Time-resolved quantitative proteome profiling of host–pathogen interactions: The response of...
-
Upload
frank-schmidt -
Category
Documents
-
view
223 -
download
10
Transcript of Time-resolved quantitative proteome profiling of host–pathogen interactions: The response of...
RESEARCH ARTICLE
Time-resolved quantitative proteome profiling of
host–pathogen interactions: The response of
Staphylococcus aureus RN1HG to internalisation by
human airway epithelial cells
Frank Schmidt1�, Sandra S. Scharf1�, Petra Hildebrandt1, Marc Burian1, Jorg Bernhardt2,Vishnu Dhople1, Julia Kalinka1, Melanie Gutjahr1, Elke Hammer1 and Uwe Volker1
1 Interfaculty Institute for Genetics and Functional Genomics, Ernst-Moritz-Arndt-University Greifswald,Greifswald, Germany
2 Institute for Microbiology, Ernst-Moritz-Arndt-University Greifswald, Greifswald, Germany
Received: January 20, 2010
Revised: April 30, 2010
Accepted: May 10, 2010
Staphylococcus aureus is a versatile Gram-positive pathogen that gains increasing importance
due to the rapid spreading of resistances. Functional genomics technologies can provide new
insights into the adaptational network of this bacterium and its response to environmental
challenges. While functional genomics technologies, including proteomics, have been
extensively used to study these phenomena in shake flask cultures, studies of bacteria from
in vivo settings lack behind. Particularly for proteomics studies, the major bottleneck is the
lack of sufficient proteomic coverage for low numbers of cells. In this study, we introduce a
workflow that combines a pulse-chase stable isotope labelling by amino acids in cell culture
approach with high capacity cell sorting, on-membrane digestion, and high-sensitivity MS to
detect and quantitatively monitor several hundred S. aureus proteins from a few million
internalised bacteria. This workflow has been used in a proof-of-principle experiment to
reveal changes in levels of proteins with a function in protection against oxidative damage
and adaptation of cell wall synthesis in strain RN1HG upon internalisation by S9 human
bronchial epithelial cells.
Keywords:
Host–pathogen interaction / In vivo proteomics / Microbiology / Pulse-chase SILAC /
S. aureus / S9 human bronchial epithelial cells
1 Introduction
Staphylococcus aureus is a Gram-positive microorganism that
permanently colonises 20% of healthy adults and transiently
colonises up to 50% of the general population [1]. This very
versatile organism is both a major human pathogen and a
ubiquitous commensal and coloniser of the skin and
mucous membranes [2]. As a result, S. aureus can cause a
wide range of diseases from superficial skin infections to
life-threatening conditions including septicaemia, endo-
carditis, and pneumonia [1, 3]. The increasing resistance of
this pathogen to almost all antibiotics including methicillin
(methicillin-resistant S. aureus (MRSA)) severely compli-
cates therapeutic intervention. Recently, the number of
invasive infections caused by community-acquired S. aureusalso increased in healthy children and young adults [4, 5].
Therefore, the emergence of highly pathogenic MRSAAbbreviations: MEM, minimal essential medium; MRSA, methi-
cillin-resistant Staphylococcus aureus; SILAC, stable isotope
labelling by amino acids in cell culture �These authors contributed equally to this work
Correspondence: Professor Uwe Volker, Interfaculty Institute for
Genetics and Functional Genomics, Ernst-Moritz-Arndt-Univer-
sity, Friedrich-Ludwig-Jahn-Strasse 15a, D-17487 Greifswald,
Germany
E-mail: [email protected]
Fax: 149-3834-8680005
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
Proteomics 2010, 10, 2801–2811 2801DOI 10.1002/pmic.201000045
strains (community-acquired MRSA) causing serious infec-
tions in otherwise healthy individuals [6] represents a major
threat and underscores the need for a comprehensive
understanding of virulence mechanisms. To date, a broad
collection of virulence factors, the repertoire of which greatly
varies among different S. aureus strains, is known [7, 8].
However, a better understanding of the adaptation of
S. aureus to the host and its molecular pathogenesis will be
important for the development of new therapeutics. Despite
its capacity for survival and persistence in various tissues,
S. aureus is classically considered as an extracellular patho-
gen [9]. Nevertheless, it is able to internalise and survive for
some days within non-professional phagocytic cells such as
epithelial and endothelial cells as well as osteoblasts [10–12].
Internalisation might constitute a bacterial strategy to evade
the host’s defence reactions and the action of antibiotics that
act mainly onto extracellular pathogens. This intracellular
niche might thus constitute a reservoir for chronic or
relapsing infections, a hypothesis that is compatible with the
high relapse rate and prolonged treatment times, which are
observed for S. aureus infections [13, 14]. Contrary to their
potential importance, so far post-invasion events have been
studied only to a limited extend.
In the post-genomic era, bacterial adaptation reactions can
now be studied at genome-wide scale. Functional genomics
technologies have been used extensively to study adaptation
reactions of in vitro grown S. aureus cultures also at the
proteomic scale. Those studies have also included conditions
that mimic the behaviour in infection-related settings, such as
iron limitation and oxidative stress [15–18]. Certainly, the shift
from extracellular to intracellular life style is linked to a fast
and extensive adaptation, the analysis of which will likely
reveal new facets of S. aureus pathophysiology because shake
flask experiments mimic the conditions the pathogen
encounters inside eukaryotic host cells only to a limited extent.
However, due to the challenge of obtaining sufficient bacterial
cells for such studies, genome-wide functional genomics
analyses of the adaptation reactions of S. aureus to the host cell
environment are rare and thus far confined to expression
profiling. Voyich et al. have studied the adaptation of methi-
cillin-resistant and methicillin-susceptible S. aureus strains to
phagocytosis by human neutrophils and revealed a gene
expression program that likely contributes to evasion of innate
host defence (e.g. upregulation of capsule synthesis, oxidative
stress, and virulence) [19]. In a second example, the response
of S. aureus to internalisation by human epithelial cells has
been studied [11]. Investigations that address the proteome of
internalised S. aureus are still lacking.
In general, proteomic studies of internalised pathogens are
still rare due to the challenge of obtaining a sufficient number
of infecting bacteria. The proteome of Francisella tularensis has
been characterised after isolation from the spleen of mice viaimmunomagnetic separation by a classical 2-DE approach [20].
However, even if this technology is still well suited for the
analysis of physiological questions particular in simple
unicellular bacteria, the number of cells required for such a
comprehensive 2-DE based proteome approach is high, often
exceeding the numbers that are available from in vivo infection
models. Therefore, a different path has been used for the study
of Salmonella enterica isolated from infected mice [21]. The
combination of flow cytometric sorting with highly sensitive
MS allowed a comprehensive analysis of the proteome of
Salmonella and indicated that its robust metabolism limits the
possibilities for the discovery of new antimicrobials [21].
However, for these experiments also large numbers of patho-
gens exceeding 108 were sorted, a number that cannot easily be
accomplished in infection models with S. aureus.In the study presented here, we now provide a first, time-
resolved in vivo proteome profile of S. aureus cells internalised
by S9 human bronchial epithelial cells. To accomplish this
analysis, we have used a pulse-chase SILAC (stable isotope
labelling by amino acids in cell culture) labelling approach in
combination with high-capacity cell sorting, direct on-
membrane digestion, and high-precision LTQ-Orbitrap-MS.
This workflow facilitated the identification of 591 and the
quantitation of 367 proteins at different time points after
invasion from only 3� 106 to 6� 106 S. aureus cells.
2 Materials and methods
2.1 Human bronchial epithelial cells
The S9 cell line (ATCCs number CRL-2778) is a human
bronchial epithelial cell line immortalised with an adeno/
SV40 hybrid virus [22]. The cystic fibrosis phenotype of its
parent cell line, 101 IB3-1, had been corrected by introduc-
tion of the gene encoding wild-type cystic fibrosis trans-
membrane conductance regulator [23]. S9 cells were grown
in minimal essential medium (MEM; PromoCell, Heidel-
berg, Germany) supplemented with 4% FBS (Biochrom,
Berlin, Germany), 1% non-essential amino acids, and 4 mM
glutamine (both PAA Laboratories, Pasching, Austria) in a
humidified incubator at 371C with 5% CO2. This medium is
referred to as eMEM.
2.2 S. aureus strain RN1HG
For SILAC, we used the S. aureus strain RN1HG [24]
carrying the plasmid pMV158GFP [25]. This strain consti-
tutively expresses GFP. Bacteria were grown in MEM
without sodium bicarbonate, arginine, and lysine (custom
formulation by PromoCell), but containing 1% of non
essential amino acids (final concentrations: 0.1 mM alanine,
asparagine, aspartic acid, glutamic acid, glycine, proline,
and serine each; PAA). Furthermore, we added to the
medium 0.6 mM heavy arginine (Arg-6, 13C6) and 0.4 mM
heavy lysine (Lys-6, 13C6; both Cambridge Isotope Labora-
tories, Andover, USA), 10 mM HEPES, 4 mM glutamine,
and 2 mM of each alanine, aspartic acid, cysteine, glutamic
acid, histidine, isoleucine, leucine, phenylalanine, proline,
2802 F. Schmidt et al. Proteomics 2010, 10, 2801–2811
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
serine, threonine, tryptophane, and valine (all PromoCell).
This medium is referred to as SILAC pMEM.
2.3 Internalisation experiments
For the internalisation of S. aureus by human S9 cells, S9 cells
were seeded in 24-well plates (8� 104 cells per well) and
cultivated for 3 days. One day before internalisation, S. aureusstrain RN1HG pMV158GFP was inoculated in SILAC pMEM
with 20mg/mL tetracyclin (Carl Roth, Karlsruhe, Germany) as
pre-culture. On the day of the internalisation experiment, a
main culture of S. aureus RN1HG pMV158GFP in SILAC
pMEM was inoculated from an exponentially growing pre-
culture to an optical density at 600 nm (OD600nm) of 0.05.
Bacteria were grown to OD600nm 5 0.4. At this time point,
bacteria were diluted in light eMEM to a final concentration of
1.5� 107 bacteria per mL. For stabilisation of pH during co-
cultivation of S. aureus RN1HG and S9 cells, sodium bicar-
bonate (PAN Biotech, Aidenbach, Germany) was added
to a final concentration of 2.2 g/L. The medium of the S9 cells
was replaced by 1 mL of diluted bacterial culture, which
corresponds to a multiplicity of infection of 25 and inter-
nalisation was allowed for 1 h in a humidified incubator at
371C with 5% CO2.
Extracellular bacteria were killed after 1 h by replacement
of the supernatant with 1 mL eMEM containing lysostaphin
(AMBI Products LLC, Lawrence, NY, USA) at a final
concentration of 10mg/mL. Plating assays were routinely
performed to assure that extracellular S. aureus cells were
efficiently killed (reduction of bacterial titre by at least four
orders of magnitude). After 25 min, the first aliquot of S9 cells
was harvested (referred to as time point 1 h) by aspirating the
medium and washing the cells twice with PBS containing
calcium and magnesium (PAA). S9 cells were lysed in 150mL
0.1% Triton X-100 (ESA Laboratories, MA, USA) per well by
incubation for 7 min at 371C with 5% CO2. Lysed cells were
detached by extensive pipetting and lysates of 12 wells were
pooled into a 15 mL tube on ice. Subsequently, four wells
were washed with 200mL FACSFlow buffer (BD Biosciences,
CA, USA) resulting in a total lysate volume of 200mL per well.
The pooled lysate was put on ice and subjected to flow cyto-
metric measurement and sorting. Sampling was performed
hourly for up to 6 h.
2.4 Flow cytometry
Flow cytometric measurements and sorting were performed
on a biosafety level 2 FACSAria high-speed cell sorter
(Becton Dickinson, CA, USA) with a 488 nm excitation from
a blue Coherent Saphire solid-state laser at 18 mW. Optical
filters were set to detect the emitted GFP fluorescence at
515–545 nm (FITC filter block). All data were recorded at
logarithmic scale with the FACSDiva v5.03 software (Becton
Dickinson). Prior to measurement of the bacteria containing
cell lysate, the proper function of the instrument was cali-
brated using Spherotech’s Rainbow calibration particles
(Spherotech, IL, USA).
2.5 Separation of bacteria by cell sorting
Prior to sorting, the drop delay was adjusted to 499%
sorting with ACCUDROP beads (Becton Dickinson). To
limit artefacts during the sorting process, cells to be sorted
were kept on ice and the filter device on which sorted cells
were collected was cooled to 41C. Sorting was performed
from the gate set in SSc versus FITC dot plot at a sort rate of
up to 3000 cells per second with the sort mode ‘‘purity’’
resulting in a sorted sample that was highly pure, at the
expense of recovery and yield. Bacteria were directly sorted
into wells of a 96-well filtration plate with a hydrophilic low
protein binding Durapore membrane (MSGVS2210, Milli-
pore, Schwalbach, Germany). Using a Millivac Maxi
diaphragm pump (Millipore) and a custom-made manifold
base (University of Greifswald, Germany) vacuum was
applied to the filtration plate during the sorting procedure
(pressure approximately 500 mbar) to accommodate three to
six million bacteria per well and remove the fluid. After
sorting, filter wells with attached bacteria were rinsed once
with 200 mL FACSFlow buffer. The membranes were then
cut from the bottom of the wells right after filtration and
stored at �201C until digestion with trypsin.
2.6 Identification of proteins by LC-LTQ-Orbitrap-
MS/MS
Membranes were cut into four pieces, dissolved in 16mL of
25 mM ammonium bicarbonate (pH 7.8) buffer, and
proteolytically digested with 4 mL of trypsin (0.25 mg/mL
Promega, Madison, WI, USA) overnight in a water bath at
371C [26]. Digestion was stopped by addition of TFA (Merck,
Darmstadt, Germany) to a final concentration of 0.1%.
Debris was removed by a centrifugation step at 16 000� gfor 10 min at room temperature. The supernatant was
transferred into a new 1.5 mL tube and peptides were
purified and desalted using C18-ZipTip columns (Millipore).
A commercial vacuum centrifuge concentrator was used to
remove ACN. MS was performed on a Proxeon nano-LC
system (Proxeon, Odense, Denmark) connected to a LTQ-
Orbitrap-MS (ThermoElectron, Bremen, Germany) equip-
ped with a nano-ESI source. For LC separation, we used a
Acelaim PepMap 100 column (C18, 3 mm, 100 A) (Dionex,
Sunnyvale CA, USA) capillary of 15-cm bed length. The flow
rate used was 300 nL/min for the nano column, and the
solvent gradient used was from 0% B (15 min) to 60% B
(290 min). Solvent A was 0.1% acetic acid, whereas aqueous
90% ACN in 0.1% acetic acid was used as solvent B. The MS
was operated in data-dependent mode to automatically
switch between Orbitrap-MS and LTQ-MS/MS acquisition.
Proteomics 2010, 10, 2801–2811 2803
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
Survey full scan MS spectra (from m/z 300 to 2000) were
acquired in the Orbitrap with resolution R 5 60 000 at m/z400 (after accumulation to a target of 1 000 000 ions in the
LTQ). The method used allowed sequential isolation of up to
five of the most intense ions, depending on signal intensity,
for fragmentation on the linear ion trap using collision-
induced dissociation at a target value of 100 000 ions. Target
ions already selected for MS/MS were dynamically excluded
for 60 s. General MS conditions were electrospray voltage,
1.5 kV; no sheath and auxiliary gas flow. Ion selection
threshold was 500 counts for MS/MS, and an activation
Q-value of 0.25 and activation time of 30 ms were also
applied for MS/MS.
2.7 Analysis of MS data
Differential analysis of SILAC MS data was performed with
Rosetta Elucidator version 3.3.0.1.39 (http://www.rosettabio.
com/products/elucidator; Rosetta Biosoftware, MA, USA).
The frame and feature annotation was done using the
following parameters: retention time minimum cut-off
41 min, retention time maximum cut-off 284 min, m/zminimum cut-off 366 and maximum 1264 m/z. An intensity
threshold of 1000 counts, an instrument mass accuracy of
10.0 ppm, and an alignment search distance of 10.0 min
were applied for binning process. For quantitative analysis,
the data pairs were built using a binning tolerance of
10 ppm, an RT location tolerance of 0.8 min, and a mass
label shift of 6.0202 Da for both arginine and lysine. For
identification, the S. aureus ssp. aureus NCTC 8325 FASTA
sequence (downloaded from NCBI repository, 2009) in
combination with SEQUEST/Sorcerer (Sorcerer version 3.5,
Sage-N Research, CA, USA) was used. MS/MS spectra were
searched with precursor ion tolerance of 20 ppm and a
fragment ion mass tolerance of 1.00 Da. Oxidation of
methionine and SILAC of arginine and lysine (6.0202 Da)
were specified as variable modification. Peptide/protein
identifications were accepted if they exceeded the Peptide-
Teller score of 0.8 and a ProteinTeller score of 0.95 in at
least one experiment [27, 28]. Proteins were considered for
further investigation when at least two peptides in at least
one experiment were identified. For quantitation, only
labelled pairs were considered reaching a Labelled Pair
Status of ‘‘good’’ (at least one of the isotope groups in the
labelled pair was annotated by peptide identification) and for
further quantitative analyses, only proteins having at least
two ‘‘good’’ pairs were taken into account. For protein
classification such as GO, TMD, and signal peptides, the
ProteinCenter (Proxeon Bioinformatics, Version 3.1.2,
Odense, Denmark) was used. Identified proteins of S. aureusssp. aureus NCTC 8325 have been further mapped by using
NCBI to S. aureus COL SACOL locus tags. According to
their function, KEGG [29, 30] orthology classifies genes/
proteins in an acyclic multihierarchical tree-graph. For a
graphical planar representation of this tree-like structure, we
have developed Voronoi Treemaps [31] according to Balzer
and Deussen [32]. Our Voronoi treemap-based layout of
KEGG’s S. aureus COL gene orthology intuitively visualises
the coverage of general cell functions (e.g. metabolism –
blue, cellular processes – yellow, environmental information
processing – red, and genetic information processing –
green) by our gel-free experimental approach.
3 Results and discussion
3.1 Experimental set-up for quantitative analysis of
the bacterial proteome
Even with optimisation of cell sorting for high purity, we
were still expecting significant amounts of contaminating
proteins from human airway epithelial cells. Such contam-
inating human proteins would probably compromise the
quantitation of bacterial proteins present in low quantities.
Therefore, we wanted to apply metabolic labelling using the
well-established SILAC workflow. However, in bacteria, usage
of amino acids labelled with stable isotopes is usually
hampered by the ability of the bacteria to synthesise amino
acids de novo, requiring the availability of mutants auxo-
trophic for arginine and/or lysine. S. aureus strains usually
need a set of amino acids for growth in synthetic medium
albeit having the full complement of enzymes required for
synthesis of amino acids. Therefore, we tested labelling of
S. aureus strain RN1HG in the adapted cell culture medium
pMEM with a mixture of 13C labelled arginine and lysine and
discovered to our surprise saturation labelling in the absence
of a true mutation causing amino acid auxotrophy. This
observation allowed application of SILAC in a pulse-chase
approach, which is an established method for quantitation of
turnover rates and PTMs [33]. Using the workflow depicted in
Fig. 1, S. aureus was initially labelled to completion during
growth in pMEM and then exponentially growing 13C labelled
bacteria were exposed to S9 human bronchial epithelial cells
in eMEM just containing the light forms of arginine and
lysine. Since only light amino acids can be incorporated into
bacterial proteins during contact and internalisation, changes
in the bacterial proteome could be monitored due to the
increase in 12C counterparts compared to 13C amino acids.
After 1 h, non-internalised bacteria were killed by treatment
with lysostaphin, which cannot enter eukaryotic cells.
Subsequently, at each of the six time points, the invasive
bacteria were subjected to flow cytometry and sorted directly
onto a filter device. This set-up allowed the efficient collection
of very dilute bacterial samples in a very small volume in a
time range of 40–50 min. After on-membrane digestion with
trypsin, bacterial proteins/peptides were identified by routine
LTQ-Orbitrap-MS, quantified and classified according to
physiological function. Using the workflow just described, we
performed two independent internalisation experiments in
which we followed the fate of bacteria up to 6 h after inter-
nalisation in hourly intervals.
2804 F. Schmidt et al. Proteomics 2010, 10, 2801–2811
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
3.2 Flow cytometric separation of S. aureus RN1HG
for proteome profiling
With the sorting approach applied, we were able to collect
up to 6.4� 106 GFP-positive bacterial events in a standar-
dised sorting time of around 50 min per sampling point.
GFP-positive bacteria were gated in a range of 104�105
intensity units in the FITC/GFP channel and 102�103
intensity units in Side or Orthogonal Scatter (SSc) (Fig. 2).
This range was used to exclude most of the remaining
debris of S9 cells. The number of sorted GFP-positive
bacterial events varied from 3.1 million after 1 h of
internalisation to 6.4 million after 6 h. Both independent
internalisation experiments showed the same trend
of a clear increase in number of events over the entire
period of 6 h, probably indicating that S. aureus RN1HG
continues to multiply after internalisation (see also general
increase of light versus heavy peptide/protein signals
described below).
3.3 Identification of proteins of internalised
S. aureus RN1HG and assignment of proteins
to cellular compartments
Direct sorting of GFP-positive bacteria onto the low protein
binding Durapore membrane facilitated easy concentration,
washing, and on-membrane digestion and concomitant
minimisation of sample loss. The sorted GFP-positive
bacteria were tryptically digested and analysed by LC-LTQ-
Orbitrap-MS using a long gradient of 290 min. Thus, we
were able to obtain from only 3� 106 to 6� 106 sorted
S. aureus cells 491 proteins in experimental series 1 and 526
proteins in experimental series 2, with an overlap of 426
proteins identified in both internalisation experiments with
at least two peptides. Sixty-five proteins were only seen in
experiment 1 and 100 only in experiment 2, which is likely
due to the fact that these proteins were present at threshold
levels sometimes not allowing detection of two peptides.
However, the quantification was not compromised by these
differences because heavy and light peptides were present in
the same sample and we used stringent selection criteria
only looking for quantitative information for proteins
detected in both experiments with at least two peptide ratios.
S. aureus RN1HG pMV158GFP culture
pMEM + C labeled arginine/lysinepMEM + labeled arginine/lysine
for internalization experiment
Level of stable isotope incorporation
h 61h 0
100 % C
Transfer to S9 monolayer
eMEM + C arginine/lysine
start of internalization and
pulse-chase
Replacement of
Lysostaphin
h 6h 0
100 % C
Replacement of
C arginine/lysine by C
cell harvest every hour
0 % C
Flow cytometry and sorting
removal of cell debris from S9
cells by sorting GFP-positive
bacteria on filter device+ -
Tryptic digestDigestion on membrane
purification and measurement
by LTQ-Orbitrap followed by
identification and
quantitation with Elucidator
Pathway mapping
Figure 1. Workflow for quantitative proteome mapping of inter-
nalised S. aureus. S. aureus RN1HG pMV158GFP grown in heavy
medium was transferred to S9 monolayers and internalisation
was allowed for 1 h. After killing of extracellular bacteria with
lysostaphin, eukaryotic cells were lysed at defined time points to
release intracellular S. aureus cells. GFP-positive bacteria were
separated via flow cytometry and collected on a filter device.
Sorted bacteria were subjected to on-membrane tryptic diges-
tion and purified peptides were measured by LTQ-Orbitrap-MS.
Data analysis was performed with Rosetta Elucidator and iden-
tified proteins were mapped onto metabolic pathways.
Figure 2. Summary of sorted bacterial events. Representative
FITC (GFP) versus SSc dot plot of internalised S. aureus RN1HG
pMV158GFP. GFP-positive bacterial events in the highlighted
area were sorted onto a filter device and subjected to analysis by
LC-ESI-MS/MS.
Proteomics 2010, 10, 2801–2811 2805
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
The ProteinCenter program package was used to calculate
the number of TMDs and signal peptides. As a result, 69% of
the proteins identified were detected with null, 20% with one,
6% with two, and 5% with more than two TMD, indicating
that the majority of the proteins belongs to the cytoplasmic
fraction. This is not surprising because due to the low
amount of material available enrichment was not feasible and
no efforts were made to increase the solubility of proteins
with extreme hydrophobicity such as integral membrane
proteins. Similarly, only 42 out of 591 proteins were predicted
to have a specific signal sequence because true extracellular
proteins were likely lost during cell sorting. To benchmark
the workflow developed, we compared the data of the in vivoproteomics approach with the most comprehensive proteome
analysis reported for S. aureus so far (Fig. 3) [34]. In that
study, complementing subproteomes were analysed for
exponentially growing and glucose-starved S. aureus cultures
to derive by combination of gel-based and gel-free technolo-
gies a coverage of close to 80% of the expressed proteome of
S. aureus COL. First of all, the number of proteins covered in
internalised S. aureus RN1HG was of course with 591 distinct
proteins much smaller than the 1703 proteins reported by
Becher et al. [34]. Second, particularly integral membrane
proteins and extracellular proteins were barely covered due to
the reasons described above. However, we also detected a
number of cell wall-associated proteins and lipoproteins
probably owing to the direct digestion of S. aureus cells
without prior cell disruption. The majority of proteins iden-
tified from S. aureus RN1HG after sorting and on-membrane
digestion were cytoplasmic (83%).
3.4 Mapping the proteome of internalised S. aureus
RN1HG onto pathways – the value of in vivo
proteomics
To investigate the coverage of known pathways in S. aureusstrain RN1HG, a Voronoi treemap-based layout of KEGG’s
S. aureus gene orthology was used to intuitively visualise
coverage of detected proteins compared to the theoretical
ones (Fig. 4). The 591 identified proteins were mainly
assigned to metabolism and genetic information processing.
Thus, particularly good coverage was accomplished for the
ribosome, amino acyl t-RNA synthetases, RNA-polymerase,
and other proteins involved in transcription as well as many
metabolic pathways including among others purine and
pyrimidine, peptidoglycan and fatty acid synthesis, and
central carbon metabolism with gylcolysis/glyconeogenesis,
pentose phosphate pathway, pyruvate metabolism, and
citrate cycle. Thus, the time-resolved proteomics snapshots
provide insights into many important branches of metabo-
lism likely because these proteins are present in rather high
levels. On the other side, information on environmental
information processing is mostly missing because integral
membrane proteins and low-abundance proteins were not
covered (Fig. 4).
Figure 3. Allocation of identified proteins to bacterial compartments and coverage of the different sub-proteomic fractions. The data of the
in vivo proteomics approach presented here are benchmarked with the most extensive proteomic investigation of S. aureus reported thus
far [34]. Starting from DNA array-based expression studies, Becher et al. defined the expressed proteome and then used extensive
proteome analysis of in vitro grown S. aureus COL to provide a proteomic inventory of 80% of the expressed proteome [34]. Six different
protein classes (cytosolic proteins, integral membrane proteins, extracellular proteins, cell wall associated proteins, lipoproteins, and
sortase substrates) are displayed. The figure displays the proportions of the different protein classes as percent of the total proteins
covered and the total number of proteins covered in the respective study. These numbers are presented for the proteins whose genes are
predicted to be expressed in vitro, the proteins identified by Becher et al. and the proteins covered in this study of internalised S. aureus
cells.
2806 F. Schmidt et al. Proteomics 2010, 10, 2801–2811
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
3.5 Quantitation of proteins via pulse-chase SILAC
labelling approach
For the quantitative analysis of the time course experiment,
S. aureus strain RN1HG was grown in pMEM containing
heavy 13C arginine and lysine and subsequently, labelled
bacteria were transferred to S9 cells growing in eMEM
containing only light amino acids. This time point was set as
null (internalisation and start of chase phase). After 1 h,
extracellular bacteria were killed and only internalised
bacteria were investigated further. Because light amino
acids accumulate in addition to their remaining heavy
counterparts, for each peptide a labelled and unlabelled
form could be observed. These pairs were further used to
quantify the proteins over a period of 6 h. Figure 5 illustrates
the rationale of such a pulse-chase experiment. At the pulse-
chase starting point, only heavy peptides are present. During
time, the intensities of light peptides continuously increase
until heavy peptides are not detected against the vast
majority of light peptides. By plotting all detected ratios, the
majority of proteins showed slightly increasing intensities of
the light peptides compared to the heavy ones (shown as
grey thick zone). This overall increase of ratios likely reflects
growth of the S. aureus population, but might in part also be
due to turnover of old proteins. Ratios significantly higher
than the median (shown as thin grey line) can be assumed
to either reflect increased levels due to upregulation of
synthesis rates or increased degradation, which would
continuously dilute heavy signal because in contrast to the
light form it would not be refilled by new synthesis. On the
other side, a significantly slower slope of the ratios than the
median (shown as thin grey line) can be interpreted as a
decreased synthesis compared to the phase prior to inter-
nalisation.
Figure 4. Overview of S. aureus pathways
plotted as Voronoi treemap. The map is
based on KEGG’s S. aureus gene orthology
and visualises the coverage of the protein
inventory. Each small field represents an
independent protein. Proteins covered in
KEGG’s S. aureus gene orthology but not
found in this study are displayed in grey and
those identified from internalised S. aureus
are displayed by the colour of the respective
branch of physiology (blue – metabolism;
red – genetic information processing; green
–environmental information processing; and
yellow – cellular processes). The figure
displays the 591 proteins identified in the
study, which are all listed in Supporting
Information Table 1.
Figure 5. Idealised principle of quantitation after pulse-chase
SILAC. S. aureus was initially grown in heavy labelled medium
for saturation labelling of all proteins with heavy amino acids.
For proteome analysis, the bacterial cells were transferred to
light medium allowing incorporation exclusively of light amino
acids and simultaneously exposed to human bronchial epithelial
S9 cells. Thus, the log-ratio of light versus heavy amino acids
allows analysis of changes in the proteome due to the exposure
to S9 cells and internalisation of S. aureus RN1HG by S9 cells.
The idealised grey zone covers the changes observed for the
majority of proteins and represents intracellular growth. If
changes for ratios of proteins in time deviate significantly (more
than twofold) from the median of all proteins (outside of the
zone flanked by thin grey lines), these candidates were consid-
ered as differentially regulated (either up- or downregulated).
Proteomics 2010, 10, 2801–2811 2807
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
Taking all time points together, we were able to identify
591 proteins from aliquots of only 3�6� 106 sorted
S. aureus cells. Of those, 511 could also be quantified with at
least two ‘‘good’’ labelled pairs; 430 proteins in experiment 1
and 458 proteins in experiment 2. When comparing the
independent experimental series, 367 proteins (72%) could
be quantified in both and those were used for the assess-
ment of the impact of internalisation on the proteome
pattern of S. aureus RN1HG. In Fig. 6, the overall ratios of
the six time points from experiment two (Fig. 6A) and the
Log10 intensities of light versus heavy peptides (Fig. 6B)
were plotted. The calculated median, shown as green line,
continuously increased from 1 h (0.9) to 6 h (4.4) indicating
approximately two doublings within 6 h. Applying a fold
change cut-off of factor two (Fig. 6A, red lines), approxi-
mately 10–15% of the proteins were up- or downregulated
over time (Supporting Information Table 2). An example of
the intensity calculation for a non-regulated protein (EF-Tu)
is given in panel 6D (Fig. 6) and Fig. 7A. After 1 h, the
intensity of the light peptide (red) was almost similar to
the intensity of corresponding heavy peptide. During the
experimental time window, this ratio continuously increased
Figure 6. Display of the time-
resolved analysis of changes in
the proteome pattern of inter-
nalised S. aureus cells. Ratio
data of stable isotope labelling
by amino acids in cell culture
(SILAC) after pulse-chase are
displayed for the time period
from 1 to 6 h after internalisa-
tion into S9 cells. (A) Protein
ratios of experiment two are
plotted. In the x-axis, all quan-
tified proteins are ranked from
the smallest to the largest ratio
to display the overall distribu-
tion of the ratios. The y-axis
displays the protein ratios (L/H)
of experiment two. (B) Dot-Plot
display of intensity values.
Dashed lines represent the
median calculated for each
time point. Dotted lines indi-
cate a twofold change. (C)
Mean intensity values of heavy
and light forms of the protein
PrsA as an example of a
protein displaying internalisa-
tion-related increase in level;
(D) mean intensity values of
heavy and light forms of the
protein EF-Tu, the level of
which only increases in the
light from due to growth of
S. aureus RN1HG in S9 cells.
2808 F. Schmidt et al. Proteomics 2010, 10, 2801–2811
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
very similar to the median. In contrast to EF-Tu, the
intensity pattern of the protein PrsA is exemplarily shown in
panel 6C. After 2 h, the intensity of the heavy peptide
compared to the light one is much lower than for EF-Tu,
indicating rapid new synthesis or degradation of PrsA post-
invasion.
3.6 Proteins of S. aureus RN1HG displaying
internalisation-associated changes in level
Considering both replicates and the six time points, it was
obvious that the level changed significantly for a number of
proteins after internalisation. Figure 7 summarises the
time-resolved changes for a number of typical examples
against the background of the majority of proteins, which
displayed only a fourfold increase in intensity during the 6-h
period monitored. Besides EF-Tu, almost all ribosomal
proteins displayed this slow increase just mentioned.
However, the ribsosomal proteins S9 (30S subunit compo-
nent) and the 50S ribosomal protein L9 (see Fig. 7A) showed
a much stronger increase in the light signal intensity which
was distinctively different from those of the other 23 50S
ribosomal proteins observed. However, in both cases, the
strong increase in the light to heavy signal very likely does
not indicate more pronounced synthesis but rather
decreased stability. L9 has been shown to display much
greater exchange in vivo by isotope-transfer experiments
already a long time ago [35, 36].
Other prime examples of regulated proteins are the foldase
protein PrsA [37] (Figs. 6C and 7B), which plays a major
role in protein secretion by assisting in the post-transloca-
tional extracellular folding of several secreted proteins
and the glucose-specific enzyme II A of the PTS system
[SAOUHSC_01430] (Fig. 7B). However, the list of proteins
displaying internalisation-associated increases in the
ratios of light to heavy signals can be extended to proteins
such as ornithine cyclodeaminase, threonine synthase,
aspartate-semialdehyde dehydrogenase, and the iron
compound ABC transporter (Supporting Information Table
2). Besides these proteins, some hypothetical proteins such as
SAOUHSC_00717 showed a very strong post-invasive
response to the host (fold change 5 21). The latter is a 146
amino acid protein consisting of a signal peptide and an
electron transfer domain 13. This domain is a component of a
novel electron-transfer system potentially involved in oxida-
tive modification of animal cell-surface proteins [38]. From a
functional point of view, the observation of fast increases in
the ratios of the peptide methionine sulphoxide reductases
MsrA2 and MsrB was particularly interesting. Peptide
methionine sulphoxide reductases have an important func-
tion as repair enzymes for proteins that have been inactivated
Figure 7. Quantitative analysis of changes in the protein level of internalised S. aureus RN1HG. (A) Total amount of quantified proteins per
experiment (grey lines). Dashed lines represent the median calculated for each time point. Dotted lines indicate a twofold change from the
median. Elongation factor Tu as an example of a non-regulated protein over time and 50S ribosomal protein L9 as the only regulated
protein of all identified 50S ribosomal proteins (see Supplorting Information Tables 1 and 2). (B) Ratio data of the response regulator VraR
and two proteins affected by this two-component regulatory system (PrsA, foldase protein; PTS Enzyme IIA, phosphotransferase system
enzyme IIA). Dashed lines represent the median calculated for each time point.
Proteomics 2010, 10, 2801–2811 2809
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
due to oxidation [39] and thus this observation might be
interpreted as a hint for the exposure to oxidative stress, a
probable defence reaction of the epithelial cells.
3.7 Regulatory adaptation of S. aureus RN1HG after
internalisation by human S9 cells
Particularly interesting was the observation of a strong
increase in the ratio of light to heavy (post to pre-inter-
nalisation) ratio of the response regulator VraR (Fig. 7B).
Dramatic changes in the lifestyle as, for example, the shift
from extracellular to intracellular life, have to be linked to
fast regulatory adaptation of the pathogen to the new
environment. In bacteria, environmental stimuli can be
sensed and transduced by two-component systems [40]. Of
the total of 17 potential two-component systems that have
been identified in S. aureus [41], in this study only the
response regulator VraR (vancomycin resistance-associated
sensor/regulator) probably increased in level over time after
internalisation into S9 cells. At least a twofold upregulation
compared to the bulk proteins could be observed 3 h after
internalisation, which constantly remained high even after
6 h (Fig. 7B). For this example, we believe in a correlation of
increased signal, synthesis, level, and activity because two of
the thirteen target genes reported in a transcriptional study
of the VraSR system [42], namely PrsA and the glucose-
specific enzyme II A of the phosphotransferase system
[SAOUHSC_01430], were shown to display strongly
increased light to heavy ratios as well (Fig. 7B). These data
are plausible from a physiological point of view because
VraSR positively modulates cell wall biosynthesis by upre-
gulation of genes associated with peptidoglycan biosynthesis
[42]. In this study, we did not observe upregulation of MurZ
and PBP2 above the threshold of two (SgtB could not be
quantified). Although the VraSR system was active during
internalisation and survival of S. aureus within S9 cells, cell
wall biosynthesis was probably not the reason for activation.
Since it is also known that VraSR responds to damage of cell
wall structure, i.e. by antibiotics that target cell wall pepti-
doglycan biosynthesis (b-lactams and vancomycin) one may
speculate that within S9 cells, cell wall damage occurs,
which is then sensed by VraSR. Because no antibiotic was
added during cultivation this might be due to host mole-
cules such as components of the complement system.
Another reason of the activity of VraSR might be a certain
rearrangement of the cell envelope.
These hypotheses can now be followed because first hints
have been provided by our study of the in vivo proteome of
S. aureus RN1HG.
4 Concluding remarks
With this study, we provide a workflow allowing time-
resolved analysis of as little as a few million internalised
S. aureus cells, thus opening the possibility of comparative
profiling of the response of wild-type strains and isogenic
mutants lacking important regulators or virulence factors.
This technology thus provides an important tool for the
deciphering of the adaptational network of S. aureus.The changes in the proteome observed for the wild-type
strain S. aureus RN1HG upon internalisation indicate a
response to oxidative stress and adjustments in cell wall
synthesis.
The authors are grateful to Ulrike Lissner for excellent technicalassistance and Leonard Menschner for construction of RN1HGpMV158GFP. Work in the Interfaculty Institute of Genetics andFunctional Genomics (U. V., F. S.) was supported within theframework of the collaborative research project SFBTRR34 by theDeutsche Forschungsgemeinschaft. The authors are grateful forthe contribution of Juliane Siebourg from the ETH Zurich for thecreation of the Voronoi treemaps and Susann M.uller, Nico Jehmlich,and Dirk Benndorf for help in establishing the cell sorting protocol.
The authors have declared no conflict of interest.
5 References
[1] Lowy, F. D., Staphylococcus aureus infections. N. Engl.
J. Med. 1998, 339, 520–532.
[2] Foster, T. J., The Staphylococcus aureus ‘‘superbug’’.
J. Clin. Invest. 2004, 114, 1693–1696.
[3] Kristinsson, K. G., Adherence of staphylococci to intravas-
cular catheters. J. Med. Microbiol. 1989, 28, 249–257.
[4] Maltezou, H. C., Giamarellou, H., Community-acquired
methicillin-resistant Staphylococcus aureus infections. Int.
J. Antimicrob. Agents 2006, 27, 87–96.
[5] Weber, J. T., Community-associated methicillin-resistant
Staphylococcus aureus. Clin. Infect. Dis. 2005, 41, S269–S272.
[6] Furuya, E. Y., Lowy, F. D., Antimicrobial-resistant bacteria in
the community setting. Nat. Rev. Microbiol. 2006, 4, 36–45.
[7] Sibbald, M. J., Ziebandt, A. K., Engelmann, S., Hecker, M.
et al., Mapping the pathways to staphylococcal pathogen-
esis by comparative secretomics. Microbiol. Mol. Bio.l Rev.
2006, 70, 755–788.
[8] Ziebandt, A. K., Kusch, H., Degner, M., Jaglitz, S. et al.,
Proteomics uncovers extreme heterogeneity in the Staphy-
lococcus aureus exoproteome due to genomic plasticity and
variant gene regulation. Proteomics 2010, 10, 1634–1644.
[9] Lowy, F. D., Is Staphylococcus aureus an intracellular
pathogen? Trends Microbiol. 2000, 8, 341–343.
[10] Sinha, B., Francois, P. P., Nusse, O., Foti, M. et al., Fibro-
nectin-binding protein acts as Staphylococcus aureus
invasin via fibronectin bridging to integrin alpha5beta1.
Cell. Microbiol. 1999, 1, 101–117.
[11] Garzoni, C., Francois, P., Huyghe, A., Couzinet, S. et al.,
A global view of Staphylococcus aureus whole genome
expression upon internalization in human epithelial cells.
BMC Genomics 2007, 8, 171.
2810 F. Schmidt et al. Proteomics 2010, 10, 2801–2811
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
[12] Jevon, M., Guo, C., Ma, B., Mordan, N. et al., Mechanisms of
internalization of Staphylococcus aureus by cultured
human osteoblasts. Infect. Immun. 1999, 67, 2677–2681.
[13] Plouin-Gaudon, I., Clement, S., Huggler, E., Chaponnier, C.
et al., Intracellular residency is frequently associated with
recurrent Staphylococcus aureus rhinosinusitis. Rhinology
2006, 44, 249–254.
[14] Clement, S., Vaudaux, P., Francois, P., Schrenzel, J. et al.,
Evidence of an intracellular reservoir in the nasal mucosa of
patients with recurrent Staphylococcus aureus rhinosinu-
sitis. J. Infect. Dis. 2005, 192, 1023–1028.
[15] Scherl, A., Francois, P., Charbonnier, Y., Deshusses, J. M.
et al., Exploring glycopeptide-resistance in Staphylococcus
aureus: a combined proteomics and transcriptomics
approach for the identification of resistance-related
markers. BMC Genomics 2006, 7, 296.
[16] Weber, H., Engelmann, S., Becher, D., Hecker, M., Oxidative
stress triggers thiol oxidation in the glyceraldehyde-3-
phosphate dehydrogenase of Staphylococcus aureus. Mol.
Microbiol. 2004, 52, 133–140.
[17] Wolf, C., Hochgrafe, F., Kusch, H., Albrecht, D. et al.,
Proteomic analysis of antioxidant strategies of Staphylo-
coccus aureus: diverse responses to different oxidants.
Proteomics 2008, 8, 3139–3153.
[18] Hochgr .afe, F., Wolf, C., Fuchs, S., Liebeke, M. et al., Nitric
oxide stress induces different responses but mediates
comparable protein thiol protection in Bacillus subtilis and
Staphylococcus aureus. J. Bacteriol. 2008, 190, 4997–5008.
[19] Voyich, J. M., Braughton, K. R., Sturdevant, D. E., Whitney,
A. R. et al., Insights into mechanisms used by Staphylo-
coccus aureus to avoid destruction by human neutrophils.
J. Immunol. 2005, 175, 3907–3919.
[20] Twine, S. M., Mykytczuk, N. C., Petit, M. D., Shen, H. et al., In
vivo proteomic analysis of the intracellular bacterial patho-
gen, Francisella tularensis, isolated from mouse spleen.
Biochem. Biophys. Res. Commun. 2006, 345, 1621–1633.
[21] Becker, D., Selbach, M., Rollenhagen, C., Ballmaier, M.
et al., Robust Salmonella metabolism limits possibilities for
new antimicrobials. Nature 2006, 440, 303–307.
[22] Zeitlin, P. L., Lu, L., Rhim, J., Cutting, G. et al., A cystic
fibrosis bronchial epithelial cell line: immortalization by
adeno-12-SV40 infection. Am. J. Respir. Cell. Mol. Biol.
1991, 4, 313–319.
[23] Flotte, T. R., Afione, S. A., Solow, R., Drumm, M. L. et al.,
Expression of the cystic fibrosis transmembrane conduc-
tance regulator from a novel adeno-associated virus
promoter. J. Biol. Chem. 1993, 268, 3781–3790.
[24] Pohl, K., Francois, P., Stenz, L., Schlink, F. et al., CodY in
Staphylococcus aureus: a regulatory link between metabo-
lism and virulence gene expression. J. Bacteriol. 2009, 191,
2953–2963.
[25] Nieto, C., Espinosa, M., Construction of the mobilizable
plasmid pMV158GFP, a derivative of pMV158 that carries
the gene encoding the green fluorescent protein. Plasmid
2003, 49, 281–285.
[26] Schmidt, F., Fiege, T., Hustoft, H. K., Kneist, S., Thiede, B.,
Shotgun mass mapping of Lactobacillus species and
subspecies from caries related isolates by MALDI-MS.
Proteomics 2009, 9, 1994–2003.
[27] Keller, A., Purvine, S., Nesvizhskii, A. I., Stolyar, S. et al.,
Experimental protein mixture for validating tandem mass
spectral analysis. OMICS 2002, 6, 207–212.
[28] Nesvizhskii, A. I., Keller, A., Kolker, E., Aebersold, R.,
A statistical model for identifying proteins by tandem mass
spectrometry. Anal. Chem. 2003, 75, 4646–4658.
[29] Kanehisa, M., A database for post-genome analysis. Trends
Genet. 1997, 13, 375–376.
[30] Kanehisa, M., Goto, S., KEGG: kyoto encyclopedia of genes
and genomes. Nucleic Acids Res. 2000, 28, 27–30.
[31] Bernhardt, J., Funke, S., Hecker, M., Siebourg, J., Sixth
International Symposium on Voronoi Diagrams 2009, pp.
233–241.
[32] Balzer, M., Deussen, O., IEEE Symposium on Information
Visualization, InfoVis 2005 pp. 49–56.
[33] Gustavsson, N., Greber, B., Kreitler, T., Himmelbauer, H.
et al., A proteomic method for the analysis of changes in
protein concentrations in response to systemic perturba-
tions using metabolic incorporation of stable isotopes and
mass spectrometry. Proteomics 2005, 5, 3563–3570.
[34] Becher, D., Hempel, K., Sievers, S., Z .uhlke, D. et al., A
Proteomic view of an important human patho-
gen-Towards the quantification of the entire Staphylo-
coccus aureus proteome. PLoS One 2009, 4, e8176.
[35] Robertson, W. R., Dowsett, S. J., Hardy, S. J., Exchange of
ribosomal proteins among the ribosomes of Escherichia
coli. Mol. Gen. Genet. 1977, 157, 205–214.
[36] Subramanian, A. R., van Duin, J., Exchange of individual
ribosomal proteins between ribosomes as studied by heavy
isotope-transfer experiments. Mol. Gen. Genet. 1977, 158, 1–9.
[37] Sarvas, M., Harwood, C. R., Bron, S., van Dijl, J. M., Post-
translocational folding of secretory proteins in Gram-posi-
tive bacteria. Biochim. Biophys. Acta 2004, 1694, 311–327.
[38] Iyer, L. M., Anantharaman, V., Aravind, L., The DOMON
domains are involved in heme and sugar recognition.
Bioinformatics 2007, 23, 2660–2664.
[39] Kryukov, G. V., Kumar, R. A., Koc, A., Sun, Z., Gladyshev,
V. N., Selenoprotein R is a zinc-containing stereo-specific
methionine sulfoxide reductase. Proc. Natl. Acad. Sci. USA
2002, 99, 4245–4250.
[40] Novick, R. P., Autoinduction and signal transduction in the
regulation of staphylococcal virulence. Mol. Microbiol.
2003, 48, 1429–1449.
[41] Benton, B. M., Zhang, J. P., Bond, S., Pope, C. et al., Large-
scale identification of genes required for full virulence of
Staphylococcus aureus. J. Bacteriol. 2004, 186, 8478–8489.
[42] Kuroda, M., Kuroda, H., Oshima, T., Takeuchi, F. et al., Two-
component system VraSR positively modulates the regu-
lation of cell-wall biosynthesis pathway in Staphylococcus
aureus. Mol. Microbiol. 2003, 49, 807–821.
Proteomics 2010, 10, 2801–2811 2811
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com