maintaining physiologic affinity, reducing the
predisposition toward off-target cross-reactivity
compared with affinity-matured TCRs.
Although the safety of engineered T cell
therapy will ultimately depend on the degree
of preferential expression of the target tumor
antigen versus healthy tissue, the strategy of
catch bond engineering to maintain physio-
logical affinity yet strong agonist signaling
responses may help to reduce the chance of
unwanted cross-reactivity with other pMHCs
for clinically directed TCRs. Enhancing the ef-
ficacy of clinical TCRs has generally involved
affinity maturation ( 15 , 17 , 41 , 42 ). However,
some affinity-matured TCRs have displayed
off-target toxicity ( 17 , 43 ). The extreme peptide
selectivity of catch bond–engineered TCRs may
even be helpful in mitigating on-target–off-
tumor reactivities—for example, by enhancing
therapeutic indices based on relative expres-
sion levels of unmutated self-tumor antigens,
or neoantigens, with very close sequence sim-
ilarity to a self-antigen expressed in healthy
versus cancerous tissues ( 43 ). Given the relativeease with which we isolated such mutants and
the simplicity of the screen, this lends itself
well to a general approach in the TCR-T clini-
cal development pipeline without requiring
specialized structural insights.Materials and methods
Cell lines
SKW3 T cells (DSMZ) were cultured in RPMI-
1640+GluMax (Thermo Fisher Scientific) com-
plemented with 10% fetal bovine serum (FBS)
(Sigma-Aldrich), 10 mM HEPES, and 50 U/mLZhaoet al.,Science 376 , eabl5282 (2022) 8 April 2022 9 of 14
Fig. 6. Cross-reactivity
screening of MAGE-A3 TCR var-
iants with pMHC libraries.
(A) Design of the single-chain HLA-
A01 yeast-display peptide library.
The DNA peptide library design
shows an NNK codon library for all
positions except anchor positions
P3 (GAK) and P9 (TAY) to
maximize peptides displayed by
HLA-A01. The single-chain trimer
construct is N-terminal to the Myc
tag fused to Aga2 for expression
on yeast. (B) Increasing myc tag
expression on yeast over rounds
of selection represents enrich-
ment of peptide HLA-A*01 and
positive selection of the library.
(C) Heatmap of round 4 selected
peptides showing peptide position
by amino acid accounting for
the number of reads detected
per peptide. Boxed amino acids
represent the MAGE-A3 peptide
EVDPIGHLY. Dark blue represents
a more enriched amino acid in
that position. (D) MAGE-A3
(red dot), TITIN (blue dot), DMSO
(black dot), and 60 predicted
peptides (MAGE-A6, cyan dot;
FAT2, orange dot) were used to
pulse 293T–HLA-A1 cells to stim-
ulate SKW3 T cells expressing
different TCRs for 14 hours.
Anti-CD69-APC staining was
performed and analyzed on flow
cytometry. (E) 293T–HLA-A1 cells
were pulsed with titrated MAGE-
A3, TITIN, MAGE-A6, or FAT2
peptides to stimulate SKW3
T cells expressing MAGE-A3 TCR
variants for 14 hours. Anti-CD69-
APC staining was performed
and analyzed on flow cytometry.
(D) Data are representative of
two independent experiments.
(E) Data are representative
of two independent experiments.
Data are shown as means ± SDs
of technical duplicates.
RESEARCH | RESEARCH ARTICLE