Blind and naïve classiﬁcation of toxicity by fish chromatophores

CHROMATOPHORE CYTOSENSOR 355

Copyright © 2004 John Wiley & Sons, Ltd. J. Appl. Toxicol. 24, 355–361 (2004)

JOURNAL OF APPLIED TOXICOLOGYJ. Appl. Toxicol. 24, 355–361 (2004)Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/jat.1022

Received 21 December 2003Copyright © 2004 John Wiley & Sons, Ltd. Accepted 15 May 2004

Blind and Naïve Classification of Toxicity byFish Chromatophores

Ljiljana Mojovic,1 Karen P. Dierksen,1 Rosalyn H. Upson,1 Bruce A. Caldwell,2 Jeannine R.Lawrence,1 Janine E. Trempy2 and Philip N. McFadden1,*1 Department of Biochemistry and Biophysics, Oregon State University, Corvallis, OR 97330, USA2 Department of Microbiology, Oregon State University, Corvallis, OR 97330, USA

Key words: cytotoxicity; biosensor; cytosensor; chromatophore; Betta splendens.

Cellular and molecular pathways involved in the ability of animals to change color have been studied previ-ously as biosensors and cytosensors of active and toxic agents, but such studies generally have been limited tojust a few standardized agents. Here we describe the performance of cultured chromatophore pigment cellsfrom the fin tissue of Siamese fighting fish as sensors of toxic agents under blind sampling conditions at theSeptember 2002 EILATox-Oregon Workshop. Detection was accomplished by monitoring motor protein-mediated movements of cellular pigment in chromatophores at both the gross population level as well as insingly imaged cells. Pigment responses were recorded both during the exposure of chromatophores to eachblind sample as well as afterwards when the cells were examined for after-effects by challenging them withclonidine, an adrenergic drug that induces standardized pigment movements. After recording all results andupon breaking the key to reveal the identities of the toxic agents, it was found that all of the toxic samples inthe study had been distinguished accurately from non-toxic controls that were included among the blindsamples. Furthermore, it was revealed that most of the toxic agents detected had never before been tested orcalibrated against chromatophores, demonstrating that detection can be achieved under naïve conditions thathave not been optimized for the analysis of any particular toxic agent. Finally, by organizing the results intocategories of pigment responses, a binary classification tree was generated that distinguished each toxic agentas having a distinct response pattern from the others. Thus, chromatophore-based cytosensors can discovertoxicity in the absence of prior knowledge of the agent in question, and the categories of responses of the cellscan be used to distinguish one toxic agent from another. Copyright © 2004 John Wiley & Sons, Ltd.

* Correspondence to: P. N. McFadden, Department of Biochemistry andBiophysics, 2011 ALS, Oregon State University, Corvallis, OR 97330, USA.E-mail: [email protected]/grant sponsor: Office of Naval Research.Contract/grant sponsor: SPAWAR.Contract/grant sponsor: DARPA.Contract/grant sponsor: Army Research Office.

aggregation’, in which the pigment is moved to the centerof the chromatophore and the overall color of the cellbecomes lightened. The opposite effect, ‘pigment disper-sion’, involves outward translocation of pigment and celldarkening, and a wide range of pigment responses beyondthese two extremes is also observed.

Chromatophores are a promising basis for a real-time,optically monitored cytosensor of biologically active sub-stances because many such substances can be expected tointeract with the complex molecular mechanisms of thesecells to cause measurable alterations in their appearanceand ability to respond to standard stimulants. Becausechromatophores are fundamentally similar to neurons andother animal cells, chromatophore detection of activeagents ranging from ion channel drugs, blood pres-sure medications, to heavy metals, polynuclear aromatichydrocarbons and bacterial toxins makes them of use ascytosensors in medical screening, pharmaceutical research,food toxicology, environmental monitoring, and so on(McFadden et al., 2001).

In this study, we present results evaluating the effectson chromatophores of 13 blind and potentially toxicsamples. This test was part of the EILATox-OregonWorkshop held in Corvallis, OR, during 9–13 September,2002, where 15 other research teams measured the samesamples in the same location at the same time using state-of-the-art toxicity monitoring methods. In the following

INTRODUCTION

Chromatophores are neural crest-derived pigment cells thatare responsible for the brilliant colors of fish, amphibians,reptiles and cephalopods, and their dynamic color adapta-tions and adjustable camouflage capabilities (Fujii, 1993,2000). Various endogenous neurotransmitters and hor-mones induce changes in the appearance of chromatophoresby interacting with cell surface receptors that triggervarious intracellular signal transduction pathways thatthen regulate motor-protein-dependent transport of pig-mented organelles along the cytoskeleton (Danosky andMcFadden, 1997; Fujii, 2000; Reese and Haimo, 2000).Many varieties of pigment color and underlying signalregulation are found among the chromatophores of atypical animal (brown to black melanophores; orange tored erythrophores or yellow to orange xanthophores). Onecommon response to physiological stimulation is ‘pigment

356 L. MOJOVIC ET AL.


study, our team evaluated the responses of chromatophoresboth by using a commercially available multiwell platereader to monitor changes in the optical density of popula-tions of several thousand chromatophores as a function ofexposure to a sample, and by using an inverted white lightmicroscope to monitor movements of pigment within indi-vidually resolved chromatophores. Toxic impairment of thecells was also evaluated after exposure to a blind sampleby challenging the chromatophores with clonidine, a stand-ardized drug that operates through the alpha2-adrenergicreceptor to induce pigment aggregation Finally, we analyzedour workshop results and compiled them into a binarydecision tree that highlights how the range of chromato-phore responses was able to distinguish between the vari-ous blind samples and may be used to classify toxicity asdetected by future chromatophore-based instruments.

MATERIALS AND METHODS

Chemicals

Chlordimeform (PS-840; Lot 287-56B), phosdrin (PS-87,Lot 273-66B), sodium arsenite (I-121, ID 55-159B),metham sodium (PS-221, Lot 281-62B) and paraquat CLtetrahydrate (PS-366; Lot 287-53B) were purchased fromChem Service, Inc. (PA, USA). Cholchicine (C-3915, Lot111K1135), mercuric chloride (M-6529, Lot 111K3649) andsodium cyanide (S-3296, Lot 12K3456) were purchasedfrom Sigma, and 4-chlorophenol (18,578-7, [106-48-49]) wasfrom Aldrich. Trimethylol propane phosphate (TMPP) wasobtained through the US Naval Research Laboratory.Blind agent samples were prepared by dissolving the chem-icals in synthetic water (NaHCO3 48 mg l−1, CaSO4 2H2O30 mg l−1, MgSO4 30 mg l−1, KCl 2 mg l−1, in distilled water)supplied by CH2MHill (Corvallis, OR) at concentrationspresented in Table 1. Synthetic water was used to prepareblank samples with no toxic activity. Each agent sample wasmarked with a number. A list of the toxic agents was pro-vided to the participants, but the agent sample identitieswere unknown until the end of the workshop when theidentities were revealed. Table 1 presents the identifica-tion number of toxic agents, their chemical identificationand concentration.

Clonidine hydrochloride (Sigma, C-7897) and α-melanocyte-stimulating hormone (MSH; Sigma, M4135)

were used as standard chromatophore aggregation anddispersion controls respectively. Enzymes used for cell cul-ture preparation were from the Worthington BiochemicalCo., Leibovitz growth medium (L-15) was from Invitrogenand fetal bovine serum (FBS) was from Hyclone Lab.Other reagents used for cell culture were from Sigma,unless otherwise stated.

Chromatophore preparation

Chromatophores were isolated from the tails and fins ofBetta splendens. Tissue was washed a minimum of six timeswith skinning solution (1mM EDTA disodium salt, 5.6 mMglucose and penicillin–streptomycin mixture 1 : 100 w/v incalcium- and magnesium-free phosphate-buffered saline)and then treated with a digestion solution containing20 mg of collagenase type 1 (178 Umg−1, Worthington) and1 mg of hyaluronidase (348 USP/NF, Worthington) in 7 mlof phosphate-buffered saline with gentle agitation (ν =80 rpm). After 5–15 min, the enzyme solution with digestedtissue was separated from undigested tissue pieces andcentrifuged for 2 min at 3250 g in a clinical centrifuge.Supernatant was removed using a transfer pipette andreapplied to the tissue. The first pellet, which containsmostly epithelial cells, was discarded. After an additional5–15 min of shaking, the above procedure was repeatedand the cell pellet, containing chromatophores, wassuspended in L-15 medium and centrifuged again. Afterremoving the supernatant, chromatophores were re-suspended in a predetermined volume of L-15 medium toensure the desired plate density and a drop was platedonto the center of each well of a 24-well plate. Fifteenminutes after plating, 1.5 ml of L-15 medium was added ineach well. After 2 h, 5% fetal bovine serum was added tothe wells. After 24 h the old medium was withdrawn andreplaced with 1.5 ml of fresh L-15 medium.

Optical density monitoring

The change in the optical density of chromatophore cellcultures at 595 nm was monitored in a microplate reader(UltraMark Ex, BioRad). Two initial readings were takenbefore injection of agents in order to determine baselineoptical density. To initiate a measurement, 500 µl of a blindsample (samples 1–12 and 15) was added to a well. Controlsamples were measured in parallel and involved the addi-tion of 500 µl of either L-15, synthetic water, 400 nMclonidine in L-15 or 40 nM MSH in L-15. All agents wereadded within 2 min of each other. The final concentrationof sample in a well was one-quarter of the initial sampleconcentration presented in Table 1. Optical density read-ings were initiated immediately after the final agentwas added and repeated at time intervals for 45 min. Inorder to get additional information on the toxicity of thesamples and whether they had impaired the cells, clonidinewas added to the wells 1 h after addition of the samples.Optical density was then monitored for an additional10 min. On the basis of raw optical density of each well,a percent of the optical density change [%OD595 change =−(ODt=0 − ODt)/ODt=0 × 100] was calculated, along withaverage and standard deviations, and graphed. A posi-tive change in OD corresponded to pigment dispersion(chromatophore darkening) and a negative OD changecorresponded to pigment aggregation (chromatophorelightening).

Table 1—List of agents tested at the ELIATox-Oregon Workshop

Agent no Agent identity Concentration (mg l−1)

1 Chlordimeformis 12502 Colchicine 4003 Blank (synthetic water) –4 Phosdrin (mevinphos) 4005 Mercuric chloride 1006 Sodium arsenite 1007 Blank (synthetic water) –8 Metham sodium 4009 Sodium cyanide 50

10 Blank (synthetic water) –11 Trimethylol propane 100

phosphate (TMPP)12 p-Chlorophenol 30015 Paraquat 100



Image analysis

This technique is based on monitoring and recordingthe translocations of pigment within microscopically re-solved chromatophores as a function of their exposureto samples, plated in 24-well plates as described above.Either of two volumes of sample (500 or 166.66 µl) wasadded to the 1500 µl of media in a cell culture well, givingfinal concentrations equal to one-quarter and one-tenth(respectively) of the initial sample concentration presentedin Table 1. A field of view containing of the order of 100chromatophores was selected under a Leica DMIL invertedmicroscope using a 10× objective and tungsten lamp. Thefield of view was not moved during the entire interval ofdigital image capture, periodically performed with a PulnixTMD-7DSP CCD video camera interfaced to a Matroxdigital image capture card. The pigment area in the micro-scopically resolved field of chromatophores was thencalculated using an in-house software tool that was builton calls to image segmentation and measurement routinessupplied by a MatLab C++ software library (Danosky andMcFadden, 1997; Pacut et al., 2001; Preston and McFadden,2001). The total pigment area in pixels (digital area units)for each captured image was determined, and the percent-age pigment area change [Pigment area change (%) =−(Area0 − Areat)/Area0 × 100] was then calculated, alongwith mean and standard deviations.

RESULTS

Optical density monitoring

We calibrated the system at the outset of the EILATox-Oregon Workshop by exposing representative chromato-phores to standard agents (Fig. 1). The chromatophoresresponded to the standard aggregating agent, clonidine,with a rapid and complete aggregation of cell pigment,resulting in a change in the population OD of −56%within 10 min. Conversely, the standard dispersion agent,MSH, caused complete pigment dispersion such that thechromatophores reached an OD of +46% within 45 min.These measurements were in the range of previousmeasurements in our laboratory. We also studied bonafide blank samples at the outset of the workshop (Fig. 1).

Figure 1. Change in OD (%) versus time for control agents in theplate reader. Control agents: clonidine, 100 nM; MSH, 10 nM;L15; and synthetic water. Data presented are mean values ofthree measurements ± standard deviation.

Addition of background chromatophore growth mediumdid not produce a significant response in the fish chro-matophores. The synthetic drinking water that was thebackground for the blind samples caused a slight initialaggregation of fish chromatophores (less than −6% ODchange), followed by a slight dispersion (ca. 3% ODchange).

We next measured the response of chromatophores to13 blind samples at a 1 : 4 dilution of the original sampleconcentrations listed in Table 1. Figure 2 illustrates theOD change (%) of chromatophores versus time for blindsamples 1–6, and Fig. 3 presents the results for blind sam-ples 7–12 and 15. As shown in Fig. 2, samples 1, 2 and 4caused partial dispersion of chromatophores. The disper-sion caused by those three samples (1, 2, 4) was less thanthe full dispersion produced by the preceding MSH stand-ard (Fig. 1) and by concurrently performed standards ofMSH (not shown). Sample 3 (Fig. 2), as well as samples 7and 10 (Fig. 3), showed initial slight aggregation (<6%OD change), followed by slight dispersion that caused areturn of the population OD approximately back to thebaseline. This kinetic response of samples 3, 7 and 10 wassimilar to the kinetic response of the synthetic water pre-sented in Fig. 1. Sample 5 (Fig. 2) and samples 9, 11 and15 (Fig. 3) caused partial aggregation of chromatophores.With continued observation, the responses to samples 5

Figure 2. Change in OD (%) versus time for agents 1–7 in theplate reader. Data presented are mean values of three measure-ments ± standard deviation.

Figure 3. Change in OD (%) versus time for agents 8–12 and 15in the plate reader. Data presented are mean values of threemeasurements ± standard deviation.



Figure 4. Response of chromatophores to clonidine after expo-sure to the agents. Chromatophores were pre-exposed to theagents for 1 h and then challenged with clonidine (100 nM). TheOD was measured in the plate reader 0 min after clonidine addi-tion. Data presented are mean values of three measurements ±standard deviation.

Figure 5. Representative field chromatophores, captured at t = 0 sand t = 1200 s after addition of blind sample 6 (diluted 1 : 4).Translocations of pigment, such as the aggregation shown here,were quantified by image analysis as described in Materials andMethods.

Figure 6. Percentage cell area change obtained by image analysisafter 20 min of exposure of cell culture to agents. Two concen-trations of agents were applied (1 : 4 and 1 : 10). Results aremean values of three measurements ± standard deviation.

chromatophore cell culture frames exposed for 20 min totwo different concentrations of agents (1 : 4 and 1 : 10 dilu-tion; Fig. 6). A positive cell area change corresponds todispersion of pigment granules, while a negative cell areachange indicates their aggregation.

As shown in Fig. 6, a strong effect of agent dose wasobserved for agents 1, 4, 5, 6, 9, 11 and 15, whereas agents2, 3, 7, 8, 10 and 12 did not show a significant dose effect.A lower dose effect, coupled with other response cate-gories, could be an indication of either very low or veryhigh toxicity. Thus, a low direct effect and an absence ofan impaired effect on chromatophores, accompanied byno dose effect (observed for agents 3, 7 and 10), indicatesnon-toxicity. However, a low direct effect, coupled witha strong impaired effect and no dose effect (observed foragent 2 and 12), indicates a strong toxic effect. If a signific-ant reaction is observed, whether direct or impaired, theabsence of a dose effect might suggest that saturationalready has been reached.

Classification of agents

It became evident that the different toxic agents couldbe distinguished from each other according to certaincategories of chromatophore responses, such as the speedof the response, whether the response involved thechromatophores becoming darker or lighter in color, themagnitude of that color response, whether the responsewas strongly dose dependent, etc. By observing Bettasplendens erythrophore responses to various agents, themain differences seen are in the following responsefeatures:

(i) Mode of response: presence of visual reaction or noreaction of chromatophore.

(ii) Type of response: dispersion or aggregation.(iii) Magnitude of response: partial aggregation, full

aggregation, partial dispersion, or full dispersion (orquantified degree of aggregation or dispersion).

(iv) Kinetics of response: fast, slow, recoverable changesor irrecoverable changes.

(v) Indirect effect: impairment of clonidine reaction orno impairment.

(vi) Effect of concentration of agent: no effect or signific-ant effect.

and 9 then tended back toward the baseline; the responsesto samples 11 and 15 did not tend back to the baselineduring the observed period. A strong aggregation (>35%OD change) was the response to samples 6 (Fig. 2) and 8(Fig. 3). Finally, sample 12 caused slight aggregation thatdeveloped slowly (Fig. 3).

Toxic impairment of the cells by the blind samples wasstudied next by adding the standard aggregating agent,clonidine. Figure 4 illustrates chromatophore responsesto clonidine 1 h after exposure to a given blind sample.The clonidine response was impaired by samples 2, 5 and12, indicating that these samples had somehow inter-fered with one or more of the molecular steps in the pig-ment aggregation pathway. The other samples tested (1,3, 4, 6, 7, 8, 9, 10, 11 and 15) did not impair clonidine-induced aggregation.

Image analysis

We also used microscopy to monitor the responses ofchromatophores. Figure 5 shows quantitative measure-ments of such microscopic views. Pigment area changes insuch fields of view were determined by image analysis of



Figure 7. Classification of agents according to chosen response features (CL = clonidine).

These features were used for the classification of responsesinduced by the tested agents. The resulting classificationtree operationally defines the toxicity of blind samples,including nontoxicity. The classification of agents tested inthis study is presented graphically in Fig. 7.

DISCUSSION

Even though the two techniques utilized for evaluationof toxic effects use different parameters (OD and cellarea), cell responses were consistent in terms of their rela-tive intensity and mode of response. Both techniques wereefficient in monitoring the toxic responses. The plate readerformat enables simple, quick and high-throughput toxicityevaluations. However, it is important to point out that theimage analysis technique has greater potential than theplate reader format. Depending on the types of agents andthe needs of a particular study, the two methods can beutilized individually or combined. We found their com-plementary use very convenient, because it can providerather comprehensive analysis in a short period. The mainbenefit of image analysis is the ability to differentiate re-sponses of particular cell subtypes such as erythrophoresand melanophores relating to their color, and/or sheet-likeand dendritic cells relating to their morphology. How-ever, this was not utilized to its full extent in this study,because only cell preparations based on erythrophores wereused for toxic evaluation. Responses of different types ofBetta splendens chromatophores to biologically activeagents have shown significant kinetic differences betweendendritic cells, usually characterized as faster responders,

and round, or sheet-like cells, usually characterized asslower responders (unpublished data). However, the dif-ferences are also agent-related. Different response patterns,suggesting differences in signal transduction pathways, wereobserved in Betta splendens and other fish melanophoreand erythrophore responses to Ca2+ channel modulators(Andersson et al., 1984; Sammak et al., 1992; Prestonand McFadden, 2001). Our current studies indicate lowersensitivity of Betta splendens melanophores comparedwith erythrophores for a panel of eight agents, includingadrenergic, dopaminergic, heavy metal and bacterial agents(unpublished data). Recent developments in chromato-phore image analysis software have made it possible todifferentiate agents according to their specific action onchromatophore subtypes.

The classification of agents presented in Fig. 7 categor-ized agents 3, 7 and 10 in the same group according totheir response features. The direct effect of these agentson chromatophores was not significant, neither was anindirect effect, suggesting their non-toxicity. Table 1 iden-tifies agents 3, 7 and 10 as synthetic water, thus confirmingour assumption. Feature response analysis and classifica-tion suggested a toxic effect for all other agents tested,with particular differences in the magnitude of the toxiceffect and mechanism of action. A very strong toxiceffect was suggested for agents 2 (colchicine), 5 (mercuricchloride), 12 (p-chlorophenol), 6 (sodium arsenite) and8 (metham sodium).

Cholchicine is a plant-derived cytoskeletal toxin thatbinds to microtubules, preventing cell division. It has beenused in chemotherapy for cancer and gout (Seagull, 1990).In chromatophores it disrupts the normal movement ofpigment granules along microtubules. A similar effect and



mechanism of action in chromatophores was produced byp-chlorophenol, a very hazardous water pollutant with anextremely toxic effect to fish and aquatic organisms (Yenet al., 2002), as well as to earthworms (Miyazaki et al.,2002). In this study, both of these agents exhibited a ratherweak direct effect on chromatophores (weak dispersionand weak aggregation, respectively) with a strong impair-ment of the clonidine response.

Mercuric chloride (agent 5) and sodium arsenite (agent6) are very dangerous environmental, especially water,pollutants causing serious toxic and long-term health-damaging and carcinogenic effects on humans, rats andmany other species or cells. The US Environmental Pro-tection Agency (EPA) set very strict regulations for thesetwo pollutants in ground and drinking water, allowing amaximum contaminant level (MCL) in drinking water of0.002 mg l−1 for mercury and 0.01 mg l−1 for arsenic. There-fore, a number of mercury- and arsenite-specific biosensorshave been studied and developed. Researchers from TheUniversity of Turku, Finland, whose technology was pre-sented at the workshop, developed a whole-cell bacterialsensor for the detection of heavy metals (specificallymercury and arsenite) by measuring luciferase-catalyzedbioluminescence (Tauriainen et al., 1999; Ivask et al., 2001;Hakkila et al., 2003). A similar concept was utilized fordesigning a fibre-optic biosensor for mercury detection,based on genetically engineered E.coli MC 1061 contain-ing the mercury-responding promoter to control the ex-pression of luxCDABE from Photorhabdus luminescens(Green et al., 2002), as well as for developing a latex-coatedmercury sensor using E.coli-harboring pRB28 with amer-lux constract (Lyngberg et al., 1999, 2001). However,the use of bioluminescence-based techniques is not limitedonly to mercury and arsenite testing. It can be used alsofor more general cytotoxicity and genotoxicity sensors,such as the Vitotox System (Verschaeve et al., 1999), alsopresented at the workshop.

Both mercury and arsenite induced aggregation ofchromatophores. Concentrations tested for both agentswere 25 and 10 mg l−1. The aggregation induced by arsenite(agent 6) was much stronger than that obtained bymercury (Fig. 7). Also, mercury caused an indirect effect,e.g. a partially inhibited clonidine response. Further testingis needed to define chromatophore sensitivity thresholdsfor these agents.

Metham sodium (agent 8), the third most commonlyused agricultural pesticide (by weight) in the USA, hasadverse ecological and human health effects (Pruett et al.,2001). The major breakdown product of metham sodiumis methylisothiocyanate (MITC), which can cause skin, eyeand respiratory irritation and also has potential as botha chemical warfare and terrorism agent. Animal studiessuggest a potential for immunological, developmental,carcinogenic and atherogenic effects (Burgess et al., 2000;Giannakou et al., 2002). Metham sodium induced a strongdirect effect on fish chromatophores, a strong and irre-coverable aggregation achieving a 35% change in opticaldensity after 15 min (Fig. 2). No significant dose effect

was observed for the 100 and 40 mg l−1 concentrations,suggesting that saturation sensitivity might be achieved withthe lower tested dose. However, further testing is neededto determine a precise dose dependence pattern and mini-mum sensitivity.

Agents 1 (chlordimeform), 4 (phosdrin) and 15(paraquat), all of them known as pesticides with differentchemical structures, were classified in different groups(Fig. 7). Chlordimeform and phosdrin caused disper-sion of chromatophores of different magnitude, whereasparaquat induced partial aggregation of the chromato-phores. It is important to point out that phosdrin, anorganophosphate agent, was the most acutely lethal of theagents tested at the workshop. It is a potent neurotoxineasily absorbed through the skin or by inhalation (Fox andCastillon, 2001). Less than seven drops may be fatal tohuman adults (Lokan and James, 1983). This pesticide ischemically related to the nerve gas sarin, and could be ofuse both for chemical warfare and terrorism. However,in contrast to the high toxicity to humans, fish chromato-phores demonstrated rather low direct (dispersion) andno indirect reaction to phosdrin.

The presence of sodium cyanide (agent 9), a verytoxic water and food pollutant, at concentrations 12 and5 mg l−1 was recognized by fish chromatophores with apartial aggregation with reversible kinetics. Low concen-trations (several mg l−1) in water may cause thyroid effectsand nerve damage in humans according to the EPA,whereas it is known that rats can tolerate up to 300 mg l−1

in their drinking water with no apparent health effects(Cassel, 1995).

Finally, TMPP (agent 11), a combustion product of fuelin airplanes and other military vehicles, is known to causeneuropathies in chronically exposed workers (Wymanet al., 1993; Keefer et al., 2001). On chromatophores, itcaused a partial aggregation that was irreversible (Fig. 7).

In conclusion, the chromatophore-based biosensorefficiently recognized and differentiated all toxic agentstested at the EILATox-Oregon Workshop, thus con-firming its potential as an activity biosensor with a broadspecificity range. Although reactivity of chromatophoresto all tested toxic agents was detected, a comparativelystrong toxic effect of cholchicine, mercuric chloride, p-chlorophenol, sodium arsenite and metham sodium wasobserved. The key to the versatility of this biosensor is inthe variety of response features that can be defined andused as a classification tool, thus enabling identificationof classes of chemical compounds according to similarresponse patterns.

Acknowledgments

This work was supported by 6.1 funding from the Office of Naval Research(Joel Davis and Eric Eisenstadt program managers) and 6.2 funding fromSPAWAR San Diego (Shannon Kasa and Cynthia Smith contract managers)and DARPA (Tissue-based Biosensor Program and the Activity DetectionTechnologies Program, Alan Rudolph program manager). The EILATox-Oregon Workshop was supported by DARPA and by the Army ResearchOffice (Stephen Lee program manager).



REFERENCES

Andersson RG, Karlson JO, Grundstrom N. 1984. Adrenergicnerves and the alpha 2-adrenoreceptor system regulatingmelanosome aggregation within fish melanophores. ActaPhysiol. Scand. 121: 173–179.

Burgess JL, Morrissey B, Keifer MC, Robertson WO. 2000.Fumigant-related illnesses: Washington State’s five-yearexperience. J. Toxicol. Clin. Toxicol. 38: 7–14.

Cassel GE. 1995. Estimation of the convulsive effect of cyanide inrats. Pharmacol. Toxicol. 77: 259–263.

Danosky TR, McFadden PN. 1997. Biosensors based on the chro-matic activities of living, naturally pigmented cells. Biosens.Bioelectron. 12: 952–936.

Fox A, Castillon K. 2001. Employee exposure to mevinphos — anorganophosphate pesticide. Appl. Occup. Environ. Hyg. 16:506–508.

Fujii R. 1993. Cytophysiology of fish chromatophores. Intern.Rev. Cytol. 143: 191–255.

Fujii R. 2000. The regulation of motile activity in fish chromato-phores. Pigment Cell Res. 13: 300–319.

Giannakou IO, Sidiropoulos A, Prophetou-Athanasiadou D. 2002.Chemical alternatives to methyl bromide for the control ofroot-knot nematodes in greenhouses. Pest. Manag. Sci. 58:290–6.

Green T, Ivask A, Virta M, Marks R. 2002. Optical fiber biosensorfor the detection of Mercury. In Proceedings of Elia-Tox-Oregon Workshop, 9–13 September, Oregon State Univer-sity, 7.

Hakkila K, Maksimow M, Rosengren A, Karp M, Virta M. 2003.Monitoring promoter activity in a single bacterial cell byusing green and red fluorescent proteins. J. Microbiol.Methods 54: 75–79.

Ivask A, Hakkila K, Virta M. 2001. Detection of organomercurialswith sensor bacteria. Anal. Chem. 73: 5168–5171.

Keefer EW, Gramowski A, Stenger DA, Pancrazio JJ, Gross GW.2001. Characterization of acute neurotoxic effects oftrimethylolpropane phosphate via neuronal networkbiosensors. Biosens. Bioelectron. 16: 513–525.

Lokan R, James R. 1983. Rapid death by mevinphos poisoningwhile under observation. Forens. Sci. Int. 22: 179–82.

Lyngberg OK, Stemke DJ, Schottel JL, Flickinger MC. 1999.A single-use luciferase-based mercury biosensor usingEscherichia coli HB101 immobilized in a latex copolymerfilm. J. Ind. Microbiol. Biotechnol. 23: 668–676.

Lyngberg OK, Ng CP, Thiagarajan V, Scriven LE, Flickinger MC.2001. Engineering the microstructure and permeability ofthin multilayer latex biocatalytic coatings containing E. coli.Biotechnol. Prog. 7: 1169–1179.

McFadden P, Chaplen F, Kolodziej W, Jovanovic G, Plant T,Liburdy J, Paul B, Trempy J, Gerwick W, Willard C. 2001.Methods of detecting bioactive compounds. Utility Patent245–59404.

Miyazaki A, Amano T, Saito H, Nakano Y. 2002. Acute toxicity ofchlorophenols to earthworms using a simple paper contactmethod and comparison with toxicities to fresh water organ-isms. Chemosphere 47: 65–69.

Pacut A, Kolodziej W, Chaplen F. 2001 Cytosensors for earlydetection of biological and chemical threats. Proce. 4thInte. Conf. on Neuronal Networks and Expert Systems inMedicine and Healthcare, NNESMED 2001, Milos Island,Greece, pp. 437–442.

Preston RR, McFadden PN. 2001. A two-cell biosensor thatcouples neuronal cells to optically monitored fish chromato-phores. Biosens. Bioelectron. 16: 447–455.

Pruett SB, Myers LP, Keil DE. 2001. Toxicology of metam sodium.J. Toxicol. Environ. Health B Crit. Rev. 4: 207–222.

Reese E, Haimo L. 2000. Dynein, dynactin, and kinesin II’s inter-action with microtubules is regulated during bi-directionalorganelle transport. J. Cell Biol. 151: 155–166.

Sammak PJ, Adams SR, Harootunian AT, Schliwa M, Tsien RY.1992. Intracellular cyclic AMP not calcium, determines thedirection of vesicle movement in melanophores: directmeasurement by fluorescence ratio imaging. J. Cell Biol.117: 57–72.

Seagull RW. 1990. The effects of microtubule and microfila-ment disrupting agents on cytoskeletal arrays and walldeposition in developing cotton fibers. Protoplasma 159: 44–59.

Tauriainen S, Virta M, Chang W, Karp M. 1999. Measurement offirefly luciferase reporter gene activity from cells and lysatesusing Escherichia coli arsenite and mercury sensors. Anal.Biochem. 272: 191–198.

Verschaeve L, Van Gompel J, Thilemans L, Regniers L, VanparysP, van der Lelie D. 1999. VITOTOX bacterial genotoxicity andtoxicity test for the rapid screening of chemicals. Environ.Mol. Mutagen. 33: 240–248.

Wyman J, Pitzer E, Williams F, Rivera J, Durkin A, Gehringer J,Serve P, von Minden D, Macys D. 1993. Evaluation of ship-board formation of a neurotoxicant (trimethylolpropanephosphate) from thermal decomposition of syntheticaircraft engine lubricant. Am. Ind. Hyg. Assoc. J. 54: 584–592.

Yen JH, Lin KH, Wang YS. 2002. Acute lethal toxicity of environ-mental pollutants to aquatic organisms. Ecotoxicol. Environ.Saf. 52: 113–116.

Documents

Blind and naïve classiﬁcation of toxicity by fish chromatophores