Microsurgical and laser ablation analysis of leaf positioning and dorsoventral patterning in tomato

The shoot apical meristem of plants is the ultimate source of all aerial parts that arise after germination(Steeves and Sussex, 1989; Weigel and Jürgens,2002). The meristem is a specialized tissue, which continuously produces lateral organs such as leaves and flowers. These are arranged in a specific pattern, called phyllotaxis(Steeves and Sussex, 1989; Reinhardt and Kuhlemeier,2002). It is generally assumed that pre-existing leaf primordia influence the site of future organ formation, thus resulting in a reiterative propagation of phyllotactic patterns (reviewed by Reinhardt and Kuhlemeier,2002).

It has been proposed that the primordia are the source of a diffusible inhibitor of organ formation (Schoute,1913) (reviewed by Steeves and Sussex, 1989). According to this idea, inhibitory fields emanate from the primordia, thus allowing new primordia to be formed only at certain minimal distances from pre-existing ones. However, recent evidence has identified an inverse mechanism in which the primordia act as sinks for the organ inducer auxin rather than as sources of an inhibitor(Reinhardt et al., 2003a). The result of this scenario is similar: new organs can only be formed at certain minimal distances from pre-existing ones, thus leading to regular arrangement of leaves.

Given the proposed role of primordia in phyllotaxis, isolating a young primordium from the meristem is expected to change the position of subsequent primordia, allowing them to arise closer to the operated site. It has been attempted to experimentally interfere with leaf positioning by separating incipient primordia from the remainder of the meristem through tangential incisions (Snow and Snow,1931). When P₁ was isolated from the meristem by a tangential incision, I₁ arose at its normal position relative to P₁, but the angle between I₁ and I₂increased. This may indicate that the position of I₁ was developmentally fixed at the time of the operation, whereas the position of I₂ could still change once the influence from P₁ was eliminated. I₁ has been well characterized by its distinct pattern of gene expression. It differs from the surrounding cells of the peripheral zone in that it expresses organ marker genes, such as PINFORMED1(Vernoux et al., 2000; Reinhardt et al., 2003a), REVOLUTA (Otsuga et al.,2001), LEAFY (Weigel et al., 1992), and ZWILLE/PINHEAD(Moussian et al., 1998; Lynn et al., 1999), whereas KNOTTED1-type transcription factors, markers for meristem identity, are repressed (Jackson et al.,1994; Long and Barton,2000). This indicates that the I₁ cells are committed to organogenesis. Gene expression analysis suggests that I₂, and perhaps even incipient primordia as early as I₃ or I₄,is distinct from the surrounding cells by the expression of organ marker genes(Otsuga et al., 2001). However, the Snow experiments have shown that the position of I₂can be changed by the isolation of P₁, and therefore these cells are not determined (even though initial steps of commitment may have been taken). Surprisingly, when I₁ was isolated by similar tangential incisions, the position of I₂ was not affected, while I₃was displaced. So, if I₂ is not determined (concluded from P₁ isolation), why did it not change its position after I₁ isolation?

Taken together, the experiments of the Snows supported a negative influence of pre-existing primordia on I₁, but they did not evaluate the relative influence of the different pre-existing primordia on I₁,and they opened the question of whether P₁ affects the positioning of I₁ at all. To address these issues, we decided to reassess the experiments of the Snows using the tomato in vitro meristem culture system(Reinhardt et al., 2003b). In a first set of experiments we repeated the original experiments. Our results are in line with the old data, but they also uncover an effect on elongation growth that may require a more cautious interpretation. Therefore, we used infrared laser technology to precisely ablate incipient primordia and to reduce the experimental interference to a minimum. Finally, we isolated meristems from the influence of all primordia but P₁, in order to assess the influence of older primordia on phyllotaxis.

While the young primordia influence organ positioning in the meristem, the meristem in its turn influences the development of organ primordia after their initiation. For example, the dorsoventral patterning of the leaves, that is the formation of different upper and lower leaf tissues, depends on the activity of the meristem. This was first demonstrated by the finding that the young or incipient leaf primordia of potato developed as radially symmetric finger-like structures when they were surgically separated from the meristem(Sussex, 1951; Sussex, 1955). However, these experiments were controversial at the time(Snow and Snow, 1954a; Snow and Snow, 1954b; Sussex, 1954), and have not been repeated since, neither in potato nor in other species.

Genetic analysis has identified several putative transcription factors that are required for the specification and development of the upper (adaxial) and lower (abaxial) leaf surface (Bowman et al., 2002). Recent evidence indicates that microRNAs (miRNAs) act in the abaxial domain of the leaf primordia by silencing adaxializing transcription factors (Emery et al.,2003; Juarez et al.,2004; Kidner and Martienssen,2004). By contrast, the proposed meristem-borne signal that instructs adaxial cells to adopt their correct identity remains elusive. Here,we report on new evidence supporting a function of the meristem in the specification of the adaxial leaf domain, and we explore the role of the L₁ layer in this process.

Tomato plants (Lycopersicon esculentum cv Moneymaker) were grown as described previously (Reinhardt et al.,1998). Shoot apices were dissected and cultured according to Fleming et al. (Fleming et al.,1997) on MS medium containing 0.01 μM giberellic acid A₃ (Fluka) and 0.01 μM kinetin (Sigma).

Separation of the site of incipient primordium formation from the remainder of the meristem was carried out with small pointed scalpel blades(Bard-Parker® #11, Becton Dickinson, New Jersey, USA). Removal of the L₁ layer was carried out as described(Reinhardt et al., 2003b). Superficial ablation of the L₁ layer between the site of primodium formation and the meristem was carried out with drawn glass needles. The ultimate tip of the needle was removed, and the sharp edge of the remaining tip was used to superficially scratch the L₁ layer. Laser ablation of the site of incipient leaf formation was performed essentially as described(Reinhardt et al., 2003b). A Q-switched Er:YAG laser emitting infrared radiation at a wavelength of 2.94μm was used to direct 10 consecutive pulses (2 Hz) of 1.5 mJ per pulse at a circular area of approximately 40 μm in diameter on the surface of the meristem.

Scanning electron microscopic analysis and time-lapse photographic analysis of living tomato apices was carried out as described(Reinhardt et al., 2003b).

Classical microsurgical studies on Lupinus albus have indicated that the isolation of young or incipient leaf primordia from the meristem leads to an altered phyllotaxis in a way that is compatible with an inhibitory influence originating from primordia (Snow and Snow, 1931). However, the Snows assessed only the final outcome of the experiment 3-5 weeks after the operation, and emphasized the circumferential aspect of phyllotaxis, while the vertical growth of the apex received little attention. In addition, the question remained open as to what extent the described effects of the operations on phyllotaxis could have been influenced by wound-related effects. We therefore decided to reassess the Snows' conclusions by repeating their experiments in tomato, and by comparing them with other surgical and laser ablation techniques.

Before we describe the experiments, it is useful to define the nomenclature used in this study. We designate the youngest primordium as P₁, and the older primordia as P₂-P_n, according to increasing age. Correspondingly, I₁ designates the first incipient primordium,and I₂-I_n designates the following primordia in the succession of their appearance. In order to avoid confusion, this nomenclature was applied to the initial situation at the beginning of the experiment(t₀) and remained unchanged during the course of the experiment.

First, we repeated the experiment of the Snows. We dissected tomato apices and separated the region of presumptive leaf formation (I₁) from the remainder of the meristem shortly before primordium emergence (indicated by the fact that P₁ was well developed; Fig. 1A,B). Despite this interference, 89% of the isolated initials grew out at the expected position(16 out of 18) (Fig. 1B-D). We then determined the successive divergence angles between I₁ and subsequent primordia (I₂, I₃, etc.), each at the time of primordium emergence. This immediate determination avoids errors due to later distortions caused by the interaction with neighbouring leaves, or errors due to wound responses (e.g. callus growth).

Like I₁, the next primordium after the operation, I₂,was initiated at the expected position(Fig. 1D, Fig. 2). However, the second next primordium, I₃, was consistently formed at an increased divergence angle relative to I₂(Fig. 1E, Fig. 2). Furthermore,I₃ grew to be larger than normal in 17% of the cases (3 out of 18; Fig. 1G). Notably, when I₃ was formed, the apex had grown considerably in the vertical direction, thus leading to an increased vertical distance between I₁ and I₃ (Fig. 1H, compare with Fig. 1I). Thus, isolating I₁ from the meristem has two effects: an increased divergence angle between I₂ and I₃, and an increase in vertical growth of the apex. In addition,the rearrangement of the apex resulted in post-meristematic changes of divergence angles (compare Fig. 1B with 1D)

Incisions such as the ones performed by Snow and Snow, and those in the present study (Fig. 1), do not only interrupt signalling between primordia and the meristem, but inevitably cause major tissue damage of the meristem. It therefore seems plausible that the effects of incisions could have been influenced by wound responses. Laser ablation has successfully been employed to study interactions between cells and tissues in meristems of the root and the shoot(Van den Berg, 1995; Van den Berg, 1997; Reinhardt et al., 2003b). The precision of the laser allows ablations to be performed with a minimum of tissue destruction. To test the effect of primordium elimination, we performed laser ablations at the meristem periphery at the position of incipient leaf primordium formation (I₁). The development of two representative examples is documented in Fig. 3. In 50% of the cases (n=26), the new primordium(I₁^*) was initiated in the vicinity of the lesion,resulting in an increased divergence angle(Fig. 3A-D), or in a decreased divergence angle (data not shown) between P₁ and I₁^*. However, this deviation was transient, and the meristem rapidly re-established normal spiral phyllotaxis(Fig. 3D). In 12% of cases,I₁^* formed at an ectopic position between the ablated I₁ position and I₂(Fig. 3F). This led to the reversal of the phyllotactic spiral (Fig. 3G,H). In the remaining cases (38%), the primordium was initiated below the lesion, with no significant deviation in divergence angle, and with no recognizable consequences for further phyllotaxis. Notably, vertical elongation, as seen after incision into the meristem(Fig. 1), was not observed after laser ablation.

In conclusion, after laser ablation of the I₁ position, the next primordium was formed either in close vicinity of the lesion, or at an ectopic position (but not at I₂). Therefore, precise removal of incipient primordia by laser ablation yielded results that were different from those obtained by tangential incisions (Fig. 1) (Snow and Snow,1931). Conceivably, laser ablation revealed a flexibility of the meristem that was unnoticed after surgical incisions, because the fragmentation of the peripheral zone (PZ), and the resulting indirect growth response, influenced the effects on phyllotaxis in specific ways, and restricted the potential of the meristem to respond with ectopic organogenesis. The tendency of I₁^* to be formed in proximity to the original I₁ position indicates that the mechanism of leaf positioning remained active after laser ablation at I₁.

An important observation after isolation of I₁, both in this study as well as in the study of the Snows, was the fact that I₂did not respond to the isolation of I₁. Could the position of I₂ have already been fixed? The fact that I₂ was displaced when P₁ was isolated(Snow and Snow, 1931) does not support this possibility. If the isolation of I₁ has no effect on the position of I₂ (although the latter is not fixed), does this mean that under natural conditions, primordia are not influenced by their next older predecessor? If this were the case, which would be the primordia that determine leaf position?

Two principal scenarios could be envisaged: A new primordium(I₁) could be influenced by the two previous primordia(P₁ and P₂), with P₁ having the stronger effect so that I₁ comes to lay closer to P₂. Alternatively, the position could be determined by the two immediate neighbours, which would be P₂ and P₃ in the case of tomato. In very large meristems, as in the case of the sunflower capitulum, it is likely that the latter mechanisms applies, as the distance between P₁ and I₁ is very large compared with the distance of P₁ to its immediate contact neighbours. By contrast, in small meristems, such as in tomato and most other plants, the meristem appears small enough to permit an influence of P₁ on I₁.

To distinguish between these two possibilities, we isolated tomato meristems from the apex in a way that left only one primordium (P₁)attached to them. Although this operation inevitably creates a large wound, it differs in two important ways from the tangential incisions carried out by the Snows and in this study. First, the meristem as such is not damaged, as the cut was made just below the peripheral zone, and secondly, any wound effect would act equally on the entire circumference of the meristem. Hence, effects like a shift of the growth axis or distortion of divergence angles are not expected to occur. Isolated meristems were cultured on MS medium and further organogenesis was observed. Eighteen of 64 isolated meristems (28%) developed in culture and continued to form leaf primordia. The first new primordium(I₁) was formed at the expected position(Fig. 4B, compare with 4A). However, in many cases the primordia were oversized (n=10; Fig. 4B,C). This was true particularly for P₁, which in five cases grew approximately to double width with two tips (Fig. 4D). Such wide and fused primordia resemble the primordia induced by the ectopic application of IAA to tomato meristems(Reinhardt et al., 2000). The next primordium (I₂) was formed approximately at the expected position, thus the direction of phyllotaxis was not reversed(Fig. 4C). However, its divergence angle was more variable, conceivably as a consequence of the changes in size of P₁ and I₁. Taken together, these results indicate that P₁ and I₁ are sufficient to determine the approximate region of I₂ (and therefore to propagate the generative spiral), whereas the older primordia(P₂-P₄), which are the direct neighbours of P₁ and I₁ (Fig. 4E), are necessary to determine the boundaries of P₁and I₁, and thereby their exact width and position. The large increase in the width of P₁ demonstrates that it was able to recruit excess cells, even after the onset of outgrowth, if its contact neighbours P₃ and P₄ were absent.

After its initiation, a leaf is subject to dorsoventral patterning, which leads to the formation of a specialized upper (adaxial) and lower (abaxial)leaf surface. In potato, surgical isolation of young or incipient primordia from the meristem leads to the formation of radialized primordia that lack a leaf blade (Sussex, 1955). This indicates that a signal from the meristem is required for dorsoventral patterning of the primordia, and for the formation of the leaf lamina. Although the initial observations were at first controversial(Snow and Snow, 1954a; Snow and Snow, 1954b; Sussex, 1954), the concept of a meristem-borne signal that confers adaxial identity to the portion of the primordium that is adjacent to the meristem became widely accepted(Steeves and Sussex, 1989; Bowman et al., 2002). Tomato has compound leaves, i.e. the leaves consist of a central rachis with a terminal leaf blade, and two rows of lateral leaflets(Sinha, 1999). It was therefore of interest to see whether compound leaf architecture also depends on signals from the meristem. Overexpression of the meristem gene Tkn1 in leaves leads to supercompound leaf architecture(Hareven et al., 1996);however, this is probably due less to the direct regulation of leaflet initiation than to the spatial and temporal extension of a state of meristematic competence that allows repetitive leaflet formation.

We surgically separated, by tangential incisions, young leaf primordia (or the site of incipient leaf formation) from the meristem(Fig. 1A), and followed their development. However, before reporting on the effect of this operation, it is useful to give a detailed description of normal leaf development in tomato(Fig. 5A). The first obvious sign of dorsoventral patterning of tomato leaf primordia is their curvature towards the centre of the meristem (e.g. P₁ in Fig. 1A). Later, the primordia form characteristic trichomes on the adaxial (towards the stem, upper side of the mature leaf) and abaxial (away from the stem, lower side of the mature leaf) surface of the primordium. The abaxial side exhibits three types of trichomes: long linear, short linear, and few globular trichomes, which are distributed relatively sparsely (Fig. 1I, Fig. 5A). By contrast, the adaxial side has only globular trichomes, which are concentrated in a dense stripe along the central axis of the primordium(Fig. 5A). Thus, adaxial and abaxial identity can clearly be distinguished by their trichomes. After the onset of trichome formation, the leaf primordia initiate lateral leaflets in a basipetal fashion, i.e. new pairs of leaflets are successively formed at the base of the primordium (Fig. 1I, Fig. 5A)(Sinha, 1999). The leaflets grow towards the adaxial side of the primordium, i.e. they point to the meristem (Fig. 1I, Fig. 5A).

When incipient tomato leaf primordia were separated from the meristem, 15 out of 23 (65%) exhibited a partial or complete loss of lateral leaflets (Fig. 5C-F, compare with 5A,B). Sometimes only one central fused leaflet was formed on the adaxial side of the primordium(Fig. 5C), while the distal portion of the leaf developed normally. In 7 out of 23 cases (30%), the primordium developed apparently normal, and one apex died (4%) (data not shown). These results are in line with earlier reports that a signal from the meristem is required for normal dorsoventral patterning(Sussex, 1951; Sussex, 1955). Cases in which the loss of dorsoventral patterning was incomplete, i.e. a small dorsoventral portion of the primordium was retained at the distal end, suggest that the adaxial domain was lost (Fig. 5D). This notion was confirmed by the morphology and distribution of the trichomes on radialized primordia. They were primarily long and linear,with fewer interspersed short linear and globular trichomes, like on the abaxial side of normal primordia (compare Fig. 5F with 5A). Thus, separation from the meristem has the same consequence as do loss-of-function mutations in the Antirrhinum MYB transcription factor PHANTASTICA(Waites and Hudson, 1995; Waites et al., 1998), and as does the overexpression of genes of the YABBY and KANADIfamilies in Arabidopsis (Bowman et al., 2002). In all of these cases, adaxial tissues were converted into abaxial tissues.

The defects in dorsoventral patterning could potentially be caused by the wound itself, rather than by the isolation from a specific patterning signal. To control for wound effects, we performed incisions of a similar extent through the centre of the meristem (Fig. 5G; n=12). Such meristems continued to develop and to form leaf primordia (Fig. 5H-J). The primordia exhibited normal dorsoventral patterns, even if they were initiated next to the lesion(Fig. 5J). This was evident by their characteristic curvature towards the meristem(Fig. 5J), and later by the formation of lateral leaflets and normal trichomes (data not shown). As a result of the incision, the two halves reorganized into two new meristems. The result of this control experiment shows that wounding is not sufficient to induce radialized primordia, indicating that the loss of dorsoventral pattern in surgically isolated primordia is a consequence of the lack of contact with the meristem, and is not due to wounding. Thus, our results are in agreement with those of Sussex (Sussex,1951; Sussex,1955), and similarly, we conclude that dorsoventral patterning is dependent on a signal from the meristem. Taken together, separation of primordia from the meristem caused two defects: the loss of lateral leaflets and the loss of dorsoventral patterning.

The meristem consists of different subdomains, the central zone (CZ) and the peripheral zone (PZ), and it can be subdivided in three clonally isolated cell layers, L₁, L₂ and L₃(Steeves and Sussex, 1989). Therefore, a certain region of the meristem could potentially play a principal role in the production or transmission of the adaxializing signal. Alternatively, the meristem as a whole could produce and release the signal. Laser ablations of the central zone (CZ) did not affect dorsoventral patterning in a measurable way (Reinhardt et al., 2003b), therefore, the CZ cannot be the exclusive source of the adaxializing signal.

The L₁ layer, which is clonally separated from the subtending cell layers by its stereotypical, anticlinal cell division pattern, plays important roles in meristem function and organ development (Baroux, 2001; Abe et al., 2003; Reinhardt,2003a; Reinhardt, 2003b). The cells that give rise to the meristem L₁ layer (and the entire epidermis) are set aside early during embryogenesis at the dermatogen stage(Jürgens, 2003). These cells attain a specific identity that is characterized by the expression of L₁-specific genes, such as AtMERISTEM LAYER1 (AtML1), FIDDLEHEAD(FDH), PROTODERMAL FACTOR1 (PDF1) and PDF2,(Lu et al., 1996; Yephremov et al., 1999; Abe et al., 2001; Abe et al., 2003). We wanted to assess the role of the L₁ layer in dorsoventral patterning of primordia.

The L₁ layer can be surgically removed from the meristem(Reinhardt et al., 2003b). Removal of the entire L₁ layer results in a gradual degeneration of basic meristem functions, and in the inhibition of organ formation(Reinhardt et al., 2003b). In order to assess the role of the meristem L₁ layer in dorsoventral patterning of adjacent leaf primordia, we removed the L₁ layer to various extents, while leaving the youngest primordium (P₁) intact. We then followed in detail the consequences for the development of P₁, and for further primordium formation(Table 1; Fig. 6). Removal of up to 50%of the L₁ layer had only local effects. P₁ developed normally and leaf formation continued, but only from the area with an intact L₁ layer. The meristem rapidly shifted its centre away from the wound, and continued to form normal leaf primordia at a normal rate(Reinhardt et al., 2003b). Owing to the ablation and to the shift of the growth centre, the phyllotactic angles were sometimes irregular (data not shown). In only one case did a primordium exhibit a dorsoventral patterning defect after removal of less than 25% of the L₁ surface (Table 1). When 50-75% of the L₁ layer was removed, 36% of the apices (8 out of 22) terminated after the formation of one new primordium(I₁). In half of these cases (4 out of 8), this primordium was radially symmetric (Fig. 6B; Table 1), while P₁developed normally. Removal of 75-100% of the L₁ layer resulted in termination of almost all meristems (26 out of 27; Table 1). However, 38% of these apices (10 out of 26) formed a last primordium (I₁) before termination, which was, in 50% of the cases, radially symmetric (5 out of 10). Of the P₁ primordia, only 11% (3 out of 27) developed as partially or entirely radialized primordia (Fig. 6C-E; Table 1),whereas the others developed normally (Fig. 6F).

These results show that even preformed leaf primordia can loose their initial dorsoventral pattern once the L₁ layer is lost. However,even small portions of the L₁ layer appear to be sufficient to promote dorsoventral patterning, even in situations where the meristem aborts(Fig. 6F). As in the case of surgical separation from the meristem (Fig. 5C,D), primordia of apices with L₁ ablations were frequently radialized only in the proximal parts (including the lack of leaflets), whereas the distal portion exhibited a normal dorsoventral pattern with the initiation of the leaf blade (Fig. 6B,D,E). Primordia that were radialized exhibited abaxial identity, as determined by the morphology and distribution of their trichomes(Fig. 6C,D compare with Fig. 5A).

It could be argued that the loss of dorsoventrality after L₁ablations was simply due to the degeneration of the meristem, and that it does not suggest a specific role of the L₁ layer. However, the similar range of patterning defects (compare Figs 5, 6), as well as the comparable frequences of dorsoventral defects between incised and L₁-ablated meristems (Table 1), indicates that the L₁ ablations had an effect equivalent to the surgical separation, which is likely to be effective immediately. By contrast, the degeneration of the meristem after L₁ ablations proceeded slowly over several days (Reinhardt et al.,2003b). These results are compatible with a direct role of the L₁ layer in the establishment and maintenance of the dorsoventral pattern in leaves. For example, a signal for dorsoventral patterning could be produced and transported to the developing primordia through the L₁layer.

To test this possibility more directly, we performed narrow, superficial ablations of the L₁ layer that left the majority of the L₁ layer intact, but separated the L₁ layer of I₁ or P₁ from the L₁ layer of the remainder of the meristem by a small corridor (further referred to as corridor ablations; Fig. 7A,B). This operation was expected to have no consequences for meristem function and maintenance, but it would interrupt the adaxializing signal, if it were transported through the L₁ layer. We operated 30 apices shortly before (I₁), and 10 shortly after, primordium initiation(P₁). Four apices operated at P₁ (40%) showed varying degrees of dorsoventral patterning defects(Fig. 7C-F). Of the apices operated at I₁, 13% showed partial defects in dorsoventral pattern(similar to Fig. 6D,E), and one was completely radialized (Fig. 7G). Because, unexpectedly, the percentage of effects was lower when the corridor ablation was performed at the I₁ stage than at the P₁ stage, we repeated the experiment with 40 apices from which 20 were operated at the I₁ stage and 20 at the P₁ stage. To assess the extent of the defects more precisely, we distinguished between partially radialized primordia, which were dorsoventral in their distal portion and radialized in their proximal portion, and fully radialized primordia, which failed to exhibit any sign of dorsoventral patterning and had only abaxial trichomes (Table 2; Fig. 7). Again,primordia operated at the I₁ stage showed defects at a lower rate and of weaker severity (35% partially, 5% completely radialized; Fig. 7H), compared with primordia operated at the P₁ stage, where 40% were fully, and a further 15% partially, radialized (Table 2; Fig. 7). Interestingly, more than half of the apices operated at I₁initiated an accessory meristem above the primordium (11 of 20; Fig. 7I). In such cases, the primordium was frequently normal (9 cases), and in only two cases was it partially radialized. Such an accessory meristem was only formed once in apices operated at the P₁ stage. It thus appears likely that the accessory meristems formed after operations at I₁ provided the dorsoventral signal, thereby lowering the frequency of radialized primordia. In conclusion, the fact that corridor ablations provoked similar defects at comparable frequencies to deep tangential incisions or complete L₁ablations demonstrates that the continuity of the L₁ layer is essential for the establishment of the adaxial domain of leaf primordia.

Early studies on Lupinus albus showed that microsurgical separation of incipient leaf primordia (I₁) from the meristem influenced the positioning of new primordia(Snow and Snow, 1931). Whereas the next primordium (I₂) was formed at the expected position with a divergence angle of approximately 137°, the following primordium(I₃) was `displaced' in the circumferential dimension towards the isolated I₁, resulting in an increased divergence angle between I₂ and I₃. This led to the interpretation that the isolation of I₁ created more space in its vicinity, which could then be occupied by the adjacent I₃ primordium.

One general limitation of the Snows' study was that they analyzed only the final outcome of the experiments in transverse sections of the shoot prepared 3-5 weeks after the operation. Hence, the divergence angles were measured several plastochrons after the initiation of the respective primordia. Therefore, some of the reported effects on phyllotaxis may have been influenced by wound reactions. We have detected, within 4 days after the operation, shifts of the meristem centre away from the wound, and this response clearly affected apparent divergence angles post-meristematically(Fig. 1B,D). To record the changes in divergence angles with confidence, we followed in our experiments the divergence angles starting immediately after the operation, thus,secondary distortions of the divergence angles can be ruled out.

Concerning the effects on divergence angles, we obtained results that are in agreement with the findings of the Snows. However, in addition to changes in divergence angles, we found major changes in the vertical growth(Fig. 1 and Fig. 2). The Snows analyzed the result of incisions at a time at which even I₅ and I₆had developed into leaves with a well-developed leaf blade(Snow and Snow, 1931). Hence,initial changes in vertical growth, such as the ones reported here, may well have occurred in Lupinus albus as well, but could have been obscured at the time of analysis.

Upon isolation of I₁, both in the experiments of the Snows and in ours, the first angle was normal and only the second primordium was displaced. One possible explanation is that the position of I₂ is already fixed. This, however, is ruled out by the P₁ isolations,which clearly demonstrate that I₂ is not fixed, because it was displaced in this case (Snow and Snow,1931). Another possibility is that the inflexibility of I₂ after I₁ isolation is related to wound effects that influence the response of the meristem. The incision causes major damage to the peripheral zone and thus restricts the space available for a new primordium. In contrast to the tangential incisions, the laser ablations deleted I₁ with a minimum of tissue damage to the meristem. Importantly, increased vertical growth was not observed. Thus, laser ablation did not affect the geometry and organisation of the meristem, and it allowed us to assess, more directly, the question of how spiral phyllotaxis responds to the specific elimination of a primordium.

After laser ablation of I₁ the next primordium(I₁^*) was formed ectopically, either in direct proximity of the lesion (50%) or at a new position between the lesion and I₂(Fig. 3). Hence, under these conditions, the position of the first primordium forming after the lesion was not fixed but displayed remarkable flexibility. Such changes in organ position are in line with theoretical predictions that elimination of a primordium should affect further organ formation immediately after ablation. Furthermore,the tendency of I₁^* to be formed at a position close to the original I₁ position indicates that I₁ normally suppresses organogenesis in its vicinity, similar to P₁(Reinhardt et al., 2000). After initial deviations caused by I₁ ablations, phyllotactic patterning rapidly returned to normal spiral phyllotaxis, either in the original, or in the reverse direction. This demonstrates the strong tendency to revert to spiral phyllotaxis in the case of disturbance, and underscores the stability of the spiral pattern, once the disturbance is overcome.

Theoretical consideration, as well as experimental evidence, suggests that the positioning of new primordia is influenced by pre-existing primordia (see above). However, it is not known to what extent the different primordia contribute to this mechanism. It seems likely that the influence of a primordium decreases with its distance to the site of incipient organ formation (I₁). In a large meristem, such as the capitulum of the sunflower, where organ formation proceeds from the edge towards the centre,the closest neighbours of a new primordium are not the predecessors in the generative spiral (P₁ and P₂), but the direct contact neighbours. If the capitulum exhibits a 34:55 phyllotactic system, the direct neighbours are P₃₄ and P₅₅, which are 34 and 55 plastochrons older than I₁, respectively. By comparison,P₁ and P₂ of such a system are positioned much more remotely. Therefore, it is likely that I₁ is positioned according to P₃₄ and P₅₅, rather than according to P₁and P₂. However, in a phyllotactic system, as in tomato and many other plants, the meristem is small enough that not only the direct neighbours(in tomato which exhibits a 2:3 phyllotaxis, this is P₂ and P₃), but also P₁ could influence the positioning of a new organ.

In the experiments of the Snows, the first primordium to arise after operations was not affected. Does this mean that the youngest primordium has no effect on the positioning of the following one? We have previously proposed that the two youngest primordia influence the positioning of the following primordium, with P₁ having the stronger effect, so that I₁ comes to lay closer to P₂ than P₁, thus resulting in the characteristic divergence angle of 137°(Reinhardt and Kuhlemeier,2002). In agreement with this idea, the youngest primordia express auxin transport proteins in a way that suggests a strong sink function for auxin (Reinhardt et al.,2003a). To distinguish between the influence of the direct predecessors and that of the direct contact neighbours on organ positioning,we isolated meristems from all influence but that of P₁. The assumption was that I₁, whose position is likely to be determined,would be formed at its normal position. However, I₂ could potentially respond to the experimental isolation with a shift. If it was positioned ectopically, this would indicate that the direct neighbours(P₁ and P₂) are important for organ positioning. If however the positioning of I₂ were not affected, then the interpretation would be that the youngest primordia (I₁ and P₁) are sufficient for leaf positioning. Taken together, our results suggest a compromise of the two possibilities. The generative spiral did not reverse in isolated meristems, indicating that I₂ was always formed in the correct direction. This finding is in accordance with surgical experiments in which isolated meristems of Primula formed leaves in continuity with the original phyllotactic pattern(Wardlaw, 1950), whereas similar isolations in Lupinus interrupted the original spiral, and led to the establishment of a new spiral system(Ball, 1952).

After isolation of the meristem, P₁ and I₁ grew considerably wider than normal (Fig. 4), indicating that in the absence of older primordia, the primordia could recruit more cells than normal. Hence, the following picture emerges for phyllotaxis under natural conditions. Because I₁appears to be fixed, it is at the I₂ stage that organ position becomes determined. We propose that I₁ and P₁ determine the approximate location of I₂, thereby dictating the direction of the generative spiral. After the approximate positioning in the meristem, the direct neighbours (P₁ and P₂) delimit its exact boundaries, hence determining its final size and its precise radial position.

The fact that P₁ grew wider in the absence of its contact neighbours indicates that lateral restriction (and fine positioning) is a prolonged process that continues after the initiation of a primordium. Although it may seem counterintuitive at first, the conclusion therefore is that organ positioning and organ outgrowth occur concomitantly, and not sequentially. Such a mechanism would allow for the feedback mechanism that is predicted to operate in phyllotaxis and other models of pattern generation in living organisms (Meinhardt, 1994).

The surgical separation of tomato leaf primordia from the meristem caused two defects. The loss of lateral leaflets and the loss of dorsoventral patterning. This raises the question whether the two processes are linked. For example, it is conceivable that leaflets, like the leaf lamina, can be formed only where the adaxial (dorsal) and the abaxial (ventral) domain of the primordium are juxtaposed. This view is compatible with the occurrence of single leaflets in a central position on the adaxial side of the rachis,instead of in a lateral position (Fig. 5C, Fig. 6E). A similar case is represented by the variably sized distal patches of adaxial identity found on primordia of phantastica mutants in Antirrhinum. In this case, the lamina of the distal leaf portion is joined on the adaxial side of the primordium, presumably marking the course of the adaxial/abaxial boundary (Waites and Hudson, 1995). This position corresponds to the position of the single central leaflets (Fig. 5C, Fig. 6E). It is therefore likely that in such cases, the central leaflet marks the boundary between the extended abaxial domain, and the reduced adaxial domain in the distal portion of the primordium. Taken together, these data indicate that lateral leaflet formation, like lamina outgrowth, occurs only at the boundary between adaxial and abaxial domains.

Dorsoventral patterning of leaves is thought to be influenced by the meristem (Bowman et al., 2002). Evidence for this notion came from early microsurgical analysis in potato(Sussex, 1951; Sussex, 1955). These experiments were challenged in the following years, and had not been repeated in other plant species since. We confirm in the present study, using three different microsurgical techniques (vertical incision, complete L₁ablation, corridor ablation), that the meristem provides information that is required for the dorsoventral patterning of the primordia (Figs 5, 6, 7). It has been proposed that a factor from the meristem induces adaxial identity in the upper part of leaf primordia, while the lower part adopts abaxial fate by default(Bowman et al., 2002). However,despite the identification of a number of putative transcription factors,which are required for the establishment of adaxial and abaxial identities(Bowman et al., 2002), the nature of the adaxializing signal remains elusive. Recently, miRNAs have been implicated in the control of abaxial identity. These are expressed just below incipient primordia (i.e. on their abaxial side), and in the abaxial side of the leaves. There, they determine abaxial cell fate by downregulating the levels of adaxializing proteins, such as PHABULOSA and ROLLED LEAF1 on the abaxial side of the primordia (Kidner and Martienssen, 2004; Juarez et al., 2004). Thus, miRNA may represent an abaxializing signal that,in concert with the adaxializing signal from the meristem, establishes dorsoventral polarity.

We have previously shown that surgical removal of the L₁ layer from the meristem leads to a progressive degeneration of the meristem(Reinhardt et al., 2003b). However, removal of L₁ also abolished dorsoventral polarity of the last one or two primordia before the meristem arrested(Fig. 6). This radialization was observed at a similar extent and frequency to in the case of surgical separation from the meristem (50% versus 65%; compare with Fig. 5), indicating that removal of L₁ is equivalent to an immediate interruption of the adaxializing signal. This is in contrast to meristem degeneration, which proceeded slowly over several days(Reinhardt et al., 2003b). Hence, the loss of dorsoventrality after L₁ ablation preceded the loss of meristem identity in the remaining tissue, rendering it improbable that the loss of dorsoventrality is due, indirectly, to the degeneration of the meristem. This evidence suggests a special role for the L₁layer in the determination of dorsoventral polarity. For example, the adaxializing signal could be transported through the L₁ layer to the young primordia to induce adaxial fate in the adjacent portion of the primordium. Indeed, our data show that the continuity of the L₁layer between the meristem and the site of primordium formation is relevant for dorsoventral patterning, as corridor ablation caused defects of a similar extent and at similar rates to the vertical incisions. Nevertheless, it remains to be clarified whether the L₂ and L₃ layer also contribute, directly or indirectly, to dorsoventral patterning.

It is noteworthy that in all experiments only a minority of the primordia were completely radialized. In most cases, the young primordia had a dorsoventral distal portion, and a proximal radialized portion of different extent. Often, this partial loss of dorsoventral pattern was acccompanied by a partial or complete loss of lateral leaflets. In tomato, the leaf primordia develop in a basipetal fashion, i.e. new leaflets are successively formed at the leaf base (Sinha, 1999). Therefore, the partially radialized primordia consist of an older distal portion with normal dorsoventrality, and a younger proximal portion that has lost its dorsoventral pattern during development. The range of dorsoventral defects observed included all possible intermediates between completely normal and completely radialized. This indicates that the adaxializing signal needs to be present during an extended period of leaf development, and that a loss of the signal during this process can abolish the dorsoventral pattern in the proximal parts at various stages of development. Thus, the establishment of dorsoventral polarity appears to be a continuous process. Such a scenario is compatible with the genetic models for dorsoventral polarization, which envisage a self-reinforcing mechanism based on the mutual inhibition of adaxial and abaxial determinants that leads to the gradual separation of the domains with adaxial and abaxial identity(Bowman et al., 2002). Furthermore, the frequent occurrence of partially radialized primordia with a normal distal portion shows that radializations are not due to the destruction of the (predetermined) adaxial domain of the I₁ position. Partially radialized primordia always exhibit a normal distal portion. As this is the part of the leaf that is formed first, the primordia must have started with an intact adaxial domain, which later lost its adaxial identity during the course of leaf development.

This work and our previous study(Reinhardt et al., 2003b)demonstrate that microsurgical techniques continue to be useful tools for studying plant development. They complement genetic analyses, and they are particularly useful in cases where it is desirable to restrict functional interference tightly in space and time. Now the challenge will be to develop new genetically based tools to elucidate the mechanisms underlying phyllotaxis and dorsoventral patterning in more detail. Such tools will be developed when we know more about the nature of the components in the signal chains. For example, once we know the nature of the adaxializing signal, its production,transport or destruction could be influenced in a tissue-specific manner. Similarly, the tissue-specific expression in subdomains of the meristem of genes involved in auxin biosynthesis, metabolism, transport and perception will allow us to rigorously test models of phyllotaxis.

This work was supported in part by grant 3100-055540 of the Swiss National Science Foundation.