Background The bone marrow (BM) microenvironment plays a critical role in acute myeloid leukemia (AML) by shaping immune surveillance and supporting leukemic cells. In pediatric AML, this immunoregulatory landscape remains poorly defined. We aimed to profile soluble immune modulators in diagnostic BM plasma from pediatric AML patients, identify key survival-related factors, and build a robust immune-based risk model.

Methods BM plasma from 77 children with newly diagnosed AML and 15 age-matched healthy donors (HSCT candidates) was analyzed using Olink proteomics to quantify 180 soluble immune proteins. Survival associations (OS, EFS, RFS) were assessed via LASSO and Elastic Net regression models in R. A multivariable risk score was built and evaluated. Bulk RNA-seq and ssGSEA, HALLMARK, and xCell analyses were conducted to explore underlying mechanisms and immune cell infiltration.

Results Compared to controls, AML patients had lower levels of SCF, Flt3L, and IL12B (p<0.001) and higher levels of LAG3, LILRB4, PDL1, CD244, ADA, uPA, and HGF (p<0.05). GO analysis revealed downregulation of T cell receptor signaling, immune effector functions, and inflammatory pathways.

LASSO and Elastic Net models identified 61, 69, and 59 immune proteins associated with OS, EFS, and RFS (p<0.05), with 11 shared. An 11-protein model for EFS showed strong predictive value (log-rank p<0.0001; C-index 0.84). The risk score stratified patients into high-risk (n=38) and low-risk (n=39) groups, with 5-year EFS of 38% vs. 92% (p<0.0001). Integration with NCCN cytogenetics improved stratification by reassigning subsets from intermediate risk into more accurate risk categories.

Clinically, high-risk patients were older (8.91 vs. 6.81 yrs, p=0.0354) and had lower platelet counts (61.05 vs. 87.13×10⁹/L, p=0.0169). Expression analyses showed reduced TNF receptor binding, innate immune response, pattern recognition, and type II interferon signaling in high-risk patients. CXCL1/5/6/11/12 and TNFSF11/12 were significantly downregulated (p<0.05).

Transcriptomic data showed suppression of oxidative phosphorylation, DNA repair, glycolysis, and p53 pathways in high-risk patients. xCell analysis revealed more CD4+ T cells, memory B cells, and CD8+ naïve T cells in the high-risk group, while CD8+ effector memory T cells were enriched in the low-risk group.

Conclusion At diagnosis, pediatric AML BM exhibits an immunosuppressive profile with elevated inhibitory mediators and diminished immune activation. An 11-protein immune signature robustly predicts 5-year EFS and refines traditional risk classification. This model may guide precision prognostication and inform immune-targeted interventions in pediatric AML.

This content is only available as a PDF.
2025
Sign in via your Institution