ABSTRACT
The oxygen affinity of the enzyme system involved in mitochondrial respiration indicates, in relation to intracellular oxygen levels and interpreted with the aid of flux control analysis, a significant role of oxygen supply in limiting maximum exercise. This implies that the flux control coefficient of mitochondria is not excessively high, based on a capacity of mitochondrial oxygen consumption that is slightly higher than the capacity for oxygen supply through the respiratory cascade. Close matching of the capacities and distribution of flux control is consistent with the concept of symmorphosis. Within the respiratory chain, however, the large excess capacity of cytochrome c oxidase, COX, appears to be inconsistent with the economic design of the respiratory cascade. To address this apparent discrepancy, we used three model systems: cultured endothelial cells and mitochondria isolated from heart and liver. Intracellular oxygen gradients increase with oxygen flux, explaining part of the observed decrease in oxygen affinity with increasing metabolic rate in cells. In addition, mitochondrial oxygen affinities decrease from the resting to the active state. The oxygen affinity in the active ADP-stimulated state is higher in mitochondria from heart than in those from liver, in direct relationship to the higher excess capacity of COX in heart. This yields, in turn, a lower turnover rate of COX even at maximum flux through the respiratory chain, which is necessary to prevent a large decrease in oxygen affinity in the active state. Up-regulation of oxygen affinity provides a functional explanation of the excess capacity of COX. The concept of symmorphosis, a matching of capacities in the respiratory cascade, is therefore complemented by ‘synkinetic’ considerations on optimum enzyme ratios in the respiratory chain. Accordingly, enzymatic capacities are matched in terms of optimum ratios, rather than equal levels, to meet the specific kinetic and thermodynamic demands set by the low-oxygen environment in the cell.
RESPIRATORY CASCADE TO RESPIRATORY CHAIN
The mitochondrial respiratory chain fulfils the kinetically and thermodynamically complex tasks of (1) catalysing a high electron flux to oxygen in active metabolic states and (2) pumping protons at high efficiency as a basis for intense aerobic ATP production. Moreover, these functions must be (3) performed at low intracellular oxygen pressure, which requires a high affinity for oxygen (Babcock and Wikström, 1992).
Oxygen in the environment is globally stable when measured at the scale of generations, but varies locally and may become limiting under diverse environmental and physiological conditions. Respiration of all organisms is a function of the partial pressure of oxygen, , below a critical environmental oxygen pressure, depending on the structural and functional capacities and regulatory mechanisms at all steps of the respiratory cascade (Fig. 1). The form of the oxygen-dependence varies among species and may be complex, yet a surprisingly simple hyperbolic equation describes the flux/pressure relationship in some cases (Gnaiger, 1993a). The rectangular hyperbolic equation with an intercept through the origin:
When studying the respiration of isolated cells, the major pathway for oxygen from the environment along the respiratory cascade (Weibel, 1984) is removed (Fig. 1), and the relevant range is reduced from more than 20 to 1 kPa. In tumour vessels, for instance, in vivo values as low as 0.7 kPa are not hypoxic (Helmlinger et al. 1997). Under the simpler conditions for oxygen transport from extracellular to the terminal oxidase, a hyperbolic oxygen-dependence (equation 1) is more generally observed (for a review, see Gnaiger et al. 1995), despite the remaining complexities of intracellular oxygen gradients and heterogeneities of lipid and aqueous phases and mitochondrial clusters (Fig. 2).
A further shortcut for oxygen delivery to the actual site of conversion of oxygen to water is afforded by the study of isolated mitochondria. A myoglobin saturation of approximately 50% is typical for the volume-averaged intracellular oxygen pressure in heart and skeletal muscle, corresponding to 0.3 kPa (2 mmHg, 3 μmol l−1; Wittenberg and Wittenberg, 1989; Wagner, 1996). In the physiologically relevant intracellular range below 1 kPa, respiration can again be described by a hyperbolic function. Measurements with mitochondria at much higher oxygen levels appear to ignore this physiological background of the respiratory cascade (Gnaiger et al. 1995).
Biochemists frequently refer to the P50 as the apparent Michaelis–Menten constant for oxygen, Km′. However, even this mitochondrial compartment at the end of the respiratory cascade is a highly complex metabolic network, harbouring the multienzyme system known as the mitochondrial respiratory chain (Fig. 1). Since the oxygen-dependence of organismic respiration is sensitive to each component of the respiratory cascade, so is the mitochondrial P50 potentially a function of each element in the respiratory chain, in contrast to the terminological implications of a Km′ ‘constant’.
In the present work, we address four aspects of mitochondrial oxygen kinetics, aiming at an integration of physiological and bioenergetic perspectives. (1) The current status of mitochondrial oxygen kinetics is reviewed in terms of experimentally determined P50 values. (2) P50 increases with the rate of oxidative phosphorylation and is a function of metabolic state (coupled versus uncoupled respiration). (3) Oxygen affinity, i.e. the reciprocal value of P50, provides insufficient information on oxygen kinetics when Jmax is a function of oxygen-independent parameters, such as ADP concentration. Under these conditions, the apparent catalytic efficiency, Jmax/P50, is physiologically relevant to describe respiratory control by low oxygen levels. While affinity decreases, the apparent catalytic efficiency increases from the resting to the active state and increases in general as the thermodynamic efficiency of energy conversion decreases. This analysis sets the stage for (4) a discussion of the oxygen kinetics of the mitochondrial respiratory chain in relation to the organismic respiratory cascade (Fig. 1), with reference to the concept of symmorphosis (Weibel et al. 1991). The excess capacity of cytochrome c oxidase (COX) relative to the capacity of the integrated respiratory chain and respiratory cascade is functionally explained as a regulatory mechanism for maintaining a high oxygen affinity by a COX turnover rate that is low even at maximum aerobic performance.
OXYGEN AFFINITY AND METABOLIC STATE/RATE
The conflicting views on the role of oxygen in respiratory control are mainly due to uncertainties about intracellular gradients and levels of oxygen in the mitochondrial microenvironment and to the difficulties of making respiratory measurements at the corresponding low oxygen concentrations (Gnaiger et al. 1995). Is intracellular limiting under physiological conditions? Negative answers to this important question refer to a value for P50 for mitochondrial respiration of 0.005 kPa, which is nearly 100 times below intracellular oxygen levels (Cole et al. 1982; Gayeski and Honig, 1991; Wittenberg and Wittenberg, 1985), whereas an affirmation relates to more recent data on P50 values that are less than one-tenth of intracellular (Wilson et al. 1988; Rumsey et al. 1990; Gnaiger et al. 1995). Up to 30-fold differences in P50 reported for mitochondrial respiration continue to separate opposing views on the role of intracellular oxygen in respiratory control and require resolution.
Intracellular oxygen gradients and mitochondrial oxygen kinetics
The study of the oxygen kinetics of isolated cells offers the advantage of observing mitochondrial function largely undisturbed in the intracellular environment (Fig. 2). An experimental example with endothelial cells is shown in Fig. 3. By selecting appropriate cell densities, the transition from kinetic oxygen saturation to anoxia is sufficiently fast to eliminate effects that are unrelated to the regulation of flux by oxygen in the physiological range, without compromising the requirement to obtain a large number of data points during the transition (Fig. 3, inset). Continuous measurement in a closed system offers the advantage of a high density of data points but presents the problem of a non-steady-state process. A quasi-steady state may be realised at each point during a sufficiently slow transition (Fig. 3), such that externally measured oxygen flux is equal to intracellular diffusional flux and equal to all enzyme-catalysed reaction rates along the respiratory chain (Fig. 1). Hyperbolic oxygen flux/pressure relationships were obtained by non-linear fitting in the oxygen range below 1.1 kPa (Fig. 4; equation 1). The distribution of residuals provides a powerful statistical evaluation (Cornish-Bowden, 1995), both in cells (Fig. 4, inset) and in isolated mitochondria (Gnaiger et al. 1995, 1998). In contrast, Petersen et al. (1974) show double-reciprocal plots of the oxygen kinetics of isolated rat liver mitochondria in the range below 1 μmol l−1 O2 (<0.1 kPa), yielding non-linear (non-hyperbolic) patterns in all metabolic states. Their method (P50 varies with enzyme concentration; Petersen et al. 1976), the narrow oxygen range and the double-reciprocal Lineweaver–Burk plots appear to be inadequate for a statistical test of deviations from a hyperbolic relationship (Cornish-Bowden, 1995).
Interpretation of the oxygen affinity of intact cells is complex owing to the diversity of metabolic states and difficulties in quantification of intracellular oxygen gradients (Jones, 1986; Wittenberg and Wittenberg, 1989; Rumsey et al. 1990). Suspended endothelial cells provide a good model for the study of oxygen kinetics because of their small size and regular shape (Fig. 2). Their respiratory rate is 95% mitochondrial (Steinlechner-Maran et al. 1996). The increase in cellular P50 with maximum respiratory rate per cell (=oxygen flow, ; Gnaiger, 1993b) is caused by intracellular oxygen gradients, which increase with oxygen flow, and by a change in mitochondrial P50 (Fig. 5). Separation of these interdependent factors is achieved by a comparison of the rate-dependence of P50 in identical ranges of oxygen flow in coupled and uncoupled/inhibited cells (Fig. 5). The shallow slope of P50 as a function of oxygen-saturated flow, Imax, in the uncoupled/inhibited state indicates that intracellular oxygen gradients are small. The difference between the slopes of coupled and uncoupled/inhibited cells, therefore, can be interpreted as the diffusion-independent mitochondrial P50 (Table 1). The lower P50 in uncoupled cells emphasises the importance of metabolic state in the regulation of P50 (Steinlechner-Maran et al. 1996). A direct test is required with isolated mitochondria.
FACT AND ARTEFACT
It is becoming increasingly clear that various potential artefacts are confounding factors in mitochondrial oxygen kinetics. These include oxygen gradients in the respirometer, back-diffusion of oxygen from Teflon stirrers, Perspex chambers and inappropriate sealing materials, the low sensitivity of oxygen sensors and insufficient time resolution. These limitations are resolved in high-resolution respirometry, based on the Oroboros oxygraph and DatLab software (Oroboros, Innsbruck, Austria) and on rigorous tests of instrumental background, correction for the exponential time constant of the sensor (3–6 s) and internal calibration of the zero oxygen signal (Gnaiger et al. 1995, 1998). The contribution of instrumental artefacts changes with mitochondrial concentration and activation state. Unfortunately, only few studies report adequate methodological tests.
INSTRUMENTAL BACKGROUND OXYGEN FLUX OVER THE EXPERIMENTAL OXYGEN RANGE (Gnaiger et al. 1995)
Background measurements yield the change in oxygen concentration over time in the closed system containing incubation medium without biological material. Background oxygen flux results from oxygen consumption by the polarographic oxygen sensor, which is directly proportional to , and back-diffusion of oxygen into the medium at declining oxygen levels, which is proportional to the difference between the oxygen source and the medium. Gradual depletion of the oxygen source gives rise to a hysteresis effect. The overall background flux is a near-linear function of oxygen concentration, which is subtracted from the experimental flux at each point of oxygen measurement. A sampling interval of 1 s is appropriate. In addition, occasional microbial contamination is detected by careful background tests.
INDEPENDENCE OF P50 FROM CONCENTRATIONS OF CELLS, MITOCHONDRIA OR ENZYMES AT CONSTANT METABOLIC STATE (Rumsey et al. 1990; Steinlechner-Maran et al. 1996; Gnaiger et al. 1998)
Oxygen flux per volume of the oxygraph chamber, (pmol s−1 cm−3=nmol l−1 s−1), is a function of specific respiratory activity and the concentration of mitochondria, cells or enzyme(s). ,V increases proportionally with enzyme concentration. In contrast, P50 must be constant, except if (i) concentration exerts an effect on metabolic state (Steinlechner-Maran et al. 1996), (ii) the varying aerobic–hypoxic–anoxic transition time influences compensatory processes (Wilson et al. 1988) or (iii) different instrumental artefacts are sensitive to different levels of ,V, particularly the time resolution, the number of data points obtained in the oxygen-dependent region and the ratio of uncorrected background flux to metabolic oxygen flux. We tested our system for concentration effects in the ,V range 20–500 pmol s−1 cm−3 at a chamber volume of 2 cm3, and observed no significant differences in P50 at various mitochondrial dilutions. At higher volume-specific fluxes, erroneously high P50 values were obtained (Gnaiger et al. 1998; B. Lassnig, A. Kuznetsov, R. Margreiter and E. Gnaiger, in preparation).
Oxygen gradients from a gas–aqueous boundary layer to the oxygen sensor yield low apparent P50 values. This artefact might explain the low P50 of 0.005 kPa in active skeletal muscle mitochondria at high protein concentration (10 mg cm−3; Cole et al. 1982) in comparison with the value of 0.015 kPa observed on the basis of intracellular myoglobin saturation in resting cardiomyocytes (Wittenberg and Wittenberg, 1985). The latter value agrees closely with the value of 0.025 kPa for skeletal muscle mitochondria determined by high-resolution respirometry (Gnaiger et al. 1995) and of 0.008 kPa for passive heart mitochondria (Sugano et al. 1974). Although Sugano et al. (1974) used a gas–aqueous transfer system, the low mitochondrial density of 0.2 mg cm−3 and the spatially dispersed bioluminescence probe reduced the interference by oxygen gradients. A higher P50 of 0.04 kPa was obtained for passive heart mitochondria in a closed respirometer using a phosphorescence probe (Rumsey et al. 1990). The difference may be due to uncorrected oxygen back-diffusion from the Teflon stirrer. Whereas high volume-specific fluxes cause an underestimation of P50 due to gradients in steady-state oxygen transfer systems, the opposite effect is produced in closed systems (P50 of 0.15 kPa in active heart and liver mitochondria; Costa et al. 1997) where fast aerobic–anoxic transitions are beyond the scope of time resolution of the oxygen sensor. In Table 1, results are excluded that appear to suffer from implicit artefacts.
Mitochondrial P50 and elasticity in passive and active states
The P50 of mitochondrial respiration increases with activation by ADP (Table 1). In heart mitochondria, P50 doubles from 0.02 to 0.04 kPa (from 0.12 to 0.26 mmHg), whereas in liver mitochondria P50 increases almost threefold from 0.02 to 0.06 kPa (0.43 mmHg) in the active state (B. Lassnig and others, in preparation; Gnaiger et al. 1998; Table 1). At an intracellular of 0.3 kPa, the maximum aerobic performance of heart mitochondria is limited to 90% of Jmax (Fig. 6), as derived from equation 1:
Catalytic efficiency versus affinity
The decrease in oxygen affinity (=1/P50) in active states, when the demand for oxygen is highest, appears to be paradoxical from a functional point of view. While the affinity decreases, however, maximum flux increases greatly, such that the ratio Jmax/P50 is actually increased by adenylate activation (Fig. 7). The scope for activation by ADP declines under hypoxia, but this decline would be even more pronounced at a constant Jmax/P50 ratio (Gnaiger et al. 1998). As seen from equation 1, Jmax/P50 is the initial slope of the hyperbolic function:
Oxygen affinity and catalytic efficiency are highly sensitive to uncoupling of oxidative phosphorylation (Fig. 8). At low concentration of carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP), uncoupled respiration of endothelial cells increased to a maximum value more than double that for coupled oxygen flow. P50 and catalytic efficiency increased concomitantly. Above an optimum uncoupler concentration, oxygen-saturated respiratory flow, Imax, declined to the inhibited state, but catalytic efficiency increased further as a function of state and independently of rate (Fig. 8). In the uncoupled/inhibited state (compare Fig. 5), further changes in catalytic efficiency were not significant. Catalytic efficiency increases as thermodynamic efficiency decreases, consistent with a model based on the thermodynamics of irreversible processes (Gnaiger et al. 1995). This inverse relationship is not restricted to uncoupling, but also applies to activation by ADP (Fig. 7) when thermodynamic efficiency decreases at maximum power output (Gnaiger, 1993c).
Oxygen affinity and excess capacity of cytochrome c oxidase
Threshold effect and flux control coefficient of cytochrome c oxidase
ADP and oxygen saturation of mitochondrial respiration at optimum conditions does not drive cytochrome c oxidase to its kinetic limit: turnover rate can be increased further by the addition of artificial electron donors. COX activity was measured in the oxygraph after blocking complex III with antimycin A (Fig. 1). Reduced N,N,N′,N′-tetramethyl-p-phenylenediamine dihydrochloride (TMPD) served as an electron donor to cytochrome c, and its concentration was held constant by the addition of ascorbate. Auto-oxidation of ascorbate and TMPD is a function of oxygen pressure and was subtracted from the measured oxygen consumption as a chemical background (B. Lassnig and others, in preparation). The activity of COX was 50% and 100% above maximally ADP-stimulated flux through the respiratory chain in liver and heart mitochondria, respectively, at kinetic oxygen saturation and 500 μmol l−1 TMPD (Fig. 9). This is an underestimation since TMPD becomes saturating only at millimolar levels in coupled mitochondria (Morgan and Wikström, 1991).
The phenomenon of an excess capacity of cytochrome c oxidase is well known and largely unexplained. The molar ratio between complexes IV and I is six in skeletal muscle (Schwerzmann et al. 1989). COX has a low flux control coefficient, , which means that a small change in COX activity plays a minor role in the control of mitochondrial respiration (Groen et al. 1982; using azide for inhibitor titrations). In titrations with cyanide, Kuznetsov et al. (1996) found a lower flux control coefficient of COX than with azide. Identical values of maximum fluxes in permeabilized muscle fibres of 50% COX-deficient mice and controls (Kuznetsov et al. 1996) might even suggest a flux control coefficient of zero. The flux control coefficient is the (infinitesimally small) change in relative flux through the pathway over the (infinitesimally small) change in relative capacity of the step (Kacser and Burns, 1973; Heinrich and Rapoport, 1974). High excess capacity is associated with a low flux control coefficient. Correspondingly, is increased in muscle fibres of the 50% COX-deficient mice (Kuznetsov et al. 1996). In a theoretical model, increases when muscle mitochondria are nearly 50% oxygen-limited at intracellular oxygen pressure (Korzeniewski and Mazat, 1996), but this lacks support from experimental oxygen kinetics.
A low flux control coefficient is directly related to the threshold effect and explains why COX can be inhibited to a large extent without effect on flux through the respiratory chain (Letellier et al. 1993, 1994). The threshold effect of COX may provide a safety margin, protecting against the consequences of accumulations of mutations in mitochondrial DNA (mtDNA) with age (Sohal and Weindruch, 1996) and counteracting genetic defects with mtDNA heteroplasmy, such that a small fraction of wild-type mtDNA remaining in a mitochondrion or cell is sufficient for normal respiratory function (Mazat et al. 1997). Importantly, the low flux control coefficient of cytochrome c oxidase needs re-evaluation under intracellular levels (Gnaiger et al. 1995).
SYMMORPHOSIS AND FLUX CONTROL ANALYSIS
Is the excess capacity of COX non-functional, merely providing a safety margin and expensive insurance against irregular events (Diamond and Hammond, 1992), or does it serve a direct function and is thus compatible with economic design of the respiratory chain? The principle of economic design is well illustrated when considering that the mitochondrial respiratory capacity in skeletal muscle exceeds just slightly the capacity for cardiovascular oxygen supply (Saltin, 1985), and the oxidative capacity of the heart is merely 10–20% above the maximum physiological performance (Mootha et al. 1997). Indeed, the simplest structure–function relationship in the respiratory cascade appears to be at the level of muscle mitochondria, the total volume of which varies in proportion to the aerobic capacity of the organism (Weibel et al. 1991). Regardless of allometric variation of body mass and adaptive variation of aerobic scope in athletic versus sedentary species, there is a constant relationship of 3.7 μmol O2 s−1 cm−3 between the maximum rate of oxygen consumption of the whole organism and mitochondrial volume in the musculature (Hoppeler and Turner, 1989). At a mitochondrial volume of 1.4 mm3 mg−1 mitochondrial protein, this is equivalent to a protein-specific maximum O2 flux of 4.1 nmol s−1 mg−1 at 30°C, corresponding to the range of 2–6 nmol s−1 mg−1 measured in ADP-activated mitochondria isolated from skeletal muscle (Schwerzmann et al. 1989) and 4.5 nmol s−1 mg−1 in active heart mitochondria (Fig. 7). In contrast to the closely matched mitochondrial capacity, there is excess lung structure in relation to maximum respiration (Weibel et al. 1991), suggesting that the flux control coefficient of the lung in the respiratory cascade must be very low under normoxic conditions.
A mitochondrial excess capacity of 10–20% (Mootha et al. 1997) is in line with an elasticity of 0.1 for active heart (Fig. 6). Under these conditions, intracellular oxygen level can be maintained high with respect to mitochondrial P50, but not high enough for full kinetic oxygen saturation of the respiratory chain (Fig. 6). This implies a fairly high mitochondrial flux control coefficient, sharing control with the cardiovascular system. Such distribution of flux control is seen directly in training and in a comparison of sedentary with athletic animals, where the gain in aerobic performance is due to cardiovascular improvement and an accompanying increase in mitochondrial density (Weibel et al. 1991).
One central hypothesis of symmorphosis is the matching of capacity for oxygen flux at each step in the respiratory cascade (Fig. 1) to the maximum flux of the physiological system (Weibel et al. 1991). At steady state, oxygen fluxes through each step in the cascade are equal. Does the matching of capacities, therefore, imply that equal capacities of all steps are optimal for economical design? Intuitively, this appears to be the logical implication of symmorphosis. The mechanistic aspects of oxygen kinetics and the mathematical formalism of flux control analysis, however, provide a quantitatively different perspective. Mitochondrial oxidative capacity is conventionally and conveniently measured at kinetic oxygen saturation (Schwerzmann et al. 1989; Mootha et al. 1997). Full kinetic oxygen saturation would be required, at each mitochondrion in the cell, at a 1:1 matching of the capacities for oxygen consumption and supply. A kinetic saturation of 99% is reached at a ratio of 100 (equation 2) or at an intracellular of 3.5 kPa (26 mmHg). Quite apart from the contradiction with experimental observation by a factor of 10, this high intracellular oxygen pressure would seriously reduce the oxygen gradient from haemoglobin to mitochondria and thereby impose additional demands on cardiovascular structure and function to compensate for the lower oxygen pressure head for diffusion. In addition, at a mitochondrial excess capacity of zero, the elasticity reduces to zero and the mitochondrial flux control coefficient is 1, exerting exclusive control over maximum aerobic performance. Equal capacities for mitochondrial oxygen consumption and cardiovascular oxygen supply are not economical.
At the other extreme of an intracellular ratio of 1 (a factor of 10 below measured in myocytes), the elasticity for oxygen is increased to 0.5 (Fig. 6). Mitochondrial oxidative capacity would have to be 100% instead of 10% in excess of oxygen supply capacity (equation 5). Such a design is not only uneconomical but impossible, because a doubling of mitochondrial volume density in cardiomyocytes would exceed the available space (Hoppeler et al. 1984). Neither zero nor a large mitochondrial excess capacity is economical. Instead, criteria for an optimum excess capacity and optimum distribution of flux control coefficients have to be developed under the constraints set by mitochondrial oxygen kinetics and the intracellular microenvironment (Fig. 6).
KINETIC TRAPPING OF OXYGEN: COX AND THE RESPIRATORY CHAIN
In these examples with the isolated enzyme, Km′ of COX is increased by decreasing the ‘internal’ rate constants, ket or k1 (equations 6 and 7). In our examples with intact mitochondria, P50 of the respiratory chain is decreased by decreasing Jmax. Since Jmax<kin (see equation 6), flux control is shifted from the internal rate-limiting electron input into COX to the external electron transport chain, thus reducing COX turnover rate at oxygen saturation. (1) COX turnover rate is reduced at low flux through the respiratory chain in the passive state. This decreases the P50 from the active to the passive state. (2) COX turnover rate is reduced at high flux through the respiratory chain by increasing the excess capacity of COX and thus distributing electron input flux to a larger number of enzymes, which yields a lower input rate per enzyme. This decreases the P50 in active mitochondria of the heart compared with the liver. Actually, equation 7 provides a similarly plausible model for the functional relationship between the excess capacity of COX and high oxygen affinity. The integrated channel area for oxygen diffusion to the reactive centre of COX is enlarged by providing a higher density of COX molecules, which then decreases the P50. This ‘synkinetic’ regulation of P50 depends merely on the quantitative relationship between COX capacity and the respiratory chain, but does not imply kinetic differences between COX in liver and heart (Bonne et al. 1993). The apparent excess capacity of COX compared with the aerobic mitochondrial capacity is a basis for the high oxygen affinity of the respiratory chain. It is the necessary price to be paid for the thermodynamic and kinetic optimum function of the respiratory chain within the constraints of oxygen supply through the organismic respiratory cascade and of intracellular oxygen pressure.
ACKNOWLEDGEMENTS
This work was supported by a research grant from the University of Innsbruck. A.K. was supported by an Oroboros research award. We thank the Department of Zoology and Limnology, University of Innsbruck (Professor Dr R. Rieger) for the use of the electron microscopy facility. We are grateful for stimulating discussions and cooperation with Professor J.-P. Mazat (Bordeaux).