Liver cancer (LC) is the sixth most common malignant tumor and the third leading cause of cancer-associated mortality worldwide. Hepatocellular carcinoma (HCC) accounts for 75%-85% of all LC, according to the GLOBOCAN 2020 estimation (1). In east Asia, especially China, a high incidence of HCC was noted, and similarly, the incidence and mortality of LC in developing countries are significantly higher than those in developed countries (2, 3). The variations in the prevalence of LC amongst different populations and regions are attributed to a variety of environmental and genetic factors, such as aflatoxin, alcohol, smoking, chronic hepatitis virus infection, and type 2 diabetes (4–6). Despite significant advances in diagnosis and treatment of LC, including surgical resection, local ablation, liver transplantation, and sorafenib–regorafenib sequential therapy, the prognosis of LC remains poor (6). Based on most recent data, 841,000 new cases and 782,000 deaths of LC around the globe was estimated to occur each year (1). Therefore, it is of great value to explore novel diagnostic and prognostic biomarkers that are sensitive and specific, and to identify potential targets for medications (7).

With the advent of next-generation sequencing (NGS) and other techniques, increasing amount of information has become available for a variety of cancer types and other diseases (8–11). Thus, the mechanisms of cancers, as well as other diseases, have become more widely investigated based on bioinformatic methods, by combining information technology and molecular biology. Bioinformatics methods, such as data-mining, are also widely applied for identification of potential biomarkers as therapeutic targets, as well as diagnostic and prognostic predictors, and to explore the pathogenesis of malignancies at the molecular level (12–14).

As an extracellular enzyme, lysyl oxidase (LOX) oxidatively deaminates specific lysine and hydroxylysine residues to form allysines in the telopeptide domains of the collagen molecule, and thus plays a critical role in covalent cross-link formation in collagen fibrils (15). LOX is highly expressed in tissues containing elastic fibers and fibrillar collagen, such as skin, lung, and the fibrous lamina propria in the small intestine, stomach, and liver (16). In addition to LOX, four LOX-like proteins (LOXL-1, -2, -3, and -4) have also been identified in the LOX family (17–19). Studies have found that the LOX family was involved in carcinogenesis and tumor metastasis, through angiogenesis promotion, formation of mature extracellular matrix at the secondary site, focal adhesion kinase (FAK) activation, and other mechanisms (19–22).

The lysyl oxidase (LOX) family consists of five members: LOX, the first described member of this family, and its four related members called lysyl oxidase-like genes (LOXL1-4). Recent evidence suggests that the LOX family play important roles in liver cancer. LOX secreted by HCC promotes tube formation of endothelial cells through upregulation of VEGF, and overexpressed LOX increases angiogenesis, whereas LOXL1 was found to be increased in liver fibrosis models (23–25). As for LOXL2, its expression level was found to be higher in HCC tissues compared with non-tumor tissue (26). Although LOXL3 has been studied in different types of cancer, studies on its roles in liver cancer are limited (25). LOXL4 was found to increase the risk of invasion and metastasis, promote angiogenesis, and play a role in the immunosuppressive microenvironment in HCC (25, 27, 28)

Although previous studies have investigated the roles of the LOX family in LC, their exact roles and mechanisms, especially for LOXL1 and LOXL3, have yet to be further investigated (25). Previous studies have shown evidence of the potential prognostic values and therapeutic values of LOX family members (25, 26). Thus, online databases were mined to analyze the expression, mutation, function, and immune infiltration of LOX family members in LC, with the goal to determine their potential oncogenic role, as well as their diagnostic and prognostic value in LC.

As a common malignancy with the third leading cause of cancer-related mortality (1), LC risk is influenced by various environmental and genetic factors (4–6). Previous studies have demonstrated that LOX is highly expressed in the fibrous lamina propria in the small intestine, stomach, and liver, as well as other tissues that contain elastic fibers and fibrillar collagen (16). LOX, LOXL, LOXL2, LOXL3 and LOXL4 were to be in intracellular locations, perinuclear regions and intranuclear locations, and are secreted to exert their functions, such as extracellular enzyme for initiating covalent cross-link formation in collagen fibrils (15, 19, 32–35). After secretion, LOX family members oxidase crosslink collagen and elastin (19, 36). LOXs were found to be involved in various physiological or pathological pathways, both in extracellular modulation and intracellular signaling (32). Studies have found LOX family members to be involved in carcinogenesis and tumor metastasis by formation of mature extracellular matrix at the secondary site, FAK activation, and promotion of angiogenesis (19–22). Overexpressed LOX was found to promote angiogenesis (23, 25), and expression level of LOXL2 was higher in HCC than in non-tumor tissue (26). LOXL4 was found to increase the risk of invasion and metastasis of LC via various mechanisms such as angiogenesis and through its involvement in creating an immunosuppressive microenvironment (25, 27, 28). However, studies on the roles of LOLX1 and LOXL3 in liver cancer are limited (25). As the role of LOX family members in LC remains inconclusive, this bioinformatic study was conducted to analyze the expression, mutation, prognostic value, and functional enrichment of LOX family in LC.

We found that all five members of LOX family are higher expressed in LC tissues than in the normal tissues and their overexpression are positively correlated with each other, which is consistent with previous findings that the expression of LOX, LOXL2, and LOXL4 are upregulated in HCC (25). A previous study found a 30-fold increase of LOXL1 level in a liver fibrosis model (24). However, the role of LOXL3 in LC was not yet clear (25), and our results provided evidence that not only LOX1 but also LOXL3 is highly expressed in LC. As shown by the cBioPortal analysis, 5.16% of LC patients were found to have genetic mutation of LOX family members, and the mutation rate of LOXL2 was the highest. Further analysis of ROC curve showed that the AUC of LOXL2 was above 0.9 with a PPV of 0.994, indicating its potential role in diagnosis. These findings are consistent with those reported by Wong et al, in which the AUC of LOXL2 to distinguish non-HCC and HCC patients was 0.896 (26). Therefore, LOXL2 is a good candidate for a diagnostic marker in LC, especially HCC.

Based on the analyses through various tools, high expression of LOX and LOXL3 was found to predict worse prognosis. This proves the previous hypothesis of upregulation of the LOX level as a predictive sign for HCC, proposed by Lin et al. (25). LOX gene, located at chromosome 5q23.1, is consist of a variable N-terminal domain and a highly conserved C-terminal domain (25). LOX itself is an extracellular, matrix-embedded protein that plays an essential role in the cross-linking of the collagen fibrils and the deposition of insoluble collagen fibers (37, 38). Previous studies indicated that LOX overexpression induced the Epithelial-Mesenchymal Transition (EMT) (39). In addition, Yang et al. proved that the overexpression of LOX activated the angiogenesis partially through increasing the VEGF and enhancing the tube formation ability of endothelial cells in tumor initiating cells (TICs)-enriched HCC, and LOX inhibitor β-aminopropionitrile (BAPN) reverses the angiogenesis (40). Zhu et al. also found that the proliferative, migratory. and invasive abilities of HCC cells were attenuated, and the expression of vascular endothelial growth factor (VEGF) was decreased by the silencing of LOX, through the p38 mitogen-activated protein kinase (MAPK) signaling pathway (30). LOX3, located at chromosome 2p13.1, plays an important role in remodeling the cross-linking of the structural extracellular matrix (ECM) of fibrotic organs such as the liver (25, 41). It was also shown that higher expression of LOXL3 was regulated by TGF-β in gastric cancer (42). However, studies on the biological function of LOXL3 in HCC are still limited (25, 39, 43). Previous literature on the prognostic role of LOXL3 in LC patients was also minimal. Therefore, meta-analysis on LOXL3 was not conducted. Nevertheless, the result of this analysis added new evidence that LOXL3 could be potentially used a prognostic biomarker in addition to LOX. A nomogram based on LOX, LOXL3, and other predictors were developed which can help predict the mortality risk for an individual LC patient. Moreover, given their negative impacts on the survival in LC patients, LOX and LOXL3 may also serve as potential therapeutic targets. Although the ULCAN and TIMER did not verify the worse prognosis associated with LOXL4, our result based on TCGA data indicated the potential clinical significance of LOXL4 for worse outcome. As part of the LOX family, LOXL4 gene is located at chromosome 10q24.2 (25). The in vitro study suggested that TGF-β might induce LOXL4 upregulation in several different HCC cell lines, and LOXL4 mediated cell-matrix adhesion and cell migration in HCC via upregulation of Src and FAK phosphorylation (43). Although LOXL4 is an important extracellular protein, the HCC cell migration was promoted more by the intracellular LOXL4 (43). In contrast, other study revealed that 5-azacytidine (5-aza-CR)-mediated overexpression of LOXL4 reactivated wild-type p53 and promoted cancer cell death, thus suppressing the development of HCC cancers, which might indicate an improved clinical outcomes of HCC patients (28, 44). These complicated and even contradicting mechanism of LOXL4 in HCC might partially explained the inconsistent findings of its role in HCC prognosis from TCGA, TIMER and ULCAN database.

Further exploration of potential drugs associated with LOX and LOXL3 in LC by using Coremine Medical found 30 drugs. The drug that demonstrated the strongest interaction was aminopropionitrile. β-aminopropionitrile (BAPN), obtained from a natural source, was the first compound found to have inhibitory effect on LOX (45). A previous study suggested the potential therapeutic value of BAPN for liver metastasis in gastric cancer (46). Another animal study demonstrated antifibrotic effect of BPAN through reducing collagen fiber bundles and LOX level, which indicates its potential role in attenuating the development of liver fibrosis (47). It was also found that that BAPN acts by reversing the angiogenesis that was activated by the overexpression of LOX (40). Although quinone and copper were found to be potential interacting drugs, they are more likely to be identified because their own function and roles in the LOX proteins. Quinone is part of the redox cofactor of LOXs, which is a functional group in the catalytic domain of LOX proteins (48). LOX family members also contain a conserved copper-binding site in the C-terminal half of the protein (49). Copper binding to key histidine residues facilitates the formation of quinone-contained redox factor which in turn leads to the oxidase activity (48, 49). Therefore, quinone and copper can be potential research targets in the future to explore any potential practical use or potential use as therapeutic target. In addition, other drugs found through the exploration, such as cetuximab, bleomycin, cisplatin, paclitaxel, were known to have anti-cancer effects in various types of cancer, including LC, and anti-fibrosis effects in other diseases such as pulmonary fibrosis (50–59). Their exact roles and effects in LC may need to be clarified in future studies.

Exploration of the PPI network of LOX family and their top interactors found that these genes are primarily related to extracellular structures and functions. GO enrichment and KEGG pathway analysis of these genes also found they are mostly involved extracellular functions and structures. This is not surprising as it is well known that members of LOX family contribute to structural integrity and increased tensile strength by their catalytic activity, and exert roles in remodeling the cross-linking of the structural extracellular matrix (ECM) of fibrotic organs such as the liver (25). In addition, LOX family members are involved in scavenger receptor activity, oxidoreductase activity, and copper ion binding. Multiple scavenger receptor cysteine-rich (SRCR) domains exist n LOXL2 and LOXL3 (60, 61). As LOX family are copper-dependent amine oxidases (25), it is not unexpected that oxidoreductase activity and copper ion binding are involved. Further analysis via LinkedOmics database also identified a significant amount of co-expressed genes associated with each LOXL family members, and found that these co-expressed genes are also largely involved in extracellular and functions and structures, or participate in human tissues that contain elastic fibers, fibrillar collagen, and organs with a great amount of fibrous lamina propria.

The growth and metastasis of tumor cells depend on a complex tumor microenvironment (TME) (62). TME comprises of cells of hematopoietic origin, such as lymphocytes and myeloid cells, cells of mesenchymal origin, including mesenchymal stem cells, endothelial cells, adipocytes, fibroblasts, and myofibroblasts, and the ECM (63). ECM is a complex network providing structural support, biochemical reagents and biomechanical signals for the growth of cancer cells, and it consists of multiple components, including collagen, integrin, laminin, fibronectin, glycosaminoglycans, matrix metalloproteinases (MMP) and secreted cysteine-rich acidic proteins (64). Further analysis on the relationship of LOX family members and tumor-infiltrating immune cells in LC found positive correlations between the infiltration of B cell, CD8+ T cells, CD4+ T cells, macrophages, neutrophils, and DCs and all LOX family members. Moreover, the infiltration of B cells, CD8+ T cells, macrophages, and DCs were associated with worse outcomes. Immune cell infiltration in HCC under different conditions, such as bile acid-mediated immune cell infiltration (65) and TP53 mutations (66), have been investigated in the past. However, evidence on the associations between LOX family members and tumor-infiltrating immune cells in LC is limited. In addition, immunomodulatory drugs are under development for various conditions and have been approved in recent years for certain tumors such as multiple myeloma (67, 68). Therefore, we explored and identified a list of immunoihibitors, immunostimulators, and MHC molecules that are positively correlated with LOX and LOXL3, the two LOX family members with poor prognosis. These immunomodulators could be potential immunotherapeutic targets. The immune environment is thought to be critical in tumor progression and may even play a crucial role in different treatments for cancers, including chemotherapy, radiotherapy, and especially immunotherapy (69–71). Our findings suggest that there is a significant role of the LOX family in the tumor microenvironment. Therefore, comprehensive studies on the association of tumor-infiltrating immune cells, as well as immunomodulators, and LOX family in LC are needed.

It is well known that tumor heterogeneity relies on the TME, including both the cancer cells themselves and different types of immune cells and the surrounding stroma. TME closely correlates with the response to immunotherapy and the prognosis in multiple cancers (72). TME tends to be involved in the immunosuppression and drug resistance, resulting in less satisfactory responses to immunotherapy. In addition, immune checkpoint blockade (ICB) relies on restoring the function of T cell to eliminate tumors (70). Moreover, as part of the adaptive immune resistance, tumor cells could upregulate the immune checkpoint gene expression to suppress T cell activity that eventually leads to immune escape (73). Thus, our findings of positive association between CD8+ T cells and CD4+ T cells, as well as other immune cells, and LOX family members suggest that ICB and other immunotherapy could have a promising potential in LC treatment as high expression of LOX family members in tumor tissues facilitates immune cells infiltration, which could induce the immune response exerting the antitumor efficiency. It would particularly helpful to investigate compounds target on immunoinhibitors and immunostimulaters identified in our study. As demonstrated in our study immunoinhibitors, CSF1R, HAVCR2, and LGALS9, were found to be associated with LOXL3, while TGFB1, TGFBR1, and VTCN1 correlated with LOX. CSF-1R plays critical roles in regulating tumor-associated macrophages in TME, and targeted inhibition of the CSF-1/CSF-1R signal axis has broad application prospects in immunotherapy of malignant tumors (74). Pexidartinib is an orally administered small-molecule tyrosine kinase inhibitor that selectively inhibits CSF1R, and is currently being assessed for other types of cancer (75). Another kinase inhibitor, Derazantinib, also found to have activity against CSF1R and is under investigation for cholangiocarcinoma (76). Their potential use in LC also deserves further exploration. HAVCR2, also known as TIM-3 and CD366, enhances T cell inhibition and apoptosis and immune-suppressive activity of Tregs (77). Antibodies against HAVCR2 disrupt the binding of the ligands to HAVCR2 are under investigation as a potential combination partner of anti-PD-1/L1 therapy (78). Previous study also found that HAVCR2 receptor limits T-cell responses by interacting with LGALS9 (79). A recent study also demonstrated that chemoradiation could induce increased expression of PD-L1 and LGALS9 in gastric cancer (80), however, whether similar result can be found in LC needs further study. TGF-B1 is a potent inhibitor of T cell growth, partly by inhibiting IL-2 expression and secretion by T cells themselves (81), and interestingly, it can also affect anti-tumor T cell responses by downregulating MHC molecules on the surface of tumor cells. Despite its critical roles, the development of TGFB1-targeting therapies has not been progressed well, probably due to concern of severe toxicities that could arise from blocking tumor suppression exerted by TGF-β1 at early stages of tumorigenesis as TGFB1 exerts potent cytostatic and pro-apoptotic activities in pre-malignant cells (82, 83). In addition, blocking TGFB1 activities on normal cells outside of the TME may also lead to toxicities (82). Nevertheless, certain antibodies are still under investigation. For example, studies on Fresolizumab, a fully human monoclonal IgG4 antibody that neutralizes mature TGFB1, were conducted for malignant pleural mesothelioma, melanoma and renal cell carcinoma (82). Galunisertib, another TGFBR1inhibitor, was found to have 16% of objective responses in glioblastoma patients with no serious treatment-related toxicities (82, 84). Another clinical trial for pancreatic cancer patients showed that combination of chemotherapy (gemcitabine) and galunisertib was associated with increased survival compared to chemotherapy alone (85), and it is now also tested for combination with anti-PD-1 antibodies (82). In addition, a new TGFBR1 kinase inhibitor called vactosertib, currently tested in early-stage clinical trial for several cancer types (83). VTCN1, also known as B7S1, is also a negative regulator of tumor immunity by various mechanisms such as dampening the anti-tumor Th1 responses (86). Recently, an early-stage clinical study of FPA150, an antibody targeted on B7S1 and other the anti-B7x family members, was started for patients with advanced solid tumors to assess preliminary efficacy of FPA150 alone or in combination anti-PD, as well its safety, tolerability, pharmacokinetics, and pharmacodynamics (87). In addition to immunoinhibitors, the immunostimulators CD86, TNSF13B, and CXCR4, were found to be associated with LOXL3, while TNFRSF9, CXCR4, and TNFSF15 correlated with LOX. CD86, also known as B7-2, is an immune checkpoint molecule of B7 family and binds to CD28 and Cytotoxic T-Lymphocyte Antigen 4 (CTLA-4). Interaction of CD86 with CTLA-4 inactivates T lymphocytes, causing the escape of tumor cells from the immune system. Therefore, immunotherapy using CTLA-4 antibodies might promote T cell activation to help eliminating tumor cells (88). Ipilimumab, an CTLA-4 antibody, is currently approved by Food and Drug Administration (FDA) for HCC treatment in combination with nivolumab. In addition, Tremelimumab, fully human immunoglobulin G2 monoclonal antibody directed against CTLA-4, is also under investigation for HCC treatment (89). TNSF14B, also known as B cell-activating factor of the TNF family (BAFF), together with its receptor, BAFFR, are important in early B-cell homeostasis and regulatory T-cell function (90). BAFF inhibitors have been tested for certain diseases. For instance, belimumab, a fully human monoclonal antibody against BAFF, has been shown to have a modest effect for active systemic lupus erythematosus (SLE), and another BAFF inhibitor Blisibimod is also under investigation for SLE (90, 91). Tabalumab is another BAFF inhibitor, and has been evaluated as a combined therapy with bortezomib for multiple myeloma (90). However, the roles of BAFF inhibitors in HCC still yet to be explored. TNFRSF9, also known as CD137, a surface glycoprotein belonging to a member of the tumor necrosis factor receptor superfamily (TNFRSF). It is expressed on activated T cells that have encountered cognate antigen, activated NK cells, and mature DCs (92). Two clinical trials have been being initiated for two anti- TNFRSF9 monoclonal antibodies urelumab (BMS-663513) and utomilumab (PF-05082566) (93, 94). TNFSF15, also called TNF-like molecule 1A (TL1A), is expressed on multiple immune cells such as DCs and B cells. It binds to DR3 receptor, leading to cell apoptosis by activating the caspase cascade through interaction with TRADD and FADD, and the activation of multiple cell survival signaling pathways including NF-kB, STAT3, JNK, p38 MAPK and ERK (95, 96). TNFSF15 can also suppress endothelial cell proliferation and angiogenesis through the binding of DR3, and this was verified in a mouse xenograft tumor model (97, 98). Moreover, TNFSF15 also can be induced in T cells, macrophages, monocytes, and DCs in response to stimulation with immune complexes, Toll-like receptor ligands, inflammatory cytokines, and T-cell receptor activator (99). Current studies mostly focus on the role of TNFSF15 in inflammatory diseases such as SLE and psoriasis (100), its potential roles in HCC yet to be further investigated. Apart from the above immunostimulators, the CXCR4 was associated with both LOX and LOXL3. It is expressed on various pro- and anti-inflammatory immune cells, especially in macrophages and T cells (101). Multiple drugs targeted on CXCR4 have been under investigation (102). AMD3100, also known as plerixafor (Mozobil), was the FDA-approved CXCR4 antagonist used for peripheral blood stem cell transplantation, but its clinical use in LC and other solid tumors is limited due to its poor pharmacokinetics and toxic adverse effects (102, 103). Recently, other CXCR4 antagonists have been developed. For example, BPRCX807 has been experimentally validated in different HCC models (103), and it deserves further investigation. The MHCs, HLA-DOA, HLA-DPA1, and HLA-DPB1, positively associated with LOXL3 all belongs to MHC Class II molecules (104). MHC class II molecules were found to be expressed by antigen-presenting cells, including antigen-presenting cancer-associated fibroblasts (apCAFs) (105, 106). Therefore, these positive association of the MHC Class II molecules might be indirect evidence of apCAFs in HCC, and drugs targeted on these apCAFs might have potential therapeutic values and future studies are needed for further clarification.

This bioinformatic study also acknowledges some limitations: First, as all data were retrieved from online databases, the results still need to be validated with other experiments and cohorts. Second, as this study was mainly aimed at exploring the potential diagnostic, prognostic, and therapeutic values of the LOX family members in LC patients, the details of their mechanisms were not comprehensively explored. Third, most of the samples on the online databases were HCC, therefore their values on other types of LC still need further investigation. Fourth, meta-analysis found that LOXL4 was associated with poor OS, while results from TIMER and UALCAN did not yield the same conclusion. However, only two studies on the survival effect of overexpressed LOXL4 were found, therefore, more studies are urgently needed to validate its effect of the prognosis for LC patients.

This bioinformatics analysis investigated the expression levels, diagnostic and prognostic values, genetic alterations, PPI network, functional enrichment, tumor microenvironment factors, and potential mechanisms of LOX family members in LC. Our results found that all LOX family members are overexpressed in LC tumors, and LOXL2 is good candidate as a diagnostic marker. LOX and LOXL3 are associated with poor prognosis and carry potential as therapeutic targets. The effect of LOXL4 on survival remains equivocal and prompts more studies. The infiltration of a variety of immune cells and a list of immunomodulators were positive correlated with LOX family members. These results highlight the need to explore the roles of LOX family in the tumor microenvironment and their potential as immunotherapeutic targets.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/supplementary material.

YZ and CS designed the research study. CS, YC, and SM selected and collected the data. CS, YC, SM, and WG analyzed the data. CS, YC, SM, and NK wrote the manuscript. ZZ participated in the statical analysis. SK, YW, WG, YC, JT, CB, NM, YH, CC, and QZ provided critical opinions and revised the manuscript. All authors read and approved the final manuscript.

The Training Program Foundation of Youth Scholars by the First Affiliated Hospital of Anhui Medical University (2020kj24).

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.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.