Rachides of Juglans regia L. (Juglandaceae) and one-year-old twigs of Evonymus latifolia (L.) Mill. (Celastraceae) were cooled in air to −25 °C, with an ultrasound detector attached to the xylem where peripheral tissues had been peeled off. Ultrasound acoustic emissions started between −4·5 and −14·3 °C, as measured with a thermocouple inserted into the xylem. The number of emissions was significantly lower from saturated plant parts than from those frozen at field water potentials. Bench-drying of saturated samples produced significantly less signals than the freezing protocols. These findings are in accordance with the hypothesis that freezing of xylem under tension induces cavitation events. They corroborate earlier work which tried to provide a logical explanation for the seemingly paradoxical cryo-scanning electron microscope observations of changing vessel contents during a daycourse in the field.
Recent objections raised by Martin Canny's group against the cohesion–tension theory of water ascent in plants were based, among other arguments, on observations of vessel contents in a cryo-scanning electron microscope (CSEM). Plant parts such as roots and rachides were frozen in situ by immersion in a bath of liquid nitrogen (LN2) or by squeezing them between the copper jaws of pliers pre-cooled in LN2. Sampling during a daycourse showed that the vessels were mostly filled with frozen water in the morning and in the evening, whereas they were largely empty around noon (Canny 1997a, b; McCully, Huang & Ling 1998; McCully 1999; Canny, McCully & Huang 2001a). The authors assume that the CSEM demonstrates the true filling status of vessels in situ, which thus appear to go through daily cycles of cavitation and refilling, with a minimum of water content in the middle of the day.
This notion is paradoxical from the point of view of the widely accepted cohesion–tension theory (Boehm 1893; Askenasy 1894; Dixon & Joly 1895), and it leaves a number of other experimental observations unexplained. For instance, sap flow measurements in tree stems show a maximum at noon-time of sunny days, when transpirational water loss from the leaves peaks (Čermák, Matyssek & Kučera 1993; Čermák et al. 1995). At the same time, the diameter of the sapwood in trees reaches a minimum; this contraction is interpreted as an effect of tensions in the water column induced by frictional potential losses (Irvine & Grace 1997; Offenthaler, Hietz & Richter 2001). It is very difficult to accept empty xylem conduits as the pathways for rapid water transport or to explain contractions of a sapwood cylinder when most of its conducting elements contain a vacuum or air.
A logical resolution of this dilemma was provided by Cochard et al. (2000) and Cochard, Améglio & Cruiziat (2001). They froze in situ walnut rachides in LN2 and corroborated the findings of variable filling status during a daycourse. Rachides cut at various times from the tree under water and thus rapidly saturated before freezing showed the vessels uniformly filled with frozen water. The authors then proceeded to demonstrate that there is no loss of conductivity of the rachide xylem at a tension of −0·7 MPa imposed by centrifugal force, which was identical to the minimum water potential observed during a daycourse. After freezing these rachides in LN2, the conductivity broke down completely. The authors concluded that the voids in the vessel contents observed in the CSEM around noon were an artefact produced by the freezing of intact water columns in a state of tension. Additional arguments for (Richter 2001) and against this view (Canny et al. 2001a, Canny, McCully & Huang 2001b) have already been reported.
During August and September 2001 material was collected from single adult specimens of walnut, Juglans regia L. (Juglandaceae), and broad-leaved spindle tree, Evonymus latifolia (L.) Mill. (Celastraceae), growing in the Botanical Garden of the University of Agricultural Sciences Vienna. Pinnate leaves of Juglans with five large and two small leaflets were detached around noon from basal sun-exposed branches, put in plastic bags and brought to the laboratory. Leaves with one additional large leaflet pair and thicker rachides were not used. (The anatomical structure of the rachide is shown in excellent micrographs by Canny et al. 2001a). The leaflets were severed and the rachides either immediately used for freezing experiments or recut under water and saturated by standing them overnight with their cut ends in distilled water inside a closed container. Leaves for the bench-drying experiments were saturated without severing the leaflets. To determine the xylem water potential of the rachides at the time of collection the terminal leaflets of leaves similarly exposed on the same twig were wrapped in situ in aluminium foil and sealed in plastic bags at least 4 h prior to sampling. Xylem water potential (Ψx) was measured with a pressure chamber (Model 3000; Soil Moisture Equipment Corporation, Santa Barbara, CA, USA). One-year-old twigs of Evonymus (about 30 cm long) were treated similarly. After collection the twigs were either immediately used for freezing experiments, with the leaves being removed, or saturated in distilled water overnight. Full saturation was checked by measuring the water potential of the uppermost pair of leaves with the pressure chamber. Then all the remaining leaves were severed and freezing started.
Xylem water potential of the stems at the time of collection was determined on the uppermost pair of leaves bagged in situ in aluminium foil and plastic for at least 4 h before measurement.
Measurements of acoustic emissions and temperature
An I15I ultrasound acoustic sensor (d = 18 mm) was connected to a 4615 Drought Stress Monitor (both by Physical Acoustics Corporation, Boston, MA, USA) by means of a cable, 10 m in length. The same sensor was used for all measurements to reduce systematic errors. The total gain of the instrument was set at 72 dB (the monitor amplifier set at 52 dB, the head-stage amplifier fixed at 20 dB), a factor often used in UAE studies on stem xylem under drought stress. The time course of UA emissions was registered with a notebook computer and later analysed by means of the Excel® software (Microsoft Corporation, Redmond, WA, USA).
The UAE sensor has an operating range from −45 to +85 °C. The deep-freeze compartment of a large laboratory refrigerator (Liebherr Lienz GmbH, Lienz Austria) was set to −25 °C. The sensor, either in air or in contact with plant xylem, plus 7 m of the coiled cable were then exposed to this temperature. About 3 m of the cable length was led through the rubber gasket of the door to the monitor on the laboratory bench. Care was taken that the cable adjacent to the sensor (20–30 cm) was kept straight, because movements in this region could produce spurious ultrasound signals.
Air temperature in the deep-freeze compartment and temperatures in the xylem of the cooled plant material were measured with a copper–constantan thermocouple mounted near the UAE sensor. Data were logged every 2 s with a CR10X measurement and control module (Campbell Scientific, Inc, Logan, UT, USA).
The peripheral tissues of freshly harvested or saturated rachides or twigs were peeled off over a length of about 25 mm and the UAE sensor at room temperature was directly clamped to the exposed xylem. The position was between the first and the second pair of leaflets in Juglans and in the centre of an internode about 15–20 cm from the tip in Evonymus. The thermocouple was inserted into the xylem about 10 cm from the UAE sensor and the whole assembly was then put into the deep-freeze compartment, where the cooling in cold air proceeded rather rapidly.
The significance of the differences in numerical results from different treatments was tested with the Mann–Whitney U-test, the Kruskal–Wallis H-test or with one-way analysis of variance (anova; SPSS 9·0 for Windows®, SPSS Inc., Chicago, IL, USA). Differences were accepted as significant if P ≤ 0·05. Values are given as means ± standard error. Vertical box plots were used to show the distribution of data. The lower boundary of the box indicates the 25th percentile and the upper boundary the 75th percentile. A line across the box marks the median. The whiskers indicate the highest and lowest values.
Some initial experiments were aimed at assessing possible sources of spurious ultrasound signals. They gave the following results:
1The number of signals from a sensor freely exposed to the air in the deep-freeze compartment was never more than three during a 3 h period, which is in the range of the spurious signals registered by a free sensor on the bench at room temperature. Obviously, cooling of the piezo crystal did not produce additional signals.
2When pieces of wet sponge cloth and wet filter paper were in contact with the sensor, it did not record more than six signals during a 20 h freezing period.
3The peripheral living tissues (epidermis, three to four layers of parenchyma without chloroplasts, eight layers of chlorenchyma, and collenchymatous ridges under the epidermis) were peeled off over the length of an internode in Evonymus. The dead sclerenchyma close to the phloem was not included. There were from 12 to 98 signals in fully saturated material and from one to 10 signals in freshly harvested samples. This difference was highly significant. Similar samples from Juglans rachides between the first and the second leaflet pair containing abundant sclerenchyma gave up to a few hundred signals and showed no difference between the fully saturated and the pre-stressed states.
4Thawing of frozen rachides and twigs did not produce any signals.
The water potentials of bagged leaflets or leaves at the time of collection were −0·72 ± 0·11 MPa (n = 14) in Juglans and −0·87 ± 0·07 MPa (n = 20) in Evonymus.
Xylem emissions followed a time course that was very similar to the pattern of stems or leaf midribs under drought. Our material never produced ultrasound during the short time necessary for cooling to temperatures below −4 °C. After the lag period the first signals appeared at distances of several seconds. They increased in number towards a maximum where the monitor sometimes registered more than a dozen events per 2 s interval, interspersed with short periods of reduced activity. After some time the production of signals petered out and stopped completely after about 3 h in Juglans and 20 h in Evonymus, respectively.
Figure 1 shows temperature courses and signal production in saturated and non-saturated rachides of Juglans. Temperatures declined steeply to values lower than −6 °C, where pronounced exotherms started. The peaks reached −3·3 °C in both samples. On average, temperatures decreased to −6·73 °C, where a fast increase was observed indicating the onset of freezing. The mean value for the peaks was −3·65 °C. Differences between the treatments were not significant. After the peak the curves declined slowly to the end value of −25 °C. Exotherms in Evonymus (not shown) were similar and the peak values statistically not different from Juglans.
Signals started in the non-saturated rachide after 3·2 min and at a temperature of −5·3 °C, in the saturated rachide after 5·9 min and at −8·0 °C, and first occurred only at intervals. After about 15 min and at a temperature around −15 °C a steep increase in the production of signals was observed (Fig. 1).
In all the combined data sets analysed the first events during concomitant UAE measurements were registered after 2·5–6·0 min at temperatures between −4·5 and −14·3 °C. In Juglans there was a significant difference in both the cooling time (P ≤ 0·001) and the temperature (P ≤ 0·05) for the first signals between saturated and non-saturated rachides. The latter started to emit UAE earlier and at higher temperatures than the saturated treatment. In Evonymus the values were not significantly different for saturated and non-saturated stems. In both species the rate of ultrasound emission increased rapidly over the next 30–40 min and slowed down thereafter.
The cumulative number of UAE obtained on the rachides of Juglans is shown in Fig. 2. There were conspicuous and highly significant differences (P ≤ 0·001) between the three treatments. The highest number (4070·7 ± 272·1; n = 13) was recorded from xylem frozen in a state of tension. Rachides previously saturated emitted a smaller number (1429·1 ± 130·4; n = 13), whereas bench-drying material had the lowest signal production (418·5 ± 46·7; n = 10).
Figure 3 gives the total number of UAE for Evonymus. Again, saturated twigs produced less signals (298·7 ± 86·8, n = 9) than those freshly harvested in the field while in a state of tension (1992·6 ± 648·3; n = 11), the difference being significant (P ≤ 0·01).
Freezing xylem, like drought-stressed xylem, emits a large number of ultrasound signals which can be recorded with appropriate equipment; water freezing in large-porous sponge cloth and in small-porous filter paper does not. Living Evonymus tissues emit a small number of signals only, and more so when they are fully re-saturated than when freshly harvested. We think that this sound production is due to friction at the sensor surface during temperature-induced contractions or expansions, and that non-turgid tissues produce fewer signals because they are ‘softer’. The removal of the peripheral tissues and the direct contact between sensor and exposed xylem makes us certain that freezing xylem is the main source of ultrasound in the plant organs investigated.
The number of signals from rachides of Juglans regia as well as from the stems of Evonymus latifolia depended on the state of tension in the water columns, as inferred from the water potential of the xylem. Scholander et al. (1965) assume that on severing a leaf the water will recede in a cut xylem element only to the next cross wall, so that the xylem fluid remains under tension. We think that this scenario is correct at the moderate tensions found in freshly harvested rachides and stems. Only with very negative xylem water potentials would we expect cavitations seeded by air entering through the pit membrane pores of the cross walls and gradually emptying the xylem.
The stems and rachides that were saturated overnight emitted a significantly (P ≤ 0·01 and P ≤ 0·001, respectively) lower total than those frozen at moderate tensions. That the rachides needed less time to reach the end of sound production than the twigs may be a consequence of the greater diameter of the xylem cylinder and the increased number of conducting elements in the twigs.
Ultrasound production during freezing and its dependance on the state of tension in the xylem somehow contradict the accepted order of events for freeze–thaw embolization. This current view has been well summarized by Lemoine, Granier & Cochard (1999): the gases in the xylem sap freeze out as bubbles, as they are insoluble in ice. If these bubbles come under tension during thawing, they grow and cause cavitation. Sperry & Sullivan (1992) calculated the tensions necessary for the growth of a bubble containing all the air saturating the water volume in a conduit. They noticed that the critical tensions calculated were several orders of magnitude lower than those necessary for actually breaking the column during a freeze–thaw cycle and gave a number of possible reasons for this observation. Among them were re-dissolution of air during the thawing process and formation of many small bubbles instead of a single large one. However, the emission of ultrasound during freezing seems to indicate that cavitation of the water column occurs well before all the water has frozen in a conduit, and this would mean that only a part of the air has been released and formed a bubble (or, more likely, several small ones) when a xylem element starts to cavitate. The significant rise in the threshold temperature of the xylem for ultrasound emission, which was observed in rachides under tension, is another indication that the freezing process is not completed when signals start. This would explain why the tensions required for a measurable loss of hydraulic conductivity are far larger than predicted by a simplified theory.
Our results agree with findings in a combined UAE-DTA (differential thermal analysis) study by Raschi et al. (1989), which in later work has not received due consideration. These authors observed the onset of acoustic signals during freezing at temperatures somewhat lower than the first ice formation and recorded very few events during thawing. We never observed signals during thawing periods of several hours. It seems therefore likely that, at least under conditions of fairly rapid cooling, cavitation occurs on freezing and not on thawing. Lo Gullo & Salleo (1993) followed the signal production from the xylem of Quercus ilex L. closely and also found copious UAE during freezing. It should, however, be emphasized that cavitation (the breaking of the water column on freezing) will be followed by embolization (the entry of air into the cavitated xylem element) only after thawing.
A puzzling result of our experiments is that the number of signals recorded during freezing, even without tension in the xylem sap, is far higher than the number recorded on parallel leaves or twigs which are drought-stressed on the laboratory bench. This cannot be due to a higher sensitivity of the I15I sensor, as this device is slightly less responsive at low temperatures (Joachim Sell, Euro Physical Acoustics, pers. comm.).
The increased number of signals recorded during freezing may be tentatively explained in different ways. First, plant tissues can be expected to shrink and expand on cooling and freezing. These movements could produce additional signals, either throughout the organ or at the point of contact between the xylem and the sensor, and thus contribute to the high base level from fully saturated organs; they can, however, not explain the very high number of additional events recorded from xylem under tension.
Second, freezing may produce ultrasound in plant structures not emitting signals on drying. For example, this could be thick-walled fibres (Lo Gullo & Salleo 1993). This assumption also will not explain the surplus UAE, because in other experiments (Kikuta, unpublished) it was observed that fibres emit signals on drying, too. This is also true for the peripheral tissues of Juglans rachides, which produce more signals than Evonymus, but about the same number on freezing and on drying. Most of these UAE in Juglans probably come from the sclerenchyma. Alternatively, living cells might be an ultrasound source (Weiser & Wallner 1988). These authors found emission peaks which they ascribed to the cavitation of ray parenchyma cells, however, only at temperatures lower than the minimum values in our experiments, where the living tissue layers of Evonymus produced only very few signals down to −25 °C and even less signals when not turgescent.
Third, ultrasound might be propagated over longer distances at subzero than at room temperatures. This is the explanation for our observations that we tentatively favour in the absence of further investigations. Cooling and freezing will gradually stiffen the walls and solidify the contents of dead and living xylem elements. It is clear that this should influence the attenuation of sound waves; single events might even be recorded more than once due to reflections on the ice front. However, the literature seems not to provide much information. Lo Gullo & Salleo (1993) recorded only a few signals from the freezing xylem of Quercus ilex at −1·5 °C despite a considerable loss of hydraulic conductivity, whereas at temperatures of −11 °C the UAE activity increased dramatically. This could also be an effect of the increasing amount of ice in the sample.
Altogether we are dealing here with yet another example for the fundamentally unclear relationship between the number of signals registered by the sensor and the number of cavitation events. This ambiguity becomes strikingly apparent in the rather tentative recipes for the selection of a ‘correct’ gain, i.e. an amplification of the original signal resulting in a 1 : 1 relationship between cavitation events and ultrasound acoustic emissions in drought-stress experiments. The gains used in most papers run from 72 to 78 dB (e.g. Weiser & Wallner 1988; Cochard 1992; Jackson et al. 1995; Salleo et al. 2001), but they are seldom adjusted to the number of cavitating conduits; in most cases, they are just set as high as possible without increasing background noise.
Our results are of course incompatible with the ideas of Canny's group about the origin of voids in vessel contents observed in the CSEM. On the contrary, we think that they support the explanation by Cochard et al. (2000). Cooling in air produced a low freezing rate compared with an immersion in LN2, but both were in the same physical range, that is, lower than the rate necessary for vitrification. The effects of the two protocols should therefore be comparable. The conspicuous daycourses observed in the CSEM are thus most likely the result of changes in the tension state of the xylem fluid, which is an important factor for cavitation during freezing, and not due to rapid cycles of cavitation and refilling in the unfrozen xylem. Cooling rates achieved by all the techniques presently available are by far too low to vitrify xylem water, which however, would be a prerequisite for observing the true filling status of conduits in the CSEM (Richter 2001). Microcrystals formed during cooling with liquid nitrogen are invisible in the CSEM, but nevertheless freeze out gas and produce solid ice surfaces with a different crystal lattice. These new phases obviously trigger cavitation events.
We thank Joachim Sell, Euro Physical Acoustics, Wolfegg, Germany, for valuable information, and Ivo Offenthaler for expert help with the temperature measurements and critical discussions. Martin Canny acted as a referee, and his comments as well as those of an anonymous second referee helped us to present some of our ideas more clearly.
Received 29 January 2002; received in revised form 15 July 2002; accepted for publication 7 August 2002