Binding of synthetic carbohydrate receptors to enveloped virus glycans: Insights from molecular dynamics simulations.

Can envelope glycans be targeted to stop viral pandemics? Here we address this question by using molecular dynamics simulations to study the binding between 10 synthetic carbohydrate receptors (SCRs) and the 33 N-glycans most commonly found on the surfaces of enveloped viruses, including Zika virus and SARS-CoV-2. Based on association quotients derived from these simulations, we classified the SCRs as weak binders, promiscuous binders, or selective binders. The SCRs almost exclusively associate at the Man3GlcNAc2 core, which is common to all N-glycans, but the binding affinity between the SCR⋅glycan pair depends on the noncovalent interactions between the heterocycle rings and the glycan antennae. Systematic variations in the glycan and SCR structures reveal relationships that could guide the design of SCRs to attain affinity and selectivity towards a chosen envelope glycan target. With these results, envelope glycans, which are currently considered "undruggable", could become viable targets for new therapeutic strategies.

Introduction

Enveloped viruses (EnV) – viruses surrounded by a glycosylated lipoprotein bilayer envelope [1] – include coronaviruses, retroviruses, flaviviruses, bunyaviruses, alphaviruses, togaviruses, filoviruses, and others [2]. Collectively, they are responsible for many recent health crises, including the HIV epidemic [3], the Zika virus (ZIKV) outbreak in 2016 [4], and the current severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic [5]. Common strategies for antiviral drug development include entry inhibition, RNA-dependent RNA polymerase inhibitors, protease inhibitors, and others [6], but none of these have resulted in FDA-approved treatments for most flaviviridae or coronaviruses, including SARS-CoV-2 or ZIKV, and only two have Emergency Use Authorization by FDA [7]. Instead treatments focus on repurposing already approved drugs, monoclonal antibodies, or palliative care [8,9]. Given the limitations of current drug design strategies to mitigate the intensity of the SARS-CoV-2 pandemic, or to protect from present or future EnV threats, there is no alternative than to consider new antiviral drug design strategies.

EnV glycoproteins, occurring as membrane-anchored peplomers or “spikes”, bind receptors on the surface of the host cell, and this step precedes virus entry and viral replication [[10], [11], [12], [13]]. Further, these glycoproteins are densely decorated with N-glycans and O-glycans which constitute up to 25% of their molecular weight [14]. The N-glycans, which share the Man3GlcNAc2 core sequence, are involved in a range of essential processes, including viral protein folding, evasion of the host's immune system, and facilitating attachment to the host receptors [15]. For instance, selective binding of HIV-1 and Ebola viruses to the host cell is facilitated by an oligomannose N-glycan mediating interactions with the DC-SIGN receptor [15], and the EnV glycans of SARS-CoV-2 are known to stabilize the open state of the spike protein‘ [[16], [17], [18]]. Thus, synthetic molecules that selectively bind EnV glycans and disrupt these processes could act as broad spectrum antivirals (BSAs) - agents that target a wide-range of viruses and which could be deployed immediately to mitigate the threat of the infection [13]. The challenge with this strategy, however, is that glycans are considered “undruggable targets”, meaning they have a known role in disease progression, but no widely adopted therapeutic strategies exploit this information [19]. Although natural lectins and antibodies recognize specific glycans, their toxicity [20] has limited their use as therapeutic or imaging agents [21]. There is only one glycan-targeting antibody that has received FDA approval – a drug of last resort for high-risk childhood neuroblastoma [22]– and there are no approved antiviral treatments whose mechanism involves binding EnV glycans.

To challenge this status quo and redefine glycans as feasible targets for therapeutics and chemosensors, several research groups [[23], [24], [25]] have developed synthetic carbohydrate receptors (SCRs) – small molecules that bind carbohydrates through noncovalent interactions. Inspired by the structure and binding thermodynamics of natural glycan binding proteins, we have recently reported a series of flexible, tetrapodal SCRs based upon a biaryl core, that are selective for non-glucosides [13,[26], [27], [28], [29], [30]], and some of these tetrapodal SCRs display nanomolar inhibition against live ZIKV, with data suggesting that the mode of action is the prevention of viral attachment/binding/entry [13,28]. Despite these encouraging results, without a molecular-level understanding of the binding of the glycans by SCRs, their structures cannot be rationally redesigned to achieve strong and selective binding towards EnV glycans. Fine-tuning the specificity of the SCRs towards selected EnV glycans would (1) improve the inhibition of the viral docking to the host receptor, thus reducing the chance of the infection, and (2) decrease the likelihood of off-site toxicity caused by binding to the host N-glycans. In this study, we report detailed molecular dynamics (MD) simulations analysis of the binding between 10 tetrapodal SCRs and 33 N-glycans common to surfaces of ZIKV [31], MERS-CoV, SARS-CoV, and SARS-CoV-2 [[32], [33], [34]]. These simulations capture how small modifications in N-glycan or SCR structures alter the binding geometries and association strengths, and provide a molecular-level rationale for the differences in the observed antiviral activity of the SCRs, which can guide the design of the next generation of EnV glycan binding-receptors.

Results and discussion

The SCRs and EnV glycans investigated in this work are shown in Fig. 1 . We selected three SCRs (SCR001, SCR005 and SCR007) that have displayed potent antiviral activity against live ZIKV [28] and seven recently synthesized SCRs (SCR007, SCR018, SCR019, SCR020, SCR021, SCR022, SCR023) that have shown high selectivity towards biologically relevant monosaccharides [29]. All tetrapodal SCRs share the same chemical design: a biaryl core decorated with four linkers terminated with a heterocycle (Fig. 1A) [26]. The investigated SCRs consist of four pairs of regioisomers, which differ by the point-of-attachment of the heterocyclic ring, and have pyrrole (SCR001 and SCR017), indole (SCR018 and SCR019), pyridine (SCR020 and SCR021) or phenol (SCR022 and SCR023) groups attached to the amine-containing linker. The other two receptors have 2-furan (SCR005) or 2-thiophene (SCR007) rings attached to the imine-containing linker.

Fig. 1

(A) 10 Tetrapodal SCRs investigated in this work. (B) A biantenary complex N-glycan showing configurations of glycosidic bonds and labelling scheme. The monosaccharides are coded according to the SNFG notation [35]. (C) 33 studied N-glycans common to EnV [[32], [33], [34]], divided into three major types. The complex N-glycan are divided into four subgroups, depending on the type of a carbohydrate at the non-reducing end.

As for the binding targets, we selected 33 N-glycans common to the surfaces of flavi- and coronaviruses [[31], [32], [33], [34]]. These N-glycans possess conserved structural features that were used to sort them into three groups (Fig. 1B and C): (1) oligomannose N-glycans bearing 3 or 5–9 mannose units (abbreviated as MX, where X = 3 or 5–9); (2) hybrid N-glycans with either LacNAc (G1) or Neu5Acα(2 → 6)LacNAc (S1) on the α 1–3 arm and a Man3 (M3) or Man5 (M5) group on the α1–6 arm; (3) complex N-glycans which were sorted into four subgroups based on the type of the carbohydrate at the antennae termini: GlcNAc-terminated (NX, where X = 1–4), Gal-terminated (GX, where X = 2–4), GalNAc-terminated (L2), and Neu5Ac-terminated (SX, where X = 1–4). Their structures can include an optional bisecting GlcNAc (-B) or fucosyl at the reducing (F) and an additional fucosyl on the non-reducing (F-3/6) GlcNAc. The attachment of a third antenna to either the α 1–3 or the α 1–6 arm is indicated by the -3 or -6 suffixes. Terminal Neu5Ac residues (S1, S2, S3-3, S3-6 and S4) are always attached to the preceding Gal with the α(2 → 6) glycosidic bond and were neutralized with K+ cations.

MD simulations were performed and analyzed using the GROMACS-2020.1 MD engine [36]. The CHARMM36 force field was selected for N-glycans because of its parametrization for carbohydrates [37] and carbohydrate-protein interactions [38]. The SCRs were parameterized using CHARMM36-compatible CGenFF [39]. For the simulation of the SCR⋅glycan pairs (1 μs each), we monitored the distance between their center-of-masses. As an example, Fig. 2A compares the binding of SCR007 (2-thiophene) to N3-3, which contains an additional GlcNAcβ(1 → 4) on the α1–3 arm, and to N2, which lacks the GlcNAc residue. The histograms of the distances reveals that SCR007 binds preferably to the tri-antennary N3-3, and this differential binding can be understood by inspection of the most abundant SCR007⋅N3-3 clusters present in the molecular trajectory. An overlay of 10 conformers that belong to the cluster is shown in Fig. 2B. The structure of the complex shows that the additional GlcNAcβ(1 → 4) in the glycan α1–3 arm engages in C–H⋯π interactions with one of the thiophene heterocycles, thus stabilizing the SCR007⋅N3-3 complex more than the SCR007⋅N2 complex, which does not have this additional stabilization. We defined the association quotient (Q ) as the ratio between a population of bound and unbound states, and use it as a metric to compare the binding of the 330 flexible host-guest complexes (Table S1 in the SI). The Q can be converted to the binding constant K via the equation K  = Q (Q + 1)/C (see SI for details), where C  = 7.69 mM, which is equivalent to one molecule per 216 nm3. With this equation, Q s of 1, 5, and 10 are equal to K s of 2.6 ⋅ 102, 3.9 ⋅ 103, and 1.4 ⋅ 104 M−1, which are in the same range as the K s determined experimentally between monosaccharides and SCRs [29]. To validate that the Q s are reproducible, we repeated the simulations of 8 SCRs and three N-glycans (G1M3, G2F and S2) three times to calculate the standard deviation (Table S2). Out of 24 SCR⋅N-glycan pairs, 17 pairs yielded a standard deviation below 0.33, six pairs yielded between 0.33 an 0.66, and one pair (SCR019⋅S2) had a standard deviation of 0.98. Thus, the Q is able to identify the receptors that bind a specific glycan, while avoiding computationally-demanding derivations of fully converged binding constants [40].

Fig. 2

(A) MD simulations of binding between SCR007 or N2 (top) and N3-3 (bottom), with histograms illustrating the time spent in bound and unbound states. The ratio of the number of bound and unbound structures defines the association quotient, Q, shown in the histogram. (B) Overlay of 10 conformers selected from the most abundant cluster of the N3-3 glycan and SCR007 (yellow) complex, showing its fluctuation around the equilibrium structure.

All plots showing the center-of-mass distance are presented in Figs. S2–S34 in the SI and the populations of the most abundant clusters are shown in Tables S3–S8. The overview of Q s of six SCRs (SCR005, SCR007, SCR018, SCR019, SCR022, SCR023) that have either strong or selective binding to 33 N-glycans is shown in Fig. 3 and illustrative examples of host⋅guest complexes are discussed in detail. An overview of three other SCRs (SCR001, SCR017, SCR020), which show weak or no biding towards, N-glycans and one SCR (SCR021), which is selective to two glycans, are shown in Fig. S35 in the SI.

Fig. 3

Association quotients (Qs) of six SCRs binding to the library of 33 N-glycans. The SCRs are sorted in regioisomeric pairs, as well as a pair of SCR005 and SCR007, to help visualize the differences in Qs between receptors. Each SCR pair is color coded. Two Qs of SCR022 are above the limit of 5.0 and have their numerical value included above the bar. The N-glycans are sorted into groups based upon their composition and the type of a carbohydrate at the non-reducing end. The data for the remaining four SCRs is shown in Fig. S35.

Oligomannose N-glycans

Only three SCRs, SCR018, SCR019, and SCR022, bind the oligomannose N-glycans with Q , and their selectivity depends on the mannose content of the glycan. Of these, SCR022 possesses the largest Q (1.59) towards M9, but the affinity of other oligomannose N-glycans towards the receptor decreases with decreasing mannose content in the α1–6 arm, to Q of 0.70 (M8) and 0.71 (M7). Further change in the number of mannose residues in the α1–3 arm in M6 and M5 increases the Q s to 1.28 and 1.28, respectively. The representative structures of the most populated complexes, shown in Fig. S36A, can be used to rationalize this trend. The receptor initially associates at the α(1 → 6) glycosidic bond in the Man5 in the α1–6 arm. This allows the phenol heterocycles to form interactions with the adjacent mannosyl residues. Reducing the mannosyl content in this arm results in a weaker association. In the complexes with M6 and M5 glycans, the receptor moves towards the central mannosyl moiety in the Man3GlcNAc2 core, which allows the heterocycles to form C–H⋯π and H-bonding interactions with the α1–3 and α1–6 arms, resulting in a similar magnitude of the association.

The other two receptors that bind oligomannose N-glycans selectively are indole-bearing SCR018 and SCR019. The first, which has a linker attached to the C2 of the indole heterocycle, is selective towards M8 (Q  = 1.12), while association with other oligomannose glycans show Q . Its regioisomer SCR019, which has the linkers attached to C3 of the indole hetorocycle, is selective towards M7 (Q  = 1.41). The most abundant clusters (Figs. 4A and S36B) reveal that in both complexes the biaryl core of the receptor aligns along the α(1 → 6) glycosidic bond connecting the two mannoses in the Man3GlcNAc2 core. In this position, the receptor associates at the more hydrophobic α-face of the central mannosyl group, opposite to the hydrophilic axial O2. Then, four heterocycle groups of the receptor engage in C–H⋯π interactions with two mannosyl groups in the α1-6 arm, one mannosyl in the α1–3 arm, and the GlcNAc moiety at the reducing end of the N-glycan. In the SCR019⋅M7 complex, these interactions with the α1–6 arm appear stronger than in SCR018⋅M8. The additional mannose residues on the α1–6 arm in M9 sterically hinder the alignment of the heterocycle rings, whereas reducing the number of mannose in the α1–3 arm diminishes the interactions of the heterocycle with this arm. Finally, we looked at the binding of the receptor with the model M3 glycan, which consists solely of the Man3GlcNAc2 core. It shows a trend similar to other oligomannose N-glycans, and only SCR022 has Q , and other receptors do not bind to this model glycan.

Hybrid N-glycans

The four hybrid N-glycans can be divided into two sets based upon the composition of the α1–3 arm (G1-and S1-) and the mannose content in the α1–6 arm (-M3 and -M5). SCR018 is selective towards the two glycans that contain three mannose residues on the α1–6 arm, as its supramolecular complex with the G1M3 and S1M3 glycans yields Q s of 2.04 and 1.76, respectively. When the number of mannoses in the α1–6 arm increases to five units, the receptor's association becomes weaker, and the Q s decrease to 0.98 (G1M5) and 1.07 (S1M5). The structures of the complexes reveal that the stronger binding to hybrid glycans with a Man3 group is driven by the interactions of the SCR with both α1–3 and α1–6 arms (Figs. 4B and S37A), as the receptor inserts into the central position above the central mannosyl in the Man3GlcNAc2 core. Increasing the mannose content on the α1–6 arm blocks this binding conformations, and the receptor instead associates on the α1–6 arm, leading to a less stable complex (Fig. S37A). The other two receptors which bind strongly to hybrid N-glycans are SCR022 and SCR023, but unlike SCR018 these receptors are promiscuous binders, meaning they will bind all four hybrid N-glycans with similar affinity. The representative structures of these complexes reveal that the receptors associate at the Man3GlcNAc2 core (Fig. S37B) which allows their linkers and heterocycles to engage in H-bond and C–H⋯π interactions with the antennae.

Complex N-glycans

GlcNAc-terminated: The smallest complex N-glycan, N2, has Q with all SCRs except for SCR022, which has Q of 1.58. Fucosylation of the reducing GlcNAc (N2F) reduces its Q to 0.88 but increases the affinity of other three receptors, SCR019 (1.15), SCR021 (1.14), and SCR023 (1.33), all of which have the heterocyclic ring attached to the linker at the 3-position. The simulations reveal that the binding, which in all cases occurs at the Man3GlcNAc2 core, is driven by hydrophobic C–H⋯π interactions between the SCR biaryl core and the β-face of the non-terminal GlcNAc residue and is boosted by analogous interactions between the 3-heterocyclic rings and GlcNAcs in two antennae. The adjacent Fuc residue further expands the hydrophobic surface of the glycan, allowing for the formation of additional C–H⋯π interactions with the SCRs containing hydrophobic 3-heterocycles (Figs. 4C and S38A). Changing the attachment of GlcNAc from the α1–6 arm (N2F) to the bisecting position (N1F-B) weakens the binding of these three receptors, a result of losing favorable interactions between one of the receptor arms and the GlcNAc residue in the flexible antenna (Fig. S38B), as GlcNAc in the bisecting position is less available for forming C–H⋯π interactions. Comparison of N2 and N2-B N-glycans shows a similar trend, as the addition of the bisecting GlcNAc weakens the overall binding by removing the H-bond that forms with the O4-hydroxy groups of the central mannosyl residue in the Man3GlcNAc2 core, which restrains the flexibility of both antennae.

Increasing the number of GlcNAc-terminated antennae in N3-3, N3-6, and N4 glycans has the largest effect on the selectivity of the SCR007 and SCR021 receptors. An additional GlcNAc residue attached to the mannose in the α1–3 arm increases the SCR007 binding affinity to N3-3 and N4 glycans from 0.60 (N2) to 1.47 and 1.78, respectively, whereas addition of the GlcNAc residue to the α1–6 arm does not affect the receptor's association. Figs. 4D and S38B reveal that the extra GlcNAc in the α1–3 arm provides some stabilizing C–H⋯π interactions with the 2-thiophene ring of SCR007, yet the interactions of the heterocyclic rings with the GlcNAc and Man in the α1–6 arm contribute more to the stability of the complex. An additional GlcNAc in the α1–6 arm disrupts these interactions, which renders the SCR007 complex with N3-6 less stable (Q  = 0.49). Another receptor, SCR021, is specific towards the N3-6 glycan, as its association quotient of 2.36 is the largest value observed for this receptor. Inspection of the most abundant cluster (Fig. S38C) shows that the receptor inserts between the two GlcNAcs in the α1–6 arm, such that each of phenyl rings forms C–H⋯π interactions with the β-face of the glycan ring. Then, this position is further stabilized by other contacts between the 3-pyridine heterocycles and same GlcNAcs. This binding mode is not observed for any other receptor or glycan, and might explain why SCR021⋅N3-6 is the only stable complex formed by SCR021. The tetra-antennary N4 glycan forms the most stable complexes with SCR007 and SCR019 (Q of 3.33, Fig. S38C). SCR007, instead of binding facially at the Man3GlcNAc2 core, inserts its biaryl core between the core and the GlcNAc in the α1–6 arm. The sandwiched complex enables C–H⋯π interactions between the α-face of the GlcNAc and the biaryl core, and the N-acetyl group and the biaryl core on the opposite side. SCR019, on the other hand, associates at the α-face of the central mannosyl between the two antennae, and this central position allows the four indole heterocycles to engage in C–H⋯π interactions with the four GlcNAc groups in both arms.

Gal-terminated: Extending the N2 glycan's antennae by two Gal residues (G2) does not alter the SCRs' binding affinities significantly, and SCR022 remains the only receptor that displays Q with G2. While the core-fucosylation in the G2F glycan nominally increases the association to all SCRs, it substantially increases the binding with SCR019 from Q  = 0.73 for G2 to Q  = 1.85 for G2F. In a complex similar to that formed between the receptors and N2F, the biaryl core of SCR019 aligns parallel to the β-face of the GlcNAc and α-face of the central mannosyl of the Man3GlcNAc2 core, which allows three out of four arms to engage in C–H⋯π interactions with two other GlcNAc residues, one in each antenna, and the fucose residue adjacent to the reducing end (S39A). Similarly, this synergy of multiple SCR-glycan contacts increases Q from 1.32 (SCR022⋅G2) to 1.71 for the SCR022⋅G2F complex (Fig. 4E ) and from 0.66 to 1.41 for SCR023⋅G2F. Further fucosylation in either of the antennae in G2F-3 and GF2-6 glycans affects their binding with SCR019 and SCR022 in an opposite manner. On the one hand, the additional fucose attached to the GlcNAc in α1–3 or α1–6 destabilizes the SCR019⋅glycan complexes by disrupting the interactions of the heterocycles with either of the antenna, yielding Q s of 1.16 and 1.01 for G2F-3 and G2F-6, respectively. On the other hand, the fucose stabilizes both SCR022⋅ glycan complexes by enhancing H-bonding interactions with the 2-phenol arms (Q s of 2.54 for G2F-3 and 3.46 for G2F-6, Fig. S39B). SCR023 experiences similar but weaker enhancement. An addition of a bisecting GlcNAc residue in the G2F-B glycan decreases the Q of the glycan complex with SCR022 to 1.35. The lower Q highlights the importance of the H-bond with the O4 atom of the central mannosyl residue for achieving the strong bonding. The same bisecting GlcNAc influences the binding of SCR018 and SCR019 in opposite ways: by decreasing Q of SCR018 from 0.98 (G2F) to 0.56 (G2F-B) but increasing Q of SCR019 from 1.85 (G2F) to 2.35 (G2F-B).

Fig. 4

Representative structures of eight SCR⋅N-glycan complexes discussed in the text. The color-coding of the glycans follows the SNFG notation, shown in the cartoon, and hydrogens were omitted for clarity. SCRs are shown in yellow, with nitrogen atoms highlighted in blue, oxygen atoms in red, sulphur atoms in orange, and polar hydrogen atoms in white. Note that the SCRs tend to bind at the Man3GlcNAc2 core of the N-glycan, facing the α-face of the central mannosyl residue. To show the binding, the N-glycans were rotated 180 with respect to the orientation shown in the SNFG representation.

Addition of the third LacNAc antenna to the α1–3 arm of G2F to form G3F either weakens or leaves the association unchanged for all SCRs except for SCR007 and SCR018, whose Q s increase, respectively, from 0.65 (G2F) to 1.94 (G3F) and from 0.98 (G2F) to 2.76 (G3F). In both complexes with G3F, the biaryl core of SCR017 or SCR018 assumes the same central position as G2F, and the three receptor heterocycles form C–H⋯π interactions with core fucose and the GlcNAc in the β1–2 antennae, but these contacts are more frequent in tri-antennary G3F (Figs. 4F and S39C). Subsequent addition of the fourth LacNAc antenna to the α1–6 arm in G4F compounds with the effect of the previous antenna. Although the association of SCR018⋅G4F decreases to 2.05, the extra antenna increases the association with SCR007, SCR023 and SCR022, complexes which display Q s of 2.58, 3.45 and 9.95, respectively. The SCR007 receptor forms a similar complex to that already observed for GlcNAc-terminated glycans, and the addition of the antennae to the α1–3 arm increases its Q , which compounds with the addition of the fourth moiety. Finally, the large Q s of the phenol-based receptors is somewhat unexpected, as their Q s with tri-antennary (N3-3, N3-6, and G3F) and tetra-antennary (N4) N-glycans are all smaller than 1.5. However, the inspection of the molecular structures of the most populated clusters of SCR022⋅G4F and SCR023⋅G4F reveals H-bond interactions between the phenol and the amine linkers and the two GlcNAc residues at the reducing end (Fig. S39D). This binding conformation is further stabilized by the bending of the β1–2 antenna in the α1–6 arm and forming C–H⋯π interactions with another phenol group.

GalNAc-terminated: Next, we investigated how changing the terminal Gal to GalNAc, which results in additional N-acetyl groups in the antennae, affects the binding of the SCRs and complex N-glycans. SCRs have a larger affinity to the smallest LacdiNAc-bearing N-glycan, L2, than to analogous Gal-terminated G2. We observed the largest Q of 2.52 and 1.56 with the two indole-bearing SCRs, SCR018 and SCR019, as well as a Q of 1.39 for binding with SCR007 (Fig. S40A). The core fucosylation in L2F decreases the association of the indole-bearing SCRs receptors to 1.45 and 0.76, respectively, and SCR007 to 0.80, but it increases the association of the phenol-bearing receptors SCR022 and SCR023 to 6.11 and 2.54, respectively. The representative structures of the most abundant cluster of the complex with SCR022 reveals that the receptor, as observed in previous complexes, aligns with the biaryl core parallel to the Man3GlcNAc2 core (Figs. 4G and S40B). One of the phenol rings engages in a C–H⋯π interaction with the core fucose and also with the GlcNAc in the α1–6 arms. Additional fucoses in either of the antennae decreases the nominal Q s of both of the phenol SCRs (Fig. S40C). However, the association of SCR022⋅L2F-3 remains strong with a Q of 4.82, whereas shifting the fucose to the α1–6 arm (L2F-6) decreases the Q to 1.92. This agrees with the observed binding mode for the L2F glycan, as extra fucosylation in the α1–6 antenna disrupts the C–H⋯π interactions with the heterocycle. Adding the fucose to the α1–6 arms brings the Q of SCR007 and SCR019 to similar values as for L2.

Neu5Ac-terminated: The last set of glycans feature bi- (S2), tri- (S3-3 and S3-6), and tetra- (S4) antennary glycans terminated with Neu5Ac residues attached with the α(2 → 6) linkages, and a G2 glycan featuring only one Neu5Ac on the α1–3 arm (S1). Only SCR022 is a strong binder towards the S1 glycan, with a Q of 2.79, whereas all other Q s are smaller than 1. Addition of the second Neu5Ac in the α1-6 antennae in S2 weakens the binding of SCR022 to 1.49, but simultaneously increases the binding of SCR019 from 0.65 to 1.63. The overall binding to these negatively charged bi-antennary glycans remains weak, but increasing the number of antennae to three results in large changes in the Q s of several receptors. First, the binding of SCR019 appears to be selective towards attachment of the antenna: additional antenna in the α1–3 arm (S3-3) increases its binding to 1.86, but adding it to the α1–6 arm (S3-6) lowers significantly the receptor affinity towards the sialylated glycan (Q  = 0.42). Second, SCR018 experiences the opposite effect, as it binds to S3-3 weaker (0.79), but its binding to S3-6 is strongly enhanced with Q of 3.72. The inspection of the structures shows that the two regiosiomeric receptors associate such that they form interactions with the new antenna in the respective arm (Fig. S41A). Both receptors then are strong binders to the tetra-antennary S4 glycan, with Q s of 2.21 and 4.26, respectively, for SCR018 and SCR019 (Figs. 4H and S41B), as the S4 glycan features the additional antenna available for the interactions with either of the receptors.

Conclusions

In conclusion, a systematic MD study of 330 SCR⋅glycan pairs revealed that the SCRs are selective binders towards N-glycans, and their specificity depends on the type of the linker, heterocyclic groups, and their point-of-attachment, consistent with our previous conclusions [29,30]. The investigated SCRs can be divided, based upon their Q s towards the library of 33 N-glycans, into three groups: weak binders, promiscuous binders and selective binders. The first group consists of weakly binding SCRs that have Q s with all N-glycans smaller than 1. The members of this group are SCR001, SCR005, SCR017, and SCR020. The second group consists of promiscuous SCRs, which have Q towards more than half of the investigated N-glycans. SCR018, SCR022 and SCR023, which have Q towards 19, 27 and 17 different N-glycans, respectively, are members of this group. SCR022 binds almost all investigated glycans, and the other two receptors show binding preferences between the types of N-glycans. For instance, SCR023 does not bind oligomannose N-glycans, but binds almost all Gal-, GalNAc-, and Neu5Ac-terminated complex N-glycans. SCR018 has more nuanced binding preferences. Although it would bind 19 different glycans with the Q , it has Q towards seven specific N-glycans, which enables the receptor to differentiate between minute structural details, such as the number of mannosyls in the α1-6 arm of the hybrid N-glycans or the position of the antenna in the Neu5Ac-terminated complex N-glycans. However, because these receptors have significant Q s towards more than half of the N-glycans, they would probably display high off-site toxicity as a consequence of binding to non-EnV glycans. The third group is composed of three selective binders, SCR007, SCR019 and SCR021, which have large Q s towards few specific N-glycans. SCR021 has only two Q s which means it would bind selectively to the N2F and N3-6 glycans. SCR007 associates with Q with 12 N-glycans, which share some common structural features such as additional antenna in the α1–3 arm. We can conclude that this receptor would bind to complex N-glycans, which have more than three Gal-terminated antennae. Finally, SCR019, which has Q for complexes with 14 different N-glycans, is strongly selective for tetra-antennary glycans N4 (3.33), and S4 (4.26) and tri-antennary glycans with the additional antennae in the α1-3 arm.

The analysis of the structures of the most abundant complexes observed in the MD simulations revealed several key features of SCR⋅glycan binding. First, we observe that the receptors associate almost exclusively at the Man3GlcNAc2 core, which is common to all EnV glycans (Fig. 4). This association is driven primarily by the C–H⋯π interactions between the SCR biaryl core and the hydrophilic surfaces of the pyranose rings. Second, the strength of the association depends on the noncovalent interactions (H-bonding and C–H⋯π) between the heterocycle rings and H-bonding linker, and the decoration of the N-glycan that involves the composition and connectivity of the antennae, optional fucosylation and/or bisecting GlcNAc. The weak binding with the M3 N-glycan, which constitutes the Man3GlcNAc2 core in other N-glycans, confirms the importance of the interactions with the antennae. However, each of these structural features affects the binding of the receptor in a non-trivial way and complicates the rational design of the receptors towards specific targets. This shortcoming, as shown in this work, can be alleviated by the computational screening the SCRs against the library of the EnV glycans.

In summary, the MD simulations illustrate how computations can guide the experimental design to address one of the most challenging and pressing problems in medicinal chemistry: developing glycan-binding antiviral agents. These data suggest that it might be possible to dial-in affinity and selectivity of SCRs towards a particular glycan target. If the computational predictions are corroborated by the experimental studies, the proposed selective binding could reclassify glycans from “undruggable” to viable targets for antivirals and helps to open new avenues of investigation in medicine, biology, sensors, and diagnostics.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Adam Braunschweig reports was provided by US Army Research Laboratory. Adam Braunschweig reports a relationship with Dultech LLC that includes: board membership. Adam Braunschweig has patent licensed to Dultech LLC.