Supporting Information

Xie et al. 10.1073/pnas.1105041108 SI Materials and Methods Real-Time Quantitative RT-PCR of Integrin mRNAs in RT2 Mice. Nor- Synthetic Peptides. Synthetic TMPFLFCNVNDCNFASRNDY- mal islets from wild-type C57BL/6 mice, hyperplastic islets from SYWL-tumstatin peptide (1) and IVRRADRAAVP-endostatin 5-wk-old RT2 mice, and angiogenic islets from 10-wk-old RT2 peptide (2) were synthesized at Tufts Core Facility and analyzed mice were collected as previously described (14). Tumors were by mass spectrometry and purified by analytic HPLC. A re- dissected from excised pancreata and the surrounding exocrine combinant protein, designated TSR2+RFK, which contains the tissues were carefully removed. RNA was prepared from pools of second type 1 repeat of human TSP1, was expressed in insect snap-frozen islets and tumors using the RNeasy kit (Qiagen). One cells as described previously (3). A mutant protein in which the microgram of total RNA was transcribed into single-stranded RFK sequence is excluded and the three conserved tryptophan cDNA using SuperScript II RNase H reverse transcriptase (In- residues were mutated to threonine (TSR2(W/T)) was prepared vitrogen). Real-time quantitative RT-PCR was performed on in the same way and used as a negative control (3). cDNA synthesized from normal, hyperplastic, angiogenic islets and tumors, or flow-sorted cells using the 5′ nuclease assay (real- In Vitro Viability Assay. In a 96-well plate, HUVEC cells were seeded time Taqman RT-PCR) with the Prism 7700 instrument (ABI) as 3 at a concentration of 6 × 10 cells per well in 100 μLDMEMsup- previously described (15). The primers for RT-PCR were in- plemented with 10% FCS. Tumstain, endostatin, and TSP1 protein ventoried probe sets from Applied Biosystems. RNA levels were fi reagents were added to the 10% FCS/DMEM media at a nal standardized to the probe/primer set for murine GAPDH. concentration of 10 μM. As a negative control, cells were incubated with 0.1% FCS in DMEM. After incubation for 40 h at 37 °C and Immunohistochemistry/Immunofluorescence. Immunohistochemical μ 5% CO2,10 L of WST-1 reagent [2-(4-Iodophenyl)-3-(4-nitro- staining was performed as previously described (16). Briefly, phenyl)-5-(2,4-disulfophenyl)-2H-tetrazolium, monosodium salt] 4-μM frozen sections were fixed in 100% ice-cold acetone for (Roche) was added to each well and incubated for 4 h at 37 °C and 5 min and then air dried. They were incubated with various 5% CO2. The absorbance at 440 nm was measured against a back- primary antibodies: rat antimouse CD31 (BD Bioscience), ground control using a microplate reader. All experiments were hamster antimouse β3 integrin (BD Bioscience), rat antimouse performed in triplicate and repeated at least three times. α5 integrin (BD Bioscience), rat antimouse CD36 (BD Bio- science), rat antimouse CD47 (BD Bioscience), rat antimouse Transgenic Mice. Wild-type C57BL/6 mice were purchased from CD34 (Abcam), mouse anti-PNCA (DAKO), goat anti–HIF-1α The Jackson Laboratory. Transgenic Rip1Tag2 mice on C57BL/6 (Santa Cruz), and mouse anti-BrdU (BD Bioscience) at 4 °C background were obtained from National Cancer Institute mouse overnight. Subsequently, they were washed three times with PBS models of human cancers consortium (MMHCC: http://www. cancer.gov). The generation and phenotypic characterization of buffer and incubated with FITC- or rhodamine-conjugated sec- Rip1Tag2 mice have been described previously (4). The standard ondary antibodies (Jackson ImmunoResearch Laboratories) at chow of all β-cell tumor-bearing mice was supplemented with room temperature for 1 h. After three washes with PBS, Vec- glucose pellets (Harlan Teklad) beginning at 6 wk of age to tashield (Vector Laboratories; App Imaging) mounting medium prevent hypoglycemia due to hyperinsulinemia from developing was applied and sections were coverslipped and imaged. Alter- insulinoma. Tumstatin-deficient mice (knockout of type IV col- natively for immunohistochemistry, specimens were processed lagen α3 chain) were described previously (5–8) and obtained using the Vectastain ABC kit (Vector Laboratories) and DAB fi from The Jackson Laboratory and backcrossed into C57BL/6 was used as the substrate reagent. For quanti cation, the number − − genetic background. The β3 integrin / mice described pre- of antigen-positive vessels/cells was counted at 20× (for blood viously (9) were backcrossed to C57BL/6 and obtained from vessels) and 40× (for cell proliferation and HIF-1α expression) the Massachusetts Institute of Technology colony or from The magnification in a blinded fashion in 10 separate fields and av- − − Jackson Laboratory. The TSP1 / mice on a C57BL/6 back- eraged. Cell death was determined by TUNEL staining using the ground were originally described by Lawler et al. (10). The ApopTag Red in situ Apoptosis Detection kit (Chemicon) ac- endostatin-deficient (deletion of type XVIII collagen) mice cording to the manufacturer’s protocol. Hypoxia was assessed by were originally described by Fukai et al. (11) and maintained on immunofluorescence with the hypoxyprobe-1 Plus kit (Chem- − − C57BL/6 background. The RT2/β3integrin/ mice, RT2/ icon) according to the manufacturer’s protocol. tumstatin-deficient mice, RT2/endostatin-deficient mice, and − − RT2/TSP1 / mice were generated by intercrossing and used Statistical Analysis. All values are expressed as mean ± SEM. − − for experimental analysis. p53 / mice C57BL/6 were described Analysis of variance (ANOVA) was used to determine statistical previously (12) and obtained from The Jackson Laboratory and differences between groups using SPSS software. Bonferroni − − − − crossed with tumstatin / mice and TSP1 / mice to generate post hoc analysis was performed, when equal variances were − − − − − − − − − − p53 / /tumstatin / mice and p53 / /tumstatin / /TSP1 / assumed. Dunnett’s T3 post hoc analysis was performed, when mice, respectively. Generation of Rip1-Tag2/Cre;VEGF fl/fl equal variances were not assumed. A level of P < 0.05 was (VEGFRIPKO) mice was described previously (13). considered statistically significant.

1. Maeshima Y, Colorado PC, Kalluri R (2000) Two RGD-independent alpha vbeta 3 4. Hanahan D (1985) Heritable formation of pancreatic beta-cell tumours in transgenic integrin binding sites on tumstatin regulate distinct anti-tumor properties. J Biol mice expressing recombinant insulin/simian virus 40 oncogenes. Nature 315:115–122. Chem 275:23745–23750. 5. Miner JH, Sanes JR (1996) Molecular and functional defects in kidneys of mice lacking 2. Wickstrom SA, Alitalo K, Keski-Oja J (2004) An endostatin-derived peptide interacts collagen alpha 3(IV): Implications for Alport syndrome. J Cell Biol 135:1403–1413. with integrins and regulates actin cytoskeleton and migration of endothelial cells. 6. Cosgrove D, et al. (1996) Collagen COL4A3 knockout: A mouse model for autosomal J Biol Chem 279:20178–20185. Alport syndrome. Genes Dev 10:2981–2992. 3. Miao WM, et al. (2001) Thrombospondin-1 type 1 repeat recombinant proteins inhibit 7. Sugimoto H, et al. (2006) Bone-marrow-derived stem cells repair basement membrane tumor growth through transforming growth factor-beta-dependent and -independent collagen defects and reverse genetic kidney disease. Proc Natl Acad Sci USA 103: mechanisms. Cancer Res 61:7830–7839. 7321–7326.

Xie et al. www.pnas.org/cgi/content/short/1105041108 1of5 8. Hamano Y, et al. (2003) Physiological levels of tumstatin, a fragment of collagen IV 13. Inoue M, Hager JH, Ferrara N, Gerber HP, Hanahan D (2002) VEGF-A has a critical, alpha3 chain, are generated by MMP-9 proteolysis and suppress angiogenesis via nonredundant role in angiogenic switching and pancreatic beta cell carcinogenesis. alphaV beta3 integrin. Cancer Cell 3:589–601. Cancer Cell 1:193–202. 9. Hodivala-Dilke KM, et al. (1999) Beta3-integrin-deficient mice are a model for 14. Parangi S, Dietrich W, Christofori G, Lander ES, Hanahan D (1995) Tumor suppressor Glanzmann thrombasthenia showing placental defects and reduced survival. J Clin loci on mouse chromosomes 9 and 16 are lost at distinct stages of tumorigenesis in Invest 103:229–238. a transgenic model of islet cell carcinoma. Cancer Res 55:6071–6076. 10. Lawler J, et al. (1998) Thrombospondin-1 is required for normal murine pulmonary 15. Elson DA, et al. (2001) Induction of hypervascularity without leakage or inflammation homeostasis and its absence causes pneumonia. J Clin Invest 101:982–992. in transgenic mice overexpressing hypoxia-inducible factor-1alpha. Genes Dev 15: 11. Fukai N, et al. (2002) Lack of collagen XVIII/endostatin results in eye abnormalities. 2520–2532. EMBO J 21:1535–1544. 16. Maeshima Y, et al. (2002) Tumstatin, an endothelial cell-specific inhibitor of protein 12. Donehower LA, et al. (1992) Mice deficient for p53 are developmentally normal but synthesis. Science 295:140–143. susceptible to spontaneous tumours. Nature 356:215–221.

Fig. S1. The initial angiogenic switch in incipient neoplasias of the endocrine pancreas is highly dependent on the angiogenesis inducer VEGF. RT2/VEGFWT mice display typical angiogenic and hyperplastic islet features at 10 wk of age. (A) The angiogenic islets are characterized by extensive vessel leakage and vascularization as shown in D as well as a hyperproliferative status as shown in G and J.(B) Hyperplastic islets of RT2/VEGFWT are also vascularized as shown in E, allowing initiation of a hyperproliferative state as shown in H and K.(C) RT2/VEGFRIPKO islets typically remained in a mildy hyperplastic state and demonstrated severe hypoxia as determined by pimonidazole staining with little to no angiogenesis as shown in F, and a moderate to low hyperproliferative status as shown in I, accompanied by enhanced apoptosis as shown in L. Islets are outlined in white dashed lines, whereas blood lakes are outlined in yellow dashed lines. (Scale bar, 200 μM.)

Xie et al. www.pnas.org/cgi/content/short/1105041108 2of5 Fig. S2. Distinctive expression profiles of α5 integrin, β3 integrin, and CD36 in endothelial cells in the stages of RT2 tumorigenesis. (A) Colocalization of α5 integrin and CD31 in RT2 mice. α5 integrin expression level remained nearly constant in blood vessels during the different stages of tumor progression in RT2 mice. (B) Colocalization of β3 integrin and CD31 in RT2 mice. β3 integrin was not detectably expressed until low levels appeared at the onset of angiogenesis and reached maximal levels in small tumors. (C) Colocalization of CD36 and CD31 in RT2 mice. CD36 expression was observed in vessels during all stages of RT2 mice tumor progression, with the highest expression level in vessels of hyperplastic tissues. (D) Quantification of α5 integrin, β3 integrin, and CD36 expression levels in CD31-positive vessels in RT2 mice. Small tumor refers to tumor with a diameter less than 2 mm; large tumor refers to those with diameter larger than 2 mm. (Scale bar, 50 μM.) See also Figs. S1 and S5.

Fig. S3. Comparisons of integrin mRNA levels in endothelial cells using quantitative real-time PCR analysis. Integrin and CD36 gene expression were measured in the distinct phases of RT2 tumorigenesis. The levels of mRNAs for these receptors were normalized to CD31 mRNA level in normal islets. Tumor endothelial cells generally expressed somewhat higher levels of integrins than did endothelial cells associated with normal, hyperplastic, and angiogenic islets.

Fig. S4. Immunohistochemical analysis of angiogenic islets isolated from SU10944/EAI-treated mice. Angiogenic islets were isolated as described in Materials and Methods. Islets were fixed and subsequently stained for CD34 (microvessel density), HIF-1α (hypoxic response), and PCNA (proliferation). Representative images are provided. Data analysis is described in Materials and Methods.

Xie et al. www.pnas.org/cgi/content/short/1105041108 3of5 Fig. S5. Lymphoma angiogenesis and vessel β3 integrin expression. Tumors, primarily lymphomas, were observed in p53−/−, p53−/−/tumstatin−/−, and p53−/−/ tumstain−/−/TSP1−/− mice at 3 mo of age. Most vessels of the of p53−/−, p53−/−/tumstatin−/−, and p53−/−/tumstain−/−/TSP1−/− lymphomas express high levels of β3 Integrin.

Fig. S6. Microvessel density of angiogenic islets in EAI-deficient and β3 knockout RT2 mice. (A) Microvessel density was assessed by counting the number of CD31 positive vessels in a high-powered 400× field. (B) Representative images of CD31 positive vessels for EAI-deficient and β3 knockout mice.

Xie et al. www.pnas.org/cgi/content/short/1105041108 4of5 Fig. S7. Hypoxia in RIP-TAg2 tumors. Hypoxia was detected by staining for the probe pimonidazole hydrochloride, which was administered to mice before killing. An increase in hypoxia was observed between 8 and 10 wk in the RT2 model. This time point corresponds to angiogenic switching and small tumor formation during RT2 tumor progression.

Table S1. Tumor spectrum in p53 deficient mice Tumor type

Genotype Tumor incidence, %* Lymphoma, % Sarcoma, % Other, %

p53−/− 25 66.67 22.22 11.11 − − − − p53 / tumstatin / 43 81.81 9.09 9.09 − − − − − − p53 / tumstatin / TSP1 / 50 83.33 8.33 8.33

*All mice were 12 wk of age.

Xie et al. www.pnas.org/cgi/content/short/1105041108 5of5