Content uploaded by Gianni Guizzunti
Author content
All content in this area was uploaded by Gianni Guizzunti
Content may be subject to copyright.
© 2006 Nature Publishing Group
Functional genomics reveals genes involved in
protein secretion and Golgi organization
Frederic Bard
1
, Laetitia Casano
1
, Arrate Mallabiabarrena
1
, Erin Wallace
1
, Kota Saito
1
, Hitoshi Kitayama
1
,
Gianni Guizzunti
1
, Yue Hu
1
, Franz Wendler
2
, Ramanuj DasGupta
3
, Norbert Perrimon
3
& Vivek Malhotra
1
Yeast genetics and in vitro biochemical analysis have identified
numerous genes involved in protein secretion
1,2
. As compared
with yeast, however, the metazoan secretory pathway is more
complex and many mechanisms that regulate organization of the
Golgi apparatus remain poorly characterized. We performed a
genome-wide RNA-mediated interference screen in a Drosophila
cell line to identify genes required for constitutive protein
secretion. We then classified the genes on the basis of the effect
of their depletion on organization of the Golgi membranes. Here
we show that depletion of class A genes redistributes Golgi
membranes into the endoplasmic reticulum, depletion of class B
genes leads to Golgi fragmentation, depletion of class C genes
leads to aggregation of Golgi membranes, and depletion of class D
genes causes no obvious change. Of the 20 new gene products
characterized so far, several localize to the Golgi membranes and
the endoplasmic reticulum.
Drosophila S2 tissue culture cells were transformed to stably
express horseradish peroxidase fused to a signal sequence (ss-HRP)
on an inducible promoter (Fig. 1a). Addition of Cu
2þ
ions to the cells
induced the production of ss-HRP, which is translocated into the
endoplasmic reticulum (ER), transported to the Golgi apparatus and
then secreted into the medium. An aliquot of the medium was
removed to measure peroxidase activity by chemiluminescence,
thereby providing a robust assay to monitor secretion in a high-
throughput format (Fig. 1b). Two controls were used to verify that
secretion of HRP occurred through the generic secretory pathway:
knockdown of Syntaxin 5 (a t-SNARE) and
b
-COP (a component of
the COP1 coat)
3–5
each caused a 100-fold reduction in HRP secretion
(Fig. 1c).
We used a genome-wide library of ,22,000 double-stranded
RNAs (dsRNAs) that have been previously used in several screens
(refs 6–9 and http://www.flyrnai.org). Cells secreting ss-HRP were
plated in 384-well plates containing dsRNA. After 5d, ss-HRP
production was induced, and 12 h later peroxidase activity released
into the supernatant was measured by chemiluminescence (Fig. 2a).
Each plate included wells with dsRNA encoding Syntaxin 5 and
GFP as positive and negative controls. The screen was carried out
in duplicate. Because most dsRNAs did not inhibit HRP secretion,
the average for a given plate was very close to that of non-treated
wells. Therefore, the z-score of each well, equal to the value of the
well (peroxidase activity) minus the average of the plate divided by
the standard deviation for the plate, was used to compare the
effects of each dsRNA on secretion across the whole set of plates.
The average of the two z-scores for each dsRNA is shown in
Fig. 2b. The scatter plot of the duplicated assay shows that most
dsRNAs did duplicate with a correlation coefficient of 0.63
(Fig. 2c). A few of the dsRNAs did not duplicate, but were
included in the next round of selection to recover potential
positives. On the basis of this analysis, 1,133 dsRNAs were selected
(Supplementary Table S1).
The genes corresponding to the 1,133 dsRNAs were analysed with
Flybase (www.flybase.org). Genes that could affect secretion
indirectly through their roles in apoptosis, transcription, protein
translation, protein degradation and basic metabolism were dis-
carded from further analysis (Supplementary Table S1). In addition,
dsRNAs that scored positively in previous cell survival screens were
removed
6
. Known components of the secretory pathway did not
score in those cell survival screens
6
. This selection reduced the
number to 284 dsRNAs that we tested further in two additional
HRP secretion assays in a 96-well plate format. The DNA-binding dye
Hoechst was used to exclude dsRNAs that could affect cell number
(Fig. 2d). The list of 284 dsRNAs tested and our reasons for excluding
154 from further analysis are given in Supplementary Table S2.
We generated a S2 cell line stably expressing mouse Mannosidase II,
a marker of the cis and medial Golgi cisternae, coupled to GFP
(MannII–GFP) to test the effect of 130 selected genes on Golgi
LETTERS
Figure 1 |HRP secretion in Drosophila S2 cells. a, S2 cells transfected with a
plasmid containing the signal sequence (ss) of Drosophila Bip appended to
HRP and a V5 tag under the influence of an inducible metallothionine
promoter (pMT). b, The peroxidase activity in the supernatant from S2 cell
culture is HRP. The supernatant from wild-type cells, cells expressing HRP
but not induced, and cells induced to produce HRP was analysed for
peroxidase activity by chemiluminescence. RLU, relative light units. c,RNAi
of known effectors of trafficking effectively blocks HRP secretion. RNAi of
the bona fide transport components
b
-COP and Syntaxin 5 was used to
monitor effects on the secretion of HRP. Cells induced to produce HRP
secrete considerable peroxidase activity, which is inhibited on depletion of
b
-COP and Syntaxin 5. Error bars in band crepresent the s.d. of triplicate
measurements from representative experiments.
1
Cell and Developmental Biology Department, University of California San Diego, La Jolla, California 92093-0634, USA.
2
National Institute for Medical Research, the Ridgeway
Mill Hill, London NW7 1AA, UK.
3
Department of Genetics, Howard Hughes Medical Institute, Harvard Medical School, 77 Avenue Louis Pasteur, Boston, Massachusetts 02115,
USA.
Vol 439|2 February 2006|doi:10.1038/nature04377
604
© 2006 Nature Publishing Group
organization. As previously reported
10
, Golgi membranes in S2 cells
are organized as several unconnected stacks of cisternae (Fig. 3a). We
incubated S2 cells expressing MannII–GFP with dsRNAs by the
procedure described above for HRP-secreting S2 cells, and then
imaged them by high-resolution deconvolution fluorescence
microscopy. The genes were classified into four groups on the basis
of the effect of their depletion on Golgi membranes in greater than
50% of the total cells. RNA-mediated interference (RNAi) of class A
genes fused Golgi membranes with the ER, as shown by the reloca-
tion of MannII–GFP in a ring around the nucleus and a diffuse
reticular network (Fig. 3b). RNAi of class B genes fragmented the
Golgi membranes into smaller elements, RNAi of class C genes
caused aggregation and swelling, and RNAi of class D genes had no
apparent effect on Golgi organization (Fig.3b). The complete list of
the four classes of genes, their potential human orthologues and their
domains are given in Supplementary Table S3. We found that 77 out
of the 130 genes had potential human orthologues and 26 had
homologues identified previously as trafficking components. The
dsRNAs that have potential off-target effects on the basis of the
presence of a 21-base-pair overlap with other genes, and which
therefore need further validation, are also listed in Supplementary
Table S3.
In mammalian cells, the Golgi membranes are fragmented and
protein transport is blocked during mitosis
11
. The list of 130 genes
includes separase,klp61F and microtubule star, which have been
linked to mitosis-specific events
12–14
. It is therefore possible that
effects on trafficking and Golgi organization in S2 cells depleted of
these genes are due to arrest in mitosis. RNAi-treated MannII–GFP
cells were visualized with antibody against tubulin and mitosis-
specific antibody against phosphorylated histone H3. We could
detect mitotic cells in control wells (Supplementary Fig. S1), but
there was no increase in the mitotic index of cells depleted of the
genes mentioned above. Consistently, the phenotypes of class B or C
genes were not associated with DNA condensation, microtubule
reorganization or increase in staining for phosphorylated histone H3
(Supplementary Fig. S1). Therefore, the fragmented Golgi phenotype
and inhibition in secretion of HRP in these cells are not due to an
arrest in mitosis. However, other caveats remain; for example, we
found that the knockdown of genes involved in fatty acid and
cholesterol biosynthesis results in a block of HRP secretion and a
class A Golgi membrane phenotype (Supplementary Table S2). To
confirm their direct role in membrane trafficking, therefore, the
104 candidate genes need further functional tests and, notably,
characterization of the intracellular localization of their products.
To this end, 20 randomly selected genes including CG14181,a
likely orthologue of Use1 (encoding a t-SNARE localized to the ER in
Saccharomyces cerevisiae
15
), were cloned in an inducible expression
vector with a V5 tag. S2 cells expressing MannII–GFP were trans-
fected with the tagged cloned genes, and the gene products were
Figure 2 |Identification of the genes involved in HRP secretion. a, Each
well of each plate containing a dsRNA corresponding to a specific Drosophila
gene was seeded with 1 £10
4
S2 cells stably expressing ss-HRP. After 5 d of
incubation, cells were induced to synthesize HRP. After 12 h, 10
m
l of culture
supernatant was transferred to another well, the HRP substrate was added,
and luminescence was measured. b, Whole-genome assay for HRP secretion.
The z-score was derived from the luminescence of each well. All dsRNAs that
inhibited secretion with a z-score of less than 21.5 (lower red line) were
selected as positive hits for further analysis. c, Scatter plot for the duplicate
screen. The two z-scores derived for each dsRNA are plotted on the xand y
axis to show the overall reproducibility. d, Flow chart showing the selection
of genes involved in secretion. From an initial collection of 1,133 genes
(Supplementary Table S1), 130 (Supplementary Table S3) were found to be
highly specific for HRP secretion. The subcellular localization of 20 of these
genes was assessed after cloning and transient transfection.
Figure 3 |Drosophila genes involved in Golgi organization. S2 cells stably
expressing MannII–GFP, a marker of medial Golgi, were subjected to
RNAi for each of the 130 genes identified to be essential for secretion.
a, Organization of the untreated control MannII–GFP in S2 cells. MannII
appears in discrete units around the nuclear periphery. b, On the basis of
their effects on organization of the Golgi membranes, the 130 secretory
components are grouped into four classes (Supplementary Table S3). Their
depletion caused the Golgi to fuse with the ER (class A), fragmented the
Golgi (class B), induced Golgi membranes to aggregate and swell (class C) or
had no effect on Golgi organization (class D).
NATURE|Vol 439|2 February 2006 LETTERS
605
© 2006 Nature Publishing Group
visualized with an antibody against V5 and imaged by deconvolution.
The localization of the gene products was compared with that of the
Golgi marker (MannII–GFP), the ER pattern, and the diffuse
cytosolic pattern of a soluble protein (examples of typical localization
are shown in Fig. 4). Of the 20 cloned genes products, 4 localized to
the Golgi membranes and 7 (including the CG14181 product)
localized to the ER, suggesting that they have a direct role in
membrane trafficking (Table 1). The gene product of CG11098,
named TANGO1 for transport and Golgi organization (Table 1),
is perfectly juxtaposed to the MannII–GFP-containing Golgi
membranes (Fig. 2a). TANGO1 contains an amino-terminal Src
homology 3 (SH3) domain followed by two coiled-coil domains,
two transmembrane domains and a proline-rich domain. The coiled-
coil domains have homology to the yeast protein Uso1p and the
golgin p115. TANGO1, however, is not the Drosophila orthologue
(encoded by CG1422) of Uso1p or p115 and has a human but not an
S. cerevisiae homologue.
Another example of a metazoan-specific gene is CG1098, whose
product localizes to the cytoplasm and Golgi membranes. CG1098
has a human orthologue, NRBP, which encodes a serine/threonine
kinase that is recruited to Golgi membranes on viral infection
16
and that, when overexpressed, perturbs early Golgi membranes in
mammalian cells
17
. Among the candidates with a yeast homologue,
TANGO2 (CG11176) contains a domain, DUF833, of unknown
function that is widely conserved, even in some bacteria. TANGO4
(CG1796) also has a potential homologue in yeast, the splicing factor
gene Prp46. Contrary to other splicing factors, however, RNAi of
TANGO4 resulted in a marked effect on Golgi membranes (class A
phenotype) and its gene product localized to the cytosol and not the
nucleus, suggesting that TANGO4 regulates secretion independently
from its potential role in RNA splicing. The challenge now is to
understand how these and the other identified components regulate
protein secretion and whether they regulate Golgi membrane organ-
ization directly. Their characterization will hopefully help to resolve
the emerging complexities of the metazoan secretory pathway.
METHODS
Constructs and cell lines. The HRP-C gene was cloned by PCR from a
construct
18
provided by D. Cutler (MRC, University College London) and
inserted into pMT/BiP/V5–His (Invitrogen). The sequence corresponding to
the 100 N-terminal amino acids of mouse MannII was fused with the GFP
sequence and inserted into pAC5/V5–His. S2 cells were cotransfected with
pHygro (Invitrogen) in a ratio of 20 to 1 and selected with 0.3mg ml
21
of
Hygromycin.
Primary screen and analysis. Two sets of 58 plates containing 0.25
m
gof
dsRNA per well were provided by the Drosophila RNAi Screening Center
(DRSC; http://flyrnai.org). 1 £10
4
ss-HRP cells were seeded in each well with
a Multidrop 384. After incubation for 2 h, 20
m
l of fetal bovine serum containing
medium was added. After 5 d, the culture medium was replaced with 50
m
lof
medium containing 500
m
M copper and the cells were incubated overnight. We
transferred 10
m
l of supernatant into a receptacle plate with a Cybio CyBi-Well
vario system and 50
m
l of ECL reagent (Perkin-Elmer Western Lightning) was
added. Luminescence was measured with an Analyst plate reader. Each receptacle
plate was assigned a number and a barcode similar to the initial set for automatic
identification by the plate reader.
Analysis of primary screen. z-Scores were derived from the log value of
luminescence and genes with a score below 21.5 were selected. By using the
DSRC database, genes that scored in cell survival screens
6
, as well as genes
involved in transcription, RNA splicing, protein translation and proteasome
function, were excluded from further analysis. The raw data for the whole
screen will be made available (http://flyrnai.org.). Identification of potential
orthologues in humans was based on information available in Flybase (http://
flybase.bio.indiana.edu/) and on the reciprocal best blast searches with the
InParanoid algorithm.
Secondary screens and morphological effects on Golgi membranes. Using
PCR templates provided by the DRSC, we resynthesized dsRNAs and tested their
effect on HRP secretion with a protocol similar to the primary screen. To
measure cell number, Hoescht was added at 5
m
gml
21
to cells for 90 min, the
cells were washed, and ultraviolet fluorescence was measured with a plate reader
(Tecan). dsRNAs resulting in greater than 50% inhibition of HRP secretion
without affecting cell number were selected as positives. MannII–GFP cells were
incubated with dsRNAs as described above; after 5 d, the cells were transferred to
concanavalin-treated 96-well glass-bottom plates, allowed to spread for 2 h and
then fixed and labelled with Hoescht at 2
m
gml
21
. Image stacks for five fields in
each well were acquired with a 60 £objective and treated for deconvolution.
Cloning of genes for expression in S2 cells. Genes were cloned by RT–PCR
from a library of poly(A) RNA from Drosophila larvae (Clontech) using the
Gateway system (Invitrogen) and subcloned into pDEST48. MannII–GFP S2
cells were transfected 2 d before gene expression was induced with Cu
2þ
for 4 h,
fixed, labelled with antibody against V5 (Invitrogen), and processed for imaging
as described above.
Received 22 August; accepted 18 October 2005.
1. Novick, P. & Schekman, R. Secretion and cell-surface growth are blocked in a
temperature-sensitive mutant of Saccharomyces cerevisiae.Proc. Natl Acad. Sci.
USA 76, 1858–-1862 (1979).
2. Rothman, J. E. Mechanisms of intracellular protein transport. Nature 372,
55–-63 (1994).
3. Nichols, B. J. & Pelham, H. R. SNAREs and membrane fusion in the Golgi
apparatus. Biochim. Biophys. Acta 1404, 9–-31 (1998).
4. Duden, R., Allan, V. & Kreis, T. Involvement of
b
-COP in membrane traffic
through the Golgi complex. Trends Cell Biol. 1, 14–-19 (1991).
5. Serafini, T. et al. A coat subunit of Golgi-derived non-clathrin-coated vesicles
with homology to the clathrin-coated vesicle coat protein
b
-adaptin. Nature
349, 215–-220 (1991).
6. Boutros, M. et al. Genome-wide RNAi analysis of growth and viability in
Drosophila cells. Science 303, 832–-835 (2004).
Table 1 |Intracellular localization of the gene products involved in HRP
secretion
CG number Gene name Golgi phenotype Localization of gene product
CG11098 TANGO1 A Golgi
CG11176 TANGO2 A cyto þGolgi
CG12444 TANGO3 AER
CG14181 Use1 AER
CG1796 TANGO4 A cyto
CG32675 TANGO5 AER
CG18398 TANGO6 B cyto þGolgi
CG8309 TANGO7 B cyto þGolgi
CG14503 TANGO8 CER
CG9191 Klp61F CERþGolgi þmicrotubules
CG10007 TANGO9 D Golgi
CG1098 Madm D cyto þGolgi
CG1841 TANGO10 D cyto
CG30404 TANGO11 DER
CG31052 TANGO12 D Golgi
CG32632 TANGO13 D Golgi
CG33553 Doa DER
CG4775 TANGO14 DER
CG7850 puckered DERþGolgi
CG8588 pastrel D cyto
Twenty genes were cloned and tagged with V5, and their subcellular localization was
monitored in S2 cells expressing MannII–GFP. The localization of the gene product is
abbreviated as follows: cyto, cytosolic; Golgi, Golgi membranes; ER, endoplasmic reticulum.
Computed genes without a previous name are labelled TANGO for transport and Golgi
organization.
Figure 4 |Localization of the products of new genes regulating secretion.
Candidate genes were cloned in an inducible vector with a V5 tag and
transiently transfected into S2 cells expressing MannII–GFP. After fixation,
cells were labelled with Hoechst (to stain DNA) and an anti-V5 antibody
coupled to Texas red. MannII–GFP is green, DNA is blue and the gene
product is red. a,CG11098 (Golgi). b,CG32675 (ER). c,CG8309 (Golgi and
cytosol).
LETTERS NATURE|Vol 439|2 February 2006
606
© 2006 Nature Publishing Group
7. DasGupta, R., Kaykas, A., Moon, R. T. & Perrimon, N. Functional genomic
analysis of the Wnt-wingless signalling pathway. Science 308, 826–-833
(2005).
8. Eggert, U. S. et al. Parallel chemical genetic and genome-wide RNAi screens
identify cytokinesis inhibitors and targets. PLoS Biol. 2, e379 (2004).
9. Kiger, A. A. et al. A functional genomic analysis of cell morphology using RNA
interference. J. Biol. 2, 27 (2003).
10. Stanley, H., Botas, J. & Malhotra, V. The mechanism of Golgi segregation
during mitosis is cell type-specific. Proc. Natl Acad. Sci. USA 94, 14467–-14470
(1997).
11. Shorter, J. & Warren, G. Golgi architecture and inheritance. Annu. Rev. Cell. Dev.
Biol. 18, 379–-420 (2002).
12. Uhlmann, F. Separase regulation during mitosis. Biochem. Soc. Symp. 70,
243–-251 (2003).
13. Sharp, D. J. Cell division: MAST sails through mitosis. Curr. Biol. 12, R585–-R587
(2002).
14. Snaith, H. A., Armstrong, C. G., Guo, Y., Kaiser, K. & Cohen, P. T. Deficiency of
protein phosphatase 2A uncouples the nuclear and centrosome cycles and
prevents attachment of microtubules to the kinetochore in Drosophila
microtubule star (mts) embryos. J. Cell Sci. 109, 3001–-3012 (1996).
15. Dilcher, M. et al. Use1p is a yeast SNARE protein required for retrograde traffic
to the ER. EMBO J. 22, 3664–-3674 (2003).
16. Chua, J. J., Ng, M. M. & Chow, V. T. The non-structural 3 (NS3) protein of
dengue virus type 2 interacts with human nuclear receptor binding protein and
is associated with alterations in membrane structure. Virus Res. 102, 151–-163
(2004).
17. De Langhe, S., Haataja, L., Senadheera, D., Groffen, J. & Heisterkamp, N.
Interaction of the small GTPase Rac3 with NRBP, a protein with a kinase-
homology domain. Int. J. Mol. Med. 9, 451–-459 (2002).
18. Connolly, C. N., Futter, C. E., Gibson, A., Hopkins, C. R. & Cutler, D. F.
Transport into and out of the Golgi complex studied by transfecting cells with
cDNAs encoding horseradish peroxidase. J. Cell Biol. 127, 641–-652 (1994).
Supplementary Information is linked to the online version of the paper at
www.nature.com/nature.
Acknowledgements We thank members of the Malhotra laboratory for
discussions; members of the DRSC for advice; the Institute for Chemistry and
Cell Biology for use of their Cybio robot; and J. Feramisco and members of the
UCSD Cancer Center imaging facility for help with microscopy. Work in the
Malhotra laboratory is supported by NIH grants and a senior investigator award
from Sandler’s Program for Asthma Research. N.P. is a Howard Hughes
investigator.
Author Information Reprints and permissions information is available at
npg.nature.com/reprintsandpermissions. The authors declare no competing
financial interests. Correspondence and requests for materials should be
addressed to V.M. (malhotra@biomail.ucsd.edu).
NATURE|Vol 439|2 February 2006 LETTERS
607
