CleavageofL1inExosomesandApoptoticMembraneVesicles ... · endometrial carcinomas in a stage-dependent manner and that L1 expression was a predictor of poor outcome (9). It was proposed

Cleavage of L1in Exosomes and ApoptoticMembraneVesiclesReleased fromOvarian Carcinoma CellsPaul Gutwein,1Alexander Stoeck,1Svenja Riedle,1Daniela Gast,1Steffen Runz,1Thomas P. Condon,4

AlexanderMarme¤ ,2 Minh-Chau Phong,3 Otwin Linderkamp,4 Alexander Skorokhod,1and PeterAltevogt1

Abstract Purpose:The L1adhesionmolecule (CD171) is overexpressed inhumanovarian and endometrialcarcinomas and is associated with bad prognosis. Although expressed as a transmembranemol-ecule, L1is released from carcinoma cells in a soluble form. Soluble L1is present in serum and as-cites of ovarian carcinoma patients.We investigated the mode of L1cleavage and the function ofsoluble L1.Experimental Design:Weusedovarian carcinoma cell lines and ascites fromovarian carcinomapatients toanalyze solubleL1andL1cleavagebyWesternblot analysis andELISA.Results:We find that in ovarian carcinoma cells the constitutive cleavage of L1proceeds in se-cretory vesicles.We show that apoptotic stimuli like C2-ceramide, staurosporine, UV irradiation,and hypoxic conditions enhance L1-vesicle release resulting in elevated levels of soluble L1.Constitutive cleavage of L1 is mediated by a disintegrin and metalloproteinase 10, but underapoptotic conditions multiple metalloproteinases are involved. L1cleavage occurs in two types ofvesicles with distinct density features: constitutively released vesicles with similarity to exosomesand apoptotic vesicles. Both types of L1-containing vesicles are present in the ascites fluids ofovarian carcinoma patients. Soluble L1 from ascites is a potent inducer of cell migration andcan trigger extracellular signal-regulated kinase phosphorylation.Conclusions:We suggest that tumor-derived vesicles may be an important source for soluble L1that could regulate tumor cell function in an autocrine/paracrine fashion.

L1 is the prototype of a neural subfamily of cell adhesionmolecules structurally belonging to the immunoglobulin (Ig)superfamily (1). In humans, L1 is a 200- to 220-kDatransmembrane glycoprotein composed of six Ig-like domainsand five fibronectin type III repeats. L1 plays a crucial role inaxon guidance and cell migration in the developing nervoussystem (2, 3). L1 is not only expressed in the nervous systembut is also found on different tumor cells like lung cancer (4),gliomas (5), melanomas (6, 7), and renal carcinoma (8). Werecently reported that L1 is overexpressed in ovarian andendometrial carcinomas in a stage-dependent manner and thatL1 expression was a predictor of poor outcome (9). It wasproposed that L1 could promote carcinoma progression byaugmenting tumor cell migration on extracellular matrixcompounds (9, 10).

Although a transmembrane molecule, L1 can be cleavedand released in a soluble form into the extracellular space(10, 11). Importantly, soluble L1 was specifically detected inserum samples of ovarian and uterine cancer patients (9).The cleavage of the ectodomain of transmembrane moleculesis termed ectodomain shedding (12), and it affects many cellsurface molecules such as growth factors like membrane-anchored heparin-binding epidermal growth factor– likegrowth factor (proHB-EGF; refs. 13, 14), transforming growthfactor-a (15), membrane receptors like Her2/neu (16), andadhesion molecules like L-selectin (17, 18) and h-amyloidprecursor protein (19, 20). Shedding can occur in aconstitutive fashion or can be induced. Common unphysio-logic activators of ectodomain shedding are phorbol esters(21–23), pervanadate treatment, and cholesterol extractingagents such as methyl-h-cyclodextrin (24–26). Several phys-iologic activators of ectodomain shedding such as chemotac-tic peptides, cytokines, and growth factors have been alsodescribed (27–29). The activation of G-protein–coupledreceptors can induce shedding of growth factors that in turnmediate the activation of tyrosine kinase receptors such asthe epidermal growth factor (EGF) receptor (27, 30). Weshowed before that shedding of L1 can be induced throughstimuli like phorbol 12-myristate 13-acetate, pervanadate,and methyl-h-cyclodextrin (10, 11, 24). A recent study hasshown that in renal carcinoma the shedding of L1 isenhanced by hepatocyte growth factor (8). Other physiologicstimuli that lead to enhanced L1 shedding have not beenidentified yet.

www.aacrjournals.orgClin Cancer Res 2005;11(7) April 1, 2005 2492

Authors’ Affiliations: 1Tumor Immunology Programme, D010, GermanCancer Research Center ; 2University Hospital for Gynecology and Obstetrics,and 3Pedriatics University of Heidelberg, Heidelberg, Germany; and 4ISISPharmaceuticals Carlsbad, CaliforniaReceived 8/24/04; revised10/13/04; accepted12/7/04.Grant support:Deutsche Krebshilfe grant10-1307-3Al (P. Altevogt).Note: P. Gutwein and A. Stoeck contributed equally to this work.The costs of publication of this article defrayed in part by the payment of pagecharges. This article must therefore be hereby marked advertisement in accordancewith18 U.S.C. Section1734 solely to indicate this fact.Requests for reprints: Peter Altevogt, Tumor Immunology Programme,G0100, German Cancer Research Center, Im Neuenheimer Feld 280, D-69120Heidelberg, Germany. Phone: 49-6221-423725; Fax: 49-6221-423702; E-mail:P.Altevogt@dkfz-de.

F2005 American Association for Cancer Research.

Human Cancer Biology

Cancer Research. on December 5, 2020. © 2005 American Association forclincancerres.aacrjournals.org Downloaded from

http://clincancerres.aacrjournals.org/

Another shedding event frequently observed in tumor cells isthe ability to release intact membrane vesicles into theenvironment (31). The precise mechanism of membraneshedding and the origin of released vesicles are not known,however, recent work has characterized at least two distincttypes of vesicles: apoptotic blebs and exosomes (32, 33). Therelease of apoptotic blebs is initiated shortly after the inductionof apoptotic cell death (33), whereas exosomes representmembrane constituents that are released from live cells viathe fusion of multivesicular endosomes with the plasmamembrane (33, 34). We reported before that spontaneousvesicles released from cells transfected with the neural form ofL1 contained L1 and the cleavage proteinase a disintegrin andmetalloproteinase 10 (ADAM10; ref. 24). In the presentcommunication we have investigated the shedding of L1 inovarian carcinoma cells in vitro and in vivo. We show that theconstitutive cleavage is mediated by ADAM10 and proceeds insecretory vesicles. Shedding is increased by apoptotic stimuli viathe release of membrane blebs in which multiple metal-loproteinases are active. We provide evidence that both types ofL1-containing vesicles occur in the ascites of ovarian carcinomapatients and that soluble L1 in the ascites can modulate thefunction of carcinoma cells. Our results suggest an importantrole for vesicles in the release of soluble L1 from tumor cells.

Materials andMethods

Cells. HeLa cells were obtained from the tumor bank of theGerman Cancer Research Center (Heidelberg, Germany). The ovariancarcinoma cell line OVMz and Chinese hamster ovary (CHO) cellsstably transfected with human L1 (CHO-hL1) were described before(10). Cells were cultivated in DMEM supplemented with 10% fetalbovine serum at 37jC, 5% CO2, and 100% humidity.

Chemicals and antibodies. Antibodies to the ectodomain [L1-11A,subclone of monoclonal antibody (mAb) UJ 127.11] or the cytoplasmicdomain (pcytL1) of human L1 were described (10). Antibodies #2547to the ectodomain and #71 to the cytoplasmic tail of ADAM10 weredescribed (24). Additional polyclonal antibodies to ADAM17 andADAM10 or mAbs to integrins were obtained from Chemicon(Hofheim, Germany). MAb Syb-1 to human CD9 was a gift from Dr.Eric Rubinstein. Antibodies to extracellular signal-regulated kinase(ERK) and phospho-ERK were purchased from BD Transduction(Heidelberg, Germany). Secondary antibody Alexa 488–conjugatedgoat anti-mouse was purchased from Molecular Probes (Leiden,The Netherlands) and Cy3-conjugated goat anti-rabbit antibodywas obtained from Dianova (Hamburg, Germany). 4V,6-Diamidino-2-phenylindole, staurosporine, a cell-permeable C2-ceramide, anddoxorubicin were purchased from Sigma (Taufkirchen, Germany).Metalloproteinase inhibitors tumor necrosis factor- protease inhibitor-0(TAPI-0) and TAPI-1 were from Calbiochem (Bad Soden, Germany),Triton X-100 was from Gerbu (Gaiberg, Germany), and caspaseinhibitor ZVAD-fmk was from BACHEM (Weil am Rhein, Germany).A stock solution of pervanadate was prepared as described before (11).

Antisense strategy. The selection of potent and specific antisenseoligonucleotide inhibitors of human ADAM10 (ISIS 100750) andADAM17 (ISIS 16337) together with the mismatched controls ISIS108030 (for ADAM10) and ISIS 17337 (for ADAM17) has beendescribed (35). Cells were transfected using oligofectamine (Invi-trogen, Karlstuhe, Germany) and were analyzed 48 hours later.

Analysis of L1 shedding and induction of apoptosis. Cells (100,000cells per well) were cultured in duplicate in six-well culture platesand after washing once with PBS incubated for 18 hours with orwithout apoptotic stimuli in serum-free medium. Staurosporine(2 Amol/L), C2-ceramide (40 Amol/L), pervanadate (200 Amol/L),

doxorubicin (5 Ag/mL), and UV irradiation were used as apoptoticreagents. Metalloproteinase inhibitors TAPI-1 and TAPI-0 (10 Amol/L)and caspase inhibitor ZVAD-fmk (25 Amol/L) were added 30 minutesbefore treatment. After 18 hours, the supernatant was removed andcells were scraped into Eppendorf tubes and centrifuged. Cell pelletwas lysed in lysis buffer [20 mmol/L Tris-HCl (pH 8.0) containing1% Triton X-100, 150 mmol/L NaCl, and 1 mmol/L phenyl-methylsulfonyl fluoride], cleared by centrifugation, and mixed with2-fold concentrated reducing SDS sample buffer. Supernatants oftreated cells were collected and centrifuged for 10 minutes at 1,200 �g and for 20 minutes at 10,000 � g to remove cellular debris.Membrane vesicles were collected by centrifugation at 100,000 � gfor 2 or 18 hours at 4jC using a Beckman SW 40 rotor. Vesicleswere directly dissolved in SDS sample buffer or processed forgradient centrifugation (see below). For hypoxia studies cells wereincubated under hypoxic conditions (1% O2) using a Reming Bio-instruments chamber and oxygen regulator (Reming Bioinstruments,Redfield, NY).

Assessment of apoptotic cell death. For the analysis of DNA

fragmentation, the cell pellet was washed once in PBS (pH 7.4) and

lysed in a hypotonic lysis buffer (0.1% sodium citrate, 0.1% Triton X-100, 50 Ag/mL propidium iodide) at 4jC overnight. The nuclei were

then analyzed for DNA content by flow cytometry (36).Isolation of vesicles from ascites fluid. Analysis of patient tumor

material was under approval of the ethics commission of the University

of Heidelberg. Ascites from ovarian carcinoma patients were clearedfrom cells and debris by two rounds of centrifugation as described

above. The cleared ascites was then overlaid on a 40% sucrose cushion

and centrifuged at 100,000 � g for 90 minutes. The interphase wasremoved and pelleted by ultracentrifugation.

Sucrose density gradient fractionation. Pelleted vesicles were resus-

pended in PBS and loaded on top of a step gradient comprising layersof 2, 1.3, 1.16, 0.8, 0.5, and 0.25 mol/L sucrose as described (24). The

gradients were centrifuged for 2.5 hours at 100,000 � g in a BeckmanSW40 rotor. Twelve 1-mL fractions were collected from the top of the

gradient and precipitated by chloroform/methanol. Samples were

analyzed by SDS-PAGE and Western blotting as described below.

Immunofluorescence. For immunofluorescent staining, cells were

grown on coverslips and fixed for 20 minutes with 4% paraformalde-hyde-PBS at room temperature. Cells were washed in PBS and

permeabilized with 0.1% Saponin in PBS containing 5% goat serum

for 15 minutes at room temperature. Cells were then incubated for 1hour with primary antibodies (L1-11A supernatant undiluted and

#2547 1:500 dilution). After three washing steps with PBS, cells wereincubated for 30 minutes in the dark with Alexa 488–conjugated goat

anti-mouse IgG or Cy3-conjugated goat anti-rabbit IgG. After washing

the cells twice with PBS, nuclei were stained with 4V,6-diamidino-2-phenylindole. Stained cells were mounted on glass slides and examined

with an epifluorescence microscope (Axioplan-2, Zeiss, Oberkochem,Germany).

Transmigration assays. This assay has been described before (10).

Briefly, fibronectin or bovine serum albumin for control was coated onthe backside of Transwell chambers (6.5 mm diameter, 5 Am pore size;

Costar, Cambridge, MA). Cells in RPMI 1640 containing 0.5 % bovine

serum albumin were seeded into the upper chamber and allowed totransmigrate to the lower compartment. For mAb blocking, the cells

were preincubated at 10 Ag/mL with the respective mAb. Transmigra-tion was quantified using crystal violet staining. Each determination

was done in quadruplicate and the data are given as mean values F SE.Biochemical analysis. The isolation of soluble L1 from culture

supernatant was described before (37, 38) and was adapted here forascites. The vesicle-cleared ascites fluid was adsorbed to Sepharose-linked mAb L1-11A and bound L1 was eluted with 0.1 mol/L glycine-HCl buffer (pH 2.8). Eluted fractions were neutralized and an aliquot ofthe samples was separated by SDS-PAGE under reducing conditions andtransferred to an Immobilon membrane using semidry blotting. Afterblocking with 5% skim milk in TBS, the blots were developed with the

www.aacrjournals.org Clin Cancer Res 2005;11(7) April 1, 20052493

Shedding of L1AdhesionMolecule



respective primary antibody followed by peroxidase-conjugated sec-ondary antibody and enhanced chemiluminescence (ECL) detection.

Statistical analysis. For the analysis of statistical significance theStudent’s t test was used.

Results

Apoptotic stimuli enhance the cleavage of L1 adhesionmolecule. Recent work has shown that adhesion moleculessuch as E-cadherin, L-selectin, and platelet/endothelial celladhesion molecule 1 are cleaved off the membrane duringapoptosis (39–41). We examined whether apoptotic stimulicould augment the release of L1. As shown in Fig. 1A, treatmentof CHO-hL1 cells with the apoptotic inducers C2-ceramide orstaurosporine significantly augmented the amount of releasedL1 in the medium. Parallel measurements of apoptosisconfirmed that both compounds induced apoptosis (notshown). Doxorubicin did not induce L1 release and did notcause apoptosis under the assay conditions. TAPI-0 treatmentpartially reduced the amount of C2-ceramide– and staurospor-ine-released L1 (Fig. 1A). Preincubating with the caspase

inhibitor ZVAD-fmk blocked the C2-ceramide–induced L1shedding only weakly (f20%), but strongly reduced thestaurosporine-induced L1 release (f65 %). As expected,ZVAD-fmk reduced the rate of apoptosis (data not shown).

In the ovarian carcinoma cell line OVMz essentially similarresults were observed (Fig. 1B). In contrast to CHO-hL1 cells,the strongest inducer was staurosporine but the amount ofsoluble L1 was only increased 2.5-fold (Fig. 1B). Importantly,all apoptotic stimuli caused a metalloproteinase-mediatedcleavage of L1 as it was blocked by TAPI (Fig. 1B).

C2-ceramide enhances the formation of membrane vesiclescontaining ADAM10 and L1. For CHO-hL1 cells the sizeheterogeneity and the partial inhibition by TAPI-0 suggested thatL1 released through apoptosis was composed of two forms: atruly soluble L1-200 form (membrane cleaved) and a full-lengthform (L1-220) in membrane vesicles. Both forms are only poorlyresolved by SDS-PAGE (24). We investigated whether membranevesicle release was indeed enhanced in cells treated withapoptotic stimuli. When L1-expressing HeLa cells were treatedwith C2-ceramide for 6 hours we noticed that the tumor cellsformed membrane blebs (Fig. 2A). Bleb formation is a well-known characteristic for apoptosis. We analyzed the location ofL1 and ADAM10 in C2-ceramide–treated cells using fluores-cence microscopy. As shown in Fig. 2B, in untreated control cellsADAM10 was expressed mostly within the cell whereas L1staining was seen both inside the cell and at the plasmamembrane. Treatment with C2-ceramide led to the accumulationof ADAM10 in the lumen of the membrane blebs (Fig. 2B , redarrows) and a prominent staining of L1 in membrane blebs.

Apoptosis-induced L1 cleavage occurs in membrane vesicles.Membrane vesicles were isolated from the supernatant of liveand apoptotic cells. As shown in Fig. 3A we detected L1-220and the cleavage fragment L1-32 in membrane vesicles fromCHO-hL1 cells. The cleavage was metalloproteinase mediatedas TAPI pretreatment reduced the amount of the L1-32fragment. The analysis of OVMz cells revealed essentiallysimilar results (Fig. 3B).

Hypoxic conditions lead to apoptosis of tumor cells and augmentthe release of soluble L1. In malignant tumors the rate ofapoptosis is high in undervascularized areas (42, 43). It isknown that low oxygen pressure or hypoxia can directly induceapoptosis of tumor cells (44). As shown in Fig. 4A, incubationof CHO-hL1 under hypoxic conditions led to enhancedapoptosis (from 17% to 69 %). Interestingly, we observed thatthe presence of TAPI increased the apoptosis rate up to 88%.Similar effects could be observed in OVMz cells. Incubationunder hypoxic conditions increased the rate of apoptosis in thelatter cells from 11% to 25 %. As shown in Fig. 4B, the presenceof TAPI rendered OVMz cells extremely sensitive to apoptosisunder hypoxic conditions (65% apoptosis). The analysis ofsupernatants revealed that hypoxic conditions enhanced thelevel of soluble L1. In CHO-hL1 (Fig. 4C) and OVMz (Fig. 4D)cells, there was an f2.5-fold increase of soluble L1 in thesupernatant. Notably, the enhanced cleavage was again blockedby TAPI-0 to basal levels (Fig. 4C and D).

Cleavage of L1 under apoptosis is mediated by multiplemetalloproteinases. ADAM10 has been identified as an im-portant metalloproteinase for L1 cleavage (10, 24). To examinethe role in apoptosis-induced shedding, we used antisenseoligonucleotides specific for ADAM10 and ADAM17. Biochem-ical analysis indicated that the antisense oligonucleotides


Fig. 1. Apoptotic stimuli induce metalloproteinase-mediated cleavage of L1. A,CHO-hL1cells were incubated for18 hours with 40 Amol/LC2-ceramide, 5 Ag/mLdoxorubicin, or 2 Amol/L staurosporine.The metalloproteinase inhibitorTAPI andcaspase inhibitor ZVAD-fmkwere incubated for 30 minutes before adding theapoptotic stimuli. Supernatants were trichloroacetic acid precipitated and analyzedfor L1usingWestern blot analysis with mAb L1-11A against the ectodomain of L1followed by peroxidase-conjugated secondary antibody and ECL detection. B,OVMz cells were incubated for18 hours with 200 Amol/L pervanadate, 40 Amol/LC2-ceramide, or 2 Amol/L staurosporine, or were UV irradiated.Themetalloproteinase inhibitorTAPI-0 (10 Amol/L) was added 30 minutes before.Supernatants were trichloroacetic acid precipitated and analyzed for L1byWesternblotting with mAb L1-11A followed by peroxidase-conjugated secondary antibodyand ECL detection.




diminished the expression level of the respective proteinase(Fig. 5A). The ADAM10 antisense oligonucleotide reduced theconstitutive L1 shedding by 80% as evidenced by a diminishedrelease of L1-200 (Fig. 5B) whereas antisense oligonucleotidesspecific for ADAM17 did not (Fig. 5B). In the presence of C2-ceramide, the effect of ADAM10-specific antisense oligonucleo-tides was clearly abrogated (Fig. 5C). These results confirmedthe role of ADAM10 in the constitutive cleavage of L1 butsuggested that under apoptosis additional metalloproteinasesare activated leading to enhanced L1 release.

Two types of L1-containing vesicles are released from ovariancarcinoma cells. To allow a more refined analysis of L1-containing vesicles, we used sucrose density gradient centrifu-gation as previously described (24). Figure 6B shows an analysisof vesicles derived from OVMz cells cultivated in the absence orpresence of C2-ceramide. Vesicles containing L1 and ADAM10were present in the middle part of the gradient and also athigher density at the bottom of the gradient. Interestingly, anadditional fragment L1-28 possibly generated by furthercleavage of L1-32 was also observed. Strikingly, C2-ceramide

led to an increase in vesicles with higher density appearing atthe bottom of the gradient. As these vesicles are enhanced underapoptotic conditions, they most likely represent apoptoticblebs. In contrast, the amount of vesicles detected in themiddle part of the gradient was unaffected by C2-ceramide.These vesicles are released constitutively and might be related toexosomes. To test this assumption, the gradient fractions wereanalyzed with a mAb to the tetraspan CD9, an establishedmarker for exosomes from dendritic and tumor cells (34).Indeed, as shown in Fig. 6B , the middle fractions of bothgradients were positive for CD9.

L1-containing vesicles in ascites of ovarian carcinoma patients.Next we isolated vesicles from the ascites fluids of ovariancarcinoma patients. The histologic type of the analyzed tumorsis given in the legend to Fig. 7. As shown in Fig. 7A, four of fiveascites contained vesicles with full-length L1-220 and werepositive for soluble L1 by ELISA (Fig. 7B). As revealed bysucrose density centrifugation, the ascites-derived vesiclescontained L1-28 and floated in the bottom of the gradientsimilar in density to apoptotic blebs (Fig. 7C). The exosome


Fig. 2. C2-ceramide induces membraneblebbing in apoptotic cells. A, HeLa cellswere plated on coverslips and incubated for6 hours with 40 Amol/LC2-ceramide.Cells were fixed with paraformaldehydeand analyzed with an epifluorescencemicroscope. Red arrows, C2-ceramideinduces membrane blebbing. B, ADAM10and L1are located inmembrane blebs.Cells were stained with polyclonal antibody#2547 to the ectodomain of ADAM10followed by Cy3-conjugated goatanti-rabbit IgGor withmAb L1-11A followedbyAlexa 488^ conjugated goat anti-mouseIgG secondary antibody.The nuclei werestainedwith 4V,6-diamidino-2-phenylindole.In nonstimulated control cells ADAM10staining is predominantly in the cytoplasm.L1shows a more membranous staining.Treatment with C2-ceramide inducesmembrane blebbing and ADAM10 and L1are now localized inmembrane blebs(red and green arrows).




marker CD9 was predominantly present in the middle part ofthe gradient (Fig. 7C). Interestingly, only in vesicles derivedfrom patient 4 the L1-32 cleavage form was present. All otherascites were devoid of L1-32 in this part of the gradient.Therefore, the ascites clearly contained two types of vesicles,one containing L1-32 being related to exosomes and theother containing L1-28 being related to apoptotic membraneblebs.

Soluble L1 from ascites binds to cells and triggers cell migrationand extracellular signal-regulated kinase phosphorylation. Weinvestigated the putative biological activity of soluble L1 inascites fluids. The affinity-purified soluble L1 was found to beintact and undegraded with a size of 150 kDa (L1-150) and200 kDa (L1-200; Fig. 8A). As shown in Fig. 8B , ascites L1 wasa potent dose-dependent inducer of cell migration. Theenhanced cell migration was blocked in the presence of mAbsto the a5h1 and avh5 integrins, but only little effect was seen

with the mAb to avh3 (Fig. 8C). To obtain evidence for thedirect binding of soluble L1 to cells, HEK293 cells wereincubated with purified L1 (10 Ag/mL). As shown in Fig. 8D ,cell surface–bound L1 was readily detected. The binding tocells was blocked in part by preincubation of cells with mAbsto the integrins a5h1, avh5, and avh3, respectively, but not toa6 integrin (Fig. 8D). In agreement with previously publisheddata (37, 45), these results suggested that L1 can bind to severalintegrins at the cell surface.


Fig. 3. Cleavage of L1in released membrane vesicles. Cells were treated with theindicated apoptosis inducers in the absence or presence ofTAPI-1at10 Amol/L for18 hours and membrane vesicles were isolated from the supernatants.Vesicles weresuspended in SDS sample buffer and analyzed byWestern blotting with pcytL1followed by peroxidase-conjugated secondary antibody and ECL detection. A,analysis of CHO-hL1cells. B, analysis of OVMz cells.

Fig. 4. Inhibitors of metalloproteinases affect rate of apoptosis and L1sheddingunder hypoxia. CHO-hL1 (A) and OVMz cells (B) were incubated for 60 hoursunder normoxic and hypoxic conditions in the presence or absence of10 Amol/LTAPI-0.The rate of apoptosis was measured by Nicoletti staining. Note that theTAPI-0 solvent DMSO did not affect the rate of apoptosis. C and D, soluble L1wasanalyzed usingWestern blot analysis with mAb L1-11A to the ectodomain of L1followed by peroxidase-conjugated secondary antibody and ECL detection.




Binding to av integrins can induce activation of the mitogen-activated protein kinase pathway (46). Indeed, when added toHEK293 cells, the soluble L1 led to a transient stimulation ofERK phosphorylation (Fig. 8E).

Discussion

Serum and ascites of ovarian carcinoma patients accumulatesignificant amounts of soluble L1 released from tumor cells. Ina previous work using L1-transfected CHO and HEK293 cells,we observed that soluble L1 was released either directly fromthe cell surface or in secretory membrane vesicles in whichsubsequent cleavage occurred. In the present report we provideevidence that the latter mechanism of L1 release seems to beoperative in ovarian carcinoma cells both in vitro and in vivo .We show that the release of L1 occurs in two types of vesicles:exosomes, which are released constitutively, and apoptoticmembrane vesicles, which are increased in the presence ofapoptotic stimuli.

The release of vesicles from living cells is considered to be anormal physiologic process (47, 48). It is particularly active inproliferating cells, such as cancer cells, where it can occurcontinuously. Here we show that cancer cells undergoingapoptosis augment the release of L1. In CHO-hL1 cells thereleased L1 was heterogeneous in size but was more homoge-neous in OVMz cells. The degree of induction of L1 release inresponse to apoptotic stimuli was higher in CHO-hL1 whencompared with ovarian carcinoma cells. This could be related todifferences in the mode of L1 release. We observed that inCHO-hL1 cells the cleavage of L1 proceeds both at the cellsurface and in released vesicles whereas in OVMz cells weobserved mainly cleavage in vesicles.5 The vesicles released

from OVMz cells could be divided into two major fractions: (a)spontaneously released vesicles that contained ADAM10, thecleavage fragment L1-32, and the tetraspan CD9. These vesiclesare most likely related to exosomes; (b) a second fraction thatwas increased following apoptotic stimuli and that containedL1-32 and ADAM10 but was nearly negative for CD9. This poolcontains most likely apoptotic membrane blebs. The formationof such apoptotic blebs containing ADAM10 and L1 could beshown by fluorescence microscopy. The observation that L1and ADAM10 were seen in exosomes deserves a further notion.We previously reported that predominantly the neural form ofL1 containing the RSLE form is recruited in spontaneouslyreleased vesicles (24). The RSLE motif facilitates interactionwith the A2 chain of the clathrin adaptor AP-2 and affects thephosphorylation and endocytosis of L1 by creating anendocytic motif allowing clathrin-mediated internalization(49, 50). Indeed, we found by reverse transcription-PCR thatOVMz cells expressed the neural form of L1.6

Using antisense oligonucleotides specific for ADAM10 andADAM17, we observed that under constitutive conditionsADAM10 was the dominant proteinase involved in L1 releasewhereas ADAM17 was not involved. This is consistent with ourprevious results showing that a dominant-negative ADAM10construct could block the constitutive cleavage of L1 in severalcell lines such as AR carcinoma cells, CHO cells, and HEK293cells (10, 24). Under apoptotic conditions, the effect ofADAM10-specific antisense oligonucleotides was abrogated,indicating that besides ADAM10, other metalloproteinases cancleave L1. This is in agreement with a recent study showing thatdying cells undergo a stress response leading to an up-regulatedcleavage of the EGF receptor ligand HB-EGF involving multipleADAMs (51).

L1-containing vesicles were also observed in vivo in theascites of ovarian carcinoma patients. Although the numberof patients studied is still small, several observations could bemade: (a) in all patients investigated soluble L1 was


5 S. Riedle, unpublished observations.6 A. Stoeck, unpublished observations.

Fig. 5. Role of ADAM10 in the constitutiveand apoptosis-induced L1shedding. A,down-regulation of ADAM10 and ADAM17after transfectionwith specific antisenseoligonucleotides (ASO) using antibodiestoADAM10 and ADAM17 followed bysecondary antibody and ECL detection.Actin was used as loading control. Band C, OVMz cells were transfected withantisense oligonucleotides specific forADAM10, ADAM17, or control antisenseoligonucleotide and the amount ofsoluble L1in the presence or absence ofC2-ceramide as apoptosis inducer wasanalyzed. Conditionedmediawere analyzedbyWestern blot with mAb L1-11A to theectodomain of L1followed byperoxidase-conjugated secondaryantibody and ECL detection.




accompanied by L1-containing vesicles; (b) ascites vesicleshad similar density features as in vitro derived vesicles andcould again be divided into a CD9-positive exosomal fractionand into a CD9 low containing fraction with a densitysimilar to apoptotic blebs; (c) only in one ascites the L1-32was detected in the CD9-positive exosomal fractions whereasthe L1-28 fragment was present in the membrane bleb


Fig. 6. Analysis of apoptotic vesicles by sucrose gradient fractionation. A,schematic representation of vesicle fractionation using sucrose density gradientcentrifugation. B, OVMz cells in serum-free mediumwere cultivated for 48 hours inthe absence or presence of C2-ceramide to induce apoptosis.Vesicles were isolatedfrom the supernatant and applied onto a sucrose density gradient. Proteins fromeach fraction were separated by SDS-PAGE under reducing conditions (L1andADAM10) or nonreducing conditions (CD9) followed byWestern blotting with theindicated antibodies and ECL detection. Fractions of the gradient referred to asexosomes andmembrane blebs are indicated. Fig. 7. Analysis of ascites vesicles from ovarian carcinoma patients. A,

vesicles from the ascites of ovarian carcinoma patients were dissolved in SDSsample buffer and probed with biotinylated mAb L1-11A followed byperoxidase-conjugated streptavidin and ECL detection. Patient 1, endometroidcystadenocarcinoma; patients 2 to 5, serous or papillary-serouscystadenocarcinomas. B, soluble L1 in ascites fluids was measured usinga sandwich ELISA specific for soluble L1. C, pelleted vesicles weresuspended in sucrose solution and applied onto a sucrose density gradient.Proteins from each fraction were separated by SDS-PAGE followed by Westernblotting with the indicated antibodies and ECL detection. *, indicated proteinbands were also reactive with the secondary antibody and are thereforenonspecific.




fraction. We suspect that the L1-28 fragment is a degradationproduct derived from L1-32 that could well be generatedduring vesicle isolation. A similar form was also seen inOVMz vesicles, ruling out the possibility that this fragmentoccurred only in vivo. (d) L1-containing vesicles and solubleL1 in the ascites fluid were variable in amount and

composition between individual patients. This could berelated to the individual status of the tumor and thetreatment history of the patients. Apoptotic conditions inascites could be caused by therapy or hypoxia leading toapoptotic vesicles. We conclude that the presence of vesiclesin the ascites containing L1-220 and the cleavage products


Fig. 8. Soluble L1from the ascites fluidtriggers cell migration. A, purification ofsoluble L1from the ascites fluids ofovarian carcinoma patients by affinitychromatography using L1-11A sepharose.An aliquot of purified L1was separated bySDS-PAGE.Themembrane was stainedwith amido black to visualize protein and L1was detected by mAb L1-11A followed byperoxidase-conjugated secondary antibodyand ECL detection. B, soluble L1at theindicated concentrations was added toHEK293 cells and haptotactic migration onfibronectin was examined.The experimentwas done thrice in quadruplicates withsimilar results; a representative studyis shown. *, P = 0.0038. C, HEK293cells stimulated with soluble L1werepreincubated with the indicated mAbs at10 Ag/mL and haptotactic migration onfibronectin was examined.The mAbswere present during the assay time.The experiment was done thrice inquadruplicates with similar results; arepresentative study is shown. D, binding ofsoluble L1to HEK293 cells in the absenceor presence of the indicated mAbs tointegrins (10 Ag/mL). Bound L1wasdetected with biotinylated mAb L1-11Afollowed by streptavidin-conjugatedphycoerythrin. E, HEK293 cells werestarved for 30 minutes under serum-freeconditions and then stimulated with solubleL1 (10 Ag/mL) for the indicated length oftimewith or without the addition of 5% fetalbovine serum before harvest to assess ERKactivation.The experiments were donethrice with similar results; a representativestudy is shown.




L1-32 and L1-28 strongly suggest that cleavage in releasedvesicles can proceed also in vivo . It is quite possible thatsoluble L1 in the serum and ascites fluid of carcinomapatients at least in part is generated by ADAM10. This issupported by the observation that L1-32, the product ofADAM10 cleavage, is present in ovarian carcinoma celllysates and that ADAM10 is expressed in nearly all advancedtumors where soluble L1 was detected (9).

It is well established that tumor cells use proteolytic activityto modify their environment, which often includes tumor cellsurface molecules (51). An intriguing question is whether thegeneration of soluble L1 is a mere byproduct or of advantagefor the tumor. In experimental studies an important link wasfound between cell migration and L1 shedding: the enhancedmigration of L1-expressing cells was dependent on metal-loproteinase activity and the released L1 ectodomain was ableto stimulate migration through autocrine/paracrine binding toavh5 integrins (10). In the present study we show that releasedL1 remains intact and can be purified from ascites fluid.Purified L1 bound to cells via integrins, triggered cell motilityon extracellular matrix proteins, and induced the phosphory-lation of ERK. It is well possible that the migration of ovariancarcinomas towards components of the extracellular matrix isdriven by L1 and other shed molecules. Not only the expressionof L1 but also the proteolytic machinery for ectodomainshedding is therefore of paramount importance for tumor cells.Silletti et al. (52) reported recently that in L1-positive comparedwith L1-negative cells a sustained ERK phosphorylation wasobserved. The expression of L1 up-regulated mRNAs for themotility-related proteins rac and rho and a variety of othergene products including the h3-integrin chain and the proteases

B- and L-cathepsin (52). Thus, beyond the augmentation of cellmotility, L1 expression may have other important effects suchas the induction of a more invasive phenotype that couldaccount for its association with bad prognosis. Whether or notsoluble L1 plays a role in gene regulation remains to beinvestigated.

Recent analysis has shown that lysophosphatidic acid ispresent in high concentration in ascites and plasma of ovariancancer patients, which stimulates cell migration via a Ras-mitogen-activated protein/ERK kinase pathway (53), inducesthe ectodomain shedding of HB-EGF (54), and causestranscriptional activity of the IL-8 gene (55). This work isrelated to our study as similar to L1, HB-EGF is cleaved byADAM10 (30, 56). Lysophosphatidic acid could therefore wellact as an upstream inducer of L1 shedding and allow the releaseof both growth factor receptor ligands (HB-EGF) as well asintegrin ligands (L1) that might cooperate to ensure tumor cellgrowth and motility.

In summary, our results underline the important role ofsecretory vesicles in the constitutive release of L1 fromcarcinoma cells both in vitro and in vivo . The data suggest thatADAM10 has a dominant role in this process. We propose thatthe release via vesicles and the subsequent cleavage could be animportant pathway of L1 release and that of other cell surfacemolecules. Soluble L1 could regulate tumor cell function in anautocrine/paracrine fashion.

Acknowledgments

We thank Caroline Marth and Alexander Strecker for excellent technical assis-tance and Dr. Eric Rubinstein (Villejuif Cedex, France) for the gift of CD9 antibody.


References1. Moos M, Tacke R, Scherer H, Teplow D, Fruh K,Schachner M. Neural adhesionmolecule L1as amem-ber of the immunoglobulin superfamily with bindingdomains similar to fibronectin. Nature 1988;334:701^3.2. Brummendorf T, Kenwrick S, Rathjen FG. Neural cellrecognition molecule L1: from cell biology to humanhereditary brain malformations. Curr Opin Neurobiol1998;8:87^97.3. SchachnerM. Neural recognitionmolecules and syn-aptic plasticity. Curr Opin Cell Biol1997;9:627^34.4. KatayamaM, Iwamatsu A, Masutani H, et al. Expres-sion of neural cell adhesionmolecule L1in human lungcancer cell lines. Cell Struct Funct1997;22:511^6.5. IzumiY, HirataM,HasuwaH, et al. Ametalloprotease-disintegrin, MDC9/meltrin-g/ADAM9 and PKCy areinvolved in TPA-induced ectodomain shedding ofmembrane-anchored heparin-binding EGF-likegrowth factor. EMBOJ1998;17:7260^72.6. Thies A, Schachner M, Moll I, et al. Overexpressionof the cell adhesion molecule L1 is associated withmetastasis in cutaneous malignant melanoma. Eur JCancer 2002;38:1708^16.7. Fogel M, Gutwein P, Mechtersheimer S, et al. L1 ex-pression as a predictor of progression and survival inpatients with uterine and ovarian carcinomas. Lancet2003;362:869^75.8. Heiz M, Grunberg J, Schubiger PA, Novak HI. Hepa-tocyte growth factor induced ectodomain shedding ofcell adhesion molecule L1: Role of the L1cytoplasmicdomain. JBiol Chem 2004;279:31149^56.9. FogelM,Mechtersheimer S,HuszarM, et al. L1adhe-sionmolecule(CD171)indevelopmentandprogressionof human malignant melanoma. Cancer Lett 2003;189:237^47.

10. Mechtersheimer S, Gutwein P, Agmon LN, et al.Ectodomain shedding of L1adhesion molecule pro-motes cell migration by autocrine binding to integrins.JCell Biol 2001;155:661^74.11. Gutwein P, Oleszewski M, Mechtersheimer S,Agmon LN, Krauss K, Altevogt P. Role of Src kinasesin the ADAM-mediated release of L1adhesion mole-cule from human tumor cells. J Biol Chem 2000;275:15490^7.12. Arribas J, Borroto A. Protein ectodomain shedding.Chem Rev 2002;102:4627^38.13. Hirata M, UmataT,Takahashi T, et al. Identificationof serum factor inducing ectodomain shedding ofproHB-EGF and studies of noncleavable mutants ofproHB-EGF. Biochem Biophys Res Commun 2001;283:915^22.14. Iwamoto R, Mekada E. Heparin-binding EGF-likegrowth factor: a juxtacrine growth factor. CytokineGrowth Factor Rev 2000;11:335^44.15. Fan H, Derynck R. Ectodomain shedding of TGF-aand other transmembrane proteins is induced by re-ceptor tyrosine kinase activation and MAP kinasesignaling cascades. EMBO J1999;18:6962^72.16. Codony SJ, Albanell J, Lopez TJ, Arribas J,Baselga J. Cleavage of the HER2 ectodomain is apervanadate-activable process that is inhibited bythe tissue inhibitor of metalloproteases-1 in breastcancer cells. Cancer Res 1999;59:1196^201.17. Feehan C, Darlak K, Kahn J, Walcheck B, SpatolaAF, Kishimoto TK. Shedding of the lymphocyte L-selectin adhesion molecule is inhibited by a hydroxa-mic acid-based protease inhibitor. Identification withan L-selectin-alkaline phosphatase reporter. J BiolChem 1996;271:7019^24.18. Fors BP, Goodarzi K, von AU. L-selectin shedding is

independent of its subsurface structures and topo-graphic distribution. J Immunol 2001;167:3642^51.19. Benjannet S, Elagoz A,WickhamL, et al. Post-trans-lational processing of h-secretase (h-amyloid-con-verting enzyme) and its ectodomain shedding. Thepro-and transmembrane/cytosolic domains affect itscellular activity andamyloid-hproduction. JBiolChem2001;276:10879^87.20. Li G, Percontino L, Sun Q, Qazi AS, Frederikse PH.h-amyloid secretases and h-amyloid degrading en-zyme expression in lens. MolVis 2003;9:179^83.21. Endres K, Anders A, Kojro E, Gilbert S, Fahrenholz F,Postina R.Tumor necrosis factor-a converting enzymeis processed by proprotein-convertases to its matureform which is degraded upon phorbol ester stimula-tion. EurJBiochem 2003;270:2386^93.22. Fitzgerald ML, Wang Z, Park PW, Murphy G,Bernfield M. Shedding of syndecan-1 and -4 ecto-domains is regulated by multiple signaling pathwaysand mediated by a TIMP-3-sensitive metalloprotei-nase. J Cell Biol 2000;148:811^24.23. Gasbarri A, Del PF, Girnita L, Martegani MP, NataliPG, Bartolazzi A. CD44s adhesive function spontane-ous and PMA-inducible CD44 cleavage are regulatedat post-translational level in cells of melanocytic line-age. Melanoma Res 2003;13:325^37.24. Gutwein P, Mechtersheimer S, Riedle S, et al. AD-AM10-mediated cleavage of L1adhesion molecule atthe cell surface and in released membrane vesicles.FASEB J 2003;17:292^4.25. vonTresckow B, Kallen KJ, von SE, et al. De-pletion of cellular cholesterol and lipid raftsincreases shedding of CD30. J Immunol 2004;172:4324^31.26. MatthewsV, Schuster B, Schutze S, et al. Cellular





cholesterol depletion triggers shedding of the humaninterleukin-6 receptor by ADAM10 and ADAM17(TACE). JBiol Chem 2003;278:38829^39.27. Baselga J, Mendelsohn J, Kim YM, Pandiella A.Autocrine regulation of membrane transforminggrowth factor-a cleavage. J Biol Chem 1996;271:3279^84.28. Lee RK,Wurtman RJ. Regulation of APP synthesisand secretion by neuroimmunophilin ligands andcyclooxygenase inhibitors. Ann N YAcad Sci 2000;920:261^8.29. Yabkowitz R, Meyer S, Black T, Elliott G,Merewether LA,Yamane HK. Inflammatory cytokinesand vascular endothelial growth factor stimulate therelease of soluble tie receptor from human endothelialcells via metalloprotease activation. Blood 1999;93:1969^79.30. Prenzel N, Zwick E, Daub H, et al. EGF receptortransactivation by G-protein-coupled receptorsrequires metalloproteinase cleavage of proHB-EGF.Nature1999;402:884^8.31.Taylor DD, Black PH. Shedding of plasmamembranefragments.Neoplastic and developmental importance.Dev Biol (NY1985) 1986;3:33^57.32. StoorvogelW, Kleijmeer MJ, Geuze HJ, Raposo G.The biogenesis and functions of exosomes. Traffic2002;3:321^30.33.Thery C, Zitvogel L, Amigorena S. Exosomes: com-position, biogenesis and function. Nat Rev Immunol2002;2:569^79.34.Andre F, SchartzNE,MovassaghM, et al.Malignanteffusions and immunogenic tumour-derived exo-somes. Lancet 2002;360:295^305.35. Condon TP, Flournoy S, Sawyer GJ, Baker BF,KishimotoTK, Bennett CF. ADAM17 but not ADAM10mediates tumor necrosis factor-a and L-selectin shed-ding from leukocyte membranes. Antisense NucleicAcid Drug Dev 2001;11:107^16.36. Nicoletti I, Migliorati G, Pagliacci MC, Grignani F,Riccardi C. A rapid and simple method for measuringthymocyte apoptosis by propidium iodide staining

and flow cytometry. J Immunol Methods 1991;139:271^9.37. RuppertM, Aigner S, HubbeM,YagitaH, Altevogt P.The L1adhesionmolecule is a cellular ligand forVLA-5.JCell Biol1995;131:1881^91.38. Beer S, Oleszewski M, Gutwein P, Geiger C,Altevogt P. Metalloproteinase-mediated release ofthe ectodomain of L1 adhesion molecule. J Cell Sci1999;112:2667^75.39. Ilan N,Mohsenin A, Cheung L,MadriJA. PECAM-1shedding during apoptosis generates amembrane-an-chored truncatedmoleculewithunique signaling char-acteristics. FASEB J 2001;15:362^72.40. Kern PM, Keilholz L, KaldenJR, Herrmann M. Apo-ptotic UV-irradiated lymphocytes undergo proteasemediated shedding of L-selectin in vitro. TransfusApheresis Sci 2001;24:99^101.41. Steinhusen U, Weiske J, Badock V, Tauber R,Bommert K, Huber O. Cleavage and shedding of E-cadherin after induction of apoptosis. J Biol Chem2001;276:4972^80.42. Barnhill RL, Piepkorn MW, Cochran AJ, Flynn E,Karaoli T, Folkman J. Tumor vascularity, proliferation,and apoptosis in human melanoma micrometastasesand macrometastases. Arch Dermatol 1998;134:991^4.43. Holmgren L, O’Reilly MS, Folkman J. Dormancy ofmicrometastases: balanced proliferation and apopto-sis in the presence of angiogenesis suppression. NatMed1995;1:149^53.44. Saikumar P, Dong Z,Weinberg JM,VenkatachalamMA. Mechanisms of cell death in hypoxia/reoxygena-tion injury. Oncogene1998;17:3341^9.45. Felding HB, Silletti S, Mei F, et al. A single immuno-globulin-like domainof thehumanneural cell adhesionmolecule L1supports adhesion by multiple vascularand platelet integrins. JCell Biol1997;139:1567^81.46. Hood JD, Frausto R, KiossesWB, Schwartz MA,Cheresh DA. Differential av integrin-mediated Ras-ERK signaling during two pathways of angiogenesis.JCell Biol 2003;162:933^43.

47.Ginestra A, La PM,Saladino F, CassaraD,NagaseH,Vittorelli ML. The amount and proteolytic content ofvesicles shed by human cancer cell lines correlateswith their in vitro invasiveness. Anticancer Res 1998;18:3433^7.48. DoloV, Ginestra A, Cassara D, et al. Selective local-ization of matrix metalloproteinase 9, h1integrins, andhuman lymphocyte antigen class Imolecules onmem-brane vesicles shed by 8701-BC breast carcinomacells. Cancer Res1998;58:4468^74.49. Kamiguchi H, LemmonV. A neuronal formof the celladhesionmolecule L1contains a tyrosine-based signalrequired for sorting to the axonal growth cone. JNeu-rosci1998;18:3749^56.50. Long KE, Asou H, Snider MD, LemmonV. The roleof endocytosis in regulating L1-mediated adhesion.JBiol Chem 2001;276:1285^90.51. Egeblad M,Werb Z. New functions for the matrixmetalloproteinases in cancer progression. Nat RevCancer 2002;2:161^74.52. Silletti S,Yebra M, Perez B, Cirulli V, McMahon M,Montgomery AM. ERK-regulated gene expressioncontributes to L1-CAM dependent motility and inva-sion. JBiol Chem 2004;279:28880^8.53. Bian D, Su S, Mahanivong C, et al. Lysophosphati-dic acid stimulates ovarian cancer cell migration via aRas-MEK kinase 1 pathway. Cancer Res 2004;64:4209^17.54. Fang X, Yu S, Bast RC, et al. Mechanisms forlysophosphatidic acid-induced cytokine productionin ovarian cancer cells. J Biol Chem 2004;279:9653^61.55. Miyamoto S, Hirata M,Yamazaki A, et al. Heparin-binding EGF-like growth factor is a promising targetfor ovarian cancer therapy. Cancer Res 2004;64:5720^7.56. Fischer OM, Hart S, Gschwind A, Prenzel N, UllrichA. Oxidative and osmotic stress signaling in tumorcells is mediated by ADAM proteases and heparin-binding epidermal growth factor. Mol Cell Biol 2004;24:5172^83.




2005;11:2492-2501. Clin Cancer Res Paul Gutwein, Alexander Stoeck, Svenja Riedle, et al. Vesicles Released from Ovarian Carcinoma CellsCleavage of L1 in Exosomes and Apoptotic Membrane

Updated version

http://clincancerres.aacrjournals.org/content/11/7/2492

Access the most recent version of this article at:

Cited articles

http://clincancerres.aacrjournals.org/content/11/7/2492.full#ref-list-1

This article cites 54 articles, 27 of which you can access for free at:

Citing articles

http://clincancerres.aacrjournals.org/content/11/7/2492.full#related-urls

This article has been cited by 12 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. (CCC)Click on "Request Permissions" which will take you to the Copyright Clearance Center's

.http://clincancerres.aacrjournals.org/content/11/7/2492To request permission to re-use all or part of this article, use this link



http://clincancerres.aacrjournals.org/content/11/7/2492.full#ref-list-1

http://clincancerres.aacrjournals.org/content/11/7/2492.full#related-urls

http://clincancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]



Documents

CleavageofL1inExosomesandApoptoticMembraneVesicles ... · endometrial carcinomas in a stage-dependent manner and that L1 expression was a predictor of poor outcome (9). It was proposed