Alternatively spliced ANLN isoforms synergistically contribute to the progression of head and neck squamous cell carcinoma

Head and neck squamous cell carcinoma (HNSCC) is a common cancer with high mortality. Anilin actin-binding protein (ANLN) has been reported to be associated with carcinogenesis in multiple tumors. However, the expression pattern and functional effects of ANLN in HNSCC remain to be unclear. Clinical data and online databases were used to analyze the expression of ANLN and its relationship with HNSCC patient survival. Expression of two major splice variants of ANLN was assessed in HNSCC tissues and cell lines. The functional effects and related mechanisms of ANLN isoforms were investigated in HNSCC in vitro and in vivo. Our study showed that patients with high expression of ANLN had a poor prognosis. The two primary isoforms of ANLN transcripts ANLN-201 and ANLN-210 were highly expressed in HNSCC tissues and cell lines. Knockout of ANLN restrained cell proliferation, migration, and invasion of SCC-9 cells. Mechanically, ANLN-201 could interact with c-Myc to keep its protein stability, thereby playing a oncogenic role in HNSCC. ANLN-210 could be transferred to macrophages via exosomes by binding to RNA-binding protein hnRNPC. Exosomal ANLN-210 promoted macrophage polarization via PTEN/PI3K/Akt signaling pathway, thus stimulating tumor growth of HNSCC. ANLN was an independent prognostic factor in patients with HNSCC. Alternatively spliced ANLN isoforms collaboratively promote HNSCC tumorigenesis in vitro and in vivo, which might provide the in-depth role and mechanism of ANLN in HNSCC development.


BACKGROUND
Head and neck squamous cell carcinoma (HNSCC) has been identified as the sixth most common cancer in the world [1,2]. Squamous cell carcinoma (SCC) is the major pathological pattern of head and neck squamous cancer. Although improvements have been made in diagnosis and treatment strategies, patients with HNSCC at advanced stages may develop visceral metastases and present a poor prognosis with low 5-year survival rates [3][4][5]. It is necessary to explore the underlying molecular mechanisms of HNSCC and identify novel effective biomarkers for diagnosis and treatment of HNSCC.
Alternative RNA splicing is a critical step of posttranscriptional gene expression regulation, which could bring about various RNA isoforms from a single gene, plays pivotal roles during normal biological processes such as tissue and organ development [17,18]. RNA splicing is a highly regulated process and accumulating evidence have shown that splicing is frequently occurred in human tumors. It is common that alternative RNA splicing owns different biological functions in cancers and might contribute to malignant cellular transformation [17,19,20]. Studies have suggested that RNA isoforms produced by alternative RNA splicing play key roles in tumor development and thus may provide novel potential therapeutic targets in human cancer treatment.
This study for the first time investigated the expression pattern, potential roles, and regulatory mechanisms of the two transcripts of ANLN in HNSCC, which may provide new perspectives for the treatment of HNSCC.

RESULTS
ANLN transcripts are specifically expressed in HNSCC To clarify the role of ANLN transcripts in HNSCC, firstly we assessed the expression levels of ANLN splice isoforms in HNSCC tumor tissues and adjacent non-tumor tissues. We observed that two splice variants of ANLN, which were ANLN-201 (ENST00000265748.6, 4731 nt) and ANLN-210 (ENST00000457743.1, 757 nt) were significantly expressed in HNSCC tumor tissues compared to the nontumor tissues ( Fig. 1A and B). Next, we examined these two ANLN transcripts in randomly selected ten pairs of HNSCC tumor tissues as shown in Fig. 1C. the expression of ANLN-201 was highly expressed in all these four pairs of tumor tissues compared to the adjacent normal tissues while ANLN-210 was expressed in tumor tissues in only two pairs of HNSCC samples. These results also indicated that ANLN-210 was not expressed in all HNSCC samples. In addition, we detected the splicing variants of ANLN in diverse tumor cell lines. Consistently, the results showed that ANLN-201 was expressed in almost all of the tested tumor cell lines whereas the ANLN-210 was only expressed in the following cell lines, including the hypopharyngeal carcinoma cell line FaDu, the two tongue squamous cell lines SCC-9 and SCC-15, and the laryngeal carcinoma cell line SNU899 (Fig. 1D). Our results suggest that ANLN-201 and ANLN-210 showed subtype-specific overexpressed in tumor cell lines and highly expressed in HNSCC cell lines such as SCC-9.
Subsequently, we analyzed the relationship between ANLN expression and clinical outcomes in 412 HNSCC samples. We found that BMI, differentiation, TNM staging and ANLN expression levels are related to overall survival (P = 0.011, P = 0.023, P < 0.001, and P < 0.001, respectively) ( Table 1). Multivariate analysis showed that BMI, TNM stage, and ANLN 201 expression levels were prognostic factors for overall survival (P = 0.041, P = 0.004, and P = 0.037, respectively) ( Table 1). Kaplan-Meier analysis showed that the upregulation of ANLN 201 expression was significantly correlated with shortened survival of HNSCC patients (P < 0.001, Fig. 1E). Whereas, we found that when the expression value of ANLN 210 increased, the patient's survival period only showed a decreasing trend (P = 0.063, Fig. 1F). This negative association may require a larger sample size to be confirmed. In addition, according to the Kaplan-Meier chart summarized by GEO, EGA, and TCGA (http://kmplot.com/analysis/), the upregulation of ANLN was also significantly correlated with poor overall survival (P < 0.05, Fig. 1G).
ANLN is critical for cell proliferation, migration, and invasion of SCC-9 To investigate the role of ANLN in HNSCC, we established the stable ANLN-knockout cell line in SCC-9 cells by using CRISPR-Cas9 (ANLN-sgRNA SCC-9) for further study ( Fig. 2A). We then assessed the effects of ANLN knockout on the cell malignancy of SCC-9. As shown in Fig. 2B-E, knockout of ANLN directly caused SCC-9 cell proliferation inhibition ( Fig. 2B and C), triggered G2/M cell cycle arrest (Fig. 2D), and suppressed cell migration and invasion capabilities (Fig. 2E). To verify the biological function of these two ANLN isoforms in HNSCC, we conducted the in vitro functional assays by overexpressing ANLN-201 or ANLN-210 tagged with GFP in ANLN-sgRNA SCC-9 cells. The results showed that ANLN-201 overexpression could rescue the suppressed cell proliferation, cell migration, and cell invasion of SCC-9 cells induced by ANLN knockout (Fig. 2F, G). Nevertheless, overexpression of ANLN-210 does not have these similar effects. Next, we observed that both ANLN-201 and ANLN-210 were highly expressed at mRNA level in ANLN-sgRNA SCC-9 cells transfected with GFP-ANLN-201 or GFP-ANLN-210 (Fig. 2H). Interestingly, ANLN-210 expression at protein level was not detected or much lower expressed compared to ANLN-201 ( Fig. 2I and J), suggesting that ANLN-210 mRNA may not be translated into protein effectively. These findings suggest that ANLN plays a key role in cell proliferation, migration, and invasion of SCC-9 and ANLN-201 and ANLN-210 may function at different patterns.
ANLN-201 is critical for the protein stability of c-Myc It is known that there is a conserved binding domain with F-actin at N-terminal of ANLN. To further clarify the variable expression pattern and functions of ANLN alternative splicing in HNSCC in vitro, we analyzed the sequences of these two ANLN splice isoforms. The results showed that compared to the structure of ANLN-201, ANLN-210 missed the domain interacting with F-actin (Fig. 3A). The co-immunoprecipitation results also validated that GFP-ANLN-201 could interact with F-actin whereas GFP-ANLN-210 could not (Fig. 3B). It was reported that ANLN could interact with myc [21]. We confirmed that there was an interaction between GFP-ANLN-201 and Myc but not GFP-ANLN-210 with Myc ( Fig. 3C). In addition, we transfected SCC-9 cells with ANLN-201 siRNA or ANLN-210 siRNA. ANLN-201/ANLN-210 mRNA expression was effectively decreased in siRNA-transfected SCC-9 cells (Fig. 3D). We observed that the expression of Myc at protein level was obviously reduced in ANLN-201-knockdown cells but not changed in ANLN-210-knockdown cells (Fig. 3E). Interestingly, neither ANLN-201 nor ANLN-210 knockdown had any effect on the expression of Myc at mRNA level (Fig. 3F). Next, we found that ANLN-201 knockdown promoted the polyubiquitination of Myc, while ANLN-210 knockdown did not have this similar effect (Fig. 3G). These results suggest that ANLN-201 mRNA but not ANLN-210 mRNA could influence the protein stability of Myc. Furthermore, we tested protein stability experiments by using protein synthesis inhibitor cycloheximide (CHX). As expected, the expression of Myc at protein level in ANLN-sgRNA SCC-9 cells was rapidly downregulated compared to the control. Overexpression of ANLN-201 could retard the protein degradation of Myc with the prolonged treating time compared to the ANLN-sgRNA cells while ANLN-210 could not (Fig. 3H). Taken together, these data indicate that ANLN-201 could interact with Myc to prevent it from degrading, and thus has an oncogenic function in SCC-9.
HNRNPC is identified to be the RNA-binding protein of ANLN-210 mRNA To understand the significance of the high levels of the ANLN-210 mRNA but not ANLN-210 protein, we detected the subcellular localization of the two ANLN splice variants. We found that ANLN-201 mRNA was dominantly located around the nucleus while ANLN-210 mRNA was almost distributed in the cytoplasm in ANLN-sgRNA SCC-9 cells transfected with GFP-ANLN-201 or GFP-ANLN-210 (Fig. 4A). What's more, the RNA stability of ANLN-210 mRNA was remarkably higher than that of ANLN-201 mRNA with actinomycin D treatment (Fig. 4B). Sequence comparison analysis showed that there was extra 54 nucleotides in the sequence that encoded ANLN-210 protein compared with ANLN-201 (Fig. 4C). This extra sequence contains three potential RNAbinding protein sites, including SRSF10, U2AF2, and HNRNPC, analyzed by RBPmap online tool (http://rbpmap.technion.ac.il/) (Fig. 4D). Next, we validated that the predicted protein hnRNPC could effectively bind to ANLN-210 mRNA by RIP assay (Fig. 4E). These data reveal that ANLN splicing isoforms have different subcellular localization and distinct existing patterns.
To further explore the relationship between ANLN-210 mRNA and HNRNPC, we performed the experiments by using si-HNRNPC, si-SRSF10, and si-U2AF2. The expression of these three predicted RNA-binding proteins HNRNPC, SRSF10 and U2AF2 was obviously decreased at mRNA and protein level ( Fig. 5A and B). As expected, the relative level of ANLN-210 mRNA was reduced only in the HNRNPC-knockdown cells, while neither SRSF10 nor U2AF2 knockdown had such effects (Fig. 5B). In addition, decreased HNRNPC expression accelerated instability of ANLN-210 mRNA in SCC-9 cells with actinomycin D treatment compared to the other three groups (Fig. 5C). We subsequently constructed the ANLN-210 mutant probe (ANLN MUT), of which the binding sites to HNRNPC were mutated (Fig. 5D). As shown in Fig. 5E, the expression of ribonucleoprotein (RNP) complex could be detected in cells co-treated with ANLN WT probe and HNRNPC and enhanced with the increased transfection concentration of HNRNPC, whereas the RNP complex disappeared with ANLN-210 MUT group (Fig. 5E). Above all, our results demonstrate that HNRNPC is the RNA-binding protein of ANLN-210 mRNA and thus ANLN-210 mRNA dominantly exists in the cytoplasm of SCC-9 steadily.
Interestingly, ANLN-210 expression was upregulated at protein level in ANLN-sgRNA and HNRNPC-knockdown SCC-9 cells transfected with GFP-ANLN-210 ( Figure S1A and S1B). These findings suggest that HNRNPC may keep ANLN-210 mRNA stability by hindering its translation. Consistently, HNRNPC was highly expressed in HNSCC tissues and cell lines (Figure S1C-S1E). Besides, neither ANLN-201 nor ANLN-210 knockdown in SCC-9 cells could influence HNRNPC expression at mRNA and protein levels ( Figure S1F and S1G). However, ANLN-210 knockdown promoted HNRNPC localized more in the nucleus ( Figure S1H). The immunofluorescence results further confirmed that HNRNPC was located in the nucleus and cytoplasm in control group SCC-9 cells; while almost distributed in the nucleus in ANLN-sgRNA cells ( Figure S1I). ANLN-210 overexpression could reverse the ANLNknockout induced the changed subcellular distribution of HNRNPC but not ANLN-201 ( Figure S1I). The above results indicate that HNRNPC protects ANLN-210 mRNA stability, and in turn, ANLN-210 affects the cellular localization of HNRNPC.
Exosomal ANLN-210 mRNA promotes M2 polarization of macrophages It has been reported that the ANLN 210 mRNA was found in colon cancer cells secreted exosomes [22]. We analyzed the expression of the two ANLN transcripts ANLN-210 and ANLN-201 in SCC-9 cells secreted exosomes. There was no significant difference in expression between ANLN-201 mRNA and ANLN-210 mRNA in SCC-9 cells whereas the expression of ANLN-210 mRNA was much more higher than ANLN-201 mRNA in SCC-9 cells secreted exosomes (Fig. 6A). What's more, the expression of ANLN-201 mRNA was not affected in HNRNPC-knockdown cells secreted exosomes compared to the control (Fig. 6B) while ANLN-210 mRNA expression in exosomes was dramatically reduced when HNRNPC expression was decreased in SCC-9 cells (Fig. 6C). Considering that HNRNPC promoted ANLN-210 mRNA stability and restained its translation, we also performed the experiments by overexpressing ANLN-210 in HNRNPC-knockdown cells. The results showed that the ANLN-210 overexpression enhanced the content of ANLN-210 mRNA in cells despite HNRNPC was knockdown whereas the expression of ANLN-210 mRNA was still downregulated in exosomes (Fig. 6C). These data indicate that HNRNPC could not only maintain the stability of ANLN-210 mRNA, but also promote the release of ANLN-210 mRNA in exosomal forms. Next, we prepared ANLN-210 mRNA enriched in exosomes in two ways. One is to transfect GFP-ANLN-210 into SSC-9 cells (210 O.E.). The other way is to electroporate GFP-ANLN-210 directly into SSC-9-derived exosomes (210 E.T.) ( Fig. 6D and E). We observed that the expression of ANLN-210 mRNA was significantly increased in ANLN-210-electroporated exosomes compared to the ANLN-210-transfected group (Fig. 6F). However, these prepared exosomes rich in ANLN-210 mRNA neither accelerated SCC-9 cell proliferation and migration after co-incubation with SCC-9 cells nor promoted the hepatocarcinoma cell line HepG-2's malignant features ( Fig. 6G and H).
Multiple studies have shown that tumor cells secrete exosomes to accelerate tumor growth by transferring signaling factors to the surrounding cells such as tumor-associated macrophages. To clarify the potential effects of exosomal ANLN-210 mRNA on macrophages, firstly we induced human leukemia mononuclear cell lines THP-1 to macrophages with PMA treatment. Macrophage-like morphological changes were observed in PMA-treated THP-1 cells. The expression of macrophage marker CD68 at mRNA level was significantly increased in THP-1 cells treated with PMA, indicating that THP-1 cells were successfully induced to the M0-state macrophages (Fig. 7A). In addition, we continued to use flow cytometry to determine the markers of macrophages. When THP-1 was treated with PMA, the expression of CD206 increased significantly, reaching 2.0% and 58.5%, respectively (Fig. 7B). Next, we incubated these M0-state macrophages with IL-4, PBS, or exosomes derived from SCC-9-control and SCC-9-210 O.E., or 210 E. T exosomes. Then we analyzed the expression of M2 indicators including Arg-1, CD206, IL-10, and TGF-β together with the M1 markers including iNOS and IL-6 as shown in Fig. 7C. Exosomes from SCC-9-210 OE or 210 E.T. exosomes enhanced the expressions of Arg-1, CD206, IL-10, and TGF-β while had no effect on iNOS and IL-6 expression compared with incubation with PBS or exosomes from SCC-9-control cells (Fig. 7C). Moreover, with locked nucleic acid (LNA) of anti-ANLN-210 mRNA treatment, the M2 polarization effects caused by exosomal ANLN-210 mRNA could be obviously blocked (Fig. 7D). These results suggest that exosomal ANLN-210 mRNA could promote macrophage M2 polarization.
PI3K/Akt signaling pathway is reported to be involved in regulating the polarization of macrophages. To investigate the  (Fig. 7D). Next, we examined the effects of M2 macrophages induced by ANLN-210-rich exosomes on the proliferation, migration, and invasion of SCC-9 cells. The results showed that, M2 macrophages induced by ANLN-210-rich exosomes could significantly promote SCC-9 cells proliferation compared to the control exosomes derived from the ANLN-sgRNA cells (Fig. 7E), which has the similar effects on cell migration and invasion (Fig. 7F). The above results reveal that ANLN-210 mRNA released in the form of exosomes secreted from SCC-9 could activate macrophages through PTEN/PI3K/Akt signaling pathway, leading to M2 polarization thus promotes SCC-9 cell proliferation, migration, and invasion in a feedback manner.

Alternatively spliced ANLN isoforms synergistically promote tumor progression in vivo
To further evaluate the role of the two splicing isoforms of ANLN in tumor development, we co-injected ANLN-sgRNA SCC-9 cells transfected with GFP-ANLN-201 or GFP-ANLN-210 or control, and macrophages treated with ANLN-210 mRNA-rich exosomes into the nude mice. As shown in Fig. 8A, the tumor volumn of mice xenografts was increased significantly. Among them, the tumor volume in the group that ANLN-sgRNA SCC-9 cells overexpressing ANLN-201 mixed with macrophages treated with ANLN-210 mRNA-rich exosomes was the most significantly increased (Fig. 8A). Consistently, Ki67-positive together with CD163-positive stainings were the most in the group that ANLN-sgRNA SCC-9 cells overexpressing ANLN-201 mixed with macrophages treated with ANLN-210 mRNA-rich exosomes (Fig. 8B). Taken together, the results reveal that the two splicing isoforms of ANLN synergistically promote tumor progression in vivo in different ways.

DISCUSSION
ANLN is expressed ubiquitously in human tissues and overexpressed in multiple human tumors. However, the expression pattern and molecular mechanism of ANLN in HNSCC remain unclear. In First of all, we proved that patients with high ANLN expression have a poor prognosis. Further research shows that ANLN at mRNA level was highly expressed in HNSCC tissues and SCC-9 cells with two splicing variants (ANLN-201 and ANLN-210). Next, we discovered that loss of ANLN inhibited cell proliferation, migration, and invasion of SCC-9 cells. Furthermore, we revealed an in-depth mechanism of these two ANLN transcripts cooperatively regulating tumorigenesis of HNSCC in two ways. That is, on one hand, Myc was identified to directly interact with ANLN-201 and keep its protein stability, thus affected cell proliferation, migration, and invasion of SCC-9 cells. On the other hand, HNRNPC was found to be an RNA-binding protein of ANLN-210 and keep RNA stability of ANLN-210. Highly expressed ANLN-210 mRNA delivered to tumor-associated macrophages (TAMs) via SCC-9 cells secreted exosomes, causing activation of macrophages polarized to M2 phenotype via the canonical PI3K/Akt signaling pathway, thus promoting tumor growth and metastasis of HNSCC. These cooperative regulations of two ANLN splicing isoforms highlighted the biological functions and regulatory mechanism of ANLN in HNSCC tumorigenesis, which might provide new perspectives for cancer therapy. HNRNPC is one of the family members of ubiquitously expressed heterogeneous nuclear ribonucleoproteins (hnRNPs), which located in the nucleus and plays critical roles in posttranscriptional regulation including roles in alternative splicing, stability, and translation [23][24][25]. It was previously reported that hnRNPC regulated cancer-specific alternative cleavage and polyadenylation (APA) profiles in colon cancer, which has a critical role in cancer progression [26]. In this study, HNRNPC was   predicted to be one of the RNA-binding proteins of ANLN-210. Next, the interaction between ANLN-210 and HNRNPC was confirmed by RNA-IP. What's more, we found that HNRNPC could maintain ANLN-210 RNA stability as well as prevent its protein translation. In turn, ANLN-210 knockdown promoted HNRNPC accumulation in the nucleus. However, whether HNRNPC is a critical novel regulator of cancer-specific APA for ANLN-210 in HNSCC remains to be explored.
It is known that tumor growth and progression are strongly associated with tumor microenvironment, of which solid tumors infiltrate around immune cells. Macrophages, called tumorassociated macrophages (TAMs) are rich in tumor tissues. Multiple studies have shown that macrophages can affect various aspects of tumor development. M2 macrophages play a pivotal role in promoting tumor growth, metastasis, and angiogenesis [27,28]. Previous studies reported that tumor-associated macrophages of the M2 phenotype contributed to progression in gastric cancer with peritoneal dissemination [29]. In bladder cancer, tumorinfiltrating M2 macrophages driven by specific genomic alterations were associated with prognosis [30]. In ovarian cancer cells, miR-21 modulated the polarization of macrophages and increased the effects of M2 macrophages on promoting the chemoresistance of ovarian cancer [31]. In this study, we observed that ANLN-210 overexpression by exosomes activated macrophage to the M2 phenotype as marked by the increased expression of M2 markers Arg-1, CD206, IL-10, and TGF-β. PI3K/Akt signaling pathway has been widely recognized in multiple cell types and functions and emerged as canonical and central regulators of macrophages activation [32]. In this study, we found that exosomal ANLN-210 overexpression regulated M2 polarization by inhibiting the common target PTEN via activating the PI3K/Akt signals.
Tumor-derived exosomes, which are found in all body fluids delivering molecular and genetic messages from tumor cells to normal or other abnormal cells around. Tumor-derived exosomes play key roles in the oncogenic transformation, which have biological and clinical significance for cancer development, cancer progression, and even cancer therapy [33]. The attributes of tumor-derived exosomes might be packaged proteins, lipids, or nucleic acids, which could be regarded as biomarkers for cancer diagnosis, prognosis, and monitoring treatment responses [34,35]. In 2008, a study reported that there were 4700 distinct mRNAs selectively packaged in exosomes derived from glioblastoma samples [36,37]. It is well known that mRNAs carried by exosomes are involved in critical cellular activities such as cell proliferation, migration, invasion, metastasis, EMT, and so on. The present study demonstrates that exosomal ANLN-210 mRNA secreted by SCC-9 promotes macrophages polarization by targeting PTEN via PI3K/ Akt. Furthermore, M2 macrophages promoted tumor growth and metastasis of HNSCC. However, it remains to be investigated whether ANLN-210 induces the activation of the PI3K/Akt pathway in the form of RNA or protein.
In summary, our findings demonstrate that the coordinate regulatory networks of alternatively spliced ANLN isoforms in HNSCC tumor growth and development, which might highlight the perspective for therapeutic strategies of HNSCC. ANLN was an independent prognostic factor in patients with HNSCC. ANLN-201 dominantly binds to Myc and exists in the nucleus in the form of protein, which can be effectively prevented from degradation and thus play a role in promoting HNSCC cell proliferation, migration, and invasion. ANLN-210 is found to bind mainly to HNRNPC, maintains its RNA stability, and secreting into the environment in the form of exosomes. Exosomal ANLN-210 mRNA promotes macrophage polarization via PTEN/PI3K/Akt signaling pathway, which in turn facilitates HNSCC tumor growth (Fig. 8C). Thus, the splicing variants of ANLN collaboratively regulate HNSCC tumorigenesis in two ways.

MATERIALS AND METHODS Patients and cancer tissues
This study was approved by the Ethics Committee of Harbin Medical University, China. All patients provided written informed consent to participate in the study. The research methodology meets the standards set out in the Declaration of Helsinki. From January 2013 to December 2015, 412 HNSCC tissue samples and some normal tissues were collected from the Affiliated Cancer Hospital of Harbin Medical University. All tumor tissues were independently diagnosed as HNSCC by two pathologists. Patients receiving chemotherapy or radiation therapy were excluded. Overall survival (OS) is defined as the time interval between the date of definitive diagnosis and the date of death or the last follow-up.

Quantitative real-time PCR (qRT-PCR)
Total RNA was extracted with TRIzol reagent and reverse transcribed to cDNA using a PrimeScript RT reagent kit (TaKaRa, Tokyo, Japan) according to the manufacturer's instructions. Quantitative real-time PCR (qRT-PCR) was performed using the SYBR Green qPCR Master Mix (Takara). The sequences of primers were listed in Table 2.

Northern blot
RNA was extracted and heat denatured at 65°C for 15 min. The prepared RNA samples were loaded in the gel in the MOPS running buffer, which was electrophoresed at 5-6 V/cm until the bromophenol blue (the fastermigrating dye) migrated at least 2-3 cm into the gel. The gel was visualized on a UV transilluminator. The 28S rRNA band should be approximately twice as intense as the 18S rRNA band.

Cell cycle analysis
The harvested SCC-9 cells were re-suspended with PBS and fixed overnight with 70% ethanol at 4°C, then incubated with PI staining reagent at room temperature for 30 min. The treated cells were analyzed using a FACS Calibur system (BD Biosciences, San Jose, CA, USA).

Cell migration and invasion assays
Cell migration and invasion assays were performed using transwell insert chambers (BD Biosciences, USA). SCC-9 cells were transfected as experiments designed. The transfected SCC-9 cells in serum-free medium were seeded into the top chamber. The medium containing 20% FBS was added to the lower chambers. After routinely cultured for the appropriate time, cells on the lower membrane surface of the top chamber were fixed with methanol and stained with DAPI. For invasion analysis, the top chambers were coated with the matrigel (BD Biosciences, USA). Immunofluorescence SCC-9 cells were fixed and permeabilized, then incubated with GFP Rabbit antibody (ab290, Abcam) for 3 h at 4°C. After washing three times with PBS, the cells were incubated with Alexa Fluor 488 conjugated Goat Anti-Rabbit IgG (H&L) (ab150077; Abcam) for 1 h at room temperature in the dark place. After washing three times with PBS, cells were treated with

Gel electrophoretic mobility shift binding assay (EMSA)
Protein-RNA interactions were evaluated using the LightShift Chemiluminescent RNA EMSA kit (Thermo Scientific). The constructed T7 promoter-ANLN-210 WT or ANLN-210 MUT DNA fragments were used as the templates for in vitro transcription to synthesize plenty of RNA. The 3′end of synthesized RNA was biotinylated using RNA 3′End Biotinylation Kit (Pierce, Rockford, IL, USA), which was used as the probe analyzed for EMSA. The recombinant HNRNPC protein was obtained from Origene (TP315956). Biotin-labeled RNA and a dose-dependent amount of HNRNPC were mixed in the reaction buffer. The samples were electrophoresised on the 5% nondenaturing polyacrylamide gel. The signals were detected using the Luminescent image analyzer (FluorChem M).

Animals
Male BALB/c nude mice (6-8 weeks) were raised strictly in the pathogenfree animal house in accordance with feeding standards, which was approved by the Institutional Animal Care and Research Advisory Committee of Affiliated Tumor Hospital of Harbin Medical University.

Xenograft models
To assess the effects of exosomal ANLN-210-activiated M2 polarized macrophages on HNSCC tumor growth in vivo, 1.5 × 10 7 SCC-9 cells were mixed in a 4:1 ratio with the macrophages transfected with exosomes with different treatment as experiment designed. Then the cell mixture was coinjected into the flanks of the nude mice. After 3 weeks, the mice were euthanized and the tumors were obtained and fixed in formaldehyde for and IHC staining.

Statistical analysis
All the experiments were repeated at least three times. The data were analyzed using SPSS 20.0 and GraphPad Prism 7. The P-value <0.05 was considered to be significant.

DATA AVAILABILITY
The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.