A research team led by Prof. SUN Cheng from the University of Science and Technology of China (USTC) of the Chinese Academy of Sciences, along with collaborators from the Agency for Science, Technology, and Research and the Chinese Academy of Agricultural Sciences, has developed a novel spatial immune-based prediction system for assessing the risk of hepatocellular carcinoma (HCC) recurrence. The study was published in Nature on March 12.

HCC is the third leading cause of cancer-related mortality globally, with postoperative recurrence rates reaching up to 70%.  Precise prediction of HCC recurrence remains challenging due to the complex spatial heterogeneity of the tumor immune microenvironment (TIME) and the dynamic interplay of tumor cells, immune cells, and other TIME components.

In this study, the researchers developed the tumor immune microenvironment spatial (TIMES) score, which quantitatively characterizes the spatial distribution patterns of immune cells within the tumor microenvironment. The TIMES scoring system was developed using the XGBoost machine learning algorithm, trained on a multiplex immunofluorescence dataset from 61 HCC patients. 

The system enables comprehensive assessment of tumor-immune interactions by integrating whole-slide imaging (WSI) with an AI-driven spatial analysis algorithm. Additionally, it enables high-precision recurrence risk prediction based on the spatial expression profiles of five key biomarkers: SPON2, ZFP36L2, ZFP36, VIM, and HLA-DRB1. 

High-dimensional analysis identified SPON2 as the most predictive biomarker, with its expression pattern in natural killer (NK) cell subsets closely correlated with HCC prognosis. Comparative spatial immune profiling demonstrated that non-recurrent HCC patients showed significant enrichment of CD57+ NK cells at the invasive tumor front compared to recurrent patients. This regional immune heterogeneity, not captured by conventional histopathological grading, provides crucial prognostic information. 

The researchers elucidated the molecular mechanism by which SPON2 regulates NK cell function. Three-dimensional migration assays confirmed that SPON2 promotes the directional migration of NK cells toward tumor cells. Cytotoxicity assays demonstrated that SPON2+ NK cells exhibit enhanced cytolytic activity and significantly promote the activation of CD8+ T lymphocytes. In NK cell-specific SPON2-knockout mouse models, the researchers observed decreased IFN-γ secretion and impaired NK cell infiltration, leading to accelerated tumor progression. These findings confirm that SPON2+ NK cells represent a highly activated subset that suppresses HCC recurrence.

When validated in an independent cohort, the TIMES system achieved an accuracy of 82.2% and a specificity of 85.7%, outperforming existing clinical prediction models.

To facilitate clinical application, the researchers developed an open-access online tool that allows clinicians to upload standard immunohistochemistry-stained images and receive comprehensive reports, including TIMES scores and personalized recurrence risk assessments. The core algorithms and computational frameworks underlying the TIMES system have been patented, and the researchers are actively pursuing industry collaboration to standardize protocols and accelerate the system’s clinical application.

This study not only provides a practical predictive tool for clinical decision-making but also advances our understanding of immune mechanisms driving HCC recurrence, thereby laying the foundation for SPON2+ NK cell-targeted immunotherapy strategies.