J. Comp. Path. 2014, Vol. 150, 474e488 Available online at www.sciencedirect.com

ScienceDirect

www.elsevier.com/locate/jcpa

DISEASE IN WILDLIFE OR EXOTIC SPECIES

Polar Bear Encephalitis: Establishment of aComprehensive Next-generation Pathogen Analysis

Pipeline for Captive and Free-living Wildlife

Cor

ber

002

http

C. A. Szentiks*, K. Tsangaras*, B. Abendroth†, M. Scheuch†,M. D. Stenglein‡, P. Wohlseinx, F. Heeger*,k, R. H€oveler{, W. Chen#,W. Sun#, A. Damiani**, V. Nikolin*,**, A. D. Gruber††, M. Grobbel*,D. Kalthoff†, D. H€oper†, G. �A. Czirj�ak*, J. DeRisi‡, C. J. Mazzoni*,k,

A. Sch€ule‡‡, A. Auexx, M. L. East*, H. Hofer*, M. Beer†, N. Osterrieder**

and A. D. Greenwood*

*Leibniz Institute for Zoo and Wildlife Research, Alfred-Kowalke-Str. 17, 10315 Berlin, †Institute of Diagnostic Virology,

Friedrich-Loeffler-Institute, S€udufer 10, 17493 Greifswald e Insel Riems, Germany, ‡Department of Biochemistry and

Biophysics, University of California San Francisco, 1700 4th St., San Francisco, CA, USA, xDepartment of Pathology,

University of Veterinary Medicine Hannover, B€unteweg 2, 30559 Hannover, kBerlin Center for Genomics in BiodiversityResearch (BeGenDiv), K€onigin-Luise-Str. 6-8, 14195 Berlin, {Chemisches und Veterin€aruntersuchungsamt

Rhein-Ruhr-Wupper (CVUA-RRW), Deutscher Ring 100, 47798 Krefeld, #Laboratory for Novel Sequencing Technology,

Functional and Medical Genomics, Berlin Institute for Medical Systems Biology, Max-Delbrueck-Center for Molecular

Medicine, Robert-Roessle-Str. 10, 13125 Berlin, **Institute of Virology, Freie Universit€at Berlin, Philippstr. 13, Haus 18,

10115 Berlin, ††Institute of Veterinary Pathology, Freie Universit€at Berlin, Robert-von-Ostertag-Straße 15, Building 12,14163 Berlin, ‡‡Berlin Zoological Garden, Budapester Str. 32, 10787 Berlin and xxBerlin-Brandenburg State Laboratory,

Invalidenstr. 60, 10557 Berlin, Germany

resp

lin.d

1-99

://d

Summary

This report describes three possibly related incidences of encephalitis, two of them lethal, in captive polar bears(Ursus maritimus). Standard diagnostic methods failed to identify pathogens in any of these cases. A comprehen-sive, three-stage diagnostic ‘pipeline’ employing both standard serological methods and new DNAmicroarrayand next generation sequencing-based diagnostics was developed, in part as a consequence of this initial failure.This pipeline approach illustrates the strengths, weaknesses and limitations of these tools in determining path-ogen caused deaths in non-model organisms such as wildlife species and why the use of a limited number ofdiagnostic tools may fail to uncover important wildlife pathogens.

� 2013 Elsevier Ltd. All rights reserved.

Keywords: diagnostic tests; encephalitis; polar bear; wildlife diseases

Introduction

Infectious diseases are suggested to be a potential keyfactor in the decline and extinction of populations andspecies of conservation concern (deCastro and Bolker,2005; Hofer and East, 2012; MacPhee and

ondence to: A. D. Greenwood or N.Osterrieder (e-mail: no.34@fu-

e).

75/$ - see front matter

x.doi.org/10.1016/j.jcpa.2013.12.005

Greenwood, 2013). In comparison with theknowledge accumulated for model organisms,livestock and domestic animals, diagnosing thesource of mortality in non-model wildlife is a massivechallenge that faces a number of problems, even whenthe species concerned arewell known, charismatic andthe clinical signs are obvious and life-threatening orfatal (de Castro and Bolker, 2005; Wyatt et al., 2008;

� 2013 Elsevier Ltd. All rights reserved.

Polar Bear Encephalitis 475

Hofer and East, 2012). Polar bears (Ursus maritimus)are illustrative of such problems associated with thestudy of wildlife diseases. Knowledge of diseases inthis and other wildlife species is limited, whichmakes the identification of causative agents duringdisease outbreaks in captive wildlife particularlyproblematic (Aguirre et al., 2012). Non-specific clin-ical signs, such as seizures, may be caused by a widerange of causative agents, even when very conspicu-ous. Furthermore, the lack of knowledge of wildlifepathogens and their reservoirs may make it difficultin many cases to pinpoint the cause of fatalities.

When several deaths occur within a short timeframe in small and threatened wildlife populationsof conservation concern, and the affected individualsdisplay similar clinical signs, the diagnostic challengeis considerable. The question arises as to whether thisis an early warning of a disease outbreak that mightdeplete the population of an endangered species orthe effect of inappropriate husbandry or managementmeasures in captive or free-ranging situations,respectively, or perhaps a consequence of genetic pre-dispositions that may emerge in small and highlystructured populations with limited genetic variation.

Such concerns were raised in the veterinary andconservation communities, the zoo world, the generalpublic and the press when two polar bears, Knut ofBerlin Zoological Garden and Jerka of WuppertalZoological Garden, died unexpectedly after seizures,with no initial clinical signs to warn of a developinghealth problem, within a limited time frame of 2years. The health of a third bear, Lars (also in Wup-pertal Zoological Garden and the father of Knut) wasseverely compromised, but this animal recovered aftersuccessful veterinary intervention.

A detailed investigation into the death of both bearswas undertaken and blood samples from the third bearwere analyzed. Possible issues concerning husbandrywere considered, but were excluded. The observedclinical signs (i.e. seizures) can be caused by a widerange of pathogens and so it was unclear whether thedeaths were caused by the same agent. Initialscreening for obvious pathogens failed to reveal a causein the case of the two dead bears, at which point thecause of death would have normally been assigned asunknown. As a consequence of the initial failure, athree-stage diagnostic ‘pipeline’ employing both stan-dard serological methods and new DNA microarrayand next generation sequencing-based diagnosticswas developed. In this report, we describe the differentstages of this pipeline and discuss the strengths, weak-nesses and limitations of these tools in determiningpathogen-associated deaths in non-model organismsandwhy the use of a limited number of diagnostic toolsmay fail to uncover important wildlife pathogens.

Materials and Methods

Case Material

Knut was born in the Berlin Zoological Garden, Ger-many, inDecember 2006, to awild-born female caughtin Canada and a captive-born male, Lars(Blaszkiewitz, 2009).Knut gainedworldwide attentionwhen, as a cub, he was rejected by his mother and sub-sequently hand-reared by a zookeeper. The zoo-keeper’s interactions with Knut were held inpublic and the bear therefore became the focus ofmuch media attention throughout its life. As a result,manyhours of professional and amateur filmarewidelyavailable of Knut’s life, including his sudden seizuresfollowed by death by drowning in themoat of the polarbear enclosure at the Berlin Zoological Garden inMarch2011.Medical files did not indicate health prob-lems prior to these seizures or in any of the co-housedpolar bears or other bear species in neighbouring enclo-sures.

Lars lived in Wuppertal Zoo from October 2009 to2012 and never had direct contact with Knut. Helived with a female polar bear, Jerka, and both suf-fered seizures in 2010. With medical interventionLars survived but Jerka died. The cause of deathwas identified as a recombinant equine herpesvirus(EHV)-1/EHV-9 derived from zebras (Greenwoodet al., 2012). The same study found no evidence thatLars and Knut’s illnesses were related, because re-combinant EHVs were not detected by genetic orserological methods (Greenwood et al., 2012).

Pathological Examinations

Knut’s body remained in the enclosure moat forapproximately 48 h before submission of the carcassfor necropsy examination. Cerebrum, brainstem,serum, blood, lung, liver, spleen, kidney, small andlarge intestine and faeces were collected from Jerkaand serum, faeces and saliva were collected fromLars. Additional clinically healthy polar bear controlsamples were collected as follows: whole blood sam-ples from four bears, brain samples from two bearsand serum samples from six bears. The experimentsundertaken in this project were approved by the In-ternal Ethics Committee of the Leibniz-Institute forZoo and Wildlife Research (IZW), Approval No.2012-04-01. Further descriptions of all polar bearsstudied can be found in Greenwood et al. (2012).

Prior to routine necropsy examination, samples ofcerebrospinal fluid (CSF) were aspirated from Knutas described for dogs (Gwalter and Wegmann-Ehrenspenger, 2006) and smears were stained withmethylene blue (B€ock, 1989). Fresh tissue samplesand samples fixed in 4% neutral buffered formalin

Table 1

PCR and immunohistochemistry screening

Pathogens IHC results PCR results

Viruses

Adenoviruses Negative NegativeBornaviruses Negative† Negative

Canine distemper virus

(CDV)*

Negative Negative

Cytomegalovirus* Not tested NegativeEncephalomyocarditis

virus

Not tested Negative

Enterovirus (ENTC)* Not tested NegativeEquine herpesvirus Negative† Negative

Feline herpesvirus* Negative† Negative

Feline leukaemia virus

(FeLV)

Negative Not tested

Hanta virus Not tested Negative

Influenza Not tested Negative

Inkoo virus Not tested Negative

Pan flavivirus* Not tested NegativePan herpes Not tested Negative

Pan morbillivirus Not tested Negative

Parvovirus Negative† Negative

Pseudorabies Not tested NegativePoxviridae Not tested Negative

Rabies* Negative Negative

Schmallenberg virus Not tested NegativeSindbis virus Not tested Negative

Tahyna virus Not tested Negative

Tettnang virus Not tested Negative

Tick borne encephalitis* Negative† NegativeVenezuelan equine Not tested Negative

476 C.A. Szentiks et al.

were taken for further investigation. Samples of cere-brum, cerebellum, brainstem, spinal cord, CSF, blood,lung, lymph nodes, thymus, spleen, liver, bile, small in-testine, large intestine, kidneys, faeces, fat, muscles andbones were frozen at �80�C for further investigations.Samples from Jerka including cerebrum, brainstem,serum,blood, lung, liver, spleen, kidney, small intestineand large intestine, which had been stored at �80�Csince her death, were obtained. After fixation, tissueswere processed routinely and embedded in paraffinwax. Sections (3 mm) from all of the organs fromKnut and of the samples from Jerka were stainedwith haematoxylin and eosin (HE). Sections from thebrain were stained by Gram, periodic acideSchiffand Grocott’s silver stain (B€ock, 1989; Stevens andFrancis, 1996). Sections from spleen and lymphnodes were stained with Perl’s Prussian blue (B€ock,1989).

For electron microscopical examination, frozenbrain tissue and formalin-fixed tissues from Knutwere fixed in 3%glutaraldehyde,washed in phosphatebuffered saline (PBS, pH 7.2) and fixed in 2% osmiumtetroxide. Dehydration was performed in ethanol andsamples were embedded in EPON812 before prepara-tion of ultrathin sections and staining with uranyl ace-tate and lead citrate.Ultrathin sectionswere examinedwith a Tecnai Spirit� transmission electron micro-scope (Fei, Eindhoven, The Netherlands).

encephalitis virus

(VEEV)

Western equineencephalitis virus

(WEEV)*

Not tested Negative

West Nile virus (WNV) Negative† NegativeParasites

Acanthamoeba Not tested Negative

Naegleria fowleri Not tested Negative

Pan-Plasmodium* Not tested NegativeToxoplasma gondii* Negative Negative

Fungi

Coccidioides spp.* Not tested Negative

Cryptococcus neoformans* Not tested NegativeHistoplasma* Not tested Negative

Bacteria

Borrelia spp. Not tested NegativeListeria Not tested Negative

Streptococcus spp. Not tested Negative

*Primers produce correct size band, but sequencing results indicated

that the sequence amplified product was either polar bear or other

carnivore related sequence.†The sample was tested by IHC for Jerka and Knut.

Immunohistochemistry

Immunohistochemistry (IHC) was performed todetect 11 different viruses, the apicomplexan parasiteToxoplasma gondii, as well as the lymphoid markersCD3 and CD79a using embedded brain tissue fromboth bears (Table 1 and Supplementary Table S3).Briefly, following dewaxing, tissue sections wereimmersed in H2O2 0.5% in methanol for 20 min.Non-specific primary antibody binding was blockedby incubation with inactivated normal serum diluted1 in 5 in PBS (pH 7.1). Primary antibody diluted inPBS was applied and tissue sections were incubatedin a humid chamber at 4�C overnight. After washing,the sections were incubated with biotinylated second-ary antibody. The antibody against influenza A viruswas directed against the nucleoprotein of the virus(Nicholls et al., 2012). For rabies, the secondary anti-body was directed against fluorescein isothiocyanate(FITC), which is coupled to the primary antibody.The detection system used was the avi-dinebiotineperoxidase complex method (VectorLaboratories, Burlingame, California, USA). Thechromogenic solution consisted of 0.05% 3,30-diami-nobenzidine tetrahydrochloride with 0.03% H2O2

as the substrate in 0.1 M Tris-buffered saline (pH

7.6). Tissue sections were counterstained withMayer’s haematoxylin. As negative controls, tissuesections from a polar bear without encephalitic lesionswere used. In addition, the primary antibody was re-placed by ascites fluid from non-immunized BALB/cJmice (Cedarlane, Ontario, Canada) or rabbit serum

Polar Bear Encephalitis 477

(SigmaeAldrich, Taufkirchen, Germany). As posi-tive controls, tissues sections from confirmed cases ofinfection with the respective agents were used.

Bacteriology and Parasitology

Bacteriological investigations were only performed onsamples from Knut. Swabs from the frontal sinus,brain surface, cerebrum, CSF and alveolar socket ofa canine tooth were taken. Samples were cultured onColumbia agar with 5% sheep blood, Gassner’sagar, chocolate agar (all from Fisher ScientificGmbH, Schwerte, Germany) and CHROMagarOrientation (MastDiagnosticaGmbH,Reinfeld,Ger-many) under aerobic, capnophilic (5% CO2) andanaerobic conditions. Dependent on the results of theGram stain, as well as the catalase and oxidase tests(BD, Heidelberg, Germany), biochemical identifica-tion of the bacteria was performed by using theAPI� Identification System (bioM�erieuxDeutschlandGmbH,N€urtingen, Germany) (B€ock, 1989). Parasito-logical analysis, primarily to exclude trichinellosis, wasperformed in the Institute for Food, Drugs and Epizo-otics according to European act No. 2075/2005(Gylstorff, 1987; European Commission, 2005).

Serology

Serum samples from Lars and six control polar bearsand whole blood samples from Knut were examinedfor the presence of antibodies against influenza virustype A nucleoprotein (ID Screen� Influenza A Anti-body Competition ELISA kit, ID-vet, Montpellier,France; IDEXX Influenza A AB test, Ludwigsburg,Germany) and against the N1 protein (ID Screen�

Influenza N1 Antibody Competition ELISA kit,ID-vet). To verify positive results from the ELISA,haemagglutination inhibition (HI) was performedaccording to the OIE Manual of diagnostic testsand vaccines for terrestrial animals (OIE, 2012a).Serum as well as whole blood samples tested againstinfluenza A H1N1 pandemic strain 2009 and anH5N1 strain reacted non-specifically. Material forHI after pre-treatment (e.g. with receptor destroyingenzyme) was not available. Blood from Knut andanother control bear (Anton) and serum from thelatter were also examined for the presence of neutral-izing antibodies to EHV-1, EHV-4 and equine arter-itis virus (EAV) using microneutralization assays aspreviously described (OIE, 2012b, c). A caninedistemper virus (CDV) serum neutralization test(SNT) was performed on the same samplesfollowing the protocol described by de Vries et al.

(1988), with the modification of CDV strain usedfor VNT. The Onderstepoort strain of CDV (GB

number: AF014953), kindly provided by V. vonMes-sling (INRS-Institut Armand-Frappier, Laval,Quebec, Canada), was used.

Polymerase Chain Reaction

Molecular analysis was performed using brain (telen-cephalon caudal to the right olfactory bulb, lateralhemisphere of the cerebellum and brainstem), liver,spleen and lung from Knut and brain (cerebrumand cerebellum) and liver from Jerka. The polarbear tissue samples were homogenized and extractedusing the Qiamp DNAMini kit and RNeasy lipid tis-sue kit (Qiagen, Hilden, Germany) following themanufacturer’s protocols. Polymerase chain reactions(PCRs) performed on DNA- and RNA-based patho-gens were analyzed by PCR or reverse transcriptase(RT)-PCR, respectively. Primer sequences aredescribed in Supplementary Table S1. PCRs andRT-PCRs were performed using My-taq (BiolineGmbH, Luckenwalde, Germany), HiFi supermix(Invitrogen, Darmstadt, Germany), Superscript III(Invitrogen) and Superscript III one-step RT-PCRaccording to the manufacturers’ protocols. All PCRswere performed at least in duplicate. PCR productswere visualized after electrophoresis in 1% agarosegels using GelRed nucleic acid gel stain (BiotiumInc., San Francisco, California, USA). Amplificationproducts with correct sizes, when detected, weresequenced using appropriate forward and reverseprimers (StarSeq GmbH, Mainz, Germany). Whenmultiple bands were generated, PCR products werecloned using the pGEM vector system (Promega,Madison, Wisconsin, USA). Positive colonies wereidentified by colony PCR and cleaned of reagentsand primers usingMSB Spin PCRapace (StratekMo-lecular GmbH, Berlin, Germany). Multiple cloneswere sequenced (Starseq) using M13 forward andreverse primers. All sequences were aligned againstGenBank sequences using BLASTn to search for ho-mology to pathogens deposited in the database(www.ncbi.nlm.nih.gov/).

DNA Microarrays

For virochip analysis RNA was first extracted fromtissue samples using Qiagen RNeasy Lipid tissuemini kit. For enrichment, total RNA extractionswere further processed following the manufacturers’protocols with Terminator exonuclease (EpicentreBiotechnologies, Madison, Wisconsin, USA) to re-move rRNA or with oligotex mRNA mini kit forpoly-A mRNA isolation (Qiagen). RNA was reversetranscribed using Superscript III (Invitrogen) withprimer 50-CGC-TCT-TCC-GAT-CTN-NNN-NN-

478 C.A. Szentiks et al.

30. Following reverse transcription, cDNA was syn-thesized using T7DNA polymerase (Fermentas,Wal-tham, Massachusetts, USA) and the same primer tosynthesize products tagged at both ends. These wereamplified by PCR using primer: 50-CTG-TCT-GGC-TCT-TCC-GAT-CT-30 and GoTaq DNA po-lymerase (Promega). Thermocycling conditionswere: 95�C for 2 min; three cycles of 95�C for 30 sec,40�C for 30 sec and 72�C for 1 min, then 25 cyclesof 95�C for 30 sec, 58�C for 30 sec and 72�C for1 min, with a final extension of 72�C for 5 min.

To fluorescently label libraries for virochip hybridi-zation, a portion of each library was amplified byPCR as above with a dNTP mixture including 75%of 5-(3-aminoallyl)-dUTP (Ambion, Carlsbad, Cali-fornia, USA) instead of dTTP, which is normally inthe mixture. PCRs were purified using a DNA Cleanand Concentrator-5 column (Zymo Research, Irvine,California, USA). Eluate was heated at 95�C for2min, cooled on ice, then labelled in reactions contain-ing 100 mM sodium bicarbonate pH 9.0, 10%dimethyl sulphoxide and 667 mM Cy3 mono-NHSester (GE Healthcare, Little Chalfont, UK) for 1 h at25�C. Labelled DNA was purified as above and addedto hybridization reactions containing 3� SSC, 25mMHEPES pH 7.4 and 0.25% sodium dodecyl sulphate(SDS). Hybridization mixtures were heated at 95�Cfor 2 min, then added to microarrays and hybridizedovernight at 65�C. Arrays were washed twice in0.57 � SSC and 0.028% SDS and twice in0.057 � SSC, then scanned on an Agilent XYZ scan-ner. Three toolswere used to analyse array data:E-pre-dict (Urisman et al., 2005), Z-score analysis (Chiu et al.,2006) and cluster analysis (Eisen et al., 1998).

For a second microarray screening using a differentchip, the pan-viral DNA-microarray platformBiochip5.1 for the detection of approximately 2,000 differentviruses was used (Gurrala et al., 2009). Isolated RNAand DNA from brain and liver of Knut and Jerkawere tested. The genetic material from Jerka wasused as a control in the investigation. For sample prep-aration, an established procedure involving a separatecDNA and second-strand synthesis of RNA and DNAwas used. Prior to labelling, the second-strand synthe-sis products of RNA and DNA were pooled and thelabelled DNA was hybridized onto the array. The mi-croarrays were scanned with an Axon GenePix 4300A(MolecularDevices, Sunnyvale, California,USA) andthe raw data were extracted with GenePixPro7 soft-ware (Molecular Devices).

Next-generation Sequencing

Extracted genetic material from brain and liver sam-ples from Knut and Jerka was used as input material.

Library preparation was performed using the SPRI-works System II (Beckman Coulter Fullerton, Cali-fornia, USA) for Roche GS FLX DNA Sequencer(Roche, Basel, Switzerland). The resulting librarieswere then prepared for 454 sequencing accordingthe manufacturer’s instructions.

Illumina sequencing was performed on 100 ng ofrRNA-depleted RNA that was fragmented. RNA-seq library preparation was carried out as describedpreviously (R Development Core Team, 2009).RNA-seq was performed on a HiSeq 2000 sequencingplatform with 1 � 100 cycles of sequencing, followingthemanufacturer’s instructions (Illumina, SanDiego,California, USA).

Bioinformatics

Statistical analysis of the array data was performed inR Development Core Team (2009). An R-packagefor the standard microarray evaluation method‘Limma’ (Smyth and Speed, 2003; Wettenhall andSmyth, 2004; Smyth, 2005; Smyth et al., 2005;Ritchie et al., 2007) and the data normalizationmethod ‘ZScore Transformation’ (Cheadle et al.,2003) were used. Polar bear-derived Illumina se-quences were removed in several steps of increasingcomputational complexity. Firstly, sequences weresearched against a custom database of bear ribosomaland mitochondrial sequences using the BLASTn al-gorithm (Altschul et al., 1997). Sequences producingalignments of $95% identity over 90 nucleotideswere removed. Next, sequences were searched againsta draft assembly of the polar bear genome (Stengleinet al., 2012) using BLASTn. Sequences producingalignments of $90% identity over 90 nucleotideswere removed. Next, low complexity sequences wereidentified by compression with the Lem-peleZiveWelch algorithm (Welch, 1984). Sequenceswhose compressed length was less than 0.49� the sizeof their uncompressed length were filtered. Remain-ing sequences were searched against the NCBI non-redundant nucleotide (nt) database using BLASTn.Sequences with hits to database sequences with E-values #1 � 10�8 were removed from further anal-ysis. Sequences were also searched against custom da-tabases of viral protein sequences using the BLASTxalgorithm.

Reads that hit any viral sequence with an E-value#0.001 were assigned to the database entry with thelowest E-value. For each of the latter hits, a specieswas determined and a P value was computed. The re-sulting dataset was then analyzed to find species thathadmultiple non-overlapping hits to different parts ofits genome. To estimate an overall probability mea-sure for each occurring species, the P values of non-

Polar Bear Encephalitis 479

overlapping hits were considered independent andthus multiplied. For each group of overlapping hits,the smallest P value was used, as overlapping hitscannot be considered independent and the minimumP value gives an upper bound for the combinedP value of the group.

In a second approach to analyse the Illuminashotgun data, the reads were assembled into contigsafter filtering out reads matching to rRNA, 7S RNAand polar bear mitochondrial DNA. The assemblywas carried out with Velvet (version 1.2.03; EuropeanBioinformatics Institute, Cambridge, UK) using stan-dard parameters and a hash length of 23. Out of theabove results, contigs longer than 200 base pairs(bp; average length 254.64 bp, minimum 200 bp,maximum 528 bp, average coverage 6.77, minimum1.33, maximum 65.35) were selected and BLASTnsearches against the NCBI nucleotide database were

Fig. 1. The research pipeline. The three steps of the pipeline are shownidentification (ID) is a possibility. When not identified at eachequivocal ID, such as with serological evidence, but lack ofthroughout. Black boxes indicate positive or equivocal ID foreach stage. Grey boxes indicate negative ID at each stage. Thfar right of the figure. HE, haematoxylin and eosin histopathtest; ELISA, enzyme linked immunosorbent assay; HIA, haem

performed. Each contig was assigned to a species byits best BLASTn hit. In contrast to the Illumina con-tig coverage, the GS FLX sequence coverage was anaverage of 2.13, a minimum of 1 and a maximum of15 reads.

Results

Overview of the Analysis Pipeline

Fig. 1 outlines the pipeline employed to examine sam-ples from Knut. In all cases, negative captive polarbear controls representing living individuals or thosewho had died of causes differing from Knut (i.e.non-encephalitic disease), were included for compar-ative purposes. In all experiments, tissue samples fromJerka, who had died from infection with a zebra-derived recombinant EHV-1/EHV-9 (Greenwoodet al., 2012), served as a positive control for similar

with the possible outcomes at each stage. For each step, pathogenindividual stage, the sum of the three stages will yield an ID or angenomic evidence for a pathogen. No ID remains a possibilityeach sub-stage. The red boxes below indicate positive outcomes ofe three possible outcomes post pipeline analysis are shown at theology; IHC, immunohistochemistry; SNT, serum neutralizationagglutination inhibition assay.

480 C.A. Szentiks et al.

tissue samples from Knut and Lars. Tissues fromJerka that tested negative for a range of pathogenswere included as negative controls for these pathogensin the pipeline analysis. As PCR and quantitativePCR (qPCR) showed that Jerka was infected with re-combinant zebra EHV-1/EHV-9, we expected thatshe should differ in her pathogen profile from Knutand be positive for EHV-1 by non-target specific ge-netic screens.

Necropsy Examination

Knut weighed 305 kg and was in good overall condi-tion. The cerebrum showedmild asymmetry of the ol-factory bulb region. Caudal to the right olfactorybulb was a small convex area. The meningeal vesselsand capillaries were congested (Fig. 2). After fixationof the brain, transverse sections were performed, butno detectable asymmetry or dilation of the lateralventricles was visible. Consistent with drowning,large quantities of water were present in the respira-tory tract and paranasal sinuses. The deciduousmaxillary canine teeth were present and both werefractured. On the right side this was associated withpurulent alveolitis. The first deciduous maxillary pre-molar and both deciduous mandibular canine teethwere also present. The lung showed marked conges-tion, alveolar oedema and moderate alveolar emphy-sema. Paltauf spots were not detected. The stomach

Fig. 2. Neuropathological findings inKnut (left) and Jerka (right). Paview of the cerebral cortex with perivascular non-suppurativearrows) and meningeal vessels in Knut only. Panel B shows a himicroglial infiltration (green arrows) and few eosinophils (red

was filled with water and the remains of a meal. Incontrast, Jerka weighed 245 kg and was in excellentcondition. The only relevant gross findings were con-gested meningeal vessels.

Histopathology and Electron Microscopy

Examination of Knut’s brain revealed severe multi-focal lymphoplasmacytic and eosinophilic panmenin-goencephalomyelitis. The cellular infiltration in thecerebrum was perivascular and perineuronal(Fig. 2). Moderate multifocal nodular microglial pro-liferation was detected. Moderate cortical neuronalnecrosis could be seen with an eosinophilic appear-ance of the neurons with central chromatolysis andpartly peripheral nuclei. In the cerebral cortex, lym-phocytic and eosinophilic inflammation of thechoroid plexus with focal perivascular haemorrhageswas identified. Cerebellar changes comprised onlyperivascular inflammation. The inflammation in thedura mater, meninges and spinal cord was highly var-iable and affected areas presented with mild to severeinflammation. Generally, inflammation was more se-vere in the cerebrum and brainstem than in the cere-bellum, spinal cord and meninges. However,inflammatory lesions did not show a predilection fora specific cerebral region, although the mild asymme-try at the olfactory bulb would suggest this. Myelitisdecreased from cranial to caudal levels. Inclusion

nel A shows histological images withHE staining. All show an over-inflammation located around blood vessels in both animals (blackgher magnification of grey matter inflammation with perineuronalarrows) in Knut.

Polar Bear Encephalitis 481

bodies, parasites, fungi, bacteria or other possibleagents were not detected.

In contrast to the brain, only minor changes unre-lated to the encephalitis were detected in other or-gans. Associated with the right deciduous maxillarycanine tooth was a focal, mild, purulent and lympho-histiocytic alveolitis. Within the submandibular,popliteal and inguinal lymph nodes and the thymus,haemorrhages and isolated macrophages containinghaemosiderin were noted. Despite congestion, thespleen showed isolated macrophages with haemosi-derin deposition, but no evidence of follicular hyper-plasia. Pulmonary lymph nodes had mildanthracosis. Mild, lymphohistiocytic rhinitis was pre-sent. The lung was filled with eosinophilic fluid andforeign bodies (e.g. small parts of plants and sand-like particles).

Electron microscopical inspection of frozen braintissue was inconclusive due to the poor condition ofthe samples. Several fixed brain tissue samples werealso examined by electron microscopy, but therewas no evidence for viral particles in the brain. Theresults were, however, consistent as no inclusionbodies were identified.

Histological examination of Jerka’s brain revealedmultifocal mild to severe non-suppurative meningo-encephalitis with gliosis and mostly perivascular lym-phohistiocytic infiltrates. There was no evidence ofinclusion bodies, parasites, fungi, bacteria or otherpossible agents. The other organs showed no abnor-malities.

Bacteriology and Parasitology

Bacteriological examination of swabs from Knut’salveolar socket and frontal sinus revealed a diverse,but non-specific aerobic and anaerobic flora in vary-ing combinations (Fig. 3). As the histopathological re-sults revealed no signs of bacterial infection orintoxication, no further bacteriological diagnosticprocedures were performed. All identified bacteriawere either components of the normal flora of the res-piratory tract mucosa and the alveolar socket or likelyoriginated from the inhaled water from the enclosuremoat, including faecal contamination (Uchida et al.,1995; Schuster, 1999). Bacteria isolated andcharacterized from the brain most likely spread tothe cranial cavity during decomposition over the48 h elapsing between death and necropsyexamination, while Knut’s carcass was floating inthe moat. These results were supported by the IHCfindings (see below). Jerka’s brain was not subjectedto bacteriological examination.

No ectoparasites or endoparasites, including Trich-

inella spp., were identified grossly in Knut. Parasites

or parasite eggs were not detected in any histologicalsections. Similarly, no parasitic infection was identi-fied in Jerka.

Immunohistochemistry and Polymerase Chain Reaction

A broad panel of viral and non-viral encephalitis-causing pathogens, representing DNA and RNA vi-ruses, bacteria, fungi and parasites were tested forby PCR and IHC (Table 1, Supplementary TableS1). All IHC results were negative. Approximately30% (and up to 50%) of the perivascular and peri-neuronal inflammatory cells were CD3+ T lympho-cytes. Labelling for CD79a+ B lymphocytes andplasma cells was more variable (approximately 10%of cells were clearly labelled with 40e50% of cellsshowing a weak reaction that was likely non-specific).

Although PCR products were obtained in somepathogen-specific reactions, the product sequencesdid not originate from pathogens (Table 1). None ofthe 37 tested viral, bacterial and parasite pathogenswere found to be associated with the encephalitis.Sensitivity is likely to be an issue where primer mis-matches exist between known and unknown or novelstrains, particularly for viruses, and this could yieldfalse-negative results. However, themajority of the as-says were ‘pan-pathogen’ PCRs designed to avoidsuch limitations (Supplementary Table S1). Thesame PCR assays were negative for all pathogens inbrain samples from Jerka, except that a zebra EHV-1 strain was identified as described in Greenwoodet al. (2012) using pan-herpesvirus primers (Table 1,Supplementary Table S1).

Screening by DNA Microarray

Two microarray platforms were employed for path-ogen identification. The first, the Virochip, has usu-ally been applied to human cases of disease (Wanget al., 2002) and has also been successfully used toidentify novel pathogens from non-human animals(Kistler et al., 2008; Stenglein et al., 2012). Itcontains sequences from approximately 3,000 mostlyanimal viruses. Recently, however, it has beendemonstrated that its sensitivity is lower than that ofshotgun sequencing from purified viral particlesfrom a study of human serum (Yozwiak et al., 2012).The second microarray, Biochip 5.1, was developedin a project within the EU Network of Excellencefor Epizootic Disease Diagnostic and Control (EPI-ZONE) (Gurrala et al., 2009) and is a diagnostictool for the detection of approximately 2,000 virusesat the genus, species and genotype level. Given themethodological similarity between both virus chips,sensitivity is likely to be similar for both arrays.

Fig. 3. Bacteria and fungi identified by shotgun sequencing andbacterial culture. Identifiedbacterial families are shownon the left. Tissuesof origin are indicated on the right and shown by respective bear within the figure. Blue lines indicate bacteria identified frommul-tiple hits to different parts of the genome by Illumina shotgun sequences. Red lines indicate bacterial sequences identified by GSFLX. Bacterial families found by sequencing and culture are indicated for Knut by a blue circle. Bacteria found by more thanone method are indicated by a blue triangle. Details about bacterial identification are shown in Supplementary Table S2.

482 C.A. Szentiks et al.

Polar Bear Encephalitis 483

Both arrays were hybridized with cDNA and DNAfrom brain and liver from Knut and Jerka. As re-ported in Greenwood et al. (2012), the Virochip didnot produce any signals above background for anyof the samples tested from either polar bear. Thus, itdid not detect EHV-1 DNA, which was easilyretrieved by PCR from Jerka. An identical resultwas obtained for Biochip 5.1.

Next-generation Sequencing

DNA libraries prepared fromRNA andDNA isolatedfrom the brains and livers of both polar bears weresequenced on a GS FLX platform and yielded83,760 reads from Knut’s brain and 66,353 fromKnut’s liver and 66,952 reads from Jerka’s brain

Fig. 4. Viruses identified by shotgun sequencing. Identified viral grouand shown by respective bear within the figure. Blue lines indgenome by Illumina shotgun sequences. Red lines indicate viralviral groups, were identified consistently and are indicated nexSupplementary Table S2.

and 97,973 from Jerka’s liver. Given that necropsy ex-amination of Knut occurred 48 h after death, viralenrichment was not performed as it was consideredlikely that host and viral degradation would havereached a point where removal of host RNA andDNA would also remove viral nucleic acids. Dataanalysis was performed by a workflow, which iscurrently under development. The workflow com-bines various alignment programs and reads are clas-sified into separate taxa based on sequencesimilarities. The results are documented in Figs. 3and 4. Of note, no viral hits were observed for Knut’sbrain with the exception of Cafeteria roenbergensis, a vi-rus that is known to infect marine invertebrates. Jer-ka’s brain displayed reads with homology to avianretroviruses and herpesviruses and further reads

ps are shown on the left. Tissues of origin are indicated on the righticate viruses identified from multiple hits to different parts of thesequences identified by 454 FLX.Retroviruses, in contrast to othert to their respective lines. Details of viral identification are shown in

484 C.A. Szentiks et al.

with homology to herpesviruses were also detected inJerka and Knut’s livers. Various reads with limitedhomology to several viruses (Megavirus chilensis, hu-man endogenous retrovirus, avian myeloblastosis-associated virus, human immunodeficiency virusand a few gallid herpesviruses) were identified in theliver reads from Knut. All reads showing homologiesto viral sequences were BLASTn searched against asequencing database generated from all previouslyperformed 454 sequencing runs. As no significant ho-mologies were found, contamination during librarypreparation, emulsion PCR and sequencing are un-likely to be the source and the sequences thereforevery likely originate from nucleic acids derived fromthe bears.

A second approach was shotgun Illumina (HiSeq2000) sequencing of cDNA from the same tissues afterpartially normalizing the libraries for ribosomalRNA, which makes up the largest fraction of all ex-pressed RNAs. In contrast to the GS-FLX, the Illu-mina reads are much shorter (100 bp), but coverageis much higher than with the GS-FLX (hundreds ofmillions of sequences as opposed to tens of thousands).The resulting data were curated by subtracting thedraft genome sequence of the polar bear (Li et al.,2011). Reads were aligned using the BLASTn algo-rithm and sequences aligning with $90% identityover 90 bases were subtracted. Sequences were simi-larly filtered using databases of bear ribosomalRNA and mitochondrial sequences. Low complexitysequences were also removed from further analysis(Welch, 1984). The numbers of reads remaining afterfiltering were 4,552,863 and 5,156,929 for Knut’sbrain and liver and 4,312,537 and 2,833,574 for Jer-ka’s brain and liver, respectively. Several methodswere employed to search for possible pathogen-derived sequences. Firstly, the filtered reads weresearched against the NCBI nucleotide (nt) databaseusing the BLASTn algorithm and an E-value cut offof 1 � 10�8 (Fig. 4). No virus sequences were identi-fied; however, the relatively short read length(100 bp) may have limited the sensitivity of this strat-egy. Secondly, the reads were searched against data-bases of viral protein sequences using the BLASTxalgorithm. Matches with an E-value <0.001 werefurther analyzed. Each read was assigned to a virusspecies based on its best match and viruses with mul-tiple non-overlapping hits were determined. Severalputative viral hits were identified (Fig. 4).

Given the possibility that short single reads hamperpathogen identification by BLASTn methods, a de-novo assembly of the reads was performed. Readsthat were not excluded by filtering for polar beargenomic DNA were assembled into contigs with Vel-vet (Zerbino and Birney, 2008). The assembly re-

sulted in 3,079,082 and 210,374 contigs for thesamples from Knut’s brain and liver, respectively,and 919,464 and 83,602 contigs for the samplesfrom Jerka’s brain and liver, respectively, with anN50 score of 132, 17, 100 and 15, respectively. Contigslonger than 200 bp were selected and a BLASTnsearch against the NCBI nucleotide (nt) databasewas performed. Most of the contigs gave matches tosequences from mammals and therefore most likelyoriginated from the polar bears. Microorganism se-quences identified by this approach are summarizedin Fig. 3.

The different sequencing methods overlappedpoorly in the viral hits identified with a few exceptionsat the viral group level. Hits to various gammaretro-viruses and betaretroviruses were observed in the tis-sues from both polar bears, particularly with Knut’sbrain sample, which had a somewhat better enrich-ment in the shotgun sequencing. Obtaining longer se-quences of the betaretrovirus by PCR and queryingthe draft polar bear genome revealed that a novel po-lar bear endogenous retrovirus had been identified. Asimilar result was obtained for the betaretrovirusidentified in Jerka’s liver. The betaretrovirus hasbeen described in detail as a newly discovered bearretrovirus elsewhere (Mayer et al., 2013). These re-sults reveal that viral sequences were detected fromthe polar bear samples by next generation sequencingmethods that were of bear origin. In terms of overallvirus similarity, retroviruses and herpesviruses wereidentified in both datasets even though the specificmatches differed. None were specific to the brain sam-ples and none to Knut’s brain. These overlappingviral groups are likely to represent species-specific vi-ruses of polar bears. Notably, none of the methodsretrieved the recombinant EHV-1 that was identifiedby PCR from Jerka. Thus, sensitivity is a critical issuefor samples that cannot be enriched for virus particlesprior to analysis irrespective of the high throughputmethod applied.

The bacterial results obtained by different methodswere more consistent than those for viruses as shownin Fig. 3. Enterobacteriaceae and Clostridiaceaewere consistently found by both sequencing ap-proaches and by bacteriological analysis. Similarly,Bacillaceae and Corynebacteriaceae were found bysequencing and by bacterial culture. However, theonly bacterium exclusively identified in Knut’s brain,but not Jerka’s brain, was Propionibacterium acnae, aspecies usually associated with topical acne.

Serology and Defined Immunohistochemistry

Blood from an unrelated polar bear (Anton) wasfrozen and thawed to mimic the state of Knut’s whole

Polar Bear Encephalitis 485

blood. Serum from Anton was also available. Assayinterference from whole blood versus serum couldtherefore be controlled in serological assays. Knuthad been vaccinated against CDV at 1 year of ageand thus had a low serum neutralizing antibody titreto this virus (not shown). All other tests were negative,except for the influenza ELISA (ID Screen� Influ-enza A Antibody Competition ELISA kit, ID-vet).This assay was run repeatedly and was positive forthe influenza nucleoprotein (NP) in Knut’s bloodonly. Whole blood samples from Anton and Jerkawere negative. Thus, we concluded that the positiveresult fromKnut was not caused by non-specific reac-tions of whole blood in the assay. In contrast, a secondNP assay from a different company (IDEXX Influ-enza AAB test) was negative. Knut’s blood also testedpositive in an ELISA assay targeting influenza virusN1 (ID Screen� Influenza N1 Antibody CompetitionELISA kit, ID-vet). Results from HI assays wereinconclusive because all bear blood and serum sam-ples showed non-specific inhibition of haemagglutina-tion, suggesting that there may be a factor in bearblood and serum that generally interferes with suchassays. A follow-up influenza-specific IHC antigentest on brain sections with particularly strong inflam-mation was negative (Nicholls et al., 2012). The testused a commercially available monoclonal antibodyHB65 (European Veterinary Laboratory, Woerden,The Netherlands) specific for the NP protein of influ-enza A virus. The test was controlled with lung sec-tions from pigs infected experimentally withinfluenza A virus, which invariably tested positive.In summary, evidence for the presence of influenzaA antigen was consistently negative, although anti-bodies against influenza A NP and N1 were detectedin assays using Knut’s samples, but not in samplesfrom the other polar bears tested.

Discussion

During 2010 and 2011 there were three documentedcases of encephalitis in captive polar bears in twozoological collections, including the related bearsLars and Knut. Additional cases of fatal encephalitisin polar and black bears have been reported previ-ously (Schrenzel et al., 2008; Wohlsein et al., 2011),but definitive diagnosis in such cases is complicatedby the extensive list of potential causative pathogens.

The extensive effort made to determine the cause ofthe observed encephalitis in these bears was requiredbecause of the gap in knowledge of polar bear diseasesspecifically and wildlife diseases in general. The anal-ysis undertaken, however, revealed the strengths andweaknesses of current methodologies when applied towildlife, but also helped us to establish a pipeline that

could be successful in achieving a diagnosis, even inproblematic cases. It became apparent that the cur-rent state of the art, advanced as it has become, is stillnot able to provide unequivocal diagnosis in all cases.Each stage of the pipeline was designed to identify oneor more probable or suspected candidates for furtherstudy while ruling out others. Comparing the resultsfor Jerka (infection with recombinant EHV-1) withthose for Knut (no conclusive pathogen) is instructivefor a prediction on how such a ‘species-blind’ pipelinewould perform in other cases.

Initial necropsy examination and histopathologicalevaluation of tissues is a necessary step. In both cases,these studies lead to a likely diagnosis of viral enceph-alitis. Eosinophilic inflammation suggesting sodiumchloride poisoning was observed, but such poisoningis associated with cerebral oedema and lamellardegeneration (Zachary, 2007), neither of whichwere detected. No parasites were identified, althoughparasites such as Acanthamoeba spp. could not beconclusively ruled out. Interestingly, mimiviruses spe-cific to Acanthamoeba spp. were among the viral readsobtained for Knut, although PCR and shotgunsequencing were negative for the parasite. As Knutwas in water for an extended period before patholog-ical examination, the DNA could have been acontamination from the water. The pathology wasinconsistent with a bacterial cause in the case ofKnut. Additionally, bacteria isolated from the brainduring bacterial investigation comprised only compo-nents of the normal flora and the inhaled water of theenclosure. Although Jerka’s brain was not specificallyinvestigated by bacteriological diagnosis, no subse-quent evidence was found for a bacterial cause ofdeath. Histology was negative for multiple pathogens.Thus, the end result of the first stage of the diagnosticprocedure in both cases was an equivocal diagnosis ofvirus infection.

Molecular diagnostics have developed rapidly fromPCR-based approaches tomicroarray-based detectionto next-generation sequencing. Each method has ben-efits and drawbacks that have prevented any one fromcompletely supplanting the others. PCR is extremelysensitive, but prone to false-negative results in thecase of primer and target mismatch, a particularlylikely situation for novel viruses or divergent viralstrains. PCR is also prone to false-positive resultswherean appropriate product is detected, but on sequencingturns out tobe anartefact. Thiswas observed in 25%ofthe cases with PCR amplicons obtained from the sam-ples from these polar bears. This increased the work-load considerably as each false-positive band must besequenced before it can be discarded.

In principle, microarrays suffer less from the possi-bility of false-positive and -negative results since the

486 C.A. Szentiks et al.

longer oligonucleotides must be homologous over alonger stretch of nucleotides and, are therefore, lessprone to mismatches. However, the conditions for hy-bridization are not easily optimized in practice andsensitivity may be substantially reduced. This hasbeen observed when results from microarray analysesof enriched viral particles from multiple samples werecompared with those of next generation sequencing.In such a comparison, the latter method detected vi-ruses in 30% of the samples and the former in 8% ofsamples (Yozwiak et al., 2012). The problem is likelyto be exacerbated for solid tissue samples that cannotbe enriched for viral particles, as was the case for tis-sues from both bears.

Next-generation shotgun sequencing either fromenriched sources or directly from infected tissue cangenerate hundreds of millions to billions of individualsequence reads. The drawback of such sequencingwithout prior pathogen enrichment is that the major-ity of obtained sequences will be host derived (over99% for Knut and Jerka). In the case of Jerka, bothhigh-density microarray and next-generationsequencing-based methods failed to identify the re-combinant EHV-1 eventually obtained with thehelp of PCR (Greenwood et al., 2012). None of thethree methods identified candidate pathogens inKnut, but interestingly, a divergent polar bear-derived betaretroviruses were identified in bothKnut and Jerka (Mayer et al., 2013), suggesting thatthe methods are able to identify previously unknownviral sequences. In combination, the three methodsclearly are more powerful than any single methodalone. In the case of Jerka, PCR, microarray andnext-generation shotgun sequencing combinedtogether identified the recombinant EHV-1 virusthat was the likely cause of death, while excludingother potential viral, bacterial and parasite pathogens(Greenwood et al., 2012). The uniform negative re-sults for a candidate causal pathogen from Knut sug-gest that the genome of the pathogen was most likelyno longer present. This may indicate either a ‘hit-and-run’ infection, post-mortem degradation of DNA/RNA or pathogen concentration below the sensitivitylevels of all the methods employed.

Serum neutralization is one of the gold standards ofpathogen diagnostics. However, it is limited by thefact that only a small number of viruses can be main-tained in culture and is not sensitive in all cases.ELISA can be effective, but can lack both sensitivityand specificity, yielding false-positive and -negativeresults. For viruses that induce haemagglutination,HI assays can serve as confirmatory tests for SNT orELISA results. However, as demonstrated in thisstudy, polar bear blood and serum may interferewith this assay, rendering its outcome inconclusive.

HI assays failed, regardless of viral strain (H1 andH5 tested) and Knut was serologically negative forall viruses that could be tested by SNT. It is unclearwhy HI assays failed in all of the polar bear samples,but the results again illustrate that there may be un-known properties of serum in specific wildlife species,which can interfere with assays that are standard fornon-wildlife species.

The results of the diagnostic pipeline presentedhere resulted in the identification of a detectable anti-body response against influenza A virus that wasrestricted to Knut and no other polar bear tested(n ¼ 6), although no confirmatory genomic evidencewas obtained. Influenza is known to instigate acascade of cytokine effects that may persist and causedeath after the virus has been cleared (Yamada, 1996;Jang et al., 2009). Therefore, infection of Knut withinfluenza virus(es) is not unrealistic given the broadhost range of influenza in general, which includesbears (Cretescu et al., 1970) and the common occur-rence of waterfowl, both migratory and resident, inzoos. Whether the infection was long ago, recent orcoincident with the encephalitis is unclear.

In summary, the present study has shown that thediagnostic tools currently available are powerful, butalso that there are still major technical problems inwildlife disease diagnostics, which cannot necessarilybe overcome even if all existing techniques areapplied. Promising new methods include solution hy-bridization capture-based methods coupled withnext-generation sequencing, which allows for strongenrichment of specific target sequences, but withoutthe introduced biases of PCR (Maricic et al., 2010).Suchmethods have been highly effective for degradedsamples (e.g. ancient DNA) and have permitted theidentification of pathogens in samples that are hun-dreds of years old (Schuenemann et al., 2011). Howev-er, their diagnostic potential in wildlife diseaseremains to be investigated.

Acknowledgements

We gratefully acknowledge access to Knut and con-trol bear samples from Dr. B. Blaszkiewitz, Dr. A.Sch€ule and H. Kl€os (Zoological Garden Berlin) andprovision of samples from Jerka and Lars by Dr. A.Lawrenz and K. Gries (Zoological Garden Wupper-tal). Furthermore, we gratefully acknowledge theprovision of control bear samples fromDr. P. Krawin-kel and Dr. J. Kitzinger (Zoological Garden Gelsen-kirchen), Dr. H. Will (Zoological GardenNuremberg) andDr. T. Knauf-Witzens and A. Kren-gel (Wilhelma Zoological Botanical Garden Stutt-gart). The authors are grateful to M. Biering, Dr. F.G€oritz, P. Gr€unig, Prof. Dr. T. B. Hildebrandt, N.

Polar Bear Encephalitis 487

Jahn, D. Krumnow, M. Lange, F. Leuser, N.Nguyen, M. Reimann, I. Sohm, D. Viertel, Dr. G.Wibbelt and J. Ziep for their assistance during nec-ropsy examination and laboratory work. The authorsthank Dr. K. M€uhldorfer for assistance during theearly phases of the analysis. The authors wish tospecially thank all the animal keepers of the BerlinZoological Garden for their cooperation and support.The first five authors contributed equally to this work.

Conflict of Interest Statement

None of the authors has any financial or personal re-lationships that could inappropriately influence orbias the content of this paper.

Supplementary data

Supplementary data related to this article can befound at http://dx.doi.org/10.1016/j.jcpa.2013.12.005.

References

Aguirre AA, Ostfeld R, Daszak P (2012) New Directions in

Conservation Medicine: Applied Cases of Ecological Health.Oxford University Press, New York.

Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Zet al. (1997) Gapped BLAST and PSI-BLAST: a newgeneration of protein database search programs. Nucleic

Acids Research, 25, 3389e3402.Blaszkiewitz B (2009) Bear enclosures. The Guide to the

Zoological Garden Berlin, Zoologischer Garten-Berlin, Ber-lin, pp. 76e78.

B€ock P (1989)RomeisMikroskopische F€arbetechnik. Urban andSchwarzenberg, M€unchen.

Cheadle C, Vawter MP, Freed WJ, Becker KG (2003)Analysis of microarray data using Z score transforma-tion. Journal of Molecular Diagnostics, 5, 73e81.

Chiu CY, Rouskin S, Koshy A, Urisman A, Fischer K et al.(2006) Microarray detection of human parainfluenzavirus 4 infection associated with respiratory failure inan immunocompetent adult. Clinical Infectious Diseases,43, E71eE76.

Cretescu L, Zilisteanu E, Niculescu L, Wagner G, CociuMet al. (1970) Influenza caused by a virus resembling A2/Hong Kong among bears in Bucharest Zoo. II. Proper-ties of the virus. Verhandlungsbericht des XII Symposium €uberdie Erkrankungen der Zootiere, Budapest, pp. 197e199.

de Castro F, Bolker B (2005) Mechanisms of disease-induced extinction. Ecology Letters, 8, 117e126.

Eisen MB, Spellman PT, Brown PO, Botstein D (1998)Cluster analysis and display of genome-wide expressionpatterns. Proceedings of the National Academy of Sciences of

the United States of America, 95, 14863e14868.European Commission (2005) Verordnung (EG) Nr. 2075/

2005 der Kommission vom 5. Dezember 2005. Amtsblatt derEurop€aischen Union, pp. 60e82, 22.12.2005. L338.

Greenwood AD, Tsangaras K, Ho SYW, Szentiks CA,Nikolin VM et al. (2012) A potentially fatal mix of her-pes in zoos. Current Biology, 22, 1727e1731.

Gurrala R, Dastjerdi A, Johnson N, Nunez-Garcia J,Grierson S et al. (2009) Development of a DNAmicroar-ray for simultaneous detection and genotyping of lyssa-viruses. Virus Research, 144, 202e208.

Gwalter RH, Wegmann-Ehrenspenger D (2006) Labora-toriums-untersuchungen. In: Praktikum der Hundeklinik,10th Edit., PF Suter, B Kohn, Eds., Parey, Stuttgart,pp. 95e115.

Gylstorff B (1987) Trichinenuntersuchungen bei Wildtie-ren in S€udw€urtemberg mit dem Magnetr€uhrverfahren.Berliner und M€unchener Tier€arztliche Wochenschrift, 100,114e119.

Hofer H, East M (2012) Stress and immunosuppression asfactors in the decline and extinction of wildlife popula-tions: concepts, evidence and challenges. In: New Direc-

tions in Conservation Medicine: Applied Cases of Ecological

Health, AA Aguirre, RS Ostfeld, P Daszak, Eds., OxfordUniversity Press, New York, pp. 82e107.

Jang H, Boltz DA, Webster RG, Smeyne RJ (2009) Viralparkinsonism. Biochimica et Biophysica Acta, 1792,714e721.

Kistler AL,GanczA,Clubb S, Skewes-CoxP, FischerK et al.(2008) Recovery of divergent avian bornaviruses fromcases of proventricular dilatation disease: identificationof a candidate etiologic agent. Virology Journal, 5, 88.

Li B, Zhang G, Willersleve E, Wang J, Wang J (2011)Genomic data from the polar bear (Ursus maritimus).Giga Science. http://dx.doi.org/10.5524/100008.

MacPhee RDE, Greenwood AD (2013) Infectious disease,endangerment, and extinction. International Journal of

Evolutionary Biology, 2013, 571939.Maricic T, Whitten M, Paabo S (2010) Multiplexed DNA

sequence capture of mitochondrial genomes using PCRproducts. PLoS One, 5(11), e14004.

Mayer J, Tsangaras K, Heeger F, Avila-Arcos M,Stenglein MD et al. (2013) A novel endogenous betare-trovirus group characterized from polar bears (Ursus

maritimus) and giant pandas (Ailuropoda melanoleuca).Virology, 443, 1e10.

Nicholls JM, Wong LP, Chan RW, Poon LL, So LK et al.(2012) Detection of highly pathogenic influenza andpandemic influenza virus in formalin fixed tissues byimmunohistochemical methods. Journal of Virological

Methods, 179, 409e413.OIE. (2012a)Avian influenza.Manual of Diagnostic Tests and

Vaccines for Terrestrial Animals, OIE, Paris, pp. 436e452.OIE. (2012b) Equine rhinopneumonitis. Manual of Diag-

nostic Tests and Vaccines for Terrestrial Animals, OIE, Paris,pp. 889e898.

OIE. (2012c) Equine arteritis virus. Manual of Diagnostic

Tests and Vaccines for Terrestrial Animals, OIE, Paris, pp.899e912.

R Development Core Team (2009) R: A Language and Envi-

ronment for Statistical Computing. R Foundation for Statis-tical Computing, Vienna, ISBN 3-900051-07-0. URL:http://www.R-project.org/.

488 C.A. Szentiks et al.

Ritchie ME, Silver J, Oshlack A, Holmes M, Diyagama Det al. (2007) A comparison of background correctionmethods for two-colour microarrays. Bioinformatics, 23,2700e2707.

Schrenzel MD, Tucker TA, Donovan TA, Busch MD,Wise AG et al. (2008) New hosts for equine herpesvirus9. Emerging Infectious Diseases, 14, 1616e1619.

Schuenemann VJ, Bos K, DeWitte S, Schmedes S,Jamieson J et al. (2011) Targeted enrichment of ancientpathogens yielding the pPCP1 plasmid of Yersinia pestisfrom victims of the BlackDeath. Proceedings of the National

Academy of Sciences of the United States of America, 108,E746eE752.

Schuster GS (1999) Oral flora and pathogenic organisms.Infectious Disease Clinics of North America, 13, 757e774.

Smyth GK (2005) Limma: linear models for microarraydata. In: Bioinformatics and Computational Biology SolutionsUsing R and Bioconductor, R Gentleman, V Carey,S Dudoit, R Irizarry, W Huber, Eds., Springer, NewYork, pp. 397e420.

Smyth GK, Michaud J, Scott HS (2005) Use of within-array replicate spots for assessing differential expressionin microarray experiments. Bioinformatics, 21,2067e2075.

SmythGK, SpeedT (2003) Normalization of cDNAmicro-array data. Methods, 31, 265e273.

Stenglein MD, Velazquez E, Greenacre C, Wilkes RP,Ruby JG et al. (2012) Complete genome sequence ofan astrovirus identified in a domestic rabbit (Oryctolaguscuniculus) with gastroenteritis. Virology Journal, 9, 216.

Stevens A, Francis RJ (1996) Micro-organisms. In: Theory

and Practice of Histological Techniques, JD Bancroft,A Stevens, Eds., Churchill Livingstone, New York, pp.291e308.

Uchida K, Yoshino T, Yamaguchi R, Tateyama S,Kimoto Y et al. (1995) Senile plaques and other senilechanges in the brain of an aged American black bear.Veterinary Pathology, 32, 412e414.

Urisman A, Fischer KF, Chiu CY, Kistler AL, Beck S et al.(2005) E-Predict: a computational strategy for speciesidentification based on observed DNA microarray hy-bridization patterns. Genome Biology, 6, R78.

de Vries P, Uytdehaag GCMF, Osterhaus DMEA (1988)Canine distemper virus (CDV) immune-stimulatingcomplexes (iscoms), but not measles virus iscoms, pro-tect dogs against CDV infection. Journal of General

Virology, 69, 2071e2083.Wang D, Coscoy L, Zylberberg M, Avila PC, Boushey HA

et al. (2002)Microarray-based detection and genotypingof viral pathogens. Proceedings of the National Academy of

Sciences of the United States of America, 99, 15687e15692.Welch TA (1984) A technique for high-performance data

compression. Computer, 17, 8e19.Wettenhall JM, Smyth GK (2004) Limma GUI: a graph-

ical user interface for linear modeling of microarraydata. Bioinformatics, 20, 3705e3706.

Wohlsein P, Lehmbecker A, Spitzbarth I, Algermissen D,Baumg€artner W et al. (2011) Fatal epizootic equineherpesvirus 1 infections in new and unnatural hosts. Vet-erinary Microbiology, 149, 456e460.

Wyatt KB, Campos PF, Gilbert MT, Kolokotronis SO,Hynes WH et al. (2008) Historical mammal extinctionon Christmas Island (Indian Ocean) correlates withintroduced infectious disease. PLoS One, 3, e3602.

Yamada T (1996) Viral etiology of Parkinson’s disease:focus on influenza A virus. Parkinsonism and Related Disor-

ders, 2, 113e121.Yozwiak NL, Skewes-Cox P, Stenglein MD, Balmaseda A,

Harris E et al. (2012) Virus identification in unknowntropical febrile illness cases using deep sequencing.PLoS Neglected Tropical Diseases, 6, e1485.

Zachary JF (2007) Nervous system. In: Pathologic Basis ofVeterinary Disease, 4th Edit., MD McGavin,JF Zachary, Eds., Mosby Elsevier, St Louis, pp.915e916.

Zerbino DR, Birney E (2008) Velvet: algorithms for denovo short read assembly using de Bruijn graphs. GenomeResearch, 18, 821e829.

½ R

A

eceived, September 13th, 2013

ccepted, December 10th, 2013

�