Identification of Escherichia coli biomarkers responsive to various
Siseon Leea, Dougu Nama, Joon Young Jungb, Min-Kyu Ohb, Byoung-In Sangc,⇑, Robert J. Mitchella,d,⇑
aSchool of Nano-Bioscience and Chemical Engineering, Ulsan National Institute of Science and Technology, 100 Banyeon-ri, Eonyang-eup, Ulsan 689-805, Republic of Korea
bDepartment of Chemical and Biological Engineering, Korea University, 5-1 Anam-Dong, Sungbuk-Gu, Seoul 136-713, Republic of Korea
cDepartment of Chemical Engineering, Hanyang University, 17 Hangdang-Dong, Seongdong-Ku, Seoul 133-791, Republic of Korea
dSchool of Urban and Environmental Engineering, Ulsan National Institute of Science and Technology, 100 Banyeon-ri, Eonyang-eup, Ulsan 689-805, Republic of Korea
a r t i c l ei n f o
Received 11 January 2012
Received in revised form 16 February 2012
Accepted 17 February 2012
Available online 24 February 2012
a b s t r a c t
Aberrations in the growth and transcriptome of Escherichia coli str. BL21(DE3) were determined when
exposed to varying concentrations of ferulic acid (0.25–1 g/L), an aromatic carboxylic acid identified
within lignin-cellulose hydrolysate samples. The expression of several individual genes (aaeA, aaeB, inaA
and marA) was significantly induced, i.e., more than 4-fold, and thus these genes and the heat shock
response gene htpG were selected as biomarkers to monitor E. coli’s responses to five additional hydro-
lysate-related compounds, including vanillic acid, coumaric acid, 4-hydroxybenzoic acid, ferulaldehyde
and furfural. While all of the biomarkers showed dose-dependent responses to most of the compounds,
expression of aaeA and aaeB showed the greatest induction (5–30-fold) for all compounds tested except
furfural. Lastly, the marA, inaA and htpG genes all showed higher expression levels when the culture was
exposed to spruce hydrolysate samples, demonstrating the potential use of these genes as biomarkers.
? 2012 Elsevier Ltd. All rights reserved.
Lignocellulose offers many benefits in that it is a renewable
source of sugar for biofuel production, a non-food feedstock and
its use aids in reducing or even reversing the carbon emissions into
the atmosphere as compared to conventional petroleum products.
However, during hydrolysis, numerous compounds are formed
from the lignin portion, including various phenolic acids and alde-
hydes. The exact composition of these compounds varies based
upon both the method used for hydrolysis and the feedstock
sources. A recent review covered their negative and beneficial
activities of hydrolysate compounds within a variety of biological
systems (Lee et al., 2012), while another recent study found that
some of these compounds actually lead to higher biomethane
production levels as they can be broken down by the microbes
(Barakat et al., 2012). Nevertheless, tests have shown that these
compounds inhibit the activity and growth of other microbes
involved in ethanol and butanol fermentative processes when
present in even small amounts (0.1–1 g/L) (Ezeji and Blaschek,
2008; Mills et al., 2009).
For this reason, numerous methods have been studied to miti-
gate the toxicity due to the presence of these chemicals. Several
groups evaluated the treatment of the hydrolysate samples in or-
der to detoxify or remove the inhibitory compounds. For example,
Jönssen et al. used the laccase and peroxidase from Trametes versi-
color to remove monoaromatic phenolics from the hydrolysate
samples (Jonsson et al., 1998). Likewise, some groups used vacuum
evaporation to remove the volatile compounds (Rodrigues et al.,
2001) which include vanillin and acetic acid, while other groups
evaluated the use of activated charcoal (Mussatto and Roberto,
2001) and even lignin residues (Bjorklund et al., 2002) to absorb
and remove the soluble inhibitory compounds. The most common
method, however, has been to treat the hydrolysates with alkali
solutions and to over-lime them (Martinez et al., 2000). All of these
methods require a significant amount of energy, expensive materi-
als or include the production of waste materials that need to be
disposed of, thereby increasing the system complexity and limiting
their effectiveness for use on a larger scale.
As an alternative, other research groups have studied methods
to make the fermentative micro-organism more robust and resis-
tant to the hydrolysate compounds. One simple manner that has
been used is to culture the strains through a series of inoculums
where the subsequent media contained more of the toxin or
hydrolysate. Several groups used this method to improve the resis-
tance of the yeast Candida guilliermondii (Sene et al., 2001) and
Escherichia coli (Geddes et al., 2011) to a variety of hydrolysate
sources. With the use of transcriptomics, it is now possible to
0960-8524/$ - see front matter ? 2012 Elsevier Ltd. All rights reserved.
⇑Corresponding authors at: Department of Chemical Engineering, Hanyang
University, 17 Hangdang-Dong, Seongdong-Ku, Seoul 133-791, Republic of Korea.
Tel.: +82 2 2220 2328; fax: +82 2 2220 4716 (B.-I. Sang), School of Nano-Bioscience
and Chemical Engineering, Ulsan National Institute of Science and Technology, 100
Banyeon-ri, Eonyang-eup, Ulsan 689-805, Republic of Korea. Tel.: +82 52 217 2513;
fax: +82 52 217 2509 (R.J. Mitchell).
(B.-I. Sang), email@example.com
Bioresource Technology 114 (2012) 450–456
Contents lists available at SciVerse ScienceDirect
journal homepage: www.elsevier.com/locate/biortech
determine the effects of various toxins and evaluate their stresses
to the microorganism through gene expression analyses. Represen-
tatively, the Papoutsakis group has studied the butanol tolerance in
Clostridium acetobutylicum ATCC 824 through transcriptomic
analyses (Borden and Papoutsakis, 2007; Tomas et al., 2004) while
another group has determined the effects of furfural, a common
inhibitory compound found in lignin hydrolysates, on E. coli
cultures (Miller et al., 2009).
This study utilized similar analyses to determine the effects of
several phenolic acids and aldehydes found in hydrolysate samples
on the transcriptome of E. coli BL21 (DE3). E. coli was selected for
this study since it is a model organism for transcriptome studies
and is recognized as one of the most promising biocatalysts for
the production of biofuels (Dien et al., 2003). Recently, the Liao
group has developed various E. coli strains to produce a wide array
of biofuel compounds, including butanol (Atsumi et al., 2008),
2-methyl-1-butanol (Cann and Liao, 2008) and 3-methyl-1-buta-
nol (Connor et al., 2010), as well as others. Downstream application
of these strains will likely involve the use of lignocellulose and,
therefore, the effects of hydrolysate-related compounds on the
growth of these and similar E. coli strains is definitely a pertinent
issue. For this reason, this study assessed the effects hydrolysate-
related compounds have on the growth and gene expression
patterns of E. coli cultures with the eventual aim of identifying
biomarker genes that can be used to evaluate plant hydrolysates
for toxic or inhibitory compounds prior to their use as fermentative
2.1. Bacterial strains and growth
The bacterial strain used for all the experiments was E. coli BL21
(DE3) unless noted otherwise. The E. coli strain is a common lab
strain and was selected since this study was undertaken to identify
biomarker genes to be used to identify harmful conditions. Like-
wise, to minimize gene expression due to starvation, the media
used was rich, i.e., LB broth, and the optical density was mid-
To ensure comparability between the samples, this strain was
grown fresh from frozen stocks (20% glycerol/?80 ?C) by streaking
it out on Luria agar plates. After growth overnight at 37 ?C, one col-
ony was transferred into 4 ml of LB within a 15 ml conical tube and
grown at 37 ?C with shaking (250 rpm) at a slant to ensure good
aeration until the optical density (OD) was 0.8 at 600 nm. The cul-
ture was then sub-cultured (1:25) into 50 ml of fresh LB media
within a 250 ml flask and grown to an OD of 0.4, at which time it
was mixed with the test media with a ratio of 1:1 (v:v). The test
media was also LB but prepared with the hydrolysate compounds
added. The final concentrations tested were 0.25, 0.5, 1.0, 1.5 and
2 g/L. The toxicity of each compound was estimated by monitoring
the growth rate of the E. coli cultures using the OD. The biolumi-
nescence (BL) of E. coli BL21 (DE3) carrying plasmid pUCDK (Mitch-
ell et al., 2006) was used to determine the effects of several
compounds on the metabolic activity of the cells. For this, the cul-
ture samples (200 ll) were added to white 96-well plates (Greiner,
USA) and the BL was determined using a Multi-Plus GloMax sys-
tem (Promega, USA).
2.2. Media preparation
The compounds tested were ferulic acid, coumaric acid, 4-hy-
droxy-3-methoxycinnmaldehyde and furfural, all purchased from
Sigma–Aldrich (USA), as well as 4-hydroxybenzoic acid (Junsei,
Japan) and vanillic acid (Alfa Aesar, USA). Each of these was the free
acid form of the compound. For the growth inhibition tests, the LB
media was autoclaved with the test chemical added. Due to the
acidic nature of some of the compounds, each test media had ster-
ile dibasic potassium phosphate buffer (1 M) added to a final con-
centration of 20 mM. The resulting pH of all the test samples was
7.0 ± 0.3 (data not shown).
2.3. RNA purification and cDNA synthesis
The total cellular RNA was purified from 5 ml cultures. Initially,
the sample was mixed RNALater (Ambion) at a ratio of 1:1 (v:v) to
stabilize the RNA according to the manufacturer’s protocol. After
spinning down the cells (4200g, 15 min) and washing them once
with sterile phosphate buffered saline (pH 7.4), the RNA was puri-
fied using the ChargeSwitch?Total RNA purification kit from Invit-
The complete removal of genomic DNA within purified RNA
sample was confirmed using PCR with the RecA primer set (Table
1), which produces a 150 bp fragment. DNA-free samples were
quantified using a Nanodrop (Fisher Scientific, USA) and run on a
denaturing gel to check the RNA quality. A total of 1 lg was used
to prepare 20 ll of cDNA for RT-qPCR using the First-Strand cDNA
Synthesis kit from Fermentec according to the manufacturer’s
2.4. Microarray experiments
Microarray analyses were performed using a two-color array
system with dye swaps. A total of 5 lg of total RNA purified from
the samples exposed to ferulic acid was used to generate labeled
cDNA using the Indirect Labeled cDNA Synthesis kit from Invitro-
gen (USA) according to the manufacturer’s protocol, except that
the cDNA synthesis was allowed to proceed overnight at 42 ?C
since longer times led to higher cDNA yields according to the
The Cy3- or Cy5-labeled control and sample cDNAs were mixed
just prior to heating the sample for 5 min at 95 ?C, cooled on ice for
1 min and then hybridized to the E. coli K-12 V2 OciChip micro-
array slide (E-biogen, South Korea) overnight at 42 ?C for 16 h.
After hybridization, the slide was washed within in pre-warmed
(50 ?C) 1 ? SSC with 1% SDS followed by washes in 25 ?C 1 ? SSC
with 1% SDS, 1 ? SSC and then three washes in 0.1 ? SSC. After dry-
ing the slide by centrifugation (2000g, 3 min), it was scanned using
an ArrayWORX scanner (Applied Precision, USA). Each microarray
experiment was performed in quadruplicate using cDNA prepared
RT-qPCR primer sequences.
16 s For
16 s Rev
S. Lee et al./Bioresource Technology 114 (2012) 450–456
from four independently grown and exposed cultures and also
with two dye-swaps to ensure that differences in the labelling effi-
ciencies were taken into account.
After synchronizing the dye swapped data sets, a total of four
replicates were used to analyze the effects of two different concen-
trations (0.25 and 0.5 g/L) of ferulic acid. We performed two kinds
of analyses for differential expression patterns of genes: gene-set
(or pathway) level and individual gene level.
The gene-set level analysis (Mootha et al., 2003; Nam and Kim,
2008) identifies differential expression patterns of collection of
genes that share common functions or other biological features.
Gene-set level analysis is known to have high sensitivity and statis-
tical power, and thereby aims at identifying ‘subtle but coordi-
nated’ expression patterns by multiple genes. For this approach,
we used the JProGO web server (Scheer et al., 2006) which was
developed for the gene-set analysis of microbe expression data.
Among the options provided, we chose the Kolomogorov–Smirnov
test that is a threshold-free method, and applied the Benjamini–
Hochberg multiple testing with the q-value set to 0.25. The GO sets
identified are listed in Table 1 with their p-values.
For the individual gene analysis, we employed the rank product
method (Breitling et al., 2004) instead of the conventional t-test.
The rank product method is applicable for and is known to perform
well with a small number of samples (two or more replicates). The
R package, RankProd (Hong et al., 2006), was used to compute the
p-values and the proportion of false positives for each gene (Elec-
tronic Annex – Tables S1–S6). Only genes with one or no missing
values were considered, and a missing value, if any, was filled in
by using the average value of the other replicates.
2.5. Real-time quantitative PCR
Based upon the microarray results, several genes that were sig-
nificantly up or down-regulated in response to ferulic acid were se-
lected for further analysis using RT-qPCR. Primers were designed
so that the Tm of the primers were 58 ± 1 ?C and the amplicon
was approximately 150–200 bp in length. Each of the primer sets
used are listed in Table 2. Each primer was designed to include a
50additional sequence flap which was shown to improve the fluo-
rescent response and reliability of RT-qPCR experiments (Afonina
et al., 2007). Three cDNA samples were prepared for each test con-
dition using 1 lg of total RNA purified from independently exposed
cultures and the RevertAid First Strand cDNA Synthesis kit from
Fermentas Co. After preparing cDNA, 2 ll were used for the
RT-qPCR experiments and analyses.
Using the cDNA prepared above, 2 ll of each cDNA samples was
mixed with 48 ll of a Sybr-Green RT-qPCR Supermix (RBC Biosci-
ences) prepared with the primers. This sample was split between
two wells (25 ll each) of a 96-well PCR plate (Roche), giving dupli-
cates for each sample. Standards were also prepared using the
same protocol but with serial dilutions of E. coli BL21 (DE3) geno-
mic DNA. These were used within this experiment to determine
the amount of cDNA present within each sample well. The plate
was sealed using sealing tape and inserted into a LightCycler
(Roche) and run using with the following program: 95 ?C for
5 min followed by 45 cycles of 95 ?C for 10 s, 56 ?C for 20 s and
72 ?C for 20 s followed by a dissociation step of 95 ?C for 5 s,
65 ?C for 1 min and then a gradual increase to 97 ?C at a rate of
0.11 ?C/s. The data was transferred to Microsoft Excel for the anal-
yses while the data plots were all prepared using SigmaPlot v.11
before being exported as GIF files.
2.6. Exposure to a spruce hydrolysate sample
A pH-adjusted spruce hydrolysate sample was kindly provided
by Dr. Kim at Kwangoon University. The contents of the hydroly-
sate were determined using Folin Coultiers reagent with gallic acid
as the standard. Likewise, the sugar, furfural and hydroxymethyl-
furfural concentrations were all determined using HPLC and GC
analyses, respectively, as previously published. Due to the color
of the hydrolysate sample, the OD could not be reliably measured
and so only the RT-qPCR data is available and was performed using
the same protocol as listed above (Section 2.5) for the single com-
3. Results and discussion
3.1. Effects of ferulic acid on E. coli str. BL21 (DE3) growth
Ferulic acid (FA) was selected as the model chemical for this
study since it was previously shown to inhibit the growth of fer-
mentative bacterial strains at concentrations between 0.25 and
2 g/L (Ezeji and Blaschek, 2008; Ezeji et al., 2007). Growth tests
with E. coli BL21 (DE3) cultures found that low concentrations of
FA led to significantly lower growth rates in a dose-dependent
manner (Fig. 1).
To better understand the toxic mechanisms, the total cellular
RNA was purified from these cultures at set times (10, 30, 60 and
90 min) and used to prepare cDNA. Based on the structural similar-
ities between FA and salicylic acid, it was presumed that FA would
lead to a significant heat shock response, as was seen previously
with salicylic acid (Mitchell and Gu, 2006). Therefore, the relative
expression levels of two heat shock genes, grpE and clpB, were
determined for each time point. It was found that a 10 min expo-
sure gave the best response characteristics (Electronic Annex –
Fig. S1) and, consequently, all subsequent transcriptional analyses
were performed using mRNA extracted at this time.
3.2. Transcriptome effects of ferulic acid: gene-set level analysis
From tests with 0.25 g/l ferulic acid, two significant gene ontol-
ogy (GO) terms were identified whose members were down-regu-
lated (Table 2). Among them, the ‘phosphate transport’ term was
most significant such that four of the ten members were also found
in the list of significantly down-regulated genes (?4-fold or higher)
while seven were down-regulated 1.5-fold or more (Table 2). The
reason for this strong response from phosphate transport genes
likely is an artefact. In the test cultures, the membranes would
be disrupted slightly by the phenolic, allowing phosphate to
‘‘transport’’ into the cell through the disrupted portion before it
can stabilize, thereby raising the intracellular phosphate concen-
trations in the test cultures as compared to the control. Further
experimentation is needed to demonstrate this phenomenon.
‘Sulfate assimilation’ was the second down-regulated GO term.
Repression of sulfate assimilation is of interest since it was recently
reported that furfural inhibits growth of E. coli by inhibiting sulfur
assimilation, resulting in the induction of many related genes,
including the cysCND and cysHIJ operons (Miller et al., 2009). In
contrast, both of these operons were repressed by ferulic acid in
this study, albeit only slightly since none of the six members in
‘sulfate assimilation’ was included in the list of significantly
down-regulated genes (Table S1). As such, this term may represent
the ‘subtle but coordinated’ expression pattern of a set of genes
(Mootha et al., 2003; Nam and Kim, 2008).
Using the same method, we also analyzed the gene expression
patterns from an exposure to 0.5 g/l ferulic acid. In this case, sev-
eral significant GO terms were down-regulated, including ‘ion
transport’ and its two descendents ‘phosphate transport’ and ‘cat-
ion transport’, ‘translation’ as well as those related to cellular local-
ization. The only two terms that were up-regulated were ‘response
to unfolded proteins’ and ‘quinone binding’ (Table 2).
S. Lee et al./Bioresource Technology 114 (2012) 450–456
3.3. Transcriptome effects of ferulic acid: individual gene level analysis
About 20 genes showed a significant change in their expression
levels (?4-fold or greater) when E. coli was exposed to 0.5 g/l feru-
lic acid (Table 3). Although the changes in expression for several of
the genes was expected, including several heat shock and mem-
(Fig. S2), the greatest change was seen from a cluster of genes that
include aaeX, aaeA and aaeB. All three of these genes are located
within a single operon and were shown previously to be induced
by and involved in the export of 4-hydroxybenzoic acid out of
the cell (Van Dyk et al., 2004).
Another operon that was strongly induced was the marRAB op-
eron, which is under dual control of the MarR repressor and MarA
inducer proteins (Hachler et al., 1991). The MarR regulon was pre-
viously shown to be induced during an exposure to xenobiotics,
including salicylic acid (Cohen et al., 1993). The results in Table 3
show that nearly one-third of the 19 genes showing a significantly
higher expression level are part of the MarR regulon, including the
inaA gene, which encodes for a gene of unknown function but is
associated with changes in the cytoplasmic pH (Slonczewski
et al., 1987). This large percentage of genes suggests that this reg-
ulon may have a significant effect on the survival of E. coli when ex-
posed to hydrolysates and their individual compounds.
Based upon the microarray results, several genes that were up-
regulated were selected as biomarkers for this study, including the
highly induced efflux protein genes aaeA and aaeB and the MarR
regulon activator gene marA. In addition, inaA and htpG, a gene
encoding for the protein chaperone involved in the heat shock re-
sponse and protein folding, were also selected based upon the GO
term and individual gene results (Table 2).
3.4. Growth effects and RT-qPCR results with other hydrolysate
Subsequent growth tests were performed with a variety of other
lignin hydrolysate-related compounds (Fig. 1). All of the com-
pounds tested, except 4-hydroxbenzoic acid, led to a dose-depen-
dent decrease in the growth of the culture for the concentrations
tested. From the growth retardations obtained, the most potent
inhibitors were 4-hydroxy-3-methoxycinnamaldehyde (or conife-
ryl aldehyde), which is the aldehyde form of ferulic acid, and furfu-
ral. Whereas the activity of 4-hydroxy-3-methoxycinnamaldehyde
is not clear, the biological activity of furfural is well documented
and includes interactions with DNA, resulting in single strand
breaks (Khan and Hadi, 1993), and the inhibition of sulfur assimila-
tion (Miller et al., 2009).
It was found that this loss in growth was due to inhibition and
not loss of viability (Electronic Annex – Fig. S3). However, a signif-
icant amount of stress was experienced by the culture, as shown by
the loss in bioluminescence from a bioreporter strain, E. coli
BL21(DE3)/pUCDK, and this stress led to a lower viable cell number
with longer exposures. Since the bioluminescence is integrally re-
lated with the metabolic state of the cell through the ATP pool and
electron carriers (Meighen, 1991), it appears that the cells are sur-
viving but at a dramatic cost to their metabolism.
As with ferulic acid, the expression of the genes selected were
monitored after an exposure to each of the compounds. Fig. 2
shows that aaeA and aaeB were both strongly induced by all of
the aromatic compounds, including both acids and aldehydes,
and this was dose-dependent with higher concentrations tending
to result in a greater expression level. Likewise, the expression of
marA was induced by most of the aromatic compounds and also
in a dose-dependent manner. The only exception was 4-hydroxy-
benzoic acid, which only induced aaeA and aaeB. This lack of re-
compound being naturally produced with E. coli through the ubi-
quinone production pathway (Van Dyk et al., 2004). Accordingly,
when the 4-hydroxybenzoic acid concentration increases, expres-
sion of the genes encoding for the AaeA and AaeB efflux proteins,
which are involved in the transport of 4-hydroxybenzoic acid out
of the cell, is induced. The fact that this compound is a natural
metabolite and that it can be effectively transported out of the cell
even when added exogenously (Van Dyk et al., 2004) helps to ex-
plain the lack of both growth loss for the concentrations tested
and responses from the other genes being studied.
Vanillic acid was also tested in this study (Fig. 1) and the growth
results indicate that this compound is much more inhibitory than
acid islikely due tothis
GO terms showing significant expression changes when E. coli str BL21(DE3) was exposed to ferulic acid. NS denotes a non-significant change in the GO term genes.
GO term (GO ID)Direction
# Genes in the GO term
(% genes showing a 1.5-
fold or higher change)
Genes showing a 1.5 or higher fold change (log2 ratio) when E. coli str. BL21(DE3) was
exposed to 0.5 g/L ferulic acid. Values in parenthesis are median values from 4
Ion transport (BP,
179 (21%)PhoU (?2.32), PstB (?2.19), YadQ (2.12), PstC (?2.03), OmpF (?2.00), PstS (?1.50),
PhoB (?1.49), EntD (?1.32), AtpE (?1.31), FhuC (?1.26), PhoR (?1.23), FepA (?1.08),
MgtA (?1.03), FeoA (?1.03), CyoB (0.91), PstA (?0.89), B4103 (0.84), Ftn (0.77), FepB
(?0.75), FecA (?0.75), FecD (?0.74), PitA (?0.72), Tsx (?0.72), AtpH (?0.69), AtpD
(?0.67), AtpB (?0.67), PhoE (?0.66), FhuB (?0.66), FeoB (?0.65), YjbB (0.65), TrkG
(?0.64), FhuA (?0.63), EntA (?0.63), PhnH (?0.63), KdpB (?0.63), CysA (?0.62), MotA
(0.61), FepC (?0.59)
HtrB (1.53), HslV (1.37), HtpG (1.22), RpoH (1.17), IbpA (0.99), ClpB (0.97), DnaK (0.92),
HslU (0.79), HtrA (0.77), HslJ (0.74), DnaJ (0.69), YrfH (0.60)
Response to unfolded
Phosphate transport (BP,
Cation transport (BP,
10 (90%)PhoU (?2.32), PstB (?2.19), PstC (?2.03), PstS (?1.50), PhoB (?1.49), PhoR (?1.23),
PstA (?0.89), PitA (?0.72), YjbB (0.65)
EntD (?1.32), AtpE (?1.31), FhuC (?1.26), FepA (?1.08), MgtA (?1.03), FeoA (?1.03),
CyoB (0.91), Ftn (0.77), FepB (?0.75), FecA (?0.75), FecD (?0.74), AtpH (?0.69), AtpD
(?0.67), AtpB (?0.67), FhuB (?0.66), FeoB (?0.65), TrkG (?0.64), FhuA (?0.63), EntA
(?0.63), KdpB (?0.63), MotA (0.61), FepC (?0.59)
NuoE (0.80), NuoK (0.80)
Quinone binding (MF,
Sulfate assimilation (BP,
Cellular localization (BP,
Down 52 (13%)InfB (?1.29), AsnA (?1.06), YjeA (?0.99), Efp (?0.73), PrfC (?0.71), PheT (?0.64), PheS
CysI (?0.72), CysD (0.67), CysJ (?0.61), CysC (?0.51), CysH (?0.40), CysN (?0.24)
Down 14 (29%)FliR (?1.47), YidC (?0.73), SecF (?0.69), SecE (?0.66)
S. Lee et al./Bioresource Technology 114 (2012) 450–456
4-hydroxybenzoic acid. This is in contrast to a previous study with
Saccharomyces cerevisiae where addition of 4-hydroxybenzoic acid
at a concentration of 1 g/L led to a 25% loss in ethanol production
while a similar addition of vanillic acid had no effect (Ando et al.,
1986). In this study, all five of the biomarker genes showed a
two-fold or higher expression when E. coli was exposed to 0.5 g/L
vanillic acid. The results show that the inaA, htpG and marA genes
give a dose-dependent response from all of the compounds tested,
except 4-hydroxybenzoic acid, but were for the most part not in-
duced very strongly, i.e., about 5-fold, unless the conditions were
overtly stressful to the cultures.
Such a condition is seen with 4-hydroxy-3-methoxycinnamal-
dehyde for all of the concentrations tested. This compound, an
aldehyde form of ferulic acid, led to much lower growth rates even
when present at only 0.25 g/L (Fig. 1), and was clearly the most
inhibitory compound tested in this study. Intact mRNA samples
could not be purified from the culture after an exposure to 1 g/L
of this compound. Regardless, expression levels for all of the genes
selected in this study were significantly higher even with only
0.25 g/L, the lowest being 4.2-fold for inaA, while marA was the
most strongly induced at this dosage. Finally, coumaric acid, which
is similar to ferulic acid in structure, led to comparable response
Fig. 1. Growth characteristics of E. coli BL21 (DE3) when exposed to various hydrolysate-related compounds, showing the relative inhibitory activity of each. Chemical
concentrations tested: d, 0 g/L (Control); s, 0.25 g/L; ., 0.5 g/L; D, 1.0 g/L. Three independent growths were performed for analysis.
Genes showing a ?4-fold or greater change in their expression levels. The fold change is the median of four independent tests. All genes listed showed a significant change in their
expression levels for at least one of the ferulic acid concentrations tested. The p-values for all genes were <0.005. Those shown in bold are part of the MarA regulon.
Gene IDAnnotationProtein encodedFerulic acid conc.
0.25 g/L0.5 g/L
Phosphate transport system protein
Phosphate ABC transporter protein
Phosphate ABC transporter protein
Predicted ABC superfamily efflux transporter
EriC chloride ion ClC channel
Predicted acyl transferase
Predicted inner membrane protein
Protein involved in stress resistance and biofilm formation
MarR DNA-binding transcriptional repressor
MarA DNA-binding transcriptional dual regulator
Multiple antibiotic resistance protein
pH-Inducible protein involved in stress response
16S rRNA m3U1498 methyltransferase
AaeAB hydroxylated, aromatic carboxylic acid efflux transport system protein B
AaeAB hydroxylated, aromatic carboxylic acid efflux transport system protein A
S. Lee et al./Bioresource Technology 114 (2012) 450–456
levels and trends as ferulic acid for each gene. These results con-
firm that the biomarker genes identified, and particularly marA,
aaeA and aaeB, are strongly expressed when E. coli is exposed to
several different hydrolysate-related phenolics.
3.5. Furfural represses aaeA, aaeB and marA expression
Furfural is another by-product formed during the acid hydroly-
sis of lignocellulose which can negatively impact fermentation
studies and yields (Ezeji and Blaschek, 2008; Ezeji et al., 2007). In
contrast to the other compounds tested in this study, furfural is
not a phenolic but is produced during the hydrolysis of samples
containing 5-C sugars, such as xylose. Whereas all of the hydroly-
sate phenolics tested led to an induced expression of the aaeA, aaeB
and inaA genes, an exposure to furfural at a concentration of 0.25 g/
L led to a 2-fold lower expression level for each of these genes
(Fig. 2). Higher concentrations (i.e., 0.5 g/L and 1 g/l) both caused
an induced htpG and marA expression but were overtly stressful
to the culture and led to a significant loss in growth (Fig. 1), making
furfural the second most inhibitory compound tested in this study.
3.6. Spruce hydrolysate induces marA and htpG expression but not
aaeA or aaeB
As noted in the introduction, this study assessed the effects
hydrolysate-related compounds have on the growth and gene
expression patterns of E. coli cultures with the eventual aim of
identifying biomarker genes that can be used to evaluate for toxic
or inhibitory compounds with plant hydrolysates prior to their use
as fermentative media. Consequently, we were interested in test-
ing the activity of the five biomarkers when exposed to an actual
hydrolysate sample. The pure hydrolysate contained 38 g/L glu-
cose, 8.1 g/L xylose, 3.4 g/L galactose, 17.5 g/L mannose, 0.23 g/L
phenolics, 0.02 g/L hydroxymethylfurfural, 0.04 g/L furfural and
2.6 g/L acetic acid. It was mixed equally (v:v) with the E. coli str.
BL21(DE3) culture for this experiment and the RNA was extracted
as above. The relative gene expression levels obtained showed the
aaeA and aaeB genes were not significantly induced (1.8- and 1.5-
fold increased mRNA concentrations) but inaA, htpG and marA all
had higher expression levels, with a 2.8-, 6.9- and 12.9-fold in-
crease in their mRNA concentrations, respectively.
Larsson et al. (1999) reported that 4-hydroxybenzoic acid and
vanillin were major constituents in untreated spruce hydrolysates.
Whereas 4-hydroxybenzoic acid showed some of the mildest ef-
fects seen on the expression of aaeA and aaeB, they were still sig-
nificant (?10-fold), while additional studies performed in our lab
found vanillin to be a strong inducer (data not shown). It is inter-
esting, therefore, that the neither of these genes were induced by
the spruce hydrolysate. One possible explanation for this is the
phenolic concentration, i.e., 0.12 g/L in the diluted sample, which
was lower than that tested above. This concentration, however, is
still expected to elicit some response based upon Van Dyk’s results
(2004) and so may be a result of some unknown phenolics or pos-
sibly an antagonistic effect of the mixture. This is an area of re-
search that can be studied further using biosensors based upon
the biomarkers selected in this study. In contrast to aaeA and aaeB,
expression of marA, inaA and htpG were all highly induced, indicat-
ing that the sample does contain some inhibitory compounds that
cause a significant heat-shock response and induction of the MarA
regulon. Using this information, our group has recently constructed
a inaA::lux bioluminescent biosensor and used this strain to sense
for both these compounds and the hydrolysate (Lee and Mitchell,
2012), demonstrating the potential use of these genes as biomark-
ers to detect for harmful or inhibitory compounds present within
This study mapped the toxicogenomic responses from E. coli
when exposed to various lignin hydrolysate-related compounds,
including ferulic acid, coumaric acid and furfural. Using microarray
and RT-qPCR results, five biomarker genes were identified and
their expression levels characterized when exposed to the com-
pounds. Two genes, aaeA and aaeB, consistently showed some of
the highest inductions seen, while the others, marA, inaA and htpG,
each showed dose-dependent responses. The use of these genes as
biomarkers was demonstrated by exposing E. coli to an acid hydro-
lysate of spruce, with the marA, inaA and htpG genes all showing
higher expression levels.
Fig. 2. Expression of the selected genes within E. coli BL21 (DE3) based upon the hydrolysate-related compound and its concentration. The genes were selected from Table 2
and were analyzed after a 10-min exposure. Chemical concentrations tested: j, 0 g/L (Control);
, 0.25 g/L;, 0.5 g/L;, 1.0 g/L. Note the differences in the y-axis scales.
S. Lee et al./Bioresource Technology 114 (2012) 450–456
The authors would like to thank the National Research Founda-
tion (NRF) of Korea for the grants provided through the Ministry of
Education, Science and Technology (Nos. NRF-2009-C1AAA001-
2009-0093479) and the research fund of Hanyang University
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.biortech.2012.02.085.
Afonina, I., Ankoudinova, I., Mills, A., Lokhov, S., Huynh, P., Mahoney, W., 2007.
Primers with 50flaps improve real-time PCR. Biotechniques 43(6), 770–+.
Ando, S., Arai, I., Kiyoto, K., Hanai, S., 1986. Identification of aromatic monomers in
steam-exploded poplar and their influences on ethanol fermentation by
Saccharomyces-Cerevisiae. Journal of Fermentation Technology 64 (6), 567–570.
Atsumi, S., Cann, A.F., Connor, M.R., Shen, C.R., Smith, K.M., Brynildsen, M.P., Chou,
K.J., Hanai, T., Liao, J.C., 2008. Metabolic engineering of Escherichia coli for 1-
butanol production. Metabolic Engineering 10 (6), 305–311.
Barakat, A., Monlau, F., Steyer, J.P., Carrere, H., 2012. Effect of lignin-derived and
furan compounds found in lignocellulosic hydrolysates on biomethane
production. Bioresource Technology 104, 90–99.
Bjorklund, L., Larsson, S., Jonsson, L.J., Reimann, A., Nilvebrant, N.O., 2002. Treatment
with lignin residue – A novel method for detoxification of lignocellulose
hydrolysates. Applied Biochemistry and Biotechnology 98, 563–575.
Borden, J.R., Papoutsakis, E.T., 2007. Dynamics of genomic-library enrichment and
identification of solvent tolerance genes for Clostridium acetobutylicum.
Applied and Environmental Microbiology 73 (9), 3061–3068.
Breitling, R., Armengaud, P., Amtmann, A., Herzyk, P., 2004. Rank products: a simple,
yet powerful, new method to detect differentially regulated genes in replicated
microarray experiments. FEBS Letters 573 (1–3), 83–92.
Cann, A.F., Liao, J.C., 2008. Production of 2-methyl-1-butanol in engineered
Escherichia coli. Applied Microbiology and Biotechnology 81 (1), 89–98.
Cohen, S.P., Levy, S.B., Foulds, J., Rosner, J.L., 1993. Salicylate induction of antibiotic-
resistance in Escherichia coli – activation of the mar operon and a mar-
independent pathway. Journal of Bacteriology 175 (24), 7856–7862.
Connor, M.R., Cann, A.F., Liao, J.C., 2010. 3-Methyl-1-butanol production in
Escherichia coli: random mutagenesis and two-phase fermentation. Applied
Microbiology and Biotechnology 86 (4), 1155–1164.
Dien, B.S., Cotta, M.A., Jeffries, T.W., 2003. Bacteria engineered for fuel ethanol
production: current status. Applied Microbiology and Biotechnology 63 (3),
Ezeji, T., Blaschek, H.P., 2008. Fermentation of dried distillers’ grains and solubles
(DDGS) hydrolysates to solvents and value-added products by solventogenic
clostridia. Bioresource Technology 99 (12), 5232–5242.
Ezeji, T., Qureshi, N., Blaschek, H.P., 2007. Butanol production from agricultural
residues: Impact of degradation products on Clostridium beijerinckii growth
and butanol fermentation. Biotechnology and Bioengineering 97 (6), 1460–
Geddes, C.C., Mullinnix, M.T., Nieves, I.U., Peterson, J.J., Hoffman, R.W., York, S.W.,
Yomano, L.P., Miller, E.N., Shanmugam, K.T., Ingram, L.O., 2011. Simplified
process for ethanol production from sugarcane bagasse using hydrolysate-
resistant Escherichia coli strain MM160. Bioresource Technology 102 (3), 2702–
Hachler, H., Cohen, S.P., Levy, S.B., 1991. Mara, a regulated locus which controls
expression of chromosomal multiple antibiotic-resistance in Escherichia-coli.
Journal of Bacteriology 173 (17), 5532–5538.
Hong, F.X., Breitling, R., McEntee, C.W., Wittner, B.S., Nemhauser, J.L., Chory, J., 2006.
RankProd: a bioconductor package for detecting differentially expressed genes
in meta-analysis. Bioinformatics 22 (22), 2825–2827.
Jonsson, L.J., Palmqvist, E., Nilvebrant, N.O., Hahn-Hagerdal, B., 1998. Detoxification
of wood hydrolysates with laccase and peroxidase from the white-rot fungus
Trametes versicolor. Applied Microbiology and Biotechnology 49 (6), 691–697.
Khan, Q.A., Hadi, S.M., 1993. Effect of furfural on plasmid DNA. Biochemistry and
Molecular Biology International 29 (6), 1153–1160.
Larsson, S., Reimann, A., Nilvebrant, N.O., Jonsson, L.J., 1999. Comparison of different
methods for the detoxification of lignocellulose hydrolyzates of spruce. Applied
Biochemistry and Biotechnology 77–9, 91–103.
Lee, S., Mitchell, R.J., 2012. Detection of toxic lignin hydrolysate-related compounds
using an inaA: luxCDABE fusion strain. Journal of Biotechnology 157, 598–604.
Lee, S., Monnappa, A.K., Mitchell, R.J., 2012. Biological activities of lignin
hydrolysate-related compounds. BMB Reports, in press.
Martinez, A., Rodriguez, M.E., York, S.W., Preston, J.F., Ingram, L.O., 2000. Effects of
Ca(OH)(2) treatments (‘‘overliming’’) on the composition and toxicity of
bagasse hemicellulose hydrolysates. Biotechnology and Bioengineering 69 (5),
Microbiological Reviews 55 (1), 123–142.
Miller, E.N., Jarboe, L.R., Turner, P.C., Pharkya, P., Yomano, L.P., York, S.W., Nunn, D.,
Shanmugam, K.T., Ingram, L.O., 2009. Furfural inhibits growth by limiting sulfur
assimilation in ethanologenic Escherichia coli strain LY180. Applied and
Environmental Microbiology 75 (19), 6132–6141.
Mills, T.Y., Sandoval, N.R., Gill, R.T., 2009. Cellulosic hydrolysate toxicity and
tolerance mechanisms in Escherichia coli. Biotechnology for Biofuels, 2.
Mitchell, R.J., Gu, M.B., 2006. Characterization and optimization of two methods in
the immobilization of 12 bioluminescent strains. Biosensors and Bioelectronics
22 (2), 192–199.
Mitchell, R.J., Hong, H.N., Gu, M.B., 2006. Induction of kanamycin resistance gene of
plasmid pUCD615 by benzoic acid and phenols. Journal of Microbiology and
Biotechnology 16 (7), 1125–1131.
Mootha, V.K., Lindgren, C.M., Eriksson, K.F., Subramanian, A., Sihag, S., Lehar, J.,
Puigserver, P., Carlsson, E., Ridderstrale, M., Laurila, E., Houstis, N., Daly, M.J.,
Patterson, N., Mesirov, J.P., Golub, T.R., Tamayo, P., Spiegelman, B., Lander, E.S.,
Hirschhorn, J.N., Altshuler, D., Groop, L.C., 2003. PGC-1 alpha-responsive genes
involved in oxidative phosphorylation are coordinately downregulated in
human diabetes. Nature Genetics 34 (3), 267–273.
Mussatto, S.I., Roberto, I.C., 2001. Hydrolysate detoxification with activated charcoal
for xylitol production by Candida guilliermondii. Biotechnology Letters 23 (20),
Nam, D., Kim, S.Y., 2008. Gene-set approach for expression pattern analysis.
Briefings in Bioinformatics 9 (3), 189–197.
Rodrigues, R.C.L.B., Felipe, M.G.A., Silva, J.B.A.E., Vitolo, M., Gomez, P.V., 2001. The
influence of pH, temperature and hydrolyzate concentration on the removal of
volatile and nonvolatile compounds from sugarcane bagasse hemicellulosic
hydrolyzate treated withactivated
evaporation. Brazilian Journal of Chemical Engineering 18 (3), 299–311.
Scheer, M., Klawonn, F., Munch, R., Grote, A., Hiller, K., Choi, C., Koch, I., Schobert, M.,
Hartig, E., Klages, U., Jahn, D., 2006. JProGO: a novel tool for the functional
interpretation of prokaryotic microarray data using gene ontology information.
Nucleic Acids Research 34, W510–W515.
Sene, L., Converti, A., Zilli, M., Felipe, M.G.A., Silva, S.S., 2001. Metabolic study of the
hydrolysate. Applied Microbiology and Biotechnology 57 (5–6), 738–743.
Slonczewski, J.L., Gonzalez, T.N., Bartholomew, F.M., Holt, N.J., 1987. Mu D-directed
Lacz fusions regulated by low Ph in Escherichia coli. Journal of Bacteriology 169
Tomas, C.A., Beamish, J., Papoutsakis, E.T., 2004. Transcriptional analysis of butanol
stress and tolerance in Clostridium acetobutylicum. Journal of Bacteriology 186
Van Dyk, T.K., Templeton, L.J., Cantera, K.A., Sharpe, P.L., Sariaslani, F.S., 2004.
Characterization of the Escherichia coli AaeAB efflux pump: a metabolic relief
valve? Journal of Bacteriology 186 (21), 7196–7204.
S. Lee et al./Bioresource Technology 114 (2012) 450–456