See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/6789598
A microscope-based screening platform for
large-scale functional protein analysis in intact
cells
Article in FEBS Letters · December 2003
DOI: 10.1016/S0014-5793(03)01197-9 · Source: PubMed
CITATIONS
READS
106
22
7 authors, including:
Urban Liebel
Vytaute Starkuviene
56 PUBLICATIONS 1,763 CITATIONS
37 PUBLICATIONS 714 CITATIONS
ACQUIFER AG
SEE PROFILE
Universität Heidelberg
SEE PROFILE
Jeremy C Simpson
Stefan Wiemann
135 PUBLICATIONS 3,947 CITATIONS
364 PUBLICATIONS 7,339 CITATIONS
University College Dublin
SEE PROFILE
German Cancer Research Center
SEE PROFILE
All content following this page was uploaded by Vytaute Starkuviene on 23 June 2014.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
FEBS Letters 554 (2003) 394^398
FEBS 27814
A microscope-based screening platform for large-scale functional protein
analysis in intact cells
Urban Liebela;1 , Vytaute Starkuvienea;1 , Holger Er£ea , Jeremy C. Simpsona ,
Annemarie Poustkab , Stefan Wiemannb , Rainer Pepperkoka;
a
Cell Biology and Cell Biophysics Programme, European Molecular Biology Laboratory, Meyerhofstrasse 1, 69117 Heidelberg, Germany
b
Molecular Genome Analysis, German Cancer Research Centre, Im Neuenheimer Feld 280/506, 69120 Heidelberg, Germany
Received 12 September 2003; accepted 30 September 2003
First published online 22 October 2003
Edited by Felix Wieland
Abstract A modular microscope-based screening platform,
with applications in large-scale analysis of protein function in
intact cells is described. It includes automated sample preparation, image acquisition, data management and analysis, and the
genome-wide automated retrieval of bioinformatic information.
The modular nature of the system ensures that it is rapidly
adaptable to new biological questions or sets of proteins. Two
automated functional assays addressing protein secretion and
the integrity of the Golgi complex were developed and tested.
This shows the potential of the system in large-scale, cell-based
functional proteomic projects.
2003 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved.
Key words: Green £uorescent protein; Proteomics;
Functional analysis; High-content screening microscopy;
Membrane tra⁄c; Golgi complex
1. Introduction
The recent completion of the sequencing of entire genomes
from various organisms provides the potential to systematically study each and every gene in turn. One of the challenges
now is to develop methods that allow large-scale functional
studies of the encoded proteins in their natural environment,
the living cell. Strategies have been devised to address a ¢rst
step in revealing protein function by systematically determining protein localisation using green £uorescent protein (GFP)tagged cDNA libraries or collections of open reading frames
(ORFs) [1^4]. These studies also provide novel tools, the
GFP-tagged cDNAs, for follow-up studies addressing protein
function by £uorescence microscopy, which itself is a powerful
technique for in vivo studies as it gives quantitative access to
the spatial distribution and dynamics of membrane-bounded
organelles, proteins or biochemical reactions in single cells
[5,6].
The problem of performing functional microscope-based
*Corresponding author. Fax: (49)-6221-387 306.
E-mail address: pepperko@embl-heidelberg.de (R. Pepperkok).
1
These authors contributed equally to this work.
Abbreviations: CFP, cyan £uorescent protein; ER, endoplasmic reticulum; GFP, green £uorescent protein; ORF, open reading frame;
YFP, yellow £uorescent protein
assays on a large set of proteins in cells is presently still a
challenge. It requires automation and coordination of various
steps such as sample preparation, image acquisition and the
handling and analysis of large sets of image data. Furthermore, integration of the results with existing knowledge on the
proteins under investigation such as it is provided by bioinformatic databases is also important. Some automated screening microscopes have already been described and applied or
are commercially available [7,8]. Limitations of commercially
available systems are often that they have been designed and
optimised for special applications, which restricts the possibilities for adaptation to new assays. Systems with ultra-high
throughput capacities are lacking the single cell or subcellular
resolution and thus provide only specialised information.
Here we describe a modular microscope-based screening
platform and its application to the development of two cellbased assays addressing protein secretion and Golgi integrity.
This demonstrates its potential for large-scale functional
screening projects.
2. Materials and methods
2.1. Materials
GFP-tagged ORFs were generated and prepared as described earlier
[3].
The yellow £uorescent protein (YFP)-tagged SAR1 mutant, SAR1GTP-YFP plasmid, was constructed by transferring the SAR1
(H79G) mutant, from the pSG5 vector into the C1-YFP Clontech
vector via EcoR1 and BamH1 restriction sites and standard cloning
methods. Recombinant adenoviruses encoding the secretory marker
protein ts-O45-G tagged with either cyan £uorescent protein (CFP) or
YFP were as described [9]. All cell culture reagents were from Gibco/
Invitrogen (Karlsruhe, Germany). LabTek eight- and 96-well plates
were from Nalge Nunc (Rochester, NY,USA). Cycloheximide and
nocodazole were from Calbiochem (San Diego, CA, USA), FuGENE6 was from Roche (Mannheim, Germany), Cy5-labelled antimouse secondary antibodies were from Amersham Biosciences (Freiburg, Germany), mouse monoclonal anti-GM130 antibody was from
BD Bioscience (San Diego, CA, USA), the mouse monoclonal anti-tsO45-G antibody VG was a gift from Kai Simons (MPI-CBG, Dresden, Germany), Hoechst 33342 stain was from Sigma (Munich, Germany).
2.2. Hardware and set-up of the screening system
Liquid handling procedures such as DNA transfection and immunostaining were performed on a Multiprobe IIex robot (Perkin Elmer,
Wellesley, MA, USA). A special specimen holder carrying four LabTek eight-well chamber slides was developed for automated bi-directional transport between the liquid handling system and the automated microscope.
The automated microscope is built with standard components of an
0014-5793 / 03 / $22.00 J 2003 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved.
doi:10.1016/S0014-5793(03)01197-9
U. Liebel et al./FEBS Letters 554 (2003) 394^398
Olympus BX system (Olympus-Europe, Hamburg, Germany) and is
mounted in a custom-built, anti-vibration aluminium frame. Images
are acquired with a cooled CCD camera (SensiCam 1280U1024 pixels, PCO, Kehlheim, Germany). A motorised xyz-stage (ASI Instruments, Eugene, OR, USA) allows positioning with a resolution of 0.1
Wm. All movable parts of the system are remote controlled by LabView1 (National Instruments, Munich, Germany) software modules
that were developed in our laboratory. The robot and microscope are
integrated into a glass cabinet to provide sterile conditions. Pictures of
the entire system and more detailed information can be viewed at
http://www.embl.de/Vliebel.
2.3. Automated transfection and immunostaining
For automated transfection and immunostaining the multiprobe
robot system was used for each step described below, except for the
plating of cells. Cells were plated at a density of 1200 cells/cm2 in 96well plates containing 100 Wl culture medium/well. Four Wl plasmid
DNA at a concentration of 50 ng/Wl was added to the transfection
master mix consisting of FuGENE6 and OPTIMEM 1, which was
prepared so that for each well 10 Wl liquid contained 9.7 Wl OPTIMEM 1 and 0.3 Wl FuGENE6. For cell ¢xation the culture medium
was removed and cells were incubated with 75 Wl of 3% paraformaldehyde for 20 min at room temperature. For immunostaining 50 Wl of
the respective antibody solutions were added per well and incubated
for 30 min at room temperature. Nuclei were labelled for 5 min with
75 Wl Hoechst solution at a concentration of 0.1 Wg/ml. All washing
steps after cell ¢xation or antibody staining were performed three
times for 5 min, each with 200 Wl of phosphate-bu¡ered saline
(PBS)/well.
2.4. Protein transport assay
HeLa cells were transfected with plasmid DNAs encoding CFP- or
YFP-tagged proteins and 6 h, thereafter they were overlaid with recombinant adenoviruses encoding the secretory marker protein tsO45-G tagged with CFP or YFP to complement the colour of the
transfected CFP- or YFP-tagged protein. After 45 min incubation the
cells were washed with culture medium and ts-O45-G was accumulated in the endoplasmic reticulum (ER) by incubating the cells at
39.5‡C for 16 h. Thereafter, cells were shifted to 32‡C, 60 min in
the presence of 100 Wg/ml cycloheximide to release ts-O45-G from
the ER. Then, cells were ¢xed with 3% paraformaldehyde and tsO45-G on the cell surface was detected by immunostaining with a
monoclonal antibody (VG) recognising an extracellular ts-O45-G epitope at the plasma membrane followed by staining with Cy5-labelled
secondary anti-mouse antibodies. Cell nuclei were labelled with
Hoechst 33342 stain diluted in PBS to a ¢nal concentration of 0.1
Wg/ml.
2.5. Golgi integrity assay
HeLa cells were ¢xed in 3% paraformaldehyde and permeabilised
with 0.1% Triton X-100 and the Golgi complex was stained with
mouse monoclonal anti-GM130 antibody followed by Cy5-conjugated
anti-mouse secondary antibodies. Cell nuclei were labelled with
Hoechst stain as described above.
3. Results and discussion
The screening platform developed consists of modules covering automated cell transfection and immunostaining, image
acquisition, data management and analysis, and collection of
bioinformatic information from databases. They were developed with the graphical prototyping software provided by
LabView and designed to operate independently of each other. Therefore, implementation of further modules or products
from di¡erent companies for the development of new assays is
easily possible. In this way the optimum hard and software
components for assay speci¢c tasks can always be used. This
is critical for scienti¢c screening platforms such as the one
described here, where di¡erent ¢elds of expertise have to be
combined.
With the protocols developed for automated cell transfec-
395
tion e⁄ciencies of up to 30% could be obtained, which was
however two to three times less compared to those obtained
by manual transfection. In contrast, immunostaining performed automatically on the robot system (see Figs. 1 and
3) showed a quality that was comparable to experiments conducted manually (not shown). Optimum expression times for
di¡erent cDNAs varied between 12 and 48 h and therefore
transfection e⁄ciencies of di¡erent cDNAs on the same 96well plate varied by a factor of up to 10. This problem could
be overcome by transfecting in parallel those cDNAs which
showed similar expression kinetics.
The image acquisition module controls automated image
acquisition of multistained samples. Screening of an entire
96-well plate when taking 15 ¢elds of view per well lasts between 45 min and 5 h depending on the number of colours in
the sample and exposure times used. Critical for obtaining
high-quality image data in such systems is the automated
identi¢cation of the focal plane for each ¢eld of view. Solutions that analyse the ¢eld of view either as one entity, or by
focusing on the plane containing the brightest object, very
often determine a focal plane with either only few of the cells
in the ¢eld or non-relevant structures with a high intensity
being in focus. Therefore, we have developed a robust software autofocus routine that analyses each individual cell in
the ¢eld separately. The plane which contains the maximum
number of objects in the ¢eld that ful¢l user de¢ned ‘in focus’
criteria is used as the focal plane. In this way, bright dead cells
or junk particles do not disturb focus identi¢cation. The procedure is independent of the objective and cell type used and
achieves similar results for glass and plastic tissue culture
plates. In test runs 92% of the images de¢ned by the autofocus
as ‘in focus’ matched also the judgment by manual inspections. More detailed information about this module can be
found at www.embl.de/Vliebel.
Currently available screening platforms use databases to
handle the acquired data. These have project speci¢c structures, which has the disadvantage that new projects or
changes in hardware may require time-consuming restructuring. Furthermore, when the platform includes several screening microscopes used in parallel, each of them can only operate when the database is fully functioning, which therefore
becomes a bottleneck for the entire platform. We have overcome these problems by using a distributed server concept for
storage, retrieval and analysis of acquired image data from
multiple independent systems without the need for a database.
Meta-information such as the experiment name or well position is saved within the ¢le name of every individual image
acquired. For data retrieval, we developed an intranet search
engine operating similar to Google1, where index computers
screen the system devices for data. In this way the system is
able to handle millions of ¢les on di¡erent computers or data
storage devices.
Once the screening has identi¢ed interesting candidate proteins, it is of vital importance to integrate the results with
existing bioinformatic knowledge on these proteins. For this
purpose we have developed a module that automatically collects for the entire human genome bioinformatic knowledge
and predictions from public webservers on a regular basis
(harvester.embl.de). In this way bioinformatic data on the
proteins under investigation are up to date and instantly available.
Two screening assays, which address protein secretion or
396
U. Liebel et al./FEBS Letters 554 (2003) 394^398
Fig. 1. Example of the automated transport assay and image analysis steps for data analysis. HeLa cells were prepared for the transport assay
as described in Section 2. Images of the Hoechst-stained cell nuclei (A), YFP-transfected cells (B), CFP-tagged ts-O45-G (C) and ts-O45-G at
the plasma membrane (D) were acquired automatically with the high-content screening microscope. The images containing the Hoechst-stained
cell nuclei were thresholded and subsequently binarised (pixels above the threshold value were set to 1, those below the threshold were set to 0)
to de¢ne a digital mask of objects corresponding to the area occupied by each nucleus (dark grey areas in E). Manual inspections showed that
on average 96 T 4% of the cell nuclei were detected in this way. Cell nuclei touching each other or the border of the image were skipped from
further analyses. The objects in this nuclear mask were then dilated (brighter ring-like grey areas in E) and the mask corresponding to the cell
nuclei only was then subtracted to de¢ne a new digital mask with ring-like structures corresponding to juxtanuclear cytoplasmic parts of the
cell. This new mask was then multiplied with image C and resulted in image F. A mask encompassing the nuclear and juxtanuclear ring-like
structures together (dark and bright areas together in E) was then multiplied with the images D (ts-O45-G at the plasma membrane) and B
(transfected YFP) resulting in images G and H, respectively. Finally, the £uorescence in images F to H was quanti¢ed for each object. Transport of ts-O45-G to the plasma membrane was then determined for each cell as the ratio (named ‘Pm/tot.’) of the £uorescences measured for
corresponding objects in images G (ts-O45-G at the plasma membrane) and F (related to the total cellular amount of YFP-tagged ts-O45-G).
Golgi morphology in cells transfected with CFP- or YFPtagged proteins, were developed. An example of the secretion
assay is shown in Fig. 1. As transport marker, the temperature sensitive transmembrane protein ts-O45-G fused to CFP
or YFP is used. It accumulates in the ER at 39.5‡C and is
transported to the plasma membrane at 32‡C where it can be
detected by a monoclonal antibody recognising an extracellular epitope of ts-O45-G (see [10]). The ratio of ts-O45-G spe-
U. Liebel et al./FEBS Letters 554 (2003) 394^398
Fig. 2. Detection of transport inhibition with the secretion assay.
Transport of ts-O45-G was analysed with the high-content screening
microscope in samples transfected with plasmid encoding SAR1GTP-YFP (A) or YFP (B) as a control. Transport of ts-O45-G
(Pm/tot. see legend to Fig. 1 for its de¢nition) and the amount of
GFP proteins expressed in each individual cell was quanti¢ed as described in the legend to Fig. 1 and in the text. Shown are scatter
plots relating the relative amounts of ts-O45-G transported to the
plasma membrane (Pm/tot.) versus the amounts of GFP proteins expressed. The latter ones were normalised and the minimum and
maximum values obtained were set to 0 and 100%, respectively. The
cells expressing less than 5% or more than 20% of the maximum
were scored as non-transfected or overexpressing cells, respectively.
The average transport of ts-O45-G was 0.65 T 0.02 in non-transfected cells and 0.61 T 0.04 or 0.36 T 0.02 in YFP or YFP-SAR1GTP overexpressing cells, respectively.
ci¢c plasma membrane £uorescence (antibody staining) and
total CFP- or YFP-ts-O45-G £uorescence determines the relative amount of ts-O45-G transported (abbreviated in the
following ‘Pm/tot.’). In order to test this assay, a YFP-tagged
SAR1 mutant (H79G, named SAR1-GTP-YFP in the following), which only slowly hydrolyses GTP and is known to
inhibit transport away from the ER [11], was transfected
and ts-O45-G transport measured as a function of the amount
of expressed SAR1-GTP-YFP. As shown in Fig. 2 ts-O45-G
transport to the plasma membrane decreased with increasing
amounts of transfected SAR1-GTP-YFP. In cells transfected
with YFP alone no such inhibition was detected. In two independent experiments with more than 700 cells analysed Pm/
tot. was determined to 0.65 T 0.02, 0.61 T 0.04 and 0.36 T 0.02
for non-transfected, YFP and SAR1-GTP-YFP overexpressing cells, respectively (see legend to Fig. 2 for de¢nitions).
This demonstrates that the automated assay developed is capable of identifying transport inhibitors such as the SAR1GTP-YFP when expressed in cells.
397
An important feature of a microscope-based screening system is that it should o¡er quantitative information on the
spatial distribution of the signals of interest in single cells.
In order to test the performance of our screening system in
this context we developed and tested a further assay, which
probes the morphology of the Golgi complex (Fig. 3). This
organelle is central to the secretory pathway and its proper
functioning is of vital importance for transport and processing
of secretory cargo. Golgi morphology changes when the £ow
of membrane transport towards or away from it is altered and
is thus often an indication of imbalanced membrane £ow
through the secretory pathway. For example, treatment of
cells with the microtubule depolymerising drug nocodazole
inhibits transport of membrane-bounded cargo carriers [12]
and the Golgi becomes scattered throughout the entire cell
cytoplasm into numerous Golgi ministacks (see Fig. 3 and
[13]). The Golgi integrity assay described here was developed
to detect such phenotype and thus scores the number, intensity and size of Golgi fragments. In order to test it, cells were
treated with 10 WM nocodazole for 8 h and Golgi morphology
was analysed in comparison to non-treated control cells (see
Fig. 3). Consistent with visual inspections the number of Golgi fragments detected by the screening microscope system increased from 3.8 T 0.7 to 50 T 7.0 in control and nocodazoletreated cells, respectively. In parallel, the average size and
intensities of Golgi ministacks in nocodazole-treated cells
were determined to decrease to 12 T 4% and 74 T 3% of the
values obtained for control cells. Such decrease in intensity
and size of Golgi fragments is consistent with the hypothesis
that during nocodazole treatment the total amount of the
Golgi marker used does not change drastically in contrast
to the Golgi morphology. Altogether, this demonstrates that
the Golgi integrity assay described here is able to detect phenotypes that resemble Golgi fragmentation in response to nocodazole.
We presently apply the system and assays described here to
systematic analyses in cells overexpressing GFP-tagged proteins of unknown function as they are identi¢ed by large-scale
gene sequencing projects such as that of the German cDNA
Consortium [14]. This has demonstrated so far that hundreds
of proteins can be analysed in parallel with our system and
more than 30 proteins which inhibit protein secretion or a¡ect
the morphology of the Golgi complex upon overexpression
have already been identi¢ed (Starkuviene et al., unpublished
results).
In conclusion, the modular microscope-based screening
platform described here has applications in large-scale functional analysis of proteins of unknown function. It automatically acquires high-quality multicolour image data, which
enables quantitative screens based on cell or organelle morphology as exempli¢ed here by the Golgi integrity assay.
An important feature of the system is that it also includes automated collection of bioinformatic data, which
is essential for systematic analyses of the functional interrelationship of the proteins of interest in larger networks.
The system can be easily adapted to implement new developments due to its modular nature and means of data organisation. Therefore, only few changes on the system will be
necessary for applications di¡erent from the ones shown here
such as in drug screening or for the systematic functional analysis of proteins by gene-knockdowns using siRNAs.
398
U. Liebel et al./FEBS Letters 554 (2003) 394^398
Fig. 3. Example of the Golgi integrity assay. HeLa cells, non-treated (A,C,E) or treated (B,D,F) with 10 WM nocodazole for 8 h, were stained
with Hoechst dye (A,B) and a monoclonal antibody recognising the Golgi matrix protein GM130 [15] (C,D). Nuclear and cytoplasmic juxtanuclear regions were determined as described in the legend to Fig. 1. The images containing Golgi labelling were further thresholded and binarised (E,F) and the number, intensity and size of thresholded Golgi fragments present in the nuclear and juxtanuclear area (rings around the
cells in E and F) were determined.
Acknowledgements: We thank Kai Simons for antibodies and Brigitte
Joggerst for technical assistance. This work has been supported by the
BMBF with grant 01GR0101 of the National Genome Research Network (NGFN); and with grants 01KW9987 (German cDNA Consortium), 01KW0012 (DKFZ) and 01KW0013 (EMBL) within the German Genome Project (DHGP).
References
[1] Sawin, K.E. and Nurse, P. (1996) Proc. Natl. Acad. Sci. USA 93,
15146^15151.
[2] Gonzalez, C. and Bejarano, L.A. (2000) Trends Cell Biol. 10,
162^165.
[3] Simpson, J.C., Wellenreuther, R., Poustka, A., Pepperkok, R.
and Wiemann, S. (2000) EMBO Rep. 1, 287^292.
[4] Ding, D.Q., Tomita, Y., Yamamoto, A., Chikashige, Y., Haraguchi, T. and Hiraoka, Y. (2000) Genes Cells 5, 169^190.
[5] Bastiaens, P.I. and Pepperkok, R. (2000) Trends Biochem. Sci.
25, 631^637.
[6] Wouters, F.S., Verveer, P.J. and Bastiaens, P.I. (2001) Trends
Cell Biol. 11, 203^211.
[7] Ghosh, R.N., Chen, Y.T., DeBiasio, R., DeBiasio, R.L., Conway, B.R., Minor, L.K. and Demarest, K.T. (2000) Biotechniques 29, 170^175.
[8] Jessen, T. (2000) Drug Discov. Today 5, 49.
[9] Keller, P., Toomre, D., Diaz, E., White, J. and Simons, K. (2001)
Nat. Cell Biol. 3, 140^149.
[10] Pepperkok, R., Lowe, M., Burke, B. and Kreis, T.E. (1998) J. Cell
Sci. 111, 1877^1888.
[11] Aridor, M., Fish, K.N., Bannykh, S., Weissman, J., Roberts,
T.H., Lippincott-Schwartz, J. and Balch, W.E. (2001) J. Cell
Biol. 152, 213^229.
[12] Presley, J.F., Cole, N.B., Schroer, T.A., Hirschberg, K., Zaal,
K.J. and Lippincott-Schwartz, J. (1997) Nature 389, 81^85.
[13] Cole, N.B., Sciaky, N., Marotta, A., Song, J. and LippincottSchwartz, J. (1996) Mol. Biol. Cell 7, 631^650.
[14] Wiemann, S., Weil, B., Wellenreuther, R., Gassenhuber, J. and
Glassl, S. et al. (2001) Genome Res. 11, 422^435.
[15] Nakamura, N., Rabouille, C., Watson, R., Nilsson, T., Hui, N.,
Slusarewicz, P., Kreis, T.E. and Warren, G. (1995) J. Cell Biol.
131, 1715^1726.