조경석 (The University of Texas MD Anderson Cancer Center) | 한빛사논문 > 한빛사

한빛사논문

조회294

The University of Texas MD Anderson Cancer Center

Science, 28 May 2026 Vol 392, Issue 6801 | DOI: 10.1126/science.adz2742 Pan-cancer spatial atlas of tertiary lymphoid structures

Kyung Serk Cho¹, Yunhe Liu¹, Guangsheng Pei¹, Jianfeng Chen², Yibo Dai^1,3, Yang Liu¹, Tieling Zhou⁴, Antoine Bougouin⁵, Alejandra Serrano⁴, Khalida Wani⁴, Akshaya Jadhav⁴, Jimin Min⁶, Sharia Hernandez⁴, Wei Lu⁴, Daiwei Zhang^7,8, Jiahui Jiang¹, Diana Shamsutdinova⁴, Enyu Dai¹, Fuduan Peng¹, Ansam Sinjab⁹, Paola A Guerrero⁴, Idania Carolina Lubo Julio⁴, Kai Yu¹, Helen Clark¹⁰, Dipen Maru^4,11, Mingyao Li¹² ,Andrew Futreal¹, Sanghoon Lee¹⁰, Luisa Maren Solis Soto⁴, Lulu Shang¹³, Pavlos Msaouel^2,3, Jaffer A. Ajani1⁴, Hannah Beird¹, Amir A. Jazaeri¹⁰, Alexander J. Lazar^1,3,4,11, Catherine Sautes-Fridman⁵, Wolf H. Fridman⁵, Anirban Maitra^6,15, Humam Kadara^3,4, Jianjun Gao^2*, Padmanee Sharma^2,15,16*, Linghua Wang^{1,3,16,17,18*}

¹Department of Genomic Medicine, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.
²Department of Genitourinary Medical Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.
³The University of Texas MD Anderson Cancer Center UT Health Houston Graduate School of Biomedical Sciences; Houston, TX, USA.
⁴Department of Translational Molecular Pathology, The University of Texas MD Anderson Cancer Center; Houston, TX, USA.
⁵Centre de Recherche des Cordeliers, Sorbonne Université, INSERM, Universite Paris Cite, Equipe labellisée Ligue Contre le Cancer, Paris, France.
⁶Laura and Isaac Perlmutter Cancer Center, Department of Medicine, New York University Grossman School of Medicine, NYU Langone Health, New York, NY, USA.
⁷Department of Biostatistics, The University of North Carolina, Chapel Hill, NC, USA.
⁸Department of Genetics, The University of North Carolina, Chapel Hill, NC, USA.
⁹Therapeutics Discovery Division, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.
¹⁰Department of Gynecologic Oncology and Reproductive Medicine, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.
¹¹Department of Pathology, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.
¹²Department of Biostatistics, Epidemiology and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA.
¹³Department of Biostatistics, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.
¹⁴Department of Gastrointestinal Medical Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.
¹⁵Department of Pathology, New York University Grossman School of Medicine, New York, NY, USA.
¹⁶Department of Immunology, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.
¹⁷James P. Allison Institute, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.
¹⁸Institute for Data Science in Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.
¹⁹Center for Cellular Language Intelligence, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.

^*Corresponding authors: L.W.; P.S.; J.G.

Abstract

INTRODUCTION
Tertiary lymphoid structures (TLSs) are ectopic immune aggregates that arise within tumors and regulate antitumor immunity. Although TLS presence has been associated with enhanced immune activity and improved outcomes in several settings, their maturation states, spatial locations relative to tumors, and context-dependent associations have not been systematically characterized at a pan-cancer scale, limiting a unified view of TLS biology and clinical utility. We asked how TLS maturation and spatial context relate to local immune programs and tumor cell states and whether these features can be quantified from routine histopathology. To address this, we combined spatial transcriptomics (ST) with deep learning applied to hematoxylin and eosin (H&E) whole-slide images (WSIs), building an atlas of TLS diversity across human cancers and a deployable framework that translates TLS spatial features into patient stratification.

RATIONALE
Whole-transcriptome spatial profiling across entire tissue sections enables unbiased identification of TLS niches, maturation trajectories, and spatial context relative to tumor regions [intratumoral (IT), peritumoral (PT), distal-tumoral (DT)], avoiding marker-limited and region of interest–constrained analyses. We uniformly processed 340 ST sections and paired histology, then extended to 3071 WSIs using an artificial intelligence (AI)–enabled framework to detect and classify TLSs at scale. We hypothesized that TLS maturation state and spatial context would be associated with distinct local immune programs and TLS proximity–linked tumor pathway patterns and that a composition-based score (capturing early, primary, or secondary TLS states and their relative proportions) would outperform binary “TLS present or absent” or dominant-state heuristics for prognosis. Cross-cohort analyses, orthogonal validations (including ultrahigh-plex imaging and single-cell spatial profiling), and sensitivity tests were used to assess robustness and generalizability.

RESULTS
Across 12 cancer types, we profiled TLSs including TLSs resolved by ST and identified from WSIs. TLSs varied substantially in prevalence, maturation, and spatial distribution across cancer types. TLSs spanned a maturation continuum [early (E-TLSs), primary follicle–like (P-TLSs), and secondary follicle–like (S-TLSs)] and occupied distinct spatial niches relative to tumors. Maturation was associated with coordinated immune organization: B and T cell zoning and activation, follicular dendritic cell networks, cytokine and chemokine signaling, and interferon (IFN)–related responses. Spatial analyses around intratumoral TLSs (IT-TLSs) revealed distance-dependent tumor pathway gradients: Immune activation and antigen-presentation programs (e.g., IFN-α and IFN-γ responses and MHC class II signatures) were enriched in tumor regions closest to TLSs and attenuated with distance, whereas proliferative and invasive programs (e.g., MYC targets, G2/M cell-cycle programs, epithelial-mesenchymal transition, and related oncogenic pathways) were relatively higher in TLS-distal tumor regions, highlighting spatial coupling between TLS niches and tumor programs. These findings were supported by single-cell–resolved spatial multimodal profiling. To scale beyond ST, we trained an AI framework to detect TLSs and predict maturation states directly from H&E WSIs and evaluated it across additional The Cancer Genome Atlas and independent therapy cohorts. We distilled these spatial features of TLSs into a maturation-aware composite score capturing within-tumor TLS state composition that robustly stratifies patients by survival and treatment response across cancers and treatment contexts.

CONCLUSION
TLSs are spatially organized, heterogeneous immune hubs whose maturation and spatial context are linked to local immune organization, tumor pathway gradients, and patient outcomes. By uniting whole-section ST with AI on standard WSIs, we provide both a pan-cancer atlas of TLS diversity and a practical route to scalable, histology-deployable TLS profiling. The composition-based TLS composite score is positioned for prospective evaluation as a stratifier or endpoint in immuno-oncology trials and may help guide strategies aimed at promoting beneficial TLS states. Future work should incorporate longitudinal sampling under therapy, improved subclass or isotype-resolved antibody profiling as platforms mature, and functional perturbation to test causal mechanisms and refine TLS-informed clinical decision-making.

논문정보