sign in sign up
<<
Home , PGDS, Prostaglandin E synthase

COX/mPGES-1/PGE2 pathway depicts an inflammatory- dependent high-risk neuroblastoma subset

Karin Larssona,1, Anna Kockb,1, Helena Idborga, Marie Arsenian Henrikssonc, Tommy Martinssond, John I. Johnsenb, Marina Korotkovaa, Per Kognerb,2,3, and Per-Johan Jakobssona,e,2,3

aRheumatology Unit, Department of Medicine, Karolinska Institutet, SE-171 76 Stockholm, Sweden; bChildhood Cancer Research Unit, Department of Women’s and Children’s Health, Karolinska Institutet, Astrid Lindgren Children’s Hospital, SE-171 76 Stockholm, Sweden; cDepartment of Microbiology, Tumor and Cell Biology, Karolinska Institutet, SE-171 77 Stockholm, Sweden; dDepartment of Clinical Genetics, Institute of Biomedicine, Sahlgrenska Academy, University of Gothenburg, SE-405 30 Gothenburg, Sweden; and eRheumatology Clinic, Karolinska University Hospital, SE-171 76 Stockholm, Sweden

Edited by Charles A. Dinarello, University of Colorado Denver, Aurora, CO, and approved May 27, 2015 (received for review January 16, 2015)

The majority of solid tumors are presented with an inflammatory muscles contraction, and protection of the intestinal mucosa, and in microenvironment. Proinflammatory lipid mediators including pros- pathological conditions such as autoimmune diseases and cancer

taglandin E2 (PGE2) contribute to the establishment of inflammation (6). are formed by the conversion of arachidonic and have been linked to tumor growth and aggressiveness. Here we acid to H2 (PGH2) by the show that high-risk neuroblastoma with deletion of 11q COX-1 and COX-2, followed by further processing by terminal represents an inflammatory subset of neuroblastomas. Analysis of , the prostaglandin synthases. Prostaglandin E2 (PGE2) enzymes involved in the production of proinflammatory lipid media- is a proinflammatory and immunomodulatory lipid mediator tors showed that 11q-deleted neuroblastoma tumors express high formed from PGH2 by microsomal synthase 1 levels of microsomal prostaglandin E synthase-1 (mPGES-1) and ele- (mPGES-1). Elevated levels of mPGES-1 and its enzymatic product PGE have been found in several different cancers, including colon vated levels of PGE . High mPGES-1 expression also corresponded to 2 2 cancer (7–9), nonsmall cell lung cancer (10), and prostate cancer (11, poor survival of neuroblastoma patients. Investigation of the tumor 12). PGE2 signaling contributes to increased proliferation (13, 14) microenvironment showed high infiltration of tumor-promoting mac- and invasiveness (15) of cancer cells, stimulates tumor angiogenesis rophages with high expression of the M2-polarization markers CD163 (16, 17), inhibit apoptosis (18), induces chemoresistance (19), and – and CD206. mPGES-1 expressing cells in tumors from different sub- mediates suppression of anti-tumor immunity (20, 21). Because tu- types of neuroblastoma showed differential expression of one or mor-promoting inflammation has been included as one of the several cancer-associated fibroblast markers such as vimentin, fibro- hallmarks of cancer (22), chronic inflammation and its impact on blast activation α, α smooth muscle actin, and PDGF tumorigenesis in adult tumors have been immensely investigated. β . Importantly, inhibition of PGE2 production with , a non- Less is known about the inflammatory component in child- selective COX inhibitor, resulted in reduced tumor growth in an in hood tumors; however, in a recent study, Asgharzadeh et al. (23) vivo model of 11q-deleted neuroblastoma. Collectively, these results showed that infiltration of immunosuppressive M2-polarized suggest that PGE2 is involved in the tumor microenvironment of spe- macrophages in metastatic MYCN-nonamplified neuroblastoma cific neuroblastoma subgroups and indicate that therapeutic strate- tumors contributed to the metastatic phenotype and worsened gies using existing anti-inflammatory drugs in combination with current treatment should be considered for certain neuroblastomas. Significance

mPGES-1 | PGE2 | neuroblastoma | tumor microenvironment | Cancer-related inflammation promotes progression and therapy cancer-associated fibroblasts resistance in tumors of adulthood. Knowledge concerning the significance of inflammation in childhood malignancies has been euroblastoma is the most common and deadliest tumor limited. Neuroblastoma is an embryonal tumor of early childhood Nof childhood. Although the survival of children with neuro- with poor prognosis despite intensified therapy, and biological blastoma has improved during the last decade, patients with high- understanding is necessary to develop novel therapies. We found risk disease still have a poor prognosis, despite advanced and high-risk neuroblastoma, in particular the therapy-resistant subset intensive treatments, with overall survival rates less than 40% (1). with chromosome 11q-deletion, to be inflammatory driven and The International Neuroblastoma Risk Group (INRG) classifi- characterized by high expression of the COX/microsomal prosta- cation system defines neuroblastoma risk groups as low, in- glandin E synthase-1 (mPGES-1)/prostaglandin E2 (PGE2) pathway termediate, and high, based on age at diagnosis, histology, and that correlates with metastatic stage and poor clinical outcome. genetic aberrations (2). Among high-risk neuroblastomas, ampli- We further detected infiltrating cancer-associated fibroblasts fication of the neuroblastoma MYC (MYCN)oncogeneisthe expressing mPGES-1, the essential for synthesis of PGE2, most frequent genetic aberration, seen in 30–40% of the patients. promoting tumor growth, angiogenesis, and metastatic spread. Another common genetic change in the high-risk group is deletion Treatment targeting this inflammatory pathway provides a of the long arm of chromosome 11 (11q-deletion). Deletion of therapeutic option for neuroblastoma and other cancers. 11q occurs in tumors with multiple genetic aberrations and MYCN Author contributions: K.L., A.K., M.A.H., J.I.J., M.K., P.K., and P.-J.J. designed research; K.L., A.K., chromosome instability but commonly without -amplifi- and H.I. performed research; K.L., A.K., H.I., T.M., J.I.J., M.K., P.K., and P.-J.J. analyzed data; and cation and therefore is a useful prognostic marker in adverse- K.L., A.K., J.I.J., P.K., and P.-J.J. wrote the paper. stage tumors lacking MYCN-amplification (3). These patients The authors declare no conflict of interest. are often older at disease onset and have slow disease pro- This article is a PNAS Direct Submission. gression but often develop therapy resistance and have poor Freely available online through the PNAS open access option. clinical outcome. Among low- and intermediate-risk neuro- 1K.L. and A.K. contributed equally to this work. “ ” blastomas are the so-called special neuroblastomas (4S) that 2P.K. and P.-J.J. contributed equally to this work. show a metastatic phenotype but are associated with sponta- 3To whom correspondence may be addressed. Email: [email protected] or Per-Johan. neous regression and a survival rate of 90% (1, 4, 5). [email protected]. Prostaglandins are bioactive lipids involved in many biological This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. processes both in physiological processes e.g., blood pressure, smooth 1073/pnas.1424355112/-/DCSupplemental.

8070–8075 | PNAS | June 30, 2015 | vol. 112 | no. 26 www.pnas.org/cgi/doi/10.1073/pnas.1424355112 Downloaded by guest on September 24, 2021 extracted tissue homogenates from 29 primary neuroblastoma A mPGES-1 COX-1 COX-2 40 80 8 samples (four 11q-deleted, eight MYCN-amplified, and 17 low- 6 risk tumors; Table S1). There were significantly higher levels of 30 60 2 PGE2 in the 11q-deleted tumors than in the MYCN-amplified 20 40 tumors (P = 0.02) or in the low-risk tumors (P = 3.0e-04) (Fig. 1 1C). No significant differences were found in the levels of other 10 20 Fold change Fold Fold change Fold (COX-2/HPRT) (COX-1/HPRT) Fold change Fold analyzed (Fig. S1 and Table S2). (mPGES-1/HPRT) 0 0 0 11q-NMA Low-risk 11q-NMA Low-risk 11q-NMA Low-risk Immunohistochemical Analysis of Prostaglandin Synthases. The BCexpression of enzymes involved in synthesis of prostaglandins in 100 Overall survival INSS 4 PGE2 neuroblastoma was analyzed using immunohistochemistry (IHC). Low mPGES-1 30 80 High mPGES-1 10 COX-1, COX-2, mPGES-1, lipocalin-type prostaglandin D synthase 0.3 (L-PGDS), hematopoietic-type prostaglandin D synthase (H-PGDS), 60 0.2 and 15-hydroxyprostaglandin dehydrogenase (15-PGDH) were an- 40 alyzed in tumors from 11 neuroblastoma patients (four 11q-deleted, 20 0.1 MYCN Percent survival Percent three -amplified, and four low-risk tumors; Fig. 2 and Fig. S2).

0 pmol/mg extracted 0.0 Expression of mPGES-1 was found in all tumors tested, with sig- 0 50 100 150 200 250 11q-NMA Low-risk nificantly higher levels in the 11q-deleted tumors than in the low-risk Follow up in month tumors alone (P = 0.02) or in the MYCN-amplified and low-risk tumors considered together (P = 0.004). mPGES-1 expression was Fig. 1. mPGES-1 expression and PGE2 levels are associated with 11q-deleted neuroblastoma. (A) Expression of mPGES-1, COX-1, and COX-2 mRNA in confirmed with IHC in additional 11q-deleted tumors (Fig. S3) neuroblastoma tissue samples was measured by the TaqMan Expres- and with Western blot. Western blot analysis clearly showed sion Assay. Expression of mPGES-1 and COX-1 was significantly higher in 11q- higher levels of mPGES-1 in the 11q-deleted tumors than in the deleted tumors than in low-risk tumors (P = 0.04 and P = 0.03, respectively). MYCN-amplified and low-risk tumors. COX-1 was detected in one No difference in COX-2 levels was found. (B) Analysis of an expression cohort of three MYCN-amplified tumors, and a weak band was seen in all shows high expression of mPGES-1 associated with worse overall survival in low-risk and 11q-deleted tumors. Of all tumors analyzed, COX-2 metastatic neuroblastoma (INSS stage 4). (C) PGE2 levels in neuroblastoma was detected in only one of the MYCN-amplified tumors (Fig. S3). samples as measured by LC-MS/MS were significantly higher in 11q-deleted Consistent with Western blot analysis, IHC analysis showed that tumors than in MYCN-amplified (NMA) (P = 0.02) and low-risk tumors (P = there was more expression of COX-1 than COX-2 in the neuro- 3.0e-04). Bars in A and C represent median values. blastoma tumors, and the difference between COX-1 and COX-2 expression was more evident in the 11q-deleted tumors, but no significant difference between the tumor subgroups was found for the outcome of these patients. Cancer-associated fibroblasts either COX-1 or COX-2 (Fig. 2B). Both prostaglandin D2 (PGD2) (CAFs) also have been found to contribute to tumor develop- synthases were present in significantly higher levels in low-risk ment and metastasis (24), and in a study by Zeine et al. (25) a tumors than in 11q-deleted tumors (L-PGDS, P = 6.0e-04; high number of CAFs was found to correlate with more aggressive H-PGDS, P = 0.03). Elevated levels of H-PGDS also were Schwannian stroma-poor neuroblastoma tumors. detected in MYCN-amplified tumors as compared with 11q- We previously reported the effect of anti-inflammatory drugs on P = – deleted tumors ( 0.02) (Fig. S2). The 11q-deleted tumors tumor growth in preclinical in vivo models of neuroblastoma (26 expressed significantly lower levels of 15-PGDH, an enzyme re- 28). We also have found high expression of the PGE2 receptors – sponsible for PGE2 and PGD2 degradation, than did either low- EP1 EP4 in neuroblastoma tumor tissue as well as effects of PGE2 risk tumors alone (P = 0.005) or MYCN-amplified and low-risk on neuroblastoma cell growth in vitro (29). These results suggest that tumors taken together (P = 0.008). inflammatory processes are important for neuroblastoma growth. Because of the relatively high levels of COX-1 compared with Therefore we investigated the importance of proinflammatory COX-2 in the neuroblastoma tumors, double staining of mPGES-1 prostaglandins and their enzymes in neuroblastoma subgroups. and COX-1 was performed. In one of the three tumors tested a considerable overlap between COX-1 and mPGES-1 was observed. Results In two of the three tumors only single cells expressed both en- mPGES-1 mRNA Expression in Primary Neuroblastoma. To study the zymes, but cells expressing the respective enzymes were in prox- expression levels of mPGES-1, COX-1, and COX-2 in primary imity to each other (Fig. 2C). neuroblastoma, quantitative real-time PCR analysis of tumor tissue was performed. When the relative expression was analyzed, signifi- IHC Analysis of the Tumor Microenvironment in Relation to mPGES-1 cantly higher expression of mPGES-1 mRNA was seen in the 11q- Expression. IHC analysis of mPGES-1 in neuroblastoma samples deleted tumors than in low-risk tumors alone (P = 0.04), and there showed positive staining of mPGES-1 in nontumorigenic cells, was a strong tendency toward a difference in mPGES-1 mRNA suggesting that stromal or tumor-infiltrating cells express mPGES-1 expression between 11q-deleted tumors vs. MYCN-amplified and (Fig. 2A). We therefore performed IHC analysis using an mPGES-1 low-risk tumors considered together (P = 0.06, Mann–Whitney; P = antibody in combination with antibodies detecting different immune 0.04, t test). The expression of COX-1 was significantly different cells, stromal and endothelial cells, and fibroblasts known to be part in 11q-deleted tumors and low-risk tumors (P = 0.03) and in 11q- of the tumor microenvironment (Table 1). The double staining was deleted tumors compared with MYCN-amplified and low-risk tu- performed on a tumor from one patient in each of the subgroups; mors considered together (P = 0.02). No significant differences were we designated these tumors “NB1” (untreated, MYCN-amplified, found for COX-2 (Fig. 1A). INSS stage 2, unfavorable prognosis), “NB4” (untreated, low-risk, Using publicly available expression cohorts, we analyzed INSS 4S, favorable prognosis), and tumor “NB43” (pretreated, 11q- mPGES-1 levels and its correlation to overall deleted, INSS stage 4, unfavorable prognosis). MEDICAL SCIENCES survival in patients with International Neuroblastoma Staging Disialoganglioside (GD2) is expressed almost exclusively on System (INSS) stage 4 tumors. Patients with tumors expressing tumors of neuroectodermal origin and is used as therapeutic high levels of mPGES-1 had an overall survival of only 10%, target in neuroblastoma (30). Double staining of GD2 with compared with 62% in patients with tumors expressing low levels mPGES-1 showed a considerably different staining pattern. No of mPGES-1 (P = 9.6e-04) (Fig. 1B). coexpression of GD2 and mPGES-1 was found in any of the three tumors analyzed (NB1, NB4, or NB43). The expression was Prostaglandin Profiling of Neuroblastoma Tumors. Prostaglandin mutually exclusive, with a clear compartmentalization of cells levels were measured by LC-MS/MS analysis of liquid–liquid expressing the respective (Fig. 3A).

Larsson et al. PNAS | June 30, 2015 | vol. 112 | no. 26 | 8071 Downloaded by guest on September 24, 2021 (M2-polarization), and MHC class II (MHCII; higher expression in A NMA 11q- Low-risk NB6 NB12 NB4 M1). There were significantly more M2-polarized macrophages in 11q-deleted and MYCN-amplified tumors than in low-risk tumors (see Fig. S5), suggesting a more immunosuppressive microenvi- ronment in high-risk neuroblastoma. Apart from macrophages, other myeloid cells, including my- mPGES-1 eloid-derived suppressor cells (MDSCs), are known to infiltrate tumor tissue. To cover all myeloid cells, double staining of NB1 NB12 NB9 mPGES-1 with CD11b was performed, but no colocalization was seen (Fig. S4). To determine if dendritic cells contribute to mPGES-1 ex- pression in neuroblastoma, double staining of mPGES-1 and

COX-1 CD11c was performed, but no colocalization was found. How- + ever, in the NB4 and NB43 tumors the CD11c cells were found + only in areas with mPGES-1 cells (Fig. S4). In the MYCN- NB6 NB12 NB4 amplified tumor, NB1, there was an overall pattern of weak CD11c staining, and no positive CD11c staining was detected in the + proximity of mPGES-1 cells. Next we studied lymphocyte infiltration in the proximity of

COX-2 + + mPGES-1 cells (Fig. S4). CD3 T cells were present in all tu- mors to a varying extent. The NB1 and NB4 tumors had no − + T cells in mPGES-1 areas. Tumor NB1 showed few CD3 cells + + B in mPGES-1 areas, whereas CD3 T cells were abundant 60 COX-1 20 COX-2 mPGES-1 50 in presence in the mPGES-1–expressing areas of the NB4 tumor. + + 40 15 These T cells also were CD4 .CD3 T cells were abundant 40 + 30 10 throughout the NB43 tumor and were slightly enriched in mPGES-1 20 20 areas. Interestingly, B cells were abundant only in the NB4 tumor and 5 10 only in proximity to mPGES-1–expressing cells (Fig. S4).

% stained area 0 0 0 Subsequently we studied the contribution of mPGES-1 expres- 11q-NMA Low-risk 11q-NMA Low-risk 11q-NMA Low-risk sion from endothelial cells. Double staining of mPGES-1 and the C NMA 11q-del Low-risk endothelial cell marker CD31 showed colocalization with mPGES- NB1 NB43 NB4 1–expressing cells in some areas of the NB4 tumor, but no colo- calization was detected in the NB1 or NB43 tumors (Fig. S4). To determine if the mPGES-1–expressing cells could be of mesenchymal origin, we performed double staining of mPGES-1 COX-1 + and vimentin (Fig. 3C). A majority of the mPGES-1 cells also mPGES-1 COX-1 mPGES-1 COX-1 mPGES-1 COX-1 showed positive staining for vimentin in the NB4 tumor. In the NB1 tumor the majority of cells were positive for vimentin, in- Fig. 2. COX/mPGES-1 expression in human neuroblastoma tumors. (A)IHC cluding cells expressing mPGES-1 (Fig. 3C). The NB43 tumor analysis of mPGES-1, COX-1, and COX-2 in primary human neuroblastoma showed weak vimentin staining, and only a few of these cells tumors with different biology: MYCN-amplification (NMA), 11q-deletion, coexpressed mPGES-1. + and low-risk. Sections were counterstained with hematoxylin. One repre- Because the majority of mPGES-1 cells in the NB1 and NB4 sentative image was selected from tumors from at least three patients in tumors also expressed vimentin, we performed IHC analysis with each subgroup. (B) Quantification was performed on tumors from 11 pa- markers for CAFs, i.e., fibroblast activation protein α (FAP), tients in three clinical subgroups (four with 11q-deleted tumors, three with β β NMA tumors, and four with low-risk tumors). Data are presented as mean ± PDGF receptor (PDGFR ), fibroblast-specific protein 1 (FSP-1), SEM. mPGES-1 expression was higher in 11q-deleted tumors than in low-risk α smooth muscle actin (αSMA), and PDGF receptor α (PDGFRα), tumors (P = 0.02), and there was a strong tendency toward higher mPGES-1 (Fig. 4). In the NB1 tumor there was a clear colocalization of levels in 11q-deleted tumors compared with NMA tumors (P = 0.07). No cells positive for mPGES-1 and CAF markers. Cells express- differences in COX-1 or COX-2 levels were seen. (C) Dual labeling of mPGES-1 ing mPGES-1 exhibited strong positive staining for FAP and (green) and COX-1 (red) counterstained with a nuclear dye, Hoechst (blue). αSMA and weak positive staining for PDGFRβ, whereas none of In the NB1 tumor, most cells expressing mPGES-1 also expressed COX-1. the cells were positive for PDGFRα or FSP-1 (Fig. 4). In the There also were cells expressing only COX-1. In the NB43 and NB4 tumors, NB4 tumor there was colocalization of mPGES-1 with PDGFRα, cells expressed either COX-1 or mPGES-1 with only a few points of colocal- PDGFRβ, and FSP-1 but no colocalization with FAP and αSMA ization. Arrowheads indicate cells coexpressing both enzymes. (Fig. 4). Only weak positive staining for CAF markers was detected in the NB43 tumor, and only FSP-1 showed colocali- Because macrophages are a sign of inflammation, are a typical zation with mPGES-1. source of mPGES-1 expression (31), and are present in tumor microenvironments as tumor-associated macrophages, double COX Inhibition of an in Vivo Xenograft Mouse Model of 11q-Deleted staining of mPGES-1 in combination with a pan-macrophage Neuroblastoma. The effect of COX inhibition on neuroblastoma marker (CD68) and a marker for M2-polarized macrophages tumors in vivo was assessed in an 11q-deleted neuroblastoma cell (CD163) was used. All analyzed tumors showed widespread ex- line (32) (SK-N-AS) xenograft mouse model treated with diclo- pression of CD68 throughout the sections, thus showing the pres- fenac (Fig. 5). Tumor volume was measured and compared with ence of macrophages (Fig. 3B). Interestingly, no colocalization untreated animals. A significant reduction in tumor growth was of mPGES-1 and CD68 was seen. CD163 also was expressed found for diclofenac-treated animals compared with control abundantly throughout the tumor sections, but no colocalization animals on day 8 (P = 0.01) and day 9 (P = 0.008) (Fig. 5A). between mPGES-1 and CD163 was found in any of the neuro- Endogenous levels of PGE2 were measured in the tumors using blastoma samples analyzed (Fig. S4). To consider the extent of LC-MS/MS of solid-phase extracted tumor tissue. There was a macrophage infiltration and macrophage polarization, tumors from significant decrease of PGE2 in tumors from diclofenac-treated three patients in each subgroup were stained with CD163, CD206 animals compared with controls (P = 0.04) (Fig. 5B).

8072 | www.pnas.org/cgi/doi/10.1073/pnas.1424355112 Larsson et al. Downloaded by guest on September 24, 2021 Table 1. Localization of different components of the tumor showed significantly higher levels of PGE2, the enzymatic prod- microenvironment in relation to mPGES-1 uct of mPGES-1, in the 11q-deleted subset than in either MYCN- amplified or low-risk tumors (Fig. 1C), confirming the observa- NB43, NB4, tions in the IHC analysis. In addition, significantly lower levels of Cell type Marker NB1, NMA 11q-deleted low-risk 15-PGDH, the enzyme responsible for metabolizing PGE2,were Neuroblastoma GD2 No No No detected in the 11q-deleted tumors, possibly contributing to the tumor high PGE2 levels found in these tumors (Fig. S2). All patients with Myeloid cells CD68 No No No high-risk neuroblastomas are subjected to the same induction CD163 No No No treatment before surgical excision of the tumor (Table S3) (33), CD11b No No No and all patients had at least 2 wk off chemotherapy before sur- CD11c No No No gery. Any effect of the chemotherapeutic drugs on prostaglandin T cell CD3 No No No levels would have affected the patients in both of the high-risk subgroups (i.e., patients with 11q-deleted tumors and those with CD4 No No No MYCN-amplified tumors) similarly. B cell CD20 No No No The mPGES-1 staining pattern in the IHC analysis suggested Epithelial CD31 No No Yes that cells located in the tumor microenvironment, and not tumor Mesenchymal Vimentin Yes No Yes cells, are responsible for the production of PGE . To confirm α 2 CAF SMA Yes No No this suggestion, double staining with mPGES-1 and GD2, a FAP Yes No No neuroblastoma tumor cell marker, was performed. In all three PDGFRβ Yes No Yes tumors analyzed a clear discrepancy in staining pattern was seen. PDGFRα No No Yes The areas with mPGES-1–expressing cells and GD2-expressing FSP-1 No Yes Yes cells were mutually exclusive, indicating that PGE2 is produced by cells in proximity to tumor cells but not by the tumor cells Colocalization was scored as negative (no) or positive (yes) in tumors from three patients, one from each subgroup [11q-deleted (tumor NB43), MYCN- themselves. To our surprise, low levels of COX-2 were detected amplified (tumor NB1), and low-risk (tumor NB4)]. in the majority of neuroblastoma samples analyzed (Fig. 2), but the pattern of COX-1 staining was similar to that of mPGES-1. Double staining with mPGES-1 and COX-1 showed coex- Discussion pression of both enzymes in some cells. We also detected cells The involvement of chronic inflammation in the initiation and with only COX-1 or mPGES-1 expression, perhaps because the enzyme was present at levels below the limit of detection for IHC progression of adult tumors of epithelial origin has been well analysis. However, COX-1 also has been shown to exert para- established (6). Much less is known about the contribution of crine distribution of PGH to mPGES-1 (34). PGH , which is inflammation and proinflammatory mediators in solid childhood 2 2 needed for conversion to PGE2, also could be supplied from cancers, and we therefore investigated the involvement of pros- circulating cells expressing COX-1 or COX-2. taglandins in different subsets of neuroblastoma. As discussed by Pistoia et al. (35), there is an immunosup- We found that mPGES-1 is expressed in human primary pressive tumor microenvironment in neuroblastoma that poten- neuroblastoma samples. Expression of mPGES-1 was present in tially enables tumor cells to evade host immune responses. This all tumors tested and was significantly higher in the 11q-deleted immunosuppressive state has been shown in human tumor sam- subset of high-risk tumors than in low-risk tumors on both the ples by the infiltration of M2-polarized macrophages (23) and in mRNA (Fig. 1A) and the protein level (Fig. 2). Our lipid analysis neuroblastoma mouse models by the infiltration of MDSCs (26, 36). Our data show predominance toward M2-polarization of the tumor-infiltrating macrophages (i.e., tumor-associated mac- rophages) in both 11q-deleted tumors and MYCN-amplified tu- NMA 11q- Low-risk NB1 NB43 NB4 mors, with elevated levels of CD206 and CD163, respectively A (Fig. S5), supporting the findings by Asgharzadeh et al. (23). + + Because CD68 and CD163 macrophages did not stain positive for mPGES-1 in the tumor samples, double staining with GD2 the dendritic cell marker CD11c was performed. No colocaliza- mPGES-1 GD2 mPGES-1 GD2 mPGES-1 GD2 tion was found; however, CD11c-expressing cells were in cell-to- – B NB1 NB43 NB4 cell contact with the mPGES-1 expressing cells in the NB4 tumor and to some extent in the NB43 tumor (Fig. S4). This finding led us to explore further different populations of immune +

CD68 cells present in the tumor samples in relation to mPGES-1 cells. + + Strikingly, in the 4S tumor NB4 there were CD3 CD4 T cells + mPGES-1 CD68 mPGES-1 CD68 mPGES-1 CD68 and CD20 B cells only in clusters together with mPGES-1– + C NB1 NB43 NB4 expressing cells and CD11c cells. Outside these immunological hotspots, no B cells or T cells were present. In the NB43 tumor, + + CD3 CD4 T cells were widespread in the tumor sections, al- though they were slightly more concentrated in the mPGES-1– Vimentin + expressing cell clusters. Nonetheless, essentially no CD20 mPGES-1 Vimentin mPGES-1 Vimentin mPGES-1 Vimentin B cells could be found in either the NB1 or the NB43 tumor. Spontaneous regression is an intriguing feature of 4S tumors, Fig. 3. Investigation of the tumor microenvironment in relation to mPGES-1. and the finding that only the 4S neuroblastoma sample had B MEDICAL SCIENCES (A) In total three human neuroblastoma tumors from one patient in each subgroup, NB1 (MYCN-amplified), NB43 (11q-deleted), and NB4 (low-risk), cells is interesting and should be investigated further. were stained with mPGES-1 (green) and a neuroblastoma tumor cell marker, Endothelial cells have been shown to express mPGES-1 and GD2(red).DuallabelingshowednocolocalizationofmPGES-1withtheneu- to secrete PGE2 (31). Colocalization between mPGES-1 and roblastoma tumor marker GD2. (B) Double staining of mPGES-1 (green) and a CD31 was detected in the 4S neuroblastoma sample NB4 (Fig. pan-macrophage marker (CD68) (red). No colocalization of mPGES-1 and CD68 S4). In the NB4 tumor and in the MYCN-amplified NB1 tumor, was found. (C) Dual labeling of vimentin (red), an intermediate filament found we also identified colocalization of mPGES-1 with vimentin. in cells of mesenchymal origin, and mPGES-1 (green) showed colocalization in Vimentin is an intermediate filament found in nerve cell pro- both the NB1 and NB4 tumors but not in the NB43 tumor. genitors and also in cells of mesenchymal origin and is used as a

Larsson et al. PNAS | June 30, 2015 | vol. 112 | no. 26 | 8073 Downloaded by guest on September 24, 2021 NMA 11q- Low-risk have been found to contribute to an immunosuppressive tumor A NB1 NB43 NB4 microenvironment (43). Next we wanted to investigate the effect of PGE2 inhibition on tumor growth in an in vivo model of 11q-deleted neuroblastoma. FAP Treatment with diclofenac, a nonselective COX inhibitor, resulted in reduced tumor growth. A reduction in PGE2 in the tumors was mPGES-1 FAP mPGES-1 FAP mPGES-1 FAP confirmed also. Based on our data, we propose that mPGES-1 is B NB1 NB43 NB4 expressed in subpopulations of CAFs; therefor only a fraction of cells in the whole tumor produce PGE2, explaining the relatively low levels of PGE2 found. Still, the decrease in PGE2 production

PDGFRβ resulting from diclofenac treatment is enough to have a significant effect on tumor progression in the in vivo 11q-deleted model, mPGES-1 PDGFRβ mPGES-1 PDGFRβ mPGES-1 PDGFRβ suggesting that targeting PGE could be of benefit for children NB1 NB43 NB4 2 C with aggressive, chemoresistant 11q-deleted neuroblastomas, as was indicated recently for bladder cancer (19). In conclusion, we have identified an activated COX/mPGES-1/ FSP-1 PGE2 pathway in 11q-deleted neuroblastoma with high expression of mPGES-1, low expression of 15-PGDH, and elevated levels of PGE mPGES-1 FSP-1 mPGES-1 FSP-1 mPGES-1 FSP-1 2 as compared with MYCN-amplified and low-risk tumors. Analysis of D NB1 NB43 NB4 expression cohorts revealed a worse outcome for high-risk patients (INSS stage 4) with high mPGES-1 expression. In addition, COX inhibition in an in vivo model of 11q-deleted neuroblastoma resulted αSMA in reduced tumor growth. These findings collectively suggest that mPGES-1 αSMA mPGES-1 αSMA mPGES-1 αSMA PGE2 signaling in 11q-deleted high-risk tumors could be targeted E NB1 NB43 NB4 using existing nonsteroidal anti-inflammatory drugs or by developing new compounds directly targeting mPGES-1. In addition there are convincing observations that mPGES-1/PGE2 may play a role in the complex network of the tumor microenvironment in general and PDGFRα specifically in CAFs (39, 44), further strengthening the suggestion mPGES-1 PDGFRα mPGES-1 PDGFRα mPGES-1 PDGFRα that mPGES-1 be targeted in neuroblastoma and in other tumors with activated COX/mPGES-1/PGE2 signaling. Fig. 4. Cellular localization of mPGES-1 expression in neuroblastoma tumors. Three human neuroblastoma tumors, NB1 (MYCN-amplified), NB43 (11q-deleted), Materials and Methods and NB4 (low-risk 4S), were stained for mPGES-1 (green) and FAP (A), PDGFRβ (B), The animal experiment was approved by the Stockholm North Animal Research FSP-1 (C), αSMA (D), or PDGFRα (E), markers for cancer-associated fibroblast Committee (approval N231/14) in accordance with the Animal Protection Law (CAFs), in red. All sections were counterstained with the nuclear dye Hoechst (SJVFS 2012:26), the Animal Protection Regulation (SFS 1988:539), and the Reg- (blue). Colocalization was seen for FAP, αSMA, and PDGFRβ in the NB1 tumor. ulation for the Swedish National Board for Laboratory Animals (SFS 1988: 541). Colocalization of PDGFRα, PDGFRβ and FSP-1 also was seen in the NB4 tumor. In , Detailed materials and methods are provided in SI Materials and Methods. the NB43 tumor mPGES-1 colocalized with FSP-1. Taken together, these results

suggest that CAFs are a source for mPGES-1 expression and PGE2 production. Patient Material. Neuroblastoma tumor tissue was obtained from the Karolinska University Hospital. Ethical approval was obtained by the Stockholm Regional marker of epithelial-to-mesenchymal transition (EMT), a pro- Ethical Review Board and the Karolinska University Hospital Research Ethics Committee (approval nos. 2009/1369-31/1 and 03/736). Tumor and patient cess promoting tumor metastasis (37). In a recent publication, characteristics are summarized in Table S1. Thirty-two neuroblastoma sam- epithelial tumor cells induced with EGF led to overexpression of ples representing all clinical subsets were used. mPGES-1 concomitant with increased vimentin expression and down-regulation of E-cadherin, initiating the phenotypic change of Prostaglandin Profiling. Approximately 40 mg of human primary tumor tissue EMT (38). Both the NB1 and the NB4 tumors had relatively high from 29 individuals was used in the analysis. Liquid–liquid extraction and vimentin expression, whereas only low, sporadic vimentin expres- analysis by LC-MS/MS was performed essentially as described previously (26). sion was found in the 11q-deleted tumor NB43, possibly because of chemotherapeutic treatment of this tumor before resection (Tables S1 and S3). Because vimentin also is expressed in fibro- blasts, we looked at CAFs. In a recent study Alcolea et al. (39) A 2.0 B showed that head and neck squamous cell carcinoma cells stim- CTRL 0.08 ulated both dermal and tumor-derived fibroblasts to express Diclofenac 0.06 1.5 0.04 mPGES-1, and they also observed PGE2 release in tumor-adjacent 0.015 mucosa (39). Erez et al. (40) showed that CAFs educated by tumor 1.0 cells mediated tumor-enhancing inflammation in an NF-κB–depen- 0.010 + dent manner. In line with these results, we found that mPGES-1 0.5 cells in all three tumors analyzed (NB1, NB4, and NB43) expressed 0.005 MYCN (mL) volume Tumor one or more markers for CAFs. The -amplified tumor 0.0 0.000

0246810pmol/mg extracted tumor NB1 coexpressed mPGES-1 and CAF markers such as vimentin, CTRL Diclofenac FAP, αSMA, and PDGFRβ but not FSP-1 and PDGFRα (41). Time (days) + Although mPGES-1 cells in NB4 expressed vimentin, PDGFRα, β α Fig. 5. COX inhibition reduces PGE2 levels and reduces tumor growth in vivo. PDGFR , and FSP-1, no colocalization with FAP or SMA was Nude mice were inoculated with SK-N-AS neuroblastoma cells harboring an detected. In the 11q-deleted tumor NB43, FSP-1 was the only + 11q-deletion and were treated with diclofenac (250 mg/L drinking water) for CAF marker present in mPGES-1 cells, and the expression of nine consecutive days. (A) Tumor volumes were measured each day, and a CAF markers was generally low (Fig. 4). Even though referred significant difference between diclofenac-treated and control mice was seen

to as one population of cells, CAFs are heterogeneous in their on day 8 (P = 0.01) and day 9 (P = 0.008). (B)PGE2 was measured in the tumors function and origin as well as in their expression of markers using solid-phase extraction and LC-MS/MS analysis. There was a significant

(41, 42), possibly reflecting the differential expression of CAF decrease in PGE2 in tumors from diclofenac-treated mice compared with markers seen in the neuroblastoma tumors. Recently, CAFs also control mice (P = 0.04). Bars represent median values.

8074 | www.pnas.org/cgi/doi/10.1073/pnas.1424355112 Larsson et al. Downloaded by guest on September 24, 2021 Quantitative Real-Time PCR Analysis. RNA was prepared from ∼30 mg of COX Inhibition of in Vivo 11q-Deleted Xenograft Tumor Growth. A xenograft primary tumor tissue, and 100 ng of RNA was used to synthesize cDNA. The mouse model harboring 11q-deletion tissue was treated with diclofenac for TaqMan Gene Expression Assay (Applied Biosystems) was used to evaluate 9 d essentially as described previously (45). the relative expression levels of mPGES-1, COX-1, and COX-2 mRNA. Statistical Analysis. For statistical comparisons a two-tailed t test and/or a The Gene-Expression Database R2. Gene-expression analysis of the impact of Mann–Whitney test were used when applicable. Tumors included in the mPGES-1 on overall survival in neuroblastoma patients was performed using a present study were collected in an ongoing national study and are part of a publicly available database [R2: microarray analysis and visualization plat- consecutive unbiased series of neuroblastoma patients. Samples available form (r2.amc.nl)]. In the public Versteeg-88 dataset, mRNA from 88 human for analysis were included in the study. neuroblastoma samples are included, 40 of whom are high-risk patients (INSS stage 4). ACKNOWLEDGMENTS. This work was supported by grants from the Swedish Research Council, The Swedish Cancer Society, the Swedish Childhood IHC Analysis of Neuroblastoma Tumor Tissue. Frozen tumor tissue, sectioned at Cancer Foundation, The Swedish Rheumatism Association, King Gustaf V 7 μm and fixed in 2% formaldehyde, was used. Staining was preformed as 80 Years Foundation, Swedish Foundation for Strategic Research, and The described earlier (31). All antibodies used in the study are listed in Table S4. Stockholm County Council.

1. Maris JM (2010) Recent advances in neuroblastoma. N Engl J Med 362(23):2202–2211. 27. Johnsen JI, et al. (2004) -2 is expressed in neuroblastoma, and non- 2. Cohn SL, et al.; INRG Task Force (2009) The International Neuroblastoma Risk steroidal anti-inflammatory drugs induce apoptosis and inhibit tumor growth in vivo. Group (INRG) classification system: An INRG Task Force report. J Clin Oncol 27(2): Cancer Res 64(20):7210–7215. 289–297. 28. Ponthan F, et al. (2007) prevents neuroblastoma tumor development and 3. Carén H, et al. (2010) High-risk neuroblastoma tumors with 11q-deletion display a potentiates the effect of chemotherapeutic drugs in vitro and in vivo. Clin Cancer Res poor prognostic, chromosome instability phenotype with later onset. Proc Natl Acad 13(3):1036–1044. – Sci USA 107(9):4323 4328. 29. Rasmuson A, et al. (2012) Autocrine prostaglandin E2 signaling promotes tumor cell 4. Brodeur GM, Bagatell R (2014) Mechanisms of neuroblastoma regression. Nat Rev Clin survival and proliferation in childhood neuroblastoma. PLoS ONE 7(1):e29331. – Oncol 11(12):704 713. 30. Yu AL, et al.; Children’s Oncology Group (2010) Anti-GD2 antibody with GM-CSF, ’ 5. D Angio GJ, Evans AE, Koop CE (1971) Special pattern of widespread neuroblastoma interleukin-2, and isotretinoin for neuroblastoma. N Engl J Med 363(14):1324–1334. – with a favourable prognosis. Lancet 1(7708):1046 1049. 31. Westman M, et al. (2004) Expression of microsomal prostaglandin E synthase 1 in 6. Wang D, Dubois RN (2010) and cancer. Nat Rev Cancer 10(3):181–193. rheumatoid arthritis synovium. Arthritis Rheum 50(6):1774–1780. 7. Yoshimatsu K, et al. (2001) Inducible microsomal prostaglandin E synthase is over- 32. Kryh H, et al. (2011) Comprehensive SNP array study of frequently used neuroblas- expressed in colorectal adenomas and cancer. Clin Cancer Res 7(12):3971–3976. toma cell lines; copy neutral loss of heterozygosity is common in the cell lines but 8. van Rees BP, et al. (2003) Expression of microsomal prostaglandin E synthase-1 in uncommon in primary tumors. BMC Genomics 12:443–453. intestinal type gastric adenocarcinoma and in gastric cancer cell lines. Int J Cancer 33. Ladenstein R, et al. (2010) Randomized Trial of prophylactic granulocyte colony- 107(4):551–556. stimulating factor during rapid COJEC induction in pediatric patients with high-risk 9. Sasaki Y, et al. (2012) Microsomal prostaglandin E synthase-1 is involved in multiple neuroblastoma: The European HR-NBL1/SIOPEN study. J Clin Oncol 28(21):3516–3524. steps of colon carcinogenesis. Oncogene 31(24):2943–2952. 34. Salvado MD, Alfranca A, Escolano A, Haeggström JZ, Redondo JM (2009) COX-2 limits 10. Yoshimatsu K, et al. (2001) Inducible prostaglandin E synthase is overexpressed in non-small cell lung cancer. Clin Cancer Res 7(9):2669–2674. production in activated HUVECs and is a source of PGH2 for transcellular – 11. Jain S, Chakraborty G, Raja R, Kale S, Kundu GC (2008) Prostaglandin E2 regulates to PGE2 by tumor cells. Arterioscler Thromb Vasc Biol 29(7):1131 1137. tumor angiogenesis in prostate cancer. Cancer Res 68(19):7750–7759. 35. Pistoia V, et al. (2013) Immunosuppressive microenvironment in neuroblastoma. Front – 12. Hanaka H, et al. (2009) Microsomal prostaglandin E synthase 1 determines tumor Oncol 3:167 174. growth in vivo of prostate and lung cancer cells. Proc Natl Acad Sci USA 106(44): 36. Santilli G, et al. (2013) Polyphenon [corrected] E enhances the antitumor immune 18757–18762. response in neuroblastoma by inactivating myeloid suppressor cells. Clin Cancer Res 13. Pai R, et al. (2002) Prostaglandin E2 transactivates EGF receptor: A novel mechanism 19(5):1116–1125. for promoting colon cancer growth and gastrointestinal hypertrophy. Nat Med 8(3): 37. Thiery JP, Acloque H, Huang RY, Nieto MA (2009) Epithelial-mesenchymal transitions 289–293. in development and disease. Cell 139(5):871–890. 14. Sheng H, Shao J, Washington MK, DuBois RN (2001) Prostaglandin E2 increases 38. Donnini S, et al. (2012) EGFR signaling upregulates expression of microsomal pros- growth and motility of colorectal carcinoma cells. J Biol Chem 276(21):18075–18081. taglandin E synthase-1 in cancer cells leading to enhanced tumorigenicity. Oncogene 15. Buchanan FG, Wang D, Bargiacchi F, DuBois RN (2003) Prostaglandin E2 regulates cell 31(29):3457–3466. migration via the intracellular activation of the epidermal growth factor receptor. 39. Alcolea S, et al. (2012) Interaction between head and neck squamous cell carcinoma J Biol Chem 278(37):35451–35457. cells and fibroblasts in the biosynthesis of PGE2. J Lipid Res 53(4):630–642. 16. Pai R, et al. (2001) PGE(2) stimulates VEGF expression in endothelial cells via ERK2/ 40. Erez N, Truitt M, Olson P, Arron ST, Hanahan D (2010) Cancer-Associated Fibroblasts JNK1 signaling pathways. Biochem Biophys Res Commun 286(5):923–928. Are Activated in Incipient Neoplasia to Orchestrate Tumor-Promoting Inflammation 17. Trompezinski S, Pernet I, Schmitt D, Viac J (2001) UV radiation and prostaglandin E2 in an NF-kappaB-Dependent Manner. Cancer Cell 17(2):135–147. up-regulate vascular endothelial growth factor (VEGF) in cultured human fibroblasts. 41. Cortez E, Roswall P, Pietras K (2014) Functional subsets of mesenchymal cell types in – Inflamm Res 50(8):422 427. the tumor microenvironment. Semin Cancer Biol 25:3–9. 18. Sheng H, Shao J, Morrow JD, Beauchamp RD, DuBois RN (1998) Modulation of apo- 42. Paulsson J, Micke P (2014) Prognostic relevance of cancer-associated fibroblasts in ptosis and Bcl-2 expression by prostaglandin E2 in human colon cancer cells. Cancer human cancer. Semin Cancer Biol 25:61–68. – Res 58(2):362 366. 43. Kakarla S, Song XT, Gottschalk S (2012) Cancer-associated fibroblasts as targets for 19. Kurtova AV, et al. (2015) Blocking PGE2-induced tumour repopulation abrogates immunotherapy. Immunotherapy 4(11):1129–1138. bladder cancer chemoresistance. Nature 517(7533):209–213. 44. Li HJ, Reinhardt F, Herschman HR, Weinberg RA (2012) Cancer-stimulated mesen- 20. Obermajer N, et al. (2012) PGE(2)-driven induction and maintenance of cancer-asso- chymal stem cells create a carcinoma stem cell niche via prostaglandin E2 signaling. ciated myeloid-derived suppressor cells. Immunol Invest 41(6-7):635–657. Cancer Discov 2(9):840–855. 21. Kalinski P (2012) Regulation of immune responses by prostaglandin E2. J Immunol 45. Baryawno N, et al. (2008) Tumor-growth-promoting cyclooxygenase-2 prostaglandin 188(1):21–28. E2 pathway provides medulloblastoma therapeutic targets. Neuro-oncol 10(5): 22. Hanahan D, Weinberg RA (2011) Hallmarks of cancer: The next generation. Cell 661–674. 144(5):646–674. 46. Idborg H, et al. (2013) Effects of mPGES-1 deletion on and fatty acid 23. Asgharzadeh S, et al. (2012) Clinical significance of tumor-associated inflammatory – cells in metastatic neuroblastoma. J Clin Oncol 30(28):3525–3532. profiles in mice. Prostaglandins Other Lipid Mediat 107:18 25. 24. Kharaishvili G, et al. (2014) The role of cancer-associated fibroblasts, solid stress and 47. Melin E, Lindroos E, Lundberg IE, Borg K, Korotkova M (2014) Elevated expression of other microenvironmental factors in tumor progression and therapy resistance. prostaglandin E2 synthetic pathway in skeletal muscle of prior polio patients. – Cancer Cell Int 14:41–48. J Rehabil Med 46(1):67 72.

’ MEDICAL SCIENCES 25. Zeine R, et al. (2009) Presence of cancer-associated fibroblasts inversely correlates 48. Pearson AD, et al.; European Neuroblastoma Study Group; Children sCancerand with Schwannian stroma in neuroblastoma tumors. Mod Pathol 22(7):950–958. Leukaemia Group (CCLG formerly United Kingdom Children’s Cancer Study Group) 26. Carlson LM, et al. (2013) Low-dose delays an inflammatory tumor progression (2008) High-dose rapid and standard induction chemotherapy for patients aged in vivo in a transgenic mouse model of neuroblastoma. Carcinogenesis 34(5): over 1 year with stage 4 neuroblastoma: A randomised trial. Lancet Oncol 9(3): 1081–1088. 247–256.

Larsson et al. PNAS | June 30, 2015 | vol. 112 | no. 26 | 8075 Downloaded by guest on September 24, 2021