NMR investigations of the static and dynamic structures of bisphosphonates on human bone: A molecular model

Article abstract of DOI:10.1021/ja0759949

We report the results of an investigation of the binding of a series of bisphosphonate drugs to human bone using ²H, ¹³C, ¹⁵N, and ³¹P nuclear magnetic resonance spectroscopy. The ³¹P NMR results show that the bisphosphonate groups bind irrotationally to bone, displacing orthophosphate from the bone mineral matrix. Binding of Pamidronate is well described by a Langmuir-like isotherm, from which we deduce an ～30-38 A² surface area per Pamidronate molecule and a ΔG = -4.3 kcal mol^-1. TEDOR of [ ¹³C₃, ¹⁵N] Pamidronate on bone shows that the bisphosphonate binds in a gauche [N-C(1)] conformation. The results of ³¹P as well as ¹⁵N shift and cross-polarization measurements indicate that risedronate binds weakly, since it has a primarily neutral pyridine side chain, whereas zoledronate (with an imidazole ring) binds more strongly, since the ring is partially protonated. The results of ²H NMR measurements of side-chain ²H-labeled Pamidronate, alendronate, zoledronate, and risedronate on bone show that all side chains undergo fast but restricted motions. In Pamidronate, the motion is well simulated by a gauche⁺/gauche^- hopping motion of the terminal -CH₂-NH₃⁺ group, due to jumps from one anionic surface group to another. The results of double-cross polarization experiments indicate that the NH₃⁺-terminus of pamidronate is close to the bone mineral surface, and a detailed model is proposed in which the gauche side-chain hops between two bone PO₄^3- sites.

Full text of DOI:10.1021/ja0759949

Published on Web 01/04/2008
NMR Investigations of the Static and Dynamic Structures of
Bisphosphonates on Human Bone: a Molecular Model
Sujoy Mukherjee,† Yongcheng Song,‡ and Eric Oldfield*,†,‡
Center for Biophysics and Computational Biology, 607 South Mathews AVenue, UniVersity of
Illinois at Urbana-Champaign, Urbana, Illinois 61801, Department of Chemistry, UniVersity of
Illinois at Urbana-Champaign, 600 South Mathews AVenue, Urbana, Illinois 61801
Received August 9, 2007; E-mail: eo@chad.scs.uiuc.edu
Abstract: We report the results of an investigation of the binding of a series of bisphosphonate drugs to
2
human bone using H, 13C, 15N, and 31P nuclear magnetic resonance spectroscopy. The 31P NMR results
show that the bisphosphonate groups bind irrotationally to bone, displacing orthophosphate from the bone
mineral matrix. Binding of pamidronate is well described by a Langmuir-like isotherm, from which we deduce
an ∼30-38 Å2 surface area per pamidronate molecule and a ∆G ) -4.3 kcal mol-1. TEDOR of [13C3, 15N]
pamidronate on bone shows that the bisphosphonate binds in a gauche [N-C(1)] conformation. The results
of 31P as well as 15N shift and cross-polarization measurements indicate that risedronate binds weakly,
since it has a primarily neutral pyridine side chain, whereas zoledronate (with an imidazole ring) binds
2
more strongly, since the ring is partially protonated. The results of H NMR measurements of side-chain
2H-labeled pamidronate, alendronate, zoledronate, and risedronate on bone show that all side chains
undergo fast but restricted motions. In pamidronate, the motion is well simulated by a gauche+/gauche-
hopping motion of the terminal -CH2-NH3+ group, due to jumps from one anionic surface group to another.
The results of double-cross polarization experiments indicate that the NH3+-terminus of pamidronate is
close to the bone mineral surface, and a detailed model is proposed in which the gauche side-chain hops
3-
between two bone PO4 sites.
1. Introduction
Bisphosphonate drugs such as pamidronate (1, Aredia),
risedronate (2, Actonel), zoledronate (3, Zometa) and alendr-
onate (4, Fosamax) are used to treat a variety of bone resorption
diseases, such as osteoporosis, Paget’s disease, and hypercal-
cemia due to malignancy.1 They are thought to act by inhibiting
the enzyme farnesyl diphosphate synthase,2 resulting in inhibi-
tion of protein prenylation in osteoclasts. In addition to their
use in treating these diseases, bisphosphonates have also been
found3 to activate γδ T cells (containing the Vγ2Vδ2 T cell
receptor) of the immune system, which then kill tumor cells;4
plus, they have direct activity against tumor cell growth,5 as
well as the growth of a variety of pathogenic protozoa6 and
some bacteria.7 There is thus interest in studying in more detail
† Center for Biophysics and Computational Biology.
‡ Department of Chemistry.
(1) Russell, R. G. Ann. N. Y. Acad. Sci. 2006, 1068, 367-401.
(2) Rondeau, J. M.; Bitsch, F.; Bourgier, E.; Geiser, M.; Hemmig, R.; Kroemer,
M.; Lehmann, S.; Ramage, P.; Rieffel, S.; Strauss, A.; Green, J. R.; Jahnke,
W. ChemMedChem. 2006, 1, 267-273.
(3) Kunzmann, V.; Bauer, E.; Wilhelm, M. N. Engl. J. Med. 1999, 340, 737-
738.
how bisphosphonates bind to human bone, since this could have
important implications for their future development in oncology
and as anti-infectives, where, in general, bone binding might
be expected to reduce efficacy.
(4) Wilhelm, M.; Kunzmann, V.; Eckstein, S.; Reimer, P.; Weissinger, F.;
Ruediger, T.; Tony, H. P. Blood 2003, 102, 200-206.
(5) Kunzmann, V.; Bauer, E.; Feurle, J.; Weissinger, F.; Tony, H. P.; Wilhelm,
M. Blood 2000, 96, 384-392.
(6) Martin, M. B.; Grimley, J. S.; Lewis, J. C.; Heath, H. T., 3rd; Bailey, B.
N.; Kendrick, H.; Yardley, V.; Caldera, A.; Lira, R.; Urbina, J. A.; Moreno,
S. N.; Docampo, R.; Croft, S. L.; Oldfield, E. J. Med. Chem. 2001, 44,
909-916.
In recent work, several groups have used chromatographic
and other methods to investigate how bisphosphonates bind to
hydroxyapatite,8-10 one component of bone mineral. However,
because bone contains ∼25-40% protein11 (mainly collagen)
(7) Leon, A.; Liu, L.; Yang, Y.; Hudock, M. P.; Hall, P.; Yin, F.; Studer, D.;
Puan, K. J.; Morita, C. T.; Oldfield, E. J. Med. Chem. 2006, 49, 7331-
7341.
9
1264
J. AM. CHEM. SOC. 2008, 130, 1264-1273
10.1021/ja0759949 CCC: $40.75 © 2008 American Chemical Society

NMR of Bisphosphonates on Bone
A R T I C L E S
2
and analysis. For H NMR, a solid echo pulse sequence (90(x - t -
90y -) was used and the signal was left shifted to the echo maximum
prior to data processing.22 13C NMR chemical shifts were referenced
with respect to the downfield peak of adamantane (taken to be 38.48
ppm from TMS); 15N shifts are reported with respect to a liquid NH3
standard taking an external reference of 1 M 15N2-urea in DMSO to be
at 79.4 ppm downfield from NH3,23 and 31P NMR chemical shifts were
referenced with respect to an external standard of 85% orthophosphoric
acid. All shifts use the IUPAC convention that high frequency,
paramagnetic, downfield or deshielded values are positive. Deuterium
line shape simulations were carried out by using the Turbopowder
program.24
and because there are only relatively low levels of hydroxya-
patite present in the mineral phase12,13 (the major component is
carbonatoapatite), it is clearly of interest to probe in a more
direct fashion how bisphosphonates bind to human bone.
Moreover, chromatographic methods give no direct information
on either the static or the dynamic structures of bound
bisphosphonates, or information on their protonation states. It
is likewise unclear whether binding is primarily a physisorption
process, or involves displacement of phosphate groups by the
anionic bisphosphonate ligands. Here, we thus report the results
of a series of liquid- and solid-state NMR investigations of the
interactions between bisphosphonates and human bone. Solid-
state NMR is an ideal technique with which to investigate the
structures (static and dynamic) of molecules bound to bone
surfaces, and has previously been used to investigate the
structures of bone mineral,12,13 as well as of a bone-seeking
peptide, statherin, interacting with hydroxyapatite.14 We show
that bisphosphonates bind tightly to human bone (displacing
PO43-); that their sidechains exhibit restricted mobility; that
protonation states can be observed directly, and that, in the case
of pamidronate, the bound drug conformation can be determined
experimentally, leading to a detailed molecular model for
pamidronate-bone binding.
2.2. Synthesis of 2H, 13C, and 15N-Labeled Bisphosphonates. 2.2.1.
General Procedure: A mixture of a carboxylic acid (1 mmol), H3PO3
(5 mmol) and toluene (4 mL) was heated to 80 °C with stirring. After
all solids melted, POCl3 (5 mmol) was added slowly and the mixture
stirred vigorously at 80 °C for 5 h.25 Upon cooling, the toluene was
decanted, and 6 N HCl (3 mL) was added to the residue. The resulting
solution was stirred for 6 h, after which most of the solvent was removed
in Vacuo. Isopropanol (25 mL) was added to precipitate a 1-hydroxy-
methylene bisphosphonate as a white powder, which was filtered,
washed with ethanol (5 × 5 mL), dried, and then further purified by
recrystallization from H2O/EtOH. All compounds had satisfactory C,
1
H, N microanalyses and H solution NMR spectra.
2.2.1.1. 1-Hydroxy-3-aminopropyl-1,1-bisphosphonic Acid-13C3,
15N (Pamidronic Acid-13C3, 15N). This compound was made from [13C3,
15N] â-alanine (Cambridge) (250 mg), following the above general
procedure (205 mg, 45%).
2. Materials and Methods
2.1. NMR Spectroscopy. Solid-state NMR spectra were obtained
by using the magic-angle sample spinning technique using 600 MHz
(1H resonance frequency) Infinity Plus spectrometers (Varian, Palo Alto,
CA) equipped with 14.1 T, 2.0 and 3.5 in. bore Oxford magnets and
Varian/Chemagnetics 3.2 and 4.0 mm T3 HXY probes. For 13C and
15N NMR experiments, spectra were obtained by using cross-polariza-
tion and with TPPM15 decoupling. Proton decoupled solid-state 31P
spectra were recorded with and without (for quantitation) cross
polarization, while liquid-state spectra were acquired with full proton
decoupling. For pamidronate, a two-dimensional 13C-13C correlation
spectrum was obtained by using radio frequency driven recoupling
(RFDR).16 A heteronuclear, broadband double cross polarization17
(DCP) experiment18 was performed for 13C-31P chemical shift cor-
relation of pamidronate on bone and TEDOR19 was used for N-C
distance determinations, on the same sample. All 13C, 15N, and 31P
spectra were obtained at ∼30 °C, with the exception of the TEDOR
and DCP spectra, which were obtained at 0 °C (to provide a slight
enhancement in sensitivity and minimize the probability of water loss
during the long data acquisition period). All 2D NMR experiments were
processed with NMRPipe20 and Sparky21 was used for visualization
2.2.1.2. 1-Hydroxy-3-aminopropyl-1,1-bisphosphonic Acid-2,2,3,3,-
2H4 (Pamidronic Acid-d4). This compound was made from â-alanine-
2,2,3,3-2H4 (C/D/N Isotopes, Quebec, Canada) (250 mg), following the
above general procedure (225 mg, 50%)
2.2.1.3. 1-Hydroxy-4-aminobutyl-1,1-bisphosphonic Acid-2,2,3,3,4,4-
2H6 (Alendronic Acid-d6). This compound was made from [2,2,3,3,4,4-
2H6]-4-aminobutyric acid (Aldrich) (250 mg), following the above
general procedure (210 mg, 47%).
2.2.1.4. 1-Hydroxy-2-(15N2]-imidazol-1-yl)-1,1-bisphosphonic Acid
(15N2-zoledronic Acid). This compound was made from [15N2]-
imidazole-1-acetic acid, which was prepared according to a published
procedure25,26 from [15N2]-imidazole (Cambridge) (260 mg), following
the above general procedure (265 mg, 26%).
2.2.1.5. 1-Hydroxy-2-([2,4,5,6-2H4]-pyridin-3-yl)ethylidene-1,1-
bisphosphonic Acid- (Risedronic Acid-d4). [2,4,5,6-2H4]-3-Pyridiny-
lacetic acid-was prepared according to a published procedure27 from
[2,4,5,6-2H4]-3-bromopyridine,28 followed by hydrolysis (1 N DCl,
reflux). [2,4,5,6-2H4]-risedronic acid was made from the acid obtained
(137 mg) following the above general procedure (185 mg, 65%).
2.2.2. Preparation of Bisphosphonate-Bone Samples. Human bone
tissue (non-demineralized bone powder, gun-shot victim, 45-125µm
particle size,) was obtained from the Pacific Coast Tissue Bank (Los
Angeles, CA) and was used without further treatment. Typically, 50
(8) Lawson, M. A.; Triffitt, J. T.; Ebetino, F. H.; Barnett, B. L.; Phipps, R. J.;
Locklin, R. M.; Russell, R. G. G. J. Bone Min. Res. 2005, 20, S396.
(9) Nancollas, G. H.; Tang, R.; Phipps, R. J.; Henneman, Z.; Gulde, S.; Wu,
W.; Mangood, A.; Russell, R. G.; Ebetino, F. H. Bone 2006, 38, 617-
627.
(10) Robinson, J.; Cukrowski, I.; Marques, H. M. J. Mol. Struct. 2006, 825,
134-142.
(20) Delaglio, F.; Grzesiek, S.; Vuister, G. W.; Zhu, G.; Pfeifer, J.; Bax, A. J.
Biomol. NMR 1995, 6, 277-293.
(11) Aerssens, J.; Boonen, S.; Lowet, G.; Dequeker, J. Endocrinology 1998,
139, 663-670.
(21) Goddard, T. D.; Kneller, D. G. University of California, San Francisco.
http://www.cgl.ucsf.edu/home/sparky/.
(12) Cho, G. Y.; Wu, Y. T.; Ackerman, J. L. Science 2003, 300, 1123-1127.
(13) Kolodziejski, W. New Techniques in Solid-State NMR. Top. Curr. Chem.
2005, 246, 235-270.
(22) Davis, J. H.; K. R. J.; Bloom, M.; Valic, M. I.; Higgs, T. P. Chem. Phys.
Lett. 1976, 42.
(23) Wishart, D. S.; Bigam, C. G.; Yao, J.; Abildgaard, F.; Dyson, H. J.; Oldfield,
E.; Markley, J. L.; Sykes, B. D. J. Biomol. NMR 1995, 6, 135-140.
(24) Wittebort, R. J.; Olejniczak, E. T.; Griffin, R. G. J. Chem. Phys. 1987, 86,
5411-5420.
(14) Goobes, G.; Goobes, R.; Schueler-Furman, O.; Baker, D.; Stayton, P. S.;
Drobny, G. P. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 16083-16088.
(15) Bennett, A. E.; Rienstra, C. M.; Auger, M.; Lakshmi, K. V.; Griffin, R. G.
J. Chem. Phys. 1995, 103, 6951-6958.
(25) Harel, Z.; Kovalevski-Liron, E.; Lidor-Hadas, R.; Lifshitz-Liron, R. Use
of certain diluents for making bisphosphonic acids. World Patent
WO03097655. November 27, 2003.
(16) Bennett, A. E.; Ok, J. H.; Griffin, R. G.; Vega, S. J. Chem. Phys. 1992,
96, 8624-8627.
(17) Hediger, S.; Meier, B. H.; Ernst, R. R. Chem. Phys. Lett. 1995, 240, 449-
(26) Gil, M. S.; Cruz, F.; Cerdan, S.; Ballesteros, P. Bioorg. Med. Chem. Lett.
1992, 2, 1717-1722.
456.
(18) Schaefer, J.; Mckay, R. A.; Stejskal, E. O. J. Magn. Reson. 1979, 34, 443-
447.
(27) Kondo, Y.; Inamoto, K.; Uchiyama, M.; Sakamoto, T. Chem. Commun.
2001, 2704-2705.
(19) Jaroniec, C. P.; Filip, C.; Griffin, R. G. J. Am. Chem. Soc. 2002, 124,
10728-10742.
(28) Sirimanne, S. R.; Maggio, V. L.; Patterson, D. G. J. Lab. Comp. &
Radiopharm. 1992, 31, 163-174.
9
J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008 1265

A R T I C L E S
Mukherjee et al.
mg of bone powder was suspended in 1 mL H2O and the bisphosphonate
was dissolved in 1 mL of H2O. The pH of both samples was adjusted
to the pH of interest, and the samples were then mixed, and the pH
adjusted (with concentrated NaOH or HCl) as needed. Samples were
incubated for 1 h and centrifuged, after which the supernatant was
removed. The bone-bisphosphonate pellets were then washed with 2
mL of Millipore water and packed into NMR rotors. In some cases,
the supernatant solutions were also investigated by solution NMR using
an external NaH2PO4 quantitation standard (doped with 10 µM FeCl3).
3. Results and Discussion
3.1. Chemisorption of Bisphosphonates to Bone. In Figure
1A-D, we show the 31P MAS NMR spectra of three different
bisphosphonates: zoledronate (3), pamidronate (1), and deoxy-
risedronate (5), bound to human bone at pH ) 7 together with,
inset, the 31P MAS NMR spectra of the pure bisphosphonates.
In each case, the bone spectra consist of two sets of resonances
(and their associated side-bands): a sharp feature at ∼3 ppm
corresponding to orthophosphate in the bone mineral, and a
second, broader and weaker feature at ∼15 ppm, due to the
bisphosphonate bound to bone. The bisphosphonate peak is
featureless, unlike those seen in the spectra of the corresponding
crystalline bisphosphonates, and has essentially the same width
and shift at 105, 104, and 103 ppm loadings, disappearing into
the noise level at ∼100 ppm (Figures 2A-D) No other peaks
were observed. These observations are similar to those made
by Josse et al.,29 who found that zoledronate (3) bound to a
variety of calcium phosphate compounds (â-tricalcium phos-
phate; calcium-deficient apatites), of interest as novel bioma-
terials for drug delivery, with a relatively broad, featureless line
shape.29 The lack of any discernible structure would, however,
be consistent with either physisorption or chemisorption. That
is, the bisphosphonates might undergo either relatively weak
electrostatic interactions with the inorganic components of bone,
together with hydrophobic interactions with the collagen
component, or they might be involved in the displacement of
anionic groups (OH-, CO32-, and PO43-), with the bisphos-
phonate PO3 groups substituting for these anions in the lattice.
And, of course, both types of interaction might occur. The
“chemical” displacement interaction might be expected to be
more likely, because bisphosphonates are known to have half-
lives in bone on the order of many years; plus, this displacement
mechanism would give at least a qualitative explanation of the
1
Figure 1. 31P MAS NMR spectra (600 MHz H resonance frequency) of
bisphosphonates bound to bone (A-D) together with (inset), spectra of the
corresponding crystalline bisphosphonates. All spectra were obtained by
using cross-polarization followed by TPPM decoupling during data
acquisition. (A) zoledronate on bone, 2 s recycle, 8192 scans and 10 kHz
spin speed; (B) pamidronate on bone, 5 s recycle, 1024 scans and
13.333 kHz spin speed; (C) deoxy-risedronate on bone, 2 s recycle, 16384
scans and 10 kHz spin speed; and (D) pamidronate on bone, 2 s recycle,
4096 scans and 8 kHz spin speed. All FIDs were processed with 50 Hz
line broadening.
3-
observation that high levels of PO4 are required to elute
bisphosphonates from hydroxyapatite in chromatographic ad-
sorption experiments.8 The NMR results support several aspects
of this strong chemisorption model. First, the overall spans of
the chemical shielding tensors are ∼120 ppm, similar to those
seen in crystalline bisphosphonates,30 indicating no significant
motional averaging of the chemical shift anisotropy. If only
physisorption were involved, then considerable narrowing the
31P spectral width would seem likely, due to fast motion on the
bone surface. Second, an analysis of the supernatants in our
experiments by solution 31P NMR shows that there is, in fact,
release of inorganic phosphate on bisphosphonate addition with,
on average, ∼1.25 bisphosphonates binding per phosphate
released, essentially the same effect as that seen with the
synthetic calcium phosphate compounds investigated by Josse
et al.29 So, bisphosphonates bind in an irrotational manner to
bone (or, at least, these are no fast, large-angle motions of the
bisphosphonate backbone), and this can be attributed, at least
in part, to displacement of inorganic phosphate by at least one
phosphonate group, enabling strong bisphosphonate interactions
with Ca2+ in the bone mineral matrix.
Interestingly, the area of the bisphosphonate surface peak
observed does not correlate in a linear way with the amount of
bisphosphonate added to the sample. Rather, our results indicate
that binding exhibits the type of sigmoid binding behavior seen
with chemisorption, as expected for example by the Langmuir-
(29) Josse, S.; Faucheux, C.; Soueidan, A.; Grimandi, G.; Massiot, D.; Alonso,
B.; Janvier, P.; Laib, S.; Pilet, P.; Gauthier, O.; Daculsi, G.; Guicheux, J.
J.; Bujoli, B.; Bouler, J. M. Biomaterials 2005, 26, 2073-80.
(30) Zhang, Y.; Oldfield, E. J. Phys. Chem. B 2006, 110, 579-586.
9
1266 J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008

NMR of Bisphosphonates on Bone
A R T I C L E S
in 1.41 mg of bisphosphonate being bound per 50 mg bone
sample, from which we deduce that the maximum coverage
corresponds to 4.68 mg of bound pamidronate. Given that the
mineral component of human bone has a surface area area of
∼110 m2/gram32,33 and that ∼60-75% of bone is composed of
this mineral component,11 we can deduce a surface area of
pamidronate on bone as: (bone mineral surface area in sample)/
(number of pamidronate molecules bound) ) (110 m2/g × 0.05
g × (0.6-0.75) × 1020 Å2)/ (6.023 × 1023 × 0.00468 g/270
g/mol (Na+pamidronate‚H2O Fwt)) ∼30-38 Å2/ molecule. This
value is close to that the 40 Å2/molecule that can be deduced
from the unit cell dimensions of monosodium pamidronate,34
and indicates that pamidronate molecules are quite highly packed
on the mineral component of human bone, at high coverage,
although as shown in Figure 1B, the 31P MAS NMR spectra of
crystalline pamidronate are of course quite different. From the
isotherm, we obtain ∆G ) -4.3 kcal mol-1 for pamidronate
binding to human bone mineral.
This type of chemisorption is only seen at pH ) 7, and
without any pH control, we find that a separate bisphosphonate
phase can form. For example, addition of the free zwitterionic
risedronic acid to bone (in H2O) results in the appearance of
new resonances at δ ) 15.0, 16.1 ppm, the same as seen in a
sample of Ca2+-risedronate prepared by adding Ca(NO3)2 to
risedronate at low pH. X-ray powder diffraction of such samples
at high bisphosphonate loading reveals sharp diffraction peaks
that are the same as those found in Ca2+-risedronate, indicating
formation of a new nanocrystalline phase, due to displacement
of bone Ca2+. This result is similar to that observed by Josse et
al.29 on zoledronate addition to a calcium phosphate phase, but
seems unlikely to be of physiological relevance. So, the results
of Figure 2 support the idea that, at pH ≈ 7, bisphosphonate
chemisorbs to human bone mineral surface, forming, at high
concentrations, a monolayer.
In addition to displacing Pi and interacting with Ca2+, there
is also now evidence for the importance of the 1-OH group on
the bisphosphonate backbone in binding to hydroxyapatite,8 as
well as the importance of side chain charge and location.8 This
is evidenced by differences in the affinity for hydroxyapatite
in chromatographic experiments by, e.g., zoledronate and
risedronate,8 whose side chains have different pKa values, as
well as by 31P NMR results, on bone. At pH ) 4.0, the pyridine
ring in risedronate is protonated, since the pKa value of the
pyridine ring is ∼5.7, and this protonation enables the ready
cross-polarization of [15N]-risedronate bound to bone, as shown
in Figure 3A. The 31P MAS NMR spectrum also indicates strong
bone binding, as seen in Figure 3B. At pH ) 6.0, however, the
15N NMR cross polarization peak intensity is much weaker
(about the same as the intensity arising from the 15N-sites in
collagen35), Figure 3C, although the 31P peak from the chemi-
sorbed bisphosphonate is still pronounced (Figure 3D). This is
consistent with a decrease in bone adsorption due to the fact
that the pyridine ring nitrogen is now (on average) far less
protonated than that at pH ) 4.0. At pH ) 8.0, the pyridinium
15N peak is no longer detectable (Figure 3E) although the surface
Figure 2. 31P MAS NMR spectra of pamidronate on bone and Langmuir
adsorption isotherm. (A), 3 × 105 ppm pamidronate solution used in sample
preparation; (B) 104 ppm solution concentration; (C) 103 ppm; (D), 102
ppm + ) bone mineral PO4 peak, * ) bisphosphonate peak. (E),
Langmuir adsorption isotherm. Y-axis at left indicates mg pamidronate
3-
bound, y-axis at right indicates θ (θ ) 1 for monolayer coverage).
like binding isotherm for the reaction:
BP(solution) + S {kk } BP•S (bound)
1
-1
where BP ) bisphosphonate, S ) bone surface, and the surface
coverage, θ, is given by the following:
θ ) K[c]/(1+K[c])
where [c] is the concentration of bisphosphonate in solution
and K ) k1/k-1. At low [c], θ [c], whereas in the limit [c] f
∞, θ)1. The binding of the salivary protein statherin to HAP
has also been shown to follow Langmuir-binding behavior,31
from which information on the thermodynamics of binding can
be deduced. We thus obtained 31P MAS NMR spectra of
pamidronate bound to bone at eight concentrations (representa-
tive spectra are shown in Figures 2A-D), then plotted the
relative bisphosphonate: bone area versus the amount of BP
(pamidronate) used and fitted the data to a rectangular hyperbolic
function, Figure 2E [the Y axes are bound pamidronate (left)
and θ (right), respectively]. The graph can be calibrated by using
our observation that the addition of 2.7 mg pamidronate results
(32) Holmes, J. M.; Davies, D. H.; Meath, W. J.; Beebe, R. A. Biochemistry
1964, 3, 2019-2024.
(33) Kolmas, J.; Slosarczyk, A.; Wojtowicz, A.; Kolodziejski, W. Solid State
Nucl. Magn. Reson. 2007, 32, 53-58.
(34) Stahl, K.; Treppendahl, S. P.; Preikschat, H.; Fischer, E. Acta Cryst. Section
E-Struct. Reports Online 2005, 61, M132-M134.
(31) Goobes, R.; Goobes, G.; Shaw, W. J.; Drobny, G. P.; Campbell, C. T.;
Stayton, P. S. Biochemistry 2007, 46, 4725-4733.
(35) Naito, A.; Tuzi, S.; Saito, H. Eur. J. Biochem. 1994, 224, 729-734.
9
J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008 1267

A R T I C L E S
Mukherjee et al.
Figure 3. 15N and 31P MAS NMR of bisphosphonates on bone. (A-F), 15N (A, C, E) and 31P (B, D, F) MAS NMR spectra of [15N]-risedronate on bone
at the pH values indicated. (G) 15N MAS NMR spectrum of [15N2]-zoledronate on bone, pH ) 7.0. The natural abundance level 15N signals from collagen
are indicated. At pH ) 7.0, there is a small resonance from N3 of the neutral imidazole side chain, at 245 ppm. Spectra were typically obtained by using
a 3.38 µs 90° pulse, a mix pulse of 6 ms (15N NMR) or 1 ms (for 31P NMR), followed by data acquisition for 1024 points with a dwell time of 20 µs. A
pulse delay of 3 s (for 15N) or 2 s (for 31P) was used. FIDs were zero-filled to 2048 points and processed with 200 Hz (15N NMR) or 50 Hz (31P NMR)
line-broadening prior to Fourier transformation. All 31P NMR spectra were obtained with 8192 scans, while the 15N spectra required from 65536 to 100 000
scans. (H) Comparison plot of retention times of bisphosphonates (zoledronate (3), ortho-risedronate (6), risedronate (2) and deoxy-risedronate (5) on HAP
obtained by column chromatography8 versus the ratio of the bisphosphonate/ bone 31P peak areas; R2 ) 0.88.
bisphosphonate peak is still observable (Figure 3F) but is clearly
less intense than that at the lower pH values. Clearly, these 15
of bound bisphosphonates might of course be guessed from pKa
values alone, this approach clearly does not work with bispho-
sphonates bound to their target FPPS enzyme, where both
risedronate and zoledronate bind in a fully protonated state36
(at pH ) 7.0), even though there are no obvious nearby proton
N
and 31P NMR results indicate that side-chain deprotonation is
associated with weaker bone binding. In the case of [15N2]-
zoledronate bound to bone at pH ) 7.0, the 15N MAS NMR
spectrum (Figure 3G) is more intense than that seen with
risedronate at either pH ) 6 or 8, consistent with the fact that
the imidazole side chain of zoledronate is expected to have a
pKa ≈ 6.7 and is, therefore, more highly protonated, resulting
in increased cross polarization as well as enhanced electrostatic
interactions with anionic groups on the bone surface. On the
basis of integrated 31P CP-MAS NMR peak areas, we find that
pamidronate and zoledronate bind most strongly to human bone
(28.7, 23% bisphosphonate: bone peak areas), followed by
ortho-risedronate (6) and alendronate (17.5, 16.9%), with weaker
binding of risedronate and deoxy-risedronate (10.4, 10.7%). For
the four compounds (2, 3, 5, 6) where chromatographic retention
times on HAP are available,8 these results correlate quite well
(R2 ) 0.88, Figure 3H) with the HAP retention times, suggesting
that binding to the collagen component in bone is not significant,
at least for these bisphosphonates. Although the protonation state
sources present in the active site, warranting therefore these 15
N
NMR experiments on bone. But what are the structuressstatic
and dynamic, of these adsorbed bisphosphonates?
3.2. Static and Dynamic Structures of Bound Bisphos-
phonates. We thus next investigated the structures (static and/
or dynamic) of several bisphosphonates bound to bone. We
elected to begin by investigating the static and dynamic structure
of pamidronate (1, Aredia), since this is the simplest nitrogen-
containing bisphosphonate drug and can be readily prepared as
the [13C, 15N]-labeled isotopomer from commercially available
[13C3, 15N] â-alanine. The 1D 13C CP-MAS NMR spectrum of
[13C3, 15N]-pamidronate is shown in Figure 4A, and consists of
the resonance of C1 (J-coupled to P1 and P2) at 73 ppm, with
(36) Mao, J. H.; Mukherjee, S.; Zhang, Y.; Cao, R.; Sanders, J. M.; Song, Y.
C.; Zhang, Y. H.; Meints, G. A.; Gao, Y. G.; Mukkamala, D.; Hudock, M.
P.; Oldfield, E. J. Am. Chem. Soc. 2006, 128, 14485-14497.
9
1268 J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008

NMR of Bisphosphonates on Bone
A R T I C L E S
Figure 4. 13C and 15N MAS NMR spectra of [13C3, 15N] pamidronate. (A) 13C NMR spectrum of pamidronate acquired by using a 3 µs 1H 90° pulse width,
2 ms mix time, 50 µs dwell time, 2048 points, 10 s recycle delay, 16 scans, and a 13.333 kHz spin speed. Spectrum was zero-filled to 4096 points and
processed with 50 Hz line broadening. (B) 13C-13C RFDR correlation spectrum of [13C3, 15N] labeled pamidronate. (C)-(E) 13C NMR spectra of (C)
pamidronate on bone; (D) bone; and (E) difference between (C) and (D). All 13C NMR spectra obtained using a 3 µs width 90° pulse, 2 ms mix time, 10
µs dwell time, 2048 points, 10 s recycle delay, 2048 scans and a 13.333 kHz spin speed. Data zero-filled to 4096 points and processed with 50 Hz line
broadening. (F) 15N NMR spectrum of [13C3, 15N]-pamidronate on bone, 4 µs 1H 90° pulse width, 5 ms mix time, 50 µs dwell time, 2048 points, 4 s recycle
delay, 20 388 scans, and a 13.333 kHz spin speed. Data zero-filled to 4096 points and processed with 50 Hz line broadening.
the C(2), C(3) peaks to high field.37 These resonances can be
unambiguously assigned from an RFDR spectrum (Figure 4B)
with C(2) being the most highly shielded peak. When bound to
bone, there is a substantial broadening of each resonance (Figure
4C); plus, there are additional small peaks observed. These arise
from the glycine, alanine, proline, and hydroxyproline residues
in collagen,38 as shown in Figure 4D. A difference spectrum
([bone + bisphosphonate]-bone) is shown in Figure 4E, and
represents the spectrum of pamidronate alone, when bound to
bone. The 13C(3) line width is small, but that of C(1) (the
bisphosphonate backbone carbon) is larger, and that of C(2) is
larger still. For C(1), much of the apparent broadening must
arise from J-coupling to 31P (see Figure 3A), but this cannot
contribute to the breath of C(2). Rather, it seems that here, there
is a chemical shift dispersion due, most likely, to γ-gauche
interactions with a heterogeneous population of phosphonate
OH/O- groups. As expected, the 15N NMR spectrum contains
only a single intense resonance, having a chemical shift of 36.5
ppm (from NH3), consistent with a fully protonated alkyl-
ammonium group (Figure 4F). So, 13C and 15N MAS NMR
spectra of pamidronate on bone can be readily obtained, opening
up the possibility of determining the structure of this bisphos-
phonate, on human bone. To do this, we employed the z-filtered
TEDOR method19 to determine the distances between C(1) and
N, which give, in principle, the backbone conformation.
In crystalline sodium pamidronate,39 the bisphosphonate side
chain adopts a close-to-trans C(1)-C(2)-C(3)-N configura-
tion, but on bone, the conformation is unknown and could, in
principle, be either a trans or a gauche configuration (gauche+
or gauche-), with the gauche conformations potentially facilitat-
ing electrostatic interactions between the terminal ammonium
group in the bisphosphonate with anionic groups on the bone
surface, due to its “bent” conformation. It also seemed possible
that the side chain might actually be quite dynamic, jumping
from one anionic bone surface site to another. To deduce the
likely backbone conformation, we used the TEDOR experi-
ment.19 We first investigated as a reference sample [13C3, 15N]-
labeled pamidronate, diluted (9%) into natural abundance
pamidronate, and cocrystallized as the sodium salt (sample
characterized via powder X-ray diffraction). In Na-pamidronate,
the 154° (∼trans) side-chain conformation corresponds to a 3.82
(37) Harris, R. K.; Merwin, L. H.; Hagele, G. Magn. Reson. Chem. 1989, 27,
470-475.
(38) Reichert, D.; Pascui, O.; deAzevedo, E. R.; Bonagamba, T. J.; Arnold, K.;
Huster, D. Magn. Reson. Chem. 2004, 42, 276-284.
(39) Vega, D.; Fernendez, D.; Ellena, J. A. Acta Crystallogr., Sect. C 2002, 58,
M77-M80.
9
J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008 1269

A R T I C L E S
Mukherjee et al.
Figure 6. 92 MHz 2H NMR quadrupole echo NMR spectra of 2H-labeled
bisphosphonates on bone and as crystalline solids. (A-D), bisphosphonates
at 30 °C; (A), [2H4]-pamidronate, (B), [2H6]-alendronate, (C), [ring-2H3]-
zoledronate, and (D), [ring-2H4]-risedronate; (E-H), bisphosphonates on
bone at -50 °C, in the same order as in (A-D); (I-L), bisphosphonates
on bone at 30 °C, in the same order as in (A-D); and, (M-P), deuterium
line shape simulations (shown in red) of bisphosphonates bound to bone at
30 °C, in the same order as in (A-D), superimposed on the experimental
30 °C spectra. A small narrow central peak in each case arises from HO2H.
The 90° pulses were typically 2 µs. The interpulse spacing was 40 µs. There
was a 25.9 µs delay prior to data acquisition, and FIDs were left shifted to
the echo maximum. A 2000 Hz linebroadening was generally applied. The
number of scans for bisphosphonates on bone were in the range from 14
810 to 102 819, recycle times were in the range 1-5 s; recycle times for
the crystalline bisphosphonates were typically 100 s with ∼100 scans being
acquired. The kinetic rate (sec-1), conformer ratio, η, QCC (kHz) and jump
angles (degrees) were as follows: pamidronate, 3 × 107, 1:1, 0.05, 175,
(65; alendronate, 5 × 106, 1:1, 0.05, 167, (60 (for both the side-chain
torsion angles); zoledronate, 5 × 106, 1:1, 0.068, 185, (35; risedronate, 3
× 106, 1:4, 0.05, 185, (60; A 50% attenuation of the Câ, Cδ CH2 groups
in alendronate was used to reflect their larger T1 values. In the case of
zoledronate, the intensity used for 2H(5) was only 20%, due to 2H/1H
Figure 5. 13C-{15N} z-filtered TEDOR results for (A) [13C3, 15N] labeled
pamidronate diluted with 9X natural abundance pamidronate and crystallized
as the Na+ salt; (B) pamidronate on bone at 0 °C. Both samples were spun
at 10 kHz, and the recycle time was 2 s. A z-filter period of 200 µs
1
(corresponding to twice the rotor period) with 10 kHz H decoupling was
1
exchange during synthesis (based on 500 MHz solution H NMR).
used while an 80 kHz 1H field was applied for both the CW and TPPM
decoupling periods. Peak intensities are normalized to those in a spectrum
obtained with a 90° pulse on 1H, followed by cross polarization to 13C and
conformers would be consistent with the NMR results and
because the two forms are mirror images, neither would be
preferred (and indeed, as discussed below, it appears that the
two forms interconvert, via 2-site jumps). In any case, the
TEDOR results strongly suggest that the pamidronate side chain
maximizes its electrostatic interaction with the bone mineral
1
data acquisition. In (A), the 90° pulse widths were 3 µs for H, 4.5 µs for
13C and 7 µs for 15N. The 15N REDOR 180°-pulse was 14 µs while the 13
C
refocusing 180° pulse was 9 µs. 8192 scans were used for each point. In
1
(B), the 90° pulse widths were 3.5 µs for H, 5 µs for 13C and 7.5 µs for
15N. The 15N REDOR 180°-pulse was 15 µs, the 13C refocusing 180° pulse
was 10 µs. 16 384 scans were used for each point. The experimental results
were fit to eqs 8-12 in Jaroniec, et al.19 by using the Mathematica program.
In both (A) and (B), the simulation in blue corresponds to the C(3)-N
distance, while red corresponds to the C(1)-N distance, and the dotted
lines indicates 95% confidence intervals for the simulation. The typical range
of the amplitude scaling factor, λ, was ∼0.2-0.5 and the relaxation rate τ
was ∼70-210 Hz. (C) NMR derived structure of pamidronate on bone.
The side chain (C1-C2-C3-N) torsion angle is (71-77°; d ) 3.2 Å.
+
surface (carbonatoapatite and hydroxyapatite) via NH3
(pamidronate)-PO4 (or CO32-) (bone mineral) interactions.
3-
That is, the bisphosphonate side chain is “bent” over, in order
to enhance this interaction.
Of course, the interactions of the bisphosphonate side chains
with bone could be quite dynamic, even though the bisphos-
phonate backbones are not highly mobile. To investigate this
Å C(1)-N distance. The results of a TEDOR experiment on this
magnetically dilute sample, Figure 5A, yielded a C(1)-N
distance of 3.90 Å, in quite good accord with the crystal-
lographic result (3.82 Å). In contrast, for pamidronate bound
to the surface of human bone mineral, the TEDOR result (Figure
5B) yielded a C(1)-N distance, d ) 3.2 Å. This is clearly very
short and translates to a close to gauche backbone conformation
(ø ) (77°, using the 1.49 Å X-ray C(3)-N bond length; (71°
using the 1.59 Å TEDOR-derived C(3)-N bond length, Figure
5B), shown in Figure 5C. Of course, both gauche+ and gauche-
2
possibility, we determined the H NMR spectra of side chain
2H-labeled pamidronate, alendronate, zoledronate, and risedr-
2
onate, bound to bone. Reference H NMR spectra for each of
the pure bisphosphonates at 30 °C are shown in Figure 6A-D,
with each clearly indicating the absence of any significant side
2
chain mobility at 30 °C, on the ∼10-5 s time scale of the H
NMR experiment. Essentially the same spectra are observed
for each bisphosphonate when bound to bone, at -50 °C (Figure
9
1270 J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008

NMR of Bisphosphonates on Bone
A R T I C L E S
Figure 7. Models for bisphosphonates on bone depicting proposed side-
chain motions: nitrogen ) cyan; deuterium ) yellow; carbon ) gray;
oxygen ) red; phosphorus ) pink. (A), pamidronate gauche+/gauche- side-
chain motion leading to a backbone torsional oscillation of ∼(65°; (B),
alendronate undergoing tetrahedral or “mirror image” jump motion; (C),
zoledronate undergoing (30° ring wobbles and, (D), risedronate undergoing
(60° ring wobbles. With the exception of risedronate, all the side chains
are shown protonated. (E), Schematic of pamidronate on bone depicting
Figure 8. Molecular models for pamidronate-bone (HAP) interactions. (A),
Top view of pamidronate bound in gauche+/gauche- conformations to a
PO43- in one of the three 31P-triangles in the unit cell of HAP. (B, C), two
views of pamidronate (carbon in brown; oxygen in red and hydrogen in
cyan) side chain NH3+ (dark blue) binding to one or other of the PO43- (P
) purple) groups (P2 or P3) in the “P3-triangle”, the bisphosphonate PO3
group occupies the third site (P1). The 1-OH group interacts with a Ca2+
(green) below. The 13Cγ-31PO43- distance is much shorter than the 13Câ-
2
motional averaging model for Hγ.
6E-H). However, at 30 °C, there are clearly spectral line shape
changes (Figure 6I-L), indicating the presence of side chain
mobility. We thus next performed 2H NMR line shape
simulations for all four 2H-labeled bisphosphonates on bone (at
30 °C), as shown in Figure 6M-P, in which the line shape
simulations (in red) are superimposed on the experimental
spectra (in black). The models used are shown in Figure 7A-
D. In the case of pamidronate, the results of the TEDOR
experiments indicated the likely existence of a gauche confor-
mation, with a C(1)-N torsion angle of ∼(71-77° (∼gauche+,
∼gauche-). As gauche+ and gauche- are equally energetically
likely, this suggested a simple two-site model (Figure 7A) in
which gauche+ and gauche- forms interconvert, and the results
of a 2H-line shape simulation in which the -CH2-NH3+ group
undergoes a 2-fold hop ((65°) between gauche+ and gauche-
3-
3-
PO4
and 13CR-31PO4
distances and the two conformations and
interactions shown are consistent with the TEDOR, 2H and DCP NMR
results.
evidence for restricted motion that freezes out at low temper-
atures (Figure 6H), with the results of a line shape simulation
(Figure 6P) suggesting a (60° ring wobble with a 1:4 conformer
population, at 30 °C. While further work with, for example,
specifically 2H-labeled bisphosphonates, will be needed in order
to obtain more detailed motional models, what is clear from
the results of Figures 6 and 7 is that the side chains of each
bisphosphonate undergo fast (but restricted) motions when
bound to bone. In contrast, none of the crystalline bisphospho-
nates undergo fast side-chain motion, and all of the side chain
motions on bone are frozen out at -50 °C. In the case of
pamidronate, fast gauche+/gauche- isomerization is indicated
at ∼3 × 107 s-1, due perhaps to the -CH2-NH3 group
+
sampling two different anionic binding sites, was found to be
in good agreement with experiment (Figure 6M). In the case of
alendronate, the line shape is very different from that seen with
2
(at 30 °C) from the H NMR line shape simulation and is
2
pamidronate, and closely resembles that seen previously in H
consistent with the TEDOR result (at 0 °C), which also indicates
the presence of a gauche conformer. In this case, only the C-3
or γ-protons are affected, as illustrated in Figure 7E. For
alendronate, the line shape is similar to that seen previously in
gel state lipids, and suggests restricted tetrahedral jumps (e.g.,
between the “mirror-images” shown in Figure 7B). For zoledr-
onate and risedronate, the situation is more complex, since both
protonated and neutral rings could be present. However, there
is clearly no evidence for rapid 180° ring flips in either
zoledronate or risedronate, since these would be expected to
result in η ≈ 0.65 line shapes, as observed, e.g., for phenyla-
lanine residues in proteins and in zwitterionic phenylalanine.41
This can of course be attributed to the presence of the basic
2
NMR spectra of H-labeled gel-state lipids,40 where the CH2
groups undergo fast but restricted two-site hopping. One possible
motion for the alendronate side chain, shown in Figure 7B,
would be a “mirror-image” jump in which all three CH2 groups
+
undergo motional averaging. In this case, the NH3 group
remains in the same location, but there are coupled rotations at
each site in the side chain and again, the results of a line shape
simulation are in good agreement with experiment, Figure 6N.
For zoledronate, we find good accord (Figure 6O) for a (30°
ring wobble (Figure 7C) and for risedronate, there is again
(40) Huang, T. H.; Skarjune, R. P.; Wittebort, R. J.; Griffin, R. G.; Oldfield, E.
J. Am. Chem. Soc. 1980, 102, 7377-7379.
9
J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008 1271

A R T I C L E S
Mukherjee et al.
Figure 9. Double cross-polarization spectrum (1H f 13C f 31P) of [13C3, 15N] pamidronate on human bone. The experiment was performed at a spinning
1
1
speed of 10 kHz with a H 90° pulse width of 3.5 µs, a H f 13C mix time of 0.8 ms followed by t1 evolution, a 13C f 31P mix time of 6 ms followed by
data acquisition using 2048 points and a dwell time of 10 µs. Power levels for the 13C f 31P transfer were matched to ωC ) (5/2)ωR and ωP ) (7/2)ωR in
order to facilitate a broadband transfer covering all the peaks of interest in both dimensions. The indirect dimension was sampled with 64 real points
(TPPI44) with a dwell time of 20 µs and 2048 scans were acquired per row with a recycle delay of 2 s. TPPM decoupling of ∼80 kHz (6 µs and 16° phase
shifted) was used during chemical shift evolution and acquisition periods while ∼75 kHz CW decoupling was used during the 13C f 31P transfer. Two such
3-
experiments with a total acquisition time of ∼6 days were added before processing. The cross-peak volume for 13Cγ-31PO4 is large and indicates close
proximity between 13Cγ and bone 31PO43- groups. There are only small or no significant cross-peak volumes for 13Câ, 13CR and 31PO43-, due to longer (3.75,
3-
3.90 Å) 13C-31P distances as compared to 13Cγ-31PO4 (3.03 Å, on average, from the models shown in Figure 8).
nitrogen atoms in both bisphosphonates, which can interact
electrostatically or via hydrogen bonds, with bone surface
groups.
as shown in Figure 8B,C. This packing arrangement gives an
immediate explanation for the presence of a gauche side chain
configuration, because in this conformation, the terminal +NH3
group can interact electrostatically with either one of the two
remaining PO43- groups (P2/P3) in the “P3-triangle”, as shown
in Figure 8B,C. The conformations of the side chains are either
3.3. Molecular Model for Pamidronate on Bone. Finally,
we propose a detailed molecular model for pamidronate binding
to the mineral component of human bone at pH ) 7 at 300 K.
As can be seen from the results presented above, there are
numerous pieces of the puzzle that need to be assembled.
Pamidronate binds in a chemisorption process, releasing ∼1
inorganic phosphate per bisphosphonate bound, and this process
follows a classical first-order Langmuir isotherm, from which
we deduce a surface area per pamidronate of ∼30-38 Å2 at
θ)1. The results of the TEDOR experiment indicate that the
pamidronate side chain adopts a gauche as opposed to a trans
2
gauche+ or gauche-, and based on the H NMR results, we
propose that these conformations can rapidly interconvert, at
30 °C. The model also reveals that the 1-OH group is adjacent
to a Ca2+, explaining the importance of this group for bone
binding since a sizable ion-dipole interaction would be expected.
To try to further test this model, we carried out a double cross
polarization experiment (1H f 13C f 31P) on [13C3, 15N]
pamidronate on bone: the result is shown in Figure 9. As can
be seen in the Figure, there is a major cross-peak between the
2
conformation (at 0 °C), whereas the H NMR results indicate
13Cγ and 31PO4 bone mineral peaks, a smaller cross-peak
3-
that at 30 °C there is ((65°) hopping between 2 states and that
this process is frozen out at -50 °C. The simplest explanation
for these observations would of course be that the side chain
conformational change is gauche+ T gauche-, and we see from
Figure 7E that this “hop” affects (or motionally averages) just
the Cγ protons.
Next, we need to see how the bisphosphonate PO3 group
might bind to bone mineral. To do this, we inspected the X-ray
crystallographic structure of hydroxyapatite.42 The structure is
characterized by several relatively close-packed triangular
arrangements of PO43- groups, and we propose that one of the
two phosphonate groups in a bisphosphonate (P atoms are shown
in purple in Figure 8A) displaces one bone mineral PO43-(P1),
between 13Câ and 31PO43-, but no observable cross-peak between
13CR and bone mineral. The 13Cγ-31P cross-peak volume is by
far the largest and is consistent with the short (average) 13CR-
31PO43- distance of ∼3.03 Å found from the model (Figure 8),
whereas the 13Câ and 13CR-31PO43- distances are clearly much
longer, ∼3.75 and 3.9 Å, respectively. These results therefore
strongly support a “bent-over” side chain conformation in which
the +NH3 group interacts with the bone surface, consistent with
the TEDOR and 2H NMR results which indicate gauche+/
gauche- side-chain conformations.
4. Conclusions
In summary, the results discussed above can be viewed
together to provide the first detailed spectroscopic picture of
how bisphosphonates bind to human bone. First, and indepen-
dent of any molecular model of binding, bisphosphonate
(41) Kinsey, R. A.; Kintanar, A.; Oldfield, E. J. Biol. Chem. 1981, 256, 9028-
9036.
(42) Wilson, R. M.; Elliott, J. C.; Dowker, S. E. P. Am. Mineral. 1999, 84,
1406-1414.
9
1272 J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008

NMR of Bisphosphonates on Bone
A R T I C L E S
3-
backbones bind to bone in a “rigid” or irrotational manner, at
least on the time scale of the inverse of the 31P chemical
shielding tensor anisotropy (∼µs). That is, the overall breadths
of the 31P shielding tensors in crystalline bisphosphonates are
very similar to those found for bisphosphonates on bone, and
there is no evidence for fast surface diffusion, which would be
expected to average the CSA. Second, we find that inorganic
phosphate is released when bisphosphonates bind to bone,
indicating a displacement mechanism in which the bisphospho-
nates are “chemically” bonded to the bone matrix. This would
be consistent with the observation that high levels of Pi are
needed for the chromatographic elution of bisphosphonates from
HAP. Third, we find that pamidronate binding can be well
described by a Langmuir adsorption isotherm, that the surface
area/pamidronate is ∼30-38 Å2 and that ∆G for binding is
∼-4.3 kcal mol-1 (at pH ) 7, 30 °C). Fourth, the results of
15N NMR measurements indicate that zoledronate binds in a
“partially” protonated form at pH 7, risedronate is non-
protonated, while pamidronate is essentially fully protonated,
consistent with known pKa values for imidazole (pKa ≈ 6.7),
pyridine (pKa ≈ 5.7) and primary amines (pKa ≈ 10.5). Fifth,
the results of TEDOR experiments show that pamidronate binds
in a close-to-gauche conformation, which suggests that the
terminal ammonium group might bind in such a way as to
maximize its electrostatic interactions with anionic groups in
the bone mineral. Sixth, the results of 2H NMR measurements
indicate that pamidronate undergoes fast conformational changes
(3 × 107 sec-1) on bone, interconverting most likely between
gauche+ and gauche- conformers. In contrast, alendronate
undergoes a series of coupled rotations, yielding lineshapes and
linewidths very similar to those seen previously with 2H-labeled
lipids.40 With the aromatic bisphosphonates, there are rapid (∼3
- 5 × 107 sec-1) ring “wobbles”, although we find no evidence
for the η ≈ 0.65 lineshapes seen with aromatic groups (i.e.,
phenylalanine) in proteins, in which 2-fold 180° ring-flips occur.
Seventh, the results of a double cross-polarization experiment
indicate that the pamidronate side chain Cγ carbon is close to
bone mineral PO43- groups, while CR and Câ carbons are more
distant. Overall, these results enable the construction of a
molecular model for pamidronate binding to bone in which the
hydroxyapatite (carbonatoapatite) triangle of PO4 groups,
enabling the gauche conformers (from TEDOR) to interconvert
(from 2H NMR), jumping from one PO43- site to another (from
DCP).
Given that bisphosphonates represent a large contribution to
the global pharmaceutical market and that there is interest in
their further development for use in oncology and as anti-
infectives, the results presented above are of general interest in
the context of tailoring new bisphosphonates to suit their
intended application. In particular, the ability to begin to probe
both the binding affinities (∆G, and in principle of course ∆H
and ∆S) as well as the static structures of bisphosphonates bound
to human bone should help with the development of novel
species that have particular attributes by using, for example,
the types of structure-based design techniques used to develop
enzyme inhibitors. For example, it has recently been shown that
bisphosphonate activity in bone resorption in ViVo can be well
described by using a simple statistical model involving enzyme
(FPPS) and HAP affinity data:43 an obvious extension of this
would be to use enzyme QSAR and structure-based bone
binding results. In addition, it should also be possible to further
develop the static (and dynamic) models described above, to
help validate computational modes for bone (HAP)-bisphos-
phonate interactions.10
Acknowledgment. This work was supported by the United
States Public Health Service (NIH Grant No. GM65307). S.M.
was supported by an American Heart Association, Midwest
Affiliate, Predoctoral Fellowship (Award Number 0615564Z).
Y.S. was supported by a Leukemia and Lymphoma Society
Special Fellowship (Award No. 3019-06). The authors would
like to thank Professor Robert Griffin and Dr. Patrick van der
Wel for providing the Turbopowder program, Professor Chad
Rienstra, for helpful advice, and Dushyant Mukkamala, for help
with the X-ray measurements.
JA0759949
(43) Ebetino, F. H.; B. L. B.; Evdokimov, A. G.; Pokross, M.; Kavanagh, K.;
Lundy, M.; McKenna, C. E.; Russell, R. G. G.; Dao, Z.; Dunford, J. Abstract
136, CERMACS ACS Regional Meeting, (Cincinatti, Ohio) 2007.
(44) Marion, D.; Wuthrich, K. Biochem. Biophys. Res. Commun. 1983, 113,
967-974.
3-
chemisorbed bisphosphonate displaces one bone PO4 in the
9
J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008 1273