25. Pisarchick, M. L., Gesty, D. & Thompson, N. L. Binding
kinetics of an anti-dinitrophenyl monoclonal Fab on
supported phospholipid monolayers measured by total
internal reflection with fluorescence photobleaching
recovery. Biophys. J. 63, 215–223 (1992).
26. Chang, P. S., Axelrod, D., Omann, G. M. &
Linderman, J. J. G protein threshold behavior in the
human neutrophil oxidant response: measurement of
G proteins available for signaling in responding and
nonresponding subpopulations. Cell. Signal. 17,
27. Mattheyses, A. L., Hoppe, A. & Axelrod, D. Polarized
fluorescence resonance energy transfer microscopy.
Biophys. J. 87, 2787–2797 (2004).
28. Lippincott-Schwartz, J., Snapp, E. & Kenworthy, A.
Studying protein dynamics in living cells. Nature Rev.
Mol. Cell Biol. 2, 444–456 (2001).
29. Gaus, K., Zech, T. & Harder, T. Visualizing membrane
microdomains by Laurdan 2-photon microscopy. Mol.
Memb. Biol. 23, 41–48 (2006).
30. Lagerholm, B. C., Weinreb, G. E., Jacobson, K. &
Thompson, N. L. Detecting microdomains in intact
cells. Annu. Rev. Phys. Chem. 56, 309–336 (2005).
31. Kenworthy, A. K., Nichols, B. J., Remmert, C. L.,
Hendrix, G. M., Kumar, M., Zimmerberg, J. &
Lippincott-Schwartz, J. Dynamics of putative raft-
associated proteins at the cell surface. J. Cell Biol.
165, 735–746 (2004).
32. Rao, M. & Mayor, S. Use of Forster resonance energy
transfer microscopy to study lipid rafts. Biochim.
Biophys. Acta 1746, 221–233 (2005).
33. Scalettar, B. A. How neurosecretory vesicles release
their cargo. Neuroscientist 12, 164–176 (2006).
34. Allersma, M. W., Bittner, M. A., Axelrod, D. &
Holz, R. W. Motion matters: secretory granule motion
adjacent to the plasma membrane and exocytosis.
Mol. Biol. Cell 17, 2424–2438 (2006).
35. Santangelo, P., Nitin, N. & Bao, G. Nanostructured
probes for RNA detection in living cells. Annals
Biomed. Eng. 34, 39–50 (2006).
36. Dirks, R. W. & Tanke, H. J. Advances in fluorescent
tracking of nucleic acids in living cells. Biotechniques
40, 489–496 (2006).
37. Brown, D. Imaging protein trafficking. Nephron. Exp.
Nephrol. 103, e55–e61 (2006).
38. Kiyokawa, E., Hara, S., Nakamura, T. & Matsuda, M.
Fluorescence (Forster) resonance energy transfer
imaging of oncogene activity in living cells. Cancer Sci.
97, 8–15 (2006).
39. Zaccolo, M., Cesetti, T., Di Benedetto, G., Mongillo,
M., Lissandron, V., Terrin, A. & Zamparo, I. Imaging
the cAMP-dependent signal transduction pathway.
Biochem. Soc. Trans. 33, 1323–1326 (2005).
40. Bai, L., Santangelo, T. J. & Wang, M. D. Single-
molecule analysis of RNA polymerase transcription.
Annu. Rev. Biophys. Biomol. Struct. 35, 342–360
41. Rosenburg, S. A., Quinlan, M. E., Forkey, J. N. &
Goldman, Y. E. Rotational motions of macromolecules
by single-molecule fluorescence microscopy. Acc.
Chem. Res. 38, 583–593 (2005).
42. Smith, L. M., McConnell, H. M., Smith Baron, A. &
Parce, J. W. Pattern photobleaching of fluorescent lipid
vesicles using polarized laser light. Biophys. J. 33,
43. Yoshida, T. M. & Barisas, B. G. Protein rotational
motion in solution measured by polarized fluorescence
depletion. Biophys. J. 50, 41–53 (1986).
44. Scalettar, B., Selvin, P., Axelrod, D., Hearst, J. &
Klein, M. P. A fluorescence photobleaching study of
the microsecond reorientational motions of DNA.
Biophys. J. 53, 215–226 (1988).
45. Velez, M. & Axelrod, D. Polarized fluorescence
photobleaching recovery for measuring rotational
diffusion in solutions and membranes. Biophys. J. 53,
46. Timbs, M. M. & Thompson, N. L. Slow rotational
mobilities of antibodies and lipids associated with
substrate-supported phospholipid monolayers as
measured by polarized fluorescence photobleaching
recovery. Biophys. J. 58, 413–428 (1990).
47. Velez, M., Barald, K. F. & Axelrod, D. Rotational
diffusion of acetylcholine receptors on cultured rat
myotubes. J. Cell Biol. 110, 2049–2059 (1990).
48. Scalettar, B., Selvin. P., Axelrod, D., Hearst, J. &
Klein, M. P. Rotational diffusion of DNA in agarose
gels. Biochemistry 29, 4790–4798 (1990).
49. Selvin, P., Scalettar, B., Axelrod, D., Langmore, J. P.,
Hearst, J. & Klein, M. P. Rotational diffusion of DNA in
intact nucleii. J. Mol. Biol. 214, 911–922
50. Yuan, Y. & Axelrod, D. Subnanosecond polarized
fluorescence photobleaching: rotational diffusion of
acetylcholine receptors on developing muscle cells.
Biophys. J. 69, 690–700 (1995).
51. Abney, J. R., Cutler, B., Fillbach, M. L., Axelrod, D. &
Scalettar, B. A. Chromatin dynamics in interphase
nucleii and its implications for nuclear structure. J. Cell
Biol. 137, 1459–1468 (1997).
52. Oheim, M. & Schapper, F. Non-linear evanescent-field
imaging. J. Phys. D Appl. Phys. 38, R185–R197
53. Huang, Z & Thompson, N. L. Theory for two-photon
excitation in pattern photobleaching with evanescent
illumination. Biophys. Chem. 47, 241–249
54. Buehler, Ch., Dong, C. Y., So, P. T. C. & Gratton, E.
Time-resolved polarization imaging by pump-probe
(stimulated emission) fluorescence microscopy.
Biophys. J. 79, 536–549 (2000).
55. Mathur, A. B., Truskey, G. A. & Reichert W. M. Atomic
force and total internal reflection fluorescence
microscopy for the study of force transmission in
endothelial cells. Biophys. J. 78, 1725–1735
56. Trache, A. & Meininger, G. A. Atomic force multi-
optical imaging integrated microscope for monitoring
molecular dynamics in live cells. J. Biomed. Optics 10,
57. Yamada, T., Afrin, R., Arakawa, H. & Ikai, A. High
sensitivity detection of protein molecules picked up on a
probe of atomic force microscope based on the fluore-
scence detection by a total internal reflection fluore-
scence microscope. FEBS Lett. 569, 59–64 (2004).
Competing interests statement
The authors declare no competing financial interests.
Daniel Axelrod’s homepage: http://www.physics.lsa.umich.
Geneva M. Omann’s homepage: http://www.biochem.med.
Access to this links box is available online.
Functional and quantitative
proteomics using SILAC
Abstract | Researchers in many biological areas now routinely characterize proteins
by mass spectrometry. Among the many formats for quantitative proteomics,
stable-isotope labelling by amino acids in cell culture (SILAC) has emerged as a
simple and powerful one. SILAC removes false positives in protein-interaction
studies, reveals large-scale kinetics of proteomes and — as a quantitative
phosphoproteomics technology — directly uncovers important points in the
signalling pathways that control cellular decisions.
Proteins are in direct control of almost
all cellular processes, and post-genomic
biology will not reach its potential until we
have tools to study proteins on a large scale.
Sadly, proteomics has lagged far behind
DNA-based technologies, mainly because
there are no protein-analysis methods
similar to oligonucleotide hybridization,
amplification and sequencing. This lag
is diminishing rapidly, however, due to
increasingly powerful mass spectrometry
(MS)-based technologies1. Mass spectro-
meters can sequence thousands of peptides
from complex mixtures in an automated
manner2. The importance of quantitation in
proteomics has recently become appreciated,
and technologies such as isotope-encoded
affinity tags (ICAT)3 have generated wide-
spread interest. Much of this interest is still
focused on the determination of the relative
levels of expression of, ideally, all proteins
between two cell or tissue states (expression
proteomics). Expression proteomics has also
been the elusive goal of the two-dimensional
gel electrophoresis community and has
essentially the same principle at the protein
level as microarrays have at the mRNA level
However, proteomics — unlike transcrip-
tomics — is not limited to measuring
whole cell or tissue expression levels. Here,
I argue that the most important contribu-
tions of quantitative proteomics to biological
understanding will come from its unique
capability to determine changes in function-
ally relevant ‘sub-proteomes’. Although many
of the strategies described here can be, and
sometimes have been, implemented with
other quantitative-proteomics techniques,
I also argue that stable-isotope labelling
by amino acids in cell culture (SILAC) com-
bined with sophisticated mass-spectrometric
and bioinformatic technology is particularly
well suited to reinvent or enhance bio-
chemistry-based approaches. This argument
will be made, in part, by describing success-
ful applications of SILAC to a wide range of
952 | DECEMBER 2006 | VOLUME 7
© 2006 Nature Publishing Group
First, I explain the principles of SILAC
and its advantages and limitations compared
with other quantitative-proteomics strate-
gies. SILAC has been used successfully for
the determination of cholesterol-dependent
lipid rafts and the elucidation of protein
interactions, side-stepping the longstanding
trade-off between assuring specificity of
the measured interaction and preservation
of weak interactions. I then provide some
examples of the quantitation of changes in
the phosphoproteome from the yeast phero-
mone pathway to stem-cell differentiation.
Furthermore, SILAC has become a prime
technology to add a time dimension to
proteo mics. Last, I describe the emerging use
of SILAC to directly study protein synthesis
and degradation. As we reach the conclusions
and perspectives, I hope that the reader will
appreciate how SILAC-based proteo mics
fundamentally transforms the questions
we can ask using biochemical approaches.
Indeed, the emergence of the SILAC strategy
and the extraordinary progress in mass-
spectrometric technology in just the past few
years now offers exciting new strategies for
advances in virtually all areas of biology.
Principles of SILAC
SILAC was first described in the literature
only four years ago4,5, later than many other
schemes for quantitative proteomics (REF. 6;
see also reviews7–11 for different aspects of
the field). It is conceptually and experimen-
tally straightforward: it involves growing two
populations of cells, one in a medium that
contains a ‘light’ (normal) amino acid and
the other in a medium that contains a ‘heavy’
amino acid. The heavy amino acid can con-
tain 2H instead of H, 13C instead of 12C,
or 15N instead of 14N. Incorporation of the
heavy amino acid into a peptide leads to a
known mass shift compared with the peptide
that contains the light version of the amino
acid (for example, 6 Da in the case of
13C6-Arg), but to no other chemical changes.
In some respects, SILAC is similar to
other metabolic labelling techniques that
have been used for decades in biological
research (for advantages and limitations of
SILAC compared with other quantitative-
proteomics technologies, see BOX 1).
For example, in pulse-chase experiments,
a radioactive form of an amino acid is added
to a medium with cells for a certain length
of time and the radioactivity that is incor-
porated into the newly synthesized proteins
is measured. However, in contrast to pulse-
chase experiments, the isotopes of SILAC
amino acids are stable — no radioactivity
is involved. Furthermore, the objective in
SILAC is to distinguish two proteomes by
the molecular weight of the light or heavy
amino acid that is used during the growth
of the two cell populations. This requires
complete labelling — the SILAC amino acid
in all proteins should be replaced — and this
is achieved after five cell doublings, even for
proteins with no significant turnover.
To ensure that cells only incorporate the
added, labelled amino acid into their pro-
teome, essential amino acids are chosen and
cells are grown with dialyzed serum, when
appropriate. Some cell types require low
molecular mass growth factors, which are
not present in dialyzed serum. In these cases,
exogenous growth factors or serum dialyzed
with a low molecular mass cut-off can be
used. Alternatively, the use of a small amount
of undialyzed serum — not sufficient to
significantly influence quantitation — has
proven useful12. Early concerns that the
isotope labels would be scrambled through
metabolic cycles that involve different amino
acids have proven unfounded. An exception
is Arg, which is converted to Pro by some
cell types when present at high levels in the
media13. Peptides that contain both Arg and
Pro can be quantified separately, or Arg in
the media can be titrated down so that Pro
conversion becomes negligible and cells can
still grow normally.
In a straightforward expression-
proteo mics experiment, one cell popula-
tion is labelled with a light amino acid
(population A) and another cell population
is labelled with a heavy amino acid (popula-
tion B). Then the cells are mixed and their
prote o mes are extracted and measured by
MS. Each peptide appears as a pair in the
mass spectra — the peptide with lower mass
contains the light amino acid and originates
from population A, and the peptide with
higher mass contains the heavy amino acid
and originates from population B. If the
SILAC peptide pair appears in a one-to-one
ratio then there is no difference in the abun-
dance of this protein between the proteomes.
A higher peak intensity from the peptide that
contains the heavy amino acid indicates
that the protein was more abundant in
population B. Because the light and heavy
amino acids are chemically identical, except
for their mass difference, the ratio of peak
intensities in the mass spectrometer directly
yields the ratio of the proteins in population
A versus population B.
Accuracy of quantitation depends on the
abundance and signal-to-noise ratio of the
peptide pair, and can be as good as a few
percent14. Usually, ratios of 1.3–2.0-fold have
been used as cut-offs for both statistical and
biological significance. For example, Everly
et al. used SILAC to compare microsomal
fractions of a more metastatic versus a
less metastatic prostate-cancer cell line15,16.
A total of 1,395 proteins were quantified, and
20% of these were differentially expressed
by more than 3-fold between the 2 cell lines.
Box 1 | Advantages and limitations of SILAC
Mass spectrometry is not inherently quantitative, as different molecules have different mass
spectrometric responses. Stable-isotope analogues of the molecule to be quantified have
therefore been used for many years for accurate quantitation in small-molecule mass spectrometry.
To differentially quantify two proteomes, a stable isotope can be introduced in various ways, most
commonly by chemical modification or by metabolic labelling6. Chemical labelling can be done on
any proteome, including body fluids and biopsis material, whereas metabolic strategies require
living cells. Chemical strategies involve a derivatization step that might not be complete, that
might introduce side products, and that might limit the sensitivity of the analysis.
Another important difference between metabolic and chemical labelling is the labelling stage. In
chemical labelling, the two proteomes to be compared have to be purified and fractionated
separately and in precisely the same way to allow relative quantitation of the same fractions. By
contrast, metabolic labelling allows mixing of labelled and unlabelled cells and therefore
subsequent fractionation and purification steps will not introduce any errors in quantitation.
Metabolic labelling for quantitation was first introduced to proteomics by the Langen group in
1998 by feeding microbes a 15N-substituted food source (described in REF. 55) and it was also used
shortly afterwards for quantitation by the Chait group56. The stable-isotope labelling by amino
acids in cell culture (SILAC) strategy has a number of advantages with respect to 15N labelling. For
instance, in SILAC, the expected mass differences are known before peptide identification,
simplifying quantitation. Furthermore, mammalian cell lines are easily labelled by providing the
SILAC amino acid instead of having to eliminate any unlabelled nitrogen source from the media
and — because only one or two amino acids in a peptide are substituted and the degree of
labelling is very high — quantitation is straightforward. SILAC does require adapting one’s cell line
to dialyzed serum, and further steps might have to be taken if cells don’t adapt well. However, this
step can be checked without labelled amino acids. The SILAC approach is not necessarily
expensive, especially if amino acids are bought directly and in gram amounts from a supplier.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 7 | DECEMBER 2006 | 953
© 2006 Nature Publishing Group
(counts per s)
(counts per s)
(counts per s)
(counts per s)
(counts per s × 10–2)
Elution time (min)
(counts per s × 10–2)
Elution time (min)
Ratio > 10Ratio = 1.73
Grown with normal Leu
filipin or nystatin
Lyse cells in 1% Triton X-100
Density gradient centrifugation to isolate
low-density fraction and enrich ra?s
Grown with LeuD3
Gene-ontology analysis revealed that 10%
of the most highly downregulated proteins
in the more metastatic cell line are involved
in cell adhesion. Similarly, Gronborg et al.
used SILAC to compare secreted proteins
from pancreatic-cancer-derived cells with
non-neoplastic pancreatic cells and found a
number of differentially expressed proteins17.
In contrast to expression proteomics,
which aims to quantify the complete prote-
ome, in a ‘functional’ SILAC experiment
one cell population functions as a control,
whereas the other one is perturbed in
some way and a sub-proteome of interest is
isolated (BOX 2). The SILAC approach itself
is straightforward, but identifying the most
elegant perturbation and control might
require considerable ingenuity.
Eliminating false-positive interactions
One of the most spectacular applications
of proteomics has been the elucidation of
protein-interaction networks. Specifically,
two large-scale studies in yeast used
immuno precipitation followed by MS,
and these were influential in promoting
the view that most proteins exist as parts
of multiprotein complexes rather than as
solitary factors18,19. Unfortunately, these
studies are extremely difficult to control for
false-positive inter actions20,21. The problem
is that MS-based proteomics is now so
sensitive that any pull-down assay, regard-
less of bait, will result in a large number
of identified background proteins. More
specific and stringent purification, such as
Box 2 | Principle of a functional SILAC experiment, exemplified by the lipid-raft proteome
Lipid rafts are areas of the plasma membrane
with specific lipid and protein composition
that are thought to organize signalling
events57; however, the concept of lipid rafts is
controversial58, partly because rafts cannot
be directly visualized or purified completely.
To investigate the existence of lipid rafts
(panel a), we used a deuterated-Leu-labelled
(LeuD3) cell population as a control and a
normal Leu-labelled population that was
treated with cholesterol-disrupting agents59.
We then combined both cell populations and
enriched for detergent-resistant lipid rafts in
the standard way. Lipid rafts are destroyed
in the treated, but not the untreated, stable-
isotope labelling by amino acids in cell
culture (SILAC)-labelled cell populations, and
cells are combined.
The actual mass spectrometry (MS) data for
peptides from two different proteins is
shown in panel b. The SILAC peptide pair at
mass-to-charge ratio (m/z) 709.89 and
712.903 had higher intensity in the heavy-
amino-acid form. Peptide sequencing
(MS/MS; middle) identified the peptide as
originating from flotillin-1, and the extracted
ion current (XIC; peptide ion signal as a
function of elution time; bottom) indicates
the total signal per peptide (see REF. 60 for
an introduction to peptide sequencing).
Clearly the integrated signal of the
light form is much less than that of the heavy
form, indicating that the cholesterol-
disrupting drug removed flotillin-1 from the
preparation. By contrast, a peptide from
β-tubulin (SILAC peaks at m/z 520.302 and
521.815) is present even when rafts are
destroyed (ratio 1.73, which is not
statistically significant in this experiment)
and this protein is therefore not a genuine
member of lipid rafts.
MS analysis identified 703 proteins, but
only 241 of these had statistically significant
fold changes, indicating that they were
depleted by removal of cholesterol.
By definition, these 241 proteins are
members of cholesterol-dependent lipid
rafts. Figure redrawn with permission from
REF. 59 © (2003) National Academy of
954 | DECEMBER 2006 | VOLUME 7
© 2006 Nature Publishing Group
Logs of SILAC ratios
with the tandem affinity purification (TAP)
tag22, is an improvement but not a panacea,
as low-affinity interactions tend to be lost.
Both the Aebersold group, using ICAT,
and my group, using SILAC, showed that
quantitative proteomics can completely
overcome the problem of nonspecifically
interacting proteins5,23,24. The basic strategy is
explained in FIG. 1 using a phosphodependent
protein–protein interaction that involves
a bacterial protein, which is injected into a
human host cell in the course of bacterial
infection. Following the synthesis of both
non-phosphorylated and phosphorylated
forms of the peptide that contains the phos-
phorylation site of the bacterial protein, the
non-phosphorylated and phosphorylated
peptides were incubated with light- and
heavy-amino-acid-containing cell lysates,
respectively. After gentle washing to preserve
weak interactions, bound proteins were eluted
and mixed. The vast majority of the proteins
that were identified in the mixed eluates were
present in a one-to-one ratio, which indicates
that they were binding nonspecifically to the
beads or to the peptide. Proteins that bind
specifically as a result of phosphorylation are
easily distinguishable because of their statisti-
cally significant ratios. As a further control,
the phosphorylated peptide can be incubated
with the heavy-amino-acid-containing lysate
and the non-phosphorylated peptide with the
light-amino-acid-containing lysate. In this
‘cross-over’ experiment, all SILAC-peptide
ratios should be inverted.
We scaled up this technology to identify
the interactome of all potentially Tyr-
phosphorylated residues of the ERBB-
receptor family. Our findings recapitulated
many of the interactions that were identified
over the past 15 years. Furthermore, some
new interactions, which implicate epidermal
growth factor receptor (EGFR) and ERBB4
in more diverse signalling roles than ERBB2
and ERBB3, were also identified25. All
interactors with a statistically significant
SILAC ratio had phosphotyrosine-binding
domains. This finding, compared with data
already available in the literature, indicated
that the screen did not result in any
Identifying binding partners of PP1. The
principle outlined above can be used with
any protein interactions as long as bait and
control are available and specific interacting
proteins are expressed in sufficient amounts
for detection by MS. Trinkle-Mulcahy et al.
tagged isoforms of protein phosphatase-1
(PP1) with green fluorescent protein (GFP)
and used SILAC to identify and distinguish
their binding partners26. Three cell states
were labelled for pull-down with control–
GFP, PP1α–GFP and PP1γ–GFP. Out of the
hundreds of the proteins that were identified
with anti-GFP antibody immunoprecipita-
tion, the SILAC ratios showed that some
proteins bind to all PP1 isoforms, whereas
others bind specifically to PP1α or PP1γ.
This study also identified a new protein,
Repo-Man, which binds to PP1γ, targets PP1
to chromatin and seems to have important
functions in the cell cycle27.
SILAC is currently used in a similar man-
ner to probe protein interactions of immo-
bilized drugs. In our laboratory we have also
used SILAC to identify transcription factors
and cofactors by incubating double-stranded
synthetic DNA with nuclear extract. Any
protein that preferentially binds to the wild-
type DNA compared to a point-mutated
sequence is a candidate sequence-specific
DNA-binding protein. Also, SILAC can be
combined with small interfering (si)RNA
knockdown of the antigen of interest in an
immunoprecipitation. By comparison of
eluates of the antibody from cells with and
without siRNA treatment, crossreactivity of
the antibody is eliminated and true interac-
tors of the endogenous protein are obtained28.
SILAC for quantitative phosphoproteomics
MS can directly measure and sequence
endogenous phosphopeptides without
the need to generate special reagents, so it
has therefore become a method of choice
for determining phosphorylation sites
(reviewed in REFS 29–31). Improvements in
phosphoproteomics technology now allow
sequencing of hundreds or even thousands
of phosphorylation sites in a single experi-
ment32. However, it has rapidly become clear
that a functional filter is needed to extract
biological insights from such experiments.
Mapping yeast pheromone responses. SILAC
is ideally suited to distinguish the phos-
phorylation events that are associated with
a particular stimulus from a vast excess of
basal phosphorylation sites. As all peptides
need to be quantifiable, both Arg and Lys
are usually labelled33, and trypsin is used as a
protease. Gruhler et al. labelled a yeast strain
by SILAC and exposed one population to
mating factor and enriched phosphopeptides
from total cell lysate34. Sequencing and quan-
titation of more than 700 phosphopeptides
showed a 2-fold change in 139 of these pep-
tides upon pheromone stimulation. These
phosphorylation sites mapped to proteins in
the entire mating-signalling pathway from
receptors to transcription factors.
Examining stem-cell differentiation.
Quantitative phosphoproteomics can
also directly identify factors that control
cellular decisions. We were interested in
understanding how growth factors influence
stem-cell differentiation, particularly the
differentiation of adult mesenchymal stem
cells to bone-forming cells35. We found that
two related growth factors, EGF and platelet-
derived growth factor (PDGF), induced
Tyr phosphorylation; however, only EGF
strongly enhanced differentiation to
To determine the cause of this difference,
we SILAC-labelled three cell populations
and exposed them to no stimulus, EGF stim-
ulation or PDGF stimulation (FIG. 2). We then
enriched the Tyr phosphoproteome with
anti-phosphotyrosine antibody. As expected,
most of the Tyr phosphoproteome was regu-
lated by both EGF and PDGF, albeit often
to different degrees. However, there were
also a few proteins that were only activated
by one of these growth factors. In particular,
phosphatidylinositol 3-kinase (PI3K)
Figure 1 | Interaction of a bacterial phospho-
protein with a human host protein. An immobi-
lized synthetic phosphopeptide that corresponds
to the phosphorylation site of the bacterial
protein CagA was incubated with heavy stable-
isotope labelling by amino acids in cell culture
(SILAC)-labelled mammalian cell extracts. As a
control, the peptide that lacks the phosphogroup
was incubated with the light-SILAC-labelled cell
extract. Eluates from phosphopeptide and non-
phosphopeptide baits were mixed and analysed
by liquid chromatography–tandem mass spec-
trometry. The proteins that were identified are
aligned according to the logs of their SILAC
ratios. More than 150 proteins were identified
with ratios close to one to one, which shows that
these proteins bind equally well to the peptide,
regardless of the presence of the phosphogroup.
Only SHP2 (Src-homology-2 domain-containing
protein tyrosine phosphatase-2) has a statistically
significant ratio, which indicates that SHP2 binds
specifically to the phosphopeptide. These data
confirm a recently described interaction61.
Figure courtesy of M. Selbach, Max-Planck
Institute for Biochemistry, Martinsried, Germany.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 7 | DECEMBER 2006 | 955
© 2006 Nature Publishing Group
423425 427 577 429
ERK2 - FRHENIIGINDIIR
PI3K, p85-β - ILLNSER
8 16 764 9
Proteins modulated by EGF and/or PDGF
Figure 2 | Systems biology of stem-cell differentiation. a | Mesenchymal adult stem cells differenti-
ated to bone cells upon epidermal growth factor (EGF), but not platelet-derived growth factor (PDGF),
treatment. To determine the proteins that control the differentiation process, stable-isotope labelling
by amino acids in cell culture (SILAC; using Arg0, Arg6 and Arg10) was used to analyse three cell
populations; EGF-treated cells, PDGF-treated cells and control cells that were untreated. Subsequently,
cells were mixed and the phosphotyrosine proteome was enriched using an anti-phosphotyrosine
antibody. b | The heavy-amino-acid SILAC forms of a peptide from mitogen-activated protein kinase-2
(MAPK2, also known as ERK2) from the EGF- and PDGF-stimulated populations are more intense than
the control, which indicates that ERK2 is stimulated by both growth factors. However, a peptide from
the p85β subunit of the phosphatidylinositol 3-kinase (PI3K) is activated over the control (present in
the Tyr phosphoproteome) only in the population that is stimulated by PDGF. This finding indicates
that PI3K might be a possible control point for the differentiation to bone cells. c | Evaluation of the
entire data set showed that 113 proteins are modulated by EGF and/or PDGF (entire oval); 76 proteins
are activated similarly by EGF and PDGF; 16 are more activated by EGF and 4 are more activated by
PDGF (up to threefold); 8 proteins are only activated by EGF, and 9 are only activated by PDGF. LC–MS,
liquid chromatography–mass spectrometry; m/z, mass-to-charge ratio. Adapted with permission from
REF. 35 © (2005) American Association for the Advancement of Science.
Mix lysates 1:1:1
Affinity purification with
Digestion with trypsin
Protein indentification and
quantitation by LC–MS
was only activated by PDGF and was there-
fore a candidate for the differential effects.
Indeed, the PI3K inhibitor wortmannin
caused PDGF stimulation to be as potent in
bone-cell differentiation as EGF in vitro and
in vivo. This finding, which might also be
of clinical interest, shows how quantitative
phosphoproteomics can help to pinpoint sites
of cellular decision making.
The Tyr phosphoproteome has also been
studied in cancer cells. ERBB2 is overex-
pressed in a subset of breast cancers and
phosphotyrosine signalling downstream of
this receptor has been studied by two groups.
Zhang et al. used ligand stimulated versus
unstimulated cells in the SILAC experiment36.
Bose et al. quantified the Tyr phospho-
proteome of untreated and phosphatase-
inhibitor treated ERBB2-overexpressing cells
and used network modelling to account for
the changes in the phosphoproteome37.
Adding the time dimension to proteomics
So far, most proteome experiments char-
acterize a static time point or an on–off
situ ation. To incorporate kinetics into
organellar proteomics, three different cell
populations were labelled, treated with an
inhibitor of transcription and harvested
at three different time points38. Cells were
subsequently mixed and their nucleoli
were isolated. In response to inhibition
of transcription, some proteins leave the
nucleoli, some remain unchanged and some
are recruited to the nucleoli. Previously,
such changes had to be visualized in a
candidate-based approach using GFP-fusion
proteins. By contrast, SILAC directly meas-
ures changes in endogenous proteins in the
entire nucleolar proteome. A peptide from
a protein recruited to the nucleolus would
show a triplet with increasing intensities,
because more of the protein is present in
the nucleolar preparations from later time
points (FIG. 3). By repeating the experiment
several times with a common time point,
up to nine point kinetics were obtained.
Interacting proteins left the nucleolus with
similar time profiles. These time profiles
also helped to group the nucleolar proteome
into interacting modules by machine learn-
ing, which is an area of artificial intelligence
that allows the computer to uncover
patterns and classify data accordingly39.
Once the experiment is set up, SILAC
time courses of a proteome can be obtained
rapidly. Use of different inhibitors of
transcription showed trafficking of dif-
ferent protein populations from or to the
nucleolus. For example, inhibition of the
proteasome resulted in largely opposite
kinetics compared to a specific inhibitor of
polymerase II transcription38. Such findings
can be followed up by further experiments
in conjunction with single-cell fluorescence
microscopy. The combination of modern
microscopy techniques with SILAC is espe-
cially powerful because SILAC is unbiased,
comprehensive and measures endogenous
proteins, whereas biochemical approaches
average over large cell populations. On the
other hand, microscopy supplies detailed
spatial, temporal and interaction information
at the single-cell level, ideally complementing
Another area in which the time dimen-
sion is of great interest is in signalling
cascades. Such kinetics are difficult to obtain
in a global and unbiased way. Using SILAC
and a time-course experiment similar to
the one described for the nucleolus experi-
ment, Blagoev et al. mapped changes in
the phosphotyrosine proteome as a result
of EGF stimulation. Activation profiles
were obtained covering all classes of Tyr
phosphorylated effectors40. Recently, we
have extended this approach to quantita-
tively measure thousands of Ser/Thr/Tyr
phosphoryl ation sites, which yielded a first
956 | DECEMBER 2006 | VOLUME 7
© 2006 Nature Publishing Group
Treatment with Act D (min)
Counts × 100
Counts × 100
Time 0 min
12C6 14N4-Arg (Arg0)
Time 20 min, Act D
13C6 14N4-Arg (Arg6)
Time 80 min, Act D
13C6 15N4-Arg (Arg10)
Mix cells 1: 1: 1
Digest with trypsin
Identify and quantify proteins
overall map of early cellular information
processing in response to stimuli, from
receptor to transcription factor41.
Studies of protein turnover
All of the examples discussed so far entail
complete labelling of each cell population.
However, as pioneered by Beynon and
Gaskel42,43, SILAC can also directly measure
protein turnover. The technique consists
of switching the SILAC amino acid in the
media at a particular time point from light
to heavy (or from heavy to light). From this
time point on, all newly synthesized proteins
incorporate the heavy SILAC amino acid
and are easily distinguishable from the pre-
existing proteome. In principle, it is straight-
forward to measure turnover of proteins in
any cellular compartment.
In a clever application of this principle,
Admon and co-workers studied major
histocompatibility complex (MHC) peptides
with the SILAC technique44. MHC class I
peptides are generally derived from cellular
proteins and are presented on the cell surface
for inspection by the immune system. The
MHC peptidome is thought to contain pop-
ulations with different turnover times: long-
lived proteins that are degraded at the end of
their useful life, short-lived proteins that are
mainly regulated by degradation, and also
newly synthesized proteins that fail to pass
the protein-quality-control machinery
and are immediately degraded45.
Indeed, the SILAC data showed that
protein classes with different turnover rates
were present. Some proteins were degraded
within minutes of ribosomal synthesis
and others persisted for days or longer.
Intriguingly, this analysis found limited
correlation between the proteome and the
MHC peptidome, highlighting one of the
reasons that mRNA levels and protein levels
are sometimes poorly linked. In another
recent study of the MHC peptidome,
Meiring et al. used SILAC to recognize and
sequence virus-derived MHC peptides after
infection of a cell line46. Direct measurement
of protein synthesis and degradation opens
up an exciting new window for cell biology
As shown above, SILAC in combination
with existing cell biological and biochemi-
cal approaches can be used to ask profound
questions, which are limited only by the
imagination of the researcher. However, the
impact of SILAC on the biological com-
munity is currently restricted by a lack of
access to the technology. Although SILAC
labelling is easy for any laboratory that
uses cell culture, the MS technology that is
required is still beyond the capabilities of
Fortunately, new mass spectrometers
such as a linear ion trap–orbitrap combina-
tion now allow very high performance in
a compact and robust format47,48. One of
the factors that contributed to the rapid
acceptance of the SILAC technology was the
availability of an open-source programme,
MSQuant, for interpreting results24.
However, because of the large amount of
data involved, interpretation of results is still
a limiting factor. Further software develop-
ments, using concepts from MS algorithms
for biomarker development (see REF. 49 as
an example) will surely advance the field.
Further automation will also help in better
defining statistical validity in quantitative-
In the future, one could imagine research-
ers using SILAC in two modes. In the
exploratory mode, experiments will be
planned, executed and analysed rapidly to
generate hypotheses and ideas for further
experimental refinement. In the explora-
tory mode, statistical validity is not the
primary concern, but rather the objective
is the discovery of a biological effect. In the
validation mode, experiments will be per-
formed several times (analytical replicates)
and perhaps with different cell populations
(biological replicates). Analysis of these
experiments should ideally also be auto-
mated so that they result in a list of proteins
with statistically validated ratios.
At the other extreme, a main application
area of SILAC might be the quantitative
analysis of modifications of particular puri-
fied proteins. A protein of interest can be
Figure 3 | Nucleolar proteome dynamics. a | Three Arg-stable-isotope labelling by amino acids in
cell culture (SILAC)-labelled cell states (Arg0, Arg6 and Arg10) were treated for various times with
actinomycin D ( Act D), an inhibitor of transcription, and nucleoli were isolated. b | Mass spectrum of
a peptide from the nucleolar protein p68 in three SILAC-labelled forms. The intensity of the peptide
is highest in the peptide derived from the nucleoli that have been isolated after 80 minutes of tran-
scription inhibition, indicating that upon inhibition of transcription, the protein is recruited to the
nucleolus. c | Mass spectrum for the same peptide after inhibition for different lengths of time.
d | Combination of the data from panels b and c leads to a kinetic curve of recruitment of p68 to
nucleoli. nanoLC–MS, nano liquid chromatography–tandem mass spectrometry; m/z, mass-to-charge
ratio. Reproduced with permission from REF. 38 © (2005) MacMillan Magazines, Ltd.
NATURE REVIEWS | MOLECULAR CELL BIOLOGY
VOLUME 7 | DECEMBER 2006 | 957
© 2006 Nature Publishing Group
enriched from differentially-treated SILAC-
labelled cells and analysed for changes
in processing or for other modifications
induced by the treatments. As an example,
histone methylation can be quantified by
using labelled Met, which is the sole methyl
donor of the cell (heavy methyl SILAC,
see REF. 50). Illustrating the biological
insights that can be obtained rapidly with
such approaches, Timmer and co-workers
expressed a tagged potassium channel,
stimulated a SILAC population with different
treatments and found regulated phosphoryla-
tion sites51. Similar experiments can be done
by any laboratory with access to MS equip-
ment and results can be analysed manually.
Will SILAC always be restricted to cell-
culture experiments? Microorganisms can
already be labelled if one uses auxotroph
strains, and Nirmalan et al. showed that the
malaria parasite can be labelled with Ile,
an amino acid that is not found in haemo-
globin, the preferential source of amino
acids for the parasite as it grows in red blood
cells52. For metazoans, Ishihama et al. have
developed culture-derived isotope tags
(CDITs)53; the authors used SILAC labelling
for a neuronal cell line, which was then used
as an internal standard to quantify brain
sections. Others have metabolically labelled
organisms as large as rats54 and we have
labelled mice with SILAC amino acids with a
defined amino-acid diet.
It is a fair bet that SILAC kits by various
companies will arrive on the market in the
next few years. With better software and
more accessible MS, SILAC might become
a routine assay for the next generation of
biochemists and cell biologists. Through its
capability to quantify both the proteome and
its modifications in response to stimuli and
perturbations, SILAC might also become an
important foundation for systems biology.
Matthias Mann is at the Department of
Proteomics and Signal Transduction, Max-Planck
Institute for Biochemistry, Am Klopferspitz 18,
D-82152 Martinsried, Germany.
Aebersold, R. & Mann, M. Mass spectrometry-based
proteomics. Nature 422, 198–207 (2003).
Washburn, M. P., Wolters, D. & Yates, J. R. 3rd. Large-
scale analysis of the yeast proteome by
multidimensional protein identification technology.
Nature Biotechnol. 19, 242–247 (2001).
Gygi, S. P. et al. Quantitative analysis of complex
protein mixtures using isotope-coded affinity tags.
Nature Biotechnol. 17, 994–999 (1999).
Ong, S. E. et al. Stable isotope labeling by amino acids
in cell culture, SILAC, as a simple and accurate
approach to expression proteomics. Mol. Cell.
Proteomics 1, 376–386 (2002).
Blagoev, B. et al. A proteomics strategy to elucidate
functional protein–protein interactions applied to EGF
signaling. Nature Biotechnol. 21, 315–318
Ong, S. E. & Mann, M. Mass spectrometry-based
proteomics turns quantitative. Nature Chem. Biol. 1,
Julka, S. & Regnier, F. Quantification in proteomics
through stable isotope coding: a review. J. Proteome
Res. 3, 350–363 (2004).
MacCoss, M. J. & Matthews, D. E. Quantitative MS for
proteomics: teaching a new dog old tricks. Anal. Chem.
77, 294A–302A (2005).
Yan, W. & Chen, S. S. Mass spectrometry-based
quantitative proteomic profiling. Brief Funct. Genomic.
Proteomic. 4, 27–38 (2005).
10. Wright, M. E., Han, D. K. & Aebersold, R. Mass
spectrometry-based expression profiling of clinical
prostate cancer. Mol. Cell. Proteomics 4, 545–554
11. Kuster, B., Schirle, M., Mallick, P. & Aebersold, R.
Scoring proteomes with proteotypic peptide probes.
Nature Rev. Mol. Cell Biol. 6, 577–583 (2005).
12. Gehrmann, M. L., Hathout, Y. & Fenselau, C.
Evaluation of metabolic labeling for comparative
proteomics in breast cancer cells. J. Proteome Res. 3,
13. Ong, S. E., Foster, L. J. & Mann, M. Mass
spectrometric-based approaches in quantitative
proteomics. Methods 29, 124–130 (2003).
14. Ong, S. E., Kratchmarova, I. & Mann, M. Properties of
13C-substituted arginine in stable isotope labeling by
amino acids in cell culture (SILAC). J. Proteome Res. 2,
15. Everley, P. A., Krijgsveld, J., Zetter, B. R. & Gygi, S. P.
Quantitative cancer proteomics: stable isotope
labeling with amino acids in cell culture (SILAC) as a
tool for prostate cancer research. Mol. Cell.
Proteomics 3, 729–735 (2004).
16. Everley, P. A. et al. Enhanced analysis of metastatic
prostate cancer using stable isotopes and high mass
accuracy instrumentation. J. Proteome Res. 5,
17. Gronborg, M. et al. Biomarker discovery from
pancreatic cancer secretome using a differential
proteomic approach. Mol. Cell. Proteomics 5,
18. Ho, Y. et al. Systematic identification of protein
complexes in Saccharomyces cerevisiae by mass
spectrometry. Nature 415, 180–183 (2002).
19. Gavin, A. C. et al. Functional organization of the yeast
proteome by systematic analysis of protein complexes.
Nature 415, 141–147 (2002).
20. von Mering, C. et al. Comparative assessment of large-
scale data sets of protein–protein interactions. Nature
417, 399–403 (2002).
21. de Hoog, C. L. & Mann, M. Proteomics. Annu. Rev.
Genomics Hum. Genet. 5, 267–293 (2004).
22. Rigaut, G. et al. A generic protein purification method
for protein complex characterization and proteome
exploration. Nature Biotechnol. 17, 1030–1032
23. Ranish, J. A. et al. The study of macromolecular
complexes by quantitative proteomics. Nature Genet.
33, 349–355 (2003).
24. Schulze, W. X. & Mann, M. A novel proteomic screen
for peptide–protein interactions. J. Biol. Chem. 279,
25. Schulze, W. X., Deng, L. & Mann, M. Phosphotyrosine
interactome of the ErbB-receptor kinase family. Mol.
Syst. Biol. 1, E1–E13 (2005).
26. Trinkle-Mulcahy, L. et al. Repo-Man recruits PP1γ to
chromatin and is essential for cell viability. J. Cell Biol.
172, 679–692 (2006).
27. Vagnarelli, P. et al. Condensin and Repo-Man–PP1
co-operate in the regulation of chromosome
architecture during mitosis. Nature Cell. Biol. 8,
28. Selbach, M. & Mann, M. Protein interaction screening
by quantitative immunoprecipitation combined with
knock-down (QUICK). Nature Methods 3, 29 Oct
29. Mann, M. et al. Analysis of protein phosphorylation
using mass spectrometry: deciphering the
phosphoproteome. Trends Biotechnol. 20, 261–268
30. Carr, S. A., Annan, R. S. & Huddleston, M. J. Mapping
posttranslational modifications of proteins by MS-based
selective detection: application to phosphoproteomics.
Methods Enzymol. 405, 82–115 (2005).
31. Loyet, K. M., Stults, J. T. & Arnott, D. Mass
spectrometric contributions to the practice of
phosphorylation site mapping through 2003: a
literature review. Mol. Cell. Proteomics 4, 235–245
32. Beausoleil, S. A. et al. Large-scale characterization of
HeLa cell nuclear phosphoproteins. Proc. Natl Acad.
Sci. USA 101, 12130–12135 (2004).
33. Ibarrola, N., Kalume, D. E., Gronborg, M., Iwahori, A.
& Pandey, A. A proteomic approach for quantitation
of phosphorylation using stable isotope labeling in
cell culture. Anal. Chem. 75, 6043–6049
34. Gruhler, A. et al. Quantitative phosphoproteomics
applied to the yeast pheromone signaling pathway.
Mol. Cell. Proteomics 4, 310–327 (2005).
35. Kratchmarova, I., Blagoev, B., Haack-Sorensen, M.,
Kassem, M. & Mann, M. Mechanism of divergent
growth factor effects in mesenchymal stem cell
differentiation. Science 308, 1472–1477
36. Zhang, G., Spellman, D. S., Skolnik, E. Y. &
Neubert, T. A. Quantitative phosphotyrosine
proteomics of EphB2 signaling by stable isotope
labeling with amino acids in cell culture (SILAC).
J. Proteome Res. 5, 581–588 (2006).
37. Bose, R. et al. Phosphoproteomic analysis of Her2/neu
signaling and inhibition. Proc. Natl Acad. Sci. USA
103, 9773–9778 (2006).
38. Andersen, J. S. et al. Nucleolar proteome dynamics.
Nature 433, 77–83 (2005).
39. Hinsby, A. M. et al. A wiring of the human nucleolus.
Mol. Cell 22, 285–295 (2006).
40. Blagoev, B., Ong, S. E., Kratchmarova, I. & Mann, M.
Temporal analysis of phosphotyrosine-dependent
signaling networks by quantitative proteomics. Nature
Biotechnol. 22, 1139–1145 (2004).
41. Olsen, J. V. et al. Global, in vivo and site-specific
phosphorylation dynamics in signaling networks. Cell
127, 635–648 (2006).
42. Pratt, J. M. et al. Dynamics of protein turnover, a
missing dimension in proteomics. Mol. Cell.
Proteomics 1, 579–591 (2002).
43. Doherty, M. K., Whitehead, C., McCormack, H.,
Gaskell, S. J. & Beynon, R. J. Proteome dynamics in
complex organisms: using stable isotopes to monitor
individual protein turnover rates. Proteomics 5,
44. Milner, E., Barnea, E., Beer, I. & Admon, A. The
turnover kinetics of major histocompatibility complex
peptides of human cancer cells. Mol. Cell. Proteomics
5, 357–365 (2006).
45. Yewdell, J. W., Reits, E. & Neefjes, J. Making sense of
mass destruction: quantitating MHC class I antigen
presentation. Nature Rev. Immunol. 3, 952–961
46. Meiring, H. D. et al. Stable isotope tagging of
epitopes: a highly selective strategy for the
identification of major histocompatibility complex
class I-associated peptides induced upon viral
infection. Mol. Cell. Proteomics 5, 902–913
47. Makarov, A. Electrostatic axially harmonic orbital
trapping: a high-performance technique of mass
analysis. Anal. Chem. 72, 1156–1162
48. Olsen, J. V. et al. Parts per million mass accuracy on
an Orbitrap mass spectrometer via lock mass injection
into a C-trap. Mol. Cell. Proteomics 4, 2010–2021
49. Leptos, K. C., Sarracino, D. A., Jaffe, J. D., Krastins, B.
& Church, G. M. MapQuant: open-source software for
large-scale protein quantification. Proteomics 6,
50. Ong, S. E., Mittler, G. & Mann, M. Identifying and
quantifying in vivo methylation sites by heavy
methyl SILAC. Nature Methods 1, 119–126
51. Park, K. S., Mohapatra, D. P., Misonou, H. &
Trimmer, J. S. Graded regulation of the Kv2.1
potassium channel by variable phosphorylation.
Science 313, 976–979 (2006).
52. Nirmalan, N., Sims, P. F. & Hyde, J. E. Quantitative
proteomics of the human malaria parasite
Plasmodium falciparum and its application to studies
of development and inhibition. Mol. Microbiol. 52,
53. Ishihama, Y. et al. Quantitative mouse brain
proteomics using culture-derived isotope tags as
internal standards. Nature Biotechnol. 23, 617–621
54. Wu, C. C., MacCoss, M. J., Howell, K. E.,
Matthews, D. E. & Yates, J. R. 3rd. Metabolic labeling
of mammalian organisms with stable isotopes for
quantitative proteomic analysis. Anal. Chem. 76,
958 | DECEMBER 2006 | VOLUME 7
© 2006 Nature Publishing Group