Københavns Universitet

Transcription

1 university of copenhagen Københavns Universitet he -SOX7 axis-modulated in pericytes and stromal cells promotes metastasis through tumour-associated macrophages Yang, Yunlong; Andersson, Patrik; Hosaka, Kayoko; Zhang, Yin; Cao, Renhai; Iwamoto, Hideki; Yang, Xiaojuan; Nakamura, Masaki; Wang, Jian; Zhuang, Rujie; Morikawa, Hiromasa; Xue, Yuan; Braun, Harald; Beyaert, Rudi; Samani, Nilesh; Nakae, Susumu; Hams, Emily; Dissing, Steen; Fallon, Padraic G; Langer, Robert; Cao, Yihai Published in: Nature Communications DOI:.38/ncomms385 Publication date: 6 Document Version Publisher's PDF, also known as Version of record Citation for published version (APA): Yang, Y., Andersson, P., Hosaka, K., Zhang, Y., Cao, R., Iwamoto, H.,... Cao, Y. (6). he -SOX7 axis-modulated in pericytes and stromal cells promotes metastasis through tumour-associated macrophages. Nature Communications, 7, 385. DOI:.38/ncomms385 Download date:. Sep. 8

2 ARICLE Received Jan 6 Accepted Mar 6 Published 6 May 6 DOI:.38/ncomms385 OPEN he -SOX7 axis-modulated in pericytes and stromal cells promotes metastasis through tumour-associated macrophages Yunlong Yang, *, Patrik Andersson, *, Kayoko Hosaka, Yin Zhang, Renhai Cao, Hideki Iwamoto, Xiaojuan Yang, Masaki Nakamura, Jian Wang, Rujie Zhuang, Hiromasa Morikawa 3, Yuan Xue,, Harald Braun 5,6, Rudi Beyaert 5,6, Nilesh Samani 7,SusumuNakae 8, Emily Hams 9, Steen Dissing, Padraic G. Fallon 9, Robert Langer & Yihai Cao,7, Signalling molecules and pathways that mediate crosstalk between various tumour cellular compartments in cancer metastasis remain largely unknown. We report a mechanism of the interaction between perivascular cells and tumour-associated macrophages (AMs) in promoting metastasis through the S-dependent pathway in xenograft mouse models of cancer. is the highest upregulated gene through activation of SOX7 transcription factor in -stimulated pericytes. Gain- and loss-of-function experiments validate that promotes metastasis through recruitment of AMs. Pharmacological inhibition of the S signalling by a soluble S significantly inhibits AMs and metastasis. Genetic deletion of host in mice also blocks -induced AM recruitment and metastasis. hese findings shed light on the role of tumour stroma in promoting metastasis and have therapeutic implications for cancer therapy. Department of Microbiology, umor and Cell Biology, Karolinska Institute, 7 77 Stockholm, Sweden. he CM Hospital of Zhejiang Province, Hangzhou, Zhejiang 36, China. 3 Unit of Computational Medicine, Department of Medicine, Center for Molecular Medicine, Karolinska Institute, 7 77 Stockholm, Sweden. Institute for Medical Engineering and Science, Massachusetts Institute of echnology, Cambridge, Massachusetts 39, USA. 5 Department of Biomedical Molecular Biology, Ghent University, B-95 Ghent, Belgium. 6 Unit of Molecular Signal ransduction in Inflammation, Inflammation Research Center VIB, B-95 Ghent, Belgium. 7 Department of Cardiovascular Sciences, University of Leicester and NIHR Leicester Cardiovascular Biomedical Research Unit, Glenfield Hospital, Leicester LE3 9QP, UK. 8 Laboratory of Systems Biology, Center for Experimental Medicine and Systems Biology, he Institute of Medical Science, he University of okyo, okyo 88639, Japan. 9 School of Medicine, rinity Biomedical Sciences Institute, rinity College Dublin, College Green, Dublin, Ireland. Department of Cellular and Molecular Medicine, Panum Institute, University of Copenhagen, N Copenhagen, Denmark. Department of Medicine and Health Sciences, Linköping University, Linköping, Sweden. * hese authors contributed equally to this work. Correspondence and requests for materials should be addressed to Y.C. ( yihai.cao@ki.se). NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms385

3 ARICLE NAURE COMMUNICAIONS DOI:.38/ncomms385 Cancer metastasis is a complex process that involves in sophisticated interactions between malignant and host cells,. Cancer cells often produce signalling molecules to manipulate host cells in the local microenvironment to facilitate their invasion, dissemination and metastasis. he PDGF-PDGFR signalling often becomes activated in the tumour microenvironment 35 and endothelial cells in angiogenic vessels are an important source for the production of 6, a pluripotent member in the PDGF family. In epithelial cell- and other cell-originated cancer types, primarily targets stromal fibroblasts and perivascular cells including pericytes and vascular smooth muscle cells 7. stimulates the proliferation and migration of perivascular cells through activation of PDGFRb although interaction with PDGFRa also occurs in fibroblasts 5,7. Although it is well known that modulates vascular remodelling and maturation by recruiting pericytes and vascular smooth muscle cells onto angiogenic vessels, activation of these perivascular cells in the tumour microenvironment in cancer invasion and metastasis is poorly understood. umour tissues often contain an exceptionally high number of inflammatory cells, which significantly alter tumour growth, angiogenesis, metastasis and drug responses 8,9. Inflammatory cytokines including GM-CSF, NF-a, IL-b, IL-6 and various chemokines are actively involved in recruitment of inflammatory cells in tumours,. However, our current understanding of recruitment of tumour-associated macrophages (AMs) and their roles in cancer invasion and metastasis are far from complete. as a relatively new cytokine belongs to IL- family and it can be produced by a broad range of cell types including fibroblasts, osteoblasts, endothelial cells, epithelial cells and adipocytes 5. exerts its biological functions through binding and activation of its receptor S, a member in the olllike receptor superfamily. is known to regulate h immune responses. However, the role of in tumour inflammation and metastasis is unknown. A recent study shows that in a mouse breast cancer model, injection of protein stimulates primary tumour growth and metastasis 6. In the present study, we show that is the most upregulated gene in -stimulated pericytes and SOX7 transcription factor mediates -induced expression. Gain-of-function and loss-of-function experiments demonstrate that pericyte- and stromal cell-derived is a crucial cytokine for recruitment of AMs in the tumour microenvironment. Importantly, in several human and mouse graft tumour models, we provide compelling evidence to demonstrate that pericyte- and stromal cell-derived -activated AMs are crucial for cancer metastasis. Finally, in in vivo tumour models, we show that -activated AMs mediate -induced cancer metastasis. hese findings shed new mechanistic lights on the crosstalk between various host cellular compartments and -stimulated pericytes in promoting cancer metastasis. Functional blocking of the --AM axis is an important approach for cancer therapy. Results -PDGFRb signalling indirectly recruits AMs. o investigate the role of in the recruitment of AMs, we screened a panel of human tumour cell lines that spontaneously express. We have found that human A3 squamous carcinoma cell line expressed a high level of endogenous PDGF- BB protein (5 pg ml ) (Fig. a). he A3 xenograft tumour contained a high number of Iba þ AMs (Fig. b). Interestingly, downregulation of by Pdgfb-specific shrna, which effectively inhibited the Pdgfb mrna level (Supplementary Fig. a), markedly ablated AMs in tumour tissues (Fig. b), suggesting that was primarily responsible for AM recruitment in this human xenograft model. o further validate these findings, we performed gain-of-function experiments in which mouse Lewis lung carcinoma (LLC) and fibrosarcoma were transfected with Pdgfb-retrovirus to stably express (Supplementary Fig. b and c). ShRNA knockdown of Pdgfb significantly inhibited A3 tumour growth (Supplementary Fig. d), whereas expression promoted tumour growth in and LLC tumours (Supplementary Fig. e and f). Notably, FACS and immunohistochemical analyses showed that -LLC and tumours contained significantly higher numbers of F/8 þ and Iba þ AMs as compared with their respective vector-transfected tumours (Fig. c,d). Of note, Iba and F/8 double immunostaining showed completely overlapping positive signals (Supplementary Fig. g), indicating that both markers detect the total macrophage population in tumour tissues. hese findings demonstrate that recruits AMs in human and mouse cell line-derived graft tumour models. o define PDGFRs that are responsible for AM recruitment, we used various PDGFR inhibitors. Imatinib, a pan PDGFR tyrosine kinase inhibitor 7, significantly inhibited AM recruitment in A3, LLC and tumours (Fig. e), suggesting that PDGFRs mediate -induced AM infiltration. o distinguish PDGFRa and PDGFRb signalling in AM recruitment, anti-mouse PDGFRa- and PDGFRb-specific neutralizing antibodies (PDGFR blockades) were used for the treatment of þ tumours. Interestingly, PDGFRb,but not PDGFRa, blockade, markedly inhibited -induced AM infiltration (Fig. e). hese findings indicate that PDGFRb is the receptor that mediates -induced AM recruitment. We next investigated the direct versus indirect role of PDGF- BB in the recruitment of AMs. Surprisingly, co-localization of PDGFRa and PDGFRb in tumours by their specific antibodies showed that AMs lacked PDGFR expression (Fig. f), suggesting an indirect role of in the recruitment of AMs. Consistent with this notion, PDGFRb was primarily localized in non-am cells including NG þ pericytes and asma þ smooth muscle cells (SMCs)/myofibroblasts (Fig. g). hese findings were further quantitatively validated by PCR with reverse transcription (RPCR), quantitative PCR (qpcr) and staining of various cell lines showing that stromal fibroblasts and pericytes expressed high levels of Pdgfrb mrna, whereas mouse Raw macrophage-like cell line and isolated AMs completely lacked Pdgfrb mrna expression (Fig. h and Supplementary Fig. h). hese findings further support our notion that recruits AMs in various tumour models through an indirect mechanism. induces pericyte- and fibroblast-derived. o identify possible genes that mediate -induced AM recruitment, we performed a genome-wide expression microarray analysis in -stimulated pericytes. Surprisingly, Il33 was the most upregulated gene product with more than an eight-fold increase among all the genes in the genome (Fig. a), and was the top one of the upregulated inflammatory cytokines (Fig. b). he -induced expression was further validated by qpcr, which revealed more than a five-fold increase of Il33 mrna expression (Fig. c). In contrast, PDGF-AA, a ligand that only binds to PDGFRa, lacked ability to induce Il33 expression (Fig. c), indicating that PDGFRb is responsible for induced Il33 expression. In addition to pericytes, stimulation of PDGFRb þ stromal fibroblasts with also led to marked upregulation of Il33 mrna (Fig. d). We further analysed the NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms385

4 NAURE COMMUNICAIONS DOI:.38/ncomms385 ARICLE protein expression from -stimulated pericytes and stromal fibroblasts. Again, the protein levels in PDGF- BB-stimulated pericytes and fibroblasts were significantly higher than those in non-stimulated cells (Supplementary Fig. a). o validate these findings in vivo, we analysed protein expression in þ tumours and found a marked increase of expression as compared with vector tumours (Fig. e). protein expression levels in A3 tumour grafts were markedly decreased by the Pdgfb-specific shrna (Fig. e). We provided further in vivo evidence by delivery of adenoviral Pdgfb (Adv-Pdgfb) into tumour-free mice. Again, delivery of Adv- Pdgfb significantly induced expression in the hepatic tissue (Fig. f). Collectively, these in vitro and in vivo findings provide compelling evidence that markedly induces expression in PDGFRb þ perivascular cells and stromal fibroblasts. o further validate the pericytes and stromal cells as the major source of production in in vivo tumours, we isolated different cell types from the tumour microenvironment. We confirmed that the PDGFRb þ cell population including stromal fibroblasts and perivascular cells were the important cells for the production of in response to (Fig. g). Furthermore, NG þ pericytes in tumours produced high levels of as compared with those isolated from the vector control tumours (Fig. g). In contrast, CD3 þ vascular endothelial cells did not significantly contribute to induced expression in tumours since levels in PDGF- BB positive population was not increased (Fig. g). Similarly, tumour cells produced negligible levels of in positive and -negative tumour cells, which remained unchanged. aken together, these findings demonstrate that pericytes and tumour stromal cells are the primary source of in the tumour microenvironment. We treated -stimulated pericytes with PDGFRa and PDGFRb blockades to monitor expression in vitro. PDGFRb, but not PDGFRa, specific blockade significantly inhibited -induced expression in pericytes (Fig. h). he combination of PDGFRb and PDGFRa blockades did not produce any additive effects. Similar to PDGFRb blockade, imatinib also produced a markedly inhibitory effect on expression (Fig. h). Likewise, PDGFRb blockade also significantly inhibited -induced expression in stromal fibroblasts (Supplementary Fig. b). Role of signalling pathways in production. Signalling pathway analysis showed that induced activation of PDGFRb by phosphorylation (Fig. 3a) and has no impact on activation of PDGFRb in pericytes. In concordance with the activation of PDGFRb, downstream signalling components including MAP kinase (Erk) and Akt also became activated in -stimulated pericytes (Fig. 3b). Signalling network analysis from cbioportal 8 showed that Akt and MAPK were correlated with expression (Supplementary Fig. 3a). Consistently, MAPK and Akt-specific inhibitors significantly and effectively inhibited Il33 mrna expression levels in stimulated pericytes (Fig. 3c and Supplementary Fig. 3b). hese findings show that induces expression in pericytes through activation of the PDGFRb signalling pathway. SOX7 mediates -induced expression. We next investigated potential mechanisms by which induces expression in PDGFRb þ pericytes and fibroblasts. Genome-wide microarray analysis of -stimulated pericytes revealed that SOX7 was the most upregulated transcription factor (about six-fold; Fig. 3d), which was ranked as the top three most upregulated gene products in the genome (Fig. a and Supplementary Fig. 3c). he qpcr analysis further validated the increased expression level of Sox7 mrna in -stimulated pericytes (Fig. 3e). Notably, PDGFRb-specific blockade significantly attenuated -stimulated expression of Sox7, whereas PDGFRa-specific blockade had no effect on Sox7 mrna expression (Fig. 3e). hese findings suggest that PDGFRb potentially mediates -induced Sox7 expression. o decipher the functional relation between SOX7 and expression, -stimulated pericytes were treated with Sox7-siRNA. Knockdown of SOX7 significantly impaired PDGF- BB-induced expression (Fig. 3f), which was correlated to the knockdown efficiency (Supplementary Fig. 3d). Similarly, Sox7-siRNA knockdown also markedly reduced production in -stimulated stromal fibroblasts (Supplementary Fig. 3e). o provide further supportive evidence of transcriptional regulation of expression by SOX7, we analysed mouse sequences of the promoter region and discovered a canonical SOX7-binding SRY box and five non-canonical binding sites (Fig. 3g). Chromatin immunoprecipitation (ChIP) assay using the Il33 promoter fragment containing the canonical binding site demonstrated that SOX7 directly bound to the Il33 promoter (Fig. 3h). However, it is possible that the non-canonical SOX7 binding sites might also mediate direct binding of SOX7. hese data show that induces expression through the PDGFRb-SOX7 signalling pathway. Signalling mechanisms of -induced Raw cell migration. As was the most upregulated cytokine in -stimulated pericytes, we investigated its functional impact on macrophages. FACS and RPCR analyses showed that Raw macrophage-like cell line expressed S receptor that mediates biological functions of (Fig. a). As a negative control, stromal fibroblasts lacked a detectable level of S expression (Fig. a). Knowing that macrophages expressed S receptor, we studied the functional impact of on macrophages. Recombinant stimulated Raw cell migration in a dose-dependent manner (Fig. b,c). In addition, induced activated morphological changes of Raw cells that manifested an elongated cell shape (Fig. d). Importantly, the St-siRNA knockdown significantly ablated the - induced Raw cell migration and morphological changes (Fig. c,d). o validate the biological effects of pericyte-derived in modulation of macrophage activities in vitro, stimulated and non-stimulated pericytes were co-cultured with Raw cells. In this co-culture assay, -stimulated pericytes induced elongated Raw cell morphological changes, which was neutralized by a soluble S receptor (Fig. e). Similarly, PDGF- BB-primed pericytes also attract Raw cell motility in a co-culture system (Fig. f). Consistent with the above-mentioned biological functions, IL- 33 stimulation of Raw cells induced marked activation of MAPK, which became hyper-phosphorylated (Fig. g). In addition, stimulated phosphorylation of p38 that is known in the regulation of cellular actin reorganization and cell morphological changes 9 (Fig. g). Notably, stimulation led to potent activation of IkBa, which became highly phosphorylated (Fig. g). However, Akt levels in -stimulated and non-stimulated cells remained unchanged (Supplementary Fig. a). o functionally link the S receptor and -activated intracellular signalling components, we used the St-siRNA knockdown technique. -induced MAPK, p38 and IkBa phosphorylation in Raw cells were largely inhibited by the St-specific sirna (Fig. h), indicating that induces the S-dependent activation of these intracellular signalling components in macrophages. In contrast to, lacked abilities to activate MAPK, p38 and IkBa, NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms

5 ARICLE NAURE COMMUNICAIONS DOI:.38/ncomms385 supporting the conclusion that does not directly act on macrophages. o define the biological functions of the -S-activated downstream signalling molecules, we treated the -stimulated Raw cells with various inhibitors that blocked the activation of a specific signalling component. As expected, the MAPK, p38 and IkBa inhibitors effectively blocked -stimulated phosphorylation of these intracellular signalling molecules (Supplementary Fig. b). reatment with a known MAPK inhibitor (U6) completely abolished -induced Raw cell a Expression of h (pg ml ) c e A3 LLC- Iba/DAPI Iba/DAPI Iba/DAPI F/8 Number of F/8 + cells per field h 5 3 U5 A59 3 P <.5 P <. P <. SF R P <. MDA-MB-3 A3 OVCAR8 CAKI- Number of F/8 + cells per field P <. P <. 3 3 PC GFP 3 3 R Macrophage P <. Iba + area per field ( 3 μm ) Relative expression of Pdgfrb NS Iba + area per field ( 3 μm ) P < A3 Iba/GFP LLC Iba + area per field ( 3 μm ) Iba/GFP Iba/GFP P <. P <. P <. Scramble control SF PC Macrophage Isolated AM LLC F/8/PDGFRα F/8/PDGFRβ NG/PDGFRβ αsma/pdgfrβ Pdgfb shrna Iba + area per field ( 3 μm ) Iba + area per field ( 3 μm ) Iba + area per field ( 3 μm ) P <. A3 Scramble control P <. A3 Pdgfb shrna LLC- LLC- LLC- LLC- P <. Imatinib 5 A3 LLC- Imatinib Imatinib PDGFRβ 5 μm 5 μm F/8 + cells (%) P <.5 P <. Imatinib A3 Anti-PDGFRα 5 5 b d P <. NS P < Anti-PDGFRβ P <. Imatinib LLC- Imatinib Anti-PDGFRα Anti-PDGFRβ 5 μm f g 5 μm 5 μm 5 μm 5 μm 5 μm NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms385

6 NAURE COMMUNICAIONS DOI:.38/ncomms385 ARICLE migration and cell shape changes (Fig. i,j). hese findings reconcile with the known functions of MAPK signalling. Similarly, p38 inhibitor also effectively inhibited -induced migration and cell shape changes of Raw cells (Fig. i,j). he treatment of -stimulated Raw cells with an NF-kB inhibitor also significantly attenuated -induced Raw cell migration and cell shape changes (Fig. i,j). hese findings demonstrate that displays direct effects on Raw cell migration and activation through MAPK, p38 and IkBa signalling pathways. increases AMs in tumours. We performed in vivo gainof-function experiments that allowed studying the effect of on tumour growth, AM recruitment and tumour invasion and metastasis. Genetic propagation of tumours with by lentiviral approach led to overexpression of protein in tumour tissues (Fig. 5a). Despite a high level of expression, -overexpressing tumours grew at similar rates in vitro and in vivo as vector-transfected control tumours (Supplementary Fig. 5a and b). However, FACS analysis showed that the number of F/8 þ AMs in -tumours was markedly increased as compared with that of vector control tumours (Fig. 5b). o further define the AM phenotypes in tumours, we implanted positive tumours in wt and St / knockout mice 3 and isolated AMs from these tumours. he isolated primary AMs from wt and St / backgrounds were subjected to genome-wide affymetrix analysis. Intriguingly, classical M markers including Cd6 (Mrc), Cd63, Pdl (Pdcdlg), Ccr3, Arg and many others were all markedly downregulated in AMs isolated from the St / background as compared with those isolated from tumours grown in wt mice (Fig. 5c and Supplementary Fig. 5c). We further investigated subpopulations of AMs using M markers by FACS analysis in vector- and - overexpressed tumours. For defining the M population of macrophages, three independent cell surface markers including CD6, CCR3 and PDL were used in our FACS analysis. Altogether, three independent analyses show that F/8 þ CD6 þ, F/8 þ CCR3 þ and F/8 þ PDL þ M subpopulations of macrophages were significantly increased (Supplementary Fig. 5df). However, the F/8 þ CD6 subpopulation was also significantly increased. -primed macrophages promote metastasis. One of the important characteristics of M macrophages is tumour promotion through various processes like metastasis. Since our study indicated increased tumour metastasis linked to, we investigated further the link between -induced AMs and tumour metastasis. AMs have been described to facilitate tumour cell invasion, intravasation and dissemination,. o functionally link -primed AMs with cancer invasion, we performed in vitro matrigel cancer invasion experiments in which tumour cells and macrophages were co-embedded in matrigel as spheroids. -treated and non-treated macrophages were mixed with GFP þ LLC tumour cells and spreading of GFP þ cells was quantitatively measured. -stimulated macrophages, but not non-stimulated cells significantly promoted cancer cell invasion in this in vitro invasion assay (Fig. 5d). hese findings show that -primed macrophages promote cancer cell invasion and possibly metastasis. We next analysed circulating tumour cells (CCs) in and vector tumour-bearing mice. Interestingly, a significantly higher number of CCs were found in -tumour-bearing mice as compared with vector tumour-bearing mice (Fig. 5e). o provide further evidence of -induced metastasis, we developed an independent metastasis model in which luciferase-expressing primary tumours were implanted in the liver of each mouse. Although no differences of primary tumour growth were observed (Fig. 5f), a higher number of -tumour-bearing mice developed luciferase þ pulmonary metastasis as compared with the vector control group (Fig. 5g). hese findings were consistent with increased CCs in -tumour-bearing mice and showed that induced tumour cell intravasation and metastasis without affecting primary tumour growth. he -S-AM axis-dependent metastasis. o mechanistically link AMs and cancer metastasis of -tumours, we used clodronate 5 as macrophage-ablating agent to deplete AMs. Expectedly, clodronate effectively ablated AM numbers in both vector and -tumours (Fig. 6a). In contrast, the treatment of tumour-bearing mice with the control liposome did not significantly affect macrophage numbers in tumours as compared with controls. Approximately 5% of tumour-bearing mice possessed visible pulmonary metastatic nodules on the surface of their lungs at week 6 after removal of primary tumours (Fig. 6b). Conversely, the vector tumour control group had a lower rate of pulmonary metastasis. he lung metastatic lesions were validated by detection of red fluorescent protein (RFP; Fig. 6b). he depletion of AMs by clodronate markedly decreased the metastatic incidences in the lungs of tumour-bearing mice, whereas the low metastatic rate in vector-tumour bearing mice remained unchanged. hese findings show that -promoted pulmonary metastasis is dependent on AMs and the low metastatic incidence in control tumour-bearing mice might be mediated through a AMindependent mechanism. o further investigate the AM-mediated metastatic potentials, we used a zebrafish metastasis model 68 that allowed detection Figure induces PDGFRb-dependent macrophage recruitment in tumour cell line grafts tumours of human and mouse origin. (a) Expression levels of in conditioned medium of various human tumour cell lines (n ¼ 3 samples per group). (b) Iba þ macrophages (red) in human scrambled shrna-a3 and Pdgfb shrna-a3 squamous carcinomas cell line grafts. Arrowheads indicate tumour-infiltrating macrophages. Scale bar, 5 mm. Iba þ AMs were quantified as areas (n ¼ 8 random fields per group). (c) Upper panels: FACS analysis of F/8 þ AMs in vector- and - fibrosarcoma cell line grafts and quantification of percentages of F/8 þ macrophages (n ¼ 6 samples per group). Lower panels: quantification of F/8 þ AMs of immunohistochemical micrographs (n ¼ 8 random fields per group). (d) Iba þ macrophages (red) in vector- and -, and vectorand -LLC tumours. he tumour cells express GFP (green). Arrowheads indicate tumour-infiltrated macrophages. Scale bar, 5 mm. Iba þ AMs were quantified as areas (n ¼ 8 random fields per group). (e) Iba þ macrophages (green) in vehicle- or imatinib-treated A3 and -LLC tumours, and in vehicle-, imatinib-, anti-pdgfra- or anti-pdgfrb-treated - tumours. issue sections were counter-stained with DAPI (blue). Arrowheads indicate tumour-infiltrated macrophages. Scale bar, 5 mm. Iba þ AMs were quantified from eight random fields per group. (f) PDGFRb (green) and F/8 (red), or PDGFRa (green) and F/8 (red) double immunostaining of A3, - and -LLC tumours. White arrowheads indicate F/8 þ macrophages and yellow arrowheads point to PDGFRa þ or PDGFRb þ cells. Scale bar, 5 mm. (g) asma (green) and PDGFRb (red), or NG (green) and PDGFRb (red) double immunostaining of vector- and - tumours. Arrowheads indicate double-positive signals (yellow). Scale bar, 5 mm. (h) RPCR and qpcr analyses of Pdgfrb in various cell types. Beta-actin was used as a standard loading (mean±s.e.m., NS, not significant, Student s t-test). NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms

7 ARICLE NAURE COMMUNICAIONS DOI:.38/ncomms385 a b c. :. Il33 Mcpt8 Sox7 Sprra Lrp8 Kif Spc5 Il3ra Pbk Mfap3l Smoc Fosb Fmod Gdpd Ogn Cxcl3 Omd Aspn Slca6 Prss35. :. Il33 Spp Mif Il6 Cxcl Vegfa Cx3c Ilr Ccl7 Il5 Il7 Il8rb Il5 nfsf3 Ilrb Cxcl Cxcl5 nfsf3b Ilrn Cxcl3 d Relative expression of Il33 Relative expression of Il P <. NS PDGF-AA PC P <. SF e.5.5 P <.5 P <.5 f 6 P <.5 Expression of (ng ml ).5 Expression of (ng ml ).5 A3- A3- Scramble Pdgfb control shrna Liver GFP Adv-Gfp 5 μm Relative expression of Il33 3 Adv- Gfp Adv- Pdgfb g Relative expression of Il P<. P< CD3 + PDGFRβ + NG + CD3 + PDGFRβ + NG + h Relative expression of Il Anti-PDGFRα Anti-PDGFRβ Imatinib P < P <. P <. P <. NS + + PC Figure -PDGFRb-signalling induces expression. (a) Heatmap of top most upregulated and downregulated genes by genome-wide expression profiling of -stimulated lung pericytes cultured in vitro. (b) Heatmap of top most upregulated and downregulated inflammationrelated signalling molecules by genome-wide expression profiling of -stimulated lung pericytes cultured in vitro. (c) qpcr quantification of Il33 mrna expression levels in PDGF-AA- or -stimulated lung pericytes cultured in vitro. (PC; n ¼ 6 samples per group). NS, not significant. (d) qpcr quantification of Il33 mrna expression levels in -stimulated bone marrow stromal fibroblasts cultured in vitro (SF; n ¼ 6 samples per group). (e) Quantification of mouse protein levels of vector- and - tumours, and scrambled and Pdgfb shrna-transfected A3 tumours (n ¼ 6 samples per group). (f) qpcr quantification of Il33 mrna of Adv-Gfp- and Adv-Pdgfb-infected liver tissues (n ¼ 6 samples per group). Adv-Gfp-infected hepatocytes were visualized by a fluorescent microscope. (g) qpcr quantification of Il33 mrna expression levels in CD3 þ, PDGFRb þ and NG þ cell populations isolated from vector and tumours (n ¼ 6 samples per group). and tumour cells served as controls. (h) qpcr analysis of Il33 mrna of vehicle-, anti-pdgfra-, anti-pdgfrb- or imatinib-treated -stimulated or non-stimulated lung pericytes cultured in vitro. (PC; n ¼ 6 samples per group; mean±s.e.m., NS, not significant, Student s t-test). of the interactions between malignant cells and macrophages at the single-cell level. his zebrafish metastasis model also permits kinetic monitoring of tumour cell invasion and metastasis in the living fish body. Moreover, the availability of certain genetic strains such as transgenic Fli:EGFP zebrafish 9,3 allows us to study the event of tumour cell intravasation with or without 6 NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms385

8 NAURE COMMUNICAIONS DOI:.38/ncomms385 ARICLE a d 5 min 3 min ppdgfrβ PDGFRβ b 5 min 3 min Phosopho-p/ MAPK (Erk) p/ MAPK (Erk) Phospho-Akt Akt c Relative expression of Il33 3 P <. P <.5 P <. U6 Akti-/. :. Sox7 Ef8 Etv5 cf9 Foxm Ef7 Zfp367 Fosl Etv Mdc Arnt Klf Aff3 cea3 Mafb Spib Klf7 Nra Fos Fosb e Relative expression of Sox7 3 Anti-PDGFRα Anti-PDGFRβ P <. NS P <. P < PC f Relative expression of Il P <. P <. + + sisox7 + Scramble + PC g PDGFRβ p p SOX7 CG G II33 Promoter II33 Exon Coding region >8% SRY-box % SRY-box Primer for ChIP assay Exon Exon 3 h ChIP efficiency (% of input) 8 P <.5 NS P <.5 IgG Control Il33 promoter Anti-Sox7 pull-down Figure 3 SOX7 transcription factor mediates the -PDGFRb-induced expression. (a) Western immunoblotting analysis of phosphorylation of PDGFRb of vehicle-, - and -treated lung pericytes cultured in vitro. Beta-actin indicates loading levels. (b) Western immunoblotting analysis of Erk phosphorylation, Akt phosphorylation of vehicle-, - and -treated pericytes. Beta-actin indicates loading levels. (c) qpcr analysis of Il33 mrna of vehicle-, U6- or Akti-/-treated -stimulated or non-stimulated lung pericytes cultured in vitro. (PC; n ¼ 6 samples per group). NS, not significant. (d) Heatmap profiling of transcription factor gene expression of vehicle- and -stimulated lung pericytes cultured in vitro. (e) qpcr analysis of Sox7 mrna expression levels of anti-pdgfra- or anti-pdgfrb-treated lung pericytes that received stimulation (n ¼ 6 samples per group). -treated pericytes served as controls (n ¼ 6 samples per group). NS, not significant; PC, pericyte. (f) qpcr analysis of Il33 mrna expression levels of vehicle- or -stimulated lung pericytes that were transfected with scrambled or Sox7 sirna (n ¼ 6 samples per group). PC, pericyte. (g) Schematic diagram of expression in pericytes regulated by the -PDGFRb signalling through SOX7 transcriptional regulation. -activated PDGFRb induces SOX7 that targets the SRY boxes located in the Il33 promoter. (h) ChIP assay of SOX7 binding to the Il33 gene promoter. Non-immune IgG and Il33 coding region served as controls (n ¼ 6 samples per group) (mean±s.e.m., NS, not significant, Student s t-test). Full-gel images for a,b are shown in Supplementary Fig. 9. co-injection of macrophages. Interestingly, the implantation of tumour cells alone in the perivitelline space did not significantly display high dissemination (Supplementary Fig. 6a and b). However, co-implantation of tumour cells and macrophages resulted in massive tumour cell dissemination from the primary sites and distal metastasis. Interestingly, a substantial number of metastatic tumour cells in distal regions of the zebrafish body including the head and truck regions were coupled with co-injected macrophages, suggesting that tumour cells hijacked stimulated macrophages for intravasation and dissemination. he injected macrophages in vector control tumours also significantly, albeit modestly, promoted tumour cell dissemination (Supplementary Fig. 6a and b). In another experimental setting, macrophages were stimulated with NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms

9 ARICLE NAURE COMMUNICAIONS DOI:.38/ncomms385 protein in vitro before implantation with tumour cells in vivo. Again, the -educated macrophages significantly increased tumour cell invasion in this zebrafish model (Supplementary Fig. 6c and d), indicating that -activated macrophages play a critical role in cancer metastasis. o encircle the functional loop between the - PDGFRb and -S signalling pathways in the AMassociated cancer invasion and metastasis, we took both pharmacological and genetic approaches to execute the loss-of-function experiments in tumours. For a F/8 Macrophage F/8 + lgg F/8 + S SF Macrophage S b Relative fold change of migration P <. P <. P <. ng ml ng ml ng ml c Relative fold change of migration P <. P <. S Scramble SiSt d Phalloidin/DAPI PBS Scramble SiSt 5 μm e Phalloidin/F/8/DAPI Macrophage + PC Macrophage primed PC primed PC primed PC + soluble S PC PC PC 5 μm f Relative fold change of macrophage migration P <. P <. ss + Co-culture with PC g Phospho- p/ MAPK (Erk) p/ MAPK (Erk) Phospho- p38 Phospho- lκbα lκbα h Scramble SiSt ll-33 Phospho- p/ MAPK (Erk) p/ MAPK (Erk) p38 i Relative fold change of migration j P <. P <. P <. P < U6 SB 358 Withaferin A U6 SB 358 Withaferin A Phospho- p38 p38 Phospho- lκbα lκbα ll-33 Phalloidin/DAPI PBS 5 μm 8 NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms385

10 NAURE COMMUNICAIONS DOI:.38/ncomms385 ARICLE pharmacological blocking of functions, we treated PDGF- BB and vector tumour-bearing mice with a soluble S receptor, which have been used to effectively block functions in other experimental settings 3. Notably, treatment of tumours with the soluble S completely blocked the -elevated Iba þ AMs that returned to the vehicle-treated control level (Fig. 6c). Similarly, -expressing LLC tumours grown in Il33 / mice also showed significantly decreased infiltration of AMs that reached to a similar level of vector tumour grown in wt mice (Fig. 6c). Furthermore, similar reduction of AMs in tumours was also seen in St / deficient mice (Fig. 6c). Collectively, these data show that the S signalling pathway mediates -recruited AMs in the tumour microenvironment. Consistent with reduction of AMs, treatment of tumours with the soluble S blocked -promoted pulmonary metastasis (Fig. 6d). o further validate these findings, we studied -promoted cancer metastasis in Il33 / mice 3. Primary tumour growth was not altered in wt and Il33 / mice (Supplementary Fig. 6e). However, tumour-bearing Il33 / mice showed attenuated metastasis as compared with tumour-bearing wt mice (Fig. 6d). he presence of pulmonary metastatic lesions was further validated by GFP positivity. hese findings provide compelling evidence that the S signalling pathway mediates -triggered cancer metastasis. Endogenous recruits AMs and promotes metastasis. o relate our findings to pathophysiological relevance, we analysed expression levels in various tumour tissues. We found that the Panc pancreatic xenograft tumour expressed endogenous at a high level (6, pg ml ) as compared with other tumour types (Fig. 7a). Surprisingly, the analysis of expression in cultured Panc cells in vitro showed only a modest expression level (o5 pg ml ), although this level was higher than other cultured tumour cells (Fig. 7a). High expression of IL- 33 in vivo tumour tissues but not in vitro cultured Panc tumour cells indicated that host cellular components in tumour tissues contributed to expression. We therefore analysed tumour tissues and found that Panc tumour tissues contained an extremely high proportion of the stromal component that constituted the majority of the tumour tissues (Fig. 7b). In contrast, other tumour tissues including those of fibrosarcoma and LLC lung cancer possessed only little stromal components (Fig. 7b). hese findings are in general agreements with pancreatic cancers that contain high stromal cellular components, which are correlated with an invasive phenotype 33. We localized PDGFRb expression in various tumour xenografts and found that the Panc tumour expressed PDGFRb at a high level as compared with other tumour types (Fig. 7b). Moreover, PDGFRb expression was restricted to stromal fibroblast components and Panc tumour cells in vitro have barely detectable levels of PDGFRb expression (Supplementary Fig. 7a), supporting the non-tumour cell expression of PDGFRb. Phosphorylation analysis showed that a substantial proportion of PDGFRb molecules were phosphorylated in Panc tumour tissue (Fig. 7c). is a known and potent ligand for the activation of PDGFRb 3. However, was barely detectable in Panc tumour cells (Supplementary Fig. 7b), suggesting an alternative mechanism for the PDGFRb activation, probably through a receptor autophosphorylation mechanism owing to the formation of receptor dimers or oligomers Consistent with the mouse fibrosarcoma model, Panc tumours also contained an exceptionally high number of AMs as compared with other tumour types (Fig. 7b), validating the causational relation between and AM recruitment. In a subcutaneous xenograft model, Panc tumour-bearing mice manifested haematogenous metastasis in several organs including lung and liver (Fig. 7d). Notably, liver metastasis was the dominant route for Panc tumour spreading, whereas pulmonary metastatic nodules were occasionally detectable (Fig. 7d). hese findings demonstrate that the Panc pancreatic tumour is a highly invasive and metastatic cancer type. AM-dependent metastasis of high tumours. o define the causational relation between AMs and Panc metastasis, Panc tumour-bearing mice were treated with clodronate to deplete AMs. Expectedly, clodronate treatment significantly ablated the total number of AMs in Panc tumour / tissues (Fig. 7e). Similarly, Panc tumours grown in Il33 mice contained a significantly less number of AMs as compared with those tumours grown in wt mice as measured by immunohistochemistry and FACS (Fig. 7e and Supplementary Fig. 7c). Consistently, significant reduction of expression in Panc tumours was detected in Il33 / mice (Supplementary Fig. 7d), supporting the fact that host cellular components are main sources of production. Importantly, both pulmonary and liver metastases were markedly inhibited in clodronate-treated and Il33 / deficient tumour-bearing mice (Fig. 7f). Finally, multiple data set network analyses of human tissues from Genemania 38 showed that expression of Pdgfrb and Il33 are positively colocalized (Supplementary Fig. 7e), supporting the existence of a regulatory pathway in humans as seen in mice. hese data further strengthen our conclusions that -induced AMs are largely responsible for metastasis. We next analysed gene expression profiles of isolated AMs from Panc tumours grown in wt and St / mice. Interestingly, Cd6 (Mrc) was among the top downregulated Figure stimulates Raw cell migration through activation of the S-intracellular signalling pathways. (a) Left panels: FACS analysis of F/8 þ S þ mouse Raw macrophage cell line (n ¼ 6 samples per group). Non-immune IgG served as a negative control. Right panel: St mrna expression in macrophages and stromal fibroblasts. (b) Dose-dependent stimulation of Raw cell migration by (n ¼ 6 samples per group). treated macrophages served as controls. (c) Inhibition of -induced Raw cell migration by a sirna specifically targeting St (n ¼ 6 samples per group). Scrambled sirna serves as a control. (d) Inhibition of -induced morphological changes of Raw cells by a sirna specifically targeting St. Scrambled sirna serves as a control. Arrowheads indicate filopodia sprouts of the -activated macrophages. Scale bar, 5 mm. (e) Cell morphologies of Raw cells co-cultured 8 h with - or buffer-stimulated lung pericytes. F/8 were shown in green (yellow overlapped with phalloidin red). Soluble S was added to block function ( ng ml ). PC, pericyte. Scale bar, 5 mm. Arrowheads indicate filopodia sprouts of the -activated macrophages. Recombinant -stimulated mouse Raw cells served as a positive control. (f) Migration of Raw cells co-cultured with - or buffer-stimulated lung pericytes in the presence of a soluble S or vehicle (n ¼ 6 samples per group). (g) induces phosphorylation of MAPK and p38 at min, and IkBa at 5 min in mouse Raw cells. Beta-actin indicates the loading level in each lane. (h) SiRNA specifically targeting St inhibited -induced phosphorylation of MAPK, p38 and IkBa in Raw cells. Beta-actin indicates the loading level in each lane. (i) Inhibition of -induced mouse Raw cell migration by MAPK, p38 and IkBa specific inhibitors (n ¼ 6 samples per group). -treated cells served as controls. (j) Inhibition of -induced Raw cell shape changes by MAPK, p38 and IkBa specific inhibitors (n ¼ 6 samples per group). -treated cells served as controls. Arrowheads indicate filopodia sprouts of the - activated Raw cells. Scale bar, 5 mm (mean±s.e.m., Student s t-test). Full-gel images for g,h are shown in Supplementary Fig. 9. NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms

11 ARICLE NAURE COMMUNICAIONS DOI:.38/ncomms385 a Expression of (ng ml ) c e.5.5 RFP P <.5 F/8 wt F/8 + St / F/ R3 b 3 3 R3 R Cd7 IIb II3a Nos lr lr nf Arg Ccr3 Cd63 Chi3I3 IIa IIa IIr IIr II6 Mrc Msr Pdcdlg gfb 3 M M d 3 GFP + C + Macrophage RFP + cells (%) R 3 Average number of migrated GFP + cells P <.5 F/8 + cells (%) P <.5 Bright field GFP h 36 h 36 h P <. μm 3 3 f -Luc -Luc Bioluminescence signal of primary tumour ( 7 ) Liver weight (g) -Luc NS NS -Luc -Luc -Luc g Bioluminescence -Luc % of mice with pulmonary metastasis 5 -Luc -Luc -Luc Figure 5 induces infiltration of M-like AMs and metastasis. (a) expression levels in vector- and - tumour xenografts (n ¼ 6 samples per group). (b) FACS analysis of the total F/8 þ macrophages in vector- and - tumour tissues (n ¼ 5 samples per group). (c) Heatmap of M and M related genes by genome-wide expression profiling of F/8 þ cells isolated from Panc tumour grafts implanted in wt and St / mice (n ¼ 3 samples per group). (d) In vitro matrigel invasion of GFP þ LLC tumours in the presence of or vehicle-stimulated macrophages (n ¼ 6 samples per group). Arrowheads point to spread GFP þ tumour cells. Scale bar, mm. (e) FACS analysis and quantification of RFP þ circulating tumour cells in the peripheral blood of vector- and - tumour-bearing mice (n ¼ 5 samples per group) at the time point of the average tumour size of.5 cm 3. (f) Bioluminescent imaging of tumour-bearing mice implanted in livers with luciferase þ vector- and - tumours. Arrowheads point to luciferase þ tumours. Quantifications of bioluminescence signals and liver weights (n ¼ 5 samples per group). NS, not significant. (g) Bioluminescent imaging of lungs of luciferase þ vector- and - tumour-bearing mice. Arrowheads point to luciferase þ metastatic nodules. Quantifications of luciferase þ pulmonary metastases (n ¼ 8 animals per group; mean±s.e.m., NS, not significant, Student s t-test). NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms385

12 NAURE COMMUNICAIONS DOI:.38/ncomms385 ARICLE genes, indicating loss of the M phenotype of macrophages in St / AMs (Supplementary Fig. 8a). Macrophage metastasisrelated genes including matrix-degradation proteases, angiogenic factors and direct tumour invasion effectors were analysed 39. Interestingly, among 56 metastasis-related gene products, many matrix metalloproteinases (MMPs) including Mmp, Mmp9, Mmp, Mmp5, Mmp9 and Mmp8 are among the top downregulated genes in St / AMs (Fig. 8a). hese findings suggest that AMs possibly promote cancer invasion through the production of MMPs. In addition, we also performed genomewide gene expression analysis of cytokines, chemokines and their receptors. However, both upregulation and downregulation of these cytokines were found (Fig. 8b). Notably, Ccl expression was not altered between St þ / þ and St / AMs. Similarly, Ccl expression is not decreased in Panc tumours implanted in St / and Il33 / mice compared with those in wt mice as validated by qpcr (Supplementary Fig. 8b). Also, chemokine receptors were not among top regulated genes between St þ / þ and St / AMs. hese data suggest that the CCL-CCR signalling is less likely involved in mediating AM-induced metastasis in our model system. However, we cannot completely exclude the possibility of involvement of chemokines and their receptors in recruiting AMs. Discussion he current work was initiated by our original surprising finding that -expressing tumours contained a high number of AMs that lacked PDGFR expression. We therefore asked a crucial mechanistic question: hrough what mechanism does recruit macrophages? he fact that monocytes and macrophages lack detectable PDGFR expression implies the existence of an indirect mechanism that underlies macrophage recruitment by in the tumour microenvironment. o uncover this indirect mechanism, we first analysed possible receptor types that mediate -induced AM recruitment and demonstrated that PDGFRb, but PDGFRa, is the crucial receptor mediating -recruited AMs. PDGFRb is primarily expressed in perivascular cells and stromal fibroblasts as epithelial LLC cancer cells completely lack PDGFRb expression 7. hus, perivascular cells and stromal fibroblasts should be the primary cell types responsible for AM recruitment. Pericytes as the main perivascular cell type often exist in tumour microvasculatures 5,7, and their functions in tumour growth, invasion and metastasis are largely unknown. Coverage of microvessels with pericytes increases maturation and stability of tumour vessels that would potentially support tumour growth. Conversely, pericyte coverage of tumour vessels may prevent intravasation of tumour cells into the circulation and thus decreases metastatic potentials. herefore, mechanisms of tumour vasculature-associated pericytes in tumour growth and metastasis may be complex and somehow paradoxical. In general, molecular mechanisms of pericyte-derived signalling molecules in modulation of the tumour microenvironment are overlooked in the field of cancer research. o date, most studies focus on characterization of signalling molecules that affect pericyte proliferation, migration and morphological changes. Unlike most other studies, we have taken a genome-wide approach to define pericyte-derived signalling molecules that affect behaviour and functions of other cellular components in the tumour microenvironment. One of the most striking discoveries of our present study is that is the most upregulated gene product in the whole genome of -stimulated pericytes. his is an unexpected discovery because is known to stimulate pericyte proliferation and migration. hus, gene products involving in cell division, motility and cytoskeleton reorganization would be expected to be within the top-listed genes of -stimulated pericytes. Further, the -induced expression is also observed in stromal fibroblasts, indicating the existence of a generally regulatory mechanism of the -PDGFRb- axis. In contrast to pericytes and stromal fibroblasts, vascular endothelial cells isolated from -positive tumours did not show elevated expression levels of. However, endothelial cells have been described as the major cellular source of in chronically inflamed tissues under other pathological conditions such as rheumatoid arthritis and Crohn s disease 3.Perhaps,the cellular sources of are different under different pathophysiological conditions. We provide mechanistic data to demonstrate that the -PDGFRb signalling pathway modulates the promoter activity through the SOX7- mediated transcriptional regulation. his seems to be a generalized regulatory mechanism existing in PDGFRb þ cells. he exceptionally high level of in -stimulated pericytes suggests the existence of a novel functional pathway since is a relatively newly identified cytokine. Despite its known functions in the regulation of immune responses,,3,3, the role of on monocytes/macrophages is relatively unexplored. Particularly, the -triggered signalling in the tumour microenvironment in relation to inflammation-associated tumour invasion is unknown. We showed that monocytes and macrophages express S receptor, which becomes activated in response to stimulation. he interaction between and S is functionally meaningful as downstream signalling components such as MAP kinase become activated, leading to macrophage migration. -induced migratory effect could be important for the recruitment of AMs in tumours from peripheral tissues such as those in surrounding tissues and peripheral blood. AMs showed a M-like phenotype characterized by expression of Cd6 (Mrc), Cd63, Pdl (Pdcdlg), Chi3i3, Arg, as well as tumour-promoting molecules involved in invasion and metastasis like MMPs. he next question is what -stimulated AMs do for tumour growth and invasion. o address this important functional issue, we have taken both gain-of-function and lossof-function approaches. Overexpression of in tumours has no impact on primary tumour growth. However, triggers extensive haematogenous metastasis, which is dependent on AMs. hese findings are in general agreement with AM functions, especially the CD6 þ M macrophages that facilitate tumour invasion and metastasis. Although AMs might affect several steps of the metastatic cascade, the - stimulated AMs are likely to increase intravasation of tumour cells into the circulation. At this early stage of metastasis, AMs may guide tumour cells to transmigrate through the vessel wall by interacting with the endothelial cells. Again, this is another example how tumour cells manipulate various host cells for invasion and metastasis. In contrast to our findings, a recent study shows that systemic injection of stimulates primary tumour growth and metastasis in a mouse tumour model 6.At this time of writing, the difference between our findings and this study is unclear. It is plausible that systemic delivery of protein in mice as shown in that study could elicit a broad immune response that favour tumour growth. Consequently, IL- 33-accelerated tumour growth rates are also coupled to increased metastasis. hus, in that study, it is unclear whether - associated metastasis is owing to large tumour sizes or other mechanisms. wo published studies also show that exhibits antitumour activity through modulation of cytolytic cells and NK cells 5,6. In addition, alarmin may also act as an immunoadjuvant to inhibit tumour growth 7. Although these findings are primarily focused on the effect of on primary NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms385

13 ARICLE NAURE COMMUNICAIONS DOI:.38/ncomms385 tumour growth, our data show that promotes metastasis through a distinct mechanism by which metastasis occurs through a primary tumour size-independent mechanism. aken together, our present work not only defines a novel mechanistic pathway of host cells in the tumour microenvironment that controls cancer metastasis, but also indicates that AMs are the primary cell types that governs the metastatic process (Fig. 8c). argeting the -PDGFRb-S signalling axis in the stromal compartment would provide a novel therapeutic option for the treatment of cancer and metastasis. a LLC Lung lba/gfp lba/gfp H&E RFP Lung lba/gfp b c d Control Control ll-33 ll-33 Clodronate LLC- ll-33 Clodronate ll-33 + wt + wt + St / Soluble S + wt wt wt Soluble S ll33 / wt 5 μm ll33 / 5 μm 5 mm 5 μm 5 μm 5 μm Iba + area per field ( μm ) % of mice with pulmonary metastasis Iba + area per field ( μm ) Iba + area per field ( μm ) 5 8 P <. P <. P <. Control Control LLC- P <. LLC- NS Clodronate Clodronate P<. P <. Soluble S wt St / P <. P <. LLC- wt Il33 / 5 mm H&E GFP 5 μm 5 μm % of mice with pulmonary metastasis 5 Soluble S wt Il33 / LLC- NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms385

14 NAURE COMMUNICAIONS DOI:.38/ncomms385 ARICLE Methods Cell culture. - and vector-transfected fibrosarcoma and LLC stable cell lines were established by a Murine Stem Cell Virus (MSCV) vector containing enhanced green fluorescent protein (GFP). he human A3 epidemoid carcinoma cell line was kindly provided by Dr Keiko Funa from the Gothenburg University, Sweden. he shrna-pdgfb-containing lentivirus (HSH856, GeneCopoeia) was amplified in 93 cells. he infected EGFP þ A3 cells were sorted by FACS and shrna efficiency was detected by qpcr. Murine pancreatic cancer cell line Panc was kindly provided by Dr Maximilian Schnurr from University of Munich, Germany. he S7 stromal cells were cultured in a Minimum Essential Medium Alpha Medium supplemented with % fetal bovine serum (FBS) 8. Mouse monocyte/macrophage RAW 6.7 cell line was kindly provided by Dr Martin Rottenberg from the Karolinska Institutet, Sweden. FACS sorting was used to isolate primary NG þ pericytes (NG antibody: Catalogue (Cat.) No. AB53, Millipore) from mouse healthy lung tissues and F/8 þ AMs (F/8 antibody: Cat No.3, Biolegend) from mouse tumours. All other cell lines were grown and maintained in Dulbecco s modified Eagle s medium (DMEM) supplemented with % FBS. All the cell lines were not authenticated after purchase or transferred from other laboratories. We routinely tested mycoplasma contaminations in all our cell lines and they were negative. Chromatin immunoprecipitation assay. Mouse primary pericytes were used for the ChIP assay, which was performed according to the manufacturer s standard protocol using an EZ-ChIP kit (Cat. No. 7-37, Millipore). In brief, the cells were fixed with % paraformaldehyde (PFA) before sonication with an agarose-blocking buffer, followed by incubation overnight with.5 mg of a non-immune sheep IgG (Cat. No. -55, Millipore) or an anti-sox7 antibody 9 (kindly provided by Dr Valerie Kouskoff, Cancer Research UK Manchester Institute, United Kingdom) per immunoprecipitation reaction. o quantitatively analyse relative levels of precipitated chromatin, quantitative PCR was used with primers directed against specific fragments of interested genes. he SRY-box containing promoter fragment of m was amplified using the following primers: forward 5 -GCAAGAAGG CAAAGCAC-3 ; and reverse 5 -AAGCGACCGCCCCCAC-3.o amplify a control fragment lacking the SRY-box consensus sequence within the coding region of, the following primers were used: Forward 5 -CACGAC GGAAACCGCAAC-3 ; and reverse 5 -AAGCCGGCCCACC-3. Fragment amplification in total input was used to adjust the enriched values after immunoprecipitation. Animals. All animal studies were approved by the North Stockholm Animal Ethical Committee. Female C57BL/6 and immunodeficient CB7/Icr-Prkdcscid/ IcrCrl (SCID) mice were provided by the breeding unit of the animal facility of Department of Microbiology, umor and Cell Biology, Karolinska Institute, Sweden. he C57BL/6-Il33 / mice were generated by Dr Susumu Nakae (University of okyo, Japan). he C57BL/6-/S / mice were generated as described 3. Zebrafish of the g(fli:egfp)y (ZFIN, Eugene) was used for metastasis assay as described 9,3. Xenograft tumour models and metastasis. Female - to 8-week-old C57/B6 or SCID mice were used. For most experiments, 6 cells per. ml tumour cells were subcutaneously injected into the middle region of the dorsal back of each mouse. In a subset of experiments, tumour cells were stably transfected with a luciferase-expressing lentiviral vector. After creating an incision.5 6 cells per.3 ml tumour cells were injected into the left liver lobe of each mouse. After tumour cell implantation, the incision was sutured. For subcutaneous tumour implantation experiments, the tumour size was measured every other day using a caliper and the tumour volumes were calculated by a standard formula: Volume ¼ Length Width Width.5 (ref. 5). For liver tumour implantation, the tumour sizes were monitored with an IVIS Spectrum C system (PerkinElmer). Briefly, tumour-bearing mice were injected with D-luciferin (5 mg kg, PerkinElmer). Luminescence positive signals were detected by IVIS Spectrum C system after min injection (PerkinElmer). Subcutaneous primary tumours were surgically removed at the approximate size of.5 cm 3.he mice were observed for 6 weeks for development of metastases. Once the mice were killed, the organs including liver and lungs were removed and surface metastases were photographed. Metastatic lesions were detected by haematoxylin and eosin (H&E) histological analysis and fluorescent microscopy. Drug treatment. For depletion of macrophages, ml control or clodronate (dichloromethylene bisphosphonate; ClodronateLiposomes, he Netherlands) were intravenously injected every days starting from 3 days before tumour implantation and continued until primary tumour removal. he mice were kept for an additional 6 weeks for the detection of metastases. For neutralization in vivo, tumour-bearing mice were daily treated by subcutaneous injection with phosphate-buffered saline (PBS) or a soluble S (ss,. mg per mouse) 3 starting from day before tumour cell injection. After surgical removal of primary tumours, the treatment was terminated and the mice continued for metastasis experiments. For specifically neutralizing PDGFRs, an anti-pdgfra (PDGFRa blockade, IH3, ImClone Inc.) or an anti-pdgfrb (PDGFRb blockade C5, ImClone Inc.) was intraperitoneally injected (.8 mg per mouse) twice per week for weeks. he tumours were collected for further experimentation. For imatinib (LC Laboratories, Woburn, MA, USA) treatment, the mice were orally administrated with imatinib (5 mg kg daily). For metastasis experiments, imatinib treatment was terminated after primary tumour removal and the experiments were continued for 6 weeks. A lethal dose of CO was used to kill the animals. Whole-mount staining. he whole-mount protocols in our laboratory were used 5,5,5. Briefly, fresh tumour tissues were fixed with % PFA at C overnight and the fixed tissues were cut into small pieces and digested with proteinase K ( mm) in a ris buffer, permeabilized with % methanol, washed and blocked overnight with 3% milk in.3% riton X- in PBS. Primary antibodies against Iba (rabbit, Cat. No. 9-97, WAKO), F/8 (Rat, Cat. No. MCA97G, AbD Serotec), F/8 (Rabbit, Cat. No. NBP-56, Novus Biologicals) and Ki67 (Rat, Cat. No , ebioscience) were incubated overnight at C, followed by washing, blocking with 3% milk and incubation with fluorescent-conjugated secondary antibodies for h at room temperature. Additional washing was performed before mounting. he stained tissues were mounted with Vectashield mounting medium (Cat. No. H-, Laboratories). Fluorescent signals were examined with a confocal microscope (Zeiss LSM5 Confocal, or Nikon C Confocal microscope) and quantitative analysis was performed with a Photoshop (CS5) software. Affymetrix gene-array analysis. he primary pericytes isolated from the lung tissues were treated with or without ng ml human for 5 days and RNA samples were prepared using an RNAeasy kit (Qiagen) and hybridized using Affymetrix. S Gene arrays. he sample preparation and analysis method for microarrays of -treated pericytes is described as follows 53. riplicates of each sample were used for gene expression analysis. Normalization and analysis for differentially expressed genes are performed using robust multi-array analysis and significance analysis of microarrays (SAM) via R statistical software packages, oligo and samr. Heatmaps were presented for up- and downregulation of gene expression using the Multiple Experiment Viewer system (version.7). Survival data and gene expression data of uterine carcinosarcoma patients and kidney renal papillary cell carcinoma patients from he Cancer Genome Atlas (CGA) is analysed for -high (above mean) and -low (below mean) groups. For breast cancer, the top 5% -high and lowest 5% -low groups were analysed. he statistical difference was analysed using KaplanMeier survival method followed by log-rank test. Figure 6 mediates -stimulated cancer metastasis through a AM-dependent mechanism. (a) Clodronate effectively inhibited Iba þ macrophage infiltration (green) in - tumours. umour cells were labelled with RFP (red). Arrowheads indicate AMs. Quantification of Iba þ macrophage in clodronate-treated and non-treated vector- and - tumours (n ¼ 8 random fields per group). (b) Lung metastasis in clodronatetreated and non-treated vector- and - tumour-bearing mice. Arrowheads indicate lung surface metastatic nodules. Dashed line marks the border between the RFP þ metastatic nodule and surrounding lung tissues., tumour. Quantification of percentage of animals with pulmonary metastasis on the surface of lungs (n ¼ 6 mice per group). (c) Detection of Iba þ macrophages (red) and tumour cells (green) in vehicle- and soluble S-treated PDGF- BB- tumours implanted in wt mice. Iba þ macrophages (red) were also detected in - tumours implanted in St / mice. - tumour serves as a control. he detection of Iba þ macrophages (red) and tumour cells (green) in -LLC tumours implanted in wt and Il33 / mice. -LLC tumour serves as a control. Arrowheads indicate Iba þ macrophages. Quantification of Iba þ macrophages (n ¼ 8 random fields per group). (d) Pulmonary metastasis in vehicle- and soluble S-treated - tumour-bearing mice. Pulmonary metastasis in -LLC tumour-bearing wt and Il33 / mice. Metastases were detected by gross examination of lung surface, fluorescent detection for GFP þ signals and histological staining with H&E. Arrowheads indicate lung and liver surface metastases in tumour-bearing mice. Dashed lines encircle the borders between tumour nodules and surrounding tissues., tumour. Quantification of lung surface metastases in tumour-bearing mice (n ¼ 8 mice per group; mean±s.e.m., NS, not significant, Student s t-test). NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms

15 ARICLE NAURE COMMUNICAIONS DOI:.38/ncomms385 Stable expression of in tumour cell lines. he full-length complementary DNA (cdna) sequence coding human was cloned into an expression vector, plvx-ires-tdomato (Cat. No. 6338, Clontech). Briefly, 93 cells were transfected with the vector containing the cdna sequence coding for human using a Lenti-X HX Packaging System (Clontech). Murine fibrosarcoma cells were cultured with the filtered viral supernatants overnight. RFP þ a Expression of (ng ml ) 5 P <. P <. P <. P <. LLC Panc MC38 C6 Expression of (ng ml ) 5 PC + P <. P <. P <. P <. P <. P <. PC + LLC Panc MC38 C6 c LLC- LLC- Panc Phospho-PDGFRβ PDGFRβ b LLC Panc d LLC Panc H&E Lung 5 μm 5 mm PDGFRβ/ Hematoxylin μm H&E 5 μm PDGFRβ/DAPI μm Liver 5 mm Iba H&E μm 5 μm e f PDGFRβ + area per field (x μm ) Iba/GFP NS Control Control P <. P <. LLC Panc Panc-GFP Panc-GFP Iba + area per field (x μm ) Clodronate Clodronate P <.5 P <. P <. LLC Panc wt wt Panc-GFP Panc-GFP % of mice with pulmonary metastasis II33/ II33/ 5 5μm LLC Panc Iba + area per field ( μm ) 8 % of mice with hepatic metastasis 5 P <. P <. Control Clodronate wt Il33 / Panc-GFP Panc-GFP LLC Panc Lung H&E 5 mm 5 μm % of mice with pulmonary metastasis 5 Control Clodronate wt Il33 / Panc-GFP Panc-GFP Liver H&E 5 mm 5 μm % of mice with hepatic metastasis 5 Control Clodronate wt Il33 / Panc-GFP Panc-GFP NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms385

16 NAURE COMMUNICAIONS DOI:.38/ncomms385 ARICLE cells were sorted by FACS and the expression level was quantified by an ELISA (enzyme-linked immunosorbent assay) assay. Intracellular signalling inhibition and immunoblotting. Pericytes and macrophages were starved in % FBS-DMEM overnight, followed by pre-treatment with selective inhibitors against MEK/ (U6, Cat. No., ocris Bioscience), p38 (SB358, Cat. No., ocris Bioscience), IkBa (Withaferin A, Cat. No. 86, ocris Bioscience) or AK/ (AKi-/, Cat. No. ab88, Abcam) for h before stimulation with either or (5 ng ml ) for min or h, and the cell lysates were collected for protein or RNA analyses. a wt F/8 + St / F/8 + wt F/8 + St / F/8 + Fgf Mmp9 Mmp9 Pdgfc Mmp8 Mmp Mmp Fgf7 Fgf8 Mmp5 b Cxcl5 llr Ccl9 llrl ll3ra ll6st Cxcl nfrsf Ccl8 Cxcl3 nfsf3b() llrb ll8bp nfsf nfsf3b() ll7r Ccl5 Cxcl Cxcl3 llra c PDGFRβ Perivascular cell Liver Lungs umour cell SL Macrophage AM-less tumour IL-RAcP Switch to invasive tumour Blood vessel PDGFRβ Stromal fibroblast AM-rich tumour Dissemination Figure 8 Genome-wide profiling of wt and St / AMs and survival correlation of expression in grafted tumours. (a) Heatmap of top metastasis-related genes by genome-wide expression profiling of F/8 þ cells isolated from Panc tumours implanted in wt and St / mice (n ¼ 3 samples per group). (b) Heatmap of top up- and top down-most cytokines, chemokines and their receptor genes by genome-wide expression profiling of F/8 þ cells isolated from Panc tumours implanted in wt and St / mice (n ¼ 3 samples per group). (c) Schematic diagram shows in pericytes and stromal fibroblasts the -PDGFRb--S axis-recruited macrophages in switching a noninvasive tumour to a highly invasive tumour. Figure 7 -recruited macrophages mediate cancer metastasis in a pancreatic tumour model. (a) ELISA detection of protein expression levels in various xenograft tumour tissues and in cultured tumour cell lines (n ¼ 6 samples per group). -stimulated and non-stimulated pericytes served as positive controls. (b) Immunostaining of PDGFRb þ signals and Iba þ macrophages in, LLC and Panc tumours. Upper two panels were counterstained with H&E, or PDGFRb and haematoxylin. Lower middle panels were double stained with DAPI (blue). Black and white arrowheads point to PDGFRb þ signals. Red arrowheads indicate Iba þ macrophages. Quantification of PDGFRb þ signals and Iba þ macrophages (n ¼ 8 random fields per group). NS, not significant. (c) Immunoblotting detection of total and phosphorylated PDGFRb in Panc tumour tissues. - and - or - LLC tumour tissues were used as controls. (d) Pulmonary and hepatic metastasis in wt, LLC- and Panc-tumour-bearing mice. Metastases were detected by gross examination of lung surface and liver surface, and histological staining with H&E. Arrowheads indicate lung and liver surface metastases in Panc tumour-bearing mice. Dashed lines encircle the borders between tumour nodules and surrounding tissues., tumour. Quantification of lung and liver surface metastases in tumour-bearing mice (n ¼ 8 mice per group). (e) Detection of Iba þ macrophages (red) and tumour cells (green) in clodronate- and vehicle-treated Panc-GFP þ tumours implanted in wt mice. Iba þ macrophages (red) were also detected in Panc-GFP þ tumours implanted in Il33 / mice. Wild type of mice serves as a control. Arrowheads indicate Iba þ macrophages. Quantification of Iba þ macrophages (n ¼ 8 random fields per group). (f) Pulmonary and hepatic metastasis in clodronate- and vehicle-treated Panc-GFP þ tumour-bearing mice, and wt or Il33 / mice with Panc-GFP þ implantation. Metastases were detected by gross examination of lung surface and liver surface, and histological staining with H&E. Arrowheads indicate lung and liver surface metastases in tumour-bearing mice. Dashed lines encircle the borders between tumour nodules and surrounding tissues., tumour. Quantification of lung and liver surface metastases in tumour-bearing mice (n ¼ 8 mice per group; mean±s.e.m., NS, not significant, Student s t-test). Full-gel images for c are shown in Supplementary Fig. 9. NAURE COMMUNICAIONS 7:385 DOI:.38/ncomms