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;
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
� 2010 Blackwell Publishing Ltd
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
� 2010 Blackwell Publishing Ltd
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.
TRANSCRIPTOMICS OF EVOLVED POLLUTION T OLERANCE 5189
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-
� 2010 Blackwell Publishing Ltd
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
5190 A. WH IT EH EAD E T A L.
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
� 2010 Blackwell Publishing Ltd
TRANSCRIPTOMICS OF EVOLVED POLLUTION T OLERANCE 5191
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
� 2010 Blackwell Publishing Ltd
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.
5192 A. WH IT EH EAD E T A L.
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
� 2010 Blackwell Publishing Ltd
(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).
TRANSCRIPTOMICS OF EVOLVED POLLUTION T OLERANCE 5193
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
� 2010 Blackwell Publishing Ltd
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.
5194 A. WH IT EH EAD E T A L.
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
� 2010 Blackwell Publishing Ltd
TRANSCRIPTOMICS OF EVOLVED POLLUTION T OLERANCE 5195
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.
� 2010 Blackwell Publishing Ltd
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-
5196 A. WH IT EH EAD E T A L.
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
� 2010 Blackwell Publishing Ltd
TRANSCRIPTOMICS OF EVOLVED POLLUTION T OLERANCE 5197
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
� 2010 Blackwell Publishing Ltd
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.
5198 A. WH IT EH EAD E T A L.
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
� 2010 Blackwell Publishing Ltd
TRANSCRIPTOMICS OF EVOLVED POLLUTION T OLERANCE 5199
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
� 2010 Blackwell Publishing Ltd
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.
5200 A. WH IT EH EAD E T A L.
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.
Toxicological Sciences, 106, 83–92.
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
Pharmacology, 70, 549–561.
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.
� 2010 Blackwell Publishing Ltd
TRANSCRIPTOMICS OF EVOLVED POLLUTION T OLERANCE 5201
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.
Toxicological Sciences, 100, 180–193.
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.
Toxicology and Applied Pharmacology, 212, 224–236.
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
� 2010 Blackwell Publishing Ltd
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
Pharmacology, 67, 444–452.
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
5202 A. WH IT EH EAD E T A L.
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.
Toxicology and Applied Pharmacology, 155, 82–95.
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.
Toxicological Sciences, 57, 229–239.
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.
Toxicological Sciences, 78, 68–77.
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.
Proceedings of the National Academy of Sciences of the United
States of America, 93, 6731–6736.
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.
Proceedings of the National Academy of Sciences of the United
States of America, 102, 12837–12842.
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.
Toxicological Sciences, 88, 242–249.
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.
� 2010 Blackwell Publishing Ltd
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.
Toxicology and Applied Pharmacology, 167, 210–221.
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
� 2010 Blackwell Publishing Ltd
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.