Macrophage-Mediated Vascular Permeability Via VLA4/VCAM1 Pathway Dictates Ascites Development in Ovarian Cancer

Macrophage-mediated vascular permeability via VLA4/VCAM1 pathway dictates ascites development in ovarian cancer

Shibo Zhang, … , Changqing Liu, Huanhuan He

J Clin Invest. 2020. https://doi.org/10.1172/JCI140315.

Research In-Press Preview Cell biology Vascular biology

Graphical abstract

Find the latest version: https://jci.me/140315/pdf Macrophage-mediated Vascular Permeability via VLA4/VCAM1 Pathway Dictates

Ascites Development in Ovarian Cancer

Shibo Zhang1*, Bingfan Xie2*, Lijie Wang1*, Hua Yang2*, Haopei Zhang1, Yuming Chen1,

Feng Wang2, Changqing Liu2#, Huanhuan He1#

1Guangdong Provincial Key Laboratory of Biomedical Imaging and Guangdong

Provincial Engineering Research Center of Molecular Imaging, The Fifth Affiliated

Hospital, Sun Yat-sen University, Zhuhai, China 519000

2Department of Gynaecology, The Fifth Affiliated Hospital of Sun Yat-sen University，

Zhuhai, China 519000

* These first authors contributed equally to this work

# The two senior authors contributed equally to this work

Correspondence to: Professor Huanhuan He ([email protected]), Guangdong

Provincial Key Laboratory of Biomedical Imaging, The Fifth Affiliated Hospital of Sun

Yat-sen University, No. 52 Mei Hua Dong Road, Zhuhai, China 519000, PR China.

Phone: (+86)756-2526143; Professor Changqing Liu ([email protected]),

Department of Gynaecology, The Fifth Affiliated Hospital of Sun Yat-sen University, No.

52 Mei Hua Dong Road, Zhuhai, China 519000, PR China. Phone:

(+86)13926947846.Conflict of interest: The authors have declared that no conflict of interest exists.

1 Abstract: The development of ascites correlates with advanced-stage disease and poor prognosis in ovarian cancer. Vascular permeability is the key pathophysiological change involved in ascites development. Previously, we provided the first evidence that perivascular M2-like macrophages protect the vascular barrier through direct contact with endothelial cells (ECs). Here, we investigated the molecular mechanism and its clinical significance in the ovarian cancer setting. We found that upon direct coculture with the endothelium, M2 macrophages tuned down their VLA4 and reduced the levels of

VCAM1 in ECs. On the other hand, ectopically overexpressing VLA4 in macrophages or

VCAM1 in ECs induced hyperpermeability. Mechanistically, downregulation of VLA4 or VCAM1 led to reduced levels of RAC1 and reactive oxygen species (ROS), which resulted in decreased phosphorylation of PYK2 (p-PYK2) and VE-cadherin (p-VE-cad), hence enhancing cell adhesion. Furthermore, targeting the VLA4/VCAM1 axis augmented vascular integrity and abrogated ascites formation in vivo. Lastly, VLA4 expression on the macrophages isolated from ascites dictated permeability ex vivo.

Importantly, VLA4 antibody acted synergistically with bevacizumab to further enhance the vascular barrier. Taken together, we reveal here that M2 macrophages regulate the vascular barrier though the VCAM1/RAC1/ROS/p-PYK2/p-VE-cad cascade, which provides specific therapeutic targets for the treatment of malignant ascites.

2 Introduction

Ovarian cancer is the fifth leading cause of cancer-related death, and it has the highest mortality rate among all gynecological cancers worldwide (1). Ascites, ﬂuid accumulation in the peritoneal cavity, has been observed in 44.1% of patients at the time of epithelial ovarian cancer (EOC) diagnosis (2, 3), and it develops more frequently in

EOC than in any other tumor types (4). Ascites can deteriorate the quality of life of patients, causing abdominal pain and respiratory distress (5, 6). The onset and progression of ascites correlate with poor prognosis and disease recurrence (7). It is believed that ascites formation is due to increasing intraperitoneal vascular permeability combined with impaired lymphatic drainage (6, 8). However, current management of malignant ascites based on these understandings is unsatisfactory, making it imperative to gain more insight into the mechanism of ascites development so that more effective treatment strategies can be developed.

Changes in vascular permeability play an important role in tissue fluid homeostasis (9).

In tumor tissues, tumor vessels are often intricate, barely covered by mural cells and loosely associated with the basement membrane. Therefore, increased capillary permeability, which is orchestrated by ample vascular endothelial growth factor (VEGF) secreted into the microenvironment, leads to the overflow of ascites (10, 11). However, our previous study suggests a new strategy to regulate vascular permeability, i.e., M2-like

3 macrophages can protect the vascular barrier through direct contact with the endothelium

(12). This observation begs the question of whether macrophage-mediated vascular permeability also participates in pathological events such as ascites development in ovarian cancer (OC).

Integrins are a large group of heterodimeric adhesion molecules that are indispensable regulators of the vascular barrier (13-15). A4β1 integrin, also known as very late antigen-4 (VLA4), is primarily expressed on leukocytes, and often binds to vascular cell adhesion molecule 1 (VCAM1) on endothelial cells (ECs) to mediate leukocyte trafficking upon inflammation. The activation of VCAM1 promotes the phosphorylation of VE-cadherin (p-VE-cad), resulting in the loss of cell adhesion (16). Utilizing this signaling pathway from the opposite perspective, we demonstrated here that M2 macrophages inhibited endothelial VCAM1 expression upon direct contact, which suppressed the downstream cascade and dampened p-VE-cad, thus, protecting the vascular barrier. Targeting the VLA4/VCAM1 axis could attenuate vascular permeability both in vitro and in vivo and, more importantly, could inhibit ascites formation in an OC animal model. Our findings uncovered the molecular mechanism behind M2 macrophage-mediated vascular permeability and provide new strategies for treating ascites in OC.

4 Results

M2 macrophages induce hypopermeability by downregulating VCAM1 in ECs

Our previous study discovered that perivascular M2-like macrophages could protect against vascular permeability, but M1s could not. To explore the mechanism underlying the effect of macrophages on vascular permeability, we first evaluated the proteins involved in the regulation of vascular integrity, such as the adherens junction protein

VE-cadherin and the tight junctional protein ZO-1. We found that the human umbilical vein endothelial cell (HUVEC) layer cocultured with M2 macrophages (THP-1) for 24 hours experienced a significant reduction in the levels of p-VE-cad at the Tyr658 site

(Y658; Supplemental Figure 1A). However, the expression of ZO-1 (Supplemental

Figure 1B) and the level of p-VE-cad at the Tyr731 site (Y731) were unchanged

(Supplemental Figure 1C).

We then compared the gene expression profile of HUVECs that had been cocultured with different subtypes of macrophages. Among the profiles of differentially expressed genes

(DEGs), we focused on cell adhesion molecules (CAMs), since macrophage-mediated permeability requires direct cell-cell contact (Figure 1A). Among all the CAMs, VCAM1 was the most markedly changed (Figure 1B). To validate this result, we isolated endothelial cells from the coculture system to detect the expression of VCAM1 mRNA and protein, and obtained consistent conclusion (Supplemental Figure 1D, E). Notably,

5 expression of VCAM1 protein in HUVECs was downregulated upon direct contact with

M2 THP-1 macrophages, detected by both western blot and immunofluorescence staining

(Figure 1C, D). This finding was validated by coculturing murine macrophages

(RAW264.7) with a murine endothelial cell line (C166; Supplemental Figure 1F, G).

Interestingly, such effect vanished in the coculture performed via a transwell system

(Supplemental Figure 1H), suggesting contact dependent regulation of endothelial

VCAM1 by M2 macrophages. We also measured the levels of soluble VCAM1

(sVCAM1), and found no difference in HUVECs cocultured with M1 versus M2 macrophages (Supplemental Figure 1I). To validate the crucial role of VCAM1 in regulating macrophage-mediated permeability, we knocked down VCAM1 expression in

HUVECs using short hairpin RNAs (shRNA; Supplemental Figure 1J) and observed attenuated p-VE-cad in EC-M1 macrophage coculture (Figure 1E). Immunofluorescence staining confirmed the down-regulation of p-VE-cad upon VCAM1 knockdown (Figure

1F). Alternatively, when applying a VCAM1 inhibitor K-7174 to block VCAM1 expression in coculture, the p-VE-cad level was clearly abrogated (Supplemental Figure

1K). Conversely, VCAM1 overexpression in HUVECs and murine ECs promoted p-VE-cad (Figure 1G, H; Supplemental Figure 1L, M).

Next, we sought to confirm that VCAM1 could functionally affect the macrophage-mediated permeability. Macrophages and ECs were cocultured in the upper

6 well of a transwell, and dextran labeled with TRITC fluorescent dye was allowed to pass through the barrier, which was then collected in the lower chamber and quantified (17).

The data revealed that there was much less dextran that passed through in the VCAM1 knockdown group relative to that of the control group (Figure 1I), whereas increased flux of dextran was observed in the VCAM1-overexpressing coculture (Figure 1J).

Together, these findings demonstrated that macrophages exert their functions of regulating the vascular barrier by affecting VCAM1 expression on the endothelium.

VLA4 expression and activation are dampened in M2 macrophages cocultured with

ECs

To determine which molecule in M2 macrophages dictates differential regulation in ECs, we scrutinized the gene expression profiles of both subtypes of macrophages isolated from coculture and found that the pathway associated with leukocyte transendothelial migration was enriched and downregulated in M2 macrophages (Figure 2A). Among all the genes participating in leukocyte transendothelial migration, it was noteworthy that the expression of VLA4, the ligand of VCAM1, decreased markedly in M2 macrophages

(Figure 2B). We confirmed this result by western blot, immunofluorescence staining and quantitative RT-PCR (Figure 2C, D; Supplemental Figure 2A, B). VLA4 protein expression was also reduced in M2-polarized murine macrophages cocultured with murine endothelium (Supplemental Figure 2C, D). To be noted, the expression of the

7 active form of VLA4 in M2 macrophages coculture was also lessened compared to that of the M1 macrophages coculture (Figure 2E).

To validate the regulatory effects of macrophage VLA4 on endothelial cell barriers,

THP-1 or RAW264.7 cells were subjected to inhibited or excessive VLA4 expression by shRNAs (Supplemental Figure 2E, F) or overexpression plasmids (Supplemental

Figure 2G, H), followed by coculturing with ECs. Consistent with the VCAM1 data,

VLA4 silencing in macrophages or pretreatment of macrophages with the VLA4 inhibitor,

CDP323, attenuated p-VE-cad upon coculture in both HUVECs (Figure 2F, G) and

C166s (Supplemental Figure 2I, J); however, increased p-VE-cad was detected when

VLA4 was overexpressed or VLA4 agonist, THI0019, was administered (Figure 2H-J;

Supplemental Figure 2K, L). Dextran permeability assay showed that suppressing

VLA4 expression in macrophages either by treating with its inhibitor (CDP323) or antibody (PS/2) blocked the dextran from crossing the EC barrier, while enhancing

VLA4 expression or applying its agonist (THI0019) exacerbated EC hyperpermeability

(Figure 2K). Overall, these results indicated that the level of VLA4 on macrophages contributed to the regulation of the vascular barrier.

ROS functions downstream of VCAM1 to regulate macrophage-mediated permeability

8 It has been reported that the accumulation of reactive oxygen species (ROS) could ultimately lead to endothelial dysfunction (18). To determine whether ROS could play a role in macrophage-mediated permeability, we first measured ROS expression in the coculture system. As shown in Figure 3A, endothelial ROS was greatly abolished when cocultured with M2 macrophages compared to that with M1s, and this difference was further enhanced following treatment with the vascular permeator VEGF (Figure 3B).

Moreover, the addition of CDP323 (Figure 3C) and K-7174 (Figure 3D) effectively blocked intracellular ROS production in HUVECs. In contrast, the ROS level in the

VLA4-overexpressing coculture was higher than that in the control (Supplemental

Figure 3A, B), suggesting that ROS was the downstream effector of M2 macrophage-mediated vascular permeability.

To confirm that the regulation of p-VE-cad by macrophages depends on ROS, a ROS inhibitor N-acetyl-l-cysteine (NAC) was added to the coculture system. To support our hypothesis, p-VE-cad was significantly abolished in ECs cocultured with macrophages in the presence of NAC (Figure 3E; Supplemental Figure 3C), whereas p-VE-cad had an evident increase after H2O2 was introduced in the EC-macrophage cocultures (Figure 3F;

Supplemental Figure 3D). We confirmed the above findings by immunofluorescence staining (Figure 3G, H; Supplemental Figure 3E). Barrier function detected by

CellZscope® device confirmed that vascular permeability was abolished upon NAC

9 treatment, while the barrier was markedly compromised with H2O2 treatment in ECs cocultured with macrophages (Figure 3I, J; Supplemental Figure 3F). Collectively, these data confirmed that the VLA4/VCAM1-initiated cascade could induce vascular hyperpermeability via enhancing ROS.

Macrophage-mediated permeability is dependent on RAC1 and phosphorylated

PYK2

RAS-related C3 botulinum substrate 1 (RAC1) and proline-rich tyrosine kinase 2

(PYK2) have been identified as potential regulators of p-VE-cad, and they can potentially regulate ROS levels up- and downstream, respectively (19). Downregulated total RAC1 and phosphorylation of PYK2 (p-PYK2; Tyr402), but not total PYK2, were observed in the coculture system with M2 macrophages compared to that in the M1 cocultures

(Figure 4A, Supplemental Figure 4A, B). Consistent with previous results, such change was not observed in the coculture without direct contact (Supplemental Figure 4C).

Immunofluorescence staining confirmed the above changes (Figure 4B). Strikingly, using VLA4 inhibitor, CDP323, to treat macrophages diminished both RAC1 and p-PYK2 levels in EC-macrophage cocultures (Figure 4C; Supplemental Figure 4D).

RAC1 expression level in the coculture was increased when treating macrophages with the VLA4 agonist THI0019 (Supplemental Figure 4E, F). Furthermore, knockdown of

VCAM1 dampened RAC1 expression, while VCAM1 overexpression promoted RAC1

10 expression (Supplemental Figure 4G-I). Concomitantly, reducing VCAM1 level by treating with its inhibitor abrogated both RAC1 and p-PYK2 levels in EC-macrophage cocultures (Figure 4D; Supplemental Figure 4J). To prove that ROS is regulated by

RAC1, we applied RAC1 inhibitor, NSC23766, and detected lower ROS level in the cocultures (Figure 4E; Supplemental Figure 4K). Since PYK2 is activated through

ROS in endothelial cells (20), it was expected that NAC treatment reduced p-PYK2 expression in HUVECs (Figure 4F), whereas H2O2 treatment increased p-PYK2 expression (Figure 4G; Supplemental Figure 4L, M). Taken together, the above results suggested that the RAC1/ROS/p-PYK2 acted downstream of VCAM1 and participated in the regulation of macrophage-mediated vascular permeability.

Blocking VLA4 reduces permeability and lessens ascites in vivo

To further investigate whether VLA4 is responsible for regulating macrophage-mediated vascular permeability in vivo, we resorted to a rescue approach. First, macrophages were largely depleted using clodronate liposomes via intraperitoneal injection in mice, and the depletion of peritoneal macrophages was verified by flow cytometry (Supplemental

Figure 5A). Then, RAW264.7 cells with either suppressed or promoted level of VLA4 were reintroduced via intraperitoneal injection. After two days of reconstruction, peritoneal vascular permeability was measured by fluorescent dye that was injected intravenously and then leaked into the peritoneal cavity (Figure 5A). In line with the in

11 vitro data, the leaky vascular barrier was reconstructed when VLA4 was either knocked down (Figure 5B) or blocked by a functional VLA4 blocking antibody, PS/2

(Supplemental Figure 5B). Furthermore, the leakage was more intense in

VLA4-overexpressing group compared to the control group (Figure 5C). These results proved that the VLA4 level was positively associated with vascular permeability in vivo, underscoring the therapeutic potential of targeting VLA4. Moreover, upon macrophage removal, the effect of PS/2 on vascular permeability was weakened (Figure 5D), which suggested that the effect of VLA4 on permeability was macrophage-dependent.

To test the efficacy of blocking VLA4 for the treatment of ascites, we resorted to a murine OC model, HM-1. PS/2 (3mg/kg) was administered to the mouse abdomen according to the scheme shown in Figure 5E. Compared to the isotype control group, ascites volume in the PS/2 treatment group was abrogated by more than 60.9% (Figure

5F, G). We also examined the alterations of the RAC1/ROS/p-PYK2 cascade in the endothelium in HM-1 tumors upon PS/2 treatment. Consistently, PS/2 treatment resulted in less ROS positive cells in endothelial cells isolated from HM-1 tumors (Figure 5H).

RAC1, p-PYK2 and p-VE-cad levels were also markedly reduced compared with controls by immunofluorescence staining (Figure 5I-M). Taken together, these findings illustrated that VLA4 indeed played a crucial role in macrophage-mediated permeability, which provided a therapeutic target for the treatment of ascites.

12 VLA4/VCAM1 expression is correlated with ascites volume

In order to verify our findings in the clinical setting, we first analyzed VLA4 expression in OC patient samples with or without signs of ascites that were found in the

Kaplan-Meier Plotter database. Intriguingly, patients with low VLA4 expression survived much longer than patients expressing higher levels of VLA4 (Figure 6A). More importantly, the expression level of VCAM1 and VLA4 in OC tissues were correlated with the ascites volume in patients observed by immunohistochemical and immunofluorescence staining (Figure 6B-E). However, the soluble VCAM1 expression detected in ascites less than 500 ml (few) was identical to that in ascites more than 1000 mL (massive; Supplemental Figure 6A). Furthermore, flow cytometry analysis showed that more macrophages expressed VLA4 in massive ascites (Figure 6F) and macrophages from massive ascites expressed higher levels of VLA4 than those with fewer ascites (Supplemental Figure 6B). Interestingly, both patients and animals with massive ascites exhibited a lower ratio of M2 to M1 macrophages than those with fewer ascites (Figure 6G; Supplemental Figure 6C).

To confirm that macrophages from massive ascites could indeed induce permeability, we isolated macrophages from patient and mouse ascites, cocultured them with ECs and measured the barrier changes by dextran permeability assay and by the electric cell

13 impedance sensing (ECIS) equipment. Both assays validated our hypothesis (Figure 6H,

I; Supplemental Figure 6D). Blocking either VLA4, VCAM1 or ROS rescued the compromised vascular barrier (Figure 6J; Supplemental Figure 6E). In summary, these results strongly suggest that the levels of VCAM1 and VLA4 positively correlate with ascites development in OC patients and may be used as therapeutic targets for ascites treatment.

VLA4/VCAM1 axis dictates permeability independent of the VEGF pathway

To further discern the role of VEGF pathway in macrophages-direct-contact-mediated permeability, we used bevacizumab to block the VEGF signaling. After adding bevacizumab to the coculture system, both permeability and p-VE-cad expression were significantly reduced in the coculture with M1 macrophages, but not altered as much in

M2 macrophage-cocultured system (Figure 7A, B), suggesting that the regulation of vascular permeability by M2 macrophages did not, at least not totally, depend on VEGF.

Furthermore, there was no correlation between VEGF and macrophage VLA4 expression in ascites of OC patients (Supplemental Figure 7A). Therefore, we speculated that the two pathways worked separately in this system. To prove this point, we combined antibodies that block VEGF and VLA4 respectively to treat the cocultures, and obtained a synergistic effect in reducing permeability (Figure 7C). Next, we compared the effects of each individual and the combinatorial treatment in HM-1 animals. B-scan

14 ultrasonography was conducted to measure the volume of ascites in the abdominal cavity.

We observed that the volumes of ascites were strikingly reduced in the combinatorial treatment group although individual treatment was also effective and comparable (Figure

7D, E). The survival data also supported the above findings (Figure 7F).

Lastly, we investigated the potential effect of blocking VLA4 on macrophage polarization, and found that PS/2 treatment could significantly enhance M1-markers such as IL6, NOS2 and IL12B, meanwhile significantly decreased the levels of M2-markers such as ARG1 and CD206 in M1 macrophages (Figure 7G). For M2 macrophages, PS/2 treatment consistently induced the expression of IL6, NOS2 and IL12B, which are often expressed by M1 macrophages and related to anti-tumor effect, meanwhile decreased the levels of M2-markers such as IL10 and ARG1 (Figure 7H). To be noted, VEGF expression was not significantly changed in both groups. These results indicate that PS/2 treatment may increase the tumor-killing capability of macrophages while exerting the barrier protection function. In summary, all the data suggest that M2 macrophages can regulate endothelial cells permeability in a pathway that is independent of VEGF and that anti-VLA4 therapy may be used as therapeutical targets to treat malignant ascites.

Discussion

15 Malignant ascites is observed in most terminal ovarian cancers, and it markedly contributes to mortality and poor quality of life. However, the formation and regulation of ascites are poorly understood, and targeted therapy to treat ascites has not been developed. In this study, we demonstrated that M2 macrophages could tune down the

VCAM1 expression in the endothelium and suppress the RAC1/ROS/p-PYK2/p-VE-cad downstream cascade, thus enhancing vascular integrity. Targeting the VLA4/VCAM1 axis could attenuate vascular permeability both in vitro and in vivo and, more importantly, inhibit ascites formation in OC animal models. Our findings uncovered a molecular mechanism of M2 macrophage-mediated vascular permeability and provide new targets to treat ascites in OC patients.

Macrophages actively participate in the maintenance of vascular homeostasis. Since we reported that macrophage depletion led to the loss of vascular integrity (12), several studies have observed similar effects in various organs. For example, when macrophages are depleted in stria vascularis of the cochlea, tight junctions between endothelial cells become unstable, resulting in increased vessel permeability and, ultimately, the loss of hearing (21). Similarly, in the blood-brain barrier (BBB), when endothelial cells isolated from the mouse brain were cocultured with macrophages, the permeability was markedly reduced (22). However, the role of macrophage-mediated permeability in OC ascites formation has not been elucidated. Moreover, tumor-associated macrophages (TAMs) are

16 believed to be more M2-like, which is detrimental to health because it promotes tumor growth and angiogenesis (23). Although a high ratio of M1/M2 macrophages predicts an improved prognosis, as reported in an OC study (24), our data suggest that a high M1/M2 ratio is also associated with more ascites in both murine OC models and OC patients, likely due to loss of the protective role of the M2 macrophages in the vascular barrier.

These findings uncovered a new role for M2 macrophages or TAMs in OC development.

Given that ascites can be detrimental to the quality of life and the progression of the disease, eliminating macrophages through techniques such as treatment with anti-CSF1R

(25) or through the reprogramming of M2 to M1 macrophages using thiazolidinediones

(TZDs) (26) may not be the wisest strategies in terms of prolonging the overall survival and quality of life of OC patients.

Direct interactions between macrophages and ECs are critical in pathological conditions such as inﬂammation and atherosclerosis. For instance, M1 macrophage-induced inflammatory responses involve monocytes tethering to the endothelium using selectins and subsequently monocytes using VLA4, also known as α4β1 integrin, to engage

VCAM1 or intercellular adhesion molecule 1 (ICAM-1) on the endothelium to form a firm adhesion; this then triggers downstream RAC1/ROS/p-PYK2/p-VE-cad signaling and ultimately leads to hyperpermeability (27-31). Interestingly, we found that M2 macrophages used the same pathway, but they use it in reverse to achieve a different, or

17 rather opposite, effect. Our data showed that M2 macrophages hampered VCAM1 expression in the endothelium and tuned down RAC1/ROS/p-PYK2/p-VE-cad signaling, leading to an enhanced vascular barrier. We think this is an efficient way for macrophages to regulate vascular permeability in a dynamic environment.

Malignant ascites presents a considerable clinical challenge to the management of OC, yet there is a lack of treatment options at present. Diuretics are frequently used in the treatment of ascites arisen from nonmalignant disease (2). However, no solid evidence has been shown to date for the efficacy of diuretics in malignant ascites (32, 33).

Alternatively, intraperitoneal administration of bleomycin, mitoxantrone, or fluorouracil in combination with cisplatin has been reported to improve the condition of ascites

(34-37). Nonetheless, the specificity for anti-ascites was not fully evaluated, and the anti-ascites mechanism has not been discussed. Another popular strategy to control ascites is to inhibit VEGF signaling, since VEGF is a well-known vascular permeator secreted by tumor cells and immune cells. Therefore, bevacizumab, an antibody that functionally blocks VEGF, has seen an increase in clinical use. Yet resistance to anti-VEGF therapy remains a key problem, and efforts need to be made to understand the underlying mechanisms (38). Here, we describe a specific and potent ascites-targeting antibody, PS/2, that blocks VLA4 and downstream signaling and shows significant efficacy in abolishing ascites development. Since the VLA4/VCAM1 pathway is distinct

18 from the VEGF signaling regarding the regulation of vascular permeability, when bevacizumab was administered in our coculture system, M2-mediated endothelial permeability was not altered as much as M1’s. In addition, when combining PS/2 and bevacizumab, we obtained a synergistic anti-permeability effect and extended survival rate in OC-bearing mice. These results suggest that the M2 macrophage-mediated

VLA4/VCAM1 pathway can regulate endothelial cell permeability independent of the

VEGF pathway.

In summary, we found that VLA4 in M2 macrophages and VCAM1 in ECs were both downregulated upon direct engagement of these two types of cells, which suppressed the downstream RAC1/ROS/p-PYK2/p-VE-cad pathway, ultimately leading to enhanced vascular integrity. Our results provide an improved understanding of the key role that macrophages play in the regulation of vasculature under the context of OC and reveal a novel therapeutic target for the treatment of malignant ascites.

19 Methods

Patients and samples

Under institutional review board approval (Medical Ethics Committee of the Fifth

Affiliated Hospital of Sun Yat-sen University, Zhuhai, China; 2018-K181-1) and with informed consent, fresh tumor, nontumor ovarian tissues and ascites were collected from

32 patients diagnosed with high-grade serous ovarian carcinoma at the Fifth Affiliated

Hospital of Sun Yat-sen University. Ascites (taken on day of surgery) from ovarian cancer patients who underwent surgical resection were used for isolating macrophages.

The clinical characteristics of all patients in two cohorts are listed in Supplemental

Table 1.

Cell lines and cell culture

Human umbilical vascular endothelial cells (HUVECs), a murine endothelial cell line

(C166) and a murine macrophage cell line (RAW264.7) were obtained from the

American Type Culture Collection (ATCC, Manassas, USA). A human macrophage cell line (THP-1) were kindly gifted by Professor Dong-Ming Kuang (SYSU). A murine ovarian carcinoma cell line, OV2944-HM-1 (HM-1) cells were kindly gifted by Professor

Nelson Teng (Stanford University). The cells were authenticated by short tandem repeat profiling and were confirmed to be mycoplasma free before use. HUVECs were cultured in ECM (ScienCell) supplemented with 5% fetal bovine serum (FBS), 1% endothelial

20 cell growth supplement (ECGS), 100 U/ml penicillin and 100 μg/ml streptomycin. THP-1 cells and HM-1 cells were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin, and 100 ug/ml streptomycin. C166 and

RAW264.7 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM;

Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (FBS), 100

U/ml penicillin, and 100 µg/ml streptomycin. All cells were maintained in a humidified incubator with 5% CO2 that was at 37 °C.

Western blot

The protein samples determined by BCA assay (Beyotime Biotechnology, Shanghai,

China) were separated by 4%-12% SDS-PAGE and transferred onto PVDF membrane.

The following primary antibodies were used in this study are listed in Supplemental

Table 2.

Permeability assay

HUVECs (1 × 105 cells) were seeded on top of a transwell insert in 24-well transwell chambers and were cultured for 24 hours till reaching confluence. THP-1 (5 × 104 cells) polarized to different subtypes of macrophages were seeded on endothelial monolayers for another 24 hours. The samples were randomly divided into several groups: control group, CDP323-treated group, PS/2-treated group, THI0019-treated group,

21 K-7174-treated group, NAC-treated group and H2O2-treated group. The lower compartments were filled with medium alone (control group), medium containing

CDP323 (2 μmol/L; CDP323 group), PS/2 (2 μmol/L, PS/2 group), THI0019 (2 μmol/L,

THI0019 group), K-7174 (2 μmol/L, K-7174 group), NAC (5 μmol/L, NAC group), or

H2O2 (2 μmol/L, H2O2 group). After 24 hours incubation, 100 μl of phosphate buffered saline (PBS) was used to wash the chambers twice. Then, 500 μl of PBS was added to the lower layer, and 200 μl of PBS containing TRITC (2 mg/ml) was added to the upper layer. The entire transwell chamber was incubated for 3 hours at 37 °C. The PBS in the lower compartments was collected and analyzed in a fluorescence spectrophotometer

(Synergy HTX, BIOTEK) with excitation and emission wavelengths of 530 nm and 540 nm, respectively.

Alternatively, HUVECs were seeded onto 24-well transwell and treated with chemicals as indicated. Then the barrier function was measured with the automated cell monitoring

CellZscope® system (NanoAnalytics, Munster, Germany) for 24h in CO2 (5 %) incubator at 37 °C.

ROS level measurements

ROS levels were measured with the dichlorodihydrofluorescein diacetate (DCFH-DA) assay (Beyotime Institute of Biotechnology, Nantong, China) according to the

22 manufacturer’s instructions. Briefly, the cells were resuspended in PBS and then were stained with 10 μM DCFH-DA for 20 min at 37 °C in the dark. After washing with PBS, the samples were detected by flow cytometry with an excitation wavelength of 488nm and an emission wave length of 525 nm.

Generation of lentiviral particles and cell transfection

VLA4 knockdown or stable overexpression in RAW264.7 cells was established by infecting cells with a lentiviral vector (U6-sh-VLA4-EGFP-IRES-puromycin) purchased from GeneCopoeia (Rockville, Md, USA). RAW264.7 cells were infected with lentivirus using 0.8 μg/ml polybrene. Stable clones were selected after two weeks using 1 µg/ml puromycin. The efficiency of RNA interference was assessed via qRT-PCR and western blot.

RNA interference

HUVECs were transfected for 24 hours with a nontargeting shRNA or a VCAM1 shRNA using Lipofectamine 3000 (Invitrogen, Thermo Fisher Scientific) and then were assayed

72 hours after transfection. Negative control shRNA (shRNA-Control) and the empty vector (pEZ-Vec) were provided as the control of shRNA-VCAM1 and pEZVCAM1 experiments, respectively. A plasmid targeting VLA4 (pEZ-VLA4, GenePharma, China) was used to upregulate VLA4 introducing it to cells. The constructs (1.5 µg) were then

23 transfected into 1×106/ml THP-1 cells (ATCC) using Lipofectamine 3000 (Invitrogen;

Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions, and then they were assayed 72 hours after transfection.

Immunofluorescence staining

Cocultures cells were seeded on chamber slides and cultured for 24 hours. Cells were washed with phosphate buffered saline (PBS), fixed with pre-cooled paraformaldehyde for 10 min at 4°C, permeabilized with 1% TritonX-100 (Solarbio Life Science, Beijing,

China) in PBS for 15 min, incubated in blocking solution for one hour at room temperature, and incubated with specific primary antibodies overnight at 4°C. Then, the samples were washed thrice with PBS, incubated with secondary antibodies for one hour at room temperature in dark place, washed thrice with PBS, and incubated for 10 min in the dark with Fluoroshield Mounting Medium with DAPI (Abcam, USA). Images were acquired using confocal microscope.

Immunohistochemical (IHC) staining

Samples of OC tissues were acquired during surgery at the Fifth Affiliated Hospital of

Sun Yat-sen University. Tissues were fixed in 10% neutral-buffered formalin overnight, dehydrated using increasing gradient concentrations, and embedded in low-melting point paraffin. Continuous 4um thick tissue sections were cut and fixed onto silicified slides.

24 Immunohistochemistry was carried out using streptavidin – peroxidase-conjugated method. Briefly, each tissue section was deparaffinized, rehydrated, immersed in antigen retrieval solution, boiled at 100°C for 10 min, and incubated with a peroxidase inhibitor.

Then, non-specific binding was blocked with normal goat serum, and tissue sections were incubated overnight at 4 ° C with the following primary antibodies: anti-VCAM1 acquired from Abcam (USA). All antibodies were used at a 1:50 dilution.

Animal experiments

All animal procedures were performed in accordance with the Guide for the Care and Use of Laboratory Animals (NIH publications Nos. 80-23, revised 1996) and Institutional

Ethical Guidelines for Animal Experiments developed by the Fifth Affiliated Hospital of

Sun Yat-sen University. Animals were maintained under pathogen-free conditions in

Guangdong Provincial Key Laboratory of Biomedical Imaging. Five groups of 5 BALB/c female nude mice (Vital River Laboratory Animal Technology Co., Ltd, Beijing, China) aged four to six weeks were subcutaneously injected with clodronate and control liposomes were injected via intraperitoneal injection (i.p.) with 200µl/mouse on day 1.

On day 3, liposomes were injected again following the same method. VLA4 knockdown, overexpressed and control macrophages (106 cells in 100ul PBS) were delivered via intraperitoneal injection (i.p.). Two days later, 200µl TRITC-dextran (Invitrogen) were injected i.v. in each mouse. After specific time intervals, animals were euthanized and

25 1ml of PBS was injected i.p. to recover the peritoneal macrophages and liquid in the peritoneal cavity. The solution was then spun down, from which the supernatant was taken for fluorescence intensity quantification.

The B6C3F1 mouse aged six to eight weeks was produced as a cross between a male

C3H mouse and a female C57BL/6 mouse. For HM-1 animal models, 106 HM1 cells were injected into the peritoneal cavity via a left lower abdominal wall injection. Mice were treated with PS/2 (3mg/kg) every other day after 2days following injection. Mice were sacrificed at 2 weeks using CO2 gas per institutional protocols.

Statistical analysis

The results were presented as mean ± standard deviation of at least three independent experiments. GraphPad Prism software (version 5.0; GraphPad Software, La Jolla, CA) was used for all statistical analyses. Differences between groups were evaluated using two Student’s t-tests when only two groups were analyzed, or by one-way ANOVA when more than two groups were compared. Pearson’s correlation was performed to analyze the correlation between protein expression of VLA4 and VEGF. P<0.05 was considered statistically significant.

Study approval

26 All animal experiments were performed with the approval of the Experimental Animal

Ethics Committee of the Fifth Affiliated Hospital of Sun Yat-sen University. For experiments using human samples, all samples were anonymously coded in accordance with local ethical guidelines (as stipulated by the Declaration of Helsinki). Written informed consent was obtained from patients, and the protocol was approved by the

Medical Ethics Committee of the Fifth Affiliated Hospital of Sun Yat-sen University.

27 Author contributions

Conception and design: S Zhang, B Xie, L Wang, F Wang, C Liu, H He.

Development of methodology: S Zhang, B Xie, L Wang, H Yang, H Zhang.

Acquisition of data: S Zhang, B Xie, L Wang, H Yang, H Zhang, Y Chen.

Analysis and interpretation of data (e.g., statistical analysis, bio-statistics, computational analysis): S Zhang, L Wang, H Zhang, H He.

Writing, review, and/or revision of the manuscript: S Zhang, B Xie, L Wang, C Liu, H

He.

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S Zhang, B Xie, L Wang, H Yang, F Wang, C Liu, H He.

Study supervision: F Wang, C Liu, H He.

S Zhang, B Xie, L Wang and H Yang provided equal contribution in conceiving the overall study, performing experiments and conducting statistical analysis. H Yang is responsible for the clinical diagnosis and follow-up of the patients. C Liu and H He provided equal contribution in designing and overseeing the overall study, and writing and reviewing the manuscript. These authors deserve equal contribution respectively. The order of the first or last authors reflects the leadership exerted in the study.

28 Acknowledgements

The authors acknowledge Professor Luisa Iruela-Arispe (Northwestern University, USA) and Professors Nelson Teng and Oliver Dorigo (Stanford University, USA) for their technical and material support and suggestions on the manuscript, and Professor Man Li for statistical consultation. This work was supported by project grants from the National

Natural Science Foundation of China (Grant Nos.31741068 and 81872113) and support from the Sun Yat-sen University (Grant No. 18ykzd02).

29 References

1. Miller KD, Siegel RL, Lin CC, Mariotto AB, Kramer JL, Rowland JH, et al. Cancer

treatment and survivorship statistics, 2016. Ca-Cancer J Clin. 2016;66(4):271-289.

2. Kipps E, Tan DSP, and Kaye SB. Meeting the challenge of ascites in ovarian cancer:

new avenues for therapy and research. Nat Rev Cancer. 2013;13(4):273-282.

3. Ayantunde AA, and Parsons SL. Pattern and prognostic factors in patients with

malignant ascites: a retrospective study. Ann Oncol. 2007;18(5):945-949.

4. Parsons SL, Lang MW, and Steele RJ. Malignant ascites: a 2-year review from a

teaching hospital. Eur J Surg Oncol. 1996; 22(3): 237-239.

5. Fleming ND, Alvarez-Secord A, Von Gruenigen V, Miller MJ, and Abernethy AP.

Indwelling Catheters for the Management of Refractory Malignant Ascites: A

Systematic Literature Overview and Retrospective Chart Review. J Pain Symptom

Manag. 2009;38(3):341-349.

6. Tan DSP, Agarwal R, and Kaye SB. Mechanisms of transcoelomic metastasis in

ovarian cancer. Lancet Oncol. 2006;7(11):925-934.

7. N A, and KL S. Getting to know ovarian cancer ascites: opportunities for targeted

therapy-based translational research. Frontiers in oncology. 2013;3:256.

8. Cavazzoni E, Bugiantella W, Graziosi L, Franceschini MS, and Donini A.

Malignant ascites: pathophysiology and treatment. Int J Clin Oncol.

2013;18(1):1-9.

30 9. Ono S, Egawa G, and Kabashima K. Regulation of blood vascular permeability in

the skin. Inflamm Regen. 2017;37:11.

10. Fushida S, Oyama K, Kinoshita J, Yagi Y, Okamoto K, Tajima H, et al. VEGF is a

target molecule for peritoneal metastasis and malignant ascites in gastric cancer:

prognostic significance of VEGF in ascites and efficacy of anti-VEGF monoclonal

antibody. Oncotargets Ther. 2013;6:1445-1451.

11. Bamias A, Koutsoukou V, Terpos E, Tsiatas ML, Liakos C, Tsitsilonis O, et al.

Correlation of NK T-like CD3+CD56+ cells and CD4+CD25+(hi) regulatory T

cells with VEGF and TNF alpha in ascites from advanced ovarian cancer:

Association with platinum resistance and prognosis in patients receiving first-line,

platinum-based chemotherapy. Gynecol Oncol. 2008;108(2):421-427.

12. He H, Mack JJ, Guc E, Warren CM, Squadrito ML, Kilarski WW, et al.

Perivascular Macrophages Limit Permeability. Arterioscl Throm Vas.

2016;36(11):2203-2212.

13. Osada T, Gu YH, Kanazawa M, Tsubota Y, Hawkins BT, Spatz M, et al.

Interendothelial claudin-5 expression depends on cerebral endothelial cell-matrix

adhesion by beta(1)-integrins. J Cerebr Blood F Met. 2011;31(10):1972-1985.

14. Obermeier B, Daneman R, and Ransohoff RM. Development, maintenance and

disruption of the blood-brain barrier. Nat Med. 2013;19(12):1584-1596.

31 15. Malinin NL, Pluskota E, and Byzova TV. Integrin signaling in vascular function.

Curr Opin Hematol. 2012;19(3):206-211.

16. Vockel M, and Vestweber D. How T cells trigger the dissociation of the endothelial

receptor phosphatase VE-PTP from VE-cadherin. Blood. 2013;122(14):2512-2522.

17. Kevil CG, et al. Expression of zonula occludens and adherens junctional proteins in

human venous and arterial endothelial cells: role of occludin in endothelial solute

barriers. Microcirculation. 1998;5:197-210.

18. Sun HJ, Wu ZY, Nie XW, and Bian JS. Role of Endothelial Dysfunction in

Cardiovascular Diseases: The Link Between Inflammation and Hydrogen Sulfide.

Front Pharmacol. 2020;10:1568.

19. Wu JR, You RI, Hu CT, Cheng CC, Rudy R, and Wu WS. Hydrogen peroxide

inducible clone-5 sustains NADPH oxidase-dependent reactive oxygen

species-c-jun N-terminal kinase signaling in hepatocellular carcinoma. Oncogenesis.

2019;8(8):40.

20. Daou GB, and Srivastava AK. Reactive oxygen species mediate

endothelin-1-induced activation of ERK1/2, PKB, and PYK2 signaling, as well as

protein synthesis, in vascular smooth muscle cells. Free Radical Bio Med.

2004;37(2):208-215.

32 21. Okano T, and Kishimoto I. Csf1 Signaling Regulates Maintenance of Resident

Macrophages and Bone Formation in the Mouse Cochlea. Front Neurol.

2019;10:1244.

22. Rodrigues SF, and Granger DN. Blood cells and endothelial barrier function. Tissue

Barriers. 2015;3(1-2).

23. Mantovani A, Sozzani S, Locati M, Allavena P, and Sica A. Macrophage

polarization: tumor-associated macrophages as a paradigm for polarized M2

mononuclear phagocytes. Trends Immunol. 2002;23(11):549-555.

24. Zhang MY, He YF, Sun XJ, Li Q, Wang WJ, Zhao AM, et al. A high M1/M2 ratio

of tumor-associated macrophages is associated with extended survival in ovarian

cancer patients. J Ovarian Res. 2014;7:19.

25. Ries CH, Cannarile MA, Hoves S, Benz J, Wartha K, Runza V, et al. Targeting

Tumor-Associated Macrophages with Anti-CSF-1R Antibody Reveals a Strategy

for Cancer Therapy. Cancer Cell. 2014;25(6):846-859.

26. Wanderley CW, Colon DF, Luiz JPM, Oliveira FF, Viacava PR, Leite CA, et al.

Paclitaxel Reduces Tumor Growth by Reprogramming Tumor-Associated

Macrophages to an M1 Profile in a TLR4-Dependent Manner. Cancer Res.

2018;78(20):5891-5900.

27. Mestas J, and Ley K. Monocyte-Endothelial Cell Interactions in the Development

of Atherosclerosis. Trends Cardiovas Med. 2008;18(6):228-232.

33 28. Huo Y, and Ley K. Adhesion molecules and atherogenesis. Acta Physiol Scand.

2001;173(1):35-43.

29. Shattil SJ, and Ginsberg MH. Perspectives series: cell adhesion in vascular biology.

Integrin signaling in vascular biology. The Journal of Clinical Investigation.

1997;100(1):1-5.

30. Elices M, et al. VCAM1 on activated endothelium interacts with the leukocyte

integrin VLA4 at a site distinct from the VLA4/fibronectin binding site. Cell.

1990;60(4):577-584.

31. Osborn L, et al. Direct expression cloning of vascular cell adhesion molecule 1, a

cytokine-induced endothelial protein that binds to lymphocytes. Cell.

1989;59(6):1203-1211.

32. Newman G, and Pudney D. A survey of current practice in the management of

recurrent malignant ascites among oncologists and palliative-care physicians in the

UK. Clin Oncol-Uk. 2006;18(2):154.

33. Becker G, Galandi D, and Blum HE. Malignant ascites: Systematic review and

guideline for treatment. Eur J Cancer. 2006;42(5):589-597.

34. Ostrowski MJ. An assessment of the long-term results of controlling the

reaccumulation of malignant effusions using intracavity bleomycin. Cancer.

1986;57(4):721–727.

34 35. Maiche AG. Management of peritoneal effusions with intracavitary mitoxantrone or

bleomycin. Anticancer Drugs. 1994;5(3):305–308.

36. Link KH, et al. Intraperitoneal chemotherapy with mitoxantrone in malignant

ascites. Surg. Oncol Clin N Am. 2003;12(3):865–872.

37. Schilsky RL, et al. Phase I clinical and pharmacologic study of intraperitoneal

cisplatin and fluorouracil in patients with advanced intraabdominal cancer. J. Clin.

Oncol. 1990;8(12):2054–2061.

38. Sennino B, and McDonald DM. Controlling escape from angiogenesis inhibitors.

Nat Rev Cancer. 2012;12(10):699-709.

35 Figure legend

Figure 1. M2 macrophages induce hypopermeability by downregulating VCAM1 in ECs

(A) KEGG pathway analysis of differentially expressed genes (DEGs) in M1 macrophage-treated HUVECs versus M2 macrophage-treated HUVECs. (B) Gene expression heatmap of differentially expressed cell adhesion molecule (CAM) genes from

(A). (C) The expression of VCAM1 in HUVECs detected by western blot in different cocultures (n=3). (D) Localization of VCAM1 protein (red) by immunofluorescence analysis. DAPI stains cell nucleus. Scale bar, 20 μm. (E) Phosphorylation of VE-cadherin protein (p-VE-cad) expression in M1 macrophage-cocultured HUVECs that were transiently transfected with VCAM1-specific shRNAs (shVCAM1-a, shVCAM1-b) or a control shRNA (shCtrl) (n=3). (F) Immunofluorescence analysis of p-VE-cad in M1 macrophage-cocultured HUVECs with VCAM1 knocked down. Scale bar, 20 μm. (G) p-VE-cad expression in M2 macrophage-cocultured HUVECs that were transiently transfected with a VCAM1-specific vector (Ove-VCAM1) or a control vector (n=3). (H)

Immunofluorescence analysis of p-VE-cad in M2 macrophage-cocultured HUVECs with

VCAM1 overexpressed. Scale bar, 20 μm. (I) TRITC-Dextran tracer fluorescence from

M1 macrophage-cocultured HUVECs that were transfected with VCAM1 knockdown shRNAs (shVCAM1-a, shVCAM1-b) or a control shRNA (shCtrl; n=5). (J)

TRITC-Dextran tracer fluorescence from M2 macrophage-cocultured HUVECs that were transfected with VCAM1-specific vector (Ove-VCAM1) or control vector (n=5). Data

36 represent three independent experiment. The results are shown as the mean ± SD, *p<

0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001, as assessed by one-way ANOVA

(C, E, I) and Student’s t-test (G, J).

Figure 2. Decreased VLA4 activation in M2 macrophages cocultured with ECs

(A) KEGG pathway analysis of differentially expressed genes (DEGs) when comparing

M2- versus M1-macrophages from cocultures. (B) Gene expression heat map of differentially expressed integrin genes from (A). (C) The expression of VLA4 in THP-1 macrophages detected by western blot in different cocultures (n=3). (D) Localization of

VLA4 protein (green) in THP-1 macrophages by immunofluorescence analysis. CD68

(red) stains macrophages. DAPI stains cell nucleus. Scale bar, 20 μm. (E) Representative graph of flow cytometric analysis of active VLA4 levels in different subtypes of THP-1 macrophages from cocultures. (F) p-VE-cad expression in HUVECs cultured with macrophages that were transiently transfected with VLA4-specific shRNAs (shVLA4-a, shVLA4-b) or a control shRNA (shCtrl) (n=3). (G) p-VE-cad expression in HUVECs from M1 macrophage-coculture system treated with CDP323 (n=3). (H) p-VE-cad expression in HUVECs cocultured with macrophages that were transiently transfected with a VLA4-specific vector (Ove-VLA4) or a control vector (n=3). (I) p-VE-cad expression in HUVECs from M2 macrophage-coculture system treated with THI0019

(n=3). (J) Immunofluorescence analysis of p-VE-cad expression in HUVECs cultured

37 with macrophages that were transiently transfected with a VLA4-specific vector

(Ove-VLA4) or a control vector. Scale bars, 20 μm. (K) TRITC-Dextran tracer fluorescence from coculture systems in which macrophages transiently transfected with a

VLA4-specific vector (Ove-VLA4) or pre-treated with THI0019, PS/2 and CDP323 compared with the respective control (n=5). Data represent three independent experiments. The results are shown as the mean ± SD, * p< 0.05, ** p < 0.01, *** p <

0.001, and **** p < 0.0001, as assessed by Student ’ s t-test (G-I, K) and one-way

ANOVA (C, F).

Figure 3. ROS functions downstream of VCAM1 in regulating macrophage-mediated permeability

(A-D) Representative graphs of flow cytometric analysis of ROS levels in HUVECs cocultured with different subtypes of macrophages (A), treated with VEGF (B) and in

HUVECs from M1 macrophage-coculture system treated with CDP323(C) or K-7174 (D).

(E) Immunoblot analysis of p-VE-cad expression in HUVECs from M1 macrophage-cocultured system treated with NAC (n=3). (F) p-VE-cad expression in

HUVECs from M2 macrophage-cocultured system treated with H2O2 (n=3). (G, H)

Immuno-fluorescence of p-VE-cad expression in HUVECs from M1 (G) or M2 (H) macrophage-cocultured system treated with NAC and H2O2. Scale bar, 20 μm. (I, J)

Effects of NAC or H2O2 treatment in HUVECs from in M1 macrophage-cocultured

38 system (I) or M2 macrophage-cocultured system (J) measured in real-time using the automated system (CellZscope®). Data represent three independent experiments. The results are shown as the mean ± SD, * p< 0.05, as assessed by Student’s t-test (E, F).

Figure 4. Macrophage-mediated permeability is dependent on RAC1 and phosphorylated PYK2

(A) Protein levels of RAC1 and phosphorylated PYK2 (p-PYK2) in HUVECs from different cocultures examined by western blot (n=3). (B) Immunofluorescence analysis of

RAC1 in HUVECs and HUVECs cultured with different subtypes of macrophages. Scale bar, 20 μm. (C) RAC1 and p-PYK2 protein expression in HUVECs from M1 macrophage-cocultured system treated with VLA4 inhibitor, CDP323 (n=3). (D) RAC1 and p-PYK2 expression in HUVECs from M1 macrophage-cocultured system treated with VCAM1 inhibitor, K-7174 (n=3). (E) Flow cytometric analysis of ROS levels in

HUVECs from M1 macrophage-cocultured system treated with RAC1 inhibitor,

NSC23766 (n=3). (F, G) p-PYK2 expression in HUVECs from M1 macrophage-cocultured system treated with NAC (F) or and M2 macrophage-cocultured system treated with H2O2 (G) (n=3). Results represent three independent experiments.

The results are shown as the mean ± SD, * p< 0.05, ** p < 0.01, *** p < 0.001, and

**** p < 0.0001, as assessed by Student’s t-test (C-G) and one-way ANOVA (A).

39 Figure 5. Targeting the VLA4/VCAM1 pathway reduces permeability and lessens ascites in vivo

(A) Schematic experimental design for the permeability assay in vivo. Clod indicates clodronate liposomes; treated cell indicates RAW264.7 cells with VLA4 knockdown or overexpression. (B, C) Evaluation of peritoneal permeability 48 hours after injection of

VLA4 knockdown (B) or overexpression (C) cells in the peritoneum of mice (n=5). (D)

Evaluation of peritoneal permeability 48 hours after injection of PS/2-treated antibody in the peritoneum from mice treated with or without clodronate liposomes (n=8 for PBS lipo group and PBS lipo+PS/2 group, n=7 for Clod Lipo, n=9 for Clod lipo+PS/2). (E)

Schematic experimental design for treating ascites in ovarian cancer animal. (F)

Representative pictures of adult mice with ovarian cancer sacrificed 10 days after PS/2 or

IgG2b treatment and the ascites each mouse produced. (G) Ascites volume from PS/2- or control antibody-treated ovarian cancer mice 12 days after HM-1 inoculation (n=10). (H)

ROS expression of endothelial cells in tissue after treatment with PS/2 in mice with ovarian cancer (OC) (n=5). (I-M) Expression of RAC1 (I-J), p-PYK2 and p-VE-cad

(K-M) in murine OC tumors detected by immunofluorescence (n=6). Scale bar, 50 μm.

Results represent three independent experiments. The results are shown as the mean ±

SD, * p< 0.05, ** p < 0.01, and *** p < 0.001, NS, not significant, as assessed by one-way ANOVA (B, D) and Student’s t-test (C, G, H, J, L, M).

40 Figure 6. VLA4/VCAM1 expression is correlated with ascites volume

(A) Kaplan-Meier Plotter analysis showed that high VLA4 expression is correlated with shorter overall survival. (B, C) Analysis of VCAM1 expression in ovarian cancer tissues from patients with few ascites (less than 500 ml) or massive ascites (more than 1000 mL) based on immunohistochemistry results (n=10). Scale bar, as indicated. (D)

Representative immunofluorescence images of VLA4 expression in human OC tumors from patients with few ascites or massive ascites (n=10). Green, VLA4; red, CD68; blue, nucleus. Scale bar, 50 μm. (E) Correlation between ascites volume and VLA4 expression in macrophages of OC tissue. Few ascites (less than 500 ml); moderate ascites (500 ml-1000ml); massive ascites (more than 1000 mL). (F) Percent of VLA4-expressing macrophages in different ovarian cancer ascites (n=11). (G) Ratio of M2- to M1- macrophages in ovarian cancer ascites (n=9 for few ascites group, n=10 for massive ascites group). (H) Permeability analysis using macrophages isolated from OC patient ascites detected by TRITC-Dextran assay (n=10). (I) Permeability assay using macrophages isolated from OC patient ascites detected by the electric cell impedance sensing (ECIS) equipment. (J) Permeability assay of macrophages isolated from OC patient ascites cocultured with HUVECs with PS/2, NAC or control treatment detected by ECIS. Results represent three independent experiments. The results are shown as the mean ± SD, * p< 0.05, ** p < 0.01, and *** p < 0.001, as assessed by one-way

ANOVA (E) and Student’s t-test (C, F-H).

41 Figure 7. VLA4/VCAM1 axis mediates permeability separately from the VEGF pathway

(A) TRITC-Dextran permeability in HUVECs cocultured with M1 or M2 macrophages before or after treatment with bevacizumab (n=3). (B) p-VE-cadherin expression in

HUVECs from M1 macrophage-cocultured system or M2 macrophage-cocultured system after treated with bevacizumab (n=3). (C) Permeability assay using M1 macrophage-coculture treated with PS/2, bevacizumab or combination of PS/2 and bevacizumab detected by the automated system (CellZscope®). (D, E) Ascites volume detected by B-mode ultrasonography after intraperitoneal injection in mice with PS/2, bevacizumab or both (n=5). (F) Kaplan – Meier survival curves of mice after intraperitoneal injection with PS/2, bevacizumab or both (n=9). (G, H) Fold change of gene expression in M1 macrophages (G) or M2 macrophages (H) treated with PS/2 compared with M1 or M2 macrophages alone (n=3). Data represent three independent experiments. The results are shown as the mean ± SD, * p< 0.05, ** p < 0.01, *** p <

0.001 and **** p < 0.0001and NS, not significant, as assessed by one-way ANOVA (A, B,

E) and Student’s t-test (G, H). The Kaplan-Meier analysis log-rank test was used to estimate the event-free survival curve between the groups.

42 Figure 1 A B EC (M1) VS EC (M2) Qvalue EC (M1) EC (M2) group 0.00000 10 Cytokine-cytokine receptor interaction MCAM 0.00002 Chemokine signaling pathway ICAM2 8 0.00004 TNFsignaling pathway PECAM1 6 0.00006 Toll-like receptor signaling pathway ICAM1 0.00008 4 SCAMP2 IL-17 signaling pathway 0.00010 2 Hematopoietic cell lineage ALCAM 0.00012 SCAMP3 0 NF-kappa B signaling pathway VCAM1 Transcriptional misregulation in cancer Gene Number NRCAM NOD-like receptor signaling pathway 9 CEACAM1 Cell adherin molecules (CAMs) 21 TICAM1 Cytosolic DNA-sensing pathway 32 CAMK2D African trypannosomiasis CAMKK2 44 Influenza A CAMSAP2 55 0 0.05 0.1 0.15 Rich Ratio C D VCAM1 DAPI Merge VCAM1

β-actin

* 2.0 ** **** 1.5

1.0

0.5

N 0.0 o E r m E C a C + l

i z E M e C E E d C C D+ irect contact 2 M +

t +

o 0 M M

1 2 β

c p-VE-cad (Y658) DAPI Merge

t G E i F

p-VE-cad

(Y658) (Y658)

E β-actin β-actin + C M + 1 M h s C 2.0 2.5 1

* t r

2.0 * l 1.5 * 1.5 1.0 1.0 0.5 0.5 N 0.0 N 0.0 o o r r + m m

+ + M a M a M l l i i 1 2 1

z C z

e t s r e

l V h d

h d e

t c t o o to C

s s r

β O h h β A - v V V - a

a e C C - M c I A A c V

t 22000 M t C

i M i n 1 1 1 *** n A - - M a b -

20000 *** 1 a + + + M M M 1 1 18000 2 +

i 16000 n t 1 e s

n 14000 h s m T V i t R y e

12000 C I a T n A C 10000 M - D H 1 e p-VE-cad (Y658) DAPI Merge x

+ - t

M b r

a 1 n s h C t rl s s h h J V V 18000 C C A **A M M 1 1 -a -b 16000 + + V M M 1 1 e

c t 14000 o r

i + n t M e n 2

s 12000 T m i t R y e I a T n C

10000 - D e + x + t M M r

a 2 2 n O

e c to e

r O - v V e -V C C A A

M M 1

+ 1 M 2 Figure 1. M2 macrophages induce hypopermeability by downregulating VCAM1 in ECs. (A) KEGG pathway analysis of differentially expressed genes (DEGs) in M1 macrophage-treated HUVECs versus M2 macrophage-treated HUVECs. (B) Gene expression heatmap of differentially expressed cell adhesion molecule (CAM) genes from (A). (C) The expression of VCAM1 in HUVECs detected by western blot in different cocultures (n=3). (D) Localization of VCAM1 protein (red) by immunofluorescence analysis. DAPI stains cell nucleus. Scale bar, 20μm. (E) Phosphorylation of VE-cadherin protein (p-VE-cad) expression in M1 macrophage-cocultured HUVECs that were transiently transfected with VCAM1-specific shRNAs (shVCAM1-a, shVCAM1-b) or a control shRNA (shCtrl) (n=3). (F) Immunofluorescence analysis of p-VE-cad in M1 macrophage-cocultured HUVECs with VCAM1 knocked down. Scale bar, 20μm. (G) p- VE-cad expression in M2 macrophage-cocultured HUVECs that were transiently transfected with a VCAM1-specific vector (Ove-VCAM1) or a control vector (n=3). (H) Immunofluorescence analysis of p-VE-cad in M2 macrophage-cocultured HUVECs with VCAM1 overexpressed. Scale bar, 20μm. (I) TRITC-Dextran tracer fluorescence from M1 macrophage-cocultured HUVECs that were transfected with VCAM1 knockdown shRNAs (shVCAM1-a, shVCAM1-b) or a control shRNA (shCtrl; n=5). (J) TRITC-Dextran tracer fluorescence from M2 macrophage-cocultured HUVECs that were transfected with VCAM1-specific vector (Ove-VCAM1) or control vector (n=5). Data represent three independent experiment. The results are shown as the mean ± SD, *p< 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001, as assessed by one-way ANOVA (C, E, I) and Student’s t-test (G, J). Figure 2

A M2 (EC) VS M1 (EC) Qvalue B M2(EC) M1(EC) 0 Pathways in cancer 8 Rap1 signaling pathway 0.0005 ITGA5 Platelet activation ITGAX Cell cycle 0.001 6 Osteoclast differentiation ITGA4 Oxytocin signaling pathway 0.0015 ITGAL 5 DNA replication 0.002 ITGA2 Glycerophospholipid metabolism ITGA7 3 Leukocyte transendothelial migration 0.0025 ITGA10 AGE-RAGE signaling pathway in 2 diabetic 0.003 ITGA2B Axon guidance 0.0035 ITGAD 0 Protein digestion and absorption ITGA8 Cytokine-cytokine receptor Gene Number ITGA9-AS1 interaction PI3K-Akt signaling pathway 17 ITGA6-AS1 Apelin signaling pathway ITGAE Focal adhesion 46 ITGA9 Parathyroid hormone synthesis 75 ITGA11 secretion MAPK signaling pathway Fc gamma R-mediated phagocytosis 103 ITGA1 Glutamatergic synapse ITGA3 132 ITGAM 0 0.05 0.1 0.15 0.2 0.25 0.3 ITGAV Rich Ratio ITGA6

C D VLA4 CD68 DAPI Merge

VLA4

β-actin

* 20μm 2.0 * *** 1.5

1.0

0.5 E N 0.0 o C r m E + a C M l

i z E 2

e C E E

d C C

M + + t

o 0 M M

1 2

β F E G H I

1a 00 p-VE-cad p-VE-cad p-VE-cad p-VE-cad EC+M2 c

n EC+M1 (Y658) (Y658) (Y658) (Y658)

E β-actin β-actin β-actin β-actin C + M 50 1 1.5 * 1.5 * 1.5 * 1.0 * * C 0.8 o

u 1.0 n 1.0 1.0

t 0.6 s 0.4 0.5 0.5 0.5 0 2 3 4 5 6 10 10 10 10 10 0.2 Active VLA4 N N N 0.0 0.0 N 0.0 0.0 o o o o r r r r m m m m +

+ a a a E a E l l l l i C i i C i z z z z E e e e E s e C

d V

C d d h d e + +

t c M t t

M t s s o t o o t r h h o o O 2

l 1

r v V V β β β E L L β e E - C - - - - C A A a V a a J 4 4 + K a + c L c p-VE-cad (Y658) Dc API Merge M - - c A M t t t a b 1 t i i i 4 i + + n 2 n n + E E C n + + C C E T D C H P I0 3 0 2 30000 1 3 9 + E C ****

i n V

t 25000 * e e

n ** c s i t t o y ** r

20000 T R 15000 I T C - D

e +

O E x C t v r E a E E C e

n C C V e + - + + M

m E c V M M C toO 2

e 1 1 v L E + r e E a M - A C C

n V + 1 L + + 4 M P A M 4 + 1 S 2 + / + +

E C 2 E T D C H C P I0 3 0 2 1 3 9 Figure 2. Decreased VLA4 activation in M2 macrophages cocultured with ECs. (A) KEGG pathway analysis of differentially expressed genes (DEGs) when comparing M2- versus M1-macrophages from cocultures. (B) Gene expression heat map of differentially expressed integrin genes from (A). (C) The expression of VLA4 in THP-1 macrophages detected by western blot in different cocultures (n=3). (D) Localization of VLA4 protein (green) in THP-1 macrophages by immunofluorescence analysis. CD68 (red) stains macrophages. DAPI stains cell nucleus. Scale bar, 20 μm. (E) Representative graph of flow cytometric analysis of active VLA4 levels in different subtypes of THP-1 macrophages from cocultures. (F) p-VE-cad expression in HUVECs cultured with macrophages that were transiently transfected with VLA4-specific shRNAs (shVLA4-a, shVLA4-b) or a control shRNA (shCtrl) (n=3). (G) p-VE-cad expression in HUVECs from M1 macrophage-coculture system treated with CDP323 (n=3). (H) p-VE-cad expression in HUVECs cocultured with macrophages that were transiently transfected with a VLA4- specific vector (Ove-VLA4) or a control vector (n=3). (I) p-VE-cad expression in HUVECs from M2 macrophage-coculture system treated with THI0019 (n=3). (J) Immunofluorescence analysis of p-VE-cad expression in HUVECs cultured with macrophages that were transiently transfected with a VLA4-specific vector (Ove-VLA4) or a control vector. Scale bars, 20μm. (K) TRITC-Dextran tracer fluorescence from coculture systems in which macrophages transiently transfected with a VLA4-specific vector (Ove-VLA4) or pre-treated with THI0019, PS/2 and CDP323 compared with the respective control (n=5). Data represent three independent experiments. The results are shown as the mean ± SD, * p< 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001, as assessed by Student’s t-test (G-I, K) and one-way ANOVA (C, F).

Figure 3 A B C D Unstained control Unstained control Unstained control Unstained control EC+M1 EC+M1+VEGF EC+M1 EC+M1 EC+M2 EC+M2+VEGF EC+M1+CDP323 EC+M1+K-7174 600 150 400 200

400 100

200 100 200 50

C 0 0 0 0 o 2 3 4 5 6 2 3 4 5 6 2 3 4 5 6 2 3 4 5 6 u 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 n t s ROS G p-VE-cad (Y658) DAPI Merge

E F p-VE-cad p-VE-cad (Y658) (Y658) E

20μm

β-actin β-actin

1.5 2.0

* 1.5 *

1.0

1.0 + E 0.5 N C

A 0.5

N N

0.0 0.0 1

o o

r r O m

m 2

a a

l l

E i E z

C z C

e e

+ + d

M d M E 2

t C t o

o + + O

β β 2 1 + - -

+ H a a

c c

A +

t C t

i i n

n C

p-VE-cad (Y658) DAPI Merge I

EC+M1

EC+M1+NAC

EC+M1+H2O2

200

150

+ 100

2 E

i n

0 2 4 6 8 10 12 14

Time (hour)

J EC+M2

E +

N EC+M2+NAC C A + EC+M2+H2O2 M C 250

200

150

2 R

100

H i

E +

50 C

2 0 2 4 6 8 10 12 14

Time (hour) Figure 3. ROS functions downstream of VCAM1 in regulating macrophage-mediated permeability. (A-D) Representative graphs of flow cytometric analysis of ROS levels in HUVECs cocultured with different subtypes of macrophages (A), treated with VEGF (B) and in HUVECs from M1 macrophage-coculture system treated with CDP323(C) or K- 7174 (D). (E) Immunoblot analysis of p-VE-cad expression in HUVECs from M1 macrophage-cocultured system treated with NAC (n=3). (F) p-VE-cad expression in HUVECs from M2 macrophage-cocultured system treated with H2O2 (n=3). (G, H) Immuno-fluorescence of p-VE-cad expression in HUVECs from M1 (G) or M2 (H) macrophage-cocultured system treated with NAC and H2O2. Scale bar, 20μm. (I, J) Effects of NAC or H2O2 treatment in HUVECs from in M1 macrophage-cocultured system (I) or M2 macrophage-cocultured system (J) measured in real-time using the automated system (CellZscope®). Data represent three independent experiments. The results are shown as the mean ± SD, * p< 0.05, as assessed by Student’s t-test (E, F). Figure 4 A B EC EC+M0 EC+M1 EC+M2

RAC1

p-PYK2 (Y402) R

A β-actin C 1

3 * *** **** 2

1 D R A

A n

0 P o C r I 1 m

e a x l i p z e r

e 2.5 d * s

t s ** o i e 2.0 o

x β n p - a r 1.5 e c s t i s n 1.0 i o n 0.5 M o n r m 0.0 e r a p g l - i P z e

e Y E d C K

t 2 o

E E E

(

β C C C Y

- + + + 4 a M M M 0 C c 0 1 2 D E 2 t EC+M1 EC+M1

i n

) 800000 EC+M1+CDP323 EC+M1+K-7174 **

1.0 0.8 RAC1 * RAC1 * * 600000 0.8 ** 0.6 p-PYK2 0.6 p-PYK2 400000 (Y402) (Y402) 0.4 0.4 200000

0.2 0.2 E

β-actin β-actin C

R N N 0.0 0.0 0 O o o r r S m m

i a a R R n l l t i i e

E z z A E A E

p n

C p e e C C C C - -

+ s + d d 1 P 1 P + M ( (Y Y M i M

Y t t t 4 1 4 1 K y

K o 1 o 20 20

2 2 β E β ) E ) - C - C

1.0 a + a + 1.5

* c M c M E t t C i 1 i 1 n

F n + 0.8 G + + C * M K 1 D p-PYK2 -7 1.0 + p-PYK2 P 0.6 1 N 3 7 2 (Y402) 4 S (Y402) 3 C 0.4 2 3 0.5 7 6 0.2 6 β-actin β-actin N N 0.0 0.0 o o 2 2 r

r O O m

m 2 2 a a l l i i z E z E E E e C e C C C d + d + + +

M M ME EM E E t t o 1C o C1 2C C2

+ + + + β M β M M M - - 2 2 1 a

1 a + + + + c N c N H H t t i

A i A n C n C Figure 4. Macrophage-mediated permeability is dependent on RAC1 and phosphorylated PYK2. (A) Protein levels of RAC1 and phosphorylated PYK2 (p-PYK2) in HUVECs from different cocultures examined by western blot (n=3). (B) Immunofluorescence analysis of RAC1 in HUVECs and HUVECs cultured with different subtypes of macrophages. Scale bar, 20μm. (C) RAC1 and p-PYK2 protein expression in HUVECs from M1 macrophage-cocultured system treated with VLA4 inhibitor, CDP323 (n=3). (D) RAC1 and p-PYK2 expression in HUVECs from M1 macrophage-cocultured system treated with VCAM1 inhibitor, K-7174 (n=3). (E) Flow cytometric analysis of ROS levels in HUVECs from M1 macrophage-cocultured system treated with RAC1 inhibitor, NSC23766 (n=3). (F, G) p-PYK2 expression in HUVECs from M1 macrophage-cocultured system treated with NAC (F) or and M2 macrophage-cocultured system treated with H2O2 (G) (n=3). Results represent three independent experiments. The results are shown as the mean ± SD, * p< 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001, as assessed by Student’s t-test (C-G) and one-way ANOVA (A). Figure 5 A B C D

400 ** 800 250 **** ** *** NS Clod Clod Treated Data 200

i 300 i n 600 i ** n n

cell analysis t t e t e e n n n

s 150 s s i i t i t y t

400 y

y 200 I.P. 100 Days 1 3 6 8 100 200 50

0 T T T 0 0

R R R I I I T T T l C ip li C p C C o tr o s - - - l V D D D h P e C

c e B e e s s h t O S lo h o x x x V V r v d t t t e L r r F r L E - a a a Control A A PS/2 G V H P C

n B n n 4 4 l - L S o

a b A d m m m 4 l l ip ip e e e o o a a a + +

n P n n P S

HM-1 cells inoculation /2 /2 in abdominal cavity 40000 (1*106 HM-1 cells/mice) 8 *** * 2 days 6 30000

PS/2 Ab or control 4 20000 Ab every other day

2 10000

c A 10 days + e s l c l

0 i R 0 t

Data analysis e O C s S

D v P

P o 3 S m S l C / 1 C u 2 / o e o 2 m n a n t t n I e ro J ro

l l i

( n

RAC1 CD31 DAPI Mm erge t e l ) n s i t y

200 *

150

100

c + e l l

R P A

C 0 C S D 1 3 /

2 1 m P

S e C /

a o 2

n n t

r i

o n l t

s K

i t

p-VE-cad (Y658) p-PYK2 (Y402) CD31 DAPI Merge

C o n t r o l P S / 2

L M 250

300

* C

o * 200 n t r

o 200

l 150

100 100 m

c c + m + e e e

50 e a l l l l a

n p p n

- - i

n V C P C i n t E D 0 Y

D 0 e t e - n K 3 3 c n s 1 2 1 a s i

t d

P i (

y P t Y

S y S ( C C

Y / 4 / o 2 o 2

0 6 n n 2 5 tr tr

o ) o 8 l l )

Figure 5. Targeting the VLA4/VCAM1 pathway reduces permeability and lessens ascites in vivo. (A) Schematic experimental design for the permeability assay in vivo. Clod indicates clodronate liposomes; treated cell indicates RAW264.7 cells with VLA4 knockdown or overexpression. (B, C) Evaluation of peritoneal permeability 48 hours after injection of VLA4 knockdown (B) or overexpression (C) cells in the peritoneum of mice (n=5). (D) Evaluation of peritoneal permeability 48 hours after injection of PS/2-treated antibody in the peritoneum from mice treated with or without clodronate liposomes (n=8 for PBS lipo group and PBS lipo+PS/2 group, n=7 for Clod Lipo, n=9 for Clod lipo+PS/2). (E) Schematic experimental design for treating ascites in ovarian cancer animal. (F) Representative pictures of adult mice with ovarian cancer sacrificed 10 days after PS/2 or IgG2b treatment and the ascites each mouse produced. (G) Ascites volume from PS/2- or control antibody-treated ovarian cancer mice 12 days after HM-1 inoculation (n=10). (H) ROS expression of endothelial cells in tissue after treatment with PS/2 in mice with ovarian cancer (OC) (n=5). (I-M) Expression of RAC1 (I-J), p-PYK2 and p-VE-cad (K-M) in murine OC tumors detected by immunofluorescence (n=6). Scale bar, 50μm. Results represent three independent experiments. The results are shown as the mean ± SD, * p< 0.05, ** p < 0.01, and *** p < 0.001, NS, not significant, as assessed by one-way ANOVA (B, D) and Student’s t-test (C, G, H, J, L, M). Figure 6 A B C

1.0 80 HR=1.56（1.03-2.36） logrank P=0.035 ** 0.8 Expression 60

low (89) F e

0.6 high (42) w 40

a P s

r 0.4 c o

i 20 b t e a b s V i a

l C i 0.2

t s y 0 A c

（89） M i t 1

（42） e

0.0 H s as

i c 0 20 40 60 80 s ite t F

o s M ew Time (months) l o a M a g sc as y s it s e iv s S s e i D Ec v o e

VLA4 CD68 DAPI Merge r e

0.3 *** ** N

o 0.2

a * s M c e i a t

e n

I 0.1 n (

D t e r n b s i t i a a t y r

y s

0.0 c u D n i t 6 i e t 8 ) s + a

/ s

V c N it

L e M o s A as 4 a M M ci

+ t a s o e s

d s s s e i F G H r ve i a v 100 * 250 40000 te

e as

* c ** ite 200 s

n 30000 t e

150 n

50 s i t y 20000 100 M 2 / M 50 10000

0 1

r a

t 0 i V T o L

R 0 A

I T 4 C

m -

a D a as s as e c c ci i ci r F te F te x te o F e s e s t e s r p w w w M a h a M a M a a n a a s s s a

c s c s c s m g i s i s s te i te iv ite i e s v s v I e J e e s e s Few Massive CDP323 NAC a

c n e ascites ascites K-7174 Control l l s

( 1.1 1.05 %

) 1 1

0.9 0.95

0.8 0.9

0.7 0.85

0.6 0.8

N 0.5 0.75 N o r o

m 0.4 0.7 r m a l a i

z 0.65 l i e 18 20 24 26 28 30 32 34 36 z 24 26 28 30 32 34 36 38 d e

d r Time (hour) Time (hour) e

r s e i s s i t s a t n a e n c e Figure 6. VLA4/VCAM1 expression is correlated with ascites volume. (A) Kaplan- Meier Plotter analysis showed that high VLA4 expression is correlated with shorter overall survival. (B, C) Analysis of VCAM1 expression in ovarian cancer tissues from patients with few ascites (less than 500 ml) or massive ascites (more than 1000 mL) based on immunohistochemistry results (n=10). Scale bar, 200μm; 50μm, as indicated. (D) Representative immunofluorescent images of VLA4 expression in human OC tumors from patients with few ascites or massive ascites (n=10). Green, VLA4; red, CD68; blue, nucleus. Scale bar, 50μm. (E) Correlation between ascites volume and VLA4 expression in macrophages of OC tissue. Few ascites (less than 500 ml); moderate ascites (500 ml- 1000ml); massive ascites (more than 1000 mL). (F) Percent of VLA4-expressing macrophages in different ovarian cancer ascites (n=11). (G) Ratio of M2- to M1- macrophages in ovarian cancer ascites (n=9 for few ascites group, n=10 for massive ascites group). (H) Permeability analysis using macrophages isolated from OC patient ascites detected by TRITC-Dextran assay (n=10). (I) Permeability assay using macrophages isolated from OC patient ascites detected by the electric cell impedance sensing (ECIS) equipment. (J) Permeability assay of macrophages isolated from OC patient ascites cocultured with HUVECs with PS/2, NAC or control treatment detected by ECIS. Results represent three independent experiments. The results are shown as the mean ± SD, * p< 0.05, ** p < 0.01, and *** p < 0.001, as assessed by one-way ANOVA (E) and Student’s t- test (C, F-H). Figure 7 A B 1.5 **** **** 0.9 ** 1.0 * 0.8 **** p-VE-cad NS 0.5 (Y658) 0.7

0.0 0.6 β-actin N

r 0.5 e o l r a m t T i a v R l e i z

I E E E E E E T c C C C C e C C h C + + + e + d + +

a e M e Me

e e - M M v v M M v t v D n v 1 a 1 v 1 2 a a c 2 a o a 2 a g c c

c e c i c

i z i β i e z i z z i x u z u z u - u

/ u u

t m

m a c m m r m a m a c o a a E E b a E Eb a E E a t n b n C b C b b C C i C C n

C t + + + m r M + D Control M + E +

o PS/2 M M 1 M M e

l 1 + 2 + 2 1 2

a + b + b + b +

n b b b

EC+M1 i

n ****

t EC+M1+PS/2 e 30 **** n EC+M1+bevacizumab s *** i t

y EC+M1+PS/2+bevacizumab 200 * 20 150

bevacizumab PS/2+bevacizumab 10 100 T A E b R d

o i n 50 m

% 0 e n

i C s T P t

0 a L S P 0 2 4 6 8 10 12 14 16 / n 2 S / Time (hour) c 2 e + b b

( e e

m v v a a

m c c iz iz

) u u m m a a b b F G H CTL TNFA TNFA PS/2 * *** * CCL2 CCL2 *** bevacizumab CD80 CD80 100 PS/2+bevacizumab * ** * M1 CD86 **** M1 CD86 * IL1B IL1B IL6 *** IL6 **** NOS2 NOS2 *** IL12B * 50 *** IL12B ** VEGF VEGF log-rank p = 0.0015 IL10 * IL10 **** ARG1 ARG1 P M2 **** M2 *** e CD36 CD36

r **** c

e CD206 CD206

n 0 * t

s 0 20 40 60 -1.0 -0.5 0.0 0.5 1.0 1.5 -2 0 2 4 6 8 u

r Time (day) M2+PS/2 v M1+PS/2 i v a l

( %

F) igure 7. VLA4/VCAM1 axis mediates permeability separately from the VEGF pathway. (A) TRITC-Dextran permeability in HUVECs cocultured with M1 or M2 macrophages before or after treatment with bevacizumab (n=3). (B) p-VE-cadherin expression in HUVECs from M1 macrophage-cocultured system or M2 macrophage- cocultured system after treated with bevacizumab (n=3). (C) Permeability assay using M1 macrophage-coculture treated with PS/2, bevacizumab or combination of PS/2 and bevacizumab detected by the automated system (CellZscope®). (D, E) Ascites volume detected by B-mode ultrasonography after intraperitoneal injection in mice with PS/2, bevacizumab or both (n=5). (F) Kaplan–Meier survival curves of mice after intraperitoneal injection with PS/2, bevacizumab or both (n=9). (G, H) Fold change of gene expression in M1 macrophages (G) or M2 macrophages (H) treated with PS/2 compared with M1 or M2 macrophages alone (n=3). Data represent three independent experiments. The results are shown as the mean ± SD, * p< 0.05, ** p < 0.01, *** p < 0.001 and **** p < 0.0001and NS, not significant, as assessed by one-way ANOVA (A, B, E) and Student’s t-test (G, H). The Kaplan-Meier analysis log-rank test was used to estimate the event-free survival curve between the groups.