Download Defining the inner membrane proteome of E coli

Defining the inner membrane
proteome of E coli and thereby
improving 51,208 topology
models of bacterial inner
membrane proteins
Erik Granseth, Mikaela Rapp, Daniel
O. Daley, Karin Melén, David Drew and
Gunnar von Heijne
The inner membrane
proteome of E coli + more
Erik Granseth1, Mikaela Rapp2, Daniel
O. Daley2, Susanna Seppälä2, Karin
Melén1, David Drew2 and Gunnar von
Heijne1,2
1Stockholm
Bioinformatics Center, AlbaNova, [email protected]
2Department
of Biophysics and Biochemistry, Stockholm University
Outline
Global topology analysis of the Escherichia coli inner
membrane proteome
Science (2005) 308, 1321-1323
(Experimentally constrained topology models for 51,208
bacterial inner membrane proteins
J. Mol. Biol. (2005), 352, 489-494)
Dual topology membrane proteins
Global topology analysis of the
Escherichia coli inner membrane
proteome
Daniel O. Daley, Mikaela Rapp,
Erik Granseth, Karin Melén, David
Drew and Gunnar von Heijne
Introduction
Cytoplasm
C
Membrane
30Å
Periplasm
Membrane proteins are found in
membranes such as the plasma
membrane, mitochondrial
membranes, chloroplast
membranes, the ER etc.
Cytoplasm
N
A topology model of
the membrane protein
above
N
Periplasm
C
Experimental technique
When expressed:
only GFP or PhoA will be active and thereby reveal the location of
the C-terminus (cytoplasm or periplasm) for a particular membrane
protein
Periplasm – outside cell
S
N
GFP is only
flourescent in
the cytoplasm
S
PhoA is only
active in the
periplasm
N
Cytoplasm – inside cell
TMHMM was used to create the topology models
• TransMembrane Hidden Markov Model
• Reliability Score (Melén, J. Mol. Biol. 2003)
– A measure of the reliability for a single prediction
– Useful for discriminating between good and bad
topology predictions
• Possibility to constrain predictions
Example of constraining TMHMM to a specific location
Original prediction
Constrained
prediction
GFP/PhoA activities are used to
determine the C-terminal location
C in
t
u
Co
C
601 proteins
for which we
can assign the
C-terminal
Distribution of membrane proteins
Most are involved in transport
in or out of the cell, but many
have unknown function
Most membrane proteins have
the C-terminal in the cytoplasm
They also prefer to have even number
of TMH, i.e. N and C in the cytoplasm
GFP flourescence provide a good estimate of the amount
of fusion protein inserted into the membrane
Overexpression
potential
Toxicity
TMHMM performance on 82
known topologies
100
Original TMHMM
prediction
Fixed Cterm
prediction
75
50
We can see an increase in topology
prediction performance if we fix the
Cterminal
25
0
Correct Cterm
Correct Topology
When comparing our results of the C-terminal location with these
known structures, just one protein is incorrect
Conclusions
• We have derived high-quality topology
models for 601 (almost all) E coli
membrane proteins with an error rate of
~1%
• We have estimated the overexpression
potential and suggest that a large fraction
can be produced in sufficient quantities for
biochemical and structural work
• The final constrained topology models are
now deposited in the uniprot database
Experimentally constrained
topology models for 51,208
bacterial inner membrane proteins
• We used the membrane proteins with
experimentally determined C-terminal
location and searched for homologs in 225
fully sequenced prokaryotic genomes.
• We created 51,208 much improved
topology models
• These cover ~30% of all predicted inner
membrane proteins in the 225 genomes
Dual topology membrane
proteins
Mikaela Rapp, Erik Granseth
Susanna Seppälä and Gunnar
von Heijne
The positive inside rule
Positive inside rule:
Cytoplasm
K
N
R
R
KKR
K
Periplasm
(K+R) bias = 7-4 = 3
R
R
R
C
The majority of (K+R) are
situated in the cytoplasm
K
The (K+R) bias is the
difference between (K+R) in
the cytoplasm and the
periplasm
Dual topology
Most membrane
proteins adopt one
unique topology...
...but there is some
evidence that dual topology
exists
Nin
Cytoplasm
Cin
+
Periplasm
Nout
Cout
Dual topology means that a membrane protein can insert either way into
the membrane. This results in two different topologies for the same
sequence.
The SMR family
Small Multidrug Resistance family
YdgE and YdgF have a clear topology
SugE and EmrE may have dual
topology
Cytoplasm
C
N
+
N
C
YdgE
YdgF
Periplasm
•
•
•
•
Small
3-4 predicted TMhelices
Few K+R residues
Very small K+R bias between
inside and outside loops
YdgF
YdgC
YnfA
CrcB
YdgE
SugE
EmrE
(K+R) alterations
SMR family
= R or K
= Charge mutation
= TM helix
The SMR family
EmrE and SugE are
highly sensitive and
(K+R) alterations lead
to big GFP or PhoA
differences
For the control, YdgE/F,
the (K+R) alterations
have little effect
CrcB, YnfA and YdgC
CrcB is also highly
sensitive to (K+R)
alterations
YnfA and YdgC are less
sensitive to alterations
because they have more
K+R than EmrE, SugE
and CrcB
Dual topology proteins have ~0 (K+R) bias
while 2 closely spaced pairs have large +/singleton
2 adjacent
copies
SMR and CrcB proteins form anti-parallel dimers
composed either of two separately expressed and
oppositely oriented homologues or of a single dual
topology protein
YnfA and YdgC do not have 2
adjacent homologs
Conclusions
• A small number of strategically located
mutations are required to redesign the
topology of a protein
• Some proteins adopt a dual topology,
others are duplicated and followed by
divergent topology evolution. This results
in pairs of oppositely oriented molecules
Final conclusions
• The work presented here is an important
framework for many future studies of
membrane proteins
• Incorporation of experimental topology
information improves the topology models
• These papers have been a nice
cooperation between experimentalists and
bioinformaticians, where both have
benefited from each others results
Acknowledgements
• Gunnar von Heijne
• Arne Elofsson
• SBC and DBB co-workers