Differential remodeling of the lipidome during cold acclimation in natural accessions of Arabidopsis thaliana

Authors


(e-mail hincha@mpimp-golm.mpg.de or giavalisco@mpimp-golm.mpg.de).

Summary

Freezing injury is a major factor limiting the geographical distribution of plant species and the growth and yield of crop plants. Plants from temperate climates are able to increase their freezing tolerance during exposure to low but non-freezing temperatures in a process termed cold acclimation. Damage to cellular membranes is the major cause of freezing injury in plants, and membrane lipid composition is strongly modified during cold acclimation. Forward and reverse genetic approaches have been used to probe the role of specific lipid-modifying enzymes in the freezing tolerance of plants. In the present paper we describe an alternative ecological genomics approach that relies on the natural genetic variation within a species. Arabidopsis thaliana has a wide geographical range throughout the Northern Hemisphere with significant natural variation in freezing tolerance that was used for a comparative analysis of the lipidomes of 15 Arabidopsis accessions using ultra-performance liquid chromatography coupled to Fourier-transform mass spectrometry, allowing the detection of 180 lipid species. After 14 days of cold acclimation at 4°C the plants from most accessions had accumulated massive amounts of storage lipids, with most of the changes in long-chain unsaturated triacylglycerides, while the total amount of membrane lipids was only slightly changed. Nevertheless, major changes in the relative amounts of different membrane lipids were also evident. The relative abundance of several lipid species was highly correlated with the freezing tolerance of the accessions, allowing the identification of possible marker lipids for plant freezing tolerance.

Introduction

Freezing injury is a major factor limiting plant growth, yield and thus productivity. The extent to which plants can tolerate low temperatures and freezing is an important determinant of the geographical distribution of plant species. Plants from temperate climates are able to increase their freezing tolerance during exposure to low but non-freezing temperatures for a period of several days to weeks, a process termed cold acclimation (see Smallwood and Bowles, 2002; Xin and Browse, 2000 for reviews).

The molecular mechanisms underlying cold acclimation, and in particular the signal transduction cascades involved in cold signaling, have been the subject of extensive research efforts (see e.g. Hua, 2009; Shinozaki et al., 2003; Thomashow, 1999 for reviews). In particular, the pathway defined by the CBF/DREB1 transcription factors regulating a plethora of downstream genes, the CBF regulon, has been characterized in detail (Shinozaki et al., 2003; van Buskirk and Thomashow, 2006; Thomashow, 2010). In addition, there is evidence from reverse genetic and transcript profiling studies for other, CBF-independent signaling pathways (Shinozaki et al., 2003; Hannah et al., 2005; van Buskirk and Thomashow, 2006).

Although many of the target genes of the cold induced CBF transcription factors are well established, their precise functions in the cold acclimation response are only poorly understood. The only COR (cold regulated) protein for which a clear function has been described is COR15A. Overexpression of this chloroplast-localized protein enhances Arabidopsis freezing tolerance without prior cold acclimation (Artus et al., 1996), presumably by stabilizing the inner chloroplast envelope membrane by direct interaction with the chloroplast-specific galactolipid monogalactosyldiacylglycerol (MGDG) (Steponkus et al., 1998; Thalhammer et al., 2010). In addition, the accumulation of compatible solutes such as raffinose during cold acclimation is, at least partially, under the control of the CBF transcription factors (Gilmour et al., 2000).

Analysis of primary metabolism using gas chromatography–time-of-flight (GC-TOF) MS-based profiling tools has, in addition, provided clear evidence for a major remodeling of the Arabidopsis metabolome during cold acclimation (see Guy et al., 2008 for a comprehensive review). This includes the accumulation of numerous compounds with cryoprotective activity, such as sugars and certain amino acids, which is, at least partially, also under the control of the CBF transcription factors (Gilmour et al., 2000). Although it has been possible to predict the freezing tolerance of different Arabidopsis genotypes (accessions and their F1 crosses) from their metabolite composition with high accuracy (Korn et al., 2010), the exact contribution of any of these metabolites to the freezing tolerance of plants has not yet been established.

It is generally accepted that damage to cellular membranes is the major cause of freezing injury in plants. This has been demonstrated for the plasma membrane (Steponkus, 1984), the chloroplast envelope (Krause et al., 1988) and thylakoid membranes (Hincha et al., 1987). In particular, detailed lipid analyses of the plasma membrane have indicated that there are no acclimation-specific lipids, but that instead complex changes in lipid composition take place during cold acclimation (Uemura and Yoshida, 1984; Lynch and Steponkus, 1987; Palta et al., 1993; Uemura and Steponkus, 1994; Uemura et al., 1995). These include an increase in the relative amounts of diunsaturated species of phosphatidylcholine in the plasma membrane that leads to increased cryostability of protoplasts due to changes in their osmotic behavior (Steponkus et al., 1988). Similarly complex changes have also been described for the lipid composition of the inner and outer chloroplast envelope membranes during cold acclimation (Uemura and Steponkus, 1997), but in this case no particular function could be assigned to any of the changes. More recently the involvement of a galactolipid:galactolipid galactosyltransferase (GGGT) in Arabidopsis freezing tolerance was established through an investigation of the sfr2 mutant (Moellering et al., 2010). While this mutant was not changed in its cold acclimation behavior, it showed a loss of lipid remodeling capability in the outer chloroplast envelope membrane during freezing. Wild-type plants showed a reduction in MGDG and a concomitant increase in digalactosyldiacylglycerol (DGDG) content, as well as the appearance of tri- and tetragalactolipids, but these changes were no longer apparent in the mutant. Monogalactosyldiacylglycerol is a non-bilayer lipid that can severely destabilize membranes during freezing, whereas DGDG is a bilayer lipid that may increase membrane stability (Webb and Green, 1991; Hincha et al., 1998). The physical behavior of the other galactolipids has not been established yet, and how these changes in lipid composition affect cellular freezing tolerance remains to be elucidated (Moellering and Benning, 2011).

While forward and reverse genetic approaches have provided a wealth of knowledge about crucial regulators of cold acclimation and freezing tolerance in plants, these loss or gain of function approaches often result in pleiotropic effects that make a quantitative analysis of the contribution of different genes impossible. An alternative ecological genomics approach therefore relies on the natural genetic variation present in a given species (Weigel, 2012). Natural selection has shaped different accessions by subtle changes at multiple loci, resulting in a balancing of allele combinations in response to a particular environment. Arabidopsis has a wide geographical range throughout the Northern Hemisphere with significant natural variation in its freezing tolerance (Hannah et al., 2006; Zhen and Ungerer, 2008; Zuther et al., 2012).

Here we present a comparative analysis of the lipidomes of 15 Arabidopsis accessions originating from close to the equator to Siberia and Scandinavia that display large differences in freezing tolerance (Zuther et al., 2012). After 14 days of cold acclimation at 4°C the plants from most accessions had accumulated massive amounts of storage lipids, with most of the changes being in long-chain unsaturated triacylglycerides (TAGs), while the total quantity of membrane lipids was only slightly changed. We nevertheless detected major changes in the relative amounts of different membrane lipids. The relative abundance of several lipid species was highly correlated with the freezing tolerance of the accessions, allowing the identification of possible marker lipids for plant freezing tolerance.

Results

We have used ultra-performance liquid chromatography coupled to Fourier-transform mass spectrometry (UPLC-FT-MS)-based high-resolution lipidome analyses (Giavalisco et al., 2011; Hummel et al., 2011) to investigate changes in Arabidopsis leaf lipid composition during cold acclimation. This has allowed us to obtain a relative quantification of the contents of 180 different lipid species, both diacylglycerides (DAGs) and TAGs (Table S1). To assess any possible functional role of changes in lipid content during cold acclimation for plant freezing tolerance, the lipidomic analyses were performed on 15 natural accessions of Arabidopsis that cover the whole range of freezing tolerance currently known in this species (LT50 (temperature of 50% electrolyte leakage) non-acclimated, −4.3°C (Can-0) to −7.7°C (Ms-0); acclimated, −5.4°C (Sah-0) to −12.1°C (N14); Zuther et al., 2012). Unlike earlier studies that specifically investigated changes in diacyl lipids in response to a short-term exposure to low temperature of up to 6 h (Burgos et al., 2011) or 3 days (Welti et al., 2002), we concentrated here on the longer-term (14 days at 4°C) cold acclimation response that also leads to a significant increase in leaf freezing tolerance.

Accumulation of storage lipids is the major quantitative change in the lipidome during cold acclimation

Figure 1 shows typical UPLC chromatograms of total leaf lipids from Arabidopsis plants before (NA) and after (ACC) 14 days of cold acclimation at 4°C. While many differences in the composition of the diacyl (membrane) lipids (elution time up to about 13 min) can be seen, the most striking differences are apparent at later elution times when the triacyl (storage) lipids elute. Both the relative amount and composition of the TAGs were massively influenced by cold acclimation. The relative amounts of all lipids can be found in Table S1.

Figure 1.

 Ultra-performance liquid chromatography elution profiles of total lipid extracted from leaves of the Arabidopsis accession Col-0.
Plants were either grown under non-acclimating conditions (20°C day/18°C night temperature) for 42 days or received an additional cold acclimation treatment at 4°C for 14 days. Mass spectral peak intensities relative to the base peak are shown. The regions where the different lipid classes eluted are indicated. MGDG, monogalactosyldiacylglycerol; DGDG, digalactosyldiacylglycerol; SQDG, sulfoquinovosyldiacylglycerol; DAG, diacylglyceride; TAG, triacylglyceride; PC, phosphatidylcholine.

A comparative analysis of the eluted lipids after FT-MS analysis confirmed these conclusions (Figure 2). While the total relative quantity of membrane lipids showed only small differences between non-acclimated and acclimated conditions in all accessions, the total relative quantity of TAGs strongly increased in 13 of the 15 investigated accessions. The only exceptions were Alc-0 and Ran, which showed no change or a small decrease, respectively. While both accessions are relatively freezing sensitive, there was no obvious relationship between freezing sensitivity and total TAG content, as very sensitive (e.g. Can-0) and very tolerant (e.g. N14) accessions showed similar TAG contents before and after cold acclimation.

Figure 2.

 Changes in the leaf contents of diacyl lipids (A) and triacyl lipids (B) during cold acclimation.
The log2-transformed peak intensities of the two groups of lipids are shown for non-acclimated (NA, light boxes) and cold acclimated (ACC, dark boxes) plants of the 15 investigated accessions. Standard box plots are shown with black horizontal bars in the boxes representing the median, bottom and top of the box indicate the 25th and 75th percentiles (lower and upper quartiles). Whiskers indicate the 1.5-fold interquartile range, circles represent potential outliers. There were two to five replicates per accession and condition. Accessions are ordered by their acclimated LT50 from low to high freezing tolerance (Zuther et al., 2012). ‘Pool’ represents pooled samples (= 8) and indicates technical variation.

The increase in TAG content during cold acclimation was not uniformly distributed across all molecular species. While under NA conditions intermediate chain-length TAGs predominated in all accessions (Figure 3A), this balance was shifted to longer chain fatty acids after cold acclimation (Figure 3B). The shorter-chain TAGs showed hardly any changes in relative abundance, while the longer-chain TAGs, and especially those with a high degree of unsaturation, were clearly increased (Figure 3C).

Figure 3.

 Content of all 61 identified triacylglycerides (TAGs) in the leaves of non-acclimated and cold acclimated plants.
Content (log2-transformed) of all 61 identified TAGs in all 15 accessions in the leaves of non-acclimated (A) and cold acclimated (B) plants. Panel (C) shows the log2 fold change in the contents of the respective TAGs during cold acclimation. Standard box plots are shown with black horizontal bars in the boxes representing the median, bottom and top of the box indicate the 25th and 75th percentiles (lower and upper quartiles). Whiskers indicate 1.5-fold interquartile range, circles represent potential outliers among the means calculated separately from the replicate data of each accession.

This is further emphasized when the TAG species are separated by accession. Figure 4 shows that while Alc-0 and Ran did not accumulate any TAGs during cold acclimation, all others did so to varying degrees, with a preference for TAGs containing long-chain fatty acids (towards the bottom of the figure). These complex accumulation patterns along different molecular species and natural accessions indicate that accumulation of TAGs in the cold is a highly regulated process that depends in an as yet unknown manner on plant genotype. The clustering also indicates that the more freezing-tolerant accessions (indicated by the darker blue labels on top of the heatmap) generally accumulated more TAGs than the freezing-sensitive accessions, although exceptions were clearly present (see above).

Figure 4.

 Hierarchical cluster analysis of the log2 fold change in triacylglyceride (TAG) content of the investigated accessions.
Log2 fold change values are color coded as indicated for each molecular species in every accession. The blue squares above the heat map indicate the relative freezing tolerance of the accessions after cold acclimation (Zuther et al., 2012), with a darker color indicating higher freezing tolerance (lower LT50 values).

Membrane lipid composition distinguishes cold acclimated from non-acclimated plants

While the relative TAG content of the leaves was strongly increased in most accessions in the cold, the total membrane lipid content of Arabidopsis leaves showed only minor changes (Figure 2). However, we detected differences in membrane lipid composition both between accessions and between plants before and after cold acclimation. To compare the compositional variability of membrane lipids and TAGs, we performed a principal components analysis (PCA) separately for the two classes of lipids (Figure 5). This indicated a clear separation between NA and ACC plants along principal component 1 (PC1) based on diacyl lipid composition (Figure 5A), while the separation was much less clear for the TAGs, where NA and ACC samples partially overlapped (Figure 5B).

Figure 5.

 Principal components analysis of the lipid composition of the 15 Arabidopsis accessions.
Panel (A) shows the results for the diacyl lipids and panel (B) the results for the triacyl lipids. Samples from non-acclimated plants are indicated by circles and samples from acclimated plants by squares. The different accessions are color coded and the size of the symbols indicates the freezing tolerance of the leaves (bigger symbols denote higher freezing tolerance). For diacyl lipid data principal components 1 and 2 (PC1 and -2) explain 26% and 13% of the variance in the data set and for triacylglyceride data 79% and 10% of the variance.
Loadings can be found in Table S3.

The changes in the relative contents of the six most abundant classes of membrane lipids during cold acclimation are depicted in Figure 6. The content of MGDG strongly decreased in the cold in almost all accessions, while the content of the other chloroplast galactolipid (DGDG) was either unchanged or slightly reduced. This reduction in DGDG content, however, was confined to the more freezing tolerant accessions. The two most abundant phospholipids [phosphatidylcholine (PC) and phosphatidylethanolamine (PE)], on the other hand, both increased in relative abundance in most accessions, while the ceramide (Cer) content generally increased and glucosylceramide (GlcCer) content decreased in the cold.

Figure 6.

 Changes in the leaf contents of six major classes of diacyl lipids in the 15 investigated Arabidopsis accessions before and after cold acclimation.
Changes in the leaf contents of six major classes of diacyl lipids in the 15 investigated Arabidopsis accessions before (light boxes) and after (dark boxes) cold acclimation. Standard box plots are shown with black horizontal bars in the boxes representing the median, bottom and top of the box indicate the 25th and 75th percentiles (lower and upper quartiles). Whiskers indicate 1.5-fold interquartile range, circles represent potential outliers. There are two to five replicates per accession and condition. Note that the scales in the panels are different, due to the large differences in the contents of the lipid classes. Accessions are ordered by their acclimated LT50 from low to high freezing tolerance (Zuther et al., 2012).

Some lipids showed a strong correlation between relative abundance and freezing tolerance

We performed Pearson correlation analyses to investigate whether the content of particular lipid species may be quantitatively related to leaf freezing tolerance, as indicated by the LT50 values (Zuther et al., 2012). Such analyses were performed separately for data obtained from non-acclimated and from acclimated plants. In addition, we also correlated the changes in lipid content during cold acclimation with the acclimation capacity, i.e. the difference in freezing tolerance before and after acclimation (delta). The results of all correlation analyses are presented in Table S2. The analysis revealed only a small number of highly significant correlations (< 0.01; Figure 7). Surprisingly, non-acclimated freezing tolerance was mostly correlated with the relative content of different highly unsaturated TAG species. Acclimated freezing tolerance, on the other hand, was mostly correlated with the relative content of different molecular species of chloroplast galactolipids (MGDG and DGDG), while acclimation capacity was correlated with changes in the content of TAGs, MGDG, DGDG and a GlcCer. The lipid species identified in these correlation analyses constitute potential metabolic markers for plant freezing tolerance. Obviously, their applicability in plant breeding needs to be tested in further experiments.

Figure 7.

 Significant Pearson correlation coefficients calculated for correlations of diacyl and triacyl lipid contents with freezing tolerance in non-acclimated and acclimated plants and difference in lipid content and freezing tolerance before and after cold acclimation.
Heatmap with significant Pearson correlation coefficients (< 0.01; > 0.65) calculated for correlations of diacyl and triacyl lipid contents with freezing tolerance (expressed as the LT50) in non-acclimated (NA) and acclimated (ACC) plants as well as the difference in lipid contents and freezing tolerance before and after cold acclimation (delta). Blue depicts positive and red negative correlations. Note that a negative correlation with LT50 indicates a positive correlation with freezing tolerance. Numbers in parentheses after the lipid names denote isoforms separated by chromatography, but containing the same number of carbon atoms and double bonds in their fatty acid chains (compare Table S1). DAG, diacylglyceride; DGDG, digalactosyldiacylglycerol; GlcCer, glucosylceramide; GalCer, galactosylceramide; lysoPC, lysophosphatidylcholine; MGDG, monogalactosyldiacylglycerol; PC, phosphatidylcholine; TAG, triacylglyceride.

Discussion

While there have been a large number of studies investigating the effects of cold temperatures on a small subset of lipids, usually phospholipids, only a few studies have used lipidomic approaches to investigate the influence of low temperatures on cellular lipid composition on a larger scale (see Wang et al., 2006 for a review). Most of these have exclusively used the Arabidopsis accession Col-0 and only one also included Wassilewskija (Ws-0) (Li et al., 2008), an accession that is only slightly more freezing tolerant than Col-0 (Zuther et al., 2012). In addition, previous work was focused on the effects of short-term exposure to low temperature (Welti et al., 2002; Burgos et al., 2011) or to freezing and on the activity of particular phospholipase enzymes (Welti et al., 2002; Li et al., 2004, 2008), or on the role of particular groups of lipids such as phosphatidic acid (Li et al., 2004) or oxidized lipids (Vu et al., 2012). Only one study was not only conducted with Arabidopsis (Zheng et al., 2011). Here the effects of drastic daily temperature cycles between −7 and 45°C were investigated in Alpine scree plants and Arabidopsis. However, the extreme temperature regime precludes direct comparison with other studies.

In the present study we have chosen a fundamentally different approach. Instead of focusing on one accession and targeted alterations in lipid composition through transgenic approaches, we investigated the lipidomic response of 15 different accessions to longer-term (14 days at 4°C) cold acclimation treatment. These accessions have previously been characterized in detail for their differential freezing tolerance both before and after cold acclimation (Zuther et al., 2012), allowing us to query how natural variation in freezing tolerance between these genotypes is reflected in the lipidome. In addition, our study provides information about the influence of cold acclimation on leaf storage lipid content and composition.

Interestingly, TAGs showed the largest changes in relative quantity in the lipidome after cold acclimation. The accumulation of storage lipids in the cold is probably not directly related to the concomitant increase in freezing tolerance. The starch content in the leaves of different Arabidopsis accessions also increases under our experimental conditions (Hannah et al., 2006; Guy et al., 2008) and increases in starch or fructan content in different plant species have previously been related to the fact that photosynthetic carbon fixation continues at a reduced rate in the cold while growth is almost completely arrested (Pollock and Cairns, 1991; Livingston et al., 2009). The present data suggest that excess fixed carbon is not only stored in starch but also in TAGs in Arabidopsis.

Triacylglycerides in plants can principally be synthesized through two different pathways. One is from MGDG via DAG, catalyzed by the enzyme GGGT (Sakaki et al., 1990). However, in Arabidopsis leaves this pathway is not activated by cold acclimation, but only by freezing (Moellering et al., 2010; Moellering and Benning, 2011). In addition, the reduction in MGDG content during cold acclimation is too small by far to quantitatively account for the accumulated TAGs. Alternatively, TAGs can be produced from de novo synthesized DAG either directly or after cycling the DAG through PC. In Arabidopsis seeds, more than 95% of all TAGs are synthesized from PC-derived DAG (Bates and Browse, 2011). Whether this is also true for leaves in the cold is not known.

In addition to the increase in total TAG content we also noted a preferential accumulation of TAGs with highly unsaturated fatty acids during cold acclimation and strong correlations between non-acclimated LT50 and the relative amounts of several highly unsaturated TAGs with four to eight double bonds in their three fatty acyl chains. The latter fact is of particular significance because non-acclimated LT50 is not correlated with the amounts of soluble sugars or proline, or the expression levels of the CBF or COR genes (Zuther et al., 2012), which are important for the increase in freezing tolerance during cold acclimation. The level of unsaturation of TAGs could be functionally connected to cellular freezing tolerance. Depending on the fatty acid composition, TAGs in the oil bodies of seeds may crystallize during low-temperature storage, leading to disruption of cellular structures and cell death (Volk et al., 2006). A similar mechanism of damage could be envisioned during the freezing of leaves, and preventing the crystallization TAGs through increased fatty acid unsaturation could be an adaptive strategy for plants in cold climates, such as the more freezing tolerant Arabidopsis accessions. While this obviously has to be tested through appropriate transgenic approaches, it currently constitutes the only hypothesis to explain the observed differences in non-acclimated freezing tolerance between different accessions.

While the total relative TAG content of Arabidopsis leaves strongly increased during cold acclimation, the total relative contents of membrane lipids remained almost unchanged. Within the membrane lipids, however, major compositional changes took place that lead to clearly distinguishable membrane lipidomes between the non-acclimated and cold acclimated plants, in agreement with earlier lipidomics studies on Col-0 and Ws-0 (Welti et al., 2002; Wang et al., 2006). The most striking effect of cold acclimation was a reduction of the MGDG content in most of the accessions. The adaptive value of such a reduction can be explained from the physical properties of MGDG as a non-bilayer lipid (Webb and Green, 1991). Pure MGDG does not form bilayers, but rather assumes a hexagonal (HII) phase. Since HII is an interbilayer structure (Seddon, 1990), it has been proposed previously that the high MGDG content of the chloroplast envelope membranes may lead to HII formation during freezing involving the plasma membrane as well (Steponkus et al., 1998).

While the total MGDG content decreased with cold acclimation, the relative amounts of two specific molecular species of MGDG (34:2 and 34:3) were nevertheless positively correlated with acclimated freezing tolerance. These particular lipids only contain two or three double bonds in their fatty acid chains, while the majority of MGDG molecular species are more highly unsaturated, with four to six double bonds. A correlation of the content of more saturated molecular species of a lipid with freezing tolerance is unexpected, because an increase in lipid unsaturation is generally thought to be important for cold acclimation, due to the need to increase membrane ‘fluidity’ to prevent liquid-crystalline to gel phase transitions in the cold (Nishida and Murata, 1996). However, MGDG is a non-bilayer lipid that shows increased propensity for formation of HII with increasing unsaturation (Gounaris et al., 1983). Therefore, saturated species of MGDG should be more resistant against forming the HII phase during freezing, making an increase in more saturated species of MGDG an adaptive response of plants to increase their freezing tolerance. An increase in the stability of photosynthetic membranes at low temperatures has also been demonstrated for the fad5 mutant with increased saturation levels of MGDG in the fab1 mutant background that was otherwise severely impaired in its low temperature growth (Barkan et al., 2006), indicating that fatty acid unsaturation is not in itself of adaptive value under low-temperature conditions.

A similar correlation was also observed for two molecular species (34:3) of the bilayer lipid DGDG. A natural mixture of DGDG species remains in the liquid-crystalline bilayer configuration even under conditions of severe dehydration (Popova and Hincha, 2003) making a functional interpretation of the observed changes difficult.

The two major phospholipids in plants are PC followed in abundance by PE. After 2 weeks at 4°C, the relative amounts of both lipids were clearly increased in most of the accessions. Increases in phospholipid content during cold acclimation have been repeatedly observed in plants, and it is frequently assumed that this is of adaptive value (see Steponkus, 1984 for a critical review). However, only the relative contents of two molecular species of PC (38:2 and 38:4) were correlated with acclimated LT50. It has previously been shown that an increase in the content of diunsaturated species of PC leads to an increased tolerance of the plasma membrane against freezing and osmotic stresses (Steponkus et al., 1988). However, in the absence of information about the localization of these particular lipids, no final conclusions can be drawn.

In addition to these changes in phospho- and galactolipid contents we also observed an increase in Cer and a decrease in GlcCer contents during cold acclimation. Only the relative abundances of one species of Cer (t18:1/h26:0 or 26:1) and of GlcCer (d18:1/h16:0 or d18:0/h16:1) were correlated negatively with leaf freezing tolerance after acclimation, while the increase in the relative content of another GlcCer (d18:1/c24:0) was positively correlated with the increase in freezing tolerance (acclimation capacity). Although we cannot at present assign specific functions to these molecules, both Cer and GlcCer have been implicated in membrane stability and in various signaling processes. In general, sphingolipids are localized in the plasma membrane, the tonoplast and the endomembrane system of plant cells (see Lynch, 2012 for a recent overview) and in rye the content of GlcCer in leaf extracts closely mirrored the content in the plasma membrane (Cahoon and Lynch, 1991). In mammalian and yeast cells, where sphingolipid metabolism has been studied in much more detail, sphingolipid homeostasis is closely regulated and also coordinated with the metabolism of phospholipids and sterols (Breslow and Weissman, 2010), indicating the importance of sphingolipid homeostasis for eukaryotic cells. In agreement with this, Arabidopsis mutants defective in sphingoid long-chain base Δ8 desaturase, which showed a 50% decrease in GlcCer levels, were severely impaired specifically in their low-temperature growth and survival (Chen et al., 2012).

In addition to such homeostatic effects, sphingolipids also play important roles in signaling across the plasma membrane. It has been generally recognized that the plasma membranes of all eukaryotic cells show segregation into lateral domains, often referred to as rafts or microdomains. Rafts are enriched in sterols, sphingolipids and specific phospholipids that form liquid-ordered domains within the more fluid liquid-crystalline membrane matrix that also harbors specific transmembrane signaling proteins (see Lingwood and Simons, 2010 for a review). During raft formation GlcCer forms stoichiometric complexes with specific phospholipids such as PE (Quinn, 2011). In the context of low-temperature signaling, phytosphingosine phosphate plays important roles in both NO- and MAP kinase-mediated signaling that was located upstream of cold-induced CBF and COR gene expression (Cantrel et al., 2011; Dutilleul et al., 2012), indicating that phytosphingosine phosphate-mediated signaling could have a direct impact on freezing tolerance during cold acclimation.

In conclusion, this paper presents a comprehensive analysis of the natural variation in the Arabidopsis lipidome and of the effects of cold acclimation in relation to leaf freezing tolerance. The data allowed the generation of novel hypotheses in relation to the function of specific lipids or lipid classes in low temperature acclimation that go well beyond the popular ‘homeoviscous adaptation’ hypothesis (Hazel, 1995). In addition, the lipids that showed significant correlations in their relative levels with LT50 constitute potential metabolic markers that may become useful for the breeding of crop plants for enhanced freezing tolerance.

Experimental procedures

Chemicals

All chemicals were purchased from Sigma-Aldrich (http://www.sigmaaldrich.com/) with the highest purity grade available, while organic solvents were purchased from BioSolve (http://www.biosolve-chemicals.com/).

Plant material

Seeds of the A. thaliana accessions Alc-0, C24, Can-0, Col-0, Cvi-0, Ita-0, Kn-0, Ms-0, N6, N13, N14, Oy-0, Ran, Rsch-0 and Sah-0 were kindly provided by the INRA, Versailles, France (http://dbsgap.versailles.inra.fr/Fichier_passport/Info_generale.php). Plants were grown for 42 days under non-acclimating conditions (20°C day, 18°C night) at 16-h day length with light supplementation to reach at least 200 μE m−2 sec−1. At this developmental stage, the plants had either no visible inflorescence, or at most a 1 to 2-mm long inflorescence. This did not change significantly during cold acclimation. For cold acclimation, plants were transferred to constant 4°C for 14 additional days (Rohde et al., 2004; Hannah et al., 2006). Leaf freezing tolerance, expressed as the LT50 (temperature of 50% electrolyte leakage), was determined in a previous study (Zuther et al., 2012).

Lipid extraction and UPLC FT-MS measurement

Lipid analyses for each accession and condition were performed on leaf samples from two to five different plants (compare Table S1). Two mature, fully expanded leaves from individual plants were harvested at mid-day (6–8 h after lights-on). Lipidome profiling was performed as described recently (Giavalisco et al., 2011; Hummel et al., 2011). Briefly, frozen leaf material (50 mg) was homogenized for 1 min at maximum speed in a ball mill (MM 301, Retsch, http://www.retsch.com/), maintaining the samples constantly frozen. Lipids were extracted and fractionated on a UPLC system using a C8 reversed phase column (100 mm × 2.1 mm, 1.7 μm particles; Waters, http://www.waters.com/). Mass spectra were acquired using an Exactive mass spectrometer (Thermo-Fisher, http://www.thermofisher.com/) using altering full scan and all-ion-fragmentation scan mode, covering a mass range from 100 to 1500 m/z. The spectra were recorded from 1 to 17 min of the UPLC gradients.

Statistical procedures

Data normalization, plotting and correlation analysis was done with the software R (http://www.r-project.org/; Gentleman et al., 2004). Relative lipid intensities were normalized to the sample weight. Afterwards, diacyl lipids and triacyl lipids were analyzed separately. Triacyl lipids were not further normalized. To capture differences in membrane composition rather than the quantity of membrane lipids, the sum of diacyl lipids per sample was normalized to the same value in all samples. In detail, the sum of diacyl lipids per sample was normalized to 1 by dividing each diacyl lipid intensity of a given sample by the sum of the diacyl lipids of that sample. Afterwards, to retain the values of the lipid species in the same orders of magnitude as before normalization, diacyl lipid intensities were multiplied by the median of the diacyl lipids sum across samples before normalization.

Principal components analysis of log2-transformed and centered data was conducted with the R package pcaMethods (Stacklies et al., 2007). Pearson correlation coefficients were calculated on mean values per cultivar and condition for lipid intensities and freezing tolerance (LT50 in °C). The significance of correlations was estimated by calculating Pearson correlation coefficients of lipid intensities and 10 000 randomized sets of the LT50 values. A correlation coefficient of 0.65 was calculated for the chosen P-value threshold of 0.01.

Acknowledgements

We thank Änne Eckhard for excellent technical assistance. This work has received financial support from GABI through the European project ANR-06-ERAPG-008 ‘Cold tolerance for the future: the CBF genes and beyond (FROSTY)’.

The authors declare no conflict of interest.

Ancillary