The control of developmental phase transitions in plants

The life cycle of flowering plants can be considered as a succession of distinct growth phases (Fig. 1), and the transition between these phases is dependent on developmental genetic programs that are triggered and modulated by both environmental and endogenous stimuli. Following germination, and before they become competent to flower and reproduce, the shoots of most plants pass through a phase of vegetative growth. During this period, plants generally rapidly increase their photosynthetic capacity and their size and mass. This vegetative mode of growth can be further divided into a juvenile and an adult vegetative phase. Although generally less conspicuous in annuals such as Arabidopsis, these phases are often recognizable by a changed growth pattern and body form, particularly in perennial species. For example, in some cephalium-bearing cacti, the different phases resemble graftings of different plant species (Mauseth, 2006). It is normally only during the adult vegetative phase that plants are capable of forming reproductive organs, and day length-dependent plants can be induced to flower by photoperiodic induction during this phase. During the juvenile-to-adult phase transition (see Glossary, Box 1), plants thus acquire reproductive competence. Simultaneously, changes in multiple traits, such as leaf size and shape, internode length and trichome distribution, result in the appearance of both early (juvenile) and mature (adult) shoots on the same plant, a condition known as heteroblasty (see Glossary, Box 1) (Goebel, 1889; Poethig, 1990; Poethig, 2010).

Eventually, another phase transition, known as the reproductive phase change (see Glossary, Box 1), occurs. During this transition, plants switch from vegetative to reproductive growth, and the vegetative shoot apical meristem (SAM, see Glossary, Box 1) takes on an inflorescence meristem (IM) identity. The developmental fate of the IM, i.e. its conversion into a floral meristem (FM) or its production of lateral meristems that may convert to FMs, is species specific and determines the type of inflorescence formed. As in other species, this transition is very obvious in Arabidopsis, and there is thus a great deal of knowledge regarding the molecular control of reproductive phase change and the determination of meristem identity (Adrian et al., 2009; Amasino, 2010; Srikanth and Schmid, 2011). The correct timing of the transition to flowering is of utmost importance to all plants, as it may have a strong impact on fitness. For example, flowering should take place when the climatic conditions are suitable and sufficiently reliable to allow completion of the process with the successful dispersal of seeds. In the case of non-self-fertile species, flowering also needs to be synchronized within a population and with the occurrence of potential pollinators.

The transition to flowering is under the control of a complex genetic network that integrates information from various endogenous and environmental cues (Amasino, 2010). Genetic analyses, mostly conducted in Arabidopsis and rice, have identified numerous genes that participate in the regulation of flowering. Endogenous factors that regulate reproductive phase change include hormones and carbohydrate assimilates, such as gibberellins and sugars (Corbesier et al., 1998; Moon et al., 2003; Ohto et al., 2001; Seo et al., 2011). Among the environmental factors that may affect plant growth, only a few seem to be specifically monitored to control flowering (Amasino, 2010; Srikanth and Schmid, 2011). The most important of these are temperature, including the ambient growth temperature a plant experiences throughout its life, as well as prolonged periods of low temperature (which cause vernalization, see Glossary, Box 1), and day length (photoperiod). Eventually, the activities of the flower inductive pathway genes converge on a small number of so-called floral integrators. Once the expression of these integrators exceeds a threshold, the plant initiates flowering, which is generally an irreversible process (Tooke et al., 2005). Besides being important to ensure the reproductive success of a species, the control of the vegetative-to-reproductive phase transition is also of considerable agronomical importance. For example, premature flowering usually results in reduced biomass and seed set. Similarly, prolonged vegetative growth might lead to an increase in biomass but at the same time often results in reduced seed number and seed filling (Demura and Ye, 2010).

In recent years, it has become increasingly clear that the networks that control the juvenile-to-adult phase transition and the reproductive phase transition share some major regulatory factors. In addition, many of these factors also affect some of the heteroblastic features that distinguish these two phases. In particular, two evolutionary highly conserved microRNAs (miRNAs), miR156 and miR172, and their targets have been identified as key components of the genetic control mechanisms that underlie plant phase changes. The miRNA miR156 targets transcripts of a subset of SQUAMOSA PROMOTER BINDING PROTEIN-LIKE (SPL) transcription factors. These were originally identified in inflorescence nuclear protein extracts by means of their conserved DNA-binding SBP domain (Klein et al., 1996) and more recently have been shown to promote the transition from juvenile to adult and to flowering (Schwarz et al., 2008; Wu and Poethig, 2006). By contrast, miR172 targets mRNAs that encode proteins with two APETALA 2 (AP2) DNA-binding domains. These proteins have been shown to regulate both the transition to flowering and flower development (Aukerman and Sakai, 2003; Jung et al., 2007; Mathieu et al., 2009; Schmid et al., 2003). Furthermore, the MADS-domain transcription factor FLOWERING LOCUS C (FLC), a key player in the Arabidopsis vernalization response (Amasino, 2010; Srikanth and Schmid, 2011), has recently been shown to delay the juvenile-to-adult transition by directly acting on some of the same targets (Willmann and Poethig, 2011; Deng et al., 2011). These findings further demonstrate that tight connections exist between the juvenile-to-adult and the reproductive phase transitions.

In this review, we discuss recent findings, mainly from Arabidopsis, that have increased our understanding of the regulation of phase transitions in plants. In particular, we focus on the roles of miR156 and miR172, and their targets, in these processes. We further highlight a possible reinforcing feedback loop between these miRNAs and their targets, and discuss the potential role of this regulatory feedback loop in establishing a sharp boundary between different floral organs. Finally, we review the roles of these miRNAs in other plant species, such as maize and poplar, and discuss the evolutionary conservation of the function of these miRNAs.

As a plant passes through the various developmental phases of its life cycle and continues to develop new organs, a number of morphological traits (such as size and shape of leaves, phyllotaxy, plastochron, internode length, adventitious root production, trichome distribution and cell size) change in accordance with the developmental stage. As a result, different parts of a plant may exist in different developmental phases. This phenomenon of heteroblasty is observed in many species. It should be noted, however, that morphological plasticity, expressed as a consequence of environmental variations, often affects the expression of the very same heteroblastic features. It can, thus, be difficult to distinguish genetically determined ontogenetic changes from those that occur due to plasticity (Diggle, 2002). Like the acquisition of reproductive competence (Mozley and Thomas, 1995), most of the heteroblastic features change gradually. These features can be classified as either juvenile or adult, and are often used as markers to monitor progression of the juvenile-to-adult phase change. Although these changes are collectively known as the vegetative phase change (see Glossary, Box 1), it is important to note that it is not yet clear how the phenomenon of heteroblasty relates to the reproductive competence of the shoot. In addition, the use of the terms `juvenile' and `adult' in describing both the heteroblastic change as well as the state of reproductive competence, may lead to confusion, as discussed by Jones (Jones, 1999) and others [Zotz et al. (Zotz et al., 2011) and references to older literature on heteroblasty therein].

Despite the appearance of similar heteroblastic traits that often accompany particular developmental phase changes in different plant species, the necessity or consequences of these relationships, from a plant-fitness perspective, remain poorly understood. Elucidating the molecular genetic mechanisms that link these phenomena is, thus, of immediate importance if we are to achieve a better understanding of plant diversification and distribution.

Large differences exist among species with regards to the extent to which particular heteroblastic changes are expressed. They are usually most apparent in perennial woody plants, in which individual developmental phases can last up to several years. Moreover, branching, in combination with seasonality in vegetative and reproductive growth, allows perennial woody plants to bear shoots in different, i.e. juvenile or adult, developmental stages, making them optimal systems for phenotypic studies of vegetative phase change (Doorenbos, 1965; Hackett, 1985; Jaya et al., 2010). Unfortunately, such plants are usually not the best suited for molecular analyses as the genomic resources available for them are still limited, and experiments with these large and slow-growing species are often space demanding and time consuming. However, building on observations made in Arabidopsis, some remarkable progress in this area has recently been achieved in poplar (Hsu et al., 2011; Wang et al., 2011).

Fortunately, vegetative phase change, albeit less dramatic, can also be observed and studied in small annual plants, such as in the model plant Arabidopsis thaliana (Steynen et al., 2001; Telfer et al., 1997; Tsukaya et al., 2000; Martinez-Zapater et al., 1995; Röbbelen, 1957).

The vegetative phase of shoot development in Arabidopsis lasts between weeks in summer-annual varieties and up to several months in winter-annual accessions that require vernalization to induce flowering (Amasino, 2010). As part of the vegetative phase transition, the juvenile-to-adult phase transition in Arabidopsis is accompanied by relatively minor changes in leaf morphology (Chien and Sussex, 1996; Telfer et al., 1997; Telfer and Poethig, 1994; Usami et al., 2009). The early rosette leaves that are produced feature `juvenile traits', including long petioles, small almost round blades and smooth margins. In addition, juvenile leaves produce trichomes on only their upper (adaxial) side. Late rosette leaves, by contrast, display `adult traits', such as shorter petioles, enlarged and elongated blades with an increased number of size-reduced cells, as well as serrated margins and trichomes on both their upper and lower (abaxial) side. With the latter as a notable exception, these features change gradually, and the so-called transition leaves may display a mix of juvenile and adult features. As the absence or presence of abaxial trichomes is easy to score, and because their production is affected by photoperiodic conditions and by mutations that affect flowering time (Telfer et al., 1997), this feature is frequently used to quantify the duration of the juvenile phase (e.g. the duration of the juvenile phase can be expressed as the number of rosette leaves without abaxial trichomes). As such, a number of recent insights into the genetic control of phase changes have been gained from studying Arabidopsis.

As mentioned above, both intrinsic and environmental conditions can affect heteroblastic features. For example, lengthening or shortening the daily light period affects the number of nodes that bear `juvenile' leaves (Chien and Sussex, 1996; Telfer et al., 1997). The importance of different regulatory systems can thus be revealed in specific mutants. For example, Arabidopsis mutants that are insensitive to the phytohormone gibberellin (GA) or those defective in its biosynthesis show a delayed appearance of the first adult leaf (Chien and Sussex, 1996; Telfer et al., 1997). Interestingly, another class of mutants that display accelerated vegetative phase changes are those associated with mutations in genes involved in the biogenesis and action of miRNAs and of trans-acting small interfering RNAs (ta-siRNAs). Whereas the effects of ta-siRNAs on heteroblasty may be limited to their ability to affect leaf polarity via their targeting of auxin response factors (Allen and Howell, 2010; Fahlgren et al., 2006; Hunter et al., 2006; Pekker et al., 2005), the effects of disturbed miRNA biogenesis are more diverse. The pleiotropic phenotypes associated with perturbed miRNA function are perfectly understandable in the light of the plethora of currently known plant miRNAs (Rubio-Somoza and Weigel, 2011). However, when the first miRNAs were identified in plants, it was soon recognized that many of the predicted targets encode transcription factors, some of which were known to affect flowering (Rhoades et al., 2002). Further insights into the role of individual miRNAs and their transcription factor targets in developmental phase changes have since been gained from misexpression studies in Arabidopsis.

One of the most abundant miRNAs in Arabidopsis is miR156, which reaches it highest levels at the seedling stage (Axtell and Bartel, 2005; Fahlgren et al., 2007; Wang et al., 2009). When constitutively overexpressed, miR156 causes a moderate delay in flowering (Fig. 2) (Schwab et al., 2005). Moreover, such transgenics produce a larger number of leaves with characteristically juvenile features (Wu and Poethig, 2006). Elevated levels of miR156 thus seem to prolong juvenility and delay the onset of the adult phase. In Arabidopsis, miR156 and some miR156 isoforms can potentially be encoded by many loci, e.g. MIR156a-f, MIR156g-h and MIR157a-d (Kozomara and Griffiths-Jones, 2011; Xie et al., 2005). The sequestration of miR156 upon the overexpression of miR156 mimicry targets (MIM156) has been used to perturb endogenous miR156 function. Such MIM156 transgenics indeed flower after producing only a few leaves, which all display adult features (Fig. 2) (Franco-Zorrilla et al., 2007; Todesco et al., 2010). miR156 is thus not only sufficient but also necessary for the expression of the juvenile phase. miR156 levels are highest in Arabidopsis seedlings and, in full agreement with its role in regulating the juvenile-to-adult phase transition, these levels decline further during development (Wang et al., 2009; Wu and Poethig, 2006). How this decline is accomplished remains unknown. In seedlings, miR156 levels remain largely unaffected in both loss- and gain-of-function studies of the core factors that act in the well-established flower-promoting pathways (Wang et al., 2009). Low temperatures may have a mild positive effect on a few MIR156 loci but none of the known ambient temperature pathway components (Lee et al., 2007) seems to affect miR156 accumulation (Lee et al., 2010). miR156 thus seems to define an independent endogenous flowering pathway that acts in an age-dependent manner. The nature of the age-dependent factor(s) that regulate the MIR156 loci remains a mystery but recent organ ablation experiments suggest that leaves act as a source of a miR156-repressing factor (Yang et al., 2011).

Whereas the molecular genetic factors that act immediately upstream of MIR156 remain to be discovered, the events downstream of MIR156 expression have already been unveiled. In Arabidopsis, the targets of miR156 include transcripts of 11 of the 17 SPL genes, and other miR156 targets that represent orthologues or likely orthologues of SPL-like genes have been identified in other species (summarized in Table 1). Experimental evidence that the miRNA recognition elements (MREs) of these transcripts are indeed responsive and functional is diverse and convincing (Franco-Zorrilla et al., 2007; Martin et al., 2010b; Park et al., 2005; Schwab et al., 2005; Shikata et al., 2009; Usami et al., 2009; Wang et al., 2009; Wang et al., 2008; Wu et al., 2009; Wu and Poethig, 2006; Xing et al., 2010).

As unique targets of miR156, SPL genes are expected to play key roles in the vegetative and possibly the reproductive phase change, but, probably owing to extensive redundancy, single mutants of SPL genes do not show strong phenotypes. The analysis of specific targets or groups of targets informed by phylogeny and, in particular, paralogous relationships (Guo et al., 2008; Riese et al., 2007; Xie et al., 2006), has helped to unveil their respective roles in developmental phase transitions.

The three smallest of all SPL genes in Arabidopsis, SPL3, SPL4 and SPL5 are exceptional in that they carry a miR156 MRE in their 3′ UTRs (Gandikota et al., 2007). They are orthologous to the very first SBP-box genes identified in snapdragon, which were originally proposed to regulate flowering (Klein et al., 1996). In Arabidopsis, constitutive overexpression of a miR156-resistant SPL3 (rSPL3) transgene, indeed results in early flowering (Fig. 2D) (Cardon et al., 1997), and similar results can be obtained for rSPL4 and rSPL5 (Wu and Poethig, 2006). Moreover, such transgenics show a precocious appearance of adult leaf traits, providing clear support that these SPL genes promote the expression of the adult phase (Gandikota et al., 2007; Wang et al., 2009; Wu et al., 2009; Wu and Poethig, 2006). Transgenics that constitutively overexpress SPL3 flower only slightly earlier than do wild-type plants, reflecting the repressing strength of endogenous miR156 (Gandikota et al., 2007; Wu and Poethig, 2006).

Single loss-of-function mutants for SPL9 or SPL15 do not show strong phenotypes (Schwarz et al., 2008). However, a double mutant for these likely paralogous genes mimics mild miR156 overexpressors, indicating a role for these genes in promoting the juvenile-to-adult phase transition (Schwarz et al., 2008; Wang et al., 2008). In fact, the expression of a miR156-insensitive form of SPL9 (rSPL9) results in plants that virtually skip the juvenile phase (Wang et al., 2009; Wu et al., 2009). Furthermore, SPL9, through miR172, promotes the characteristic appearance of abaxial trichomes on adult leaves, possibly through the direct transcriptional activation of MIR172b (Wu et al., 2009). In turn, miR172 promotes abaxial trichome production at least in part by repressing TARGET OF EAT1 (TOE1) and TOE2 (discussed in more detail later). SPL9 also exerts control over the acropetal change in trichome density along the shoot, another hallmark of vegetative phase change (Telfer et al., 1997), by directly regulating two negative regulators of trichome formation: TRICHOMELESS1 and TRIPTYCHON (Yu et al., 2010).

SPL9 and SPL15 control not only phase dimorphisms, but also the competence to flower in response to the appropriate photoperiod (Schwarz et al., 2008). Furthermore, an spl9 spl15 double mutant partly phenocopies transgenics that constitutively overexpress miR156, such that they both exhibit shortened plastochron (Schwarz et al., 2008; Wang et al., 2008). Interestingly, Wang et al. (Wang et al., 2008) could detect SPL9 expression in leaf anlagen and primordia but not in the shoot apical meristem (SAM). In addition, the authors observed that elevated expression of rSPL9 under leaf primordia-specific promoters increased plastochron (see Glossary, Box 1) length, and concluded that the effect of SPL9 on the initiation of new leaf primordia is due to its expression in existing leaf primordia. The nature of this signalling remains unknown but may be related to cell proliferation and expansion, features affected by the more and smaller cells (msc) mutants, of which one (the msc1 mutant) represents a miR156-resistant allele of SPL15 (Usami et al., 2009).

SPL9 and SPL15 seem to share functional redundancy with SPL3/4/5 with respect to their effects on abaxial trichome production and petiole length, but not with regard to their effects on leaf shape (Wu et al., 2009) or plastochron length (Wang et al., 2008). Some functional specialization with respect to different features of shoot development may thus already be recognizable between these subgroups of miR156-targeted SPL genes.

Loss of SPL2 function weakly enhances the spl9 spl15 double mutant phenotype (Schwarz et al., 2008). Phylogenetic analysis places SPL2 in the same clade as SPL10 and SPL11 (Riese et al., 2007). Single mutants, as well as the double spl2 spl10 and spl2 spl11 mutants, look similar to wild type (Shikata et al., 2009; Wu et al., 2009). The SPL10 and SPL11 loci are in close physical proximity, complicating the generation of triple mutants to study further redundancy. Shikata et al. (Shikata et al., 2009) circumvented this problem by generating plants that express dominant repressor versions of the respective genes. In these transgenics, the rosette leaves were narrower but abaxial trichome initiation and flowering time remained unaffected. However, the expression of miR156-insensitive versions of SPL10 and SPL11 (rSPL10 and rSPL11) accelerated the expression of adult traits, producing phenotypes that are similar, but not identical, to those manifested by rSPL9 (Wu et al., 2009). In normal development, SPL2, SPL10 and SPL11 thus only weakly affect leaf shape, indicating that they have a minor role in the vegetative phase change. Actually, this subgroup of miR156-targeted SPL genes may play its major role in early embryonic patterning by redundantly regulating the embryonic morphogenesis-to-maturation phase transition (Nodine and Bartel, 2010). Within this context, it is interesting to note that Luo et al. (Luo et al., 2006) observed a striking increase in miR156 levels associated with differentiating rice embryonic calli (see Glossary, Box 1) in vitro, suggesting a role for miR156, and thereby for its targets, in embryogenesis.

Our current knowledge on the roles of the remaining miR156-targeted SPL genes, SPL6 and SPL13, is still very limited. To date, loss- or gain-of-function SPL6 mutant phenotypes have not been reported. SPL13, however, may play a role in the cotyledon to vegetative-leaf stage switch, as the expression of a miR156-resistant SPL13 transgene (rSPL13) results in a slight delay in the emergence of the first true leaves, an effect mediated by the miR172 target SCHNARCHZAPFEN (SNZ) (Martin et al., 2010a; Martin et al., 2010b).

As direct targets of miR156, genes such as SPL3, SPL4 and SPL5, as well as SPL9 and SPL15, have been shown to be important determinants in both the vegetative phase transition and flowering competence. Accordingly, and concomitantly with decreasing miR156 activity, their transcript levels gradually increase during aging (Cardon et al., 1999; Schmid et al., 2005; Wu and Poethig, 2006).

As mentioned before, vernalization, photoperiod and GA-dependent flowering pathways do not have an obvious effect on miR156 levels in seedlings. In addition, these flowering pathways also leave the expression levels of miR156 targets such as SPL3 and SPL9 largely unaffected in seedlings (Wang et al., 2009). Upon photoperiodic induction of adult plants, however, SPL3, SPL9 and other miR156-targeted SPL genes become strongly upregulated in the shoot apex (Schmid et al., 2003), indicating an additional role in establishing inflorescence meristem or floral meristem (see Glossary, Box 1) identity.

One of the major events in photoperiodic induction in Arabidopsis, i.e. the generation of a flower-promoting and leaf-derived signal that is sensed by the SAM, involves the activation of FLOWERING LOCUS T (FT) in the leaf. FT protein is subsequently translocated to the shoot apex where its interaction with the bZIP transcription factor FD results in the activation of SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1) and a meristem identity switch from vegetative into inflorescence. Together, FT/FD and SOC1 are able to activate LEAFY (LFY), APETALA1 (AP1) and FRUITFULL (FUL), which subsequently show mutual activation but negatively feed back on SOC1, thereby promoting the switch from IM to FM identity (Amasino, 2010; Srikanth and Schmid, 2011). The Arabidopsis inflorescence, however, remains indeterminate as TERMINAL FLOWER1 prevents the activation of these floral meristem identity genes in the IM itself such that only its lateral meristems become transformed into flowers (Preston, 2010).

Different lines of evidence show that SPL3 and SPL9, despite being from different SPL clades, target the same floral meristem identity genes. Using X-ChIP, Yamaguchi et al. (Yamaguchi et al., 2009) showed that SPL3 can bind to promoter and intragenic regions of LFY, AP1 and FUL in vivo. In addition, Wang et al. (Wang et al., 2009) reported SPL9 to be a direct regulator of FUL and AP1, and identified SOC1 and its likely paralogue AGL42 as further direct targets.

In order to promote flowering, a photoperiod-dependent pathway and an age-dependent pathway (i.e. a pathway activated by the seemingly autonomous decline in miR156 activity), thus converge through FT/FD and miR156-targeted SPL genes, respectively, on a set of target genes that determine inflorescence and floral meristem identity. The absence of SPL activity, as a consequence of high miR156 levels early in shoot development, renders the plant insensitive to FT/FD-dependent flower induction. However, the ability of these photoperiod-dependent factors to induce flowering becomes established when miR156 levels decline and the expression of its targets increases. Notably, when miR156-targeted SPL activity continues to rise, plants will eventually flower without the requirement for photoperiod-dependent FT/FD activity (Wang et al., 2009). In fact, overexpression of a rSPL3 transgene may result in a complete consumption of the IM in creating floral organs and thereby in determinate growth of the inflorescence (Gandikota et al., 2007).

The first few flowers formed in Arabidopsis often produce significantly fewer seeds than do later flowers, which may be due to an inflorescence architectural effect, i.e. dependent on the position of the flower in the inflorescence and not only to environmental variation (Diggle, 2002). Interestingly, miR156-targeted SPL genes function redundantly with the SPL8 gene, which is not targeted by miR156, in promoting Arabidopsis flower fertility (Xing et al., 2010). Furthermore, miR156-targeted SPL gene activity has recently been coupled to increasing flavonol to anthocyanin ratios within the developing Arabidopsis shoot (Gou et al., 2011). Anthocyanins, like trichomes, may discourage small herbivores at early stages of development, whereas flavonoids are required for full fertility at later stages of development (Thompson et al., 2010). An increase in miR156-targeted SPL gene activity even after plants have bolted and initiated their first flowers may thus explain the ontogenetic shift observed in the degree of flower fertility and seed set.

In addition to miR156, the miRNA miR172 and its targets have been implicated in phase transitions in Arabidopsis. miR172 levels increase as plants age, in a pattern that is complementary to that of miR156 (Fig. 3). In contrast to miR156, the abundance of which is controlled by plant age, miR172 expression appears to be under photoperiodic control (Jung et al., 2007). miR172 has been found to target several transcripts that encode transcriptions factors involved in the repression of flowering (Table 2). For example, the transcripts of six genes in Arabidopsis that all encode AP2-type transcription factors are targeted by miR172 (Chen, 2004). One of these is AP2 itself, a gene well known for its role in floral patterning (Bowman et al., 1991; Jofuku et al., 1994). The other five targets, TOE1, TOE2, TOE3, SCHLAFMÜTZE (SMZ) and SNZ, also act as repressors of flowering (Aukerman and Sakai, 2003; Jung et al., 2007; Mathieu et al., 2009; Schmid et al., 2003). However, only recently has it been shown that miR172 and its targets also control the juvenile-to-adult transition in Arabidopsis. In particular, the overexpression of miR172 and its targets (Aukerman and Sakai, 2003; Chen, 2004; Jung et al., 2007; Mathieu et al., 2009; Schmid et al., 2003) results in effects that are opposite to those caused by miR156 overexpression (Fig. 2C,E,G). Wu et al. (Wu et al., 2009) also reported that in transgenic plants that constitutively overexpress miR172, abaxial trichomes are produced earlier than is normal. Conversely, the formation of abaxial trichomes is delayed in the mir172a mutant. In either case, leaf shape is morphologically normal, indicating that MIR172 genes affect only a subset of the traits (e.g. epidermal patterning) that are normally associated with juvenility.

MIR172b is a direct transcriptional target of SPL9, as demonstrated by chromatin immunoprecipitation and by the induction of miR172 in transgenics in which SPL9 activity can be conditionally induced (Wu et al., 2009). This finding suggested that SPL proteins might promote adult epidermal identity by indirectly repressing miR172-targeted AP2-like transcription factors. Indeed, further analyses revealed that the loss of SPL9 delayed the appearance of abaxial trichomes in a toe1 toe2 double mutant. However, the observed effect was rather mild, indicating that TOE1 and TOE2 only contribute to, but are not solely responsible for, this epidermal phenotype (Wu et al., 2009).

In summary, it appears that, via their antagonistic function, miR156 and miR172 ratios determine the timing of the juvenile-to-adult transition, and that the vegetative phase change in particular is controlled by their sequential activity (Fig. 3).

In addition to their rather subtle role in promoting juvenile epidermal traits, miR172 and its targets have an additional function in regulating the transition to flowering. This was first made apparent when Aukerman and Sakai (Aukerman and Sakai, 2003) isolated an extremely early flowering mutant from an activation-tagging screen and identified MIR172b as the causal gene. The same screen also identified the late flowering toe1-D mutant, which was found to overexpress the miR172 target TOE1. Analyses of loss-of-function alleles revealed that toe1 mutants flowered significantly earlier than did wild-type plants, and that this phenotype was enhanced in a toe1 toe2 double mutant that also lacked the closely related TOE2 gene (Aukerman and Sakai, 2003; Jung et al., 2007). However, even the double mutant flowered significantly later than miR172b-overexpressing lines, suggesting that other factors act redundantly with TOE1 and TOE2 to repress flowering.

Schmid and colleagues (Schmid et al., 2003) reported that overexpression of SMZ also significantly delayed flowering, particularly under the condition of long days. In contrast to toe1, plants mutant for SMZ or for its paralog SNZ, did not flower earlier than control plants. Loss of either SMZ or SNZ function, however, did further accelerate flowering when combined with toe1 and toe2. Nevertheless, even in the quadruple toe1 toe2 smz and snz mutant, flowering occurred significantly later than in transgenics constitutively overexpressing miR172 (Mathieu et al., 2009), suggesting further redundancy within this gene family. This was later confirmed when it was found that AP2, besides its well-described role in floral patterning, also functions as a floral repressor (Yant et al., 2010). Eventually, a hextuple mutant lacking the functions of all six miR172 targets was found to essentially phenocopy the effect of constitutive miR172 expression (Yant et al., 2010).

The transition to flowering is under the control of a complex genetic network that perceives and integrates various endogenous and environmental cues. Boss and colleagues (Boss et al., 2004) have proposed that the different pathways that regulate flowering can be viewed as either pathways that actively promote flowering, such as the photoperiod, light quality, gibberellic acid and ambient temperature pathways, or pathways that enable flowering by removing floral repressors from the system. According to this hypothesis, miR156-targeted SPL genes and miR172 can be considered `enabling factors', as their activity ultimately leads to the shutting down of the AP2-like floral repressors that regulate `meristem competence' (Bernier, 1988). In this light, miR156, which has been shown to prevent precocious flowering (Wang et al., 2009) until plants have reached a permissive age, can be considered as an `enabling factor'. However, the miR156-targeted SPL genes do not only regulate the expression of miR172 and thus the AP2-like floral repressors, SPL proteins have also been shown to directly bind to and promote the expression of floral integrator genes, such as SOC1 (Wang et al., 2009) and floral meristem identity genes such as LFY, FUL and AP1 (Yamaguchi et al., 2009). As often occurs, the more we learn, the more complex the situation becomes.

Besides its role in regulating phase transitions, miR172 and its targets appear to also play a role in floral patterning. The angiosperm prototype flower consists of four different organs that are organized in concentric whorls. The patterning of the flower is explained by the `classical' ABC model (see Box 2), according to which the combinatorial interaction of three classes of homeotic functions (A, B and C), provides the positional information that determines the fate of the emerging floral organs (Causier et al., 2010; Lohmann and Weigel, 2002). In brief, in the outermost whorl of the Arabidopsis flower, sepals are specified by the activity of AP1 and AP2 (A class). In the second whorl, the activity of A-class proteins overlaps with those of the B-class proteins PISTILLATA and APETALA3 to establish petals. Similarly, overlapping activity of the B-class proteins with that of the C-class protein AGAMOUS (AG) induces stamen fate in the third whorl. Finally, in the innermost whorl, AG controls the formation of carpels. An important feature of the ABC model is that A-class and C-class function are mutually exclusive. More recently, four highly redundant genes, SEPALLATA1 (SEP1), SEP2, SEP3 and SEP4, that are required for the pattering of all four whorls of the emerging flower have been identified and have been added to the ABC model as E-class function genes.

Early reports had suggested that AP2 transcripts could be detected in all parts of the emerging flower (Jofuku et al., 1994), which was puzzling given the A-class function of AP2. However, it has recently been reported that AP2 expression is actually restricted to the outer two whorls (Wollmann et al., 2010). Interestingly, the expression of miR172 was shown to be restricted to the centre of young floral primordia. As it was recently shown that AP2 can directly bind to and repress the expression of MIR172b (Yant et al., 2010), it would appear that AP2 and miR172 are engaged in a negative-feedback loop that presumably helps to establish a sharp boundary between the inner and the outer whorls of the emerging flower (Fig. 3). In this scenario, AP2 not only directly binds to AG and prevents its transcription in the outer two whorls, but it is also at the same time cleared from the centre of the floral primordium by the activity of miR172.

Both miR156 and miR172 belong to a subset of evolutionary conserved miRNAs that are present throughout the angiosperms (Axtell and Bowman, 2008; Cuperus et al., 2011). However, one of the best-studied miR172 targets is AP2, and its orthologues in other species seem to have very divergent roles. This raises the issue of whether the roles of miR156 and miR172 in regulating

phase transitions can be generalized to other plant species. Results obtained in maize, rice, tomato, potato and, more recently, in trees, however, indicate that these miRNAs are not only conserved in sequence but also in function.

In maize, mutations that affect phase transitions have been mapped to miR156 and AP2-like genes. For example, the dominant maize Corngrass1 (Cg1) mutant affects multiple developmental traits, including leaf initiation and floral architecture. In particular, the Cg1 mutant initiates more juvenile leaves than do control plants (Chuck et al., 2007a). This was found to be caused by a STONER retrotransposon insertion in the upstream regulatory region of a tandem cluster of two miR156 genes, zma-miR156b and zma-miR156c (Chuck et al., 2007a). Maize contains at least 17 SBP-box genes that are potential targets of miR156 (Hultquist and Dorweiler, 2008); seven of these show reduced expression in Cg1. Among them is teosinte glume architecture 1, a gene that has been implicated in the domestication of maize (Wang et al., 2005). Another SBP-box gene that is misregulated in Cg1 is tasselsheath4, which is known to regulate the development of bracts and meristem boundaries (Bensen et al., 1995; Chuck et al., 2010). From this, it would appear that the very strong phenotype of the Cg1 mutant is the result of the reduced expression of several SBP-box genes that regulate various aspects of maize development. Similar to what has been described in miR156-overexpressing Arabidopsis, miR172 expression is reduced in the Cg1 mutant (Chuck et al., 2007a), suggesting that at least some MIR172 genes are among the targets of the maize SBP-domain proteins. Taken together, these findings have lead to the speculation that changes in the expression levels and temporal/spatial expression patterns of MIR156 genes and their targets contributed to the evolution of grasses (Chuck et al., 2007a; Wang et al., 2005).

Another gene that regulates the juvenile-to-adult transition in maize is Glossy15 (Gl15). Gl15 has been cloned and encodes an AP2-like transcription factor that was shown to promote juvenile leaf epidermal traits (Moose and Sisco, 1996). More recently, it has been found that Gl15 is a target of miR172, and that downregulation of Gl15 promotes vegetative phase change in maize (Lauter et al., 2005), similar to the way in which downregulation of TOE genes in Arabidopsis promotes vegetative phase change (Wu et al., 2009). Besides Gl15, another AP2-like gene and target of miR172, Indeterminate spikelet1 (Ids1), which specifies determinate spikelet meristem fate, has been identified in maize. Mutations in the miR172-binding site have been shown to result in the dominant allele of Ids1, tasselseed6 (Chuck et al., 2007b). Furthermore, it has been shown that Ids1 is the key target of Tasselseed4, which represents the MIR172e gene of maize (Chuck et al., 2007b). Interestingly, maize Ids1 is orthologous to the Q locus of wheat, dominant alleles of which have been shown to play an important role in wheat domestication (Simons et al., 2006). Similarly, the maize flowering time quantitative trait locus Vegetative to generative transition 1 has been mapped to a cis-regulatory region upstream of an AP2-like gene (Salvi et al., 2007). These results from wheat and maize suggest that changes in miRNAs and their targets may have contributed to plant genetic adaptation throughout breeding and evolution.

Interestingly, MIR172 overexpression has been shown to promote flowering and also tuberization in potato (Martin et al., 2009). Moreover, the effect of miR172 on tuberization has been found to be graft transmissible (Martin et al., 2009). Whether long-distance movement of miR172 itself was responsible for this phenomenon is unclear. The finding that overexpression of MIR172 in potato roots was insufficient to induce tuberization seems to argue against a non cell-autonomous function of miR172. However, as MIR172 is expressed in the vasculature of potato and as miR172 has been detected in the phloem exudate from Brassica napus (Buhtz et al., 2008), a non-cell-autonomous effect of miR172 on the control of tuberization is still possible.

As mentioned above, morphological differences between the juvenile and the adult phase tend to be more pronounced in woody perennial plants than in annuals such as Arabidopsis or maize. Given their slow development, data on the molecular nature of the factors that regulate phase transitions in trees are sparse. However, most recently Wang et al. (Wang et al., 2011) investigated the role of miR156, miR172 and their targets in the control of vegetative phase change in trees. In several species, they observed sequential changes in miR156 and miR172 expression levels, similar to those observed in Arabidopsis and maize (Wang et al., 2011). Furthermore, the overexpression of miR156 in a transgenic poplar lead to the reduced expression of SPL genes and of miR172, and drastically prolonged the juvenile phase. From these data it can be concluded that, at least within the angiosperms, miR156/miR172 and their respective targets can be considered to be general regulators of phase transitions, irrespective of whether a plant follows an annual or perennial lifestyle.

miR156 appeared early in land plant evolution, and the mature miRNA has been detected in various mosses, ferns and gymnosperms (Arazi et al., 2005; Axtell and Bartel, 2005; Cuperus et al., 2011; Zhang et al., 2006). The role of miR156 in these plants remains to be determined. In contrast to miR156, miR172 appears to be angiosperm specific and it has not been cloned from other land plants (Axtell and Bowman, 2008; Cuperus et al., 2011), even though the expression of miR172 has been detected by microarrays of RNA extracted from ferns (Axtell and Bartel, 2005) and has been computationally predicted in Physcomitrella (Fattash et al., 2007). Regardless of the exact time of miR172 emergence, the question remains as to how miR172 and AP2-like transcription factors were adopted to function in the regulation of phase change. As miR172 apparently arose with the appearance of flowering plants and as miR172/AP2 contribute to patterning the young flower primordium, it seems possible that this reflects the ancient function of this regulatory module.

Over the past few years, two evolutionary conserved miRNAs, miR156 and miR172, and their targets have emerged as key regulators of various phase transitions in plants. Although we have learned a lot about the function of these miRNAs and the genes regulated by their targets, the SPL and AP2-like transcription factors, we know very little about the factors that regulate the temporal and spatial expression of the miRNA genes themselves. For example, levels of miR156 diminish with increasing age of the plant. However, it is largely unclear how plants determine `age' and how this information is used to regulate gene expression. Identifying the signals and signalling pathways that regulate miR156 expression would not only further our understanding of how plants regulate vegetative phase transitions but would also probably provide important insight into the molecular basics of plant `aging'.

One characteristic feature of plants is that different shoots on the same plant can exist in different developmental phases, a condition called heteroblasty. In addition, many plants display a remarkable phenotypic plasticity, i.e. the ability to adapt their growth rate and development, allowing them to thrive and survive under varying environmental conditions. However, during any given developmental phase, phenotypic plasticity is limited by homeostatic mechanisms that ensure a minimal degree of developmental robustness. Unfortunately, heteroblasty and plasticity can result in very similar phenotypic outcomes, which has sometimes led to confusion between these two phenomena. However, it is reasonable to assume that the homeostatic mechanisms that limit a plant's plasticity need to be reprogrammed or to be overcome in order to allow developmental phase changes to occur. The recent insight into the molecular mechanisms that shape heteroblastic traits, i.e. the role of miR156, miR172 and their respective targets, will allow us to better discriminate between these two phenomena. At the same time, studying the molecular genetic mechanisms that underlie these phase transitions will lead to a better understanding of what drives plant diversification and distribution in relation to ecological factors.

From a molecular genetic point of view, the most dramatic phase change in the life cycle of plants is probably the diploid to haploid phase change, i.e. between the sporophytic and gametophytic generations. In fact, the ultimate goal of the process of flowering is the production of the gametophytic generation through meiosis in the floral reproductive organs. It will be interesting to see whether central components of the pathways that lead to flowering, i.e. miR156, miR172 and their targets, also contribute to the diploid to haploid phase transition. miR172 appears with the evolution of the angiosperms and it seems unlikely that its ancestral function is in regulating the sporophytic-to-gametophytic phase transition. By contrast, miR156 can be found in all land plants, including mosses, liverworts, ferns and gymnosperms, making it a good candidate for comparative studies.

The function of miRNAs in regulating phase transitions is, of course, not limited to plants. As a matter of fact, the very first miRNAs isolated from Caenorhabditis elegans, lin-4 and let-7, have been shown to play important roles during C. elegans post-embryonic development (Kato and Slack, 2008; Vella and Slack, 2005). In brief, lin-4 and one of the targets of lin-4, LIN-14, are required for the larval stage 1 (L1)-to-L2 transition. Similarly, let-7 and its key target LIN-41 are essential for the L4-to-adult transition. lin-4 and let-7 belong to a subgroup of evolutionary highly conserved miRNAs, which is indicative of a conserved function. Interestingly, both lin-4 and let-7 need to be upregulated during development to allow phase transitions to occur. This temporal regulation resembles that of miR172 in plants. Although the observation that miRNAs play a role in phase transitions in the different kingdoms is an interesting one, it begs the question as to what it is that makes miRNAs so particularly useful for controlling phase changes and developmental timing? One hypothesis is that the temporal activation of miRNAs results in a rapid clearing of their target mRNA from cells, allowing for a swift and irreversible phase transition. Another possibility is that organisms employ miRNAs against regulators of phase transition to confer robustness to the developmental programs between the transitions. This may be achieved by repressing leaky transcripts in the case of high miRNA to target ratios or to buffer against fluctuations in target expression in the case of co-expression at intermediate levels (Hornstein and Shomron, 2006). The latter scenario, in particular, may enable organisms to manage duplication events or altered expression patterns that might occur in the course of the evolution of key developmental control genes, the altered expression levels of which would otherwise be detrimental. The co-evolution of such key developmental genes with their targeting miRNAs would allow more gradual changes and could open the opportunity for the evolution of new varieties and speciation to evolve. Clearly, no matter the particular biological system we are working on, we can still expect many exiting new and surprising findings from the analysis of miRNAs and their role in regulating phase transitions.

We thank D. Weigel, J. W. Wang, L. Yant and members of our labs for comments on the manuscript.