henhaobio
Sign in

ScienceImmunologyPeer-reviewed journal article

De novo design and structure of a peptide-centric TCR mimic binding module

Karsten D. HouseholderiD, Xinyu XiangiD, Kevin M. JudeiD, Arthur DengiD, Matthias ObenausiD, Yang ZhaoiD, Steven C. WilsoniD, Xiaojing Tina CheniD, Nan WangiD, K. Christopher GarcíaiD

PDFDOIPreprint
In-depth full-text digest available · Jump to digest ↓

Abstract

T cell receptor (TCR) mimics offer a promising platform for tumor-specific targeting of peptide-MHC in cancer immunotherapy. Here, we designed a de novo -helical TCR mimic (TCRm) specific for the NY-ESO-1 peptide presented by HLA-A*02, achieving high on-target specificity with nanomolar affinity (Kd = 9.5 nM). The structure of the TCRm/pMHC complex at 2.05 [A] resolution revealed a rigid TCR-like docking mode with an unusual degree of focus on the up-facing NY-ESO-1 side chains, suggesting the potential for reduced off-target reactivity. Indeed, a structure-informed in silico screen of 14,363 HLA-A*02 peptides correctly predicted two off-target peptides, yet our TCRm maintained a wide therapeutic window as a T cell engager. These results represent a path for precision targeting of tumor antigens with peptide-focused -helical TCR mimics.

In-depth summary

Claude Sonnet 4.6 · max effort · 8 Nov 2026 · from the full text of version 2 (14 Apr 2025) · CC BY 4.0

A de novo four-helix bundle designed with RFdiffusion and ProteinMPNN binds NY-ESO-1/HLA-A*02 at Kd 9.5 nM, with a 2.05 Å crystal structure and peptide-focused docking.

Antibody-derived TCR mimics contact the MHC helices extensively and use flexible CDR loops, which makes clean peptide specificity hard to achieve. This work tests rigid, computationally designed α-helical scaffolds as a peptide-focused alternative.

Key results

  1. ShownThe pipeline yielded a nanomolar, peptide-specific binder: few scaffolds passed the computational filter, and the strongest yeast-displayed design, mini-TCRm 1.1, bound NY-ESO-1 but not the MART-1 control.

    4 of 100 scaffolds had a design with iPAE under 10.0; Scaffold #1 had 129/500 passing designs; 2 of 5 yeast-displayed designs stained with NY-ESO-1 but not MART-1 tetramer; mini-TCRm 1.1 Kd 9.5 nM by SPR, none detectable for MART-1. · Design of α-helical “mini”-TCR mimics with…; Experimental validation of de novo TCR mimics…

  2. ShownThe crystal structure shows a diagonal, TCR-like docking mode centered on the up-facing Met4-Trp5 bulge of the peptide, and AlphaFold2 reproduced the overall pose but missed some peptide contacts.

    2.05 Å resolution; buried surface area 1216.7 Å^2 with approximately 31% from peptide contacts; 40-degree docking angle; AlphaFold RMSD 0.410 Å for 128 Cα atoms; three peptide-contact discrepancies. · High-resolution crystal structure reveals…

  3. ShownCompared with an existing NY-ESO-1 Fab and a natural TCR, the mini-TCR mimic is more compact, contacts more MHC residues, and buries a smaller fraction of peptide surface.

    19 MHC residues contacted vs 14 for the Fab and 13 for the TCR; buried surface 1207.7 Å^2 (TCR) and 1055 Å^2 (Fab); peptide fraction buried 36.5% (TCR) and 38.5% (Fab). Kd comparison is in a figure not reproduced in the text. · Comparison to existing antibody TCR mimic and…

  4. SuggestedA structure-conditioned ProteinMPNN score ranked two of seven candidate human-proteome off-target peptides as likely binders, and both stained T2 cells more weakly than NY-ESO-1.

    Hamming search over 14,363 MS-detected HLA-A*02 9-mers gave 5 peptides at distance 4 plus 2 Met4-Trp5 matches from 14,357; off-targets 3 and 5 had the lowest scores; predicted buried surface 1213.6 Å^2 (off-target 5) vs 940.2 Å^2 (off-target 3). · Structure-guided identification of off-target…

  5. ShownAs a Jurkat T cell engager, the binder potently activated cells against NY-ESO-1-pulsed T2 cells and left a dose window with no off-target activation.

    EC50 9.1 nM with nearly 70% activation at 100 nM; off-target 5 about 25% at 100 nM with EC50 174 nM; no activation for off-target 3; zero off-target activity at 10 nM. · Peptide-specific T cell engagers provide a…

Methods, data and reproducibility

RFdiffusion fold-conditioning over NY-ESO-1/HLA-A*02, ProteinMPNN sequence design and AlphaFold2 iPAE ranking, then yeast display, SPR and crystallography. Structure-conditioned ProteinMPNN scoring flagged off-targets, tested in T2 and Jurkat assays.

  • Computationally designed four-helix bundles docked on NY-ESO-1/HLA-A*02, with ProteinMPNN sequences ranked by AlphaFold2 iPAE · n = 100 scaffolds; 600 designs; 2000 designs for 4 hit scaffolds
  • Crystal structure of mini-TCRm 1.1 with NY-ESO-1 HLA-A*02 and AD01 nanobody · human (HLA-A*02 peptide-MHC complex) · n = 1 structure at 2.05 Å · PDB 9MIN
  • MS-detected 9-mer HLA-A*02 peptides from the MHC Motif Atlas, used for the Hamming-distance off-target search · human · n = 14,363 peptides
  • Reference structures: 3M4E5 Fab and 1G4 TCR bound to NY-ESO-1 HLA-A*02 · n = 2 structures · PDB 3HAE; PDB 2BNQ
  • Candidate off-target peptides tested by T2 cell staining and Jurkat engager co-culture · n = 7 candidate peptides; off-targets 3 and 5 in the engager assay
Methods & tools
RFdiffusion, ProteinMPNN, AlphaFold2, Yeast display, Surface plasmon resonance, X-ray crystallography, Size-exclusion chromatography, MHC Motif Atlas, Hamming distance search, Flow cytometry
Reagents & environment
Expression in BL21 E. coli; AD01 anti-β2M nanobody as crystallization chaperone; T2 peptide-pulsing cells; Jurkat NFAT-eGFP reporter; engager built from blinatumomab anti-CD3ε scFv with a 5 amino acid Gly-Ser linker.
Experimental conditions
Yeast staining with 100 nanomolar tetramer; SPR; crystal data to 2.05 Å resolution; T2-Jurkat co-culture at 1:1 ratio with flow cytometry readout next day. No computational hardware reported.
Reproducibility gaps
No code repository stated; raw diffraction images are in the SBGrid databank but no URL is given; other data are in the manuscript or supplement.
Limitations (5)
  • Off-target prediction was only a proof-of-concept; the authors say broader validation is needed to define how accurate the structure-guided approach is and where it fails. · stated by the authors
  • The molecule may need further optimization and off-target screening before it could be considered a clinical candidate. · stated by the authors
  • AlphaFold differed from the crystal structure in three peptide-contact details, which the authors cite as showing that structural validation helps refine the specificity of designed mimics. · stated by the authors
  • Only one target (NY-ESO-1) and one binder (mini-TCRm 1.1) were characterized, and the off-target search used a single MS-derived 9-mer dataset (7 test peptides), limiting generality of the design and screening claims. · noted in this summary
  • Functional readouts use peptide-pulsed T2 cells and a Jurkat reporter line, not tumor cells or primary T cells; off-target 3 stained but gave no activation, so staining and function diverged. · noted in this summary

Read this if you want a structural and computational case for designing peptide-centric helical pMHC binders and screening their off-targets in silico.

Citations
43
FWCI
11.60
Field percentile
top 1%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISMotmaen '25Sinclair '25Visani '25Schuster '26Johansen '25Pang '18
70% semantic matchImmunology
Amir Motmaen, Kevin M. Jude, Nan Wang+9 more·bioRxiv·Nov 19, 2025· 6 citations