Anal. Chem. 1999, 71, 265-267
Silicon-29 Nuclear Magnetic Resonance Spectroscopy Detection Limits Christopher T. G. Knight* and Stephen D. Kinrade†
School of Chemical Sciences, University of Illinois at UrbanasChampaign, 600 South Mathews Avenue, Urbana, Illinois 61801, and Department of Chemistry, Lakehead University, 955 Oliver Road, Thunder Bay, Ontario, Canada P7B 5E1
We show that an unmodified, commercially available highfield (17.61T) NMR spectrometer using the DEPT pulse sequence is capable of detecting silicon-containing species down to concentrations of 150 ng/mL (150 ppb) per spin site. This is in the range given for the concentration of silicon in the blood of silicone breast implant recipients, as determined by ICP analysis, and demonstrates that, contrary to the view expressed in the literature, in theory 29Si NMR may be sufficiently sensitive to be of use in determining the nature of the silicon-containing species present. A summary of the factors affecting the detection limits in NMR spectroscopy is given. There has been some discussion1 as to whether the signals apparent in the 29Si NMR spectra of blood from silicone (poly(dimethylsiloxane), PDMS) breast implant recipients are real. These signals were reported by Garrido and co-workers2 and purport to show that a variety of silicon-containing species are present in measurable amounts. The results of more traditional means of chemical analysis, such as inductively coupled plasma atomic emission spectroscopy3 (ICP-AES) and graphite furnace atomic absorption spectroscopy4 (GFAAS), as reported by several research groups, reveal silicon blood levels in the nanograms per milliliter range in both implant recipients and non-implanted volunteers alike. This is some 4-5 orders of magnitude less than the silicon blood levels suggested by Garrido et al. on the basis of 29Si NMR spectroscopy. At least two groups1,4 have cited this difference as evidence that 29Si NMR spectroscopy may be ill suited to detecting silicon-containing species in human blood. It has been strongly implied that 29Si NMR is not sufficiently sensitive to observe silicon concentrations in the nanograms per milliliter range. We have been working with 29Si NMR of aqueous siliconcontaining solutions for many years.5 Using appropriate experi* Address correspondence to this author at Box 61-1, School of Chemical Sciences, University of Illinois at UrbanasChampaign, 600 S. Mathews Ave., Urbana, IL 61801. Tel: (217) 398-6390. Fax: (217) 333-8868. E-mail: blindpig@ rbc6000.cvm.uiuc.edu. † Lakehead University. (1) Taylor, R. B.; Kennan, J. J. Magn. Reson. Med. 1996, 36, 498-499. (2) Garrido, L.; Pfleiderer, B.; Jenkins, B. G.; Hulka, C. A.; Kopans, D. B. Magn. Reson. Med. 1994, 31, 328-330. (3) Teuber, S. S.; Saunders, R. L.; Halpern G. M.; Bruckner, R. F.; Conte, V.; Goldman B. D.; Winger, E. E.; Wood, W. G.; Gershwin, M. E. Biol. Trace Elem. Res. 1995, 48, 121-130. (4) Macdonald, P.; Plavac, N.; Peters, W.; Lugowski, S.; Smith, D. Anal. Chem. 1995, 67, 3799-3801. 10.1021/ac980547k CCC: $18.00 Published on Web 11/06/1998
© 1998 American Chemical Society
mental parameters, the technique is certainly capable of detection down to nanograms per milliliter levels, as we show below. Whether NMR spectroscopy is capable of identifying any actual silicon-containing species that may be present in human blood, however, depends not only on the detection limits of the spectrometer but also on the aqueous chemistry of silicon in blood, a topic about which little is known. Of course, systems in which silicon is present in only a single chemical environment will be more amenable to analysis than systems in which the silicon signal is distributed over a variety of chemical sites. Likewise, chemical exchange and relaxation considerations will also prove crucial to determining the final applicability of NMR spectroscopy. We realize that the chemistry of the model system we use here is likely to be significantly different from that of any silicon-containing species encountered in human blood. Our aim, however, is to show that, given favorable chemistry, 29Si NMR is at least capable of providing information on systems at biologically relevant concentrations. This is contrary to the views expressed in the literature.1,4 It must be said, however, that the experimental methodology used by Garrido and co-workers2 is entirely inappropriate in the context of maximizing NMR detection limits in solution. This is primarily due to their use of small MAS rotors, which are designed for solids and have sample volumes less than 10% that of the 10 mm o.d. NMR tubes used in this work. The use of a lower magnetic field (9.4 vs 17.6 T used here) also contributes to a decrease in the sensitivity of the NMR experiment. Additionally, they appear to have made no attempts either to shorten the long T1 (29Si) relaxation times that might be expected for PDMS-related compounds by doping samples with paramagnetic relaxation agents, nor to increase sensitivity by employing polarizationtransfer sequences. RESULTS AND DISCUSSION Establishing the NMR detection limits6 for small mobile molecules in solution is straightforward, as is establishing experimental conditions that will maximize NMR sensitivity. The (5) See, for example: Knight, C. T. G. J. Chem. Soc., Dalton Trans. 1988, 1457-1460 and references therein. Kinrade, S. D.; Syvitski, R. K.; Marat, K.; Knight, C. T. G. J. Am. Chem. Soc. 1996, 118, 4196-4197 and references therein. (6) We are aware of the IUPAC definition of the term “limits of detection” (see Long, G. L.; Winefordner, J. D. Anal. Chem. 1983, 55, 712A-724A, for example) and note that in this work we are simply referring to the term in the wider context of the practical sensitivity of a commercial NMR spectrometer, no statistical analysis being implied.
Analytical Chemistry, Vol. 71, No. 1, January 1, 1999 265
factors that govern the sensitivity of the NMR experiment are well established.7 The detection limits depend on a variety of factors, related to both the spectrometer itself and the system under investigation. Essentially, for each scan acquired, the signal-tonoise ratio of an NMR signal is proportional to
γ4 B02NB1g(n)T-1
(1)
where γ is the magnetogyric ratio of the observed nucleus, B0 is the external magnetic field, N is the number of magnetically active nuclei in the sample and is a function of the concentration of the solute and the level of isotopic enrichment, B1 is the strength of the observing rf magnetic field, g(n) is the line shape factor, and T is the temperature of the sample. To obtain the maximum sensitivity for a given nucleus, the experiment should be conducted (a) at the highest possible external magnetic field, (b) with the greatest number of magnetically active spins in the sample, (c) over the smallest possible sweep width, (d) under conditions that give the sharpest resonances, such as the most homogeneous B0 field possible, and (e) at the lowest temperature possible. Additionally, the achievable signal-to-noise ratio increases with the square of the number of scans acquired; consequently, to double the signal-to-noise ratio of a given resonance, the experiment must be run for 4 times as long. The system under investigation also influences the achievable signal-to-noise ratio for a given spin site. The relevant factors are the longitudinal (T1) and transverse (T2) relaxation time constants associated with the spin sites present, the existence of other coupled magnetically active nuclei, and the chemical lifetime of the site being observed. For optimum sensitivity, the observed spin site must (a) exhibit the shortest possible T1 relaxation time, to allow the largest number of scans to be accumulated in a given time, (b) exhibit the longest possible T2 relaxation times, to yield the sharpest signal, (c) be decoupled from other magnetically active centers, so as to yield only a single sharp line, and (d) be stable on the NMR time scale. Signal-to-noise ratios may be further enhanced, where applicable, by transferring magnetization from adjacent scalar coupled nuclei to the spin site under observation. In the case of silicone and related compounds, magnetization can be transferred from the coupled methyl protons using pulse techniques such as the distortionless enhanced polarization transfer (DEPT) sequence,8 commonly used in carbon chemistry as an assignment tool. We show in Figure 1 29Si NMR spectra of the small, mobile silicon-containing molecule hexamethyldisiloxane (HMDS) in chloroform. The spectra were chosen to illustrate the point that observing nanograms per milliliter concentrations of siliconcontaining molecules is routine, and the spectra were obtained on a commercially available, unmodified spectrometer using a standard 10 mm tunable broadband probe. Both spectra were obtained at an external magnetic field of 17.61 T (750 MHz 1H (7) Harris, R. K. Nuclear Magnetic Resonance Spectroscopy; Pitman: London, 1983; pp 9-11. Freeman, R. A Handbook of Nuclear Magnetic Resonance; Wiley: New York, 1988; pp 216-223. (8) Dodrell, D. M.; Pegg, D. T.; Bendall, M. R. J. Magn. Reson. 1982, 48, 323327. Blinka, T. A.; Helmer, B. J.; West, R. Adv. Organomet. Chem. 1984, 23, 193-218.
266 Analytical Chemistry, Vol. 71, No. 1, January 1, 1999
Figure 1. The 17.61 T 29Si NMR spectra of hexamethyldisiloxane in chloroform. Both samples contain ∼20% (v/v) deuterochloroform for field frequency locking purposes. The spectra and chemical shift scale have been set in register using the peak in the lower trace, at a chemical shift of 7.066 ppm, and have been scaled vertically by maintaining a constant noise level. The spectra were externally referenced to a solution of TMS in deuteroacetone. An exponential linebroadening function of 1.0 Hz has been applied to both spectra. (A) The solution contains 18 ppm silicon and 0.03 M Cr(acac)3. The spectrum was acquired at 298 K over a sweep width of 10 kHz, using 39 936 points, resulting in an acquisition time of 1.997 s/scan. An interpulse delay of 8 s was used, with 6784 90° pulses being accumulated, giving a total run time of ∼17 h. The 1H decoupler was enabled during the acquisition time. (B) The sample contains 1 ppm silicon, as measured by ICP analysis (standard deviation 0.25 ppm). The spectrum was acquired at 266 K over a sweep width of 8941 Hz, using 65 536 points, resulting in an acquisition time of 3.665 s/scan. An interpulse delay of 1 s was used, and 9488 90° pulses were accumulated, giving a total run time of ∼13 h. A DEPT-45 transmitter pulse sequence was used, with 2JSi-C-H ) 6.8 Hz.
resonance frequency) using 10 mm o.d. glass NMR tubes. These have a functional volume of about 3 mL, more than 10 times the volume used by Garrido and co-workers.2 The upper trace (Figure 1A) is that of a 50 ppm HMDS solution, which therefore contains 18 ppm of silicon. The shiftless relaxation compound chromium(III) acetylacetonate (Cr(acac)3) has been added to reduce the T1 (29Si) relaxation time. The spectrum required 17 h to accumulate and has a signal-to-noise ratio of 18:1. Taking an observed S/N ratio of 2:1 as being the minimum necessary to define a signal, the spectrum in Figure 1A implies a detection limit of 2 ppm. The lower trace is the 29Si NMR spectrum of a solution containing 1 ppm of silicon, as measured using ICP-AES analysis. It required 13 h to obtain, and in this case the DEPT-45 transmitter pulse sequence was used. Here the sensitivity advantage is 2-fold. First, signal intensity is gained by transferring polarization from the methyl protons, a process that gives a theoretical enhancement factor of approximately 5 (although in practice this is seldom attained; indeed, for spin systems with multiple heteronuclear J-couplings, determination of the optimal DEPT parameters can be complex, as noted recently by Alam9). Second, the recycle time in a DEPT experiment depends on the proton, not the silicon, T1 value and consequently permits much faster recycling. Indeed, it is possible to recycle even faster than in solutions to which relaxation agents have been added, while avoiding the deleterious linebroadening effects associated with them. In this case the S/N ratio is 7:1, implying a detection limit of ∼0.3 ppm. Of course, (9) Alam, T. M. Spectrochim. Acta 1997, A53, 545-552.
running the experiment for 4 times as long would halve the attainable detection limit, allowing detection down to 0.15 ppm (150 ppb), which corresponds to 150 ng/mL in aqueous solution. The use of isotopically enriched compounds allows further signal enhancements, up to a maximum of 21-fold for materials containing 100% 29Si, resulting in a theoretical detection threshold of ∼7 ng/mL.
sufficiently sensitive to be of relevance to the debate about the nature of silicon-containing species found at biologically realistic concentrations. We note that the use of dedicated rf coils, larger volume probes, and higher magnetic fields will, of course, extend the detection limit further. Work aimed at gaining a better understanding of the chemistry of silicon-containing species in human blood is in progress.
CONCLUSION We stress that we have chosen ideal circumstances and a welldocumented system, for which the NMR parameters are known, to make our point. Solutions containing silicon present simultaneously in multiple spin sites, bound to large molecules, or undergoing rapid inter- or intramolecular chemical exchange will present more difficulties. The chemistry of silicon in blood is unknown and may well be far from the ideal conditions used here, thus limiting the practical application of NMR spectroscopy. The figure of 150 ng/mL is well below the maximum value of 870 ng/mL, determined by ICP-AES for silicon in blood serum,3 although it is greater than the maximum value of 57 ng/mL determined by GFAAS.4 The reason for the discrepancy between the two ICP techniques is unclear. Given favorable chemistry, however, 29Si NMR spectroscopy may, in principle at least, be
ACKNOWLEDGMENT We thank Dr. F. Lin of the Varian Oxford Instrument Center for Excellence in NMR Laboratory at the University of Illinois in UrbanasChampaign for obtaining the NMR spectra. Funding was provided in part by the W. M. Keck Foundation, the National Institutes of Health (PHS 1 S10 RR 10444-01; GM-42208 and RR01811), the National Science Foundation (NSF CHE 96-10502), and the Natural Sciences and Engineering Council of Canada. Facilities were provided by the Illinois EPR Research Center, an NIH-supported research center.
Received for review May 20, 1998. Accepted September 29, 1998. AC980547K
Analytical Chemistry, Vol. 71, No. 1, January 1, 1999
267
