BGPT: Paper Review: Transcriptional programs of neoantigen-specific TIL in anti-PD-1-treated lung cancers

Explore by Goal

Quick Explanation Copied

Concise verdict: Caushi et al. 2021 present a careful single-cell/TCR-linked dissection of neoantigen (MANA)-specific CD8+ TIL in neoadjuvant anti‑PD‑1 NSCLC showing MANA-TIL are TRM-like, checkpoint-high, IL7R-low and transcriptionally distinct from viral-specific T cells, with signatures that correlate with major pathologic response; the study is high-quality, data-rich, and provides experimentally-validated TCR functional data but is limited by small numbers of validated MANA-TIL per response group and potential selection/conflict risks (see long review)

Long Explanation

Visual summary — core results (data-driven)

Visualize counts and key contrasts reported in the paper (source: Caushi et al. 2021).

Transcriptional phenotype contrasts (simplified)

Key gene/program differences comparing MANA‑ vs influenza‑ vs EBV‑specific CD8 TIL (qualitative; numbers from the paper).

Evidence: MANA-TIL displayed the highest checkpoint/exhaustion composite, lower IL7R/TCF7 and incomplete cytolytic activation; influenza TRM had higher IL7R and stem-like markers; EBV cells were effector-like (see main paper for statistics and tests).

Critical appraisal — strengths

Method integration: MANAFEST/ViraFEST antigen mapping coupled to scRNA‑seq + scTCR‑seq provides direct linkage of antigen specificity to transcriptional state, avoiding many inferences from bulk or phenotype-only studies
Scale and depth: half‑million T cells profiled, refined CD8 analysis (235k) gives power to identify clusters and TRM heterogeneity (15 clusters; 6 TRM subsets).
Functional validation: TCR cloning and ligand dose–response (Jurkat NFAT luciferase) and peptide‑HLA binding/stability assays for key peptides provide mechanistic support for functional avidity claims.
Public data & code: processed single-cell data (GEO GSE176022), raw EGA controlled access, and analysis code available (GitHub), facilitating reproducibility.

Critical appraisal — limitations, biases, and blindspots

Sample & cell numbers for MANA-TIL comparisons: key response-group comparisons of MANA-specific TIL rest on 45 MPR transcriptomes vs 885 non-MPR transcriptomes derived from only 3 responding and 3 non-responding patients respectively, increasing risk of patient-level confounding and overinterpretation of per-cell statistics (paper notes this)
Selection & detection bias from MANAFEST: MANAFEST relies on ex vivo peptide stimulation and expansion, which can preferentially detect higher‑proliferative clonotypes and misses non-expanding or low-frequency MANA‑TIL; some true intratumoral MANA-TIL may be undetected.
Conflicts of interest & funding: multiple authors report industry funding, consulting, equity and patents related to MANAFEST and neoantigen technologies (declared in COI); such ties require careful interpretation and independent replication to exclude subtle biases in candidate selection or interpretation
Generalizability: cohort is resectable NSCLC under neoadjuvant nivolumab — extrapolation to metastatic disease, other tumor types, and different ICB regimens requires validation (see TSTScope and MANAscore follow-ups that extend and test generalizability)
Statistical approach caveat: many comparisons use cell-level statistical tests (Wilcoxon) with multiple-correction; when cells cluster by patient, patient-level bootstrapping/pseudobulk would better control Type I error — authors used pseudobulk in parts but some antigen-specific tests appear cell-level (see Methods).

Where the findings fit in the field (selective literature links)

Demonstrates and extends prior observations that many intratumoral T cells are bystander (Simoni et al.) but provides direct MANA-specific transcriptomes mapped to TCRs
Consistent with later algorithmic work (MANAscore / TSTScope) that extracts tumor‑specific signatures from scRNA/TCR data and shows predictive value for ICB response (external validation efforts underway)

Concrete conclusions, confidence, and falsifiability

Primary supported conclusions: (1) MANA-specific CD8 T cells in neoadjuvant anti‑PD‑1 NSCLC predominantly occupy TRM-like transcriptional states and have higher checkpoint/exhaustion signatures and lower IL7R/TCF7 than influenza-specific TRM, and (2) MANA-TCRs from non-responders have lower ligand-dependent signalling (functional avidity) than responder-derived MANA-TCRs — these are supported by mapped scRNA/TCR data plus TCR functional assays and peptide‑HLA binding experiments
Confidence level: Moderate–high for the descriptive transcriptional findings and per-TCR functional assays (direct measurements); moderate for the causal role of those features in determining clinical response because of limited patient numbers and potential confounders (e.g., STK11/KRAS in one non-responder)
How to falsify: Show, in larger, independent cohorts with paired scRNA/TCR and validated antigen mapping, that (A) MANA-specific TIL do not systematically differ from influenza/viral TIL in checkpoint or IL7R/TCF7 expression, or (B) MANA-specific TCR functional avidity does not associate with pathologic/clinical response.

Practical takeaways & follow-up experiments (testable)

Validate signatures in independent, larger cohorts (scRNA+scTCR) using patient-level pseudobulk statistics and MANAscore/TSTScope classifiers to control for patient-level confounding (computational replication).
Prospective measurement: pair pre-treatment peripheral blood MANAFEST with intratumoral single-cell profiling in a larger neoadjuvant/ metastatic cohort; test whether pre-existing high-avidity MANA-TCRs predict response.
Intervention experiment: in preclinical models (murine tumors with known neoantigens), adoptively transfer T cells transduced with low- vs high-avidity MANA TCRs and measure PD-1 blockade synergy to test causality of TCR avidity in PD-1 response (functional, falsifiable).
IL-7 axis test: short controlled trial or preclinical study delivering IL-7 (or IL-7 superagonists) combined with PD-1 blockade to assess whether IL‑7 can reprogram IL7R-low MANA‑TIL to durable effectors — the paper shows supraphysiological IL‑7 can induce gene programs in MANA-TIL in vitro but functional in vivo data are missing.

Novel hypotheses inspired by the paper

Hypothesis 1: Low IL7R expression on MANA-TIL is an active, antigen-driven adaptation enforced by chronic tumor antigen exposure and local cytokine milieu that reduces homeostatic responsiveness; transient IL-7 dosing during PD‑1 blockade will selectively expand functional high‑avidity MANA clones and improve response.
Hypothesis 2: MANA-specific TRM subsets in non-responders co-express killer inhibitory receptors (KIRs) and checkpoint modules controlled by PRDM1/TOX2; combined blockade of PD-1 plus KIR/TIGIT or epigenetic modulation of TOX2 will resuscitate effector programs and restore cytolytic function in low-avidity MANA-TIL.

How this review can be improved / next steps

Provide patient-level pseudobulk re-analyses, re-run antigen‑specific differential tests with patient as blocking factor, and integrate TSTScope/MANAscore on the raw processed GEO single-cell matrices to produce ROC and survival analyses in independent cohorts.

Data & tools to run next (one-click options)

Author Review: Srinivasan Yegnasubramanian Author Review: Hongkai Ji Author Review: Drew M. Pardoll Author Review: Kellie N. Smith

Key citations (primary)

Run status: review complete

Confidence: 8/10

Feedback:

Updated: March 13, 2026