TOXICOLOGICAL SCIENCES, 143(2), 2015, 277–295
doi: 10.1093/toxsci/kfu248
Advance Access Publication Date: November 18, 2014
Identification of Specific mRNA Signatures as
Fingerprints for Carcinogenesis in Mice Induced by
Genotoxic and Nongenotoxic Hepatocarcinogens
Nadine Kossler*, Katja A. Matheis*,1, Nina Ostenfeldt†, Dorthe Bach Toft†,
Stéphane Dhalluin‡, Ulrich Deschl*, and Arno Kalkuhl*
*Boehringer Ingelheim Pharma GmbH & Co. KG, 88400 Biberach an der Riss, Germany, †H. Lundbeck A/S, 2500
Valby, Denmark and ‡UCB Pharma S.A., 1070 Brussels, Belgium
1
To whom correspondence should be addressed at Boehringer Ingelheim Pharma GmbH & Co. KG, Birkendorfer Straße 65, 88397 Biberach an der Riss,
Germany. Fax: þ0049 (0) 7351 54 98686. E-mail: [email protected].
ABSTRACT
Long-term rodent carcinogenicity studies for evaluation of chemicals and pharmaceuticals concerning their carcinogenic
potential to humans are currently receiving critical revision. Additional data from mechanistic studies can support cancer
risk assessment by clarifying the underlying mode of action. In the course of the IMI MARCAR project, a European
consortium of EFPIA partners and academics, which aims to identify biomarkers for nongenotoxic carcinogenesis, a
toxicogenomic mouse liver database was generated. CD-1 mice were orally treated for 3 and 14 days with 3 known
genotoxic hepatocarcinogens: C.I. Direct Black 38, Dimethylnitrosamine and 4,40 -Methylenedianiline; 3 nongenotoxic
hepatocarcinogens: 1,4-Dichlorobenzene, Phenobarbital sodium and Piperonyl butoxide; 4 nonhepatocarcinogens:
Cefuroxime sodium, Nifedipine, Prazosin hydrochloride and Propranolol hydrochloride; and 3 compounds that show
ambiguous results in genotoxicity testing: Cyproterone acetate, Thioacetamide and Wy-14643. By liver mRNA expression
analysis using individual animal data, we identified 64 specific biomarker candidates for genotoxic carcinogens and 69 for
nongenotoxic carcinogens for male mice at day 15. The majority of genotoxic carcinogen biomarker candidates possess
functions in DNA damage response (eg, apoptosis, cell cycle progression, DNA repair). Most of the identified nongenotoxic
carcinogen biomarker candidates are involved in regulation of cell cycle progression and apoptosis. The derived biomarker
lists were characterized with respect to their dependency on study duration and gender and were successfully used to
characterize carcinogens with ambiguous genotoxicity test results, such as Wy-14643. The identified biomarker candidates
improve the mechanistic understanding of drug-induced effects on the mouse liver that result in hepatocellular adenomas
and/or carcinomas in 2-year mouse carcinogenicity studies.
Key words: toxicogenomics; genotoxic carcinogens; nongenotoxic carcinogens; mouse liver; biomarker candidates; individual
animal fold changes
Evaluation of the carcinogenic potential of new drugs is a key
part of the drug development process. Drug-induced carcinogenicity can be caused by genotoxic as well as nongenotoxic
mechanisms. Genotoxic carcinogens (GCs) directly interact with
DNA as either parent chemical or reactive metabolite. They are
considered to be detectable in the standard genotoxicity battery
of short-term tests including Ames test, mouse lymphoma assay, in vitro micronucleus and/or chromosomal aberration test
(ICH guidance on genotoxicity testing S2 (R1), 2011). Despite its
broad application, the current classification process for carcinogenic substances is questionable. The specificity of the standard
genotoxicity battery was shown to be highly debatable and a
C The Author 2014. Published by Oxford University Press on behalf of the Society of Toxicology.
V
All rights reserved. For Permissions, please e-mail: [email protected]
277
278
|
TOXICOLOGICAL SCIENCES, 2015, Vol. 143, No. 2
large number of false-positive results during the safety assessment of new chemicals and drugs might be the result (Kirkland
et al., 2005, 2006). Therefore, it has been suggested that the classification of several compounds as “Ames test negative GCs”
would be more accurate than classification as “non-genotoxic
carcinogens” (NGCs) (Waters et al., 2010).
Contrary to the DNA-reactive carcinogens, NGCs are thought
to stimulate tumor formation or promotion by secondary or epigenetic mechanisms, primarily by promoting cell survival and
proliferation of the already mutated cell, or indirectly enhance
the cell mutation rate via reactive oxygen species (ROS) (Shaw
and Jones, 1994). Multiple mechanisms such as increased secretion of trophic hormones (eg, estrogens), receptor activation (eg,
peroxisome proliferators), enhanced oxidative stress, inhibition
of apoptosis, or immunosuppression, were suggested as root
causes of nongenotoxic carcinogenesis (Benigni et al., 2013;
Lima and Van der Laan, 2000). These processes are not yet completely understood. Currently, the 2-year carcinogenesis bioassay in rats and mice is the standard test to identify potential
NGC effects for humans. The obvious disadvantage of this test
is the requirement for large numbers of animals, high costs,
and its long duration.
Understanding the mechanism by which genotoxic and nongenotoxic effects may be induced, as well as identification of
mechanism-based biomarkers after short-term treatment,
would greatly improve the risk assessment of new drugs and
the evaluation of whether an effect is relevant for the whole animal or human. Gene expression profiling reflects underlying
modes of mechanism, providing additional information that
might be helpful in predicting the carcinogenic potential of substances in conjunction with conventional short-term mutagenicity assays.
For the rat, several mechanistic analyses have been performed and relevant biomarker candidate gene signatures identified that appear to discriminate direct versus indirect GCs,
NGCs and noncarcinogens in short-term assays (Waters et al.,
2010). Liu and colleagues applied toxicogenomics in a shortterm study to monitor toxicological events and to predict nongenotoxic hepatocarcinogenicity (Liu et al., 2011). Further, a rat
toxicogenomics approach has been used, including data from
various known carcinogenic compounds and time points derived from the publicly available Japanese TG-GATEs database
with the aim of selecting biomarker genes for hepatocarcinogenesis (Fielden et al., 2011; Gusenleitner et al., 2014; Uehara
et al., 2011). For the mouse, as the second rodent species used
for 2-year carcinogenicity testing, genomics data with mechanistic analysis of GCs and NGCs are very limited so far (Melis
et al., 2014; Thomas et al., 2007a).
Therefore, within the IMI MARCAR project a genomic mouse
database was generated (for further information see Gene
Expression Omnibus, GEO Accession number: GSE44783).
Substances that show ambiguous results in genotoxicity testing,
such as Cyproterone acetate (CPA), Thioacetamide (TAA) and
Pirinixic acid (Wy-14643) were termed as “unclassified” compounds (UCs). Concerning dose selection for these studies, a low
multiple of the hepatocarcinogenic dose in the respective
2-year mouse cancer study was used. In addition, doses were
selected to induce exaggerated pharmacology or toxicological
changes after 14 days of administration. In this study, we focused on liver mRNA expression profiling and mechanistic
analysis of male mice dosed for 14 days using Affymetrix
Microarray. For identification of mRNA candidate biomarkers
from animal studies, several publications describe the use
of toxicogenomics analysis by machine learning methods
(Melis et al., 2014; Nie et al., 2006). These methods have limitations, as they might lead to the identification of unspecific biomarkers in those cases where gene expression data are derived
from studies with unequipotent dose selection or equivocal carcinogenic compound classification. Furthermore, individual differences between animals in their susceptibility to developing
cancer are not considered by these methods. Therefore, we
used an approach based on individual animal values and identified specific and mechanistically relevant mouse biomarker
candidates for GCs and NGCs in the liver of male mice at day 15.
The derived biomarker candidates were then compared to biomarker candidates of male mice at day 4 and female mice at
days 4 and 15 in order to gain insights in their dependency on
time and gender. Finally, they were applied to investigate the
gene expression profile of CPA, TAA, and Wy-14643.
MATERIALS AND METHODS
Animal treatment. Groups of 5–6 male and female CD-1 mice (CDR
1V
Mouse Crl: CD-1(ICR)), purchased from Charles River Wiga
GmbH in Germany, were orally treated for 3 and 14 days with
CIDB, DMN, MDA, DCB, PB, PBO, CFX, Nif, Praz, Prop, CPA, TAA,
Wy-14643, or vehicle. These compounds were classified as GCs,
NGCs, or NHCs as indicated in the Table 1. Three compounds
for which an appropriate classification were considered uncertain were included and termed as UCs. For compounds categorized as GC and NGCs, the selected dose was a low multiple of
the hepatocarcinogenic dose tested in the 2-year mouse carcinogenicity study (about 10-fold TD50). In addition, the selected
dose was expected to induce exaggerated pharmacology or toxicological changes in the liver after 14 days of administration
(Table 2). For UCs dose selection, histopathology data and a
multiple of the hepatocarcinogenic dose were considered. For
compounds categorized as NHCs, a dose of 10- to 15-fold of the
human therapeutic dose was selected. The administration volume was 10 ml/kg. As vehicles, carboxymethylcellulose (CMC)
or corn oil (CO) were used. All information regarding compound
class, compound name, CAS number, vehicle, information
about compound doses from literature and databases, as well
as the applied compound dose in this study are all listed in
Table 1. After treatment, the animals were anaesthetized with
isoflurane, exsanguinated, and subjected to necropsy. The animal study has also been briefly described in a previous publication (Eichner et al., 2013).
Tissue sampling, RNA preparation, and Affymetrix Microarray hybridization. The left lateral liver lobe was dissected in 4–5 mm sized
cubes, placed in Wheaton Cryovials, snap frozen and stored
under 80 C conditions. Single animal liver samples were
homogenized with Qiazol (Qiagen, Hilden, Germany). Isolation
of total RNA was performed using the miRNeasy 96 Kit (Qiagen,
Hilden, Germany) according to the manufacturer’s instructions.
Further steps including cRNA synthesis, target hybridization,
washing, staining, and subsequent probe array scanning were
performed using the 30 IVT Express Kit according to the manufacturer’s instructions (User Manual Affymetrix). Finally, microarray analysis was performed for each sample using the
GeneChip Mouse Genome 430 2.0 Array from Affymetrix. The
data were analyzed using Genedata Expressionist Refiner Array
(Version 7.5, Genedata AG, 4016 Basel, Switzerland) and
Genedata Expressionist Analyst (Version 7.5.4d) software. Raw
data and results were stored by Genedata Expressionist
Database Browser software (Version 7.5) with Oracle Database
11.2.0.1.
KOSSLER ET AL.
|
279
TABLE 1. Compounds
Compound
Class
Compound Name
(abbreviation), CAS Number
Vehicle
Mouse TD50 and Compound Doses Tested in Long-Term
Animal Studies (Databases /References from Literature)
Compound Dose
(mg/kg/day)
GC
C.I. Direct Black 38 (CIDB),
1937-37-7
CO
2500
Dimethylnitrosamine
(DMN), 62-75-9
4,40 -Methylenedianiline
(MDA), 101-77-9
CO
TD50: 71.6 mg/kg/day (CPDB); 2500 mg/kg/day for 13
weeks leads liver degeneration in 3/10 animals
(NCTP)
TD50: 0.189 mg/kg (CPDB)
1,4-Dichlorobenzene (DCB),
106-46-7
CO
Phenobarbital sodium (PB),
57-30-7
CMC
Piperonyl butoxide (PBO),
51-03-6
CMC
Cefuroxime sodium (CFX),
56238-63-2
Nifedipine (Nif), 21829-25-4
CMC
NGC
NHC
UC
Prazosin hydrochloride
(Praz), 19237-84-4
Propranolol hydrochloride
(Prop), 318-98-9
Cyproterone acetate (CPA),
427-51-0
CO
CMC
CMC
CMC
CO
Thioacetamide (TAA),
62-55-5
CMC
Pirinixic acid (Wy-14643),
50892-23-4
CMC
TD50: 32.4 mg/kg/day (CPDB); 29.4 mg/kg for 24 months
induced liver hpc in 6/50 animals (CPDB); 24.5 mg/kg
for 24 months resulted in liver hpc in 33/50 animals
(CPDB) data refer to 4,40 -Methylenedianiline dihydrochloride (CAS number: 13552-44-8)
TD50: 323 mg/kg/day (CPDB); 212 mg/kg for 24 months
resulted in liver hpc in 5/50 animals (CPDB); 212 mg/
kg for 24 months resulted in liver hpc in 11/50 animals (CPDB); 31.3 mg/kg for 24 months induced liver
hpa in 9/49 animals and liver hpc in 17/49 animals
(CPDB); Nagano et al. (1998) (37,6 mg/kg for 24 months
resulted in liver hpa in 10/50 animals and liver hpc in
4/50 animals)
TD50: 29.7 mg/kg/day (CPDB); 85 mg/kg/day for 24
months resulted in hpc formation in 6/125 animals
(CPDB); 85 mg/kg/day for 80 weeks resulted in hpc
formation in 4/5 animals (CPDB); 56 mg/kg/day for 60
weeks resulted in hpc formation in 4/22 animals
(CPDB); 85 mg/kg/day for 91 weeks resulted in hpc
formation in 26/90 animals (CPDB); Herren-Freund
et al. (1987) (83.3mg/kg for 62 weeks induced liver hpa
in 2/22 animals); Evans et al. (1992) (85 mg/day for 61
weeks induced liver hpc in 2/20 animals)
TD50: 291 mg/kg/day (CPDB); Butler et al. (1998) (300 mg/
kg for 79 weeks induced liver hpa in 12/60 animals);
Takahashi et al. (1994) (720 mg/kg for 52 weeks
induced liver hpa in 7/53 animals and liver hpc in 6/53
animals)
4- to 10-fold of human dose (mg/kg); Max. 1000 mg/day
human (15 mg/kg); 17-fold of human dose
5- to 10-fold of human dose (mg/kg); Max. 120 mg/day
human (1.7 mg/kg); 30-fold of human dose
25-fold of human dose (mg/kg); Max. 40 mg/day
(0.6 mg/kg); 8-fold of human dose
Max. 320 mg/day (4.6 mg/kg); 17-fold of human dose
TD50: 21.9 mg/kg/day (CPDB); 96 mg/kg for 24 months
resulted in liver hpa in 7/39 animals and liver hpc in
12/39 animals (CPDB); Tucker et al. (1996) (104 mg/kg/
day for 97 weeks induced liver hpa in 2/37 animals
and liver hpc in 8/37 animals)
TD50: 8.81 mg/kg/day (CPDB); Akao et al. (1990)
(19.1 mg/kg/day for 40 weeks induced liver hpc in
11/24 animals)
TD50: <10.8 mg/kg/day (CPDB); Reddy et al. (1979)
(76.3 mg/kg/day for 53 weeks induced liver hpc in
18/18 animals)
2
75
600
80
600
250
50
5
80
160
20
200
Long-term toxicity test results and the TD50 rate were used as references for GCs, NGCs, and UCs dose selection. A dose reflects a multiple of the hepatocarcinogenic
dose known to induce hepatocellular carcinomas (hpc) and hepatocellular adenomas (hpa) in a 2-year mouse cancer study (about 10-fold of TD50). For NHCs a 10-fold
multiple of the human therapeutic dose was selected. As vehicles CMC or CO were used.
Preprocessing of raw hybridization data. For preprocessing of raw
hybridization data, the multichip condensing method “GCRMA
Condensing Algorithm” was applied. For background correction,
GCRMA adjusts intensities, including optical noise and nonspecific binding.
Fold change calculation. For statistical analyses, a two-sided
Student’s t-Test (P .05) was applied and q-values were
adjusted for multiple comparisons using false discovery rate
(FDR) according to Benjamini and Hochberg. In addition, the
ratios of medians were calculated for each treatment group
280
|
TOXICOLOGICAL SCIENCES, 2015, Vol. 143, No. 2
TABLE 2. Histopathology Results of Liver Tissue
Compound
Class
Compound
Abbreviation
Compound
Dose (mg/kg/day)
Histopathological Changes Considered to
be Treatment Related (Number of Animals)
GC
CIDB
DMN
MDA
2500
2
75
NGC
DCB
PB
PBO
CFX
Nif
Praz
Prop
CPA
TAA
600
80
600
250
50
5
80
160
20
Wy-14643
200
6/6 Occasional to generalized deposition of yellow-brown pigment in Kupffer cells
2/6 Foci of paravascular mononuclear cell infiltration associated with necrotic cells
6/6 Scattered peribiliary inflammatory cell infiltration, consisting predominantly
of mononuclear cells in 1/6 associated with pigment-laden cells
5/6 Periacinar hepatocellular hypertrophy
6/6 Periacinar hepatocellular hypertrophy
1/6 Small focal hepatocellular necrosis in 1 area
6/6 N.a.d.
5/5 N.a.d.
6/6 N.a.d.
6/6 N.a.d.
3/6 Generalized, predominantly centroacinar, vacuolization of hepatocellular cytoplasm
6/6 Periacinar homogenization of hepatocellular cytoplasm and scattered single cell necrosis.
Many or scattered mitoses
6/6 Generalized hepatocellular hypertrophy with increased density and granular appearance
of the cytoplasm
4/6 Occasional or scattered single cell necrosis
4/6 Many or scattered mitoses
NHC
UC
Male mice were treated with GCs (CIDB, DMN, or MDA), NGCs (DCB, PB, or PBO), NHCs (CFX, Nif, Praz, or Prop), or UCs (CPA, TAA, or Wy-14643). After 14 days of administration liver was dissected and histopathologically examined. Liver changes in treated animals are listed in the table. N.a.d., nothing adverse detected.
with their respective control group. We considered probe sets
showing a fold change deregulation (1.5 FC or FC 1.5) in the
treatment group compared with the respective control group
(based on group medians) with a q-value of 0.1.
Individual animal fold change calculation and definition of criteria for
biomarker candidate detection. The evaluation of mRNA signatures
was performed on the basis of gene expression deregulations in
individual animals, compared with the median of the corresponding vehicle control group for both genders and time
points. Detailed analysis results are only described for male
mice administered for 14 days.
• Identification of GC biomarker candidates: a gene expression
change detected by a probe set was considered to be a potential
biomarker candidate for GCs when the following criteria were
fulfilled: the deregulation of the probe set had to be 1.5-fold
higher or lower in at least 50% of GC-treated animals when compared with the median of the corresponding vehicle control
group. The male mice day 15 GC group included a total of 18 animals treated with CIDB, DMN, or MDA. Therefore, 50% corresponds to 9 animals of this group. Furthermore, an expression
change of the GC candidate biomarker probe sets should not be
observed in more than 1 animal of the NGC group (16 animals
treated with DCB, PB, or PBO) or the NHC group (23 animals
treated with CFX, Nif, Praz, or Prop) (Table 3).
• Identification of NGC biomarker candidates: a gene expression
change detected by a probe set was considered to be a potential
biomarker candidate NGCs if the deregulation was 1.5-fold
higher or lower in at least 50% of NGC-treated animals when
compared with the median of the corresponding vehicle control
group. In all, 16 animals of the male mice day 15 NGC group
were treated with DCB, PB, or PBO. Thus, 50% corresponds to 8
animals of this carcinogenic class. Furthermore, gene expression
deregulation of the NGC candidate biomarker probe sets
should not be seen in more than 1 animal of the GC group
(18 animals administrated with CIDB, DMN, or MDA) or the NHCtreated animals (23 mice treated with CFX, Nif, Praz, or Prop)
(Table 3).
Categorization of probe sets.. All detected probe sets from male
mice administered for 14 days were specified with respect to
their biological function (Tables 4 and 5). Based on our own literature search, Ingenuity Pathway Analysis (IPA) and wording
from Doktorova and Ellinger-Ziegelbauer, we used different toxicological terms to categorize them (Doktorova et al., 2012;
Ellinger-Ziegelbauer et al., 2004). All “unknown” probe set IDs
were uploaded in NetAffxTM Analysis Center (Affymetrix) and
annotated with their transcript assignments (Tables 4 and 5).
RESULTS
Histopathology
To investigate whether the selected compound doses induced
primary cellular changes in the liver, histopathology examinations were performed (Table 2). Among the GCs, CIDB induced
occasionally to generalized deposition of yellow-brown pigment
in the Kupffer cells in all treated animals. After DMN treatment,
foci of paravascular mononuclear cell infiltration associated
with some necrotic cells were detected. All mice given MDA
exhibited scattered peribiliary inflammatory cell infiltration,
consisting predominantly of mononuclear cells. Periacinar hepatocellular hypertrophy was the common denominator for the
lesions in groups administered with the NGCs PB or DCB. In 1
animal administered with the NGC PBO, focal necrosis was
observed after 14 days. In 3 animals given CPA, liver changes in
terms of generalized cytoplasmic vacuolization, predominantly
located in centro-acinar hepatocytes were detected, which
might lead to glycogen accumulation. We found that Wy-14643
treatment resulted in generalized hepatocellular hypertrophy
and increased density and granular appearance of the cytoplasm in all animals. These changes were accompanied by
occasional or scattered single cell necrosis and mitotic activity
in most of the animals, indicating a treatment-related effect. All
mice treated with TAA showed periacinar homogenization of
hepatocellular cytoplasm, scattered single cell necrosis and
many or scattered mitoses. In conclusion, applied doses of GC,
NGC, and UCs resulted in initial treatment-related cellular
KOSSLER ET AL.
|
281
TABLE 3. Criteria for Biomarker Candidates
Biomarker
Candidates
Number of GC-treated Animals
Showing a Biomarker Candidate
Fold Change Deregulation 1.5 FC
or FC 1.5 (Corresponding %)
Number of NGC-treated Animals Showing
a Biomarker Candidate Fold Change
Deregulation 1.5 FC or FC 1.5
(Corresponding %)
Number of NHC-treated
Animals Showing a
Biomarker Candidate Fold
Change Deregulation
1.5 FC or FC 1.5
(Corresponding %)
GC
NGC
9 out of 18 (50%)
1 out of 18 (5.6%)
1 out of 16 (6.3%)
8 out of 16 (50%)
1 out of 23 (4.3%)
1 out of 23 (4.3%)
A total of 18 mice were administrated with GCs (CIDB, DMN, or MDA). Biomarker candidates for GCs (1) have to be deregulated (1.5 FC or FC 1.5) in at least 50% of
all GC-treated animals (corresponding to 9 animals) and (2) a deregulation of the probe set should not be seen in more than 1 animal of the NGC (corresponding to
6.3%) and NHC (corresponding to 4.3%) treated groups, respectively. In all, 16 animals were treated with NGC compounds (DCB, PB, or PBO). A potential NGC biomarker
candidate have (1) to show a gene expression change (1.5 FC or FC 1.5) in at least 50% of all NGC administrated mice (corresponding to 8 animals) and (2) the probe
set deregulation should not be seen in more than 1 animal of GC (corresponding to 5.6%) and NHC (corresponding to 4.3%) treated groups, respectively.
effects in liver after short-term administration. In contrast, no
lesions indicating a treatment-related effect were found in any
of the animals given NHCs such as CFX, Nif, Prop, or Praz.
Number of Statistically Significant Deregulated Probe Sets
Initially, the number of statistically significant deregulated
probe sets in the different compound groups was determined.
Thereby, we considered gene expression deregulation of probe
sets with 1.5-fold change higher or lower than the median of
the corresponding vehicle group and calculated the FDR (q 0.1)
(Supplementary Table 1). For GC and NGC groups, we calculated
some significantly deregulated probe sets, ranging between 28
and 517 in number. In contrast, no significant gene expression
changes were observed in NHC-treated mice with the exception
of CFX. Compounds of the UCs group exhibited a higher number
of significantly deregulated probe sets, indicating a stronger
compound-related effect. Altogether, the compound treatmentrelated effects detected by the number of significantly deregulated probe sets showed high variability between the various
compound administrations indicating the difficulty to select
equipotent dose levels for gene expression analysis based on
classical toxicological measures.
Functional Analysis of GC Biomarker Candidates
To meet concerns about the use of classical statistical
approaches for toxicogenomics evaluation, an individual animal fold change calculation approach was applied for this
study. A representative list of 64 differentially regulated probe
sets as biomarker candidates for GCs (day 15) is shown in Figure
1. Fifty-five out of 64 probe sets displayed a fold change deregulation FC 1.5 when compared with the median of the corresponding vehicle control group, highlighted in red. Nine probe
sets showed a deregulation with a FC 1.5 when compared
with the median of the corresponding vehicle controls, highlighted in blue. Seven out of 64 probe sets have no assigned
gene symbol and 14 probe sets are currently unknown in function (Fig. 1 and Table 4). Some of the “Unknown” probe sets
were identified as noncoding RNA elements including fulllength transcript noncoding RNAs and predicted noncoding
elements (Table 4). The identified GC biomarker candidates
were analyzed in a principal component analysis (PCA)
(Supplementary Table 2A). The PCA plot shows a clear separation of GC compound expression data when compared with
NGC and NHC compound expression data.
The identified list of GC biomarker candidates confirms the
validity of our approach, as the function of the corresponding
genes involves known mechanisms of GC carcinogenicity.
Within the “DNA damage response” genes, we identified “Cell
cycle progression” involved probe sets including Bcor, Ccng1,
Emp3, Iqgap1 (2 probe sets), Top2a, Tspan13, and Zeb2 in response
to GC exposure. In detail, the cellular growth and proliferation
markers Ccng1, Emp3 and Iqgap1 have been described in several
publications as proliferative cellular responses in context with
tumorigenesis (Alaminos et al., 2005; Jensen et al., 2003; White
et al., 2009).
A modification of DNA replication genes in response to GC
administration became apparent with the upregulation of
Top2A and the downregulation of Bcor gene expression. An
induction of Top2A gene expression in response to GC exposure
was also reported by Koufaris and colleagues, demonstrating an
important role for the nuclear enzyme in the processes of DNA
replication, transcription, recombination, and chromatin
remodeling (Koufaris et al., 2012; Watt and Hickson, 1994).
Further, within the “DNA damage response” genes we detected
“Apoptosis” probe sets including Bax, Bcl2a1, Zak, Phlda3, and
Siva1 in response to all 3 GCs. Interestingly, Bax, Siva1, and
PHLDA3 have been directly linked to tumor suppressor gene p53
signaling (Barkinge et al., 2009; Kawase et al., 2009; Miyashita
and Reed, 1995). In addition, it has been shown previously that
BAX and PHLDA3 are induced by GC stress (Kawase et al., 2009;
Watanabe et al., 2012). Genes described to have functions in
“Oxidative stress” responses such as Enc1 and Ddit4l, as well as
the “DNA repair” gene Mgmt, have also been specified as “DNA
damage response” genes. Several studies with DNA-damaging
agents have been shown to induce MGMT associated repair
activity via p53 signaling upon treatment with GCs (eg,
Grombacher and Kaina, 1995). Notably, Bax, Ccng1, and Mgmt,
candidate genes were also detected as candidate genes for GC in
a short-term rat study performed by Ellinger-Ziegelbauer et al.
(2004).
A couple of derived probe sets such as Col1a2, Fbn1, Fstl1,
Loxl2, Nisch, Plekha2, Tagln2, Tuba1A, and Tmsb10/Tmsb4x that
are grouped within the category “Cellular assembly and organization” were deregulated in all 3 GC-treated animal groups,
which might indicate a stress response initiating cell protection
and resulting in cytoskeletal reorganization and stabilization.
Among them, Loxl2, Plekha2, and Tmsb10/Tmsb4x have been
described as cell migration genes that are mainly involved in
the development of tumors (Cano et al., 2012; Li et al., 2013;
Sribenja et al., 2013). TUBA1A, Plekha2, and Nisch affect the cellular organization of the actin cytoskeleton (Alahari et al., 2000; Li
et al., 2013; Spalice et al., 2009). Meanwhile, Ccr2, Cd34, Fgl2, H2DMa, H2-DMb2, Lck, and Mbl2 are grouped as “Immune
response” genes and deregulation of those genes is in
Immune response
Cellular assembly
and organization
Bax
Bcl2a1
Siva1
Zak
DNA damage
response
H2-DMb2
H2-DMa
Fgl2
Ccr2
Mbl2
Cd34
Lck
Tagln2
Fstl1
Col1a2
Fbn1
Loxl2
Tmsb10/Tmsb4x
Tuba1A
Nisch
Plekha2
Phlda3
Bcor
Ccng1
Top2a
Emp3
Iqgap1
Iqgap1
Tspan13
Zeb2
Mgmt
Ddit4l
Enc1
Gene Symbol
Toxicological Category
Major histocompatibility complex, class II, DM beta
Major histocompatibility complex, class II, DM alpha
Fibrinogen-like 2
Chemokine (C-C motif) receptor 2
Mannose-binding lectin (protein C) 2, soluble
CD34 molecule
Lymphocyte-specific protein tyrosine kinase
Collagen, type I, alpha 2
Fibrillin 1
Lysyl oxidase-like 2
Thymosin beta 4, X-linked
Tubulin, alpha 1a
Nischarin
Pleckstrin homology domain containing, family A
(phosphoinositide binding specific) member 2
transgelin 2
Follistatin-like 1
BCL2-associated X protein
BCL2-related protein A1
SIVA1, apoptosis-inducing factor
Sterile alpha motif and leucine zipper containing
kinase AZK
Pleckstrin homology-like domain, family A, member 3
BCL6 corepressor
Cyclin G1
Topoisomerase (DNA) II alpha 170kDa
Epithelial membrane protein 3
IQ motif containing GTPase activating protein 1
IQ motif containing GTPase activating protein 1
Tetraspanin 13
Zinc finger E-box binding homeobox 2
O-6-Methylguanine-DNA methyltransferase
DNA-damage-inducible transcript 4-like
Ectodermal-neural cortex 1 (with BTB-like domain)
Gene Name
1418638_at
1422527_at
1421855_at
1421186_at
1418787_at
1416072_at
1425396_a_at
1426529_a_at
1448259_at
1450857_a_at
1460208_at
1431004_at
1437185_s_at
1418884_x_at
1433757_a_at
1417288_at
1449002_at
1429438_at
1420827_a_at
1454694_a_at
1417104_at
1434998_at
1417380_at
1418643_at
1454200_at
1421309_at
1444139_at
1420965_a_at
1416837_at
1419004_s_at
1452020_a_at
1435029_at
Affymetrix
Probe Set ID
Antigen presentation
Antigen presentation
Antigen presentation
Chemotaxis
Complement activation/association with cancer
Others/association with cancer
T cell maturation
Tissue development
Tissue development
Connective tissue formation
Connective tissue formation
Connective tissue formation
Cytoskeleton
Cytoskeleton
Cytoskeleton/association with cancer
Cytoskeleton/association with cancer
Apoptosis/association with cancer
Cell cycle progression
Cell cycle progression
Cell cycle progression
Cell cycle progression/association with cancer
Cell cycle progression/association with cancer
Cell cycle progression/association with cancer
Cell cycle progression/association with cancer
Cell cycle progression/association with cancer
DNA repair
Oxidative stress
Oxidative stress
Apoptosis/association with cancer
Apoptosis/association with cancer
Apoptosis
Apoptosis
Biological Function
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
;
:
:
:
:
:
:
:
:
:
:
:
:
:
:
Direction of
Deregulation
|
TABLE 4. Toxicological Categories and Biological Functions of GC Biomarker Candidates
282
TOXICOLOGICAL SCIENCES, 2015, Vol. 143, No. 2
Ces2e
G6pd
Moxd1
Exoc4
Ccdc80
Ltn1
Zfp54
Snx6
Akap13
Srprb
Atp6v1d (includes EG:299159)
Atp6v1d (includes EG:299159)
Atp6v1d (includes EG:299159)
Acot9
Cox6b2
Zdhhc14
Zdhhc14
Ggta1
Znf878
9330175E14Rika
9230114K14Rika
Pqlc3
Dleu2a,b
C9orf116
BC021614
Detoxification
response
Others
Unknown
Zinc finger protein 878
RIKEN cDNA 9330175E14 gene
RIKEN cDNA 9230114K14 gene
PQ loop repeat containing 3
Deleted in lymphocytic leukemia, 2
Chromosome 9 open reading frame 116
cDNA sequence BC021614
Glucose-6-phosphate dehydrogenase
Monooxygenase, DBH-like 1
Exocyst complex component 4
Coiled-coil domain containing 80
Listerin E3 ubiquitin protein ligase 1
Zinc finger protein 54
Sorting nexin 6
A kinase (PRKA) anchor protein 13
Signal recognition particle receptor, B subunit
ATPase, Hþ transporting, lysosomal 34kDa, V1 subunit D
ATPase, Hþ transporting, lysosomal 34kDa, V1 subunit D
ATPase, Hþ transporting, lysosomal 34kDa, V1 subunit D
Acyl-CoA thioesterase 9
Cytochrome c oxidase subunit VIb polypeptide 2 (testis)
Zinc finger, DHHC-type containing 14
Zinc finger, DHHC-type containing 14
Glycoprotein, alpha-galactosyltransferase 1 pseudogene
Carboxylesterase 2E
Gene Name
1460369_at
1460134_at
1456643_at
1424906_at
1427411_s_at
1428705_at
1424953_at
1438245_at
1441052_at
1442233_at
1442905_at
1447527_at
1457814_at
1458249_at
1448354_at
1422643_at
1422686_s_at
1424186_at
1452611_at
1419239_at
1425148_a_at
1443923_at
1450089_a_at
1416951_a_at
1416952_at
1438993_a_at
1418073_at
1435275_at
1438975_x_at
1437614_x_at
1418483_a_at
1427137_at
Affymetrix
Probe Set ID
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Carbohydrate metabolism
Catecholamine metabolism
Cellular signaling
Cellular signaling/association with cancer
Cotranslational protein quality control
Embryonic development
Intracellular trafficking
Intracellular trafficking
Intracellular trafficking/association with cancer
Intracellular trafficking/association with cancer
Intracellular trafficking/association with cancer
Intracellular trafficking/association with cancer
Lipid metabolism
Mitochondrial respiration
Others/association with cancer
Others/association with cancer
Pseudogene
Drug metabolism
Biological Function
:
:
:
:
;
:
:
:
:
:
;
:
:
;
:
;
:
:
;
:
:
:
;
:
:
:
:
:
;
;
:
:
Direction of
Deregulation
Noncoding RNA elements within the group “Unknown” by predicted noncoding element.
Noncoding RNA elements within the group “Unknown” by full-length noncoding element.
a
b
(17 probe sets), and “Unknown” (14 probe sets).
Derived GC biomarker candidates (including Affymetrix Probe set ID, gene symbol, gene name, and probe set deregulation labeled by an arrow) were assigned to a toxicological category and the probe set function is specified. The
toxicological categories contain the following terms “DNA damage response” (16 probe sets), “Cellular assembly and organization” (9 probe sets), “Immune response” (7 probe sets), “Detoxification response” (1 probe set), “Others”
a
Gene Symbol
Toxicological Category
TABLE 4. (continued)
KOSSLER ET AL.
|
283
Nebulette
Nebl
Hip1r
Akr1d1
Adrbk2
Fndc5
Tmem98
Slc25a32
Tulp2
Rasal2
Camkk2
Cebpb (includes EG:1051)
Atxn10
Gnat1
Pkp2 (includes EG:287925)
Cellular assembly and
organization
Immune response
Others
Plakophilin 2
Cyp2c50
Cyp2c65
Gstm1
Ces2a
Detoxification response
CCAAT/enhancer binding protein (C/EBP), beta
Ataxin 10
Guanine nucleotide binding protein (G protein), alpha
transducing activity polypeptide 1
Tubby like protein 2
RAS protein activator like 2
Calcium/calmodulin-dependent protein kinase
kinase 2, beta
Adrenergic, beta, receptor kinase 2
Fibronectin type III domain containing 5
Transmembrane protein 98
Solute carrier family 25 (mitochondrial folate
carrier), member 32
Huntingtin interacting protein 1 related
Aldo-keto reductase family 1, member D1
(delta 4-3-ketosteroid-5-beta-reductase)
Cytochrome P450, family 2, subfamily C, polypeptide 18
Cytochrome P450, family 2, subfamily C, polypeptide 19
Glutathione S-transferase mu 5
Carboxylesterase 2A
Tnfrsf1b
Pgap2
Pgap2
Nolc1
Apoptosis
Fgl1
Aig1
Rorc
Tumor necrosis factor receptor superfamily, member 1B
Post-GPI attachment to proteins 2
Post-GPI attachment to proteins 2
Nucleolar and coiled-body phosphoprotein 1
NSL1, MIND kinetochore complex component, homolog
(Saccharomyces cerevisiae)
Non-SMC condensin II complex, subunit G2
ATPase family, AAA domain containing 2
TEA domain family member 1 (SV40 transcriptional
enhancer factor)
Fibrinogen-like 1
Androgen-induced 1
RAR-related orphan receptor C
Nsl1 (includes EG:25936)
1425553_s_at
1425771_at
1440801_s_at
1453135_at
1424133_at
1453149_at
1417276_at
1436910_at
1455401_at
1427844_a_at
1450666_s_at
1460212_at
1438452_at
1460511_at
1418653_at
1429994_s_at
1416416_x_at
1449375_at
1418099_at
1424616_s_at
1424614_at
1428869_at
1424599_at
1420679_a_at
1425792_a_at
1417926_at
1436174_at
1429556_at
1447705_at
1419397_at
1436808_x_at
1434334_at
Affymetrix
Probe set ID
Intracellular trafficking
Lipid metabolism
Cellular signaling
Energy metabolism
Energy metabolism
Intracellular trafficking
Cellular signaling
Cellular signaling
Cellular signaling
Acute phase resopnse
Cellular signaling
Cellular signaling
Tissue development
Cytoskeleton
Drug metabolism
Drug metabolism
Drug metabolism
Lipid metabolism/drug metabolism
Apoptosis
Lipid metabolism
Lipid metabolism
Nucleogenesis/associated with cancer
Proliferation/association with cancer
Proliferation/association with cancer
Transcription factor/association with cancer
Proliferation
Proliferation/association with cancer
Proliferation/apoptosis/association with cancer
DNA replication
DNA replication
Intracellular trafficking/proliferation/
association with cancer
Proliferation
Biological Function
;
:
;
:
;
;
:
;
;
;
:
;
:
;
:
:
:
:
;
:
:
;
;
;
:
;
;
;
:
;
;
;
Deregulation
|
Ncapg2
Atad2
Tead1
Polymerase (DNA directed), alpha 1, catalytic subunit
Minichromosome maintenance complex component 5
Protein kinase D2
Pola1
Mcm5
Prkd2
Cell cycle progression
Gene Name
Gene Symbol
Toxicological Category
TABLE 5. Toxicological Categories and Biological Functions of NGC Biomarker Candidates
284
TOXICOLOGICAL SCIENCES, 2015, Vol. 143, No. 2
Diacylglycerol kinase, alpha 80kDa
Phospholipase A2, group XVI
Pancreatic lipase-related protein 1
Chromosome 1 open reading frame 31
Zinc finger protein 472
Nuclear receptor subfamily 2, group C, member 1
Zinc finger protein 394
Zinc finger protein 750
General transcription factor IIB
Grainyhead-like 1 (Drosophila)
Sterile alpha motif domain containing 4A
Chromosome 6 open reading frame 211
Chromosome 17 open reading frame 101
Family with sequence similarity 171, member A1
Family with sequence similarity 214, member A
PR domain containing 15
Chromosome 11 open reading frame 54
RIKEN cDNA 5430416B10 gene
RIKEN cDNA A930036K24 gene
RIKEN cDNA 0610008F07 gene
RIKEN cDNA 2810433D01 gene
RIKEN cDNA 4930597L12 gene
Chromosome 6 open reading frame 72
Chromosome 6 open reading frame 72
Chromosome 9 open reading frame 91
Predicted gene 10419
Predicted gene 2011
Transmembrane protein 181C, pseudogene
Dgka
Pla2g16
Pnliprp1
C1orf31
Zfp472
Nr2c1
Znf394
Znf750
Gtf2b
Grhl1
Samd4a
C6orf211a
C17orf101
Fam171a1
Fam214a
Prdm15
C11orf54
5430416B10Rika
A930036K24Rika
0610008F07Rika
2810433D01Rika,b
4930597L12Rika
C6orf72
C6orf72
C9orf91a
Gm10419a
Gm2011a
Tmem181c-psa,b
a
a
a
Gene Name
Gene Symbol
1421495_a_at
1418331_at
1438402_at
1436033_at
1455459_at
1454067_a_at
1454558_at
1431174_at
1425246_at
1445896_at
1432438_at
1424026_s_at
1424025_at
1426315_a_at
1436265_at
1452477_at
1435948_at
1444343_at
1441230_at
1446621_at
1438068_at
1436198_at
1440337_at
1456808_at
1457470_at
1442427_at
1418578_at
1445597_s_at
1415777_at
1429061_at
1425058_at
1418605_at
1418607_at
1437469_at
1451135_at
1424030_at
1436356_at
Affymetrix
Probe set ID
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Unknown
Lipid metabolism
Lipid metabolism
Lipid metabolism
Mitochondrial respiration
Transcription factor
Transcription factor
Transcription factor
Transcription factor
Transcription factor
Transcription factor
Translation
Biological Function
;
:
:
:
;
:
;
;
:
:
:
:
:
;
:
:
;
;
;
;
:
;
;
;
;
:
;
;
:
;
;
;
:
;
;
;
;
Deregulation
Noncoding RNA elements within the group “Unknown” by predicted noncoding element.
Noncoding RNA elements within the group “Unknown” by full-length noncoding element.
a
b
“Detoxification response” (4 probe sets), “Cellular assembly and organization” (2 probe sets), “Immune response” (1 probe set), “Others” (22 probe sets), and 26 probe sets assigned to “Unknown”) and the biological functions of the
probe sets were specified.
NGC biomarker candidates (including Affymetrix Probe set ID, gene symbol, gene name and the direction of deregulation : or ;) were toxicological categorized (“Cell cycle progression” (10 probe sets), “Apoptosis” (4 probe sets),
Unknown
Toxicological Category
TABLE 5. (continued)
KOSSLER ET AL.
|
285
286
|
TOXICOLOGICAL SCIENCES, 2015, Vol. 143, No. 2
FIG. 1. Heatmap of GC biomarker candidates. In total, 64 probe sets were identified as potential biomarker candidates for GCs. Deregulation of each GC probe set in individual animals treated with GC (CIDB, DMN, or MDA), NGC (DCB, PBO, or PB), or NHC (Nif, CFX, Praz, or Prop) compounds is highlighted in blue (FC 1.5) and red
(FC 1.5). Color version of this figure is available at http://toxsci.oxfordjournals.org.
accordance with the observed histopathology findings (eg,
mononuclear cell infiltration, please refer to Table 2). It was previously reported that the macrophage surface protein Fgl2
showed increased expression in a number of human malignant
tumors (Liu et al., 2012). An upregulation of Lck was also
observed in human colon and small cell lung carcinomas
(McCracken et al., 1997), whereas human CD34 positive hematopoietic progenitor cells were described as sensitive targets for
GC-inducible DNA damage response (Abernethy et al., 2004). We
detected 1 gene involved in “Detoxification response” (Ces2e)
that was upregulated in the GC-treated animals. Studies of
increased Ces2a expression in a human lung cancer cell line and
its candidacy as a p53 target gene has been studied, pointing out
p53-mediated activation in response to DNA damage (Kannan
et al., 2001). In addition to the listed toxicological categories
above, a category named “Others” was added that includes
Acot9, Akap13, Atp6v1d (3 probe sets), Ccdc80, COX6B2, Exoc4,
G6pd, Ggta1, Ltn1, Moxd1, Snx6, Srprb, Zdhhc14 (2 probe sets), and
Zfp54. Within this group, Acot9, G6pd, and Modx1 have different
important functions in cellular metabolisms. Interestingly,
deregulation of G6pdx was in accordance with published data
supporting G6pdx as a biomarker candidate in response to GC
administration (Lioi et al., 1998). Additionally, Snx6, Akap13,
Atp6v1d, and Srprb were assigned as being involved in cellular
intracellular trafficking mechanisms. Finally, the listed genes
Atp6v1d (3 probe sets), Bax, Bcl2a1, Ccdc80, CD34, Emp3, Iqgap1
(2 probe sets), Mbl2, Nisch, Phlda3, Plekha2, Srprb, Tspan13,
Zdhhc14 (2 probe sets), and Zeb2 were previously reported to be
involved in carcinogenesis.
In this mouse study, we also identified GC treatment related
biomarker candidates after 3 days of drug administration
(necropsy at day 4) as well as mRNA expression candidate
KOSSLER ET AL.
biomarker profiles in female mice at day 4 and day 15
(Supplementary Table 5). We identified 11 probe sets (representing 9 genes) as comparable deregulated GC biomarker candidates in females and males at day 15 and 2 probe sets at day 4.
One probe set was found to be deregulated in male mice at day
4 and day 15 and 1 other probe set was deregulated in female
mice at day 4 and day 15 (Table 6 and Supplementary Table 5).
A detailed mechanistic description of the identified marker
candidates for male day 4, female day 15 and female day 4 mice
would go beyond the scope of this article. For this regard, the
mechanistic analysis was focused on the identified biomarker
candidates from male mice at day 15.
Functional Analysis of NGC Biomarker Candidates
As previously mentioned, mechanisms behind NGCs are not yet
completely understood. To gain further insights into NGC
modes of action, we analyzed gene expression profiles of animals treated with known NGCs such as 1,4-DCB, PB or PBO.
Meeting the biomarker criteria listed in Table 3, we identified 69
probe sets as marker candidates for NGC (Fig. 2 and Table 5).
Twenty-eight probe sets were upregulated (FC 1.5) when compared with the median of the corresponding vehicle control
group, highlighted in red. Forty-one probe sets were downregulated (FC 1.5) when compared with the median of the corresponding vehicle control group, highlighted in blue (Fig. 2). The
identified NGC biomarker candidates were analyzed in a PCA
(Supplementary Table 2B). Expression data of NGCs such as PBO
and PB were clearly separated from GC and NHC data. GC and
NHC expression data formed an overlapping cluster. Data for
DCB were located between the NGC and the GC/NHC clusters.
In total, 9 out of 69 probe sets were not assigned to a gene
symbol and 26 probe set derived genes are presently not functionally characterized in the literature. Based on literature and
IPA, the remaining 43 probe sets (encoding 42 genes) were
grouped into toxicologically relevant groups (Table 5). Several
probe sets grouped as “Unknown” were identified as noncoding
RNA elements. We differentiated between full-length noncoding elements and predicted noncoding RNAs (Table 5).
Interestingly, the biggest group of NGC biomarker candidates
comprises genes that are directly or indirectly involved in regulation of “Cell cycle progression” (Aig1, Atad2, Fgl1, Mcm5, Nsl1,
Ncapg2, Pola1, Prkd2, Rorc, Tead1). Among these genes, 2 encoded
proteins are characterized to be involved in “DNA replication”
(Mcm5, Pola1), 2 play important roles in chromosomal segregation processes during mitosis (Ncapg2, Nsl1), and 6 genes are
described as having an impact on cellular proliferation (Aig1,
Atad2, Fgl1, Prkd2, RORC, Tead1).
Both genes that were assigned to “DNA replication” were
downregulated in response to NGC treatment. Interestingly, it
has been shown that a downregulation of Pola1 by siRNA
increases the delay in DNA replication initiation that may lead
to DNA damage due to secondary mechanisms (Conti et al.,
2010). Furthermore, Nsl1 encodes a protein which is part of a
protein complex that bridges centromeric heterochromatin
with the outer kinetochore structure during mitosis. In this
study, it was upregulated in NGC-dosed mice.
We detected genes that have different functions during proliferation of cells, some of them directly described to be linked
with cancer progression in different organs, such Aig1, Atad2,
Fgl1, Tead1, Prkd2, and Rorc. Among them, transcription factor
Rorc and Fgl1 have been suggested to be involved in the development of hepatocellular carcinomas (eg, Ou et al., 2013).
As shown in Table 5, a set of 3 genes including Tnfrsf1b,
Pgap2 (2 probe sets), and Nolc1 have different functions during
|
287
“Apoptosis”. Interestingly, it has been demonstrated that Nolc1
together with transcription factor GTF2B (category “Others” in
Table 5) activates the acute phase response gene CEBPB grouped
in the category “Immune response” in Table 5 (Miau et al., 1997);
all of them were downregulated in NGC-treated animals.
A group of 4 genes with roles of the respective proteins in
“Detoxification response” (Ces2a, Cyp2c50 (homolog to human
Cyp2C18), Cyp2c65 (homolog to human Cyp2C19) and Gstm1)
were strongly upregulated in mice treated with NGCs. In a variety of experiments, an induction of CYP2C enzymes in response
to known NGCs PB or PBO has been clearly pointed out
(Okamiya et al., 1998; Phillips et al., 1997). Another upregulated
metabolic enzyme Gstm1 has also been previously described as
a candidate predictive biomarker gene in lung tissue of mice
treated for 13 weeks with compounds that showed increased
lung tumor incidence in a 2-year rodent study (Thomas et al.,
2007b).
We identified deregulation of Nebl and Pkp2, which were
assigned to “Cellular assembly and organization.” Interestingly,
the expression of desmosome configuration factor Pkp2 has
been proposed as an early prognostic marker for gastric carcinoma progression in patients (Demirag et al., 2011).
The remaining 22 NGC specific protein encoding genes were
summarized within the category “Others” and comprised
diverse cellular functions like “Cellular signaling” (6 probe sets),
“Transcription factor” (6 probe sets), “Lipid metabolism” (4
probe sets), “Intracellular trafficking” (2 probe sets), “Energy
metabolism” (2 probe sets), “Mitochondrial respiration” (1 probe
set), and “Translation” (1 probe set). Several probe sets including
Akr1d1, Dgka, Pla2g16, and Pnliprp1 have different functions
within “Lipid metabolism.” Interestingly, it has already been
demonstrated in rodents and humans that PB induced liver
damage with effects on serum total cholesterol levels (Kiyosawa
et al., 2004).
A comparison of NGC treatment-related biomarker candidates after 3 days of drug administration (necropsy at day 4) as
well as mRNA expression candidate biomarker profiles in
female mice at day 4 and day 15 (Supplementary Table 5)
revealed some overlaps (Table 6). We identified 7 probe sets
(representing 6 genes) as comparable deregulated NGC biomarker candidates in females and males at days 15 and 17
probe sets representing 14 genes at day 4. In all, 12 probe sets
were found to be deregulated in male mice at day 4 and day 15
and 15 probe sets representing 12 genes were deregulated in
female mice at day 4 and day 15 (Table 6 and Supplementary
Table 5).
Definition of GC and NGC Responder Animals based on the Identified
Gene Signatures
The above described biomarker candidate lists for GC and NGC
were identified by using an individual animal approach.
Detection of the identified probe sets in all animals included in
this study helps to define which animals are so called “GC or
NGC responder animals.” Therefore, we counted the number of
our derived GC biomarker candidates per treated animal for
each compound class. In this approach, 6 mice per compound
were treated with CIDB, DMN, MDA, PBO, CFX, Praz, Prop, CPA,
TAA, or Wy-14643 and 5 animals per compound were treated
with DCB, PB, or Nif. The number of GC biomarker candidate
probe sets with a fold change deregulation 1.5 FC or FC 1.5
per mouse was counted and listed as percent (Supplementary
Table 3A). In none of the GC-treated mice were all 64 biomarker
candidates deregulated. The highest GC marker detection rate
of 54 probe sets (corresponding to 84.4%) was achieved in 1
288
|
TOXICOLOGICAL SCIENCES, 2015, Vol. 143, No. 2
FIG 2. Heatmap of NGC biomarker candidates. The heatmap shows the deregulation of the 69 derived NGC biomarker candidates for each animal. The deregulation of
each probe set in each animal treated with GCs (CIDB, DMN, or MDA), NGCs (DCB, PBO, or PB), or NHCs (Nif, CFX, Praz, or Prop) is highlighted in blue (FC 1.5) and red
boxes (FC 1.5), compared with the median of the corresponding vehicle control group. Color version of this figure is available at http://toxsci.oxfordjournals.org.
TABLE 6. Overview about Common Candidate Biomarker Genes for GC or NGC between Different Time Points and/or Genders
Intersection Groups
Gene Symbol
Overlapping GC biomarker candidates day 15 females and males
Acot9, Atp6v1d (3 probe sets), Bax, C9orf116, Ccng1, Cox6b2, Loxl2,
Pqlc3, Tuba1a
C11orf54, Cyp2c50, Fndc5, Nebl, Pgap2 (2 probe sets) and Pnliprp1
Ccnd1 and Ccng1
Ang (2 probe sets), C11orf54 (2 probe sets), Ces2a, Chaf1b, Chid1,
Cyp2c50 (2 probe sets), Emc1, Gstt3, Fndc5, Foxa3, Pglyrp2, Pnliprp1,
Samd4a and Syncrip
Ccng1
9130409123Rik, Adrbk2, C11orf54 (2 probe sets), Cyp2c50 (2 probe sets),
Fndc5 (2 probe sets), Foxa3, Pnliprp1, Samd4a, Slc17a1, Slc26a6,
Sult2a2 and Wdr18
Exoc4
C11orf54, Ces2a, Cyp2c50, Cyp2c65, Fgl1, Fndc5, Gm10419, Nebl, Nsl1,
Pgap2, Pnliprp1 and Rorc
Overlapping NGC biomarker candidates day 15 females and males
Overlapping GC biomarker candidates day 4 females and males
Overlapping NGC biomarker candidates day 4 females and males
Overlapping GC biomarker candidates day 4 and day 15 females
Overlapping NGC biomarker candidates day 4 and day 15 females
Overlapping GC biomarker candidates day 4 and day 15 males
Overlapping NGC biomarker candidates day 4 and day 15 males
Overview about different gene intersection groups between the 4 NGC and 4 GC candidate biomarker lists: female and male, day 4 and day 15, respectively. Genes
observed as deregulated in all 4 NGC intersection groups are labeled with bold characters.
KOSSLER ET AL.
CIDB-treated mouse. In all other GC-dosed animals, the number
of GC biomarker candidates varied between 14 and 52 probe
sets (corresponding to 21.9% and 81.3%). In only 2 MDA-treated
animals was the number of GC biomarker candidates reduced,
to 14 and 15 GC-specific probe sets (corresponding to 21.9% and
23.4%). In NGC-treated mice, the number of deregulated GC-specific probe sets reached maximally 4 GC biomarker candidates
(corresponding to 6.3%), with the exception of 1 PBO treated animal showing a deregulation of 6 GC-specific probe sets (corresponding to 9.4%). In 2 out of 23 NHC-treated mice, 5 and 6 GC
biomarker candidates (corresponding to 7.8% and 9.4%) were
deregulated. In all other NHC-dosed animals, the number of GCspecific probe sets varied between 0 and 4 sets (corresponding
up to 6.3%). Based on these results, we suggest defining an animal as a GC responder if 20% of GC biomarker candidates were
deregulated in its liver. Thus, all GC-treated animals were
defined as responders.
In order to define the number of NGC biomarker candidates
that have to be changed in a NGC responder animal, the NGC
biomarker candidates were counted for individual animals
treated with NGCs, GCs, and NHCs (Supplementary Table 3B). In
none of the 16 NGC-treated animals were all NGC-specific probe
sets deregulated. However, in mice administered PBO or PB,
42–57 probe sets (corresponding to 60.9% and 82.6%) were
deregulated. In contrast, the number of deregulated NGC biomarker candidates in DCB-treated mice varied between 12 and
24 NGC-specific probe sets (corresponding to 17.4% and 34.8%).
The highest NGC biomarker candidate detection rate of 6 probe
sets (corresponding to 8.7%) was determined in 1 animal administered with MDA. All other GC-treated mice exhibited at most a
deregulation of 4 NGC-specific probe sets (corresponding to
5.8%). In all mice administered with NHC, a deregulation of
maximally 3 NGC-specific probe sets (corresponding to 4.3%)
was observed, with the exception of 1 Praz-dosed animal, which
showed a gene expression change of 5 NGC probe sets (corresponding to 7.2%). Based on these data, we suggest defining an
animal as a NGC responder if 20% of NGC biomarker candidates
were deregulated in its liver. Thus, only 1 animal out of 16
NGC-treated mice was defined as a nonresponder animal,
showing a gene expression change of 12 NGC biomarker candidates (corresponding to 17.4%).
Investigation of CPA, TAA and WY-14643 Expression Profiles Using
the Newly Identified GC and NGC Biomarker Candidates
In the present study, we termed CPA, TAA, and WY-14643 as
UCs, owing to ambiguous results in genotoxicity testing listed in
Supplementary Table 4. All of them are Ames test negative but
showed positive or unclear results in 1 or more tests for genotoxicity in vitro and in vivo. Our approach resulted in 2 probe set
biomarker candidate lists specifically discriminating GC from
NGC substances. Based on these mRNA signatures, a categorization of CPA, TAA, and WY-14643 was performed (Figs. 3 and 4).
In CPA-treated mice, a range between 22 and 26 (corresponding to 31.9% and 37.7%) NGC biomarker candidates were deregulated (Supplementary Table 3B). In contrast, only 0–6 GC
biomarker candidates probe sets (corresponding to 9.4%) were
deregulated in CPA-dosed animals (Supplementary Table 3A).
Together with the PCA plot data illustrating a clustering in the
vicinity of NGC substances CPA might be categorized as NGC
(Supplementary Table 2).
In contrast, TAA-treated mice displayed similar gene expression changes as GCs (Fig. 3). Only 1 TAA-dosed animal
responded with a deregulation of 15 NGC-specific probe sets
(corresponding to 21.7%), whereas in all other mice a range
|
289
between 29 and 39 (corresponding to 45.3% and 60.9%) GC biomarker candidates were deregulated (Supplementary Table 3).
Expression data of TAA-dosed animals in the context of data
derived from GC-treated animals could be visualized in a PCA
plot (Supplementary Table 2A). Based on our data, we suggest
categorizing TAA as a GC compound.
A comparison of our derived biomarker candidate lists with
gene expression profiles of Wy-14643 treated animals showed
overlaps with both GC and NGC biomarker candidate lists
(Figs. 3 and 4). In all Wy-14643 dosed animals, 17–37 (corresponding to 26.6% to 57.8%) GC-specific probe sets
(Supplementary Table 3A) and 21–29 (corresponding to 30.4% to
42%) NGC-specific probe sets were differentially expressed
(Supplementary Table 3B). Thus, it seems that Wy-14643 has
both GC and NGC compound characteristics at a dose of 200 mg/
kg/day.
DISCUSSION
Identification of Specific Biomarker Candidates for GCs and NGCs
Using an Approach Considering Individual Animal Data
The aim of this study was to identify specific biomarker candidates differentiating between GCs and NGCs. A similar number
of significantly deregulated probe sets were calculated in GCand NGC-treated mice (Supplementary Table 1). In addition to
GCs and NGCs, UCs and NHC compound groups were tested in
this study. UCs might have been overdosed, as histopathology
data revealed strong treatment-related effects and all compounds showed a higher number of statistically significant
deregulated probe sets compared with data from animals
treated with GC or NGC agents (Table 5 and Supplementary
Table 1). In contrast, histopathology data as well as the calculation of statistically significant probe sets of NHC-treated animals showed no treatment-related changes in the liver. In order
to have similar numbers of significantly deregulated gene sets
between GCs, NGCs, UCs, and NHCs, higher doses of the NHC
compounds tested, as well as lower doses for the UCs, should be
used in future studies.
Compound dose selection, based on dose range finding studies or literature reports, are not always optimal for the prediction of the right level of gene expression deregulation. The issue
of an appropriate dose selection exists for all supervised learning models that are not able to distinguish independently
between unequipotent dose effects. Another limitation of toxicogenomics investigations by machine learning methods is an
insufficient compound classification. In the past, compound
classification was mainly done based on genotoxicity analysis
with the focal point on Ames test results (Kirkland et al., 2006;
Waters et al., 2010). However, it is highly questionable to use
compounds with ambiguous genotoxicity test results as clearly
categorized in an evaluation of mRNA expression profiles with
machine learning methods. This might negatively influence the
selection of specific marker candidates for GCs and NGCs.
In order to meet these concerns, we used a study approach
that is more flexible with respect to unequipotent compound
doses and animal dose response. The keystone of this approach
is the focus on individual animal fold changes instead of animal
group mean values. Description of individual animal findings is
also the conventional procedure in histopathology investigations in the course of drug exposure studies (Crissman et al.,
2004). The use of single animal expression values in microarray
measurement evaluations of toxic effects was examined in a
limited number of animal studies, which all demonstrated a
very good correlation between gene expression profiling and
290
|
TOXICOLOGICAL SCIENCES, 2015, Vol. 143, No. 2
FIG 3. Investigation of CPA, TAA, and Wy-14643 liver gene expression profiles based on derived GC biomarker candidates. The identified 64 GC biomarker candidate
probe sets were visualized for all GC, NGC, and UC animals via heat map. They were grouped according to their classification as follows: GCs (CIDB, DMN, or MDA),
NGCs (DCB, PBO, or PB), or UCs (CPA, TAA, or Wy-14643). Each colored box represents 1 single animal. Colored red boxes display an upregulation of gene expression
(FC 1.5) and blue colored boxes display a downregulation of gene expression (FC 1.5) when compared with the median of the corresponding vehicle control group.
TAA and Wy-14643 showed a similar gene expression profiling to GCs. Color version of this figure is available at http://toxsci.oxfordjournals.org.
histopathology analysis (Ellinger-Ziegelbauer et al., 2011).
Furthermore, single animals responded differently to drug
administration, which can be monitored on mRNA expression
level within 1 treatment group. In addition, only clearly classified compounds with unambiguous genotoxicity test results
were included in this approach, which underlines the specificity
of the biomarker candidate probe sets identified in this study.
Looking at the intersection of GC and NGC biomarker candidate lists between male and female mice at the same time
points using the same compounds and statistical approach is
relatively small (Table 6). We get a similar picture when looking
at the comparisons of GC and NGC biomarker candidates
between time points day 4 and day 15. Interestingly, the number
of overlapping genes was much higher between NGC lists when
compared with GC lists and 4 of the identified NGC biomarker
candidates appear in both genders and both time points:
Cyp2c50, C11orf54, Fndc5, and Pnliprp1 (bold in Table 6). In conclusion, most of the identified biomarker candidate genes show
a high dependency with regard to time point and gender.
Promising New NGC Biomarker Candidates Reveal Functions in Cell
Cycle Progression, Apoptosis, and Detoxification Response—Is There
a Common Denominator for the Various NGC Mechanisms?
Whereas GCs act via direct DNA reactive pathways, NGCs show
diverse mechanisms. Principal categories are NGC-cytotoxic
and NGC-mitogenic modes of action. The characteristic mechanism of NGC-cytotoxic compounds is induction of chronic cell
injury leading to necrosis and subsequent regenerative
KOSSLER ET AL.
|
291
FIG 4. Investigation of CPA, TAA, and Wy-14643 liver expression profiles based on derived NGC biomarker candidates. The identified 69 NGC biomarker candidate probe
sets were visualized for all GC, NGC, and UC animals via heat map. They were classified as follows: GCs (CIDB, DMN, or MDA), NGCs (DCB, PBO, or PB), or UCs (CPA,
TAA, or Wy-14643). Gene expression deregulations are highlighted in blue (FC 1.5) and red boxes (FC 1.5), compared with the median of the corresponding vehicle
control group. CPA and WY-14643 showed overlapping gene expression patterns with NGCs. Color version of this figure is available at http://toxsci.oxfordjournals.org.
hyperplasia, serving as a preneoplastic lesion for tumor development after numerous initiation and promotional activities
(Butterworth et al., 2007; Dietrich and Swenberg, 1991). In
contrast, NGC-mitogenic/promotional compounds reversibly
stimulate tissue/organ growth as well as growth of pre-cancerous lesions through different mechanisms, including immunosuppression, increased secretion of trophic hormones (eg,
estrogens), receptor activation (eg, peroxisome proliferation),
and others (eg, CYP450 induction) (Butterworth et al., 2007; Lima
and Van der Laan, 2000; Schulte-Hermann et al., 1983).
In our approach, we included PB, PBO, and 1,4-DCB as known
NGCs with different mechanisms of action. There is evidence
that 1,4-DCB and PB act through a NGC-mitogenic/promotional
mode of action, as they show no cytotoxicity (Butterworth, 1990;
Butterworth et al., 2007). Induction of cancer by 1,4-DCB in the
rodent bioassay could be mediated by forcing the growth of
spontaneous precancerous lesions upon long-term treatment
(Butterworth et al., 2007; Elridge et al., 1992). Concerning PB toxicity, it has been demonstrated that chronic treatment leads to
increased oxidative stress due to constant induction of metabolism and biotransformation genes (CYPs) (Elrick et al., 2005). CYP
induction also results in promotion of hepatocellular and
smooth endoplasmic reticulum proliferation (Waxman, 1999).
Contrary to PB and 1,4-DCB, PBO belongs to a category of NGCcytotoxic agents leading to DNA-damaging side effects and promotional activity at the site of injury, such as oxidative stress
and regenerative cell proliferation (Tasaki et al., 2012).
Interestingly, the tumor promoting mechanism of PB and PBO
seem to have some similarities as both compounds induce
CYPs leading to generation of ROS and inhibition of gap
292
|
TOXICOLOGICAL SCIENCES, 2015, Vol. 143, No. 2
junctional intercellular communication (Muguruma et al., 2006;
Okamiya et al., 1998).
Tumor development is a complex multistep process. On the
one hand, it could be argued that a transcriptomics analysis of a
14 day short-term study cannot hope to predict the outcome of
a 2-year carcinogenicity bioassay in mice. On the other hand, all
NGC mechanisms mentioned above have in common that they
lead to an imbalance between cell replication and cell death by
apoptosis, most likely via molecular mechanisms like mRNA or
epigenetic deregulation. This is clearly reflected in the specific
gene expression profiles identified for NGC, where we show the
deregulation of gene groups involved in cell cycle deregulation,
apoptosis as well as cellular assembly and organization.
Alterations of these pathways have the potential to result in
changes in cell proliferation and differentiation, influences on
signal transduction pathways and inhibition of intercellular
communication, as well as neoplastic transformation. The identified detoxification response marked by CYP enzyme induction
supports the previously suggested mechanism for NGC of the
tested compounds PB and PBO, comprising generation of ROS
via CYP induction.
Taken together, we conclude that despite the fact that there
are some mechanistic differences behind the selected NGCs
used for this study, identification in a short-term 2 week in vivo
mouse assay of a common denominator serving as a predictive
and specific biomarker signature for NGCs for all of these mechanisms is a realistic aim.
NGC expression analysis within this study revealed new as
well as known candidate biomarkers reflecting possible common features of NGCs. Despite these informative results, the
number of NGCs in this study was not sufficient to cover the
multifaceted modes of action. Application of the identified candidate gene lists to further mouse short-term studies with other
known NGC mechanisms, like immunosuppression or
increased secretion of trophic hormones (eg, estrogens), as well
as analysis of comparable data in female mice, would be one of
the next steps.
Specific GC and NGC Biomarker Candidates—New Tool for
Evaluation of Ambiguous Genotoxicity Test Results
The conventional genotoxicity test battery includes various
short-term assays screening for gene mutation, clastogenicity,
and aneuploidy and is demonstrably sensitive in the identification of presumptive carcinogens. Nevertheless, its specificity
has been questioned and conclusions concerning mutagenicity
can often not be drawn based on ambiguous genotoxicity test
results (Waters et al., 2010). Therefore, additional tools for more
detailed characterization of compounds with unclear test
results are urgently required. The GC biomarker candidates
identified in this study could serve as such a tool. They clearly
reflect the expected DNA damage response, including genes
with biological functions in apoptosis, cell cycle progression,
DNA repair, and oxidative stress response.
Despite intensive literature review, it was not possible to
group CPA, TAA, and Wy-14643 as clear GC or NGC
(Supplementary Table 4). Applying our newly derived biomarker
candidate lists, we propose to classify TAA as a GC and CPA as a
NGC. Furthermore, our results indicate that Wy-14643 has characteristics of both carcinogenic classes (Figs. 3 and 4). This could
be the result of a very strong biological response following highdose application of Wy-14643 for 14 days. Induction of oxidative
DNA damage and subsequent DNA repair via a secondary
mechanism might be the consequence, marked by a high
number of deregulated probe sets. A proper follow-up doseresponse study could address this issue.
In conclusion, a mouse liver genomic dataset with mechanistic information from 13 known compounds including GCs,
NGCs, NHCS, and UCs was published in the course of the IMI
MARCAR project. It was possible to get an improved mechanistic understanding of genotoxic and nongenotoxic carcinogenesis and to identify specific biomarker candidates for GCs and
NGCs. Mechanistic investigation of test compounds, potential
genotoxic metabolites, and impurities using specific gene
expression profiles for NGC and GC in a short-term toxicogenomics design might lead to a better assessment of carcinogenic risk for humans early in toxicological testing.
In the future, approaches using the biomarker candidate
lists for GC and NGCs need to be validated. Follow-up should
focus on comparison with the present NGC- and GC-specific
mouse candidate biomarker mRNA expression data, using the
same approach with more compounds as well as with the other
rodent species, the rat.
SUPPLEMENTARY DATA
Supplementary data are available online at http://toxsci.
oxfordjournals.org/.
FUNDING
The IMI MARCAR project was funded by Innovative Medicine
Initiative Joint Undertaking (IMI JU) under grant agreement
number [115001].
ACKNOWLEDGMENTS
We want to thank all IMI MARCAR consortium members
who had a role in animal study design, data collection
and analysis, or preparation of the article. We are also
thankful to Natalie Motsch for sharing her genotoxicity
expertise with us. The IMI MARCAR project was evaluated
by independent experts on request of the European
Commission. There are no conflicts of interest of authors
or consortium members since the project was partially
funded by public (EU) funds. The MARCAR project is dedicated to basic research and there are no commercial
interests.
REFERENCES
Abernethy, D. J., Kleymenova, E. V., Rose, J., Recio, L., and Faiola,
B. (2004). Human CD34þ hematopoietic progenitor cells are
sensitive targets for toxicity induced by 1,4-benzoquinone.
Toxicol. Sci. 79, 82–89.
Alahari, S. K., Lee, J. W., and Juliano, R. L. (2000). Nischarin, a
novel protein that interacts with the integrin alpha5 subunit
and inhibits cell migration. J. Cell Biol. 151, 1141–1154.
Alaminos, M., Davalos, V., Ropero, S., Setien, F., Paz, M. F.,
Herranz, M., Fraga, M. F., Mora, J., Cheung, N. K., Gerald, W. L.,
and Esteller, M. (2005). EMP3, a myelin-related gene located
in the critical 19q13.3 region, is epigenetically silenced and
exhibits features of a candidate tumor suppressor in glioma
and neuroblastoma. Cancer Res. 65, 2565–2571.
Amacher, D. E. and Turner, G. N. (1982). Mutagenic evaluation of
carcinogens and non-carcinogens in the L5178Y/TK assay
KOSSLER ET AL.
utilizing postmitochondrial fractions (S9) from normal rat
liver. Mutat. Res. 97, 49–65.
Barkinge, J. L., Gudi, R., Sarah, H., Chu, F., Borthakur, A.,
Prabhakar, B. S., and Prasad, K. V. (2009). The p53-induced
Siva-1 plays a significant role in cisplatin-mediated apoptosis. J. Carcinog. 8, 2.
Benigni, R., Bossa, C., and Tcheremenskaia, O. (2013).
Nongenotoxic carcinogenicity of chemicals: Mechanisms of
action and early recognition through a new set of structural
alerts. Chem. Rev. 113, 2940–2957.
Butterworth, B. E. (1990). Consideration of both genotoxic and
nongenotoxic mechanisms in predicting carcinogenic potential. Mutat. Res. 239, 117–132.
Butterworth, B. E., Aylward, L. L., and Hays, S. M. (2007). A mechanism-based cancer risk assessment for 1,4-dichlorobenzene. Regul. Toxicol. Pharmacol. 49, 138–148.
Cano, A., Santamaria, P. G., and Moreno-Bueno, G. (2012). LOXL2
in epithelial cell plasticity and tumor progression. Future
Oncol. 8, 1095–1108.
Chieli, E., Aliboni, F., Saviozzi, M., and Malvaldi, G. (1987).
Induction of micronucleated erythrocytes by primary thioamides and their metabolites in the mouse. Mutat. Res. 192,
141–143.
Chilakapati, J., Korrapati, M. C., Hill, R. A., Warbritton, A.,
Latendresse, J. R., and Mehendale, H. M. (2007).
Toxicokinetics and toxicity of thioacetamide sulfoxide: A
metabolite of thioacetamide. Toxicology 230, 105–116.
Conti, C., Leo, E., Eichler, G. S., Sordet, O., Martin, M. M., Fan, A.,
Aladjem, M. I., and Pommier, Y. (2010). Inhibition of histone
deacetylase in cancer cells slows down replication forks, activates dormant origins, and induces DNA damage. Cancer
Res. 70, 4470–4480.
Craddock, V. M. and Henderson, A. R. (1978). De novo and repair
replication of DNA in liver of carcinogen-treated animals.
Cancer Res. 38, 2135–2143.
Crissman, J. W., Goodman, D. G., Hildebrandt, P. K., Maronpot, R.
R., Prater, D. A., Riley, J. H., Seaman, W. J., and Thake, D. C.
(2004). Best practices guideline: Toxicologic histopathology.
Toxicol. Pathol. 32, 126–131.
Demirag, G. G., Sullu, Y., Gurgenyatagi, D., Okumus, N. O., and
Yucel, I. (2011). Expression of plakophilins (PKP1, PKP2, and
PKP3) in gastric cancers. Diagn. Pathol. 6, 1.
Deutsch, W. A., Kukreja, A., Shane, B., and Hegde, V. (2001).
Phenobarbital, oxazepam and Wyeth 14,643 cause DNA damage as measured by the Comet assay. Mutagenesis 16,
439–442.
Dietrich, D. R., and Swenberg, J. A. (1991). Preneoplastic lesions
in rodent kidney induced spontaneously or by non-genotoxic
agents: predictive nature and comparison to lesions induced
by genotoxic carcinogens. Mutat. Res. 248, 239–260.
Doktorova, T. Y., Ellinger-Ziegelbauer, H., Vinken, M.,
Vanhaecke, T., van, D. J., Kleinjans, J., Ahr, H. J., and Rogiers,
V. (2012). Comparison of hepatocarcinogen-induced gene expression profiles in conventional primary rat hepatocytes
with in vivo rat liver. Arch. Toxicol. 86, 1399–1411.
Eichner, J., Kossler, N., Wrzodek, C., Kalkuhl, A., Bach, T. D.,
Ostenfeldt, N., Richard, V., and Zell, A. (2013). A
toxicogenomic approach for the prediction of murine hepatocarcinogenesis using ensemble feature selection. PLoS One
8, e73938.
Eldridge, S. R., Goldsworthy, T. L., Popp, J. A., and Butterworth, B.
E. (1992). Mitogenic stimulation of hepatocellular proliferation in rodents following 1,4-dichlorobenzene administration. Carcinogenesis 13, 409–415.
|
293
Ellinger-Ziegelbauer, H., Adler, M., Amberg, A., Brandenburg, A.,
Callanan, J. J., Connor, S., Fountoulakis, M., Gmuender, H.,
Gruhler, A., Hewitt, P., et al. (2011). The enhanced value of
combining conventional and “omics” analyses in early assessment of drug-induced hepatobiliary injury. Toxicol. Appl.
Pharmacol. 252, 97–111.
Ellinger-Ziegelbauer, H., Stuart, B., Wahle, B., Bomann, W., and
Ahr, H. J. (2004). Characteristic expression profiles induced by
genotoxic carcinogens in rat liver. Toxicol. Sci. 77, 19–34.
Elrick, M. M., Kramer, J. A., Alden, C. L., Blomme, E. A., Bunch, R.
T., Cabonce, M. A., Curtiss, S. W., Kier, L. D., Kolaja, K. L., Rodi,
C. P., et al. (2005). Differential display in rat livers treated for
13 weeks with phenobarbital implicates a role for metabolic
and oxidative stress in nongenotoxic carcinogenicity. Toxicol.
Pathol. 33, 118–126.
Feser, W., Kerdar, R. S., Blode, H., and Reimann, R. (1996).
Formation of DNA-adducts by selected sex steroids in rat
liver. Hum. Exp. Toxicol. 15, 556–562.
Fielden, M. R., Adai, A., Dunn, R. T., Olaharski, A., Searfoss, G.,
Sina, J., Aubrecht, J., Boitier, E., Nioi,P., Auerbach, S., et al.
(2011). Development and evaluation of a genomic signature for the prediction and mechanistic assessment of
nongenotoxic hepatocarcinogens in the rat. Toxicol. Sci.
kfr202.
Greenwood, S. K., Hill, R. B., Sun, J. T., Armstrong, M. J., Johnson,
T. E., Gara, J. P., and Galloway, S. M. (2004). Population doubling: a simple and more accurate estimation of cell growth
suppression in the in vitro assay for chromosomal aberrations that reduces irrelevant positive results. Environ. Mol.
Mutagen. 43, 36–44.
Grombacher, T. and Kaina, B. (1995). Constitutive expression and
inducibility of O6-methylguanine-DNA methyltransferase
and N-methylpurine-DNA glycosylase in rat liver cells exhibiting different status of differentiation. Biochim. Biophys.
Acta 1270, 63–72.
Gupta, R. C., Goel, S. K., Earley, K., Singh, B., and Reddy, J. K.
(1985). 32P-postlabeling analysis of peroxisome proliferatorDNA adduct formation in rat liver in vivo and hepatocytes
in vitro. Carcinogenesis 6, 933–936.
Gusenleitner, D., Auerbach, S. S., Melia, T., Gómez, H. F., Sherr, D.
H., and Monti, S. (2014). Genomic models of short-term exposure accurately predict long-term chemical carcinogenicity
and identify putative mechanisms of action. PloS One 9,
e102579.
ICH Expert Working Group. (2011). Guidance on Genotoxicity
Testing and Data Interpretation for Pharmaceuticals
Intended for Human Use. In International Conference on
Harmonisation of Technical Requirements for Registration
of Pharmaceuticals for Human Use, Geneva, ICH S (Vol. 2).
Jensen, M. R., Factor, V. M., Fantozzi, A., Helin, K., Huh, C. G., and
Thorgeirsson, S. S. (2003). Reduced hepatic tumor incidence
in cyclin G1-deficient mice. Hepatology 37, 862–870.
Kannan, K., Amariglio, N., Rechavi, G., Jakob-Hirsch, J., Kela, I.,
Kaminski, N., Getz, G., Domany, E., and Givol, D. (2001). DNA
microarrays identification of primary and secondary target
genes regulated by p53. Oncogene 20, 2225–2234.
Kawachi, T., Yahagi, T., Kada, T., Tazima, Y., Ishidate, M., Sasaki,
M., and Sugiyama, T. (1980). Cooperative programme on
short-term assays for carcinogenicity in Japan. IARC Sci. Publ.
27, 323–330.
Kawase, T., Ohki, R., Shibata, T., Tsutsumi, S., Kamimura, N.,
Inazawa, J., Ohta, T., Ichikawa, H., Aburatani, H., Tashiro, F.,
and Taya, Y. (2009). PH domain-only protein PHLDA3 is a
p53-regulated repressor of Akt. Cell 136, 535–550.
294
|
TOXICOLOGICAL SCIENCES, 2015, Vol. 143, No. 2
Khadapkar, S. V., Chavan, B. G., and Bhide, S. V. (1973). Studies
on the radioactivity in glycogen from mice treated with 35Slabelled thioacetamide. Chem. Biol. Interact. 7, 189–194.
Kirkland, D., Aardema, M., Henderson, L., and Muller, L. (2005).
Evaluation of the ability of a battery of three in vitro genotoxicity tests to discriminate rodent carcinogens and
non-carcinogens I. Sensitivity, specificity and relative predictivity. Mutat. Res. 584, 1–256.
Kirkland, D., Aardema, M., Muller, L., and Makoto, H. (2006).
Evaluation of the ability of a battery of three in vitro genotoxicity tests to discriminate rodent carcinogens and noncarcinogens II. Further analysis of mammalian cell results, relative predictivity and tumour profiles. Mutat. Res. 608, 29–42.
Kiyosawa, N., Tanaka, K., Hirao, J., Ito, K., Niino, N., Sakuma, K.,
Kanbori, M., Yamoto, T., Manabe, S., and Matsunuma, N.
(2004). Molecular mechanism investigation of phenobarbitalinduced serum cholesterol elevation in rat livers by microarray analysis. Arch. Toxicol. 78, 435–442.
Kizer, D. E., Cox, B., Howell, B. A., and Shirley, B. C. (1969). Effect
of hepatocarcinogens on hepatocyte DNA synthesis and cortisone induction of tryptophan oxygenase. Cancer Res. 29,
2039–2046.
Koufaris, C., Wright, J., Currie, R. A., and Gooderham, N. J. (2012).
Hepatic microRNA profiles offer predictive and mechanistic
insights after exposure to genotoxic and epigenetic hepatocarcinogens. Toxicol. Sci. 128, 532–543.
Lefevre, P. A., Tinwell, H., Galloway, S. M., Hill, R., Mackay, J. M.,
Elcombe, C. R., Foster, J., Randall, V., Callander, R. D., and
Ashby, J. (1994). Evaluation of the genetic toxicity of the peroxisome proliferator and carcinogen methyl clofenapate, including assays using Muta Mouse and Big Blue transgenic
mice. Hum. Exp. Toxicol. 13, 764–775.
Li, H., Hou, S., Wu, X., Nandagopal, S., Lin, F., Kung, S., and
Marshall, A. J. (2013). The tandem PH domain-containing
protein 2 (TAPP2) regulates chemokine-induced cytoskeletal
reorganization and malignant B cell migration. PLoS One 8,
e57809.
Lima, B. and Van der Laan, J., (2000). Mechanisms of nongenotoxic carcinogenesis and assessment of the human hazard. Regul. Toxicol. Pharmacol. 32. 135–143.
Lioi, M. B., Scarfi, M. R., Santoro, A., Barbieri, R., Zeni, O., Di, B. D.,
and Ursini, M. V. (1998). Genotoxicity and oxidative stress induced by pesticide exposure in bovine lymphocyte cultures
in vitro. Mutat. Res. 403, 13–20.
Liu, Z., Kelly, R., Fang, H., Ding, D., and Tong, W. (2011).
Comparative analysis of predictive models for nongenotoxic
hepatocarcinogenicity using both toxicogenomics and quantitative structure-activity relationships. Chem. Res. Toxicol. 24,
1062–1070.
Liu, Y., Xu, L., Zeng, Q., Wang, J., Wang, M., Xi, D., Wang, X., Yang,
D., Luo, X., and Ning, Q. (2012). Downregulation of FGL2/
prothrombinase delays HCCLM6 xenograft tumour
growth and decreases tumour angiogenesis. Liver Int. 32,
1585–1595.
Mattioli, F., Garbero, C., Gosmar, M., Manfredi, V., Carrozzino, R.,
Martelli, A., and Brambilla, G. (2004). DNA fragmentation,
DNA repair and apoptosis induced in primary rat
hepatocytes by dienogest, dydrogesterone and 1,4,6androstatriene-17beta-ol-3-one acetate. Mutat. Res. 564,
21–29.
McCracken, S., Kim, C. S., Xu, Y., Minden, M., and Miyamoto, N.
G. (1997). An alternative pathway for expression of p56lck
from type I promoter transcripts in colon carcinoma.
Oncogene 15, 2929–2937.
Melis, J. P., Derks, K. W., Pronk, T. E., Wackers, P., Schaap, M. M.,
Zwart, E., van Ijcken, W. F., Jonker, M. J., Breit, T. M., Pothof, J.,
et al. (2014). In vivo murine hepatic microRNA and mRNA
expression signatures predicting the (non-)genotoxic carcinogenic potential of chemicals. Arch. Toxicol. 88, 1023–1034.
Miau, L. H., Chang, C. J., Tsai, W. H., and Lee, S. C. (1997).
Identification and characterization of a nucleolar phosphoprotein, Nopp140, as a transcription factor. Mol. Cell Biol. 17,
230–239.
Mirkova, E. T. (1994). Activity of the rodent carcinogen 1,4-dioxane in the mouse bone marrow micronucleus assay. Mutat.
Res. 322, 142–144.
Miyashita, T. and Reed, J. C. (1995). Tumor suppressor p53 is a direct transcriptional activator of the human bax gene. Cell 80,
293–299.
Muguruma, M., Nishimura, J., Jin, M., Kashida, Y., Moto, M.,
Takahashi, M., Yokouchi, Y., and Mitsumori, K. (2006).
Molecular pathological analysis for determining the possible
mechanism of piperonyl butoxide-induced hepatocarcinogenesis in mice. Toxicology 228, 178–187.
Muller-Tegethoff, K., Kasper, P., and Muller, L. (1995). Evaluation
studies on the in vitro rat hepatocyte micronucleus assay.
Mutat. Res. 335, 293–307.
Nie, A. Y., McMillian, M., Parker, J. B., Leone, A., Bryant, S., Yieh,
L., Bittner, A., Nelson, J., Carmen, A., Wan, J., et al. (2006).
Predictive toxicogenomics approaches reveal underlying
molecular mechanisms of nongenotoxic carcinogenicity.
Mol. Carcinog. 45, 914–933.
Okamiya, H., Mitsumori, K., Onodera, H., Ito, S., Imazawa, T.,
Yasuhara, K., and Takahashi, M. (1998). Mechanistic study on
liver tumor promoting effects of piperonyl butoxide in rats.
Arch. Toxicol. 72, 744–750.
Ou, Z., Shi, X., Gilroy, R. K., Kirisci, L., Romkes, M., Lynch, C.,
Wang, H., Xu, M., Jiang, M., Ren, S., et al. (2013). Regulation of
the human hydroxysteroid sulfotransferase (SULT2A1) by
RORalpha and RORgamma and its potential relevance to human liver diseases. Mol. Endocrinol. 27, 106–115.
Phillips, J. C., Price, R. J., Cunninghame, M. E., Osimitz, T. G.,
Cockburn, A., Gabriel, K. L., Preiss, F. J., Butler, W. H., and
Lake, B. G. (1997). Effect of piperonyl butoxide on cell replication and xenobiotic metabolism in the livers of CD-1 mice
and F344 rats. Fundam. Appl. Toxicol. 38, 64–74.
Schulte-Hermann, R., Schuppler, J., Timmermann-Trosiener, I.,
Ohde, G., Bursch, W., and Berger, H. (1983). The role of growth
of normal and preneoplastic cell populations for tumor promotion in rat liver. Environ. Health Perspect. 50, 185–194.
Shaw, I. C. and Jones, H. B. (1994). Mechanisms of non-genotoxic
carcinogenesis. Trends Pharmacol. Sci. 15, 89–93.
Sina, J. F., Bean, C. L., Dysart, G. R., Taylor, V. I., and Bradley, M. O.
(1983). Evaluation of the alkaline elution/rat hepatocyte assay as a predictor of carcinogenic/mutagenic potential.
Mutat. Res. 113, 357–391.
Singh, V. K., Ganesh, L., Cunningham, M. L., and Shane, B. S.
(2001). Comparison of the mutant frequencies and mutation
spectra of three non-genotoxic carcinogens, oxazepam, phenobarbital, and Wyeth 14,643, at the lambdacII locus in Big
Blue transgenic mice. Biochem. Pharmacol. 62, 685–692.
Spalice, A., Parisi, P., Nicita, F., Pizzardi, G., Del, B. F., and Iannetti,
P. (2009). Neuronal migration disorders: Clinical, neuroradiologic and genetics aspects. Acta Paediatr. 98, 421–433.
Sribenja, S., Wongkham, S., Wongkham, C., Yao, Q., and Chen, C.
(2013). Roles and mechanisms of beta-thymosins in cell migration and cancer metastasis: an update. Cancer Invest. 31,
103–110.
KOSSLER ET AL.
Storer, R. D., McKelvey, T. W., Kraynak, A. R., Elia, M. C., Barnum,
J. E., Harmon, L. S., Nichols, W. W., and DeLuca, J. G. (1996).
Revalidation of the in vitro alkaline elution/rat hepatocyte
assay for DNA damage: Improved criteria for assessment of
cytotoxicity and genotoxicity and results for 81 compounds.
Mutat. Res. 368, 59–101.
Suzuki, T., Jin, M., Dewa, Y., Ichimura, R., Shimada, Y.,
Mizukami, S., Shibutani, M., and Mitsumori, K. (2010).
Evaluation of in vivo liver genotoxic potential of Wy-14,643
and piperonyl butoxide in rats subjected to two-week repeated oral administration. Arch. Toxicol. 84, 493–500.
Takasawa, H., Suzuki, H., Ogawa, I., Shimada, Y., Kobayashi, K.,
Terashima, Y., Matsumoto, H., Aruga, C., Oshida, K., Ohta, R.,
et al. (2010). Evaluation of a liver micronucleus assay in young
rats (III): A study using nine hepatotoxicants by the
Collaborative Study Group for the Micronucleus Test
(CSGMT)/Japanese Environmental Mutagen Society (JEMS)Mammalian Mutagenicity Study Group (MMS). Mutat. Res.
698, 30–37.
Tasaki, M., Kuroiwa, Y., Inoue, T., Hibi, D., Matsushita, K., Ishii,
Y., Maruyama, S., Nohmi, T., Nishikawa, A., and Umemura,
T. (2013). Oxidative DNA damage and in vivo mutagenicity caused by reactive oxygen species generated in the
livers of p53-proficient or -deficient gpt delta mice treated
with non-genotoxic hepatocarcinogens. J. Appl. Toxicol. 33,
1433–1441.
Thomas, R.S., O’Connell, T.M., Plita, L., Wolfinger, R.D., Yang, L.,
and Page, T.J. (2007a). A comparison of transcriptomic and
metabonomic technologies for identifying biomarkers predictive of two-year rodent cancer bioassays. Toxicol. Sci. 96,
40–46.
Thomas, R. S., Pluta, L., Yang, L., and Halsey, T. A. (2007b).
Application of genomic biomarkers to predict increased lung
tumor incidence in 2-year rodent cancer bioassays. Toxicol.
Sci. 97, 55–64.
|
295
Topinka, J., Oesterle, D., Reimann, R., and Wolff, T. (2004a). Noeffect level in the mutagenic activity of the drug cyproterone
acetate in rat liver. Part I. Single dose treatment. Mutat. Res.
550, 89–99.
Topinka, J., Oesterle, D., Reimann, R., and Wolff, T. (2004b). Noeffect level in the mutagenic activity of the drug cyproterone
acetate in rat liver. Part II. Multiple dose treatment. Mutat.
Res. 550, 101–108.
Uehara, T., Minowa, Y., Morikawa, Y., Kondo, C., Maruyama, T.,
Kato, I., Nakatsu, N., Igarashi, Y., Ono, A., Hayashi, H., et al.
(2011). Prediction model of potential hepatocarcinogenicity
of rat hepatocarcinogens using a large-scale toxicogenomics
database. Toxicol. Appl. Pharmacol. 255, 297–306.
Watanabe, T., Suzuki, T., Natsume, M., Nakajima, M., Narumi, K.,
Hamada, S., Sakuma, T., Koeda, A., Oshida, K., Miyamoto, Y.,
et al. (2012). Discrimination of genotoxic and non-genotoxic
hepatocarcinogens by statistical analysis based on gene expression profiling in the mouse liver as determined by quantitative real-time PCR. Mutat. Res. 747, 164–175.
Waters, M. D., Jackson, M., and Lea, I. (2010). Characterizing and
predicting carcinogenicity and mode of action using conventional and toxicogenomics methods. Mutat. Res. 705, 184–200.
Watt, P. M. and Hickson, I. D. (1994). Structure and function of
type II DNA topoisomerases. Biochem. J. 303(Pt 3), 681–695.
Waxman, D. J. (1999). P450 gene induction by structurally diverse
xenochemicals: Central role of nuclear receptors CAR, PXR,
and PPAR. Arch. Biochem. Biophys. 369, 11–23.
White, C. D., Brown, M. D., and Sacks, D. B. (2009). IQGAPs in cancer: a family of scaffold proteins underlying tumorigenesis.
FEBS Lett. 583, 1817–1824.
Wolff, T., Topinka, J., Deml, E., Oesterle, D., and Schwarz, L. R.
(2001). Dose dependent induction of DNA adducts, gene mutations, and cell proliferation by the antiandrogenic drug cyproterone acetate in rat liver. Adv. Exp. Med. Biol. 500,
687–696.