Synthesis and characterization of 1,4,7-triazacyclononane derivatives with methylphosphinate and acetate side chains for monitoring free MgII by31P and1H NMR spectroscopy

Article abstract of DOI:10.1021/ja953771p

A series of 1,4,7-triazacyclononane-based ligands containing one, two, or three methylphosphinate (MP) side chains has been prepared. The macrocycle with three methylpbosphinates, NOTMP, appeared to be a very promising ligand for monitoring Mg^II in biological samples by ³¹P NMR spectroscopy. The ³¹P resonances of the free ligand and the Mg^II complex were well resolved, in slow exchange, and well downfield of typical tissue phosphate and phosphate ester metabolite resonances. The K_d value of the Mg(NOTMP) complex was 0.35 mM at physiological pH and temperature, a value which is optimal for accurate assessment of free Mg^II in cells and blood plasma. The latter was confirmed experimentally. Furthermore, the selectivity of NOTMP for Mg^II over Ca^II exceeded 2 orders of magnitude. The K_d values of the Mg^II complexes were sensitive to pH and temperature. The pH effects could be easily predicted from potentiometric pH titration data. Examination of the Mg^II ligand equilibria by ³¹P NMR over a wide temperature range provided complexation enthalpies and entropies. It was shown that the smaller K_d values at higher temperatures were largely due to a temperature effect on the highest ligand protonation constant. Only a small endothermic complexation enthalpy contributed to this phenomenon.

Full text of DOI:10.1021/ja953771p

4396
J. Am. Chem. Soc. 1996, 118, 4396-4404
Synthesis and Characterization of 1,4,7-Triazacyclononane
Derivatives with Methylphosphinate and Acetate Side Chains
31
1
for Monitoring Free MgII by P and H NMR Spectroscopy
Jurriaan Huskens† and A. Dean Sherry*,†,‡
Contribution from the Department of Chemistry, UniVersity of Texas at Dallas,
P.O. Box 830688, Richardson, Texas 75083-0688, and Department of Radiology,
The Rogers Magnetic Resonance Center, UniVersity of Texas Southwestern Medical Center,
5801 Forest Park Road, Dallas, Texas 75235-9085
ReceiVed NoVember 9, 1995X
Abstract: A series of 1,4,7-triazacyclononane-based ligands containing one, two, or three methylphosphinate (MP)
side chains has been prepared. The macrocycle with three methylphosphinates, NOTMP, appeared to be a very
promising ligand for monitoring MgII in biological samples by 31P NMR spectroscopy. The 31P resonances of the
free ligand and the MgII complex were well resolved, in slow exchange, and well downfield of typical tissue phosphate
and phosphate ester metabolite resonances. The Kd value of the Mg(NOTMP) complex was 0.35 mM at physiological
pH and temperature, a value which is optimal for accurate assessment of free MgII in cells and blood plasma. The
latter was confirmed experimentally. Furthermore, the selectivity of NOTMP for MgII over CaII exceeded 2 orders
of magnitude. The Kd values of the MgII complexes were sensitive to pH and temperature. The pH effects could
be easily predicted from potentiometric pH titration data. Examination of the MgII ligand equilibria by 31P NMR
over a wide temperature range provided complexation enthalpies and entropies. It was shown that the smaller Kd
values at higher temperatures were largely due to a temperature effect on the highest ligand protonation constant.
Only a small endothermic complexation enthalpy contributed to this phenomenon.
Introduction
intracellular conditions, thus making the determination of the
ratio between free and bound ATP extremely sensitive to experi-
mental error. More recently, chelators containing a fluorine
reporter functionality have been used to measure [Mg]f in cells
and perfused organs by 19F NMR.9,10 Due to line broadening
of the 19F resonances after loading the ligands into cells, the
observed signal-to-noise ratios were not as high as expected.
Since 31P NMR spectroscopy is quite valuable for monitoring
intracellular pH, phosphorylation potentials, [ATP]/[ADP] ratios,
and a wide variety of other energy status signals in cells,
perfused organs, and intact animals, our objective was to design
ligands containing one or, preferably, more NMR equivalent
31P nuclei whose chemical shifts do not overlap with those
typical phosphorus-containing metabolites, such as (poly)-
phosphates and phosphate esters, but are sensitive to complex-
ation with MgII. In previous studies,11-13 we demonstrated the
usefulness of the triazacyclononane ring, for both its synthetic
versatility and its selectivity for binding MgII over CaII. The
trisubstituted phosphonate monoester derivative NOTPME (see
Figure 1) demonstrated the potential of these ligands for
monitoring [Mg]f in isolated cells;11 the free ligand and the MgII
complex were in slow exchange, thus allowing the observation
of both resonances separately and providing the ratio of their
peak areas by signal integration. The conditional stability
Magnetic resonance spectroscopy (MRS) is an attractive tool
for monitoring physiological events in ViVo. The continuing
growth of MRS in clinical diagnosis raises interest in the
development of auxiliary ligands for determining levels of
biologically important ions in intact, functioning tissue. Among
the divalent alkaline earth ions, most interest has focused on
CaII, largely because of its well-established role as an intra-
cellular second messenger.1 Magnesium(II) has received less
attention in this regard because its concentration is typically
much higher in cells and, consequently, it is generally thought
of as a relatively nonspecific, hard acid cation that binds with
and helps stabilize phosphorus-containing metabolites like ATP.
However, there has been evidence mounting in recent years that
MgII may play a pivotal role in regulation of a much wider
variety of intracellular events.2
The most common method for monitoring levels of free MgII
([Mg]f) uses the chemical shift separation of the 31P R- and
γ-resonances of ATP.3 Accurate assessment of [Mg]f is com-
plicated by other factors contributing to the chemical shift of
the observed resonances, including pH and coordination of
monovalent cations.4,5 The dissociation constant of Mg(ATP)
(30-80 µM)6-8 also is not favorable for accurate assessment
of [Mg]f, since ATP is more than 90% bound under typical
* To whom correspondence should be addressed at either address.
Phone: (214) 883-2907 or (214) 648-5877. FAX: (214) 883-2925 or (214)
648-5881. E-mail: sherry@utdallas.edu.
(7) Williams, G. D.; Smith, M. B. Magn. Reson. Med. 1995, 33, 853-
857.
(8) Golding, E. M.; Golding, R. M. Magn. Reson. Med. 1995, 33, 467-
474.
† University of Texas at Dallas.
‡ University of Texas Southwestern Medical Center.
X Abstract published in AdVance ACS Abstracts, April 15, 1996.
(1) Rasmussen, H.; Barrett, P. Q. Physiol. ReV. 1984, 64, 938-984.
(2) Grubbs, R. D.; Maguire, M. E. Magnesium 1987, 6, 113-127.
(3) Gupta, R. K.; Moore, R. D. J. Biol. Chem. 1980, 255, 3987-3993.
(4) Halvorson, H. R.; Vande Linde, A. M. Q.; Helpern, J. A.; Welch, K.
M. A. NMR Biomed. 1992, 5, 53-58.
(5) Williams, G. D.; Mosher, T. J.; Smith, M. B. Anal. Biochem. 1993,
214, 458-467.
(6) Burt, C. T.; Cheng, H.-M.; Gabel, S.; London, R. E. J. Biochem.
1990, 108, 441-448.
(9) Levy, L. A.; Murphy, E.; Raju, B.; London, R. E. Biochemistry 1988,
27, 4041-4048.
(10) Murphy, E.; Steenbergen, C.; Levy, L. A.; Raju, B.; London, R. E.
J. Biol. Chem. 1989, 264, 5622-5627.
(11) Ramasamy, R.; Lazar, I.; Brucher, E.; Sherry, A. D.; Malloy, C. R.
FEBS Lett. 1991, 280, 121-124.
(12) Lazar, I.; Ramasamy, R.; Brucher, E.; Geraldes, C. F. G. C.; Sherry,
A. D. Inorg. Chim. Acta 1992, 195, 89-93.
(13) Van Haveren, J.; DeLeon, L.; Ramasamy, R.; Van Westrenen, J.;
Sherry, A. D. NMR Biomed. 1995, 8, 197-205.
S0002-7863(95)03771-1 CCC: $12.00 © 1996 American Chemical Society

Synthesis of 1,4,7-Triazacyclononane DeriVatiVes
J. Am. Chem. Soc., Vol. 118, No. 18, 1996 4397
4, 7, and 10) used for electrode calibration were obtained from Ricca.
All other chemicals were obtained from Aldrich. tert-Butyl alcohol
was dried over zeolite NaA. All other chemicals were used as received.
1,4,7-Triazacyclononane-N-(methylenemethylphosphinic acid)-
N′,N′′-bis(acetic acid) Hydrochloride (NO2A-MP). 1,4,7-Triazacy-
clononane (0.482 g, 3.74 mmol) was dissolved in 10 mL of tBuOH.
Trifluoromethanesulfonic acid (TfOH; 0.588 g, 3.92 mmol) was added,
and the solution was heated to 50 °C. Bromoacetic acid tert-butyl ester
(1.457 g, 7.47 mmol in 6 mL of tBuOH) and KOtBu (0.838 g, 7.47
mmol in 6 mL of tBuOH) were added simultaneously in 20 equal
aliquots over a period of 30 h. The reaction was stirred overnight at
50 °C. After cooling to room temperature, a 13C NMR spectrum
revealed that the mixture contained tBuNO1A, tBu2NO2A, and tBu3-
NOTA in a 1:5:1 ratio. Water was added, the pH was lowered to 5,
and the mixture was washed with diethyl ether. A 13C NMR spectrum
of the ether extract showed the presence of tBu3NOTA. The water
layer was adjusted to pH 8 and extracted twice with diethyl ether. After
drying over Na2SO4, evaporation of the ether gave tBu2NO2A. TfOH.
This was redissolved in water, and the pH was raised to 13 with aqueous
NaOH. Extraction with diethyl ether gave the free base, tBu2NO2A
(0.914 g, 68%), which was judged 95% pure by 13C NMR. Traces of
tBu3NOTA that were present were removed by repeating the extraction
procedure, finally yielding 0.662 g (1.85 mmol, 50%) of the white solid,
tBu2NO2A. (Extraction of the original water layer at pH 13 yielded
77 mg of tBuNO1A (0.32 mmol, 9%) which was used for the NMR
titration experiment; see below.) 13C NMR (CDCl3) (tBu2NO2A,
ppm): δ 171.4 (C5), 80.5 (C6), 57.8 (C4), 52.8 and 52.6 (C1 and C2),
46.7 (C3), 28.0 (C7) (numbering according to Scheme 1).
A mixture of tBu2NO2A (0.399 g, 1.12 mmol) and diethoxymeth-
ylphosphine, MeP(OEt)2 (0.263 g, 1.93 mmol), was prepared using the
phosphine as the solvent. Nitrogen was bubbled into the flask for 5
min, after which the mixture was heated to 80 °C. Paraformaldehyde,
(CH2O)n (0.060 g, 2.0 mmol), was added, and this dissolved within 1
min. The mixture was stirred for 1 h. After cooling to room
temperature, water was added and the pH lowered to 5. The solution
was washed twice with diethyl ether. The pH of the water layer was
raised to 13, and the product separated from the water layer. Diethyl
ether was added, and the water layer was extracted twice with diethyl
ether. The combined ether layers were dried over Na2SO4. Evaporation
gave tBu2NO2A-EtMP (0.450 g, 0.94 mmol, 84%) as a yellow oil. It
appeared pure by NMR. 1H NMR (CDCl3) (tBu2NO2A-EtMP, ppm):
δ 3.96 (2H, POCH2, “quintet”, 3JPH ) 3JHH ) 7 Hz), 3.15 (4H, H4, s),
2.70-2.80 (10H, H2, H3, and PCH2N, m), 2.70 (4H, H1, s), 1.40 (6H,
Figure 1. Ligands discussed in this study.
2
PCH3, d, JPH ) 12 Hz), 1.31 (18H, H7, s), 1.17 (3H, POCH2CH3, t,
constant, Kd ) [Mg]f[L]f/[MgL], was too high, however, for
accurate determination of the [MgL]/[L]f ratio. Ideally, this ratio
should be about 1, which means that, for typical intracellular
[Mg]f levels of about 0.5 mM, the Kd for the ideal ligand should
also be about 0.5 mM. In a more recent study,13 we demon-
strated that changing the ligating side chain groups attached to
the triazacyclononane ring allowed fine-tuning of the Kd. In
that study, we concluded that a 1,4,7-triazacyclononane deriva-
tive containing two acetate and one ethylphosphinate side chains,
NO2A-EP, was most promising for in ViVo application.
In the present study, we investigated the possibility of
changing the ethyl group of NO2A-EP to a methyl group to
provide an easily monitored 1H resonance as well. Surprisingly,
this relatively minor structural alteration yielded ligands with
significantly higher affinities for MgII. We report here the
synthesis and characterization of the complete methylphosphi-
nate family of ligands containing mixed acetate and meth-
ylphosphinate side chains. We show that the fully substituted
methylphosphinate ligand NOTMP has an appropriate Kd for
monitoring [Mg]f, and demonstrate its utility for measuring
[Mg]f in blood plasma.
3JHH ) 7 Hz). 13C NMR: δ 171.2 (C5), 80.5 (C6), 59.9 (POCH2, d,
2
2JPC ) 8 Hz), 59.6 (C4), 57.2 (C3, d, JPC ) 8 Hz), 57.1 (PCH2N, d,
1JPC ) 115 Hz), 55.5 (C1 and C2), 28.0 (C7), 16.4 (POCH2CH3, d,
3JPC ) 6 Hz), 12.9 (PCH3, d, 1JPC ) 90 Hz) (numbering (H1-H4, H7,
C1-C7) according to tBu2NO2A, Scheme 1).
The ester product tBu2NO2A-EtMP (0.406 g, 0.85 mmol) was
dissolved in 1 M aqueous HCl and heated at 80 °C for 15 h. This
removed the tBu groups completely, but about 50% of the Et groups
were still intact. Therefore, the pH was raised to 12, and the mixture
was heated at 50 °C for 15 h. A small fraction (10%) showed cleavage
of the methylphosphinate group to NO2A. The product was purified
on a cation exchange column (DOWEX-H+) eluted by a gradient of
0.1-2.0 M HCl. Evaporation gave NO2A-MP (0.264 g, 65%) as a
white solid, also containing HCl and H2O, which was judged pure by
1H and 31P NMR. Potentiometry gave the molecular weight (477) and
the approximate formula NO2A-MP‚2HCl‚3.6H2O. 1H NMR (D2O)
(NO2A-MP, ppm): δ 4.01 (4H, H4, s), 3.53 (4H, H1, s), 3.37-3.43
2
(8H, H2 and H3, br d), 3.33 (2H, H6, d, JPH ) 6 Hz), 1.44 (3H, H7,
2JPH ) 14 Hz). 13C NMR: δ 173.1 (C5), 59.6 (C4), 57.8 (C6, d, 1JPC
1
) 97 Hz), 54.1 and 53.6 (2:1) (C1, C2, and C3), 17.5 (C7, d, JPC
)
91 Hz) (numbering according to Scheme 1).
1,4,7-Triazacyclononane-N,N′-bis(methylenemethylphosphinic
acid)-N′′-(acetic acid) Hydrochloride (NO1A-2MP). tBuNO1A was
prepared in a similar procedure using 1,4,7-triazacyclononane (0.215
g, 1.67 mmol), 5 mL of tBuOH, TfOH (0.266 g, 1.77 mmol),
bromoacetic acid tert-butyl ester (0.492 g, 2.52 mmol in 3 mL of
tBuOH), and KOtBu (0.281 g, 2.50 mmol in 3 mL of tBuOH). The
reaction was stirred overnight at 50 °C. Subsequent extractions of the
Experimental Section
Synthesis. Chemicals. Diethoxymethylphosphine was obtained
from Alfa and tert-butyl alcohol from Baker. Standard solutions for
potentiometry (0.1000 M HCl and 1.000 M KOH) and the buffers (pH

4398 J. Am. Chem. Soc., Vol. 118, No. 18, 1996
Huskens and Sherry
Scheme 1
water layer at pH 7 and 13 with diethyl ether gave an oily mixture of
tBuNO1A and tBu2NO2A (7:3, 0.113 g, 0.29 mmol of tBuNO1A, 17%),
which was used without further purification. 13C NMR (CDCl3)
(tBuNO1A, ppm): δ 171.6 (C5), 81.0 (C6), 57.8 (C4), 52.8 (C1), 46.8
and 46.2 (C2 and C3), 28.2 (C7) (numbering according to Scheme 1).
Phosphorylation and subsequent hydrolysis were performed in a
procedure analogous to the one described for NO2A-MP, using the
mixture of tBuNO1A and tBu2NO2A (0.113 g), MeP(OEt)2 (0.082 g,
0.60 mmol), and (CH2O)n (0.018 g, 0.60 mmol). NO1A-2MP was
obtained (51 mg, 33%), also containing HCl and H2O, which was judged
All solutions were prepared from distilled, demineralized, and
degassed water, and were saturated with and stored under N2. Ligand
solutions were typically 2-3 mM, and excess KOH was added. To
determine the ligand protonation constants, these solutions were titrated
with standardized 0.1000 M HCl. For the metal ligand stability
determinations, a concentrated metal chloride solution (0.2 M) was
added to the cup solution to reach a metal ligand ratio of about 1, after
which the solution was also titrated with 0.1 M HCl. Exact ligand
concentrations (and molecular weights of the isolated salts) were
evaluated after multiple titrations of ligand and metal ligand solutions.
All titrations were performed at least twice.
1
pure by H and 31P NMR. Potentiometry gave the molecular weight
(535), and the approximate formula NO1A-2MP‚3HCl‚3H2O was
derived. 1H NMR (D2O) (NO1A-2MP, ppm): δ 3.99 (2H, H4, s), 3.39
(8H, br s) and 3.29 (8H, br s) (H1, H2, H3, and H6), 1.38 (6H, H7,
The calculation of the constants was performed using a general
spreadsheet program with an approach described earlier.14,15 Only the
three highest protonation constants could be evaluated in the pH range
of 2-12 for all ligands. Formation of polynuclear complexes MnL
was not observed in any system.
2JPH ) 15 Hz). 13C NMR: δ 172.2 (C5), 59.3 (C4), 57.4 (C6, d, 1JPC
1
) 96 Hz), 54.3, 53.9, and 53.5 (C1, C2, and C3), 17.6 (C7, d, JPC
)
92 Hz) (numbering according to Scheme 1).
NMR Spectroscopy. All analytical 1H (200 MHz) and 13C (50
MHz) NMR spectra, as well as the protonation titrations of tBuNO1A,
tBu2NO2A, and NO2A-MP, were recorded on a JEOL FX200 NMR
spectrometer at 295 K. The protonation titration of NO1A-2MP was
monitored by 1H NMR spectroscopy (500 MHz), using a 5 mm 13C/1H
probe. 1H NMR spectra of the hydrolyzed ligands were recorded in
D2O using HDO (4.80 ppm) as the internal standard, while their 13C
NMR spectra were recorded in 10% aqueous D2O with tBuOH (31.2
ppm) as the internal standard. More experimental details about the
protonation titrations are provided in the supporting information together
1,4,7-Triazacyclononane-N,N′,N"-tris(methylenemethylphosphin-
ic acid) Hydrochloride (NOTMP). The phosphorylation of 1,4,7-
triazacyclononane (0.634 g, 4.91 mmol) was performed in a proce-
dure analogous to the one for NO2A-MP, using MeP(OEt)2 (2.429 g,
17.86 mmol) and (CH2O)n (0.536 g, 17.87 mmol). The ester product
Et3NOTMP was purified on a silica column eluted with MeOH/CH2-
Cl2 (1:1 v/v). The pure ester was obtained as a yellow oil (1.705 g,
3.49 mmol, 71%). 1H NMR (CDCl3) (Et3NOTMP, ppm): δ 4.00 (6H,
3
3
POCH2, “quintet”, JPH ) JHH ) 7 Hz), 2.81-2.92 (18H, ring CH2
2
and PCH2N, m), 1.46 (9H, PCH3, d, JPH ) 14 Hz), 1.24 (9H,
1
with all H and 13C resonance data and assignments.
POCH2CH3, t, 3JHH ) 7 Hz). 13C NMR: δ 60.2 (POCH2, d, 2JPC ) 8
Hz), 57.9 (PCH2N, d, 1JPC ) 113 Hz), 57.6 (ring CH2, d, 2JPC ) 7 Hz),
All 31P NMR spectra (202 MHz), as well as the 1H (500 MHz)
monitored Kd determinations of NO2A-MP and NOTMP, were recorded
on a General Electric GN500 NMR spectrometer using a 10 mm broad-
band probe. Spectra were recorded at room temperature (295 K) or at
another temperature using the standard GN variable temperature control
unit. For the 31P NMR spectra, aqueous 85% phosphoric acid (0 ppm)
was used as the external standard, and 1H decoupling was applied unless
stated otherwise. For the Kd determinations, performed at 295 and 310
K, 3 mL samples were prepared of 2-3 mM ligand in an aqueous
solution containing 10% D2O, 115 mM of KCl, 20 mM of NaCl, 10
mM of glucose, and 10 mM of HEPES. The pH was adjusted to 7.40
(at room temperature) using aqueous NaOH. Aliquots of 0.2 M MgCl2
were added, and the 31P spectrum was recorded after each addition.
The pH was checked and adjusted if necessary. Resonance areas were
determined by peak integration using standard GE software. Evaluation
of the Kd values was performed by treating the ligand concentration as
a variable, as described in the text. Calculated ligand concentrations
always appeared to match the molecular weights obtained by potenti-
3
1
16.6 (POCH2CH3, d, JPC ) 6 Hz), 13.3 (PCH3, d, JPC ) 91 Hz).
The hydrolysis to and purification of NOTMP was performed
similarly as described above to yield NOTMP (1.58 g, 87%), also
containing HCl and H2O, which was judged pure by 1H and 31P NMR.
Potentiometry gave the molecular weight (557), and the approximate
formula NOTMP‚3HCl‚2H2O was derived. 1H NMR (D2O) (NO1A-
2MP, ppm): δ 3.37 (6H, PCH2N, d, 2JPH ) 8 Hz) and 3.34 (12H, ring
2
CH2, s), 1.38 (9H, PCH3, JPH ) 14 Hz). 13C NMR: δ 56.1 (PCH2N,
1
1
d, JPC ) 94 Hz), 52.8 (ring CH2), 16.5 (PCH3, d, JPC ) 92 Hz).
Potentiometry. Potentiometric titrations were conducted at 298 and
310 K using an Accumet 925 pH meter (Fisher), an Orion 8103 Ross
combination electrode, and a Metrohm 665 Dosimat automatic buret
(Brinkman Instruments). The ionic strength was adjusted to 0.1 M
using KCl in all titrations, and all measurements were performed under
a N2 atmosphere. Hydrogen ion activities were obtained from millivolt
readings, and the electrode was calibrated using Ricca high-precision
buffers, which are valid from 278 to 318 K. Hydrogen ion activities
were translated into concentrations using pKw ) 13.79 (298 K) and
13.40 (310 K) and an activity coefficient of 0.82, which were
determined from separate titrations of 0.001 and 0.01 M KOH with
0.1 M HCl.
(14) Van Westrenen, J.; Khizhnyak, P. L.; Choppin, G. R. Comput. Chem.
1991, 15, 121-129.
(15) Huskens, J.; Van Bekkum, H.; Peters, J. A. Comput. Chem. 1995,
19, 409-416.

Synthesis of 1,4,7-Triazacyclononane DeriVatiVes
J. Am. Chem. Soc., Vol. 118, No. 18, 1996 4399
ometry. The 31P NMR shifts of the ligands were measured as a function
of pH, using 3-5 mM samples.
sitated a careful tuning of the [H+]/[9N3] ratio in the synthesis
of tBu2NO2A. To achieve the maximal selectivity, this ratio
has to be 1 during the course of the reaction. A larger ratio
will lead to protection of tBuNO1A, and therefore a competitive
reaction rate between tBuNO1A and tBu2NO2A, while a smaller
ratio fails to protect tBu2NO2A from the consecutive reaction
toward tBu3NOTA. This selectivity was reached by using 1
equiv of TfOH, and by adding tert-butyl bromoacetate and
KOtBu simultaneously in small equal quantities.
The usual hydrolysis procedure of refluxing in 6 M HCl
appeared unsuitable for this ligand, since more than 20% of
the methylphosphinate side chain was cleaved off after 15 h at
120 °C. Even lower acid concentrations and lower temperatures
yielded small amounts of NO2A. The product NO2A-MP
was further purified on a cation exchange column eluted
with aqueous HCl, finally yielding the free acid. The ligand
as well as its MgII complex appeared to be quite stable in
neutral aqueous solution. No degradation was observed for
several weeks; the same was true for the other ligands discussed
here.
Two other compounds, NO1A-2MP and NOTMP, were
synthesized using similar methods. NO1A-2MP was prepared
from tBuNO1A isolated from a reaction of 9N3 with 1.5 equiv
of tert-butyl bromoacetate, applying 1 equiv of TfOH. We also
attempted to use the protective strategy described above for tBu2-
NO2A in the selective synthesis of tBuNO1A, by using 2 equiv
of TfOH. However, the reactivity of the diprotonated 9N3
appeared to be too low to protect the diprotonated tBuNO1A
from consecutive reaction toward tBu2NO2A.
The intermediate tBuNO1A was reacted with excess MeP-
(OEt)2 and (CH2O)n, followed by hydrolysis and purification
with cation exchange column chromatography. NOTMP was
prepared in one step as the triethyl ester, starting from 9N3 and
excess MeP(OEt)2 and (CH2O)n, similar to a known procedure.17
Hydrolysis in aqueous NaOH, followed by purification on a
cation exchange column, gave the free acid. These latter two
ligands were prepared to measure the Kd of their MgII
complexes, because the Kd of Mg(NO2A-MP) was, surprisingly,
too low (see below) for our intended application.
For the studies of the behavior of Keq (for the equilibrium MgL +
H+ h MgII + HL, see below) Versus temperature, similar samples
were used containing 2.00 mM ligand, and 2.00 mM (for NOTMP) or
1.62 mM (for NO1A-2MP and NO2A-MP) MgCl2. Spectra were
measured from 283 to 353 K in steps of 10 K. For NOTMP, Keq values
were calculated from the peak areas, and the enthalpy, ∆H, and entropy,
∆S, were determined from a plot of -R ln Keq Versus 1/T. For NO1A-
2MP and NO2A-MP, the [Mg]tot/[L]tot ratio was treated as a variable
in the calculation of ∆H, while ∆S was fixed at -132 J/(mol K) (see
text). In both cases, the calculated [Mg]tot/[L]tot ratio appeared to be
equal to the experimental one (0.81). No corrections for a possible
change of pH Versus temperature were made.
The spin-lattice relaxation times (T1) of NOTMP and its MgII
complex were determined using the inversion-recovery pulse sequence
with a three-parameter curve-fitting procedure.16 The T1 of the free
ligand was also determined in a sample without MgCl2. Selective
saturation of either the free ligand or the MgII complex resonance, using
the former sample, was performed by applying a low-power long pulse
(1 s). In both cases, no effect on the peak area of the nonirradiated
resonance was observed, thus excluding spin transfer between the
resonances.
Plasma from one of the authors (A.D.S.) was used for a determination
of the Kd of Mg(NOTMP) monitored by 31P NMR, as described above.
An accurately known amount of ligand (3 mM) was used, and 10%
D2O was added for locking. The pH (7.80) was measured and appeared
to be stable during the titration.
Results and Discussion
Synthesis. Previous work13 has shown that the ethylphos-
phinate derivative NO2A-EP (Figure 1) had an appropriate Kd
(0.43 mM) for in ViVo purposes, and the 31P NMR spectrum
showed two well-separated (∆δ ) 2.4 ppm) signals for the free
ligand and the MgII complex. Since our goal was to design a
1
ligand that might have a H resonance also sensitive to MgII
binding, our first target compound was NO2A-MP. This ligand
was synthesized via a new strategy, in which no protecting
groups are required. Starting from 1,4,7-triazacyclononane
(9N3), two acetate groups (tBu esters) were attached by reaction
with 2 equiv of tert-butyl bromoacetate in tBuOH, giving
tBuNO1A and tBu2NO2A (Scheme 1). The HBr produced in
the reaction was neutralized by the simultaneous addition of
KOtBu. The formation of the trisubstituted product tBu3NOTA
was inhibited by protonating a single macrocyclic nitrogen using
1 equiv of trifluoromethanesulfonic acid (TfOH). This method
using H+ as the “protecting group” relies on the fact that there
is a large gap between the stability constants (in water) of the
first and the second protonation steps of 9N3 and its derivatives.
In this way, a selectivity of 70% toward tBu2NO2A was reached.
Earlier attempts without TfOH showed a selectivity of only 30%.
The use of TfOH as a protection agent in the synthesis of
tBu2NO2A was further investigated by titrating tBuNO1A and
tBu2NO2A with TfOH, while monitoring their 1H and 13C NMR
spectra. In general, protonation of an amine induces a large
downfield shift on R-hydrogen nuclei and a large upfield one
on â-carbons. Therefore, H4 and C5 were expected to be good
indicators for the protonation sites. Both nuclei exhibited
marginal shifts for both protonations of tBuNO1A and for the
first protonation step of tBu2NO2A, while large effects were
observed for the second protonation step of the latter compound.
From these results it was concluded that protonation takes place
preferentially at a secondary nitrogen atom. A more elaborate
NMR Characterization of the Ligands and their MgII
Complexes. All three ligands NOTMP, NO1A-2MP, and
NO2A-MP displayed a single 31P NMR resonance in the 35-
45 ppm range, well downfield of resonances from biologically
important phosphates. For all ligands, the resonances of the
free ligand and the MgII complex were well resolved and in
slow exchange at 202 MHz. There was no line broadening of
either resonance up to 80 °C. We also used selective saturation
of either the free ligand or the MgII complex resonance of
NOTMP to investigate the time scale of the interchange between
them. In both experiments, saturation of either resonance had
no effect on the peak area of the nonsaturated resonance. Thus,
exchange between ligand and complex was slow relative to their
spin-lattice relaxation rates (1/T1 ) 0.63 s-1 for these ligands).
Accurate assessment of [ligand]/[complex] integral ratios is
vital for measuring free MgII levels using the strategy employed
1
here. The H coupled and decoupled 31P line widths, nuclear
Overhauser effects (NOE), and longitudinal relaxation times (T1)
were measured for NOTMP and its MgII complex. Since the
T1 values were essentially equal for the free ligand (1.6 ( 0.1
s) and the MgII complex (1.7 ( 0.1 s), short delays without
complete relaxation may be applied without loss of integral
accuracy. Similarly, equal values of the NOEs (1.3 and 1.2,
1
description, together with all assignments of the H and 13C
resonances, is given in the supporting information. The strong
preference for protonation at secondary nitrogen atoms neces-
(17) Broan, C. J.; Cole, E.; Jankowski, K. J.; Parker, D.; Pulukkody, K.;
Boyce, B. A.; Beeley, N. R. A.; Millar, K.; Millican, A. T. Synthesis 1992,
63-68.
(16) Canet, D.; Levy, G. C.; Peat, I. R. J. Magn. Reson. 1975, 18, 199.

4400 J. Am. Chem. Soc., Vol. 118, No. 18, 1996
Huskens and Sherry
1
Table 1. 31P (202 MHz) and H (500 MHz) Chemical Shiftsa of
This microscopic protonation sequence was confirmed by 1H
NMR shifts (see the supporting information). The first two
protonation steps occurred at ring nitrogens, and were selective
for protonation at a nitrogen atom near an acetate. This
selectivity can probably be ascribed to a stronger hydrogen bond
formed between a protonated nitrogen and an adjacent acetate
Versus a methylphosphinate group. The third protonation
produces very little change in chemical shift of the ring protons.
This indicates that this protonation does not take place at a
nitrogen, but probably at the carboxylate group of an acetate
side chain. This may lead to a rearrangement of a proton from
the N atom adjacent to the protonated acetate to the N atom
adjacent to the still deprotonated methylphosphinate group.
Also, the large increase of the 31P NMR chemical shift of
all three ligands during the third protonation (Figure 2) sug-
gests that a conformational change occurs, which may be
ascribed to a rotation of the acetate group. The protonation
sequence for NO2A-MP is schematically illustrated by Scheme
2. The microscopic protonation sequence, observed here for
the methylphosphinate family, also explains the 31P NMR
chemical shift differences observed for the ethylphosphinate and
the phosphonate ethyl ester derivatives (see Table 1).
NOTMP, NO1A-2MP, and NO2A-MP and Their MgII Complexes at
pH 7.4 and 295 K
31P NMR
1H NMR
L
δL
δMgL
∆δ
δL
δMgL
∆δ
NOTMP
36.8
39.6
40.1
40.5
42.9
42.7
17.5
21.3
42.3
42.1
41.8
45.5
45.3
45.1
23.5
23.4
5.5
2.5
1.7
5.0
2.4
2.4
6.0
2.1
1.33
1.30
1.30
1.40
1.41
1.41
0.07
0.11
0.11
NO1A-2MP
NO2A-MP
NOTEPb
NO1A-2EPb
NO2A-EPb
NOTPMEc
NO2A-PMEb
a In parts per million; measured with 85% phosphoric acid (0 ppm)
as the external and HDO (4.80 ppm) as the internal standard for the
31P and 1H NMR spectra, respectively. b Reference 13. c Reference 11.
The methyl proton resonances of NOTMP, NO1A-2MP, and
NO2A-MP appear as a sharp doublet (2JPH ) 13 Hz) while their
MgII complexes show slightly broadened resonances. However,
∆δ ()δMgL - δL) is only 0.07 (NOTMP) or 0.11 ppm (NO1A-
2MP and NO2A-MP; see Table 1). Although these resonances
are well resolved at high field, this small shift difference may
be too small to be useful in ViVo. The observation that the shift
difference is smaller for NOTMP compared to NO1A-2MP and
NO2A-MP can be explained by the fact that protonation of a
nitrogen atom near a methylphosphinate side chain causes a
(small) downfield shift of the δL of the methyl protons, in
contrast to the upfield effect on the δL of the 31P resonance. If
the free and the MgII-bound ligand resonances could be resolved
in ViVo, then spin-echo heteronuclear editing or multiple
quantum NMR techniques should allow monitoring of the
Figure 2. 31P chemical shift of the free ligand resonances, δL, of
NOTMP, NO1A-2MP, and NO2A-MP as a function of pH, as
determined at 202 MHz at 295 K.
respectively) allow proton decoupling without loss of integral
accuracy. The line width of the free NOTMP resonance
decreased from 30 Hz (multiplet) to 4 Hz (singlet) upon proton
decoupling, while the line width of the Mg(NOTMP) resonance
decreased from 37 to 12 Hz. So, in both cases the resonances
of the complex were somewhat broader than for the free ligand,
probably due to the formation of diastereomeric complexes due
to the methylphosphinate groups becoming chiral upon coor-
dination to MgII. Assuming that exchange between the dia-
stereomeric MgL species is slow or intermediate on the NMR
time scale, the chemical shift difference between such species
must be small.
1
methyl H resonance of the methylphosphinates without inter-
1
ference from other H signals.
Thermodynamics. Potentiometry was used to measure the
protonation constants of the ligands NOTMP, NO1A-2MP, and
NO2A-MP and their metal ligand stability constants with MgII
and CaII. All constants were measured in 0.1 M KCl to mimic
intracellular conditions, and all were determined at 25 and 37
°C. Typical titration curves for NOTMP (at 25 °C) are
presented in Figure 3. All results are summarized in Table 2,
together with previous results for NOTA and the ethylphosphi-
nate derivatives NOTEP, NO1A-2EP, and NO2A-EP.
The NOTMP protonation constants and stability constant with
MgII differed considerably from values reported previously by
others,18 especially the second protonation constant which they
report to be near pH 7. Figure 3, however, clearly shows the
absence of a pKa in this pH region, suggesting that the previous
literature values are in error.
The first protonation constants of the ligands described herein
are all somewhat higher than the values for the corresponding
ethylphosphinate derivatives.13 Also, the complexes with MgII
and CaII are more stable for the methylphosphinates than for
the corresponding ethylphosphinates. The selectivity for MgII
over CaII (log KMgL - log KCaL) decreased with an increase of
the number of acetate side chains, as reported previously.13 At
37 °C, the first protonation constants decreased as expected (see
below), while the MgII-ligand stability constants were some-
The 31P NMR chemical shifts of NOTMP, NO1A-2MP, and
NO2A-MP and of their MgII complexes are listed in Table 1,
together with some values of previously described ligands.11,13
The most striking phenomenon is the large shift difference, ∆δ
()δMgL - δL), for NOTMP, NOTEP,13 and NOTPME.11 Since
the chemical shifts of the MgII complexes, δMgL, were nearly
equal for all members of the same family, the large ∆δ values
for the triphosphinate and tris(phosphonate ester) ligands must
be due to differences in the values of δL. Therefore, we
investigated the behavior of δL Versus pH for NOTMP, NO1A-
2MP, and NO2A-MP (Figure 2). It appeared that the values
of δL were about the same for all three ligands at high pH,
where the ligands are nonprotonated. Upon addition of 1 equiv
of acid, the NOTMP resonance shifted to lower frequency (4.1
ppm), while the effects on the shift of NO1A-2MP (1.2 ppm)
and NO2A-MP (0 ppm) were much smaller. This indicates that
the first protonation of NO1A-2MP and NO2A-MP occurred
largely at a nitrogen atom with an acetate side chain, while in
NOTMP protonation must occur at a site near a phosphinate
side chain. This results in shielding of the 31P nuclei that are
in a â-position relative to the protonated N atom.
(18) Broan, C. J.; Jankowski, K. J.; Kataky, R.; Parker, D. J. Chem.
Soc., Chem. Commun. 1990, 1738-1739.

Synthesis of 1,4,7-Triazacyclononane DeriVatiVes
J. Am. Chem. Soc., Vol. 118, No. 18, 1996 4401
Scheme 2
) [Mg]f + [MgL]). Plots of [Mg]tot/y Versus 1/(1 + y) gave
straight lines with [L]tot as the slope and Kd as the intercept.
The results of the 31P NMR titrations at 25 °C for NOTMP,
NO1A-2MP, and NO2A-MP are shown in Figure 4. The
calculated [L]tot values obtained from these plots agreed well
with those calculated using the molecular weights of the ligands,
as determined later by potentiometry. The intercepts of Figure
4 indicate that the Kd values for the three ligands are in the
order NO2A-MP < NO1A-2MP < NOTMP. The Kd of Mg-
(NO2A-MP) estimated from this curve was 0.07 mM, a value
that is much smaller than the actual [L]tot (3 mM) used in the
experiment. These experimental conditions gave rise to a large
relative error in the Kd for this complex. However, the
concentration of NO2A-MP could not be lowered in this
experiment because this ligand has only one 31P nucleus per
molecule and, consequently, the signal-to-noise ratio became
limiting. The ligand concentrations of NO1A-2MP and NOT-
MP (2 mM) were of the same order of magnitude as the Kd
values of their MgII complexes estimated from these plots.
Figure 3. Potentiometric titration curves for NOTMP (2 mM) in the
absence and presence of 1 equiv of MgII or CaII, as determined at
298 K.
what higher. The species MgHL was observed in most of the
experiments, but was a minor species throughout the titrations,
while CaHL was not observed in any of the titrations.
For biological applications, the more critical binding constant
is that given by the conditional dissociation constant, Kd )
[Mg]f[L]f/[MgL], near physiological pH values. Here, [Mg]f
is the free MgII concentration, and [L]f is the total concentration
of uncomplexed (free) ligand, i.e., the sum of the concentrations
of (nonprotonated) L and its protonated forms. Since the first
protonation constants for all these ligands are much higher than
physiological pH values while the second protonation constants
are much lower, [L]f can conveniently be written as [HL] at
this pH.
1
Similar experiments were performed using H NMR spec-
troscopy. Table 2 lists the Kd values of the methylphosphinate
1
derivatives measured experimentally by 31P and H NMR and
the corresponding values estimated from the potentiometric data
(log Kd ) log KHL - pH - log KMgL). All Kd values appeared
to be equal to their potentiometric estimates within experimental
error. Somewhat to our surprise, the Kd value for Mg(NO2A-
MP) (0.07 mM) was about 6-fold lower than that for Mg(NO2A-
EP) (0.43 mM), too low for measurement of typical intracellular
free MgII levels. However, NO1A-2MP and even NOTMP
proved to have Kd values in a useful range for monitoring [Mg]f.
An additional advantage of these ligands is their higher 31P NMR
sensitivity, because they have more equivalent 31P nuclei per
molecule. NOTMP is especially suited, since it combines the
Kd values can be assessed at physiological pH from 31P NMR
titrations performed under buffered conditions. The experiments
described below were performed in HEPES-buffered solutions
(pH 7.40), containing KCl, NaCl, and glucose to mimic
intracellular conditions. In general, 2-3 mM ligand solutions
were titrated with small aliquots of a concentrated MgCl2
solution. The 31P NMR spectrum was recorded after each
addition, and the peak areas corresponding to L and MgL were
measured to give the integral ratio y ) [MgL]/[L]f. Since the
exact ionic form and, hence, molecular weight of each ligand
were unknown when these titrations were begun (the potentio-
metric titrations were performed later), the total ligand concen-
tration, [L]tot, was treated as a variable in each NMR titration.
In order to present the results graphically, the definition of Kd
1
highest 31P (and H) NMR sensitivity with a relative ease of
preparation and the largest 31P NMR shift difference between
L and MgL (see above).
A remarkable effect was the temperature dependence of Kd.
Both NMR and potentiometry indicated that the Kd of Mg-
(NOTMP) decreased by about a factor of 3 between 25 and 37
°C (see Table 2). In order to investigate this effect further, 31
P
NMR spectra of samples containing mixtures of L and MgL
were measured as a function of temperature. For a sample
containing 2 mM NOTMP and 2 mM MgCl2, the temperature
effect on the integral ratio was dramatic: at 10 °C there was
more free ligand than MgII complex (y ) 0.8), while at 80 °C
the free ligand peak was barely visible (y > 20). From
measurements over this temperature range, it became possible
to calculate thermodynamic parameters (∆H and ∆S) for the
was rewritten as eq 1, using the mass balances for L ([L]tot
[Mg]tot/y ) [L]tot/(1 + y) + Kd (1)
[L]f + [MgL]) and MgII ([Mg]tot (the total MgII concentration)
)

4402 J. Am. Chem. Soc., Vol. 118, No. 18, 1996
Huskens and Sherry
Table 2. Stepwise Protonation and Metal-Ligand Stability Constants of NOTMP, NO1A-2MP, and NO2A-MP with MgII and CaII As
Determined by Potentiometry at 298 and 310 K (I ) 0.1 M KCl), and Conditional MgII-Ligand Dissociation Constants, Kd, Measured from 31P
1
(202 MHz) or H (500 MHz) NMR Spectra at pH 7.40
a
log Kia
log KML
MgL
Kd of MgL (mM)
log âMHLb
MgHL
L
T (K)
HL
H2L
H3L
CaL
pH-metryc
31P NMR
1H NMR
1.2(1)d
NOTMP
298
310
298
310
298
310
298
298
298
298
10.92(2)
10.68(4)
11.7(1)
11.4(2)
12.0(1)
11.7(1)
10.27(5)
10.45(5)
11.15(4)
11.41
3.97(3)
3.98(2)
4.24(5)
4.1(1)
4.99(3)
5.04(3)
4.72(3)
4.42(1)
4.89(4)
5.6
2.09(2)
2.05(3)
2.10(2)
2.0(2)
2.73(3)
2.90(3)
3.26(3)
2.76(2)
2.82(4)
2.9
6.66(5)
6.67(5)
8.0(1)
8.1(1)
8.9(2)
9.0(2)
5.54(2)
5.57(3)
7.26(3)
9.69
4.45(4)
4.40(4)
6.3(1)
6.6(1)
7.5(1)
7.7(1)
3.79(2)
3.56(4)
6.43(1)
8.92
12.76(15)
e
14.5(3)
14.0(2)
14.8(2)
14.4(2)
0.8
0.4
0.2
0.08
0.05
0.02
1.1(1)d
0.35(3)
0.40(3)d
NO1A-2MP
NO2A-MP
0.07(3)d
0.06(3)d
NOTEPf
NO1A-2EPf
NO2A-EPf
NOTAg
b
a Standard deviations derived from two or three separate titrations: Ki ) [HiL]/[Hi-1L][H], KML ) [ML]/[M][L]. âMHL ) [MHL]/[M][H][L].
c Estimated values from potentiometry: log Kd ) log KHL - pH - log KML at pH 7.40. d T ) 295 K. e Not observed. f Reference 13. g Reference
19.
Figure 5. -R ln Keq (Keq is defined for the equilibrium MgL + H+
h
Figure 4. [Mg]tot/y (y ) [MgL]/[L]f) Versus the free ligand fraction,
[L]f/[L]tot ()1/(1 + y)), for NOTMP, NO1A-2MP, and NO2A-MP, as
determined by 31P NMR (202 MHz) at 295 K and pH 7.4. The slopes
represent [L]tot values, whereas the intercepts give Kd values.
MgII + HL) Versus 1/T for NOTMP ([L]tot ) 2.00 mM, [Mg]tot ) 2.00
mM), and NO1A-2MP and NO2A-MP ([L]tot ) 2.00 mM, [Mg]tot
)
1.62 mM), as determined by variable temperature 31P NMR (202 MHz)
at pH 7.4. The slopes represent ∆H values for this equilibrium.
Table 3. Enthalpies, ∆H (in kJ/mol), and Entropies, ∆S (in J/(mol
K)), of the Equilibrium MgL + H+ h MgII + HL for NOTMP,
NO1A-2MP, and NO2A-MP As Measured by Temperature
Dependent 31P NMR, and Resulting Values for the Protonation and
MgII Complexation As Derived in Combination with Potentiometric
Data
equilibrium MgL + H+ a HL + MgII. The equilibrium
constant, Keq ()[HL][Mg]f/[MgL][H]), could be obtained from
the NMR data once the ligand and MgII concentrations were
accurately known. At physiological pH, [L]f ) [HL] as
described above. Therefore, the relationship between Keq and
Kd is straightforward (Keq ) Kd/[H]). Since the equilibrium is
the decomplexation of MgL to MgII and L followed by the
protonation of L to HL, the free energy of the equilibrium, ∆G,
can be written as the difference in free energy of the protonation,
∆GHL, and the complexation, ∆GML, as given in eq 2. Plots of
L
∆H
∆S
∆HHL ∆SHL ∆HMgL ∆SMgL
NOTMP
-62(3) -126(8) -47 (52)a
15
11
4
178(8)
(184)a
(184)a
184
NO1A-2MP -62(4) (-132)a
-51 (52)a
NO2A-MP
NOTAb
9N3b
-57(3) (-132)a
-53 (52)a
-46
-44
50
54
0
a Value estimated from literature values for 9N3 and/or NOTA.
b Reference 19.
∆G ) -RT ln Keq ) ∆H - T∆S ) ∆GHL - ∆GML (2)
-R ln Keq Versus 1/T gave straight lines with ∆H as the slope
and -∆S as the intercept, as shown in Figure 5. The enthalpy
T∆SHL term contributes only 25% to the ∆GHL term at 298 K.
∆HHL values for the methylphosphinate ligands were calculated
from the protonation constants (at 298 K) as given in Table 2.
(∆H ) ∆HHL - ∆HML) and entropy (∆S ) ∆SHL - ∆SML
)
changes determined from such plots are obviously related to
the protonation and complexation values.
For the NOTMP data shown in Figure 5, no further assump-
tions were necessary: ∆HML and ∆SML were calculated from
∆H and ∆S. The value of ∆SML of NOTMP (178 ( 8 J/(mol
K)) was equal to the literature value for NOTA (184 J/(mol
K)) within experimental error. This suggests that the MgII ion
in both complexes is coordinated in a similar manner, very likely
with three nitrogens and three side chains. The endothermic
enthalpy of Mg(NOTMP) (∆HML ) 15 kJ/mol) reflects the
difference in coordinating strength of the phosphinate group
compared to the acetate group in NOTA (∆HML ) 0 kJ/mol).
This contributes to lower MgII-ligand stability constants for
the methylphosphinate family compared to NOTA (see Table
2) and to a (slightly) higher stability constant of Mg(NOTMP)
In order to obtain the complexation enthalpy, ∆HML, and
entropy, ∆SML, from the data shown in Figure 5, we fixed the
protonation entropy, ∆SHL, at a value of 52 J/(mol K) for all
three ligands NOTMP, NO1A-2MP, and NO2A-MP. This
approximation seems valid since the literature values19 for 9N3
(54 J/(mol K), Table 3) and NOTA (50 J/(mol K)) are very
close to each other, even though protonation in NOTA is
probably assisted by hydrogen bonding. Furthermore, the
(19) Martell, A. E.; Smith, R. M.; Motekaitis, R. J. NIST Critical Stability
Constants of Metal Complexes Database, NIST Standard Reference
Database 46; NIST Standard Reference Data, Gaithersburg, MD 20899,
1993.

Synthesis of 1,4,7-Triazacyclononane DeriVatiVes
J. Am. Chem. Soc., Vol. 118, No. 18, 1996 4403
with an increase in temperature from 25 to 37 °C. However,
the largest contributing factor to the decrease in Keq (and Kd)
with T comes from the exothermic protonation enthalpy (-47
kJ/mol). Thus, the greater tendency of HL to dissociate at higher
temperatures contributes most to the greater overall stability of
Mg(NOTMP) at physiological temperature than at room tem-
perature.
For NO1A-2MP and NO2A-MP, the measurements of Keq
Versus temperature were subject to the same problem as in the
experiment in which the Kd of Mg(NO2A-MP) at 25 °C was
determined (see above): the concentration of the ligand had to
be (much) higher than the Kd itself, providing relatively large
uncertainties in the [Mg]f values. This problem was overcome
by using another approximation: since the ∆SML values were
the same for Mg(NOTMP) and Mg(NOTA), we fixed the values
for Mg(NO1A-2MP) and Mg(NO2A-MP) at 184 J/(mol K). This
set the intercept (-∆S) in Figure 5 for these two ligands at 132
J/(mol K). The values of ∆H were then obtained from a fit of
the data presented in Figure 5, while the ∆HHL values were
calculated from the protonation constants at 298 K and ∆SHL
(see above). Thus, endothermic ∆HML values of 11 and 4 kJ/
mol for NO1A-2MP and NO2A-MP, respectively, were ob-
tained. An overview of the ∆HML data (Table 3) for the
methylphosphinate derivatives and NOTA shows that substitut-
ing a methylphosphinate side chain for an acetate group (from
NOTA to NOTMP) leads to an endothermic contribution to
∆HML of about 5 kJ/mol per group.
Figure 6. [Mg]add - [MgL] Versus y ()[MgL]/[L]f) for NOTMP (3
mM) as determined in human blood plasma (pH 7.8) by 31P NMR (202
MHz) at 310 K. The slope represents the Kd value, whereas the intercept
gives -[Mg]ini. Spectrum: 31P NMR spectrum of the plasma sample
before addition of MgII.
remaining 24% (0.26 mM) must be bound to the proteins present
in the plasma with a much higher affinity than that of NOTMP.
Simulation of a 1:1 binding model for MgII plus protein showed
that the Kd values of the protein binding sites must be less than
0.1 mM. The NMR titration results presented in Figure 6 are
inconsistent with a large number of weaker MgII-protein sites.
Performance of NOTMP in Blood Plasma. To investigate
the behavior of NOTMP in blood plasma, we performed an in
Vitro titration of (human) blood plasma with MgCl2 at 37 °C
while monitoring the sample by 31P NMR. For locking, a small
amount of D2O was added, yielding a sample containing 81%
plasma (pH 7.8). At any point of the titration, the MgII mass
balance was described by eq 3. [Mg]ini represents the free MgII
Conclusions
[Mg]ini + [Mg]add ) [Mg]f + [MgL]
(3)
The main goal of this research was to prepare a MgII chelator
which contained easily monitored 31P and 1H resonances.
Switching from ethyl- to methylphosphinate side chains proved
to be a significant advantage in that the number of phosphorus-
containing side chains could be increased without sacrificing
the targeted Kd values of the ligands with MgII. In fact, the
tris(methylphosphinate) derivative NOTMP appeared to have
exactly the desired Kd (0.35 mM) at 37 °C and pH 7.4. Not
only does this derivative have a 3-fold greater NMR sensitivity
compared to NO2A-EP, but its synthesis is much easier and it
is more selective for MgII over CaII. The latter appears to be a
general phenomenon for triazacyclononane derivatives contain-
ing fewer acetate side chains. Furthermore, NOTMP shows a
much larger chemical shift difference between the 31P resonances
of the free ligand and those of the MgII complex, owing to
shielding of the 31P nuclei that takes place upon protonation of
a single amine. This is in contrast to the mixed side chain
derivatives, in which protonation occurs preferentially at a N
atom near an acetate group.
The relative concentrations of free NOTMP and its MgII
complex can be assessed accurately from integration of 31P
resonance areas without rigorous experimental conditions, since
their NOEs and longitudinal relaxation rates are equal. The
effects of pH and temperature on the Kd can be evaluated using
the thermodynamic data presented here, allowing an accurate
calculation of free MgII levels under a variety of solution
conditions. The pH effects can be easily evaluated using the
concentration initially present, while [Mg]add represents the
amount of MgII added during the experiment. It was assumed
that the initially bound MgII, by other components of the plasma,
would not be released during the experiment. Therefore, the
definition for Kd can be rewritten as eq 4. Plotting [Mg]add
-
[Mg]add - [MgL] ) yKd - [Mg]ini (4)
[MgL] Versus y will give linear plots with Kd as the slope and
-[Mg]ini as the intercept. The results for the titration of the
blood plasma (pH 7.8) are presented graphically in Figure 6,
together with the 31P spectrum of the plasma sample before
addition of MgII.
Figure 6 shows that the Kd (slope) of NOTMP was lower in
the blood plasma (0.14 ( 0.01 mM, pH 7.8) than in the aqueous
buffer experiments (0.35 ( 0.03 mM at pH 7.4, Table 2). The
difference can be completely ascribed to the pH difference in
the two experiments. The most remarkable feature of Figure
6, however, is the intercept (-0.66 mM). Since the plasma
was diluted to 81% at the beginning of the experiment, this
means that [Mg]ini in pure plasma is 0.82 ( 0.02 mM. The
same value (0.82 mM) was obtained from the initial sample
without added MgII, by measuring the 31P peak integrals and
using a Kd ) 0.14 mM. The total MgII concentration in pure
plasma was 1.08 ( 0.01 mM, as measured by atomic absorption
spectroscopy. This indicates that the free MgII concentration
in plasma is 76% of the total MgII concentration, identical to
results reported recently using an ion specific electrode.20 The
(20) Altura, B. T.; Shirey, T. L.; Young, C. C.; Hiti, J.; Dell’Orfano,
K.; Handwerker, S. M.; Altura, B. M. Methods Find. Exp. Clin. Pharmacol.
1992, 14, 297-304.

4404 J. Am. Chem. Soc., Vol. 118, No. 18, 1996
Huskens and Sherry
equation Kd(pH) ) 10(7.4-pH)Kd7.4, because of the large gap
between the first and second protonation steps.
Acknowledgment. This research was supported in part by
grants from the Robert A. Welch Foundation (AT-584) and the
NIH Biotechnology Research Program (P41-RR02584).
The methyl 1H resonance of the methylphosphinate group is
a sharp doublet in Vitro, for both the free ligand and the MgII
complex. The chemical shift difference, however, is small (0.07
ppm). For in ViVo applications, NO1A-2MP might be more
useful because its shift separation is somewhat larger (0.11 ppm).
Still, 31P decoupling and the use of multiple pulse sequences
might be necessary to obtain reliable free MgII levels using 1H
in ViVo NMR spectroscopy.
Supporting Information Available: Text describing the
detailed synthesis of NO1A-2MP and NOTMP, figure showing
the 1H NMR spectrum of NOTMP and Mg(NOTMP), and tables
giving the NMR shift data for tBuNO1A and tBu2NO2A (1H
and 13C), and for NO1A-2MP and NO2A-MP (1H) (7 pages).
This material is contained in many libraries on microfiche,
immediately follows this article in the microfilm version of the
journal, can be ordered from the ACS, and can be downloaded
from the Internet; see any current masthead page for ordering
information and Internet access instructions.
The results presented here suggest that the ligand NOTMP
may prove useful for determining free MgII levels in plasma or
for monitoring changes in [Mg]f in perfusates during 31P NMR
studies of perfused organs. For intracellular use, derivatization
will be necessary.
JA953771P