Improving NMR sensitivity by use of salt-tolerant cryogenically cooled probes

The field of metabolomics relies heavily on information-rich techniques such as LC–MS and NMR. The advantages of NMR are that it is nondestructive, highly reproducible, and furnishes quantitative information. It does, however, usually suffer from poor resolution and sensitivity. Use of stronger magnetic fields can help solve both problems, especially resolution, but is usually cost prohibitive. Advances in NMR probe technology have improved the sensitivity measurement of the ethylbenzene standard by approximately 40% every three years. The most recent advance is the development of the cryogenically cooled probes [1, 2]. If the RF (radio frequency) coil and preamplifier are cooled to cryogenic temperatures the thermal noise of the system is reduced such that the measured signal-to-noise ratio increases approximately fourfold for samples dissolved in organic solvents. Chemists and biochemists have made extensive use of cryogenically cooled probes for several applications [3, 4]. As samples become more “lossy”, however, the increased sensitivity resulting from use of a cryogenically cooled probe is compromised. This can be best explained by examining the signal-to-noise-ratio equation [5]:

$$\frac{S}{N} \propto \frac{1}{{{\sqrt {R_{s} (T_{s} + T_{{pa}} ) + R_{c} (T_{c} + T_{{pa}} )} }}}$$

where the resistance of the coil is R _c, the temperature of the coil is T _c, the noise temperature of the preamplifier is T _pa, the temperature of the sample is T _s, and the resistance of the sample to RF is R _s. The preamplifier and the coil are at cryogenic temperatures, typically 10–25 K, which reduce the coil resistance, R _c, and increase the signal-to-noise ratio. The sample is usually at, or close to, room temperature. The remaining variable, R _s, the resistance of the sample to RF penetration, can easily become the dominant term if the sample is conductive. The lossiness of the sample is driven by both its ionic and dielectric conductivity. Thus, even polar organic solvents can reduce the signal-to-noise ratio. Figure 1 illustrates the effect on signal-to-noise ratio as the sample solvent becomes increasingly lossy. These data were acquired on a 600 MHz cryogenically cooled probe and are consistent with those presented by Horiuchi et al. [6]. Lossy samples are, essentially, regarded an additional source of noise. Such samples, for example urine, will cause increased 90° pulse lengths, because much of the pulse power is absorbed by creation of electric fields in the sample. The power dissipated by the sample depends not only on its electrical properties but also on its volume or geometry [7]. Use of a smaller diameter tube will reduce the overall sample resistivity to RF by the fourth power of the sample radius. In metabolomics, lossy or salty samples are common. Transferring the sample (without concentration) from a 5-mm tube to a 3 or 4-mm tube will not only improve the tuning and 90° pulse lengths of the cold probe but may also improve the signal-to-noise ratio for the sample in question. Greater improvements in sensitivity may be realized at higher fields, because the contribution of dielectric conductivity to the noise of the sample depends on the field [6].

Fig. 1

Plot of sample lossiness against expected decrease in signal-to-noise ratio compared with an empty probe. From this graph it is evident that both ionic and dielectric conductivity contribute to the noise created by the sample

“Going cold” may, however, have other effects. One question recently explored was how principle-component analysis (PCA) of a series of rat urine spectra collected by use of a cryogenically cooled probe equipped with a flow cell (the equivalent of a 3-mm NMR tube) compared with the same data collected by use of a room-temperature flow probe. The concern was that PCA can be sensitive to variances present in the data, for example baseline and phase distortions [8]. The PCA map of the first two principle components is shown in Fig. 2. For the fifteen rat-urine samples measured, the cold probe and room-temperature probe patterns were in agreement within experimental error. This confirms that databases collected with room-temperature probes need not be recollected using the cryogenically cooled probe. It is worth mentioning that for data collected using the cold probe the signal-to-noise ratio was usually two or more times greater than for data collected using the room-temperature probe. Solvent suppression for the two probes was either comparable, or slightly better for the cold probe.

Fig. 2

PCA map of data from fifteen rat-urine samples acquired automatically using either the room-temperature IFC probe (gray circles) or the cold probe in flow mode (black circles). Results show there is virtually no difference between overall spectral profiles

New cryogenically cooled probe technology has recently taken another step to improve sensitivity for lossy NMR samples [7]. As mentioned above, the sample itself is a significant source of noise. This is a result of RF power dissipated by the sample. Not all parts of the traditional cylindrical sample geometry contribute equally to the losses, however (Fig. 3). In fact, some regions of the sample are in “Electric-field hot spots” whereas other regions of the sample are in quite benign regions. For an optimized coil it has been shown that much of the noise generated by the sample is because of generation of electric fields parallel to the B₁ axis whose effects are most pronounced at the two parallel edges of the sample (Fig. 3). Changing the tube geometry to rectangular and properly orienting it in the probe substantially reduces the amount of sample in the “Electric-field hot spots” in which the sample would otherwise absorb much of the power and generate electric fields. This geometry also places the largest volume of the sample in the more benign regions of the RF field. The Salt-Tolerant cold probe (Varian) comprises an optimized coil circuit, a rectangular Salt-Tolerant tube (S-tube), and a sample holder that automatically orients the sample in the probe. This probe will accept both standard round NMR tubes and the rectangular Salt Tolerant tubes. Because the sample noise is reduced, the signal-to-noise ratio is expected to increase for lossy samples. NMR data collected from a rat urine sample in 3, 4, and 5-mm NMR tubes and in an S-tube are compared in Fig. 4. Comparison of the relative signal-to-noise ratios of the TSP peak normalized to the 5-mm tube reveals that placing the urine sample in smaller-diameter NMR tubes does not significantly reduce the signal-to-noise ratio. A 45% gain in signal-to-noise ratio is, however, observed when the S-tube is used, even though the sample volume of the S-tube is 77% that of the 5-mm NMR tube. Figure 4 also shows that the quality of solvent suppression and line shape is maintained despite the novel tube geometry. Shimming of the S-tube is, in fact, no different from that of a standard round tube after the high-order axial-shims (X, Y, ZX, ZY) have been set.

Fig. 3

Schematic diagram showing where the electric-fields are generated in a cross-section of a cylindrical NMR tube (a). Changing the geometry of the NMR tube to a squashed or S-tube (b) helps avoid the “hot-spot” regions of the RF field and places more sample in the lower electric-field-generating regions of the NMR tube

Fig. 4

Comparison of results obtained from tubes of different geometry in a Salt-Tolerant cold probe. The spectral region between −0.5 and 8.0 ppm is shown for rat urine diluted with phosphate buffer (∼66 mmol L⁻¹). Homogeneity was comparable for all sample tubes. A marked improvement in sensitivity was observed when the S-tube was used. The sample was transferred from one NMR tube to the next maintaining sample length. The signal-to-noise ratio was measured on the basis of the TSP (chemical shift reference peak at 0.00 ppm) and 200 Hz noise was used to calculate the r.m.s noise. Signal-to-noise ratios presented have been normalized relative to the 5-mm tube signal-to-noise measurement

In conclusion, sensitivity in NMR is constantly increasing. We have changed from room-temperature probes, to cryogenically cooled probes, and, currently, to Salt Tolerant Cold probes with novel sample geometry. The transition from room temperature-style probes to cold probes can be made without having to re-collect extensive databases. Use of a cold probe in metabonomics results in substantially improved sensitivity compared with a room-temperature probe while still enabling effective water suppression and homogeneity. An additional gain in sensitivity is observed when the Salt Tolerant cold probe equipped with an oriented S-tube is used. As the discipline of metabolomics evolves from binning approaches to more sophisticated and quantitative ways of interpreting the NMR data [9], high-quality and high-sensitivity NMR data is of paramount importance.