WO2001001165A1

WO2001001165A1 - Use of non-lossy solvents as a means to improve the high-field nmr sensitivity of biological samples

Info

Publication number: WO2001001165A1
Application number: PCT/US2000/016979
Authority: WO
Inventors: Robert A. Wind; Michael A. Kennedy; Paul D. Ellis
Original assignee: Battelle Memorial Institute
Priority date: 1999-06-25
Filing date: 2000-06-20
Publication date: 2001-01-04
Also published as: AU5880800A

Abstract

The present invention comprises a method and apparatus to substantially decrease the dielectric and inductive losses in the sampling of biological materials by nuclear magnetic resonance spectroscopy, thereby increasing the speed and sensitivity of NMR measurements. Biological material carried in a biologically-conducive solution is intermixed with a non-lossy solvent for characterization in the NMR. Biological materials having a size of at least 40kD may be rapidly characterized with minimal dielectric losses.

Description

USE OF NON-LOSSY SOLVENTS AS A MEANS TO IMPROVE THE HIGH-FIELD NMR SENSITIVITY OF BIOLOGICAL SAMPLES

FIELD OF THE INVENTION

The present invention relates generally to a process for increasing the rate of data set acquisition and sensitivity of nuclear magnetic resonance (NMR) spectroscopy in large external fields for biological molecules.

BACKGROUND OF THE INVENTION

The analysis and characterization of various samples can be performed by a number of disparate analytical techniques. However, for structure determinations of large biological molecules, and in particular, proteins, nucleic acids and their complexes, are often best characterized by way of either X-ray crystallography or by NMR spectroscopy.

However, the structure determination by NMR methods is more complicated by large molecules because they tend to reorient slowly in aqueous solvents relative to smaller molecules resulting in a shorter spin- spin relaxation times (T₂). In order to overcome the T₂ problem, a number of approaches have been proposed. For example, Farmer et al (Farmer II, B.T., Benters, R.A., Metzler, W.J., Farmer, B.T., Spicer, L.D. & Mueller, L. (1995) J. Am. Chem. Soc. 117, 9592-9593) have used deuteration to reduce the dipolar field so that deuterium-decoupled triple resonance could be utilized.

Wand et al (Wand, A.J., Ehrhardt, M.R., & Flynn, P.F. (1998) Proc. Natl. Acad. Sci. USA Vol 95, pp15299-15302) have produced proteins in reverse micelles dissolved in low-viscosity fluids in order to conduct NMR analysis of large biological proteins. The protein ubiquitin was encapsulated in reverse micelles prepared in an alkane solvent (liquid propane) without significant structural distortion. Gaemers et al (Gaemers, S., Elsevier, C.J., & Bax, A., (1999) Chem. Phys. Letters 301 , 138-144) utilized liquid C0₂ to make NMR measurements on peptides and proteins. When possible, determinations of protein structure by X-ray crystallography has heretofore been the standard, since the data sets needed to obtain the protein structure could be determined in a matter of hours or days rather than the weeks it took with conventional NMR procedures. Historically, sensitivity of NMR determinations has suffered from a number of different factors, including resistive losses in the leads and capacitors in the LC circuit; dielectric losses in the sample due to the finite imaginary portion of the dielectric permitivity; dielectric and inductive losses in the sample due to the finite electrical conductivity; efficiency of the NMR coil; and losses in the NMR coil itself and the noise factor of the receiver.

Therefore, there is a critical need for NMR processes that not only make the characterization of large proteins possible (Wand, et al.), but that also radically increase the rate at which such the NMR data can be collected.

SUMMARY OF THE INVENTION

The present invention comprises a method and apparatus to substantially decrease the dielectric and inductive losses in the sampling of biological materials by nuclear magnetic resonance spectroscopy, thereby increasing the speed and sensitivity of NMR measurements. The process comprises the steps of preparing a non-lossy solvent, preparing the biological molecules in a biologically-conducive solutions, adding the solvent of step "a" to the solution of step "b" to create a sample solution, and determining the structure of the biological material by NMR spectroscopy. The biological materials of primary concern herein are large complex molecules such as proteins and nucleic acid, wherein the size of the molecule is on the order of at least 40kD. Preferably, the non-lossy solvent comprises a non-polar, non-conductive solution which has the dielectric and inductive loss characteristics of air. Examples of non-lossy solvents include hydrocarbons (alkanes) or supercritical fluids (such as C0₂). The biological material may also be constrained within reverse micelles. Generally speaking, in order to prevent degradation of the biological material, it must be carried in a biologically-conducive solution, such as an aqueous salt solution.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

Non-lossy solvents as used herein refer to solvents which mimic the dielectric and inductive losses characteristic for that of air in an NMR; that is, there are no significant dielectric and inductive losses experienced in the characterization of the sample as a result of the solvent within which the solute is contained. For example, water is a highly polar solvent, and because most biological material (including proteins) is water soluble and their native environment is water-based, heretofore protein determinations were carried out in a solvent that create significant dielectric losses. In addition, biological systems contain a number of salts (typically < 200mM) which further exacerbate the problem of dielectric losses. This solvent combination produces inductive losses that arise from the formation of eddy currents in the sample.

In typical prior-art NMR determinations, these sample losses dominate the probe characteristics utilized in the NMR experiment. The losses increase dramatically with the frequency of the spectrometer and the sample diameter. Hence, the higher the field or the larger the molecule at a fixed field, the worse the problem becomes.

In addition to characterization of the human genome, there is intense scientific interest in identifying the structure and function of the approximate 100,000 gene products (proteins) for by the genome.

Whereas X-ray crystallography and NMR are the two preferred methods of making this determination, NMR spectroscopy has several unique capabilities compared to X-rays: proteins can be examined under physiological solution-state conditions; dynamic regions of the proteins can be well-characterized; and intermolecular complexes can be easily studied as a function of pH, ionic strength, etc. However, data collection of a single protein by NMR can take as long as approximately 60 days (as opposed to 2-4 hours in some instances with X-rays), and the size of proteins amenable to NMR characterization is currently limited to approximately 40kD.

In the preferred embodiment of the invention, the use of non-lossy solvents will significantly increase the NMR sensitivity and, therefore, the speed with which NMR determinations may be made. In addition, the sensitivity of the NMR may be increased by: (i) reducing losses in the NMR coil and the noise factor of the receiver; (ii) reducing resistive losses in the leads and capacitors in the LC circuit; (iii) reducing dielectric losses due to the finite imaginary part of the dielectric permittivity;

(iv) reducing dielectric and inductive losses in the sample due to the finite electrical conductivity; (v) reducing efficiency loss from sub-optimal coil volume and type; (vi) reducing losses from sub-optimal sample volume; and (vii) increasing the external magnetic field to as large as possible. As important as these factors are, however, they are dwarfed by the losses experienced from the use of aqueous salt solutions as the solvent for the biological protein macromolecules. Therefore the use of non-lossy solvents will have the greatest impact on improvements in the sensitivity of NMR structure determinations of large biomolecules.

One embodiment of the present invention utilizes non-polar, non- conductive solvents containing the protein molecules in reverse micelles. Incorporating all of the improvements noted above will reduce the NMR data set collection time by one to two orders of magnitude. Furthermore, the use of the non-polar solvents permits the NMR characterization of proteins having a size of at least 40kD, and preferably as large as 100kD. The speed with which these extremely large molecules can be processed decreases the cost of each characterization from approximately $100,000 to approximately $1 ,000 per protein.

It is imperative that processes enable the characterization of proteins or other biological material without the artificially-imposed constraint of avoiding aqueous carriers for the biological material. Because the biological material is preferably carried in an aqueous salt solution, heretofore such polar solutions would have caused extreme and unmanageable losses in NMR characterizations.

EXPERIMENTAL RESULTS I. We calculated sensitivity of an optimized NMR and compared these calculations with experimental results.

1. The sample was contained in a 5 mm o.d., 4 mm i.d. capillary tube, having a length of 15 mm.

2. The solenoid used had a diameter d_c of 6 mm, and a length l_c of 15 mm. The coil volume V_c is 4.2x10^"7m³. 3. The quality factor of the coil Q was compared to the measured quality factor of the circuit, thereby yielding information about the effects of the various loss mechanisms set forth above. Q_L may be calculated as

QL = 4 (f ^1/2)a_cd_c f in [Hz] and d_c in [m] (1 )

a_c is a numerical factor depending on the coil dimensions:

a_c = 100 (lc/dc) / [100(lcJd_c) + 45] (2)

Eq. (1) predicts the quality factor of a well-constructed solenoid with 10- 20%. The self-inductance L of such coil can be approximated as:

L = 10^"6 a_cn_c ²d_c ² / l_c L in [H], d_c and l_c in [m] (2A)

where nc is the number of turns. It follows that the losses of the coil can be translated into a resistance r in series with the coil, where r is given by:

Q_L = 1.6 10-^b n_c 2^ i(ff 1 "/2 π_ = ω L / ) d_c / l_c (2B)

For example, take n_c = 3. Then for this coil with the dimensions given above it follows: a_c = 0.85, L = 18 nH. Then at 800 MHz Q = 577, r_L = 0.163 ohm.

4. The power, P, necessary to produce a 90° pulse in the middle of the coil with a length t₉₀. This power is given by

P = 6x10^"6 . a_c .Q- . (t₉₀)-² . H₀ . V_c . γ ¹ . (1 + d_c ² / l_c ²) [W] (3) Where H₀ is the external field in [Am-1],γ is the proton gyromagnetic ratio = 336.5 A^"1ms^"1 , and tgo is in [sec].

5. The SNR is calculated after (NT) scans of 90° pulses, assuming that the magnetization is in thermal equilibrium before each pulse, and that no line broadening is applied. A lengthy but reasonably straight-forward calculation gives as a result that for protons at room temperature the SNR of a specific spectral line is given as:

SNR = 9.26 x lO^"4 (H₀)^3/2 n_sM V_S{Q(NT) / [a_c V_c (1 +d_c ²/l_c ²) T_acF]}^1/2(LW)-¹ (4)

Where n_s is the amount of chemical equivalent protons per molecule, divided by the amount of lines produced by these protons (equal intensity of these lines is assumed). In our case, where we measure a doublet caused by two para protons, n_s = 1. Furthermore, M is the mol fraction in [Mole], T_ac is the acquisition time in [sec], F is the noise figure of the receiver, and (UN) is the full width of the NMR line in [HZ]. It is measured at 500 and 750 MHz that F = 1.2, and we shall use the same number for 800 MHz.

6. Dielectric losses in a solenoid of 20 nH at 800 MHz in water at 20 oC, doped with 100 mM NaCl can be calculated. If τ is 9.36 ps, ε_r = 80.2 = ε_r (0), ε_r (oo) = 5.6, and the dc conductivity σ= 1 Sm^"1 . It follows that the intrinsic imaginary permittivity in water is 3.5, and from

ε_r (ω) = ol (ω ε₀) (5)

that the imaginary permittivity due to the conductivity is 22.6. Hence at this frequency the conductive dielectric losses are dominating. II. Example 1.

We performed an experiment wherein the sample used was 1 mM phenylalinine, dissolved in water + 20 mM sodiumphosphate. Two samples were prepared, one without additional NaCl (the non-lossy sample) and one with 100 mM of NaCl added, thereby increasing the conductivity in the sample to about 1 Sm^"1 (the lossy sample). The samples were placed in a 4 mm i.d., 5 mm o.d. tube and surrounded by susceptibility plugs. The sample length was 15 mm, hence the sample volume, V_s, was 1.9 x 10^"7 m³.

1. Results at 500 MHz:

The signal-to-noise ratio was measured after 8 scans, using an acquisition time of 1.64 sec. The NMR linewidth was 3.75 Hz. H₀ = 9.3 x 10 ⁶ Am^"1. The experimental and theoretical results are summarized in Table 1.

TABLE 1

Experimental Theoretical

Q_c (non-lossy) 253

Q_c (lossy) 224

QL 456 t₉₀ (micros), non-lossy 5.8

P (Watt), non-lossy 50 8.1 (Q=253)

P (Watt), Q and t₉₀=5.8mi cros 4.5 tgo (micros), lossy 5.9

P (Watt), lossy 50 8.8 (Q=224)

SNR (for Q_c=253) 33.7 66.3

SNR (for Q_c=224) 31.3 62.4

SNR (for Q_c=456) 89.0

2. Results at 800 MHz:

The signal-to-noise ratio was measured after 8 scans, using an acquisition time of 1.024 sec. The NMR linewidth was 4 Hz. H₀ = 1.49 x 10 ⁷ Am^"1. The experimental and theoretical results are summarized in Table 2.

TABLE 2

Experimental Theoretical

Q_c (non-lossy) 197

Q_c (lossy) 166

QL 577 tgo (micros), non-lossy 5.0

P (Watt), non-lossy 36 22 (Q=197)

P (Watt), Q_L and t₉₀=5.0 micros 7.5 t₉o (micros), lossy 6.0

P (Watt), lossy 36 18 (0=166)

SNR (for Q_c=197) 100 141

SNR (for Q_c=166) 80 129

SNR (for Q_c=577) 240

The results from Table 1 illustrates that the probe efficiency of the experimental NMR coil and LC circuit is poor. For both the lossy and non-lossy sample the experimental SNR is a factor of 2 smaller than the theoretical, and the experimental power needed to obtain a particular pulse width is a factor of 6 larger. There is likewise a significant reduction in circuit quality. The reduction in the Q value from 456 to 253 using the non-lossy sample is primarily a function of capacitive losses, which are proportional to (f)^{3 2}. This is illustrated in that the quality factors at 500 MHz are actually larger than those at 800 MHz, whereas the Q at 500 MHz is a factor of 1.26 lower than at 800 MHz if the coil is determining the losses. The additional reduction in Q from 253 to 224 using the lossy sample reflects sample losses at this frequency and the effect becomes even more severe if the unloaded circuit Q can be enhanced. The probe efficiency at 800 MHz (Table 2) is substantially better than at

500 MHz. The experimental SNR is a factor of 1.4-1.6 times smaller than the theoretical, and the experimental power is a factor of 1.6-2.0 times larger than the theoretical power using the experimental circuit Q's. The NMR coil provides a factor of 3 reduction in SNR.

Because the sample losses dominate, the other factors that can improve NMR performance provide a significantly lower return on system performance. It is clear that sample losses may be reduced by: a. using coils with minimal E-fields in the sample; b. using thick-wall sample tubes or larger coil diameters to reduce the E- fields in the sample; c. minimizing the inductive losses in the sample (e.g. reduce sample diameter and maximize sample length, keeping total sample volume constant); d. using a porous medium (such as a sponge) in the sample tube so that eddy currents flow over a relatively smaller distance; e. using samples that are not conductive (i.e. distilled water; encapsulate proteins in reverse micelles).

III. Example 2

If one assumes that the determination of the structure of a protein dissolved in a lossy medium takes 60 days to characterize in the present 500 MHz environment, one can calculate the reduction in this time period resulting from the various improvements set forth above, both at 500 MHz and 800 MHz. We assume all experimental parameters are the same at both frequencies, therefore the only effects are improvements in the LC circuit. Table 3 illustrates the results of this analysis:

TABLE 3

Type of Improvement 500 MHz 800 MHz

SNR MT SNR MT incr. incr.

1. PRIOR ART 1.00 60 days 1.00 9.2 days

2. eliminate sample losses

(incl. Non-lossy solvents) 1.06 53.1 days 1.09 7.7 days

3. increase sample volume (x2) 1.41 26.5 days 1.41 3.9 days

3. eliminate LC circuit losses

(except coil losses) 1.34 14.8 days 1.71 1.3 days

4. improve circuit efficiency 1.97 3.8 days 1.41 15.9 hrs

5. reduce noise in receiver 1.1 3.1 days 1.1 13.1 hrs

6. use cryo-cooled or superconducting coil 3.0 8.4 hrs 3.0 88 min

7. use dual-sample probe 1.41 4.2 hrs 1.4 44 min

Having described a preferred and alternative embodiments of the invention, it should be understood that numerous modifications and adaptations may be made to the invention without departing from the spirit of the invention. Accordingly, the scope of the present invention should be considered limited solely by the scope of the claims appended hereto.

Claims

CLAIMSWe Claim:

1. A method of decreasing the dielectric losses and increasing the speed and sensitivity of nuclear magnetic resonance (NMR) spectroscopy measurements of biological molecules, comprising the steps of: a. preparing a non-lossy solvent; b. preparing the biological molecules in a biologically-conducive solution; c. adding the solution of step "b" to the solvent of step "a" to create a sample solution having substantially reduced dielectric losses; and d. determining the structure of the biological molecules by processing the product of step "c" in an NMR.

2. The method of Claim 1 , wherein the biological molecules comprise proteins.

3. The method of Claim 2, wherein the size of the protein is greater than about 40kD.

4. The method of Claim 1 , wherein the non-lossy solvent comprises a non-polar, non-conductive solution.

5. The method of Claim 1 , wherein the non-lossy solvent comprises a hydrocarbon.

6. The method of Claim 5, wherein in addition to a hydrocarbon, the non-lossy solvent comprises a supercritical fluid.

7. The method of Claim 5, wherein in addition to a hydrocarbon, the non-lossy solvent comprises a solution comprised of reverse micelles.

8. The method of Claim 1 , wherein the time required for determination of the structure of the protein is less than one hour.

9. The method of Claim 1 , wherein the biologically-conducive solution comprises water.

10. The method of Claim 9, wherein the biologically-conducive solution further comprises biologically active salts.