Protein Domain Mimics as Modulators of Protein-Protein Interactions

Abstract

Protein-protein interactions (PPIs) are ubiquitous in biological systems and often misregulated in disease. As such, specific PPI modulators are desirable to unravel complex PPI pathways and expand the number of druggable targets available for therapeutic intervention. However, the large size and relative flatness of PPI interfaces make them challenging molecular targets. This Account describes our systematic approach using secondary and tertiary protein domain mimics (PDMs) to specifically modulate PPIs. Our strategy focuses on mimicry of regular secondary and tertiary structure elements from one of the PPI partners to inspire rational PDM design. We have compiled three databases (HIPPDB, SIPPDB, and DIPPDB) of secondary and tertiary structures at PPI interfaces to guide our designs and better understand the energetics of PPI secondary and tertiary structures. Our efforts have focused on three of the most common secondary and tertiary structures: α-helices, β-strands, and helix dimers (e.g., coiled coils). To mimic α-helices, we designed the hydrogen bond surrogate (HBS) as an isosteric PDM and the oligooxopiperazine helix mimetic (OHM) as a topographical PDM. The nucleus of the HBS approach is a peptide macrocycle in which the N-terminal i, i + 4 main-chain hydrogen bond is replaced with a covalent carbon-carbon bond. In mimicking a main-chain hydrogen bond, the HBS approach stabilizes the α-helical conformation while leaving all helical faces available for functionalization to tune binding affinity and specificity. The OHM approach, in contrast, envisions a tetrapeptide to mimic one face of a two-turn helix. We anticipated that placement of ethylene bridges between adjacent amides constrains the tetrapeptide backbone to mimic the i, i + 4, and i + 7 side chains on one face of an α-helix. For β-strands, we developed triazolamers, a topographical PDM where the peptide bonds are replaced by triazoles. The triazoles simultaneously stabilize the extended, zigzag conformation of β-strands and transform an otherwise ideal protease substrate into a stable molecule by replacement of the peptide bonds. We turned to a salt bridge surrogate (SBS) approach as a means for stabilizing very short helix dimers. As with the HBS approach, the SBS strategy replaces a noncovalent interaction with a covalent bond. Specifically, we used a bis-triazole linkage that mimics a salt bridge interaction to drive helix association and folding. Using this approach, we were able to stabilize helix dimers that are less than half of the length required to form a coiled coil from two independent strands. In addition to demonstrating the stabilization of desired structures, we have also shown that our designed PDMs specifically modulate target PPIs in vitro and in vivo. Examples of PPIs successfully targeted include HIF1α/p300, p53/MDM2, Bcl-xL/Bak, Ras/Sos, and HIV gp41. The PPI databases and designed PDMs created in these studies will aid development of a versatile set of molecules to probe complex PPI functions and, potentially, PPI-based therapeutics.

Figure 1

Examples of (A) α-helices, (B) β-strands, and (C) helix dimers at protein–protein interaction interfaces. PDB codes: (A) 1BXL (Bcl-xL/Bak), 1YCR (p53/MDM2); (B) 1OY3 (NF-κB homodimer), 1F3U (Rap30/Rap74); (C) 2IW5 (LSD1/CoREST), 3CL3 (vFLIP/IKKγ).

Figure 2

Distribution of α-helix hotspots at PPI interfaces. Hotspot frequency is plotted as a bar graph (A) and on an idealized α-helix (B). The data show increases in hotspot frequency every 3–4 residues, corresponding to alignment of hotspots on a single α-helical face (helical faces indicated with black arcs). The color legend depicts fractional occurrence of hot spot residues (ΔΔG for alanine mutation > 1 kcal/mol) at each positon.

Figure 3

Design rationale for hydrogen bond surrogate (HBS) helices and oligooxopiperazine helix mimetics (OHMs). (A) (Left) An α-helix projects its i, i + 4, and i + 7 side chains on a single helical face. A tetrapeptide can be modeled to mimic these side chain projections (center left) and synthetically realized using half-chair oxopiperazine rings (center). Overlay of the OHM model with an idealized α-helix (center right) shows excellent mimicry of a single helical face. The general synthetic scheme for OHMs (right): (a) oNBS-Cl, 2,4,6-collidine; (b) 2-bromoethanol, triphenylphosphine (PPh₃), diisopropylazodicarboxylate (DIAD); (c) base (e.g., 1,8-diazabicyclo(5.4.0)undec-7-ene (DBU)); (d) 2-mercaptoethanol, DBU. (B) (Left) The helix-nucleating N-terminal hydrogen bond can be transformed into a covalent bond (center left), specifically a carbon–carbon bond (center). Overlay of the HBS helix crystal structure with an α-helix shows excellent mimicry. The general synthetic scheme for HBS helices (right): (a) SPPS for secondary amine, (b) SPPS, (c) 4-pentenoic acid, N,N′-diisopropylcarbodiimide (DIC); (d) ring-closing metathesis (RCM).

Figure 4

In vivo modulation of hypoxia signaling by designed α-helix PDMs. The center image shows the CBP CH1 domain (surface representation) interacting with HIF1α CTAD (ribbon) (PDB 1L8C). The upper panels show the HIF1α helix (center) with hotspots mimicked using OHM (left) and HBS (right) PDMs and K_d values for each PDM and negative controls. The K_d value for the unconstrained peptide counterpart, Ac-ELARALDQ-NH₂, is 6 μM. The lower panels show reduction in mouse xenograft tumor volume after a sequence of parenteral 13–15 mg/kg injections with the HIF1α-derived OHM (left) or HBS (right) PDMs.

Figure 5

Distribution of β-strand hotspots at PPI interfaces. Hotspot frequency is plotted as a bar graph (A) and on an idealized β-strand (B). With the exception of position i + 2, the data show a minimal increase in hotspot frequency every 2 residues, corresponding to alignment of hotspots on a single β-strand face. The color legend depicts fractional occurrence of hot spot residues (ΔΔG for alanine mutation > 1 kcal/mol) at each positon.

Figure 6

Triazolamers as β-strand mimetics. NMR (A) and X-ray crystallographic (B) data support a zigzag geometry for triazolamers that mimics β-strands. The solid and dashed arrows in panel (A) indicate strong and weak ROESY cross peaks, respectively, in the NMR spectrum. (C) The interaction of the tetrapeptide inhibitor L-400,417 (stick representation) with HIV-1 protease (HIVPR) dimer (green cartoon) was used to guide triazolamer HIVPR inhibitor design (overlay on right with L-400,417 in green and a triazolamer design in purple). (D) Examples of a triazolamer inhibitor (left) and negative control (right), showing IC₅₀’s for HIVPR protease inhibition.

Figure 7

Distribution of helix dimer hotspots at PPI interfaces. The frequency of particular spacings between the first and last hotspots is plotted as a bar graph (A) and on an idealized coiled coil (B). The spacing between first and last hotspots is usually shorter than 3 heptads. The color legend depicts fractional occurrence of hot spot residues (ΔΔG for alanine mutation > 1 kcal/mol) at each spacing.

Figure 8

Salt bridge surrogate (SBS) helix dimers as PPI inhibitors. (A) Different bis-triazole SBS linkages (middle left) were synthesized as part of an idealized coiled coil (top left) and monitored by circular dichroism (CD, bottom left), demonstrating that a propargyl ether-azidolysine linkage (green) favored helix formation. A helical wheel diagram and structural model based on NMR restraints are shown on the right. (B) The SBS approach was applied to the NHR2 coiled coil (top left). Though the SBS alone did not stabilize the helix dimer as monitored by CD, combining the SBS with optimization of the helix dimer interface yields a minimal helix dimer with significant helicity and native-like binding affinity for N2B (NHR2-N2B K_d = 356 μM). Binding affinity was further enhanced with a disulfide bridge.