Ascites-induced compression alters the peritoneal microenvironment and promotes metastatic success in ovarian cancer

The majority of women with recurrent ovarian cancer (OvCa) develop malignant ascites with volumes that can reach > 2 L. The resulting elevation in intraperitoneal pressure (IPP), from normal values of 5 mmHg to as high as 22 mmHg, causes striking changes in the loading environment in the peritoneal cavity. The effect of ascites-induced changes in IPP on OvCa progression is largely unknown. Herein we model the functional consequences of ascites-induced compression on ovarian tumor cells and components of the peritoneal microenvironment using a panel of in vitro, ex vivo and in vivo assays. Results show that OvCa cell adhesion to the peritoneum was increased under compression. Moreover, compressive loads stimulated remodeling of peritoneal mesothelial cell surface ultrastructure via induction of tunneling nanotubes (TNT). TNT-mediated interaction between peritoneal mesothelial cells and OvCa cells was enhanced under compression and was accompanied by transport of mitochondria from mesothelial cells to OvCa cells. Additionally, peritoneal collagen fibers adopted a more linear anisotropic alignment under compression, a collagen signature commonly correlated with enhanced invasion in solid tumors. Collectively, these findings elucidate a new role for ascites-induced compression in promoting metastatic OvCa progression.


Results
Artificial ascites-induced compression potentiates OvCa cell adhesion to peritoneum. The peritoneal MC monolayer represents a barrier to OvCa cell access to the sub-mesothelial extracellular matrix, wherein OvCa cells anchor and proliferate to form secondary lesions 15 . Successful adhesion of OvCa cells to the peritoneal surface induces mesothelial cell retraction and exposure of the sub-mesothelial matrix [16][17][18][19] . A previous study showed that elevated IPP (via CO 2 pneumoperitoneum) in a murine OvCa model enhanced peritoneal metastases 14 . To investigate whether ascites-induced IPP influences early events in adhesion of OvCa cells to peritoneum, an in vivo artificial ascites assay was developed in which C57Bl/6 female mice were injected i.p. with RFP-tagged OVCAR5 or OVCAR8 cells in a large volume (5 mL) of PBS to mimic the ascites condition, relative to control mice that received tumor cells in a small volume (1 mL) of PBS (Fig. 1a) 20 . After allowing time for tumor cell i.p. adhesion (5-8 h), mice were euthanized and the peritoneum was dissected and imaged to enable quantitation of adherent tumor cells (Fig. 1b). Adhesion to omental tissue was rapid (30 min-1 h) and did not differ between control and artificial ascites conditions (data not shown). However a 20-30 fold increase in adhesion of both OVCAR5 and OVCAR8 cells to peritoneal surfaces in vivo was observed in mice with artificial ascites compared to control mice (Fig. 1c), suggesting that fluid induced changes in IPP enhance peritoneal metastasis through potentiating OvCa cell adhesion.
Compression alters peritoneal mesothelial cell surface ultrastructure and induces TNT formation. During OvCa progression, peritoneal MC undergo morphological and molecular modifications that contribute to disease progression 21 . As the MC monolayer lines the inner surface of the peritoneum and covers peritoneal organs, these cells are directly subjected to ascites-induced compression. To evaluate whether ascites-induced compression alters MC morphology or ultrastructure, we used a Flexcell Compression Plus System to apply controlled compressive force (~ 3 kPa; ~ 22 mmHg, comparable to IPP in OvCa patients with tense ascites) to a human peritoneal MC cell line (LP9) or to primary human peritoneal MC (HPMC) 7,8 . Strikingly, compressed LP9 and HPMC cells exhibited dramatic changes in morphology, as evidenced by retraction, acquisition of a more mesenchymal phenotype and formation of sub-micrometer scale intercellular projections (Fig. 2a). Similar results were seen when intact murine peritoneal tissue explants were subjected to compression (~ 3 kPa; ~ 22 mmHg) ex vivo. Scanning electron microscopy (SEM) analysis of the compressed peritoneum showed alterations in the mesothelial surface consistent with the formation of tunneling nanotubes (TNT) connecting distant MC (Fig. 2b). The compression-induced TNTs are actin-based, have a length of ~ 3-100 µm (Fig. 2c), a diameter of 30-60 nm and are not attached to the surrounding substratum 22,23 .
To further evaluate compression-induced TNT formation in vivo, C57Bl/6 female mice were injected i.p with either 5 or 10 mL of PBS for 5 h (Fig. 3a) 20 , euthanized and the peritoneal tissue was dissected and processed for SEM imaging. Retraction of MC was observed in mice injected with 5 mL PBS relative to control mice (Fig. 3b, c). When injected with a higher fluid volume to mimic tense ascites (10 mL), abundant TNT formation was observed on the peritoneal mesothelial surface (Fig. 3d), suggesting a pressure threshold required for induction of TNT formation in vivo. Similar to structures induced by injection of high volume artificial ascites in mice, peritoneal tissue obtained from human OvCa patients with ascites displayed remarkably similar TNT structures (Fig. 3e).
To assess the minimum threshold required to form TNT in peritoneal MC, a compressive gradient (1-3 kPa; 7.5-22 mmHg) was applied to LP9 cells using the Flexcell Compression Plus System. Under these conditions, a minimum pressure of 2.5 kPa (~ 18.5 mmHg) was needed to induce TNT formation in LP9 cells (Fig. 4a). Furthermore, TNT formation was observed as early as 3 h in compressed LP9 cells (~ 3 kPa; ~ 22 mmHg) (Fig. 4b).

Compression promotes TNT formation between OvCa cells and peritoneal MCs. As adhesion
of OvCa cells to the mesothelial surface is a key event in metastasis, we sought to model how ascites-induced compression may impact OvCa cell adhesion to peritoneal mesothelium. Compression (~ 3 kPa; ~ 22 mmHg) was applied to OVCAR5 or OVCAR8 cells in contact with murine peritoneal explants ex vivo, using the Flexcell Compression Plus System. Strikingly, both compressed OVCAR5 and OVCAR8 cells formed extensive nanoscale projections contacting the peritoneal mesothelial surface as early as 30 min (Fig. 5a, b). As previous data showed high level expression of the non-canonical Wnt ligand Wnt5a in OvCa ascites 9 , the effect of compression on Wnt5a expression was examined. Compression upregulated expression of WNT5A mRNA and Wnt5a protein (Suppl. Fig. 1a, b). Addition of exogenous Wnt5a to OvCa cells induced formation of nanoscale projections visible by fluorescence and scanning electron microscopy (Suppl. Fig. 1c, d), suggesting a role of Wnt5a signaling in TNT formation. In an in vivo model, TNT formation with peritoneal mesothelial cells was observed in an artificial ascites mouse model injected with either OVCAR5 or OVCAR8 cells (Fig. 5c). Similar to the murine in vivo model, human peritoneal tissues with metastatic lesions, collected from OvCa patients and imaged with SEM, also exhibited TNT between tumor cells and the peritoneal mesothelial surface (Fig. 5d). It has been previously demonstrated that TNTs mediate the transfer of cellular cargo and organelles between cells under mechanical stress 22,23 . Interestingly, the compression-induced TNTs exhibit variations in the number and shape of distensions in the nanotube, suggesting a transfer of cargo between compressed cells (Fig. 6a). As mitochondrial transport through TNT has been reported in several malignant cells [24][25][26][27][28] , we investigated the possibility of mitochondrial transport in the compression-induced TNTs using GFP-tagged LP9 human MC with fluorescently labeled mitochondria (MitoTracker Red). These LP9 cells were then cultured with either OVCAR5 or OVCAR8 cells under compression (~ 3 kPa; ~ 22 mmHg). TNTs containing fluorescent (red) mitochondria were observed in nanotubes formed between MC and both OVCAR5 and OVCAR8 cells (Fig. 6b), indicating the transport of mitochondria from LP9 cells to OvCa cells.
Compression alters peritoneal collagen fiber alignment. Disseminating OvCa cells adhere to the peritoneal mesothelial surface, intercalate within the MC layer, and induce MC retraction to facilitate invasion MicroCT Scans showing C57Bl/6 female mice injected i.p. with 1 mL (control) or 5 mL (artificial ascites, or AA) PBS containing 10 6 RFP-tagged OVCAR5 or OVCAR8 cells, as indicated, for 5 or 8 h, respectively. (B) Mice were sacrificed, peritoneal tissue was collected and images of adherent cells to peritoneum were obtained using Echo Revolve fluorescent microscope at × 20 magnification. (C) Adherent cells were quantified using ImageJ. All experiments in were performed as triplicates with three independent biological replicates per cell line. All results are presented as mean ± s.e.m. and P-values were calculated using a Student's two-tailed t-test. P < 0.05 is statistically significant. www.nature.com/scientificreports/ of the underlying type I collagen-rich ECM, wherein they proliferate to form secondary lesions 11,16,17,19 . In the ECM of normal tissues, collagen fibers are wavy and isotropic [29][30][31] . During cancer progression, collagen fibers are realigned, becoming straightened and anisotropic. This alteration in collagen quaternary structure has been shown to enhance cancer cell motility and invasion, leading to tumor progression and poor survival in several neoplasms 29,32 .
To determine whether ascites-induced compression alters the ultrastructure of peritoneal collagen, we injected C57Bl/6 female mice with either RFP-tagged ID8 Trp53 −/− or ID8 Trp53 −/− BRCA −/− , which can induce ascites formation in 5-7 weeks post injection ( Fig. 7a) 33 . Mice were divided into two groups; the first group was sacrificed 2 weeks post-injection to allow i.p. metastasis without formation of ascites and the second group was sacrificed 5-6 weeks post-injection after ascites accumulation. MicroCT scans were performed to monitor ascites progression and to visualize peritoneal cavity expansion (Fig. 7a). Mice were sacrificed and the peritoneum was imaged with Second Harmonic Generation (SHG) microscopy to visualize collagen fiber alignment. Z-stacks of  www.nature.com/scientificreports/ the peritoneal collagen were taken and pictures were evaluated to determine collagen fiber alignment (Fig. 7b). Collagen fibers in mice with i.p. metastasis in the absence of ascites were wavy with isotropic alignment (Fig. 7b, week 2). In contrast, peritoneal collagen fibers in mice bearing ascites (week 5-6) were significantly anisotropic and straight (Fig. 7b, c), indicating that ascites accumulation is associated with alterations in the sub-mesothelial matrix ultrastructure.

Discussion
Malignant ascites accumulation in OvCa results from obstruction of peritoneal lymphatic drainage by disseminating OvCa cells, in addition to enhanced vascular permeability in the peritoneal cavity 2 . The accumulation of malignant ascites can dramatically elevate IPP from sub-atmospheric (5 mmHg) to as high as 22 mmHg 3 , changing the loading environment in the peritoneal cavity and causing compressive and shear stress. Cells adaptively transduce mechanical stimuli by converting them into biochemical signals, resulting in altered cytoskeletal organization and cellular behavior [34][35][36] . While the contribution of the cellular and molecular components of malignant ascites to OvCa progression is well studied 2 , the role of ascites-induced increases in IPP in OvCa progression is unknown. Herein, we aimed to model ascites-induced compressive effects on the peritoneal microenvironment and the resulting impact on OvCa peritoneal seeding. OvCa cell metastasis to the peritoneum involves the adhesion of OvCa cells to the peritoneal mesothelial surface, migration into and anchoring within the sub-mesothelial collagen-rich matrix and proliferation to establish secondary lesions 37,38 . While tumor cells initially rapidly home to the omentum, other mesotheliallylined surfaces are populated later in metastatic progression. In models of artificial ascites, our data show rapid www.nature.com/scientificreports/ omental homing accompanied by significantly enhanced OvCa cell adhesion to the peritoneum. Similar results were obtained in a previous study that generated high IPP (8 mmHg) using CO 2 pneumoperitoneum, demonstrating increased abdominal invasion of pre-injected ID8 OvCa cells in the high pressure group relative to low IPP (2 mmHg) or controls 14 . Together these data suggest that the increase in intraperitoneal adhesion observed in the current study is related to compression-induced effects rather than enhanced dispersion of cells within the peritoneal space due to enhanced fluid volumes. This is supported by experiments using ex vivo peritoneal explants which demonstrate enhanced receptivity to OvCa cell adhesion under conditions of compression. Additional research indicated similar effects of compression on cancer cell behavior and metastatic potential 39 . For example, compression altered cytoskeletal organization and acquisition of an invasive morphology in pancreatic and breast cancer cells, leading to increased migration and invasion 40,41 . Interstitial fluid pressure regulated www.nature.com/scientificreports/ collective invasion in compressed breast cancer cellular aggregates 42 . Additionally, compression-induced alterations in cells in the tumor micro-environment that also contribute to disease progression have also been reported. This is exemplified by a study demonstrating that compression activated pancreatic fibroblasts, which in turn promoted pancreatic cancer cell migration 43 . Here, we show that compression induced dramatic alterations in peritoneal MC morphology, including MC retraction and acquisition of a mesenchymal phenotype. Moreover, we previously reported compression-induced changes in OvCa multicellular aggregates correlated with upregulation of epithelial-mesenchymal transition (EMT)-associated genes 11 . Together these data suggest that ascites-induced compression may play a role in mesothelial-mesenchymal transition 21,44 . A limitation of the current study is the lack of information regarding the uniformity of ascites-induced compression throughout structures of the peritoneal cavity impacted by OvCa metastases. Given the high degree of complexity of this compartment, future studies employing computational models that take into account peritoneal fluid dynamics to assess regional differences in compressive loading are warranted. In addition to inducing ultrastructural changes that result in a more receptive peritoneal surface, novel findings show the induction of TNTs under compression in cultured cells, in intact murine peritoneum ex vivo and in vivo, and in human peritoneum from OvCa patients. Cells form TNTs as an intercellular communication conduit to withstand environmental stress, including metabolic stress and inflammatory conditions, via the intercellular transport of cytoplasmic organelles and molecules that promote survival 45,46 . To the best of our knowledge, this is the first report of mechanically-regulated TNT induction in intact tissues. Mitochondrial trafficking through TNT has been shown in several cell types and can contribute to drug resistance and metabolic reprogramming in cancer cells [24][25][26]47 . Our results indicate that compression-induced TNTs mediate the transfer of mitochondria from MC to OvCa cells, which may contribute to OvCa cell survival under compression. TNTmediated mitochondrial transfer can change the behavior and fate of the recipient cells 27 . A previous study determined that TNT-transferred mitochondria from mesenchymal stem cells to breast cancer cells enhanced basal and maximal oxygen consumption, reduced glycolysis and lactate production and increased the concentrations of both the endogenous mitochondrial DNA and the produced ATP 48 . Also, ATP levels increased after TNTmediated mitochondria transfer from bone marrow stromal cells to acute myeloid cells during chemotherapy, leading to myeloid cell survival 47 . In light of these observations, future investigations will address the impact of TNT-mediated mitochondrial transfer from MC to OvCa cell on OvCa cellular metabolism and survival.
Mechanical stimuli can remodel the extracellular matrix, impacting tumor cell behavior 30,41,[49][50][51] . Normal peritoneal tissues exhibit a random, wavy and isotropic arrangement of collagen fibers, while collagen fiber remodeling into an organized, straight and anisotropic alignment is observed under pathologic conditions and in several epithelial tumors which has been termed 'tumor-associated collagen signatures' (TACS) 30,52 . Anisotropic collagen fibers facilitate cancer cell motility and invasion to the surrounding stroma 53,54 . Here, we show that ascites accumulation in two different in vivo OvCa murine models is associated with an alteration in the alignment of peritoneal collagen fibers to a linear anisotropic arrangement. Together these data suggest that ascites-induced compression may contribute to OvCa progression by promoting OvCa cell adhesion to peritoneum, altering the integrity of the peritoneal MC monolayer, and enhancing interactions between OvCa cells and peritoneal MC via TNT formation. While results implicate Wnt5a as a potential molecular mediator, additional mechanistic studies are needed to elucidate the mechanosensory receptors and/or pathways that mediate the observed effects and to modulate the activity of these effectors in relevant pre-clinical models of metastatic disease.

Human specimens. De-identified fresh peritoneum tissue was obtained through University of Kansas
Cancer Center's Biospecimen Repository Core Facility under Institutional Review Board approved protocol HSC#5929, abiding with the US Common Rule, and studies using these tissues were conducted in accordance with the relevant guidelines and regulations of this committee. Written patient consent was obtained for use of the specimen for research purposes. At the time of cytoreductive surgery, once optimal debulking was achieved, grossly normal appearing peritoneum was identified and a 2 cm diameter section removed. The peritoneum was transferred to the laboratory immediately upon retrieval from the patient, rinsed in sterile PBS, and marked with a tissue marker to indicate the outer side of the peritoneum. The tissue was then placed in SEM fixative solution (below) and incubated at 4C for 2-3 days prior to further processing as described below.
Scientific RepoRtS | (2020) 10:11913 | https://doi.org/10.1038/s41598-020-68639-2 www.nature.com/scientificreports/ X-Ray MicroCT scans. The microCT experiments were conducted using an Albira CT System (Bruker Xtreme In Vivo Imaging system, Billerica, MA, USA) in the Notre Dame Integrated Imaging Facility as previously described 56 . Mice were first anesthetized with isoflurane (2.5% flow rate) prior to imaging. Scans of the mice were performed with a FOV of 160 mm at low dose CT intensity (0.2 mA) and a high CT voltage (45 kVp).

Artificial ascites in vivo adhesion assay.
To model the effect of ascites on OvCa cell adhesion to peritoneum, C57Bl/6 female mice were i.p injected with 5 × 10 6 RFP-tagged OVCAR5 or OVCAR8 in 5 mL or 10 mL PBS as an artificial ascites model or in 1 mL PBS as control. Mice were imaged using microCT, as described above, to demonstrate abdominal distension. The IPP changes correlated with these injection volumes is unknown. Mice injected with RFP-tagged OVCAR5 or OVCAR8 were euthanized by CO 2 inhalation after 5 and 8 h, respectively, followed by cervical dislocation and then rapidly dissected using a ventral midline incision.
In some experiments, only PBS (5 mL or 10 mL) was injected in the absence of OVACR cells and peritoneal tissues were processed for scanning electron microscopy as described below. After skin removal, the parietal peritoneum was dissected and vigorously washed in PBS (five times) and mounted onto a glass coverslip. Adherent cells were imaged with either the Echo Revolve fluorescent microscope or EVOS FL digital inverted fluorescence microscope and cells were counted manually using ImageJ software. The assay was performed in three experimental replicates and repeated in three biological replicates for all conditions.
Fluorescence microscopy. To  Compression of ex vivo murine peritoneum explants. C57Bl/6 female mice were euthanized by CO 2 inhalation followed by cervical dislocation and then rapidly dissected using a ventral midline incision. After skin removal, the parietal peritoneum lining the ventral abdominal wall was dissected to remove a 1.2 × 1.2 cm 2 piece of peritoneal tissue immediately adjacent to the midline in the lower two abdominal quadrants. The tissue explants were placed into the foam sample holders of six-well BioPress culture plates with silicone elastomer well bottoms (Flexcell International Corporation, Hillsborough, NC, USA). Fresh complete culture medium (4 mL) was added to each well of a BioPress plate, stationary platens were inserted into the culture plate. Samples were incubated at 37 °C in 5% CO 2 under static compression (~ 3 kPa; ~ 22 mmHg) applied for 1 h. The tissue explants were fixed and processed for scanning electron microscopy imaging as described below. The assay was performed in three experimental replicates and repeated in three biological replicates. Peritoneal tissues from three mice were used in the experiment and TNT were quantified in 10 images per mouse peritoneum using ImageJ. Data are presented as mean ± s.d. or standard error of the mean (s.e.m.). Comparison between groups was performed using Student two-tailed t-test to determine p values. P value < 0.05 was considered significant.
Scanning electron microscopy. Murine peritoneal explants were dissected from 4-6 month old C57Bl/6J mice as described above and processed for scanning electron microscopy (SEM). Tissues were fixed for SEM using a fixative containing 2% glutaraldehyde and 2% paraformaldehyde in 0. www.nature.com/scientificreports/ as described above. TNT were quantified in 10 cells under control or compressed conditions using ImageJ. Experiments were conducted in triplicate with three biological replicates. Data are presented as mean ± s.d. or standard error of the mean (s.e.m.). Comparison between groups was performed using Student two-tailed t-test to determine p values. P value < 0.05 was considered significant.
Mitochondria transport in LP9-OvCa cell co-culture. Mitochondria in GFP-tagged LP9 cells were labeled with MitoTracker Red CMXRos according to manufacturer protocol (ThermoFisher Scientific) then cocultured with either OVCAR5 or OVCAR8 cells at 1:1 ratio atop of rat tail collagen I-coated coverslips (10 µg/ mL). Co-cultured cells were then placed in Flexcell compression plates under compression (~ 3 kPa; ~ 22 mmHg) for 24 h. Cells were washed in PBS, fixed and imaged with a Leica DM5500 fluorescence microscope (Leica, Biosystems, Inc.).
Fluorescence/SHG imaging of murine peritoneal collagen. Female C57Bl/6 mice were i.p injected with 5 × 10 6 RFP-tagged ID8-Trp53 −/− or ID8-Trp53 −/− BRCA −/− cells. Mice were divided into two groups of 3 mice each. One group was euthanized 2 weeks post injection to allow formation of peritoneal metastasis without ascites formation; the second group was euthanized 5-6 weeks post injection after formation of peritoneal metastasis and accumulation of ascites (6-9 mL). Mice were sacrificed by CO 2 inhalation followed by cervical dislocation. The parietal peritoneum was dissected, rinsed with PBS and placed between coverslips for imaging with the mesothelium side facing the objective (25X XLPlanN, 1.05na WATER) of the 2-Photon confocal microscope (Olympus FV1000, software FLUOVIEW FV1000). Using a Mai Tai DeepSee titanium-sapphire 690-1040 nm laser, RFP-tagged ID8 metastatic implants and peritoneal collagen were visualized using the RFP and SHG signals, respectively. At 12% laser power, the 2-photon laser was set to 860 nm and emission was simultaneously collected at 425-465 nm and 575-625 nm for SHG and RFP, respectively. Quantification of fiber anisotropy from SHG images of collagen was done using the FibrilTool plugin in ImageJ 60 . Data are presented as mean ± standard error of the mean (s.e.m.).

Statistical analysis.
All experiments were conducted in a minimum of three independent replicates. The statistical analysis of the data was done using GraphPad Prism software or Excel software. Data are presented as mean ± s.d. or standard error of the mean (s.e.m.). Comparison between groups was performed using Student two-tailed t-test to determine p values. P value < 0.05 was considered significant.