Embed Size (px)
Citation preview
Resource
Absolute Proteome Composition and Dynamics
during Dormancy and Resuscitation ofMycobacterium tuberculosisGraphical Abstract
Highlights
d A SWATH MS-based strategy can estimate genome-wide
absolute protein concentrations
d This technique revealed absolute protein abundances in Mtb
during growth and dormancy
d The DosR regulon makes up 20% of the total protein biomass
during dormancy
d Absolute protein abundances reveal maximal enzymatic
reaction velocities in metabolism
Schubert et al., 2015, Cell Host & Microbe 18, 96–108July 8, 2015 ª2015 Elsevier Inc.http://dx.doi.org/10.1016/j.chom.2015.06.001
Authors
Olga T. Schubert, Christina Ludwig,
Maria Kogadeeva, ...,
Stefan H.E. Kaufmann, Uwe Sauer,
Ruedi Aebersold
In Brief
Schubert et al. develop a strategy to
estimate absolute cellular protein
concentrations on a global scale based
on the proteomic technique SWATH MS.
They use this strategy to generate a
resource of absolute protein
concentrations in Mycobacterium
tuberculosis during exponential growth,
induction of dormancy, and resuscitation.
Cell Host & Microbe
Resource
Absolute Proteome Compositionand Dynamics during Dormancyand Resuscitation of Mycobacterium tuberculosisOlga T. Schubert,1,5 Christina Ludwig,1,5 Maria Kogadeeva,1 Michael Zimmermann,1 George Rosenberger,1
Martin Gengenbacher,2,3 Ludovic C. Gillet,1 Ben C. Collins,1 Hannes L. Rost,1 Stefan H.E. Kaufmann,2 Uwe Sauer,1
and Ruedi Aebersold1,4,*1Department of Biology, Institute of Molecular Systems Biology, ETH Zurich, Zurich, CH-8093, Switzerland2Department of Immunology, Max Planck Institute for Infection Biology, Berlin, D-10117, Germany3Department of Microbiology, Yong Loo Lin School of Medicine, National University of Singapore, Singapore SG-117545, Singapore4Faculty of Science, University of Zurich, Zurich CH-8057, Switzerland5Co-first author*Correspondence: [email protected]
http://dx.doi.org/10.1016/j.chom.2015.06.001
SUMMARY
Mycobacterium tuberculosis remains a healthconcern due to its ability to enter a non-replicativedormant state linked to drug resistance. Understand-ing transitions into and out of dormancy will informtherapeutic strategies.We implemented a universallyapplicable, label-free approach to estimate absolutecellular protein concentrations on a proteome-widescale based on SWATH mass spectrometry. Weapplied this approach to examine proteomic reor-ganization of M. tuberculosis during exponentialgrowth, hypoxia-induced dormancy, and resuscita-tion. The resulting data set covering >2,000 proteinsreveals how protein biomass is distributed amongcellular functions during these states. The stress-induced DosR regulon contributes 20% to cellularprotein content during dormancy,whereas ribosomalproteins remain largely unchanged at 5%–7%. Abso-lute protein concentrations furthermore allow proteinalterations to be translated into changes in maximalenzymatic reaction velocities, enhancing under-standing of metabolic adaptations. Thus, global ab-solute protein measurements provide a quantitativedescription of microbial states, which can supportthe development of therapeutic interventions.
INTRODUCTION
To date, most quantification of biomolecules is relative,
describing the fold change of molecules between samples.
However, in biological and biomedical research, there is an
increasing demand for absolute quantification to gain insight
into cellular composition and to enable mathematical modeling
of cellular processes. Proteins are the main molecular effectors
of a cell, and knowledge of their absolute abundance is therefore
crucial to gain a quantitative view of cellular functions. The
method of choice for high-throughput proteome-wide quantifi-
96 Cell Host & Microbe 18, 96–108, July 8, 2015 ª2015 Elsevier Inc.
cation is liquid chromatography-tandem mass spectrometry
(LC-MS/MS) (Aebersold and Mann, 2003). The gold standard
method to determine absolute quantities of proteins by MS is
to compare endogenous peptide levels to spiked-in, isotope-
labeled standards of known concentration (Desiderio and Kai,
1983; Gerber et al., 2003). However, such accurately quantified
reference peptides or proteins are expensive, and therefore,
this approach is not easily implemented on a larger scale.
An alternative strategy for absolute protein quantification uses
fluorescent tagging, but this approach requires genetic manipu-
lation of the cell and is therefore not readily applicable to every
organism (Taniguchi et al., 2010). Recently, it has been shown
that absolute protein concentrations in fast-growing microbes
can also be indirectly estimated by determining absolute protein
synthesis rates using ribosomal profiling (Li et al., 2014).
To enable large-scale and high-throughput absolute abun-
dance estimation of proteins in a label-free manner and without
requiring genetic manipulation, several MS-based methods
have been developed (Ahrne et al., 2013). Generally, these
approaches assume a linear correlation between absolute con-
centration of a protein and spectral counts or measured MS
signal intensities of all or selected peptides of that protein. This
is a non-trivial assumption, as the MS response factor of a pep-
tide depends on its sequence, charge, and other physico-chem-
ical properties (Mallick et al., 2007), and theMS intensities of two
different peptides of identical concentration can vary by several
orders of magnitude (Picotti et al., 2007). It could, however, be
shown that summarizing peptide MS intensities to protein level
in a defined way yields a robust estimator that linearly correlates
with the actual protein concentration. This summarization func-
tion can, for example, be the average or sum of the highest
peptide intensities per protein (Silva et al., 2006), or the sum of
all peptide intensities per protein, divided by the theoretical
number of tryptic peptides (Schwanhausser et al., 2011).
To date, all proteomic studies reporting proteome-wide abso-
lute abundance estimates have been based on untargeted, dis-
covery (shotgun) MS and can therefore be affected by sampling
inconsistency of peptides of lower abundance due to semi-sto-
chastic selection of precursors during data-dependent acquisi-
tion (Domon and Aebersold, 2010). Furthermore, precursor
ion-based quantification of lower intensity signals bears a higher
A
B
C D
Figure 1. A Strategy to Estimate Absolute Cellular Protein Concentrations in Mtb on a Proteome-Wide Scale by SWATH MS
(A) Hypoxia-induced dormancy and resuscitation in Mtb H37Rv was studied by applying gradual oxygen depletion over 20 days followed by re-aeration for
2 days. Protein samples were measured by SWATH MS.
(B) A comprehensive SWATH assay library for Mtb was generated based on various sample types.
(C) Using this assay library, quantitative signals were extracted from the SWATH data. To obtain absolute protein abundance estimates, a set of anchor proteins
was used to identify the best approach to combine peptide into protein intensities and to achieve an optimal linear correlation between protein intensity and actual
absolute protein concentration.
(D) With the resulting absolute complete data matrix, we could perform unique analyses, including investigations of proteome biomass composition and maximal
enzymatic reaction velocity calculations.
risk of interferences from overlapping signals than fragment ion-
based quantification, particularly in complex samples (Gillet
et al., 2012; Rost et al., 2014). Recently, we described an abso-
lute label-free quantification method for targeted proteomics by
selected (ormultiple) reactionmonitoring (S/MRM) (Ludwig et al.,
2012). In contrast to discovery MS, S/MRM allows more consis-
tent and sensitive quantification, but it is limited in the number of
peptides and proteins that can be analyzed in a single MS injec-
tion. The proteomic technique SWATH MS overcomes the
limited capacity of S/MRM while maintaining major advantages
of targeted proteomics, such as a high degree of reproducibility,
large dynamic range, and low limit of detection (Gillet et al.,
2012). SWATH MS is thus optimally suited to acquire prote-
ome-wide, quantitative data reproducibly over many samples.
In the present study, our goal was to exploit the favorable perfor-
mance characteristics of SWATH MS and combine these with a
strategy to estimate absolute protein concentrations for signifi-
cant fractions of a proteome in a label-free manner.
The human pathogenMycobacterium tuberculosis (Mtb) is the
causative agent of tuberculosis and claims 1.5 million lives annu-
ally (WHO, 2014). Not least due to the continuous rise of drug-
resistant Mtb strains, there is an urgent need to develop new
and more efficient intervention measures to fight tuberculosis.
The marked phenotypic drug resistance of Mtb and its success
in evading the host immune system and persisting in the human
body are linked to its ability to enter a non-replicating state (Gen-
genbacher and Kaufmann, 2012). A detailed molecular view of
this dormant state and of the transition into and out of it should
facilitate the quest to findmore efficient anti-infectives to prevent
or cure difficult-to-treat latent tuberculosis infections that are
related to dormant Mtb. Dormancy can be induced by subjecting
Mtb bacilli to hypoxia, a condition that has been shown to be
prevalent in vivo during infection of the host (Via et al., 2008). A
widely used in vitro hypoxia model for Mtb is the Wayne model,
C
where bacilli are gradually deprived of oxygen (Wayne and
Hayes, 1996; Wayne and Sohaskey, 2001). Other models of hyp-
oxia include the defined hypoxia model, where a constant flow of
low-oxygen gas over the surface of a stirred culture is used to
deplete the oxygen rapidly and reproducibly (Rustad et al.,
2008; Sherman et al., 2001). In the Wayne model, the slow
decline in oxygen allows the otherwise obligate aerobe Mtb to
adapt to hypoxic conditions by readjusting its metabolism and
arresting growth (Boshoff and Barry, 2005; Wayne and Sohas-
key, 2001). The underlying genetic program facilitating adapta-
tion to hypoxia is controlled by the dormancy survival regulon,
which is under the control of the transcription factor DosR and
comprises roughly 50 proteins (Boon and Dick, 2002; Park
et al., 2003; Sherman et al., 2001; Voskuil et al., 2003). The
DosR regulon has been associated with ribosome stability, a
shift of metabolism away from aerobic respiration to fermenta-
tion, and maintenance of basal energy levels and redox balance
during hypoxic conditions (Leistikow et al., 2010; Trauner et al.,
2012). Many aspects of the DosR regulon and other molecular
mechanisms involved in the adaptation of Mtb to hypoxic stress,
for example, metabolic rearrangements, remain unclear. With
the label-free, SWATH MS-based absolute quantification strat-
egy developed in this study, we could obtain absolute cellular
concentrations of the majority of Mtb proteins in the Wayne
model and thereby gain insight into these adaptations from a
so far unexplored perspective.
RESULTS AND DISCUSSION
Generation of a Nearly Complete SWATH Assay LibraryAs SWATH MS is a targeted proteomic technique, it requires an
assay library containing the precise coordinates of every peptide
and protein to be quantified (Figure 1) (Schubert et al., 2015). We
built a comprehensive, high-quality fragment ion spectral library
ell Host & Microbe 18, 96–108, July 8, 2015 ª2015 Elsevier Inc. 97
A B C
D E
Figure 2. Global Absolute Cellular Protein Concentrations in Mtb by SWATH MS
(A) Coverage of the Mtb SWATH assay library and the SWATH MS data sets compared to the annotated proteome of Mtb (TubercuList v2.6 R27).
(B) SWATH assay library coverage compared to the entire proteome.
(C) Pearson correlation coefficients of peptide feature intensities in all 24 samples compared against each other. Black boxes highlight comparisons among
replicates.
(D) Linear correlation between absolute concentration and SWATH MS intensity of 30 anchor proteins that were accurately quantified using isotope-labeled
spike-in peptides. The figure is based on the average over all samples.
(E) The cross-validated mean fold error determined using the 30 anchor proteins was stable over all 24 SWATH MS injections. Error bars represent the 95%
confidence interval. The secondary axis shows summed protein abundances (in mg) over all quantified proteins per sample. Startingmaterial was 1 mg total protein
extract.
See also Figures S1 and S2 and Tables S1, S2, and S3.
for Mtb covering 32,344 tryptic peptides representing 3,818 pro-
teins (i.e., 96% of the 3,990 annotated Mtb proteins) (Figures 2A
and 2B). This SWATH assay library is a definitive and portable
resource that can be used by other researchers for SWATH
MS analysis of Mtb on a TripleTOF or any other mass spectrom-
eter that produces similar peptide fragmentation patterns
(Toprak et al., 2014). It is available as a public resource in the
SWATHAtlas database (www.SWATHAtlas.org).
Consistent, High-Coverage ProteomeProfiling ofMtb bySWATH MSTo studyMtb proteome composition and dynamics during expo-
nential growth, hypoxia-induced dormancy, and resuscitation,
we grew Mtb H37Rv cultures in triplicates under gradually
decreasing oxygen levels using the Wayne model (Wayne and
Hayes, 1996). We harvested cultures before the stress was initi-
ated (i.e., during exponential growth [day 0]) at three time points
during hypoxia (day 5, day 10, and day 20) and at two time points
after re-aeration, which we started at day 20 (day 20 + 6 hr and
day 20 + 48 hr) (Figure 1). In this study, we focused on the intra-
98 Cell Host & Microbe 18, 96–108, July 8, 2015 ª2015 Elsevier Inc.
cellular proteome composition, and all measurements were thus
performed onwhole-cell lysates, while culture supernatants con-
taining secreted proteins were discarded. Of note, some bacilli
may fail to adapt to gradual oxygen depletion while contributing
to the overall protein biomass analyzed, therefore accounting as
a source of inaccuracy and potentially masking or attenuating re-
sponses of the responsive cells.
We acquired 18 samples (triplicates of six time points) in
SWATH MS mode on a TripleTOF mass spectrometer, with
one set of biological replicates injected twice to assess technical
reproducibility of the method. Overall, in the 24 SWATH MS in-
jections, we confidently and reproducibly quantified 2,458 Mtb
proteins (Table S1). This corresponds to 62% coverage of the
annotated proteome and to �80% coverage of the expressed
proteome reported to date (Kelkar et al., 2011; Schubert et al.,
2013). Over the entire data matrix, 93% of all protein intensities
were obtained by direct quantification while the remaining 7%
were inferred from background values and represent maximally
possible signal intensities (Figures S1A and S1B). We found
that the SWATH MS data was highly reproducible with a median
A B
C
Figure 3. Protein Abundance Range and
Biomass Distribution across Functional Classes
(A) Abundance range of all absolutely quantified pro-
teins in Mtb averaged over all samples.
(B) Biomass contributions of functional classes.
Fractional proteome contributions were calculated by
summing up protein copies per cell for all proteins
present in a given functional class and dividing by the
sum of all quantified proteins in a given sample.
(C) Absolute cellular protein concentrations of
specific functional classes (red) as compared to all
other proteins of which absolute concentrations were
estimated (gray).
See also Figure S3.
Pearson correlation coefficient of R = 0.95 between biological
replicates and R = 0.97 between technical replicates (Figure 2C).
In summary, the SWATH MS technique, together with the com-
plete, high-quality SWATH assay library, allowed us to achieve
accurate and consistent quantification of 2,458 proteins over
many samples of Mtb during exponential growth, hypoxia-
induced dormancy and resuscitation.
Proteome-Wide Estimates of Absolute Cellular ProteinConcentrations by SWATH MSPrincipally, SWATHMSmeasurements are used to quantify how
protein abundances change between samples. Our goal was
to extend this relative quantification by developing a method
to estimate absolute cellular concentrations of all measured
proteins (i.e., on a proteome-wide scale) by SWATH MS. We
implemented a strategy based on the linear correlation between
summed fragment ion MS intensities and actual protein concen-
trations (Ludwig et al., 2012). This linear correlation was estab-
lished using 30 selected anchor proteins, which we accurately
quantified in ‘‘fmol per mg of protein extract’’ using stable-
isotope-labeled synthetic peptide standards (Table S2). The
concentrations of these anchor proteins span almost the entire
abundance range of the Mtb proteome (Figure 2D). Using
cross-validated model selection (Rosenberger et al., 2014), we
found that summed intensities of the five most intense fragment
ions of the three most intense peptides per protein represent
actual protein concentrations most accurately (Figure S2A).
We determined repeatedly a 1.5-fold average error of the esti-
mated anchor protein abundances over all 24 samples using
Monte Carlo cross-validation (Figure 2E). The obtained errors
were normally distributed (Figure S2B) with an average 95%
confidence interval of 0.3 (error bars in Figure 2E). Our approach
could be validated on a completely independent set of 48 human
proteins (UPS2 standard) spiked into Mtb whole-cell lysates
(Figure S2C).
Cell Host & Microbe 1
Subsequently, we applied the method to
estimate absolute concentrations of all pro-
teins identified by at least two peptides in
the SWATH MS data set. This resulted in
2,023 absolutely quantified Mtb proteins
over the entire hypoxic time course (Fig-
ure 2A; Table S3). Having quantified a large
fraction of the expressed proteome, all pro-
tein abundances should sum up to the total
amount of protein that the analysis was started with. Indeed,
the average amount of 0.8 mg total protein mass, calculated
from the sum of all quantified proteins (considering the specific
molecular weight of each protein), approximated well the 1 mg
protein mass that was injected into the mass spectrometer (Fig-
ure 2E). The validity of our strategy was further supported by
good correlations between absolute abundance estimates and
expected stoichiometry of members of several well-studied
multi-protein complexes (Figure S2D). Finally, we converted
the absolute protein concentration unit ‘‘fmol per mg of extracted
protein’’ into ‘‘protein copies per cell’’ by assuming a cellular pro-
tein density value of 0.2 g/ml (Milo, 2013) and an average cellular
volume of 0.5 mm3. Further, we assumed that protein extraction
was complete and that protein density and cell volume did not
change significantly between cells in exponential growth and
dormant state.
In conclusion, we established a SWATH-based approach to
estimate absolute protein abundances on a proteome-wide
scale and thereby we were able to acquire an unprecedented
data set of absolute protein copies per cell for over 2,000 Mtb
proteins over a large number of samples.
Protein Concentration Range and Protein BiomassCompositionGlobal absolute protein abundance estimates reveal that Mtb
protein concentrations span at least four orders of magnitude,
with lowest concentrations below 10 copies per cell andmaximal
concentrations at about 100,000 copies per cell (Figure 3A). The
median abundance of all Mtb proteins was 215 copies per cell.
The most abundant proteins were the chaperonines GroES
and GroEL2; the acyl carrier AcpM, which is involved in mycolic
acid biosynthesis; and the translation elongation factor Tuf.
Proteins essential for optimal growth (Griffin et al., 2011) were
on average significantly more abundant than all other proteins,
not only for exponentially growing bacilli (day 0) but also during
8, 96–108, July 8, 2015 ª2015 Elsevier Inc. 99
A
0
1
2
3
4
5
0 1 2 3 4 5
Log1
0 (B
CG
pro
tein
cop
ies/
cell)
Log10 (Mtb protein copies/cell)
0
1000
2000
3000
4000
5000
6000
Prot
ein
copi
es/c
ell Mtb
BCG
B CIsocitrate lyase (Icl1) DosR regulon
0
50000
100000
150000
200000
250000
300000
Pro
tein
cop
ies/
cell
Mtb BCG
Figure 4. Absolute Quantitative Comparison of Mtb and BCG
(A) Correlation of absolute protein abundances (log10) obtained for Mtb and BCG during exponential growth (day 0). Generally, the two strains showed a high
correlation in exponential growth (R2 = 0.76) as well as during hypoxia (data not shown).
(B) Protein levels of the DosR regulon.
(C) Protein levels of the enzyme isocitrate lyase (Icl1).
Error bars indicate SD of three biological replicates for Mtb. For BCG, no biological replicates were measured.
dormancy (Figure S3A). In general, the extent of protein regula-
tion appears independent of protein abundance (Figure S3B).
We next analyzed the protein biomass share in different func-
tional classes of Mtb during exponential growth, hypoxia-
induced dormancy, and resuscitation. During exponential
growth, ribosomal proteins constituted only 7% of the total pro-
tein biomass (Figure 3B), which is low compared to the >40%
ribosome fraction of rapidly growing E. coli (Li et al., 2014).
Consistent with the extraordinarily high lipid content of Mtb,
the protein biomass share of lipid metabolism was almost 12%
(Figure 3B). Proteins associated with regulatory functions were,
as expected, generally of low abundance with a median of 174
copies per cell (Figures 3B and 3C). Despite the dramatic
changes in environmental conditions and the following growth
arrest, we observed surprisingly small changes in the overall pro-
tein biomass share of functional classes between exponential
growth and dormancy (Figures 3B and 3C). Ribosomal proteins,
for example, which are clearly associated with growth and repli-
cation, were only insignificantly reduced during dormancy. A
striking exception is the DosR regulon. Most of its approximately
50 proteins were highly induced in response to hypoxic stress,
and together, they accounted for almost 20% of total protein
biomass during dormancy, while their initial protein biomass
share during exponential growth was below 1% (Figures 3B
and 3C). In dormant Mtb, the DosR-regulated stress proteins
HspX and TB31.7 became some of the most abundant proteins
of the bacterial cell, with 90,000 and 30,000 copies per cell,
respectively (Figure 3C). The quantitative extent of DosR regulon
induction on protein level and the stability of protein levels over
almost 3 weeks has not been shown before in the Wayne model
of hypoxia.
An alternative model of hypoxia suggests a rather moderate
role of the DosR regulon during prolonged hypoxia (Rustad
et al., 2008). The striking dominance in absolute protein abun-
dance in our experiments indicates, however, that DosR mem-
bers may be potentially interesting latency-specific antigens
for vaccine development (Andersen and Kaufmann, 2014;
100 Cell Host & Microbe 18, 96–108, July 8, 2015 ª2015 Elsevier Inc
Leyten et al., 2006). To explore this aspect further, we recorded
a SWATH MS data set of the vaccine strain M. bovis BCG
subjected to the same hypoxic conditions as used for Mtb.
Generally, BCG exhibited similar protein abundances as Mtb
(Figure 4A). Protein levels of the DosR regulon in BCG, however,
were about 2-fold lower during hypoxia than in Mtb (Figure 4B),
although on mRNA level, DosR regulon members were shown to
be induced to the same extent in both strains (Lin et al., 2007).
Another striking discrepancy between Mtb and BCG was the
enzyme isocitrate lyase (Icl1), which was present at 2,000 copies
per cell in exponentially growing BCG as compared to 500
copies per cell in Mtb. Remarkably, protein levels in BCG re-
mained unchanged during hypoxia-induced dormancy, whereas
in Mtb Icl1 was induced to over 4,000 copies per cell (Figure 4C).
Dynamic Remodelling of the Mtb Proteome duringHypoxia-Induced Dormancy and ResuscitationOver the hypoxic stress time course, we found 631 proteins that
changed significantly (p value < 0.01, fold change > 1.5) in at
least one time point compared to exponential growth at day 0.
Two days after re-aeration of the cultures, the vast majority of
proteins (94%) returned back to their state before initiation of
the stress. The temporal profiles of proteins that changed signif-
icantly in response to hypoxic stress showed very distinct
patterns as revealed by unsupervised k-means clustering (Fig-
ure 5A; Table S1). Clusters 1–3 all represent proteins decreasing
in response to stress. Cluster 1 protein levels were decreasing
relatively quickly and returned back to their initial levels within
2 days after stress release. Cluster 2 proteins were reduced
more slowly but also returned to initial levels within 2 days after
re-aeration. Cluster 3 proteins dropped quickly but did not
recover within 2 days after re-aeration. Most of the >200
stress-induced proteins were assigned to a single cluster
(cluster 4), with the vast majority of proteins showing constant
levels between day 5 and day 20 of the hypoxic stress. This
cluster contains 90% of all DosR regulon proteins. Induction of
the more recently described enduring hypoxic response was
.
A B
C
D
Figure 5. Dynamic Remodeling of the Mtb Proteome in Response to Hypoxic Stress
(A) Relative regulation of proteins compared to day 0. Unsupervised k-means clustering was applied to obtain clusters of significantly changing proteins with
similar dynamics.
(B) Intracellular ATP levels in Mtb. Error bars indicate SD of two biological replicates. For day 20 + 6 hr, no replicate was available.
(C) Log2 fold changes compared to day 0 (exponential growth) of the F0F1 ATP synthase.
(D) Log2 fold changes compared to day 0 (exponential growth) of components of the aerobic and anaerobic electron transport chain. All significant changes are
marked with an asterisk (fold change > 1.5, p value < 0.05).
less prominent, with only 43 of the 135 detected proteins (32%)
in clusters 4 and 5 (Rustad et al., 2008). Other strongly (>4-fold)
induced proteins in cluster 4 include alanine dehydrogenase Ald,
which is involved in cell wall synthesis, several proteins involved
in lipid metabolism (FadE5, DesA1/2, Tgs1/4, and Icl1), as well
as the copper stress-related enzymes MymT (copper toxicity
protection) and CsoR (copper-sensitive operon repressor). Clus-
ter 5 is noteworthy because it contains proteins that responded
transiently within 6 hr to re-aeration but returned to pre-aeration
levels within 2 days. Proteins in this cluster include PrpC and
PrpD, two enzymes of the methylcitrate cycle, as well as the
sigma factors SigE and SigB and the transcriptional regulator
ClgR, suggesting their involvement during resuscitation. Cluster
6 contains proteins induced by hypoxia but not returning to
basal levels within 2 days.
The strong induction of the DosR regulon and other proteins in
cluster 4 represents a considerable biosynthetic investment that
has to be raised while the cell loses its primary terminal electron
acceptor, oxygen.We therefore took a closer look at the electron
transport chain and ATP synthase. Unexpectedly, despite the
marked decrease in intracellular ATP levels (Figure 5B), the
F0F1 ATP synthase protein levels did not significantly change
Ce
and remained abundant throughout the experiment (Figure 5C).
These findings underscore previous reports on the requirement
of ATP synthase to maintain ATP homeostasis and viability in
dormant Mtb (Gengenbacher et al., 2010; Leistikow et al.,
2010; Rao et al., 2008). In contrast to the stable ATP synthase
levels, many components of the electron transport chain and
energy-generating machinery changed in abundance, indicating
reorganization of Mtb’s energy metabolism in response to hyp-
oxia (Figure 5D). In summary, 26% of the quantified Mtb proteins
were significantly regulated in response to hypoxic stress with
distinct patterns of dynamic responses.
Coordinated Responses of Connected MetabolicReactionsSince about a quarter of all hypoxia-responding proteins have
metabolic functions, we next focused on this category. We em-
ployed a genome-scale metabolic model to identify coordinated
regulation of biochemically connected enzymes. 80% of all
significantly altered metabolic enzymes (fold change > 1.5, p
value < 0.01) could be assigned to strings of at least four or
more connected enzymes, leading to metabolic subnetworks
that are independent of classical pathway definitions (Figure 6A).
ll Host & Microbe 18, 96–108, July 8, 2015 ª2015 Elsevier Inc. 101
A
B
C
D
E
F
G
Figure 6. Regulation of Metabolic Subnetworks
(A) Network representing connected reactions (path lengthR 4) for which all corresponding enzymes were changing significantly in at least one time point (fold
change R 1.5, p value % 0.01). accoa: acetyl-CoA; malcoa: malonyl-CoA; mmalcoa: methylmalonyl-CoA; ACP: acetyl carrier protein; mqn: menaquinone; glu:
glutamate; ala: alanine; tre: trehalose.
(B–D) Subnetworks were extracted based on coordinated regulation of their constituting proteins in the same direction. Such co-regulated subnetworks were
identified for (B) branched-chain fatty acid metabolism with methylmalonyl-CoA in the center, (C) fatty acid and mycolate biosynthesis, and (D) 2-methylcitrate
cycle. Coloring represents z scores of protein abundances (edges) and metabolite intensities (nodes).
(E–G) Subnetworks illustrating antagonistic temporal regulation, indicating interplay between connected reactions and pathways: (E) medium chain fatty acid
metabolism with malonyl-CoA in the center; (F) alanine, aspartate, and glutamate metabolism; and (G) trehalose sulfate, sulfolipid, and thioredoxin synthesis.
See also Figure S4 (dynamics of the entire network) and Tables S4 and S5.
102 Cell Host & Microbe 18, 96–108, July 8, 2015 ª2015 Elsevier Inc.
As expected for a Gram-positive bacterium that mainly uses
menaquinone as respiratory co-factor, the clustering of menaqui-
nonebiosynthesis enzymes indicates their regulatory connectivity
upon oxygen depletion (Dhiman et al., 2009). Another metabolic
subnetwork contained enzymes of amino acid metabolism,
such as glutamate, aspartate, and alanine. Although alterations
in the metabolism of these amino acids have previously been
mentioned in the context of hypoxic mycobacteria, our genome-
wide data integration reveals their regulatory connectivity upon
oxygen depletion (Hutter andDick, 1998; Shi et al., 2010). Consis-
tent with the prevalent lipid metabolism, we also found subnet-
works of metabolic enzymes around the lipid building blocks
acetyl-CoA, malonyl-CoA, and methylmalonyl-CoA.
Next, we investigated whether the identified metabolic sub-
networks show uniform protein regulation (Figures 6 and S4;
Table S4). Intriguingly, subnetworks found in lipid metabolism
showed distinct protein dynamics during entry into hypoxia
(Figures 6B–6D). While enzymes involved in malonyl-CoA and
methylmalonyl-CoA metabolism were rapidly reduced upon
oxygen depletion (Figure 6B), the cluster dominated by acetyl
carrier proteins reacted slower in response to the decrease of
oxygen (Figure 6C). Strikingly, protein levels of the methylcitrate
cycle peaked at 6 hr after re-aeration (Figure 6D), suggesting
an early role of this pathway during resuscitation before other
metabolic processes are fully operational. The methylcitrate
cycle is involved in the catabolism of propionyl-CoA derived
from branched chain amino acids and lipids and odd-chain fatty
acid degradation (Upton and McKinney, 2007).
In obligate aerobic mycobacteria, hypoxia triggers re-routing
of metabolic fluxes through alternative enzymatic transforma-
tions of a given metabolite (Boshoff and Barry, 2005; Shi et al.,
2010; Watanabe et al., 2011). Hence, antagonistic temporal
profiles of enzymes within our metabolic subnetworks should
indicate such flux changes. We found differential regulation of
enzymes in the subnetworks of acetyl-CoA (Figure 6E), alanine/
glutamate (Figure 6F), and trehalose metabolism (Figure 6G).
Consistently, competing incorporation of acetyl-CoA either into
TCA cycle or triacylglycerol biosynthesis was previously
described for hypoxic tubercle bacilli (Baek et al., 2011).
To obtain further evidence for differential pathway usage,
we correlated abundance changes of enzymes with connected
metabolites, whichweremeasured by untargetedMS in samples
collected from the same cultures as used for proteomic analyses
(Table S5) (Fuhrer et al., 2011). Temporal profiles of some
metabolites correlated with concentration dynamics of adjacent
enzymes—for example, in the methylcitrate cycle and quinone
biosynthesis (Figures 6D and S4). In the case of the methylma-
lonyl-CoA subnetwork, metabolites reacted much faster to
re-aeration than corresponding enzymes (Figure 6B). Methylma-
lonyl-CoA is derived from propionyl-CoA that is presumably
produced after re-aeration leading to the expression of the
detoxifying methylcitrate cycle. Furthermore, it is attractive to
speculate that methylmalonyl-related compounds accumulated
as a thermodynamic consequence of increased intracellular
levels of propionyl-CoA after re-aeration. Subsequently, these
accumulating metabolites could potentially induce expression
of genes in their utilizing subnetwork.
In summary, the high proteome coverage obtained by the
SWATH MS technique and the dynamic nature of this data set
Ce
enabled us to integrate the protein data with a genome-scale
metabolic network. This revealed the regulation of metabolic
genes as functional subnetworks, which was also directly re-
flected in corresponding metabolite levels.
Absolute Protein Concentrations Improve Conclusionson Flux Changes and Protein Resource BalancingRelative enzyme levels are generally used for comparison of
metabolism under different conditions. However, depending on
an enzyme’s basal expression level, a given fold change might
have either an important or a negligible contribution to the global
flux reconfiguration; hence, such relative data are of limited pre-
dictive power. The availability of absolute protein levels allowed
us to assess the maximal reaction rate (Vmax) of an enzyme,
which is critical information to constrain metabolic fluxes in a
genome-scale flux balancing model. We calculated Vmax by
multiplying absolute protein abundances with reported turnover
numbers (kcat) of given enzymes. The relevance of this analysis
can be illustrated with the isoenzymes phosphofructokinases A
and B (PfkA/Rv3010c and PfkB/Rv2029c). Upon hypoxia, PfkB
was induced more than 5-fold to an absolute level comparable
to PfkA, which did not respond to oxygen depletion (Figure 7A).
While these results suggest a major contribution of PfkB to
the glycolytic flux, the very low turnover frequency of PfkB
compared to PfkA (Phong et al., 2013) reveals that the Vmax of
PfkB remains insignificant compared to PfkA (Figure 7A). This
conclusion is consistent with the genetic evidence that PfkB
cannot functionally complement a pfkA-null mutant of Mtb
during hypoxia, despite strong transcriptional induction (Phong
et al., 2013).
Another example highlighting the relevance of absolute pro-
tein abundances to derive the Vmax of a reaction is the bifurca-
tion between glyoxylate shunt and oxidative TCA cycle that is
important for mycobacterial pathogenesis (McKinney et al.,
2000). The branch point is established by the isocitrate lyases
Icl1 and Icl2 (Rv0467 and Rv1915) and the isocitrate dehydroge-
nases Icd1 and Icd2 (Rv3339c andRv0066c) (Figure 7B) (Munoz-
Elıas and McKinney, 2005). Upregulation of only Icl1 upon
hypoxia suggested major functional relevance, but calculated
reaction velocities demonstrate that even with a 6-fold increase
in Icl1 protein levels, the Vmax of Icl1 only matches the Vmax of
Icd2 during hypoxia (Figure 7B; Table S6). Since the Vmax values
suggest that neither Icd1 nor Icl2 carry important carbon fluxes
during hypoxia, again the relevance of the fold change for this
critical branch point could easily be over-interpreted. Despite
the simplified flux regulation model that ignores, for example,
post-translational modifications, the two examples demonstrate
the value of absolute protein concentrations due to their direct
causal link to carbon fluxes.
For a more systematic genome-wide identification of meta-
bolic capacity changes, we sorted all proteins according to
the change to day 0 (day 5, day 10, and day 20) or day 20
(day 20 + 6 hr and day 20 + 48 hr) for the transition into hypoxia
or re-aeration, respectively, in their (i) relative change, (ii) abso-
lute protein abundance changes, or (iii) calculated Vmax
changes (Table S6; if no kcat was reported for Mtb, the median
of all available values in the Brenda database was used instead).
On the set of significantly changing proteins (fold changeR 1.5,
p value % 0.01), we performed metabolic pathway enrichment
ll Host & Microbe 18, 96–108, July 8, 2015 ª2015 Elsevier Inc. 103
A
B
C
Figure 7. Analysis of Maximal Reaction Velocities (Vmax)
(A) Time profiles of isoenzymes PfkA and PfkB: (i) relative protein concentration changes compared to day 0, (ii) estimated absolute protein abundances, and (iii)
estimated Vmax.
(B) Same time profiles as above, but for the isoenzymes Icl1/Icl2 and Icd1/Icd2.
(C) Significantly enrichedmetabolic pathways (p value% 0.1) in the set of significantly decreased or increased proteins. Proteins were sorted by decreasing (i) fold
change, (ii) absolute protein abundance change, or (iii) absolute change in Vmax (reference for day 5, 10, and 20 was day 0; reference for day 20 + 6 hr and day
20 + 48 hr was day 20).
See also Tables S6 and S7.
analogous to gene set enrichment analysis of transcriptomic
data (Figure 7C; Table S7) (Subramanian et al., 2005).
Only three pathways were enriched in the protein list sorted
by relative abundance change, whereas numerous pathways
were enriched in the protein list sorted by absolute abundance
change and Vmax change. In line with the observed regulatory
subnetwork dynamics of metabolic enzymes involved in
lipid metabolism (Figure 6), several lipid pathways were signifi-
cantly enriched (p value < 0.05) in the protein set with large
104 Cell Host & Microbe 18, 96–108, July 8, 2015 ª2015 Elsevier Inc
negative absolute abundance change during the transition
into hypoxia and large positive absolute abundance change
following re-aeration (FASI-III, mycolic acid, and methyl-
branched fatty acid biosynthesis).
This enrichment analysis demonstrates the additional infor-
mation content in absolute protein concentrations compared
to protein fold changes. Biologically, enrichment based on abso-
lute protein changes can be interpreted as a cellular resource
investment for the modulation of protein levels, whereas the
.
Vmax changes can be directly linked to metabolic flux changes
as discussed above. Taken together, even in the absence of spe-
cific enzyme parameters for less well characterized organisms
such as Mtb, more meaningful conclusions on functional
changes are possible from absolute than from relative protein
data. The exemplary analysis of key enzymes in glycolysis and
TCA cycle, and the conceptual analysis on a more global scale,
illustrate the importance of absolute cellular protein concentra-
tions and maximal reaction velocities for a truly quantitative
understanding of an organism’s metabolism.
CONCLUSIONS
Relative abundance changes of proteins are informative for the
comparison of cellular states. They are, however, also limited,
because they do not support quantitative comparisons across
different proteins and have limited value for supporting mathe-
matical models of cellular processes. In this study, we adapted
the proteomic technique SWATH MS to allow proteome-wide
estimation of cellular protein concentrations and show that
absolute concentrations of biomolecules are an important com-
plement for a comprehensive, more detailed, and quantitative
understanding of any cellular system. The method estimates in
a label-free manner absolute cellular concentrations of all
quantified proteins in a SWATH data set. Together with a com-
plete SWATH assay library for Mtb, it enabled us to perform a
system-level analysis of absolute cellular protein concentrations
in Mtb during exponential growth, hypoxia-induced dormancy,
and resuscitation. The resulting molecular inventory of cellular
protein concentrations for 2,023 Mtb proteins under these
clinically highly relevant growth states is complementary to exist-
ing proteomic and transcriptomic data sets and can serve as a
reference for further quantitative analyses and mathematical
modeling, as exemplified by key enzymes in the central carbon
metabolism. The data set transforms our understanding of meta-
bolic adaptations by allowing us to translate effects of protein
regulation into actual changes in maximal enzymatic reaction
velocities.
Knowledge of absolute concentrations of proteins in Mtb
might also assist in the development of treatments and preven-
tive measures to fight tuberculosis. The discovery of highly
upregulated and abundant proteins can support the identifica-
tion of important pathways and processes and indicate where
the system might be vulnerable to drug interventions. More-
over, knowledge of absolute protein concentrations might
help to prioritize immune-dominant target antigens of the
human T cell response for subunit vaccines against tubercu-
losis. To prevent disease outbreak in individuals with latent
tuberculosis, emphasis is now given to so-called multi-stage
vaccines that express both antigens of dormant and replicating
bacilli (Andersen and Kaufmann, 2014). The here-described
method and resulting resource of Mtb protein concentrations
during dormancy and resuscitation will facilitate identification
of appropriate antigens for rational design of future multi-stage
vaccines.
EXPERIMENTAL PROCEDURES
Detailed descriptions are given in the Supplemental Experimental Procedures.
Ce
Reference Proteome
As a reference annotation of theMtb proteome, we used TubercuList v2.6 R27.
After removal of 41 identical protein sequences, the protein database con-
tained 3,990 proteins.
Bacterial Cultures
Wayne experiments in Mtb H37Rv and M. bovis BCG were performed as
described earlier (Wayne and Hayes, 1996). Samples were taken before the
start of the experiment in aerobic cultures (day 0) and at 5, 10, and 20 days
of the hypoxic time course (day 5, day 10, and day 20). After sampling at
day 20, remaining glass vials were opened, and cultures were transferred
into new vessels for aerated incubation without addition of new culture
medium. Further time points were sampled from cultures re-aerated for 6 or
48 hr (day 20 + 6 hr and day 20 + 48 hr).
Proteomics Sample Preparation
After protein extraction and digest into peptides, 51 isotopically labeled
synthetic reference peptides (AQUA QuantPro, Thermo Fisher Scientific) for
absolute quantification of 30 anchor proteins and 11 iRT retention time normal-
ization peptides (RT-kit WR, Biognosys) (Escher et al., 2012) were added to
every sample.
SWATH Assay Library Generation
As an input for the SWATH assay library, the following samples were acquired
in data-dependent acquisition mode on a TripleTOF 5600 mass spectrometer
(AB Sciex): (i) 19 pools of �1,000 crude synthetic peptides each (Schubert
et al., 2013), (ii) 24 off-gel electrophoresis fractions of an Mtb lysate consisting
of bacteria from exponential and stationary phase cultures (Schubert et al.,
2013), and (iii) 12 unfractionated whole-cell lysates of Mtb and M. bovis BCG
during hypoxic stress. To each of the 55 samples, iRT peptides were added.
The resulting spectral library was used to build a comprehensive high-quality
SWATH assay library (Schubert et al., 2015).
SWATH MS Analysis
All samples of the hypoxic time course were measured in SWATH mode on
a TripleTOF 5600 mass spectrometer using data-independent acquisition
settings described earlier (Gillet et al., 2012). Resulting data was analyzed
using the software OpenSWATH (Rost et al., 2014), including run-to-run
alignment and extraction of representative background signal in case a
feature could not be identified with high confidence in a particular sample.
Due to chromatographic issues caused by column overloading observed
for the most abundant proteins, the 21 most abundant proteins were manually
requantified (http://panoramaweb.org/labkey/Mtb_requant.url) using the soft-
ware Skyline (MacLean et al., 2010). To obtain fold changes and correspond-
ing p values, the software MSstats was used (Choi et al., 2014).
Absolute Protein Abundance Estimations
To derive absolute protein abundance estimates, 30 anchor proteins were
quantified using 51 synthetic isotopically labeled reference peptides that
were spiked into all samples during sample preparation at concentrations
similar to the endogenous peptides (Table S2). To ensure accurate absolute
quantitation, a dilution series experiment was performed for each synthetic
reference peptide (http://panoramaweb.org/labkey/Mtb_dilutions.url). Abso-
lute concentrations of all anchor proteins were obtained by manual analysis
(http://panoramaweb.org/labkey/Mtb_anchors.url). The optimal model to
combine label-free SWATH MS intensities to a single protein MS signal was
determined by Monte Carlo cross-validation using the aLFQ R package
(Rosenberger et al., 2014), and the model with highest accuracy (summing
the five most intense transitions of the threemost intense peptides per protein)
was used to estimate proteome-wide concentrations of all proteins quantified
with two or more peptides.
Metabolic Profiling
To determine intracellular ATP levels, bacilli were sampled by fast filtration
(Watanabe et al., 2011) and metabolites were extracted in chloroform:metha-
nol (2:1). ATP was quantified by LC-MS/MS on a triple quadrupole tandem
mass spectrometer (Buescher et al., 2010). Untargeted measurements of me-
tabolites were performed by direct injection MS using a quadrupole-coupled
ll Host & Microbe 18, 96–108, July 8, 2015 ª2015 Elsevier Inc. 105
time of flight instrument (Agilent 6550) (Fuhrer et al., 2011). A genome scale
metabolic model of Mtb was used to compile a metabolite reference list,
according to which ions were assigned to metabolites applying a mass toler-
ance of 1 mDa and an intensity cut-off of 1,500 counts.
Subnetwork Analysis
Basis for the subnetwork analysis was a genome-scale metabolic model of
Mtb. Only reactions of which at least one enzyme was changing significantly
(fold change R 1.5, p value % 0.01) were considered. Two reactions that
share a metabolite were considered as connected. H2O, CO2, and common
cofactors were not considered as connecting metabolites (Table S4). Shortest
path search (Yen, 1971) was performed between each two reactions in the
resulting reaction-reaction network, and those with length R4 were kept.
Network edges were colored according to protein abundances (x) normalized
to the mean values across conditions with the Matlab z score function
(z score= ðx�meanðxÞÞ=stdðxÞ). Subnetworks were assigned to metabolic
classes according to TBDB annotations (www.tbdb.org) and manually group-
ed according to their temporal profiles. Metabolites involved in each reaction
were assigned to the corresponding enzymes, and Pearson correlation
coefficients were calculated for the z scores with time shifts ±1 to capture
delayed dependencies.
Maximal Enzymatic Reaction Velocity Estimation
and Enrichment Analysis
Maximal enzymatic reaction velocity (Vmax) was estimated per definition as
absolute protein abundance multiplied by enzyme turnover frequency (kcat)
obtained from the Brenda enzyme database (www.brenda-enzymes.org).
Median values of all reported kcats (or only those reported for Mtb, if available)
were taken (Table S6). If no kcat was available, 1 was set as default value. For
the PfkA and PfkB isoenzymes, kcat (PfkA = 0.015 s�1, PfkB = 0.001 s�1) was
calculated based on Mtb-specific activity values (PfkA = 25.0 [nmol min�1
purified protein mg�1], PfkB = 1.7 [nmol min�1 purified protein mg�1]) (Phong
et al., 2013) and enzyme molar masses (PfkA = 36,879.99 g/mol, pfkB =
35,401.37 g/mol).
Pathway enrichment was performed for significantly changing proteins
according to pathway definitions in TBDB (www.tbdb.org). The gene set
enrichment analysis method from Subramanian et al. (2005) was adapted for
proteins. All significantly changing proteins were sorted by one of the following
characteristics: fold change, absolute protein concentration change, and
Vmax change. Subsequently, a hypergeometric test was applied for each
subset of size 1 to N (N is the number of all significantly changing proteins).
For each pathway, the lowest adjusted p value from tests of different sizes
was taken.
Data Availability
The Mtb SWATH assay library is available through the SWATHAtlas data-
base (www.SWATHAtlas.org). All SWATH MS raw data has been made
available through the PeptideAtlas database (www.peptideatlas.org/PASS/
PASS00655). Manual analyses of selected peptides and proteins can be
visualized in and downloaded from Panorama (https://panoramaweb.org;
see Experimental Procedures for direct links).
SUPPLEMENTAL INFORMATION
Supplemental Information includes four figures, seven tables, and Supple-
mental Experimental Procedures and can be found with this article online at
http://dx.doi.org/10.1016/j.chom.2015.06.001.
AUTHOR CONTRIBUTIONS
O.T.S. and C.L. designed the study, developed methods, and performed pro-
teomic analyses. M.K. and M.Z. performed metabolomic analyses and inte-
grated proteomic and metabolomic data sets. M.G. performed Mtb and
BCG experiments. G.R. and H.L.R. provided support for aLFQ and Open-
SWATH analyses, respectively. L.C.G. and B.C.C. provided support for
mass spectrometric measurements. O.T.S., C.L., M.K., and M.Z. wrote the
manuscript. R.A., U.S., M.G., and S.H.E.K. discussed the manuscript and pro-
vided important feedback. R.A. supervised the project.
106 Cell Host & Microbe 18, 96–108, July 8, 2015 ª2015 Elsevier Inc
ACKNOWLEDGMENTS
We thank Mary Louise Grossman for critically reading the manuscript and
Doris Lazar and Frank Siejak for technical support. We furthermore thank
David Campbell for support of the SWATHAtlas database and Dennis Vitkup
for his valuable contribution to the genome-scale metabolic model. This
work was financially supported by the SysteMTb Collaborative Project (grant
No. 241587, Framework Programme 7 of the European Commission), the
ERC advanced grant ‘‘Proteomics v3.0’’ (grant No. 233226), and TbX, which
is part of the Swiss initiative for systems biology, SystemsX.ch.
Received: February 1, 2015
Revised: April 20, 2015
Accepted: May 18, 2015
Published: June 18, 2015
REFERENCES
Aebersold, R., and Mann, M. (2003). Mass spectrometry-based proteomics.
Nature 422, 198–207.
Ahrne, E., Molzahn, L., Glatter, T., and Schmidt, A. (2013). Critical assessment
of proteome-wide label-free absolute abundance estimation strategies.
Proteomics 13, 2567–2578.
Andersen, P., and Kaufmann, S.H.E. (2014). Novel vaccination strategies
against tuberculosis. Cold Spring Harb. Perspect. Med. 4, http://dx.doi.org/
10.1101/cshperspect.a018523.
Baek, S.-H., Li, A.H., and Sassetti, C.M. (2011). Metabolic regulation of myco-
bacterial growth and antibiotic sensitivity. PLoS Biol. 9, e1001065.
Boon, C., and Dick, T. (2002). Mycobacterium bovis BCG response regulator
essential for hypoxic dormancy. J. Bacteriol. 184, 6760–6767.
Boshoff, H.I.M., and Barry, C.E., 3rd. (2005). Tuberculosis - metabolism and
respiration in the absence of growth. Nat. Rev. Microbiol. 3, 70–80.
Buescher, J.M., Moco, S., Sauer, U., and Zamboni, N. (2010). Ultrahigh perfor-
mance liquid chromatography-tandem mass spectrometry method for fast
and robust quantification of anionic and aromatic metabolites. Anal. Chem.
82, 4403–4412.
Choi, M., Chang, C.-Y., Clough, T., Broudy, D., Killeen, T., MacLean, B., and
Vitek, O. (2014). MSstats: an R package for statistical analysis of quantitative
mass spectrometry-based proteomic experiments. Bioinformatics 30, 2524–
2526.
Desiderio, D.M., and Kai, M. (1983). Preparation of stable isotope-incorpo-
rated peptide internal standards for field desorption mass spectrometry
quantification of peptides in biologic tissue. Biomed. Mass Spectrom. 10,
471–479.
Dhiman, R.K., Mahapatra, S., Slayden, R.A., Boyne, M.E., Lenaerts, A.,
Hinshaw, J.C., Angala, S.K., Chatterjee, D., Biswas, K., Narayanasamy, P.,
et al. (2009). Menaquinone synthesis is critical for maintaining mycobacterial
viability during exponential growth and recovery from non-replicating persis-
tence. Mol. Microbiol. 72, 85–97.
Domon, B., and Aebersold, R. (2010). Options and considerations when
selecting a quantitative proteomics strategy. Nat. Biotechnol. 28, 710–721.
Escher, C., Reiter, L., MacLean, B., Ossola, R., Herzog, F., Chilton, J.,
MacCoss, M.J., and Rinner, O. (2012). Using iRT, a normalized retention
time for more targeted measurement of peptides. Proteomics 12, 1111–1121.
Fuhrer, T., Heer, D., Begemann, B., and Zamboni, N. (2011). High-throughput,
accurate massmetabolome profiling of cellular extracts by flow injection-time-
of-flight mass spectrometry. Anal. Chem. 83, 7074–7080.
Gengenbacher, M., and Kaufmann, S.H.E. (2012). Mycobacterium tubercu-
losis: success through dormancy. FEMS Microbiol. Rev. 36, 514–532.
Gengenbacher, M., Rao, S.P.S., Pethe, K., and Dick, T. (2010). Nutrient-
starved, non-replicating Mycobacterium tuberculosis requires respiration,
ATP synthase and isocitrate lyase for maintenance of ATP homeostasis and
viability. Microbiology 156, 81–87.
.
Gerber, S.A., Rush, J., Stemman, O., Kirschner, M.W., and Gygi, S.P. (2003).
Absolute quantification of proteins and phosphoproteins from cell lysates by
tandem MS. Proc. Natl. Acad. Sci. USA 100, 6940–6945.
Gillet, L.C., Navarro, P., Tate, S., Rost, H.L., Selevsek, N., Reiter, L., Bonner,
R., and Aebersold, R. (2012). Targeted data extraction of the MS/MS spectra
generated by data-independent acquisition: a new concept for consistent and
accurate proteome analysis. Mol. Cell Proteomics 11, http://dx.doi.org/10.
1074/mcp.O111.016717.
Griffin, J.E., Gawronski, J.D., Dejesus, M.A., Ioerger, T.R., Akerley, B.J., and
Sassetti, C.M. (2011). High-resolution phenotypic profiling defines genes
essential for mycobacterial growth and cholesterol catabolism. PLoS
Pathog. 7, e1002251.
Hutter, B., and Dick, T. (1998). Increased alanine dehydrogenase activity dur-
ing dormancy in Mycobacterium smegmatis. FEMSMicrobiol. Lett. 167, 7–11.
Kelkar, D.S., Kumar, D., Kumar, P., Balakrishnan, L., Muthusamy, B., Yadav,
A.K., Shrivastava, P., Marimuthu, A., Anand, S., Sundaram, H., et al. (2011).
Proteogenomic analysis of Mycobacterium tuberculosis by high resolution
mass spectrometry. Mol. Cell Proteomics 10, http://dx.doi.org/10.1074/mcp.
M111.011445.
Leistikow, R.L., Morton, R.A., Bartek, I.L., Frimpong, I., Wagner, K., and
Voskuil, M.I. (2010). The Mycobacterium tuberculosis DosR regulon assists
in metabolic homeostasis and enables rapid recovery from nonrespiring
dormancy. J. Bacteriol. 192, 1662–1670.
Leyten, E.M.S., Lin, M.Y., Franken, K.L.M.C., Friggen, A.H., Prins, C., van
Meijgaarden, K.E., Voskuil, M.I., Weldingh, K., Andersen, P., Schoolnik,
G.K., et al. (2006). Human T-cell responses to 25 novel antigens encoded by
genes of the dormancy regulon of Mycobacterium tuberculosis. Microbes
Infect. 8, 2052–2060.
Li, G.-W., Burkhardt, D., Gross, C., and Weissman, J.S. (2014). Quantifying
absolute protein synthesis rates reveals principles underlying allocation of
cellular resources. Cell 157, 624–635.
Lin, M.Y., Geluk, A., Smith, S.G., Stewart, A.L., Friggen, A.H., Franken,
K.L.M.C., Verduyn, M.J.C., van Meijgaarden, K.E., Voskuil, M.I., Dockrell,
H.M., et al. (2007). Lack of immune responses to Mycobacterium tuberculosis
DosR regulon proteins following Mycobacterium bovis BCG vaccination.
Infect. Immun. 75, 3523–3530.
Ludwig, C., Claassen, M., Schmidt, A., and Aebersold, R. (2012). Estimation
of absolute protein quantities of unlabeled samples by selected reaction moni-
toring mass spectrometry. Mol. Cell Proteomics 11, http://dx.doi.org/10.1074/
mcp.M111.013987.
MacLean, B., Tomazela, D.M., Shulman, N., Chambers, M., Finney, G.L.,
Frewen, B., Kern, R., Tabb, D.L., Liebler, D.C., and MacCoss, M.J. (2010).
Skyline: an open source document editor for creating and analyzing targeted
proteomics experiments. Bioinformatics 26, 966–968.
Mallick, P., Schirle, M., Chen, S.S., Flory, M.R., Lee, H., Martin, D., Ranish, J.,
Raught, B., Schmitt, R., Werner, T., et al. (2007). Computational prediction
of proteotypic peptides for quantitative proteomics. Nat. Biotechnol. 25,
125–131.
McKinney, J.D., Honer zu Bentrup, K., Munoz-Elıas, E.J., Miczak, A., Chen, B.,
Chan, W.T., Swenson, D., Sacchettini, J.C., Jacobs, W.R., Jr., and Russell,
D.G. (2000). Persistence of Mycobacterium tuberculosis in macrophages
and mice requires the glyoxylate shunt enzyme isocitrate lyase. Nature 406,
735–738.
Milo, R. (2013). What is the total number of protein molecules per cell volume?
A call to rethink some published values. Bioessays 35, 1050–1055.
Munoz-Elıas, E.J., and McKinney, J.D. (2005). Mycobacterium tuberculosis
isocitrate lyases 1 and 2 are jointly required for in vivo growth and virulence.
Nat. Med. 11, 638–644.
Park, H.-D., Guinn, K.M., Harrell, M.I., Liao, R., Voskuil, M.I., Tompa, M.,
Schoolnik, G.K., and Sherman, D.R. (2003). Rv3133c/dosR is a transcription
factor that mediates the hypoxic response of Mycobacterium tuberculosis.
Mol. Microbiol. 48, 833–843.
Phong, W.Y., Lin, W., Rao, S.P.S., Dick, T., Alonso, S., and Pethe, K. (2013).
Characterization of phosphofructokinase activity in Mycobacterium tubercu-
Ce
losis reveals that a functional glycolytic carbon flow is necessary to limit the
accumulation of toxic metabolic intermediates under hypoxia. PLoS ONE 8,
e56037.
Picotti, P., Aebersold, R., and Domon, B. (2007). The implications of proteo-
lytic background for shotgun proteomics. Mol. Cell. Proteomics 6, 1589–
1598.
Rao, S.P.S., Alonso, S., Rand, L., Dick, T., and Pethe, K. (2008). The protonmo-
tive force is required for maintaining ATP homeostasis and viability of hypoxic,
nonreplicating Mycobacterium tuberculosis. Proc. Natl. Acad. Sci. USA 105,
11945–11950.
Rosenberger, G., Ludwig, C., Rost, H.L., Aebersold, R., and Malmstrom, L.
(2014). aLFQ: An R-package for estimating absolute protein quantities from
label-free LC-MS/MS proteomics data. Bioinformatics 30, http://dx.doi.org/
10.1093/bioinformatics/btu200.
Rost, H.L., Rosenberger, G., Navarro, P., Gillet, L., Miladinovi�c, S.M.,
Schubert, O.T., Wolski, W., Collins, B.C., Malmstrom, J.A., Malmstrom, L.,
and Aebersold, R. (2014). OpenSWATH enables automated, targeted analysis
of data-independent acquisition MS data. Nat. Biotechnol. 32, 219–223.
Rustad, T.R., Harrell, M.I., Liao, R., and Sherman, D.R. (2008). The enduring
hypoxic response of Mycobacterium tuberculosis. PLoS One 3, http://dx.
doi.org/10.1371/journal.pone.0001502.
Schubert, O.T., Mouritsen, J., Ludwig, C., Rost, H.L., Rosenberger, G., Arthur,
P.K., Claassen, M., Campbell, D.S., Sun, Z., Farrah, T., et al. (2013). The Mtb
proteome library: a resource of assays to quantify the complete proteome of
Mycobacterium tuberculosis. Cell Host Microbe 13, 602–612.
Schubert, O.T., Gillet, L.C., Collins, B.C., Navarro, P., Rosenberger, G.,Wolski,
W.E., Lam, H., Amodei, D., Mallick, P., MacLean, B., and Aebersold, R. (2015).
Building high-quality assay libraries for targeted analysis of SWATH MS data.
Nat. Protoc. 10, 426–441.
Schwanhausser, B., Busse, D., Li, N., Dittmar, G., Schuchhardt, J., Wolf, J.,
Chen, W., and Selbach, M. (2011). Global quantification of mammalian gene
expression control. Nature 473, 337–342.
Sherman, D.R., Voskuil, M., Schnappinger, D., Liao, R., Harrell, M.I., and
Schoolnik, G.K. (2001). Regulation of theMycobacterium tuberculosis hypoxic
response gene encoding alpha -crystallin. Proc. Natl. Acad. Sci. USA 98,
7534–7539.
Shi, L., Sohaskey, C.D., Pfeiffer, C., Datta, P., Parks, M., McFadden, J., North,
R.J., and Gennaro, M.L. (2010). Carbon flux rerouting during Mycobacterium
tuberculosis growth arrest. Mol. Microbiol. 78, 1199–1215.
Silva, J.C., Gorenstein, M.V., Li, G.-Z., Vissers, J.P.C., and Geromanos, S.J.
(2006). Absolute quantification of proteins by LCMSE: a virtue of parallel MS
acquisition. Mol. Cell. Proteomics 5, 144–156.
Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L.,
Gillette, M.A., Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., and
Mesirov, J.P. (2005). Gene set enrichment analysis: a knowledge-based
approach for interpreting genome-wide expression profiles. Proc. Natl.
Acad. Sci. USA 102, 15545–15550.
Taniguchi, Y., Choi, P.J., Li, G.W., Chen, H., Babu, M., Hearn, J., Emili, A., and
Xie, X.S. (2010). Quantifying E. coli proteome and transcriptome with single-
molecule sensitivity in single cells. Science 329, 533–538.
Toprak, U.H., Gillet, L.C., Maiolica, A., Navarro, P., Leitner, A., and Aebersold,
R. (2014). Conserved peptide fragmentation as a benchmarking tool for mass
spectrometers and a discriminating feature for targeted proteomics. Mol. Cell.
Proteomics 13, 2056–2071.
Trauner, A., Lougheed, K.E.A., Bennett, M.H., Hingley-Wilson, S.M., and
Williams, H.D. (2012). The dormancy regulator DosR controls ribosome stabil-
ity in hypoxic mycobacteria. J. Biol. Chem. 287, 24053–24063.
Upton, A.M., and McKinney, J.D. (2007). Role of the methylcitrate cycle in pro-
pionate metabolism and detoxification in Mycobacterium smegmatis.
Microbiology 153, 3973–3982.
Via, L.E., Lin, P.L., Ray, S.M., Carrillo, J., Allen, S.S., Eum, S.-Y., Taylor, K.,
Klein, E., Manjunatha, U., Gonzales, J., et al. (2008). Tuberculous granulomas
are hypoxic in guinea pigs, rabbits, and nonhuman primates. Infect. Immun.
76, 2333–2340.
ll Host & Microbe 18, 96–108, July 8, 2015 ª2015 Elsevier Inc. 107
Voskuil, M.I., Schnappinger, D., Visconti, K.C., Harrell, M.I., Dolganov, G.M.,
Sherman, D.R., and Schoolnik, G.K. (2003). Inhibition of respiration by nitric
oxide induces a Mycobacterium tuberculosis dormancy program. J. Exp.
Med. 198, 705–713.
Watanabe, S., Zimmermann, M., Goodwin, M.B., Sauer, U., Barry, C.E., 3rd,
and Boshoff, H.I. (2011). Fumarate reductase activity maintains an energized
membrane in anaerobic Mycobacterium tuberculosis. PLoS Pathog. 7,
e1002287.
108 Cell Host & Microbe 18, 96–108, July 8, 2015 ª2015 Elsevier Inc
Wayne, L.G., and Hayes, L.G. (1996). An in vitro model for sequential study of
shiftdown of Mycobacterium tuberculosis through two stages of nonreplicat-
ing persistence. Infect. Immun. 64, 2062–2069.
Wayne, L.G., and Sohaskey, C.D. (2001). Nonreplicating persistence of myco-
bacterium tuberculosis. Annu. Rev. Microbiol. 55, 139–163.
Yen, J.Y. (1971). Finding the KShortest Loopless Paths in a Network. Manage.
Sci. 17, 712–716.
.
