Coronavirus receptor switch explained from the stereochemistry of protein-carbohydrate interactions and a single mutation

Abstract

Hemagglutinin-esterases (HEs) are bimodular envelope proteins of orthomyxoviruses, toroviruses, and coronaviruses with a carbohydrate-binding "lectin" domain appended to a receptor-destroying sialate-O-acetylesterase ("esterase"). In concert, these domains facilitate dynamic virion attachment to cell-surface sialoglycans. Most HEs (type I) target 9-O-acetylated sialic acids (9-O-Ac-Sias), but one group of coronaviruses switched to using 4-O-Ac-Sias instead (type II). This specificity shift required quasisynchronous adaptations in the Sia-binding sites of both lectin and esterase domains. Previously, a partially disordered crystal structure of a type II HE revealed how the shift in lectin ligand specificity was achieved. How the switch in esterase substrate specificity was realized remained unresolved, however. Here, we present a complete structure of a type II HE with a receptor analog in the catalytic site and identify the mutations underlying the 9-O- to 4-O-Ac-Sia substrate switch. We show that (i) common principles pertaining to the stereochemistry of protein-carbohydrate interactions were at the core of the transition in lectin ligand and esterase substrate specificity; (ii) in consequence, the switch in O-Ac-Sia specificity could be readily accomplished via convergent intramolecular coevolution with only modest architectural changes in lectin and esterase domains; and (iii) a single, inconspicuous Ala-to-Ser substitution in the catalytic site was key to the emergence of the type II HEs. Our findings provide fundamental insights into how proteins "see" sugars and how this affects protein and virus evolution.

Keywords: coronavirus; crystal structure; hemagglutinin-esterase; sialate-O-acetyl esterase; sialic acid.

Fig. 1.

(A) Stick representation of 9-O-Ac-Sia and 4-O-Ac-Sia. O-Ac moieties are depicted with carbon atoms in cyan. (B) Substrate specificity of MHV-DVIM HE (red circles) and RCoV-NJ HE (blue squares). BSM (Left) and HSG (Right) were coated in MaxiSorb plates and incubated with twofold serial dilutions (starting at 100 ng/µL) of enzymatically active HE-Fc fusion proteins. Loss of 4-O- and 9-O-Ac-Sias (indicated by percentual depletion on the y axis) was assessed by solid-phase lectin-binding assay with enzymatically inactive virolectins MHV-S HE⁰-Fc and PToV-P4 HE⁰-Fc, respectively, with virolectin concentrations fixed at 50% maximal binding. (C) Cartoon representation of the crystal structures of the RCoV-NJ HE and MHV-DVIM HE dimers. The Left monomer is colored gray, the other by domain: lectin domain (L, blue); esterase domain (E, green) with Ser-His-Asp active site triad (cyan sticks); membrane proximal domain (red).

Fig. S1.

Sequence alignment of a representative set of type I and II HEs. The membrane-proximal domain is indicated in red, the esterase domain in green, and the lectin domain in blue. The type I/II distinctive elements, i.e., the α1β2 cysteine-loop (orange circles), the Ala-Ser substitution (red arrowhead), and the β16α6 segment (green circles) are indicated. Catalytically important residues are annotated in red below. A comprehensive alignment of all published HE sequences that was used to identify consistent differences between type I and II esterases is available upon request.

Fig. S2.

(A) Alignment of the lectin domains of MHV-S (carbon in purple) and RCoV-NJ HE (carbon in gray) with 4-O- and 4-N-Ac-Sia bound, respectively. (B) Side-by-side comparison of two type II HE lectin metal-binding sites. In MHV-S HE the site is occupied by K⁺ (magenta sphere), in RCoV-NJ HE by Na⁺ (purple sphere).

Fig. S3.

Electron density of the 4-N-Ac-Sia-substrate analog bound in the RCoV-NJ HE⁰ esterase catalytic site (A) and lectin site (B). The model used for difference density calculation did not contain the substrate analog and had been obtained by refining the crystal structure of free RCoV-NJ HE⁰ against the diffraction data of the substrate analog complex without manual model building (R_free = 20%). The contour level is 3.0 σ.

Fig. 2.

(A) Superposition of residues lining the P1 pocket of Influenza C HEF (carbon atoms, cyan), MHV-DVIM HE (carbon atoms, green), and RCoV-NJ HE (carbon atoms, salmon). Surface representation is that of MHV-DVIM HE. Conserved residues within the SGNH family of hydrolases are underlined. (B) Surface representation of the catalytic center of MHV-DVIM HE with the P1 and P2 pockets indicated. The Ser-His-Asp catalytic triad is shown as sticks. (C) 9-N-Ac-Sia binding in the HEF catalytic site as observed in the crystal complex (18). Contacting amino acid side chains are shown in stick representation and colored by atom type (oxygen, red; nitrogen, blue; carbons, gray or green for amino acid side chains and 9-N-Ac-Sia, respectively). Oxyanion hole hydrogen bonds and the bidentate hydrogen bond interaction between Arg³⁰⁵ and the Sia carboxylate moiety are shown as black, dashed lines. (D) Model of 9-N-Ac-Sia binding in the MHV-DVIM HE catalytic site based on superposition with the HEF–inhibitor complex (carbon atoms, green) and on automated molecular docking (carbon atoms, salmon), represented as in Fig. 2C. (E) Catalytic activity of MHV-DVIM HE toward glycosidically bound 9-O-Ac-Sia is abrogated by substitution of Arg³⁰⁵ by Ala. Receptor destruction was assessed as in Fig. 1B. (F) Arg³⁰⁵Ala substitution in MHV-DVIM HE does not affect activity toward the synthetic substrate pNPA. Ser⁴⁴Ala is a catalytically inactive mutant. Enzymatic activity shown as percentage of wild-type activity.

Fig. 3.

(A) Surface representation of the MHV-DVIM HE (Left) and RCoV-NJ HE (Right) catalytic sites in complex with 9-O-Ac-Sia [docked with Autodock4 (55)] and 4-N-Ac-Sia (crystal complex), respectively. (B) Surface representation of the catalytic sites of MHV-DVIM HE (Left) and RCoV-NJ HE (Right). The active-site Ser⁴⁴ in MHV-DVIM HE already adopts the “active” rotamer observed in HEF (18, 39, 40); For RCoV-NJ HE crystallized as an inactive Ser-to-Ala mutant, a Ser side chain with active rotamer was introduced using COOT. The P1 and P2 pockets are highlighted by dashed circles; approximate distances between pockets, as measured from the centers, are indicated. (C) Binding topology of αNeu4,5,9Ac₃ in type I (Left) and type II (Right) esterases. The P1 and P2 pockets accommodating the O- and N-acetyl moieties are shown schematically. αNeu4,5,9Ac₃ is shown in stick representation and colored as in Fig. 2C. Asterisks indicate the position of the O₂ atom through which Sias are glycosidically linked. The distances between 5-N- and 9-O- or 4-O-Ac methyl groups are shown. (D) RCoV-NJ HE Arg³⁰⁷ is not essential for sialate-4-O-acetylesterase activity. Ser⁴⁰Ala is a catalytically inactive mutant. Receptor destruction was assessed as in Fig. 1B. For a comparison with type I HEs, see Fig. 2E. (E) Arg³⁰⁷Ala substitution in RCoV-NJ HE does not affect activity toward the synthetic substrate pNPA. Enzymatic activity shown as percentage of wild-type activity. (F) Hydrogen bonding of the sialate-5-N-acyl carbonyl oxygen and amide nitrogen with RCoV-NJ HE Ser⁷⁴ and His³³⁶, respectively, as observed in the crystal complex, indicated as in Fig. 2C. Hydrophobic contacts between Tyr⁴⁶ and the Sia-5-N-acyl methyl group are shown as thin gray lines.

Fig. 4.

(A) Partial sequence alignment of MHV-DVIM and RCoV-NJ HE, highlighting consistent differences between type I and type II HEs (Fig. S1). Aligned sequences, with residue numbering presented Left and Right, cover the α1β2-cysteine-loop, the β2α3 segment (single Ala⁷⁸Ser substitution), and the β16α6 segment. Catalytic residues (Ser, Asp, His) are marked with asterisks. (B) Overlay of cartoon representations of the active-site regions of MHV-DVIM HE (gray) and RCoV-NJ HE (blue). Side chains of catalytic triad residues are depicted as sticks. The three type I/II distinctive elements are colored as in A. (C) Cartoon representation of the novel metal-binding site near the RCoV-NJ HE active site, formed by Glu⁴⁸, His⁵², Asp⁵⁶, and His³³⁶. The catalytic triad is shown for reference. Side chains are depicted as sticks, the Zn²⁺ ion as a gray sphere. (D) A type II HE converted into a type I enzyme. An RCoV HE-based chimera with all three type I/II distinctive elements replaced by those of MHV-DVIM displays strict sialate-9-O-acetylesterase activity. The enzyme activity of the recombinant protein (“Type I chimera”) was compared with that of the parental proteins (MHV-DVIM and RCoV-NJ HE) on BSM (Left) and HSG (Right). Cleavage of 9-O- and 4-O-Ac-Sias was assessed as in Fig. 1B, but now starting at 10 ng/µL. (E) Contribution of the three type I/II distinctive elements to esterase activity and substrate specificity. The type I chimera was subjected to mutational analysis entailing systematic reintroduction of RCoV-NJ segments. Esterase activities of chimeric proteins toward 9-O-Ac- (blue bars) and 4-O-Ac-Sias (red bars) were determined in twofold dilution series as in Fig. 1B. Data are shown as percentages of specific esterase activity, calculated at 50% receptor depletion, relative to that of the type I chimera (for 9-O-Ac-Sia) or of wild-type RCoV-NJ HE (for 4-O-Ac-Sia). The error bars represent the SD over six measurements (two biological replicates, each of which performed in technical triplicates). (F) The type II esterase metal-binding site is required for full 4-O-AE activity. Note that disruption of metal binding by either Glu⁴⁸Gln or Asp⁵⁶Asn substitution reduces sialate-4-O-acetylesterase activity by 75% (comparable to the amount of type II activity conferred by the Ala⁷⁴Ser substitution alone). Enzymatic activity measured as in Fig. 1B and presented as in Fig. 4E. (G) 4-O- and 9-O-Ac-Sias are abundantly expressed in the mouse colon. Paraffin-embedded mouse colon tissue sections were stained for 4-O-Ac-Sia with MHV-S HE⁰-Fc, and for 9-O-Ac-Sia with PToV-P4 HE⁰-Fc.

Fig. S4.

(A) Cartoon representation of the RCoV-NJ HE monomer. The structure is colored according to secondary structure with helices in yellow and β-strands in purple. Secondary structure elements are numbered sequentially. (B) Amino acid sequence of RCoV-NJ HE with secondary-structure elements colored and labeled as in A.

Fig. S5.

X-ray fluorescence spectrum of RCoV-NJ HE crystals near the Zn K-edge absorption energy. The observed absorption edge at 9.665 keV is close to the theoretical K-edge absorption energy of 9.6607 keV and indicative of the presence of Zn²⁺ in RCoV-NJ HE crystals.

Fig. S6.

CBS versatility resulting from the stereochemistry of protein–carbohydrate interactions. (A) Close-ups of the CBSs of legume lectins complexed with their receptors. Note that, in each case, monosaccharide binding relies on a strictly conserved 3-aa assembly, but that unique specificity is attained by binding of the dedicated ligands (galactose or mannose and glucose) in different orientations (through hydrogen bonding with 3- and 4-OH or 4- and 6-OH, respectively) (–48). (B) Close-ups of the CBSs of mammalian C-type lectins mannose-binding proteins A and C. Note the similarity between binding sites with hydrogen-bonding and Ca²⁺ coordination of the vicinal, equatorial 3- and 4-OH groups, and the difference in mannose binding topologies, with the sugar’s pyranose ring rotated 180° (44).

Fig. S7.

Synthesis of α-4-N-Ac-Sia-2Me. (A) Overview of the synthesis. (Step i) a. Dowex 50wx8, MeOH; b. Ac₂O, AcOH, H₂SO₄, 61% over two steps; (step ii) a. TMSN₃, t-BuOH; b. NBS, MeOH, 44% over two steps; (step iii) a. tributyltin hydride, AIBN, 1,4-dioxane; b. Ac₂O, NEt₃, THF; c. NaOH, MeOH/H₂O, 27% over three steps. (B) ¹H-NMR spectrum of final compound.

Fig. S8.

Crystal structure of RCoV-NJ HE⁰ in complex with 4-N-Ac-Sia compared with the model of 4-O-Ac-Sia bound in RCoV-NJ HE as determined with autodock4. The lowest energy solution from 10 independent runs is shown. Both structures show binding of the 4-Ac group in the P1 pocket and binding of the 5-N-Ac group in the P2 pocket.