The-K segment of maize DHN1 mediates binding to anionic phospholipid vesicles and concomitant structural changes

Dehydrins (DHNs, LEA D11 family) are a family of intrinsically unstructured plant proteins that accumulate in the late stages of seed development, and in vegetative tissues subjected to water deficit, salinity, low temperature or abscisic acid (ABA) treatment. We demonstrated previously that maize ( Zea mays ) DHNs bind preferentially to anionic phospholipid vesicles, this binding is accompanied by an increase in α –helicity of the protein, and adoption of α –helicity can be induced by SDS. All DHNs contain at least one “K-segment”, a lysine-rich 15 amino acid consensus sequence. The K-segment is predicted to form a class A2 amphipathic α –helix, a structural element known to interact with membranes and proteins. Here, three K-segment deletion proteins of maize DHN1 were produced. Lipid vesicle binding assays revealed that the K-segment is required for binding to anionic phospholipid vesicles, and adoption of α –helicity of the K-segment accounts for most of the conformational change of DHNs upon binding to anionic phospholipid vesicles or SDS. The adoption of structure may help stabilize cellular components including membranes under stress conditions.


Lipid Binding Activity of K-Segment Deletion Proteins
The lipid binding activity of maize DHN1 and the deletion proteins was assayed using gel filtration chromatography and western blot of column fractions as described previously (Koag et al., 2003). Based upon our previous finding that DHN1 has a strong interaction with anionic liposomes, a 1:1 mixture of PA (phosphatidic acid; anionic):PC (phosphatidylcholine; neutral) large unilamellar vesicles (LUV) and 100 % PC LUV were prepared using an extrusion method described previously (Hope et al., 1985) and the LUV suspensions were incubated with DHN1 proteins at a 15:1 mass ratio of phospholipid to protein for 3 h at 25 o C, as described previously (Koag et al., 2003). The intensity of the immunoblot of ΔK1 by anti K-segment antibody was lower compared with normal DHN1, and detection of ΔK2 and ΔK3 protein was even less apparent than that of ΔK1 protein. Therefore, antiserum raised against whole DHN1 (RAB17) protein of maize was used. As expected, there was no binding of any of the four proteins to LUV composed solely of the neutral phospholipid PC ( Figure 4A). In contrast, normal DHN1 and the ΔK1 and ΔK2 proteins bound LUVs containing the anionic phospholipid PA (fraction 8 and 9) ( Figure 4B). The ΔK3 protein, which is missing both of the K-segments, eluted at a specific fraction volume (fraction 18) of free protein on the precalibrated Superose 6 gel-filtration column ( Figure 4B) and therefore appears not to bind to LUV of any composition. Together, these results show that either K-segment in DHN1 is sufficient for LUV binding and the binding of DHN1 to LUV is dependent on the presence of at least one K-segment.
helical conformation when DHN1 binds to vesicles containing PA. This established that the binding of DHN1 to anionic phospholipid-rich lipid vesicles is associated with a structural change in the protein ( Figure 5A), as previously reported for DHN1 derived from maize kernels (Koag et al. 2003).
The ΔK1 and ΔK2 deletion proteins, each of which contains only one K-segment, also displayed a shift in CD values upon incubation with 1:1 PA:PC SUV, but not with 100% PC SUV ( Figure 5B and 5C). The CD spectrum of both ΔK proteins changed significantly in the far-UV range (190 to 210 nm). The CD spectrum of ΔK2 also exhibited a shift in the 210 to 230 nm range in the presence of 1:1 PA:PC SUV. These spectral shifts are consistent with adoption of secondary structure by the K1 and K2 segments, and some differences exist between the two.
The ΔK3 mutant, which is missing both K-segments, had no significant shift in the CD spectrum in the presence of 1:1 PA:PC SUV or 100% PC SUV ( Figure 5D). Therefore, similar to the lack of in vitro lipid vesicle binding described above, a conformational change seems also to depend on the K-segments. Analysis of the CD spectra of each DHN1 protein with anionic lipid vesicles using the structure analysis program CDSSTR and reference protein data set #7 showed that wild type DHN1 adopts 5.0 % α -helicity the and ΔK2 deletion protein 4.2 %, while ΔK1 displays much less α -helical structure (Whitmore & Wallace, 2004) We note also that if the shift in CD spectra of the ΔK1 and ΔK2 proteins with PA:PC SUV are added together, then the sum very nearly reconstitutes the shift in the CD spectrum of normal DHN1 under the same conditions ( Figure 5E). This is consistent with the overall conformational change of DHN1 in association with anionic vesicles being attributed to a gain of structure of both K-segments.
about double. The CD spectra of wild type DHN1 purified from E.coli in the presence of 10 mM SDS and PA:PC SUV are very similar to those of DHN1 purified from maize seed in their pattern and intensity (Koag et al., 2003).
The ΔK1 and ΔK2 deletion proteins, each of which contains only one K-segment, also underwent a shift in CD values upon incubation with 10mM SDS ( Figure 6B and C). The CD spectral changes of ΔK2 in the presence of 10mM SDS (7.5 % increase in α -helicity) were larger than those of ΔK1 (2% increase of α -helix). These spectral shifts indicate adoption of secondary structure by the K1 and K2 segments with some differences and additivity between the two.
The ΔK3 mutant, which is missing both K-segments, showed only a minor shift in the CD spectrum and no significant increase of α -helical structure in the presence of 10 mM SDS ( Figure   6D). If the shifts in CD spectra of the ΔK1 and ΔK2 proteins with SDS are added together, then the sum is nearly equal to the shift in the CD spectrum of normal DHN1 under the same conditions ( Figure 6E).
When the structural changes of each protein in the presence of 10mM SDS were compared with those upon liposome binding, even though there are some differences such as intensity in the ellipticity around 220nm, the effect of anionic phospholipid and SDS on the structural changes of DHN1 looks similar.
These observations are consistent with the idea that the overall conformational change of DHN1 in association with anionic vesicles and SDS is largely attributed to a gain of structure of both K-segments.

Secondary structure of synthetic K-segment peptide in the presence of SDS and liposomes
Because the CD data from each protein in the presence of anionic lipid vesicles and SDS imply that the structural changes of DHN1 protein are due to interaction of K-segment with lipid vesicles or SDS, we also examined the CD spectra measurement of the synthetic K-segment peptide in the presence of SDS and liposomes. No structural transition was induced when the Ksegment was incubated with 100% PC SUV ( Figure 7). However, when the K-segment was incubated with either PA:PC liposomes or 10 mM SDS a prominent induction of α-helix was observed. Similar to the wild type and mutant DHN1 proteins, SDS causes a more substantial structural change of the K-segment than do vesicles containing anionic phospholipids. Analysis

Discussion
To assess the structural and functional role of the highly conserved K-segment of DHNs, a deletion mutation approach was used. Three K-segment deletion mutants (ΔK1, ΔK2 and ΔK3) were produced from maize DHN1 using site-directed mutagenesis and expression in E.coli. As a result of these deletions, the mutant proteins are more acidic than the wild type protein (Table   S1). Mutant proteins were purified by a combination of ion exchange and hydrophobic interaction chromatography ( Figure S2A, B and C). Each fraction containing the protein in each purification step was examined by SDS-PAGE and western blotting as described in "Materials and Methods." The identity of each mutant protein was confirmed by MALDI-spectrometry and amino acid composition analysis ( Figure 3, Table S1 and S2). MALDI-spectrometry of each protein gave molecular weights within 0.05 to 0.1% of predicted values, which is well within an acceptable range of accuracy (Hillenkamp et al., 1991). Three satellite peaks from the ΔK2 mutant have additional mass of 98 Da each. This may be explained by the formation of H 2 SO 4 adducts (+98 Da) during the phenyl superose chromatographic step of purification. The intensity and numbers of SO 4 adducts can be increased by the amount of (NH 4 ) 2 SO 4 used for the elution step (Prinz et al., 1999). The low peak intensity of the ΔK3 mutant may be explained by inefficient ion formation resulting from deletion of the highly charged K-segments.
The properties of high temperature solubility, slow mobility in SDS-PAGE and high solubility in 50% (NH 4 ) 2 SO 4 are characteristic of the mutant proteins, much like wild type DHN1. However, migration of the ΔK2 and ΔK3 deletion proteins in SDS-PAGE is even slower than for wild type DHN1. This could be due to lower charge-charge association of the mutant proteins with SDS resulting in less electrostatic pulling force, or retention of a less flexible structure that retards migration relative to wild type DHN1 due to greater impedance with the polyacrylamide matrix.
It has been presumed that the K-segment is an essential unit relevant to the function of DHNs in response to dehydration-affiliated stresses (Close, 1996;Svenssen et al., 2002), and that the Ksegment can form an amphipathic α-helix, a biochemically important element involved in protein-protein and protein-lipid interaction (Dure, 1993;Epand et al., 1995). In our previous studies with wild type maize DHN1 purified from the maize seed, we demonstrated that this protein binds to anionic phospholipid rich vesicles, and that this binding is accompanied by a gain of structure of the protein (Koag et al., 2003). In the lipid-binding assay with the K-segment deletion mutant proteins, the binding was still observed for the ΔK1 and the ΔK2 single deletion mutants, while it was abolished for the ΔK3 double deletion mutant. These results indicate that the K-segment is necessary and sufficient for the binding of DHNs to anionic phospholipid vesicles ( Figure 4A and B). The binding of nonphosphorylated forms of the protein is consistent with previous results from other research group (Kovacs et al., 2008).
As shown in Figure 5, the induction from unstructured to partially α-helical structure upon incubation with PA-containing anionic phospholipid vesicles is consistent with the binding of wild type DHN1, ΔK1 and ΔK2 proteins to LUV. In contrast, there was no detectable transition in the CD spectrum of the ΔK3 double deletion protein with PA-containing vesicles, indicating that the change in structure detected by CD was dependent on the K-segments. Interestingly, when the changes in CD spectra of the ΔK1 and ΔK2 protein upon incubation of PA SUV are combined, then the combination reconstitutes the change in the CD spectra of wild type DHN by PA SUV ( Figure 5E). So it seems that the K-segments of DHN1 may have a cumulative effect on the overall structure of the protein and that the structural change of DHNs due to association with anionic phospholipids may be derived from conformational changes in K-segments.
Additionally, as described in the Results, the CD spectrum of maize DHN1 prepared from maize seeds in a phosphorylated state (Koag et al., 2003) is nearly identical to the CD spectrum of maize DHN1 prepared from E. coli in a non-phosphorylated state. This is in good agreement with two recent reports that phosphorylation has no or only a marginal effect on structural changes in vitro (Mouillon et al., 2008) and on membrane binding (Kovacs et al., 2008).
However, a recent study on the phosphoproteome showed evidence for mediation of phosphorylation effects through (i) conformational change coupled to phosphorylation, and (ii) modulation of charge-charge interaction between protein and other cellular targets (Kitchen et al., 2008). Therefore, although phosphorylation does not lead to conformational changes in DHNs detectable in the present study, it should not be ruled out that phosphorylation may affect lipid binding or can have some other functional role. In fact, prior studies of DHHs have shown that phosphorylation regulates nuclear localization (Jensen et al., 1998) and the binding of calcium ions (Heyen et al., 2002;Alsheikh et al., 2003;Brini et al., 2006 Taken together, even though physico-chemical conditions in plant cells under dehydrative stresses are very different from those typically used to characterize biochemical properties in vitro, the results of this study suggest that the K-segments of DHN1 may be responsible for binding to negatively charged membranes in vivo, and that such binding is causally related to the adoption of structure of DHN1. It then follows also that the membrane-bound, structured form of DHN1 is a relevant element of the response of plant cells to environmental stresses which typically evoke DHN1 production. While some LEA proteins (generally group 3 and 4) adopt predominantly α-helical structure upon drying or addition of SDS or TFE (Goyal et al., 2003;Shih et al., 2004;Tolleter et al., 2007), others (generally group 1 and 2) assume only partial (about 10%) α-helical structure upon addition of TFE, SDS, liposomes and other helical inducers (Soulages et al., 2003;Koag et al., 2003;Shih et al., 2004;Mouillon et al., 2006Mouillon et al., , 2008. In accordance with those previous results, CD spectra of full length DHN1 displayed low overall percentages of α-helical structure with SDS and liposomes, while a more prominent percentage of α-helical transition was observed with the K-segment alone. So, in line with one recent report (Mouillon et al., 2008), we refine the view of group 2 LEAs by clarifying that the structural change of full length DHN can be attributed significantly to the conserved K-segment, which can adopt sufficient structure to account for most of the α-helicity of the entire DHN1 protein (Fig. 7). To confirm the structural changes of DHN1 bound to SDS and anionic phospholipid vesicle, multi-dimensional NMR spectra including 1 H-15 N HSQC (heteronuclear single quantum coherence), known as a 'fingerprint spectrum' of proteins, of DHN1 alone and of DHN1 in the presence of SDS have been acquired (Koag, 2002). In the presence of SDS, nearly all of the cross-peaks for the Ksegment of DHN1 in 1 H-15 N HSQC disappeared or shifted. Compared to the spectrum of DHN1 alone, the positions of DHN1 cross-peaks that do not belong to the K-segment remained unchanged after binding to SDS, indicating that these cross-peaks represent amino acids that do not bind to anionic membrane vesicles or SDS and remain unfolded and mobile in the presence of them.
The plasma membrane of plants is known to be a major site of physical strain and damage under dehydrative stresses such as freezing and drought (Steponkus, 1998). It has been suggested that LEA proteins can contribute to membrane stabilization against those stresses (Artus et al., 1996;Danyluk et al., 1998;Koag et al., 2003;Tolleter et al., 2007;Kovacs et al., 2008 Recently, it has been shown that phosphatidic acid (PA), which acts as a second messenger in stress signaling pathways (osmotic and oxidative stress, ABA treatment, wounding, and pathogen attack) accumulates over several hours of stress (Frank et al., 2000;Katagiri et al., 2001). PA-enriched membranes are induced to form a hexagonal phase II membrane structure under acidic pH conditions or with elevated concentrations of Ca 2+ (Cullis et al., 1986). By binding to membranes through the K-segment amphipathic α-helix, DHNs may stabilize PAenriched membranes such that they are less prone to transition into hexagonal phase II structures under stress conditions.
The role of DHN1 stabilization of membranes may be quite general given that all membrane trafficking processes, such as transport between endoplasmic reticulum and Golgi, and DHNs bind to a range of metal ions with multiple tandem His residues (Svensson et al., 2000;Kruger et al., 2002) and can bind a large number of solute ions (Tompa et al., 2006). For example, the citrus CuCOR19 DHN protein binds Cu 2+ ions and scavenges reactive oxygen species, presumably to protect membrane from oxidative damage caused by water deficit (Hara et al., 2003(Hara et al., , 2005. Members of the acidic sub-group of DHNs (ERD14, ERD10 and COR47) were observed to increase Ca 2+ binding when phosphorylated in the poly-Ser motif (Alsheikh et al., 2003). It was also recognized that acidic residues of DHNs are involved in ion-binding (Alsheikh et al., 2005). Water deficit can trigger an increase in metal ion concentration, which may cause stress to the plant directly and indirectly. An acidic DHN was observed to interact with membranes under cold stress (Danyluk et al., 1998). Recently, the nonphosphorylated forms of two acidic DHNs (ERD10 and ERD14) were reported to bind membranes through a peripheral electrostatic interaction with phospholipid head groups (Kovacs et al., 2008). More extensive studies on lipid-and metal-binding activities clearly are needed to determine whether the two affinities work synergistically.
In summary, our results support the hypothesis that the K-segments of DHNs constitute the interface through which DHNs bind surface of membranes enriched in anionic phospholipids, and that, upon such binding, the K-segments adopt an α-helical conformation whereas most of the remainder of the protein remains unstructured. More extensive studies on the binding partners of DHNs and structural aspects of the disordered to structured transition will help refine our understanding of the role of these proteins under environmental stresses.

Construction of K-segment deletion proteins
Deletion mutant forms of maize DHN1 were constructed by introducing deletions of either or both K-segments. Oligonucleotide-directed mutagenesis was carried out following the manufacturer's protocol for the GeneEditor TM in vitro Site-Directed Mutagenesis System (Promega, Madison, WI, USA), briefly as follows. The pET19b-Dhn1 (maize) (Jepson & Close, 1995) plasmid was used as the starting material. Primers containing SacII and NheI restriction site were designed and used as mutagenic oligonucleotides ( Figure S1). Denatured pET19b-Dhn1 DNA was annealed to phosphorylated mutagenic oligonucleotides along with another oligonucleotide to modify the ampicillin resistance gene to confer resistance to Selection Antibiotic Mix TM . The mutant DNA strand was synthesized using T4 DNA polymerase and T4 DNA ligase. The reaction mixture was used to transform competent cells of E. coli strain BMH71-18 mutS (repair defective) to establish a culture containing the desired mutant plasmid.
Transformants were grown as a mixed culture in LB medium supplemented with 20 μ g/ml Selection Antibiotic Mix TM . Strain DH5α was then transformed with plasmid DNA isolated from the BMH71-18 mutS mixed culture and colonies were selected on LB agar containing 50 μg/ml ampicillin and 20 μg/ml Selection Antibiotic Mix TM . This second round of transformation separated wild type from mutant plasmids. Several individual colonies were then analyzed by restriction enzyme digestion of plasmid DNA and sequencing with a T7 primer.
To generate a mutant missing the first K-segment (ΔK1), a pair of primers including a SacII restriction site (5'-GACGACGGCATCCGCGGAAGGAGGAAG-3' and 5'-GGAGAAGCTGCCCCGCGGGCCACAAGGACG-3') was used ( Figure S1). For a mutant missing the second K-segment (ΔK2), a set of primers containing a NheI restriction enzyme site (5'-GAGGGCACCGGCTAGCAGAAAGGCATTATC-3' and 5'-GAGAAGCTGCCCGGCTAGCACTGAGCGCCC-3') was used ( Figure S1). For a double mutant lacking both K-segments (ΔK3), both primer pairs were used simultaneously ( Figure S1). sequence. For the expression of ΔK1, construct MK1-2 was used. For the ΔK2 mutant, three of three colonies (MK201, MK202 and MK203) selected from the second round transformation for site-directed mutagenesis were verified to include the NheI restriction site. MK202 was chosen to digest with NheI restriction enzyme and ligated with T4 ligase. Then the ligated construct was used to transform the E. coli DH5α competent cells. Three selected clones (MK2-1, 2-2 and 2-3) were verified to have the correct sequence. For the expression of ΔK2, construct MK2-2 was used.
For the ΔK3 mutant, two of five clones (MK301and MK304) were proven to contain both the SacII and NheI restriction sites by sequencing. MK301 was digested to with SacII and ligated with T4 ligase. Then the ligation reaction was used for transformation of E. coli DH5α competent cells to identify a ΔK1 deletion. Four selected colonies (MK31-1, 31-2, 31-3 and 31-4) contained the correct sequence. Subsequently, MK31-4 was digested by NheI and re-ligated.
The re-ligated construct was used for transformation of strain DH5α to screen for a double deletion mutant. Two of the chosen colonies (MK32-1 and 32-2) were confirmed to possess the correct sequence. The construct MK32-2 was used for expression of ΔK2.
Two amino acids of ΔK1 and ΔK3 at positions 88 and 89 were changed from Met/Gly to Ile/Arg as a consequence of the site directed mutagenesis as shown in Figure 1. In addition, the molecular mass and isoelectric points of each protein were altered as shown in Table S1.

Protein Expression in Escherichia coli
BL21-CodonPlus (DE3)-RIL competent cells (Stratagene, La Jolla, CA) were transformed with mutant constructs MK 1-2, MK 2-2 and MK 32-2 according to manufacturer protocol. The transformation mixture was grown on LB agar plates (100 µg/ml ampicillin, 20 µg/ml chloramphenicol) to select for transformants. LB medium containing 100 µg/ml ampicillin, 20 µg/ml chloramphenicol was inoculated with a fresh overnight culture of the E. coli expression strain with each DHN protein and grown to a cell density of 0.5 OD at 600 nm. IPTG was added to a concentration of 1 mM to induce protein expression and the culture was grown at 250 rpm, 37°C for an additional two hours to a cell density of 1.0 OD at 600 nm. Cells were harvested by centrifugation at 6,000 x g for 15 min at 4°C (Beckman J2-21, JS7.5 rotor). Cell pellets were stored at -80°C.

Purification of Deletion Mutant Proteins
Cell pellets were thawed on ice and resuspended in ice cold lysis buffer (25 mM MES, pH 6.0, 20 mM NaCl, 1mM PMSF). Cells were lysed by passing twice through a French pressure cell (Thermo Fisher Scientific, Inc. Waltham, MA) with an internal pressure of 25,000 psi.

Immunoblotting
Immunoblotting was used for analysis of fractions at each purification step and for the lipid binding assay. For wild type DHN1, polyclonal antibody against the conserved K-segment was used as described previously (Close et al., 1993). To detect K-segment deletion proteins, anti-RAB17 antiserum (a gift of M. Pages, Barcelona, Spain) was used at a 1:500 dilution (Jensen et al., 1998).

Matrix assisted laser desorption ionization time of flight (MALDI-TOF) spectrometry
Mass spectra of normal and deletion mutant proteins were determined by matrix-assisted   The primary amino acid sequences of wild type DHN1 and each K-segment mutant are shown.
There are two altered amino acids in the ΔK1 and ΔK3 proteins (italics). Hyphens indicate deleted positions. The wild type DHN1, ΔK1, ΔK2, and ΔK3 proteins were purified as described in the text. Each