SL.

To add to the complexity of SL–mediated branching inhibition, it was recently reported that OsTB1/FC1 may be involved in regulating the expression of the D14 gene (Guo et al., 2013). The transcription factor OsMADS57 is a transcriptional repressor of D14, but OsTB1/FC1 can interact with OsMADS57 to permit D14 expression. Additionally, it was shown that miRNAs contribute to this regulation, whereby OsMIR444a targets OsMADS57. These findings place BRC1 upstream of D14, or perhaps into a positive feedback loop, further locking buds into an inactive state. This may be important in the context of the canalization-based model for bud activation, which has at its heart a positive feedback loop, making it inherently unstable under stochastic fluctuations.

Shoot-branching control and SL–regulated degradation targets

As described above, there are at least three different protein families that are proposed to be targets for SL–regulated protein degradation: DELLAs, BES1 and D53/SMAX. In each case, a strong argument can be made that degradation of these proteins would mediate transcriptional responses. Proteins in the DELLA and BES1 families are well known transcriptional regulators, and D53 contains ethylene-responsive element binding factor-associated amphiphilic repression (EAR) motifs, known to interact with TOPLESS (TPL) proteins in transcriptional co-repression (Szemenyei et al., 2008; Pauwels et al., 2010). Consistent with this idea, a TPL-related protein in rice, TPR2, can interact with D53 (Jiang et al., 2013). These results have led to the hypothesis that D53 sequesters TPL proteins, thereby altering transcription. The transcriptional regulatory function associated with these three target proteins suggest gene expression changes are an important and direct effect of SL signalling; however, compared with the effects that modulation of BES1, DELLAs or TPL levels have on the transcriptome, SL has a relatively modest effect (Yin et al., 2002; Cao et al., 2006; Mashiguchi et al., 2009; Kagale and Rozwadowski, 2011). It is possible that this results from the relatively limited expression domain of SL signalling components, leading to more limited transcriptional changes than those observed when BES1, DELLA or TPL levels are changed more widely by other means.

Although there is a clear precedence for these protein families in regulating transcription, they could also be involved in the nuclear sequestration of proteins that could influence PIN1 endocytosis when released into the cytoplasm. The putative chaperonin function of the D53/SMAX1 family members is consistent with this idea. Furthermore, the EAR domain specifically interacts with the TPL Lissencephaly type–1-like homology motif (LisH)/C–terminal to LisH motif (CTLH) (Kobayashi et al., 2007; Szemenyei et al., 2008), which is shared by proteins in the CTLH complex. This is a widely conserved complex with roles in vesicle trafficking (Kobayashi et al., 2007; Tomaštíková et al., 2012), components of which could be sequestered in the nucleus by D53. Similarly, both DELLA and BES1 proteins can participate in diverse protein complexes (Locascio et al., 2013; Shigeta et al., 2013). Therefore, the discovery of these target proteins is consistent with either or both transcriptional and non-transcriptional downstream effects.

The Evolution of SL Signalling

Throughout this review, we have emphasized the role of SL in shoot-branching control, but it is clear that useful insights have come from examining the other processes in which SL is involved, such as germination and seedling establishment. Extending these comparisons to include an evolutionary perspective is also likely to be informative. An understanding of the ancestral role(s) of SLs will no doubt provide invaluable information about their mechanisms of action in extant angiosperms: for example, whether they function primarily in the regulation of transcription, the regulation of protein trafficking or both.

One possibility is that the ancestral role of SL was in the recruitment of AM fungi, allowing for more efficient nutrient uptake and thus facilitating the colonization of land. AM symbiosis is thought to have been essential for this transition, and there is evidence that it was present in the first land plants 360–450 million years ago (Simon et al., 1993); however, SL has been found in charophyte algae, which are the sister group to the land plants and are typically aquatic (Delaux et al., 2012). SL treatment of charophyte algae results in increased rhizoid elongation, a response that is shared with mosses and liverworts (Delaux et al., 2012). Interestingly, this response is apparently independent of MAX2, because homologues have not been found outside land plants (Delaux et al., 2012; Challis et al., 2013). There are hints of MAX2-independent responses in higher plants too, where quite modest levels of GR24 can inhibit root growth even in a max2 mutant background (Ruyter-Spira et al., 2011; Shinohara et al., 2013). A D14 family member is present in charophytes, but it is distinct from D14 and more closely related to KAI2 (Waters et al., 2012b; Delaux et al., 2012; Challis et al., 2013) (Figure 4). An attractive hypothesis is therefore that the interaction of SLs with this protein can trigger downstream responses independent of MAX2 (Figure 3).

Figure 4.

Speculative schematic of the evolution of the strigolactone (SL) signalling pathway. An ancestral pathway involving a D14L protein and possibly a D53-related protein (here called D53R) is present in algae, where exogenous SL application triggers rhizoid elongation. With the emergence of land plants the duplication of clades D53R and D14L, along with the rapid evolution of MAX2, provided the signalling components required for endogenous SL functions, such as shoot-branching control.

Interestingly, the MAX2 and D14 clades appeared rapidly upon the emergence of land plants (Figure 4). For D14, this presumably resulted from duplication within the clade. A second such duplication event within the D14 lineage is likely to have occurred after the divergence of the Lycophytes and Euphyllophytes, resulting in a third class of α/β–hydrolases, the D14-LIKE2 (DLK2) group (Waters et al., 2012b). These duplications are consistent with functional diversification within the pathway, concurrent with the diversification of land plants. Initial analysis of D53 phylogeny suggests a similar pattern, with moss D53-like genes being more closely related to SMAX1 than to the D53/SMAXL7 clade, with further subsequent duplications in both the SMAX1 and D53/SMAXL7 clades (Zhou et al., 2013) (Figure 4). The appearance of MAX2 is particularly remarkable because of its speed, requiring a rate of two substitutions per site since diverging from its nearest sister gene (Challis et al., 2013); however, since this rapid divergence there has been tight conservation within the MAX2 clade. This is particularly evident in the C–terminal domain, which is predicted to be involved in substrate recognition, suggesting an equally conserved MAX2 interaction domain in the targets/target.

The timing of MAX2 recruitment into the SL pathway is unknown; however, it has been hypothesized that the ancestral role of MAX2 was in mycorrhization, with its role in SL signalling coming later (Challis et al., 2013). Evidence for this comes from the d3 mutant in rice, which is defective in AM colonization. This phenotype has not been observed in any of the biosynthetic SL mutants or in d14 (Yoshida et al., 2012). It will be interesting to compare the phenotypes of mutants in MAX2 and D14 family members in basal land plants with each other, and with those of SL biosynthesis mutants (Proust et al., 2011). Analysis of potential downstream targets such as PIN protein accumulation and SMAX homologues in these species could also provide insights into SL signalling.

Concluding Remarks

The SLs have demonstrated functions in many developmental processes, and will almost certainly be assigned more roles in the future. Examination of the shoot-branching phenotype of SL–impaired mutants in diverse species has proven to be particularly valuable in providing a basic pathway for SL synthesis, as well as some strong candidates for downstream mediators of SL signalling. Many questions remain about the exact mechanism of SL action and perception, however. The array of SL responses and associated mutant phenotypes should provide powerful tools to address these knowledge gaps, and indeed have already yielded some promising candidates. Comparative analyses of the signalling pathways of SL and its structurally related molecules also provide important avenues for unravelling some of the complexities of strigolactone biology.

Acknowledgements

The authors' research is supported by the European Research Council (grant no. 294514 – EnCoDe) and the Gatsby Foundation (GAT3272C). We thank Tom Bennett for useful discussions on D53 evolution.

Ancillary