WDR5 drives the development of cervical squamous cell carcinoma by inducing epithelial-mesenchymal transition and cancer-associated broblasts formation

WD repeat domain 5 (WDR5) has been indicated to be involved in tumor progression, however, its role in cervical cancer (CC) has not been investigated yet. A total of 350 pairs of CC tissues and para-carcinoma tissues (PCT) were collected. Primary human cervical epithelial cells (hCECs) and cancer-associated broblasts (CAFs) were isolated from PCT and cancer tissues. MM102 was used to block the interaction between WDR5 and mixed lineage leukemia protein-1 (MLL1), and it was used in vivo to investigate its therapeutic value. WDR5 was up-regulated in cervical squamous cell carcinoma (CSCC) tissues compared to that in PCT. C-X-C motif chemokine ligand 8 (CXCL8) was indicated to be the target gene of WDR5. Highly expressed CXCL8 promoted epithelial-mesenchymal transition (EMT) to form CAFs, and enhanced the cytokine secretions in CAFs to promote CSCC progression. CXCL8 expression was regulated by the interaction between WDR5 and MLL1, and blocking the interaction between these two proteins using MM102 signicantly suppressed tumor growth in mice models. WDR5 plays a key role in CSCC progression by inducing CXCL8 expression and promoting the transformation of CAFs from epithelial cells.


Introduction
Cervical cancer (CC) is the third most common cancer and the fourth leading cause of cancer death in women [1]. Cervical squamous cell carcinoma (CSCC) is the most common pathology subtype, accounting for about 80-90% of total CC [1]. The incidence of CC is much higher in developing countries due to the unavailability of HPV vaccines and unprotected sexuality [2]. It is estimated that more than 85% of CC and deaths were found in developing countries, including China [2]. Although early diagnosis can increase the survival rate of CC, the overall treatment e cacy is still poor because of its unclari ed pathogenesis [3].
Cancer-associated broblasts (CAFs) are one of the most abundant stromal components in tumor stroma and have prominent roles in the pathogenesis of many, if not all, solid tumors [4]. Mechanistically, CAFs build up and remodel the tumor microenvironment (TME) by secreting growth factors and cytokines [5].
Recent studies indicate that CAFs are involved in the progression of CC [6,7], however, limited study has investigated how the formation of CAFs in CC tissues is regulated.
WD repeat domain 5 (WDR5) is an important component of histone methyltransferase (HMT), which is responsible for the catalyzation of trimethylation of histone 3 lysine 4 (H3K4me3) [8]. WDR5 has been proved to play crucial roles in cancer pathogenesis [9], however, no study has explored its role in CC development. In this study, we proposed a hypothesis that WDR5 may drive the development of CC by promoting CAFs formation. Based on this hypothesis, we analyzed the expression pattern of WDR5 in CC tissues and the para-carcinoma tissues (PCT) using clinical samples. We also investigated the function of WDR5 by in vitro studies and explored the therapeutic value by targeting the WDR5 function using a mouse model.

Isolation and culture of cells
The primary normal cervical epithelial cells (hCECs) were isolated as the prior report [10]. Brie y, the fresh PCT was minced and digested with type I collagenase at 37 °C for 60 min. The digested mixture was then ltrated through a stainless-steel strainer (0.5-1.0mm). Tell suspension was centrifuged at 1,500 rpm for 5 min, and hCECs were collected and cultured in Dulbecco's Modi ed Eagle's Medium (DMEM, Thermo Fisher Scienti c, USA) supplemented with 10% (v/v) fetal bovine serum (FBS) (Clark, USA). The purity of hCECs was identi ed by > 90% positive immuno uorescence staining for cytokeratin 7 (Supplementary Figure 1A). Cells with passages > 10 were used in the following cell experiments.
We also isolated primary CAFs from cancer tissues of CSCC as previously described [11]. Brie y, fresh cancer tissues were minced and digested with type I collagenase, and the digested mixture was ltrated through a nylon mesh of 38 um pore size. The CAFs were in the ltrate. The ltrate was centrifuged at 1000 g for 5 min, and CAFs were resuspended in DMEM supplemented with 10% FBS, and were then incubated in a fresh culture medium for 4 weeks to allow cell attachment and grow out. The purity of CAFs was identi ed by > 90% positive immuno uorescence staining for broblast activation protein (FAP) (Supplementary Figure 1B). Cells with passages > 10 were used in the following cell experiments. CSCC cell line CaSki was also used in this study. CaSki cells were purchased from the American Type Culture Collection (ATCC) and were cultured in DMEM supplemented with 10% FBS. Cells were all placed in the 37 °C incubators supplemented with 20% oxygen and 5% carbon dioxide.

Cell transfections
Small interfering RNAs (siRNAs) for WDR5 (si-WDR5) and CXCL8 (si-CXCL8) (VectorBuilder, China) were used to knock down the target transcripts. siRNAs having no target (si-NC) were served as the control. Plasmid overexpressing WDR5 (Vector-WDR5) or CXCL8 (Vector-CXCL8) were also used to express exogenous proteins, and the empty vectors (Vector-NC) were used as the control. The transfection was conducted using Lipofectamine 3000® Transfection Reagent (Invitrogen, USA) according to the manufacturers' instructions. The sequence of siRNAs was provided in Table 1.
Chromatin immunoprecipitation (ChIP) and ChIP-sequencing (ChIP-seq) ChIP assays were performed as described previously [12]. Brie y, cells were cross-linked, lysed, and sheared by sonication to produce DNA fragments with an average length of approximately 500 bp. Then, 1% of the chromatin fragments were used as the input. Chromatin was immunoprecipitated using antibodies against WDR5 (13105, Cell Signaling Technology, USA), H3K4me3 (ab8580, Abcam, USA), and MLL1 (34907, Cell Signaling Technology, USA). Normal rabbit IgG was added as the control. Then DNA fragments immunoprecipitated were puri ed and were further analyzed by quantitative real-time PCR (RT-qPCR). The sequence of primers used is provided in Table 1. ChIP-seq analyses were performed using the ChIP-IT High Sensitivity Kit (ab185908, Abcam, USA). The model-based analysis of the ChIP-Seq peaknding algorithm was used to normalize ChIP against the input control.

Immunohistochemistry
Immunohistochemistry was performed on para n-embedded tissue sections. Tissues on the sections were blocked with 5% BSA, and were then incubated overnight at 4 °C with primary antibodies for the detection of the following: WDR5 (ab178410, Abcam, USA), CXCL8 (ab106350, Abcam, USA), and FAP (ab207178, Abcam, USA). After being incubated with the secondary antibodies, slides were visualized using DAB-Substrate (Beyotime, China) and photographed using the Aperio ePathology Scanner (Leica, Germany). Protein expression was quanti ed by H-score, which was calculated by the formula: H-score = Pi (i), where i is the intensity of staining with a value of 1, 2, or 3 (weak, moderate, or strong, respectively) and Pi is the percentage of stained cells for each intensity in the range of 0-100%. Confocal Microscope (Leica, Germany) and results were analyzed by Leica Application Suite X (Leica, Germany).

Western blotting (WB)
Cells lysates containing total proteins were collected and prepared using a loading buffer. An equal amount of protein from each group was loaded into SDS-PAGE (10% gel) and separated by electrophoresis. Proteins in the gel were then transferred to a polyvinyl di uoride membrane. After immersion in quick blocking buffer (Beyotime, China) for 30 min, membranes were incubated overnight at 4 °C with the primary antibodies for the detection of the following: E-cadherin (ab40772, Abcam, USA), Ncadherin (ab76011, Abcam, USA), β-catenin (ab32572, Abcam, USA), snail (ab216347, Abcam, USA), and tubulin (ab215037, Abcam, USA). Membranes were then incubated with the secondary antibodies (ab6721, Abcam, USA) at room temperature for another 30 min. Protein bands in the membrane were visualized using ECL plus kit (Beyotime, China), and the relative expressions of target proteins were quanti ed using Image J (National Institutes of Health, USA).

RT-qPCR
Total RNA was extracted by the Trizol method (Thermo Fisher Scienti c, USA) and were reversely transcript into cDNA using PrimeScriptTM RT Master Mix (Takara, Japan). RT-qPCR was performed on the Bio-Rad CFX96 (Bio-Rad Laboratories, USA) using SYBR Premix Ex TaqTM (Takara, Japan). The sequence of primers was provided in Table 1. The expression level of each transcript was calculated using the comparative threshold cycle (Ct), based on the using the 2 -△△Ct formula.

Invasion Assay
Invasion chambers (Corning, USA) were used to perform invasion assay. Brie y, 100 ul of Matrigel (BD, USA) was used to coat the inner side of the chamber, and then cells suspended in a culture medium were loaded. DMEM containing 20% FBS was added to the lower chamber as the chemoattractant. The chambers were incubated at 37°C for 48 hours. The non-invasive cells on the upper side of the chamber were removed gently, and chambers were stained using crystal violet to locate the invaded cells. The number of invaded cells was observed and photographed under an inverted optical microscope (Olympus, Japan).

Enzyme-linked immunosorbent assay (ELISA)
Levels of growth factors [granulocyte-macrophage colony-stimulating factor (GM-CSF), interleukin-6 (IL-6), CC-chemokine ligand 2 (CCL2), and transforming growth factor-β (TGF-β)] in cancer tissues and cell cultures were analyzed using ELISA kits (mlbio, China) according to the manufacturer's protocols. The intra-and inter-assay coe cients of variation were all less than 10%. Samples were adjusted for total protein concentrations before detection to ensure that the amounts of total protein in each group were equal.

Tumor xenograft mouse model
Four-week male nude BALB/c mice (Cyagen, China) were housed in a facility with a 12 h light/dark cycle maintained at 25 ± 0.5°C and 50% to 60% humidity. The xenograft CC model was established as described before [13]. Brie y, CaSki cells were prepared as a single-cell suspension in Matrigel, and 1×10 6 prepared cells were injected subcutaneously into the right axillary fossa. Mice were euthanized 14 days after modeling (pelltobarbitalum natricum, 100mg/kg, intravenous administration), and tumor lesions were enucleated for further analysis.

Statistical Analysis
Statistical analyses were performed using GraphPad Prism 6.0 (GraphPad Software, USA). The student's t-test was used to analyze the difference between two groups, and one-way ANOVA was used for the comparison among multiple groups. The Chi-square test was used to analyze the categorical variables. In this study, results were obtained from three independent experiments, and P<0.05 was considered to be statistically signi cant.

WDR5 is up-regulated in CSCC tissues compared to that in PCT
We rst analyzed the expression pattern of WDR5 in CC tissues and the corresponding PCT. Results showed that compared to PCT, WDR5 was up-regulated CSCC tissues ( Fig.1 A and B), but had no signi cant change in cancer tissues of cervical adenocarcinoma and cervical adenosquamous cell carcinoma ( Supplementary Figure 2A and B). We next analyzed the correlation between WDR5 expression and the TNM stage of CSCC patients, and found that WDR5 expression was positively correlated with the TNM stage-patients in III-IV stages had markedly higher WDR5 levels compared to patients in I-II stages ( Fig. 1A and B).

CXCL8 is a target of WDR5
To explore the function of WDR5, we performed ChIP-seq to nd its target gene. Results of ChIP-seq showed that a large amount of DNA fragment of CXCL8 was immunoprecipitated by anti-WDR5 antibody ( Fig. 2A), indicating that CXCL8 is a target gene of WDR5. Results of WB showed that overexpression WDR5 in hCECs signi cantly up-regulated the expression of CXCL8, while the knock-down of WDR5 in CAFs markedly inhibited CXCL8 expression (Fig. 2B). Results of ChIP analysis also indicated that WDR5 could be recruited to the promoter region of CXCL8 (Fig. 2C). These results all suggested that CXCL8 expression was regulated by WDR5. Consistent with the expression pattern of WDR5, analysis on human tissues showed that CXCL8 was up-regulated in CSCC tissues compared to that in PCT, and its expression level was positively correlated with the TNM stage of patients (Fig. 2D).
CXCL8 promotes epithelial-mesenchymal transition (EMT) and induces the formation of CAFs Using hCECs from PCT, we next investigated the function of CXCL8 in the progression of CSCC. Results showed that overexpression of CXCL8 markedly changed the shape of hCECs, and made it present the shape of stromal cells (Fig. 3A). The markers of EMT such as E-cadherin, N-cadherin, β-catenin, and snail were all changed markedly (Fig. 3B). To identify the characteristic of these transformed cells, we analyzed the expression of FAP, a marker of CAFs [14]. Results showed that CXCL8 overexpression signi cantly up-regulated the expression of FAP (Fig. 3C), indicating that CXCL8 is an enhancer of CAFs formation. Besides, the secretions of cytokines such as GM-CSF, IL-6, CCL2, and TGFβ, which were proved to promote malignancy progression, were all up-regulated upon CXCL8 overexpression (Fig. 3D), and the invasive ability of CaSki cells was signi cantly enhanced when they were incubated with the culture medium from cells overexpressing CXCL8 (Fig. 3F).

Loss of CXCL8 function makes CAFs transformed into epithelial cells
Using CAFs from CSCC tissues, we then performed knockout experiments to verify the role of CXCL8. Results showed that knock-down of CXCL8 made CAFs present the shape of epithelial cells (Fig. 4A), and changed the expressions of EMT markers (Fig. 4B). Besides, the knock-down of CXCL8 suppressed the expression of FAP (Fig. 4C), attenuated the secretions of cytokines (Fig. 4D). Additionally, compared to the culture medium from CAFs treated with si-NC, the culture medium from hCSCs with CXCL8 knockout signi cantly attenuated the invasive ability of CaSki cells (Fig. 4E). These results all proved that CXCL8 was involved in CSCC progression by enhancing the transformation of CAFs from epithelial cells.

CXCL8 is regulated by the interaction between WDR5 and MLL1
In mammals, WDR5 interacts with MLL1 to form the HMT complex and catalyzes the H3K4me3 in gene promoters [8]. We next investigated the regulation pathway of CXCL8 by using MM102, a molecule that can speci cally block the interaction between WDR5 and MLL1. Results showed that in hCSCs from CSCC tissues, the expression of CXCL8 is down-regulated after WDR5 knockout, while the knockout of WDR5 had no impact on CXCL8 expression when MM102 was used, and MM102 markedly inhibited CXCL8 expression ( Fig. 5A and C). Additionally, in hCECs from PCT, overexpression of WDR5 markedly induced CXCL8 expression, whereas WDR5 overexpression had no impact on CXCL8 expression when MM102 was used ( Fig. 5B and D). ChIP analysis revealed that the distributions of WDR5, MLL1, and H3K4me3 in the CXCL8 promoter region were similar, and the recruitments of WDR5 and MLL1 in CXCL8 promoter were signi cantly higher in hCSCs from CSCC tissues than that in hCECs from PCT, as well as the H3K4me3 level of CXCL8 promoter ( Fig. 5E and F).

Blocking the interaction between WDR5 and MLL1 suppresses CC development
We then performed in vivo study to explore the therapeutic value for CSCC by targeting the interaction between WDR5 and MLL1. Results showed that using MM102 could signi cantly inhibit tumor growth in mice model (Fig. 6A). Analysis on tumor body showed that MM102 could markedly attenuate the expressions of CXCL8 and FAP ( Fig. 6B and C), and the markers of EMT (Fig. 6D), as well as the cytokines involved in tumor growth (Fig. 6E).
Overall, all the above results indicate that WDR5 plays a key role in the development of CC. Highly expressed WDR5 promotes the transformation of CAFs from epithelial cells, and enhanced the secretions of cytokines such as GM-CSF, VEGF, PDGF, and TGFβ, thus driving the progression of CC (Fig. 7).

Discussion
Tumor recurrence, metastasis, and the intractable TME remain the three key unsolved issues that hamper effective cancer treatment in clinical practice [15]. TME creates a protective niche that is bene cial for tumor cells proliferation and invasion, and protects tumor cells from conventional interventions, leading to therapeutic failure [16]. It is well known that the TME is a multicellular system with complex tumorstromal interactions [16]. Stromal support is essential for multi-step cancer progression including tumor growth, metastatic dissemination, and ectopic colonization [17]. However, normal broblasts have been indicated to function as an inhibitor of cell proliferation and tumorigenesis [18]. Therefore, reprogrammed broblasts, also known as CFAs, are proposed to explain the pro-tumorigenic ability of broblasts in tumor stroma.
As the most abundant components in the tumor stroma, CAFs play a key role in the regulation of TME [15]. The role of CAFs has been extensively studied in vitro. Compared to normal broblasts, CAFs express different proteins and usually exhibit enhanced proliferative and migratory properties [19]. The most unique feature of CAFs is their ability to remodel the TME by their high capacity for cytokine secretions, including GM-CSF, TGFβ, IL-6, and CCL2 [20]. These cytokines can further recruit immunosuppressive cells into the tumor stroma, and lead to immune evasion [20]. One of the most important sources of CAFs is epithelial or endothelial cells adjacent to cancer cells undergoing EMT to form stromal cells, and further, transform into CAFs [21]. During the development of CSCC, normal epithelial cells gradually disappeared and are replaced by cancer cells and tumor stromal cells. Do these epithelial cells undergo apoptosis in the process of tumor dilatation? Or do they differentiate into another kind of cells to regulate CSCC progression? Previous studies have indicated that EMT occurs during CSCC oncogenesis and progression [22,23]. So, can these normal epithelial cells be transformed into CAFs via EMT and further be involved in CSCC progression? No answer has been given in previous reports.
As a key regulator of HMT, WDR5 is crucial for H3K4me3, chromatin remodeling, and transcriptional activation of target genes [8]. It was also proven to function as an oncogenic protein and might serve as a novel epigenetic target in cancer treatment [9]. The role of WDR5 has been investigated in pancreatic cancer [24], breast cancer [25], prostate cancer [26], etc. However, its role in CC has not been explored yet. In this study, we proved that WDR5 expression was up-regulated in CSCC, and its level in cancer tissue is positively correlated with the TNM stage of patients, suggesting that WDR5 is an oncogenic protein that promotes the development of CSCC. A recent study reveals that WDR5 facilitates EMT and metastasis of cholangiocarcinoma by changing chromatin opening and target gene expression [27], however, no study has investigated the role of WDR5 in EMT in CSCC and its relationship with CFAs formation. Again, our results revealed that highly expressed WDR5 facilitates EMT of normal cervical epithelial cells, and promotes the transformation of CAFs from epithelial cells, so as to drive the progression of CSCC. More importantly, we found a way to block the function of WDR5 by hindering the interaction between WDR5 and MLL1, and proved that loss of WDR5 function suppressed CSCC growth in vivo, and this may provide us a novel approach for CSCC treatment.
In conclusion, for the rst time, this study investigated the role of WDR5 in CSCC development from the perspective of CAFs, and proved that highly expressed WDR5 induces EMT, promotes the formation of CAFs from epithelial cells, and enhanced the secretions of cytokines, thus driving the progression of CSCC.    TGFβ were all attenuated after CXCL8 knock-down. (E) The invasive ability of CaSki cells was attenuated markedly after being incubated with the culture medium from CAFs with CXCL8 knock-down. Data were shown as mean ± SEM. ** P<0.01, ***P<0.001. Figure 5 CXCL8 is regulated by the interaction between WDR5 and MLL1. (A and C) Expression of CXCL8 in CAFs was down-regulated after WDR5 knock-down, while the knock-down of WDR5 had no impact on CXCL8 expression when MM102 was used, and MM102 markedly inhibited CXCL8 expression in CAFs. (B and D)

Declarations
Expression of CXCL8 in hCECs was up-regulated after WDR5 overexpression, while the overexpression of WDR5 had no impact on CXCL8 expression when MM102 was used (E and F) ChIP analysis revealed that the distributions of WDR5, MLL1, and H3K4me3 in CXCL8 promoter were overlapping, and the recruitments of WDR5 and MLL1 in CXCL8 promoter were signi cantly higher in CAFs than that in hCECs, as well as the H3K4me3 level in this region. Data were shown as mean ± SEM. ** P<0.01, ***P<0.001.