Professional Documents
Culture Documents
201000289 3935
REVIEW
Developments in subcellular fractionation strategies have provided the means to profile and Received: April 30, 2010
analyze the protein composition of organelles and cellular structures by proteomics. Here, we Revised: July 14, 2010
review the application of classical (e.g. density gradient centrifugation) and emerging Accepted: August 3, 2010
sophisticated techniques (fluorescent-assisted organelle sorting) in the fractionation, and
statistical/bioinformatics tools for the prediction of protein localization in subcellular
proteomics. We also review the validation methods currently used (such as microscopy, RNA
interference and multiple reaction monitoring) and discuss the importance of verification of
the results obtained in subcellular proteomics. Finally, the numerous challenges facing
subcellular proteomics including the dynamics of organelles are being examined. However,
complementary approaches such as modern statistics, bioinformatics and large-scale inte-
grative analysis are beginning to emerge as powerful tools to proteomics for analyzing
subcellular organelles and structures.
Keywords:
Cell biology / Organelle / Subcellular fractionation
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
3936 Y. H. Lee et al. Proteomics 2010, 10, 3935–3956
steps of the organelles. This is particularly crucial in the Recent developments in methodological and technological
detection of proteins that are present in low abundance advances have the potential to yield organelles of higher
against the complex background of whole cells or crude purity, and as technologies progress, the effectiveness will
tissue homogenates [3–5]. Classical methods of enriching improve, albeit at a cost of higher complexity or prohibitive
organelles and cellular structures such as density-gradient resource requirements. Here we discuss both classical and
centrifugation have been used extensively but these orga- emerging techniques that may be stand-alone or comple-
nellar preparations often contain contaminating proteins. mentary to conventional methods to improve subcellular
This major limitation provided the opportunities for cell fractionation for high-resolution organellar proteomics.
biologists and biochemists to improve on the enrichment
techniques, which can then lead to an increase in the
proteome coverage and at the same time being able to 2.1 Conventional methods
provide reliable data on the distribution of bona fide orga-
nellar proteins. Indeed, much progress has been made in Specific proteomes of cellular compartments and organelles
the recent years in uncovering many novel organellar can be isolated via several approaches, and the most
proteins with their spatial distributions and new biological commonly used strategy is subcellular fractionation [8, 9].
paradigms. Such novel discoveries can potentially lead to the The concept of subcellular fractionation was pioneered by De
systematic identification of new causative proteins for Duve and coworkers in 1955 using rat liver tissue [10]. The
various human diseases [6, 7]. Thus in this review, we aim to procedure for subcellular fractionation first involved disrup-
discuss, evaluate and summarize the various strategies and tion of cells and tissues using physical methods such as
technologies, conventional as well as emerging innovative mechanical or liquid shear or non-physical methods such as
approaches that have been used to verify and validate the detergents or hypo-osmotic shock. The cell disruption step
proteome of eukaryotic organelles. aimed to obtain high protein yield, and maintain structural
and functional integrity of intracellular organelles. The effi-
ciency of cell disruption is often monitored with immuno-
2 Strategies for subcellular fractionation chemical and microscopic studies. Conventional methods
involved in subcellular fractionation, typically by differential
Figure 1 shows the organelles and cellular structures and density-gradient centrifugation, employed a series of
commonly found within higher organisms and plants. The centrifugation steps to separate different populations of
approaches used to isolate target organelles from the rest of cellular compartments or organelles from cell homogenates
the cellular compartments will be important prior to based on their mass and/or density [2].
proteomics. Classically, organellar proteomics depend on
biochemical methods to isolate the diverse organelles found
in cells. However, the complete global profiling of organelle 2.1.1 Differential centrifugation
proteomes demands robust sample preparation and orga-
nelle enrichment methods to ensure the correct annotation Differential centrifugation operates via sequential centrifu-
of protein distribution. Such classical approaches have gation of the cell or tissue homogenate to obtain subcellular
unfortunately, been plagued by the problem of co-purifying organelles such as nuclei, mitochondria and lysosomes.
contaminants. While this may be true, classical methods This separation method is based on differences in size and
remain easy to adopt and modify in most laboratories and density, whereby larger and denser organelles sediment at
they also lay the ground for more sophisticated approaches. lower centrifugal forces. However, this method has poor
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
Proteomics 2010, 10, 3935–3956 3937
resolving power and thus may result in fractions containing membrane organelle-localized, soluble and DNA-associated
a different organelle type that has similar sedimentation nuclear and cytoskeletal proteins) [14, 15], and is commer-
velocities. cially available as the ProteoExtract Subcellular Proteome
Extraction kit (Calbiochem, San Diego, CA, USA). Abdol-
zade-Bavil et al. described a robust DDF method with high
2.1.2 Density-gradient centrifugation efficiency and selectivity, which was supported by 2-D gel
electrophoresis results, morphological, immunological and
The more commonly applied density-gradient centrifuga- enzymatic analyses [15]. Recently, this method has been
tion method separates organelles based on continuous or used to obtain comprehensive proteomes (including hydro-
discontinuous gradients using various media, such as phobic membrane proteins) for functional genome annota-
sucrose of different osmolarities, viscosities or densities. In tion [16]. It has also been applied on frozen archival tissues
general, the cell homogenate or post-nuclear supernatant is that showed that there is no loss of proteome coverage or
first added to the top of the medium and centrifuged. Upon information about protein localization [17]. DDF has also
centrifugation, the organelle focuses in the gradient where been used to fractionate neurons in culture plates to enrich
its density equals the density of the surrounding medium for cytosolic, membrane-bound or enclosed, nuclear, and
(isopycnic point). The rate of sedimentation of an organelle cytoskeletal proteins for differential proteomics analysis [18].
in density-gradient centrifugation thus depends on the Lastly, Holden and Horton established a protocol to
differences in organelle density relative to medium, and that sequentially extract proteins from cultured mammalian cells
in turn is determined by the organelle content, lipid/protein into fractions that were enriched for cytosolic, membrane-
ratio, size and shape [11]. For example, high-density mito- bound organellar, nuclear and insoluble proteins, using
chondria and endoplasmic reticulum have high protein digitonin and detergent [19].
content, whereas endosomes have a low density due
to their lipid-rich membranes. Sucrose is the most
commonly used medium for density-gradient centrifugation 2.2 Emerging methods
as sucrose is biologically inert, inexpensive and dialyzable.
Other media include Ficoll, Percoll, Nycodenz and 2.2.1 Free-flow electrophoresis
Metrizamide [12].
In a continuous gradient, the density increases linearly Free-flow electrophoresis (FFE) offers an alternative method
along the tube but in a discontinuous gradient, the gradient to isolate and fractionate organelles such as peroxisomal
is divided into fixed portions consecutively. During discon- membranes [20], mitochondria [21], secretory vesicles [22],
tinuous gradient centrifugation, different organelles are Arabidopsis thaliana tonoplasts, plasma membrane vesicles
enriched at different interphases of the medium. On the and peroxisomes [23, 24]. Organelle isolation by FFE is
other hand, in continuous gradient centrifugation, equili- based on separation by their net global isoelectric charges or
brium separation occurs whereby organelles distribute electrophoretic mobilities. In addition to fractionation of
throughout the gradient and are focused at their isopycnic organelles, FFE has been used for protein separation [25].
points. This provided a better resolution of organelles than During FFE, the carrier buffer is flowing continuously and
discontinuous gradients. However, preparation of contin- laminarly in a narrow gap between two cooling plates, with
uous gradient can be time consuming and required special an electric field being applied perpendicularly to the laminar
apparatus. Density-gradient centrifugation remained as a flow of the buffer and sample, thus resulting in the differ-
popular choice of researchers for subcellular fractionation as ential deflections of the sample molecules [26].
this method is well developed and referenced, and applic- Some of the advantages of FFE include the wide choice of
able for many organellar proteomics studies [13]. pH, conductivity and composition of electrolytes solutions
[26] and the high sample recoveries (approximately in the
range of 95%) that arose from the injection of samples as
2.1.3 Differential detergent fractionation thin film which precluded the use of any matrix. The
continuous application and fractionation of the samples and
Besides differential and density-gradient centrifugation, one collection of the samples at the other end of the FFE
approach advocated to reduce contaminating proteins from chamber lend itself well for preparative organelle separa-
other organelles and to avoid time-consuming centrifuga- tions. In addition, the purified organelles retain their
tion steps is by using buffers of increasing stringency in the intactness and functionality, which makes FFE an attractive
subcellular fractionation steps. One such method is differ- technique not only for proteomic analyses but also
ential detergent fractionation (DDF), which employed complementary functional studies. Also, the preservation of
sequential extraction of organelle proteins using different proteins in their native state by FFE is complementary to
detergent-containing buffers. DDF is a simple method that native or top-down proteomics [27, 28]. To avoid the problem
can fractionate proteins in their native state according to of Joule heating, which arose from the heating of the
four subcellular localizations (cytosolic, membrane and separation medium by electric currents to focus the larger
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
3938 Y. H. Lee et al. Proteomics 2010, 10, 3935–3956
drag force of organelles, micro-FFE that utilized lower different groups have suggested conflicting hypotheses
electric currents and higher surface area-to-volume ratio to [38–42]. Hence complementary and supportive biochemical
dissipate heat, has been employed [29, 30]. and microscopic techniques will aid in defining the
One challenge of FFE is when organelles or cellular contribution of ER to phagosomes [43] in subcellular frac-
structures have similar isoelectric points, which would tionation. Recently, a different method of purifying phago-
result in comigration. For example, it was found that in the somes has been developed which entailed the use of a
FFE separation of neutrophil secretory vesicles, contamina- combination of differential and density-gradient sedi-
tions with ER and mitochondria were evident, probably as a mentation on sucrose and iodixanol gradient media. The
result of the similarity in isoelectric points. Similarly, method did not use latex beads and it showed a high degree
mitochondrial and microsomal proteins can be found in of purification [44].
peroxisomal preparations [20, 24] and hence other orthogo- Recently, more interest has been generated in the char-
nal approaches are required to resolve them. In spite of this, acterization of exosomes, given their pleiotropic effects on
FFE was shown to be useful in resolving a minor fraction of paracrine and endocrine processes [45]. Immunoaffinity
peroxisomes from the heavy mitochondrial fraction, in purification of exosomes is an attractive way of obtaining
addition to the more commonly enriched light mitochon- more homogenous exosomal preparations without the
drial fraction [31]. possibility of capturing contaminants of similar density and
size. It should be noted that exosomes and HIV-1 particles
share a number of similar biophysical properties such as
2.2.2 Immunoaffinity purification size, density and export/transport processes. To this end,
exosomes purified using immunoaffinity capture has been
Immunoaffinity purification is a powerful method to isolate shown to be virion free [46, 47]. For verification, flow cyto-
organelles with specificity and in adequate yields [32]. The metry combined with exosomal-specific markers was used to
two major methods, affinity purification and immunopre- demonstrate the purity of enriched exosomes [48].
cipitation that are used to isolate organelles are based on
binding between ligands (antibodies) immobilized on solid
supports and targets (organelle of interest). For proteomic 2.2.3 Fluorescent-assisted organelle sorting (FAOS)
analysis, affinity purification is a good option to start for its
ability to do repeated purification using a single sample and Besides its use as a verification method, cell sorter flow
hence enrichment of the organelle. Ligands are no longer cytometry can also be used to isolate and enrich organelles in
limited to antibodies, but also include chemical ligands, a method termed ‘‘fluorescent-assisted organelle sorting’’
combinatorial peptide libraries, aptamers and others [33]. (FAOS) [49, 50]. Only simple amendments to the flow
Limitations to affinity purification include its high cost and cytometer are required, thus making FAOS readily accessible
the increased experimental duration required for effective to most laboratories attempting subcellular proteomics. This
purification. Lastly, the signal-to-noise generated in MS may approach has been used to isolate a number of subcellular
be inadequately low for complex biological samples, espe- organelles such as phagosomes [51], mitochondria [52],
cially with the initially less well-fractionated samples. endocytic vesicles [53, 54] and secretory granules [50].
Therefore, to reduce the complexity of the sample, very often There are a few things to observe when sorting organelles
density centrifugation is performed prior to immunopur- of interest. Intact organelles are important for successful
ification. For example, Nycodenz density-gradient centrifu- sorting and proteomic analysis. A protein marker that is
gation followed by immunoaffinity purification has been specific to the organelle of interest will be mandatory for
reported to isolate peroxisomes from rat liver [34]. Likewise, their successful sorting. This protein tagged with green
such a strategy has been used to enrich for secretory fluorescent protein (GFP) can be transfected into cells,
synaptic vesicles from synaptosomes [35]. enabling its fluorescence to be discriminated by the cell
In a bid to enrich phagosomes, opsonized latex beads sorter [54]. Alternatively, if organelle-specific surface
have been used. Cells engulf these targets and the phago- markers are available, FAOS can be employed in a fashion
somic membranes remain adherent and can be isolated by analogous to FACS for cell sorting. Information on the size
flotation of the beads. Opsonized latex beads are particularly of organelles is important and as long as an appropriate
useful because they are inert, amenable to laboratory sorting window is set within the known dimensions of the
manipulations and their different sizes can be suited to the organelle of interest, their enrichment can be achieved.
different targets of individual and differing phagocytic However, in cases where there is heterogeneity of organelle
systems [36]. The ER is generally dense so it should not sizes, it can be difficult to ascertain if FAOS sorted the
sediment with the latex beads in density gradients [37]. organelles accurately. One recent application of FAOS is in
However, opsonized latex beads enrichment has consis- the identification of more than 150 secretory granule
tently presented ER proteins in the phagosomal prepara- proteins [50]. It was noted that a high degree of purity was
tions [36]. The issue surrounding the contribution of ER achieved, including FAOS’s ability to distinguish two other
membranes to phagosomes is contentious as several fluorescing organelles, the ER and Golgi apparatus.
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
Proteomics 2010, 10, 3935–3956 3939
2.2.4 Proteomics in combination with statistical (LOPIT) in A. thaliana [55, 56]. In principle, the LOPIT
analyses workflow can be summarized as follows: (i) preparative
isolation of organelles by density centrifugation, (ii) selec-
In order to address the problem of co-purifying proteins, tion of fractions for analysis based on known marker protein
many groups have looked at the use of statistical analysis distribution across the density gradients (Figs. 2A, B and D)
and bioinformatic predictions to aid the assignment of and (iii) identification and quantitation by iTRAQ labeling
known and unknown proteins to their respective subcellular and MS analysis. The strength of LOPIT lies in that it works
compartments. Multivariate analysis in general offers great around the problem of imperfect methods to isolate pure
versatility in analyzing proteomic data, especially where organellar preparations. Proteins were sorted according to
there are a large number of parameters (proteins) and since their subcellular distributions by PCA and tight clusters of
dimension reduction is required to predict and annotate proteins belonging to similar compartments can be
proteins of unknown function and localization. observed visually on PCA plots with 1 being the perfect
Multivariate analysis such as principle component match to a group. For the purpose of classifying proteins of
analysis (PCA) and partial least-squares discriminant unknown distributions, PLS-DA predicts group member-
analysis (PLS-DA) were applied by Dunkley and Lilley to ship by the use of the set of distinctly clustered proteins
predict and classify proteins to their organelles in a method already assigned by PCA. Where proteins cannot be discri-
termed ‘‘localization of organelle protein by isotope tagging’’ minated, such proteins may be found in (i) multiple
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
3940 Y. H. Lee et al. Proteomics 2010, 10, 3935–3956
locations, (ii) transiting cellular compartments such as tion. The investigators then restrict their analyses to
endosomes or (iii) where few or no proteins of known uncharacterized open reading frames that were only found
compartments are detected in the MS data such that no in the subtracted nuclear envelope fractions, thereby anno-
distinctive clusters are formed [57]. tating 67 previously unknown putative proteins to the
Another major advancement in subcellular proteomic nuclear envelope, in addition to the 13 known proteins [6].
strategy that is independent of organelles being purified to Of these proteins, eight of them were validated using
homogeneity and which also utilizes statistical measures is immunofluorescence. Their findings led them to propose
protein correlation profiling (PCP) [37, 58–60]. In PCP, that 62 of these nuclear envelope proteins are potentially
organelles are recovered from fractions generated in density associated with dystrophies in humans. While these
centrifugation and the abundance of each peptide in each hypotheses remain to be tested [63], many dystrophy disea-
gradient fraction is calculated from the area of its extracted ses have been linked to mutations in nuclear envelope
ion current peak. Adjacent fractions are also taken into proteins [64]. In summary, this highlighted that the
account, hence signals of the same peptide are normalized, subtractive proteomics approach is capable of assigning and
correlated with elution times of different fractions and identifying proteins to their specific subcellular locations
compiled together to form protein correlation curves and generating new models to link protein mutations to
(Fig. 2C). Using the established resident organellar proteins human diseases.
peaking in different density fractions, one can define the
distinctive profiles of these markers by using the w2 value
that then acts as a basis to discern subcellular localization of 2.2.6 Large-scale integrative analyses
other proteins [59]. This strategy is similar to the more
commonly used immunoblotting profile of fractions Increasingly, integrative analyses are being applied to better
(Figs. 2A and B). Notably, PCP was able to distinguish annotate protein localization within organelles. MS-based
dynamic transiting organelles including endosomes and proteomics are employed as the main tool of detecting
ER/Golgi-derived vesicles [37]. PCP can also be used to train organellar proteins, integrated with diverse data sets and
for false-positive sets with proteins known to be localized robust statistics and bioinformatics to systemically assign
outside the organelle of interest [58, 59]. For example, Wiese proteins their spatial locations [7, 37, 65, 66].
et al. applied PCP on kidney peroxisomes fractionation, and The mitochondrion is an ideal organelle to start such
found 15 novel peroxisomal candidates, 5 of which were integrative efforts. Its many core cellular functions includ-
validated by immunocytochemistry [58]. Peroxisomes and ing oxidative phosphorylation, fatty acid oxidation and
mitochondria possess similar densities and such character- others are well conserved during evolution. Phylogenetic
istics compound the problem of correctly assigning proteins evolution is a useful dimension in elucidating the proteome
to either of each fraction. As such, unsupervised k-means complement of an organelle – sequence homology of
clustering was employed as an added criterion to better mitochondrial proteins in other organisms offers a level of
differentiate unique peroxisomal, mitochondrial or dual inference of the same organelle localization of proteins in
localized proteins. humans [67]. There are two caveats to this approach, first,
only proteins that have homology to other species or are
conserved can be inferred; second phylogenetic analyses
2.2.5 Subtractive proteomics may be difficult to specify subcellular localization for large
protein families such as SNAREs and Rab GTPases that
One potential error of organellar proteomics is the inad- have diverse functions, structures and diverged evolution
vertent contamination of target organelles with co-purifying [68]. It is thought that the mitochondrion was derived from
organelles from the commonly used classical methods. This an endosymbiotic relationship with the primordial eukar-
problem may be circumvented through the use of subtrac- yotic cell [69]. Rickettsia prowazekii, an a-proteobacterium, is
tive proteomics [6, 61, 62]. In subtractive proteomics considered the closest living ancestor to the modern day
(Fig. 2E), the method compares and ‘subtracts’ the identified mitochondria and the homology of its constituent proteins
protein constituents of the contaminated fraction containing provides strong clues to the modern day mitochondrial
the organelle of interest against that of a crude preparation proteome (specificity 14%, sensitivity 30%). Similarly, the
[1]. With the aim of identifying membrane proteins from the genetically tractable Saccharomyces cerevisiae provides
nuclear envelope, Yates and colleagues first prepared experimental ease for the systematic identification of genes
microsomal membranes from livers that were rich in ER in its relatively small genome, estimated to be 60%
proteins, contained mitochondrial proteins, but free of similar to humans [66]. To achieve significant success in
nuclear membranes. Next they collected crude nuclear identifying genes and known functionality in yeast, yeast
envelope fraction that contained contaminating mitochon- orthologs provide strong clues to mitochondrial protein
drial membranes. To obtain nuclear envelope-specific localization in higher organisms (specificity 35–47%,
proteins, they subtracted proteins of the microsomal sensitivity 40–48%) [70]. In addition, proteins targeted to the
membrane fraction from the crude nuclear envelope frac- mitochondrion generally possess N-targeting sequences
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
Proteomics 2010, 10, 3935–3956 3941
[71] that are co-expressed via transcriptional induction upon subject to artifacts and/or provide low coverage of the full
mitochondrial biogenesis [72] or GFP-localization [73], and proteome [77]. These experimental limitations have
these are useful information (data set) in identifying bona fide encouraged the ongoing efforts to develop reliable compu-
mitochondrial proteins. Each data set is heterogeneous and tational predictors to analyze large-scale genome sequence
incomplete yet complementary; thus by integrating diverse for protein localization [78–83]. These computational
data sets of varying specificity and sensitivity using naı̈ve methods are automated, faster and provide wider proteome
Bayes, linear predictors or support vector machine (SVM), coverage and hints to the potential biological roles of
proteins can be scored as a function of its probability as a uncharacterized proteins to direct further experiments [84].
mitochondrial protein [67]. A predictive score, termed Maes-
tro was generated that permitted the ranking of the likelihood
of a truly mitochondrial protein. To validate their predictions, 3.1 Sequence-based and annotation-based
localization of 131 predicted novel mitochondrial proteins classifiers
were demonstrated unambiguously to reside within the
mitochondrion [7]. Subtractive proteomics (Fig. 2E) was Many in silico prediction tools for subcellular localization
employed to distinguish mitochondrial proteins from false have been developed using different data sets and features,
positives (non-mitochondrial co-purifying proteins such as and they are accessible via the Internet [85]. A comprehen-
ER proteins). While the current MS lacks the sensitivity of sive summary of prediction methods was published by Chou
detecting lower abundance proteins and that the composition and Shen [86]. Some predictors predict for all cellular
of the mitochondrial proteome changes with its cell state, compartments, but many are specific for one organelle or
proteomics alone will be insufficient in confirming the true organism. Organism-specific prediction tools have been
localization of proteins. Hence, large-scale integrative studies shown to perform better than those designed for general
that leverage various aspects of organelle biology are powerful eukaryotic organisms [78].
strategies to characterize the resident proteins in a highly The two important factors in predicting subcellular
confident manner. localization are the extraction of crucial biological features
The phagosome or autophagosome, a crucial double relevant to subcellular localization, and the computational
membrane-bound organelle in immune responses, apopto- method applied in the system. These computational
sis and tissue remodeling, is another organelle whose predictors are mainly grouped as sequence-based or anno-
protein components were expanded and identified through tation-based classifiers [87]. Sequence-based predictors are
the use of a combination of MS, RNA-mediated interference based on the amino acid sequence of the protein, by iden-
(RNAi) and bioinformatics [74]. The elegant use of a systems tifying protein sorting signals [88, 89] such as N-terminal
biology approach offered a powerful framework to under- targeting peptides [79, 90–92], mitochondrial targeting
stand the dynamic nature of the Drosophila phagosome [36]. peptides [93], chloroplast transit peptides [94] and nuclear
This set of data was used to build the interactome, which in localization signals [95] or analyzing whole sequence
turn generated potential phagosome associating proteins features [96–98] such as amino acid composition [99, 100],
that were validated functionally. The authors extended the dipeptide composition [101] and gap amino acid pair
macromolecular complex, exocyst to the phagosome, thus compositions [100]. Annotation-based methods detect func-
further increasing the inventory of phagosomal proteins. tional domains and motifs [81, 102, 103], textual information
Such an effort enhances our understanding of proteome in databases [104–108], homology search [109–111] or
composition during host–pathogen response and under- phylogenetic profiling [103, 112].
scores the utility of integrative analysis in combination with There are hybrid predictor systems [96, 113] that applied
simpler fractionation techniques to confidently ascribe a mixture of strategies such as those based on sequence and
proteins to their cellular localizations. annotation [82] or sorting signals and sequence composition
[114, 115]. Some models incorporate results from different
modules to integrate signal peptide or whole sequence
3 Bioinformatics prediction information with features such as physical and chemical
properties [116] or protein domain information [117]. For
Uncovering protein localization would help to annotate example, PSORT-B is based on several features including
genomes, infer potential roles and discover diagnostics and signal peptide, amino acid composition, transmembrane
drug targets [75]. However, annotation of protein localiza- a-helices and motifs, and homology to proteins of known
tion has not been keeping pace with the rapidly accumu- location [118].
lating genomic databases generated from large-scale These hybrid systems are more robust and often achieve
sequencing projects. A large proportion of proteins in the higher prediction performance, and hence are preferred for
Swiss-Prot and Gene Ontology databases do not have novel proteins. The incorporation of evolutionary informa-
experimentally validated localization annotations [76]. tion with machine learning technique in RSLpred often
Many of the ‘wet-lab’ methods used to determine protein resulted in a better performance than other methods such as
localization are time consuming, expensive, laborious, TargetP, Wolf-PSORT, PA-SUB, Plant-Ploc and ESLpred
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
3942 Y. H. Lee et al. Proteomics 2010, 10, 3935–3956
[119]. SherLoc2 is another hybrid subcellular localization apparatus. Not all organelle proteins contain sorting signals,
prediction system, which covers 11 main eukaryotic thus the major drawback of prediction models based on
subcellular locations for animal, fungal and plant proteins targeting signals is that there are many subcellular sites for
with an estimated accuracy of 93% [120]. Small et al. which targeting sequence is not known, such as the nuclear
recommended the use of high-throughput screening tools membranes and the Golgi apparatus. Post-translational
such as Predotar [92] and MITOPRED [121, 122] to identify modifications and protein–protein interactions may also
potential proteins for subsequent testing using hybrid influence protein localization.
programs (e.g. SherLoc and PA-SUB) [123].
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
Proteomics 2010, 10, 3935–3956 3943
functional interaction network of human proteome to map sites [146]. Euk-mPLoc is one of the predictors developed
known disease associations and identify novel disease-asso- that can analyze eukaryotic proteins with multiple location
ciated genes. However, this model can only predict for five sites including those not annotated in the Swiss-Prot data-
major subcellular localizations due to limited training and base [147]. Recently, Lin et al. [148] described KnowPredsite,
testing data sets for other cellular compartments. Indeed, a knowledge-based system, which can be used for proteome-
the performance of prediction tools depends heavily on the wide prediction of proteins that are single-localized or multi-
redundancy of data set used for training. The distribution of localized. They showed that KnowPredsite performed better
proteins in different cellular locations is imbalanced, which than the ngLOC and Blast-hit method, with an overall
results in certain locations with a smaller number of train- accuracy of 91.7 and 72.1% for single-localized and multi-
ing data. Many of the computational prediction methods localized proteins, respectively.
have limited location sites coverage or predicts for many
locations but with low accuracy. For example, due to the
small peroxisomal training data set, the performance of 3.8 Organelle database
predictions tools such as PSORT II [80], pTARGET [140] or
PTS1 predictor [141] have generated inconsistent data, even With the increasing number of proteins including novel
for the known peroxisomal proteins. uncharacterized gene products being reported by subcellular
In addition, the predictive performance of any computa- proteomics, there are currently a large number of publicly
tional methods is highly associated with the degree of accessible organelle databases [2, 149]. For example, there
similarity between the training and test sets. One of the have been a number of comprehensive organellar proteomic
limitations of prediction models is that the training data sets studies being performed, and several systematic studies of
encompass proteins with high sequence homology to others subcellular localizations using cellular fractionation have
located in the same cellular compartment. For example, been reported [150]. The proteome profiles of most organelles
PSORT-B is a widely used prediction method but its training including mitochondria [151], chloroplast [152] and nucleolus
set contains a strong homology bias. On the other hand, [153] have now been identified [1]. However, many of the
Gneg-PLoc has a training set where none of the proteins in a proteome databases are still incomplete. For example, despite
subcellular location has more than 25% sequence homology the completion of the human genome sequence, the exact
to each other [142]. Unlike PSORT-B that covers five location number of protein-coding genes is unknown. Furthermore,
sites, Gneg-PLoc can predict eight subcellular location sites. the exact genomic structure of human protein-coding genes
Despite the exclusion of homologous proteins and the has only been correctly predicted for 50–60% of the genes
inclusion of more subcellular locations sites, Gneg-PLoc [154, 155]. Incomplete proteome datasets would thwart the
achieves higher success rates than PSORT-B. Chou et al. identification of novel organelle proteins from subcellular
developed Cell-PLoc that contains six predictors: Euk- proteomics studies. For example, several proteomics studies
mPLoc, Hum-mPLoc, Plant-PLoc, Gpos-PLoc, Gneg-PLoc of mitochondria such as those of yeast [156], human heart
and Virus-PLoc, which are specialized for eukaryotic, [157] and mouse [158] have been performed but the exact
human, plant, Gram-positive and negative bacterial and viral number of mammalian mitochondrial proteins is still
proteins, respectively [76]. None of the training proteins unknown as a fraction of the mitochondrial proteome has not
had425% sequence identity to other protein in the same been identified [159, 160]. False positives in databases due to
subcellular location. A high accuracy was demonstrated by contaminants from impure subcellular fractions would
Cell-PLoc based on cross-validation tests on the benchmark further complicate organellar proteome analyses.
datasets that involved up to 22 subcellular location sites. Other limitations of public-accessible databases include
variations in data sets and restricted data sets [161]. For
example, MitoProteome [162] is restricted to human records,
3.7 Protein with multiple localizations and thus has limitations to integrate results derived from
model organisms. On the other hand, MitoMiner is a public
A large number of eukaryotic proteins, up to more than database of experimental information from the published
35%, are localized at multiple subcellular locations [143, mitochondrial proteome studies [161], and includes data sets
144]. For example, some proteins are dual targeted to from seven species including Homo sapiens, S. cerevisiae and
mitochondria and plastids [145]. Proteins with multiple Drosophila melanogaster to allow cross-species comparisons.
locations or dynamic localizations are important in both MitoP2 is a database of mitochondrial proteins of yeast,
basic research and drug discovery. Proteome-scale epitope mouse, A. thaliana, neurospora and human [163], and links
tagging, immunolocalization strategies and improved mitochondrial proteins to functional data sets and includes
subcellular fractionation procedures have generated direct information about mouse models and human diseases [164].
evidence for the multiple localizations of many proteins. However, the incompleteness in proteome databases would
One of the limitations of most existing computational hinder inferences of proteins’ subcellular localization
predictors of localizations is the inability to predict proteins between species such as in correlating model organism
that are targeted to or traffic between two or more location proteome with the human proteome.
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
3944 Y. H. Lee et al. Proteomics 2010, 10, 3935–3956
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
Proteomics 2010, 10, 3935–3956 3945
4.3 Assay for enzymatic activity of organelle While fluorescence microscopy is of low throughput, since it
markers is mostly targeted to one protein of interest each time and
being semi-quantitative, and thus less suitable for large-
Besides Western blotting, the other widely used practice of scale quantitative proteomics, it is useful in validating a
evaluating the composition of subcellular fractions is the small number of protein localizations. For example, GFP
measurement of enzymatic activities. Biochemical moni- fusions and confocal microscopy was used to image the
toring of subcellular fractionation requires at least one localization of several uncharacterized target proteins to
reliable marker enzyme per organelle to determine the validate subcellular assignments obtained from organellar
biochemical homogeneity of each fractions. Evaluating the proteomics [6, 158, 167].
amount and activity of characteristic enzyme markers in More importantly, some studies have demonstrated
different cellular compartments will allow quantitative that the presence of tags can induce artifacts such as
assessment of the fractions obtained. It should be noted that mislocalization by obscuring targeting motif, or interfering
measurement of enzymatic activity is semi-quantitative and with folding or protein–protein interactions [176–179]. The
subjected to possible interference from in vitro environment. tag may contain localization signal that overwhelms
Likewise, the criterion for marker selection is similar to the original targeting sequence information of the
immunoblotting verification. However, the presence of protein [180, 181] or become activated in different species
activity may not indicate intact organelles. For example, the [182, 183]. The addition of tag may result in differential
presence of succinate dehydrogenase activity could indicate processing or interfere with post-translational modifications
intact mitochondria or fragments of mitochondrial inner of protein, thus changing the localization [184]. Moreover,
membrane. In order to confirm the presence of intact overexpression of tagged fusion proteins to determine
mitochondria, morphological evidences or assay of markers subcellular localization of gene products has been
of outer mitochondria and intermembrane space must also reported [178, 185] which often results in mislocalization.
be performed. Changes in protein abundance that are dependent on
Recently, Andreyev et al. fractionated RAW264.7 cells different stimuli or conditions are also likely to be lost with
into six fractions that represent the major organelles (nuclei, overexpression.
mitochondria, cytoplasm, endoplasmic reticulum, plasma To circumvent the problem of overexpression, the fluor-
membrane and a dense microsomal fraction) [174]. The escence-based verification technique termed CD-tagging can
authors validated these subcellular fractions by analyzing be applied to verify proteomic data. CD-tagging is a random
the distribution of enzymatic markers. Subsequently, they tagging approach whereby the coding sequence of GFP is
identified 50-member organelle-specific ‘‘marker ensem- inserted randomly into a genomic DNA, generating a large
bles’’ for each subcellular compartments using quantitative library of subcellular images after all proteins are tagged
LC-MS proteomics. The quantitative composition of the [186]. In contrast to cDNA modification that uses constitu-
fractions was based on distribution of the marker ensem- tive promoter and results in higher expression level, this
bles. The use of panels of markers avoided the problem of method retains the regulatory sequence and expression level
multiple locations of some markers, thereby preventing bias of the protein [187]. Chromosomally integrated GFP
from single ‘‘best’’ protein, and the difficulty in obtaining constructs under the control of endogenous promoter would
completely pure organelle fractions. lead to low protein expression [75, 188]. Limitations of CD-
tagging include the need for complete coding sequence of
cDNA, such as inclusion of signal peptides for correct
4.4 Fluorescent microscopy localization and the generation of antibodies for novel gene
products can be difficult and time consuming.
Novel localization assignments by organellar proteomics can In addition to CD-tagging, there are other methods
be validated by visualizing the subcellular protein locations designed to minimize the problems of overexpression.
using the orthogonal technique of fluorescence microscopy. New generation vectors have been designed to include:
This method can be applied to show co-localization of target (i) Flp recombination target site that facilitates rapid
protein with the known organellar markers. Endogenous or isogenic integration in cell lines containing a single FRT
recombinantly expressed proteins can be detected immu- (flanked with heterologous recombinase target) site [189];
nochemically or by tagging with a recombinant epitope or (ii) Tetracycline induction that allows the control of
fluorescent protein. One of the most widely used methods expression levels [190]; (iii) A tag that facilitates protein
for this is transfection of cells with cDNA clones fused to a detection. These new generation vectors combined with
fluorescent reporter such as the GFP [175]. GFP is widely large ‘‘orfeome’’ collections are a very powerful system
used in bioimaging as this protein can be easily tagged to [191–193]. The use of these vectors circumvents the pitfalls
almost any protein to determine its localization. The loca- of CD tagging, that requires the generation of new anti-
tion of proteins fused to non-fluorescent tag, such as FLAG, bodies, while at the same time allowing the scientist to
could be determined using a fluorescent-labeled antibody express the tagged protein uniformly and at close to
targeted to the tag, known as indirect immunofluorescence. physiological levels.
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
3946 Y. H. Lee et al. Proteomics 2010, 10, 3935–3956
Strategies that could avoid the possible artifacts of protein multiple subcellular compartment(s), and many corre-
fusions include the use of fluorescent probes or antibodies sponded to Gene Ontology localization prediction model.
conjugated with fluorescent dyes to visualize protein locali- They used three organelle probes: a DNA stain, DAPI for
zation within organelles. There are several available fluor- the nucleus, tubulin for cytoskeleton as well for internal
escent markers for cell imaging such as the ERTracker, control for fixation quality and homogeneity and calreticulin
MitoTracker, LysoTracker, DAPI or phallicidin for ER, for the endoplasmic reticulum. In summary, these large-
mitochondria, lysosome, nuclei and filamentous actin scale protein localization approaches are well adapted to
cytoskeleton, respectively. However, these approaches evaluate the large volumes of data generated from proteo-
require cells that are fixed and permeabilized for probe mics and to validate their cellular locations.
entry, which thus disallow temporal study of protein loca-
lization. The fixation step may mask antigens or alter the
conformation of proteins. Fluorescence-based methods also 4.6 Automated assignment of localization using
have the disadvantages of pH- and concentration-dependent fluorescent microscopy
fluorophore quenching and photobleaching. Other
problems include non-specific binding and aggregation of Traditionally, protein locations are assigned using visual
organelles by antibodies. inspection and manual evaluation which often contained
Of note, in SILAC-based experiment on adenovirus- bias and varied with the experimenter’s training and
infected HELA cells showed that a subset of proteins enri- experience. Automated, high-throughput fluorescent
ched in the nucleolus was detectable by proteomics but not microscopy can generate images for tens of thousands of
detected by fluorescence microscopy [194]. This highlighted fluorescent-tagged proteins, which is well suited for the
that some proteins may be present in such low abundance validation of high volumes of data generated from large-
that they are below the limit of the current microscopic scale proteome studies [199–201]. In addition, publicly
detection methods. This could be due to the fact that available databases of images for protein localizations would
fluorescence-based microscopy focuses on single cells, provide information about organellar locations and help to
whereas proteomics analyses are based on purified orga- assign locations to unknown proteins in organelle proteo-
nelles obtained from millions of cells. Despite certain mic experiments.
limitations most laboratories have access to fluorescence Systematic study of protein locations in cultured cells has
microscopy to perform spatial confirmation of at least been accomplished by coupling fluorescence microscopy
several proteins. with pattern recognition and machine-learning techniques,
such as ANN and SVM [202, 203]. Methods for systematic
study of protein locations commonly extract subcellular
4.5 Large-scale protein localization location features from images to recognize subcellular
patterns [199]. There are pattern recognition systems avail-
Large-scale protein localization experiments based on GFP able to describe new location patterns or subtle changes.
tagging or immunofluorescence are an important step Cluster analysis of images derived from a library of tagged
forward in complementing and validating cellular spatial protein clones would group proteins into location families.
organization of proteins [185, 195] despite its high cost. In Clustering of images has also been used to group drugs
the first comprehensive study of proteome-wide protein effects upon subcellular patterns [204]. This method has also
localization of a eukaryotic system using systematic GFP allowed building of generative models to create protein
tagging and microscopy, 4156 S. cerevisiae proteins were localization images for the proteome, unlike the conven-
localized to one or more of 22 subcellular organelles or tional microscopy that focused on a few proteins each time
compartments [75]. High-throughput screening of GFP- [205]. High-resolution microscopy images and training sets
tagged proteins has also been reported in yeast on a with sufficient number of well-annotated proteins are
genome-wide scale [196]. Natter et al. studied localization required to allow proper determination of protein location.
pattern of proteins involved in lipid metabolism in S. cere- In particular, the recent technological advances in auto-
visiae by tagging 493 proteins with GFP and using high- mated imaging have dramatically improved the speed of
resolution fluorescence microscopy [197]. The results for bioimaging [206, 207]. Newberg et al. developed a system to
many proteins with known localization were consistent with assign automatically subcellular locations on a proteome-
other reports using cell fractionation or large-scale localiza- wide scale for collections of tissue images such as the
tion approaches. Barbe et al. reported the pilot large-scale Human Protein Atlas [208]. The Human Protein Atlas
antibody-based study of 466 proteins in 3 human cell lines currently contained images for 46000 proteins [209].
using fluorescent-labeled antibodies and confocal micro- One of the current challenges is to apply such compu-
scopy [198]. They generated approximately 3000 high-reso- tational method for all proteins in all cell types in different
lution images, which are available on the Human Protein conditions. Time-lapse images are difficult to acquire since
Atlas portal. Using these techniques, they showed that many microscopes do not maintain a viable environment for
480% of these proteins could be classified in one or cell cultures. Repeated excitations of the fluorescent dyes
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
Proteomics 2010, 10, 3935–3956 3947
would also cause photobleaching, reduced signal and samples, EM will remain an important technique to
phototoxicity to the cells. To circumvent such challenges, a conclusively demonstrate sufficient purity of enriched
system to infer correlation of subcellular protein localization organelles.
with cell cycle by analyzing only static, asynchronous images
using a one-dimensional manifold computed on simple
nuclear image features was developed. For example, Buck 4.8 RNAi
et al. associated protein pattern variation in asynchronous,
static cell images with cell cycle progression [210]. Helmuth The correlation of decreased protein expression in specific
et al. also presented a novel, model-based image analysis organelles upon RNAi knockdown of target proteins can
algorithm to reconstruct outlines of subcellular structures also be used to determine the validity of the protein location
using a sub-pixel representation [211]. They validated the assignment. Coene et al. studied the localization of BRCA1
reconstruction performance on synthetic data and applied to using confocal and immune-EM to show that BRCA1 is
fluorescence microscopy images of endosomes. This new present in mitochondria of several cancer cell lines as well
algorithm allowed quantification of endosome morphology as primary breast and nasal epithelial cells [214]. They
and capture the dynamics of endosome fusion. While large- showed that siRNA knockdown of BRCA1 in human cancer
scale antibody-based protein localization study provides cells led to decreased nuclear, cytoplasmic and mitochon-
valuable temporal and spatial information, such work is drial immunofluorescence staining, validating the data
often limited by antibody specificity [212]. obtained from the microscopic analyses. The authors also
applied subcellular fractionation, dephosphorylation and
enzyme protection experiments to show the enrichment of a
4.7 Electron microscopy phosphorylated isoform of BRCA1 in mitochondrial and
nuclear subcellular fractions. Using these independent
EM allows the direct morphological characterization of the approaches, they showed that phosphorylation relocalization
samples, providing valuable information on the purity and of BRCA1 plays a role in the maintenance of genome
integrity of the enriched organelle of interest. EM uses integrity in mitochondria and nucleus.
particle beam of electrons to illuminate and produce a Many studies have reported nuclear cytoplasmic shuttl-
magnified image of a specimen. EM enables examination of ing of tumor suppressor APC. Using immunofluorescence
samples with a high magnification and a resolution of microscopy, cell fractionation, siRNA and Cre/Fox silencing
approximately 1 nm. Additionally, the location of a protein technology, Brocardo et al. found that some commonly used
can be localized to the organelles using immuno-EM to antibodies against APC showed non-specific nuclear stain-
confirm fluorescence microscopic data. ing pattern in immunofluorescence despite being effective
One application of EM in organellar studies is to analyze in immunoblotting [215]. Despite inactivation of APC
the isolated mitochondria with regards to structural integrity expression by siRNA or Cre/Flox, staining of APC was still
of the inner and outer membranes and mitochondrial observed by some antibodies in immunofluorescence
matrix, so as to distinguish between intact and damaged microscopy. Their results refuted the hypothesis that APC
mitochondria [213]. Owing to its high resolution, EM can nuclear export is inhibited by truncation of C-terminal NES,
detect the presence of double layer membrane and matrix. but demonstrated that this mutant of APC shuttle between
Subsequently, functional integrity of the organelle was nucleus and cytoplasm. This study demonstrated the need
determined based on assay for JC-1 dye uptake. Immuno- for several independent validations and cautions in inter-
gold particles staining and EM was also used to confirm the preting localization from immunofluorescence microscopy
protein localization along and/or in the mitochondrial inner alone.
and outer membranes.
Electron microscopic visualization of target proteins in
the sample is achieved by electron optical contrast, which 5 Perspectives: Challenges of organelle
involves selective electron scattering of the atoms of sample purification exclusiveness or dynamic
and stain. Several staining procedures have been employed localization?
in EM such as immunoantibody staining using colloidal
gold or enzymes such as horse-radish peroxidase, and It is becoming increasingly clear that organelles are dynamic
enzyme cytochemistry which target site of enzyme activity in cellular entities, remodeling and interfacing with other
situ. Although EM allows a direct assessment of organelle organelles at different stages of development, physiological
ultrastructure, this method is often criticized for being time conditions and tissues [146]. Translocation of certain
consuming and inconvenient as it requires skilled operators proteins may also be crucial during the development of
and access to sophisticated equipments, and permitting the diseases, and thus comparative analysis of the dynamic
study of small quantities of sample each time. Artifacts of changes in organellar proteome would aid in understanding
EM, such as shrinkage and precipitation are also widely the molecular mechanism of diseases. It has been observed
known. However, due to its direct, visual comparison of that proteins can be dual or multiple localized in different
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
3948 Y. H. Lee et al. Proteomics 2010, 10, 3935–3956
organelles. It has been reported that 39% of proteins are not This led to the authors proposing that the dynamic change
unique to one cellular compartment but exhibit multiple in protein localization in different compartments may have
localizations [37, 216] and this distribution of proteins functional roles in breast cancer.
complicates organellar proteomics. However, different studies sometimes throw up inter-
Indeed, one of the factors that contributed to the esting and conflicting results and alternative conclusions are
conflicting results obtained from different subcellular suggested. For instance, PCP argues that GAPDH and
proteomics studies is shuttling of proteins between cellular aldolase, both predominantly cytosolic glycolytic enzymes
compartments. This includes proteins that translocate are not localized to the mitochondrion [37]. In contrast,
between organelles, shuttle between cytoplasm and nucleo- other proteomic studies (MudPIT, 2-D native PAGE and 3-D
plasm or cycle between subcellular pools. A well-known denaturing PAGE) have suggested that these two enzymes
dynamic proteome system is the endomembrane system and glucokinase (another glycolytic enzyme) physically
where proteins trafficking en route to their final destination interact with the mitochondrion [157, 223]. Not surprisingly,
are in a constant state of flux through several organelles. organelle preparation may be the early causes of such
Intracellular trafficking of endosomes and secretory vesicles arguments. Alternatively, the different physiological states
occurs continuously between the Golgi and ER to achieve at which the samples were sampled might cause this
dynamic equilibrium [217, 218]. Isolation of these organelles discrepancy.
involved in the endomembrane system is challenging due to Another mitochondria-related controversy is catalase.
the similar buoyant densities of vesicular structures. The Catalase is thought to be confined to the peroxisomal matrix.
dynamic exchange of materials between vesicles would also Indeed in the most complete mitochondrial proteome
alter their densities [219]. Some plasma membrane proteins compendium catalogued, catalase is considered a non-
are endocytosed and then retrieved to the plasma mitochondrial protein [7]. However, catalase has been found
membrane. For example, subtypes of glucose transporters in rat heart mitochondria [224]. More interestingly, yeast
shuttle from intracellular pools to cell surface upon insulin catalase (catalase A) has been noted to distribute to the
signaling [220]. Isolation and analysis of a single organelle is peroxisome as well as mitochondrion, a phenomenon
thus unable to fully understand the dynamic processes dependent on the carbon source [225]. Do take note that
between organelles. To understand the dynamic change of yeast homology is considered part of the strategy in several
the proteome at the subcellular level, several subcellular mitochondrial protein annotative studies [7, 65]. Therefore
structures should be monitored in parallel [8]. this reiterates that the state of the cell’s physiology at a
In fact, comparative analysis at the subcellular level is particular point in time determines the fraction of proteins
necessary to confirm spatial rearrangement of proteins to that are distributed across the various compartments.
provide insights into their biogenesis and cross-talk. It is a The interconnectivity of organelles and their subsequent
challenge to study the biological significance of proteins exchange of proteins is increasingly thought to heavily
found in multiple localizations in different conditions, influence cell function, and their dysregulations can lead to
particularly with a complex intracellular communication pathological conditions. To facilitate Ca21 exchange, the
between organelles [156]. There have been new develop- mitochondria are organized in networks physically linked to
ments in proteomics to facilitate the study of organelle the ER [226]. The interaction between the ER and mito-
dynamics in a high-throughput manner. For example, chondria is important in modulating the Ca21 signaling for
Andersen and colleagues applied SILAC on the human apoptosis and cell metabolism, and mutations in Ca21
nucleoli and in vivo fluorescent imaging to characterize the signaling can lead to Charcot-Marie-Tooth-type 2a neuro-
flux of endogenous nucleolar proteins in response to three pathy, type 2 diabetes and obesity [227, 228]. As such, this
metabolic inhibitors known to affect nucleolar morphology highlighted the need to consider the pathological and
[221]. They demonstrated that the nucleolar proteome biological state when analyzing organellar proteomics data
exhibited temporal compositional and kinetic changes in and do not simply regard the presence of other organellar
response to different cellular growth conditions. These data proteins as contaminants. Autophagy is another highly
suggested that there is no unique proteome for any orga- complex biological mechanism that reflected the dynamism
nelle (nucleoli at least); instead the proteomes overlap in of protein localization. Since autophagy engaged the lyso-
different cell states or conditions and this showed the some and vacuole with the target organelle or cellular
significance of determining the relative abundance of structure destined for degradation [229], it is foreseable that
proteins in different cellular organelles in different physio- when autophagy is induced, an intertwining of the phago-
logical states. Qattan et al. combined sucrose density-gradi- cytic and the to-be-degraded organelle proteomes would
ent subcellular fractionation with quantitative, MS/MS- occur. Selective autophagy of organelles is known to
based shotgun proteomics to study the distribution pattern degrade peroxisomes, secretory granules and mitochondria
of proteins fractionated from MCF-7 breast cancer cells [230, 231]. For example, Brunner and colleagues observed a
[222]. Out of the 2184 identified proteins, 1249 proteins large number of lysosomal proteins and attributed their
possessed multiple location sites of which many of these presence to the degradation and sorting of old and imma-
multi-localized proteins were implicated in breast cancer. ture insulin granules respectively [175]. It has been
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
Proteomics 2010, 10, 3935–3956 3949
demonstrated in vitro that mitochondrial autophagy has a lular proteomes, it is expected that new and valuable data
role during hepatocyte remodeling [232]. Proteomic studies will be obtained soon.
in this area is still lacking, perhaps due to the difficulty in
‘locking’ this dynamic biological event for sampling, but we The authors have declared no conflict of interest.
expect this area of organellar proteomics to grow rapidly
since there is increasing evidence of the contribution of
dysfunction lysosomal activities in human diseases [233].
The dynamic localization of proteins could also be 7 References
resulted from post-translational modifications (PTMs) of
proteins. Spatial and temporal PTMs of proteins are [1] Yates, J. R., 3rd., Gilchrist, A., Howell, K. E., Bergeron, J. J.,
Proteomics of organelles and large cellular structures. Nat.
important aspects in the intracellular signaling systems
Rev. Mol. Cell. Biol. 2005, 6, 702–714.
associated with human physiology. For example, PTMs of
mitochondrial proteome due to changes in cell’s redox [2] Dreger, M., Subcellular proteomics. Mass Spectrom. Rev.
2003, 22, 27–56.
status and increase in reactive oxygen and nitrogen species
are known to play key roles in signal transduction, and also [3] Liu, H., Sadygov, R. G., Yates , J. R., 3rd, A model for
disruption of mitochondrial functions in cancer and random sampling and estimation of relative protein
abundance in shotgun proteomics. Anal. Chem. 2004, 76,
Parkinson’s disease [234]. Thus, proteomics analysis of
4193–4201.
oxidative and nitrosative stress on mitochondria would
identify redox-based mitochondrial proteome alterations [4] Washburn, M. P., Wolters, D., Yates , J. R., 3rd, Large-scale
analysis of the yeast proteome by multidimensional
that are associated with the disease progression [235–237].
protein identification technology. Nat. Biotechnol. 2001,
Phosphorylation of proteins is another well-established PTM
19, 242–247.
that results in redistribution of the modified proteins into
[5] Gramolini, A. O., Kislinger, T., Alikhani-Koopaei, R., Fong,
different subcellular compartments, as well as regulation of
V. et al., Comparative proteomics profiling of a phospho-
signal transduction [238, 239]. The development of proteo-
lamban mutant mouse model of dilated cardiomyopathy
mics technologies to detect PTMs of organelle proteome that reveals progressive intracellular stress responses. Mol.
contributes to the translocation of proteins between Cell. Proteomics 2008, 7, 519–533.
subcellular locations is an important aspect to be addressed.
[6] Schirmer, E. C., Florens, L., Guan, T., Yates , J. R., 3rd,
Gerace, L., Nuclear membrane proteins with potential
disease links found by subtractive proteomics. Science
6 Concluding remarks 2003, 301, 1380–1382.
[7] Pagliarini, D. J., Calvo, S. E., Chang, B., Sheth, S. A. et al.,
Organellar proteomics aims to determine protein localiza- A mitochondrial protein compendium elucidates complex I
tion and the associated biological functions of these disease biology. Cell 2008, 134, 112–123.
proteins. Classical subcellular fractionation methods that [8] Jung, E., Heller, M., Sanchez, J. C., Hochstrasser, D. F.,
have been widely applied in many studies to isolate cellular Proteomics meets cell biology: the establishment
organelles have limited resolution, thus resulting in erro- of subcellular proteomes. Electrophoresis 2000, 21,
neous assignment due to cross-contaminating proteins. 3369–3377.
Despite the developments in bioinformatics and computa- [9] Huber, L. A., Pfaller, K., Vietor, I., Organelle proteomics:
tional tools to predict localization of proteins, these are not implications for subcellular fractionation in proteomics.
totally reliable and may not be applicable for all gene Circ. Res. 2003, 92, 962–968.
products. It is thus necessary to validate assignment of [10] De Duve, C., Pressman, B. C., Gianetto, R., Wattiaux, R.,
protein localization derived from subcellular fractionation Appelmans, F., Tissue fractionation studies. 6. Intracellular
and in silico predictions using independent techniques. distribution patterns of enzymes in rat-liver tissue.
However, the lack of high-throughput methods for this Biochem. J. 1955, 60, 604–617.
purpose is a bottleneck for the validation of proteomics data. [11] de Araujo, M. E., Huber, L. A., Stasyk, T., Isolation of
It has been decades since cellular organelles were studied endocitic organelles by density gradient centrifugation.
using microscopy and subcellular fractionation but protein Methods Mol. Biol. 2008, 424, 317–331.
cataloguing of every cellular compartment is not yet [12] Graham, J. M., Ford, T., Rickwood, D., Isolation of the
complete. One of the challenges in this field is to fully major subcellular organelles from mouse liver using
characterize all subcellular proteomes, and understand the Nycodenz gradients without the use of an ultracentrifuge.
dynamic changes in protein localization in response to Anal. Biochem. 1990, 187, 318–323.
various perturbations. With the recent advancement of new [13] Graham, J. M., Rickwood, D. (Eds.), Subcellular fractiona-
proteomics technologies and strategies, such as FFE, tion: A Practical Approach, IRL Press at Oxford University
immunopurification, FAOS, LOPIT, PCP and subtractive Press, Oxford, New York 1997.
proteomics, which are designed specifically for the [14] Ramsby, M. L., Makowski, G. S., Khairallah, E. A., Differential
systematic isolation and characterization of various subcel- detergent fractionation of isolated hepatocytes: biochemical,
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
3950 Y. H. Lee et al. Proteomics 2010, 10, 3935–3956
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
Proteomics 2010, 10, 3935–3956 3951
phagosome proteome. Mol. Cell. Proteomics 2010, 9, [59] Andersen, J. S., Wilkinson, C. J., Mayor, T., Mortensen, P.
32–53. et al., Proteomic characterization of the human centro-
[45] Simpson, R. J., Jensen, S. S., Lim, J. W., Proteomic some by protein correlation profiling. Nature 2003, 426,
profiling of exosomes: current perspectives. Proteomics 570–574.
2008, 8, 4083–4099. [60] Forner, F., Foster, L. J., Campanaro, S., Valle, G., Mann, M.,
[46] Cantin, R., Diou, J., Belanger, D., Tremblay, A. M., Gilbert, Quantitative proteomic comparison of rat mitochondria
C., Discrimination between exosomes and HIV-1: purifica- from muscle, heart, and liver. Mol. Cell. Proteomics 2006,
tion of both vesicles from cell-free supernatants. J. 5, 608–619.
Immunol. Methods 2008, 338, 21–30. [61] Zhou, Z., Licklider, L. J., Gygi, S. P., Reed, R., Compre-
[47] Coren, L. V., Shatzer, T., Ott, D. E., CD45 immunoaffinity hensive proteomic analysis of the human spliceosome.
depletion of vesicles from Jurkat T cells demonstrates that Nature 2002, 419, 182–185.
exosomes contain CD45: no evidence for a distinct [62] Neubauer, G., King, A., Rappsilber, J., Calvio, C. et al.,
exosome/HIV-1 budding pathway. Retrovirology 2008, 5, Mass spectrometry and EST-database searching allows
64. characterization of the multi-protein spliceosome
[48] Valadi, H., Ekstrom, K., Bossios, A., Sjostrand, M. et al., complex. Nat. Genet. 1998, 20, 46–50.
Exosome-mediated transfer of mRNAs and microRNAs is a [63] Wilkie, G. S., Schirmer, E. C., Guilt by association: the
novel mechanism of genetic exchange between cells. nuclear envelope proteome and disease. Mol. Cell.
Nature Cell Biol. 2007, 9, 654–659. Proteomics 2006, 5, 1865–1875.
[49] Bock, G., Steinlein, P., Huber, L. A., Cell biologists sort [64] Worman, H. J., Courvalin, J. C., The nuclear lamina and
things out: analysis and purification of intracellular orga- inherited disease. Trends Cell Biol. 2002, 12, 591–598.
nelles by flow cytometry. Trends Cell Biol. 1997, 7,
[65] Calvo, S., Jain, M., Xie, X., Sheth, S. A. et al., Systematic
499–503.
identification of human mitochondrial disease genes
[50] Gauthier, D. J., Sobota, J. A., Ferraro, F., Mains, R. E., through integrative genomics. Nat. Genet. 2006, 38,
Lazure, C., Flow cytometry-assisted purification and 576–582.
proteomic analysis of the corticotropes dense-core secre-
[66] Perocchi, F., Jensen, L. J., Gagneur, J., Ahting, U. et al.,
tory granules. Proteomics 2008, 8, 3848–3861.
Assessing systems properties of yeast mitochondria
[51] Dhandayuthapani, S., Via, L. E., Thomas, C. A., Horowitz, through an interaction map of the organelle. PLoS Genet.
P. M. et al., Green fluorescent protein as a marker for 2006, 2, e170.
gene expression and cell biology of mycobacterial
[67] Aiyar, R. S., Gagneur, J., Steinmetz, L. M., Identification of
interactions with macrophages. Mol. Microbiol. 1995, 17,
mitochondrial disease genes through integrative analysis
901–912.
of multiple datasets. Methods 2008, 46, 248–255.
[52] Cossarizza, A., Ceccarelli, D., Masini, A., Functional
[68] Kloepper, T. H., Kienle, C. N., Fasshauer, D., An elaborate
heterogeneity of an isolated mitochondrial population
classification of SNARE proteins sheds light on the
revealed by cytofluorometric analysis at the single orga-
conservation of the eukaryotic endomembrane system.
nelle level. Exp. Cell Res. 1996, 222, 84–94.
Mol. Biol Cell 2007, 18, 3463–3471.
[53] Wilson, R. B., Murphy, R. F., Flow-cytometric analysis of
[69] Gray, M. W., Burger, G., Lang, B. F., Mitochondrial evolu-
endocytic compartments. Methods Cell Biol.1989, 31,
tion. Science 1999, 283, 1476–1481.
293–317.
[70] Perocchi, F., Mancera, E., Steinmetz, L. M., Systematic
[54] Cao, Z., Li, C., Higginbotham, J. N., Franklin, J. L. et al., Use
screens for human disease genes, from yeast to human
of fluorescence-activated vesicle sorting for isolation of
and back. Mol. Biosys. 2008, 4, 18–29.
Naked2-associated, basolaterally targeted exocytic vesi-
cles for proteomics analysis. Mol. Cell. Proteomics 2008, 7, [71] Neupert, W., Herrmann, J. M., Translocation of proteins
1651–1667. into mitochondria. Annu. Rev. Biochem. 2007, 76, 723–749.
[55] Dunkley, T. P., Hester, S., Shadforth, I. P., Runions, J. et al., [72] Mootha, V. K., Handschin, C., Arlow, D., Xie, X. et al., Erra
Mapping the Arabidopsis organelle proteome. Proc. Natl. and Gabpa/b specify PGC-1a-dependent oxidative
Acad. Sci. USA 2006, 103, 6518–6523. phosphorylation gene expression that is altered in
diabetic muscle. Proc. Natl. Acad. Sci. USA 2004, 101,
[56] Dunkley, T. P., Watson, R., Griffin, J. L., Dupree, P., Lilley,
6570–6575.
K. S., Localization of organelle proteins by isotope tagging
(LOPIT). Mol. Cell. Proteomics 2004, 3, 1128–1134. [73] Ozawa, T., Umezawa, Y., A genetic method to identify
mitochondrial proteins in living mammalian cells. Meth-
[57] Sadowski, P. G., Dunkley, T. P., Shadforth, I. P., Dupree, P.
et al., Quantitative proteomic approach to study subcel- ods in Mol. Biol. 2007, 390, 119–130.
lular localization of membrane proteins. Nat. Protoc. 2006, [74] Stuart, L. M., Boulais, J., Charriere, G. M., Hennessy, E. J.
1, 1778–1789. et al., A systems biology analysis of the Drosophila
[58] Wiese, S., Gronemeyer, T., Ofman, R., Kunze, M. et al., phagosome. Nature 2007, 445, 95–101.
Proteomics characterization of mouse kidney peroxisomes [75] Huh, W. K., Falvo, J. V., Gerke, L. C., Carroll, A. S. et al.,
by tandem mass spectrometry and protein correlation Global analysis of protein localization in budding yeast.
profiling. Mol. Cell. Proteomics 2007, 6, 2045–2057. Nature 2003, 425, 686–691.
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
3952 Y. H. Lee et al. Proteomics 2010, 10, 3935–3956
[76] Chou, K. C., Shen, H. B., Cell-PLoc: a package of Web transit peptides and their cleavage sites. Protein Sci. 1999,
servers for predicting subcellular localization of proteins in 8, 978–984.
various organisms. Nat. Protoc. 2008, 3, 153–162. [95] Cokol, M., Nair, R., Rost, B., Finding nuclear localization
[77] Davis, T. N., Protein localization in proteomics. Curr. Opin. signals. EMBO Rep. 2000, 1, 411–415.
Chem. Biol. 2004, 8, 49–53. [96] Hoglund, A., Donnes, P., Blum, T., Adolph, H. W., Kohl-
[78] Schneider, G., Fechner, U., Advances in the prediction of bacher, O., MultiLoc: prediction of protein subcellular
protein targeting signals. Proteomics 2004, 4, 1571–1580. localization using N-terminal targeting sequences,
[79] Emanuelsson, O., Nielsen, H., Brunak, S., von Heijne, G., sequence motifs and amino acid composition. Bioinfor-
Predicting subcellular localization of proteins based on matics 2006, 22, 1158–1165.
their N-terminal amino acid sequence. J. Mol. Biol. 2000, [97] Wang, J., Sung, W. K., Krishnan, A., Li, K. B., Protein
300, 1005–1016. subcellular localization prediction for Gram-negative
[80] Nakai, K., Horton, P., PSORT: a program for detecting bacteria using amino acid subalphabets and a combina-
sorting signals in proteins and predicting their subcellular tion of multiple support vector machines. BMC Bioinfor-
localization. Trends Biochem. Sci. 1999, 24, 34–36. matics 2005, 6, 174.
[81] Scott, M. S., Thomas, D. Y., Hallett, M. T., Predicting [98] Chang, J. M., Su, E. C., Lo, A., Chiu, H. S. et al., PSLDoc:
subcellular localization via protein motif co-occurrence. protein subcellular localization prediction based on
Genome Res. 2004, 14, 1957–1966. gapped-dipeptides and probabilistic latent semantic
analysis. Proteins 2008, 72, 693–710.
[82] Shatkay, H., Hoglund, A., Brady, S., Blum, T. et al.,
SherLoc: high-accuracy prediction of protein subcellular [99] Cedano, J., Aloy, P., Perez-Pons, J. A., Querol, E., Relation
localization by integrating text and protein sequence between amino acid composition and cellular location of
data. Bioinformatics 2007, 23, 1410–1417. proteins. J. Mol. Biol. 1997, 266, 594–600.
[83] Bhasin, M., Raghava, G. P., ESLpred: SVM-based method [100] Park, K. J., Kanehisa, M., Prediction of protein subcellular
for subcellular localization of eukaryotic proteins using locations by support vector machines using compositions
dipeptide composition and PSI-BLAST. Nucleic Acids Res. of amino acids and amino acid pairs. Bioinformatics 2003,
2004, 32, W414–W419. 19, 1656–1663.
[84] Sprenger, J., Fink, J. L., Teasdale, R. D., Evaluation and [101] Huang, Y., Li, Y., Prediction of protein subcellular locations
comparison of mammalian subcellular localization using fuzzy k-NN method. Bioinformatics 2004, 20, 21–28.
prediction methods. BMC Bioinformatics 2006, 7, S3. [102] Chou, K. C., Cai, Y. D., Using functional domain composi-
[85] Donnes, P., Hoglund, A., Predicting protein subcellular tion and support vector machines for prediction of protein
localization: past, present, and future. Genomics Proteo- subcellular location. J. Biol. Chem. 2002, 277, 45765–45769.
mics Bioinformatics 2004, 2, 209–215. [103] Mott, R., Schultz, J., Bork, P., Ponting, C. P., Predicting
[86] Chou, K. C., Shen, H. B., Recent progress in protein subcel- protein cellular localization using a domain projection
lular location prediction. Anal. Biochem. 2007, 370, 1–16. method. Genome Res. 2002, 12, 1168–1174.
[87] Emanuelsson, O., Brunak, S., von Heijne, G., Nielsen, H., [104] Huang, W. L., Tung, C. W., Ho, S. W., Hwang, S. F., Ho,
Locating proteins in the cell using TargetP, SignalP and S. Y., ProLoc-GO: utilizing informative Gene Ontology
related tools. Nat. Protoc. 2007, 2, 953–971. terms for sequence-based prediction of protein subcellular
localization. BMC Bioinformatics 2008, 9, 80.
[88] Emanuelsson, O., von Heijne, G., Prediction of organellar
targeting signals. Biochim. Biophys. Acta 2001, 1541, 114–119. [105] Nair, R., Rost, B., Mimicking cellular sorting improves
prediction of subcellular localization. J. Mol. Biol. 2005,
[89] Jagla, B., Schuchhardt, J., Adaptive encoding neural
348, 85–100.
networks for the recognition of human signal peptide
cleavage sites. Bioinformatics 2000, 16, 245–250. [106] Chou, K. C., Cai, Y. D., Predicting protein localization in
budding yeast. Bioinformatics 2005, 21, 944–950.
[90] Bendtsen, J. D., Nielsen, H., von Heijne, G., Brunak, S.,
Improved prediction of signal peptides: SignalP 3.0. J. [107] Bhasin, M., Garg, A., Raghava, G. P., PSLpred: prediction
Mol. Biol. 2004, 340, 783–795. of subcellular localization of bacterial proteins. Bioinfor-
matics 2005, 21, 2522–2524.
[91] Petsalaki, E. I., Bagos, P. G., Litou, Z. I., Hamodrakas, S. J.,
PredSL: a tool for the N-terminal sequence-based predic- [108] Gardy, J. L., Laird, M. R., Chen, F., Rey, S. et al., PSORTb
tion of protein subcellular localization. Genomics Proteo- v.2.0: expanded prediction of bacterial protein subcellular
mics Bioinformatics 2006, 4, 48–55. localization and insights gained from comparative
[92] Small, I., Peeters, N., Legeai, F., Lurin, C., Predotar: a tool proteome analysis. Bioinformatics 2005, 21, 617–623.
for rapidly screening proteomes for N-terminal targeting [109] Nair, R., Rost, B., Sequence conserved for subcellular
sequences. Proteomics 2004, 4, 1581–1590. localization. Protein Sci. 2002, 11, 2836–2847.
[93] Claros, M. G., Vincens, P., Computational method to [110] Yu, C. S., Chen, Y. C., Lu, C. H., Hwang, J. K., Prediction of
predict mitochondrially imported proteins and their protein subcellular localization. Proteins 2006, 64, 643–651.
targeting sequences. Eur. J. Biochem. 1996, 241, 779–786. [111] Lu, Z., Szafron, D., Greiner, R., Lu, P. et al., Predicting
[94] Emanuelsson, O., Nielsen, H., von Heijne, G., ChloroP, a subcellular localization of proteins using machine-learned
neural network-based method for predicting chloroplast classifiers. Bioinformatics 2004, 20, 547–556.
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
Proteomics 2010, 10, 3935–3956 3953
[112] Marcotte, E. M., Xenarios, I., van Der Bliek, A. M., Eisen- [128] Christophe, D., Christophe-Hobertus, C., Pichon, B.,
berg, D., Localizing proteins in the cell from their phylo- Nuclear targeting of proteins: how many different signals?
genetic profiles. Proc. Natl. Acad. Sci. USA 2000, 97, Cell Signal. 2000, 12, 337–341.
12115–12120. [129] Gould, S. J., Keller, G. A., Hosken, N., Wilkinson, J.,
[113] Horton, P., Park, K. J., Obayashi, T., Fujita, N. et al., WoLF Subramani, S., A conserved tripeptide sorts proteins to
PSORT: protein localization predictor. Nucleic Acids Res. peroxisomes. J. Cell Biol. 1989, 108, 1657–1664.
2007, 35, W585–587. [130] Nielsen, H., Engelbrecht, J., Brunak, S., von Heijne, G., Iden-
[114] Bannai, H., Tamada, Y., Maruyama, O., Nakai, K., Miyano, tification of prokaryotic and eukaryotic signal peptides and
S., Extensive feature detection of N-terminal protein sort- prediction of their cleavage sites. Protein Eng. 1997, 10, 1–6.
ing signals. Bioinformatics (Oxford, England) 2002, 18, [131] Ni, L., Heard, T. S., Weiner, H., In vivo mitochondrial import. A
298–305. comparison of leader sequence charge and structural rela-
[115] Drawid, A., Gerstein, M., A Bayesian system integrating tionships with the in vitro model resulting in evidence for co-
expression data with sequence patterns for localizing translational import. J. Biol. Chem. 1999, 274, 12685–12691.
proteins: comprehensive application to the yeast genome. [132] Kleffmann, T., Russenberger, D., von Zychlinski, A.,
J. Mol. Biol. 2000, 301, 1059–1075. Christopher, W. et al., The Arabidopsis thaliana chlor-
[116] Feng, Z. P., Zhang, C. T., A graphic representation of oplast proteome reveals pathway abundance and novel
protein sequence and predicting the subcellular locations protein functions. Curr. Biol. 2004, 14, 354–362.
of prokaryotic proteins. Int. J. Biochem. Cell. Biol. 2002, 34, [133] Eisenhaber, F., Bork, P., Wanted: subcellular localization of
298–307. proteins based on sequence. Trends Cell Biol. 1998, 8,
[117] Chou, K. C., Cai, Y. D., Prediction of protein subcellular 169–170.
locations by GO-FunD-PseAA predictor. Biochem. [134] Takatalo, M. S., Kouvonen, P., Corthals, G., Nyman, T. A.,
Biophys. Res. Commun. 2004, 320, 1236–1239. Ronnholm, R. H., Identification of new Golgi complex
[118] Nakai, K., Kanehisa, M., Expert system for predicting specific proteins by direct organelle proteomic analysis.
protein localization sites in Gram-negative bacteria. Proteomics 2006, 6, 3502–3508.
Proteins 1991, 11, 95–110. [135] Reinhardt, A., Hubbard, T., Using neural networks for
[119] Kaundal, R., Raghava, G. P., RSLpred: an integrative prediction of the subcellular location of proteins. Nucleic
system for predicting subcellular localization of rice Acids Res. 1998, 26, 2230–2236.
proteins combining compositional and evolutionary [136] Cai, Y. D., Chou, K. C., Predicting subcellular localization of
information. Proteomics 2009, 9, 2324–2342. proteins in a hybridization space. Bioinformatics 2004, 20,
[120] Briesemeister, S., Blum, T., Brady, S., Lam, Y. et al., 1151–1156.
SherLoc2: a high-accuracy hybrid method for predicting [137] Bairoch, A., Apweiler, R., The SWISS-PROT protein
subcellular localization of proteins. J. Proteome Res. 2009, sequence data bank and its supplement TrEMBL. Nucleic
8, 5363–5366. Acids Res. 1997, 25, 31–36.
[121] Guda, C., Fahy, E., Subramaniam, S., MITOPRED: a [138] Xu, Q., Hu, D. H., Xue, H., Yu, W., Yang, Q., Semi-super-
genome-scale method for prediction of nucleus-encoded vised protein subcellular localization. BMC Bioinformatics
mitochondrial proteins. Bioinformatics 2004, 20, 1785–1794. 2009, 10, S47.
[122] Guda, C., Guda, P., Fahy, E., Subramaniam, S., MITOPRED: [139] Park, S., Yang, J. S., Jang, S. K., Kim, S., Construction of
a web server for the prediction of mitochondrial proteins. functional interaction networks through consensus locali-
Nucleic Acids Res. 2004, 32, W372–374. zation predictions of the human proteome. J. Proteome
[123] Gaston,D., Tsaousis,A. D., Roger,A. J., Predicting Res. 2009, 8, 3367–3376.
proteomes of mitochondria and related organelles from [140] Guda, C., Subramaniam, S., pTARGET [corrected] a new
genomic and expressed sequence tag data. Methods method for predicting protein subcellular localization in
Enzymol. 2009, 457, 21–47. eukaryotes. Bioinformatics 2005, 21, 3963–3969.
[124] Shen,H. B., Chou,K. C., PseAAC: a flexible web server for [141] Neuberger, G., Maurer-Stroh, S., Eisenhaber, B., Hartig, A.,
generating various kinds of protein pseudo amino acid Eisenhaber, F., Prediction of peroxisomal targeting signal
composition. Anal. Biochem. 2008, 373, 386–388. 1 containing proteins from amino acid sequence. J. Mol.
[125] Cherian, B. S., Nair, A. S., Protein location prediction using Biol. 2003, 328, 581–592.
atomic composition and global features of the amino acid [142] Chou, K. C., Shen, H. B., Large-scale predictions of
sequence. Biochem. Biophys. Res. Commun. 2010, 391, gram-negative bacterial protein subcellular locations.
1670–1674. J. Proteome Res. 2006, 5, 3420–3428.
[126] Hua, S., Sun, Z., Support vector machine approach for [143] Zhang, S., Xia, X., Shen, J., Zhou, Y., Sun, Z., DBMLoc: a
protein subcellular localization prediction. Bioinformatics Database of proteins with multiple subcellular localiza-
2001, 17, 721–728. tions. BMC Bioinformatics 2008, 9, 127.
[127] Pfanner, N., Geissler, A., Versatility of the mitochondrial [144] King, B. R., Guda, C., ngLOC: an n-gram-based Bayesian
protein import machinery. Nat. Rev. Mol. Cell. Biol. 2001, method for estimating the subcellular proteomes of
2, 339–349. eukaryotes. Genome Biol. 2007, 8, R68.
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
3954 Y. H. Lee et al. Proteomics 2010, 10, 3935–3956
[145] Millar, A. H., Whelan, J., Small, I., Recent surprises in [162] Cotter, D., Guda, P., Fahy, E., Subramaniam, S., MitoPro-
protein targeting to mitochondria and plastids. Curr. Opin. teome: mitochondrial protein sequence database and
Plant Biol. 2006, 9, 610–615. annotation system. Nucleic Acids Res. 2004, 32, D463–467.
[146] Small, I., Wintz, H., Akashi, K., Mireau, H., Two birds with [163] Elstner, M., Andreoli, C., Klopstock, T., Meitinger, T.,
one stone: genes that encode products targeted to two or Prokisch, H., The mitochondrial proteome database:
more compartments. Plant Mol. Biol. 1998, 38, 265–277. MitoP2. Methods Enzymol. 2009, 457, 3–20.
[147] Chou, K. C., Shen, H. B., Euk-mPLoc: a fusion classifier for [164] Prokisch, H., Andreoli, C., Ahting, U., Heiss, K. et al., MitoP2:
large-scale eukaryotic protein subcellular location predic- the mitochondrial proteome database – now including
tion by incorporating multiple sites. J. Proteome Res. mouse data. Nucleic Acids Res. 2006, 34, D705–711.
2007, 6, 1728–1734. [165] Sant’Anna, C., Nakayasu, E. S., Pereira, M. G., Lourenco, D.
[148] Lin, H. N., Chen, C. T., Sung, T. Y., Ho, S. Y., Hsu, W. L., et al., Subcellular proteomics of Trypanosoma cruzi
Protein subcellular localization prediction of eukaryotes reservosomes. Proteomics 2009, 9, 1782–1794.
using a knowledge-based approach. BMC Bioinformatics [166] Casey, T. M., Meade, J. L., Hewitt, E. W., Organelle
2009, 10, S8. proteomics: identification of the exocytic machinery
[149] Boore, J. L., Requirements and standards for organelle associated with the natural killer cell secretory lysosome.
genome databases. OMICS 2006, 10, 119–126. Mol. Cell. Proteomics 2007, 6, 767–780.
[150] Simpson, J. C., Pepperkok, R., The subcellular localization [167] Kislinger, T., Cox, B., Kannan, A., Chung, C. et al., Global
of the mammalian proteome comes a fraction closer. survey of organ and organelle protein expression in
Genome Biol. 2006, 7, 222. mouse: combined proteomic and transcriptomic profiling.
[151] Lopez, M. F., Kristal, B. S., Chernokalskaya, E., Lazarev, A. Cell 2006, 125, 173–186.
et al., High-throughput profiling of the mitochondrial [168] Song, Y., Hao, Y., Sun, A., Li, T. et al., Sample preparation
proteome using affinity fractionation and automation. project for the subcellular proteome of mouse liver.
Electrophoresis 2000, 21, 3427–3440. Proteomics 2006, 6, 5269–5277.
[152] Peltier, J. B., Friso, G., Kalume, D. E., Roepstorff, P. et al., [169] Mann, M., Can proteomics retire the western blot? J.
Proteomics of the chloroplast: systematic identification Proteome Res. 2008, 7, 3065.
and targeting analysis of lumenal and peripheral thylakoid [170] Anderson, L., Hunter, C. L., Quantitative mass spectro-
proteins. Plant Cell 2000, 12, 319–341. metric multiple reaction monitoring assays for major
[153] Andersen, J. S., Lyon, C. E., Fox, A. H., Leung, A. K. et al., plasma proteins. Mol. Cell. Proteomics 2006, 5, 573–588.
Directed proteomic analysis of the human nucleolus. Curr. [171] Sandhu, C., Hewel, J. A., Badis, G., Talukder, S. et al.,
Biol. 2002, 12, 1–11. Evaluation of data-dependent versus targeted shotgun
[154] International Human Genome Sequencing Consortium. proteomic approaches for monitoring transcription factor
Finishing the euchromatic sequence of the human expression in breast cancer. J. Proteome Res. 2008, 7,
genome. Nature 2004, 431, 931–945. 1529–1541.
[155] Pennisi, E., Genetics. Working the (gene count) numbers: [172] Tribl, F., Asan, E., Arzberger, T., Tatschner, T. et al., Iden-
finally, a firm answer? Science 2007, 316, 1113. tification of L-ferritin in neuromelanin granules of the
[156] Sickmann, A., Reinders, J., Wagner, Y., Joppich, C. et al., human substantia nigra: a targeted proteomics approach.
The proteome of Saccharomyces cerevisiae mitochondria. Mol. Cell. Proteomics 2009, 8, 1832–1838.
Proc. Natl. Acad. Sci. USA 2003, 100, 13207–13212. [173] Miller, C. A., Horn, D., Torvi, S., Tang, N., Waddell, K.,
[157] Taylor, S. W., Fahy, E., Zhang, B., Glenn, G. M. et al., Transfer of optimized acquisition parameters between mass
Characterization of the human heart mitochondrial analyzer types for improved protein identification and quan-
proteome. Nat. Biotechnol. 2003, 21, 281–286. tification. J. Am. Soc. Mass Spectrom. 2009, Poster MPA015.
[158] Mootha, V. K., Bunkenborg, J., Olsen, J. V., Hjerrild, M. [174] Andreyev, A. Y., Shen, Z., Guan, Z., Ryan, A. et al., Applica-
et al., Integrated analysis of protein composition, tissue tion of proteomic marker ensembles to subcellular organelle
diversity, and gene regulation in mouse mitochondria. Cell identification. Mol. Cell. Proteomics 2010, 9, 388–402.
2003, 115, 629–640. [175] Brunner, Y., Coute, Y., Iezzi, M., Foti, M. et al., Proteomics
[159] Taylor, S. W., Fahy, E., Ghosh, S. S., Global organellar analysis of insulin secretory granules. Mol. Cell. Proteo-
proteomics. Trends Biotechnol. 2003, 21, 82–88. mics 2007, 6, 1007–1017.
[160] Basu, S., Bremer, E., Zhou, C., Bogenhagen, D. F., [176] Seibel, N. M., Eljouni, J., Nalaskowski, M. M., Hampe, W.,
MiGenes: a searchable interspecies database of mito- Nuclear localization of enhanced green fluorescent protein
chondrial proteins curated using gene ontology homomultimers. Anal. Biochem. 2007, 368, 95–99.
annotation. Bioinformatics (Oxford, England) 2006, 22, [177] Piedras, P., Rivas, S., Droge, S., Hillmer, S., Jones, J. D.,
485–492. Functional, c-myc-tagged Cf-9 resistance gene products
[161] Smith, A. C., Robinson, A. J., MitoMiner, an integrated are plasma-membrane localized and glycosylated. Plant J.
database for the storage and analysis of mito- 2000, 21, 529–536.
chondrial proteomics data. Mol. Cell. Proteomics 2009, [178] Simpson, J. C., Wellenreuther, R., Poustka, A., Pepperkok,
8, 1324–1337. R., Wiemann,S., Systematic subcellular localization of
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
Proteomics 2010, 10, 3935–3956 3955
novel proteins identified by large-scale cDNA sequencing. [195] Matsuyama, A., Arai, R., Yashiroda, Y., Shirai, A. et al.,
EMBO Rep. 2000, 1, 287–292. ORFeome cloning and global analysis of protein localiza-
[179] Rial, D. V., Lombardo, V. A., Ceccarelli, E. A., Ottado, J., tion in the fission yeast Schizosaccharomyces pombe.
The import of ferredoxin-NADP1reductase precursor into Nature Biotechnol. 2006, 24, 841–847.
chloroplasts is modulated by the region between the [196] Proszynski, T. J., Klemm, R. W., Gravert, M., Hsu, P. P.
transit peptide and the mature core of the protein. Eur. J. et al., A genome-wide visual screen reveals a role
Biochem. 2002, 269, 5431–5439. for sphingolipids and ergosterol in cell surface delivery
[180] Zhou, J., Weiner, H., The N-terminal portion of mature in yeast. Proc. Natl. Acad. Sci. U.S.A. 2005, 102,
aldehyde dehydrogenase affects protein folding and 17981–17986.
assembly. Protein Sci. 2001, 10, 1490–1497. [197] Natter, K., Leitner, P., Faschinger, A., Wolinski, H. et al.,
[181] Chew, O., Whelan, J., Millar, A. H., Molecular definition of The spatial organization of lipid synthesis in the yeast
the ascorbate-glutathione cycle in Arabidopsis mitochon- Saccharomyces cerevisiae derived from large scale green
dria reveals dual targeting of antioxidant defenses in fluorescent protein tagging and high resolution micro-
plants. J. Biol. Chem. 2003, 278, 46869–46877. scopy. Mol. Cell. Proteomics 2005, 4, 662–672.
[182] Ma, C., Mitra, A., Intrinsic direct repeats generate consis- [198] Barbe, L., Lundberg, E., Oksvold, P., Stenius, A. et al.,
tent post-transcriptional gene silencing in tobacco. Plant J. Toward a confocal subcellular atlas of the human
2002, 31, 37–49. proteome. Mol. Cell. Proteomics 2008, 7, 499–508.
[183] Rohila, J. S., Chen, M., Cerny, R., Fromm, M. E., Improved [199] Boland, M. V., Murphy, R. F., A neural network classifier
tandem affinity purification tag and methods for isolation capable of recognizing the patterns of all major subcellular
of protein heterocomplexes from plants. Plant J. 2004, 38, structures in fluorescence microscope images of HeLa
172–181. cells. Bioinformatics 2001, 17, 1213–1223.
[184] Haynes, P. A., Cross, G. A., Differential glycosylation of [200] Conrad, C., Erfle, H., Warnat, P., Daigle, N. et al., Automatic
epitope-tagged glycoprotein Gp72 during the Trypanosoma identification of subcellular phenotypes on human cell
cruzi life cycle. Mol. Biochem. Parasitol. 1996, 83, 253–256. arrays. Genome Res. 2004, 14, 1130–1136.
[185] Kumar, A., Agarwal, S., Heyman, J. A., Matson, S. et al., [201] Hamilton, N. A., Pantelic, R. S., Hanson, K., Teasdale, R. D.,
Subcellular localization of the yeast proteome. Genes Dev. Fast automated cell phenotype image classification. BMC
2002, 16, 707–719. Bioinformatics 2007, 8, 110.
[186] Jarvik, J. W., Fisher, G. W., Shi, C., Hennen, L. et al., In vivo [202] Chen, X., Velliste, M., Murphy, R. F., Automated interpretation
functional proteomics: mammalian genome annotation of subcellular patterns in fluorescence microscope images for
using CD-tagging. Biotechniques 2002, 33, 852–854, 856, location proteomics. Cytometry A 2006, 69, 631–640.
858–860 passim. [203] Glory, E., Murphy, R. F., Automated subcellular location
[187] Jarvik, J. W., Adler, S. A., Telmer, C. A., Subramaniam, V., determination and high-throughput microscopy. Dev. Cell
Lopez, A. J., CD-tagging: a new approach to gene and protein 2007, 12, 7–16.
discovery and analysis. Biotechniques 1996, 20, 896–904. [204] Perlman, Z. E., Slack, M. D., Feng, Y., Mitchison, T. J. et al.,
[188] Hazbun, T. R., Malmstrom, L., Anderson, S., Graczyk, B. J. Multidimensional drug profiling by automated micro-
et al., Assigning function to yeast proteins by integration scopy. Science 2004, 306, 1194–1198.
of technologies. Mol. Cell 2003, 12, 1353–1365. [205] Zhao, T., Murphy, R. F., Automated learning of generative
[189] Nehlsen, K., Schucht, R., da Gama-Norton, L., Kromer, W. models for subcellular location: building blocks for
et al., Recombinant protein expression by targeting pre- systems biology. Cytometry A 2007, 71, 978–990.
selected chromosomal loci. BMC Biotechnol. 2009, 9, 100. [206] Schubert, W., Bonnekoh, B., Pommer, A. J., Philipsen, L.
[190] Walker, D., Htun, H., Hager, G. L., Using inducible vectors to et al., Analyzing proteome topology and function by
study intracellular trafficking of GFP-tagged steroid/nuclear automated multidimensional fluorescence microscopy.
receptors in living cells. Methods 1999, 19, 386–393. Nat. Biotechnol. 2006, 24, 1270–1278.
[191] Wiemann, S., Arlt, D., Huber, W., Wellenreuther, R. et al., [207] Starkuviene, V., Liebel, U., Simpson, J. C., Erfle, H. et al.,
From ORFeome to biology: a functional genomics pipe- High-content screening microscopy identifies novel
line. Genome Res. 2004, 14, 2136–2144. proteins with a putative role in secretory membrane traffic.
[192] Ohara, O., ORFeome cloning. Methods Mol. Biol. 2009, Genome Res. 2004, 14, 1948–1956.
577, 3–9. [208] Newberg, J., Hua, J., Murphy, R. F., Location proteomics:
[193] Nagase, T., Yamakawa, H., Tadokoro, S., Nakajima, D. systematic determination of protein subcellular location.
et al., Exploration of human ORFeome: high-throughput Methods Mol. Biol. 2009, 500, 313–332.
preparation of ORF clones and efficient characterization of [209] Uhlen, M., Bjorling, E., Agaton, C., Szigyarto, C. A. et al.,
their protein products. DNA Res. 2008, 15, 137–149. A human protein atlas for normal and cancer tissues based
[194] Lam, Y. W., Evans, V. C., Heesom, K. J., Lamond, A. I., on antibody proteomics. Mol. Cell. Proteomics 2005, 4,
Matthews, D. A., Proteomics analysis of the nucleolus in 1920–1932.
adenovirus-infected cells. Mol. Cell. Proteomics 2010, 9, [210] Buck, T. E., Rao, A., Coelho, L. P., Fuhrman, M. H. et al.,
117–130. Cell cycle dependence of protein subcellular location
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com
3956 Y. H. Lee et al. Proteomics 2010, 10, 3935–3956
inferred from static, asynchronous images. Conf. Proc. [225] Petrova, V. Y., Drescher, D., Kujumdzieva, A. V., Schmitt,
IEEE Eng. Med. Biol. Soc. 2009, 721–1078, 1016–1019. M. J., Dual targeting of yeast catalase A to peroxisomes
[211] Helmuth, J. A., Burckhardt, C. J., Greber, U. F., Sbalzarini, and mitochondria. Biochem. J. 2004, 380, 393–400.
I. F., Shape reconstruction of subcellular structures from [226] Rizzuto, R., Pozzan, T., Microdomains of intracellular Ca21
live cell fluorescence microscopy images. J. Struct. Biol. : molecular determinants and functional consequences.
2009, 167, 1–10. Physiol. Rev. 2006, 86, 369–408.
[212] Agaton, C., Galli, J., Hoiden Guthenberg, I., Janzon, L. [227] Bach, D., Naon, D., Pich, S., Soriano, F. X. et al., Expression
et al., Affinity proteomics for systematic protein profiling of Mfn2, the Charcot-Marie-Tooth neuropathy type 2A
of chromosome 21 gene products in human tissues. Mol. gene, in human skeletal muscle: effects of type 2 diabetes,
Cell. Proteomics 2003, 2, 405–414. obesity, weight loss, and the regulatory role of tumor
[213] Hartwig, S., Feckler, C., Lehr, S., Wallbrecht, K. et al., A critical necrosis factor a and interleukin-6. Diabetes 2005, 54,
comparison between two classical and a kit-based method 2685–2693.
for mitochondria isolation. Proteomics 2009, 9, 3209–3214. [228] Baloh, R. H., Schmidt, R. E., Pestronk, A., Milbrandt, J.,
[214] Coene, E. D., Hollinshead, M. S., Waeytens, A. A., Altered axonal mitochondrial transport in the pathogen-
Schelfhout, V. R. et al., Phosphorylated BRCA1 is predo- esis of Charcot-Marie-Tooth disease from mitofusin 2
minantly located in the nucleus and mitochondria. Mol. mutations. J. Neurosci. 2007, 27, 422–430.
Biol.Cell 2005, 16, 997–1010. [229] Xie, Z., Klionsky, D. J., Autophagosome formation: core
[215] Brocardo, M., Nathke, I. S., Henderson, B. R., Redefining machinery and adaptations. Nat. cell Biol. 2007, 9,
the subcellular location and transport of APC: new insights 1102–1109.
using a panel of antibodies. EMBO Rep. 2005, 6, 184–190.
[230] Dunn , W. A., Jr, Cregg, J. M., Kiel, J. A., van der Klei, I. J.
[216] Hall, S. L., Hester, S., Griffin, J. L., Lilley, K. S., Jackson, et al., Pexophagy: the selective autophagy of peroxisomes.
A. P., The organelle proteome of the DT40 lymphocyte cell Autophagy 2005, 1, 75–83.
line. Mol. Cell. Proteomics 2009, 8, 1295–1305.
[231] Lemasters, J. J., Selective mitochondrial autophagy, or
[217] Hanton, S. L., Bortolotti, L. E., Renna, L., Stefano, G., mitophagy, as a targeted defense against oxidative stress,
Brandizzi, F., Crossing the divide – transport between the mitochondrial dysfunction, and aging. Rejuvenation Res.
endoplasmic reticulum and Golgi apparatus in plants. 2005, 8, 3–5.
Traffic 2005, 6, 267–277.
[232] Rodriguez-Enriquez, S., Kai, Y., Maldonado, E., Currin,
[218] Jurgens, G., Membrane trafficking in plants. Annu. Rev. R. T., Lemasters, J. J., Roles of mitophagy and the mito-
Cell Dev. Biol. 2004, 20, 481–504. chondrial permeability transition in remodeling of cultured
[219] Manunta, M., Izzo, L., Duncan, R., Jones, A. T., Establish- rat hepatocytes. Autophagy 2009, 5, 1–8.
ment of subcellular fractionation techniques to monitor [233] Lubke, T., Lobel, P., Sleat, D. E., Proteomics of the lyso-
the intracellular fate of polymer therapeutics II. Identifica- some. Biochimica Biophysica Acta 2009, 1793, 625–635.
tion of endosomal and lysosomal compartments in HepG2
[234] Gibson, B. W., The human mitochondrial proteome: oxida-
cells combining single-step subcellular fractionation with
tive stress, protein modifications and oxidative phosphor-
fluorescent imaging. J. Drug Target 2007, 15, 37–50.
ylation. Int. J. Biochem. Cell. Biol. 2005, 37, 927–934.
[220] Bryant, N. J., Govers, R., James, D. E., Regulated transport
of the glucose transporter GLUT4. Nat. Rev. Mol. Cell. Biol. [235] Bailey, S. M., Landar, A., Darley-Usmar, V., Mitochondrial
2002, 3, 267–277. proteomics in free radical research. Free Radic. Biol. Med.
2005, 38, 175–188.
[221] Andersen, J. S., Lam, Y. W., Leung, A. K., Ong, S. E. et al.,
Nucleolar proteome dynamics. Nature 2005, 433, 77–83. [236] Brunet, S., Thibault, P., Gagnon, E., Kearney, P. et al.,
Organelle proteomics: looking at less to see more. Trends
[222] Qattan, A. T., Mulvey, C., Crawford, M., Natale, D. A.,
Cell Biol. 2003, 13, 629–638.
Godovac-Zimmermann, J., Quantitative organelle proteo-
mics of MCF-7 breast cancer cells reveals multiple [237] Lopez-Sanchez, L. M., Muntane, J., de la Mata, M., Rodri-
subcellular locations for proteins in cellular functional guez-Ariza, A., Unraveling the S-nitrosoproteome: tools
processes. J. Proteome Res. 2010, 9, 495–508. and strategies. Proteomics 2009, 9, 808–818.
[223] Danial, N. N., Gramm, C. F., Scorrano, L., Zhang, C. Y. [238] Kalume, D. E., Molina, H., Pandey, A., Tackling the phos-
et al., BAD and glucokinase reside in a mitochondrial phoproteome: tools and strategies. Curr. Opin. Chem. Biol.
complex that integrates glycolysis and apoptosis. Nature 2003, 7, 64–69.
2003, 424, 952–956. [239] Spickett, C. M., Pitt, A. R., Morrice, N., Kolch, W., Proteo-
[224] Radi, R., Turrens, J. F., Chang, L. Y., Bush, K. M. et al., mic analysis of phosphorylation, oxidation and nitrosyla-
Detection of catalase in rat heart mitochondria. J. Biol. tion in signal transduction. Biochim. Biophys. Acta 2006,
Chem. 1991, 266, 22028–22034. 1764, 1823–1841.
& 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com