BMSD : 10.65048
BMSD : 10.65048
https://doi.org/10.65048/ijbmr
Search
About Us
Authors
Reviewers
Editorial Board
Contact Us

International Journal of Biological and Medical Research (IJBMR)

Original Article
Year : 2026  |  Volume : 17  |  Issue : 2  |  Page : 8431-8437 Full Version
Integrated gene expression and network analysis identifies molecular differences between flt3-positive and flt3-negative aml
Int J Biol Med Res. 2026; 17(2): 8431-8437  |  DOI: 10.65048/ijbmr.2026.17.2.8431.1019
Yagmur Kiraz
Department of Genetics and Bioengineering, Faculty of Engineering, Izmir University of Economics, 35330, ?zmir, Türkiye
Correspondence Address :
Yagmur Kiraz PhD,
Department of Genetics and Bioengineering
Faculty of Engineering, Izmir University of Economics,
35330, Izmir, Türkiye. E-mail: yagmur.kiraz@ieu.edu.tr
Received: December 30, 2025 Accepted: February 2, 2026 Published: April 5, 2026

Abstract
FLT3 mutations, particularly internal tandem duplications (FLT3-ITD), occur in a substantial proportion of acute myeloid leukemia (AML) cases and are associated with aggressive disease biology and poor clinical outcomes. Although the introduction of FLT3 inhibitors has improved response rates, relapse remains frequent, highlighting the need to identify additional molecular targets that cooperate with FLT3 signaling. In this study, we performed an integrative transcriptomic and network-based analysis comparing FLT3-mutant (FLT3+) and FLT3-wildtype (FLT3-) AML to uncover reproducible molecular alterations and functionally relevant gene networks. Differential gene expression analysis was conducted using two independent public AML datasets (GSE10358 and GSE14468). Shared differentially expressed genes (DEGs) were identified and subjected to protein–protein interaction and pathway enrichment analyses to define coordinated functional modules. This approach revealed a robust set of common DEGs (n = 29) consistently dysregulated across cohorts, with enrichment of genes involved in cell adhesion, migration, and cytoskeletal organization. Network analysis highlighted a highly interconnected module related to cytoskeletal dynamics and cell–microenvironment interactions, suggesting altered niche engagement in FLT3-mutant AML. From these analyses, four network-prioritized genes (EPAS1, IQGAP1, PTPN4, and CTHRC1) were selected for experimental validation. Quantitative real-time PCR confirmed significantly increased expression of all four genes in FLT3+ AML compared to FLT3- controls. Furthermore, survival analysis identified EPAS1, encoding the hypoxia-inducible factor HIF-2?, as prognostically relevant, with higher expression associated with shorter overall survival. Together, these findings demonstrate that FLT3-mutant AML is characterized by coordinated dysregulation of cytoskeletal, adhesion, and microenvironment-adaptive pathways. The identified gene network provides mechanistic insight into FLT3-driven disease biology and highlights candidate targets that may complement existing FLT3-directed therapies to improve patient outcomes.
Keywords
Acute myeloid leukemia (AML), FLT3-ITD, Differential gene expression, Network analysis, qRT-PCR validation

Introduction

Acute myeloid leukemia (AML) is a heterogeneous clonal malignancy of hematopoietic precursors characterized by various genetic mutations. One of the most frequent and clinically significant mutations in AML involves the FMS-like tyrosine kinase 3 (FLT3) gene. FLT3 internal tandem duplication (FLT3-ITD) mutations occur in roughly 15–35% of AML cases, with an additional 5–10% harboring FLT3 tyrosine kinase domain point mutations [1]. These activating mutations result in ligand-independent dimerization and constitutive signaling through FLT3, leading to aberrant activation of multiple downstream pathways including PI3K/AKT, MAPK/ERK, and STAT5 [1]. The clinical consequences are aggressive disease features: FLT3-mutant AML is associated with high leukocyte counts, elevated bone marrow blast percentages, and an increased risk of relapse and treatment failure [2,3]. Indeed, FLT3 mutations are recognized as high-risk factors in AML conferring poorer overall survival outcomes, especially in cases with high mutant allelic burden [2,4,5].

The advent of targeted therapies has partly improved the outlook for FLT3-mutant AML. FLT3 inhibitors (such as midostaurin and gilteritinib) incorporated into therapy have demonstrated improved remission rates and survival in both newly diagnosed and relapsed FLT3+ AML [6]. However, responses to FLT3-directed therapy are not universally durable. Many patients eventually relapse due to emerging drug resistance mechanisms, which include secondary kinase mutations that abrogate inhibitor binding, activation of alternative signaling pathways, and protective effects from the bone marrow microenvironment. These challenges highlight that FLT3 mutation alone is not an isolated driver – it operates within a broader oncogenic network. The leukemogenic mechanism of mutant FLT3 and its cooperation with other molecular pathways remain areas of active investigation. There is a pressing clinical need to identify co-occurring dysregulated genes and pathways in FLT3-mutant AML that could serve as novel therapeutic targets or biomarkers beyond FLT3 itself. By discovering the downstream or parallel pathways that FLT3-ITD leukemic cells depend on, we may devise combination therapies to prevent or overcome resistance and improve patient outcomes [7,8].

Prior studies have suggested that different AML patients exhibit a distinct gene expression profile [9]. For example, computational analyses of large cohorts have identified FLT3-associated expression signatures and hinted at enrichment of genes related to stemness, quiescence, and oncogenic signaling in FLT3-ITD cases. Nonetheless, relatively few studies have integrated multiple datasets to find a consensus set of differentially expressed genes (DEGs) that are consistently dysregulated by FLT3 mutations. Integrative bioinformatic approaches – including cross-cohort DEG overlap, network modeling, and pathway enrichment – can help filter out random cohort-specific findings and pinpoint robust, functionally significant molecular changes associated with FLT3+ AML. Such changes may unveil FLT3-mutant leukemia cells that could be exploited therapeutically [10–13].

In this study, we performed a comprehensive transcriptomic analysis focusing on FLT3+ vs FLT3- AML. Using two independent patient cohorts with gene expression data, we identified DEGs associated with FLT3 mutations in each dataset and then determined the overlap of dysregulated genes common to both. We next applied protein–protein interaction and pathway enrichment network analyses to these shared DEGs to discover interconnected hubs and pathways that are consistently altered in FLT3-mutant AML. This systems biology approach allowed us to prioritize a handful of key genes within those networks as candidate therapeutic targets. We then validated the differential expression of four such candidate genes by quantitative PCR (qPCR) in an independent sample set. Finally, we assessed the prognostic relevance of one top candidate gene by correlating its expression with patient survival. By combining in silico discovery and experimental validation, our work sheds light on critical biological pathways in FLT3-driven leukemogenesis and proposes novel targets (e.g. EPAS1/HIF-2α and IQGAP1) that could complement FLT3-directed therapy in AML.

Materials and Method

2.1. Data Sources and Differential Expression Analysis

We analyzed two independent public gene expression datasets of AML patients, obtained from the NCBI Gene Expression Omnibus (GEO) [14]. Dataset 1 (GSE10358) [15,16] comprises expression profiling of 304 AML samples (Affymetrix Human Genome U133 Plus 2.0 arrays), and Dataset 2 (GSE14468) [17] includes 526 AML samples (Affymetrix U133A/B arrays). Both datasets contain annotations of FLT3 mutational status (FLT3-ITD vs wildtype). We focused on adult AML cases and excluded any samples lacking FLT3 status information. Raw microarray data were background-corrected and normalized using the Robust Multi-array Average (RMA) method. For each dataset, we performed differential gene expression analysis between FLT3-mutant (FLT3+) and FLT3-wildtype (FLT3-) sample groups. In Dataset GSE10358, 91 samples were FLT3+ and 213 FLT3- (as per GEO metadata), while in Dataset GSE14468, 130 samples were FLT3+ and 394 FLT3- (including only cytogenetically normal AML as per the study design). Differential expression testing was conducted using the limma package (for microarrays) in R, fitting linear models with empirical Bayes moderation of variance. We obtained a list of significantly differentially expressed genes (DEGs) for each dataset, defined by a false discovery rate (FDR) < 0.05 and an absolute fold change threshold (|Log2FC| ≥ 1.0) between FLT3+ and FLT3- groups. DEGs were further classified as upregulated or downregulated in FLT3-mutant AML based on the direction of fold change [18].

2.2. Identification of Shared Dysregulated Genes

We next identified commonly dysregulated genes between the two datasets. We compared the DEG lists from GSE10358 and GSE14468 and took the intersection (overlap) of genes significantly dysregulated in the same direction in both cohorts. This yielded a set of shared DEGs that represent a robust FLT3-associated expression signature reproducible across independent patient populations. We recorded the number of shared upregulated and downregulated genes and examined their expression patterns. A Venn diagram was used to illustrate the overlap between the DEG sets from the two studies [19]. These overlapping genes served as the focus for subsequent network and pathway analyses. By concentrating on genes repeatedly implicated in FLT3+ AML, we aimed to filter out dataset-specific noise and highlight core biological changes linked to FLT3 mutations.

2.3. Network Analysis of Shared DEGs

We explored the functional interactions among the shared DEGs by constructing a protein–protein interaction (PPI) network. First, we retrieved known and predicted PPIs for the overlapping gene products using the GeneMANIAtool  for visualization and analysis [20]. In the network graph, nodes represented the proteins encoded by the shared DEGs, and edges represented direct or indirect interactions between them (such as physical binding or participation in the same complex/pathway). Key hub nodes (genes with many connections) were also observed, as these could represent critical regulatory points in the network. In parallel, we performed a pathway enrichment analysis on the shared DEGs to gain a high-level view of the dysregulated biological processes. This pathway network was then clustered to reveal functional modules, wherein groups of related pathways aggregate together. We observed, for example, a module of pathways all pertaining to cell migration, adhesion, and cytoskeletal dynamics. By collapsing individual gene lists into pathway clusters, this analysis helped prioritize coherent functional themes affected by FLT3 mutation.

2.4. Selection of Key Genes for Validation

From the combined network analyses (gene-level PPI and pathway networks), we selected four key genes for further validation. The selection was based on several criteria: [1] the gene occupied a central (high-degree) position in the PPI network or belonged to an important network cluster; [2] the gene was implicated in one of the top dysregulated pathways or modules (e.g., cytoskeletal organization or signaling); and [3] the gene had potential relevance to leukemia biology or was a plausible therapeutic target. Based on these considerations, we chose EPAS1, IQGAP1, PTPN4, and CTHRC1 for validation. We designed gene-specific primers for each and proceeded with experimental validation as described below.

2.6. qPCR Validation of Gene Expression

2.6.1. Cell Lines:Human acute myeloid leukemia (AML) cell lines HL-60 (FLT3-wild-type) and MOLM-13 (FLT3-ITD mutant) were used to model distinct molecular subtypes of AML. Cells were maintained in RPMI-1640 medium (Capricorn, RPMI-A) supplemented with 10–20% fetal bovine serum (FBS; Capricorn, FBS-11A), 1% penicillin-streptomycin (Gibco, Cat#15140122), and L-glutamine. Cultures were incubated under standard conditions at 37°C in a humidified atmosphere containing 5% CO?[21].

2.6.2. Quantitative PCR: Total RNA was extracted from HL60 and MOLM13 cells by using TRIzol reagent (Sigma-Aldrich, Cat#T9424), and quantified using NanoDrop 2000 (Thermo Fisher Scientific). RNA integrity was measured by 260/280 ratio (1.8–2.0). First-strand cDNA synthesis was performed using M-MLV Reverse Transcriptase (Sigma-Aldrich, Cat#M1302) by using oligo(dT) primers. Gene expression of the four selected candidates (EPAS1, IQGAP1, PTPN4, and CTHRC1) was measured by quantitative real-time PCR (qRT-PCR) (BioRad CFX Connect). Each reaction used SYBR Green dye for detection and was run in triplicate. We normalized the expression of target genes to a housekeeping gene (GAPDH) as an internal control, using the ΔΔCt method to calculate relative expression.

2.7. Survival Analysis

To evaluate the prognostic impact of the identified genes, we conducted a survival analysis by using GEPIA 2.0 [22] tool with a particular focus on EPAS1, which emerged as a strong candidate. We leveraged clinical outcome data from The Cancer Genome Atlas AML cohort (TCGA-LAML, which has overall survival information) to assess whether expression of the candidate genes correlated with patient survival. For EPAS1, AML patients were stratified in the TCGA cohort into “high EPAS1” vs “low EPAS1” expression groups (using the median EPAS1 expression (TPM) as the cutoff). A Kaplan–Meier survival curve was generated for overall survival, comparing FLT3-wildtype AML patients with high EPAS1 expression to those with low EPAS1 expression.

Results

3.1. Differential Expression in FLT3+ vs FLT3- AML

Transcriptomic differences between FLT3-mutant and FLT3-wildtype AML were apparent in both independent patient cohorts. In GSE10358, we identified a total of 499 DEGs (probes) (at FDR <0.05) differentiating FLT3+ from FLT3- samples. Among the annotated genes, 129 genes were upregulated in FLT3+ AML (higher expression relative to FLT3-wildtype), and 215 were downregulated in FLT3+ AML. Likewise, in GSE14468, we found 61 DEGs, with 40 upregulated and 7 downregulated in FLT3-mutant cases. To identify transcriptomic differences associated with FLT3 mutational status in acute myeloid leukemia (AML), differential gene expression (DEG) analysis was performed comparing FLT3-positive (FLT3?) and FLT3-negative (FLT3?) samples in two independent publicly available datasets: GSE10358 and GSE14468. In both cohorts, normalized expression data were analyzed using the limma framework, and genes with an adjusted p-value (Padj) < 0.05 were considered significantly differentially expressed (Figure 1).

In the GSE10358 cohort, volcano plot visualization revealed a substantial number of genes significantly dysregulated between FLT3+ and FLT3- AML samples, with both upregulated and downregulated transcripts observed (Figure 1A). Mean–difference (MA) plot analysis further demonstrated that these DEGs were distributed across a wide range of expression levels, indicating that FLT3-associated transcriptional changes were not restricted to lowly or highly expressed genes (Figure 1B). Unsupervised dimensionality reduction using Uniform Manifold Approximation and Projection (UMAP) showed partial but discernible separation between FLT3+ and FLT3- cases in the GSE10358 cohort, suggesting that FLT3 mutational status contributes to global transcriptomic variation while also overlapping with other sources of heterogeneity inherent to AML (Figure 1D). The DEG analysis was independently replicated in the GSE14468 cohort, which focuses on gene expression profiling in cytogenetically normal AML. Volcano plot representation again demonstrated distinct gene expression differences between FLT3+ and FLT3- samples (Figure 1E), while MA plot analysis confirmed consistent expression-dependent patterns of differential regulation (Figure 1F). UMAP analysis of the GSE14468 dataset revealed a similar trend to GSE10358, with FLT3+ and FLT3- samples exhibiting partial clustering but without complete segregation, reflecting both FLT3-driven transcriptional effects and broader AML heterogeneity (Figure 1H). Collectively, these analyses demonstrate that FLT3 mutational status is associated with reproducible yet context-dependent transcriptional alterations across independent AML cohorts. The DEG sets identified in each dataset formed the basis for subsequent overlap analysis, network construction, pathway-based interrogation, and experimental validation (Figure 1).

 

Figure 1. Differential gene expression and global transcriptomic patterns in FLT3-positive versus FLT3-negative AML across two independent datasets. (A) Volcano plot showing differentially expressed genes (DEGs) between FLT3-positive and FLT3-negative AML samples in the GSE10358 dataset using limma analysis (Padj< 0.05). Red dots indicate significantly upregulated genes, blue dots indicate significantly downregulated genes, and gray dots represent non-significant genes. (B) Mean–difference (MA) plot for GSE10358 illustrating log2 fold change as a function of average gene expression (log2Exp), highlighting differentially expressed genes between FLT3-positive and FLT3-negative samples. (C) Venn diagram summarizing the number of significantly differentially expressed genes identified in GSE10358 (Padj< 0.05). (D) UMAP projection (n_neighbors = 15) of GSE10358 samples colored by FLT3 mutational status, demonstrating partial separation between FLT3-positive and FLT3-negative AML cases. (E) Volcano plot depicting DEGs between FLT3-positive and FLT3-negative AML samples in the GSE14468 dataset (Padj< 0.05). (F) Mean–difference (MA) plot for GSE14468 showing log2 fold change versus average expression. (G) Venn diagram indicating the number of significantly differentially expressed genes identified in GSE14468 using limma (Padj< 0.05). (H) UMAP projection (n_neighbors = 15) of GSE14468 samples colored according to FLT3 mutational status, revealing partial transcriptomic clustering.

3.2. Shared Dysregulated Genes across Datasets

Intersecting the DEG lists from GSE10358 and GSE14468, we found 29 genes that were significantly differentially expressed in FLT3-mutant vs wildtype AML in both cohorts and with concordant direction of change. Out of the 29 shared DEGs, 25 were upregulated in FLT3+ AML across both datasets, whereas fifteen 4 were downregulated in FLT3+ AML, relative to FLT3-wildtype. The presence of these overlapping genes in independent patient groups strengthens their relevance and reliability as true FLT3-associated genes (Figure 2).

Figure 2. Overlap of differentially expressed genes between independent AML cohorts stratified by FLT3 mutational status. Venn diagrams illustrating the overlap of significantly upregulated (left) and downregulated (right) differentially expressed genes (DEGs) identified in FLT3-positive versus FLT3-negative AML samples from the GSE10358 (green) and GSE14468 (blue) datasets using limma analysis (Padj< 0.05).

Table 1 lists the shared DEGs, which include several intriguing candidates. Among the upregulated shared genes were: NKX2-3, HOXB-AS3, COL4A5, MAP1A, MMP2, CTSG, CPA3, MRC1, MIR155, SOCS2, ENPP2, IGFBP2, KCNK5. Several genes previously implicated in leukemia biology or hematopoietic regulation, including MSI2, IGFBP2, and CLU, were also part of this shared upregulated signature. In contrast, four genes were consistently downregulated in FLT3-positive AML relative to FLT3-negative cases across both datasets. These included GNG2 and PDE4B, which are associated with intracellular signaling pathways, as well as SNURF and HBG2, genes linked to transcriptional regulation and erythroid differentiation, respectively.

Table 1. List of genes consistently differentially expressed between FLT3-positive and FLT3-negative acute myeloid leukemia across the GSE10358 and GSE14468 datasets (Padj< 0.05). Genes are shown with their direction of regulation in FLT3-positive AML.

Upregulated Genes

FAM92A1

NKX2-3

C3orf80

GOLGA8A

TARP

COL4A5

STON2

CTSG

SEPP1

MSI2

HOXB-AS3

CLU

MIR155

MIR10A

KCNK5

IQCJ-SCHIP1

ADGRG1

SOCS2

ENPP2

CPA3

IGFBP2

C10orf128

MRC1

MAP1A

MMP2

Downregulated Genes

GNG2

PDE4B

SNURF

HBG2

 

3.3. Interaction Network of Shared DEGs

To explore the functional relationships among genes consistently dysregulated in FLT3-positive AML, we performed a network-based analysis using the shared differentially expressed genes identified across the GSE10358 and GSE14468 datasets. Protein–protein interaction and functional association networks were constructed to identify highly interconnected gene modules and prioritize biologically relevant hubs. The global interaction network revealed a densely connected structure, indicating that the shared DEGs do not act independently but rather participate in coordinated biological processes (Figure 3A). Several genes exhibited high connectivity within the network, suggesting potential regulatory or structural roles. Notably, genes involved in extracellular matrix organization, immune signaling, and cytoskeletal dynamics formed interconnected clusters, consistent with the functional diversity observed in the shared DEG list. Further inspection of the network enabled the identification of functionally distinct modules.

Figure 3. Network analysis of shared differentially expressed genes identifies functional modules in FLT3-positive AML. (A) Global interaction network constructed from genes consistently differentially expressed between FLT3-positive and FLT3-negative AML across the GSE10358 and GSE14468 datasets. Nodes represent genes, and edges represent functional or protein–protein associations. Node size reflects relative connectivity within the network. (B) Subnetwork highlighting a module associated with tissue interaction and cellular motility, enriched for extracellular matrix components and matrix remodeling genes. IQGAP1 and CTHRC1 are indicated as central nodes within this module (red circles). (C) Subnetwork representing a microenvironment and stress response module, containing genes related to hypoxia signaling and cellular adaptation. EPAS1 is highlighted as a key hub within this cluster (red circle).

One prominent module was associated with tissue interaction and cellular motility, characterized by a dense interaction pattern among extracellular matrix components and matrix remodeling enzymes (Figure 3B). Within this module, IQGAP1 and CTHRC1 emerged as central nodes, indicating potential roles in cytoskeletal regulation and cell migration–related processes in FLT3-positive AML. This module also included multiple collagen genes and matrix metalloproteinases, supporting its annotation as a tissue interaction and motility-related network. A second major module was enriched for genes linked to microenvironmental adaptation and stress response (Figure 3C). This cluster contained genes associated with hypoxia signaling, metabolic regulation, and cellular stress pathways. Within this module, EPAS1 (encoding HIF-2α) was identified as a highly connected node, suggesting a central role in coordinating stress and microenvironmental responses in FLT3-positive AML. EPAS1 by taking its relation with stress response; PTPN4 gene, a phosphatase that plays a role in immunity, learning, synaptic plasticity or cell homeostasis was taken into consideration. Together, these network analyses highlight that FLT3-associated transcriptional changes converge on specific functional programs, particularly those related to cell–matrix interactions, cytoskeletal organization, and microenvironmental stress adaptation. Based on their network centrality and functional relevance, IQGAP1, CTHRC1, PTPN4 and EPAS1 were prioritized for further experimental validation and prognostic evaluation (Figure 3).

2.4. Experimental Validation Confirms Upregulation of Network-Prioritized Genes in FLT3-Positive AML Cells

To validate the bioinformatically identified candidate genes, we assessed the expression of EPAS1, IQGAP1, PTPN4, and CTHRC1 by quantitative real-time PCR (qRT-PCR) in AML cell line models. The FLT3-negative HL60 cell line and the FLT3-ITD–positive MOLM13 cell line were used to represent FLT3-wildtype and FLT3-mutant AML, respectively. qRT-PCR analysis revealed a marked upregulation of EPAS1 expression in MOLM13 cells compared to HL60 cells (Figure 4A), consistent with the transcriptomic and network-based findings. Similarly, IQGAP1 expression was significantly increased in MOLM13 cells relative to HL60 cells (Figure 4B), supporting its identification as a central node within the tissue interaction and motility-related network module. In addition, PTPN4 and CTHRC1, both of which were implicated through pathway and network analysis, also exhibited higher expression levels in the FLT3-positive MOLM13 cell line compared to the FLT3-negative HL60 cell line (Figure 4C–D). The direction of change observed for all four genes was concordant with the differential expression patterns identified in patient-derived transcriptomic datasets. Collectively, these results provide experimental confirmation that genes prioritized through integrative network analysis are upregulated in FLT3-positive AML cells, supporting their potential functional relevance in FLT3-driven leukemogenesis and justifying further investigation of their biological and clinical significance.

Figure 4. qRT-PCR validation of network-prioritized genes in FLT3-positive and FLT3-negative AML cell lines. (A–D) Relative mRNA expression levels of EPAS1 (A), IQGAP1 (B), PTPN4 (C), and CTHRC1 (D) were measured by quantitative real-time PCR in HL60 (FLT3-negative) and MOLM13 (FLT3-ITD–positive) AML cell lines. Gene expression levels are shown as fold change relative to HL60 cells after normalization to a housekeeping gene GAPDH. Bars represent mean values, and individual data points indicate biological replicates.

2.5. Survival Analysis of Network-Prioritized Genes in AML

To assess the potential prognostic relevance of the network-prioritized genes, overall survival analysis was performed using transcriptomic and clinical data from an independent TCGA-LAML cohort. Patients were stratified into high and low expression groups for each gene based on median transcript per million (TPM) expression levels, and overall survival was evaluated using Kaplan–Meier analysis in GEPIA 2.0 tool. For EPAS1, patients with higher expression levels showed a trend toward improved overall survival compared to those with lower expression; however, this difference did not reach statistical significance (log-rank p = 0.14; hazard ratio [HR] = 0.66; Figure 5A). Similarly, IQGAP1 expression levels were not significantly associated with overall survival (log-rank p = 0.46; HR = 1.2), with largely overlapping survival curves between high and low expression groups (Figure 5B).

Analysis of PTPN4 revealed a non-significant trend toward reduced overall survival in patients with higher expression (log-rank p = 0.18; HR = 1.5; Figure 5C). In contrast, CTHRC1 expression did not demonstrate a clear association with overall survival, with comparable survival distributions between high and low expression groups (log-rank p = 0.31; HR = 0.75; Figure 5D).

Overall, while none of the examined genes showed a statistically significant association with overall survival in this cohort, the observed trends suggest that expression of selected network-derived candidates may influence clinical outcome in a context-dependent manner. These findings indicate that larger cohorts or stratified analyses may be required to fully elucidate the prognostic impact of these genes in AML.

Figure 5. Kaplan–Meier overall survival analysis based on expression levels of network-prioritized genes in AML by using GEPIA 2.0. (A–D) Kaplan–Meier curves depicting overall survival stratified by high versus low gene expression (median TPM cutoff) for EPAS1 (A), IQGAP1 (B), PTPN4 (C), and CTHRC1 (D). Blue lines represent low-expression groups, and red lines represent high-expression groups. Dotted lines indicate 95% confidence intervals. Log-rank p-values, hazard ratios (HR), and group sizes (n) are shown within each panel.

Discussion

FLT3-ITD–positive AML appears to have specific cellular programs that may underlie its aggressive clinical behavior. Our transcriptomic and network analyses indicate that FLT3-mutant blasts upregulate genes involved in adapting to hypoxic stress, remodeling the cytoskeleton, and modulating oncogenic signaling. Notably, the four genes validated (EPAS1, IQGAP1, CTHRC1, PTPN4) highlight these themes. Together, they suggest that FLT3-ITD AML cells evolve to survive in the bone marrow niche and potentially acquire greater motility and resistance to therapy, even if our survival analyses did not reach significance due to limited cohort size.

The genes consistently upregulated in FLT3-positive AML across independent cohorts indicate activation of pathways related to cell–microenvironment interaction, cytoskeletal organization, and stress adaptation. Several of these genes, including IGFBP2, MSI2, and CLU, have been previously linked to leukemic stemness, proliferation, and adverse disease features, suggesting that FLT3-mutant AML may reinforce self-renewal and survival programs beyond canonical kinase signaling. Upregulation of extracellular matrix– and cytoskeleton-associated genes (COL4A5, MMP2, MAP1A, and CTHRC1), together with the network-central scaffold protein IQGAP1, supports a shift toward enhanced cellular motility and altered adhesion to the bone marrow niche. Additionally, increased expression of immune- and inflammation-related genes (MIR155, CTSG, CPA3, and MRC1) and the hypoxia-responsive gene EPAS1 suggests that FLT3-positive AML cells may adapt to microenvironmental stress and exploit inflammatory signaling to support disease progression. Overall, the upregulated DEG signature highlights coordinated transcriptional programs that may contribute to the aggressive and treatment-resistant phenotype of FLT3-mutant AML.

EPAS1 encodes HIF-2α, a key hypoxia-inducible factor. Constitutive FLT3 signaling is known to stabilize HIF proteins via PI3K/AKT, enhancing hypoxia response pathways. In FLT3-ITD AML, elevated EPAS1 may promote leukemic cell survival under low-oxygen conditions and contribute to chemoresistance. Prior studies have linked hypoxia signaling to worse outcomes in AML, although the prognostic impact of HIF-2α itself has been variable. Interestingly, one recent analysis found low EPAS1 expression correlated with poorer survival, suggesting higher EPAS1 might reflect an active hypoxia response in less proliferative, more quiescent blasts. In our FLT3-ITD cohort, EPAS1 upregulation did not significantly predict survival, indicating that its effect might be context-dependent. Biologically, HIF-2α can support leukemic stem cell maintenance and drug resistance, but also has roles in differentiation and cell cycle control that could mitigate aggressive growth. These nuances underscore the need for larger studies and possibly FLT3-specific subgroup analysis to clarify EPAS1’s prognostic value. Clinically, the hypoxic niche is an appealing target – hypoxia-activated prodrugs are being tested to selectively kill AML cells in low-oxygen bone marrow regions [23]. Such approaches, or direct HIF pathway inhibitors, could potentially complement FLT3 inhibitors by eradicating dormant FLT3-ITD clones that evade conventional therapy.

Two of the upregulated genes, IQGAP1 and CTHRC1, point toward cytoskeletal reorganization and enhanced migratory or invasive properties in FLT3-ITD AML. IQGAP1 is a scaffolding protein that interacts with actin and multiple signaling pathways (Rac1/Cdc42, MAPK, Wnt, etc.), thereby coordinating cell motility, adhesion, and proliferation [24]. We found IQGAP1 highly expressed in FLT3-ITD+ cells, which aligns with recent reports that primary AML blasts overexpress IQGAP1 compared to normal hematopoietic cells [24]. Functionally, IQGAP1 appears critical for leukemia cell growth: knockdown of IQGAP1 in AML models significantly impairs proliferation and colony formation. Its overexpression may thus facilitate FLT3-driven signaling crosstalk and leukemic expansion. While our survival analysis showed only a non-significant trend, the direction was consistent with an adverse impact (higher IQGAP1 associated with shorter survival). This trend, also noted by others (p = 0.063 in a larger cohort), suggests IQGAP1 could contribute to a more aggressive phenotype, though it may not act as an independent prognostic marker without considering other factors. Importantly, IQGAP1’s druggability offers translational promise: a first-in-class small molecule inhibitor of IQGAP1 (UR778Br) was recently shown to selectively induce apoptosis in AML cells while sparing normal cells. Targeting the cytoskeletal signaling path in FLT3-ITD AML, perhaps via IQGAP1 or its downstream effectors, could inhibit leukemic cell migration and proliferation that are not fully addressed by FLT3 inhibitors alone. CTHRC1 (collagen triple helix repeat-containing 1) further reinforces the motility/invasion axis in FLT3-mutant AML. CTHRC1 is a secreted protein implicated in tissue remodeling and cell migration in solid tumors. Although it is not well-studied in AML, pan-cancer analyses identify CTHRC1 as an oncogene that promotes tumor progression through Wnt pathway activation, extracellular matrix modulation, and even immune evasion. Its upregulation in FLT3-ITD AML may reflect a shift toward a more motile, stem cell-like phenotype that can relocate or interact differently with the microenvironment. For instance, high CTHRC1 could conceivably facilitate AML cell egress from the bone marrow or create a protective niche by remodeling local stroma. These hypotheses align with clinical observations that FLT3-ITD blasts have high CXCR4/CXCL12 signaling interplay and possibly greater extramedullary involvement [25]. While speculative, the presence of CTHRC1 in FLT3+ leukemic cells invites further research into its role in AML cell homing, invasion of sanctuaries (like the central nervous system), or immune suppression in the marrow milieu. From a therapeutic standpoint, CTHRC1 could be explored as a biomarker of aggressive disease or as a targetable pathway (for example, via Wnt pathway inhibitors or immunotherapies) if its function in leukemia is confirmed. In contrast to the pro-oncogenic roles of the above genes, PTPN4 may represent a cellular attempt to calibrate or counterbalance hyperactive signaling in FLT3-ITD AML. Our data (and prior studies [26] show PTPN4 expression is markedly elevated in AML blasts, particularly in certain subsets such as FLT3-mutated cases. Notably, Kabir et al. identified PTPN4 among the top phosphatases overexpressed in AML, suggesting it might even help define specific AML subtypes. High PTPN4 could theoretically dampen certain pathways (possibly JAK/STAT or others) and confer a survival advantage by preventing oncogenic overstimulation. This nuanced role might explain why PTPN4 itself was not a clear prognostic factor in our cohort; its impact may depend on which downstream signals are dominant in a given patient’s leukemia. Further mechanistic studies are needed to clarify whether PTPN4 offers any tumor-suppressive brake in AML or if, paradoxically, leukemic cells leverage it to fine-tune signaling for optimal growth. Clinically, PTPN4’s consistent overexpression in AML could make it a useful diagnostic or disease-monitoring marker [26]. Although none of the four genes demonstrated a statistically significant impact on patient survival in our analysis, the trends observed are hypothesis-generating. These results echo the literature, where gene expression prognostic effects in AML often depend on disease context (molecular subtype, treatment received, etc.) and sample size. It is plausible that in a larger FLT3-mutant patient cohort or in specific molecular backgrounds, these genes’ expression levels could stratify risk. For example, EPAS1 or CTHRC1 might predict outcomes in normal karyotype FLT3-ITD AML, or IQGAP1 might be more prognostic in the presence of concomitant mutations. The present findings therefore warrant validation in expanded cohorts and possibly in a stratified manner (e.g. examining EPAS1 or PTPN4 within FLT3-ITD patients vs FLT3-wildtype). Furthermore, functional studies in FLT3-ITD models (knockdown or overexpression of these genes) would help confirm their mechanistic roles – such as whether reducing EPAS1 impairs FLT3-ITD cell survival under hypoxia, or if silencing IQGAP1/CTHRC1 diminishes the invasive behavior of FLT3-mutant blasts.

From a translational perspective, our study identifies potential molecular vulnerabilities in FLT3-ITD AML beyond FLT3 itself. The introduction of FLT3 tyrosine kinase inhibitors (TKIs) like midostaurin and gilteritinib has improved outcomes, yet relapse due to residual disease and microenvironment-mediated resistance remains a challenge [25] Targeting the pathways highlighted here could augment current therapies. For instance, combining FLT3 inhibitors with approaches that disrupt leukemic cell–niche interactions is a promising strategy. Notably, a 2019 study suggested that FLT3-ITD patients might benefit from CXCR4 inhibition added to FLT3 blockade to overcome microenvironment protection [25]. Similarly, HIF pathway inhibitors or hypoxia-activated cytotoxins could be combined with standard regimens to kill hypoxia-adapted AML cells that escape chemotherapy [23]. Our findings also raise the prospect of novel targeted therapies: IQGAP1 inhibitors could restrain AML cell proliferation and motility in FLT3-ITD cases, and indeed the preclinical success of an IQGAP1 inhibitor in AML is an encouraging proof-of-concept [24].

Conclusions

In conclusion, the differential expression profile of FLT3+ vs FLT3- AML underscores biologically meaningful differences: FLT3-ITD leukemic cells are primed for hypoxia tolerance (EPAS1 upregulation), show cytoskeletal and migratory adaptations (IQGAP1, CTHRC1), and alter phosphatase signaling feedback (PTPN4). These changes likely contribute to the clinical hallmarks of FLT3-ITD AML – rapid proliferation, extramedullary spread, and therapy resistance. While the individual genes did not reach independent prognostic significance in our dataset, they offer insights into the disease mechanism and suggest avenues for intervention. Future studies in larger patient cohorts and experimental models will be crucial to validate the functional importance of EPAS1, IQGAP1, CTHRC1, and PTPN4 in AML progression. Such work could pave the way for integrated treatment approaches – for instance, combining FLT3 inhibitors with hypoxia-targeted therapy, CXCR4 blockade, or cytoskeletal inhibitors – to improve outcomes for FLT3-ITD patients. Ultimately, harnessing these findings could inform both better risk stratification and the development of adjunct therapies that address the “non-FLT3” vulnerabilities of FLT3-mutant AML [25], moving closer to a more durable remission for this high-risk subset of patients.

Data availability statement

The source data underlying the submitted manuscript are available to other researchers through NCBI GEO repository [15].

Acknowledgements

Abbreviations

AML – Acute myeloid leukemia
FLT3 – FMS-like tyrosine kinase 3
FLT3-ITD – FLT3 internal tandem duplication
FLT3? – FLT3-mutant
FLT3? – FLT3-wildtype
DEG – Differentially expressed gene
PPI – Protein–protein interaction
qRT-PCR – Quantitative real-time polymerase chain reaction
qPCR – Quantitative polymerase chain reaction
GEO – Gene Expression Omnibus
TCGA – The Cancer Genome Atlas
TCGA-LAML – The Cancer Genome Atlas Acute Myeloid Leukemia cohort
GO – Gene Ontology
KEGG – Kyoto Encyclopedia of Genes and Genomes
UMAP – Uniform Manifold Approximation and Projection
RMA – Robust Multi-array Average
FDR – False discovery rate
FC – Fold change
Padj – Adjusted p-value
OS – Overall survival
TPM – Transcripts per million
PCA – Principal component analysis
MAPK – Mitogen-activated protein kinase
PI3K – Phosphoinositide 3-kinase
AKT – Protein kinase B
STAT5 – Signal transducer and activator of transcription 5
CXCR4 – C-X-C chemokine receptor type 4
HIF-2α – Hypoxia-inducible factor 2 alpha
EPAS1 – Endothelial PAS domain-containing protein 1
IQGAP1 – IQ motif containing GTPase activating protein 1
CTHRC1 – Collagen triple helix repeat containing 1
PTPN4 – Protein tyrosine phosphatase non-receptor type 4
MIR155 – MicroRNA-155
GAPDH – Glyceraldehyde-3-phosphate dehydrogenase

Funding:

No grants, financial support, or other assistance were obtained.




Conflicts of Interest:

There is no conflict of interest to declare.


References

[1] Cheng J, Qu L, Wang J, Cheng L, Wang Y. High expression of FLT3 is a risk factor in leukemia. Molecular Medicine Reports. 2018 Feb 1;17(2):2885–92.

[2] Meng F, Xiang M, Xie H, Liu Y, Qi Y, Zeng D. Retrospective study on the clinical outcomes and characteristics of acute myeloid leukemia: different outcomes in the same risk group. PeerJ. 2025 Dec 5;13:e20436.

[3] Chin KK, Ball BJ, Abaza Y, Altman JK, Thakur RK, Ali M, et al. Outcomes of relapsed or refractory acute myeloid leukemia after menin inhibition failure. Blood Adv [Internet]. 2025 Dec 6; Available from: http://dx.doi.org/10.1182/bloodadvances.2025018178

[4] Stancioaica MC, Coriu D, Ghiaur G. FLT3: A 35-Year Voyage from Discovery to the Next Generation of Targeted Therapy in AML. Cancers (Basel) [Internet]. 2025 Oct 23;17(21). Available from: http://dx.doi.org/10.3390/cancers17213415

[5] Haladik B, Maurer-Granofszky M, Zoescher P, Jimenez-Heredia R, Frohne A, Segarra-Roca A, et al. Image-based drug screening combined with molecular profiling identifies signatures and drivers of therapy resistance in pediatric AML. Cell Rep Med. 2025 Sep 16;6(9):102304.

[6] Orgueira AM, Raíndo AP, López MC, Rodríguez BA, Arias JÁD, Ferro RF, et al. Gene expression profiling identifies FLT3 mutation-like cases in wild-type FLT3 acute myeloid leukemia. PLOS ONE. 2021 Feb 16;16(2):e0247093.

[7] Dao B, Trinh VN, Nguyen HV, Nguyen HL, Le TD, Luu PL. Crosstalk between genomic variants and DNA methylation in FLT3 mutant acute myeloid leukemia. Brief Funct Genomics [Internet]. 2025 Jan 15;24. Available from: http://dx.doi.org/10.1093/bfgp/elae028

[8] Zhang C, Wen P, Zeng Y, Chen T, Guo Q, Liu C, et al. HOXA5-mediated spatial remodeling of tumor-immune interfaces across cancers promotes AML pathogenesis. Front Immunol. 2025 Oct 23;16:1677713.

[9] Ergun C, Kiraz Y, Ayna Duran G. Determination of survival associated genetic biomarkers to discover novel therapeutic targets for acute myeloid leukaemia. J Chemother. 2025 Jul 5;1–17.

[10] Banerjee J, Samanta S, Afiadenyo M, Ahmed R, Das S, Hasan SN, et al. Evaluation of antileukemic potentials of maslinic acid through targeting responsive proteins of human leukemia: a mechanistic perception on molecular modelling and dynamic stimulations. In Silico Pharmacol. 2025 Nov 21;13(3):191.

[11] Carranza-Aranda AS, Jave-Suárez LF, Flores-Hernández FY, Huizar-López MDR, Herrera-Rodríguez SE, Santerre A. and study of FLT3 inhibitors and their application in acute myeloid leukemia. Mol Med Rep [Internet]. 2024 Dec;30(6). Available from: http://dx.doi.org/10.3892/mmr.2024.13353

[12] Xu ML, Gru A, Geyer JT, Evans AG. Novel Diagnostic Tools and Biomarkers in Hematologic Malignancies. Frontiers Media SA; 2022. 218 p.

[13] Palma CA, Al Sheikha D, Lim TK, Bryant A, Vu TT, Jayaswal V, et al. MicroRNA-155 as an inducer of apoptosis and cell differentiation in Acute Myeloid Leukaemia. Mol Cancer. 2014 Apr 5;13:79.

[14] Ismail HD. Bioinformatics: A Practical Guide to NCBI Databases and Sequence Alignments. CRC Press; 2022. 492 p.

[15] Home - GEO - NCBI [Internet]. [cited 2025 Dec 14]. Available from: https://www.ncbi.nlm.nih.gov/geo/

[16] Tomasson MH, Xiang Z, Walgren R, Zhao Y, Kasai Y, Miner T, et al. Somatic mutations and germline sequence variants in the expressed tyrosine kinase genes of patients with de novo acute myeloid leukemia. Blood. 2008 May 1;111(9):4797–808.

[17] Wouters BJ, Löwenberg B, Erpelinck-Verschueren CAJ, van Putten WLJ, Valk PJM, Delwel R. Double CEBPA mutations, but not single CEBPA mutations, define a subgroup of acute myeloid leukemia with a distinctive gene expression profile that is uniquely associated with a favorable outcome. Blood. 2009 Mar 26;113(13):3088–91.

[18] Budak B, Tükel EY, Turanl? B, Kiraz Y. Integrated systems biology analysis of acute lymphoblastic leukemia: unveiling molecular signatures and drug repurposing opportunities. Ann Hematol. 2024 Oct;103(10):4121–34.

[19] Dashboard - Genomics Biotools [Internet]. [cited 2025 Dec 14]. Available from: https://www.biotools.fr/

[20] Warde-Farley D, Donaldson SL, Comes O, Zuberi K, Badrawi R, Chao P, et al. The GeneMANIA prediction server: biological network integration for gene prioritization and predicting gene function. Nucleic Acids Res. 2010 Jul;38(Web Server issue):W214–20.

[21] Synergistic Apoptotic Effects of Bortezomib and Methylstat on Multiple Myeloma Cells. Archives of Medical Research. 2020 Apr 1;51(3):187–93.

[22] Website [Internet]. Available from: https://gepia2.cancer-pku.cn/

[23] Zhong J, Wu H, Bu X, Li W, Cai S, Du M, et al. Establishment of Prognosis Model in Acute Myeloid Leukemia Based on Hypoxia Microenvironment, and Exploration of Hypoxia-Related Mechanisms. Front Genet. 2021 Oct 26;12:727392.

[24] Sahasrabudhe DM, Liesveld JL, Minhajuddin M, Singh NA, Nath S, Kumar VM, et al. In silico predicted compound targeting the IQGAP1-GRD domain selectively inhibits growth of human acute myeloid leukemia. Sci Rep. 2024 Jun 4;14(1):12868.

[25] Cao T, Jiang N, Liao H, Shuai X, Su J, Zheng Q. The FLT3-ITD mutation and the expression of its downstream signaling intermediates STAT5 and Pim-1 are positively correlated with CXCR4 expression in patients with acute myeloid leukemia. Sci Rep. 2019 Aug 21;9(1):12209.

[26] Liu Y, Zhang J, Du Z, Huang J, Cheng Y, Yi W, et al. Comprehensive analysis of family expression and prognosis in acute myeloid leukemia. Front Genet. 2022;13:1087938.


 


IJBMR -ISSN: 0976:6685. ©Copyright 2026 This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Published by BioMedSciDirect Publications