Adhesion between growing cells
Intercellular adhesion in plants arises in a different way from adhesion between animal cells. With exceptions such as pollen on the stigmatic surface (Zinkl & Preuss 2000), most plant cells do not come together and stick: they are formed in an adherent state and remain so throughout their lifetime. After meristematic cells divide, the growth of the daughter cells is a concerted process in which neighbouring cells expand in a co-ordinated manner and adherent walls remain fused along the line of the middle lamella as they expand (Knox 1992). A schematic view of cell plate formation, plant cell adhesion and its loss is shown in Fig. 1.
Figure 1. Schematic diagrams showing plant cell cytokinesis and intercellular space formation. A membrane-bound cell plate grows from the centre of a cell to fuse with the plasma membrane/primary cell wall to produce two daughter cells. As primary cell wall is deposited on either side of the cell plate it becomes a middle lamella and region of intercellular attachment. For a intercellular space to form a region of the primary wall of the parent cell must be degraded allowing the new middle lamella to link up with the older middle lamella of the parent cell. This space may be filled with material or open up to form an intercellular air space (a distinctive feature of parenchyma tissues).
Download figure to PowerPoint
It is not clear how the extension of the two adherent cell walls is co-ordinated. A simple biophysical feedback loop may be all that is required: if one of the two walls expands less it will then carry more of the load and be further above the threshold stress for expansion, provided that the middle lamella resists shear. That is not to say that no mechanism exists for the co-ordinated development of walls of adjacent cells. The absence of shear at the middle lamella is necessary to prevent rupture of plasmodesmatal connections between the adjacent cells during growth. The textured wall structure and patterns of pectic epitopes observed around plasmodesmata on both plasma-membrane faces of adherent tomato pericarp cell walls (Orfila & Knox 2000) may reflect some, as yet unknown, mechanism of the co-ordination of cell wall extension in the region of pit fields.
Isotropic expansion of tissues is the exception rather than the rule. Any anisotropy in growth means that two walls of a single cell must extend at different rates, even though each wall is co-ordinated in its extension with the adherent wall of an adjacent cell. It may be easier to think of the directional expansion of plant tissues as being quantized not into single cells, but into pairs of adherent cell walls.
Do plant cells always adhere to the same cells?
It might be thought that intrusive growth is an exception to the rule that plant cells do not slide past one another. Examples of intrusive growth are the penetration of the pollen tube through stylar tissue (Jauh & Lord 1996), the differentiation of tracheids (Kalev & Aloni 1998) and laticifers (Serpe, Muir & Keidel 2001) and the elongation of flax fibre cells into the cortical parenchyma (Roland et al. 199). However, the intrusive cell expands between the invaded cells by tip growth (Yang 1998; Ryan et al. 1998), so it is only the intruding apex that actually moves, separating the walls of the cells on either side. Along all the rest of its length the intrusive cell remains firmly attached to its new neighbours. At the growing tip therefore the original cell–cell adhesion is dismantled and a new cell–cell bond is established. Although intrusive growth is not a case of whole cell movement it is a case of the formation of new cell contacts that are not formed at cytokinesis.
A further exception to the rule that plant cells remain in contact with the same cells throughout development is the behaviour of pollen on the stigma surface. Here adhesion occurs de novo. It can take different forms. In species such as Arabidopsis with dry stigmas, the initial adhesion event is rapid (seconds), species-specific and mediated by lipophilic molecules so that it can function in the absence of free water (Zinkl et al. 1999). Pollen hydration and alterations to the underlying stigmatic cells then follow. Compatible pollen tubes growing down the hollow style of lily flowers adhere to its inner surface through the co-operative action of a number of water-soluble macromolecules including a pectic fraction and a small lipid transfer protein (Park et al. 2000; Mollet et al. 2000)
Why do epidermal cells not adhere on contact?
When two leaves or stems of the same plant come into chance contact they do not fuse, but remain distinct. Likewise pollen tubes that germinate on an epidermal surface other than the stigma do not interact with it. However during flower formation in many species, fusion of floral organs such as carpels is part of the normal process of development. It also seems that some form of adhesion of parenchymatous tissues is an early event in the formation of a graft union between cut faces of two compatible plants. This stage of the grafting process has been little studied because the key events determining compatibility occur later when vascular continuity is becoming established (Jeffree & Yeoman 1983). Thus the ability of a plant to delineate its own surface, the boundary between itself and non-self, seems to reside in the epidermis but to be capable of being inactivated under developmental control in special circumstances such as the fusion of floral parts. Some understanding of the molecular articulation of these processes has recently emerged from experiments on mutant plants with defects in cuticle formation.
In certain of these mutants, plant surfaces fuse that would not be expected to do so (ectopic organ fusion) and pollen can become hydrated on these surfaces when normally it hydrates only on the receptive stigma surface (ectopic pollen hydration). This fusion of normally separate juvenile organs is a feature of the eceriferum (cer) mutants in Arabidopsis and the glossy mutants in maize, which were originally identified from defects in cuticular wax formation (Post-Beittenmiller 1998). The FIDDLEHEAD gene in Arabidopsis, identified originally from ectopic organ fusion and ectopic pollen hydration in the mutant fdh1 phenotype, encodes a probable β-ketoacyl CoA synthase involved in chain lengthening of fatty acids, a requirement for cuticular lipid synthesis (Yephremov et al. 1999; Pruitt et al. 2000). Altered cuticle formation was likewise found in the adherent1 mutant in maize, previously recognized by postgenital organ fusion and pollen hydration like fdh1 (Sinha & Lynch 1998).
With some exceptions, the mutated genes encode enzymes involved in long-chain lipid synthesis (Xu et al. 1997; Fiebig et al. 2000, Rashotte, Jenks & Feldmann 2001). CER2 encodes a novel, possibly regulatory protein (Xia, Nikolau & Schnable 1997), although the cer2 phenotype shows aberrant cuticular wax composition as in other cer mutants (Negruk et al. 1996). The Arabidopsis ALE1 gene, necessary both for normal cuticle formation in embryos and to prevent organ fusion, encodes a probable serine protease (Tanaka et al. 2001). It might be suggested that ectopic organ fusion is normally prevented by some non-cuticular function of long-chain lipids, for example in signalling. This is unlikely in view of the observation that transgenic Arabidopsis plants expressing a fungal cutinase (Sieber et al. 2000) showed strong postgenital organ fusion despite having no abnormality in the synthesis of lipid monomers.
How does the cuticle maintain the distinctness of vegetative tissues in contact with one another? It is not simply a matter of preventing water movement (Kerstiens 1996). Transmission of molecular signals across the divide is certainly involved in both pollen hydration and the normal fusion of floral organs (Lolle et al. 1997). Only at a later stage in floral organ fusion are symplastic connections formed (). It has been suggested that small polar molecules are transmitted across the epidermal walls to signal contact between floral organs that will fuse (Siegel & Verbeke 1989), and that the permeability of the cuticle may be tuned to control their passage (Lolle et al. 1997). Whatever the signal molecules whose passage from cell to cell is prevented by the cuticle, its presence seems to be the key factor that prevents adhesion between plant cells when none is programmed to occur.
In addition to the phenomenon of organ fusion in lipid elongation mutants, some of them show other developmental defects. One such mutant, fdh1 in Arabidopsis, is affected in trichome differentiation (Yephremov et al. 1999), whereas hic mutants in Arabidopsis (Gray et al. 2000) and a range of cer mutants in barley (Post-Beittenmiller 1998) are altered in stomatal abundance. These observations prompted suggestions that the transport of signal molecules involved in differentiation is influenced by cuticular composition. The HIC gene (Gray et al. 2000) is particularly interesting because its product may form part of a signal cascade sensing ambient CO2 concentrations and controlling stomatal abundance in response (Retallack 2001). Since the cuticle is also a CO2 barrier, a reasonable hypothesis might be that a shared CO2 sensing mechanism forms part of a feedback loop controlling the development of the cuticle itself.
Biomechanics of intercellular adhesion
Although adhesion between plant cells restricts the capacity for individual cell movement, cell adhesion makes it possible for plants to grow into structures of greater height and mechanical robustness than anything in the animal kingdom. Plant tissues become mechanically strong by two strategies, both of which are fundamentally dependent on adhesion between cells (Jarvis & McCann 2000).
One strategy is typical of wood, in which compressive as well as tensile stresses are carried by the cell walls. Because of scaling factors this strategy is the only one possible in a tree, but it requires a much greater investment of material in thick, rigid secondary cell walls that can withstand compressive buckling. Bending stresses on a tree are also translated into shear stresses between adjacent files of cells, and these are resisted by lignification of the pectic middle lamella (Hafren, Daniel & Westermark 2000). Wood under heavy bending loads frequently splits at this point (Thuvander & Berglund 2000). However, in this review we shall not concern ourselves further with intercellular adhesion in wood.
In the other strategy, typical of soft plant tissues in rapid growth, turgor pressure carries all compressive stresses on the tissue and the primary cell walls are kept permanently in tension. This makes it possible for the cell walls to be thin and flexible, allowing economy in structural material. However, because a sphere is the shape of lowest energy for a flexible-walled, pressurized cell, turgor generates secondary stresses that tend to tear the cell away from its neighbours at each corner, pulling it towards a spherical shape. The magnitude of these cell separation stresses is comparable with the tensile stress within each cell wall, and to withstand them considerable intercellular adhesion strength is necessary at the cell corners (Jarvis 1998).
It is common to think of the middle lamella as a line of adhesive gluing the cells together, and certainly the question of intercellular adhesion is linked to the structure of the middle lamella. However the idea that the middle lamella glues two cells together is misleading in more than one respect. We use glue to stick together surfaces that were initially apart, but two plant cells joined along the middle lamella have never been apart. Furthermore, as explained above, the stresses that tend to separate cells are not distributed evenly over the cell surface but are concentrated at the cell corners (tricellular junctions) and at the corners of intercellular spaces (Jarvis 1998). Precisely at these points there are reinforcing zones which can be distinguished under the electron microscope and which, as will be seen later, differ in polymer composition from both the primary cell walls and the middle lamella. It is these reinforcing zones that carry the turgor-imposed stress and are the first line of defence against cell separation (Parker et al. 2001). We must focus on their mechanical properties if we wish to understand intercellular adhesion and separation from a biomechanical point of view. It follows that the bulk composition of the cell wall will normally tell us little about the strength of intercellular adhesion.
Ontogeny of cell junctions in plants
In general terms, one of two possible histories can be ascribed to an adherent pair of cell walls occurring in a plant. These contrasting origins are exemplified in epidermal and cortical cells where anticlinal cell walls (at right angles to the plant surface) originate in cell division and the periclinal cell walls (parallel to the plant surface) originate largely from cell wall assembly during cell expansion. When the cell plate is formed within a dividing cell, it expands radially by the accretion of material at the edges until it reaches the existing cell walls (Heese, Mayer & Jurgens 1998). At the edges of the expanding cell plate, nascent microfibrils and matrix polymers arriving from both daughter cells have the opportunity to mingle before becoming consolidated, in ways that are still unclear, into the completed structure of the newly inserted wall (Heese et al. 1998; Otegui & Staehelin 2000; Verma 2001). However, when two adherent cell walls expand together, the non-cellulosic polymers that will provide the increased area of middle lamella must diffuse through the interstices in the outermost microfibril layer of each of the cell walls.
We may anticipate that where a new cross-wall meets a pre-existing wall, the adhesion of walls at the tricellular junction may have different characteristics. Structural elements of the existing wall will remain intact across the edges of the cross-wall unless modified biochemically at the junction point, and these existing and new cross walls may have different adhesion characteristics. This is indeed observed: enzymic cell separation in elongating tissues, such as hypocotyls, releases long files of cells that have separated along longitudinal walls but remain attached to one another at the ends by means of transverse walls (Cocking 1960) and see also Fig. 2.
Figure 2. Regulation of plant cell separation. (a, b) In parenchyma systems, cells separate to a limited extent producing intercellular space at tricellular junctions (indicated by arrows). (a) A transverse section of a pea stem immunolabelled with antipectin monoclonal antibody JIM5 indicating all primary cell walls. (b) An equivalent section immunolabelled with antipectin monoclonal antibody LM7 indicates a pectic homogalacturonan epitope that is restricted to cell walls lining intercellular space and particularly the corners of tricellular junctions at points of cell adhesion/separation. Micrographs courtesy of Dr Bill Willats. For details see Willats et al. (2001). (c) Three cells from a cell file of mung bean hypocotyl isolated by treatment with pectin lyase (0.9 U mL−1) remain adhered at transverse cell walls (arrows). Scale bars (a, b) = 100 µm. (c) = 250 µm.
Download figure to PowerPoint
Structural polymers responsible for intercellular adhesion
Except in graminaceous plants the principal macromolecules of the middle lamella are pectins, accompanied by proteins (Jarvis 1984; Carpita & Gibeaut 1993). It is possible to consider the middle lamella and reinforcing zones simply as regions of the extracellular matrix from which microfibrils are absent, and which therefore lack cellulose. The microfibrils of the primary cell walls are themselves arranged in a more or less lamellate structure (Carpita & Gibeaut 1993) and it has been suggested that pectins attach each of the microfibril layers to the next (Jarvis 1998). In tobacco cells habituated to the herbicide dichlobenil, which inhibits cellulose biosynthesis, conventional microfibrils were absent and the middle lamella could not be distinguished ultrastructurally from the primary walls on either side of it (Sabba, Durso & Vaughn 1999)
That is not to say that the polymers of the middle lamella and those of the reinforcing zones, at tricellular junctions and the corners of intercellular spaces, are identical with the polymers of the interfibrillar matrix of the primary cell wall. They are not, and the differences are characteristic. In a wide range of dicotyledonous plant tissues linear, low-ester pectic homogalacturonans are most abundant in the middle lamella, and especially in the reinforcing zones (e.g. Knox et al. 1990; Roy, Vian & Roland 1992; Liners & Van Cutsem 1992; Willats et al. 1999; Parker et al. 2001).
Recently the introduction of the monoclonal antibody LM7 (Willats et al. 2001) has made it possible to distinguish the pectic galacturonans of the reinforcing zones from those of the middle lamella and primary cell wall. LM7 labelling in Pisum and other angiosperms was restricted to the reinforcing zones and, at lower intensity, to the adjacent cell wall surface lining intercellular spaces (see Fig. 2). LM7 binds to a partially methyl-esterified epitope of homogalacturonan that is abundant in samples with a random pattern of methyl-esterification (Willats et al. 2001). Random esterification is characteristic of the process by which methyl ester groups are added to pectins in the Golgi, whereas de-esterification of galacturonans of higher initial ester content, by the action of most plant pectin methylesterases, generates long galacturonan blocks completely free from methyl ester groups (Goldberg et al. 1996). However, plant pectin methyl esterases have varied action patterns (Catoire et al. 1998; Goldberg et al. 2001) and the LM7 epitope may be generated in muro.
Cross-linking of pectic polysaccharides
There must be a mechanism for the middle-lamella and reinforcing-zone polymers (contributed by both cells) to be linked into a coherent network. Otherwise there would be nothing to hold the two cells together. Whatever their pattern of esterification, low-ester pectic polymers have a relatively high affinity for calcium ions. Both the middle lamella and, in particular, the reinforcing zones show elevated levels of calcium imaged by SIMS (Rihouey et al. 1995) or EELS microscopy (Huxham et al. 1999). Low-ester galacturonans gel in vitro with calcium but the mechanical properties and porosity of these gels depend strongly on whether the esterification pattern is random or blockwise, independent of the degree of esterification (Willats et al. 2001).
A reasonable hypothesis might be that calcium-linked gels of low-ester pectins, rich in the LM7 homogalacturonan epitope, might be responsible for the cohesion of the reinforcing zones and hence for intercellular adhesion. However there is evidence that this is not the only mechanism involved. Extraction of dicot cell wall preparations with calcium-chelating agents brings pectin into solution, but normally much less than half of the total pectin present (Goldberg et al. 1996), and only limited cell separation occurs under these conditions (Cocking 1960; McCartney & Knox 2002). The presence of further, presumably covalent, intermolecular linkages must therefore be proposed. Borate diester links between RGII segments are unlikely because RGII was not detected in these locations (Matoh et al. 1998). Whatever the nature of these intermolecular cross-links, the enzymes catalysing their formation must reside in the middle lamella of expanding cells and in the reinforcing zones as they are formed – although similar cross-links and similar enzyme systems may also be found in the primary cell wall.
The three-dimensional cross-linked network formed in this way could be purely pectic, or it might include pectic chains covalently linked to other insoluble polysaccharides on both sides of the middle lamella. The formation of either of these kinds of network would also explain the insolubility of native pectins (‘protopectin’) in the primary walls of dicot cells (Goldberg et al. 1996).
A covalent polymer network can be disrupted either by breaking the cross-links between the chains or by cleaving the chains themselves between crosslinks. Enzymic cleavage of pectic galacturonan chains is an efficient method for pectin solubilization (Keegstra et al. 1973), and generally separates dicot cells (Ramana & Taylor 1994; Zhang, Henriksson & Johansson 2000) Interestingly, in a detailed study of polygalacturonase action on carrot cell walls, even though endo-polygalacturonase degradation removed most of the polymer material from the middle lamella of carrot and left it greatly weakened, a sparse reticulum of unidentified material remained (Tamura & Senda 1992). Enzymic cleavage of the galactan side-chains characteristic of some pectins neither solubilized these pectins nor separated cells with convincing efficiency (Redgwell & Harker 1995). These observations support the central role of cross-linked galacturonans in intercellular adhesion, but do not show how they are cross-linked.
Mild alkaline extraction, under conditions suitable for cleaving ester or labile amide linkages, solubilizes a substantial pectic fraction (Goldberg et al. 1996) and also separates some of the cells that are not separated by chelators. The nature of the alkali-labile bonds is unclear. Galacturonoyl ester links to hydroxyl groups on other chains have repeatedly been suggested (Jarvis 1982; Fry 1986; Kim & Carpita 1992; Brown & Fry 1993; MacKinnon et al. 2002), although such ester-linked fragments have not so far been isolated and characterized. The total fraction of substituted pectic carboxyl groups significantly exceeded the fraction esterified with methanol in cell walls of maize (Kim & Carpita 1992) and potato (Mackinnon et al. 2002). Although these additional substituents have not been identified, other hydroxyls on the same galacturonate residue (giving a lactone) can be ruled out on grounds of stability, suggesting that an inter-residue or intermolecular linkage is present. Amide linkages to peptides or polyamines could also have suitable properties for cross-linking pectic chains (Perrone et al. 1998). Bound polyamines are present in cell walls (Geny et al. 1997) and inhibition of polyamine biosynthesis affects cell wall integrity and intercellular adhesion (Berta et al. 1997), whether directly or indirectly.
All these observations are consistent with the idea that dicot cells of many types adhere to one another through the formation of a three-dimensional pectic gel network, which has features in common with the pectic network of the primary cell wall and is responsible for the insolubilization of native pectins (Goldberg et al. 1996). This network appears to contain cross-links of at least three types: chain aggregates held together by calcium and possibly other cations; covalent links having the alkali lability of esters or possibly amides; and alkali-resistant covalent links with characteristics similar to glycosidic bonds.
Beet, spinach and related species differ from other dicots in having an additional cross-linking mechanism in their cell walls, which involves the peroxidase-catalysed formation of dimeric feruloyl esters between the pectic side-chains, stabilizing the cells against separation (Waldron et al. 1997b) In grasses, cereals and related monocots the abundance of pectins in the primary cell wall is relatively low and cell separation by chelating agents is not generally possible. Feruloyl and other hydroxycinnamoyl esters are abundant, however. Their dimers cross-link cell wall polymers, in this case arabinoxylans (Ng, Greenshields & Waldron 1997), and fluorescence microscopy readily demonstrates their presence at tricellular junctions, like calcium-bound pectic galacturonans in dicots (Waldron et al. 1997a).
The spatial pattern of polymer cross-linking matches the location of mechanical stresses imposed by turgor pressure (Jarvis 1998). In many tissues of most dicots that have been studied, LM7-binding pectins and associated calcium ions are concentrated exactly where the stress is greatest, at the tricellular junctions and the corners of the intercellular spaces. In beets a similar spatial pattern is shown by ferulate esters (M. Marry, unpublished results). The small amount of low-ester pectic galacturonan in oat roots and the abundant ferulate in water chestnut (Cyperaceae, with cell walls like a cereal) also show this spatial distribution (Waldron et al. 1997a). This evidence is consistent with the idea that intercellular adhesion depends on the formation of a specific type of cross-linked pectic network at a specific, mechanicallly stressed extracellular location, the reinforcing zones. This is the case even though the polysaccharides concerned differ between dicots and graminaceous plants, and though the enzymes responsible for covalent cross-linking must also differ.
Intercellular adhesion, secretion and plant morphogenesis
It seems reasonable to infer the existence of a mechanism for secreting polysaccharides and proteins to precisely the required locations through the plasma membrane and the separate barrier of the primary cell wall. How this happens is unknown. The Arabidopsis gene EMB30 (GNOM) encodes a protein required for the targeting of secretion to specific parts of the cell surface. Mutants in EMB30 are defective in cell adhesion as well as in polar cell expansion and the control of plant form (Shevell, Kunkel & Chua 2000). The defect in cell adhesion was associated with accumulation of pectic polysaccharides in the intercellular spaces, as if the pectins themselves were secreted normally but a factor that retained them in the reinforcing zones of the wild-type was absent (Shevell et al. 2000).
In the tomato mutant Cnr, which shows defective cell adhesion in the pericarp of ripe fruit, altered spatial patterns of galacturonan esterification were accompanied by blocked secretion, across the plasma membrane, of a pectic polymer rich in α(1,5)-l-arabinan (Orfila et al. 2001). A Nicotiana callus with reduced cell adhesion similarly showed low levels of pectic arabinan, and lost galacturonan from the middle lamella into the culture medium (Iwai, Ishii & Satoh 2001) There was no suggestion that the tomato arabinan was directly involved in cell adhesion as the LM6 epitope was restricted to the primary cell wall, and was absent from the middle lamella and reinforcing zones, in the wild type (Orfila et al. 2001). This epitope is characteristic of the primary walls of proliferating cells, rather than walls that have expanded (Willats et al. 1999; McCartney et al. 2000; Bush et al. 2001). However it is possible that the secretory defect also affected other polymers more directly involved in network formation at the points of maximum stress.
Reduced intercellular adhesion is a common feature of a group of Arabidopsis mutants in which cytokinin responses are disrupted (Delarue et al. 1997; Faure et al. 1998). It is also found in the tumorous shoot development (tsd) mutants that have defects in what appears to be an intercellular signalling system involving class I knox homeobox genes and cytokinin sensitivity (Frank et al. 2002). In some tissues of these mutants the reduced intercellular adhesion results in ‘vitreous’ texture as enlarged intercellular spaces become filled with apoplastic liquid, an effect that can be phenocopied by excess cytokinin (Faure et al. 1998). In the tsd mutants the organization of the shoot apical meristem is drastically disrupted, apparently due to failure of interlayer signalling. It was pointed out by Frank et al. (2002) that the developmental disorganization and the loss of intercellular adhesion seemed to be connected, and Shevell et al. (2000) suggested that normal cell adhesion was required for the intercellular signalling needed to co-ordinate morphogenesis (Gisel et al. 1999). However the chain of cause and effect in the complex emb30 and tsd phenotypes is not clear.