Chilling and freezing: two different stresses requiring different solutions
Identification of a major cis-element in the control of cold gene expression
The CBF transcription factors (TFs) and their regulation
Events downstream of CBFs
A post-genomic view on global transcript changes in response to low temperature
The effect of light and circadian signals on cold gene expression
A receptor for cold?
What are the characteristics of plant cell thermometer(s)?
Low-temperature signalling downstream of perception
Plant species exhibit a range of tolerances to low temperatures, and these constitute a major determinant of their geographical distribution and use as crops. When tolerance is insufficient, either chilling or freezing injuries result. A variety of mechanisms are employed to evade the ravages of extreme or sub-optimal temperatures. Many of these involve cold-responsive gene expression and require that the drop in temperature is first sensed by the plant. Despite intensive research over the last 100 yr or longer, we still cannot easily answer the question of how plants sense low temperature. Over recent years, genomic and post-genomic approaches have produced a wealth of information relating to the sequence of events leading from cold perception to appropriate and useful responses. However, there are also crucial and significant gaps in the pathways constructed from these data. We describe the literature pertaining to the current understanding of cold perception, signalling and regulation of low-temperature-responsive gene expression in higher plants, raising some of the key questions that still intrigue plant biologists today and that could be targets for future work. Our review focuses on the control of gene expression in the pathways leading from cold perception to chilling and freezing tolerance.
‘He withers all in silence, and in his hand unclothes the earth, and freezes up frail life.’
From the poem ‘To Winter’ (1783) by William Blake
I. Chilling and freezing: two different stresses requiring different solutions
Some plant species can survive the ravages of extreme cold (Nilsson-Leissner, 1929); some are capable of enduring temperatures as low as − 196°C (Sakai & Larcher, 1987). At the other extreme, mild chilling temperatures prove lethal to other species. Chilling stress is typically experienced by plants from tropical or sub-tropical zones, and is usually encountered when temperatures fall below c. 10°C, but may occur at higher temperatures than this (Levitt, 1980). Chilling injury is associated with a complex array of cellular dysfunctions, and symptoms include loss of vigour, wilting, chlorosis, sterility and even death (Levitt, 1980). Damage can be reduced in plants that have been acclimated to chilling through gradual exposure to low temperatures. Sub-zero temperatures lead to freezing damage, largely attributable to ice formation, and, to a lesser extent, to the adverse effects of very low temperatures on protein function. The two major sources of freezing damage result from cellular dehydration and membrane injury (reviewed in Thomashow, 1999). When ice forms in the intercellular compartment, the reduction in water potential causes unfrozen water to move from the cytoplasm to the intercellular space. The immediate consequences are similar to those of drought, reducing the amount of water available for cellular processes. On thawing, water moves rapidly back into the cell and this, too, has consequences; expansion-induced lysis (EIL) occurs when the reduced surface area of the cell protoplast is insufficient to contain the new cell volume (Uemura & Steponkus, 1989). Membranes are also subject to injury through fracture jump lesions and lamellar to hexagonal II phase transitions (Webb et al., 1994), making them a major site of freezing damage.
Most temperate plants are chilling tolerant and can be categorized as freezing sensitive or freezing tolerant. Many of the latter exhibit constitutive freezing tolerance (FT), but, in some species, further improvements are acquired through cold acclimation. Cold acclimation is the process whereby plants increase their FT in response to a period of exposure (usually days or weeks) to low positive temperatures in the range 0–5°C (Thomashow, 1999). During this period, transcriptional, biochemical and morphological changes occur that allow the plant to avoid the injuries described above.
1. Cold-regulated gene expression associated with FT
2. A role for cold-regulated gene expression in chilling tolerance
Arabidopsis is a chilling-tolerant species capable of cold acclimation, and therefore has established itself as an ideal model for the study of gene regulation leading to FT (Thomashow, 1994). It is frequently assumed that all cold-upregulated genes in Arabidopsis play a role in survival of freezing temperatures; however, there is no reason why, in such species, chilling tolerance could not be achieved by reactive changes in expression during the early stages of chilling. A positive correlation between the degree of membrane desaturation and chilling tolerance was apparent for some time (reviewed in Nishida & Murata, 1996) before evidence was gained for the transcriptional upregulation of membrane fatty acid desaturases (FADs) by low temperature, suggesting that such changes may contribute to chilling tolerance (Gibson et al., 1994; Berberich et al., 1998; Matteucci et al., 2011). In addition, transcriptomic studies suggest that many genes upregulated early after transfer to chilling conditions may play a role in combating the effects of oxidative stress (Gibson et al., 1994). Chilling in high light causes photoinhibition of photosystem II (PSII; Allen & Ort, 2001), the accumulation of reactive oxygen species (ROS) and resultant photo-oxidative damage (Prasad et al., 1994). A number of common genes are expressed in response to either chilling or the ROS H2O2 (Fowler & Thomashow, 2002), suggesting that tolerance of ROS accumulation during chilling is mediated via transcriptional changes. Altered expression profiles of cold-regulated genes involved in lipid metabolism, chloroplast function, carbohydrate metabolism and free radical detoxification, observed in chilling-sensitive Arabidopsis mutants, supports the suggestion that at least some cold-induced genes play a role in acclimation to chilling temperatures (Provart et al., 2003).
3. A possible role for cold-regulated gene expression in metabolic adjustment to low temperatures
Much attention has been focused on the metabolic changes that occur during cold acclimation to promote subsequent FT (Cook et al., 2004; Guy et al., 2008). Also worthy of consideration is the transcriptional response to low temperature that effects an adjustment of photosynthetic and respiratory activity to levels appropriate to the conditions experienced during acclimation (Stitt & Hurry, 2002). During cold acclimation, growth ceases and available sugars are used to maintain cellular function. A number of genes involved in photosynthetic energy production are downregulated (Strand et al., 1997; Fowler & Thomashow, 2002), presumably to protect against a greater risk of photo-oxidative damage (Hannah et al., 2005).
II. Identification of a major cis-element in the control of cold gene expression
Early in the study of transcriptional responses to low temperature, a region of the COR15A gene promoter was identified that conferred cold-, abscisic acid (ABA)- and drought-responsive expression (Baker et al., 1994). It contained a cis-acting element responsible for drought and low-temperature responsivity, subsequently named the drought-responsive element (DRE; Yamaguchi-Shinozaki & Shinozaki, 1994). This motif (also known as the C-repeat or CRT) was later found in the promoters of many other cold-regulated genes, leading ultimately to the identification of the DREB1 and DREB2 trans-acting factors which bind to it (Stockinger et al., 1997; Liu et al., 1998). In Arabidopsis, the DREB1 (also known as the c-box binding factors, or CBFs) and DREB2 families effect expression in response to low temperatures and drought, respectively (Stockinger et al., 1997; Liu et al., 1998).
III. The CBF transcription factors (TFs) and their regulation
In Arabidopsis, the CBF gene family comprises three closely related members (Gilmour et al., 1998; Shinwari et al., 1998; Medina et al., 1999). CBF1, 2 and 3 (also known as DREB1B, C and A, respectively) are rapidly upregulated by low temperature at the transcriptional level and have overlapping effects on Cold-on regulated (COR) gene regulation (Gilmour et al., 2004). (The term ‘COR’ has been subject to different usage over the years. Here, we use it to denote genes that comprise the DRE-controlled regulon specifically, rather than adopting its more general use to describe all cold-upregulated genes.) A less closely related member of the family, CBF4, is involved in the control of expression in response to drought and ABA, but not cold (Haake et al., 2002).
Transcription of the CBF genes themselves is controlled by positive and negative regulators acting via specific cis-elements in CBF gene promoters, the instability of CBF transcripts at ambient temperatures contributing further to their accumulation under cold conditions (Zarka et al., 2003). Positive regulation is achieved via the MYC TFs ICE1 and ICE2, encoded by genes that are not themselves subject to transcriptional regulation. In higher plants, there appears to be strong evolutionary pressure to achieve regulation of TF activity at the level of protein stability (Park et al., 2011). ICE1 is ubiquitinated by HOS1 (Dong et al., 2006), which leads to its proteosomal degradation under ambient conditions to prevent CBF-mediated expression of COR genes. Consequently, the hos1 mutant shows constitutive COR gene expression (although, interestingly, no increase in FT; Ishitani et al., 1998). HOS1 plays a role in protein ubiquitination that is not limited to low-temperature responses, and other targets include the key photoperiodic flowering time regulator CONSTANS (Lazaro et al., 2012).
Conversely, sumoylation (ligation to small ubiquitin-like modifiers, SUMOs) stabilizes ICE1 and, consequently, siz1 and siz2 mutants (deficient in SIZ1 and SIZ2 SUMO E3 ligase activity) show reduced cold gene expression and FT (Miura et al., 2007). This system of antagonistic regulation of ICE1 (Fig. 1) appears to be conserved in monocots, with functionally equivalent SIZ1 and SIZ2 proteins found in rice (Park et al., 2010). SIZ1 also regulates the expression of the floral regulator FLC through changes in chromatin structure, further support for the suggestion that the pathways leading to vernalization and FT may involve common proteins (Jin et al., 2008).
Negative regulation of the CBFs appears to be important in maintaining an optimal cold-induced transcriptome and minimizing the potentially detrimental effects of sustained CBF expression, including growth retardation (Jaglo-Ottosen et al., 1998). In Arabidopsis, CBFs are negatively regulated by an upstream TF, MYB15 (an R2R3-MYB family protein), which binds to MYB recognition elements in CBF gene promoters (Agarwal et al., 2006). Enhanced CBF expression and FT are observed after cold acclimation in myb15 knockout mutants, whereas overexpression of MYB15 has the opposite effect (Agarwal et al., 2006).
The results of a number of studies have shown that overexpression of CBF1, 2 or 3 in the absence of low temperature mimics the effects of cold acclimation, causing high levels of target gene expression, accumulation of proline and soluble sugars and increasing FT (Jaglo-Ottosen et al., 1998; Liu et al., 1998; Gilmour et al., 2000; Vogel et al., 2005). Early reports that the three Arabidopsis CBFs do not alter the transcription of one another were based on the absence of the DRE/CRT cis-element in CBF gene promoters and the observation that overexpression of CBF1 did not alter levels of CBF3 transcript (Gilmour et al., 1998). However, subsequent studies suggested that, although CBF1 and CBF3 do not regulate one another’s transcription, they are both subject to negative regulation by CBF2 (Novillo et al., 2004, 2007). A mutant lacking expression of CBF2 exhibited higher levels and more sustained expression of CBF1 and CBF3 transcripts, greater expression of some CBF target genes and increased FT. This is difficult to reconcile with the fact that the overexpression of CBF2 also increases FT (Vogel et al., 2005); however, it has been proposed that the duration rather than the absolute amount of CBF1 and CBF3 expression is the determining factor. Further challenge to the idea that the three Arabidopsis CBFs are functionally equivalent came with the discovery that some CBF targets require both CBF1 and CBF3 for their expression (Novillo et al., 2007).
Steady-state levels of CBF transcripts are controlled by the circadian clock (Harmer et al., 2000), and their cold induction is gated by the clock (Fowler et al., 2005). Although the above study found little consequence of this on downstream COR gene expression, other published work has demonstrated that cold-induced RD29A expression levels vary depending on the time of day that cold is experienced (Dodd et al., 2006). This, however, cannot necessarily be attributed to circadian variations in CBF levels, as steady-state transcript levels for many COR genes are themselves subject to circadian oscillations. COR gene transcript levels are highest just before subjective dusk, possibly in anticipation of cooler overnight temperatures (Dodd et al., 2006; Covington et al., 2008). Recent work by Michael Thomashow’s group (Michigan State University, USA) has shown that the key circadian MYB TFs, LATE ELONGATED HYPOCOTYL (LHY) and CIRCADIAN CLOCK ASSOCIATED1 (CCA1), regulate cold-inducible expression of CBFs and confer responsiveness to circadian gating (Dong et al., 2011). Mutants lacking LHY and CCA1 exhibited poor induction of CBF target genes and reduced FT with or without cold acclimation, demonstrating their importance in cold responsiveness and supporting the suggestion that an input of information from the clock is important in cold gene regulation (Dong et al., 2011). It is interesting to note that, although mutants lacking the expression of both LHY and CCA1 lost all circadian regulation of CBF1 and CBF3, some control over CBF2 remained, further support for the assertion that CBF2 is functionally distinct from the other two Arabidopsis CBFs.
CBF and CBF-like TFs exist in numerous species of both monocots and dicots (Jaglo et al., 2001; Qin et al., 2004; El Kayal et al., 2006; Ito et al., 2006; Shan et al., 2007; Pennycooke et al., 2008) with up to 25 possible CBF genes in wheat (Badawi et al., 2007); interestingly, CBFs are present in species that are incapable of FT, including even those that are chilling sensitive. Although the CBF proteins are conserved, the genes they activate appear to play different roles in the low-temperature response, dependent on the species. This is very nicely illustrated by the observation that the heterologous expression of CBFs from chilling-sensitive tomato in Arabidopsis can effect FT (Zhang et al., 2004), whereas Arabidopsis CBFs increase tolerance of chilling and oxidative stress in tomato (Hsieh et al., 2002). The ice1 mutant in Arabidopsis, which is defective in CBF3 expression, is sensitive to chilling (Chinnusamy et al., 2003), supporting the suggestion that CBFs contribute to chilling tolerance in tolerant species. Recent evidence has also emerged for a role in the regulation of dormancy (Kendall et al., 2011).
IV. Events downstream of CBFs
1. Chromatin remodelling
Research into eukaryotic gene regulation has revealed the importance of chromatin remodelling. Chromatin folding and DNA association with nucleosomes affect the accessibility of promoter motifs to which TFs bind. Transcriptional activators recruit chromatin remodelling complexes to overcome these barriers to transcription, either by the removal and destabilization of histones or their modification by a variety of post-translational marks, including acetylation and methylation (Strahl & Allis, 2000). Nucleosome depletion and hyperacetylation of histones H2B, H3 and H4 are generally associated with transcriptionally active chromatin (Struhl, 1998), whereas reduced acetylation and increased nucleosome occupancy correspond to inactive genes.
In the yeast Saccharomyces cerevisiae, GCN5 and ADA2 are components of histone acetyl transferase (HAT) complexes involved in the acetylation of nucleosome histones H3 and H2B, and are associated with transcriptional activation. Arabidopsis knockouts in GCN5 and ADA2b showed normal CBF expression, but reduced COR gene expression, suggesting a role for histone acetylation downstream of the CBFs (Vlachonasios et al., 2003). Histone H3 acetylation and nucleosome occupancy at COR gene promoters increase during cold acclimation, the pattern reversing under deacclimating conditions. These changes are mediated by the CBFs themselves, rather than in direct response to low temperature, demonstrated by the fact that they also occur in CBF overexpressers at ambient temperature (Pavangadkar et al., 2010). Although histone acetylation accompanies COR gene activation, it is not wholly responsible for it; plants overexpressing a truncated CBF2 protein lacking the transcriptional activation domain responded to low temperature with H3 acetylation without alteration in nucleosome occupancy or COR gene activation (Pavangadkar et al., 2010).
Control of histone acetylation/deacetylation has been implicated in cold gene regulation through a number of other studies. HOS15 in Arabidopsis encodes a histone deacetylase that controls the expression of COR genes; hos15 mutant plants accumulate higher levels of some COR gene, but not CBF, transcripts (Zhu et al., 2008). This suggests that HOS15 acts independently or downstream of CBF expression. hos15 mutant plants failed to show alterations in FT resulting from elevated COR gene transcript levels; however, this may be explained by the fact that not all CBF targets were affected. A screen for components involved in cold gene regulation led to the identification of the altered cold-responsive gene 1 (agc1) Arabidopsis mutant, which exhibited normal CBF transcript levels but greater DRE/CRT-mediated expression. These results indicate that AGC1, identified as synonymous with FVE, is a negative regulator of COR gene expression (Kim et al., 2004). FVE is a component of the autonomous pathway leading to flowering and shows high homology to the human retinoblastoma-associated protein RbAp, which, in mammals, represses transcription through the recruitment of histone deacetylation complexes (HDACs). A recent study has confirmed that AGC1/FVE associates with the promoters of COR genes (Jeon & Kim, 2011).
2. The Mediator transcriptional co-activator complex
Mediator acts as a bridge between TFs and RNA polymerase II, transmitting stimulus-specific information to the transcriptional machinery (Conaway & Conaway, 2011). Well studied in other eukaryotes, but only recently identified in plants (Bäckström et al., 2007), Mediator is composed of subunits (c. 34 in plants) that comprise head, middle and tail modules and a dissociable regulatory kinase module (Kornberg, 2005). SFR6 is homologous to the yeast subunit MED16/SIN4 (Li et al., 1995) and, as a tail subunit, is predicted to interact with TFs directly, although no evidence for an interaction between SFR6/MED16 and the CBFs has yet emerged. SFR6/MED16 does not control all cold-regulated gene regulons (Knight et al., 1999), indicating that it effects the actions of specific TFs only. It remains to be seen which of the other subunits composing plant Mediator play roles in the expression of the same or other low-temperature-controlled regulons. Demonstration that MED16 from freezing-sensitive rice is functionally orthologous to that of freezing-tolerant Arabidopsis (Wathugala et al., 2011) suggests that the transcriptional machinery and TFs for responding to low temperature are well conserved across higher plants, but the regulons controlled differ between species.
V. A post-genomic view on global transcript changes in response to low temperature
Microarray studies on Arabidopsis, as well as crop species, have revealed the extent of the massive transcriptional reprogramming that occurs on transfer of plants to low temperatures (Fowler & Thomashow, 2002; Gulick et al., 2005; Vogel et al., 2005; Kilian et al., 2007; Robinson & Parkin, 2008). Extensive analysis of genes whose expression was altered in response to cold suggested that levels of as many as 10 000 transcripts could alter in Arabidopsis, revealing previous estimates to have been conservative (Hannah et al., 2005). The authors of this study re-examined other available data alongside their own, and the consensus indicated that between 4% and 12% of the transcriptome alters in its expression after hours, days or weeks of cold treatment; these figures include both up- and downregulated genes (Hannah et al., 2005).
Cold upregulation occurs in waves; some groups of genes (clusters) show very early upregulation after transfer to cold, whereas others respond much more slowly (Fowler & Thomashow, 2002; Robinson & Parkin, 2008). In addition, some genes are induced transiently, whereas the levels of other transcripts remain high for days (Fowler & Thomashow, 2002; Hannah et al., 2005). These temporal differences in expression pattern may reflect the different functions of the genes; those expressed earliest appear likely to encode TFs or components required either for signalling in response to cold or for chilling tolerance.
The role of ABA in cold-regulated expression has been debated for many years, its study hampered by the fact that the COR genes chosen for examination were also inducible by drought and contained both DRE and ABRE motifs. Although ABA levels increase slightly in response to temperature reductions (Lång et al., 1994), genetic studies fail to agree on whether its perception or accumulation is essential for cold gene expression and FT (Gilmour & Thomashow, 1991; Mantyla et al., 1995). Promoter deletion analysis demonstrated that the cold inducibility of native and artificial reporter constructs did not require the involvement of ABRE, although it should be noted that this study did not examine its possible role in the response to longer exposures (Narusaka et al., 2003). Data from these experiments suggested that, although DRE may act as a coupling element to ABRE in drought-responsive expression (a possible reason for their frequent association), ABRE has no role in cold responsiveness. ABA is capable, however, of eliciting a modest degree of activation via DRE, and has been shown to increase levels of CBF1 transcript and protein, a reminder that the two pathways are not entirely independent of one another (Knight et al., 2004).
Work led by Thomashow sought to identify the means by which genes lacking DRE and unresponsive to ABA could be upregulated by cold; here, ABRE was found to play a surprising supporting role. This work revealed that the EE or EE-like (EEL) elements can impart cold upregulation independently of CBFs, and that their action can be augmented by the coupling of ABRE-like (ABREL) elements with EELs. The incidence of ABRE and EE coupling is higher in the promoters of cold-regulated genes than elsewhere in the genome, suggesting that this may be an important regulatory mechanism in the control of cold gene expression (Mikkelsen & Thomashow, 2009). These data show that the circadian clock can influence other cold-responsive regulons as well as CBF targets, possibly controlling processes other than FT.
A large proportion of cold-inducible genes encode TFs (Hannah et al., 2005), a figure estimated at 25% (Robinson & Parkin, 2008). An alternative approach to the identification of novel cis–trans-activating systems for cold gene upregulation has been to identify cold-inducible TFs. Clearly, this cannot be an exhaustive approach, as not all TFs involved in low-temperature gene regulation are themselves transcriptionally responsive to cold (for instance, ICE1, 2). TF gene families upregulated by cold include the CONSTANS (CO)-like, heat-shock factor (HSF), auxin response factor (ARF) and TCP-domain families, suggesting cross-talk between low temperature and a variety of activities (Hannah et al., 2005). Six TF genes whose upregulation by cold was temporally similar to that of CBF2 were overexpressed in one study. Five had little effect on cold gene expression when overexpressed individually; however, one, ZAT12, reduced CBF transcript levels, although this had little impact on downstream COR gene expression (Vogel et al., 2005). This lack of effect points to a greater level of complexity than might previously have been anticipated, and may suggest a need for co-ordinate upregulation of multiple TFs, as well as their post-transcriptional regulation, in the control of some low-temperature-regulated genes.
VI. The effect of light and circadian signals on cold gene expression
The effects of diurnal and circadian factors on the cold-responsive transcriptome are extensive (Bieniawska et al., 2008) and the results of numerous studies link light perception and the circadian clock with the control of cold-regulated gene expression. Light sensing has been associated with COR gene regulation, although some conflicting findings have been reported. One study identified the red light photoreceptor phytochrome B (PHYB) as a positive regulator of cold-inducible DRE-controlled expression (Kim et al., 2002). However, results of a subsequent study indicated a repressive role for PHYB and PHYD in the control of the CBF-responsive regulon (Franklin & Whitelam, 2007). A number of differences in experimental protocol may account for these apparently contrasting observations. Franklin & Whitelam (2007) also showed that the expression of CBF genes could be activated by low red to far red (R/FR) ratio (an indicator of shade or low light levels) under the influence of the clock, and contributed to FT.
A repressive effect of light, mediated by PHYB, has been confirmed in recent interesting work from Kazuko Yamaguchi-Shinozaki's group (University of Tokyo, Japan). In this study, Phytochrome Interacting Factor 7 (PIF7) was shown to negatively regulate the expression of DREB1C (CBF2) transcripts. PIF7 activity was controlled by TOC1, a component of the circadian oscillator, as well as by PHYB (Kidokoro et al., 2009). Studies such as these, together with the discovery of a role for EE in low-temperature transcriptional regulation, emphasize the importance of context in achieving an appropriate response to cold.
VII. Post-transcriptional regulation
Although this review focuses largely on the low-temperature regulation of expression at the transcriptional level, it is worth noting the influence of temperature post-transcriptionally. The extent to which alternative splicing (AS) of pre-mRNAs occurs in plants is now emerging (Marquez et al., 2012). AS creates mRNA variants that encode novel protein isoforms with altered function, and can be responsible for the generation of large increases in the diversity of translated proteins. Low temperature has been shown to influence the generation of splice variants in genes involved in the control of the circadian clock and the production of proteins which themselves control splicing (Palusa et al., 2007; James et al., 2012).
MicroRNAs (miRNAs) have received a great deal of attention over the past few years as post-transcriptional regulators of expression, their role in plant stress responses becoming evident. Plant miRNAs are of around 21 nucleotides in length and act as repressors by promoting the degradation of mRNA species to which they are complementary, or reducing their translation (Sunkar et al., 2012). Profiling of miRNAs in Arabidopsis (Sunkar & Zhu, 2004; Liu et al., 2008) and rice (Lv et al., 2010) has identified a number of cold-regulated miRNAs. Arabidopsis miR319 targets two classes of TF (Sunkar & Zhu, 2004; Liu et al., 2008). Some miRNAs, for example miR397 and miR169, are upregulated in response to cold across plant species, whereas others behave differently between the plant models studied (Zhang et al., 2009). In one case, at least, a specific miRNA has been linked to stress tolerance: miR398 is downregulated by cold, allowing the increase of two transcripts, CSD1 and CSD2, encoding superoxide dismutases and, in doing so, improving chilling-induced oxidative stress tolerance (Sunkar & Zhu, 2004).
VIII. A receptor for cold?
In searching for cold receptors in higher plants, it would be normal to begin by looking at orthologous systems in other eukaryotes. In mammals, specific receptors for low temperature have been identified, including the menthol receptor (Peier et al., 2002) and, more recently, TRPA1 (Karashima et al., 2009). Mammalian cold receptors are TRPs (transient receptor potential cation channel family members), a specific class of ion channel with characteristic activities and protein domains. Unfortunately, this information is of little assistance in understanding cold sensing in plants, as bioinformatic analysis of genomic sequence data indicates that plants do not possess TRP channels. Since the pioneering work of Minorsky in 1989 (Minorsky & Spanswick, 1989), good evidence of cold-sensitive calcium (Ca2+) channels in plants has accumulated (Ding & Pickard, 1993; Carpaneto et al., 2007), and the idea that they are primary sensors of cold postulated (Plieth, 1999; White, 2009), but no channel has been identified at the molecular level. There are clear functional parallels between animal and plant systems, as elevations in cytosolic Ca2+ concentration ([Ca2+]c) via Ca2+ channels occur very early in both cases (Knight et al., 1991; Peier et al., 2002), and thus are likely to be physically and temporally close to the primary temperature-sensing event.
Interestingly, although the mammalian literature has argued that the Ca2+ channel itself acts as a cold sensor, the evidence in higher plants is that perception occurs upstream of Ca2+ channels. In blue–green algae, it has been demonstrated that membrane fluidity plays an essential role in sensing cold (Murata & Los, 1997; Los & Murata, 2000). Low temperature, through direct physical effects, causes a reduction in membrane fluidity, and this rigidification can be sensed and can activate cold gene expression. In this system, membranes act as cellular thermometers. This led to the idea that membrane fluidity might also act as a sensing mechanism in higher plants. Little indication has emerged, however, of a parallel role in higher plants; the majority of evidence is from the laboratory of Raj Dhindsa (McGill University, Canada). Chemical agents that either increase or reduce membrane fluidity have the predicted effect on cold sensing; benzyl alcohol fluidizes membranes and leads to cold insensitivity, whereas dimethyl sulfoxide (DMSO) rigidifies membranes and thus acts as a ‘cold mimic’, triggering downstream events (cold gene expression and FT) in the absence of cold (Orvar et al., 2000). It must be pointed out that there will always be caveats when using such chemical agents, as they are likely to cause effects in the cell in addition to those intended. Most recently, the work of Harriet McWatters’ group (University of Oxford, UK), using a new method to measure membrane fluidity, provided experimental evidence for the change in membrane fluidity that occurs in response to low temperature in higher plant cells. Genetic data from this study helped to explain how temperature ‘measurements’ acquired through such sensing mechanisms can be used to modulate clock function (Martinière et al., 2011). Measurement of cold-induced gene expression in mutant or transgenic plants with altered membrane lipid saturation or sterol content would be needed to conclude, with any certainty, that membrane fluidity is part of cold sensing in higher plants.
In our own transcriptomic experiments (http://affy.arabidopsis.info/narrays/RefSearch.pl?ref_number=133), we observed no substantial difference in cold-induced gene expression between wild-type, single Fatty Acid Biosynthesis 1 (fab1), FAD 2-2 (fad2-2) mutants and triple fad3/fad7/fad8 mutants, nor a line overexpressing FAD3, despite significantly altered saturation of membrane lipids in all cases. In blue–green algae, membrane fluidity is sensed by specific histidine kinases, for example Hik33 in Synechocystis (Mikami et al., 2002). Again, here the trail goes cold, as no higher plant orthologues for these specific histidine kinases exist.
With a simplistic model in which membrane fluidity acts as a sensor, it is difficult to envisage how such a cellular thermometer could report absolute temperature, as the biophysical changes are likely to be discrete phase transitions. One solution to this conundrum could be that different domains within the membrane are responsible for reporting within particular temperature ranges. In mammalian cells, different temperature ranges are reported by distinct receptors; furthermore, there are changes in lipid raft formation and composition in response to cold (Bali et al., 2009). It is not beyond the realms of possibility that such higher order organization in plant cells occurs and is part of cold sensing. This might be an area worthy of future research attention. Finally, it should be noted that there is some preliminary evidence that enzymatic modification of specific sphingolipids, occurring in response to a reduction in temperature, may be linked to downstream gene expression (Dutilleul et al., 2012). This might imply that such changes, rather than membrane fluidity per se, are important for cold perception.
The work of Raj Dhindsa’s group also reveals a role for the cytoskeleton, downstream of cold perception and upstream of Ca2+ (Fig. 2). Drugs that stabilize microfilaments reduce cold sensitivity, whereas drugs that destabilize microfilaments can induce cold-dependent downstream processes in the absence of cold (Orvar et al., 2000). This work also placed the dynamic changes in microfilament structure upstream of Ca2+ influx into cells. This is consistent with direct measurements of Ca2+ dynamics using recombinant aequorin, in which drugs destabilizing microtubules increased the magnitude of Ca2+ response to cold (Mazars et al., 1997). The observed kinetics of microfilament/microtubule depolymerization in response to cold (Pokornáet al., 2004) seem unlikely to be sufficiently fast, however, to account for the very rapid increases in [Ca2+]c levels that occur (Knight et al., 1991). Stabilizing microfilaments with drugs did not completely inhibit downstream cold gene expression (Orvar et al., 2000), suggesting that cytoskeletal remodelling is not an absolute requirement for cold-induced responses. It seems likely, therefore, that although the cytoskeleton performs some function in temperature detection, it is not the primary cold sensor. More data examining the effect of mutations leading to alterations in cytoskeletal remodelling, for example using the tonneau2 (ton2) mutant (Mazars et al., 1997), would help significantly in resolving this issue in the future, particularly by measuring the cold-induced [Ca2+]c responses in these genetic backgrounds. This would bolster the previous, largely pharmacological, data.
Despite the dominance of the membrane fluidity hypothesis in the literature, membranes are not the only possible cellular thermometers. Two very interesting papers published by Phil Wigge’s group (University of Cambridge, UK) propose that temperature could be sensed via chromatin remodelling (Kumar & Wigge, 2010; Kumar et al., 2012). This is a very attractive hypothesis as it forms a relatively direct link between cold and gene expression. Although this work does not relate to the temperature ranges relevant to cold acclimation and chilling, it is possible that other histone remodelling proteins (e.g. HOS15 and AGC1, described above) or histone subunits may play an equivalent role at lower temperatures. Putative candidates could be obtained from future proteomic analysis of chromatin at different temperatures to determine the cold-conditional presence of specific proteins, or by testing the chromatin remodelling proteins identified in previous mutant screens for altered cold gene expression.
Finally, we must accept that there is scope for more than one thermometer to operate in plant cells, and thus the different hypotheses outlined above are not mutually exclusive. For instance, although there is a wealth of data supporting a role for Ca2+, Ca2+ has no known role in the regulation of gene expression by temperature via specific nucleosomal proteins (Kumar & Wigge, 2010; Kumar et al., 2012). We discuss below how plant cells have the ability to detect different parameters within a temperature drop, which makes it tempting to suggest that there will be several different thermometers.
IX. What are the characteristics of plant cell thermometer(s)?
When attempting to identify higher plant cell cold receptor(s), it is useful to review the literature for evidence of the parameters monitored by such thermometers (Table 1). Evidence that the thermometer detects absolute temperature, rate of change in temperature or another variable might eliminate some candidates. Indeed, the literature provides examples showing that plant cells are able to measure multiple parameters: absolute temperature and its rate of change, as well as the duration of cold. Levels of CBF transcripts in Arabidopsis increase in response to a specific temperature, irrespective of the rate at which cooling occurs (Zarka et al., 2003), providing evidence of absolute temperature sensing. Despite this, it is clear that plant cells can also measure the rate of cooling; electrophysiological analysis of cucumber root cells showed that the magnitude of depolarization in response to cooling was positively correlated with the cooling rate (Minorsky & Spanswick, 1989). This electrophysiological work also asserted a role for Ca2+ in the detection of cooling rate (Minorsky & Spanswick, 1989), later proven by direct measurement of [Ca2+]c kinetics (Plieth et al., 1999). Cooling rates have a very clear effect on the magnitude of [Ca2+]c elevation in response to cold. These data link the two processes, and the striking similarity between the depolarization profiles and the Ca2+ responses (Fig. 3; Minorsky & Spanswick, 1989; Plieth et al., 1999) suggests that they share the same cellular thermometer. These data are consistent with the hypothesis that the primary sensors for cooling rate in plant cells are Ca2+ channels, whereas those sensing absolute temperature are Ca2+ efflux transporters (Plieth, 1999).
Table 1. Evidence for the ability of plants to sense different parameters associated with temperature change
Proposed parameter sensed
Temperature change (°C)
The table shows examples of plant responses to low temperature and lists the likely parameter sensed in each case and the evidence for this. ‘Temperature change’ refers to the initial and final temperature experienced by plants. One example of ambient temperature sensing is included for comparison.
Arabidopsis: [Ca2+]c elevation in roots
Magnitude of [Ca2+]c elevation was dependent on the absolute temperature reached, but to a lesser extent than on cooling rate
Plants transferred from ambient growth temperature to high ambient temperature for 2 h showed reduced histone H2A.Z occupancy at the heat-responsive HSP70 promoter accompanied by altered gene expression. No evidence that this occurs under cold conditions
Gradual cooling over several weeks was accompanied by a reduction in light level and day length, to mimic cold acclimation. These parameters may have contributed to temperature sensing. No response to rapid cold shock
A good example of plant cells having the capacity to measure ‘area under the curve’ (the duration for which temperature remains low) is the suppression of FLC gene expression in Arabidopsis in the vernalization response to prolonged cold (Sung & Amasino, 2005). The mechanism by which this occurs, that is expression of antisense FLC transcript and epigenetic control of expression, is well understood, but the primary ‘integrator’ of temperature × time is still not known.
X. Low-temperature signalling downstream of perception
Before discussing cold signalling in detail, it is important to ask at what level one would expect differences between species that can and cannot cold acclimate? As discussed above, the CBF TFs are conserved between tomato (chilling sensitive) and Arabidopsis (chilling tolerant and able to cold acclimate), and activate genes appropriate to the physiology and environment of each plant species. This fact strongly implies that signalling upstream of CBFs (and possibly upstream of most TFs involved in cold-induced expression) is conserved between species, irrespective of the level of tolerance achieved by the species. Consistent with this idea is the fact that the Ca2+ responses of Arabidopsis (chilling tolerant) and tobacco (chilling sensitive) are very similar (Knight et al., 1996).
If not the primary event in response to cold, then the [Ca2+]c elevation in response to cold, is one of the earliest signalling events. The closest events downstream of Ca2+, temporally and spatially, are the activation of anion channels (Lewis et al., 1997) and Ca2+-dependent proteins (Monroy et al., 1993). It has been shown through numerous studies using different approaches that Ca2+ activates, and is necessary for, the expression of cold genes (Knight et al., 1996; Tähtiharju et al., 1997; Sangwan et al., 2001). However, it is unlikely that Ca2+ simply acts as a ‘switch’ for these processes (Scrase-Field & Knight, 2003), and it has been suggested that more complex information is encoded through the ‘Ca2+ signature’, the precise kinetics of the [Ca2+]c elevation. Ca2+ signatures in response to cold differ in plants that have encountered low temperature previously (Knight et al., 1996), a phenomenon referred to as ‘cold memory’. In some cases, altered Ca2+ signatures occurring after stress memory are associated with modified gene expression patterns, making it possible that they encode information that defines the response (Knight et al., 1998). We suggest that the cytoskeleton could play a role in integrating the information relating to previous environmental experiences; stress treatments alter the cytoskeletal configuration, potentially resulting in a modified Ca2+ signature. It would be particularly interesting to observe the consequence of drugs affecting cytoskeletal rearrangements (Mazars et al., 1997; Orvar et al., 2000) on stress memory (Knight et al., 1996).
Most recently, it has been demonstrated that Ca2+ regulates the two most common promoter motifs in cold-induced genes, CRT and ABRE, and that these are likely to respond differentially to specific Ca2+ signatures (Whalley et al., 2011). There is some evidence that CBF gene expression is transcriptionally regulated by Ca2+ (Cataláet al., 2003; Doherty et al., 2009), but this does not preclude Ca2+ regulation at the post-transcriptional level. Further work will be required to elucidate at what level Ca2+ controls gene expression, and whether additional promoter motifs are involved. Even more interesting will be the determination of how the different ‘Ca2+ signatures’ are ‘read’ by the cell to induce the correct specific TF. It should be noted that, when the rate of cooling is low, there is no detectable Ca2+ response (Plieth et al., 1999), most probably because Ca2+ movement is localized in the cell to just around the cytosolic face of the plasma or tonoplast membranes (Knight et al., 1996). Yet, under these conditions, cold induction of COR gene expression is still dependent on Ca2+ (Tähtiharju et al., 1997). This argues that the ‘decoders’ of the Ca2+ signatures will also be localized specifically, and it will thus be important to discover the locations of individual members of families of such decoders in the future, as carried out for all 10 calcineurin B-like calcium sensors (CBLs) in Arabidopsis (Batistic et al., 2010).
The activation of anion channels by cold-induced [Ca2+]c elevations is likely to further enhance the depolarization of the membrane in response to cold (Lewis et al., 1997), but there is no clear link to downstream cold gene expression. The role for Ca2+-binding proteins activated by cold-induced [Ca2+]c increases is much clearer. It has been known for some time that genes encoding the Ca2+-responsive protein calmodulin (CaM) are themselves cold upregulated (Polisensky & Braam, 1996). Early (largely pharmacological) work suggested that CaM activity is necessary for the expression of cold genes in a number of systems, for example Arabidopsis (Tähtiharju et al., 1997), although overexpression of CaM has been reported to inhibit the expression of COR genes (Townley & Knight, 2002). These studies did not examine whether CaM affected TF activity directly; however, later work has shown that CBF2 expression is regulated by CAMTA3 and CAMTA1, two TFs that bind CaM (Doherty et al., 2009). The camta1camta3 double mutant is consequently unable to achieve full cold acclimation (Doherty et al., 2009). Most recently, a GT-2 subfamily TF, shown to be a CaM-binding protein, whose expression increases in cold, has been demonstrated to be capable of inducing cold gene expression when overexpressed (Xi et al., 2012).
Apart from CaM, two other Ca2+-binding protein systems are worth noting; Ca2+-dependent protein kinases (CPKs) and CBLs, which, in turn, regulate Ca2+ sensor-associated protein kinases (CIPKs). At the global level, there is good evidence for Ca2+ regulation of the phosphoproteome in response to cold. An early proteomic study revealed a significant effect of cold on the phosphoproteome of alfalfa, and that Ca2+ was responsible for regulating the majority of these events (Monroy et al., 1993). As with CaM, there is ample evidence of genes encoding CPKs themselves being cold induced (Ludwig et al., 2004). The possibility of cold-induced genes being regulated by CPK activity specifically was raised early on through pharmacological approaches (Tähtiharju et al., 1997) and the ectopic overexpression of constitutive CPK constructs (Sheen, 1996; Saijo et al., 2000). However, loss-of-function genetic studies provided no evidence for the necessity of CPKs in cold gene expression, most probably because of genetic redundancy. However, loss-of-function experiments showed that CIPK3 is necessary for the induction of cold genes in Arabidopsis (Kim et al., 2003). Work on Ca2+-dependent phosphorylation in response to cold has not been limited to protein kinases, and there are examples of cold-responsive Ca2+-dependent protein phosphatase activity (Monroy et al., 1998).
There is also evidence of a mitogen-activated protein (MAP) kinase cascade leading from cold to gene expression. The MAP kinase kinase MKK2 phosphorylates MAP kinases MPK4 and MPK6 in response to cold (Teige et al., 2004). The importance of this event is evidenced from the reduced ability of mkk2 mutants to cold acclimate (Teige et al., 2004). Interestingly, overexpression of MKK2 leads to the induction of many cold genes in the absence of cold, consistent with MKK2 being a positive regulator of cold gene expression (Teige et al., 2004). There appears to be no evidence that this MAP kinase cascade activation by cold occurs downstream of Ca2+ and, indeed, there is evidence to suggest that MAP kinase pathways might act independently of Ca2+, and specifically that MPK4 and MPK6 operate independently of CPKs, albeit in response to salt (Mehlmer et al., 2010; Wurzinger et al., 2011).
It is not clear how the Ca2+-dependent and Ca2+-independent (MAP kinase) pathways leading from cold perception impinge on the post-translational modifications of TFs responsible for the regulation of cold gene expression (Fig. 1). This must surely be a focus for future research. It would be particularly interesting to determine whether HOS1 or SIZ1/SIZ2, or other components of the ubiquitin/sumo-ligase complexes, interact with Ca2+-binding signalling proteins. It would also be interesting to test whether Ca2+-binding signalling proteins interact directly with the TFs regulating cold gene expression, for example CBFs, as has been shown for other TFs, for example ABF4 which binds CPK32 (Choi et al., 2005).
XI. Unresolved questions
1. What is the molecular identity of the cold-regulated Ca2+ channel(s) in plants?
Although electrophysiological identification of Ca2+ channel activities has been possible, for instance through the work of Carpaneto et al. (2007), none of the genes encoding the proteins responsible for these activities have yet been cloned. This is in stark contrast with the situation in animals (Peier et al., 2002; Karashima et al., 2009). A great number of mutant screens for cold-induced gene expression and cold tolerance have failed to identify any plant Ca2+ channels. It is thus possible that their identification requires a more targeted screen, searching directly for altered cold-induced Ca2+ responses in mutants expressing the Ca2+ reporter, aequorin, an approach that has yielded results for elicitor responsiveness (Ranf et al., 2012). However, it is possible that plants use an array of Ca2+ channels to respond to low temperature, and this may be the reason why none have been identified through a loss-of-function screen. Alternatively, functional cloning in a heterologous system has been successful in the identification of a Ca2+-sensing receptor in Arabidopsis (Han et al., 2003) and could be employed here.
2. How do different Ca2+-binding signalling proteins decipher the cold Ca2+ response?
[Ca2+]c elevations occur in response to a wide range of temperature reductions, including those well above the range required to trigger cold acclimation. Indeed, drops from 40 to 30°C can elicit a [Ca2+]c elevation (Larkindale & Knight, 2002). We have still to learn how plants distinguish between these and the changes in [Ca2+]c that occur when the temperature drops to below 5°C during cold acclimation (Knight & Knight, 2000). The activity of Ca2+-sensing proteins (CaM, CPKs, CBLs) could be responsive to the precise Ca2+ signature generated. It would be interesting to measure the activity of Ca2+-responsive proteins in vitro in response to different Ca2+ signatures, a method successfully employed to determine the optimal frequency of Ca2+ oscillation for CaMKII (De Koninck & Schulman, 1998). It is also possible that, in the longer term, the expression of Ca2+-sensing proteins is controlled directly by temperature, sensed by other thermometers or that their cellular targeting changes in response to temperature. An alternative view is that [Ca2+]c changes inform the cell only of a reduction in temperature, and that other Ca2+-independent signalling mechanisms overlay information about absolute temperature to inform the plant as to the type of response required. Future identification of Ca2+ channels will enable a genetic approach to be applied to the question of what information is encoded purely through Ca2+.
3. Why is there a need for multiple pathways leading from cold perception to response?
The existence of parallel mechanisms (e.g. Ca2+-dependent and MAP kinase-dependent pathways; Fig. 1) to mediate the regulation of gene expression may suggest that each conveys a different type of information, for instance cooling rate and absolute temperature, and that together they provide a complete picture of the temperature reduction. It would be interesting to discover whether mutations in components for each pathway confer differential sensitivity to particular temperature parameters. An alternative explanation could be that different groups of genes fall under the control of each sensing system. Evidence that specific genes respond selectively to particular forms of temperature reduction comes from a study of wheat CBF genes, in which some were transcriptionally responsive to rapid cold shock, whereas others were upregulated by slow cooling (Winfield et al., 2009, 2010). It will be interesting, in the future, to determine which groups of genes are regulated by these different mechanisms and, from their role and function, to determine the significance. Whether this analysis will distinguish between the genes involved, for instance, in chilling and FT, or vernalization, remains to be seen.
4. How can we distinguish between the cold-regulated genes involved in chilling and FT?
Genes expressed in response to temperatures below 5°C in species capable of cold acclimation may be required for the gain and maintenance of chilling tolerance or the acquisition of FT. Transcriptomic profiling in species such as Arabidopsis does not distinguish between the two, and further steps are required in order to identify the genes specifically required for chilling tolerance. Comparison of the transcriptome of chilling-sensitive and freezing-tolerant species may be of assistance here. Interestingly, one study has shown that less divergence exists between the cold-inducible transcriptome of Arabidopsis and rice than might have been anticipated (Narsai et al., 2010). A recent study investigated the cold-responsive transcriptome in two species of potato, with and without the ability to cold acclimate, and compared these with Arabidopsis. This study not only confirmed the utility of this approach to identify genes associated with one particular form of cold tolerance; it also added to the growing evidence that there is a high level of conservation of components between chilling and FT pathways with only the expression of the target genes showing alterations (Carvallo et al., 2011).
5. If plants are so responsive to temperature changes, how do they distinguish between those that are worthy of note and those that should be ignored?
Plants can detect temperature changes of as little as 1°C (Knight & Knight, 2000), but such fluctuations would be expected throughout a normal day; indeed, leaf temperatures may change by > 10°C when the sun becomes obscured by cloud. It seems imperative that the plant does not engage a full cold acclimation response every time this is experienced; therefore, a system of checks is likely to exist to distinguish harmless fluctuations from important warning signals. This might involve sensing the duration for which temperatures remain low, and it may be that, to provoke the cold acclimation response, more than one independent confirmation of low temperature is required. This could be resolved by a genetic approach to knocking out individual cold signalling pathways. There can be no doubt that the circadian clock and light quality signals provide important contextual information on how relevant low-temperature changes are to the plant. A sudden drop to 5°C at midday requires a very different interpretation to the same temperature drop experienced at dusk or during the night. The detection of multiple temperature parameters through parallel signal transduction systems may be one safeguard against wasteful and inappropriate responses.
We acknowledge the work of our colleagues that we did not have space to describe.