Chemosphere 144 (2016) 2150e2157
Contents lists available at ScienceDirect
Chemosphere
journal homepage: www.elsevier.com/locate/chemosphere
Classification and toxicity mechanisms of novel flame retardants
(NFRs) based on whole genome expression profiling
Miao Guan a, Guanyong Su a, John P. Giesy a, b, c, d, e, Xiaowei Zhang a, *
a
State Key Laboratory of Pollution Control and Resource Reuse & School of the Environment, Nanjing University, Nanjing, China
Department of Veterinary Biomedical Sciences and Toxicology Centre, University of Saskatchewan, Saskatoon, Saskatchewan, Canada
c
Department of Zoology, and Center for Integrative Toxicology, Michigan State University, East Lansing, MI, USA
d
School of Biological Sciences, University of Hong Kong, Hong Kong, China
e
Department of Biology, Hong Kong Baptist University, Hong Kong, China
b
h i g h l i g h t s
Assess mechanisms of toxic modes of action of six NFRs in genome-wide level.
NFRs were clustered based on expression of multiple genes that responded.
Clustering by molecular descriptors was consistent with that by gene profiles.
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 8 August 2015
Received in revised form
25 October 2015
Accepted 26 October 2015
Available online xxx
Recently some novel alternative flame retardants (NFRs), which have been widely applied to meet demands for mandated flame retardation of products, have been detected in various matrices of the
environment. However, knowledge on toxic effects and associated molecular mechanisms of these
chemicals was limited. Here, toxic mechanisms of action of six NFRs, bis (2-ethylhexyl) phosphate
(BEHP), chlorendic acid (Het acid), 2,2-bis (bromomethyl)-1,3-propanediol (BMP), tris (2-butoxyethyl)
phosphate (TBEP), triethyl phosphate (TEP), tributyl phosphate (TBP) were investigated by use of a library containing ~1820 modified green fluorescent protein (GFP) expressing promoter reporter vectors
constructed from Escherichia coli K12(E.coli). BEHP, Het acid, BMP, TBEP, TEP, TBP inhibited growth of
E. coli with 4 h 10%-inhibition concentrations of 53.0e3102.3 mM. A total of 119, 44, 26, 131, 62, 103 genes
out of 336 genes selected during preliminary screening were significantly altered with fold-changes
greater than 1.5 by BEHP, Het acid, BMP, TBEP, TEP and TBP, respectively. GO analyses of responsive
genes suggested that RNA and primary metabolism process were involved in molecular mechanisms of
toxicity. Chemical clustering based on expression of 62 multi-responsive genes showed that BEHP, TBP
and TBEP were grouped together, which is consistent with similarity of their chemical structures,
especially for BEHP and TBP. Clustering by molecular descriptors and molecular activity by use of the
multivariate classification system ToxCast was consistent with that by profiles of multi-responsive genes.
The results of this study demonstrated the utility of the E. coli, whole-cell assay for determining
mechanisms of toxic action of chemicals.
© 2015 Elsevier Ltd. All rights reserved.
Keywords:
NFRs
E coli
GFP reporter
Genomics
Toxic mechanism
Clustering
1. Introduction
Flame retardants are ubiquitous in the environment due to their
* Corresponding author. School of the Environment, Nanjing University, Nanjing,
210046, China.
E-mail
addresses:
zhangxw@nju.edu.cn,
howard50003250@yahoo.com
(X. Zhang).
http://dx.doi.org/10.1016/j.chemosphere.2015.10.114
0045-6535/© 2015 Elsevier Ltd. All rights reserved.
use to inhibit or resist the spread of fire, in thermoplastics, thermosets, foams, textiles, electronics and coatings. Since the global
banishing of some brominated flame retardants that were used
historically, such as polybrominated diphenyl ethers (PBDEs),
because of their high-performance and low-cost, production of
some novel flame retardants (NFRs) like organophosphate flame
retardants have been increased to meet demand required by
various jurisdictions (van der Veen and de Boer, 2012). Moreover,
some alternative flame retardants have been widely detected in
M. Guan et al. / Chemosphere 144 (2016) 2150e2157
indoor air (Marklund et al., 2003; Hartmann et al., 2004), waters
(Andresen et al., 2004; Teo et al., 2015) and soils (Ingram et al.,
1996). This has raised concerns about potential toxicity of novel
alternative flame retardants to wildlife and humans.
Previously, toxicity data for NFRs have been based primarily on
in vivo tests with animals. Bis (2-ethylhexyl) phosphate (BEHP)
induced oxidative stress by proliferation of both peroxisomes and
mitochondria in rat liver (Lundgren and DePierre, 1987). Chlorendic
acid (Het acid) was determined to be a clastogen by use of the
in vitro mouse lymphoma assay and mutagenic in the L5178Y/
TK ± mouse lymphoma assay in the absence of S9 activation
(McGregor et al., 1988; Sofuni et al., 1996). 2,2-Bis (bromomethyl)1,3-propanediol (BMP) can damage DNA in-vivo and might be
associated with oxidative stress (Kong et al., 2011). Tris (2butoxyethyl) phosphate (TBEP) caused developmental toxicity by
inhibiting degradation and utilization of nutrients and inducing
apoptosis in zebrafish (Han et al., 2014). Triethyl phosphate (TEP)
increased activity of reductase in rat liver microsomal preparations
(Noboru et al., 1987). Triethyl phosphate (TBP) induced lung damage probably via the depression of key antioxidant enzymes and
elevation of lipid peroxidation (Salovsky et al., 1998). However,
information on molecular mechanisms of toxicity for these NFRs
was limited, especially genome-wide information.
Sequencing of the complete genome of Escherichia coli K-12 and
fusion of stress promoters to fluorescent transcriptional reporters
prompted development of a useful toxicogenic approach to characterize toxic modes of action of chemicals or samples (Su et al.,
2012; Zhang et al., 2011; Fu et al., 2013; Su et al., 2013; Hug et al.,
2015). For each strain, fusion of stress promoters to the GFP protein gene provides a mechanism for detection of modulation of
cellular signaling, which makes analyses of differential expression
of genes easier and more accurate (Elad et al., 2010). Compared
with microarray technology, live cell arrays avoid complex protocols of pre-treatment and high-costs of experimental materials
have fewer interferences and can provide temporal resolution
(Onnis-Hayden et al., 2009). Furthermore, the short time required
to complete a test makes use of live cell arrays rapid, economical,
high-throughput biosensor systems for detecting toxicity and
determining effects on specific signaling pathways.
Profiles of expression of genes can reveal mechanisms of toxic
actions of chemicals, which are correlated to both the structure of
chemicals (molecular descriptors) and structure of the target of test
organism, which can be assessed by use of E. coli reporter genes as
demonstrated by the results presented here. Chemicals with similar
mechanisms of toxic action produced similar profiles of transcriptional expression (Waring et al., 2001), which was useful for clustering of compounds or samples based on changes in patterns of
expression of genes caused by each chemical (Su et al., 2014; Hug
et al., 2015).
In this study the E.coli, microbial reporter gene assay was used
to: 1) assess mechanisms of toxic actions of six NFRs; 2) identify
multi-responsive genes which were responsive to multiple chemicals and based on expression of these multi-responsive genes to
cluster six NFRs based on their effects on expression of genes in
multiple pathways.
2. Materials and methods
2.1. Chemicals
Bis (2-ethylhexyl) phosphate (BEHP), chlorendic acid (Het acid),
2,2-bis (bromomethyl)-1,3-propanediol (BMP), tris (2-butoxyethyl)
phosphate (TBEP), triethyl phosphate (TEP), tributyl phosphate
(TBP) were obtained from Sigma Aldrich (St. Louis, MO, USA). Stock
solutions of six chemicals were prepared in dimethyl sulfoxide
2151
(DMSO), and other test concentrations were made by serial dilution
with DMSO. Structures and potencies for cytotoxicity information
are given (Table 1 and Fig. 1).
2.2. Microbial live cell array
The collection of microbial promoters developed by the Weizmann Institute of Science, which includes most of the genome of
E. coli K12 strain MG1655 (1820/2500) was used to assess dynamic
expression of genes (Zaslaver et al., 2006). Each of the reporter
strains is coupled with a bright, fast-folding green fluorescent
protein (GFP) fused to a full-length copy of an E. coli promoter in a
low-copy plasmid. This makes quantification of expression of genes
easier and faster without the need to extract DNA/RNA. All clones
were grown separately using 96-well plate (Corning, NY, USA) at
37 C in LB-Lennox media plus 25 mg/L kanamycin.
2.3. Cytotoxicity
For each of the NFRs studied, six concentrations (n ¼ 3) were
selected for use in tests of cytotoxicity to E. coli in 96-well plate, and
the maximum concentration which was determined by chemical
cytotoxicity and maximum solubility is listed in Table 1. Here, the
vital stain, alamar blue was used as an indicator of cytotoxicity.
Alamar blue is not toxic to cells and is a more sensitive measure
than OD600 (Su et al., 2012). After 3 h incubation at 37 C, alamar
blue was used to assess whether cells had sufficient capacity to
proliferate. After dyeing for 1 h with alamar blue, blue-red fluorescence was quantified by use of a Synergy H4 hybrid microplate
reader (excitation at 545 nm and emission at 590 nm) (Bio Tek
Instruments Inc., Winooski, VT).
2.4. High throughput screening
Strains of E. coli were inoculated into a fresh 96-well plate from a
96-well stock plate by use of disposable replicators (Genetix, San
Jose, CA, USA). Cells were incubated at 37 C for 3 h in 96-well plate
and then transferred into 384-well plate (NUNC, Rochester, NY,
USA). Finally, 3.75 mL of DMSO (solvent control) or chemical solutions were added into individual wells on the 384-well plate to
make a final concentration of 10% inhibition concentration (IC10).
Intensity of fluorescence of GFP in each well was consecutively
monitored every 10 min for 4 h by use of a Synergy H4 hybrid
microplate reader (excitation/emission: 485 nm/528 nm). Differential expressions of genes of 1820 E. coli reporter strains exposed
to the six NFRs, which were BEHP, Het acid, BMP, TBEP, TEP and TBP,
were obtained in two batches, where each batch contained 21 96well plates. The response measured as fluorescence of GFP was
fitted to a function of time for each promoter reporter strain with a
p value less than 0.001. Genes that changed in response to exposure
to the six NFRs with maximum fold changes greater than 2-fold
were selected to be monitored in a series of concentrations of
each of the individual NFRs.
For validation, all selected E. coli reporter strains were exposed
to each of three concentrations representing 0.01*IC10, 0.1*IC10
and IC10. To select the final promoter reporter genes that were
significantly differentially expressed in response to NFRs, a linear
regression model was applied. Changes in expressions of genes
which exhibited a significant correlation between magnitude of
response and time and also concentrations with p values less than
0.001 and a maximum fold change greater than 1.5 or 2 were
considered to be significant. Details of the analyses applied to the
data have been previously described (Su et al., 2012; Zhang et al.,
2011; Gou et al., 2010). Lists of genes which derived from three
series concentrations validation test were developed for analysis of
2152
M. Guan et al. / Chemosphere 144 (2016) 2150e2157
Table 1
Cytotoxicity of six novel flame retardants (NFRs) to E. coli.
No.
Chemicals
Abbreviation
CAS
Test concentration range (mM)
IC10b (mM)
IC50c (mM)
1
2
3
4
5
6
Bis (2-ethylhexyl) phosphate
Chlorendic acid
2,2-Bis(bromomethyl)-1,3-propanediol
Tris(2-butoxyethyl) phosphate
Triethyl phosphate
Tributyl phosphate
BEHP
Het acid
BMP
TBEP
TEP
TBP
298-07-7
115-28-6
3296-90-0
78-51-3
78-40-0
126-73-8
0e17114.3
0e8381.3
0e19874.8
0e15511.8
0e39224.9
0e36678.5
145.8
53.0
871.6
549.9
3102.3
375.1
5867.8
2484.3
6721.8
NAa
NAa
NAa
a
b
c
NA means not achieved within the test concentration.
IC10 means 10% inhibitory concentration of a NFR after a 4-hr exposure.
IC50 means median inhibition concentration of a NFR after a 4-hr exposure.
Fig. 1. Structures of six novel flame retardants (NFRs).
mechanisms of toxic action based on a cutoff of 1.5 fold-changes.
Assessment of mechanisms of toxic action was conducted by use
of GO gene set enrichment analysis by use of the R package clusterProfiler (Yu et al., 2012). P-values which were adjusted for
multiple comparisons less than 0.01 and q-values less than 0.05
were also calculated as cutoff for FDR control (Storey, 2003). Annotations of responsive genes were obtained from website (www.
ecogene.org).
2.5. Clustering of NFRs
Chemicals with similar patterns of expression of genes were
clustered together. To avoid inclusion of unaltered genes that did
not contribute to categorization of NFRs, only those altered by at
least three NFRs (p value < 0.001& fold change > 1.5) in the validation test were used to classify NFRs. Classification of NFRs based
on similarities of differentially expressed genes was accomplished
by use of ToxClust (Zhang et al., 2009). Dissimilarities among genes
were calculated by use of Manhattan distances between expressions among genes across three concentrations and 25 time points.
Other data on molecular toxicity for the six NFRs by U.S. Environmental Protection Agency (EPA) were obtained from ToxCast
(Dix et al., 2007) (http://actor.epa.gov/dashboard/). There were 174,
382, 384, 170, 167 and 404 tested assays and 42, 1, 1, 15, 1 and 16
active assays for BEHP, Het acid, BMP, TBEP, TEP and TBP, respectively. Of 431 assays conducted, 61which were active by at least one
NFR were chosen for toxicity mechanism comparison. NFRs with
more overlap active assays were considered having more similar
toxicity mechanism.
Molecular descriptors for the six NFRs were calculated by use of
E-Dragon (Mauri et al., 2006), which provides more than 1600
molecular descriptors that are divided into 20 logical blocks (http://
www.vcclab.org). Molecular descriptors of constant expression
among the six NFRs were removed. The final dataset consisted of
1492 descriptors which contained 31 constitutional indices, 10 ring
descriptors, 72 topological indices, 42 walk and path counts, 37
connectivity indices, 48 information indices, 492 2-D matrix-based
descriptors, 211 2-D autocorrelations, 91 burden eigenvalues, 33 PVSA-like descriptors, 23 ETA indices, 303 edge adjacency indices, 10
atom-centered fragments, 10 atom-type E-state indices, 14 CATS
2D, 37 2-D atom pairs, 20 molecular properties and 8 drug-like
indices
(http://www.talete.mi.it/products/dragon_molecular_
descriptor_list.pdf). Molecular descriptors were normalized by
range standardization (X-Xmin)/(Xmax-Xmin). The Cluster Affinity
Search Technique (CAST) was applied to 1492 descriptors and 81
clusters with an inclusion threshold of 0.8 identified (Ben-Dor et al.,
1999). The median profile of each cluster of molecular descriptors
(MCP) was chosen to represent each cluster. Hierarchical clustering
M. Guan et al. / Chemosphere 144 (2016) 2150e2157
2153
53.0, 871.6, 549.9, 3102.3 and 375.1 mM (Hamilton et al., 1977). IC50
values could be calculated for only BEHP, Het acid and BMP and
were 5867.8, 2484.3 and 6721.8 mM, respectively. This result suggested that these NFRs can cause toxicity to E. coli but only at very
high concentrations. The IC10 was selected as the test concentration to be used in studies of expression of genes for use in classification of the six NFRs.
3.2. Profiles of expression of genes
Fig. 2. Inhibition of growth of E. coli growth inhibition profiles by novel flame retardants (NFRs) at different concentrations (Data points represent mean and standard
error).
of the six NFRs by use of MCPs was conducted by use of Spearman
distances.
3. Results and discussion
3.1. Cytotoxicity
After a 4-h exposure of E. coli reporter strains to NFRs, different
profiles of term were observed (Fig. 2, Table 1). Three NFRs, BEHP,
Het acid and BMP, were cytotoxic and inhibited E. coli cells in a
concentration-dependent manner, with maximum inhibitions of
100%, 95% and 81%, respectively. For TBEP, TEP and TBP, slight inhibitions were observed, with maximum inhibitions 33% 33%, and
43%, respectively. After exposure to BEHP, Het acid, BMP, TBEP, TEP
or TBP, IC10 values based on probit model analyses, were 145.8,
After exposure to BEHP, Het acid, BMP, TBEP, TEP or TBP, 336
genes were modulated by at least 2-fold by at least one of the NFRs.
Of these 336 genes, 119, 44, 26, 131, 62, 103 were significantly
altered by 1.5-fold or greater by BEHP, Het acid, BMP, TBEP, TEP and
TBP, respectively (Table S2eS7). Furthermore, 62, 24, 9, 75, 37, 56
genes were significantly altered by 2-fold or greater for the six
NFRs, respectively. More genes were up-regulated than downregulated due to 4 h exposure to the six NFRs with cutoffs of
either 1.5 or 2 (Fig. S1eS12). The number of genes altered by
exposure to NFRs with maximum fold change greater than 1.5 was
proportional to concentrations of the individual NFRs. Strains that
responded to lesser or moderate concentrations were responsive to
greater concentrations as well (Fig. S13). Genes that were responsive to at least one concentration were selected for use in determining toxicity mechanisms of toxic actions. Genes modulated by
the greatest concentration, which represented the most comprehensive potential toxicity mechanism, were chosen for further
clustering NFRs. Meanwhile, 62 genes which were modulated by at
least three NFRs were selected as multi-responsive genes for
further clustering NFRs.
3.3. Mechanisms of toxic action
Gene ontology (GO) biological processes were inferred and used
to understand biochemical pathways that were modulated by
exposure to each of the six NFRs. GO biological processes modulated by the five NFRs (except BMP) included primary metabolic
Fig. 3. Biological processes (BP) GO of responsive genes (fold change > 1.5, IC10) modulated by bis (2-ethylhexyl) phosphate (BEHP), chlorendic acid (Het acid), tris(2-butoxyethyl)
phosphate (TBEP), triethyl phosphate (TEP) and tributyl phosphate (TBP) (A) ~ (E). The Venn diagram displays the overlap of five NFRs' BP GO terms (F).
2154
M. Guan et al. / Chemosphere 144 (2016) 2150e2157
process, cellular metabolic process, organic substance metabolic
process, cellular process, metabolic process and biological process
by responsive genes of NFRs (p < 0.01, q < 0.05). There was no
significant GO pathway enriched with BMP. The common pathways
affected by the five NFRs were primary processes occurring in cell
and important for maintaining homeostasis (Maughan and
Nicholson, 2011). Modification of RNA; tRNA metabolism process,
tRNA aminoacylation for protein translation, ncRNA metabolic
process, amino acid activation, tRNA aminoacylation and organonitrogen compound metabolic process; small molecular metabolism process; macromolecular biosynthetic process and cellular
macromolecule biosynthetic process were the pathways only
modulated by BEHP, TBP, TBEP and Het acid, respectively. Gene
expression was the pathway modulated by Het acid and BEHP.
Translation, cellular amino acid metabolic process, protein metabolic process, cellular biosynthetic process and cellular protein
metabolic process were modulated by BEHP and TBP. Nucleobasecontaining compound metabolic process, cellular aromatic compound metabolic process, RNA metabolic process, cellular nitrogen
compound metabolic process, heterocycle metabolic process,
nucleic acid metabolic process and organic cyclic compound
metabolic process were enriched by BEHP, Het acid and TBP; BEHP,
Het acid, TBEP and TBP and common pathway of nitrogen compound metabolic process. BEHP, Het acid, TEP and TBP have common pathways of macromolecule metabolic process and cellular
macromolecule metabolic process. 21 common GO pathways of
Fig. 4. Clustering of time-dependent expression of six novel flame retardants (NFRs) based on profiles of 62 multi-responsive genes. Classification and visualization of the gene
expression were derived by use of ToxClust (Zhang et al., 2009). Color gradient represent fold change of gene expression and the time course of expression changes were indicated
form left to right. Bis (2-ethylhexyl) phosphate (BEHP), tributyl phosphate (TBP) and tris(2-butoxyethyl) phosphate (TBEP) were classified together, and the other three NFRs,
chlorendic acid (Het acid), triethyl phosphate (TEP) and 2,2-bis(bromomethyl)-1,3-propanediol (BMP), were different from others and classified into unique clusters. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
M. Guan et al. / Chemosphere 144 (2016) 2150e2157
BEHP (23 pathways) and TBP (27 pathways) indicated that the
mechanisms of toxicity of the two OPFRs were more similar to the
other four NFRs; Six GO pathways of TBEP (8 pathways) and TEP (8
pathways) indicated that mechanisms toxicity of the two OPFRs
were more similar to the other four NFRs based on the GO
enrichment analysis (Fig. 3). Potential mechanisms of toxicity of
BEHP, TEP and Het acid were primary processes of metabolism
those that influence transcription of RNA. Potential key mechanisms of toxic actions of TBEP and TEP were primary metabolism
processes.
3.4. Clustering of NFRs
Based on differential expression of genes, a chemical classification was conducted. A total of 62 genes which were significantly
altered by at least three NFRs (p < 0.001 & fold change > 1.5) were
selected and data from the greatest concentration was used for
2155
clustering of the six NFRs (Fig. 4). BEHP, TBP and TBEP, which have
similar chemical structures, were classified together (Fig. 1). Especially for BEHP and TBP, patterns of modulations of genes caused by
these two chemicals were quite similar, while the patterns of
expression caused by the other three NFRs, Het acid, TEP and BMP,
were different from the others.
Multi-responsive genes that were altered by exposure to at least
three of the NFRs and thus might provide superior power of
discrimination are listed. These 62 genes were classified into four
groups: enzymes or putative enzymes; regulatory proteins or putative regulator proteins; structural proteins and those of unspecified function, accounted for 48.4%, 32.3%, 14.5%, and 4.8% of the 62
modulated genes, respectively (Table S1). Among these genes, iscR,
pspB, ycaC and yncC were differentially expressed due to exposure
among five NFRs. Genes expression of dksA, ecfG, efp, galU, hisS, slyA,
stpA, tolB, trmA, xseA, yadF, ycfQ, yciA, ydiV, yeaT, yhcO, yjjK and znuA
were altered by exposure to four NFRs. The other 40 genes were
Fig. 5. Parallel heat map of six novel flame retardants (NFRs) based on the activity of ToxCast test assays. Yellow color represented active assays, blue color represented inactive
assays and gray color represented assays not tested with this chemical. Bis (2-ethylhexyl) phosphate (BEHP), tributyl phosphate (TBP) and tris(2-butoxyethyl) phosphate (TBEP)
have similar active assays, while the other three NFRs, triethyl phosphate (TEP), 2,2-bis(bromomethyl)-1,3-propanediol (BMP) and chlorendic acid (Het acid), which has very
different active assay. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
2156
M. Guan et al. / Chemosphere 144 (2016) 2150e2157
altered by exposure to three NFRs. iscR the transcript of which is a
protein that is a repressor of DNA-binding transcription, which
regulates transcription of several operons and genes involved in
biogenesis of FeeS clusters was modulated by exposure to BEHP,
Het acid, TBEP, TEP and TBP. PspB, psp operon transcription coactivator was responsive for BEHP, TBEP, TEP, TBP and Het acid.
YcaC, putative isochorismatase family hydrolase was responsive for
BEHP, Het acid, BMP, TBEP and TBP. YncC, coding colonic acid and
biofilm gene transcriptional regulator was responsive for BEHP, Het
acid, BMP, TBEP and TEP. Of these 62 genes, 11 genes which were
stpA, yncC, iscR, allR, yjeB, slyA, yahB, ycfQ, yeaT, ygfI and yegE were in
the term for regulation of transcription, regulation of RNA metabolic process. 12 genes which were deaD, yncC, iscR, allR, yjeB, papB,
lsrR, slyA, yahB, ycfQ, yeaT and ygfI were in the term for transcription
based on gene ontology biological process analysis. That might
implied that these six NFRs could be distinguished in the expression of genes coding transcription.
Sixty one molecular toxicity assays from ToxCast also indicated
that BEHP, TBP and TBEP have similar toxicity activity. Three assays
which were ATG_PPARg_TRANS_up (target gene: PPARG),
ATG_PXRE_CIS_up (target gene: NR1I2) and ATG_VDRE_CIS_up
(target gene: VDR) were active for BEHP, TBEP and TBP. Four assays
which were Tox21_ARE_BLA_agonist_ratio (target gene: NFE2L2),
Tox21_Aromatase_Inhibition (target gene: CYP19A1), Tox21_TR_LUC_GH3_Antagonist (target gene: THRB) and ACEA_T47D_80hr_Negative were active for BEHP and TBEP. Three assays which were
NVS_MP_hPBR (target gene: TSPO), NVS_MP_rPBR (target gene:
Tspo) and OT_FXR_FXRSRC1_0480 (target gene: NR1H4) were
active for TBEP and TBP. One assay which was ATG_PXR_TRANS_up
(target gene: NR1I2) was active for BEHP and TBP. While the other
three NFRs, which were TEP (only active by ACEA_T47D_80hr_Positive assay, ESR1), BMP (only active by NVS_ENZ_oCOX2 assay,
PTGS2) and Het acid (only active by NVS_GPCR_gLTB4 assay, Ltb4r),
have very different active assays (Fig. 5). The grouping of six NFRs
by molecular toxicity data from ToxCast was consistent with that
obtained by use of profiles of differential expression of genes.
Molecular descriptors of NFRs can be used to gain insights into
potential modes of toxic action of NFRs. Based on the 81 median
cluster profiles of molecular descriptors, the six NFRs were clustered into four clusters: Cluster 1 contained BEHP, TBP and TBEP
while Cluster 2 contained: Het acid and TEP and BMP were clustered alone in Clusters 3 and 4 (Fig. 6). Clustering by use of molecular descriptors was also consistent with that obtained by use of
profiles of differential expression of genes.
4. Conclusions
Cytotoxicity of six NFRs, BEHP, Het acid, BMP, TBEP, TEP and TBP,
expressed as the IC10, were 145.8, 53.0, 871.6, 549.9, 3102.3 and
375.1 mM, respectively. Of the 336 genes identified in the initial
screening, expression of 119, 44, 26, 131, 62, 103 genes were
modulated by a factor of 1.5 by BEHP, Het acid, BMP, TBEP, TEP and
TBP, respectively. Analysis of biological processes based on GO
helped elucidate potential mechanisms of toxic actions of the six
NFRs. BEHP, TBP and TBEP were clustered together based on both
gene expression of multi-responsive genes and molecular descriptors. Het acid, BMP and TEP separated into separate clusters.
Flame retardants with similar mode of action trend to have
similar gene expression profiles. In the future, we can investigate
mode of action of some other novel flame retardants by clustering
them with flame retardants with known mode of action by gene
expression profiles to predict mode of action of novel flame
retardants.
Fig. 6. Clustering of six novel flame retardants (NFRs) based on median cluster profiles of molecular descriptors. Bis (2-ethylhexyl) phosphate (BEHP), tributyl phosphate (TBP) and
tris(2-butoxyethyl) phosphate (TBEP) were classified together, while the other three NFRs, chlorendic acid (Het acid), triethyl phosphate (TEP) and 2,2-bis(bromomethyl)-1,3propanediol (BMP), which were clustered separately (cutoff was represented by the red line). (For interpretation of the references to color in this figure legend, the reader is
referred to the web version of this article.)
M. Guan et al. / Chemosphere 144 (2016) 2150e2157
Acknowledgments
We thank the National Science Foundation of China for the
Excellent Yong Scientist Grant (21322704) and Research Fund for
the Doctoral Program of Higher Education of China (Grant:
20120091110034). This work is also supported by the Seventh
Framework Programme (the SOLUTIONS Project, FP7-ENV-2013) of
the European Union under grant agreement no. 603437. Prof. Giesy
was supported by the program of 2012 “High Level Foreign Experts”
(#GDT20143200016) funded by the State Administration of Foreign
Experts Affairs, the P.R. China to Nanjing University and the Einstein
Professor Program of the Chinese Academy of Sciences. He was also
supported by the Canada Research Chair program.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.chemosphere.2015.10.114.
References
Andresen, J.A., Grundmann, A., Bester, K., 2004. Organophosphorus flame retardants
and plasticisers in surface waters. Sci. Total Environ. 332, 155e166.
Ben-Dor, A., Shamir, R., Yakhini, Z., 1999. Clustering gene expression patterns.
J. Comput. Biol. 6, 281e297.
Dix, D.J., Houck, K.A., Martin, M.T., Richard, A.M., Setzer, R.W., Kavlock, R.J., 2007.
The ToxCast program for prioritizing toxicity testing of environmental chemicals. Toxicol. Sci. 95, 5e12.
Elad, T., Lee, J.H., Gu, M.B., Belkin, S., 2010. Microbial Cell Arrays. Whole Cell Sensing
Systems I. Springer, pp. 85e108.
Fu, J., Hu, X., Tao, X., Yu, H., Zhang, X., 2013. Risk and toxicity assessments of heavy
metals in sediments and fishes from the Yangtze River and Taihu Lake, China.
Chemosphere 93, 1887e1895.
Gou, N., Onnis-Hayden, A., Gu, A.Z., 2010. Mechanistic toxicity assessment of
nanomaterials by whole-cell-array stress genes expression analysis. Environ.
Sci. Technol. 44, 5964e5970.
Hamilton, M.A., Russo, R.C., Thurston, R.V., 1977. Trimmed Spearman-Karber
method for estimating median lethal concentrations in toxicity bioassays. Environ. Sci. Technol. 11, 714e719.
Han, Z., Wang, Q., Fu, J., Chen, H., Zhao, Y., Zhou, B., Gong, Z., Wei, S., Li, J., Liu, H.,
2014. Multiple bio-analytical methods to reveal possible molecular mechanisms
of developmental toxicity in zebrafish embryos/larvae exposed to tris (2butoxyethyl) phosphate. Aquat. Toxicol. 150, 175e181.
Hartmann, P.C., Bürgi, D., Giger, W., 2004. Organophosphate flame retardants and
plasticizers in indoor air. Chemosphere 57, 781e787.
Hug, C., Zhang, X., Guan, M., Krauss, M., Bloch, R., Schulze, T., Reinecke, T., Brack, W.,
2015. Microbial reporter gene assay as a diagnostic and early warning tool for
the detection and characterization of toxic pollution in surface waters. Environ.
Toxicol. Chem. 34, 2523e2532.
Ingram, J., Groenewold, G., Appelhans, A., Dahl, D., Delmore, J., 1996. Detection limit
and surface coverage determination for tributyl phosphate on soils by static
SIMS. Anal. Chem. 68, 1309e1316.
Kong, W., Kuester, R.K., Gallegos, A., Sipes, I.G., 2011. Induction of DNA damage in
human urothelial cells by the brominated flame retardant 2,2bis(bromomethyl)-1,3-propanediol: role of oxidative stress. Toxicology 290,
271e277.
2157
Lundgren, B., DePierre, J., 1987. Induction of xenobiotic-metabolizing enzymes and
peroxisome proliferation in rat liver caused by dietary exposure to di (2ethylhexyl) phosphate. Xenobiotica 17, 585e593.
Marklund, A., Andersson, B., Haglund, P., 2003. Screening of organophosphorus
compounds and their distribution in various indoor environments. Chemosphere 53, 1137e1146.
Maughan, H., Nicholson, W.L., 2011. Increased fitness and alteration of metabolic
pathways during Bacillus subtilis evolution in the laboratory. Appl. Environ.
Microbiol. 77, 4105e4118.
Mauri, A., Consonni, V., Pavan, M., Todeschini, R., 2006. Dragon software: an easy
approach to molecular descriptor calculations. MATCH Commun. Math. Comput. Chem. 56, 237e248.
McGregor, D.B., Brown, A., Cattanach, P., Edwards, I., McBride, D., Riach, C.,
Caspary, W.J., 1988. Responses of the L5178Y tkþ/tk mouse lymphoma cell
forward mutation assay: III. 72 coded chemicals. Environ. Mutagen. 12, 85e154.
Noboru, F., Hitoshi, N., Masuo, S., Yasuo, S., 1987. Induction of the hepatic microsomal cytochrome P-450 system by trialkyl phosphorothioates in rats. Biochem.
Pharmacol. 36, 1291e1296.
Onnis-Hayden, A., Weng, H., He, M., Hansen, S., Ilyin, V., Lewis, K., Gu, A.Z., 2009.
Prokaryotic real-time gene expression profiling for toxicity assessment. Environ. Sci. Technol. 43, 4574e4581.
Salovsky, P., Shopova, V., Dancheva, V., 1998. Antioxidant defense mechanisms in
the lung toxicity of tri-n-butyl phosphate. Am. J. Ind. Med. 33, 11e15.
Sofuni, T., Honma, M., Hayashi, M., Shimada, H., Tanaka, N., Wakuri, S., Awogi, T.,
Yamamoto, K., Nishi, Y., Nakadate, M., 1996. Detection of in vitro clastogens and
spindle poisons by the mouse lymphoma assay using the micro well method:
interim report of an international collaborative study. Mutagenesis 11,
349e355.
Storey, J.D., 2003. The positive false discovery rate: a Bayesian interpretation and
the q-value. Ann. Stat. 2013e2035.
Su, G., Yu, H., Lam, M.H., Giesy, J.P., Zhang, X., 2014. Mechanisms of toxicity of hydroxylated polybrominated diphenyl ethers (HO-PBDEs) determined by toxicogenomic analysis with a live cell array coupled with mutagenesis in
Escherichia coli. Environ. Sci. Technol. 48, 5929e5937.
Su, G., Zhang, X., Liu, H., Giesy, J.P., Lam, M.H., Lam, P.K., Siddiqui, M.A., Musarrat, J.,
Al-Khedhairy, A., Yu, H., 2012. Toxicogenomic mechanisms of 6-HO-BDE-47, 6MeO-BDE-47, and BDE-47 in E. coli. Environ. Sci. Technol. 46, 1185e1191.
Su, G., Zhang, X., Raine, J.C., Xing, L., Higley, E., Hecker, M., Giesy, J.P., Yu, H., 2013.
Mechanisms of toxicity of triphenyltin chloride (TPTC) determined by a live cell
reporter array. Environ. Sci. Pollut. Res. 20, 803e811.
Teo, T.L.L., Nald, J.A., Coleman, H.M., Khan, S.J., 2015. Analysis of organophosphate
flame retardants and plasticisers in water by isotope dilution gas
chromatography-electron ionisation tandem mass spectrometry. Talanta 143,
114e120.
van der Veen, I., de Boer, J., 2012. Phosphorus flame retardants: properties, production, environmental occurrence, toxicity and analysis. Chemosphere 88,
1119e1153.
Waring, J.F., Ciurlionis, R., Jolly, R.A., Heindel, M., Ulrich, R.G., 2001. Microarray
analysis of hepatotoxins in vitro reveals a correlation between gene expression
profiles and mechanisms of toxicity. Toxicol. Lett. 120, 359e368.
Yu, G., Wang, L.-G., Han, Y., He, Q.-Y., 2012. clusterProfiler: an R package for
comparing biological themes among gene clusters. Omics a J. Integr. Biol. 16,
284e287.
Zaslaver, A., Bren, A., Ronen, M., Itzkovitz, S., Kikoin, I., Shavit, S., Liebermeister, W.,
Surette, M.G., Alon, U., 2006. A comprehensive library of fluorescent transcriptional reporters for Escherichia coli. Nat. Methods 3, 623e628.
Zhang, X., Newsted, J.L., Hecker, M., Higley, E.B., Jones, P.D., Giesy, J.P., 2009. Classification of chemicals based on concentration-dependent toxicological data
using ToxClust. Environ. Sci. Technol. 43, 3926e3932.
Zhang, X., Wiseman, S., Yu, H., Liu, H., Giesy, J.P., Hecker, M., 2011. Assessing the
toxicity of naphthenic acids using a microbial genome wide live cell reporter
array system. Environ. Sci. Technol. 45, 1984e1991.