Oncogene (2010) 29, 4766–4778 & 2010 Macmillan Publishers Limited All rights reserved 0950-9232/10 www.nature.com/onc ORIGINAL ARTICLE L1CAM–integrin interaction induces constitutive NF-jB activation in pancreatic adenocarcinoma cells by enhancing IL-1b expression

H Kiefel1, S Bondong1, N Erbe-Hoffmann1, J Hazin1, S Riedle1, J Wolf1, M Pfeifer1, A Arlt2, H Scha¨fer2, S Sebens Mu¨erko¨ster2 and P Altevogt1

1Translational Immunology, German Cancer Research Center, Heidelberg, Germany and 2Clinic for Internal Medicine, Laboratory of Molecular Gastroenterology and Hepatology, UKSH-Campus, University of Kiel, Kiel, Germany

L1 cell adhesion molecule (L1CAM) overexpression is 2004; Jemal et al., 2007). Resection of the tumour is still often associated with bad prognosis in various human the only option that offers a chance for cure (Schneider carcinomas. Recent studies also suggest a role of L1CAM et al., 2005) hence, the prognosis for patients is poor and in pancreatic ductal adenocarcinomas (PDAC). To more than 80% die within 5 years after surgery. further address its contribution, we expressed functional Pancreatic cancer is often refractory to conventional domains of L1CAM in PT45-P1 PDAC cells. We found chemotherapeutic treatment because of the frequent that L1CAM that is full length (L1-FL), but neither the development of profound chemoresistance. This che- soluble ectodomain (L1ecto) nor the cytoplasmic part moresistant phenotype is often mediated by the con- (L1cyt), could enhance cell proliferation or tumour growth stitutive activation of the transcription factor NF-kB, in mice. Expression of L1-FL resulted in constitutive resulting from the autocrine production of interleukin- activation of NF-jB, which was abolished by L1CAM 1b (IL-1b) (Arlt et al., 2002; Muerkoster et al., 2005). knockdown. We showed that the expression of IL-1b was An important goal is to understand the molecular selectively upregulated by L1-FL, and increased IL-1b parameters involved not only in chemoresistance but levels were instrumental for sustained NF-jB activation. also in pancreatic carcinoma progression. IL-1b production and NF-jB activation were abolished by The L1 cell adhesion molecule (L1CAM) is a knockdown of a5-integrin and integrin-linked kinase, but 200–220 kDa transmembrane glycoprotein of the insensitive to depletion of L1CAM cleavage proteinases. immunoglobulin superfamily composed of six immuno- Supporting these data, PT45-P1 cells transduced with globulin-like domains and five fibronectin type III an L1CAM mutant deficient in integrin binding (L1-RGE) repeats, followed by a transmembrane region and a did not support the described L1-FL functions. Our highly conserved cytoplasmic tail (Moos et al., 1988). results suggest that membranous L1CAM interacts with L1CAM can interact with itself (homophilic) and with RGD-binding integrins, leading to sustained NF-jB activa- heterophilic ligands such as integrins, CD24, neurocan, tion by IL-1b production and autocrine/paracrine signalling. neuropilin-1 and other members of the neural cell The unravelling of this novel mechanism sheds new light on adhesion family (Brummendorf et al., 1998; Hortsch, the important role of L1CAM expression in PDAC cells. 2000). Many binding sites in the L1CAM molecule have Oncogene (2010) 29, 4766–4778; doi:10.1038/onc.2010.230; already been mapped. The RGD site(s) in the sixth published online 14 June 2010 immunoglobulin domain (one in human, two in mouse) support a5- and av-integrin-mediated cell binding Keywords: pancreatic tumour; L1CAM signalling; (Ruppert et al., 1995; Montgomery et al., 1996; integrins Oleszewski et al., 1999). Homophilic L1CAM interac- tion has been recently mapped to immunoglobulin domains 1–4 (Gouveia et al., 2008). L1CAM was initially described in the nervous system, Introduction in which it is important for cell migration and axon outgrowth. Subsequent work has shown overexpression Pancreatic cancer is a highly aggressive disease that is of L1CAM in a variety of human tumours, in which it is mostly detected in an advanced stage. Owing to its associated with bad prognosis and metastases formation exocrine functions, it frequently develops multiple (Thies et al., 2002; Fogel et al., 2003a, b; Boo et al., metastases and is the fourth leading cause of cancer- 2007; Kaifi et al., 2007). Furthermore, L1CAM was related death in the United States. (Neoptolemos et al., detected in tumour-associated endothelium (Issa et al., 2009; Maddaluno et al., 2009). Expression of L1CAM was also described in pancreatic ductal adenocarcino- Correspondence: Professor P Altevogt, Translational Immunology, mas (PDACs) (Sebens Muerkoster et al., 2007, 2009; D015, German Cancer Research Center, Im Neuenheimer Feld 280, Geismann et al., 2009). Accumulating evidence suggests Heidelberg D-69120, Baden Wu¨rtemberg, Germany. E-mail: [email protected] an important role of L1CAM in the malignancy of Received 28 October 2009; revised 19 April 2010; accepted 7 May 2010; this tumour (Sebens Muerkoster et al., 2007, 2009; published online 14 June 2010 Geismann et al., 2009). The mechanism of L1CAM- L1CAM and NF-kappaB H Kiefel et al 4767 mediated tumour progression is not clearly established. processed further in both PT45-P1-L1-FL and Panc-1 Previous studies showed that L1CAM augments tumour cells (Supplementary Figure 1b). There was no effect of growth in NOD/Scid mice, enhances cell motility on treatment observed in L1cyt-expressing cells (data not extracellular matrix proteins and increases matrigel shown). invasion (Mechtersheimer et al., 2001; Thelen et al., L1ecto was secreted in large amounts into the medium 2002; Silletti et al., 2004; Gast et al., 2005; Gavert et al., (Figure 1e) and was detected by western blot (Figure 1d) 2005; Meier et al., 2006). Interference with L1CAM in cell lysates and L1CAM-specific capture enzyme- expression by genetic manipulation or by selective linked immunosorbent assay (ELISA) (Figure 1e). As knockdown was found to be growth inhibitory (Primiano expected, L1-FL cells released significantly less soluble et al., 2003; Bao et al., 2008). Studies from different L1CAM into the medium (Figure 1e). The amount laboratories showed L1CAM-dependent gene expres- of soluble L1CAM from L1-FL in the supernatant sion in mouse and human carcinoma cells (Silletti et al., was increased by PMA treatment, whereas there was 2004; Gavert et al., 2007; Gast et al., 2008a) that is no effect on the release of L1ecto (Supplementary relevant for understanding the mechanisms of L1CAM Figure 1a). These results implicate that L1CAM tumour-promoting effects. processing in pancreatic cells is similar to that of other To further analyse the mechanisms of L1CAM carcinoma cells. signalling, we expressed functional domains of L1CAM in PT45-P1 PDAC cells. We show that only L1-FL shows nuclear translocation that is increased L1CAM that is full-length (L1-FL), but neither the by PMA treatment soluble L1CAM (L1ecto) nor the cytoplasmic part In a recent study, we have shown that the C-terminal (L1cyt), can promote cell proliferation and tumour L1CAM fragment translocates into the nucleus depend- growth in vivo. Only L1-FL-expressing cells support ing on proteolysis by ADAM10 and presenilin (Riedle NF-kB activation and upregulate various genes includ- et al., 2009). We examined nuclear translocation in ing the prototype NF-kB inducer IL-1b. The L1CAM- PT45-P1 cells expressing L1-FL and L1cyt by fluores- mediated upregulation of IL-1b expression seems to be cence microscopy, using pcytL1 to the cytoplasmic tail instrumental for NF-kB activation and requires the of L1CAM (Riedle et al., 2009). Increased nuclear RGD motive in L1CAM and integrin signalling. Our staining of L1CAM was observed following treatment work identifies important novel downstream effects of of L1-FL cells with PMA (Figure 1f). PMA-induced L1CAM that help to understand the molecular mechan- nuclear translocation of L1-FL was clearly blocked by isms leading to tumour progression promoted by pretreatment with DAPT (Figure 1f and Supplementary L1CAM. Figure 2). In contrast, no effect of the inhibitor or PMA was observed in PT45-P1 cells expressing L1cyt, suggesting that L1cyt could reach the nucleus without Results proteolytic processing (Figure 1f and Supplementary Figure 2). Characterization of PT45-P1 cells overexpressing L1CAM domains We transduced PT45-P1 pancreatic adenocarcinoma PT45-P1 expressing L1CAM full-length augments cell cells with functional domains of L1-FL, cytoplasmic proliferation and tumour growth in vivo L1CAM (L1cyt) or L1CAM ectodomain (L1ecto), as Next, we analysed the impact of L1CAM domains on outlined (Figure 1a). The selected cells stably expressed cellular functions in PDAC cells. Therefore, we tested L1CAM fragments as detected by quantitative real-time cell proliferation of PT45-P1 cells expressing various (qRT)–PCR (Figure 1b) and by fluorescence-activated L1CAM constructs. A clear growth-promoting effect of cell sorting analysis using antibodies to the cytoplasmic L1-FL compared with mock-transfected cells was noted (mAb 74 5H7) or the extracellular portion of L1CAM after 96 h (Figure 2a). Despite the significantly higher (mAb L1-11A and L1-9.3) (Figure 1c). L1-FL and L1cyt amounts of released L1ecto, only a marginal effect was were visualized by western blot analysis of cell lysates observed in L1ecto cells. (Figure 1d). L1-FL also showed detectable amounts of To test whether enhanced proliferation in vitro L1-32 (Figure 1d), representing the membrane-retained resulted in augmented tumour growth in vivo,we cleavage fragment resulting from ADAM10-mediated injected the different cell lines into NOD/Scid mice. ectodomain shedding of L1CAM (Gutwein et al., 2000; Moreover, in vivo, only PT45-P1 cells expressing L1-FL Mechtersheimer et al., 2001). Further analysis showed showed significantly enhanced tumour growth com- that cleavage of L1-FL was augmented by Phorbol 12- pared with mock-transduced cells (Figures 2b and c). myristate 13-acetate (PMA) treatment and was blocked In contrast, tumour growth of L1ecto and L1cyt cells by the metalloproteinase inhibitor BB-3644 (Supple- was comparable to that of PT45-P1 mock-transduced mentary Figures 1a and b). Similar results were obtained cells (Figures 2b and c). with Panc-1 cells endogenously expressing L1CAM (Supplementary Figure 1b right panel). As shown PT45-P1 cells expressing L1-FL promote gene regulation previously in ovarian carcinoma cells (Riedle et al., Previous research has shown that L1CAM expression 2009), treatment with presenilin inhibitor DAPT in- supports gene regulation in tumour cells. To establish creased the amount of L1-32 by preventing it from being expression profiles for PT45-P1 L1CAM-expressing

Oncogene L1CAM and NF-kappaB H Kiefel et al 4768 primer: L1CAM cyt PT45-P1 1000 EC mock L1-FL L1ecto 800 ** Actin β 600 % of Max NH NH NH 2 2 2 400 mRNA level

Ig1 L1CAM/ 200 Ig2 fluorescence: L1 ecto Ig3 0 L1-220 L1-200 Ig4 IC mock L1-FL L1ecto Ig5 mock L1-FL L1cyt Ig6 RGE L1ecto FNIII1 FNIII2 primer: L1CAM ecto % of Max FNIII3 800 ** FNIII4 ** EC FNIII5 600 Actin fluorescence: L1 ecto IC β 400 IC L1-32 300 mock L1-FL L1cyt

L1-FL L1ecto L1cyt L1-RGE mRNA level 100 L1CAM/

0 % of Max

mock L1cyt L1-FLL1ecto PT45-P1 fluorescence: L1 cyt

*** mockL1-FLL1ectoL1cyt 35 NS DMSO L1-220 * *** PMA L1-85 30 4.0 *** PMA/DAPT * 3.5 25 L1-32 3.0 α L1 cyt 2.5 *** 20 α β hu Actin 0.3 PT45-P1 15

[OD450] 0.2 0.1 10 soluble L1CAM soluble 0.0 5 mockL1-FLL1ectoL1cyt

L1-220 mean fluorescence intensity mock L1cyt 0 α L1 ecto L1-FLL1ecto α hu βActin PT45-P1 L1cyt PT45-P1 L1-FL PT45-P1 Figure 1 Characterization of PT45-P1 cells expressing L1CAM domains (a) Schematic illustration of L1CAM constructs used in this study. (b) qRT–PCR analysis using primers specific for the cytoplasmic or the ectodomain part of L1CAM. (c) Cytofluorographic analysis of PT45-P1 cells retrovirally transduced with the indicated constructs. Cells were stained directly with antibodies to the ectodomain of L1CAM (L1-11A) (extracellular (EC)) or after permeabilization and fixation with mAb to the intracellular domain (74 5H7) (intracellular (IC)), followed by staining with a PE-conjugated goat anti-mouse IgG. (d) Western blot analysis of tumour cell lysates using pcytL1 to the cytoplasmic portion of L1CAM (top) and mAb L1-9.3 to the ectodomain of L1CAM (bottom). *marks an unspecific band. (e) Soluble L1CAM in the tissue culture supernatant of dense cultures was analysed by L1CAM-specific capture ELISA. (f) Nuclear translocation of L1CAM. Cells were pretreated overnight with DMSO (vehicle control) or DAPT (10 mM)at371C and than stimulated with PMA for 2 h. Cells were fixed, permeabilized with methanol (À20 1C) and stained with pcytL1 and Alexa488- conjugated anti-rabbit IgG. Note that methanol treatment reduces membranous L1CAM staining but leaves intact intracellular and nuclear staining. **Po0.01, ***Po0.001.

cells, we performed a DNA chip screen. Schematically, IL-11 and IL-1b were upregulated in tumours derived the expression differences between PT45-P1 mock from L1-FL-expressing PT45-P1 cells compared with and L1-FL are summarized in Figure 3a. A total mock, L1cyt- or L1ecto-expressing cells. The effects of 226 genes were selectively regulated in L1-FL- were even more apparent in the in vivo background expressing cells. (Figure 3c). In contrast, the expression of MDK and To analyse whether differential gene expression had S100A4 was also increased in PT45-P1 L1ecto-derived an impact on cellular function of PDACs cells, we tumours (Figure 3c). Cross-reactivity of qRT–PCR selected candidate genes that were described to be primers to mouse sequences was excluded (data not involved in tumour progression and were expressed in shown). As expected, levels of IL-1b protein, as pancreatic tumours. We verified the results for the measured by IL-1b-specific ELISA, were elevated in selected genes by qRT–PCR and included L1ecto and L1-FL compared with mock cells both in vitro (super- L1cyt cells into the analysis. We observed that the natant) and in vivo (tumour lysate) (Figure 3d). expression of IL-11, IL-1b, S100A4 and MDK was significantly upregulated by L1-FL, but not by L1ecto or L1cyt (Figure 3b). To analyse whether the same genes L1-FL induces constitutive NF-kB activation in PT45-P1 were also upregulated under in vivo growth conditions, cells we isolated RNA from tumours grown in NOD/Scid We noted that many of the identified genes were related mice. Using qRT–PCR analysis, we confirmed that to the NF-kB signalling pathway. Most remarkable was

Oncogene L1CAM and NF-kappaB H Kiefel et al 4769 PT45-P1 3.0 * mock 2.5 L1-FL 2.0 1.5 L1ecto 1.0 L1cyt

proliferation 0.5 0.0 fold increase to d0 fold 0h 48h 96h time

*** *** 400 ** mock 800 mock L1-FL L1-FL 300 L1ecto 600 L1ecto 200 L1cyt L1cyt 400 100 200

0 tumour mass (mg) tumour volume (mm^3) tumour volume 0 d0 d6 d9 d13 d16 d20 d24 mock L1cyt time L1-FL L1ecto PT45-P1 Figure 2 L1-FL expression promotes cell proliferation and tumour growth (a) Cells were starved for 48 h in serum-free medium and then grown in fresh medium containing 10% FCS. Cell proliferation was measured using MTT assay. (b) Growth of tumour cells in NOD/Scid mice. 5 Â 106 cells were injected s.c. into the flanks of 6-week-old NOD/Scid mice, (44 animals per group), and tumour growth was monitored for 24 days. Results represent mean tumour volume for n 44 mice/group. (c) At the end of the experiments (day 24), tumours were excised and the weights were determined. Data are pooled from three independent experiments. *Po0.05, **Po0.01, ***Po0.001.

4.0 2 IL-11 MDK 3.5 2.5 * 2.5 *** 2 *** 3.0 2.0 2.0 4 2.5 Actin 1.5 1.5 β genes

9 Actin

2.0 β upregulated 131 1.0 1.0 MDK/ 1.5 mRNA level 0.5 0.5 mRNA level -1.5 IL-11/ 66 -2.0 0.0 0.0 9 -2.5 IL-1β S100A4 3 -3.0 genes 4.0 ** 3.5 *** fold regulation *** 3.0 downregulated 3.0 2.5 Actin β Actin 2.0

IL-11 MDK β

/ 2.0

35 2.0 β 1.5 30 1.0 mRNA level mRNA level

IL-1 1.0 1.5 25 S100A4/ 0.5

20 Actin 0.0 0.0 β 1.0 15 10 IL-11/Actin 0.5 mock L1cyt mock L1cyt mRNA level L1-FL L1-FL MDK/ 5 level mRNA L1ecto L1ecto PT45-P1 0 0.0 PT45-P1 IL-1β IL-1β IL-1β S100A4 6 ** mock 35 12 5 mock L1-FL 30 10 L1-FL 4 4 25 8 Actin β 20 Actin 3 level β / 6 A 15 β 2 2 4 10 pg/µg protein mRNA level IL-1 mRN 1 pg/1*10^5 cells 5

2 S100A4/ 0 0 0 0 PT45-P1 PT45-P1 supernatant tumour lysate mock L1cyt mock L1cyt L1-FL L1ecto L1-FL L1ecto PT45-P1 PT45-P1 Figure 3 Expression profiling of PT54-P1-transfected cell lines (a) Total RNA was isolated from PT45-P1-FL or mock cells and used for expression profiling on Illumina human Sentrix-6 v3 BeadChip. (b) Selected genes identified in (a) (IL-11, IL-1b, MDK and S100A4) were confirmed by qRT–PCR analysis. A representative experiment of n>3 is shown. (c) mRNA was isolated from tumours in Figure 2 and cDNAs were analysed by qRT–PCR as in (b). Data shown represent the mean±s.d. for n ¼ 4 tumours. (d) IL-1b protein content was determined by an IL-1b-specific ELISA in supernatant from densely grown cells or in tumour lysate from in vivo grown tumours. A representative experiment from n ¼ 4 is shown. *Po0.05, **Po0.01, ***Po0.001.

Oncogene L1CAM and NF-kappaB H Kiefel et al 4770 the role of IL-1b as a prototype inducer of NF-kB (Figure 4a). To confirm the enhanced NF-kB activation, activation. Furthermore, the autocrine production we analysed phosphorylation levels of the NF-kB of IL-1b was previously described to induce sustained subunit p65 in PT45-P1-mock and L1-FL cells by NF-kB activation in PDAC cells (Arlt et al., 2002). western blot analysis of total cell extracts (Figure 4b). Therefore, we compared the NF-kB activity between We observed increased phosphorylation of p65 in the PT45-P1 cells expressing various L1CAM domains, lysate from L1CAM-expressing PT45-P1 cells using an NF-kB-luciferase reporter plasmid. We ob- (Figure 4b). served significantly increased NF-kB activation in To show that the intrinsic NF-kB activity was due to L1-FL versus mock-transduced cells (Figure 4a). L1ecto enhanced IL-1b expression, we used a neutralizing mAb showed an intermediate activation, whereas L1cyt- to IL-1b and to IL-1 receptor that efficiently inhibits expressing cells were similar to mock-transduced cells binding and signalling of IL-1b. Both compounds

NF-κB activity 200 ***

150 mock L1-FL α P-p65 100 RLU α p65 50 α hu βActin % of SV40 control 0 PT45-P1

mock L1cyt L1-FL L1ecto PT45-P1

κ NF- B activity 2.5 mock 175 mock ** 2.0 L1-FL 150 ** L1-FL 125 1.5 100 1.0

RLU 75 50 proliferation 0.5

25 increase to d0 fold % of SV40 control 0.0 0 β β IgG1 IgG1 α IL-1 α IL-1 α IL-1 RI α IL-1 RI Antibody Antibody

5 mock 4 3 L1-FL mock 3 L1-FL 2 2 proliferation 1 1 fold increase to d0 fold proliferation 0 fold increase to d0 fold 0

DMSO DMSO DMSO GFP p65 parthenolide sulfasalazine Bay11-7082 siRNA Inhibitor Figure 4 L1-FL supports constitutive NF-kB activation (a) The indicated cells were transfected with an NF-kB-Luciferase reporter plasmid, together with a Renilla plasmid for transfection efficiency control. Luciferase activity was determined after 48 h. A representative experiment from n ¼ 3 is shown. (b) Western blot analysis of cell lysates from the indicated cells with an Ab to phospho-p65. (c) PT45-P1 mock or L1-FL cells were incubated with neutralizing mAb to IL-1b (anti-IL-1b,10mg/ml) or IL-1 receptor (IL-1RI, 250 ng/ml) or an IgG1 isotype control. NF-kB activation was measured by a luciferase reporter assay. Data shown represent the mean±s.d. for n ¼ 2 determinations. (d) PT45-P1 mock or L1-FL cells were incubated with neutralizing mAb to IL-1b (10 mg/ml) or IL-1 receptor (250 ng/ml) or an IgG1 isotype control. Cell proliferation was determined by MTT assay. A representative experiment of n ¼ 3 is shown. (e) The indicated cells were treated with the NF-kB inhibitors parthenolide (1 mM), sulfasalazine (100 mM), or Bay11- 7082 (1 mM), and cell proliferation was determined by MTT assay. A representative experiment of n43 is shown. (f) The indicated cells were transfected with a p65-specific siRNA or GFP siRNA for control, and cell proliferation was determined by MTT assay. A representative experiment of n ¼ 3 is shown. **Po0.01, ***Po0.001.

Oncogene L1CAM and NF-kappaB H Kiefel et al 4771 could block NF-kB activity in L1-FL-expressing We examined whether L1CAM was involved in NF- cells (Figure 4c). When cultivated in the presence kB activity in Panc-1 PDAC cells that endogenously of antagonists, L1CAM-dependent proliferation was express L1CAM. Indeed, the knockdown of L1CAM clearly reduced (Figure 4d). Similarly, the pharmacolo- (Figure 5d) almost completely suppressed the expression gical inhibitors of NF-kB, parthenolide, sulfasalazine of IL-1b (Figure 5e) and abolished the constitutive NF- and Bay11-7082, as well as small interfering RNA kB activation (Figure 5f). Furthermore, depletion of (siRNA)-mediated depletion of the p65-subunit, re- L1CAM in Panc-1 resulted in a significant inhibition of duced the enhanced cell proliferation of L1-FL cells, proliferation (Figure 5g). These data underline the fact but had little effect on mock-transduced cells (Figures 4e that L1-FL has an important role in the constitutive and f). These results suggest that an autocrine loop NF-kB activation of pancreatic carcinoma cell lines. involving IL-1b and IL-1 receptor mediates sustained NF-kB activation in cells expressing L1-FL. This activity can drive other cellular functions such as cell Role of integrins in the constitutive NF-kB activation proliferation. L1CAM is a ligand for several integrins such as a5b1, avb3, aIIbb3 and avb5. L1CAM on tumour cells can interact with integrins on neighbouring tumour cells or L1CAM knockdown reverts constitutive NF-kB on stromal cells of the surrounding tissue. As suggested activation before, interaction between neighbouring tumour cells To further show the role of L1CAM in constitutive NF- could lead to bidirectional signalling by L1CAM and kB activation, we used siRNA mediated knockdown. integrins, as many cell types express both binding Depletion of L1CAM in PT45-P1-L1-FL cells led to a partners (Gast et al., 2008b). To address the involve- clear reduction of L1CAM as detected by qRT–PCR ment of integrins in NF-kB activation, we examined the and western blotting (Figure 5a), accompanied by a effect of integrin knockdown on IL-1b expression. The drop in IL-1b expression (Figure 5b). L1CAM knock- heavy chains of a5- and av-integrins were efficiently down also reduced NF-kB activity as determined by depleted as shown by qRT–PCR analysis (Figures 6a luciferase reporter assay but had only a nonsignificant and b) and fluorescence-activated cell sorting analysis effect in mock cells (Figure 5c). with specific antibodies (data not shown). With respect

L1CAM IL-1β NF-κB activity 400 GFP siRNA 5.0 250 ** L1CAM siRNA#1 GFP siRNA GFP siRNA L1CAM siRNA#1 300 L1CAM siRNA#2 4.0 200 NS L1CAM siRNA#2 Actin L1CAM siRNA#2 β

GFP siRNAL1CAM siRNA #2 Actin 3.0 150

200 β /

α L1 ecto β 2.0 RLU 100

mRNA level 100 α β

hu Actin mRNA level IL-1 L1CAM/ 1.0 50 PT45-P1 0 0.0 % of SV40 control L1-FL 0 mock L1-FL mock L1-FL mock L1-FL PT45-P1 PT45-P1 PT45-P1

L1CAM IL-1β 1.4 GFP siRNA 1.2 GFP siRNA 1.2 L1CAM siRNA#1 1.0 L1CAM siRNA#1 1.0 L1CAM siRNA#2 L1CAM siRNA#2

Actin 0.8 Actin β 0.8 GFP siRNAL1CAM siRNA #2 β

/ 0.6

0.6 α L1 ecto β 0.4 α hu βActin 0.4 mRNA level IL-1 mRNA level

L1CAM/ 0.2 Panc-1 0.2 0.0 0.0 Panc-1 Panc-1

κ NF-κB activity NF- B activity 600 GFP siRNA 3x105 GF siRNAP 4 GFP siRNA 500 ** L1CAM siRNA#2 L1CAM siRNA#2 L1CAM siRNA#2 3 400 2x105 300 2 RLU P-p65 200 arbitrary 1x105 1 100 proliferattion % of SV40 control fold increase to d0 fold 0 luminescence units 0 0 Panc-1 Panc-1 Panc-1 Figure 5 Knockdown of L1CAM abolishes constitutive NF-kB activation (a) Cells were transfected with two L1CAM-specific siRNAs or GFP siRNA for control and the depletion was verified by qRT–PCR and western blot analysis using mAb L1-9.3 to the L1CAM ectodomain. (b) L1CAM depletion is accompanied by a reduction in IL-1b expression. A representative experiment of n ¼ 4is shown. (c) L1CAM depletion is paralleled by a loss in NF-kB activity. A representative experiment of n ¼ 3 is shown. (d) L1CAM was knocked down in Panc-1 cells as described in (a). (e) L1CAM depletion is accompanied by a reduction of IL-1b expression in Panc-1 cells. A representative experiment of n ¼ 4 is shown. (f) L1CAM depletion is paralleled by a loss in NF-kB activity in Panc-1 cells as determined by luciferase reporter assay. Data shown represent the mean±s.d. for n ¼ 2 experiments and a phospho-p65-specific ELISA (a representative of n ¼ 3 experiments is shown). (g) Panc-1 cells were transfected with a L1CAM-specific siRNA or GFP siRNA for control and cell proliferation was determined by MTT assay. A representative experiment of n ¼ 3 is shown. *Po0.05, **Po0.01.

Oncogene L1CAM and NF-kappaB H Kiefel et al 4772 α5-integrin IL-1β NF-κB activity 1.5 8 250 ** GFP siRNA α5-integrin siRNA Actin 6 200 β 1.0 Actin 150 β

/ 4 β 0.5 RLU 100 2 mRNA level mRNA level IL-1 50 5-integrin/ α 0.0 0 % of SV40 control 0 mock L1-FL mock L1-FL mock L1-FL PT45-P1 PT45-P1 PT45-P1

αv-integrin IL-1β NF-κB activity 1.2 15 250 GFP siRNA α 1.0 200 v-siRNA Actin β 0.8 10 150 Actin β /

0.6 RLU

β 100 0.4 5 50 mRNA level mRNA level 0.2 IL-1 v-integrin/ % of SV40 control

α 0 0.0 0 mock L1-FL mock L1-FL mock L1-FL PT45-P1 PT45-P1 PT45-P1

κ ILK IL-1β NF- B activity 1.5 5 150 *** GFP siRNA * 4 ILK siRNA 1.0 100 Actin Actin 3 β β / RLU 0.5 β 2

ILK/ 50 mRNA level mRNA level IL-1 1

0.0 0 % of SV40 control 0 mock L1-FL mock L1-FL mock L1-FL PT45-P1 PT45-P1

ADAM10 IL-1β NF-κB activity 1.2 3.5 150 GFP siRNA 1.0 3.0 ADAM10 siRNA Actin 0.8 2.5 β 100 Actin 2.0 β

0.6 / RLU β 1.5 0.4 1.0 50 mRNA level mRNA level 0.2 IL-1 0.5 ADAM10/ 0.0 0.0 % of SV40 control 0 mock L1-FL mock L1-FL mock L1-FL PT45-P1 PT45-P1 PT45-P1

integrin IL-1β NF-κB activity 1.2 1.2 400 GFP siRNA 1.0 1.0 α5-siRNA 300 Actin 0.8 0.8 αv-siRNA β Actin

0.6 β / 0.6 200 RLU 0.4 β 0.4 100 mRNA level mRNA level 0.2 IL-1 0.2 integrin/

0.0 0.0 % of SV40 control 0 α5 αv Panc-1 Panc-1 Panc-1 Figure 6 IL-1b expression in PT45-P1-L1 requires a5-integrin and ILK but not av-integrin or ADAM10. (a–d) a5-integrin, av- integrin, ILK or ADAM10 were knocked down in the indicated cell lines using specific siRNAs (left panel), and the impact on IL-1b expression was determined by qRT–PCR (middle panel). Data shown are representative of n>4 experiments. The effect of specific siRNA on NF-kB was measured by luciferase reporter assay (right panel). Data shown represent the mean±s.d. for n ¼ 3 experiments. (e) a5-integrin or av-integrin was depleted in Panc-1 cells using specific siRNAs; the effect on IL-1b expression was determined by qRT–PCR and NF-kB actvity was measured by luciferase reporter assay. Data shown are representative of n ¼ 3 experiments. *Po0.05, **Po0.01, ***Po0.001.

to IL-1b expression, there was a threefold reduction depletion of ILK led to a reduction in IL-1b levels and observed after a5-integrin depletion (Figure 6a), accom- concomitantly abolished sustained NF-kB activation in panied by a substantial loss of NF-kB activity PT45-P1-L1-FL cells (Figure 6c). It is noteworthy (Figure 6a). In contrast, knockdown of av-integrins that, in Panc-1 cells, the knockdown of both a5- and had no significant effect on IL-1b message (Figure 6b). av-integrin led to a decrease in IL-1b expression and a NF-kB has been described as a downstream target of reduction in NF-kB activity (Figure 6e), and ILK integrin-mediated integrin-linked kinase (ILK) signal- depletion in Panc-1 cells showed similar effects (data ling (Legate et al., 2006). To analyse whether ILK has a not shown). Together, these data support a central role role in L1CAM-dependent NF-kB activation, ILK for integrin signalling in L1CAM-mediated activation of was knocked down by siRNA transfection. Indeed, the NF-kB pathway in pancreatic cancer cells.

Oncogene L1CAM and NF-kappaB H Kiefel et al 4773 NF-kB activation by L1CAM is independent of L1CAM IL-1β *** cleavage 6 mock It is known that nuclear signalling of L1CAM involves 5 L1-FL mock L1-FL L1-RGE 4 proteolytic activity of ADAM10 and presenilins (Riedle Actin β / 3 α β L1 ecto et al., 2009). We silenced ADAM10 to address the 2 α β mRNA level influence of L1CAM cleavage on IL-1b expression. The IL-1 1 hu Actin depletion of ADAM10 affected neither IL-1b expression 0 PT45-P1 (Figure 6d) nor NF-kB activity (Figure 6d). IgG1 L1-9.3 L1CAM can also be cleaved by ADAM17 (Maretzky L1-35.9 L1-11a et al., 2005; Stoeck et al., 2006a). To formally rule out a possible function of ADAM17, we depleted ADAM17 PT45-P1 EC by specific siRNA. Although ADAM17 was efficiently mock L1-FL L1-RGE depleted in PT45-P1 cells, there was no effect on IL-1b expression (Supplementary Figure 3). Similar results % of Max were obtained when presenilin-1 or presenilin-2 was depleted from cells by specific knockdown (Supplemen- fluorescence: L1CAM tary Figure 3) or when the cells were treated with TAPI- 0 or DAPT to pharmacologically inhibit ADAM and IL-1β NF-κB activity presenilin activity (Supplementary Figure 3). Collec- 5 80 * * 70 tively, these results suggest that the effect of L1CAM on 4 60 50 IL-1b expression and NF-kB activity is independent of Actin 3 β 40 / RLU proteolytic processing by ADAM10/17 and presenilins. β 2 30 20 Instead, it requires integrins and integrin signalling. mRNA level IL-1 1 10

0 % of SV40 control 0

The RGD motif in L1CAM is required for IL-1b mock mock L1-FL L1-FL L1-RGE production and NF-kB activation L1-RGE To further investigate the role of integrin binding in PT45-P1 PT45-P1 L1CAM-mediated NF-kB activation, we first analysed Figure 7 L1-RGE mutant prevents constitutive NF-kB activation whether mAbs to distinct domains of L1CAM could (a) PT45-P1 cells were incubated for 48 h in the presence of L1- block IL-1b expression. MAbs that bind to the first Ig mAb (20 mg/ml) to the indicated domains of L1CAM or an IgG1 domain (L1-9.3), the sixth Ig domain (L1-35.9) or the isotype control. IL-1b expression was measured by qRT–PCR. A representative experiment of n ¼ 3 is shown. (b, c) PT45-P1 cells FNIII repeats (L1-11A) of L1CAM were selected. were retrovirally transduced with mutant L1-RGE. Expression of (Wolterink et al., 2010). A significant effect on IL-1b mutant protein compared with that of L1CAM wild type was production was only observed with mAb L1-39.5 to the analysed by western blot using mAb L1-9.3, and cell surface sixth Ig domain (Figure 7a). This domain contains RGD expression was analysed by fluorescence-activated cell sorting. (d) PT45-P1 cells expressing mock, L1-FL or L1-RGE were site(s) (one in human and two in mice) that are known to compared for IL-1b expression using qRT–PCR analysis. A support a5- and av-integrin-mediated cell binding representative experiment of n ¼ 3 is shown. (e) The same cells (Ruppert et al., 1995; Montgomery et al., 1996; were analysed for NF-kB activity using luciferase assay. Data Oleszewski et al., 1999). To further investigate the role shown represent the mean±s.d. for n ¼ 3 experiments. *Po0.05, of the RGD motif, we expressed an L1-RGE mutant in ***Po0.001. PT45-P1 cells (see Figure 1a). The expression levels, as determined by western blot or fluorescence-activated cell sorting analysis, were comparable between L1-FL and Finally, we analysed IL-1b expression and NF-kB the L1-RGE mutant (Figures 7b and c). To show that activity in PT45-P1-L1-RGE cells and observed that the L1-RGE mutant was indeed impaired in integrin both were severely impaired (Figures 7d and e). These binding, we isolated mutant and wild-type L1CAM results suggest that the L1-RGD motive has to interact protein from cellular lysate and carried out cell binding with RGD-binding integrin to provoke the observed assays on immobilized protein. As expected, wild-type biological effects. L1CAM supported cell binding that was integrin dependent as it was blocked in the presence of an RGDS peptide (Supplementary Figure 4). In contrast, the L1-RGE mutant was unable to support cell binding Discussion by integrins (Supplementary Figure 4). To show that RGE mutagenesis had not altered other In this paper, we generated PT45-P1 cell lines expressing L1CAM-binding functions, we examined the homophi- various functional domains of L1CAM, namely, L1-FL, lic binding to immobilized L1-Fc. Although PT45-P1 L1ecto and L1cyt. As expected, L1ecto cells released mock cells showed weak adhesion, the number of bound large amounts of soluble L1CAM into the medium. We L1-FL- and L1-RGE-expressing cells was significantly observed a clear difference between the effects of L1-FL- increased (Supplementary Figure 4), suggesting that versus L1ecto-expressing cells. Only L1-FL cells sup- homophilic binding of L1CAM is not impaired in the ported enhanced cell proliferation and tumour growth L1-RGE mutant. in vivo and were clearly distinct in the pattern of

Oncogene L1CAM and NF-kappaB H Kiefel et al 4774 regulated genes in vitro. Although L1-FL upregulated integrin ligation are integrated into IL-1b-mediated NF- the expression of IL-11, MDK, S100A4 and IL-1b, this kB activation and which cellular signals are involved. was not seen in cells expressing L1ecto. In ex vivo- On integrin engagement, activated PI3-K has been derived tumours, there was an upregulation of MDK reported to stimulate ILK activity and induce down- and S100A4 by L1ecto observed for reasons presently stream signalling of Akt and GSK3b (Delcommenne unknown. L1-FL- and L1ecto-expressing cells also et al., 1998; Attwell et al., 2003; Hannigan et al., 2005; differed in the level of NF-kB activation. Although Qian et al., 2005). Downstream targets of ILK-mediated L1-FL cells were consistently high in NF-kB activity as Akt signalling include NF-kB (Legate et al., 2006). Our compared with mock or L1cyt cells, the effects of L1ecto results support the involvement of ILK/integrin signal- were more diverse. ling in L1CAM-mediated activation of NF-kB, as ILK IL-1b was described as a potent inducer of NF-kB depletion results in reduced IL-1b expression and activity in PDAC cells (Arlt et al., 2002). Therefore, it NF-kB activity in PT45-P1-L1-FL cells. These data was important to establish a link between elevated suggest that an interaction of L1CAM with RGD- levels of IL-1b in L1-FL cells and the state of NF-kB binding integrins, but not L1CAM cleavage and nuclear activation in these cells. Using neutralizing antibodies to translocation, is crucial for IL-1b expression and the IL-1b and IL-1 receptor, we could provide induction of NF-kB activity. compelling evidence that IL-1b indeed functioned in In support of this conclusion, we observed that an an autocrine manner to induce NF-kB activation. We mAb to the sixth Ig domain of L1CAM (which contains also showed that specific inhibition of NF-kB could the RGD integrin-binding site) was most potent in the revert the augmented proliferation of L1-FL cells inhibition of NF-kB activity in PT45-P1-L1-FL cells. in vitro. Thus, the failure of PT45-P1 cells expressing We also used antibodies to a5-integrins, but the results L1ecto, L1cyt or mock cells to induce massive tumour were rather variable. In addition, an L1-RGE mutant growth in vivo could be due to the lack of IL-1b with an impaired integrin-binding ability could not production and subsequent NF-kB activation. We support IL-1b expression and NF-kB activity in PT45- conclude that the L1CAM-mediated induction P1 cells. A model to summarize our findings is presented of NF-kB activity is an important downstream event in Figure 8. responsible for cell proliferation in vitro and tumour In previous work, we have already shown that the formation in vivo and most likely has a role in RGD integrin-binding site in human L1CAM is other cellular features, such as enhanced cell motility important for nuclear signalling (Gast et al., 2008b). and invasion. Indeed, all these features are promoted by It is also involved in the resistant phenotype of PT45-P1- L1CAM expression in tumours. It should be noted L1-FL cells towards chemotherapeutic drugs (Sebens that L1CAM expression and NF-kB activation are Muerkoster et al., 2009). We reported that L1-RGE- abundant in pancreatic cancer (Muerkoster et al., 2005). expressing human embryonic kidney (HEK293) cells However, it remains to be investigated whether a showed reduced cell–cell binding, cell motility, invasive- correlation between L1CAM expression and NF-kB ness and tumour growth in NOD/Scid mice. The RGE activation is observed in human pancreatic adenocarci- mutation also impaired L1CAM-dependent gene noma in vivo. regulation (Gast et al., 2008b) and abolished the How can L1-FL lead to the activation of IL-1b L1CAM-dependent enhanced proliferation of HEK293 expression? We recently showed that L1CAM cleavage cells (J Wolf unpublished results). In the light of our new by ADAM10 and presenilins is required for nuclear findings, we can explain these results by the failure translocation of L1CAM and L1CAM-dependent gene of L1-RGE to induce efficient NF-kB activity (see regulation (Riedle et al., 2009). We first investigated Figure 8). These results are reminiscent of an earlier whether cleavage and nuclear translocation were intact study showing that binding of endothelial cells to the in PT45-P1 cells. As expected, L1CAM was cleaved by immobilized sixth Ig domain of L1CAM could induce ADAMs and presenilins and was translocated into the NF-kB activity that was blocked by anti-integrin mAbs nucleus. The translocation was enhanced when L1CAM (Reidy et al., 2006). cleavage was induced by PMA treatment of cells. In in this study, we want to propose two distinct contrast to our expectation, we could not show a role for mechanisms of L1CAM signalling. We term the first L1CAM cleavage in the induction of IL-1b expression. mechanism, based on L1CAM-regulated intramem- The knockdown of ADAM10, ADAM17, presinilin-1 brane proteolysis and nuclear translocation of the L1- and presinilin-2, as well as the pharmacological inhibi- cytoplasmic fragment, as ‘forward signalling’. Indeed, tion of shedding proteinases and presenilins by TAPI-0 we showed previously that PT45-P1-L1-FL cells are and DAPT, respectively, did not affect the level of IL- capable of regulating L1CAM-dependent genes, such as 1b. In contrast, the depletion of a5-integrin (but not av- b3-integrin or CRABPII, by ‘forward signalling’ (Riedle integrin) had a profound effect on IL-1b expression in et al., 2009). In this study, we describe a second pathway PT45-P1-L1-FL cells that was similar to the knockdown of L1CAM-mediated signalling by NF-kB activation of L1CAM. In Panc-1, the knockdown of a5-integrin that is cleavage independent, addresses different genes and av-integrin led to a reduction in IL-1b expression and relies on the function of L1-FL as a membrane- and NF-kB activity, suggesting the role of both integrins bound ligand to integrins. We specify this mechanism as in NF-kB activity in these cells. However, little is known ‘reverse signalling’. Recent evidence showed a crucial about the exact mechanisms by which L1CAM and role of this signalling by L1CAM and integrins in

Oncogene L1CAM and NF-kappaB H Kiefel et al 4775 Materials and methods

Cells and DNAs Human PDAC cells Panc-1 or PT45-P1 (Sebens Muerkoster et al., 2007) retrovirally transduced with an empty vector (mock), L1-FL or a fragment encoding the cytoplasmic part of

α5-integrin L1CAM L1CAM (L1cyt) from position 1148G to the C-terminus were RGD described (Riedle et al., 2009). A fragment encoding the

IL-1β mRNA L1CAM ectodomain (L1ecto) from the N-terminus to 1108G was constructed by PCR and inserted into the retroviral vector P pBM-Ires-Puro. The site-directed mutagenesis of L1CAM in NFκB position D566E (RGD to RGE) was described previously (Gast et al., 2008b). All constructs were verified by sequencing. The transduction of cell lines with retroviral vectors and selection with puromycin were described (Stoeck et al., 2006b). IL-1β All cell lines were cultivated in RPMI1640 supplemented with 10% FCS at 37 1C, 5% CO and 100% humidity. α IL-1β Ab 2 IL-1 R α IL-1R Ab Chemicals and antibodies Antibodies to the ectodomain (monoclonal antibody (mAb) L1-11A, a subclone of UJ127.11, mAb L1-9.3, L1-35.9) or cytoplasmic domain (polyclonal Ab pcytL1, mAb 745H7) of human L1CAM were described (Mechtersheimer et al., 2001; Gast et al., 2008a). The antibodies to phospho-p65 or p65 were from Santa Cruz Biotechnologies (Heidelberg, Germany). Anti IL-1b and IL-RI antibodies were acquired from R&D Systems (Wiesbaden, Germany). PMA, metalloprotease inhibitors (BB- 3364 and TAPI-0) and the presenilin inhibitor DAPT were α5-integrin L1-RGE RGE described previously (Riedle et al., 2009). All other compounds were from Calbiochem (Darmstadt, Germany) or purchased from Sigma-Aldrich (Munich, Germany).

Analysis of L1CAM shedding For analysis of L1CAM ectodomain shedding, 1 Â 105 cells were seeded into six-well plates. Where indicated, cells were treated with PMA, metalloprotease inhibitor or the presenilin inhibitor DAPT. The release of the L1CAM ectodomain into the supernatant was measured by ELISA (Mechtersheimer et al., 2001).

Figure 8 A model for L1CAM-dependent NF-kB activation Biochemical analysis and the role of integrins. (a) Cells expressing both L1CAM and a5-integrin at the cell surface show elevated levels of IL-1b Cells were seeded into six-well plates and treated as indicated. expression at the mRNA and protein level. Secreted IL-1b binds For cell lysis, 100 mlof2Â NuPage LDS sample buffer back in an autocrine manner to the IL-1b-receptor. Triggering of (Invitrogen, Karlsruhe, Germany) was added per well and the the receptor leads to constitutive NF-kB activation. Soluble lysates were sonicated. Protein concentrations were determined L1CAM alone does not support this constitutive activation. using the BCA protein assay (Thermo Fisher Scientific, (b) The depletion of a5-integrin or L1CAM abrogates signalling Schwerte, Germany). Sodium dodecyl sulphate-poly acryla- in a manner similar to the mutation of the L1CAM integrin mide gel electrophoresis and transfer of separated proteins to binding site from RGD to RGE. Immobilon membranes using semi-dry blotting have been described before (Gutwein et al., 2003). tumour progression. Blockade of L1CAM binding to Flow cytometry Staining of cells with mAbs and with PE-conjugated secondary integrins abrogates L1CAM-induced tumour growth antibodies has been described (Mechtersheimer et al., 2001; and reduces L1CAM-mediated chemoresistance (Sebens Issa et al., 2009). Muerkoster et al., 2007; Gast et al., 2008a). Both modes of L1CAM signalling may be functionally active in siRNA transfection tumour cells that often express both L1CAM and PT45-P1 or Panc-1 cells were transfected with Interferin integrins. The mode of L1CAM signalling that is used (Polyplus, Illkirch, France) according to the manufacturer’s by tumour cells under various conditions remains to be protocol. siRNA sequences are listed in the Supplementary studied in more detail. The data presented in this study Material. are an important hallmark in understanding the different L1CAM signalling mechanisms and their DNA chip analysis impact on the malignancy of PDACs and possibly of The expression analysis of PT45-P1 mock and L1CAM- other cancers. domain-transduced cells using the human Sentrix-6 v3

Oncogene L1CAM and NF-kappaB H Kiefel et al 4776 BeadChip was performed by the expression profiling core In vivo experiments facility (Dr Bernhard Korn) of the German Cancer Research Six-week-old NOD/Scid female mice were injected subcuta- Center, Heidelberg. The detailed procedure has been described neously with 5  106 PT45-P1 cells. At different time points, before (Wolterink et al., 2010). Two independent biological the tumour size was measured and volume was calculated replicates were analysed. Genes with fold change of p –1.5 or using the formula Volume ¼ (Length  Width2)p/6. Tumour X 1.5 were regarded as regulated and the gene expression of growth was monitored for the indicated time period; the mice selected candidates was verified using qRT–PCR. were then killed and tumours were collected for RNA extraction and lysis. All experiments were carried out under the German animal protection law and were approved by local Quantitative real-time PCR authorities. qRT–PCR was performed as described before (Riedle et al., 2009). Primers for qRT–PCR were designed using the IDT Statistical analysis primer quest programme and were produced by MWG For the analysis of statistical significance, Student’s t-test was (Ebersberg, Germany). b-Actin was used as an internal used. P-values in the figures are indicated as follows: * 0.05, standard. The sequences of primers used are available on o ** 0.01 *** 0.001. request. o o

Luciferase assay Abbreviations Cells (1 Â 105) were seeded into six-well plates and grown for 24 h and then subjected to jetPEI transfection (Polyplus, ADAM, A Disintegrin And Metalloprotease; IL-1b, Inter- Illkirch, France) using 1 mg pGL3-(kB)3-Luc (firefly luciferase; leukin 1 beta; L1CAM, L1 cell adhesion molecule; L1-FL, T. Hofmann, DKFZ, Heidelberg, Germany) or pGL3-SV40- L1CAM full length; L1ecto, the ectodomain of L1CAM; Luc (firefly luciferase) with 0.02 mg pRLTK (Renilla luciferase; L1cyt, the cytoplasmic fragment of L1CAM; L1-RGE, human Promega, Mannheim, Germany). The assay was carried out as L1CAM with mutations of RGD to RGE; NF-kB, nuclear described by the manufacturer (Promega). factor kappa B; PDAC, pancreatic ductal adenocarcinoma.

ELISA assay Cells (1 Â 106) were seeded into 10 cm plates and were grown Conflict of interest for 24 h. Phospho-NF-kB levels were quantified using PathS- can Phospho-NF-kB p65 ELISA (Cell Signalling, Frankfurt, The authors declare no conflict of interest. Germany). The assay was performed according to the manufacturer’s instructions. Levels of secreted human IL-1b were quantified using the Quantikine-HS human IL-1b Acknowledgements immunoassay (R&D Systems, Wiesbaden, Germany). ELISA 0 was performed according to the manufacturer s instructions. We acknowledge the competent help of Dr Alex Stoeck Measured IL-1b concentrations were normalized to cell (University of Michigan, Ann Arbor) in retroviral expression numbers. in the early phase of the study. We also thank Ramona Straub and Dagmar Leisner for excellent technical assistance. This Cell adhesion assay study was supported by grants from Deutsche Krebshilfe The cell adhesion assay has been described before (Schwerpunktprogramm: Invasion and Migration), the (Mechtersheimer et al., 2001). Deutsche Forschungsgemeinschaft project nr. SE-1831/2-1 to S.S. and the EU-FP6 framework program OVCAD project nr. PE-14034 to PA. Further financial support was received MTT proliferation assay from a collaborative research grant from Medigene Inc. This assay was described before (Novak-Hofer et al., 2008). (Munich) to PA.

References

Arlt A, Vorndamm J, Muerkoster S, Yu H, Schmidt WE, Folsch UR Brummendorf T, Kenwrick S, Rathjen FG. (1998). Neural cell et al. (2002). Autocrine production of interleukin 1beta confers recognition molecule L1: from cell biology to human hereditary constitutive nuclear factor kappaB activity and chemoresistance in brain malformations. Curr Opin Neurobiol 8: 87–97. pancreatic carcinoma cell lines. Cancer Res 62: 910–916. Delcommenne M, Tan C, Gray V, Rue L, Woodgett J, Dedhar S. Attwell S, Mills J, Troussard A, Wu C, Dedhar S. (2003). Integration (1998). Phosphoinositide-3-OH kinase-dependent regulation of of cell attachment, cytoskeletal localization, and signaling by glycogen synthase kinase 3 and protein kinase B/AKT by the integrin-linked kinase (ILK), CH-ILKBP, and the tumor suppressor integrin-linked kinase. Proc Natl Acad Sci USA 95: 11211–11216. PTEN. Mol Biol Cell 14: 4813–4825. Fogel M, Gutwein P, Mechtersheimer S, Riedle S, Stoeck A, Smirnov Bao S, Wu Q, Li Z, Sathornsumetee S, Wang H, McLendon RE et al. A et al. (2003a). L1 expression as a predictor of progression and (2008). Targeting cancer stem cells through L1CAM suppresses survival in patients with uterine and ovarian carcinomas. Lancet glioma growth. Cancer Res 68: 6043–6048. 362: 869–875. Boo YJ, Park JM, Kim J, Chae YS, Min BW, Um JW et al. (2007). L1 Fogel M, Mechtersheimer S, Huszar M, Smirnov A, Abu DA, Tilgen expression as a marker for poor prognosis, tumor progression, and W et al. (2003b). L1 adhesion molecule (CD 171) in development short survival in patients with colorectal cancer. Ann Surg Oncol 14: and progression of human malignant melanoma. Cancer Lett 189: 1703–1711. 237–247.

Oncogene L1CAM and NF-kappaB H Kiefel et al 4777 Gast D, Riedle S, Issa Y, Pfeifer M, Beckhove P, Sanderson MP et al. Meier F, Busch S, Gast D, Goppert A, Altevogt P, Maczey E et al. (2008a). The cytoplasmic part of L1-CAM controls growth and gene (2006). The adhesion molecule L1 (CD171) promotes melanoma expression in human tumors that is reversed by therapeutic progression. Int J Cancer 119: 549–555. antibodies. Oncogene 27: 1281–1289. Montgomery AM, Becker JC, Siu CH, Lemmon VP, Cheresh DA, Gast D, Riedle S, Kiefel H, Sebens Muerko¨ster S, Scha¨fer H, Pancook JD et al. (1996). Human neural cell adhesion molecule L1 Scha¨fer MKE et al. (2008b). The RGD integrin binding site in and rat homologue NILE are ligands for integrin alpha v beta 3. human L1-CAM is important for nuclear signaling. Exp Cell Res J Cell Biol 132: 475–485. 314: 2411–2418. Moos M, Tacke R, Scherer H, Teplow D, Fruh K, Schachner M. Gast D, Riedle S, Schabath H, Schlich S, Schneider A, Issa Y et al. (1988). Neural adhesion molecule L1 as a member of the (2005). L1 augments cell migration and tumor growth but not immunoglobulin superfamily with binding domains similar to beta3 integrin expression in ovarian carcinomas. Int J Cancer 115: fibronectin. Nature 334: 701–703. 658–665. Muerkoster S, Arlt A, Sipos B, Witt M, Grossmann M, Kloppel G Gavert N, Conacci-Sorrell M, Gast D, Schneider A, Altevogt P, et al. (2005). Increased expression of the E3-ubiquitin ligase receptor Brabletz T et al. (2005). L1, a novel target of beta-catenin signaling, subunit betaTRCP1 relates to constitutive nuclear factor-kappaB transforms cells and is expressed at the invasive front of colon activation and chemoresistance in pancreatic carcinoma cells. cancers. J Cell Biol 168: 633–642. Cancer Res 65: 1316–1324. Gavert N, Sheffer M, Raveh S, Spaderna S, Shtutman M, Neoptolemos JP, Stocken DD, Friess H, Bassi C, Dunn JA, Hickey H Brabletz T et al. (2007). Expression of L1-CAM and ADAM10 et al. (2004). A randomized trial of chemoradiotherapy and in human colon cancer cells induces metastasis. Cancer Res 67: chemotherapy after resection of pancreatic cancer. N Engl J Med 7703–7712. 350: 1200–1210. Geismann C, Morscheck M, Koch D, Bergmann F, Ungefroren H, Novak-Hofer I, Cohrs S, Grunberg J, Friedli A, Schlatter MC, Arlt A et al. (2009). Up-regulation of L1CAM in pancreatic duct Pfeifer M et al. (2008). Antibodies directed against cells is transforming growth factor beta1- and slug-dependent: role L1-CAM synergize with Genistein in inhibiting growth and survival in malignant transformation of pancreatic cancer. Cancer Res 69: pathways in SKOV3ip human ovarian cancer cells. Cancer Lett 261: 4517–4526. 193–204. Gouveia RM, Gomes CM, Sousa M, Alves PM, Costa J. (2008). Oleszewski M, Beer S, Katich S, Geiger C, Zeller Y, Rauch U et al. Kinetic analysis of L1 homophilic interaction: role of the first four (1999). Integrin and neurocan binding to L1 involves distinct Ig immunoglobulin domains and implications on binding mechanism. domains. J Biol Chem 274: 24602–24610. J Biol Chem 283: 28038–28047. Primiano T, Baig M, Maliyekkel A, Chang BD, Fellars S, Sadhu J Gutwein P, Mechtersheimer S, Riedle S, Stoeck A, Gast D, Joumaa S et al. (2003). Identification of potential anticancer drug targets et al. (2003). ADAM10-mediated cleavage of L1 adhesion molecule through the selection of growth-inhibitory genetic suppressor at the cell surface and in released membrane vesicles. Faseb J 17: elements. Cancer Cell 4: 41–53. 292–294. Qian Y, Zhong X, Flynn DC, Zheng JZ, Qiao M, Wu C et al. (2005). ILK mediates actin filament rearrangements and cell migration Gutwein P, Oleszewski M, Mechtersheimer S, Agmon-Levin N, and invasion through PI3K/Akt/Rac1 signaling. Oncogene 24: Krauss K, Altevogt P. (2000). Role of Src kinases in the ADAM- 3154–3165. mediated release of L1 adhesion molecule from human tumor cells. Reidy M, Zihlmann P, Hubbell JA, Hall H. (2006). Activation of cell- J Biol Chem 275: 15490–15497. survival transcription factor NFkappaB in L1Ig6-stimulated en- Hannigan G, Troussard AA, Dedhar S. (2005). Integrin-linked kinase: dothelial cells. J Biomed Mater Res A 77: 542–550. a cancer therapeutic target unique among its ILK. Nat Rev Cancer Riedle S, Kiefel H, Gast D, Bondong S, Wolterink S, Gutwein P et al. 5: 51–63. (2009). Nuclear translocation and signalling of L1-CAM in human Hortsch M. (2000). Structural and functional evolution of the L1 carcinoma cells requires ADAM10 and presenilin/gamma-secretase family: are four adhesion molecules better than one? Mol Cell activity. Biochem J 420: 391–402. Neurosci 15: 1–10. Ruppert M, Aigner S, Hubbe M, Yagita H, Altevogt P. (1995). The L1 Issa Y, Nummer D, Seibel T, Muerkoster SS, Koch M, Schmitz- adhesion molecule is a cellular ligand for VLA-5. J Cell Biol 131: Winnenthal FH et al. (2009). Enhanced L1CAM expression on 1881–1891. pancreatic tumor endothelium mediates selective tumor cell Schneider G, Siveke JT, Eckel F, Schmid RM. (2005). transmigration. J Mol Med 87: 99–112. Pancreatic cancer: basic and clinical aspects. Gastroenterology 128: Jemal A, Siegel R, Ward E, Murray T, Xu J, Thun MJ. (2007). Cancer 1606–1625. statistics, 2007. CA Cancer J Clin 57: 43–66. Sebens Muerkoster S, Kotteritzsch J, Geismann C, Gast D, Kruse ML, Kaifi JT, Reichelt U, Quaas A, Schurr PG, Wachowiak R, Altevogt P et al. (2009). alpha5-integrin is crucial for L1CAM- Yekebas EF et al. (2007). L1 is associated with micrometastatic mediated chemoresistance in pancreatic adenocarcinoma. Int J spread and poor outcome in colorectal cancer. Mod Pathol 20: Oncol 34: 243–253. 1183–1190. Sebens Muerkoster S, Werbing V, Sipos B, Debus MA, Witt M, Legate KR, Montanez E, Kudlacek O, Fassler R. (2006). ILK, PINCH Grossmann M et al. (2007). Drug-induced expression of the cellular and parvin: the tIPP of integrin signalling. Nat Rev Mol Cell Biol 7: adhesion molecule L1CAM confers anti-apoptotic protection and 20–31. chemoresistance in pancreatic ductal adenocarcinoma cells. Onco- Maddaluno L, Verbrugge SE, Martinoli C, Matteoli G, Chiavelli A, gene 26: 2759–2768. Zeng Y et al. (2009). The adhesion molecule L1 regulates Silletti S, Yebra M, Perez B, Cirulli V, McMahon M, transendothelial migration and trafficking of dendritic cells. J Exp Montgomery AM. (2004). Extracellular signal-regulated Med 206: 623–635. kinase (ERK)-dependent gene expression contributes to L1 cell Maretzky T, Schulte M, Ludwig A, Rose-John S, Blobel C, adhesion molecule-dependent motility and invasion. J Biol Chem Hartmann D et al. (2005). L1 is sequentially processed by two 279: 28880–28888. differently activated metalloproteases and presenilin/gamma-secre- Stoeck A, Keller S, Riedle S, Sanderson MP, Runz S, Le Naour F et al. tase and regulates neural cell adhesion, cell migration, and neurite (2006a). A role for exosomes in the constitutive and outgrowth. Mol Cell Biol 25: 9040–9053. stimulus-induced ectodomain cleavage of L1 and CD44. Biochem Mechtersheimer S, Gutwein P, Agmon LN, Stoeck A, Oleszewski M, J 393: 609–618. Riedle S et al. (2001). Ectodomain shedding of L1 adhesion Stoeck A, Schlich S, Issa Y, Gschwend V, Wenger T, Herr I et al. molecule promotes cell migration by autocrine binding to integrins. (2006b). L1 on ovarian carcinoma cells is a binding partner for J Cell Biol 155: 661–674. Neuropilin-1 on mesothelial cells. Cancer Lett 239: 212–226.

Oncogene L1CAM and NF-kappaB H Kiefel et al 4778 Thelen K, Kedar V, Panicker AK, Schmid RS, Midkiff BR, Maness associated with metastasis in cutaneous malignant melanoma. Eur PF. (2002). The neural cell adhesion molecule L1 potentiates J Cancer 38: 1708–1716. integrin-dependent cell migration to extracellular matrix proteins. Wolterink S, Moldenhauer G, Fogel M, Kiefel H, Pfeifer M, Luttgau S J Neurosci 22: 4918–4931. et al. (2010). Therapeutic antibodies to human L1CAM: functional Thies A, Schachner M, Moll I, Berger J, Schulze HJ, Brunner G et al. characterization and application in a mouse model for ovarian (2002). Overexpression of the cell adhesion molecule L1 is carcinoma. Cancer Res 70: 2504–2515.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene