Identification of an apple juice compound directly in the 1H-spectrum of the mixture with hyphenated techniques and push button NMR: example fruit juice screening

Abstract

Nuclear Magnetic Resonance (NMR) has been used for structure analysis of pure compounds over decades. Opposite to this, the extraordinary potential of NMR has just started to be recognized and exploited in the analysis of mixtures in the context of biofluids and food beverages. In our contribution, we will show results of studies done in the field of beer and the already developed BRUKER-SGF-Profiling Method, the NMR Fruit-Juice Screener. We will show, that from the very same set of spectra several specific parameters can be derived. 1H NMR in combination with pattern recognition: 1. can distinguish between direct juices and juices from re-diluted concentrates, 2. can identify the geographical origin of the fruits used to produce the juices, 3. can even identify mixing of direct juices and concentrates, 4. can identify artificial addition of different sugars, glucose syrup and RTK. Two approaches for quantification of individual metabolites in juices and beer were successfully applied. The first is based on integration of individual lines in the individual spectra after identification which can be done by combining spectral match technology with pure compound spectra from a reference spectral data base. The second approach is Partial Least Squares Regression (PLS), a supervised prediction approach which uses calibration models obtained from a training data set. The latter approach proved most successful even for the prediction of parameters which are not directly related to individual metabolites like the gravity of beer from proton NMR spectra. LC-NMR cannot be used in a high throughput screening scenario, as it is not sensitive enough in the on-flow-mode, thus comparison between NMR results and LC-MS is important. Therefore combining flow-injection-NMR and LC-MS is a promising approach. The two most important analytical tools can be hyphenated and in its most stringent form also time synchronised. This allows to create NMR and LC-MS data for either individual statistical evaluations or for combined methods such as the NMR/LC-MS covariance. This for example allows combining molecular ion information directly with the corresponding NMR signals of the same compound and as such gives so far unreached structural information. In our contribution we will show the power of the hyphenated techniques, the application of the MetabolicProfiler™. As an example we identify the structure of the most differentiating metabolite in apple juice concentrates from China and Poland directly from the statistical evaluation of the mixture spectra and the SBASE database.