Correlation of Matrisome-Associatted Gene Expressions with LOX Family Members in Astrocytomas Stratified by IDH Mutation Status

Tumor cell infiltrative ability into surrounding brain tissue is a characteristic of diffusely infiltrative astrocytoma and is strongly associated with extracellular matrix (ECM) stiffness. Collagens are the most abundant ECM scaffolding proteins and contribute to matrix organization and stiffness. LOX family members, copper-dependent amine oxidases, participate in the collagen and elastin crosslinking that determine ECM tensile strength. Common IDH mutations in lower-grade gliomas (LGG) impact prognosis and have been associated with ECM stiffness. We analyzed the expression levels of LOX family members and matrisome-associated genes in astrocytoma stratified by malignancy grade and IDH mutation status. A progressive increase in expression of all five LOX family members according to malignancy grade was found. LOX, LOXL1, and LOXL3 expression correlated with matrisome gene expressions. LOXL1 correlations were detected in LGG with IDH mutation (IDHmut), LOXL3 correlations in LGG with IDH wild type (IDHwt) and strong LOX correlations in glioblastoma (GBM) were found. These increasing correlations may explain the increment of ECM stiffness and tumor aggressiveness from LGG-IDHmut and LGG-IDHwt through to GBM. The expression of the mechanosensitive transcription factor, β-catenin, also increased with malignancy grade and was correlated with LOXL1 and LOXL3 expression, suggesting involvement of this factor in the outside–in signaling pathway.

Astrocytomas, or astrocytic gliomas, are the most common brain tumor. These tumors are classified according to their histologic and molecular characteristics and, more recently, according to isocitrate dehydrogenase (IDH) mutation status, where the presence of this mutation confers a more favorable prognosis [23]. Low-grade astrocytic gliomas (LGG), encompassing grades 2 and 3, are more well-differentiated, slower-growing tumors than grade 4 glioblastoma (GBM), with a median overall survival (OS) of 15 months [24]. Therefore, tumor aggressiveness progressively increases from LGG with IDH mutation (LGG-IDH mut ) to LGG with IDH wild type (LGG-IDH wt ), through to GBM. Tumor recurrence occurs independently of other factors, but time to recurrence varies according to malignancy grade and IDH mutational status, proving shorter in IDH wt than IDH mut tumors [25]. A growing body of evidence indicates that denser ECM leads to more aggressive tumor progression. Recently, an association between ECM stiffness oscillation and IDH mutation status was established [26].
Our group has previously reported a correlation between increased LOX expression and higher malignancy grade in human astrocytomas, and also a correlation of LOX expression with IDH1 mutation status [27]. In the present study, the expression of the other members of the LOX family in astrocytomas, stratified by malignancy grade and IDH mutation status, was explored. The differentially expressed LOX member genes were subsequently correlated with matrisome-associated gene expression in LGG-IDH mut , LGG-IDH wt , and GBM groups. Additionally, a search for mechanosensitive transcriptional factors, whose expression correlated with the group of differentially expressed genes, was performed. The study objective was to identify the LOX family members, along with matrisome components, that impact astrocytoma ECM stiffness. This knowledge can help identify LGG-IDH wt that are more prone to tumor progression, together with new candidates for interventions aimed at reducing tumor invasiveness.

LOX Family Expression Levels in Different Malignant Grades of Astrocytomas
Gene expression analysis by quantitative real-time PCR (RT-qPCR) for LOX, LOXL1, LOXL2, LOXL3, and LOXL4 showed increased expression levels in astrocytoma samples compared to non-neoplastic (NN) samples. Moreover, these expressions increased with malignancy, proving the highest in GBM. Pairwise comparisons of NN samples and diffusely infiltrative astrocytoma grades 2, 3, and 4 (AG2, AG3, and GBM, respectively) were significant (p < 0.0001) for all five LOX family genes. The paired comparisons relative to GBM were also significant, except for the LOXL3-AG3 comparison ( Figure 1A). The results for this cohort were confirmed in an independent larger cohort of The Cancer Genome Atlas (TCGA) and Genotype-Tissue Expression (GTEx), showing significant differences among groups for the five genes analyzed (p < 0.0001), with the highest significance found in the paired comparisons with GBM ( Figure 1B). Expression profiles of LOX, LOXL1, LOXL2, LOXL3, and LOXL4 in the TCGA cohort were also evaluated in low grade astrocytic glioma (LGG), encompassing AG2 and AG3, according to IDH mutation status and for GBM cases ( Figure 1C). Gene expression levels were higher in LGG-IDH wt than in LGG-IDH mut for all genes except LOXL2. Comparisons for the LGG-IDH wt -GBM pair showed significant differences for all members of the LOX family, with higher expression in GBM, except for LOXL4. Significant LOX, LOXL1, and LOXL3 differential expressions were observed in both LGG-IDH mut vs. LGG-IDH wt and LGG-IDH wt vs. GBM comparisons. Therefore, these three genes were selected for further analysis. Expression increased with malignancy grade of astrocytomas for the four genes analyzed. Statistical analysis was relative to GBM cases. (C) Expression levels of LOX and LOXL1-4 in TCGA RNAseq database, for LGG (AG2 and AG3) with (IDH mut ) and without (IDH wt ) IDH mutation, and GBM cases. Differences were statistically significant (p < 0.0001) as determined by Kruskal-Wallis test for all groups of genes in both cohorts. Middle lines represent median of groups. Top and the bottom lines represent first and third quartiles, respectively. The post-hoc Dunn's multiple comparison test was used to calculate differences between two groups (* p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001). Abbreviations: AG2, low-grade astrocytoma; AG3, anaplastic astrocytoma; GBM, glioblastoma; GTEx, Genotype-Tissue Expression; IDH mut , isocitrate dehydrogenase gene with mutation; IDH wt , isocitrate dehydrogenase gene wild type; LGG, lower-grade astrocytic gliomas; NN, non-neoplastic; RT-qPCR, real-time quantitative polymerase chain reaction; TCGA, The Cancer Genome Atlas.

LOX, LOXL1, and LOXL3 Protein Expression Analyses and Gene Expression Impact on Prognosis
Protein expressions of LOX, LOXL1, and LOXL3 were investigated by immunohistochemistry in LGG-IDH mut , LGG-IDH wt , GBM-IDH mut , and GBM-IDH wt . A progressive Expression increased with malignancy grade of astrocytomas for the four genes analyzed. Statistical analysis was relative to GBM cases. (C) Expression levels of LOX and LOXL1-4 in TCGA RNAseq database, for LGG (AG2 and AG3) with (IDH mut ) and without (IDH wt ) IDH mutation, and GBM cases. Differences were statistically significant (p < 0.0001) as determined by Kruskal-Wallis test for all groups of genes in both cohorts. Middle lines represent median of groups. Top and the bottom lines represent first and third quartiles, respectively. The post-hoc Dunn's multiple comparison test was used to calculate differences between two groups (* p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001). Abbreviations: AG2, low-grade astrocytoma; AG3, anaplastic astrocytoma; GBM, glioblastoma; GTEx, Genotype-Tissue Expression; IDH mut , isocitrate dehydrogenase gene with mutation; IDH wt , isocitrate dehydrogenase gene wild type; LGG, lower-grade astrocytic gliomas; NN, non-neoplastic; RT-qPCR, real-time quantitative polymerase chain reaction; TCGA, The Cancer Genome Atlas.

LOX, LOXL1, and LOXL3 Protein Expression Analyses and Gene Expression Impact on Prognosis
Protein expressions of LOX, LOXL1, and LOXL3 were investigated by immunohistochemistry in LGG-IDH mut , LGG-IDH wt , GBM-IDH mut , and GBM-IDH wt . A progressive increase in expression of these proteins was observed from LGG-IDH mut , LGG-IDH wt , and through to GBM, proving highest in GBM and confirming the results of the gene expression analyses ( Figure 2 increase in expression of these proteins was observed from LGG-IDH mut , LGG-IDH wt , and through to GBM, proving highest in GBM and confirming the results of the gene expression analyses ( Figure 2). The impacts of LOX, LOXL1, and LOXL3 expression levels were analyzed for OS in LGG and GBM, based on the TCGA database. In LGG, patients with higher levels of LOX (p = 0.033) and LOXL1 (p = 0.002) expression had shorter OS than patients with lower expression ( Figure 3A). LGG with highest and lowest levels of LOX expression had OS of 49.91 ± 7.38 months and 57.89 ± 10.59 months (p = 0.033), respectively. LGG with the highest and lowest levels of LOXL1 expression had OS of 19.87 ± 5.25 months and 57.89 ± 8.06 months (p = 0.0034), respectively. Similarly, OS of GBM patients was analyzed using Kaplan-Meier curves. Only LOXL1 expression levels impacted the prognosis of GBM, where patients with high expression had shorter OS ( Figure 3B). GBM patients with LOXL1 overexpression had mean survival of 10.28 ± 1.11 months, while those with lower expression had mean survival of 13.76 ± 1.09 months (p = 0.0034). Multivariate Cox regression analysis with age at diagnosis identified only LOXL1 expression as an independent variable for predicting prognosis in LGG (p = 0.027) ( Figure 3C) and GBM patients (p = 0.032) ( Figure 3D). In addition, radiotherapy was indicated to 45 (31.5%) out of 143 LGG patients presenting low LOXL1 expression, in contrast to 55 (49.5%) indications among The impacts of LOX, LOXL1, and LOXL3 expression levels were analyzed for OS in LGG and GBM, based on the TCGA database. In LGG, patients with higher levels of LOX (p = 0.033) and LOXL1 (p = 0.002) expression had shorter OS than patients with lower expression ( Figure 3A). LGG with highest and lowest levels of LOX expression had OS of 49.91 ± 7.38 months and 57.89 ± 10.59 months (p = 0.033), respectively. LGG with the highest and lowest levels of LOXL1 expression had OS of 19.87 ± 5.25 months and 57.89 ± 8.06 months (p = 0.0034), respectively. Similarly, OS of GBM patients was analyzed using Kaplan-Meier curves. Only LOXL1 expression levels impacted the prognosis of GBM, where patients with high expression had shorter OS ( Figure 3B). GBM patients with LOXL1 overexpression had mean survival of 10.28 ± 1.11 months, while those with lower expression had mean survival of 13.76 ± 1.09 months (p = 0.0034). Multivariate Cox regression analysis with age at diagnosis identified only LOXL1 expression as an independent variable for predicting prognosis in LGG (p = 0.027) ( Figure 3C) and GBM patients (p = 0.032) ( Figure 3D). In addition, radiotherapy was indicated to 45 (31.5%) out of 143 LGG patients presenting low LOXL1 expression, in contrast to 55 (49.5%) indications among 111 LGG patients with high LOXL1 expression. Nevertheless, the survival time was shorter for those with high LOXL1 expression ( Figures 3E and S1).  111 LGG patients with high LOXL1 expression. Nevertheless, the survival time was shorter for those with high LOXL1 expression ( Figures 3E and S1).  LGG, lower-grade astrocytic gliomas.

Matrisome Analysis
An in silico analysis of LOX family genes and of genes coding for matrisome proteins was performed to evaluate correlations among the expression profiles using the TCGA and GTEx databases, stratifying cases into LGG and GBM, according to IDH mutation status. Initially, the expression levels of the genes coding for the 1027 matrisome proteins and the of LOX family genes in LGG and GBM compared to NN and LGG compared to GBM, were compared using Limma-voom, as per the workflow depicted in Figure 4. Genes with a log fold change (FC) > |1| and an adjusted p ≤ 0.05 in any of the comparisons were selected, resulting in 439 upregulated genes and 233 downregulated genes, including LOX, LOXL1, and LOXL3. Expression of LOXL2 and LOXL4 did no differ significantly on any of the comparisons. Subsequently, a total of 672 up and downregulated genes in LGG-IDH mut , LGG-IDH wt , and GBM groups were analyzed using Kruskal-Wallis and Dunn's post hoc tests (LGG-IDH wt vs. LGG-IDH mut and LGG-IDH wt vs. GBM). This analysis led to the identification of 70 differentially expressed genes, comprising 9 genes coding for collagens, 17 for ECM glycoproteins, 23 for ECM regulators, 5 for ECM-affiliated proteins, 3 for proteoglycans, 10 for secreted factors, and 3 LOX genes (LOX, LOXL1, and LOXL3) (Supplementary Table S1). The workflow of the selection is presented in Figure 4. The heat map of these 70 genes is presented in Figure 5A. Higher gene expression levels were observed in LGG-IDH wt than in LGG-IDH mut , whereas the highest expression levels were detected in GBM cases, particularly in the mesenchymal molecular subtype.

Discussion
ECM stiffness contributes significantly to cancer progression by promoting tumor cell proliferation, invasion, and drug resistance. In fact, the matrix stiffness of glioma is higher than that of normal brain tissue and increases with malignancy [28,29]. The LOX family is widely known as amine oxidase enzymes responsible for crosslinking of ECM collagens and/or elastin that determines ECM tensile strength, remodeling, and integrity [6,7,18,[30][31][32][33]. LOX is the most extensively studied member of this family and, in a previous study by our group, a progressive increase in LOX expression according to malignancy was observed in human astrocytomas [34]. In the present study, the role of the other LOX family members in the progression of astrocytomas was confirmed. LOXL1, LOXL2, LOXL3, and LOXL4 expression levels were also found to increase progressively in astrocytomas from grades 2 to 4, proving highest in GBM, and significantly higher relative to non-neoplastic brain tissue. These results were subsequently confirmed in silico for the TCGA glioma RNAseq dataset. Similar results have been observed for other tumors, suggesting a key role of the LOX family members in tumor progression [9,18,21,[35][36][37][38][39][40]. In a previous study by our group, a correlation of LOX expression level with IDH1 mutation status was found in diffusely infiltrative astrocytomas (grades 2 to 4) [27]. Similarly, significant differential expression of the other LOX family members according to IDH mutational status was also observed in the present investigation. Oncogenic IDH mutations lead to decreased levels of α-ketoglutarate (α-KG) and accumulation of 2-hydroxyglutarate (2HG) [41], an oncometabolite linked to increased reactive oxygen species (ROS) production [42] and decreased mitochondrial respiration in GBM cells [43]. Moreover, 2HG is a competitive inhibitor of α-KG dependent dioxygenases, including histone demethylases and the ten eleven translocation (TET) family of 5-methylcytosine (5mC) LGG (AG2 and AG3) with (IDH mut ) and without (IDH wt ) IDH mutation, and GBM cases. Differences were statistically significant (p < 0.0001) as determined by Kruskal-Wallis test. Middle lines represent median of groups. Top and bottom lines represent first and third quartiles, respectively. The post-hoc Dunn's multiple comparison test was used to calculate differences in expression between two groups (*** p < 0.001; **** p < 0.0001). (B) Pearson's correlation coefficient test of CTNNB1 expression levels and LOX, LOXL1, and LOXL3 in groups of LGG-IDH mut , LGG-IDH wt , and GBM. Significative correlations (in red) between LOXL1 and LOXL3 with CTNNB1 were observed in LGG-IDH mut and LGG-IDH wt . Abbreviations: GBM, glioblastoma; IDH mut , isocitrate dehydrogenase gene with mutation; IDH wt , isocitrate dehydrogenase gene wild type; LGG, lower-grade astrocytic glioma; NN, non-neoplastic.

Discussion
ECM stiffness contributes significantly to cancer progression by promoting tumor cell proliferation, invasion, and drug resistance. In fact, the matrix stiffness of glioma is higher than that of normal brain tissue and increases with malignancy [28,29]. The LOX family is widely known as amine oxidase enzymes responsible for crosslinking of ECM collagens and/or elastin that determines ECM tensile strength, remodeling, and integrity [6,7,18,[30][31][32][33]. LOX is the most extensively studied member of this family and, in a previous study by our group, a progressive increase in LOX expression according to malignancy was observed in human astrocytomas [34]. In the present study, the role of the other LOX family members in the progression of astrocytomas was confirmed. LOXL1, LOXL2, LOXL3, and LOXL4 expression levels were also found to increase progressively in astrocytomas from grades 2 to 4, proving highest in GBM, and significantly higher relative to non-neoplastic brain tissue. These results were subsequently confirmed in silico for the TCGA glioma RNAseq dataset. Similar results have been observed for other tumors, suggesting a key role of the LOX family members in tumor progression [9,18,21,[35][36][37][38][39][40]. In a previous study by our group, a correlation of LOX expression level with IDH1 mutation status was found in diffusely infiltrative astrocytomas (grades 2 to 4) [27]. Similarly, significant differential expression of the other LOX family members according to IDH mutational status was also observed in the present investigation. Oncogenic IDH mutations lead to decreased levels of α-ketoglutarate (α-KG) and accumulation of 2-hydroxyglutarate (2HG) [41], an oncometabolite linked to increased reactive oxygen species (ROS) production [42] and decreased mitochondrial respiration in GBM cells [43]. Moreover, 2HG is a competitive inhibitor of α-KG dependent dioxygenases, including histone demethylases and the ten eleven translocation (TET) family of 5-methylcytosine (5mC) hydroxylases, which promote marked epigenetic alterations [44]. Given α-KG is also required by prolyl hydroxylases to promote the degradation of hypoxia-inducible factor 1 alpha (HIF-1α), the IDH mutation with decreased α-KG level leads to HIF-1α stability [45]. Thus, increased HIF-1α promotes ECM remodeling through regulation of collagen deposition by tumor cells [46] and upregulation of metalloproteinases and collagen-modifying enzymes in stromal cells [47]. Glioma aggressiveness and patient outcomes have also been found to correlate with HIF-1α levels and tenascin C-enriched ECM stiffness. IDH mut has been shown to restrict tumor aggression by decreasing HIF1α-dependent tenascin C expression, thereby decreasing ECM stiffness and mechanosignaling [28]. In fact, ECM stiffness has been described as significantly lower in IDH mut than in IDH wt tumors [28] and, clinically, LGG-IDH wt has a poorer prognosis than LGG-IDHmut [48]. Moreover, reports show LOX [49], LOXL1, and LOXL3 [50] expression levels are modulated by HIF-1α. Regulation of ECM stiffness and glioma cell migration by LOX expression have been shown in drosophila and mouse models [51], and LOX activity was reduced by HIF-1α knockdown [50]. Correlations of LOX expression with epithelial mesenchymal transition (EMT) and IDH1 mutation status have also been described [52]. Similarly, LOXL1 has been associated with tumor invasion, metastasis, and extracellular accumulation of lactate [53], linked to integrin α11, a stromal collagen receptor [54]. Associations of LOXL3 expression with EMT (through E-cadherin transcription repression by SNAIL) [55,56], cell invasion in breast cancer [57], and with GBM cell adhesion and invasion [10] have been described. Interestingly, genes related to EMT were overexpressed in IDH wt [58].
In the present study, a progressive increase in levels of LOX, LOXL1, and LOXL3 expression was observed for LGG-IDH mut , LGG-IDH wt , and GBM in the TCGA glioma RNAseq dataset, suggesting their role in determining ECM composition and stiffness enhancement in these phenotypes.
ECM stiffness is related to matrix composition, matrix contraction, and matrix crosslinking, where LOX family members play a major role [59]. Concerning matrix composition, collagen is an ECM scaffolding protein that contributes significantly to the tensile strength of tissue, binding cells by forming specialized extracellular networks [60]. Fibrillar collagen types I-III, V, and XI are the most common [61], and LOX, LOXL1, and LOXL3 are preferentially associated with fibrillar collagen types I and III [6]. In this study, COL1A1, COL1A2, COL3A1, and COL5A1 expression levels were correlated with all three LOXs in GBM, whereas COL6A1-3 expression levels correlated with LOXL1 and LOXL3 in LGG, and with all three LOXs in GBM. Fibrillar collagen deposition was described in the adventitia of remodeled large vessels and glomeruloid vascular structures of GBM [62]. High COL6A1 expression levels were observed in astrocytoma of different malignancy grades, especially in higher grade tumors associated with poor prognosis [63]. COL6A1 was detected in perivascular regions and pseudopalisading cells, while COL6A1 expression was associated with hypoxia [64] and VEGF expression [65].
Results showed overexpression of two genes coding for lysyl hydroxylases, procollagenlysine, and 2-oxoglutarate 5-dioxygenase (PLOD) [66], PLOD1, and PLOD2, involved in collagen biosynthesis and crosslinking [66,67] in LGG-IDH1 wt and GBM. Similar results were recently reported by other authors [68,69] and in other cancer types with aggressive phenotype [70]. We also found upregulation of another gene related to collagen biosynthesis, the serpin family H member 1 (SERPINH1), which was significantly correlated with LOXL1 in LGG-IDH wt and GBM, and with LOX in GBM. SERPINH1 encodes a heat shock protein (HSP47) localized in the endoplasmic reticulum, participates in the correct folding of collagen [71], and facilitates its secretion and deposition [72]. Increased expression of HSP47 has been associated with high malignancy grade of glioma [73]. Transforming growth factor β1 (TGF-β1) is also an inducer of collagen biosynthesis [74], and the gene coding for the transforming growth factor beta induced (TGFBI) was overexpressed and correlated significantly with LOX expression in GBM. TGFBI binds to collagens I, II, and IV and, therefore, mediates cell-collagen interaction. TGFBI also inhibits cell adhesion, promotes cell migration in glioma cells [75] and has been associated with the expression signature of mesenchymal GBM [76], the molecular subtype with the poorest prognosis.
The organization of collagen fibrils, including modulation of their diameters and interfibrillar spacing [77], is performed by lumican (coded by LUM), a class II leucine rich proteoglycan. The ECM organization process assembles adhesion plaque complexes and allows integrins to transduce cues from the ECM by activating a signaling cascade that induces cytoskeletal remodeling and regulates cell behavior [78]. Thus, LUM may enhance cancer growth through integrin β1 activating the β-catenin/focal adhesion kinase (FAK) [79]. Fibronectin [80] also induces α5β1 signaling and rapidly activates the downstream cascade through FAK, providing tensile support for motility of cancer cells in the invasion process [81]. In addition, β1 integrins and tumor adhesion to FN1 mediate resistance to radiotherapy [82]. Previous studies have reported that radiotherapy increases LOX secretion [83]. In the current analysis, both LUM and FN1 were upregulated and showed significant correlation with LOX in GBM. A high abundance of FN1 protein in the ECM of GBM was previously demonstrated by our group [84].
Metabolically, IDH wt possesses a cytosolic substrate composition that is better suited for collagen biosynthesis. Proline (Pro) constitutes about 10% of total amino acids in collagen [85], and can be synthesized from arginine, glutamine (Gln), and glutamate (Glu) [86]. The presence of IDH mut in glioma leads to a significant reduction in Gln and Glu levels, whereas IDH wt gliomas contain high levels of intracellular Glu [87,88], providing Pro for collagen biosynthesis through the activity of two enzymes: aldehyde dehydrogenase 18 family member A1, coded by ALDH18A1, and pyrroline-5-carbohydrate reductase 1, coded by PYCR1. These two enzymes synthesize Glu to glutamate Υ-semialdehyde and then to Pro, respectively. Interestingly, both these genes were upregulated in both LGG and GBM (data not shown).
In addition to Glu accumulation, an increase in lactate level can be observed in IDH wt due to the Warburg effect, with a shift of the oxidative tricarboxylic acid cycle towards glycolysis [89]. Glycolytically derived acids are transported to the extracellular microenvironment by membrane ion pumps and transporters, substantially lowering extracellular pH [90]. The acidification of tumoral environments produces favorable conditions for tumor cell invasion, as it induces the formation and maturation of invadopodia and activates proteases to focally degrade the ECM [91], through metalloproteinase 3 activity [92] and MMP9 secretion [93]. Moreover, acid-activated protease, such as cathepsin B (CTSB) [94], is secreted into the ECM at low pH. CTSB is located predominantly in secretory vesicles [95] and participates in ECM remodeling by degrading ECM components such as collagen, laminin, fibrin, elastin and TNC [96][97][98]. In gliomas, CTSB can bind to ANXA2 and induce the expression of vascular endothelial growth factor C, TGF-β, and MMP9 to promote angiogenesis [99,100]. CSTB upregulation has been described in GBM [101] and GBM stem cells [102], particularly in IDH wt , and has been strongly associated with the mesenchymal subtype and immunosuppressive conditions in gliomas [94]. Interestingly, CTSB was highly correlated with LOXL3 in LGG-IDH wt and with LOX in GBM. The 2-HG generated in IDH mut enhances angiogenesis through HIF1α stabilization, partially by decreasing levels of endostatin, an HIF1α antagonist, which in turn increases vascular VEGF signaling [103]. VEGFA was upregulated and significantly correlated with LOX in GBM. Additionally, the expression level of the collagen triple helix repeat containing-1 gene (CTHRC1), an ECM glycoprotein inducer of angiogenesis, migration, and cell invasion, showed the highest correlation with LOXL1 in LGG-IDH mut . FN1, significantly correlated with LOX expression in GBM in the present analysis, may also promote angiogenesis in cancer by providing a ridged structure for vessel development and signaling for endothelial cell migration [104].
Lastly, the ways in which ECM stiffness can trigger mechanotransduction were analyzed by searching for correlations between the differentially expressed genes and mechanosensitive transcription factors. Interestingly, there was a progressive increase in CTNNB1, which codes for β-catenin, according to astrocytoma malignancy grade, and its expression correlated significantly with LOXL1 and LOXL3 expression in both LGG-IDH mut and LGG-IDH wt . Collagen deposition and crosslinking promote ECM stiffness and clustering of integrins, leading to FAK activation with subsequent PI3K/AKT activation, inhibition of GSK-3β, and stabilization of β-catenin [105]. Nuclear accumulation of β-catenin and induction of stemness by matrix stiffness was observed in glioma cells [29]. LOXL1 has previously been shown to regulate cell migration and apoptosis via Wnt/β-catenin signaling [106]. However, this is the first report of correlation of LOXL3 with β-catenin expression level in cancer. Recently, our group identified upregulation of LOXL1 and involvement of Wnt/β-catenin signaling in the malignant transformation of normal astrocytes by anoikis [107]. We speculate that LOXL1, and its interactions with the matrisome-related genes, may be involved in this first step of transformation towards an astrocytic tumor. Additionally, elements of TCF/LEF transcription factors were identified in the promoter region of LOXL1, suggesting that the β-catenin signaling pathway may regulate positively LOXL1 expression [108].
Overall, the results showed a significant correlation of increased gene expression of LOXL1, LOXL3 and LOX with matrisome-related gene expression, possibly explaining the progressive increase in ECM stiffness from LGG-IDH wt , LGG-IDH wt , and through to GBM. Figure 7 summarizes our findings. More specifically, the LOXL1 correlation net was detected in less malignant astrocytomas. Connectivity with LOXL3 was prominent among more aggressive LGG-IDH wt , and strong connectivity with LOX was detected in GBM, demonstrating the complexity of ECM components involved in astrocytoma progression. More specifically, level of LOXL1 expression impacted the outcome of LGG and GBM patients, with shorter OS in individuals exhibiting higher expression. In addition, radiotherapy was indicated more frequently in LGG patients with high LOXL1 expression, yet this group had shorter survival than LGG patients with low LOXL1 expression. FAK activation with subsequent PI3K/AKT activation, inhibition of GSK-3β, and stabilizatio of β-catenin [105]. Nuclear accumulation of β-catenin and induction of stemness by matr stiffness was observed in glioma cells [29]. LOXL1 has previously been shown to regulate c migration and apoptosis via Wnt/β-catenin signaling [106]. However, this is the first report correlation of LOXL3 with β-catenin expression level in cancer. Recently, our group identifi upregulation of LOXL1 and involvement of Wnt/β-catenin signaling in the malignant tran formation of normal astrocytes by anoikis [107]. We speculate that LOXL1, and its interactio with the matrisome-related genes, may be involved in this first step of transformation towar an astrocytic tumor. Additionally, elements of TCF/LEF transcription factors were identifie in the promoter region of LOXL1, suggesting that the β-catenin signaling pathway may reg late positively LOXL1 expression [108]. Overall, the results showed a significant correlation of increased gene expression LOXL1, LOXL3 and LOX with matrisome-related gene expression, possibly explaining th progressive increase in ECM stiffness from LGG-IDH wt , LGG-IDH wt , and through to GBM Figure 7 summarizes our findings. More specifically, the LOXL1 correlation net was d tected in less malignant astrocytomas. Connectivity with LOXL3 was prominent amon more aggressive LGG-IDH wt , and strong connectivity with LOX was detected in GBM demonstrating the complexity of ECM components involved in astrocytoma progressio More specifically, level of LOXL1 expression impacted the outcome of LGG and GB patients, with shorter OS in individuals exhibiting higher expression. In addition, radi therapy was indicated more frequently in LGG patients with high LOXL1 expression, y this group had shorter survival than LGG patients with low LOXL1 expression.  LGG-IDH wt , and through to GBM, and have been attributed to the IDH mutation status and ECM stiffness. ECM composition in turn was directly related to differential expression levels of LOX, LOXL1, and LOXL3, and their correlations with matrisome-associated genes, especially fibrillar collagens type I, III, and V. LOXL1 expression level correlated significantly with expression of CTHRC1 in LGG-IDH mut . In LGG-IDH wt , strong correlations were observed between LOXL3 and cathepsin B expression levels. LOX expression correlated significantly with PLOD1 and PLOD2 (related to collagen biosynthesis) and with serpin H1 (involved in collagen folding). In IDH wt , Glu is accumulated and converted to Pro favoring collagen biosynthesis. The Warburg effect, with increased glucose uptake and glycolysis, leads to lactate production with lowering of extracellular pH, which promotes cathepsin B secretion and contributes to ECM degradation and remodeling. In GBM, LOX expression correlated with genes coding for lumican, TGFI, and fibronectin expression levels. ECM stiffness triggers mechanotransduction through α5β1 integrins and Wnt/β-catenin signaling with translocation of β-catenin to nucleus and transcription genes related to tumor malignant progression. Abbreviations: 2HG, 2-hydroxyglutarate; α-KG, α-ketoglutarate; CTHRC1, collagen triple helix repeat containing-1; ECM, extracellular matrix; FAK, focal adhesion kinase; GBM, glioblastoma; Gln, glycine; Glu, glutamine; GSK-3β, glycogen synthase kinase 3 beta; HIF-1α, hypoxia-inducible factor 1 alpha; IDH mut , isocitrate dehydrogenase gene with mutation; IDH wt , isocitrate dehydrogenase gene wild type; LEF, lymphoid enhancer factor; LGG, lower-grade astrocytic gliomas; PHD, prolyl hydroxylase; Pro, proline; TCA, tricarboxylic acid; TCF, T cell factor; TET, ten eleven translocation; Wnt, wingless and Int-1. Created with BioRender.com.

Tissue Samples
The casuistic samples consisted of 130 diffusely infiltrative astrocytomas (grades II to IV). Tumors were graded according to the WHO classification AG2 (n = 26, mean age at diagnosis 34.0 ± 8.1 years, 15 males and 11 females), AG3 (n = 18, mean age at diagnosis 35.0 ± 12.3 years, 11 males and 7 females), and GBM (n = 86 mean age at diagnosis 54.0 ± 13.9 years, 58 males and 28 females). The non-neoplastic control group consisted of samples from individuals undergoing temporal lobe resection during epilepsy surgery (n = 22, mean age at diagnosis 38.0 ± 7.6 years, 10 males and 12 females). All samples were collected during surgical procedures by the Neurosurgery Group of the Department of Neurology at the Hospital das Clinicas of the School of Medicine of University of Sao Paulo. Fresh surgical samples were immediately snap-frozen in liquid nitrogen upon surgical removal. Before RNA extraction, a 4-µm-thick section of each sample was obtained for histological assessment using a cryostat at −25 • C. Sections were stained with hematoxylin and eosin and examined under light microscope. Necrotic and non-neoplastic areas were removed from the frozen block of tumoral tissue by microdissection prior to RNA extraction. Grey matter was avoided in the control samples. Written informed consent was obtained from all patients according to the ethical guidelines approved by the Ethical Committee of the School of Medicine, University of São Paulo (0599/10).

Total RNA Extraction and cDNA Synthesis
Total RNA was extracted from frozen tissues using the RNeasy Mini kit (Qiagen, Hilden, Germany). RNA concentration and purity were determined by measuring absorbance at 260 and 280 nm. Ratios of 260/280 measures ranging from 1.8 to 2.0 were considered acceptable for purity standards. Denaturing agarose gel electrophoresis was used to evaluate the integrity of samples. A conventional reverse transcription reaction was performed to yield single-stranded cDNA. The first strand of cDNA was synthesized from 1000 ng of total RNA previously treated with one unit of DNase I (FPLC-pure, GE Healthcare, Uppsala, Sweden) using random and oligo (dT) primers, RNase inhibitor, and SuperScript III reverse transcriptase according to the recommendations of the manufacturer (Thermo Fisher Scientific, Carlsbad, CA, USA). The resulting cDNA was subsequently treated with one unit of RNase H (GE Healthcare), diluted with TE buffer, and stored at −20 • C until later use.

Reverse Transcription Quantitative Real Time PCR
Analysis of relative expression levels of LOXL1, LOXL2, LOXL3, and LOXL4 were performed by RT-qPCR using the Sybr Green I approach. Quantitative data were normalized in relation to the geometric mean of three reference genes: glucuronidase beta (GUSB), hypoxanthine phosphoribosyltransferase (HPRT), and TATA box binding protein (TBP). Table 1 shows the specific primer sequences and concentrations used for RT-qPCR. The minimum primer concentrations necessary were determined to give the lowest quantification cycle (Cq) and maximum amplification efficiency, while minimizing non-specific amplification. Primer concentrations used were 200 nM for all primers, except for GUSB which was 400 nM. Standard curves were established to ensure amplification efficiency, and analysis of melting curves demonstrated a single peak for all PCR products. Additionally, agarose gel electrophoresis was employed to check the size of the PCR products amplified. Sybr Green I amplification mixtures (12 µL

Immunohistochemistry
The immunohistochemical procedures for LOX, LOXL1, and LOXL3 expression analyses were performed on 4-µm slices of paraffin-embedded tissues of 6 cases each of AG2, AG3, and GBM. Tissue sections were first subjected to antigen retrieval in 10 mM citrate buffer, pH 6.0, and incubated at 122 • C for 3 min using an electric Pascal (BioCare Medical, Walnut Creek, CA, USA). Specimens were then blocked and further incubated with primary antibodies against human LOX (ab31238, Abcam, Cambridge, UK), LOXL1 (HPA 042111, Sigma-Aldrich, St Louis, MO, USA), and LOXL3 (ARP60280, Aviva, San Diego, CA, USA) at 16-20 • C for 16 h. Slices were then incubated for endogenous peroxidase blocking (Novolink Polymer Detection System, Novocastra, Newcastle upon Tyne, UK) at 16-20 • C for 16 h. Antibodies localization was visualized using diaminobenzidine and Harris hematoxylin. Table 2 show the positive controls were used for all reactions, as well as the dilutions for each antibody.

Statistical Analysis
The distribution of gene expression data was analyzed by the normality test of Kolmogorov-Smirnov and Shapiro-Wilk test. For gene expression, non-parametric Kruskal-Wallis test was used to compare multiple groups, followed by Dunn's post-hoc test. Kaplan-Meier survival curves were analyzed using the log-rank test. Multivariate Cox proportional regression analysis was performed using age as a covariate with gene expression level. Receiver Operating Characteristic curves were used to determine the high and low expression groups to Kaplan-Meier analysis for OS of LGG and GBM cases. The cut off of the expression values for ROC analysis was determined for the LGG cases were divided into IDH mut and IDH wt , and for GBM samples, the cases were divided into proneural and mesenchymal subtype. Limma package with voom method in R package was used for differentially expressed gene analysis. Correlation analyses between gene expression values were assessed by the non-parametric Pearson's correlation test. Correlation coefficient R value used was R > 0.5. Statistical significances were considered when p ≤ 0.05. All analyses were performed using SPSS version 20.0 (IBM Corporation, Armonk, NY, USA), R software, and plots were made using the program GraphPad Prism version 8.0 (GraphPad Software, San Diego, CA, USA).

Conclusions
The LOX family was increasingly expressed according to malignancy of astrocytoma, exhibiting the highest expression in GBM. The progressive gene expression connectivity among LOX, LOXL1, and LOXL3 and matrisome-related genes from LGG-IDH mut , LGG-IDH wt , and through to GBM, reinforce the impact of ECM components contributing for matrix stiffness on the malignant progression and prognosis of astrocytoma.

Conflicts of Interest:
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.