Design and recognition of cucurbituril-secured platinum-bound oligopeptides†

Platinum terpyridyl complexes, stacked on top of one another and secured as dimers with cucurbit[8]uril (CB[8]) in aqueous medium, were functionalized quantitatively and in situ with a pair of pentapeptides Phe-(Gly)3-Cys by grafting their cysteine residues to the Pt centers. The resulting CB[8]·(Pt·peptide)2 assemblies were used to target secondary hosts CB[7] and CB[8] via their pair of phenylalanine residues, again in situ. A series of well-defined architectures, including a supramolecular “pendant necklace” with hybrid head-to-head and head-to-tail arrangements inside CB[8], were obtained during the self-sorting process after combining only 3 or 4 simple building units.


Introduction
Cucurbit [8]uril (CB [8]), a member of the cucurbituril family of macrocycles, 1-5 typically forms ternary complexes in aqueous medium with pairs of charged guests by distributing the positive charges over both portals of the macrocycle, in a head-to-tail (HT) arrangement. However, we recently showed that platinum terpyridyl (tpy) complexes bearing a CB [8]-binding unit at the tpy 4 0 -position assemble with CB [8] into a head-to-head (HH) motif. Both positive Pt centers sit on top of each other at one CB [8] portal, leaving the other void of any guest interaction. 6 Favorable dispersive interactions between the stacked tpy ligands and possible metal-metal bonding through d z 2 À d z 2 orbital overlap were proposed as driving forces for the recognition pattern. We showed later that a variety of thiolates, including cysteine and glutathione, can be graed in situ to CB [8]-secured Pt chloride dimers to form large dynamic libraries of homo-and heteroternary assemblies. 7 Graing peptides onto the CB [8]-secured Pt dimer scaffold is a step towards two distinct longer term objectives: (1) the design of new well-dened motifs for protein recognition, and (2) the rational design of synthetic oligopeptides with specic or unusual conformations. The well-known cytotoxic properties of Pt complexes [8][9][10][11][12][13] are also seen as an exploitable feature. In this proof-of-concept study, we wanted to test whether it was indeed possible to (1) form well-dened HH CB [8]-secured Pt/peptide dimers, and (2) use those assemblies to target a subsequent host, again in a well-dened manner. We show that both hypothesis are valid using pentapeptide Phe-(Gly) 3 -Cys (FGGGC) and either CB [7] or CB [8] as secondary hosts. We also show that elegant structures such as "pendant necklaces" with both HH and HT features can be obtained in situ from just a few readily available building blocks in aqueous medium. Pentapeptide FGGGC was chosen as terminal phenylalanines (Phe) form tight binary complexes with CB [7] (binding affinity up to 3.1 Â 10 7 M À1 ), 1,14,15 and HT ternary assemblies with CB [8] (binding affinity up to 1.5 Â 10 11 M À2 ; 15,16 see ESI † section for the titration of the pentapeptide with CB [8]). This feature was discovered by Urbach and coworkers, [15][16][17][18][19][20][21][22][23] and was exploited by the same group to selectively encapsulate into CB [7] the N-terminal Phe residue of proteins, including the insulin B-chain, 24 the human growth hormone (hGH), 25 ubiquitin 26 and myoglobin. 26 Brunsveld and coworkers [27][28][29][30][31][32][33] showed that protein dimerization is possible by exploiting the recognition ability of CB [8] towards pairs of Phe residues, leading to applications such as enzyme activity modulation. 29, 31 Liu also showed that the same motif can be used to engineer large protein-based nanostructures. [34][35][36][37] In our system, the C-terminal cysteine residue of FGGGC binds to the Pt centers, and the three glycines act as spacers that confer exibility to the side chain. For the sake of clarity, we will add "Pt" or "Phe" superscripts to CB[n] to indicate which portion of the Pt/peptide complex the macrocycles interact with.

Results and discussion
Pt/peptide assembly CB [8] Pt $1 2 (see Fig. 1) was readily obtained in situ by substitution of the chloride ligand from the parent Pt chloride assembly with pentapeptide FGGGC in deuterium oxide. No buffer was used as the charge of Pt/peptide complex 1 is expected to remain singly positive on a wide pH (or pD) range (approximately 2-9). Upon dimer formation at the Pt tpy site, terpyridyl hydrogens H 6 are shied downeld by 0.33 ppm and split into two doublets, as they become diastereotopic in the presence of the two chiral peptides (see Fig. 1, spectrum b). This behavior was also observed for aromatic hydrogens at positions 3-5 and 3 0 , as expected from one of our earlier studies. 7 Signals pertaining to peptide FGGGC are barely affected except for diastereotopic hydrogens H a , which split into two multiplets separated by 0.08 ppm. On the other hand, uorine atoms located at the tpy head experience an upeld shi (0.19 ppm) upon peptide binding.
Treatment of the CB[7]-bound Pt chloride assembly with the pentapeptide, however, afforded mostly assembly 1$CB [7] Phe , with CB [7] switching from the Pt tpy to the Phe station (see Fig. 1, spectrum c). In the presence of an excess amount of CB [7] (>2 equiv.), both units were encapsulated by the macrocycle (see Fig. 1, spectrum d).
We then titrated assembly CB [8] Pt $1 2 with CB [7] to test the stability of the CB[8]-secured Pt dimer in the presence of a competing target host. [5]Pseudorotaxane CB [8] Pt $(1$CB [7] Phe ) 2 was formed exclusively aer addition of 2.0 equiv. CB [7] (see Fig. 2). No association between the tpy head and CB [7] was observed, i.e. no disassembly of the CB[8]-secured dimer took place. Signicant upeld shis were observed for aromatic hydrogens H h , H i and H j (0.64-0.88 ppm) and methylene hydrogens H g (0.64 ppm), and downeld shis for hydrogens H f (0.10 ppm). Those shis conrm the inclusion of the benzyl moiety within CB [7], with the a-carbonyl hydrogen H f located near the rim of the macrocycle. As expected, no signicant shis were observed for the remote uorine nuclei on the tpy heads.
To quantify the CB [8] preference for the Pt tpy sites over the Phe residues, the various recognition events at play were analyzed by isothermal titration calorimetry (ITC) in pure water.   The affinities of CB [7] towards the Phe residue of pentapeptide FGGGC and the CB[8] Pt $1 2 complex are 1.2 (AE0.2) Â 10 7 M À1 and 6.3 (AE0.2) Â 10 6 M À1 (see Fig. 3), respectively, in excellent agreement with the binding affinities measured by Urbach for the FGG peptide (3.1 Â 10 7 M À1 in a 10 mM sodium phosphate buffer, pH 7.0). 15,20 The proximity of both peptide chains in complex CB [8] Pt $1 2 thus does not cause any cooperativity effect, i.e. graing the peptide to the CB[8]-secured Pt dimer scaffold does not signicantly impact the binding affinity of the terminal Phe unit. To determine the affinity of CB [7] towards the diuoroaryl substituent of the tpy ligand without perturbation from the Phe binding site, titrations were carried out using truncated Pt/cysteine complex 2 highlighted with a dashed blue box in Fig. 3. The affinity of this complex towards CB [7] is only 1.9 (AE0.1) Â 10 5 M À1 . The affinities of pentapeptide FGGGC towards CB [8] were 1.9 (AE0.6) Â 10 7 M À1 and 1.2 (AE0.6) Â 10 6 M À1 for the formation of the binary and HT ternary complexes, respectively, again in excellent agreement with reported binding affinities of N-terminal Phe residues in short peptides 15,18,[38][39][40] and proteins. 29,31,34 Cooperativity is quantied using eqn (1), where K 1 and K 2 are the equilibrium constants for the formation of binary and ternary complexes, and a is an interaction parameter; positive and negative cooperativities are observed when a > 1 and a < 1, respectively. 41,42 In our case, cooperativity in the CB [8] encapsulation of the pair of Phe residues is slightly negative (a ¼ 0.26 AE 0.15).
A word of caution is warranted, however. As shown by Urbach and coworkers in the case of tripeptide FGG and CB [8], 18 as well as Cistola and coworkers with small molecule/protein interactions, 43 K 1 and K 2 constants obtained by ITC can be strongly correlated, i.e. (1) reasonable ts of ITC enthalpograms can be obtained when setting K 1 as a constant while tting K 2 ; and (2) ternary binding constant b ¼ K 1 K 2 (in M À2 ) is rather insensitive to the value of K 1 . In our case, a plot of the goodnessof-t value c 2 as a function of K 1 returns a clear minimum at the K 1 constant mentioned above (see ESI section, Fig. S33a †). Furthermore, the error on parameter a is small enough to ascertain that negative cooperativity is much more likely than not.
The dimerization constant K Pt-Pt of assembly 2 corresponds to a free energy term of À5.9 (AE0.3) kcal mol À1 , in excellent agreement with the typical strength of Pt-Pt interactions. 44 We note that this dimerization in the absence of CB [8] prevents us from extracting separate binding constants K 1 and K 2 towards the macrocycle. A reliable ternary binding constant b ¼ K 1 K 2 was obtained, however (1.3 (AE0.6) Â 10 13 M À2 ). In other terms, dimer 2 2 can be considered as a standalone guest forming a 1 : 1 complex with CB [8], with a binding affinity K 0 of 7 (AE5) Â 10 8 M À1 , obtained from eqn (4).
The combined equilibrium constant for the formation of assemblies [CB [7] Fig. 4, spectrum b), showing the formation of a new assembly in a slow exchange regime. A very well-dened set of new signals is present aer addition of exactly 1.0 equiv. CB [8] relative to assembly CB [8] Pt $1 2 (spectrum c). This led us to suspect that only one of the assemblies outlined above is present in the mixture. The formation of a new, unique 19 F NMR signal at 1.0 equiv. CB [8] (À107.06 ppm) is consistent with the observations made from the 1 H NMR spectra.
A comparison of spectra a and c in Fig. 4 Fig. S31 †). This clearly indicates that the recognition event is taking place at the Phe residue. The 5.3-6.0 ppm portion of the 1 H NMR spectrum displays a feature that also allows us to discard assembly CB [8] Pt $1 2 $CB [8] Phe HH with the double HH arrangement (see Fig. 4, scenario 2): the CB [8] rims in such an assembly would all be non-equivalent, and would afford 2 pairs of doublets for the pseudo-equatorial hydrogens. This is not the case; addition of 0.50 equiv. CB [8] to assembly CB [8] Pt $1 2 shows the growth of only one doublet at 5.85 ppm (see Fig. 4, green signal in spectrum c close-up), in addition to the expected signal of the new CB [8] Fig. 4, scenarios 1 and 3), we compared the diffusion coefficients of the unknown assembly with that of complex CB [8] Pt $1 2 using diffusion-ordered spectroscopy (DOSY) experiments. 40,[45][46][47][48][49][50][51][52][53] Diffusion constants in D 2 O were 1.51 (AE0.01) Â 10 À10 m 2 s À1 and 1.58 (AE0.04) Â 10 À10 m 2 s À1 , respectively. A difference as low as 7 (AE5) Â 10 À12 m 2 s À1 indicates that the hydrodynamic radius of the unknown assembly is very similar to CB[8] Pt $1 2 (within 0.7 (AE0.4)Å). 40 Assembly CB [8] Pt $1 2 $CB [8] Phe HT , with hybrid HH and HT arrangements at the tpy ligand and Phe residues, respectively (see Fig. 4, scenario 1) thus becomes the only plausible option. It is preferred over the entropically penalizing 4 : 4 assembly [CB [8] Pt $1 2 $CB [8] Phe ] 2 . Furthermore, in two earlier studies, 47,50 we showed that the molecular weights M of metal tpy complexes that self-assemble into rigid dynamic oligomers in the presence of CB [8] are linked to their diffusion coefficients D by the empirical power law (5), with m ¼ 0.41 (AE0.03) and b ¼ 8.28 (AE0.09). A coefficient m close to 1/3 would characterize spherical particles according to the Stokes-Einstein equation, while m usually ranges from 0.3 to 0.6 in non-spherical systems like oligomers. 54 Power law (5) returns M ¼ 5200 (AE600) g mol À1 for putative assembly CB [8] Pt $1 2 $CB [8] Phe HT , in remarkable agreement with its actual molecular weight (4617 g mol À1 ). Finally, mass spectrometry analysis also conrmed the formation of this "pendant necklace" (see ESI section †).
Binding constants for the formation of the binary and ternary complexes at the Phe residue were K 1 ¼ 2.4 (AE0.3) Â 10 6 M À1 and K 2 ¼ 1.7 (AE0.4) Â 10 6 M À1 , respectively. Again, a plot of the goodness-of-t value c 2 as a function of K 1 returns a clear minimum at the K 1 constant mentioned above (see ESI section, Fig. S33b †). While attempts to rationalize the mild differences with the CB [8] encapsulation of free pentapeptide FGGGC would be putative, two trends deserve mentioning: (1) positive cooperativity is now observed (a ¼ 2.8 AE 0.7) while cooperativity is negative in the case of free peptide FGGGC (a ¼ 0.26 AE 0.15); and (2) the formation of the "pendant necklace" (i.e. the formation of the ternary complex at the Phe residues) is entropically neutral (TDS 2 ¼ À0.1 kcal mol À1 ), while ternary complex formation with the free peptide is entropically penalizing (TDS 2 ¼ À2.0 kcal mol À1 ). Both trends suggest that the entropically favorable "intra-assembly" necklace formation overcompensates the concomitant, entropically-penalizing restriction of conformational mobility.
We note here that we use the term "pendant necklace" for the lack of a better word to qualify structure CB [8] [8] Pt $1 as one unit, without taking into account that it is itself a [3]pseudorotaxane. 55,56 Ultimately, we propose the term "pendant necklace", as the CB[8]-secured Pt tpy dimer reminds us of the pendant unit of the jewelry, and the HT ternary complex between the pair of Phe residues and CB [8] Phe its clasp behind the neck. To the best of our knowledge, based in part on the recent review by Wang, Kermagoret and Bardelang on oligomeric CB [8] complexes, 57 the formation of this hybrid HH/HT necklace is unprecedented.
As obtaining crystals of the pendant necklace for X-ray diffraction analysis was unsuccessful, we explored its structure using the most up-to-date combination of molecular dynamics, semi-empirical methods and density functional theory (DFT) being currently developed by Grimme and coworkers. Conformational screening was rst carried out using an approximate geometrical analog of the pendant necklace bearing only one CB [8] unit (see Fig. 5a and S34 †) with Grimme's Conformer-Rotamer Ensemble Sampling Tool (CREST) 58,59 and the built-in generic GFN-Force-Field (GFN-FF), 60 in conjunction with the GBSA solvation model. 61,62 The CB [8]-secured Pt dimer surrogate was designed to mimic the distance between the sulfur atoms and the direction of its substituent in the pendant necklace, while increasing computing efficacy. The 37 most stable candidates (out of approximately 12 000; a 25 kcal mol À1 cut-off was applied) were isolated, and the surrogate fragment was replaced by the CB[8]secured Pt dimer motif. The structures were then reoptimized with the semiempirical tight-binding method GFN2-xTB 63,64 in conjunction with the GBSA solvation model. 61,62 The four most stable structures (based on a 10 kcal mol À1 cut-off) were nally reoptimized by DFT using a functional well suited for supramolecular systems (B97-3c; 65 basis sets are def2-mTZVP) 66,67 in conjunction with the COSMO solvation model. 68,69 The relative stabilities (DG) of the four assemblies were calculated using eqn (6), where DE B97-3c is the electronic contribution at 0 K calculated by DFT in the gas phase, DG T,xTB is the vibrational contribution at 25 C obtained at the GFN2-xTB level, and DG solv,xTB is the solvation energy calculated with the GBSA model at the GFN2-xTB level.
The most stable conformation of the pendant necklace is presented in Fig. 5. Its compact arrangement suggests that through-space interactions between hydrogens of the peptide and of CB [8] might be observable. A 1 H-1 H NOESY experiment indeed revealed cross-peaks between equatorial hydrogens of CB [8] Phe and tpy hydrogen H 6 , as well as peptidic hydrogens H a and H b (see Fig. 5). Favorable interactions between the peptide carboxylate units and the outer wall of CB [8] Phe might also further compact the assembly. 50

Conclusions
We showed that we could functionalize CB [8]-secured Pt dimers in situ and quantitatively with a pair of cysteine-containing peptides, and use the Pt/peptide/CB [8] assembly to target secondary hosts CB [7] and CB [8] site-selectively. The most elegant outcomes of the study are (1) the formation of pendant necklace CB [8] Pt $1 2 $CB [8] Phe HT with a new hybrid HH and HT arrangement at the tpy head and Phe residues, respectively, and a thorough quantication of all recognition forces at play, (2) the successful in silico screening and isolation of a plausible geometry for this pendant necklace, and (3) the quantitative switching of CB [7] and CB [8] from the tpy head and Phe residues, respectively, when attempting to attach the Cys units of ternary complex 1 2 $CB [8] to the Pt centers of binary assembly CB [7] Pt $(Pt$Cl). This study now paves the way for the recognition of proteins by these CB [8]-secured Pt dimers, and for the design of rationally constrained oligopeptides.

Data availability
All analytical data is provided in the ESI. All details about computational methods and associated coordinates are available in the ESI.

Author contributions
EM and HB conceived the project. HB conducted all experiments. EM and HB wrote the manuscript.

Conflicts of interest
There are no conicts to declare.