Role of prostaglandin D2 receptor DP as a suppressorof tumor hyperpermeability and angiogenesis in vivoTakahisa Murataa,1, Michelle I. Linb, Kosuke Aritakec, Shigeko Matsumotoa, Shu Narumiyad, Hiroshi Ozakia,Yoshihiro Uradec, Masatoshi Horia, and William C. Sessab

aDepartment of Veterinary Pharmacology, Graduate School of Agriculture and Life Sciences, University of Tokyo, Tokyo 133-8657, Japan; bDepartment ofPharmacology and Vascular Biology and Therapeutics Program, Yale University School of Medicine, New Haven, CT 06511; cDepartment of MolecularBehavioral Biology, Osaka Bioscience Institute, Osaka 565-0874, Japan; and dDepartment of Pharmacology, Faculty of Medicine, Kyoto University, Kyoto606-8315, Japan

Edited by Sergio Henrique Ferreira, University of Sao Paulo, Ribeirao Preto, Brazil, and approved October 21, 2008 (received for review May 30, 2008)

Although COX-dependent production of prostaglandins (PGs) isknown to be crucial for tumor angiogenesis and growth, the roleof PGD2 remains virtually unknown. Here we show that PGD2

receptor (DP) deficiency enhances tumor progression accompaniedby abnormal vascular expansion. In tumors, angiogenic endothelialcells highly express DP receptor, and its deficiency acceleratesvascular leakage and angiogenesis. Administration of a syntheticDP agonist, BW245C, markedly suppresses tumor growth as well astumor hyperpermeability in WT mice, but not in DP-deficient mice.In a corneal angiogenesis assay and a modified Miles assay, host DPdeficiency potentiates angiogenesis and vascular hyperpermeabil-ity under COX-2-active situation, whereas exogenous administra-tion of BW245C strongly inhibits both angiogenic properties in WTmice. In an in vitro assay, BW245C does not affect endothelialmigration and tube formation, processes that are necessary forangiogenesis; however, it strongly improves endothelial barrierfunction via an increase in intracellular cAMP production. Ourresults identify PGD2/DP receptor as a new regulator of tumorvascular permeability, indicating DP agonism may be exploited asa potential therapy for the treatment of cancer.

tumorigenesis � vascular permeability � prostaglandin

Angiogenesis, the growth of new blood vessels from preex-isting vasculature, is crucial for both tumor progression and

metastasis (1). VEGF is a well-recognized angiogenic cytokinethat promotes tumor angiogenesis and growth. VEGF alsopotently induces vascular leakage (2, 3). The resulting extrava-sation of plasma proteins through the microvasculature providesa provisional matrix to sequester growth factors and supportendothelial and tumor cell growth.

Numerous epidemiological, laboratory animal, and clinicalstudies have provided evidence that non-steroidal anti-inflammatory drugs that inhibit COX and prostaglandin (PG)synthesis can significantly reduce the risk of cancer development(4, 5). Two isoforms of COX have been characterized, COX-1and COX-2. COX-1 is expressed constitutively in various tissues,whereas COX-2 is inducible by mitogens, cytokines, and tumorpromoters. Selective COX-2 inhibitors and COX-2 gene disrup-tion in mice suppress tumor progression (6, 7), suggesting PGspromote aspects of tumor growth. The main PGs responsible fortumor progression are being explored and PGE2 has beendocumented to promote tumor growth and angiogenesis (8–10).

PGD2 is another COX metabolite produced by activated mastcells, macrophages, and Th2 cells. Its biological actions aremediated through the G protein-coupled receptor, named DP(11, 12). Although there is evidence that PGD2-DP signalingoccurs in various inflammatory diseases including asthma (13),the role of PGD2 in tumor growth remains unknown. Becausevarious blood and immune cells possess the capacity to generatePGD2, we hypothesized that PGD2 produced by infiltrating cellsmodulates aspects of tumorigenesis, thus affecting tumor pro-gression. In the present study, we investigated whether DP gene

disruption and DP agonism influence tumor angiogenesis andprogression in mice.

ResultsDP Receptor Signal Exerts a Suppressive Effect on Tumor Growth.Initially, we examined tumor growth in PGD2 receptor DP-deficient mice (DP�/�). As shown in Fig. 1A, Lewis lungcarcinoma (LLC) cells implanted onto the backs of DP�/� micegrew faster than those implanted onto WT mice, and thusdecreased their survival rate (Fig. 1B). Although similar con-centrations of PGD2 were detected in these tumors (WT,1072.4 � 54.4 pg/g; DP�/�, 1,200.7 � 80.4 pg/g on day 14, n �4–5), their levels were much higher than those in normal intacttissues (lung, 98.1 � 9.3 pg/g; liver, 102.1 � 13.1 pg/g, n � 4 each;P � 0.05 with tumor on WT mice), implying a substantial roleof PGD2 in tumor growth. Next, we administered either a DPreceptor blocker, BWA868C (3 mg/kg i.p., twice a day), or a DPagonist, BW245C (50 �g/kg i.p., twice a day), into WT mice andour findings were consistent with the idea that DP blockageaccelerates tumor growth (Fig. 1C) and shortens lifespan (Fig.1D). Importantly, the anti-tumor effects of BW245C were ab-rogated by host DP deficiency (final tumor volume on day 14:DP�/�, 1,608.6 � 106.8 mm3; DP�/��BW245C, 1,581.6 � 125.3mm3, n � 5 each), showing that the effects of BW245C weremediated via the DP receptor. During the treatment periods,vehicle, BWA868C, or BW245C did not influence normal bodilyfunctions and behavior including weight gain, appetite, orgrooming behavior. Based on the data showing that the im-planted LLC cells grow faster in DP�/� mice and that BW245Cexhibits no detectable anti-proliferative effects on LLC [sup-porting information (SI) Fig. S1 A], we hypothesized that theeffect of DP activity on tumor growth may be dependent on theresponse of host cells that are deficient in DP receptor, e.g.,immune and vascular cells, and not directly on tumor cells.

DP Receptor Signal Modulates Environment for Tumor Growth. Tofurther analyze the effect of DP deficiency on tumor growth, weexamined end-point tumors (day 14 after implantation) histo-logically. H&E staining of tumor cross sections from WT micerevealed the commonly observed pockets of necrosis spreadthroughout the tumor mass (arrowheads in Fig. 2A) in contrastto those seen in DP�/� mice, which had smaller necrotic areas

Author contributions: T.M. and W.C.S. designed research; T.M., M.I.L., K.A., and S.M.performed research; T.M., S.N., H.O., Y.U., and M.H. contributed new reagents/analytictools; T.M. and K.A. analyzed data; and T.M. and W.C.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

1To whom correspondence should be addressed. E-mail: murata@mail.vm.a.u-tokyo.ac.jp.

This article contains supporting information online at www.pnas.org/cgi/content/full/0805171105/DCSupplemental.

© 2008 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0805171105 PNAS � December 16, 2008 � vol. 105 � no. 50 � 20009–20014

PHA

RMA

COLO

GY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

(summarized in Fig. 2B). Also, tumor sections from BW245C-treated mice had larger areas of necrosis (Fig. 2A Right, Inset)throughout the tumor mass than those from vehicle-treatedmice. As tumor growth is the net result of tumor cell prolifer-ation and apoptosis, we next determined tumor cell proliferationby Ki-67 immunohistochemical staining on tumor tissue sections.As shown in Fig. 2C, neither DP deficiency nor treatment withBW245C affected the number of proliferating Ki-67-positivecells (summarized in Fig. 2D). In contrast, by using the TUNELcell death detection assay, we found that tumors grown in aDP-deficient environment had decreased TUNEL-positive apo-ptotic cells (representative pictures are in Fig. 2E and summa-rized in Fig. 2F), whereas activation of DP receptor via theBW245C agonist significantly increased the number of apoptoticcells. These studies suggest that host DP deficiency leads to afavorable environment for tumor growth by limiting tumor celldeath, whereas stimulation of the DP receptor, either on host orcancer cells themselves, can lead to increased cell death, thusexplaining decreased tumor growth.

DP Protein Is Localized in Tumor Endothelial Cells. By using doubleimmunofluorescent staining, we found that the DP receptor(green fluorescence) is highly expressed in tumor endothelialcells (labeled by platelet endothelial cell adhesion molecule-1[PECAM-1], red fluorescence) on WT mice but not DP�/� mice(Fig. 3A). Whereas an equivalent amount of mRNA expressionof DP receptor was detected regardless of tumor growth stage(days 6–14; Fig. S1B), their expression levels were higher thanthose in normal intact tissues (i.e., lung and liver). These datasuggest continuous involvement of DP receptor in tumor growth.Furthermore, abundant DP mRNA expression was detected inthe endothelium-intact—but not in endothelium-denuded—vascular tissue (e.g., mouse aorta, pulmonary artery, and carotidartery), and in isolated endothelial cell lines (Fig. S1 C and D,respectively). There is another G protein-coupled receptor,CRTH2, which mediates PGD2-indcued bioactivities. As shownin Fig. S1E, mRNA expressions of CRTH2 were detected invascular cells and growing tumor.

DP Receptor Signal Modulates Tumor Vascular Leakage and Tumori-genesis. We examined the effect of DP deficiency or receptoractivation on tumor endothelial cells. Similar to what has often

been observed in solid tumors (1, 14), the LLC implanted in WTmice exhibits substantial neovascularization (PECAM-1-positivestaining; Fig. 3B, Upper) and vascular leakage (fibrinogen dep-osition; Fig. 3C, Middle). Tumor sections from DP�/� micerevealed increased tumor angiogenesis, marked by increasedPECAM-1-positive blood vessels per unit area (Fig. 3C). Thiscorrelated positively with its increase in tumor growth, whereastreatment with the DP agonist BW245C in WT mice had theopposite effect. In addition, tumor sections from DP�/� miceshowed increased vascular leakage compared with WT mice,measured as the ratio of fibrinogen to PECAM-1 immunopos-itivity (1.44 fold increase from WT; Fig. 3D). In comparison withthe angiogenic vasculature in WT mice, the tumor vasculature ofDP�/� mice exhibited reduced tubule length and more branchpoints but less smooth muscle coverage (representative picturesin Fig. S1F; summarized in Fig. S1G), indicative of a less maturevasculature (1, 15). Thus, our data support the concept thatendogenous PGD2-DP signaling may reduce vascular leakage,angiogenesis, and tumor growth.

To examine if DP signal can directly and acutely influencetumor vascular leakage, extravasation of Evans blue dye wasmonitored as a measure of albumin leakage. We infused Evansblue intravenously into mice bearing equal-sized tumors and theamount of dye extravasation was quantified spectrophotometri-

Fig. 1. (A and B) Tumors implanted into DP�/� mice have increased tumorgrowth rates and decreased survival rates. LLCs were injected into WT or DP�/�

mice, and tumor growth and survival were monitored (n � 25 each, *, P � 0.05vs. WT). (C and D) DP agonism (BW245C) suppressed tumor growth andextended life span. Four days after LLC implantation (quoted as day 0), micewere injected i.p. with saline solution, 50 �g/kg of BW245C, or 3 mg/kgBWA868C twice a day (n � 10, *, P � 0.05 vs. vehicle treatment).

Fig. 2. DP deficiency decreased and BW245C increased necrotic/apoptoticregions in implanted tumors without affecting tumor proliferation. Tumorsfrom WT, DP�/�, or BW245C-treated WT mice were excised on day 14 andprocessed for immunostaining. Representative microscopic images of LLCtumors are shown here on day 14. (A) H&E staining, (Scale bar, 200 �m.)(C) Ki67 immunostaining. (Scale bar, 10 �m.) (E) TUNEL labeling. (Scale bar, 50�m.) Ki67 immunostaining (brown) was counterstained with hematoxylin(blue). TUNEL was counterstained with DAPI (blue) for nuclear labeling.Pictures are randomly taken from four fields each at a magnification of �100(H&E) or �200 (Ki67 and TUNEL) from eight independent sections. Necroticareas in the tumors were quantified relative to total pixel density (B). Thenumber of Ki67-positive or TUNEL-positive cells were normalized to a totalnuclei number (D and F, respectively; n � 6 each).

20010 � www.pnas.org�cgi�doi�10.1073�pnas.0805171105 Murata et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

cally. Fig. 3E demonstrates that tumors implanted in DP�/� micehad increased vascular permeability compared with those in WTmice. This observation is in line with the increased fibrinogendeposition found in end-point tumors implanted into DP�/�

mice, supporting the idea that DP receptor is required forconstitutive maintenance of tumor vascular barrier function. Intumors implanted onto WT mice, administration of the DPagonist BW245C significantly reduced Evans blue extravasationinto tumors compared with vehicle-treated mice.

The Role of DP Receptor on VEGF- or IL-1�-Induced Corneal Angio-genesis. Tumor angiogenesis is a consequence of complex inter-actions effected by a set of growth factors and cytokines. IL-1�is a principal inducer of inflammation mediated by COX-2,multiple prostanoids, chemokines, and angiogenic mediatorsincluding VEGF (16). Previous studies reported that IL-1�promotes COX-2-dependent angiogenesis and carcinomagrowth (17, 18). Other groups separately provided evidence thatCOX-2 activation was required for VEGF production and tumorangiogenesis (7, 8). Thus, we examined the role of the PGD2-DPsignaling pathway on IL-1�- and VEGF-induced corneal angio-genesis. As shown Fig. 4A, hydron pellets containing VEGF (100ng) or IL-1� (30 ng) were implanted into corneas. Cornealneovasculature extended from the limbus toward the pellets inWT and DP�/� animals (Fig. 4B), but responses to both cyto-kines were augmented in corneas of DP�/� mice. Notably,whereas DP�/� corneas exhibited an �1.3-fold increase inVEGF-induced angiogenesis compared with WT, IL-1� stimu-lation resulted in a more than twofold increase in the DP�/� mice(Fig. 4B). Given that the COX-2 inhibitor CAY10404 (5 �l of0.37 ng/ml twice a day) prevented IL-1�-induced neovascular-ization in WT and DP�/� mice, endogenous PGD2-DP signalingappears to promote angiogenesis when COX-2 is induced. Asshown in Fig. 4 C and D, exogenous administration of the DPagonist BW245C (5 �l of 0.37 ng/ml twice a day) greatlysuppressed both VEGF- and IL-1�-induced angiogenesis. Theinhibitory actions by DP agonism on angiogenesis may explainwhy the DP agonist BW245C retarded tumor growth.

The Role DP Receptor on VEGF- or IL-1�-Mediated Vascular Perme-ability In Vivo. To further define the potential role of theendogenous DP receptor in endothelial barrier protection,

VEGF- and IL-1�-mediated Evans blue dye extravasation wasassessed in a modified Miles assay. As seen in Fig. 5A andsummarized in Fig. 5B, the amount of dye extravasation causedby local administration with VEGF (30 ng, 5 min before dyeinjection) was identical in WT and DP�/� mice, whereas theIL-1�-mediated (10 ng, 1 h before dye injection) dye extrava-sation was more abundant in DP�/� mice compared with WTmice. This result was blocked by the COX-2 inhibitor CAY10404(100 �g/kg i.p. 30 min before IL-1� treatment). Exogenous DPstimulation by BW245C (50 �g/kg i.p. 10 min before VEGF orIL-1� treatment) significantly inhibited both VEGF- and IL-1�-induced Evans blue extravasation (Fig. 5C; summarized in Fig.

Fig. 3. (A) DP receptor localized in tumor-vascular endothelial cells. Sections were double-labeled for DP receptor (green) and vascular endothelium (PECAM-1,red), and counterstained with DAPI (blue; n � 5). (Scale bar, 20 �m.) (B–D) DP deficiency accelerated, and daily DP agonism suppressed fibrinogen depositionand angiogenesis in tumors. Excised tumors were subjected to double immunostaining for PECAM-1 (green) and fibrinogen (red), and then counterstained withDAPI (blue). (Scale bar, 50 �m in B.) To quantify the angiogenic areas in the tumors, PECAM-1 staining was quantified relative to total pixel density (C). Fibrinogendeposition was normalized to total PECAM-1-positive pixel density (D; n � 6 each). (E) DP deficiency and DP agonism on tumor permeability. Mice bearing LLCtumors (on day 10 after the LLC implantation) were treated with vehicle (10 min), BW245C (50 �g/kg; i.v., 10 min). Tumors were excised and the Evans blue dyecontent was quantified (n � 6–8; *, P � 0.05 vs. WT and †, P � 0.05 vs. vehicle treatment).

Fig. 4. DP receptor modulates VEGF- and IL-1�-induced angiogenesis. Hy-dron pellets containing VEGF (100 ng) or IL-1� (30 ng) were implanted into thecorneas of C57B/6 (WT or DP�/�) or BalbC mice. Five microliters of 0.37 ng/mlBW245C or CAY10404 (CAY) was dropped onto the mice corneas twice daily.On day 6, corneal neovasculature was photographed (A and C) and areas werequantified (B and D; n � 9–12; *, P � 0.05 vs. WT or vehicle and †, P � 0.05 withIL-1�-treated mice corneas).

Murata et al. PNAS � December 16, 2008 � vol. 105 � no. 50 � 20011

PHA

RMA

COLO

GY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

5D). As DP�/� mice exhibited greater hyperpermeability whentreated with IL-1� (Fig. 5B), DP antagonism (BWA868C, 3mg/kg, 30 min before IL-1� injection) also enhanced IL-1�-induced vascular leakage while, at the same time, it was inef-fective in augmenting VEGF-mediated vascular permeability.These results suggest that endogenous PGD2-DP signaling me-diates inflammation-linked vascular leakage.

In vivo, the tumor vascular barrier function is determined byseveral factors, including systemic and microvascular hemody-namics and permeability. Intraperitoneal administration ofBW245C only slightly decreased systemic blood pressure whenadministered in a 50 �g/kg bolus (77.8 � 6.2 mmHg 5 min afteradministration vs. 83.6 � 4.1 mmHg in vehicle-treated mice; n �4 per group).

DP Receptor Agonism Strongly Suppresses Endothelial Permeability,but Does Not Influence Endothelial Proliferation, Migration, or TubeFormation. To see if DP agonism directly affected endothelial celljunctions, we performed an in vitro permeability assay. Conflu-ent monolayers of bovine aortic endothelial cell (BAEC) platedinto trans-well plates had minimal FITC-dextran flux across themonolayer under non-stimulated condition, whereas adminis-tration of VEGF greatly increased FITC-dextran flux into thelower chamber in a dose-dependent manner (Fig. 6A). Pretreat-ment with both BW245C (0.1–1 �M, 5 min) and PGD2 (0.1–1�M, 5 min) minimized VEGF (10 ng/ml)-mediated changes inbarrier function. The effect of BW245C was reversed by the DPantagonist BWA868C (10 �M, 30 min before BW245C), as wellas by DP siRNA. Previous reports have shown that DP receptoractivation leads to Gs-mediated increases in intracellular cAMP(11). Elevation in intracellular cAMP tightens endothelial junc-tions via cAMP-dependent protein kinase (PKA)-dependentpathway (19, 20) and/or an independent pathway includingEpac-dependent signaling (21). As seen in Fig. 6A, exogenousadministration of a cell-permeable cAMP analogue, db-cAMP(1–10 �M), also counteracted VEGF-induced dextran leakage.Conversely, a potent PKA inhibitor, KT5720 (10 �M, 30 min),

only partially attenuated BW245C-induced barrier protection.These results suggest cAMP is indispensable for barrier protec-tion through the mechanism of DP agonism, whereas PKA mightonly be a partial downstream mediator of cAMP.

One possible mechanism which BW245C treatment couldattenuate tumor growth is by direct anti-angiogenic effects (i.e.,by blocking endothelial cell proliferation, migration, and/ororganization). This could lead to the observed decrease invascular permeability (i.e., decreased overall accumulation ofEvans blue dye). To directly test this, we performed in vitroassays to monitor the effects of BW245C on BAECs. VEGF (50ng/ml) or serum (2%) triggered robust endothelial cell migration(assessed by a modified Boyden chamber migration assay; seeFig. S1H), proliferation (indexed by BrdU uptake; see Fig. S1I),and in vitro tube formation in collagen gels (representativepictures in Fig. 6B; summarized in Fig. 6C). Whereas theangiogenic prostanoid PGE2 potentiates endothelial cell prolifer-ation, migration, and organization as previously described (22, 23),BW245C did not influence any of these angiogenic processes.

DP Agonism Strongly Increases Intracellular cAMP Concentration inthe Endothelial Layer but Not in the Smooth Muscle Layer. As with theDP receptor, other prostanoid receptors (PGE receptor, EPs

Fig. 5. The DP receptor regulates vascular permeability. VEGF (30 ng, 5 min)or IL-1� (10 ng, 60 min) was administered to DP�/� and WT mice intradermallyon the ear. Evans blue dye (30 mg/kg) was injected and circulated for 30 min.Evans blue dye was extracted and the content was determined. (Representa-tive pictures in A and summary in B, n � 7 each; *, P � 0.05 vs. WT and †, P �0.05 vs. IL-1� single treatment). Mice were treated with the agents intrave-nously according to the following protocols: CAY10404 (CAY), 100 �g/kg, 30min; BW245C, 50 �g/kg, 10 min; BWA868C, 3 mg/kg, 30 min. Then, after VEGF(30 ng, 5 min) or IL-1� (10 ng, 60 min) administration, Evans blue dye (30mg/kg) was injected and circulated for 30 min. (Representative picture in C andsummary in D, n � 7 each; *, P � 0.05 vs. vehicle treatment and †, P � 0.05 vs.BW245C single treatment).

Fig. 6. (A) DP agonism diminished in vitro EC permeability via the cAMP-PKApathway. Confluent monolayers of BAECs were pretreated with the followingtest agents before the stimulation: BW245C and PGD2, 1 �M, 5 min; db-cAMP,10–100 nM, 10 min; BWA868C, 5 �M; and KT5720, 10 �M, 30 min. After VEGFstimulation (10 ng/ml, 20 min), 1% FITC-dextran was added on the upperchamber and incubated for 10 min. The amount of FITC-dextran fluxed intothe lower chamber was measured. For DP knock-down, DP receptor-targeted,or control, siRNA was transfected into BAECs 48 h before the experiments (n �5–6; *, P � 0.05 vs. 10 ng/ml VEGF treatment and †, P � 0.05 vs. VEGF�BW245Ctreatment). (B and C) DP agonism did not affect in vitro endothelial tubeformation. BAECs were sandwiched between gelatinized collagen and incu-bated. After a 4-day incubation, images were taken from four areas in one gel,and tube length was quantified. (B) Representative picture. (Scale bar, 35 �m.)(C) Summary (n � 5–6; *, P � 0.05 vs. vehicle). (D and E) DP receptor stimulationincreased cAMP production in endothelial layer. BAECs and mouse thoracicaortas with (E�) or without (E-) endothelium were stimulated with specifiedtest agents (1 �M PGD2, PGE2, PGI2, BW245C, and 1–10 �M forskolin for 5 min;5 �M BWA868C for 10 min before BW245C treatment) and assayed forintracellular cAMP concentration (n � 4 each; *, P � 0.05 vs. BW245C treat-ment and †, P � 0.05 vs. endothelium intact aorta).

20012 � www.pnas.org�cgi�doi�10.1073�pnas.0805171105 Murata et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

and PGI2 receptor, IP) are also coupled with Gs, acting throughintracellular cAMP production. Finally, therefore, we comparedthe capacity of intracellular cAMP increases by these PGs (Fig.6D). Of interest, 1 �M PGD2 or 1 �M BW245C prominentlyincreased intracellular cAMP in BAECs. The amount inducedwas much more than that stimulated by PGE2 (1 �M), PGI2(1 �M), or 10 �M of forskolin. The effect of BW245C wasabolished by the DP antagonist BWA868C (10 �M, 30 minbefore BW245C), as well as by DP siRNA. Given that thisabundant cAMP elevation by BW245C or PGD2 was not seen inendothelium-denuded intrapulmonary artery (Fig. 6E), endo-thelial DP receptor activation is assumed to be the main sourceof cAMP production in vascular tissue. This data are consistentwith the distribution of DP mRNA levels in intact (E�) orendothelium-denuded (E-) arteries (see Fig. S1C). AlthoughPGI2 also increased cAMP in BAEC (Fig. 6D), it increasedintracellular cAMP content in both endothelium denuded andintact arterial segments (Fig. 6E). This suggested IP receptordistribution both in endothelial and smooth muscle layers.

DiscussionIn the present study, we showed that PGD2 receptor DP islocalized in endothelial cells in tumor vasculature and its defi-ciency accelerates vascular leakage and tumorigenesis. Further-more, we demonstrated exogenous administration of a DPagonist strongly suppresses tumor expansion by a concomitantdecrease in tumor vascularity.

Given that an elevated expression of COX-2 is observed invarious malignancies including lung, esophageal, breast, head,and prostate cancer, investigators have focused their attentionon the role of each prostanoid in tumor growth. Several groupshave reported that the major prostanoids, PGE2 and thrombox-ane A2 (TXA2), accelerate angiogenesis through their distinctreceptors, EPs and TXA2 receptor. Amano et al. reported thathost PGE2-EP3 signal is a prerequisite for stromal vasculargrowth factor expression and tumor angiogenesis in lung carci-noma (10), whereas Kamiyama et al. suggested PGE2-EP2 signalalso accelerates tumor angiogenesis affecting endothelial motil-ity in breast carcinoma (23). Furthermore, the other groupreported that TXA2 receptor agonism promotes angiogenesisand its antagonism blocks IL-1�-induced corneal angiogenesis(17). These results indicating angiogenic properties of majorprostanoids may partially explain the anti-tumor effects ofnon-steroidal anti-inflammatory drugs or COX-2 inhibitors. Inthis article, we provide data documenting the importance of thePGD2-DP receptor signaling pathway as an endogenous negativeregulator of vascular permeability and tumorigenesis. Moreover,DP agonism strongly prevented vascular leakage in implantedLLC tumors and in cultured endothelial cells.

There has been argument about multifaceted biological ac-tions of PGD2 as a pro-inflammatory or anti-inflammatorymediator (11). PGD2 exerts its biological action via binding DPor CRTH2. Furthermore, PGD2 is quickly metabolized into15d-PGJ2 which exhibit anti-inflammatory response via theactivation of PPAR�. These distinct signal pathways on varioustarget-cells and/or -tissues are assumed to complicate PGD2-induced physiological responses. Although, we detected mRNAexpression in vascular cells and LLC tumor, we did not focus therole of CRTH2 in tumorigenesis. Further investigations areneeded to settle the argument.

It is widely recognized that solid tumors require angiogenesisto grow beyond a certain size. This process involves a wide rangeof soluble mediators including VEGF, fibroblast growth factor,angiopoietin, IL-1, and tumor necrosis factor. In an in vivopermeability assay (Fig. 5 A and B) and an angiogenesis assay(Fig. 4 A and B), we demonstrated that host DP deficiency hasmore influence on COX-2-dependent vascularity when com-pared to that induced by VEGF. Thus, the impact of endogenous

PGD2-DP signal in tumor growth is assumed to vary with COX-2activity. The opposite effect was seen with exogenous DPreceptor agonism. This showed robust suppression of the angio-genic response (Fig. 4 C and D and Fig. 5 C and D) and its effectwas irrespective of host COX activity. These features maybroaden the clinical application spectrum of DP agonism.

Previous reports have suggested that vascular leakage, one ofthe first angiogenic responses of the peripheral vasculature toVEGF, may modulate tumor progression (24). To illustrate this,an anti-permeability peptide, cavtratin (25), and TNP-470 (26)have both been shown to acutely suppress vascular leakage andsubsequent tumor growth. Mechanistically, these agents inter-fere with VEGF-dependent nitric oxide production, looseningendothelial cell junctions. Unlike these pathways, we haveproposed a potential strategy to protect the vascular barrier intumors by increasing endothelial-specific cAMP productionthrough the agonistic activation of DP receptor.

The reduced leakage in vivo led to decreased tumor angio-genesis and tumor growth. Because vascular leakage and angio-genesis are important pathophysiological features of variousdiseases including cancer, ischemic injury, and inflammation,our findings implicate the therapeutic potential of DP agoniststo reduce the untoward consequences of enhanced vascularleakage.

Materials and MethodsDP-Deficient Mice and Tumor Implantation. All experiments were approved bythe institutional animal care and use committees of the University of Tokyo.DP-deficient mice were generated, back-crossed more than 10 generations toC57BL/6CrSlc mice, and bred as described (13), and respective control WT miceof the same generation were used (8–10 weeks old). LLC was donated byYoshikazu Sugimoto (through the Riken BRC Cell Bank; Tsukuba, Japan). LLCcells (1 � 106 cells in 100 �l) were injected s.c. on the backs of mice. Tumorvolume was determined daily using a caliper and applying the formula toapproximate the volume of a spheroid (0.52 � [width]2 � [length]).

PGD2 Enzyme Immunoassay. Tumors were harvested and immediately frozen inliquid nitrogen. They were homogenized in ethanol containing 0.02% HCl.After centrifugation, 3H-labeled PGD2 (New England Nuclear) were added astracers for estimation of the recovery to the supernatant. The PGD2 wereextracted with ethyl acetate, and the samples were separated by HPLC. Thequantification was performed with a PGD2-MOX enzyme immunoassay kit(Cayman Chemicals).

Immunofluorescence and TUNEL Assay. Paraffin-embedded sections were usedfor H&E and Ki67 immunostaining. Cryosections were used for all otherimmunostaining. The primary antibodies used were DP (1:1,000, CaymanChemicals), CD31 (1:1000, BD Biosciences), and fibrinogen (1:1,000, Santa CruzBiotechnology). For apoptotic cell labeling, we applied the TUNEL assay kit(Roche) according to the manufacturer’s instructions.

Tumor Permeability and Modified Miles Assay. As a tumor permeability assay,LLC tumors were grown to �800–1,000 mm3 (10 days after implantation) onthe backs of mice. After the mice were injected i.p. with test agents followingeach regimen, Evans blue (30 mg/kg) was injected through the tail vein andcirculated for 30 min. For the modified Miles assay, male C57B/6 (WT andDP�/�) or BalbC mice were used. After treatment with test agents, animalswere anesthetized with ketamine/xylazine and VEGF (30 ng), IL-1� (10 ng), orsaline solution was injected intradermally (30 �l total) into the dorsal ear skinbefore Evans blue dye (30 mg/kg) injection and circulation for 30 min. For bothassays, animals were killed and perfused with 0.5% paraformaldehyde beforetissues were excised, dried, and weighed. Evans blue dye was extracted informamide and its content was quantified by reading at 610 nm in a spectro-photometer.

Corneal Micropocket Assay and Quantification of Corneal Neovascularization.Hydron pellets (0.3 �l) containing VEGF (100 ng/pellet) or IL-1ß (30 ng/pellet)were prepared and implanted in mice corneas. BW245C (dissolved in salinesolution at 0.37 ng/ml; 5 �l) was administered onto the test animal’s eye twicea day. On day 6, the mice were anesthetized and their corneal vessels photo-graphed. Areas of corneal neovascularization were analyzed using the ImageJ

Murata et al. PNAS � December 16, 2008 � vol. 105 � no. 50 � 20013

PHA

RMA

COLO

GY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

1.37 software package and expressed in mm2. The neovascular area wasdetermined by subtracting non-stimulated vascular area from vascular area.

Measurement of Endothelial Permeability and siRNA Preparation. BAECs wereisolated from bovine thoracic aortas. Confluent monolayers of BAEC formedon the transwell inserts with 8 �m pores (Falcon; BD) and serum-starved. Testagents were treated to the upper and lower chamber following each protocol,and then VEGF was added to the upper chamber for 20 min. FITC-dextran (1%)was added to the upper chamber and the entire chamber was incubated for10 min. Aliquots (100 �l) were then retrieved from the lower chamber and FITCconcentrations measured using a fluorescence spectrophotometer. For siRNAtransfection, we purchased siRNA targeted against bovine DP receptor (targetsequence 5�-AAC ATG GAA TCC AGT CTA TAA-3� and 5�-CAC GTC GGT GGAGAA GGG CAA-3�) and negative control siRNA (target sequence 5�-GCG CGCUUU GUA GGA UUC G-3�) from Qiagen. BAECs were transfected with siRNA (30nM) using Lipofectamine 2000 (0.15% vol/vol), following the manufacturer’sprotocols. The cells were then incubated for 48 h to form a monolayer andserum-starved before the assay.

In Vitro Angiogenesis Assay. Collagen gel was prepared by mixing type Icollagen (Nitta Gelatin) solution with MEM and neutralizing buffer. Col-

lagen solution (500 �l) was transferred to each well of 24-well plates andgelatinized. BAECs were plated onto the base layer at a density of 2 � 105

cells/well. Additional collagen solution (500 �l/well) was layered onto thecells at and gelatinized. Media containing 2% FBS was added to each well,and 5 days later cultures were photographed for evaluation of tubeformation and the mean length of tube-like structure was quantified infour fields per group.

Measurement of cAMP Content. BAECs and dissected mouse aortas werestimulated by different treatment agents and immediately homogenized in6% trichloroacetic acid solution. After centrifugation, the supernatants wereapplied to a cAMP enzyme-immunoassay system (Amersham Pharmacia).

Data Representation. All data are shown as mean � SEM.

ACKNOWLEDGMENTS. This work was partly supported by Grant-in-Aid forScientific Research from The Ministry of Education, Culture, Sports, Scienceand Technology and the Japan Society for the Promotion of Science, theNational Institute of Biomedical Innovation of Japan, and ONO ResearchFoundation.

1. Carmeliet P, Jain RK (2000) Angiogenesis in cancer and other diseases. Nature 407:249–257.2. Dvorak HF, et al. (1991) Distribution of vascular permeability factor (vascular endo-

thelial growth factor) in tumors: concentration in tumor blood vessels. J Exp Med174:1275–1278.

3. Senger DR, et al. (1983) Tumor cells secrete a vascular permeability factor that pro-motes accumulation of ascites fluid. Science 219:983–985.

4. Baron JA, Sandler RS (2000) Nonsteroidal anti-inflammatory drugs and cancer preven-tion. Annu Rev Med 51:511–523.

5. Coussens LM, Werb Z (2002) Inflammation and cancer. Nature 420:860–867.6. Masferrer JL, et al. (2000) Antiangiogenic and antitumor activities of cyclooxygenase-2

inhibitors. Cancer Res 60:1306–1311.7. Williams CS, Tsujii M, Reese J, Dey SK, DuBois RN (2000) Host cyclooxygenase-2

modulates carcinoma growth. J Clin Invest 105:1589–1594.8. Chang SH, et al. (2004) Role of prostaglandin E2-dependent angiogenic switch in cyclo-

oxygenase 2-induced breast cancer progression. Proc Natl Acad Sci USA 101:591–596.9. ChangSH,AiY,BreyerRM,LaneTF,HlaT(2005)TheprostaglandinE2receptorEP2 is required

for cyclooxygenase 2-mediated mammary hyperplasia. Cancer Res 65:4496–4499.10. Amano H, et al. (2003) Host prostaglandin E(2)-EP3 signaling regulates tumor-

associated angiogenesis and tumor growth. J Exp Med 197:221–232.11. Kostenis E, Ulven T (2006) Emerging roles of DP and CRTH2 in allergic inflammation.

Trends Mol Med 12:148–158.12. Monneret G, Gravel S, Diamond M, Rokach J, Powell WS (2001) Prostaglandin D2 is a

potent chemoattractant for human eosinophils that acts via a novel DP receptor. Blood98:1942–1948.

13. Matsuoka T, et al. (2000) Prostaglandin D2 as a mediator of allergic asthma. Science287:2013–2017.

14. Brown LF, Dvorak AM, Dvorak HF (1989) Leaky vessels, fibrin deposition, and fibrosis:a sequence of events common to solid tumors and to many other types of disease. AmRev Respir Dis 140:1104–1107.

15. Chen J, et al. (2005) Akt1 regulates pathological angiogenesis, vascular maturation andpermeability in vivo. Nat Med 11:1188–1196.

16. Voronov E, et al. (2003) IL-1 is required for tumor invasiveness and angiogenesis. ProcNatl Acad Sci USA 100:2645–2650.

17. Kuwano T, et al. (2004) Cyclooxygenase 2 is a key enzyme for inflammatory cytokine-induced angiogenesis. FASEB J 18:300–310.

18. Nakao S, et al. (2005) Infiltration of COX-2-expressing macrophages is a prerequisite forIL-1 beta-induced neovascularization and tumor growth. J Clin Invest 115:2979–2991.

19. Mehta D, Malik AB (2006) Signaling mechanisms regulating endothelial permeability.Physiol Rev 86:279–367.

20. Moore TM, Chetham PM, Kelly JJ, Stevens T (1998) Signal transduction and regulationof lung endothelial cell permeability. Interaction between calcium and cAMP. Am JPhysiol 275:L203–L222.

21. Fukuhara S, et al. (2005) Cyclic AMP potentiates vascular endothelial cadherin-mediated cell-cell contact to enhance endothelial barrier function through an Epac-Rap1 signaling pathway. Mol Cell Biol 25:136–146.

22. Shao J, Sheng GG, Mifflin RC, Powell DW, Sheng H (2006) Roles of myofibroblasts inprostaglandin E2-stimulated intestinal epithelial proliferation and angiogenesis. Can-cer Res 66:846–855.

23. Kamiyama M, et al. (2006) EP2, a receptor for PGE2, regulates tumor angiogenesisthrough direct effects on endothelial cell motility and survival. Oncogene 25:7019–7028.

24. Weis SM, Cheresh DA (2005) Pathophysiological consequences of VEGF-induced vas-cular permeability. Nature 437:497–504.

25. Gratton JP, et al. (2003) Selective inhibition of tumor microvascular permeability bycavtratin blocks tumor progression in mice. Cancer Cell 4:31–39.

26. Satchi-Fainaro R, et al. (2005) Inhibition of vessel permeability by TNP-470 and itspolymer conjugate, caplostatin. Cancer Cell 7:251–261.

20014 � www.pnas.org�cgi�doi�10.1073�pnas.0805171105 Murata et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0