On July 17, 2025, the research team led by Prof. Zemin Zhang at the Peking University Biomedical Pioneering Innovation Center (BIOPIC), in collaboration with the team in Shenzhen Bay Laboratory, Guangdong Provincial People’s Hospital, Changping Laboratory, and Chongqing Medical University, published an article titled “Integrated Single-Cell and Spatial Transcriptomics Uncover Distinct Cellular Subtypes Involved in Neural Invasion in Pancreatic Cancer” in Cancer Cell. The study sheds light on the cellular and molecular mechanisms driving neural invasion (NI) in pancreatic ductal adenocarcinoma (PDAC), one of the deadliest types of cancer with extremely poor survival outcomes.

 

Figure 1. Research strategy and key findings

 

Neural invasion is a hallmark of PDAC and a major contributor to poor prognosis and diminished quality of life. While clinical evidence shows that removing tumor-associated nerves during surgery can lower the risk of recurrence, the underlying biological basis remained poorly understood.

 

To address this, the researchers profiled 62 tumor samples from 25 PDAC patients using single-cell and single-nucleus RNA sequencing (sc/snRNA-seq) and spatial transcriptomics (ST), combined with pathological analysis, multiplex immunohistochemistry (mIHC), and in vitro experiments. By comparing tumor tissues with and without neural invasion, the researchers identified notable shifts in immune and stromal cell composition. Low-NI tumors were enriched with B cells, plasma cells, and Temra CD8⁺ T cells, suggesting the presence of tertiary lymphoid structures (TLSs). In contrast, high-NI tumors showed higher levels of regulatory T cells (Tregs), NLRP3⁺ inflammatory macrophages, and myofibroblastic cancer-associated fibroblasts (myCAFs).

 

Figure 2. Single-cell and spatially resolved profiling of human PDAC tissues with low/high neural invasion status

 

Aiming to determine whether these immune and stromal cell populations are spatially associated with neural invasion, the researchers integrated pathological morphology with spatial transcriptomic projections of sc/snRNA-seq data to compare the cellular microenvironment surrounding invaded and non-invaded nerves. Strikingly, non-invaded nerves were notably encircled by tertiary lymphoid structures (TLSs), with local fibroblasts exhibiting high CXCL12 expression, potentially mediating the recruitment of CXCR4⁺ B cells and promoting TLS formation. In contrast, invaded nerves lacked TLSs but showed substantial accumulation of NLRP3⁺ macrophages and myofibroblastic CAFs (myCAFs), accompanied by elevated TGFβ signaling.

 

Figure 3. Variation in cellular composition between low-NI and high-NI TME

 

In addition to immune and stromal populations, the presence of cancer cells and nerve-associated cells at the invasion sites was expected. The researchers further dissected the cellular composition within nerve bundles. Beyond conventional myelinating Schwann cells, a novel TGFBI⁺ Schwann cell subpopulation with unique molecular features was identified. These cells uniquely expressed growth factors, extracellular matrix components, and axon guidance cues, implicating the role in promoting tumor growth, directing axonal tropism toward tumor cells, and facilitating neural invasion. Spatially, TGFBI⁺ Schwann cells aligned along the invading edge of the invaded nerves.

 

Figure 4. Characteristics and spatial mapping of Schwann cell subpopulations in PDAC

 

To uncover the upstream signals that induce this Schwann cell phenotype, the researchers performed ligand-receptor analysis and identified TGFβ1, which was primarily secreted by NLRP3⁺ macrophages and myCAFs. In vitro experiments confirmed that TGFβ1 induces Schwann cells to differentiate into the S02_Schwann-TGFBI phenotype, which significantly enhanced the invasiveness and migratory capacity of PDAC cell lines. These findings reveal that NLRP3⁺ macrophages and myCAFs enrich around invaded nerves and secrete TGFβ1, which in turn induces Schwann cell phenotypic reprogramming to promote cancer neural invasion.

 

Figure 5. Induction and functional roles of TGFBI+ Schwann cells

 

To elucidate the relationship between PDAC malignant subpopulations and neural invasion, the researchers integrated sc/snRNA-seq, ST, and morphology to pinpoint two distinct malignant ductal subpopulations with the highest prevalence adjacent to the invaded nerves: D09_Ductal-CEACAM6 and D04_Ductal-GABRP. The former displayed basal-like, poorly differentiated characteristics and high epithelial-mesenchymal transition (EMT) activity, while the latter one highly expressed neurotransmitter receptors and appeared as moderately differentiated, ill-defined glands. Notably, the molecular signatures of these subtypes corresponded with pathological features, providing a cross-dimensional validation. These results lay the molecular groundwork for AI-driven pathology, whereby digital histopathology images may in the future be used to rapidly predict tumor subtypes and guide precise anti-neural invasion strategies.

 

Figure 6. Identification of malignant subpopulations associated with neural invasion

 

This study constructs a high-resolution single-cell spatial atlas of PDAC by integrating spatial transcriptomics, single-cell deconvolution, histological analysis and mIHC, providing a comprehensive characterization of the tumor microenvironment (TME) across varying stages of neural invasion. These pioneering insights will likely inspire future research into ultimately decoding the dynamic interplay among the nervous system, immune compartments, and malignant cells during cancer initiation, progression, and metastasis.

 

Paper link: https://doi.org/10.1016/j.ccell.2025.06.020

 

Academician Zemin Zhang (BIOPIC, Peking University/Chongqing Medical University), Prof. Rufu Chen and Prof. Qingling Zhang (Guangdong Provincial People’s Hospital), and Dr. Min-Min Chen (Shenzhen Bay Laboratory) are co-corresponding authors. Dr. Min-Min Chen, PhD student Qinhang Gao, Dr. Huiheng Ning, research assistant Kang Chen, Dr. Yang Gao, physician Min Yu and Dr. ChaoQun Liu are co-first authors. This research was supported by Major Program of Shenzhen Bay Laboratory, National Key Research and Development Program of China, the National Natural Science Foundation of China, and the Open Program of Shenzhen Bay Laboratory.