, 1083 (2008);
et al.Shiho Tanaka,
Atomic-Level Models of the Bacterial Carboxysome
www.sciencemag.org (this information is current as of October 28, 2008 ):
The following resources related to this article are available online at
version of this article at:
including high-resolution figures, can be found in the online
Updated information and services,
can be found at:
Supporting Online Material
, 11 of which can be accessed for free:
cites 27 articles
2 article(s) on the ISI Web of Science.
This article has been
1 articles hosted by HighWire Press; see:
This article has been
This article appears in the following
in whole or in part can be found at:
permission to reproduce
of this article or about obtaining
Information about obtaining
registered trademark of AAAS.
is aScience 2008 by the American Association for the Advancement of Science; all rights reserved. The title
CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005.
(print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience
on October 28, 2008
26. B. D. Santer et al., Proc. Natl. Acad. Sci. U.S.A. 104,
27. A. Wood et al., Clim. Dyn. 16, 755 (2004).
28. E. P. Maurer, H. G. Hidalgo, Hydrol. Earth Syst. Sci.
Discuss. 4, 3413 (2007).
29. X. Liang, D. Lettermaier, A. Wood, S. Burges, J. Geophys.
Res. 99, 14415 (1994).
30. T. P. Barnett et al., Science 309, 284 (2005); published
online 2 June 2005 (10.1126/science.1112418).
31. X. Zhang et al., Nature 448, 461 (2007).
32. The choices were CCSM3-FV noise for significance testing,
PCM fingerprint, and statistical downscaling with the
CA method. In the multivariable case, PCM noise was
used for normalization.
33. The MIROC data were generously supplied by the
National Institute for Environmental Studies, Onogawa,
Tsukuba, Ibaraki, Japan. The PCM simulation had
previously been made available to the Scripps Institution
of Oceanography (SIO) by the National Center for
Atmospheric Research for the Accelerated Climate
Prediction Initiative project. Supported by the Lawrence
Livermore National Laboratory (LLNL) through a
Laboratory-Directed Research and Development grant to
SIO via the San Diego Super Computer Center for the
LUCSiD project; the U.S. Department of Energy and
NOAA through the International Detection and
Attribution Group (T.P.B.); Program of Climate Model
Diagnoses and Intercomparison grant DOE-W-7405-ENG-
48 (C.B., B.D.S., G.B., A.A.M.); the U.S. Geological Survey
and SIO (D.R.C., M.D.D.); and the California Energy
Commission (D.W.P., H.G.H.).
Supporting Online Material
Figs. S1 to S3
2 November 2007; accepted 23 January 2008
Published online 31 January 2008;
Include this information when citing this paper.
Atomic-Level Models of the Bacterial
Shiho Tanaka,1Cheryl A. Kerfeld,3,4Michael R. Sawaya,2Fei Cai,5Sabine Heinhorst,5
Gordon C. Cannon,5Todd O. Yeates1,2*
The carboxysome is a bacterial microcompartment that functions as a simple organelle by
sequestering enzymes involved in carbon fixation. The carboxysome shell is roughly 800 to 1400
angstroms in diameter and is assembled from several thousand protein subunits. Previous studies
have revealed the three-dimensional structures of hexameric carboxysome shell proteins, which
self-assemble into molecular layers that most likely constitute the facets of the polyhedral shell.
Here, we report the three-dimensional structures of two proteins of previously unknown function,
CcmL and OrfA (or CsoS4A), from the two known classes of carboxysomes, at resolutions of 2.4 and
2.15 angstroms. Both proteins assemble to form pentameric structures whose size and shape are
compatible with formation of vertices in an icosahedral shell. Combining these pentamers with
the hexamers previously elucidated gives two plausible, preliminary atomic models for the
ing the key enzymes ribulose-1,5-bisphosphate
he carboxysome enhances CO2fixation
inside many photosynthetic and chemo-
autotrophic bacterial cells by encapsulat-
carboxylase-oxygenase (RuBisCO) and carbonic
anhydrase (1–3). In contrast to membrane-bound
eukaryotic organelles, carboxysomes and related
bacterial microcompartments (4, 5) have a pro-
teinaceous outer shell, which is roughly polyhe-
carboxysomes by electron microscopy date back
more than 40 years (6, 7). Subsequent genetic
and biochemical studies have provided essential
(3, 8–10); structural studies (11–14) have begun
to illuminate the functional mechanisms and ar-
chitectural details of the carboxysome.
Carboxysomes are found in all cyanobacteria
and in some chemoautotrophic bacteria. Two types
of carboxysomes have been defined by patterns
comparisons (15). They are represented in two
model organisms, Halothiobacillus neapolitanus
1Department of Chemistry and Biochemistry, University of
California at Los Angeles (UCLA), Los Angeles, CA 90095,
USA.2UCLA–Department of Energy Institute for Genomics
and Proteomics, Los Angeles, CA 90095, USA.3Department
of Energy–Joint Genome Institute, Walnut Creek, CA 94598,
USA.4Department of Plant and Microbial Biology, University
of California, Berkeley, CA 94720, USA.5Department of Chem-
istry and Biochemistry, University of Southern Mississippi,
Hattiesburg, MS 39406, USA.
*To whom correspondence should be addressed. E-mail:
Fig. 1. Carboxysome ar-
chitecture and operon or-
ing (A) (left) a section
through a dividing cyano-
bacterial cell, Syn. 6803
(scale bar, 200 nm), and
(right) an enlargement
of a single carboxysome
(scale bar, 50 nm) and
neapolitanus cells with
their carboxysomes high-
lighted by arrows and
(right) purified carboxy-
somes (scale bars, 100 nm). Their polyhedral shape helps
distinguish carboxysomes from other cytoplasmic inclusions.
(C) A diagram showing the construction of a large icosa-
hedron from many smaller hexagons and 12 pentagons at
of 75 (29). (D) Genomic arrangement of carboxysome-
from ref. (11); (B) is adapted from ref. (12).
VOL 31922 FEBRUARY 2008
on October 28, 2008
(a chemoautotroph containing a-carboxysomes)
and Synechocystis sp. strain PCC (Pasteur
Culture Collection) 6803 (Syn. 6803, a cyano-
bacterium containing b-carboxysomes); potential
functional differences between the two types are
not clear. The carboxysome shell is composed
predominantly of CsoS1 proteins (homologous
proteins CsoS1A, B, and C) in H. neapolitanus
(16, 17) and of CcmK proteins (homologous
have previously determined the structures offour
of these shell proteins and found that they are all
hexamers and that they tend to form extended,
tightly packed molecular layers hypothesized to
represent the flat facets of the polyhedral car-
boxysome shell. The structures suggest that diffu-
is limited by the small hexameric pores and gaps
between the hexamers (11, 12).
Two recent electron microscopy studies have
confirmed that carboxysomes are approximate-
of large icosahedral structures typically requires
a combination of hexameric and pentameric
units (18, 19). Pentamers generate curvature in
an otherwise flat hexagonal sheet and occupy
the vertices of an icosahedral shell (Fig. 1C).
Candidates for such pentameric components of
the shell had not been identified. Functions
have been elucidated for most of the proteins
encoded within well-characterized carboxysome
operons (3). However, functions had not been
assigned to the gene products of two open
reading frames in H. neapolitanus, designated
orfA and orfB, or their homolog in Syn. 6803,
CcmL (Fig. 1D).
The CcmL protein from Syn. 6803 was ex-
pressed in Escherichia coli cells, and its crystal
structure was determined by selenomethionine
An atomic model was built and refined at a
resolution of 2.4 Å, with a final model consisting
of 96 of the 100 residues of the complete CcmL
protein (20). OrfAwas expressed in E. coli cells,
and its crystal structure was determined by mo-
lecular replacement using the CcmL structure as
a search model; the amino acid sequence identity
between the two proteins is 37%. The OrfA
structure was refined at a resolution of 2.15 Å,
yielding a final model that contains 82 of the 83
amino acid residues in that protein (20). The
structures of CcmL and OrfA are highly similar;
their backbones superimpose with a root mean
square deviation (RMSD) of only 1.0 Å (Fig. 2).
CcmL consists of seven b strands and one a
helix,withfiveoftheb strandsformingab barrel
(Fig. 2). OrfA contains the same central five-
stranded b barrel and the a helix, but it lacks the
to its smaller size.
Both CcmL and OrfA formed symmetric
pentamers in their respective crystal structures
(Fig. 2). The individual subunits of the penta-
mers are held tightly together by their protrud-
ing C-terminal regions, which form extensive
interactions with neighboring subunits. The large
subunit interfaces and the structural agreement
between the two proteins argue that the observed
pentameric structures represent their natural bio-
logical forms. The CcmL pentamer is shaped
roughly like a pentagonal disk 30 to 35 Å thick.
One side of the disk, where both the N and C
pentamer the appearance of a truncated pyramid.
Its base has a pentagonal edge of ~42 Å (Fig. 2),
which narrows to ~35 Å in the middle of the
the axis of symmetry, leaving only a narrow pore
through the center, with a diameter of ~5 Å in
CcmL and ~3.5 Å in OrfA. This tight packing
around a narrow pore is reminiscent of the hexa-
meric carboxysome shell proteins. The tight
and out of the carboxysome (11, 12).
The database of known protein structures
was searched for proteins similar to CcmL and
OrfA. A single protein, EutN from E. coli, was
identified as having a similar three-dimensional
fold. The protein backbone of EutN super-
imposes on that of CcmL with an RMSD of
only 1.3 Å (Fig. 2C). The structure of EutN had
been determined as part of a structural genomics
program (21), so detailed biological interpretation
was not provided. It is known, however, that the
typhimurium (22), encodes several proteins in-
volved in ethanolamine utilization, which takes
place inside the eut microcompartment. Remark-
ably, however, the reported EutN structure is
hexameric rather than pentameric (Fig. 2). The
difference between the oligomeric state of EutN,
compared with those of CcmL and OrfA, pre-
sumably reflects structural differences between
the eut microcompartment and the carboxysome,
as well as a different functional role for EutN.
Aside from EutN, no other homolog of CcmL or
which suggests that some other potentially un-
related protein might serve as a pentameric shell
the eut microcompartment could lack pentamers,
which might explain why the microcompart-
icosahedral shapes than carboxysomes.
The pentameric organization of CcmL and
OrfA suggests that these proteins serve as
Fig. 2. Crystal structures of the carboxysome proteins CcmL and OrfA revealing pentagonal symmetry.
(A) Structure of the CcmL monomer from Syn. 6803. (B) A comparison of similar structures: CcmL (blue),
OrfA (or CsoS4A) from H. neapolitanus (yellow), and EutN from E. coli (pink) (PDB 2Z9H). The RMSD
between the protein backbones of CcmL and OrfA is 1.0 Å, and 1.3 Å between CcmL and EutN. (C) CcmL
and OrfA assemble as natural pentamers. EutN, which is part of the eut operon that encodes proteins
views of the CcmL pentamer showing a pentagonal disk with slanted sides.
22 FEBRUARY 2008 VOL 319
on October 28, 2008
vertices in their respective carboxysome shells.
Whether there are significant functional differ-
ences between OrfA and its adjacent paralog in
H. neapolitanus, OrfB, is unknown. The role
proposed here for CcmL and OrfA is consistent
with the earlier observation that deleting the ccmL
(9), as would be expected if the component re-
quired for vertex formation is lost. In large icosa-
hedral shells, 12 pentamers are present among
a much greater number of hexamers. In the car-
boxysome, only 60 copies of the CcmL or OrfA
(and possibly OrfB) subunits would be expected
to be present among about four or five thousand
hexameric shell subunits (3, 12, 23) and a similar
number of total RuBisCO subunits (14). This ex-
OrfA or OrfB in preparations of carboxysomes.
Knowing the structures of both the hexa-
meric and pentameric components of the car-
of the complete shell. The size and shape of the
CcmL and OrfA pentamers make them suitable
for insertion into the hexagonally packed molec-
ular layers previously elucidated for the carboxy-
some shell. The hexameric units (i.e., of CcmK
subunits) in the Syn. 6803 shell are packed to-
gether with an edge length of 40 Å (20) [sup-
porting online material (SOM) text]. This size is
consonant with the edge length of the CcmL
pentamers. The suggestion that the CcmL and
OrfA pentamers generate curvature by being
inserted into an otherwise flat hexagonal molec-
ular layer is consistent with recent electron mi-
croscopy investigations in which carboxysomes
exhibit a relatively simple polyhedral shape at
their vertices (13, 14). Some large viral capsids
at their pentameric vertices (24, 25). The lack of
such protrusions at the vertices of the carboxy-
some places useful constraints on atomic models
of the carboxysome that can be built from the
known hexameric and pentameric components.
CcmL (or OrfA) pentamers can be fit into
vacancies created by folding up hexameric mo-
lecular layers of CcmK (or CsoS1) proteins in
four distinct ways (Fig. 3). It has not yet been
possible either by biophysical methods or by
electron microscopy to define which side of the
hexagonally packed protein layers represents the
inside of the carboxysome and which side faces
outward into the bacterial cytosol (12). The hex-
agonal layers, therefore, could be folded up with
either side facing outward (fig. S1). Likewise,
which side of the CcmL or OrfA pentamer faces
inward and which faces outward is unknown.
Computational attempts were made to fit the
CcmL and OrfA pentamers into their respective
hexagonal layers, evaluating all four distinct
(inside versus outside) orientations of the pen-
tamers and hexamers. In addition, rotation and
translation of the pentamer about its central axis
was allowed, but given the tightness of the fit,
only a narrow range of rotations and translations
was feasible. The best-fitting solutions were
identified by evaluating geometric fit (26) and
empirical energy functions (20, 27). Among the
and could be rejected. The other three led to
potential packing solutions whose fits could be
compared (fig. S2). One of these provided a
much lower amount of buried surface area be-
tween the pentamer and its surrounding hexamers,
and left largergaps in the resulting shell and was,
therefore, judged unlikely to be correct. Two
plausible solutions for building complete models
of the shell remained (Fig. 3). Both candidate
Fig. 3. Modelsofthecarboxysomeshellbasedonpentamerandhexamercomponents.(A)Aflatlayerof
hexagons can be folded to give pentagonal vertices by removing one sector at each vertex. Twelve such
vertices are present in an icosahedral shell. (B) Taken in combination, alternate choices for the curvature
1 to 4 according to the quality of fit. Combination 4 led to impossible steric collisions. The structures are
colored according to calculated electrostatic potential, from negative (red) to positive (blue). (C)
Illustration of the best packing solutions for constructions 1 to 3. EN, calculated packing energies (27)
(with more negative values being favorable); SC, surface complementarity (26); and SA, buried surface
area between a pentamer and a single neighboring hexamer (with higher values of these parameters
being favorable). (D) Two alternate models for the complete carboxysome shell, based on the two
constructions, 1 and 2, judged to be most plausible. There are 740 hexamers and 12 pentamers in a T =
75 arrangement. The packing of hexamers is derived from multiple consistent crystal structures. The two
models differ with respect to the orientation of the hexameric layer. The hexagonal layer is colored
magenta. The diameter from vertex to vertex is 1150 Å.
VOL 31922 FEBRUARY 2008
on October 28, 2008
solutions orient the pentamer with its broader Download full-text
base facing outward. This is consistent with the
role the pentamer appears to play in introducing
curvature into the hexagonal layer. A notable
feature of the hexagonal layer of the carboxy-
some shell is the presence of a bowl-shaped
depression or concavity on one side of the hex-
americ building block (figs. S3 and S4). This
depression is situated on the same side of the
hexamer as both the N and C termini of the pro-
tein chain, and the prominence of the depression
is affected by the disposition of the C terminus,
which tends to vary in conformation between
different homologs of the carboxysome hexamer
(11, 12). The narrow pore through the carboxy-
some hexamer tends to be positively charged,
but the bowl-shaped depression surrounding the
pore has a partial hydrophobic character, giving
one side on the hexameric layer a distinctive ap-
pearance (Fig. 3). The model building does not
or outward. In one of the plausible models con-
structed here, it faces inward and could interact
with RuBisCO, carbonic anhydrase, or possibly
other protein components.
The present findings clarify the roles of pre-
viously uncharacterized proteins in the carboxy-
some; accordingly, we propose that the genes
orfA and orfB be renamed csoS4A and csoS4B,
consistent with other known shell protein genes
in a-type carboxysomes. The results also lead to
atomic models for the carboxysome shell, but
these are incomplete in numerous respects, and
considerably more work will be required to
complete our understanding of this structure.
For instance, the details of how the pentamers
and hexamers fit together are only approximated
here, and how the edges of the icosahedral shell
are formed where a hexagonal layer bends is
certain to exist between the shell and the other
enzymatic components of the carboxysome are
only beginning to be elucidated (28).
The emerging structure of the carboxysome
terial microcompartments and certain previously
characterized viral capsids are constructed and
illustrates the ability of evolution to produce a
diversity of highly complex molecular structures
basic building blocks. The elucidation of the
component structures of the carboxysome shell
also opens up strategies for designing new kinds
of molecular containers or reaction chambers on
the mid-nanometer scale.
References and Notes
1. J. M. Shively, R. S. English, S. H. Baker, G. C. Cannon,
Curr. Opin. Microbiol. 4, 301 (2001).
2. M. R. Badger, G. D. Price, J. Exp. Bot. 54, 609 (2003).
3. S. Heinhorst, G. C. Cannon, J. M. Shively, in Complex
Intracellular Structures in Prokaryotes, J. M. Shively, Ed.
(Springer, Berlin, 2006), vol. 2, pp. 141–165.
4. T. A. Bobik, Appl. Microbiol. Biotechnol. 70, 517
5. J. T. Penrod, J. R. Roth, J. Bacteriol. 188, 2865 (2006).
6. J. M. Shively, F. L. Ball, B. W. Kline, J. Bacteriol. 116,
7. E. Gantt, S. F. Conti, J. Bacteriol. 97, 1486 (1969).
8. R. S. English, S. C. Lorbach, X. Qin, J. M. Shively,
Mol. Microbiol. 12, 647 (1994).
9. G. D. Price, S. M. Howitt, K. Harrison, M. R. Badger,
J. Bacteriol. 175, 2871 (1993).
10. A. K. So et al., J. Bacteriol. 186, 623 (2004).
11. C. A. Kerfeld et al., Science 309, 936 (2005).
12. Y. Tsai et al., PLoS Biol. 5, e144 (2007).
13. M. F. Schmid et al., J. Mol. Biol. 364, 526 (2006).
14. C. V. Iancu et al., J. Mol. Biol. 372, 764 (2007).
15. M. R. Badger, D. T. Hanson, G. D. Price, Funct. Plant Biol.
29, 161 (2002).
16. J. M. Shively et al., Can. J. Bot. 76, 906 (1998).
17. G. C. Cannon et al., Appl. Environ. Microbiol. 67, 5351
18. J. E. Johnson, J. A. Speir, J. Mol. Biol. 269, 665 (1997).
19. D. L. Caspar, A. Klug, Cold Spring Harbor Symp. Quant.
Biol. 27, 1 (1962).
20. Materials and methods are available as supporting
material on Science Online.
21. Z. Wunderlich et al., Proteins 56, 181 (2004).
22. E. Kofoid, C. Rappleye, I. Stojiljkovic, J. Roth, J. Bacteriol.
181, 5317 (1999).
23. J. M. Shively, R. S. English, Can. J. Bot. 69, 957 (1991).
24. J. T. Huiskonen, V. Manole, S. J. Butcher, Proc. Natl.
Acad. Sci. U.S.A. 104, 6666 (2007).
25. P. L. Stewart, R. M. Burnett, M. Cyrklaff, S. D. Fuller, Cell
67, 145 (1991).
26. M. C. Lawrence, P. M. Colman, J. Mol. Biol. 234, 946
27. J. J. Gray et al., J. Mol. Biol. 331, 281 (2003).
28. B. M. Long, M. R. Badger, S. M. Whitney, G. D. Price,
J. Biol. Chem. 282, 29323 (2007).
29. See the Virus Particle Explorer, http://viperdb.scripps.edu/.
30. The authors thank D. Cascio for crystallographic advice
and the staff at the Advanced Light Source beamline
8.2.2 for technical assistance. This work was supported by
a grant from the Biological and Environmental Research
program of the Department of Energy Office of Science.
Support was also provided by NSF (grants MCB-0444568
and DMR-0213883 to G.C.C. and S.H., respectively).
Protein structure coordinates and structure factors have
been deposited in the Protein Data Bank (PDB); accession
numbers are 2QW7 (CcmL), 2RCF (OrfA or CsoS4A), and
Supporting Online Material
Materials and Methods
Figs. S1 to S4
Tables S1 and S2
9 October 2007; accepted 3 January 2008
Differential Regulation of Dynein and
Kinesin Motor Proteins by Tau
Ram Dixit, Jennifer L. Ross,* Yale E. Goldman, Erika L. F. Holzbaur†
Dynein and kinesin motor proteins transport cellular cargoes toward opposite ends of microtubule
tracks. In neurons, microtubules are abundantly decorated with microtubule-associated proteins
(MAPs) such as tau. Motor proteins thus encounter MAPs frequently along their path. To determine
the effects of tau on dynein and kinesin motility, we conducted single-molecule studies of motor
proteins moving along tau-decorated microtubules. Dynein tended to reverse direction, whereas
kinesin tended to detach at patches of bound tau. Kinesin was inhibited at about a tenth of the tau
concentration that inhibited dynein, and the microtubule-binding domain of tau was sufficient to
inhibit motor activity. The differential modulation of dynein and kinesin motility suggests that
MAPs can spatially regulate the balance of microtubule-dependent axonal transport.
this process are associated with dysfunction and
disease (1). Much of the active transport in cells
depends on the molecular motor proteins cyto-
plasmic dynein and kinesin-1, which transport
ctive transport of cytoplasmic material
along microtubules is critical for cell or-
ganization and function, and defects in
cargo toward the minus-end (toward the cell
center) and plus-end of microtubules (toward the
cell periphery), respectively. Dynein and kinesin
have very different structures and translocation
mechanisms (2). Kinesin has a compact motor
domain and walks unidirectionally along single
protofilaments with 8-nm steps (2). In contrast,
and is capable of variable step sizes, lateral steps
across the microtubule surface, and processive
runs toward both the minus- and plus-end of the
microtubule (3–5). Cytoplasmic dynein function
in vivo also requires an accessory complex, dy-
to facilitate dynein processivity (6) and may also
regulate dynein activity (5). Within the cell, the
In the crowded cell environment, dynein and
kinesin compete with nonmotile microtubule-
associated proteins (MAPs) for binding to the
microtubule surface. MAPs bound to micro-
tubules might also block the path of motor
proteins. Thus, MAPs can provide spatio-
Department of Physiology and Pennsylvania Muscle Institute,
University of Pennsylvania, Philadelphia, PA 19104, USA.
*Present address: 302 Hasbrouck Laboratory, Department
of Physics, University of Massachusetts at Amherst, Amherst,
MA 01003, USA.
†To whom correspondence should be addressed. E-mail:
22 FEBRUARY 2008VOL 319
on October 28, 2008