Molecular Modeling of Single- and Double-Hydrocarbon-Stapled Coiled-Coil Inhibitors against Bcr-Abl: Toward a Treatment Strategy for CML

Abstract

The chimeric oncoprotein Bcr-Abl is the causative agent of virtually all chronic myeloid leukemias and a subset of acute lymphoblastic leukemias. As a result of the so-called Philadelphia chromosome translocation t(9;22), Bcr-Abl manifests as a constitutively active tyrosine kinase, which promotes leukemogenesis by activation of cell cycle signaling pathways. Constitutive and oncogenic activation is mediated by an N-terminal coiled-coil oligomerization domain in Bcr (Bcr-CC), presenting a therapeutic target for inhibition of Bcr-Abl activity toward the treatment of Bcr-Abl⁺ leukemias. Previously, we demonstrated that a rationally designed Bcr-CC mutant, CCmut3, exerts a dominant negative effect upon Bcr-Abl activity by preferential oligomerization with Bcr-CC. Moreover, we have shown that conjugation to a leukemia-specific cell-penetrating peptide (CPP-CCmut3) improves intracellular delivery and activity. However, our full-length CPP-CCmut3 construct (81 aa) is encumbered by an intrinsically high degree of conformational variability and susceptibility to proteolytic degradation relative to traditional small-molecule therapeutics. Here, we iterate a new generation of CCmut3 inhibitors against Bcr-CC-mediated Bcr-Abl assembly designed to address these constraints through incorporation of all-hydrocarbon staples spanning i and i + 7 positions in α-helix 2 (CPP-CCmut3-st). We utilize computational modeling and biomolecular simulation to evaluate single- and double-stapled CCmut3 candidates in silico for dynamics and binding energetics. We further model a truncated system characterized by the deletion of α-helix 1 and the flexible loop linker, which are known to impart high conformational variability. To study the impact of the N-terminal cyclic CPP toward model stability and inhibitor activity, we also model the full-length and truncated systems devoid of the CPP, with a cyclized CPP, and with an open-configuration CPP, for a total of six systems that comprise our library. From this library, we present lead-stapled peptide candidates to be synthesized and evaluated experimentally as our next iteration of inhibitors against Bcr-Abl.

Introduction

Protein oligomerization is pivotal for a diverse range of cellular processes, including signal transduction, gene expression, enzyme activity and regulation, and protein binding.¹⁻¹⁴ Such processes reflect but a fraction of the complete human interactome, composed of an estimated 650,000 unique protein–protein interactions (PPIs) responsible for regulating physiological functions.¹⁵ Consequently, dysregulation within the interactome is also responsible for the development of disease states.¹⁶⁻¹⁹ Indeed, aberrant oligomerization has been implicated in various pathological conditions and conformational diseases, such as diabetes mellitus,^20,21 Alzheimer’s disease,²²⁻²⁴ and Bcr-Abl⁺ cancers including chronic myeloid leukemia (CML) and acute lymphoblastic leukemia (ALL).²⁵⁻²⁷

Approximately 95% of CML and 20% of ALL cases are characterized by the so-called Philadelphia chromosome (Ph), a genetic abnormality which manifests due to a balanced reciprocal translation t(9;22) (q34;g11).²⁵⁻²⁷ In this event, the ABL1 gene of chromosome 9 is fused to the breakpoint cluster region gene of chromosome 22, producing the BCR-ABL1 oncogene which encodes for the chimeric oncoprotein Bcr-Abl, a nonreceptor tyrosine kinase with constitutive activity. Analogous to receptor tyrosine kinases, Bcr-Abl is activated by trans-autophosphorylation of tyrosine residues between kinase domains, clustered together upon homodimerization.²⁸⁻³⁰ These phosphotyrosine residues then act as docking sites for SH2 domain substrates, activating RAS, PI3K/AKT, JAK/STAT, and MAPK signaling pathways which promote cell proliferation, inhibit apoptosis, and enhance cell motility—conferring growth advantage for leukemic cells.³¹⁻³³

Bcr-Abl assembles as a dimer of dimers, mediated by an N-terminal oligomerization, or coiled-coil, domain in Bcr (Bcr-CC). Monomers associate as a dimer through formation of an antiparallel coiled-coil and domain swapping, further associating as a tetramer through dimer stacking³⁰ (Figure 1A). Bcr-CC (72 aa) consists of a short N-terminal helix (α1; residues 5–15) and a long C-terminal helix (α2; residues 28–67) connected by a flexible loop linker region. α-helix 2 predominantly composes the dimeric interface, while α-helix 1 domain swaps against α-helix 2 of the opposing monomer—packing the interface (Figure 1B).³⁰ Deletion of Bcr-CC demonstrably diminishes transformation potential and interleukin-3 dependence imparted upon lymphoid cells, while heterodimerization between Bcr-Abl and an isolated coiled-coil domain diminishes transformation and abrogates transphosphorylation.³⁴⁻³⁶ Collectively, these findings substantiate Bcr-CC as an integral feature of leukemogenesis and highlight the dimeric interface as a therapeutic target for treating Bcr-Abl⁺ leukemias.

Figure 1.

Structure and design of CCmut3, a Bcr-CC oligomerization inhibitor. (A) Bcr-Abl associates as a dimer through the formation of an antiparallel coiled-coil and domain swapping, further associating as a tetramer through dimer stacking. Oligomerization and consequent oncogenic activation are mediated by the Bcr-CC domain, the target of CCmut3. (B) CCmut3 (blue) and Bcr-CC (gray) assemble as an antiparallel coiled-coil, promoted by six salt bridges (55R-E34, 41R-E48, 34E-R55, 46E-R53, 53R-E46, and 60E-K39) in the dimer interface. (C) CCmut3 (blue) incorporates five mutations (magenta) in α-helix 2 which are designed to promote preferential heterodimer formation with Bcr-CC (S41R and Q60E) while dissuading formation of native and mutant homodimers (K39E, E48R, and L45D). A sixth C38A mutation is included for crystallization purposes.

Previously, we designed and developed a dominant-negative Bcr-CC mutant (CCmut3) through a combinatorial approach utilizing biomolecular simulation, rational mutagenesis, and in vitro experimentation.³⁷⁻³⁹ CCmut3 incorporates a set of six α-helix 2 mutations (K29E, C38A, S41R, L45D, E48R, and Q60E) (Figure 1C) which promote the formation of preferential heterodimers with wild-type Bcr-CC while dissuading the formation of deleterious Bcr-CC and CCmut3 homodimers. This CCmut3 construct demonstrably reduces transformation and proliferation, induces apoptosis, and diminishes phosphorylation of Bcr-Abl and its substrates in Bcr-Abl⁺ cells, including those with the clinically significant T315I point-contact mutant and the T315I/E255 V compound mutant.³⁷⁻⁴⁰ In a successive design iteration, a leukemia-specific cell-penetrating peptide (LS-CPP) was conjugated to the N-terminus of CCmut3 (CPP-CCmut3) to improve intracellular delivery and activity.^40,41 This CPP (CAYHRLRRC), characterized by an arg-rich cell-penetrating motif (RLRR) and a lymph node homing motif (CAY), was discovered by phage display, conferring greatly improved binding of directed phage to a panel of leukemia/lymphoma cells and patient samples compared against nonleukemic cells.⁴¹ Indeed, we have shown that CPP-CCmut3 exhibits markedly improved internalization and activity against Bcr-Abl compared against wild-type Bcr-CC and untargeted CCmut3.⁴⁰ However, translation of the full-length CPP-CCmut3 construct (81 aa) necessitates further design iteration to address the intrinsically high degree of conformational variability, relative to “traditional” small-molecule therapeutics, which predisposed it to loss of helicity and proteolytic degradation in serum and intracellularly.

Herein, we iterate a new generation of our inhibitor against Bcr-CC-mediated Bcr-Abl assembly that is designed to reduce conformational variability and improve proteolytic stability through the incorporation of single- and double-all-hydrocarbon “staples” (Figure 2A). Molecular staple cross-links have emerged as an effective strategy to reinforce α-helices and stabilize peptides by constraining their helical conformation. Moreover, hydrocarbon staples have been widely reported to confer proteolytic resistance, enhance serum stability and half-life, and improve cellular uptake.⁴²⁻⁵¹ Single-stapled peptides have shown to improve half-life by up to 6–8 fold, while double-stapled peptides have shown up to a 24-fold enhancement in protease resistance.⁴² Staples between i and i + 7 positions span approximately 2 helical turns, providing maximal helical stability and protease resistance; thus, our stapled constructs exclusively incorporate staples at i and i + 7 positions within α-helix 2 (CPP-CCmut3-st) with the objective of stabilizing and protecting the primary binding interface. Examples of “ideal” i and i + 7 staple locations (used individually as single staples or as a double staple) include G29 and E36 and N50 and I57, which are on the backside of the helix and not in the dimerization interface surface. Sequence-scanning of α-helix 2, with omission of all residues key to preserving the dimer interface, yields 10 viable staple sites (Figure 1 and Tables 1 and 2). An additional staple site (F54-T61) occluded by sequence-scanning is included based on previous all-atom molecular dynamics (MD) simulations performed by the Cheatham group using a truncated CCmut3 sequence.³⁷ Here, we model these 11 staple sites as individual single-hydrocarbon staples (SHC) and in combinations as double-hydrocarbon staples (DHC).

Figure 2.

Structural representations of CCmut3 systems and hydrocarbon staples. (A) Hydrocarbon staples are applied to α-helix 2 of CCmut3 (blue) to stabilize and protect the dimer interface. (B) CCmut3 full-length (CCmut3F) consists of both α-helices 1 (light blue) and 2 (blue), with the LS-CPP conjugated to the N-terminal of α-helix 1. (C) CCmut3 truncated (CCmut3T) is characterized by deletion of α-helix 1, consisting only of α-helix 2 (blue), to which the LS-CPP is conjugated. In both full-length and truncated systems, α-helix 2 comprises the central dimer interface with Bcr-CC (gray).

Table 1. Modeled System Variants without Staples, from Which Stapled Systems Are Built Upon^a.

unstapled systems	system ID	heterodimer segments	residues	total aa
CCmut3 truncated	CCmut3T	CCmut3T	28–67	112
		Bcr-CC	68–139
CCmut3 truncated with cyclic CPP	CCmut3TC	CPP	19–27	121
		CCmut3T	28–67
		Bcr-CC	68–139
CCmut3 truncated with open CPP *	CCmut3TC′	CPP	19–27	121
		CCmut3T	28–67
		Bcr-CC	68–139
CCmut3 full-length	CCmut3F	CCmut3F	01–72	144
		Bcr-CC	73–144
CCmut3 full-length with cyclic CPP	CCmut3FC	CPP	(−)09–0	153
		CCmut3F	01–72
		Bcr-CC	73–144
CCmut3 full-length with open CPPb	CCmut3FC′	CPP	(−)09–0	153
		CCmut3F	01–72
		Bcr-CC	73–144

^aTruncated and full-length system variants are divided into three configurations: without CPP, with an open CPP, and with a closed, cyclic CPP.

^bFor single-stapled systems, the CPP part was tested in cyclic and opened forms.

Table 2. Comprehensive Library of All Single- and Double-Stapled Candidates for Modeling^a.

single-hydrocarbon-stapled peptides (SHC)
truncated systems (SHCT/SHCTC/SHCTC′)	stapled residues	full-length systems (SHCF/SHCFC/SHCFC′)	stapled residues
G29-E36	Gly29 and Glu36	G29-E36	Gly29 and Glu36
D30-R37	Asp30 and Arg37	D30-R37	Asp30 and Arg37
Q33-A40	Gln33 and Ala40	Q33-A40	Gln33 and Ala40
E36-R43	Glu36 and Arg43	E36-R43	Glu36 and Arg43
R37-R44	Arg37 and Arg44	R37-R44	Arg37 and Arg44
A40-Q47	Ala40 and Gln47	A40-Q47	Ala40 and Gln47
R43-N50	Arg43 and Asn50	R43-N50	Arg43 and Asn50
R44-Q51	Arg44 and Gln51	R44-Q51	Arg44 and Gln51
N50–I57	Asn50 and Ile57	N50–I57	Asn50 and Ile57
I57-A64	Ile57 and Ala64	I57-A64	Ile57 and Ala64
		F54-T61	Phe54 and Thr61

double-hydrocarbon-stapled peptides (DHC)
truncated systems (DHCT/DHCTC)	stapled residues	full-length systems (DHCF/DHCFC)	stapled residues
G29-E36_R37-R44	Gly29 and Glu36	G29-E36_R37-R44	Gly29 and Glu36
	Arg 37 and Arg 44		Arg 37 and Arg 44
G29-E36_A40-Q47	Gly29 and Glu36	G29-E36_A40-Q47	Gly29 and Glu36
	Ala40 and Gln47		Ala40 and Gln47
G29-E36_R43-N50	Gly29 and Glu36	G29-E36_R43-N50	Gly29 and Glu36
	Arg43 and Asn50		Arg43 and Asn50
G29-E36_R44-Q51	Gly29 and Glu36	G29-E36_R44-Q51	Gly29 and Glu36
	Arg44 and Gln51		Arg44 and Gln51
G29-E36_N50–I57	Gly29 and Glu36	G29-E36_N50–I57	Gly29 and Glu36
	Asn50 and Ile57		Asn50 and Ile57
G29-E36_I57-A64	Gly29 and Glu36	G29-E36_I57-A64	Gly29 and Glu36
	Ile57 and Ala64		Ile57 and Ala64
Q33-A40_R43-N50	Gln33 and Ala40	G29-E36_F54-T61	Gly29 and Glu36
	Arg43 and Asn50		Phe54 and Thr61
		Q33-A40_R43-N50	Gln33 and Ala40
			Arg43 and Asn50

^aFor single-staple systems, 10 and 11 staple sites were selected for truncated and full-length variants, respectively. From these sites, sets of residue pairs were selected to produce 7 and 8 sets of staple sites for truncated and full-length double-stapled variants, respectively.

The affinity and specificity of PPIs are intrinsically sequence-specific, so caution must be exercised in the selection of staple sites such that these features are not meaningfully diminished by substitution. Additionally, CCmut3 is characterized by a multitude of potential staple sites, economically and temporally precluding the possibility of synthesizing a comprehensive library. Strategies commonplace in stapled peptide design such as sequence-scanning and rational selection informed by high-resolution structures are consequently insufficient as individual methods in their inability to predict behavioral outcomes and screen a reasonably sized library.^49,52 Instead, we employ a combinatorial approach utilizing rational mutagenesis, sequence-scanning, and molecular modeling to design and characterize single- and double-stapled candidates, building upon our seminal work modeling staples applied to a truncated Bcr-CC sequence.³⁷ In addition to full-length CPP-CCmut3, we model a truncated system characterized by deletion of α-helix 1 and the flexible loop linker, which are known to impart high conformational variability. To study the impact of the N-terminal cyclic CPP on model stability and inhibitory activity, we also model the full-length (Figure 2B) and truncated (Figure 2C) systems without CPP, with cyclized CPP, and with linear CPP, for a total of six systems.

Methods

Model Building

Construction of the modified Bcr-Abl oligomerization domain was informed by the crystal structure of the N-terminal oligomerization domain of Bcr-Abl [Protein Data Bank (PDB) entry 1K1F], utilizing residues 1–67 of chain A and chain B to generate the models.³⁰ To align with the wild-type protein, the UCSF Chimera⁵³ swapaa tool was used to convert selenomethionine residues and residue 38 into methionine residues and cysteine, respectively. The swapaa tool was further utilized to generate modified CCmut3 constructs by sequential implementation of mutations C38A, K39E, S41R, L45D, E48R, and Q60E on Chain A (residues 1–67). No modifications were implemented for Chain B (residues 68–134) beyond previously described changes required for reversion to the wild-type form. Truncated CCmut3 constructs (residues 28–67) were generated by deletion of the highly conformationally variable CCmut3 α1 region (residues 1–27), while full-length CCmut3 constructs retained the CCmut3 α1 region. Full-length and truncated model variants with the leukemia-specific CPP were produced by conjugation of the 9 aa sequence (CAYHRLRRC)¹⁷ to the N-terminus, allowing investigation of the targeting moiety’s effect upon overall function and behavior of the protein. Heterodimers were divided into sequence segments corresponding to CCmut3T (residues 28–67) and Bcr-CC (residues 68–139) for truncated constructs and CCmut3F (residues 1–72) and Bcr-CC (residues 73–144) for full-length constructs. An additional sequence segment corresponding to the N-terminal conjugated CPP was also included for truncated (residues 19–27) and full-length (residues (−)9–0) (Table 1). For both truncated and full-length systems, the CPP was tested in a cyclic and opened form in single-stapled systems (with the closed form being the expected conformation in vitro/in vivo).⁴¹ Staples were incorporated into the molecular structure using the AMBER general Amber force field⁵⁴ and restrained electrostatic potential⁵⁵ charges to produce our single- and double-stapled construct library (Table 2).

Staple sites were incorporated into CCmut3 coils by residue replacement within the PDB file. Models were constructed using AMBER ff14SB force-field parameters⁵⁶ and were explicitly solvated in a truncated octahedron of 10 Å surrounding buffer of TIP3P water molecules.⁵⁷ Net-neutralizing counterions (Na⁺/Cl^–) were incorporated into the system using the ion parameters developed by Joung and Cheatham⁵⁸ to maintain overall charge neutrality. Additionally, 50 extra Na⁺/Cl^– atoms were included to simulate a physiological ion concentration of approximately 200 mM. To enhance the efficiency of MD simulations, hydrogen masses of the solute were repartitioned to a value of 3.024 Da to double the MD time integration step and enable a time step of 4 fs.⁵⁹

Minimization and Equilibration

A 10-step simulation preparation protocol for explicitly solvated biomolecules, as developed and described by Roe and Brooks,⁶⁰ was applied to all models to ensure system stability in subsequent production MD. The first nine steps of this protocol consist of a series of energy minimization and relaxation stages, in which there are 4,000 total minimization steps and 40,000 total relaxation steps for a total 45 ps. In the final step, production simulation settings are applied to the NPT ensemble and are performed continuously until final density plateau or “stabilization” criteria are achieved (Supporting Information: Full Minimization and Equilibration Protocol). All steps were performed using double precision on central processing unit codes. Constant pressure was controlled using a Monte Carlo barostat,⁶¹ and constant temperature was controlled using a Langevin thermostat⁶² with a collision frequency of 10 ps^–1.

Production MD Simulations

All production MD simulations were performed with the graphics processing unit implementation of the Amber20 modeling suite using NVIDIA GeForce GTX 1080ti GPUs.⁶³ MD simulations were performed in an explicit solvent using a 4 fs time step and a Langevin thermostat with a collision frequency of 10 ps^–1 to regulate constant temperature and pressure. A 10 Å cutoff was used to calculate nonbonded interactions, the particle mesh Ewald method⁶⁴ was used to treat electrostatics, and the SHAKE algorithm⁶⁵ was used to constrain bond lengths involving hydrogen atoms. Three independent copies of each system were simulated for statistical purposes and to account for variability using a random seed for each to prevent synchronization artifacts. Each simulation for each system in this study was conducted for approximately 5 μs of simulation time.

Analysis Protocols

Candidate peptides were analyzed using the CPPTRAJ toolset, an integral component of the AmberTools23 package. Clustering was conducted upon the last 1000 frames of trajectories using the DBScan algorithm and an epsilon value of 2.0. This algorithm is well suited for identifying clusters within complex and noisy data sets.⁶⁶

Energetic analyses were performed using the molecular mechanics Poisson–Boltzmann surface area (MM-PBSA) methodology,⁶⁷ by which the relative binding energetics between construct peptides and the Bcr-CC target peptide were estimated using single trajectories. Relative overall free energies were calculated by using the MM-PBSA.py program within the AmberTools23 suite. Binding energetics between each unique construct (stapled and unstapled CCmut3) and native Bcr-CC were calculated from trajectory snapshots of the final 2000 frames with an ionic strength of 0.1 M. Simulations were performed in triplicate by using independent simulation copies.

Salt bridges were analyzed within a distance threshold of 4.0 Å using MDAnalysis.⁶⁸ Specific salt bridge interactions include: Glu with side chain atoms OE1 and OE2, Asp with OD1 and OD2, Lys with NZ, and Arg with NH1 and NH2.

Results and Discussion

Using extensive all-atom MD simulations, we investigated the structural characteristics and dynamics of multiple variants of the stapled full-length and truncated CCmut3 sequences in complex with Bcr-CC. Both full-length and truncated sequences were constructed without the CPP (CCmut3F/CCmut3T), with the CPP in a closed configuration (CCmut3FC/CCmut3TC), and with the CPP in an open configuration (CCmut3FC′/CCmut3TC′), for a total of 3 full-length and 3 truncated variants (Table 1). These variants were modified by incorporating either a single-hydrocarbon (SHCF/SHCT) staple or a double-hydrocarbon (DHCF/DHCT) staple at amino acid positions i and i + 7, spanning two helical turns. Staple sites were meticulously chosen to mitigate disruption of the CCmut3:Bcr-CC binding interface and to preserve any crucial binding partners, such as salt bridges or contacts vital in maintaining the dimerization process. Of each full-length variant, 11 SHC staples were modeled; from each truncated variant, 10 SHC staples were modeled. DHC staple sites were selected from previous SHC staple sites and were modeled only for non-CPP and closed CPP variants. Of the two full-length variants, 8 DHC staples were modeled; of the two truncated variants, 7 DHC staples were modeled. An exhaustive list of both single- and double-staple sites, identified numerically and by residue pairs, is provided in Table 2. In total, 93 unique stapled constructs were modeled and investigated.

Simulations of Stapled Constructs in Complex Provide Dynamic Insights

To assess structural changes induced by incorporation of hydrocarbon staples, stapled constructs in complex with Bcr-CC were evaluated against the native heterodimer for residue-specific flexibility, as determined by root-mean-square fluctuation (RMSF) analysis (Figures 3 and 4). Heterodimer systems were divided into two sequence segments corresponding to each monomer: CCmut3T (residues 28–67) and Bcr-CC (residues 68–139) for truncated constructs and CCmut3F (residues 1–72) and Bcr-CC (residues 73–144) for full-length constructs. An additional sequence segment corresponding to the N-terminal conjugated CPP was also included for truncated (residues 19–27) and full-length (residues (−)9–0) constructs. With a variety of staple sites and residue pairs implemented across the 93 constructs, this approach allows for a comparison of interspecific differences while considering proximity to staple sites.

Figure 3.

RMSF histograms examining single-staple constructs in the bound state report minimal deviations in the residue flexibility. (A–C) Residues 19–27 correspond to the N-terminal LS-CPP, residues 28–67 correspond to truncated CCmut3, and residues 68–139 correspond with Bcr-CC. (D–F) Residues (−)9–0 correspond with the N-terminal LS-CPP, residues 1–72 correspond with full-length CCmut3, and residues 73–144 correspond with Bcr-CC.

Figure 4.

RMSF histograms examining double-staple constructs in the bound state report minimal deviations in residue flexibility. (A,B) Residues 19–27 correspond with the N-terminal LS-CPP, residues 28–67 correspond with truncated CCmut3, and residues 68–139 correspond with Bcr-CC. (C,D) Residues (−)9–0 correspond to the N-terminal LS-CPP, residues 1–72 correspond to full-length CCmut3, and residues 73–144 correspond with Bcr-CC.

CCmut3 Truncated with Single Staple (SHCT/SHCTC/SHCTC′)

In the truncated CCmut3 systems with single staples, minimal changes were observed in overall fluctuation compared to the unstapled CCmut3T analogue (Figure 3A). For all constructs, fluctuations were expectedly greatest at the N and C-termini of CCmut3T (residues 28 and 67, respectively) and the C-terminus of Bcr-CC (residue 139), with increased fluctuation also apparent in α-helix 1 of Bcr-CC (residues 73–81), which is of particular interest due to important π-stacking interactions that occur between Bcr-CC α1 and CCmut3 α2 to stabilize the dimer. These heightened fluctuations persist into the flexible loop linker region between α-helices 1 and 2 (residues 82–94). Closely examining α-helix 2 of CCmut3T, in which staple sites were incorporated for all constructs, we observe no apparent divergence between constructs and control in the RMSF space. Constructs of note include SHCT-D30-R37, which exhibits an enhanced degree of fluctuation and thus indicates increased system dynamics, and SHCT-R43-N50, which comparatively exhibits a diminished degree of fluctuation and thus indicates greater configuration stability.

To the N-terminal of this truncated CCmut3 system, a self-cyclizing CPP was conjugated to produce two more system variants—the first with a closed configuration (Figure 3B) and the second with an open configuration (Figure 3C). Within each configuration, distinct fluctuation patterns were observed. Of note, a higher degree of fluctuation is observed for the SHCTC-A40-Q47 construct, particularly at the staple site (aa 40–47). These fluctuations may indicate an ongoing attempt by the system to adapt or adjust to the introduced changes, as such behavior is not observed in the system with the open-configuration CPP. Constructs SHCTC-R44-Q51 and SHCTC′-R44-Q51 appreciably deviate in the RMSF space around the Bcr-CC α1 region and around the staple site, suggesting distinctive regional interactions or structural features that contribute to the observed fluctuations, indicative of the staple site.

CCmut3 Full-Length with Single Staple (SHCF/SHCFC/SHCFC′)

In the full-length CCmut3 systems with single staples, fluctuations are generally increased compared to truncated systems due to inclusion of α-helix 1 in CCmut3F (Figure 3D). α-Helix 1 of CCmut3F (amino acids 1–27), which is absent in truncated CCmut3T systems, increases interactions between CCmut3 and Bcr-CC, consequently increasing overall fluctuation. Fluctuations in the α1 region are particularly pronounced for the SHCF-A40-Q47 construct, which also exhibits a significant fluctuation in the α2 region. In the SHCF-R44-Q51 construct, we observe a substantial fluctuation in the α1 region of Bcr-CC, indicating significant movement and dynamics in this specific region. The smallest fluctuation is observed in the SHCF–N50-I57 system, suggesting a more stable and well-accommodated state for the system. This reduced fluctuation indicates that the SHCF–N50-I57 system may accommodate a more favorable conformation or interaction, contributing to its enhanced stability.

As performed with the truncated CCmut3 system, a self-cyclizing CPP was conjugated to produce a closed configuration system (Figure 3E) and an open-configuration system (Figure 3F). The behaviors of these CPP-conjugated systems are consistent between the truncated and full-length variants, with open SHCFC′ constructs exhibiting generally higher fluctuations in the RMSF space compared to closed SHCFC constructs, which we attribute to the formation of a loop between the α-helix regions. Figure 3E shows that the region corresponding to α1 in CCmut3 experiences the most significant fluctuations in the SHCFC-D30-R37 and SHCFC-R43-N50 systems. Similarly, residues 116–130 in the α2 region of Bcr-CC also exhibit substantial fluctuations, suggesting a dynamic interaction between α1 CCmut3 and α2 Bcr-CC crucial in maintaining dimer stability. In Figure 3F, the largest fluctuation in the α1 region of CCmut3 is observed in the SHCFC′-D30-R37 system. However, we do not observe a similar trend in the α2 regions of Bcr-CC. Interestingly, both the SHCFC-F54-T61 and SHCFC′-F54-T61 systems display increased fluctuations in the α2 region of Bcr-CC, indicating more movement in this region.

CCmut3 Truncated with Double Staples (DHCT/DHCTC)

Relative to their single-stapled counterparts, truncated CCmut3 systems with double staples characteristically exhibit greater fluctuations in the α1 and flexible loop linker regions of Bcr-CC (residues 73–94), indicating enhanced motility in this region (Figure 4A). Consistent with our observations in single-staple systems, CCmut3T and Bcr-CC termini assume loop conformations, with a consistent peak in fluctuation between residues 68–77 corresponding to the CCmut3T C-terminus and Bcr-CC N-terminus. Of particular interest, we also observe greater variability in the RMSF space around the α2 region of CCmut3, although only minimal losses in helicity are observed in this region (Figure S1A). Closely examining all DHCT systems, we note constructs DHCT-G29-E36_I57-A64 and DHCT-Q33-A40_R43-N50 which exhibit fewer fluctuations overall and, most crucially, in these Bcr-CC α1 and CCmut3 α2 regions, which are otherwise more volatile.

For the closed configuration CPP-conjugated systems (Figure 4B), similar fluctuation patterns were observed, although inclusion of the CPP produces an apparent upward shift in fluctuation for the control, particularly within the Bcr-CC α1 region. For these systems, stapled constructs consistently exhibit few RMSF fluctuations, with exception to construct DHCTC-G29-E36_I57-A64 in the Bcr-CC α1 region and construct DHCTC-G29-E36_A40-Q47 in the CCmut3 α2 region, the latter corresponding to the Gly29-Glu36 staple site. An appreciable loss in helicity across and around staple site Gly29-Glu36 is also manifested for construct DHCTC-G29-E36_A40-Q47, indicating this specific staple’s culpability in destabilizing the structure (Figure S1B). The large RMSF discrepancies between DHCT-G29-E36_I57-A64 and DHCTC-G29-E36_I57-A64 are particularly curious, with addition of the CPP producing a disproportionately destabilizing effect relative to its unconjugated analogue, which may be due to interactions between the CPP and helix α1 of Bcr-CC. We have previously shown the propensity for staples in the α2 region to interact with the flexible tail of α1 (residues 1–27), to which we have now conjugated the LS-CPP. It is not unreasonable to us then that addition of this 9 aa CPP might further influence or augment these interactions for select staple sites. Conversely, construct DHCTC-Q33-A40_R43-N50 presents the most promising for this system subset, exhibiting fewer fluctuation overall and also within the Bcr-CC α1 and CCmut3 α2 regions specifically.

CCmut3 Full-Length with Double Staples (DHCF/DHCFC)

The full-length double-stapled CCmut3 systems generally exhibit RMSF profiles similar to those of their single-stapled counterparts, although with less predictable aberrant shifts (Figure 4C). Specifically, constructs DHCF-G29-E36_A40-Q47 and DHCF-G29-E36_R44-Q51 exhibit large upward shifts in the CCmut3 α1 and Bcr-CC α1 regions, respectively, which are inconsistent with other full-length constructs. Of particular interest, DHCF-G33-A40_R43-N50 exhibits aberrant upward fluctuation shifts in the CCmut3 α2 region corresponding to the Gln33-Ala40 and Arg43-Asn50 staple sites, as well as in the Bcr-CC α2 region which interfaces with CCmut3 α2. A similar shift is apparent for DHCF-G29-E36_R43-N50 in the Bcr-CC α2 region. Helical losses are apparent in the Bcr-CC α2 region for all four listed constructs, the most pronounced of which to the detriment of approximately 40 and 12% for constructs DHCF-G33-A40_R43-N50 and DHCF-G29-E36_R43-N50 at staple sites Arg43-Asn50 and Gly29-Gly36, respectively. Moreover, DHCF-G29-E36_R43-N50 exhibits further helical losses at its second staple site Arg43-Asn50, which is shared with DHCF-G33-A40_R43-N50 (Figure S2A). Constructs DHCF-G29-E36_I57-A64 and DHCF-G29-E36_F54-T61 present as the most promising for this system subset with fewer fluctuations overall and specifically within the α-helical regions.

Similar fluctuation patterns are generally observed for the cyclic CPP-conjugated systems (Figure 4D). Remarkably, the DHCFC-Q33-A40_R43-N50 system stands out with the lowest overall RMSF fluctuation—a stark contrast to its counterpart lacking the N-terminal CPP, which reflects heightened motility and reduced stability due to the destabilizing influence of its staple site. Construct DHCFC-G29-E36_A40-Q47 is noteworthy for exhibiting greater fluctuation in the region of its first staple, as well as increased fluctuation in the α-helix 1 of CCmut3, which is substantiated by a total loss of helicity across the Gly29-Gly36 staple site (Figure S2B).

Binding Affinity and Thermodynamic Insights

We incorporate hydrocarbon staples into our CCmut3 constructs with pharmacokinetics in mind—ultimately performed with an objective to constrain and stabilize the α-helices for delivery. However, it is imperative that the hydrophobic staples do not disrupt α-helical interactions between CCmut3 and Bcr-CC and reduce the potency. Toward this end, MM-PBSA calculations were performed to evaluate relative binding affinities between different systems and between lead CPP-CCmut-st and CPP-CCmut-dst candidates (Figures 5 and 6). To evaluate statistical significance, Tukey’s honestly significant difference (HSD) test was applied as a post hoc method to precisely compare specific pairs of groups and allow for more nuanced comparisons (Tables 3 and 4).

Figure 5.

Binding energy (kcal/mol) for CCmut3 systems with single staples and their unstapled analogues. Null CPP systems are shown in gray, open CPP systems are shown in blue, and closed CPP systems are shown in orange. (A) For truncated CCmut3 systems, null CPP constructs are generally less energetically favorable than either open or closed CPP constructs regardless of staple sites. N50–I57 and I57-A64 staple sites are consistently among the most favorable in all groups. (B) For full-length CCmut3 systems, null CPP constructs are generally less energetically favorable than either open or closed CPP constructs, although extraordinary exceptions are observed for staple sites A40-Q47 and N50-I57, with N50-I57 consistently among the most favorable in all groups. Values are reported as mean ± SEM.

Figure 6.

Binding energy (kcal/mol) for full-length CCmut3 systems with single staples and their unstapled analogues. Null CPP systems are shown in gray, open CPP systems are shown in blue, and closed CPP systems are shown in orange. Null CPP systems are generally less energetically favorable than either open or closed CPP systems, although extraordinary exceptions are observed for staple sites A40-Q47 and N50-I57, with N50-I57 consistently among the most favorable in all groups. Values are reported as mean ± SEM.

Table 3. Mean Difference and Statistical Significance between Binding Free Energies (ΔG) of Various CCmut3 Systems, as Determined by Tukey’s HSD Post Hoc Test^a.

A	CCmut3 truncated with single-staple multiple comparison of mean—Tukey HSD, FWER = 0.05
comparison group 1	comparison group 2	Δ Mean	p-value	lower CI	upper CI	reject
SHCT	SHCTC	–11.53	0.001	–17.69	–5.37	true
SHCT	SHCTC′	–11.06	0.001	–17.22	–4.89	true
SHCTC	SHCTC′	0.47	0.9	–5.69	6.63	false

B	CCmut3 full-lengthwith single-staple multiple comparison of mean—Tukey HSD, FWER = 0.05
comparison group 1	comparison group 2	Δ Mean	adjusted p-value	lower CI	upper CI	reject
SHCF	SHCFC	–18.75	0.001	–27.05	–10.44	true
SHCF	SHCFC′	–19.33	0.001	–27.63	–11.02	true
SHCFC	SHCFC′	–0.58	0.9	–8.88	7.73	false

C	CCmut3 full-length and truncated with double staples multiple comparison of mean—Tukey HSD, FWER = 0.05
comparison group 1	comparison group 2	Δ mean	adjusted p-value	lower CI	upper CI	reject
DHCF	DHCFC	–18.89	0.001	–27.31	–10.47	true
DHCF	DHCT	12.70	0.0025	3.98	21.42	true
DHCF	DHCTC	–1.87	0.9p	–10.59	6.84	false
DHCFC	DHCT	31.59	0.001	22.87	40.31	true
DHCFC	DHCTC	17.01	0.001	8.30	25.73	true
DHCT	DHCTC	–14.57	0.001	–23.58	–5.57	true

^a(A,B) For truncated and full-length CCmut3 systems with single staples, CPP-null constructs are significantly less favorable than those of CPP-conjugated systems. (C) For CCmut3 systems with double staples, truncated and CPP-null constructs are significantly less favorable than full-length and CPP-conjugated constructs. A low adjusted p-value (-adj < 0.05) indicates a significant difference in means, while a greater adjusted p-value indicates insignificance. The “Reject” column is a binary indicator, where “True” indicates rejection of the null hypothesis, and “False” indicates no significant difference between groups.

Table 4. Mean Difference and Statistical Significance between Binding Free Energy ΔG of Lead-Stapled Candidates, as Determined by Tukey’s HSD Post Hoc Test^a.

lead candidate comparisons multiple comparison of mean—Tukey HSD,α = 0.05
comparison group 1	comparison group 2	Δ mean	p-value	lower CI	upper CI	reject
SHCF	SHCT	9.61	0.023	1.24	17.96	True
SHCF	SHCTC	–1.92	0.619	–6.04	2.19	False
SHCFC	DHCFC	2.39	0.467	–2.46	7.23	False

^aA low p-value (p-adj < 0.05) indicates a significant difference in means, while a greater p-value indicates insignificance. The reject column is a binary indicator, where “True” indicates rejection of the null hypothesis, and “False” indicates no significant difference between groups.

CPP-Conjugated Systems Exhibit Superior Binding Affinity to Nonconjugated Analogues

MM-PBSA calculations report significant improvements in binding affinity for single-staple CPP-conjugated systems with respect to both truncated (Figure 5A) and full-length (Figure 5B) CCmut3 variants when compared to unconjugated analogues regardless of the CPP configuration. While binding affinity is significantly different between unconjugated (SHCT/SHCF) and CPP-conjugated truncated CCmut3 systems, no significant difference is observed between the open and cyclic CPP variants (Table 3A). This trend is consistent in the full-length CCmut3 systems with no reported significance between closed and cyclic CPP variants despite significant differences in affinity between CCmut3F and each system (Table 3B). For double-stapled CCmut3 systems, in which only the cyclic CPP configuration was modeled, binding affinities are significantly different compared to unconjugated analogues for both truncated (Figure 6 and Tables S1 and S2) and full-length (Figure 6 and Tables S3 and S4) systems (Table 3C). These findings elucidate the CPP targeting moiety as a pivotal determinant of binding affinity for all pertinent systems and underpin its viability for empirical applications, as it does not appear to impair, but rather ameliorate binding between our construct and the native Bcr-CC target.

Full-Length CCmut3 Systems Exhibit Superior Binding Affinity to Truncated Analogues

Of all single-stapled systems, the SHCFC′ and SHCFC subsets are the most thermodynamically favorable. Moreover, each full-length CCmut3 subset (SHCF/SHCFC′/SHCFC) is characterized by generally lower free energies than their respective truncated CCmut3 subset counterparts (SHCT/SHCTC′/SHCTC). This trend is consistent with MM-PBSA calculations for the double-staple systems, as well, of which DHCFC is the most thermodynamically favorable. These findings reflect an apparent and consistent difference in binding affinity between full-length and truncated CCmut3 systems regardless of the CPP or staple states. Indeed, Tukey’s HSD mean comparison test reveals that all full-length single-staple system subsets are significantly more energetically favorable than even the lead truncated single-staple system subset SHCTC (Table 4). An important exception includes a comparison between SHCF and SHCTC, for which no significant difference is reported. Of all double-stapled systems, the DHCFC subset is the most thermodynamically favorable. Like single-stapled systems, each full-length CCmut3 system subset (DHCF/DHCFC) exhibits superior and statistically significant binding affinity relative to their truncated system subsets (DHCT/DHCTC). Moreover, Tukey’s HSD mean comparison test reveals that each full-length double-stapled system subset is significantly more energetically favorable than even the lead truncated double-staple system subset DHCTC, with exception to the comparison between DHCF and DHCTC, for which no significant difference is reported.

Key Residue Interactions between Stapled CCmut3 Systems and Bcr-CC

While our stapled CCmut3 constructs generally improve the binding affinity for Bcr-CC against unstapled analogues, we also observe several individual constructs that deviate from the behavior of their sister constructs of the same subsystem. Examining closely the SHCF subsystem for example, we identify SHCF-A40-Q47 and SHCF-R44-Q51 as model constructs which exhibit significant energetic disparities, on the order of ∼40 kcal/mol from each other and ∼20 kcal/mol from the unstapled CCmut3F analogue (Figure 5B). To further analyze potential molecular interactions that might influence these disparities, we conducted clustering analysis to capture representative conformations and feature notable salt bridge interactions—allowing us to identify and characterize specific salt bridge configurations that may be altered by the presence of our hydrocarbon staples (Table 5 and Figure 7).

Table 5. Salt Bridge Analysis Reveals Highly Sampled Interactions Glu@OE1/OE2, Asp@OD1/OD2, Lys@NZ, and Arg@NH1/NH2 between Bcr-CC and (A) Unstapled CCmut3F, (B) SHCF-A40-Q47, and (C) SHCF-R44-Q51.

A	CCmut3F		B	SHCF-A40-Q47		C	SHCF-R44-Q51
AA 1	AA 2	%	AA 1	AA 2	%	AA 1	AA 2	%
Glu46	Arg127	76.72	Glu34	Arg129	91.82	Glu36	Arg129	77.41
Glu34	Arg129	63.35	Glu24	Arg129	29.30	Glu46	Arg127	53.85
Glu66	Arg100	23.23	Glu19	Arg118	25.51	Glu32	Lys141	27.19
Glu60	Lys113	22.86	Glu66	Arg100	23.78	Asp104	Arg98	20.61
Glu32	Lys141	21.65	Glu60	Lys113	21.77	Glu60	Lys113	15.53
			Glu32	Lys141	21.18
			Asp104	Arg100	21.04

Figure 7.

Cluster analysis reveals the most highly sampled conformations and interactions for model CCmut3F systems (blue) (A) SHCF-A40-Q47 and (B) SHCF-R44-Q51 in complex with Bcr-CC (gray). Atoms participating in salt bridges are highlighted as follows: Glu@OE1/OE2 (lime), Asp@OD1/OD2 (orange), Lys@NZ (black), and Arg@NH1/NH2 (magenta). Staples are depicted in licorice-cyan along the CCmut3F α2 backbone. Clusters are identified in order of prominence (c0 > c1 > c2) with a population of 64.3% and 25.8% for dominant SHCF-A40-Q47 and SHCF-R44-Q51 clusters, respectively. Secondary and tertiary clusters are indicated accordingly. In all clusters, the A40-Q47 staple placement appears to kink CCmut3F toward Bcr-CC, producing a greater population of salt bridges above the staple site. Staple R44-Q51 conversely rigidifies the peptide backbone and produces fewer salt bridges in this region.

To identify salt bridge interactions, we emphasize key glutamic acid, arginine, aspartic acid, and lysine residues that regularly engage in electrostatic interactions. SHCF-A40-Q47, which is characterized by markedly enhanced binding affinity against unstapled CCmut3F, notably reflects a salt bridge disruption between Glu46 and Arg127 as a consequence of staple placement between Ala40 and Gln47. However, abrogation of this interaction appears to be offset by stabilization of the salt bridge between Glu34 and Arg129, in addition to the emergence of three new interactions between Glu19 and Arg118, Glu24 and Arg129, and Asp104 and Arg100 (Table 5). In contrast, SHCF-R44-Q51, which is characterized by markedly diminished binding affinity against unstapled CCmut3F, notably reflects a massively diminished interaction between Glu46 and Arg127, with the elimination of the Glu34-Arg127 and Glu66-Arg100 interactions altogether. New interactions between Glu36-Arg129 and Asp104-Arg98 are observed, with the first at a high frequency of 77.41% but the latter at only 20.61%. While it is difficult to conclude what effects these alterations may have based on population data alone, it is unsurprising to us that we observe emergent and enhanced interaction in more favorable constructs such as SHCF-R44-Q51.

Structural Insights: Staple Interactions and Influence

While RMSF and helical analyses have provided valuable insights into the selection of optimal stapled peptides, nuanced relationships between specific staple sites and peptide structure have also been illuminated. Across all constructs deviant in RMSF and/or helical space, staple sites Asp30-Arg37, Ala40-Gln47, and Arg43-Asn50 emerge as particularly deleterious selections, with nearly all notable helical losses and RMSF fluctuation shifts in the CCmut3 α2 region associated with these residues and staple sites. Within the span of these sites exist Glu34 and Glu46, previously identified as key residues in the formation of salt bridges between CCmut3 and Bcr-CC. We speculate that these interactions may be disrupted by a variety of molecular mechanisms, including (i) direct obstruction sterically and electronically due to draping of the hydrocarbon staple across or proximal to key contacts in the dimer interface or (ii) indirect obstruction by induction of local and global conformational changes in the peptide, altering the dimer interface altogether. Indeed, salt bridge analysis of construct DHCFC-G29-E36_A40-Q47 indicates diminished interaction between Glu34 and Arg139 and complete abrogation of the interaction between Glu46 and Arg127, in addition to higher fluctuations and loss of helicity around its staple site (Table S5). Moreover, clustering analysis of constructs SHCF-A40-Q47 and SHCF-R44-Q51 highlights the impact different staple sites might have on complex conformation, where the former induces kinks which pack the interface and promote electrostatic interactions while the latter produces a more rigid and distant interface with reduced salt bridges. The propensity for these staples to interact hydrophobically with loops at the termini must also be considered, with RMSF shifts in the termini regularly observed across stapled system types, exacerbated at the N-terminus by CPP conjugation (Figures 3 and 4). We observe such an interaction between a staple and the CPP in construct DHCTC-G29-E36_I57-A64, which is curiously destabilized by the presence of the CPP unlike its sister constructs and unstapled analogue. In all scenarios, the dynamics of the dimer must change to accommodate the hydrocarbon staple, resulting in thermodynamic gains or losses unique to each construct and staple site.

Selection of Ideal CPP-CCmut3-st Candidates

These findings illuminate two important trends among all CCmut3 system subsets: (i) CPP conjugation, whether open or cyclic, enhances binding affinity regardless of staple quantity and (ii) full-length systems consistently exhibit higher binding affinity than truncated systems, even those enhanced by the N-terminal CPP. We are thus able to qualitatively assess the energetic contributions dictated by the α1-helix of CCmut3 and the CPP individually, where inclusion of the α1-helix in full-length CCmut3 confers the greater energetic contribution compared to inclusion of the CPP. This is not unexpected, as α-helix 1 of CCmut3 is known to engage in π-stacking interactions with α2 of Bcr-CC, stabilizing the dimer beyond the central α2–α2 interface. Notably, energetics between equivalent single- and double-stapled subsets vary minimally, with no significant difference observed between means by Tukey’s HSD when evaluating lead SHCTC and DHCTC systems (Table 4). Ideal candidates for synthesis include the cyclic CPP targeting moiety which is crucial for delivery.⁴⁰ As our CPP-conjugated systems have demonstrated not only that the CPP does not generally impair binding or system stability but rather enhances these features, we present lead-stapled CCmut3 candidates selected exclusively from our cyclic CPP-conjugated system subtypes (Table 6). Curiously, we see a high correlation in ideal staple sites between lead SHC and DHC candidates. Of the seven unique sites stapled in our lead DHC candidates, five (Gly29-Glu36, Gln33-Ala40, Arg37-Arg44, Asn50-Ile57, and Ile57-Ala64) also present in both lead SHCTC and SHCFC candidates. Morever, the only set of staple sites to yield leads in both DHCTC and DHCFC systems (Gly29-Glu36 and Asn50-Ile57) also yields leads in both SHCTC and SHCFC systems for both staples when isolated and applied as single staples. This suggests that residue pairs that are productive as single-staple sites are likely to also be productive when included as one of two double-staple sites—specifically highlighting Gly29-Glu36 and Asn50-Ile57 as ideal sites for single and double-stapled constructs.

Table 6. Lead SHC and DHC Constructs for Synthesis as Determined by the RMSF and Binding Energies^a.

^aIdeal candidates for biological studies include the cyclic LS-CPP targeting moiety. Of the four cyclic CPP-conjugated systems, two include CCmut3 truncated and two include CCmut3 full-length. Single- and double-staple systems are presented. Lead candidates are designated by green boxes with an X.

Conclusions

In these extensive all-atom MD simulation studies, we investigated 93 unique stapled CCmut3 analogues comprising 6 system variants to characterize and select lead constructs for synthesis and further investigations. These systems encompass both the full-length and truncated CCmut3 sequences, the presence or absence of cell-penetrating peptides (CPP), and both single- and double-stapled configurations. While we did not anticipate inclusion of the N-terminal CPP to disrupt the binding interface between CCmut3 and Bcr-CC, we were surprised to see that this addition, particularly in the anticipated closed configuration, generally enhanced the binding for our constructs. Moreover, while we anticipated full-length constructs to generally excel both dynamically and in binding, several truncated CCmut3 candidates emerged as viable lead candidates for synthesis, including SHCTC-R43-N50, DHCTC-G29-E36_N50–I57, and DHCTC-Q33-A40_R43-N50, among others. Collectively, our RMSF, energetics, and helicity data suggest that while key system characteristics (full-length vs truncated, open CPP vs closed CPP) are the greatest predictors for construct performance, staple placement must still be considered meticulously and carefully. Indeed, a handful of deviant constructs were observed in both the RMSF space and ΔG, including DHCTC-G29-E36_N50–I57, DHCFC-G29-E36_A40-Q47, and DHCF-G33-A40_R43-N50. These findings are substantiated by marked losses in helicity for DHCFC-G29-E36_A40-Q47 in its second staple site (Ala40-Gln47) and DHCF-G33-A40_R43-N50 around/through both of its staple sites (Gln33-Ala40 and Arg43-Asn50).

Peptide helicity was found to generally correlate with RMSF data for double-stapled systems, although exceptions are noted such as in the case of DHCTC-G29-E36_A40-Q47, which exhibited greater fluctuations around α-helix 1 of Bcr-CC with no marked losses in helicity observed. Loss in secondary structure has been directly correlated with diminished proteolytic resistance due to facile digestion by intracellular proteases, and thus, these findings are imperative to the selection of stapled CCmut3 candidates which mitigate these risks in drug delivery.^{37,42−46,69−71} RMSF analysis unveiled critical insights into the influence of staples on the structural fluctuations within these peptide systems. While staples did not induce significant changes in overall fluctuations, the terminal regions, particularly loops and regions connecting α-helix 1 and α-helix 2 of Bcr-CC, displayed the greatest flexibility. Furthermore, fluctuations in the α1 region of Bcr-CC were observed, vital for π-stacking interactions with α-helix 2 of CCmut3. These fluctuation patterns varied between different systems, highlighting the influence of a cyclic or open CPP on the structural dynamics. We originally hypothesized that deletion of α-helix 1 in CCmut3 might produce a stabilizing effect on our CCmut3: Bcr-CC complexes by reducing interactions between the staple site and terminal residues/α-helix 1 of CCmut3. We observe little distinction between the helicity and RMSF data for our truncated and full-length systems in this way but rather that interactions with the hydrocarbon staples largely manifest between them and α-helix 2 of Bcr-CC.

Binding energetics, as measured within a given system type, were found to vary depending upon staple quantity and placement. While staple sites were rationally selected to mitigate the disruption of critical residue interactions, we observe enhancement, diminishment, and emergence of electrostatic interactions in model system SHCF depending on the location of staple inclusion. Notably, Glu34 is occluded for construct SHCF-R44-Q51, which exhibits a ΔG to the detriment of 17.7 kcal/mol against its unstapled analogue, whereas SHCF-A40-Q47 exhibits a ΔG to the benefit of −21.01 kcal/mol with the emergence of three altogether new electrostatic interactions. While these data alone do not illustrate the itemized energetic contributions of each interaction gained or lost, it does highlight the complex influence these hydrocarbon staples can impart upon the binding interface and local landscape.

Peptide and protein therapeutics present a promising alternative to small-molecule drugs in their ability to inhibit “undruggable” targets in the protein–protein interactome, emerging at the forefront of contemporary research for novel, targeted therapies in a wide array of disease cases where small-molecule drugs have proven inadequate. However, these drugs continue to be encumbered by poor in vivo stability and membrane permeability, and moreover, their design is challenged by their intrinsic conformational, and thus functional, sensitivity to even minor alterations. Using a combinatorial approach with rational mutagenesis and molecular simulations, we constructed a library of lead-stapled peptide candidates designed to reduce conformational variability and improve proteolytic stability. By characterizing the structural dynamics and binding energetics of our 93 unique stapled constructs, we have identified those that are measurably enhanced or impaired by inclusion of hydrocarbon staples. This computational approach allows us to comprehensively consider the conformational sensitivities of our expansive library and select lead candidates for synthesis and further experimental studies in an accessible time frame occluded experimentally.

Here, we present 12 lead single-stapled peptide candidates and 6 lead double-stapled peptide candidates for consideration in synthesis and biological experimentation. While our primary objective in this work is to construct, characterize, and select our next iteration of CCmut3 constructs against Bcr-Abl for the treatment of CML, we also present our platform and findings toward the purpose of expanding literature pertaining to the molecular stapling of larger biological peptides/small proteins. It is our hope that this research will be insightful for the design of stapled peptides in contexts beyond our own, therapeutic, or otherwise.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jpcb.4c02699.

• α-Helicity for double-stapled systems; change in free binding energy for DHCT systems; change in free binding energy for DHCTC systems; change in free binding energy for DHCF systems; change in free binding energy for DHCFC systems; salt bridge occupancy for model double stapled systems; and full minimization and equilibration protocol (PDF)

Author Contributions

^§ M.C.P.L. and B.D.H. equally contributed to this study and are co-first authors.

The authors declare no competing financial interest.