Structure of a Conserved Golgi Complex-targeting Signal in Coronavirus Envelope Proteins

Background: Coronavirus envelope (CoV E) proteins have a predicted β-coil-β motif reported to target the Golgi complex. Results: This conserved domain forms β-structure on its own but is α-helical in the context of full-length SARS-CoV E protein. Conclusion: This domain is potentially involved in large conformational transitions. Significance: This is the first structural data of the extramembrane domain of any coronavirus E protein.

The main coronavirus structural proteins are S (spike), E (envelope), M (membrane), and N (nucleocapsid), where S, E, and M are integral membrane proteins. E proteins are ϳ100residue-long polypeptides that are minor components in virions but are abundantly expressed inside infected cells (13). They have a short hydrophilic N terminus, at least one predicted terminal transmembrane (TM) domain, and a less hydrophobic C-terminal tail.
Co-expression of E and M proteins is sufficient for formation and release of virus-like particles (14 -19), and E proteins have been proposed to participate in inducing membrane curvature or in the scission of particles (20). Mutations at their C-terminal extramembrane domain impair viral assembly and maturation in the murine hepatitis virus (20). In the transmissible gastroenteritis virus, the absence of E protein resulted in a blockade of virus trafficking in the secretory pathway and prevention of virus maturation (21,22). In the case of the severe acute respiratory syndrome coronavirus (SARS-CoV), viruses lacking the E gene showed attenuation and did not grow in the central nervous system (23,24), whereas stress response genes were up-regulated and cell apoptosis increased (25). These results suggested a role of E protein in both tissue tropism and pathogenicity, where modulation of stress responses contributes to viral attenuation. In fact, SARS-CoV ⌬E attenuated viruses constitute promising vaccine candidates (26 -28). Thus, although E proteins are not absolutely essential for in vitro or in vivo coronavirus replication, their absence is clearly deleterious.
Immunofluorescence experiments have shown that SARS-CoV E has a cytoplasmically oriented C terminus and a lumenal N terminus (i.e. an N exo C cyto orientation) (13), which is consistent with the presence of a single TM domain. SARS-CoV E protein and other CoV E proteins have channel activity in synthetic membranes (29 -31). This channel activity is mediated by formation of pentameric oligomers (32)(33)(34) and is only very mildly selective for cations (35). The only available structural data for CoV E proteins, obtained using synthetic TM peptides, is derived from the channel-forming TM domain in SARS-CoV E (32,34,36). No structural data are available for the predicted N-or C-terminal extramembrane domains despite the latter being critical for viral assembly (37,38), although some results have been obtained using shorter synthetic peptides encom-passing this domain (39). The C-terminal domain of E proteins contains a totally conserved proline residue, which in ␤and ␥-coronaviruses is at the center of a predicted ␤-coil-␤ motif (Fig. 1A), reminiscent of viral internal fusion peptides (40 -42). In SARS-CoV E, this motif was found to be responsible for redirecting a plasma membrane protein to the Golgi region. Conversely, mutations designed to increase its ␣-helical propensity disrupted localization to membranes (43).
The C-terminal tail of E proteins is also important for its interaction with the C-terminal domain of M protein (18,44,45) at the cytoplasmic side of the endoplasmic reticulum-Golgi intermediate compartment, the budding compartment of the host cell. These interactions are the major drivers for envelope formation (46). The C-terminal tail of SARS-CoV E protein also interacts with the cellular protein PALS1 (47), which is thus depleted from the tight junctions in epithelial cells. Finally, SARS-CoV E interacts with the seven-domain (48) SARS nonstructural protein 3 (Nsp3) (49).
Thus, it is crucial to determine the structure of the C-terminal predicted extramembrane domain of E proteins, especially in the context of a large construct that includes the TM domain. Until now, these structural studies have been hampered by problems in expression, purification, and stabilization of E proteins. We report herein the first detailed structure of a truncated form of the SARS-CoV E monomer that includes both its TM domain and its predicted ␤-coil-␤ motif at the C-terminal tail.

MATERIALS AND METHODS
SARS CoV E Protein Constructs-Full-length SARS CoV E and a truncated version were used in this work. Full-length SARS-CoV E protein (E FL ) was cloned into pTBMalE plasmid downstream of the MBP fusion tag and tobacco etch virus cleavage sequence. The plasmid was transformed into Escherichia coli strain BL21(DE3) codon plus for protein expression. The truncated form, E TR , was cloned into pNIC28-Bsa4 with an N-terminal His 6 tag followed by a tobacco etch virus cleavage sequence. The plasmid was transformed into E. coli strain BL21(DE3) Rosetta T1R for protein expression. In both constructs, all three native cysteines (Cys-40, Cys-43, and Cys-44) were mutated into alanines. In addition, two E FL mutants previously described by Cohen et al. (43) were prepared by sitedirected mutagenesis: (i) P54A (E P54A ) and (ii) V56A/Y57A/ V58A/Y59A (E 4ALA ).
Protein Expression and Purification-Non-labeled E protein was produced by growing the culture in Terrific broth medium at 37°C until the culture density reached an A 600 of 2. Protein expression was induced by adding 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside and growing the culture overnight at 18°C. The cells were harvested by centrifugation at 7,500 ϫ g and stored at Ϫ80°C.
Stable isotope-labeled E protein was produced by growing the culture in LB medium at 37°C. When the culture density reached an A 600 of 0.7, the medium was exchanged to M9 minimal medium at 25% of the initial volume to achieve a high density culture, as described previously (50). The M9 medium was appropriately supplemented with 15 N-NH 4 Cl and [ 13 C]glucose (Cambridge Isotope Laboratories) to produce 15 N-labeled and 15 N/ 13 C-labeled protein. Cultures were further grown for 1 h before inducing protein expression with 0.5 mM isopropyl 1-thio-␤-Dgalactopyranoside at 18°C. After 6 h, cells were harvested as described above and stored at Ϫ80°C.
Gel Electrophoresis-Standard SDS-PAGE was performed in 13.5% Tris-glycine gel with TGS running buffer and stained with Coomassie Blue G-250. SDS-NuPAGE was performed in 4 -12% NuPAGE BisTris gel (Invitrogen) with NuPAGE MES SDS running buffer and stained with SimplyBlue TM SafeStain (Invitrogen) according to the manufacturer's protocol.
To perform electrophoresis in the presence of perfluorooctanoic acid (PFO) detergent, we modified Invitrogen's SDS-NuPAGE protocol by replacing SDS with PFO. Lyophilized peptide was dissolved in sample buffer containing 4% PFO and heated at 65°C for 5 min prior to loading. The gel was run at 80 V for 2-3 h with MES running buffer containing 0.5% PFO.
Blue native PAGE was performed as described previously (51). Lyophilized peptide was solubilized (0.1 mM) in sample buffer containing 25 mM SDS and either 25, 50, or 100 mM DPC. Aquaporin Z in 20 mM SDS (heated at 65°C for 10 min) was included as an additional molecular weight marker.
Fourier Transform Infrared Spectroscopy-Sample preparation, data collection, and H/D exchange were performed essentially as described (32) on a Nicolet Nexus spectrometer (Madison, WI). The peptides were incorporated in multilamellar liposomes by dissolving a dry mixture of 1,2-dimyristoyl-snglycero-3-phosphate (DMPC; Avanti Polar Lipids) and lyophilized peptide in HFIP at a 50:1 molar ratio. Fourier self-deconvolution was performed for some spectra using the following parameters: full width at half height, 20 cm Ϫ1 ; narrowing factor, k ϭ 1.5) (52).
Analytical Ultracentrifugation-Sedimentation equilibrium experiments were performed using a Beckman XL-I analytical ultracentrifuge at 20°C (53) and monitored by measuring the absorbance at 280 nm. Lyophilized E TR peptides were dissolved at A 280 of 0.3, 0.5, and 0.8 (12-mm path length cell) in 20 mM sodium phosphate, pH 5.5; 50 mM NaCl; and for detergent either 5 mM C14SB, 100 mM DPC, or 12.5, 25, or 50 mM SDS. To match the density of the SDS-DPC mixture, D 2 O was added at 61.6, 65, and 72.4%, respectively, to each SDS concentration. The samples were centrifuged in six-channel charcoal-filled Epon centerpieces using quartz windows. A radial distribution profile was acquired after sufficient time to reach equilibrium, as tested by HeteroAnalysis. The data were processed and fitted to several monomer/n-mer models in SEDFIT and SEDPHAT (54).
Circular Dichroism-CD data were acquired on a Chirascan CD spectrometer (Applied Photophysics) using a 0.2-mm quartz cuvette (Hellma). E TR peptide samples were dissolved at 1 mg/ml in 20 mM sodium phosphate, 50 mM NaCl, pH 5.5, 100 mM DPC, with or without 50 mM SDS. CD spectra were acquired from 180 to 260 nm with a 1-nm spectral bandwidth and 3 replicates/spectra. Data points with excessive absorbance were excluded. After baseline subtraction, the data were analyzed in Dichroweb (55) by using the CDSSTR method (56) and the SMP180 reference set (57).
NMR Sample Preparation-Approximately 1.2 mg of lyophilized E TR protein was solubilized in 100 l of methanol and dried under a dry stream of N 2 gas, resulting in a thin protein film deposit. The tube was placed in a vacuum lyophilizer overnight to remove any residual methanol. The thin protein film was then solubilized with sample buffer containing 20 mM sodium phosphate, pH 5.5, 50 mM NaCl, and 50 mM SDS. The sample was vortexed and sonicated several times until a clear solution was obtained, indicating protein reconstitution into detergent micelles.
For paramagnetic relaxation enhancement (PRE) experiments, a single point mutation (S60C) was introduced into E TR by site-directed mutagenesis using appropriate sets of primers. Expression and purification protocol of the E TR -S60C mutant was the same as that of E TR protein. For labeling, 0.3 mM 15 Nlabeled E TR -S60C was dissolved in 20 mM sodium phosphate, 50 mM NaCl, 200 mM SDS, and 0.8 mM DTT at pH 5.5 and split into two equal portions for parallel labeling with (1-oxyl-2,2,5,5-tetramethyl-⌬ 3 -pyrroline-3-methyl) methanethiosulfonate (MTSSL) (TorontoResearch Chemicals Inc.) and a diamagnetic analog of MTSSL: (1-acetyl-2,2,5,5-tetramethyl-⌬ 3 -pyrroline-3-methyl) methanethiosulfonate (dMTSSL; Toronto Research Chemicals Inc.). A 10-fold molar excess of both reagents was added from 75 mM stocks in methanol. The sample was vortexed for 30 min at high speed and incubated overnight at room temperature. A centrifugal filter unit (10,000 molecular weight cut-off; Millipore Corp.) was used to remove excess of both reagents. Labeled samples were washed four times by concentrating to 100 l. After a fourth wash, the sample was concentrated to 180 l for NMR measurements.
Partial alignment of the E TR protein-micelle complexes relative to magnetic field was obtained by using stretched polyacrylamide hydrogels (58,59). A 7% polyacrylamide gel was polymerized in a gel chamber of 5.4-mm inner diameter. After complete polymerization, gels were washed in H 2 O overnight and then twice with sample buffer containing 20 mM sodium phosphate and 50 mM NaCl at pH 5.5. The gels were then completely dried at room temperature. The protein solution containing E TR /SDS was soaked into the dried gels over 2 days to ensure complete rehydration. The hydrated 7% gel was then compressed into a 4.2-mm inner diameter open-ended tube using the gel press assembly (New Era Enterprise, Inc.).
NMR Spectroscopy-NMR experiments were performed at 308 K using an Avance-II 700 NMR spectrometer with cryogenic probe. Sodium 2,2-dimethyl-2-silapentane-5-sulfonate was used as the internal reference for 1 H nuclei. The chemical shifts of 13 C and 15 N nuclei were calculated from the 1 H chemical shifts. The NMR data were processed using TopSpin version 3.1 and analyzed using CARA. Sequence-specific assignment of backbone 1 H N , 15 N, 13 CЈ, and 13 C ␣ was achieved by using two-dimensional 1 H-15 N TROSY-HSQC, three-dimensional HNCO, HN(CA)CO, HNCA, HN(CO)CA, and HNCACB experiments on a 15 N/ 13 C-labeled E TR protein.
CSP ϭ ͱ⌬␦H 2 ϩ ͑0.23 ϫ ⌬␦N͒ 2 (Eq. 1) Structure Calculation-NOE distance restraints were obtained from 15 N NOESY-HSQC (mixing time 80, 100, and 150 ms) and 13 C NOESY-HSQC (mixing time 120 ms) spectra, respectively. Backbone dihedral angle restraints ( and ) were derived from 13 CЈ, 13 C ␣ , 13 C ␤ , 1 H ␣ , and 1 H ␤ chemical shift values using TALOSϩ (61). The short range and medium range NOE connectivities were used to establish the sequence-specific 1 H NMR assignment and to identify elements of the regular secondary structure. Hydrogen bonds were derived from the H/D exchange experiment and NOE connectivity.
Distance restraints were obtained from the measured PRE effect using the procedures described previously (62)(63)(64). The intensities of cross-peaks in the MTSSL (Ip) and dMTSSL (Id) were calculated in CARA. The correlation time was set to 10 ns. The ratios of intensities (Ip/Id) were normalized against a set of the eight highest Ip/Id ratios, which were assumed to belong to peaks unaffected by PRE. For peaks with ratios below 0.15, no lower distance restraints were used, whereas upper restraints were set to 15 Å. For peaks with ratios above 0.9, only upper restraints of 25 Å were utilized. For peaks with ratios between 0.15 and 0.9, upper and lower distance restraints were generated using Ϯ3 Å margins.
Structure calculations were performed using CYANA version 3.0 (65, 66) and visualized using PyMOL (Schroedinger LLC, New York). CNS 1.3 (67,68) was used to refine the structure using the standard simulated annealing protocol. All of the restraints used in the calculations to obtain a total of 15 structures and all of the structure statistics are summarized in Table 1.
Single Channel Activity Measurement-Ion channel activity of E TR was measured by using Nanion Port-a-Patch. Briefly, giant unilamellar vesicles of 1,2-diphytanoyl-sn-glycero-3phosphocholine containing 10% cholesterol were prepared in 1 M sorbitol using Nanion Vesicle Prep Pro, following the manufacturer's protocol. Into 100 l of a giant unilamellar vesicle solution, 0.5 l of E TR peptide solution in ethanol (0.1 mg/ml) was added and incubated for 1 h at room temperature. E TR -containing giant unilamellar vesicles were subsequently deposited onto 6 -8 megaohm NPC-1 chips (Nanion). Conductance was measured under symmetrical buffer conditions (10 mM HEPES, 500 mM NaCl, pH 5.5).
Surface Plasmon Resonance-The nsp3a sequence was subcloned from pcDNA3(ϩ) into pET28b upstream of a C-terminal His 6 tag for expression in E. coli. The protein was expressed and purified as described previously by Serrano et al. (69). A negative control, consisting of C-terminal peptide from the small hydrophobic (SH) protein of human respiratory syncytial virus (RSV SH(45-65)), was synthesized by standard solid phase and purified by reverse-phase HPLC. Surface plasmon resonance measurements were performed on a Biacore 3000 system (GE Healthcare) using 10 mM phosphate buffer at pH 6.5, 100 mM NaCl, 3 mM EDTA, 0.05% n-octyl-␤-D-glucopyranoside, and 0.27% C14-betaine at 25°C. E TR was immobilized to 15,000 RU onto a research grade CM5 sensor chip (GE Healthcare) using standard amine-coupling chemistry. Briefly, a buffer-equilibrated carboxymethyl dextran surface was activated with a 10-min injection of a 1:1 mixture of 0.05 M N-hydroxysuccinimide and 0.2 M N-ethyl-N- [3-(diethylamino)propyl]carbodiimide. E TR peptides dissolved in 10 mM sodium acetate, 15 mM DPC (pH 5.0) were passed over the activated surface to achieve the desired response level. Another 10-min injection of 1 M ethanolamine-HCl (pH 8.5) was used to deactivate the surface and remove any non-covalently bound protein. Kinetic measurements of immobilized E TR association with Nsp3a and SH(45-65) (49 nM to 25 M in 10 2-fold serial dilutions) were performed with a 1-min association phase and 5-min dissociation phase at a 30 l/min flow rate. Each concentration was tested in duplicate. No regeneration was necessary because all complexes dissociated within the monitored time. Sensorgrams were double-referenced (70) and globally fit to a steadystate model to obtain affinity values.

RESULTS AND DISCUSSION
Expression and Purification of SARS-CoV E Protein-Initially, we successfully expressed and purified full-length SARS-CoV E protein (E FL , Fig. 1B) by using either a ␤-barrel (71) or MBP as fusion tags. However, the yield of pure protein was low due to the presence of truncations (not shown). Nevertheless, the pure sample obtained (Fig. 1, C and D) was sufficient for backbone assignment in SDS micelles, although not to produce a three-dimensional model. Therefore, a series of hexahistidine-tagged SARS-CoV E constructs were screened to obtain an expressing and well behaved sample. The best construct, encompassing residues 8 -65, was successfully expressed in E. coli and purified by affinity chromatography in milligram amounts without any enzymatic cleavage steps. This truncated construct (E TR ; see Fig. 1B) has an N-terminal His tag and a 16-residue linker that connects it to residues 8 -65. The purified peptide appeared in MALDI-TOF MS as a single-charged peak at 8,997 Da and a double-charged peak at 4,512 Da, consistent with the calculated molecular mass of E TR , 8,995 Da, and a small proportion of larger (dimer to pentamer) oligomers (Fig.  1E). After HPLC, the E TR monomer (9 kDa) showed anomalous migration in standard SDS-PAGE (Fig. 1F), as reported previously for the full-length protein (31).
Identification of the Membrane-embedded Region of E TR by NMR-Screening of reconstitution conditions identified SDS as the best environment to achieve good peak dispersion in both dimensions ( Fig. 2A). Comparison of 1 H-15 N TROSY-HSQC spectra in water and in 99% D 2 O (Fig. 2, A and B) identified a stretch of 21 residues, from Leu-18 to Leu-39, protected from H/D exchange. The same residues were protected from H/D exchange in E FL (not shown), which indicates that both E TR and E FL have only one TM domain.
The topology of E TR was also delineated by introducing the paramagnetic probes 5-DSA and 16-DSA. Upon the addition of 5-DSA, a slight intensity reduction was observed around residues 11-20 and 40 -55 (Fig. 2C), suggesting that these residues are located at or near the micelle surface. The addition of 16-DSA caused pronounced intensity reduction in the stretch of TM residues 19 -40 and periodically in the stretch 48 -61 (Fig. 2D). Together with the H/D exchange data (Fig. 2, A and  B), these results again demonstrate the presence of a single TM domain and suggest that the stretch of residues 55-65 corresponds to a domain bound to the surface of the micelle. Finally, a 1 H-15 N steady-state heteronuclear NOE experiment (Fig. 2E) showed that the protein forms a well folded structure, with high 1 H/ 15 N NOE values. However, more dynamic regions (lower 1 H/ 15 N NOE values) can be observed (i) at the N terminus and (ii) at residues 46 -55, which connects the TM domain to the proposed membrane-bound domain. The latter is predicted to be ␣-helical (Fig. 2F). Table 1 were used in a calculation to obtain a total of 15 structures, with a root mean square deviation of 0.27 Ϯ 0.11 and 0.70 Ϯ 0.13 Å for backbone and all heavy atoms, respectively (Fig. 3A). A longer ␣-helix (residues 15-45) encompasses the TM domain, which is connected to another shorter C-terminal ␣-helix (residues 55-65) by a flexible (see Fig. 2E) linker domain (residues 46 -54), forming an L-shape. The short extramembrane helix may be partially bound to the micelle surface, as suggested by the pattern of intensity attenuation of paramagnetic reagents (see Fig. 2, B and C). The most affected residues, which would face the micelle surface, are Val-52, Thr-55, Tyr-59, and Lys-63 (Fig. 3B).

NMR Structure Determination of E TR in SDS Micelles; Structure Calculation-The restraints shown in
The TM ␣-helix has a slight bend at residues 26 -30, consistent with previous results obtained for the synthetic TM The cysteine residues are underlined, the conserved proline is highlighted (gray), and the four residues mutated to alanine in the E 4ALA mutant (see "Materials and Methods") are shown in red. For these four proteins, the prediction of secondary structure is shown below in a color code, with the TM domain indicated as a black line; B, proteins used in the present work: a His-tagged construct (E TR ) encompassing residues 8 -65 (boldface type, underlined), and full-length SARS-CoV E (E FL ). In E FL , the fragment SNA results from the cleavage of the tag. In both proteins, the native cysteines were mutated to alanine (C40A, C43A, and C44A; see asterisks); C and D, MALDI-TOF MS spectra (C) and standard SDS-PAGE (D) of pure E FL with the species labeled; E and F, same as for purified E TR ; the identities of various single-and double-charged species are indicated. The calculated mass of E TR is 8,995 Da. domain in DPC micelles (34). A kink near this location is suggested by the short distance (2.0 Ϯ 0.1 Å) between Thr-30 H␥ and the carbonyl oxygen at Phe-26, in the range of a hydrogen bond (Fig. 3C).
Effect of Truncation and Environment on E TR Secondary Structure-To assess the effect of the truncation and the presence of a His tag on the E TR secondary structure, E TR and E FL were compared. The possible effect of the reconstitution environment was also determined.

CD and IR Spectra of E TR and E FL in Detergent and Lipid
Membranes-The CD spectra of E TR in DPC, SDS, and mixed (1:2 molar ratio) SDS/DPC micelles are almost superimposable (Fig. 4A) with minima at 209 and 222 nm. Also, these data are entirely consistent with the CD spectra of E FL , which was predominantly ␣-helical in both SDS and DPC micelles (71). When reconstituted in DMPC membranes, E TR and E FL produced an almost identical spectrum, with an amide I band centered at 1,655 cm Ϫ1 (Fig. 4B) characteristic of a predominantly ␣-helical conformation. Overall, these data show that both E TR and E FL (i) are predominantly ␣-helical and (ii) have a secondary structure that is not significantly affected by the reconstitution environment, supporting the relevance of the E TR structure (Fig. 3). 13 C␣ Chemical Shifts-13 C␣ chemical shifts are highly correlated with secondary structure (72,73). Comparison of 13 C␣ chemical shifts of E TR and E FL in SDS (Fig. 4C) shows that E TR shifts (red) are almost identical to their counterpart residues (residues 8 -65) in E FL (blue). In E FL , the shifts for the last 10 residues (residues 66 -76) are predicted to correspond to random coil, which is consistent with results obtained previously with a synthetic peptide spanning residues 59 -76 (39). This peptide produced a broad amide I band in the IR spectrum centered around 1,650 cm Ϫ1 and immediately experienced complete H/D exchange.
When the 13 C shifts for E TR were compared for SDS and an SDS/DPC mixture (1:4 molar ratio), only the N-terminal region (residues 7-15) showed minor differences. Finally, a comparison between E FL and E TR in an SDS/DPC mixture (1:4 molar ratio) also revealed an almost identical pattern in the C-terminal tail (not shown), although the spectral resolution for E FL was reduced in other regions. Overall, these results are consistent with those shown above (Fig. 4, A and B), indicating that the detergent used, the tag, and the truncation has a minimal on E TR secondary structure.
Ion Channel Activity of E TR and Its Inhibition by HMA-Purified E FL has channel activity (71), which is inhibited by the drug HMA (30). An I/V plot obtained in a symmetrical 0.5 M NaCl experiment for purified E TR in 1,2-diphytanoyl-snglycero-3-phosphocholine (Fig. 5A) was used to determine a conductance of 0.39 Ϯ 0.02 nanosiemens. For comparison, synthetic full-length SARS-CoV E and E TM (residues 7-38) produced single channel conductances of 0.19 Ϯ 0.06 picosiemens and 0.18 Ϯ 0.12 nanosiemens in 1 M NaCl (35), although the values were higher in 1 M KCl, with 0.37 Ϯ 0.16 and 0.31 Ϯ 0.12 for full-length SARS-CoV E and E TM , respectively. The lower conductance observed in synthetic samples may be due to extraneous modifications or impurities resulting from exposure to harsh chemicals. Representative traces of E TR channel activity (Fig. 5B) and complete inhibition after the addition of 10 M HMA (Fig. 5C) suggest that E TR is entirely functional.
To determine the binding site for HMA, we measured the differences in CSP values before and after the addition of the drug (Fig. 5, D and E). In SDS, the average CSP value was low, 0.006 ppm (Fig. 5F), even at an HMA/E TR 10:1 molar ratio, suggesting no significant binding. However, the same panel shows that the addition of HMA to E TR in mixed SDS/DPC (1:4 molar ratio) micelles produced an average CSP value of 0.013 ppm, even at an HMA/E TR 2:1 molar ratio.
Except for Asn-64 and Leu-65, the residues that showed significant CSP (CSP Ն 0.025 ppm) clustered near the membrane interface regions of the TM domain (Fig. 5F, see arrows). Near the N-terminal side of the TM domain, the most affected were Glu-8, Gly-10, Thr-11, Val-14, Asn-15, and Ser-16, the latter two consistent with observations made on the TM channel in DPC (34). At the C-terminal end of the TM, Leu-37 was the most affected, suggesting that the interaction of HMA at H⑀ of Arg-38 reported previously (34) may have been an artifact due to the use of a TM peptide.
Oligomeric State of E TR Gel Electrophoresis-The localized changes in chemical shifts observed after HMA addition to E TR in mixed DPC/SDS micelles (Fig. 5E), but not in SDS micelles (Fig. 5D), suggest that binding sites for HMA may have been induced after DPC addi-   tion to SDS, possibly through E TR increased oligomerization and population of the pentameric form. Thus, we examined the oligomeric state of E TR using gel electrophoresis performed in the presence of SDS or an SDS/DPC mixture. In contrast with the results obtained in normal SDS-PAGE (Fig. 1), we used NuPAGE, where E TR migrated with its expected monomer molecular weight, as a single band at 9 kDa (Fig. 6A). No oligomerization is consistent with the lack of interaction between HMA and E TR observed in SDS micelles.
To test the effect of DPC, in a blue native polyacrylamide gel, a constant SDS concentration (25 mM) was titrated with increasing DPC (Fig. 6B), from 1:1 to a 1:4 molar ratio. Bands  . Equivalence in secondary structure of E TR and E FL . A, CD spectra of E TR in DPC (black), 1:2 molar ratio SDS/DPC mixture (blue), and SDS (red). B, infrared amide I band of E TR (red) and E FL (blue) in DMPC lipid bilayers and their respective Fourier self-deconvolved spectra (dotted lines). C, comparison of secondary 13 C␣ chemical shifts (deviation from tabulated random coil 13 C␣ chemical shift values) for E TR (red dots) and E FL (blue dots) in SDS micelles and for E TR in (1:4 molar ratio) mixed SDS/DPC micelles (white dots). For the latter, Pro-54 and Thr-55 were excluded from the analysis due to significant line broadening; Arg-38 was excluded from the analysis due to the peak overlapping. consistent with dimers and trimers were observed at 1:1 and 1:2 molar ratios, whereas tetramers and pentamers were observed at a 1:4 molar ratio. This increasing oligomerization is again consistent with the binding of HMA observed in the presence of DPC.
Sedimentation Equilibrium, SDS/DPC Micelles-For a sample corresponding to a 1:4 molar ratio of SDS/DPC, the equilibrium sedimentation data (Fig. 7A) could be fitted to several oligomeric models, from trimers to heptamers (Fig. 7B). Similar ambiguous results were obtained for a 1:2 molar ratio SDS/DPC mixture (Fig. 7C), suggesting that E TR forms a mixture of oligomers in SDS/DPC micelles, consistent with electrophoresis results (Fig. 6). The fit of the traces (Fig. 7A) to a monomerpentamer model produced an apparent K a of 1.25 ϫ 10 15 M Ϫ4 (i.e. 6 ϫ 10 3 M Ϫ1 ) or a K d for monomer-monomer interaction of 0.17 mM and a molar fraction standard free energy change (74), ⌬G x 0 , of approximately Ϫ4 kcal⅐mol Ϫ1 . Overall, the above results show that higher oligomeric states, including pentamers, are observed when increasing the DPC concentration, consistent with the larger shifts observed after adding HMA to mixed DPC/SDS micelles.
E TR Forms Pentamers in C14 Betaine and PFO-We have shown previously that SARS-CoV E TM domain and E FL form pentamers (36,71). When E TR was solubilized in C14 betaine detergent (Fig. 7D), data could be optimally fitted to a monomer-pentamer model (K a ϭ 10 16 M Ϫ4 ) with a significantly lower 2 than that obtained in SDS/DPC. Also, the mobility of E TR in PFO-NuPAGE (Fig. 7E) corresponds to a molecular mass of 45 kDa (i.e. a pentameric form). Thus, E-TM, E TR , and E FL show a similar oligomerization behavior, being able to form pentameric channels that are inhibited by HMA.
Interaction of E TR with Nsp3a-The interaction of Nsp3 with SARS-CoV E protein, thought to be related to E ubiquitination, was previously mapped to an N-terminal acidic domain, Nsp3a (49). Although the site of interaction with E is not known, it is likely to involve the C-terminal tail because it is the largest extramembrane domain. Thus, to test that the conformation of  E TR is structurally equivalent to the corresponding sequence in E FL , we tested its ability to interact with Nsp3a (49) using surface plasmon resonance.
As a negative control, the mock peptide RSV SH(45-65) (see "Materials and Methods") was used to bind to immobilized E TR in detergent (see "Materials and Methods") (Fig. 8A). Binding and unbinding of RSV SH(45-65) was very slow (Ͼ1 min), and the data could not be fitted to any model, suggesting a nonspecific interaction.
In contrast, rapid and reversible interaction was observed using purified Nsp3a (Fig. 8B). These data could be fitted with a stoichiometry of 1:1, although the end point could only be esti-mated. From that model, affinity was determined as 1.6 mM, although we note that binding is already evident at concentrations as small as 1-10 M (see Fig. 8B, inset).
The residues involved in the interaction between Nsp3a and E TR were identified by the differences in CSP values before and after the addition of Nsp3a to E TR in SDS/DPC micelles (1:4 molar ratio). Large chemical shift changes at Leu-39, Val-49, and Leu-65 (Fig. 8C) indicated a potential binding site at the C-terminal tail, whereas the central TM region did not experience any change. Overall, these data suggest that E TR has a structure in SDS/DPC micelles that is similar to the native fold of E FL in biological membranes. Structural Discordance at the C-terminal Tail-Having determined the suitability of the construct used by comparison with the full-length protein and shown that the nature of the detergent did not affect the secondary structure significantly, we investigated the structure of the C-terminal extramembrane domain of E TR and its apparent contrast with secondary structure predictions.
The secondary structure predictions clearly suggest the presence of a ␤-turn-␤ motif in the C-terminal region, not only in SARS-CoV E but also in other representative sequences of other coronaviruses (see Fig. 1A). Indeed, we have reported previously that the synthetic peptide E 46 -60 , which encompasses this putative ␤-hairpin, folds as ␤-strands, is completely resistant to H/D exchange, and has a very high tendency to aggregate in solution (39). However, we have shown here that in the context of E FL or E TR , this domain does not adopt a ␤-structure. To test if at least some part of the population adopts this structure in lipid membranes, we mutated E FL residues 56 -59 to increase the predicted helicity of this part of the molecule (43), from ␤-branched Val or bulky Tyr to small side chain Ala, FIGURE 9. Structural flexibility at the putative ␤-coil-␤ motif. A, amide I band corresponding to E FL , E P54A , and E 4ALA in DMPC bilayers. Regions that change after mutation of the four residues indicated in Fig. 1A (E 4ALA ) are shown as arrows. B, Fourier self-deconvolved spectra corresponding to the amide I bands shown in A. C, infrared amide I and II bands corresponding to SARS-CoV E peptide E 46 -60 in DMPC lipid bilayers before (blue) and after (red) being exposed to D 2 O (39). D, possible equilibrium between two conformations at the putative ␤-coil-␤ motif, shifted toward an ␣-helical form, between the model determined experimentally for E TR and E FL and one built with prediction tools (PEP FOLD) (3) and consistent with data obtained with synthetic fragment E 46 -60 shown in C. to obtain the construct E 4ALA . Last, to test the effect of the conserved Pro-54 at the center of this putative motif, we generated the mutant E P54A .
The IR spectrum of these mutants when reconstituted in DMPC membranes (Fig. 9A) shows a similar amide I band. However, for the mutant E 4ALA , two shoulders are eliminated (see arrows). This indicates that these shoulders in E FL do not represent misfolded protein, but they may correspond to the bona fide ␤-structure conformation present in a small part of the population. However, the significance of these mutations is not completely clear because, although the mutations introduced in E 4ALA prevented Golgi complex accumulation when the C terminus tail of SARS-E was coupled to VSV-G (43), a similar effect was also observed for the P54A mutant, which in our IR spectrum showed no obvious differences with respect to E FL .
As stated above, the synthetic peptide E 46 -60 (39), which encompasses the predicted ␤-hairpin in SARS-CoV E, was found to produce ϳ100% ␤-structure and was completely resistant to H/D exchange (Fig. 9C). Combined with the effect observed for the E 4ALA mutant, we propose that this ␤ structure may be in dynamic equilibrium with the much more abundant ␣-helical form (Fig. 9D). A delicate balance between these two forms may alter processes in the infected cell (e.g. membrane scission, binding to protein partners, or E protein localization).
Finally, the HMA titration results showed large CSP values for residues Val-49 and Leu-65, which are far apart in the sequence. Using the SymmDock server (75,76), a reconstructed E TR pentameric model was obtained based on the published structure of the E-TM pentamer (34) and the current E TR structure. The model suggests that these two residues may be in fact spatially close (Fig. 10) and belong to different monomers, providing a rationale for the observation above.
CoV E proteins have been proposed to have at least two roles. One is related to their TM channel domain. This would be active in the secretory pathway, altering lumenal environments and rearranging secretory organelles and leading to efficient trafficking of virions (38,77). The other would be related to their extramembrane domains, particularly the C-terminal domain. This is involved in protein-protein interactions and targeting, among other roles. E protein participates in M-M and E-M interactions (17,44), which are interesting targets for drug discovery. Also, formation of viral particles appears to be facilitated by a broad range of E sequences (78), which suggests that a common topology is more important than sequence requirements. E proteins have been suggested to act as chaperones during packaging (79), but the precise mechanism by which this takes place is not known. In this context, the structure determined here sheds light on a critical domain present in most CoV E proteins.