ANALYST, MARCH 1993. VOL. 118 24 1 Detection of Aluminium(iii) Binding to Citrate in Human Blood Plasma by Proton Nuclear Magnetic Resonance Spectroscopy Jimmy D. Bell NMR Unit, Hammersmith Hospital, Du Cane Road, London, UK W12 OHS Gina Kubal, Stojan Radulovic, Peter J. Sadler" and Alan Tucker Department of Chemistry, Birkbeck College, University of London, Gordon House and Christopher lngold Laboratories, 29 Gordon Square, London, UK WCIH OPP Reactions of AP+ (50-500 pmol 1-1) with intact blood plasma and its low relative molecular mass ultrafiltrate (<5 kDa) have been studied by proton nuclear magnetic resonance spectroscopy. Binding t o citrate was detected and was reversed by addition of desferrioxamine. The use of combined exponential and sine-bell functions for the resolution enhancement of spectra of plasma is illustrated.Keywords: Proton nuclear magnetic resonance spectroscopy; aluminium; citrate; desferrioxamine; blood plasma There is a great deal of current interest in the speciation of AF+ in the body. Aluminium is not thought to be an essential element for mammals, and has only toxic effects.',* The speciation of AF+ determines its biological availability,3 and there is a need to understand the processes by which Al3+ is absorbed into the blood stream and either excreted, or, when the kidneys do not function efficiently, is deposited in the brain." The main carriers of AP+ in blood plasma are thought to be citrate , transferrin and possibly albumin.5-10 Calculations of the distributions of AP+ species are usually based on thermodynamic stability constants and do not take kinetic factors into account. 1 1 Speciation via the separation and isolation of complexes can be complicated by changes in the equilibria during the separation process.We have there- fore used nuclear magnetic resonance (NMR) spectroscopic methods to study AF+ complexation as intact blood plasma can be studied, and kinetic measurements can be made. Blood plasma is a complicated heterogeneous mixture of fat particles (the lipoproteins) , proteins (such as immunoglobu- lins, albumin and transferrin), and small molecules and ions.12 The lipoproteins are organized particles consisting of triacylgl ycerols, phospholipids, free and esterified cholesterol and proteins; some of the lipids are relatively mobile within the core of the particles.The smallest and densest particle HDL (high density lipoprotein) contains the highest protein and phospholipid content. Several plasma proteins are in- volved in binding functions. For example, albumin (66 kDa), the major plasma protein (concentration approximately 0.65 mmol 1-I), binds fatty acids, metal ions such as Zn2+ and Ca", hormones and various drugs,l3 and transferrin (an 80 kDa glycoprotein) transports iron as FeT+ along with carbo- nate anions.14 The details of many of these biologically- important binding processes, including the mechanisms of metal ion and small molecule uptake and release by plasma proteins, are currently poorly understood. Proton (1H) NMR spectra of blood plasma consist of a complicated mixture of broad and sharp resonances, and normal single-pulse spectra are often difficult to interpret.Previously the broad resonances have been filtered out using Hahn or CPMG (Carr-Purcell-Meiboom-Gill) spin- echo seque'nces. This leaves the resonances from the mobile small molecules and a few highly mobile parts of macromole- cules.12.1"16 We have assigned resonances for the N-acetyl groups of mobile glycan chains of acute-phase plasma proteins,17 and resonances for the mobile parts of chylo- microns, very low density lipoproteins (VLDL) , low density * To whom correspondence should be addrcssed. lipoproteins (LDL) and HDL.18 It is even possible to distinguish between these various classes of lipoproteins. I n some situations the peaks for these mobile protons are relatively intense and can be used, for example, to monitor disease states such as diabetes or lipidaemia,ly and to investigate the binding of paramagnetic Cu2+ ions.") In this paper the reactions of AP+ with intact blood plasma and a low relative molecular mass ( M , ) ultrafiltrate, and the reversal of the binding processes with the clinically-used chelating agent desferrioxamine, are studied.The use of combined exponential and sine-bell functions for resolution enhancement of 1H NMR spectra is illustrated and discussed. Experimental Materials Blood plasma samples were prepared from freshly drawn heparinized venous blood taken from normal healthy volun- teers. Cells were removed by centrifugation at 277 K. Sodium hydrogen carbonate (25 mmol 1-1) was added to maintain the pH at 7.4.Protein-free low M , ultrafiltrates (<5 kDa) were prepared using Amicon Centrifree filtration devices, which had been thoroughly washed with water ( 5 ~ ) to remove preservatives, and centrifugation for 1 h at 277 K. Aluminium was added in microlitre aliquots of an aqueous stock solution of AIC13 (Fluka). Desferrioxamine (Desferral) was obtained from Ciba-Geigy and trisodium citrate was purchased from Aldrich. NMR Spectroscopy The 'H NMR spectra were recorded on Bruker AM500 and JEOL GSX500 instruments at 500 MHz, using 0.55 ml of solution in a 5 mm tube, ambient temperature, 128 transients, 45" pulses, relaxation delay 1.6 s, 16 k data points (zero-filled to 32 k), 6 kHz spectral width, and gated or continuous secondary irradiation of HOD.For Hahn spin-echo spectra a 90"-~-180"-~-collect free induction decay (FTD) (typically T = 60 ms) sequence was used." Exponential functions equivalent to a line-broadening of 0.5-2 Hz were used for processing where necessary. For resolution enhancement, FTDs were typically processed using exponential functions equivalent to line-broadenings of 1-2 Hz combined with unshifted sine-bell functions. Data processing was carried out on a Bruker Aspect 1000, or SUN SPARC2 (Varian VNMR software), or Silicon Graphics personal IRIS 4D35TG using FELIX software (D. Hare).242 ANALYST, MARCH 1993, VOL. 118 The effect of enhancement functions on peak heights was investigated using the PANIC simulation program (Bruker Spectrospin). Spectra containing single lines of different widths at half height were simulated followed by inverse Fourier transformation to produce the FIDs.Results and Discussion In Fig. 1, 500 MHz 1H NMR spectra of the methyl and methylene regions of heparinized human blood plasma are compared. In the normal, single-pulse spectrum [Fig. l(A)], some sharper resonances superimposed on a broad envelope of resonances can just be discerned. The broad peaks arise predominantly from fatty acids (triacylglycerols and phospho- lipids) and the major protein albumin. Two methods of filtering out the broad peaks are shown. First, the Hahn NAc I :i" A W I ! 'HB 3.0 Fig. 1 Aliphatic region of 500 MHz 1H NMR spectra of human blood plasma. A, Single-pulse spectrum: FID multiplied by an exponential function (equivalent to a line broadening of 1 Hz) prior to Fourier transformation. B.Resolution-enhanced single-pulse spectrum: FID multiplicd by an cxponcntial function (equivalent to a line broadening of 2 Hz) followed by an unshifted sine-bell function prior to Fourier transformation. C, Hahn spin-echo spectrum: T = 60 ms, FID multiplied by an exponential function (equivalent to a line broadening of 2 Hz) prior to Fourier transformation. NAc = N-acetyls of glycoproteins; Gln = glutamine; Val = valine; HB = hydroxybuty- rate; and Lac = lactate. P = lipoprotein pcaks: PI, HDL and LDL CH,; P,, VLDL and chylomicron CH,; P3, HDL and LDL CH,; and P4, VLDL and chylomicron CH2 spin-echo spectrum Fig. 1(C). With this procedure the contribution of magnetic field inhomogeneity to linewidths is removed, but linewidths are not otherwise reduced.Broad resonances are filtered out because the magnetization asso- ciated with them decays before acquisition begins (spin-spin relaxation times, T2, short compared with the total delay 2 X T). Intensities are also reduced by diffusion in local inho- mogeneous fields in the sample during the delay,22 and by phase modulation of the multiplets.23 The choice of t is a compromise between removing sufficient of the broad signals to achieve the simplification required, and retaining interpret- able phase modulation.16 With t = 60 ms, most of the doublets are inverted (T = 1/2 J ) . The largest peaks (P) are those for the CH3 (P2) and CH2 (P4) groups of chylomicrons and VLDL,18 and the N-acetyl groups of the glycan chains of acute-phase glycoproteins such as al-acid glycoprotein.17 Well-resolved sharp multiplets include those for valine, hydroxybutyrate, lactate, alanine and citrate. Secondly, a normal spectrum is shown after enhancement using combined exponential and sine-bell functions. The over-all appearance of this is similar to the spin-echo spectrum, except that there is no phase modulation, and partly as a result of this, some peaks are enhanced in intensity, e.g., P1 and P3. The peaks for glutamine are now interpretable, whereas in the spin-echo spectrum they are severely distorted. The advantage of resolution enhance- ment is that it can be carried out after the single-pulse spectrum has been acquired, and does not require acquisition of further experimental data. Resolution enhancement is a well-known procedure in NMR spectroscopy,24.2~ but the combined application of exponential and sine-bell functions, although proposed ,26,27 has been little used in the past.The sine-bell function26 is a commonly uscd window function, especially for studies of proteins. In the simplest use of this function, the FID is multiplied by a half cycle of a sine function with a period of twice the acquisition time. However, the improvement in resolution obtained with this function is often accompanied by an excessive degradation of the signal-to-noise ratio and distortion of baselines. We have countered this problem by pre-multiplication of the FID by a line-broadening exponen- tial function, a procedure that we have found useful for the study of the large plasma proteins albumin and trans- ferrin .28-30 The function begins at zero (ensuring suppression of the broadest lines), rises to a maximum, which might be at an earlier point in the time domain than for the sine-bell (depending on the exponential function used), and then decays smoothly to zero.Mathematically the sine-bell func- tion produces an FID that, after Fourier transformation, is equivalent to the discrete differential of the dispersion spectrum ,27 yielding a spectrum with inherently narrower lines, but with a poorer signal-to-noise ratio than the absorption spectrum. The introduction of either a phase-shift of the sine-bell function or pre-multiplication by an exponen- tial function effectively adds a fraction of the absorption spectrum to the dispersion differential, improving the signal- to-noise ratio and broadening the lines slightly.The phase- shifted sine-bell does not begin at zero and is not as effective in suppression of broad resonances. We investigated the effect of sine-bell and exponential sine- bell resolutim enhancements on the intensity of resonances using simulated spectra. The FIDs for Lorentzian lines of varying widths at half height (2-30 Hz) were simulated and transformed with and without pre-multiplication with en- hancement functions. The reduction of pcak heights in enhanced spectra resulting from linewidth changes was determined. The curve for exponential sine-bell enhancement was much shallower than that for sine-bell, i.e., enhanced spectra are less affected by linewidth changes.For example, if a spectrum is enhanced using the sine-bell function, then an increase in peak width from 2 to 5 Hz leads to a 45% reduction in peak intensity, and an increase from 2 to 10 Hz to a 71%ANALYST, MARCH 1993. VOL. 118 243 DFO I 1 I 1 I I I I 1 1 I I I I II I I II II II I I I I I I I1 II II Citrate 3.0 2.0 1 .o h (PPm) Fig. 2 Aliphatic region of resolution-enhanced 500 MHz 1H NMR spectra of human blood plasma. A. Control; B. after addition of AF+ (SO pmol 1-1); and C, after further addition of desferrioxamine (100 pmoll-l). Exponential functions (equivalent to a line broadening of 2 Hz) followed by unshifted sine-bell functions were applied to the FIDs prior to Fourier transformation. DFO = desferrioxamine reduction, whereas the corresponding reductions i n peak intensity are 28% (increase in linewidth from 2 to 5 Hz) and 56% (from 2 to 10 Hz), respectively, as a result of using exponential sine-bell enhancement. The choice of acquisition time greatly affects the enhancement procedure when sine- bell functions arc used, as the period of the function is normally set to twice that of the acquisition time, i.e., the maximum of the function is at half the acquisition time.Hence the choice of acquisition time determines the portion of the FID that is maximally enhanced. The effect of AP+ on the resolution-enhanced spectrum of blood plasma i s shown in Fig. 2. After addition of 50 pmol 1-1 AP+ the intensities of the peaks forming the AB quartet for citrate near 2.6 ppm, present at a concentration of 0.11 mmol 1-1 in this sample of plasma (i.e., within the normal range), markedly and selectively decrease intensity [Fig.2(B)]. No further change was observed in the spectrum for the next 24 h. The citrate peaks disappear from the spectrum completely after addition of 200 pmol 1-1 AP+, and no other resonances were affected up to a concentration of 500 pmol l-1 t - m lz m m .- I I 1 I I I I I I I I I I I B I Fig. 3 Aliphatic region of 500 MHz 'H NMR spectra of a low M , ultrafiltrate (<5 kDa) of human blood plasma. A , Control; B, after addition of 50 pmol I-' A13+; and C, after further addition of desferrioxamine (100 vmol 1- I ) . Most of the multiplets between 3.2 and 3.9 ppm are assignable to glucose Al3+ (data not shown). Addition of dcsferrioxamine to the plasma sample containing 50 pmol 1-1 A P + led to the rapid (minutes) reappearance of citrate peaks, Fig.2(C), although in a slightly broader form than they were originally. Addition of desferrioxamine alone to plasma had no effect on the peaks for plasma components. Single-pulse IH NMR spectra of the low M , (<5 kDa) ultrafiltrate of blood plasma are shown in Fig. 3. Now there are no broad peaks in the spectrum because the high M , macromolecules have been separated out. It can be seen that addition of Al3+ (50-300 pmol 1-1) causes a specific decrease in intensity of the citrate peaks, whilst there is little effect on peaks for other small molecules such as lactate and alanine. Addition of desferrioxamine restored the intensity of the citrate peaks again, Fig.3(C). The 1H NMR spectra of model systems containing A13+ and citrate were studied. The spectrum of a 0.75 + 1.0 mixture of Al3+ and citrate at pH* 6.6 (pH* = pH meter reading in D20 solution; adjusted with NaDC03) for example, gave rise to about 20 quartets covering the range 2.2-3.5 ppm. Evidently244 ANALYST, MARCH 1993, VOL. 118 there are a large number of different coordination modes for citrate in AP+-citrate complexes. This conclusion has already been drawn by others from 13C NMR work,31 and from X-ray crystallography of isolated complexes,32 and appears to explain why new peaks for an aluminium citrate complex are not readily observable in spectra of plasma or its ultrafiltrate. The crystalline trinuclear complex [AI3(H- ,Cit),(OH)- (H20)]+ isolated from 1 + 1 mixtures of AP+ and citrate (Cit) in the pH range 7-9,32 contains three similar but distinct six-coordinate AP+ ions bridged by tetradentate citrate anions; the hydroxyl and carboxyl groups are deprotonated.Curiously, the 1H NMR spectrum of this 1 + 1 cluster was reported to consist of a single multiplet at 4.71 ppm (i.e. , very close to the water resonance). As the citrate ligands are magnetically-inequivalent, a total of six quartets arising from four magnetically-inequivalent protons per citrate might have been expected. Desferrioxamine is known to bind AF+ more strongly than citrate o r transferrin, and is used in the clinic for decreasing plasma aluminium levels after exposure to toxic levels .33 These 'H NMR data suggest that AP+ added to heparinized blood plasma (or plasma ultrafiltrate) in vitro initially binds to citrate.Citrate, which has three ionized carboxylate groups at neutral pH (pK values = 2.87,4.35 and 5.69),34 is known to be a strong ligand for A13+,11 and probably plays a role in promoting the absorption of aluminium from food into the blood stream. Here much of the AP+ at the lowest dose studied (SO pmol 1-1) could have been transferred to trans- ferrin, which is normally present at a concentration of 30-40 pmol l-1 in plasma and usually has only about one third of the Fe3+ sites (two per protein molecule) occupied. 14 This would leave about 4&53 pmol l-I of vacant sites available for binding Ali+. Perhaps citrate to transferrin transfer is very slow under the conditions used.There is good evidence that transferrin is the ultimate binding agent for AP+ in vivo,s.~ and AP+ is known to bind strongly to transferrin in vitro.h.7 An alterna- tive, and less likely, explanation for our results is that AP+ binding to transferrin (or another protein) induces a confor- mational change that in turn leads to citrate binding to the protein and broadening of resonances because of slow tumbling. This is difficult to rule out at the moment but might be clarificd when experiments on AF+ binding to transferrin in the presence of citrate are studied by NMR spectroscopy. So far our own attempts,35 and those of othcrs,'h to detect Al?+ binding to transferrin by 27Al NMR have not been successful, presumably due to quadrupolar broadening. The IH NMR methods such as those described here should be useful for investigating the effectiveness of potential thera- peutic chelating agents designed to remove aluminium from blood plasma.We thank the MRC, SERC, Wellcome Trust, Wolfson Foundation and ULlRS for their support for this work. We also thank Dr. M. C. Grootveld (London Hospital Medical School) and Dr. R. W. Evans (Guy's Hospital) for stimulating discussions. This paper is based on work presented at the XXVII-CSI Post-Symposium in Loen, Norway, June 16-18, 1991. References 1 Aluminum in Chemistry, Biology and Medicine, eds. Nicolini, M., Zatta, P. F., and Corain, B . , Cortina International. Verona. 1991. 2 Aluminium in Biology and Medicine, eds. Chadwick. D., and Whelm, J., CIBA Foundation Symposium 169, Wiley, New York. 1992.Daydk, S . , Filella, M., and Berthon, G., J . Inorg. Riochem., 1991,38,241. Candy, J . M., Oakley, A. E., Klinowski, J . , Carpenter, T. A., Perry, R. H., Atack, J . R., Perry, E. K., Blcsed, G.. Fairbairn, A , , and Edwardson, J. A., Lancet, 1986, i , 354. Martin. R. B., Clin. Chem. ( Winston-Salem N.C.), 1986, 32, 1797. Harris, W. R.. and Sheldon, J . , Inorg. Chem., 1990, 29, 119. Fatemi, S. J . A . , Kadir, F. H. A., and Moorc. G. R., Biodzem. J . , 1991, 280, 527. Day, J. P., Barker. J.. Evans, I . J. A . , Perk\, J . , Seabright, P. J . , Ackrill, P., Lilley, J . S . , Drumm. P. V., and Newton, G. W. A.. Lancet, 1991, 337, 1345. 9 Venturini, M., and Berthon, G.. J . Inorg. Bioclzem., 1989, 37, 69. 10 Fatemi, S. J . A., Williamson, D. J., and Moore.G. R., J . Inorg. Riochem., 1992, 46, 35. 11 Martin, R. B., in Aluminum in Ciicwzistry, Biology und Medicine, eds. Nicolini, M., Zatta, P. F., and Corain, B., Cortina International, Verona, 1991, pp. 3-20. 12 Bcll, J. D., Brown, J . C . C., and Sadlcr, P. J.. Chem. Br. 1988, 24, 1021. 13 Peters, T., Jr., Adv. Prolein Chem., 1985, 37, 161. 14 Crichton, R. R., Inorgunic Biochc~mistry of Iron Metubolism, Ellis Horwood, Chichester, 1991. 15 Nicholson. .J. K., Buckingham, M. J., and Sadler. P. J., Biochem. J., 1983, 211, 605. 16 Bell, J. D., Brown, J . C. C., and Sadler, P. J . , NMK Biomed., 1989. 2, 246. 17 Bcll, J . D.. Brown, J . C. C . , Nicholson, J . K., and Sadlcr, P. J . , FEBS Lctt., 1987. 215, 311. 18 Bell, J . D . , Sadler, P. .I.. Macleod, A. F., Turner, P.R.. and LaVille, A., FEBS Lett., 1987, 219, 239. 19 Nicholson, K. K., O'Flynn, M. P.. Sadler, P. J . , Macleod, A. F., Juul, S. M . , and Stinksen, P. H., Biochem. J . , 1984,217, 365. 20 Bligh, S. W. A., Boyle. H . A., McEwen, A. B., Sadler, P. J . , and Woodham, R. H., Biocliem. Phurm., 1992, 43. 137. 21 Hahn, E. L., Piiys. Rev., 1950, 80, 580. 22 Brown. F. F., and Campbell, I . D., Proc. R. Soc. London Ser. R, 1980, 289, 395. 23 Rabenstein, D. L., and Naka\hima, T. T., Anal. Chem., 1979, 51, 1465. 24 Lindon, J. C., and Ferrige, A . G., Prog. NMR Spectrosc'., 1981, 14, 27. 25 Sanders, J . K. M., and Hunter. B. K., Modern NMR-Spectro- scopy. A Guide for Chemists, Oxford University Press, Oxford, 1987. 26 De Marco. A., and Wuthrich, K., J . Magn. Reson., 1976, 24, 201. 27 Gudron, M., J . Magn. Reson., 1978, 30, 515. 28 Kubal, G., Sadler, P. J . , and Evans, R., J . Am. Chem. Soc., 1992, 114, 1117. 29 Kubal, G., Mason, A. B., Sadler, P. J., Tucker, A . , and Woodworth, R. C., Biochem. J . , 1992, 285, 71 1. 30 Sadler, P. J.. and Tucker, A., Eur. J. Riochem., 1992,205,631. 31 Fatemi, S. J . A., Kadir, F. H. A., Williamson, D. J., and Moore, G. R., Adv. Inorg. Chem., 1991, 35, 409. 32 Freng, T. I>.. Gurian, P. L., Healy, D., and Barron, R., Znorg. Chem., 1990, 29, 408. 33 Leung, F. Y . , Hodsman. A. B., Muirhcad, N., and Henderson, A. R., Clin. Chem. ( Winston-Salem N. C.), 1985. 31, 20. 34 Martell, A. E., and Smith, R. M., Critical Sfahifity Constants. Plenum, New York, 1977, vol. 3, p. 161. 35 Kubal, G., Kiang, W., and Sadler, P. J . , unpublished work. 36 Fatemi, S. J . A., Williamson, D. J . , and Moore, G. R., J . lnorg. Biochem., 1992, 46, 35. 3 4 5 6 7 8 Puper 2/03] 47G Received June 16, I992 Accepted October 28, 1992