AI-based characterization of lymphoid structures and the tumor microenvironment reveals lower mature tertiary lymphoid structures, lymph-node follicles, and plasma cells after neoadjuvant therapy in lung squamous cell carcinoma

Arne Schmidt, Justin Nieder, Cornelius Boehm, Julika Ribbat-Idel, Marius Teodorescu, Marie-Lisa Eich, Simon Schallenberg, Gabriel Dernbach, Katja Lingelbach, Miriam Haegele, Timo Milbich, Adam Narai, Maria Stoelben, Hans Pinckaers, Lukas Ruff, Kai Standvoss, Christian Watermann, Christiane Kuempers, Sven Perner, Verena Sailer, Jasmin Tiebach, Torsten Goldmann, Sebastian Marwitz, Johannes Braegelmann, Frederick Klauschen, Marija Pezer, Rosemarie KrupararXiv 2026 · arXiv

This study leverages AI-driven digital pathology to map the "loco-regional" immune axis in lung cancer—the connection between the tumor and its draining lymph nodes. The researchers discovered that neoadjuvant therapy (chemotherapy and radiation) doesn't just suppress the immune system; it specifically disrupts the maturation of lymphoid structures and decouples the spatial relationship between plasma cells and lymphocytes, a finding that offers a new candidate biomarker for patient survival.

TL;DR

Neoadjuvant therapy in lung squamous cell carcinoma (LUSC) selectively depletes mature tertiary lymphoid structures and disrupts the spatial coordination of the humoral immune response. Using AI-based whole-slide analysis, researchers identified that the spatial co-localization of stromal plasma cells and lymphocytes is a potent prognostic biomarker, outperforming simple cell counts.

What Problem Does This Paper Address?

In the fight against non-small cell lung cancer (NSCLC), Tertiary Lymphoid Structures (TLS)—ectopic aggregates of immune cells within the tumor—act as local "command centers" that coordinate anti-tumor immunity. While we know that mature, germinal center-bearing TLS are associated with better survival and immunotherapy response, we don't fully understand how standard neoadjuvant therapies (chemo/radiation given before surgery) affect these structures.

Most studies look at the tumor in isolation. This paper addresses the gap by examining the entire "loco-regional axis"—the tumor, the proximal lymph nodes, and the distal lymph nodes—as a single linked system. The goal was to determine if treatment-induced immune changes are a "scorched earth" depletion of all immune cells or a more surgical disruption of specific immune architectures.

Methods

The authors employed a sophisticated AI-driven pathology workflow:

  • Aignostics Atlas HETA (H&E Tissue Analysis): A deep-learning pipeline that performs tissue segmentation (identifying stroma, carcinoma, etc.) and classifies individual cells into nine types (e.g., lymphocytes, plasma cells, macrophages) based on nuclear morphometry.
  • TLS/SLS Maturity Grading: A dedicated segmentation model built on a pathology foundation model was trained to detect TLS in tumors and Secondary Lymphoid Structures (SLS) in lymph nodes. Crucially, it graded these as "mature" (containing germinal centers) or "immature."
  • Spatial Phenomics: Beyond simple densities, the team used neighborhood analysis (within a 40 µm radius) to quantify cell-cell co-localization, effectively measuring how "well-connected" different immune cell types were within the tissue.

Datasets

  • Discovery Cohort: 40 LUSC patients (20 neoadjuvant-treated, 20 therapy-naïve) from the UKSH Campus Lübeck and Research Center Borstel. This cohort was unique because it included matched triplets: the primary tumor, a proximal (N1) lymph node, and a distal (N2/3) lymph node for each patient.
  • Validation Cohort: A large, independent tumor microarray (TMA) dataset of 458 LUSC patients (46 treated, 412 naïve) used to test the prognostic value of the AI-derived readouts.

Key Insights and Ideas

  • Selective Maturity Depletion: Interestingly, the total number of TLS didn't change significantly after therapy. However, the number of mature TLS plummeted. This suggests that therapy doesn't prevent immune cells from aggregating, but it does prevent those aggregates from developing into fully functional "command centers."
  • The Humoral Disconnect: In untreated patients, the density of plasma cells in the "ring" around a TLS correlated strongly with the number of lymphocytes inside the tumor. In treated patients, this correlation vanished. The systemic immune response (in the nodes) and the local response (in the tumor) became "decoupled."
  • Myeloid Resilience: While B-cells and plasma cells (the humoral arm) were hit hard by therapy, macrophages and granulocytes remained largely stable. This indicates that the "humoral axis" is particularly sensitive to the cytotoxic effects of neoadjuvant regimens.
  • Spatial Organization as a Biomarker: The study found that where cells are matters as much as how many there are. The spatial co-localization of lymphocytes and plasma cells was a stronger predictor of survival than the density of lymphocytes alone.

Key Results

  • Cell Density: In treated tumors (neoLUSC), stromal plasma cell density was 72% lower (p=0.004) and stromal lymphocyte density was 48% lower (p=0.015) compared to naïve tumors.
  • Lymph Node Impact: Mature SLS area in distal lymph nodes was significantly lower in treated patients (p=0.0033).
  • Prognostic Power: In the validation cohort, high stromal plasma cell density (Hazard Ratio [HR] 0.69, p=0.011) and high lymphocyte–plasma cell co-localization (HR 0.66, p=0.004) were significantly associated with longer overall survival.
  • Treatment Modality: The immune disruption was most pronounced in patients receiving radiation-containing regimens compared to chemotherapy alone.

Limitations and Future Work

  • Sample Timing: The study analyzed tissue resected only weeks after therapy. This captures the "maximal injury" phase but doesn't account for long-term immune reconstitution or "lymphoid neogenesis" that might occur months later.
  • The Immunotherapy Era: The cohorts studied largely predated the current standard of care, which combines neoadjuvant chemotherapy with PD-1/PD-L1 inhibitors. Future work is needed to see if immunotherapy can "rescue" the maturation of TLS that chemotherapy and radiation seem to disrupt.
  • Cohort Size: The discovery cohort was relatively small (n=40), though the findings were robustly confirmed in the much larger validation set.

By providing a quantitative, AI-based map of the immune landscape, this research highlights the stromal plasma cell/co-localization axis as a critical candidate biomarker for evaluating treatment response and predicting long-term outcomes in lung cancer.

Summary updated

Abstract

Mature, germinal centre-bearing tertiary lymphoid structures (TLS) and mature lymph node follicles (secondary lymphoid structures, SLS) sustain the plasma cell response that shapes outcome and immunotherapy benefit in lung cancer. How this loco-regional axis differs between neoadjuvant-treated and therapy-naive tumors is poorly defined. We analyzed H&E whole-slide images of resected tumors and paired proximal and distal lymph nodes from matched neoadjuvant-treated (neoLUSC, n=20) and therapy-naive (naivLUSC, n=20) lung squamous cell carcinoma (LUSC) patients. Aignostics Atlas H&E Tissue Analysis (HETA) models quantified cell densities per tissue compartment and cell-cell co-localization; a dedicated deep-learning model validated against pathologist annotations detected TLS and SLS and graded their maturity. Prognostic value was tested in an independent cohort (n=458; 46 treated). In neoLUSC tumors, stromal plasma cell density was 72% (p=0.004) and stromal lymphocyte density 48% (p=0.015) lower than in naivLUSC; macrophages and granulocytes did not differ significantly. Total and immature TLS did not differ, whereas mature TLS in the tumor (p=0.034) and mature SLS in the tumor-distal lymph node (p=0.003) were lower in neoLUSC patients. Peri-TLS plasma cell density correlated with tumor lymphocyte density in naivLUSC (rho=0.54, p=0.017) but not in neoLUSC tumors (rho=0.07). In the validation cohort, stromal plasma cell density (HR 0.69, p=0.011) and lymphocyte-plasma cell co-localization (HR 0.66, p=0.004) were associated with longer survival, and neoadjuvant status with shorter survival (HR 1.50, p=0.026). Neoadjuvant therapy is thus associated with fewer mature lymphoid structures and weaker coupling of the humoral compartment to the tumor rather than uniformly lower immune readouts. The stromal plasma cell/co-localization axis is a candidate prognostic biomarker.
BibTeX
@misc{schmidt2026ai,
  author = {Arne Schmidt and Justin Nieder and Cornelius Boehm and Julika Ribbat-Idel and Marius Teodorescu and Marie-Lisa Eich and Simon Schallenberg and Gabriel Dernbach and Katja Lingelbach and Miriam Haegele and Timo Milbich and Adam Narai and Maria Stoelben and Hans Pinckaers and Lukas Ruff and Kai Standvoss and Christian Watermann and Christiane Kuempers and Sven Perner and Verena Sailer and Jasmin Tiebach and Torsten Goldmann and Sebastian Marwitz and Johannes Braegelmann and Frederick Klauschen and Marija Pezer and Rosemarie Krupar},
  title = {{AI-based characterization of lymphoid structures and the tumor microenvironment reveals lower mature tertiary lymphoid structures, lymph-node follicles, and plasma cells after neoadjuvant therapy in lung squamous cell carcinoma}},
  year = {2026},
  eprint = {2611.12855},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.12855},
}
Figure