Microgravity simulation as a probe for understanding early Xenopus pattern specification

The amphibian egg displays a dramatic response to gravity: once activated, it rotates within the perivitelline space which surrounds the egg so that the darkly pigmented animal hemisphere opposes gravity. Internal cytoplasmic rearrangements follow egg rotation (Ubbels, Hara, Koster & Kirschner, 1983). Novel gravity orientation (e.g. egg inversion) has recently been employed as a probe for understanding the general features of the organization of the egg cytoplasm (Neff, Malacinski, Wakahara & Jurand, 1983). From those studies (e.g. Neff, Wakahara, Jurand & Malacinski, 1984), as well as others which employed 90° rotation (e.g. Gerhart et al. 1981), it appears that gravity may play a role in determining the manner in which the radial symmetry of the egg cytoplasm is reorganized into a bilaterally symmetrical format following gravity-driven rotation.

One way to test whether gravity plays a role in bilateral symmetrization would be to subject fertile eggs to a gravity-free environment. Besides the freefall weightless environment of space, eggs can be subjected to simulated microgravity conditions by rotation on a horizontal clinostat. A horizontal clinostat is an instrument which rotates a biological specimen about a horizontal axis at a slow constant speed. Horizontal clinostat rotation simulates microgravity by compensating for the influence of the gravity vector. Amphibian eggs subjected to horizontal clinostat rotation at a speed that does not in itself create appreciable centrifugal forces should theoretically experience a time-averaged force vector of 0 and therefore should not be able to get a biological fix on the g vector. In the appropriate speed range, the egg should respond to horizontal clinostat rotation much as it would respond to the hypogravity of outer space. Comparisons of various plant responses to clinostat rotation and weightlessness (space flight) have shown that there are very striking qualitative similarities between the clinostat and space flight (as well as small, perhaps insignificant quantitative differences) (Brown, 1979; and Brown, Chapman & Lin, 1974).

Clinostat experiments have been carried out by previous workers. Most of the earlier efforts were, however, inadequate -for one reason or another. Table 1 lists several of the earlier clinostat experiments. Limitations included a narrow range of clinostat speeds, rotation beginning after fertilization, failure to monitor dorsal/ ventral polarity, and failure to consider batch-to-batch variability.

The ideal clinostat experiment would have the following characteristics: rotation beginning at the time of sperm penetration; eggs oriented on horizontal axis of clinostat with respect to the sperm entrance site (SES); employment of complete range of rotation speeds; comparison of the effects of short- and longterm rotation; performance of vertical axis control experiments; and analyses of egg cytoplasm characteristics included in data interpretations.

Unfortunately, all of those features cannot be incorporated into the design of a single clinostat experiment. Four clinostat procedures (see Materials and Methods) were therefore employed. Several improvements over previous microgravity simulations were achieved: (1) the use of four separate versions of the clinostat permitted appropriate comparisons with data collected in earlier studies; (2) an expanded speed range (96 000 fold) was tested; (3) in one version, eggs were fertilized on the clinostat; (4) the effect of very slow speeds (e.g. 0·45r.p.h.) on egg polarization was monitored; (5) Xenopus laevis eggs, which have been chosen for future space shuttle experimentation, were employed; (6) characterization of the egg cytoplasm was performed to control batch-to-batch variability; and (7) in contrast to most previous clinostat investigations, the present authors consider the normal development of eggs on horizontal clinostats as a ‘positive result’.

The results of the present studies serve two purposes: first, they provide a data base for further refining egg polarity models that have been proposed to account for amphibian egg bilateral symmetrization; and second, they allow a prediction as to whether Xenopus eggs will display normal pattern specification in the microgravity of the space shuttle experiments presently contemplated by the European and American space agencies (see Materials and Methods).

Xenopus laevis eggs were chosen as an experimental system. They have recently been employed extensively for studies on egg polarity (reviewed by Malacinski, 1984b), and are scheduled to be employed in space shuttle experiments by the European and American space agencies (ESA: Ubbels et al. 1983; NASA: Souza, K. A., personal communication). Collection of eggs, fertilization, and egg manipulations followed standard procedures (e.g. see Neff et al. 1984).

The key features of each of the four clinostat protocols are illustrated in Fig. 1. It was considered essential to examine the performance of eggs in each of the four different clinostat protocols. Variables among the protocols were: time of clinostat rotation initiation; whether eggs were immobilized or allowed to rotate freely; and rotation speeds. The nomenclature for each of the three general procedures is: FreRo (free rotation, within jelly coats) -loaded into cellophane bags and fertilized; NaRReO (natural rotation, re-orientation) eggs fertilized, dejellied, and then immobilized in gelatin prior to initiation of rotation; PreFO (prefertilization orientation, within jelly coats) -eggs oriented prior to fertilization.

The FreRo protocol has several advantages over the others for the following reasons: easy to perform; can be carried out with large numbers of eggs; and rotation can be initiated very early (version (a) approx. T = 0·1 and version (b) T = 0·0). Because the eggs are not immobilized and thus are free to rotate in their perivitelline spaces, this procedure is not valid at speeds slower than the egg’s intrinsic rotation rate (< 30r.p.h.) or at the higher speeds, where eggs may orient with the centrifugal field (>24r.p.m.). In the FreRo (a) version, newly stripped eggs were placed in small dry cellophane bags (Nalgene cat. no. 500–0100) and fertilized. Ten minutes after fertilization the bags were filled with 20 % Steinberg’s solution and heat sealed. The bags were then attached with their long axis parallel to the clinostat shafts. The eggs remained attached to the inner surface of the bag (via the jelly membranes) throughout the rotation. The bags contained enough 20 % Steinberg’s solution for embryonic survival. In the FreRo (b) version, newly stripped eggs in cellophane bags were placed on the shaft of the clinostat. A 1 ml plastic syringe containing 0·-5 ml sperm suspension and a 5 ml plastic syringe containing 4-0ml 20% Steinberg’s solution (both with no. 27 needles) were then attached to the clinostat shaft so as to pierce the sides of bag. After 2 min of clinostat rotation the eggs were fertilized on the rotating clinostat. At 10 min postfertilization, the 20 % Steinberg’s solution was added. Air bubbles were removed with the 5 ml syringe.

The NaRReO protocol immobilizes eggs by shrinking their perivitelline spaces after fertilization and their natural rotation response. This procedure permits a wider range of clinostat speeds and allows the experimenter to orient eggs on the clinostat with respect to the egg’s SES. The disadvantages of this procedure are that eggs cannot be rotated on the clinostat before T = 0·35, and large numbers of eggs cannot be manipulated. Eggs were dejellied in 2·5 % cysteine-HCl (pH 7·6) after they had naturally rotated within the perivitelline space (following fertilization). Eggs with SES were then placed, animal hemisphere opposing gravity, in plastic petri dishes (3·5 cm diam.) and flooded with 6% (w/v) gelatin (2 parts 175 Bloom: 1 part 300 Bloom in 20 % Steinberg’s solution). The dishes were placed on melting ice for approximately 2·5 min to solidify the gelatin. The dishes were then mounted on the horizontal axis of the clinostat. Eggs were removed from the gelatin prior to gastrulation by placing the dishes on a 32°C warm plate for approximately 5 min to liquify the gelatin. Eggs were then permitted to develop in 20% Steinberg’s solution. For polarity studies, NaRReO eggs were oriented with respect to the SES, and were scored for the dorsal lip location in gelatin.

The PreFO protocol immobilizes eggs after fertilization but before the egg’s natural rotation response. This procedure allows for a wide range of clinostat speeds and for an early initiation of clinostat rotation (approx. T = 0·1). Eggs in their jelly coats were manually oriented (forceps) with animal hemispheres opposing gravity in the plastic dish, then fertilized and immobilized in gelatin as in the NaRReO protocol. The gelatin was removed manually with forceps prior to gastrulation.

All experiments were performed at 15°C in an environmentally controlled room. At that temperature, survival is enhanced (Neff et al. 1983), and the interval between fertilization and first cleavage is prolonged. That expanded interval permitted more precise timing and the use of larger numbers of eggs than would be possible at room temperature. In tabulating the data, uncleaved eggs were not considered.

Non-clinostated FreRo, PreFO, and NaRReO eggs as well as eggs clinostated on a vertical axis at 1, 2, 6,12, and 300r.p.m. (depending on experiment) were used for controls.

The ‘failure speed’ was defined as the speed at which 50 % of the clinostated embryos failed to complete gastrulation.

The external morphogenesis (gross morphology) of the clinostated embryos was observed at the tailbud stage. Clinostated embryos that were morphologically indistinguishable with respect to axial development were considered normal. Internal morphogenesis was not monitored.

Eggs inverted prior to fertilization were immobilized in 20 % Ficoll 6 min after fertilization as previously described (Neff et al. 1984). At T = 0·50 inverted PreFO eggs fixed in PBFG (4% formalin, 2·5 % glutaraldehyde in 0-05 M-phosphate buffer, pH 7-4) for 4 to 18 h and were sliced (razor blade) midsagittally along the plane of egg tilt in 100% Steinberg’s solution. The thickness of the subcortical vitelline layer (SVL) and residual vitelline mass (RVM) (Neff et al. 1984) at the gravity opposing apex of 150° to 180° inverted eggs was measured directly at x38 (dissecting microscope with low-angle illumination). The following classification of cytoplasmic mobility was utilized: High-mobility cytoplasm (HMC) inverted eggs had a combined RVM and SVL thickness of <67 μm. Intermediate mobility cytoplasm eggs (IMC) had a RVM and SVL thickness of between 67 and 133 μm. Low-mobility cytoplasm eggs (LMC) had a RVM and SVL thickness of >133 μm.

In order to determine the range of useful speeds which mimic hypogravity, analyses were carried out on the rate at which internal cytoplasmic components (e.g. yolk) shift in response to abnormal gravity orientation. It was assumed that in order to mimic microgravity, eggs immobilized on the clinostat (PreFO and NaRReO protocols) should rotate at speeds faster than the rate at which internal cytoplasmic components shift in response to novel gravity orientation.

Inverted eggs that were fertilized in the inverted position and then immobilized in that position were chosen as a model system for these preliminary determinations. The rate of shift of the major yolk compartments in a batch of intermediate cytoplasmic mobility (IMC) eggs at 15 °C was analysed by histological methods. Fig. 2A indicates that the large yolk compartment shifted at an average of 0·19r.p.h. (range: 0·15–0·24). Therefore, clinostat speeds faster than 0·2r.p.h. were employed for the PreFO and NaRReO protocols.

In the case of the FreRO clinostat procedure, a different assumption is necessary. Eggs are free to rotate within their vitelline membranes in the FreRo protocol. It was assumed that the minimum clinostat speed needed to mimic microgravity should be faster than the speed at which eggs rotate in their perivitelline spaces. Fig. 2B shows that inverted eggs rotated within their perivitelline spaces at an average speed of approximately 0·25 r.p.m. (15 r.p.h.) (range: 1·56– 31·7r.p.h.). Accordingly, clinostat speeds faster than 0-5r.p.m. (30r.p.h.) were employed in the FreRo version.

The optimal clinostat speed which simulates microgravity is difficult to calculate because of the vast number of unknowns concerning the microenvironment and physical properties of cellular components. Evidence indicates that crucial properties such as cytoplasmic viscosity, for example, change during egg activation (Elinson, 1983). That observation suggests that the optimal clinostat speeds may actually change during development. A whole range of clinostat speeds was therefore employed. The upper and lower limits of the proper clinostat speed range was determined by observation. The upper effective speed was 24 r.p.m. At 24r.p.m. the eggs in a few batches (3/50) oriented with respect to the centrifugal field, while at 100 r.p.m. all FreRo eggs oriented with respect to the centrifugal field. The lower effective speed is 1 r.p.m. At speeds slower than 1 r.p.m. the egg polarity was influenced by the direction of rotation (see Figs 5 and 6 below).

Neff et al. (1984) have emphasized the variability in response to novel gravity orientation (e.g. inversion) of different spawnings of eggs. The extent to which early pattern formation is reversed varied in proportion to the mobility of the major cytoplasmic density compartments. Three classes of eggs were designated according to internal cytoplasmic mobility: LMC (low-mobility cytoplasm); IMC (intermediate); and HMC (high).

Since the extent of pattern formation reversal displayed by inverted eggs varied among these three categories (Neff et al. 1984), a special effort was made to insure that each egg type, including both HMC and LMC versions, was tested in each of the three major clinostat protocols.

The frequency of normal external morphogenesis (tailbud stage) among approximately 3000 eggs which were clinostat rotated over a wide range of speeds by the PreFO protocol from approximately T = 0·1 to T = 3·0 is shown in Fig. 3. Because the PreFO rotated eggs were immobilized, clinostat rotation at low (< 1 r.p.m.) and high speeds (>24r.p.m.) was also possible. Substantial numbers of eggs developed normally over a wide range of clinostat speeds. From 0·25 r.p.h. to 12r.p.m. more than 50% of the embryos developed normally. At speeds of 24 r.p.m. and greater, however, more than 50 % displayed developmental abnormalities. At 24 r.p.m. approximately 60% of the embryos were abnormal at the tailbud stage. At 100r.p.m. approximately 70 % were abnormal. A wide range of abnormalities were observed. These included cleavage-furrow defects which resulted in cytolysis prior to gastrulation, incomplete gastrulation, endoderm oedema, and axis deformities such as microcephaly and tail defects. Since the major significance of the clinostat results is the observation that large proportions of the eggs develop normally, detailed descriptions of the nature of developmental abnormalities are not provided.

Of the approximately 1500 T = 0·1 fertile eggs clinostat rotated by the FreRo protocol, a substantial number showed normal external morphogenesis (Fig. 4). Because FreRo eggs are not immobilized, they can rotate within their perivitelline spaces below 1 r.p.m. and above 24 r.p.m. Eggs rotated by the FreRo method at low speeds, for example 15r.p.h., kept adjusting their animal vegetal axes to gravity on the horizontal clinostats. At 100 r.p.m. virtually all eggs oriented their vegetal poles centrifugally away from the horizontal clinostat axis. Therefore, over the useful range of clinostat speeds (1 r.p.m. to 24 r.p.m.) for microgravity simulation, Xenopus eggs typically display normal external morphogenesis.

Since they are unable to rotate to accommodate the enhanced centrifugal forces, PreFO eggs may undergo a breakdown in the cytoplasmic component(s) which generate pattern specification. Such breakdown may account for the lower failure speed of the PreFO protocol (i.e., PreFO failure speed = 24r.p.m.; FreRo failure speed = 200 r.p.m.).

A series of specific questions were answered in order to facilitate analyses and comparisons of data generated by the various clinostat protocols:

1. Do each of the four clinostat protocols generate similar data re. normal/ abnormal morphogenesis and failure speed? Generally, yes. The frequency of normal external morphogenesis was greater than 50 % for each protocol up to the speed of approx. 24 r.p.m. In the PreFO and NaRReO protocols the failure speed was usually 24 r.p.m. For the FreRo protocol eggs often displayed >50 % normal development up to 200 r.p.m.

2. The T = 0·l to T = 0·15 time interval that the typical FreRo and PreFO clinostated eggs spend at 1 g before gravity compensation does not affect normal external development. Eggs fertilized on the rotating clinostats (2·5, 6, 8, 10, 12, and 24r.p.m.) according to the FreRo (b) protocol developed normally. These results demonstrate that the typical PreFO and FreRo experiments in which clinostat rotation begins by T = 0·l to T = 0·15 mimic the ideal clinostat experiment.

3. Do all three cytoplasmic mobility classes of eggs respond similarly to clinostat treatment? Yes. At least two or more batches of each egg type (HMC, IMC, LMC) were tested in each of the three major clinostat protocols. No appreciable difference in survival frequency was detected among the three protocols. Likewise, the nature of the defects in morphogenesis observed at higher speeds (>24r.p.m.) was similar among the three protocols.

4. Does the amount of time on the clinostat affect the results? No. Eggs were taken off the clinostats (four versions; 1 r.p.m. –400r.p.m.) at either the first or third cleavage, or after overnight clinostating (embryos at late blastula stage). Their development was carefully monitored at the gross external morphological level. Neither the ‘failure speed’, nor the frequency of abnormalities, nor the pattern of development differed appreciably with regard to the time at which the clinostat treatment ceased.

5. Do eggs which are rotated around a vertical rather than horizontal axis respond differently to clinostat treatment? No. Eggs of each type (HMC, IMC, LMC) were routinely mounted on vertical axis clinostats (FreRo, PreFO, NaRReO). Speeds of 1, 6,10, and 300r.p.m. were usually employed. At speeds of 1–10r.p.m. development was typically normal. At 300r.p.m., however, virtually 100 % of the vertical clinostated eggs cytolysed during cleavage stages. These data indicate that the failure of eggs at high clinostat speeds may, in part, be due to an effect of the high-speed clinostats other than gravity compensation.

6. Do very slow (e.g. <1 r.p.m.) clinostat speeds mimic the polarization effect of 90° rotation? Yes. Previous reports (Scharf & Gerhart, 1980; Chung & Malacinski, 1980, 1981) revealed that a brief 90° rotation of uncleaved eggs establishes the dorsal side in the region of the egg which opposed gravity during the rotation period. Using the NaRReO protocol it was possible to orient eggs with respect to the sperm entrance site on the horizontal clinostat. Eggs were placed on clinostats with their animal hemispheres facing gravity and their SESs perpendicular to the horizontal clinostat axis. Rotation was carried out at 0·1,0·5, 1·0, 2·0, 6·0 and 12·0 r.p.h. with the SES either leading or trailing in a plane perpendicular to the clinostat axis. At the dorsal lip stage, the eggs were scored for location of the involution site with respect to the SES. Fig. 5 shows that at the slow speeds (0·1 to 12·0r.p.h.), the polarity (involution site) was determined by the direction of rotation.

7. At clinostat speeds that simulate microgravity, is the invagination site (d/v polarity) strictly determined by the SES? No. Using the NaRReO protocol at speeds of 2r.p.h.-12r.p.m., the location of the dorsal lip with respect to the SES was monitored. Fig. 6 reveals that the very slow clinostat speed (2r.p.h.) mimics the effect of 90° rotation (clinostat rotation determined egg polarity). At higher speeds (1 and 2 r.p.m.) the natural polarization of the eggs with the invagination site 180° opposite the SES was observed. At speeds of 8 and 12r.p.m., however, the site of the dorsal lip was dissociated (randomized) with respect to the SES. The following controls indicate that this randomization of polarity was not a side effect of clinostat rotation: (1) the shift of polarity at 2r.p.h. and 2 r.p.m. provided the internal controls; (2) eggs rotated on vertical clinostats at 2r.p.h., r.p.m., and 12r.p.m. displayed the majority of involution sites 180° opposite the SES. It appears that over the wide range of clinostat speeds that best mimic microgravity (1–12r.p.m.), fertile eggs behave in two distinct ways with respect to the d/v polarity. At low speeds (1–2r.p.m.), the polarity was normal with the dorsal side of the embryos 180° opposite the SES. At the higher clinostat speeds (8–12 r.p.m.), the eggs displayed abnormal (random) polarization. This randomization or reorientation of polarity was not determined by either the SES or the direction of clinostat rotation.

Because the fertilized amphibian egg displays a distinct rotation response to gravity, it has been considered feasible that gravity might drive internal cytoplasmic rearrangements. Previous attempts at using the clinostat to cancel the effects of gravity have, however, been incomplete for several reasons, including employment of a narrow range of clinostat speeds, rotation beginning after fertilization, and dorsal/ventral polarity not being monitored. In the present study a variety of clinostat protocols designed to correct those inadequacies were employed, including a novel one (FreRo (b)) which involved actually fertilizing the eggs during clinostat rotation. By comparing the results of the various procedures it is apparent that at the useful clinostat speeds which gravity compensate, early pattern specification is normal in a majority (>50%) of eggs. These positive results taken together with the previous published clinostat experiments (see Table 1) and with the normal development of Rana eggs placed in the microgravity environment of space at around T = 1·0 (Young & Tremor, 1968) predict that Xenopus eggs fertilized and allowed to develop in the microgravity of space will develop normally.

Since Xenopus eggs display differences in the apparent viscosity of their cytoplasm (Neff et al. 1984), it was considered important that all types of eggs (HMC to LMC) be represented in the data. Accordingly, the cytoplasmic mobility of each batch of eggs was quantified. Even HMC eggs, which might be expected to show the greatest sensitivity to gravity effects, developed normally throughout the same broad range of speeds at which LMC eggs developed normally with respect to external morphogenesis.

Various models have been proposed which explain the manner in which the radial symmetry of unfertilized egg cytoplasm about the animal/vegetal axis is reformated to generate bilateral symmetry (dorsal-ventral polarity) in the zygote. The salient features of three of those models are illustrated in Fig. 7. In each instance a rearrangement of major cytoplasmic regions (‘density compartments’; Neff et al. 1984) occurs. In each model the effects of the zygote’s dorsal/ventral polarity on later pattern specification (e.g. mesoderm induction) are visualized to emerge from the novel compartment/compartment interactions which result from the breaking of the radial symmetry of the egg derived from oogenesis (reviewed by Malacinski, 1984a). The predicted behaviour under microgravity conditions is included. Should significant gravity-driven rearrangements of the cytoplasm be a prerequisite for egg polarization, as, for example, the ‘gravity propulsion’ model B in Fig. 7 predicts, early embryogenesis of clinostat rotated eggs would be abnormal. Conversely, normal pattern specification in clinostated eggs would favour the ‘push-pull’ and ‘compartment shift’ models A and C shown in Fig. 7.

In the ‘gravity propulsion’ model B (Fig. 7), some compartments built into the egg cytoplasm during oogenesis are not stratified according to buoyant density. Following egg activation (fertilization), the fluidity change of the cytoplasm (Elinson, 1983) permits gravity-driven compartment rearrangements to occur (Malacinski, 1984b ; Neff et al. 1984). The data derived from the PreFO and FreRo clinostat experiments, which showed substantial normal development of clinostated eggs over the clinostat range that mimics microgravity, reveal that the ‘gravity propulsion’ model is not appropriate for describing the forces which rearrange the egg cytoplasm during the interval between activation and early cleavage.

In the ‘push pull’ model A (Fig. 7), the growth of the sperm aster away from the sperm entrance site (SES) physically pushes the cytoplasm compartments. Subsequent cortical contraction toward the SES pulls up yolk compartments on the future dorsal side (e.g. Gerhart et al. 1981; Ubbels et al. 1983). Although this model may be applicable at 1g (Ubbels et al. 1983) and at the lower clinostat speeds (1 and 2 r.p.m.) this mechanism does not necessarily provide indispensable driving forces for the cytoplasmic rearrangements which are necessary for providing the eggs with a dorsal/ventral axis. Clinostated eggs at 8 and 12 r.p.m. (Fig. 6) displayed an uncoupling of the SES from the future dorsal side. Clinostating does not represent the only way in which D/V polarity can be uncoupled from the SES. In inverted eggs the direction of shift of the major yolk components rather than the SES determines D/V polarity (Neff et al. 1984). In Rana escalenta the tilt of the animal hemisphere with respect to the animal/vegetal axis, rather than the SES, determined the dorsal/ventral polarity (Pasteels, 1937) of selected normally oriented eggs. Other examples of uncoupling of SES and D/V polarity include: 90° rotation and lateral centrifugation (Gerhart et al. 1981); and as little as 15° tilt (Malacinski & Neff, 1985).

The data collected in this series of clinostat experiments are valuable in the following ways: They have predictive value with respect to future microgravity space flight experiments. They exclude model B and place model A in doubt. They cause us to think about amphibian egg pattern specification in novel ways such as model C (Fig. 7). This model is heavily based on the data presented in this paper, on previous compartment data produced by this laboratory (Neff et al. 1984), and on experimental manipulations of D/V polarity such as egg rotation (Gerhart et al. 1981). This model requires no single indispensable driving force to bring about the compartment shifts that allow for the asymmetric compartment localizations and/or interaction (activations) (Neff et al. 1984). Rather, multiple, overlapping, sequential forces, some of which may be dispensable, establish the necessary compartment asymmetry. The desired end result, compartment asymmetry, is the end state that the egg requires to establish a normal dorsal/ventral axis. Therefore, several combinations of egg properties and/or forces such as intrinsic egg asymmetries (Pasteels, 1937), viscosity changes (Elinson, 1983), aster growth, cortical contraction (Ubbels et al. 1983), egg tilt (Malacinski & Neff, 1985), and gravity orientation (Gerhart et al. 1981) etc. may be sufficient. Although this model might be considered vague and perhaps less susceptible to simple direct experimentation, it emphasizes the complexity of the problem of understanding egg polarization.

This investigation was supported in part by NASA grant NAGW-60, and an international scientific collaboration grant from NSF (INT 82-17344) and KOSEF (Korea Science and Engineering Foundation) to G.M.M. and H.-M.C. Mr Myeong Sok Lee provided excellent technical assistance. Drs K. Souza (Ames), G. A. Ubbels (Utrecht), and J. Gerhart (Berkeley) kindly provided reviews of the manuscript.