Infrared differential thermal analysis (IDTA) was used to study ice propagation and whole plant freezing patterns in dehardened intact individuals of various alpine plant species, including a shrub (Rhododendron ferrugineum), a herbaceous plant (Senecio incanus), a cushion plant (Silene acaulis), and two graminoids (Poa alpina and Juncus trifidus). Freezing patterns differed markedly among species and reflected peculiarities of the shoot structure and the vascular system. In graminoids, each single leaf required a separate ice nucleation event, as the polystele prevents ice propagation between leaves via the stem. Additionally, enhanced supercooling resulted in a temperature range of whole plant freezing of up to 10°C, which corroborates the high summer frost resistance of graminoids. This could have ecological significance for frost survival. In contrast, in dicotyledonous species one nucleation event was usually sufficient for whole plant freezing. Controlled ice-seeding experiments on leaves with droplets of water and bacterial water suspension (Pseudomonas syringae) showed that ice propagation into the leaf tissue from the surface was inhibited as long as the leaves were undamaged. The rate of ice propagation in veins was significantly higher at lower temperatures and reached up to 24 cm s−1 in J. trifidus, which is much higher than reported in earlier findings. Ice propagation in graminoids was much faster, which may indicate that ice propagates within the protoxylem lacunae of large vessels.
Freezing stress is the first environmental filter a species has to pass through to become alpine (Körner, 1999). In European temperate alpine regions, sufficient frost hardening and snow cover (Sakai and Larcher, 1987) usually ensure frost survival in winter with predictable frost events. The situation is different throughout the growing period in summer when a low frost resistance coincides with unpredictable freezing events (Taschler and Neuner, 2004). After bud burst, during leaf expansion, alpine plant species are most frost susceptible and may even be damaged by frost at their natural growing sites (Taschler and Neuner, 2004). Small plants dominate the alpine habitat; the smallness causes a decoupling of plant body temperature from ambient (Körner, 1999). During the day, this usually produces a warmer microclimate, advantageous in a usually cold alpine environment. During the night, radiative heat loss from alpine plants is significant (Jordan and Smith, 1994) with leaves cooling by up to 8°C below ambient air temperature. Under calm conditions with low convection, small plants can be exposed to significantly lower temperatures than tall species due to long wave thermal energy dissipation. This increases the likelihood of frost damage and hence increases the need for specific frost survival mechanisms. Ice propagation throughout the plant can be considered ecologically significant for frost survival, particularly for the alpine species that may be exposed to freezing temperatures during their active growing period.
In earlier studies ice propagation in woody plants (Wisniewski et al., 1997; Workmaster et al., 1999; Carter et al., 2001; Ball et al., 2002; Sekozawa et al., 2004) and graminoids (Pearce and Fuller, 2001; Stier et al., 2003; Wisniewski et al., 2002a) was mainly studied by infrared video thermography (IRVT). IRVT investigations on ice propagation are rather rare for herbaceous plants (Wisniewski et al., 1997, 2002a), and are lacking totally for alpine plants. The recently developed technique of infrared differential thermal analysis (IDTA) of infrared images captured by a digital infrared camera has shown to be an efficient method for investigating freezing at the plant tissue level in leaves (Hacker and Neuner, 2007). In earlier investigations, ice propagation was mostly studied in detached plant parts, but rarely in whole plants. The smallness of alpine plants is hence advantageous as it allows an investigation of the freezing process in plants at the whole organism level. We investigated intact individuals from the predominating growth forms found in alpine habitats (woody shrub: Rhododendron ferrugineum; graminoids: Juncus trifidus and Poa alpina; herbaceous dicotyledonous species: Senecio incanus and Silene acaulis [cushion]) to determine ice propagation patterns at the whole organism level, employing the recently developed technique of IDTA (Hacker and Neuner, 2007). In the woody species investigated by Hacker and Neuner (2007), ice spread into the leaves after ice nucleation in the stem. Ice propagation was always associated with the leaf venation. This led to the suggestion that ice propagation initially occurs in the vascular tissue within vessels and tracheids. This implies a crucial role of the structure of the vascular system for ice propagation within the plant.
Freezing of plants can be initiated either by intrinsic or extrinsic ice nucleation (Wisniewski et al., 1997). Important extrinsic ice nucleators are ice nucleating active (INA) bacteria, which induce ice formation at the surface of plants at high temperatures (Maki et al., 1974). Leaves are a large microbial habitat and are colonized by a variety of species with bacteria by far the most numerous (Lindow and Brandl, 2003). The existence of INA bacteria is not sufficient for extrinsic ice nucleation, as nucleators are only active in aqueous solutions and thus surface moisture is a prerequisite for extrinsic ice nucleation (Wisniewski et al., 2002a). In nature, plants are usually exposed to radiative cooling and are thus colder than the surrounding air. This causes the formation of dew on the cold plant surfaces when air is cooled below the dew point. External ice does not necessarily induce freezing inside the plant, as the external ice has to come into contact with internal spaces of the plant. Ice may propagate through stomates, damaged cuticle or other lesions into the plant (Wisniewski et al., 2002a). If there is no direct contact of extrinsic ice with surface water in the apoplast, ice propagation into the plant is inhibited and can hence cause no harm to the plant. As the experiments were conducted under convective cooling conditions, the influence of extrinsic ice nucleators present under radiative cooling conditions was additionally investigated. Droplets of water and bacterial suspension of INA bacteria were applied on intact and manually damaged leaf surfaces to simulate extrinsic ice nucleation.
Material and Methods
Ice propagation was studied in various alpine plant species: an herb (Senecio incanus L.), a woody shrub (Rhododendron ferrugineum L.), two graminoids (Poa alpina L. and Juncus trifidus L.), and a cushion plant (Silene acaulis [L.] Jacq.). Plants of the investigated species were collected at natural sites in the central alps (around Mount Patscherkofel, 2240 m a.s.l., 47°12′32″N, 11°27′38″E) and potted in plastic containers (8 × 8 × 8.5 cm) filled with an alpine soil mixture. The potted plants were cultivated under natural environmental conditions in the Botanical Garden of the University of Innsbruck (600 m a.s.l., 47°16′4″N, 11°22′46″E). All experiments were conducted during summer, when the species were not frost hardened and exhibited their lowest frost resistance.
Potted plant samples were frozen in a convective freezing chamber with controlled temperature lowering at moderate rate of 4 K h−1. Two additional measurements were done under radiative conditions for R. ferrugineum and S. incanus. Investigations were restricted to the aboveground parts of the plants. As the soil is a tremendous reservoir of heat, soil temperature lagged behind the temperature of aboveground plant parts and did not influence the freezing process. Thermocouples connected to a datalogger (CR10X, Campbell Scientific, Logan, Utah, U.S.A.) were placed at the surface of the analyzed organs to measure absolute temperatures. Plant samples were measured mainly with a dry surface to identify action sites of intrinsic ice nucleators. For testing the influence of extrinsic ice nucleation on whole plant freezing, leaf surfaces of J. trifidus, P. alpina, R. ferrugineum, and S. incanus were wetted either with tap water or with a suspension of INA bacteria in water for controlled ice nucleation at −2.5°C. The water and INA suspension were applied as 50 µL droplets either on intact or damaged leaf surfaces. The latter was produced manually with a razor blade, generating a grid of deep cuts enabling the contact of the external ice with internal tissues. The experiments included four surface treatments, and the number of individuals tested are as follows (dry/water or INA suspension on an intact leaf surface/water on a damaged leaf surface/bacterial suspension on a damaged surface): J. trifidus (6/1/0/1), P. alpina (11/3/2/3), R. ferrugineum (14/2/2/1), S. incanus (8/2/1/1), and S. acaulis (3/0/0/0).
A strain of INA bacteria with high ice nucleating activity (Pseudomonas syringae, strain 5176, DSMZ Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, Braunschweig, Germany) was used for the bacterial suspension. The bacteria were cultivated monthly on nutrient agar (Nutrient Agar No 2, Fluka, Sigma-Aldrich Chemie GmbH, Buchs, Switzerland) in petri dishes at 22°C for 24 h and then stored at 5°C. The INA bacteria were applied as a bacterial suspension in distilled water containing several millions of bacterial cells (a count with the standard plate count method resulted in 144,000 bacterial cells µL−1 suspension).
Ice propagation in plants was measured with a digital infrared camera (FLIR Systems ThermaCAMTM S60, FLIR Systems AB, Danderyd, Sweden) that was equipped with a close-up lens (LW 64/150) to achieve a high spatial resolution of 200 µm. The system was connected via a FireWire interface with a control computer for data transfer. Control of the infrared camera and data analysis was conducted with the software package ThermaCAMTM Researcher (FLIR Systems AB, Danderyd, Sweden). The original infrared data were further analyzed by performing an infrared differential thermal analysis (IDTA) as described in Hacker and Neuner (2007). IDTA works in a fashion analogous to differential thermal analysis (DTA), where the sample temperature is compared with a reference temperature during the freezing treatment to detect the smallest temperature changes (Burke et al., 1976). IDTA is based on the subtraction of a reference image, captured just before the occurrence of freezing, from a sequence of images. The resulting differential images show only changes of temperature due to the release of heat during freezing; background temperature fluctuations are canceled out and smallest temperature changes can be visualized. In the resulting IDTA images, freezing is indicated by brightening of the tissue, while unfrozen areas remain black.
Measurement of the Rate of Ice Propagation
The rate of ice propagation in the veins of leaves was calculated by analysis of time series of IDTA images. The distance ice had propagated in the time span between IDTA images was measured by an image analyzing software (Optimas 6.5, Media Cybernetics, Bethesda, Maryland, U.S.A.). The rate of ice propagation was calculated as the ratio between distance and time.
Statistical Data Analysis
Differences between means, tested by analysis of variance followed by the Bonferroni multiple comparisons post-hoc test (P < 0.01) and correlation analysis were performed by the statistical software SPSS (SPSS, Chicago, Illinois, U.S.A.). Means and standard errors were calculated for every species based on the number of replicates mentioned in the above section.
The sequence of IDTA images in a shoot of the herbaceous plant S. incanus (Fig. 1A) shows initial ice nucleation at −4.3 °C in the mesophyll of a leaf (Fig. 1B). Ice immediately spreads into the vascular bundles of this leaf and therein down the stem (Figs. 1C–1F). From the stem ice is propagated to all other leaves of the plant within 5.2 s (Figs. 1G–1J). During vascular ice propagation a second ice nucleation event was detected in the mesophyll of a second leaf (Fig. 1G). The reticulate venation pattern is only visible close to the moving ice front, but soon afterwards, it is blurred due to mesophyll freezing which starts immediately after ice has spread throughout the vessels. The veins end freely in the areole (Esau, 1965) as indicated by illumination of the outermost tips of the lobes of a leaf blade when ice has reached it (Figs. 1C–1G). At the beginning of mesophyll freezing, a brindled freezing pattern can be observed, indicating a faster water efflux in some regions of the leaf blade compared to others (Fig. 1H). This pattern is diminished within 30 s due to the release of heat in the whole leaf lamina (Fig. 1L). Intrinsic ice nucleation (dry leaf surface) occurred either in the leaf lamina (6 out of 8 individuals) or in the leaf petiole (2 out of 8 individuals), but never in the stem. Ice nucleation occurred in seven individuals in the mesophyll tissue and only once in the vascular tissue. This resulted in a time lag between ice nucleation and further ice propagation in the vascular tissue towards the leaf margin and the stem. Vascular freezing started mainly within 24.4 s after ice nucleation, but lasted up to 294 s in one individual. Freezing of the whole plant, i.e. further spread of ice from the stem to the unfrozen leaves, was finished in five individuals within 1 min, but could be delayed up to 35 min. Both the time lag between ice nucleation and vascular ice propagation and the time of whole shoot freezing was highly correlated with the freezing temperature. At higher ice nucleation temperatures (especially above −4°C), both processes were significantly delayed.
Freezing of droplets of water or INA suspension on an intact leaf lamina had no effect on the freezing process of the plant, as the ice was restricted to the frozen droplet at the surface and could not propagate into the plant. When the leaf lamina was damaged, ice of a frozen droplet of water or bacterial suspension could easily propagate into the plant and cause whole shoot freezing. The freezing pattern consisting of vascular and mesophyll freezing was not affected by the extrinsic source of ice nucleation.
The investigation of the freezing process of one individual of S. incanus under radiative cooling conditions in situ, showed the same freezing patterns as observed under convective cooling conditions (Fig. 2). During the ongoing experiment, snowfall set in, but had no effect on ice nucleation and the freezing pattern. Intrinsic ice nucleation occurred in a leaf lamina at −3.5°C and caused the freezing of two leaves (Figs. 2D–2K). The other leaves of the shoot remained unfrozen down to a leaf temperature of −4.5°C at the end of the experiment, when the plant was fully snow-covered.
In the woody species R. ferrugineum, ice propagated throughout the entire shoot within 2 s, after ice nucleation had occurred in last year's stem at −7.3°C (Fig. 3). During ice propagation in the vascular tissue, the leaf venation pattern became visualized down to second-order side veins (Figs. 3B–3F). While in this year's leaves (the arrow in Fig. 3A marks the border between this and last year's leaves) the freezing pattern was clearly visible, in the last year's leaves IDTA images could not resolve the freezing process with the same resolution. Fully developed leaves are much thicker and the vascular bundles are embedded in a more compact tissue, hence heat, emitted during freezing of vascular bundles, is dispersed during the transmission to the leaf surface. In this year's leaves, the IDTA images did not differ, whether the lower or upper leaf side was exposed to the infrared camera. Immediately after vascular freezing the onset of mesophyll freezing blurred the venation pattern and showed a brindled freezing pattern at the beginning (Figs. 3G, 3H) and within one minute these differences were diminished (Fig. 3I).
The freezing pattern in R. ferrugineum was identical, when freezing was induced by extrinsic ice nucleation (Figs. 3J–3R). Ice nucleation occurred at −5.3°C in a water droplet, which was placed on a leaf with a damaged surface (Fig. 3J). Only 400 ms after freezing of the water droplet, vascular freezing of the leaf could be detected (Fig. 3L, indicated by an arrow). Subsequently ice spread into the stem and from there throughout the whole shoot (Figs. 3M–3R). A water droplet located on another leaf with damaged surface started to freeze as soon ice had spread into that leaf (Fig. 3Q). Frozen droplets of water or bacterial suspension on an intact leaf surface did not act as extrinsic nucleators and had no influence on the freezing process of the plant. This was also observed for the other investigated species with this surface treatment.
IDTA images determined during ice propagation in leaves of the grass P. alpina show the parallel venation pattern typical of monocotyledonous plants (Fig. 4). Intrinsic ice nucleation occurred typically in the vascular tissue either at the boundary between the leaf sheath and the leaf blade (Fig. 4A, ice nucleation temperature −5.9°C) or at the base of the leaf blade (Fig. 4E, ice nucleation temperature −10.4°C). From there, ice propagated throughout the vascular system towards the leaf tip and the sheath base, and vascular freezing was followed by mesophyll freezing resulting in a brindled freezing pattern (Fig. 4D). Ice propagation was sometimes delayed for several seconds at the boundary between the leaf blade and leaf sheath, independently of which direction the ice front was moving in. The boundary between the leaf blade and leaf sheath is not present in early stages of leaf development, but is established when the ligule develops (Esau, 1965).
The vascular system of grass leaves consists of longitudinal bundles which are laterally interconnected by small bundles in a ladder-like manner (Esau, 1965); in P. alpina these lateral connections consist of single tracheids (Figs. 4K, 4L). This specific vascular system results in a combination of longitudinal and lateral vascular ice propagation (Figs. 4E–4J). As the distance between the parallel bundles is too small (92 ± 10 µm) to be spatially resolved by the infrared camera, lateral ice propagation cannot be visualized in IDTA images. However, occurrence of lateral ice propagation can be detected in continued ice propagation in a neighboring bundle (Fig. 4I and detail in Fig. 4J). The measured rate of lateral ice propagation is 20 times slower than the longitudinal rate, because it consists of a longitudinal and a lateral component.
Ice propagation was always restricted to a single leaf. Ice did not propagate further to other leaves via the stem. All leaves attached to a stem froze independently of each other requiring an autonomous ice nucleation event in each leaf. The influence of extrinsic nucleators was tested by applying water droplets and water droplets containing INA bacteria to intact and artificially damaged (razor blade incisions) leaves of P. alpina. Ice nucleation in water droplets occurred at −6.2 ± 0.2°C and in water droplets containing INA bacteria at the much higher temperature of −2.8 ± 0.2°C. After extrinsic ice nucleation in droplets applied to an intact leaf surface, ice always remained restricted to the plant surface. Similarly, ice propagation from frozen plant tissue to a droplet placed on the intact leaf surface was not detected. When the leaf surface below the droplet was artificially damaged, ice could easily propagate either from the droplet into the leaf after extrinsic ice nucleation or from the leaf tissue into the droplet after intrinsic ice nucleation. The site of ice nucleation and the temperature of freezing did not affect the general freezing pattern (combination of longitudinal and lateral ice propagation).
The sequence of IDTA images during freezing of a tussock of J. trifidus is shown in Figure 5. All shoots of the tussock froze independent from each other, requiring an ice nucleation event in each shoot. In contrast to P. alpina the freezing pattern was affected by the degree of supercooling (Fig. 6). While intrinsic ice nucleation at −6.6°C caused freezing of a shoot in a single freezing event affecting all vascular bundles, extrinsic ice nucleation at −1.7°C initiated by INA bacteria resulted in multiple freezing events with ice propagating in both directions. These multiple freezing events either did not show a spatial and temporal relation (Figs. 6E–6G) or seemed to be related, as only 10 s after ice had spread from the base to the tip, ice propagated in the opposite direction (Figs. 6C, 6D).
IDTA images during freezing of a whole cushion of S. acaulis are shown in Figure 7. Ice nucleation occurred somewhere inside the cushion (but not within the plant tissues themselves) at a temperature of −0.8°C. The exact location of this extrinsic ice nucleation event could not be deduced from the IDTA images; it may have occurred in the soil, in the litter, or on the plant surface. After only 10 min, the first intrinsic freezing event within a single cushion shoot was detectable. Freezing of the whole cushion then took 52.5 min, but 30 min after intrinsic ice nucleation, 90% of the cushion was already frozen. Within the freezing process, 30 freezing events, including up to five shoots, could clearly be differentiated. As the stem system was not in focus of the infrared camera, it is difficult to tell whether ice propagated from one shoot to the next or several intrinsic ice nucleation events occurred. At the end of the freezing treatment all parts of the cushion were frozen.
Whole Plant Freezing Pattern
Ice nucleation temperatures varied within a broader temperature range in the investigated graminoids J. trifidus and P. alpina as compared to the investigated dicotyledonous species R. ferrugineum and S. incanus (Fig. 8). While ice nucleation did not occur below −6.3°C and −7.6°C in S. incanus and R. ferrugineum, respectively, supercooling down to −11.2°C and −13°C was observed for P. alpina and J. trifidus, respectively. The temperature range of whole plant freezing, i.e. the temperature span from the initial ice nucleation event until ice was found in the whole plant was significantly higher in the graminoids. While in R. ferrugineum and S. incanus one or two nucleation events resulted in freezing of the whole plant, the graminoids required a nucleation event in every leaf, as the single leaves freeze autonomously. The requirement of many nucleation events and the capability for high supercooling enables the graminoids to avoid freezing of the whole plant, when the exposition temperature was above the minimum ice nucleation temperature. Hence, leaves without ice nucleation could survive the freezing treatment (shown for J. trifidus in Fig. 5).
Rate of Ice Propagation
The rate of ice propagation in the veins of leaves increased significantly with decreasing temperature in all investigated species (Fig. 9). When freezing occurred at temperatures of −2°C, the rate of ice propagation was low in all species with values between 2 and 3 cm s–1. With increasing supercooling, i.e. lower freezing temperatures, ice propagated much faster, reaching values of up to 24 cm s–1 in J. trifidus. The linear correlation between the degree of supercooling and the rate of ice propagation was highly significant. The incline of the regression lines was identical for R. ferrugineum and S. incanus, but was significantly higher in P. alpina and even more in J. trifidus. This resulted in higher rates of ice propagation in the graminoids at similar freezing temperatures.
Ice Propagation and Freezing Patterns
The determination of ice propagation and freezing patterns at the whole organism level using IDTA revealed substantial differences among the investigated species. In plants with plane leaf blades such as P. alpina, R. ferrugineum, and S. incanus the leaf venation pattern could clearly be visualized during vascular freezing, which was also shown for woody species in Hacker and Neuner (2007). In leaves of J. trifidus and S. incanus, this was not possible due to the small leaf size (1 mm). Ice propagation patterns in the two investigated graminoids J. trifidus and P. alpina reflected the influence of the vascular structure on the freezing process, at the leaf level and the whole organism level as well. In P. alpina single tracheids laterally interconnect the longitudinal bundles in a ladder-like manner; hence, vascular freezing consists of a longitudinal and a lateral component. Lateral ice propagation cannot be visualized by IDTA, as the distance between longitudinal veins is only 92 µm and thus below the spatial resolution of the infrared camera; additionally, the released heat resulting from freezing of single tracheids is very small. In contrast, the rush J. trifidus lacks these vascular interconnections and the longitudinal bundles are isolated from each other. At high nucleation temperatures induced by application of a bacterial INA suspension, freezing of bundles seemed to be facilitated by ice propagation within non-vascular tissue or an upturn of ice propagation at the leaf tip.
Leaves and shoots of graminoid tussocks froze independent from each other as ice could not propagate from one leaf to another via the stem. This has already been reported for other graminoid species (Pearce and Fuller, 2001; Stier et al., 2003). The IDTA records reveal ice propagating throughout the vascular bundles of the leaf blade and leaf sheath into the stem. The polystele of monocotyledonous stems prevents ice propagation via the vascular system into other leaves. Leaf vascular bundles do not end blindly in the stem, but are usually linked with strands of older leaves via nodal plexus branches (Esau, 1965; Hitch and Sharman, 1971; Shane et al., 2000). Although this linkage allows the conduit of water, ice propagation seems to be blocked.
In S. acaulis, initial freezing was always extrinsic and occurred inside the cushion (litter, soil). Only after a distinct delay, freezing of single shoots was detected; the freezing pattern of the whole cushion consisted of many separate freezing events of single shoots. Whether these freezing events were induced by independent intrinsic ice nucleation or by delayed ice propagation between shoots could not be determined by IDTA due to the highly complex aboveground plant structure of the cushion. Leaf and branch junctions are known as potential water flow constrictions and segmentation points in the hydraulic architecture (Schulte and Brooks, 2003). The sectoriality of the vascular system describes the degree of hydraulic coupling between plant parts, which may range from full coupling to limited interconnectedness (Brooks et al., 2003; Schulte and Brooks, 2003; Orians et al, 2005; Ellmore et al, 2006; Zanne et al., 2006). Hydraulic sectoriality may explain the freezing pattern in S. acaulis with its highly branched shoot and S. incanus with leaves emerging from a compressed stem.
The number of ice nucleation events necessary for whole plant freezing and the temperature range wherein whole plant freezing occurred increased from woody and herbaceous plants to cushions and increased significantly further to the graminoids. In herbaceous and woody plants, one ice nucleation event caused immediate freezing of the whole plant, as there were no ice barriers. No specific effect on frost survival may be derived from this ice propagation pattern and at least for some alpine plants it is known that they depend more on repair and replacement of vegetative tissue lost after frost damage than on high summer frost resistance (Körner, 1999; Taschler and Neuner, 2004). The rate of ice propagation and mesophyll dehydration is distinctly slowed down at higher ice nucleation temperatures (Hacker and Neuner, 2007) and might even result in less frost damage as extended supercooling may cause more severe frost damage (Sakai and Larcher, 1987).
For cushion plants during summer night frosts, thermal buffering by freezing of water externally sucked up by litter and debris within the compact cushion could be an important frost survival mechanism (Sakai and Larcher, 1987). The heat of crystallization released during freezing of this external water inside compact cushion plants such as S. acaulis can keep the plant temperature at moderate freezing temperatures (> −2°C) for several hours until all of the adhering water is frozen. Hence, the living plant parts that freeze in situ at mean temperatures of −3.5°C (Taschler and Neuner, 2004) may be protected from lethal freezing temperatures throughout a whole summer night frost (Sakai and Larcher, 1987). In graminoids, each leaf needs an autonomous ice nucleation event. This freezing pattern of the graminoids combined with a high supercooling ability may explain why graminoids had the highest summer frost hardiness among alpine plant species (Taschler and Neuner, 2004). Air temperature minima indicate that the investigated plants may suffer frost damage at their natural growing sites (see Fig. 4 in Taschler and Neuner, 2004). Hence, the freezing pattern observed in graminoids may be advantageous for survival at alpine sites.
Intrinsic Versus Extrinsic Nucleation
In natural habitats, the efficiency of supercooling as a frost survival mechanism will also depend on the question of formation of surface ice. Bacterial ice nucleators are commonly present on plant surfaces (Hirano and Upper, 2000). They are only active in aqueous solutions and thus surface moisture is a prerequisite for extrinsic ice nucleation (Wisniewski et al., 2002b). In a natural setting during radiative cooling plants can be exposed to dew formation on the plant surface. This dew provides the surface moisture required for the activation of INA bacteria that are likely to cause external ice nucleation at higher temperatures. Hence supercooling as observed in our laboratory experiments will only be possible at natural growing sites if ice growth from extrinsic nucleation sites into the plant is impaired. Frozen droplets of water or bacterial suspension applied on the leaf surface of P. alpina, S. incanus, and R. ferrugineum could not induce intrinsic ice nucleation as long as the leaf epidermis was undamaged. In contrast, frozen droplets applied on damaged leaf surfaces induced freezing of the plant in all investigated species. The freezing patterns were independent of the ice nucleation source (either water or INA bacteria). When INA bacteria at −2.5°C induced extrinsic nucleation, the freezing pattern was not altered, but the rates of ice propagation were lower. This corresponds to earlier observations that external ice does not necessarily induce intrinsic freezing, as the external ice first has to propagate into the plant (Wisniewski et al., 2002a). Ice propagation can be through stomata in some species (Phaseolus vulgaris, Prunus domestica, Malus domestica: Wisniewski et al., 1997; Wisniewski and Fuller, 1999), but only via a damaged cuticle or other lesions in Rhododendron sp. (Wisniewski et al., 1997), Lolium perenne, and Poa supina (Stier et al., 2003). Leaf wettability is also an important factor in protecting the leaf from extrinsic ice nucleation. Applying a hydrophobic particle film can prevent extrinsic ice nucleation, which prevents ice growth from the surface into the plant and hence frost damage (Wisniewski et al., 2002b; Fuller et al., 2003). Preliminary results indicate that leaf wettability is reduced in most alpine plants (Aryal and Neuner, unpublished data).
Rate of Ice Propagation
Our results corroborate earlier findings (Hacker and Neuner, 2007; Kitaura, 1967; Pearce and Fuller, 2001) that a major determinant of the rate of ice propagation is the ice nucleation temperature, i.e. supercooling will markedly speed up ice propagation in plants. The rate of ice propagation increased significantly with decreasing freezing temperature in the investigated plant species (J. trifidus, P. alpina, R. ferrugineum, and S. incanus). Hacker and Neuner (2007) found the same relationship for woody species (B. sempervirens, C. camphora, F. sylvatica, Q. robur, and T. baccata). Ice growth in an aqueous solution depends mainly on thermal phenomena occurring at the solid-liquid interface (Ayel et al., 2006; Liu et al., 2004; Shibkov et al., 2004). Three factors limit the ice growth rate: (1) the diffusion of heat and solute away from the solid-liquid interface, (2) the tension between the ice and the solution, and (3) the kinetics of attachment of water molecules to the ice surface (Liu et al., 2004). The latter can be neglected, as the cooling rates in biological systems are not high enough.
The investigated graminoids J. trifidus and P. alpina showed significantly higher rates of ice propagation compared with other herbaceous and woody species at similar temperatures. A peculiarity of monocotyledonous leaves is the existence of protoxylem besides metaxylem vessels in some of the larger longitudinal bundles (Dannenhoffer and Evert, 1994; Martre et al., 2000). Protoxylem vessels ensure vascular continuity in the growth zone of grass leaves. During further growth, the protoxylem vessels mainly collapse and are replaced by a lacunae. This protoxylem lacunae is water filled and functions as a water-conducting conduit continuous throughout the leaf (Dong et al., 1997; Martre et al., 2000; Canny, 2001).
In intact vessels, moving water has to pass the pit membranes at the end walls. The small pore sizes (estimated average pore diameter in dicotyledonous trees: 5–20 nm (Choat et al., 2003) generate a significant hydraulic resistance (Choat et al., 2006) which may account for up to 87% total vascular resistance (Schulte and Gibson, 1988; Sperry et al., 2005; Choat et al., 2006). Pit membranes may also represent an efficient barrier for ice propagation that is known to be suppressed in capillaries below 5 nm (Muldrew et al., 2000). In the water filled protoxylem lacunae, fast ice propagation is possibly enabled by the lack of pit membranes. Hence, ice propagation in the two investigated graminoids may occur in the protoxylem lacunae, indicated by the faster ice propagation.
IDTA has proven to be an efficient method to investigate freezing on a whole plant level and revealed distinct freezing patterns in the investigated species. The various freezing patterns and ice propagation velocities reflect differences in the hydraulic systems of the plants. Additional investigations of the sectoriality and synchronization with the freezing pattern could further elucidate the pathway of ice propagation. Whole plant freezing patterns can explain the results of summer frost resistance of alpine plants. Graminoids may gain an advantage for frost survival due to enhanced supercooling ability combined with independent freezing of single leaves. Ice at the plant surface acted as an extrinsic ice nucleator only, when the surface was damaged, enabling the contact between ice and inner tissues; an intact epidermal layer prevented freezing of the plant. The freezing patterns themselves were not affected by the source of ice nucleation. Convective cooling has been shown to be a viable method for investigation of freezing patterns, as the freezing patterns were similar under radiative cooling conditions. The main difference between these treatments is dew formation on the plant surface under radiative conditions that will possibly enable extrinsic ice formation. Under convective conditions due to a dry surface plant freezing is always initiated by intrinsic ice nucleation. When the epidermal layer is intact, extrinsic ice will have no influence, but in case of injuries, it will be the starting point of freezing. Although the freezing pattern did not change, freezing at higher temperatures will cause freezing and withdrawal of water from cells at more moderate rates, which can be advantageous for plant survival.
This work was funded by the Austrian Science Fund FWF (P17188-B03). Thanks to Rosa Margesin (University of Innsbruck, Institute of Microbiology) for assistance with the cultivation of INA bacteria. We wish also to thank the reviewers and editors, and especially Katy Sommerville for valuable comments on the manuscript.