Type VI secretion requires a dynamic
contractile phage tail-like structure
M. Basler1*, M. Pilhofer2,3*, G. P. Henderson2, G. J. Jensen2,3& J. J. Mekalanos1
related to components of bacteriophage tails. Here we show that protein secretion by the type VI secretion system of
Vibrio cholerae requires the action of a dynamic intracellular tubular structure that is structurally and functionally
homologous to contractile phage tail sheath. Time-lapse fluorescence light microscopy reveals that sheaths of the
type VI secretion system cycle between assembly, quick contraction, disassembly and re-assembly. Whole-cell
electron cryotomography further shows that the sheaths appear as long tubular structures in either extended or
contracted conformations that are connected to the inner membrane by a distinct basal structure. These data support
a model in which the contraction of the type VI secretion system sheath provides the energy needed to translocate
proteins out of effector cells and into adjacent target cells.
Secretion systems allowbacteria totransport macromolecules such as
proteins out of effector cells or into either target host cells during
pathogenesis or target bacterial cells during competition in various
ecological settings. The type 6 secretion systems (T6SS) are encoded
byaclusterof 15–20genesthatispresent inatleast onecopyinabout
25% of all sequenced Gram-negative bacteria. Although linked to
virulence during host infection, recent studies showed that T6SS of
Pseudomonas, Burkholderia and Vibrio species can kill bacterial as
thought to kill target cells through translocation of toxic effector
proteins in a cell–cell contact-dependent process1–3,8. Little is known,
however, about how T6SS transport toxic proteins through their own
cell membranes or across target cell membranes.
T6SS components are structural homologues of components present
in contractile phage tails. For example, secreted VgrG proteins are
Another highly conserved T6SS gene product is predicted to be a
homologue of gp25, a major component of the T4 phage tail base-
plate10,12,13. Two T6SS gene products of V. cholerae, VipA and VipB,
form tubular structures that can be depolymerized by another T6SS
visually resemble T4 contracted tail sheath and were the first to
propose that a sheath-like structure might power T6SS translocation
by a phage tail-like contraction mechanism. Here we show that
T6SS-dependent secretion of Hcp and killing of Escherichia coli by
V. cholerae correlates with the activity of a dynamic intracellular
structure that indeed appears structurally and functionally related
to contractile phage tail sheath.
Fluorescence microscopic imaging of the T6SS
constructed a carboxy (C)-terminal fusion of VipA protein with super-
folder green fluorescent protein (sfGFP)17. As shown in Supplementary
Fig. 1a, VipA–sfGFP complements a chromosomal in-frame deletion
of vipA for Hcp secretion when the fusion protein is expressed from
microscopy revealed that the VipA–sfGFP fusion is associated with
and length of the cell. The number of visible structures in a single cell
Critically, these structures were not visible in vipB mutant cells (Sup-
of VipA with VipB10. Because VipA is not secreted and resides within
cellular fractions (Supplementary Fig. 3) and the expression level of
VipA–sfGFP was comparable to VipA under the conditions used to
visualize these sheath structures (Supplementary Fig. 4), we conclude
that the fluorescent structures were within the cytosol of imaged cells.
Time-lapse imaging revealed these putative sheath structures to be
different subcellular locations and then contracted and disassembled
also within tens of seconds. Most of the extended sheath structures
visible in cells stretched from one lateral side of the cell to the other,
perpendicular to the membrane, and thus had lengths approximately
equal to the width of the cell (about 0.75–1mm). As shown in Fig. 1a,
these sheaths assembled at speeds of approximately 20–30smm21.
Contraction was very fast, occurring in approximately 5ms or less
(unresolvable at frame rates of approximately 200 frames per second;
see Fig. 1b, c, Supplementary Fig. 5 and Supplementary Video 3).
Sheaths contracted to about 50% of their extended length (Sup-
plementary Fig. 6), and then disassembled over the next 30–60 s
(Fig. 1a). The disassembly of the contracted sheath is most probably
a ClpV-dependent event because ClpV is known to disassemble
do not disassemble VipA–sfGFP-labelled contracted sheaths (Sup-
plementary Fig. 2d and Supplementary Video 8). A similar number
of VipA–sfGFP-labelled sheaths were seen in strains 2740-80 and
V52 and at various levels of VipA–sfGFP expression (Supplementary
1Departmentof Microbiology and Immunobiology, Harvard MedicalSchool, 200 Longwood Avenue,Boston, Massachusetts02115, USA.2Division of Biology,California Institute of Technology,1200 East
California Boulevard, Pasadena, California 91125, USA.3Howard Hughes Medical Institute, California Institute of Technology, 1200 East California Boulevard, Pasadena, California 91125, USA.
*These authors contributed equally to this work.
1 8 2 | N A T U R E | V O L 4 8 3 | 8 M A R C H 2 0 1 2
Macmillan Publishers Limited. All rights reserved
Figs 2c, g and 4 and Supplementary Videos 4–6). The sheaths in wild-
type cells displayed the same extension–contraction–disassembly
cycles as sheaths observed in complemented vipA mutant cells (Sup-
plementary Videos 4 and 5) and when mCherry2 (ref. 18) was sub-
that contain exclusively VipA–sfGFP fusion protein, but also of
sheaths composed largely of wild-type VipA or other VipA fusion
proteins. VCA0109 encodes a member of a family of phage base-plate
was dispersed in the cytosol with only a rare, VipA–sfGFP-containing
VCA0109 therefore plays a critical role in the formation of functional
Electron cryotomographic imaging of the T6SS
To visualize the sheaths directly, we imaged wild-type and mutant
whole cells with electron cryotomography (ECT). ECT has been
shown to preserve and reveal bacterial cytoskeletal structures directly
in three dimensions in a near-native, life-like state19. ECT analyses of
wild-type 2740-80 cells showed straight, tubular structures which
10 s30 s 0 s310 s 340 s 270 s
0 s10 s
30 s90 s120 s140 s
0 s20 s50 s250 s290 s410 s
0 ms 150 ms 300 ms
Figure 2 | Electron cryotomographic imaging of T6SS structures inside
intact cells. Shown aredifferenttomographic slices(19nm in a, e, c, g; 9.5nm
in b, f; 190nm in d, h) of an extended (a–d) and a contracted (e–h) structure
imaged in two different wild-type cells (contracted/extended structures, T6SS;
SG, polyphosphate storage granule). b, f, Each part shows three slices at the
same orientation but at different z-heights. Compared with extended
(pitch angle 87u) and a smaller diameter (indicated in the perpendicular views
in d, h). c, g, Segmentations of densities observed in the extended (c) and
contracted (g) structures. Densities shown in h originate from a contracted
structure from a different tomogram. Segmented are putative densities
corresponding to sheath (green), base plate (pink and yellow) and membranes
(applies to c, d, g, h).
Figure 1 | Fluorescence light microscopy of VipA–sfGFP. a, Individual
3mm33 mm frames from a time-lapse imaging with a frame rate of 10s per
frame show three frames of extension of VipA–sfGFP structure in DVipA
background from one side of the cell to another (arrows) followed by a
contraction event and apparent disassembly (shown on three frames) of the
contracted VipA–sfGFP structure (arrows). Scale bar on the first frame
represents 1mm. The whole 10 min time-lapse sequence is shown in
Supplementary Video 1 with another 17 similar events; a larger field of cells is
length of the VipA–sfGFP structure. Projection of signal intensity in time at a
rate of 200 frames per second along the axis of the maximal intensity on an
length and increase in maximal intensity of the contracted structure (30 frame
average shown on c right). Arrows indicate contracting VipA–sfGFP structure
and mark the start and end of a line for generating the kymogram. Scale bar
shown on c left represents 1mm. Gaussian blur filter (sigma radius51) was
applied toindividualframesbefore generating thekymogram.All60 frames of
the time-lapse sequence are shown in Supplementary Video 3 (video segment
number 3) with four more contraction events imaged at the same or lower
8 M A R C H 2 0 1 2 | V O L 4 8 3 | N A T U R E | 1 8 3
Macmillan Publishers Limited. All rights reserved
appeared to exist in two different conformations: a longer, thinner
conformation (Fig. 2e–h). Both structures were oriented roughly per-
pendicular to the cytoplasmic membrane and were clearly located
exclusively in the cytosol (Supplementary Video 10). Tubular struc-
tures were observed in 26 of 90 imaged wild-type cells. Some cells
exhibited more than one tubule and on occasion both extended and
contracted conformations were seen in the same cell (Supplementary
in a vipB mutant (0of 53cells), a VCA109 mutant (0of 10cells)and a
VCA0109/ClpV double mutant (0 of 8 cells), strongly suggesting that
both types of tubule are T6SS-related structures.
Consistent with the dynamic sheaths in the two-dimensional fluor-
and 372656nm (n516), respectively. Although extended tubes had
diameters of 11.660.7nm, dense interiors, and a homogeneous
surface, we observed that contracted tubes were thicker (14.660.7nm
diameter), hollow, and had helical ridges (87u pitch angle) spaced 6nm
apart (Fig. 2). The tubular structures of both conformations were
usually found with one end in close proximity to the cytoplasmic
but instead appeared to be connected to it by a flaredbell-shaped base
(Fig. 2c, g, pink highlights). Distal to the flared base of extended, but
not contracted tubes, there was an additional conical-shaped density
(Fig. 2c, yellow highlights) that crossed the periplasm and protruded
through the outer membrane. Given that various T6SS components
have been localized to the inner membrane, periplasm and outer
membrane including a lipoprotein unique to T6SS as well as proteins
Purification of T6SS sheath from V. cholerae
To prove that the dynamic fluorescent structures observed in VipA–
sfGFP expressing cells and the tubes observed by ECT were indeed
structures from disrupted cells. Negative stain electron microscopic
analysis of macromolecular fractions purified from cell lysates of
wild-type cells revealed straight, hollow tubular structures similar
to, but more uniform than, the VipA/VipB tubules produced previ-
and Supplementary Fig. 1b left). No sheath-like structures were
detected in identically prepared samples from mutants defective in
Electron microscopic analysis of sheath preparations made from the
VipA–sfGFP complemented vipA mutant strain revealed sheath
structures similar to those produced by wild-type cells except that a
diffuse coatwaslaterally displayedon thefilament’ssurface, probably
and Supplementary Fig. 1b right).
To identify proteins that were associated with these sheaths, the
structures were purified from a non-flagellated mutant (flgG) of
V. cholerae 2740-80 (Supplementary Fig. 9a). Two major proteins
were enriched in these preparations with apparent molecular masses
of 55 and 20kDa, respectively (Supplementary Fig. 9b). Mass spec-
trometry analysis revealed that the 55kDa band was VipB and the
20kDa band was VipA (Supplementary Table 2). Interestingly, we
also identified four additional T6SS proteins in the sheath samples:
ClpV, VCA0109 (a gp25-like protein), and two other proteins within
the T6SS gene cluster encoded by genes VCA0111 and VCA0114.
in its polymerized state with VipA15. As noted earlier, VCA0109 is a
homologue of T4 base-plate protein gp25 (ref. 10) and a T6SS gp25-
like protein was recently shown to localize to the cytoplasm of
Pseudomonas aeruginosa13. The function of VCA0111 and VCA0114
are currently unknown, but they are essential components of the V.
as the contracted tubes seen previously inside cells (14.4nm diameter,
angle 87u). Interestingly, in addition to helical surface ridges, purified
sheaths exhibited cogwheel-like cross-sections with 12 surface
and thus are structurally similar to contracted T4 phage sheaths12
states of the dynamic VipA–sfGFP-labelled T6SS sheath that was
visualized using fluorescence light microscopy.
Contractile phage tails consist of a contractile outer sheath and an
The T6SS of V. cholerae is known to possess antibacterial activity
Figure 3 | Images of purified VipA/VipB sheaths and comparison with
VipA–sfGFP-labelled sheath (right) are highly similar except for flared extra
densities on the outside of the VipA–sfGFP-labelled structure. Cryotomograms
ofwild-type sheath (b, showing three 12.6-nm slices at different z-heights) were
surface pitch angle of 87u seen in tomographic slices (b) and an isosurface of a
subtomogram average (c; map deposited in the Electron Microscopy Data Base
under accession EMD-2045). The negatively stained perpendicular view of a
purified wild-type sheath showed the characteristic ‘cog-wheel’ like structure
with 12 paddles (d) and is similar to the perpendicular view of a contracted T4
phage sheath (e, left; two rings of six gp18 subunits, created in Chimera from
EMDB 1086 map). Similar to T6SS sheath (c), also the surface of the contracted
T4 phage sheath appears helical (e, right) though with a different pitch angle.
Scalebars: b, 20nm (applies to a, b); e, 10nm (applies to c–e). Note that protein
densities appear white in negative stain images and black in cryotomograms.
1 8 4 | N A T U R E | V O L 4 8 3 | 8 M A R C H 2 0 1 2
Macmillan Publishers Limited. All rights reserved
strain 2740-80 secretes abundant Hcp and this secretion is com-
pletely abolished by deletion of T6SS genes vipA, vipB and VCA0109
(whichencodesa gp25-like protein), ashasbeenpreviouslyshownfor
V. cholerae strain V52 (refs 7, 14). Although the material inside the
could not be resolved as a separate ‘inner tube’ per se, its diameter
was similar to the diameter of Hcp tubes described at either the crys-
tallographic or microscopic level16,24. Furthermore, contracted tubes
were clearly hollow (Fig. 2h and Supplementary Fig. 10c–f). Thus, we
propose that the thinner extended tubule found in whole cells is an
uncontracted ‘extended T6SS sheath’ whose VipA/VipB subunits are
probably wrapped around a thinner inner tube composed of Hcp
protein. Unfortunately, the uncontracted, extended T6SS sheaths
because of spontaneous sheath contraction during cellular disruption
and purification. Because Hcp was not found in purified contracted
T6SS sheaths, we conclude that the postulated inner Hcp tube of
extended sheaths is largely expelled from the cell at the moment of
ClpVand T6SS sheath recycling
Like strain V52 (refs 4, 7), V. cholerae 2740-80 also rapidly kills E. coli
strain V52 (ref. 7), the ClpV mutant of 2740-80 showed 90% loss of
well above background during incubation for 4h. Thus, ClpV is not
essential for T6SS function in V. cholerae. Because ClpV has been
shown to disassemble in vitro a tubular structure that is produced in
E. coli expressing VipA and VipB14, we asked whether ClpV affected
didnotobserve anypolymerization ordisassemblyeventsintheClpV
mutant background; rather, we found that most VipA–sfGFP existed
in static punctate structures (Supplementary Fig. 2d), which were
probably contracted T6SS sheaths because ClpV mutants produce
contracted sheath-like structures (Supplementary Fig. 1b).
Because our fluorescence microscopic analysis showed that con-
tracted sheath forms from extended sheath, it follows that ClpV may
is responsible for recycling VipA and VipB from contracted T6SS
the original site of extension (that is, the T6SS base-plate complex) and
then continued disassembly in a random fashion throughout the cell
new extended sheaths in many cells (Supplementary Videos 1 and 2).
steps in the functional cycle of the T6SS apparatus that will serve as a
framework for further studies (Fig. 4 and Supplementary Video 11, a
narrated animation of our data and model). Although analogous to
translocation events mediated by contractile phage tails, the proposed
T6SS process is different because it occurs in a topologically reversed
orientation and compartmentalization (within the cytosol), and further
undergoes efficient recycling through the action of other T6SS com-
ponents such as ClpV. Collectively, the data presented here provide
strong evidence that energy captured from conformational changes
in polymeric structures can rapidly transport proteins through cell
V. cholerae strains and genetic manipulation have been described previously7,25,26.
Full-length VipA was fused at its C terminus to either sfGFP or mCherry2 genes
(separated by a 33Ala 33Gly linker) and expressed from an arabinose-inducible
promoter27present on either plasmid pBAD24-VipA–sfGFP or pBAD24-VipA-
mCherry2. Cellsgrowntoanattenuanceat600nm(D600)of1.0inthe presenceof
various concentrations of arabinose were spotted on a thin pad of 1% agarose in
PBS and imaged at room temperature. Fluorescence micrographs were captured
using microscopes and camera combinations that are described in detail in
Methods and image analysis was performed using ImageJ 1.45 software. For
electronmicroscopy,sheath samples preparedasdescribedinMethodswerespot-
ted on carbon-coated grids, and stained with 1% uranyl formate. The grids were
examined in a JEOL 1200EX transmission electron microscope and images were
recorded with an AMT 2k CCD (charge-coupled device) camera. For ECT, V.
cholerae cells were grown to D60051.5–2.2, mixed with 10nm colloidal gold,
applied to an electron microscope grid and plunge-frozen in a liquid ethane–
propane mixture28. Tilt series were collected using a Polara 300 kV FEG transmis-
coupled 4k3 4kUltraCam. Pixels on the CCD represented 0.95nm (322,500) or
0.63nm(334,000) atthe specimen level.Leginon29orUCSFTomo30was usedfor
automatic tilt-series acquisition. Three-dimensional reconstructions were calcu-
lated using the IMOD software package31or Raptor32. IMOD31was used to model
the centre of the sheath, PEET33to align and average repeating sub-volumes, and
Chimera34to do isosurface rendering of the sub-volume averages.
Full Methods and any associated references are available in the online version of
the paper at www.nature.com/nature.
Received 3 October 2011; accepted 9 January 2012.
Published online 26 February 2012.
103, 1528–1533 (2006).
ATP + H2O
ADP + P
base plate complex
Figure 4 | Model of T6SS action. OM, outer membrane; PG, peptidoglycan;
IM, inner membrane. a, Assembly. The first step is a base-plate complex
probably composed of gp25, VgrG and other T6SS proteins that define a bell-
shaped cytoplasmic component (black objects) and periplasmic component
(brown objects), which together span the inner membrane, peptidoglycan and
VipB heterodimers) around the Hcp tube in an extended conformation.
b, Extended T6SS apparatus in extended ‘ready to fire’ conformation. The
membrane distal end may be capped by an unknown protein or VipAB
conformational state. c, Contraction. Upon an unknown extracellular signal, a
conformational change in the base-plate complex triggers sheath contraction
that leads to the translocation (secretion) of the VgrG/Hcp tube complex
through effector cell membranes and penetration of adjacent target cell
membrane. Translocation of additional effector proteins might then follow
new extended T6SS apparatus at either the original or a newly formed base-
plate complex. In the absence of target cell penetration (c), Hcp and VgrG
proteins are released into the extracellular space as secreted proteins.
8 M A R C H 2 0 1 2 | V O L 4 8 3 | N A T U R E | 1 8 5
Macmillan Publishers Limited. All rights reserved
2.Ma, A. T., McAuley, S., Pukatzki, S. & Mekalanos, J. J. Translocation of a Vibrio
Host Microbe 5, 234–243 (2009).
Russell, A. B. et al. Type VI secretion delivers bacteriolytic effectors to target cells.
Nature 475, 343–347 (2011).
MacIntyre, D. L., Miyata, S. T., Kitaoka, M. & Pukatzki, S. The Vibrio cholerae type VI
secretion system displays antimicrobial properties. Proc. Natl Acad. Sci. USA 107,
Schwarz, S. et al. Burkholderia type VI secretion systems have distinct roles in
eukaryotic and bacterial cell interactions. PLoS Pathog. 6, e1001068 (2010).
Hood, R. D. et al. A type VI secretion system of Pseudomonas aeruginosa targets a
toxin to bacteria. Cell Host Microbe 7, 25–37 (2010).
Zheng, J., Ho, B. & Mekalanos, J. J. Genetic analysis of anti-amoebae and anti-
bacterial activities of the type VI secretion system in Vibrio cholerae. PLoS ONE 6,
USA 107, 4365–4370 (2010).
system translocates a phage tail spike-like protein into target cells where it cross-
links actin. Proc. Natl Acad. Sci. USA 104, 15508–15513 (2007).
10. Leiman, P. G. et al. Type VI secretion apparatus and phage tail-associated protein
complexes share a common evolutionary origin. Proc. Natl Acad. Sci. USA 106,
11. Pell, L. G., Kanelis, V., Donaldson, L. W., Howell, P. L. & Davidson, A. R. The phage
lambda major tail protein structure reveals a common evolution for long-tailed
phages and the type VI bacterial secretion system. Proc. Natl Acad. Sci. USA 106,
12. Leiman, P. G., Chipman, P. R., Kostyuchenko, V. A., Mesyanzhinov, V. V. &
Rossmann, M. G. Three-dimensional rearrangement of proteins in the tail of
bacteriophage T4 on infection of its host. Cell 118, 419–429 (2004).
Microbiology 157, 3292–3305 (2011).
14. Bonemann, G., Pietrosiuk, A., Diemand, A., Zentgraf, H. & Mogk, A. Remodelling of
VipA/VipB tubules by ClpV-mediated threading is crucial for type VI protein
secretion. EMBO J. 28, 315–325 (2009).
in type VI protein secretion. J. Biol. Chem. 286, 30010–30021 (2011).
16. Mougous, J. D. et al. A virulence locus of Pseudomonas aeruginosa encodes a
protein secretion apparatus. Science 312, 1526–1530 (2006).
17. Pedelacq, J. D., Cabantous, S., Tran, T., Terwilliger, T. C. & Waldo, G. S. Engineering
24, 79–88 (2006).
18. Cho, H., McManus, H. R., Dove, S. L. & Bernhardt, T. G. Nucleoid occlusion factor
SlmA is a DNA-activated FtsZ polymerization antagonist. Proc.Natl Acad. Sci. USA
108, 3773–3778 (2011).
19. Pilhofer, M., Ladinsky, M. S., McDowall, A. W. & Jensen, G. J. Bacterial TEM: new
insights from cryo-microscopy. Methods Cell Biol. 96, 21–45 (2010).
Escherichia coli. J. Bacteriol. 190, 7523–7531 (2008).
21. Aschtgen, M. S., Thomas, M. S. & Cascales, E. Anchoring the type VI secretion
system to the peptidoglycan: TssL, TagL, TagP.what else? Virulence 1, 535–540
22. Aschtgen, M. S., Gavioli, M., Dessen, A., Lloubes, R. & Cascales, E. The SciZ protein
anchors the enteroaggregative Escherichiacoli type VIsecretion system to the cell
wall. Mol. Microbiol. 75, 886–899 (2010).
of contraction. Nature Struct. Mol. Biol. 12, 810–813 (2005).
24. Ballister, E. R., Lai, A. H., Zuckermann, R.N., Cheng, Y.& Mougous,J. D.Invitroself-
assembly of tailorable nanotubes from a simple protein building block. Proc. Natl
Acad. Sci. USA 105, 3733–3738 (2008).
25. Goldberg, S. & Murphy, J. R. Molecular epidemiological studies of United States
Gulf Coast Vibrio cholerae strains: integration site of mutator vibriophage VcA-3.
Infect. Immun. 42, 224–230 (1983).
26. Bina, J. E. & Mekalanos, J. J. Vibrio cholerae tolC is required for bile resistance and
colonization. Infect. Immun. 69, 4681–4685 (2001).
27. Guzman, L. M., Belin, D., Carson, M. J. & Beckwith, J. Tight regulation, modulation,
and high-level expression by vectors containing the arabinose PBAD promoter.
J. Bacteriol. 177, 4121–4130 (1995).
28. Tivol, W. F., Briegel, A. & Jensen, G. J. An improved cryogen for plunge freezing.
Microsc. Microanal. 14, 375–379 (2008).
29. Suloway, C. et al. Fully automated, sequential tilt-series acquisition with Leginon.
J. Struct. Biol. 167, 11–18 (2009).
30. Zheng, S. Q. et al. UCSF tomography: an integrated software suite for real-time
J. Struct. Biol. 157, 138–147 (2007).
31. Mastronarde, D. N. Correction for non-perpendicularity of beam and tilt axis in
tomographic reconstructions with the IMOD package. J. Microsc. 230, 212–217
32. Amat,F.etal.Markovrandomfield based automaticimagealignment for electron
tomography. J. Struct. Biol. 161, 260–275 (2008).
tomography. Science 313, 944–948 (2006).
and analysis. J. Comput. Chem. 25, 1605–1612 (2004).
Supplementary Information is linked to the online version of the paper at
Acknowledgements We thank T. G. Bernhardt and N. T. Peters for assistance with
fluorescence microscopy, discussions and for a gift of plasmids carrying sfGFP and
mCherry2 genes. We thank the Nikon Imaging Center at Harvard Medical School for
help with fluorescence microscopy, and Research Precision Instruments and
Hamamatsu for lending an ORCA-Flash2.8 camera. We thank the Harvard Medical
School Electron Microscopy Facility for help with and supervision of transmission
electron microscopy. We thank M. K. Waldor for a V. cholerae 2740-80 strain and
discussions. We thank D. Ewen Cameron for a knockout construct pWM91-flgG. We
thankB.WenandZ.Lifor initial cryotomographicstudies.Thisworkwassupported by
J.J.M. and National Institute of General Medical Sciences grant GM094800B to G.J.J.
Author Contributions All authors helped design and analyse experiments; M.B., M.P.
and G.P.H. performed experiments, and M.B., M.P., G.J.J. and J.J.M. wrote the paper.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interests.
Readers are welcome to comment on the online version of this article at
www.nature.com/nature. Correspondence and requests for materials should be
addressed to G.J.J. (Jensen@caltech.edu) or J.J.M.
1 8 6 | N A T U R E | V O L 4 8 3 | 8 M A R C H 2 0 1 2
Macmillan Publishers Limited. All rights reserved