Phosphate activation in the ground state of purine nucleoside phosphorylase

Article abstract of DOI:10.1021/ja0570281

Phosphate and ribose 1-phosphate (R1P) bound to human purine nucleoside phosphorylase (PNP) have been studied by FTIR spectroscopy for comparison with phosphate bound with a transition state analogue. Bound phosphate is dianionic but exists in two distinct binding modes with similar binding affinities. The phosphate of bound R1P is also dianionic. Bound R1P slowly hydrolyzes to ribose and phosphate even in the absence of nucleobase. The C-OP bond is cleaved in bound R1P, the same as in the PNP-catalyzed reaction. Free R1P undergoes both C-OP and CO-P solvolysis. A hydrogen bond to one P-^..O group is stronger than those to the other two P-^..O groups in both the PNP·R1P complex and in one form of the PNP·PO₄ complex. The average hydrogen bond strength to the P-^..O bonds in the PNP·R1P complex is less than that in water but stronger than that in the PNP·PO₄ complex. Hydrolysis of bound R1P may be initiated by distortion of the phosphate moiety in bound R1P. The unfavorable interactions on the phosphate moiety of bound R1P are relieved by dissociation of R1P from PNP or by hydrolysis to ribose and phosphate. The two forms of bound phosphate in the PNP·PO₄ complex are interpreted to be phosphate positioned as the product in the nucleoside synthesis direction and as the reactant in the phosphorolysis reaction; their interconversion can occur by the transfer of a proton from one PO bond to another. The electronic structure of phosphate bound with a transition state analogue differs substantially from that in the Michaelis complexes.

Full text of DOI:10.1021/ja0570281

Published on Web 05/26/2006
Phosphate Activation in the Ground State of Purine
Nucleoside Phosphorylase
Hua Deng,* Andrew S. Murkin, and Vern L. Schramm
Contribution from the Department of Biochemistry, Albert Einstein College of Medicine,
1300 Morris Park AVenue, Bronx, New York 10461
Received October 15, 2005; E-mail: hdeng@aecom.yu.edu
Abstract: Phosphate and ribose 1-phosphate (R1P) bound to human purine nucleoside phosphorylase
(PNP) have been studied by FTIR spectroscopy for comparison with phosphate bound with a transition
state analogue. Bound phosphate is dianionic but exists in two distinct binding modes with similar binding
affinities. The phosphate of bound R1P is also dianionic. Bound R1P slowly hydrolyzes to ribose and
phosphate even in the absence of nucleobase. The C-OP bond is cleaved in bound R1P, the same as in
the PNP-catalyzed reaction. Free R1P undergoes both C-OP and CO-P solvolysis. A hydrogen bond to
one P-O group is stronger than those to the other two P-O groups in both the PNP‚R1P complex and
in one form of the PNP‚PO4 complex. The average hydrogen bond strength to the P-O bonds in the
PNP‚R1P complex is less than that in water but stronger than that in the PNP‚PO4 complex. Hydrolysis of
bound R1P may be initiated by distortion of the phosphate moiety in bound R1P. The unfavorable interactions
on the phosphate moiety of bound R1P are relieved by dissociation of R1P from PNP or by hydrolysis to
ribose and phosphate. The two forms of bound phosphate in the PNP‚PO4 complex are interpreted to be
phosphate positioned as the product in the nucleoside synthesis direction and as the reactant in the
phosphorolysis reaction; their interconversion can occur by the transfer of a proton from one PO bond to
another. The electronic structure of phosphate bound with a transition state analogue differs substantially
from that in the Michaelis complexes.
Introduction
It has been suggested that this reaction proceeds by a mechanism
termed “nucleophilic displacement by electrophile migration”
Phosphoryl transfer enzymes function by either breaking the
bridging RO-P bond so that a PO3 group is transferred or
breaking the R-OP bond so that a PO4 group is transferred.
The phosphate distortions in the reaction intermediate by several
PO3 transferring enzymes have been studied. For example, the
structure of a cryo-temperature trapped, pentacovalent phos-
phorus intermediate in â-phosphoglucomutase has been observed
by X-ray crystallography.1,2 By using photolysis of caged GTP
bonds, the distortions of the phosphate moiety during the
hydrolysis of GTP in Ras protein activated by GTPase activating
protein or in a hyper-active mutant have been studied by FTIR.3,4
Similar studies have been performed on the phosphate distortion
in ATP hydrolysis by Ca2+-ATPase using caged ATP com-
pounds.5 These studies revealed how the phosphate moieties in
Ras and Ca2+-ATPase are distorted to facilitate the hydrolysis
reaction.
(see below). Previous X-ray crystallography studies on a
transition state analogue complex have shown that in this
mechanism the nucleophilic inorganic phosphate is fixed in the
catalytic site pocket by at least nine hydrogen bonds and ionic
interactions with the protein and an additional hydrogen bond
and ionic interaction with Immucillin-H, a transition state
analogue.7,8 These interactions decouple the stretch motion of
one P-O stretch of the bound phosphate from the other two,
resulting in a highly asymmetric bound phosphate.9 Here we
investigate the nature of bound phosphate in the Michaelis
complexes of phosphate and ribose 1-phosphate without the
transition state analogue. Comparison of these complexes
suggests changes to the phosphate environment on conversion
from Michaelis to transition state analogue complexes.
Purine nucleoside phosphorylase catalyzes the reversible
phosphorolysis of the N-ribosidic bond of 6-oxypurine nucleo-
sides and deoxynucleosides (Scheme 1). In humans, the
metabolic role of PNP is to remove deoxyguanosine that
accumulates from DNA turnover. The genetic deficiency of PNP
In this study, we report the observation of phosphate distortion
in ground state complexes of another class of phosphoryl
transferring enzyme, purine nucleoside phosphorylase (PNP).6
(1) Tremblay, L. W.; Zhang, G.; Dai, J.; Dunaway-Mariano, D.; Allen, K. N.
J. Am. Chem. Soc. 2005, 127, 5298-9.
(7) Fedorov, A.; Shi, W.; Kicska, G.; Fedorov, E.; Tyler, P. C.; Furneaux, R.
H.; Hanson, J. C.; Gainsford, G. J.; Larese, J. Z.; Schramm, V. L.; Almo,
S. C. Biochemistry 2001, 40, 853-60.
(2) Lahiri, S. D.; Zhang, G.; Dunaway-Mariano, D.; Allen, K. N. Science 2003,
299, 2067-71.
(3) Allin, C.; Gerwert, K. Biochemistry 2001, 40, 3037-46.
(4) Du, X.; Frei, H.; Kim, S. H. J. Biol. Chem. 2000, 275, 8492-500.
(5) Liu, M.; Krasteva, M.; Barth, A. Biophys. J. 2005, 89, 4352-63.
(6) Abbreviations: PNP, human purine nucleoside phosphorylase; R1P, R-D-
ribose 1-phosphate; ImmH, Immucillin-H.
(8) Lewandowicz, A.; Shi, W.; Evans, G. B.; Tyler, P. C.; Furneaux, R. H.;
Basso, L. A.; Santos, D. S.; Almo, S. C.; Schramm, V. L. Biochemistry
2003, 42, 6057-66.
(9) Deng, H.; Lewandowicz, A.; Schramm, V. L.; Callender, R. J. Am. Chem.
Soc. 2004, 126, 9516-7.
9
10.1021/ja0570281 CCC: $33.50 © 2006 American Chemical Society
J. AM. CHEM. SOC. 2006, 128, 7765-7771
7765

A R T I C L E S
Deng et al.
Scheme 1. PNP-Catalyzed Reaction and lmmH-Phosphate
Interactions in the PNP‚lmmH‚PO4 Complex
in the active site might be expected. Previous studies have shown
that bovine PNP catalyzes the hydrolysis of inosine in the
absence of PO4 and binds tightly to the hypoxanthine (Kd ∼ 2
pM) product,14 supporting the formation of a ribooxacarbenium
ion and its subsequent hydrolysis. However, the observation of
stable inosine in the crystal structures of a PNP‚inosine complex
seems to be inconsistent with the SN1-like mechanism and in
favor of a SN2-like mechanism.15 The recent transition state
analysis of human PNP provides unequivocal evidence that the
arsenolysis reaction catalyzed by PNP has a transition state
closely related to a fully dissociated ribooxacarbenium ion.16
Vibrational spectroscopy is an ideal tool to probe the
electronic state of enzyme-bound phosphate and its interactions
with proteins.4,17-21 Activation of phosphate in a transition state
analogue complex of PNP‚ImmH‚PO4 has been studied by
vibrational spectroscopy and ab initio normal mode analysis.9
To investigate how PNP may activate the phosphate nucleophile
in the ground state, we have extended our vibrational studies
to the PNP‚PO4 and PNP‚R1P complexes. Our results show that
the phosphate dianions in both complexes are noticeably
distorted upon binding to PNP, established by significant
frequency changes in the phosphate P-O stretch modes.
Furthermore, we have found that bound R1P is slowly hydro-
lyzed to ribose and phosphate by attack of the water (hydroxyl)
nucleophile at C1 of R1P, rather than at the phosphorus atom.
This reaction resembles the PNP-catalyzed nucleoside synthesis
reaction. Thus, activation of the substrate R1P toward the
ribooxacarbenium ion in the PNP‚R1P complex is evidence for
the ground state activation expected in the electrophile migration
mechanism. Substrate activation arises from the unfavorable
interactions on the phosphate moiety of PNP-bound R1P
compared to either PNP-bound phosphate or R1P in solution.
causes a T-cell immunodeficiency due to dGTP accumulation
in dividing T-cells.10 Inhibition of PNP inhibits the growth of
activated T-cells, providing a clinical means to ameliorate T-cell
proliferative disorders.11 The catalytic acceleration of PNP is
achieved through formation of a transition state with oxacar-
benium ion character and specific leaving group interactions to
the purine. Ribosides with better leaving groups, such as
4-nitrophenyl-â-D-riboside, are poor substrates, establishing
leaving group activation as a major catalytic interaction.12
Catalytic site contacts to the purine fix it in one position through
the reaction coordinate and facilitate ribosyl electrophile migra-
tion from the leaving group to a phosphorus nucleophile also
immobilized at the catalytic site.7 Although the reaction equi-
librium favors nucleoside synthesis, the enzyme operates in the
phosphorolysis direction in vivo because of the rapid metabolic
removal of products by purine phosphoribosyl transferases.
Recently, nucleophilic displacement by electrophilic migration
has been proposed for PNP, hypoxanthine-guanine (-xanthine)
phosphoribosyl transferase, lysozyme, and uracil DNA glyco-
sylase.7,13 Key features of this SN1 reaction mechanism include
(1) activation of both leaving group and nucleophile, (2)
activation of the nucleobase leaving group by protonation or
by hydrogen bonding to N7, (3) formation of a fully dissociated
ribooxacarbenium ion at some point (not necessarily the
transition state) during the reaction, and (4) neighboring group
assistance by stacking the ribose O5′ over O4′ as a driving force
for the electrophile (ribooxacarbenium) migration in PNP. This
reaction mechanism is significantly different from the SN2-like
mechanisms previously proposed for a wide range of glycosyl
transferases, wherein activation of the nucleophile in the ground
state is not required and no ribooxacarbenium ion is formed.
Furthermore, the phosphate group is fixed in position relative
to the enzyme, unlike that in the PO3 transferring enzymes.
The formation of an enzyme-stabilized ribooxacarbenium ion
requires significant substrate activation in the ground state, and
hydrolytic quenching of oxacarbenium ions by water molecules
Materials and Methods
7-Methyl inosine was obtained from Sigma. H3P18O4 was prepared
according to the published procedure.22 R1P or 18O4-labeled R1P was
prepared by mixing PO4 or P18O4 and 7-methyl inosine in a 1:1.5 molar
ratio at pH 7.5 in the presence of PNP. After completion of the reaction,
the unreacted nucleoside and base were removed by charcoal. The final
products were checked by 1H and 31P NMR to be >99% pure and free
of inorganic phosphate. Human PNP was prepared as described
previously.23 To remove tightly bound hypoxanthine that co-purifies
with human PNP, the enzyme (at concentration <1 mg/mL, determined
by UV using an extinction coefficient of 30 mM-1 cm-1 at 280 nm)
was dialyzed for 3 days against 50 mM phosphate buffer at pH 7.2, 4
°C in the presence of 2 g of charcoal in the dialysate. Removal of
bound hypoxanthine was verified with proton NMR.24 Hypoxanthine-
(14) Kline, P. C.; Schramm, V. L. Biochemistry 1992, 31, 5964-73.
(15) Canduri, F.; dos Santos, D. M.; Silva, R. G.; Mendes, M. A.; Basso, L. A.;
Palma, M. S.; de Azevedo, W. F.; Santos, D. S. Biochem. Biophys. Res.
Commun. 2004, 313, 907-14.
(16) Lewandowicz, A.; Schramm, V. L. Biochemistry 2004, 43, 1458-68.
(17) Mertz, E. L.; Leikin, S. Biochemistry 2004, 43, 14901-12.
(18) Martinez-Liarte, J. H.; Iriarte, A.; Martinez-Carrion, M. Biochemistry 1992,
31, 2712-19.
(19) Wang, J. H.; Xiao, D. G.; Deng, H.; Webb, M. R.; Callender, R.
Biochemistry 1998, 37, 11106-16.
(20) Cheng, H.; Sukal, S.; Deng, H.; Leyh, T. S.; Callender, R. Biochemistry
2001, 40, 4035-43.
(10) Stoeckler, J. D. Developments in Cancer Chemotherapy. In DeVelopments
in Cancer Chemotherapy; Glazer, R. J., Ed.; CRC Press Inc.: Boca Raton,
FL, 1984; pp 35-60.
(11) Bzowska, A.; Kulikowska, E.; Shugar, D. Pharma. Ther. 2000, 88, 349-
425.
(12) Mazzella, L. J.; Parkin, D. W.; Tyler, P. C.; Furneaux, R. H.; Schramm,
V. L. J. Am. Chem. Soc. 1996, 118, 2111-2.
(13) Schramm, V. L.; Grubmeyer, C. Prog. Nucleic Acid Res. Mol. Biol. 2004,
78, 261-304.
(21) Chakrabarti, P. P.; Suveyzdis, Y.; Wittinghofer, A.; Gerwert, K. J. Biol.
Chem. 2004, 279, 46226-33.
(22) Deng, H.; Wang, J.; Ray, W. J.; Callender, R. J. Phys. Chem. B 1998, 102,
3617-23.
(23) Deng, H.; Lewandowicz, A.; Cahill, S. M.; Furneaux, R. H.; Tyler, P. C.;
Girvin, M. E.; Callender, R. H.; Schramm, V. L. Biochemistry 2004, 43,
1980-7.
(24) Deng, H.; Cahill, S. M.; Abad, J. L.; Lewandowicz, A.; Callender, R. H.;
Schramm, V. L.; Jones, R. A. Biochemistry 2004, 43, 15966-74.
9
7766 J. AM. CHEM. SOC. VOL. 128, NO. 24, 2006

Phosphate Activation by Purine Nucleoside Phosphorylase
A R T I C L E S
stretches of the bound phosphate and positive or negative
features due to the protein vibrational modes that are shifted
upon phosphate binding (Figure 1B). Since Figure 1B and 1D
clearly shows more than three phosphate modes, which is the
maximum number of P-O stretch bands that can be observed
for a phosphate in this spectral region, we conclude that there
are more than one binding mode for the PNP-bound phosphate.
The simplest interpretation of the spectra of PNP-bound
phosphate is that there are two binding modes, resulting in a
maximum of six P-O stretches. P-O stretch bands can be
identified by their frequency shifts upon 18O labeling of the
phosphate with the constraint that the isotope shift should be
in the range of 25-45 cm-1. On the basis of these criteria, six
bands at 980, 990, 1035, 1075, 1096, and 1105 cm-1 in Figure
1B can be assigned to the phosphate P-O stretches on the basis
of their 18O shifts (Figure 1D). Since the intensities of the two
symmetric P-O stretches at 980 and 990 cm-1 are similar,
indicating similar populations, the binding affinities of the two
types of phosphate should also be similar. Further attempts to
assign these P-O stretches to one or the other form of bound
phosphate by exploring different PO4 binding affinities did not
yield conclusive results because of insufficient signal-to-noise
at low phosphate concentrations. Nevertheless, the observation
of a P-O stretch band at 1035 cm-1, which is 45 cm-1 lower
than the asymmetric P-O stretch in solution, is consistent with
one form of bound phosphate having a P-O bond more strongly
hydrogen bonded than the other two P-O bonds (see below).
In the other form of bound phosphate, all three P-O bonds
experience similar hydrogen bonding strength, resulting in two
near degenerate asymmetric P-O stretches with similar fre-
quencies.
2-
Figure 1. (A) FTIR spectrum of 50 mM HPO4 at pH 9.5. (B) FTIR
difference spectrum of PNP‚PO4 complex minus PNP at pH 7.5 at room
2-
temperature. (C) FTIR spectrum of 50 mM [18O4]HPO4 at pH 9.5. (D)
Phosphate Binding in the PNP‚R1P Complex. Our first
attempt to obtain the isotope-edited difference FTIR spectrum
of PNP‚R1P failed, yielding a difference spectrum that was
similar to the bound phosphate spectrum shown in Figure 1B.
PNP slowly decomposes R1P to ribose and phosphate, as
confirmed by 1H and 31P NMR measurements (data not shown).
This reaction is not due to trace contaminants of phosphomo-
noesterase since the reaction is stopped by the addition of the
specific PNP inhibitor Immucillin-H. Fortunately, R1P hydroly-
sis is slow enough to permit FTIR experiments to be completed
with appropriate precautions.
FTIR difference spectrum of PNP‚[18O4]PO4 complex minus PNP at pH
7.5 at room temperature. Only nonbridging P-O stretch modes are observed
in the 900-1200 cm-1 spectral range. The enzyme samples were prepared
in 300 mM NaCl. The concentration of PNP:PO4 in these complexes was
4:3.5 mM.
free PNP was dialyzed against 0.3 M NaCl at pH 7.0 to remove
phosphate and concentrated to ∼4 mM by Centricon30 for IR
measurements.
FTIR spectroscopy was performed on a Magna 760 Fourier transform
spectrometer (Nicolet Instrument Corp., WI) using a MCT detector.
BaF2 windows with a 25 µm Teflon spacer were used. Spectra were
collected in the range of 900-4000 cm-1 with 2 cm-1 resolution. A
Blackman-Harris three-term apodization and a Happ-Genzel apodiza-
tion were applied, respectively. Omnic 4.1a (Nicolet Instruments, Corp.)
software was used for data collection and analysis.
Solution 31P NMR spectra were obtained at 121.5 MHz on a Bruker
DRX300 spectrometer; 2k scans were collected in each spectrum with
a sweep width of 14 ppm, sampled with 8k points, and a recycle delay
of 3.5 s.
Figures 2A and 3A show the FTIR spectra of R1P and 18O-
labeled R1P, respectively, in solution. At pH 9.0, R1P was stable
for the time of the measurement, and at this pH, >99% of the
phosphate group is dianionic. The R1P bands at 950 and 1095
cm-1 (Figure 2A) can be assigned to the symmetric and the
asymmetric P-O stretch modes, respectively, based on the 18
O
shifts (28 cm-1 for both modes; Figure 3A). It is interesting to
note that these frequencies are significantly different from those
of the dianionic phosphate P-O stretches in solution (Figure
1A). Figure 2B shows the time evolution of the PNP‚R1P
spectra from 1 to 20 min after mixing. These spectra were
obtained by collecting sequential FTIR spectra and by dividing
each spectrum by the last one. Division (rather than subtraction)
provides more consistent baselines to identify both positive and
negative bands, whose intensities decrease with time using this
difference procedure. The positive features are related to the
PNP‚R1P complex, and the negative features are related to the
PNP‚PO4 complex. Thus, the phosphate P-O stretch bands of
Results
2-
Phosphate Binding in the PNP‚PO4 Complex. For HPO4
in water (at pH 9.5, >99% of PO4 will be present as the dianion),
the three P-O stretches are coupled to form a symmetric stretch
at 991 cm-1 (Figure 1A) and a 2-fold degenerate asymmetric
stretch mode at 1080 cm-1 (Figure 1A). Upon 18O4 labeling of
the phosphate, these vibrational frequencies show -34 and -32
cm-1 shifts for symmetric and asymmetric modes, respectively
(Figure 1C). The difference spectrum between the PNP‚PO4
complex and PNP contains positive features due to P-O
9
J. AM. CHEM. SOC. VOL. 128, NO. 24, 2006 7767

A R T I C L E S
Deng et al.
18O shift (to 1105 cm-1 in Figure 3B). A component of the
1065 cm-1 band is assigned to another asymmetric P-O stretch
based on its decreased intensity and the appearance of a new
broad band near 1025 cm-1 upon 18O labeling of R1P (Figure
3B). The 960 cm-1 band in the spectra of PNP‚R1P is assigned
to the symmetric P-O stretch. Due to strong absorbance of
the IR window material and water below 950 cm-1, this stretch
band could not be observed for the PNP‚R1P complex with 18O-
labeled R1P. However, disappearance of this band upon 18O
labeling of R1P is consistent with the assignment (Figure 3B).
Many of the negative bands in Figure 2B are associated with
the P-O stretch modes of bound phosphate. Their counterparts
are readily identified in the bound phosphate spectrum (PNP‚
PO4; Figure 1B). Thus, both kinds of difference spectra (Figures
1B and 2B) are consistent with two phosphate binding modes
in the PNP‚PO4 complex. The symmetric P-O stretch of R1P
shifts up by 10 cm-1 upon binding to PNP, and the two
asymmetric P-O stretches shift in opposing directions by -30
and +41 cm-1 in the PNP‚R1P complex (Figure 2A and 2B).
The time evolutions for the major asymmetric P-O stretch
band at 1136 cm-1 of unlabeled R1P and the 1105 cm-1 band
of the 18O-labeled R1P fit to a single-exponential function (insets
of Figures 2B and 3B), yielding a rate constant of 0.023 ( 0.005
s-1 for PNP-catalyzed hydrolysis of R1P.
Figure 2. (A) FTIR spectrum of 50 mM R1P at pH 9.0. R1P is stable at
this pH for days. (B) The time-dependent, serial difference FTIR spectra
of the PNP‚R1P complex at pH 7.0. The first spectrum was recorded about
1 min after mixing, the following seven spectra were recorded at 14 s
intervals, and the subsequent 18 spectra were recorded at 55 s intervals.
Difference spectra were constructed by dividing each of the spectra by the
last one. Selected spectra are presented here. In this form, difference spectra
present the vibrational modes of bound R1P as positive bands, while the
vibrational modes of the ribose and phosphate products appear as negative
bands. All band intensities decrease with time and disappear as the reaction
reaches equilibrium. The inset to B shows a single-exponential curve fitted
to the time dependence of the band intensity at 1136 cm-1. The average
time constant determined from several measurements, including those with
[18O4]R1P, is 0.023 ( 0.005 s-1. The concentration of PNP:R1P in the
complex was 4:4 mM.
The Nature of the PNP-Catalyzed Decomposition of R1P.
Two mechanisms could account for R1P conversion to ribose
and phosphate: the bond cleavage of the CO-P bond or the
C-OP bond. It is important to determine the reaction mecha-
nism since, in the latter case, the distortion of the phosphate
moiety of R1P in the PNP‚R1P complex should be similar to
that in the PNP‚base‚R1P complex in the direction of nucleoside
synthesis.
These two cases can be distinguished by 31P NMR measure-
ments of the phosphate product from the PNP-catalyzed R1P
decomposition using 18O-labeled R1P.25 Each 18O labeling of
the four phosphate oxygens causes a ∼0.02 ppm upfield shift
of the 31P chemical shift.26 The 31P spectrum for a mixture of
unlabeled and 18O-labeled phosphate (3:1 molar ratio) shows a
small resonance ∼0.02 ppm downfield of the major 18O-labeled
phosphate resonance, consistent with <10% of the 18O-labeled
phosphate with three 18O atoms (Figure 4A). The number of
the 18O atoms on the phosphate product from R1P with PNP
was determined by a 31P measurement of the mixture of this
product (Figure 4B) and the known phosphate sample (Figure
4A). The major resonance of the products of 18O-labeled R1P
from PNP overlaps with the P18O4 resonance in the mixture,
establishing that it contains phosphate with four 18O atoms
(Figure 4C). Thus, the PNP-catalyzed decomposition of R1P
occurs by the cleavage of the C-OP bond, as in the PNP-
catalyzed nucleoside synthesis reaction. The nonenzymatic
hydrolysis of R1P in aqueous solution, pH 9.0, at 80 °C, involves
cleavage at both C-OP and CO-P bonds, resulting in equal
amounts of P18O4 and PO18O3 (Figure 4D and 4E).
Figure 3. (A) FTIR spectrum of 50 mM [18O4]R1P at pH 9.0. (B) Same
as in Figure 2B except [18O4]R1P is used.
Discussion
bound R1P are positive, while the P-O stretch bands of bound
PO4 are negative. Protein bands can appear either as positive
or negative because some protein vibrational modes change
when R1P is hydrolyzed to ribose and phosphate.
The major positive band at 1136 cm-1 is assigned to the
asymmetric P-O stretch of bound R1P based on its 31 cm-1
Phosphate Structure from Vibrational Spectroscopy.
Vibrational spectroscopy is a powerful tool to determine the
(25) Jordan, F.; Patrick, J. A.; Salamone, S., Jr. J. Biol. Chem. 1979, 254, 2384-
6.
(26) Cohn, M.; Hu, A. Proc. Natl. Acad. Sci. U.S.A. 1978, 75, 200-3.
9
7768 J. AM. CHEM. SOC. VOL. 128, NO. 24, 2006

Phosphate Activation by Purine Nucleoside Phosphorylase
A R T I C L E S
between PO4 and active site residues also contribute.9 In the
following sections, similar analyses are performed on the data
obtained with the PNP‚PO4 and PNP‚R1P complexes.
2-
Comparison of the Structures of the -PO3 Moiety in
R1P and Phosphate. The P-O stretch frequencies in dianionic
phosphate and in R1P are drastically different; the symmetric
stretches differ by 41 cm-1 and asymmetric stretches differ by
15 cm-1 (Figures 1A and 2A). However, the average P-O
stretch frequencies in these two compounds as defined above
differ by only 2 cm-1, corresponding to a very small change in
the average bond order or bond length. The average P-O
stretch frequency will be strongly influenced by the hydrogen
bonding strength on the P-O bonds and, to a smaller extent,
other factors such as the pKa of the alcohol substituent.29
Analysis using the methods given in Deng et al.22 suggests that
2-
the -PO3 moiety is more flattened in R1P than in dianionic
phosphate, and this geometry is largely responsible for the P-O
stretch frequency pattern difference in these molecules. The
geometry change does not cause a large change in the average
P-O stretch frequency but does cause a significant change in
the difference between symmetric and asymmetric P-O stretch
frequencies. When the -PO32- moiety becomes more flattened,
the frequency difference becomes larger. Molecular modeling
calculations based on ab initio methods suggest that a hydrogen
bond between the ribose C2-OH and one of the P-O bonds
in R1P is likely to cause this geometry change.
Figure 4. (A) 30:10 mM mixture of phosphate and 18O-labeled phosphate.
Sample contains 0.3 M NaCl and 2 mM EDTA. (B) The decomposition
product of R1P by PNP. EDTA was added to a final concentration of 2
mM. The sample concentration was 10 mM. (C) 31P spectrum of the sample
mixture of A and B. (D) The hydrolysis product of 10 mM R1P at 80 °C
at pH 9.0 for 18 h. EDTA was added to a final concentration of 2 mM
prior to NMR measurements. (E) 31P spectrum of the sample mixture of A
and D.
PO4 Distortion in the PNP‚PO4 Complex. Phosphate is
distorted in the ternary complex of PNP with ImmH,9 and the
current study investigates distortion of phosphate and R1P in
the binary complexes. From the six observed P-O stretch
frequencies, we detect two dianionic forms of PO4 bound to
PNP. Although not all of the vibrational modes could be
assigned, several important observations are noted as follows.
Normal mode calculations from phosphate model complexes
indicate that a strong hydrogen bond to one of the P-O bonds
causes the degenerate asymmetric P-O stretch band to split
into two bands with frequencies higher and lower than the
degenerate frequency. The stronger the hydrogen bonding, the
larger the splitting becomes.9 The P-O stretch mode at 1035
cm-1 in the complex is 45 cm-1 lower than the degenerate P-O
stretch mode at 1080 cm-1 in solution. Thus, one of the P-O
bonds in the first form of bound phosphate forms a stronger
hydrogen bond than the rest of the P-O bonds. In the second
form of the bound phosphate, the hydrogen bonds to all P-O
bonds are similar so that the asymmetric P-O stretches are
nearly degenerate. In the first form of bound phosphate, the
average P-O stretch frequency is approximately 6 cm-1 lower
than that in solution, establishing that the average hydrogen
bonding to bound P-O bonds is stronger.30 Since X-ray
structures of PNP with bound phosphate reveal only one binding
site for the PO4 group (see below), the two forms of bound
dianionic phosphate detected in our FTIR studies must locate
ionization state of phosphate under various environments,
especially in proteins, since the vibrational frequencies of the
P-O stretches are significantly different in each ionic state (for
recent applications, see refs 17 and 18). For P-O bonds of
phosphates, vibrational normal mode analysis has provided very
accurate empirical relationships for bond lengths (RP-O) and
bond orders (SP-O) as a function of vibrational frequency.22
They are given by RP-O ) [0.2838 × ln(224500/υ)] and SP-O
) [0.175 × ln(224500/υ)]-4.29, where υ is the average frequency
of the phosphate stretches defined as υ2 ) (υs2 + naυa )/(na +
2
1), where υs, υa, and na are the symmetric modes, asymmetric
modes, and the degeneracy of υa, respectively. For dianionic
phosphate, na ) 2. The error in the bond length/frequency
relationships in determining RP-O is estimated to be about
(0.004 Å, and bond orders (in valence units) to (0.04 vu.19,22
In addition, bond angle changes, as a measure of the flatness
2-
of the -PO3 moiety, can also be estimated from the P-O
stretch frequency changes.19,22
Phosphate Structure in a Transition State Analogue
Complex. In a previous study, the bond lengths of the P-O
bonds for HPO42- in solution and in complex with a transition
state analogue, ImmH, were characterized.9 One of the P-O
bonds in the PNP‚ImmH‚PO4 complex is strongly polarized and
is ∼0.027 Å longer than the other two P-O bonds. The bond
polarization is attributed to a ∼3.0 Å ion pair interaction with
N4′ of ImmH,8,27 which is a cation.28 Additional hydrogen bonds
(28) Sauve, A. A.; Cahill, S. M.; Zech, S. G.; Basso, L. A.; Lewandowicz, A.;
Santos, D. S.; Grubmeyer, C.; Evans, G. B.; Furneaux, R. H.; Tyler, P. C.;
McDermott, A.; Girvin, M. E.; Schramm, V. L. Biochemistry 2003, 42,
5694-705.
(29) Cheng, H.; Nikolic-Hughes, I.; Wang, J. H.; Deng, H.; O’Brien, P. J.; Wu,
L.; Zhang, Z. Y.; Herschlag, D.; Callender, R. J. Am. Chem. Soc. 2002,
124, 11295-306.
(30) This conclusion assumes that the higher symmetric stretch frequency and
the highest asymmetric stretch are associated with this phosphate. If other
frequencies are associated with this phosphate, the average frequency will
be even lower, corresponding to even stronger hydrogen bonding.
(27) Shi, W.; Basso, L. A.; Santos, D. S.; Tyler, P. C.; Furneaux, R. H.;
Blanchard, J. S.; Almo, S. C.; Schramm, V. L. Biochemistry 2001, 40,
8204-15.
9
J. AM. CHEM. SOC. VOL. 128, NO. 24, 2006 7769

A R T I C L E S
Deng et al.
Scheme 2. Hydrogen Bonding Distances to Phosphate in Selected Crystal Structures of PNP
Scheme 3. Relative Hydrogen Bond Energy to -PO32- in Various Complexes Based on the Average P-O Stretch Frequency Changes
from Solution
at the same place but differ only in their proton positions as
discussed below.
in this reaction strongly suggests that this reaction contains
features that are characteristic of transition state stabilization
leading to nucleophilic displacements at the cationic carbon of
ribose in normal PNP catalysis. The FTIR spectra indicate that
the phosphate moiety of bound R1P is dianionic. Of the three
P-O stretch frequencies, the symmetric stretch is blue shifted
Coordination with PNP Crystal Structures. The X-ray
structures of PNP active sites reveal that several PNP active
site residues and a structural water molecule interact with bound
anions (Scheme 2; refs 7, 8, 31, and 32). One P-O bond is
hydrogen bonded to the Nꢀ2 of His86, while the carboxyl group
of Glu89 is hydrogen bonded to the Nδ1 of His86 (Scheme 2).
Similar triads, which involve an Asp/Glu-His pair and another
titratable group, such as C-OH or P-OH, have been found in
other enzymes and are proposed to form a proton relay channel.
In PNP, this triad pattern allows the formation of a favorable
hydrogen bond to the P-O1 and also allows a change of the
protonation pattern for the bound phosphate dianion by proton
relay to the Glu89 carboxyl group (Scheme 2). The protonation
sites of phosphate are important since, in the phosphorolysis
direction, the phosphate nucleophile requires the O4 to be an
anion adjacent to the incoming oxacarbenium ion, while in the
nucleoside synthesis direction, the PO4 leaving group of the
PNP‚R1P complex must accommodate extra electron density
as the ribosyl group accepts the hypoxanthine nucleophile (see
below).
by 10 cm-1, one asymmetric stretch is blue shifted by 41 cm-1
,
and the other is red shifted by 30 cm-1, compared to their
solution values. Normal mode analysis based on ab initio
frequency calculations of phosphate model compounds indicates
that the breakdown of the frequency degeneracy of the two
asymmetric stretches can be explained by different hydrogen
bond strengths to the three P-O bonds in the PNP‚R1P
complex.9 The hydrogen bond to the P-O1 group from the
Nꢀ2 of His86 is proposed to be stronger than those to the other
P-O bonds. The average P-O stretch frequency in the PNP‚
R1P complex is ∼7 cm-1 higher than that of R1P in solution.
Considering that the average P-O bond order of R1P is higher
when bound to the enzyme than in solution, the average
hydrogen bonding strength to the P-O bonds is weaker in the
enzyme than in water. On the basis of leaving group potential,
this interaction would be expected to stabilize R1P from loss
of phosphate when bound to the enzyme. It is illuminating to
see that this does not occur since PNP catalyzes its hydrolysis.
PO4 Distortion in the PNP‚R1P Complex. The slow
decomposition of R1P catalyzed by PNP provides a useful probe
into the catalytic mechanism. The specific C-OP bond cleavage
Hydrogen Bond Patterns to PO4 and R1P. The interaction
pattern of the phosphate moiety in the PNP‚R1P complex is
similar to that of the first form of the bound phosphate in the
PNP‚PO4 complex, meaning that the hydrogen bonding to one
(31) Erion, M. D.; Takabayashi, K.; Smith, H. B.; Kessi, J.; Wagner, S.; Honger,
S.; Shames, S. L.; Ealick, S. E. Biochemistry 1997, 36, 11725-34.
(32) Mao, C.; Cook, W. J.; Zhou, M.; Koszalka, G. W.; Krenitsky, T. A.; Ealick,
S. E. Structure 1997, 5, 1373-83.
9
7770 J. AM. CHEM. SOC. VOL. 128, NO. 24, 2006

Phosphate Activation by Purine Nucleoside Phosphorylase
A R T I C L E S
of the P-O bonds is stronger than to the others. This similarity
protein active site residues at full strength. After the decomposi-
tion of R1P, the PO4 is fully hydrogen bonded and all hydrogen
bond energy with the phosphate moiety is released. During this
process, a proton from the active site water replaced the ribose
group of R1P and the hydroxyl group of the water forms a bond
with the leaving oxacarbenium ion. In our hypothesis, the first
form of bound phosphate is the phosphate product in the purine
nucleoside synthesis direction. By transferring a proton from
the protonated Nꢀ2 of His86 to its hydrogen bonded P-O1,
and with the aid from water molecules in the active site, the
bound phosphate is converted to the second form in which the
P-O4 oxygen anion points toward the ribose C1, as suggested
from the PNP‚PO4 complex (Scheme 3). The second form of
bound phosphate serves as the nucleophile substrate in the
phosphorolysis direction, and the P-O4 bond is available for
the formation of the C1-OP bond. The incoming oxacarbenium
ion polarizes the bond during the reaction, as established in the
studies of the transition state complexes of PNP‚ImmH‚PO4.7,9
Thus, our vibrational studies of PNP ground state complexes
and a transition state complex provide unique information on
how the phosphate is activated along the reaction coordinate.
2-
suggests that the -PO3 moiety in these complexes may
experience the hydrogen bonds from the same active site
residues. However, the average hydrogen bond strength to the
P-O bonds in the PNP‚R1P complex is weaker than to bound
phosphate, as indicated by the >12 cm-1 higher average P-O
stretch frequency compared to the first form of the bound
phosphate. It is also weaker than that of the R1P in solution, as
indicated by its 7 cm-1 higher average P-O stretch frequency.
2-
Unfavorable hydrogen bonding interactions to the -PO3
moiety indicate high potential energy relative to solution thus
a thermodynamically unstabilized phosphate leaving group.
Implications for the Reaction Mechanism. Crystal struc-
tures of PNP suggest that the protonated Nꢀ2 of His86 may
form a variable-strength hydrogen bond with the P-O1 bond,
while hydrogen bonds formed with other residues and a
structural water molecule complete the catalytic site contacts
for a stable, bound PO4 moiety (Scheme 2).7,16,33,34 The observed
P-O frequency patterns in the complexes studied here suggest
that, in the PNP‚R1P complex, the PO4 group is unfavorably
bound, presumably due to the steric crowding imposed by the
ribose moiety of R1P, thus unable to form hydrogen bonds with
Acknowledgment. Supported by research grants GM068036,
EB001958 (Robert Callender), and GM41916 (V.S.) from the
NIH.
(33) de Azevedo, W. F., Jr.; Canduri, F.; dos Santos, D. M.; Silva, R. G.; de
Oliveira, J. S.; de Carvalho, L. P.; Basso, L. A.; Mendes, M. A.; Palma,
M. S.; Santos, D. S. Biochim. Biophys. Res. Commun. 2003, 308, 545-52.
(34) Mao, C.; Cook, W. J.; Zhou, M.; Federov, A. A.; Almo, S. C.; Ealick, S.
E. Biochemistry 1998, 37, 7135-46.
JA0570281
9
J. AM. CHEM. SOC. VOL. 128, NO. 24, 2006 7771