to different strains of toxic dino ﬂ agellates, Alexandrium spp. , Distinctly different behavioral responses of a copepod, Temoralongicornis Harmful Algae

(1)

Distinctly different behavioral responses of a copepod, Temora

longicornis, to different strains of toxic dino ﬂ agellates, Alexandrium spp.

Jiayi Xu

^a,b,

*, Per Juel Hansen

^c

, Lasse Tor Nielsen

^a

, Bernd Krock

^d

, Urban Tillmann

^d

, Thomas Kiørboe

^a

aCentreforOceanLife,NationalInstituteforAquaticResources,TechnicalUniversityofDenmark,2920Charlottenlund,Denmark

bKeyandOpenLaboratoryofMarineandEstuaryFisheries,MinistryofAgricultureofChina,EastChinaSeaFisheriesResearchInstitute,ChineseAcademyof FisheriesSciences,200090Shanghai,China

cMarineBiologicalSection,UniversityofCopenhagen,3000Helsingør,Denmark

dAlfredWegenerInstitut-HelmholtzZentrumfürPolar-undMeeresforschung,27570Bremerhaven,Germany

ARTICLE INFO

Articlehistory:

Received19August2016

Receivedinrevisedform29November2016 Accepted29November2016

Availableonlinexxx

Keywords:

Feedingbehavior Paralyticshellﬁshtoxins GoniodominA Lyticactivity

ABSTRACT

Zooplanktonresponsestotoxicalgaearehighlyvariable,eventowardstaxonomicallycloselyrelated speciesordifferentstrainsofthesamespecies.Here,theindividuallevelfeedingbehaviorofacopepod, Temora longicornis, was examined which offered 4 similarly sized strains of toxic dinoflagellate Alexandrium spp.andanon-toxiccontrolstrainofthedinoflagellateProtoceratiumreticulatum.The strainsvariedintheircellulartoxinconcentrationandcompositionandinlyticactivity.High-speedvideo observationsrevealedfourdistinctlydifferentstrain-specificfeedingresponsesofthecopepodduring4h incubations:(i)the‘normal’feedingbehavior,inwhichthefeedingappendageswerebeatingalmost constantlytoproduceafeedingcurrentandmost(90%)ofthecapturedalgaewereingested;(ii)the beatingactivityofthefeedingappendageswasreducedbyca.80%duringtheinitial60minofexposure, afterwhichveryfewalgaewerecapturedandingested;(iii)captureandingestionratesremainedhigh, but ingested cellswere regurgitated;and (iv)thecopepod continued beatingitsappendages and capturedcellsatahighrate,butafter60min,mostcapturedcellswererejected. Thevariousprey aversionresponsesobservedmayhaveverydifferentimplicationstothepreyandtheirabilitytoform blooms:consumedbutregurgitatedcellsaredead,capturedbutrejectedcellssurviveandmaygivethe preyacompetitiveadvantage,whilereducedfeedingactivityofthegrazermaybeequallybeneficialto thepreyanditscompetitors.Thesebehaviorswerenotrelatedtolyticactivityoroverallparalytic shellfishtoxins(PSTs)contentandcompositionandsuggestthatothercuesareresponsibleforthe responses.

1.Introduction

Zooplanktonplaysacrucialroleinmarinefoodwebs,bothby channeling primary production tohigher trophic levelsand by controllingphytoplanktonpopulations.Algalbloomsoccurwhen algal growth exceeds zooplankton grazing. Thus, harmful algal bloomsarethoughttobefacilitatedbyreducedgrazingduetothe algaeproducingtoxicsubstances(Jonssonetal.,2009)that,inturn, are believed mainly to function as grazer deterrents. Reported grazerresponsestoharmfulalgaearediverse.Theresponsesof

copepods,forexample,totoxicalgaemayvarywithinandbetween speciesofboththegrazersandalgae,andresponsesrangefrom unaffectedto substantial(Turner,2014).Evendifferentpopula- tionsofthesamecopepodspeciesmayshowdifferentresponsesto thesamestrainofa toxicalgadue toacclimationoradaptation (Colin and Dam, 2002; Engström-Öst et al., 2002; Kozlowsky- Suzukietal.,2003).Afurthercomplicatingfactoristhatdifferent strainsandnaturalpopulationsofthesamealgalspeciesmayvary intheirtoxicityandwithitsgrowthconditions(Burkholderand Glibert,2006;Cembella,1998).Withafewexceptions(e.g.,Hong et al., 2012), only macroscopic responses (e.g. mortality and feedingrate)ratherthanbehavioralresponsesareexamined,and in most cases it is not possible to establish a mechanistic relationship between the algal toxin proﬁle and its effects on thecopepodgrazer.

* Correspondingauthorat:CentreforOceanLife,NationalInstituteforAquatic Resources,TechnicalUniversityofDenmark,2920Charlottenlund,Denmark.

E-mailaddress:sjxu@aqua.dtu.dk(J.Xu).

ContentslistsavailableatScienceDirect

Harmful Algae

j o u r n a lh o m e p ag e :w w w . e l s e vi e r . c o m / l o c a t e / h al

(2)

ThegenusAlexandriumisfoundworldwideandisoneofthe most studied toxic dinoflagellates (Anderson et al., 2012). It includes33describedspecies,ofwhich11areknowntoproduce paralytic shellfish toxins (PST) (Moestrup et al., 2009). The chemical structuresof this group of toxins,including saxitoxin (STX)andapproximately57derivatives,arewelldescribedfrom thegenusand fromseafood (Munday,2014).Paralytic shellfish toxins are sodium-ion channel blockers that can cause potent neurotoxic syndromes in humans as well as fish, seabirds and marine mammals (Cembella, 1998; Turner and Tester, 1997;

Turner,2014).ReportedeffectsoncopepodsofferedPST-contain- ingAlexandriumspp.,however,rangefromnonetoadverseeffects on ingestion rate, egg production, egg hatching and offspring development duration (Dutz, 1998; Frangopulos et al., 2000;

Guisandeetal.,2002).Thesevariationsinresponsesarenotrelated totheoveralltoxicityofthecells(Teegardenetal.,2008)andraises thequestionofwhetherornotthePSPtoxinsactuallyfunctionasa grazerdeterrents.Could othercompoundsproduced byAlexan- driumspp.beresponsiblefortheobservedeffectsoncopepods?

In fact,anumberofdifferenttoxinshavebeenfoundamong Alexandriumspp.inadditiontothePSTs,makinginterpretationsof past reports difﬁcult: spiroimines (spirolides, gymnodimines), goniodominAandlyticcompounds.Thespiroimines arepotent fast-actingneurotoxinsthat havesofaronlybeenfoundinthe European and North Atlantic A. ostenfeldii but not Baltic A.

ostenfeldii(Krempetal.,2014;Sopanenetal.,2011).Goniodomin Aisalsoaneurotoxinthathasbeenreportedtoaffectvertebrates (Kleinetal.,2010)aswellasinvertebrates(Murakamietal.,1988).

ThistoxinhasonlybeenreportedforA.hiranoi,A.monilatumandA.

pseudogonyaulax(Hsiaetal.,2006;Murakamietal.,1998,1988) andthesespeciesdonotproducePSTs.Finally,manyAlexandrium speciesandstrainsalsohavetheabilitytoproduceextracellular allelochemical compounds, which are still poorly examined chemically(Ma et al.,2009).These extracellular allelochemical compoundshavebeen demonstratedtoaffectprotistan grazers (Legrand et al., 2003; Tillmann and John, 2002), competitors (Granéliand Hansen, 2006;Legrand etal., 2003;Tillmann and Hansen,2009),orparalyzepreycells(Blossometal.,2012),while effectsonmetazoangrazersarestillunknown.Thus,studiesusing experimentalandcontrolAlexandriumstrainscharacterizedasPST and non-PST strains might be misleading, as they may differ

substantially in the presence/absence of other toxins/bioactive compounds.

Here,authorsexaminedtheinitialbehavioralresponseofthe copepodTemoralongicornisto3differentstrainsofAlexandrium tamarense,asinglestrainofA.pseudogonyaulaxandtoastrainof Protoceratiumreticulatumthatcontainsnoknowntoxins.Species andstrainswereselectedduetotheirsimilarsizeandshapebut differenttoxincontentandproﬁle(PSTs,lyticactivityofthecells, Goniodomin A). Direct, high-speed video was used to describe feeding behaviors (activity, prey capture, rejection, ingestion, regurgitation).Awide prey-speciﬁcbehavioral repertoireof the copepodsweredemonstratedthatleadtoavariationofingestion rateandwithdistinctlydifferentimplicationstothepreyandtheir abilitytoformblooms.Thebehavioralresponsewasunrelatedto thecomposition orcontentthecompounds analyzedfor theA.

tamarensestrains,suggestingthatothercompoundsmaytrigger theavoidancebehaviorobservedtowardssomeoftheprey.

2.Materialsandmethods 2.1.Algalcultures

A strain of Protoceratiumreticulatum and 4 clonal strainsof Alexandrium spp. wereused in the experiments(Table 1). The cultureofP.reticulatumCCMP1889obtainedfromNationalCenter for Marine Algae and Microbiota, A. pseudogonyaulax CAWD138 obtained from Cawthron Institute, and A. tamarense Alex2,A.tamarenseAlex5,andA.tamarenseAlexH5obtainedfrom AlfredWegenerInstitute.Thedifferentalgaewereofsimilarsize butvariedintheirtoxinproﬁles(Tables1and2).Algalcultures were maintained on B1 medium prepared with pasteurized, ﬁlteredseawaterat16Candasalinityof32.Thecultureswere exposedtoanirradianceof150

m

^mol^photons^m ²^s ¹^on^a¹²^h:

12hlight:darkcycle.Allphytoplanktonusedintheexperiments wereinexponentialgrowth.

2.2.Toxinanalyses

Paralyticshellﬁsh toxins and lyticactivity of the cells were quantiﬁed. Meanwhile, the presence of goniodomin A, and yessotoxins (YTX) weretested. For cell content analyses,10 to

Table1

ListofthealgaeusedaspreyspeciesforTemoralongicornisinvideoobservations,includingthestrainnumber,theisolationlocation,andtheequivalentsphericaldiameter (ESD).

Algae Strain Origin ESDSD(m^m) ^Reference

Protoceratiumreticulatum CCMP1889 FridayHarbor,USA 32.02.3 (Howardetal.,2009)

Alexandriumtamarense Alex2 NorthSeaoffScotland 31.32.5 (TillmannandHansen,2009;Tillmannet al.,2009) Alex5 NorthSeaoffScotland 33.80.5 (TillmannandHansen,2009;Tillmannetal.,2009) AlexH5 GulfofSanJorge,Argentina 31.60.7 (Krocketal.,2015)

Alexandriumpseudogonyaulax CAWD138 Kerikeri,NewZealand 33.80.9

Table2

Toxinprofilesandcontentsofthealgae.PST=paralyticshellfishtoxins;YTX=Yessotoxin;GA=GoniodominA;LA=Lyticactivity.+=toxindetectedbutnotquantified;–=not detected.

Algae Strain PSTs(fmolcell ¹) CellToxicity

(pgSTXeqcell ¹)

YTX GA LALC50

(cellsml ¹) C1/C2 GTX1/4 dcGTX2/3 GTX2/3 NEO STX total

Protoceratiumreticulatum CCMP1889 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 – – –

Alexandriumtamarense Alex2 4.7 0.0 0.0 0.4 4.2 2.1 11.4 2.3 – – 511

Alex5 43.0 40.6 2.3 3.9 27.5 10.8 128.1 29.1 – – –

AlexH5 119.5 40.1 0.1 2.9 8.5 0.0 171.0 22.9 – – 544

Alexandriumpseudogonyaulax CAWD138 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 – + –

(3)

20mlofexponentiallygrowingAlexandriumspp.andP.reticulatum cells(around2000cellsml ¹)werecentrifuged(2150g,15min).

After removing most of the supernatant, the algae were re-suspended in 1ml B1 medium and transferred to a micro- centrifugetube.Sampleswerecentrifugedagainat3200gfor 15min.Allthesupernatantwasremoved.Thedrycellpelletswere keptat 20C.Bothcellconcentrationsoftheinitialalgalculture and the supernatant were enumerated to calculate the exact numberofcellsinthepelletsfortoxinanalysis.

Paralyticshellﬁshpoisoningtoxins(PSTs)wereextractedwith 500

m

^l^0.03^mMâceticâcid^byultrasonication(sonotrodeHD2070, Bandelin,Berlin,Germany;1min,cycletime50%,10%power).The sampleswerecentrifugedat16,100xgandsupernatantsfiltered overcentrifugationfilters(pore-size0.45mm,MilliporeUltrafree, Eschborn,Germany)at1,500xgfor30s.filtratesweretransferred toautosamplervialsandmeasuredbyion-pairliquidchromatog- raphy coupled to post-column derivatization and fluorescence detectionasdescribedindetailinSuikkanenetal.(2013).Inorder to make the data comparable to other literature values, the combined cell toxicity was calculated as saxitoxin equivalents (STXeq) by multiplying toxin concentration values from HPLC chromatograms by toxin-specific toxicity equivalency factors (TEFs, Alexanderet al.,2009)Since dinoflagellates arebelieved to exclusively produce the betamers of enantiomeric pairs (Cembella,1998)andthecorrespondingalphamersareregarded asextraction artifacts,only TEFsof thebetamerswereused to calculatetotaltoxicityasSTXequivalents.

Lipophilic toxins (YTX and goniodomin A) were extracted with 300

m

^l ^methanol ând ânalyzed ^for ^YTX âs ^described ⁱⁿ

detailinSala-Pérezetal.(2016).Alllipophilicextractswerealso screenedfortwo pseudomolecularionsofgonodomin Ainthe positivemode:m/z786([M+NH4]⁺)andm/z791([M+Na]⁺)that were reported by Hsia et al. (2006). For the positive samples product ion spectra of both pseudomolecular ions were also recorded.

To quantify allelochemical (lytic) activity, the method of Blossom et al. (2014) was used, in which the concentrationof dinoflagellate cells that cause 50% mortality of target cells (Teleaulaxacuta)isdetermined(Table2and AppendixFig.1).A targetcellconcentrationof3250cellsml ¹andrelativefluores- cencewereusedtoquantifytargetcellconcentration.Vialswith targetcellsand 10to15differentconcentrationsofsupernatant obtainedaftercentrifugation(2150g,15min) ofdinoflagellate cultureswereplacedat16C inthedarkfor 3h,atwhich time target cell survival was quantified flourometrically (TD-700 Fluorometer,TurnerDesigns,SanJose,California,US).

2.3.Copepodfeedingbehavior

A feeding-currentfeedingcopepod, Temora longicornis, were isolated from the Øresund, Denmark, and used to establish a continuouscultureat16C,salinity32.Theculturewasfedamixed phytoplankton diet including Akashiwo sanguinea, Heterocapsa triquetra, Prorocentrum minimum, Thalassiosira weissﬂogii, and Rhodomonassalina.

Adultfemalesusedforvideoexperimentsweretetheredtothe dorsalsurfacewithashortlengthofhumanhairusingasmalldrop ofsuperglue(CowlesandStrickler,1983)andplacedovernightin filteredseawaterinadark,thermo-constant room(16C).The subsequentmorning,theotherendofthehairwasattachedtoa micromanipulatorandthecopepodwasplaced ina101010 cm³ transparent container filled with filtered sea water in a thermo-constantroom.Thetetheredcopepodsmayliveformany daysandappearunaffectedbythetether.

Phytoplanktonwasadded attime0,and thebehaviorof the copepod recorded during the subsequent 4h. Copepods were

offeredoneofthefourstrainsofAlexandriumspp.orP.reticulatum at one of 3 different concentrations (40, 80 and 200 cells ml ¹;<10%).Three individual copepods weretested foreach strain/concentrationtreatment,totaling 9individualsperstrain.

Samples(3ml)for algalenumerationwereremovedduringthe beginning, middle and end of filming to check the prey concentrations. Also, the water was gently stirred throughout the experiment to prevent sedimentation of the algae. The tetheredcopepodwas filmedwitha PhantomV210highspeed camerausinginfraredilluminationshinedthroughtheaquarium towardsthecamera.ThecamerawasequippedwithNikonlenses to yield a field of view of approximately 2.51.6mm² (varied slightly between experiments). Both high speed (resolution:

1280800pixels;framerate:2200Hz)andlowspeed(resolution:

720576pixels;framerate:25Hz)videosweresaved simulta- neously fromthecamera.The lowspeed videowas settosave automaticallytheﬁrst30minandthenfor10minevery¹/2hourto describe feeding activity and prey interactions. Several 2.5s sequencesofhighspeedrecordingsweresavedthroughtheentire experimentaldurationtoquantifyappendagebeatfrequenciesand describepreyresponsebehaviors.

ThefeedingcurrentofT.longicornisiscreatedbytheregular beatingofthesecondantenna(A2)andthemaxillipeds(MXP)as wellasoftheotherfeedingappendages(Gonçalvesetal.,2014;

Paffenhöferetal.,1982;Tiseliusetal.,2013)(AppendixVideo1).

Whenapreyparticlewithinthefeedingcurrenttouchesthesetae on oneof thefeedingappendages and is detected, the regular beatingofthefeedingappendagesischangedtoguidetheprey particlenexttomandibles.Aneventasa‘capture’wasclassiﬁed when the prey particle was handled by the copepod (AppendixVideo1).Afterbeingcaptured,thepreywasgenerally handledforashortperiodandadjustedtoacertainpositionbefore either being swept into the mouth, an ‘ingestion’ event (Appendix Video 1), or being rejected (a ‘rejection’ event;

AppendixVideo2).Insomecases,allorpartsofapreyparticle were regurgitated afteringestion, which was recorded as both

‘ingestion’and‘regurgitation’events(AppendixVideo3).

The low speed recordings wereused to enumerate capture, ingestion,rejection,andregurgitationeventsandtoquantifythe fractionoftimetheanimalwasbeatingitsfeedingappendages.

Thefractionoftimebeatingwasestimatedbycountingthenumber offramesthatthecopepodwasbeatingitsappendagesduringthe last1minofevery10minsequence.

The high speed video was usedmainly to quantify thebeat frequencyof theappendages.Characteristicsequenceswerealso saved to illustrate the various types of feeding behaviors (AppendixVideo1–4).Preypositionsandbeatingfrequencieswere measuredusingImageJ(Version1.48;NationalInstitutesofHealth, USA)andPhantomCineViewer(Version2.6;VisionResearch).

Waterbornecuesfromcopepodgrazerscaninduceincreased PSTsproductioninAlexandriumspp.(Selanderetal.,2006),butthe fullinductiontakes2–4days(Selanderetal.,2012),andislowwith thelowconcentrationofcopepodsusedhere(1per800ml).Thus, itisassumedthatthechemicalproﬁleofcellsfromthecultureis representativefortheexperiments.

2.4.Statisticalanalysis

Differences in appendage beat frequency, fraction of time beating, capture rate, ingestion rate, and fraction of captured cells rejected between treatments were tested using two-way ANOVA with prey concentration and prey species as factors.

MeanvalueswerecomparedusingHolm-SidakTestandcarried out in SigmaPlot 13.0. Normality was tested according to Shapiro-Wilk.

(4)

3.Results

3.1.Algaltoxincontent

6 different species of PSTs wereidentiﬁed (Table 2). Strain Alex5containedmainlyC1/C2,GTX1/4,STXandNEO.AlexH5also hadhighcelltoxincontent,butmainlyC1/C2andGTX1/4;itlacked STX. Strain Alex2 had fewer PST derivatives and an order of magnitude lower cellular PST content. A compound with the molecular mass of goniodomin A was only detected in A. pseudogonyaulax, which did not have PSTs. The strain of P. reticulatum contained neither YTX nor other toxins (lyticcompounds, PSTsor goniodominA) abovedetectionlevel andhenceworkedasanon-toxiccontrol.

2strainsofA.tamarense(Alex2andAlexH5)bothproducedand excretedcompoundswithlyticeffectsonthetestorganismT.acuta (Table2).

3.2.Appendagebeatfrequency

ThecephalicappendagesofT.longicornisproduceacontinuous repetitivebeating.Theappendagebeatfrequencyvariedbetween

22 and 34Hz between the5 diets(Fig.1A). The variationwas independent of prey concentrations (P=0.905), and time (P=0.380; data in Appendix Fig. 2), but differed signiﬁcantly betweenprey(P<0.05).Thebeatfrequencywashighest(33Hz) whenfedonA.tamarenseAlexH5atallpreyconcentrations,while thebeatfrequenciesofcopepodsexposedtotheotherfourpreys weresimilartooneanotherandaveraged 26Hz. Time-resolved patternsinbeatfrequenciesaregiveninAppendixFig.2.

3.3.Fractionoftimethefeedingappendagesbeat

Initially all the copepods were using their appendages constantly.Mostofthemkeptbeatingatnear100%of thetime during all 4 observation hours (Appendix Video 1–3), except copepodsexposedtoA.tamarenseAlex5(AppendixVideo4).With thisprey,thebeatingactivityofthecopepodsdecreasedrapidly during theﬁrst hourto reachabout20% of thetime and then remainedatthatlevelduringtheremaining3h(Figs.1Band2A).

Thedeclinewasstatisticallysigniﬁcant(p<0.05)butindependent of prey concentration (P=0.222). Since several aspects of the feedingbehaviorchangedduringtheﬁrsthourbutsubsequently remained relatively stable, all the statistical analyses below

Fig.1.FeedingbehaviorsofTemoralongicornisfedonProtoceratiumreticulatum,Alexandriumtamarense,andA.pseudogonyaulaxatthreepreyconcentrations(40,80and 200cellsml ¹).N=3,errorbarsrepresentstandarddeviation.(A)Averagebeatingfrequency(Hz)ofT.longicornisduringallfourhoursofvideoexperiments.(B)Average fractionoftimewhenT.longicorniswasbeatingitsfeedingappendagesduringthelastthreehoursofvideoexperiments.(C)Averagepreycellsobservedtobecapturedper copepodperhourduringthelastthreehoursofvideoexperiments.(D)AveragefractionofpreycellsthatwasrejectedafterbeingcapturedbyT.longicornisduringthelast threehoursofvideoexperiments.(E)Averagepreycellsobservedtobeingestedpercopepodperhourduringthelastthreehoursofvideoexperiments.(F)Averagefractionof preycellsthatwasregurgitatedafterbeingingestedbyT.longicornisduringthelastthreehoursofvideoexperiments.

(5)

consider only the last 3h of each experiment. Time resolved patterns in appendage activity are shown for all prey in AppendixFig.3.

3.4.Capturerate

Preycapturerateincreasedwithincreasingpreyconcentration forallpreytypes(Fig.1C,timeresolvedinAppendixFig.4)and differed signiﬁcantly between prey species and concentrations (P<0.05).Withthesamepreyconcentration,A.pseudogonyaulax werecapturedatthehighestrate andA.tamarenseAlex5atthe lowestrate.Theotherstrainswerecapturedatintermediateand similarrates.

3.5.Rejection

Captured prey may be ingested or rejected. Initially, all copepodsrejectedonlyasmallfraction(20%)ofcapturedcells.

After60min,theproportionofrejectedA.pseudogonyaulaxcells increasedto80%andremainedatthis leveltilltheend ofthe observation period (Figs.1D and 2B, Appendix Video 2, time- resolved pattern in Appendix Fig.5). Withthe otherfour prey strains the fraction of rejected cells remained stable and low (AppendixFig.5).Thus,notincludingthefirsthour,thefractionof rejected cells was significantly higher with A.pseudogonyaulax (0.90.1) compared to the four other prey strains (0.20.1) (p<0.05),whilepreyrejectionwasindependentofpreyconcen- trationforallfiveprey(P=0.152).

3.6.Ingestionrate

Theingestionrateofpreyistheproductofcapturerateandthe fractionof accepted(i.e., not rejected)cells. The ingestion rate increasedwiththeincreasingofpreyconcentrationwithallprey (p<0.05),andtherewerealsosigniﬁcantdifferencesbetweenprey

strains(P<0.05)(Fig.1E).Non-toxiccontrol,P.reticulatum,cells wereconsistentlyingestedatthehighestrate(8060cellsh ¹at 200cells ml ¹),andA.pseudogonyaulaxandA.tamarenseAlex5 (17.17.3and23.18.0cellh ¹at200cellsml ¹)atthelowest rates, with the other strains in between (time resolved in AppendixFig.6).

3.7.Regurgitation

Someingestedcellswererapidly(within1s)regurgitated.This was in particular evident with A. tamarense Alex2 as prey (AppendixVideo3).Theproportionofcellsregurgitatedincreased withincreasingA.tamarenseAlex2concentrationtomorethan30%

atthehighestconcentration(Fig.1FandAppendixFig.7),butwas independentoftime(P=0.670).Asmallproportion(2%and13%)of ingested cells wasalsoregurgitatedwhen copepodswerefeda highconcentrationofA.tamarenseAlexH5andA.pseudogonyaulax, respectively.

3.8.IngestionofPSTsbythecopepods

Basedontheingestionandregurgitationofphytoplanktoncells by copepods (Fig.1) and the toxin content of each algal prey (Table2),thetotalingestionofPSTs(inSTXequivalents)andSTXon the three diets of A. tamarense were calculated (Fig. 3 and Appendix Fig.8 –10). The cumulated amount of ingested PSTs increasedovertimeandwithcellconcentration,andwashighest in copepods offered AlexH5, and lowest when offered Alex2 (Fig.3A).DuetoalackofSTXinAlexH5,Thehighestaccumulation ofSTXwasincopepodsofferedAlex5andfollowedbycopepods offeredAlex2(Fig.3B).

4.Discussion

4.1.Repertoireofcopepodfeedingbehaviorsandimplicationstoprey populations

4distinctlydifferentbehavioralresponsesofthecopepodwere observedtovariouspreycells,viz.:(i)normalfeedingbehavior– thefeedingappendagesarebeatingmoreorlessconstantlyand mostcapturedcellsareingested(controlalga P.reticulatumand AlexH5);(ii)thecopepodsigniﬁcantlyreducesthefractionoftime it isbeatingitsappendagesin thecourseoftheﬁrsthourafter introducing prey cells and beating activity then remains low;

capturedcellsarehowevermainlyingested,althoughatalowrate (Alex5);(iii)appendage beatactivity remainshighand cellsare capturedandingested ata highrate, butalargefractionofthe ingested materialis subsequentlyregurgitated(Alex2); and(iv) feeding activity and prey capture rate remains high, but an increasingfractionofcapturedcellsarerejectedduringthefirst hour,andrejectionrateremainshighduringtheremainderofthe observationperiod(A.pseudogonyaulax).Mostpreviousstudiesof theresponseofcopepodstotoxicalgaeareincubationstudies,in which the net outcome of the copepod-prey interaction is quantified in terms of feeding rate, prey selection, growth or eggproductionrate,orothersimilarbulkmeasures(reviewedby Turner,2014).Thedirectvideoobservationofindividualresponses andofdirectcopepod-preycellinteractionsprovidedbythisanda fewotherstudies(Brunoetal.,2012;Hongetal.,2012;Tiselius etal.,2013)areinnovativeandallowustodisentanglethepossible mechanisms underlying the diverse outcome of ‘black box’ incubation experiments and to better evaluate their ecological significance.

Several studies have reported that copepods may select between toxic and non-toxic cells in a prey mixture (DeMott andMoxter,1991;Huntleyetal.,1986;SchultzandKiørboe,2009;

Fig.2.TemporalbehavioralvariationofTemoralongicornisfedAlexandriumspp.

duringthefourhoursvideoexperiments.(A)ThefractionoftimebeatingofT.

longicornis fed on A. tamarense (Alex5) decreased independently of prey concentration.(B)ThefractionofrejectedofT.longicornisfedonA.pseudogonyaulax increasedindependentlyofpreyconcentration.

(6)

Selanderetal.,2006;Teegarden,1999).Thefactthatthecopepods can distinguish between cells of very similar size and shape suggests that selection is mediated by chemical information.

SchultzandKiørboe(2009)suggestedthatcopepodspossessthe abilitytoremotelydiscriminatenon-toxicandtoxicalgaebefore capture. Recently however theability of copepods to remotely detectphytoplanktonbasedontheirchemicalcharacteristicshas beenquestioned(Gonçalvesand Kiørboe,2015), and ourobser- vations suggest that prey selection is based on post-capture discriminationand thatunwantedcellsarerejectedfollowing a handling time. Vanderploeg et al. (1990) reported a similar observationinafreshwatercopepod.Thus,preyselectionappears tobebasedongustation(taste)ratherthanolfaction(smell).

Multiple studies have reported reduced feeding rates on toxiccomparedtosimilarlysizedandshapednon-toxicalgaein single-prey experiments (Turner, 2014), and our observations suggesttwo possible mechanismsbehind sucha response,i.e., rejectionofcapturedcellsbeforeingestionandreducedfeeding activity. An increased cell rejection was observed only with A. pseudogonyaulax as prey, and the increasing rejection rate duringtheﬁrsthour,suggestingthatthecopepodwouldneedto learn that these cells are un-wanted. The reduced feeding activity response (reduced appendage beat activity) was only observed with A. tamarense Alex5 as food. Since the lytic compoundsdetectedfromAlexH5hadnoeffectonthefeeding behaviorofT.longicornis,theextracellularcompoundswerenot thetrigger.Thereduced feedingactivity onlymaterializesafter thecopepodhasingestedsomecells, and soislikely mediated by substances released during processing of food in the gut.

Subsequent tothis transitionperiod thecopepodkeepsbeating its feedingappendages intermittentlyand capturesand ingests preycells at a low rate, allowing thecopepod tocontinuously sampletheenvironmentand–presumably–topickupfeeding athighrateifthepreyenvironmentchanges.Asimilarbehavior isobservedinthecopepodAcartiatonsa.Thiscopepodmodiﬁes its appendage beat activity and feeding current production in response tothe concentrationand type (size,motility) of prey cellsintheenvironment(JonssonandTiselius,1990),allowingit toswitchbetweenfeedingcurrentfeedingandambushfeeding.

InthepresenceoftoxicKareniaspp.cellsitalso(within10min) reduces appendage beat activity to only sample the environ- ment,andresumesmoreactivefeedingifthepreyenvironment becomes favorable(Honget al.,2012).

Some studies have demonstrated reduced growth and egg productionrates(ColinandDam,2007;Dutz,1998;Guisandeetal., 2002;Roncallietal.,2016;Sopanenetal.,2011;Teegardenetal., 2008)or elevatedmortality (Averyet al.,2008; Sopanen et al., 2011) in copepods exposed totoxic algae compared to control algae. Such responses may be mediated by the behaviors considered above that both lead toreduced prey ingestion, or by the regurgitation of consumed algae, as described here for T. longicornis feeding on A. tamarense Alex2. From our video observations,mostoftheregurgitatedcellsweresmashedandsoit isimpossibletoquantifytheexactamountoffoodlostthrough regurgitation,butitmaybesignificant.SykesandHuntley(1987) reportedasimilarobservationofthecopepodCalanusfinmarchicus regurgitatingProtoceratiumreticulatum(=Gonyaulaxgrindleyi),and foundthat thecopepod was unabletofill itsgut, suggestinga Fig.3.ThecumulativeingestedPSTsinTemoralongicorniswhenfedAlexandriumtamarensestrains:Alex5(solidline),Alex2(dashedline),andAlexH5(dottedline).The cumulativetotalingestedPSTbycopepodsexposedtopreyconcentrationsof40cellsml¹(A-1),80cellsml ¹(A-2),and200cellsml ¹(A-3).Thecumulativeingested saxitoxin(STX)atpreyconcentrationsof40cellsml ¹(B-1),80cellsml ¹(B-2),and200cellsml ¹(B-3).Valuesaremeans(n=3).

(7)

signiﬁcantreductionin netfoodintake.While theregurgitation response observed by Sykes and Huntley only occurred 45–120minafterinitiationoffeeding,theresponsereportedhere isimmediateandspeciﬁctothecelljustconsumed.

Allthe preyaversion responsesobserved here, viz.reduced feedingactivity,rejectionofcapturedcells,and regurgitationof ingestedcells,mayallleadtoreducedenergyuptake,growth,and eggproductionofthecopepod.Theecologicaland evolutionary implications of theresponses differin several importantways.

First,apreycuethatleadstopreyrejection,thatis,atruefeeding deterrent,isbeneﬁcialtothealgalcellasitsurvivestheinteraction with the copepod, allows the copepod tocontinue to feed on competingbutpalatablecells,andmayleadtotheformationofa bloom.Itisalsoeasytoenvisagehowsuchfeedingdeterrentcan evolve as it gives the individual cell a competitive advantage.

Second, a cue that leads toreduced feeding activity, although beneficialtothecompoundsproducer,maybeequallybeneficialto its competitors, and cheaters that do not pay the price of compound production may flourish. It appears not to be an evolutionarystablestrategyandmaynotleadtotheformationofa bloom of the compound producer. Thirdly, an ingested but regurgitatedcellisnotbeneficialtotheindividualcell(itisdead), and although it mayreduce grazing due toreduced growthor survivalofthepredators,itdoessoonlyonitssiblingcellsand equallyonitscompetitors.

4.2.Potentialroleofcellulartoxinquantityandcompositionon copepodbehavioralresponses

It is unclear from our resultsand from data in the literature exactly whatelicitstheverydifferentbehavioral responsesin copepods exposedtothevariousstrainsofAlexandrium,exceptthatitmost likelyisachemicalcuecontainedinorreleasedbythecells.

Withrespect toPSTs, theirmodeof action in vertebratesis knowntobeabindingtovoltage-gatedsodiumchannelsinhibiting action potential, nerve transmission, and ultimately muscle contraction (Cusick and Sayler, 2013), and it thus could be expected that voltage-gated sodium channels of invertebrates arelikewiseaffected.Althoughanumberofinvertebratesretain and accumulatePSTs in theirtissues,many species includinga numberofbivalvemollusksare–contrarytopopularbelief–not immunetoPSTs(GainesandShumway,1988;KvitekandBeitler, 1991;Robineauetal.,1991).Paralyticshellfishtoxinsresistancein soft shell clams has been identified to be caused by a single mutation in the saxitoxin binding site in the sodium channel (Bricelj etal.,2005).Copepodshavealsobeendemonstratedto adapt to PSTs and become immune after some generations of exposure (Colin and Dam, 2007, 2004), but the mechanism of adaptationremains unknown.Like in Bivalves, sodiumchannel mutantshave beenidentifiedin the copepodA. hudsonica, but theseturnedout not toaccountfor achieved immunityin this copepodspecies(Finiguerraetal.,2015).

Considering the toxin composition in the PSTs containing strains used in our experiments, the most distinct response (intermsofingestionrates)isfoundwhenfedonAlex5(reduced beatingactivity)followedbyAlex2(regurgitation)andwithno responsetoAlexH5.Thispatternofbehavioralresponsescannot beeasilyexplainedbycellularcontentoringestionoftotalPSTs (inSTXeq)(Table2andFig.3A).AlexH5amongallPST-producing strainstestedhadthehighestcellulartoxincontent,bothastotal compounds per cell and when calculated as STX equivalents (Table 2), but failed to cause obvious copepod behavioral responses. The common conversion of all quantities of single congenerstooneestimateoftotaltoxicity(inSTXeq)isjustiﬁed byhowtoxictheyaretohumans,andthusToxicity-Equivalent- Factorsbasedonthestandardmousebioassay(Munday,2014)are

used.Itcanbeanreasonableassumptionthatrelativepotencyof each singlePST compoundmay varydramatically for different sodium channeltypes even for vertebrate cells (Alonso et al., 2016)anditisunknownwhetherandhowinvertebratesodium channels are differentially affected. Thus, copepod behavior might be related to single toxin compounds, and copepod behavior was indeed correlated with cell content and total ingestion of the STX molecule (Fig. 3B), suggesting that speciﬁcallySTXplaysanimportantrolefortheobservedcopepod behavioralchanges.Suchaview,however,islikelytoosimplistic, as all A.tamarense strainsalso had signiﬁcant amounts of the nearly identical molecule neosaxitoxin (NEO). While Alex5 (reducedbeatingactivity)alsohadthemostSTXandNEOwhen combined,AlexH5(noeffect)hadmorethanAlex2(regurgitation).

Sotheobservedbehaviorcannotberelatedtoestimatesoftotal STX+NEO ingestion(see Appendix Fig. 8–10). The inconclusive pattern of behavioral responsesin ourexperiments alignswell withtheresultsoftheincubationexperimentsof(Teegardenetal., 2008):thevery differentfeedingrates in4 differentspecies of copepodstoAlexandriumspp.strainsofdifferentPSTtoxicitythat theyobservedwereunrelatedtotheleveloftotaltoxicity(asSTX eqpercell)andonlyrelatedtowhetherornotthecellsweretoxic.

Thus,theresolutionofthefeedingresponseintothemorediverse behavioralresponsesandtheinformationonthecompositionof toxinsreportedheredonotappeartoprovideclearanswerstothe identity and nature of the cues that are responsible for the responses.

Thepresenceof lyticcompoundsalsodidnotcorrelatewith behavioralchangesofT.longicornis.AllelochemicalsfromAlexan- driumareassumedtoprimarilyactdestructivelyontheexternal plasma membrane and have been shown to have a high lysis potentialforsingleprotistancells(Tillmannetal.,2008).Authors herefortheﬁrsttimeprovideevidencethattheselyticcompounds, atleastattheconcentrationsappliedinourexperiments,donot affecttheshorttermfeedingbehaviorofT.longicornis.

TheresponseofthecopepodtoonestrainofA.pseudogonyaulax wasalsoexamined.Thisspecieshasbeendescribedtoproducethe neurotoxin goniodominA, but noPSTs. The A.pseudogonyaulax strainusedinthisstudycontainedadominantpeakofacompound withthemassofgoniodominAandwastheonlystrainthatelicited strongpreyrejectionresponses.

Inconclusion,thehigherresolutionofthebehavioralresponses revealed by direct observations compared to incubation approacheshasdemonstratedahighdegreeofstrain-specificity, not only in bulk grazing reduction but also in how grazing reduction is achieved. Our comparative approach of using a numberofAlexandriumstrains,whichareconsideredtoxicfroma humanhealthpointofview,differingintheamountandtypeof toxinswassuccessfulinprovidingfirstevidencethatgoniodominA plays a role as a true grazer deterrent. Moreover, there is no evidence that lytic compounds affect T. longicornis feeding behavior.Ontheotherhand,behavioralresponseofT.longicornis tothree PST-producingstrains, evenwhenacknowledgingtheir differencesintotalamountsandPSTprofile,wastoodifferentto acceptauniversalroleofPSTsinaffectingT.longicornisfeeding,and othersubstancesmayprovidethecuesforthediversebehavioral responsesobservedhere.Onepromisingavenuetopursuemaybe tocombinedirectlyobservedresponseswithmetabolicprofilingof thephytoplanktonasappliedtoresolveotherplanktonchemical cues(Selanderetal.,2016).

Acknowledgements

JiayiissupportedbyaPhDfellowshipfromChinaScholarship CouncilandTheCentreforOceanLifeissupportedbytheVillum Foundation.[SS]