A common mechanism controls the life cycle and architecture of plants

The life cycle of most organisms can be divided into several developmental phases, from an early juvenile period when resources are built up, through to a reproductive phase (Poethig, 1990; Schultz and Haughn, 1993; Telfer et al., 1997). In the case of Arabidopsis, wild-type plants pass through three main phases, reflecting a sequence of identities in the shoot apical meristem (Figs 1, 2). Following germination the meristem has a vegetative (V) identity and produces leaf primordia on its flanks to form a rosette. The length of this V phase depends on the environmental conditions, and is regulated through a complex network of flowering-time genes (Martinez-Zapater et al., 1994). On receipt of favourable signals, reproductive development is initiated and the apical meristem acquires an inflorescence identity (I), which itself has 2 distinct phases. During the first-inflorescence phase (I₁), 2-3 cauline leaf primordia are produced which subtend axillary inflorescence meristems. This is followed by a second-inflorescence phase (I₂), during which determinate floral meristems are made on the periphery of the apex (Figs 1, 2). According to an alternative viewpoint, I₁ can be considered as part of the V phase, such that the apex switches directly to floral meristem production at the end of V (Hempel and Feldman, 1994). In this study, however, to allow a precise description of different plant phenotypes, it was essential to distinguish the period of cauline leaf initiation from other phases of growth, and for that reason we defined this period as I₁. In both descriptions, the fate of floral meristems depends on the action of meristem identity genes such as LEAFY (LFY) and APETALA1 (AP1) (Mandel et al., 1992; Weigel et al., 1992; Bowman et al, 1993; Gustafson-Brown, 1994; Mandel and Yanofsky, 1995; Weigel and Nilsson, 1995). The inflorescence apex itself remains indeterminate and continues to form floral meristems on its periphery until senescence occurs.

TERMINAL FLOWER 1 (TFL1) is a key gene which affects the developmental phases and architecture of Arabidopsis (Shannon and Ry Meeks-Wagner, 1991; Schultz and Haughn, 1991; Alvarez et al, 1992; Ray et al., 1996). The most notable feature of the tfl1 mutant is that a flower forms at the tip of the inflorescence, unlike the wild-type apex which grows indeterminately (Fig. 1). In accordance with this, TFL1 expression is observed at high levels in the wild-type inflorescence apex (Bradley et al., 1997). Additionally, in tfl1 mutants, LFY and AP1 expression can be observed in the inflorescence apical meristem, correlating with its floral identity, whereas in wild type, expression of both genes is restricted to young floral meristems on the periphery of the apex (Weigel et al., 1992; Bowman et al., 1993; Gustafson-Brown et al., 1994; Bradley et al., 1997).

In addition to the terminal flower phenotype, the tfl1 mutant also has a significantly shorter vegetative phase than wild type, due to an earlier commitment to reproductive growth (Schultz and Haughn, 1991; Shannon and Meeks-Wagner, 1991; Bradley et al., 1997). This correlates with TFL1 expression in vegetative apices, although this is weaker than in infloresence meristems. The TFL1 gene product therefore has two effects: firstly it delays the switch to reproductive growth (flowering time) and secondly it prevents the formation of a terminal flower (Bradley et al., 1997).

The effects of TFL1 could be explained in two ways. In one model, TFL1 may have a dual function and act differently at separate times in the life cycle, affording a link between the regulation of flowering time and meristem identity. First it acts to control flowering time and secondly it acts by a separate mechanism to inhibit the activity of the floral meristem identity genes, LFY and AP1, in the apex (Shannon and Meeks-Wagner, 1991, 1993; Hicks et al., 1996; Bradley et al., 1997). An alternative view is that TFL1 could have a single function influencing the rate of progression through all phases of the shoot apical meristem (Bowman et al., 1993; Schultz and Haughn, 1993). According to this second model, the tfl1 apical meristem progresses more rapidly through each phase (V, I₁ and I₂) such that it then enters a fourth terminal phase with a floral (F) identity, which is not normally attained before senescence occurs in wild type.

To distinguish between these possibilities, we produced transgenic Arabidopsis plants in which TFL1 was ectopically expressed at all stages of the life cycle. If the dual function model was applicable, we anticipated that these plants might show a considerable delay in flowering time and form an inflorescence comparable to that of a lfy;ap1 double mutant, which never produces normal flowers (Weigel et al., 1992; Schultz and Haughn, 1993). By contrast, if TFL1 fulfilled a single function influencing the rate of phase change, then there would be a consistent lengthening of all phases, but normal flowers would eventually be produced.

We show by a combination of molecular and physiological analyses that TFL1 affects all phases of the Arabidopsis life cycle. Transgenic plants expressing TFL1 throughout development display extended V and I₁ phases compared to wild-type controls but do ultimately produce a normal I₂ phase and fertile flowers. Consistent with this, we show by in situ hybridisation, that the upregulation of LFY and AP1 is delayed in comparison to wild-type plants. We therefore conclude that TFL1 does not function by directly restricting the activity of the meristem identity genes LFY and AP1, but rather modulates the mechanism of shoot phase transitions. These results are consistent with TFL1 having a single function, over the entire life cycle, to maintain the duration of growth phases of the Arabidopsis shoot apex.

The TFL1 gene (EST 129D7T7) was obtained from the Arabidopsis Biological Resource Centre at Ohio State University in pZL1. This was digested with SmaI and XbaI to release the TFL1 cDNA. The TFL1 cDNA fragment was ligated into plasmid SLJ4K1 (Jones et al., 1992) containing the CaMV 35S promoter, after this had been cut with ClaI, filled in, and then digested with XbaI. The product of the ligation was named pSR15. The 35STFL1 fusion was released from pSR15 by digestion with EcoRI and HindIII and ligated into the binary vector SLJ 44024A (Jones et al., 1992), which conferred kanamycin resistance, to give pJAM 2076.

All transformation experiments were performed on Arabidopsis plants of the ecotype Columbia. Agrobacterium mediated transformation with pJAM 2076 was attempted by both root transformation (Valvekans et al., 1988) and vacuum infiltration of flowers (Bechtold et al., 1993). Approximately 30 primary transformants derived from root transformation were established on soil. These often displayed excessive vegetative growth and sometimes produced aerial rosettes. However, it was difficult to establish whether these were genuine features or artifacts of tissue culture. Three primary root transformants set seed and could be analyzed in future generations. A further 3 primary transformants were obtained from vacuum infiltration, all of which set seed. A total of 6 lines were therefore available and all displayed similar phenotypes in subsequent generations. Two representative independent lines, JI.At1 and JI.At2, each derived from a T₂ plant which contained the 35STFL1 insertion at a single locus, were selected for further study. All subsequent analysis described was performed on the T₃ population from these plants.

For all experiments, seeds were imbided and stratified for 5 days at 4°C in the dark, then germinated and grown on soil in growth cabinets at 20°C In the photoperiod shift experiments to assess commitment, 35STFL1 plants of line JI.At1 were grown in long days (16 hours light/8 hours dark, 90-120 μE m⁻² s⁻¹) and transferred to short days (8 hours light/16 hours dark) after 8, 12, or 19 days. Plants were also grown in continuous long days and short days as controls. Numbers of rosette leaves were counted for approximately 10 plants in each batch. Errors indicate standard error about the mean with 95% confidence limits attached.

Shoot apices were moulded as described by Green and Linstead (1990). Casts of the moulds were then sputter coated with gold and viewed and photographed using a scanning electron microscope.

Longitudinal sections of plant tissue were probed with digoxigenin-labelled TFL1, LFY, or AP1 antisense RNA. These probes were made using plasmids pJAM 2045 (Bradley et al.,1997), pDW122 (kindly provided by Detlef Weigel; Weigel et al., 1992), and pKY89 (kindly provided by Martin Yanofsky; Mandel et al., 1992) respectively. RNA signal was detected as a dark blue/black colour on a light blue background when viewed under the light microscope. In situ hybridisation was carried out as described by Coen et al. (1990).

We established six independent lines in which TFL1 was expressed from the 35S CaMV promoter (Odell et al., 1985), all of which displayed similar dominant and heritable phenotypes (Fig. 1). These plants had greatly extended V and I₁ phases: under long day (LD) conditions, the V phase was 2-3 times longer, and the I₁ phase 15-20 times longer than in wild type (Fig. 2). The latter two thirds of the I₁ phase displayed a novel region (I₁*), not observed in wild type, in which axillary shoots were not subtended by cauline leaves (Fig. 3A,C,D). These shoots became progressively more flower-like towards the apex; the uppermost typically comprising small clusters of flowers surrounded by leaf-like organs arranged in a whorled phyllotaxy (Fig. 3C). Sometimes such structures consisted of a flower with organ abnormalities and additional flowers buds in the axils of those organs (Fig. 3B). Eventually the 35STFL1 apex entered an I₂ phase during which it produced apparently normal fertile flowers (Fig. 3C). Occasionally, when the plants were very old, the shoot apex passed into a floral (F) phase, terminating in a flower (data not shown). The growth pattern of the primary apex was repeated by secondary shoots, (i.e. those growing from the axils of cauline and rosette leaves) giving the plant a highly branched architecture. This extensive branching resulted in the 35STFL1 plants becoming much larger and producing a much greater quantity of seed than wild type and tfl1 mutants. Since they all grew at approximately the same rate, it took proportionately longer for the 35STFL1 lines to set seed: the tfl1 mutant produced ripe seed by about 30 days, wild type by about 45 days, whilst the transgenics required 80–100 days.

In short day (SD) conditions, the 35STFL1 lines produced approximately twice as many rosette leaves as long day grown plants indicating that they were still sensitive to photoperiod. However, after bolting the inflorescence phenotypes were very variable between individuals of the same line. Some plants produced an inflorescence similar to those grown in long days but with a slightly increased number of nodes in the I₁ and I₁* phases before making normal flowers. One individual, however, remained in the I₁ phase, making over 50 cauline leaf nodes, and did not form flowers even after seven months. Other individuals also failed to produce flowers in a seven month period but produced large aerial rosettes during the I₁ phase. These aerial structures sometimes consisted of more that 20 leaves and developed at each axillary position instead of a single cauline leaf bearing a secondary inflorescence.

The eventual appearance of normal flowers in long day grown 35STFL1 plants indicated that TFL1 most likely acted through a common mechanism to extend the duration of phases. The appearance of an extra phase, I₁*, intermediate between I₁ and I₂ could also be accounted for by a general retardation in the rate of phase change.

To test this model further, patterns of meristem identity gene expression were examined in relation to the various phases. The timing of phase transitions was determined by monitoring the total number of nodes made by the primary shoot apex of 35STFL1 plants over time, by dissection and scanning electron microscopy (Fig. 4). By comparison with the total numbers of different structures visible on the main axis of adult plants (Fig. 2), it was possible to deduce time-points at which the phase transitions had taken place.

This analysis showed that 35STFL1 plants switch from making V to I₁ nodes at approximately 13-15 days after sowing, a delay of about 7 days compared to wild type (Bradley et al., 1997). To check that this was due to a delay in commitment to reproductive growth, batches of plants were moved from inductive (long days, LD), to non-inductive (short days, SD) conditions, at various times during their development. In SD regimes, 35STFL1 plants made about twice as many rosette leaves (41±6) compared to similar plants in LD conditions (21.9±0.9 rosette leaves). Those batches that received 8 or 12 inductive days had not been committed and produced 44±7 and 38±4 rosette leaves respectively. This was not significantly fewer than for the non-induced controls. Eventually 35STFL1 plants did become committed to reproductive growth, as revealed by batches receiving 19 LD which produced 21±2 rosette leaves. Wild-type controls given the same treatment made about 10 rosette leaves for the 8 and 12 day time-points.

The 35STFL1 plant apex continued in the I₁ phase until about day 20, after which it entered the novel I₁* phase, during which it produced shoots lacking cauline leaves. Finally at around 45-50 days, it switched to I₂, and initiated normal floral meristems. The I₁^*-I₂ transition was not an abrupt switch, but occurred gradually, as indicated by the increasingly flower-like nature of structures produced by the inflorescence meristem (Fig. 3). To examine how this related to the activity of floral meristem identity genes, expression of LFY and AP1 was compared to the timing of the different phases in wild-type and transgenic plants.

In wild type, small quantities of LFY RNA were detected in rosette and cauline leaf primordia of the V and I₁ phases (data not shown; Bradley et al., 1997; Blazquez et al., 1997, Hempel et al., 1997). Strong LFY expression was first observed at 10-12 days from sowing when the plant apex was in early I₂ (Fig. 5). AP1 was upregulated slightly later than LFY, at 12-14 days from sowing, and was not observed during wild-type vegetative growth.

In 35STFL1 plants, only weak LFY expression was noted during the V and I₁ phases, comparable to the levels observed in the V and I₁ phases of wild type. Control sections probed with TFL1 showed that regions expressing LFY overlapped with ectopic TFL1, although expression from the 35S promoter was somewhat patchy (Fig. 5). This suggested that TFL1 did not have a distinct role as an overall repressor of LFY expression. At 26 days, just after the start of the I₁* phase (Fig. 4), AP1 expression was not detected and LFY expression was still weak. Later in the I₁* phase, at the 35 day time-point, strong LFY expression was observed but AP1 RNA was still absent (Fig. 5). The gradient of structures observed in the I₁* phase was consistent with such a gradual increase in meristem identity gene activity. Eventually strong expression of both LFY and AP1 was observed in young floral meristems during I₂, as seen at the 59 day time-point. Again these genes were co-expressed in many regions containing TFL1 RNA. It appears, therefore, that TFL1 delays the upregulation of LFY and AP1 expression but cannot directly block their activity.

We have shown that TFL1 regulates the duration of growth phases of the shoot apex, and as a consequence, the overall morphology of Arabidopsis plants. Previous models for the action of TFL1 have suggested interactions with the flowering-time genes during the vegetative phase and a second role antagonising LFY and AP1 in the shoot apical meristem. We propose that TFL1 acts by influencing a central mechanism controlling the identity of shoot apical meristem at all stages of development throughout the life cycle.

Our main analyses were conducted under inductive long day conditions. However, the phenotype of the 35STFL1 lines did become more severe in short days, confirming that other factors act to extend the Arabidopsis life cycle, independently of TFL1, in non-inductive conditions.

Furthermore, the increase in rosette leaf number of 35STFL1 lines in SD compared to LD conditions, indicated that their extended vegetative phase and delayed flowering was not merely due to inhibition of flowering-time genes such as CONSTANS which mainly function to promote reproduction in long days (Koornneef et al., 1991; Putterill et al., 1995; Hicks et al., 1996; Simon et al., 1996).

LFY is expressed in young leaf primordia at low levels during the wild-type vegetative phase and is upregulated during reproductive development (Bradley at al, 1997; Blazquez et al., 1997; Hempel et al., 1997). In the 35STFL1 lines low level LFY expression persisted much longer, and upregulation was not observed until around 23 days after that of the wild type. AP1 upregulation was even later and did not occur until about 40-45 days after that in wild type. Therefore, TFL1 delays upregulation of floral meristem identity genes, rather than acting as a direct inhibitor throughout the life cycle.

It is also noteworthy that the upregulation of AP1 was significantly delayed compared to LFY. These genes are induced via separate pathways with AP1 closely following LFY induction (Simon et al., 1996). That AP1 upregulation was uncoupled from LFY upregulation, suggests that TFL1 impinges more stringently on the pathway governing AP1 activation than on that controlling LFY. Recent data suggests that this pathway could involve the activation of AP1 by genes such as FT and FWA (Ruiz-Garcia et al., 1997).

Newly initiated floral meristems arise from the periphery of the inflorescence meristem. It has been suggested that floral meristem identity genes such as LFY and AP1 act as mutual antagonists with TFL1, to ensure that floral meristems pursue a separate determinate growth pattern from the indeterminate inflorescence meristem which produced them (Shannon and Meeks-Wagner, 1993; Schultz and Haughn, 1993). This stand-point is supported by the complementary phenotypes of the tfl1 mutant (where flowers replace inflorescences) when compared to those of lfy and ap1 mutants (where inflorescences replace flowers). Additionally, lines which ectopically express LFY or AP1 from a 35S promoter, broadly phenocopy the tfl1 mutant (Weigel and Nilsson, 1995; Mandel and Yanofsky, 1995). Although we have demonstrated that TFL1 does not directly inhibit LFY or AP1, it is still not clear if LFY or AP1 can directly inhibit TFL1. Answering this question awaits a study of TFL1 expression in the 35SLFY and 35SAP1 plants.

A key factor which has become apparent from this work is that TFL1 does not have two separate functions, of first delaying reproduction and then inhibiting LFY and AP1 in the apex. Rather, the results show that TFL1 has a unified effect on the rate of progression through all the different phases of the shoot apex. This implies that there is a common mechanism underlying phase transitions during the life cycle, rather than there being unrelated mechanisms specific to each transition. In a similar way, common components, such as cdc2, are involved in progression through different phases of the yeast cell cycle (Forsburg et al., 1991). The complementary effects of accelerating phase transitions in the tfl1 mutant, and of retarding transitions in the 35STFL1 lines, indicate that TFL1 activity is both necessary and sufficient to influence this common process. By acting in this way, TFL1 can have profound consequences for the life cycle and overall architecture of the plant, modulating its degree of branching and determinacy.

The TFL1 protein is similar to phosphatidylethanolamine binding proteins (PBPs) of animals which can bind to membrane protein complexes (Bradley et al., 1996, 1997; Ohshima et al., 1997). This suggests that TFL1 might be involved in a signalling process, perhaps influencing the response to signals arriving at the plant apex. The effects of TFL1 might eventually be overcome if such signals were to reach a sufficiently high level; possibly accounting for the transition to a floral identity that sporadically occurs in very old 35STFL1 apices. Centroradialis (cen), a homologue of TFL1 which also has similarity to PBPs, is found in another flowering plant, Antirrhinum majus (Bradley et al., 1996). However, cen only appears to prevent the transition of the apex from an inflorescence to floral identity, and does not notably affect earlier developmental phases. It is possible that cen had an ancestral role similar to that proposed for TFL1 but that its involvement in other phase transitions has become redundant.

The production of a terminal flower is thought to be an ancestral state from which the indeterminate condition evolved (Stebbins, 1974). By accelerating progression through growth phases, the tfl1 mutant, in effect, recapitulates this aspect of the ancestral form. The evolution of an indeterminate inflorescence from an ancestor with a determinate inflorescence might be regarded as an example of neoteny (De Beer, 1940); a process in which juvenille traits persist into later periods of the life cycle. According to this view, the indeterminate growth of the wild-type Arabidopsis apex may have arisen from TFL1 activity retarding its progression and ensuring that it never reaches the mature determinate floral phase, exhibited by the ancestor.

The tfl1 mutant and the 35STFL1 plants illustrate different reproductive strategies. The tfl1 mutant passes through its life cycle very rapidly but produces relatively small amounts of seed. By contrast 35STFL1 plants grow for much longer and eventually produce a larger quantity of seed. In the same way, some plants have a very short life-span, whereas others, such as trees, accumulate substantial reserves over many years before reproducing. It is possible that variation in levels of TFL1-like gene activity underlies some of these differences and it will therefore be interesting to examine how TFL1-like genes function in diverse species.

We are grateful to Mark Chadwick, Pilar Cubas, Sandra Doyle and our colleagues for useful discussions. O. J. R., C. A. V., R. C. and E. S. C. were suported by the BBSRC, I.A. by the E. C. and D. J. B. by the BBSRC and Sainsbury Laboratory.