LncRNA SNHG16 Facilitates Liver Metastases of Colorectal Cancer by Modulating Circulating Tumor Cells (CTCs) Epithelial-Mesenchymal Transition (EMT) via a Positive Feedback Loop with YAP1/TEAD1 Complex

Background: Circulating tumor cells are important precursor of colorectal cancer metastasis, which attributes to the main cause of cancer-related death. The ability to adopt epithelial-mesenchymal transition (EMT) process facilitates CTCs generation, thereby overcoming metastatic bottlenecks and realizing distant metastasis. However, the potential molecular mechanism of CRC EMT remains largely unknown. Methods: RT-qPCR, immunohistochemical staining, and western blot were used to detect the expression of mRNA and protein in CRC. Loss- and gain-of-function approaches were performed to investigate the effect of SNHG16 on CRC cell phenotypes. Function assays, including wounding healing, transwell assay, and clone formation were used to assess the effect of SNHG16 on tumor biological behavior. Then, RNA immunoprecipitation, Chromatin Immunoprecipitation, Co-Immunoprecipitation, GST-pull down, biotin-labeled miR-195-5p pull down, and dual-luciferase assay were performed to uncover the underlying mechanism for molecular interaction. Finally, CRC nude mice xenograft model experiment was performed to evaluate the inuence of SNHG16 on tumor progression in vivo Results: Compared with normal tissue and cell line, SNHG16 was signicantly upregulated in CRC. Clinical investigation revealed that SNHG16 high expression was correlated with advanced TNM stage, distant metastasis, and poor prognosis of cancer patients. According to Loss- and gain-of-function experiment, SNHG16 could promote CRC proliferation, migration, invasion, EMT, mesenchymal-type CTCs ( M CTCs) generation, and liver metastasis through YAP1 in vitro and in vivo. Mechanistic research indicates that, SNHG16 could act as miRNA sponge to sequester miR-195-5p on Ago2, thereby protecting YAP1 from repression and facilitating CRC liver metastasis and tumor progression. Moreover, YAP1 could combine with TEA Domain Transcription Factor 1 (TEAD1) to form a YAP1/TEAD1 complex, which could in turn bind to the promoter of SNHG16 and regulate its transcription. In addition, both of YAP1 and TEAD1 indispensable during this process. we demonstrated that YAP1 signicantly promoted the tumor progression, and SNHG16 could rescue the effect of YAP1 on tumor progression USA). incubation with diluted primary antibodies and HRP-conjugated secondary antibodies, blots were blocked non-fat milk dilution folds Immunoprecipitation Kit (Millipore, Billerica, MA, USA), we performed the RIP to investigate the binding of microRNA-195-5p to lncRNA SNHG16. The anti-Ago2 antibodies (CST, USA) were used for immunoprecipitation, and the species anti-IgG antibodies and total RNA (input controls) were used for control. The co-precipitated RNAs was reverse transcribed to cDNA and detected by qRT-PCR. cell of Lv-anti-SNHG16-NC Lv-anti-SNHG16 (HCT116), Lv-anti-SNHG16/Lv-Oe-YAP1 (HCT116) thereby protecting YAP1 from repression and facilitating tumor progression. Moreover, YAP1 could also combine with TEAD1 to form a YAP1/TEAD1 complex, which could in turn bind with the promoter of SNHG16 to regulate its transcription. Finally, we revealed that YAP1/TEAD1-SNHG16 loop signicantly promoted the tumor progression, and SNHG16 could rescue the effect of YAP1 on tumor progression. Thus, there exists a positive feedback loop between YAP1/TEAD1 and SNHG16. These novel results suggested that SHNG16-YAP1 positive feedback loop may be a potential therapeutic target for CRC.

Currently, accumulating evidence demonstrated that the Hippo signaling pathway plays important roles in cancers' metastasis [34][35][36][37] . As a downstream transcriptional coactivator of the Hippo pathway, the abnormal expression of Yes Associated Protein 1 (YAP1) will promote the malignant tumor's proliferation and metastasis, induce EMT and produce possible drug resistance [38][39][40] . YAP1 in its active form, on the other hand, could function as a transcriptional co-activator predominantly mediated by an interaction with TEAD transcription factors 41 . MiR-195-5p has been demonstrated be a tumor suppressor in various cancers 42 . In a previous study, we also have demonstrated that miR-195-5p could potently suppress the expression of YAP1, and thus inhibit the process of EMT in CRC 43 .
Intriguingly, using bioinformatics, we found same miR-195-5p response elements on SNHG16 and YAP1. Therefore, we wonder whether lncRNA-SNHG16 could interplay with miR-195-5p/YAP1 axis and regulate the state of CTCs' EMT, thereby affecting liver metastasis of colorectal cancer. More importantly, exploring the underlying molecular mechanism of the YAP1-lncRNA-microRNA interplay has great signi cance.
In present study, we found that SNHG16 was upregulated in CRC and signi cantly associated with poor prognosis of CRC patients. Coxregression analysis revealed that SNHG16 may represent an independent prognostic biomarker in CRC. According to Loss-and gain-of function analysis, YAP1 was the most signi cant changed EMT-related transcription factors, and SNHG16 could affect tumor progression through YAP1 in vivo and in vitro. Mechanistically, SNHG16 could act as miRNA sponge to sequester miR-195-5p on Ago2, thereby protecting YAP1 from repression and facilitating CRC liver metastasis. Moreover, YAP1 could combine with TEAD1 to form a complex, which in turn bind with the promoter of SNHG16 to regulate its transcription. Herein, we clari ed a hitherto unexplored positive feedback loop between SNHG16 and YAP1/TEAD1. These nding provides a novel mechanism for CRC liver metastasis, and it may be a potential candidate in the treatment of CRC.

Method And Material
Patient samples This study was approved by the Research Ethics Committee of Wuhan University (Wuhan, Hubei, PR China). Informed consents were obtained from all participating patients. The participating patient must be diagnosed with primary CRC by histopathologic diagnosis and undergone surgeries with complete prognostic information. No local or systemic neoadjuvant radiotherapy, or/and chemotherapy and targeted therapy were managed. One-hundred and eleven human CRC tissues and PANT (distance to cancer > 5 cm) were randomly obtained from patients in the Zhongnan Hospital of Wuhan University between January 2014 and December 2015. Each sample was snap-frozen in liquid nitrogen and then stored at -80 °C.

CTC isolation and identi cation
The isolation and enrichment of CTC was performed by CTCBIOPSY device (Wuhan YZY Medical Science and Technology Co., Ltd., Wuhan, China), which was described in our previous research 44 . According to the manufacture's instruction, we diluted 1 ml mice blood into 5 ml with 0.9% sodium chloride solution, the total liquid was then transferred to ISET tubes with 8 µm diameter aperture membrane. Through the positive pressure from 12 to 20 mmHg in ISET tubes, candidate CTC was adhered to the membrane of ISET tube and identi ed by three-color immuno uorescence.

Cell culture and treatment
Human colon cancer cell lines DLD1, Caco2, SW480, SW620, HT29, lovo, Hct116 and normal intestinal epithelium cell line NCM460 were obtained from the Cell Bank of Wuhan University. Cells were maintained in an incubator at 37℃ and 5% CO 2 . Cells were cultured in DMEM medium (Gibco, USA) Supplemented with 2 mmol/L glutamine and 10% fetal calf serum (Gibco, USA).

Transient transfection and stable transfection
To induce miR-195-5p overexpression and inhibition, cells were transfected with miR-195-5p mimics and inhibitors (RiboBio Co., Ltd) using Lipofectamine 2000. To verify the transfection e ciency, miR-195-5p mimic negative control and miR-195-5p inhibitor negative control were included in each transfection experiment. For SNHG16, a pool of three siRNAs were purchased from genepharma and transfected into cells via lip2000 to select the best knockdown effect.
According to the instruction of manufacture, we get the MOI of HCT116 (MOI = 10) and DLD1 (MOI = 20). For virus infection, virus and polybrene ( nal 7 µg/ml, Sigma Aldrich, Cat#107689) was added to 25% con uent cells. Fresh media was added 16 hours after infection. Media was changed with media containing appropriate antibiotics 48 hours after infection. After Puro selection, cells were maintained for at least one day without drug for further experiments.
Isolation of RNA and performance of qRT-PCR Total RNA was extracted from tissue samples and cell lines using a the Trizol Reagent (Invitrogen, USA) according to the manufacturer's instructions. Nanodrop2000 was used to quantify the concentration of RNA and 1 µg of total RNA was reverse transcribed into cDNA. We performed the cDNA synthesis of mRNA by using Primescript™ RT reagent Kit (Vazyme, Nanjing, China). cDNA synthesis for microRNA detection was carried out using miRNA 1st Strand cDNA Synthesis Kit (Vazyme, Nanjing, China). cDNA was used for subsequent qRT-PCR using the SYBR-Green PCR Master Mix (Vazyme, Nanjing, China). Each reaction was run on the BioRad IQ5 Real time PCR machine (BioRad, USA). Relative expression was calculated using the 2-ΔΔCt method. The primer sequences are listed in Table S1 (supplementary Table 1).

Western blot and antibodies
Cells were lysed using RIPA buffer supplemented with complete proteinase inhibitor cocktail (Thermo Scienti c, USA). Then, the lysate was homogenized using sonication and quanti ed using BCA reaction. Total proteins were separated by SDS-PAGE and subsequently transferred to PVDF membranes (Millipore, USA). Prior to incubation with diluted primary antibodies and HRP-conjugated secondary antibodies, blots were blocked in non-fat milk for an hour. The primary antibodies and dilution folds are as follows: anti-E-cadherin

RNA immunoprecipitation (RIP) assay
According to the instruction of Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, Billerica, MA, USA), we performed the RIP to investigate the binding of microRNA-195-5p to lncRNA SNHG16. The anti-Ago2 antibodies (CST, USA) were used for immunoprecipitation, and the species anti-IgG antibodies and total RNA (input controls) were used for control. The co-precipitated RNAs was reverse transcribed to cDNA and detected by qRT-PCR.
Co-immunoprecipitation CRC cells were lysed in RIPA buffer and the protein concentration of lysates quanti ed using BCA reaction. After impurities removed, the clari ed lysates was incubated with rabbit anti-YAP1 or anti-TEAD1 for overnight end-over-end shaking at 4 °C. Subsequently, we use the Protein A/G sepharose beads to capture the antigen-antibody complexes in 4 °C for 12 h. The operation were followed by the wash of beads with PBS, and the complexes were boiled to separate antigens and antibodies from the beads. After elution from the beads, the expression level of YAP1 and TEAD1 in the IP was analyzed by western blot.

Pulldown assay with biotinylated miRNA
To assay whether miR-195-5p bind to lncRNA SNHG16 or not, a biotin-avidin pulldown system was applied. The biotinylated miR-195-5p was synthesized by RioBio (RiboBio Co., Ltd), we called it miR-195-5p-Bio. Then, the SNHG16 binding sites mutated biotinylated miR-195-5p was also synthesized in RiboBio, termed as miR-195-5p-mut-Bio. We also used the NC-Bio as a negative control to ensure the accuracy of the result, and the operation process are as follows. Put it simply, CRC cells were transfected with biotinylated miR-195-5p by Lipofectamine 2000, and cells were rinsed and lysed in a buffer 48 hours later. Then, the cell lysates were treatment with streptaviden magnetic beads. Then, The bound RNAs were isolated by TRIzol LS reagent for further RT-PCR analysis.

ChIP
We performed ChIP using the SimpleChIP® Enzymatic Chromatin IP Kit (ChIP) Kit (Cell Signaling, #9003, USA), as per the instruction of manufacturer. Put simply, after the cells were crosslinked and lysed, the cells become lysates. Subsequently, the lysates were sonicated into 300 bp to 600 bp fragments and immunoprecipitated by anti-TEAD1 antibody (Abcam). Besides, we used Normal rabbit immunoglobulin G (IgG) as a negative control. Then, we performed the reverse crosslinking and DNA puri cation. Finally, the precipitated DNA was analyzed by qRT-PCR via the SYBR-Green PCR Master Mix (Vazyme, Nanjing, China). Primers used for SNHG16 promoter regions were shown in supplementary material.

Gst-pull down
To test whether YAP1 could interact with TEAD1 directly, we performed the GST pull-down assay as follows. First, after the transformant, cloning and expression of GST-YAP1 fusion protein in Escherichia coli, GST-tagged YAP1 fusion protein was further puri cation using glutathione-Sepharose 4B beads. For pull-down assay, the GST or GST-tagged YAP1 fusion protein were incubated with puri ed His tagged TEAD1 from Escherichia coli for 4-8 h in 4℃. Then, the protein bound glutathione-Sepharose 4B beads were washed four times and the proteins were recovered. Finally, the recovered proteins were analyzed by SDS-PAGE.

Colony formation and wound healing assay
For the colony formation detection, HCT116 were placed in 6-well plates at a density of 1000 cells per well. In the same way, DLD1 were placed at a density of 1000 cells per well. After two week's culture in incubator, the cells were xed with 4% paraformaldehyde and stained with 0.5% crystal violet. A wound-healing assay was also used to measure the migratory capacity of CRC cells. Cells were seeded at 6-well plate and allowed to grow until the con uence reached by 90%.The 20 ul pipette tip was used to scrap several lines across the cell surface, and then plates were washed three times to remove non-adherent debris. Finally, the cells were incubated at 37℃ with serum-free medium and photos were taken per 12 hours.

Xenograft assays
The animal experiment were approved by Zhongnan hospital of Wuhan University ethical committee, and performed in accordance with the Guide for the Care and Use of Laboratory Animals of Wuhan University. To perform tumor proliferation assay, 12 BALB/c nude mice (4-6 weeks old) (Hubei Research Center of Laboratory Animals, China) were randomly divided into three groups (n = 4 per group): NC (HCT116), Lv-anti-SNHG16 (HCT116), Lv-anti-SNHG16/ Lv-Oe-YAP1 (HCT116). In total, 5 × 10 6 HCT116 cells suspended in 200 ul medium were subcutaneously injected into the right ank of mice. After ten days, tumor size was measured by digital Vernier calipers every 1 week, and the volume of tumor was calculated by the following formula: volume = 1/2 x (length x width 2 ). After thirty days, the transplantable tumor and 1 ml mice blood was collected for further research. For metastasis experiment, the mice was randomly divided into three groups (n = 6 per group). The stable cell line were injected into the mice via tail vein, and the metastasis was assessed 30 days after injection by dissection. Finally, the mice were euthanized and necropsied to assess metastatic burden. Then, we further examined the tumor tissues, liver and lung tissues by H&E, IHC staining and RT-PCR assay.

Statistical analysis
All the data derived from at least three times independent experiment, and p values < 0.05 were considered statistically signi cant. Through the SPSS statistical software and GraphPad Prism software, all the statistical analyses were performed. To assess the association between SNHG16 expression and YAP1, and miR-195-5p expression, we used the Pearson's correlation analysis. Then, we applied Chi-square test to analysis the expression of SNHG16 and clinicopathological status of CRC patients. Comparisons between 2 Result SNHG16 is upregulated in colorectal cancer tissues and indicates poor prognosis of colorectal cancer patients Initially we measured the expression of lncRNA-SNHG16 in 45 matched normal colorectal and cancerous colorectal tissues via qRT-PCR. Compared with paired adjacent normal tissue (PANT), lncRNA-SNHG16 was signi cantly upregulated in CRC tissues (Fig. 1A) and correlated with advanced TNM stage. We then measured the expression level of 111 CRC tissues and separated the samples into two groups based on the median value of expression level. To explore whether the expression of lncRNA-SNHG16 correlated with CRC progression, we investigated the correlation between lncRNA-SNHG16 expression and clinicopathological signi cance (Table 1) meanwhile, we detected the highest expression level of SNHG16 in HCT116 and the lowest expression of SNHG16 in DLD1, and thus these two cell lines were selected for further research. Additionally, ISH showed that SNHG16 was mainly located in the cytoplasm of CRC tumor cells, but not in PANT (Fig. 1F). Taken together, these results strongly indicated that SNHG16 upregulated in CRC and was an independent prognostic factor. groups were performed using 2-tailed, unpaired student's t test. In addition, Kaplan-Meier survival curve and log-rank test was used for survival analysis. Finally, univariate and multivariate Cox-regression analyses were applied to identify the independent factors of prognosis.  The ectopic expression of lncRNA-SNHG16 affects proliferation, migration, invasion, and EMT of CRC cells To investigate the role of SNHG16 in tumor progression, the loss-and gain-of-function approaches was applied. First, three sequences of siRNA against SNHG16 were designed and qRT-PCR was used to verify the most e cient sequence ( Fig S1A). Thus, the sequence of si-SNHG16 (si-3) and its negative control were used to construct the lentiviral-based SNHG16 knockdown stable cell line, termed as Lvanti-SNHG16 and Lv-anti-SNHG16-NC, respectively ( Fig. 2A). Then, we constructed a lentiviral-based DLD1 cell line which will stably overexpress SNHG16, termed as Lv-Oe-SNHG16 (Fig. 2B). Finally, the knockdown and overexpression effects were validated by PCR ( Fig. 2A and Fig. 2B).
As shown in Fig. 2C, the SNHG16-downregulated HCT116 displayed fewer clones than control group, indicating that the cell proliferation was inhibited by SNHG16 knockdown. In contrast, the clones were much more numerous in the SNHG16 overexpressed DLD1 than their negative control (Fig. 2D). Hence, the overexpression of SNHG16 promoted the proliferation of CRC cells. According to transwell migration assay, the knockdown of SNHG16 signi cantly reduced the cells crossing the membrane (Fig. 2E), whereas the increased number of cells crossing the membrane occurred after the overexpression of SNHG16 (Fig. 2F). Wound healing assays also revealed the positive effect of lncRNA-SNHG16 on CRC migratory capacity. Evidently, the knockdown of SNHG16 inhibited the migratory speed (Fig. 2G); yet the migratory speed was faster in the SNHG16-overexpressed DLD1 cells than in their control group (Fig. 2H).
Furthermore, matrigel-coated transwell invasion assay con rmed the promotive effect of SNHG16 on CRC invasion ( Fig. 2E and Given the important role of EMT in CRC cellular migration and invasion 45 , we next investigated whether SNHG16 could induce EMT in CRC cells. Western blot and immuno uorescence revealed that, knockdown of SNHG16 could signi cantly reduce the expression level of Vimentin but increase the level of E-cadherin in HCT116 ( Fig. 2I and Fig. 2J). In contrast, SNHG16 overexpression promoted the expression of mesenchymal markers but inhibited the expression of epithelial markers in DLD1 ( Fig. 2I and Fig. 2J).
Altogether, all these results demonstrated that SNHG16 could regulate CRC cellular migration and invasion through affecting the process of EMT.
LncRNA SNHG16 facilitates CRC cellular proliferation, migration, invasion and EMT in a YAP1-dependent manner To explore the mechanism by which SNHG16 regulates EMT in CRC cells, we focused on the potential EMT-related transcriptional factor, which could regulate the progression of CRC 46 . As shown in Fig. 3A and 3B, a panel of transcription factors was screened, and YAP1 showed the most signi cant change among them. Meanwhile, western blot also revealed that SNHG16 could positively regulate the protein level of YAP1 (Fig. 3C), indicating a potential interaction between SNHG16 and YAP1. In addition, the positive association between the expression of SNHG16 and YAP1 in CRC tissues further demonstrated that YAP1 was the potential target of SNHG16.
Thus, YAP1 was selected for further research.
To investigate whether the function of SNHG16 on tumor progression was dependent on YAP1 or not, rescue experiment was performed. The result showed that SNHG16 knockdown could inhibit the proliferation, migration, invasion, and EMT of CRC cells.
Evidently, YAP1 overexpression could rescue SNHG16-downregulation-mediated inhibitory effect on proliferation, migration, and invasion (Fig S2A, Fig. 3E, Fig S2C). Meanwhile, SNHG16-overexpression facilitated proliferation, migration, invasion, and EMT of DLD1, and the function of SNHG16-overexpression on tumor progression could be abrogated by YAP1 knockdown (Fig S2B, Fig. 3F, Fig   S2D). Immuno uorescence and western blot revealed that the overexpression of SNHG16 could promote the EMT of CRC cells, and the promotive effect on EMT could be rescued by YAP1 knockdown; meanwhile, the SNHG16-knockdown-mediated inhibitory effect on EMT could be rescued by YAP1 overexpression (Fig. 3G and Fig. 3H). These results revealed that YAP1 was functional mediator of SNHG16 in CRC cell lines.
To further investigate how the EMT of CRC cells might be controlled dynamically by SNHG16/YAP1 axis during tumor metastasis, we focused on the involvement of microRNAs (miRNAs) 47,48 . According to the hypothesis of competing endogenous RNAs (ceRNAs), certain lncRNAs have many miRNA response elements (MRE) and can function as microRNA sponge to impose a posttranscriptional regulation of target genes in cytoplasm, thus playing an important role in the progression of CRC 49,50 . As shown in Fig. 1C, SNHG16 is mainly located in the cytoplasm, indicating a potential posttranscriptional regulation. Through starBase 51 and LncBase Predicted 52 , we identi ed several miRNAs which have potential SNHG16 binding sites, one of which was miR-195-5p. Furthermore, we found the same miR-195-5p response elements between SNHG16 and YAP1 (Fig. 4A), which indicate a potential interplay between these three factors.
Next, we sought to examine whether SNHG16-mediated miR-195-5p regulation occurs through direct binding with miR-195-5p or not.
Therefore, RIP, biotin-avidin pulldown, and luciferase assay were applied to conduct further research.
miRNAs are present in the form of miRNA ribonucleoprotein complexes (miRNPs) which contain Argounaute2 (Ago2), the core component of RNA-induced silencing complexes (RISCs) 53 . To validate whether SNHG16 associated with miR-195-5p on Ago2, we performed RNA-binding protein immunoprecipitation (RIP) with Ago2 on HCT116 cells transfected with miR-195-5p inhibitor or transfected with negative control (Fig. 5A). As shown in Fig. 5B, Ago2 protein was successfully immunoprecipitated from HCT116 extracts by Ago2 antibody. Clearly, SNHG16 was signi cantly decreased in Ago2 immunoprecipitates puri ed from miR-195-5p inhibitor transfected cells, compared with its level in Ago2 immunoprecipitates puri ed from their negative control (Fig. 5C). Thus, the RIP assay revealed that SNHG16 was likely in the miR-195-5p-RISC and interacts with miR-195-5p. Additionally, the sequence-speci c binding between miR-195-5p and SNHG16 was further validated by a biotin-avidin pulldown system. As Fig. 5D shows, the expression of SNHG16 in miR-195-5p-Bio group was about 30 times higher than that of miR-195-5p-Mut-Bio group or control group. Consistent with our results, the luciferase reporter assays revealed that miR-195-5p signi cantly inhibited the luciferase activity of SNHG16-WT but not of SNHG16-Mut (Fig. 5E-F). Collectively, these results revealed that SNHG16 could act as miRNA sponge to directly bind with miR-195-5p and exert inhibitory effect on miR-195-5p expression.
Furthermore, the inhibited luciferase of miR-195-5p group on YAP1-WT could be rescued by SNHG16 (Fig. 5G). We did not nd signi cant changes when we conducted the same experiment on YAP1-mut (Fig. 5G). The luciferase assay among these three factors revealed that, the overexpression of SNHG16 could prevent miR-195-5p from targeting YAP1. Through western blot and qRT-PCR in CRC cell line, we could also found that SNHG16 could positively regulate the expression of YAP1. Furthermore, miR-195-5p could rescue SNHG16-mediated regulation on YAP1 (Fig. 5I-K).
Altogether, SNHG16 could function as miRNA sponge to directly bind and sequester endogenous miR-195-5p, thereby preventing it from inhibiting YAP1 expression and affecting tumor progression.
YAP1 could directly bind with TEAD1, forming a complex that binds to the promoter region of SNHG16 and activates its transcription.
Since YAP1 functions as an important transcriptional co-activator, the positive correlation between the expression of SNHG16 and YAP1 drew us to investigate the regulatory role of YAP1 in SNHG16 transcription. Then, we constructed a lentiviral-based HCT116 cell line which will stably overexpress or knockdown YAP1, termed as Lv-Oe-YAP1 or Lv-anti-YAP1, respectively. Evidently, overexpression of YAP1 signi cantly promoted SNHG16 expression (Fig. 6A), whereas knockdown of YAP1 dramatically suppressed the expression level of SNHG16 (Fig. 6B). Furthermore, knockdown of YAP1 could inhibit the proliferation (Fig S3A), migration (Fig S3C and Fig S3E) ,invasion (Fig S3C), and EMT (Fig. 6C) of CRC cell lines. In addition, we observed the promotive effect on tumor progression in Lv-Oe-YAP1 group (Fig S3B, S3D, S3F, and Fig. 6C).
Although we have demonstrated the important role of YAP1 in SNHG16 expression and tumor progression, the mechanism of YAP1-lncRNA interaction remains largely unknown. Based on existing research, YAP1 cannot bind directly with DNA alone. However, YAP1 could combine with transcription factors to regulate the transcription of downstream genes 54 . Among these transcription factors, TEAD1-4 is one of the most common binding molecules 55 . Once bound to TEAD, YAP1 formed a YAP1/TEAD complex to initiate downstream gene transcription 54,56 . Through bioinformatic analysis, We found the potential transcription factors TEAD1. Further study showed that the knockdown of either YAP1 or TEAD1 reduced the expression of SNHG16 (Fig. 6D). And evidently, the inhibitory effect on the expression of SNHG16 was most signi cant when we simultaneously knocked down both of YAP1 and TEAD1 (Fig. 6D). To investigate whether YAP1 and TEAD1 act synergistically, we overexpressed TEAD1 in the Lv-anti-YAP1 cell line and we did not nd signi cant change on the expression of SNHG16 (Fig. 6D). Similarly, when we overexpressed TEAD1 after YAP1 knockdown, the expression level of SNHG16 was not rescued (Fig. 6D). To further verify the cross-talk between TEAD1 and YAP1, coimmunoprecipitation (co-IP) analysis was applied. The co-IP con rmed that YAP1 could bind to TEAD1 directly or indirectly in HCT116 (Fig. 6E). To further investigate whether YAP1 interact with TEAD1 through directly binding or not, GST-pull down assay was performed.
After the GST-tagged YAP1 (Fig. 6F) and His-tagged TEAD1 (Fig. 6G) were successfully induced and puri ed from Escherichia coli, we incubated these two proteins in 4℃ for 4-8 h. Finally, western blot was performed to examine the expression of these two proteins.
GST-pull down further demonstrated the direct interplay between YAP1 and TEAD1 (Fig. 6H). Collectively, these experiments con rmed that YAP1 could directly interact with TEAD1 and form a complex, which regulated the expression of SNHG16. However, whether YAP1/TEAD1 binds to the promoter of SNHG16 remains largely unclear.
Therefore, the interplay between YAP1/TEAD1 complex and SNHG16 was selected for further research. Signi cantly, dual-reporter luciferase assays showed that overexpression of either YAP1 or TEAD1 in HCT116 cells stimulated the promoter activity of SNHG16 (Fig. 6J). And evidently, the promoter activity was most signi cant when we simultaneously overexpress both of YAP1 and TEAD1 (Fig. 6J). To verify the binding of SNHG16 and TEAD1, we generated a series of 5' deletion constructs of SNHG16 promoter. According to the result of luciferase assay, the regulatory region between − 1794 and − 1367 was responsible for YAP1/TEAD1-mediated promoter regulation (Fig. 6K). As shown in Fig. 6I, two binding sites are located in this region. Compared with wild-type (WT), YAP1/TEAD1 failed to stimulate the mutants of both predicted sites in the promoter region of SNHG16 in HCT116 (Fig. 6L). Furthermore, in the ChIP assay, we designed two primer sets containing site 1 and site 2, respectively. Then, we used the primer and puri ed DNA among ChIP to amplify part of the promoter region. ChIP assays revealed that TEAD1 directly bound to both site1 and site2 of SNHG16 promoter in HCT116 (Fig. 6M).
The results indicated that YAP1 could directly bind with TEAD1 to form a complex, which could, in turn, bind to the promoter of SNHG16 and regulate its transcription. During this process, YAP1 is indispensable to TEAD1 and these two act synergistically.

The effect of YAP1-SNHG16 positive feedback loop on the progression of tumor
We have demonstrated that SNHG16 was a direct target of YAP1/TEAD1 complex, and there existed a positive feedback regulation between SNHG16 and YAP1. As aforementioned, YAP1 could regulate the expression of SNHG16 and promote tumor progression. To investigate the YAP1/TEAD1-SNHG16 positive feedback loop on tumor progression, the following experiment was performed. Clone formation revealed that the overexpression of YAP1 signi cantly promoted the tumor proliferation, whereas the enhanced proliferation of HCT116 was impaired by simultaneous knockdown of SNHG16 (Fig S4A). Similarly, SNHG16 knockdown could partially attenuate the enhanced migration, invasion, and EMT of Lv-Oe-YAP1 stable cell line (Fig S4C, Fig S4E, and Fig. 6Q). In addition, the overexpression of SNHG16 could rescue the inhibitory effect of Lv-anti-YAP1 on tumor progression (Fig S4B, Fig S4D, Fig S4F and   Fig. 6N).
The alteration of lncRNA-SNHG16 expression in uenced the CRC tumorigenesis and M CTC generation in vivo To verify the above results in vivo, nude mouse xenograft experiment was performed. The stable cell lines of Lv-anti-SNHG16-NC (HCT116), Lv-anti-SNHG16 (HCT116), and Lv-anti-SNHG16/Lv-Oe-YAP1 (HCT116) were injected into the anks of nude mice separately.
During the period of tumor growth, both the tumor weight and tumor volume was measured. Evidently, knockdown of SNHG16 inhibited the growth of tumor. Furthermore, the inhibitory effect of SNHG16 on tumor growth was rescued by YAP1 overexpression (Fig. 7A, B and C).
Similarly, Ki67 staining also decreased in Lv-anti-SNHG16 group, indicating the inhibited proliferation of tumor. Meanwhile, overexpression of YAP1 signi cantly rescued the inhibited expression of Ki67 (Fig. 7D). Signi cantly, the IHC staining also demonstrated the in vitro result that SNHG16 could regulate the expression of YAP1 and the process of EMT. Moreover, YAP1 overexpression could rescue SNHG16-knockdown-mediated inhibitory effect on EMT (Fig. 7D). Through the process of EMT, cancer cells could shed from tumor and invade into blood, thereby forming CTCs and facilitating tumor metastasis. Thus, we detected the CTC from mouse blood (Fig. 7E). Obviously, Lv-anti-SNHG16 signi cantly decreased the generation ratio of M CTC, compared with other groups (Fig. 7F); meanwhile, YAP1 overexpression could rescue SNHG16-knockdown-mediated inhibitory effect on M CTC generation (Fig. 7F).
The stable cell lines of Lv-anti-SNHG16 NC (HCT116), Lv-anti-SNHG16 (HCT116), Lv-anti-SNHG16/Lv-Oe-YAP1 (HCT116) were separately injected into tail vein to con rm the effect of SNHG16 on tumor metastasis. The representative images of CRC liver metastasis and lung metastasis were presented in Fig. 7G and 7H, respectively. Further analysis revealed that the knockdown of SNHG16 signi cantly inhibited the metastasis of tumor. In addition, the overexpression of YAP1 could rescue the inhibitory effect of SNHG16 knockdown on tumor metastasis ( Fig. 7G and 7H). Collectively, these results demonstrated that SNHG16 could promote the tumor growth, metastasis, and M CTC generation. And what's more, YAP1 could rescue the effect of SNHG16 on tumor metastasis.
We summarized our mechanistic ndings in a schematic (Fig. 7I). Our study illustrated a positive feedback loop between lncRNA SNHG16 and YAP1/TEAD1. Mechanistically, SNHG16 could act as miRNA sponge to sequester miR-195-5p on Ago2, thereby protecting YAP1 from repression and facilitating CRC liver metastasis. Moreover, YAP1 could also combine with TEAD1 to form a YAP1/TEAD1 complex, which could bind with the promoter of SNHG16 to regulate its transcription. In a positive feedback loop, SNHG16 promoted the migration, invasion, M CTC generation, lung metastasis, and liver metastasis of CRC.

Discussion
Extensive efforts in past have been made to explore the cellular and molecular mechanism of CRC liver metastasis, one of the main cause for the failure of advanced CRC. Recently, lncRNAs have gained enormous attention as a critical regulator of tumor invasion and metastasis 57,58 . Tang et al. also found that, Rho/ROCK signaling-associated lncRNAs is useful for cancer metastasis 59 . In present study, we discovered a positive feedback loop of SNHG16-YAP1/TEAD1 in CRC. Through this positive feedback loop, SNHG16 could promote CRC cellular proliferation, migration, invasion, EMT, M CTC generation, and liver metastasis of CRC. In turn, YAP1 could combine with TEAD1 to form a complex, which in turn bind with the promoter of SNHG16 and activates its transcription. Therefore, our research clari ed a hitherto unexplored mechanism for both SNHG16 and YAP1/TEAD1 in CRC progression.
Clinically, elevated level of lncRNA-SNHG16 located in CRC tissues was positively correlated with TNM stage, LVI, PNI, distant metastasis, and poor prognosis of cancer patients, indicating their potential regulation on CRC dissemination and invasion. Recently, accumulating evidences revealed that lncRNA was associated with the EMT of tumor cells 60,61 . In this study, our results potently demonstrated that SNHG16 exerted oncogenic role in CRC through promoting cell proliferation, mobility, and EMT. CTCs, which was considered as the precursors of metastases 13 . According to recent research, CTCs shed from primary tumor could gain more mesenchymal traits by EMT program, which allow them to have stronger invasive ability and metastatic ability 62  Currently, mounting evidence suggested that lncRNAs could act as sponges of miRNAs to impact the expression of mRNA 49 . For instance, lncRNA-ATB could act as miRNA sponge to inhibit miR-200, thus regulate the expression of ZEB1 and induce epithelialmesenchymal transition 72 . In addition, lncRNA-HOXD-AS1, which is highly expressed in HCC, could act as a 'sponge' to prevent SOX4 from miR-130-3p-mediated degradation, leading to the HCC metastasis 58 . Subsequent ISH experiment in our research showed that SNHG16 is mainly located in cytoplasm in CRC, which indicating a potential role in posttranscriptional regulation. Thus, we speculated that SNHG16 could act as microRNA sponge to prevent YAP1 from the microRNA mediated inhibition. Here, the inverse correlation between SNHG16 expression and miR-195-5p further indicated the potential targeted regulation between SNHG16 and miR-195-5p. Lamar also found that, YAP could promote metastasis through its TEAD-interaction domain 69    Supplementary Files