Supplementary Figure 1. Analysis of hallmarks of cancer in Csf1op/op RT2 tumors. WT and Csf1op/op RT2 mice were sacrificed at 13.5 weeks of age, and pancreatic tissues isolated and prepared for frozen histology as previously described (Gocheva et al 2010a). (a) For proliferation analysis, mice were injected intraperitoneally with 100 mg/g of bromodeoxyuridine (BrdU; Sigma) 2 hours prior to sacrifice. Pancreatic tissues were stained with rat-anti BrdU antibodies (clone BU1/75(ICR1), 1:200, Serotec or Abcam) with DAB detection. The percentage of BrdU+ nuclei relative to the total number of hematoxylin+ nuclei per tumor was quantitated with HistoQuest analysis software (TissueGnostics). Graph depicts the mean percentage of BrdU+ cells per mouse (n=7 mice per genotype). (b) For analysis of apoptosis, pancreatic tissues were immunofluorescently stained for cleaved caspase-3 (CC3; rabbit polyclonal anti-cleaved caspase 3 (Asp175), 1:500, Cell Signaling Technology) and the percentage of CC3+ cells relative to the total number of DAPI+ cells per tumor was quantitated with TissueQuest analysis software as previously described (Gocheva et al 2010a). Graph depicts the mean percentage of CC3+ cells per mouse (n=5 mice per genotype). (c-e) Pancreatic tissues (n=5 mice per genotype) were immunofluorescently co-stained for vessels (rat monoclonal anti-mouse CD31, clone MEC 13.3, 1:100, BD Pharmingen) and pericytes (rabbit polyclonal anti-NG2 chondroitin sulfate proteoglycan, 1:200, Chemicon/Millipore) for angiogenesis analysis. (c) Overviews of pancreatic sections from WT RT2 (left) and Csf1op/op RT2 mice (right) stained for angiogenesis analysis were generated by TissueGnostics acquisition software by stitching together individual 200x images. Tumors (T) and the surrounding normal exocrine pancreas (E) are indicated. The CD31+ vessel area (d) and pericyte coverage (e), the percentage of NG2+ staining overlapping CD31+ staining, were quantitated using Metamorph (Molecular Devices), as previously described (Gocheva et al 2010a). Each dot represents an individual mouse, and the horizontal line the mean. Statistical analysis by unpaired Student’s t-test revealed no significant differences in (a, b, d, e). (f) For invasion analysis, pancreatic tissues from WT RT2 (n=8) and Csf1op/op RT2 (n=14) were stained with hematoxylin and eosin, and the tumors were graded as previously described (Lopez and Hanahan 2002). The distribution of tumor types between genotypes was compared using a cumulative logit model with generalized estimating equations to correct for correlations within mice (not significant).

Supplementary Figure 2. Macrophage quantitation in RT2 tumors with the cell type specific marker Iba1. (a) Iba1 staining labels macrophages. Pancreatic tissue from WT RT2 mice was co-stained for the macrophage cell type specific markers F4/80 (red; rat anti-F4/80, clone CI:A3-1, 1:1000, Serotec) and Iba1 (green; rabbit anti-Iba1, 1:1000, Wako Pure Chemical Industries, Japan). Images were analyzed using TissueQuest analysis software (TissueGnostics) and the percentage of F4/80+ cells that were positive for Iba1 was quantified and graphed for individual RT2 tumors (n=13). The horizontal red line indicates the mean for the data set. (b) Representative images of immunostaining for Iba1+ macrophages (red) in 13.5 week old WT RT2 (left panel) and Csf1op/op RT2 (right panel) pancreatic tumors. Magnification x200. (c) The percentage of Iba1+ macrophages relative to the total number of DAPI+ cells per islet lesion (excluding macrophages in the normal exocrine pancreas) was quantified with TissueQuest Analysis software as in Figure 4. A total of 4 WT RT2 (black bars) and 5 Csf1op/op RT2 (gray bars) mice were analyzed. Bars, mean per mouse+SEM. *P <0.05, **P <0.005, by unpaired Student’s t-test. Note that the pancreatic tissue from only 1 of the 4 WT RT2 mice analyzed contained neoplastic lesions scored as angiogenic islets, and as such no statistical comparisons with this stage are possible.

Supplementary Figure 3. Cytokine expression in Csf1op/op RT2 tumors. qRT-PCR analyses of whole WT and Csf1op/op RT2 tumors. Tumors from independent mice (n=7 WT RT2, n=6 Csf1op/op RT2) were isolated and snap frozen in liquid nitrogen. RNA was isolated with Trizol, DNase treated, and 0.5 µg of RNA was used for cDNA synthesis. (a) Taqman probes (Applied Biosystems) for CCL3 (Mm00441258_m1), CCL4 (Mm00443111_m1), and CXCL12 (Mm00445552_m1), were used for q-PCR. Assays were run in triplicate and expression was normalized to ubiquitin C (Mm01201237_m1) for each sample. Data for individual tumors are represented as circles, and horizontal lines correspond to the mean, **P <0.005, by unpaired Student’s t-test. (b) mRNA expression data from Figure 4 and part (a) is depicted for each individual tumor for each Taqman probe set. Each mouse is assigned a number 1 through 7, and for tumors from the same mouse, individual tumors are indicated after the decimal point. Tumors from WT RT2 mice are on the left and tumors from Csf1op/op RT2 mice are on the right. The relevant Taqman probe set is marked in the upper right hand corner of each graph.

Supplementary Table 1. Relevant clinicopathological information for PNET patient samples used for CD68+ macrophage immunostaining and analysis in Figure 1 and Supplementary Tables 2, 3.

Supplementary Table 2. Comparison of clinical and histopathologic parameters that were significantly associated with increased levels of CD68 immunostaining in PNET patients shown in Figure 1.

Supplementary Table 3. Comparison of clinical and histopathologic parameters for association with presence or absence of liver metastasis in PNET patients.

4