Fischer Verena, Ignatius Anita, Schmidt-Bleek Katharina, Duda Georg, Haffner-Luntzer Melanie
Institute of Orthopaedic Research and Biomechanics, University Medical Center Ulm, Ulm, Germany.
Julius Wolff Institute, Berlin Institute of Health, Charité, Berlin, Germany.
PLoS One. 2025 Apr 29;20(4):e0322542. doi: 10.1371/journal.pone.0322542. eCollection 2025.
It is well established that the early inflammatory response following fracture is essential for initiating subsequent bone regeneration. An imbalance in inflammation, whether within the innate or adaptive immune response, can result in impaired fracture healing. In our previous studies, we demonstrated that, for example, mice with ovariectomy-induced osteoporosis exhibit altered immune cell populations in the early fracture hematoma and bone marrow, leading to delayed healing. These analyses were conducted using conventional FACS/flow cytometry software, where surface marker expression was assessed using a single threshold based on isotype controls-a binary "yes or no" decision. Recent advances have highlighted that immune cell populations are often more heterogeneous, with distinct phenotypic subgroups depending on their polarization status. This has been particularly well documented for macrophage subpopulations (M1, M2, and intermediate polarization states). In light of this, we employed a commercially available artificial intelligence-based clustering software (Cytolution) to more accurately and objectively identify immune cell subpopulations. We re-analyzed flow cytometry raw data from fracture hematoma and bone marrow of non-osteoporotic and osteoporotic mice at day 1 after fracture. Our findings revealed distinct subclusters for granulocytes (27 subclusters), macrophages (7 subclusters), B cells (4 subclusters), and T cells (6 subclusters) within the fracture hematoma and bone marrow. Comparing osteoporotic and non-osteoporotic mice, we observed an increased abundance of a specific B cell subpopulation in osteoporotic mice, alongside a significant reduction of a particular granulocyte subpopulation in the early fracture hematoma. Several subclusters of granulocytes, T cells, and macrophages were also altered in the bone marrow. The specific role of these immune cell subclusters remains to be investigated in the future. These results suggest that AI-based clustering may provide a powerful tool for identifying immune cell phenotypes during bone regeneration, offering a more nuanced understanding of flow cytometry data.
骨折后的早期炎症反应对于启动随后的骨再生至关重要，这一点已得到充分证实。炎症失衡，无论是先天免疫反应还是适应性免疫反应中的失衡，都可能导致骨折愈合受损。在我们之前的研究中，我们证明，例如，卵巢切除诱导骨质疏松的小鼠在骨折早期血肿和骨髓中表现出免疫细胞群体的改变，导致愈合延迟。这些分析是使用传统的FACS/流式细胞术软件进行的，其中基于同型对照使用单个阈值评估表面标志物表达——一个二元的“是或否”判断。最近的进展突出表明，免疫细胞群体通常更加异质，根据其极化状态有不同的表型亚组。巨噬细胞亚群（M1、M2和中间极化状态）的情况尤其如此。有鉴于此，我们使用了一种市售的基于人工智能的聚类软件（Cytolution）来更准确、客观地识别免疫细胞亚群。我们重新分析了骨折后第1天非骨质疏松和骨质疏松小鼠骨折血肿和骨髓的流式细胞术原始数据。我们的研究结果揭示了骨折血肿和骨髓内粒细胞（27个亚群）、巨噬细胞（7个亚群）、B细胞（4个亚群）和T细胞（6个亚群）的不同亚簇。比较骨质疏松和非骨质疏松小鼠，我们观察到骨质疏松小鼠中特定B细胞亚群的丰度增加，同时早期骨折血肿中特定粒细胞亚群显著减少。骨髓中粒细胞、T细胞和巨噬细胞的几个亚簇也发生了改变。这些免疫细胞亚簇的具体作用仍有待未来研究。这些结果表明，基于人工智能的聚类可能为识别骨再生过程中的免疫细胞表型提供一个强大的工具，从而对流式细胞术数据有更细致入微的理解。
