Structure of a bacterial type IV secretion core complex
at subnanometer resolution
Angel Rivera-Calzada1, Rémi Fronzes2, Christos G. Savva1, Vidya Chandran1, Pei W Lian1,
Toon Laeremans3, 4, Els Pardon3, 4, Jan Steyaert3, 4, Han Remaut3, 4, Gabriel Waksman1,5
and Elena V. Orlova1,5
1Institute
of Structural and Molecular Biology, UCL and Birkbeck, Malet Street, WC1E 7HX
London, UK.
2Institut
Pasteur, G5 “biologie structural de la sécrétion bactérienne”, UMR 3528, 25 Rue
du Dr. Roux, 75015 Paris, France.
3Vrije
Universiteit Brussel, Structural Biology Brussels, Pleinlaan 2, 1050 Brussels,
Belgium.
4VIB, Department of Structural Biology, VIB, Pleinlaan 2, 1050, Brussels, Belgium
5Authors
for correspondence:
[email protected] and [email protected]
Supplementary Information
Supplementary methods
Induction of a humoral immune response in llama and Nanobody (NB) identification.
Llama Claudine (Llama glama) was immunized with 6 subcutaneous injections of adjuvant
(Gerbu LQ, GERBU biotechnik) emulsified pKM101 full-length core complex (FLCC).
Immunogens were administered in weekly intervals and four days after the final boost,
llamas were bled and total RNA was extracted from collected peripheral blood
mononuclear cells according to Domanska et al, 2011. Starting from total RNA, cDNA was
synthesized and the NB repertoire was amplified and cloned following the method
published by Conrath et al, 2001, except that phagemid pMESY4 was used as the display
vector allowing the expression of C-terminally His6-EPEA tagged NBs. The resulting library
consisted of 1.7108 independent clones and 95% of these clones contained an insert
corresponding to the size of a NB. To identify FLCC-specific binders, 1µg of the antigen
was solid phase immobilized in sodium bicarbonate buffer pH 8.2 in 96-well Maxisorp
plates (Nunc). Microwells were subsequently blocked with PBS containing 2% skimmed
1
milk powder. Following incubation with NB displaying phage, unspecific phage was
removed by extensive washing with PBS 0.05% Tween-20 and bound phage was eluted
after trypsin treatment. Two rounds of selections were performed and 92 monoclonal NBs
randomly picked from the first and second round outputs were expressed in the periplasm
of E. coli. Specific NBs were identified via ELISA by coating 0.1µg of the FLCC in
Maxisorp plates. Bound NBs were detected via the EPEA tag using an in house prepared
NbSyn2 alkaline phosphatase conjugate (De Genst et al, 2010). All FLCC-specific NBs
were sequenced resulting in 40 unique NB sequences. The NBs could be separated into
12 distinct sequence groups that are expected to bind distinct epitopes on the FLCC.
Labelling of TraO/VirB9NT using NBs. The FLCC eluted from the StrepTrap HP (GE
Healthcare) was incubated at 4°C for two hours with purified NBCA4304 in a molar ratio
approximately equivalent to 7 molecules of the NB bound to one FLCC. The mix was then
loaded onto Superose 6 10/300 (GE Healthcare) equilibrated in buffer A (50 mM Tris-HCl
pH 8.0, 200 mM NaCl, 10 mM LDAO). Fractions were analyzed by SDS-PAGE, and those
corresponding to the peak containing the FLCC and NBCA4304 were pooled together and
concentrated using a centrifugal filter of 100 kDa cut-off (Millipore). The concentrated
sample was incubated with a 5-fold molar excess of anti-His antibody (GE Healthcare) for
one hour at 4°C, and then loaded onto Superose 6 10/300 (GE Healthcare) equilibrated in
buffer A. A new clear band of 150 kDa appeared associated with the peak of the
FLCC+NB complex (Fig. 5b). This band reacts with a secondary antibody targeted against
mouse IgG and the band disappears when reducing agent is present in the loading buffer
(Fig. 5b). All this confirms that the top band corresponds to the IgG bound to the His-tag of
NBCA4304 in the FLCC+NB complex. The fraction corresponding to the purified complex
made of FLCC, NBCA4304 and anti-His antibody was used to prepare negative stain grids.
Electron microscopy of NB-labelled FLCC. Samples of the FLCC bound to
NBCA4304 and anti-His antibodies were stained with Nano-W (Nanoprobes) to avoid the
drop in pH associated with other staining agents that might fall apart the complex on the
grid. Images were taken on a F20 FEI microscope at a magnification of 55,000X. Image
processing was done using IMAGIC-5 (van Heel et al, 2000) including multivariate
statistical analysis (MSA) and multi-reference alignment (MRA) (Fig. 5d). 444 images of
CC+NBCA4304+anti-His were used for classification. The class-averages shown in Fig. 5d
represent the averages of 5-7 images per class.
Electron microscopy and image analysis of the FLCC and CCelastase complexes. The
FLCC (0.5 mg/ml) sample was applied to lacey carbon coated grids that had been glow
discharged for 30 sec (Fronzes et al, 2009). The grids were plunge frozen and kept at
2
liquid nitrogen temperatures. Samples of the CCelastase complex at a concentration of 0.01
mg/ml were applied on glow discharged continuous carbon-coated grids (Agar Scientific).
Samples were plunge-frozen using a Vitrobot (FEI) and visualised on a F20 FEI
microscope operating at a voltage of 200 kV, a magnification of 68,100X, low dose
conditions and a defocus range of 1.5-3.5 μm. Image frames were recorded on a 4k x 4k
Gatan CCD camera. The pixel size at the specimen level was 2.22 for the FLCC and 2.08
for the CCelastase complex.
The CTF parameters were assessed using CTFFIND3 (Mindell & Grigorieff, 2003),
and the phase flipping was done using Bshow 1.6 (Heymann, 2001; Heymann & Belnap,
2007). Particles were selected from CTF-corrected micrographs using BOXER (EMAN 1.9
(Ludtke et al, 1999)). Images with astigmatism more than 5% were discarded. A total of
4,877 particle images of the FLCC were selected from 420 CCD frames and windowed
into 160x160 pixel boxes. For the CCelastase complex, we have collected 262 frames from
which 10,642 images were selected (160x160 pixels box size). Images of both samples
were normalised to the same mean and standard deviation and high-pass filtered at a lowresolution cut-off of ~100 Å. They were centred, and subjected to the first round of multistatistical analysis (MSA). A very first reference set was obtained using reference free
classification in IMAGIC-5 (van Heel et al, 2000). The best classes corresponding to
characteristic views were used as references for the MRA. Reconstructions for both
CCelastase and FLCC samples were obtained independently. The first 3D model was
calculated from the best 10 characteristic views (with good contrast and well defined
features) of the complex with orientations determined by angular reconstitution (van Heel
et al, 2000). The 3D maps were refined by several rounds of MRA, MSA and anchor set
refinement, gradually reducing the number of images per class from 15 to three. After
excluding distorted images, a total of 3,805 images of the FLCC were taken for the final
reconstruction. The final 3D map of the CCelastase complex was calculated from 5,430
particles. 14-fold symmetry was applied throughout the entire procedure of structural
analysis. Handedness of both EM structures was checked by visual inspection of the OL
sections in the three structures: FLCC, CCelastase and the atomic structure of the OL
(Chandran et al, 2009). Then it was verified by assessment of cross-correlation
coefficients between the atomic model of the OL and the non-mirrored and mirrored
versions of the FLCC and the CCelastase structures (Supplementary Fig. 2). The resolution
was estimated by Fourier Shell Correlation (FSC) accordingly to the 0.5 criteria level as
12.4 Å and 8.5 Å for the FLCC and CCelastase structures respectively (Supplementary Fig.
1c). The final CCelastase and FLCC 3D reconstructions were sharpened according to the B
3
factor of ~250 and 100 respectively and rendered to show 100% of the protein mass
assuming an average density of 1.35 g/cm3.
To compare the two cryo-EM maps in detail we have adjusted the level of low
frequencies in the structures and the resolution of both maps to 12.4 Å. The structures
were aligned using tools available in Chimera (Pettersen et al, 2004). The crosscorrelation coefficient between the aligned cryo-EM maps of the CCelastase and FLCC
complexes is 0.72. A difference map was calculated between the FLCC and the CCelastase
complexes using cryo-EM maps aligned and normalised to the same sigma. Visualization
of maps and figures were done in Chimera (Pettersen et al, 2004).
Modelling of domain structures for the FLCC and CCelastase complexes. The
sequence of TraO/VirB9NT (Q46704 UniProtKB) and TraF/VirB10NT (Q46705 UniProtKB)
were analysed using the structure prediction servers Phyre2, HHpred and I-TASSER
(Kelley & Sternberg, 2009; Roy et al, 2010; Soding, 2005; Zhang, 2008). The five top
models generated by I-TASSER were considered for the further analysis.
Models for the first 160 amino acids of TraF/VirB10 were generated using ITASSER (Roy et al, 2010; Zhang, 2008). This region is predicted to be predominantly
disordered; however, three areas were predicted to be helices with high confidence by
several servers. The areas correspond to helix αA (amino acids K40AF-LVF53), helix αB
(amino acids A107RA-QAA113) and helix αC (amino acids P138EE-QRR146).
Fitting of atomic data into the cryo-EM maps of the FLCC and CCelastase complexes.
The initial fitting was done manually in Chimera (Pettersen et al, 2004). The crystal
structure of the OL complex (PDB 3JQO (Chandran et al, 2009)) was initially fitted in the
OL of the CCelastase cryo-EM map using the rigid body approach. This fit was further
optimized using the flexible fitting software Sculptor (Birmanns et al, 2011) and Flex-EM
(Pandurangan & Topf, 2012; Topf et al, 2008), and local refinement in Chimera (Pettersen
et al, 2004). The N-terminal lever arm of TraF/VirB10 (Chandran et al, 2009) was manually
adjusted to optimize the fitting to the corresponding area in the cryo-EM map. The crosscorrelation coefficient of the final atomic model and the original atomic structure (3JQO)
fitted in the CCelastase cryo-EM map is 0.71 and 0.69, respectively. For the FLCC cryo-EM
map, these cross-correlation coefficients were 0.63 in both cases.
The model of the N-terminus (L24EVG to FIET135) of TraO/VirB9 was fitted firstly
manually in the EM density and then further optimized with local refinement in Chimera
(Pettersen et al, 2004). This initial result was subsequently improved using the Flex-EM
software for flexible fitting (Pandurangan & Topf, 2012; Topf et al, 2008). The final crosscorrelation coefficient for the 14-fold symmetrised model of TraO/VirB9NT in the cryo-EM
4
map of the CCelastase complex is 0.82. The equivalent fitting in the FLCC cryo-EM map has
a cross-correlation coefficient of 0.72. The models generated for helix αA, helix αB and
helix αC located in TraF/VirB10NT were fitted manually in one monomer of the difference
map and then symmetrised in Chimera. Supplementary Table I shows the crosscorrelation coefficients for the different fittings.
Supplementary references
Birmanns S, Rusu M, Wriggers W (2011) Using Sculptor and Situs for simultaneous
assembly
of
atomic
components
into
low-resolution
shapes.
J Struct Biol 173: 428-435
Chandran V, Fronzes R, Duquerroy S, Cronin N, Navaza J, Waksman G (2009) Structure
of the outer membrane complex of a type IV secretion system. Nature 462: 1011-1015
Conrath KE, Lauwereys M, Galleni M, Matagne A, Frere JM, Kinne J, Wyns L,
Muyldermans S (2001) Beta-lactamase inhibitors derived from single-domain antibody
fragments elicited in the camelidae. Antimicrobial agents and chemotherapy 45: 28072812
De Genst EJ, Guilliams T, Wellens J, O'Day EM, Waudby CA, Meehan S, Dumoulin M,
Hsu ST, Cremades N, Verschueren KH, Pardon E, Wyns L, Steyaert J, Christodoulou J,
Dobson CM (2010) Structure and properties of a complex of alpha-synuclein and a singledomain camelid antibody. J Mol Biol 402: 326-343
Domanska K, Vanderhaegen S, Srinivasan V, Pardon E, Dupeux F, Marquez JA, Giorgetti
S, Stoppini M, Wyns L, Bellotti V, Steyaert J (2011) Atomic structure of a nanobodytrapped domain-swapped dimer of an amyloidogenic beta2-microglobulin variant. Proc
Natl Acad Sci U S A 108: 1314-1319
Fronzes R, Schafer E, Wang L, Saibil HR, Orlova EV, Waksman G (2009) Structure of a
type IV secretion system core complex. Science 323: 266-268
Heymann JB (2001) Bsoft: image and molecular processing in electron microscopy. J
Struct Biol 133: 156-169
Heymann JB, Belnap DM (2007) Bsoft: image processing and molecular modeling for
electron microscopy. J Struct Biol 157: 3-18
Kelley LA, Sternberg MJ (2009) Protein structure prediction on the Web: a case study
using the Phyre server. Nat Protoc 4: 363-371
Ludtke SJ, Baldwin PR, Chiu W (1999) EMAN: semiautomated software for high-resolution
single-particle reconstructions. J Struct Biol 128: 82-97
Mindell JA, Grigorieff N (2003) Accurate determination of local defocus and specimen tilt in
electron microscopy. J Struct Biol 142: 334-347
5
Pandurangan AP, Topf M (2012) Finding rigid bodies in protein structures: Application to
flexible fitting into cryoEM maps. J Struct Biol 177: 520-531
Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, Meng EC, Ferrin TE
(2004) UCSF Chimera--a visualization system for exploratory research and analysis. J
Comput Chem 25: 1605-1612
Roy A, Kucukural A, Zhang Y (2010) I-TASSER: a unified platform for automated protein
structure and function prediction. Nat Protoc 5: 725-738
Soding J (2005) Protein homology detection by HMM-HMM comparison. Bioinformatics
21: 951-960
Topf M, Lasker K, Webb B, Wolfson H, Chiu W, Sali A (2008) Protein structure fitting and
refinement guided by cryo-EM density. Structure 16: 295-307
van Heel M, Gowen B, Matadeen R, Orlova EV, Finn R, Pape T, Cohen D, Stark H,
Schmidt R, Schatz M, Patwardhan A (2000) Single-particle electron cryo-microscopy:
towards atomic resolution. Q Rev Biophys 33: 307-369
Zhang Y (2008) I-TASSER server for protein 3D structure prediction. BMC Bioinformatics
9: 40
6
Supplementary
Figures
Supplementary Figure 1. Cryo-EM analysis of the CCelastase complex. (a) Micrograph
of the vitrified sample. Clear single molecules corresponding to side and top views are
boxed with circles and squares respectively. The scale bar corresponds to 100 nm. (b)
Gallery of single particles (SP) used during the refinement. Representative pairs of final
classes (C) and the equivalent projections (P) are also shown. The scale bar on the image
galleries corresponds to 10 nm. (c) Fourier Shell Correlation plot of the CCelastase (blue)
and FLCC (red) cryo-EM maps.
7
Supplementary Figure 2. Determination of the CCelastase cryo-EM map handedness.
Here we show central cross-sections of the OL part of the final and the mirrored CCelastase
structure (left and right panels, respectively) and the OL atomic structure filtered to the
equivalent resolution of the CCelastase cryo-EM map (central panel). The handedness of the
overall features observed in the central sections is indicated with white arrows. The
characteristic features of structural elements are schematically represented above the
cross-sections. The correlation coefficient for the fitting of the atomic structure of the OL in
the CCelastase cryo-EM maps with the final and mirrored handedness are shown below the
panels with the cross sections.
8
a
9
b
Supplementary Figure 3. Secondary structural elements predicted for the N-terminal
domains of TraO/VirB9 (a) and TraF/VirB10 (b). The confidence of prediction is shown
in tables. In (a) β-strands identified are shown in yellow. According to Phyre2 the highest
confidence is 77 % for this prediction (the average score is 73%). In (b) the α-helices
identified by I-TASSER and Phyre2 are shown in green. For those α-helical regions
identified with high confidence (Normalized Z-score = 0.76) the corresponding amino acid
sequence is shown in red.
10
Supplementary Figure 4. A pseudo atomic model predicted for the N-terminal
domain of TraO/VirB9 (from residues 24 to 135). The model obtained using I-TASSER
and chosen for the subsequent flexible fitting analysis is shown in blue. The final model
optimised by the flexible fitting in the experimental density of the CC elastase cryo-EM map is
shown in red. Location of the C-termini is indicated by a red-circle.
Supplementary Figure 5. Fitting of the OL crystal structure into the CCelastase cryoEM map. Slices of the OL at different heights of the CCelastase cryo-EM map with the fitted
atomic structure of the OL (shown in cyan). (a) Slice at the level of the top of TraN/VirB7.
(b) Slice at the level of the lever arm of TraF/VirB10.
11
Supplementary Figure 6. Nanobody screening for the FLCC, CCelastase, and OL
complexes. ELISA results show different affinities of the nanobodies for the FLCC (black),
the OL (clear grey) and the CCelastase (dark grey) complexes. The experiment was done
three times using duplicates in each experiment obtaining equivalent results for all of them.
The results shown correspond to the normalized average. NBCA4304 was chosen for the
labeling experiments.
FLCC
CCelastase
OL
(3JQO)
Flexible
fitting of the
OL
Atomic model of
TraO/VirB9
N-terminus (from
L24EVG to FIET135)
0.63
0.63
0.72
0.69
0.71
RMSD
Crosscorrelation
between
both EM
maps
1.21 Å
0.72
0.82
Supplementary Table I. Cross-correlation coefficients of the models fitted into
EM maps.
12