*For correspondence (fax +44 1789 470552; e-mail email@example.com). †These authors contributed equally to this manuscript and should be regarded as interchangeable, joint first authors.
Ion transport processes at the plasma membrane of plant cells are frequently studied by applying membrane-patch voltage-clamp (patch–clamp) electrophysiological techniques to isolated protoplasts. As plants are composed of many tissues and cell types, and each tissue and cell type may be specialized to a particular function and possess a unique complement of transport proteins, it is important to certify the anatomical origin of the protoplasts used for patch–clamp studies. This paper describes a general molecular genetic approach to marking specific cell types for subsequent patch–clamp studies and presents a specific example: a comparison of the K+ currents in protoplasts from cortical and stelar cells ofArabidopsisroots. TransgenicArabidopsiswere generated in which the expression of green fluorescent protein (GFP) fromAequoria victoriawas driven by the CaMV 35S promoter (line mGFP3). In roots of the transgenic mGFP3 line, visible fluorescence was restricted to the stele. Protoplasts were generated from roots of the mGFP3 line and K+ currents in non-fluorescent (cortical/epidermal) and fluorescent (stelar) protoplasts were assayed using patch–clamp techniques. It was found that both the frequency of observing inward rectifying K+ channel (IRC) activity and the relative occurrence of IRC compared to outward rectifying K+ channels were significantly lower in protoplasts from cortical/epidermal cells compared to cells of the stele. The presence of GFP did not affect the occurrence or biophysical properties of K+ channels. It is concluded that the generation of transgenicArabidopsisexpressing GFP in a cell-specific fashion is a convenient and reliable way to mark protoplasts derived from contrasting cell types for subsequent patch–clamp studies.
The application of membrane-patch voltage-clamp (patch–clamp) electrophysiological techniques has revolutionized our understanding of both passive (channels) and active (carriers and pumps) transport systems. Consequently, an ever increasing number of laboratories are employing patch–clamp techniques to study the mechanisms of ion transport across plant cell membranes ( Sanders & Tester 1997). The patch–clamp technique requires that the plasma membrane of an appropriate cell is clear of cell-wall material, and, for studies using intact cells, that the cell is symplastically isolated ( Hedrich & Schroeder 1989). These conditions are usually met by performing experiments on protoplasts.
Frequently protoplasts are generated by the digestion of bulk plant tissue. However, plant organs are composed of many cell types. Each cell type may be specialized to a particular function, and hence possess a unique complement of ion transport proteins. Each cell type is likely to express contrasting genes and to respond differently to environmental or developmental cues. To interpret patch–clamp experiments in a physiological context, it is necessary to identify the origins of the protoplasts studied.
Some attempts have been made to identify specific cell types for patch–clamp studies, relying on positional, morphological or physiological characteristics. In the leaf, guard-cell protoplasts prepared from strips of epidermis can be identified from their small diameter ( Schroeder et al. 1984 ) and protoplasts from epidermal cells can be distinguished from mesophyll cells by the abundance of chloroplasts ( Elzenga et al. 1991 ). Furthermore, protoplasts from motor cells can be separated ( Moran et al. 1988 ;Stoeckel & Takeda 1993) and advantage can be taken of lignification patterns to isolate protoplasts from xylem contact cells ( Keunecke et al. 1997 ). In the root, protoplasts from root hairs can be individually selected ( Gassmann & Schroeder 1994;Kurkdjian et al. 1993 ) and advantage can be taken of the Casparian strip to separate stele from cortex prior to making protoplasts ( Roberts & Tester 1995) or to isolate protoplasts from xylem parenchyma cells ( Wegner & Raschke 1994). Protoplasts can be made from pollen tubes ( Obermeyer & Kolb 1993) and isolated from the aleurone of seeds ( Bush et al. 1988 ). However, these are only a small fraction of the total number of cell types in a plant, and methods for the isolation of protoplasts from specific tissues are generally cumbersome, time-consuming and restricted to certain plant species.
In this paper, we describe a molecular genetic cell-marking approach suitable for patch–clamp studies. The approach is based on controlling the expression of marker genes by tissue-specific promoters. A number of plant genes exhibit tissue-specific expression and their promoters can be used to control marker gene expression. Alternatively, (uncharacterized) endogenous promoters can be used to drive tissue-specific expression of a marker gene in ‘promoter-trapped’ or ‘enhancer-trapped’ mutants ( Topping & Lindsey 1995). This allows a great variety of cell types to be marked (and studied) individually.
The marker gene we have chosen to use is that encoding the green fluorescent protein (GFP) of Aequoria victoria, which has been modified (i) to prevent the excision of a cryptic intron, (ii) to have improved GFP fluorescence and (iii) to be retained in the endoplasmic reticulum (mgfp5-ER;Haseloff et al. 1997 ). The usefulness of GFP to mark cells for physiological experiments is that tissues can be imaged directly. This contrasts with the traditional marker gene in plants, β-glucuronidase (GUS) for which: (i) standard GUS substrates are generally membrane-impermeant and can produce cytotoxic products, (ii) use of the non-toxic, membrane-permeant substrate 5-dodecanoylamino fluorescein di-β- d-glucuronide (C12FGlcU;White et al. 1996 ), whose fluorescent product remains within the cell, is compromised by photobleaching and by autofluorescence, and (iii) endogenous, physical damage-induced or micro-organism-related GUS activities are difficult to avoid during the preparation of protoplasts for patch–clamp experiments.
Our studies have utilized transgenic Arabidopsis thaliana (L.) Heynh., ecotype Wassilewskija, expressing GFP under the control of the CaMV 35S promoter. In the roots of these plants, visible GFP fluorescence was restricted to cells within the stele. This is consistent with the reported expression patterns of both GUS ( Benfey et al. 1989 ) and mgfp4-ER ( Haseloff et al. 1997 ) in transgenic Arabidopsis when controlled by the CaMV 35S promoter. Protoplasts of root cells from the transgenic lines were amenable to patch–clamp techniques, which allowed us to compare the dominant, time-dependent, inward and outward K+ currents in fluorescent (stelar) and non-fluorescent, non-stelar (cortical/epidermal) protoplasts. The relative frequency with which the inward and outward K+ currents were observed differed between cell types. However, the frequency of observing, and the biophysical properties of K+ currents were similar in both transgenic and wild-type plants, suggesting that these parameters were not affected by the presence of GFP. Tissue-specific expression of GFP was found to be a convenient way to identify protoplasts for patch–clamp studies and, as more tissue- and cell-specific promoters become available, it will be possible to use the cell-marking approach to study the transport properties of many different cell types.
Expression of GFP in roots of the mGFP3 transgenic line
Ten transformants were selected which showed strong CaMV 35S-driven GFP expression. Variability of expression is frequently observed between individual transgenic lines transformed with the same construct, even under the control of the widely expressed CaMV 35S promoter ( Wilkinson et al. 1997 ). We therefore chose a homozygous line (mGFP3) for further experimentation which showed high levels of GFP expression consistent with that previously described for CaMV 35S-driven reporter genes ( Benfey et al. 1989 ).
The mGFP3 line showed no visible signs of deficiency or damage, and no differences in root morphology were apparent between the mGFP3 line and wild-type Arabidopsis. No fluorescence was observed in the roots of wild-type plants. In the primary and lateral roots of transgenic plants, GFP was observed predominantly in the vasculature ( Fig. 1). The GFP was located in all living cells of the stele, including the pericycle. For the most part, the occurrence and location of fluorescence was independent of physiological plant age. However, fluorescence did appear to be greatest in plants younger than 5 weeks old, and additional fluorescence was found in the root tips of very young plants, which declined over the first two weeks of growth.
Fluorescent protoplasts from roots of the mGFP3 transgenic line
Protoplasts were generated from roots of the mGFP3 transgenic line using conventional procedures ( Maathuis & Sanders 1995). No fluorescence was observed in protoplasts from the roots of wild-type plants when illuminated at an excitation wavelength of 490 nm. By contrast, significant numbers of protoplasts from roots of the mGFP3 transgenic line were highly fluorescent (22%, n = 137;Fig. 2). Fluorescence was confined to the cytoplasm, and high-resolution confocal microscopy suggested the presence of highly fluorescent cytoplasmic domains (cf. Haseloff et al. 1997 ). These are likely to coincide with the endoplasmic reticulum since the mGFP5-ER sequence includes an N-terminal signal peptide sequence from Arabidopsis thaliana basic chitinase and a C-terminal HDEL sequence ( Haseloff et al. 1997 ). The average sizes of fluorescent and non-fluorescent protoplasts from roots of the mGFP3 transgenic line were 11 ± 5 μm (n = 39) and 24 ± 7 μm (n = 41), respectively. This reflects the smaller size of stelar cells in Arabidopsis roots compared to non-stelar cells (e.g. Figures 1 and 2), and, combined with the data shown in Fig. 1, strongly supports a stelar origin for the fluorescent protoplasts. Henceforth, fluorescent protoplasts will be referred to as ‘stelar’ protoplasts. The non-fluorescent protoplasts assayed from roots of the mGFP3 transgenic line originate from the cortical and epidermal tissues of the root and were selected on the basis of their large diameters ( Maathuis & Sanders 1995).
K+currents in protoplasts from stelar and cortical cells in the roots of the mGFP3 transgenic line
For patch–clamp studies, it is important to have a seal resistance in excess of 1 GΩ between the patch pipette and the plasma membrane. Obtaining giga-ohm seals with plant protoplasts can be problematic and may be affected by a large number of unknown factors. However, the presence of large amounts of GFP in the cytoplasm of fluorescent protoplasts did not appear to have a detrimental effect on sealing. We found no significant difference in the success rates of sealing to protoplasts from cells expressing GFP compared to non-fluorescent cells from the same plant (30–40%), and a comparison between the success rates of sealing to protoplasts from roots of the mGFP3 transgenic line and those from wild-type roots also showed no significant difference.
The dominant inward and outward K+ currents in protoplasts from Arabidopsis roots have been characterized previously ( Maathuis & Sanders 1995, 1997). These studies were performed on large protoplasts (> 20 μm diameter) and are therefore thought (see below) to represent channel activities in the cortical/epidermal cells. A time-dependent, outward-rectifying K+ channel (ORC) with a unitary conductance of 15 ± 2.4 pS (in 10/10 m m[K+]in/[K+]out) and 38 ± 4.7 pS (in 100/10 m m[K+]in/[K+]out) was observed in most protoplasts from wild-type roots (approximately 65%). The ORC exhibited flickery kinetics at deactivating membrane potentials and had a low but significant open probability at hyperpolarizing potentials, allowing it to mediate inward K+ current under these ionic conditions. The dominant, time-dependent, inward-rectifying K+ channel (IRC) had a unitary conductance of 5 ± 1.7 pS (in 10/10 m m[K+]in/[K+]out) and 6 ± 2.3 pS (in 100/10 m m[K+]in/[K+]out) and was less frequently observed. This channel strictly catalysed an inward K+ current.
Both cortical/epidermal protoplasts and stelar protoplasts from the roots of the mGFP3 transgenic line exhibited ORC-and IRC-mediated K+ currents similar to those recorded in protoplasts from root cortical cells in wild-type Arabidopsis ( Figs 3 and 4). The time courses of activation of both the inward and the outward currents ( Fig. 3) were comparable to those observed in protoplasts from wild-type roots ( Maathuis & Sanders 1995, 1997), with time constants ranging from 200 to 500 msec. The voltage dependencies of neither the IRC nor the ORC open probabilities (e.g. Figure 4) differed significantly between protoplasts from cortical/epidermal cells and stelar cells (results not shown). An equivalent gating charge of about 2 for the ORC and 0.7 for the IRC were obtained. These values agree well with those reported for wild-type Arabidopsis ( Maathuis & Sanders 1995). In protoplasts from transgenic plants, the unitary conductances of the ORC and IRC were 15.5 ± 2.8 pS (n = 15) and 5.7 ± 2.0 pS (n = 18) in 10/10 m m[K+]in/[K+]out, respectively ( Fig. 4). These values are comparable to the unitary conductances previously estimated for ORC and IRC in protoplasts from wild-type plants ( Maathuis & Sanders 1995, 1997).
A large number of cells were surveyed for ORC and IRC activity, using the cell-attached configuration to retain cells in a physiological state. There appeared to be a marked difference in the relative frequency with which the inward and outward K+ currents were observed in non-fluorescent and fluorescent protoplasts from roots of the mGFP3 transgenic line ( Table 1). Outward K+ currents were observed more frequently than inward K+ currents in cortical/epidermal protoplasts. However, stelar protoplasts from roots of the mGFP3 transgenic line showed a significant (P < 0.001) reversal of this pattern and displayed inward K+ currents more frequently than outward K+ currents.
Table 1. Occurrence of the dominant, time-dependent, inward (IRC) and outward (ORC) rectifying K+ channels in protoplasts from roots of wild-type Arabidopsis and the transgenic mGFP3 line. Results are expressed as a percentage of the total population of protoplasts where IRC or ORC type currents were observed in the cell-attached patch–clamp configuration, with the total number of protoplasts in parentheses
Average diameter (μm)
a,b,c Superscripts with different letters denote significant differences at the 5% level.
The relative prevalence of inward K+ currents in protoplasts from stelar cells as compared to cortical/epidermal cells was also reflected in the excised-patch ( Fig. 4) and whole-cell configurations ( Fig. 3). In extreme cases this resulted in a virtual absence of inward current in cortical/epidermal cells ( Fig. 3a), whereas the reverse was observed in stelar cells ( Fig. 3c).
K+ currents in protoplasts from stelar and cortical cells in the roots of wild-type plants
High levels of GFP expression might alter the activity of membrane transport systems directly. Parallel experiments were therefore performed to study the occurrence of time-dependent IRC and ORC in size-selected protoplasts from roots of wild-type plants ( Table 1). It was assumed that smaller protoplasts originated from the stele and that larger protoplasts were of cortical and epidermal cells ( Maathuis & Sanders 1995). As was observed for the mGFP3 transgenic line, the ratio of IRC/ORC activity was significantly (P < 0.05) higher in small ‘stelar’ protoplasts than in large ‘non-stelar’ protoplasts from roots of wild-type plants. Furthermore, there were no significant differences between wild-type and transgenic plants in the occurrence of IRC or ORC K+ currents in protoplasts from specific cell types. These observations indicate clearly that the presence of GFP had no effect on the expression of K+ channel activities.
A major objective in the study of ion transport phenomena is to integrate knowledge of molecular mechanisms and cellular details into an understanding of the physiological functioning of a multi-cellular organism. To realise this objective, identification of the cellular location, and physiological regulation, of particular transport processes within complex tissue(s) is of absolute importance.
In principle, any cell type for which there exists a specific promoter can be identified unequivocally from the expression of a marker gene. Here GFP was used as a marker gene and it proved ideal for identifying cell types for patch–clamp studies: protoplasts from cells expressing GFP could be identified using a fluorescence microscope and targeted for subsequent patch–clamp studies. Furthermore, the presence of GFP did not affect the occurrence or biophysical characteristics (activation kinetics, voltage dependence of the channel open probability or unitary conductance) of the channels studied. The utility of the cell-marking approach to patch–clamp studies is exemplified by our comparison of K+ currents in cortical and stelar cells of Arabidopsis roots.
Comparison of K+ currents in protoplasts from stelar and cortical cells
It is generally accepted that the uptake of K+ by plant roots is confined to the cortical and epidermal cells, that subsequent radial transport of K+ occurs through the root symplasm, and that K+ is released into xylem vessels via the stelar apoplast in a process mediated by the xylem parenchyma cells ( Marschner 1995). It has been suggested that the contrasting functions of cortical and stelar cells might necessitate contrasting complements of K+ channels ( Roberts 1998;Roberts & Tester 1995, 1997;White 1997). To investigate this hypothesis, transgenic Arabidopsis plants were generated which expressed GFP in a tissue-specific manner under the control of the CaMV 35S promoter (mGFP3). This allowed fluorescent (stelar) and non-fluorescent (cortical/epidermal) protoplasts to be identified for patch–clamp studies.
In protoplasts from cortical/epidermal cells of wild-type Arabidopsis roots, ORC were observed more frequently than IRC (53% versus 35% of cells, respectively;Table 1). These data are somewhat different from those previously reported (65% and 25% for ORC and IRC, respectively, see Maathuis & Sanders 1995, 1997), and may reflect the use of a different ecotype (Wassilewskija versus Columbia) and/or slightly different experimental conditions such as the routine inclusion of ATP in the cytoplasmic medium.
A comparison between transgenic plants expressing GFP and wild-type plants showed that there was no significant difference between (i) the occurrence of IRC and ORC activity in cortical/epidermal root protoplasts derived from either transgenic plants or wild-type plants, and (ii) the occurrence of IRC and ORC activity in stelar protoplasts from either transgenic plants or wild-type plants.
Interestingly, the dominance of ORC-mediated K+ currents in cortical/epidermal cells is lost in cells derived from stelar tissue, where IRC activity was observed as frequently (wild-type) or more frequently (mGFP3 plants) than ORC activity ( Table 1). This observation was made irrespective of whether protoplasts were from transgenic or wild-type plants, indicating that it was not the result of GFP expression itself. Clearly the stele is composed of several different cell types. However, it is not possible at present to determine whether the increased occurrence of IRC is a property of all stelar cells or a restricted subset of stelar cell types. The IRC and ORC single-channel characteristics such as unitary conductance, voltage dependence and time dependence were similar in stelar and epidermal/cortical protoplasts, suggesting that the same types of channel function in these different tissues.
It is unclear how the different IRC/ORC ratios of stelar and cortical/epidermal cells reflect the physiological function associated with each tissue. A high prevalence of inward current in the stelar cells was previously found in xylem parenchyma cells of barley ( Wegner & Raschke 1994), and in that study it was suggested that inward-rectifying channels may play a role in resorption of ions from the xylem. Conversely, in roots of well-watered maize plants, inward currents were dominant in cortical cells and almost absent in stelar cells, which frequently contained outward currents ( Roberts & Tester 1995). Results from the latter study were interpreted as reflecting the K+ uptake function of the root cortex and the K+ release function of the stele. However, the relative activities of inward and outward K+ currents were later found to depend critically on the water status of the maize plants ( Roberts 1998).
In the stelar cells of Arabidopsis roots, ORC may be involved in the regulation of membrane voltage and/or cation release and the IRC may play a role in K+ or cation resorption from the xylem. Since a large amount of K+ is continuously recycled from the shoot to the root through the phloem ( Marschner 1995), and phloem unloading is potentially apoplastic ( van Bel 1993), an additional function of IRC in stelar cells might be to facilitate re-entry of phloem-released K+ into the symplast. Transgenic lines expressing GFP under the control of promoters with a higher degree of cell specificity within the stele will allow us to examine the role of contrasting membrane transporters in different stelar cell-types.
Exploiting the potential of marked cells for physiological studies
Cell marking is useful for the genetic, biochemical and physiological characterization of single cells. The utility of cell marking in patch–clamp studies comparing the ion transport properties in contrasting cell types provides one such example.
The cell-marking approach, in combination with patch–clamp electrophysiology, could be extended further to investigate how ion transport processes respond to developmental or environmental stimuli. For example, ionic fluxes involved in developmental processes such as lateral root initiation can be studied in protoplasts expressing GFP under the control of cell-specific and developmental stage-specific promoters ( Malamy & Benfey 1997). As a second example, cell types in the elongation zone which have differential sensitivities (and responses) to auxin during gravitropism ( Ishikawa & Evans 1995) can be distinguished in Arabidopsis mutants produced by promoter trapping ( Masson et al. 1993 ), allowing the ionic currents associated with gravitropic responses in each cell type to be investigated.
The technique of cell marking can also be used to target cells for the production of single-cell ( Karrer et al. 1995 ) or tissue-typed cDNA libraries. Tissue-typed cDNA libraries might be obtained by separating fluorescent cells by flow or imaging cytometry using a fluorescence-activated cell sorter ( Sheen et al. 1995 ). Such libraries could be used to obtain rare or cell-specific cDNA and would allow the characterization of genes involved in cell-specific biological processes. Appropriate cells to target for transport-related phenomena could be identified from prior patch–clamp studies of marked cells.
Finally, it should be noted that there is the potential to identify any cell type for analysis, provided a suitable cell-specific or physiology-specific promoter can be obtained. Many such promoters are already available and more will become available as extensive promoter- or enhancer-trap collections are assembled. Recently, a collection of Arabidopsis lines with distinct and stable patterns of GFP expression in the root has been generated by Dr J. Haseloff (MRC Laboratory of Molecular Biology, Cambridge, UK; personal communication). The cell-marking technique described here is not restricted to studies using Arabidopsis, but may be applied to any plant species which can be successfully transformed genetically.
The plasmid pBIN m-gfp5-ER ( Haseloff et al. 1997 ) was obtained from Dr K. Siemering (MRC Laboratory of Molecular Biology, Cambridge, UK). It was transformed into Agrobacterium tumefaciens C58 pGV3850 by electroporation ( Wen-jun & Forde 1989), and subsequently into Arabidopsis thaliana (ecotype Wassilewskija) by vacuum infiltration ( Bechtold et al. 1993 ). The infiltration procedure was modified from the original method. Our experience shows that incorporation of sucrose in the infiltration medium encourages fungal contamination of the plants and does not appear to enhance the infiltration efficiency. Similarly, neither inclusion of hormones nor centrifugation of the Agrobacterium culture and re-suspension in MS medium ( Murashige & Skoog 1962) appear to be necessary for successful transformation. Our standard protocol is therefore to vacuum infiltrate a 200 ml turbid overnight culture of Agrobacterium in LB (or 2 × YT) broth under sufficient vacuum to cause Arabidopsis thaliana inflorescences to exude bubbles (normally 700–750 mmHg). Inclusion of antibiotics in the broth did not appear either to inhibit or to enhance the transformation frequency (data not shown).
Selection of transformants
Surface-sterilized seeds (NaOCl, 1% active chlorine; rinsed in 70% (v/v) ethanol/water and air-dried) from pBIN m-gfp5-ER-infiltrated plants, were sown on MS medium with 1% (w/v) sucrose, 1% (w/v) agar and 50 mg l–1 kanamycin sulphate (Sigma), then stratified at 4°C in the dark for 48 h and grown at 25°C under continuous light. Plants were selected after production of secondary leaves and transplanted onto soil. Ten primary transformants from the pBIN m-gfp5-ER infiltration were driven to homozygosity via selection through three generations and observation of correlated GFP expression. One homozygous line (designated mGFP3) was used for subsequent analysis.
Protoplast isolation and patch–clamp electrophysiology
Seeds of mGFP3 were germinated in soil. Seedlings were transferred to plastic containers 10–12 days after germination and grown hydroponically on quarter-strength MS medium ( Maathuis & Sanders 1995).
Protoplasts were generated from roots of 4–6-week-old plants as described previously ( Maathuis & Sanders 1995). In both intact roots and protoplasts, GFP fluorescence was detected using a fluorescence microscope equipped with an excitation filter of 460–490 nm and an emission filter at 520 nm. None, or very little, autofluorescence was detectable with these filter settings. After digestion, fluorescent and non-fluorescent protoplasts were separated and collected individually using a suction pipette. Fluorescent protoplasts, which were derived from stelar cells of the mGFP3 transgenic line, and non-fluorescent protoplasts with large diameters, which were derived from cortical/epidermal cells ( Maathuis & Sanders 1995), were transferred to the patch–clamp rig.
Electrophysiological studies were performed as described by Maathuis & Sanders (1995). The standard extracellular solution contained 10 m m KCl, 2 m m CaCl2, 2 m m MgCl2, 3 m m MES/Tris, pH 5.5. The standard solution facing the cytoplasmic side contained either 10 m m or 100 m m KCl, 2 m m CaCl2 buffered with EGTA (free Ca2+ 200–400 n m), 2 m m MgCl2, 2 m m ATP and 2 m m MES/Tris, pH 7.5. All solutions were adjusted with mannitol to 500 mOsm l–1 and filtered (0.2 μm pore size) prior to use.
Hand-cut root sections from 4–6-week-old plants were placed in a drop of water onto a microscope slide. Confocal fluorescence imaging was performed with a laser scanning confocal microscope (Bio-Rad MCR-1000 LSCM) controlled using CoMOS software (Version 7.0a). The confocal system interfaced with a Nikon Optiphot microscope. The 488 nm line of an argon laser (3% intensity) in conjunction with a 527 DRLP excitation filter was used for image acquisition. Fluorescence images were collected using a 522/35 nm bandwidth filter. No fluorescence was detected from wild-type root sections. Bright-field images were collected via a transmission device which gathered transmitted laser light and delivered it directly to the dedicated photomultiplier detector.
We thank Professor K. Lindsey (University of Durham, UK) for stimulating our interest in cell-marking techniques, and Dr K. Siemering and Dr J. Haseloff (MRC Laboratory of Molecular Biology, Cambridge, UK) for the gift of plasmids. S.T.M., N.G., H.C.B., M.J.B. and P.J.W. were funded by the Biotechnology and Biological Sciences Research Council (UK). F.J.M.M. and T.J. were supported by BBSRC grant 87/P04043. P.B.T. holds a Horticulture Research International/University of Warwick postgraduate studentship.