Schwann Cell-Derived CCL2 Promotes the Perineural Invasion of Cervical Cancer

Perineural invasion (PNI) has guiding significances for nerve preservation in cervical cancer, but there is no definite marker indicating PNI. Two cervical cancer cell lines (HeLa and ME-180) showed significant abilities to migrate along neurites in vitro and in vivo. Morphological observation revealed that Schwann cells (SC) arrived at the sites of cervical cancer cells before the onset of cancer metastasis. We used high-throughput antibody array to screen the signals mediating the interaction of nerve cells and cancer cells and found the high expression of CCL2 in dorsal root ganglion (DRG). Meanwhile, serum CCL2 showed a notable raise especially in cervical adenocarcinoma. SC-derived CCL2 bound to its receptor CCR2 and promoted the proliferation, migration, invasion, and epithelial-mesenchymal transition (EMT) of cervical cancer cells. In turn, cancer cell-derived signals triggered the expression of metalloproteinases (MMPs) including MMP2, MMP9, and MMP12 in SCs, promoting SCs to dissolve matrix. These data demonstrated that the cancer-nerve crosstalk formed a tumor microenvironment (TME) that facilitated to PNI. We identified the CCL2/CCR2 axis as a potential marker to predict the PNI and affect the nerve preservation for cervical cancer.

and cancer cells, and lead to both cancer progression and nerve expansion, thus causing PNI (13). The abnormal release of chemotactic factors might be correlated with PNI in cancers (14,15). Chemokine (C-C motif) ligand 2 (CCL2), also known as monocyte chemoattractant protein-1 (MCP-1), recruits monocytes and macrophages to the sites of inflammation (16). Previous studies have revealed that CCL2 secreted by the nerves facilitates pancreatic ductal adenocarcinoma invasion by binding to its receptor (CCR2) expressed on the membrane of cancer cells or macrophages (17,18). Moreover, CCL2 was regarded as a molecular target of PNI in prostate cancer (18). CCL2 is related to both primary tumor development and metastasis in various cancers including cervical cancer (14,19). Thus, the role of CCL2 and its receptor in PNI of cervical cancer is an emerging field of research.
Schwann cells (SCs) are a major component of the peripheral nerves and play an important role in promoting axon regeneration during repair (20). A recent study showed that SCs were highly cancer-affine cells that migrated toward cancer cells even before invasion in pancreatic cancer. After contacting with SCs, pancreatic cancer cells move along neurites, during which physical contact and cellular signals play significant roles (21). Previous studies have reported that SCs secreted soluble factors including chemokine and adhesion molecules, affecting pancreatic cancer cells movement (22)(23)(24); however, no studies to date have explored the potential role of SCs in cervical cancer.
In this study, we explored the role of CCL2, identified its association with SCs and elucidated its potential clinical value in predicating PNI and selecting the appropriate surgical method for cervical cancer.

Gene Set Enrichment Analysis
All raw genomic data and clinical data were obtained from the GEO (Gene Expression Omnibus) database. We obtained four RNA sequence datasets with the keyword "perineural invasion" and "tumor" (GSE103479, GSE86544, GSE102238 and GSE7055). Gene Set Enrichment Analysis (GSEA) was performed to seek common pathways of different cancers using PNI as the criteria for sample classification.

Cell Migration and Invasion Assays
Cell migration and invasion assays were performed using 8.0 µm pore transparent polyethylene terephthalate inserts (Corning Inc., Glendale, AZ, USA) in 24-well plates. 1 × 10 5 cells in 0.2 mL of FCS free media were added to each of the inserts, with or without Matrigel, while DRG or RSC were placed in the bottom with 0.6 ml of medium supplemented with 10% FCS. After 20 h for migration assay and 36 h for invasion assay, the membranes were fixed with 4% polyoxymethylene at room temperature for 30 min and then stained with crystal violet staining solution (YEASEN, Shanghai, China) for 30 min. Five random fields were counted at ×10 magnification. The cells for each membrane were quantified by counting five random fields at ×20 magnification.

Wound Healing Assay
Cancer cells were cultured with complete medium until 80% confluence in 6-well plates. Then slowly scratch the monolayer across the center of the well with a 200 µl pipette tip. After washing twice, fresh medium was added into the well. Images of a scratch were captured at a 0, 24, and 48 h and processed using Image J software.

In vitro Neural Invasion Assay
A Matrigel/DRG model in vitro was constructed by Huyett et al. (25) and was frequently used to investigate the interaction between nerve cells and cancer cells. DRG are harvested from the spinal column of a sacrificed Sprague Dawley rat and placed in the center of 2.5 µl of matrix. Cancer cell lines were placed peripherally around the matrix 2 days later. Cellular movement was detected by confocal microscopy at a 24 h interval.

Histological Analysis
The acquisition protocol was approved by the Institutional Ethics Committee of the International Peace Maternity and Child Health Hospital (IPMCH). Twenty samples with PNI and 36 samples without PNI collected between 2012 and 2018 were utilized in this research. These tissues were embedded in paraffin and then cut into 4 µm sections. The sections were stained with haematoxylin & eosin (H&E). For immunohistochemical assay, sections were incubated with a CCR2 antibody at 4 • C overnight followed by secondary antibody conjugated with HRP. The images were obtained by microscopy (Leica, Germany). The positive nerve fibers were counted in a blinded fashion in 10 representative fields.

Flow Cytometry
The HeLa or ME-180 cells were incubated in 1 mL of diluted CCR2 (357208, Biolegend) and Ki67 antibody (CST-9449S, Cell Signaling Technology) on ice for 30 min after being harvested, fixed, washed, and blocked. Then, secondary antibodies conjugated with Alexa Fluor R 488 and Alexa Fluor R 594 were added into the buffer and the samples were measured by FACS Calibur flow cytometry (BD, NJ, USA). Data were processed by FlowJo software (LLC, Ashland, USA).
For immunocytochemistry, the cells were fixed, washed, blocked, and incubated with primary antibodies against pancytokeratin or CCL2 (ab9899, Abcam) overnight at 4 • C. Staining was detected with Alexa Fluor R 594. The cells were counterstained with DAPI.
The slides from mice were incubated with MMP9 (10375-2-AP, Proteintech) and MMP2 (10373-2-AP, Proteintech) and then a secondary antibody conjugated with Alexa Fluor R 594 was added in. The images were obtained by fluorescence microscope (Leica, Germany).

ELISA Assay
Serum samples from 15 normal cervical samples and 33 cervical cell carcinoma samples were obtained from IPMCH. Besides, cell culture media under different conditions were collected and prepared. The detailed procedures were conducted in accordance with the protocol in the MCP-1 immunoassay kit (EK0902, BOSTER).

Lentivirus Infection
HeLa and ME-180 cells were implanted in the 6-well plates at appropriate densities. The short hairpin RNAs (shRNA) against CCR2 (sh7732, sh7733, sh7734) and the negative control shRNA were obtained from Obio Technology (Shanghai, China) and were added into the wells. Puromycin was used to select stable transfected populations of cells.

In vivo Perineural Invasion Assay
The sciatic nerves were surgically exposed after anesthetizing athymic nude mice with isoflurane. The mice were grouped randomly into five groups (PBS, HeLa, ME-180, HeLa shNC, and HeLa shCCR2) of 9 mice each. 3 × 10 5 cells of every type in 3 µL PBS was injected into the distal sciatic nerve under the epineurium using a 10 µL Hamilton syringe. Five weeks later, the mice were euthanized to observe the metastasis and excise their sciatic nerves to measure PNI-related indexes. Then these sciatic nerves were cut into 8-µm-thick sections. This study was approved by the Institutional Ethics Committee of the International Peace Maternity and Child Health Hospital (IPMCH), number: [GKLW] 2017-125.

FITC-Phalloidin Staining and CellTracker CM-DiI Labeling
FITC-phalloidin staining was performed to detect the microfilaments and cytoskeletal reorganization insides the cells. After being washed, cells were fixed in 4% formaldehyde at room temperature for 15 min and permeabilized using 0.1% Triton X-100 in PBS for another 10 min. Cancer cells were incubated with FITC-phalloidin working solution (C1033, Beyotime) for 1 h. Cells were counterstained with DAPI. ME-180 cells were incubated with red fluorescent CM-DiI (40718ES50, YEASEN) for 30 min, and then the CM-DiI was washed with PBS. All images were captured via confocal microscopy.

Statistical Analysis
Data was processed using GraphPad Prism 7 software and presented as mean ± standard error of mean (SEM). Comparisons between different groups were using two-tail unpaired Student's t-test or one-way analysis of variance (ANOVA). Differences were considered significant if the P < 0.05.

The Occurrence of PNI in Cervical Cancer
Perineural invasion, a phenotype that cancer cells invade into the perineural space of local peripheral nerves and contact to the endoneurium directly, could be observed in specimens of cervical cancer ( Figure 1A). We assessed the potential of PNI in seven cervical cancer cell lines including HeLa, ME-180, SiHa, CaSki, C33A, MS751, and HCC94 by cocultivation with DRG separately. After 2 days of cocultivation, HeLa and ME-180 cells showed their notable ability to interact with DRG neurites, whereas the other five cell lines had little or no interaction with DRG (Supplementary Figure 1A). We counted the contact of neurites and cancer cell clusters after 3 days of cocultivation ( Figure 1B, Supplementary Figure 1B). HeLa and ME-180 were prone to PNI, SiHa, and CaSki had inferior abilities to interact with nerve cells. In contrast, C33A, MS751, and HCC94 demonstrated low contact with neurites. Confocal imaging of neurites stained with the pan neuronal marker PGP9.5 and cancer cells stained with pan-cytokeratin (pan-CK) revealed the close touch between the DRG and HeLa cells. After 3 days of cocultivation, HeLa cells had already arrived at the DRG in the center of Matrigel along the neurites (Figure 1C). Marking ME-180 cells with CellTracker CM-DiI and staining neurites with neurofilament-heavy (NF-H) antibody, we found that ME-180 cells could also spread along neurites toward DRG (Supplementary Figure 1C).
In order to further analyze the frequency of PNI of cervical cancer cells, HeLa and ME-180 cells were injected into the mouse sciatic nerves. Four of nine and three of nine mice injected with HeLa and ME-180 cells, respectively showed perineural invasion. Cancer cells colonized and spread along the nerve fibers, which were validated by H&E staining of sciatic nerves and double immunofluorescent staining of PGP9.5 and pan-CK ( Figure 1D). These results showed the occurrence of PNI from different levels.

The Interaction Between Nerve Cells and Cervical Cancer Cells
The signals in the DRG medium (DM) markedly enhanced the migration and invasion of HeLa and ME-180 cells (Figure 2A,  Supplementary Figure 2A). In turn, the continuous confocal   recording over a 12 h period indicated that SCs were firstly triggered and moved toward HeLa cells before the onset of cancer invasion. Upon contact, SCs induced HeLa cell dispersion and movement directionally toward the DRG (Figure 2B). Coculturing with ME-180 cells, SCs were activated and arrived at the sites of cancer cells (Supplementary Figure 2B). There was some difference that SCs linked with each other and more time was needed to induce ME-180 cells movement comparing to HeLa cells. Based on these observations, SCs as primary, non-neoplastic cells were activated, arrived at the sites of cancer cells, attracting cancer cells spread toward DRGs.
Rat Schwann cells (RSC96) were identified with s100β antibody and used in experiments (Supplementary Figure 3A). PNI is involved in SCs and cancer cells motility, which requires an adaptable microenvironment to degrade the ECM. ECM integrity is regulated by the balance between matrix metalloproteinases (MMPs) and their inhibitors (26). In the co-culture supernatant, the concentration of MMP9 markedly increased (Supplementary Figure 3B). SCs displayed increased expression of MMP-2, MMP-9, and MMP-12 at the protein level after co-cultivation with cervical cancer cells, whereas no change in the expression of these MMPs was detected in HeLa and ME-180 cells, differing from previous report about pancreatic adenocarcinoma cells (23) (Figure 2C,  Supplementary Figure 3C). Immunofluorescent analysis also displayed the increased expression of MMP9 and MMP2 of sciatic nerves after being injected with HeLa cells or ME-180 cells (Supplementary Figure 3D). The metalloproteases, especially MMP2 and MMP9 were reported to degrade type IV collagen (27). Therefore, we detect the expression of type IV collagen in both the PNI and non-PNI sites (Supplementary Figure 3E). Cancer cells next to a peripheral nerve had a lighter staining of type IV collagen than that distant from nerves. These results indicated that SCs and cervical cancer cells interacted with each other.

The CCL2/CCR2 Axis Enhances the Proliferation, Migration, and Invasion of Cervical Cancer Cells
To identify common pathways involved in PNI, the raw gene expression data of several PNI-related cancers including colon cancer, head and neck cancer, pancreatic ductal adenocarcinoma, and prostate cancer (GSE103479, GSE86544, GSE102238, and GSE7055) were downloaded and conducted GSEA. Cytokine receptor interaction was selected for further study not only for its significant P-value in colon cancer and prostate cancer ( Figure 3A), but also the enrichment in cytokine production and cytokine-mediated signaling pathway in head and neck cancer patients diagnosed with PNI (28). A cytokine array kit was used to screen molecules involved with PNI and CCL2 was identified as the target among 34 cytokines for its upregulation in cocultivation group (Figure 3B). The monocytes recruited by CCL2 were found to mediate PNI of pancreatic cancer (17). CCL2 was mainly produced by DRGs and the levels of CCL2 increased after coculturing ME-180 cells and DRGs (Supplementary Figure 4A). SCs upregulated CCL2 after nerve injury, which then attracted macrophages into the nerves (29). To identify the source of CCL2 during PNI, we assessed the expression of CCL2 in SCs. After cocultivation with HeLa and ME-180 cells, the expression of CCL2 in SCs and its receptor CCR2 on the membrane of cancer cells both increased (Figures 3C-F, Supplementary Figure 4B).
We next used flowcytometry to detect HeLa and ME-180 cells marked by CCR2 and Ki67 antibodies, and found significant increases in the proportion of CCR2 − KI67 + , CCR2 + KI67 − , and CCR2 + KI67 + cells in the co-culture group (Figures 4A,B). The majority of CCR2 + positive cells were in the proliferative phase, which means that SCs might affect cellular proliferation through CCR2. CCL2 at different concentrations (50, 100 ng/ml) resulted in a rise in the wound healing percentages of ME-180 cells, suggesting CCL2 as a chemoattractant of ME-180 cells (Supplementary Figures 4C,D). To investigate whether CCL2 is required for movement of HeLa and ME-180 cells, we added CCL2 and CCR2 antagonist, RS102895 into the lower layer of the Boyden chamber as required. The ectopic CCL2 stimulation (50 ng/ml) promoted the migration and restored the mobility interrupted by RS102895 of ME-180 cells rather than HeLa cells, whereas RS102895 could reduce the number of SCs-induced HeLa and ME-180 cells (Figures 4C-E). These results reveal that CCL2 plays an important role in inducing cervical cancer cells movement toward nerve cells.

CCL2 Induces Cervical Cancer Cells EMT by Binding CCR2
EMT is regarded as a key step during the cancer progression from primary site toward metastasis (30). SCs induced the downregulation of ZO1 and the upregulation of Snail in HeLa cells, while SCs mainly upregulated the expression of Slug and Twist as well as downregulated the expression of ZO1 in ME-180 cells ( Figure 5A). As shown in Figure 5B, CCL2 at the concentration of 50 ng/ml stimulating HeLa cells increased the expression of Snail and decreased the expression of ZO1 at the protein level. In contrast, both 25 and 50 ng/ml of CCL2 decreased the expression of ZO1 and 10 ng/ml of CCL2 increased the expression of Slug and Twist in ME-180 cells. Transfection of CCR2 shRNA significantly reduced CCR2 expression ( Figure 5C). The EMT process were nearly reversed after interfering with the expression of CCR2, indicating that the CCL2-induced EMT was CCR2 dependent. A slight difference in the expression of E-cadherin, N-cadherin and Claudin-1 was also found in ME-180 cells (Figure 5D).
HeLa and ME-180 cells both showed morphological changes after co-cultivation with SCs, becoming slenderer with a fibroblast-like appearance ( Figure 5E). FITC-phalloidin staining was performed to detect the cytoskeleton. ShRNA-mediated CCR2 interference caused the collapse of the cytoskeletal organization and reduced the pseudo foot formation compared to the shNC group in both cell lines (Figure 5F).

CCL2/CCR2 Contributed to PNI and Induced EMT of Cervical Cancer in vivo
The ability of CCR2 blocking PNI in vivo was assessed using a neural invasion model with balb/c nude mice. HeLa shNC and HeLa shCCR2 cell suspensions were injected into the sciatic nerves in the left. Five weeks after injection, 4 of 9 mice in two groups showed perineural invasion and the sciatic nerve scores and sciatic nerve index of the left hind limb were assessed. In the shCCR2 group, the sciatic nerve score was dramatically decreased compared with the shNC group and accompanied by an increase in the sciatic nerve function index (Figures 6C,D). Following euthanasia, the tumors were resected for histologic examination (Figure 6A). The tumor diameters at 2, 4, and 6 mm from the implantation site were measured the mean values of which were regarded as the sciatic nerve diameter. The diameter of sciatic nerve in shNC group was significantly larger compared to the shCCR2 group (Figure 6B), suggesting that the CCL2-CCR2 axis mediated the progression of cervical cancer cells along the sciatic nerve.
Next, we analyzed the expression of CCR2 and EMT-markers including E-cadherin, N-cadherin, Slug, and Snail. After CCR2 was silenced, the expressions of Snail and N-cadherin decreased and E-cadherin increased (Figure 6E). These results demonstrate that blocking the CCL2/CCR2 signal pathway could inhibit cervical cancer growth and decrease the invasion distance along the sciatic nerve.

Expression and Clinical Significance of CCL2/CCR2 in Cervical Cancer
The serum values of CCL2 in normal cervical samples and cervical cell carcinoma were 48.33 ± 2.618 (pg/ml) and 233.7 ± 54.26 (pg/ml), respectively ( Figure 7A). Due to a fluctuating range in CCL2 values, we further subdivided the tumor type to cervical SCC and AC. The levels of CCL2 in the serum of cervical AC samples were significantly upregulated compared to normal samples (417.9 ± 115.5, P = 0.0014), whereas no significant change in the CCL2 values was detected between normal samples and cervical SCC, which might indicate that CCL2 is a suitable marker of PNI in cervical AC (Figure 7B).
The results obtained from the cancer genome atlas (TCGA) showed that enhanced expression of CCR2 might correlate with poor overall survival in human cervical cancer ( Figure 7C). IHC staining of CCR2 protein was performed on tissue sections from 16 human normal cervical samples, 20 cervical non-PNI cancer samples, and 20 with PNI. The expression of CCR2 was increased in the samples with PNI compared to the non-PNI and normal groups (Figures 7D,E). These data suggested that the detection of CCL2/CCR2 might contribute to early detection of PNI in cervical cancer.

DISCUSSION
PNI is gradually paid more attention to with its clinical significance. Neoplastic cells infiltrating into the perineurium space were spared by tumor resection, leading to local recurrence (31). PNI has been reported as an important danger factor for independent survival, indicating poor prognosis for many malignancies, including prostate cancer (32)(33)(34), colon cancer (35)(36)(37), head and neck cancer (38), gastric cancer (39), and cervical cancer (40). In cervical cancer, PNI could be considered as an indication guiding the operation and postoperative adjuvant therapy (41). Nerve-sparing radical hysterectomy (NSRH) preserves the pelvic autonomic nerves, thus leading to a much improved quality of life, comparing with Radical hysterectomy (RH) (42,43). Given the association between PNI and poor prognosis of cervical cancer patients, an appropriate marker of PNI is in an urgent need. Our results demonstrated that the CCL2/CCR2 axis played a crucial role in PNI and promised to be a marker guiding the selection of NSRH. Patients with AC had a worse survival compared to SCC (44,45). Our results found that the serum CCL2 levels were increased in cervical cancer samples, especially in patients with AC. Moreover, the CCL2/CCR2 axis mediated PNI of cervical cancer in vitro and in vivo. Previous studies have reported that the CCL2 mRNA expression in cervical carcinoma cells were related to local recurrence and distant metastasis significantly and the absence of CCL2 expression indicated an increased survival (46). The upregulation of CCL2 in the serum of patients of AC may partially explain its poor prognosis in contrast with SCC and detailed mechanism needs further elaboration.
Chemokines contribute to the mechanisms underlying PNI. The chemokine CCL2, not only recruits immune cells including monocytes, dendritic cells, natural killer (NK) cells memory T cells and dendritic cells (47), but also promotes changes in neuronal excitability and nerve repair (48). CCL2 and its receptor CCR2 were reported to play crucial roles in cancer metastasis, accelerate cancer cell growth, and promote their colonization at metastasis sites (49). CCL2 is mainly secreted by SCs to attract macrophages in response to nerve injury (50). The expression of CCR2 could be detected in the membrane of cancer cells (51,52). Our results indicated that the migration and invasion of HeLa and ME-180 cells were both significantly increased after co-cultivation with SCs, which could be interrupted by a CCR2 antagonist. CCL2 also influenced cellular proliferation and morphological changes to support cervical cancer cell migration toward SCs. Blocking CCR2 induced the collapse of the cytoskeletal organization and mesenchymal to epithelial transition in vitro and in vivo. These results identified CCL2 as a chemotactic factor attracting cancer cells toward nerves and ultimately causing their metastasis along the neurites.
PNI is the result of an active crosstalk between nerves and cancer involving in many molecules, resulting in neuron outgrowth and fueling tumor progression (12,53,54). We found that SCs could arrive at cervical cancer cell sites and then induce them metastasis. Meanwhile, we detected high expression of CCL2 in SCs, while cervical cancer cells expressed lowly. Our study demonstrated that SC-derived CCL2 modulated the PNI of cervical cancer. As the observation of the movement of SCs, there must be signals derived from cervical cancer cells attracting SCs movement. In a recent study, CXCL12 secreted by cancer cells bound to its receptors and attracted receptor-positive SCs to pancreatic cancer cells, thereby initiating PNI (55). A similar effect on nerves by cancer cells were reported in the brain metastasis of breast cancer. Microglia emerged at the site of breast cancer and enhanced the invasion of cancer cells into the brain and then spread away when co-cultured with breast cancer cells (56). The signals from cancer cells during PNI require further investigation.
As endopeptidases, the MMP family plays a role in ECM degradation and tissue remodeling (11). MMP2 and MMP9 belong to the gelatinase subgroup of MMPs, which degrade type IV collagen (57). MMP12 is a proteinase mainly secreted by macrophages and inhibits inflammation through regulation of the CCL2/CCR2 signal axis (58). A previous study reported that MMP2 and MMP9 were mainly secreted by pancreatic cancer cells to modulate the neural cancerous microenvironment (23). SCs could also secret MMP2 and MMP9 to support axonal regeneration (59). Here, we confirmed that cervical cancer cells upregulated the expression of MMP2, MMP9, and MMP12 of SCs. The degradation of matrix could not only provide tunnels or tracks for SC movement but also eliminate tissue barriers and form a TME for cancer metastasis.

CONCLUSIONS
In summary, a reciprocal interaction between SCs and cervical cancer cells is revealed. SCs arrive at the site of cancer cells and secrete CCL2 as a strong chemoattractant which then induces CCR2 + cervical cancer cells to move along neurites. Conversely, cervical cancer cells upregulate the expression of MMPs in SCs to generate a suitable TME for SCs movement (Figure 7F). The CCL2/CCR2 axis might provide a prospective marker to predict PNI and affect NSRH indications.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/ Supplementary Material.