1 Introduction
Acute myeloid leukemia (AML), characterized by blocked myeloid lineage differentiation and accumulation of leukemic blast cells, is the most common acute leukemia in adults [
1]. AML is a highly heterogeneous disease driven by different types of mutations affecting multiple biological processes [
2]. Combination chemotherapy and hematopoietic stem cell transplantation are mainstream therapies for patients with AML, and treatment selection is based on clinical, hematological, and genetic indicators. Currently, over 70% of AML patients in therapeutics cannot live for more than 5 years [
3]. Hence, other effective therapeutic approaches must be developed urgently to improve the overall survival (OS) of AML.
AML was previously considered to be driven solely by genetic or epigenetic lesions. However, the niches, which are a specialized bone marrow (BM) microenvironment that regulates and maintains the normal production of blood and immune cells, are now increasingly acknowledged to play a vital role in the pathogenesis of AML [
4]. At present, the niches serve as predisposition events that contribute to mutant hematopoietic cell expansion, tumor progression, and adaptive resistance to chemotherapy, ultimately leading to relapse [
5]. Multiple potential mechanisms, including increased hypoxia, angiogenesis, inflammation, and metabolic reprogramming by stromal niche cells, are involved in the cancer-promoting role of niches. Consequently, remodeling of BM niches leads to immune evasion, support in leukemia pathological functions, and protection from chemotherapy [
4].
The fibroblast growth factor (FGF) family containing 22 polypeptides plays an important role in tumor microenvironment-related processes, including cell proliferation, angiogenesis, wound healing and repair, metabolic regulation, and embryonic development [
6]. FGF11, FGF12, FGF13, and FGF14 are members of the FGF homologous factor (FHF) subfamily of the FGF superfamily. FHFs differ from other FGFs because they cannot be secreted to the extracellular milieu and interact with FGF receptors (FGFRs) due to the lack of a secretory signal peptide [
7]. Thus, FHFs are intracellular proteins executing functions in a manner independent of FGFRs.
FGF13 plays a pivotal role across multiple tissues. In the cardiovascular system, FGF13 can protect the heart against cardiac hypertrophy during pressure overload by regulating the voltage-gated Na
+ channel [
8]. In the nervous system, FGF13 acts intracellularly as a microtubule-stabilizing protein leading to neuronal polarization and migration during brain development. FGF13-deficient mice display neuronal migration defects and weakened learning and memory [
9]. In the skeletal muscle system, FGF13 inhibits the proliferation and differentiation of skeletal muscle cells by downregulating Spry1, indicating that FGF13 plays a negative regulatory role in skeletal muscle development [
10]. Recently, its role in cancers has gained attention. Overexpression of FGF13 in HeLa cells increases resistance to cisplatin and decreases intracellular platinum concentration [
11]. FGF13 is mainly located in the cytoplasm and highly expressed in A549 cells; it promotes A549 proliferation by activating the AKT-GSK3 signaling pathway and inhibiting the activities of p21 and p27 [
12]. It also promotes the metastasis of triple-negative breast cancer and may represent a novel therapeutic target for blocked metastatic outgrowth [
13]. However, the definite role of FGF13 in AML remains unclear.
In the present study, we found that FGF13 was lowly expressed in patients with AML and represented a good prognosis indicator. FGF13 was associated with several clinicopathologic parameters, including white blood cell (WBC) count and cytogenetic risk, but not relevant to AML driver gene mutation. FGF13 also served as an independent prognostic factor for patients with AML. Results of functional analysis, protein–protein interaction (PPI) analysis, and immune infiltration analysis indicated that FGF13 suppressed AML progression by regulating BM niches. Finally, results of gain-of-function experiments demonstrated that FGF13 inhibited the proliferation and promoted the apoptosis of AML cells, and FGF13 overexpression prolonged the survival of recipient mice.
2 Methods and materials
2.1 Data acquisition
The RNA sequencing data (transcripts per million reads) of pan-cancer and normal control were downloaded from UCSC Xena based on the TCGA and GTEx databases. Data were normalized according to the description from UCSC Xena. The expression analysis of normal controls and patients with AML was obtained from GEPIA. Clinical information was obtained from the TCGA database (Table S1).
2.2 Evaluation of the FGF13 gene in AML
According to the median of FGF13 expression, patients were divided into high-FGF13 and low-FGF13 groups. Kaplan–Meier analysis was performed using the “survminer” and “survival” packages.
2.3 Establishment of a nomogram
Univariate and multivariate Cox analyses using the “survival” R package were performed to determine the independent prognostic factors. Then, a nomogram was established using the “rms” R package. Calibration curves were used to evaluate the accuracy of this nomogram. The “survivalROC” package in R software was used to calculate AUC values and draw the receiver operating characteristic (ROC) curve.
2.4 Functional enrichment analysis
The “limma” R package was used to determine differentially expressed genes (DEGs) between the high-FGF13 and low-FGF13 groups. The significance criteria were |logFC| > 1 and P-value < 0.05. Gene Ontology (GO) functional analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were implemented using the “ClusterProfiler” R package. Gene set enrichment analysis (GSEA) was implemented using GSEA software (version 4.1.0).
2.5 PPI network
The online tool STRING was used to construct a PPI network of the top 300 DEGs. Cytoscape software (v3.7.2) was used to visualize the interaction results from STRING. The cytoMCODE plugin was used to identify key subnetworks.
2.6 Immune infiltration analysis
ssGSEA using the “GSVA” R package was used to perform the immune infiltration analysis of FGF13. Twenty-four types of infiltrating immune cells were acquired as previously described [
14]. The association between FGF13 and enrichment scores of 24 immune cell types was determined through Spearman correlation analysis. Wilcoxon rank-sum test was used to analyze the enrichment scores between the high-FGF13 and low-FGF13 groups.
2.7 Association of FGF13 with immune-related factors
A co-expression analysis of FGF13 with immune checkpoint genes and cytokines was conducted through Spearman analysis. TMB scores were acquired through Strawberry Perl and corrected by dividing by the total length of exons. MSI scores were determined based on somatic mutation data. Spearman’s method was used to evaluate the association of FGF13 with TMB and MSI. Drug sensitivity processed data were downloaded from the CellMiner™ database. The drugs from this database in clinical trials or approved by FDA were included in this study. The “impute,” “limma,” “ggplot2,” and “ggpubr” R packages were used to process data and visualize results.
2.8 Cell culture
THP1 and NOMO1 cells were obtained from DSMZ and maintained in RPMI-1640 (Gibco, United States) supplemented with 10% fetal bovine serum (Gibco, United States) and 1% penicillin–streptomycin (Invitrogen, United States). Human bone marrow mononuclear cells (MNCs) were obtained from LONZA and maintained in IMDM (Gibco, United States) supplemented with 20% fetal bovine serum (Gibco, United States) and 10 ng/mL human cytokines, including SCF, TPO, FLT3L, IL-3, and IL-6 (PeproTech).
2.9 Xenograft studies
The xenograft mouse model was established by injecting 5 × 106 THP1 cells expressing oe-NC or oe-FGF13 into sub-lethally irradiated NOD mice. The human CD45+ cells in BM were detected through flow cytometry. Animal experiments were approved by the Ethics Committee of Ruijin Hospital Clinical Research Center, Shanghai Jiao Tong University School of Medicine.
2.10 Quantitative real-time PCR
Total RNA from cell lines was extracted using TRIzol reagent (Invitrogen, United States). Quantitative real-time PCR (qRT-PCR) was performed as previously described [
15]. The primers used for qRT-PCR were as follows: 5′-GTTACCAAGCTATACAGCCGAC-3′ (forward) and 5′-ACAGGGATGAGGTTAAACAGAGT-3′ (reverse).
2.11 Western blot
Western blot (WB) was performed as previously described [
15]. The primary antibodies (FGF13 and β-actin) were purchased from Abcam, United States. The anti-mouse or anti-rabbit secondary antibodies were obtained from Cell Signaling Technology, United States.
2.12 Transfection
The lentivirus containing FGF13 overexpression or a negative control sequence (NC) was obtained from OBIO (Obio TechnologyCorp, China). Transduction was performed in THP1 cells. Pools of stable transductants were generated by selection using puromycin (1 μg/mL) for 2 weeks.
2.13 Colony-forming assay
Cells were transduced with lentivirus and seeded into MethoCult H4434 Classic medium (StemCell Technologies) at a density of 1000 cells/well. After 10 days, the number of cells was counted.
2.14 Cell proliferation/growth and apoptosis assays
Cell growth was assessed using Cell Counting Kit-8 (CCK-8) proliferation assay (Dojindo, Japan) following the manufacturer’s instructions. For apoptosis assays, the annexin V-FITC/PI cell apoptosis kit (Cat. No.: KGA108, KeyGEN BioTECH) was used following the manufacturer’s instructions.
2.15 Statistical analysis
All statistical analyses were performed using the R software (version 4.1.1). All data are represented as the mean ± SD of three independent experiments. Student t-test or one-way ANOVA was applied to evaluate differences between groups. Statistical significance was considered at P < 0.05.
3 Results
3.1 FGF13 was lowly expressed in patients with AML
A pan-cancer analysis indicated that FGF13 was expressed heterogeneously across 33 cancer types, i.e., highly expressed in 15 cancer types and lowly expressed in 16 cancer types compared with normal controls (Fig.1). The abbreviations of 33 cancer types are listed in Table S2. Interestingly, the most striking FGF13 expression difference was found in AML. FGF13 was lowly expressed in AML cells compared with normal controls (human BM MNCs) (Fig.1). In addition, the patients in the high-FGF13 group showed a favorable prognosis (Fig.1). Considering the particularity of patients with acute promyelocytic leukemia (M3), we performed a prognostic analysis in this subgroup separately. Results showed that the patients with high FGF13 expression displayed a favorable prognosis in the M3 group (Fig. S1). Considering that only 15 patients with M3 were included to perform the prognostic analysis, we should be cautious with the conclusion. We also explored the association between FGF13 and prognosis at a pan-cancer level. Results showed that FGF13 was a favorable factor in LAML, SKCM, LGG, and OV but an adverse factor in LIHC and UVM (Fig. S2). Other FHFs were studied using Cox analysis. As expected, the results were heterogeneous and indicated FHFs as biomarkers but with both positive and negative prognostic values (Fig. S3).
3.2 Association between FGF13 expression and clinical features
We explored whether FGF13 expression was associated with the clinical features of patients with AML. Results showed that FGF13 was lowly expressed in the groups with WBC counts > 20×10 9/L and intermediate/poor cytogenetic risk; its expression was not significantly influenced by age and gender (Fig.2). For gene mutation factors, FGF13 was not related to AML driver gene mutation (Fig.2). In addition, we explored the association between FGF13 and clinical features in the normal karyotype subgroup. The results were similar to the results in the whole cohort (Fig. S4). Overall, low FGF13 expression was positively associated with poor prognosis clinical features.
3.3 Novel nomogram for AML prognosis
To determine whether or not FGF13 expression is a dependent prognostic factor, we conducted univariate and multivariate Cox analyses. Results showed that FGF13 can predict AML prognosis dependently (univariate Cox: hazard ratio (HR), 0.539; 95% confidence interval (95% CI), 0.351–0.829; P = 0.005; multivariate Cox: HR, 0.588; 95% CI, 0.374–0.927; P = 0.022; Table S3). To provide a clinically associated quantitative method to estimate OS for patients with AML, we developed an individualized nomogram based on the results of univariate and multivariate Cox analyses (Table S3). In this nomogram, FGF13 was integrated with age and cytogenetic risk as prognostic factors (Fig.3). Calibration curves of actual 1-, 3-, and 5-year OS versus predicted probabilities of 1-, 3-, and 5-year OS demonstrated an excellent concordance in the TCGA-LAML cohort, which suggested that the nomogram was accurate and reliable (Fig.3). In addition, the AUC values of 1-, 3-, and 5-year OS were 0.743, 0.809, and 0.916, respectively, indicating that the nomogram was reliable (Fig.3).
3.4 Functional enrichment analysis
To explore the definite role of FGF13 in AML development, we performed a series of enrichment analyses. First, we divided the patients into high-FGF13 and low-FGF13 groups. We found 689 differentially expressed protein-coding genes, including 122 downregulated genes and 566 upregulated genes (Fig. S5). GO and KEGG functional enrichment analyses showed that multiple gene sets were associated with the function of FGF13 in AML. KEGG included focal adhesion, ECM–receptor interaction, and protein digestion and absorption; biological process (BP) included skeletal system morphogenesis, extracellular matrix (ECM) organization, and extracellular structure organization; cellular components (CC) included a complex of collagen trimers, collagen trimer, and collagen-containing ECM; molecular function (MF) included glycosaminoglycan binding, ECM structural constituent conferring tensile strength, and ECM structural constituent (Fig.4). GSEA results showed that the terms of AML with PML RARA fusion, collagen formation, and upregulated genes in CD133+ cells (hematopoietic stem cells) compared with the CD133− cells were enriched in the high-FGF13 group, whereas the terms of glycolysis and disease of the immune system were enriched in the low-FGF13 group (Fig.4). A PPI analysis of the top 300 DEGs was performed to further understand the interactions of the DEGs. A network including 219 nodes and 511 edges was constructed (Fig.4). Then, 10 hub genes were identified in this network by using the cytoHubba plugin. Correlation analysis showed that FGF13 was positively associated with the 10 hub genes (Fig.4). These results suggest that FGF13 plays a role in AML development through multiple biological processes.
3.5 Association between FGF13 expression and immunity
Considering that AML arises in the context of the immunosuppressive BM microenvironment, we analyzed the relationships between FGF13 and immunity. Specifically, we explored the association of FGF13 and 22 immune checkpoint molecules in LAML and found 10 significantly related molecules (Fig.5). Cytokines CXCL8, CCL5, and TGFB1 were negatively related to FGF13, whereas IL12B was positively associated with FGF13 (Fig.5). A total of 24 infiltrating immune cell types were included in the correlation analysis (Fig.5). T cells and T helper cells were significantly enriched in the high-FGF13 group, indicating that FGF13 plays a protective role in AML through activating T cells and T helper cells (Fig.5 and 5E).
3.6 Associations of FGF13 with MSI, TMB, and drug sensitivity
We further investigated whether FGF13 is relevant to MSI and TMB. Results suggested that FGF13 was not associated with MSI (Fig.6) but was significantly related to TMB (Fig.6). In addition, FGF13 was significantly associated with 15 drugs (Table S4), and it was positively correlated with the drug sensitivity of bafetinib, okadaic acid, pipamperone, and raltiterxed, indicating that FGF13 could exert a synergistic effect on these anti-tumor drugs (Fig.6).
3.7 Biological effects of forced expression of FGF13 on human AML cells
The direct effect on AML cells must be tested to further evaluate the role of FGF13 in AML. An FGF13-overexpressing cell model was established (Fig.7 and 7B). Gain-of-function studies were performed, and overexpression of FGF13 substantially inhibited cell growth (Fig.7 and 7D) and significantly induced early apoptosis in the THP1 cell line (Fig.7 and 7F). Bcl-2 protein suppresses cell apoptosis [
16]. Preclinical studies showed that targeting Bcl-2 has emerged as an efficacious and well-tolerated clinical strategy in AML [
17]. In the present study, we investigated whether or not Bcl-2 expression is involved in FGF13-induced apoptosis in THP1 cells. WB results showed that FGF13 overexpression reduced Bcl-2 expression (Fig.7). In addition, FGF13 overexpression significantly prolonged survival in xenograft recipient mice (Fig.7). FGF13 overexpression also decreased CD45
+ cells in BM and reduced spleen weight (Fig.7 and 7J).
4 Discussion
Low FGF13 expression in patients with AML was associated with adverse prognosis. FGF13 was related to several clinicopathologic parameters and acted as an independent prognostic factor. FGF13 plays an oncogenic role in some solid tumors, including lung cancer [
12], breast cancer [
13], glioma [
18], and cervical cancer [
11]. However, we obtained contradictory results in AML. Some studies showed that FHFs can both favor and inhibit tumor growth. Overexpression of FGF14 in lung cancer cells decreases cell proliferation, colony formation, migration, and mesenchymal-to-epithelial transition
in vitro [
19]. Transcriptional silencing of FGF14 in colorectal cancer is regulated by DNA methylation. Re-expression of FGF14 in colorectal cancer cells induces cell apoptosis and inhibits cell viability and colony formation via the PI3K/AKT/mTOR pathway [
20]. Cox survival analysis indicated that FGF11 was also a favorable prognostic biomarker. FGF11 promotes non-small cell lung cancer progression [
21]. In bladder cancer, FGF11 is considered a cancer-promoting factor [
22]. However, the role of FGF11 in hematological tumors is unclear. Future studies could analyze the role of FGF11 in AML and investigate possible interactions between FHF family members.
FGF13 was associated with WBC count and cytogenetic risk classification but not with AML driver gene mutations. Our functional analyses showed that FGF13 was associated with multiple BM niche-related biological processes, including protein digestion and absorption, focal adhesion, ECM or collagen-related process, and skeletal system morphogenesis. BM niches include endothelial cells, osteoblasts, adipocytes, mesenchymal stem cells, and immune cells. The cross-talk of these cells determines the fate of AML cells. Obtained GSEA results revealed that “hematopoietic stem cell up”-related biological processes were enriched in the high-FGF13 group, indicating that FGF13 plays a role in normal hematopoiesis. Hence, the decreased expression of FGF13 in AML may be due to the impaired hematological cell function or transformation of normal HSCs.
The results of the PPI network indicated that 10 hub genes were associated with FGF13, including COL1A1, COL3A1, BMP4, MMP2, COL4A2, COL4A1, COL1A2, POSTN, BGN, and COL2A1. Meanwhile, FGF13 was positively associated with all top 10 hub genes. MMP2 is a member of the MMP family that plays a role in physiologic processes and pathological conditions, including cancer development [
23]. Decreased MMP2 expression in patients with AML is contrary to various solid tumors where increased expression is usually observed, which is similar to our findings for FGF13 [
24]. Reduced activation of the WNT canonical pathway induces low expression of BMP4 in hMSC-AML, contributing to leukemogenesis [
25]. COL1A1, COL3A1, COL4A2, COL4A1, COL1A2, and COL2A1 belonging to the collagen family regulate intercellular adhesion and differentiation and play vital roles in tumor formation and metastasis [
26,
27]. POSTN and BGN are essential ECM constituents that regulate cell proliferation and weaken cell adhesion via interacting with proteins in the intracellular and extracellular matrixes [
28,
29]. Hence, FGF13 may play a tumor-suppressive role by regulating BM niches.
Considering that remodeling of BM niches can facilitate immune evasion, we would like to ascertain the association between FGF13 and immunity. FGF13 was significantly associated with several checkpoint molecules in AML, such as PVR, TNFRSF18, LGALS9, HAVCR2, TNFRSF4, TNFSF4, TNFSF9, BTNL9, BTNL3, and BTN2A2. A previous study reported that TNFSF4, HAVCR2, and LGALS9 are considered immune-inhibitors while TNFRSF4, TNFRSF18, TNFSF9, and PVR are described as immune-stimulators [
30]. Co-expression analysis showed that the expression of these checkpoint genes could partially explain why FGF13 plays a suppressive role in AML. BTN2A2 is an MHC-associated gene that can inhibit T cell metabolism and cytokine secretion (IL-2 and IFN-γ); thus, it may be a tumor stimulator [
31]. In the present study, FGF13 was negatively associated with BTN2A2. BTNL9 suppresses invasion and correlates with a favorable prognosis of uveal melanoma [
32]. BTNL9 is frequently downregulated and inhibits breast cancer cell proliferation and metastasis via the P53-CDC25C/GADD45 pathways [
33]. We showed that FGF13 was positively related to BTNL9. For BTNL3, we failed to find its association with cancers from previous reports. Overall, FGF13 possibly exerts its suppressive function in AML by regulating checkpoint genes. FGF13 was related to cytokines CXCL8, CCL5, TGFB1, and IL12B. IL-8 (CXCL8) is a pro-inflammatory cytokine that is overexpressed in many solid tumors and has several well-documented tumor-promoting activities [
34]. CCL5 is a ligand of CCR5, and CCR5 signaling has multiple effects in promoting tumor survival through immune suppression, metabolic pathway reprogramming, promoting angiogenesis, and expansion of stem cells [
35]. TGFB1 is overexpressed in many tumors and facilitates tumor cells to overcome the growth-inhibitory effect [
36]. The IL12B gene encodes a subunit of IL12, a cytokine that acts on T and natural killer cells to eradicate tumor cells [
37]. In addition, our immune infiltration analysis indicated that FGF13 was positively associated with the enrichment score of T cells, including T helper cells, Th1 cells, and CD8
+ T cells. Btn2a2(−/−) mice exhibit enhanced effector CD4
+ and CD8
+ T cell responses and potentiate antitumor responses [
38]. IL-12 is a heterodimeric pro-inflammatory cytokine that induces the production of IFN-γ and favors the differentiation of T helper 1 cells, thereby enhancing antigen presentation [
39]. IL-12 also stimulates CD8
+ T cell proliferation and increases their cytotoxic activity to kill tumor cells [
40]. CXCL8
+ naive T cells are preferentially enriched CD31
+ T cells and do not express T cell activation markers or typical Th effector cytokines, including IFN-γ, IL-4, IL-17, and IL-22. Naive T cell-derived CXCL8-mediated neutrophil migration promotes tumor growth in an
in vivo human xenograft model [
41]. This culminated in the elegant hypothesis that T cell infiltration in BM could be controlled by FGF13-regulated cytokines. The possible role of FGF13 in bone marrow niches is proposed in Fig.8. In the future, wet laboratory experimentations should be performed to explore the definite molecular mechanisms of the immune landscape in BM niches regulated by FGF13.
FGF13 was positively correlated with the drug sensitivity of bafetinib, okadaic acid, pipamperone, and raltitrexed, indicating that FGF13 could exert a synergistic effect on these anti-tumor drugs. Bafetinib can enhance the activity of several pro-apoptotic Bcl-2 homology proteins and induce the apoptosis of Ph
+ leukemia cells [
42]. Okadaic acid provokes the mitotic arrest and apoptosis of leukemia cells via inhibiting protein phosphatase 2A [
43]. In a Phase I trial, raltiterxed is well tolerated when administered as a single agent to children with recurrent or refractory leukemia, and preliminary evidence of antileukemic activity has been observed [
44]. Pipamperone is rarely reported in patients with leukemia.
To a certain extent, several limitations of this study should not be ignored: (1) clinical information and data analyses were limited due to the retrospective data from TCGA; (2) the number of samples in this study were relatively small, and more patients need to be included in the future; (3) AML is a multifactorial disease. Thus, the investigation of a single gene cannot comprehensively interpret the association of AML risk. Hence, conclusions generated from this study would need future validation.
In conclusion, FGF13 is a favorable independent prognostic factor for AML. It plays a suppressive role in AML development by regulating BM niches.
PDF
(4927KB)
