Structural and Functional Insights into the Cryoprotection of Membranes by the Intrinsically Disordered Dehydrins*

Background: Genetic evidence supports a protective role for plant dehydrins against drought and cold. Results: Dehydrins prevent membrane fusion and lower the transition temperature without altering membrane accessibility and fluidity. Conclusion: The lysine-rich segments are important for protecting membranes from freeze-thaw damage. Significance: An amphipathic helix with positively charged residues flanking the hydrophobic face may be a common motif for disordered stress proteins. Dehydration can be due to desiccation caused by a lack of environmental water or to freezing caused by a lack of liquid water. Plants have evolved a large family of proteins called LEA (late embryogenesis abundant) proteins, which include the intrinsically disordered dehydrin (dehydration protein) family, to combat these abiotic stresses. Although transcription and translation studies have shown a correlation between dehydration stress and the presence of dehydrins, the biochemical mechanisms have remained somewhat elusive. We examine here the effect and structure of a small model dehydrin (Vitis riparia K2) on the protection of membranes from freeze-thaw stress. This protein is able to bind to liposomes containing phosphatidic acid and protect the liposomes from fusing after freeze-thaw treatment. The presence of K2 did not measurably affect liposome surface accessibility or lipid mobility but did lower its membrane transition temperature by 3 °C. Using sodium dodecyl sulfate as a membrane model, we examined the NMR structure of K2 in the presence and absence of the micelle. Biochemical and NMR experiments show that the conserved, lysine-rich segments are involved in the binding of the dehydrin to a membrane, whereas the poorly conserved φ segments play no role in binding or protection.

Water deficiency can be lethal because it leads to damage to DNA, membranes, and proteins, including an impairment of enzyme function. Many poikilothermic organisms express proteins that protect them from the detrimental effects of freezing and dehydration. In plants, this includes dehydrins (dehydration proteins) whose up-regulated transcription and translation has been associated with protection from several abiotic stresses, including drought, cold, and high salinity (for reviews on dehydrins, see Refs. [1][2][3][4][5][6]. Dehydrins are a member of a large family of proteins, known as LEA (late embryogenesis abundant) proteins (7)(8)(9).
Dehydrins themselves have been found in a number of different intracellular locations and are often found in the cytoplasm and the nucleus. Other locations include at the mitochondrial, chloroplast, and plasma membranes. The diverse localization may also be reflected in the diverse protective effects of dehydrin in the plant. These include preventing electrolyte leakage across cold-stressed membranes (10,11), preventing lipid peroxidation (11), and a reduction in stomatal density to prevent dehydration (12).
An examination of the dehydrin sequence reveals that they are modular in nature and contain a variable number of conserved sequence motifs. This modularity results in a range of dehydrin sizes . By definition, dehydrins must contain at least one K segment (1). This 15-residue, Lys-rich motif (EKKGIMDKIKEKLPG) can be found in 1-11 copies in a single dehydrin sequence. The S segment, which consists of 5-7 Ser residues, is not present in all dehydrins. When present, it is normally present only once. The Y segment consists of the sequence motif (V/T)D(E/Q)YGNP and is usually present in one or two copies. Residues not found in one of these three motifs are said to be located in the segment. This segment is not conserved in terms of length but is rich in Gly and polar/ charged amino acids such as Glu, Thr, Lys, and His (6). The lack of hydrophobic residues explains the lack of tertiary structure in dehydrins. Dehydrins are intrinsically disordered proteins (IDPs) 2 (13,14), a property that prevents them from being denatured during dehydration and has been used to purify them (15).
The roles of the various conserved motifs are not yet fully clear. Dehydrins that contain multiple copies of the K segment * This work was supported by an Natural Sciences and Engineering Research tend to be more expressed during cold-stress, whereas the Y segment dehydrins tend to be correlated with desiccation and salt stress (5,6,16). In terms of localization within the cell, dehydrins containing K segments or K and S segments are found in the cytoplasm and at the membrane (17), with the possibility that residues in and near the S segment are responsible for nuclear localization (18).
To better characterize the physiological role of dehydrins, a number of in vitro assays have been performed. One of the most extensively performed assays is the enzyme cryoprotection. The most commonly used enzyme is lactate dehydrogenase (19,20), although the starch degradation enzyme ␣-amylase has also been used (21). The repeated freezing and thawing of lactate dehydrogenase results in its loss of activity, most likely because of aggregation (22). The addition of dehydrin results in the recovery of activity, with the longer the dehydrin the more efficient the recovery (23). Dehydrins have been shown to bind metals, which may prevent the formation of reactive oxygen species. These proteins have also been found to bind DNA and RNA in a nonspecific manner (24), which was postulated to help protect nucleic acids from desiccation damage. The localization studies showing that dehydrins can be located near the membrane (25) prompted several studies to examine whether binding can occur in vitro, using both liposomes and micelles (5, 26 -30).
Many IDPs can gain structure upon binding to a ligand. The same is true for dehydrins, because CD studies showed that the protein became ␣-helical in the presence of SDS micelles and liposomes consisting of phosphatidylcholine and a second lipid (phosphatidic acid, phosphatidylglycerol, or phosphatidylserine) that must be negatively charged (27,30). Dehydrins have also been shown to interact with liposomes whose lipid composition mimics that of the plasma membrane (31). Deletion studies with the maize dehydrin ZmDHN1 showed that the K segments are likely responsible for binding to these different membranes, although which residues were responsible was not determined, nor if there is a weaker interaction with the segment (28). A need to closely examine the role of the segment is driven by a study of the Arabidopsis dehydrin Lti30, which is able to bind liposomes, whereas the K peptide (that is, a peptide consisting of only the 15-residue K segment) failed to bind (32).
The mechanisms by which dehydrins may protect the membrane from desiccation and cold stress also require further study. A study on the Arabidopsis dehydrins ERD10 and ERD14 used diphenylhexatriene (DPH) to examine the mobility of a membrane in the presence of these two dehydrins (30). No effect was observed, but this may not be surprising because this probe partitions in the acyl chain region of the liposome, whereas the polar dehydrins are likely to interact at the liposome surface. The study using Arabidopsis Lti30 dehydrin also examined membrane fluidity by measuring the phase transition temperature of 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC):1,2-dimyristoyl-sn-glycero-3-phospho-L-serine liposomes (32). The protein was able to lower the transition temperature of the liposomes, which would allow the plant to maintain membrane fluidity at lower temperatures (32). However, this does not address the problem of how dehydrins could protect membranes at subzero temperatures.
We examine here the ability of the minimal dehydrin K 2 to protect liposomes from freeze-thaw and cold stress damage, what effect binding has on some membrane properties, and what effect binding has on K 2 structure. We also see whether the gain in ␣-helicity is restricted to the K segments and whether the segments play any role in binding to the liposome.

Experimental Procedures
Expression, Purification, and Labeling of the Dehydrin-The Vitis riparia K 2 protein was expressed and purified as described previously (15,22), with the following modifications for isotopically labeled proteins: Escherichia coli BL21(DE3) cells transformed with pET-22B expression vector containing the K 2 gene were grown in Luria-Bertani medium until an A 600 of ϳ0.8 was reached. Cell cultures were centrifuged at 5,000 ϫ g for 30 min, and the cell pellet was resuspended in 1 liter of modified M9 minimal with [ 13 C]glucose (3 g) and/or [ 15 N]NH 4 Cl (1 g) for protein labeling. The protein was extracted by boiling the resuspended pellet in water for 20 min; sodium acetate, pH 5.0 buffer was added to give a final buffer concentration of 20 mM. Purification by cation exchange and desalting by reversed phase HPLC was performed as previously described (22). The protein was lyophilized to dryness and stored at Ϫ20°C until use. K peptide was synthesized and purified as described previously (23).
Preparation of Liposomes-L-␣-Lyso-phosphatidylcholine (egg PC), L-␣-phosphatidic acid (egg PA), L-␣-phosphatidylglycerol (egg PG), and L-␣-lysophosphatidylserine (brain PS) were obtained from Avanti Polar Lipids (Birmingham, AL). The lipids were dissolved in a 4:1 chloroform:methanol solution at a concentration of 100 mg/ml. To form a lipid cake, 12.5 l of both PC and PA, PG, or PS were dispensed in the bottom of a glass vial, mixed, and subsequently dried under a stream of nitrogen gas. After 1 h of further evaporation of solvents under vacuum, the lipids were resuspended in 50 mM phosphate buffer, pH 7.4. After five liquid nitrogen freeze-thaw cycles, the solution was extruded 21 times through a 100-nm polycarbonate membrane that had been warmed to 45°C (33). Liposome size was confirmed using dynamic light scattering (described below), and preparations were used for experiments within 4 days.
Liposome Freeze-Thaw Damage and Recovery by K 2 -For the fusion assays, liposome solutions were diluted to a concentration of 0.4 mg/ml in 5 mM phosphate buffer, pH 7.4, with and without K 2 , lysozyme (egg white, white crystalline, ϳ100% pure; Fisher Scientific), or PEG 3350 (Ն99.95% pure, M r 3,000 -3,700; Sigma-Aldrich). Samples (100 l) were chilled to Ϫ20°C in an enzyme cooler box for ϳ2 min. To nucleate the flashfreeze process, the base of each tube was then firmly pressed against a metal spatula that had been chilled by liquid nitrogen. Once frozen, tubes were stored at Ϫ20°C for 12-16 h. Tubes were then thawed at room temperature, and 400 l of buffer was added before size measurement by dynamic light scattering. Size distributions for liposomes were measured using a ZetaSizer Nano S (Malvern Instruments, Worcestershire, UK). Analysis was done in the Malvern Dispersion Technology software suite, version 3.3, using the default settings. Measure-ments were made at 25°C, and the data were analyzed using the discrete peak calculation method with polystyrene latex as the reference standard.
Measurement of Lipid Accessibility and Mobility-Surface accessibilities in the presence of K 2 , lysozyme, and PEG 3350 were assessed using the dye merocyanine 540 (34). Liposomes (0.3 mg/ml) were prepared in 50 mM phosphate buffer, pH 7.4. K 2 , lysozyme, or PEG 3350 were added to the liposome containing solutions, and samples were equilibrated for 30 min. A stock solution of merocyanine 540 was added to a final concentration of 4 M, and the samples were allowed to equilibrate for an additional 30 min. The absorbance ratio (570 nm/540 nm) was measured using a Cary 100 Bio UV-visible spectrophotometer (Agilent Technologies) and corrected for absorbance at 570 and 540 nm in the absence of the fluorescent dye.
For the lipid mobility experiments (35), liposomes were constituted in the usual manner, with DPH or trimethylammonium diphenylhexatriene (TMA-DPH) added to the lipid mixture in the first step to produce a lipid:probe molar ratio of 200:1. The probed lipid cake was dissolved in 50 mM phosphate buffer, pH 7.4, and extruded as described previously. Lipids were diluted to experimental concentrations of 0.1 mg/ml for measurement in a PTI QuantaMaster C-61 steady-state fluorometer with an electronic polarized excitation filter and two manual emission polarizing filters (Photon Technology International, London, Canada). The G-factor was obtained first for each sample using the equation, where I VH and I HH are the intensities of the vertical and horizontal excited light, respectively, with the emission polarizer set in the horizontal position. The emission polarizer was then set in a vertical position, and the measurements were repeated. Anisotropy (r) was calculated by the equation, where I VV is the intensity of the vertical excited light with the emission polarizer set in the vertical position.
Liposome Transition Temperature Measurement-The transition temperature (T m ) of the PC:PA liposomes was measured by differential scanning calorimetry (DSC). Liposomes consisting of a 1:1 (w:w) ratio of DMPC to 1,2-dimyristoyl-sn-glycero-3-phosphate (DMPA) were prepared as described above except the extrusion process was performed at 55°C to maintain fluid lipids. K 2 dehydrin, lysozyme, or PEG 3350 were added at a concentration of 75 M to liposomes at a concentration of 0.4 mg/ml. The mixtures were allowed to incubate for 30 min before being degassed for 15 min. Measurements were made on a MicroCal VP differential scanning calorimeter (MicroCal, Northampton, MA). To account for the heat capacity of the buffer alone, the buffer was scanned against a water reference. DSC measurements were performed with up-scans at a rate of 45°C/h, using a 5-s filtering period and a 15-min prescan equilibrium and with high gain feedback mode enabled. High gain feedback was used because of the relatively low signal intensity observed at these lipid concentrations. Because the noise was generated in this mode, the filtering period was set to 5 s to provide a more reliable signal. To account for the "thermal history" of a sample, 15 scans were performed, with only the final scan kept for analysis. K 2 -Liposome Binding-The binding of K 2 to liposomes was assessed by separating bound and unbound protein by centrifugation (36). PC:PA liposomes in 5 mM sodium phosphate, pH 7.4, were added to protein to create a range of 0.5:1 to 10:1 liposome:protein (w/w) concentrations. Samples (200 l) were incubated with and without K 2 (0.2 g/l) at room temperature for 1 h before centrifugation at 80,000 ϫ g for 30 min, after which the supernatant and pellet were separated. Pellets were resuspended in the same binding buffer, and both supernatant and pellets were subjected to SDS-polyacrylamide gel electrophoresis. Quantification of the bound and unbound protein was made using Bio-Rad Quantity One software, trace quantity command. The amount of protein was quantified by comparison to a gel with known amounts of K 2 . To determine the affinity of the protein-liposome interaction and to estimate the stoichiometry, the data were fit according to the equation (37), where [K 2 ] b is the amount of bound K 2 , [K 2 ] total is the total K 2 concentration, B max is the maximal bound signal, K d app is the apparent dissociation constant, L is the total PA concentration, and h is the Hill coefficient.
Secondary Structure Analysis-CD data of the dehydrin proteins were collected using a Jasco-815 CD spectropolarimeter (Easton, MD). All protein samples were dissolved in 10 mM sodium phosphate, pH 7.4, at a protein concentration of 0.16 mg/ml (30 M). A quartz cuvette with a 2-mm pathlength (Hellma, Concord, Canada) containing the protein sample with and without liposome (0.65 mM) or SDS (10 mM) was scanned from 250 to 190 nm. In the SDS titration experiment, the concentration of the detergent was varied between 0 and 1,000 M. For scans containing liposomes, the spectra were averaged over 25 accumulations. For scans containing SDS, the spectra were averaged over 8 accumulations. All CD experiments were performed at 25°C.
Titration of SDS into K 2 -A sample containing 0.2 mM 15 N-K 2 and 9 mM 2 H-SDS was prepared. The SDS titration solution included 0.2 mM 15 N-K 2 to prevent dilution of the protein signal. Changes in chemical shift in the presence of SDS micelles were followed for all of the assigned residues, and the weighted chemical shift change for an individual residue was based on the following equation, where ⌬␦ weighted is the average chemical shift change, ⌬␦ H is the amide proton chemical shift change, and ⌬␦ N is the amide nitrogen chemical shift change.
To determine the affinity of K 2 interacting with the SDS micelle, resonances undergoing fast exchange were fitted to the following equation assuming there is one type of binding site (43), where ⌬ obs is the observed chemical shift, ⌬ max is the maximum chemical shift, K d is the dissociation constant, [L] (42) by fitting the cross-peak intensity decay to a two parameter, single exponential decay function, where I 0 is the intensity at time t ϭ 0 and I(t) is the intensity after a delay time t for the R 1 or R 2 experiment. The errors were estimated from the uncertainty of the nonlinear fits to the data. The { 1 H}-15 N steady-state NOE values represent the intensity ratio from the two HSQC spectra with and without 1 H saturation applied prior to the 15 N excitation pulse.

Results
The Protective Effects of K 2 on Stressed Liposomes-Several previous reports have shown that dehydrins are able to interact with membranes (5, 26 -30). We wished to examine the ability of the minimal dehydrin K 2 to protect from cold stress. Liposomes of 100 nm in diameter were created using an equal weight mixture of PC and PA. The freezing and thawing of a membrane can cause different problems that all lead to a loss of membrane integrity. In our case, we examined the fusion of liposomes by measuring the distribution of liposome diameters before and after freezing in the presence of increasing concentrations of K 2 (Fig. 1, A and B). The unfrozen liposomes have a range of sizes as expected using the extrusion method, with an average diameter of ϳ100 nm. In the absence of any dehydrins, the freeze-thaw-treated liposomes became very large, with diameters averaging ϳ800 nm. K 2 was added over a range of concentrations before freezing. In the presence of the dehydrin, at low to mid-concentrations of protein (5-50 M), the average diameter first increased to 1,000 nm, but at higher concentrations (75 and 100 M), there was a downward shift in the diameter. At the highest concentration (250 M), there is a distribution of sizes clustered around the original diameter (Fig. 1A, left panel). To determine whether the protective effect is specific to PA, the fusion assays were repeated using PC:PS and PC:PG. The results are shown in Fig. 1A (center and right panels) and indicate that K 2 is able to protect liposomes that contain a mixture of neutral and negatively charged lipids.
To examine whether the protective effects of K 2 are specific to dehydrins or are a generic effect that can occur with any protein or disordered polymer, the fusion assay was repeated in the presence of lysozyme and PEG 3350. Neither lysozyme nor PEG 3350 showed any protective effects whatsoever, even at a concentration of 250 M (Fig. 1B). Lysozyme was chosen as a control because it is a small protein with approximately the same hydrodynamic radius as K 2 (23) and is also known to bind to liposomes and promote their aggregation (44). As with dehydrins, lysozyme has previously been shown to bind to negatively charged liposomes (45). This protein did not prevent fusion at moderate concentrations (25-100 M), and at concentrations of 250 M, the liposomes began to show visible signs of aggregation (data not shown), demonstrating that lysozyme did not provide any protection from freeze-thaw damage (Fig. 1B, left  panel) and that the binding of a protein to liposomes could promote aggregation. PEG 3350 was used because it is a disordered polymer with a similar hydrodynamic radius to K 2 . With PEG 3350, the average diameter of the liposomes becomes ϳ1,000 nm. Even at the higher concentrations of PEG 3350, the liposomes remained large, and fusion was not prevented (Fig.  1B, right panel).
The effect of K 2 on membrane surface properties was examined next. We examined whether K 2 coats the surface of a liposome to prevent another liposome from coming in close contact with the surface (i.e. by causing some protective steric hindrance). Surface accessibility was measured using the dye merocyanine 450, which partitions itself between the hydrophilic and hydrophilic phases (34). The presence of 50 or 250 M K 2 or PEG 3350 did not measurably alter accessibility ( Fig.  2A, left and right panels). However, the presence of lysozyme did lower the accessibility of the liposomes as evident in the 4.5-fold decreased fluorescence ratio, possibly because of its ability to cause liposome aggregation (44).
A previous report showed that an Arabidopsis dehydrin was able to lower the transition temperature (T m ) of a liposome (32). However, the authors found that dehydrin membrane binding required His-His sequences flanking the K segment. To see whether this is also required for a minimal dehydrin like K 2 , which lacks the His-His dipeptides, we measured the T m by DSC. The acyl chain diversity in the egg PC and PA preparations prevents the detection of a distinct transition peak (data not shown); to overcome this problem4 liposomes were made with DMPC and DMPA to ensure cooperative transition processes that are easily observed by this technique. Liposomes alone exhibited a sharp gel fluid phase transition at 39.5°C (Fig.  2B). The addition of 75 M K 2 (Fig. 2B, left panel) caused the T m to shift to 36.6°C, with a premelt peak at 33.7°C, whereas the addition of PEG 3350 caused the T m to remain at 39.7°C (Fig.  2B, right panel). The addition of lysozyme in the DSC measurements prevented a major enthalpy transition and caused the formation of precipitate, suggesting that the liposomes may have been damaged by this protein (Fig. 2B, middle  panel). We examined whether the presence of the K 2 dehydrin altered the mobility of lipids in the liposome using DPH and TMA-DPH probes using steady-state fluorescence anisotropy (35). DPH, being a highly hydrophobic molecule, probes the mobility in the alkyl chain region of the bilayer, whereas the TMA-DPH probes mobility near the lipid headgroups. The same concentration range for the compounds was used as in the fusion assay. As can be seen in Fig. 2C, none of these compounds (K 2 , lysozyme, or PEG 3350) significantly altered the mobility of the lipids.
Lipid Binding, Disorder to Order Transition, and the Importance of Charge-The cryoprotective effects of K 2 on the enzyme lactate dehydrogenase was previously shown to occur by a molecular shield effect (23). To see whether the same is true for the cryoprotection of liposomes, we first examined whether K 2 is able to directly bind to the lipid vesicles using a differential centrifugation binding assay (36). The results in Fig.  3A (left column) show that K 2 bound to liposomes consisting of PC:PA and that molecular shielding is unlikely to be the mechanism because it requires the protein to remain unbound. Lysozyme was used as a control because other studies have shown that it binds to negatively charged lipid (44), and BSA was used as a negative control because it only interacts weakly with negatively charged liposomes (46). Fig. 3A (left column) shows that lysozyme did bind to the liposomes under our assay conditions, whereas BSA did not.
Two studies on other dehydrins had shown that negatively charged lipids are important for the interaction (27,28). To test whether this is also true for K 2 , we examined binding in the presence of PC-only liposomes. As can be seen for K 2 in Fig. 3A (right column), neither K 2 nor lysosome were able to bind to these zwitterionic lipids because almost all of the protein is found in the unbound supernatant fraction. This shows that the presence of phosphatidic acid, most likely caused by the nega-  NOVEMBER 6, 2015 • VOLUME 290 • NUMBER 45 tive charge of the phosphate headgroup, is important for the binding of K 2 to the membrane surface. Because PA needs to be present for K 2 to bind to the liposome, the importance of electrostatic forces in the binding interaction was directly examined with increasing amounts of NaCl. As the concentration of salt was increased, the amount of K 2 bound to the liposome decreased (Fig. 3B). The largest drop in the percentage of protein bound occurred between 0 and 200 mM NaCl (ϳ30% decrease), after which there is a further drop but in much smaller decrements (3-10% decreases). To further biochemically characterize the interaction, the apparent binding affinity was measured using the pulldown method (Fig. 3C). The ligand concentration is expressed in terms of total PA concentration because K 2 does not interact with PC-only liposomes. Because of the sigmoidal shape, the binding curve was fit to Equation 3.

Dehydrin Structure and Membrane Protection
From the fit, the apparent K d was determined to be 19 M with a cooperativity coefficient of 6.8.
Many IDPs gain structure when they bind to a ligand. To examine what changes may have occurred in K 2 , we compared the CD spectra of this dehydrin in the presence and absence of PC:PA liposomes (Fig. 3D). In the absence of lipids (Fig. 3D, filled circles), the spectrum shows the typical pattern of a disordered protein with a strong negative ellipticity at 198 nm and a weak shoulder near 222 nm. After the addition of a 70ϫ molar excess of lipid, it can be seen that the minimum at 198 nm has shifted to 205 nm and become less negative and that the ellipticity at 222 nm has decreased (Fig. 3D, open squares). These changes suggest that the K 2 protein has lost ϳ6% coil structure and gained ϳ2% ␣-helicity in the presence of a membrane (47). Based on the CD data from previous dehydrin deletion studies  (48 -50) prediction of residues that could form a helix. F, helical wheel projection of the K segment when bound to the liposome. The figure was drawn using Heliquest (68). The beginning and end of the sequence are marked with a red N and a red C, respectively. The arrow indicates the hydrophobic moment ϽHϾ, which is 0.578. Hydrophobic residues are colored yellow, basic residues are blue, acidic residues are red, and glycine is gray. that suggested that K segments have a role in membrane binding (28) and the importance of electrostatics in the K 2 /liposome interaction we observed here, we present a possible helical model of the K segment sequence and how it may interact with the surface of a membrane. Using a combination of the helix/ coil transition prediction algorithm Agadir (48 -50) (Fig. 3E) with the NMR data from K 2 in the presence of micelles (see below), we propose a model where 11 residues of the K segment interact with the membrane surface such that there are four hydrophobic residues (two Met, Leu, and Ile) that weakly interact with the acyl chains and four flanking Lys residues that have favorable electrostatic interactions with the phosphate head groups (Fig. 3F).
Interaction of K 2 Dehydrin with a Micelle Membrane-Our goal was to determine the site-specific secondary structure of K 2 when bound to the membrane using several biophysical techniques, including solution state NMR. Because of the large size of a liposome, NMR studies on bound proteins can be challenging because of the large molecular weight of the liposome that causes a dramatic increase in the tumbling rate of the bound protein. SDS micelles with their smaller size are especially suitable to the study membrane proteins by NMR (51,52). To that end, we used SDS micelles as a membrane because it is regarded as a good membrane substitute to mimic the interactions that occur between proteins and lipids (53).
We acquired the CD spectrum in the presence and absence of SDS micelles (Fig. 4A) to see whether the structural changes of K 2 upon binding were similar to those observed in the presence of liposomes. As with K 2 /liposome interaction shown in Fig.  3D, the K 2 signal minimum shifted from 198 to 205 nm with a concomitant decrease in the CD signal, and a significant gain in the CD signal was observed at 222 nm. The latter shows that K 2 has lost ϳ15% coil structure and gained ϳ7% ␣-helicity when bound to the micelle surface (47).
To facilitate an analysis of the changes in the structure and dynamics that occur when the dehydrin binds to a membrane, we determined the chemical shifts of K 2 in the presence of SDS micelles. The changes that occurred are highlighted by the overlap of an 15 N HSQC of K 2 in the presence and absence of SDS (Fig. 4B). The K 2 alone spectrum shows very little dispersion in the amide proton and nitrogen dimensions, with all of the backbone HN resonances located between 8.0 and 8.5 ppm. This observation is typical for intrinsically disordered proteins. The addition of SDS shows the increased dispersion of many resonances, with HN resonances now spread over 7.3 to 8.7 ppm. These findings are in agreement with the CD data in that there is some structuring and/or interaction of the protein because of the micelles. Some resonances show larger changes than others, suggesting that specific regions of K 2 are interacting with the detergent.
To determine which K 2 residues were affected by SDS binding, we plotted the chemical shift differences of the backbone A, CD spectra of K 2 alone (filled circles) and K 2 in the presence of SDS micelles (open squares). B, 15 N HSQC of K 2 in the presence (blue peaks) and absence (orange peaks) of SDS micelles. Peak assignments are shown as the residue number followed by its single-letter amino acid code. C, weighted chemical shift differences between bound and unbound forms of K 2 . The differences were calculated according to Equation 4. The diagram at the top of the panel shows the location of the various segments in the K 2 sequence. D, ␦2D analysis of chemical shifts. The probability of secondary structure on per residues basis was calculated for K 2 in the presence (bottom panel) and absence (top panel) of micelles. Light blue, coil; dark blue, ␣-helix; red, ␤-strand; green, polyproline type II helix. NOVEMBER 6, 2015 • VOLUME 290 • NUMBER 45

JOURNAL OF BIOLOGICAL CHEMISTRY 26907
protein and nitrogen amide atoms of the bound and unbound forms of K 2 on a per residue basis (Fig. 4C). As this figure shows, the largest changes occurred in residues 4 -11 and residues 37-44. These residues are found in the K segments (defined in K 2 as residues 1-11 and 32-46). The largest chemical shift changes (Ͼ2 ppm) are located in the ends of the K segment (residues 6 -9 and 41-44). The segments, located between residues 12-31 and 47-48, showed generally little chemical shift change (0.5 Ϯ 0.3 ppm). These results suggest it is the conserved K segments and not the segments that are interacting with the SDS.
These localized chemical shift effects in the presence of SDS are also reflected in gains in the secondary structure in K 2 . The HN, H␣, C␣, C␤, CЈ, and backbone N chemical shifts of K 2 alone and in the presence of SDS were analyzed using the program ␦2D (54). The method uses these chemical shifts to predict the presence of transient secondary structures in disordered proteins. The program discriminates between ␣-helices, ␤-strands, type II polyproline helices, and coil on the basis of the combined chemical shifts. The top panel of Fig.  4D shows the secondary structure probability (P ss ) for K 2 alone in buffer. The structure is predominantly coil, with a small amount of polyproline II helix located in the residues 4 -6, 16 -20, 27-30, and 36 -43, and a small amount of ␤-strand in residues 9, 24 -26, and 44 -45 (i.e. adjacent to the three proline residues). In the presence of SDS micelles (Fig.  4D, bottom panel), one can see that a considerable amount of ␣-helicity has been gained in residues 3-9 and residues 34 -44, with remaining residues in the K segments showing a small amount of polyproline II helix structure, whereas a small amount of strand is still seen in residues 24 -25.
To examine the K 2 -micelle interaction in greater detail, a 15 N HSQC titration experiment was performed (Fig. 5A). Resonances that were very well separated from all others (Glu-7, Arg-14, Thr-21, Glu-42, Leu-44, and Ala-47) were used to further characterize the protein-detergent micelle binding interaction (Fig. 5, B-D). The residues cover the three different segments of K 2 (first K segment, central segment, second K segment, and C-terminal segment). In addition, these various residues were undergoing exchange on three different time scales (slow, intermediate, and fast). The data from residues undergoing fast exchange were fitted to the binding equation (Equation 5) to determine the affinity and apparent stoichiometry of K 2 binding to SDS. The results show that the apparent K d is 16 M (i.e. a moderately weak affinity), and the stoichiometry is 7. We mapped the different exchange rates on the K 2 sequence (Fig. 5E) depending on how the peaks changed during the titration (i.e. the single peak that shifted without a loss in intensity is considered to be a residue in fast exchange; peaks with decreasing intensity in one location and appearing in another location are considered to be in slow exchange; and peaks that both displayed a change in intensity and shifted are considered to be in intermediate exchange). Proline residues and residues that could not be followed in the titration because of spectral overlap remain uncolored in the figure panel. Residues located in the middle of the K segments were in slow exchange. This suggests that they are likely binding to the micelles and changing structure (gaining ␣-helicity). Residues in the central segment and at the C-terminal segment (the two C-terminal residues of the protein) were in fast exchange, suggesting that they do not interact with the membrane surface. Residues located at the interface between fast and slow exchanging regions were undergoing intermediate exchange because they represent transitioning from parts of the protein that are strongly interacting with the micelle to those that have no interaction.
To further characterize the dehydrin membrane surface interaction dynamics, we measured the R 1 , R 2 , and 15 N NOE relaxation parameters of K 2 in the presence and absence of SDS (Fig. 6, A-C). In the absence of micelles (filled circles), the two K segments showed slightly higher R 1 values than the mean (R 1 average 1.3 Ϯ 0.1 s Ϫ1 ), whereas the center of the segment showed slightly lower values. This suggests that the K segments are slightly more rigid than the segment, which may be reflected in their propensity to form secondary structure in the presence of a membrane surface (see below and Ref. 28). The R 2 and NOE values showed similar patterns with the segment being more flexible (Fig. 6, B and C). Note that the R 2 plot (R 2 average 2.0 Ϯ 0.2 s Ϫ1 ) appears flat (Fig. 6B, filled circles), but an expanded y axis scale reveals the same pattern of increased flexibility in the segment region (see Fig. 2 in Ref. 22). In the absence of SDS, the NOE values were negative over the entire protein (ϽϪ0.25), with the region showing NOE ratios of Ϫ1.5 to Ϫ2.
In the presence of micelles (open squares), not all three parameters show the same pattern. For the R 1 data ( Fig. 6A; R 1 average 1.5 Ϯ 0.1 s Ϫ1 ), the plot is rather featureless, showing only a small increase in flexibility at the C terminus. R 1 is not affected by conformational exchange, so the gain in helicity is not reflected in this relaxation parameter. In contrast, the NOE ratio (Fig. 6C) and the R 2 data (Fig. 6B) showed a difference once again between the different segments, with the NOE ratio approaching zero for the region and 0.5-1.0 for the K segments. The largest range of values are in the R 2 measurements (R 2 average of 7 Ϯ 3 s Ϫ1 ), where the core of the K segment residues have R 2 values of 10 -13 s Ϫ1 , whereas the segment residues flanking the K segments have R 2 values of ϳ7 s Ϫ1 . The center of the segment has R 2 values of ϳ4 s Ϫ1 , approaching the R 2 values of K 2 alone in solution.

Discussion
Our results show that even a minimal dehydrin such as K 2 , with only two K segments and a central segment, is able to prevent membrane damage during cold stress. The type of cold protection provided appears to be 2-fold: by preventing freeze/ thaw-induced fusion (Fig. 1A) and by keeping membranes fluid at a lower temperature (Fig. 2B). These two effects have also been observed for two other plant stress IDPs. The dehydrin Lti30 from Arabidopsis has also been shown to lower the T m (32). In that report, the authors suggested that the decreased T m allows the membrane to remain fluid at lower temperatures, a property that is important for proper membrane function (32,55). A mitochondrial LEA protein (LEAM) was also shown to lower the T m of model liposomes, and to reduce fusion caused by dehydration (56). The yeast Hsp12 protein, a yeast stress IDP, in contrast, increased the T m of a model liposome (57). Another difference between Hsp12 and K 2 is that Hsp12 decreased membrane fluidity as measured by fluorescence ani-sotropy, which the authors say is the cause of the increased T m (57). Functionally this makes sense, because Hsp12 is a heat shock protein that would need to keep the membrane at the same fluidity at higher temperatures. Hsp12 has been shown to gain four helices in the presence of SDS micelles. Although helix 4 is similar to the K segment sequence, the other three helices are different, with helix 2 having a Lys residue located on the hydrophobic face (see Fig. 3 in Ref. 57). It is possible that a different arrangement of the residues in these amphipathic helices and/or a different insertion depth could decrease lipid mobility rather than keep it fluid.
The dramatic flexibility of the segment even in the presence of a membrane surface (Fig. 6B) led us to hypothesize that the segment could act as a steric shield to protect the membrane from fusion. This could prevent two liposomes from coming close enough together to fuse. Such a model would be similar to what we proposed with K 2 preventing the aggregation of lactate dehydrogenase after freeze-thaw denaturation (22,23). To test this, we used the fusion assay with the PC:PA liposomes and measured the T m of DMPC:DMPA liposomes using a K peptide (Fig. 7). The 15-residue peptide is identical to the second K segment found in V. riparia K 2 . For both experiments, the peptide was used at twice the concentration to compensate for the presence two K segments in K 2 . The prevention of fusion and the decrease in T m are very similar to what is seen for the intact K 2 , showing that it is the K segments that are important for membrane protection and that the segment has no role.
Our binding assay shows that the presence of a negative charge on the liposome lipid is important for K 2 binding (Figs. 1A and 3B). The somewhat weaker protection with the PC:PS liposomes is likely a function of the dehydrin binding more weakly to the PS (27). Although it is true that PA is a relatively low abundant lipid in normal plant membranes (58), there are several reasons as to why it may be important in the protective function of dehydrin from abiotic stresses (27). Freezing temperature treatment of Arabidopsis plants was shown to cause a 6-fold increase in PA content of the plasma membrane (58), and the authors speculated that during cold-stress the increased PA can cause an increased chance of membrane fusion. In vivo, fusion can be a problem during cold stress; it has been shown that plasma membranes can fuse at lower temperatures (59). A contributing factor may be that PA can form hexagonal II phase membranes, where the acyl tails point outwards and the dehydrated headgroups face inwards (60). Outward pointing tails would likely increase the chance of membrane fusion. Dehydration of the membrane headgroups is therefore linked with an increase in the probability of fusion (61). In addition, dehydration of headgroups has also been linked with an increase in T m (62,63). Previous work by Tompa et al. (64) showed that dehydrins are able to bind a large amount of water compared with a BSA control and keep this water in an unfrozen state even at subzero temperatures. A possible mechanism to explain how K 2 functions to both reduce T m and reduce fusion may be that dehydrins could bind to the periphery of the membrane, whereas their high level of hydrophilicity ensures that many water molecules are present to help hydrate the headgroups. The error bars represent the error in fitting the relaxation decay curves as described in CCPNMR (42). The diagram at the top of the figure shows the location of the various segments in the K 2 sequence.
In addition to the common protective effects shown by several cold and desiccation stress IDPs (i.e. reduced fusion and a lowering of the T m ), evidence points to a common structural motif that allows these proteins to bind to the membrane. Several stress IDPs, including K 2 discussed here and other dehydrins, gain ␣-helical structure in the presence of a membrane (28,55,57,65). The helical wheel projections for some of these stress IDPs are shown in Fig. 8. The amphipathic helices have a similar pattern of arrangement of amino acid types: hydrophobic residues are clustered on the face that points toward the membrane surface, positively charged residues, predominantly Lys, flank the hydrophobic face, whereas the negative residues (Asp and Glu) are located opposite the hydrophobic face. In all of the examples the proteins were found to preferentially interact with the negatively charged phosphates on lipids and/or negatively charged sulfates on SDS, involving Lys residues (28,55,57,65). These results also show that it is not necessarily true that any flexible, cationic peptide could provide membrane protection. First, it is Lys and not Arg that is found flanking the hydrophobic face. We speculate that this difference is due to the ability of Lys to snorkel on a membrane surface (66), although this proposal requires further experimentation for confirmation. Second, the use of polylysine polymers has been shown to increase the T m of a liposome rather than decrease it (67). The T m of DMPC:DMPA liposomes was measured using differential scanning calorimetry in the presence (solid line) and absence (dashed line) of K peptide. C p , specific heat capacity. Helical wheel representation of disordered proteins that bind negatively charged membranes. The beginning and end of the sequence are marked with a red N and a red C, respectively. The arrow indicates the hydrophobic moment ϽHϾ. Hydrophobic residues are colored yellow, basic residues are blue, acidic residues are red, serine and threonine are purple, asparagine and glutamine are pink, and alanine and glycine are gray. A, dehydrin K segment. B, Hsp12 helix 4. C, LEAM helix motif. D, CDeT11-24 lysine-rich sequence element.