Solution Structure of the Severe Acute Respiratory Syndrome-Coronavirus Heptad Repeat 2 Domain in the Prefusion State^*

EXPERIMENTAL PROCEDURES

Expression and Purification—The SARS-CoV S2-HR2 domain consisting of residues 1141-1193 was subcloned into the BamHI/HindIII restriction sites of a modified pQE30 expression vector (Qiagen). The resulting construct, termed His-PG-S2-HR2, consists of a N-terminal polyhistidine tag followed by protein G (the IgG-binding domain of streptococcus protein G) (41), a tobacco etch virus cleavage site (sequence, ENLYFQGS) (42) for removal of the expression tag, and S2-HR2. For simplicity, from here on S2-HR2 will be numbered 1-55 corresponding to residues 1141-1193 of the intact protein (the N-terminal glycine and serine residues are a cloning artifact). Protein expression of ¹⁵N-labeled and ¹³C/¹⁵N-labeled S2-HR2 was achieved by growing Escherichia coli strain SG13009 as previously described (43). Briefly, the cells were grown in 4 liters of LB medium at 37 °C until they reached an A ₆₀₀ of 0.8. Subsequently, the cells were pelleted, washed one time with M9 minimal medium, resuspended in 1 liter of M9 minimal medium supplemented with 1 g/liter of ¹⁵NH₄Cl (Martek Biosciences, Columbia, MD) and 4 g/liter of ¹³C-glucose (Isotec, Miamisburg, OH), set to recover for 1 h at 37°C, induced with 0.8 mm isopropyl β-d-thiogalactopyranoside, and grown for an additional 4-5 h at 37 °C. The His-PG-S2-HR2 fusion protein was purified from the soluble fraction using a Ni²⁺ fast flow Sepharose column (Qiagen). The protein was then cleaved using tobacco etch virus protease and run once more over the Ni²⁺ column to remove His-PG and tobacco etch virus protease, which also contains a polyhistidine tag. The flow through fraction containing S2-HR2 was then dialyzed extensively against 10 mm NaPO₄, pH 7.0, and concentrated by ultrafiltration (YM3; Amicon, Billerica, MA). The purity and identity of S2-HR2 were confirmed by SDS-PAGE and matrix-assisted laser desorption ionization time-of-flight mass spectrometry.

NMR Spectroscopy—For the NMR experiments, the experimental conditions were 1 mm S2-HR2 in 10 mm NaPO₄, pH 7.0, 30% TFE-d3 (Cambridge Isotope Laboratories, Andover, MA), and 10% ²H₂O. NMR spectra were recorded at 25 °C on a Buker AVANCE 800 MHz spectrometer equipped with a cryogenic triple resonance probe. The spectra were processed by NmrPipe and visualized with NmrDraw (44). Backbone assignments were determined using a standard set of three-dimensional heteronuclear NMR experiments including: HNCO, CBCA-(CO)NH and CBCANH (45, 46). The side chain resonances were assigned by three-dimensional ¹⁵N-edited TOCSY-HSQC, threedimensional HCC(CO)NH, and three-dimensional CC(CO)NH experiments. A three-dimensional ¹⁵N-edited NOESY-HSQC (mixing time, 120 ms) was acquired for structural restraint information as well as confirmation of side chain assignments. An additional HNHA experiment was performed to add ³J_HNα information (47). The intersubunit NOEs were identified by recording a three-dimensional F1-filtered, F2-edited ¹H-¹³C NOESY-HSQC with a mixing time of 120 ms (48) on a 1 mm sample comprised of a mixture of 50% ¹⁴N/¹²C-labeled and 50% ¹⁵N/¹³C-labeled monomers. Exchangeable amide protons were identified by the CLEANEX-PM experiment with a mixing time of 100 ms (49). HNOE values were measured by a standard pulse sequence (reviewed in Ref. 50). The HNOE values for each residue were calculated as the intensity ratio (I/I ₀)ofthe ¹⁵N-¹H correlation peak in the presence (I) and absence (I ₀) of proton saturation during the relaxation delay of 5 s. Spectral processing and analysis was performed using the program NMRPipe (44).

Structure Determination—NOEs were manually assigned and classified as: strong (1.8-2.7 or 1.8-2.9 Å for the NOE of NH), medium (1.8-3.3 or 1.8-3.5 Å for NOE of NH), weak (1.8-5.0 Å), or very weak (1.8-6.0 Å). For distances involving methyl protons, 0.5 Å was added to the upper limit to account for the higher apparent intensity of methyl protons. ψ and φ torsion angle restraints were derived from ³J_HNA (51) and a data base analysis of chemical shifts (N, H_N,C_α,C_β,C′,H_α) using the program TALOS (52). Minimum error ranges for ψ and φ were set to ± 40 and ±20, respectively. H-bonds, which were identified by the ¹³C backbone chemical shifts, short range NOE patterns, and nonex-changeable protons, were incorporated for the helical region, residues 17-48, as two restraints per H-bond where r_NH-O = 1.5-2.8 Å and r_N-O = 2.4-3.5 Å. The structures were calculated by simulated annealing in torsion angle space (53) starting from three extended strands, followed by conventional simulated annealing (54), using the program CNS (55). The program was adapted to incorporate a conformational data base (56) and a pseudopotential term for noncrystallographic symmetry (17, 57). A family of the 30 lowest energy structures was chosen, and a minimized mean structure was calculated. The overall quality of the final structures was assessed using the program PRO-CHECK (58). The figures were generated using the program MOL-MOL (59).

RESULTS AND DISCUSSION

Optimization of Experimental Conditions—Previous analytical ultracentrifugation studies have demonstrated that SARS CoV S2-HR2 is a trimer (40). Accordingly, our initial NMR experiments on S2-HR2 were performed under varying conditions of pH and salt to optimize the solubility of the construct. However, at protein concentrations greater than 400 μm, which are typically necessary for NMR structural studies, S2-HR2 tends to form higher molecular mass species (larger than that of a trimer, 18 kDa) as evidenced by the presence of additional broad peaks in the ¹⁵N-edited HSQC (data not shown). The additional broad peaks of the higher molecular mass species are only observed at relatively high concentrations and therefore believed to be nonphysiologically relevant. The co-solvent TFE has previously been used to break up aggregates formed between coiled coil complexes without breaking up the true coiled coil quaternary structure (60). By the addition of 30% TFE, we were able to obtain NMR spectra with one set of resonance peaks at a concentration of 1 mm, which was suitable for high resolution structural studies.

Resonance Assignment and Secondary Structure—A two-dimensional ¹⁵N-edited HSQC spectrum of S2-HR2 (buffer conditions 10 mm NaPO₄, pH 7.0, 30% TFE-d3, 10% ²H₂O) is shown in Fig. 1 . The backbone assignments were obtained using the standard set of three-dimensional triple resonance experiments as described under “Experimental Procedures” (the resonance assignments have been deposited in the Biological Magnetic Resonance Databank under entry 6969). The ¹⁵N-¹H assignments are shown in Fig. 1. The ¹³C′ and ¹³C_α chemical shifts with respect to random coil values are shown in Fig. 2a . The secondary chemical shift suggests the presence of helix from residues 17 to 51. As shown by Fig. 2b, the HNOE profile further supports the presence of the helical region encompassing this region. For example, residues 3-16 and 52-55 clearly exhibit higher mobility with respect to the central region. The presence of exchangeable amide protons was next probed by CLEANEX-PM experiments (49). As shown in Fig. 3 , the amides of residues 7-13, 49, and 51-55 are in relatively fast exchange with solvent. Based on the ¹³C chemical shifts, HNOE, and amide exchange experiment, as well as the short range NOE patterns, S2-HR2 consists of a continuous helix from residues Asn¹⁷ to Leu⁴⁷.

Structure Determination—The calculation of a symmetric trimer using NMR is a challenging task and requires the use of isotopic-filtered NOE experiments (reviewed in Ref. 61). In the present work, intermolecular NOEs were identified by a three-dimensional F1-filtered F2-edited ¹H-¹³C NOESY-HSQC (48) experiment recorded on a 1:1 mixture of ¹³C/¹⁵N: ¹²C/¹⁴N-labeled S2-HR2. Based on previous analytical ultracentrifugation experiments, S2-HR2 is a trimer; hence one is not able to distinguish between intermolecular NOEs for example between subunit A and B or subunit A and C. We therefore assigned the NOEs in such a way that a proton on one subunit is allowed to interact with protons of the other subunits as (∑r^-6)^-1/6 sums (17). Examples of the quality of the filtered NOE spectrum are shown for selected residues in Fig. 4 . In total, all intermolecular NOEs were found between residues in the a, d, e, and g positions of the helical wheel, consistent with the interface of a trimeric coiled coil structure, to be discussed below. The solution structure of S2-HR2 was determined using information from a total of 344 NOEs (including 85 intermolecular NOEs), 30 H-bonds, 19 J-coupling constants, and 110 dihedral restraints/subunit (Table 1 ). The structure statistics of the family of the 30 lowest energy structures and the final minimized mean structure are summarized in Table 1. The superposition of the backbone of the final 30 lowest energy simulated annealing structures is shown in Fig. 5a . The final ensemble shows no NOE violations over 0.5 Å and no dihedral angles in the disallowed regions of the Ramachandran plot (Table 1). The root mean square deviation of residues Asn¹⁷-Leu⁴⁷, which comprise the helical regions, relative to the mean is 0.46 Å for the backbone atoms and 1.04 Å for the heavy atoms.

S2-HR2 Solution Structure—A ribbon representation of the minimized mean structure of S2-HR2 is shown in Fig. 5b. In this representation, subunits A, B, and C are colored red, green, and blue, respectively. S2-HR2 forms a symmetric trimeric coiled coil with an overall length of ∼96 Å and width of ∼13 Å. There are extensive contacts between the three subunits with ∼6150 Ų buried upon complex formation. Interestingly, the side chains of the two conserved glycosylation sites of HR2, Asn¹⁷ and Asn³⁸, are exposed to the surface as expected. A surface electrostatic representation of S2-HR2 shown in Fig. 5c reveals that the charge is evenly distributed along the entire helix. The main interface of the S2-HR2 coiled coil trimer was determined to include residues Asn¹⁷-Leu⁴⁷.

A coiled coil structure is a bundle of α-helices that interact with each other in such a way that they will form a left-handed superhelix. Residues in the a and d positions are often hydrophobic residues that will interact with each other to form the hydrophobic core. This packing is often referred to as knobs-into-holes and is the hallmark for a true coiled coil structure (62, 63). The packing of the amino acid side chains within S2-HR2 coiled coil is shown in Fig. 6 . The hydrophobic residues in the a position (Val²⁰, Ile²⁷, Ala³⁴, and Leu⁴¹) are colored purple, and those in the d position (Ile²³, Leu³⁰, Leu³⁷, and Leu⁴⁴) are colored green to show the relative orientation and positioning. The stacking of these residues is arranged in a way that creates a tightly packed core within the S2-HR2 trimer with the hydrophobic interface buried. Positions e and g within a heptad repeat usually contain charged amino acids placed in such a way that the charges are complementary to each other to allow electrostatic interaction. In the case of S2-HR2, there is only one charged amino acid, in position g (Glu²⁶), and therefore it is not likely that the e and g positions play important roles in the stability of the S2-HR2 trimer. On the other hand, positions b, c, and f are surface exposed in the coiled coil structure and generally encompass polar amino acids that contribute to the solubility of the protein. In the coiled coil region of S2-HR2, we note that there are two surface exposed hydrophobic residues, Val²¹ and Ile⁴², that are conserved among coronaviruses. Interestingly, in the fusion structure (25), post-Val and Ile form contacts with the HR1 that presumably stabilize the six-helix bundle conformation.

Conclusions—In conclusion SARS-CoV S2-HR2 forms a coiled coil structure in solution consisting of three helices coming together to form a parallel trimer. As mentioned in the introduction, SARS-CoV S2 is thought to exist in two conformations: a prefusion conformation in which HR1 and HR2 do not interact with each other and a fusion conformation in which HR2 interacts with the trimeric coiled coil of HR1 in an antiparallel fashion to form the six-helix bundle, which was also observed for HIV and simian immunodeficiency virus (SIV) gp41 and Ebola glycoprotein. In light of our results, it is enticing to speculate that the present structure represents a prefusion intermediate of the S2-HR2 region, which has never before been observed for coroniviridae. Importantly, the information provided in this study may be exploited for the design of novel antivirals. For example, the relatively poor antiviral activity of SARS-CoV S2-HR2 may be a consequence of its tendency to form the trimeric coiled coil, and thus one potential strategy would be to mutate residues that are important to the formation of a HR2 trimer while at the same time maintaining the interactions necessary for the HR2-HR1 interaction.