*2.2. Gadolinium Provides Enhanced Sensitivity*

The addition of PRE agents has been previously used to accelerate NMR data acquisition [13,17,22–25]. PRE agents contain unpaired electrons and decrease T1 relaxation times for all nuclei in a sample due to dipolar interactions between nuclear and electron spin states. The PRE effect is very large, owing to the large magnetic moment of an unpaired electron, and can be tunable by adjusting the concentration of the PRE agent [17]. By combining NUS with the relaxation enhancing agent, Cu(EDTA), Rai and colleagues demonstrated a 22-fold reduction in the 2D 1H-13C HSQC data collection time to quantify a handful of urine metabolites [13]. Gadolinium-based contrast agents have been widely used in MRI diagnostic imaging, for studying soluble proteins, for characterizing protein-protein, protein-oligosaccharides, and protein-nucleic acid complexes, and for investigating membrane proteins using NMR spectroscopy [26,27]. Sakol et al. have also shown the utility of the Gd-based contrast agent, Gd-DOTA, for cellular localization studies using NMR spectroscopy [28]. Similarly, Mulder and colleagues utilized gadolinium-based PRE agents and achieved a 3- to 4-fold improvement in acquisition time for quantifying several plasma metabolites [17]. We sought to expand upon these findings by focusing on parameters to increase spectral sensitivity for a fixed-time experiment (1 h ± 4 min) instead of accelerating acquisition times.

We first assessed the 1D 1H spectral changes for Reference 1 (Supplementary Material Table S1) with a d1 of 0.8 s with increasing concentrations (0.25 mM to 1 mM) of Cu (EDTA) and Gadobutrol (Gd) (Figure S1), a gadolinium-containing macrocyclic that has previously been shown to enhance the relaxation rates of urine metabolites [17,25] (Figure S2). In general, contrast agents containing Gd shorten T1 and T2 relaxation rates through a dipole– dipole interaction between the unpaired electron of Gd and nuclei in the compound. The decrease in T1 and T2 rates depends on the contrast agent used and its concentration, the charge state of the compound, the viscosity of the solution, and the protein affinity of either the compound or contrast agent, among other issues. The typical range of T1 values for nuclei of common metabolites such as glucose, lactate, citrate, acetate, glutamine, and alanine are between 0.9 and 4 s. Similarly, T2 values range from 100 to 600 ms [17,28,29]. In the presence of Gd, T1 values can decrease from 2- to 10-fold depending on the concentration of Gd. A similar reduction is observed for T2, but is more pronounced at higher Gd concentrations. Accordingly, NMR resonances will significantly broaden into the baseline with the increase in Gd concentration [28]. Experimentally, we observed that a concentration of Gd at 0.25 mM allowed us to decrease our recycle delay to 0.8 s and achieve an overall increase in sensitivity while avoiding substantial line broadening. As the concentration of the Gd agent increased, the decrease in T2 and the associated peak broadening eventually eclipsed the reduction in T1 and negated any intensity gains from a larger number of scans [17,28,30,31]. In agreement with these observations, at 0.25 mM Gd, we noted an increase in intensities for the majority of metabolite resonances. As the concentration of Gd increased to 0.5 mM, a handful of metabolite resonances continued to show an increase in intensity, while others began to broaden. At 1 mM Gd, the majority of resonances were diminished compared to the control that lacked Gd. Interestingly, our results are in line with the theoretical optimal recycle delay predictions of Rovnyak et al. [32]. To perform the comparison, we identified NMR relaxation times reported in the literature for metabolites included in our study. For example, the work by Mulder et al. [17] demonstrated that the addition of Gd at a concentration of 0.5 mM to a mixture of small molecules (glucose, creatinine, citrate, glutamine, acetate, alanine, etc.) greatly reduced the T1 relaxation times by 2- to 10-fold, resulting in an average T1 relaxation time of ~0.6 s. Using the equation derived from Rovnyak et al., in the presence of Gd the theoretical optimal recycle delay would be ~0.8 s (1.26 × 0.6 s), which is in perfect agreement with our experimental findings of an optimal d1 of 0.8 s. In the presence of Cu (EDTA), we observed a decrease in NMR resonance intensities and significant line broadening at all concentrations tested. These results suggest that the addition of 0.25 mM of Gd may offer an optimal improvement in S/N. Indeed, when we recorded a 25% NUS 1H-13C HSQC spectrum with a d1 of 0.8 s in

the presence of 0.25 mM Gd, we observed an increase in both the mean intensity and mean S/N for Reference 1 (Figure 2). While the average fold-change increase in peak intensity due to the addition of Gd was modest (1.25-fold), we observed large fold-change increases (>2-fold) for the lowest intensity resonances (Figure 2c). Thus, the addition of Gd could improve the ability to detect low abundant metabolites. Of note, significant differences were observed in the intensity for individual metabolites, suggesting that Gd affects each metabolite to a different extent. Previous studies have suggested that a charge distribution, especially anionic metabolites, may be more affected by Gd [22,24]. We also verified that, for 0.25 mM Gd, the optimal d1 remained at 0.8 s as measured by both an increase in mean peak intensity and mean S/N (Figure S3). Taken together, our results suggest that the addition of Gd can improve both S/N and peak intensities, which will result in an overall sensitivity improvement, leading to a higher accuracy and precision in the measurement of metabolite concentrations.

**Figure 2.** Gadolinium improves S/N and mean intensity of metabolite resonances. 2D 1H-13C HSQC spectra (**a**) without and (**b**) with the addition of Gd. (**c**) Fold change of the median normalized peak intensity in the presence of Gd.
