Transgenic potato (Solanum tuberosum) plants simultaneously over-expressing a pea (Pisum sativum) glucose-6-phosphate/phosphate translocator (GPT) and an Arabidopsis thaliana adenylate translocator (NTT1) in tubers were generated. Double transformants exhibited an enhanced tuber yield of up to 19%, concomitant with an additional increased starch content of up to 28%, compared with control plants. The total starch content produced in tubers per plant was calculated to be increased by up to 44% in double transformants relative to the wild-type. Single over-expression of either gene had no effect on tuber starch content or tuber yield, suggesting that starch formation within amyloplasts is co-limited by the import of energy and the supply of carbon skeletons. As total adenosine diphosphate-glucose pyrophosphorylase and starch synthase activities remained unchanged in double transformants relative to the wild-type, they cannot account for the increased starch content found in tubers of double transformants. Rather, an optimized supply of amyloplasts with adenosine triphosphate and glucose-6-phosphate seems to favour increased starch synthesis, resulting in plants with increased starch content and yield of tubers.
Potato plants are one of the most important crops in the world. They are grown in many countries, and their tubers are a source of human nutrition in almost all countries of the world. The major constituent of potato tubers is starch, which makes up 80% of the tuber dry weight (Kruger, 1997). In addition to its function in food production, starch forms the basis for a wide variety of industrial applications. About 600 food and non-food products have been reported to rely on starch produced in plants (Ball et al., 1996). Plant breeding and applied biotechnological research have developed in two main directions in recent years: (i) methods to increase the yield of the usable parts of plants, i.e. potato tubers (Stark et al., 1992; Sweetlove et al., 1996a,b; Tjaden et al., 1998; Regierer et al., 2002; Jenner, 2003); and (ii) techniques to tailor the composition of starch, which consists of about 80% branched amylopectin and 20% helical amylose. Using transgenic approaches, starches with modified compositions, i.e. alterations in the amylose to amylopectin ratio, have been achieved, and starch that lacks amylose completely has been produced (Jobling, 2004).
In this study, in order to achieve an increase in tuber yield, focus was placed on the enhancement of the sink strength of tubers. To this end, the availability of substrates for adenosine diphosphate (ADP)-glucose pyrophosphorylase (AGPase), the key enzyme of starch biosynthesis, was increased within amyloplasts, the exclusive location of the formation of ADP-glucose in potato tubers (Preiss, 1991; Neuhaus et al., 2005). For net starch production, both glucose-6-phosphate (Glc-6-P), the precursor of the AGPase substrate glucose-1-phosphate (Glc-1-P), and adenosine triphosphate (ATP) need to be imported into the stroma. These import processes are mediated by two metabolite transporters, the glucose-6-phosphate/phosphate translocator (GPT) and the adenylate translocator (nucleotide translocator, NTT), both residing in the inner envelope membrane of amyloplasts.
GPTs represent a subfamily of the phosphate translocator (PT) family (Fischer and Weber, 2002); the other subfamilies characterized to date consist of the triose phosphate/phosphate translocators (TPTs) (Flügge et al., 1989; Flügge, 1999), the phosphoenolpyruvate/phosphate translocators (PPTs) (Fischer et al., 1997) and the xylulose-5-phosphate/phosphate translocators (XPTs) (Eicks et al., 2002). The PT family itself belongs to the nucleotide sugar transporter (NST)/TPT superfamily (Knappe et al., 2003), and its members mediate an antiport of phosphorylated intermediates with inorganic phosphate across the plastid inner envelope membrane (reviewed in Fischer and Weber, 2002). Substrate specificities for GPTs have been determined in detail for a GPT from pea roots (Kammerer et al., 1998). In addition to Glc-6-P and inorganic phosphate, triose phosphates and, to a lesser extent, 3-phosphoglycerate (3-PGA) serve as transport substrates. In chloroplasts, the main metabolite transporter is a TPT, which exports triose phosphates during the light period as precursors for sucrose biosynthesis in the cytosol. Plastids of non-green tissues, e.g. amyloplasts of potato tubers, rely on a different strategy of metabolite exchange. They have to import carbon skeletons in the form of hexose phosphates rather than triose phosphates. This is because: (i) plastidic fructose-1,6-bisphosphatase (FBPase) activity is absent in amyloplasts (Entwistle and ap Rees, 1990; Neuhaus et al., 1993); and (ii) a TPT, which could deliver triose phosphates is, if at all, only weakly expressed in heterotrophic tissues (Schulz et al., 1993). Amyloplasts import hexose phosphates in the form of Glc-6-P via a GPT in counterexchange with either inorganic phosphate (derived from starch biosynthesis, i.e. from the cleavage of pyrophosphate generated by AGPase) or triose phosphates (the end-products of the oxidative pentose phosphate pathway; Kammerer et al. 1998).
In order to analyse the impact of a modified supply of energy and carbon skeletons in amyloplasts on tuber starch formation and yield, both GPT and NTT were over-expressed, either individually or in combination, under the control of the B33 patatin promoter. This promoter is mainly active in tubers, and thus minimizes or avoids unwanted effects on aerial parts of the plants (Rocha-Sosa et al., 1989).
As both ATP and carbon skeletons are required for the formation of storage starch in potato tubers, the consequences of a single or double over-expression of GPT (PsGPT, pea GPT gene expressed under the control of the B33 promoter) and NTT (AtNTT1, Arabidopsis NTT1 gene expressed under the control of the B33 promoter) on tuber yield, starch content and total tuber starch yield per plant were analysed. The aerial parts of all transgenic lines generated in this study lacked any particular phenotype and appeared to be indistinguishable from the wild-type.
Over-expression of GPT
Potato plants (Solanum tuberosum cv. Desirée) were transformed with the B33::PsGPT construct. From the resulting transformants, three lines (BG1, BG6 and BG12) were selected, which significantly over-expressed the PsGPT gene in tubers. As indicated by Northern blot analysis (Figure 1a), PsGPT-specific transcripts were highly abundant in BG1 and BG6, detectable in BG12 and absent in the wild-type. Moreover, determination of the transport activities after reconstitution of the membrane proteins in proteoliposomes supported the general pattern of PsGPT transcript abundance (Figure 1c). Using inorganic phosphate, Glc-6-P and 3-PGA as counterexchange substrates, line BG1 exhibited the greatest increase in Glc-6-P-specific transport rates (1.8-fold compared with the wild-type), followed by BG6 with a moderate increase in Glc-6-P transport activity. The transport rates of line BG12 were indistinguishable from those of the wild-type.
Over-expression of NTT
Furthermore, potato plants (S. tuberosum cv. Desirée) were transformed with the B33::AtNTT1 construct. The resulting transgenic lines BA8 and BA24 displayed a high abundance of AtNTT1 transcripts in tubers, with line BA12 showing a lower abundance. Moreover, slight cross-hybridization was observed, most probably with an endogenous NTT detectable in wild-type tubers (Figure 1b). As revealed by reconstitution experiments, the NTT activity was increased significantly in transgenic tubers (Figure 1d). Transgenic tubers displayed a 50% (line BA12) to 100% (line BA8) increase in NTT activity.
Simultaneous over-expression of GPT and NTT
Over-expression of the B33::AtNTT1 construct in the GPT over-expressor background (line BG1) also led to an increase in NTT transcript levels and NTT transport activities in tubers (Figure 2a,b). Each of the three double transformants, BGA24, BGA-31 and BGA-32, exhibited an almost two-fold increase in NTT activity in tubers compared with the wild-type; the increase is therefore in a similar range as in single NTT over-expressors. The second transformation had no effect on the level of GPT over-expression, which remained high in the double transformants (Figure 2c).
Tuber yield and starch content are strongly increased in lines over-expressing both GPT and NTT
In order to assess the impact of PsGPT and/or AtNTT1 expression on the biosynthetic performance of transgenic potato plants, the tuber yield, starch content of tubers and total tuber starch yield per plant were analysed in replicated harvests.
Individual over-expression of either PsGPT (Figure 3a–c) or AtNTT1 (Figure 3d–f) had no detectable effect on tuber yield or starch content. In contrast, simultaneous over-expression of both translocators led to an increased tuber yield and starch content (Figure 3g–i). Tuber yield was increased in transgenic lines by 10% (BGA24), 15% (BGA32) and 19% (BGA31). Moreover, the starch content of tubers [expressed as micromoles of C6 unit per gram fresh weight (FW)] was increased by 20% (BGA32), 23% (BGA31) and 28% (BGA24). This resulted in overall increases in starch yield, expressed as tuber starch per plant, of 33% (BGA32), 43% (BGA24) and 44% (BGA31) compared with the wild-type. Furthermore, the question was raised as to whether the increase in tuber and starch yield observed in tubers of 10-week-old double transformants occurred continuously during tuber development. Data gathered from independent experiments obtained a similar proportional change in the above parameters when they were harvested from 8- and 12-week-old plants (Table 1). The absolute values of tuber starch content, tuber yield and total tuber starch yield of 8-week-old wild-type plants were 745.3 ± 52.0 µmol C6 unit/g FW (SE, n = 4), 139.4 ± 7.2 g FW/plant (SE, n = 5) and 108.9 ± 11.6 mmol/plant (SE, n = 4), respectively. The corresponding values of 12-week-old wild-type plants were 821.0 ± 85.5 µmol C6 unit/g FW (SE, n = 4), 179.5 ± 6.8 g FW/plant (SE, n = 4) and 140.2 ± 14.2 mmol/plant (SE, n = 3), respectively. The 12-week-old wild-type and transgenic plants were fully mature. It was therefore concluded that both over-expressed genes exert an effect on tuber starch content and yield throughout plant development.
Table 1. Tuber starch content and yield of transgenic lines as a percentage of the wild-type (WT)
% of WT
Tubers of the independent experiments 1, 2 and 3 were harvested after 12, 10 and 8 weeks, respectively. The results of experiment 2 were calculated from the data also displayed in Figure 3. Experiments 1, 3 and BG and BA plants of experiment 2, n = 3–5; WT and BGA plants of experiment 2, n = 6–10. *P < 0.1,†P < 0.05,‡P < 0.02,§P < 0.01.
These findings strongly suggest that starch formation within amyloplasts is co-limited by the import of energy (ATP) and the supply of carbon skeletons (Glc-6-P).
Notably, the tuber sugar (glucose, fructose and sucrose) and sugar phosphate (Glc-6-P, fructose-6-phosphate) contents of the transformants remained unchanged relative to those of wild-type plants (Figures S1 and S2, see ‘Supplementary material’).
Tuber starch composition is altered in lines over-expressing both GPT and NTT
The double transformants over-expressing PsGPT in combination with AtNTT1 were analysed for their tuber starch composition. The proportion of amylose in tuber starch was increased from 19.4% in the wild-type to 21.6% (BGA24), 22.2% (BGA31) and 22.6% (BGA32) in the double transformants (Figure 4). This slightly increased amylose content of tuber starch indicates an improved availability of ADP-glucose for starch synthases, in particular for granule-bound starch synthases (GBSSs), which are responsible for amylose synthesis (Visser et al., 1991; Kuipers et al., 1994). GBSS isoforms have a lower affinity than soluble starch synthases (which are involved in the synthesis of amylopectin) for ADP-glucose (Smith et al., 1997), i.e. an increase in the amylose to amylopectin ratio can be explained by an increased ADP-glucose level or availability (Tjaden et al., 1998).
Adenylate contents are unchanged in lines over-expressing both GPT and NTT
Increased starch accumulation may be caused by the increased availability of ADP-glucose as the substrate for starch synthases. In order to test this assumption, the ADP-glucose content was determined in the double transformants in comparison with the wild-type. There was some variation in the pool sizes of ADP-glucose, but no significant differences could be detected between the transgenic lines and wild-type (Figure 5a). Furthermore, the transport substrates of NTT, ADP (Figure 5b) and ATP (Figure 5c) were analysed. Similar to ADP-glucose, none of the transgenic lines displayed a significant difference in ATP or ADP content compared with the wild-type.
Taken together, adenylate contents cannot be held responsible for the increased tuber starch content and yield of double transformants. This finding does not necessarily exclude ADP-glucose availability from playing an important role in starch synthesis.
AGPase and starch synthase activities remain unaffected in lines over-expressing both GPT and NTT
An increased availability of ADP-glucose may be brought about by a higher activity of AGPase. However, total extractable AGPase activity was unaltered in the double transformants (Figure 6a), indicating that the amount of AGPase protein remains unaffected by increased activities of both transporters. These data suggest that substrate availability for AGPase apparently limits ADP-glucose formation in the wild-type, and that this limitation can be at least partially overcome in the double transformants.
To rule out a possible pleiotropic effect of increased starch synthase activity as a cause for the higher tuber starch content and yield of the double transformants, total starch synthase activity was analysed and was shown to be unchanged compared with the wild-type (Figure 6b). Thus, unchanged amyloplast enzyme activities of starch synthesis argue against pleiotropic effects as the cause of the increased tuber starch content and yield of double transformants.
Our data indicate an increased flux of carbon and energy into tuber starch, mediated by GPT and NTT, without affecting the pool sizes of the respective transport substrates or ADP-glucose. The double transformants manage to do this without increasing total AGPase or starch synthase activities.
Amyloplasts of potato tubers rely on the import of carbon precursors, mediated by GPTs, as substrates for both starch biosynthesis and the oxidative pentose phosphate pathway (which delivers redox equivalents for a number of biosynthetic processes, e.g. fatty acid biosynthesis) (Martin and Ludewig, 2007). They also differ from chloroplasts in that they rely on the supply of ATP via NTT (Schünemann et al., 1993; Neuhaus et al., 1997), because a full glycolytic pathway starting from Glc-6-P, and yielding ATP and pyruvate, appears either not to be installed in amyloplasts or proceeds too slowly for a significant contribution to starch biosynthesis. The importance of ATP supply to amyloplasts has been demonstrated previously by antisense inhibition of a potato NTT (Tjaden et al., 1998). In these transgenic lines, tuber yield and starch contents were decreased. Moreover, tuber morphology (Tjaden et al., 1998) and the structure of starch granules (Geigenberger et al., 2001) were altered. However, an increase in NTT activity in transgenic potato plants over-expressing the Arabidopsis NTT1 gene delivered inconsistent data with respect to starch contents (Tjaden et al., 1998). A non-significant (0.05 < P < 0.1) increase in starch content was observed in only one of three potato lines. Furthermore, in this particular line, the starch content was increased at the expense of the tuber yield. Calculating the total tuber starch content per plant using the mean values of the tuber FW per plant and the amount of starch per gram FW, the three lines displayed 117%, 75% and 80% total tuber starch compared with the wild-type, i.e. the starch yield per plant was not consistently increased in these transgenic lines (all data were calculated from Tjaden et al., 1998). For NTT over-expression, the authors used the cauliflower mosaic virus (CaMV) 35S promoter, which may perturb metabolism in the aerial parts of the plants. To avoid such a complication, we used the B33 patatin promoter, which is mainly active in tubers (Rocha-Sosa et al., 1989), to over-express both a pea GPT gene and the Arabidopsis NTT1 gene, either individually or in combination. Tuber-specific expression of either PsGPT or AtNTT1 had no marked effect on tuber yield or tuber starch content in single over-expressing lines (Figure 3a–f).
Strikingly, the combination of both approaches, i.e. the expression of AtNTT1 in the PsGPT over-expressor background, led to increased tuber yield and starch content (Figure 3g–i). This finding demonstrates that, in wild-type plants, both transporters share control of the starch content and tuber yield, most probably by limiting substrate supply (i.e. Glc-6-P or ATP) for AGPase. Hence, a decrease in either transport activity should result in a decrease in starch content and/or tuber yield as a result of substrate limitation. Indeed, antisense plants with reduced NTT activity are strongly impaired in tuber starch content and yield (Tjaden et al., 1998). However, it remains to be elucidated whether lower GPT activities in potato tubers have a similar effect. Moreover, there is evidence of the importance of GPT in starch synthesis. In Arabidopsis thaliana, the absence of one of two GPTs (AtGPT1) resulted in the lethality of gametophytes, presumably as a result of a restriction of the oxidative pentose phosphate pathway by limited Glc-6-P supply (Niewiadomski et al., 2005). For instance, in comparison with wild-type pollen, mutant pollen grains contain much less lipid bodies, and lack starch granules completely. Moreover, in seeds of antisense GPT Vicia narbonensis plants, the starch content is reduced (Rolletschek et al., 2007).
It has been reported that an increase in tuber yield and starch content may also be brought about by manipulation of the adenylate pools (Regierer et al., 2002) or de novo pyrimidine synthesis (Geigenberger et al., 2005). Our results imply that tuber starch content and yield increase as a consequence of an enhanced availability of both AGPase substrates, i.e. ATP and hexose phosphates. In the case of adenylate kinase antisense potato plants (Regierer et al., 2002), an increased availability of ATP in the stroma as a consequence of decreased adenylate kinase activity has been proposed. To date, however, there is no information on the altered supply of carbon skeletons in these lines. Thus, it is conceivable that a pleiotropic up-regulation of a GPT and concomitant enhanced hexose phosphate supply, together with increased ATP availability, may be responsible for the improved yield, an assumption waiting to be tested. In the case of the inhibition of de novo pyrimidine synthesis in potato tubers, the authors proposed a compensatory stimulation of the pyrimidine salvage pathway, which, in turn, may lead to an increase in biosynthetic performance (Geigenberger et al., 2005). Currently, however, there is no obvious explanation for the improved AGPase substrate availability, and a pleiotropic up-regulation of GPT and NTT cannot be ruled out. Furthermore, the starch contents of the wild-type tubers analysed in both publications were, on average, quite low. The starch contents per tuber FW of 8.6% and 3.6% reported by Regierer et al. (2002) and Geigenberger et al. (2005), respectively, are lower than the values obtained in this report (13.8%) and by others (average: 13.8% ± 2.5% SD; see Sonnewald, 1992; Zrenner et al., 1993; Abel et al., 1996; Tjaden et al., 1998; Lloyd et al., 1999; Sweetlove et al., 1999, 2001; Veramendi et al., 1999; Kauder et al., 2000; Tiessen et al., 2002, 2003; Hajirezaei et al., 2003; Leggewie et al., 2003; Lytovchenko et al., 2005; Skirycz et al., 2005). The featured high-starch tubers contain 10.4%–13.7% starch (Regierer et al., 2002), values that hardly reach wild-type levels in our and other analyses. However, data sampled from a field trial in the same publication revealed 11.9% starch for wild-type and 13.8%–14.1% starch for adenylate kinase antisense tubers, values that better meet the expectations. Taken together, the increased biosynthetic performance of the antisense adenylate kinase plants seems to be mainly a result of an increased tuber yield than of an increased starch content. The same holds true for antisense uridine monophosphate synthase plants. ‘High-starch’ antisense tubers contain 4.1%–4.9% starch (Geigenberger et al., 2005), values that fit quite well with antisense AGPase plants with a low starch content (2.6%–7.8%; Sweetlove et al., 1999). This conflict of data may be explained by the extremely different growth conditions; however, differences in the extraction or determination of starch may also be considered. An even more dramatic example of conflicting data concerning potato tuber starch contents has been published by McKibbin et al. (2006) and Sweetlove et al. (1996b). McKibbin et al. (2006) featured ‘high-starch’ potato tubers that contained 2.8% starch per gram FW (compared with 2.0% for the corresponding wild-type), whereas Sweetlove et al. (1996b) published 10.7% starch-containing, β-glucuronidase (GUS)-expressing tubers of the same cultivar in a different publication. The increase in starch content was brought about by an over-expression of SnRK1 (McKibbin et al., 2006). The authors discussed a role for SnRK1 in the regulation of carbohydrate metabolism and partitioning via an activation of sucrose synthase and AGPase.
An approach aimed directly at an increase in starch biosynthesis in potato tubers has been reported previously (Stark et al., 1992). The most obvious concept behind this approach is the over-expression of the key enzyme of starch biosynthesis, AGPase, in potato plants. For this purpose, Stark et al. (1992) used a mutant AGPase of Escherichia coli (glgC16), which is, unlike the enzyme of higher plants, not allosterically regulated, and reported an increase in tuber starch content in AGPase over-expressing lines. However, this finding could not be corroborated by Sweetlove et al. (1996a,b), who used a similar approach, albeit in the background of a different potato cultivar. The reported increase in tuber AGPase activity had no lasting effect on tuber starch content, because starch degradation was also increased, i.e. the beneficial effects of increasing the flux through AGPase were reversed by enhanced starch turnover. The two analyses were carried out with the two potato cultivars Russett Burbank and Prairie, respectively. However, our data sampled from a third potato cultivar (Desirée) strongly support the proposal that a shortage of substrates (i.e. Glc-6-P and ATP), in addition to AGPase activity, may control tuber yield and starch content. Moreover, similar to the chloroplast enzyme, amyloplast AGPase is subject to regulation by thioredoxin (Tiessen et al., 2002), and can therefore sense the redox status of the amyloplast stroma. This feedback control mechanism allows an adaptation of the fluxes of hexose phosphates into the direction of the oxidative pentose phosphate pathway or starch biosynthesis. There is increasing evidence emerging on redox-regulated target enzymes in amyloplasts (Balmer et al., 2006), including α-glucan, water dikinase and probably α-1,4-glucan phosphorylase. Such regulatory loops also need to be considered when interpreting transgenic approaches that aim to increase the biomass and starch production in potato tubers. In the case of our approach, the data demonstrate that an increased flux into starch synthesis may be achieved by a concomitant increased supply of both AGPase substrates: Glc-1-P via Glc-6-P and ATP. A possible change in the redox activation state of AGPase in the double transgenic plants cannot be ruled out.
The combined over-expression of GPT and NTT also caused changes in tuber starch composition, such that the content of amylose relative to amylopectin was increased. Higher amylose contents may indicate an increased availability of ADP-glucose as substrate for starch synthases. In particular, GBSS isoforms, which synthesize amylose (Visser et al., 1991; Kuipers et al., 1994), have a lower affinity than soluble isoforms (which are involved in the synthesis of amylopectin) for ADP-glucose (Smith et al., 1997). For Arabidopsis leaves, Zeeman et al. (2002) were able to positively correlate high starch with high amylose contents in both wild-type plants of different ages and in mutants that accumulate starch. In addition, they found that the GBSS activity was increased in mutant leaves, whereas soluble starch synthase activity remained similar to the wild-type. In potato tubers, Flipse et al. (1996) found a positive correlation of GBSS activity and amylose content of starch up to a certain threshold level. A further increase in GBSS activity did not lead to an increase in amylose content in tuber starch. These findings, taken together with data on starch from other organisms (Van den Koornhuyse et al., 1996; Clarke et al., 1999), indicate that substrate availability is critical for starch synthesis and, in particular, for amylose synthesis.
Our analyses were performed using the high-yield, medium-starch-containing potato cultivar Desirée. It remains to be elucidated whether the breeding process during recent decades or even centuries has accidentally focused on substrate availability for AGPase to yield high-starch cultivars, and whether or not the starch content of high-starch cultivars can be further increased by an approach similar to that reported in this study.
Construction of AtNTT1 and PsGPT
To generate the AtNTT1 construct, the 1872-bp coding sequence of the Arabidopsis adenylate translocator NTT1 gene was cloned between the B33 promoter (Rocha-Sosa et al., 1989) and the ocs (octopine synthase) terminator of plasmid pBinB33(Kan), which was created by exchanging the CaMV 35S promoter of plasmid pBinAR (Höfgen and Willmitzer, 1990) for the B33 patatin promoter. The construct confers kanamycin resistance to transformed cells. The construct PsGPT was generated by cloning the 1623-bp coding sequence of the pea GPT gene between the B33 promoter and the ocs terminator of plasmid pBinB33(Hyg), which was created by cutting the B33 promoter – polylinker –ocs terminator cassette out of pBinB33(Kan) using EcoRI and HindIII, and cloning this fragment into the equally cut plasmid pBIB-Hyg (Becker, 1990). The construct confers hygromycin resistance to transformed cells.
Wild-type and transgenic potato plants were transferred from axenic cultures to soil, and grown under glasshouse conditions from spring to autumn with additional lighting (16 h light/8 h dark) in 3.5 L pots for the indicated time periods. The temperature was kept at 22 °C during the light period and 18 °C during the dark period. The plants perceived an average photon flux density of approximately 250 µmol/m2/s.
For the analysis of adenylate contents, samples were extracted as in Häusler et al. (2000). Nucleotides were detected using a highly sensitive fluorescence method according to Haink and Deussen (2003): adenosine nucleotides were specifically converted into fluorescent etheno-adenosine nucleotides. Prior to high-performance liquid chromatography (HPLC) separation, an aliquot of the samples was derivatized with 10% (v/v) chloracetaldehyde in 62 mm sodium citrate and 76 mm potassium dihydroxide phosphate, pH 5.2. The mixture was incubated for 40 min at 80 °C, cooled immediately on ice, centrifuged at 20 000 g for 1 min and used for HPLC analysis. Separation was carried out with a reversed phase HPLC system (Waters Alliance 2795; Waters, Milford, MA, USA), consisting of a gradient pump, a degassing module, an integrated microtitre plate autosampler and a fluorescence detector. The gradient was accomplished using a buffer containing 5.7 mm tetrabutylammonium bisulphate (TBAS) and 30.5 mm potassium dihydroxide phosphate, pH 5.8 and an eluent containing pure acetonitrile (Roti C Solv HPLC; Roth, Karlsruhe, Germany). A single run was set to 4.5 min, followed by a reconditioning of 3 min with the eluent. The excitation wavelength was set to 280 nm and the emission wavelength to 410 nm. In all cases, chromatograms were integrated using the software package Empower (Waters, Milford, MA, USA).
We would like to thank Iris Schmitz and Wolfgang Jeblick for excellent technical assistance. This work was supported by the Deutsche Forschungsgemeinschaft (Priority Programme 1108).