Guard cells in intact leafs display light-induced membrane potential changes, which alter the direction of K+-transport across the plasma membrane (Roelfsema et al., 2001). A beam of blue light, but not red light, directed at the impaled guard cell triggers this response, while both light qualities induce opening of stomata. To gain insight into this apparent contradiction, we explored the possible interaction between red light and CO2. Guard cells in the intact plant were impaled with double-barrelled electrodes and illuminated with red light. Cells that were hyperpolarized in CO2-free air, depolarized after a switch to air with 700 µl l−1 CO2, in a reversible manner. As a result, K+-fluxes across the plasma membrane changed direction, to favour K+ extrusion and stomatal closure in the presence of CO2. Concurrent with the depolarization, an inward current across the plasma membrane appeared, most likely due to activation of anion channels. Guard cell responses to CO2 could be recorded in darkness as well as in red light. However, in darkness some cells spontaneously depolarized, these cells hyperpolarized again in red light. Here, red light was projected on a large area of the leaf and decreased the intracellular CO2 concentration by about 250 µl l−1, as measured with a miniature CO2 sensor placed in the substomatal cavity. We conclude, that in intact leaves the red light response of guard cells is mediated through a decrease of the intercellular CO2 concentration.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Stomata enable the uptake of CO2 into leaves, to reach the sites of photosynthesis. In accordance with this function, the opening of stomata is triggered by light. However, light is not the only factor affecting stomatal movement. Stomata are sensitive to a number of signals, which are either received from the environment, or produced within the plant itself (Assmann and Shimazaki, 1999). To study signal perception and transduction of guard cells in their natural environment, we have developed a method to record the membrane potential of guard cells in an intact plant (Roelfsema et al., 2001). Using this technique, white light was shown to hyperpolarize the plasma membrane. This change in membrane potential allowed the guard cell to take up K+ via inward rectifying channels. In the dark, guard cells extruded K+ via outward rectifying K+ channels. The light-induced changes in K+-fluxes across the plasma membrane fuel the osmotic motor of guard cells and finally lead to stomatal movement (Raschke et al., 1988).
A plasma membrane response could also be evoked by a small beam of blue light projected on a single guard cell (Roelfsema et al., 2001). Hyperpolarization of the plasma membrane correlated with a blue light-induced outward current across the plasma membrane. The current peaked 3 min after the onset of the blue light, a time course similar to that found for proton extrusion. This indicates that the current is due to the activation of plasma membrane H+-ATPases (Kinoshita and Shimazaki, 1999). A small red light beam, applied in the same way as blue light, did not trigger a response at the plasma membrane (Roelfsema et al., 2001). Patch-clamp experiments with guard cell protoplasts confirmed this red light insensitivity. In the latter experiments blue light also triggered an outward pump-current across the plasma membrane (Assmann et al., 1985), but again red light was ineffective (Taylor and Assmann, 2001).
In contrast to the finding that red light applied to single guard cells is ineffective, transpiration measurements show that stomata open in response to blue as well as red light (Karlsson, 1986). This apparent contradiction may emerge from differences in the way red light illumination was applied. In gas exchange experiments red light was projected on a large area of a leaf, while in electrophysiological studies red light was applied either to a small section of the leaf (Roelfsema et al., 2001), or to a single guard cell (Taylor and Assmann, 2001). This indicates that the guard cell response to red light may be mediated via a signal generated by mesophyll cells, possibly the CO2 concentration of the substomatal cavity.
Here, we study the plasma membrane response of guard cells in intact plants to changes in the atmospheric CO2-concentration. Based on the results obtained with CO2- and red light treatments, we conclude that guard cells do not respond to red light directly, but rather to a red light-induced drop of the CO2 concentration in the intercellular space.
CO2-dependent changes in ion-channel activity
The experimental set-up described by Roelfsema et al. (2001) was adapted, so that guard cells could be exposed to rapid changes in CO2 concentration. Guard cells in intact plants were observed through a long distance (dry) objective and illuminated with red light provided by the microscope lamp (Figure 1a). A stream of air was directed on the stoma of interest, which was either CO2-free or at a CO2 concentration of 700 µl l−1. A glass capillary filled with 50 mm KCl was placed on the cuticle and served as a reference electrode (Figure 1a). The potential difference between the guard cell wall and reference electrode was determined with blunt electrodes brought in contact with the cell wall via an open stoma. The youngest fully expanded leaves, used in our studies, had an average surface potential of 16 mV (sd 9, n = 14). This surface potential was sensitive to changes in the CO2 concentration. Upon exposure to CO2-free air a transient change of 7 mV (sd 6, n = 14) was observed, while a transient change of −6 mV (sd 6, n = 14) occurred after returning to 700 µl l−1 CO2 (Figure 1b). Changes in the CO2 concentration, however, did not affect the series resistance between the guard cell wall and reference electrode (RS = 5 MΩ, sd 2, n = 13). Guard cells were impaled with sharp double-barrelled electrodes and based on their free running membrane potential could be classified as hyperpolarized, depolarized or far depolarized (Roelfsema et al., 2001). Cells that were hyperpolarized in CO2-free air, depolarized after application of 700 µl l−1 CO2 (Figure 1c). The average membrane potential of hyperpolarized cells was −101 mV (sd = 28, n = 16) and dropped to −46 mV (sd = 17, n = 16), with 700 µl l−1 CO2.
Changes in the conductance of the plasma membrane that underlie the CO2-induced depolarization, were studied in the voltage-clamp mode. Guard cells displaying a CO2-induced depolarization (Figure 2a), were clamped from a holding potential of −80 mV to test potentials ranging from −160 mV to 0 mV. Figure 2b displays the typical time dependent inward- and outward currents that are conducted by K+ selective channels. The outward currents remained constant, while inward currents slightly decreased during the experiment. The presence of CO2 had no obvious effect on the magnitude of time dependent currents (Figure 2b). However, when displayed on a larger scale, a CO2-induced change of instantaneous current became apparent (Figure 2c).
CO2-induced conductance changes at the plasma membrane were further analysed, averaging the data obtained for 6 cells (Figure 3). Steady state currents were sampled at the end of the 2 s-test pulses (Ist, Figure 2b), while instantaneous currents were determined for 10 ms directly after termination of the capacitance compensation peak (Iins, Figure 2c). The steady state currents were only significantly altered at those potentials were no time dependent K+ channels were active (Figure 3a). A significant change of instantaneous current was found over the whole voltage range (Figure 3b), which lead to a shift of the IV-curve along the current axis to more negative values.
In a second series of experiments we tested the sensitivity of guard cells to changes in the CO2-concentration of the electrode. Guard cells in the intact plant were impaled with electrodes filled with electrolyte, equilibrated either with CO2-free air, or air containing 700 µl l−1 CO2. Lowering the CO2 concentration in the electrode increased the probability to find guard cells in the hyperpolarized state. Out of 17 guard cells impaled without CO2 in the electrode, 9 became hyperpolarized, while only 1 out of 11 guard cells became hyperpolarized with high CO2 in the electrode. The difference in membrane potential correlated with a CO2-dependent change of instantaneous currents (Figure 3d). Larger inward currents were found in the presence of CO2. In line with the results with externally applied CO2, CO2 in the electrode did not cause large changes in steady state currents (Figure 3c).
Interaction between red light and CO2 signalling
In the experiments described above guard cells were exposed to CO2, while red light was projected on a large section of the leaf (diameter 20 mm, Figure 6a, beam b). We tested if the response to CO2 was dependent on the red light background. Switching off red light caused 4 cells to depolarize while 8 cells remained hyperpolarized. Cells that had remained hyperpolarized in darkness depolarized upon exposure to CO2 (Figure 4a,b). A subsequent removal of CO2 from the air stream (n = 6) caused a hyperpolarization in some cells (Figure 4a, n = 2), showing that the response to CO2 does not strictly depend on red light. Other cells remained depolarized after the removal CO2 (Figure 4b, n = 4) and only hyperpolarized after the red light was switched on again. This indicates that red light and low CO2 act synergistically on guard cells. High CO2 concentrations therefore may act antagonistically to red light. This hypothesis was tested with guard cells that hyperpolarized upon illumination with red light (Figure 4c). Here, the application of CO2 reversed the effect of red light and the cells became depolarized. All cells that were hyperpolarized in red light (n = 13), depolarized upon exposure to 700 µl l−1 CO2.
The cells that depolarized directly after switching off red light were further analysed. These cells hyperpolarized again after switching on red light (Figure 5a). Clamping these cells from −80 mV to potentials ranging from −160 mV to 0 mV (Figure 5b), revealed that the response to red light was similar to that triggered by low CO2. Turning off red light stimulated an instantaneously activating conductance (Figure 5c). Alike the CO2-response (Figures 2 and 3), the change in conductance was caused by a shift of all currents along the current axis, to more negative values (Figure 5d). Averaged data of 6 cells responding to red light, revealed that red light did not significantly alter the time dependent inward- and outward K+ currents (data not shown).
Red light induced changes of the intracellular CO2 concentration
The similarity of the guard cell response to red light and CO2, prompted us to consider the possibility that red light acts via a decrease of the intracellular CO2 concentration. A miniature CO2 sensor was recently developed by Hanstein et al. (2001) which can be placed in the substomatal cavity via an open stoma. Switching off the red light provided by the microscope lamp caused an increase of intercellular CO2 (Figure 6b), on average the CO2 concentration increased with 244 µl l−1 (sd 90, n = 6). In the latter experiments red light was projected on an area of 20 mm in diameter with a photon flux density of 125 µmol m−2 s−1 (beam b, Figure 6a). For comparison, the leaf was illuminated in a way similar to that in our previous report (Roelfsema et al., 2001), in which we did not observe a red light response. A background of red light at an intensity of 55 µmol m−2 s−1 was provided by the microscope lamp (beam b). Additional red light was projected via the objective at an area of 0.3 mm in diameter with a photon flux density of 350 µmol m−2 s−1 (beam c, Figure 6a). Switching on and off the additional light (beam c) did not significantly change the CO2 concentration in the substomatal cavity (Figure 6c), the average change was −17 µl l−1 (sd 38, n = 6).
Nature of the CO2-dependent conductance
Finally, we analysed the nature of conductance changes caused by CO2 with guard cells in epidermal strips. Here, the ionic composition of the extracellular solution was controlled by bath perfusion. Guard cells were impaled with microelectrodes containing electrolyte equilibrated either with CO2-free air or with air containing 700 µl l−1 CO2. In a bath solution with 5 mm KCl, guard cells showed typical time dependent inward- and outward K+ currents (Figure 7a). In contrast to cells impaled in the intact plant, intracellular CO2 did not cause a significant change in the instantaneous current voltage relation (Figure 7d). Replacing K+ in the bath solution with 2.5 mm Cs+ and 2.5 mm TEA+, reduced the current carried by outward K+-channels and blocked inward channels (Figure 7b). Subsequently the plasma membrane was clamped from a holding potential of 0 mV, to more negative test voltages (Figure 7c). With this protocol, CO2 in the electrode enhanced inward currents at the most negative test potentials (Figure 7e). Apparently, CO2 stimulated a channel activating at 0 mV and conducting an inward current. The inward current slowly deactivated in time (Figure 7c), the velocity of deactivation was slowed down by the presence of CO2 in the electrode. In the absence of CO2 the half time of deactivation at −180 mV was 0.46 sec (se 0.11, n = 11), which increased to 0.91 sec (se 0.22, n = 11) in the presence of CO2.
Here, we explored the possibility that CO2 acts as an intermediate link in the red light response of guard cells. The hypothesis that guard cells respond to light via changes in the intercellular CO2 concentration was already proposed by Stahl (1920). The theory has remained accepted (Heath, 1959) until several authors found that guard cells could directly respond to light (Kuiper, 1964; Sharkey and Raschke, 1981). The direct response to light is at least partially mediated via blue light receptors encoded by PHOT1 and PHOT2 (Kinoshita et al., 2001), which leaves the possibility that the response to red light is indirect and only affects guard cells via changes in the CO2 concentration.
Guard cell responses to CO2
High concentrations of CO2 lower the conductance of leaves via long- as well as short-term mechanisms. Long-term exposure of plants to high CO2, decreases the stomatal density of leaves, a process that involves activity of the HIC gene (Gray et al., 2000). In the short term, CO2 provokes stomatal closure, a response studied with electrophysiological methods in epidermal peels, by Brearley et al. (1997). The latter authors found that CO2 induced a depolarization of the plasma membrane from −60 to −49 mV. Although such a depolarization may enhance the efflux of K+ across the plasma membrane, it does not alter the direction of the K+-flux. Guard cells in intact plants were considerably more responsive, showing an average depolarization from −101 to −46 mV. Membrane potential changes in the latter range reverse the flux of K+ across the plasma membrane and therefore tune the osmotic machinery that causes stomatal movement (Roelfsema et al., 2001; Thiel et al., 1992).
For guard cells in the intact plant, changes in the CO2 concentration did not affect voltage dependent K+ channels in the plasma membrane, but increased the activity of instantaneously activated channels. Guard cells responded both to CO2 supplied via the atmosphere and to CO2 applied via the microelectrode. Experiments with epidermal strips indicated that CO2 applied via the electrode stimulated anion channels. Brearley et al. (1997) described a transient activation of anion channels by CO2. In contrast, we found a steady activation of these channels by CO2. In line with the results of Brearley et al. (1997), CO2 did not only enhance the conductance, but also slowed down the deactivation of anion channels.
Although CO2 stimulates the anion channels in guard cells both in epidermal strips and in the intact plant, their gating properties were different in both systems. In epidermal strips, CO2-stimulated anion channels were only resolved when voltage pulses were applied from a holding potential of 0 mV, but not from −100 mV. In intact plants, the CO2 effect was observed with holding potentials of −100 and −80 mV. This difference may be due to the loss of apoplastic solutes from epidermal strips. Organic anions are present at significant concentrations in the intact plant (Lohaus et al., 2001; Roelfsema and Hedrich, 2002), but not in the bath solution of epidermal strips. The presence of organic anions shifts the activation potential of anion channels to more negative values (Dietrich and Hedrich, 1998). In this respect malate may play an important role, since it is synthesized in large quantities in guard cells during stomatal opening (Raschke and Schnabl, 1978). Malate feeds in to the apoplastic pool during stomatal closure, which becomes elevated at high atmospheric CO2 (Hedrich et al., 1994). The elevated level of malate in the apoplast will enhance the conductance of anion channels and thus may modulate the response to CO2 (Hedrich and Marten, 1993; Hedrich et al., 1994).
CO2 signalling pathway
The signalling pathway, through which CO2 affects anion channels, probably comprises a change of cytoplasmic Ca2+. Webb et al. (1996) found that the cytoplasmic Ca2+ concentration of guard cells of Commelina communis increases in the presence of CO2. This may explain the stimulation of anion channels, since both the rapid- as well as the slow activating anion channel are stimulated by Ca2+ (Hedrich et al., 1990; Schroeder and Hagiwara, 1989). However, an increase of the cytoplasmic Ca2+ level would also inhibit inward rectifying K+-channels (Grabov and Blatt, 1999; Schroeder and Hagiwara, 1989). The latter response was not observed, indicating that the activation of anion channels is not just mediated by a simple rise of cytoplasmic Ca2+. Possibly, CO2 triggers a signal specific Ca2+ wave that stimulates the anion channel, but does not inhibit the inward K+ channel (Allen et al., 2001; McAinsh and Hetherington, 1998). Alternatively, the anion channel could have been stimulated through a Ca2+-independent pathway. Rapid anion channels are sensitive to cytoplasmic pH changes (Schulz-Lessdorf et al., 1996) and thus could be modulated via a CO2-induced pH-signal.
Red light- and CO2-responses are interconnected
Guard cells were capable to respond to CO2 in complete darkness, well in agreement with gas exchange measurements (Hedrich et al., 2001). It is therefore unlikely that the response to CO2 depends on guard cell photosynthesis. Some cells, however, depolarized in response to CO2, but failed hyperpolarize again after the removal of CO2. These cells hyperpolarized only after the onset of the red light illumination. Red light, applied to a larger area of the leaf, lowers the intercellular CO2 concentration, which is in turn sensed by the guard cells. A beam of red light, projected on guard cells only, did not trigger a response at the plasma membrane (Roelfsema et al., 2001; Taylor and Assmann, 2001). In agreement with these results, this small beam of light could not evoke a change in the CO2 concentration of the substomatal cavity (Figure 6c). This indicates that guard cells do not respond directly to red light, but to a red light induced lower CO2 concentration, instead. This is in line with the function of stomata, providing a pathway for CO2 uptake when the CO2 concentration is limiting photosynthesis.
Photosynthetic CO2 fixation does occur in guard cells (Goh et al., 1999; Gotow et al., 1988), but the rate is too low to support the accumulation of carbohydrates during stomatal opening (Outlaw, 1989; Raschke et al., 1988). The maximum rate of CO2 fixation has been determined for Vicia faba guard cell protoplasts. The rate varied from 240 to 54 fmol CO2 h−1 per protoplast (see Experimental procedures). Based on chlorophyll fluorescence measurements, it is questionable if guard cells in intact plants have the same photosynthetic capacity. Guard cell protoplasts have a quantum yield which is only slightly lower than that of mesophyll protoplasts (Goh et al., 1999). However, in intact plants the difference in quantum yield between guard cells and mesophyll cells was much more pronounced (Lawson et al., 2002). Furthermore, the present experiments show that guard cells respond to small amounts of CO2 leaking from microelectrodes. The amount of CO2 diffusing from the microelectrode was estimated at 1 fmol h−1 (see Experimental procedures), apparently these small quantities can cause a significant increase of the cytoplasmic CO2 concentration.
Although the photosynthetic rate of CO2 fixation is relatively low for guard cells in intact plants, the situation might be slightly different for isolated guard cells. Red light may lower the cytoplasmic CO2 concentration of guard cells in epidermal strips, favouring red light induced stomatal opening. In intact leaves, however, the photosynthetic capacity of guard cells is negligible compared to that of the mesophyll. Here, red light most likely acts primarily on the mesophyll cells and lowers the CO2 level in the substomatal cavity. Subsequently, the CO2 concentration in guard cells will drop and stimulate K+-uptake, causing stomata to open.
Based on these results we conclude that guard cells are equipped with two mechanisms to respond to altered light conditions. First, a response to red light, via decreased CO2 levels that inhibits anion channels and second, a direct response to blue light, which stimulates the H+-ATPase. The red light/CO2-response couples the increased demand for CO2 to stomatal opening, while the blue light action provides for a rapid response to fast changes in light conditions, such as sunflecks (Assmann and Shimazaki, 1999).
Guard cell measurements on intact plants
Broad bean (V. faba L. cv. Grünkernige Hangdown, Gebag, Hanover, Germany) plants were grown in a green house. 4- to 6-week-old plants were placed next to an upright microscope (Axioskop 2FS, Carl Zeiss, Göttingen, Germany) The adaxial side of a leaf was mounted on a Plexiglas holder in the focal plane using double sided adhesive tape. A glass capillary (inner Ø, 0.58 mm, GC100F-10, Harvard Instruments, Edenbridge, Kent, UK) filled with 50 mm KCl was placed on the abaxial cuticle and served as a reference electrode. The reference was connected to ground via an agar bridge and a Ag/AgCl half-cell filled with 300 mm KCl. Water evaporating from the tip of the capillary was automatically replaced, keeping the spot wet, at which the capillary touched the cuticle.
The leaves were illuminated, at an area with a diameter of 20 mm, with red light from the microscope lamp (HAL 12V/100W, Carl Zeiss) filtered by a longpass (edge wavelength 610 nm) glass filter (RD 610, Schott, Mainz, Germany). The photon flux density was 125 µmol m−2 s−1 at the adaxial side, which corresponds to a density of 3.5 µmol m−2 s−1 at the abaxial side of the leaf. The guard cells were observed through a dry objective (Epiplan LD 50×/0.50, Carl Zeiss). A stream of air (0.5 l min−1) was guided towards the leaf via a PVC suspensor (split dimensions 14 mm × 2 mm). CO2 free air was obtained with soda lime, while the CO2 concentration could be increased to 700 µl l−1, by direct injection of CO2 to the air steam via a capillary (inner Ø, 0.1 mm, length 10 cm, TPS100170, Composite Metal Services, Worcester, UK). Electrodes were moved towards an open stomate with a micromanipulator (type 5171, Eppendorf, Hamburg, Germany) equipped with a Piezo translator (P-280.30, Physik Instrumente, Waldbronn, Germany). Guard cells under investigation were 41 µm (sd = 3, n = 60) in length and 14 µm (sd = 1, n = 60) in diameter.
Guard cell measurements in epidermal strips
The abaxial epidermis was peeled from the leaves and attached to a microscope slide using Medical Adhesive (VM 355, Ulrich AG, St. Gallen, Switzerland). The microscope slide was mounted in a perfusion chamber that allowed a solution exchange rate of 5 bath volumes min−1. The bath solutions contained 1 mm CaOH buffered to pH 6.0 with Mes and either 5 mm KCl or 2.5 mm TEACl and 2.5 mm CsCl. The solutions were aerated with CO2 free-air (passed through soda lime) starting at least 1 h previous to the experiments. Guard cells were impaled using the same microscope and manipulators as described above, but now the microscope was equipped with a water immersion objective (Achroplan 40×/0.80W, Carl Zeiss).
Electrodes and electrical configuration
Blunt- and fine-tipped-double barrelled electrodes were pulled as described previously (Roelfsema et al., 2001). The electrodes were filled with 300 mm KCl, which was equilibrated with CO2-free air or air containing 700 µl l−1 CO2, for at least 1 h. The electrodes were connected via Ag/AgCl half cells to the same electrical circuit as described previously.
Measurements with miniature CO2 sensor
The miniature CO2 sensors were made as described by Hanstein et al. (2001). The sensor is based on a small pH-electrode placed in a capillary containing a carbonate buffer with carbonic anhydrase. The volume between the tip of the sensor and the pH electrode was kept small to detect CO2-dependent pH-changes. The CO2 sensor was introduced into the substomatal cavity, via an open stoma, using the same set-up as for the impalement measurements. Red light was provided by the microscope lamp and projected on the abaxial side at an area of 20 mm in diameter and a photon flux density of 125 µmol m−2 s−1. For comparison, red light was applied at conditions similar to a previous paper (Roelfsema et al., 2001). Now, red light provided by the microscope lamp was given at a photon flux density of 55 µmol m−2 s−1, while additional red light was obtained with a second light source (KL 1500, Schott, Mainz, Germany). The second beam of red light was guided through the objective and projected on the abaxial side of the leaf. Note that for the present experiment a dry objective was used, while a water immersion objective was used in the previous paper (Roelfsema et al., 2001). The second beam was projected at an area of 0.3 mm in diameter with a photon flux density of 350 µmol m−2 s−1.
The slow inactivation of anion channels at −180 mV (Figure 4b) was fitted with the following equation:
where, Im is the plasma membrane current, Ifin the steady state current, Itime1 and Itime2 time dependent changes in current and τ1 and τ2, respectively, fast and slow time constants. Current traces at −180 mV were fitted using Prism software (version 3, Graphpad, San Diego, CA, USA).
The photosynthetic rate of CO2 fixation determined for V. faba guard cell protoplasts ranges from 120 (Goh et al., 1997) to 27 nmol µ−1g chlorophyll h−1 (Gotow et al., 1988). Taken into account a chlorophyll content of 2 pg per guard cell (Shimazaki et al., 1983), this corresponds to CO2 fixation rates of single guard cells of 240 and 54 fmol h−1, respectively.
The tip dimensions of double-barrelled electrodes used for impalement were measured with an electron microscope. The electrodes had a tip with two pores of approximately 50 nm diameter and a taper angle of 9°. The double-barrelled electrodes can thus be regarded as two cones with a taper angle of 9° that are aligned. The cones are truncated at their tip, here the smallest radius (r), r0 = 25 nm. The diffusion rate of CO2 from such an electrode can be calculated according to Purves (1981). The [CO2] at any point in the electrode would be:
Supposed that the CO2 concentration at the start of the cone is 25.3 nmol cm−3 and at the end of the cone is 0. Differentiating Eqn. 2 and calculating for r = r0 reveals:
The diffusion rate of CO2 from the tip of the electrode (θr) is now:
Where DCO2 is the diffusion coefficient of CO2 in water (1.7 × 10−5 cm2 s−1, Nobel, 1970) and A the area of the electrodes tip. For double-barrelled electrodes we calculated an efflux of 0.94 fmol CO2 h−1.
We thank Ralf Steinmeyer, Petra Dietrich and Natalya Ivashikina (University of Würzburg) for help with the preparation of the manuscript and Rainer Wolf (University of Würzburg) for assistance concerning electron microscopy. This work was funded by grants from the Deutsche Forschungsgemeinschaft to R.H.