INTRODUCTION
Septicemia, the presence of bacteria in the bloodstream, is among the leading causes
of death worldwide, and its incidence is on the rise (1). In the past decade, there
has been a rapid increase in the rates of hospitalization and mortality from severe
sepsis, mainly because of the escalation of antibiotic resistance (2). While most
bacteria are unable to endure the strong bactericidal effects of serum, several pathogens
have evolved mechanisms that enable them to subvert the host defense systems and successfully
survive in this hostile niche.
In order to survive and even proliferate in serum, bacteria must overcome two major
obstacles, the nutritional immunity and innate immunity of the host. Nutritional immunity
is the process by which nutrients are kept in various storage molecules that make
them unavailable to pathogens (3). Thus, an invading bacterium has to pass a metabolic
barrier to survive. Iron sequestration by the host is the best-studied case of nutritional
immunity. Although iron is an abundant nutrient in nature, serum contains very little
free iron because iron is bound to storage molecules such as ferritin and hemosiderin.
Therefore, it is not surprising that iron acquisition systems and receptors were found
to play a pivotal role in the virulence of numerous pathogens. For instance, TonB,
a protein that provides energy for the transport of iron compounds, was found to be
required for the pathogenicity of several Gram-negative bacteria (4
–
8). An additional factor that was shown to be important for growth in serum is the
ability to synthesize nucleotides, as the inactivation of nucleotide biosynthesis
genes was shown to hamper the growth of Escherichia coli, Salmonella enterica serovar
Typhimurium, and Bacillus anthracis in human serum (9).
The second barrier that bacteria need to overcome to establish sustained bacteremia
is the one drawn by the immune system of the host. The complement system serves as
the first line of defense against invading bacteria and acts on the outer membrane.
In Gram-negative bacteria, the complement complex mediates direct killing by the formation
of pores in the cell membrane. To avoid the highly bactericidal effect of serum, pathogens
evolved structural features that inhibit complement-dependent killing. Many of these
adaptations are in surface-exposed components, such as the outer membrane lipopolysaccharide
(LPS) and the bacterial capsule (10). It was previously shown that variation in the
length of the O antigen, as well as its type (E. coli for instance, has more than
190 types of O antigens), influences the level of serum resistance (11). Even a single
nucleotide change that consequently results in a truncated gene coding for O antigen,
is sufficient to turn a serum-resistant strain to a serum-sensitive one (12). However,
not all of the adaptations are structural; some bacteria can secrete complement-binding
proteins and specific complement inhibitors. Staphylococcus aureus, for instance,
employs an arsenal of proteins to interfere with and block the activation of the innate
immune response (13). Successful evasion of the innate immune system, coupled with
proper metabolic adaptations, enables pathogenic bacteria to persist and multiply
in the bloodstream, a process that often leads to sepsis and, if not treated successfully,
death.
Among the leading causative agents of sepsis are extraintestinal pathogenic E. coli
(ExPEC) strains, which bring about high morbidity and mortality rates. E. coli infections
also result in a heavy economic burden. ExPEC strains are characterized by the ability
to resist the bactericidal effect of human serum, and this study was aimed at understanding
the functional genomic basis of this resistance. To this end, we studied E. coli O78:H19
sequence type 88 (ST88) isolate 789 (O78-9), an ExPEC isolate able to grow, and even
multiply, in serum. We performed comprehensive system-wide analyses of both its transcriptome
and its proteome in response to serum. To distinguish between genes that respond to
the metabolic challenge presented by serum and those that respond to the bactericidal
effect of the complement system, we compared the effect of serum depleted of immune
complement by heat treatment (inactive serum) with that of active serum. Here we show
that the bacterial response to heat-inactivated serum included a change in the expression
of a vast spectrum of known virulence determinants and iron acquisition genes, as
well as transcriptional regulators. In contrast, the distinctive response to active
serum (complement system) involved a much smaller group of genes, some of which are
involved in the response to pH changes and others of which have unknown functions.
One clear conclusion that emerged from these studies is that the key regulator of
the response to serum—active and inactive—is the iron regulator Fur, which controls
the expression of more than 80% of the serum-upregulated genes.
RESULTS AND DISCUSSION
ExPEC growth in serum.
Septicemic bacteria are relatively resistant to the bactericidal effect of serum.
We studied a septicemic strain, E. coli serotype O78:H19 ST88 isolate 789 (O78-9),
that is capable of growing in serum. The results presented in Fig. 1A show the growth
of E. coli O78-9 in 40% serum in comparison with that of a nonpathogenic E. coli strain
(K-12) that undergoes lysis under the same conditions (Fig. 1B). The serum component
significant in growth inhibition is the immune complement, as its inactivation by
heat (inactive serum, 40 min at 56°C) enables the growth of E. coli K-12 (Fig. 1B,
inset).
FIG 1
ExPEC O78-9 is serum resistant. Cultures of E. coli strains O78-9 (A) and K-12 (B)
were grown overnight and diluted in MOPS medium or in MOPS medium containing 40% serum,
and growth curves were generated by measuring OD600 with the BIO-TEK Eon platform.
Viability was estimated by dropping 10 µl of a culture treated with serum for 60 min
on LB medium plates and incubating them overnight at 37°C (insets). S, serum; IS,
inactive serum.
Transcriptome analyses of ExPEC genes upon exposure to serum.
As pointed out, the response to serum is driven by two major stress conditions—nutritional
stress and the bactericidal effect of complement. To study the nutritional stress,
we investigated the response of ExPEC O78-9 to inactive serum. Because serum inactivation
is achieved by mild heat conditions that destroy the complement system without affecting
other properties such as pH and nutrient availability, the response to inactive serum
represents the response to nutritional immunity. The experiments were carried out
with cultures grown in minimal, salt-glucose, media. In these media, all of the biosynthetic
genes are active and it is possible to study metabolic stress.
To obtain transcriptional expression data for the complete genome, we used the Illumina
transcriptome sequencing (RNA-seq) platform. RNA was obtained from biological replicates
of bacteria grown in minimal morpholinepropanesulfonic acid (MOPS) medium supplemented
with glucose and with or without exposure to 40% inactive serum. Total RNA was extracted
and rRNA depleted (as described in Materials and Methods). Two independent biological
replicates and deep sequencing analyses were performed, and the data were analyzed
by using strict criteria, as described in Materials and Methods. We examined only
genes that were expressed the same way in the two independent biological experiments
by using a 2-fold cutoff. The expression data of selected genes were confirmed by
quantitative reverse transcription-PCR.
To obtain a comprehensive overview of the transcriptional response to serum, we compiled
Voronoi tree maps (14, 15) from the quantitative transcriptomic data. In these tree
maps, the transcripts are clustered according to their functions as obtained from
the TIGR classification. Thus, functionally related elements are localized in close
proximity to each other. The categories can then be subdivided sequentially down to
the level of a single gene. A general representation of all functional pathways is
shown in Fig. 2, upper panel. The log2 ratios (i.e., treated/control) of expression
data were color coded by using a divergent color gradient (from blue for downregulated
to red for upregulated). The Voronoi tree maps of the transcriptome induced by inactive
serum and the pathways derived from it are presented in Fig. 2, lower panel, left
side, which shows that serum affects the expression of a broad spectrum of genes with
a variety of functions. Most of the serum-induced changes are metabolic; the upregulated
pathways were mainly those of cations, iron-carrying compounds, amino acids, and amine
biosynthesis, especially in the aspartate family. In contrast, pyruvate, glutamate,
and aromatic amino acid biosynthesis pathways were downregulated. Other significantly
downregulated genes included those for sulfur metabolism, the trichloroacetic acid
cycle, and metabolism of polysaccharides. Most striking is the downregulation of biosynthetic
genes involving the synthesis of several amino acids, mainly cysteine. The downregulated
genes (Table 1) appeared to be affected directly by two major transcription factors,
CysB and BasR. The cysB gene product downregulates 11 genes involved in cysteine metabolism
in addition to that for the transcription factor Cbl, which is responsible for sulfate
starvation. The downregulated genes also include the six members of the sugar nucleoside
metabolism arn operon (16
–
18) (arnB, arnC. arnA, arnD, arnT, arnE), which are under the control of the BasRS
two-component system. The latter consists of two proteins that are involved in iron
metabolism and protect against high concentrations of external iron (19, 20). Generally,
it appears that the downregulated genes encode mainly metabolic enzymes (aminotransferases,
glycosyltransferases, and formyltransferases).
FIG 2
Tree map of differentially expressed genes in ExPEC O78-9 based on transcriptome analysis.
Functional Voronoi tree maps of E. coli strain O78-9 were obtained from transcriptomes
of bacteria exposed to serum for 15 min. Each cell represents one quantified transcript,
and transcripts encoding functionally related proteins are subsumed in convex-shaped
parental categories of increasing levels based on TIGR gene classification. The ratios
of expression data were color coded by using a divergent color gradient. The bluish
and orange tiles represent down- and upregulated pathways, respectively. Top panel:
General representation of all functional pathways, levels 1 to 4 (left to right).
Lower panel, left side: Transcription patterns in response to nutritional challenge.
Lower panel, right side: Immune complement-induced gene fraction.
TABLE 1
Genes downregulated (at least 2-fold) in response to inactive serum
Operon
Gene(s) differentially expressed
Function(s)
Regulator(s)
cysPUWAM
cysP, cysU, cysW, cysA
Sulfate/thiosulfate transporter
CysB, H-NS
cysDNC
cysD, cysN, cysC
Sulfate adenylyltransferase and adenylyl-sulfate kinase
CysB
cysJIH
cysJ, cysI, cysH
Sulfite and 3′-phosphoadenylylsulfate reductases
CysB, IHF
cysK
cysK
Cysteine synthase A
CysB
cbl
cbl
CysB-like transcriptional activator
CysB, NtrC
arnBCADTEF
arnB, arnC, arnA, arnD, arnT, arnE
Modification of LPS
BasR
yibD
yibD
Predicted glycosyltransferase
BasR, NsrR, PhoB
glnK-amtB
glnK
Nitrogen-regulatory protein
Fur, NtrC, FNR, GadX
ftnA
ftnA
Cytoplasmic ferritin iron storage protein
Fur, H-NS
basS-basR
bass, basR
Response to excess external iron
ais
ais
Protein induced by aluminum
yjdB
yjdB
Metal-dependent hydrolase
ugd
ugd
UDP-glucose 6-dehydrogenase
yeeED
yeeE, yeeD
Unknown
yedEF
yedF
Unknown
ydjN
ydjN
Unknown
ydjO
ydjO
Unknown
The metabolic changes in response to serum are characterized by a massive upregulation
of operons attributed to various iron receptors, binding proteins, transporters, and
siderophore genes (Table 2). We show that the vast majority of these genes are involved
in the maintenance of iron equilibrium between the bacterium and its surroundings.
Interestingly, many of the upregulated genes have basic metabolic functions. For instance,
the entire operons encoding iron hydroxamate (fhuACDB), iron dicitrate (fecABCDE),
FeS cluster assembly protein (sufABCDSE), an electron transport system (nrdHIEF),
and enterobactin (fepA-entD) formed the scaffold of the ExPEC O78-9 iron response
to serum. Notably, these operons are also found in commensal E. coli strains such
as K-12.
TABLE 2
Genes upregulated (at least 2-fold) in response to inactive serum
Operon
Gene(s) differentially expressed
Function(s)
Regulator(s)
fhuACDB
fhuA, fhuC, fhuD, fhuB
Iron-hydroxamate transporter
Fur
fepA-entD
fepA, entD
Enterobactin
Fur, CRP
fes-ybdZ-entF-fepE
fes, ybdZ, entF
Enterobactin
Fur, CRP
fepDGC
fepG, fepC
Enterobactin transporter
Fur
entCEBAH
entC, entE, entB, entA, entH
Enterobactin
Fur, CRP
sufABCDSE
sufA, sufB, sufC, sufD, sufS
FeS cluster assembly protein
Fur, IHF, IscR, NsrR, OxyR
nrdHIEF
nrdH, nrdI, nrdE, nrdF
Electron transport system
Fur, IscR, NrdR
fecABCDE
fecA, fecB, fecC, fecD, fecE
Iron-dicitrate transporter
Fur, CRP, PdhR, YjiE
fecIR
fecI, fecR
Signal transducer, sigma 19
Fur
exbBD
exbB, exbD
Energy transduction system
Fur
feoABC
feoB
Ferrous iron transporter
Fur, NagC, FNR
fhuE
fhuE
Ferric-rhodotorulic acid transporter
Fur
fhuF
fhuF
Ferric iron reductase
Fur, OxyR
entS
entS
Enterobactin exporter
Fur
fepB
fepB
Enterobactin transporter
Fur, RutR
ybiX
ybiX
Fe(II)-dependent oxygenase
Fur
Fiu
Fiu
Catecholate siderophore receptor
Fur
bfd-bfr
bfd
Bacterioferritin, iron storage, detoxification
Fur
a
yncE
yncE
Unknown
Fur,
a
MarA
ybtA
ybtA
Yersiniabactin transcriptional regulator
Fur
a
iron
iroN
Outer membrane siderophore receptor
Fur
a
yddAB
yddA, yddB
Multidrug transporter
Fur
a
(yddA)
efeU
efeU
Ferrous ion transporter
Fur,
a
CpxR
efeOB
efeO, efeB
Ferrous ion transporter
yjjQ-bglJ
bglJ
DNA-binding transcriptional activator
H-NS, LeuO
hokE
hokE
Toxic polypeptide
LexA
ybiI
ybiI
Unknown
yrbL
yrbL
Unknown
BasR, PhoP
pqqL
pqqL
Predicted peptidase
78900613
78900613
Ferric enterochelin esterase
Fur
a
78901042
78901042
Phage-like element
78903310
78903310
Unknown
78904620
78904620
Unknown
a
Fur binding box computationally detected in promoter region.
Most interesting is the upshift in the expression of bglJ, a component of the yjjQ-bglJ
transcription unit. These genes detoxify the α-oxoaldehyde methylglyoxal (MG), an
intracellular metabolism product; overproduction of MG results in cell death. The
gene products contain the helix-turn-helix motif typical of the LuxR transcript regulator
family (21) and are therefore suspected of being transcription regulators. It was
previously suggested that the yjjQ gene of E. coli serotype O2:H5 (22) and S. enterica
serovar Typhimurium (23) may be important for virulence in long-term systemic infections
of mice.
Another upregulated gene is hokE, a homologue of hok (host killing) that is part of
the hok-sok toxin-antitoxin system, an associated gene member of the SOS regulon (24).
Five additional genes that are upregulated in the presence of serum have unknown functions
(yrbL pqqL, 78901042, 78903310, and 78904620).
Complement-dependent transcriptomic response.
The genes responding to the bactericidal effect of serum could be identified by exposure
to active serum and comparison of the transcriptome data to those obtained in the
presence of inactive serum. This comparison revealed a small set of genes that were
induced only in the presence of active serum—probably by the presence of complement.
In contrast to the metabolic stress (Fig. 2, lower panel, left side), the complement-induced
gene fraction is characterized by upregulation of the serine family, sulfur metabolism,
and pyruvate and glutamate biosynthesis (Fig. 2, lower panel, right side). The complement-induced
response consists of genes involved in iron metabolism (Table 3)—the ferric enterobactin
transport system gene fepE and the iron storage bacterioferritin gene bfr. The presence
of complement also induces galP, a gene that codes for a transporter, a member of
the major facilitator superfamily of transporters. The rest of the upregulated genes
have unknown functions and code for the putative proteins YcfJ (25, 26), YpfG (27),
and YgaC (28, 29). Exposure to complement repressed the transcription of the enzyme-encoding
genes yfaO, sodB, sorC, and fimA in addition to one conserved hypothetical protein
with high homology to iraM. The observation that FimA, the major E. coli fimbrial
subunit protein, is downregulated in response to serum is in line with previous findings
suggesting that under low-iron conditions, E. coli decreases the expression of type
I fimbriae (30, 31). The iraM gene is induced under magnesium starvation and serves
as a stabilizer of RpoS, the sigma factor of the general stress response (32, 33).
The sodB gene, which encodes the Fe-dependent superoxide dismutase FeSOD, is also
downregulated in the presence of serum and is connected with iron metabolism (34).
TABLE 3
Complement-dependent genes
Operon
Genes differentially expressed
Function(s)
Response to serum
b
Regulation
fes-ybdZ-entF-fepE
fepE
Enterobactin
Up
Fur, CRP
ygaC
ygaC
Response to cytoplasmic pH stress
Up
Fur
ypfG
ypfG
Unknown
Up
Fur
a
ycfJ
ycfJ
Biofilm formation
Up
bfd-bfr
bfr
Bacterioferritin, iron storage, detoxification
Up
galP
galP
Galactose-proton symport of transport system
Up
CRP, GalR, GalS, NagC
78901228
78901228
Enhancement of lycopene biosynthesis protein 1
Down
ArcA, FNR, PhoP
fimAICDFGH
fimA
Major type 1 fimbrial subunit
Down
H-NS, IHF, Lrp
sodB
sodB
Superoxide dismutase
Down
NsrR, CRP, IHF
yfaO
yfaO
Putative nudix hydrolase YfaO
Down
sorC
sorC
Transcriptional regulator of sorbitol uptake and utilization
Down
a
Fur binding box computationally detected in promoter region.
b
Up, upregulation. Down, downregulation.
It should be noted that in addition to the genes whose expression changes in response
to complement, there are genes for other systems essential for complement survival,
such as LPSs or capsule. These genes were studied in recent papers (10, 35) that looked
at the effect of serum on an E. coli strain involved in urinary tract infections.
Clearly, these surface components are essential for pathogenesis and survival in the
bloodstream. Some of these genes are constitutive, and therefore, their expression
is not affected by the presence of serum. However, the genes involved in the biosynthesis
of the exopolysaccharide colanic acid are induced by serum (35) but were not detected
in our system. These genes are under the control of the Rcs regulator, whose level
is greatly affected by the growth medium (36). Thus, the experiments of Miajlovic
et al. were conducted with cultures grown in Luria-Bertani (LB) medium, where the
expression levels of the Rcs-regulated genes are low. Our experiments were carried
out with cultures grown in minimal medium, where the Rcs-regulated genes are already
induced and the addition of serum probably does not result in significant further
induction.
Proteome analyses of the serum response.
The effect of serum on gene expression was determined at the proteome level by using
mass spectrometry (MS)-based proteomics as described in Materials and Methods. As
in the transcriptome experiments, the proteins were extracted from cultures exposed
to active or inactive serum and an untreated reference culture. A total of 1,037 proteins
were identified (see Appendix SA in the supplemental material), and many of them could
be detected only under one condition—in the presence of serum or in the untreated
culture. Two hundred proteins were significantly differentially expressed, at least
2-fold, in the presence of serum (113 were upregulated, and 87 were downregulated).
Of these proteins, 155 were also differentially expressed in response to inactive
serum (70 were upregulated, and 85 were downregulated). The proteins that are present
at very low levels in the presence of serum are probably encoded by genes that were
significantly repressed in the presence of serum. However, it is possible that the
level of some of them goes down in the presence of serum because of degradation. In
an effort to achieve a global view of the functions altered in response to innate
or nutritional immunity, we mapped all of the proteins to their corresponding pathways
on the basis of the Voronoi tree map representation (Fig. 3, top panel).
FIG 3
Tree map of differentially expressed genes in ExPEC O78-9 based on proteome analysis.
The tree map shown is as described in the legend to Fig. 2 but is based on proteomic
data.
The group of proteins that were induced by serum included a number of conserved hypothetical
proteins. However, many of the proteins that were present in very low concentrations
without serum and induced by serum (inactive and active) were involved in iron acquisition.
This finding is not surprising, as serum is devoid of free iron because of the presence
of human iron-binding proteins. There were also proteins that could not be detected
once serum (active or inactive) was added. This group includes proteins that participate
in various biosynthesis pathways, the arn operon (arnC, arnA, and arnD [16
–
18]), transcription regulatory factor BasR (19, 20), and conserved hypothetical proteins.
A comparison of the genes differentially expressed in the presence of active serum
versus inactive serum is shown in the bottom part of Fig. 3, right side. This group
contained only a few proteins, as most of the responding proteins behave the same
in active and inactive serum. The unique group of complement-induced proteins contained
proteins involved in energy metabolism, amino acid synthesis, and the biosynthesis
of cofactors, purines pyrimidine, and nucleosides. This group also contained proteins
involved in cell envelope maintenance (MurD, MurE, MreB, GlmU, Alr, GalU). The latter
observation is in line with the knowledge that the innate immune system causes membrane
stress and that the integrity of the bacterial cell envelope is crucial for survival
in serum and circumvention of the host’s membrane attack complex (37
–
39).
For the complete list of proteins whose expression depends on the presence of serum,
see Appendix SA in the supplemental material.
Comparison of mRNA levels and protein abundance.
An analysis of the mRNA levels and protein concentrations in response to exposure
to inactive and activated serum (Fig. 4A and B, respectively) shows that under both
conditions, there is a linear relationship between RNA levels and protein abundance
(active serum, R
2 = 0.3 and Spearman’s rank correlation coefficient = 0.45; inactivated serum, R
2 = 0.25 and Spearman’s rank correlation coefficient = 0.47). These results are consistent
with previous observations that mRNA levels explain one- to two-thirds of the variance
in protein levels (40).
FIG 4
Comparative analysis of RNA and protein profiles of E. coli O78-9 in response to human
serum. The results show the correlation between changes in protein (log2 fmol/ng ratio)
and transcript (log2 RPKM ratio) levels in bacteria grown in 40% serum over bacteria
grown in MOPS minimal medium. A 2-fold change between the conditions was used as the
cutoff for up- and downregulation of RNA and protein levels. (A) After exposure to
inactive serum. (B) After exposure to active serum.
Fur is the major regulator of the serum response.
In an attempt to discover the regulatory network underlying serum adaptation, we systematically
analyzed which transcription factor controls each of the differentially expressed
genes involved in ExPEC’s serum response. To this end, we used a literature survey
and RegulonDB, a database of the regulatory interactions of E. coli K-12 (41). As
shown in Table 2, in accordance with the proteome results, Fur appears to be the key
regulator of serum adaptation, controlling more than 70% of all upregulated genes.
Fur, the ferric uptake regulator, serves as the main iron regulator in a wide range
of bacterial species (29, 42
–
51). Fur was found to play a role in the virulence of many bacteria, including Staphylococcus
aureus, Vibrio cholerae, Neisseria species, Yersinia species, and E. coli (52, 53).
Fur repression occurs under iron-depleted conditions and is achieved by binding of
Fur dimers to Fur boxes of target promoters. In the absence of free iron, the repression
is removed and the genes are transcribed, as occurs in serum, an iron-limiting environment.
While RegulonDB provides comprehensive data on transcription factor-gene interactions
in E. coli K-12, it cannot detect new regulatory interactions that appear in other
strains. To overcome this obstacle, we scanned the region upstream of each of the
serum-responsive genes to identify the presence of the 19-bp Fur binding consensus
box (Fig. 5, upper part) by using weight matrices taken from PRODORIC (54). The Fur
binding box is composed of three adjacent hexamers that bind dimeric Fur (55). Fur
binding boxes were detected in seven additional genes that were upregulated in response
to inactive serum, efeU, iroN, yncE, ybtA, bfd, yddA, and the ferric enterochelin
esterase-encoding gene 78900613 (Fig. 5, lower part), and in one gene, ypfG, that
was induced only in the presence of complement. As expected, most of these genes are
attributed to iron transporters, acquisition systems, and storage. It may be significant
that two out of the eight genes found to have a Fur box, iroN and ybtA, were acquired
by lateral gene transfer from other pathogens (S. enterica and Yersinia pestis, respectively)
(56, 57). Thus, all in all, Fur is predicted to control more than 80% of the genes
involved in the adaptation of E. coli to serum (Fig. 6 and Table 2).
FIG 5
The Fur binding box motif and predicted E. coli O78-9 genes with Fur binding boxes
in their promoter regions. The Fur binding box 19-bp consensus sequence was created
in WebLogo (upper part). Eight O78-9 gene sequences containing a Fur binding box (highlighted)
that were computationally derived from the PRODORIC database are shown at the bottom.
FIG 6
The majority of the genes upregulated in response to serum are controlled by Fur.
Genes upregulated following exposure to inactive serum that contain a Fur binding
box (blue) and genes upregulated under another transcription factor (red) are shown.
Asterisks indicate genes predicted by this study to have a Fur binding box.
Fur is essential for survival in serum.
As Fur is clearly a major regulator of growth in serum, we constructed a deletion
of the fur gene to study its effect on survival in serum. The results presented in
Fig. 7 indicate that the removal of Fur had no effect on growth in minimal medium.
In contrast, the deletion-containing bacteria grew very poorly in the presence of
serum. Growth in serum was restored by complementation with the fur gene on a plasmid.
Although there was an induction of many proteins involved in iron metabolism, the
concentration of Fur was unchanged upon exposure to serum. These results further support
the notion that the effects of Fur are not solely concentration dependent and there
must be additional factors involved, such as iron-dependent changes in its activity
(58, 59).
FIG 7
Fur is essential for growth of E. coli O78-9 in serum. Bacterial cultures were grown
as described in the legend to Fig. 1. Strain 078-9 Δfur grew in MOPS medium (continuous
line) or in the presence of 40% serum (dotted line). Growth arrest due to serum exposure
was restored in strain 078-9 Δfur/pBADfur (dashed line). Viability was estimated as
described in the legend to Fig. 1, and the results are shown on the right. S, serum;
IS, inactive serum.
Our results reflect the complex role of Fur in bacterial metabolism. As Fur is the
regulator of iron metabolism, its removal would disrupt the delicate balance of the
function involved in this important system. Actually, one would expect that a fur
deletion would have a significant effect under all growth conditions. Yet, the bacteria
appear to be able to control growth in the absence of Fur but are not capable of doing
so under extreme iron concentration changes, such as in the presence of serum. The
effect of fur deletion is not restricted to E. coli, as it was also previously shown
in Neisseria meningitidis, deletion mutants of which acquired serum sensitivity (60).
Conclusions.
Here we present data, obtained from proteomic and transcriptomic experiments, on the
global effect of exposure to serum on septicemic E. coli. We focused on genes required
to overcome the nutritional immunity of serum and therefore conducted our studies
with minimal media, in which the metabolic genes are active. Under such conditions,
it is possible to see the serum-induced metabolic changes and the regulatory networks
that control them. The most striking finding is the identification of Fur as the major
regulator during serum exposure. Thus, >80% of the upregulated genes are under the
control of Fur, the major regulator of iron metabolism (61, 62).
Genes that encode virulence factors are essential for pathogenesis. However, recent
studies indicate that bacterial virulence is highly dependent on the regulation of
metabolic genes (63
–
66), which are no less important than genes that encode virulence factors. The results
presented here strongly support this notion and indicate that the major changes in
bacterial gene expression upon exposure to serum involve metabolic genes, and these
changes are essential for survival in serum.
MATERIALS AND METHODS
Bacterial strains and growth media.
All of the E. coli strain and plasmids used in this study are listed in Table 4. Unless
stated otherwise, all E. coli strains were grown while shaking at 30°C in minimal
defined MOPS medium (with 0.2% glucose added). When required, antibiotics were added
to the medium (ampicillin at 300 µg/ml, kanamycin at 50 µg/ml, chloramphenicol at
20 µg/ml). To determine growth curves, log-phase cultures were diluted to an optical
density at 600 nm (OD600) of 0.04 as determined with a BioTek Eon plate reader, and
turbidity at 600 nm was measured every 20 min. Viable counts were determined after
1 h of incubation with serum or inactive serum by plating 10-µl drops of increasing
dilutions on LB medium plates and incubating them overnight.
TABLE 4
Plasmids and strains used in this study
Plasmid or strain
Description
Reference
Plasmids
pKD4
Template for kanamycin resistance cassette
76
pBAD24
Ampr, arabinose-inducible plasmid
69
pBADfur
Ampr, arabinose-inducible plasmid containing fur ORF
This study
E. coli strains
MG1655
Wild-type K-12 strain
078-9
Wild-type ExPEC 078 strain isolated from a turkey with sepsis
078-9 Δfur
This study
078-9 Δfur/pBADfur
This study
RNA isolation and purification.
Overnight cultures were diluted to an OD600 of 0.04 and allowed to grow to an OD600
of 0.3; cells were then introduced to serum, inactive serum, or saline to a final
volume of 40%. Cells were pelleted by 10 min of centrifugation at 3,000 rpm at 4°C.
Pellets were resuspended in 6 µl of freshly prepared lysozyme solution (1 mg/ml) in
10 mM Tris-HCl, pH 8.0, and incubated at 37°C for 7 min with occasional mixing. One
milliliter of TriReagent was added to each sample, and the homogenate was stored at
room temperature for 5 min. A 200-µl volume of chloroform was added for 15 s of incubation,
and the resulting mixture was stored at room temperature for 2 to 15 min and centrifuged
at 12,000 rpm for 15 min at 4°C. RNA was precipitated from the aqueous phase by mixing
with isopropanol (0.5 ml/ml of the TriReagent used for the initial homogenization),
stored at room temperature for 5 to 10 min, and centrifuged at 12,000 rpm for 8 min
at 4 to 25°C. The RNA pellet was washed with 75% ethanol, centrifuged at 7,500 rpm
for 5 min, and air dried. The pellet was dissolved in 30 to 50 µl of ultrapure water
(or treated with diethyl pyrocarbonate). rRNA was depleted by using the MICROBExpress
kit (Ambion) according to the manufacturer’s protocol. DNase treatment was done according
to the manufacturer protocol with the Ambion TURBO DNA-free kit.
RNA-seq.
Removal of 16S and 23S rRNAs from total RNA was performed with the MICROBExpress bacterial
mRNA purification kit (Ambion) according to manufacturer’s protocol (67). RNA concentrations
and quality were determined by using the Bioanalyzer 2100 (Agilent). For whole-transcriptome
sequencing (RNA-seq), cDNA libraries were prepared by the Illumina mRNA-seq TruSeq
protocol according to the manufacturer’s protocol without the poly(A) isolation stage.
Sequencing was performed with an Illumina HiSeq 2500 sequencing machine. Sequencing
generated 20 to 28 million raw reads per sample, of which 2 to 6 million were mapped
to protein-encoding genes. The RNA-seq data were analyzed by a standard protocol based
on the number of reads per kilobase of coding sequence length per million reads analyzed
(RPKM) (68).
Construction of deletions and recombinant plasmids.
All site-specific gene knockouts were obtained as described by Datsenko and Wanner
(69). Briefly, competent wild-type O78-9 bacteria were transformed with plasmid pKD46.
The transformants were grown in ampicillin-containing LB medium, induced with arabinose,
made competent for electroporation, and stored at −70°C until used. A linear PCR product
was made on the template of a kanamycin resistance cassette flanked by FLP recognition
target (FRT) sequences from the pKD4 plasmid according to the region to be deleted.
The primers were designed to contain 36 nucleotides from the flanking region of the
sequence to be deleted from strain O78-9. Kanamycin-resistant recombinants were screened
by means of colony PCR with primers k2 and kt. Unless stated otherwise, the pKD46
plasmid was cured by growth on LB medium at 42°C. The resulting bacteria were transformed
with pCP20 and grown on LB medium at 42°C to promote both FRT-specific recombination
and curing of the plasmid. The final deletion was verified by sequencing. Recombinant
plasmids for complementation were constructed by cloning PCR-amplified DNA fragments
into a pBAD24 vector with Fast-Link DNA ligase (Epicentre). The ligation products
were transformed into competent 078-9 bacteria, and the resulting clones were screened
by colony PCR.
Human serum.
Sterile and filtered human male AB plasma was acquired from Sigma-Aldrich. Human serum
was inactivated by incubation for 40 min at 56°C to eliminate all of the active immune
complement system.
Clustering and enrichment analysis.
E. coli 078-9 is still available in a roughly annotated version. HMMER (70) was used
for the prediction of gene functions according the TIGRFAM classification. For this
purpose, HMMER used the Hidden Markov file containing the definitions derived from
sequences of already assigned TIGR protein families from other bacteria, which can
be downloaded from the website ftp://ftp.jcvi.org/pub/data/TIGRFAMs. In the resulting
list, locus tags (RACOXXXX) were assigned to one or more TIGRFAMs, which were assigned
to (metabolic) subroles, which were then summarized in hierarchically parent main
roles. TIGRFAMs that were assigned to generic categories such as hypothetical proteins,
general, and so on were, if possible, manually reclassified into more-specific categories.
By using Voronoi tree maps (14, 15), all functionally assigned genes were shown in
a hierarchically organized and space-filling manner, as shown in Fig. 2. The space-filling
approach divides the two-dimensional plane into subareas according the main rules,
which are then subdivided into subsubareas of the subrules and so forth. The polygons
on the last level represent all of the functionally assigned genes. Transcriptional
data were mapped by using a divergent color gradient starting with gray in the middle
and ending with blue (for repression) and orange (for induction) on the sides.
Identification of Fur binding sites.
Fur binding site position weight matrices (PWMs) were downloaded from PRODORIC version
8.9 (54). The score of each sequence is the log of the ratio of the likelihood of
the sequence given the PWM model to the likelihood of the sequence given a background
model. The background model assumes that at each position, the probability of each
character equals its frequency in a concatenated regulatory sequence of 078-9. A significant
hit is defined as a score of >4.6. The position-specific scoring matrix for the consensus
logo was created with WebLogo.
Gel-free absolute proteome quantification.
E. coli 078-9 was grown in 50 ml of MOPS medium in shaking flasks under agitation
at 37°C. At an OD600 of 0.5, human serum, inactivated human serum, or saline was added
to a final concentration of 40% and the bacteria were incubated for 45 min. Cells
were harvested by centrifugation (8,000 × g, 15 min, 4°C). The resulting cell pellet
was resuspended and washed twice in 50 mM triethylammonium bicarbonate. Cell disruption
was carried out in a RiboLyzer (Thermo Fisher Scientific, Waltham, MA) for 2 × 30 s
at 6,800 rpm. Two centrifugation steps were used to remove glass beads and cell debris.
The supernatant was used for the following analysis.
Protein extracts (250 µg) were digested with trypsin as described by Muntel et al.
(71); this was followed by a desalting step via stage tips using a standard protocol
described by Rappsilber et al. (72). For absolute quantification, an internal standard
protein (tryptic digest of alcohol dehydrogenase; Waters, Milford, MA) was used to
spike the samples at a final concentration of 50 fmol/µl.
Data acquisition via LC-IMSE setup.
Peptide samples were analyzed with a nanoACQUITY ultraperformance liquid chromatography
(UPLC) system (Waters) coupled to a Synapt G2 mass spectrometer (Waters). For each
sample condition, three biological replicates were acquired with three technical replicates
each. The samples were loaded directly onto the analytical column (nanoACQUITY UPLC
column, BEH130 C18, 1.7 µm, 75 by 200 mm; Waters) with 99% buffer A (1% acetic acid),
1% buffer B (1% acetic acid in acetonitrile) at a flow rate of 300 nl/min within 35 min.
Separation of peptides was achieved in 165 min by applying the following gradient:
to 18% buffer B in 102 min, to 26% buffer B in 22 min, to 99% buffer B in 16 min,
at 99% buffer B for 10 min, and equilibration for 15 min with 99% buffer A. The Synapt
G2 mass spectrometer (Waters) was equipped with a NanoLockSpray source in positive
mode and operated with the MassLynx V 4.1 software (Waters). The analyzer was set
to resolution mode and operated with the ion mobility separation. For analysis of
peptide ions, a mass range of 50 to 2,000 Da Th was used, low-energy (MS) scans were
set to 4 V (collision energy [CE]), for elevated-energy ion MS (IMSE) scans, the CE
was ramped from 25 to 40 V, the scan time was set to 1 s, the wave velocity was ramped
from 1,000 to 400 m/s, and the wave height was set to 40 V. For lock mass correction,
[Glu1]-fibrinopeptide B solution (m/z 785.8426 Da, 500 fmol/µl in 50% [vol/vol] acetonitrile–0.1%
[vol/vol] formic acid) was constantly infused at a flow rate of 500 nl/min and scans
were acquired at intervals of 30 s.
Analysis of LC-IMSE data.
For identification and quantification of proteins, raw data were imported into ProteinLynx
Global Server (PLGS) 2.5.3 and processed via an Apex3D algorithm. The following processing
parameters were used. The chromatographic peak width and MS time-of-flight resolution
were set to automatic, the lock mass for charge 2 was 785.8426 Da/e, the lock mass
window was 0.25 Da, the low-energy threshold was 250 counts, the elevated-energy threshold
was 30 counts, the retention time window was set to automatic, and the precursor/fragment
ion cluster intensity threshold was 1,000 counts. A database search was carried out
by the ion accounting algorithm implemented in PLGS (67) by using a database that
consisted of 4,628 E. coli O78-9 entries in a raw database (73–75; Huja et al., unpublished
data) and 51,767 human entries in a human reviewed database uploaded from UniProtKB.
This database was complemented with common laboratory contaminants and the yeast ADH1
sequence. The following parameters were used for positive protein identification.
The peptide tolerance was set to automatic; the protein tolerance was set to automatic;
the minimum number of fragment ion matches per peptide was set to 1; the minimum number
of fragment ion matches per protein was set to 5; the minimum number of peptide matches
per protein was set to 1; the primary digest reagent was trypsin; the number of missed
cleavages was 2; the variable modifications were carbamidomethylation C (+57.0215),
deamidation N, Q (+0.9840), and oxidation M (+15.9949); the false-discovery rate (FDR)
was 5%; and the calibration protein was yeast ADH1.
All of the proteins identified in only one or two of the nine replicates (three technical
replicates of each of the three biological replicates) of a sample were discarded.
This filter procedure has revealed protein results with an FDR of <2% on the protein
level. Proteins were considered to have significant expression differences if they
showed at least a 2-fold change in serum or inactivated serum relative to the MOPS
control. Only genes that were detected by both LC-IMSE and RNA-seq were used for the
comparison of mRNA levels and protein abundance.
SUPPLEMENTAL MATERIAL
Appendix SA
Effects of active and inactive serum on protein levels. Protein levels were determined
as described in Materials and Methods. Download
Appendix SA, XLSX file, 0.1 MB