~ LINZ 2006-~-------------

Neuronal Cell Culture from HumanEmb-:yonic Stem Cells as in vitro Modelfor NeuroprotectionAndre Schrattenholz and Martina KlemmProteoSys AG, Mainz, Germany

SummaryIn the context of efficacy testing ofpharmacological compoundsin animal models, replacement of some of these models with arelevant human in vitro system appears attractive, in particularwith regard to large scale screening. Here, we show resultsfrominitial phases of a project, which attempts to explore theoutstanding potential of human embryonie stem cell (hESC)-based in vitro models with special regard to neuronal stress asa potential replacement of animal models for human neuro-degenerative diseases.We show the functionality of neurons derived from hESCprecursors by calcium imaging, mitochondrial potentialmeasurements and Western blots and moreover demonstratethat this model reproduces crucial mechanistic aspectsobserved during ischemia and excitotoxicity that are thought tobe at the core of some neurodegenerative diseases. Also,the broader possibilities for refining surrogate molecularinformation emerging from the detailed analysis of this modelare discussed.

Zusammenfassung: Stammzellbasierte menschliche in vitroModelle zum Ersatz der Wirksamkeitsprüfung bei TierenIm Kontext von Wirksamkeitsstudien von pharmakologischenFormulierungen und Medikamentenkandidaten in Tiermodellenbesteht das starke Bedürfnis Ersatzmodelle auf der Basismenschlicher in vitro Systeme zu entwickeln. Wir zeigen hierErgebnisse aus den frühen Phasen eines Projekts, welches dasbesondere Potential humaner embryonaler Stammzell- (hESC)-Modelle als entsprechende in vitro Systeme erforschen möchte,mit einem speziellem Fokus auf neuronalen Stress und mit demZiel Tiermodelle für humane neurodegenerative Krankheiten zuersetzen.Wir konnten die Funktionalität von Neuronen, die aus hESCVorläufern differenziert wurden, durch Calcium Imaging,Potentialmessungen der mitochondrialen Membran undWestern Blots zeigen. Wir konnten auch nachweisen, dass dasModell wesentliche mechanistische Aspekte, wie sie währendder Ischämie oder unter erregungstoxischen Bedingungen auf-treten, gut widerspiegelt. Es wird vermutet, dass diese Prozesseeiner Reihe von menschlichen neurodegenerativen Krankheitenzugrunde liegen. Auch weitergehende Möglichkeiten zur Ent-wicklung besserer und genauerer; auch molekularer Endpunkte,die sich aus der detaillierten Analyse dieser Modelle ergebenkönnten, werden diskutiert.

Keywords: human embryonie stem cells, excitotoxicity, in vitro models, neurodegeneration

1 Introduction

Examples for efficacy models includexenograft models for various human can-cers; models for stroke, in which severecerebral ischemia is induced by perma-nent occlusion of cerebral arteries in testanimals (MCAO); various transgenic ro-dent models for neurodegenerative dis-eases, or extremely tough experimentalmodels for autoimmune encephalomyeli-tis induced by vaccination of experimen-tal animals with autoantigens as preclini-

cal models of multiple sclerosis (MOG-EAE) (Bähr, 2004; Schrattenholz andKlemm, 2006). Despite a high degree ofsophistication, there is a general sense ofcaution towards these models, becausetheir read-outs are often insufficient ormisleading.Generally, in the area of efficacy test-

ing, there is an urgent need for the devel-opment of novel in vitro methods, whichcan balance disadvantages in terms of or-gan specific barriers and metabolism withadvantages regarding the higher relevance

Received 11 September 2006; received in final form and accepted for publication 15 December 2006

ALTEX 24, 1/07

of humanised systems, more preeise func-tional and molecular read-outs and the po-tential of higher throughput. In the givensituation of regulatory frameworks in theEC and elsewhere, the purely diagnosticuse of human embryonie stern cell modelspromises to potentially provide highly at-tractive alternatives (Schrattenholz,Klemm and Cahill, 2004). The cells canbe differentiated to organotypic cell cul-tures and serve as a flexible substrate for avariety of functional molecular endpoints.Moreover, they are ideal for validationpurposes using modern silencing tech-nologies in a framework of genetically ho-mogeneous differentiations to various

9

~WM ~----------------------------------------~---

"tissue" -like cell culture substrates likeneural cells, cardiomyocytes, varioustypes of muscle cells and adipocytes(Schrattenholz and Klemm, 2006).

Related results from correspondingmurine ESC-screening systems, in com-bination with quantitative differentialproteomic display techniques, have al-ready provided biomarkers for the poten-tial replacement of some related animalmodels (Buesen et a1., 2004; Genschowet a1., 2004; Schillo et al., 2005; Schrat-tenholz et a1., 2005; Sommer et a1.,2004). Here we show the first results ofthe adaptation of mESC models to corre-sponding hESC systems.

2 Materials and methods

Culture and differentiation of hES cellsUndifferentiated hES cells (human em-bryonic stem cell line (hES) H9 fromWiCell Research Institute, NIH CodeWA09) were cultured on a feeder layerof irradiated CFl mouse embryonie fi-broblasts (MEF) in hES-medium[DMEMfFI2 supplemented with 20%(v/v) Knockoutr» Serum Replacement(defined, serum-free formulation), 0.1mM ß-mercaptoethanol, 200 mM L-glu-tamine, 0.1 mM non-essential aminoacids, penicillin (100 IU/ml), strepto-mycin (100 ug/ml) from Gibco and 4ng/ml basic fibroblast growth factor(bFGF, Invitrogen)] with a daily changeof medium and passage once a week at aratio betweenl:3 and 1:6.

For differentiation procedures seeFigure 1. As differentiation mediumDMEMlF12 was supplemented with N2(Gibco), penicillin (100 IU/ml, Gibco),streptomycin (100 ug/ml, Gibco) and hep-arin (2 ug/ml) in the presence of bFGF (20ng/ml) or BDNF (10 ng/ml) and finally re-placed by neurobasal medium.

Routinely APase staining was per-formed to characterise differentiation(Fig. 2A) by fixing hES cells in 4%paraformaldehyde at 4°C for 20 min,washing several times in Tris-maleatebuffer and staining with naphthol AS MXphosphate/Fast Red TR salt mix.

Western Blot AnalysisThe protein concentration was deter-mined by BCA assay according to Smith

10

Daily change ofhESmedium

DayO

Cultivation of hEScells onirradiated CF1mouse

fibroblasts

1 Dispase detachment ofhESceli colonies

4 DaysDerivation of embryoldbodies In hES-medium

without bFGF

1 Seeding of embryoidbodies in

differentiation medium

5to 10Days

Embryoid body attachmentand neural precursor

outgrowth

1Dispasedetachment of embyroid

body derived neural precursor cells

7 Days Proliferation of neuralprecursor cells in the presence

ofBDNF

Neural precursorseeding in culture

dlshes

1- 2 weeks

Neuronaldifferentiation in thepresence of BDNF

Fig. 1: Differentiation of hEse from totipotent precursor cells to mixed neural cellcultures including functional neurons; bFGF, basic fibroblast growth factor, BDNF,brain derived neurotrophic factor.

(Smith et a1., 1985) with adaptation formicroplates. Sampies were mixed withsodium dodecyl sulphate polyacrylamidegel electrophoresis- (SDS-PAGE) or XT-sample buffer and heated (95°C) for 5min. Individual samples were run on12% Criterion-XT Bis-Tris gels (Biorad,Cat # 345-0119) and transferred toPVDF membranes (Biorad). After block-ing with 30 mM Tris-HCI (pH 7.5), 100mM NaCl and 0.1% Tween 20 (TBS-T)containing 5% fat-free milk powder for 1 h

at room temperature, the membraneswere incubated with the primary anti-body: anti synaptophysin (SYP, D-4,Santa Cruz Biotechnology, Santa Cruz,CA, USA) overnight at 4°C, followed bya peroxidase-labelled goat anti-mousesecondary antibody for 1 h at room tem-perature. Visualisation was performedusing the ECL method and a CCD cam-era (Diana system, raytest, Strauben-hardt). Immunoreactive bands werequantified using the AIDA Software

ALTEX 24, 1107

~ ~~---~----------------------------------------

C

100908070605040302010o

Control 200 IJMNMDA

Fig. 2: A, human embryonie stem cells (hESC) stained by alkaline phosphatase. B,embryoid bodies differentiated to cells with neuronal morphology (V41 differentiatedcells) in the presence of BDNF. These neural cells had morphologieal, functional andmolecular characteristics typical of neurons. C, cells were tested for neuronalsurvival of a standard excitotoxic insult by application of 200 IJM NMDA. They thusprovide a model for testing efficacies of neuroprotective substances.

package (raytest) as described elsewhere(Schillo et al., 2005).

Assessment 01 in vitro functionality01hES-derived neurons by calcium-imagingThe calcium imaging procedure forquantification of functional hES-derivedneuronal responses was described previ-ously (Sommer et al., 2004). Generally,the neurons were kept in washing buffer(130 ruM NaCI, 5 ruM KCI, 2 ruM Ca-

ALTEX 24, 1/07

C12,0.5 ruM MgCI2, 10 ruM HEPES, 10ruM Glucose, pH 7.3) and stimulated byincubation for 30 s in stimulation buffer(130 ruM NaCI, 5 ruM KCI, 2 ruM Ca-CI2>10 ruM HEPES, 10 ruM glucose and200 f.lM NMDA or 10 f!M glutamate).

lmaging 01mitochondrial membranepotential (~'tjJ) changesFluorescent probe rhodamine-l23 wasused to monitor qualitative changes ofmitochondrial membrane potential (~'tjJ).

Slices were loaded with rhodamine-123were excited at 480 nm and emission wasmonitored at 565 nm. Mitochondrial rho-damine-123 accumulation and quenchingresults in a fiuorescence decrease, where-as depolarisation provokes signal de-quenching and an increase of fiuores-cence after mitochondrial dye release(Metzger et al., 2005).

3 Results

As shown in Figure 2, human embryon-ic stern cells (hESC, APase stainingFig. 2A, H9 from NIH registry, NIHCode WA09) were differentiated viaembryoid bodies to cells with neuronalmorphology (V41: neuronal differenti-ated hES cells, Fig 2B), which weretested for neuronal survival upon astandard excitotoxic insult, i.e. applica-tion of 200 f.lM NMDA (Fig 2C). Theseconditions were similar to experimentalprotocols used for description of surro-gate protein biomarkers in mESC mod-els and produced a comparable effecton the viability and functional respons-es of neurons (Sommer et al. , 2004),thus providing the prerequisite for in-vestigating neuroprotective compoundsand conditions.On the molecular level, the matura-

tion of neuronal derivatives from hESCprecursors was confirmed by stainingcorresponding cell extracts with an anti-body against the human form of thepresynaptic marker pro tein synapto-physin (SYP, D-4, a synaptic vesiclemarker). As shown in Figure 3, Westernblots frorn I-dimensional poly acryl-amide gels of proteins from undifferen-tiated precursor cells (hESC prec) andneuronal differentiated derivatives (V41diff), stained with a specific antibodyagainst human synaptophysin, showedthe strong emergence of this protein atthe expected molecular mass of 38 kDa(lanes 9 and 10 in Fig. 3).Calcium-driven excitotoxicity and

mitochondrial apoptosis, which followsthe opening of the mitochondrial perme-ability transition pore, contribute to es-sential pathomechanisms underlyingmost human neuro degenerative diseases(Schrattenholz and Soskic, 2006).Therefore, during development of new

11

LINZ 2006 m.....-------------~-

lead structures for treatment of e.g.Alzheimer's disease or stroke, the test-ing of their efficacy with regard to thesemechanisms can play a pivotal role.Moreover, as outlined in the schemeshown in Figure 4H, it is of advantage ifa variety .of converging pathomecha-nisms can be addressed with similarread-outs. Here, excitotoxic conditions(NMDA insult), which are considered toplay a role in Alzheimer's and Parkin-son's diseases, certain aspects of stroke,traumatic brain injury and amyotrophiclateral sclerosis (ALS); ischemia, whichis a key event in stroke related disorders,and massive calcium-overload of neu-rons induced by brief applications ofneurotoxic amyloidogenic peptides, likeAßI-40 (relevant for Alzheimer-relatedmechanisms) (Cappai and White, 1999;Pereira et al., 2004), were all tested andquantified on a single cell level,

As an exarnple, in Figure 4 A-D, con-secutive Fura-2 measurements of intra-cellular calcium changes were moni-tored on the single cell level of hESCneurons. All cells responded to a moder-ate stimulus (pulse application of 10 11Mglutamate, Fig. 4A) with clearly quan-tifiable calcium transients, which re-turned to steady state levels within sec-onds (not shown). Based on thisviability control, the very same cellswere subsequently challenged with anexcitotoxic concentration of the gluta-matergic agonist N-methyl-D-aspartate(200 11M NMDA for 10 min), a treat-ment which itself evokes calcium tran-sients (Fig. 4B), but with different ki-netic characteristics, leading to neuronsthat are no Ion ger functional (Fig. 4C).Neurons that were not exposed to exci-totoxic conditions or, if so, in the pres-ence of neuroprotective agents, contin-ued to res pond to the viability control ina reproducible manner far up to severalho urs (Fig. 4D).

In Figure 4E, a phase contrast imageof typical hESC neurons used for thefunctional experiments described aboveis shown (scale: E and G: 100 11M,F: 2011M). As examples of mechanism-relat-ed read-outs, in Figure 4F the imagingof the opening of the mitochondrialtransition pore (change of mitochondri-al membrane potential) was quantifiedby a fluorescent assay employing Rho-

12

Protein Synaptophysin

15eL10eL75-

SeL

31-

25-

2eL

1 2 3 4 5 6 7 8 9 10

Fig. 3: Neuronal derivatives of hEse precursors expressed protein markers formature neurons.In extracts from cultured hESC-derived neural cells (as shown in Fig. 1B), the human formof the presynaptic marker protein synaptophysin (a synaptic vesicle marker) was detectedby Western blot, thus demonstrating differentiation to a neuronal end point. Left panel,silver-stained 1-dimensional polyacrylamide gels of proteins from undifferentiatedprecursor cells (hESC prec) and neuronal differentiated derivatives (V41 diff) (Ianes 2-5);lane 1, molecular mass standards. Right panel, Western blots of these sam pies (Ianes 7-10) show a strong expression of the protein at the expected molecular mass of 38 kDa(lanes 9 and 10) exclusively in the cultures with cells of neural morphology and NMDA-responses (compare to Fig. 2B,C).

damin-123 (Metzger et al., 2005) and inFigure 4G, a typical Fura-2 calcium im-age of neurons from hESC after expo-sure to 10 11M glutamate is shown asa measure of functional viability. (Fig. 4A,C,D)

4 Discussion

Taken together, we could show that thehuman stern cell model represents cru-cial mechanistic aspects that are thoughtto playamajor role in a variety of neu-rodegenerative diseases. At the mornent,

ALTEX 24, 1/07

re, LINZ 2006--~--------------------------

A500

+50

Functional responses .00

Normal amplitudes 350350

10 26 30 40 50time[,]

B c D500 500

,30<)!~ lSO

,300!« 250

100

100INMDA,200uMI

Insult Non·functionalNeurons after insult

400 Functional responses350 Normal amplitudes

400

350

200200

m 150

100

50 . 50+---,--.---r-.---,---.1210 1220 1230 1240 USO 1260 mo 1620 1630 1610 1650 1660 1670 1680

time [s] time [s)SOD 1000

öm.[,jISOD

H

c •Calcium0• NMDA :;::; transients Calcium imagingcu •

"i:• Hypoxia cu Neurotransmitter>• ß-Amyloid Q) • Mitochondrial release

I/J0 transition pore • MorphologyQ

InsultlIntervention

Response Readout

Fig. 4: Calcium-driven excitotoxicity and mitochondrial apoptosis, which 101l0w the opening 01 the mitochondrial permeabilitytransition pore, are connected to a variety 01 human pathomechanisms and were tested in vitro using neural hESC derivatives.A, all viable cells responded to pulse applications of 10 [JM glutamate. B, subsequent excitotoxic challenge (200 [JM NMDA for 10min).C, non-functional neurons after excitotoxic challenge. D, controls that were not exposed to excitotoxic conditions or, if so, in the presenceof neuroprotective agents, still continued to respond to the viability control in a reproducible manner for up to several hours. E, phasecontrast image of typical hESC neurons (scale: E and G: 100 [JM, F: 20 [JM). F, imaging of the opening of the mitochondrial transition poreis quantified by Rhodamin-123 fluorescence. G, typical Fura-2 calcium image of neurons from hESC after exposure to 10 [JM glutamateas functional viability.H, scheme summarising the sequence of functional insults and related read-outs, which can be easily extended to corresponding assaysemploying ß-amyloid-related neurotoxic peptides or hypoxie conditions.

ALTEX 24, 1/07 13

~~ ~-------------~-

hESC and mESC models are employedin early preclinical phases of drug devel-opment and a parallel molecular explo-ration contributes significantly to under-standing the underlying mechanisms,which in the future could help to replaceanimal tests in efficacy studies (Schrat-tenholz and Klemm, 2006).Our work represents only one exam-

ple among others and a first step enroute to further and eventually human-ised in vitro models (Kim et al., 2005;Huuskonen, 2005; Spielmann, 2005).The development of the full potential ofthese nove1 models requires a gradualincorporation of novel concepts con-cerning in vitro pharmacokinetics andbioavailability into related projects(Spielmann, 2005; Whitebread et al.,2005; Li et al., 2005; Hareng et al.,2005). The questions around systemicbarriers, endocrine disruptors andmetabolism, will certainly be exclusive-ly addressed by in vivo models in theforeseeable future. But probably themost important aspect is to convincing-ly connect the molecular pathologies ofvarious neuro degenerative diseases withcorresponding molecular signatures inthese in vitro models. Here it is note-worthy that various pathological pro-ces ses converge on the level of calcium-related stress and thus provide commonprinciples of neuroprotection, whichcan be tested in vitro under conditionsof highly reduced complexity and well-defined functional read-outs, In particu-lar, calcium-related excitotoxicity and

14

the intrinsic mitochondrial apoptoticpathway, which are targeted by some ofthe most successful current treatments(Schrattenholz and Soskic, 2006), aredirectly accessible by fluorescent cell-based imaging technologies, as shownin Figure 4.In addition, emphasis is on the post-

translational protein information emerg-ing from ESC models, which in connec-tion with parallel investigations ofappropriate animal models for certainhuman diseases could help to form thebasis of a future replacement of some ofthese models. Stern cell technologies,maintenance and characterisation havereached certain international standards(Li et al., 2005; Moon et al., 2006; Pla-ia et al., 2005) suitable for validated al-ternative methods in toxicity and effica-cy testing (Hareng et al., 2005; Seiler etal., 2004; Spielmann, 2005).The purely diagnostic use of hESC

suggested here not only offers attractiveexperimental advantages and, in partic-ular, the major advantage of being basedon human cells, but also avoids ethicalimplications around the generation ofnew cell lines and therapeutic risks.Moreover hESC-based in vitro modelsare superior to any type of primary neu-ronal cell culture, because they arebased on human cells. Primary culturesfrom rodent embryos are not only proneto various types of cell culture artefactsand great variability, but also require thesacrifice of huge numbers of animals,or, more precisely, pregnant animals and

embryos. Human neuroblastoma celllines are not terribly popular, becausethey are too far off the native physiolo-gy, too variable and often too reduced intheir repertoire of cellular responses(for a comprehensive overview seeBähr, 2004).

ReferencesBuesen, R., Visan, A., Genschow, E. etal. (2004). Trends in improving theembryonie stern cell test (EST): anoverview. ALTEX 21, 15-22.

Bähr, M. (2004). Neuroprotection: Mod-els, Mechanisms and Therapies. Wein-heim, New York: Wiley-VCH. Cappai,R. and White, A. R. (1999). Amyloidbeta. Int. 1. Biochem. Cell Bio!. 31,885-889.

Genschow, E., Spielmann, H., Scholz, G.et al. (2004). Validation of the embry-onic stern cell test in the internationalECVAM validation study on three invitro embryotoxicity tests. Altern. Lab.Anim. Sep. 32, 209-244.

Hareng, L., Pellizzer, C., Bremer, S. et al.(2005). The integrated project RePro-Tect: a novel approach in reproductivetoxicity hazard assessment. Reprod.Toxico!. 20,441-452.

Huuskonen, H. (2005). New models andmolecular markers in evaluation of de-velopmental toxicity. Toxicol. Appl.Pharmacol. 207, 495-500.

Kim, J. J., Patton, W. c., Corselli, 1. et al.(2005). Mouse embryonie stern cellsfor quality control testing in assisted

ALTEX 24. 1107

~ LINZ 2006---~----------------------------------------

reproductive technology programs. 1.Reprod. Med. 50, 533-538.

Li, Y., Powell, S., Brunette, E. et al.(2005). Expansion of human embryon-ic stern cells in defined serum-freemedium devoid of animal-derivedproducts. Biatechnol. Bioeng. 91, 688-698.

Metzger, F., Klapproth, N., Kulik, A. etal. (2005). Optical assessment of mo-toneuron function in a "twenty-four-hour" acute spinal cord slice modelfrom fetal rats. 1. Neurosci. Methods141,309-320.

Moon, S. Y., Park, Y. B., Kim, D. S. et al.(2006). Generation, culture, and differ-entiation of human embryonie sterncells for therapeutic applications. Mol.Ther. 13,5-14.

Pereira, c., Ferreiro, E., Cardoso, S. M.,and de Oliveira, C. R. (2004). Cell de-generation induced by amyloid-betapeptides: implications for Alzheimer'sdisease.1. Mol. Neurosei. 23,97-104.

Plaia, T. w., Josephson, R., Liu, Y. et al.(2005). Characterization of a New NIHRegistered Variant Human EmbryonieStern CeIl Line BGOIV: A Tool forHuman Embryonie Stern Cell Re-search. Stern Cells.

Schillo, S., Pejovic, v., Hunzinger, C. etal. (2005). Integrative proteomics:functional and molecular characteriza-tion of a particular glutamate-relatedneuregulin isoform. J. Proteome. Res.4,900-908.

Schrattenholz, A. and Klemm, M.(2006). How Human Ernbryonie Stern

ALTEX 24, 1/07

Cell Research Can Impact In VitroDrug Screening Technologies of theFuture. In U. Marx and V. Sandig(eds.), Drug Testing in vitro: Break-throughs and Trends in Cell Culturetechnology (pp. 205-228). Weinheim,New York: WileyNCH.

Schrattenholz, A., Klemm, M. andCahill, M. A. (2004). Potential ofComprehensive Toxico- Proteomics:Quantitative and Differential Miningof Functional Proteomes from NativeSamples. ATLA 32 (1), 123-13I.

Schrattenholz, A., Wozny, W., Klemm,M. et al. (2005). Differential and quan-titative molecular analysis of ischerniacomplexity reduction by isotopic la-beling of pro teins using a neural em-bryonic stern cell model. 1. Neurol.Sei. 229-230,261-267.

Schrattenholz, A. and Soskic, V. (2006).NMDA receptors are not alone: dy-narnic regulation of NMDA receptorstructure and function by neuregulinsand transient cholesterol-rich mem-brane domains leads to disease-specif-ic nuances of glutamate-signalling.Curr. Top Med. Chem. 6,663-686.

Seiler, A., Visan, A., Buesen, R. et al.(2004). Improvement of an in vitrostern cell assay for developmental tox-icity: the use of molecular end points inthe embryonie stern cell test. Reprod.Toxicol. 18,231-240.

Srnith, P. K., Krohn, R. 1., Hermanson,G. T. et al. (1985). Measurement ofprotein using bicinchoninic acid. Anal.Biochem. 150, 76-85.

Sommer, S., Hunzinger, C., Schillo, S. etal. (2004). Molecular analysis of ho-mocysteic acid-induced neuronalstress. 1. Proteome. Res. 3, 572-58I.

Spielmann, H. (2005). Predicting the riskof developmental toxicity from in vitroassays. Toxicol. Appl. Pharmacol. 207,375-380.

Whitebread, S., Hamon, J., Bojanic, D.,and Urban, L. (2005). Keynote review:in vitro safety pharmacology profiling:an essential tool for successful drugdevelopment. Drug Discov. Today 10,1421-1433.

AcknowledgementsThis work was supported by a grant fromthe SET (Stiftung zur Förderung der Er-forschung von Ersatz- und Ergänzungs-methoden zur Einschränkung von Tier-versuchen), Mainz, Germany and theDoerenkamp-Zbinden Foundation, Zurich,Switzerland. We thank Katrin Kulpe forexcellent technical assistance.

Correspondence toProf. Dr. Andre SchrattenholzProteoSys AGCarl-Zeiss-Str.5155129 Mainz, GermanyTel.: 0049-6131-5019215Fax: 0049-6131-5019211e-mail:andre. schrattenholz @proteosys.com

15