Background. The classical mammalian senescence marker CDKN1A (p21) was recently reported absent from SA-betaGal-high senescent CD8 effector memory cells (Turano et al., bioRxiv Dec 2025), challenging the standard panel design for blood-based CD8 immunosenescence biomarkers. Here, at orders-of-magnitude larger single-cell scale (1.82 million CD8 T cells across large public blood immune datasets, 2,004 donors total spanning healthy aging and disease-context cohorts), we replicate that finding and identify what panel architecture remains predictive after p21 exclusion. Main result. Across 35+ independent validation strata - cross-cohort leave-one-cohort-out (LODO), six AIDA ethnicities, sex stratification, CD4 / monocyte / NK compartments, 10,000-shuffle permutation null, five external Census cohorts, and the canonical Sade-Feldman melanoma anti-PD-1 cohort (GSE120575) - removing CDKN1A from the previously proposed 5-gene panel improved or matched cross-cohort age correlation in 26 of 29 strata (median delta-rho = +0.034). Architecture ablation identified LGALS1 as the dominant single transcriptional marker of CD8 immunosenescence (patient-level mean per-cell-score AUC = 0.900; toxic-fraction-score AUC = 0.822; label-permutation p = 0.008; n = 19 melanoma patients). CDKN2A and LMNB1 (down) provide a senescence-context scaffold. CD47 mRNA is dispensable (bootstrap AUC 95% CI -0.091, +0.067); CD47 protein is a biologically supported candidate readout (Hayflick proteomics +0.92 sen-specific, AUF1 PAR-CLIP eCLIP at the CD47 3'UTR) but currently lacks direct living-donor CD8 flow validation, which we identify as the critical wet-lab next step before clinical assay deployment. Cell-context specificity and mechanism. Whole-tumor bulk RNA-seq scoring (Riaz GSE91061) reversed the direction of association (AUC = 0.378), confirming that the panel reflects CD8-intrinsic rather than generic tumor biology. Mechanism analysis in 161,051 healthy-donor effector memory CD8 cells identified a coordinated NRF2 / HIF-1alpha / ARNT / AHR / TOX transcription factor program enriched in T0-high cells. Translation and limits. We propose an LGALS1-centered CD8 immunosenescence panel (CDKN2A + LGALS1 up / LMNB1 down at the mRNA level, plus CD47 protein for multi-omic clinical assays). The melanoma anti-PD-1 result is hypothesis-strengthening, not definitive clinical validation; replication on independent ICB single-cell cohorts is required before clinical translation. We do not claim a universal aging clock, an optimal three-gene combination (15.7% of random three-gene panels match the T0 patient-level AUC), or superiority over published senescence signatures: the 125-gene SenMayo signature (Saul et al. 2022) reached toxic-fraction AUC = 0.822 in our pipeline, equivalent to LGALS1 alone. Code and data. Code: https://github.com/webmaxstudio/senex-report. Datasets: CELLxGENE Census 2025-11-08; GSE120575 (Sade-Feldman); GSE115978 (Jerby-Arnon); GSE91061 (Riaz). Provisional patent USPTO 64/047,904 (filed 23 Apr 2026, M. Izralevich sole inventor) covers the immune-evasion-centered architecture.
Maksim Izralevich (Thu,) studied this question.