-binding fibromodulin can determine stroma matrix structure and fluid balance in experimental carcinoma

Åke Oldberg*, Sebastian Kalamajski*, Alexei V. Salnikov†‡§, Linda Stuhr¶, Matthias Mo¨ rgelinʈ, Rolf K. Reed¶, Nils-Erik Heldin**, and Kristofer Rubin†,††

*Department of Experimental Medical Sciences, BMC, B-12, and ʈDepartment of Clinical Sciences, BMC B14, University of Lund, SE-221 84 Lund, Sweden; †Department of Medical Biochemistry and Microbiology, Uppsala University, BMC, Box 582, SE-751 23 Uppsala, Sweden; ‡Oncology Clinic, University Hospital Lund, SE-221 85 Lund, Sweden; ¶Department of Biomedicine, University of Bergen, N-5009 Bergen, Norway and **Department of Genetics and Pathology, Uppsala University Hospital, Rudbeck Laboratory, SE-751 85 Uppsala, Sweden

Edited by Erkki Ruoslahti, Burnham Institute for Medical Research, Santa Barbara, CA, and approved July 16, 2007 (received for review March 7, 2007) Research on the biology of the tumor stroma has the potential to Clara, CA) of Fc:T␤RII-treated KAT-4 carcinomas revealed lead to development of more effective treatment regimes enhanc- down-regulations predominantly of macrophage-related ing the efficacy of drug-based treatment of solid malignancies. (17). Furthermore, inhibition of IL-1 reduces IFP in these Tumor stroma is characterized by distorted blood vessels and carcinomas (17). These findings indicate a role of inflammation activated connective tissue cells producing a collagen-rich matrix, in the generation of an elevated IFP. Tortuous and leaky blood which is accompanied by elevated interstitial fluid pressure (IFP), vessels, and absence of lymph drainage in tumors have also been indicating a transport barrier between tumor tissue and blood. suggested to be involved in the generation of the elevated tumor Here, we show that the collagen-binding proteoglycan fibromodu- IFP. Inhibition of plasma leakage by interfering with lin controls stroma structure and fluid balance in experimental vascular endothelial growth factor reduces IFP in experimental carcinoma. ablation or inhibition of expression by anti- carcinomas (18, 19) and human rectal carcinoma (15). Little is inflammatory agents showed that fibromodulin promoted the known, however, about how the structure and composition of the formation of a dense stroma and an elevated IFP. Fibromodulin- ECM in the stroma influence carcinoma IFP. deficiency did not affect vasculature but increased the extracellular The small leucine-rich repeat proteoglycan fibromodulin has fluid volume and lowered IFP. Our data suggest that fibromodulin a known role in collagen assembly and maintenance (20). controls stroma matrix structure that in turn modulates fluid Fibromodulin is expressed in dense regular connective tissues, convection inside and out of the stroma. This finding is particularly such as tendons, ligaments, and , but is absent or important in relation to the demonstration that targeted modula- expressed at low levels in , bone, and visceral organs (21, 22). tions of the fluid balance in carcinoma can increase the response to Fibromodulin controls collagen matrix structure in tendons and cancer therapeutic agents. ligaments, and deficient mice show altered tissue organization with fewer and structurally abnormal collagen fiber bundles. The physiology ͉ TGF-␤ ͉ tumor physiology ͉ inflammation ͉ collagen fibrils are thinner (23), tendons and ligaments have interstitial fluid pressure reduced tensile strength, and the animals develop knee-joint instability and osteoarthritis (24–26). Fibromodulin mRNA has espite advances in our knowledge of causative factors and been detected in a variety of clinical malignancies, such as lung, Dgenetic changes during development of the malignant ge- breast, and prostate carcinomas (27–29). Here, we report that notype, treatment results and mean survival time for patients fibromodulin has a role in regulating stroma ECM structure and suffering from advanced cancer have improved only marginally. fluid balance in carcinoma. In recent years, partly in response to disillusion with the available Results cytotoxic cancer chemotherapeutic agents, attention has turned to the stromal elements of tumors. Abnormal blood vessels and We were led to fibromodulin by the finding that in addition to ␤ connective tissue cells that produce a fibrotic collagen-rich lower expression of macrophage-related genes, Fc:T RII sig- matrix characterize the stroma of a carcinoma (1–4). Invasive- nificantly down-regulated mRNA encoding fibromodulin. No- ness and growth of malignant cells are affected by interactions tably, mRNA levels encoding other ECM , such as with the stroma (5), and in this respect a carcinoma resembles an collagen type I, fibronectin, and were unchanged [sup- inflammatory lesion (6) or a wound that will not heal (7). A porting information (SI) Table 3]. Initially, we determined that mathematical model of cancer invasion recently suggested that fibromodulin showed a wide-ranging expression in experimental a heterogeneous extracellular matrix (ECM) with varying ECM carcinomas, such as chemically induced rat mammary carci- component densities selects aggressive tumor cell phenotypes, whereas tumors with a homogenous ECM are less aggressive and Author contributions: Å.O., S.K., and K.R. designed research; Å.O., S.K., A.V.S., L.S., R.K.R., invasive (8). N.-E.H., and K.R. performed research; Å.O., S.K., A.V.S., L.S., M.M., R.K.R., N.-E.H., and K.R. Carcinomas have a pathologically elevated interstitial fluid analyzed data; Å.O., R.K.R., and K.R. wrote the paper. pressure (IFP) indicating a transport barrier between tumor The authors declare no conflict of interest. tissue and blood (9–11). Indeed, pharmaceutical treatments that This article is a PNAS Direct Submission. lower IFP in experimental carcinoma (12–14) and human can- Abbreviations: ECM, extracellular matrix; EBD, Evans’ blue dye; IFP, interstitial fluid pres- cers (ref 15 and references therein) also increase efficacy of sure; ECV, extracellular fluid volumes. cytostatic treatment. Treatment of mice carrying a xenografted §Present address: Division of Molecular Immunology, German Cancer Research Center, human anaplastic thyroid carcinoma (KAT-4) with the soluble D-69120 Heidelberg, Germany. TGF-␤ receptor type II-murine Fc:IgG2A chimeric protein ††To whom correspondence should be addressed. E-mail: [email protected]. (Fc:T␤RII), which inhibits TGF-␤1 and -␤3, lowers IFP, reduces This article contains supporting information online at www.pnas.org/cgi/content/full/ plasma protein leakage, and enhances the anti-tumor effect of 0702014104/DC1. doxorubicin (16, 17). Microarray analyses (Affymetrix, Santa © 2007 by The National Academy of Sciences of the USA

13966–13971 ͉ PNAS ͉ August 28, 2007 ͉ vol. 104 ͉ no. 35 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0702014104 Downloaded by guest on October 1, 2021 A B

C D

EF

G H

Fig. 1. Expression of fibromodulin in carcinomas. (A–C) Immunoperoxidase staining of fibromodulin in PROb rat colonic carcinoma (A), DMBA-induced rat breast carcinoma (B), and mouse RIP-Tag insulinomas (C). Quantitative real- time PCR of KAT-4 tumors from mice treated with control IgG2A or Fc:T␤RII. (D) The difference in mRNA levels between treated and untreated tumors was significant (P Ͻ 0.05, n ϭ 3). (E and F) Staining of fibromodulin in control IgG2A-treated (E) and Fc:T␤RII-treated (F) mice carrying KAT-4 carcinomas. (G and H) Fibromodulin stainings of KAT-4 carcinoma xenografted in WT (G) and Fig. 2. Tumor stroma ultrastructure. (A–D) Transmission electron micro- Fmod-null (H) mice. Mice carrying KAT-4 carcinoma were treated with vehicle graphs of KAT-4 carcinomas grown in WT (A), fibromodulin-deficient (B), or Fc:T␤RII (10 mg/kg) 10 days before killing (16). Tumors were frozen, sec- control IgG2A-treated (C), and Fc:T␤RII-treated (D) mice. (Magnification, A–D, tioned, stained with a rabbit anti-bovine fibromodulin antiserum, and coun- ϫ25,000; scale bar, 200 nm.) (E) Morphometric analyses of collagen fibrils in terstained with hematoxylin (23). (Scale bar, 100 ␮m.) carcinomas grown in WT (open bars) and fibromodulin-deficient (filled bars) mice. (F) Collagen fibril diameters in control IgG2A-treated (open bars) and ␤

Fc:T RII-treated (filled bars) carcinomas. (G and H) Scanning electron micro- CELL BIOLOGY noma, mouse RIP-Tag insulinoma, transplantable syngeneic graphs of carcinomas grown in WT (G) and Fmod-null (H) mice. (I and J) Control ␤ ϫ colonic carcinoma (PROb), and KAT-4 carcinoma growing as a IgG2A-treated (I) and Fc:T RII-treated (J) mice (Magnification, G–J 300; scale xenograft in nude mice (Fig. 1). Staining was restricted to the bar, 100 ␮m). Quantification by image analyses of entire whole-mount tumors revealed that the collagen scaffold density was significantly higher (P Ͻ 0.008) stroma compartment in the carcinomas. This finding is consis- Ϯ ϭ tent with the absence of fibromodulin in normal murine skin, in WT (60 4 pixels per area unit, n 4) (G) compared with Fmod-null carcinomas (45 Ϯ 0.5 pixels per area unit, n ϭ 4) (H). Collagen densities in thyroid, and colon (SI Fig. 5 A–C), which is in agreement with control carcinomas (IgG2A-treated) were 62 Ϯ 4 pixels per unit area (n ϭ 3) (I) the reported distribution of fibromodulin (21, 22). Cultured compared with 49 Ϯ 3 pixels per unit area (n ϭ 3) in Fc:T␤RII-treated carcino- KAT-4 and PROb cells did not express fibromodulin mRNA or mas (J)(P Ͻ 0.03). protein (data not shown). Quantitative PCR analyses showed that treatment of KAT-4 carcinomas with Fc:T␤RII reduced Ϸ fibromodulin mRNA by 50% (Fig. 1D). Immunohistochemical Stroma collagen fibrils in WT and heterozygous mice were analyses showed that fibromodulin protein expression was also similar with an average diameter of 43 nm, whereas those in reduced (Figs. 1 E–F). No staining was seen in carcinomas from KAT-4 carcinomas grown in Fmod-null mice were significantly Fmod-null mice (Fig. 1H). thinner (30 nm) (Fig. 2 A, B, and E). The gross structures of the Fmod-null nude/nude mice were used further to study the role collagen scaffold also differed, with thicker and more abundant of fibromodulin in tumors. Human KAT-4 carcinoma cells were collagen fiber bundles in WT stroma (Fig. 2 G and H). Analyses transplanted s.c. in WT, Fmod-null, and heterozygous nude/nude of entire whole-mount KAT-4 carcinomas showed that the littermates. Tumor take, length of lag phase, and temporal Ͻ increase in external sizes were identical in the three genotypes. densities of the collagen scaffolds were 25% (P 0.008) higher Ϯ ϭ Total carcinoma weights 31 days after transplantation averaged in WT (60 4 pixels per area unit, n 4) compared with 0.54 Ϯ 0.26 g (average Ϯ SD, n ϭ 8) in WT and 0.54 Ϯ 0.21 g Fmod-null carcinomas (45 Ϯ 0.5 pixels per area unit, n ϭ 4). A (n ϭ 6) in fibromodulin-deficient mice. DNA content averaged similar dependence on fibromodulin for collagen scaffold struc- 5.8 mg DNA/g wet weight Ϯ 0.6 (Ϯ SD, n ϭ 5) in WT and 5.8 Ϯ ture was also observed in PROb rat colorectal adenocarcinoma 0.3 (n ϭ 7) in fibromodulin-deficient mice, indicating that (data not shown). Average hydroxyproline contents in pepsin- fibromodulin had no influence on tumor cellularity. resistant triple-helical collagen solubilized from KAT-4 carci-

Oldberg et al. PNAS ͉ August 28, 2007 ͉ vol. 104 ͉ no. 35 ͉ 13967 Downloaded by guest on October 1, 2021 Fig. 3. Effect of fibromodulin deficiency on IFP in carcinomas. KAT-4 (A)or PROb (B) carcinomas were grown in WT, heterozygous (ϩ/Ϫ), and Fmod-null (Ϫ/Ϫ) mice. IFPs were determined by the ‘‘wick-in-the-needle’’ technique described in ref. 16. Average IFP values in carcinomas grown in fibromodulin- deficient animals were significantly different from carcinomas grown in WT animals for both KAT-4 (A) and PROb (B)(P Ͻ 0.05).

nomas were 2.79 Ϯ 0.72 and 1.73 Ϯ 0.13 ␮g/mg dry weight tumor Fig. 4. Analyses of vessel density, macrophage infiltration, and regulation of tissue (Ϯ SD, n ϭ 3, P ϭ 0.066) in carcinomas from WT and fibromodulin mRNA by anti-inflammatory drugs in KAT-4 carcinomas. (A–C) Fmod-null animals, respectively. This 38% lowering of hy- KAT-4 carcinomas grown in WT (open bars), fibromodulin-heterozygous droxyproline content in fibromodulin-deficient carcinomas is in (stripped bars), and fibromodulin-deficient (filled bars) mice were subjected agreement with the difference in the gross structure of the to immunohistochemistry to detect CD31-positive vessels (A), tumor- collagen scaffold and the analyses of collagen densities of whole- infiltrating macrophages (B), and MHC class II-expression by macrophages (C). ␤ Athymic mice carrying KAT-4 carcinomas were treated with 30 mg/kg dexa- mount tumors (Fig. 2 G and H). Fc:T RII treatment, which methasone (n ϭ 3) or rh-IL-1Ra (Kinaret; injected s.c. twice daily for 10 days reduces fibromodulin, resulted in similar collagen fibril diame- with 7.5 mg; n ϭ 2) (17). (D) Fibromodulin mRNA levels were determined by ters and scaffold structures as in the stroma of KAT-4 carcino- real-time PCR, using ␤-actin or GAPDH as an internal standards. mRNA levels mas grown in fibromodulin-deficient mice (Fig. 2 C, D, and F). are presented as percent of the levels in IgG2A-treated control tumors (mean Ϯ Two discrete populations arose after treatment with Fc:T␤RII, range). characterized by thin fibrils similar to those in Fmod-null stroma and wider fibrils, the latter presumably formed before treatment. The collagen scaffold gross structure after Fc:T␤RII treatment were, however, similar (Table 2). Lymph vessels and cells and in fibromodulin-deficient stroma were similar (Fig. 2 I and expressing LYVE-1 were not detected in tumor tissue but were J). Fc:T␤RII treatment reduced collagen density by 21% (P Ͻ abundant in the s.c. tissues surrounding tumors. No apparent differences in distribution or intensity of LYVE-1-stained struc- 0.03). Collagen densities in control carcinomas (IgG2A-treated) were 62 Ϯ 4 pixels per unit area (n ϭ 3) compared with 49 Ϯ 3 tures were detected around KAT-4 carcinomas grown in WT or pixels per unit area (n ϭ 3) in Fc:T␤RII-treated carcinomas. Fmod-null mice (SI Fig. 6). Similar results were obtained when These results are consistent with our observation of a 35% lymphatic vessels were detected by anti-podoplanin antibodies reduction in hydroxyproline after Fc:T␤RII treatment of KAT-4 (data not shown). carcinomas (16) and suggest a role for fibromodulin in the Fibromodulin deficiency had no effect on the number of assembly of the collagen scaffold. macrophages or their expression of MHC class II in KAT-4 IFP was lower in KAT-4 and PROb experimental carcinomas tumor viable tissue (Fig. 4 B and C). Anti-inflammatory agents, grown in Fmod-null mice (Fig. 3). Average IFP values were 5.7 Ϯ such as dexamethasone and recombinant human IL-1 receptor 2.0 mmHg (1mmHg ϭ 133 Pa) (Ϯ SD, n ϭ 10) in WT, 4.7 Ϯ 1.2 antagonist (rh-IL-1Ra) efficiently down-regulated the expres- mmHg (n ϭ 9) in heterozygous, and 3.7 Ϯ 1.3 mmHg (n ϭ 15, sion of fibromodulin mRNA in KAT-4 carcinomas (Fig. 4D). P Ͻ 0.05) in fibromodulin-deficient KAT-4 carcinomas. PROb Immunoreactive fibromodulin in KAT-4 carcinomas was also carcinomas showed IFP values of 5.6 Ϯ 1.7 mmHg (n ϭ 4) when reduced in rh-IL-1Ra-treated carcinomas (SI Fig. 5D). Ϯ ϩ Ϫ ϭ grown in WT mice, 4.3 1.6 mmHg in Fmod / (n 4), and Discussion 2.6 Ϯ 1.8 mmHg in Fmod Ϫ/Ϫ (n ϭ 5). The differences between Here, we show that fibromodulin promotes the formation of a WT and fibromodulin-deficient mice were significant (P Ͻ 0.05) dense collagen scaffold in experimental carcinoma. Absence or assuming a Gaussian distribution of the values. IFP averages in tumors grown in heterozygous mice were intermediate but not significantly different from IFP averages recorded in any of the Table 1. Plasma protein leakage and dextran-perfused vessel other genotypes, indicating a dose-dependent effect of fibro- density in KAT-4 tumors grown in fibromodulin-deficient mice modulin. Fibromodulin-deficiency had no apparent effect on blood EBD leakage, pixels Dextran-perfused vessels, ϭ 2 ϭ vessels in KAT-4 carcinomas. Thus, the densities of CD31- Mouse per area unit; n 3 no. per mm ; n 3 expressing endothelial structures (Fig. 4A), leakage of Evans’ Fmod ϩ/ϩ 74 Ϯ 543Ϯ 10 blue dye (EBD)-labeled albumin, and number of perfused Fmod Ϫ/Ϫ 70 Ϯ 345Ϯ 3 vessels per area unit were not significantly different in KAT-4 carcinomas grown in WT or Fmod-null mice (Table 1). Further- Plasma protein leakage from KAT-4 tumor blood vessels was determined by more, plasma volumes in KAT-4 carcinomas, determined from measuring the extent of EBD-labeled albumin extravasation. One area unit corresponds to 69 ␮m2. FITC-dextran-perfused blood vessels were quantified the distribution volume of radio-labeled albumin, were unaf- under low (ϫ200) magnification according to Chalkley point-overlap mor- fected (Table 2). KAT-4 carcinomas grown in Fmod-null mice phometry. There was no statistically significant difference between vascular had, however, significantly increased extracellular fluid volumes leakage or density of dextran-perfused vessels between wild type (Fmod ϩ/ϩ) (ECV) (Table 2). The ECV in the skin of the two genotypes and Fmod Ϫ/Ϫ carcinomas (P Ͼ 0.05).

13968 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0702014104 Oldberg et al. Downloaded by guest on October 1, 2021 Table 2. Tumor extracellular fluid and plasma volumes ECV, ml/g dry weight; Plasma volume, ml/g Total tissue water, ml/g n ϭ 5 dry weight; n ϭ 5 dry weight; n ϭ 5

Mouse Tumor Skin Tumor Skin Tumor Skin

Fmod ϩ/ϩ 1.78 Ϯ 0.42* 1.91 Ϯ 0.69 0.13 Ϯ 0.12 0.06 Ϯ 0.05 5.27 Ϯ 0.20 2.69 Ϯ 0.42 Fmod Ϫ/Ϫ 2.59 Ϯ 0.63 1.88 Ϯ 0.68 0.11 Ϯ 0.06 0.04 Ϯ 0.02 5.77 Ϯ 0.67 2.62 Ϯ 1.23

Extracellular fluid and plasma volumes in KAT-4 carcinomas grown in wild-type (Fmod ϩ/ϩ) and Fmod-null (Fmod Ϫ/Ϫ) mice. Comparison between Fmod ϩ/ϩ and Fmod Ϫ/Ϫ for the same parameter in the two groups. *, P Ͻ 0.05 (Student’s t test rank, 0.043; Mann–Whitney, 0.028).

reduction of fibromodulin resulted in a lowering of IFP. The The composition of the ECM determines hydraulic conduc- lower IFP recorded in Fmod-null mice could not be ascribed to tivity in tissues (37), and irradiation of experimental tumor enhanced lymphatic drainage because LYVE-1-positive struc- increases collagen content and reduces hydraulic conductivity tures were absent in carcinoma stroma independent of fibro- (38). The sparser collagen scaffold and reduced fibril diameters modulin. Furthermore, we did not detect any vascular differ- in fibromodulin-deficient carcinoma would by itself raise the ences between WT and Fmod-null KAT-4 carcinoma. Taken hydraulic conductivity. The increased ECV will also contribute together, these observations strongly suggest that factors beyond to increased conductivity in the stroma by effectively diluting the lymphatics and the vasculature are responsible for the matrix components (37). The increased hydraulic conductivity reduction in IFP and increase in ECV in fibromodulin-deficient will favor a flow of fluid from parts of the tumor with a high IFP KAT-4 carcinoma. Our results thus indicate that the changes in to the periphery where IFP is lower. Such a redistribution of fluid balance are caused by alterations in stroma collagen fluid should result in a general reduction of IFP in the carcinoma architecture. as seen in fibromodulin-deficient KAT-4 carcinomas. The in- It is likely that the architecture and functional properties of the creased ECV in fibromodulin-deficient carcinomas also suggests collagen scaffold correlate with malignancy (8), although des- that the collagen scaffold is less stiff, allowing tumors to swell moplasia as a prognostic marker remains controversial (30, 31). because of the high plasma protein content in the carcinoma In carcinomas resembling wounds, collagen turnover is generally interstitium that results from the leaky blood vessels. Indeed, high (32). Fibromodulin and other small leucine rich repeat treatment of KAT-4 carcinomas with a specific inhibitor of bind and protect collagen fibrils from degradation carcinoma cell-derived vascular endothelial growth factor (Be- by collagenases in vitro (33). Recent studies in our laboratories vacizumab) decreases plasma protein leakage and reduces the suggest that collagen turnover in tendons is increased in Fmod- ECV (19). An increased hydraulic conductivity and fluid flow in null mice indicating a protective role for fibromodulin against carcinomas may also explain the increased efficacy of chemo- degradation also in vivo (A.O., S.K., and K.R. unpublished data). therapy in KAT-4 carcinomas pretreated with Fc:T␤RII (17). Alternatively, fibromodulin may affect the collagen fibril assem- Several agents reduce IFP, and some increase the delivery and bly by binding and aligning monomeric triple-helical collagen efficacy of chemotherapeutics (11). Here, we show that one on and thin fibrils in the growing fibril. A proper alignment of these agents, Fc:T␤RII, reduces fibromodulin expression. Fur- collagen is essential for interactions and the formation of thermore, the collagen scaffold restricts diffusive transport of cross-links in the growing fibril. An increased turnover may in macromolecules in the stroma of experimental carcinoma (10, part be due to misaligned collagen fibrils that are rapidly 39) and thus most likely the delivery of high-molecular weight degraded in fibromodulin-deficient tendons. The present find- anti-cancer drugs. The unexpected role for fibromodulin in ings of fibromodulin-directed regulation of collagen assembly in controlling carcinoma fluid balance may provide a basis for a carcinoma suggests that fibromodulin synthesis constitutes a treatment to improve delivery of anti-cancer drugs by modula- CELL BIOLOGY mechanism providing a stable collagen scaffold that otherwise tion of fibromodulin expression. would be rapidly degraded. The tumor stroma is generated by carcinoma cell-driven Methods activation of normal, i.e., nontransformed host cells, such as Animals and Tumors. Fibromodulin-deficient mice were generated endothelial cells, pericytes, and fibroblasts. Carcinoma cells may and maintained in 129 SV background (23). Fmod-null mice activate such stroma target cells directly and/or indirectly via were crossed with C57BL nu/nu mice to generate male (nude/ infiltrated immune cells. In recent years much focus has been nude, Fmod ϩ/Ϫ) and female (nude/wild type, Fmod ϩ/Ϫ). All directed to the role of macrophages for the development of a experiments were conducted on nu/nu littermates generated by tumor stroma. Activated macrophages in carcinoma modulate breeding male (nude/nude, Fmod ϩ/Ϫ) and female (nude/wild their microenvironment (6) and are conversely modulated by a type, Fmod ϩ/Ϫ). Mice were genotyped as described in ref. 26. hypoxic microenvironment (34). Our data show that fibromodu- KAT-4 anaplastic thyroid carcinoma cells (5 ϫ 106) or rat PROb lin does not affect macrophage numbers or their expression of (5 ϫ 106) colorectal cells were injected s.c. in the left flank of the MHC class II antigens, the latter reportedly reflecting macro- mice and treated with Fc:T␤RII (10 mg/kg) as described in refs. phage activation (35). Treatment of KAT-4 carcinomas with 14 and 16. All animal experiments were approved by the Ethical Fc:T␤RII reduces inflammatory characteristics (17) and down- Committee for Animal Experiments in Uppsala and Lund, regulated fibromodulin expression (present findings). In agree- Sweden. The number of animals was minimized to comply with ment, the anti-inflammatory agents dexamethasone and IL-1 guidelines from the Ethical Committee. receptor antagonist down-regulated fibromodulin. The same agents and Fc:T␤RII reduce IFP in KAT-4 carcinomas (ref. 16 In Vivo Perfusion, Vascular Permeability, Extracellular Fluid, and and unpublished data). Dexamethasone reduces IFP also in Plasma Volumes. Perfused tumor vessels were visualized by using other types of experimental carcinomas (36). Thus, our findings FITC-dextran (molecular mass 2,000 kDa) (Sigma, St. Louis, suggest that inflammatory processes in carcinomas promote MO). Vascular permeability was assessed by determining leak- fibromodulin synthesis by stroma cells, leading to the formation age of EBD into the tumor tissue interstitium. EBD (30 mg/kg) of a dense and stiff collagen scaffold and a high IFP. (Sigma) was administered i.v. 30 min and FITC-dextran (100

Oldberg et al. PNAS ͉ August 28, 2007 ͉ vol. 104 ͉ no. 35 ͉ 13969 Downloaded by guest on October 1, 2021 mg/kg) 2 min before animal killing. To evaluate the perfused TRIZOL solution (Invitrogen, San Diego, CA) and with DNase area of tumor vasculature and leakage of EBD, 20-␮m frozen I, and a first-strand cDNA synthesis was performed [first-strand sections were analyzed by fluorescent microscopy. Images were synthesis kits (Roche Diagnostics, Penzberg, Germany) or obtained from 16 fields of vision from eight sections per tumor iScript (Bio-Rad Laboratories, Sundbyberg, Sweden)], using taken 100 ␮m apart and assessed under low magnification. The oligo(dT) primers. The cDNA samples were mixed with specific extent of EBD leakage was analyzed by using NIH Image 1.62 primers and SYBR Green and amplified in an Applied Biosys- software (National Institutes of Health, Bethesda, MD). Digital tems Prism instrument (Applied Biosystems, Foster City, CA), images were analyzed in a gray-scale mode, and dye density in starting with an initial 10-min heating at 95°C followed by 40 tumor sections was presented as number of pixels per area of cycles at 95°C for 15 s, 60°C for 30 s, and 72°C for 30 s. The data tumor tissue. The density of FITC-dextran-perfused vessels was were analyzed by using SDS software, Version 2.1 (Applied quantified in 16 fields of vision with high vascular density under Biosystems). The calculated threshold cycle values were normal- low magnification according to Chalkley point-overlap mor- ized to the threshold cycle-value for ␤-actin or glyceraldehyde- phometry. Using this method, the mean of perfused vessel 3-phosphate dehydrogenase (GAPDH) used as internal densities of the most vascular areas within the tumor (ϫ200 standard. The primers used were: Fmod, 5Ј-AGCAGTCCAC- magnification) is calculated manually. Selected areas (‘‘hot CTACTACGACC-3Ј and 5Ј-CAGTCGCATTCTTGGG- spots’’) were identified after scanning the whole section at low GACA-3Ј; ␤-actin, 5Ј- GGCACCCAGCACAATGAAG-3Ј and magnification. This technique has been established as a standard 5Ј-GCCGATCCACACGGAGTACT-3Ј; GAPDH, 5Ј-AGGT- method for quantification of angiogenesis in sections from solid CGGTGTGAACGGATTTG-3Ј and 5Ј-TGTAGACCATG- tumors. ECV in KAT-4 carcinomas were measured as extravas- TAGTTGAGGTCA-3Ј. cular distribution space of 51Cr-EDTA after functional nephrec- tomy by the bilateral ligation of the renal pedicles via flank Electron Microscopy. Three similarly sized tumors from WT, incisions. Plasma volumes were measured by using 125I-labeled ␤ 125 fibromodulin-deficient, IgG2A-treated, and Fc:T RII-treated human serum albumin ( I-HSA; Institute for Energy Tech- mice were investigated by transmission electron microscopy. niques, Kjeller, Norway). Both techniques have been detailed Carcinomas were fixed in 0.15 M sodium cacodylate-buffered elsewhere (19). 2% glutaraldehyde, postfixed in 0.1 M-collidine-buffered 2% osmium tetroxide, and embedded in epoxy resin (23). Ultrathin Immunohistochemistry and Immunofluorescence. Six-micrometer sections were analyzed in Philips CM-10 electron microscope frozen sections of KAT-4 carcinomas were subjected to immu- (Philips, Amsterdam, The Netherlands). Scanning electron mi- nohistochemistry with anti-mouse CD31/PECAM-1 antibody croscopy was performed on whole-mount specimens prepared by (BD PharMingen, San Diego, CA) detecting endothelial cells alkali maceration (41) and analyzed in a Philips 515 electron and anti-mouse F4/80 antibody (Serotec, Oxford, U.K.) detect- microscope. For quantification of matrix amount, the images ing macrophages. Blood vessel density and macrophage infiltra- ϫ ϫ were converted to white and black pixels that represented matrix tion of tumor tissue were analyzed under 200 and 500 and empty areas, respectively. Pixels were quantified with Im- magnification by using a counting grid as described in ref. 19. ageJ software in three whole-mount carcinomas of similar size Pericyte coverage of CD31-positive blood vessels was deter- from control and Fc:T␤RII-treated carcinomas and from carci- mined by double immunofluorescence with anti-smooth muscle nomas grown in WT and fibromodulin-deficient mice. Collagen actin antibody (Sigma). MHC class II antigen expression by densities are expressed as percent of area occupied by matrix. F4/80-positive intratumoral macrophages was analyzed with anti-mouse FITC-labeled I-A/A-E (2G9) antibody (BD Phar- Mingen). Lymphatic vessels were detected by double immuno- Statistical Methods. The unpaired, two-tailed Student’s t test was used. The Mann–Whitney U test was used when data failed to fluorescence with rabbit anti-LYVE-1 IgG (Abcam, Cambridge, Ͻ U.K.) and anti-CD31. fulfill the normality criteria. P 0.05 was considered statistically significant. Standard deviations of means are indicated in the Quantitation of Hydroxyproline and Quantitative Real Time PCR. figures unless otherwise specified. Lyophilized tumors were treated with 2 mg/ml porcine stomach mucosa pepsin (Sigma) in 0.5 M acetic acid. Released triple We thank Ann-Marie Gustafson, Annika Hermansson, and Gerd Salvesen for technical assistance. This work was supported by grants helical collagen was precipitated with 1 M NaCl. The collagen from the Swedish Cancer Foundation (N.-E.H. and K.R.), Swedish precipitate was hydrolyzed by 6 M HCl at 110°C for 18 h. Research Council (Å.O., N.-E.H., and K.R.), Gustaf V:s 80-Anniversary Hydroxyproline content was determined in the hydrolysate as Fund (Å.O. and K.R.), Greta and Johan Kocks and Alfred O¨ sterlunds described in ref. 40. Total RNA was extracted from KAT-4 Funds (Å.O.), Gunnar Nilssons Cancer Organization (K.R.), and the tumors by using RNAeasy kit (Qiagen, Valencia, CA) and Norwegian Research Council (L.S. and R.K.R.).

1. Hashizume H, Baluk P, Morikawa S, McLean JW, Thurston G, Roberge S, Jain 13. Pietras K, Rubin K, Sjo¨blom T, Buchdunger E, Sjo¨quist M, Heldin C-H, RK, McDonald DM (2000) Am J Pathol 156:1363–1380. O¨ stman A (2002) Cancer Res 62:5476–5484. 2. Mueller MM, Fusenig NE (2004) Nat Rev Cancer 4:839–849. 14. Salnikov AV, Iversen VV, Koisti M, Sundberg C, Johansson L, Stuhr LB, 3. Bhowmick NA, Neilson EG, Moses HL (2004) Nature 432:332–337. Sjo¨quist M, Ahlstro¨m H, Reed RK, Rubin K (2003) FASEB J 17:1756– 4. Liotta LA, Kohn EC (2001) Nature 411:375–379. 1758. 5. Hanahan D, Weinberg RA (2000) Cell 100:57–70. 15. Willett CG, Boucher Y, Di Tomaso E, Duda DG, Munn LL, Tong RT, Chung 6. Coussens LM, Werb Z (2002) Nature 420:860–867. DC, Sahani DV, Kalva SP, Kozin SV, et al. (2004) Nat Med 10:145–147. 7. Dvorak HF (1986) N Engl J Med 315:1650–1659. 16. Lammerts E, Roswall P, Sundberg C, Gotwals PJ, Koteliansky VE, Reed RK, 8. Anderson AR, Weaver AM, Cummings PT, Quaranta V (2006) Cell 127:905–915. Heldin N-E, Rubin K (2002) Int J Cancer 102:453–462. 9. Jain RK (1987) Cancer Res 47:3039–3051. 17. Salnikov AV, Roswall P, Sundberg C, Gardner H, Heldin N-E, Rubin K (2005) 10. Netti PA, Berk DA, Swartz MA, Grodzinsky AJ, Jain RK (2000) Cancer Res Lab Invest 85:512–521. 60:2497–2503. 18. Lee CG, Heijn M, di Tomaso E, Griffon-Etienne G, Ancukiewicz M, Koike C, 11. Heldin C-H, Rubin K, Pietras K, O¨ stman A (2004) Nat Rev Cancer 4:806– Park KR, Ferrara N, Jain RK, Suit HD, Boucher Y (2000) Cancer Res 813. 60:5565–5570. 12. Emerich DF, Dean RL, Snodgrass P, Lafreniere D, Agostino M, Wiens T, 19. Salnikov AV, Heldin N-E, Stuhr LB, Wiig H, Gerber H, Reed RK, Rubin K Xiong H, Hasler B, Marsh J, Pink M, et al. (2001) J Pharmacol Exp Ther (2006) Int J Cancer 119:2795–2802. 296:632–641. 20. Ameye L, Young MF (2002) Glycobiology 12:107R–116R.

13970 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0702014104 Oldberg et al. Downloaded by guest on October 1, 2021 21. Heinegård D, Larsson T, Sommarin Y, Franze´n A, Paulsson M, Hedbom E 30. Ueno H, Jones AM, Wilkinson KH, Jass JR, Talbot IC (2004) Gut 53:581–586. (1986) J Biol Chem 261:13866–13872. 31. Sis B, Sarioglu S, Sokmen S, Sakar M, Kupelioglu A, Fuzun M (2005) J Clin 22. Saamanen AM, Salminen HJ, Rantakokko AJ, Heinegård D, Vuorio E (2001) Pathol 58:32–38. Biochem J 355:577–585. 32. Hotary KB, Allen ED, Brooks PC, Datta NS, Long MW, Weiss SJ (2003) Cell 23. Svensson L, Aszodi A, Reinholt FP, Fa¨sslerR, Heinegård D, Oldberg Å (1999) 114:33–45. J Biol Chem 274:9636–9647. 33. Geng Y, McQuillan D, Roughley PJ (2006) Matrix Biol 25:484–491. 24. Gill MR, Oldberg Å, Reinholt FP (2002) Osteoarthritis Cartilage 10:751–757. 34. Murdoch C, Muthana M, Lewis CE (2005) J Immunol 175:6257–6263. 25. Jepsen KJ, Wu F, Peragallo JH, Paul J, Roberts L, Ezura Y, Oldberg A, Birk 35. Gordon S (2003) Nat Rev Immunol 3:23–35. DE, Chakravarti S (2002) J Biol Chem 277:35532–35540. 36. Kristjansen PE, Boucher Y, Jain RK (1993) Cancer Res 53:4764–4766. 26. Ameye L, Aria D, Jepsen K, Oldberg Å, Xu T, Young MF (2002) FASEB J 37. Levick JR (1987) Q J Exp Physiol 72:409–437. 16:673–680. 38. Znati CA, Rosenstein M, McKee TD, Brown E, Turner D, Bloomer WD, 27. van’t Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M, Peterse HL, van der Kooy K, Marton MJ, Witteveen AT, et al. (2002) Nature 415:530–536. Watkins S, Jain RK, Boucher Y (2003) Clin Cancer Res 9:5508–5513. 28. Welsh JB, Sapinoso LM, Su AI, Kern SG, Wang-Rodriguez J, Moskaluk CA, 39. Brown E, McKee T, diTomaso E, Pluen A, Seed B, Boucher Y, Jain RK (2003) Frierson HF, Jr, Hampton GM (2001) Cancer Res 61:5974–5978. Nat Med 9:796–800. 29. Garber ME, Troyanskaya OG, Schluens K, Petersen S, Thaesler Z, Pacyna- 40. Berg RA (1982) Methods Enzymol 82:372–398. Gengelbach M, van de Rijn M, Rosen GD, Perou CM, Whyte RI, et al. (2001) 41. Ohtani O, Ushiki T, Taguchi T, Kikuta A (1988) Arch Histol Cytol Proc Natl Acad Sci USA 98:13784–13789. 51:249–261. CELL BIOLOGY

Oldberg et al. PNAS ͉ August 28, 2007 ͉ vol. 104 ͉ no. 35 ͉ 13971 Downloaded by guest on October 1, 2021