Crystallographic analysis of counterion effects on subtilisin enzymatic action in acetonitrile

Article abstract of DOI:10.1021/ja908703c

When enzymes are in low dielectric nonaqueous media, it would be expected that their charged groups would be more closely associated with counterions. There is evidence that these counterions may then affect enzymatic activity. Published crystal structures of proteins in organic solvents do not show increased numbers of associated counterions, and this might reflect the difficulty of distinguishing cations like Na⁺ from water molecules. In this paper, the placement of several Cs⁺ and Cl^- ions in crystals of the serine protease subtilisin Carlsberg is presented. Ions are more readily identified crystallographically through their anomalous diffraction using softer X-rays. The protein conformation is very similar to that of the enzyme without CsCl in acetonitrile, both for the previously reported (1SCB) and our own newly determined model. No fewer than 11 defined sites for Cs ⁺ cations and 8 Cl^- anions are identified around the protein molecule, although most of these have partial occupancy and may represent nonspecific binding sites. Two Cs⁺ and two Cl^- ions are close to the mouth of the active site cleft, where they may affect catalysis. In fact, crosslinked CsCl-treated subtilisin crystals transferred to acetonitrile show catalytic activity several fold higher than the reference crystals containing Na⁺. Presoaking with another large cation, choline, also increases the enzyme activity. The active site appears only minimally sterically perturbed by the ion presence around it, so alternative activation mechanisms can be suggested: an electrostatic redistribution and/or a larger hydration sphere that enhances the protein domain.

Full text of DOI:10.1021/ja908703c

Published on Web 01/25/2010
Crystallographic Analysis of Counterion Effects on Subtilisin
Enzymatic Action in Acetonitrile
Michele Cianci,*,† Bartlomiej Tomaszewski,‡ John R. Helliwell,§ and
Peter J. Halling*,‡
European Molecular Biology Laboratory, Hamburg Outstation, c/o DESY, Building 25a,
Notkestrasse 85, 22603 Hamburg, Germany, WestCHEM, Department of Pure & Applied
Chemistry, UniVersity of Strathclyde, Glasgow G1 1XL, United Kingdom, and Department of
Chemistry, UniVersity of Manchester, Manchester M13 9PL, United Kingdom
Received October 20, 2009; E-mail: m.cianci@embl-hamburg.de; p.j.halling@strath.ac.uk
Abstract: When enzymes are in low dielectric nonaqueous media, it would be expected that their charged
groups would be more closely associated with counterions. There is evidence that these counterions may
then affect enzymatic activity. Published crystal structures of proteins in organic solvents do not show
increased numbers of associated counterions, and this might reflect the difficulty of distinguishing cations
like Na+ from water molecules. In this paper, the placement of several Cs+ and Cl- ions in crystals of the
serine protease subtilisin Carlsberg is presented. Ions are more readily identified crystallographically through
their anomalous diffraction using softer X-rays. The protein conformation is very similar to that of the enzyme
without CsCl in acetonitrile, both for the previously reported (1SCB) and our own newly determined model.
No fewer than 11 defined sites for Cs+ cations and 8 Cl- anions are identified around the protein molecule,
although most of these have partial occupancy and may represent nonspecific binding sites. Two Cs+ and
two Cl- ions are close to the mouth of the active site cleft, where they may affect catalysis. In fact, cross-
linked CsCl-treated subtilisin crystals transferred to acetonitrile show catalytic activity several fold higher
than the reference crystals containing Na+. Presoaking with another large cation, choline, also increases
the enzyme activity. The active site appears only minimally sterically perturbed by the ion presence around
it, so alternative activation mechanisms can be suggested: an electrostatic redistribution and/or a larger
hydration sphere that enhances the protein domain.
Introduction
be strongly favored. Hence, they may well have important
effects on function. There is evidence8,9 that counterions affect
There is interest, both preparative and fundamental, in the
behavior of enzymes in low-water media, such as those based
on organic solvents.1-7 Some aspects can be qualitatively
different from aqueous solution, such as the role of counterions
to charged groups in proteins. Proteins carry a number of
different charged groups, some of which are usually essential
for proper function. Under most conditions, the molecule will
have a net charge, and electroneutrality requires that this must
be balanced by counterions. In aqueous solution, the high
dielectric allows these counterions to be some distance from
the protein, so their presence and identity can normally be
disregarded in discussions of function. In low dielectric media,
however, close approach and ion-pairing of counterions will
the activity of enzymes in mainly organic media. Essentially,
protein ionization (and hence enzymatic activity) becomes
dependent on the availability of counterions as well as of H+.
However, direct evidence for the location of counterions has
been lacking. A number of 3D structures have been determined
for protein crystals that have been rinsed or soaked with organic
solvents.10-15 With nonpolar organic solvents having little
capacity to dissolve water, like hexane, it is likely that much of
the water in the crystal is not removed by the rinsing procedure.
Hence, in these cases the protein environment remains substan-
tially aqueous and counterions may be relatively distant and
(8) Zacharis, E.; Halling, P. J.; Rees, D. G. Proc. Natl. Acad. Sci. U.S.A.
1999, 96, 1201–1205.
† Hamburg Outstation.
(9) Halling, P. J. Curr. Opin. Chem. Biol. 2000, 4, 74–80.
(10) Deshpande, A.; Nimsadkar, S.; Mande, S. C. Acta Crystallogr. 2005,
D61, 1005–1008.
‡ University of Strathclyde.
§ University of Manchester.
(1) Klibanov, A. M. Nature 2001, 409, 241–246.
(2) Castro, G. R.; Knubovets, T. Crit. ReV. in Biotech 2003, 23, 195–
231.
(11) English, A. C.; Groom, C. R.; Hubbard, R. E. Protein Eng. 2001, 14,
47–59.
(12) Gao, X. G.; Maldonado, E.; Perez-Montfort, R.; Garza-Ramos, G.;
De Gomez-Puyou, M. T.; Gomez-Puyou, A.; Rodriguez-Romero, A.
Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 10062–10067.
(13) Mattos, C.; Bellamacina, C. R.; Peisach, E.; Pereira, A.; Vitkup, D.;
Petsko, G. A.; Ringe, D. J. Mol. Biol. 2006, 357, 1471–1482.
(14) Schmitke, J. L.; Stern, L. J.; Klibanov, A. M. Proc. Natl. Acad. Sci.
U.S.A. 1998, 95, 12918–12923.
(3) Gupta, M. N.; I.Roy, I. Eur. J. Biochem. 2004, 271, 2575–2583.
(4) Hudson, E. P.; Eppler, R. K.; Clark, D. S. Curr. Opin. Biotechnol.
2005, 16, 637–643.
(5) Hobbs, H. R.; Thomas, N. R. Chem. ReV. 2007, 107 (6), 2786–2820.
(6) Roosen, C. P.; Muller, P. L.; Greiner, L. Appl. Microbiol. Biotechnol.
2008, 81 (4), 607–614.
(7) van Rantwijk, F.; Sheldon, R. A. Chem. ReV. 2007, 107 (6), 2757–
2785.
(15) Zhu, G. Y.; Huang, Q. C.; Zhu, Y. S.; Li, Y. L.; Chi, C. W.; Tang,
Y. Q. Biochim. Biophys. Acta 2001, 1546, 98–106.
9
10.1021/ja908703c  2010 American Chemical Society
J. AM. CHEM. SOC. 2010, 132, 2293–2300 2293

A R T I C L E S
Cianci et al.
mobile. In contrast, with more polar solvents like acetonitrile
or dioxane, the removal of weakly associated water is likely to
be effective16 (the soaking medium will reach a low water
activity). Although we might expect ion-pairing to be promoted
under these conditions, the structures reported in polar organic
solvents have not shown this. However, the cations most
commonly used in protein crystallization, Na+ and NH4+, are
isoelectronic with water and may often be modeled as water
oxygens, although their coordination with other atoms is
monitored. A further issue is that counterions may be rather
more mobile, or perhaps have several alternative sites with
partial occupancies, which would make them hard to resolve.
Counterion sites can be located more clearly if heavier elements
are used.17-23
and, thereby, of the enhanced catalytic activity in the presence
of CsCl in acetonitrile.
Material and Methods
Crystallization. Subtilisin Carlsberg was purchased from Sigma
(product code: P5380) and used for crystallization without further
purification. The protein powder was dissolved in 330 mM Na-
cacodylate buffer at pH 5.6 to a concentration of 10 mg/mL.
Subtilisin was crystallized by the batch method from buffer solution
saturated with Na2SO4 ∼13% (w/v) as precipitant.32 Long needle-
like crystals grew over a period of two weeks of typical dimensions
50 × 50 × 400 µm3. They were stored in mother liquor for up to
6 weeks, with no evidence for difference in catalytic activity over
this storage period.
Cs Soaking, Cross-Linking, and Acetonitrile Wash. Soaking
with CsCl or other salts and glutaraldehyde cross-linking was
performed in a single step, to reduce the number of washes and
transfers. This reduces the chances of osmotic shocks to the crystals
(very often fatal for the quality of diffraction). So the soak solution
was prepared freshly by mixing 2 volumes of 25% aqueous
glutaraldehyde with 3 volumes of 250 mM Na-cacodylate buffer
pH 6.5 containing 25% v/v glycerol and, where required, 1.5 M
salt. Crystals were soaked in this for 15 min, with vortexing in the
case of suspensions of multiple crystals for activity assay. Washing
in acetonitrile followed immediately. For crystallography, individual
crystals were washed 5 times with acetonitrile for 10 s, again
following the protocols described elsewhere.25 For catalytic activity
measurements, washing required vortexing and settling of the
suspension, with each wash taking about 90 s. There were two
washes in pure acetonitrile, then two in the reaction medium,
acetonitrile containing 1% (v/v) H2O and 1 M 1-propanol. Finally,
crystals were resuspended in reaction medium and preincubated at
25 °C with shaking for 20 min, with vortexing and if necessary
gentle poking to break up aggregated crystals.
Catalytic Activity Determination. A solution of n-acetyl
tyrosine ethyl ester in the reaction medium was added to the
suspension of crystals to give a final concentration of 10 mM. The
reaction mixture was incubated with shaking at 25 °C, with periodic
removal of small supernatant samples for analysis. Conversion to
the propyl ester was followed by HPLC as described,33 and the
initial rate was estimated from at least 5 points on the linear progress
curve. After the reaction, the mixture was centrifuged and the
supernatant was decanted. The pellet of crystals was solubilized
by heating in 1 M NaOH at 100 °C for 30 min, then the protein
content was estimated from the A280 of the cooled solution. This
value was used to estimate the initial rate in specific activity units.
Crystallography Data Collection and Structure Refinement.
Data collection was performed on SRS BL10,34 Daresbury Labora-
tory. A fluorescence scan showed a peak at 5721.51 eV. Diffraction
data were collected using softer X-rays35,36 at the Cs L1 edge (2.167
Å) and processed using the HKL2000 package.37
The structure was solved by molecular replacement using
MolRep,38 and refined with Refmac 5.0.39 Finally, anomalous
difference Fourier maps were calculated and electron density peaks
contoured at 4sigma level or higher were taken into consideration.
Ions were assigned on the basis of coordination geometry, map
peak height, B-factor, and occupancy of each site. Final Rfactor and
Rfree were 0.16 and 0.23 respectively for the Cs+ ion model (Cs-
ACN).
The protease subtilisin Carlsberg has the status of a model
in low-water enzymology, and structures have been reported
for crystals soaked in dioxane24 and acetonitrile.25,14 It has a
catalytic Asp-His-Ser triad, which must bear a net negative
charge for good catalytic activity, as found in alkaline aqueous
solutions. The possible counterions in low dielectric media and
their effects have been discussed.8 Counterion effects may also
contribute to a much studied phenomenon when lyophilized
enzyme powders are used as catalysts in organic media. Activity
of subtilisin and other enzymes is increased enormously by
drying with an excess of neutral salts like KCl,26-31 and
specificity can also be favorably altered.29
In this paper, the placement of ions in crystals of subtilisin
Carlsberg soaked with CsCl and acetonitrile is investigated. In
parallel, the catalytic activity of the crystals suspended in
acetonitrile has been determined. Interestingly, treatment with
CsCl was found to increase the activity several fold. Using X-ray
crystallography with anomalous dispersion, clear sites for Cs+
and Cl- ions are identified around the protein molecule, where
they may be affecting catalysis. Correct identification of the
ion sites helps the modeling of electrostatic potential surfaces
(16) Halling, P. J. Phil. Trans. R. Soc. B 2004, 359, 1287–1297.
(17) Basu, S.; Rambo, R. P.; Strauss-Soukup, J.; Cate, J. H.; Ferre’-
D’Amare, A. R.; Strobel, S. A.; Doudna, J. A. Nat. Struct. Biol. 1998,
5, 986–992.
(18) Tereshko, V.; Wilds, C. J.; Minasov, G.; Prakash, P. T.; Maier, M. A.;
Howard, A.; Wawrzak, Z.; Manoharan, M.; Egli, M. Nucleic Acids
Res. 2001, 29, 1208–1215.
(19) Dauter, Z.; Dauter, M.; Rajashankar, R. J. Acta Crystallogr. 2000,
D56, 232–237.
(20) Korolev, S.; Dementieva, I.; Sanishvili, R.; Minor, W.; Otwinowski,
Z.; Joachimiak, A. Acta Crystallogr. 2001, D57, 1008–1012.
(21) Nagem, R. A. P.; Dauter, Z.; Polikarpov, I. Acta Crystallogr. 2001,
D57, 996–1002.
(22) Evans, G.; Bricogne, G. Acta Crystallogr. 2002, D58, 976–991.
(23) Schultz, D. A.; Friedman, A. M.; White, M. A.; Fox, R. O. Protein
Sci. 2005, 14, 2862–2870.
(24) Schmitke, J. L.; Stern, L. J.; Klibanov, A. M. Proc. Natl. Acad. Sci.
U.S.A. 1997, 94, 4250–4255.
(25) Fitzpatrick, P. A.; Steinmetz, A. C.; Ringe, D.; Klibanov, A. M. Proc.
Natl. Acad. Sci. U.S.A. 1993, 90, 8653–8657.
(26) Khmelnitsky, Y. L.; Welch, S. H.; Clark, D. S.; Dordick, J. S. J. Am.
Chem. Soc. 19943, 116, 2647–2648.
The plot of 〈|∆Fanom|/σ∆Fanom〉 versus resolution for the Cs-ACN
model indicates a strong anomalous scattering signal across the
(27) Ru, M. T.; HiroKane, S. Y.; Lo, A. S.; Dordick, J. S.; Reimer, J. A.;
Clark, D. S. J. Am. Chem. Soc. 2000, 122, 1565–1571.
(28) Altreuter, D. H.; Dordick, J. S.; Clark, D. S. J. Am. Chem. Soc. 2002,
124, 1871–1876.
(32) Petsko, G. A.; Tsernoglou, D. J. Mol. Biol. 1976, 106, 453–456.
(33) Fernandes, J. F. A.; Halling, P. J. Biotechnol. Prog. 2002, 18, 1455–
1457, 2002.
(34) Cianci, M.; Antonyuk, S.; Bliss, N.; Bailey, M. W.; Buffey, S. G.;
Cheung, K. C.; Clarke, J. A.; Derbyshire, G. E.; Ellis, M. J.; Enderby,
M. J.; Grant, A. F.; Holbourn, M. P.; Laundy, D.; Nave, C.; Ryder,
R.; Stephenson, P.; Helliwell, J. R.; Hasnain, S. S. J. Synchrotron.
Radiat. 2003, 12, 442–445.
(29) Eppler, R. K.; Komor, R. S.; Huynh, J.; Dordick, J. S.; Reimer, J. A.;
Clark, D. S. Proc. Natl. Acad. Sci. U. S. A. 2006, 103, 5706–5710.
(30) Eppler, R. K.; Hudson, E. P.; Chase, S. D.; Dordick, J. S.; Reimer,
J. A.; Clark, D. S. Proc. Natl. Acad. Sci. U.S.A. 2008, 105 (41), 15672–
15677.
(31) Serdakowski, A. L.; Dordick, J. S. Trends Biotechnol. 2008, 26 (1),
48–54.
9
2294 J. AM. CHEM. SOC. VOL. 132, NO. 7, 2010

Subtilisin Enzymatic Action in Acetonitrile
A R T I C L E S
diffraction resolution range with an f′′ of 10.42 e- at 5767 eV.
At the same energy a Ca2+ ion has an f′′ of 2.68 e- and Cl- has
an f′′ of 1.57 e-. An anomalous difference Fourier map was
calculated using the phases of just the refined model without
any ions. Display at the 4 sigma level allowed identification of
a number of sites showing favorable association of ions with
the protein, as expected. However, even the low occupancy sites
could be modeled with confidence, because of the clear electron
density derived from the anomalous dispersion measurements
from these atoms. The sites were further confirmed by inspecting
the isomorphous difference Fourier map. Given that Cs+ and
Cl- atoms cannot be assigned only on the basis of the presence
of an anomalous difference Fourier map peak, additional
attention was paid to the coordination geometries, B-factors and
occupancies in order to distinguish between Cs+ and Cl-. The
overall occupancy was modeled to match the average B-factor
of the residues surrounding the site, computed using the program
BAVERAGE.40 Rather surprisingly, there are no less than 11
Cs sites identified (full details in Supplementary Materials 2,
Supporting Information). All of these sites are modeled with
partial occupancy, varying from 0.6 to 0.1. Moreover, an
additional 8 Cl- sites are identified, with occupancies varying
from 1.0 to 0.2. Cs+ ions bound to subtilisin presenting two
types of coordination. The most commonly present is octahedral
often slightly distorted as shown by the Cs-1 for instance (Figure
2a). Another two Cs+ ion sites present a cation-π coordination
with the Cs+ ion held above the aromatic ring of a tyrosine
(Figure 2b). The coordination of chloride ions is mostly
triangular or square pyramidal or octahedral where the chlorine
occupies one vertex (Figure 2c).
Figure 1. Plot of 〈|∆Fanom|/σ∆Fanom〉 versus 1/d2, where d is the interplanar
spacing, for Cs-ACN. The diffraction anomalous signal is above one across
the whole resolution range.
Table 1. Effect of Salt Soaks on Catalytic Activity of Subtilisin
Crystals in Acetonitrile
added salt in soak solution
initial ratea (nmol min-1mg-1
)
None
KCl
Choline-Cl
CsCl
2.4 ( 0.8
5.6 ( 2.6
10.8 ( 3.7
13.7 ( 5.2
a Rates are shown as mean ( standard deviation.
whole diffraction resolution range (Figure 1). A native model (ACN)
was also collected and data refined as before. Native crystals were
cross-linked and soaked in acetonitrile, preparing solutions as before
without the addition of salt. Final Rfactor and Rfree were 0.18 and
0.23, respectively. Coordinates and structure factors have been
submitted to the Protein Data Bank (2wuv and 2wuw PDB codes).
Data collection, processing, and refinement statistics are reported
in Supplementary Materials 1, Supporting Information.
Results and Discussion
The partial occupancies show that each individual site does
not have a very high affinity. However, a plausible interpretation
is that electroneutrality demands that several of these sites are
occupied around every protein molecule but that there is some
freedom for the ions to select between alternative sites. Ions
with low occupancies also have a less defined geometry,
probably due to a higher turnover of their ligands or a more
variable position of the ion itself.
Counterion Sites Close to the Active Site. The active form of
subtilisin has a negative charge on the catalytic triad, formally
located on the Asp-32 carboxylate. There are no obvious
balancing positive sites close by in the protein, so the need for
associated counterions in nonaqueous media has been proposed.8
Figure 3a,b shows that there are two Cs+ sites (3 and 10) near
the entrance to the active site cleft. The Cs-3 site is 6.6 Å away
from the Cl-7 site, and Cs-10 is closely associated with Cl-5
with the overall net charge modeled as weakly negative. Under
the relatively acidic conditions that are needed for subtilisin
crystallization, the catalytic triad is likely to have only a partial
negative charge. The presence of counterions near the entrance
to the active site is however a plausible reason for effects on
catalytic activity.
Catalytic Activity Is Enhanced after Soaking in CsCl. Crystals
for activity measurements were treated essentially identically
to those used for crystallography, to allow comparison of the
observations. Hence, crystals had a 15 min soak in a glutaral-
dehyde cross-linking solution, which also contained the added
salts when tested, followed by washing in pure acetonitrile. As
Table 1 shows, the inclusion of CsCl in the cross-linking soak
solution led to significantly increased activity in acetonitrile. It
appears that a KCl soak may also increase the rate somewhat, while
choline-Cl has a larger effect. All of these specific activities are
rather lower than for the cross-linked crystals normally used,33
perhaps reflecting mass transfer limitations in the larger crystals
grown for crystallography. The catalytic activity measured was also
less reproducible than for other catalyst forms of subtilisin in
acetonitrile. This may reflect variation in crystal sizes in different
reactions, due to random sampling and/or the effect of soaking
treatments. The crystals aggregated on transfer to acetonitrile, and
the aggregates could be quite hard to disperse. Microscopic
examination of crystals in the final acetonitrile suspensions showed
that they were essentially dispersed back to individual crystals, but
that the long needles observed for the original aqueous crystals
had been broken to shorter lengths.
Most of the Closest Cation Interactions are with Carbonyl
Oxygens, Not Formal Negative Charges. Turning to the remain-
ing Cs+ sites, the remarkable observation is how few of them
are actually very close to potential formal negative charges. Only
Cs-5 appears to be coordinated by a protein charge (the
carboxylate of Glu-54). Cs-1 and Cs-10 are coordinated by
chloride ions (Cl-8 and Cl-5), and Cs-8 is rather further (4.7
Å) from Cl-2. For all other sites negative charges are more than
Soaking Crystals in CsCl Reveals Many Nonspecific Sites
for Counterion Association. Using softer X-rays, Cs+ ions give
a strong anomalous signal from their L1 edge across the whole
(35) Helliwell, J. R. J. Synchrotron. Radiat. 2004, 11, 1–3.
(36) Mueller-Dieckmann, C.; Panjikar, S.; Schmidt, A.; Mueller, S.; Kuper,
J.; Geerlof, A.; Wilmanns, M.; Singh, R. K.; Tucker, P. A.; Weiss,
M. S. Acta Crystallogr. 2007, D63, 366–380.
(37) Otwinowski, Z.; Minor, W. Methods Enzymol. 1997, 276, 307–326.
(38) Vagin, A.; Teplyakov, A. J. Appl. Crystallogr. 1997, 30, 1022–1025.
(39) Murshudov, G. N.; Vagin, A. A.; Dodson, E. J. Acta Crystallogr. 1997,
D53, 240–255.
(40) Collaborative Computational Project, Number 4. Acta Crystallogr.
1994, D50, 760-763.
9
J. AM. CHEM. SOC. VOL. 132, NO. 7, 2010 2295

A R T I C L E S
Cianci et al.
Figure 2. Stereoview of selected counterion sites identified. The anomalous difference Fourier map is shown cut at a 4 sigma contour level. Sites are Cs-1 (a); Cs-6 (2b);
Cl-1 (2c); Cs-2 (2d); and Cs-3 (2e). The yellow atoms in (c) come from an adjacent protein molecule in the crystal. Graphics were made with Molecular Graphics.40
5 Å away, often more than 7 Å, with formally neutral atoms
much closer. From Cs-7, water-46 may be bridging an interac-
tion with the Asp-120 carboxylate at 5.3 Å. All the cations must
be neutralizing partial negative charges spread around in the
9
2296 J. AM. CHEM. SOC. VOL. 132, NO. 7, 2010

Subtilisin Enzymatic Action in Acetonitrile
A R T I C L E S
Figure 3. (a) Cesium and Chloride sites around the active site of subtilisin Carlsberg. View down the cleft to the catalytic triad (in yellow). Cs are in red,
Cl in blue and the Ca and Na in gray, all as space-filling spheres; (b) view at 180°. (c) Increase in protein surface for the Cs soaked subtilisin (area 9988.7
Å2, in blue) when compared to the native one (area 9959.5 Å2, in white). In particular, the region around the active site is altered; (d) view at 180°.
crystal, but it is clear that reasonable separation of charges is
possible even in the acetonitrile environment.
distances to carbons below being 4 Å. This type of cation-π
interaction in protein structure and function has been receiving
more and more interest.42,43
In two cases, Cs+ sites are found near the C-termini of alpha-
helices, where the helix dipole would make this a favored
position for positive charge. Cs-6 is close to the helix terminus
at residues 142-145, and Cs-8 is close to the terminus at
residues 251-254.
Cs-4, Cs-5, and Cs-8 show close interactions with two adjacent
protein molecules in the crystal. These sites have generally a very
distorted octahedral geometry. Thus they reflect the particular
packing in the crystals studied, and may well not be favored sites
for cation interactions in other crystal forms or noncrystalline
protein, as often used catalytically in nonaqueous media.
WASP Analysis. The WASP program (WAter Screening
Program)44 was used to assess the specificity of the observed
Cs ion positions. The observed Cs+ ion positions were assigned
to putative waters and Cs-specific valence values calculated. A
The most common close interaction with the protein is in
fact with formally neutral backbone carbonyl oxygens. Figure
2a shows the strongest peak (Cs-1) appearing on the anomalous
difference Fourier map at a 39 sigma level as a distorted
octahedral coordination with four carbonyl oxygens, with a water
and a chloride ion. The last of these was assigned in part because
it showed a positive anomalous signal. Cs-2 (Figure 2d) again
shows distorted octahedral coordination by four carbonyl
oxygens, together with a Ser side-chain oxygen and one or two
waters. For Cs-3 (Figure 2e) the geometry is again distorted
octahedral, with coordination by two carbonyl oxygens and four
waters. Similar coordination is seen with Cs-4 and Cs-8 and
the remaining Cs sites with different degrees of distortion. Cation
coordination by carbonyl oxygens appears to be a common
feature of proteins.41 Two cations, Cs-5 and Cs-6 (Figure 2b),
appear to be interacting with aromatic rings of Tyr residues,
with placement more or less over the face of the ring and
(42) Gromiha, M. M. Biophys. Chem. 2003, 103, 251–258.
(43) Ma, J. C.; Dougherty, D. A. Chem. ReV. 1997, 97, 1303–1324.
(44) Nayal, M.; Di Cera, E. J. Mol. Biol. 1996, 256, 228–234.
(41) Torrance, G. M.; Leader, D. P.; Gilbert, D. R.; Milner-White, E. J. J.
Mol. Biol. 2009, 385, 1076–1086.
9
J. AM. CHEM. SOC. VOL. 132, NO. 7, 2010 2297

A R T I C L E S
Cianci et al.
valence value g1 is the condition necessary and sufficient to
define a monovalent cation binding site.45 The ligation distance
was set to 4.0 Å, with only oxygen, waters and nonprotein atoms
treated as ligating atoms. None of the Cs+ sites has a valence
value g1, meaning that they are not specific (values in
Supplementary Material 2, Supporting Information). For instance
Cs-3 site (Figure 2e), close to the active site, while it scores
highly (valence value 0.685), is still likely to be nonspecific.
This may indicate that these sites are not of very high affinity,
in line with our findings that crystallographic occupancies are
less than 1. Nevertheless, these sites could be occupied by
similar monovalent cations like Na+ or K+, especially for sites
with valence >0.5, and this needs to be considered in discussing
function. This is particularly relevant when the enzyme is used
in the solid state as a catalyst in nonaqueous media, especially
as crystals.46-49
The ionic radius of Cs+ (1.66 Å) is considerably larger than
that for K+ (1.33 Å) and particularly Na+ (0.95 Å), the alkaline
ions more likely to be relevant to normal enzyme behavior.
However, there are indications that these relatively weaker
counterion sites are quite flexible in terms of size that can be
accommodated, as might be expected. In fact Rb+ (1.48 Å) has
been successfully used to replace and identify the position of
Na+ in thrombin,44 while sites in RNA can be identified with
Cs+, albeit at lower occupancy than with the smaller Tl+.17
putative Ca2+ site (pdb codes: 1CSE,48 1C3L,49 1SCD25 and
2SEC51) is, from our data, a possible Na+ site (Na-2 in both
our models), given that there is no peak in the anomalous
difference Fourier maps. The second Ca2+ site in the previous
structures has either partial occupancy or a B-factor a bit higher
than the surrounding atoms, suggesting it may be a mistakenly
assigned Na+. Finally the assignments of sites equivalent to the
Cs ion positions can be compared to other subtilisin Carlsberg
structures. Site Cs-1 in Cs-ACN is confirmed as a genuine
monovalent cation site, assignable to a Na+ ion (Na-1) in the
ACN structure. In the previously described acetonitrile structures
1SCB and 1SCD, a water is found in this position, while in the
structure 1SCA, as well as 1SCN and 2SEC, this is assigned to
a Ca ion. As noted, it is entirely possible that, in those structures,
the site was occupied by Na+ ions that were not clearly
identifiable from the electron density maps. In our case the
assignment is made easier by the presence or absence of
anomalous difference Fourier map peaks in the data. Full details
of the site comparisons are in Supplementary Material 3,
Supporting Information.
Chloride Ion Sites. Cl-1 (Figure 2c) is placed between two
protein molecules, at the vertex of a distorted triangular pyramid,
given by the coordination with two side chain atoms (OH Tyr
263, NH Asn 185) and a water. Cl-5 and Cl-8 are clearly
coordinating Cs+ as counterions, while Cl-2 is also 4.7 Å away
from Cs-8. Similarly to Cl-1, Cl-3 is on the vertex of a pyramid
whose base is given by NH Gln 245 at 2.91 Å, and two waters
at over 4 Å distance. The remaining Cl- ions are organized in
similar coordinations. Interestingly, all of the Cl- ions have
carbonyl groups in proximity making the ion identification
process more difficult. Moreover, there appears not to be many
comprehensive studies of negative ion coordination in proteins
to assist in such distinction.
Comparison with Other Structures. Data were collected and
refined from crystals treated with ACN, with (model Cs-ACN)
and without (ACN) CsCl soak. By superposing the 1SCB, the
Cs-ACN and the ACN models, the calculated C-alpha rmsd’s
are: ACN vs 1SCB (25), 0.331 Å; Cs-ACN vs ACN 0.188 Å;
Cs-ACN vs 1SCB, 0.332 Å. Considering that the average
Cruickshank DPI or coordinate error50 of these structures at 2.3
Å resolution is ∼0.39 Å we conclude that the peptide conforma-
tion of the backbone conformation is essentially identical in all
3 models.
There are a few changes in side-chain conformations between
Cs-ACN and ACN, where, overall, their functional groups had
moved between 2 and 4.4 Å. One move, of the carboxyl group
of Asp-172 away from Cl-4, has a clear link to counterion
interaction.
Ca2+ and Na+ Ion Sites. Previous subtilisin crystal structures
reported between two to three Ca2+ binding sites (pdb codes:
1CSE,48 1C3L,49 1SCA,25 1SCB,25 1SCN25 and 2SEC51). By
comparing the native and the Cs-soaked anomalous difference
Fourier maps, it is possible to confirm only one genuine Ca2+
binding site, where strong peaks can be seen in the anomalous
difference Fourier maps for both Cs-ACN and ACN. This site
is modeled as a Ca2+ in all the previous structures. The second
Counterion Total Charge. It is interesting to consider the total
counterion charge found in the Cs-ACN model. The occupancies
of the 11 Cs sites sum to a total of 2.90 positive charges, to
which must be added the 3+ of one Ca and one Na. The 8
chloride sites have a summed occupancy of 4.65. In the
acetonitrile soaked crystals it is most likely that there are no
freely mobile counterions that are missed by the anomalous
dispersion measurements. Hence, we estimate a net charge on
the polypeptide of -1.25. From the amino acid composition of
subtilisin Carlsberg, the ExPASy server calculates an expected
isoelectric point of 6.57, close to the pH 5.6 used for crystal-
lization. However, the net charge on the protein molecules in
the crystal may be somewhat different after soaking with Cs-
solution and washing with acetonitrile, reflecting preferential
counterion association in the crystal. Note that the experimental
pI of subtilisin Carlsberg is 9.4 (Sigma reference standard, http://
www.sigmaaldrich.com/sigma/product%20information%20sheet/
p5380pis.pdf), which presumably reflects one or two bound Ca2+
that comigrate with the protein in electrophoresis.
(45) Nayal, M.; Di Cera, E. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 817–
821.
(46) Margolin, A. L.; Navia, M. A. Angew. Chem., Int. Ed. 2001, 40, 2205–
2222.
Acetonitrile Molecule Positions. The structure 1SCB25 in-
cludes 12 acetonitrile molecules bound to subtilisin, while 5
acetonitriles have been identified in each of our own structures
(Table 2). Relatively small molecules like acetonitrile often
cannot be identified with certainty in medium resolution
(2.3-2.4 Å) structures like these. We used the following
approach, similar to guidelines already reported:25 (a) 2Fo-Fc
maps at one sigma level were searched for oblate shapes of
2.6-2.7 Å length (overall length of an acetonitrile molecule);
(b) geometric and environmental constraints like proximity and
type of surrounding residues were considered; (c) occupancy
(47) Roy, J. J.; Abraham, T. E. Chem. ReV. 2004, 104, 3705–3721.
(48) Noritomi, H. A.; Sasanuma, A. S.; Kato, S. K.; Nagahama, K.
Biochemical Eng. J. 2007, 33, 228–231.
(49) Rajan, A.; Abraham, T. E. Bioprocess. Biosyst. Eng. 2008, 31 (2),
87–94.
(50) Cruickshank, D. W. J. Acta Crystallogr. 1999, D55, 583–601.
(51) McPhalen, C. A.; Schnebli, H. P.; James, M. N. FEBS Lett. 1985,
188, 55–58.
(52) Prange, T.; Schiltz, M.; Pernot, L.; Colloc’h, N.; Longhi, S.; Bourguet,
W.; Fourme, R. Proteins 1998, 30, 61–73.
(53) McPhalen, C. A.; James, M. N. Biochemistry 1988, 27, 6582–6598.
(54) Steinmetz, A. C.; Demuth, H. U.; Ringe, D. Biochemistry 1994, 33,
10535–10544.
9
2298 J. AM. CHEM. SOC. VOL. 132, NO. 7, 2010

Subtilisin Enzymatic Action in Acetonitrile
A R T I C L E S
Table 2. Occupation of Acetonitrile Sites in the 3 Modelsa
tials at the protein surface. In general there is an increase in
the surface with weak positive potential. Figure 4 shows the
electrostatic potential for the ACN and Cs-ACN models. A
patch of weakly negative surface near the active site His-64
in the ACN model, becomes positive after taking into account
counterions in the region (in particular Cs-3). It is also evident
that the presence of negative or weakly negative patches on
the unliganded protein model is not a sufficient condition
for attracting positive counterions charges directly. It is
evident that charge neutralization does not happen locally
and directly, but rather with an averaged charge network
spread over the whole surface.
Surface electrostatic potential maps and molecular dynamics
are commonly used in discussing protein function. They are
normally calculated on the basis of charges assigned to protein
atoms, and perhaps the influence of associated waters. However,
if there are actually associated counterions, these will have a
major effect on electrostatic potentials. This should be taken
into account if correct understanding of protein behavior in non-
aqueous media is to be achieved.
Counterions will also affect protein interactions with hydration
water. As shown above, Cs+ and Cl- ions find and exploit
coordination sites not suitable for water molecules, like the
cation-π ones, or short H bonds to anions. These ions, in turn,
create new coordination opportunities for water molecules,
thereby increasing the radius of the water binding shell around
the protein. In Figure 3c and d, it is possible to see that the Cs+
soaked protein shows a small increase (0.03%) in surface (in
blue) when compared to the native one (in white). In particular,
the region around the active site is affected.
Effect of Counterions on Catalytic Activity. As shown above,
the type of counter-cations present has a significant effect on
the catalytic activity of subtilisin crystals in acetonitrile. The
crystallographic evidence points to the mechanism behind
thissthe counterions do indeed become closely associated with
the protein in the nonaqueous medium. The clearly identified
counterion sites not far from the active site are the most obvious
candidates for the main effect. They might interfere with
catalytic turnover either by electrostatic or steric effects, and
these may influence approach of the substrate or the catalytic
reaction.
It is less obvious why the rate should be increased when Cs+
ions are bound. At first sight, their larger size would be more
likely to interfere sterically with substrate binding, while their
electrostatic effect would be similar to Na+. It is possible that
the large size of the Cs+ ions forces them to lie rather further
away from the active site, and hence actually interfere less. As
noted, although Na+ ions were not identified close to the active
site in either our ACN model or the previously published 1SCB,
the crystallographic data were not sufficient to distinguish the
small cation from water. A link with countercation size is
supported by the catalytic activity trend of crystals soaked with
KCl and choline-Cl (Table 1).
It should be noted that acetonitrile has a relatively high
dielectric (37.5) for an organic solvent. Hence, the driving force
for close association of charged groups and counterions will
not be as strong as in less polar solvents (although more than
twice that in aqueous solution). The thermodynamics of separat-
ing balancing charges will also be influenced by the effective
dielectric action of the protein structure itself.
1SCB
ACN
Cs-ACN
CCN-404
CCN-405
CCN-406
CCN-407
CCN-408
CCN-409
CCN-410
CCN-411
CCN-413
CCN-414
CCN-415
CCN-416
W-383
Empty
CCN-1b
CCN-2c
Empty
CCN-3d
CCN-5
CCN-6e
Empty
W-133
Empty
Empty
Empty
W-107
W-99
CCN-1
CCN-2
Empty
CCN-3d
Cs-8 and W-113
NH2’s of Arg-145
W-69
CCN-4f
Empty
Empty
Empty
CCN-5
a ACN is the usual abbreviation for acetonitrile in enzymatic studies and
CCN is the conventional PDB id. Where there are significant differences in
the positions of the acetonitrile molecules (by more than about 1 Å), this is
indicated by footnotes. The electron density offers essentially no
information about the direction of the linear acetonitrile molecule, and
suggestions can only be made on the basis of putative interactions with
other atoms. Hence reversal of assigned direction is probably not so
significant. Refinement of some of the acetonitrile molecules in our
structures led to significant deviations from linearity (with energetic cost),
e.g. CCN-5 in model ACN. This may be an indication that the electron
density is actually due to waters or counterions. b Reversed direction.
c Moved about 2 Å in direction of CH3. d Reversed direction and tilted (45°
in ACN). e Tilted by 60° and moved 1.4 Å. f N in same place, but tilted at 90°.
was estimated on the basis of B-factors of the acetonitrile
molecule (two carbons and one nitrogen) and their respective
surrounding atoms, and no fractional occupancies. With these
restraints reasonably reliable identification is possible, but the
number of small molecules identified may well increase with
the quality of the data as well as the resolution.
As Table 2 shows, three acetonitrile molecules are conserved
across all 3 models, while three more are present in 1SCB and
one of our models. The fully conserved sites include CCN-405
and CCN-406 (numbering scheme of 1SCB) which are present
near the active site, with the former probably accepting a
hydrogen bond from the catalytic Ser-221 hydroxyl.
CCN-413 (1SCB) and CCN-4 of model Cs-ACN have
nitrogens in similar positions, but with the rest of the molecule
in very different orientations. In 1SCB, there is a plausible
hydrogen bond accepted from the backbone amide of Ala-52
(2.49 Å between nitrogens). Given that W-108 of Cs-ACN is
in a similar position to the CH3 of CCN-413 in 1SCB, it might
be thought that the axis of the acetonitrile in Cs-ACN could be
turned to match 1SCB. However, this interpretation did not fit
the electron density as well. An interesting case is CCN-409
(1SCB), which agrees with CCN-5 (ACN), but where Cs-8 and
W-113 are found in Cs-ACN (although not exactly in the same
line as the CCN axis). This may be a real difference in site
occupancy, but one can also speculate that a Na+ ion and a
water in the first two structures could have been modeled instead
as an acetonitrile.
Sites at which organic solvent molecules are located may give
clues about potential binding regions for natural effectors or
designed ligands such as drug leads, as already shown in other
studies.13 The comparison above shows that careful attention to
site identification is important if this approach is to be viable.
Correct identification of counterion sites will have important
implications in understanding organic ligand binding in soaked
protein crystals.
Implications of the Presence of Counterions on the Electro-
static Potential Maps. Taking into account the counterions
leads to substantial effects on calculated electrostatic poten-
Alternatively, it is possible to suggest a mechanism of
electrostatic redistribution, in which polarization of water
molecules or amino acid side chains cause effective spreading
9
J. AM. CHEM. SOC. VOL. 132, NO. 7, 2010 2299

A R T I C L E S
Cianci et al.
Figure 4. Change of electrostatic potential of the protein surface. Red areas have a negative potential, blue areas a positive potential, white areas are neutral.
(a) subtilisin soaked in acetonitrile only, without Cs salt addition; (b) view at 180°; (c) subtilisin soaked in acetonitrile after Cs salt addition; (d) view at
180°.
of the counterion charge. The redistribution of the charge would
be greater with ions like Na+ and K+ for which the water
coordination sphere is more strongly polarized, while the
polarization, and hence the redistribution of the charge, would
be weaker for Cs+ and choline ions. In a low dielectric medium
a stronger redistribution of a positive charge for the smaller
ions might partially compensate the overall negative charge of
the catalytic triad thus reducing the enzymatic catalytic activity,
as observed.
has been concerned with the nonphysiological situation of a
protein surrounded by acetonitrile, similar effects could also
apply in low-dielectric environments like lipid bilayers within
the biological cell.
Acknowledgment. M.C. is especially grateful to European
Molecular Biology Laboratory (EMBL) during which time he
has performed the analysis and prepared the manuscript. We
thank the anonymous referees for their valuable comments. Data
collection was conducted by M.C. when employed at STFC
Daresbury Laboratory as North West Structural Genomics
(NWSGC) beamline scientist, on BBSRC grant (719/B15474)
to PI Professor S.S. Hasnain with Co-I Prof. J.R. Helliwell.
NWSGC MAD10 beamline at SRS (UK) was funded by BBSRC
grant (719/B15474) and an NWDA project award (N0002170).
J.R.H. is especially grateful to STFC Daresbury Laboratory for
the provision of beamtime.
Conclusions
The enzymatic activity of subtilisin crystals in acetonitrile
is sensitive to the type of counterions present before transfer
to the organic solvent. Larger cations increase the enzyme
activity. Anomalous dispersion from Cs+ and Cl- counterions
at longer X-ray wavelength allows several of their sites to
be much more confidently identified in protein crystals. This
confirms that subtilisin Carlsberg in acetonitrile has a large
number of associated ions. Correct identification of counterion
sites appears to be important for a proper understanding of
enzyme activity in nonaqueous media. Moreover, the knowl-
edge of counterion sites allows for a better description of
the electrostatic interactions of proteins. Although this study
Supporting Information Available: Supplementary materials
1-3 (tables). This material is available free of charge via the
Internet at http://pubs.acs.org.
JA908703C
9
2300 J. AM. CHEM. SOC. VOL. 132, NO. 7, 2010