Assessment of serum biomarkers in rats after exposure to pesticides of different chemical classes

Toxicology and Applied Pharmacology 282 (2015) 161–174

Contents lists available at ScienceDirect

Toxicology and Applied Pharmacology

j ourna l homepage: www.e lsev ie r .com/ locate /ytaap

Assessment of serum biomarkers in rats after exposure to pesticides ofdifferent chemical classes☆

Virginia C. Moser a,⁎, Nicholas Stewart a, Danielle L. Freeborn a, James Crooks b, Denise K. MacMillan b,Joan M. Hedge c, Charles E. Wood c, Rebecca L. McMahen d, Mark J. Strynar e, David W. Herr a

a Neurotoxicology Branch/Toxicity Assessment Division, National Health and Environmental Effects Research Laboratory, Office of Research and Development, US Environmental Protection Agency,Research Triangle Park, NC 27711, USAb Analytical Chemistry Research Core/Research Cores Unit, National Health and Environmental Effects Research Laboratory, Office of Research and Development, US Environmental ProtectionAgency, Research Triangle Park, NC 27711, USAc Integrated Systems Toxicology Division, National Health and Environmental Effects Research Laboratory, Office of Research and Development, US Environmental Protection Agency,Research Triangle Park, NC 27711, USAd ORISE fellow, Human Exposure and Atmospheric Sciences Division, National Exposure Research Laboratory, Office of Research and Development, US Environmental Protection Agency, ResearchTriangle Park, NC 27711, USAe Human Exposure and Atmospheric Sciences Division, National Exposure Research Laboratory, Office of Research and Development, US Environmental Protection Agency, Research Triangle Park,NC 27711, USA

☆ Disclaimer: The views expressed in this paper are thosessarily reflect the views or policies of the US Environmen⁎ Corresponding author at: MD B105-04, US EPA, Res

USA.E-mail address: [email protected] (V.C. Moser).

http://dx.doi.org/10.1016/j.taap.2014.11.0160041-008X/© 2014 Published by Elsevier Inc.

a b s t r a c t

Article history:Received 10 September 2014Revised 3 November 2014Accepted 26 November 2014Available online 9 December 2014

Keywords:BiomarkersCytokinesMetabolomicsHormonesPesticides

There is increasing emphasis on the use of biomarkers of adverse outcomes in safety assessment and translationalresearch. We evaluated serum biomarkers and targeted metabolite profiles after exposure to pesticides(permethrin, deltamethrin, imidacloprid, carbaryl, triadimefon, fipronil) with different neurotoxic actions.Adult male Long–Evans rats were evaluated after single exposure to vehicle or one of two doses of each pesticideat the time of peak effect. The doses were selected to produce similar magnitude of behavioral effects acrosschemicals. Serumor plasmawas analyzedusing commercial cytokine/protein panels and targetedmetabolomics.Additional studies of fipronil used lower doses (lacking behavioral effects), singly or for 14 days, and includedadditional markers of exposure and biological activity. Biomarker profiles varied in the number of alteredanalytes and patterns of change across pesticide classes, and discriminant analysis could separate treatmentgroups from control. Low doses of fipronil produced greater effects when given for 14 days compared to a singledose. Changes in thyroid hormones and relative amounts of fipronil and its sulfone metabolite also differedbetween the dosing regimens. Most cytokine changes reflected alterations in inflammatory responses, hormonelevels, and products of phospholipid, fatty acid, and amino acidmetabolism. These findings demonstrate distinctblood-based analyte profiles across pesticide classes, dose levels, and exposure duration. These results showpromise for detailed analyses of these biomarkers and their linkages to biological pathways.

© 2014 Published by Elsevier Inc.

Introduction

There is great promise in the use of blood-based biomarkers as indica-tors of adverse health outcomes in safety assessment, environmental ex-posure, and translational research. The search for bioindicators of toxicityhas been aidedwith expanding technologies capable of measurements ofhundreds to thousands of analytes, frommRNAandmicroRNA transcriptsto proteins and metabolites (Collings and Vaidya, 2008). In clinical trials,there is a critical need for biomarkers of effect that can serve as predictors

e of the authors and do not nec-tal Protection Agency.earch Triangle Park, NC 27711,

or surrogate endpoints for adverse outcomes, especially if such monitor-ing can detect toxicity at an early, preclinical stage (BiomarkersDefinitionsWorking Group, 2001). To be relevant for humanmonitoring,such markers must be measurable in accessible sample types such asserum/blood or urine. In safety pharmacology and toxicology studies, bio-marker research has been used to provide a basis for understandingmechanisms and adverse outcome pathways for specific manifestationsof toxicity. These efforts highlight the continuing need for biomarkerdiscovery, evaluation, and interpretation.

Research is ongoing to identify and validate biomarkers of nervoussystem changes that relate to the underlying molecular pathways(e.g., Wiesinger et al., 2012). One option is to evaluate biochemicalmarkers of neurotransmission function aswell as neuronal injury; how-ever, these endpoints are typically specific for certain modes of action(pathways) and particular chemicals (Manzo et al., 1996). On theother hand, exposures to a variety of environmental agents may

162 V.C. Moser et al. / Toxicology and Applied Pharmacology 282 (2015) 161–174

produce disease outcomes that are distinct from the primary site of tox-icity, someofwhichmaynot be knownor predicted a priori. In epidemi-ological and/or occupational studies, typical neurological complaintsinclude cognitive, emotional, neuromuscular, or sensory changes.These conditions may be the result of a generalized stress or off-targetresponses in the body, but it is as yet unclear whether specific patternsof such responses may vary with differentmanifestations of toxicity, in-cluding neurotoxicity. Thus, an understanding of biomarkers that couldbe related to the body's response to neurotoxicants could be importantfor evaluating health impacts in toxicological and community-basedstudies.

Assessments of biomarkers linked to physiological or pathologi-cal changes, including cytokines and other biological proteins, holdadditional promise as predictive or indicators of some adverse out-come. Such markers should be relatively specific and sensitive, beamenable to rapid measurement from accessible body fluids, and re-flect underlying biological processes. To address this need, biomark-er panels have been recommended for both preclinical safetyassessment and environmental monitoring (e.g., Duramad andHolland, 2011; Duramad et al., 2007; Tarrant, 2010). Furthermore,biomarker patterns may be mined for pathways consistent with spe-cific biological states, providing additional information on the toxi-cological effects of these chemicals.

Biomonitoring studies such as the National Health and NutritionExamination Survey (NHANES) (Crinnion, 2010) demonstrate thathumans are routinely exposed to a large number of pesticides. Cumula-tive exposures to these chemicals in the general population occur viafood, air, dust, and water. Pesticides in use today represent a numberof different chemical classes, including organophosphates, carbamates,pyrethroids, and others. Insecticides in particular often target the ner-vous system, with potentially greater sensitivity in the targeted pestsconferring species selectivity. While exposure is especially of concernfor children's health, the general population includes other sensitiveor vulnerable populations for which pesticide exposuresmay also resultin adverse effects. Widespread low-dose exposures may be difficult todocument when using existing biomarkers of exposure such as levelsof parent chemical or metabolite in urine or blood.

Prior work has shown distinct biomarker profiles in rats 24 h afterexposure to different classes of acetylcholinesterase inhibiting pesti-cides (carbaryl and chlorpyrifos) (Gordon andWard, 2009). Itwas of in-terest, therefore, to investigate changes in biomarker patterns followingexposure to a wider range of acutely neurotoxic pesticides. In thesestudies, we evaluated single exposures to six pesticides (permethrin,deltamethrin, imidacloprid, carbaryl, triadimefon, fipronil) havingdifferent modes of nervous system toxicity. We cast a broad net to ex-plore patterns of serum biomarkers and metabolites, using targetedpanels of cytokines, hormones, enzymes, and metabolites, in additionto in vivo neurophysiological measures (changes in electroencephalog-raphy, or EEG). Resultant toxicological pathways could then be furthermined using bioinformatic approaches. Biomarker outcomes and pro-files are presented here, whereas a companion paper presents EEGdata (Freeborn et al., in press). Single and 14-day exposures to fipronilwere also evaluated to compare changes in relation to exposure dura-tion, and to link serum biomarkers and EEGmeasures with standard in-dicators of toxicity and exposure. The overall goal of this work was toassess the experimental approach (proof of concept) and provide pre-liminary data on potential biomarker patterns associated with pesti-cides having different mechanisms of toxicity. Distinct patterns couldbe a result of pesticide-specific biological actions, whereas a generalizedstress response could present a similar pattern across chemicals. Poten-tial uses of the data are twofold: 1) provide patterns of biomarkerchanges from which may be derived putative adverse outcomepathways that may or may not be related to the primary CNS targetsof these pesticides, and 2) suggest possible screening approaches forchanges in serum biomarkers that could be further developed forpossible application in human population biomonitoring.

Materials and methods

Chemicals. Test chemicals, listed in Table 1,were all obtained fromChem-Service (West Chester, PA). The chemicals represent type I (permethrin)and II (deltamethrin) pyrethroids, neonicotinoid (imidacloprid), N-methyl carbamate (carbaryl), triazole (triadimefon), and phenylpyrazole(fipronil) pesticidal classes. All chemicals were suspended in corn oil,except for imidacloprid, which was suspended in 0.5% methylcelluloseand 0.4% Tween® 80 in distilled water.

Animals. Adult (60–90 days of age)male Long–Evans rats (Charles RiverLaboratories, Wilmington, MA) were used for all studies. They werehoused individually on heat-treated hardwood chip bedding in anAAALAC-International approved facility maintained at 22 ± 2 °C, withhumidity at 40 ± 20%, and a light:dark 12:12 h cycle period (lights onat 6:00 am). All procedures were approved by the National Health andEnvironmental Effects Research Laboratory (NHEERL) Animal Care andUse Committee. Rats were randomly assigned to treatment groups(n = 9–18/dose/chemical) for EEG testing, and tissues from some orall of the rats were used for the various measurements (describedbelow). Each chemical studywas conducted separately, with the excep-tion of the pyrethroid studies. In the pyrethroid studies, one cohortreceived vehicle or low doses of deltamethrin or permethrin, and thesecond cohort received vehicle or higher doses of these chemicals(Freeborn et al., in press). A third cohort received vehicle or oneof eitherdoses of each pyrethroid.

Experimental design. Details of the different experiments are presentedin Freeborn et al. (in press) and summarized in Table 1. Two doseswere selected that were approximately equi-effective in terms ofin vivo changes as measured by motor activity, an apical behavioralmeasure that is altered by all these pesticides, using data obtainedfrom the literature and regulatory study submissions. The doses select-ed produce decreases in activity (except triadimefon, which increasesactivity) of approximately 30% (low dose) or 50% (high dose; exception,carbaryl produced 80% effect at high dose). Data were collected at thetime of peak effect, using time-coursemotor activity data fromavailableinformation. Pilot studies were conducted in-house for imidaclopridand fipronil to verify both the dose selection and the estimated timeto peak behavioral signs of toxicity. EEG testing was conducted at thefollowing times after a single oral dose: permethrin, deltamethrin,imidacloprid, 2 h; carbaryl, 30 min; triadimefon, 1 h; and fipronil, 6 h.

The high doses (25, 50 mg/kg) selected for fipronil met the abovebehavioral criteria. Fipronil was chosen for further examination on abroader range of endpoints (described below) to integratewith the bio-marker data, including in-house human and environmentalmonitoring,and to compare effects of single and repeated exposures. The additionalstudies were conducted using lower doses (which did not produceacute changes in motor activity: 5, 10 mg/kg) administered as bothsingle and repeated (14-day) exposures. The 6-h test time was main-tained in these acute and repeated studies (6 h after the last dose onthe 14th day).

Almost all rats were tested using EEG recordings and colonictemperature, lasting about 20 min, shortly before blood was taken forbiomarker analyses. The EEG data are presented elsewhere (Freebornet al., in press). One of the pyrethroid cohorts did not undergo surgeryor EEG testing, but otherwise this cohort was dosed and treated in thesame manner as the other rats.

Serum/plasma preparation. About 5–10 min after EEG testing, the ratswere euthanized by decapitation and whole blood was collected intotubes with or without anticoagulants for the various biochemical mea-surements. Blood was separated into plasma (Meso Scale Discoveryand Biocrates assays, see below) and serum (Myriad RBM assays, seebelow, for all groups, and thyroid, liver enzymes, and levels of fiproniland its sulfone metabolite for fipronil groups). Serum was prepared

Table 1Experimental information for each study, including biological action, dosing, and testing times.

Chemical Permethrina Deltamethrina Imidacloprid Carbaryl Triadimefon Fipronila

CAS 52645-53-1 52918-63-5 138261-41-3 63-25-2 43121-43-3 120068-37-3Purity 46.1% cis, 53.2% trans 99.5% 99.5% 99.5% 99.5% 98.5%-98.8%Class Type I pyrethroid Type II pyrethroid Neonicotinoid N-methyl carbamate Triazole PhenylpyrazoleNeurological action Prolonged Na+

channel activationand depolarizationb

Lengthens Na+

channel activationand repetitive firingb

Activates and blocksnicotinic receptorsc

Inhibits acetylcholinesterased

Inhibits dopaminereuptakee

Inhibits GABAreceptorf

Vehicle Corn oil Corn oil 0.5% methyl-cellulose/0.4% Tween 80/water

Corn oil Corn oil Corn oil

Doses (mg/kg) 0, 43 0, 100 0, 43, 100 0, 2.5 0, 5.5 0, 2.5, 5.5 0, 50, 100 0, 10, 50 0, 75, 150 0, 25, 50 0, 5, 10 0, 5, 10 for14 days

Time of testing (h) 2 2 2 2 2 2 2 0.5 1 6 6 6 (after last dose)

a Three separate cohorts of rats.b Narahashi et al. (2007).c Matsuda et al. (2001).d Casida (1963).e Ikaiddi et al. (1997).f Bloomquist (2003).

163V.C. Moser et al. / Toxicology and Applied Pharmacology 282 (2015) 161–174

after storing the blood on ice for 1–1.5 h. Inmost studies, the tubeswerecentrifuged at 13,000 ×g for 2 min at 4 °C. To accommodate additionalendpoints in the single and repeated low-dose fipronil studies, thesesamples were centrifuged at 1300 ×g for 30min at 4 °C. For all samples,the serum supernatant was collected and frozen on dry ice. Inmost sin-gle dose studies, plasma was prepared by collecting blood into tubescontaining heparin (100 IU in 5 μl) and stored on ice. In the single andrepeated low-dose fipronil studies, K2EDTA was used as the anticoagu-lant (3.6 mg K2EDTA/2ml blood in 10 μl). All plasma samples were cen-trifuged at 950–1000 ×g for 10 min at 4 °C, and the supernatant wascollected and frozen on dry ice. All tissues were stored at −80 °C untilassays were performed.

Myriad RBM assays. Panels of serum analytes for potential biomarkerswere commercially analyzed by Myriad RBM, Inc (Austin, TX) usingtheir RodentMAP® (Multi-Analyte Profile) and RatMetabolicMAP®Luminex® bead-based multiplex immunodetection platforms. Duringthe course of this study, the composition of these panels changed sothat data are not available for all analytes for all chemicals (https://rbm.myriad.com/myriad-rbm-assay-change-log/). The manufacturerdescribes extensive quality assurance procedures for each assay(https://rbm.myriad.com/scientific-literature/data-quality/). Theseanalyses provided data on about 80 proteins and hormones for eachrat in each study, except for the pyrethroid study in which two cohortswere analyzed only with the RodentMAP® and the third cohort onlywith the RatMetabolicMAP® (cohort data were combined for simplici-ty). Reported values below the least detectable dose (LDD; mean ± 3standard deviations of 20 blank readings) were set to missing for dataanalysis.

This analysis was conducted on a subset of serum samples: n = 5/dose (single dose studies) or n = 10–11/dose (fipronil repeated-dosestudies). The sample selection was based on the EEG outcomes. Inorder to maximize potential for obtaining effects with a small samplesize, for each chemical the EEG measure showing the greatesttreatment-related effectwas evaluated. Serum samples from rats show-ing effects at the magnitude of the group median and greater wereanalyzed.

Meso Scale Discovery assays. In addition to the RBM analyses, plasmafrom all rats in all studies was analyzed using the Rat Demonstration7-plex Ultra-sensitive Kit, a multi-assay electrochemiluminescence sys-tem, from Meso Scale Discovery (MSD; Rockville, MD). Assays wereconducted on 96-well plates according to the kit protocol, but with alarger volume of serum (50 μl) tomaximize detection, and then quanti-fied using a Sector photodetector/imager 2400. This assay provided datafor seven analytes, six of which were also evaluated in the RBM

analyses. The MSD platform provided lower limits of detection (LOD;mean ± 2.5 standard deviations above zero calibrator on standardcurve) than the RBM assays for these six analytes.

Biocrates assays. Targetedmetabolomic analyses were also conducted onplasma from almost all rats using the AbsoluteIDQ™ p180 kit fromBiocrates Life Sciences (Innsbruck, Austria), a mass spectrometry (MS)-based analysis of 186 metabolites of different classes (acylcarnitines,amino acids, hexoses, biogenic amines, glycerophospholipids, andsphingolipids). Analyses were performed according to kit instructions,using an AB Sciex (Framingham, MA) 4000 Qtrap linear ion trap massspectrometer. Plasma samples (10 μl) were derivatized and extractedon 96-well plates. Analytes were detected by LC/MS using an Agilent(Santa Clara, CA) Zorbax Eclipse XDB C18 column. Analysis by flow injec-tion (FIA)/MS followed. Analytes were quantitated using multiplereaction monitoring (MRM) transitions and internal standards. Datawere validated and processed using Biocrates MetIDQ™ software. Be-cause plasma was used for acetylcholinesterase activity assays(Freeborn et al., in press), there was insufficient plasma from carbaryl-treated rats to conduct these assays.

Additional fipronil measures. In the low-dose single and repeated-dosefipronil studies, additional endpoints were quantified in all rats (n =9–12/group). Serum total thyroxine (T4) and total triiodothyronine(T3) were measured in duplicate by standard solid-phase Coat-A-Count® radioimmunoassay (RIA) according to kit directions (SiemensMedical Solutions Diagnostics, Los Angeles, CA). Assay variation wasassessed using the multivalent control module (Siemens Medical Solu-tions Diagnostics, Los Angeles, CA; CON6, lot 23) before and after mea-suring the experimental samples; overall intra-assay coefficients ofvariationwere 3.2% for T3 and 4.6% for T4. Serum levels of alanine trans-aminase (ALT), aspartate transaminase (AST), lactate dehydrogenase(LDH), and sorbitol dehydrogenase (SDH) were quantified as bio-markers of liver injury using assays modified for use on the KonelabArena 30 clinical chemistry analyzer (Thermo Clinical Labsystems,Espoo, Finland). Kits and controls for LDH and ALT were obtained fromThermo Fisher Scientific (Middletown, Virginia), SDH from Sekisui Di-agnostics (Charlottetown, Prince Edward Island, Canada), and ASTfrom TECO Diagnostics (Anaheim, CA). To evaluate overt liver toxicity,liver samples were collected at necropsy from all of the low doseacute and repeated-dose animals andfixed in 70% ethanol for histopath-ological evaluation. Fixed tissues were paraffin-embedded, sectioned at5 μm, and stained with hematoxylin and eosin (H&E) using standardhistologic procedures. Three sections per animal from the left lateral,right median, and caudate lobes were evaluated by a board-certifiedpathologist.

164 V.C. Moser et al. / Toxicology and Applied Pharmacology 282 (2015) 161–174

Quantitative analysis for fipronil in serum was carried out using anAgilent 1100 HPLC (Agilent Technologies, Palo Alto, CA) interfacedwith a Sciex 3000 triple quadrupole mass spectrometer (AppliedBiosystems/MDS Sciex, Foster City, CA) fitted with an electrosprayionization source (ESI) operated in the negative ionization mode.The HPLC system consisted of a Phenomenex Luna C18 column(50 × 3 mm, 5 μm; Torrance, CA, USA) with a Security-guard guardcolumn (Phenomenex). Compounds screened for in the LC/triple-quadmethod (fipronil, fipronil sulfone, fipronil sulfide, fipronil amide, andmonochloro fipronil) were optimized on a compound-specific basis.Sample proteins were precipitated with acetonitrile, centrifuged, andextracted with ammonium acetate buffer as described in McMahenet al. (submitted for publication). Fipronil des F3 was used as internalstandard, and limit of quantitation (LOQ) of both fipronil and fipronilsulfonewas determined to be 10ng/ml. Urinary and some serum resultsare reported inMcMahen et al. (submitted for publication); only serumdata are reported here.

Statistical analyses. Statistical analyses of the RBM andMSD analyte datawere conducted using SAS (v9.1; SAS Institute Inc., 2004). The ap-proaches used for statistical analyses of treatment effects differedbased on the dependent variable. Serum biomarker data were firstfiltered by setting all values less than the LDD as missing. Analysis ofvariance (ANOVA) was used for data in which at least 90% of the sam-ples were above the LDD. Given the small sample size, we accepted anuncorrected alpha level of 0.05 to be less conservative. Following a sig-nificant treatment effect, Tukey's HSD test was used to compare groups(Kramer, 1956). For analytes that had fewer than 90% of the samplesabove the LDD, the data were converted into a binary response asbeing above or below the LDD. These binary data were analyzed usingcategorical analysis (PROC CATMOD), followed by t-test contrasts withp-value thresholds b0.05. This approach required that at least 50% ofat least one dose group had values greater than the LDD. The incidenceof liver histopathological changes in the acute and repeated-dosefipronil groups was compared to respective control groups using aFisher's Exact Test.

The two cohorts for deltamethrin and permethrin that underwentEEG testing were combined to enable a dose–response (two doses) foreach chemical. The data were expressed as percent of their respectivecontrols and then combined, resulting in a larger control group. Becausethese results were percentages, the data were arcsin square-root

Table 2Analytes that showed no significant effects or were not detectable in sufficient number of sam

Analytes with no significant treatment effects

Adiponectin Interleukin-13Angiotensin-converting enzymea Interleukin-18Apolipoprotein A-1 Leukemia inhibitoComplement C3 α des arga Luteinizing hormFactor VII Macrophage inflamFibroblast growth factor-9 Macrophage inflamFibroblast growth factor basic MyeloperoxidaseGalanina MyoglobinGlucagon-like peptide-1, total Oncostatin-MGlutathione S-transferase-αa Peptide YYGranulocyte-macrophage colony-stimulating factor Plasminogen actiGrowth hormone Prolactina

Haptoglobin ResistinImmunoglobulin A Secretina

Insulin-like growth factor-1 Serum glutamic oInterferon-γb

Interleukin-1α T-cell specific proInterleukin-3 ThrombopoietinInterleukin-4b Tissue factora

Interleukin-5 (MSD) Tumor necrosis faInterleukin-7 Vascular cell adhe

a Not measured for all chemicals.b Both RBM and MSD assays.

transformed prior to analysis (Sokal and Rohlf, 1981). For sevenanalytes, the assay data were very different in the two cohorts, in thatmost or all of the data were below the LDD in one cohort and not theother. For these, the data could not be combined and no further analysiswas performed.

For descriptive purposes, all analytes having a significant (p b 0.05)treatment effect following ANOVA analyses were entered into adiscriminant analysis using JMP® (v10.02; SAS Institute Inc., 2012)to examine the multivariate canonical variables that summarizebetween-groupvariation. Because of the exploratory nature of this anal-ysis, and the small sample sizes, we did not use statistically-based vari-able selection (Maugis et al., 2011; Pacheco et al., 2006; Qiao et al.,2009) to optimize the canonical functions. However, as a preliminaryinvestigation into which variables might be useful for further confirma-tory analysis, a forward stepwise selection was used to determineall variables that contributed to the discriminant function with a p-value b 0.05. For most pesticides, all analytes that had a significantANOVA treatment effect were included in the final discriminantmodel. For carbaryl, however, due to the large number of endpointsthat was altered (coupled with the relatively small number of samples)only variables that had a p-value b 0.05 in the forward selection processare indicated in the discriminant analysis. It was hypothesized thattreatment could alter the relationships between the bioindicatorsresulting in different covariances between the groups. Therefore,we used a quadratic discriminant method that allows for differentbetween-group covariances.

Analyses of the Biocrates data followed approaches used for micro-array data (Shi, 2006). The concentrations were analyzed on the log2-scale. All calculations were performed in R version 2.15.2. To accountfor metabolite concentrations falling below their limit of detection, acensored Gaussian error distribution (implemented by the Surv() func-tion from the survival package) was used in our linear regression. Giventhe resulting set of parameter estimates and standard errors for eachmodel coefficient and each metabolite, the eBayes() function in thelimma package was used to stabilize the t-statistics by partially shrink-ing the standard errors toward a commonvalue. A Benjamini–Hochbergadjustment was then applied to the resulting p-values. Differentiallyexpressed metabolites were defined as those having an adjustedp-value less than 0.05 and an absolute fold-change greater than 1.3(meaning the absolute value of the regression coefficient was greaterthan log2(1.3)).

ples for analysis.

Analytes with concentrationsbelow LDD in all datasets

CD40CD40 ligand

ry factor Epidermal growth factoronea Endothelin-1matory protein-1α Fibrinogenmatory protein-1γ Glucagon

Interleukin-1βInterleukin-2Interleukin-5 (RBM)Interleukin-17A

vator inhibitor-1 Interleukin-12 subunit p70Matrix metalloproteinase-9Monocyte chemotactic protein-5

xaloacetic transaminasea (SGOT)

tein RANTES

ctor-αb

sion molecule-1 Interleukin-10

165V.C. Moser et al. / Toxicology and Applied Pharmacology 282 (2015) 161–174

Levels of serum T3, T4, and liver enzymes were analyzed using one-way ANOVA. Tukey's test was used to compare each dose group to con-trol in the event of a significant (p b 0.05) overall dose factor. Due tonon-normal distributions in some of the fipronil and fipronil sulfonetissue level data, non-parametric analysis of the overall data was con-ductedusing theKruskal–Wallis test, followed byone-way comparisonsbetween each dose group and control. For thyroid hormones andpesticide/metabolite levels in the fipronil studies, the data for eachsingle-dose group was compared to the group receiving the same dosefor 14 days (t-test for T3 and T4, Kruskal–Wallis for chemical levels)to test if the duration of dosing made significant differences.

Results

No observable signs of toxicity occurred in any of the single-dosestudies. With repeated fipronil, three of the high-dose rats showed ap-parent audiogenic seizureswhen dosed in a laboratorywith loud equip-ment: when the noise was reduced no more abnormal behaviorsoccurred. The high-dose rats also showed weight loss in the first fewdays, but this effect recovered after the first week of dosing (Freebornet al., in press).

Table 3Analytes showing significant differences by chemical treatment. Doses different from cHI = high dose, LO= lowdose. In some cases, the overall significant dose effectwasnot followedvalueswere too low to analyze, ‘–’means that therewasnooverall significance, andNT=not incluassays.

Class Biomarker Permethrin Deltamethrin I

Chemokine CXCL1: growth-regulated alphaprotein (GRO/KC)

– HI ↑ (178%) –

Chemokine CXCL2: macrophage inflammatoryprotein-2 (MIP-2)

– NDC –

Chemokine CXCL6: granulocyte chemotacticprotein-2 (GCP-2)

– HI ↓ (81%) L

Chemokine CXCL10: interferon gamma inducedprotein-10 (IP-10)

– – –

Chemokine CCL2: monocyte chemotacticprotein-1 (MCP-1)

– – –

Chemokine CCL4: macrophage inflammatoryprotein-1 beta (MIP-1β)

– – –

Chemokine CCL7: monocyte chemotacticprotein-3 (MCP-3)

– – –

Chemokine CCL11: eotaxin – – –

Chemokine CCL19: macrophage inflammatoryprotein-3 beta (MIP-3β)

– – N

Chemokine CCL22: macrophage derivedchemokine (MDC)

– – –

Chemokine XCL1: lymphotactin – – –

Interleukin Interleukin-6 (IL-6)Interleukin Interleukin-11 (IL-11) HI ↑ (264%) – –

Growth factor Macrophage colony-stimulatingfactor-1 (MCSF-1)

– – –

Growth factor Vascular endothelial growthfactor A (VEGF-A)

– HI ↓ (81%) –

Acute phase protein Serum amyloid P-component (SAP) – HI ↑ (121%) –

Acute phase protein C-reactive protein (CRP) – – –

Signal transduction Stem cell factor (SCF) LO ↓ (15%) – –

Matrixin inhibitor Tissue inhibitor ofmetalloproteinase-1 (TIMP-1)

– – –

Hemostasis Von Willebrand factor (VWF) – – –

Hormone Adrenocorticotropic hormone (ACTH) – – –

Hormone Cortisol – – –

Hormone Insulin – – –

Hormone Angiotensinogen – – NHormone Leptin – – –

Hormone Progesterone – – –

Hormone Testosterone, total – – H

a Significantly increased the number of samples above LDD, i.e., number of detectable value

RBM and MSD assays

With the Myriad RBM biomarker panels (RodentMAP® plusRatMetabolicMAP®; see Materials and methods), there were 13analytes with values that never, or almost never, provided data greaterthan the LDD and therefore did notmeet the stated criteria for statisticalanalyses. Therewere another 43 analytes thatmet criteria andwere sta-tistically analyzed, but showed no significant group differences in anystudy. These analytes, listed in Table 2, represented more than half ofthe total biomarker panel.

The number of analytes in these Myriad RBM assays that showedsignificant effects of chemical treatment ranged from two (permethrin)to 18 (carbaryl) (Table 3). There was no single analyte that was alteredby all chemicals. Thus, a different biological response profile existed foreach chemical, including the type I and II pyrethroids. The three bio-markers that were altered most often were changed in only about halfof the treatments: granulocyte chemotactic protein-2 (GCP-2), macro-phage colony-stimulating factor-1 (MCSF-1), and stem cell factor(SCF). Of all the 50 significant findings, post-hoc analyses for sixanalytes revealed no differences when comparing either dose group tocontrol. For the remaining 44 instances with significant pairwise find-ings, most showed differences between controls and either the high

ontrol and direction of change are indicated, with percent change in parentheses.by either dose groupbeingdifferent fromcontrol, indicatedby ‘NDC’. Empty cells indicate thatded in panel. All data fromRBMpanels exceptGRO/KC, forwhich thedata came fromtheMSD

midacloprid Carbaryl Triadimefon Fipronil acutehigh dose

Fipronil acutelow dose

Fipronilrepeated

– – LO ↓ (55%) – –

HI ↓ (60%) – – – –

O ↓ (87%) HI ↓ (72%) – NDC LO ↑ (116%) –

HI ↓ (66%) – – – LO, HI ↑(128%, 127%)

HI ↓ (69%) – – – –

– HI ↓ (69%) HI ↓ (77%) – –

HI ↓ (67%) – – – –

HI ↓ (64%) – – LO ↑ (123%)DC – – – – –

HI ↓ (61%) – – – –

HI ↓ (49%) – – – –

– HI ↑ a

– – – LO ↓ (70%) –

HI ↓ (73%) – LO, HI ↓(68%, 74%)

– LO, HI ↓(79%, 78%)

HI ↓ (55%) – – – –

– – – NDC –

– – – – HI ↑ (125%)HI ↓ (52%) – LO, HI ↓

(71%. 68%)– –

HI ↓ (60%) – – – HI ↑ (138%)

HI ↓ (73%) NDC – – –

HI ↑ (223%) HI ↑ (206%) – NT NT– – – – LO, HI ↑

(123%, 135%)HI ↓ (66%) LO ↓ (50%) – – –

DC HI ↓ (84%) – – NT NTLO, HI ↓(56%, 60%)

– – – –

LO ↑ (122%) – – – HI ↑ (147%)I ↓ (32%) – – HI ↓ (57%) – –

s, using categorical analysis (see Materials and methods).

166 V.C. Moser et al. / Toxicology and Applied Pharmacology 282 (2015) 161–174

dose group alone, or both dose groups. In only nine cases was the low,but not high, dose group different. For one cytokine (interleukin, or IL,6), the treatment groups showed a higher number of detectable valuescompared to control (which were below the LDD), and this wassignificant with categorical analyses. While the standard ANOVA analy-ses indicated significant treatment-related effects formany of these bio-markers individually, discriminant analyses presented a multivariatecomposite of these relationships across affected biomarkers and illus-trated that the patterns of responses could separate the treatmentgroups. Data from the fipronil studies are presented in detail here tohighlight these relationships as well as to compare the effects of singledoses (low and high) and repeated dosing.

Six biomarkers (GCP-2, growth-regulated alpha protein (GRO/KC),MCSF-1, macrophage inflammatory protein-1 beta (MIP-1β), SCF, andtestosterone) showed a significant overall dose effect following a singledose of 25 or 50 mg/kg fipronil (Fig. 1). These analytes were decreased,but formost therewas little or no difference in themagnitude of changeat the two doses. GCP-2 had a significant dose factor in the ANOVA, butneither groupwas different from control. For GRO/KC, only the lowdosewas significantly different from control. This information is alsodepicted in the discriminant analysis graph (Fig. 1), which shows over-lap in the 95% confidence limits of the two dose groups, both of whichare separated from control. The low degree of separation of the dosegroups suggests a relatively flat slope of the dose–response curve, orperhaps asymptotic changes over the 25 to 50 mg/kg dose range.

In contrast to the clear changes observedwith 25 or 50mg/kg singlefipronil doses, single exposure at the lower doses (5, 10 mg/kg)produced few changes. These data are provided in Fig. 2. While threebiomarkers showed a significant treatment effect (GCP-2, interleukin-11 (IL-11), and serumamyloid P-component (SAP)), two showed signif-icant differences from control only in the low dose (5 mg/kg), whereasthe third biomarker (SAP) showed an overall significant dose effect butno groups differed from control. The discriminant groupings (Fig. 2)show that the degree of separation for the low dose group is greaterthan the high dose. By evaluating the multivariate response profiles,the discriminant analyses showed that the high dose was also differentfrom control, providing support for the effects that were not evident inthe individual biomarker data. Overall, themagnitude of effects was notas great with these lower doses compared to higher doses shown inFig. 1.

A different pattern emerged when these low doses of fipronil wereadministered for 14 days. The significantly affected biomarkers areshown in Fig. 3. Dose–response was suggested by significant effects inthe high dose group only with interleukin-6 (IL-6), C-reactive protein(CRP), tissue inhibitor of metalloproteinase-1 (TIMP-1), and progester-one. Both dose groups were different from control for three markers(cortisol, MSCF-1, IP-10), but for the latter two the doses produced ap-proximately equal effects. Only the low dosewas significant for eotaxin.Discriminant analyses (Fig. 4) confirmed a change in multivariateresponse profiles (no overlap in treatment groups) as well as clear dif-ferences from control. Comparison of these data shows that while a sin-gle administration of either 5 or 10 mg/kg fipronil had minimal effect,repeated administration of the same doses produced more clear dose-responsive changes. Single administration of higher doses (25 or50 mg/kg fipronil) produced more changes than the lower doses, butMCSF-1 was the only biomarker that was altered by both the singlehigh-dose treatment and the repeated low-dose treatments.

Discriminant analyses for the other pesticide treatments are provid-ed in Supplemental Figs. 1–5. As with the fipronil examples, for all theseanalyses the loading into the canonical variables served to separate thetreatment groups, mostly without overlap. This was even the case withpermethrin, forwhich the fewest number of biomarkerswas altered. Onthe other hand, the control and low dose deltamethrin group had someoverlap. The scatter of individual data points by dose was the least withthe high dose of carbaryl, even though not all of the biomarker datacould be included in the discriminant analysis. These graphs can be

compared to the univariate analysis of biomarkers and magnitude ofchange presented in Table 3.

Biocrates assay

As with the analyte data, the metabolomic profiles and number ofmetabolites significantly altered by treatment differed across pesticides,as shown in Table 4. Overall, only about 20% of the measured metabo-lites met the statistical criteria for one or more treatments. The pyre-throids produced almost no changes (one amino acid each), while thehigh doses of fipronil produced the most. Dose-related changes weresuggested in almost all cases, since either only the high dose showedchanges, or else both low and high doses were significantly differentfrom control. Triadimefon also had few effects, and they were mostlychanges in the carnitine group. Imidacloprid altered mostly carnitinesand amino acids as well as a few biogenic amines. For fipronil, compar-ison across studies showed that few metabolites were altered by thelower doses (and mostly only at 10 mg/kg), whereas differences onmany of the same metabolites were produced by repeated exposureto both doses (5 and 10 mg/kg/d). The single high doses of fipronil al-tered even more metabolites, especially in the glycerophospholipidclass, which were not affected as much by the lower doses (single orrepeated).

Additional fipronil measures

Treatmentwith fipronil reduced serum thyroid hormone levels aftereither single or repeated treatments of 5 or 10mg/kg (Fig. 5). Serum T3levels were reduced in the high-dose group, and themagnitude of effect(~21%) was similar between single and repeated dosing regimens.Following a single dose, T4 was significantly decreased only by thehigh dose (~22%), whereaswith repeated dosing both treatment groupshad lower levels (~48–67% decrease). In both dose groups, the valuesafter repeated dosing were significantly lower compared to thosefollowing a single dose.

There were no significant changes in serum levels of liver injurymarkers (ALT, AST, LDH, or SDH) following either single or repeatedfipronil treatments (data not shown). Note that AST (also known asserum glutamic oxaloacetic transaminase, SGOT) was included in theRBM assays and again was not altered. Additionally, no treatment-related effects were observed on liver histopathology following eithersingle or repeated-dose fipronil treatments (data not shown).

Fipronil and/or fipronil sulfone were detected in the serum from alltreated animals (Fig. 6). A few control samples showed fipronil levelsgreater than the LOQ, and the source is not known; however, mediansacross all control groups were lower than the LOQ. Following a singledose, both fipronil and fipronil sulfone were detected at significantlyhigher levels than in control rats, and levels increased with dose. Withrepeated low-dose exposure, fipronil was almost undetectable, andwas therefore significantly lower than the levels after a single dose.On the other hand, levels of fipronil sulfone were significantly higherthan control and also higher than the levels measured after one dose.In other words, fipronil sulfone was found in higher concentrationscompared to fipronil in all treatments, and appeared to accumulatewith repeated dosing, whereas fipronil itself was evident after a singledose but was quickly converted to the sulfone metabolite (and was es-sentially undetectable) with repeated dosing. More information on anumber of fipronil metabolites is presented in McMahen et al.(submitted for publication).

Discussion

Findings from this series of experiments showed that exposure tovarious pesticides produces different patterns of activation and sup-pression of cytokines, hormones, andmetabolites at the timeof peakbe-havioral changes (determined by the literature and/or pilot studies).

Fig. 1.A: Bar graphs present groupmeans (± SEM) for each biomarker significantly affected by a single administration of high doses (25, 50mg/kg) of fipronil. * indicates treatment groupsignificantly different from control; for GCP-2 there was an overall effect of dose but no significant post-hoc comparison. Y-intercept shows the LDD for each assay. B: Multivariatediscriminant analyses of these effects showing the separation of individual rats along two canonical functions. Variables having a p-value b 0.05 in the forward selection process for thediscriminant analysis are indicated with dotted vector lines. The length of each vector reflects the relative degree of impact of a variable in the canonical functions.

167V.C. Moser et al. / Toxicology and Applied Pharmacology 282 (2015) 161–174

Distinctive profiles were also evident when comparing a range offipronil doses, and single compared to repeated dosing with fipronil.About half of the analytes were responsive to one or another treatment,but in the individual studies the number of significant changes rangedfrom about 2% to 22% of the total number of analytes. Likewise, pesti-cides differed in terms of metabolite changes, ranging from essentially

no changes (less than 1% measured) to about 9% of the totalmetabolomic analytes.

Neurophysiological measurements in these same animals also pro-duced different profiles of EEG changes (Freeborn et al., in press). Exci-tation and reductions in neuronal activity in specific EEG bands showeddifferent patterns for permethrin, deltamethrin, carbaryl, and acute and

Fig. 2. A: Bar graphs present group means (± SEM) for each biomarker significantly affected by a single administration of low doses (5, 10 mg/kg) of fipronil. * indicates treatment groupsignificantly different from control; for SAP there was an overall effect of dose but no significant post-hoc comparison. Y-intercept shows the LDD for each assay. B: Multivariate discriminantanalyses of these effects showing the separation of individual rats along two canonical functions. Variables having a p-value b 0.05 in the forward selection process for the discriminant analysisare indicated with dotted vector lines. The length of each vector reflects the relative degree of impact of a variable in the canonical functions.

168 V.C. Moser et al. / Toxicology and Applied Pharmacology 282 (2015) 161–174

Fig. 3. Bar graphs for each biomarker significantly affected by a 14-day dosingwith low doses (5, 10mg/kg) of fipronil. Each graph presents the groupmeans (± SEM), with the exceptionof IL-6, which is presented as the percent of each dose group showing values above the LDD. * indicates treatment group significantly different from control. Y-intercept shows the LDD forthe assays (for all but IL-6, which was 2.6 pg/ml).

169V.C. Moser et al. / Toxicology and Applied Pharmacology 282 (2015) 161–174

Fig. 4. Multivariate discriminant analyses of analytes following repeated dosing withfipronil showing the separation of individual rats along two canonical functions. Variableshaving a p-value b 0.05 in the forward selection process for the discriminant analysis areindicatedwith dotted vector lines. The length of each vector reflects the relative degree ofimpact of a variable in the canonical functions.

170 V.C. Moser et al. / Toxicology and Applied Pharmacology 282 (2015) 161–174

repeated fipronil, while imidacloprid was essentially inactive andtriadimefon showed only marginal effects. These distinct patterns inboth biomarkers and neurophysiology potentially reflect the differentmodes of action for these pesticides rather than generalized responses(e.g., stress), although additional work is needed to confirm pesticide-specific profiles.

Table 4Effects of pesticides on metabolomic panel. Doses significantly different from control (p b 0.05

Class Metabolite Permethrin Deltameth

Acylcarnitines CarnitineAcylcarnitines AcetylcarnitineAcylcarnitines PropionylcarnitineAcylcarnitines ButyrylcarnitineAcylcarnitines ValerylcarnitineAcylcarnitines Hexadecanoylcarnitine C16Acylcarnitines Octadecanoylcarnitine C18Acylcarnitines Octadecenoylcarnitine C18:1Acylcarnitines Octadecadienylcarnitine C18:2Amino acids Alanine HI ↓Amino acids ArginineAmino acids AsparagineAmino acids CitrullineAmino acids Glutamate HI ↑Amino acids MethionineAmino acids OrnithineAmino acids ProlineAmino acids ThreonineAmino acids TyrosineBiogenic amines PutrescineBiogenic amines SpermidineSugars HexoseGlycerophospholipids Lysophosphatidylcholine acyl C16:0Glycerophospholipids Lysophosphatidylcholine acyl C16:1Glycerophospholipids Lysophosphatidylcholine acyl C18.1Glycerophospholipids Lysophosphatidylcholine acyl C18.2Glycerophospholipids Lysophosphatidylcholine acyl C20.3Glycerophospholipids Lysophosphatidylcholine acyl C20.4Glycerophospholipids Phosphatidylcholine diacyl C38.3Glycerophospholipids Phosphatidylcholine diacyl C38.4Glycerophospholipids Phosphatidylcholine diacyl C40.5Glycerophospholipids Phosphatidylcholine diacyl C40.6Glycerophospholipids Phosphatidylcholine acyl-alkyl C38.4Glycerophospholipids Phosphatidylcholine acyl-alkyl C40.1Glycerophospholipids Phosphatidylcholine acyl-alkyl C40.4Glycerophospholipids Sphingomyeline C16:0Glycerophospholipids Sphingomyeline C20:2

While there were unique biomarker profiles for these pesticides,there were no clear clusterings for specific analytes. The chemokines,which are involved in stimulation of inflammation and host defense re-sponses (Graves and Jiang, 1995), were altered by the largest number ofpesticides. Carbaryl altered the greatest number of these cytokines, all ofwhich were decreased by the high dose with a similar magnitude of ef-fect (~49–72%). This class of cytokines was also decreased by delta-methrin, triadimefon, and fipronil (high dose acute and low doserepeated). On the other hand, increases in select chemokines werealso noted with deltamethrin and fipronil (low dose). Interleukin levelsare generally low (Tarrant, 2010), and indeed in these studies mostwere lower than the detectable limits. In only one instance (IL-6),were there significantly more detectable values in the treated groupcompared to control, which had no measurable values. Only IL-11 hadhigh enough baseline concentrations that both increases and decreaseswere detected. These chemokines have numerous and overlappingfunctions, but they can be considered collectively as markers of inflam-mation. Despite the number of altered cytokines, someof themost com-mon primary pro-inflammatory signals (e.g., IFN-γ and TNF-α) werenot altered. Basal levels of these cytokines are normally very low, andeven their stimulation may have produced levels lower than the assayLDD (Tarrant, 2010). In addition, given the temporal patterns tocytokine cascades, it is also possible to have missed the peak levels.

There are few reports in the literature with which to directlycompare these specific findings. Gordon and Ward (2009) also usedthe RBM Multi-analyte Profile to evaluate rats dosed with carbaryl;however, a higher dose (75 mg/kg) and longer time points (24 h, 7days) were used. They reported changes in eight analytes at 24 h. De-spite the experimental differences, it was of interest to note that four

and N1.3-fold change), and direction of change, are listed for each treatment.

rin Imidacloprid Triadimefon Fipronil acute highdose

Fipronil acute lowdose

Fipronilrepeated

LO, HI ↓ HI ↓LO, HI ↓ LO, HI ↓LO, HI ↓ LO, HI ↓LO, HI ↓ HI ↓ LO, HI ↓ HI ↓ HI ↓LO, HI ↓ HI ↓LO, HI ↑ HI ↑ HI ↑HI ↑LO, HI ↑HI ↑ LO ↑ HI ↓

HI ↓ HI ↓HI ↓HI ↓

LO, HI ↓HI ↓HI ↓HI ↓

LO, HI ↑HI ↓

HI ↓HI ↓ HI ↓

LO, HI ↑LO, HI ↓

LO ↓ HI ↓ LO, HI ↓LO, HI ↓ HI ↓ LO, HI ↓LO, HI ↓ HI ↓ LO, HI ↓LO, HI ↓ HI ↓ LO, HI ↓LO, HI ↓ LO, HI ↓ LO, HI ↓LO, HI ↑LO ↑LO, HI ↑LO, HI ↑ LO, HI ↑LO ↑

LO ↓LO ↑

LO ↑HI ↑

Fig. 5. Serum T3 and T4 levels (mean ± SEM) following acute and repeated dosing with fipronil (5, 10 mg/kg). * indicates treatment group significantly different from control for eachstudy; # indicates significant difference between effects after one dose compared to repeated doses. Y-intercept shows the LOQ for the assays.

171V.C. Moser et al. / Toxicology and Applied Pharmacology 282 (2015) 161–174

of those analytes (lymphotactin, vascular endothelial growth factor A(VEGF-A), IP-10, and von Willebrand factor (VWF)) were also alteredin the present study, albeit mostly in the opposite direction of change.

It is now known that neurotransmitter systems are directly involvedin immunomodulation, with findings of neurotransmitter synthesis,storage, and release capability as well as effector receptors (Levite,2008). The influence of cholinergic neurotransmission is multi-faceted,but in general stimulation of nicotinic and/or muscarinic receptors inimmune cells results in decreased pro-inflammatory factors (Nizri andBrenner, 2013). Cholinomimetic effects produced by the reversible ace-tylcholinesterase inhibitors (pyridostigmine, edrophium) suppressedpro-inflammatory cytokine production in response to treatment withimmune stimulation (Nizri et al., 2006). In agreement with those find-ings, in this study carbaryl, a carbamate acetylcholinesterase inhibitor,decreased a number of inflammatory chemokines, suggesting impairedimmune responsiveness or decreased inflammatory processes. Whilemost studies typically assess cholinergic influences in response to im-mune stimulation, the high acute dose of carbaryl (50–67% inhibitionof cholinesterase in brain, plasma, and erythrocytes; Freeborn et al., inpress) might have been sufficient to disrupt the normal balance.Imidacloprid also stimulates the cholinergic system through the nico-tinic receptor, and nicotine has also been reported to inhibit T-cell

proliferation. However, there were almost no chemokine effects in thisstudy. This lack of effects could be partially due to imidacloprid's lowpotency on vertebrate nicotinic receptors (as opposed to insect recep-tors) (Tomizawa and Casida, 2005), and/or a function of dose selection.Similarly, triadimefon (dopaminergic) decreased only one cytokine,which could suggest little inflammatory influences or simply selectionof doses below the effective range.

There is also considerable influence of the GABAergic system onimmune cells, mediated through actions on both GABA-A and GABA-Breceptors (Jin et al., 2013). Pharmacological stimulation of GABA recep-tors leads to decreased cytokine production (Bhat et al., 2010). Fipronil,an antagonist at the GABA-A receptor, could be expected to have pro-inflammatory influences, and this was the case with increasedchemokines following 14 days of exposure, but not with a single dose.Indeed, a few chemokines were decreased following an acute dose ofone of the higher doses (25 or 50 mg/kg), but this profile was mixedand did not show a dose–response with the lower single doses (5 or10 mg/kg). Repeated exposure to fipronil was the only treatment toincrease IL-6, a notably pleiotrophic cytokine that, in addition to its im-mune actions, also induces acute phase proteins, activates the hypotha-lamic–pituitary–adrenal (HPA) axis, and inhibits thyroid-stimulatinghormone secretion (Papanicolaou et al., 1998). These actions may be

Fig. 6. Serum fipronil and fipronil sulfone levels (median ± interquartile range) followingacute and repeated dosing with fipronil (5, 10 mg/kg). * indicates treatment group signif-icantly different from control; # indicates significant difference between effects after onedose compared to repeated doses. Y-intercept shows the LOQ for the assays.

172 V.C. Moser et al. / Toxicology and Applied Pharmacology 282 (2015) 161–174

related to other fipronil effects measured herein, including increasedCRP and cortisol, and decreased thyroid hormone levels, respectively.

In addition to various chemokines, two cytokines that are consideredgrowth factors were altered by a few pesticides. MCSF-1 and VEGF-Astimulate angiogenesis and osteoclastogenesis as well as promote variousinflammatory responses (Neufeld et al., 1999; Sweet and Hume, 2003).One or the other of these was decreased by deltamethrin, carbaryl,and high doses of fipronil (single dose of 75 mg/kg, repeated dosing at10mg/kg/d). Inflammatory response activationwas also suggested by in-creases in the acute phase proteins CRP and SAP by deltamethrin andrepeated-dose fipronil. Overall, these changes suggest dysregulation ofthe immune system produced to a varying degree by these pesticides.

A few proteins involved in cellular development and maintenance(SCF, TIMP-1, VWF) were mostly decreased (permethrin, carbaryl, andsingle high dose, 75 mg/kg, of fipronil) but TIMP-1 was also increased(repeated 10 mg/kg/d fipronil). Among several biological functions,SCF plays an especially important role in hematopoiesis (Broudy,

1997), TIMP-1 regulates extracellular matrix turnover (St-Pierre andPotworowski, 2000), and VWF is critical for hemostasis (Sadler, 1998).TIMP-1 is considered to be highly inducible (St-Pierre and Potworowski,2000), yet levels were only increased with repeated exposure tofipronil, and decreased by carbaryl.

Hormone levels were also modulated by a few of the pesticides.ACTH and cortisol were increased with repeated fipronil, carbaryl, andtriadimefon. This may reflect direct actions of the pesticides on theHPA axis, which is known to be stimulated by dopaminergic and cholin-ergic agents (e.g., triadimefon, carbaryl) and IL-6 (increased withfipronil) (Locatelli et al., 2010; Papanicolaou et al., 1998). In addition,fipronil (GABA antagonist) could be expected to increase hormone se-cretion by blocking the GABAergic inhibitory influences (Locatelliet al., 2010). On the other hand, this could reflect a stress response. Car-baryl and triadimefon had the shortest times of peak effect (30 min and1 h), and the ACTH increase following those single doses could reflectthe transient spike in response to cholinesterase inhibition or otherstress produced by treatment. Repeated fipronil produced cortisol in-creases, suggesting a more protracted stress response. Unfortunately,ACTH was not included in the RBM panel at the time of the repeatedfipronil experiment, preventing direct comparisons between studies.However, effort was taken to reduce stress of treatment, including twodays of vehicle dosing and habituation to the test procedures, beforethe actual test chemicals were administered.

Pesticide-induced increases in progesterone, and decreased testoster-one, suggest acute disturbances of reproductive endocrine homeostasis,for which there is some support in the literature. Fipronil induces testos-teronemetabolism (Das et al., 2006) and increases progesterone levels infemale rats (Ohi et al., 2004). These findings agree with the decreasedtestosterone (single, high doses) and increased progesterone (repeated-dosing) studies. Likewise, in this study imidacloprid decreased testoster-one, in agreement with an earlier study (Bal et al., 2012). No in vivostudies could be identified for carbaryl effects on progesterone withwhich to compare our findings, although it has been reported to inhibitprogesterone biosynthesis in vitro (Cheng et al., 2006). Homeostatic hor-mones (angiotensinogen, leptin, insulin) showed a dose–response onlywith carbaryl and triadimefon.

There were differences observed when the low doses of fipronil (5,10 mg/kg) were given once or repeatedly for 14 days; specifically,greater impact on thyroid hormones, more metabolic changes, and adifferent pattern of biomarkers for the repeated-dose regimen. Further-more, the biomarker/metabolomic pattern following repeated lowdoses did not appear qualitatively similar to the effects of single treat-ment with higher doses (25 or 50 mg/kg). These differences could bedue in part to physiological and compensatory responses, as well as dif-fering levels of fipronil and its metabolites, with these different dosingparadigms. Thyroid hormone effects of fipronil were evident in boththe lower-dose single and repeated-dosing studies. Others have also re-ported lowered thyroid hormones with repeated dosing (APVMA,2009; Leghait et al., 2009; Roques et al., 2012). However, we are notaware of reports of decreased thyroid hormones after a single treatmentwith fipronil. Liver toxicity has also been reported in rats and mice fol-lowing subchronic and chronic exposures (APVMA, 2009; De Oliveiraet al., 2012), but was not observed in the current short-term studies.Others have also shown that, with repeated dosing, the sulfone metab-olite accumulates to a much greater extent than does fipronil (Lacroixet al., 2010; Leghait et al., 2009), as was clearly observed in our study.The increased levels of the sulfone metabolite may be related to induc-tion of hepatic metabolizing enzymes (Das et al., 2006) as well as itshigher biological persistence (APVMA, 2009). The sulfone metaboliteis biologically active, and has been shown to have qualitatively similareffects as fipronil on decreasing T4 and hepatic enzyme induction(Roques et al., 2012), and on interactions with GABA receptors (Hainzlet al., 1998).

The metabolomic panel showed different profiles for the five pesti-cides, with considerable overlap in fipronil-induced changes across

173V.C. Moser et al. / Toxicology and Applied Pharmacology 282 (2015) 161–174

the single and repeated dosing studies.While imidacloprid had very feweffects on the biomarker panel, it showedmanymore changes inmetab-olite levels. Imidacloprid producedmost of the changes in carnitine andits acyl esters, followed by triadimefon and fipronil. In general, carnitineand most metabolites were decreased, while the hexadecanoyl- andoctadecenoylcarnitines were increased. As these metabolites are in-volved in fatty acid transport and oxidation, this could suggest alteredmitochondrial metabolism, energy production and/or acute oxidativestress (Reuter and Evans, 2012).

The amino acids reflect glucogenic and/or ketogenic energy produc-tion, protein and neurotransmitter turnover, as well as other metabolicand regulatory functions (Wu, 2009). At least some of these werealtered by treatment with all pesticides, especially imidacloprid, and afew amino acid-derived polyamines were also altered. Most of theselevels were decreased, suggesting increased turnover. While singletreatment with higher doses of fipronil (25, 50 mg/kg) produced somechanges, lower doses (5 or 10 mg/kg) only showed effects with repeat-ed treatment. These changes could reflect the acute effects of thesepesticides on various neurotransmitter systems and/or gluconeogenesis.

Only fipronil altered glycerophospholipid and sphingolipids levels,decreasing several lysophosphatidylcholines but increasing diacyl andacyl-alkyl phosphatidylcholines and sphingomyelines. In addition, afew sphingomyelines were increased by fipronil. These chemicals area major component of cellular membranes and lipoproteins and areinvolved in a number of biological functions (Cole et al., 2012;Hermansson et al., 2011). Neuronal membrane integrity is especiallyimportant for signal transduction, and changes in lipid compositionhave been suggested to play a role in neurodegeneration (Farooquiet al., 2000).

It should be noted that regulatory levels (reference doses, acceptabledaily intakes) for these pesticides are set at doses much lower thanthose used in this study, although spikes in exposures could occurwith high contamination or poisonings. The behaviorally active dosesused herein were meant to identify biological changes and biomarkerpatterns. Logical follow-up studies would assess lower doses as well aslonger exposure durations, to begin to approach environmentallyrelevant exposures.

In summary, these studies demonstrate differential profiles ofcirculating cytokines, proteins, hormones, and metabolites followingacute exposure to pesticides with different modes of neurologicalaction. For fipronil, these biomarker profiles differed between single-and repeated-dose exposures, and in addition repeated dosing withfipronil produced lower thyroid hormones and altered proportions ofthe parent and sulfone metabolite. Major biological processes linkedto these changes include inflammation, mitochondrial metabolism,membrane lipid dynamics, hormone homeostasis, and tissue mainte-nance. Changes in cytokines in particular could reflect actions of someof these pesticides on the different neurotransmitter systems involvedin neuroimmunomodulation. These patterns of biomarker changesmay provide a basis for gene mining and follow-up studies of adverseoutcomes that may or may not be related to the specific CNS targets.In addition, serum biomarkers hold promise for screening approaches;however, further delineation and validation of appropriate biomarkersare necessary. These findings may suggest directions for additionalstudies to understand biomarker patterns and pathways.

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.taap.2014.11.016.

Conflict of interest

The authors declare that there are no conflicts of interest.

Acknowledgments

The authors gratefully acknowledge the assistance of Ms JudyRichards for liver enzyme assays, and Drs Steve Edwards and Brian

Chorley for their review of this manuscript. This research was fundedby the intramural research program of the Office of Research andDevelopment, U.S. Environmental Protection Agency.

References

Australian Pesticides and Veterinary Medicines Authority (APVMA), 2009. Review ofMammalian Toxicology and Metabolism/Toxicokinetics of Fipronil. (http://apvma.gov.au/sites/default/files/fipronil-phase-5-prf-vol2-toxicology.pdf (accessed July 2014)).

Bal, R., Turk, G., Tuzcu, M., Yilmaz, O., Kuloglu, T., Gundogdu, R., Gür, S., Agca, A., Ulas, M.,Cambay, Z., Tuzcu, Z., Gencoglu, H., Guvenc, M., Ozsahin, A.D., Kocaman, N., Aslan, A.,Etem, E., 2012. Assessment of imidacloprid toxicity on reproductive organ system ofadult male rats. J. Environ. Sci. Health B 47, 434–444.

Bhat, R., Axtell, R., Mitra, A., Miranda, M., Lock, C., Tsien, R.W., Steinman, L., 2010. Inhibi-tory role for GABA in autoimmune inflammation. Proc. Natl. Acad. Sci. U. S. A. 107,2580–2585.

Biomarkers Definitions Working Group, 2001. Biomarkers and surrogate endpoints:preferred definitions and conceptual framework. Clin. Pharmacol. Ther. 69, 89–95.

Bloomquist, J.R., 2003. Chloride channels as tools for developing selective insecticides.Arch. Insect Biochem. Physiol. 54, 145–156.

Broudy, V.C., 1997. Stem cell factor and hematopoiesis. Blood 90, 1345–1364.Casida, J.E., 1963. Mode of action of carbamates. Ann. Rev. Entomol. 8, 39–58.Cheng, S., Chen, J., Qiu, Y., Hong, X., Xia, Y., Feng, T., Liu, J., Song, L., Zhang, Z., Wang, X.,

2006. Carbaryl inhibits basal and FSH-induced progesterone biosynthesis of primaryhuman granulosa-lutein cells. Toxicology 220, 37–45.

Cole, L.K., Vance, J.E., Vance, D.E., 2012. Phosphatidylcholine biosynthesis and lipoproteinmetabolism. Biochim. Biophys. Acta 1821, 754–761.

Collings, F.B., Vaidya, V.S., 2008. Novel technologies for the discovery and quantitation ofbiomarkers of toxicity. Toxicology 245, 167–174.

Crinnion, W.J., 2010. The CDC fourth national report on human exposure to environmen-tal chemicals: what it tells us about our toxic burden and how it assist environmentalmedicine physicians. Altern. Med. Rev. 15, 101–109.

Das, P.C., Cao, Y., Cherrington, N., Hodgson, E., Rose, R.L., 2006. Fipronil induces CYPisoforms and cytotoxicity in human hepatocytes. Chem. Biol. Interact. 164, 200–214.

De Oliveira, P.R., Bechara, G.H., Denardi, S.E., Oliveira, R.J., Mathias, M.I., 2012. Cytotoxicityof fipronil on mice liver cells. Microsc. Res. Tech. 75, 28–35.

Duramad, P., Holland, N.T., 2011. Biomarkers of immunotoxicity for environmental andpublic health research. Int. J. Environ. Res. Publ. Health 8, 1388–1401.

Duramad, P., Tager, I.B., Holland, N.T., 2007. Cytokines and other immunologicalbiomarkers in children's environmental health studies. Toxicol. Lett. 172, 48–59.

Farooqui, A.A., Horrocks, L.A., Farooqui, T., 2000. Glycerophospholipids in brain: their me-tabolism, incorporation into membranes, functions, and involvement in neurologicaldisorders. Chem. Phys. Lipids 106, 1–29.

Freeborn, D.L., McDaniel, K.L., Moser, V.C., Herr, D.W., 2014. Use of electroencephalogra-phy (EEG) to discriminate pesticides with different modes of action: differential ef-fects of carbaryl, permethrin, deltamethrin, fipronil, imidacloprid, and triadimefon.Toxicol. Appl. Pharmacol. (in press) http://dx.doi.org/10.1016/j.taap.2014.11.011.

Gordon, C.J., Ward, W.O., 2009. A multianalyte profile of serum proteins to screen fortoxicological effects of anticholinesterase insecticides in the rat. Neurotoxicology30, 377–381.

Graves, D.T., Jiang, Y., 1995. Chemokines, a family of chemotactic cytokines. Crit. Rev. OralBiol. Med. 6, 109–118.

Hainzl, D., Cole, L.M., Casida, J.E., 1998. Mechanisms for selective toxicity of fipronil insec-ticide and its sulfone metabolite and desulfinyl photoproduct. Chem. Res. Toxicol. 11,1529–1535.

Hermansson, M., Hokynar, K., Somerharju, P., 2011. Mechanisms of glycerophospholipidhomeostasis in mammalian cells. Prog. Lipid Res. 50, 240–257.

Ikaiddi, M.U., Akunne, H.C., Soliman, K.F., 1997. Behavioral and neurochemical effects ofacute and repeated administration of triadimefon in the male rat. Neurotoxicology18, 771–780.

Jin, Z., Mendu, S.K., Birnir, B., 2013. GABA is an effective immunomodulatory molecule.Amino Acids 45, 87–94.

Kramer, C.Y., 1956. Extension of multiple range tests to group means with unequal num-bers of replications. Biometrics 12, 307–310.

Lacroix, M.Z., Puel, S., Toutain, P.L., Viguie, C., 2010. Quantification of fipronil and its me-tabolite fipronil sulfone in rat plasma over a wide range of concentrations by LC/UV/MS. J. Chromatogr. B Anal. Technol. Biomed. Life Sci. 878, 1934–1938.

Leghait, J., Gayrard, V., Picard-Hagen, N., Camp, M., Perdu, E., Toutain, P.L., Viguie, C., 2009.Fipronil-induced disruption of thyroid function in rats is mediated by increased totaland free thyroxine clearances concomitantly to increased activity of hepatic enzymes.Toxicology 255, 38–44.

Levite, M., 2008. Neurotransmitters activate T-cells and elicit crucial functions via neuro-transmitter receptors. Curr. Opin. Pharmacol. 8, 460–471.

Locatelli, V., Bresciani, E., Tamiazzo, L., Torsello, A., 2010. Central nervous system-actingdrugs influencing hypothalamic–pituitary–adrenal axis function. Endocr. Dev. 17,108–120.

Manzo, L., Artigas, F., Martinez, E., Mutti, A., Bergamaschi, E., Nicotera, P., Tonini, M., Candura,S.M., Ray, D.E., Costa, L.G., 1996. Biochemical markers of neurotoxicity. A review ofmechanistic studies and applications. Hum. Exp. Toxicol. 15 (Suppl. 1), S20–S35.

Matsuda, K., Buckingham, S.D., Kleier, D., Rauh, J.J., Grauso, M., Sattelle, D.B., 2001.Neonicotinoids: insecticides acting on insect nicotinic acetylcholine receptors. TrendsPharmacol. Sci. 22, 573–580.

Maugis, C., Celeux, G., Martin-Magniette, M.-L., 2011. Variable selection in model-baseddiscriminant analysis. J. Multivar. Anal. 102, 1374–1387.

174 V.C. Moser et al. / Toxicology and Applied Pharmacology 282 (2015) 161–174

McMahen, R., Strynar, M.J., Dagnino, S., Herr, D.W., Moser, V.C., Garantziotis, S., Andersen,E.M., Lyke, D.F., McMillan, L., Lindstrom, A.B., 2014. Identification of fipronilmetabolites in rodents by time-of-flight mass spectrometry for application in ahuman exposure study. Environ. Intern. (Submitted for publication).

Narahashi, T., Zhao, X., Ikeda, T., Nagata, K., Yeh, J.Z., 2007. Differential actions of insecti-cides on target sites: basis for selective toxicity. Hum. Exp. Toxicol. 26, 361–366.

Neufeld, G., Cohen, T., Gengrinovitch, S., Poltorak, Z., 1999. Vascular endothelial growthfactor (VEGF) and its receptors. Faseb J. 13, 9–22.

Nizri, E., Brenner, T., 2013. Modulation of inflammatory pathways by the immunecholinergic system. Amino Acids 45, 73–85.

Nizri, E., Hamra-Amitay, Y., Sicsic, C., Lavon, I., Brenner, T., 2006. Anti-inflammatory prop-erties of cholinergic up-regulation: a new role for acetylcholinesterase inhibitors.Neuropharmacology 50, 540–547.

Ohi, M., Dalsenter, P.R., Andrade, A.J., Nascimento, A.J., 2004. Reproductive adverse effectsof fipronil in Wistar rats. Toxicol. Lett. 146, 121–127.

Pacheco, J., Casado, S., Nuñez, L., Gómez, O., 2006. Analysis of new variable selectionmethods for discriminant analysis. Comput. Stat. Data Anal. 51, 1463–1478.

Papanicolaou, D.A., Wilder, R.L., Manolagas, S.C., Chrousos, G.P., 1998. The pathophysio-logic roles of interleukin-6 in human disease. Ann. Intern. Med. 128, 127–137.

Qiao, Z., Zhou, L., Huang, J.Z., 2009. Sparse linear discriminant analysis with applications tohigh dimensional low sample size data. IAENG Int. J. Appl. Math. 39, 48–60.

Reuter, S.E., Evans, A.M., 2012. Carnitine and acylcarnitines: pharmacokinetic, pharmaco-logical and clinical aspects. Clin. Pharmacokinet. 51, 553–572.

Roques, B.B., Lacroix, M.Z., Puel, S., Gayrard, V., Picard-Hagen, N., Jouanin, I., Perdu, E.,Martin, P.G., Viguie, C., 2012. CYP450-dependent biotransformation of the insecticide

fipronil into fipronil sulfone can mediate fipronil-induced thyroid disruption in rats.Toxicol. Sci. 127, 29–41.

Sadler, J.E., 1998. Biochemistry and genetics of von Willebrand factor. Annu. Rev.Biochem. 67, 395–424.

SAS Institute Inc., 2004. SAS/STAT® 9.1. SAS Institute, Inc., Cary, NC.SAS Institute Inc., 2012. JMP® 10.02. SAS Institute, Inc., Cary, NC.Shi, L., 2006. Executive Summary: The MicroArray Quality Control (MAQC) Project.

(http://www.fda.gov/downloads/ScienceResearch/BioinformaticsTools/MicroarrayQualityControlProject/UCM132150.pdf (accessed July 2014)).

Sokal, R., Rohlf, J.F., 1981. Biometry: The Principles and Practice of Statistics in BiologicalResearch. 2nd ed. W.H. Freeman and Company, San Francisco, pp. 1–859.

St-Pierre, Y., Potworowski, E.F., 2000. T cell control of extracellular matrix degradation.Dev. Immunol. 7, 171–177.

Sweet, M.J., Hume, D.A., 2003. CSF-1 as a regulator of macrophage activation and immuneresponses. Arch. Immunol. Ther. Exp. (Warsz) 51, 169–177.

Tarrant, J.M., 2010. Blood cytokines as biomarkers of in vivo toxicity in preclinical safetyassessment: considerations for their use. Toxicol. Sci. 117, 4–16.

Tomizawa, M., Casida, J.E., 2005. Neonicotinoid insecticide toxicology: mechanisms ofselective action. Annu. Rev. Pharmacol. Toxicol. 45, 247–268.

Wiesinger, M., Mayer, B., Jennings, P., Lukas, A., 2012. Comparative analysis of perturbedmolecular pathways identified in in vitro and in vivo toxicology studies. Toxicol. inVitro 26, 956–962.

Wu, G., 2009. Amino acids: metabolism, functions, and nutrition. Amino Acids 37, 1–17.