2.1 Examination of Lactylation-Dependent Differentially Expressed Genes (Lac-DEGs) in Acute Myeloid Leukemia (AML) with KMT2A Rearrangement
The study design and key findings are summarized in Fig. 1.. After addressing batch effects in the GEO dataset (Figure S1), our study included 70 samples of KMT2Ar-AML, as well as 52 control specimens. Comparative analysis identified 646 DEGs, with 360 upregulated and 286 downregulated. This observation highlights substantial molecular variances specific to KMT2Ar-AML, as illustrated in Fig. 2A.
Our analysis subsequently focused on DEGs related to lactylation, leading to the identification of 12 crucial lactylation-dependent DEGs, with 7 showing upregulation and 5 showing downregulation. This characterization highlights the significant role of these genes in the lactylation pathways associated with AML, as depicted in Figs. 2B-C. To provide a more comprehensive understanding of the expression patterns of the Lac-DEGs, we utilized heatmaps and histograms to offer a detailed representation of their differential expression between KMT2Ar-AML and CN-AML cohorts, as illustrated in Figs. 2D-E.
Moreover, a functional clustering analysis of DEGs and Lac-DEGs was conducted, revealing insights into their role and association within the lactylation mechanisms pertinent to AML, as evidenced in Figure S2-3.
2.2 Identification and Validation of Diagnostic Markers
To uncover potential diagnostic markers, a multifaceted machine-learning strategy was implemented. Initially, LASSO regression analysis identified 9 significant genes from the statistically relevant univariate variables (Fig. 3A). Following this, Random Forest analysis (Fig. 3B) highlighted the top 10 genes based on their importance. Furthermore, SVM-RFE (Fig. 3C) was employed to refine the identification of significant genes. The intersection of these methodologies culminated in the identification of 6 core genes, designated as LactoKey Genes: PFN1, S100A6, CBR1, LDHB, LGALS1, and PRDX1 (Fig. 3D). We investigated the interrelationships among these specific genes, utilizing red to signify positive correlations and green to indicate negative correlations, as illustrated in Fig. 3E. In order to assess the diagnostic relevance of these genes, a receiver operating characteristic (ROC) curve analysis was performed (Fig. 3F). This analysis highlighted the predictive efficacy of the LactoKey Genes concerning disease manifestation, yielding the following area under the curve (AUC) values: PFN1: AUC = 0.77, S100A6: AUC = 0.855, CBR1: AUC = 0.791, LDHB: AUC = 0.806, LGALS1: AUC = 0.863, and PRDX1: AUC = 0.74. The AUC values suggest that the LactoKey genes hold considerable promise for the diagnosis of KMT2Ar-AML. Notably, LGALS1 and S100A6 demonstrate the most significant diagnostic precision, with AUC values recorded at 0.863 and 0.855, respectively. These findings suggest that the LactoKey genes are not only essential for elucidating the biological mechanisms underlying KMT2Ar-AML but also serve as valuable biomarkers for its diagnosis.
2.3 Immune Infiltration Analysis of LactoKey Genes
The examination of immune cell infiltration indicated a positive association among diverse immune cell populations. Nevertheless, specific subsets, particularly Type 2 T helper cells, exhibited negative correlations with other immune cell types, notably Activated B cells, Type 17 helper cells, CD56dim natural killer cells, Natural killer cells, and Immature dendritic cells (refer to Fig. 4A). In samples from patients with KMT2Ar-acute myeloid leukemia (AML), there was a marked increase in the proportions of several immune cell populations, which included CD56dim natural killer cells, Eosinophils, Gamma delta T cells, Myeloid-derived suppressor cells (MDSCs), Macrophages, Monocytes, Natural killer cells, and Neutrophils, when compared to control AML samples. Notably, MDSCs exhibited elevated levels in both KMT2Ar-AML and control AML, demonstrating distinct immune cell subset profiles (Fig. 4B). Correlation analysis between the LactoKey Genes (CBR1, PFN1, S100A6, LGALS1, LDHB, and PRDX1) and immune cell infiltration indicated that CBR1, PFN1, S100A6, and LGALS1 generally exhibited positive correlations with most immune cells, whereas LDHB and PRDX1 tended to show negative correlations, as illustrated in Fig. 4C-H.
2.4 Biological Functions of LactoKey Genes
Through the execution of correlation analysis and Gene Set Enrichment Analysis (GSEA) of Reactome pathways pertaining to LactoKey genes, facilitated by the clusterProfiler package in R, we uncovered distinct functional relationships for each gene. This underscores their contributions to the pathogenesis of KMT2Ar-AML. The detailed functional associations are as follows: CBR1 is primarily linked to mitochondrial functions, energy metabolism, and oxidative phosphorylation, underscoring its involvement in energy metabolism. LDHB shows strong correlations with glycolysis/gluconeogenesis, pyruvate metabolism, and protein synthesis, highlighting its role in carbohydrate metabolism. PRDX1 exhibits strong associations with ROS detoxification, antioxidant activity, and cell cycle regulation, suggesting its role in cellular protection and regulation. PFN1 is linked to actin cytoskeleton organization, cell migration, and signal transduction, implicating it in cytoskeleton dynamics and signaling. S100A6 demonstrates associations with calcium signaling, inflammatory responses, and the regulation of apoptosis, underscoring its significance in both calcium signaling and inflammatory processes. Conversely, LGALS1 is implicated in the modulation of immune responses, cell adhesion, and the regulation of apoptosis, thereby highlighting its crucial function in immune regulation and cellular attachment. These in-depth characterizations enhance our comprehension of the molecular mechanisms implicated in KMT2Ar-AML and underscore the pivotal contributions of each LactoKey gene to disease advancement and prospective therapeutic interventions. This thorough examination elucidates the manner in which these genes partake in the pathogenesis of KMT2Ar-AML, as depicted in Fig. 5.
2.5 Examination of Gene-miRNA and Gene-TF Interaction Networks
The utilization of network analysis facilitated the construction of interaction networks involving gene-miRNA and gene-transcription factor (TF) associations for the LactoKey Genes, as illustrated in Fig. 5. The gene-miRNA network encompassed a total of 103 interactions, highlighting specific miRNAs such as hsa-miR-361-3p, hsa-miR-548b-5p, hsa-miR-548c-5p, hsa-miR-548d-5p, and hsa-miR-96 in relation to these genes. Furthermore, several common regulators were identified, including CLEC5A, E2F4, MAX, MYC, TP53, USF1, and YY1. A subsequent correlation analysis encompassing all genes was performed, which identified the top 50 genes that exhibited positive correlations with each of the key genes. Notably, LDHB demonstrated the most robust positive correlation with NPM1, as depicted in Fig. 6.
2.6 Lactylation Subtypes in KMT2Ar-AML: Divergent Expression Profiles and Pathway Associations
Our analysis identified two distinct lactylation subtypes in KMT2Ar-AML, referred to as Cluster A and Cluster B. Cluster A is distinguished by increased expression of PFN1 and S100A6, suggesting a unique lactylation profile. In contrast, Cluster B is defined by elevated expression of LDHB, LGALS1, and PRDX1, indicating a contrasting gene expression pattern across clusters, highlighting the heterogeneity within KMT2Ar-AML(Fig. 7A-C).
The biological differences between these lactylation subtypes were investigated through pathway analyses. Cluster A showed significant associations with pathways such as KEGG endocytosis and Notch signaling, while Cluster B demonstrated positive correlations with pathways related to proteasome activity, Parkinson's disease, protein export, folate-mediated one carbon metabolism, and pyruvate metabolism. These findings offer valuable insights into the unique biological processes and disease associations of each subtype (Fig. 7D).
In the Reactome pathway analysis, Cluster A demonstrated significant relationships with the modulation of cell death-related genes via FOXO-mediated transcription, the innate immune response to cytosolic DNA, and interleukin signaling pathways. In contrast, the associations identified in Cluster B suggested the suppression of Notch4 signaling, the destabilization of mRNA mediated by AUF1 HNRNP D0, the stabilization of the p53 protein, the metabolism of polyamines, and the cross-presentation of soluble exogenous antigens within endosomal compartments. These observations offer valuable insights into the molecular mechanisms that distinguish the two lactylation subtypes present in KMT2Ar-AML. (Refer to Fig. 7E).
2.7 Drug Sensitivity Analysis of LactoKey Genes
In our comprehensive examination of drug sensitivity for the six LactoKey genes in the KMT2Ar-AML context, we screened 198 compounds and identified 43 with significant sensitivity profiles. To ensure a targeted and effective analysis, these compounds were chosen due to their well-characterized mechanisms of action and their potential to specifically interact with the identified LactoKey genes. Notably, among the top ten most responsive drugs were PI3K inhibitors, specifically Dactolisib and GNE-317. The top five drugs in terms of sensitivity, ranked by their inhibitory concentration values, were Sepantronium bromide, a CDK9 inhibitor, Luminespib, Dactolisib, and Epirubicin (Fig. 8A-B). KMT2Ar-AML samples with high expression of the six key LactoKey genes (PFN1, S100A6, CBR1, LDHB, LGALS1, PRDX1) exhibited increased sensitivity to Dactolisib, Epirubicin, AZD7762_1022, Trametinib_1372, and Pevonedistat_1529(Fig. 8C-H).