Computational design of dynamic receptor-peptide signaling complexes applied to chemotaxis

The study addresses a core challenge in synthetic biology: creating biosensors that detect flexible peptide ligands and couple their recognition to precise intracellular signaling outputs. Traditional protein design methods emphasize static binding to rigid targets, limiting applications where both receptors and ligands are conformationally dynamic, such as peptidergic GPCR systems. The authors aim to model and engineer dynamic receptor–peptide complexes that not only bind with high sensitivity but also trigger robust allosteric signaling. Focusing on the chemokine receptor CXCR4 and its native ligand CXCL12, which regulate key processes like chemotaxis, the work proposes that preserving conformational flexibility at the binding interface while optimizing allosteric transmission can enhance signaling efficacy and enable programmable cell behaviors.

Prior computational protein design successes largely involve optimizing static protein–protein interfaces to create inhibitors and vaccines. However, peptide–receptor systems exhibit high conformational plasticity and induced-fit behavior, complicating structure prediction and design due to limited structural data and vast peptide conformational space. Peptidergic GPCRs have larger binding cavities, broader ligand promiscuity, and often undergo significant rearrangements upon binding. Existing deep-learning and physics-based peptide docking approaches face challenges from scarce peptide-bound structures. Attempts to engineer CXCR4 ligands have predominantly yielded antagonists, potentially stabilizing inactive receptor states. Thus, a gap remains for rationally designing dynamic peptide agonists and receptors that promote active signaling via robust allosteric pathways.

The authors developed a multi-step computational and experimental framework to model and design dynamic GPCR–peptide signaling complexes (CAPSens):

• Modeling (steps i–v): (i) Build active-state hybrid transmembrane scaffolds for CXCR4 using fragments from homologs (inactive CXCR4 PDB 4RWS and active US28:CX3CL1 PDB 4XT1) via Rosetta hybridization; (ii) Thread the N-terminal CXCL12 peptide onto CX3CL1 and perform flexible peptide docking with FlexPepDock while allowing receptor side-chain flexibility; (iii) Diversify bound peptide poses by clustering and binning in a feature space of peptide position, orientation, and shape to capture conformational breadth; (iv) Rebuild extracellular loops near the peptide de novo (Rosetta loop modeling) for each pose; (v) Relax receptor–peptide complexes with side-chain repacking/minimization, cluster models, and select cluster centers for design.
• Design strategies (steps vi–vii): (vi) Conformational selection (Csel): combinatorial interface design on single conformations using RosettaMembrane to stabilize one favored bound state and reduce interface energy; (vii) Conformational dynamism (Cdyn): design mutations compatible with multiple conformations by constructing a computationally guided point-mutant library at first-shell binding/allosteric sites that avoid steric clashes across the ensemble.
• Experimental validation and refinement (steps viii–ix): (viii) Measure intracellular signaling in HEK293T cells using Gαi dissociation BRET and calcium mobilization assays; (ix) Refine the receptor–peptide ensemble by comparing predicted interface energy shifts to experimental potency (EC50) shifts, selecting models best matching data and performing additional docking/refinement.
• Peptide super-agonist design (step x): Introduce peptide mutations at positions P3 and P7 to optimize packing and activation with designed receptors; validate with signaling assays.
• Allosteric analysis (step xi): Run extensive all-atom MD simulations of selected complexes in lipid bilayers (GROMACS, CHARMM36) and apply the AlloDy method to compute mutual information-based allosteric pathways from peptide-binding sites to G-protein interfaces, clustering into pipelines and identifying conserved transmission hubs. Additional methods include expression ELISA, generation of retroviral vectors and transduced human T cells, and Boyden chamber chemotaxis assays to assess migration.

• Design success with native ligand sensing: Conformational selection designs (Csel1/Csel2) enhanced CXCL12 sensing. Csel2 showed ~3.1-fold higher potency in Ca2+ mobilization and ~3.2-fold in Gαi coupling over WT. Success rate: 37% (19 designs tested).
• Dynamism-based designs outperform: Cdyn achieved ~11-fold enhanced Gαi potency and ~20% efficacy increase versus WT; success rate: 33% (15 designs tested). Combining approaches (Csedy) yielded >9-fold potency increase.
• Peptide super-agonists: At P7, Y7L improved selectivity for Csel2 (Gαi efficacy +30%) while reducing WT response by ~18%. At P3, bulky aromatics (e.g., V3Y) strongly activated designed receptors: Cdyn:V3Y showed >80-fold potency increase and ~25% efficacy gain vs WT:WT. Combining P3/P7 mutations with Csedy (Csedy:V3Y-Y7L) delivered >100-fold potency increase and ~34% efficacy gain.
• Chemotaxis in primary human T cells: Engineered receptors (Csel2, Cdyn, Csedy) drove significantly enhanced migration in Boyden chamber assays toward full-length chemokine, with multiple receptor–ligand pairs showing higher migration indices than WT:WT and statistically significant improvements (p-values reported, many ≤ 0.01 or 0.001).
• Conformational diversity and contacts: MD revealed increasing binding-interface dynamism from Csel2:Y7L to Csedy:V3Y-Y7L to WT to Cdyn:V3Y. Csel2 engaged 16 static vs 5 dynamic contacts; Csedy had 12 static/11 dynamic; Cdyn had 11 static/13 dynamic. Receptor pocket cross-sectional area adapted by up to ~52% at ~10.25 Å cavity depth across variants.
• Allosteric signaling pathways: AlloDy identified conserved transmission hubs (F87^2.53, L120^3.36, H203^5.42, W252^6.48, N298^7.49) common to WT and designs. Variants used distinct sets of ‘allosteric triggers’ at the binding pocket to connect to these hubs, indicating substantial rewiring at the interface while maintaining robust core transmission. Greater interface dynamism correlated with stronger signaling efficacy.

The findings demonstrate that explicitly designing for conformational flexibility at the receptor–peptide interface enables superior sensing and signaling compared to stabilizing a single bound conformation. Although different receptor–peptide variants use distinct binding contacts and allosteric triggers, they funnel activation through a conserved set of allosteric transmission hubs in the receptor core, explaining how diverse binding modes can produce potent signaling. This challenges traditional static lock-and-key design strategies and supports a model in which high conformational entropy facilitates mutual induced fit, access to multiple activating sub-states, and robust long-range allosteric coupling. The work extends design methodologies to dynamic peptide–GPCR systems, providing orthogonal receptor–peptide pairs with high potency/efficacy and functional outcomes such as enhanced chemotaxis, which are valuable for synthetic biology and therapeutic applications (e.g., improving immune cell homing).

The study introduces a computational framework (CAPSens) that models and designs dynamic receptor–peptide complexes, achieving ultrasensitive, efficacious GPCR signaling and robust chemotaxis. By preserving conformational dynamism and optimizing allosteric transmission, the designs surpass traditional conformational selection approaches. The framework successfully engineered CXCR4 variants and peptide super-agonists with up to >100-fold potency increases and higher efficacy, validated via cell signaling assays, MD-based allosteric analyses, and primary human T cell migration. These results highlight a generalizable principle: a flexible sensing layer coupled to a robust allosteric transmission network underlies evolvability in peptidergic GPCRs. Future work could develop fully orthogonal, genetically encodable receptor–peptide pairs, expand to other GPCR scaffolds and signaling pathways, integrate neural-network structure predictors, and apply in vivo for programmable cell trafficking in immunotherapy and other biomedical contexts.

• Structural uncertainty: Designs and MD simulations rely on hybrid homology models and designed complexes rather than experimentally determined active-state structures of CXCR4; detailed binding geometries may be affected by model inaccuracies.
• Scope of validation: Functional validation focuses on selected variants, in vitro signaling assays, and ex vivo chemotaxis; comprehensive in vivo performance and long-term stability were not assessed.
• Generality: While principles likely extend to other peptidergic GPCRs, transferability to diverse receptors/peptides remains to be demonstrated.
• Bias toward active-state modeling: MD analyses start from active-state complexes and do not capture inactive-to-active transitions or full activation kinetics.