Downregulation of SNX5 By KLF9 Leads to Clear Cell Renal Cell Carcinoma Progression By Inducing CD44 Internalization and Suppressing Epithelial-to-Mesenchymal Transition

Background: Aberrant expression of SNX5 can contribute to tumourigenesis, invasion, and metastasis of several human cancers. However, the clinicopathological and biological signicance of SNX5 in clear cell renal cell carcinoma (ccRCC) remain unclear. The aim of this study was to examine the role of SNX5 in the progression of ccRCC. Methods: Immunohistochemical (IHC), Western blot, qRT-PCR, western blot, ow cytometry and immunouorescence were used to detect the expression of indicated molecules. The biological role of SNX5 in ccRCC cells was evaluated by CCK8, colony formation, transwell assay, subcutaneous tumor formation as well as veil tail injection. ChIP assay and luciferase reporter assay were used to determine the direct binding of KLF9 to the promoter of the SNX5 gene. Results: SNX5 expression was downregulated in human ccRCC tissues. SNX5 expression was negatively correlated with tumor size, AJCC stage, tumor thrombus of inferior vena cava (IVC) and poor prognosis of ccRCC. Ectopic expression of SNX5 inhibited ccRCC cell proliferation and metastasis whereas knockdown of SNX5 increase these activities both in vitro and in vivo. Mechanistically, overexpression of SNX5 blocked internalization and intracellular tracking of CD44 in ccRCC cells. Exogenous expression of CD44 partially rescued the inhibitory effects of SNX5 on the proliferation and invasion activity of ccRCC cells. Knockdown of SNX5 in ccRCC cells was associated with epithelial mesenchymal transition (EMT), including the down-regulation of E-cadherin, ZO-1 and Claudin-1 and the concomitant up-regulation of Snail and N-cadherin. In addition, SNX5 inhibited TGF-β-induced migration, invasion and EMT in ccRCC cells. Moreover, we observed a signicant correlation between SNX5 expression and E-cadherin levels in ccRCC patients. In addition, KLF9 directly bound to the SNX5 promoter and increased SNX5 transcription. SNX5 expression was closely correlated with KLF9 expression in ccRCC. Moreover, we found that the combination of SNX5 and CD44 or E-cadherin or KLF9 was a more powerful predictor of poor prognosis than either parameter alone. Conclusion: Collectively, our data reveal a mechanism that KLF9-mediated SNX5 expression was associated with poor prognosis via tracking of CD44 and promoting EMT in ccRCC. SNX5 may be a potential prognostic biomarker and therapeutic target for patients with ccRCC. and (A) Transwell analysis showed that overexpression inhibited ccRCC cells migration and invasion. (B) Transwell analysis showed that knockdown of SNX5 promoted ccRCC cells migration and invasion.(C) The expression of MMP9 was detected by Western blot in SNX5-overexpressing cells and (D)SNX5 knocked down cells.(E) H&E staining showing that lung metastasis in mice injected via tail vein.


Introduction
Kidney cancer is among the top ten malignant tumors, representing 4.1% of all new cancer cases, and it is estimated that 73820 people will be diagnosed in 2019 in the United States [1].Clear cell renal cell carcinoma (ccRCC) is the most common form of kidney cancer, accounting for up to 85% of cases [2]. In recent years, there has been further in-depth understanding of the underlying pathways driving ccRCC biology, resulting in notable improvements in diagnosis and therapeutics [3]. However, owing to recurrence and metastatic diseases, in additional to lacking of effective treatments, the prognosis of ccRCC remains still poor [4].Therefore, it is urgent to elucidate the molecular mechanism of ccRCC and explore new therapeutic targets.
Sorting Nexin (SNXs) is a family that regulates the endocytosis of eukaryotic cells discovered in recent years [5].This family is a relatively conservative in the process of biological evolution, and its structure contains the Phoxhomology (PX) domain, which is tightly combined with phosphoinositol in the membrane structure [6]. As part of a retromer complex, SNXs play a critical part in transmembrane transport, protein sorting, intracellular signal transduction and organelle movement [7]. Therefore, SNXs family members are closely related to the occurrence and development of tumors.SNX16 expression was signi cantly upregulated in colorectal cancer (CRC) tissues. Upregulated SNX16 predicted poor survival of CRC patients [8]. SNX10 acts as a tumor suppressor which inhibits colorectal cancer initiation and progression [9].SNX1 expression is signi cantly downregulated in colon cancer [10]. Knockdown of SNX1 expression can signi cantly increase the phosphorylation level of EGFR and promote colon cancer cells proliferation [11].SNX6 predicts poor prognosis and contributes to the metastasis of pancreatic cancer cells via activating epithelial-mesenchymal transition (EMT) [12].SNX27 promotes breast cancer metastasis and the elevated expression of SNX27 could be related to the marginally shorter survival of the patients [13].Our previous research shows that SNX5 predicts poor prognosis and promotes hepatocellular carcinoma progression by modulating the EGFR-ERK1/2 signaling pathway [14].
However, the biological role and molecular mechanism of SNX5 in ccRCC remain unknown. In the present study, we analyzed the expression levels and clinical signi cance of SNX5 in ccRCC patients, and investigated the molecular mechanism underlying its role.

Materials And Methods
Cell lines 786-O, 769-P, Caki-1and ACHN were purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). HEK-293T cell lines were purchased from the American Type Culture Collection (Manassas, VA, USA). 786-O, 769-P were cultured in RPMI 1640 medium, Caki-1 were cultured in McCoy's 5A medium, ACHN and HEK-293T were cultured in DMEM medium containing 10% fetal bovine serum (FBS), 100 U/ml penicillin and 100 U/ml streptomycin at 37 °C with 5% CO 2.

Immunohistochemistry (IHC)
This study was conducted in accordance with the International Ethical Guidelines for Health-related Research Involving Humans and was approved by the Research Ethics Committee of Renji Hospital. A total of 150 pathologically con rmed ccRCC and 30 adjacent normal kidney tissues were collected by Shanghai National Engineering Research Center from Taizhou Hospital from 2008 to 2015. IHC assays were conducted as reported previously. Brie y, the sections were depara nized with xylene and rehydrated before being heated to just below boiling temperature in Tris/EDTA buffer pH 9.0 for 20 minutes in a microwave oven for antigen retrieval. The primary antibodies used for the IHC assay were against SNX5. Scores of staining intensity were: 0, negative; 1, weak; 2, moderate; 3, strong. Scores of positively stained cell proportion were: 0, no positive; 1, <10%; 2, 10%-35%; 3, 35%-75%; 4, >75%.The results were scored 0 to 4 by two independent investigators. The individual scores for the staining intensity and percentage of positivity cells were then multiplied to calculate the immunoreactivity score for each sample. Samples having a nal staining score of ≤4 were considered to exhibit low expression and those with a score of >4 were considered to be exhibit high expression.

Plasmids construction and lentivirus production
The plasmids coding sequences for human SNX5 and CD44 were constructed in our laboratory. The SNX5 lentiviral plasmid was supplied by GeneCopoeia (Guangzhou, China). psPAX2 and pMD2.G plasmids were purchased from Addgene (USA). According to manufacturer's protocol, 293T cells were transfected with a mixture of overexpressing or interfering plasmid, psPAX2 and pMD2.G plasmids for lentivirus packaging using Lipofectamine™ 2000 (Invitrogen, USA).The viruses were collected after harvesting 48-72h and added to ccRCC cells with 1×10 6 recombinant lentivirus-transducing units in the presence of 6 μg/ml polybrene (Sigma, USA).The target sequences are listed in Supplementary Table S1.
Quantitative real-time polymerase chain reaction (qRT-PCR) Total RNA was retracted with TRIzol reagent (Invitrogen, USA) and was reverse-transcribed with a PrimeScript TM RT Reagent kit (TaKaRa, China).RT-qPCR was performed with an ABI Prism 7500 System (Applied Biosystems, USA) with SYBR Green Maste (Takara, China). All primer sequences are listed in Supplementary Table S2.

Western blot
Total proteins extracted from the cells were lysed with RIPA buffer (Thermo Scienti c, USA) and separated on 8% -12% SDS-PAGE gels, transferred onto polyvinylidene uoride (PVDF) membranes (Millipore, USA), incubated with primary antibodies overnight at 4°C, probed with HRP-conjugated secondary antibodies, and visualized using enhanced chemiluminescence reagent (Pierce, USA) via a chemiluminescence analyzer (Bio-Rad, USA). Information on the antibodies is listed in Supplementary   Table S3.
Cell proliferation and colony formation assays Brie y, for CCK8 assay 600-800 cells/well were seeded in 96-well plates. After overnight, the CCK8 solution and the culture solution are mixed at a ratio of 1:9, the mixture is 100μL/well, incubated at 37°C 2 hours, the absorbance value is at A450nm. For colony formation assays, cells were seeded in each well of a 6-well plate and incubated at 37°C. After 1-2 weeks, discarded the culture medium and xed it with formalin for 30 minutes. Added the Giemsa staining solution and placed it on a shaker overnight. Scanned and counted the number of clones using Image J software.
Transwell assay Inoculated 4×10 4 cells in 200 μl of serum-free medium, put in the upper chamber of a transwell (8 μm pore size) or Matrigel coated transwell upper cavity. The lower chamber contains RPMI 1640medium with 10% fetal bovine serum as a chemotactic agent. After incubating for 6-8 hours at 37°C, took out the transwell and xed it with formalin, stain with crystal violet for 10-15 minutes, used a cotton swab to gently wipe off the non-migrated or non-invaded cells from the upper chamber. And choosed ve random elds of view to count under the microscope.

Flow cytometry analysis
To detect the population of E-cadherin + or the population of CD44 + , cells were plated in a 6-well plate and incubate at 37°C overnight, digested with trypsin the next day, washed the cells twice with PBS, added the corresponding antibodies and incubated on ice for 45 minutes, avoid light. Afterwards, the supernatant was discarded, and the cells were washed twice with PBS for ow cytometry analysis. The antibodies used are shown in Supplementary Table S3.
Immuno uorescence confocal imaging The cells were inoculated on Lab-Tek laboratory slides one day in advance, and xed with 4% paraformaldehyde at room temperature for 30 minutes the next day, washed three times with PBS; then 0.1% Triton X-100 was permeated for 5 minutes. Washed three times with PBS, added the primary antibody, 4 °C overnight, washed three times with PBS, incubated in Alexa Fluor 594-conjugated and Alexa Fluor 534-conjugated secondary antibodies and 4′, 6-diamidino-2-phenylindole (DAPI) in blocking solution for 30 minutes at 37°C in a humidi ed chamber, washed three times and took pictures on Leica TCS SP8 confocal system (Leica, Microsystems). Information on the antibodies is listed in Supplementary Table S3.

Internalization and recycling assays
Internalization and recycling assays were conducted as reported previously [14].Firstly, Seeded cells in 6cmdishes, the next day reached 80%-100%con uence. Secondly For recycling assays, the biotin-labeled surface protein was internalized at 37°C for 30 minutes. The surface biotin was removed as described above. After treatment with stripping buffer, the cells were washed once in cold medium. Added pre-warmed serum-free growth medium and incubated the cells at 37°C for the speci ed time point. Finally, incubated the cell lysate with streptavidin-coupled agarose beads at 4°C for 2 hours. The agarose bead mixture was washed three times in RIPA buffer, and then the biotinylated protein was detected by Western blot analysis.

Luciferase assay
The SNX5 promoter (-1297 bp /40 relative to the transcription start site) was cloned into the luciferase reporter gene vector pGL3-Basic (Promega).. The delity of the constructs were con rmed by sequencing.
The primer sequences are listed in Supplementary Table S2.
Cells were cotransfected with the corresponding reporter plasmid and the indicated plasmids in each experiment according to the standard protocol. The pRL-TK reporter construct was used as the internal control. Luciferase activity was detected using a Dual-Luciferase Report Assay (Promega) system in accordance with the manufacturer's instructions. [15].
In vivo growth and metastasis assays Invivo growth, nude mice were injected with a total of 6× 10 6 cells suspended in 200μl of a mixture of serum-free RPMI 1640/Matrigel (1:1 volume) (BD Biosciences, MA) into the left axillary fossa. For metastasis assay, 2× 10 6 cellssuspended in 200μl ofserum-free RPMI 1640 injected into nude mice via tail vein injection. Eight weeks later, the mice were sacri ced, and the tumors and lung tissues were excised and xed with 4% phosphate-buffered neutral formalin for at least 72 h. Metastatic tissues were analyzed by H&E staining. All of the experiments were approved by the Shanghai Medical Experimental Animal Care Commission.

Statistical analysis
Statistical analyses were performed using SPSS16.0 software. All data are presented as the mean ± SD. The correlation between SNX5 expression and E-cadherin was performed using Pearson's coe cient tests. Statistical comparisons of the data were performed using two-tailed Student's t-test or one-way ANOVA for multiple comparisons. Chi-square tests or sher's exact tests were used to analyze the correlation between the expression of SNX5 and clinicopathological features. Overall survival curves were calculated using the Kaplan-Meier method and compared using the log-rank test. Univariate and multivariate analyses were performed using the Cox proportional hazard regression model in a stepwise manner. p< 0.05 was considered statistically signi cant.

SNX5 is down-regulated and correlates with prognosis of ccRCC patients
To explore the role of SNX5 in ccRCC, we rst detected the expression of SNX5 using data sets from TCGA (The Cancer Genome Atlas).The results showed that the expression of SNX5 was downregulated in ccRCC tissues compared with noncancerous tissues (Fig. 1A). Furthermore, expression of SNX5 was downregulated in ccRCC based on the GSE15641 and GSE126964 datasets ( Fig. 1B-1C). To investigate these ndings further,we next analyzed SNX5 protein expression using immunohistochemistry and a tissue microarray of patient-derived ccRCC samples (n = 30) and matched normal kidney samples (n = 30).Of the 30 pairs, 24 (80%) had higher SNX5 protein expression in nontumor tissues than in tumor kidney tissues, 4 (13.3%) had similar expression, whereas only 2 (6.7%) had lower expression (Fig. 1D).In addition, expression of SNX5 was also down-regulated in ccRCC tumor tissue and noncancerous tissues using the Clinical Proteomic Tumor Analysis Consortium (CPTAC) datasets ( Supplementary Fig. S1). Taken together, the results of these analyses show that SNX5 is downregulated in ccRCC.
We next detected the expression of SNX5 in 150 cases of ccRCC using IHC. According to the IHC results, the expression intensity of SNX5 protein was scored. A score of 0 to 4 was considered to represent low expression and a score of >4 was considered to represent high expression (Fig. 1E). Our results showed that expression of SNX5 was closely associated with tumor size, American Joint Committee on Cancer (AJCC) stage and tumor thrombus of inferior vena cava (IVC) ( Table 1).Kaplan-Meier survival analysis revealed that lower levels of SNX5 were associated with a shorter overall survival (OS) time (p = 0.009; Fig. 1F). The same results were also found from the TCGA ccRCC cohort (p = 0.023; Fig. 1G). Moreover, univariate and multivariate COX regression analysis showed that a low level of SNX5 was associated with worse survival of ccRCC patients (Table 2). Therefore these ndings indicate that SNX5 may serve as a valuable prognostic factor for ccRCC patients after surgery.

SNX5 inhibits the proliferation and tumourigenicity of ccRCC cells
Due to the fact that expression of SNX5 is associated with tumor size and tumor thrombus in ccRCC led us to rationalize that SNX5 might be important for ccRCC tumor growth and metastasis. To determine this possibility, 769-P, Caki-1 and 786-O cells were chosen for loss-or gain-of-function studies due to their high or low endogenous SNX5 levels ( Supplementary Fig. S2).CCK8 and colony formation assays showed that the overexpression of SNX5 inhibited ccRCC cell proliferation and colony formation ability, while the knockdown of SNX5 signi cantly promoted cell proliferation and colony formation ability ( Fig.  2A-2F). Next, we examined the effect of SNX5 on the tumourigenicity of ccRCC in vivo by using subcutaneous tumor model in nude mice. As shown in Fig. 2G, compared with the control group, 786-O overexpressing SNX5 inhibited tumor growth as determined by tumor volume and tumor weights. In contrast, tumors with shSNX5 exhibited larger volumes and higher weights than tumors with shNC in Caki-1 cells (Fig. 2H). Furthermore, expression of Ki-67 and PCNA was signi cantly reduced in ccRCC cells with SNX5 overexpression and enhanced in ccRCC cells with SNX5 knockdown (Fig. 2I-2J). Taken together, these results provide strong evidence that SNX5 inhibits the tumorigenic ability of ccRCC cells.

SNX5 inhibits invasion and metastasisof ccRCC in vitro and in vivo
To explore whether SNX5 is necessary for ccRCC metastasis, we rst analyzed the effects of SNX5overexpression or knockdown on ccRCC cell migration and invasion. Our results showed that overexpression of SNX5 inhibited ccRCC migration and invasion, while knockdown of SNX5 resulted in signi cant increase in ccRCC cells migration and invasion ( Fig. 3A-3B). In addition, we found that overexpression of SNX5 decreased MMP9 expression in ccRCC cells, whereas the knockdown of SNX5 increased MMP9 expression in ccRCC cells (Fig.3C-3D).
To determine whether SNX5 expression plays a role in ccRCC metastasis in vivo, we injected ccRCC cells expressing Vector or SNX5 into the tail veins of mice to mimic lung metastasis. Histological examination of lung tissues indicated that SNX5 overexpression tumor-bearing mice had signi cantly lower numbers of lung metastatic nodules than the control group (Fig. 3E). Taken together, our ndings demonstrate that SNX5 play an important role in the metastasis of ccRCC cells.

SNX5 negatively regulated epithelial to mesenchymal transition (EMT) in ccRCC cells
Epithelial-to-mesenchymal transition (EMT) is a key process by which cancer cells acquire invasive and metastatic properties [16].Our results demonstrated that SNX5 overexpression in 786-O and 769-P cells resulted in a morphological change in epithelial mesenchymal transition in ccRCC cells, from ber to round shape, suggesting the inhibition of EMT transition (Fig. 4A). Then, we detected EMT markers in SNX5-overexpressing and SNX5 knockdown ccRCC cells. The results showed that epithelial markers Ecadherin, ZO-1 and Claudin-1 were increased in SNX5 overexpression 786-O and 769-P cells, whereas mesenchymal markers N-cadherin and transcription factors Snail were decreased. On the contrary, knockdown of SNX5 in 769-P and Caki-1 cells decreased the expression of E-cadherin, ZO-1 and Claudin-1, and was accompanied by increased the expression of N-cadherin and snail (Fig. 4B-4C).And, that is the same as the result of western blot on tumor tissues overexpressing and knockdown SNX5 (Fig. 4D-4E). Moreover, we used immuno uorescence to detect the expression levels of E-cadherin and N-cadherin in 769-P cells with stable overexpression of SNX5, the results were consistent with the ndings obtained by Western blot (Fig. 4F).
We also detected the cell surface expression of E-cadherin in the SNX5-overexpressing and SNX5 knockdown ccRCC cells by owcytometry. The results showed that SNX5 overexpression increased the expression of E-cadherin compared to the control group in 769-P cells. Conversely, knockdown of SNX5 signi cantly decreased the expression of E-cadherinin 769-P cells (Fig. 4G). Collectively, these results revealed that SNX5 negatively regulates EMT in ccRCC cells.
To further explore potential clinical applications, we next evaluated the expression of E-cadherin in ccRCC using data sets from TCGA. The results showed that expression of E-cadherin was downregualted in ccRCC tumor tissues compared with normal tissues (Supplementary Fig. S3A). In addition, E-cadherin protein expression was also down-regulated in ccRCC tumor tissue and noncancerous tissues using CPTAC datasets (Supplementary Fig. S3B). Then, we analyzed the relationship between SNX5 and Ecadherin using data sets from TCGA data. The results showed that there was a positive correlation between the expression of SNX5 and E-cadherin in ccRCC tissue ( Supplementary Fig. S3C). Furthermore, Kaplan-Meier survival analysis revealed that lower levels of E-cadherin were associated with a shorter overall survival (OS) time based on TCGA (p = 0.000; Supplementary Fig. S3D). Moreover, the patients with low expression of SNX5 and E-cadherin displayed a worse prognosis than the high SNX5 and Ecadherin groups based on TCGA (Fig. 4H).

TGF-β-induced EMT was partially reversed by SNX5 overexpression in ccRCC cells
Transforming growth factor β (TGF-β) is a secreted cytokine and may function as a tumor promoter by facilitating cancer cells to undergo EMT [17].Therefore, we investigated whether the SNX5 affects TGF-βinduced EMT in ccRCC cells. Our results showed that morphologically, overexpression of SNX5 partially inhibited TGF-β-induced EMT (Fig. 5A). Furthermore, elevated N-cadherin and Snail expression and reduced E-cadherin, ZO-1 and Claudin-1 expression in response to SNX5 overexpressing were alleviated by TGF-β treatment (Fig.5B). In addition, we used immuno uorescence to detect the expression levels of E-cadherin and N-cadherin in SNX5 overexpressing 769-P cells-treated with TGF-β. The results were consistent with the ndings obtained by Western blot (Fig. 5C). Transwell assays showed that overexpression of SNX5 partly reversed the effect of TGF-β on promoting ccRCC cells migration and invasion (Fig. 5D). Thus, these data support the view that SNX5 negatively regulates EMT in ccRCC cells.

SNX5 inhibited ccRCC cell proliferation, migration and invasion by inducing CD44 internalization
Cancer cells that undergo an EMT acquire cancer stem cell-like properties and show an increase in CD44 expression [18].Therefore, we determined the effects of overexpression or knockdown of SNX5 on the marker of potential cancer stem cell (CSC). Our results showed that overexpression of SNX5 inhibited CD44 and Oct4 mRNA and protein expressionin 786-O and 769-P cells. Knockdown of SNX5 increased CD44 and Oct4 mRNA and protein expression in 769-P and Caki-1 cells (Fig. 6A-6D). And, that is the same as the result of western blot on tumor tissues overexpressing and knockdown SNX5 (Fig. 6E-6F).
In addition, our results showed that SNX5 overexpression decreased the cell surface expression of CD44 compared to the control group in 769-P cells, whereas knockdown of SNX5 signi cantly increased the cell surface expression of CD44 in 769-P cells (Fig.7A). Furthermore, confocal colocalization analysis also revealed that SNX5 overexpression had lower expression of CD44 in 786-O and 769-P cells (Fig. 7B). IHC demonstrated that CD44 was down regulated in murine xenografts from SNX5-overexpressing 786-O cells (Fig. 7C).
Considering SNX-mediated endosomal transport, we investigated whether the overexpression of SNX5 affects the internalization and recycling of CD44 in ccRCC cells. Our results showed that SNX5 overexpression caused a signi cant increase in CD44 internalization and affected its recycling to the cell surface ( Fig. 7D-7F).
Next, we investigated whether the SNX5 inhibited ccRCC cell proliferation, migration and invasion through regulation of CD44.To con rm the role of CD44 in SNX5-inhibted cell proliferation, migration and invasion, we constructed CD44 expression vector to rescue its expression in SNX5 overexpression ccRCC cells. The results showed that the SNX5 overexpression inhibited cell proliferation, migration and invasion was partially reversed by the overexpression of CD44 (Fig. 8A-8D).Therefore, all results show that SNX5 inhibited ccRCC cell proliferation, migration and invasion by regulation of CD44.
To further explore potential clinical applications, we next evaluated the expression of CD44 in ccRCC using data sets from TCGA. The results showed that expression of CD44 was upregualted in ccRCC tumor tissues compared with normal tissues (Supplementary Fig. S4A). In addition, CD44 protein expression was also upregulated in ccRCC tumor tissue and noncancerous tissues using CPTAC datasets( Supplementary Fig. S4B).Furthermore, Kaplan-Meier survival analysis revealed that high levels of CD44 were associated with a shorter overall survival (OS) time based on TCGA (p = 0.005. (Fig. 8E). In addition, the patients with low expression of SNX5 and high expression of CD44 played a worse prognosis than the high SNX5 and low CD44 groups according to TCGA, indicating that the combination of SNX5 and CD44 has better prognostic value than SNX5 or CD44 alone (Fig. 8F).

SNX5 is upregulated by KLF9 in ccRCC cells
Our analyses of previously published data from ccRCC samples revealed decreased SNX5 at the mRNA level, leading us to hypothesize that the reduced expression was the result of altered transcriptional regulation. In order to identify potential transcriptional regulators of SNX5, we performed a reporter assay to identify the regulatory elements that control SNX5 transcription. We analyzed the SNX5 promoter via websites that predicted transcription factor binding sites (JASPAR database). Krüppel-like factor 9 (KLF9) transcription factor binding sites were observed in the promoter of SNX5 (Fig. 9A-9B). KLF9 generally function as transcriptional repressors [19]. Luciferase assay showed that knockdown of KLF9 inhibited the activity of SNX5 (Fig. 9C). In addition, knockdown of KLF9 inhibited the expression of SNX5 in ccRCC cells (Fig. 9D-9E). The binding of KLF9 to the promoter of SNX5 was further con rmed by ChIP assay (Fig. 9F). Furthermore, we found that there a signi cant positive correlation between the expression of KLF9 and SNX5 in ccRCC tissue (Fig. 9G, r=0.36, P<0.01). Therefore, all results suggested that KLF9 binds to the SNX5 promoter and increases its expression in ccRCC cells.

Low expression of SNX5 and KLF9 in ccRCC predicts a poor prognosis
To further explore potential clinical applications of the experimental data, we next assessed the expression of KLF9 in ccRCC based on the TCGA. We found that the expression of KLF9 was downregulated in the ccRCC tissues compared with noncancerous tissues (Fig. 9H). Analysis of the GES15641 cohort showed that the KLF9 mRNA level was downregulated in the ccRCC tissues compared with noncancerous tissues (Fig. 9I). Furthermore, the expression of KLF9 was negative associated with tumor stage, nodal metastasis and overall survival and disease free survival (Fig. 9J-9L).
Based on the expression of SNX5 and KLF9, ccRCC patients were classi ed into the following four groups: low SNX5 and low KLF9 group (n=154), low SNX5 and high KLF9 group (n=107), high SNX5 and low KLF9 group (n=118), high SNX5 and high KLF9 group (n=149). The patients with low expression of SNX5 and KLF9 displayed a worse prognosis than the low SNX5 or KLF9 alone, indicating that the combination of SNX5 and KLF9 has better prognostic value than SNX5 or KLF9 alone (Fig. 9M).

Discussion
In the present study, we found that expression of SNX5 is downregulated and associated with a worse prognosis in ccRCC patients. We found that SNX5 is an inhibitor of proliferation and metastasis in ccRCC cells. Mechanistically, SNX5 facilitates the internalization and inhibits the recycling of CD44 in ccRCC cells, thereby inhibiting EMT of ccRCC cells and exerting a tumor suppressor function.
As a member of the SNX family, SNX5 is involved in endocytosis and internalization of membrane proteins and lipids [20,21]. Endocytosis is an essential component of cell motility. Therefore, endocytosis can in uence the transformed cell behavior, including proliferation and metastasis [22,23].However, role of SNX5 is different in various tumors. Previous studies showed that loss of SNX5 stabilizes internalized growth factor receptors to promote thyroid cancer progression [24]. However, SNX5 decreases the FBW7mediated oncoproteins degradation to promote head and neck squamous cell carcinoma (HNSCC) progression [25]. Our previous studies showed that expression of SNX5 was upregulated and related with poor prognosis in HCC [14]. Therefore, SNX5 plays different roles in different types of human tumors. In this study, we found that expression of SNX5 was downregualted in ccRCC. The expression of SNX5 was closely associated with tumor size, AJCC stage tumor thrombus of inferior vena cava (IVC) and prognosis. Moreover, univariate and multivariate COX regression analysis showed that a low level of SNX5 was associated with worse survival of ccRCC patients. Therefore these ndings indicate that SNX5 may serve as a valuable prognostic factor for ccRCC patients after surgery.
CD44, a non-kinase transmembrane glycoprotein, is widely expressed on cell membrane implicated in many physiological and pathological processes, which include development, in ammation, immune responses, wound healing and cancer progression [26,27]. CD44 overexpression or alternative splicing was described for various types of cancers including ccRCC and associated with invasion, metastasis and resistance to chemotherapeutic drugs [28][29][30].CD44 is a known cargo of clathrin-independent endocytosis [31]. Acylation of CD44 is a critical driving force to CD44 association with lipid rafts, which is essential for the rates of hyaluronan endocytosis and CD44 turnover from cell surface [32].Yu,K.S et al show that receptor-meditated endocytosis by hyaluronic acid for targeting of CD44-overexpressing cancer cells has promising therapeutic prospects [33]. A large endocytic receptor LRP-1mediated internalization of CD44, which plays an important role in the adhesive properties of tumor cells [34]. Others present that chondrocytes regulate the internalization of CD44, and CD44 turnover is modulatedby occupancy with hyaluronan [35]. Here, we demonstrate that overexpression of SNX5 decreased the cell surface expression of CD44 and accelerated internalization of CD44 into the cytoplasm in ccRCC cells. Furthermore, overexpression of CD44 rescued the effects of overexpression of SNX5 on ccRCC cell proliferation and invasion. In addition, the patients with low expression of SNX5 and high expression of CD44 played a worse prognosis than the high SNX5 and low CD44 groups according to TCGA, indicating that the combination of SNX5 andCD44has better prognostic value than SNX5 or CD44 alone. Therefore, together our data suggest that downregulated SNX5 facilitated the localization of CD44 to the cell surface by recycling which is an important mechanism to promote ccRCC cell invasion and metastasis.
By regulating multiple important cellular signaling pathways including β-catenin, TGF-β, etc, CD44 is known to promote tumorigenesis and EMT [36][37][38]. Upregulation of CD44 facilitates EMT-phenotypic change at acquisition of resistance to EGFR kinase inhibitors in lung cancer [39].Our data demonstrate that downregulation of SNX5 decreased the expression of E-cadherin, ZO-1, Claudin-1, and was accompanied by increased the expression of N -cadherin and snail. Furthermore, knockdown of SNX5 promoted the expression of MMP9 in ccRCC cells. Therefore, the knockdown of SNX5 induces an EMTlike phenotypic transition and promotes metastasis in ccRCC. Furthermore, we found that overexpressed SNX5 inhibited TGF-β-induced EMT in ccRCC cells. TGF-β-induced cell migration and invasion was partially reversed by overexpression of SNX5, suggesting that TGF-β may contribute to these processes downstream of SNX5.
Based on the database analyses and our experiments, we nd that KLF9 binds to the SNX5 promoter and increases its expression in ccRCC cells. KLF9 is a zinc nger-containing transcription factor that belongs to the KLF family, and is expressed in various human tissues. KLF proteins have important roles in the regulation of a wide variety of biological processes as well as human diseases, such as cancer [40]. KLF9 acts as a tumor suppressor which has been reported to suppress cell proliferation, migration, invasion [41][42][43]. Previous study also shows that KLF9 is downregulated and negatively associated with poor prognosis in human ccRCC [44]. In this study, we nd that low expression of KLF9 is one of the reasons for the down-regulation of SNX5 in ccRCC. Furthermore, we nd that combination of SNX5 and KLF9has better prognostic value for ccRCC patients. Therefore, downregulated SNX5 by the loss of KLF9 facilitates ccRCC proliferation and metastasis.

Conclusion
Our study shows that SNX5 is a tumor suppressor in ccRCC. Mechanistically, knockdown of SNX5 triggers CD44 internalization and recycling, resulting in increased cell surface CD44 expression and promoting cell proliferation and metastasis. Furthermore, expression of SNX5 is regulated by KLF9 in ccRCC cells. The combination of low SNX5, high CD44 and low KLF9 expression may serve as a predictor of poor prognosis in patients with ccRCC (Fig. 10). These ndings shed new insights into the pathogenesis of ccRCC and may lead to the development of novel targeted therapies. Availability of data and materials All data generated or analyzed during this study are available from the corresponding author upon reasonable request.

Competing Interests
The authors declare that they have no con ict of interest.