The pathological signicance of LOXL2 in premetastatic niche formation of HCC and its related molecular mechanism

Background: The mechanisms underlying the contribution of primary tumor to pre-metastatic niche formation remains largely unknown in HCC. We previously reported that the released Lysyl oxidase like 2 (LOXL2) from HCC cells under higher stiffness stimulation facilitated the formation of lung premetastatic niche. Here we further claried the pathological role of LOXL2 in promoting lung premetastatic niche formation in HCC and its relevant molecular mechanism. Methods: Lung pre-metastatic niche nude mouse model and LOXL2-overexpressed xenograft HCC models were respectively developed to validate the signicance of LOXL2 in pre-metastatic niche formation and pulmonary metastasis. Simultaneously, a gel-based cell culture system mirroring lung tissue stiffness in normal and pathological state was constructed to investigate the underlying mechanism of LOXL2-induced pre-metastatic niche formation. Results: LOXL2 by tail vein injection had obvious inducting effects on CD11b + /CD45 + BMDCs recruitment and bronectin expression in lung tissue during the development of lung pre-metastatic niche. Tumor-secreted LOXL2 also promoted the occurrence of pulmonary metastasis and the number of metastasis lesions remarkably in LOXL2-overexpressed xenograft HCC models. Applying a gel-based cell culture system, we discovered that LOXL2 and LOXL2-caused matrix stiffening all upregulated the expressions of MMP9 and bronectin in lung broblasts signicantly, but little effect on MMP2 expression. Moreover, the activation of PI3K-AKT pathway participated in the regulation of LOXL2-upregulated bronectin. Additionally, LOXL2 and LOXL2-caused matrix stiffening also increased the number of adherent HCC cells evidently, as well as the expression of chemokine CXCL2. Clinical data analysis demonstrated that HCC patients in High-LOXL2 group had higher ratio of microvascular invasion (OR=10.563, P=0.005) and tumor recurrence (OR=8.556, P=0.013) than HCC patients in Low-LOXL2 group, also revealing a signicance of LOXL2 in HCC progression and unfavorable outcome Conclusion: Primary tumor-released LOXL2 contributes to the formation of lung pre-metastatic niche and the occurrence of lung metastasis. LOXL2 and LOXL2-caused matrix stiffening work together to induce pre-metastatic niche formation through inuencing matrix remodeling and cell colonization. This study sheds light on the pathological signicance of LOXL2 in lung pre-metastatic niche formation in HCC, also implicates a new intervention approach for HCC metastasis through reversal of lung pre-metastatic niche. cell ow cytometry Simultaneously, the expression of the related molecules to pre-metastatic niche formation were detected in the collected lung tissues immunohistochemistry, immunouorescence and qRT-PCR. (2) Xenograft HCC model A subcutaneous HCC tumor derived from LV-LOXL2-OE-MHCC97H cells was established in nude mice by subcutaneous inoculation, and another subcutaneous HCC tumor from LV-NC-MHCC97H cells was also established as the Small tumor masses with the size of 2×2×2 mm 3 obtained from two subcutaneous HCC tumors were orthotopically transplanted into the livers of new mice. Six weeks later, two groups of xenograft HCC models were formed successfully including LOXL2 overexpression group and the control group (6 nude mice in each group). HCC tissues, lung tissues and serum samples of xenograft HCC model were collected for evaluating the roles of LOXL2 in promoting HCC metastasis.


Background
Although three classical theories including seed and soil hypothesis [1][2][3], circulatory theory [3,4], and the cascade theory of metastatic spread [5] have well explained the metastasis mechanism of solid tumor from different aspects such as non-randomness of metastasis, metastasis pathway, multi-stages of metastasis, they are still unable to answer how primary tumor chooses the target organ and determine the formation of metastatic foci. The hypothesis of pre-metastatic niche, proposed by Kaplan in 2005 [6,7], builds an ideal bridge for the rst time between primary tumor and metastatic lesion in the target organ, and exhibits an obvious theoretical innovation in organotropism metastasis over three classical metastasis theories. Primary tumor-released soluble factors or exosomes enter the blood stream and reach distant target organ to induce a permissive environment before the arrival of metastatic tumor cells, termed the pre-metastatic niche. The induced pre-metastatic niche not only offers a favorable soil environment for the adherence and colonization of disseminated tumor cells, but also accelerates the completion of organ-speci c metastasis. Currently, the existence of pre-metastatic niche has been validated in multiple tumor animal models including breast cancer, colon cancer, pancreatic cancer, melanoma and so on [8,9]. Matrix remodeling, Bone marrow derived cells (BMDCs) recruitment, in ammation, immunosuppressive and nutrient rich microenvironment, identi ed as typical pathological alterations during the development of pre-metastatic niche [10][11][12][13], have gradually become potential indicators to judge the occurrence of this pathological event. Some secreted factors from primary tumor such as CXCL17, G-CSF, OPN, CXCL12, TNF-α, TGF-β, VEGF-A, and PIGF have been characterized to contribute to the formation of pre-metastatic niche through in uencing CD11b + /Gr-1 + myeloid-derived suppressor cells (MDSC)-mediated angiogenesis, inducing in ammation, remodeling extracellular matrix (ECM), altering niche cells, and recruiting immune cells [8,12,14]. Similarly, the released exosomes form primary tumor also reveal to trigger premetastatic niche formation via educating the resident cells in distant target organ [15]. MicroRNA-21-5p in small extracellular vesicles derived from colon cancer cells regulates macrophage polarization in the liver niche to drive an in ammatory pre-metastatic niche and then promote colon cancer liver metastasis [16]. In addition to tumor cells, other cells in tumor microenvironment like macrophages also participate in pre-metastatic niche formation in colorectal cancer liver metastases by producing CXCL1 to recruit MDSCs with CXCR2(+) [17]. Recently, some hypoxia-induced factors like Lysyl oxidase (LOX) present stronger abilities in inducing the formation of pre-metastatic niche [18], indicating an important pathological role of biophysical cue in pre-metastatic niche formation. Increased matrix stiffness is a remarkable physical property of Hepatocellular carcinoma (HCC) and its clinical signi cance has been highlighted in prediction of HCC progression and prognosis [19][20][21]. Other studies in vitro also suggest that higher matrix stiffness strengthens malignant characteristics of HCC cells and facilitates HCC invasion and metastasis by upregulating invasion/metastasis-associated gene expression [22,23], enhancing stemness characteristics [24], triggering the occurrence of epithelial-mesenchymal transition [25,26], promoting therapeutic resistance [27]. However, little effort has been dedicated to exploring the relationship between matrix stiffness and pre-metastatic niche in HCC. Our previous research demonstrated that the secreted Lysyl oxidase like 2 (LOXL2) from HCC cells under higher stiffness stimulation educated lung broblast and recruited BMDCs to induce lung pre-metastatic niche formation in vitro [28]. Here we further clari ed the pathological roles of LOXL2 in premetastatic niche formation in two animal models and HCC tissues, simultaneously applied an in vitro system of mechanically tunable gel simulating lung tissue stiffness to investigate its underlying mechanism.

Materials And Methods
Cells and cell culture MHCC97H was established at the Liver Cancer Institute of Zhongshan Hospital, Fudan University. Human embryo lung broblasts (HELFs) were obtained from the Cell Bank of Shanghai Institute of Biochemistry and Cell Biology, CAS. MHCC97H cells and HELF cells were all cultured in Dulbecco's Modi ed Eagle's Medium (DMEM, Gibco) supplemented with 10% fetal bovine serum (FBS, Biowest) and 1% penicillin/streptomycin (Gibco).
Preparation of HCC cells with LOXL2 overexpression and their CM-LV-LOXL2-OE MHCC97H cells with LV-LOXL2-OE and MHCC97H cells with LV-scramble were constructed by lentiviralmediated RNA overexpression technology. After the transfected cells were grown in serum-free DMEM for 24h, their culture supernatants including CM-LV-LOXL2-OE and CM-LV-NC were collected, respectively. The detailed experiment procedures were same as the method reported previously [28].
Animal models (1) Nude mouse model of pre-metastatic niche established by tail vein injection of CM-LV-LOXL2-OE. BALB/C nude mice (male, 4 weeks old) were purchased from SLAC Laboratory Animal Co., Shanghai, China. All the nude mice were divided into 4 groups such as group 1 (CM-LV-NC), group 2 (CM-LV-LOXL2-OE), group 3 (Tumor/CM-LV-NC) and group 4 (Tumor/CM-LV-LOXL2-OE). Approximate 2×10 7 MHCC97H cells were inoculated subcutaneously into the right ank of each mouse in group 3 and group 4 to form subcutaneous tumor. On the 7 th day after the inoculation, all the nude mice began to receive tail vein injection. The nude mice in group 1 and 3 were injected with CM-LV-NC (200 μl/d, LOXL2 <0.0002μg/d), and group 2 and 4 with CM-LV-LOXL2-OE (200 μl/d, LOXL2=0.045μg/d). Fifteen days later, the injection volume was changed to be 350μl/d. The injection was performed every other day. On the 14 th , 22 nd and 24 th day after injection, the mice were executed by euthanasia. Their lungs were douched slowly from the right ventricle with D-Hank's equilibrium solution. When the color of the lung turned to pale, the lung was collected and cut into small pieces of 1-2mm 3 . Subsequently, they were dissociated into single cell suspension in buffer containing type 1 collagenase (2mg/ml) using a GentleMACS Octo Dissociator (Miltenyi Biotec). The collected cells were ltered using a cell strainer with 70-μm (Millipore) and treated in erythrocyte lysis buffer (Beyotime). After washed three times with PBS, the puri ed primary cells from lung tissue were used for ow cytometry analysis. Simultaneously, the expression of the related molecules to pre-metastatic niche formation were detected in the collected lung tissues by immunohistochemistry, immuno uorescence and qRT-PCR. (2) Xenograft HCC model A subcutaneous HCC tumor derived from LV-LOXL2-OE-MHCC97H cells was established in nude mice by subcutaneous inoculation, and another subcutaneous HCC tumor from LV-NC-MHCC97H cells was also established as the control. Small tumor masses with the size of 2×2×2 mm 3 obtained from two subcutaneous HCC tumors were orthotopically transplanted into the livers of new mice. Six weeks later, two groups of xenograft HCC models were formed successfully including LOXL2 overexpression group and the control group (6 nude mice in each group). HCC tissues, lung tissues and serum samples of xenograft HCC model were collected for evaluating the roles of LOXL2 in promoting HCC metastasis.
All animal care and experiment approved by Animal Care Ethical Committee of Fudan University (Permit Number: 2016-120) met the requirements of the guideline for the Care and Use of Laboratory Animals published by the National Academy of Science.
In vitro system of mechanically tunable gel simulating lung tissue stiffness Taking use of dual biological function of LOXL2 in lung premetastatic niche formation and matrix crosslinking/tissue stiffness, we mixed COL1 (3.58 mg/ml, Corning), basic DMEM and CM-LV-LOXL2-OE (310 ng/ml) in 1:10:5 to construct high stiffness substrate mirroring the stiffness of lung premetastatic niche. Brie y, 2 ml of the mixture was rst prepared, and then placed into a 6 well plate and incubated in 37℃ incubator for 1.5 hours, subsequently the gel substrate was placed at 4℃for another 12 hours and form a gel-based culture platform with stiffness of 926.18 Pa. Simultaneously, we prepared a low stiffness substrate by mixing COL1 and basic DMEM in 1:20 as the control substrate (202.43 Pa) to mirror the stiffness of lung tissue. HELF cells were grown on different stiffness substrates mentioned above and treated with CM-LV-LOXL2-OE (310 ng/ml). Stiffness of the gel substrate was measured by TA.
Detection of CD11b + /CD45 + BMDCs in lung tissue by Flow Cytometry (FCM) Approximate 1×10 6 of the puri ed primary cells from lung tissue were prepared and resuscitated?? with 100ul of 1% BSA to block unspeci c binding, then they were incubated with FITC-conjugated CD11b antibody (1:100, BD Pharmingen) and PE-conjugated CD45 antibody (1:100, BD Pharmingen) at 4℃ for 30 minutes, ultimately the treated cells were subjected to ow cytometry analysis. Unstained cells, CD11b single-stained cells, and CD45 single-stained cells were set as gating in the experiment.
The levels of serum LOXL2 in xenograft HCC models detected by Enzyme-linked immunosorbent assay (ELISA) The levels of serum LOXL2 in xenograft HCC model were measured by ELISA (human LOXL2 ELISA Kit, Boster China) in accordance with the manufacturer's instructions. Brie y, 100 μl of serum sample was added into each well and incubated at 37℃ for 90 min, then reacted with 100 μl biotin-labeled antihuman LOXL2 antibody at 37℃ for 60 min. Subsequently, the captured antigen-antibody complex was further incubated in 100 μl ABC working solution for 30 min at 37℃. Finally, substrate solution was used for color rendering. The absorbance was detected by a microplate reader (Thermo) at 450nm wavelength.

Cell adhesion assay
When they were grown on the gel substrate in each well to reach 60% con uence, lung broblasts were treated separately with CM-LV-LOXL2-OE and CM-LV-NC. Twenty-four hours later, 2×10 5 GFP-labeled HCC cells were evenly seeded into the well. After 24 hours culture, the attached HCC cells on the surface of HELF cells were photographed using a uorescent microscopy (Olympus). The average number of attached HCC cells was measured by counting six random microscope elds (×200).

qRT-PCR
The detailed procedure of qRT-PCR was same as the methods described previously [28]. The primer sequences of the target genes are listed in Table1.
Table1. Primer sequence used for qRT-PCR

Results
The roles of LOXL2 in the formation of lung pre-metastatic niche in nude mouse Previously, we reported that the released LOXL2 from HCC cells under higher stiffness stimulation educated lung broblast and recruited BMDCs to promote lung pre-metastatic niche formation in vitro [28]. Here we further validated the pathological signi cance of LOXL2 during the formation of premetastatic niche in animal model. We injected CM-LV-LOXL2-OE into tail vein, which simulated primary tumor-released TDSFs, to induce lung pre-metastatic niche formation in nude mouse. Based on the state of BMDCs recruitment in lung tissues at different time points, we found that CM-LV-LOXL2-OE injection resulted in an obvious increase in the proportion of CD11b + / CD45 + cells in lung tissues, regardless of the presence of subcutaneous tumor (Fig.1B), highlighting a signi cant role of blood LOXL2 in lung BMDCs recruitment. Simultaneously, on the 24 th day, lung BMDCs recruitment reached a peak value under CM-LV-LOXL2-OE induction, suggesting that tumor-derived LOXL2 present strong ability to induce lung premetastatic niche formation successfully in nude mouse. Analysis of double staining immuno uorescence also demonstrated that the number of cells with CD11b + /CD45 + in lung tissue section of the injected mice with CM-LV-LOXL2-OE was much more than that of the injected mice with CM-LV-NC (Fig.1C), in accordance with the results of FCM analysis. Considering that bronectin production increase implicates occurrence of matrix remodeling in lung tissue, we continued to detect the expression of bronectin in the lung tissues at both mRNA and protein level, and found that lung tissues of the injected mice with CM-LV-LOXL2-OE all exhibited higher expression in bronectin as compared with the controls (Fig.1D-E), revealing that LOXL2 can increase the production of bronectin in the lung tissues and in turn in uence lung matrix remodeling. Additionally, the injected LOXL2 also showed to reach the lung tissue in immunohistochemistry analysis (Fig.1A). Taken together, the released LOXL2 from HCC cells participated in the formation of lung pre-metastatic niche by promoting BMDCs recruitment and matrix remodeling.
LOXL2 overexpression facilitated lung metastasis of xenograft HCC model Subsequently, we established two groups of xenograft HCC nude mouse models using LV-LOXL2-OE-MHCC97H cells and LV-NC-MHCC97H cells to clarify the roles of LOXL2 in growth and metastasis of HCC. As shown in Fig.2A, there was no signi cant changes in the size of xenograft HCC tumors between the two groups, indicating that LOXL2 has little effects on the growth of xenograft HCC tumors. High expression of LOXL2 in tumor tissues (Fig.2B) and high level of LOXL2 in serum (Fig.2C) in LOXL2overexpressed xenograft HCC model group all supported that LOXL2-overexpressed xenograft HCC models were successfully developed and tumor-secreted LOXL2 was released into the blood. Importantly, the pulmonary metastasis ration of LOXL2-overexpressed xenograft HCC models (6/6) was obviously higher that of the controls (2/6) (Fig.2F), and the number of metastasis lesions in LOXL2-overexpressed xenograft HCC models (19 sites) was much more than that in the control group (7 sites). These results suggest that LOXL2 derived from primary tumor can promote lung metastasis in HCC remarkably, in agreement with the induction roles of LOXL2 in the formation of pre-metastatic niche. Meanwhile, we analyzed the state of BMDCs recruitment in lung tissues and found that the number of cells with CD11b + /CD45 + was increased evidently in LOXL2-overexpressed xenograft HCC models (Fig.2D), consistent with the results of nude mice model of pre-metastatic niche, further suggesting that the increase of CD11b + /CD45 + cell number is closely related to lung metastasis. Combining the results of pre-metastatic niche models and xenograft HCC models, we concluded that LOXL2 derived from primary HCC tumor participated in the formation of lung pre-metastatic niche and ultimately promoted the completion of lung metastasis in HCC.
LOXL2 and LOXL2-caused matrix stiffening participate in pre-metastatic niche formation through in uencing matrix remodeling and cell colonization The stiffness of lung tissue (approximate 200 Pa) is far lower than that of Petri dish (2-4 GPa) [30], thereby, the biological characteristics of cells grown on Petri dish are not able to exactly re ect the real pathological and physiological states of the resided cells in lung. Besides, matrix remodeling itself in pathological state also means the changes of matrix stiffness. Tumor-released LOXL2 can not only induce the formation of lung premetastatic niche formation by remodeling matrix and recruiting BMDCs [28], but also promote matrix crosslinking and increase matrix stiffness [31]. We here speculated that LOXL2 and LOXL2-induced matrix stiffening all participated in lung pre-metastatic niche formation. We respectively constructed high stiffness substrate (H, 926.18Pa, Fig. 3A  we observed the morphological characteristics of lung broblasts and explored the effects of LOXL2 and LOXL2-caused matrix stiffening on the expressions of matrix remodeling-associated genes, and chemokine as well as the activation of related pathways in lung broblasts for better understand premetastatic niche formation. Lung broblasts (HELFs) grew well on high or low stiffness substrates. Lung broblasts grown on substrate H exhibited an extension and expansion morphology, whereas on substrate L, they presented an appearance of round and small. Simultaneously, on same stiffness substrate, the size of lung broblasts treated with CM-LV-LOXL2-OE was more extended than that of the cells treated with CM-LV-NC (Fig.3B), implying that LOXL2 and LOXL2-caused matrix stiffening can obviously alter the morphology of lung broblasts and further in uence their biological function.
The expression levels of matrix metalloproteinnases (MMPs) and bronectin generally indicate the state of matrix remodeling in lung tissue [6,32]. We further analyzed the expression levels of MMP9, MMP2, and bronectin in lung broblasts grown on substrate H and L in the presence of CM-LV-LOXL2-OE or CM-LV-NC. MMP9 and bronectin highly expressed in lung broblasts grown on substrate H compared with the cells on substrate L under same CM stimulation (Fig3C (a, c), Fig 3D (a, b)), demonstrating that LOXL2-caused matrix stiffening participates in regulation of matrix remodeling. Meanwhile, on the same stiffness substrate, CM-LV-LOXL2-OE stimulation signi cantly upregulated the expressions of MMP9 and bronectin in lung broblasts ( Fig. 3C (a, c), Fig 3D (a, b)), supporting the roles of LOXL2 in lung matrix remodeling. However, there was no signi cant changes in the expression of MMP2 among the four groups of cells (Fig. 3C (b), Fig.3D (a)), in consistence with the results of our previous work [28], indicating that MMP2 is not involved in the process of LOXL2-induced matrix remodeling. On the other hand, CM-LV-LOXL2-OE stimulation remarkably upregulated bronectin expression and the phosphorylation level of AKT (Fig.3D (b)) in lung broblasts grown on substrate H and L. A recombinant protein LOXL2 (rhLOXL2) also showed same effects on the expression of bronectin and p-AKT (Fig.3D  (c)), indicating a linkage between PI3K-AKT pathway and LOXL2 expression. We applied PI3K inhibitor LY294002 to inactive phosphorylation of AKT, and found that it sharply attenuated the expression of bronectin and p-AKT in lung broblasts grown on substrate H (Fig.3D (d)), validating that PI3K-AKT pathway participates in LOXL2-induced bronectin upregulation.
Premetastatic niche created a suitable "soil" for colonization of circulating HCC cells. We used a modi ed cell adhesion assay to evaluate whether LOXL2-induced lung matrix remodeling was suitable for cell adhesion and colonization of HCC cells. No matter the lung broblasts were grown on high or low stiffness substrate, the numbers of adherent HCC cells on surface of lung broblasts treated with CM-LV-LOXL2-OE were all signi cantly higher than those of CM-LV-NC group (Fig.3E). Additionally, LOXL2caused matrix stiffening also promoted the number of adherent HCC cells evidently (Fig.3E) These results indicates that LOXL2-induced matrix remodeling and LOXL2-caused matrix stiffening all assist HCC cell adhesion and colonization, ultimately promote the completion of lung metastasis. Except that, chemokine CXCL2 highly expressed in lung broblasts grown on substrate H treated with CM-LV-LOXL2-OE, which will be bene cial to the chemotaxis of HCC cells to site of lung pre-metastatic niche. Accordingly, the above data suggest that LOXL2 and LOXL2-caused matrix stiffening all contribute to premetastatic niche formation through in uencing matrix remodeling and cell colonization.

Clinical signi cance of LOXL2 expression in HCC cohort
We rst de ned median expression level of LOXL2 in tumor tissues as a threshold to classify HCC patients into LOXL2 high expression group (High-LOXL2 group) and LOXL2 low expression group (Low-LOXL2 group). By a retrospective analysis of pathological parameters of HCC patients between High-LOXL2 group and Low-LOXL2 group, we found that HCC patients in High-LOXL2 group had higher ratio of microvascular invasion (OR=10.563, P=0.005) and tumor recurrence (OR=8.556, P=0.013) than HCC patients in Low-LOXL2 group ( Table 2). The above results validate a clinical signi cance of LOXL2 upregulation in HCC progression and unfavorable outcome. Given that the expression levels of LOX and COL1 can indicate the grade of liver matrix stiffness, we further strati ed HCC patients into LOX high /COL1 high group (group H) and LOX low /COL1 low group (group L) for analyzing the relationship between LOXL2 and matrix stiffness. HCC patients in High-LOXL2 had higher matrix stiffness (OR=5.76, P=0.038).
As shown in Fig.4 (A-C), the expression of LOXL2 was obviously increased in HCC tissue in LOX high /COL1 high group, as compared with that in LOX low /COL1 low group, and positively correlated with the expression of LOX and COL1, respectively ( Figure.4D), revealing that LOXL2 expression is positively correlated with matrix stiffness, in agreement with the results described previously [28].

Discussion
Metastasis is still the major obstacle to improving therapeutic effect and prognosis of HCC patients. It is undoubtedly of great signi cance to identify the metastasis driver molecules for development of novel treatment strategy against HCC metastasis. Despite some progress has been made in the understanding of tumor metastasis mechanism during the last two decades, how primary tumor in uences and determines the occurrence of metastatic lesion in the target organ remains largely dismal. The theory of pre-metastatic niche, which builds up a bridge between primary tumor and metastatic foci, offers a new research direction for better understand organotropism metastasis [9,33]. The formation of pre-metastatic niche can facilitate the completion of distant metastasis obviously through constructing a favorable soil environment for colonization and growth of circulating tumor cells (CTC) [15], and gradually become an important speed-limiting step at the late stage of metastasis. So far, the identi ed tumor-derived soluble factors are still very limited for triggering pre-metastatic niche formation in HCC. LOXL2 upregulation has been identi ed to promote metastasis of gastric, breast and colorectal cancer [34][35][36], and implicates in less differentiation [37] and poor prognosis [35,38], implying a positive correlation between LOXL2 level and tumor metastasis. Except that, hypoxia-associated LOXL2 and LOX also exhibit a potential induction role in the formation of pre-metastatic niche in HCC and breast cancer [39,40]. High expression of LOXL2 alters tissue stiffness to promote the occurrence of intrahepatic metastasis in HCC, simultaneously enhances recruitment of BMDCs to metastatic site and facilitates completion of extrahepatic metastasis [40]. Our previous research demonstrated that the released LOXL2 from HCC cells under high stiffness stimulation facilitated the formation of lung pre-metastatic niche in vitro [28], highlighting the pathological signi cance of LOXL2 during matrix stiffness-induced pre-metastatic niche formation.
Besides, LOXL2 presents an important biological function in matrix remodeling and tissue stiffening due to its catalyzing ability in the cross-linking of collagen and elastin [41]. Fibronectin deposition and MMPs all appear to be critical factors for remodeling matrix and regulating the formation of lung pre-metastatic niche [6,15,33] On the other hand, bronectin matrix modulates LOX catalytic activity by providing speci c microenvironments [42]. Therefore, in addition to LOXL2 itself, we speculate that LOXL2-caused matrix stiffening may be also involved in the formation of lung pre-metastatic niche.
BMDCs recruitment and matrix remodeling are the most common pathological alterations during the formation of pre-metastasis niche. Firstly, we injected CM-LV-LOXL2-OE into the blood by tail vein simulating primary tumor-released TDSFs and validate the induction roles of LOXL2 in the formation of lung pre-metastatic niche formation. Our results showed that the injected LOXL2 signi cantly enhanced CD11b+/CD45+ BMDCs recruitment and bronectin expression in lung tissue, suggesting that the injected LOX2 can induce the formation of lung pre-metastatic niche successfully. Subsequently, we applied LOXL2-overexpressed xenograft HCC models to further evaluate the the pathological role of LOXL2 in promoting lung metastasis in HCC. Our data demonstrated that high level of serum LOXL2 released from primary tumor obviously promoted the occurrence of pulmonary metastasis and the number of metastasis lesions, meanwhile increased the number of CD11b + /CD45 + BMDCs, in agreement with the induction roles of LOXL2 in the formation of pre-metastatic niche and the results of other study [40], implying that the increase of BMDCs recruitment is closely associated with the occurrence of lung metastasis. Accordingly, the above data in vivo con rmed that LOXL2 derived from primary tumor participated in the formation of lung pre-metastatic niche and accelerate the occurrence of lung metastasis.
To investigate whether LOXL2-caused matrix stiffening also contribute to the formation of lung premetastatic niche, we constructed high and low stiffness substrate gel mirroring lung tissue stiffness in pathological state and normal state to explore the relevant mechanism by which LOXL2-caused matrix stiffening participate in lung pre-metastatic niche formation. We mainly analyzed the effects of LOXL2 and LOXL2-caused matrix stiffening on the expressions of matrix remodeling-associated molecules, and chemokine as well as the activation of related pathways in lung broblasts, and found that both LOXL2 and LOXL2-caused matrix stiffening prominently upregulated MMP9 and bronectin expressions in lung broblasts, supporting that they participates in regulation of matrix remodeling together. Like CM-LV-LOXL2-OE stimulation, recombinant protein LOXL2 intervention also showed same effects on the expression of bronectin and p-AKT, indicating a linkage between LOXL2 expression and PI3K-AKT pathway. PI3K inhibitor evidently suppressed the expressions of bronectin and p-AKT in lung broblasts grown on high stiffness substrate, further illustrating that LOXL2 stimulation promotes bronectin upregulation in lung broblasts via activating PI3K-AKT pathway. Additionally, no signi cant roles were found for MMP2 in LOXL2-induced matrix remodeling, in consistence with the results of our previous work [28] Our results also indicated that LOXL2-induced matrix remodeling and LOXL2-caused matrix stiffening provide a favorable soil environment for adhesion and colonization of HCC cells. Except that, LOXL2 stimulation obviously increased the expression of chemokine CXCL2 in lung broblasts grown on high stiffness substrate, which will be bene cial to the chemotaxis of liver cancer cells to site of lung premetastatic niche. Clinical signi cance of LOXL2 upregulation in HCC progression and unfavorable outcome was validated in a retrospective analysis of pathological parameters in HCC patients between High-LOXL2 group and Low-LOXL2 group. Additionally, the expression of LOXL2 was positively correlated with matrix stiffness, in agreement with in vitro ndings described previously [28].

Conclusion
Primary tumor-released LOXL2 contributes to the formation of lung pre-metastatic niche and the occurrence of lung metastasis. LOXL2 and LOXL2-caused matrix stiffening work together to induce lung pre-metastatic niche formation through in uencing matrix remodeling and cell colonization. This study sheds light on the pathological signi cance of LOXL2 in lung pre-metastatic niche formation in HCC, also implicates a new intervention approach for HCC metastasis through reversal of lung pre-metastatic niche.    Schematic diagram of the mechanism underlying LOXL2-induced lung pre-metastatic formation in HCC.