Content uploaded by Sharon Grayer Wolf
Author content
All content in this area was uploaded by Sharon Grayer Wolf on Aug 19, 2016
Content may be subject to copyright.
Nature © Macmillan Publishers Ltd 1998
8
letters to nature
NATURE
|
VOL 393
|
14 MAY 1998 191
general Pol II transcription factors7,20,25,26. All protein fractions were dialysed
against buffer B (25 mM Tris-HCl, pH 7.9, 50 mM KCl, 0.5 mM dithiothreitol,
0.1 mM EDTA and 20% glycerol (v/v)). 25-ml reactions contained either 25 ng
17M/5pAL7 and pG1 (ref. 5) or 100 ng pTEF(D-138) or pTEF(D-138
TATA
)15,
with aliquots of TFTC, TFIIDband recombinant TBP5. Where indicated,
200 ng purified anti-TBP monoclonal antibody 1C2 was also included in the
reactions before the other factors were added. GAL-VP16-activated transcrip-
tion was performed as described5. After the preincubation steps (30 min),
transcription was initiated by addition of nucleoside triphosphates to 0.5 mM
and MgCl
2
to 5 mM. Transcriptions were incubated at 25 8C for 45 min.
Correctly initiated transcripts from the different promoters were analysed by
quantitative S1 nuclease analysis15,27.
DNase I footprinting. DNase I footprinting was performed as described18,19.
The labelled AdMLP-containing probes were amplified by polymerase chain
reaction on either the 17M5/pAL7 (ref. 28) (Fig. 3a) or the pM677 (ref. 29)
(Fig. 3b) templates. For the footprinting experiments, ten times more TBP,
TFIIDb, and TFTC was used than in the transcription reactions.
Received 4 December 1997; accepted 6 March 1998.
1. Zawel, L. & Reinberg, D. Common themes in assembly and function of eukaryotic transcription
complexes. Annu. Rev. Biochem. 64, 533–561 (1995).
2. Roeder, R. G. The role of general initiation factors in transcription by RNA polymerase II. Trends
Biochem. Sci. 21, 327– 335 (1996).
3. Verrijzer, C. P. & Tjian, R. TAFs mediate transcriptional activation and promoter selectivity [see
comments]. Trends Biochem. Sci. 21, 338–342 (1996).
4. Chang, M. & Jaehning, J. A. A multiplicity of mediators: alternative forms of transcription complexes
communicate with transcriptional regulators. Nucleic Acids Res. 25, 4861–4865 (1997).
5. Brou, C. et al. Distinct TFIID complexes mediate the effect of different transcriptional activators.
EMBO J. 12, 489– 499 (1993).
6. Jacq, X. et al. Human TAFII30 is present in a distinct TFIID complex and is required for
transcriptional activation by the estrogen receptor. Cell 79, 107– 117 (1994).
7. Bertolotti, A., Lutz, Y., Heard, D. J., Chambon,P. & Tora, L. hTAFII68 a novel RNA/SSDNA-binding
protein with homology to the pro-oncoproteins TLS/FUS and EWSis associated with both TFIID and
RNA polymerase II. EMBO J. 15, 5022– 5031 (1996).
8. Mengus, G. et al. Cloning and characterization of hTAFII118, hTAFII20and hTAFII28; three subunits
of the human transcription factor TFIID. EMBO J. 14, 1520–1531 (1995).
9. Lescure, A. et al. The N-terminal domain of human TATA-binding protein plays a role in transcription
from TATA-containing RNA polymerase II and III promoters. EMBO J. 13, 1166–1175 (1994).
10. Usheva, A. & Shenk, T. TATA-binding protein-independent initiation: YY1, TFIIB, and RNA
polymerase II direct basal transcription on supercoiled template DNA. Cell 76, 1115–1121 (1994).
11. Timmers, H. T. M., Meyers, R. E. & Sharp, P. A. Composition of transcription factor B-TFIID. Proc.
Natl Acad. Sci. USA 89, 8140– 8144 (1992).
12. Van der Knaap, J. A., WillemBorst, J., van der Vliet, P. C., Gentz, R. & Timmers, H. T. M. Cloning of
the cDNA for the TATA-binding protein-associated factor
II
170 subunit of transcription factor B-
TFIID reveals homology to global transcription regulators in yeast and Drosophila.Proc. Natl Acad.
Sci. USA 94, 11827– 11832 (1997).
13. Hansen, S. K., Takada, S., Jacobson, R. H., Lis, J. T. & Tjian, R. Transcriptionproperties of a cell type
specific TATA-binding protein, TRF. Cell 91, 71–83 (1997).
14. Crowley, T.E., Hoey, T.,Liu, J. K., Jan, Y.N., Jan, L. Y. & Tjian,R. A new factor related to TATA-binding
protein has highly restricted expression patterns in Drosophila.Nature 361, 557 –561 (1993).
15. Boam, D. S., Davidson, I. & Chambon, P. A TATA-less promoter containing binding sites for
ubiquitous transcription factors mediates cell type-specific regulation of the gene for transcription
enhancer factor-1 (TEF-1). J. Biol. Chem. 270, 487–494 (1995).
16. Nakajima, N., Horikoshi, M. & Roeder, R. G. Factors involved in specific transcription by mammalian
RNA polymerase II: purification, genetic specificity, and TATA box-promoter interactions of TFIID.
Mol. Cell. Biol. 8, 4028–4040 (1988).
17. Pugh, B. F. & Tjian, R. Transcription from a TATA-less promoter requires a multisubunit TFIID
complex. Genes Dev. 5, 1935–1945 (1991).
18. Purnell, B. A., Emanuel, P. A. & Gilmour, D. S. TFIID sequence recognition of the initiator and
sequences further downstream in Drosophila class II genes. Genes Dev. 8, 830– 842 (1994).
19. Oelgeschlager, T., Chiang, C. M. & Roder, R. G. Topology and reorganization of a human TFIID-
promoter complex. Nature 382, 735–738 (1996).
20. Dubrovskaya, V. et al. Distinct domains of hTAF
II
100 are required for functional interaction with
transcription factor TFIIFb (RAP30) and incorporation into the TFIID complex. EMBO J. 15, 3702–
3712 (1996).
21. Ruppert, S. & Tjian, R. Human TAFII250 interacts with RAP74: implications for RNA polymerase II
initiation. Genes Dev. 9, 2747–2755 (1995).
22. Hisatake, K. et al. Evolutionary conservation of human TATA-binding-polypeptide-associated factors
TAFII31 and TAFII80 and interactions of TAFII80 with other TAFs and with general transcription
factors. Proc. Natl Acad. Sci. USA 92, 85–89 (1995).
23. Lavigne, A. C. et al. Multiple interactions between hTAFII55 and other TFIID subunits. Requirements
for the formation of stable ternary complexes between hTAFII55 and the TATA-binding protein.
J. Biol. Chem. 271, 774–780 (1996).
24. Lennon, G., Auffray, C., Polymeropoulos, M. & Soares, M. B. The I.M.A.G.E. Consortium: an
integrated molecular analysis of genomes and their expression. Genomics. 33, 151 –152 (1996).
25. Gerard, M. et al .Purification and interaction properties of the human RNA polymerase B(II) general
transcription factor BTF2. J. Biol. Chem. 266, 20940– 20945 (1991).
26. De Jong, J. & Roeder, R. G. A single cDNA, hTFIIA/alpha, encodes both the p35 and p19 subunits of
human TFIIA. Genes Dev. 7, 2220–2234 (1993).
27. Tora, L. et al. The human estrogen receptor has two independent nonacidic transcriptional activation
functions. Cell 59, 477–487 (1989).
28. Brou, C. et al. Different TBP-associated factors are required for mediating the stimulation of
transcription in vitro by the acidic transactivator GAL-VP16 and the two nonacidic activation
functions of the estrogen receptor. Nucleic Acids Res. 21, 5–12 (1993).
29. Moncollin, V., Miyamoto, N. G., Zheng, X. M. & Egly, J. M. Purification of a factor specific for the
upstream element of the adenovirus-2 major late promoter. EMBO J. 5, 2577 –2584 (1986).
Acknowledgements. We thank P. Chambon for support; J. C. Dantonel for help in identification and
cloning of hTLF; E. Scheer for technical assistance; Y. Lutz for antibodies; D. Boam, V. Dubrovskaya, A. C.
Lavigne, G. Mengus, I. Davidson and the IMAGE Consortium for reagents; H. T. M. Timmers for
antibodies and for discussing unpublished results; A. Bertolotti for discussions; D. J. Heard for discussions
and reading the manuscript; P. Eberling for peptide synthesis; the cell culture group for HeLa cells;
R. Buchert, J.-M. Lafontaine and B. Boulay for illustrations; and A. Ozyhar for his contribution to the
training of E.W.E.W. was supported by a fellowship from the Ministe
`re de l’Enseignement Supe
´rieur et de
la Recherche. Research wassuppor ted by grants fromthe CNRS, the INSERM, the Ho
ˆpital Universitaire
de Strasbourg, the Ministe
`re de la Recherche et Technologie, the Fondation pour la Recherche Me
´dicale
and the Association pour la Recherche contre le Cancer.
Correspondence and requests for materials should be addressed to L.T. (e-mail: laszlo@titus.u-strasbg.fr).
corrections
Structure of the ab
tubulin dimer by
electron crystallography
Eva Nogales, Sharon G. Wolf & Kenneth H. Downing
Nature 391, 199–203 (1998)
..................................................................................................................................
In this Letter, the numbers for the secondary structure elements
involved in Taxol binding are incorrect (page 202, second-to-last
paragraph of main text). The sentences giving the correct numbers
are, ‘‘In our model, the C-39is near the top of helix H1 (that is,
between b:15–25), and the C2 group near H6 and the H6–H7 loop
(that is, between b:212–222). The main interaction of the taxane
ring is at L275, at the beginning of the B7–H9 loop.’’ M
Spatial and temporal
organization during
cardiac fibrillation
Richard A. Gray, Arkady M. Pertsov & Jose
´Jalife
Nature 392, 75–78 (1998)
..................................................................................................................................
The x-axis of Fig. 1d was mislabelled: the frequency values should
instead read 0, 10, 20, 30, 40 Hz. M
Nature © Macmillan Publishers Ltd 1998
8
Structure of the ab
tubulin dimer by
electron crystallography
Eva Nogales, Sharon G. Wolf*& Kenneth H. Downing
Life Science Division, Lawrence Berkeley National Laboratory, Berkeley,
California 94720, USA
.........................................................................................................................
The ab tubulin heterodimer is the structural subunit of micro-
tubules, which are cytoskeletal elements that are essential for
intracellular transport and cell division in all eukaryotes. Each
tubulin monomer binds a guanine nucleotide, which is non-
exchangeable when it is bound in the asubunit, or N site, and
exchangeable when bound in the bsubunit, or E site. The a- and
b-tubulins share 40% amino-acid sequence identity, both exist in
several isotype forms, and both undergo a variety of post-
translational modifications1. Limited sequence homology has
been found with the proteins FtsZ2and Misato3, which are
involved in cell division in bacteria and Drosophila, respectively.
Here we present an atomic model of the ab tubulin dimer fitted to
a 3.7-A
˚density map obtained by electron crystallography of zinc-
induced tubulin sheets. The structures of a- and b-tubulin are
basically identical: each monomer is formed by a core of two b-
sheets surrounded by a-helices. The monomer structure is very
compact, but can be divided into three functional domains: the
amino-terminal domain containing the nucleotide-binding
region, an intermediate domain containing the Taxol-binding
site, and the carboxy-terminal domain, which probably constitu-
tes the binding surface for motor proteins.
In the presence of zinc ions, purified tubulin assembles into two-
dimensional sheets that are ideal samples for electron crystallogra-
phy studies4. In these sheets, protofilaments appear to be similar to
those in microtubules, but associated in an antiparallel fashion. The
zinc-induced tubulin sheets used here are cold-labile and require
GTP for assembly. Addition of taxol stabilizes the sheets against low-
temperature depolymerization and ageing4, an effect similar to that
on microtubules. Taxol binds to a single site on the dimer in the
sheets, near lateral contacts between protofilaments5.
We have previously described three-dimensional density maps of
tubulin at 6.5 and 4 A
˚(refs 5, 6). The present model has been built
into a 3.7-A
˚map derived from a data set that includes 93 electron-
diffraction patterns (providing structure factor amplitudes) and
159 images (providing experimental phases). Table 1 and Fig. 1
summarize the data. The high quality of the phases produced a clean
map, with well defined connectivity, which is readily interpretable
in terms of secondary-structure elements. The present model of
tubulin has been built in the raw density map without refinement
and includes all but the last 10 and 18 C-terminal residues of a- and
b-tubulin, respectively. Figure 2 shows the density map and the
model in several regions of the dimer.
The density maps for a- and b-tubulin are almost superimpos-
able. Differences are limited to the length and conformation of some
loops, very slight displacements (,1A
˚) of some of the secondary-
structure elements, and differences in side-chain densities. Owing to
the similarity between monomers, our description and comments
on the model apply to both a- and b-tubulin unless indicated
otherwise. Residue numbers correspond to the aligned sequences
of a- and b-tubulin as shown in Fig. 3, and include gaps in the
sequence of b-tubulin.
Figure 4 shows the ribbon diagram of the tubulin dimer model
(see later for ‘dimer’ definition). The core of the structure contains
two b-sheets of 6 and 4 strands, flanked by 12 a-helices. Although
there is no clear division of the density into domains in the map, it
makes sense functionally to divide the structure of each monomer
into three sequential domains.
The N-terminal domain includes residues 1–205 and forms a
Rossmann fold, which is typical of nucleotide-binding proteins, in
which parallel b-strands alternate with a-helices. Helices H1 and
H2 are on one side of the sheet, whereas helices H3, H4 and H5 are
on the other. Strands B2 and B3 are poorly defined in the unrefined
density map. The residues in the loops connecting H1 and B2, and
H2 and B3, are included for completeness, but were built in very
weak density. These segments are long predicted loops on the
putative inside surface of the microtubule, and correspond to a
region of the sequence that can accommodate insertions and
deletions7.
Strand B6 leads to the intermediate domain, residues 206– 381,
containing a mixed b-sheet and five surrounding helices. The
domain starts with helices H6 and H7, followed by a long loop
and helix H8 at the longitudinal interface between monomers. B7 is
a long b-strand that interacts with the b-sheet in the N-terminal
domain. The loop connecting B7 and H9 is more ordered in a-tubulin,
where it is involved in strong lateral contacts. Following a long loop
letters to nature
NATURE
|
VOL 391
|
8 JANUARY 1998 199
Table 1 Electron crystallographic data
.............................................................................................................................................................................
Two-dimensional crystals
Two-sided plane roup P12
1
Unit cell a¼80, b¼92 ˚
A
Sampling thickness c¼90 ˚
A
Sample thickness 62 A
˚
Experimental data set
Resolution cut-off 3.7 A
˚
Number of structure factors 12,000
Electron diffraction
Number of patterns by tilt angle 18 (08), 57 (458),19 (558)
R
sym
19%
R
merge
25%
Image/phase data
Number of images by tilt angle 12 (08), 51 (458), 86 (608)
Phase residual by resolution zones 368(5–4 A
˚), 468(4–3.7 A
˚)
.............................................................................................................................................................................
-500
500
1,500
2,500
3,500
Image phase
Lattice line (8,13)
Res. (
Z
*=0) 5.78 Å
Max. res. 3.55 Å
Diffraction intensity
-180
0
180
-500
1000,
2,500
4,000
5,500
Image phase
Diffraction intensity
-180
0
180
-0.25 -0.2 -0.15
Lattice line (12,12)
-0.1 -0.05
Res. (
Z
*=0) 5.03 Å
Max. res. 3.40 Å
0 0.05 0.1
Z
* (Å
–1
)
0.15 0.2 0.25
Figure 1 Experimental phase and intensity data and fitted curves for two
representative reciprocallattice lines. Error bars for the intensities are Friedel-pair
differences. The resolution of each lattice line at the equator (z¬¼0) and its
furthest point are indicated.
* Present address: Electron Microscopy Unit, Weizmann Institute of Science, Rehovot 76100, Israel.
Nature © Macmillan Publishers Ltd 1998
8
come B8 and H10, which unravels at its C terminus in a-tubulin,
followed by a very well defined short loop into B9. Finally, the loop
between B9 and B10 includes an 8-residue insertion in the a-
subunit which occludes the site that in bis occupied by taxol. The
present model is compatible with the observation that C241 and
C356 in the b-subunit can be crosslinked8(,8 A
˚apart in the model;
Fig. 4b).
The C-terminal domain is formed by helices H11 and H12. These
helices overlay the previous domains, sitting on the surface of the
molecule that we have identified as the outside surface in the
microtubule9(Fig. 4, legend). They are probably involved in the
binding of MAPs and motor proteins. The loop connecting H11 and
H12 is important for the interaction with the next monomer along
the protofilament.
The last C-terminal residues of each monomer, missing from the
model, correspond to the hypervariable part of the sequence where
most of the differences between isotypes and across species occur. As
the tubulin preparation we used included several aand bisotypes
that are found in bovine brain, the inhomogeneity of the sample
could contribute to the poor visibility in this part of the map.
However, we do not see any significant difference in projection
maps among abIII, abII, and undifferentiated tubulin (our unpub-
lished results). This tubulin segment is highly acidic in both
monomers and likely to be disordered, irrespective of the isotype
composition.
The nucleotide in tubulin is positioned at the base of the
Rossmann fold. The B1–H1, B2–H2 and B3 –H3 loops, and the
glycine-rich B4–H4 loop, appear to contact the phosphates. The
letters to nature
200 NATURE
|
VOL 391
|
8 JANUARY 1998
Figure 2 Sections of the experimental density map
with the fitted model for different regions in the a- and
b-tubulin molecules. a,b, Sections from b-tubulin;
c,d, sections from a-tubulin. The map was calcu-
lated with a 3.7-A
˚cutoff. In the direction perpendi-
cular to the crystals, the resolution is less than 3.7A
˚,
owing to the limit of 608in tilt angle, but still sufficient
that individual b-strands can be distinguished. The
initial assignment of the sequence to the backbone
trace was based mostly on the positions of aromatic
side chains and on the connectivity of the density.
Comparing corresponding densities in the a- and b-
subunits, and relating the differences and similarities
to those in their amino-acid sequences, were
extremely helpful in tracing the chain. The sequence
used for the model, in the absence of that from
bovine brain tubulin, corresponds to porcine brain
tubulin28. The numbering of residues is based on the
alignment of the a- and b-tubulin sequences and so
includes gaps in the sequence of b-tubulin (Fig. 3).
10 20 30 40 50 60 70 80 90 100
MRECISIHVGQAGVQIGNACWELYCLEHGIQPDGQMPSDKTIGGGDDSFNTFFSETGAGKHVPRAVFVDLEPTVIDEVRTGTYRQLFHPEQLITGKEDAA
MREIVHIQAGQCGNQIGAKFWEVISDEHGIDPTGSYVGDSDLQL..ERINVYYNEAAGNKYVPRAILVDLEPGTMDSVRSGPFGQIFRPDNFVFGQSGAG
110 120 130 140 150 160 170 180 190 200
NNYARGHYTIGKEIIDLVLDRIRKLADQCTGLQGFSVFHSFGGGTGSGFTSLLMERLSVDYGKKSKLEFSIYPAPQVSTAVVEPYNSILTTHTTLEHSDC
NNWAKGHYTEGAELVDSVLDVVRKESESCDCLQGFQLTHSLGGGTGSGMGTLLISKIREEYPDRIMNTFSVVPSPKVSDTVVEPYNATLSVHQLVENTDE
210 220 230 240 250 260 270 280 290 300
AFMVDNEAIYDICRRNLDIERPTYTNLNRLIGQIVSSITASLRFDGALNVDLTEFQTNLVPYPRAHFPLATYAPVISAEKAYHEQLSVAEITNACFEPAN
TYCIDNEALYDICFRTLKLTTPTYGDLNHLVSATMSGVTTCLRFPGQLNADLRKLAVNMVPFPRGHFFMPGFAPLTSRGSQQYRALTVPELTQQMFDAKN
310 320 330 340 350 360 370 380 390 400
QMVKCDPRHGKYMACCLLYRGDVVPKDVNAAIATIKTKRSIQFVDWCPTGFKVGINYEPPTVVPGGDLAKVQRAVCMLSNTTAIAEAWARLDHKFDLMYA
MMAACDPRHGRYLTVAAVFRGRMSMKEVDEQMLNVQNKNSSYFVEWIPNNVKTAVCDIPP........RGLKMSATFIGNSTAIQELFKRISEQFTAMFR
410 420 430 440 450
KRAFVHWYVGEGMEEGEFSEAREDMAALEKDYEEVGVDSV.E.GEGEEEGEEY..
RKAFLHWYTGEGMDEMEFTEAESNMNDLVSEYQQYQDATADEQGEFEEEGEEDEA
B1 H1 B2 H2 B3
H3 B4 H4 B5 H5
B6 H6 H7 H8 B7 H9
B8 H10 B9 B10 H11
H12
N-terminal domain
Intermediate domain
C-terminal domain
β−ΤΒ
α−ΤΒ
β−ΤΒ
α−ΤΒ
β−ΤΒ
α−ΤΒ
β−ΤΒ
α−ΤΒ
β−ΤΒ
α−ΤΒ
Figure 3 Sequences of pig brain a- and
b-tubulin28 used in the model (in the
absence of tubulin sequences from
cow we have used its closest known
relative). Secondary structure elements
are indicated and labelled as for Fig. 4.
The tubulin preparations used in our
experiments contained a mixture of iso-
types. Most of the differences between
isotypes are located at the extreme C
terminus, which is not visible in our den-
sity. In most of the other positions of
isotype differences, we arbitrarily chose
the residue most similar to the other
monomer.
Nature © Macmillan Publishers Ltd 1998
8
B5– H5 loop is near the ribose, and N206 in H6, and Y224 and N228
in H7 interact with the nucleotide base. The position of the
nucleotide agrees with models, based on comparison with other
nucleotide-containing proteins, that position the phosphates near
the glycine-rich loop. It is consistent with photocrosslinking experi-
ments that locate C12 in b-tubulin near the guanine base10, the
peptide b157–176 near the ribose11, and b65–79 by the g-
phosphate12. The position is also consistent with mutations in the
region b105– 111 affecting the hydrolysis of GTP13. Finally, C12 and
either C203 or C213 can be crosslinked in b-tubulin, but only in the
absence of the nucleotide14. In our model, the nucleotide is between
those residues, but the distances between the cysteines are larger
than 9 A
˚. However, the position of H6, containing C213, could
easily be affected by the absence of nucleotide, bringing that cysteine
closer to C12.
The nucleotide inone monomer interacts with the next monomer at
the longitudinal interface. The interfaces between consecutive
monomers in our density map are too similar to allow us to
letters to nature
NATURE
|
VOL 391
|
8 JANUARY 1998 201
B1
B2 B3
B4
B5
B6
B7
B8
B9
B10
GTP
TAX
H3
H4
H5
H6
H8
H9
H10
H11
H12
H7
H1
H2
GDP
α-TB
β-TB
+
–
a
Figure 4 Ribbon diagram of the tubulin dimer showing a-tubulin with bound GTP
(top), and b-tubulin containing GDP and taxotere (bottom). Labels for strands (in
the a-subunit) and helices (in the b-subunit) are included. The arrow indicates the
direction of the protofilament and microtubule axis. a, Stereo front view from the
putative outside of the microtubule; b, back view from the putative inside of the
microtubule; c, side view. Figures produced with AVS (Advanced Visual; ribbon
module from M. Carson and A. Shah). The in–out orientation was determined by
reference to reconstructions of intact microtubules9. Such reconstructions show
prominent longitudinal ridge on the outside, which in our model would be formed
by H11, H12 and the loop between H10 and B9, and shallow inside grooves giving
the protofilament a bumpy appearance, corresponding in our model to H1, B3 and
the long loops in the N-terminal domain. This represents the most likely arrange-
ment of the dimer, because it buries the nucleotide that is at the non-exchange-
able site in a(see text). For the nucleotide in bto be exchangeable at the plus end
of a microtubule, the bottom of the figure would correspond to the plus end. We
previously presumed the opposite orientation, based on a comparison of the zinc
sheets in negatively stained, stain–glucose, and tannin –glucose embedding, with
projection maps of open microtubules of known polarity in negative stain9. Some
ambiguity in that determination may be introduced by uncertainty about the exact
rotational alignment of the protofilament in the sheets with respect to those in
open microtubules and by stain artefacts. The polarity with the plus end down
would be consistent with experiments that located the b-subunit at the plus end of
the microtubule29 and the a-subunit at the minus end30. Circles in bindicate the
positions of bCys 241 and bCys356, separated by about 8 A
˚.
B1
B2
B3
B4
B5 B6
B7
B8
B9
B10
H10
H9
H8
H6
H7
H3 H4
H2
H1
H5 H11
H12
GTP
TAX
GDP
H12
H11
H5
H2
H9
H6
H7
H10
TAX
GDP
H8
B1
B2
B3
B4 B5B6
B7
B8
B9
B10
GTP
Inside
Outside
b c
•• C356
•• C241
Nature © Macmillan Publishers Ltd 1998
8
distinguish unequivocally between inter- and intradimer contacts
(the interface near the N site is only slightly tighter than that by the E
site). On the basis of the difference of exchangeability of nucleotide
in the aand bmonomers, we identified the intradimer interface as
that in which the N site in the a-subunit is buried. The GTP at the E
site in the b-subunit is partially exposed in the dimer but would be
buried in the microtubule, where it becomes non-exchangeable. The
assignment of the dimer shown in Fig. 4 thus provides the simplest
explanation for nucleotide exchangeability in tubulin. Further
support comes from data on the binding of colchicine. The main
binding site of this drug is in the b-subunit, residues C356 and C241
(ref. 15) and region 1–36 (ref. 16) having been identified as part of
the binding site (see the position of these residues in Fig. 4b). In
addition, colchicine binds close to the ab intradimer interface17. On
the other hand, antibodies against b241–256, near the top of the b-
monomer (Fig. 4b), and to a214–226, at the bottom of the a-
monomer, bind to the dimer but are unable to bind to the
microtubule18. An alternative dimer to that shown in Fig. 4 would
more readily explain these results, although the present model
would be compatible if changes in the dimer occurred on binding
to the microtubule.
On the basis of the similar behaviour of our sheets and micro-
tubules, we believe that hydrolysis at the E site occurs upon sheet
polymerization. In agreement with this, we find extra density in a-
tubulin that corresponds to the g-phosphate. The similarity
between the two monomers, in spite of their being in two different
nucleotide-bound states, is not surprising. The current model is that
the GDP-containing dimer buried in the microtubule is kept locked
in a GTP-like conformation by the microtubule lattice, and that
only upon depolymerization does the GDP dimer ‘spring’ into a
different conformation19. Comparison by electron microscopy of
microtubules formed in the presence of GTP or of non-hydrolysable
GTP analogues have shown only a change of 1.5 A
˚in the axial repeat
(less than 2% of the subunit length)20,21.
The favoured polymers for tubulin-GDP are rings formed by
curved protofilaments22. The conformation of tubulin-GDP is thus
referred to as ‘curved’, as opposed to the ‘straight’ conformation of
tubulin-GTP (or tubulin-GDP held locked in the microtubule
lattice)23. In two simplified models, the curved conformation
could either be one in which longitudinal contacts between
dimers are at an angle (consistent with the dimer definition in
Fig. 4), or one in which the dimer is bent at the monomer–
monomer interface (alternative dimer definition). A more general
model would involve extensive allosteric effects following hydro-
lysis, which could affect points distant from the nucleotide-binding
pocket.
Our study was done on tubulin sheets that were stabilized by
taxol. The model in Fig. 4 includes a molecule of taxotere, a Taxol
analogue whose structure has been solved by X-ray crystallography24
(in taxotere, the C-10 acetyl and the C-39benzamide groups of Taxol
have been replaced respectively by a hydroxyl and a N-t-BOC
group). In our map, there is clear density for the taxane ring, the
most dense and least flexible part of Taxol. Some density is also clear
for one of the side chains. The position in the model is based on the
best visual fit to the observed density. This position is in good
agreement with photocrosslinking results placing the C-39group
near the sequence b: 1–31 (ref. 25), and the C2 group near the
sequence b: 217–231 (ref. 26). In our model, the group at C-39is
near the top of helix H1 (that is, between b: 15–25), and the C2
group near H5 and the H5–H6 loop (that is, between b: 212–222).
The main interaction of the taxane ring with tubulin is at L275, at
the beginning of the B8–H9 loop.
Our model of tubulin shows a compact molecular structure with
three functional domains: namely, GTP-binding, drug-binding and
motor/MAP-binding domains. The interaction between domains is
very tight, so the effects that nucleotides, drugs and other proteins
in the cell have on tubulin are firmly linked. The assemblyof tubulin
and its regulation through dynamic instability results from the fine
tuning of the three components. Knowledge of the structure of
tubulin should be invaluable for understanding the microtubule
system in the cell.
The structure of the bacterial FtsZ protein is reported in this
issue27. Comparison of the tubulin and FtsZ models indicates that
they have a common structural core of identical fold, which includes
10 b-strands surrounded by 10 a-helices: more detail will be
revealed by careful comparison of there two structures. M
.........................................................................................................................
Methods
Crystalline tubulin sheets were polymerized in the presence of zinc from bovine
brain tubulin (Cytoskeleton Inc.) and stabilized with taxol as described4.
Samples were prepared by tannin–glucose embedding and examined at
liquid-nitrogen temperature in a JEOL 4000 electron microscope at 400 kV
following previously described procedures4,5. Images were taken using spot-
scan imaging and dynamic focus correction. Images and electron diffraction
patterns were processed and merged as described4,9.
Received 8 August; accepted 27 October 1997.
1. Ludven
˜a, R. F.The multiple forms of tubulin: different gene products and covalent modifications. Int.
Rev. Cyt. 178, 207–275 (1998).
2. Mukherjee, A. & Lutkenhaus, J. Guanine nucleotide-dependent assembly of FtsZ into filaments.
J. Bacteriol. 176, 2754–2758 (1994).
3. Gabor Miklos, G. L., Yamamoto, M., Burns, R. G. & Maleszka, R. An essential cell division gene of
Drosophila, absent from Saccharomyces, encodes an unusual protein with tubulin-like and myosin-like
peptide motifs. Proc. Natl Acad. Sci. USA 94, 5189– 5194 (1997).
4. Nogales, E., Wolf, S.G., Zhang, S. X. & Downing, K. H. Preservation of 2-D crystals of tubulin for
electron crystallography. J. Struct. Biol. 115, 199–208 (1995).
5. Nogales, E., Wolf, S. G., Khan, I. A., Luduen
˜a, R. F.& Downing, K. H. Structure of tubulin at 6.5 A
˚and
location of the taxol-binding site. Nature 375, 424–427 (1995).
6. Nogales, E., Wolf, S. G. & Downing, K. H. Visualizing the secondary structure of tubulin: three-
dimensional map at 4 A
˚.J. Struct. Biol. 118, 119– 127 (1997).
7. Burns, R. G. & Surridge, C. D. in Microtubules (eds Hyams, J. S. & Lloyd, C. W.) 3–32 (Wiley, New
York, 1993).
8. Little, M. & Luduen
˜a, R. F. Structural differences between brain b1- and b2-tubulins: implications for
microtubule assembly and colchicine binding. EMBO J. 4, 51– 56 (1985).
9. Wolf, S. G., Nogales, E., Kikkawa, M., Gratzinger, D., Hirokawa, N. & Downing, K. H. Interpreting a
medium-resolution model of tubulin: comparison of zinc-sheet and microtubule structure. J. Mol.
Biol. 263, 485–501 (1996).
10. Shivanna, B. D., Mejillano, M. R., Williams, T.D. & Himes, R. H. Exchangeable GTP binding site of b-
tubulin—identification of cysteine 12 as the major site of cross-linking by direct photoaffinity
labeling. J. Biol. Chem. 268, 127–132 (1993).
11. Hesse, J., Thierauf, M. & Ponstingl, H. Tubulin sequence region b155– 174 is involved in binding
exchangeable guanosine triphosphate. J. Biol. Chem. 262, 15472– 15475 (1987).
12. Linse, K. & Mandelkow, E.-M. The GTP-binding peptide of b-tubulin. Localization by direct
photoaffinity labeling and comparison with nucleotide-binding proteins. J. Biol. Chem. 263,
15205–15210 (1988).
13. Davis, A., Sage, C. R., Dougherty, C. A. & Farrell, K. W. Microtubule dynamics modulated by
guanosine triphosphate hydrolysis activity of b-tubulin. Science 264, 839– 842 (1994).
14. Little, M. & Luduen
˜a, R. F.Location of two cysteines in brain b
1
-tubulin that can be cross-linked after
removal of exchangeable GTP. Biochim. Biophys. Acta 912, 28–33 (1987).
15. Bai, R. et al. Identification of cysteine 354 of b-tubulin as part of the binding site for the A ring of
colchicine. J. Biol. Chem. 271, 12639–12645 (1996).
16. Uppuluri, S., Knipling, L., Sackett, D. L. & Wolff, J. Localization of the colchicine-binding site of
tubulin. Proc. Natl Acad. Sci. USA 90, 11598– 11602 (1993).
17. Shearwin, K. E. & Timasheff, S. N. Effect of colchicine analogs on the dissociation of ab tubulin into
subunits: the locus of colchicine binding. Biochemistry 33, 894 –901 (1994).
18. Andreu, J. M. Site-directed antibodies to tubulin. Cell Motil. Cytoskel. 26, 1– 6 (1993).
19. Caplow, M., Ruhlen, R. L. & Shanks, J. The free energy of hydrolysis of a microtubule-bound
nucleoside triphosphate is near zero: all of the free energy for hydrolysis is stored in the microtubule
lattice. J. Cell Biol. 127, 779–788 (1994).
20. Vale, R. D., Coppin, C. M., Malik, F., Kull, F.J. & Milligan, R. A. Tubulin GTP hydrolysis influencesthe
structure, mechanical properties, and kinesin-driven transport of microtubules. J. Biol. Chem. 269,
23769–23775 (1994).
21. Hyman, A. A., Chre
´tien, D., Arnal, I. & Wade,R. H. Structural changes accompanying GTP hydrolysis
of microtubules: information from a slowly hydrolyzable analog guanylyl-(a,b)-methylene-
diphosphonate. J. Cell Biol. 128, 117– 125 (1995).
22. Dı
´az, J. F., Pantos, E., Bordas, J. & Andreu, J. M. Solution structure of GDP-tubulin double rings to
3 nm resolution and comparison with microtubules. J. Mol. Biol. 238, 214 –225 (1994).
23. Mandelkow, E. & Mandelkow, E.-M. Microtubules and microtubule-associated proteins. Curr. Opin.
Cell Biol. 7, 72– 81 (1995).
24. Gueritte-Voegelein, F. et al. Structure of a synthetic taxol precursor: N-tert-butoxycarbonyl-10-
deacetyl-N-debenzoyltaxol. Acta Crystallogr. C 46, 781–784 (1990).
25. Rao, S., Krauss, N. E., Heerding, J. M., Orr, G. A. & Horwitz, S. B. 39-( p-Azidobenzamido)taxol
photolabels the N-terminal 31 amino acids of b-tubulin. J. Biol. Chem. 269, 3132–3134 (1994).
26. Rao, S., Orr, G. A., Chaudhary, A. G., Kingston, D. G. I. & Horwitz, S. B. Characterization of the taxol
binding site on the microtubule. J. Biol. Chem. 270, 20235–20238 (1995).
27. Lo
¨we, J. Y. & Amos, L. A. Crystal structure of the bacterial cell-division protein FtsZ complexed with
GDP. Nature 391, 203 –206 (1998).
28. Ponstingl, H., Krauhs, E., Little, M., Kempf, T., Hofer-Warbinek, R. & Ade, W. Amino acid sequence of
a- and b-tubulins from pig brain: heterogeneity and regional similarity to muscle proteins. Cold
Spring Harbor Symp. Quant. Biol. 46, 191–197 (1982).
29. Mitchison, T. J. Localization of an exchangeable GTP binding site at the plus end of microtubules.
Science 261, 1044–1047 (1993).
letters to nature
202 NATURE
|
VOL 391
|
8 JANUARY 1998
Nature © Macmillan Publishers Ltd 1998
8
30. Fan, J., Griffiths, A. D., Lockhart, A., Cross, R. A. & Amos, L. A. Microtubule minus ends can be
labeled with a phage display antibody specific to a-tubulin. J. Mol. Biol. 259, 325– 330 (1996).
Acknowledgements. Wethank R. F. Luduen
˜a for isotypically purified abII and abIII tubulin, M. Le for
help with electron diffraction processing, and R. M. Glaeser and Y. L. Han for comments on the
manuscript. Taxol was provided by the Drug Synthesis and Chemistry Branch, Division of Cancer
Treatment of the National Cancer Institute. This work was supported by the NIH.
Correspondence and requests for materials should be addressed to E.N. Coordinates referred to in this
Letter have been deposited in the Brookhaven Protein Data Bank with ID 1tub and will be accessible
within one year.
letters to nature
NATURE
|
VOL 391
|
8 JANUARY 1998 203
Crystal structure of the
bacterial cell-division
protein FtsZ
Jan Lo
¨we & Linda A. Amos
MRC Laboratory of Molecular Biology, Cambridge CB2 2QH, UK
.........................................................................................................................
Bacterial cell division ends with septation, the constriction of the
cell wall and cell membranes that leads to the formation of two
daughter cells1,2. During septation, FtsZ, a protein of relative
molecular mass 40,000 which is ubiquitous in eubacteria and is
also found in archaea and chloroplasts3, localizes early at the
division site to form a ring-shaped septum. This septum is
required for the mechanochemical process of membrane
constriction4. FtsZ is a GTPase5,6 with weak sequence homology
to tubulins7. The nature of FtsZ polymers in vivo is unknown, but
FtsZ can form tubules, sheets and minirings in vitro8,9. Here we
report the crystal structure at 2.8 A
˚resolution of recombinant
FtsZ from the hyperthermophilic methanogen Methanococcus
jannaschii. FtsZ has two domains, one of which is a GTPase
domain with a fold related to one found in the proteins p21
ras
and elongation factor EF-Tu. The carboxy-terminal domain,
whose function is unknown, is a four-stranded b-sheet tilted by
908against the b-sheet of the GTPase domain. The two domains
are arranged around a central helix. GDP binding is different from
that typically found in GTPases and involves four phosphate-
binding loops and a sugar-binding loop in the first domain, with
guanine being recognized by residues in the central connecting
helix. The three-dimensional structure of FtsZ is similar to the
structure of a- and b-tubulin10.
Two FtsZ genes (named after filamenting temperature-sensitive
mutant Z) from the archaeon M. jannaschii have been characterized
by the genome project11. One gene, MJ0370, was amplified by
genomic polymerase chain reaction (PCR) and expressed in
E. coli/C41, a mutant of BL21 capable of expressing toxic genes12.
Proteolysis during cell disruption was minimized by using heat-
shock treatment. Cubic crystals were obtained and the structure was
solved by multiple isomorphous replacement and density modifica-
tion (see Methods and Table 1). The model (Fig. 1) contains
residues 23–356, 116 water molecules, and one molecule of GDP;
weak density for residues 1– 22 was visible as an extension from
helix H0.
FtsZ consists of two domains with a long, 23-residue, helix H5
(Figs 1a, 2) connecting them. The N-terminal portion of the
molecule, containing residues 38–227, has GDP obtained from
the expression host bound to it and will be called the GTPase
domain. It consists of a six-stranded parallel b-sheet surrounded by
two and three helices on both sides. The overall fold of the GTPase
domain of FtsZ is related to typical GTPases and can be super-
imposed on the p21
ras
–GDP complex (Protein Data Bank (PDB)
entry 1Q21; ref. 13) using 52 Caatoms (S1, H1, S2, H2, S4, H3 and
S5) to give a root-mean-squared (r.m.s.) deviation of 1.88A
˚. The
topology of the b-sheet in FtsZ is 321456, which is slightly different
from the topology in p21
ras
(ref. 13), where it is 231456, but,
together with the arrangement of five helices (H1, HL1, H2, H3
and H4), is consistent with typical Rossmann-fold topology14. Helix
H2A is unique to FtsZ. Numbering of secondary structure elements
(Fig. 2) follows the corresponding elements of p21
ras
proteins.
The C-terminal domain, spanning residues 228 –356, consists of a
mainly parallel four-stranded central b-sheet supported by two
helices on one side. The topology of the sheet is 1423, with strand 4
antiparallel to the others. The uncovered side of the sheet makes
contacts with helix H5 and is otherwise open to the solvent. The
fold of the C-terminal domain is related to chorismate mutase of
Bacillus subtilis and can be superimposed on PDB entry 1COM15
with an r.m.s. deviation of 1.83 A
˚over 52 Caatoms (SC1, HC2, SC2,
HC3, SC3 and SC4). Additionally, sequence comparisons give
similarities to calmodulins in three loop regions (Swissprot CALM-
TRYCR; loops between H5/HC1, SC1/HC2, and SC2/HC3) and to
adenylyl cyclase (CYA1_HUMAN; residues 620– 740), making a role
in calcium binding feasible. The electrostatic potential on the
Figure 1 Ribbon drawings of FtsZ (residues 23–356) from M. jannaschii.a, View
showing the GTPase domain inblue/green, the C-terminal domain in red/orange,
and the connecting helix H5 in yellow. GDP is represented by a space-filling
model. b, View of FtsZ rotated by ,908from that in a. GDP is represented by a ball-
and-stick model. Figures were prepared with POVSCRIPT (D. Peisach, personal
communication)28.
Nature © Macmillan Publishers Ltd 1998
8
letters to nature
NATURE
|
VOL 392
|
5 MARCH 1998 75
bootstrap replicates. ML analyses of these two trees showed that they are not
significantly different.
Effects of long-branch taxa. To identify taxa with long apparent branch
lengths, we performed a four-taxon NJ analysis (using gamma-corrected
Kimura distances) using Tripedalia (a diploblast), Antedon (a deuterostome)
and Glycera (a protostome) with each nematode taxon in turn. We recorded the
inferred distance from the protostome–nematode node to the nematode taxon.
MP distances were derived from the phylogeny presented in Fig. 1. These long-
branch-length taxa often have extreme base-composition biases, but not all
taxa with extreme base compositions have long branch lengths (for example,
Brugia has an AT content of 79% but one of the shorter inferred branch
lengths). NJ and MP analyses were re-performed with successive trimming of
the long-branch taxa (distance .0.19 from root in four-taxon NJ analysis)
from the dataset. As would be expected29, exclusion of these taxa had effects on
the bootstrap support for some clades. In particular, re-analysis excluding one
or all of Panagrellus,Panagrolaimus, and Strongyloides yielded stronger boot-
strap support for the cephalobid– steinernematid clade IV (,50% to 68%), and
analyses excluding the long-branch rhabditid taxa Bunonema,Teratorhabditis
and Pellioditis gave increased support for the Diplogasterida–Rhabditina clade
V (51% to 93%). Figure 2a shows a consensus of these analyses: branchpoints
that were supported by .60% in bootstrap long-branch taxa resampling of NJ
or MP trees from the trimmed datasets were accepted. When there was no
support for a resolved branching order, we collapsed nodes to form polytomies.
Major clades supported by all analytical methods are shown and are numbered
I– V. The association of Secernentea (clades II, IVand V) with the Plectidae has
not been unequivocally resolved and is shown as a polytomy. We could not
place the long-branch-length taxa in our trees with any certainty. We assessed
statistical support for the placement of the vertebrate-parasitic taxa into four
clades by calculating ML values for six-taxon subsets from the data. Each of the
placements was strongly supported.
Received 16 July; accepted 12 December 1997.
1. Luc, M., Sikora, R. A. & Bridge, J. Plant Parasitic Nematodes in Tropical and Subtropical Agriculture
(CAB International, Wallingford, UK, 1990).
2. Anderson, R. C. Nematode Parasites of Vertebrates. Their Development and Transmission (CAB
International, Wallingford, UK, 1992).
3. Lambshead, J. Recent developments in marine benthic biodiversity research. Oceanis 19, 5 –24 (1993).
4. Riddle, D., Blumenthal, T., Meyer, B. & Priess, J. (eds) C. elegans II (Cold Spring Harbor Laboratory
Press, NY, 1997).
5. Ellis, R. E., Sulston, J. E. & Coulson, A. R. The rDNA of C. elegans: sequence and structure. Nucleic
Acids Res. 14, 2345–2364 (1986).
6. Zarlenga, D. S., Stringfellow, F., Nobary, M. & Lichtenfels, J. R. Cloning and characterisation of
ribosomal RNA genes from three species of Haemonchus (Nematoda: Trichostrongyloidea) and
identification of PCR primers for rapid differentiation. Exp. Parasitol. 78, 28–36 (1994).
7. Fitch, D. H. A., Bugaj-gaweda, B. & Emmons, S. W. 18S ribosomal gene phylogeny for some
rhabditidae related to Caenorhabditis elegans.Mol. Biol. Evol. 12, 346 –358 (1995).
8. Baldwin, J. G., Frisse, L. M., Vida, J. T.,Eddleman, C. D. & Thomas, W.K. An evolutionary framework
for the study of developmental evolution in a set of nematodes related to Caenorhabditis elegans.Mol.
Phylogenet. Evol. 8, 249–259 (1997).
9. Baldwin, J. G. et al. The buccal capsule of Aduncospiculum halicti (Nemata: Diplogasterina): an
ultrastructural and molecular phylogenetic study. Can. J. Zool. 75, 407– 423 (1997).
10. Swofford, D. L., Olsen, G. J., Waddell,P. J. & Hillis, D. M. in Molecular Systematics (eds Hillis, D. M.,
Moritz, C. & Mable, B. K.) 407–514 (Sinauer, Sunderland, MA, 1996)
11. Aguinaldo, A. M. A. et al. Evidence for a clade of nematodes, arthropods and other moulting animals.
Nature 387, 489–493 (1997).
12. Lorenzen, S. The Phylogenetic Systematics of Free-Living Nematodes (The Ray Society, London, 1994).
13. Malakhov, V. V. Nematodes. Structure, Development, Classification and Phylogeny (Smithsonian
Institution Press, Washington, 1994).
14. Maggenti, A. R. in Concepts in Nematode Systematics (eds Stone, A. R., Platt, H. M. & Khalil, L. F.) 25–
40 (Academic, London, 1983).
15. Baldwin, J. G. & Eddleman, C. D. Buccal capsule of Zeldia punctata (Nemata: Cephalobidae): an
ultrastructural study. Can. J. Zool. 73, 648–656 (1995).
16. Etzinger, A. & Sommer, R. The homeotic gene lin-39 and the evolution of nematode epidermal cell
fates. Science 278, 452– 455 (1997).
17. Poinar, G. Origins and phylogenetic relationships of the entomophilic rhabditids, Heterorhabditis and
Steinernema.Fund. Appl. Nematol. 16, 332–338 (1993).
18. Siddiqi, M. R. Phylogenetic relationships of the soil orders Dorylaimida, Mononchida, Triplonchida
and Alaimida, with a revised classification of the subclass Enoplia. Pak. J. Nematol. 1, 79–110 (1983).
19. Poinar, G. O. The Natural History of Nematodes (Prentice-Hall, Englewood Cliffs, NJ, 1983).
20. De Ley, P., van de Velde, M. C., Mounport, D., Baujard, P. & Coomans, A. Ultrastructure of the stoma
in Cephalobidae, Panagrolaimidae and Rhabditidae, with a proposal for a revised stoma terminology
in Rhabditida. Nematologica 41, 153–182 (1995).
21. Winnepenninckx, B. et al. 18S rRNA data indicate that Aschelminthes are polyphyletic in origin and
consist of at least three distinct clades. Mol. Biol. Evol. 12, 1132–1137 (1995).
22. Blaxter, M. L. et al. Genes expressed in Brugia malayi infective third stage larvae. Mol. Biochem.
Parasitol. 77, 77–96 (1996).
23. Swofford, D. L. PAUP: Phylogenetic Analysis Using Parsimony, Version 3.1 (Illinois Natural History
Society, Champaign, 1993).
24. Maddison, W. & Maddison, D. MacClade v3.0 (Sinauer, Sunderland, MA, 1993).
25. Kumar, S., Tamura, K. & Nei, M. MEGA: Molecular Evolutionary Genetics Analysis. Version 1.0
(Pennsylvania State Univ., 1993).
26. Van de Peer, Y., Rensing, S., Maire, U.-G. & De Wachter, R. Substitution rate calibration of small
subunit subunit RNA identifies chlorarachniophyte nucleomorphs as remnants of green algae. Proc.
Natl Acad. Sci. USA 93, 7732–7736 (1996).
27. Van de Peer,Y. & De Wachter, R. TREECON for Windows: a softwarepackage for the construction and
drawing of evolutionary trees for the Microsoft Windows environment. Comput. Appl. Biosci. 10,
569–570 (1994).
28. Yang, Z. Phylogenetic Analysis by Maximum Likelihood (PAML)Version 1.2 (Univ. California, Berkeley,
1996).
29. Felsenstein, J. Cases in which parsimony and compatibility methods will be positively misleading. Syst.
Zool. 27, 401–410 (1978).
Acknowledgements. Wethank our colleages for donations of nematode material, and D. Swofford for use
of prerelease versions of PAUP*4.0. This work was supported by grants from the Wellcome Trust, the
Linnean Society of London, the Belgian National Fund for Scientific Research,the NSF, the NIH and the
United States Department of Agriculture.
Correspondence and requests formaterials should be addressed to M.L.B. (e-mail: mark.blaxter@ed.ac.uk).
Spatial and temporal
organization during
cardiac fibrillation
Richard A. Gray*†, Arkady M. Pertsov*& Jose
´Jalife*
*Department of Pharmacology, SUNY Health Science Center, 766 Irving Avenue,
Syracuse, New York 13210, USA
†Department of Biomedical Engineering and Department of Medicine Division of
Cardiovascular Disease, University of Albama at Birmingham, 1670 University
Blvd, Birmingham, Alabama 35294-0019, USA
.........................................................................................................................
Cardiac fibrillation (spontaneous, asynchronous contractions of
cardiac muscle fibres) is the leading cause of death in the
industrialized world1,yetitisnotclearhowitoccurs.Ithasbeen
debated whether or not fibrillation is a random phenomenon.
There is some determinism during fibrillation2,3, perhaps result-
ing from rotating waves of electrical activity4–6. Here we present a
new algorithm that markedly reduces the amount of data required
to depict the complex spatiotemporal patterns of fibrillation. We
use a potentiometric dye7and video imaging8,9 to record the
dynamics of transmembrane potentials at many sites during
θ
01234
0
50
250
200
Time (s)
0 50 100 150 200
F
(
t
)
250 012 3 4
Frequency (Hz)
ab
cd
150
100
0
50
250
200
150
100
40
30
20
10
0
Separatrix
Power
F
(
t
+t)
Figure 1 Temporal organization. a, Phase portrait of an excitable element
incorporating two state variables30. A stable fixed point occurs at the intersection
of the nullclines (dotted lines)30.b, Fluorescence signal (F) from a site on the
surface of a rabbit heart during fibrillation. c, Phase portrait reveals trajectories
circling around a centre ( F
mean
,F
mean
), shown as a circle. d, The fluorescence
signal exhibited a periodic component centred near 8 Hz, as observed in the
corresponding power spectra. The frequency band 8 63 Hz was different from
equivalent white noise; P,0:00001 for each heart (all sites combined).
Nature © Macmillan Publishers Ltd 1998
8
letters to nature
76 NATURE
|
VOL 392
|
5 MARCH 1998
fibrillation. Transmembrane signals at each site exhibit a strong
periodic component centred near 8 Hz. This periodicity is seen as
an attractor in two-dimensional-phase space and each site can be
represented by its phase around the attractor. Spatial phase maps
at each instant reveal the ‘sources’ of fibrillation in the form of
topological defects, or phase singularities10, at a few sites. Using
our method of identifying phase singularities, we can elucidate
the mechanisms for the formation and termination of these
singularities, and represent an episode of fibrillation by locating
singularities. Our results indicate an unprecedented amount of
temporal and spatial organization during cardiac fibrillation.
It is still uncertain whether rotors underlie cardiac fibrillation.
Self-organized rotors giving rise to spiral waves have been observed
in various excitable media11– 13 including cardiac muscle8. Although
stationary spiral waves occur in isolated thin pieces of cardiac tissue,
in the whole heart, as in many excitable media, they tend to move
throughout the heart. If these spiral waves move rapidly (at .30%
of the wave speed), they give rise to fibrillatory activity4. The
mechanisms of cardiac fibrillation vary4,15, however, and fibrillation
is usually the result of multiple three-dimensional electrical waves,
sometimes described as meandering wavelets, propagating through-
out the heart16,17. Cardiac fibrillation has been described in terms of
Figure 2 Snapshots of phase from the heart surface of
the rabbit and sheep during sustained fibrillation.
a, Rabbit; b, sheep. We classify rotor chirality as ‘+’
for clockwise and ‘–’ for anticlockwise25. At these
instants, three phase singularities (two clockwise and
one anticlockwise) were observed on the rabbit heart
and nine (five clockwise and four anticlockwise) on the
larger sheep heart. Signals ( F) demonstrate (c) low
amplitude and (d) remain near the centre of their phase
portraits when a spatial phase singularitysite is nearby.
Dashed line and red circle indicate the time of the
corresponding snapshot. Vertical white line represents
1 cm.
Figure 3 Initiation of a pair of spatial phase
singularities. Snapshots of phase before (a),
during (b), and after (c) the formation of a pair of
spatial phase singularities during sustained
fibrillation in the sheep heart. d,e, Transmem-
brane signals (F) measured at sites a– e and 1– 5
labelled in b.f, A pair of singularities form when
the local phase gradient becomes large (in
other words, the excitation wave, v<0 (green),
approaches regions not fully recovered,v<6p
(red)). The excitation wave cannot proceed into
the recovered region, and hence breaks, form-
ing two phase singularities. The two excitation
waves rotate around these newly formed sin-
gularities. Sustained rotation in the form of a pair
of rotors occurs only if this excitation wave
causes type 0, or even, phase resetting at the
site of the initial wave break. Type 0 resetting
(suprathreshold) advances the phase of this
region into a new cycle, generating a new exci-
tation wave (in the opposite direction to the
previous wave; see arrow in c), resulting in the
formation of a pair of self-sustaining rotors.Type
1 resetting (subthreshold) does not create this
new excitation wave, and the phase singularity
pair lasts less than one rotation. Notice the
‘extra’ cycle in the central region of block, sites
3 & c in dand e, indicative of type 0 resetting. g,
Trajectories of ‘+’ and ‘ –’ rotors following their
initiation plotted in x,y,tspace. Vertical line
between aand brepresents 1 cm.
Nature © Macmillan Publishers Ltd 1998
8
letters to nature
NATURE
|
VOL 392
|
5 MARCH 1998 77
rotors4–6,14,15, on the basis of the long-held view that the heart is an
example of a generic excitable medium10,18–20. There are many
theories about fibrillation in excitable media, but only recently
have experimental techniques become available to study the com-
plex spatial patterns observed during sustained fibrillation4,6,21–23.
Results from these experiments indicate that rotating waves are
observed during fibrillation; however, they appear infrequently, and
their initiation, termination and interaction have not been char-
acterized.
An excitable element (for example, a cell, a patch of membrane, or
a localized region in a spatially distributed system) can often be
represented in phase space, which the element spends most of its
time at a fixed point. A suprathreshold stimulus pushes the state of
the excitable element past the separatrix and it continues along a
closed-loop trajectory; however, if the stimulus is subthreshold, the
state of the element does not cross the separatrix (Fig. 1). In periodic
dynamics, it is simple and useful to represent the state of an element
by its phase (v) around the loop. The responses of single elements to
external stimuli have been extensively studied by analysing the
induced changes in v(that is, phase resetting). Two fundamentally
different responses to stimuli occur, namely, type 0 or even resetting,
where a suprathreshold response gives rise to a new cycle, and type 1
or odd resetting, which effects a subthreshold response to
stimuli10,24. In spatially extended excitable systems, the stimulation
of individual elements are provided by neighbouring elements
(usually through diffusion).
A rotor is composed of a wave of excitation propagating around a
topological defect, which is known as a phase singularity. A spatial
phase singularity is a site in an excitable medium at which the phase
of the site is arbitrary; the neighbouring elements exhibit a con-
tinuous progression of phase that is equal to 62paround this site.
As shown in Fig. 1, transmembrane signals (F) recorded from the
surface of rabbit and sheep hearts during fibrillation exhibited
attractors in reconstructed two-dimensional phase space, when
F(t) was plotted against Fðtþt) where tis time and tis the
embedding delay (see Methods). Periodicity of each site was near
8 Hz. Although trajectories for subsequent cycles did not coincide in
phase space, the trajectories circulated around a central region,
allowing us to construct a new variable, the phase along the
attractor, v. The phase variable (v) calculated at each site rotated
clockwise (dv,0) 89 64% of the time, indicating that the trajec-
tories were encircling the centre. The distance of points from the
centre was smaller for dv$0 compared with dv,0 (P,0:0001),
as would be expected if the phase were ambiguous at the centre10.
This new variable, v, has certain advantages over the fluorescence
signal (F) that simplify the analysis of fibrillation. First, use of v
eliminates the need to pick activation times, which is difficult
during fibrillation, especially in the important regions of slow
propagation and block. Second, we can test directly whether spatial
phase singularities exist and are necessary to maintain fibrillation.
With this new method to represent fibrillation by phase, vðx;y;tÞ,
we could study directly the detailed dynamics of spatial phase
singularities and rotors, including their initiation and termination.
The spatial phase patterns during sustained fibrillation (Fig. 2)
concurred with theoretical predictions (for example, isophase lines
connect phase singularities of opposite chirality or end on a no-flux
boundary)10,25. Spatial phase singularities are easily identified as sites
at which all phase values (−pto p) converge. The continuous spatial
phase changes reflect waves propagating on the heart surface as a
result of processes of excitation, recovery, and diffusion, and
indicate that each site of the heart surface can be represented by
its phase around a two-dimensional attractor. We elucidated the
mechanism of rotor formation and termination by analysing
successive frames of vðx;y;tÞ, (representative examples are shown
in Figs 3 and 4). Movement of existing phase singularities created
Doppler-shifted4,8,26 short cycle lengths, and thus created large local
phase gradients in front of moving singularities. The formation of
phase singularities was necessary, although not sufficient, for
sustained rotation (that is, for rotor formation). In addition to
the formation of phase singularities, a new excitation wave must be
generated27 (that is, type 0 or even resetting must occur)10,24 to form
a rotor. Lifespan histograms for both rabbits and sheep indicate that
the majority (80% for rabbit and 84% for sheep) of phase singula-
rities lasted ,100 ms, which is less than one rotation4,14. Therefore,
Figure 4 Dynamics of phase singularities. a–e, Snapshots of phase illustrating
rotor dynamics on the surface of the rabbit heart during sustained fibrillation. f,
Trajectoriesof the clockwise and anticlockwise rotors shown in a– e plotted in x, y,
t space. Each rotor is numbered (1–4) and coloured (clockwise, blue; anti-
clockwise, red). The xand yprojections in time are shown in black in f(except for
rotors 1 and 2 for clarity). a, Rotors 1 and 2 formed separately before the time
interval shown here (0.75 –1.17 s). b, At t ¼0:842 s a clockwise rotor (3) enters the
field of view from the left and moves rapidlytoward the right (grey arrow in band f).
This movement creates a convergenceof phase values ahead of rotor 3, and (c) a
pair of phase singularities (4) form (#in cand f) when the excitation wave (v<0)
reaches the high-phase-gradient region. Both of these newly formed singularities
move; the clockwise one collides with anticlockwise-rotating rotor 3, resulting in
mutual annihilation (asterisk in d), whereas the anticlockwise one survives in the
form of a rotor. Vertical white line represents 1 cm.
Nature © Macmillan Publishers Ltd 1998
8
letters to nature
78 NATURE
|
VOL 392
|
5 MARCH 1998
only ,20% of phase singularities formed rotors.
Rotor termination occurred when rotors of opposite chirality
merged or when a single rotor collided with a boundary (both of
these occurrences are topologically equivalent for a non-flux
boundary)10. Specifically, rotor pairs were mutually annihilated if
the phase gradient between the rotors, perpendicular to the line
connecting the phase singularities, was sufficiently large to stop
propagation. On the basis of the number of rotors observed in our
recording array, each rotor occupied on average 12 64 cm2.
According to rough measurements of heart surface area, we estimate
that the total number of rotors during fibrillation would be
approximately 1–2 for rabbits, 5 for sheep, and 15 for humans
(assuming the rotor density is the same in humans).
These results indicate that analysing the complex spatiotemporal
patterns seen during fibrillation on the surface of the heart can be
greatly simplified by identifying and analysing phase singularities.
This analysis reveals topological restrictions to the dynamics of
fibrillation10: first, phase lines do not intersect; second, phase
singularities are joined via isophase lines to two other singularities
with opposite chirality (or a boundary); and third, phase singula-
rities form and terminate as oppositely rotating pairs (Fig. 4). Under
certain conditions, phase singularities give rise to rotors, which
sustain fibrillation. The direct observation of phase singularities has
led, for the first time, to the quantification of fibrillation in terms of
rotor number and lifespan, and to the elucidation of the mechan-
isms underlying the formation and termination of rotors. M
.........................................................................................................................
Methods
The experimental protocols, video imaging-recording system, and signal
processing have been described previously14,28. We acquired video images
(typically 200 3100 pixels) from the ventricular surface at a rate of 120
frames s
−1
(Dt¼0:00833 s). We applied spatial and temporal filtering to
improve the signal-to-noise ratio28. Two-dimensional phase portraits were
obtained by plotting F(t) versus FðtþtÞ(ref. 29), where tþnDtand nis the
frame number. The value of twas chosen to be roughly one-quarter of the cycle
length during fibrillation (t¼25 ms); this value roughly corresponds to the
first zero crossing of the autocorrelation of F, indicating linear independence. A
new variable phase, v(t), was computed as atanðFðtþtÞ2Fmean;FðtÞ2FmeanÞ.
Isolated hearts from rabbits of ,3 kg (n¼3) and from sheep of ,20 kg
(n¼3) body weight were maintained at 36–388C. Ventricular fibrillation was
initiated by rapid pacing, and 4-s recordings were obtained at least 5 min after
initiation (perfusion was maintained during fibrillation). In Figs 2–4, white
and black bars near images reflect a distance of 1 cm. Values are presented as
mean 6s:d:m:Comparisons were made with paired student t-tests.
Received 30 September; accepted 8 December 1997.
1. Myerburg, R. J. et al. in Cardiac Electrophysiology, From Cell to Bedside (W. B. Saunders, Philadelphia,
1990).
2. Garfinkel, A. et al. Quasiperiodicity and chaos in cardiac fibrillation. J. Clin. Invest. 99, 305 –314
(1997).
3. Witkowski, F. X. et al. Evidence for determinism in ventricular fibrillation. Phys. Rev. Lett. 75, 1230–
1233 (1995).
4. Gray, R,. A. et al. Mechanisms of cardiac fibrillation. Science 270, 1222– 1223 (1995).
5. Ikeda, T. et al. Mechanism of spontaneous termination of functional reentry in isolated canine right
atrium. Circulation 94, 1962–1973 (1996).
6. Lee, J. J. et al. Reentrant wavefronts in Wiggers’ stage IIventricular fibrillation. Circ. Res. 78, 660 –675
(1996).
7. Salzberg, B. M., Davila, H. V. & Cohen, L. B. Optical recordings of impulses in individual neurons of
an invertebrate central nervous system. Nature 246, 508–509 (1973).
8. Davidenko, J. M., Pertsov, A. M., Salomonsz, R., Baxter, W. T.& Jalife, J. Stationary and drifting spiral
waves of excitation in isolated cardiac muscle. Nature 355, 349–351 (1991).
9. Pertsov, A. M., Davidenko, J. M., Salomonsz, R., Baxter, W. T. & Jalife, J. Spiral waves of excitation
underlie reentrant activity in isolated cardiac muscle. Circ. Res. 72, 631 –650 (1993).
10. Winfree, A. T. When Time Breaks Down. (Princeton Univ. Press, 1987).
11. Winfree, A. T. Scroll-shaped waves of chemical activity in three dimensions. Science 181, 937 –939
(1973).
12. Goldbeter, A. Mechanism for oscillatory synthesis of cAMP in Dictyostelium discoideum.Nature 253,
540–542 (1975).
13. Lechleiter, J., Girdad, S. & Peralta, E. Spiral calcium wave propagation and annihilation in Xenopus
laevis oocytes. Science 252, 123 –125 (1991).
14. Gray, R. A. et al. Non-stationary vortex-like reentry as a mechanism of polymorphic ventricular
tachycardia in the isolated rabbit heart. Circulation 91, 2454–2469 (1995).
15. Winfree, A. T. Electrical turbulence in three-dimensional heart muscle. Science 266, 1003–1006
(1994).
16. Moe, G. K. & Abildskov, J. A. Atrial fibrillation as a self-sustaining arrhythmia independent of focal
discharge. Am. Heart J. 58, 59–70 (1959).
17. Allessie, M. A., Lammers, W., Bonke, F.I. M. & Hollen, J. in Cardiac Electrophysiology and Arrhythmias
265–275 (Grune and Stratton, Orlando, 1985).
18. Krinsky, V.I. Mathematical models of cardiac arrhythmias (spiral waves). Pharmacol. Ther. B 3, 539–
555 (1978).
19. Zykov, V.S. Simulation of Wave Processes in Excitable Media. (University Press, New York/Manchester,
1987).
20. Gray, R. A. & Jalife, J. Spiral waves and the heart. Int. J. Bifurc. Chaos 6, 415–435 (1996).
21. Bayly, P. V. et al. Efficient electrode spacing for examining spatial organization during ventricular
fibrillation. J. Cardiovasc. Electrophysiol. 4, 533– 546 (1993).
22. Bove, R. T. & Dillon, S. M. A new high performance system for imaging cardiac electrical activity.
Circulation 94, I–714 (1996).
23. Cha, Y. M., Birgersdotter-Green, U., Wolf, P. L., Peters, B. B. & Chen, P. S. The mechanism of
termination of reentrant activity in ventricular fibrillation. Circ. Res. 74, 495–506 (1994).
24. Glass, L. & Mackay, M. C. From Clocks to Chaos (Princeton Univ. Press, 1988).
25. Agladze, K. I. & Krinsky, V. I. Multi-armed vortices in an active chemical medium. Nature 296, 424 –
426 (1982).
26. Fast, V.G. & Pertsov, A. M. Drift of a vortex in the myocardium. Biophysics 35, 489–494 (1990).
27. Gotoh, M. et al. Cellular graded responses and ventricular vulnerability to reentry by a premature
stimulus in isolated canine ventricle. Circulation 95, 2141– 2154 (1997).
28. Gray, R. A., Ayers, G. & Jalife,J. Video imaging of atrial defibrillation in the sheep heart. Circulation 95,
1038–1047 (1997).
29. Takens, F. in Dynamical Systems and Turbulence (eds Rand, D. A. & Young, L. S.) Lecture Notes in
Mathematics 898, 366–381 (Springer, Berlin, 1981).
30. Fitzhugh, R. Impulses and physiological states in theoretical models of nerve membrane. Biophys. J. 1,
445–446 (1961).
Acknowledgements. We thank O. Berenfield, Z. Silverman, J. Jiang and M. Flanagan for technical
assistance and M. Vinson for reading the manuscript. This work was supported by grants from the
Whitaker Foundation and the N.I.H.
Correspondence and requests for materials should be addressed to R.A.G. (e-mail: rag@crml.uab.edu).
Spatiotemporal evolution of
ventricular fibrillation
Francis X. Witkowski*, L. Joshua Leon†,
Patricia A. Penkoske‡, Wayne R. Giles§, Mark L. Spanok,
William L. Ditto¶& Arthur T. Winfree#
*Department of Medicine, and ‡Department of Surgery, University of Alberta,
Edmonton, Alberta T6G 2R7, Canada
†Ecole Polytechnique, Montreal, Quebec H3C 3J7, Canada
§Department of Physiology and Biophysics, University of Calgary, Calgary,
Alberta T2N 4N1, Canada
kNaval Surface Warfare Center, West Bethesda, Maryland 20817, USA
¶Applied Chaos Laboratory, School of Physics, Georgia Institute of Technology,
Atlanta, Georgia 30332, USA
#Department of Ecology and Evolutionary Biology, University of Arizona, Tucson,
Arizona 85721, USA
.........................................................................................................................
Sudden cardiac death is the leading cause of death in the indus-
trialized world, with the majority of such tragedies being due to
ventricular fibrillation1. Ventricular fibrillation is a frenzied and
irregular disturbance of the heart rhythm that quickly renders the
heart incapable of sustaining life. Rotors, electrophysiological
structures that emit rotating spiral waves, occur in several systems
that all share with the heart the functional properties of excit-
ability and refractoriness. These re-entrant waves, seen in numer-
ical solutions of simplified models of cardiac tissue2, may occur
during ventricular tachycardias3,4. It has been difficult to detect
such forms of re-entry in fibrillating mammalian ventricles5– 8.
Here we show that, in isolated perfused dog hearts, high spatial
and temporal resolution mapping of optical transmembrane
potentials can easily detect transiently erupting rotors during
the early phase of ventricular fibrillation. This activity is char-
acterized by a relatively high spatiotemporal cross-correlation.
During this early fibrillatory interval, frequent wavefront colli-
sions and wavebreak generation9are also dominant features.
Interestingly, this spatiotemporal pattern undergoes an evolution
to a less highly spatially correlated mechanism that lacks the
epicardial manifestations of rotors despite continued myocardial
perfusion.
Ventricular fibrillation is a complicated, often lethal, but poorly
understood, high-frequency mode of electrical activity. It can be
