Templated chemistry for bioorganic synthesis and chemical biology

In light of the 2018 Max Bergmann Medal, this review discusses advancements on chemical biology–driven templated chemistry developed in the author's laboratories. The focused review introduces the template categories applied to orient functional units such as functional groups, chromophores, biomolecules, or ligands in space. Unimolecular templates applied in protein synthesis facilitate fragment coupling of unprotected peptides. Templating via bimolecular assemblies provides control over proximity relationships between functional units of two molecules. As an instructive example, the coiled coil peptide–templated labelling of receptor proteins on live cells will be shown. Termolecular assemblies provide the opportunity to put the proximity of functional units on two (bio)molecules under the control of a third party molecule. This allows the design of conditional bimolecular reactions. A notable example is DNA/RNA–triggered peptide synthesis. The last section shows how termolecular and multimolecular assemblies can be used to better characterize and understand multivalent protein‐ligand interactions.


| INTRODUCTION
The molecules used in chemical biology studies face a specificity challenge: The envisioned probe/drug ought to recognize and exert action only upon a selected target that will be a minority compound in the vast space of cell's biomolecules. This problem becomes most pronounced when the probe/drug must not only recognize or bind to its biomolecule target but must also initiate a chemical reaction. Such reactions are used for labelling of biomolecules with reporter groups or affinity tags to visualize their localization and understand their functions in the native environment. [1][2][3][4][5] However, the functional groups offered by biomolecules do not lend themselves to target specific covalent labelling unless a particular microenvironment such as an enzyme's active site facilitates a regioselective reaction that is the basis of activity-based profiling. 6,7 Most biomolecules, however, lack such reactivity enhancing microenvironments. In this case, artificial functional groups can be introduced by metabolic engineering or expanded genetic code technologies. [8][9][10][11][12] Specificity problems also complicate the targeting of specific binding sites that are common to a family of related but functionally distinct biomolecules. This situation frequently emerges for protein-based receptor targets that rely on multivalent interactions with low affinity ligands such as carbohydrates or peptide motive repeats. [13][14][15] One solution to this lack of target specificity is to utilize templates, designer scaffolds that orient functional units in space. The spatial arrangement and not the functional unit per se is crucial for the template action. For example, a template can organize ligands to match the orientation of binding sites of a receptor. Templating can bring functional groups into proximity to direct or enable reactions that would otherwise not occur or at least not in a target-specific fashion.
Prominent examples can be found in nucleic acid directed chemistry where the sequence-specific interactions between complementary oligonucleotides increase the effective molarity of functional groups and enable chemical reactions under dilute conditions where nontemplated biomolecular reactions cannot occur. [16][17][18][19] Focusing on own work, this article intends to categorize and showcase different templating approaches used in bioorganic synthesis and chemical biology.

| TEMPLATES
One way to categorize templating strategies used in bioorganic chemistry and chemical biology refers to the molecularity of the template complex. For simplicity, only two functional units X and Y are shown in Figure 1A. These units may represent (i) functional groups that engage in a templated reaction, (ii) dyes or other reporter groups that engage in distance-dependent interactions, or (iii) ligands or biomolecules.
A unimolecular template connects the functional units within a single molecular entity that may serve as a stand-alone scaffold or may be conjugated with another molecular unit. Typical examples ( Figure 1B) for stand-alone templates are small molecule scaffolds such as cyclodextrins (1), 20 calixarenes (2), 21 and many more as well as larger molecules such as dendrimers (3), [22][23][24] peptides (including oligoproline 4), [25][26][27][28][29] or even entire proteins (8) 30 or polymers (5). 31,32 In another application scenario, conjugated unimolecular templates are frequently used as reaction scaffolds to facilitate chemical reactions such as ligation auxiliaries in protein synthesis via native chemical ligation (NCL) (vide infra). [33][34][35] Templating via bimolecular assembly relies on the mutual interactions between two components. Each component presents a functional unit X or Y. The recognition event defines the proximity relationship between the two components. Instructive examples for bimolecular templates include dual pharmacophore DNA-encoded libraries, 36 proximity-triggered methods for protein labelling, 37,38 and chemical noncovalent protein dimerization. [38][39][40] Nucleic acid-and protein-based molecules can be designed to engage in termolecular or even multimolecular template complexes.
If only two components carry functional units, the third component typically serves as a landing hub that orchestrates the proximity between X and Y. The larger the landing hub component the easier it is to extend this approach to multimolecular complexes. For example, long DNA-type strands have been used to construct multivalent ligand assemblies by recruitment of multiple ligand-modified oligonucleotide strands. 41,42 In some applications, each component interacts with at least two other components. This principle has been used for the design of multivalent carriers on the basis of multimolecular protein assemblies such as viral capsids (9). 43,44 The use of the term "template" implies that the molecular architecture provides a certain degree of structural integrity required for orienting the functional units in space. This is comparably easy to achieve with small molecules such as benzene derivatives, monosaccharides, calixarenes, cyclodextrins, and other macrocycles, which allow confinement of functional units within <20 Å distance.
Dendrimeric structures (3 in Figure 1B) have frequently been used for clustering. [45][46][47][48] A precise presentation of functional units over larger distances is more difficult to achieve. This is the domain of nucleic acid-and protein-based scaffolds, which form accurately defined tertiary structures. For example, DNA duplex (6), triplex, and quadruplex structures are rigid and can be fashioned to have high persistence length over >150 Å distances. 42 DNA origami allows the construction of nanosized templates with sequence-programmed three-dimensional shape. 49,50 Peptide-based coiled coils (7) 51 and oligoproline scaffolds 29 fold into helical structures that enable the spatially defined arrangement of functional units along a two-dimensional track. Although de novo design of protein 3D objects is rapidly improving, 52 folded protein scaffolds (8) from natural sources and viral capsids (9) are, perhaps, currently preferable for a three-dimensional presentation of functional units.

| UNIMOLECULAR TEMPLATES AS LIGATION SCAFFOLDS IN PROTEIN SYNTHESIS
Modern chemical synthesis of proteins depends upon ligations of unprotected peptide fragments, which are easier to handle than protected segments because of their higher solubility. Until the development of NCL 53 chemistry, the selective chemical coupling of two unprotected peptides was a major challenge. NCL reactions involve a C-terminal peptide thioester 10 and a cysteinyl residue 11 at the Nterminus of the C-terminal segment ( Figure 2A). The cysteine residue adopts the function of an intrinsic template. Owing to a chemoselective thiol exchange reaction, the side chain mercapto group captures the acyl component in the form of an intermediary thioester 12, which subsequently rearranges via an intramolecular S-N acyl shift to the ligation product 13. For proteins that lack cysteine, the side chain may be emulated by means of an auxiliary (14) 54 that is appended to the peptide. Mercapto groups for acyl capture have also been anchored via ester bonds to glutamate 55 and glycan residues. 56 The most versatile chemistry has been introduced by Offer et al 57 and Botti et al 34 who connected benzyl-type scaffolds 18 and 19 to the N-terminal amine. The approach is attractive because the ligation auxiliary is introduced in the last step of solid-phase peptide synthesis via reductive alkylation. Preformed amino acid building blocks are not required. Electron-donating substituents at the aryl part facilitate the removal of the ligation template upon acid treatment. However, ligation at the benzyl-type templates proceeds with rather low rates, which restricts their application to glycine-containing ligation junctions. This probably is the reason why an alternative approach, the ligation-desulfurization method, 58-61 gained popularity. The mercapto group is attached directly to the side chain of the N-terminal amino acid in 20 and is removed after ligation by a radical desulfurization 62 reaction. This method allowed faster ligation reactions. A drawback is the additional workload required for the preparation of the thiolated amino acid building blocks.
Motivated by the prospect of extending NCL without the need for new amino acid building blocks, we rethought the ligation auxiliary approach. A suitable ligation template should provide high reactivity for both the thiol exchange and the S → N acyl migratory steps and must lend itself to facile removal under conditions that leave the ligation product unharmed. We systematically analyzed the architecture of the ligation template and found that the substituents at the αand β-positions in 15′ play key roles. 35  concerning because in traditional ligation auxiliaries, α-substitution was a requirement for cleavability (phenyl substitution in benzyl-type template). However, we discovered a potentially general approach for the removal of N-amide-linked mercaptoethyl groups. Nuclear magnetic resonance (NMR) analysis of a C 13 -labelled ligation template suggested that treatment with phosphine in aqueous morpholine at pH 8.5 triggers a radical-induced oxidative fragmentation ( Figure 2B). 35 The investigation led us to the 2-mercapto-2phenethyl (MPE) template, which is the first ligation auxiliary that enables NCL at ligation sites beyond glycine. We have proven the FIGURE 2 A, Native chemical ligation (10 + 11) and auxiliary mediated native chemical ligation (10 + 14). B, Proposed radical-induced oxidative fragmentation of the MPE auxiliary. Products highlighted in green were detected by high-performance liquid chromatography-mass spectrometry (HPLC-MS) and nuclear magnetic resonance (NMR). C, One-pot ligation auxiliary removal in the synthesis of opistoporin-2 (GnHCl, guanidinium hydrochloride; TCEP, triscarboxyethylphosphine) usefulness of MPE in the chemical total synthesis of the 48 amino acid residue peptide opistoporin-2 ( Figure 2C) 63  Genetic fusions with autofluorescent proteins or self-labelling enzymes make protein fluorescent labelling readily available for nonchemists. 65,66 While useful in many cases, the large size (18-33 kDa) of these reporters can impair the functional properties of the protein of interest. This limitation motivated intense research efforts geared towards the development of labelling methods that proceed with smaller tags. 37 We have introduced a peptide-templated labelling reaction that relies on the formation of a peptide coiled coil complex. 67,68 The coiled coil motif involves two (or more) α-helices that wrap around each other. Our work was inspired by contributions from Matsuzaki, who applied the artificial heterodimeric coiled coil peptides E3 and K3, initially developed by Litowski and Hodges, 69 for the noncovalent labelling of cell surface proteins. 70 We equipped the 21 aa long E3 peptide with a N-terminal cysteine residue. This peptide was used as the genetically encoded tag (see Cys-E3-GPCR, Figure 3A). A modified K3 peptide served as the labelling agent. For this purpose, a fluorophore is connected via a thioester linkage with the N-terminus (see F-CO-S-K3). The formation of the E3-K3 parallel coiled coil brings the thioester unit into close proximity with the cysteinyl residue. This arrangement templates an acyl transfer reaction, which proceeds in analogy to an NCL reaction. The end-of-helix arrangement of the functional groups results in a very high effective molarity.
Furthermore, arylmercapto-linked thioesters are known to react rapidly in NCL reactions. 71 Therefore, formation of the coiled coil triggers an almost instantaneous labelling reaction. The E3-K3 coiled coil has a stability in the nanomolar range (K d = 70 nM). As a result, the E3-K3templated labelling reaction proceeds rapidly at 100 nM concentration of reactants. At concentrations this low, nontemplated acylation reactions are negligible. The labelling reaction occurs with high target specificity, proceeds within seconds to minutes, and offers a free choice of the transferred reporter group.
We have used the coiled coil-templated acyl transfer for the labelling of cell surface proteins expressed in HEK293 cells and Chinese hamster ovary (CHO) cells. Labelling of G-protein coupled receptors (GPCRs) such as human neuropeptide Y receptors 1, 2, 4, 5, human dopamine receptor and human neuropeptide FF receptors 1 and 2 succeeded within 2 to 5 minutes by using 100nM labelling agent only ( Figure 3B). We demonstrated labelling with AF350, ATTO488, TAMRA, and biotin. The method facilitates the analysis of GPCR trafficking. In one example, we followed the intracellular transport of  internalized human neuropeptide Y2 receptors (hY 2 R). 72 Unstimulated HEK cells expressing the tagged hY 2 R were incubated with a TAMRAlabelling agent. Treatment with the neuropeptide Y2 triggered internalization. At the concentration applied internalization was not quantitative. The remaining receptors were labelled with ATTO488. After a second stimulation with neuropeptide Y2, the spatial distribution of vesicles containing TAMRA and/or ATTO488 labels was followed by confocal fluorescence microscopy. The two-color pulse-chase experiment revealed that rather than traveling separately, vesicles from separately internalized hY 2 R begin to fuse already after 10 to 12 minutes.
Fusion was complete after 30 minutes, and the receptors were pooled in Rab-4-positive vesicles for fast recycling.
In ongoing work, we explore the reversibility of coiled coil formation 73 and develop methods for tagging proteins with peptide nucleic acid (PNA) strands, which is expected to provide a universal platform for labelling and manipulation of cell surface proteins.

| TERMOLECULAR ASSEMBLIES FOR NUCLEIC ACID-PROGRAMMED PEPTIDE SYNTHESIS
It is a fascinating vision to allow chemical reactions to only occur in or on a selected subset of cells. Such a reaction could lead to the formation of cell toxic molecules that would be formed only in diseased cells, eg, cancer cells. To achieve such cell-type specificity, the chemical reaction must be conditional and require a specific trigger that occurs only in the selected cell type. Given that a cell's RNA expression profile encodes its phenotype and considering the breadth of DNA-encoded chemistries available today, nucleic acid templates seem like an appropriate trigger. Taylor was among the first to describe a reaction system that could, in principle, allow a nucleic acid triggered drug release. 74 The reaction was based on the hydrolysis of a nitrophenyl ester. Today, many more nucleic acid-templated reactions are available including, among others, nucleophilic displacements, 75,76 tetrazine-triggered reactions, 77,78 Staudinger reductions, [79][80][81] bisarsenic thioester formation, 82 olefination reactions, [83][84][85] and photoinduced oxidative and reductive dissociation reactions. [86][87][88][89] The mentioned reactions are highly chemoselective and have been demonstrated to proceed in complex environments such as cells.
Driven by the prospect of enabling a nucleic acid instructed peptide synthesis, we explored NCL reactions. In our first example of such a reaction, we equipped PNA molecules with a C-terminal glycine thioester 21 or an N-terminal cysteine residue 22 ( Figure 4A). [91][92][93] The nonionic DNA analogue PNA was chosen owing to its compatibility with peptide synthesis. Adjacent hybridization of two PNA strands with the nucleic acid template brings the reactive groups in close proximity and accelerated the NCL by 10 3 -fold in initial reaction rates. The reaction showed a remarkable target specificity: A mismatch of a single nucleotide within the target strand nearly abolished the template effect, and the NCL ceased to proceed ( Figure 4A′). 92 To demonstrate the method's exquisite chemoselectivity, we interfaced the templated ligation with the polymerase chain reaction (PCR, Figure 4B). 90 The PCR provides a formidable challenge to NCL chemistry. The thioester should remain intact despite the high temperatures and slightly basic pH applied in the PCR process. Nevertheless, the templated reaction must occur rapidly within the rather short time available when the target strand is accessible in the primer extension phase of PCR. We found that the use of β-alanine (rather than glycine) thioesters greatly decreased the vulnerability against hydrolysis without detriment to the speed of templated ligation. The "in-PCR set-up" allowed DNA template synthesis starting from attomolar concentrations of target ( Figure 4B′). We used the templated ligation to determine the number of triplet repeats in Huntington DNA, which at lengths >36 repeats cause chorea Huntington disease. 94 Middel et al used photocleavable PNA templates to direct NCL to a glutamic acid side chain. 95 Very recently, Sayers et al reported a templated ligation between PNA-linked selenoesters and selenocysteine. 96 The reaction is the fastest nucleic acid-templated reaction to date and has been used to detect miRNA within cell lysates by means of a paper strip assay.
The need for stoichiometric amounts of target strand is a key issue of templated ligation chemistries. In real-world scenarios, nucleic acid molecules occur in rather low quantities. PCR is not an option for reactions that ought to proceed in biological samples. Ideally, the templated reaction should provide for turnover in the target strand. 97,98 In this case, each target molecule would instruct the formation of many product molecules. However, nucleic acid-templated ligation reactions suffer from product inhibition. The ligation products contain more nucleobases, and therefore, ligation products typically have higher affinity for the target strand than the reactive conjugates prior to reaction. One approach to reduce product inhibition in nucleic acid-programmed ligation reaction involves the integration of units that destabilize the product-target complex. Careful optimization of the ligation site such as the replacement of a glycine-cysteine by a glycine-isocysteine junction allowed improvements of turnover numbers. 93 In another report, we described bifunctional PNA conjugates, which reacted in a two-step reaction. 99 Templated NCL was followed by a cyclization reaction. The cyclic products had lower template affinity than the ligation products prior to cyclization. Under thermocycling conditions, the two-step ligation-cyclization reaction provided two to three times more product than the "ligation only" reaction. However, thermal cycling is, again, not an option for reactions designed to occur in experiments that include cellular material such as in lysates or fixed/live cells.
To ease the problem of product inhibition, we conceived a templated NCL-like reaction that avoids ligation of the two reactive PNA strands. Instead, the reaction involved the transfer of a thioesterlinked acyl unit from a donor conjugate to an acceptor conjugate ( Figure 5). 100 The reactions still proceed in analogy to the NCL reac- In subsequent studies, we demonstrated the versatility of the acyl transfer reaction. We showed a DNA-promoted transfer of pyrene and an RNA-promoted transfer of biotin onto a His 6 -tagged acceptor. 101,102 After immobilization onto Ni-coated microtiter plates, transferred biotin was quantified by means of a horseradish peroxidase-streptavidin. The setup allowed the detection of 500 attomol RNA target. Recently, we used the reaction in the RNA target-promoted transfer of fluorophores onto semiconductor quantum dots. 103,104 We showed that the acyl transfer reaction is not restricted to PNA-based reaction systems but also proceeds with reactive DNA conjugates. 105 At the outset of this chapter, I described the idea of developing systems that read RNA and translate the recognition event into a reaction product that interferes with cellular processes. Most of the reactions pertinent to this idea belong to the category of cleavage reactions.
Typically, bioactive molecules are released from inactive, prodrug-like forms by some kind of dissociative chemistry such as hydrolysis, tetrazine-mediated cleavage, reduction-triggered fragmentation, or photoinduced cleavage. 16 The aforementioned acyl transfer offers the prospect of building up drug-like molecules by bond-forming rather than bond-cleaving reactions. The first example was described by Erben et al. 106,107 The idea was to translate nucleic acid information into the output of peptide molecules that interfere with disease-related protein-protein interactions. In this example, a DNA target strand triggered the transfer of an alanine residue from thioester-linked PNA conjugate 27 onto tripeptides or hexapeptides ( Figure 6A). Interestingly, the length of the acceptor peptide 28 had little effect on the transfer rate. More than 60% product was obtained in less than 30 minutes.
The formed peptide was designed to disrupt the interaction between    The extent of the binding enhancement provided by multivalency depends, among other factors (vide infra), on the number and orientation of ligands on the multivalent display that should match the arrangement of binding pockets offered by the receptor system. 32,116 Often, the structure of the receptor system is unknown, and as a result, it is unclear how a multivalent ligand display should be designed in order to allow for tight interactions at acceptable ligand economy.
We reasoned that DNA would be an ideal scaffold for controlling the number and orientation of ligands because (a) hybridization of complementary strands provides full control over the valency of the ligand display and (b) sequence-programmed self-assembly of duplex and higher order structures is well established allowing Ångstrom-precise positioning of ligands. In an approach, which we termed DNA-programmed spatial screening, the distance between the ligands is systematically varied ( Figure 8A). 117 Assemblies that arrange the ligands in an orientation that matches the arrangement of binding sites will provide the highest affinity for the receptor system under scrutiny. Therefore, spatial screening provides structural information about the receptor system.
Pioneering studies from Baird involved the DNA-programmed bivalent presentation of haptens to characterize bivalent binding by antibodies. 118 Matsuura et al examined the interactions between high molecular weight DNA-galactose cluster on lectin recognition. 41 Gorska et al described the DNA-programmed presentation of bivalent and monovalent dimannosides and trimannosides. 119 By examining binding to the antibody 2G12, which broadly neutralizes human immunodeficiency virus (HIV), the authors concluded that DNA-guided presentation of carbohydrates emulates the complex carbohydrate epitopes found on gp120 of HIV. We conjugated a single N-acetyllactosamine (LacNAc) unit with PNA and used hybridization with DNA to align one to four PNA strands. 117 Each nick site between the annealed PNA strands can be regarded as a hinge that enables torsions around the helical axis and facilitates bending. The torsional flexibility is important to avoid a scenario where the distance between the two ligands would be optimal for bivalent interaction with the receptor but access would be blocked owing to presentation on opposite sides of the helix. We used the sequence-programmed assembly to position the glycan residues in distances between 42 and 146 Å. The when the glycan residues were displayed by using a 104-to 127-Å long connector region ( Figure 8B, top). 117 120 According to crystal structure analysis, the RCA 120 tetramer arranges the sugar binding sites in 120 Å Euclidian distance ( Figure 8B, bottom). The DNA-programmed spatial screen revealed that the highest (250-fold) enhancement of binding affinity was observed when the two LacNAc residues were separated by 146 Å, which, again, matches the distance required to adapt to the convex surface.
The DNA-based spatial screening was used to identify selective binders of two closely related members of the Src family of tyrosine kinases. 121 The spleen tyrosine kinase (Syk) and the ζ-chain associated EG_n (n = 7, 10, 13) is characterized by RBA = 5%-10% ( Figure 10A).
NMR and UV spectroscopic analyses hint at hydrophobic interactions between the ER ligands that may penalize binding to the ER.
Furthermore, tethers such as oligoethyleneglycol can fold back and wrap around the ER ligand. 123 Such interactions should not occur with DNA-based spacers. Indeed, termolecular DNA assemblies, in which two appended raloxifene units were positioned in three or six nucleotides distance (Ral 2 -DNA_k, k = 0, 3), bind the ERα with 120% and 80% RBA, respectively. 124 Molecular modeling suggested that the six-nucleotide-spaced display can bridge the 35 Å distance between the canonical estrogen binding sites ( Figure 10B). However, the three-nucleotide spacer is too short to span this distance. Crystal structure analysis and docking studies suggested the presence of secondary binding sites 17 Å away from the canonical binding site. We speculated that the three-nucleotide-spaced arrangement picks up interactions with this hydrophobic patch. On the basis of this assumption, we tethered two raloxifene units in Ral 2 -EG_1 via a spacer too short for bridging the canonical binding sites but of sufficient length to allow engagement of one canonical and one secondary binding site.
Spectroscopic measurements suggested that the short tether does not permit homophilic raloxifene interactions. The knowledge obtained in these studies was used for the design of high affinity fluorescent ERα binders. 128 Viruses take advantage of multivalency-enhanced interactions between protein-based receptors and sugars in order to facilitate adherence to host cells. A well-studied example is the influenza A virus (IAV) that offers hundreds of hemagglutinin (HA) trimers for enhanced recognition of cells displaying multiple sialylated galactose sugar units ( Figure 11A). Driven by the medical need to prevent pandemic influenza infections, a variety of multivalent HA binders have been developed. A typical approach relies on multivalent presentation of glyco ligands from polymers, 131,132 dendrimers, 45,133 or nanoparticles. 134,135 It remains unknown which and how many of the several glyco ligands engage on binding and, therefore, these scaffolds do not provide information about the optimal spacing of HA ligands. In order to identify the criteria for enhanced binding at high ligand economy and learn about the arrangement of binding sites on the IAV particle, we used DNA-programmed bivalent screening. 129 We figured that a range of far reaching scaffolds would be required to assess the potential for enhanced interactions upon bridging of two sugar binding sites within an HA trimer and across HA trimers on the IAV surface ( Figure 11B). The study also included a comparison between distance-affinity relationships provided by two kind of scaffolds, i.e., rigid, sequence-programmable DNA-type architectures and flexible polyethylene glycol (PEG), which presented the sialyl-LacNAc ligands in 23 to 101 Å averaged distance. The combination of distance-affinity measurements by microscale thermophoresis and hemagglutination inhibition assays with a theoretical analysis by statistical mechanics models revealed that PEG-based scaffolds fail to provide affinity enhancements ( Figure 11C, red curve). PEG is too flexible to raise the effective molarity of the glyco ligands to the millimolar concentrations required to enable the interaction with the 42 Å-spaced, low affinity binding site (K mono = 3 mM). The situation is entirely different with the DNA-based scaffolds. The DNA-based spatial screening exposed a bimodal distance-affinity relationship for both soluble HA and HA on the IAV surface ( Figure 11C, blue curve). One of the binding optima, indicated by 10 3 -fold enhanced binding, was obtained when the two ligands were separated by 52 to 59 Å. This probably is the spacer length required to bend over the slightly convex protein surface. A second binding optimum was observed for complexes that presented the glyco ligands at a 26 Å distance. This pointed to a secondary binding site, which corroborated previous results from crystal structure analysis. 136 The spatial screen also revealed a preference for bivalent recognition of sugar binding sites within a HA trimer rather than binding across HA trimers on the IAV surface. This conclusion was drawn from a comparison of experimentally observed binding data and modeling data. The latter would have suggested an affinity enhancement for complexes displaying glyco ligands in >80 Å distance, which was not observed experimentally. In a separate study, we explored the reach of bivalency-enhanced binding (vide infra, Figure 12). 137 We found that the reach is critically controlled by the strength of the monovalent interaction. This suggests that the millimolar K d provided by the interaction between a single sialyl-LacNAc and a single HA binding site is not sufficient to enable bridging across two HA trimers.
Recently, we extended the DNA scaffolds to allow the oligomerization of distance-optimized binders with micromolar affinity. 130 For this purpose, we assembled DNA sequence repeat motifs by means of rolling circle amplification (RCA). In this method, a DNA polymerase extends primers according to the information on a circular DNA template. RCA of 39 or 50 nt long circular DNA provided single strands containing ≈46 or ≈15 repeats, respectively. These strands were used for the concatenation of sialyl-LacNAc conjugates ( Figure 11D). A). 138 The heterotetrameric adaptor complex 2 (AP-2) recruits clathrin to membrane regions destined for clathrin-mediated endocytosis. 139 AP-2 harbors two ear (also termed appendage) domains: the   137 We concluded that the distance between the αand β2-ear domains is too large to allow bivalency-enhanced interactions.
To test this assumption, we designed a model system, which would provide full control over both, the receptor and the ligand system. 137 For this purpose, we attached two cucurbit 7 uril CB 7 units to one DNA scaffold complex and two adamantane ligands to another DNA scaffold ( Figure 12B). Adamantane and CB 7 form host-guest complexes at nanomolar concentrations. The study included two distinct adamantane guests that differed by the CB 7 affinity. By systematically varying the distance between the adamantane and CB 7 on DNA scaffolds, we explored the distance reach of bivalencyenhanced interactions between bivalent guests and bivalent hosts ( Figure 12C). The study revealed that the affinity gain provided by bivalency is critically controlled by the strength of the monovalent interaction. The higher the stability of the host-guest (or in other words, receptor-ligand) interaction the longer the linker can be.
Importantly, the reach of the bivalency enhancement is reduced with increasing flexibility of both the receptor and the ligand scaffolds.  Another actively pursued approach pertains to the development of methods that hijack natively expressed biomolecules and repurpose their function to serve as templates that instruct the formation druglike molecules. In this approach, a drug would be formed in situ only in those cells that express the instructing templates. At the first glance, this sounds like science fiction. But an impressive report from the Winssinger group has shown that cell endogenous RNA molecules can be used to drive a fluorogenic reaction inside whole organisms. Now, the art is to conceive a templated chemistry that induces the formation of a highly potent drug-like molecule. We, and others, are working on it.
Specificity also is a key issue in peptide synthesis and live cell protein labelling. It is, for example, desirable to be able to append any kind of fluorescence label within minimal time. An ideal method would allow multiplexing in order to follow the localization of and interaction between two or more proteins in real time. Labelling would ideally extend beyond the appendage of fluorophores. Imagine the opportunities if the labelling reaction was reversible or created a handle that provides control over the localization of a protein and its interaction with other proteins or ligands. Again, templates will be of help. Short peptides and small molecule scaffolds provide unique microenvironments, which allow chemical reactions to proceed with exquisite site specificity. For example, end-of-helix arrangements at coiled coil peptides provide high effective molarities and enable labelling reactions within seconds to minutes at low concentration of labelling agent.
Currently, we are exploring orthogonal coiled coils and the use of PNA-based tags as generically addressable landing platforms.
I have sketched research problems that drive the research in my lab. In hindsight, so many things seem to follow a logical plan.
However, as much as I would like to be able to follow a strict plan, I have to admit that serendipity has its place and very often it is chemistry and biology itself which drags us towards certain research topics.
This has happened, for example, with our unexpected excursion to radical chemistry in ligation auxiliaries. Our model studies on the limits of bivalency were the result of fruitless and frustrating attempts with biological material. I am grateful to the many individuals who made this never-ending journey to specificity possible and to the award of the Max Bergmann Medal that recognizes their work.