Emergent Role of the Fractalkine Axis in Dissemination of Peritoneal Metastasis from Epithelial Ovarian Carcinoma

Abstract

Epithelial ovarian carcinoma is the most common cause of death from gynecologic cancers largely due to advanced, relapsed, and chemotherapy-resistant peritoneal metastasis, which is refractory to the currently used treatment approaches. Mechanisms supporting advanced and relapsed peritoneal metastasis are largely unknown, precluding development of more effective targeted therapies. In this study we investigated the function of a potentially targetable fractalkine axis in the formation and the development of advanced and relapsed peritoneal metastasis and its impact on patients’ outcomes. Our mouse model studies support a role for the fractalkine receptor (CX₃CR1) in the initiation of peritoneal adhesion important for recolonization of relapsed peritoneal metastasis. We show that downregulation of CX₃CR1 results in reduction of metastatic burden at several peritoneal sites commonly colonized by advanced and relapsed metastatic ovarian carcinoma. We show that the chemokine fractalkine (CX₃CL1), an activating ligand of CX₃CR1, regulates organ-specific peritoneal colonization. High expression of CX₃CR1 correlates with significantly shorter survival, specifically in post-menopausal patients with advanced and terminal stages of the disease. Taken together, our studies support a key regulatory role for the fractalkine axis in advanced and relapsed peritoneal metastasis in epithelial ovarian carcinoma.

Introduction

Advanced, relapsed, and chemotherapy-resistant metastatic epithelial ovarian carcinoma is a major clinical challenge, because the currently used treatment approaches, mainly consisting of surgery and cytotoxic chemotherapy, fail to cure the disease at these terminal stages. For the past several decades epithelial ovarian carcinoma, of which high grade serous is the most predominant and deadliest histotype, was the leading cause of mortality from gynecologic malignancies and the fifth cause of death from all cancers affecting women; only in the US more than 15,000 women die annually from this disease (1–7). To improve survival it is imperative to develop new treatment approaches that specifically target the metastatic disease. To this end, identification of the mechanisms that promote metastatic success of ovarian carcinoma could yield better targets for the development of novel therapeutics.

The majority of ovarian carcinoma patients are first diagnosed when peritoneal metastases have already spread (8, 9). In patients with metastatic disease progression from the time of the initial diagnosis to death is often very quick; according to current statistics, the 5-year survival for this cohort is less than 30%. Initial peritoneal lesions as well as recolonization by the recurrent disease occur as a result of lodging the malignant cells shed from the primary tumor on to peritoneal mesothelium, the lining of the peritoneum that consists of parietal peritoneum and visceral peritoneum continuous with each other. Once the disseminating cells have successfully adhered to the mesothelium, they invade the submesothelial matrix and establish secondary lesions in the stroma of intraperitoneal tissues and organs. Patients diagnosed with disseminated ovarian carcinoma undergo surgery, including hysterectomy, bilateral salpingo-oophorectomy, omentectomy, and debulking of peritoneal metastasis; however, complete tumor resection often cannot be achieved (10–12). Moreover, cytotoxic chemotherapy regimens to which patients are subjected fail to eliminate all tumor cells due to acquired and intrinsic resistance to chemotherapy (13–16). Consequently, metastases often recur in the peritoneal wall and other remaining peritoneal organs. The microenvironment at these secondary sites likely supports organ-specific colonization, although the underlying mechanisms are not well understood (17). Uncontrolled proliferation of metastatic lesions leads to their expansive growth, which eventually causes death through malnutrition and bowel obstruction.

Hence, peritoneal adhesion and metastatic cell proliferation are among the key processes in the metastatic cascade of ovarian carcinoma that could be exploited for anti-metastatic intervention by novel targeting agents.

Targeted therapies directed against tumor-specific pathways and molecules have been more successful in treating metastasis compared to the conventional cytotoxic chemotherapy (18). We reasoned that the chemokine family G protein-coupled receptor fractalkine (CX₃CR1) might be such a molecule, because it is expressed in 64% of metastatic ovarian carcinoma specimens (19) and in vitro studies showed that CX₃CR1-positive ovarian carcinoma cells adhered to mesothelial monolayer and proliferated in CX₃CL1-dependent manner (19, 20); moreover, we previously reported that CX₃CR1 was expressed in all tested types of stromal and epithelial ovarian carcinoma (19, 21, 22). Chemokine receptors are seven transmembrane G protein-coupled receptors (GPCR) that are activated upon ligand (chemokine) binding resulting in activation of intracellular signaling networks that regulate processes vital for both normal and cancer cells (23–25). Approximately 50% of all currently used drugs are directed against GPCRs, underscoring their value as successful drug targets (26). Fractalkine receptor is uniquely and specifically activated by the only ligand, the chemokine CX₃CL1 (fractalkine) (27, 28). CX₃CL1 is a unique member of the CX₃C subfamily of chemokines, as it contains a transmembrane domain with which it anchors on the cell membrane and supports cell-cell adhesion. Proteolytic cleavage of CX₃CL1 yields a soluble form capable of inducing cell migration and proliferation (29–31). CX₃CL1 is expressed by epithelial ovarian carcinoma cells and it can be detected in malignant ascites from patients (19, 20), but its expression status across the peritoneal organs and tissues serving as metastatic sites is largely unknown. The fractalkine axis has previously been shown to regulate adhesion and migration of leukocytes and signaling from neurons to glia (29, 32). Notably, a role in supporting metastatic colonization of breast, prostate, and pancreatic cancers has also been established (33–36). Due to the prominent role of CX₃CR1 in various pathological conditions extensive efforts have been invested to design and develop CX₃CR1 inhibitors. Such inhibitors have proven specific and effective in the pre-clinical setting (37–39), providing a strong rationale for potential CX₃CR1-directed therapies in ovarian carcinoma.

Here we build on our in vitro mechanistic studies (19) and show that in patients and in a syngeneic mouse model expression of CX₃CR1 strongly correlates with mortality and metastatic development at CX₃CL1-positive peritoneal organs. Our data validate CX₃CR1 as a potential therapeutic target in late-stage metastatic ovarian carcinoma.

Results

Status of CX₃CR1 expression predicts survival of patients with serous EOC

Our previous studies suggested that the CX₃CL1/CX₃CR1 axis supports cell-cell adhesion, migration, and proliferation (19), all of which are essential for successful colonization and metastatic expansion. These properties are directly linked to the aggressiveness of cancer metastasis; however, the relationship between expression of CX₃CR1 and patients’ survival has not yet been investigated. We used gene expression data from The Cancer Genome Atlas (TCGA) database for serous ovarian adenocarcinoma (n=557) and the Oncomine platform (40). We found that patients’ survival inversely correlated with the level of CX₃CR1 expression (FIGURE 1A, left panel). Moreover, patients with high CX₃CR1 expression lived significantly shorter compared to their counterparts with low CX₃CR1 expression (FIGURE 1A, center panel). Furthermore, we asked whether CX₃CR1 expression status correlated with survival of patients at advanced and terminal FIGO stages. We divided specimens in three groups, including those at The International Federation of Gynecology and Obstetrics (FIGO) Stage I and II, III, and IV (Supplementary FIGURE 1A). In the group containing patients with Stage III disease, those with high CX₃CR1 expression survived significantly shorter compared to the patients with low CX₃CR1 expression (FIGURE 1B, left and right panels). For patients with Stage IV disease the difference in survival as a function of CX₃CR1 expression was even more pronounced, as patients with high CX₃CR1 lived twice shorter compared to those with low CX₃CR1. In fact, all patients with Stage IV disease and high CX₃CR1 expression died by a little over 2000 days, while at least one third of patients with Stage IV disease and low CX₃CR1 expression were still alive at this time point (FIGURE 1B, center and right panels). As ovarian carcinoma is associated with age, we also asked whether menopausal status in a combination with CX₃CR1 expression played a role in survival. To address this, the specimens were divided into pre- and post-menopausal groups (Supplementary FIGURE 1B). Interestingly, in the pre-menopausal cohort CX₃CR1 expression did not affect survival (FIGURE 1C, left and right panels). In contrast, in post-menopausal patients’ cohort, high CX₃CR1 expression significantly correlated with worse survival compared to the patients with low CX₃CR1 expression (FIGURE 1C, center and right panels). Thus, not only that patients with high expression of CX₃CR1 survived significantly shorter than those with low CX₃CR1 expression, this decrease in survival was significantly more pronounced in post-menopausal patients at terminal stages of the disease. These data suggest that high expression of CX₃CR1 in primary ovarian carcinoma, occurring in a quarter of the tested patients’ population, is a predictor of worse survival for the late stage post-menopausal women with serous epithelial ovarian cancer.

FIGURE 1. High expression of CX₃CR1 correlates with an overall shorter survival.

(A) High vs low CX₃CR1 expression predicts an overall shorter survival in patients with serous ovarian carcinoma. Data from The Cancer Genome Atlas (TCGA) database for serous ovarian adenocarcinoma were analysed in Oncomine and plotted as CX₃CR1 expression Log_{2(median centered ratio)} vs percent survival (green line; left panel). Data points with incomplete information were removed from the analysis; total number of specimens analysed was 557. Based on the value of median CX₃CR1 expression, defined as CX₃CR1 expression Log_{2(median centered ratio)} at 50% survival, the data set was divided in two (center panel), including specimens with CX₃CR1 expression > 3.32, termed “high CX₃CR1 expression” (n=143; red line), and specimens with CX₃CR1 expression < 3.32, termed “low CX₃CR1 expression” (n=414; blue line). Overall survival of patients expressing high and low levels of CX₃CR1 was plotted and analysed with Kaplan-Meier plot using Prism software (center panel). Average overall survival of patients with high and low CX₃CR1 expression is indicated in the Table (right panel). *p=0.0020, Log-rank (Mantel-Cox) test; *p=0.0027, Gehan-Breslow-Wilcoxon test. (B) High vs low CX₃CR1 expression predicts an overall shorter survival in patients with serous ovarian carcinoma at FIGO Stages III and IV. The specimens in groups containing Stage III (n=429) and Stage IV (n=83) specimens were divided in two groups with “high” and “low” CX₃CR1 expression (red and blue lines, respectively) corresponding to cases with higher than median CX₃CR1 expression and lower than median CX₃CR1 expression, respectively (3.23 for Stage III and 3.77 for Stage IV). Overall survival of patients with Stage III disease expressing high (n=100) and low (n=329) levels of CX₃CR1 was plotted and analysed with Kaplan-Meier plot using Prism software (left panel); *p=0.0205, Log-rank (Mantel-Cox) test, *p=0.0170, Gehan-Breslow-Wilcoxon test. Overall survival of patients with Stage IV disease expressing high (n=23) and low (n=60) levels of CX₃CR1 was plotted and analysed with Kaplan-Meier plot using Prism software (center panel);*p=0.0036, Log-rank (Mantel-Cox) test, *p=0.0197, Gehan-Breslow-Wilcoxon test. Average overall survival of patients at Stage III and IV disease with high and low CX₃CR1 expression is indicated in the Table (right panel). (C) High vs low CX₃CR1 expression predicts an overall shorter survival in postmenopausal patients with serous ovarian carcinoma. Overall survival of patients of pre-menopausal age with high (n=22) and low (n=108) CX₃CR1 expression was plotted and analysed using Kaplan-Meier plots (left panel); p=0.8421, Log-rank (Mantel-Cox) test, p=0.5221, Gehan-Breslow-Wilcoxon test. Overall survival of patients of post-menopausal age with high (n=120) and low (n=307) CX₃CR1 expression was plotted and analysed using Kaplan-Meier plots (center panel); *p=0.0046, Log-rank (Mantel-Cox) test, *p=0.0215, Gehan-Breslow-Wilcoxon test. Average overall survival of patients of both pre- and post-menopausal age with high and low CX₃CR1 expression is indicated in the Table (right panel).

Downregulation of the CX₃CR1 expression reduces peritoneal adhesion and metastasis formation in a syngeneic model of ovarian cancer

Analysis of patients’ survival indicated that high expression of CX₃CR1 correlated with shorter survival (FIGURE 1) and previous in vitro studies have suggested a role of CX₃CR1 in CX₃CL1-dependent peritoneal adhesion and proliferation (19, 20); however, this did not explain whether CX₃CR1 played an active role in aggressiveness of the disease or simply correlated with its progression. To test this, we employed the ID8/C57BL/6 syngeneic mouse model of ovarian carcinoma to determine the role of tumoral CX₃CR1 in the disease progression in the presence of a functional immune system. ID8 has been derived by spontaneous transformation of the murine ovarian epithelium and, thus, provides a model to study metastatic dissemination in the presence of the native immune system (41); however, it has been recently shown that ID8 does not carry mutations in TP53 and in other genes commonly mutated in ovarian carcinoma (42). We reasoned that CX₃CR1-directed therapies may not be able to affect all of the CX₃CR1-positive cells and/or may not be able to completely block CX₃CR1 activity in the context of the involved organs due to various complications, including drug permeability and its intratumoral concentration. To better model such a scenario, we created ID8 clones in which CX₃CR1 was downregulated rather than completely knocked-out. These clones were generated by transfecting a pool of four CX₃CR1-specific shRNAs following puromycin selection. Downregulation of CX₃CR1 of an average of 30% was confirmed by Western blot and immunofluorescence staining in four subclones, termed cx2, cx3, cx4, and cx6 (Supplementary FIGURE 2A,B). Significant reduction of CX₃CR1-dependent cell proliferation and migration in the selected clones was confirmed using Transwell cell migration and proliferation assays, confirming the loss of function as a consequence of reduction of CX₃CR1 expression. Non-effective 29-mer scrambled shRNA cassette in retroviral untagged vector and an empty untagged vector were used to create control cell lines (designated “scr sh” and “vector”, respectively), in which neither CX₃CR1 expression nor CX₃CL1-dependent cell migration and proliferation were significantly altered compared to the parental ID8 (Supplementary FIGURE 2C,D). Further, a small molecule specific inhibitor of CX₃CR1, compound 18a (37), was able to efficiently block CX₃CL1-mediated cell proliferation (Supplementary FIGURE 3).

To determine the requirement of CX₃CR1 in peritoneal adhesion we used a short-term in vivo adhesion assay. Prior to i.p. injection both control (ID8 and scr sh) and experimental (cx2 and cx4) cells were transfected with GFP. Following i.p. injection of GFP-labeled cells into abdomens of C57BL/6 mice the intraperitoneal adhesion was allowed to take place for 4 hours. Adhered ovarian carcinoma cells were found mainly on the surfaces of omentum and anterior peritoneal wall, consistent with the pattern of metastatic seeding of early peritoneal metastasis. Experimental groups showed significantly reduced number of adhered cells compared to the controls (FIGURE 2A), further supporting a key role of the fractalkine axis in peritoneal adhesion of ovarian carcinoma cells.

FIGURE 2. CX₃CR1 supports peritoneal adhesion and tumor formation in a syngeneic model of ovarian carcinoma.

(A) Short-term in vivo adhesion assay demonstrates that CX₃CR1 is important for peritoneal adhesion. Immunofluorescence images (green fluorescence) demonstrate cells adherent to omentum and peritoneal wall, as indicated, 4 hours following i.p. injection of the GFP-labeled control (scr sh group) and experimental (cx2 clone) ID8 cells into abdomens of C57BL/6 (5/group). Adherent cells were visualized under the fluorescence microscope in tissues excised from sacrificed control (injected with ID8 and scr sh) and experimental (injected with cx2 and cx4 clones) animals, enumerated using Zeiss AxioVert software, and plotted. Red squares – mouse #1, purple checks – mouse #2, blue flakes – mouse #3, green triangles – mouse #4, blue diamonds – mouse #5. Bars show an average number of adherent cells in each group. The differences between groups (cx2 vs scr sh and cx4 vs scr sh) were statistically analysed using both Mann-Whitney U test and Students’ t-test; * p<0.05. Differences between ID8 and scr sh groups were not statistically significant using both Mann-Whitney U test and Students’ t-test. All groups were statistically analysed using one-way ANOVA; p=0.0026. (B) Downregulation of CX₃CR1 leads to significant reduction of tumor formation. The number of animals bearing metastasis on the organs covered by the parietal peritoneum and the visceral peritoneum in both control (ID8 and scr sh combined to a total of 20 animals; designated “control”) and experimental (clones cx2, cx3, and cx6 combined to a total of 30 animals; designated “CX₃CR1sh”) groups was calculated and plotted. Animals were considered positive for metastasis at each site if lesions visible with the naked eye were present upon dissection when animals reached humane endpoints. The data are shown as average ± standard deviation and were statistically analyzed using Students’ t-test.; * p<0.05. (C) Normal human mesothelium is strongly CX₃CL1-positive. Expression of CX₃CL1 in human mesothelium was determined using immunohistochemistry and anti-CX₃CL1-specific antibodies. A typical image of the CX₃CL1 immunostaining in normal mesothelium from intestine/colon of a 55 year old woman is shown (core B1, Supplementary Table 1). Black arrow points to the mesothelial monolayer. Brown, CX₃CL1; blue, hematoxylin. Bar, 200 micron. Insert is a 5-fold magnification of the area outlined with dashed lines.

To determine the impact of reduction of peritoneal adhesion on tumor formation we i.p. injected both control (ID8 and scr sh) and experimental (clones cx2, cx3, and cx6) cells into abdomens of C57BL/6 mice and allowed animals to reach humane endpoints. Downregulation of CX₃CR1 expression resulted in a significant reduction of the number of animals bearing tumors at all peritoneal mesothelial surfaces, such as parietal peritoneum and visceral peritoneum (FIGURE 2B). These data suggest that impairment of peritoneal adhesion via downregulation of CX₃CR1 could result in reduction of metastasis formation.

To validate our findings in human-derived mesothelial cell lines we measured CX₃CL1 expression in human mesothelium from healthy donors. All specimens of female visceral peritoneum containing mesothelial cells (n=5) displayed very strong (nearly 3+) immunoreactivity against CX₃CL1 antibodies in mesothelial cells (FIGURE 2C; Supplementary TABLE 1), suggesting that CX₃CL1/CX₃CR1-dependent peritoneal adhesion may occur in human disease.

Downregulation of CX₃CR1 expression reduces metastatic burden in a syngeneic model of ovarian cancer

Previous findings indicating a role of CX₃CL1/CX₃CR1 in ovarian carcinoma cell proliferation prompted us to examine the effect of downregulation of CX₃CR1 on the tumor burden in the ID8/C57BL/6 syngeneic mouse model. We analyzed the overall tumor burden and found that the total tumor volume and size for all experimental groups (cx2, cx3, and cx6) was significantly lower than that for the control groups (ID8 or scr sh; FIGURE 3A,B), suggesting that expression of CX₃CR1 is tightly linked to tumor burden possibly via supporting CX₃CL1-dependent cell proliferation. Importantly, expression of CX₃CR1 in the clone cells (score 3+) remained lower than in the parental ID (score 2+) at the end of the experiments (FIGURE 3C). Moreover, parental cells displayed strong Ki67 and PCNA staining (3+) compared to the cx2 clone cells (2+), suggesting that proliferation of clone cells with reduced CX₃CR1 expression remained lower than in the parental cells throughout the experiment. Reduced proliferation in the clone cells was accompanied by the evidence of apoptosis, as determined by staining with CD95 (score 1+), while the parental ID8 cells were not undergoing apoptosis, as these tissues scored negatively for CD95 staining (FIGURE 3D).

FIGURE 3. Downregulation of CX₃CR1 expression results in significantly reduced overall tumor burden in a syngeneic model of ovarian carcinoma.

(A) An overall tumor burden was calculated by adding the weights of individual tumor nodules at all sites (omentum, peritoneal wall, diaphragm, liver, mesenthery, kidney, ovary&uterus, and pancreas) and plotted. Orange circles – mouse #1, red squares – mouse #2, purple checks – mouse #3, lime dashes – mouse #4, blue flakes – mouse #5, green triangles – mouse #6, green crosses – mouse #7, blue diamonds – mouse #8, orange dashes – mouse #9, lilac diamonds – mouse #10. Color coding for the control and experimental animal groups: red – ID8, green – scr sh, purple – cx2, orange – cx3, black– cx6. Bars show an average tumor burden in each group. Differences between two groups were analysed using both Mann-Whitney U test and Students’ t-test; * p<0.05. All groups were statistically analysed using one-way ANOVA; p=0.0030. (B) Typical images of gross tumor dissemination throughout the abdomen for ID8 and CX2 groups are shown (bottom panels). Yellow dashed lines outline tumor nodules. Specimens of tumors generated by ID8 and CX2 groups at the time of sacrifice were immunohistochemically stained for CX₃CR1 (C) as well as Ki67, PCNA, and CD95 (D). Primary antibody omission was used as negative control; sections were incubated with secondary anti-mouse and anti-rabit antibodies only, as indicated. Bar, 200 micron. Brown: CX₃CR1, Ki67, PCNA, or CD95, as indicated. Blue: hematoxylin.

Next, to evaluate the role of the fractalkine signaling in organ-specific tumor growth we analyzed distribution of the tumor burden by site. Interestingly, tumor burden at sites covered by the parietal peritoneum, such as the anterior and the posterior peritoneal wall and diaphragm, was significantly reduced when CX₃CR1 in the cancer cells was downregulated (FIGURE 4A). Tumor burden on liver, covered by the visceral peritoneum, was also reduced in response to downregulation of CX₃CR1 (FIGURE 4B). However, growth of tumors underlying the mesentery anchored in small intestines was not affected by reduction of CX₃CR1 expression (FIGURE 4C). Similarly, although average tumor burden in one of the experimental groups (cx2) was nearly statistically significantly lower compared to the control scrsh group (p=0.064), overall, omental tumors in both control and experimental conditions did not significantly differ in size when animals have reached moribund conditions (FIGURE 4D). However, complete omentectomy is the standard of care for patients undergoing surgery for ovarian carcinoma; hence, management of the relapsed disease will not be complicated by the omental metastasis. We also assessed tumor formation at retroperitoneal sites, and found that downregulation of CX₃CR1 did not affect tumor burden at pancreas, whereas it significantly reduced it at the kidney (FIGURE 5). These data indicate that growth of the CX₃CR1-expressing ovarian carcinoma cells depends on the context of the host tissue.

FIGURE 4. Effect of CX₃CR1 downregulation on development of intraperitoneal metastasis at (A) peritoneal wall and diaphragm, (B) liver, (C) the mesentery, and (D) the omentum.

Tumors at each site were excised weighted, measured, and the results were plotted (left and center panels). Orange circles – mouse #1, red squares – mouse #2, purple checks – mouse #3, lime dashes – mouse #4, blue snowflakes – mouse #5, green triangles – mouse #6, green crosses – mouse #7, blue diamonds – mouse #8, orange dashes – mouse #9, lilac diamonds – mouse #10. Color coding for the control and experimental animal groups throughout the figure: red – ID8, green – scr sh, purple – cx2, orange – cx3, black– cx6. Bars show an average tumor burden in each group. Differences between two groups (scr sh and clones) were analysed using both Mann-Whitney U test and Students’ t-test; * p<0.05. Tumors were preserved, sectioned, and stained with hematoxylin and eosin. Dashed yellow lines separate tumors from stroma. Images were generated using Aperio ScanScope (right panels). Bar, 100 micron.

FIGURE 5. Effect of CX₃CR1 downregulation on development of retraperitoneal metastasis at pancreas and kidney.

Tumors at each site were excised weighted, measured, and the results were plotted (left and center panels). Orange circles – mouse #1, red squares – mouse #2, purple checks – mouse #3, lime dashes – mouse #4, blue snowflakes – mouse #5, green triangles – mouse #6, green crosses – mouse #7, blue diamonds – mouse #8, orange dashes – mouse #9, lilac diamonds – mouse #10. Color coding for the control and experimental animal groups throughout the figure: red – ID8, green – scr sh, purple – cx2, orange – cx3, black– cx6. Bars show an average tumor burden in each group. Differences between two groups (scr sh and clones) were analysed using both Mann-Whitney U test and Students’ t-test; * p<0.05. Tumors and outlying stromal regions were preserved, sectioned, and stained with hematoxylin and eosin. Dashed yellow lines separate tumors from stroma. Images were generated using Aperio ScanScope (right panels). Bar, 100 micron.

It is important to note that these animal model studies were designed to determine the role of CX₃CR1 in tumor growth rather than survival. Paracentesis was not performed; instead, animals were sacrificed when they became moribund, resulting in comparable survival among the control and experimental groups (Supplementary FIGURE 4A). Volumes of ascites did not significantly vary among all groups (Supplementary FIGURE 4B) and lymph node tissues were CX₃CL1-negative (Supplementary FIGURE 4C, Supplementary TABLE 2), indicating that CX₃CL1/CX₃CR1 does not contribute to ascites formation.

Expression of CX₃CL1 in intraperitoneal organs targeted by metastatic dissemination correlates with tumor growth of CX₃CR1-positive cells

Previous studies suggested that once ovarian carcinoma cells attach to the mesothelial monolayer, they disrupt it by invading between the intercellular spaces followed by establishing secondary lesions in submesothelial stroma (43, 44). Hence, interaction with the matrix and cells lying beneath the mesothelium could become more important for cell proliferation and growth of metastasis than the initial interaction of metastasizing ovarian carcinoma cells with the mesothelial cells. The local microenvironment surrounding invading metastatic cells varies according to the host site; consequently, the paracrine signaling regulating proliferation of peritoneal metastases could be driven by different growth factors. As downregulation of CX₃CR1 affected tumor growth at peritoneal wall, diaphragm, liver, and kidney, we hypothesized that cell proliferation at these sites largely depended on fractalkine signaling. High levels of CX₃CL1 expression at these would be consistent with such a model of CX₃CL1-dependent proliferation. On the other hand, tumor formation at omentum, pancreas, and small intestine did not depend on CX₃CR1, indicating that tumor cell proliferation at these sites is regulated by other growth factors. Indeed, Western blot analysis showed low expression of CX₃CL1 in omentum and pancreas, while the peritoneal wall and diaphragm had comparatively high expression levels. Major immunoreactive bands with apparent molecular weight of about 55 kDa were observed in both the control and tested specimens (FIGURE 6A), confirming previously reported findings in pancreas (45). Further, using immunohistochemistry we identified that both tumor and host tissues were CX₃CL1-positive (FIGURE 6B,C). However, the intensity of CX₃CL1 immunostaining in skeletal muscle cells of diaphragm (100% of cells, score 2+) and tubular cells of kidney (100% of cells, score 3+) was stronger than that in the tumor itself (score 1+; FIGURE 6B). On the other hand, both host tissues and tumors developed at pancreas and omentum had similar intensity of CX₃CL1 immunostaining (FIGURE 6C). These findings could argue in favor of a significant role of host CX₃CL1 in supporting the growth of peritoneal lesions seeded by CX₃CR1-positive ovarian cancer cells in our mouse model. These data also suggest that the role of tumoral CX₃CL1 in metastatic cell growth is minimal; additionally, analysis of CX₃CL1 expression in human serous ovarian carcinoma specimens showed a weak correlation with CX₃CR1 expression and no correlation between its expression levels and patients survival (Supplementary FIGURE 5). In contrast, recent studies suggest a prominent role for the autocrine fractalkine signaling in colorectal carcinoma, as tumor cells that co-expressed both CX₃CR1 and CX₃CL1 did not disseminate leading to reduced metastasis and longer survival (46). Perhaps, distinct modes of metastatic dissemination and intertumoral differences may result in dissimilar outcomes related to the role of the fractalkine axis in advanced and relapsed metastasis from different cancer types. Additionally, the level of CX₃CR1 reduction in our model may not have been sufficient to affect tumor formation at other organs, such as omentum, in which case only animals in cx2 group showed trends toward reduced omental tumor burden.

FIGURE 6. Role of tumoral and stromal CX₃CL1 in CX₃CR1-dependent metastatic dissemination.

(A) Expression of CX₃CL1 in specimens of tumor-naive mouse omentum, pancreas, peritoneal wall, and diaphragm. Expression of CX₃CL1 in total cell lysates (10 µg protein/lane) of peritoneal wall, diaphragm, omentum, and pancreas obtained from tumor-naive C57BL/6 mice (n=2) was detected using Western blot. Left lane shows positions of the molecular weight marker proteins. Mouse recombinant CX₃CL1 (10 ng) served as a positive control. Actin served as a loading control. (B,C) Tumoral and stromal CX₃CL1 expression. CX₃CL1 in tumor and invaded stroma, as specified for each host site, was visualized with immunostaining in organ sites dissemination to which correlated with CX₃CR1 expression in tumor cells (diaphragm, kidney, (B)) and where it did not depend on CX₃CR1 expression in the tumor cells (pancreas, omentum, (C)). Images were generated using Aperio ScanScope; bar, 100 micron. Dashed yellow lines separate tumors from stroma. Brown, CX₃CL1; blue, hematoxylin.

To determine the relationship between CX₃CL1 expression in peritoneal tissues and the propensity for hosting metastatic ovarian carcinoma cells in human subjects, we analyzed its expression in a subset of normal peritoneal tissues serving as hosts for metastasis. We conducted immunohistochemistry using specimens of two normal human tissues, liver and kidney, dissemination to which was CX₃CR1-dependent in the mouse model. These tissues were characterized by high intensity of CX₃CL1 immunostaining, which was detected on the cell membrane, the cytoplasm, and the nucleus (FIGURE 7A, Supplementary TABLES 3 and 4). Nearly 100% of hepatocytes and 100% of Kupffer cells in the liver tissue were strongly CX₃CL1-positive (average score in female specimens 2.6+). Distal tubular cells of kidney were strongly CX₃CL1-positive (score 3+), while CX₃CL1 expression in proximal tubular cells of kidney and interstitial stromal cells was lower (score 1+); in total CX₃CL1 score in kidney averaged to 2.4+. In the specimens of intestine/colon and omentum, dissemination to which was not CX₃CR1-dependent in the mouse model, CX₃CL1 immunoreactivity localized to sparsely present fibroblasts and endothelial cells and it was predominantly nuclear (FIGURE 7B, Supplementary TABLE 1). Average immunoscores for kidney and liver were significantly higher compared to those for intestine/colon (FIGURE 7C), suggesting that in patients, proliferation of lesions lodged on liver and kidney could, to a large extent, depend on CX₃CL1 expressed in these tissues.

FIGURE 7. Expression of CX₃CL1 in human kidney and liver is higher compared to that in intestine/colon and omentum.

Immunohistochemical staining with CX₃CL1-specific antibodies was used to determine CX₃CL1 expression and intracellular localization in normal human specimens of liver and kidney (A), intestine/colon, and omentum (B). Brown, CX₃CL1; blue, hematoxylin. Typical images of CX₃CL1-immunoreactive staining in specimens of liver, kidney, intestine/colon, and omentum are shown. Image of liver corresponds to the case C6 (Supplementary Table 3), image of kidney corresponds to the case A4 (Supplementary Table 4), image of intestine/colon corresponds to the case C06 (Supplementary Table 1) and image of normal human specimen of epiploon (greater omentum) belongs to a 21 year old female. Brown, CX₃CL1; blue, hematoxylin. Images were generated using Aperio ScanScope; bar, 100 micron. (C) Comparison of the CX₃CL1 immunoscores among the tested tissues. The scoring was performed for the entire tissue, including all cell types present (of note, specimens of intestine/colon contained stroma underlying mesothelium, consistent with the microenvironment surrounding ovarian cancer cells anchoring in submesothelial stroma). The normal liver TMA contained 9 female and 15 male specimens; the average staining in female and male samples was 2.6 and 2.4, respectively (Supplementary Table 3). The normal kidney TMA contained 24 female and 56 male specimens; the average staining score in female and male samples was 2.1 and 2.3, respectively (Supplementary Table 4). Distal tubules displayed stronger immunoreactivity against CX₃CL1 antibodies compared to the proximal tubules. The visceral peritoneum TMA contained both stomach and intestine/colon specimens. Among those there were 13 female and 14 male intestine/colon specimens with average staining score of 1.2 and 1.0, respectively. Immunoscores of CX₃CL1 expression in female specimens of normal human liver, kidney, and intestine/colon were averaged and plotted. The data were analysed using both Mann-Whitney U test and Students’ t-test; * p<0.05.

Discussion

Our studies suggest that the fractalkine axis actively contributes to the development of both initial and recurrent ovarian carcinoma metastasis at several crucial points, including peritoneal adhesion and tumor formation as well as organ-specific colonization and growth of tumor lesions, ultimately impacting patient survival. However, the ID8 cell line has now been extensively characterized (42) and it does not have mutational patterns characteristic of human ovarian cancers and in particular does not have a TP53 mutation which is pathognomonic of high grade serous ovarian cancers. Thus the relevance of the current studies to human ovarian cancers will need to be determined from studies of additional relevant murine models and of human ovarian cancer samples.

We show that downregulation of CX₃CR1 impairs peritoneal adhesion and reduces formation of tumor lesions in a syngeneic model of the disease. Peritoneal adhesion is a key event for both the development of initial metastatic lesions and recolonization of the abdominal tissues by the recurrent disease. The role of the fractalkine axis in de novo formation of ovarian carcinoma metastasis seems to parallel that previously described in cancers of other primary sites, including breast and pancreatic carcinomas (33–35). Our findings suggest that in recurrent metastatic disease, recolonization of the abdomen could be, at least partially, prevented and blocked if the fractalkine axis is inactivated.

Our findings suggest a prominent role for the CX₃CL1/CX₃CR1 interaction in advanced and relapsed ovarian carcinoma metastasis. Our data indicate that the fractalkine axis regulates development and growth of lesions at peritoneal wall, diaphragm, liver, and kidney, i.e. sites commonly colonized by the relapsed disease. Further, our data indicate that the stromal CX₃CL1 plays a prominent role by supporting cell proliferation, as the tumor burden at organs in which CX₃CL1 expression was higher than that in tumors themselves was the most sensitive to downregulation of CX₃CR1 in the cancer cells. A major clinical challenge in the successful treatment of ovarian carcinoma is posed by the advanced and relapsed peritoneal metastasis characterized by rampant expansion, uncontrolled proliferation, and chemotherapy resistance, leading to patients’ death (9, 47). Optimal debulking of peritoneal metastases carries significant survival benefits, suggesting that optimization of therapeutic approaches focused on reduction of the residual tumor burden will extend survival (48). Our studies indicate that downregulation of CX₃CR1 expression could significantly reduce both formation and growth of recurrent peritoneal lesions, thus providing rational for targeting the fractalkine axis in treatment of ovarian carcinoma.

Importantly, we found a significant correlation between the levels of CX₃CR1 expression and patients’ survival, supporting a major role for the fractalkine axis in the aggressiveness of ovarian carcinoma.

Taken together, our data indicate that CX₃CR1 is a key regulator of formation and relapse of peritoneal metastasis in ovarian carcinoma. Not only that its high expression correlates with worse prognosis in postmenopausal patients at advanced and terminal disease stages, but CX₃CR1 could also play an active role in seeding secondary lesions as well as promoting metastasis by supporting CX₃CL1-dependent organ-specific metastatic growth. These data suggest that CX₃CR1 could be an optimal target for the retardation of metastatic expansion, leading to increasing survival, by negatively affecting multiple aspects of the peritoneal metastatic disease, from preventing recolonization to blocking metastatic growth at main sites relevant to the relapsed metastatic disease. At least a quarter of patients with serous ovarian carcinoma express CX₃CR1 at high levels that correlate with significantly shorter survival. This represents a large fraction of patients potentially benefitting from such a treatment, considering the high degree of molecular heterogeneity commonly displayed by this disease. Furthermore, our previous studies demonstrated that CX₃CR1 is expressed in all tested histotypes of epithelial and non-epithelial ovarian carcinoma, suggesting that metastatic progression of all types of ovarian carcinoma may depend on CX₃CR1. Given that the ID8 cell culture model used in this study has been derived from murine ovarian surface epithelium, future studies involving models representing other histotypes of ovarian carcinoma are needed to firmly establish the role of CX₃CR1 in peritoneal adhesion and growth of metastatic lesions as it pertains to all types of ovarian carcinoma. The fractalkine axis has been shown to support a very limited number of normal physiological processes, including conception, regulation of resident macrophage population in the intestine, and glia-neuron interaction in the brain, negative effect on which could be managed (32, 49, 50). Encouragingly, a handful of CX₃CR1 targeting agents has been developed and successfully applied using various preclinical models (37, 39, 51). Given the results of our studies indicating a key role for CX₃CR1 in tumor growth and organ-specific peritoneal colonization, further studies are needed to assess the potential efficacy of these agents in metastatic ovarian carcinoma.

Materials and Methods

Materials

Mouse-derived ovarian carcinoma cell line ID8 (REF) was obtained from Dr. Katherine Roby (University of Kansas Medical Center) and maintained in minimal essential media supplemented with 10% fetal bovine serum. The cell line was cultured for no longer than 15 consecutive passages and was routinely assessed by cell morphology and the average doubling time. IMPACT III PCR profiling test (IDEXX RADIL) showed that the cell line was free of contamination by Mycoplasma spp., Mycoplasma pulmonis, mouse hepatitis virus, minute virus of mice, mouse parvovirus, Theiler’s murine encephalomyelitis virus, Sendai virus, pneumonia virus of mice, murine norovirus, reovirus 3, mouse rotavirus, Ectromelia virus, Lymphocytic choriomeningitis virus, polyomavirus, and lactate dehydrogenase-elevating virus. Human ascites-derived ovarian adenocarcinoma cell line SKOV-3 was obtained from the NCI Tumor Bank Repository and maintained in minimal essential media supplemented with 10% fetal bovine serum. The cell line was cultured for no longer than 10 consecutive passages and was routinely assessed by cell morphology and the average doubling time; a Short Tandem Repeat (STR) analysis of this cell line revealed a 100% match to the DNA profile of SKOV-3 at ATCC. Four unique 29-mer CX₃CR1 (mouse)-specific shRNA constructs (#1 5’AGTGCAGCACGGTGTCACCATTAGTCTGG3’, #2 5’GCCTTTGGAACCATCTTCCTGTCCGTCTT3’, #3 5’AGTGGCGTTCAGCCACTGTTGCCTCAACC3’, #4 5’CGTGCTCCGCAACTCGGAAGTCAACATCC3’) and non-effective 29-mer scrambled shRNA cassette (5'GCACTACCAGAGCTAACTCAGATAGTACT3’) in retroviral untagged vector were obtained from Origene. Six- to eight-weeks old C57BL/6 and athymic nude (FOXN1NU) mice were purchased from Jackson Laboratories (Sacramento, CA) and maintained at the AAALAC-approved animal facility at UIC in accordance with approved IACUC protocols. Tissue microarrays (TMA) containing human specimens of normal lymph node (LY481), normal kidney (KD803), normal liver (LVN241), and a single specimen of normal epiploon (greater omentum) from a healthy female were obtained from US Biomax (Rockville, MD). TMA containing normal human intestinal and stomach mesothelium (GIP541) was obtained from Pantomics (Richmond, CA). Anti-CX₃CL1, anti-CX₃CR1, anti-rabbit, and anti-mouse biotin-conjugated antibodies were obtained from Abcam (Cambridge, MS). Anti-Ki67, anti-PCNA, and anti-CD95 antibodies were obtained from Novus Biologicals (Littleton, CO), Santa Cruz Biotechnologies (Dallas, TX), and BioLegend (San Diego, CA), respectively. ABC and DAB kits as well as animal-free blocker were obtained from Vector Laboratories (Burlingame, AL). DharmaFECT was obtained from Dharmacon (Lafayette, CO). Monoclonal anti-β-tubulin antibody was purchased from the Developmental Studies Hybridoma Bank (East Iowa City, IA). Transwell cell migration chambers with eight-micron pores were purchased from Corning (Corning, NY). CX3CR1 antagonist 18a (37) was purchased from Axon Medchem (Reston, VA). The WST-1 proliferation kit was obtained from Takara Bio (Madison, WI).

Short-term in vivo cell adhesion

10⁶ cells of GFP-labeled parental ID8 and its subclones scr sh, cx2, and cx4 suspended in PBS in a total volume of 150 µl were i.p. injected into abdominal cavities of 6–8 weeks old female C57BL/6 mice (n=5). Animals were sacrificed 4 hours post injection following dissection of abdominal organs and tissues. Adhered fluorescent cells were visualized using inverted Axiovert Zeiss fluorescence microscope and enumerated.

Long-term tumor formation

For generation of intraperitoneal tumors 10⁶ cells of parental ID8 and its subclones stably expressing CX₃CR1shRNA (clones cx2, cx3, cx4, and cx6) and scrambled shRNA suspended in PBS in a total volume of 150 µl were injected intraperitoneally (i.p.) into 6–8 weeks old female C57BL/6 mice (n=10). Sample size was chosen based on our previous in vivo studies (52, 53). Animals were randomly assigned to either control or experimental groups. In vivo experiments were performed by H.G.M. and J.X., who were blinded to the outcomes of the study. Animals were monitored three times weekly until they reached humane endpoints. Animals were sacrificed due to their poor health condition, a large volume of ascites, or 20% higher weight compared to the control group (animals injected with PBS). One animal in the “PBS” group was excluded due to malocclusion. Ascitic fluid was collected and measured. Animals were sacrificed, dissected, and the abdominal region was examined for visible tumors. Tumors associated with a specific organ were excised, measured, preserved in paraffin and examined by immunohistochemistry with hematoxylin&eosin (H&E).

Western blot and immunofluorescence staining

Western blot was used to detect the expression of CX₃CR1, CX₃CL1, and β-tubulin as previously described [38,39,41]. Immunofluorescence staining and imaging was performed as previously described (REF).

Immunohistochemical staining and scoring

Procedures were performed as we described before [43,46,47]. Briefly, slides containing paraffin-preserved human tissue sections were rehydrated by incubation in xylenes and ethanol solutions followed by blocking of peroxidase activity. Antigen retrieval was achieved with a 10-min microwaving in 10 mM tris(hydroxymethyl)aminomethane (TRIS), 1 mM ethylenediaminetetraacetic acid (EDTA), pH 9.0. Primary mouse anti-human- CX₃CL1 antibody was used at 1 µg/ml overnight at 4 °C. Biotin-conjugated goat anti-mouse secondary antibodies were used at a dilution of 1:200 for 30 min at RT. Vectashield ABC and DAB reagent were prepared and used as suggested by the manufacturer. Slides were stained with hematoxylin, dehydrated, and mounted with Permount. Specimens of adrenal gland and ovarian carcinoma were used as a positive control, and exclusion of primary antibody was used as a negative control. Staining was evaluated by A.A.K.-B., O.E., and H.X., who were blinded to the experimental outcomes of the study. Staining was scored based on the intensity and percentage of positive cells as follows: “0” for negative cases, “1” for weakly positive cases, “2” for moderately positive cases, and “3” for highly positive cases. Staining was assessed in the membrane, the cytoplasm, and the nucleus.

Statistical analyses

For analyses of survival two non-parametric tests, the log-rank (Mantel-Cox) test and the Gehan-Breslow-Wilcoxon test, were employed using GraphPad Prism software. Comparisons between two data sets with normal distribution were conducted using Student’s t-test and Microsoft Excel software. Mann-Whitney U test and Tukey HSD test were used to compare two data sets with abnormal distribution. Data belonging to three or more independent groups were analysed using one-way ANOVA. The findings were considered statistically significant at p<0.05.