A C-Repeat Binding Factor Transcriptional Activator (CBF/DREB1) from European Bilberry (Vaccinium myrtillus) Induces Freezing Tolerance When Expressed in Arabidopsis thaliana

Freezing stress affects all plants from temperate zones to the poles. Global climate change means such freezing events are becoming less predictable. This in turn reduces the ability of plants to predict the approaching low temperatures and cold acclimate. This has consequences for crop yields and distribution of wild plant species. C-repeat binding factors (CBFs) are transcription factors previously shown to play a vital role in the acclimation process of Arabidopsis thaliana, controlling the expression of hundreds of genes whose products are necessary for freezing tolerance. Work in other plant species cements CBFs as key determinants in the trait of freezing tolerance in higher plants. To test the function of CBFs from highly freezing tolerant plants species we cloned and sequenced CBF transcription factors from three Vaccinium species (Vaccinium myrtillus, Vaccinium uliginosum and Vaccinium vitis-idaea) which we collected in the Arctic. We tested the activity of CBF transcription factors from the three Vaccinium species by producing transgenic Arabidopsis lines overexpressing them. Only the Vaccinium myrtillus CBF was able to substantially activate COR (CBF-target) gene expression in the absence of cold. Correspondingly, only the lines expressing the Vaccinium myrtillus CBF were constitutively freezing tolerant. The basis for the differences in potency of the three Vaccinium CBFs was tested by observing cellular localisation and protein levels. All three CBFs were correctly targeted to the nucleus, but Vaccinium uliginosum CBF appeared to be relatively unstable. The reasons for lack of potency for Vaccinium vitis-idaea CBF were not due to stability or targeting, and we speculate that this was due to altered transcription factor function.


Introduction
Freezing stress affects all plants from temperate zones to the poles. The incidence of freezing stress is largely predictable: generally ensuing at specific times of the year. However, the strength of this prediction is now being eroded by climate change, which is leading to more chaotic patterns of climate, including unexpected cold episodes e.g. during the summer [1], as well as extreme winter temperature events due to altered patterns of landscape snow cover [2]. This is becoming a serious agricultural problem e.g. in recent years there have been a number of examples worldwide of crops being lost to freezing temperatures. Multiple states in the USA lost fruit crops in April 2012 [3]. In 2011 Mexico lost 90% of its corn crops to frost [4], which in turn caused food prices in that country to triple that year. In the same year India had its coldest winter in 30 years [5] and Zimbabwe (a major exporter of flowers) lost a large number of flower crops to frost [6]. Such problems arise because levels of freezing tolerance of plant/ crop species are not fixed, but require an active process known as cold acclimation, which needs to be timed to precede freezing events.
Temperate and Arctic plants usually survive winter by cold acclimating, triggered by temperatures slowly becoming lower and the day length decreasing. This combination of light and temperature acts as a signal to the plant that freezing events are likely in the near future. Depending on the plant species, the acclimation process can take from a few days to several weeks to complete. Acclimation involves metabolic and cellular changes to allow survival of freezing temperatures, including the production of compatible solutes and reconfiguration of membrane components [7]. The majority of these responses are mediated by changes in gene expression during the acclimation process.
In the model plant system Arabidopsis, key players in this molecular response are the CBF/DREB1 (C-repeat Binding Factor/Drought Response Element Binding Factor) transcription factors [8]. These transcription factors are induced in expression in response to low, non-freezing temperatures. These transcription factors in turn activate the expression of hundreds of genes (COR (cold-regulated) genes) whose products effect the changes needed for cold acclimation. In most species tested, CBFs occur as multigene families. In the case of Arabidopsis, the family of CBFs which regulates cold gene expression total three members [9]. Reports are mixed in terms of equivalence of activity, with most studies concluding they have equal function e.g. [10] although there is some evidence in Arabidopsis of a hierarchy of regulation, whereby CBF2 acts as a negative regulator of CBF1 and CBF3 [11,12]. Studies of other species of plants strongly support the importance of CBF/DREB1 transcription factors in cold acclimation of many species, including the genus Vaccinium [4,13]. The importance of CBF/DREB1 to freezing tolerance can be demonstrated by overexpressing these transcription factors, which leads to constitutive expression of COR genes and constitutive freezing tolerance, in the total absence of physiological cold acclimation. Recently, overexpression of Vaccinium corymbosum CBF in Arabidopsis was demonstrated to lead to induced COR expression, and constitutive freezing tolerance [13], indicating a functional CBF system in this closely related Vaccinium species. Additionally, overexpression of CBF from Vaccinium corymbosum in transgenic native background (same species) was demonstrated to lead to increased freezing tolerance [14]. This study demonstrates that CBF genes are indeed functional in a Vaccinium background in achieving freezing tolerance. This has also been demonstrated for other plant species e.g. Brassica napus [4]. Natural variation in CBF function has been postulated to be a substrate for selection by evolution e.g. the Cape Verde accession of Arabidopsis has a greatly reduced CBF function and is found in a tropical climate. Another important aspect of the cold acclimation response is developmental, i.e. there is altered (reduced) growth. This has been shown to be, at least partially, mediated by the DELLA growth-regulation system i.e. DELLA protein levels accumulate at cold temperatures, leading to reduced growth. Interestingly, CBF/DREB1 have also been shown to be important in regulating this process, by suppressing flux through the gibberrellin pathway, which promotes stability of DELLA proteins [15].
We were interested in assessing the potency of CBF/DREB1 transcription factors from three plant species from the genus Vaccinium, namely V. myrtillus, V. uliginosum and V. vitis-idaea that are key components of circumpolar Arctic tundra heath communities [16,17,18]. In such communities, these species have to tolerate significantly low temperatures, often for extended periods of the yearly growth cycle.

Plant Materials and Sampling Locations
V. myrtillus, V. uliginosum and V. vitis-idaea samples were collected in Abisko, Sweden (68u20'N 18u49'E), 400 m above sea level and 50 km from the nearest sea coast [19]. Temperatures fluctuate between 235uC and +18uC over the year.
Seeds for Arabidopsis thaliana accession Columbia (Col-0) were obtained from Lehle seeds (Round Rock, Texas, USA). Seeds for Nicotiana benthamiana were a kind gift from Dr Tim Hawkins (Durham University, UK).

Plant Growth Conditions
Sterilised Arabidopsis seeds were grown on 16 MS medium, 0.8% (w/v) agar plates [20] and grown at a temperature of 20uC and 16/8 h photoperiod with a light intensity of 150 mmol m 22 s 21 in a Percival plant growth cabinet (Model: CU-36L5D, CLF plant climatics, Emersacker, Germany) for 7 days. After this time seedlings were transferred to peat plugs (Jiffy products international AS, Norway) and grown in a bespoke walkin growth chamber with approximately the same conditions. Nicotiana benthamiana was grown in a similar bespoke walk-in growth chamber at a daytime temperature of 24uC and a night Genomic DNA Extraction, Sequencing and Analysis DNA was extracted from Vaccinium leaf tissue using the CTAB DNA extraction method [21] with an added phenol-chloroform clean-up stage: this involved resuspending in phenol chloroform (1:1 DNA aqueous solution to phenol chloroform (v/v)) followed by centrifugation, removal of the aqueous top layer and repeating until no more precipitate was formed at the interface between the aqueous top layer and phenol chloroform bottom layer. DNA was precipitated by adding one tenth of the volume of the aqueous layer of 3 M sodium acetate (pH 4.5) and two volumes of 100% ethanol at 280uC. The pelleted DNA was washed in 1 ml of 80% ethanol and resuspended in TE buffer. CBF coding regions were amplified by PCR using RedTaq (Bioline Reagents Ltd, London, UK) and the CBF-VI primers listed below. PCR products were separated by TBE-agarose gel electrophoresis, cut-out of the gel and purified using a Qiaquick gel extraction kit (Qiagen Ltd. Crawley, UK). Sequencing was performed by the Durham University sequencing service. Sequences were aligned using Clustal X and Jalview [22,23]. Sequences for the Vaccinium myrtillus, Vaccinium uliginosum and Vaccinium vitis-idaea DREB1 cDNA clones were submitted to GenBank (accession numbers, JN254610, JN866911 and JN866912, respectively).

Vaccinium RNA Extraction
Leaf tissue from the three Vaccinium species was cold treated on ice for two hours with an ambient room temperature control. Total RNA was extracted using the method described in Jaakola et al., 2001 [24].

Production of Plant Transformation Constructs
For overexpression of Vaccinium CBFs (vCBFs) in Arabidopsis, CBF coding regions were amplified from Vaccinium genomic DNA using the following primers: CBF-VI-R: 59-ATCTAACTCCA-CAAGAGACC-39 and CBF-VI-F: 59-CACCGAAGAGTTA-GATGTGCGCAGC-39 which amplify sequences from all three Vaccinium species. Sequences were amplified and sequenced on two separate occasions to ensure that the correct sequence was obtained without PCR errors. The resulting PCR products were cloned into pENTR-D-TOPO (Invitrogen) and then into the PB7WG2 destination binary vector [25] using a Gateway LR Clonase II kit (Invitrogen). This produced constructs where each of the three Vaccinium CBF coding regions were inserted after the 35S promoter for overexpression. Each CBF coding region started at the start codon and ended at the stop codon, so the constructs did not have any native CBF 59 or 39 untranslated sequences. In this way the constructs were only different to each other in the CBF coding region itself, allowing for proper comparison between the activity of each CBF protein when expressed. For production of GFP-CBF fusions, vCBF coding regions were amplified using the following primers: Myrtillus CBF-F: 59-CACCATGGAA-TATTACTCAAGTCC-39 or Uli-vitis-CBF-F: 59-CACCATG-GAATATAACTCTAGTCC-39 as forward primers and CBF-VI-R: 59-ATCTAACTCCACAAGAGACC-39 as reverse primer. PCR products were cloned into pENTR-D-TOPO (Invitrogen, Life technologies Ltd. Paisley UK) then into the pK7WGF2 destination binary vector [25] using LR Clonase and a DH5a E. coli strain. After completion of binary constructs, the constructs were sequenced to confirm there were no errors and, where relevant, that the CBF coding regions were in frame with GFP. The binary vectors also conferred resistance to BASTA to allow for selection of transgenic lines (see below). For transactivation studies in Nicotiana benthamiana, a binary construct to express firefly luciferase driven by 4 copies of the CRT/DRE promoter motif [26] was used in conjunction with the GFP constructs.

Plant Transformation
For stable transformation of Arabidopsis, Arabidopsis Col-0 was transformed by Agrobacterium-mediated floral dip [27] using Agrobacterium tumefaciens strain C58C1. Transformants were grown on a mixture of sand and Perlite and selected by spraying the seedlings from 7 days onwards with 150 mM BASTA (herbicide) at 5 day intervals. Individual plants were bulked up and segregating M1 seed used to grow seedlings for gene expression experiments, and M2 homozygotes selected for freezing and growth assays. For transient expression of Nicotiana benthamiana, Agrobacterium tumefaciens infiltration method was used [28]. Bacteria were resuspended in 10 mM MgCl 2 and infiltrated at an OD of 0.3. Plants were grown for a further for 2 days before imaging.
Fluorescence and Luminescence Imaging GFP fluorescence was imaged using a Zeiss Meta 510 CLSM (confocal laser scanning microscope) and software at a magnification of x40 (oil immersion lens). Luciferase imaging was performed after spraying leaves with 5 mM potassium luciferin in 0.01% Triton X-100 (v/v) using a Photek photon-counting camera (Photek, Hastings, UK) [29].

Western Blot Analysis
Total protein was extracted from Nicotiana benthamiana leaf disks (as described in [30]) using a BioRad D c Protein assay. 50 ng of the extracted protein was boiled in SDS loading buffer and run on a 15% polyacrylamide gel for ,3 hours. Proteins were then transferred onto PVDF membrane (BioRad laboratories Ltd. Hertfordshire, UK) using a semi-dry blotting technique. The membrane was then blocked in 5% milk powder in TBST and incubated with a 1 in 200 dilution of aGFP antibody (#G1112 Santa Cruz Biotechnology inc. California, USA) in 10% (w/v) milk powder in TBST overnight at 4uC. Anti-rabbit IgG HRP secondary antibody (#E1012 Santa Cruz Biotechnology inc. California, USA) was incubated for 1 hour at room temperature as a 1 in 5000 dilution in 5% (w/v) milk in TBST. After this, the filter was then washed and incubated with a WestDura Femto ECL kit (Fisher Scientific, Leicestershire, UK) and bands visualised using a photo counting camera (Photek, Hastings, UK).

qPCR Analysis of Gene Expression
Quantitative real-time PCR was use to analyse gene expression levels as described previously [31]. cDNA was synthesised from Arabidopsis RNA using a high capacity cDNA reverse transcription kit (Applied Biosystems, California, USA). cDNA was synthesised from Vaccinium RNA using a QuantiTect reverse transcription kit (Qiagen, Sussex, UK). qPCR was carried out using an AB 7300 real time PCR system (Applied Biosystems California USA) and Go Taq qPCR master mix (Promega, Wisconsin,USA). Primers were designed by eye and synthesised by Invitrogen. Sequences

Freezing Assays
Transgenic Arabidopsis lines overexpressing vCBFs were tested for vCBF and COR gene expression in 4 independent lines for each construct. Two transgenic lines produced from using each of the three different Vaccinium constructs were chosen for freezing assays: one showing relatively high expression, the other relatively low expression, giving 6 lines in total. Seeds from each of the 6 chosen lines and Col-0 wild type control were germinated and grown for 5 weeks (in the same way as described above) then transferred straight to 27uC for 24 h in a Sanyo MIR 254 incubator in darkness. The plants were then left to thaw and returned to their original growth conditions. Before being transferred to the freezing chamber each plant was individually photographed (using a Cannon EOS 550D DSLR on Raw format to produce TIFF images), then photographed again 3 days after freezing. The average diameter of each plant rosette was calculated from 3 measurements between the points of widest diameter using AxioVison software (Carl Zeiss, Cambridge, UK) and then standardised between photos using the scale bar in the corner of each photo (not shown).

F v /F m Measurements
Damage due to stress can be diagnosed as a reduced ratio of variable fluorescence (F v ) over maximal fluorescence (F m ) of plant photosystems [32,33]. To test the effect of freezing on plants lines we therefore determined F v /F m . Before being transferred to the freezing chamber the F v /F m value for each plant was measured using a FluorCam 700 mf (Photon Systems instruments, Brno, Czech Republic) on the F o , F m and Kautsky effect setting. Plants were dark-acclimated for 30 min prior to measurements. Measurements were repeated 3 days after freezing using the same parameters for F o and F m measurement, Kautsky induction, and dark relaxation before and after Kautsky induction.

Results
Protein Sequence Variation in CBF from Three Different Vaccinium Species V. myrtillus, V. uliginosum and V. vitis-idaea are freezing-tolerant in the wild, albeit to different extents [34,35,36]. CBF transcription factors have been shown to be key genetic determinants of freezing tolerance in many plants [3,4,37]. For at least one other species of Vaccinium, Vaccinium corymbosum, it has been demonstrated that a functional CBF is present and capable of promoting freezing tolerance when overexpressed in Arabidopsis [13]. We wished to determine if this was the case for V. myrtillus, V. uliginosum and V. vitis-idaea. We first cloned and sequenced the coding region of a CBF gene from each species, utilising the sequence information already published for V. vitis-idaea [38]. Comparing the predicted protein sequences (Figure 1) to Arabidopsis, the CBF sequences from all three Vaccinium species were closest to the Arabidopsis CBF3 (DREB1A) which has been clearly demonstrated to have a role in cold acclimation [39]. Amino acid sequences in regions deter- mined by previous reports [4,40] to be important for either binding to DNA, or activation of transcription, were well conserved in all three Vaccinium species (Figure 1). Comparing the predicted CBF protein sequences for each of the three Vaccinium species we cloned with the previously reported sequences for V. vitis-idaea [38] and V. corymbosum [13] showed 9-12 amino acid differences between species and 5 between our Northern Swedish V. vitis-idaea CBF sequence as compared to V. vitis-idaea from China [38]. It was also notable that the V. vitis-idaea CBF which we cloned and sequenced, and that from China [38] had an extra 19 amino acid N-terminal extension, not present in either V. myrtillus, V. uliginosum or V. corymbosum. To test the possibility that the 3 CBF genes we cloned were involved in cold responses in the host species, we performed qPCR experiments to measure the expression of all three in response to short (2 h) exposure to cold (2 h cold is reported at the peak for CBF expression in Arabidopsis [9]). As can be seen in Figure S1, all 3 CBF genes appeared to be cold-induced. Without full genome sequence information on the three Vaccinium species, one cannot formally discount that primers used for qPCR might not crosshybridise with other CBF transcripts, however, these data suggest it is possible that all three CBF genes we cloned are cold inducible in their native genetic backgrounds.

Testing the Functionality of CBF from Three Vaccinium Species
Overexpression of functional CBF in Arabidopsis leads to the ectopic induction of COR genes, in the absence of a cold signal [37]. Therefore to test if the CBFs cloned from the three Vaccinium species were all functional, and whether, given the differences in protein sequence described above (Figure 1), there were differences in their activity, we expressed them all in Arabidopsis using genetic transformation. Lines which expressed Vaccinium CBF transcripts to levels detectable above background were identified using real-time PCR. Figure 2A shows typical results with 4 lines obtained from each of the three different Vaccinium constructs. It can be seen in Figure 2A that the levels of CBF transcripts achieved tended, on average, to be lower when the V. myrtillus sequence was expressed compared to the other 2 species (the primers and amplicon for the real time PCR were identical in all cases, so such quantitative comparisons are possible). Testing the expression of the most well-characterised CBF target genes KIN2, LTI78 and GOLS3 ( Figures 2B, 2C and 2D) in the transgenic Arabidopsis lines demonstrated clear overexpression of all three genes in between 2-3 transgenic lines expressing V. myrtillus CBF. However in the other 2 types of transgenic lines (expressing V. uliginosum and V. vitis-idaea CBF) only very weak increases in expression of KIN2 compared to wild type were observed and no significant induction of LTI78 or GOLS3 above wild type levels was observed. As it was reported that V. corymbosum CBF activated expression of KIN2 and LTI78 but was not able to activate expression of COR15A and COR414, we also tested the expression of COR15A and COR414 in our lines overexpressing V. myrtillus, V. uliginosum and V. vitis-idaea CBF. Similarly to KIN2, LTI78 and GOLS3 ( Figures 2B, 2C and 2D), clear overexpression of both COR414 and COR15A was seen in between 2-3 transgenic lines expressing V. myrtillus CBF (Figure 3). In the other 2 types of transgenic lines (expressing V. uliginosum and V. vitis-idaea CBF) there was no significant induction of COR414 ( Figure 3A), but interestingly, COR15A was induced in all lines expressing V. uliginosum CBF, and 2 of the lines expressing V. vitis-idaea CBF ( Figure 3B), relative to wild type.

Testing Freezing Tolerance and Development in Arabidopsis Expressing Vaccinium-derived CBF
It has been demonstrated previously that overexpression of CBF in Arabidopsis leads to constitutive freezing tolerance, in part due to COR gene activation [37] and in part due to developmental growth arrest via DELLA proteins [41]. Given that the data from Figures 2 and 3 indicated the ability of V. myrtillus CBF to strongly activate the expression of Arabidopsis COR genes, we sought to determine whether the lines overexpressing V. myrtillus CBF were more freezing tolerant than lines overexpressing CBF from V. uliginosum and V. vitis-idaea and wild type. Two independent lines for each of the three overexpressing constructs were tested, along with untransformed Arabidopsis wild type. Individual plants were grown for 5 weeks, and then treated to a temperature of 27uC for 24 h before being returned to the original growth conditions. Figure 4 shows representative photographs of individual plants before, and after, freezing. It is clear that the transgenic lines overexpressing either V. uliginosum or V. vitis-idaea CBF were as susceptible to freezing as wild type Arabidopsis (Figure 4

[vi] -4[viii]). However, both independent lines expressing V. myrtillus CBF were robustly freezing tolerant (Figures 4[ix] and 4[x]
). As well as visual inspection, survival of freezing was tested by measuring Fv/Fm as an indicator of photosynthetic capacity ( Figure 5). Imaging of chlorophyll fluorescence for individuals is shown in Figure 5A and average values for F v /F m in Figure 5B. Before freezing, the F v /F m values were very similar for all 7 genotypes. However, after freezing, it was clear that the wild type Arabidopsis and Arabidopsis expressing either V. uliginosum or V. vitis-idaea CBF suffered such severe damage that F v /F m was unmeasurable ( Figure 5B). For the 2 independent lines expressing V. myrtillus CBF, however, whilst there was more variability in the F v /F m of individual plants compared to before freezing (compare size of error bars in Figure 5B), the average values were not significantly different to before freezing ( Figure 5B).
The activation of COR genes is only one aspect of CBF function that is important for freezing tolerance, the other being DELLAmediated grown inhibition [41]. It was clear that whilst there was little difference in overall plant size between wild type controls and lines overexpressing V. uliginosum and V. vitis-idaea CBF ( Figure 4B), the lines overexpressing V. myrtillus showed significantly reduced diameters of rosette, as described previously when Arabidopsis CBF was overexpressed in Arabidopsis [41]. Figures 2-3 that even though the transcripts of V. uliginosum and V. vitis-idaea CBFs were expressed to equal (and in most cases higher) levels that V. myrtillus in transgenic Arabidopsis, target COR genes were not strongly activated by these two transcription factors. This leads to the question: what makes the V. myrtillus CBF more potent that CBFs from V. uliginosum and V. vitis-idaea? Data in Figure 2, however, do not rule out differences in protein expression, or targeting of protein to the nucleus which may account for this phenomenon. In order to test if there were differences in levels of proteins or targeting to the nucleus when expressing the three Vaccinium CBFs we produced GFP tagged versions that could be visualised in cells using fluorescence microscopy, and levels of protein determined by western blot analysis using an antibody to GFP. We tested these parameters by transiently expressing these constructs in Nicotiana benthamiana. Firstly, however, confirmation that the GFP tag did not affect the differential activation of target gene expression by the Vaccinium CBFs was required. We therefore co-expressed each of the Vaccinium CBF-GFP fusions together with a construct which consisted of 4 copies of the CRT/DRE (CBF-binding motif) fused to luciferase [26]. The effects of expression of Vaccinium CBF upon CRT/DRE-controlled LUC expression was determined by comparing LUC expression obtained by co-expressing a free-GFP control. As shown if Figure 6, the V. myrtillus CBF-GFP construct was able to induce LUC expression to a higher relative level than V. myrtillus/V. vitis-idaea CBFs, mimicking the effect of non-tagged CBFs in transgenic Arabidopsis (Figures 2 and 3). Interestingly, in transient expression in Nicotiana benthamiana, both the V. vitis-idaea and V. uliginosum CBFs were able to transactivate the DRE/CRT, unlike in stable transgenic Arabidopsis. The levels of transactivation in Nicotiana benthamiana, however, were not as high as for V. myrtillus, with V. vitis-idaea and V. uliginosum CBFs inducing to only roughly 30 and 50% of the levels seen with V. myrtillus CBF, respectively. To test for localisation of the protein fusions in the cell we imaged the GFP fusions using confocal microscopy. As shown in Figure 7, the localisation of all three CBF-GFP fusions was exclusively nuclear. As expected the free GFP control was localised in both the cytosol and the nucleus. We tested protein expression of all 3 CBFs 24, 48 and 72 h after infiltration, by extracting proteins and performing western blot analysis using a GFP polyclonal antibody. These data showed that there was a peak of CBF-GFP levels after 48 h for both V. myrtillus and V. vitis-idaea, and levels were very similar between the two

Discussion
Our main aim was to assess the potency of CBF/DREB1 transcription factors from three Arctic plant species from the genus Vaccinium, namely V. myrtillus, V. uliginosum and V. vitis-idaea, as these species have the ability to tolerate significantly low temperatures [16,17,18]. To determine whether the predicted CBF protein sequences from V. myrtillus, V. uliginosum and V. vitis-idaea might reveal differences that could be correlated to altered transcription factor activity, we cloned a single CBF coding from each species. It is important to bear in mind that the three Vaccinium species are likely to have multiple CBF genes as do other plant species. None of the three Vaccinium species has had their genomes sequenced yet, so we cannot yet gauge the complexity of the CBF gene families in these three species. It is important, therefore, to bear in mind that conclusions regarding differences in CBF activity in the three species in a environmental context must be circumspect, as other (not revealed) Vaccinium CBF proteins may have totally different activity. Lining up the three Vaccinium CBF protein sequences with existing CBF/DREB1 sequences from Arabidopsis [9], another accession of V. vitis-idaea from China [38] and V. corymbosum [13] allowed us to compare three important protein domains for these proteins, namely the AP2/EREBP signature domain [40], the DSWAR region [40] and the COOH region [42]. In Arabidopsis the AP2/EREBP signature domain has been shown to be important to binding of CBF to its cognate DNA element, the DRE/CRT [40]. As can be seen in figure 1 the AP2/EREBP, which in all three Arabidopsis CBF genes is ''PKKPAGRKK-FRETRHP'', shows two changes in the Vaccinium genus namely ''PKKRAGRKKFKETRHP''. These proline to arginine and arginine to lysine substitutions were found in all five Vaccinium sequences. As it had been previously demonstrated that V. corymbosum CBF could activate COR gene expression in transgenic Arabidopsis [13] and Vaccinium [14], it is clear that these changes do not inhibit binding of CBF to the DRE/CRT DNA motif. Similarly for the DSAWR sequence, there was one alanine to valine substitution that was common in all five Vaccinium sequences in comparison to Arabidopsis. By the same logic, this substitution cannot be responsible for inhibiting binding to the DRE/CRT element. However, strikingly, the sequence for V. vitis-idaea from the Swedish accession we collected, had one substitution not present in any of the other sequences, be they Arabidopsis or Vaccinium. This was an arginine to glutamine. This ''DSVWR'' to ''DSVWQ'' change was not found in the V. vitis-idaea accession from China [38]: clearly resulting from a polymorphism in the Northern Swedish population. Whilst the DSVWR motif is highly conserved in plant CBF/DREB1s [40] it has not, to our knowledge, been demonstrated to be essential for binding to the DRE/CRT motif. However, the amino acid substitution in our Northern Sweden accession of V. vitis-idaea might be responsible for reduced binding activity. Comparing the COOH regions, which are responsible for transcriptional activity [42], whilst there were numerous differences in sequences between Arabidopsis and Vaccinium CBF sequences, there was only a single difference between the five Vaccinium sequences. This difference was in a single species, V. corymbosum, where a leucine was present where in all four other Vaccinium species an isoleucine was present. Interestingly, the V. corymbosum sequence was identical to the Arabidopsis sequence at this point. Given that it has been demonstrated that all three Arabidopsis CBF sequences, and V. corymbosum CBF can activate COR genes in transgenic Arabidopsis [9,13] it was a formal possibility that in the other three Vaccinium species the leucine to isoleucine substitution was sufficient to render them incapable of transactivation. However, the similarity in properties between leucine and isoleucine rendered this possibility relatively unlikely.
To empirically test these sequence-based predictions, we overexpressed CBF cloned from V. myrtillus, V. uliginosum and V. vitis-idaea all collected in Northern Sweden. As can be seen in Figure 2A transgenic lines were obtained for the three different Vaccinium CBF sequences. Interestingly, comparison of the levels of CBF overexpression showed than on average, the levels of expression were lower for V. myrtillus than the other two species. This implied some sort of fitness penalty for CBF overexpression for this species, selecting against highest expressers. When the expression of CBF target COR genes KIN2, GOLS3 and LTI78 ( Figure 2B, 2C and 2D, respectively) was measured, it could be clearly seen that on average, only V. myrtillus CBF expression resulted in greatly elevated COR gene expression compared to wild type. For V. corymbosum CBF it had been reported that it was incapable of inducing KIN2, GOLS3 and LTI78 when overexpressed in Arabidopsis, but was able to induce COR15A and COR414 [13]. As seen in Figure 3, V. myrtillus was also able to induce both of these COR genes. This implies that the selectivity shown by V. corymbosum CBF [13] is not a feature of V. myrtillus CBF. How such selectivity could occur in V. corymbosum is not clear, but this property must reside in the differences between the CBF protein myrtillus CBF overexpression did not correlate with levels of COR gene expression. This implies that the levels of CBF transcripts do not correlate with CBF protein level/activity directly, and implies posttranslational control. It is very clear that the CBF sequences from V. uliginosum and V. vitis-idaea were very ineffective at inducing the expression of COR genes, despite very high levels of CBF transcripts (Figures 2 and 3). In the case of V. vitis-idaea this could be due to poor binding to the CRT/DRE element if the amino acid substitution in the DSAWR domain described above was significant. However, in the case of V. uliginosum the prediction would be that it would bind CRT/DRE as well as CBF from V. myrtillus and V. corymbosum. Similarly, the poor induction of COR genes by V. uliginosum and V. vitis-idaea CBF could not be due to poor transactivation as the COOH domains were identical to V. myrtillus CBF (Figure 1). It was therefore possible that the differences in potency were as a result of differences in  To test these possibilities directly, we produced GFP-tagged versions of CBF from the three Vaccinium species, and transiently expressed them in Nicotiana benthamiana. As the GFP tag might affect the function of the CBF we firstly tested the ability of all three GFP fusions to induce a DRE/CRT::LUC construct by coexpressing them. As can be seen in Figure 6, the pattern of activation seen in stable Arabidopsis expressing non-tagged CBF was replicated: the V. myrtillus CBF was far more effective at activating expression via DRE/CRT promoter motif than the CBFs from the other two species. Interestingly, in the transient expression system V. vitis-idaea CBF was capable of transactivating CRT/DRE::LUC ( Figure 6), even though it had been unable to induce COR genes in transgenic Arabidopsis (Figures 2 and 3). There are several possibilities for this difference, for example the levels of overexpression in Nicotiana benthamiana might be sufficiently greater to overcome any potential reduced binding to CRT/DRE binding-efficiency as a result of the ''DSVWR'' to ''DSVWQ'' change (see above). We examined the cellular localisation of the GFP-CBF fusions using confocal microscopy ( Figure 7). Our results showed that there was no difference in the localisation of all three CBF GFP fusions: all were discreetly localised to the nucleus. To determine the relative levels of the three proteins, we performed western blot analyses on proteins extracted from Nicotiana benthamiana tissue in which the GFP fusions were expressed. We analysed the levels of protein expression 24, 48 and 72 h after infiltration. Both V. vitis-idaea and V. myrtillus CBF proteins showed the same behaviour: peaking in expression at 48 h, with relatively high levels of expression (compared to 24 and 72 h timepoints, and compared to V. uliginosum). The situation with V. uliginosum was considerably different, with relatively low levels of protein, and no appreciable peak. Thus, the lack of potency for V. uliginosum at inducing COR gene expression (Figures 2 and 3) or transactivating the DRE/COR element ( Figure 6) could be due to low protein levels. Given that the constructs we used were identical apart from the three coding regions, it is highly unlikely that this low level of protein is due to lower levels of translation, thus we conclude that the V. uliginosum CBF is relatively unstable compared to CBF from the other two Vaccinium species. This difference is most likely due to differences in protein sequence e.g. presence of sites for ubiquitination in the V. uliginosum sequence. For instance, lysine78 in the V. uliginosum CBF sequence is not present in the other two Vaccinium species. Whether this is a site for translational modification e.g. polyubiquitination leading to proteosomal degradation would be an interesting topic for future study. On the other hand, the lack of a difference in protein expression ( Figure 8) and targeting (Figure 7) for V. vitis-idaea compared to V. myrtillus does not explain the inability of the former to induce COR genes (Figures 2 and 3) or fully transactivate DRE/CRT ( Figure 6). Along with the fact that the sequence data render the likelihood of poor binding to DRE/CRT or ineffective transactivation domain low, one can only surmise that the difference is due to reduced ability to interact with other components of the transcriptional machinery e.g. Mediator. In such a scenario, the amino acid differences between V. vitis-idaea and V. myrtillus CBF might be responsible, and it would be interesting to test mutations of V. vitis-idaea CBF towards the V. myrtillus sequence to see if activity can be rescued.
The ability of V. myrtillus CBF to activate COR gene expression in transgenic Arabidopsis indicated the possibility that there would be an increase in the freezing tolerance of these lines. This was tested and it was confirmed that transgenic Arabidopsis overexpressing V. myrtillus CBF was more freezing tolerant than either wild type or transgenic Arabidopsis expressing CBF from the other two Vaccinium species. Also importantly, the growth phenotype associated with CBF overexpression, dwarfism, was also only manifest when V. myrtillus was overexpressed.
V. myrtillus in the Arctic can only survive the winter under a deep insulating snow cover: loss of this cover results in extreme frost damage [43]. V. vitis-idaea and V. uliginosum can both survive under much shallower, and thus colder, snowpacks in more exposed parts of the natural landscape. It is thus clear that V. vitis-idaea and V. uliginosum have evolved relatively greater freezing tolerance with respects to V. myrtillus. If the behaviour of the Vaccinium CBFs observed in Arabidopsis and Nicotiana represent the behaviour in the host genetic background, this would suggest the higher level of freezing tolerance found in V. vitis-idaea and V. uliginosum is not CBF-based, but resulting from another pathway/trait. One has to apply caution, however, when trying to extrapolate CBF function in the native species using data based upon prediction of CBF function by expressing in a heterologous system (Arabidopsis and Nicotiana) and behaviour of the CBF proteins could be different in their natural genetic backgrounds. Also, as in Arabidopsis, the CBF protein of the three Vaccinium species may be encoded by a multigene family, and more potent versions of CBF found in the V. uliginosum and V. vitis-idaea genomes. In the future, taking a whole genomic approach in these Vaccinium species e.g. full genome sequencing and assembly, will answer these questions. It will be important along with this to develop proteomic and transcriptomic platforms for these species to fully elucidate the molecular basis of freezing tolerance in these remarkable species. Figure S1 Expression of CBF in V. myrtillus, V. uliginosum and V. vitis-idaea. Relative CBF transcript abundance was measured in either V. myrtillus, V. uliginosum or V. vitis-idaea (''M'', ''U'' and ''I'', respectively). ''A'' corresponds to ambient and ''C'' is a 2 h treatment at 5uC. Data was normalised using beta-tubulin expression. Error bars represent RQ MIN and RQ MAX and constitute the acceptable error level for a 95% confidence level according to Student's t-test [31]. (TIF)