Plant acclimation to freezing temperatures is very complex. Many temperate plants increase in freezing tolerance upon exposure to a period of low but non-freezing temperatures, an adaptive process known as cold acclimation. This acclimation phenomenon has encouraged investigations of physiological, biochemical, and molecular changes that are associated with the development of freezing tolerance. Although many biochemical and gene-expression changes occur during cold acclimation, few have been unequivocally demonstrated to contribute to the development of freezing tolerance. However, in the last few years, exciting new progress has been made through the use of mutational analysis and molecular genetic approaches. We now recognize that several interacting signal pathways are activated to bring about cold acclimation and ensure the winter survival of plants. The challenge for the future is to understand these pathways at a mechanistic level. Facile map-based cloning in Arabidopsis and techniques (such as DNA micro-arrays) for transcript profiling will provide the tools needed for this task.
Freezing temperature represents a major environmental constraint limiting growth, development, and distribution of plants. Most tropical and subtropical plant species lack the ability to adapt to freezing temperature and are typically injured by temperatures below 10 °C. In contrast, temperate plants have evolved mechanisms by which they can increase their ability to withstand the subsequent freezing temperatures in response to a period of low but non-freezing temperatures. This process is called cold acclimation (CA) ( Levitt 1980). In nature, cold acclimation is initiated by the decreasing temperatures in late autumn or early winter. It can be duplicated in the laboratory simply by exposing plants to low temperatures (2–6 °C) under appropriate light condition. Depending on the plant species, it may take a few days to several weeks to reach maximum levels of freezing tolerance and the tolerance produced ranges from −10 °C to below −60 °C ( Gilmour, Hajela & Thomashow 1988; Webb, Uemura & Steponkus 1994). Cold acclimation is very complex, involving many biochemical and physiological changes. It has been estimated that the expression of hundreds of genes can be altered ( Guy 1990). These characteristics, and indeed the existence of an inducible cold acclimation process, indicate that the biological costs of constitutive freezing tolerance must be high enough to be a negative factor in evolutionary selection. Despite intensive research for nearly a century, fundamental questions of this field still remain to be answered. How do plants sense the decreasing temperature? What are the signalling cascades that transduce the low temperature cue to the cell nucleus and bring about changes in gene expression? What is the relative importance, to freezing tolerance, of each of the changes in metabolism and gene expression that occur during cold acclimation? More complete answer to these questions is of basic scientific interest and has enormous practical applications as well.
In the last few years, exciting progress has been achieved through the introduction of molecular and genetic technologies, especially the adoption of Arabidopsis as a model plant in studying cold acclimation. Here, we summarized the recent progress and discuss the major challenges in this field. More detailed information on cold acclimation and freezing tolerance can be found in the following excellent reviews ( Steponkus 1984; Guy 1990; Thomashow 1994, 1999).
FREEZING INJURY IN PLANTS
Under most circumstances, freezing temperatures induce the formation of ice in the intercellular spaces and cell walls of plant tissue. This extracellular ice formation occurs because the intercellular fluid has a higher freezing point than the cytoplasm. In addition, the intercellular liquid normally contains heterogeneous ice nucleating agents, such as dust and ice-nucleating bacterial proteins ( Brush, Griffith & Mlynarz 1994). (In the absence of heterogeneous ice-nucleating agents, pure water remains as a supercooled liquid until −39 °C – the homogeneous nucleation temperature at which water freezes in absence of any heterogeneous nucleating event.) Upon freezing, the ice-state of water has a much lower water potential than liquid solution and this difference increases as temperature decreases ( Guy 1990). Thus, when ice forms extracellularly, there is a sudden drop in water potential outside the cell. Consequently, water from the cytoplasm moves through the plasma membrane by osmosis, leading to cellular dehydration. The net amount of water removed from a cell depends on both the initial solute concentration of the cytoplasm and the freezing temperature which directly determines the water potential of the extracellular ice. For example, freezing at −10 °C results in a water potential of −11·6 MPa. This will remove 90% of the osmotically active water from the cell, assuming the initial osmotic potential of cell is −1 MPa. If the initial osmotic potential of cell is −2 MPa due to accumulation of solutes during cold acclimation, the same freezing only removes 80% of cellular water. Thus, freezing injury is largely an injury caused by cellular dehydration and, for this reason, freezing stress, drought stress and salt stress share many features (and genes) in common.
Dehydration may damage cellular functions in a variety of ways. However, at least in the case of freezing stress, injury normally involves effects on membrane structure and function ( Webb et al. 1994 ; Uemura, Joseph & Steponkus 1995). The plasma membrane has been considered as the primary site of freezing injury since 1912 ( Levitt 1980). The works from Steponkus and colleagues ( Webb et al. 1994 ; Uemura et al. 1995 ) provide evidence that freezing-induced destabilization of plasma membrane involves three different lesions. In protoplasts from non-acclimated rye leaves, the reduction in cell volume at temperatures down to −5 °C is accompanied by loss of plasma membrane surface area due to invagination of the plasma membrane followed by budding off of endocytotic vesicles. Upon rewarming, the melted water is drawn back into the cells. Because of the irreversible loss of plasma membrane, the protoplast bursts before it regains the original volume due to the hydrostatic pressure created by the incoming water. This type of lesion is known as ‘expansion-induced lysis’ ( Webb et al. 1994 ). In contrast, protoplasts prepared from cold-acclimated leaves do not form endocytotic vesicles. Instead, the plasma membrane is retained as exocytotic extrusions that allow re-expansion of the cell upon thawing.
At colder temperatures (and greater dehydration), different cellular membranes are brought into close apposition. Membrane lipids in non-acclimated tissues undergo lateral phase separations. Certain lipids aggregate to form an inverted structure with hexagonal packing symmetry, called HexII phase. The lipid molecules are arranged in cylinders with the head groups oriented toward an aqueous core, 20 Å in diameter, which disrupts the membrane bilayer. The plasma membrane becomes permeable to water and solutes upon rewarming and loses osmotic responsiveness. Formation of the HexII phase is considered as an interbilayer event and may involve the participation of two or more bilayers because the HexII phase is only observed in multilamellar or stacked bilayer regions. Most frequently, the HexII phase is observed in regions where plasma membrane is brought into close apposition with the outer membrane of chloroplasts ( Uemura et al. 1995 ). These two types of freezing-induced membrane destablization are associated with protoplasts isolated from non-acclimated tissues and are largely precluded by cold acclimation ( Webb et al. 1994 ; Uemura et al. 1995 ).
In cold-acclimated protoplasts, freezing injury is associated with ‘fracture-jump lesion’. This type of membrane lesion is exemplified by localized deviation in the freeze-fracture plane of the plasma membrane during cryo-electron microscopy, probably due to the localized fusion of plasma membrane with other cellular membranes, especially the chloroplast envelopes ( Webb et al. 1994 ). It is believed that both HexII and fracture-jump lesions are formed from a common structural intermediate of membranes ( Uemura et al. 1995 ). However, it is unclear why HexII lesions are observed only in protoplasts isolated from non-acclimated tissues whereas fracture-jump lesions are observed exclusively in protoplasts from cold-acclimated tissues. Furthermore, the temperatures at which fracture-jump lesions are observed vary greatly among plant species ( Webb et al. 1994 ; Uemura et al. 1995 ). Little is known about the physiological, biochemical, and molecular basis accounting for this variation, but it is considered a measure of the relative freezing tolerance that can be achieved in different species. It is worth noting that these freezing-induced membrane lesions are primarily observed in isolated protoplasts, and may not reflect what occurs in intact plant cells.
COMPARATIVE BIOCHEMICAL AND BIOPHYSICAL APPROACHES HAVE IDENTIFIED MANY COMPONENTS OF FREEZING TOLERANCE
A common approach to determine the biochemical basis of freezing tolerance has been to compare the metabolite pools or structural components from cold-acclimated and non-acclimated plants, or from plant varieties with different abilities to cold acclimate. Numerous physiological and biochemical changes are known to occur during cold acclimation, and these are shown schematically in Fig. 1. The most notable changes include a reduction or cessation of growth, reduction of tissue water content ( Levitt 1980), transient increase in abscisic acid (ABA) levels ( Chen, Brenner & Li 1983), changes in membrane lipid composition ( Lynch & Steponkus 1987; Uemura & Steponkus 1994), the accumulation of compatible osmolytes such as proline, betaine, polyols and soluble sugars, and increased levels of antioxidants ( Koster & Lynch 1992; Kishitani et al. 1994 ; Murelli et al. 1995 ; Nomura et al. 1995 ; Dörffling et al. 1997 ; Tao, Oquist & Wingsle 1998). These complicated responses have made it very difficult to separate the processes responsible for freezing tolerance from those that merely represent responses to low, non-freezing temperatures. However, it is clear from these comparative approaches that cold acclimation probably requires many changes in cell biology and metabolism. Here, we will briefly review evidence for some of the most important changes that have been revealed by such approaches.
Since cell membranes are thought to be the primary sites of freezing injury, changes in membrane behavior during cold acclimation must be critical to the development of freezing tolerance. Ultrastructural changes in the plasma membrane can be observed within 6 h of the start of cold acclimation in Arabidopsis ( Ristic & Ashworth 1993). The underlying biochemical bases for these changes in membrane behavior are not completely known. Alterations in membrane lipid composition are correlated with membrane cryostability as they have been observed during cold acclimation in all species examined ( Steponkus 1984; Uemura & Steponkus 1994; Uemura et al. 1995 ). There is also direct evidence for the involvement of membrane lipids in freezing and dehydration tolerance. For example, when non-acclimated rye protoplasts are placed in hyperosmotic medium, their plasma membrane buds off endocytotic vesicles as they do during freezing. However, Steponkus et al. (1988) demonstrated that when non-acclimated protoplasts were pre-incubated with mono-unsaturated or di-unsaturated species of phosphatidylcholine so that the phospholipid was incorporated into the plasma membrane, hyperosmotic treatment resulted in the formation of exocytotic extrusions. Disaturated species of phosphatidylcholine did not induce this change. On the other hand, alteration in lipid composition typically lags far behind increases in freezing tolerance ( Uemura et al. 1995 ; Wanner & Junttila 1999), suggesting that other changes must also contribute to membrane stability ( Lineberger & Steponkus 1980; Steponkus et al. 1998 ).
Accumulation of soluble sugars during cold acclimation is well documented in many plants including Arabidopsis, and the time course of sugar accumulation correlates well with development of freezing tolerance during cold acclimation ( Ristic & Ashworth 1993; Wanner & Junttila 1999). In addition, genetic evidence is available to support the role of soluble sugars in freezing tolerance. For example, a mutant of Arabidopsis, sfr4 (see below), impaired in its ability to cold acclimate, does not accumulate sugars in response to low temperature ( McKown, Kuroki & Warren 1996). In contrast, esk1, a constitutively freezing-tolerant mutant, accumulates sugars at warm temperatures ( Xin & Browse 1998). Several roles for sugars in protecting cells from freezing injury have been proposed, including functioning as cryoprotectants for specific enzymes ( Carpenter et al. 1986 ), as molecules promoting membrane stability ( Lineberger & Steponkus 1980), and as osmolytes to prevent excessive dehydration during freezing ( Steponkus 1984). However, soluble sugars alone are insufficient for full freezing tolerance. Several sfr (sensitive to freezing) mutants that accumulate soluble sugars normally during cold acclimation are, nevertheless, defective in freezing tolerance. Transformation of tobacco with a bacterial pyrophosphatase or invertase gene increased the levels of soluble sugars but did not provide any increase in freezing tolerance ( Hincha et al. 1996 ).
Accumulation of betaines has been reported in several plant species in response to low temperature ( Kishitani et al. 1994 ; Nomura et al. 1995 ). Arabidopsis does not accumulate betaines. However, overexpression of a single bacterial gene, choline oxidase, in Arabidopsis results in accumulation of betaine to between 0·1 and 1 μmol g−1 fresh weight ( Hayashi et al. 1997 ). Although low compared to the levels (20–300 μmol g−1 fresh weight) found in plants that naturally accumulate betaines, this increase significantly improved the tolerance to various stresses including cold and freezing ( Hayashi et al. 1997 and Chen T.H.H., personal communication). Even though betaines improved freezing tolerance in Arabidopsis, it is also true that many chilling- and freezing-sensitive species also accumulate betaines ( Yang et al. 1995 ).
Free proline also increases in plants in response to many stresses ( Delauney & Verma 1993). However, its role in stress tolerance remains equivocal. Selection of somatic mutants that accumulate proline has provided correlations with enhanced freezing tolerance in potato ( Van Swaaij et al. 1986 ) and winter wheat ( Dörffling et al. 1997 ). Arabidopsis accumulates proline during cold acclimation but the increase in proline content lags behind the development of freezing tolerance ( Wanner & Junttila 1999). The strongest genetic evidence that proline may contribute to the increased freezing tolerance comes from the isolation of several constitutively freezing-tolerant mutants in Arabidopsis that accumulate proline in the absence of low-temperature treatment ( Xin & Browse 1998). Proline content increases by 10-fold during two days of cold acclimation at 4 °C in wild-type Arabidopsis. In the absence of acclimation, the esk1–1 mutant contains proline at levels 30-fold higher than non-acclimated wild-type plants. At least two other freezing-tolerant mutants also contain high levels of proline (Xin & Browse unpublished) suggesting that proline does play an important role in freezing tolerance. However, proline accumulation is not required for freezing tolerance since some constitutively freezing-tolerant mutants contain the same low levels of proline as non-acclimated wild-type plants.
Apparently, plants employ multiple mechanisms to ensure freezing tolerance. At present, our knowledge of the metabolic changes that contribute to freezing tolerance is incomplete, but information about the biochemical processes contributing to freezing tolerance is essential for successful engineering of freezing tolerance in crop plants. Fortunately, the genetic approaches outlined below offer new ways to evaluate the contribution of individual components toward overall freezing tolerance.
DIFFERENTIAL SCREENING HAS UNCOVERED FAMILIES OF COLD-INDUCED GENES
For the cold-induced genes as a whole, the challenge is to assess the relative contribution of each of them in increasing freezing tolerance. The first question is whether a particular gene has a significant role in freezing tolerance. The answer, clearly, is that some cold-induced genes can be eliminated without measurably affecting tolerance. For example, phenylalanine ammonia lyase and chalcone synthase, which are involved in biosynthesis of anthocyanin, are induced to high levels of expression during cold acclimation. Arabidopsis mutants deficient in either of these genes are not measurably affected in their ability to develop full levels of freezing tolerance ( Leyva et al. 1995 ). Similarly, a null mutant of Arabidopsis, which lacks alcohol dehydrogenase can develop the full level of freezing tolerance ( Jarillo et al. 1993 ).
The second question is to identify the genes that do contribute to freezing tolerance and to determine their relative importance. To date, only one gene, COR15a, has been demonstrated to offer some protection to chloroplasts and protoplasts derived from non-acclimated transgenic plants overexpressing this gene ( Artus et al. 1996 ). Being localized in the chloroplast stroma, the mature COR15a protein is proposed to function by deferring freezing-induced hexagonal II phase formation to lower temperature through the alteration of intrinsic curvature of the inner membrane of the chloroplast envelope ( Steponkus et al. 1998 ). The enhancement of freezing tolerance in a COR15a transgenic line is subtle since no obvious improvement is observed at the whole plant level ( Artus et al. 1996 ; Jaglo-Ottosen et al. 1998 ). Overexpression of other cold-induced genes in transgenic plants also showed little or no enhancement in freezing tolerance ( Zhu, Chen & Li 1996; Kaye et al. 1998 ).
This limited success in demonstrating a role for cold-induced genes could be explained if multiple genes act in concert to increase freezing tolerance. This hypothesis was recently tested by Jaglo-Ottosen et al. (1998) and Liu et al. (1998) . CBF1 (CRT-repeat binding factor) is a transcription activator that binds to a low-temperature-inducible DNA element, the CRT-repeat, which has a core sequence AAGAC in promoters of several COR genes ( Stockinger, Gilmour & Thomashow 1997). (CBF1 is also called DREB1B for drought-regulated element binding factor ( Liu et al. 1998 ).) Overexpression of CBF1 in transgenic Arabidopsis induces the expression at warm temperatures of the entire battery of COR genes that have the common DNA-elements in their promoter region ( Jaglo-Ottosen et al. 1998 ; Liu et al. 1998 ). When tolerance was assayed by an ion-leakage test, the transgenic plants showed 3·3 °C improvement in freezing tolerance over non-transgenic control plants in the absence of cold acclimation. This increase in freezing tolerance is much more dramatic than that achieved by overexpression of COR15a alone. Indeed, overexpression of CBF1 increased freezing tolerance at the whole-plant level ( Jaglo-Ottosen et al. 1998 ), although no detectable increase in freezing tolerance is observed in transgenic plants overexpressing COR15a alone. These experiments demonstrate that CBF1 and CBF1-mediated cold-induced genes play significant roles in freezing tolerance. However, it is by no means clear that the COR genes are the major contributors to this constitutive tolerance since CBF1 very probably induces other genes that contain the CRT element in their promoters. Indeed, many lines of evidence now indicate that the CBF1-COR pathway response is only one of several signalling pathways involved in cold acclimation (see below).
FREEZING TOLERANCE AND PLANT BREEDING
Freezing tolerance is a complex trait with multigenetic inheritance ( Thomashow 1990). Plants vary in inherent freezing tolerance before cold acclimation and in the potential to acquire freezing tolerance during cold acclimation. Genetic analysis of freezing tolerance in crosses of two potato species with contrasting freezing tolerance demonstrated that these two traits are controlled by different sets of genes ( Stone et al. 1993 ). This finding is important as it implies that it may be possible to genetically manipulate different aspects of freezing tolerance and combine them to make a significant improvement in freezing tolerance. Although it is not clear how many genes are involved in freezing tolerance, recent advancement in mapping of quantitative trait loci has permitted the identification of major loci that have a large effect on freezing tolerance ( Pan et al. 1994 ; Galiba et al. 1995 ). Four of the five major loci that control vernalization and freezing tolerance in winter wheat have been mapped on chromosome 5A (Vrn1), 5D (Vrn3), 5B (Vrn4), and 7B (Vrn5) ( Galiba et al. 1995 ). It is interesting to note that the location of Vrn1 on chromosome 5A is homologous to the location of barley gene Sh2 on chromosome 7 (5H) and Sp1 of rye on chromosome 5R. Although the genes that control vernalization response are closely linked with those that control freezing tolerance, these two traits are controlled by separate genes ( Galiba et al. 1995 ). Further fine-mapping of these major loci may eventually lead to the identification of genes contributing to freezing tolerance. Identification of these genes will permit the study of naturally evolved mechanisms of freezing tolerance. Due to the complexity of the genome and paucity of genetic and physical markers in these crop plants, the cloning of these quantitative trait loci could be a huge endeavour.
MUTATIONAL ANALYSIS OF FREEZING TOLERANCE
Recently, three mutational screens have been employed in Arabidopsis to dissect the mechanisms of freezing tolerance. One genetic approach, as reported by Ishitani et al. (1997) , was to generate transgenic plants harbouring a construct of the COR 78 (RD29) gene promoter fused to a reporter gene, firefly luciferase. After mutagenesis of the transgenic plants, mutants with aberrant expression of the reporter gene were isolated. Since the promoter of COR78 has cold-, drought-, and ABA-responsive elements, this approach allows the simultaneous study of the signal pathways for cold, drought, and ABA. Hundreds of mutants with altered luciferase activity have been identified. These mutants fall into three major categories: the cos mutants show constitutive expression of osmotically responsive genes; the los mutants show loss of expression of these genes; whereas hos mutants show hyper-expression after induction by drought, cold or ABA. Analysis of these mutants suggests that multiple signalling pathways cross-talk and converge to activate the COR78 gene. Freezing tolerance has been determined for two of the hos mutants, hos1 and hos2 ( Ishitani et al. 1998 ; Lee et al. 1999 ). Compared with wild-type plants, these mutants are either less freezing tolerant or are less capable of developing freezing tolerance after cold acclimation even though they hyper-accumulate COR gene transcripts when exposed to cold. The defect in freezing tolerance is unlikely to be due to other mutations that decrease plant vigor in general because similar phenotypes are observed in multiple alleles of mutants isolated from different mutagenized populations.
A second approach involved isolating mutants that are defective in developing freezing tolerance after cold acclimation. Warren et al. (1996) have isolated seven Arabidopsis mutants that fail to develop full freezing tolerance even after 2 weeks of cold acclimation. These mutants are named sfr for sensitive to freezing. This approach offers the potential to identify a wide range of signalling components mediating cold acclimation. Some of the sfr mutants are deficient in accumulation of soluble sugars during cold acclimation ( McKown et al. 1996 ). One of these freezing-sensitive mutants, sfr6, was shown to be deficient in CBF1-mediated induction of COR genes ( Knight, Trewas & Knight 1999). Characterization of sfr6 confirms the importance of the CBF1 pathway in cold acclimation. However, except for sfr6, the remaining classes of sfr mutants all show strong induction of the COR genes even though they are partly deficient in the ability to cold acclimate. Thus, COR gene induction is only one component of freezing tolerance.
A third approach involved the isolation of constitutively freezing-tolerant (cft) mutants, i.e. mutants that are more freezing tolerant than wild-type plants in the absence of cold acclimation. Twenty-six constitutive freezing-tolerant mutants were isolated from 800 000 EMS-mutagenized M2 seedlings of Arabidopsis ( Xin & Browse 1998). Due to the complexity of cold acclimation, a single gene mutation that results in significant increases in freezing tolerance is most likely a mutation in a signalling component of cold acclimation, which mediates to the expression of a suite of terminal genes to provide increased freezing tolerance. One of the best characterized cft mutants, eskimo1, tolerates freezing to −10·6 °C without cold acclimation. This improvement in freezing tolerance (5·1 °C over non-acclimated wild type) represents 70% of the freezing tolerance found in fully acclimated wild-type plants. The esk1 mutants contain high levels of proline and soluble sugars but do not express the COR genes in the absence of cold acclimation ( Xin & Browse 1998). This implies that CBF1-regulated genes are not required for the development of certain aspects of freezing tolerance. Interestingly, cold acclimation of esk1 produces plants that are more than 2 °C more freezing tolerant than fully acclimated wild type. Apparently the esk1 mutation is hyperactivating some aspects of freezing tolerance.
Since the cft mutants are freezing tolerant when grown at warm temperatures, the genes that are constitutively activated in these mutants may contribute the freezing tolerance observed. It is possible that genes which are merely responsive to low temperature but play no role in freezing tolerance are not constitutively expressed in the cft mutants at warm temperature. Thus, these mutants may also provide unique opportunities to separate genes that are essential to freezing tolerance from those that are merely responsive to low-temperature exposure. The biochemical and molecular basis for the constitutive freezing tolerance is unknown in any of the cft mutants. Molecular cloning of the genes defined by the cft mutations and extensive analysis of the biochemical and molecular basis of the freezing tolerance in each mutant will likely result in new insight into the mechanisms of plant acclimation to freezing temperatures.
SIGNALLING PROCESSES ARE KEY TO UNDERSTANDING COLD ACCLIMATION
The diverse responses of plants to low temperatures and the complicated biochemical and molecular changes associated with cold acclimation indicate that a complete understanding of freezing tolerance will not be possible until the signalling cascades specific to cold acclimation are defined. In broad terms, it can be assumed that cold acclimation involves a temperature sensor to perceive the low temperatures and a signal transducer to activate the biochemical and gene expression events required for increased freezing tolerance. In this respect, major advances have been made by the identification of DNA sequences, such as CRT, DRE, and ABRE in the 5′-regulatory regions of several COR genes ( Baker, Wilhelm & Thomashow 1994; Yamaguchi-Shinozaki & Shinozaki 1994; Wang & Cutler 1995), and by the cloning of transcription factors that bind to these CRT/DRE elements (CBF or DREB) upstream of COR15a and COR78 ( Stockinger et al. 1997 ; Liu et al. 1998 ). These transcriptional factors belong to small gene families consisting of three members, CBF1, CBF2, and CBF3, organized in a direct repeat on chromosome 4 ( Gilmour et al. 1998 ; Shinwari et al. 1998 ). The transcript levels for all three CBF genes increased within 15 min after plants are exposed to cold temperatures, which is followed by the induction of CBF1-mediated COR genes at about 3 h. Distinct signal pathway(s) must be involved in the induction of the CBF genes since these genes do not have CRT/DRE elements in their promoter region.
It is still unclear how plants perceive the decreasing temperature and relay this signal to the nucleus to activate these transcription factors. Several lines of evidence indicate that changes in cellular calcium levels may be involved in temperature sensing. First, a transient increase in cytosolic calcium occurs almost immediately when plants are exposed to low temperatures ( Monroy & Dhindsa 1995; Knight et al. 1996 ; Sheen 1996), implying that calcium may be involved in the transduction of the low temperature signal. Second, the temperature that increases calcium influx coincides with the temperature that induces cold acclimation ( Monroy & Dhindsa 1995). Furthermore, calcium channel blockers or chelators inhibit the expression of COR (CAS) genes at low temperature whereas calcium ionophore causes calcium influx and induces the expression of COR (CAS) genes in warm temperatures ( Monroy & Dhindsa 1995; Knight et al. 1996 ). Other cellular processes may be required to relay the low-temperature signal since similar calcium increases are also observed in chilling-sensitive plants which are injured by exposure to low temperature ( Knight et al. 1996 ; Sheen 1996). The unsaturation level of membrane lipids or membrane fluidity, has been suggested as a temperature-sensing mechanism in cyanobacteria ( Vigh et al. 1993 ). Whether this process occurs in higher plants is still subject to debate ( Gibson et al. 1994 ).
The plant hormone ABA has also been shown to mediate the development of freezing tolerance. Four lines of evidence suggest that ABA may play a central role in the signal transduction of cold acclimation. First, ABA treatment at normal growth temperatures can increase the freezing tolerance of a wide range of plants including Arabidopsis ( Lang et al. 1994 ). Second, endogenous ABA levels increase in certain plants in response to low temperatures ( Chen et al. 1983 ). Third, ABA-deficient mutants are impaired in developing freezing tolerance during cold acclimation, however, the freezing tolerance can be restored to the wild-type level by adding ABA into the culture medium ( Heino et al. 1990 ; Gilmour & Thomashow 1991). Fourth, ABA treatment can induce all the COR genes ( Heino et al. 1990 ; Gilmour & Thomashow 1991). In contrast to these results, all of the ABA-insensitive mutants examined can cold acclimate to the same level as wild type Arabidopsis ( Gilmour & Thomashow 1991). One possible explanation for these observations is that cold acclimation is mediated by a different ABA receptor or pathway from that regulating other ABA responses. COR genes can be also induced by low temperature in ABA-deficient and ABA-insensitive mutants ( Gilmour & Thomashow 1991; Nordin, Heino & Tapio Palva 1991), indicating that there are low-temperature signalling pathways independent of ABA. Indeed, the CRT/DRE elements present in the promoter regions of COR genes are not responsive to ABA treatment through promoter analysis ( Yamaguchi-Shinozaki & Shinozaki 1994). Rather, a different DNA element, ABRE, is responsible for the ABA inducibility of these COR genes.
Increasing evidence indicates that protein kinases and phosphatases are also involved in transduction of low-temperature signals during cold acclimation in plants. Several protein kinases responsive to low temperature have been isolated from higher plants based on sequence homology or cross-reaction of antibodies ( Anderberg & Walker-Simmons 1992; Jonak et al. 1996 ; Mizoguchi et al. 1996 ; Tahtiharju et al. 1997 ). Using specific inhibitors, Monroy, Sangwan & Dhindsa (1998) demonstrated that protein kinase inhibitor, stauosporine, could prevent the induction of CAS15 by low temperature whereas a protein phosphatase inhibitor, okadaic acid, could induce the expression of CAS15 at 25 °C. They further identified a protein phosphatase 2A as an early target for cold-inactivation. These experiments demonstrate that plants may use many components that are similar to those associated with other signalling pathways to control cold acclimation.
The biochemical and genetic studies discussed above indicate that the signal cascades controlling cold acclimation are likely to be very complex. It is not appropriate to consider cold acclimation as a simple, linear signalling pathway activating the full set of processes required for increased freezing tolerance. Instead, we need to consider a model for cold acclimation in which parallel or branched signalling pathways activate distinct suites of cold acclimation responses ( Fig. 2). Constitutive activation of one of these pathways can result in considerable freezing tolerance without support from other components. Previous studies have shown that the expression of COR genes is induced by both the CBF1 (ABA-independent) pathway and by the bZIP-mediated ABA-dependent pathway ( Gilmour & Thomashow 1991; Nordin et al. 1991 ; Mantyla, Lang & Palva 1995). As discussed above, it is likely that each of these pathways induces other genes as well. It is likely that these two pathways induce overlapping sets of genes rather than providing parallel induction of a single set of genes. The constitutively freezing-tolerant mutant esk1 accumulates high levels of proline but does not constitutively express the COR genes. Proline accumulation in esk1 is achieved by both constitutive activation of Δ-pyrroline-5-carboxylate synthase and the prevention of proline oxidase gene induction by proline, indicating that ESK1 is a signalling or regulatory component. Since esk1 mutant plants do not exhibit induction of the COR genes, ESK1 must define a third signal pathway ( Fig. 2). Several of the cft mutants neither accumulate proline nor activate COR genes (unpublished results). They must define one or more additional pathways that are distinct from the ones described above. Thus, there are at least four separate signalling pathways involved in cold acclimation. In each of the constitutively freezing-tolerant mutants, only one of the signal pathways is activated, and therefore, only partial freezing tolerance is achieved. This conclusion is also supported by the analysis of sfr mutants that are not able to fully acclimate ( Warren et al. 1996 ). Most of the sfr mutants retain over 50% capacity to cold acclimate. The simplest explanation is that each sfr mutation blocks one signalling pathway. Therefore, each mutant is still able to partially cold acclimate through signalling pathways that are not disrupted in the mutant plant. Cloning of the genes defined by the sfr and cft mutants will provide unique insights into the regulation of signalling pathways mediating the development of freezing tolerance.
At present, it is unclear which genes or biochemical processes are essential to the development of freezing tolerance and which are general responses to low, non-freezing temperatures but are not required for freezing tolerance. Except for the induction of a few COR genes, the signal cascades mediating most aspects of cold acclimation, such as increases in ABA, synthesis of compatible osmolytes, and changes in membrane lipid composition, are unknown. The isolation of several series of mutants with altered freezing tolerance now opens new routes to study the processes required for freezing tolerance and to identify components of the signalling pathways that mediate these processes. For example, using micro-arrays or other genomics techniques for transcript profiling, it will be possible to compare expression patterns in cft and sfr mutants with those in wild-type Arabidopsis and thereby identify genes and proteins required for freezing tolerance. At the same time, map-based cloning of the cft and sfr loci, using information from the Arabidopsis genome sequence, will allow characterization of additional steps in the signalling cascade. We thank Vicki Racicot for her critical reading of the manuscript.
This work was supported by US National Science Foundation grant no. IBN-940790 and by the Agricultural Research Center, Washington State University. We would like to thank the two anonymous reviewers for their constructive suggestions.