Abstract

Background

   Osteosarcoma is an aggressive bone malignancy with high metastatic
   potential and poor prognosis, primarily affecting children and
   adolescents. Although lysyl oxidase-like 4 (LOXL4) has been implicated
   in tumor progression, its functional role and mechanistic contributions
   in Osteosarcoma remain unclear.

Methods

   We performed integrated bioinformatics analysis using GTEx, TARGET, and
   TCGA datasets to evaluate LOXL4 expression and prognostic significance
   across cancers. Genetic alteration, immune infiltration, and RNA
   methylation analysis were carried to explore the different roles of
   LOXL4 in tumors. Functional assays, including Cell Counting Kit-8
   (CCK-8), colony formation, and Matrigel transwell assays, were
   conducted in Osteosarcoma cell lines. Besides, western blotting and
   gene set enrichment analysis (GSEA) were used to explore LOXL4’s
   mechanistic roles.

Results

   LOXL4 was significantly upregulated in Osteosarcoma tissues and
   associated with poor patient survival. Pan-cancer analysis revealed
   LOXL4 is upregulated in multiple cancer types and exhibits
   tumor-type-specific genetic alteration patterns, most frequently
   mutated in melanoma and amplified in endometrial carcinoma. Besides,
   LOXL4 expression significantly correlated with immune infiltration
   levels, the expression of immune checkpoint molecules, and RNA
   methylation across multiple cancers. Functional experiments
   demonstrated that LOXL4 knockdown suppressed cell proliferation,
   invasion, and epithelial–mesenchymal transition (EMT), whereas LOXL4
   overexpression enhanced these malignant phenotypes. Mechanistically,
   LOXL4 activated the Wnt/β-catenin signaling pathway. Inhibition of
   Wnt/β-catenin signaling with XAV-939 reversed LOXL4-induced oncogenic
   effects.

Conclusion

   LOXL4 promotes Osteosarcoma progression via Wnt/β-catenin-mediated EMT
   and cell proliferation. Its pan-overexpression and associations with
   the tumor immune microenvironment underscore its potential as a
   therapeutic target. Targeting LOXL4 or its downstream pathway may offer
   novel therapeutic strategies for Osteosarcoma.

Supplementary Information

   The online version contains supplementary material available at
   10.1007/s12672-025-03761-z.

   Keywords: LOXL4, Osteosarcoma, Progression, Wnt/beta-catenin signaling
   pathway, Bioinformatics

Introduction

   Osteosarcoma primarily impacts children and adolescents [[44]1]. The
   disease is characterized by strong local invasion and high
   susceptibility to distant metastasis, significantly contributing to its
   poor survival outcomes [[45]2]. Studies indicate that 10% to 15% of
   individuals with Osteosarcoma eventually develop metastases
   [[46]3–[47]5]. Although advancements in adjuvant chemotherapy have
   improved survival outcomes for many cancers, the metastatic potential
   of Osteosarcoma continues to pose a major therapeutic challenge,
   contributing significantly to mortality [[48]6, [49]7]. Treatment
   options for Osteosarcoma remain limited. This underscores an urgent
   need to understand its progression mechanisms and develop more
   effective therapies.

   The lysyl oxidase (LOX) family comprises five members (LOX and lysyl
   oxidase-like 1–4). These enzymes maintain tissue architectural
   stability by promoting collagen and elastin cross-linking in the
   extracellular matrix (ECM) [[50]8]. Collagen is a critical element of
   the ECM and a primary component of the basement membrane (BM). Notably,
   aberrant collagen expression has been identified as an early marker of
   invasive cancer characteristics [[51]9–[52]11]. Collagen changes are
   commonly seen in cancer cells from different origins. These alterations
   can impact tumor behavior, such as invasion and metastasis, influencing
   the progression and treatment outcomes of cancers including
   Osteosarcoma [[53]12]. Understanding ECM components and their
   interactions with tumor cells is crucial for developing targeted
   therapies to potentially inhibit the aggressive and invasive nature of
   Osteosarcoma. Accumulating evidence highlights the diverse biological
   functions of the LOX family, particularly its critical roles in tumor
   development, epithelial-mesenchymal transition (EMT), and metastasis
   [[54]13–[55]15]. Among its members, lysyl oxidase-like (LOXL4), a
   secretory lysyl oxidase, has been increasingly associated with tumor
   growth and progression [[56]16, [57]17] and is also classified among BM
   genes (BMGs) [[58]18]. LOXL4 contributes to gastric cancer progression
   through FAK/Src pathway activation [[59]8] and regulates metastasis,
   angiogenesis, and cell-matrix adhesion in liver cancer [[60]19]. These
   studies indicate that LOXL4 is an oncogene. Additionally, LOXL4
   catalyzes cross-linking of cell surface annexin A2 into multimers,
   inhibiting integrin β-1 internalization, and promoting cancer cell
   outgrowth [[61]20]. Despite these findings, the roles of LOXL4 in
   different cancers remain inconsistent and even contradictory
   [[62]21–[63]23], and its broader biological functions are still not
   fully elucidated. In particular, the role of LOXL4 in Osteosarcoma has
   not been previously explored. We hypothesize that LOXL4 may influence
   BM properties and facilitate Osteosarcoma progression.

   In this study, we demonstrate that LOXL4 is significantly overexpressed
   in Osteosarcoma tissues and is associated with poor clinical outcomes.
   Mechanistically, we show that LOXL4 promotes invasion, proliferation,
   and EMT in Osteosarcoma cells via activation of the Wnt/β-catenin
   signaling pathway. Our findings identify LOXL4 as a potential oncogene
   in Osteosarcoma and suggest its therapeutic relevance as a candidate
   target for intervention.

Materials and methods

Gene expression and survival prognosis analysis

   The Osteosarcoma sample data and clinical sources were acquired from
   the University of California, Santa Cruz (UCSC) Xena platform
   ([64]https://xena.ucsc.edu/). Patients with Osteosarcoma without
   survival data were excluded, resulting in a final cohort of 85 patients
   with Osteosarcoma having complete survival and gene expression data for
   analysis. Differentially expressed genes (DEGs) between Osteosarcoma
   and normal samples were identified on the count data using DESeq2,
   applying a significance threshold of p < 0.05 and |log2fold change
   (FC)| >1. A volcano plot and heatmap of DEGs were generated using the
   ggplot2 and ComplexHeatmap packages, respectively. To investigate the
   effect of LOXL4 gene on the survival of Osteosarcoma patients,
   univariate Cox regression analysis was conducted using the survival
   package. Kaplan-Meier (K-M) survival analysis was then employed to
   compare survival differences between risk groups. Abbreviations and
   full names of the corresponding tumors are given in Supplementary Table
   1. The CCLE dataset ([65]https://portals.broadinstitute.org/ccle) was
   utilized to obtain the mRNA expression matrix in tumor cell lines. The
   analysis was conducted using the ggplot2 package in R software.

Single-cell analysis

   Besides, this study estimated the LOXL4 expression level in cell types
   across numerous cancers via Tumor Immune Single-cell Hub (TISCH)
   database, an online platform designed for multiple single-cell analysis
   ([66]http://tisch.comp-genomics.org/home/). We download the
   Osteosarcoma single-cell data ([67]GSE152048) in GEO dataset, and the
   influence of unique molecular identifiers (UMIs) and mitochondrial
   content (%) was eliminated using Seutat’s ScaleData function.
   Subsequently, the batch effect was removed using the Harmony R package.
   The top 30 principal components and top 2000 variable genes were
   selected for cell clustering and the uniform manifold approximation and
   projection (UMAP) visualization [[68]24]. Re-cluster the cells using
   the t-distributed stochastic neighbor embedding method [[69]25].

Analysis of pan-cancer gene expression

   We downloaded the uniformly normalized pan-cancer dataset from UCSC:
   TCGA, TARGET, GTEx (PANCAN, N = 19131, G = 60499). Further, we
   extracted the expression data of the LOXL4 gene in each sample from it.
   In addition, we also filtered the samples with an expression level of
   0. Further, we performed a log2(x + 0.001) transformation on each
   expression value. Finally, we also eliminated the Cancer types with
   less than 3 samples in a single cancer type. Finally, the expression
   data of 34 cancer types were obtained. We used R software (version
   3.6.4) to calculate the expression differences between normal samples
   and tumor samples in each tumor, and used unpaired Wilcoxon Rank Sum
   and Signed Rank Tests for significance analysis of the differences.
   Then, the different expression situations of LOXL4 were visualized by
   the Sangerbox website ([70]http://www.sangerbox.com/) [[71]26]. At the
   same time, we use UALCAN online tools ([72]http://ualcan.path.uab.edu/)
   and the Gene Expression Profiling Interactive Analysis (GEPIA2) website
   to analyze the generic cancer data in the Osteosarcoma and LOXL4
   expression in normal tissue. Besides, we used the “Gene_DE” function of
   the Tumor Immune Estimation Resource, version 2 (TIMER2.0,
   [73]http://timer.cistrome.org/) website to observe the expression of
   LOXL4 in tumor tissues and corresponding normal tissues.

Analysis of pan-cancer gene survival prognosis

   We use the R package maxstat (Maximally selected rank statistics with
   several p-values approximations), Version 0.7–25 calculated the optimal
   cutoff value of LOXL4. Based on this, the patients were divided into
   high and low groups. Further, the survfit function of the R software
   package survival was used to analyze the prognostic differences between
   the two groups. The log-rank test method was used to evaluate the
   significance of prognostic differences among samples in different
   groups.

Analysis of protein structure and subcellular localization

   AlphaFold is an effective artificial intelligence (AI) technology that
   uses deep learning on both genetic and structural data to predict
   protein structures from amino acid sequences, and it also assesses the
   amount of trust in these forecasts to assure their dependability
   [[74]27]. In this study, we used AlphaFold to identify the
   three-dimensional (3D) structures of LOXL4. The Human Protein Atlas
   (HPA) ([75]http://www.proteinatlas.org) includes high-resolution images
   of immunohistochemistry (IHC) [[76]28]. The HPA database provides
   comprehensive insights into the spatial distribution of proteins within
   human cells, tissues, and organs. To determine the subcellular
   localization of LOXL4, we employed the HPA databases.

Genetic alterations and methylation levels analysis

   The analysis of genetic alterations was conducted by cBioPortal
   ([77]https://www.cbioportal.org/), and we downloaded the detailed
   information of alteration frequencies and types. Moreover, we also
   download all the level4 TCGA samples of Simple Nucleotide mutation
   Variation data set from GDC database
   ([78]https://portal.gdc.cancer.gov/) and then integrate the sample
   data. The information of the genetic alterations was obtained by using
   the R software package maftools. We extracted the expression data of
   the LOXL4 gene along with 44 marker genes representing three types of
   RNA modifications—m1A (10 genes), m5C (13 genes), and m6A (21
   genes)—from each tumor sample. Next, we calculated the Spearman
   correlation between LOXL4 and the marker genes of RNA modifications.

Immune analysis

   We analyzed the relationship between LOXL4 and the tumor immune-related
   situation using SangerBox. We conducted an investigation to determine
   the correlation between LOXL4 and 60 immune checkpoints, marker genes
   of 5 types of immune-associated biomarkers, and 3 immune infiltration
   scores (StromalScore, ImmuneScore, and EstimateScore). These scores
   predicted the tumor immune microenvironment by calculating the
   abundance of stroma and immune cells in tumor tissues. Among them,
   StromalScore reflects the interstitial components (such as fibroblasts
   and blood vessels), ImmuneScore represents the infiltration level of
   immune cells (such as lymphocytes and macrophages), and ESTIMATEScore
   is a combination of the StromalScore and ImmuneScore, which can
   indirectly predict tumor purity [[79]29]. The higher the score, the
   more non-tumor cell components are present and the lower the tumor
   purity. Studies have shown that a higher StromalScore is usually
   associated with lower tumor purity and a relatively better prognosis
   [[80]30].

Gene enrichment analysis

   Gene set enrichment analysis (GSEA) was used to explore the potential
   biological functions and signaling pathways regulated by LOXL4 in
   Osteosarcoma. To achieve this, the “gmt” data of the hallmark gene set
   (h.all.v7.4.symbols.gmt) was extracted from the MSigDB database
   ([81]https://www.gsea-msigdb.org/gsea/index.jsp). The analysis process
   was carried out via R packages such as “clusterProfiler,” “enrichplot,”
   and “ggplot2,” and the results were visualized.

Cell culture and transfection

   The hFOB1.19, MNNG/HOS, MG63, and U2OS cell lines were sourced from the
   Cell Bank of the Shanghai Chinese Academy of Sciences. Culture media
   were differentially employed: hFOB1.19 cells in hFOB-specific medium,
   MNNG/HOS and MG63 in Dulbecco’s modified Eagle medium, and U2OS in
   McCoy’s 5A medium. GenePharma (Shanghai, China) supplied the pcDNA3.4-
   LOXL4, pcDNA3.4-NC, two LOXL4-targeting siRNAs (si-LOXL4#1:
   5’-GCUGAAGAGCCUGACGAAUTT-3’; si-LOXL4#2: 5’-CCAAGUCUGCGGAUCACAUTT-3’)
   and corresponding negative control (si-NC). Cellular transfection was
   conducted following standard protocols using Lipofectamine™ 3000
   (Thermo Fisher Scientific, Waltham, MA, USA).

Western blotting and reagents

   Cellular lysis was performed using RIPA buffer (Beyotime, China) after
   phosphate-buffered saline (PBS) washing. Lysates underwent
   centrifugation (13,000 rpm, 15 min, 4 °C), and protein quantification
   was conducted via bicinchoninic acid (BCA) assay (FudeBiologic
   Technology, China). Electrophoretic separation using 4–10% SDS-PAGE
   preceded protein transfer to polyvinylidene difluoride membranes.
   Blocking with Protein Free Rapid Blocking Buffer (EpiZyme, China) was
   followed by overnight incubation with target-specific primary
   antibodies at 4 °C. Anti-β-actin (22957, 1:2000, Signalway Antibody),
   anti-LOXL4 (88186, 1:1000, Abcam), anti-Vimentin (60330-1-Ig, 1:50000,
   Proteintech), anti-N-cadherin (66219-1-Ig, 1:50000, Proteintech),
   anti-E-cadherin (20874-1-AP, 1:20000, Proteintech), anti-β-Catenin
   (51067-2-AP, 1:20000, Proteintech), anti-Cyclin D1 (26939-1-AP,
   1:20000, Proteintech), anti-c-MYC (10828-1-AP, 1:10000, Proteintech),
   Goat anti-mouse IgG H&L (HRP) (ab205719, 1:10000, Abcam), and Goat
   anti-rabbit IgG H&L (HRP) (ab205718, 1:30000, Abcam). MCE (USA)
   provided the pathway inhibitor XAV-939 (HY-15147).

Colony formation assays

   A cell suspension containing 200 quantified viable cells was plated
   into 12-well culture plates post-transfection [[82]31]. Throughout the
   14-day culture period, fresh medium was replaced every 3 days. The
   wells were cleaned with PBS, the colonies were fixed with methanol for
   30 min, colored with crystal violet, and photographed using a camera.

Cell viability assay

   In 96-well plates seeded with 3 × 10³ cells/well, Cell Counting Kit-8
   (CCK-8) reagent (FudeBiologic Technology, China) was introduced at 24,
   48, or 72 h post-transfection, followed by 2 h incubation. Optical
   density measurements at 450 nm were conducted using an ELISA microplate
   reader.

Matrigel transwell assays

   Transwell upper chambers (Corning, USA) were loaded with 2 × 10⁴
   cells/plate suspended in 200 µl serum-free medium, coated with 60 µl
   Matrigel matrix (Corning, USA). Lower compartments contained 600 µl
   serum-supplemented medium (10% FBS). Subsequent to 48 h incubation at
   37 °C, chamber membranes underwent paraformaldehyde fixation and 20-min
   crystal violet staining. Migratory cell quantification was performed
   using Olympus light microscopy (Japan).

Statistical analysis

   Statistical evaluations were conducted using GraphPad Prism (version
   10.0; USA) and SPSS software, with significance thresholds defined at
   p < 0.05.

Results

Identification of hub genes in osteosarcoma

   We analyzed gene expression using GTEx and TARGET datasets, comprising
   386 normal tissues and 88 Osteosarcoma samples, respectively.
   Differential expression was visualized via a heatmap and a volcano plot
   (Fig. [83]1A and B). We identified intersecting genes among those
   upregulated in tumors (p < 0.05), prognostic genes (p < 0.05), and
   BMGs, yielding 4 hub genes (Fig. [84]1C). Among these, COL13A1, ROBO1,
   and EFEMP2 have established roles in Osteosarcoma progression and
   prognosis [[85]32–[86]35], supporting the validity of our screening
   approach. While LOXL4 was also identified as a prognostic hub gene, its
   functional significance in Osteosarcoma remained unclear, prompting
   further investigation.

Fig. 1.

   [87]Fig. 1
   [88]Open in a new tab

   Identification of key differentially expressed genes in Osteosarcoma. A
   Heatmap illustrating upregulated and downregulated genes in
   Osteosarcoma compared to normal tissue. B Volcano plot highlighting the
   identification of genes with significant expression changes. C Venn
   diagram showing common genes between upregulated genes in tumor tissue,
   genes highly expressed with poor prognosis, and BMGs. D Analysis of
   GTEx and TARGET datasets reveals significantly elevated expression of
   LOXL4 in Osteosarcoma tissues. E Left: Dimensionality reduction
   visualization displays cellular subpopulations through t-SNE
   clustering, with color-coded annotations distinguishing distinct
   cellular subtypes. Center: Spatial mapping of transcriptional profiles
   illustrates target gene expression gradients across cell clusters using
   a nonlinear dimensionality reduction approach. Expression intensity is
   represented by a gradient color scale (low: dark hues; high: light
   chromatic values), correlating with transcript abundance. Right:
   Quantitative expression analysis demonstrates differential
   transcriptional activity of target genes among cellular subgroups via
   categorical comparison, visualized through a barplot format encoding
   expression magnitudes. F Kaplan–Meier survival curves correlating the
   expression of LOXL4 with Osteosarcoma prognosis. G The GTEx, TCGA, and
   TARGET databases show the expression levels of LOXL4 in both cancerous
   and healthy tissues. H Expression of LOXL4 in Osteosarcoma cell lines
   and hFOB1.19 cells at translational levels. I Three-dimensional protein
   structure of LOXL4. J-K The subcellular location of LOXL4 is shown by
   HPA database and immunofluorescence staining. *p < 0.05, **p < 0.01,
   ***p < 0.001, ****p < 0.0001

   The results of bioinformatics analysis indicate that LOXL4 was
   significantly upregulated in Osteosarcoma (Fig. [89]1D). Single-cell
   analysis of [90]GSE162454 data across 8 cell types confirmed higher
   LOXL4 expression in Osteosarcoma compared to osteoblasts (Fig. [91]1E).
   Kaplan–Meier survival analysis indicated that elevated LOXL4 expression
   was associated with poorer prognosis (Fig. [92]1F).

   Pan-cancer analysis revealed differential LOXL4 expression across
   multiple tumor types compared to normal tissues (Fig. [93]1G and
   Supplementary Figure. 1). Furthermore, high LOXL4 expression correlated
   with worse prognosis in GBM, GBMLGG, LGG, OV, STAD, and UVM
   (Supplementary Figure. 2).

   Western blotting confirmed higher LOXL4 protein levels in Osteosarcoma
   cell lines (MNNG/HOS, MG63, U2OS) than in normal osteoblasts
   (Fig. [94]1H). The predicted 3D structure of LOXL4 is shown in
   Fig. [95]1I. Utilizing data from the HPA database, LOXL4 was found to
   predominantly localize to vesicles within U2OS cells, as depicted in
   Fig. [96]1J. Notably, the subcellular localization of LOXL4 in U2OS
   cells was determined by employing nucleus, microtubule, and endoplasmic
   reticulum (ER) markers (Fig. [97]1K).

Genetic alteration analysis

   Genomic analysis via cBioPortal revealed tumor-specific LOXL4
   alterations (Fig. [98]2A). Melanoma showed the highest alteration
   frequency (8.41%), primarily missense mutations. Soft tissue sarcoma
   exhibited 5.88% alterations, mainly deep deletions, while endometrial
   carcinoma had 5% frequency, predominantly amplifications. Comprehensive
   genomic characterization data, including alteration categories,
   molecular positions, and affected case numbers, are systematically
   detailed in Fig. [99]2B.

Fig. 2.

   [100]Fig. 2
   [101]Open in a new tab

   Genomic alteration profiles of LOXL4 across different cancer types. A
   The frequency and types of LOXL4 genetic alterations from the
   cBioPortal database, and alteration types mainly include missense
   mutations, amplifications, and deep deletions. B Structural mapping and
   mutation profile of LOXL4 across various cancers

Immune related analysis

   Given the therapeutic relevance of immune checkpoint blockade, we
   performed the correlational analysis of LOXL4-immune checkpoint
   proteins (ICP) interactions. LOXL4 showed positive associations with
   ICPs in many tumors, reinforcing its potential as a novel
   immunotherapeutic target (Fig. [102]3).

Fig. 3.

   [103]Fig. 3
   [104]Open in a new tab

   Correlation analysis between LOXL4 expression and immune checkpoint
   (ICP) genes across multiple cancer types. The heatmap depicts
   correlation coefficients of LOXL4 expression with ICPs, and LOXL4 shows
   widespread positive correlations with immune checkpoints. *p < 0.05

   Pan-cancer analyses of LOXL4 revealed its broad immunomodulatory roles.
   Systematic co-expression profiling demonstrated significant
   associations between LOXL4 and diverse immune-related
   biomarkers—including chemotactic factors, receptor systems, major
   histocompatibility complex (MHC) components, immunosuppressive
   mediators, and immune-activating signals—across multiple malignancies
   (Supplementary Figure. 3, Supplementary Table 2).

   Furthermore, we also evaluated the correlation between LOXL4 expression
   and 3 immune scoring (StromalScore, ImmuneScore, and EstimateScore). In
   most tumor types, the expression of LOXL4 is positively correlated with
   StromalScore, ImmuneScore, and EstimateScore (Supplementary Table 3).

Analysis of LOXL4 methylation levels in pan-cancer

   It is reported that mRNA modifications regulate the occurrence and
   development of tumors, such as RNA methylation, especially m6A
   (N6-adenylate methylation) modification, and abnormal methylation may
   lead to tumorigenesis [[105]36–[106]39]. LOXL4 is correlated with 21
   M6A-related genes in tumors, and the figure indicates that the m6A
   methylation level is mainly positively correlated with LOXL4 in various
   malignant tumors, including FTO, METTL3, ALKBH5, and METTL14, etc.
   (Fig. [107]4 and Supplementary Table 4).

Fig. 4.

   [108]Fig. 4
   [109]Open in a new tab

   Correlation between LOXL4 expression and RNA methylation regulators
   across multiple cancer types. The heatmap displays correlation between
   LOXL4 and regulators of m1A, m5C, and m6A RNA modifications, including
   writers, readers, and erasers. *p < 0.05

LOXL4 promotes the proliferation of osteosarcoma cells in vitro

   Given the close association of LOXL4 with tumor progression and
   immunity across multiple cancers based on pan-cancer analyses, we
   further investigated its specific role in Osteosarcoma. Transcriptional
   expression profiles of 8 Osteosarcoma cell lines were obtained from the
   CCLE database (Fig. [110]5A). Based on these expression patterns,
   MNNG/HOS and U2OS cells were selected for knockdown experiments, while
   MNNG/HOS and MG63 cells were chosen for overexpression studies.
   Following transfection with si-LOXL4 or overexpression (OE) -LOXL4
   plasmid, knockdown and overexpression of LOXL4 were confirmed at the
   protein level by western blotting (Fig. [111]5B and E). Functional
   assays revealed that LOXL4 knockdown significantly impaired colony
   formation ability in U2OS and MNNG/HOS cells, whereas its
   overexpression enhanced clonogenic growth in MG63 and MNNG/HOS cells
   (Fig. [112]5C and F). Consistent with these findings, CCK-8 assays
   showed that LOXL4 downregulation reduced cell viability, while its
   upregulation promoted proliferation (Fig. [113]5D and G). Collectively,
   these results demonstrate that LOXL4 facilitates the proliferation of F
   cells.

Fig. 5.

   [114]Fig. 5
   [115]Open in a new tab

   Osteosarcoma cell growth is regulated in vitro by LOXL4. A
   Transcriptomic expression of LOXL4 genes in Osteosarcoma cell lines. B
   and E Validation of the transfection effect of LOXL4 in Osteosarcoma
   cell lines. C and F The effect of LOXL4 on Osteosarcoma cells’ ability
   to colony formation was assessed (black scale bar = 2 mm). D and G The
   effect of LOXL4 on Osteosarcoma cell line proliferation was assessed
   using CCK-8 assays. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

LOXL4 facilitates osteosarcoma cell invasion and induces EMT

   To assess LOXL4’s impact on Osteosarcoma cell invasion, Matrigel
   transwell assays were performed. LOXL4 silencing significantly reduced
   invasion in U2OS and MNNG/HOS cells, whereas LOXL4 overexpression
   enhanced invasion in MG63 and MNNG/HOS cells (Fig. [116]6A and B).
   Copper-dependent monoamine oxidase enzymes belong to LOXL4 and these
   enzymes facilitate the oxidative deamination of lysine residues in
   collagen and elastin, which is essential for the crosslinking of
   soluble collagen and elastin during the ECM remodeling process
   [[117]40]. This process is pivotal for EMT, a biological phenomenon
   enabling polarized epithelial cells to acquire mesenchymal
   characteristics, thereby promoting tumor migration and invasion
   [[118]41, [119]42]. GSEA of Osteosarcoma data revealed enrichment of
   EMT signaling in the LOXL4 high-expression group, suggesting its role
   in EMT activation (Fig. [120]6C–E). Consistently, western blotting
   showed that LOXL4 silencing increased E-cadherin and decreased vimentin
   and N-cadherin in U2OS and MNNG/HOS cells, while overexpression
   produced the opposite effect (Fig. [121]6F–G). These results indicate
   that LOXL4 promotes both invasion and EMT in Osteosarcoma cells.

Fig. 6.

   [122]Fig. 6
   [123]Open in a new tab

   The Osteosarcoma cells’ capacities for invasion and EMT are affected by
   LOXL4. A-B Matrigel transwell assays were used to determine if LOXL4
   affected the Osteosarcoma cell lines’ capacity for invasion (black
   scale bar = 400 μm). C-E The results of the gene set enrichment
   analysis (GSEA) show a positive association between EMT processes and
   genes in the LOXL4 high expression group. F-G The effect of LOXL4 on
   the EMT of Osteosarcoma cell lines was evaluated. * p < 0.05, **
   p < 0.01, *** p < 0.001, **** p < 0.0001

LOXL4 regulates WNT/β-catenin signaling in osteosarcoma

   To elucidate the mechanisms by which LOXL4 influences Osteosarcoma
   progression, we performed single-sample GSEA (ssGSEA) and GSEA using
   transcriptomic data from the TARGET database. Patients were stratified
   into high- and low-LOXL4 groups based on median expression levels. The
   high-LOXL4 expression group exhibited significant activation of the
   Wnt/β-catenin signaling pathway (Fig. [124]7A, B). Kyoto Encyclopedia
   of Genes and Genomes (KEGG) pathway analysis further revealed
   upregulation of key markers including Cyclin D1, c-Myc, and β-catenin
   in this group (Supplementary Figure. 4). Moreover, LOXL4 expression
   positively correlated with WNT/β-catenin signaling marker
   β-catenin/catenin beta 1 (CTNNB1) mRNA levels (Supplementary Figure.
   5). Experimental validation confirmed that LOXL4 knockdown
   downregulated Cyclin D1, c-Myc, and β-catenin protein expression in
   U2OS and MNNG/HOS cells (Fig. [125]7C), while overexpression elevated
   their levels (Fig. [126]7D). These results demonstrate that LOXL4
   activates the Wnt/β-catenin signaling pathway in Osteosarcoma.

Fig. 7.

   [127]Fig. 7
   [128]Open in a new tab

   LOXL4 regulates WNT/β-catenin signaling in Osteosarcoma. A The pathway
   score between the low and high expression groups of LOXL4 across ssGSEA
   (Osteosarcoma patients were classified into low and high expression
   groups based on the median LOXL4 expression level). B Genes in the
   LOXL4 high-expression group have a positive connection with
   WNT/β-catenin signaling according to the results of the gene set
   enrichment analysis (GSEA). C LOXL4 affected the Osteosarcoma cell
   lines’ levels of Cyclin D1, c-Myc, and β-catenin protein expression. D
   The impacts of LOXL4 and Wnt/β-catenin signaling pathway inhibitor on
   the protein expression level of Cyclin D1, c-Myc, and β-catenin of
   Osteosarcoma cell lines were assessed by western blotting. * p < 0.05,
   ** p < 0.01, *** p < 0.001

Osteosarcoma cell proliferation, invasion, and EMT are regulated by LOXL4 via
WNT/β-catenin signaling

   To further validate the functional role of Wnt/β-catenin signaling in
   LOXL4-mediated oncogenicity, we treated MG63 and MNNG/HOS cells with
   the Wnt/β-catenin inhibitor XAV-939. XAV-939 significantly suppressed
   pathway activity, as evidenced by reduced expression of Cyclin D1,
   c-Myc, and β-catenin. Subsequent overexpression of LOXL4 partially
   restored the levels of these proteins (Fig. [129]8A). Cell
   proliferation assays revealed a reduction in cell viability and growth
   in MG63 and MNNG/HOS cells upon suppression of the Wnt/β-catenin
   signaling. Notably, LOXL4 overexpression reversed these inhibitory
   effects. (Fig. [130]8B and C). Similarly, Matrigel transwell assays
   showed that XAV-939 impaired cell invasion, which was rescued by LOXL4
   upregulation (Fig. [131]8D). Furthermore, Wnt/β-catenin inhibition
   downregulated vimentin and N-cadherin while upregulating E-cadherin,
   and these effects were likewise attenuated by LOXL4 overexpression
   (Fig. [132]8E). These results collectively indicate that LOXL4 promotes
   malignant phenotypes in Osteosarcoma cells through activation of the
   Wnt/β-catenin pathway Fig. [133]9.

Fig. 8.

   Fig. 8
   [134]Open in a new tab

   LOXL4 regulates the biological functions of Osteosarcoma cells through
   the Wnt/β-catenin signaling. A The impacts of LOXL4 and Wnt/β-catenin
   signaling pathway inhibitor XAV-939 on the protein expression level of
   Cyclin D1, c-Myc, and β-catenin of Osteosarcoma cell lines. B The
   impacts of LOXL4 and Wnt/β-catenin signaling pathway inhibitor XAV-939
   on the proliferation of Osteosarcoma cell lines. C The impacts of LOXL4
   and Wnt/β-catenin signaling pathway inhibitor XAV-939 on the
   colony-forming ability of Osteosarcoma cell lines (black scale bar
   = 2 mm). D The impacts of LOXL4 and Wnt/β-catenin signaling pathway
   inhibitor XAV-939 on the invasion capacity of Osteosarcoma cell lines
   (black scale bar = 400 μm). E The impacts of LOXL4 and Wnt/β-catenin
   signaling pathway inhibitor XAV-939 on the EMT process of Osteosarcoma
   cell lines. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001

Fig. 9.

   [135]Fig. 9
   [136]Open in a new tab

   LOXL4 Promotes Osteosarcoma Progression via Wnt/β-catenin Signaling

Discussion

   Osteosarcoma’s high metastatic rate is a major driver of its poor
   prognosis [[137]43]. The lack of effective therapies underscores the
   need to identify novel molecular drivers. While LOXL4 is a known
   oncogene in other cancers [[138]8, [139]19, [140]20], its role in
   Osteosarcoma was unclear. Our study addresses this gap and reveals that
   LOXL4 is significantly upregulated in Osteosarcoma and associated with
   poor survival, positioning it as a critical prognostic marker. Beyond
   Osteosarcoma, our pan-cancer analysis uncovered that LOXL4’s genomic
   alterations are cancer-type specific. Furthermore, we found that LOXL4
   expression correlates with immune infiltration and m6A regulators,
   suggesting its potential role in shaping the tumor immune
   microenvironment—a previously unexplored aspect of its function. Many
   cancers are involved in complex interrelationships with the immune
   microenvironment [[141]44–[142]47]. Additionally, LOXL4 shows
   significant associations with m6A RNA methylation regulators,
   indicating potential epigenetic regulatory mechanisms [[143]48]. These
   multifaceted findings showed LOXL4 may be a key regulator of the tumor
   immune microenvironment and a promising therapeutic target across
   cancers.

   Next, we demonstrated that LOXL4 enhances Osteosarcoma cell
   proliferation, invasion, and EMT through experiments. These functional
   effects are consistent with the known roles of LOX family enzymes in
   ECM remodeling and EMT induction via their catalytic activity, which
   includes the production of hydrogen peroxide—a signaling molecule that
   can further modulate tumor cell behavior [[144]49–[145]55]. GSEA
   further supported the involvement of LOXL4 in EMT. Our findings not
   only corroborate previous reports of LOXL4’s oncogenic role in other
   malignancies [[146]8, [147]19, [148]20], but also significantly extend
   upon the work of Tan et al. (who identified LOXL4 as a prognostic hub
   gene [[149]56]) by elucidating its functional mechanistic impact in
   Osteosarcoma. However, the molecular mechanisms through which LOXL4
   exerts its oncogenic effects were not explored. Previous studies
   reported that the Wnt/β-catenin signaling plays a role in cancer
   progression, metastasis, and chemotherapy resistance [[150]57–[151]61],
   and the activation of this pathway can promote EMT through the nuclear
   accumulation of β-catenin and subsequent transcriptional downregulation
   of E-cadherin [[152]62]. To explore whether LOXL4 affects the
   progression of Osteosarcoma through Wnt/β-catenin signaling, we
   performed pathway enrichment analysis and found that the Wnt/β-catenin
   pathway may be a key mediator of LOXL4-driven oncogenicity. The
   subsequent western blotting experiments also confirmed that LOXL4
   expression positively correlated with the activation of Wnt/β-catenin
   signaling. Besides, inhibition of Wnt/β-catenin signaling with XAV-939
   attenuated the pro-tumorigenic effects of LOXL4 overexpression,
   confirming the functional dependency of LOXL4 on this pathway.

   Several limitations of our study should be acknowledged. First,
   although we establish a link between LOXL4 and Wnt/β-catenin
   activation, the exact molecular intermediates remain to be fully
   elucidated. LOXL4 may regulate the Wnt/β-catenin signaling pathway
   either by directly modulating key components within the pathway or by
   indirectly affecting other molecules that, in turn, influence
   Wnt/β-catenin activity. This requires further experimental exploration
   in the future. Second, while we employed extensive bioinformatic
   (including the use of AlphaFold, an effective AI technology for
   structural prediction) and functional validations, our study did not
   incorporate broader AI approaches. This is a significant consideration
   given that AI-integrated methods are revolutionizing biomedical
   research by identifying biomarkers and improving precision oncology
   [[153]63]. For instance, AI-driven analysis of transcriptomic data
   could in the future be leveraged to uncover complex relationships
   between LOXL4 and the tumor microenvironment or to predict its
   prognostic relevance across cancers with higher precision [[154]63].
   From a translational medicine perspective, our findings posit LOXL4 as
   a promising therapeutic target in Osteosarcoma. Targeted inhibition of
   LOXL4 or its downstream Wnt/β-catenin pathway could suppress metastatic
   progression and improve patient outcomes. However, to advance this
   potential toward clinical application, future research must include
   validation using animal models, as well as assessment of LOXL4
   inhibition alone or in combination with standard chemotherapy.
   Furthermore, the integration of LOXL4 expression into clinical
   decision-making could be powerfully aided by emerging AI-based tools.
   AI is not only revolutionizing biomarker discovery but also enhancing
   therapeutic strategies by optimizing choices, as demonstrated by
   Espinosa et al. [[155]64]. Therefore, future work should also explore
   developing AI-powered models to incorporate LOXL4 status with other
   clinical variables, thereby improving risk stratification and
   personalizing treatment plans for patients. Beyond these scientific and
   technical considerations, it is also essential to acknowledge the
   ethical dimensions associated with biomarker discovery and the
   development of targeted therapies. As precision oncology increasingly
   relies on biomarker-driven patient stratification and predictive
   modeling, issues such as equitable access to novel therapies, data
   privacy in handling large-scale genomic and transcriptomic datasets,
   and the interpretability of AI-assisted decision-making tools become
   critical. Addressing these challenges will be necessary to ensure that
   the translation of LOXL4-targeted strategies into clinical practice is
   both responsible and inclusive, ultimately benefiting a broader patient
   population.

Conclusion

   In summary, our study provides comprehensive evidence that LOXL4
   promotes Osteosarcoma progression through Wnt/β-catenin-mediated
   proliferation and EMT. Its widespread overexpression and association
   with the tumor immune microenvironment further support its role as a
   multifunctional oncogene. Targeting the LOXL4/Wnt/β-catenin axis may
   offer a novel therapeutic avenue for Osteosarcoma patients.

Supplementary Information

   Below is the link to the electronic supplementary material.
   [156]Supplementary Material 1.^ (137.2MB, docx)
   [157]Supplementary Material 2.^ (9.9KB, xlsx)
   [158]Supplementary Material 3.^ (144.8KB, xlsx)
   [159]Supplementary Material 4.^ (17.3KB, xlsx)
   [160]Supplementary Material 5.^ (51.6KB, xlsx)

Author contributions

   Wenjie Chen: Writing – original draft, Visualization, Validation,
   Software, Methodology, Investigation, Conceptualization. Fujie Xie:
   Writing – original draft, Visualization, Validation, Software,
   Methodology, Investigation. Jie Lv: Writing – review & editing,
   Resources, Investigation. Dixi Huang: Writing – review & editing,
   Resources, Investigation. Ronghao Zhong: Writing – review & editing,
   Resources, Investigation, Data curation. Zhijia Wen: Writing – review &
   editing, Resources, Investigation. Jiangsen Sun: Writing – review &
   editing, Resources, Investigation. Shaowei Zheng: Writing – review &
   editing, Resources, Investigation. Weile Liu: Writing – review &
   editing, Resources, Investigation. Haobo Zhong: Writing – review &
   editing, Resources, Investigation. Shoubin Huang: Writing – review &
   editing, Resources, Investigation.

Funding

   This work was supported by the Foundation of Guangdong Basic and
   Applied Basic Research Foundation (2023A1515140045 & 2023A1515140129 &
   2023A1515140183 & 2023A1515140034).

Data availability

   The datasets supporting this study are publicly available in the
   following repositories: UCSC (https://xenabrowser.net/) database
   provides the Pan-cancer dataset
   (https://xenabrowser.net/datapages/?cohort=TCGA%20TARGET%20GTEx&removeH
   ub=https%3 A%2 F%2Fxena.treehouse.gi.ucsc.edu%3A443), GTEx dataset
   (https://xenabrowser.net/datapages/?cohort=GTEX&removeHub=https%3 A%2 F
   %2Fxena.treehouse.gi.ucsc.edu%3A443), and osteosarcoma dataset
   (https://xenabrowser.net/datapages/?cohort=GDC%20TARGET-OS&removeHub=ht
   tps%3 A%2 F%2Fxena.treehouse.gi.ucsc.edu%3A443).Gene Expression Omnibus
   (GEO) datasets [161]GSE152048 were accessed via the NCBI GEO database:
   https://www.ncbi.nlm.nih.gov/geo/download/?acc=[162]GSE152048.

Declarations

Ethics approval and consent to participate

   Not applicable.

Consent for publication

   Not applicable.

Competing interests

   The authors declare no competing interests.

Footnotes

   Publisher’s note

   Springer Nature remains neutral with regard to jurisdictional claims in
   published maps and institutional affiliations.

   Wenjie Chen, Fujie Xie, and Jie Lv have contributed equally to this
   work and share first authorship.

Contributor Information

   Haobo Zhong, Email: zhonghaobohz78@163.com.

   Shoubin Huang, Email: ShoubinHuang81@163.com.

References