*2.3. Gadolinium Maintains Linearity*

Metabolomics requires quantification across a broad range of concentrations and the ability to accurately detect changes in metabolite levels [33,34]. We previously demonstrated that NUS 1H-13C HSQC metabolite profiling is highly linear in the 0.05 μM to 2 mM range [15]. Rai and colleagues also observed that the addition of a PRE agent, Cu(EDTA), maintained linearity for an NUS 1H-13C HSQC experiment that measured four amino acids (glycine, alanine, valine and methionine) over a concentration range of 24 to 78 mM [13]. We first sought to confirm that the addition of Gd maintained linearity over a broad concentration range. A series of six NUS 1H-13C HSQC spectra were recorded

for a mixture containing 29 metabolites (Reference 2) with concentrations ranging from 50 μM to 2 mM (Table S2). For each NMR resonance, the peak intensity was plotted as a function of concentration and the data were fit to a linear regression model (Figure S4). Example plots of the four NMR resonance peaks for leucine and the single resonance peak for pyruvic acid are shown in Figure 3. More than 98% of the metabolite resonances displayed a correlation coefficient of R2 > 0.9, indicating excellent linearity (Table 1). Interestingly, glucose resonances, which can be affected by isomers and conformational changes, had an R2 > 0.99 that was an improvement from our previous findings without Gd, where we observed an R<sup>2</sup> of ~0.8 [15]. Overall, this analysis demonstrated that peak intensities are highly linear as a function of metabolite concentration for NUS 1H-13C HSQC spectra in the presence of Gd.

**Figure 3.** Linear regression analysis of NMR resonance intensities for (**a**) leucine and (**b**) pyruvic acid peaks.

