In vivo-in vitro toxicogenomic comparison of TCDD-elicited gene expression in Hepa1c1c7 mouse hepatoma cells and C57BL/6 hepatic tissue.

Abstract

In vitro systems have inherent limitations in their ability to model whole organism gene responses, which must be identified and appropriately considered when developing predictive biomarkers of in vivo toxicity. Systematic comparison of in vitro and in vivo temporal gene expression profiles were conducted to assess the ability of Hepa1c1c7 mouse hepatoma cells to model hepatic responses in C57BL/6 mice following treatment with 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD).
Gene expression analysis and functional gene annotation indicate that Hepa1c1c7 cells appropriately modeled the induction of xenobiotic metabolism genes in vivo. However, responses associated with cell cycle progression and proliferation were unique to Hepa1c1c7 cells, consistent with the cell cycle arrest effects of TCDD on rapidly dividing cells. In contrast, lipid metabolism and immune responses, representative of whole organism effects in vivo, were not replicated in Hepa1c1c7 cells.
These results identified inherent differences in TCDD-mediated gene expression responses between these models and highlighted the limitations of in vitro systems in modeling whole organism responses, and additionally identified potential predictive biomarkers of toxicity.

Full text

Page 1

BioMed Central
Page 1 of 18
(page number not for citation purposes)
BMC Genomics
Open Access
Research article
In vivo – in vitro toxicogenomic comparison of TCDD-elicited gene
expression in Hepa1c1c7 mouse hepatoma cells and C57BL/6
hepatic tissue
Edward Dere1,2, Darrell R Boverhof1,2,3, Lyle D Burgoon1,2,3 and
Timothy R Zacharewski*1,2,3
Address: 1Department of Biochemistry & Molecular Biology, Michigan State University, East Lansing MI, 48824-1319, USA, 2National Food Safety
& Toxicology Center, Michigan State University, East Lansing MI, 48824-1319, USA and 3Center for Integrative Toxicology, Michigan State
University, East Lansing MI, 48824-1319, USA
Email: Edward Dere - dereedwa@msu.edu; Darrell R Boverhof - boverho5@msu.edu; Lyle D Burgoon - burgoonl@msu.edu;
Timothy R Zacharewski* - tzachare@msu.edu
* Corresponding author
Abstract
Background: In vitro systems have inherent limitations in their ability to model whole organism
gene responses, which must be identified and appropriately considered when developing predictive
biomarkers of in vivo toxicity. Systematic comparison of in vitro and in vivo temporal gene expression
profiles were conducted to assess the ability of Hepa1c1c7 mouse hepatoma cells to model hepatic
responses in C57BL/6 mice following treatment with 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD).
Results: Gene expression analysis and functional gene annotation indicate that Hepa1c1c7 cells
appropriately modeled the induction of xenobiotic metabolism genes in vivo. However, responses
associated with cell cycle progression and proliferation were unique to Hepa1c1c7 cells, consistent
with the cell cycle arrest effects of TCDD on rapidly dividing cells. In contrast, lipid metabolism and
immune responses, representative of whole organism effects in vivo, were not replicated in
Hepa1c1c7 cells.
Conclusion: These results identified inherent differences in TCDD-mediated gene expression
responses between these models and highlighted the limitations of in vitro systems in modeling
whole organism responses, and additionally identified potential predictive biomarkers of toxicity.
Background
Advances in microarray and related technologies continue
to revolutionize biomedical research and are being incor-
porated into toxicology and risk assessment. These tech-
nologies not only facilitate a more comprehensive
elucidation of the mechanisms of toxicity, but also sup-
port mechanistically-based quantitative risk assessment
[1-5]. In addition, these technologies are being used to
develop predictive toxicity screening assays to screen drug
candidates with adverse characteristics earlier in the devel-
opment pipeline in order to prioritize resources and max-
imize successes in clinical trials [6-8]. Comparable
screening strategies are also being proposed to rank and
prioritize commercial chemicals, natural products, and
environmental contaminants that warrant further toxico-
logical investigation. Traditionally, rodent models or sur-
Published: 12 April 2006
BMC Genomics 2006, 7:80doi:10.1186/1471-2164-7-80
Received: 19 January 2006
Accepted: 12 April 2006
This article is available from: http://www.biomedcentral.com/1471-2164/7/80
© 2006 Dere et al; licensee BioMed Central Ltd.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0),
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Page 2

BMC Genomics 2006, 7:80http://www.biomedcentral.com/1471-2164/7/80
Page 2 of 18
(page number not for citation purposes)
rogates for ecologically-relevant species are typically used
in regulatory testing. However, public and regulatory pres-
sure, especially in Europe, seek to minimize the use of ani-
mals in testing [9]. Similar policies in the US, such as the
ICCVAM Authorization Act of 2000, provide guidelines to
facilitate the regulatory acceptance of alternative testing
methods. These initiatives combined with the need to
assess an expanding list of drug candidates and commer-
cial chemicals for toxicity, have increased demand for the
development and implementation of high-throughput in
vitro screening assays that are predictive of toxicity in
humans and ecologically-relevant species.
Various in vitro hepatic models including the isolated per-
fused liver, precision cut liver slices, isolated primary liver
cells and a number of immortalized liver cell lines, have
been used as animal alternatives [10]. In addition to pro-
viding a renewable model, in vitro systems are a cost-effec-
tive alternative and are amenable to high-throughput
screening. These models, particularly immortalized cell
lines, also allow for more in-depth biochemical and
molecular investigations, such as over-expression, knock-
down, activation or inhibition strategies, thus further elu-
cidating mechanisms of action. However, inherent limita-
tions in the ability of cell cultures to model whole
organism responses must also be considered when identi-
fying putative biomarkers for high-throughput toxicity
screening assays, and elucidating relevant mechanisms of
toxicity that support quantitative risk assessment. Despite
several in vitro toxicogenomic reports [11-13], few have
systematically examined the ability of in vitro systems to
predict in vivo gene expression profiles in response to
chemical treatment [10,14].
2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) is a wide-
spread environmental contaminant that elicits a number
of adverse effects including tumor promotion, teratogene-
sis, hepatotoxicity, and immunotoxicity as well as the
induction of several metabolizing enzymes [15]. Many, if
not all of these effects, are due to alterations in gene
expression mediated by the aryl hydrocarbon receptor
(AhR), a basic-helix-loop-helix-PAS (bHLH-PAS) tran-
scription factor [15,16]. Ligand binding to the cytoplas-
mic AhR complex triggers the dissociation of interacting
proteins and results in the translocation of the ligand-
bound AhR to the nucleus where it heterodimerizes with
the aryl hydrocarbon receptor nuclear translocator
(ARNT), another member of the bHLH-PAS family. The
heterodimer then binds specific DNA elements, termed
dioxin response elements (DREs), within the regulatory
regions of target genes leading to changes in expression
that ultimately result in the observed responses [17].
Although the role of AhR is well established, the gene reg-
ulatory pathways responsible for toxicity are poorly
understood and warrant further investigation to assess the
Number of genes differentially regulated (P1(t) > 0.9999 and Ifold changel > 1.5-fold) as measured by microarray analysis for the (A) time course and (B) and dose-response studies in mouse hepatoma Hepa1c1c7 cells
Figure 1
Number of genes differentially regulated (P 1(t) > 0.9999 and Ifold changel > 1.5-fold) as measured by microarray analysis for
the (A) time course and (B) and dose-response studies in mouse hepatoma Hepa1c1c7 cells. For the time course study, cells
were treated with 10 nM TCDD and harvested at 1, 2, 4, 8, 12, 24 or 48 hrs after treatment. Cells for the 12 hr dose-response
study were treated with 0.001, 0.01, 0.1, 1.0, 10 and 100 nM of TCDD
Time course
1 2 4 8 12 24 48
0
50
100
150
time (hrs)
0 0
10-4
20 20
40
40
60
60
80
80
100
100
Dose response
-4 -3
10-3
-2
10-2
-1
10-1
0
1
1 2 3
TCDD concentration (nM)
A) B)

103
10 102
EC 50 = 88.5 pM

Page 3

BMC Genomics 2006, 7:80http://www.biomedcentral.com/1471-2164/7/80
Page 3 of 18
(page number not for citation purposes)
potential risks to humans and ecologically relevant spe-
cies.
Hepa1c1c7 cells and C57BL/6 mice are well-established
models routinely used to examine the mechanisms of
action of TCDD and related compounds. In this study,
TCDD-elicited temporal gene expression effects were sys-
tematically compared in order to assess the ability of
Hepa1c1c7 cells to replicate C57BL/6 hepatic tissue
responses. Our results indicate that several phase I and II
metabolizing enzyme responses are aptly reproduced.
However, many responses were model-specific and reflect
inherent in vitro and in vivo differences that must be con-
sidered in mechanistic studies and during the selection of
biomarkers for developing toxicity screening assays.
Results
In vitro microarray data analysis
Temporal gene expression profiles were assessed in
Hepa1c1c7 wild type cells following treatment with 10
nM TCDD using cDNA microarrays with 13,362 spotted
features. Empirical Bayes analysis of the in vitro time
course data identified 331 features representing 285
unique genes with a P1(t) value greater than 0.9999 at one
or more time points, and differential expression greater
than ± 1.5 fold relative to time-matched vehicle controls.
The number of differentially regulated genes gradually
increased from 1 to 24 hrs, followed by a slight decrease
at 48 hrs (Figure 1A). In vitro dose-response data per-
formed at 12 hrs with TCDD covering 6 different concen-
trations (0.001, 0.01, 0.1, 1.0, 10 and 100 nM), identified
181 features representing 155 unique genes (P1(t) >
0.9999 and an absolute fold change > 1.5 at one or more
doses; Figure 1B). Complete in vitro time course and dose-
response data are available in Additional file 1 and 2,
respectively.
As a control, the gene expression effects elicited by 10 nM
TCDD in ARNT-deficient c4 Hepa1c1c7 mutants [18]
were examined at 1 and 24 hrs (data not shown). Only
ATPase, H+ transporting, V1 subunit E-like 2 isoform 2
(Atp6v1e2) and SUMO/sentrin specific peptidase 6
(Senp6) exhibited a significant change in expression using
the same criteria (P1(t) > 0.9999 and an absolute fold
change > 1.5). Neither Atp6v1e2 nor Senp6 were among
the active genes in wild-type Hepa1c1c7 cells or in C57BL/
6 liver samples [19]. These results provide further evi-
dence that the AhR/ARNT signaling pathway mediates
TCDD-elicited gene expression responses, which are con-
sistent with in vivo microarray results with AhR knockout
mice [20].
Hierarchical clustering of the genes expressed in
Hepa1c1c7 time course indicate that 2 and 4 hrs were
most similar, as were 8 and 12 hrs, and 24 and 48 hrs,
while the 1 hr time point was segregated (Figure 2A). A
strong dose-response relationship was also evident with
clusters sequentially branching out with increasing con-
centration (Figure 2B). At 12 hrs, 117 genes were differen-
tially expressed with 112 exhibiting a dose-dependent
response. Moreover, the fold changes measured in both
the time course and dose-response studies using 10 nM
TCDD were comparable. For example, xanthine dehydro-
genase (Xdh) and NAD(P)H dehydrogenase, quinone 1
(Nqo1) were induced 2.39- and 4.89-fold respectively in
the time course and 2.93- and 4.71-fold in the dose-
response study. There is a strong correlation (R = 0.97)
between the differentially expressed genes at 12 hrs in the
time course with the differentially regulated genes in the
dose-response study at 10 nM, demonstrating the repro-
ducibility between independent studies and providing
further evidence that these genes are regulated by TCDD.
The list of temporally regulated genes was subjected to k-
means clustering using the standard correlation distance
metrics. Five k-means clusters best characterized the data-
set and identified clusters representing A) up-regulated
early and sustained, B) up-regulated intermediate and sus-
tained, C) up-regulated intermediate, D) up-regulated
immediate and E) down-regulated late (Figure 3). These
were comparable to the k-means clusters identified in
hepatic tissue of C57BL/6 mice following treatment with
30 µg/kg TCDD [19]. Although, no discernable functional
category is over-represented in any one cluster, the sus-
tained up-regulation of early (Cluster A) and intermediate
(Cluster B) responding genes include classic TCDD-
responsive genes such as cytochrome P450, family 1, sub-
family a, polypeptide 1 (Cyp1a1), Xdh and Nqo1. Many
down-regulated late genes were associated with cell cycle
regulation such as myelocytomatosis oncogene (Myc).
Additionally, targets of Myc, including cyclin D1 and orni-
thine decarboxylase (Odc1), were also down-regulated
suggesting a mechanism for cell cycle arrest [21-23], a
common in vitro response to TCDD.
Classification of gene expression responses for common
regulated genes
Using the same filtering criteria (P1(t) > 0.9999 and an
absolute fold change > 1.5), 678 features representing 619
unique genes were differentially expressed as previously
reported in a time course study conducted in hepatic tis-
sue from C57BL/6 mice orally gavaged with 30 µg/kg
TCDD [19]. The number of responsive in vivo genes and
their temporal expression patterns closely paralleled the
results from this in vitro study. The fewest number of active
genes was observed at 2 hrs, followed by a large increase
at 4 hrs, which was sustained to 72 hrs. However, the sub-
stantial increase in expressed in vivo genes at 168 hrs was
attributed to triglyceride accumulation and immune cell
infiltration, which was not observed in Hepa1c1c7 cells.

Page 4

BMC Genomics 2006, 7:80http://www.biomedcentral.com/1471-2164/7/80
Page 4 of 18
(page number not for citation purposes)
This list of 619 of in vivo genes served as the basis for sub-
sequent comparisons against TCDD-elicited in vitro
responses.
Comparison of in vitro and in vivo differentially expressed
gene lists identified common and model specific
responses (Figure 4A). TCDD treatment resulted in a total
of 838 regulated genes in either model and with 67 com-
mon to both. TCDD elicited 218 gene expression changes
unique to Hepa1c1c7 cells while 552 genes were specific
to C57BL/6 hepatic samples. Although 67 genes were reg-
ulated in both models, not all possessed similar temporal
patterns of expression. Contingency analysis using a 2 × 2
table and the χ2 test resulted in a p-value < 0.001 (α =
0.05) illustrate a statistically significant association
between the lists of differentially regulated genes in vitro
and in vivo. Further stratification revealed genes that were
either induced in both models (class I), repressed in both
models (class II), induced in vivo while repressed in vitro
(class III), or repressed in vivo while induced in vitro (class
IV; Figure 4B). Genes regulated in a similar fashion in
both models (classes I and II) accounted for 49 of the 67
common active genes, while the remaining genes exhib-
ited divergent expression profiles (classes III and IV).
Hierarchical clustering of the temporal expression values
for the 67 overlapping genes identified the same four
classes (Figure 4C). The pattern across model and time
illustrates that the earliest time points (i.e. 1 hr in vitro and
2 hr in vivo time points) cluster together while the remain-
ing clusters branch into in vitro or in vivo clusters according
to time. These results suggest that potential biomarkers of
acute TCDD-mediated responses may best be predicted by
the immediate-early in vitro gene responses.
In vitro and in vivo induced genes (class I) include xenobi-
otic and oxidoreductase enzymes such as abhydrolase
domain containing 6 (Abhd6), Cyp1a1, dehydrogenase/
reductase (SDR family) member 3 (Dhrs3), Nqo1, pros-
taglandin-endoperoxide synthase 1 (Ptgs1), UDP-glucose
dehydrogenase (Ugdh) and Xdh (Table 1). These genes
have previously been reported to be TCDD-responsive
[19,24], with Cyp1a1 and Nqo1 being members of the
"AhR gene battery" [25]. Glutathione S-transferase, alpha
4 (Gsta4) was also induced in vitro and in vivo, 1.7- and
2.0-fold respectively, consistent with TCDD-mediated
induction of phase I and II metabolizing enzymes. Of the
35 genes responding similarly in both models, approxi-
mately 71% of were similarly up-regulated (class I) while
the remaining genes were repressed across both models
(class II). Repressed class II genes include minichromo-
some maintenance deficient 6 (Mcm6), glycerol kinase
(Gyk) and ficolin A (Fcna) (repressed 1.6-, 1.6- and 1.7-
fold in vitro, respectively). Overall, repressed genes did not
share any common discernable biological function.
Forty-two of the 67 common differentially expressed
genes were dose responsive at 12 and 24 hrs in vitro and in
vivo, respectively, further suggesting the role of the AhR in
mediating these responses. Microarray-based EC50 values
spanned at least 3 orders of magnitude ranging from 0.05
µg/kg to >150 µg/kg in vivo, and 0.00118 nM to 2.4 nM in
vitro (Table 1). Cyp1a1, the prototypical marker of TCDD
exposure, had EC50 values of 0.05 µg/kg and 0.014 nM, in
vivo and in vitro respectively, and was induced 38-fold in
both time course studies. Complete data sets for the in vivo
time course and dose-responses experiments are available
in Additional file 3 and 4.
Of the 67 overlapping genes, 18 exhibited divergent tem-
poral profiles (classes III and IV). Class III contains 12
genes induced in vivo but repressed in vitro, while 6 were
repressed in vivo and induced in vitro (class IV). Example
genes include Myc (class III) and B-cell translocation gene
2 (Btg2, class IV) which are both involved in regulating
cell cycle progression [23,26-30]. Myc was induced 3.7-
fold in vivo and repressed 2.2-fold in vitro, while Btg2 was
repressed 1.8-fold in vivo and induced 1.5-fold in vitro.
In addition to the regulated genes common to both mod-
els, 218 in vitro- and 559 in vivo-specific genes were iden-
tified. Many of the unique in vitro responses are involved
in cell cycle regulation, including cyclins D1 and B2
(Table 2). Cyclin D1, which complexes with cyclin-
dependent kinase 4 (Cdk4) to regulate the progression
from G1 to S phase [31,32], was down-regulated early and
repressed 1.7-fold to 48 hrs. Furthermore, cyclin B2 and
cell division cycle 2 homolog A (Cdc2a) which interact to
form an active kinase required for G2 promotion, were
down-regulated, 1.8-fold and 1.5-fold, respectively. In
addition to cell cycle related genes, UDP glucuronosyl-
transferase 1 family, polypeptide A2 (Ugt1a2), a phase II
metabolizing enzyme, was induced 2.8-fold in vitro, but
not significantly regulated in vivo.
Analysis of the C57BL/6 hepatic time course identified
552 unique genes that were solely regulated in vivo. This
included TCDD induced transcripts for microsomal epox-
ide hydrolase 1 (Ephx1) and carbonyl reductase 3 (Cbr3)
which both function as xenobiotic metabolizing enzymes.
Notch gene homolog 1 (Notch1) and growth arrest spe-
cific 1 (Gas1) which are both associated with develop-
ment and differentiation but serve undetermined roles in
the liver, were also induced by TCDD (Table 3). Genes
related to immune cell accumulation were also specific to
the in vivo study, coincident with immune cell accumula-
tion at 168 hr as determined by histopathological exami-
nation [19].

Page 5

BMC Genomics 2006, 7:80http://www.biomedcentral.com/1471-2164/7/80
Page 5 of 18
(page number not for citation purposes)
Comparison of basal gene expression levels in Hepa1c1c7
cells and hepatic tissue
In order to further investigate differences in gene expres-
sion levels, Hepa1c1c7 cells and C57BL/6 liver samples
were directly compared by competitive hybridization on
the same array, to identify basal gene expression level dif-
ferences. Subsequent linear regression analysis of the
mean normalized signal intensities from the untreated
samples resulted in a correlation value of R = 0.75 (Figure
5), which is consistent with basal gene expression com-
parisons of various in vitro rat hepatic systems against
whole livers, where correlation values decreased between
liver slices (R = 0.97), primary cells (R = 0.85), BRL3A (R
= 0.3) and NRL clone 9 (R = 0.32) rat liver cell lines [10].
Overall, the correlation illustrates reasonable concord-
ance in basal gene expression levels between the two mod-
els. However, data points which deviate from the fitted
line indicate differences in the basal expression of individ-
ual genes between the Hepa1c1c7 cells and hepatic tissue
from C57BL/6 mice. Although there are differences, they
may be negligible if the TCDD-elicited responses are con-
served in vitro and in vivo. Complete microarray data for
the untreated comparisons are available in Additional file
5.
The relative basal expression of the 67 common active fea-
tures was further investigated (Figure 5). In general, class
I (i.e. induced in both models) genes fell close to the
regression line, indicating that the basal expression of
induced genes were comparable as were their in vitro and
in vivo responses to TCDD. In contrast, basal expression
levels of class III genes (i.e. induced in vivo while repressed
in vitro) were generally higher in the Hepa1c1c7 cells,
while levels in class II and IV (i.e. repressed in both mod-
els and repressed in vivo while induced in vitro, respec-
tively) genes were scattered around the fitted linear line in
Figure 5.
Hierarchical clustering of the differentially regulated gene lists for A) temporal and B) dose-response microarray studies in mouse hepatoma Hepa1c1c7 cells
Figure 2
Hierarchical clustering of the differentially regulated gene lists for A) temporal and B) dose-response microarray studies in
mouse hepatoma Hepa1c1c7 cells. The results illustrate time and dose-dependent clustering patterns. From the A) temporal
results, the early (2 hr and 4 hr), intermediate (8 hr and 12 hr) and late (24 hr and 48 hr) time points cluster separately while
the 1 hr time point clusters alone. Results from the B) dose-response show that highest doses clustered together, while the
remaining doses branched out in a dose-dependent manner
4.0
3.0
2.0
1.5
1.0
0.7
0.5
0.3
Expression Ratio
A) B)
0.001 0.01 0.1 1.0 10 100 1 2 4 8 12 24 48
Time (hrs)
Dose (nM)