(Received for publication, January 28, 1997, and in revised form, March 6, 1997)
From Arris Pharmaceutical Corporation, South San Francisco, California 94080
The pH dependences of the affinities for
streptavidin of linear and cyclic peptide ligands containing the HPQ
sequence discovered by phage display were determined by plasmon
resonance measurements. At pH values ranging from 3.0 to 9.0, the
Kd values for Ac-AEFSHPQNTIEGRK-NH2,
cyclo-Ac-AE[CHPQGPPC]IEGRK-NH2, and
cyclo-Ac-AE[CHPQFC]IEGRK-NH2, were determined by
competition, and those for
cyclo-[5-S-valeramide-HPQGPPC]K-NH2 were
determined directly by equilibrium affinity measurements. The
Kd values of the ligands increase by an average
factor of 3.0 ± 0.8 per decrease in pH unit between pH ~4.5 and
pH ~6.3. Below pH ~4.5 there is a smaller increase in
Kd values, and above pH ~6.3 the
Kd values become relatively pH-independent. We
determined the crystal structures of complexes of streptavidin with
cyclo-[5-S-valeramide-HPQGPPC]K-NH2 at pH
1.5, 2.5, 3.0, and 3.5, with cyclo-Ac-[CHPQFC]-NH2 at pH
2.0, 3.0, 3.6, 4.2, 4.8, and 11.8, with
cyclo-Ac-[CHPQGPPC]-NH2 at pH 2.5, 2.9, and 3.7, and with
FSHPQNT at pH 4.0 and compared the structures with one another and
with those previously determined at other pH values. At pH values
from 3.0 to 11.8, the electron density for the peptide His side chain
is strong, flat, and well defined. A hydrogen bond between the
N1 atom of the His and the peptide Gln amide group indicates the His
of the bound peptide in the crystals is uncharged at pH
3.0. By
determining selected structures in two different space groups, I222
with two crystallographically inequivalent ligand sites and
I4122 with one site, we show that below pH ~3.0, the
pKa of the bound peptide His in the crystals is
influenced by crystal packing interactions. The presence of the
N
1His-NGln hydrogen bond along with pH
dependences of the peptide affinities suggest that deprotonation of the
peptide His is required for high affinity binding of HPQ-containing
peptides to streptavidin both in the crystals and in solution.
Screening of peptide libraries either displayed on phage by
molecular biology or produced by combinatorial chemical synthesis is an
effective method for discovery of peptide ligands for diverse protein
targets. Owing to the remarkably high stability and high affinity of
streptavidin for its natural ligand biotin (Kd ~1015 M) (1), together with widespread
bioanalytical, diagnostic, and therapeutic applications (2-5), this
protein has been extensively used to develop and validate such
screening methodologies. High affinity unnatural ligands have been
discovered by screening linear (6-11) and cyclic (6) peptide
libraries. Streptavidin also provides an ideal paradigm for probing the
structural basis of high affinity protein-ligand interactions (12-16)
and for introducing or improving properties by protein engineering (2,
17, 18). Finally, the high resolution crystal structures of
apostreptavidin and of streptavidin-ligand complexes provide a powerful
basis for developing structure-based ligand design strategies (12, 19-23).
Recently we described the structures of streptavidin-bound linear and cyclic peptide ligands containing the HPQ sequence (13) discovered by phage display and probed the structural basis for the higher affinities of the cyclic ligands compared with the corresponding linear ones. These structures enabled the successful design of cyclic peptide ligands conformationally constrained with designed thioether cross-links (22), of streptavidin dimerizing peptide ligands (19-21), and of a streptavidin-binding small molecule ligand (12).
The binding to streptavidin and avidin of certain small molecule and peptide ligands is pH-dependent. The affinities for avidin of biotin derivatives, 2-iminobiotin and diaminobiotin, decrease dramatically as the pH is lowered (24), as does the affinity for streptavidin of a linear peptide discovered by phage display, FSHPQNT (25). The extent of topochemical catalysis of disulfide formation and the resulting dimerization of designed streptavidin-bound HPQ-containing ligands whose thiols are presented next to one another in the crystal lattice also depend on pH (20).
Determination of the pH dependence of ligand binding or of changes in properties incurred by ligand binding often yields insight into the mode of action (26-28) or mechanism of binding (29-36) in biological and chemical processes. Appraisal of the ionization states of groups in a protein-bound ligand or at the ligand binding site of the protein target may reveal some of the determinants of high affinity binding crucial for structure-based ligand design. To this end we determine the pH dependences of binding to streptavidin of linear and cyclic peptide ligands containing the HPQ sequence and probe the structural basis for the dependences through crystallographic determination of complexes at multiple pH values. The affinities for streptavidin at pH 3.0, 4.0, 5.0, 6.0, 6.2, 6.4, 7.0, 7.3, 8.0, and 9.3 of linear Ac-AEFSHPQNTIEGRK-NH2, cyclo-Ac-AE[CHPQGPPC]IEGRK-NH2, cyclo-Ac-AE[CHPQFC]IEGRK-NH2, and cyclo-[5-S-valeramide-HPQGPPC]K-NH2 are determined by plasmon resonance measurements. We also determine the streptavidin-bound crystal structures of cyclo-[5-S-valeramide-HPQGPPC]K-NH2 at pH 1.5, 2.5, 3.0, and 3.5 and of smaller versions of the other ligands, cyclo-Ac-[CHPQFC]-NH2 at pH 2.0, 3.0, 3.6, 4.2, 4.8, and 11.8, cyclo-Ac-[CHPQGPPC]-NH2 at pH 2.5, 2.9 and 3.7, and FSHPQNT at pH 4.0. Several structures are determined in two space groups, I222 and I4122. The structures of the complexes are compared with those previously determined at other pH values. The crystal structures of the streptavidin-bound linear and cyclic peptide ligands determined over a range of pH values show that the pKa of the peptide His is greatly reduced in the crystals. Together with the pH dependences of the peptide affinities, the structures of the complexes over a range of pH values suggest that deprotonation of the peptide His is required for high affinity binding both in the crystals and in solution.
Peptides for crystallography, FSHPQNT, cyclo-Ac-[CHPQGPPC]-NH2, cyclo-Ac-[CHPQFC]-NH2, and cyclo-[5-S-valeramide-HPQGPPC]K-NH2 were synthesized as described (6, 13, 22). For BIAcore analysis, corresponding peptides Ac-AEFSHPQNTIEGRK-NH2, cyclo-Ac-AE[CHPQGPPC]IEGRK-NH2, cyclo-Ac-AE[CHPQFC]IEGRK-NH2 and cyclo-[5-S-valeramide-HPQGPPC]K-NH2 were synthesized as described (6, 22). The first three of these are longer than the corresponding ones for crystallography; they contains N-terminal AE sequence and C-terminal IEGRK for reasons independent of the present study (6).
Crystallization of Streptavidin-Peptide ComplexesApostreptavidin, purchased from Calbiochem, was
crystallized by vapor diffusion in 40 sitting drops under
conditions described for crystals of space group I222 (13, 37).
Streptavidin-FSHPQNT was cocrystallized (13) in the I222 and
I4122 (a = 58.0 Å, c = 174.5 Å) space groups under similar conditions, pH 4.0, at an initial
peptide concentration of 15.0 mg/ml.
Streptavidin-cyclo-Ac-[CHPQFC]-NH2 crystals, pH 2.0, 3.0, 3.6, 4.2, 4.8, and 11.8 were prepared by soaking I4122
streptavidin-cyclo-Ac-[CHPQFC]-NH2 cocrystals (13) in
synthetic mother liquor (75% saturated
(NH4)2SO4, 25% buffer) at peptide
concentrations ranging from 15.0 to 28.0 mg/ml for several days. For
the pH 2.0 and 3.0 structures the buffer was 1.0 M sodium
formate. At pH 3.6, 4.2, and 4.8 the buffer was 1.0 M
potassium acetate, and at pH 11.8 it was 0.10 M
CAPS.1 Crystals of
streptavidin-cyclo-[5-S-valeramide-HPQGPPC]K-NH2, and -cyclo-Ac-[CHPQGPPC]-NH2 were prepared by soaking
I222 or I4122 apostreptavidin crystals in 75% saturated
(NH4)2SO4, 0.25 M
sodium formate, 20 mg/ml peptide, pH 1.5, 2.5, or 2.9, or in 75%
(NH4)2SO4, 0.25 M
potassium acetate, pH 3.7.
The target pH values were obtained by adjusting 2-ml volumes of
(NH4)2SO4/buffer solution with NaOH
or HCl using a calibrated Corning pH meter. The peptides were
subsequently dissolved in smaller (~100-250 ) volumes whose final
target pH values were confirmed with pH paper designed to measure
integral pH values as well as pH values 0.5 units from integral.
Because of the relatively large molarity of the buffers in the
synthetic mother liquors and the smaller molarity of peptides, the
effect of dissolving the peptide in the
(NH4)2SO4/buffer solutions is
insignificant.
Some x-ray diffraction data sets of streptavidin-peptide complexes were collected on a Siemens IPC area detector coupled to a Siemens three-circle goniometer mounted on a Rigaku rotating anode target copper tube operating at 50 kV, 60 mA. Data were indexed and reduced to produce integrated intensities and structure factors with the programs Sadie and Saint supplied by Siemens as described (13).
Other x-ray diffraction data sets were obtained with an R-AXIS IV image
plate system equipped with mirrors mounted on the Rigaku generator
upgraded to operate at 50 kV, 100 mA. The nonmonochromated mirror-focused x-rays were filtered with 0.004 cm of nickel. The image
plate and associated data reduction software (Biotex) were from
Molecular Structure Corporation (The Woodlands, TX).
Crystal-to-detector distances ranged from 69.6 to 85.0 mm, 2 = 0.0 °, oscillations in
(
) from 1.00 to 1.50 °/frame,
20-45 min/frame. For highly diffracting crystals, data were also
collected at a distance of 75.7 mm, 2
= 7.77 °. Data collection
statistics are summarized in Table I.
|
The previously determined crystal structures of streptavidin-peptide
complexes (13, 22) provided the starting structures for refinement of
the complexes at other pH values. Structures were refined with X-PLOR
(38) and with difference Fourier methods (39). In (|Fo| |Fc|)
c maps, positive and negative
peaks whose magnitudes were greater than 2.8
were systematically
identified with the program Peak-pick, which was written in house, and
analyzed. Water structure was determined with Peak-pick, X-sight, or
X-solvate from Molecular Simulations, Inc. (San Diego, CA) and refined
according to published procedures (40). Waters with temperature factors greater than 50 Å2 were examined in the context of the
corresponding 2(|Fo|
|Fc|)
c map and kept only if there was significant density for
them. The coordinates corresponding to the conformations of discretely disordered residues (such as the peptide histidine in some structures) were simultaneously refined along with the rest of the structure, followed by simultaneous refinement of temperature factors and then of
occupancies. Refinement of coordinates, temperature factors, and
occupancies was iterated until the parameters and R-factor converged.
Refinement statistics are given in Table I. The
structures have been deposited into the Brookhaven Data Bank.
The affinities of linear and cyclic streptavidin
binding peptides were determined as a function of pH by surface plasmon
resonance. The BIAcore 2000 system, sensor chip, amine coupling kit
containing N-hydroxysuccinimide,
N-ethyl-N-(3-diethylaminopropyl)-carbodiimide, and ethanolamine-HCl were from Pharmacia Biotech Inc. Buffers containing 150 mM NaCl, 3.4 mM EDTA, and
0.0050% surfactant P20 ranging from pH 3.0 to 8.0 were 10 mM potassium acetate, pH 3.0; 10 mM sodium
acetate, pH 4.0, 5.0, and 6.0; 10 mM bis-Tris, pH 6.2 and
6.4; 10 mM HEPES, pH 7.0 and 7.3; and 10 mM
Tris, pH 8.0 and 9.0.
The peptide, cyclo-Ac-AE[CHPQGPPC]IEGRK-NH2, was coupled
directly to the sensor chip surface via the -amino group of the C-terminal lysine residue under a continuous HEPES buffer flow of 10 µl/min. The sensor chip surface was activated with a 2-min pulse of a
solution of 50 mM N-hydroxysuccinimide, 200 mM
N-ethyl-N
-(3-diethylaminopropyl)-carbodiimide. The peptide surface was made by injection of 100 mM peptide
in 100 mM sodium borate, pH 8.5. To inactivate any
remaining N-hydroxysuccinimide ester groups, the
immobilization procedure was completed by an 8-min injection of 1.0 M ethanolamine-HCl followed by a 2-min pulse of 6.0 M guanidine-HCl, pH 2.1, to wash out any noncovalently bound peptide. The Kd values of
cyclo-Ac-AE[CHPQGPPC]IEGRK-NH2, cyclo-AE[CHPQFC]IEGRK-NH2, and
Ac-AEFSHPQNTIEGRK-NH2 were determined by described
procedures (13, 20, 22) as a function of pH through competition for
streptavidin with cyclo-Ac-AE[CHPQGPPC]IEGRK-NH2 immobilized on the surface.
Direct Kd values were determined for cyclo-[5-S-valeramide-HPQGPPC]K-NH2 immobilized on the BIAcore chip. At each pH, two flow cell surfaces with different relative amounts of immobilized peptide were made by varying the peptide injection time. The cell with the lower density surface was used as a blank. At each pH, triplicate data sets were analyzed by the method of multispot sensing (41), which relies on equilibrium affinity measurements on two surfaces of different peptide ligand densities. To minimize perturbations to the apparent affinity constants caused by avidity or ligand binding to the surface more than once, it was necessary to use surfaces whose peptide densities corresponded to below 15 resonance units. Affinities within experimental error of those previously determined by other methods (6) were thus obtained.
Because of the large amounts of peptides required for both crystallography and BIAcore at low pH where the binding is weak, two sets of peptides were used for most of this study: core peptides for crystallography and longer peptides synthesized for a previous study (6) for BIAcore. The core peptides were used for crystallography because the C- and N-terminal extensions in the streptavidin-bound longer versions are expected to be disordered or to make crystal cracking from peptide soaking more likely. The affinities of the longer peptides were determined to be within experimental error of the corresponding core versions; the Kd of cyclo-Ac-[CHPQGPPC]-NH2, pH 7.3, is 310 nM compared with 230 nM for cyclo-Ac-AE[CHPQGPPC]IEGRK-NH2, pH 7.3 (22), and the Kd of FSHPQNT is 78 mM (6) or 125 mM (25) at pH 7.3 compared with 150 mM for FSHPQNTK-NH2, pH 7.3 (12) or 160 mM determined here for Ac-AEFSHPQNTIEGRK-NH2, pH 7.3. Similarly the Kd values of the other core peptides in this study were determined to lie within experimental error of the corresponding longer versions. An engineered thioether cross-linked ligand (22), cyclo-[5-S-valeramide-HPQGPPC]K-NH2, was used both for crystallography and BIAcore.
pH-dependent Affinities Implicate Ionization of a Group with a pKa of ~6.3 upon BindingThe pH dependences of affinities of Ac-AEFSHPQNTIEGRK-NH2, cyclo-Ac-AE[CHPQGPPC]IEGRK-NH2, cyclo-Ac-AE[CHPQFC]IEGRK-NH2, and cyclo-[5-S-valeramide-HPQGPPC]K-NH2 determined by plasmon resonance measurements are shown in Table II and plotted in Fig. 1. The affinity of the linear peptide is several hundred-fold lower than the affinities of the cyclic peptides at all pH values of this study. The Kd values of the ligands increase at roughly the same rate as the pH is lowered from ~6.3 to ~4.5, by an average factor of 3.0 ± 0.8 per decrease in pH unit. Below pH ~4.5 the increase in Kd values becomes smaller, and above pH ~6.3 the Kd values become relatively pH-independent. The apparent pKa of ~6.3 is most clearly seen in the data for cyclo-[5-S-valeramide-HPQGPPC]K-NH2, for which direct Kd values (and associated standard deviations from triplicate data sets) at pH 5.0, 6.0, 6.2, 6.4, 7.0, 7.3, 8.0, and 9.0 were determined. Thus ionization of a group (or groups) with an apparent pKa of ~6.3 in the unbound state is implicated in the binding process.
|
An Intrapeptide Hydrogen Bond between N
The
(2|Fo| |Fc|)
c map
superimposed on the refined structures of I222
streptavidin-cyclo-[5-S-valeramide-CHPQGPPC]K-NH2, pH 2.5, at one of the two crystallographically independent ligand binding sites is shown in Fig. 2A. The
refined N
1His-NGln distance and associated
angles (Table III) indicate a hydrogen bond between NGln and the unprotonated N
1 atom of the uncharged His.
The imidazole density is strong, well defined, and flat, typical for a
well ordered histidine. Similar density and an
N
1His-NGln distance and associated bond
angles indicating a hydrogen bond are also obtained for the
I4122 complex at pH 2.5 and for the complexes of other
linear and cyclic HPQ-containing peptides at pH values > 2.5 (Table III).
|
The density
for the peptide imidazole in streptavidin-cyclic peptide complexes at
pH values 2.0 indicates disorder. In I4122 streptavidin-cyclo-Ac-[CHPQFC]-NH2, pH 2.0, the imidazole
density becomes somewhat cylindrical through disorder involving an
additional protonated conformer that is rotated by ~20 ° about the
peptide histidine C
-C
bond compared with the single unprotonated
conformation in the same complex at pH 3.0 or in I222
streptavidin-cyclo-[5-S-valeramide-HPQGPPC]K-NH2, pH 2.5 (Fig. 2A). For the charged conformer the
N
1His-NGln hydrogen bond is no longer
possible, and the N
1His-NGln distance
increases to 3.25 Å. The occupancy of the charged conformer in
I4122 cyclo-Ac-[CHPQFC]-NH2, pH 2.0, is 26%,
yielding a calculated pKa of 1.5. Accompanying the
disorder in the peptide imidazole in this complex is an increase in its
average temperature factor from 28.4 ± 0.5 Å2 at pH
3.0 to 61.7 ± 1.4 Å2 at pH 2.0.
Disorder and mobility in the peptide His are also apparent at both
ligand binding sites in I222
streptavidin-cyclo-[5-S-valeramide-HGPQFC]K-NH2, pH 1.5, in which the temperature factor of the peptide imidazole is
high, ~63 Å2. The density for the peptide is weak and/or
broken, reflecting weaker binding at this low pH, but the density of
the imidazole is well defined enough to resolve two imidazole
components. The N1 atom of one component is not within hydrogen
bonding distance of NGln and is surely protonated. Although
the refined N
1His-NGln distance (2.92 ± 0.13 Å) of the other component suggests a hydrogen bond, the
NGln-HGln-N
1His angle is poor
(123 ± 5°), lower by 4.1
than the average for the other
complexes above pH 2.0 (Table III). Thus this other component may be
largely protonated as well, despite the small
N
1His-NGln distance.
In I222 streptavidin-peptide complexes one of the two
crystallographically independent binding sites (site 1) is near a
2-fold related crystallographically equivalent site and is thus more shielded from solvent than site 2. The unique site in I4122
streptavidin is also solvent-shielded from a nearby 2-fold related
equivalent site. At many pH values the density of the bound peptide in
the complexes is better defined at the unique site in I4122
complexes and at site 1 in I222 complexes than at site 2. The
conclusions regarding the protonation states of the His of the bound
peptides are essentially the same for site 1 and site 2 of I222
complexes and for the unique site of I4122 complexes at pH
values > 2.5 where the density is well defined, the temperature
factors relatively low, and the N1His-NGln
distances and associated angles (Table III) clearly indicate an
N
1His-NGln hydrogen bond at all three sites.
However, at pH values
pH 2.5, differences in the protonation state of the peptide His are observed at crystallographically different
sites.
Fig. 2 (A and B) compares the structures of the
bound peptide at site 1 and site 2, respectively, for I222
streptavidin-cyclo-[5-S-valeramide-HPQGPPC]K-NH2, pH 2.5. The imidazole density at site 2 is distinctly different from
that at site 1; it is elongated in a direction corresponding to the
presence of a second, protonated histidine conformer that is rotated
17° about the C-C
bond with respect to the unprotonated conformer. The distance and angle parameters of the unprotonated conformer indicate an N
1His-NGln hydrogen
bond (Table III) shown in yellow in Fig. 2B. The
occupancies of the protonated and unprotonated conformers are 20 and
80%, respectively, yielding a calculated pKa at
site 2 of 1.9. For the same I222 complex at pH 3.0, similar
inequivalence of the two sites is observed, with resolvable protonated
and unprotonated conformers at site 2. The occupancies of the
protonated and unprotonated conformers at site 2 are 15 and 85%,
respectively, at pH 3.0, yielding a calculated pKa
of 2.2. By contrast, at site 1, the occupancies of any unprotonated
components at pH
2.5 are not high enough to observe or to
resolve. Thus the pKa of the bound peptide His
is detectably lower at site 1 than at site 2.
In the conformation of the protonated component at site 2 in I222
streptavidin-cyclo-[5-S-valeramide-HPQGPPC]K-NH2,
pH 2.5, shown in Fig. 2B, the side chain is rotated 180 °
about the C-C
bond with respect to the conformation of the
unprotonated component to allow N
1 to make a hydrogen bond (shown in
cyan) with a water molecule. This change in the His
conformation increases the N
2His-O
Ser88 distance from 3.05 to 3.21 Å and decreases the
N
2His-H
2His-O
Ser88 angle
from 163 to 108 °, indicating a weakening or loss of the hydrogen
bond involving N
2His and O
Ser88. However
an alternate hydrogen bond, C
1His-O
Ser88 = 2.98 Å,
C
1His-H
1His-O
Ser88 = 126 ° (shown in cyan in Fig. 2B) similar to
those observed in other protein crystal structures (42) is now
possible.
Inequivalence of the two peptide binding sites in I222 complexes is also manifested by differences in temperature factors for the bound ligands at many pH values. Temperature factors are lower at site 1 than at site 2 except at low pH where they become large at both sites. For example, in the I222 cyclo-[5-S-valeramide-HPQGPPC]K-NH2 complex the average temperature factor of the imidazole at pH 2.5, 3.0, and 3.5 is lower at site 1 (21 ± 4 Å2) than at site 2 (36 ± 5 Å2), whereas at pH 1.5 it increases dramatically at both sites, becoming the same at site 1 (63 ± 2 Å2) as at site 2 (64 ± 2 Å2). For the I4122 complexes, the ligand temperature factors are as low as at site 1 in the I222 complexes.
The pH dependences of
affinities for streptavidin of linear and cyclic HPQ-containing
peptides are consistent with ionization of a group with a
pKa of ~6.3 in the unbound state involved in
binding. The unperturbed pKa for the His side chain fully exposed to water is 6.3 (43). The observation of the
N1His-NGln hydrogen bond over a range of pH
values in the bound linear and cyclic peptides shows that the peptide
His is uncharged and unprotonated at pH values as low as 2.5 in the
crystalline cyclic peptide complexes; the pKa of the
peptide His is thus reduced to a value of less than 2.5. Therefore the
decrease in binding affinity in solution as the pH decreases is
attributed to the cost of deprotonating the peptide His at low pH. The
deprotonation and formation of the
N
1His-NGln hydrogen bond are required for
high affinity binding both in the crystals and in solution.
Table III provides the N1His-NGln hydrogen
bond lengths and associated angles for various HPQ-containing peptide
ligands at various pH values for this and other investigations (13,
19-22). This hydrogen bond is also observed in complexes of
streptavidin with HHPQGPPH, linear (reduced)
Ac-CHPQGPPC-NH2, and linear
Ac-CHPQFC-NH2.2 These data
suggest that for other HPQ-containing linear peptide ligands such as
(HDHPQNL and SHPQGPPS) and disulfide-bonded cyclic peptide ligands such
as cyclo-[CHPQFSNC], cyclo-[CHPQFPC], and cyclo-[CHPQFNC] (6),
the same hydrogen bond forms in the complexes with streptavidin at pH
values above 2.5.
Inequivalence of the two crystallographically independent sites in I222 streptavidin results in significant differences in some of their properties. The temperature factors of bound peptide ligands are often lower and the density better defined at site 1 than at the more solvent exposed site 2. Topochemical lattice-mediated disulfide interchange occurs between neighboring bound cyclo-Ac-[CHPQGPPC]-NH2 ligands at site 1, which is close to a 2-fold related equivalent site, but not at site 2 (21). Two of the hydrogen bonds to the ureido oxygen of biotin are systematically shorter over a range of pH values at site 1 than at site 2.2 In this investigation site 1 and site 2 were also shown to differ with respect to the protonation state of the His in bound HPQ-containing peptide ligands. The pKa of the peptide His is lower at the more solvent-shielded site 1 than at site 2. In the I4122 space group the pKa at the unique, solvent-shielded ligand binding site is also lower than at the solvent exposed site 2 in the I222 space group. The protonation state of the peptide His at the more solvent exposed site 2 in I222 streptavidin-peptide crystals more accurately reflects its state in solution. However, because of the observed effect of crystal packing on the pKa of the peptide His at site 1 in I222 streptavidin complexes and at the unique site in I4122 streptavidin complexes, solvent shielding to a lesser extent by crystal packing at site 2 in I222 streptavidin complexes must also be considered as a potential factor that could perturb the pKa of the peptide His at this site from the pKa in solution. Thus the crystallographically determined pKa of the peptide His at site 2 in I222 streptavidin should be taken as a lower limit to the corresponding value in solution.
Directionalities of Hydrogen Bonds Involving Linear and Cyclic Peptide Ligands Are UnambiguousBecause the proton atoms of protein-ligand complexes are normally not visible by x-ray crystallography, directionalities of hydrogen bonding interactions can not always unambiguously be determined by this technique. In some cases, however, from the environments of residues or groups participating in hydrogen bond networks, directionalities of some hydrogen bonds can be inferred (44). For the complexes of streptavidin with the linear and cyclic peptide ligands discussed here, the directionality of every hydrogen bond involving every atom of each ligand is uniquely determined at each pH value > 2.5.
In complexes of streptavidin with HPQ-containing peptide ligands, the
orientation of the peptide Gln side chain amide group is unambiguous
based in part on the better geometry of hydrogen bonds in one
orientation (Fig. 3A) versus the
alternate one in which this amide is rotated by 180 ° (Fig.
3B). In the less favorable orientation (Fig. 3B),
the vector between N2Gln and O
2Thr90 is
directed right between the Gln N
2 hydrogens, and there are two
proton donors to O
2Thr90 (N
2Gln and
N
1Trp79). In the more favorable orientation, the O
2
proton of Thr90 is directed at O
1Gln, whereas one of the
N
2Gln hydrogens is directed at the oxygen of
Wat600. In this arrangement O
2Thr90 receives
a proton from N
1Trp79 and donates a proton to
O
1Gln. The angles associated with these hydrogen bonds
are more favorable in this orientation than in the alternate one. The
angle between the hydrogen bonds received and provided by
O
2Thr90 is ~90 °. The peptide Gln side chain is
also uniquely oriented because of its interaction (d = 3.51 ± 0.15 Å, determined from 12 structures) with the
Trp108 ring, which reflects an NH
aromatic ring
system hydrogen bond similar to those described (45).
FSHPQNT
Fig. 4 shows the hydrogen bonding
network that connects the linear peptide to streptavidin at pH 4.0 and
5.6. The orientation of the peptide Asn side chain is unambiguous
because of the hydrogen bond between its O1 atom and
N
2Asn23. Asn23 is in turn uniquely oriented
due to the hydrogen bond between its O
1 atom and NSer27.
The peptide Asn side chain is also uniquely oriented because of an
N
2Asn23
Trp120 interaction
(d = 3.64 ± 0.04 Å, determined from four
structures). Because the peptide His N
1 atom accepts a proton from
the peptide Gln main chain NH group, the peptide His N
2 atom must
donate a proton to O
Ser88. The two other peptide groups,
OGln and NThr, involved in hydrogen bonds must
be an acceptor and donor, respectively. Thus the directionalities of
all 11 peptide-protein hydrogen bonds in streptavidin-FSHPQNT are
unambiguous, as well as the directionality of the intrapeptide hydrogen
bond.
Cyclic Peptides
Fig. 5 shows the hydrogen
bonding interactions for
streptavidin-cyclo-Ac-[CHPQFC]-NH2, pH 3.0. The
directionalities of the hydrogen bond interactions involving this
cyclic peptide are also unambiguous based on considerations similar to
those described above for the streptavidin-bound linear peptide. The
same hydrogen bonding network is observed at pH 11.8, 7.5, 4.8, 4.2, 3.6, and 2.0 with a protonated His component at the latter pH.
Likewise, the hydrogen bond directionalities involving the binding of
cyclo-Ac-[CHPQGPPC]-NH2 to streptavidin at pH 2.5, 2.9, 3.7, 5.0, and 7.5 are unambiguous (Fig. 6). The same
network is observed in the complex with
cyclo-[5-S-valeramide-HPQGPPC]K-NH2 at pH 1.5, 2.5, 3.0, and 3.5;
cyclo-[5-S-valeramide-HPQGPPC]-NH2, pH 6.0 (22); cyclo-[4-S-toluamide-HPQGPPC]-NH2, pH
6.0 (22); head-to-head cyclo-[CHPQGPPC]-NH2 dimer, pH
2.5, 3.5, 5.0 (21), and 6.0 (21); and head-to-tail
cyclo-[CHPQGPPC]-NH2 dimer, pH 2.5, 3.5, and 7.3 (19),
with protonated peptide His components at pH values 2.5 in
these complexes.
In a previous study of the binding of streptavidin to FSHPQNT, it was suggested that enhanced binding at neutral compared with acidic pH reflects an equilibrium between alternate unprotonated states of the peptide His in alternate equi-energetic hydrogen bond networks involving the protein at neutral pH (25). Thus an increase in binding at neutral pH was proposed to reflect an increase in entropy (25). In the present study, however, we were able to establish one unique hydrogen bonding network wherein the directionalities of all hydrogen bonds between FSHPQNT and streptavidin, even those mediated by water molecules, are essentially unambiguous at pH 4.0 and 5.6. Similarly, unique, lowest energy hydrogen bond networks were delineated for all the streptavidin-cyclic peptide complexes at pH values > 2.5. In this investigation any disorder in the His occurs at low pH (< 2.5), not neutral or basic pH. Therefore the pH dependence of binding of any of these peptides to streptavidin is probably not due to such an entropy effect. The structures and affinities of the linear and cyclic HPQ-containing peptide ligands determined over a large range of pH values in two space groups are most consistent with deprotonation of the peptide His upon binding as a major determinant of the pH dependence of ligand binding.
Structural Basis for the Large Perturbation of the pKa of the Peptide HistidineThe experimentally determined pKa values of certain noncatalytic residues in naturally occurring proteins are shifted by as much as 2.5 units (46). Larger shifts in the pKa values often occur in functionally important residues, such as those at active sites where buried charged residues often participate in catalysis (47-51). For example, the pKa of active site Asp26 in reduced thioredoxin is elevated by more than 5 units (52). The pKa values of charged residues engineered into hydrophobic cores of protein mutants are also perturbed by as much as 3.9 units (53-55). Much of the theoretical work that involves prediction of pH-dependent properties of proteins is based on the assumption that ionization equilibria in proteins are influenced primarily by electrostatic interactions (43, 56-58). Hydrogen bonding interactions involving nonionizable groups (59) and hydrophobic interactions (60) may also play roles in determining pKa values. Large shifts in pKa values can also be effected simply by desolvation, the dominant factor shifting the pKa of the buried lysine introduced into staphylococcal nuclease (53).
Although the peptide His side chain of streptavidin-bound
HPQ-containing peptide ligands is not involved in salt bridge
interactions, it makes two hydrogen bonds
(N1His-NGln and
N
2His-O
2Thr90) at pH values
2.5. Thus, hydrogen bonding interactions together with desolvation within a
protein cavity of low dielectric constant are among the factors
perturbing the pKa in the
streptavidin-HPQ-containing peptide complexes. Although the C
2 atom
of the imidazole of the peptide His is solvent-accessible from one
direction, the other imidazole atoms are shielded from solvent by the
rest of the bound peptide and by Trp79, Leu110,
Ser88, Ala86, and Trp120 of a
neighboring subunit. Upon protonation at low pH a small rotation about
1 of the peptide His and a 180 ° rotation about
2 allows N
1
to hydrogen bond with a solvent molecule (Fig. 2B).
Through plasmon resonance measurements combined with crystallography at multiple pH values on a set of HPQ-containing ligands, features of the mechanism of high affinity binding to streptavidin have been delineated. High resolution crystal structures at pH values as low as 1.5 yield insight into the nature of the structural rearrangements that occur in the bound peptide upon protonation of the His at three crystallographically different binding sites in two space groups. Observation of perturbations to the pKa of the peptide His from long range crystal packing interactions should be taken as a caveat in extrapolation of pKa values determined in crystals to the corresponding ones in solution. The determination of the greatly reduced pKa of the His in streptavidin-bound HPQ-containing peptides and of the difference in pKa at sites with different extents of solvent shielding should provide valuable structural data for testing and improving theoretical models directed at predicting pH-dependent properties of proteins and of protein-ligand complexes.