Spatial Screening of Hemagglutinin on Influenza A Virus Particles

Oct 20, 2017 - In an idealistic scenario, rigid and distance-matched scaffolding should provide for highest affinity enhancements. However, exact matc...
2 downloads 22 Views 2MB Size
Subscriber access provided by University of Florida | Smathers Libraries

Article

Spatial screening of hemagglutinin on Influenza A virus particles: SialylLacNAc displays on DNA and PEG scaffolds reveal the requirements for bivalency enhanced interactions with weak monovalent binders Victor Bandlow, Susanne Liese, Daniel Lauster, Kai Ludwig, Roland R. Netz, Andreas Herrmann, and Oliver Seitz J. Am. Chem. Soc., Just Accepted Manuscript • DOI: 10.1021/jacs.7b09967 • Publication Date (Web): 20 Oct 2017 Downloaded from http://pubs.acs.org on October 20, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of the American Chemical Society is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

Spatial screening of hemagglutinin on Influenza A virus particles: Sialyl-LacNAc displays on DNA and PEG scaffolds reveal the requirements for bivalency enhanced interactions with weak monovalent binders Victor Bandlow,† Susanne Liese,‡ Daniel Lauster,# Kai Ludwig,§ Roland R. Netz,‡ Andreas Herrmann,# Oliver Seitz*,† †

Institute of Chemistry, Humboldt-Universität zu Berlin, Berlin, Germany



Institute of Theoretical Physics, Freie Universität Berlin, Berlin, Germany

#

Institute of Biology, Humboldt-Universität zu Berlin, Berlin, Germany

§

Institute of Chemistry and Biochemistry, Freie Universität Berlin, Berlin, Germany

DNA-programmed display, influenza, multivalency, secondary binding sites, statistical mechanics.

ABSTRACT: Attachment of the Influenza A virus onto host cells involves multivalent interactions between virus surface hemagglutinin (HA) and sialoside-containing glyco ligands. Despite the development of extremely powerful multivalent binders of the Influenza virus and other viruses, comparably little is known about the optimal spacing of HA ligands, which ought to bridge binding sites within or across the trimeric HA molecules. To explore the criteria for ligand economical high affinity binding, we systematically probed distance-affinity relationships by means of two differently behaving scaffold types based on i) flexible polyethylene glycol (PEG) conjugates and ii) rigid self-assembled DNA·PNA complexes. The bivalent scaffolds presented two sialyl-LacNAc ligands in 23-101 Å distance. A combined analysis of binding by means of microscale thermophoresis measurements and statistical mechanics models exposed the inherent limitations of PEG-based spacers. Given the distance requirements of HA, the flexibility of PEG scaffolds is too high to raise the effective concentration of glyco ligands above a value that allows interactions with the low affinity binding site. By contrast, spatial screening with less flexible, self-assembled peptide nucleic acid (PNA)·DNA complexes uncovered a well-defined and, surprisingly, bimodal distance-affinity relationship for interactions of the Influenza A virus HA with bivalent displays of the natural sialyl-LacNAc ligand. Optimal constructs conferred 103-fold binding enhancements with only two ligands. We discuss the existence of secondary binding sites and shine light on the preference for intramolecular rather than intermolecular recognition of HA trimers on the virus surface.

Introduction Multivalency is a principle employed by nature when monovalent interactions between receptors and ligands are too weak to elicit a biological effect.1-6 For example, the Influenza A virus (IAV) assembles hundreds of hemagglutinin (HA) trimers on its envelope to trigger attachment to host cells by recognition of specific sialic acid – galactose linkages.7-9 A single HA ligand binding site has only mM affinity for the sialylated sugar, but the multivalent engagement of receptor-ligand interactions allows the virus to attach at much lower concentrations.10 In attempts to prevent virus adhesion to host cells, a variety of multivalent HA binders have been prepared by chemical synthesis. Clustering of glyco ligands on dendrimers18-20 and nanoparticles21-24 polymers11-17, provided highly potent inhibitors. Owing to the statistical distribution of the glyco ligands, little can be learned about the structural requirements governing high affinity binding to virus surfaces and a substantial number of the expensive glyco units may in fact be redundant or even interfere with high affinity binding. By contrast, exquisite ligand economy was obtained by rationally designed glyco

systems. In a noteworthy example25, the scaffolds were designed to match the C3-symmetry of HA. This approach has required information about the precise arrangement of binding sites on the multivalent receptor. Such knowledge may not always be available when biological targets change upon evolution. Furthermore, given the high number of HA trimers on a given viral surface, it is difficult to rationally account for binding modes that bridge two neighboring HA trimers. Here we hypothesize that the spatial screening with bivalent systems sharing a natural α2,6-linked sialylLacNAc ligand will provide clues about the arrangement of ligand binding sites on HA and thereby reveal the requirements for high affinity binding to IAV without prior knowledge of structure and spatial organization. Such an approach appears particularly attractive for discovering additional, low affinity binding sites, which occasionally have been observed in crystal structure analyses of lectin-type recognition but typically go unnoticed when interactions are probed by means of low precision multivalent binders.26-30 The ideal probe would also allow the exploration of intramolecular vs.

ACS Paragon Plus Environment

Journal of the American Chemical Society

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 17

scaffold architecture is not obvious. In an idealistic scenario, rigid and distance-matched scaffolding should provide for highest affinity enhancements. However, exact matching of distances may be difficult. While most studies focused on a single scaffold type, we decided to assess distance-affinity relationships by means of two entirely different scaffold architectures based on 1) DNAtype scaffolds and 2) polyethylene glycol (PEG). In early seminal work, Knowles et al. described the application of PEG-based spacers.31 They employed rather short (37-49 atoms long) spacers to probe bivalent recognition of IAV. To facilitate enhanced recognition over long distances we explored longer PEG spacers, which include up to 144 monomer units and show high flexibility. On the contrary, the use of DNA-based materials leads to rigid architectures. DNA-based scaffolds allow a defined arrangement of the sialyl-LacNAc ligands and the sequence-programmable self-assembly facilitates distance variations.27,32-44 Herein we present a comparative binding analysis for IAV X31 and discuss the results in light of modeling of interaction of binders with HA by statistical mechanics models. The study exposes the inherent limitations of PEG-based spacers and illustrates the advantages provided by the use of high precision scaffolds such as nucleic acid duplexes (Fig. 2). The bimodal distanceaffinity relationship uncovered by the DNA-programmed spatial screening provides criteria for the design of ligand economical, high affinity IAV inhibitors. Figure 1. (A) Electron micrograph of the human Influenza A virus (X31) presenting the trimeric hemagglutinin on the viral surface. The average distance of the midpoints of two adjacent HA trimers has been determined as 101.7±0.6 Å (see Supporting Information for further details). The surface displays of the HAs show the canonical sugar binding sites in yellow. The 42 Å distance between two canonical binding pockets on one HA refers to the Eucledian distance and is not accounting for the impenetrability of the protein surface. The 49 − 154 Å span refers to the distance between two canonical binding pockets on two adjacent, freely rotating HAs. (B) Crystal structure (Pdb 1HGG) of the HA trimer (one monomer in blue) with bound sialosides (yellow: bound to canonical binding sites, red: bound to secondary binding sites. (C) Schematic but to scale display of trimeric HAs on viral membrane. As the rotational orientation of the HA trimers is unknown the HA surface displays were cylindrically averaged. For illustration, a bivalent PNA·DNA scaffold is shown (upper right).

intermolecular recognition by HA trimers on the virus. For this purpose, long-reaching scaffolds are required, which enable the assessment of distance-affinity relationships over a wide range of distances to bridge binding sites within and, potentially, across HA trimers. Cryo electron microscopy suggested a 102 Å average distance between the centers of two neighboring HA trimers (Figure 1A). However, owing to rotational degrees of freedom of HA trimers on the virus surface two sugar binding sites on two HA trimers may span a range of distances. Given these uncertainties, the choice of the

Figure 2. (A) Hybridization of modified (blue, red) and unmodified peptide nucleic acid (PNA) oligomers (green) with DNA templates provides bivalent sialyl-LacNAcPNA·DNA complexes with a defined sugar-sugar spacing along the nucleic acid backbone. Structure of (B) a bivalent sialyl-LacNAc-PNA·DNA complex and (C) bivalent sialylLacNAc-PEG conjugates.

2

ACS Paragon Plus Environment

Page 3 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

Results and Discussion Design and synthesis. For the spatial screening of HA with rigid scaffolds we used 39 nucleotides (nt) long DNA template strands and selected three out of 6 available 13 nt long peptide nucleic acid45 (PNA) oligomers (Figure 2A). Five PNA oligomers contained a single sialyl-LacNAc residue at position 3, 4 (2x), 6 or 9 and one PNA strand was applied in unmodified form. After hybridization with the DNA template, the two sugar appendages will be spaced at 7−31 nucleotide distance, which corresponds to 23 – 101 Å on the basis of the PNA·DNA duplex structure.46 Owing to the high affinity interactions between PNA and sequence complementary DNA a 13 nt long PNA·DNA duplex has sufficient stability (TM > 50°C, Table S2, Fig. S22) to maintain integrity at submicromolar concentrations. The nick sites provide rotational degrees of freedom to avoid spiralization stress upon bivalent interactions.47 For attachment of the sialyl-LacNAc residue via 1,4-addition to maleimides, γmercaptomethylated PNA monomers37 were included in the PNA sequence (Figure 2B, Scheme S5). To ensure comparability between the two scaffold types, we used the thiol-maleimide conjugation also for the synthesis of PEG spaced bivalent sialyl-LacNAc systems and appended

terminal thymidine units to PEG diamines (Figure 2C, Scheme S4). From light scattering experiments and atomistic simulations it is known that the average end-toend distance rete of PEGs follows a Flory-behavior; rete = aF·N3/5, with aF = 4.0 Å and N the number of PEG monomers.48,49 The incremental gain in length provided by an additional ethyleneoxy unit decreases with increasing oligomer length. As a result, rather long PEG chains are needed to separate the glyco ligands by > 50 Å averaged distance. The bivalent PEG sugar conjugates PEG29, PEG50 and PEG79 display the ligands at averaged distances of rete = 29 Å, 50 Å and 79 Å, respectively. For control measurements, we also prepared monovalent PEG sugar conjugates of varying length (Scheme S3). Spatial screening of soluble hemagglutinin trimers. The soluble trimeric ectodomain bHA was prepared by bromelaine cleavage of HA of intact X31 virus.50 The affinities of the DNA- and PEG-linked sialylLacNAc for fluorescence labeled bHA (Fig. S24) were determined by means of a microscale thermophoresis (MST)-based assay previously described by Xiong et al.51 To validate the assay and confirm the integrity of the bHA, we tested the interaction with N-Cbz-protected sialyl-LacNAc glycoside α2,6-SLN. The dissociation

Table 1: Binding affinities of the trisaccharide ligand, trisaccharide PNA·DNA duplexes and PEG-trisaccharide conjugates for bHA and IAV X31 determined by thermophoresis and hemagglutination inhibition assay. complex

rete / Å

KD,app / µM (bHA)

KD,app / µM (virus)

KiHAI/ µM (virus)

α2,6-SLN

monovalent

3100±250

1760±340

60000

α2,6-SL

monovalent

n.d.

n.d.

50000

, NAm

monovalent

863±67

n.d.

>2000

, NA23

23

244±99

128±33

>500

, NA26

26

208±18

68±7

83±21

, NA33

33

277±183

127±46

>500

, NA42

42

187±15

35±4

>500

, NA52

52

29±5

10±1

63

, NA59

59

104±11

16±2

44±6

, NA62

62

240±15

32±5

63

, NA68

68

543±79

136±21

78±18

, NA85

85

-

-

>500

, NA101

101

588±47

340±42

>500

PEG (N=3), PEGm1

monovalent

5900±600

n.d.

-

PEG (N≈45), PEGm2

monovalent

n.d.

n.d.

-

PEG (N=27), PEG29

29

2160±70

n.d.

5000

PEG (N≈67), PEG50

50

4300±1400

n.d.

5000

PEG (N≈144), PEG79

79

4600±1900

n.d.

>6000

a

, , , , modified PNA oligomers, unmodified PNA oligomer, DNA template, α2,6a sialyl-LacNAc, n.d.= not determinable. Data are minimum concentrations required for inhibition of hemagglutination. A KiHAI > x means that no inhibition of viral hemagglutination was detected at the given concentration x.

3

ACS Paragon Plus Environment

Journal of the American Chemical Society

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

constant Kd = 3100±300 µM is in reasonable agreement with the Kd = 6100±200 µM reported for unmodified α2,6linked sialyl-LacNAc from commercial sources.51 The analysis of the bivalent DNA·PNA-sugar assemblies NA23NA101 revealed, perhaps surprisingly, a bimodal distanceaffinity relationship (Table 1 and Fig. 4, green). Rather low affinities were determined for constructs that presented the sialyl-LacNAc ligands in distances ≥ 68 Å. A welldefined bivalency enhancement (106-fold over the trisaccharide SLN and 30-fold over monovalent PNA·DNA complex NAm) was observed when the two sugar ligands were separated by 52 Å (Kd = 29±5 µM). Crystal structure analysis shows a 42 Å Eucledian distance between canonical carbohydrate binding sites (Pdb 1HGG). A bivalent ligand must accommodate the slightly convex protein surface (Fig. S29) and should therefore present the glyco ligands at distances > 45 Å, which is in agreement with the observed maximum in binding affinity at 52 Å. Of note, spatial screening revealed a further, albeit less preferred arrangement for sugars separated by 26 Å (Kd = 210 µM). We note that previous reports have discussed crystallographic evidence for a second, very low affinity carbohydrate binding site at approx. 22 Å distance from the nearest canonical binding site through protein space (see also Fig. S29).26 Next, we assessed the PEG-linked sialyl-LacNAc conjugates. In sharp contrast to the nucleic acid-based bivalent displays, the bivalent PEG sugar conjugates showed only weak binding affinities being in the range of the monovalent interaction. To characterize the contrasting behavior of DNA- and PEG-based scaffolds in more detail we assessed the interaction of monovalent constructs with bHA. The DNA·PNA-sugar conjugate selfassembly NAm had a 7-fold higher affinity for bHA than the best monovalent PEG-linked trisaccharide PEGm1. On the one hand, the PEG chain seems to hamper interactions between the glyco ligand and the bHA. On the other hand, the DNA·PNA scaffold seems to slightly increase (3.6-fold) the affinity for bHA, which may be due to favorable hydrophobic interactions or, perhaps, a phosphodiester mediated increase of the Ca2+ concentration in the vicinity of the calcium-dependent sugar binding sites. However, regardless of the reasons for the differential binding affinity upon changes in the periphery of the glyco ligand, an important question is why bivalency induces binding enhancements for the matched DNA·PNA construct NA52 but not for the PEGconjugate PEG50, even though both systems arrange the two sialyl-LacNAc residues at approx. identical mean distances. Analysis of bivalent interactions with soluble hemagglutinin trimers by a statistical mechanics model. In order to understand the role of spacer length and flexibility we modeled the bivalent binding by means of simple polymer models. For a quantitative analysis, all possible binding modes between HA and the bivalent compound have to be considered. We examined three different binding modes by which a bivalent compound can bind to canonical sugar binding sites or to a canonical

Page 4 of 17

Figure 3. Schematic cartoon of the interactions between a bivalent ligand and a trimeric receptor (A-C). The bivalent ligand may bind (A) in a monovalent fashion, (B) connect two neighboring canonical binding sites or (C) bridge canonical and putative secondary binding sites. (D) Effective concentration ceff for intratrimeric interactions of HA (see (B)) at a fixed separation (dcan = 52.1 Å) of binding sites with ligands displayed by PNA·DNA spacers (blue line) or PEG spacers (red) as a function of the average end-to-end distance. The dashed lines refer to bridging interactions between two adjacent HA trimers. In subfigure (D) and (E) the same data is shown, but the plotted range of ceff differs.

and a potentially existing low affinity secondary binding site of HA (Fig. 3). A bivalent compound can bind to a single HA trimer in a monovalent (Figure 3A) or bivalent fashion (Figure 3B,C). We only consider monovalent complexes that involve binding of the canonical binding sites. Monovalent interactions with a putative secondary binding site are neglected, due to their expectedly weak binding affinity. For bivalent interactions, we consider bridging of two canonical binding sites on the top of HA (Figure 3B) as well as a link between canonical and putative secondary binding sites (Figure 3c), where the large dissociation constant of the secondary binding site would be compensated by the bivalency effect. Each of the binding modes exhibits a different dissociation constant, which we briefly discuss in the following.52,53 The dissociation constant KD,mv of the monovalently bound conjugate is given by K D,mv =

K mono (1), ∆Gspacer exp � �α kB T

where Kmono is the dissociation constant of the monovalent ligand in unconjugated form and ∆Gspacer describes interactions between spacer and receptor. The parameter α accounts for the steric repulsion between spacer and receptor. The factor α adopts values between 0 and 1, in which 0 corresponds to the case that the spacer completely inhibits the binding, while α = 1 describes the hypothetical case that the spacer does not at all inhibit binding. For a rigid spacer such as the PNA·DNA duplex

4

ACS Paragon Plus Environment

Page 5 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

the factor α is approximated by α = 0.5. This corresponds to a rigid rod that is attached to a plane and can only move in the upper half space. From the measurements of the trisaccharide in conjugated (KD (NAm) = 0.86 mM) and unconjugated form (KD(SLN) = 3.1 mM) we extract an additional interaction energy of ∆Gspacer =1.98kBT for conjugates with PNA·DNA spacers. In case of a monovalent PEG complex, the factor α depends on the spacer length. We model the monovalently bound PEG complex as a flexible chain at an impenetrable wall, which leads to the following expression (c.f. SI Section 12) for the factor α: 3

2 2 𝑎𝑎 (2), α = 6� � 3𝜋𝜋 rete

with a the length of the linker between the trisaccharide and the PEG chain, which we approximate by a = 10 Å and the mean end-to-end distance of the chain rete which is estimated from the Flory model. The experimentally measured dissociation constant of the monovalent PEG complexes is well reproduced without an additional interaction energy (∆Gspacer = 0). Inserting equation (2) into equation (1) with ∆Gspacer = 0, we calculate a dissociation constant for the monovalent PEG complex with 3 PEG monomers of KD(PEGm1) = 4.0 mM, which is in the same range as the experimental value (KD(PEGm1) = 5.9 mM, Table 1). The second binding mode involves bivalent binding to two canonical binding sites (Figure 3B). The dissociation constant reads: K D,bv,can =

K 2mono

∆Gspacer exp � � 2π ceff (dcan ) kBT

(3),

where ceff(dcan) is the effective concentration that describes the probability to bridge two canonical binding sites and dcan the distance between the binding sites. The factor 2π accounts for the rotational constraints of the trisaccharides with respect to each other. While two monovalent trisaccharides can adopt any orientation with respect to each other, they are pre-orientated in the bivalent complex due to the connection with the spacer. If the ligand is connected to a spacer via a chemical bond, the angle of this bond is fixed. We therefore approximate the reduction of the rotational degrees of freedom of the second trisaccharide with respect to the first trisaccharide by 2π. In analogy to the second binding mode, the dissociation constant of the third binding mode, which describes the bridging of a canonical and a secondary binding site (Figure 3C), reads: 𝐾𝐾𝐷𝐷,𝑏𝑏𝑏𝑏,𝑠𝑠𝑠𝑠𝑠𝑠 =

𝐾𝐾𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 𝐾𝐾𝑠𝑠𝑠𝑠𝑠𝑠 (4), ∆𝐺𝐺𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑒𝑒𝑒𝑒𝑒𝑒 � � 2𝜋𝜋 𝑐𝑐𝑒𝑒𝑒𝑒𝑒𝑒 (𝑑𝑑𝑠𝑠𝑠𝑠𝑠𝑠 ) 𝑘𝑘𝐵𝐵 𝑇𝑇

with Ksec the dissociation constant of the secondary binding site and dsec the distance between the canonical

and secondary binding sites. Note, that there are two possible distances dsec between the binding sites, corresponding to the bridging of two binding sites on the same HA subunit (Figure 3C right) or neighboring HA subunits (Figure 3C left). To evaluate the dissociation constant in each binding mode quantitatively we have to discuss the effective concentration, which is the underlying cause for bivalency enhancement and depends on the length and flexibility of the bivalent compound.52 We describe the flexible PEG spacer by a Gaussian chain with mean endto-end distance rete and the rigid PNA·DNA scaffold by a harmonic spring with an average length rete which can fluctuate by ∆r around the average length.52 A detailed derivation of the effective concentration is presented in the Supporting Information. For the rigid PNA·DNA scaffold the effective concentration reads:

𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑐𝑐𝑒𝑒𝑒𝑒𝑒𝑒 (𝑑𝑑) =

1

3

2(2𝜋𝜋)2

𝑒𝑒𝑒𝑒𝑒𝑒 �−

1 (𝑑𝑑 − 𝑟𝑟𝑒𝑒𝑒𝑒𝑒𝑒 )2 � 2 𝛥𝛥𝛥𝛥 2 (5), 2 𝛥𝛥𝛥𝛥 𝑟𝑟𝑒𝑒𝑒𝑒𝑒𝑒

with d the distance between the binding sites, rete the distance between the ligands along the bivalent compound and ∆r the length fluctuations of the bivalent compound. For the flexible PEG scaffold we obtain: flex (d) = ceff

3 d 2 8√6a2 exp �− � � � (6). 5 2 rete 𝜋𝜋 7/2 rete

In Figures 3D and 3E the effective concentration of a PNA·DNA and a PEG scaffold is shown for d=dcan = 52.1 Å and ∆r = 4.1 Å. As is explained below, these parameters were extracted from experimental data obtained by DNAbased spatial screening. The effective concentration of the PNA·DNA scaffold reaches values in the mM range, while the effective concentration of the PEG scaffold is limited to the µM range (see Fig. 3E). These considerations lead to an important conclusion. A PEG-based spacer will only provide for bivalency enhancements when the ligands have HA affinities in the lower µM range, which is not the case for the sialylLacNAc ligand. The modeling of the distance-affinity relationships provides a more detailed picture. Knowing the dissociation constant in each binding mode (given by equations 1,3 and 4), we calculated the concentration of occupied binding pockets by summing over all possibilities by which one to three ligands can bind to HA. A detailed summary of combinations that contribute to the concentration of occupied binding pockets is presented in the Supporting Information (Figure S28). To describe the MST measurements, we assume that the apparent dissociation constant is reached, when half of all canonical binding sites are occupied. This definition reflects that the MST signal is proportional to the concentration of bound receptor. A comparative evaluation must consider potential interactions of the linker with the protein surface. The analysis of crystal

5

ACS Paragon Plus Environment

Journal of the American Chemical Society

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

structure data revealed a rather even distribution of charge and hydrophobicity on the HA surface (Figure S29), suggesting rather weak linker-surface interactions. This is in line with the analysis of the dissociation constant of monovalent ligands in unconjugated form and conjugated to PNA·DNA scaffold (NAm) from which we extract the additional interaction energy ∆Gspacer =1.98kBT, which is applied in the calculations. The remaining unknown parameters dcan, dsec, ∆r and Ksec were determined from a fit (solid green lines in Figure 4) of the apparent dissociation constant, based on the effective concentration (equation 5) to the experimental data (Figure 4, green squares). The shaded areas of the

Figure 4. Distance-affinity relationship for interactions between bivalent PNA·DNA complexes (green) or bivalent PEG conjugates (red) and soluble hemagglutinin trimers. Filled squares represent data points obtained by MST measurements. Solid lines depict the results of the analysis by statistical mechanics. The shaded areas visualize the uncertainty resulting from the statistical error of the measured dissociation constant Kmono.

modeled (for PEG conjugates, red line) and fitted (for PNA·DNA complexes, green line) apparent dissociation constant reflect the uncertainty of the measured dissociation constant Kmono. We assign the deep minimum in Fig. 4 (green curve) to bivalent binding of two canonical binding sites on the top of HA with dcan = 52±1 Å and ∆r = 4.1±0.2 Å. This concurs with the 45−52 Å spacer length required to present the anomeric centers of the two glyco ligands across the slightly convex HA surface (Figure S30). In light of a previous report on secondary binding sites of HA26, the shallow minimum at 26 Å may be attributed to a bridging interaction between the canonical and a secondary binding site. Fitting suggests dsec1 = 28±3 Å for bridging of binding sites on the same HA subunit and dsec2 = 63±20 Å for bridging two binding sites on neighboring HA subunits. Based on these premises the dissociation constant of the secondary binding site is two orders of magnitude larger than the dissociation constant of the canonical binding site (Ksec=200±90 mM vs. Kmono = 3 mM). Given this weak affinity, it would be challenging to detect binding to the secondary binding site by using monovalent compounds. The analysis suggests a potentially generic method for estimating the dissociation constant of low affinity binding sites.

Page 6 of 17

With Kmono = 3100±250 µM known and using average endto-end distances provided by the elastic polymer model, the apparent dissociation constant of bivalent ligands with PEG spacer (Fig. 4, red data points) is modeled parameter free, based on the effective concentration (equation 6). An important result is that − on the basis of theoretical considerations − there will be no PEG spacer providing for bivalency enhanced recognition of sialylLacNAc by soluble bHA (red line, Figure 4). The increase in effective molarity that can be induced by PEG bridges is not sufficient to overcome the weak affinity of this glyco ligand. It is instructive to compare this result with data obtained by Knowles et al.31 They conjugated sialic acid to 37-49 atoms long PEG-type spacers. Based on the elastic polymer model, the PEG conjugates used by Knowles at al. will arrange the sialic acid residues in 11-13 Å averaged distance. Like us, they found no bivalency enhancement of interactions with bHA. While this seems quite plausible in light of the 42 Å distance (through protein space) between two adjacent sugar binding sites (and 23 Å distance between a canonical and the nearest secondary binding site, Pdb: 1HGG, see also Figure S30), our results with extended PEG chains point to a general limitation when PEG spacers are used to bridge distant low affinity binding sites. Given the rather modest increases in effective molarity provided by PEG, the second binding site is too weak to afford bivalency enhancements. Spatial screening of hemagglutinin on IAV by microscale thermophoresis and statistical mechanics models. We next applied the spatial screening to a less defined environment, i.e. HAs on the viral surface. CryoTEM measurements revealed a 102 Å distance between the centers of two HA trimers on the X31 virus (see Supporting Information for details). Given the uncertainty of the orientation of the HA trimers around their vertical axes, the distance between two sugar binding sites of two adjacent HA trimers may range anywhere between 49 Å (min) and 154 Å (max). Nearest neighbors will be arranged in 49 − 66 Å distance. The DNA-programmed bivalent probing unraveled, again, a distinct bimodal distance-affinity relationship (Table 1, see also blue squares in Figure 5A) showing affinity optima at 8 and 16 nucleotides distance (~26 and ~52 Å, respectively) between the two trisaccharides. Compared to the measurements with bHA trimers (see Figure 4) the binding curve appeared broader around the second optimum at 52 Å and high affinity for HA in the X31 virus (KD < 35 µM) was obtained up to 62 Å distance. The first affinity optimum at 26 Å may be caused by recognition of a canonical and a secondary binding site. The second optimum at 52 Å very likely is the hallmark of intramolecular bivalent recognition of the canonical sugar binding sites. However, the broadening of the binding curve may indicate contributions from intermolecular binding modes (HA-HA bridging). One noticeable observation is that the bivalent complexes NA23-NA101 had higher affinity for the virus than for bHA. However, a similar effect was observed for the monovalent sialyl-LacNAc derivative α2,6-SLN

6

ACS Paragon Plus Environment

Page 7 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

Figure 5. Distance-affinity relationship for binding of bivalent conjugates to Influenza X31 virus. (A) Filled squares show KD values for interactions of bivalent PNA·DNA complexes determined by MST measurements (see Table 1). The solid blue line represents the statistical mechanics calculation based on intratrimeric binding. The shaded areas correspond to the uncertainty of the experimentally measured monovalent dissociation constant. (B) KiHAI values for inhibition of Influenza X31-induced hemagglutination by bivalent PNA·DNA complexes (blue) or bivalent PEG conjugates (red) (see Table 1). The open squares represent the highest measured concentrations at which inhibition of hemagglutination was still not achieved.

(Kd,app (X31) = 1760±340 µM vs. Kd,app (bHA) = 3100±250 µM). IAVs also display neuraminidase (NA) which has a sialic acid binding site. However, owing to the low density of NA on the viral surface (40 - 50 copies NA vs. 300 – 400 copies HA)8 and in account of previous investigations, in which activity of HA binders remained unchanged upon inactivation of neuraminidase inhibitors31,54, this enzyme should contribute little, if anything, to the binding of the trisaccharide to the virus.31,54 The additional binding opportunities provided by bridging of two adjacent HA trimers could, in principle, explain the improved affinity of the bivalent systems for the virus. To compare intra- and inter-HA binding, we 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 show in Figure 3D the effective concentration 𝑐𝑐𝑒𝑒𝑒𝑒𝑒𝑒 for the bridging of two adjacent HA trimers by PNA·DNA spacers (black, dashed line). We averaged the effective concentration (equation 5) over all possible relative orientations around the vertical HA axes, assuming an average distance of 102 Å between neighboring HA trimers. The effective concentration indeed exhibits a

maximum around 50 Å very similar to the intratrimeric curve, which might contribute to the broadening of the 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 affinity curve in Figure 5A. Furthermore, 𝑐𝑐𝑒𝑒𝑒𝑒𝑒𝑒 exhibits a broad second maximum around 90 Å (Figure 3D). The analysis suggests that the bivalent complex NA101, which arranges the sugar ligands within rete = 101 Å, should 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 provide for a 𝑐𝑐𝑒𝑒𝑒𝑒𝑒𝑒 above the KD of the monovalent interaction, and hence should result in a low KD,app. However, the KD,app determined by MST measurements is the highest within the series of bivalent DNA·PNA sugar conjugates (Table 1, Figure 5A). We surmise that the impact of the large effective concentration is diminished by additional effects. Individual HA trimers diffuse along the virus surface55, which would be impaired by bridging of two HA trimers. Quantifying the entropic penalty that is associated with the restriction of lateral diffusion would allow to estimate the binding affinity of inter-HA binding. Such an analysis would require a detailed understanding of the conformational space available to two cross-linked HA trimers, which is beyond the scope of this work. To evaluate the impact of intra-HA binding, we use the same parameters for dcan, dsec and ∆r that were fitted for binding to bHA in Fig. 4. Furthermore, we assume that the binding energy ∆Gspacer = 1.98kBT and the ratio between the dissociation constants of canonical and the secondary binding sites Kmono/Ksec is the same for bHA and the entire virus. The only difference in the theoretical curves in Fig. 4 and 5A results from the difference of the monovalent dissociation constant measured for bHA and the X31 virus. The theoretical dissociation constant, which we determine without additional fit parameters, is shown in Fig. 5A as solid line. The shaded areas correspond to the uncertainty that arises due to the deviation of the monovalent dissociation constant obtained by experiment. The theoretical model, which only accounts for intra-HA binding, describes the experimentally obtained binding properties remarkably well. Apparently, inter-HA interactions seem to play only a negligible role. Rather, we believe that the improved interaction of the glyco ligand is due to differences of the solvation energies between the relatively small HA in solution and HA on IAVs, which influence the overall binding energy.56 Spatial screening of hemagglutinin on IAVs by the hemagglutination inhibition assay. For confirmation of the results obtained with the MST assay, we also explored compounds in the hemagglutination inhibition (HAI) assay (Figure 5B). The monovalent, unconjugated sugar required a 50-60 mM concentration to fully inhibit virus-induced hemagglutination of human erythrocytes. The inhibitory potency of the bivalent DNA·PNA sugar displays NA23-NA101 was up to three orders of magnitude higher. Most efficient inhibition was achieved when the two sugar residues were separated by 7 (~26 Å) or 15 – 20 nucleotides (~52 Å – 68 Å). Complex NA59 showed at least one order of magnitude better inhibition than the complexes NA85 and NA101 with significantly longer distances between sugar ligands. We note, though, that the optimal binding occurred at 59 Å rather than 52 Å. We assign these differences to the experimental setups.

7

ACS Paragon Plus Environment

Journal of the American Chemical Society

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The MST assay monitors the formation of complexes between virus particles and sugar ligands. In contrast, the HAI assay monitors inhibition of a lattice structure that forms when virus particles agglutinate erythrocytes. These processes may be affected by steric shielding in different ways. Regardless of the differences, the similarity of distance-affinity profiles revealed by the two distinct assays is noteworthy (compare Fig. 5A with Fig. 5B). In addition, both, the MST assay and the HAI assay show the lack of a clear distance-dependent interaction for PEG-bridged sialyl-LacNAc conjugates. According to the HAI assay, the PEG-linked bivalent constructs PEG29 and PEG50 were 10-fold more potent than the unconjugated sugars. Though this may be an indicator for a minor bivalency enhancement in this assay, the PEGbased systems inhibit with less than 1% of the potency of the best PNA·DNA sugar construct. Although the overall structure is preserved, evolution of IAV is associated with modifications of the HA ectodomain including the receptor binding site. This has an effect on interactions with sialic acid containing sugars. As a result, multivalent glyco ligands may be able to discriminate between different IAV strains. We tested viral strains from the antigenic type A belonging to the H3N2, H1N1 and the H7N1 viral serotypes. Hemagglutinin from A/strain mute swan (H7N1) preferentially binds α2,3-linked sialyl-LacNAc. As expected, the bivalent α2,6sialyl-LacNAc construct NA59 did not show any inhibitory activity at the highest concentration (500 µM) tested (Table 2). In this assay, the unconjugated α2,6Table 2: NA59-induced inhibition of hemagglutination by different viral strains from the antigenic type A. Influenza strains

Receptor pref.

KiHAI / µM (NA59)

KiHAI / µM (2,6 SL)

A/Aichi/1968/H3N2(X31)

α2,6

44

50000

A/Panama/1999/ H3N2

α2,6

63

50000

α2,6

>500

50000

>500

>50000

A/Puerto Rico/8/1934/H1N1

α2,3 A/Mute swan/Germany/2006/H7N1L P

α2,3

sialyl-lactose served as a positive control for inhibiting hemagglutination by the strains X31 (H3N2), Panama (H3N2) and PR8 (H1N1). With a Ki(HAI) = 50 mM, the sugar did not distinguish between these virus strains (Table 2). For inhibition of X31 and Panama, both belonging to H3N2, the bivalent construct NA59 was equally efficient. Surprisingly, hemagglutination by PR8 (H1N1) was not inhibited by NA59 within the tested concentration range (0.2-500 µM). The at least 10-fold specificity for inhibition of H3N2 is noteworthy given the non-selective interactions with the unconjugated α2,6-

Page 8 of 17

sialyl-lactose ligand. The results are consistent with reports from Baker et al., who reported improvements of strain specificity upon presentation of 64 sialic acid residues on a polyamidoamine (PAMAM) dendrimer.19 Our data shows, for the first time, strain specific binding can also be achieved by using only two, appropriately arranged glyco ligands. We surmise that the increase in strain specificity may be due to differences in the surface topographies. It remains to be shown, whether adjustment of distances between the two 2,6-sialylLacNAc residues on the PNA·DNA scaffold will provide for binders with specificity for the PR8 virus. In future work, we will evaluate a variety of IAV strains including the PR8 by DNA-programmed spatial screening to clarify whether different virus strains arrange the sialoside binding sites in different distances. Conclusion In our work, we assessed the distance-affinity relationship for the interaction of trimeric Influenza A virus HA with bivalent displays of sialyl-LacNAc, which is the natural ligand on host cells recognized by the HA. The study is the first to cover a 20 – 100 Å range of distances required to probe the feasibility of binding enhancement upon simultaneous engagement of i) canonical and secondary binding sites within a single HA trimer, ii) two intratrimeric canonical binding sites or iii) two canonical binding sites across adjacent HA trimers. We compared two entirely different scaffold architectures based on flexible, 27 − 144 monomer units long polyethylene glycol spacers or rigid, self-assembled peptide nucleic acid (PNA)·DNA complexes. Measurements of microscale thermophoresis (MST assay) revealed a lack of bivalency enhancement for interactions of bivalent PEG conjugates with soluble HA ectodomain and HA on the virus surface. A rather insignificant binding enhancement by one order of magnitude was obtained when the bivalent interactions were assessed by means of a hemagglutination assay (HAI assay). A statistical mechanics model pointed to the general limitations of PEG spacers. We modeled the effective concentration of the sugar ligand at the second binding site after engagement of the first, monovalent interaction. A 67 monomer units long PEG scaffolds presented the sialyl-LacNAc ligands with an average endto-end distance rete = 50 Å, which should, in principle, suit the distance between the canonical sugar binding sites on HA. However, the flexibility of the PEG spacers diminishes the effective concentration to values ceff ≤ 102 µM which is significantly lower than the KD = 3 mM of the monovalent interaction. As a result, PEG spacers are inappropriate to provide bivalency enhancements when the ligands have only millimolar affinity. In stark contrast, 52 Å long spacers based on DNA scaffolds are expected to furnish effective concentration in the millimolar range. According to this analysis, a size-matched DNA scaffold affords 102-fold higher effective concentration (at 52 Å distance between binding sites) than PEG spacers, despite similar end-to-end length. The spatial screening with DNA·PNA scaffolds unraveled a bimodal distance-affinity relationship, which

8

ACS Paragon Plus Environment

Page 9 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

was observed with both soluble HA and HA on the virus surface. The highest binding affinity (MST assay: KD,app (biv) = 10 µM vs. KD,app (mono) = 1760 µM; HAI assay: Ki,HAI (biv) = 44 µM vs. Ki,HAI (mono) = 60.000 µM) was obtained when the sialyl-LacNAc ligands were separated by a 52-59 Å long DNA·PNA scaffold. This binding mode was assigned to the simultaneous engagement of two canonical sugar bindings within a single HA trimer. Perhaps surprisingly, DNA-programmed spatial screening unraveled a second affinity maximum at 26 Å distance between the sugar attachment points at the PNA·DNA scaffold. The binding optimum was observed for soluble HA ectodomain and for virus-surface HA (in two different assays). We assumed the existence of a secondary binding site, reported earlier by Sauter et al. A fit of the apparent dissociation constant based on the distance dependence of the effective concentration would suggest a KD (sec) = 200 mM, which is two orders of magnitude larger than the dissociation constant of the canonical binding site. Such weak interactions would be unnoticeable by using flexible scaffolds. However, we wish to emphasize that unambiguous proof for the existence of secondary binding sites is still awaiting. In this study, we used PNA·DNA complexes for spatial screening. Given our previous exploration of bivalency enhanced interactions between peptides or small molecules with tandem Src homology 2 domains or estrogen receptor32,38,57, respectively, we expect that spatial screening of carbohydrate-protein interactions should also extend to DNA·DNA complexes. This would facilitate the design of structurally defined multivalent displays based on, for example, branched DNA, DNA three-way or four-way junctions. Our data suggests that precision scaffolds may allow for the design of subtype-specific virus inhibitors. We found that an optimized bivalent binder inhibited hemagglutination induced by X31 (H3N2) at least ten times better than of PR8 (H1N1). At this stage, we do not know whether the discrimination is due to slight variations of the arrangement of binding sites or to differential interactions between the linker and the protein surface. However, it seems feasible to develop specific inhibitors or detection systems, which are able to discriminate between different HA serotypes. We combined DNA-programmed spatial screening and modeling in order to explore which role intertrimeric HA-HA bridging may play in providing for bivalent affinity enhancements. However, we found that bivalent interactions within a single HA trimer adequately describe the binding of bivalent DNA·PNA complexes to HA on the viral surface. From these results, we infer that multivalent presentation of optimized intratrimeric HA binders should provide for highest binding enhancements. In current work, we evaluate the DNAcontrolled multivalent presentation of optimized bivalent HA binders.

ASSOCIATED CONTENT Supporting Information. Experimental procedures and characterization of the sugar ligand, PNA- and PEG-sugar-

conjugates and nucleic acid complexes; preparation of viruses and bHA; analysis of microscale thermophoresis and hemagglutination inhibition assay measurements; details about the statistical mechanics models and determination of the distances of the HA trimers by Cryo-TEM. This material is available free of charge via the Internet at http://pubs.acs.org.

AUTHOR INFORMATION Corresponding Author *E-mail: [email protected]

Author Contributions The manuscript was written through contributions of all authors. / All authors have given approval to the final version of the manuscript.

Funding Sources This work was financially supported by the Deutsche Forschungsgemeinschaft (CRC 765).

Notes

The authors declare no competing financial interest.

ACKNOWLEDGMENT We appreciate funding from Forschungsgemeinschaft (CRC 765).

the

Deutsche

REFERENCES (1) Wittmann, V.; Pieters, R. J.. Chem. Soc. Rev. 2013, 42, 4492. (2) Fasting, C.; Schalley, C. A.; Weber, M.; Seitz, O.; Hecht, S.; Koksch, B.; Dernedde, J.; Graf, C.; Knapp, E. W.; Haag, R. Angew. Chem. Int. Ed. 2012, 51, 10472. (3) Kiessling, L. L.; Gestwicki, J. E.; Strong, L. E. Angew. Chem. Int. Ed. 2006, 45, 2348. (4) Muller, C.; Despras, G.; Lindhorst, T. K. Chem. Soc. Rev. 2016, 45, 3275. (5) Mammen, M.; Choi, S. K.; Whitesides, G. M. Angew. Chem. Int. Ed. 1998, 37, 2755. (6) Bhatia, S.; Cuellar, L.; Haag, R. J. Am. Chem. Soc. 2016, 138, 8654. (7) Wiley, D. C.; Skehel, J. J. Annu. Rev. Biochem. 1987, 56, 365. (8) Harris, A.; Cardone, G.; Winkler, D. C.; Heymann, J. B.; Brecher, M.; White, J. M.; Steven, A. C. Proc. Natl. Acad. Sci. USA 2006, 103, 19123. (9) Weis, W.; Brown, J. H.; Cusack, S.; Paulson, J. C.; Skehel, J. J.; Wiley, D. C. Nature 1988, 333, 426. (10) Sauter, N. K.; Bednarski, M. D.; Wurzburg, B. A.; Hanson, J. E.; Whitesides, G. M.; Skehel, J. J.; Wiley, D. C. Biochemistry 1989, 28, 8388. (11) Lees, W. J.; Spaltenstein, A.; Kingerywood, J. E.; Whitesides, G. M. J. Med. Chem. 1994, 37, 3419. (12) Gambaryan, A. S.; Tuzikov, A. B.; Chinarev, A. A.; Juneja, L. R.; Bovin, N. V.; Matrosovich, M. N. Antivir. Res. 2002, 55, 201. (13) Totani, K.; Kubota, T.; Kuroda, T.; Murata, T.; Hidari, K. I. P. J.; Suzuki, T.; Suzuki, Y.; Kobayashi, K.; Ashida, H.; Yamamoto, K.; Usui, T. Glycobiology 2003, 13, 315. (14) Wang, H.; Huang, W.; Orwenyo, J.; Banerjee, A.; Vasta, G. R.; Wang, L. X. Bioorg. Med. Chem. 2013, 21, 2037. (15) Umemura, M.; Itoh, M.; Makimura, Y.; Yamazaki, K.; Umekawa, M.; Masui, A.; Matahira, Y.; Shibata, M.; Ashida, H.; Yamamoto, K. J. Med. Chem. 2008, 51, 4496. (16) Sigal, G. B.; Mammen, M.; Dahmann, G.; Whitesides, G. M. J. Am. Chem. Soc. 1996, 118, 3789.

9

ACS Paragon Plus Environment

Journal of the American Chemical Society

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(17) Choi, S. K.; Mammen, M.; Whitesides, G. M. J. Am. Chem. Soc. 1997, 119, 4103. (18) Kwon, S. J.; Na, D. H.; Kwak, J. H.; Douaisi, M.; Zhang, F.; Park, E. J.; Park, J. H.; Youn, H.; Song, C. S.; Kane, R. S.; Dordick, J. S.; Lee, K. B.; Linhardt, R. J. Nat. Nanotechnol. 2017, 12, 48. (19) Landers, J. J.; Cao, Z. Y.; Lee, I.; Piehler, L. T.; Myc, P. P.; Myc, A.; Hamouda, T.; Galecki, A. T.; Baker, J. R. J. Infect. Dis. 2002, 186, 1222. (20) Roy, R.; Zanini, D.; Meunier, S. J.; Romanowska, A. J. Chem. Soc. Chem. Comm. 1993, 1869. (21) Zhang, Z.; Schepens, B.; Nuhn, L.; Saelens, X.; Schotsaert, M.; Callewaert, N.; De Rycke, R.; Zhang, Q.; Moins, S.; Benali, S.; Mespouille, L.; Hoogenboom, R.; De Geest, B. G. Chem. Commun. 2016, 52, 3352. (22) Weissleder, R.; Kelly, K.; Sun, E. Y.; Shtatland, T.; Josephson, L. Nat. Biotechnol .2005, 23, 1418. (23) Papp, I.; Sieben, C.; Ludwig, K.; Roskamp, M.; Bottcher, C.; Schlecht, S.; Herrmann, A.; Haag, R. Small 2010, 6, 2900. (24) Vonnemann, J.; Sieben, C.; Wolff, C.; Ludwig, K.; Bottcher, C.; Herrmann, A.; Haag, R. Nanoscale 2014, 6, 2353. (25) Waldmann, M.; Jirmann, R.; Hoelscher, K.; Wienke, M.; Niemeyer, F. C.; Rehders, D.; Meyer, B. J. Am. Chem. Soc. 2014, 136, 783. (26) Sauter, N. K.; Glick, G. D.; Crowther, R. L.; Park, S. J.; Eisen, M. B.; Skehel, J. J.; Knowles, J. R.; Wiley, D. C. Proc. Natl. Acad. Sci. USA 1992, 89, 324. (27) Scheibe, C.; Wedepohl, S.; Riese, S. B.; Dernedde, J.; Seitz, O. Chembiochem 2013, 14, 236. (28) Singh, D. D.; Saikrishnan, K.; Kumar, P.; Surolia, A.; Sekar, K.; Vijayan, M. Glycobiology 2005, 15, 1025. (29) Lee, R. T.; Shinohara, Y.; Hasegawa, Y.; Lee, Y. C. Bioscience Rep. 1999, 19, 283. (30) Silva-Martin, N.; Bartual, S. G.; Ramirez-Aportela, E.; Chacon, P.; Park, C. G.; Hermoso, J. A. Structure 2014, 22, 1595. (31) Glick, G. D.; Knowles, J. R. J. Am. Chem. Soc. 1991, 113, 4701. (32) Abendroth, F.; Bujotzek, A.; Shan, M.; Haag, R.; Weber, M.; Seitz, O. Angew. Chem. Int. Ed. 2011, 50, 8592. (33) Englund, E. A.; Wang, D. Y.; Fujigaki, H.; Sakai, H.; Micklitsch, C. M.; Ghirlando, R.; Martin-Manso, G.; Pendrak, M. L.; Roberts, D. D.; Durell, S. R.; Appella, D. H. Nat. Commun. 2012, 3. (34) Matsuura, K.; Hibino, M.; Yamada, Y.; Kobayashi, K. J. Am. Chem. Soc. 2001, 123, 357. (35) Machida, T.; Novoa, A.; Gillon, E.; Zheng, S. S.; Claudinon, J.; Eierhoff, T.; Imberty, A.; Romer, W.; Winssinger, N. Angew. Chem. Int. Edit. 2017, 56, 6762.

Page 10 of 17

(36) Gorska, K.; Huang, K. T.; Chaloin, O.; Winssinger, N. Angew. Chem. Int. Ed. 2009, 48, 7695. (37) Scheibe, C.; Bujotzek, A.; Dernedde, J.; Weber, M.; Seitz, O. Chem. Sci. 2011, 2, 770. (38) Eberhard, H.; Diezmann, F.; Seitz, O. Angew. Chem. Int. Ed. 2011, 50, 4146. (39) Diezmann, F.; von Kleist, L.; Haucke, V.; Seitz, O. Org. Biomol. Chem. 2015, 13, 8008. (40) Paar, J. M.; Harris, N. T.; Holowka, D.; Baird, B. J. Immunol. 2002, 169, 856. (41) Rinker, S.; Ke, Y.; Liu, Y.; Chhabra, R.; Yan, H. Nat. Nanotechnol .2008, 3, 418. (42) Williams, B. A.; Diehnelt, C. W.; Belcher, P.; Greving, M.; Woodbury, N. W.; Johnston, S. A.; Chaput, J. C. J. Am. Chem. Soc. 2009, 131, 17233. (43) Gorska, K.; Beyrath, J.; Fournel, S.; Guichard, G.; Winssinger, N. Chem. Commun. 2010, 46, 7742. (44) Janssen, B. M. G.; Lempens, E. H. M.; Olijve, L. L. C.; Voets, I. K.; van Dongen, J. L. J.; de Greef, T. F. A.; Merkx, M. Chem. Sci. 2013, 4, 1442. (45) Nielsen, P. E.; Egholm, M.; Berg, R. H.; Buchardt, O. Science 1991, 254, 1497. (46) Eriksson, M.; Nielsen, P. E. Nat. Struct. Biol. 1996, 3, 410. (47) Protozanova, E.; Yakovchuk, P.; Frank-Kamenetskii, M. D. J. Mol. Biol. 2004, 342, 775. (48) Devanand, K.; Selser, J. C. Macromolecules 1991, 24, 5943. (49) Liese, S.; Gensler, M.; Krysiak, S.; Schwarzl, R.; Achazi, A.; Paulus, B.; Hugel, T.; Rabe, J. P.; Netz, R. R. ACS Nano 2017, 11, 702. (50) Brand, C. M.; Skehel, J. J. Nat. New. Biol. 1972, 238, 145. (51) Xiong, X. L.; Coombs, P. J.; Martin, S. R.; Liu, J. F.; Xiao, H. X.; McCauley, J. W.; Locher, K.; Walker, P. A.; Collins, P. J.; Kawaoka, Y.; Skehel, J. J.; Gamblin, S. J. Nature 2013, 497, 392. (52) Liese, S.; Netz, R. R. Beilstein J. Org. Chem. 2015, 11, 804. (53) Diestler, D. J.; Knapp, E. W. J. Phys. Chem. C 2010, 114, 5287. (54) Janakiraman, M. N.; White, C. L.; Laver, W. G.; Air, G. M.; Luo, M. Biochemistry 1994, 33, 8172. (55) Reddy, T.; Shorthouse, D.; Parton, D. L.; Jefferys, E.; Fowler, P. W.; Chavent, M.; Baaden, M.; Sansom, M. S. P. Structure 2015, 23, 584. (56) Sedlmeier, F.; Netz, R. R. J. Chem. Phys. 2013, 138. (57) Marczynke, M.; Groger, K.; Seitz, O. Bioconjugate Chem. 2017, 28, 2384.

10

ACS Paragon Plus Environment

Page 11 of 17

Journal of the American Chemical Society

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 ACS Paragon Plus Environment

11

Journal of the American Chemical Society

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(A) Electron micrography of the human Influenza A virus (X31) presenting the trimeric hemagglutinin on the viral surface. The average distance of the midpoints of two adja-cent HA trimers has been determined as 101.7±0.6 Å (see Supporting Information for further details). The surface displays of the HAs show the canonical sugar binding sites in yellow. The 42 Å distance between two canonical binding pockets on one HA is determined through space. (B) Crystal structure (Pdb 1HGG) of the HA trimer (one monomer in blue) with bound sialosides (yellow: bound to canonical binding sites, red: bound to secondary binding sites). (C) Schematic but to scale display of trimeric HAs on viral mem-brane. As the rotational orientation of the HA trimers is unknown the HA surface displays were cylindrically aver-aged. For illustration, a bivalent PNA·DNA scaffold is shown (upper right). 84x146mm (300 x 300 DPI)

ACS Paragon Plus Environment

Page 12 of 17

Page 13 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

Figure 2. (A) Hybridization of modified (blue, red) and unmodified peptide nucleic acid (PNA) oligomers (green) with DNA templates provides bivalent sialyl-LacNAc-PNA·DNA complexes with a defined sugar-sugar spacing along the nucleic acid backbone. Structure of (B) a bivalent sialyl-LacNAc- PNA·DNA complex and (C) bivalent sialyl-LacNAc-PEG conjugates. 84x88mm (300 x 300 DPI)

ACS Paragon Plus Environment

Journal of the American Chemical Society

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure 3. Schematic cartoon of the interactions between a bivalent ligand and a trimeric receptor (A-C). The bivalent ligand may bind (A) in a monovalent fashion, (B) connect two neighboring canonical binding sites or (C) bridge canonical and putative secondary binding sites. (D) Distance dependence of the effective concentration ceff for intratrimeric interactions of HA (see (B)) with ligands displayed by PNA·DNA spacers (blue line) or PEG spacers (red) in varied average end-to-end distance. The dashed lines refer to bridging interactions between two adjacent HA trimers. In subfigure (D) and (E) the same data is shown, but the plotted range of ceff differs. 84x55mm (300 x 300 DPI)

ACS Paragon Plus Environment

Page 14 of 17

Page 15 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

Figure 4. Distance-affinity relationship for interactions be-tween bivalent PNA·DNA complexes (green) or bivalent PEG conjugates (red) and soluble hemagglutinin trimers. Filled squares represent data points obtained by MST measure-ments. Solid lines depict the results of the analysis by statistical mechanics. The shaded areas visualize the uncertainty resulting from the statistical error of the measured dissocia-tion constant Kmono. 84x60mm (300 x 300 DPI)

ACS Paragon Plus Environment

Journal of the American Chemical Society

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure 5. Distance-affinity relationship for binding of bivalent conjugates to Influenza X31 virus. (A) Filled squares show KD values for interactions of bivalent PNA·DNA complexes determined by MST measurements (see Table 1). The solid blue line represents the statistical mechanics calculation based on intratrimeric binding. The shaded areas correspond to the uncertainty of the experimentally measured monovalent dissociation constant. (B) KiHAI values for inhibition of Influenza X31-induced hemagglutination by bivalent PNA·DNA complexes (blue) or bivalent PEG conjugates (red) (see Table 1). The open squares represent the highest measured concentrations at which inhibition of hemagglutination was still not achieved. 84x130mm (300 x 300 DPI)

ACS Paragon Plus Environment

Page 16 of 17

Page 17 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of the American Chemical Society

83x44mm (300 x 300 DPI)

ACS Paragon Plus Environment