Encoding Specificity in Plant Calcium Signalling: Hot‐spotting the Ups and Downs and Waves

Abstract

Calcium ions function as intracellular second messengers in regulating a plethora of cellular processes from acclimative stress responses to survival and programmed cell death. The generation of specificity in Ca²⁺ signals is dependent on influx and efflux from the extracellular milieu, cytosol and intracellular organelles. One aspect of plant Ca²⁺ signalling that is currently attracting a great deal of interest is how ‘Ca²⁺‐signatures’, specific spatio‐temporal changes in cytosolic‐free Ca²⁺, encode the necessary information to bring about this range of physiological responses. Here, current information is reviewed on how Ca²⁺‐signatures are generated in plant cells and how stimulus‐specific information can be encoded in the form of Ca²⁺‐signatures.

INTRODUCTION

The Ca²⁺ ion is now firmly established as a ubiquitous signalling molecule in plants. Numerous plant signal transduction pathways have been shown to use Ca²⁺ as an integral signalling component (Sanders et al., 1999, 2002). The universality of the Ca²⁺ ion in signalling highlights the importance of understanding how specificity can be encoded in elevations in the cytosolic concentration of this ion (Sanders et al., 1999, 2002; Berridge et al., 2000). Perhaps the best example of this in plants is the response of stomata to the plant hormones abscisic acid (ABA) and auxin, which bring about the diametrically opposite effects of stomatal closure and opening, respectively, via changes in guard cell turgor mediated in both cases through increases in the concentration of cytosolic‐free Ca²⁺ ([Ca²⁺]_cyt) (McAinsh et al., 1990; Irving et al., 1992). One plausible explanation is that each stimulus generates a unique increase in [Ca²⁺]_cyt. The spatial and temporal components of this increase in [Ca²⁺]_cyt, or ‘Ca²⁺‐signature’ as it is sometimes called, then dictate the outcome of the final response (McAinsh et al., 1997; McAinsh and Hetherington, 1998; Ng et al., 2001b). It has been suggested that the key to generating stimulus‐specific Ca²⁺‐signatures lies in the ability to access differentially the cellular machinery controlling Ca²⁺ influx and release from intracellular stores (McAinsh et al., 1997; Blatt, 2000; Evans et al., 2001; Ng et al., 2001b; Schroeder et al., 2001).

Ca²⁺‐mobilizing signalling intermediates that have been implicated in mediating elevations in [Ca²⁺]_cyt in plants include, inositol‐1,4,5‐trisphosphate (InsP₃), inositol hexakisphosphate (InsP₆), nicotinic acid adenine dinucleotide phosphate (NAADP), phosphatidylinositol 3‐ and 4‐phosphate (PI3P and PI4P), cyclic adenosine 5′‐diphosphoribose (cADPR), hydrogen peroxide (H₂O₂) and sphingosine‐1‐phosphate (S1P) (Gilroy et al., 1990; McAinsh et al., 1996; Leckie et al., 1998; Staxén et al., 1999; Lemtiri‐Chlieh et al., 2000; Navazio et al., 2000; Pei et al., 2000; Ng et al., 2001a; Jung et al., 2002). At the cellular level, it is possible that this complexity may contribute to the spatio‐temporal variations in the Ca²⁺‐signatures, needed for specifying stimulus‐specific responses. Furthermore, the locale within the cell of such elevations in [Ca²⁺]_cyt, whether the appropriate response elements are present in a given region of the cell to decode the information encrypted in the Ca²⁺ signal, may also be important (Trewavas and Malhó, 1997; McAinsh and Hetherington, 1998; Malhó et al., 1998; Evans et al., 2001; Ng et al., 2001b). It is not the intention here to provide an exhaustive review of Ca²⁺ signalling in plants; for this information, the reader is referred to Sanders et al. (1999, 2002). Instead, the purpose of this review is to draw attention to the different types of Ca²⁺ elevations that occur in plants, which can take the form of hot‐spots, puffs, sparks, oscillations and waves, and the importance of stimulus‐specific Ca²⁺‐signatures for encoding information necessary for eliciting the appropriate physiological responses in plant cells. The various forms of Ca²⁺ elevations are discussed below, with examples given in the figures: hot‐spots (Fig. 2), puffs and sparks (Figs 1 and 3), oscillations (Fig. 5) and waves (Fig. 1).

Fig. 2. Spatial heterogeneities in guard cell [Ca²⁺]_cyt in response to 100 nM ABA (A) and 1 mm [Ca²⁺]_ext (B). [Ca²⁺]_cyt levels are colour‐coded; blue indicates low [Ca²⁺]_cyt, red indicates high [Ca²⁺]_cyt. These data suggest that plant cells have the capacity to encode specificity in the Ca²⁺ signal in the form of localized increases in [Ca²⁺]_cyt. Reproduced, with permission, from McAinsh et al. (1992, 1995).

Fig. 1. Elemental Ca²⁺ elevations during Ca²⁺ wave propagation in a Fucus rhizoid cell. (A), Single‐line confocal scans of Ca²⁺ Green fluorescence along the longitudinal axis of the cell during the initation of Ca²⁺ waves are displayed sequentially to show the relative change in fluorescence following hypo‐osmotic treatment (to 50 % sea water). (B), Three‐dimensional plot shows a non‐uniform increase in Ca²⁺ during the onset of the Ca²⁺ wave in the rhizoid apex. (C), Elemental Ca²⁺ elevations in the perinuclear region. (D), Elemental Ca²⁺ events appear to arise repetitively at the same location in perinuclear region. Reproduced, with persmission, from Goddard et al. (2000).

Fig. 3. (Opposite) Elemental Ca²⁺ elevations during Ca²⁺ wave propagation in a Fucus rhizoid cell and patterns of Ca²⁺ elevations following hypo‐osmotic treatment in dividing cells. (A), Elemental Ca²⁺ events in the perinuclear region either occur individually or appeared to cluster into more prolonged elevations. (B), Number of discrete elevations at the rhizoid apex increased initially during the first 1·5 s of wave propagation and then declined with the appearance of more prolonged elevations. (C), Ca²⁺ Green to Teax Red ratio images during the onset of a hypo‐osmotically induced Ca²⁺ waves (100 % sea water to 50 % sea water) shows an initial elevation of Ca²⁺ in the rhizoid apex which declines before the onset of Ca²⁺ elevation arising in the apical nucleus region. (D), A minority of cells shows a variation of the pattern shown in part C where the Ca²⁺ elevations were observed to arise in the subapical nucleus simultaneous with the apical Ca²⁺ elevation. Reproduced, with permission, from Goddard et al. (2000).

Fig. 5. Encoding signalling information in plant Ca²⁺ signatures. (A), In guard cells the strength of the stimulus has been correlated directly with the pattern of [Ca²⁺]_cyt oscillations (i.e. the period, frequency and amplitude) which, in turn, dictates the resultant steady‐state stomatal aperture. (B), Guard cells are able to integrate signalling information from a number of stimuli that induce oscillations in [Ca²⁺]_cyt applied simultaneously to generate a novel Ca²⁺ signature when formulating the final stomatal aperture. Reproduced, with permission, from Evans et al. (2001).

HOT‐SPOTTING: ELEMENTAL EVENTS

Studies using animal cells have shown that elevations in [Ca²⁺]_cyt induced by agonists, e.g. InsP₃ and cADPR, through their respective Ca²⁺ release‐channels, InsP₃R and ryanodine receptors (RyR), respectively, involve a hierarchically distinguishable series of sub‐threshold events (Berridge et al., 2000; Bootman et al., 2001). Low levels of stimulation result in single channel events, leading to Ca²⁺ elevations known as ‘quarks’ and ‘blips’. The Ca²⁺ release associated with higher levels of stimulations is termed ‘puffs’ and ‘sparks’. Quarks, blips, puffs and sparks constitute what is collectively termed ‘elemental events’ and form the fundamental building blocks for global Ca²⁺ signalling, e.g. Ca²⁺ waves. In plants, the best examples of elemental events in Ca²⁺ signalling were obtained using Fucus serratus embryos by Brownlee and co‐workers (Goddard et al., 2000; Fig. 1). They showed that hypo‐osmotic shock induced unitary Ca²⁺ elevations in discrete domains within the cytosol of Fucus embryos. These discrete elevations in [Ca²⁺]_cyt are in the range 200–300 nm and last between 15 and 30 ms (Fig. 1C). Interestingly, these unitary increases in [Ca²⁺]_cyt occur repetitively and are spatially separated by regions with fewer events (Fig. 1D). Unitary elevations in Ca²⁺ of similar magnitudes, amplitudes and spatial dimensions were also induced by UV‐photolysis of caged‐InsP₃, and suggest a role for InsP₃ in the generation of osmotically induced Ca²⁺ signals, although direct evidence for this is lacking (Goddard et al., 2000). The observation of these unitary increases in [Ca²⁺]_cyt in Fucus provides the first evidence in support of elemental events as fundamental building blocks in plant Ca²⁺ signalling.

In stomatal guard cells, stimulus‐induced elevations in [Ca²⁺]_cyt also show marked spatial heterogeneities (Fig. 2). McAinsh et al. (1992) observed that ABA‐induced elevations in [Ca²⁺]_cyt were unevenly distributed and appeared as ‘hot‐spots’ and Ca²⁺‐quiescent regions. Spatial heterogeneity in guard cell [Ca²⁺]_cyt has also been reported by Gilroy et al. (1991) and McAinsh et al. (1995). It is possible that the spatial heterogeneities in [Ca²⁺]_cyt elevations could result from (a) differential accessibility of the primary stimulus to only a subset of the signalling machinery, or (b) the non‐uniform distribution of the intracellular signalling machinery. These observations suggest the potential for encoding specificity in the form of localized increases in [Ca²⁺]_cyt. It is tempting to suggest that these localized Ca²⁺ ‘hot‐spots’ observed in guard cells represent elemental events (puffs and sparks). However, due to the spatial and temporal resolution used in these studies, it is likely that these localized elevations in [Ca²⁺]_cyt represent longer transients in [Ca²⁺]_cyt, as opposed to the unitary Ca²⁺ elevations observed in Fucus in response to osmotic stress and UV‐photolysis of caged‐InsP₃.

TRIGGERING WAVES

Ca²⁺ waves represent a form of global increase in [Ca²⁺]_cyt that are triggered by the clustering of unitary Ca²⁺ elevations leading to Ca²⁺‐induced Ca²⁺‐release (CICR). Ca²⁺ waves have been extensively studied in animal cells (Berridge et al., 2000; Bootman et al., 2001). It is now well established that Ca²⁺ signalling in plants can also take the form of a propagating Ca²⁺ wave from studies using Fucus (Taylor et al., 1996). Goddard et al. (2000) showed an increase in the number of unitary Ca²⁺ elevations following exposure to hypo‐osmotic stress, leading subsequently to clustering and more prolonged elevations (Fig. 3A and B). Interestingly, they observed that clustering and prolonged elevations preceded the formation of Ca²⁺ waves. This suggests that the clustering of unitary Ca²⁺ elevations acts as the trigger for the generation of a Ca²⁺ wave (Fig. 3C). In this respect, the unitary Ca²⁺ elevations in Fucus are reminiscent of Ca²⁺ quarks/blips reported in animal cells and suggest that the clustering of such elemental events may be representative of Ca²⁺ puffs and sparks.

Importantly, the authors showed that the spatio‐temporal changes in [Ca²⁺]_cyt can vary with the strength of the hypo‐osmotic shock treatment, suggesting that variations in the signature of the Ca²⁺ wave can determine downstream physiological responses (Goddard et al., 2000). In rhizoid cells that had undergone nuclear division but not partition wall formation, Ca²⁺ waves were observed to propagate from two nuclear regions, with most cells showing an initial transient Ca²⁺ elevation in the rhizoid apex, followed by Ca²⁺ elevations in the apical nuclear region (n1) that subsequently spread to the sub‐apical nucleus (n2) (Fig. 3C). However, in 20 % of the cells, Ca²⁺ elevations were observed to occur in the region of the sub‐apical nucleus (n2) either before or simultaneously with Ca²⁺ elevations in the rhizoid apex followed by Ca²⁺ elevations in the apical nucleus (n1) (Fig. 3D). Additionally, the authors showed that these Ca²⁺ elevations were highly correlated with the distribution of endoplasmic reticulum (ER) and that these specific patterns of Ca²⁺ wave generation may encode the necessary information for differential regulation of cell volume changes and the rate of cell division (Goddard et al., 2000).

Recently, Coelho et al. (2002) showed that the generation of reactive oxygen species (ROS) are tightly linked to the generation of Ca²⁺ waves in Fucus and suggested the following sequence of events may occur during osmotic stress (Fig. 4). (a) Sensing of osmotic changes by an unidentified osmosensor results in extracellular production of ROS. (b) This initial ROS burst then activates the influx of Ca²⁺ through Ca²⁺‐permeable channels on the plasma membrane. Interestingly, H₂O₂ has also been reported to activate plasma membrane Ca²⁺‐permeable channels in guard cells (Pei et al., 2000). (c) Ca²⁺ influx then leads to elevations in [Ca²⁺]_cyt, triggering CICR from the ER via InsP₃‐sensitive release mechanisms. (d) This InsP₃‐ dependent release of Ca²⁺ from the ER results in the formation of a Ca²⁺ wave followed by mitochondrial uptake of Ca²⁺ and subsequent mitochondrial production of ROS. Although the significance of the subsequent production of ROS by the mitochondria remains to be established, this study nevertheless highlights the importance of peripheral ROS production and Ca²⁺ influx through plasma membrane Ca²⁺‐permeable channels in the generation of Ca²⁺ waves during acclimative responses to osmotic stress in Fucus.

Fig. 4. Proposed signalling pathway during hyperosmotic stress in Fucus embryos. Osmotic change is sensed by an unidentified sensor leading to the production of extracellular ROS production (which may be important for strengthening the cell wall). The H₂O₂ formed also diffuses into the cell, leading to localized peripheral intracellular increase. The extracellular H₂O₂ also increases Ca²⁺‐channel activity. This leads to Ca²⁺‐induced Ca²⁺ release from intracellular stores (ER) resulting in Ca²⁺ wave propagation and mitochondrial ROS production. Reproduced, with permission, from Coelho et al. (2002).

OSCILLATIONS: SIMPLY UP AND DOWN?

Since the initial discovery of external Ca²⁺ ([Ca²⁺]_ext)‐induced oscillations in [Ca²⁺]_cyt in guard cells (McAinsh et al., 1995), stimulus‐induced oscillations in [Ca²⁺]_cyt have also been observed in other cell types, including pollen tubes (Holdaway‐Clarke et al., 1997), roots (Kiegle et al., 2000) and root hairs (Ehrhardt et al., 1996). Oscillations in [Ca²⁺]_cyt result from the dynamic balance of fluxes of Ca²⁺ into and out of the cytosol and include release and uptake from intracellular stores (vacuole and ER) and nucleus as well as influx and efflux across the plasma membrane. These fluxes are mediated through the regulated activities of Ca²⁺ channels and Ca²⁺‐pumps located on the plasma and endo‐membranes (Bunney et al., 1999; Sanders et al., 1999, 2002; Pauly et al., 2000; Harper, 2001; Miedema et al., 2001). Oscillations in [Ca²⁺]_cyt allow for information to be encoded in both the amplitude and frequency (Fig. 5) (Berridge et al., 1988; Fewtrell, 1993; McAinsh et al., 1997; McAinsh and Hetherington, 1998; Evans et al., 2001; Ng et al., 2001b). Of the various cell types that have been used to study stimulus‐induced oscillations in [Ca²⁺]_cyt, the guard cell has emerged as the most intensively used system.

Studies in the Hetherington and McAinsh laboratories using manganese quenching and imaging have shown that [Ca²⁺]_ext‐induced oscillations in [Ca²⁺]_cyt in guard cells of Commelina communis result from both Ca²⁺ influx through the plasma membrane and release from intracellular stores (McAinsh et al., 1995). Later studies from the Blatt and Schroeder laboratories using Vicia faba and Arabidopsis thaliana guard cells demonstrated that Ca²⁺ influx through the plasma membrane is gated by hyperpolarization‐dependent Ca²⁺ channels (Grabov et al., 1998, 1999; Hamilton et al., 2000; Pei et al., 2000). Interestingly, Blatt and co‐workers showed that the influx of Ca²⁺ through the plasma membrane Ca²⁺ channels are coupled to oscillations in plasma membrane potentials, and that ABA can regulate this influx of Ca²⁺ by increasing the probability of channel opening and by shifting the voltage sensitivity of these channels to more depolarizing potentials. Together these results suggest that ABA‐induced oscillations in [Ca²⁺]_cyt is the result of influx of Ca²⁺ through the plasma membrane due in part to channel gating by oscillations in membrane potentials and greater probabilities of Ca²⁺ channel opening (Grabov et al., 1998, 1999; Hamilton et al., 2000). More recently, the Schroeder laboratory showed that ABA induced H₂O₂ production, and subsequent H₂O₂‐activation of Ca²⁺ influx through hyperpolarization‐activated guard cell plasma membrane Ca²⁺‐permeable channels can contribute to increases in [Ca²⁺]_cyt (Pei et al., 2000).

In addition to the role of plasma membrane Ca²⁺ influx, a variety of second messenger systems exist for the release of Ca²⁺ from intracellular stores, for the generation of increases in [Ca²⁺]_cyt. One of the earliest indications of an intracellular route for Ca²⁺ release in guard cells was reported by the Trewavas laboratory (Gilroy et al., 1990). They showed that UV‐photolysis of caged InsP₃ microinjected into the cytosol of guard cells elevated [Ca²⁺]_cyt and stimulated stomatal closure. This artificial elevation of cytosolic InsP₃ could also reversibly inactivate inward‐rectifying K⁺ channels (I_K,in), whilst at the same time activating an inward current that depolarizes the plasma membrane (Blatt et al., 1990). Many elements of an InsP₃‐mediated signalling pathway have been identified in plants, and Lee et al. (1996) have shown that ABA induced the rapid turnover of phosphoinositides in guard cell protoplasts. Staxén et al. (1999) have also demonstrated the involvement of a phosphoinositide‐specific phospholipase C (PI‐PLC)‐InsP₃ Ca²⁺‐mobilizing pathway in the regulation of guard cell turgor by ABA but not [Ca²⁺]_ext. Their data suggest that ABA‐induced oscillations in [Ca²⁺]_cyt have the potential to encode information on the strength of the initial ABA stimulus and for maintaining steady‐state stomatal apertures (Staxén et al., 1999). Hunt et al. (2003) have subsequently confirmed the role of PI‐PLC in the control of stomatal aperture by ABA. Another inositol phosphate that has been implicated in ABA‐mediated changes in guard cells is InsP₆ (Lemtiri‐Chlieh et al., 2000). ABA has been shown to induce a rapid production of InsP₆ and that loading of InsP₆ into the cytosol of guard cell protoplasts inhibits I_K,_in in a Ca²⁺‐dependent manner. It will be of interest to determine the exact nature of the Ca²⁺‐dependence of InsP₆‐mediated inhibition of I_K,in and the modulation of [Ca²⁺]_cyt by InsP₆.

More recently, Jung et al. (2002) used a pharmacological approach to demonstrate that PI3P and PI4P are components of the ABA signal transduction pathway mediating stomatal closure, and that inhibitors of PI3‐kinase and PI4‐kinase (wortmanin and LY294002) inhibited ABA‐induced oscillations in [Ca²⁺]_cyt. Together, these data suggest that inositol phosphates are important regulatory components in guard cell ABA signalling. Other Ca²⁺‐mobilizing second messengers known to function in regulation of guard cell turgor include cADPR and the recently identified sphingolipid metabolite, S1P. Leckie et al. (1998) observed that most cells microinjected with cADPR showed a sustained increase in [Ca²⁺]_cyt lasting in excess of 10 min while oscillations were observed in some cells. The increase in [Ca²⁺]_cyt elicited by cADPR ranged from 0·05 to 0·4 µm. In the case of S1P, the increase in [Ca²⁺]_cyt take the form of oscillations that have a period and amplitude that is characteristic of the concentration of S1P used. The differential pattern of [Ca²⁺]_cyt induced by different concentrations of S1P can be correlated with the rate of stomatal closure and is further evidence in support of oscillations in the control of guard cell turgor (Ng and Hetherington, 2001; Ng et al., 2001a).

The removal of Ca²⁺ from the cytosol is also an important factor in the generation of oscillations in [Ca²⁺]_cyt causing the downward phase of the oscillation. The importance of cytosolic Ca²⁺ removal is highlighted by Allen et al. (2000) who showed that the V‐type ATPase activity is required for returning [Ca²⁺]_cyt to resting levels. The authors showed in the det3 (de‐etiolated 3) mutant of arabidopsis (which exhibits decreased endomembrane energization due to a 60 % decline in the expression of the C‐subunit of the V‐type ATPase) that a correlation exists between stomatal closure and [Ca²⁺]_ext‐induced [Ca²⁺]_cyt oscillations. Guard cells of det3 mutants do not close in response to [Ca²⁺]_ext and do not exhibit oscillations in [Ca²⁺]_cyt compared with wild‐type guard cells. Interestingly, [Ca²⁺]_cyt oscillations induced by cold treatment and ABA were not affected in guard cells of the det3 mutants and suggest a scenario where multiple pathways of cytosolic Ca²⁺ removal exist and that interact differentially to bring about the complexities observed in stimulus‐induced oscillations in [Ca²⁺]_cyt.

The mechanism by which stimulus‐specific information may be encoded in [Ca²⁺]_cyt oscillations has been studied in the Schroeder laboratory. In an elegant study, Allen et al. (2001) showed that the information encrypted in elevations in [Ca²⁺]_cyt can be separated into two components: (1) ‘Ca²⁺‐reactive’ and (2) ‘Ca²⁺‐programmed’. These two distinct components encode information relating to short‐term and long‐term acclimative responses, respectively. The ‘Ca²⁺‐reactive’ component of the elevation in [Ca²⁺]_cyt occurs rapidly and encodes information for the initiation of stomatal closure. This rapid reactive component usually takes the form of the initial transient elevation in [Ca²⁺]_cyt. On the other hand, the ‘Ca²⁺‐programmed’ component encodes information related to the maintenance of steady‐state stomatal apertures and is usually represented by a series of oscillations. The number of transients (a complete cycle consisting of an upward and downward phase) determines the final steady‐state aperture: the greater the number of transients, the smaller the steady‐state stomatal aperture (Allen et al., 2001). Taken together, these data make a compelling case for the encryption of important physiological information in the form of [Ca²⁺]_cyt oscillations.

CONCLUSIONS AND FUTURE PROSPECTS

The potential for encoding specificity in the Ca²⁺ signal has been demonstrated by the ability of guard cells to generate stimulus‐specific temporal changes in [Ca²⁺]_cyt in the form of oscillations. Additional information can also be encoded in the form of complex patterns of [Ca²⁺]_cyt oscillations generated in response to two or more stimuli (Hetherington et al., 1998; McAinsh et al., 2000; Ng et al., 2001b). The ability to modulate the pattern of the Ca²⁺‐signature suggests the presence of a complex cellular machinery capable of integrating stimuli perception into a meaningful physiological response. While [Ca²⁺]_cyt oscillations play an important role in turgor regulation in stomatal guard cells, stimulus‐specific patterns of Ca²⁺ wave formation and propagation appear to be important in specifying the information necessary for differential regulation of cell volume changes and the rate of cell division. The ability for encryption of stimulus‐specific information immediately raises the question of how the cell is able to decode the information into a meaningful physiological response. It has been suggested that a variety of Ca²⁺‐binding proteins, Ca²⁺‐dependent protein kinases and protein phosphatases may provide the diversity needed for decoding the encrypted Ca²⁺ signal (Luan et al., 1993; Berkowitz et al., 2000; Guo et al, 2002). Webb and Hetherington (1997) have suggested that signalling components may be organized in the form of a ‘cassette’ and function as focal points for signal convergence and integration. Trewavas and Malhó (1997) also suggested a similar concept of a ‘transducon’, large protein complexes where signalling components may be spatially organized within the cellular milieu. The hypotheses that signalling components are organized into spatially distinct cassettes/transducons was lent credence by the recent observation of Guo et al. (2002) who showed the physical interaction of ScaBP5, a Ca²⁺‐binding protein with PKS3, its interacting protein kinase. PKS3 in turn interacts with ABI2 and ABI1. This suggests that ScaBP5, PKS3 and ABI2 and/or ABI1 are spatially organized, perhaps in a cassette/transducon, for efficient transduction of the ABA signal. Insights into how stimulus‐specific signals are encrypted and the underlying mechanism(s) for decoding the Ca²⁺ signal into meaningful physiological responses are likely to be gained through the systematic identification of genes, their products and the interactions between the various signalling components.

ACKNOWLEDGEMENTS

We are grateful to the Biotechnology and Biological Sciences Research Council, UK (M.R.M.), The Royal Society, UK (M.R.M.) and University College Dublin (C.K.‐Y.N.) for research funding.

Received: 8 May 2003; Returned for revision: 2 June 2003; Accepted: 15 June 2003    Published electronically: 21 August 2003