Comparative transcriptomics implicates mechanisms of ... · transcriptome responses to embryonic...

Molecular Ecology (2010) 19, 5186–5203 doi: 10.1111/j.1365-294X.2010.04829.x

Comparative transcriptomics implicates mechanisms ofevolved pollution tolerance in a killifish population

A. WHITEHEAD,* D. A. TRIANT,† D. CHAMPLIN‡ and D. NACCI‡

*Department of Biological Sciences, 202 Life Sciences Building, Louisiana State University, Baton Rouge, LA 70803, USA,

†Department of Biochemistry and Molecular Genetics, University of Virginia, Charlottesville, VA 22908, USA, ‡Atlantic

Ecology Division, National Health and Environmental Effects Research Laboratory, Office of Research and Development, US

Environmental Protection Agency, 27 Tarzwell Drive, Narragansett, RI 02882, USA

Corresponde

E-mail: andre

Abstract

Wild populations of the killifish Fundulus heteroclitus resident in heavily contaminated

North American Atlantic coast estuaries have recently and independently evolved

dramatic, heritable, and adaptive pollution tolerance. We compared physiological and

transcriptome responses to embryonic polychlorinated biphenyl (PCB) exposures

between one tolerant population and a nearby sensitive population to gain insight into

genomic, physiological and biochemical mechanisms of evolved tolerance in killifish,

which are currently unknown. The PCB exposure concentrations at which developmental

toxicity emerged, the range of developmental abnormalities exhibited, and global as well

as specific gene expression patterns were profoundly different between populations. In

the sensitive population, PCB exposures produced dramatic, dose-dependent toxic

effects, concurrent with the alterations in the expression of many genes. For example,

PCB-mediated cardiovascular system failure was associated with the altered expression

of cardiomyocyte genes, consistent with sarcomere mis-assembly. In contrast, genome-

wide expression was comparatively refractory to PCB induction in the tolerant

population. Tolerance was associated with the global blockade of the aryl hydrocarbon

receptor (AHR) signalling pathway, the key mediator of PCB toxicity, in contrast to the

strong dose-dependent up-regulation of AHR pathway elements observed in the

sensitive population. Altered regulation of signalling pathways that cross-talk with

AHR was implicated as one candidate mechanism for the adaptive AHR signalling

repression and the pollution tolerance that it affords. In addition to revealing

mechanisms of PCB toxicity and tolerance, this study demonstrates the value of

comparative transcriptomics to explore molecular mechanisms of stress response and

evolved adaptive differences among wild populations.

Keywords: adaptation, comparative biology, ecotoxicology, natural selection and contemporary

evolution, transcriptomics

Received 12 May 2010; revision received 20 July 2010; accepted 23 July 2010

Introduction

Upon drastic or rapid environmental change, popula-

tion persistence may emerge if either inherent pheno-

typic plasticity enables physiological adjustment to new

conditions or natural selection sorts among genotypes

of varying fitness resulting in derived local adaptation.

nce: Andrew Whitehead, Fax: +225 578 2597;

[email protected]

Many estuarine habitats on the Atlantic coast of North

America have experienced recent (post-World War II),

dramatic and persistent change related to environmen-

tal releases of chemically stable, bioaccumulative and

toxic industrial pollutants. Populations of the killifish

Fundulus heteroclitus that persist in many of these toxic

industrially polluted habitats provide unique examples

of contemporary convergent local adaptation (Burnett

et al. 2007; Van Veld & Nacci 2008). Many populations

demonstrate inherited pollution tolerance, yet targeted

� 2010 Blackwell Publishing Ltd

TRANSCRIPTOMICS OF EVOLVED POLLUTION T OLERANCE 5187

mechanistic studies have failed to reveal the genetic,

physiological and biochemical alterations that support

tolerance in any population. Here, we compare tran-

scriptome responses to pollutant exposure in tolerant

and sensitive killifish populations to provide insight

into the genomic mechanisms enabling contemporary

adaptation in the wild.

Chemical pollution tolerance has been characterized

in F. heteroclitus populations inhabiting at least four dif-

ferent polluted sites in Atlantic coast estuaries, includ-

ing New Bedford Harbor, MA, Bridgeport, CT, Newark,

NJ, and the Elizabeth River near Norfolk, VA (Nacci

et al. 2010). Although the chemical compositions pollut-

ing these sites vary considerably, one unifying feature

is widespread contamination with a group of mechanis-

tically related pollutants: specifically those that act

partly or wholly through the aryl hydrocarbon receptor

(AHR), including some polyhalogenated aromatic

hydrocarbons (polychlorinated dibenzodioxins, or diox-

ins, and polychlorinated biphenyls, PCBs) and polycy-

clic aromatic hydrocarbons (PAHs). Here, we compared

physiological and transcriptomic responses of one of

these populations (Bridgeport, CT) with those of a

nearby reference population from a clean environment

(Flax Pond, NY) following exposure to the model

dioxin-like compound (DLC) and AHR agonist, PCB

congener 3,3¢4,4¢,5 pentachlorobyphenyl or PCB126.

Laboratory studies show that lethal concentrations of

PCB126 are two orders of magnitude higher for Bridge-

port than Flax Pond killifish in early life stages (Nacci

et al. 2010).

Killifish from tolerant populations also demonstrate

resistance to the sublethal effects of DLCs. DLCs are

teratogens that stereotypically induce morphological

and functional cardiovascular system abnormalities in

developing fish embryos (Antkiewicz et al. 2005) and in

developing birds and mammals (Walker & Catron 2000;

Thackaberry et al. 2005). Yet, killifish from New Bedford

and Newark are unaffected by exposure to DLCs at con-

centrations at least two orders of magnitude higher than

those that induce characteristic deformities and death in

embryos and larvae from sensitive populations (Prince

& Cooper 1995; Nacci et al. 1999). Dioxin-like com-

pounds are ligands for the AHR transcription factor,

and AHR signalling is the key mediator of this develop-

mental toxicity in fish embryos (Prasch et al. 2003, 2006;

Carney et al. 2004; Antkiewicz et al. 2006; Jonsson et al.

2007), although the precise biochemical pathway linking

AHR activation to developmental toxicity has yet to be

defined. Similarly, all tolerant killifish populations stud-

ied to date are also resistant to AHR-mediated cyto-

chrome P450 1A (CYP1A) induction (Nacci et al. 2010).

These data suggest a common evolved mechanism of

tolerance associated with disruption of AHR-mediated


signalling. Although polymorphisms in AHR can

account for species and strain-specific differences in

dioxin sensitivity in avian, mouse and rat models (Tho-

mas et al. 1972; Pohjanvirta et al. 1999; Head et al.

2008), preliminary population genetic and functional

studies have not isolated a single component of the

AHR pathway that can account for the profound toler-

ance in F. heteroclitus (Hahn et al. 2004, and M. Hahn,

personal communication of unpublished data).

The ecological relevance and fitness advantage of the

derived tolerance phenotype for polluted site residents

is clear, yet the genomic basis of this trait is unknown.

We adopt a comparative transcriptomics approach to

characterize evolved population differences in response

to PCB challenge across a wide range of doses. Individu-

als from the polluted and reference sites were main-

tained for up to two generations in the laboratory under

the same ‘common garden’ clean conditions. This was

important to minimize the likelihood that observed pop-

ulation differences are physiological (habitat induced)

and that detected population differences in response to

PCB challenge are primarily heritable and therefore

genetically based. Experiments were designed to expose

ecologically relevant population-by-environment inter-

actions with respect to toxicity and gene expression. The

transcriptomic status of developing embryos was char-

acterized for each individual within a treatment group

(rather than for pools of individuals). This enabled iden-

tification not only of group effects of PCB exposure and

population effects on gene expression, but also to link

gene expression to measures of individual-specific toxic-

ity. These experiments yield insights into mechanisms of

PCB toxicity and importantly into the potential mecha-

nisms of evolved pollution tolerance.

Methods

Animals

Source populations for this comparative experiment

were collected from Bridgeport, Connecticut (tolerant

population; 41.1470�N, 73.2189�W) and from Flax Pond,

Stony Brook, New York (sensitive population;

40.9637�N, 73.1342�W). Like many other urban harbours

of the northeastern United States, Bridgeport Harbor

sediment pollution increased dramatically throughout

the 20th century in relation to increasing human popu-

lation density and industrialization (http://www.

preserveamerica.gov/PAC-bridgeport.html). By the 1980s,

sediments and biota from the Bridgeport site were

highly contaminated with organic and inorganic chemi-

cals including metals (Cu, Cr, Zn, Pb, Ni, Cd, Hg),

PAHs, PCBs and dioxins (Rogerson et al. 1985) and

were shown to be toxic to marine organisms (e.g.,

5188 A. WH IT EH EAD E T A L.

Gardner et al. 1992). The Bridgeport fish collection site

is over three orders of magnitude more contaminated

with PCBs, alone, than the Flax Pond reference site

(Nacci et al. 2010), which has been used as a reference

site in other toxicological studies using killifish (e.g.,

Arzuaga & Elskus 2002).

To minimize habitat-induced differences in gene

expression and to isolate the heritable portion of popu-

lation-specific responses to PCB exposure, experiments

were conducted with fish that were maintained in the

laboratory under clean and identical conditions for one

or two generations. As described in greater detail else-

where (Nacci et al. 2002), adult killifish (�100–200 fish

per site) were collected using baited traps and trans-

ported alive to the U.S. EPA Atlantic Ecology Division,

Narragansett, RI. Fish were held in laboratory aquaria

with flowing, ambient temperature sea water for over

6 months or up to 2 years to minimize the transfer of

maternal contaminants to embryos, and then used as

breeding stock. These conditions provided seasonally

reproductive breeding stocks, and embryos were col-

lected following unstimulated spawning events.

Embryos were used in laboratory tests to characterize

sensitivity to PCB126, as described below, or allowed to

grow to maturity (at least 1 year) in the laboratory.

Experimental fish from the polluted site (Bridgeport,

CT) were two generations removed from field-collected

fish, and experimental fish from the clean reference site

(Flax Pond, Stony Brook, NY) were one generation

removed from the field.

PCB exposure experiments

Responsiveness of killifish embryos to the most toxic

dioxin-like PCB congener, PCB 126 (supplied by Accu-

Standard, New Haven, CT, USA), was evaluated using

laboratory test procedures developed to compare the

DLC sensitivity of U.S. Atlantic coastal killifish popula-

tions (e.g., Nacci et al. 2010). Briefly, 1-day post-fertil-

ization (dpf) embryos were individually transferred to

20 -mL glass vials containing sea water with either

0.01% acetone carrier (supplied by Sigma Chemical, St.

Louis, MO, USA), identified as level 0 for the purposes

of statistical analyses, or PCB126-acetone stocks at nom-

inal concentrations ranging in log increments from two

through 200 000 ng PCB126 ⁄ L, which were identified as

levels 1–6, respectively. Individual embryos (up to 20

per treatment) remained in sealed vials for 7 days, after

which embryos were transferred to vials containing

clean sea water until 10 days post-hatching. During

these 10 days, embryos were held in incubators at

23 �C, with 12:12 light cycle and observed daily for sur-

vival. At 10 dpf, embryos were rated for developmental

anomalies related to dioxin toxicity, then flash frozen in

liquid nitrogen and archived at )80 �C until transcrip-

tomic analysis. To assess hatching and survival effects

of PCB126, two replicate tests (similar and concurrent

with tests used for generating samples for microarray

analysis) were performed for each population; however,

rather than sacrificing biological samples at 10 dpf as

for samples for microarray analysis, they were main-

tained and observed until 7 days post-hatching

(<28 days in total). Survival up through 7 days post-

hatching was used to estimate concentrations lethal to

50% of the tested population (LC50), using models

described elsewhere (Nacci et al. 2010). General linear

modelling methods were used to compare treatment

effects on phenotypic rating, and relationships between

survival and phenotypic rating between populations

(using SAS software; SAS Inc., Cary, NC, USA).

Scoring of developmental abnormalities

The DLC-tolerant phenotype was defined operationally

by observing embryos microscopically at 10 dpf in late

organogenesis, or approximately stage 34 (Armstrong &

Child 1965). Because DLC-mediated developmental tox-

icity is not observable until early organogenesis (stage

30) and becomes more pronounced as development pro-

ceeds, we selected a mid-developmental stage to assess

embryonic phenotypes for tolerance: at earlier stages,

sensitive and tolerant embryos were indistinguishable,

and at later stages, some sensitive embryos had died

(D. Nacci, unpublished data). At 10 dpf, each embryo

was observed using a dissecting stereoscope (50–80·)

and scored for the presence of the abnormalities charac-

teristic of DLC toxicity (Antkiewicz et al. 2005). The

summed score of abnormalities, defined as the pheno-

typic abnormality (PA) rating, describes a gradient of

toxicity ranging from 0, reflecting no observable abnor-

malities, to 4, reflecting multiple co-occurring and often

increasingly severe abnormalities. Examples of some of

the developmental anomalies characteristic of dioxin-

like toxicity are shown in Fig. 1A–E and include those

associated with mild toxicity such as mild pericardial

oedema and those representing increasing circulatory

failure such as severe pericardial oedema, abnormal

heart size or chambering (e.g., a severely deformed

heart can appear chamberless, which is referred to as

‘tube heart’) and tail and head haemorrhaging. Other

anomalies such as small body ⁄ head size and other mis-

cellaneous abnormalities were also noted and were

included in the phenotypic abnormality rating.

Transcriptomics data collection

RNA was extracted and purified from frozen whole

embryos (n = 4 or 5 per treatment) using chaotropic


Bridgeport population

Ph

eno

typ

ic a

bn

orm

alit

y ra

tin

g

Ph

eno

typ

ic a

bn

orm

alit

y ra

tin

g

0 101 103 104 105 106102

(0) (1) (2) (3) (4) (5) (6)

0

20

40

60

80

100

PCB126, log ng/L(treatment level)

0

1

2

3

4

5

% s

urv

ival

up

to

7-d

ays

po

st-h

atch

Flax pond population

0

20

40

60

80

100

PCB126, ng/L (treatment level)

0

1

2

3

4

5

0 101 103 104 105 106102

(0) (1) (2) (3) (4) (5) (6)

% s

urv

ival

up

to

7-d

ays

po

st-h

atch

(a)

(f) (g)

(b) (c) (d) (e)

Fig. 1 Phenotypic effects of polychlorinated biphenyl (PCB)126 exposure in embryos 10 days post-fertilization. (a–e) Characteristic

embryonic deformities following exposure. (a) Unaffected embryo (phenotypic rating = 0); (b) low dioxin toxicity, some tail haemor-

rhage or some pericardial oedema, top arrow indicates slight tail haemorrhaging (one deformity, phenotypic rating = 1); (c) moderate

dioxin toxicity, pericardial oedema and tail fin ray haemorrhage (two deformities, phenotypic rating 2); (d) high dioxin toxicity,

extensive tail haemorrhage and extensive pericardial oedema and moderate body size (three deformities, phenotypic rating 3); (e)

severe dioxin anomalies, including extreme cardiac oedema or ‘tube heart’ in addition to other deformities (phenotypic rating = 4).

(f, g) Early life stage survival (circles, primary Y-axis) and developmental abnormalities (triangles, secondary Y-axis) in Flax Pond

population (f) and Bridgeport population (g) killifish exposed to PCB126 ⁄ acetone in sea water at these treatment levels: 0 (acetone

only, dose 0), 2 ng ⁄ L (dose 1), 20 ng ⁄ L (dose 2), 200 ng ⁄ L (dose 3), 2000 ng ⁄ L (dose 4), 20 000 ng ⁄ L (dose 5, and 200 000 ng ⁄ L (dose

6, Bridgeport population only). Fifty per cent lethal concentrations for PCB126 (LC50): Flax (F1), 59 ng ⁄ L; Bridgeport (F2), 9041 ng ⁄ L.


buffer and spin column technology (for protocol details,

see Triant & Whitehead 2009). High-quality total RNA

for each sample was verified by microfluidic gel electro-

phoresis (Experion; Bio-Rad Inc., Hercules, CA, USA)

and then antisense RNA (aRNA) prepared using the

MessageAmp II aRNA amplification kit (Applied Bio-

systems ⁄ Ambion, Austin, TX, USA). Samples were then

coupled with Alexa fluor� dyes (Alexa fluor 555 and

647; Molecular Probes, Inc.) and competitively hybrid-

ized to Agilent custom oligonucleotide microarrays

(Whitehead et al. 2010) and incubated at 60 �C for 18 h

in a SureHyb microarray hybridization chamber (Agi-

lent Technologies, Santa Clara, CA, USA). Arrays

included duplicate printings for each of 6800 putative

genes, plus control elements, and probes were designed

using OLIGOWIZ software (Wernersson et al. 2007) pri-

marily from F. heteroclitus ESTs (Paschall et al. 2004)

and other Fundulus transcript sequences available

through GenBank, with position preference towards the

3¢ end of target sequences. Four to five individual

embryos (biological replicates) within each population-


by-dose treatment were hybridized in a balanced loop

design (e.g., Oleksiak et al. 2002) in duplicate including

a dye-swap. Following hybridization, slides were

washed following the manufacturer’s protocols, slide

images were captured using the Packard Bioscience

ScanArray Express (PerkinElmer Inc., Waltham, MA,

USA), and spot intensities were assayed for each chan-

nel using ImaGene software (Biodiscovery Inc., El Seg-

undo, CA, USA).

Transcriptomics data analysis

Of the 6800 unique probes on the microarray, 962 spots

were excluded because they were either too bright in

any array (median spot intensity >65 000) or too faint

(median spot intensity <2 standard deviations above

background) in most (90%) arrays, leaving 5838 ele-

ments that were included in normalization and statisti-

cal analyses. Median raw spot intensity data were

log2 transformed, lowess-normalized to correct for

intensity bias (JMP-Genomics; SAS Institute Inc.), then


normalized using linear mixed models to account for

fixed (dye) effects and random (array) effects. Our

objective was to detect PCB-induced changes in gene

expression that differ between tolerant and sensitive

populations. To specifically isolate PCB-induced effects,

our experimental design compared PCB dose exposures

to vehicle (acetone) control exposures. We avoided

direct comparison of absolute gene expression levels

between populations in the control sample because it

was not possible to disentangle constitutive baseline dif-

ferences in expression among populations from a vehi-

cle exposure–induced difference between populations.

Accordingly, expression level for each gene and for

each sample was further normalized relative to the pop-

ulation-specific vehicle control average, so that average

gene expression for each population’s vehicle control

treatment was zero.

Normalized data were used for statistical analysis of

PCB-induced changes in gene expression using mixed

linear models with ‘population’ and ‘dose’ as main

effects, including an interaction term. Dye was mod-

elled as a fixed effect and array and replicate individu-

als within treatment were modelled as random effects,

and denominator degrees of freedom were 48 (6

doses · 2 populations · 5-1 individuals per treatment).

P-values were adjusted for multiple tests using Q-value

(Storey & Tibshirani 2003). The interaction term identi-

fied the genes with PCB-induced alterations in gene

expression that differed between populations (q < 0.1,

equivalent to P < 0.0032). As gene expression response

to dose may be nonlinear (Andersen et al. 2002; Calzo-

lai et al. 2007), mixed linear models may lack sensitivity

for detecting many genes with expression responses

that differ between populations. Accordingly, we used

spline-fitting models using the software EDGE (using the

natural cubic spline option with ‘population’ specified

as the class covariate; Storey et al. 2005) to further iden-

tify genes with different dose responses among popula-

tions (q < 0.1). A second mixed model was used to test

for the association between gene expression and devel-

opmental abnormality index scores, by specifying ‘pop-

ulation’ and ‘phenotype score’ (PA rating) as main

effect terms, including an interaction term that identi-

fied genes differentially associated with developmental

phenotype in different populations (q < 0.1). Note that a

fully specified model including main effect terms for

‘dose’, ‘population’ and ‘phenotype’ was not possible

because phenotype categories were not equally repre-

sented within each dose-by-population group. Treat-

ment means for all genes differentially expressed

among doses (‘dose’ term from mixed model signifi-

cant, q < 0.1) or with different dose responses among

populations (interaction term from mixed model signifi-

cant, q < 0.1; or EDGE analysis significant, q < 0.1) were

included for principal components analysis (PCA) per-

formed using MeV (Saeed et al. 2006; median centring

mode specified). Identification of co-regulated gene sets

was performed by hierarchical clustering based on

Euclidian distance and average linkage clustering. Gene

ontology enrichment analysis of co-regulated gene sets

was performed using the Functional Annotation Clus-

tering tool within the DAVID Bioinformatics Database

(Huang et al. 2009; http://david.abcc.ncifcrf.gov/home.

jsp), and inference of gene interaction networks was

performed using Ingenuity Pathway Analysis (Inge-

nuity Systems Inc, Redwood City, CA, USA) and

refined manually with input from recent literature.

Results and discussion

Developmental toxicity

This study was designed to characterize responses to

increasing concentrations of PCB126 of 10 dpf embry-

onic killifish from Flax and Bridgeport. Survival data

from replicate exposures, where embryos were

observed until 7 days post-hatching (or 28 days post-

fertilization, in the event of nonhatching) shown in

Fig. 1f,g, resulted in LC50 values of 59 and 9041 ng

PCB126 ⁄ L for Flax Pond F1 and Bridgeport F2 embryos,

respectively. This represents >150-fold difference in

chemical tolerance, consistent with a previously pub-

lished report (Nacci et al. 2010). In general, the occur-

rence of individuals with developmental anomalies and

the magnitude of individual PA ratings increased with

increasing PCB exposure in both populations (Fig. 1).

Phenotypic abnormality ratings were positively corre-

lated with dose (P < 0.0001), yet populations differed in

phenotypic responses (P = 0.0008), as they differed with

respect to survivorship responses (Fig. 1). Phenotypic

anomalies spanned the entire continuum of severity in

the sensitive population, whereas tolerant population

individuals did not express the highest phenotypic

abnormality ratings (at least as observed at 10 dpf),

even at doses that were lethal later in development

(Fig. 1). Analysis of variance indicated that phenotypic

abnormality rating was negatively correlated with

embryonic and larval survival (P = 0.0002), although

this relationship differed between populations

(P = 0.0263). In summary, groups from each population

differed in sensitivity to PCB126 measured both by

appearance of toxicity and by survivorship in develop-

ing embryos.

Dioxin-like compounds are extremely toxic to verte-

brates, and especially toxic to the early development of

many fish species including killifish (reviewed in Van

Veld & Nacci 2008). In developing fish embryos (Ant-

kiewicz et al. 2005), and also in birds (Walker & Catron



2000) and mammals (Thackaberry et al. 2005), toxicity is

largely mediated by disrupting the cardiovascular sys-

tem. Anomalies observed in developing killifish

exposed to DLCs include reduced heart size, oedema,

and haemorrhaging (Fig. 1a–e), similar to other fish

species (Antkiewicz et al. 2005). Multiple, concurrent

anomalies within individuals are consistent with cardio-

vascular failure and contribute to the increase in aver-

age PA ratings at exposure concentrations that are

ultimately lethal (Fig. 1). Despite a reputation for hardi-

ness (Burnett et al. 2007), killifish from unpolluted sites

such as the Flax population studied here are relatively

sensitive to DLCs compared to other fish species (Van

Veld & Nacci 2008). In contrast, embryos from the pop-

ulation resident to the highly polluted Bridgeport site

are orders of magnitude more resistant to DLC toxicity

compared to the population from the clean Flax site

(Fig. 1). Not only are many killifish populations from

polluted sites more than three orders of magnitude

more tolerant to DLCs than clean site populations (Nac-

ci et al. 2010), they are also more than two orders of

magnitude more tolerant than the most tolerant species

of fish reported so far (Van Veld & Nacci 2008). That is,

the DLC sensitivity range exhibited by populations of

F. heteroclitus far exceeds the sensitivity range of all fish

species.

General transcriptome response

Differences in biological responsiveness to PCB126

between populations were evident in comparisons of

transcriptome responses: the transcriptional response to

PCB exposure was much more dramatic in the sensitive

population than in the tolerant population (Fig. 2; a list

of genes with results from statistical tests is in Table S1).

Many transcripts were up- or down-regulated in the

sensitive population at each dose relative to the control

(931 transcripts @ q < 0.1, 1167 transcripts @ P < 0.05),

and the number of genes and magnitude of expression

response tended to increase with dose (Fig. 2a). At

higher (ultimately lethal) exposures where sampled

embryos were highly abnormal, expression response was

skewed towards down-regulation (rather than up-regu-

lation). Intriguingly, most of the putative transcripts that

were down-regulated upon PCB exposure could be

assigned homology to known genes in UniProt (80% of

transcripts with significant similarity; BLASTx threshold

E-value = 1e)5), yet identities for most of the PCB up-

regulated genes are unknown (only 40% of transcripts

with significant BLASTx similarity). PCA indicated that the

largest transcriptome changes occurred between mid-

range doses for the sensitive population (Fig. 2c).

In contrast, far fewer genes were regulated in

response to PCB exposure in the tolerant population (25


transcripts @ q < 0.1, 444 transcripts @ P < 0.05), and

the number of genes and magnitude of expression

response did not tend to increase with dose (Fig. 2b).

Furthermore, expression responses tended to be skewed

towards up-regulation at each dose relative to the con-

trol. In the tolerant population, most transcriptome vari-

ation among doses was accounted for by PC1 (81%)

(Fig. 2c). Little variation was partitioned between the

control and low doses up to dose 4 (especially along

PC1), and only at extreme (and ultimately lethal) doses

did transcriptome characteristics start to diverge from

control in the tolerant population. When dose responses

of each population were plotted together in principal

components space, the dramatic response of the sensi-

tive population relative to the minimal response of the

tolerant population is particularly apparent (Fig. 2c),

especially along PC1.

Genes down-regulated only in the sensitive popula-

tion were enriched for gene ontology terms contractile

fibre, response to wounding, angiogenesis, cell adhesion,

oxidative phosphorylation, glycolysis, calcium ion bind-

ing and protease inhibitor. Genes up-regulated only in

the sensitive population were enriched for gene ontol-

ogy terms cytochrome P450 and haeme binding. All

major co-regulated sets of genes, with associated gene

ontology enrichment categories, can be found in Fig. S1.

Gene expression is more predictive of individualphenotype than of dose, particularly in the tolerantpopulation

Dose accounted for the expression patterns of more

genes in sensitive than tolerant embryos, which is con-

sistent with differences in biological responsiveness to

PCB exposures between populations. In contrast, phe-

notype accounted for many gene expression patterns in

both the sensitive and tolerant populations. In the sensi-

tive population, grouping gene expression by pheno-

type or by dose implicates a largely overlapping set of

genes (Fig. 3a). Furthermore, the distributions of F-

ratios (which illustrate of the proportion of variation

partitioned among groups compared to the proportion

partitioned among individuals within groups), whether

grouping individuals by dose or phenotype, are largely

the same (P = 0.1, Kolmogorov–Smirnov test; Fig. 3b).

However, in the tolerant population, much more

individual variation in gene expression is partitioned

among phenotypes than among doses (Fig. 3b),

reflected by distributions of F-ratios that are signifi-

cantly different between dose and phenotype (P <

0.001, Kolmogorov–Smirnov test; Fig. 3b). This remains

true even for effects-matched doses relative to the con-

trol for each population (for doses 2 and 3 for Flax, and

doses 4 and 5 for Bridgeport) (data not shown). For

4

6

8

270 242 256 100 637 536524 346419 274

0-1 0-2 0-3 0-4 0-5 0-6

0

2

–4 –2 0 2 4 –4 –2 0 2 4 –4 –2 0 2 4 –4 –2 0 2 4

401 319

–4 –2 0 2 4

–4 –2 0 2 4 –4 –2 0 2 4 –4 –2 0 2 4 –4 –2 0 2 4 –4 –2 0 2 4–4 –2 0 2 4

335 164 761 680634 454543 364Sens

itive

8

0

2

4

6

288 211

33192 204

51

158 214

98

193 286

9678 89

62142 208

41

Tole

rant

00 0

4 5

PC1: 81%PC1: 66% PC1: 63%

S1S2

S3

S4 S5

1

23

4

6

51

2

T1

T2–6

Sensitivepopulation

PC2: 11%PC2: 16%

Sensitive populationPC2: 9%

Tolerant population

Tolerantpopulation

Sensitive + Tolerant

(a)

(b)

(c)

Fig. 2 Volcano plots and PCA illustrating general transcriptome response to polychlorinated biphenyl (PCB) exposures in tolerant

and sensitive populations. Panels (a) and (b) Volcano plots of gene expression change for each of five or six PCB exposure levels

compared to the control for the sensitive Flax Pond population (a) and the tolerant Bridgeport population (b). Genes that are up- or

down-regulated relative to the control are represented by right and left arms of each plot, respectively. X-axis is the mean log2 differ-

ence in gene expression in the PCB-exposed group relative to the control. Y-axis is )log10 P-value of PCB exposure effect from ANOVA.

Solid line indicates P = 0.05, dashed line indicates q = 0.1. Top pair of numbers beside right and left arms of volcano plot within each

figure indicates number of genes significantly up- or down-regulated, respectively, at q < 0.1. Bottom pair of numbers indicates num-

ber of genes significant at P < 0.05. Panel (c) First two principal components of the transcriptome response to PCB exposure for each

PCB exposure level (numbers associated with spots; ‘0’ is ‘control’) for the sensitive population (open circles), the tolerant population

(shaded circles) and both populations’ response plotted in the same PC space. Treatment means for all genes differentially expressed

among PCB exposure levels (‘PCB’ term from mixed model, q < 0.1) or with different dose responses among populations (interaction

term from mixed model significant, q < 0.1; or EDGE analysis significant, q < 0.1) were included for PCA.


example, at the most extreme dose (dose 6:

200 000 ng ⁄ L), a significant subset of genes is up-regu-

lated in the two ‘normal’ individuals, but not regulated

above control levels for the three developmentally

compromised individuals (Fig. 4, set 4; Fig. S1). We

interpret this pattern of expression as ‘protective’,

where up-regulation above control doses is correlated

with a normally developing embryo, but failure to up-

regulate is associated with developmentally compro-

mised individuals. Expression of these genes may be

associated with the evolved mechanism of pollution tol-

erance in the Bridgeport population (see further Results

and discussion below).

Gene expression patterns at the level of individuals,

beyond simple analysis of mean group effects, were

highly informative in this study in that many more

genes with PCB-induced toxicity-related patterns of

expression were identified especially in the tolerant

population. Extensive variation in gene expression is

partitioned among individuals within populations of

F. heteroclitus (Oleksiak et al. 2005; Whitehead & Craw-

ford 2005) and likely in most wild outbred species

(Whitehead & Crawford 2006). Much of this variation is

heritable and of physiological and evolutionary rele-

vance (Crawford & Oleksiak 2007). For example, gene

expression patterns distinguished individual killifish

with cardiac metabolism differences (Oleksiak et al.

2005) and distinguished individual honey bees with dif-

ferent behaviours (Whitfield et al. 2003). Linking the

phenotypic variation observed among individuals

within populations to individual variations in gene

expression is likely to enable more nuanced exploration


(a)

(b)

F-Phenotype

89

279499

Phenotype-T (504)Phenotype-S (368)

65220

5

Dose-S (931) Dose-T (25)

1600

2000

1600

2000 TolerantSensitive

F-PCB exposure800

1200

800

1200

Fre

qu

ency

0

400

0

400

F-ratio

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21F-ratio

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

Fig. 3 Gene expression is highly predictive of phenotype, especially in the tolerant population. (a) Venn diagrams represent sets of

genes significantly (q < 0.1) associated with polychlorinated biphenyl (PCB) exposure level (blue) or phenotype (red) for sensitive (S)

population (solid fill) and tolerant (T) population (grid fill). (b) Curves are smoothed frequency histograms of ANOVA F-ratios for PCB

exposure term (blue) and for phenotype term (red) for each of sensitive and tolerant populations. Exposure and phenotype distribu-

tions are not different for the sensitive population, but these distributions are significantly different for the tolerant population

(P < 0.001, Kolmogorov–Smirnov test).


of the functional links between genotype and pheno-

type, especially for studies in wild species that are

emerging as important models for ecological and evolu-

tionary genomics research.

Cardiovascular gene expression is associated withtoxicity in the sensitive population only

Dioxin-like compounds are well-known developmental

toxicants (Theobald et al. 2003) where exposure stereo-

typically causes cardiovascular system failures, espe-

cially in fish where exposure causes pericardial

oedema, reduction in cardiomyocyte proliferation, heart

size reduction and atrial elongation (Guiney et al. 2000;

Antkiewicz et al. 2005). Consistent with the occurrence

of these abnormalities (Fig. 1), a large set of co-regu-

lated genes were down-regulated upon PCB exposure

in the sensitive population but did not change with

exposure in the tolerant population (Fig. 4, set 1), and

this set was enriched for genes associated with gene

ontology categories ‘muscle contraction’, ‘sarcomere’

and ‘actin cytoskeleton’ (P < 0.01, FDR <0.1). Indeed,

many genes showing this pattern of expression form an

interaction network, are localized to the sarcomere, are

involved in angiogenesis or are known regulators of

myogenesis (Fig. 5, network model). Mutants or mis-

regulation of many of these genes has been associated


with various cardiomyopathies (Fig. 5—blue spots),

including for FHL1 (Knoblauch et al. 2010), FHL2

(Arimura et al. 2007), MLP (Knoll et al. 2002; Geier

et al. 2003), MYH6 (Niimura et al. 2002), MYBPC (Mar-

ian & Roberts 2001), MYL3 (Olson et al. 2002), MYOZ2

(Chen et al. 2008b), PDLIM7 (Camarata et al. 2010),

TMOD1 (Fritz-Six et al. 2003), TnT (Marian & Roberts

1995; Sehnert et al. 2002), TPM3 (Marian & Roberts

1995) and TTN (Itoh-Satoh et al. 2002; Carmignac et al.

2007; Seeley et al. 2007). Fly homologs for several of

these genes have also been implicated by RNA knock-

down as necessary for normal sarcomere assembly

(Fig. 5—green spots; Schnorrer et al. 2010). Other

intriguing genes within this co-expressed group include

basigin (EMMPRIN) and blood vessel epicardial sub-

stance (BVES). Knockdown of basigin, which is impor-

tant for intercellular recognition, caused sarcomere

disassembly in Drosophila (Schnorrer et al. 2010). BVES

is known to be expressed in cardiac tissues especially

during development, and expression knockdown is

associated with defects in muscle repair (Andree et al.

2002), wound healing (Ripley et al. 2004) and embry-

onic development (Lin et al. 2007), and expression of

mutant BVES is associated with inhibition of myogene-

sis (Smith et al. 2008). This regulation of cardiovascular

system genes is consistent with the stereotypical devel-

opmental consequences of dioxin toxicity and with how

Fig. 4 Scores of phenotypic developmental abnormalities for each embryo (bar graph) associated with gene expression responses of

individuals (heat maps). Bar graph illustrates phenotypic abnormality score for each individual from each treatment group. Treat-

ment groups are clustered by alternate vertical panels of white and grey shading, and the identifier at the top of each panel indicates

treatment group [‘S’ and ‘T’ represent sensitive and tolerant populations, respectively, and numbers 0–6 represent polychlorinated

biphenyl (PCB) exposure level]. X-axis identifiers of bar graph represent replicate individual within each treatment group. Increasing

numbers along bar graph Y-axis represent increasing severity of deformity (see Methods). Heat maps illustrate expression level for

each individual (columns) for specific genes (rows) identified on the right side of each heat map set. Colour of each cell represents

expression level: yellow and blue indicate up- and down-regulation, respectively, relative to the mean of control group expression

level (black) calculated separately for each population. Sets 1 and 4 were defined by hierarchical clustering, and sets 2 and 3 were

defined by network interaction. Intensity is scaled by log2 fold change relative to control for each heat map set (see scale legend

below each set). Red line graph between heat map sets 3 and 4 represents expression level of nitric oxide synthase (NOS) super-

imposed on the bar graph of individual deformities. The secondary Y-axis represents log2 fold expression level relative to average

control group expression for each population. Gene names are as follows: AHNAK, neuroblast differentiation-associated protein

AHNAK; AHR1, aryl hydrocarbon receptor 1; AHR2, aryl hydrocarbon receptor 2; ANF, atrial natriuretic factor; ANGPTL4, angio-

poietin-related protein 4; ARID1A, AT-rich interactive domain-containing protein 1A; ARNT2, aryl hydrocarbon receptor nuclear

translocator 2; ATP2A1, sarcoplasmic ⁄ endoplasmic reticulum calcium ATPase 1; B3GNT5, beta-1,3-N-acetylglucosaminyltransferase

5; BVES, blood vessel epicardial substance; CAPZA2, F-actin-capping protein subunit alpha-2; CD9, CD9 antigen; CYB5, Cytochrome

b5; CYP1A1, cytochrome P450 1A1; CYP1B1, cytochrome P450 1B1; CYR61, cysteine-rich angiogenic inducer 61; EMMPRIN, basigin;

FHL1, four and a half LIM domains protein 1; FHL2, four and a half LIM domains protein 2; GCH1, GTP cyclohydrolase 1; HIF1A,

hypoxia-inducible factor 1 alpha; HSP90AA1, heat shock protein 90 alpha 1; IGFBP1, insulin-like growth factor–binding protein 1;

IGFN1 (probes )1776 and )5607), immunoglobulin-like and fibronectin type III domain-containing protein 1; JUN, proto-oncogene

c-jun; MLP, muscle LIM protein; Murc, muscle-related coiled-coil protein; MURF2, muscle-specific RING finger protein 2; MYBPC,

myosin-binding protein C; MYH10, myosin heavy chain 10; MYH6, myosin heavy chain 6; MYL3, myosin light chain 3; MYL4, myo-

sin light chain 4; MYOZ2, myozenin-2; NOS, putative neuronal nitric oxide synthase; NPPB, brain natriuretic peptide; PDLIM3, PDZ

and LIM domain protein 3; PDLIM7, PDZ and LIM domain protein 7; REN, renin; S100-A11, S100 calcium-binding protein A11;

TMOD1, tropomodulin 1; TnT, troponin T; TPM3, tropomyosin alpha-3 chain; TTN (probes )1476, )8099, )9045), titin; TTN-10541,

titin a; UDPGT, UDP-glucuronosyltransferase; WISP1, WNT1-inducible-signalling pathway protein 1.


this toxicity is clearly differentially manifest among tol-

erant and sensitive populations of F. heteroclitus.

Altered transcript abundance for this group of genes

could be a cause or a consequence of developmental

abnormalities. Low abundance could be a consequence of

the developmental phenotype, where proteins preferen-

tially expressed in the heart appear in lower abundance

mainly because cardiac tissue contributes less to the



overall tissue composition of the embryo in develop-

mentally compromised individuals; in severely deformed

embryos (developmental index >3), the heart is small. If

low abundance of these transcripts was a consequence

of developmental abnormality, one would expect low

abundance to be confined to individuals exhibiting

reduced heart size. This is not observed in our data set.

Alternatively, altered expression of these genes could

contribute to causing the observed developmental abnor-

malities. Many of these genes are down-regulated in

individuals from the sensitive population at the lowest

doses, even in individuals that do not yet exhibit signs

of cardiac abnormality. Intriguingly, these genes do not

appear down-regulated in those tolerant population

individuals that exhibit cardiovascular abnormalities at

extreme doses (Fig. 4, set 1). These data indicate that

mis-regulation of sarcomere development is a compo-

nent of the mechanism leading to PCB-induced cardiac

developmental abnormalities in the sensitive population

but that similar cardiac abnormalities observed at

extreme PCB doses in the tolerant population emerge

from other mechanisms.

Expression of some structural cardiac sarcomere pro-

teins is interdependent, where alteration in the expres-

sion of one protein affects the expression of others

(Sehnert et al. 2002); this mis-regulation can result in

altered sarcomere assembly during development and

manifest as cardiac dysfunction in zebrafish. Although

the data presented here strengthen the evidence that

disrupted sarcomere assembly is mechanistically related

to DLC-induced cardiac toxicity in developing fish

(Handley-Goldstone et al. 2005), the intervening links

between AHR activation by DLCs and regulation of

myocyte protein expression remain to be determined

(though see recent progress by Carney et al. 2006; Chen

et al. 2008a).

Global disruption of AHR pathway in the tolerantpopulation

The AHR signal transduction pathway is activated

upon binding of the DLC agonist to the cytosolic

HSP90-associated AHR. The ligand-bound complex

translocates to the nucleus where it dissociates from

HSP90 and dimerizes with aryl hydrocarbon receptor

nuclear translocator (ARNT), then binds to promoter

elements to initiate transcription of a battery of genes,

and transcription is feedback-regulated by aryl hydro-

carbon receptor regulator (AHRR) (Rowlands & Gu-

stafsson 1997; Hahn et al. 2009). The AHR signalling

pathway is the key mediator of the developmental

toxicity of DLCs; morpholino knockdowns for AHR2

(Prasch et al. 2003; Carney et al. 2004; Antkiewicz et al.

2006) and ARNT1 (Antkiewicz et al. 2006; Prasch et al.


2006) are protective of dioxin-induced developmental

toxicity in zebrafish embryos, and mutant AHR is pro-

tective in mice (Bunger et al. 2008). AHR2 knockdown

also protects against embryo toxicity for other AHR

ligands including PCB126 (Jonsson et al. 2007).

One might propose three broad categories of mecha-

nisms whereby tolerance to dioxins is derived in F. het-

eroclitus populations such as Bridgeport. First, the AHR

signalling pathway may be generally blocked, thereby

protecting cells from diverse downstream transcrip-

tional and biochemical responses to signalling that initi-

ate and propagate toxicity. Second, the AHR signalling

pathway may be generally functional, but the AHR-

mediated signal to the subset of AHR targets that

directly mediate toxicity may be selectively blocked.

Third, the AHR signalling pathway could be entirely

functional, but the cells protected by downstream signal

disruption or efficient prevention or repair of responses

that normally result in developmental abnormalities.

Our data best support the first scenario, in that the

target genes for direct AHR activation are stereotypi-

cally activated in the sensitive population, but not acti-

vated in the tolerant population (Fig. 4, set 2). In the

tolerant Bridgeport population, the stereotypical AHR-

regulated genes appear up-regulated, if at all, only at

extreme doses. For example, the following are genes

known to be regulated directly by the AHR mechanism,

in that expression of these genes is consistently up-regu-

lated by AHR agonists in many species and many have

been shown to contain AHR-binding elements in

upstream promoters: CYP1A (Nebert et al. 2004), CYP1B

(Nebert et al. 2004), UDPGT (Nebert et al. 2000), CYB5

(Andreasen et al. 2006; Franc et al. 2008), IGFBP-1

(Marchand et al. 2005), S100A4 (Cao et al. 2003) and

JUN (Weiss et al. 2005). These genes are significantly

up-regulated in the sensitive population, most even at

the lowest doses. In contrast, in the tolerant population,

these genes are not responsive to PCB exposure, indicat-

ing that AHR signalling is blocked. For many of these

genes, including the well-characterized AHR-mediated

DLC responsive genes (e.g., the CYP450s and UDPGT),

up-regulation in the tolerant population is only

observed at the most extreme doses. Similar to the mini-

mal transcriptional response in the tolerant Bridgeport

killifish (Fig. 2), Ahr-null mice demonstrated a minimal

transcriptional response to dioxin exposure relative to

mice with intact AHR signalling (Tijet et al. 2006).

Population differences in the expression of genes that

directly mediate AHR signalling were subtle (Fig. 4, set

3) and do not appear to implicate differential regulation

of these genes as part of the adaptive mechanism.

Expression of AHR1, the AHR paralog that does not

appear to bind dioxin in zebrafish (Karchner et al. 2005),

was not associated with PCB exposure in either popula-


tion, nor was expression of ARNT2. Expression of AHR2,

the AHR paralog that does mediate DLC toxicity in ze-

brafish (Prasch et al. 2003; Carney et al. 2004; Antkiewicz

et al. 2006; Jonsson et al. 2007), was subtly increased in

the tolerant population (P = 0.003, q = 0.19), but not in

the sensitive population. The AHR repressor can regu-

late and suppress AHR signalling (Hahn et al. 2009), but

hybridization to probes for this gene did not occur above

background on our microarray. Expression of these same

genes (AHR, ARNT, AHRR) did not explain repression of

AHR signalling in Elizabeth River and New Bedford

Harbor tolerant F. heteroclitus populations (Powell et al.

2000; Karchner et al. 2002; Meyer et al. 2003), though

experiments in the Newark Bay tolerant population find

lower levels of AHR protein in livers of adult fish com-

pared to sensitive fish (Arzuaga & Elskus 2002). Expres-

sion of HSP90, the protein that stabilizes cytosolic AHR,

was slightly decreased in the tolerant population

(P = 0.01, q = 0.29), but not in the sensitive population

relative to control exposure levels. This is intriguing

given that HSP90 can act as an evolutionary capacitor

where altered regulation can expose previously hidden

phenotypic variation in populations to natural selection

(Rutherford & Lindquist 1998). It is not clear whether the

subtle increase in AHR2 and decrease in HSP90 expres-

sion we observe in the Bridgeport population may be

functionally associated with resistance to PCBs, or

whether these differences in expression may be a conse-

quence of some feedback regulation from AHR signal

impairment achieved by other means.

It is well known that derived tolerant populations of

F. heteroclitus tend to be refractory to CYP1A induction

by DLCs (Nacci et al. 1999, 2010; Bello et al. 2001;

Meyer et al. 2003). However, this is the first study to

indicate that most known targets of the AHR gene bat-

tery are refractory to induction at least in one tolerant

(Bridgeport) killifish population. These data indicate

that the mechanism of evolved tolerance in this popula-

tion is associated with generalized blockade of AHR-

mediated signalling, and only at the most extreme

doses does this blockade become leaky resulting in ste-

reotypical AHR-mediated gene expression responses

and developmental abnormalities, which are observed

at much lower doses (more than two orders of magni-

tude lower) in the sensitive population.

The AHR signalling pathway provides many candi-

date molecular targets to effect adaptive signal impair-

ment. Whereas mammals have one AHR gene, teleost

fishes including F. heteroclitus have at least two func-

tional copies of the AHR gene designated AHR1B and

AHR2 (Hahn et al. 1997; Karchner et al. 2005), and these

paralogs exhibit developmental stage–specific and tis-

sue-specific expression (Karchner et al. 1999; Powell

et al. 2000) and diversified function; notably, in zebra-

fish, AHR2 appears to mediate DLC toxicity (Prasch

et al. 2003; Carney et al. 2004; Antkiewicz et al. 2006;

Jonsson et al. 2007), whereas AHR1B is not dioxin

inducible (Karchner et al. 2005). Similarly, two paralogs

of the AHR nuclear translocator (ARNT) differentially

mediate dioxin-induced toxicity in zebrafish embryos

(Antkiewicz et al. 2006), and paralogs of the aryl hydro-

carbon receptor repressor (AHRR) differentially regulate

AHR signalling (Jenny et al. 2009).

Although the physiological role of AHR signalling is

often considered in the context of defence against xeno-

biotics, a physiological role in mediating embryonic

development has also been implicated by studies using

AHR knockout mice (Fernandez-Salguero et al. 1995;

Schmidt et al. 1996). Because tolerant killifish develop

normally, we can speculate that the AHR may act

through different signalling pathways in development

than in xenobiotic responses. Alternately, functional

redundancy facilitated by duplicate copies of many of

the proteins involved in AHR signalling in teleost fishes

may enable adaptive blockage of DLC-induced signal-

ling while retaining AHR signalling necessary for devel-

opment in tolerant F. heteroclitus killifish.

Polymorphisms in AHR can account for species and

strain-specific differences in dioxin sensitivity in some

avian, mouse and rat models (Thomas et al. 1972;

Pohjanvirta et al. 1999; Head et al. 2008) but not all

(Kawakami et al. 2006). Although population genetic

analyses have revealed differences in haplotype frequen-

cies between the New Bedford and a nearby sensitive

killifish population, polymorphisms for AHR1 (Hahn

et al. 2004) (or AHR2 or ARNT, M. Hahn, personal com-

munication) are not fixed between populations and vari-

ants do not differ functionally when assessed using in

vitro mammalian systems. However, absent tests of func-

tionality in vivo, the candidacy of these genes to explain

DLC tolerance cannot be definitively ruled out. Alterna-

tively, other AHR-interacting genes [such as AHRR,

HSP90 and AIP (Bell & Poland 2000)], and genes in sepa-

rate but interacting signalling pathways [such as hypoxia

signalling (Prasch et al. 2004); see discussion below],

remain as plausible targets for selective modification of

AHR signalling in derived tolerant populations.

Genes ‘protective’ in the tolerant population only:potential mechanisms of adaptive AHR signaldisruption

Our data indicate that the mechanism of evolved toler-

ance to PCBs in the Bridgeport killifish population

involves generalized blockage of AHR-mediated signal-

ling. This blockage could emerge from either altera-

tions in the proteins that are directly involved in AHR

signalling (such as AHR, ARNT, AHRR, HSP90, AIP) or



AHR signalling interference mediated by other path-

ways that cross-talk with the AHR pathway. Differences

in nitric oxide synthase gene expression between toler-

ant Bridgeport and sensitive Flax Pond populations are

suggestive of the latter scenario.

Nitric oxide synthase (NOS) exhibits a ‘protective’ pat-

tern of gene expression in the tolerant population, where

the gene is up-regulated relative to control at the lowest

PCB doses but not in the few individuals that exhibit

PCB-induced developmental abnormalities at extreme

doses (Fig. 4—line graph, and top line of heat map set

4). In contrast to the tolerant population, NOS is only

slightly up-regulated with respect to dose in the sensi-

tive population. The potential role of NOS in the mecha-

nism underlying evolved tolerance in the Bridgeport

population derives from four sources. First, the pattern

of expression suggests a protective role in the tolerant

population. Second, nitric oxide is a signalling molecule

that is a key mediator of cardiac remodelling (Massion

& Balligand 2007), and disruption of cardiovascular

system development is evident in our data and is the

stereotypical toxic response to DLCs in developing fish

embryos (Antkiewicz et al. 2005). Third, pre-exposure to

hypoxia is protective of dioxin toxicity in developing

fish embryos (Prasch et al. 2004), and nitric oxide

synthesis is induced by hypoxia signalling in fish

(McNeill & Perry 2006) (Fig. 6). Finally, Kim & Sheen

(2000) provide evidence that nitric oxide blocks dioxin-

induced AHR-mediated CYP1A promoter activity

(Fig. 6). Although dioxin-induced developmental toxic-

ity appears independent of CYP1A (Carney et al. 2004),

perhaps the mechanism whereby nitric oxide blocks

AHR regulation of CYP1A (which is currently unknown)

is generally effective at blocking AHR induction of other

genes in the AHR battery, including those through

which developmental toxicity is typically mediated

(which are also currently unknown). NOS up-regulation

is possibly a consequence (rather than a cause) of AHR

MYL3

MurcCD9ANFNPPB

FHL2

MYBPC

RENAHNAKWISP1

TGFB1

MYH6

ANGPTL4 CYR61

FHL1

MYL4

MYL3

MlcPDLIM7

Fig. 5 Interaction network includes proteins from the Set 1 (Figure 4

the sensitive population only. Solid lines represent direct interaction

represent genes not included on our array. Blue, green, and red sp

with RNAi-induced defects in sarcomere assembly, and genes associa

in the associated heat map but not included in the network, but are lo


blockage, but NOS was not among the small set of

dioxin-regulated genes in AHR-null mice (Tijet et al.

2006). Our data and literature promote the intriguing

possibility that signalling pathways associated with

increased nitric oxide synthesis are a component of the

adaptive mechanism of tolerance to pollution that is

recently derived in some populations of killifish.

The mechanism underlying NOS up-regulation fol-

lowing PCB126 exposure in resistant individuals from

the Bridgeport population is unclear, though AHR sig-

nalling pathways cross-talk with multiple signalling

pathways including those that regulate NOS expression

(Fig. 6). For example, in human endothelial cells, expo-

sure to PCB126 is correlated with increased eNOS

expression and depends on both AHR and oestrogen

receptor (ER) signalling pathways (Omori et al. 2007).

Also, dioxin stimulates cellular production of growth

factors and cytokines like TNFa (Matsumura 2003),

which in turn can regulate NOS (Li et al. 2002).

Intriguingly, pre-exposure to hypoxia diminishes

dioxin induction of CYP1A and diminishes dioxin toxic-

ity in developing zebrafish embryos (Prasch et al. 2004),

but pre-exposure to dioxin does not diminish the

hypoxia response (Prasch et al. 2004). Both AHR (the

dioxin sensor) and HIF-1a (the hypoxia sensor) are

members of the same family of transcription factors (Gu

et al. 2000), and both form heterodimers with ARNT to

become transcriptionally active (Fig. 6). Perhaps tolerant

population-specific HIF-1a activation sequesters avail-

able ARNT, thereby preventing transcriptional activation

by AHR (Prasch et al. 2004). Indeed, HIF-1a expression

is highly correlated with NOS expression (Fig. 4),

although up-regulation of HMOX, a classic biomarker of

HIF-1a activation (Lee et al. 1997), is not evident in the

tolerant population (data not shown). Alternatively,

other genes that are activated by hypoxia such as NOS

(McNeill & Perry 2006) may be responsible for the

protective effect of hypoxia on embryonic DLC toxicity.

TPM3

PDLIM3SarcomereZ-discMYOZ2

MLP

CAPZA2

ATP2A1SarcomereI-band

Z-disc

MURF2TTN

Angiogenesis

Cardiomyopathy

RNAi sarcomere defectsmyosin

TMOD1

TPM3

TnT

) cardiac genes that are down-regulated upon PCB exposure in

, dashed lines indicate indirect interactions, and dashed circles

ots highlight proteins associated with cardiomyopathies, genes

ted with angiogenesis, respectively. Three proteins are included

calized to the sarcomere. Gene names are as in Figure 4.

Dioxins/PCBs HypoxiaER

HIF-1aARNTAHRGrowth factors

NOSARID1A

Cytokines

Nitricoxide

TTN DREDRE HRE

CYP1A HMOX

MYH10

Cardiac toxicityCardiac remodeling

HRE

AHRgenebattery HIF-1agenebattery

Fig. 6 Model illustrating interactions between genes and signalling pathways. Solid blue circles and dashed blue circles represent

genes included or not present in our analysis, respectively. AHR, aryl hydrocarbon receptor; ARID1A, AT-rich interactive domain-

containing protein 1A; ARNT, aryl hydrocarbon receptor nuclear translocator 2; CYP1A, cytochrome P450 1A; DRE, dioxin response

element (AHR complex DNA binding sites) ER, estrogen receptor; HIF-la, hypoxia-inducible factor 1 alpha; HMOX, haeme oxyge-

nase 1; HRE, hypoxia response element (HIF1a complex DNA binding site); MYH1O, myosin heavy chain 10; NOS, nitric oxide

synthase; TTN, titin.


Increased NO has been associated with inhibition of

dioxin induction of CYP1A (Kim & Sheen 2000),

although it is currently unknown whether NO can block

induction of other transcriptional targets of the AHR.

Over 70 genes are co-expressed with NOS (Fig. S1),

and this group is enriched for gene ontology categories

including nucleic acid binding, RNA metabolic process,

chromosome organization and biogenesis, and anatomi-

cal structure development. Some of these genes with

‘protective’ patterns of gene expression strongly corre-

lated with NOS are highlighted in Fig. 4 (set 4) and

include myosin heavy chain 10 (MYH10), titin (TTN)

and other titin-like proteins including IGFN1 and ARID

domain-containing protein 1A (ARID1A). Both TTN and

IGFN1 are sarcomeric, and members of the same protein

family (Furst & Gautel 1995; Mansilla et al. 2008). TTN,

the largest known protein, is a major structural constitu-

ent of the cardiac sarcomere, and abnormal titin is asso-

ciated with cardiac disease and muscular dystrophy in

humans (Hackman et al. 2002; Itoh-Satoh et al. 2002;

Carmignac et al. 2007; Pollazzon et al. in press). The

titin protein is considered a key integrator of mechanical

and stress-induced signalling pathways during cardio-

myocyte development. In mammals, one copy of the

TTN gene is present in the genome, and alternate splic-

ing of the protein adjusts cardiomyocyte stiffness during

development and disease (Kruger & Linke 2009). In tele-

ost fishes, two titin copies exist (Seeley et al. 2007).

In zebrafish, the paralogs TTNa and TTNb are differen-

tially expressed during development, and expression

knockdown of TTNa (but not TTNb) disrupts sarcomere

assembly (Seeley et al. 2007).

The ARID1A gene also exhibits this ‘protective’ pat-

tern of expression (Fig. 4 set 4) and is intriguing

because the protein is a key member of the SWI ⁄ SNF

chromatin remodelling complex (Wang et al. 2004),

which promotes the activation of androgen, oestrogen

and glucocorticoid receptor signalling pathways (Inoue

et al. 2002). The ER and AHR signalling pathways inter-

act (Matthews & Gustafsson 2006; Ahmed et al. 2009),

knockdown of ER expression reduces dioxin-induced

expression of CYP1A (Matthews et al. 2005), and both

pathways may contribute to DLC-induced regulation of

NOS (Omori et al. 2007) (Fig. 6).

Summary and conclusions

Much variation in gene expression is partitioned among

individuals, populations and species, and much variation

in gene expression is induced by environmental change.

Importantly, some adaptive patterns of gene expression

may only emerge in specific environments as taxon-by-

environment interactions (Aubin-Horth & Renn 2009).

Indeed, little transcriptome variation was detected

between developing embryos from pollution-tolerant and

pollution-sensitive populations of F. heteroclitus raised in

a common clean environment (Bozinovic & Oleksiak in

press). In contrast, population comparative experiments

reported here that included challenge to PCBs revealed

extensive, evolved, functional genomic variation between



tolerant and sensitive populations. Studies that are

designed to test for taxon-by-environment interactions

may accelerate the discovery of adaptively important

functional genomic variation and accelerate functional

annotation of the many currently uncharacterized, but

potentially ecologically relevant, genes.

We found that gene expression was highly correlated

with individual phenotype and that individual-specific

measures were powerful for detecting PCB exposure–

related expression patterns that were not detected with

measures of mean group (exposure level) effects. This

was particularly noticeable in the tolerant population.

Individual-specific measurements enabled greater scope

for inference and highlighted the potential benefits of

measuring gene expression in individuals especially in

studies including wild outbred populations that may

express extensive inter-individual phenotypic variation.

Consistent with the cardiovascular system toxicity phe-

notype, a large group of genes were down-regulated and

predictive of cardiac toxicity in the sensitive population

only. The interaction network, putative function and sub-

cellular localization of these proteins further strengthens

evidence that sarcomere mis-assembly is the mechanism

through which PCB exposure induces cardiovascular sys-

tem toxicity in developing fish embryos. Intriguingly,

these genes were not similarly regulated in the few

Bridgeport individuals exhibiting cardiovascular system

abnormalities at extreme doses, implying that similar

phenotypes may emerge through alternate mechanisms

in tolerant versus sensitive populations.

Our data indicate that derived tolerance in the

Bridgeport population is associated with global block-

ade of the AHR signalling pathway (the key pathway

that mediates DLC toxicity) but that the blockade

becomes leaky at the most extreme PCB exposures

where toxicity starts to emerge. Several previous studies

have demonstrated that CYP1A, one target of AHR sig-

nalling, is refractory to PCB induction in tolerant killi-

fish populations. However, data presented here are the

first to indicate that most known gene targets of AHR

signalling are refractory to induction in a tolerant killi-

fish population. Global disruption of AHR signalling

could be a result of mutations in AHR signalling pro-

teins (such as AHR, ARNT, AHRR, AIP), although pop-

ulation genetic and functional assays to date have not

provided support for any single candidate. Alterna-

tively, altered signalling in cross-talking pathways

could contribute to derived disruption of AHR signal-

ling. The derived ‘protective’ pattern of nitric oxide syn-

thase expression is a candidate for cross-talk disruption

of AHR signalling. Expression of this gene is highly

predictive of normal development in tolerant popula-

tion individuals only, it is regulated by pathways

known to interact with AHR signalling and repress


DLC toxicity, and nitric oxide is important for cardio-

vascular system function and appears to block some

AHR-mediated gene induction. The normally DLC-

inducible AHR signalling pathway is refractory to

induction in the derived tolerant population, which

appears adaptive to canalize development in the face of

extreme environmental stress. Comparative studies

including additional tolerant and sensitive populations

are ongoing and may yield key insights into molecular

mechanisms underlying convergence of pollution toler-

ance in killifish.

Authors’ contributions

DN and AW contributed to study design; DN, DC and DT con-

tributed to data collection; DN, DC, DT and AW contributed to

data analysis and interpretation; DN and AW wrote the manu-

script.

Acknowledgements

We appreciate the helpful advice from reviewers of early

drafts, including Dina Proestou (U.S. Environmental Protection

Agency), John Battista (Louisiana State University) and Patrick

Flight (Brown University). This is contribution number AED-

10-041 of the U.S. Environmental Protection Agency, Office of

Research and Development, National Health and Environmen-

tal Effects Research Laboratory, Atlantic Ecology Division,

which partially supported this research. This manuscript has

been reviewed and approved for publication by the U.S. EPA.

Approval does not signify that the contents necessarily reflect

the views and policies of the U.S. EPA. Mention of trade

names, products, or services does not convey and should not

be interpreted as conveying, official U.S. EPA approval,

endorsement, or recommendation. Funding for some of this

project was provided by the National Science Foundation grant

BES-0652006 to AW. The funders had no role in study design,

data collection and analysis, decision to publish or preparation

of the manuscript.

References

Ahmed S, Valen E, Sandelin A, Matthews J (2009) Dioxin

increases the interaction between aryl hydrocarbon receptor

and estrogen receptor alpha at human promoters.

Toxicological Sciences, 111, 254–266.

Andersen ME, Yang RSH, French CT, Chubb LS, Dennison JE

(2002) Molecular circuits, biological switches, and nonlinear

dose-response relationships. Environmental Health Perspectives,

110, 971–978.

Andreasen EA, Mathew LK, Tanguay RL (2006) Regenerative

growth is impacted by TCDD: gene expression analysis

reveals extracellular matrix modulation. Toxicological Sciences,

92, 254–269.

Andree B, Fleige A, Arnold HH, Brand T (2002) Mouse Pop1 is

required for muscle regeneration in adult skeletal muscle.

Molecular and Cellular Biology, 22, 1504–1512.


Antkiewicz DS, Burns CG, Carney SA, Peterson RE, Heideman

W (2005) Heart malformation is an early response to TCDD

in embryonic zebrafish. Toxicological Sciences, 84, 368–377.

Antkiewicz DS, Peterson RE, Heideman W (2006) Blocking

expression of AHR2 and ARNT1 in zebrafish larvae protects

against cardiac toxicity of 2,3,7,8-tetrachlorodibenzo-p-

dioxin. Toxicological Sciences, 94, 175–182.

Arimura T, Hayashi T, Matsumoto Y et al. (2007) Structural

analysis of four and half LIM protein-2 in dilated

cardiomyopathy. Biochemical and Biophysical Research Commu-

nications, 357, 162–167.

Armstrong PB, Child JS (1965) Stages in normal development

of Fundulus heteroclitus. Biological Bulletin, 128, 143–168.

Arzuaga X, Elskus A (2002) Evidence for resistance to

benzo[a]pyrene and 3,4,3 ‘4‘-tetrachlorobiphenyl in a

chronically polluted Fundulus heteroclitus population. Marine

Environmental Research, 54, 247–251.

Aubin-Horth N, Renn SCP (2009) Genomic reaction norms:

using integrative biology to understand molecular mechanisms

of phenotypic plasticity. Molecular Ecology, 18, 3763–3780.

Bell DR, Poland A (2000) Binding of aryl hydrocarbon receptor

(AhR) to AhR-interacting protein. Journal of Biological

Chemistry, 275, 36407–36414.

Bello SM, Franks DG, Stegeman JJ, Hahn ME (2001) Acquired

resistance to Ah receptor agonists in a population of Atlantic

killifish (Fundulus heteroclitus) inhabiting a marine superfund

site: in vivo and in vitro studies on the inducibility of

xenobiotic metabolizing enzymes. Toxicological Sciences, 60,

77–91.

Bozinovic G, Oleksiak MF (2010) Embryonic gene expression

among pollutant resistant and sensitive Fundulus heteroclitus

populations. Aquatic Toxicology, 98, 221–229.

Bunger MK, Glover E, Moran SM et al. (2008) Abnormal liver

development and resistance to 2,3,7,8-tetrachlorodibenzo-

p-dioxin toxicity in mice carrying a mutation in the DNA

binding domain of the aryl hydrocarbon receptor.


Burnett KG, Bain LJ, Baldwin WS et al. (2007) Fundulus as

the premier teleost model in environmental biology:

opportunities for new insights using genomics. Comparative

Biochemistry and Physiology Part D: Genomics and Proteomics, 2,

257–286.

Calzolai L, Ansorge W, Calabrese E et al. (2007) Transcriptomics

and proteomics. Applications to ecotoxicology. Comparative

Biochemistry and Physiology Part D: Genomics and Proteomics, 2,

245–249.

Camarata T, Krcmery J, Snyder D et al. (2010) Pdlim7 (LMP4)

regulation of Tbx5 specifies zebrafish heart atrio-ventricular

boundary and valve formation. Developmental Biology, 337,

233–245.

Cao ZJ, Tanguay RL, McKenzie D, Peterson RE, Aiken JM

(2003) Identification of a putative calcium-binding protein as

a dioxin-responsive gene in zebrafish and rainbow trout.

Aquatic Toxicology, 63, 271–282.

Carmignac V, Salih MAM, Quijano-Roy S et al. (2007)

C-terminal titin deletions cause a novel early-onset myopathy

with fatal cardiomyopathy. Annals of Neurology, 61, 340–351.

Carney SA, Peterson RE, Heideman W (2004) 2,3,7,8-

tetrachlorodibenzo-p-dioxin activation of the aryl

hydrocarbon receptor ⁄ aryl hydrocarbon receptor nuclear

translocator pathway causes developmental toxicity through

a CYP1A-independent mechanism in zebrafish. Molecular

Pharmacology, 66, 512–521.

Carney SA, Chen J, Burns CG et al. (2006) Aryl hydrocarbon

receptor activation produces heart-specific transcriptional

and toxic responses in developing zebrafish. Molecular


Chen J, Carney SA, Peterson RE, Heideman W (2008a) Compa-

rative genomics identifies genes mediating cardiotoxicity in

the embryonic zebrafish heart. Physiological Genomics, 33,

148–158.

Chen SN, Lombardi R, Rodriguez G, Marian AJ (2008b)

Abstract 1427: pathogenesis of hypertrophic cardiomyopathy

caused by the myozenin 2 mutations involves calcineurin-

dependent and -independent mechanisms. Circulation, 118,

S_323.

Crawford DL, Oleksiak MF (2007) The biological importance of

measuring individual variation. Journal of Experimental

Biology, 210, 1613–1621.

Fernandez-Salguero P, Pineau T, Hilbert D et al. (1995)

Immune system impairment and hepatic fibrosis in mice

lacking the dioxin-binding Ah receptor. Science, 268, 722–726.

Franc M, Moffat I, Boutros P et al. (2008) Patterns of dioxin-

altered mRNA expression in livers of dioxin-sensitive versus

dioxin-resistant rats. Archives of Toxicology, 82, 809–830.

Fritz-Six KL, Cox PR, Fischer RS et al. (2003) Aberrant

myofibril assembly in tropomodulin1 null mice leads to

aborted heart development and embryonic lethality. Journal

of Cell Biology, 163, 1033–1044.

Furst DO, Gautel M (1995) The anatomy of a molecular giant:

how the sarcomere cytoskeleton is assembled from

immunoglobulin superfamily molecules. Journal of Molecular

and Cellular Cardiology, 27, 951–959.

Gardner GR, Pruell RJ, Malcolm AR (1992) Chemical induction

of tumors in oysters by a mixture of aromatic and

chlorinated hydrocarbons, amines and metals. Marine

Environmental Research, 34, 59–63.

Geier C, Perrot A, Ozcelik C et al. (2003) Mutations in the

human muscle LIM protein gene in families with

hypertrophic cardiomyopathy. Circulation, 107, 1390–1395.

Gu Y-Z, Hogenesch JB, Bradfield CA (2000) The PAS

superfamily: sensors of environmental and developmental

signals. Annual Review of Pharmacology and Toxicology, 40,

519–561.

Guiney PD, Walker MK, Spitsbergen JM, Peterson RE (2000)

Hemodynamic dysfunction and cytochrome P4501A mRNA

expression induced by 2,3,7,8-tetrachlorodibenzo-p-dioxin

during embryonic stages of lake trout development.

Toxicology and Applied Pharmacology, 168, 1–14.

Hackman P, Vihola A, Haravuori H et al. (2002) Tibial

muscular dystrophy is a titinopathy caused by mutations in

TTN, the gene encoding the giant skeletal-muscle protein

titin. American Journal of Human Genetics, 71, 492–500.

Hahn ME, Karchner SI, Shapiro MA, Perera SA (1997)

Molecular evolution of two vertebrate aryl hydrocarbon

(dioxin) receptors (AHR1 and AHR2) and the PAS family.

Proceedings of the National Academy of Sciences of the United

States of America, 94, 13743–13748.

Hahn ME, Karchner SI, Franks DG, Merson RR (2004) Aryl

hydrocarbon receptor polymorphisms and dioxin resistance

in Atlantic killifish (Fundulus heteroclitus). Pharmacogenetics,

14, 131–143.



Hahn ME, Allan LL, Sherr DH (2009) Regulation of constitutive

and inducible AHR signaling: complex interactions

involving the AHR repressor. Biochemical Pharmacology, 77,

485–497.

Handley-Goldstone HM, Grow MW, Stegeman JJ (2005)

Cardiovascular gene expression profiles of dioxin exposure

in zebrafish embryos. Toxicological Sciences, 85, 683–693.

Head JA, Hahn ME, Kennedy SW (2008) Key amino acids in

the aryl hydrocarbon receptor predict dioxin sensitivity in

avian species. Environmental Science & Technology, 42, 7535–

7541.

Huang DW, Sherman BT, Lempicki RA (2009) Systematic and

integrative analysis of large gene lists using DAVID

bioinformatics resources. Nature Protocols, 4, 44–57.

Inoue H, Furukawa T, Giannakopoulos S et al. (2002) Largest

subunits of the human SWI ⁄ SNF chromatin-remodeling

complex promote transcriptional activation by steroid

hormone receptors. Journal of Biological Chemistry, 277, 41674–

41685.

Itoh-Satoh M, Hayashi T, Nishi H et al. (2002) Titin mutations as

the molecular basis for dilated cardiomyopathy. Biochemical

and Biophysical Research Communications, 291, 385–393.

Jenny MJ, Karchner SI, Franks DG et al. (2009) Distinct roles of

two zebrafish AHR repressors (AHRRa and AHRRb) in

embryonic development and regulating the response to

2,3,7,8-tetrachlorodibenzo-p-dioxin. Toxicological Sciences, 110,

426–441.

Jonsson ME, Jenny MJ, Woodin BR, Hahn ME, Stegeman JJ

(2007) Role of AHR2 in the expression of novel cytochrome

p450 1 family genes, cell cycle genes, and morphological

defects in developing zebra fish exposed to 3,3¢,4,4¢,5-

pentachlorobiphenyl or 2,3,7,8-tetrachlorodibenzo-p-dioxin.


Karchner SI, Powell WH, Hahn ME (1999) Identification and

functional characterization of two highly divergent aryl

hydrocarbon receptors (AHR1 and AHR2) in the teleost

Fundulus heteroclitus. Journal of Biological Chemistry, 274,

33814–33824.

Karchner SI, Franks DG, Powell WH, Hahn ME (2002)

Regulatory interactions among three members of the

vertebrate aryl hydrocarbon receptor family: AHR repressor,

AHR1, and AHR2. Journal of Biological Chemistry, 277, 6949–

6959.

Karchner SI, Franks DG, Hahn ME (2005) AHR1B, a new

functional aryl hydrocarbon receptor in zebrafish: tandem

arrangement of ahr1b and ahr2 genes. Biochemical Journal,

392, 153–161.

Kawakami T, Ishimura R, Nohara K et al. (2006) Differential

susceptibilities of Holtzman and Sprague-Dawley rats to

fetal death and placental dysfunction induced by 2, 3,7,8-

teterachlorodibenzo-p-dioxin (TCDD) despite the identical

primary structure of the aryl hydrocarbon receptor.


Kim JE, Sheen YY (2000) Nitric oxide inhibits dioxin action for

the stimulation of Cyp1a1 promoter activity. Biological &

Pharmaceutical Bulletin, 23, 575–580.

Knoblauch H, Geier C, Adams S et al. (2010) Contractures and

hypertrophic cardiomyopathy in a novel FHL1 mutation.

Annals of Neurology, 67, 136–140.

Knoll R, Hoshijima M, Hoffman HM et al. (2002) The cardiac

mechanical stretch sensor machinery involves a Z disc


complex that is defective in a subset of human dilated

cardiomyopathy. Cell, 111, 943–955.

Kruger M, Linke WA (2009) Titin-based mechanical signalling

in normal and failing myocardium. Journal of Molecular and

Cellular Cardiology, 46, 490–498.

Lee PJ, Jiang B-H, Chin BY et al. (1997) Hypoxia-inducible

factor-1 mediates transcriptional activation of the heme

oxygenase-1 gene in response to hypoxia. Journal of Biological

Chemistry, 272, 5375–5381.

Li H, Wallerath T, Forstermann U (2002) Physiological

mechanisms regulating the expression of endothelial-type

NO synthase. Nitric Oxide, 7, 132–147.

Lin S, Zhao DB, Bownes M (2007) Blood vessel ⁄ epicardial

substance (bves) expression, essential for embryonic

development, is down regulated by Grk ⁄ EFGR signalling.

International Journal of Developmental Biology, 51, 37–44.

Mansilla F, Dominguez CAG, Yeadon JE et al. (2008)

Translation elongation factor eEF1A binds to a novel myosin

binding protein-C-like protein. Journal of Cellular

Biochemistry, 105, 847–858.

Marchand A, Tomkiewicz C, Marchandeau JP et al. (2005)

2,3,7,8-tetrachlorodibenzo-p-dioxin induces insulin-like

growth factor binding protein-1 gene expression and

counteracts the negative effect of insulin. Molecular


Marian AJ, Roberts R (1995) Recent advances in the molecular

genetics of hypertrophic cardiomyopathy. Circulation, 92,

1336–1347.

Marian AJ, Roberts R (2001) The molecular genetic basis for

hypertrophic cardiomyopathy. Journal of Molecular and

Cellular Cardiology, 33, 655–670.

Massion PB, Balligand J-L (2007) Relevance of nitric oxide for

myocardial remodeling. Current Heart Failure Reports, 4, 18–

25.

Matsumura F (2003) On the significance of the role of cellular

stress response reactions in the toxic actions of dioxin.

Biochemical Pharmacology, 66, 527–540.

Matthews J, Gustafsson J-A (2006) Estrogen receptor and aryl

hydrocarbon receptor signaling pathways. Nuclear Receptor

Signaling, 4, e016.

Matthews J, Wihlen B, Thomsen J, Gustafsson J-A (2005) Aryl

hydrocarbon receptor-mediated transcription: ligand-

dependent recruitment of estrogen receptor a to 2,3,7,8-

tetrachlorodibenzo- p-dioxin-responsive promoters. Molecular

and Cellular Biology, 25, 5317–5328.

McNeill B, Perry SF (2006) The interactive effects of hypoxia

and nitric oxide on catecholamine secretion in rainbow trout

(Oncorhynchus mykiss). Journal of Experimental Biology, 209,

4214–4223.

Meyer JN, Wassenberg DM, Karchner SI, Hahn ME, Di Giulio

RT (2003) Expression and inducibility of aryl hydrocarbon

receptor pathway genes in wild-caught killifish (Fundulus

heteroclitus) with different contaminant-exposure histories.

Environmental Toxicology and Chemistry, 22, 2337–2343.

Nacci D, Coiro L, Champlin D et al. (1999) Adaptations of

wild populations of the estuarine fish Fundulus heteroclitus to

persistent environmental contaminants. Marine Biology, 134,

9–17.

Nacci D, Champlin D, Coiro L, McKinney R, Jayaraman S

(2002) Predicting the occurrence of genetic adaptation to

dioxin like compounds in populations of the estuarine fish


Fundulus heteroclitus. Environmental Toxicology and Chemistry,

21, 1525–1532.

Nacci D, Champlin D, Jayaraman S (2010) Adaptation of the

estuarine fish Fundulus heteroclitus (Atlantic Killifish) to

polychlorinated biphenyls (PCBs). Estuaries and Coasts, 33,

853–864.

Nebert DW, Roe AL, Dieter MZ et al. (2000) Role of the

aromatic hydrocarbon receptor and [Ah] gene battery in the

oxidative stress response, cell cycle control, and apoptosis.

Biochemical Pharmacology, 59, 65–85.

Nebert DW, Dalton TP, Okey AB, Gonzalez FJ (2004) Role of

aryl hydrocarbon receptor-mediated induction of the CYP1

enzymes in environmental toxicity and cancer. Journal of

Biological Chemistry, 279, 23847–23850.

Niimura H, Patton KK, McKenna WJ et al. (2002) Sarcomere

protein gene mutations in hypertrophic cardiomyopathy of

the elderly. Circulation, 105, 446–451.

Oleksiak MF, Churchill GA, Crawford DL (2002) Variation in

gene expression within and among natural populations.

Nature Genetics, 32, 261–266.

Oleksiak MF, Roach JL, Crawford DL (2005) Natural variation

in cardiac metabolism and gene expression in Fundulus

heteroclitus. Nature Genetics, 37, 67–72.

Olson TM, Karst ML, Whitby FG, Driscoll DJ (2002) Myosin

light chain mutation causes autosomal recessive

cardiomyopathy with mid-cavitary hypertrophy and

restrictive physiology. Circulation, 105, 2337–2340.

Omori N, Fukata H, Sato K et al. (2007) Polychlorinated

biphenyls alter the expression of endothelial nitric oxide

synthase mRNA in human umbilical vein endothelial cells.

Human & Experimental Toxicology, 26, 811–816.

Paschall JE, Oleksiak MF, VanWye JD et al. (2004) FunnyBase:

a systems level functional annotation of Fundulus ESTs for

the analysis of gene expression. BMC Genomics, 5, 96.

Pohjanvirta R, Viluksela M, Tuomisto JT et al. (1999)

Physicochemical differences in the AH receptors of the most

TCDD-susceptible and the most TCDD-resistant rat strains.


Pollazzon M, Suominen T, Penttila S et al. (2010) The first

Italian family with tibial muscular dystrophy caused by a

novel titin mutation. Journal of Neurology, 257, 575–579.

Powell WH, Bright R, Bello SM, Hahn ME (2000)

Developmental and tissue-specific expression of AHR1,

AHR2, and ARNT2 in dioxin-sensitive and -resistant

populations of the marine fish Fundulus heteroclitus.


Prasch AL, Teraoka H, Carney SA et al. (2003) Aryl

hydrocarbon receptor 2 mediates 2,3,7,8-tetrachlorodibenzo-

p-dioxin developmental toxicity in zebrafish. Toxicological

Sciences, 76, 138–150.

Prasch AL, Andreasen EA, Peterson RE, Heideman W (2004)

Interactions between 2,3,7,8-tetrachlorodibenzo-p-dioxin

(TCDD) and hypoxia signaling pathways in zebrafish:

hypoxia decreases responses to TCDD in zebrafish embryos.


Prasch AL, Tanguay RL, Mehta V, Heideman W, Peterson RE

(2006) Identification of Zebrafish ARNT1 Homologs: 2,3,7,8-

Tetrachlorodibenzo-p-dioxin Toxicity in the Developing Ze-

brafish Requires ARNT1. Molecular Pharmacology, 69, 776–787.

Prince R, Cooper KR (1995) Comparisons of the effects of

2,3,7,8-tetrachlorodibenzo-p-dioxin on chemically impacted

and nonimpacted subpopulations of Fundulus heteroclitus .1.

TCDD toxicity. Environmental Toxicology and Chemistry, 14,

579–587.

Ripley AN, Chang MS, Bader DM (2004) Bves is expressed

in the epithelial components of the retina, lens, and

cornea. Investigative Ophthalmology and Visual Science, 45,

2475–2483.

Rogerson PF, Schimmel SC, Hoffman G (1985) Chemical and

Biological Characterization of Black Rock Harbor Dredged

Material. U.S. Army Corps of Engineers, Vicksburg.

Rowlands JC, Gustafsson JA (1997) Aryl hydrocarbon receptor-

mediated signal transduction. Critical Reviews in Toxicology,

27, 109–134.

Rutherford SL, Lindquist S (1998) Hsp90 as a capacitor for

morphological evolution. Nature, 396, 336–342.

Saeed AI, Bhagabati NK, Braisted JC et al. (2006) TM4

microarray software suite. Methods in Enzymology, 411, 134–

193.

Schmidt JV, Su GH, Reddy JK, Simon MC, Bradfield CA (1996)

Characterization of a murine Ahr null allele: involvement of

the Ah receptor in hepatic growth and development.



Schnorrer F, Schonbauer C, Langer CCH et al. (2010)

Systematic genetic analysis of muscle morphogenesis and

function in Drosophila. Nature, 464, 287–291.

Seeley M, Huang W, Chen Z et al. (2007) Depletion of

zebrafish titin reduces cardiac contractility by disrupting the

assembly of Z-discs and A-bands. Circulation Research, 100,

238–245.

Sehnert AJ, Huq A, Weinstein BM et al. (2002) Cardiac

troponin T is essential in sarcomere assembly and cardiac

contractility. Nature Genetics, 31, 106–110.

Smith TK, Hager HA, Francis R et al. (2008) Bves directly

interacts with GEFT, and controls cell shape and movement

through regulation of Rac1 ⁄ Cdc42 activity. Proceedings of the

National Academy of Sciences, 105, 8298–8303.

Storey JD, Tibshirani R (2003) Statistical significance for

genomewide studies. Proceedings of the National Academy of

Sciences of the United States of America, 100, 9440–9445.

Storey JD, Xiao WZ, Leek JT, Tompkins RG, Davis RW (2005)

Significance analysis of time course microarray experiments.



Thackaberry EA, Nunez BA, Ivnitski-Steele ID, Friggins M,

Walker MK (2005) Effect of 2,3,7,8-tetrachlorodibenzo-p-

dioxin on murine heart development: alteration in fetal and

postnatal cardiac growth, and postnatal cardiac chronotropy.


Theobald HM, Kimmel GL, Peterson RE (2003) Developmental

and reproductive toxicity of dioxins and related chemicals.

In: Dioxins and Health (eds Schecter A, Gasiewicz TA). pp.

327–432, Wiley-Interscience, Hoboken, NJ, USA.

Thomas PE, Kouri RE, Hutton JJ (1972) Genetics of aryl

hydrocarbon hydroxylase induction in mice—single gene

difference between C57bl ⁄ 6j and Dba ⁄ 2j. Biochemical Genetics,

6, 157–168.

Tijet N, Boutros PC, Moffat ID et al. (2006) Aryl hydrocarbon

receptor regulates distinct dioxin-dependent and dioxin-

independent gene batteries. Molecular Pharmacology, 69, 140–

153.


TRANSCRIPTOMICS

Triant DA, Whitehead A (2009) Simultaneous extraction of

high-quality RNA and DNA from small tissue samples.

Journal of Heredity, 100, 246–250.

Van Veld PA, Nacci DE (2008) Toxicity resistance. In: The

Toxicology of Fishes (eds Di Giulio RT, Hinton DE). pp. 597–

641, Taylor and Francis, Boca Raton, FL.

Walker MK, Catron TF (2000) Characterization of cardiotoxicity

induced by 2,3,7,8-tetrachlorodibenzo-p-dioxin and related

chemicals during early chick embryo development.


Wang XM, Nagl NG, Wilsker D et al. (2004) Two related ARID

family proteins are alternative subunits of human SWI ⁄ SNF

complexes. Biochemical Journal, 383, 319–325.

Weiss C, Faust D, Durk H et al. (2005) TCDD induces c-jun

expression via a novel Ah (dioxin) receptor-mediated p38-

MAPK-dependent pathway. Oncogene, 24, 4975–4983.

Wernersson R, Juncker AS, Nielsen HB (2007) Probe selection

for DNA microarrays using OligoWiz. Nature Protocols, 2,

2677–2691.

Whitehead A, Crawford DL (2005) Variation in tissue-specific

gene expression among natural populations. Genome Biology,

6, R13.

Whitehead A, Crawford DL (2006) Variation within and

among species in gene expression: raw material for

evolution. Molecular Ecology, 15, 1197–1211.

Whitehead A, Galvez F, Zhang S, Williams LM, Oleksiak MF

(2010) Functional genomics of physiological plasticity and

local adaptation in killifish. Journal of Heredity, doi: 10.1093/

jhered/esq077.

Whitfield CW, Cziko AM, Robinson GE (2003) Gene expression

profiles in the brain predict behavior in individual honey

bees. Science, 302, 296–299.

Andrew Whitehead is an Assistant Professor in the Depart-

ment of Biological Sciences at Louisiana State University with


interests in ecological and evolutionary functional genomics,

population genomics, ecophysiology and genome evolution.

Deborah Triant is a postdoctoral research associate in the

Department of Biochemistry and Molecular Genetics at the

University of Virginia. This study was part of her postdoctoral

research in the Whitehead Lab at Louisiana State University.

Denise Champlin is a Research Biologist at the U.S. Environ-

mental Protection Agency’s Office of Research and Develop-

ment in Narragansett, RI, with interests in the experimental

design and conduct of toxicological studies using marine

organisms. Diane Nacci is a Research Biologist at the U.S.

Environmental Protection Agency’s Office of Research and

Development in Narragansett, RI, with interests in the ecologi-

cal and evolutionary responses of wildlife to chemical expo-

sures and other human-mediated stressors.

OF EVOLVED POLLUTION T OLERANCE 5203

Supporting information

Additional supporting information can be found in the online

version of this article.

Fig. S1 Major groups of co-regulated genes.

Table S1 List of genes included in the analysis, including

probe sequence, putative annotation, results from statistical

tests, and average expression level for individuals and popula-

tions.

Please note: Wiley-Blackwell are not responsible for the

content or functionality of any supporting information supplied

by the authors. Any queries (other than missing material)

should be directed to the corresponding author for the article.

Comparative transcriptomics implicates mechanisms of ... · transcriptome responses to embryonic...

Documents

Transcript of Comparative transcriptomics implicates mechanisms of ... · transcriptome responses to embryonic...