Datasets on RNA Sequencing (RNA-seq) and mRNA expression profiles of 510 samples were obtained from TCGA-GTEx on 10 March 2021, which included the information on 173 AML tumorous tissue samples and 337 normal blood samples. Among the AML tumorous tissue samples, the survival time of 12 samples was 0, and 12 samples had no survival information, which have been all removed. Finally, a total of 149 AML samples were included in this study. Also, genes with an expression value of 0, which occupied more than 60% of all the samples, were excluded. RGene expression data and phenotype data of 337 normal whole blood samples were downloaded from the GTEx (https://www.gtexportal.org/home/) database. The abovementioned databases were obtained from UCSC XENA (https://xenabrowser.net/) (Figure 1A).

The mRNA expression levels between tumor and normal samples were analyzed by using the DESeq2 package in R software. First, genes with an expression value of 0 in more than half of the samples were excluded. Differentially expressed genes were screened with the cutoff value of adjusted p-value < 0.05 and | log2-fold change [FC] | > 2.

To figure out the prognostic genes of the AML samples, the related hazard ratios (HRs), 95% confidence intervals (CIs) of the HRs, and p-values were analyzed using univariate and multivariate Cox regression. Least absolute shrinkage and selection operator (LASSO) Cox regression analysis, which could reduce the dimensionality and select the most robust markers to predict prognosis, was used after the univariate Cox analysis to further identify the candidate genes that could be used to construct the multivariate Cox regression model. Survival analysis was conducted using the R Bioconductor. The survival package was used for univariate and multivariate Cox regression analyses, while the glmnet package was used for the LASSO regression analysis.

Genes in each module were subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes Genomes (KEGG) pathway analysis to understand their biological function better. The GO enrichment analysis and KEGG signal pathway analysis of differential expression genes (DEGs) were carried out by clusterProfiler in R software, and the results were visualized. For this study, we analyzed the gene set named h.all.v6.2 symbols.gmt with the cutoff value of normal p < 0.05.

The Gene Set Enrichment Analysis (GSEA) software was used for single-gene enrichment analysis aiming at LYPD3. Based on LASSO regression and Cox regression, survival analysis was performed to screen out the prognostic genes. Finally, the receiver operating characteristic (ROC) curve was plotted, and the 1-, 3, and 5-year survival rates were calculated.

HL-60 cells (non-adherent human acute myelogenous leukemia cells), A549 cells (adherent human lung cancer cells), HCT 116 cells (adherent human colon cancer cells), EC cells (human umbilical vein endothelial cells), and CEM cells (non-adherent human lymphoma cells) were all obtained from the ScienCell (San Diego, California, United States). They were all incubated with RPMI Medium Modified medium (Cytiva, Dana Hector, United States) containing 10% fetal bovine serum (Gibco, Thermo Fisher Scientific, Waltham, MA, United States). The cells were cultured using a previously described method (Chen et al., 2018).

We obtained LYPD3 siRNA and scrambled negative control siRNA for transfection from Shanghai GenePharma (China). The siRNA sequences were LYPD3 siRNA (sense: 5′- GCU​GUA​ACU​CUG​ACC​UCC​GCA​ACA​A-3′; antisense: 5′ UUG​UUG​CGG​AGG​UCA​GAG​UUA​CA GC-3′) and scrambled negative control siRNA (sense, 5′-UUC​UCC​GAA​CGU​GUC​ACG​UTT-3′; antisense, 5′-ACG​UGA​CAC​GUU​CGG​AGA​ATT-3′). HL-60 cells (0.4×10⁶ cells/well) were seeded in six-well plates and grown to 30% confluence. The HL-60 cells were divided into three groups: a control group, a blank group (transfected with negative control siRNA), and an Si-LYPD3 group (transfected with LYPD3 siRNA). Lip2000 (Invitrogen, China) was used to improve the transfection efficiency. The transfection mixture was replaced after 24 h with 1640 with 10% fetal bovine serum (FBS). Then, the cells were incubated for another 24 h and subjected to Western blot analysis. Also, LYPD3 knockdown was confirmed by Western blot.

The effect of LYPD3 on HL-60 proliferation was assayed using the Cell Counting Kit-8 (CCK8, Beyotime, China). In summary, for the CCK8 assay, cultured HL-60 were suspended in a culture medium with 0.1% FBS and inoculated in a 96-well plate (1 × 10⁴cells/well) along with 0, 5 μM siRNA(Shanghai GenePharma, China). After 0, 24, 48, and 72 h incubation, 10 μL of CCK8 solution was added to each well. The plate was incubated for an additional 1.5 h before measuring the absorbance at a 450 nm wavelength using a microplate reader (Thermo MK3, United States).

For the analysis of HO-1, Nrf2, Cas-3, Cas-1, PARP-1, akt-1, P-akt, P53, and β-actin at the protein level, HL-60 cells were seeded in six-well plates with a density of 7.35 × 10⁵cells per well. After 24 h, the medium was changed. Each sample was homogenized in 3 mL of lysis buffer [50 mm Tris (pH 8.0), 150 mm NaCl, and 0.5% NP40] with protease inhibitors (1 mm phenylmethylsulfonyl fluoride, 10 μg/ml aprotinin, and 10 μg/ml leupeptin), followed by incubation on ice for 30 min. After centrifugation at 15,000g for 10 min at 4°C, the protein content of the cell lysates was determined by the Bio-Rad protein assay (Hercules, CA). Twenty-five micrograms of protein per lane (unless otherwise noted) was resolved by 7.5% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to the nitrocellulose membrane. The membranes were blocked with 1% non-fat dried milk in 50 mM Tris (pH 7.5) with 150 mM NaCl and 0.05% Tween-20 and sequentially incubated with antibodies against LYPD3 (EPR9107, #ab1517-9, ABCAM, MA, United States), β-actin (SANYING, Wuhan, China), Cas-1(PROTEINTECH, United States), Cas-3(PROTEINTECH, United States), PARP-1(PROTEINTECH, United States), P-akt(PROTEINTECH, United States), akt-1(PROTEINTECH, United States), P53(PROTEINTECH, United States), and the appropriate horseradish peroxidase-conjugated, secondary anti-mouse (for ERα and β-actin), or anti-rabbit (for ERβ) antibodies (Amersham Biosciences, Piscataway, NJ). Blots were visualized using enhanced chemiluminescence (ECL Plus) reagents as recommended by the manufacturer (Amersham Biosciences). Densitometric analysis of band intensities was conducted using optical scanning and qualification with ImageJ. After the analysis was completed, protein expression was normalized to β-actin and compared to the corresponding vehicle controls.

All experiment data are expressed as mean ± standard deviation (SD). Graph Prism 8 software was used for statistical analysis. Mean values of the experimental groups were compared by the t-test and Chi-square test, and p-value < 0.05 was accepted as statistically significant. Western blot results were analyzed with ImageJ software. Experiments were repeated in triplicate, with similar results each time, and the figures show representative experimental results.

The DEGs of 149 AML samples obtained from the TCGA database and 337 normal whole blood controls from the GTEx database were identified. To sum up, a total of 11,490 DEGs were identified. Among them, 4164 genes were upregulated and 7756 genes were downregulated (Figure 1B).

The univariate Cox regression analysis and LASSO regression analysis showed that 28 genes were associated with OS (Figures 1C,D). In order to identify the prognosis genes of AML, the multivariate Cox regression model was performed. Also, multivariate Cox analysis further narrowed these candidates into 10 genes, including AC108479.3, FAM207A, GS1-304P7.1, LINC00540, LYPD3, NDST3, RP11-379P1.4, RP11-615I2.1, RP11-672L10.6, and TREML2, which were significantly correlated with OS in AML; AC108479.3, NDST3, RP11-379P1.4, and RP11-672L10.6 benefited AML, while FAM207A, GS1-304P7.1, LINC00540, LYPD3, RP11-615I2.1, and TREML2 had a poor impact on AML (Supplementary Table S1). Surprisingly, LYPD3 stands out from these 10 genes and shows a potentially negative impact on AML (p <0.05).

To validate the expression of the LYPD3 protein with the LYPD3 GPI Ab in cancer lines (see Figure 2A), a Western blot analysis was performed. This antibody gave prominent bands around 76 kDa in the A549, hct 116, and HL-60 cells, while no bands were shown in EC and CEM cells, and the results of these are consistent with the study reported by Ping Hu et al. in 2020, the research reported by Willuda J et al. in 2017, and the study reported by Wang L in 2017 (Wang et al., 2017; Willuda et al., 2017; Hu et al., 2020). As shown in Figure 2A, the color of HL-60 cell bands was darker than that of the other two bands (that is, HCT116 cell bands and A549 cell bands), indicating that the expression of LYPD3 was the highest in the HL-60 cells. It is well documented that increased LYPD3 expression is related to lung adenocarcinoma and colon cancer, which is the same as our experimental results mentioned above. In general, the results are expected to provide a research basis for the screening of markers related to the treatment and prognosis of AML.

To investigate the function of LYPD3, proliferation and apoptosis assays were carried out in HL-60 transfected with siRNA-LYPD3. As shown in Figure 2C, LYPD3 expression was significantly suppressed in HL-60 cells by siRNA-LYPD3. It is significant because it has to do with one of the mechanisms of leukemia, abnormal cell proliferation, which might provide a molecular basis for future clinical and basic research. After that, one is that the proliferation of HL-60 cells was evaluated more distinctly compared with the HL-60 cells by siRNA-LYPD3; that is to say, LYPD3 gene knockdown mediated by siRNA-LYPD3 suppressed proliferation. Furthermore, the expression of apoptotic markers including CAS-1, CAS-3, and PARP-1 was significantly upregulated, indicating that obvious inducing of cell apoptosis occurred in HL-60 cells (Figure 2B). This result demonstrated that knocking down LYPD3 induced multiple cell death modes including apoptosis, pyroptosis, and parthanatos, which suppressed the growth of HL-60 cells.

According to the above results, we have definitely confirmed that LYPD3 should play an important role in AML. Thus, we further explored the possible related pathways through which LYPD3 affected the pathogenesis and prognosis of AML. As shown in Figure 2D, the significantly enriched pathways in AML samples with high LYPD3 expression were analyzed by GSEA. We identified 25 pathways that were significantly upregulated in AML samples with high LYPD3 expression (Figure 2D, normal p < 0.05), including the E2F signaling pathway, the p53 signaling pathway, the PI3K_AKT signaling pathway, and so forth. Among them, the two most significantly enriched pathways were the P53 signaling pathway and the PI3K_AKT signaling pathway (Figures 2E,F). In conclusion, we thought that it was doubtful that LYPD3 might affect the clinical features of AML by regulating the P53 and/or PI3K_AKT signaling pathway. To further explore our conjecture, Western blot was performed (Figure 2G). As shown, the expression of LYPD3 gene knockdown mediated by the siRNA group was obviously increased in p53 and PI3K_AKT signaling, which was consistent with our predictions. However, their specific mechanisms and the upstream and downstream molecules that regulate them need to be further studied.