EGFL6 Promotes Bone Metastasis of Lung Adenocarcinoma by Enhancing Osteoclast Differentiation and Stimulating Cancer Cell Metastasis

Lung cancer is still the most fatal cancer today with approximately 30%-40% of these patients will develop bone metastasis. Lung adenocarcinoma (LUAD) is the most common and aggressive type of lung cancer. The relationship between LUAD and bone metastasis and its underlying mechanism remains unclear. This study proved that epidermal growth factor-like domain multiple 6 (EGFL6) was highly expressed in LUAD specimens of patients, and the expression level was positively correlated with bone metastasis of LUAD. Wnt/β-catenin 37 PI3K/AKT/mTOR 38 , MAPK 39 , Notch 40 In Wnt/β-catenin β-catenin glycogen synthase (GSK3), and polyposis (APC). Phosphorylation GSK3 The effect of EGFL6 on EMT markers (E-cadherin, N-cadherin and Vimentin) in A549 detected by ICC. Over-expression of EGFL6 increases the expression of N-cadherin and Vimentin in cytoplasm but attenuates the expression of E-cadherin while knockdown of EGFL6 displays the adverse. b Western blotting analysis is utilized to detect the levels of p-GSK-3, GSK-3 and β-catenin under the inuence of over-expression and knockdown of EGFL6 in A549 cells. c Western blotting analysis is utilized to detect the levels of p-PI3K, PI3K, p-AKT, AKT, p-mTOR and mTOR under the inuence of over-expression and knockdown of EGFL6 in A549 cells. Results were presented as mean ± SD. P <0.05 was considered statistically signicant. * P <0.05, ** P <0.01, *** P <0.001.


Results
EGFL6 is elevated in LUAD and is associated with bone metastasis. In vivo, implantation of human adenocarcinoma A549 cells with a higher expression of EGFL6 not only increased tumor growth rate but also bone resorption of tibias in nude mice. In vitro, the secretion of EGFL6 from A549 cells increased osteoclast differentiation but had little effect on osteogenic differentiation. To reveal the underlying mechanism, we demonstrated that EGFL6 enhanced osteoclast differentiation through activating nuclear factor-kappa B (NF-κB) and downstream c-Fos/NFATc1 signaling pathways, and in addition promoted the proliferation, migration and invasion of A549 cells through enhancing the epithelial mesenchymal transformation (EMT) and promoting Wnt/β-catenin and PI3K/AKT/mTOR signaling pathways.

Conclusions
We unveil EGFL6 as a predictor in bone metastasis of LUAD and underscore the relevance of EGFL6 as a therapeutic target.

Background
Lung cancer is the leading cause of cancer deaths worldwide, among which lung adenocarcinoma (LUAD) is the most common and aggressive type 1, 2 . The skeletal system is one of the most common metastatic sites in lung cancer patients, accounting for approximately 30%-40% of lung cancer patients 3, Page 4/22 4 . Bone metastasis patients often suffer skeletal-related events (SREs) including pathologic fractures, spinal cord compression, bone marrow aplasia, and hypercalcemia 5 . Among all types of lung cancer, LUAD has the highest incidence of bone metastases 6 . However, the underlying mechanism of bone metastasis in lung cancer, especially in LUAD, remains unclear. Studies were carried out to identify the interaction of lung cancer cells with bone marrow cells. Nakamura ES et al demonstrated that osteoclasts might promote bone metastasis of lung cancer cells by producing macrophage-derived chemokine CCL22, which upregulated receptor activator of the NF-κB ligand (RANKL) in osteoclast-like cells 7 . The expression of bone remodeling cytokines or receptors, such as RANKL, receptor activator of NF-κB (RANK), and osteoprotegerin (OPG), were found to be upregulated in bone metastatic tissues from primary non-small cell lung cancer (NSCLC) patients. In addition, the increased expression of RANKL and OPG was correlated with clinical features of the patients 8 . As a collagen receptor, Discoidin domain receptor-1 (DDR1) was found to associate with poor clinical outcome in lung cancer. In a nude mice model of tumor with bone metastasis, the lack of DDR1 showed decreased metastatic activity and osteolytic lesions 9 . A representative study determined that the activation of the WNT/TCF signaling pathway mediates the metastasis of LUAD to bones and brain 10 . However, the signaling pathway regulates the LUAD metastasis is still largely unknown.
Tumor growth and metastasis require the formation of new blood vessels, which is termed angiogenesis. Epidermal growth factor-like domain multiple 6 (EGFL6) is an exocrine protein that plays an important role in promoting endothelial cell migration and angiogenesis 11 . Ectopic expression of EGFL6 has been shown to promote the invasion of cancer cells in breast, gastric, and nasopharyngeal cancers [12][13][14] . In addition, cumulated evidence proved that EGFL6 exerted a pivot role in the proliferation, migration, and invasion of several tumor cells, such as breast, ovarian, and colorectal cancer cells 15,16 . Clinically, immunohistochemical staining of 150 LUAD specimens showed that high EGFL6 expression was associated with poor prognosis of LUAD, especially in young patients (<69 years) 17 . However, this study focused on the relationship between the survival rate and the expression level of EGFL6 in LUAD. It has not been studied whether there is a relationship between the expression of EGFL6 and the occurrence of bone metastasis.
Here, we proved that EGFL6 was highly expressed in LUAD samples and was positively correlated with bone metastasis and TNM stage. We demonstrated that EGFL6 enhanced osteoclast differentiation and bone resorption through activating NF-κB and downstream c-Fos/NFATc1 signaling pathways but exerted little effect on the formation of osteoblasts. Additionally, EGFL6 also promoted the proliferation, migration and invasion of LUAD cells through epithelial mesenchymal transformation (EMT) and Wnt/βcatenin and PI3K/AKT/mTOR signaling pathways. Therefore, we propose that EGFL6 can not only be used as a predictor of the prognosis of LUAD, but also play a role in promoting LUAD bone metastasis and subsequent bone destruction.

Specimen collection
We collected 30 LUAD tissues and matched adjacent normal tissues from the Biological Resource Center of Enze Medical Center, Taizhou Hospital. The specimens were staged according to the 8th manual of the American Joint Committee on Cancer Staging Manual. The time distribution of patients undergoing lung surgery was from 2004 to 2012, and the clinical data and follow-up information of the patients were fully recorded. This study was approved by the Medical Ethics Committee of Taizhou Hospital, and the informed consent of all patients was obtained.

Immunohistochemistry (IHC)
Para n sections were baked in a 60°C in an incubator at 60°C for 2h before processing, and then taken out and cooled to room temperature. After depara nization and rehydration, the tissue sections were heated with citrate buffer to perform antigen retrieval in an autoclave. Then, 3% H 2 O 2 was added to the slices to block endogenous peroxidase. Normal sheep serum (Gibco) was used to block nonspeci c binding sites. The sections were incubated with anti-EGFL6 antibody (1:100, ab140079, Abcam, MA, USA) overnight. After that, the secondary antibody was added to the sections. Finally, the tissue sections were sequentially treated with hydrochloric acid, hematoxylin counterstain, dehydration and sealing.
These sections were graded by professional pathologists and were based on the staining intensity of and the percentage of positively stained cells. The intensity was divided into four grades: 0 (negative), 1 (weak), 2 (medium) and 3 (strong). The percentage of positive staining cells was graded as follows: 1 (0-25%), 2 (26-50%), 3 (51-75%), and 4 (76-100%). We multiplied the two scores, the nal score falling in the range of 0-6 was de ned as low EGFL6 expression, and 7-12 was de ned as high EGFL6 expression.

Cell culture
The human A549 LUAD cell line was purchased from the Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). The authenticity of A549 cells (RRID: CVCL_0023) has been veri ed using Short Tandem Repeat (STR) pro ling within the last 3 years. A549 cells were cultured in RPMI 1640 medium containing 1% penicillin/streptomycin, 10% fetal bovine serum (FBS) (Gibco, Grand Island, NY, USA). Cells were maintained in an incubator at 37°C with a humidity of 5% CO 2 . All experiments were performed with cells without mycoplasma.

Cell transfection
The siRNA targeting EGFL6 (si-EGFL6) and the paired negative control (si-Ctrl) were synthesized by GenePharma (Shanghai, China). A549 cells were seeded in six-well plates and transfected with 5μl siRNA and 5μl Lipofectamine 2000 (Invitrogen). As for the overexpression of EGFL6, the LV-EGFL6-RNAi (sh-EGFL6) was obtained from GeneChem (Shanghai, China). A549 cells were grown in 24-well plates and transfected with lentivirus and 20μl HitransG according to the manufacturer's instructions.
Cell proliferation assay Cell Counting Kit-8 (CCK-8, Dojindo, Japan) was used to measure cell proliferation ability. The cells were cultured into 96-well plates, and 10μl CCK8 was added to each plate at 0, 24, 48, 72, 96h. Then, after 2h incubation at 37°C, the determination of cell proliferation was replaced by the absorbance value at 450nm.

Colony formation assay
The transfected cells were seeded in 6-well plates at a density of 1000 cells per well. After culturing for 10d, the colonies were xed with 4% paraformaldehyde for 30min and then stained with 0.1% crystal violet for 30min. The number of colonies was counted and compared in each group.
Migration and invasion assay A549 cells were transfected on the 24-well transwell plates (Corning Inc., Corning, NY). For migration assays, a serum free cell suspension (100μl) containing 5×10 4 cells was added into the upper chamber, and medium containing 10% FBS was added to the lower chamber. After incubating for 12h at 37°C, the cells at the bottom of the upper chamber were xed with 4% paraformaldehyde for 30min and stained with 0.1% crystal violet for 30min. Then, a cotton swab was used to remove the non-migrating cells in the upper chamber. For the invasion assay, the upper chamber was pre-coated with Matrigel (BD Biosciences, San Jose, CA), and the cell number of the cell suspension was changed to 1×10 5 . The remaining steps were the same as the migration assay. Finally, we counted and compared the number of cells of each group.

Wound healing assay
The wound healing assay was performed by culturing insert (ibidi GmbH, Martinsried, Germany) in 24well plates. The cell suspension (70μl) at a density of 5×10^4 cells/ml was placed into each well of the culture-insert. After incubating at 37°C and 5% CO2 for 24h, we used sterile tweezers to gently remove the culture-insert and photographed under an optical microscope. Then, we added serum free RPMI 1640 medium (300μl) to each well and cultured for 24h. Time lapse images were captured at the same position at 0 and 24h. Immunocytochemistry A cell suspension containing 1×10 4 cells was seeded on a coverslip in 12-well plates. After the cells were attached to the coverslips, they were xed with 4% paraformaldehyde for 30min. The cells were permeabilized in 0.5% Triton X-100 for 10min, then blocked with 10% FBS for 30min. The primary antibodies E-cadherin, N-cadherin and Vimentin (CST, Beverly, MA, USA; diluted 1:250) were used to incubate with the cells overnight at 4°C. On the next day, the cells were cultured with the uorescently labeled secondary antibody (diluted 1:250) for 1h in the dark. Later, the cells were stained with DAPI for 5min and then mounted. Images were captured under confocal microscope.
RNA isolation, reverse transcription and qPCR Total RNA was extracted from transfected A549 cells by TRIzol reagent (Invitrogen) following the manufacturer's instructions. Then, the total RNA was reverse transcribed into cDNA using the HiFiScript cDNA Synthesis Kit (CWBIO, Beijing, China). Quantitative real-time PCR analysis was performed by an ABI 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) with ChamQ Universal SYBR qPCR Master Mix (Vazyme, China) in triplicate. The GAPDH gene was utilized as housekeeping gene. All sequences were listed in Table S1.

Western blotting analysis
Total cellular protein of transfected cells was extracted using Radio-Immunoprecipitation Assay buffer (RIPA) containing protease inhibitor PMSF (100:1) and phosphatase inhibitor (Beyotime, Shanghai, China) (100:1). Quantitative of protein concentration was detected by BCA assay (Beyotime, Shanghai, China). The protein samples at the same quantity were separated by 10% SDS-PAGE and then transferred to PVDF membranes (Bio-Rad, Hercules, CA, USA). After blocking with Western Blocking Buffer (Biosharp, China) for 2h, the PVDF membranes were incubated with primary antibodies at 4°C overnight. The membranes were washed by TBS-Tween and incubated with HRP-conjugated secondary antibodies for 1h at room temperature. Chemiluminescent HRP Substrate (Millipore Corporation, Billerica, MA, USA) and ImageQuant LAS 500 (GE Health Care, Fair eld, CT, USA) were used to observe the immunoreactivity.
ImageJ software was utilized to quantify the intensity of the bands.

Quantitative analysis of medium EGFL6 level
An ELISA assay was used to verify that EGFL6 play a role through exocrine. The culture medium of the cells cultured for 2d was extracted and the concentration was detected by an ELISA kit (AMEKO, China). The optical density was tested at 450nm using the Multiskan FC Microplate Photometer (Thermo Fisher Scienti c, Waltham, MA, USA). The standard curve with OD value as the ordinate and standard concentration as the abscissa. The concentration of EGFL6 in each medium could be obtained by the measured OD value against the standard curve.

Osteoclast differentiation in vitro
BMMs were extracted from the femur and tibia of C57BL6 mice (8-week-old). After culturing for 4d or reaching 90% con uence in a 10cm cell culture dish, the BMMs were seeded into a 96-well plate at a density of 8000 cells/well in complete α-MEM containing 30ng/ml macrophage colonies stimulating factor (M-CSF). After 24h, we replaced the medium with complete α-MEM containing 30ng/ml M-CSF, 50ng/ml RANKL and supplemented with the culture medium of si-Ctrl, si-EGFL6, sh-Ctrl and sh-EGFL6 respectively. The osteoclast induction medium was changed every 2d until the 7th day. The cells were xed by 4% paraformaldehyde for 1h, and then stained by TRAP activity kit (Sigma-Aldrich, St. Louis, MO, USA). ImageJ software was used to quantify the area (percentage of a well) of TRAP-positive cells (≥3 nuclei).

Osteoblast differentiation in vitro
Bone marrow mesenchymal stem cells (BMSCs) were extracted from the femur and tibia of C57BL6 mice (8 weeks old), and cultured in complete α-MEM for 6d or reached 90% con uence in a 10cm cell culture dish. For osteoblast differentiation, BMSCs were re-seeded into a 12-well plate at a density of 1×10 4 cells/well with osteogenic medium (Complete DMEM supplemented with 10mM β-glycerophosphate, 50μM ascorbic acid and 100nM dexamethasone). The medium was changed every 2d. On day 7 and 21, the cells were stained with alkaline phosphatase (ALP), and the mineralized bone nodules were stained with Alizarin Red.

Bone absorption assay in vitro
Bone marrow mononuclear macrophage (BMMs) were seeded into 100μm bovine bone slices (Rongzhi Haida Biotech Co., Ltd, Beijing, China) at a density of 1×10^4 cells/well in α-MEM supplemented with 100 ng/ml RANKL and the culture medium of si-Ctrl, si-EGFL6, sh-Ctrl and sh-EGFL6 respectively. The osteoclast induction medium was changed every 2d until the 14th day. The cells were eliminated by mechanical agitation and sonication. Then, the resorption pits were examined by scanning electron microscope (Field Environmental Instruments Inc., Hillsboro, OR). ImageJ software was utilized to quantify the area of resorption pits.

Nude mouse tumor-induced bone destruction model in vivo
All animal experiments followed the guidance of the Institutional Animal Ethics Committee of Taizhou Hospital. 6-week-old female nude mice were randomly divided into si-Ctrl, si-EGFL6, sh-Ctrl and sh-EGFL6 group (6 mice/group). After induction of anesthesia, 20μl of transfected tumor cells (1×10^6 cells/ml) suspended in Phosphate Buffered Saline (PBS) were injected into the tibial marrow cavity through the tibial plateau of nude mice. All mice were euthanized on day 14. We removed the hind limbs of nude mice and compared the size of tumor. Then, the hind limbs were xed in 4% paraformaldehyde and collected for further experiments.

Histological and histomorphometric analysis
After decalci cation in 10% EDTA (ethylenediaminetetraacetic acid) for 14d, the xed hind limbs were embedded in para n and then sectioned. The tissue slices were utilized for H&E and TRAP staining. We acquired images of each sample through a high-quality microscope (Olympus, Japan) and counted the number of TRAP-positive cells in each sample.

Statistical analysis
The signi cant differences between groups were analyzed by Student's t-test or one-way ANOVA or Kruskal Wallis test using GraphPad Prism® version 8. 3

EGFL6 is elevated in lung adenocarcinoma and is associated with bone metastasis
We collected 30 random cases of LUAD samples and clinical data after surgery. IHC analysis was utilized to evaluate the expression of EGFL6 between LUAD tissues and adjacent normal tissues. The results showed that, in general, the expression of EGFL6 in LUAD tissues was higher than that in neighboring normal tissues (Fig. 1A, B). Through statistical analysis of clinical parameters, we found that the expression of EGFL6 was closely correlated with the TNM stage and bone metastasis of LUAD (Fig. 1C,  D). However, the expression of EGFL6 was not signi cantly related to the patients' age, histological grade, and lymph node metastasis ( Table 1). Effects of high and low expression of EGFL6 on the proliferation, migration and invasion of LUAD cells Through the analysis of clinical samples, we speculated that EGFL6 played a role in promoting the progression of LUAD. In order to explore the biological role of EGFL6 in LUAD, we used human LUAD cell line A549 cells for further study. A549 cells obtained overexpression (sh-EGFL6) and knockdown (si-EGFL6) phenotypes under lentivirus and siRNAs respectively. EGFL6-siRNA and null vector were applied to transfect A549 cells. WB and RT-PCR veri ed that the knockdown e ciency of EGFL6-siRNA-899 was the highest (Fig. S1A, B). the lentivirus was transfected into A549 cells and veri ed by uorescence (Fig.  S1C). Finally, we proved the knockdown and overexpression e ciency by WB and RT-PCR (Fig. S1D, E).
We performed CCK8 assays to test the relationship between the proliferation ability of A549 cells and the expression of EGFL6. The enhanced expression of EGFL6 signi cantly promoted the proliferation of A549 cells, while silencing EGFL6 weakened the viability of A549 cells ( Fig. 2A). Besides, the colony formation assays also showed the same trend with CCK-8 (Fig. 2B). In order to explore the migration and invasion abilities mediated by EGFL6, we conducted a transwell assay and a scratch test in vitro. As shown in Fig. 2C, the number of migrated and invading cells in EGFL6 overexpression group was signi cantly greater than that in the control group. On the contrary, the EGFL6 knockdown group had the opposite phenomenon. In addition, the cell scratch experiment found that the healing speed of the EGFL6 overexpression group was signi cantly faster than that of the control group, and depleting EGFL6 expression slowed the healing rate to an extent (Fig. 2D).
High expression of EGFL6 promotes EMT of lung adenocarcinoma cells EMT is an important biological process for LUAD cells to acquire the ability of migration and invasion. It  (Fig. S2). The WB results were consistent with RT-PCR results (Fig. 2E).
To further prove that EGFL6 can promote the EMT of A549 cells, we performed immunocytochemistry (ICC) to detect the expression of E-cadherin, N-cadherin and Vimentin in the si-Ctrl, si-EGFL6, sh-Ctrl and sh-EGFL6 groups. The results showed that the expression of N-cadherin and Vimentin in the sh-EGFL6 group was higher than that in the sh-Ctrl group, while the expression in the si-EGFL6 group was lower than that in the si-Ctrl group. The expression of E-cadherin was opposite to the above results (Fig. 3A).

EGFL6 partially regulates LUAD cells by activating Wnt and PI3K/AKT/mTOR signaling pathways
Wnt/β-catenin signaling pathway and PI3K/Akt/mTOR signaling pathway both play key roles in regulating the process of proliferation, migration and invasion of tumor cells. Through Western blotting, we observed that overexpression of EGFL6 increased the levels of GSK-3 and β-catenin compared to the control. Silencing EGFL6 signi cantly downregulated the levels of these signaling proteins (Fig. 3B). Moreover, the forced expression of EGFL6 increased the phosphorylation levels of PI3K, AKT and mTOR, while the low expression of EGFL6 did the opposite, with no difference in total protein level (Fig. 3C). Taken together, these evidences indicated that EGFL6 promoted the proliferation, migration and invasion of A549 cells at least in part through Wnt/β-catenin signaling pathway and PI3K/AKT/mTOR signaling pathway.

EGFL6 increases tumor growth and bone destruction in bone metastasis of LUAD cells in nude mice
To further validate the role of EGFL6 in promoting the migration and invasion of A549 cells, we established a bone metastasis model by implanting LUAD cells in nude mice. The suspensions of A549 cells of EGFL6 overexpression, knockdown, and their respective control were injected into nude mice through the tibial plateau. After 14d, all nude mice were euthanized, and the differences in tumor appearance size and bone destruction in each group were compared. The tumor growth rate of the sh-EGFL6 group was much faster than that of the sh-Ctrl group, while the si-EGFL6 group exhibited the opposite result (Fig. 4A).
In addition, we observed by X-ray that the sh-EGFL6 group had more severe tibial bone destruction than the sh-Ctrl group, while the si-EGFL6 group showed the opposite result (Fig. 4B). Then we performed para n sections on the tibia of nude mice and stained them with H&E and TRAP. H&E analysis supported that the osteolysis in sh-EGFL6 group was more severe than that of sh-Ctrl group. However, the opposite result was observed in the si-EGFL6 group compared with the si-Ctrl group (Fig. 4C). In addition, based on X-ray and H&E results, TRAP staining proved that osteoclasts in the bone surface were enhanced in the sh-EGFL6 group, but decreased in the si-EGFL6 group (Fig. 4D).
Since the results of animal models showed that the bone destruction in the EGFL6 overexpression group was more serious, in order to study the effect of EGFL6 on bone formation, we extracted the culture medium of each cell group and added them to the BMM and MC3T3 cells. Compared with the sh-Ctrl and si-Ctrl groups, BMM using sh-EGFL6 medium showed enhanced osteoclast differentiation (Fig. 4E), and MC3T3 using sh-EGFL6 medium had no signi cant change in osteogenic formation (Fig. 4F). According to the existence of the above-mentioned cell experiment, the in vitro bovine bone section resorption test showed that the bone resorption capacity of the sh-EGFL6 group was more potent than that of the control, and vice versa (Fig. 4G).
EGFL6 promotes osteoclast differentiation by activating NF-κB and downstream c-Fos/NFATc1 signaling pathways In order to clarify the potential mechanism of EGFL6 promoting osteoclast differentiation, we studied several classic signaling pathways in osteoclastogenesis by Western blotting. We rst divided BMMs into two groups. One group was treated with RANKL and a partial medium extracted from sh-EGFL6 (EGFL6 medium) to explore the levels of c-Fos and NFATc1. The other group was stimulated with RANKL and the medium extracted in sh-Ctrl for 0, 1, 3, and 5d. The results showed that under the stimulation of RANKL and control medium, the expression of c-Fos and NFATc1 reached the peak on the 5th day. In addition, when RANKL and EGFL6 medium were used together, the expression of c-Fos and NFATc1 was signi cantly enhanced on the 5th day (Fig. 5A).
Besides, we further studied the short-term signaling pathways of MAPKs, NF-κB and AKT under the in uence of EGFL6 medium and control medium, respectively. Under the pretreatment of EGFL6 medium or control medium, BMMs was induced by RANKL, and the phosphorylation of each protein in the MAPK, NF-κB and AKT signaling pathways was detected at 0, 5, 10, 20, 30 and 60min level. As shown in Fig. 5B, there was no signi cant difference in the activation and phosphorylation levels of AKT and MAPK members ERK, JNK and p38 between BMMs pretreated with EGFL6 medium and BMMs pretreated with control medium. In contrast, when induced by EGFL6 medium, the phosphorylation of NF-κB member IκBα increased at 10 min. Then, the phosphorylation of p65 in the trimeric structure was therefore similarly enhanced at 10min (Fig. 5C).

Discussion
It is well accepted that EGFL6 plays an important role in the formation of new blood vessels of epithelial cells, which is a necessary process for tumor growth and metastasis 11,16,19,20 . In this study, we found that EGFL6 expression was higher in LUAD tissues than that in adjacent normal tissues and EGFL6 staining in LUAD tissues was positively correlated with TNM stage and bone metastasis. Our results were consistent with a retrospective study of LUAD 17 , which indicated that high EGFL6 expression might serve as a marker of poor clinical outcome of LUAD, especially in younger patients 17 . In that study, the focus was mainly on the relationship between the expression level of EGFL6 and the prognosis of LUAD. Our study further discovered that the high expression level of EGFL6 in LUAD tissues was associated with bone metastasis. It has been proven that LUAD is the most common pathological type of lung cancer with a survival rate of 4-17% 21 . Compared with patients without bone metastases, the overall survival of LUAD patients with bone metastases was signi cantly shorter (5.8 months vs 10.2 months, P<0.05) 22 . Therefore, the poor prognosis in EGFL6-positive LUAD patients may be the result of enhanced bone metastasis. The expression of EGFL6 has also been shown to be positively correlated with the invasion, metastasis and poor prognosis of other cancers, such as ovarian cancer, breast cancer, colorectal cancer, gastric cancer, oral squamous cell carcinoma, nasopharyngeal carcinoma and benign meningioma [12][13][14][23][24][25][26] .
We also found that EGFL6 might promote bone metastasis of LUAD by stimulating bone destruction by osteoclasts. Bone is the preferred metastasis site for many tumor cells, such as breast, lung and prostate cancer cells [27][28][29] . We rst veri ed that EGFL6 was secreted to the culture medium by A549 cells. The medium containing a higher concentration of EGFL6 enhanced the osteoclast differentiation of BMMs.
The bone resorption assay also con rmed that EGFL6 signi cantly increased the area of bone resorption pits. Our in vivo experiments demonstrated that tartrate-resistant acid phosphatase (TRAP)-positive osteoclasts were increased in the tibia of nude mice invaded by A549 cells with high expression of EGFL6. Moreover, studies have shown that when osteoblasts and bone metastatic breast cancer MDA-MB-231 cells were co-cultured, they can reduce the secretion of OPG and promote the secretion of monocyte chemotactic protein 1 (MCP1). This effect can be partially blocked by EGFR inhibitors 30 .
EGFL6 may promote bone metastasis and osteolytic lesions in the tumor microenvironment by decreasing OPG expression and increasing MCP1 expression in an EGFR-dependent manner 30 . In our study, the medium containing EGFL6 had little effect on the osteoblast differentiation of MC3T3 cells. In brief, our results suggest that the metastatic cells of LUAD may secret EGFL6 through paracrine, and stimulate bone resorption by stimulating osteoclast differentiation of BMMs.
Furthermore, we explored the underlying mechanism of EGFL6 promoting osteoclast differentiation. After the stimulation of RANKL in BMMs, NFATc1 and c-Fos, which play an essential role in the downstream pathways of osteoclast differentiation, were activated to control the cytokine expression 31,32 . We found that EGFL6 increased the expression of c-Fos and NFATc1 in a time-dependent manner. In addition, the MAPKs, NF-κB and AKT signaling pathways have been called the most classic signaling pathways in osteoclast formation. In the classic pathway of NF-κB, phosphorylation of IκBα induces the release of the RelA(p65)/p50 heterodimer, which causes nuclear translocation of p65 and ultimately leads to the activation of NF-κB 33 . In this study, we found that EGFL6 enhanced the phosphorylation of IκBα and subsequently the phosphorylation of p65, which means that EGFL6 promoted osteoclast differentiation in part by activating NF-κB and downstream c-Fos/NFATc1 signaling pathway. However, MAPKs members ERK, JNK, p38 and AKT were almost unaffected by EGFL6. It has been shown that EGF-like ligands are related to the bone environment. As a member of EGF-like ligands, EGFL6 is preferentially expressed in osteoblast-like cells, but has no effect on osteoblast proliferation and mineralization 11 . In addition, it is well known that osteoblasts are the main source of several cytokines responsible for regulating the differentiation and formation of osteoclasts, such as OPG, RANKL, M-CSF, and MCP1 34,35 .
Studies have shown that EGF-like ligands can stimulate the formation of TRAP-positive osteoclasts by indirectly reducing OPG expression and increasing MCP1 expression, which may contribute to the effect of metastatic breast cancer cells on bone 30 . However, the role of EGFL6 in bone metastasis of LUAD remains to be resolved.
Our in vitro studies have shown that knocking-in EGFL6 promoted the proliferation, migration and invasion of A549 cells, while silencing EGFL6 did the opposite. EMT is an important process for tumor cells to acquire the ability of migration, invasion and anti-apoptotic stimulation. EMT gives tumor cells the characteristics of metastasis, accompanied by the decrease of epithelial markers and the increase of mesenchymal markers 36 . Our study found that the mesenchymal markers N-cadherin and vimentin were signi cantly enhanced in knocking-in EGFL6 cells, but inhibited in EGFL6 silenced cells. This phenomenon indicated that EGFL6 promoted the ability of A549 cells to obtain migration and invasion and metastasize to bone, which might be partly through the enhancement of EMT. Besides, as a complex biological process of tumorigenesis and progression, EMT is regulated by a variety of signaling  [42][43][44][45] . Similarly, we found that knocking-in EGFL6 enhanced the expression of p-PI3K, p-AKT and p-mTOR in LUAD. The above results collectively proved that EGFL6 promoted the proliferation, migration and invasion of A549 cells at least in part through the enhanced EMT process and the activation of Wnt/β-catenin and PI3K/AKT/mTOR pathways. However, due to the characteristics of PI3K signaling pathway in regulating proliferation, migration and metastasis of tumor cells, the existing evidence was still insu cient to prove that EGFL6 drives the EMT of LUAD exactly through Wnt/β-catenin and PI3K/AKT/mTOR pathways. Further experiments remain to be carried out to explore the potential molecular mechanism of EGFL6 to promote LUAD metastasis.

Conclusion
Our ndings provide evidence that EGFL6 is highly expressed in LUAD tissues and is positively correlated with bone metastasis in LUAD. Knocking-in EGFL6 promotes osteoclast differentiation through NF-κB signaling pathway and promotes bone destruction in nude mice. As attested by the experiments in vivo and in vitro, we demonstrated that overexpression of EGFL6 signi cantly enhanced the proliferation, migration and invasion of A549 cells. On the contrary, the low expression of EGFL6 attenuated these capabilities. Further mechanism studies have found that EGFL6 promotes EMT, at least in part through the Wnt and PI3K/AKT/mTOR signaling pathways to promote the above-mentioned characteristics of A549 cells. Therefore, we speculate that EGFL6 tends to be a predictor of bone metastasis in LUAD. Availability of data and materials

Abbreviations
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Con ict of interest
All authors declare no potential con ict of interest.    EGFL6 moderates the effect of A549 cells on skeleton system of nude mice and enhances the osteoclast differentiation of BMMs. a The tumor size in the sh-EGFL6 group is much larger than the sh-Ctrl group, and the si-EGFL6 group exhibited a smaller size than the si-Ctrl group. b Compared with each control, Xray shows that the bone destruction of tibia in sh-EGFL6 group is much severe while in sh-EGFL6 group is slight. c The tibial tumor sections of nude mice were stained with H&E. d The tibial tumor sections of nude mice were stained with TRAP. e EGFL6 enhances the osteoclast differentiation of BMMs. f EGFL6 has no signi cant in uence on osteoblast differentiation of MC3T3. g The bone resorption pits assay.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. TableS1.docx