Desiccation Tolerance and Water-Retentive Mechanisms in Tardigrades

Several workers have discussed the importance of humidity-related parameters in tardigrade ecology (Beasley, 1978; Hallas, 1977,1978; Nelson, 1975; Ramazzotti & Maucci, 1983). However, no attempts have been made by these or other workers to relate ecological trends to aspects of tardigrade physiology. The possibility of interspecific differences relating to their extreme desiccation tolerance or anhydrobiosis (cryptobiosis) (see Keilin, 1959; Crowe, 1975) is especially attractive since this process has allowed the primarily aquatic phylum to adapt to increasingly xeric habitats.

The induction of anhydrobiosis involves evaporation of liquid water surrounding the tardigrade and ensuing desiccation. It has long been known that tardigrades will only enter anhydrobiosis successfully if desiccated in high humidities (Pouchet, 1959). In a typical habitat such as a moss cushion, static air trapped between the individual gametophytes will probably make the dispersion of water vapour a very slow process. However, when exposed to direct insolation, mosses may desiccate very rapidly, especially if growing on a low heat capacity substrate, and transpiration is quickly arrested by leaf-folding (von Zanter, 1974). It is therefore likely that the transition from 100% relative humidity (RH) to low humidity may, on occasion, be quite rapid. Desiccation tolerance during the initial stages of dehydration is thus likely to be of critical importance in survival and habitat determination. Unfortunately, the technology required to record humidity within the confines of a moss cushion is not yet available.

A true experimental mimic of the natural desiccation regime within a moss cushion would involve gradual reductions of RH from 100%, and probably in nonlinear patterns. Such graded humidity falls would be too exacting to reproduce in any sort of experimentally standardized way. Crowe (1972) studied desiccation tolerance in Macrobiotus areolatus Murray and used a range of controlled humidities to dry animals at constant rather than increasing rates. In addition, using different durations of drying in 95% RH and then transferring animals to dry air, he was able to assess the period of high-humidity desiccation necessary to confer resistance to low RH. Similar procedures are employed here to compare desiccation tolerance in a range of common tardigrade species.

This study also investigates aspects of anhydrobiosis that may determine desiccation tolerance. Structural adaptations of anhydrobiosis have been recognized for over a century (Broca, 1860). In the tardigrades and the bdelloid rotifers, induction of anhydrobiosis involves an anteroposterior contraction to form the tun. Tardigrades dried in low relative humidities or under anoxia do not form tuns, collapsing into an irregular, flattened form, and do not revive (Crowe, 1972), suggesting that contraction into the tun is metabolically dependent (see also Collin & May, 1950). Crowe (1975) has suggested that tun formation serves to reduce the transpiring surface area and to remove more permeable intersegmental cuticle from the exterior.

Comparative cytological studies of anhydrobiotic and hydrated tardigrades (Hickernell, 1917; May, 1946; Crowe, 1975) have failed to demonstrate any fine structural changes related specifically to anhydrobiosis. However, Crowe (1975) reported biochemical changes in anhydrobiotic tardigrades, with breakdown of lipid and glycogen in the cavity cells (‘coelomocytes’) and a concomitant increase in intracellular concentrations of trehalose and glycerol. Subsequent studies have detected trehalose and glycerol in hugely elevated concentrations in many anhydrobiotic organisms (Crowe et al. 1984a,b), and it is now well established that these sugars serve as dehydration protectants and cryoprotectants of proteins an phospholipid bilayers (Crowe et al. 1987). When dehydrated, phospholipid bilayers undergo liquid crystalline-gel phase transitions (Chapman et al. 1967; Kodama et al. 1985) with fusion of adjacent bilayers and displacement of integral membrane proteins (Crowe et al. 1983). Trehalose and other membrane protectants appear to replace the integral water of lipid bilayers, stabilizing lipid transition temperatures and preventing gel formation (Crowe et al. 1987).

Although the quantitative significance of trehalose, glycerol and the other membrane protectants is poorly understood, work on the nematode Aphelenchus (Crowe & Madin, 1975) has shown that accumulation of trehalose continues for at least 72 h during desiccation in 97% RH and glycerol continues to accumulate after this. Trehalose has been reported to comprise over 20% of the dry mass of desiccated cysts of the brine shrimp Artemia salina (Clegg, 1967), which also demonstrate anhydrobiosis. The time and metabolism required to mobilize reserves and synthesize these protectants makes it imperative that the animals are able to retain their normal, hydrated state, and thus sustain normal metabolism, for several hours at least. Therefore, dehydration must either be very slow, or anhydrobiotic organisms must possess mechanisms for retaining water in more severely desiccating conditions.

Given the considerable desiccation rates tolerated by tardigrades (Crowe, 1975), the physical process of tun formation and/or permeability changes in the cuticle during drying are likely to play a crucial role in water retention. Such processes may thus provide clues to the factors underlying differences in desiccation tolerance among species.

The following tardigrade species were studied: Echiniscus testudo Doyère, Macrobiotus hufelandi Schultze, Macrobiotus richtersi Murray, Milnesium tardigradum Doyère, Hypsibius (H.) dujardini Doyère, Hypsibius (H.) oberhaeuseri Doyère and Hypsibius (I.) prosostomus Thulin. Routine abbreviations are used in the tables: E.t., M.h., M.r., M.t., H.d., H.o. and H.p., respectively. All animals were hydrated in habitat plants (mosses or lichens) for a minimum of 24 h before each experiment.

The dehydration apparatus consisted of a 5 ml watch glass, with a mimimal volume of water (0·05–0·1 ml) containing the tardigrades, placed in a 0·121 glass desiccating chamber holding 20 ml of KOH solution. A glass lid was then sealed onto the chamber using high-vacuum silicone grease (BDH chemicals Ltd, Poole, England). The KOH solutions were made up according to Solomon (1951) to give a graded series of humidity controls between 50% and 90%. Freshly made KOH solutions were sealed and cooled to room temperature, then poured into the desiccating chamber.

Preliminary trials using a capacitance hygrometer probe [HMI31 indicator and HMP31UT probe, Vaisala (UK) Ltd], connected to a nine-channel d.c. chart recorder [Grant Instruments (Cambridge) Ltd] showed that the equilibration of humidity (RH) within each chamber required up to 1 h for high humidities and up to 2 h below 80% RH. The volume of water containing the tardigrades was judged to give adequate time for RH equilibration before it had evaporated. All trials were conducted at laboratory temperature (18·7–21·4°C).

Following evaporation of water in the watch glass (discernible by eye), the tardigrades were left exposed to the controlled RH in the chamber for 1h. The watch glass was then removed and 2 ml of distilled water immediately added to the desiccated animals. Revival, observed using a 20x binocular microscope, was judged as the return of the normal, direct locomotor rhythm of the lobopodia, protraction of each lobopodium beginning at the posterior and progressing anteriorly, opposite lobopodia 180° out of phase. Four experiments, each using eight animals per watch glass, were performed for each species at each RH and revival times were recorded.

Tardigrades were dehydrated, as described above, in 85% RH. Following various periods of this high RH desiccation, watch glasses were removed from the chamber and the animals subjected to further dehydration in the laboratory at 25–31% RH. (Transferring the watch glasses to different KOH solutions is not a practicable method of changing the RH because of the long equilibration times involved). Animals were allowed to dehydrate for 15h in low RH before being rehydrated with distilled water as before and observed for revival. Two experiments, each using eight animals, were performed simultaneously for each species over each time range. Mean laboratory humidity during each pair of experiments was recorded using the Vaisala capacitance hygrometer as before.

Scanning electron micrographs (SEMs) of tuns of Hypsibius oberhaeuseri are shown in Fig. 1A,B. Tun formation is clearly a complex process, involving retraction of the lobopodia, infolding of the intersegmental cuticle, and a degree of longitudinal infolding as well. These processes appear to be fairly standard in all species studied, with the exception of Echiniscus testudo, where the inflexible and heavily sclerotized dorsal plates impose certain constraints on tun morphology. SEMs of extended and partially contracted (preliminary stages of tun formation) animals are shown in Fig. 2A,B for comparison. Again, the proportions are fairly standard across the species studied, with the exception of E. testudo. Extended animals show an oval profile in transverse section, with a height: width ratio of approximately 0·7:1.

On the basis of such SEM studies, extended animals and tuns were considered as hemicylinders; the additional surface area attributable to the eight small lobopodia in extended animals was ignored. Active animals were placed in individual drops of water on a microscope slide and then allowed to desiccate in a sealed chamber, using aqueous potassium hydroxide to control the relative humidity at 85%. Following evaporation of the water drops, the animals were left to dehydrate for 4h and the tuns then measured under 400× magnification using a calibrated eyepiece graticule. Following this, the animals were rehydrated and allowed to revive. Once active, they were narcotized by addition of a small quantity of 10% aqueous magnesium chloride and remeasured.

The procedure adopted for measurement was as follows. Length (L) was measured from the anterior tip to the junction between the posterior lobopodia. Width (d) was recorded as the mean of the intrasegmental regions of segments 2,3 and 4, corresponding to the first, second and third pairs of lobopodia. The surface area (SA) was then calculated as that of a hemicylinder, excluding the ventral surface which is in contact with the substrate and thus isolated from a desiccating atmosphere. Twenty-four replicates for tun SA and extended SA were measured for the seven tardigrade species studied.

Samples of 15–30 hydrated tardigrades were allowed to desiccate on preweighed foil dishes in the weighing chamber of a Perkin-Elmer AD-4 autobalance (sensitive to 0·1 μg). Relative humidity within the chamber was regulated at 80% RH, again using aqueous KOH, and allowing a minimum of 3h for equilibration. The initial volume of the water drop containing the tardigrades was prejudged to allow at least this period in the balance chamber before complete evaporation.

Net mass of the tardigrade sample was recorded at 15-s intervals for the first 10 min, from just prior to the disappearance of the water drop, and subsequently at 5-min intervals until 30 min had elapsed. (Loss of standing water was discernible by eye to an accuracy of approximately ± 10 s and was confirmed by the arrested movement of the animals.) After 30 min, specimens were removed with the foil, and balance drift was then subtracted from/added to the initial foil mass; drift never exceeded 5 μg. Mean individual masses of the tardigrades were then calculated and dry masses were recorded after 60 h at 70°C in a desiccating oven. Five such weighing trials were performed for each of five species (E. testudo, M. hufelandi, M. richtersi, H. oberhaeuseri and H. dujardini).

Similar methods were adopted in recording mass loss, during desiccation in 80% RH, of dead animals, unable to form tuns. These had been freshly killed by immersion in water at 50°C for 10 s. This should cause minimal damage since all species could survive prolonged immersion in water at 42 °C (personal observation) and temperature-induced phase-changes of lipids, which may have consequences for permeability, are generally reversible on cooling (Matabayasi et al. 1986).

To assess the lethal effects of very rapid desiccation, animals were weighed during drying in low humidities (25–30%), by leaving the balance chamber open to equilibrate with laboratory humidity. The humidifying effect of the water drop could be almost eliminated by using a minute amount of water since it was not necessary to allow time for humidity equifibration following closure of the balance chamber. Four such experiments were performed in this RH range using one species, Macrobiotus richtersi. Subsequently, dry masses were also recorded for these animals.

To investigate the effect of high RH dehydration on subsequent water losses in low RH, tardigrades were desiccated in 80% RH in the balance chamber, and mass losses recorded for 100 min, before removal of the regulating KOH solution and exposure to low RH. The high-humidity air in the chamber was flushed with a small fan during which time the animals were covered with a small shield to ensure that this air movement did not contribute to desiccation. Once the RH inside the chamber had equilibrated with that of the laboratory (measured with the Vaisala HMI31 probe), it was again sealed and the mass recordings continued in low RH. Two such experiments were performed using M. richtersi; these were then repeated allowing just 15 min of dehydration in 80% RH before flushing the chamber to low RH.

For all weighing data, results were converted to percentage of initial (individual) mass, taking this value (W) as the mean individual mass at the moment when standing water was judged to have disappeared. In every experiment, the mass had stabilized after 20 min from this moment and subsequent mass loss was undetectable. Mass loss curves of the percentage of the initial mass against time showed a hyperbolic fall, linear in semilogarithmic plot, with the rate of mass loss suddenly declining very sharply after a few minutes. Such a semilogarithmic curve is shown in Fig. 3A. Xd is the point where the rate of mass loss (as log% W/time) begins to decline, and R is the stable residual mass (taken at 30 min) giving an indication of the mass of water retained in the tun. T_1/2 values taken from the initial minute o| dehydration give a measurement of the initial rate of mass loss. All T_1/2 values are measured from the semilogarithmic plots, but other values (Xd, R) are backtransformed and expressed as percentages of the initial mass.

The rate of decline of water loss between Xd and R is a function of the permeability fall as the animal dries. This decline of water loss was calculated by subtracting the R value from the percentage weight loss values from point Xd onwards. A semilogarithmic plot of the residuals gives the profile shown in Fig. 3B. This curve was observed for all species and suggests a two-stage decline in water loss after Xd (phases DI and D2). DI represents a very rapid decline in cuticle permeability or permeability slump. D2 represents the subsequent, much lower, rate of permeability fall. Best-fit curves were plotted for these two regions for the species studied, and T values of the curves calculated to give their rates of decline in water loss (T_1/2 DI and T_1/2 D2).

VD internal may be assumed as the saturated vapour density of water at the animal’s temperature since changes of vapour density with large changes in osmotic pressure are very small (Gilby, 1980). The surface area estimates of the extended animals and tuns are described above; tun formation was assumed to be complete by point Xd. Vapour density values were taken from the tables given in Unwin (1980), assuming a mean laboratory temperature of 18°C during the experiments. The value obtained for R comprises the transpiration barrier provided by the cuticle and by the still layer of high vapour pressure air surrounding the animal, and is used as an exaggerated estimate of the cuticle permeability. The experimental precautions taken to reduce accumulations of saturated air in the vicinity of the animals are described above.

R (% of W), final mass as a percentage of initial mass (equivalent to water content).

Xd (% of W), mass at which permeability slump begins (as a percentage of initial mass).

R/Xd, the ratio of the final mass to the mass at which the permeability slump begins; an estimate of the fraction of water retained following Xd.

Time to Xd, time between initiation of dehydration and Xd (seconds).

T_1/2 initial, half-time of the initial rate of desiccation (seconds).

T_1/2 DI, half-time of the rate of decline in mass loss during the permeability slump (DI) (seconds).

T_1/2 D2, half-time of the rate of decline in mass loss following the permeability slump (D2) (seconds).

P₁₅, permeability after 15 s of dehydration.

P_A, permeability at the beginning of the permeability slump (beginning of DI).

P_B, permeability at the end of the permeability slump (end of DI).

PT_1/2, rate of decline of permeability during the permeability slump.

P_A/P₁₅, the ratio of P_A to P₁₅, an approximation of the factor of permeability decline between initial desiccation and Xd.

P_B/P_A, the ratio of P_B to P_A, an approximation of the factor of permeability decline during the permeability slump.

Clear differences in tolerance of desiccation are demonstrated among species (Table 1). Lower lethal humidities for initial dehydration are 75–80% RH for H. dujardini, and are only slightly lower for M. hufelandi, M. richtersi and H. prosostomus, whereas M. tardigradum and E. testudo can withstand 60% and H. oberhaeuseri can tolerate as low as 50%. Estimates of the humidities causing 50% mortality (LD₅₀ values) were taken from plots of these data and are shown in the same table. Mean revival times were plotted for each species and an example (M. richtersi) is shown in Fig. 4. Following high-humidity desiccation, most animals revive in 8-16min. However, revival times show considerable variation and become distinctly longer and more variable after more rapid dehydration. As the lower lethal humidities are approached, recovery sometimes takes several hours. Small intraspecific differences between the lower lethal humidities thus result in very large variances about the mean recovery times.

The durations of dehydration in 85% RH required to confer resistance to low (laboratory) RH are shown in Table 2 and show a similar trend to the initial desiccation tolerances. Those species least tolerant of high initial desiccation rates also require more prolonged pretreatment in high RH before they are able to survive exposure to much lower humidities. H. dujardini requires 150–200 min of dehydration in 85% RH before it can withstand exposure to low RH whereas, at the other extreme, H. oberhaeuseri requires only 40–60 min.

An example plot of mean revival times (M. richtersi) is shown in Fig. 5. Recovery of animals took significantly longer than in the first batch of experiments, but shows the same trend of increasing time and variability as the lower lethal period of pretreatment at 85% is approached.

The ratios of tun surface area: extended surface area for each species are shown in Table 3. The heterotardigrade E. testudo shows a significantly smaller surface area reduction than is seen in the other (eutardigrade) species, with a tun SA: extended SA ratio of 0·77:1. This can probably be attributed to the very thick and inflexible cuticular plates that cover the dorsal surface. Eutardigrade species show SA ratios of the order of 0·5:1, and this ratio decreases with increasing desiccation tolerance. If E. testudo is omitted from the data, there is a significant positive linear regression of surface area reduction against lower lethal humidity for instantaneous drying, shown in Fig. 6 (t = 3·076, P = 0·044). It thus appears that those species most tolerant of desiccation undergo the greatest reduction in surface area during tun formation.

Fig. 7A shows a semilogarithmic plot of mass loss curves for M. richtersi dried is 80% RH. All species showed the abrupt decline in rate of water loss described in the analysis section. A semilogarithmic plot of residuals of this decline in water loss following point Xd is shown in Fig. 7B; data from all such curves are listed in Table 4. Cross-species comparisons of these data were studied by means of /-tests (Table 5A-D). There were no significant differences between the species when the dry masses were expressed as a percentage of initial (wet) mass (W), so the differences between species in the other weighing data are assumed to represent temporal differences in water content.

In Table 5A, R values indicate that H. oberhaeuseri and E. testudo retain Significantly more water than M. hufelandi, H. dujardini and M. richtersi. M. richtersi retains significantly more water than M. hufelandi and H. dujardini. No significant differences are seen between H. oberhaeuseri and E. testudo. The relationship between water retention and desiccation tolerance was tested using a linear regression analysis (Fig. 8) and found to be significant (t=–3·508, P = 0·038).

Relating to the differences in water content among the species are differences in the Xd value where the permeability slump is initiated (Table 5B). Again, the values for H. oberhaeuseri and E. testudo are significantly higher than for M. hufelandi, and that for H. oberhaeuseri significantly higher than that for H. dujardini. It thus appears that H. oberhaeuseri and E. testudo retain more water by effecting a permeability decfine in their cuticles at an earlier stage in their dehydration. Very little difference is apparent among species in the R/Xd ratios (Table 5C), suggesting that following initiation of the permeability slump, similar water losses occur in all species.

No clear trend emerges from the initial rates of desiccation of the species (Table 5D). M. richtersi is generally significantly slower to lose water at this stage than the other species and H. dujardini always significantly faster. The duration of dehydration before initiation of the permeability slump (time to Xd) gives a similar picture. These trends would appear to correlate with the sizes of the species -M. richtersi being the largest and H. dujardini the smallest species on average (see dimensions in Table 3) - though this cannot be rigorously tested since the animals used in each weighing trial were not individually measured beforehand. The higher surface area: volume ratio of a smaller species would, however, account for more rapid dehydration when measured as percentage wet mass per unit time.

Rates of decline in mass loss following Xd (T_1/2 DI and T_1/2 D2) are rather variable and do not show any clear relationship to the other variables. T_1/2 DI values are smallest for H. dujardini - usually significantly lower than for the other species - which may relate to the rapid dehydration observed in this species. This would be the expected situation if the permeability decline were, itself, dehydration-dependent. However, M. richtersi, which shows the slowest desiccation, also has a relatively rapid T_1/2 DI value. T_1/2 D2 values are similarly ambiguous, showing no clear trend across species, though by this stage the mass loss rates of all species are very low.

To express this information as cuticle permeabilities, a representative permeability curve for M. richtersi is shown in Fig. 9. The methods for calculation of such curves are described in the analysis section. All species show a characteristic permeability slump corresponding to stage DI on the dehydration curves (Fig. 7A). Variables taken from such permeability curves are fisted in Table 6. Permeabilities are listed at 15 s into desiccation (P15), at point A marking the onset of the permeability slump (PA), and at point B marking the end of the permeability slump (PB). Cross-species comparisons were not made since the variables were only taken from the dehydration curves of single animals, and corresponding data are shown in Table 5. All species, however, show a high initial permeability, which falls during tun formation (up to point A) and then slumps rapidly to point B, declining by a factor of about 10². PT_1/2 values giving the rate of permeability decline during the slump vary from 13 to 43 s and correspond to the DI values described above. Permeability following point B is extremely low and continues to decline thereafter at rates of T_1/2 ≈ 5min. These T_1/2 value correspond to the T_1/2 D2 values listed in Table 4.

Semilogarithmic plots of mass loss for dead animals in 80% RH are shown in Fig. 10. Again, the examples shown are of M. richtersi. All species show similar curves to those resulting from dehydration of active animals (see Fig. 7A). However, the mass loss profile now shows an increasing rate of decline in semilogarithmic plot and the abrupt reduction in rate of water loss is seen at Approximately 18% of the initial mass. Extrapolated plots of residuals following Xd were very similar to those obtained for the corresponding dehydration of active animals (see Fig. 7B). Variables from these curves are listed in Table 7 and compared between species using t-tests in Table 8A-E. Where figures are not included, masses could not be reliably interpreted owing to balance drift.

Again, H. oberhaeuseri and E. testudo when dead retain the most water and H. dujardini and M. hufelandi the least (Table 8A). Similar trends are seen in the Xd values (Table 8B) although the initial desiccation rates (T_1/2 initial) and those at Xd (T_1/2 at Xd) suggest a relationship to size of species, as with active animals. Variables taken from permeability curves for dead animals are listed in Table 6. Initial permeabilities of dead tardigrades are higher than for living animals, with lower A values (corresponding to the Xd values). However, following this point, the permeability slump shows a virtually identical pattern in living and dead specimens.

Comparisons are made within species with the corresponding data for active animals in Table 9, and show that mean values fo Xd and R are lower than the corresponding values for active animals. This may relate to the higher desiccation rate prior to Xd which can be attributed to the lack of tun formation and hence retention of a greater surface area exposed to desiccation. The initial desiccation rates of the dead and active animals are similar since, at this point, dead and active animals are extended alike, exposing similar surface areas. However, unlike live animals forming tuns, dead animals show an increase in rates of water loss up to Xd and at this stage their desiccation rates are generally greater than those of live specimens (T_1/2 at Xd values).

The T_1/2 values measured immediately prior to Xd (T_1/2 at Xd) can be compared for tuns and extended (dead) animals to give an estimate of the reduction in desiccation rate per unit mass - and thus an independent estimate of the reduction in surface area - concomitant with tun formation. [In active animals, T_1/2 at Xd values are given as ‘T_1/2 initial’ since the logarithmic mass loss rates remain constant up to Xd (i.e. during tun formation).] Table 9 shows that the T_1/2 at Xd values for tuns are significantly lower than for extended animals in three of the species studied, demonstrating that tun formation reduces the desiccation rate. The mean ratios of these T_1/2 at Xd values between tuns and extended animals are shown in Table 10. The trials for living and dead animals used different specimens, so differences in size and disposition of the animals during drying constitute compounding sources of error. However, the ratios indicate relatively small Reductions in desiccation rate - approximately twofold in M. richtersi and H. oberhaeuseri, and less for the other species - and thus show reasonable agreement with the measured estimates of surface area reduction (1·75-fold to 2·22-fold for the eutardigrade species; 1·30-fold for E. testudo).

The four dehydration curves obtained for M. richtersi dried in low RH are shown in Fig. 11. Variables from these curves and their extrapolations are shown in Table 11. In comparison with animals dried in 80% RH, there is a much more rapid water loss attributable to the low RH, and the permeability slump occurs at a much later stage in the dehydration. This is shown by the lower Xd values (15–21% initial mass; P = 0·005) and the reduced water retention after 20 min (R = 8·4–12·3% initial weight; P= 0·005), although the R/Xd ratios show no significant difference from the corresponding values for animals desiccated in high RH. The T_1/2 extrapolation of the curve following Xd is also significantly more rapid (P = 0·05) with lower T_1/2 D1 values. Again, if the permeability change is function of cuticle hydration, this may be related to the greater desiccation rate.

The results for M. richtersi desiccated in low RH (20–25% ) following 15 min and 100 min of desiccation in high RH (80%) are shown in Table 12. In all trials, animals had passed through the permeability slump during the high RH desiccation and had thus attained low permeability. The results show that this low permeability following the slump is subsequently effective as a desiccation barrier to low humidities, since negligible mass loss is observed after transfer to low RH.

The results presented here illustrate dramatic differences in desiccation tolerance among species. Lower lethal humidities for initial dehydration range from 78 to 59%. A similar trend is shown by the times taken to acquire resistance to low RH: those species least tolerant of high initial desiccation rates also require more prolonged dehydration at high humidity before they are able to withstand exposure to much lower humidities. Thus, in ecophysiological terms, species tolerating rapid initial dehydration also tolerate a more rapid fall in humidity of the environment. Both suggest adaptation to xeric conditions.

Surface area reduction during tun formation is expected to be an important adaptation conferring desiccation tolerance. This study demonstrates a large reduction in surface area during tun formation in eutardigrade species. These show a significant positive correlation between surface area reduction and desiccation tolerance, thus supporting the hypothesis that tun formation serves an adaptive function in reducing the cuticular area available for transpiration. The heterotardigrade Echiniscus testudo shows a significantly smaller reduction in surface area, which can be partly attributed to the cuticular plates covering the dorsal surface. These plates show a greatly expanded and fenestrated epicuticle and are much thicker than any eutardigrade cuticle that has been measured (see Wright, 1988a, b).

The most dramatic features illustrated by the weighing experiments performed here are the extensive water loss incurred on dehydration and the abrupt permeability decline (the permeability slump) which arrests this water loss. Following 20 min of desiccation in 80% RH, all species had reached virtually constant mass, retaining 15–24% of their initial (wet) mass, of which 7–18% is water (dry masses are 6–8% of the wet mass). [Some of this will be ‘bound’ water, hydrogen-bonded to polar residues, but differential scanning calorimetry estimates of the bound water contents of M. richtersi and H. oberhaeuseri give values of only 3–4% of the initial (wet) mass (Wright, 1988c).] The amount of water retained correlates significantly with the desiccation tolerance, suggesting that water retention is a critical factor in determining the ability of a tardigrade to enter anhydrobiosis successfully.

Tardigrades dehydrated in an extended posture (dead and unable to form tuns) desiccate at an almost constant rate prior to Xd and thus show increasing T_1/2 values in semilogarithmic plots. Active animals, however, show linear semilogarithmic plots with a constant T_1/2 value prior to Xd. Since they do not reduce their surface area during drying, dead animals are expected to dry at a uniform rate. Tun formation, by contrast, reduces the surface area of active animals and the effect of this is to produce an exponentially decreasing rate of water loss (constant T_1/2)- Otherwise, however, the mass loss and humidity curves for the two groups are very similar. Dead, extended animals show the same abrupt decline in permeability after a short period and virtual cessation of water loss thereafter, although the permeability decline occurs at a significantly later stage in the dehydration for four of the five species studied. A similar but more dramatic picture is seen for animals desiccated in low RH (25–30% RH). Though initially active, these animals are unable to form tuns during such rapid dehydration and remain in an extended posture after drying; they do not revive on subsequent rehydration. Initial desiccation is very rapid and the permeability decline, although still present, occurs at a very late stage in the dehydration. Rapid desiccation thus delays the initiation of the permeability slump although, once effected, this arrests water loss just as rapidly as when the animals are desiccated in high humidities (shown by the unchanged R/Xd values). If, however, retention of a certain minimal volume of water is vital for survival, as the earlier results suggest, the later initiation of the permeability slump during more rapid drying will set an upper lethal limit to dehydration rates. This could explain why those species which effect the permeability slump at the earliest stage in their dehydration (H. oberhaeuseri and E. testudo) are also the most tolerant of rapid drying.

If the permeability slump provides an effective permeability barrier in low humidities, it should be the critical factor in conferring protection against such low humidities following high-humidity desiccation. This idea is substantiated by the results obtained from drying M. richtersi in low humidities following different durations of high-humidity desiccation. In all cases, animals which had effected the permeability slump during high-humidity desiccation showed negligible (undetectable) water loss on subsequent exposure to low humidity. Prolonged exposure to high humidities following the permeability slump does not appear to confer any extra protection. In a steadily falling humidity, characteristic of desiccating conditions in the field, the permeability slump will be increasingly delayed and so those species which effect it earlier will withstand the most rapidly falling humidities, just as they withstand the most rapid instantaneous drying rates.

The initial desiccation rates of the different species do not show any clear relationship to the other variables measured but would appear to relate quite closely to the size of the species, larger species desiccating less rapidly as a function of their initial (wet) mass. This is obviously to be expected given the increase in surface area: volume ratio with decreasing size. A larger animal should thus be able to survive more rapid transpiration since it will retain proportionally more water for a given temporal delay in initiation of the permeability slump. However, the behavioural importance of exploiting the smallest and most sheltered Compartments of a desiccating habitat confers a disadvantage on larger animals and may offset other advantages.

Somewhat similar results to these were obtained by Crowe (1972) who investigated rates of water loss from M. areolatus dried in different humidities. At high relative humidities (>80%), water loss was reported to be rapid during the first hour, then slowing considerably with equilibrium being attained after about 3 days. The long period of drying prior to the permeability decline can, in part, be attributed to the high humidities used for these experiments since the initial drying rates are considerably slower. However, the disparity between the results is rather surprising given that the species dehydrated in 80% RH in this study all effected their permeability decline within the first 10 min of drying. Crowe’s estimates of water retention are also considerably higher than those measured here, the water contents recorded after 40 h being 33–36% (equivalent to 41–44% of wet mass) for animals desiccated in humidities between 80 and 95%. Since Crowe’s drying experiments involved large numbers of animals, it seems likely that the crowding together of individuals caused substantial amounts of water to be trapped, thereby slowing subsequent dehydration. Given the negative correlation between water retention and desiccation rate demonstrated here, retarded desiccation would also explain Crowe’s higher estimates of water retention.

Crowe (1975) also studied water loss in lower humidities using M. areolatus. The results were similar to those presented here, with more rapid dehydration and an earlier decline in permeability. These features were also noted for asphyxiated animals dehydrated in high humidities. Crowe attributed the increased desiccation rate of asphyxiated animals to the lack of tun formation. He reported that the intersegmental areas are drawn into the body during tun formation and that these ‘cuticular folds’ are about one-twelfth the thickness of the rest of the cuticle (but see Wright, 1988b). Tun formation would thus appear to be an adaptation for reducing the mean permeability, as well as the surface area, of the desiccating animal. Crowe also reported that external dyes permeated the thin, intersegmental cuticle more rapidly. (However, given the ability of dehydrated animals to take up external water, whilst remaining impermeable to water loss, a permeability asymmetry clearly exists. Movements of dyes or tracers passing inwards across the cuticle should not, therefore, be used to assess outflow permeabilities.) Comparing transpiration rates in living and in dead animals prior to the permeability slump indicates that tun formation only reduces transpiration by approximately twofold. Since the reduced transpiration of tuns can therefore be attributed almost entirely to the surface area changes, the significance of removing the more permeable intersegmental cuticle, stressed by Crowe, appears to be small.

Tun formation clearly cannot explain the 100-fold decline in the rate of water loss which occurs during the permeability slump. In any case, the permeability slump is also observed in dead, asphyxiated and rapidly desiccated animals which do not form tuns. Crowe (1975) has postulated that phospholipids in the cuticle undergo a hydration-dependent phase change, resulting in an impermeable layer. Similar hypotheses have been put forward to explain the hydration-dependent permeability changes in other groups (Edney, 1951; Auzou, 1953; Bursell, 1955,Beament, 1961). Unfortunately, no-one has tested this idea using X-ray diffraction or electron diffraction and such techniques cannot be used to study tardigrade cuticles, less than 2μm in thickness. The results presented here do, however, provide clues to how such a phase change in cuticular lipids might operate. Experiments demonstrating the role of intracuticular lipids in conferring a dehydration-dependent permeability barrier during anhydrobiosis are discussed in a separate paper (Wright, 1988c).

The permeability of tardigrade cuticles in the later stages of anhydrobiosis cannot be estimated from this study since mass losses soon after the permeability slump become indistinguishable from balance drift and, for reasons already discussed, reliable mass loss data cannot be collected if large numbers of animals are dried together. However, animals clearly continue to transpire since Crowe & Cooper (1971) have demonstrated water contents below 3% in tardigrades subjected to prolonged desiccation. Comparative permeability data for insects are quoted in Gilby (1980), re-analysing data from Beament (1958). The nymph of Periplaneta americana, an insect of considerable desiccation tolerance, has an estimated permeability of 5·0×10⁻⁵ m s⁻¹. This is comparable to the calculated permeability of tardigrade cuticles at the end of the permeability slump (2·38×10⁻⁴ to 3·15×10⁻³ms⁻¹). Following this point, tardigrade permeabilities clearly drop further during the D2 stage of the permeability decline. Thus an animal essentially adapted for aquatic life (and of relatively high initial permeability) is able during anhydrobiosis to achieve a permeability similar to, and eventually probably lower than, that of a fully terrestrial insect.

I would like to express my thanks to Dr Mark Pagel for his statistical advice and to Dr Pat Willmer for her invaluable assistance and encouragement throughout this study.