© 2006 Nature Publishing Group
Atomic structure of a Na1- and K1-conducting
Ning Shi1, Sheng Ye1, Amer Alam1, Liping Chen1& Youxing Jiang1
Ion selectivity is one of the basic properties that define an ion
channel. Most tetrameric cation channels, which include the K1,
Ca21, Na1and cyclic nucleotide-gated channels, probably share a
similar overall architecture in their ion-conduction pore, but the
structural details that determine ion selection are different.
Although K1channel selectivity has been well studied from a
structural perspective1,2, little is known about the structure of
other cation channels. Here we present crystal structures of the
NaK channel from Bacillus cereus, a non-selective tetrameric
cation channel, in its Na1- and K1-bound states at 2.4A˚and
2.8A˚resolution, respectively. The NaK channel shares high
sequence homology and a similar overall structure with the
bacterial KcsA K1channel, but its selectivity filter adopts a
different architecture. Unlike a K1channel selectivity filter,
which contains four equivalent K1-binding sites, the selectivity
filter of the NaK channel preserves the two cation-binding sites
equivalent to sites 3 and 4 of a K1channel, whereas the region
in which ions can diffuse but not bind specifically. Functional
analysis using an86Rb flux assay shows that the NaK channel can
conduct both Na1and K1ions. We conclude that the sequence of
the NaK selectivity filter resembles that of a cyclic nucleotide-
gated channel and its structure may represent that of a cyclic
nucleotide-gated channel pore.
All Kþchannels contain the highly conserved signature sequence
TVGYG, which is essential for Kþion selectivity3,4. The cyclic
with Kþchannels, but is non-selective and permeable to most group
and Kþchannels is that CNG channels lack a tyrosine and glycine
residue from the conserved TVGYG sequence (bold) present in Kþ
results in the loss of ion selectivity in CNG channels.
In a search of the microbial genome, we identified two two-
transmembrane channels from Bacillus cereus and Bacillus anthracis
that have sequences very similar to the KcsA Kþchannel, except for
their selectivity filters, which resemble those of CNG channels with
the sequence TVGDG or TVGDA (Supplementary Fig. S1). On the
basis of this similarity, we hypothesized that these channels can non-
selectively conduct both Naþand Kþ. This proved to be correct
according to our functional assay. Here we present the crystal
structure of the channel from Bacillus cereus in complex with Naþ
at 2.4A˚(Table 1 and Supplementary Table S1), and we call this
channel NaK (conducting both Naþand Kþions).
NaK shares several common features with KcsA (Fig. 1a, b). It
contains two membrane-spanning segments, M1 and M2, corre-
helices form an ‘inverted teepee’. The ion conduction pathway has a
water-filled cavity near the centre of the membrane, with four pore
architecture shared by NaK and KcsA presumably underlies their
function to conduct cations across the cell membrane.
Two notable differences distinguish NaK and KcsA from each
other. First, the amino-terminal 19 amino acids of NaK form an
interfacial helix (M0 helix) parallel to the membrane. Phe4, Leu8
and Met11 at the N-terminal region of each M0 helix form
hydrophobic contacts with Val26, Val29 and Leu30 at the
N terminus of helix M1 from a neighbouring subunit (Fig. 1b).
The four M0 helices form a cuff that encircles the inner helix bundle
crossing, and seem to be positioned appropriately to affect the
opening and closing of the pore.
The second structural difference between NaK and KcsA is
observed in the selectivity filter (Fig. 2). The KcsA filter is formed
sequence (Fig. 2b, left). The backbone carbonyl oxygen atoms from
the TVGY residues, along with the hydroxyl oxygen atom from
Thr75, point towards the centre of the pathway and form four
equivalent binding sites (numbered 1 to 4) for dehydrated Kþions.
The eight oxygen atoms surrounding each site mimic the hydration
sit at the perimeter of the filter entrance and point directly into the
extracellular solution, generating an electronegative environment
that stabilizes a half-dehydrated Kþion. The side chains from the
four Tyr78 residues form specific packing interactions with neigh-
bouring aromatic residues from the pore helices that surround the
In comparison, the NaK filter has a sequence of63TVGDG67, with
the hydroxyl oxygen atom from Thr63 and backbone carbonyl
oxygen atoms from Thr63 and Val64 forming two ion-binding
sites equivalent to sites 3 and 4 in KcsA (Fig. 2b, right). For
comparative purposes, these two sites are also numbered 3 and 4 in
Fig. 2b. Sites 1 and 2 of KcsA do not exist in NaK owing to the
Table 1 | Refinement statistics
Number of atoms
Bond lengths (A˚)
Bond angles (8)
1Department of Physiology, University of Texas Southwestern Medical Center, Dallas, Texas 75390-9040, USA.
Vol 440|23 March 2006|doi:10.1038/nature04508
© 2006 Nature Publishing Group
different backbone conformations adopted by the GDG residues at
these two positions. The carbonyl groups from Gly65 and Asp66
align tangentially to the ion conduction pathway, and therefore their
oxygen atoms do not point towards the centre to coordinate ions.
Furthermore, the Ca of Asp66 is directed away from the centre and
its side chain protrudes upward and is exposed to the extracellular
specifically. At the entryway to the filter from the extracellular
solution, the main chain at Gly67 pinches inwards so that its
carbonyl oxygen points towards the pore axis, forming an ion-
binding site, which, as we discuss below, can have important
Superimposition of the NaK filter with that of KcsA gives rise to a
root-mean-square deviation (r.m.s.d.) of 1.6A˚in their main-chain
positions. These differences are caused mainly by the last three
residues, which have an r.m.s.d. of 2.1A˚. The observed differences
between NaK and KcsA are not due to crystal packing, as the NaK
filter is in the centre of the channel tetramer and is not involved in
protein packing in the crystal.
In KcsA, Kþions are required to stabilize the selectivity filter: the
structure of KcsA in conditions of high Naþ, low Kþreveals
significant changes at the filter region9. In contrast, the structure of
NaK does not depend on the ion composition: we determined a
structure in KCl (rather than NaCl) at 2.8A˚and found no apparent
differences in the filter structure (Fig. 3a–c). Thus, NaK retains its
proper conformation for ion conduction in both Naþand Kþ.
Electron density maps of NaK show that ions can bind at the
extracellular entrance, along the selectivity filter, and in the central
that these ions bind at all the same sites in the pore. The crystal-
lization solutions also contained 200mM CaCl2in addition to
100mM NaCl or KCl. To deduce whether it is the monovalent or
divalent cations that bind at specific sites within the pore, we carried
out soaking experiments in which the NaK crystals were soaked in
stabilization solutions containing various monovalent and divalent
salts at the same concentrations as those in the crystallization solu-
tions. To minimize non-isomorphism, a native reference crystal was
subjected to the same soaking procedures in a stabilization solution
containing CaCl2and NaCl (Supplementary Table S2). Difference
electron density maps using Fourier coefficients (Fsoak–Freference)
were calculated with phases from the model (Fig. 3d–h).
The electron density difference between crystals soaked with
Na/Ca and Cs/Ca, Na/Ca and Rb/Ca, or Na/Ca and Tl/Ca all give
the binding of Csþ, Rbþand Tlþat these sites (Fig. 3d–f, red mesh).
The difference map between Na/Ca- and Na/Mg-soaked crystals
reveals only one Ca2þ-binding site, at the extracellular entrance to
the filter (Fig. 3g, green mesh). This suggests that only monovalent
cations bind at sites 3, 4 and the central cavity, and that divalent
cations can bind at the external site outside the selectivity filter. On
the basis of these results, looking back at the ion omit maps we can
conclude that it is Naþand Kþ, but not Ca2þ, inside the selectivity
filter and central cavity (Fig. 3a, b).
The Ca2þ-binding site formed by the four carbonyl oxygen atoms
from the Gly67 residues may not be specific for divalent cations—
Ca2þbinding might have prevailed simply because it is at a higher
concentration in the crystallization solutions. A 2Fo–Fcmap of a
native crystal soaked in a stabilizing solution containing only NaCl
still shows reasonably strong electron density at this position (data
not shown), possibly owing to Naþbinding. Monovalent cation
binding at this position probably occurs as ions conduct through
the pore. Divalent cations might bind at this position to block
the conduction of monovalent cations through the pore, a property
relevant to CNG channels, for which a divalent cation block
of monovalent currents is important for physiological channel
The difference map between Na/Ca and Na/Ba soaked crystals
reveals two Ba2þ-binding sites, one at the filter entrance (equivalent
to the Ca2þ-binding site) and another at site 3 in the selectivity filter
(Fig. 3h, green mesh). Ba2þcan block a Kþchannel16,17bybinding to
site 4 in the selectivity filter18. Ba2þmight also serve as a blocker for
in KcsA) of the selectivity filter or at the external entrance. Indeed,
our functional assay showed that Ba2þreduces the Rbþflux at
micromolar concentrations (data not shown).
Neither the ion-omit maps of native crystals nor the difference
maps between various soaked crystals reveal any specific ion binding
in the vestibule of the filter. The 2Fo–Fcmaps of native crystals
(Fig. 2a) and the difference map between Na/Ca- and Tl/Ca-soaked
crystals reveal weak electron density in the vestibule, indicating the
presence of an ion with low occupancy. Ion binding in the vestibule
surface lining main-chain carbonyl groups from Gly65 and Asp66.
However, such binding is not as specific as that of sites 3 and 4, and
the bound ion distributes over a larger volume, giving rise to weak
electron density on both 2Fo–Fcand difference maps.
Based on the premise that NaK is capable of conducting Rbþ
(among other monovalent cations), we performed an
assay19using NaK channels reconstituted into liposomes in order
Figure 1 | Overall structure of NaK. a, b, Ribbon representation of NaK
from the intracellular side in b. Green spheres representions in the channel.
Each subunit is individually coloured. Green ball-and-stick representations
show side chains from residues involved in hydrophobic contacts between
M0 and neighbouring M1 helices.
NATURE|Vol 440|23 March 2006
© 2006 Nature Publishing Group
to study channel ion-selectivity properties. We also reconstituted a
truncated form of the channel, NaKND19, in which the N-terminal
M0 helix-forming residues were removed, as we suspected that the
four M0 helices around the helix bundle might lock the channel in a
liposome membrane (high concentration of NaCl inside) so that any
outflow of Naþthrough NaKwould leave a deficit of positive charge
inside and drive the influx of external radioactive86Rb, which could
bemonitored using a scintillation counter. Figure4a showsthe time-
dependent accumulation of86Rb inside liposomes, confirming that
NaK is capable of conducting both Naþand Rbþ. As a much higher
flux ratewasobservedfor NaKND19,thistruncated formofNaKwas
used in all other flux assays. As expected, the flux rate was negligible
for control liposomes (containing no reconstituted protein) and for
liposomes reconstituted with KcsA, which is a Kþ-selective channel.
Addition of gramicidin A, a Group 1A metal ionophore, to the
control liposomes increases Rbþinflux, confirming that it is the
outflow of Naþthat causes Rbþinflux.
We performed the same flux assay using liposomes loaded with
KCl instead of NaCl. As shown in Fig. 4b, liposomes reconstituted
with KcsA or NaKND19 both show86Rb accumulation in a time-
dependent manner, indicating the conduction of Kþin both chan-
nels. The difference in flux rates of NaKND19 in the presence of Naþ
or Kþcould be the result of different batches of liposomes used for
each set of experiments.
different concentrations in addition to86Rb. NaKND19-containing
10min before radioactivity levels in the liposomes were measured.
Any of the test ions that the channel was able conduct were expected
to compete with86Rb and decrease its influx into liposomes. The data
channel is able to conduct both Naþand Kþ, which result in lowering
of the86Rb signal compared to control (86Rb only, in a background of
,35mM NaCl). Liþand NMGþ(N-methyl-D-glucamine) do not
lower the86Rb signal and hence we conclude that the channel is
unable to conduct these cations. Competition assays repeated in
liposomes with KCl yielded the same pattern of results (data not
From the competition assays, it seems that Kþhas a greater effect
conclusions before a more thorough and quantitative functional
analysis of the channel is performed. What is clear from the data is
that the NaK channel is not very selective among Naþ, Kþand Rbþ.
This result is consistent with our crystallographic studies showing
cavity of the NaK channel.
One of the more intriguing conclusions we draw from the NaK
channel has to do with ion-binding sites 3 and 4. Chemically, these
Figure 2 | Structural comparison of the selectivity filters in NaK and
KcsA. a, Stereo view of a 2Fo–Fcmap at 1j contour (blue mesh), showing
the selectivity region of NaKwith the front subunit removed. Green spheres
represent ions in the filter and cavity. b, Structural details of the selectivity
filters from KcsA (left) and NaK (right). The front and rear subunits have
been removed for clarity. Only discrete ion-binding sites in the filter are
numbered (1–4 from the extracellular side in KcsA, 3 and 4 in NaK).
NATURE|Vol 440|23 March 2006
© 2006 Nature Publishing Group
sites are indistinguishable from the corresponding positions in the
KcsA Kþchannel, yet their ion-binding properties are clearly
different from those observed in Kþchannels. Specifically, in Kþ
channels, Naþis never observed to bind at position 3, whereas in the
NaK channel, Naþbinds there without difficulty. On the basis of
these experimental data, we conclude that structural differences in
otherwise chemically identical sites must account for their different
ion-binding properties. These structural differences must arise from
differences in amino acid packing around the binding site.
Our structural study of NaK also addresses a fundamental issue
concerning ion selectivity in Kþchannels. A theory-based compu-
tational assay has suggested that Kþselectivity in Kþchannels
originates not from geometric constraints provided by the protein,
but rather from electrostatic repulsion between carbonyl oxygen
atoms—the repulsion was said to prevent the oxygen atoms from
approaching close enough to each other to form a cage small enough
to coordinate the smaller Naþion20. Comparison of NaK and KcsA
provides a clear experimental demonstration that electrostatic
repulsion between carbonyl oxygen atoms is not the origin of Kþ
over Naþselectivity in Kþchannels, and that protein atoms
surrounding the ion-binding site must confer size-selectivity
through geometric constraints.
We also find that the external site in the NaK selectivity filter is
distinct from the ion-binding site just outside the selectivity filter of
divalent cations, whereas in NaK it can bind divalent cations that
could serve to block monovalent cation conduction. This difference
can be understood on the basis of their different structural proper-
very constrained. An apparent greater flexibility of the main chain in
this region of the NaK selectivity filter probably allows the carbonyl
oxygen atoms to conform to an ion with more degrees of freedom.
Figure 4 |86Rb flux assay. a, Time-dependent86Rb influx into liposomes
prepared in NaCl. The liposomes contain NaK, NaKND19, KcsA or no
protein (as a control). Arrow indicates86Rb influx into control
liposomes upon addition of 10mgml21gramicidin A. b, Time-dependent
86Rb influx into KcsA or NaKND19-containing liposomes prepared in KCl.
monovalent cations. Three final concentrations of each cation (0.1, 0.5 and
1.0mM) were tested in the assay. No external cations were added in the
control experiment. c.p.m., counts per minute.
Figure 3 | Ion binding in the NaK channel. a, b, 2Fo–Fcion-omit maps of
Naþ(a) and Kþ(b) complexes of NaK contoured at 6j (red mesh).
c, Superimposition of NaK selectivity filters in Naþ- (green) and Kþ-
(yellow) bound states. d–h, Fsoak–Freferencedifference maps between the
reference crystal and crystals with various soaking conditions reveal the
binding of cations (labelled underneath each panel) in NaK. All maps are
contoured at 10j except for the densityof Ca2þbinding, which is contoured
at 6j. Densities for monovalent and divalent cations are coloured red and
NATURE|Vol 440|23 March 2006
© 2006 Nature Publishing Group
reveals the structural basis of external divalent cation blockage in
CNG channels. The physiological significance of this blockage to
visual transduction was first revealed nearly 20years ago21. Here for
the first time we observe its chemical basis in the form of a precise
binding site at the entryway.
The NaK channel from Bacillus cereus was cloned into the pQE60 vector and
expressed in E. coli XL1-Blue cultures. The protein was purified as a tetramer in
n-decyl-b-D-maltoside with NaCl or KCl as the monovalent salt. Crystals were
grown by sitting-drop vapour diffusion at 208C by mixing equal volumes of
protein solution at 30–35mgml21and reservoir solution containing 36–42%
4% t-butanol. The crystals were of space group C2221with cell dimensions
a ¼ 81.5A˚, b ¼ 85.5A˚, c ¼ 129.6A˚, a ¼ b ¼ g ¼ 908, and contained two
subunits per asymmetric unit. The four-fold axis of the channel tetramer
coincides with one of the crystallographic dyads.
All data were collected at the Advanced Photon Source (APS) and processed
usingHKL200022. The structure wasdeterminedbysingle isomorphousreplace-
sites were determined with SHELXD23and initial phases were improved by
solvent flattening in n-decyl-b-D-maltoside24. The modelwas constructed in O25
and was refined in CNS26to 2.4A˚with Rwork¼ 23.5% and Rfree¼ 26.0%, and
structure was determined by molecular replacement and was refined to 2.8A˚
with Rwork¼ 24.1% and Rfree¼ 28.0%.
For soaking experiments, crystals of Naþcomplexes were soaked in a
stabilization solution of 40% PEG400, 100mM Tris HCl pH8.0, 20mM
n-decyl-b-D-maltoside, 4% t-butanol, 100mM XCl and 200mM YCl2, where
X and Y represent a monovalent cation (Naþ, Rbþor Csþ) and a divalent cation
(Ca2þ, Mg2þor Ba2þ), respectively. For Tlþ-soaking, 100mM TlNO3and
200mM Ca(NO3)2were used.
All channel proteins used in the flux assay were reconstituted into lipid
vesicles composed of 1-palmitoyl-2-oleoyl-phosphatidylethanolamine and
described27. The86Rb flux assay was performed as described19. For competition
Received 28 July; accepted 5 December 2005.
Published online 8 February 2006.
1.Zhou, Y., Morais-Cabral, J. H., Kaufman, A. & MacKinnon, R. Chemistry of ion
coordination and hydration revealed by a Kþchannel– -Fab complex at 2.0A˚
resolution. Nature 414, 43– -48 (2001).
Doyle, D. A. et al. The structure of the potassium channel: molecular basis of
Kþconduction and selectivity. Science 280, 69– -77 (1998).
Heginbotham, L., Abramson, T. & MacKinnon, R. A functional connection
between the pores of distantly related ion channels as revealed by mutant Kþ
channels. Science 258, 1152– -1155 (1992).
Heginbotham, L., Lu, Z., Abramson, T. & MacKinnon, R. Mutations in the Kþ
channel signature sequence. Biophys. J. 66, 1061– -1067 (1994).
Yau, K. W. & Baylor, D. A. Cyclic GMP-activated conductance of retinal
photoreceptor cells. Annu. Rev. Neurosci. 12, 289– -327 (1989).
Kaupp, U. B. & Seifert, R. Cyclic nucleotide-gated ion channels. Physiol. Rev. 82,
769– -824 (2002).
Matulef, K. & Zagotta, W. N. Cyclic nucleotide-gated ion channels. Annu. Rev.
Cell Dev. Biol. 19, 23– -44 (2003).
Zagotta, W. N. & Siegelbaum, S. A. Structure and function of cyclic nucleotide-
gated channels. Annu. Rev. Neurosci. 19, 235– -263 (1996).
Zhou, Y. & MacKinnon, R. The occupancy of ions in the Kþselectivity filter:
charge balance and coupling of ion binding to a protein conformational
change underlie high conduction rates. J. Mol. Biol. 333, 965– -975 (2003).
10. Haynes, L. W., Kay, A. R. & Yau, K. W. Single cyclic GMP-activated channel
activity in excised patches of rod outer segment membrane. Nature 321, 66– -70
11.Stern, J. H., Knutsson, H. & MacLeish, P. R. Divalent cations directly affect the
conductance of excised patches of rod photoreceptor membrane. Science 236,
1674– -1678 (1987).
12. Colamartino, G., Menini, A. & Torre, V. Blockage and permeation of divalent
cations through the cyclic GMP-activated channel from tiger salamander
retinal rods. J. Physiol. (Lond.) 440, 189– -206 (1991).
13. Zimmerman, A. L. & Baylor, D. A. Cation interactions within the cyclic GMP-
activated channel of retinal rods from the tiger salamander. J. Physiol. (Lond.)
449, 759– -783 (1992).
14. Frings, S., Seifert, R., Godde, M. & Kaupp, U. B. Profoundly different calcium
permeation and blockage determine the specific function of distinct cyclic
nucleotide-gated channels. Neuron 15, 169– -179 (1995).
15. Zufall, F., Firestein, S. & Shepherd, G. M. Cyclic nucleotide-gated ion channels
and sensory transduction in olfactory receptor neurons. Annu. Rev. Biophys.
Biomol. Struct. 23, 577– -607 (1994).
16. Armstrong, C. M. & Taylor, S. R. Interaction of barium ions with potassium
channels in squid giant axons. Biophys. J. 30, 473– -488 (1980).
17. Armstrong, C. M., Swenson, R. P. Jr & Taylor, S. R. Block of squid axon K
channels by internally and externally applied barium ions. J. Gen. Physiol. 80,
663– -682 (1982).
18. Jiang, Y. & MacKinnon, R. The barium site in a potassium channel by X-ray
crystallography. J. Gen. Physiol. 115, 269– -272 (2000).
19. Heginbotham, L., Kolmakova-Partensky, L. & Miller, C. Functional
reconstitution of a prokaryotic Kþchannel. J. Gen. Physiol. 111, 741– -749 (1998).
20. Noskov, S. Y., Berneche, S. & Roux, B. Control of ion selectivity in potassium
channels by electrostatic and dynamic properties of carbonyl ligands. Nature
431, 830– -834 (2004).
21. Haynes, L. & Yau, K. W. Cyclic GMP-sensitive conductance in outer segment
membrane of catfish cones. Nature 317, 61– -64 (1985).
22. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in
oscillation mode. Methods Enzymol. 276, 307– -326 (1997).
23. Schneider, T. R. & Sheldrick, G. M. Substructure solution with SHELXD. Acta
Crystallogr. D Biol. Crystallogr. 58, 1772– -1779 (2002).
24. Collaborative Computational Project, Number 4, The CCP4 suite: programs for
protein crystallography. Acta Crystallogr. D Biol. Crystallogr. 50, 760– -763 (1994).
25. Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard. Improved methods for
building protein models in electron density maps and the location of errors in
these models. Acta Crystallogr. A 47, 110– -119 (1991).
26. Brunger, A. T. et al. Crystallography & NMR system: A new software suite for
macromolecular structure determination. Acta Crystallogr. D Biol. Crystallogr.
54, 905– -921 (1998).
27. Heginbotham, L., LeMasurier, M., Kolmakova-Partensky, L. & Miller, C. Single
Streptomyces lividans Kþchannels: functional asymmetries and sidedness of
proton activation. J. Gen. Physiol. 114, 551– -560 (1999).
Supplementary Information is linked to the online version of the paper at
Acknowledgements We thank R. MacKinnon for discussion and critical review
of the manuscript. Use of the Argonne National Laboratory Structural Biology
Center beamlines at the Advanced Photon Source was supported by the US
Department of Energy, Office of Energy Research. We thank the beamline staff
for assistance in data collection. This work was supported by grants from the
David and Lucile Packard Foundation (to Y.J.) and the Searle Scholars Program
Author Contributions S.Y. and A.A. contributed equally to this work. S.Y. helped
with the structure determination and A.A. performed the86Rb flux assay.
Author Information Atomic coordinates of the Naþand Kþcomplexes of the
NaK channel have been deposited in the Protein Data Bank with accession
numbers of 2AHY and 2AHZ, respectively. Reprints and permissions
information is available at npg.nature.com/reprintsandpermissions. The authors
declare no competing financial interests. Correspondence and requests for
materials should be addressed to Y.J. (email@example.com).
NATURE|Vol 440|23 March 2006