**4. Summary and concluding remarks**

*Mass Spectrometry - Future Perceptions and Applications*

ing strategy) as previously commented [35].

non-targeted LC-HRMS workflow.

discrimination was achieved.

not provide fragmentation data of the detected ions, metabolomics data can solve authenticity problematics without the identification of any compound (fingerprint-

The first step of non-targeted LC-HRMS approaches data treatment is the conversion of raw data in a matrix built by retention time, *m/z* values and the area or signal of each peak detected. Sometimes, chemical interferences are removed from the matrix by fixing some parameters to be achieved such as mass tolerance for peak alignment, total intensity threshold, maximum peak shift, and S/N threshold. At this point, the generated matrix can be treated by univariate or multivariate data analysis. For instance, d'Urso et al. [32], who aimed to compare wild strawberry samples of different geographical origin (Sarno and Petina, Italy), growing conditions (spontaneous and cultivated populations), and germplasm (autochthonous and non-autochthonous), created a unique data matrix from raw data obtained in both positive and negative ionization mode in the performed LC-HRMS analysis following a data fusion procedure. PCA was then applied to the data matrix obtaining a scores plot that clearly discriminates between spontaneous and cultivated samples regardless the other variables. Moreover, a good classification was also observed for the five groups of samples studied, which were different combinations of the above geographical origin, growing conditions and germplasm mentioned variables. Anyways, when the objective of the study is the identification of molecular features that could behave as a biomarker in food integrity and authenticity, the matrix needs to be reduced. Thus, measures like the elimination of those molecular features that are not detected in a minimum percentage of the samples or of those that are not observed in the quality controls, which usually consists in a mix formed by a constant volume of all the analyzed samples, are normally implemented in the

As an example, Cavanna et al. [33], whose objective was the identification and selection of biomarkers responsible of the freshness of egg products, proposed a first reduction of data matrix by establishing some critical parameters values: (i) precursor ion deviation of 5 and 10 ppm for negative and positive runs, respectively, (ii) maximum peak shift of 0.3 min, (iii) a total intensity threshold of 1,000,000 AU, and (iv) a 30% of relative intensity tolerance used for isotope search. The authors removed the molecular features that showed a coefficient of variation bigger than 40% in the quality control sample, which was prepared by mixing 10 μL of each extract sample and was injected at the beginning of the sequence as well as every 10 samples analyzed. As a clear separation between fresh and non-fresh egg samples was observed when making a PCA study on positive and negative ionization modes with the reduced matrixes, the authors then applied supervised OPLS-DA. As can be seen in **Figure 5**, an expected increase in the

**18**

**Figure 5.**

*ESI + OPLS-DA scores plot of the fresh samples against the "1 day" samples. Left area dots (0 h), fresh samples;* 

*right area dots (1D), "1 day" samples. Reproduced from Ref. [33]. Open Access Journals.*

The role of LC-MS and LC-HRMS methodologies to address food integrity and authenticity have been presented and discussed by means of some selected applications published in the last years.

Most of the methods described in the literature opt for RPLC with mainly C18 columns, with gradient elution using acidified aqueous solutions and methanol or acetonitrile as mobile phase components, probably due to the strong capacity of this separation mode when dealing with low molecular weight chemicals with a relatively wide range of polarities. The use of other stationary phases such as C18 amide or perfluorinated columns are also proposed in some specific applications.

ESI continues to be the ionization source of choice when dealing with LC-MS and LC-HRMS analysis of food products, although in some cases other API sources are also employed. APPI has shown to provide similar or slightly better sensitivity for some specific applications, such as in the case of the determination of polyphenols, but it resulted in a very feasible option when addressing the characterization and classification of natural extracts due to the higher robustness of APPI source in the presence of matrix effect. Therefore, although it has not been widely exploited in food integrity and authenticity issues up to now, it is strongly recommended because of the sample matrix complexity of foodstuffs.

Regarding the mass analyzers, QqQ and IT instruments are the chosen ones when LRMS is employed, and TOF and Orbitrap analyzers for HRMS applications. However, the selection of LC-MS or LC-HRMS methods usually depends on the targeted or non-targeted approach. When targeted strategies are proposed, some specific biochemical food components are determined as food features to address food integrity and authenticity, requiring a quantitation step using standards for each targeted component. In those cases, LC-MS(/MS) methodologies, mainly using QqQ instruments, are very appropriate due to the low sensitivity attainable with these analyzers, and their good performance for quantitative analysis. Obviously, LC-HRMS methods providing higher resolution and accurate mass measurements are also a very good option for targeted food analysis, although it is more expensive and requires a more specialized staff. In order to achieve sample characterization and authentication, the comparison of the content and distribution of the targeted chemicals is sometimes enough, but the use of chemometric methods to try to find food feature similarities between the analyzed samples is highly recommended, especially when both the number of samples and the number of targeted bioactive substances increase.

In many applications, the quantitation of some chemicals may be a difficult task due to food matrix complexity, especially due to the possibility of unknown interfering compounds. In those cases, non-targeted approaches (based on metabolomic fingerprinting) using LC-HRMS have shown to be the best option to address food integrity and authenticity. As non-targeted analysis is performed, the high resolution and accurate mass measurements attainable with TOF and Orbitrap instruments are required. In non-targeted approaches, the measurement of peak intensity values as a function of *m/z* and retention times is frequently enough to

achieved food integrity and authenticity. Obviously, due to the huge amount of data obtained, especially when working in full-scan mode, the use of chemometrics is mandatory. Nevertheless, it has been reported that when dealing with metabolomic HRMS methodologies, the final annotated metabolites are strongly dependent on the global experimental approach employed (sample treatment, separation and detection, instrumentation employed, etc.). This is very important when searching for possible food biomarkers, as those will depend on the methodology used.

In conclusion, targeted and non-targeted LC-MS and LC-HRMS methodologies, especially in combination with multivariate chemometric methods, are powerful tools to address a hot topic nowadays such as food integrity and authenticity, and the number of publications in this field will continue to increase in the near future.
