Identification of SFR6, a key component in cold acclimation acting post-translationally on CBF function


  • Heather Knight,

    Corresponding author
    1. School of Biological and Biomedical Sciences, Durham University, South Road, Durham, DH1 3LE, UK, and
    Search for more papers by this author
  • Sarah G. Mugford,

    1. School of Biological Sciences, Royal Holloway, University of London, Egham, Surrey, TW20 0EX, UK
    Search for more papers by this author
    • Present address: Department of Metabolic Biology, John Innes Centre, Norwich Research Park, Colney, Norwich, NR4 7UH, UK.

  • Bekir Ülker,

    1. School of Biological and Biomedical Sciences, Durham University, South Road, Durham, DH1 3LE, UK, and
    Search for more papers by this author
  • Dahai Gao,

    1. School of Biological Sciences, Royal Holloway, University of London, Egham, Surrey, TW20 0EX, UK
    Search for more papers by this author
    • Present address: Key Laboratory of Arid and Grassland Ecology, Lanzhou University, Lanzhou 730000, China.

  • Glenn Thorlby,

    1. School of Biological Sciences, Royal Holloway, University of London, Egham, Surrey, TW20 0EX, UK
    Search for more papers by this author
  • Marc R. Knight

    1. School of Biological and Biomedical Sciences, Durham University, South Road, Durham, DH1 3LE, UK, and
    Search for more papers by this author

(fax +44 191 3341201; e-mail


The sfr6-1 mutant of Arabidopsis thaliana was identified previously on the basis of its failure to undergo acclimation to freezing temperatures following exposure to low positive temperatures. This failure is attributed to a defect in the pathway leading to cold on-regulated (COR) gene expression via CBF (C-box binding factor) transcription factors. We identified a region of chromosome 4 containing SFR6 by positional mapping. Fine mapping of the sfr6-1 mutation proved impossible as the locus resides very close to the centromere. Therefore, we screened 380 T-DNA lines with insertions in genes within the large region to which sfr6-1 mapped. This resulted in the identification of two further mutant alleles of SFR6 (sfr6-2 and sfr6-3); like the original sfr6-1 mutation, these disrupt freezing tolerance and COR gene expression. To determine the protein sequence, we cloned an SFR6 cDNA based on the predicted coding sequence, but this offered no indication as to the mechanism by which SFR6 acts. The SFR6 gene itself is not strongly regulated by cold, thus discounting regulation of SFR6 activity at the transcriptional level. We show that over-expression of CBF1 or CBF2 transcription factors, which constitutively activate COR genes in the wild-type, cannot do so in sfr6-1. We demonstrate that CBF protein accumulates to wild-type levels in response to cold in sfr6-1. These results indicate a role for the SFR6 protein in the CBF pathway -downstream of CBF translation. The fact that the SFR6 protein is targeted to the nucleus may suggest a direct role in modulating gene expression.


Plants encounter a range of environmental conditions during their lifetime, including reductions in temperature, and need to be able to adjust their metabolism and physiology to cope with these changes. In temperate climates, plants that over-winter require the ability to acclimate to freezing temperatures in order to survive. Cold acclimation results in significant increases in the frost tolerance of plants, and the minimum temperature tolerated by a plant can be lowered considerably by acclimation (Thomashow, 1999; Smallwood and Bowles, 2002).

The process of cold acclimation is triggered by exposure to low positive temperatures, usually for a period of days or weeks (Guy, 1990). Exposure to temperatures typically in the range 2–5°C leads to increased expression of a battery of genes known as COR (Cold On-Regulated) genes, which together contribute towards improved freezing tolerance and changes in both the physical structure of the cell and its metabolism (Guy et al., 2008). The individual functions of many of these genes are unclear, but a number are known to encode enzymes involved in the production of compatible solutes that reduce the impact of osmotic stress resulting from withdrawal of water from the cell during freeze–thaw cycles. By increasing the osmotic potential of the cytosol, expansion-induced lysis is reduced (Steponkus, 1984). It is unsurprising, therefore, that most COR genes are also inducible by osmotic and drought stresses. Other COR genes encode gene products that are active in stabilizing membranes to withstand freeze-induced damage, altering the lipid composition of membranes, combating oxidative stress or acting as chaperones to protect cellular processes during periods of low temperature (Thomashow, 1999).

Arabidopsis thaliana is a good model plant for cold acclimation as it is a temperate over-wintering species, and it has been used for many years to dissect this complex process (Thomashow, 1994, 2001). Research into the potential for individual Arabidopsis COR genes to influence frost tolerance in isolation has, with some exceptions, been fruitless, and it is generally agreed that the COR genes can only exert their positive effects in concert (Thomashow, 1990, 1994). For this reason, a forward genetics approach is particularly suitable for identification of key components in the pathways leading to cold acclimation. sfr (sensitive to freezing) mutants of Arabidopsis were isolated by laborious lethal screening of families produced from EMS-mutagenized Columbia seed (Warren et al., 1996). The eight sfr mutants that were isolated survived positive acclimating temperatures but failed to increase their freezing tolerance to maximum levels after the acclimation period (Warren et al., 1996). Of the SFR genes, SFR2 has been cloned (Thorlby et al., 2004) by classical mapping techniques, as has SFR3 (A. Lytovchenko, A.R. Fernie (Max-Planck Institute of Molecular Plant Physiology, Potsdam-Golm, Germany), G. Warren and G. Thorlby, unpublished results), but the others remain unidentified. These include SFR6, which is a particularly intractable subject for classical mapping as it resides close to the centromere of chromosome 4, an area in which recombination rates are particularly low (Schmidt et al., 1995; Drouaud et al., 2006). For this reason, the identity of SFR6 has remained impossible to ascertain.

The CBF family of AP2-type transcription factors (also known as the DREB1 family) has three members (Gilmour et al., 1998; Shinwari et al., 1998) that activate COR gene expression via the well characterized CRT (C-repeat) promoter element, the core sequence of which is CCGAC (Baker et al., 1994). This element is also activated in response to osmotic stress by a second group of AP2-type transcription factors (TFs) known as the DREB2 family, and for this reason is also described as the drought-responsive element (DRE) (Liu et al., 1998). Over-expression of CBF1, 2 or 3 induces ectopic COR gene expression and a concomitant increase in freezing tolerance (Jaglo-Ottosen et al., 1998; Liu et al., 1998; Gilmour et al., 2004). There is a fourth member of the CBF family (CBF4) that plays no role in cold acclimation but mediates responses to ABA and drought (Haake et al., 2002). It was originally suggested that CBF1–3 could not be auto-regulated as these genes do not themselves contain CRT elements (Gilmour et al., 1998); however, more recent work suggests that CBF1 and 3 are negatively regulated by CBF2 (Novillo et al., 2004, 2007).

The CBFs are very widespread in the plant kingdom, and functional homologues have been identified in a variety of species including Brassica napus, wheat (Triticum aestivum), rye (Secale cereale) and tomato (Lycopersicon esculentum) (Jaglo et al., 2001), eucalyptus (Eucalyptus gunnii) (El Kayal et al., 2006), chilli pepper (Capsicum annuum) (Kim et al., 2004) and grape (Vitis vinifera) (Xiao et al., 2006). Interestingly, the target genes for the CBF TFs vary considerably between species, consistent with the fact that CBFs are present even in species that originate from climates without frost and that do not normally cold-acclimate (Zhang et al., 2004). In the case of tomato, AtCBF1 over-expression is associated with increased chilling tolerance (Hsieh et al., 2002). Over-expression of AtCBF1 increases freezing tolerance in species such as Brassica napus (Jaglo et al., 2001) and Solanum tuberosum (Pino et al., 2007). The widespread occurrence and functionality of CBF-like genes makes the CBF pathway an obvious target for exploitation to engineer stress tolerance in plants.

Our previous work showed that the sfr6 mutant is severely impaired in its ability to express COR genes in response to cold or drought stimuli (Knight et al., 1999), and that COR protein levels are accordingly reduced (Boyce et al., 2003), demonstrating clearly the necessity of COR proteins for cold acclimation. We specifically showed that SFR6 controls expression of the CBF-regulated cold genes via the CRT (Boyce et al., 2003). A number of cold-inducible genes do not contain the CRT element and are not regulated by CBFs; these are unaffected by the sfr6 mutation. Our conclusion from bioinformatic analysis indicating that only those COR genes comprising the CBF regulon are affected in sfr6 was strengthened by data from experiments with a CRT–LUC reporter showing that the mutant is specifically impaired in its ability to induce gene expression via this element (Boyce et al., 2003). We showed that all three CBF genes are expressed normally in the mutant (Knight et al., 1999), suggesting that SFR6 affects COR gene transcript levels by acting post-transcriptionally of the CBF genes. In the present study, we show that SFR6 is a large nuclear protein that acts post-translationally of the CBFs, and identify SFR6 using an alternative to classical gene mapping.


Identification of the SFR6 gene

Coarse mapping of SFR6 determined that its position was within an interval of approximately 4.4 Mbp between the markers GA1 and nga8 (Figure 1a). Thus SFR6 is located on chromosome 4 close to the centromeric region. To fine map this interval, we generated a large mapping population by crossing the sfr6 mutant (in the Col-0 background) with Landsberg erecta (Ler) wild-type. In total, 1500 individuals of this mapping population were screened. Initially, 180 mapping lines were screened with the CAPS markers m228 (cer433228) and m973 (cer428973) (details of these and all further markers used are available in Table S1.) These detected 32 recombination events: 31 to the right and one to the left of SFR6 (Figure 1b). Two further markers to the left of m973 were used to further refine the map position of SFR6, and resulted in 19 recombinations to the left of m717 (cer442717) and four to the left of m349 (cer433349) (Figure 1b). An additional 1320 plants were screened using m228 and m349 as flanking markers. This gave a further 32 recombinants, resulting in a total of 13 on the left of SFR6 and 24 on the right (Figure 1c). These were tested with newly generated markers in an attempt to further reduce the region containing SFR6. When the recombinants to the left were tested using markers closer to SFR6 (m005, m602, m696 and m930), none were still recombinant. Testing of the recombinants to the right slightly reduced the region, in that m586 was separated from SFR6 by four recombinants (Figure 1c). Additional markers to the left of m586 were, however, no longer recombinant. Therefore, despite screening a population of 1500 mapping lines, we were only able to reduce the mapping interval to around 2.6 Mbp (between markers m228 and m586), due to the low frequency of recombination in the centromeric region. It was thus clear that it would be impossible to generate the numbers of lines required to identify SFR6 by conventional mapping.

Figure 1.

 Map-based cloning of SFR6.
(a) Region of chromosome IV containing SFR6, showing the position of the centromere (indicated by the oval shape) in relation to known markers.
(b) The same region, showing the position of molecular markers used for mapping (the markers are described by the last three digits of their number rather than the full name for reasons of clarity), and the number of recombinants obtained from the first 180 individuals of the mapping population.
(c) As in (b), showing additional markers used for genotyping the remaining 1320 individuals from the mapping population.

We therefore turned to an alternative approach. Under the growth conditions we use, sfr6 is clearly recognisable within 10–14 days of germination on MS agar, as the mutant seedlings are paler and have larger cotyledons than the wild-type (Figure 2). We therefore sought to re-isolate an sfr6 mutant, from a T-DNA-tagged population to achieve gene identification using these criteria. Using the TAIR database (, we searched for T-DNA insertions in genes within the region between the two flanking markers m228 and m586 (Figure 1). At the time the work was performed, this region was predicted to contain 585 annotated genes. We ordered 429 insertion lines that were listed as having insertions in genes within this interval. We were able to screen 380 of these lines successfully (49 lines could not be screened due to difficulty in germination or growth, or them being unavailable from the stock centre). The vast majority of these were SALK lines from the laboratory of Joseph Ecker, with some originating from the SAIL collection. We screened the lines for the visible phenotype described above (Figure 2). Only one line amongst the 380 tested displayed a convincing sfr6 phenotype; this line (SALK_048091, NASC ID N048091) contained an insertion into the coding region of At4g04920 (Figure 2). Consequently, we ordered another line (WiscDsLox504A08, NASC ID N859103) from a separate collection, in which the T-DNA was predicted to insert into At4g04920. This line also displayed the visible sfr6 phenotype (Figure 2).

Figure 2.

 Twelve-day-old Columbia wild-type (WT) and sfr6 mutants.
(a) sfr6 mutants show paler colouring.
(b) Cotyledons (indicated by arrows) are larger in sfr6 mutant alleles than WT. sfr6-1 is the original EMS mutant, sfr6-2 is line SALK_048091, and sfr6-3 is line WiscDsLox504A08.

In light of these results, we decided to sequence the At4g04920 genomic locus from the original sfr6 EMS mutant, to search for mutations. Using the predicted At4g04920 coding region pertaining to the genomic DNA in the TAIR database (, we designed DNA primers to amplify overlapping sections of genomic DNA (each comprising approximately 1000 bp). We amplified genomic DNA from both Columbia (Col-0) and the sfr6 mutant, and, using the same primers, sequenced each of these sections. Two independent sets of PCR products were produced for each genotype, and each were sequenced twice. The primers used are listed in Table S1. Comparison of the two sets of DNA samples revealed a point mutation roughly a third of the way into the sequence of the SFR6 gene isolated from the original sfr6 mutant (Figure 3). This was a G→A mutation, causing amendment of the tryptophan-encoding codon UGG to a UGA stop codon, resulting in a truncated predicted protein (Figure 3 and Figure S1). We sequenced the point of insertions in the two T-DNA lines (Figure S2), and the positions are also marked on Figure 3 and Figure S1. We named the two insertion alleles sfr6-2 and sfr6-3 (SALK and Wisconsin lines, respectively), and the original allele sfr6-1 (Figure 3).

Figure 3.

 Map of At4g04920 showing the T-DNA insertion sites and the original EMS point mutation.
(a) Representation of the At4g04920 genomic coding sequence in which black blocks represent exons and thin lines represent introns. The sfr6-1 EMS mutation is caused by a premature stop codon in exon VIII. The insertion sites of T-DNA causing the mutations in sfr6-2 and sfr6-3 are shown.
(b) Predicted protein sequence with sites of T-DNA insertion and the premature stop codon (sfr6-1) marked.

Characterization of two further mutant alleles confirms the locus

Homozygous sfr6-2 and sfr6-3 plants (genotyping performed by PCR, results are shown in Figure S3) displayed the visible phenotype associated with sfr6-1 (Figure 2). Allelism tests with sfr6-1 that involved examination of the visible phenotype of segregating F2 populations of crosses indicated that sfr6-1, sfr6-2 and sfr6-3 are allelic. Analysis of the visible phenotype of progeny from crosses between sfr6-2 and sfr6-3 and Col-0 wild-type confirmed these were both recessive, as is the sfr6-1 mutation (Warren et al., 1996).

We have shown previously that expression of COR genes such as KIN1/2 in response to low temperature is markedly reduced in sfr6-1 (Knight et al., 1999). Here we compared COR gene expression in sfr6-2 and sfr6-3 with expression in sfr6-1 and Col-0 wild-type. Seven-day-old seedlings were treated on agar plates at 4°C for 6 h before harvesting. Replicate samples of Col-0 wild-type and sfr6-1 were harvested as controls. Three independent homozygous segregating lines were chosen for each of the T-DNA mutants. Figure 4 shows an RNA gel blot of samples derived from the plants described above, and probed for expression of the COR genes KIN1/KIN2, LTI78 and COR15a. Expression of β-TUBULIN was used as a control for the amount of RNA loaded on the gel. sfr6-2 and sfr6-3 showed reduced levels of KIN1/2, LTI78 and COR15a expression in response to cold compared with Col-0; the degree of expression was similar to that seen in the sfr6-1 mutant (Figure 4). Although sfr6-2 might have been expected to be a stronger mutant allele than sfr6-1 (the insertion being closer to the 5′ end of the gene), these data indicate that its effect on COR gene expression was no more severe than that of the original mutation.

Figure 4.

 RNA gel blot of COR gene expression in three sfr6 mutants.
RNA was extracted from 7-day-old seedlings subjected to 5°C for 6 h. The blot was probed with 32P-dCTP-labelled probes for KIN1/2, LTI78, COR15A and β-TUBULIN (loading control). Two biological replicates samples are shown for Col-0 WT and sfr6-1. Three independent homozygous lines are shown for the newly isolated sfr6-2 and sfr6-3 alleles.

Finally, as COR gene expression was reduced in the sfr6-2 and sfr6-3 mutants to a level similar to that seen in sfr6-1, we assessed the effect of the two new mutations on freezing tolerance. Five-week-old plants grown under short-day conditions were acclimated for 11 days at 4°C. After this, they were transferred to a freezer and maintained at −7°C for 16 h. After this time, the plants were returned to ambient temperature and the development of freeze–thaw-induced damage was monitored. Figure 5 shows that, after 5 days, all three mutants showed severe damage following freezing and thawing, but Col-0 showed no visible signs of damage. Together, these COR gene expression and freezing assay data confirm that SFR6 is At4g04920.

Figure 5.

 Freezing tolerance of sfr6 alleles.
The three sfr6 mutants and Col-0 wild-type (WT) controls were grown on soil for 5 weeks at 20°C before transfer to 4°C for 11 days of cold acclimation. Plants were then frozen at −7°C for 16 h, and returned to pre-acclimation conditions to allow recovery to be assessed.
(a) Cold-acclimated plants immediately before freeze testing.
(b) The same plants 5 days after freezing.

Expression of SFR6 transcripts

In order to learn more about the possible role of SFR6 in low temperature, we examined the expression of SFR6 in wild-type Arabidopsis responding to cold. Figure 6 shows real-time PCR data for samples derived from seedlings treated for 3, 6 or 24 h at 4°C (all in constant light). SFR6 transcript levels changed little throughout the course of the cold treatment, indicating that any regulation of SFR6 activity that may occur in response to cold is post-transcriptional (KIN2 was induced over 30-fold over the same time course; Figure 6). Data from the Genevestigator database ( support this observation and indicate that SFR6 transcript levels remain stable in response to cold for over 24 h, whereas levels of CBF transcripts and the COR gene transcripts KIN1, COR15A and LTI78 are greatly increased (Figure S4).

Figure 6.

 Real-time PCR of SFR6 transcripts in cold-treated tissue.
Real-time PCR was performed on cDNA samples originating from 7-day-old seedlings subjected to 5°C for 0, 3, 6 and 24 h. The values shown are relative quantification (RQ) values for SFR6 (grey bars) and KIN2 (open bars). Each value is the mean of four separate quantitative RT-PCRs normalized to contemporaneous β-TUBULIN4 expression. These results are for one of two biological replicates showing similar results. Expression was calculated using the comparative Ct method which calculates an RQ value and an estimate of statistical variability (SV) for each sample ( The error bars represent RQMIN and RQMAX, and constitute the acceptable error for a 95% confidence limit according to Student’s t test.

SFR6 protein sequence

In order to determine the predicted SFR6 protein sequence, cDNA was synthesized from cold-treated wild-type Col-0 Arabidopsis tissue, and the full-length SFR6 cDNA was cloned by PCR amplification. The SFR6 cDNA was cloned on two separate occasions involving separate PCR reactions. The coding sequence was determined by sequencing using a combination of primers used for genomic sequencing and some purpose-designed primers (Table S1). The predicted protein is 1268 amino acids long (Figure 3) with a predicted molecular mass of 137 kDa. SFR6 is a unique gene in the Arabidopsis genome. Close orthologues are present in a number of higher-plant species, including brassica, potato, wheat and rice (which is freezing-sensitive). The predicted protein sequences deduced from potato and wheat ESTs show a high degree of conservation (61 and 71% protein sequence identity over the sequence that is available, respectively). Searches for similar proteins revealed no matches outside the plant kingdom. We attempted to use bioinformatic tools (data not shown) to elucidate the possible function of the predicted SFR6 protein. However, these approaches yielded no information on putative function, and no known protein domains are predicted. In the absence of any known protein domains that might have indicated the possible function of SFR6, we continued to examine its mode of action by studying its effect in planta.

SFR6 acts on the CBF pathway downstream of CBF translation

Our previous work showed that mutation of SFR6 affects the accumulation of COR gene transcripts in response to cold (as demonstrated in Figure 4), but that CBF transcripts are induced to wild-type levels (Knight et al., 1999). Thus, it is possible that CBF transcription is not efficiently coupled to downstream COR gene expression in sfr6 mutants. We tested, therefore, whether over-expressing CBF transcripts to much higher than wild-type levels could overcome this limitation and increase COR gene expression in the mutant. sfr6-1 plants were crossed with wild-type lines expressing CBF1 or CBF2 under the control of the 35S constitutive promoter (Gilmour et al., 2004) (a kind gift from Sarah Gilmour and Michael Thomashow, Michigan State University). These over-expressing lines have been shown previously to exhibit high levels of COR gene expression in the absence of low temperature (Jaglo-Ottosen et al., 1998; Gilmour et al., 2004). Homozygous sfr6-1 segregants over-expressing CBF1 or CBF2 were isolated from the F3 generation and subjected to analysis of COR gene expression. The RNA gel blot in Figure 7 shows that, as expected, high levels of CBF transcript were detectable in both wild-type and sfr6 lines harbouring the constructs, but not in untransformed plants of either genotype. However, whilst CBF over-expression in wild-type resulted in high levels of KIN1/2 and LTI78 expression (as reported previously), no such effect was seen in sfr6-1 (Figure 7). Neither CBF1 nor CBF2 over-expression caused elevated expression of target COR genes in the mutant, and their transcript levels remained almost undetectable, at levels similar to those seen in non-transformed wild-type. The fact that an over-abundance of CBF transcripts does not restore normal COR gene expression in sfr6 indicates that SFR6 acts downstream of CBF transcripts.

Figure 7.

 RNA gel blot showing the effect of CBF over-expression in sfr6 and wild-type.
RNA was extracted from 10-day-old wild-type (WT) and sfr6 seedlings expressing 35S::CBF1 or 35S::CBF2 and untransformed WT and sfr6 controls (three biological replicate samples). The blot was probed with 32P-dCTP-labelled probes for CBF1/2, KIN1/2 and β-TUBULIN (loading control).

The results above led us to consider the possibility that expression of CBF protein is defective in sfr6 mutants, e.g. that CBF translation may be inefficient or the CBF protein may be unstable. To test this possibility directly, we sought to measure CBF protein levels in sfr6 and wild-type plants responding to low temperature. Several groups have attempted to raise antibodies to the CBF proteins but with little success (M.F. Thomashow, personal communication), and this may be a consequence of high turnover and low abundance of these TFs. We therefore adopted an alternative approach. Wild-type plants expressing a CBF–aequorin conjugate under the control of the native CBF1 promoter (Knight et al., 2004) were crossed with sfr6-1 plants, and the protein levels were assayed in segregating lines. Aequorin is a sensitive luminescent reporter (Knight et al., 1991; Knight and Knight, 1995) that can be used to detect low levels of protein in planta. We have used this system in the past to report CBF protein levels, and have shown that the luminescent fusion protein accumulates after increasing time in the cold (Knight et al., 2004). The level of CBF–aequorin fusion protein was measured by recording luminescence emitted in extracts from cold-treated seedlings. We compared aequorin levels in segregating sfr6-1 and wild-type lines from the F2 generation of three independent crosses, and observed that luminescence (and thus protein levels) was not reduced in sfr6 segregants, but was in fact slightly higher in sfr6 (Figure 8). This result shows that CBF protein translation and stability are unlikely to be altered in sfr6 mutants, suggesting that SFR6 acts in the CBF pathway after CBF translation.

Figure 8.

 Expression of CBF–aequorin in wild-type and sfr6.
The chart shows the mean number of luminescence counts recorded by luminometry of discharged aequorin over a 10 sec count period in samples originating from wild-type (WT) and sfr6-1 plants expressing CBF–aequorin fusion protein under the control of the CBF1 promoter. Twelve-day-old plants were subjected to treatment at 5°C for 48 h before harvesting. The data shown are means of six replicate samples taken from a total of three independent crosses between sfr6-1 and WT plants expressing the construct.

Targeting of SFR6

As bioinformatic investigation of the SFR6 protein sequence yielded no information as to its function, we determined its subcellular localization empirically to provide an insight into its function. Preliminary data from fusions of SFR6 to GFP expressed transiently in leek (Allium porrum) cells using particle bombardment showed clear nuclear localization, but GUS fused to GFP was located in the cytosol (Figure S5). Confocal microscopy of stably transformed Arabidopsis roots showed that N-terminal GFP fusions were localized similarly to those in leek cells. GFP–GUS showed an expression pattern indicative of a cytosolic localization (Figure 9a,b), but the GFP–SFR6 fusion was localized predominantly in the nucleus (Figure 9c,d), with the ring shaped appearance of the nuclear fluorescence suggesting that the protein is borne on the nuclear envelope. These data indicate that SFR6 is localized in the nucleus.

Figure 9.

 Localization of SFR6 by GFP.
Confocal microscopy of N-terminal GFP-tagged GUS control (a,b) and N-terminal GFP-tagged SFR6 (c,d) in the root tip. Images were captured using a 40 × objective. One optical section (a,c) and the complete z-stack (b, d) are shown for each construct.


Cloning SFR6

Since first identifying the effect of the sfr6 mutation on freezing tolerance and COR gene expression (Warren et al., 1996; Knight et al., 1999), progress towards mapping and cloning the gene has been slow. Initial mapping by classical approaches indicated that the gene locus was close to the centromere of chromosome 4 (Figure 1), a region in which recombination is very low (Schmidt et al., 1995; Drouaud et al., 2006). In accordance with this, when we genotyped 1500 mapping lines generated from an Ler/sfr6 cross, we did not detect a single cross-over event between flanking markers that spanned a region of 2.6 Mbp. We concluded that little further progress would be achieved by isolation of additional mapping lines, so we changed our approach to one that is not reliant on natural recombination rates.

The interval was too large for us to attempt gene identification by complementation; therefore an alternative approach was required. Although gene density in this region is low (approximately 1 gene per 12 kb), and approximately 33% of the genes present are highly conserved copies of transposable elements (Wright et al., 2003) that are unlikely to provide a unique host function and thus unlikely to be SFR6, approximately 400 genes remained as possible candidates.

We chose, therefore, to attempt to re-isolate an sfr6 mutant allele from available T-DNA mutant collections. There are a number of Arabidopsis T-DNA insertion collections and databases that may be searched for loss-of-function mutations (reviewed by Alonso and Ecker, 2006). Consulting the records for individual genes in the TAIR database ( allowed us to identify 429 lines that carried T-DNA insertions within genes present in the interval of interest. The phenotype that we screened for comprised larger and paler cotyledons and paler true leaves in seedlings. Work with many generations of sfr6 crosses has indicated strongly that this visible phenotype is indeed linked to the sfr6 genotype. This approach led to re-isolation of the SFR6 locus, and provided two additional alleles exhibiting the phenotype of reduced COR gene expression and loss of freezing tolerance with which we confirmed that we had cloned SFR6. Despite the time-consuming growth and observation of hundreds of individual T-DNA lines, the approach led us to gene identification significantly more quickly than the alternatives would have done; complementation of this large region would have required a great number of plant transformations and further analysis. Our approach may be worthy of consideration in any situation in which an EMS mutation of particular interest proves difficult to map by conventional means, and where the mutant phenotype is readily apparent upon visual examination or very simple to score.

The SFR6 gene and predicted protein

Identification of SFR6 as At4g04920, and cloning of the coding sequence from cDNA, revealed that it is a large gene comprising 16 exons with a predicted coding region of 3807 bp. Results from our own experiments and examination of the Genevestigator database suggest that SFR6 transcript levels alter little in response to cold. This suggests that control of SFR6 activity in response to cold does not occur by changes in transcript levels, but must occur at the protein level. Given that SFR6 appears to affect developmental processes such as chlorophyll biosynthesis/degradation (indicated by the yellow colour of the mutant), flowering (Knight et al., 2008) and morphology (as described in Results), it may also regulate groups of genes other than the CBF regulon.

To our surprise, we were unable to identify any predicted domains in the 1269 amino acid protein sequence, despite searching a number of protein prediction databases. All three mutant alleles are the result of interruptions in the coding sequence: the original allele sfr6-1 is a point mutation resulting in a premature stop codon, sfr6-2 is an insertion into the 4th intron, and sfr6-3 is an insertion into the 8th exon. All of these mutations occur in approximately the first third of the protein, with the EMS point mutation occurring at 1452 bp (484 amino acids) and sfr6-3 very close to this. It is therefore clear that the first 485 amino acids are insufficient for SFR6 function, but the role of the C-terminal part (and majority) of the protein (the remaining 785 amino acids) cannot yet be deduced. It is possible that the nuclear localization signal (NLS) occurs at the C-terminal end of the protein, and that its absence in the three mutant versions is responsible for loss of function. This may be addressed empirically in the future.

As a protein that influences the transcription of a vast number of genes (Boyce et al., 2003), we might expect SFR6 to act in the nucleus; this hypothesis is supported by our data showing that SFR6–GFP protein fusions are localized in the nucleus. One question to address in the future is whether the protein is constitutively translocated to the nucleus, or whether this occurs only under particular stress conditions. Nuclear localization in unstressed cells, together with the fact that the sfr6 mutation affects resting levels of COR gene transcripts under non-inducing conditions (Knight et al., 1999), suggest that this is not the case, and that SFR6 is constitutively active.

Where does SFR6 act in the CBF pathway?

Our previous work indicated that COR genes under the control of the CBF TFs were mis-regulated in sfr6 (Knight et al., 1999), but that CBF transcripts themselves were expressed normally. This indicated that SFR6 acts either downstream of CBF transcription (possibly in conjunction with the CBF protein) or independently of the CBF pathway. Several pieces of evidence have led us to the conclusion that the latter is not the case. Firstly, our work with a luciferase reporter indicated that SFR6 affects transcription that is specifically reliant upon the CRT/DRE motif (Boyce et al., 2003), the specific cis-element target of the CBF TFs. Secondly, our microarray analysis of genes affected in sfr6 under cold stress (NASC microarray service array number 73; showed a strong correlation between COR genes that were mis-regulated in sfr6 and genes containing the CRT/DRE element. The data we present here show that over-expression of CBF1 or CBF2 in sfr6 does not ameliorate this defect in COR gene expression, suggesting that SFR6 acts after CBF transcription in the pathway leading to COR gene expression and freezing tolerance. These data confirm our earlier deduction that SFR6 does not act on a pathway separate from the CBF pathway, as, were this the case, over-expression of CBFs would be expected restore the COR gene defect. Therefore these data, together with our previous results, confirm that SFR6 acts on the CBF-controlled pathway to regulate COR gene expression. Future work will test the hypothesis that SFR6 and CBF proteins interact directly or indirectly to facilitate expression of COR genes.

We have used CBF–aequorin fusion proteins previously to reveal changes in CBF levels in Arabidopsis (Knight et al., 2004), in the absence of a reliable antibody to CBF. In this study, we used detection of CBF–aequorin luminescence as a measure of the relative levels of CBF protein in sfr6 and wild-type plants. Our observation that CBF–aequorin fusion protein accumulates in sfr6 to at least wild-type levels in response to cold indicates that SFR6 acts on the pathway after CBF translation, i.e. it affects CBF activity. This is consistent with the finding that over-expression of CBF in an sfr6 mutant background does not lead to increased COR gene expression. It must be borne in mind that accumulation of the CBF–aequorin fusion protein may not exactly mirror accumulation of the native CBF protein, but any stabilizing or destabilizing effect of the C-terminal aequorin would be expected to occur in both wild-type and sfr6; therefore we believe that the data do indicate that levels of CBF protein are equivalent in both.

Concluding remarks

SFR6 may act post-translationally on the CBFs to modulate their activity by any of a number of means. The activity of proteins, especially TFs, may be regulated by processes such as ubiquitination or sumoylation (Downes and Vierstra, 2005). Polypeptides are added to proteins as post-translational modifiers and modulate the target protein’s activity, thus controlling processes that include developmental and stress responses in plants. Sumoylation involves SUMO (small ubiquitin modifier peptides) becoming attached to other proteins to affect the functions, interactions and/or stability of the modified targets. For example, SIZ1 facilitates the sumoylation of inducer of CBF expression 1 (ICE1), the TF that controls CBF3 expression (Miura et al., 2007). ICE1 is also subject to control by ubiquitination-mediated proteosomal protein degradation (Zhu et al., 2007), evidence that these processes play a role in cold signal transduction. However, to date, no protein has been identified that acts as a post-translational modifier of the CBFs. SFR6 may play such a role, and our future work will be aimed at elucidating its mode of action in the CBF pathway. Alternatively, SFR6 may be instrumental in recruiting CBF protein into the nucleus to facilitate its action. Given the lack of indications from the SFR6 protein sequence as to possible activities, identifying the function of SFR6 is likely to reveal new mechanisms involved in the control of TF activity.

Experimental procedures

Plant materials and treatments

Columbia wild-type (Col-0) and sfr6 mutant allele plants were grown on 1 × Murashige and Skoog/0.8% agar plates as described previously (Knight et al., 1999). Lighting was maintained at 150–200 μE m−2 sec−1 for all experiments, and a 16 h photoperiod with a temperature of 20 ± 1°C was used for growth of plants for all experiments apart from freeze testing. For freeze testing, plants were grown on soil with a 9 h photoperiod for 5 weeks at 20°C, before cold acclimation at 4°C for 11 days under the same daylength conditions. Plants were subsequently placed in a freezer with a minimum air temperature of −7°C for 16 h before returning them to pre-acclimation growth conditions to allow recovery and observation of freezing damage. For gene expression and protein expression measurements, plants grown on plates were transferred to 5°C for the appropriate amount of time in constant light. Seedlings were frozen in liquid nitrogen in Eppendorf tubes immediately after collection.

Mapping and sequencing

sfr6 was crossed with an Ler wild-type plant, and 1500 F2 progeny were analysed. Homozygous sfr6 plants were identified by their yellowish colouring. DNA was extracted and recombinants selected using CAPS and SSLP markers as described previously (Thorlby et al., 1999) and 11 new CAPS and SSLP markers designed using SNP and INDEL information available in the Monsanto CEREON database ( Details of the new markers are given in Table S1. T-DNA-tagged lines were obtained from the Nottingham Arabidopsis Stock Centre (NASC; and from the Arabidopsis Biological Resource Center at Ohio State University ( These were scored by growing on agar plates as above for 10–14 days and selecting any showing visible yellowing of the cotyledons and first true leaves comparable to that seen in control sfr6 plants grown alongside. Any plant line exhibiting a physical appearance similar to sfr6 was grown to maturity and the progeny re-examined.

The SFR6 gene was amplified from genomic DNA by PCR using PYROBEST Taq polymerase (Takara; and primers that are detailed in Table S1. All sequencing was performed by DBS Genomics (the in-house sequencing service provided in the School of Biological Sciences, Durham University) using an Applied Biosystems 3730 DNA analyser (

Measurement of gene expression

Gene expression was monitored by RNA gel-blot analysis exactly as reported previously for CBF (Knight et al., 2004), KIN1/2, LTI78, COR15A and β-TUBULIN (loading control) (Knight et al., 1999). The CBF probe corresponded to the entire CBF1 coding sequence, and hybridized to both CBF1 and CBF2 transcripts. COR gene expression was measured in the three mutant alleles in 7-day-old plants treated at 5°C for 6 h.

Real-time PCR was used to measure SFR6 gene expression and KIN2 expression in the same samples, using an Applied Biosystems 7300 machine. A high-capacity cDNA reverse transcription kit (Applied Biosystems) was used to reverse-transcribe cDNA from 1.5 μg total RNA extracted using a Qiagen RNeasy plant mini kit ( in conjunction with Qiagen RNase-free DNase to remove any genomic contamination. A 1:50 dilution of cDNA (10 μl aliquot) was used with TaqMan Universal PCR mix (Applied Biosystems) in a 25 μl reaction in an optical 96-well plate, and four technical replicates were used for each sample. Taqman probes for SFR6 (probe identifier At02209654_g1), KIN2 (At02354775_s1) and β-TUBULIN4 (endogenous control for normalization; At02337699_g1) were used. Two biological repeats were carried out. Relative quantification was performed by the ΔΔCt method (Livak and Schmittgen, 2001).

Measurement of CBF–aequorin fusion protein levels

Levels of CBF–aequorin fusion protein were detected as described previously (Knight et al., 2004). Five 8-day-old seedlings per replicate were harvested after cold or ambient treatment, with six wild-type and six sfr6 replicates from F2 lines originating from three independent crosses. Plant homogenates were centrifuged at 13 000 g for 10 min. The aequorin co-factor, coelenterazine (Lux Biotechnology;, was added to the supernatants at a concentration of 1 μm, and the mixture was incubated at room temperature for 2 h. The amount of reconstituted aequorin protein was measured in a luminometer immediately after the addition of 25 mm calcium chloride, which discharges all available aequorin. Four biological repeats were performed.

Localization of SFR6 by fusion to GFP

The full-length SFR6 coding sequence including introns was amplified from Columbia (Col-0) genomic DNA using oligonucleotide primers 5′-TTCAGTCGACATGAATCAGCAAAACCCAGAAGAAG-3′ (forward) and 5′-CCGCGAATTCACAACACGGACCCACGTTCCACCAC-3′ (reverse) with PYROBEST high-fidelity DNA polymerase. The amplified sequence was inserted into the Gateway entry vector pENTR1A (Invitrogen, using the SalI and EcoRI restriction sites, and then recombined into the binary Gateway destination vectors Pk7FWG2 and pK7WGF2 (Karimi et al., 2002) to produce N- or C-terminal GFP fusions under the control of the 35S constitutive promoter. Control plasmids were produced in parallel using the same vectors but with the SFR6 coding sequence substituted by a GUS coding sequence derived from the GUS entry vector pENTR-GUS (Invitrogen,

Plasmid DNA was purified from Eschrichia coli for use in biolistic bombardment experiments using a Qiagen maxi kit. Biolistic bombardment experiments were performed according to the manufacturer’s instructions (Bio-Rad, Briefly, gold particles were coated with DNA and applied to macrocarrier discs. Colourless leaves from fresh leek (Allium porrum) were cut into 5 × 5 cm squares and bombarded with DNA prepared as above, using Bio-Rad biolistic particle delivery system model PDS-1000/He with a 1100 psi rupture disc. Bombarded tissue explants were placed in Petri dishes with added water to maintain high humidity, and were left for 48 h in a Sanyo MLR-350 growth chamber (Sanyo Biomedical, (16 h/8 h light/dark cycle at 20°C) before visualization using a Nikon Ecl IPSE TE300 fluorescence microscope ( with excitation at 490 nm and emission at 505–530 nm wavelength. Images were captured digitally using openlab image capture software (Bucher Biotec AG;

Arabidopsis Col-0 was transformed with GFP fusion constructs using the floral dip method with the Agrobacterium C58C1 strain (Holsters et al., 1978; Clough and Bent, 1998), and transformants were selected for using kanamycin. GFP was visualized in root tips using a Zeiss LSM50 confocal laser scanning microscope with excitation at 488 nm using an argon laser and emission at 505–530 nm. Images were captured and processed using the integral LSM software.


This work was completed with the help of funding from the UK Biotechnology and Biological Sciences Research Council (BBSRC) (grants P20470, P18613 and BB/D009162/1). We would like to thank NASC for provision of the numerous T-DNA tagged lines, and Pauline White (Department of Plant Sciences, University of Oxford, UK) for her patient sowing of the seeds. We are grateful to Margaret Pullen, Mike Deeks and Tim Hawkins for advice and help with fluorescence microscopy, and to the sequencing team at DBS Genomics. Finally, we would like to acknowledge our gratitude to our late colleague and collaborator Gary Warren, who isolated the original sfr6 mutant and whose intellectual contribution to the pursuit of sfr6 was invaluable.