Protein Recognition and Assembly by a Phosphocavitand

Abstract

Controlled protein assembly is an enabling technology, in particular, for biomaterials fabrication. Here, we report protein recognition and assembly by a phosphate-containing macrocycle (pctx). We show that the C ₃-symmetric phosphocavitand is a versatile receptor for N-terminal residues or arginine but not lysine. Using atomic resolution X-ray diffraction data, we reveal the precise details of N-terminal complexation in the β-propeller protein lectin (RSL). In some cocrystal structures, a tetrahedral cluster of the phosphocavitand occupies one end of the β-propeller fold, providing a node for protein assembly. The macrocycle cluster is compatible with different types of precipitants, a broad pH range, and zinc complexation. We demonstrate system control with an arginine-enriched RSL that alters the overall assembly due to selective arginine complexation by pctx. A lysozyme–pctx cocrystal structure also demonstrates arginine complexation by the macrocycle. An alternative macrocycle cluster occurs with an engineered RSL bearing an extended N-terminus. In this structure, involving zinc ligation at the N-terminus, the macrocycle forms trimeric clusters and four such clusters form cage-like substructures within the tetrahedral protein framework. Thus, N-terminal complexation in combination with phosphocavitand self-assembly provides new routes to protein crystal engineering.

Introduction

The selective noncovalent modification of proteins by synthetic receptors has diverse applications from sensing to assembly and biomaterials fabrication. − Receptors that selectively bind small sites on protein surfaces, including unique features such as the N-terminus, are desirable. ,, Here, we describe the protein recognition and assembly properties of a phosphate-containing macrocycle. We show that, in addition to complexing the N-terminus, the phosphocavitand forms clusters that act as nodes for protein assembly. Multiple crystal structures reveal how the macrocycle clusters can engage different protein features, as well as complexing zinc ions, resulting in distinct assemblies. This approach complements the major developments in protein assembly and crystal engineering , utilizing computationally designed, , metal-mediated, , or ligand-mediated interfaces. ,−

The macrocycle of interest is the recently reported phosphate derivative of C ₃-symmetric cyclotrixylohydroquinoylene, dubbed pctx and produced in 3 reaction steps. , Compared with the bowl-shaped, C ₄-symmetric sulfonato-calix­[4]­arene (sclx ₄ , a known protein receptor), the saucer-shaped pctx is smaller, shallower, less polar and less anionic (Figure , Table ). While both rims of sclx ₄ are polar, the lower rim of pctx is manifestly hydrophobic. Another distinction between these two hosts is their differing flexibility. The singly bridged sclx ₄ bowl can breathe and the sulfonate substituents rotate freely. In contrast, the doubly bridged pctx saucer is rigidly constrained and each phosphate points one oxygen atom toward the macrocycle center. The resulting cavity has a portal dimension of P–O···O–P ∼ 6 Å. These anionic hydrogen bond acceptors, oriented toward the macrocycle center and delimiting the cavity, appear to be pivotal in guest binding by the shallow pctx. Considering the phosphate charge distribution, the −P–O groups are equivalent. For example, the related molecule diphenyl phosphate forms hydrogen bonded ionic chains with imidazole (CCDC 1226648), in which each −P–O group contributes equally. We note also the intriguing structural similarity between pctx and compound 1 in ref . This bisaryl phosphate has strong hydrogen bonding capability, again involving equivalent −P–O groups.

1.

Molecular structures of the pctx and sclx ₄ sodium salts and 3D representations of the corresponding saucer and bowl. Refer to main text for consideration of charge distribution in pctx. Color code: carbon in lilac, oxygen in red, phosphorus in orange, sulfur in yellow.

1. Physico-Chemical Properties of pctx and sclx ₄ .

receptor	MW t (Da)	surface area (Å2)	shape	cavity Ø (Å)	cavity depth (Å)	charge
pctx	636	∼670	saucer	∼6	∼3.5	–3
sclx 4	745	∼800	bowl	∼6	∼6	–5

^aAssuming complete deprotonation.

Considering the utility of sclx ₄ for N-terminal, lysine or arginine complexation and protein assembly, ,− we envisaged that pctx would be a useful receptor, especially suited to arginine complexation.

The test protein we studied is lectin (RSL, ∼30 kDa), a 6-bladed β-propeller with C ₃ and pseudo C ₆ symmetry (Figure ). The cocrystallization of pctx with RSL and two variants of RSL resulted in five cocrystal structures. Remarkably, in each structure pctx binds the accessible N-terminus and forms clusters. The macrocycle clusters are compatible with zinc, an important bridging ion for directing protein assembly. ,,, One of the protein–pctx–zinc cocrystals incorporates two types of cages, including a substructure similar to metal organic cages. , None of the structures involve lysine encapsulation, and simple modeling reveals why this residue is not a suitable guest for pctx. An arginine-enriched variant of RSL (RSL-R₆) yields a striking modification of the protein assembly due to pctx–Arg complexation. A cocrystal structure with chicken egg white lysozyme provides further evidence of pctx–Arg interactions. This work expands our knowledge of biomolecular recognition by phospho-containing synthetic receptors, ,− and provides new components for the protein assembly toolkit. ,

2.

Surface representations of C ₃-symmetric RSL showing the toroidal structure and the funnel-shaped water channel, viewed from (a) below, (b) above and (c) the side. A side view cross-section emphasizes the channel. The N-termini (S1) are indicated in panels b and c.

Results

Protein–pctx Cocrystallization and Space Groups

A sparse matrix cocrystallization screen of RSL and up to 50 mM pctx yielded crystals in six out of 96 conditions (Table , Figure a–c, and SI Methods). Using these crystals and diffraction data collected at the PROXIMA-2A beamline of SOLEIL synchrotron, we solved three RSL–pctx cocrystal structures in space groups H3, P6₃, and H32 (Tables , , and S1). All three structures involve the same cluster arrangement of pctx (vide infra). The P6₃ and H32 structures contain similar RSL–pctx assemblies but the latter structure also includes zinc ions. Apparently, the macrocycle cluster is compatible with different types of precipitant (high salt or PEG), a relatively broad pH range, and high concentrations of metal ions such as Li⁺ or Zn²⁺ (Table ). We obtained two additional crystal forms with variants of RSL. The arginine-enriched RSL-R₆, with all lysine residues replaced by arginine (Lys25Arg, Lys34Arg, and Lys83Arg), cocrystallized with pctx in space group H3 (Figure d). The variant MK-RSL, with the macrocycle-binding N-terminal Met-Lys tag, , cocrystallized with zinc and a different type of pctx cluster in space group P4₁2₁2 (Figure e).

2. RSL–pctx Cocrytals Obtained in a Commercial Screen and Characterized by X-ray Diffraction .

#	precipitant	buffer	additive	space group
C11	2 M ammonium sulfate	0.1 M Na acetate pH 4.6		H3
F2	3.15 M ammonium sulfate	0.1 M Na citrate pH 5.0		H3
F10	1.1 M sodium malonate	0.1 M HEPES pH 7.0	0.5% Jeffamine ED-2001	P63
C2	20% PEG 6000	0.1 M Na citrate pH 4.0	1 M lithium chloride	P63
D1	24% PEG 1500		20% glycerol	P63
E6	20% PEG 3000	0.1 M imidazole pH 7.8	0.2 M zinc acetate	H32

^aAlphanumeric code indicates the condition from JCSG HTS++ (Jena Biosciences).

^bRSL-R₆–pctx cocrystals (space group H3) grew in similar conditions.

^cMK-RSL–pctx cocrystals (space group P4₁2₁2) grew in this condition only.

3.

pctx cocrystals with (a–c) RSL, in space groups H3, P6₃, and H32, respectively, (d) RSL-R₆ in space group H3, (e) MK-RSL in P4₁2₁2 and (f) lysozyme in P3₁21. Two cocrystals (c, e) also contain zinc. Refer to Table for crystallization conditions. Scale bar = 200 μm.

3. Unit Cell Parameters of Protein–pctx Cocrystal Structures.

protein	space group	res. (Å)	a (Å)	b (Å)	c (Å)	%SC	PDB
RSL	H3	1.0	46.709	46.709	102.923	40	9HRV
RSL-R6	H3	1.5	43.569	43.569	122.997	42	9HRZ
RSL	P63	1.1	46.078	46.078	139.565	39	9HRW
RSL	H32	1.5	45.893	45.893	421.816	39	9HRX
MK-RSL	P41212	1.5	112.158	112.158	99.312	50	9HRY
lysozyme	P3121	1.9	86.307	86.307	72.738	53	9HRU

^a% solvent content from Matthew’s calculation, accounting for protein and macrocycle masses.

While neutral RSL (isoelectric point pI ∼ 6.5) remains soluble in the presence of up to 50 mM pctx, the cationic, arginine-rich lysozyme (pI ∼ 10) is soluble only up to 5 mM of the anionic macrocycle. Heavy precipitates form at pctx concentrations above 5 mM. Sparse matrix crystal screening of pctx and lysozyme gave one hit. These unusual ellipsoid crystals, obtained in 1.6 M MgSO₄, 0.1 M HEPES pH 6.5 (Figure f), proved to be a lysozyme–pctx complex in space group P3₁21 (Table S2).

pctx Encapsulates the N-Terminus

RSL–pctx cocrystals in space group H3 diffracted to atomic resolution and the electron density maps clearly reveal encapsulation of the N-terminus (Figures a and S2). ∼115 Ų of Ser1, or ∼65% of the residue, is buried by the macrocycle. Table lists the four noncovalent bonds crystallographically discernible between Ser1 and pctx. Ser1-N^α forms a salt bridge to one of the phosphates, while Ser1-C^α and Ser1-C^β each make CH-π bonds with phenyl rings of the macrocycle. The fourth interaction is a hydrogen bond between the Ser1 hydroxyl (O^γ) and a phosphate of pctx. Another phosphate of pctx forms hydrogen bonds with the Ser2 amide backbone and the Gln4 side chain. Considering solvation, eight water molecules form hydrogen bonds to the three phosphates (H₂O···O–P-pctx = 2.6–3.1 Å). A ninth water is hydrogen bonded to the N-terminus (Ser1-N^α···H₂O = 2.8 Å). Thus, multiple noncovalent bonds mediate the RSL–pctx complex. Similar features of N-terminal complexation occur in the P6₃ cocrystal structure, which also includes the N-terminal ammonium making a cation−π bond with pctx (Ser1-N^α···centroid = 3.2 Å). A different type of N-terminal encapsulation occurs with MK-RSL (vide infra).

4.

(a) An X-ray crystal structure at 1 Å resolution shows pctx encapsulating Ser1 of RSL. The refined 2Fo–Fc electron density map (green mesh) is contoured at 1σ. (b) pctx forms a tetrahedral cluster located at the narrow end of the RSL β-propeller trimer. Shown are two views of the assembly, related by 90° rotation, with pctx as spheres and RSL as C^α trace. Ser1 highlighted in yellow. Note the 3-fold phosphate junctions in the pctx cluster.

4. RSL–pctx Noncovalent Bond Lengths Determined in a 1 Å Resolution Crystal Structure.

residue-atom	pctx	distance (Å)	interaction
Ser1-Nα	O–P1	2.7	salt bridge
Ser1-Cα	centroid	3.8	CH−π (cation−π)
Ser1-Cβ	centroid	3.4	CH−π
Ser1-Oγ	O–P2	2.5	hydrogen bond
Ser2-Nα	O–P3	2.8	hydrogen bond
Gln4-Nε	O–P3	2.9	hydrogen bond
Lys34-Nζ	O–P2	2.8	salt bridge
Tyr37-centr	Me	3.7	CH−π

^aO–P^X indicates arbitrary numbering of the three phosphates.

^bResidue from another RSL (symmetry mate) in the crystal packing.

Plugging the Toroid with a Macrocycle Cluster and Assembly Node

β-propeller proteins are toroidal like a Bundt cake. , In trimeric RSL, the central water-filled channel is funnel-shaped, with a wide end and a narrow end (Figure ). The three solvent exposed N-termini are collocated at the narrow end. Each N-terminus of RSL binds a macrocycle, the three of which form the base of a tetrahedron (Figure b). A fourth capping macrocycle completes the pctx tetrahedral cluster and is devoid of a guest. The tetrameric pctx cluster plugs the narrow end of the RSL β-propeller toroid (Figure b). In addition to the RSL binding at the base of the pctx tetrahedron, three other RSL trimers assemble on the periphery of the pctx cluster.

The 3-fold rotational axes of the tetrahedral cluster comprise three phosphates, one from each pctx, flanked by pairs of methyl substituents (Figure b). Cationic features on the protein complement these anionic patches of the cluster, contributing to charge neutralization in the crystalline assembly. Depending on the crystal form, different cationic features of RSL engage pctx in exo interactions (Figure ). With a 6-bladed β-propeller topology and pseudo C ₆ symmetry (Figure ), RSL can pack in layers with a honeycomb-like arrangement. The H3 cocrystal comprises protein layers interspersed by pctx clusters nestling in the approximately tetrahedral interstitial sites (Figure a). The Lys34 side chains from three RSL molecules each form a salt bridge with a phosphate of the pctx cluster (Table aand Figure d). In the P6₃ (Figure c) and H32 cocrystal structures, the packing incorporates “bilayers” of RSL interspersed with pctx clusters. Here, Lys25 forms the salt bridges to pctx, with the ammonium group approximately equidistant from the three phosphates (Figure f). These exo interactions between the pctx cluster and different lysines of RSL are part and parcel of the distinct crystal packing arrangements obtained. Apparently, the pctx cluster acts like a node around which proteins can assemble in different ways depending on which cationic feature (e.g., Lys25 or Lys34) is engaged. Solvent-bound counterions likely contribute further to charge neutralization of pctx, but none was evident in the X-ray data.

5.

Macrocycle cluster mediated assembly with crystal packing in (a) RSL (space group H3), (b) RSL-R₆ (H3) and (c) RSL (P6₃). Shown are pctx clusters as spheres, protein as C^α traces and unit cell axes in blue. The double-headed arrow indicates a c axis of ∼10 nm. (d–f) Protein–cluster interactions corresponding to panels a–c. The detail shows the pctx tetrameric cluster (sticks) bound at the N-termini of one RSL trimer (c.f. Figure ) and the contributions of Lys34, Arg34, or Lys25 from another RSL in the crystal packing. For clarity, only one additional RSL monomer is shown. Green dashed lines indicate salt bridges.

We were intrigued by the distinct binding modes between the phosphate patches of the pctx cluster and the lysine side chains in the different cocrystals. With Lys25, there is a neat match between the tetrahedral ammonium group and all three phosphates on a C ₃ symmetry axis, while for Lys34 only one phosphate makes a direct salt bridge (Figure ). These observations prompted us to question how the planar guanidinium , of arginine might interact with the cluster. To answer the question we tested the RSL-R₆ variant, in which all lysines are replaced by arginines. Remarkably, RSL-R₆ cocrystallized with pctx in the H3 space group, similar to RSL. However, instead of the peripheral interaction between Lys34 and the pctx cluster (Figure d), the guanidinium of Arg34 binds the capping macrocycle (Figure e). Apparently, the planar guanidinium has higher affinity for the pctx portal than for the exo phosphate patch. The consequent rearrangement of the proteins relative to the pctx cluster results in a ∼20% elongated c axis, and minor contractions of the a and b axes. Thus, a single residue change, Lys34Arg, leads to expansion of the unit cell in one dimension.

pctx Is an Arginine Chelator

Besides demonstrating the possibility of crystal engineering, the RSL-R₆–pctx cocrystal structure provides high-resolution detail of arginine complexation by pctx (Figures and S3). The guanidinium group is perched 0.8 Å above the plane of the three −P–O^– acceptors, with each N atom positioned 2.6 Å from the nearest phosphate oxygen, forming a chelate-type complex (Table S2). The high charge density of the pctx portal likely enables an ion pair contribution to the host–guest complexation of arginine. Furthermore, each guanidinium N atom is ≤3.4 Å from the nearest phenyl centroid indicative of cation−π bonding. Overall, the binding mode harks back to the early arginine receptors based on aromatic bis- and trisphosphonates. In addition to clamping the guanidinium group, the pctx receptor engages the alkyl portion of the side chain between two upper rim methyls.

We used this X-ray crystal structure to build a model of pctx complexing Lys, with the ammonium positioned equidistant from the phosphate acceptors and from the aromatics (Figure S3). The putative bond lengths are 25–30% longer in the model than in the X-ray structure with arginine (Table S3). This observation suggests that pctx is unsuited to binding lysine, consistent with the millimolar host–guest complexation of butylamine, a lysine side chain analogue.

A lysozyme–pctx cocrystal structure provides further evidence for arginine selectivity (Figure ). Lysozyme contains 11 arginines and 6 lysines including the N-terminus Lys1. This protein is readily crystallizable and of the >1000 PDB entries, 88% are in space group P4₃2₁2. Interestingly, this crystal form can accommodate lysozyme–ligand complexes with relatively large components such as crystallophore (MW _t 560 Da) or adenosine triphosphate (MW _t 507 Da). , Therefore, we were surprised to find the lysozyme–pctx cocrystal structure is trigonal in space group P3₁21. This rare space group for lysozyme was attributed previously to the precipitant action of magnesium sulfate. In the structure, pctx complexes Arg73 (Figure b) with the same binding mode as observed in the cocrystal structure with RSL-R₆ (Figure S3). The binding site also includes the guanidinium of Arg61 forming a salt bridge with one of the phosphates. A second pctx demonstrates an alternative binding mode. Here, the side chain of Arg112 is oriented perpendicular to the plane of the phosphate acceptors (Figure c). While the macrocycle does not form clusters in this structure, the Arg73 binding site is at a C ₂ crystallographic axis such that two molecules of pctx make van der Waals contact.

6.

Lysozyme–pctx cocrystal structure. (a) Crystal packing in space group P3₁21. (b) pctx binding at Arg73, on a C ₂ crystallographic axis. (c) Peripheral pctx–Arg112 complexation. Other interacting side chains shown as sticks. The approximate locations of Lys1 (panel b) and Arg125 (panel c) are indicated.

Lys1 is near the pctx binding site (Figure b). However, both of the Lys1 ammonium groups are tucked away forming intramolecular noncovalent bonds within lysozyme. Apparently, the stability of this tertiary structure feature outweighs any free energy gain possible in Lys1–pctx complexation. The ∼90 Ų accessible surface area of Lys1 is significantly lower than the buried surface area of Ser1 (115 Ų) in the RSL–pctx structure. The inaccessibility of Lysozyme Lys1 is also suggested by the lack of reaction with dansyl chloride. Solution state NMR spectroscopy provides further evidence of pctx binding to arginine rather than the N-terminus of Lysozyme. Using ¹H–¹⁵N HSQC measurements (Figure S4) and the known resonance assignments, we identified Arg125 as the principal binding site for pctx. Chemical shift perturbations for the resonances of Val120, Ala122, Trp123 (indole), Leu124, Arg125 and Gly126 suggest pctx complexation of Arg125 (Δδ ¹H ∼ 0.08 ppm, Figure S4). Interestingly some of these residues occur in the crystallographically defined binding site at Arg5 (Figure c), which also has a small chemical shift perturbation. Possibly, pctx preferentially binds Arg125 in solution but has higher affinity for the Arg5/Trp123 site in the solid state due to the additional interactions arising in the crystal packing, such as the Arg112 contribution. The resonances of Val2 and Phe3 were unperturbed suggesting that the N-terminus is not a binding site (Figure S4).

A pctx–Tris cocrystal structure lends further insights into host–guest chelation by the phosphocavitand. Drops containing 1 mM protein and 20–50 mM pctx yielded small molecule crystals in several conditions. Crystals from a PEG/ammonium sulfate mixture were solved in the triclinic system P1̅. In this structure (CCDC 2442403, Figure S5), despite the presence of >0.3 M ammonium counterions, the C ₃-symmetric pctx accommodates a Tris guest originating from the protein sample. Some features of the Tris binding are similar to Ser1 encapsulation (Figure a). Equivalent hydrogen bonds occur between each of the phosphates and the three hydroxyls of Tris (pctx–P–O···O–Tris <2.7 Å). There are also three equivalent CH−π bonds between the Tris methylenes and the aromatic rings (Tris–C···centroid = 3.5 Å). There is a cation-π character to the latter interaction. Unlike the Ser1-N^α, the Tris ammonium group points away from the pctx cavity and engages in hydrogen bonding with the hydroxyl group of another Tris and two water molecules. The Tris ammoniums and water molecules form a hydrogen bond network to the phosphates of pctx. Seemingly, the symmetry matched chelate-type complex takes precedence over a possible salt bridge between the ammonium and a pctx-phosphate.

Protein–pctx–Zn²⁺ Assemblies

The RSL–pctx cocrystals obtained in the presence of 0.2 M zinc acetate (Figure c) were solved in space group H32 with a similar structure to the P63 crystals (Figure ) but including 8 zinc ions in the asymmetric unit. Zinc single-wavelength anomalous dispersion (SAD) experiments confirmed the positions of these ions (Figure S6). Remarkably, the pctx cluster bearing 12 phosphates is compatible with zinc complexation (Figure a). Ser1, encapsulated by pctx, coordinates Zn²⁺ via the N-terminal amino and the backbone carbonyl (Ser1-N^α···Zn = 2.2 Å and Ser1-CO···Zn = 2.7 Å). The amine of Ser1 is also bonded to the macrocycle (Ser1-N^α···O–P-pctx = 2.6 Å) similar to that described in Figure and Table . Two water molecules complete the Zn²⁺ coordination sphere. While this zinc does not bind the macrocycle directly, there is an attractive charge–charge interaction with the pctx phosphate via Ser1. In addition to the zinc ions binding to each Ser1, a fourth Zn²⁺ coordinates three phosphates of the macrocycle cluster at a 3-fold symmetry axis (Figure a). These data suggest the possibility of complementing protein–pctx clusters with metal ions for aiding structure determination or catalysis. ,

7.

Protein–pctx–Zn²⁺ cocrystals. (a) The pctx cluster bound to RSL and zinc (space group H32). Note the fourth zinc ion binding to three phosphates at the C ₃ symmetry axis. Zinc ions and Ser1 shown as spheres and sticks, respectively. (b) In MK-RSL, each N-terminus binds one zinc and one pctx. Met0 and Lys1 shown as sticks. (c) Macrocycle clustering arises as four MK-RSL–pctx–Zn complexes (shown in b) assemble with tetrahedral geometry, forming a cage of 12 macrocycles. One of the four trimeric pctx clusters is in full-color. (d) The overall MK-RSL–pctx–Zn crystal packing (space group P4₁2₁2) viewed along the c axis, illustrating the macrocycle and protein cages. pctx and zinc represented as spheres, proteins as C^α traces and unit cell axes in blue.

Zinc complexation takes on another aspect in the assembly with MK-RSL (Figure ). This variant of RSL has a cationic, extended and disordered N-terminus containing Met0 and Lys1 that functions as a macrocycle binding site. ,, MK-RSL–pctx–Zn cocrystals grew in the same conditions as the RSL–pctx–Zn cocrystals (Table ), but in space group P4₁2₁2 and involving a different type of pctx cluster. Each of the three N-terminal methionine residues coordinates a Zn²⁺ ion, and each Met–Zn complex is encapsulated by a pctx. The penta-coordinate Zn²⁺, confirmed by Zn SAD experiments (Figure S6), binds the N-terminal amino and carbonyl groups (Met0-N^α···Zn = Met0-CO···Zn ≤ 2.1 Å), a phosphate (pctx–P–O···Zn = 2.0 Å), a carboxylate (Asp46-O^δ···Zn = 2.0 Å) and a water molecule. The metal–amino complex binds pctx via salt bridges and cation−π bonds. In addition to the N^α and CO, the C^α and C^β of Met0 are encapsulated and form CH−π bonds with the pctx cavity. The thioether portion of Met0 remains outside the macrocycle. Lys1 makes numerous exo interactions with pctx (Figure d), including a van der Waals contact (Lys1-C^β···C-pctx = 3.6 Å), a hydrogen bond (Lys-N^α···O–P-pctx = 2.7 Å) and salt bridges. The side chain ammonium extends away to salt bridge with two phosphates in an adjacent pctx cluster (Lys-N^ζ···O–P-pctx = 3.1–3.3 Å). This interaction is similar to the roles played by Lys25/Lys34 in the RSL–pctx structures (Figure ).

In contrast to the tetrameric clusters in cocrystals with RSL, the cocrystal with MK-RSL incorporates trimeric clusters of pctx. The trimer is an approximately trigonal planar arrangement of pctx held together by coplanar π–π stacking (centroid···centroid = 3.6 Å) and CH−π bonds with the methylene bridges. Each component of the cluster binds a different MK-RSL trimer and four such trimeric clusters coalesce forming a supramolecular cage of macrocycles (Figure c,d). The resulting tetrahedral assembly of MK-RSL is a porous structure with 50% solvent content, comprising protein cages and protein–Zn–macrocycle cages. The latter can be described as a Zn₁₂ pctx ₁₂(protein)₄ assembly, with similarities to metal organic cages. ,

Discussion

Similar to sclx ₄ and prototypical arginine receptors, , the physicochemical properties of pctx (Figure and Table ) inspired us to investigate protein recognition and assembly by this novel phosphocavitand. A cocrystal structure and a solution NMR study confirmed that pctx binds arginine in lysozyme (Figures and S4). The Arg side chain complements the pctx receptor with size and charge matching between the guanidinium and the three portal phosphates (Figures e, , and S3). In contrast, there was no evidence for lysine encapsulation suggesting that, unlike sclx ₄ and the molecular tweezers, ,, pctx is selective for arginine. Apparently, this selectivity arises from the complementarity of the guanidinium (but not the ammonium) with the pctx portal phosphates. A cocrystal structure with MK-RSL, containing the highly accessible Lys1, emphasizes the lack of lysine encapsulation by pctx. Selective Arg binding is desirable considering the importance of arginine and the cation−π bond in protein–protein interfaces. , pctx could be used as a protein modulator, for example, with peptide appendages that enhance the interaction capability.

The experiments with lysozyme hint at significantly lower binding affinity of pctx compared to sclx ₄ . Mixtures of lysozyme and up to 5 mM pctx are soluble, while even micromolar concentrations of sclx ₄ precipitate lysozyme. Another interesting contrast is that lysozyme–sclx ₄ complexation involves the two most solvent exposed arginines, Arg128 and Arg14. The pctx binding sites include Arg5, Arg61, Arg73, Arg112 and Arg125, all of which are less accessible (Figure S7) than the residues bound by sclx ₄ . Geometric constraints may be responsible for these differences in arginine binding. The shallow pctx may select “buried” arginine side chains, as exo interactions are possible with the surrounding residues (Figure ). Highly accessible arginine side chains that protrude into the solvent lack additional contributions to the binding enthalpy as there are no neighboring groups for exo complexation.

RSL–pctx cocrystal structures revealed the surprising result of N-terminal recognition. Such binding is not possible with an inaccessible N-terminus, as in lysozyme. The small, polar Ser1 of RSL proves to be a suitable guest as pctx both encapsulates the side chain and forms a salt bridge with the ammonium group. While Ser1 dominates the complex, Ser2 and Gln4 each contribute a hydrogen bond. Aspects of the Ser1–pctx complex are similar to (1) serine encapsulation by a tetra-phosphonate cavitand (CCDC 1415492) and by sclx ₄ (CCDC 202923), and (2) cucurbituril interactions with N-terminal aliphatic or aromatic residues (e.g., CCDC 628234). Receptors that specifically bind the N-terminus are advantageous as binding is conferred by a few residues rather than a protein surface patch. ,, Such site selectivity means that proteins with a 1–3 residue “tag” are amenable to purification, sensing, assembly etc by the receptor. It appears that compared with the larger and deeper sclx ₄ , the dimensions of pctx are favorable for Ser1 encapsulation (including enthalpically favorable interactions with surrounding residues). The structure with MK-RSL further suggests a preference for small residues as the thioether portion of the methionine side chain is outside the pctx cavity. Future studies will determine what other N-terminal residue types pctx can complex/encapsulate. Interestingly, protein–pctx complexation is compatible with Zn²⁺ chelation at the N-terminus, enabling dual metal- and macrocycle-mediated assembly (Figure ). Zinc is highly prevalent in proteins and is the metal of choice for directing protein assembly, including protein-based MOFs. ,,,

The salient result with the RSL–pctx cocrystals is the occurrence of macrocycle clusters. A ∼2.5 kDa tetrahedral cluster of the phosphocavitand, approximating a dimpled sphere with shallow cavities and a formal net charge of minus 12, sits in the channel at the narrow end of RSL (Figure ). The three macrocycles at the base of the cluster bind to the three cationic N-termini, and the fourth capping macrocycle is empty. pctx recognition of the N-terminus, and concomitant clustering effectively plugs the RSL toroid. Previously, a designed 6-bladed β-propeller (Pizza, PDB 5CHB) was plugged with a cadmium chloride nanocrystal. And we have shown that sulfonato-calix[8]­arene can block the wide end of the designed β-propeller Pent (PDB 8R3B). Together these structures suggest strategies for modulating the β-propeller fold which is widespread in different enzyme classes.

In cocrystals with RSL, three additional proteins interact exo to the pctx cluster. Different assemblies form depending on which lysine residues engage the cluster (Figure ). Replacing all three lysines of RSL with arginine, results in selective complexation of Arg34 by the capping macrocycle and consequent reorganization of the crystal packing. Thus, crystal engineering was possible by a single residue modification that altered protein binding to the macrocycle cluster. As polymorph selection occurs at the earliest stages of protein crystallogenesis, it is likely that protein–pctx clusters (Figure b) act as precursors for crystal nucleation. Furthermore, it appears that the multivalency of RSL and collocation of the three N-termini are propitious to cluster formation. In MK-RSL, the longer N-termini result in a different type of clustering that depends to a lesser extent on multivalency (compare Figures b and b). The cocrystals of MK-RSL and pctx also include zinc, which mediates macrocycle binding (Met–Zn²⁺–pctx complex). In the crystal, four complexes of MK-RSL laden with zinc and pctx coalesce in a tetrahedral geometry with trimeric pctx clusters at the vertices (Figure c). The cage-like substructure of macrocycles, similar to metal organic cages, has an internal diameter of ∼2 nm. The voids in the protein assembly also have an internal diameter of ∼2 nm (Figure d). This porous material relies on the N-terminal Met-Lys motif combined with zinc chelation, pctx encapsulation and macrocycle clustering. Including the Met-Lys motif in other proteins may provide a straightforward route to new types of materials directed by Zn²⁺–pctx complexation.

While pctx–lysozyme complexation and arginine binding are evident by NMR spectroscopy (Figure S4), similar experiments with RSL or MK-RSL did not yield binding in solution (data not shown). Under the conditions tested, pctx interaction with the neutral RSL may be too weak for detection, while pctx binding to the cationic and arginine-rich lysozyme is sufficient. These contrasting data suggest that in solution pctx–arginine complexation (chelation) is more favorable than pctx encapsulation of the N-terminal serine (one salt bridge). There is also a disparity in protein binding between sclx ₄ and pctx, which may be attributed to differences in the macrocycle cavity volume. For example, with MK-RSL, sclx ₄ encapsulation of the N-terminal Met is evidenced in both a crystal structure (PDB 9GR3) and an NMR study. In this case sclx ₄ wholly encapsulates the thioether side chain. As the expulsion of water from the cavity (hydrophobic effect) is the driving force for host–guest encapsulation, , weaker binding is expected with pctx, which has a shallow cavity (depth ∼ 3 Å) compared to sclx ₄ (∼6 Å, Table ). This difference may partly explain why micromolar sclx ₄ , in contrast to millimolar concentrations of pctx, induce lysozyme precipitation.

The hydrophobic base of pctx (Figure ) is conducive to cluster formations that are not possible with sclx ₄ , with major implications for protein assembly. Evidence is accumulating in favor of protein crystal engineering by discrete macrocycle oligomers or supramolecular synthons. , The tetrameric pctx cluster appears to be a pivotal example, acting as an assembly node with four shallow cavities for single residue encapsulation and four exo anionic sites for complexing additional cationic residues or metal ions. While discrete capsules or cages of calixarenes rely on intermolecular hydrogen bonding or metal chelation, , aromatic macrocycle assembly more often results in π-stacked supramolecular polymers (open/infinite assemblies). Examples of π-stacked discrete oligomers (closed assemblies) are comparatively rare and in some cases are controlled by guest binding, including protein complexation. − In pctx, the tetrahedral closed assembly arises from back-to-back π-stacking, as originally observed with the related macrocycle cyclotriveratrylene. , The coplanar π–π stacking of pctx (centroid···centroid = 3.7 Å) buries ∼245 Ų per macrocycle suggesting a significant contribution of the hydrophobic effect to cluster stability. While pctx has intrinsically lower binding affinity compared to sclx ₄ , the self-assembly capabilities of pctx confer more interesting possibilities for fabricating protein-based materials.

Conclusions

The noncovalent modification of proteins and the fabrication of (porous) assemblies have diverse applications from biopharmaceuticals to biomaterials. − This first study of pctx–protein interactions demonstrates selectivity for arginine or for small and accessible N-terminal residues. The phosphocavitand, readily synthesized in three steps, is a potentially valuable modulator of protein–protein interfaces which frequently involve arginine residues. The modification of pctx with additional protein binding features (such as peptide appendages) may be necessary, and will benefit from existing synthetic strategies. , The ease of introducing a Met-Lys N-terminal motif into proteins of interest, suggests a route to Zn²⁺/pctx controlled assembly. More generally, the N-terminal complexation and the alternative pctx cluster formations (trimeric versus tetrameric) indicate a promising versatility and scope for assembling different protein types. In some crystal structures with tetrameric clusters, only three of the cavities are utilized, while the capping macrocycle is empty. Arginine complexation by the capping macrocycle (in cocrystals with RSL-R₆) suggests the possibility of heterogeneous assemblies where different protein types with distinct binding features assemble on the same cluster. It is easy to envisage arrangements with only two or all four cavities utilized for protein binding, as well as the possibility of one or more cavities accommodating other guest types. Future research will focus on (1) the repertoire of N-terminal residues that are suitable guests, (2) how protein–pctx interactions lead to macrocycle clusters that mediate assembly/nucleate crystallization and (3) mixed protein assemblies on pctx clusters.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.5c08121.

• Methods; X-ray data statistics; noncovalent bond lengths; structure diagrams; HSQC spectra; and Zn SAD Fourier maps (PDF)

The authors declare no competing financial interest.