J. Mol. Biol.

(2008) 379, 545553

doi:10.1016/j.jmb.2008.04.004

Available online at www.sciencedirect.com

Physical Basis of Structural and Catalytic Zn-Binding

Sites in Proteins
Yu-ming Lee and Carmay Lim
Institute of Biomedical Sciences,
Academia Sinica, Taipei 115,
Taiwan
Department of Chemistry,
National Tsing Hua University,
Hsinchu 300, Taiwan
Received 8 January 2008;
received in revised form
30 March 2008;
accepted 1 April 2008
Available online
8 April 2008

Zn2+, an element that is essential to all life forms, can play a catalytic or a
solely structural role. Previous works have shown that Zn2+ binds preferentially to water molecules and His in catalytic sites, but to Cys instructural
sites, but the molecular basis for the observed ligand preference is unclear.
Here, we show that the different Zn2+ roles are also reflected in the different
bond distances to Zn2+ in structural and catalytic sites. We reveal the
physical basis for the observed differences between structural and catalytic
Zn sites: In most catalytic sites, water is found bound to Zn2+ as it transfers the
least charge to Zn2+ and is less bulky compared to the protein ligands,
enabling Zn2+ to serve as a Lewis acid in catalysis. In most structural sites,
however, 2 Cys are found bound to Zn2+, as Cys transfers the most
charge to Zn2+ and reduces the Zn charge to such an extent that Zn2+ can no
longer act as a Lewis acid; furthermore, steric repulsion among the bulky Cys
(S) prevents Zn2+ from accommodating another ligand. Based on the
observed ligand preference and Znligand distance differences between
structural and catalytic Zn sites, we present a simple method for distinguishing the two types of sites and for verifying the catalytic role of Zn2+.
Finally, we discuss how the physical bases revealed aid in designing potential drug molecules that target Zn proteins.
2008 Elsevier Ltd. All rights reserved

Edited by D. Case

Keywords: catalytic sites; structural sites; zinc; Zn proteins; bondvalence

sum

Introduction
One of the most important trace metal ions in living
organisms is Zn2+, an essential cofactor in many metabolic enzymes and regulatory proteins.17 Zn2+ can
play a catalytic role (catZn) or a solely structural role
(strZn) in proteins. In catZn sites, Zn2+ can act as a
Lewis acid (i.e., an electron-pair acceptor or an emptyorbital donor) and facilitate Zn-bound water to ionize
to a nucleophilic hydroxide in enzymes such as
*Corresponding author. Institute of Biomedical Sciences,
Academia Sinica, Taipei 115, Taiwan. E-mail address:
carmay@gate.sinica.edu.tw.
Abbreviations used: catZn, catalytic Zn; strZn,
structural Zn; HCA2, human carbonic anhydrase II; 3D,
three-dimensional; PDB, Protein Data Bank; ColE7, colicin
E7; ColE9, colicin E9; DNase, deoxyribonuclease; CSD,
Cambridge Structure Database; CN, coordination
number; BVS, bondvalence sum; HIV, human
immunodeficiency virus; IE, ionization energy; EA,
electron affinity.

human carbonic anhydrase II (HCA2)8 or to stabilize

the negatively charged intermediate of enzymes such
as carboxypeptidase A.9 In strZn sites, Zn2+ can induce the folding of proteins such as Zn fingers or
stabilize the protein structure.6
Several features distinguishing catZn and strZn sites
have been revealed from statistical analyses of these
sites in three-dimensional (3D) structures and from
experimental redesign of Zn sites. In catZn sites, Zn2+
is found bound usually to water and preferentially to
His, followed by Glu, Asp, and then Cys.1,3,8,1012 In
str
Zn sites, however, Zn2+ is found tetrahedrally coordinated to amino acid residues and bound preferentially to Cys rather than to His.11,12 A proposed
feature distinguishing catZn and strZn sites is the presence of activated water or hydroxide in catZn sites,
but not in strZn sites.1,6 However, certain strZn sites
contain a typical catZn core (e.g., human interferon-
[Protein Data Bank (PDB) entry 1au1] contains a [Zn
(His)3H2O]2+ core characteristic of enzymes such as
HCA2).13 Furthermore, some typical Zn-finger sites
can be catalytically active (e.g., the Escherichia coli
ADA enzyme14 contains a catalytic [Zn(Cys)4]2 core).

0022-2836/$ - see front matter 2008 Elsevier Ltd. All rights reserved

Structural Versus Catalytic Zn Sites in Proteins

546
This suggests that the protein environment, in addition to Zn ligands, determines Zn's role in certain
proteins sharing the same metal complex.
The question of whether Zn2+ plays primarily a
structural or a catalytic role may still be uncertain
even when the respective protein structure is known.
For instance, it was unclear whether Zn2+ plays a
purely structural and/or a catalytic role in colicin E7
(ColE7) and colicin E9 (ColE9), which are protein
toxins made by bacteria to kill other bacteria via an
endonuclease domain.15 The X-ray structures of
ColE7 and ColE9 deoxyribonuclease (DNase) domains (1m08 and 1fsj) show Zn2+ tetrahedrally coordinated to three His residues and a phosphate
oxygen molecule. The Zn2+ in ColE9 has been hypothesized to play a structural role based on experimental results showing the Zn-bound ColE9 DNase
domain to be more thermally stable than the apoprotein and to be active with Ni2+, Co2+, and Mg2+,
but inactive with Zn2+.16 However, the Zn2+ in
ColE7, which shares a ~ 70% sequence identity with
ColE9, has been postulated to play a catalytic role
based on experimental results showing the Zn2+ to
be dispensable for DNA binding but essential for
DNA hydrolysis.17
The debatable role of Zn2+ in bacterial endonuclease enzymes, the Zn ligand preference observed in
cat
Zn and strZn sites, and the role of Zn2+ as a Lewis
acid in catZn sites raise several intriguing questions
that, to the best of our knowledge, have not been
addressed in previous works: (1) Are the different
Zn2+ roles reflected in different distances between
Zn2+ and a given ligand (L) atom in the two types of
sites? (2) If so, what is the physical basis for the observed differences in ligand preference and ZnL
distances between catZn and strZn sites (e.g., why do
most strZn sites strongly prefer Cys to other amino
acid ligands and how does this affect the metal's
properties/function?)? (3) Can the feature(s) governing the different Zn2+ roles be exploited to distinguish
cat
Zn from strZn sites, especially in 3D structures
where the Zn-bound water position is unknown/

uncertain? (4) Can they resolve the debatable role of

Zn2+ in the ColE7/E9 DNase domain?
To address these questions, we have created a new
set of nonredundant strZn and catZn sites by combining previous datasets1012,18 derived from the
PDB19 structures. We computed the mean ZnL distances (ZnL) in these sites and compared them
with those found in Zn complexes from the Cambridge Structure Database (CSD).20 A significant
difference in the ZnL distance between strZn and
cat
Zn sites was found. To provide a physical basis for
the observed Znligand preference and ZnL distance differences between strZn and catZn sites, we
employed global and local reactivity descriptors to
elucidate the effect of the Zn ligands and the relative
Znligand distances on the metal's Lewis acidity.
The Znligand preference and ZnL distance differences were then used to distinguish catZn sites from
structural ones and to predict Zn's role in the ColE7/
E9 DNase domain. The physical makeup and distinction between catZn and strZn sites revealed herein aid
in elucidating Zn's role and in designing Zn sites or
inhibitors of target Zn proteins.

Results
ZnL distances in

str

Zn and

cat

Zn sites

To evaluate whether the ZnL distance in strZn

sites differs from that in catZn sites, we computed the
ZnL distances in the two types of sites in Supplementary Tables S1 and S2, and compared them
with those found in CSD structures for a given metal
coordination number (CN). As the ZnL distance
depends on the metal's CN, they were grouped
according to whether Zn2+ was 4-coordinated (T4), 5coordinated (T5), or 6-coordinated (T6) in Table 1. A
key difference between catZn and strZn sites is the
metal CN, which seems to be more variable in the
former than in the latter. In catZn sites, 4-coordinated,
5-coordinated, and 6-coordinated Zn2+ are present,

Table 1. Average distances () between Zn2+ and its coordinating atoms in PDB structures, as compared to those in CSD
structures and BVS distances
Zn type
T4 CSD
T4 BVS
T4 Strc
T4 Cat
T5 CSD
T5 BVS
T5 Catg
T6 CSD
T6 BVS
T6 Cat

ZnS
2.32 0.05
2.35
2.33 0.09
2.38 0.18
2.42 0.13
2.43

ZnO

ZnN
(246)
(137)
(16)
(46)
(2)

2.04 0.04
2.03
2.05 0.16
2.16 0.12
2.10 0.08
2.11
2.08 0.07
2.16 0.05
2.18
2.10 0.10

(603)
(36)
(64)
(38)
(38)
(409)
(7)

1.97 0.03
1.96
2.09 0.17
2.09 0.04
2.04
2.16 0.19
2.10 0.06
2.11
2.19 0.17

ZnOwater
(109)
(2)
(36)
(12)
(46)
(103)
(11)

2.00 0.04
1.96
2.28 0.23
2.01 0.04
2.04
2.24 0.25
2.11 0.06
2.11
2.27 0.11

(21)
(1)
(27)
(5)
(24)
(274)
(6)

The number after the distance is the standard deviation, and the number in parentheses is the number of observations. When the number
of observations is b 5, the mean distance is not given as it is statistically unreliable.
CSD: distances derived from CSD structures (Kuppuraj and Lim, in preparation); BVS: distances calculated using Eq. 3 (see Materials
and Methods); Strc: distances derived from the strZn sites in Supplementary Table S1; Cat: distances derived from the catZn sites in
Supplementary Table S2.

Structural Versus Catalytic Zn Sites in Proteins

reflecting perhaps the need for Zn2+ to adopt different CNs during an enzymatic reaction. In strZn sites,
however, Zn2+ is tetracoordinated: no T5 site and one
T6 site (1enr) was found. Hence, only the ZnL
distances in T4 strZn and catZn sites were compared.
The different roles of Zn2+ are reflected not only in
the different metal CNs but also in the different ZnL
distances in strZn and catZn sites. In strZn sites, the
ZnN and ZnS distances (2.05 and 2.33) are
similar to those found in CSD structures (2.04 and
2.32). In catZn sites, however, the ZnN distance
(2.16) is longer than that in strZn sites; the ZnO
and ZnOwater distances (which cannot be compared with those in strZn sites due to lack of statistics)
are longer than those in the T4 CSD structures. The
longer ZnOwater distance in catZn sites relative to
CSD structures has been attributed to hydrogenbonding interactions with the Zn-bound water in the
enzyme, pulling it away from Zn2+.10
Effect of the Zn ligands on Zns Lewis acidity
To account for the observed preference of Cys to His
in strZn sites and vice versa in catZn sites, the electronegativity () of Zn complexes modeling common
str
Zn and catZn cores (Fig. 1) was evaluated (see Materials and Methods). The values in Table 2 provide
a measure of Zn's ability to act as a Lewis acid, since
only Zn2+ can accept electrons in the metal complex.
As the number of Zn-bound Cys decreases, the of
the complex increases (from 2.50 to 6.76eV), implying that electrons will flow more easily from an
electron donor (e.g., substrate) to catZn cores than to
structural ones. Thus, Zn's Lewis acidity can be modulated by its bound ligands, in particular by the
number of Zn-bound Cys, and Zn2+ would be a
better Lewis acid in catZn sites than in strZn sites.
To provide a physical basis for the greater values
of catZn sites, as compared to strZn sites, the net charge
transferred by all the ligands to Zn2+ was computed
from the Zn's charge. As the of the complex
increases, the net charge transferred by the ligands to
Zn2+ decreases (Table 2). Comparison of the charge
transfer between [Zn(CH3S)4]2 and [Zn(CH3S)3(ImH)],
between [Zn(CH3S)3(ImH)] and [Zn(CH3S)2(ImH)2]0,
or between [Zn(CH3S)2(ImH)2]0 and [Zn(OH)(ImH)3]+
shows that a negatively charged ligand transfers more
charge to Zn2+ than a neutral one. Notably, even
when the net charge of the complex is the same, the
charge transfer in the [Zn(CH3S)2(ImH)2]0 model
structural core (0.68e) is greater than that in the [Zn
(OH)(HCOO)(ImH)2]0 model catalytic core (0.43e),
implying that Cys transfers more charge to Zn2+
than Asp /Glu or OH. Indeed, Cys has been
shown to transfer the most charge to Zn2+, while
water has been shown to transfer the least charge.21
Effect of the ZnL changes on Zns Lewis acidity
To evaluate whether the observed differences in
the ZnL distances affect Zn's Lewis acidity, the
+
condensed Fukui functions at Zn2+ (f Zn
), which
reflect Zn's electron-acceptor ability, were computed

547
from the fully optimized geometry of the metal complex and from the constrained geometry of the respective complex, with the ZnL distances fixed to
those in b 2.5- PDB and CSD structures, as described in Materials and Methods (see Supplementary
Table S3). The PDB-constrained geometries reflect
the effect of the protein matrix on the isolated Zn
+
values derived from
complex structures. The f Zn
various conformations of the same metal complex
and metal CN were compared: Those derived from
the PDB-constrained T4 geometries were compared
+
derived from the fully optimized geowith the f Zn
metry (whose ZnL distances were similar to those in
the CSD structures). Those derived from the PDBconstrained T5 geometries were compared with the
+
derived from the respective CSD-constrained T5
f Zn
geometry, as full optimization would yield a tetra+
values for the
hedral complex. The relative f Zn
str
cat
common Zn and Zn cores are depicted in Fig. 2,
along with the sum of the deviations of the ZnL
distances in the PDB site (diPDB ) from those in the
fully optimized or CSD-constrained reference structure (diREF); (diPDB diREF) values that are b 1 are in
gray, while those that are 1 are in black.
The net bond length changes (diPDB diREF) in strZn
sites containing 2 Cys, where Zn2+ is already a
poor Lewis acid, negligibly affect Zn's electron-acceptor ability, but those in catZn sites alter Zn's Lewis
acidity. In strZn sites, the net bond length changes are
+
changes are
as large as 0.53, but the respective f Zn
+
negligible (f Zn 0.009). This trend is found in the
three most common strZn sites, regardless of the metal
complex's net charge. Thus, relative to the isolated Zn
complex, the different bond lengths in the strZn sites
in protein structures negligibly affect Zn's Lewis
acidity. In contrast, in catZn cores, the net bond length
+
values are at
changes are 0.21.6, while the f Zn
least an order of magnitude greater than those in strZn
sites, implying that the ZnL distance changes can
modulate Zn's Lewis acidity.
Distinguishing strZn and
ligand preference

cat

Zn sites based on

As Zn-bound water may not be seen or resolved

accurately in the 3D structure, our results suggest
another way of distinguishing strZn and catZn sites
depending on the number of Zn-bound Cys: When
Zn2+ is bound to 2 Cys, it is a poor electron acceptor and plays a structural role, but when it is
bound to 1 Cys, it can serve as a Lewis acid,
enabling it to play a catalytic role. For the set of
nonredundant Zn proteins studied, 41 of the 45 strZn
sites have 2 Cys, and 58 of the 64 catZn sites have
1 Cys (Fig. 3a). Thus, counting the number of Znbound Cys predicts Zn's role correctly in ~91% of
the 109 Zn sites. This accuracy is comparable to that
obtained by counting the number of Zn-bound water
ligands, which predicts the correct role of Zn in ~90%
of the Zn sites (Fig. 3b).
In well-resolved X-ray structures, we propose
counting both the number of Zn-bound Cys and
the number of water ligands to predict the role of

548

Structural Versus Catalytic Zn Sites in Proteins

Fig. 1. The S-VWN/631+G(d) fully optimized structures modeling common structural and catalytic Zn cores.

Zn2+. When Zn2+ is tetracoordinated to N 1 Cys , it is

predicted to play a structural role if there is no Znbound water, but it is predicted to play a catalytic
role if there is Zn-bound water. A catZn is also
predicted if it is bound to 1 Cys . Based on these

criteria, Zn's role is correctly predicted in ~ 95% of

the Zn sites (Fig. 4). Four strZn sites lacking Cys are
wrongly predicted to be catalytic: the Zn-DHHH site
in stromelysin (1hy7) and ferredoxin (1xer), the ZnDDEHww site in concanavalin A (1enr), and the Zn-

Structural Versus Catalytic Zn Sites in Proteins

549

Table 2. Electronegativity () values of common Zn cores

Zn core

Model Zn complex
2

[Zn Cys4]
[Zn Cys3 His]
[Zn Cys2 His2]0
[Zn Cys Asp/Glu
His H2O]0
[Zn OH Asp/Glu
His2]0
[Zn OH His3] +

[Zn (CH3S)4]
[Zn (CH3S)3 ImH]
[Zn (CH3S)2 (ImH)2]0
[Zn CH3S HCOO
ImH H2O]0
[Zn OH HCOO (ImH)2]0
[Zn OH (ImH)3] +

(eV)a

CT (e)b

2.50
0.48
3.29
3.64

0.81
0.76
0.68
0.58

4.11

0.43

6.76

0.40

Evaluated at the MP2/631+G(d)S-VWN/631+G(d) level

using Eq. 1.
b
Net charge transfer (CT) from all the ligands to Zn2 +, which is
equal to 2 minus the S-VWN/631+G(d) NBO charge on Zn2 +.

HHHw site in human interferon- (1au1). Thus,

although Zn2+ could still act as a Lewis acid in these
four proteins, the protein matrix probably prevents
Zn2+ from playing a catalytic role. Only the catalytic
Zn-CCDH site in -carbonic anhydrase (1i6p) was
wrongly predicted to be structural.
Identifying
distances

cat

Zn sites based on the ZnL

Even when the Zn ligands are identical, Zn2+ can

play a catalytic role in one protein (e.g., HCA2; 2cba)
but a structural role in another (e.g., human interferon-; 1au1), implying that the protein matrix,
rather than the Zn ligands, dictates its specific role.
Since the ZnL distances in the X-ray structures
reflect the effect of the protein matrix (e.g., interactions between the Zn ligands and side-chain or backbone groups in the second shell), we examined the
possibility of using them to identify Zn's role by
computing the bondvalence sum (BVS) values (see
Materials and Methods). The ZnL distances in
cat
Zn sites are longer than those in strZn sites and
CSD structures, which are similar to the respective
ideal BVS distances (Table 1); the longer distances
allow space for binding another ligand during the
enzymatic reaction. According to Eq. 3 below, ZnL
distances longer than the respective BVS distances
would yield a BVS value of b2, suggesting that a
cat
Zn site would generally be characterized by a BVS
value of b2.
However, the ZnL distances depend on the accuracy of the PDB structure. Thus, we first examined
the sensitivity of the BVS to the resolution of the PDB
structure by computing the catZn site's BVS values
in the 1.50- to 2.20- X-ray structures of HCA2. Regardless of the X-ray structure resolution and the
ZnL distance variations, the BVS values of the
catalytic [Zn(His)3H2O]2+ cores are all b 2 (Supplementary Table S4).
To further verify that catZn sites are characterized
by BVS b 2, BVS values were computed for all Zn
sites (Supplementary Table S5). The mean BVS of the
cat
Zn sites is 1.61 0.41, whereas that of the strZn sites
is 2.12 0.40. Fifty-eight of the 64 catZn sites (91%)
have BVS values of b 2. Notably, the BVS values of
the Zn sites can predict the correct role of Zn2+
bound to the same set of ligands [e.g., the BVS of the

Zn-HHHw site in human interferon- (1au1) is 2.58,

but it is b 2 in all the other enzymes containing the
same set of Zn ligands (1akl, 1b6z, 1hy7, 1iag, 1lml,
1sml, 1thj, 1znb, and 2cba)]. Furthermore, the BVS of
the Zn-DHHH site in ferredoxin (1xer) containing a
str
Zn is 2.32, but that in endonuclease IV (1q0e) or
elastase (2aps) containing a catZn is b 2. However, six
of the catZn sites have BVS values of N 2, out of which
four (2c6g, 2a7m, 1cg2, and 1qtw) belong to binuclear Zn enzymes. In the 1qtw structure, both ZnN
distances (2.00 and 1.98) are shorter than the respective BVS distance (2.03), while in the other three
binuclear Zn enzymes, the bridging ZnO distance is
shorter than the BVS ZnO distance (1.96), thus
resulting in BVS N 2. The other two wrongly predicted catZn sites are the Zn-CCCw site in 5-aminolaevulinic acid dehydratase (1h7n) and the Zn-EHHH
site in L-fuculose-1-phosphate aldolase (1fua), where
the first step of the reaction mechanism is known to
involve the departure of a Zn ligand.22 This seems
consistent with the N 2 BVS values in these two enzymes, which indicate a crowded Zn2+ coordination
sphere, thus favoring ligand dissociation.
Zns role in ColE7/E9
As the X-ray structures of the ColE7/E9 DNase
domain show Zn2+ bound to three His residues and a
phosphate oxygen molecule rather than to a water
molecule, it was unclear whether the Zn site is
structural or catalytic. Since Zn2+ is not bound to
any Cys, it is predicted, according to Fig. 4, to play a
catalytic role. To further confirm that Zn2+ plays a
catalytic role in the ColE7/E9 DNase domain, the
BVS values for the respective Zn sites were computed.

Fig. 2. Effect of bond length changes on Zn's Lewis

+
acidity, as measured by fZn
, for common strZn and catZn
+
cores with 2 and 1 Cys , respectively. The fZn
calculations were performed at the S-VWN/631+G(d)
level; their values, along with the net bond length changes
dREF
i
i
dPDB
, are listed in Supplementary Table S3.

550

Structural Versus Catalytic Zn Sites in Proteins

Fig. 3. The distribution of strZn

and catZn sites according to the number of (a) Zn-bound Cys or (b) Znbound water. The number of strZn
and catZn sites with 2 or 1 Cys
(a) and 0 or 1 Zn-bound water
molecules (b) is shown above each
histogram.

The resulting values in Table 3 support a catZn in the

ColE7/E9 DNase domain, as they are b 2, except for
the BVS derived from the 1m08 chain A structure,
which is 2.48. The latter is due to the unusually short
ZnN(His544) distance in the 1m08 chain A structure
(1.86), as compared to the ZnN distances in the
other structures (1m08/B: 2.05; 7cei: 2.06; 1mz8:
1.982.09; 1fsj: 2.062.11). Notably, the b 2 BVS
values in the ColE7/E9 X-ray structures imply a
relatively expanded Zn-chelating sphere that could
accommodate an extra ligand acting as a general acid
in the DNA hydrolysis reaction.

Discussion
Whereas the different roles of Zn2+ have been found
in previous works1,1012 to be reflected in the different

Znligand preferences, they are found herein to be

also reflected in the different ZnL distances and metal
CN preferences in strZn and catZn sites: A structural
Zn2+ strongly prefers to be tetracoordinated with
ZnL distances similar to those in CSD structures,
whereas a catalytic Zn2+ could adopt a CN of 4, 5, or
6, with ZnL distances generally longer than those in
the respective CSD structures.
More importantly, our results provide a physical
basis for the observed ligand preference and ZnL
distance differences between strZn and catZn sites.
Most strZn sites are Cys-rich, with a CN no larger
than 4, as Cys transfers more charge to Zn2+ than the
other amino acid ligands and water/OH,21,23 thereby reducing the Zn's charge and electron-accepting
ability; the steric repulsion among the bulky Cys(S)
further inhibits Zn2+ from binding a fifth ligand. In
this way, the Cys ligands protect the Zn2+ from un-

Fig. 4. Distinguishing strZn and catZn sites based on ligand preference. (a) Flowchart showing how catZn sites can be
distinguished from structural ones by counting the number of Zn-bound Cys and water ligands. (b) The distribution of
predicted strZn and catZn sites.

Structural Versus Catalytic Zn Sites in Proteins

551

Table 3. BVS values of the Zn sites in X-ray structures of

the ColE9/E7 DNase domain

domain, which is found to play a catalytic role, in

accordance with the absence of Zn-bound Cys.

PDB code/
chain

Implications in drug design

1fsj/B
1fsj/C
1fsj/D
1fsj/E
1m08/A
1m08/B
7cei
1mz8/B
1mz8/D
a
b

Molecule(s)
ColE9

ZnN ZnN ZnN ZnOa BVS

2.08
2.06
2.08
2.11
ColE7
1.86
2.05
ColE7 + Im7 2.06
ColE7 + Im7 1.98
2.09

2.08
2.09
2.17
2.14
2.10
2.08
1.92
2.22
2.10

2.07
2.08
2.08
2.07
2.03
2.09
2.57
2.07
2.15

1.94
2.00
1.95
1.99
1.79
1.90
2.16b
2.04
2.07

1.83
1.75
1.71
1.67
2.48
1.90
1.52
1.71
1.56

The distance between Zn2 + and phosphate oxygen.

The distance between Zn2 + and water oxygen.

wanted reactions. In contrast, most catZn sites contain

Zn-bound water, which transfers less charge to Zn2+
than amino acid ligands.21,23 Compared to the strZn
sites, the preference for water and His in catZn sites
results in a reduced net charge transfer from the Zn
ligands and a higher positive charge on the metal,
making Zn2+ a good Lewis acid and enabling it to
play a catalytic role. Thus, the dramatic charge transfer and size differences between Cys and water,
which affect Zn's electron-acceptor ability, provide a
rationale for the observed preference for water in
cat
Zn sites and for Cys in strZn sites.
The principles revealed herein not only help to
explain why the nonredox Zn2+ can play a dual role in
proteins and how proteins can tune the activity of its
native Zn2+ cofactor but also help to explain why no
natural Zn-finger domains with Zn-CHHH or ZnHHHH cores exist and why Cd2+, which is often used
as a substitute for the spectroscopically silent Zn2+,
rarely plays a catalytic role similarly to Zn2+ in
proteins. The latter is probably because Cd2+ is a soft
metal ion that prefers soft polarizable Cys to harder
water or His ligands. Binding of 2 Cys to Cd2+
would prevent it from serving as a Lewis acid in
enzyme catalysis. Although Zn-finger proteins contain
structural Zn-CCCC, Zn-CCCH, and Zn-CCHH sites,
they do not possess Zn-CHHH or Zn-HHHH cores
because if Zn2+ were to bind to 1 Cys, it could act
as a Lewis acid and bind another ligand. In other
words, Zn-finger proteins employ 2 Cys to bind
Zn2+ to prevent a coordination geometry change.
As the different roles of Zn2+ are reflected in the
different compositions of the Zn ligands and the different ZnL distances in strZn and catZn sites, these
features have been exploited to distinguish between
the two types of sites. We propose counting the
number of Zn-bound Cys and water ligands as in
Fig. 4, or only the former if the Zn-bound water is
absent or ambiguous, to differentiate catZn cores
from strZn sites. The BVS method, which has conventionally been used to analyze the compatibility of
a given set of measured metalligand distances with
a certain metal oxidation state,2426 is used here in a
novel way to support a catalytic role for Zn2+. Notably, it has been used to predict the different roles of
Zn2+ bound to the same set of ligands and to resolve
the debatable role of Zn2+ in the ColE7/E9 DNase

The physical bases provide valuable knowledge in

designing potential drug molecules that target Zn
proteins: We propose that the target of drug design is
the metal ion for Zn sites with 1 Cys, but the Zn
ligands for sites containing 2 Cys. This is because
in Zn sites with 1 Cys, Zn2+ is a good Lewis acid
and can accommodate an extra ligand. Indeed, for
several Zn enzymes containing Zn sites with 1
Cys , inhibitors have been found to bind directly to
the metal ion [e.g., inhibitors have been found to
bind the Zn2+ monodentately in HCA2 (1oq5, 1kwr,
2foq, and 2nng), carboxypeptidase A (1cbx, 1cps,
1iy7, and 4cpa), metallo--lactamase (1hlk, 1m2x,
1vgn, and 2doo), and peptide deformylase (1bsk),
and bidentately in HCA2 (1i9o and 1i9p), metallo-lactamase (1jje), peptide deformylase (1szz), and
alkaline protease (1jiw)]. On the other hand, in Zn
sites with 2 Cys, Zn2+ is a poor Lewis acid and
would not make a good drug target, as it cannot bind
another ligand tightly. However, Zn-bound thiolates
lacking electrostatic or steric shielding could serve as
drug targets.27 This is supported by the fact that the
Zn-bound thiolates in the two highly conserved Zn
fingers of the human immunodeficiency virus (HIV)
nucleocapsid p7 protein are specific targets of disulfide benzamides. These anti-HIV agents covalently
bind to the Zn-bound thiolates, causing Zn2+ ejection, loss of native viral protein function, and thus
HIV replication, but do not affect the functions of
cellular Zn-finger proteins.

Materials and Methods

Datasets of nonredundant

cat

Zn and

str

Zn sites

Statistical analyses were based on four datasets derived

from previous studies,1012,18 which had surveyed the PDB
for 3D structures of Zn sites in proteins and had categorized
them into strZn and catZn+ in each protein. Our analyses
were restricted to b 3.0- X-ray structures of the wild-type
protein containing Zn2+ bound to 4 native ligands; NMR
structures and sites with Zn2+ bound to b 4 ligands or to
nonnative ligands were excluded. The protein sequences
were aligned using the ClustalW program.28 Those with a
sequence identity of N 30% and the same Zn ligands were
considered to belong to the same family; the best-resolution
structure was chosen to represent the family. When the PDB
structure contained N 1 chain, only one chain (chain A) was
chosen. When this chain contained multi-Zn sites, all
nonredundant Zn sites were included. These criteria resulted
in 79 nonredundant Zn protein families comprising 109
native Zn sites, of which 45 were strZn sites and 64 were
cat
Zn sites (Supplementary Tables S1 and S2).
Definition of first-shell ligands
Analyses of high-resolution (R-factor 0.065) X-ray
structures of small metal complexes in the CSD have shown

Structural Versus Catalytic Zn Sites in Proteins

552
that the mean ZnN, ZnOwater, ZnO, and ZnS distances do not exceed 2.16, 2.11, 2.10, and 2.42, respectively
(Table 1). On the other hand, analyses of b 2.5- X-ray
structures of the same protein (HCA2) have shown that a
given ZnL distance could vary by as much as 0.4. This
uncertainty would be larger for exchanging water ligands.
Thus, to account for the lower resolution of the PDB structures, as compared to the CSD structures, 0.4 was added
to the mean ZnN, ZnO, and ZnS CSD distances, while
0.8 was added to the ZnOwater CSD distance to locate
the first-shell Zn ligands in the PDB structures. For a given
metal CN, if the ZnOwater distance in the PDB structure is
below 1 S.D. of the respective ZnOwater CSD distance
(i.e., less than 1.96, 1.97, and 2.05 for T4, T5, and T6 sites,
respectively), it is treated as a ZnO distance.

probes the reactivity of different sites within a molecule. It

can be defined to reflect Zn's electron-acceptor ability in a
Zn complex by estimating it in terms of the change in the
Zn's charge upon adding an electron:35

fZn
qZn N1qZn N

where qZn(N + 1) and qZn(N) represent the Zn charge in

a molecule with N + 1 and N electrons, respectively. The
f +Zn values in the common Zn cores were estimated from SVWN/631+G(d) natural population analyses 36 of the N +
1 and N electron molecules. Natural population analysis
was chosen over other methods such as MerzKollman
and CHelp/CHelpG schemes because it had been recommended37,38 and successfully applied to Fukui function
calculations39,40 and it does not depend on empirically
adjusted parameters such as van der Waals radii.

Models used
We modeled the most common strZn ([Zn(Cys)4]2, [Zn
(Cys)3His], and [Zn(Cys)2(His)2]0) and catZn ([Zn(Cys)(Asp/
Glu)(His)(H2O)]0, [Zn(OH)(Asp/Glu)(His)2]0, and [Zn(OH)
(His)3]+) cores in proteins (Fig. 1). In the latter two complexes,
the Zn-bound water was assumed to be deprotonated in the
catalytically active state, in accordance with the measured
pKa values of the Zn-bound water in HCA28 and carboxypeptidase A.9 The side chains of Cys, His, and Asp/Glu
were modeled by methyl thiolate (CH3S), imidazole (ImH),
and formate (HCOO), respectively.
Geometry optimization
The model strZn and catZn cores were fully optimized or
constrained optimized by fixing only the ZnL distances to
the values observed in the b 2.5- PDB structures (Supplementary Table S3); the overall fully optimized and constrained optimized geometries are similar on visual
inspection. The S-VWN functional set and the 631+G(d)
basis set had been found adequate for geometry optimization of Zn complexes containing methyl thiolate, imidazole,
and/or formate ligands. Since S-VWN/631+G* reproduces
the experimentally observed ZnL distances in related Zn
complexes better than other functional set/basis set
combinations,2931 it was used to optimize the geometries
of the Zn complexes using the Gaussian 03 program.32
Computing the global reactivity descriptor
The electronegativity was estimated by averaging the
molecule's ionization energy (IE) and electron affinity (EA),33



IE EA
vi
2

with IE approximated by the highest occupied molecular

orbital energy and with EA approximated by the lowest
unoccupied molecular orbital energy.34 The energies were
evaluated at the MP2/631+G(d)S-VWN/631+G(d) level,
which was chosen as it could reproduce the experimental EA
and IE values of CO2 and HCONH2.29 Furthermore, MP2/
631+G(d) calculations yield a value for the free Zn2+
(29.2eV) that is close to the experimental number (28.8eV).33
Computing the local reactivity descriptor f(r)
Unlike , which is a global reactivity descriptor characterizing the molecule as a whole, the Fukui function f(r)

Computing BVS values

The BVS2426 is given by:
X
X
expR0  Ri =b
BVS
si

where si is the valence of a single bond formed by atom i,

Ri is the distance from atom i to the metal ion, R0 is the
corresponding parameter derived from X-ray structures,
and b is a universal constant equal to 0.37.25 In computing BVS values using Eq. 3, the R0 values for ZnS (2.09),
ZnN (1.77), and ZnO (1.70) bonds were obtained
from published tables.25,26 To ensure that the R0 values are
suitable for CNs of 4, 5, and 6, ideal BVS ZnL distances
were calculated for a given metal CN from R0 b ln si,
where si = 0.50, 0.40, and 0.33 for a metal CN of 4, 5, and 6,
respectively. A comparison with the corresponding ZnS,
ZnN, and ZnO distances in the CSD shows that the ideal
BVS distances are close to the respective CSD values for
T4, T5, and T6 geometries (Table 1).

Acknowledgements
We thank H.S. Yuan, T. Dudev, T.W. Chang, I.
Tunell, B. Chen, and G. Kuppuraj for helpful discussions. This work was supported by the National
Science Council, Taiwan (contract no. 95-2113-M001-001).

Supplementary Data
Supplementary data associated with this article
can be found, in the online version, at doi:10.1016/
j.jmb.2008.04.004

References
1. Vallee, B. L. & Auld, D. S. (1990). Active-site zinc
ligands and activated H2O of zinc enzymes. Proc. Natl
Acad. Sci. USA, 87, 220224.
2. Christianson, D. W. (1991). Structural biology of zinc.
Adv. Protein Chem. 42, 281355.
3. Coleman, J. E. (1992). Zn proteins: enzymes, storage
proteins, transcription factors and replication proteins. Annu. Rev. Biochem. 61, 897946.

Structural Versus Catalytic Zn Sites in Proteins

4. Lippard, S. J. & Berg, J. M. (1994). Principles of Bioinorganic Chemistry. University Science Books, Mill
Valley, CA.
5. Lipscomb, W. N. & Strater, N. (1996). Recent advances
in zinc enzymology. Chem. Rev. 96, 23752433.
6. Cox, E. H. & McLendon, G. L. (2000). Zinc-dependent
protein folding. Curr. Opin. Chem. Biol. 4, 162165.
7. Dudev, T. & Lim, C. (2003). Principles governing Mg,
Ca, and Zn binding and selectivity in proteins. Chem.
Rev. 103, 773787.
8. Christianson, D. W. & Cox, J. D. (1999). Catalysis by
metal-activated hydroxide in zinc and manganese
metalloenzymes. Annu. Rev. Biochem. 68, 3357.
9. Christianson, D. W. & Lipscomb, W. N. (1989). Carboxypeptidase A. Acc. Chem. Res. 22, 6269.
10. Alberts, I. L., Nadassy, K. & Wodak, S. J. (1998). Analysis of zinc binding sites in protein crystal structures.
Protein Sci. 7, 17001716.
11. Auld, D. S. (2001). Zinc coordination sphere in biochemical zinc sites. BioMetals, 14, 271313.
12. Dudev, T., Lin, Y. L., Dudev, M. & Lim, C. (2003).
Firstsecond shell interactions in metal binding sites
in proteins: a PDB survey and DFT/CDM calculations. J. Am. Chem. Soc. 125, 31683180.
13. Karpusas, M., Nolte, M., Benton, C. B., Meier, W.,
Lipscomb, W. N. & Goelz, S. (1997). The crystal structure of human interferon beta at 2.2- resolution. Proc.
Natl Acad. Sci. USA, 94, 1181311818.
14. Myers, L. C., Terranova, M. P., Ferentz, A. E., Wagner,
G. & Verdine, G. L. (1993). Repair of DNA methylphosphotriesters through a metalloactivated cysteine
nucleophile. Science, 261, 11641167.
15. James, R., Kleanthous, C. & Moore, G. R. (1996). The
biology of E colicins: paradigms and paradoxes.
Microbiology, 142, 15691580.
16. Pommer, A. J., Cal, S., Keeble, A. H., Walker, D.,
Evans, S. J., Kuhlmann, U. C. et al. (2001). Mechanism
and cleavage specificity of the HNH endonuclease
colicin E9. J. Mol. Biol. 314, 735749.
17. Ku, W. Y., Liu, Y. W., Hsu, Y. C., Liao, C. C., Liang,
P. H., Yuan, H. S. & Chak, K. F. (2002). The zinc ion in
the HNH motif of the endonuclease domain of
colicin E7 is not required for DNA binding but is
essential for DNA hydrolysis. Nucleic Acids Res. 30,
16701678.
18. Yang, T.-Y., Dudev, T. & Lim, C. (2008). Mononuclear
vs. binuclear metal-binding sites: metal binding affinity and selectivity from PDB survey and DFT/CDM
calculations. J. Am. Chem. Soc. 130, 38443852.
19. Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G.,
Bhat, T. N., Weissig, H. et al. (2000). The Protein Data
Bank. Nucleic Acids Res. 28, 235242.
20. Allen, F. H. (2002). The Cambridge structural database: a quarter of a million crystal structures and
rising. Acta Crystallogr. Sect. B, 58, 380388.
21. Garmer, D. R. & Gresh, N. (1994). A comprehensive
energy component analysis of the interaction of hard
and soft dications with biological ligands. J. Am. Chem.
Soc. 116, 35563567.
22. Parkin, G. (2004). Synthetic analogues relevant to the
structure and function of zinc enzymes. Chem. Rev.
104, 699768.

553
23. Dudev, T. & Lim, C. (2001). Metal selectivity in metalloproteins: Zn2+ vs. Mg2+. J. Phys. Chem. B, 105, 44464452.
24. Pauling, L. (1929). The principles determining the
structure of complex ionic crystals. J. Am. Chem. Soc.
51, 10101026.
25. Brown, I. D. & Altermat, D. (1985). Bondvalence
parameters obtained from a systematic analysis of the
inorganic crystal structure database. Acta Crystallogr.
Sect. B, 41, 244247.
26. Brese, N. E. & O'Keeffe, M. (1991). Bondvalence parameters for solids. Acta Crystallogr. Sect. B, 47, 192197.
27. Maynard, A. T. & Covell, D. G. (2001). Reactivity of
zinc finger cores: analysis of protein packing and electrostatic screening. J. Am. Chem. Soc. 123, 10471058.
28. Thompson, J. D., Higgins, D. G. & Gibson, T. J. (1994).
CLUSTAL W: improving the sensitivity of progressive
multiple sequence alignments through sequence
weighting, position specific gap penalties and weight
matrix choice. Nucleic Acids Res. 22, 46734680.
29. Lin, Y.-L., Lee, Y.-M. & Lim, C. (2005). Differential
effects of the ZnHisBkb vs. ZnHis[Asp/Glu] triad
on Zn-core stability and reactivity. J. Am. Chem. Soc.
127, 1133611347.
30. Lin, Y.-L. & Lim, C. (2004). Factors governing the
protonation state of Zn-bound histidine in proteins: a
DFT/CDM study. J. Am. Chem. Soc. 126, 26022612.
31. Dudev, T. & Lim, C. (2002). Factors governing the
protonation state of cysteines in proteins: an ab initio/
CDM study. J. Am. Chem. Soc. 124, 67596766.
32. Frisch, M. J., Trucks, G. W., Schlegel, H. B., Scuseria,
G. E., Robb, M. A., Cheeseman, J. R. et al. (2003).
Gaussian 03, Rev. B.03. Gaussian, Inc., Pittsburgh, PA.
33. Pearson, R. G. (1988). Absolute electronegativity and
hardness: application to inorganic chemistry. Inorg.
Chem. 27, 734740.
34. Koopmans, T. (1933). Ordering of wave functions and
eigenenergies to the individual electrons of an atom.
Physica, 1, 104113.
35. Yang, W. & Mortier, W. J. (1986). The use of global and
local molecular parameters for the analysis of the gasphase basicity of amines. J. Am. Chem. Soc. 108,
57085711.
36. Reed, A. E., Curtiss, L. A. & Weinhold, F. (1988). Intermolecular interactions from a natural bond orbital,
donoracceptor viewpoint. Chem. Rev. 88, 899926.
37. De Proft, F., Martin, J. M. L. & Geerlings, P. (1996). On
the performance of density functional methods for
describing atomic populations, dipole moments and
infrared intensities. Chem. Phys. Lett. 250, 393401.
38. Nguyen, L. T., Le, T. N., De Proft, F., Chandra, A. K.,
Langenaeker, W., Nguyen, M. T. & Geerlings, P.
(1999). Mechanism of [2 + 1] cycloadditions of hydrogen isocyanide to alkynes: molecular orbital and
density functional theory study. J. Am. Chem. Soc. 121,
59926001.
39. Perez, P., Simon-Manso, Y., Aizman, A., Fuentealba, P.
& Contreras, R. (2000). Empirical energydensity
relationships for the analysis of substituent effects in
chemical reactivity. J. Am. Chem. Soc. 122, 47564762.
40. Rai, V. & Namboothiri, I. N. N. (2006). A theoretical
evaluation of the Michael-acceptor ability of conjugated nitroalkenes. Eur. J. Org. Chem. 2006, 46934703.