Combination of Low-Temperature Electrosurgical Unit and Extractive Electrospray Ionization Mass Spectrometry for Molecular Profiling and Classification of Tissues

Real-time molecular navigation of tissue surgeries is an important goal at present. Combination of electrosurgical units and mass spectrometry (MS) to perform accurate molecular visualization of biological tissues has been pursued by many research groups. Determination of molecular tissue composition at a particular location by surgical smoke analysis is now of increasing interest for clinical use. However, molecular analysis of surgical smoke is commonly lacking molecular specificity and is associated with significant carbonization and chemical contamination, which are mainly related to the high temperature of smoke at which many molecules become unstable. Unlike traditional electrosurgical tools, low-temperature electrosurgical units allow tissue dissection without substantial heating. Here, we show that low-temperature electrosurgical units can be used for desorption of molecules from biological tissues without thermal degradation. The use of extractive electrospray ionization technique for the ionization of desorbed molecules allowed us to obtain mass spectra of healthy and pathological tissues with high degree of differentiation. Overall, the data indicate that the described approach has potential for intraoperative use.


Introduction
In computer-assisted surgery, the surgeon's manipulations are largely guided by a navigation system. There is great interest in using mass spectrometry (MS)-based navigation because MS allows molecular visualization of operated tissues.
Mass spectrometry molecular imaging by matrix-assisted laser desorption ionization (MALDI) is a useful method in tissue-based studies [1,2], but tissue sample preparation is required. Laser desorption methods [3], including picosecond infrared laser desorption method [4,5], do not need a preanalytical stage, which makes analysis less time-consuming. One of the most popular MS methods for tissue analysis is desorption electrospray ionization mass spectrometry (DESI-MS), in which tissue molecules are desorbed from their surface by electrosprayed solvent droplets (water, methanol, acetonitrile,

Materials and Methods
Methanol and water, both of HPLC grade, were purchased from Sigma-Aldrich (St. Louis, MO, USA). The LTEU-EESI-MS setup was based on the commercially available quantum molecular resonance (QMR) electrosurgery unit (Vesalius, Telea Inc., Vicenza, Italy) and the Maxis Impact (Bruker Daltonics, Bremen, Germany) mass spectrometer with a home-built EESI source (Figure 1). A low-frequency electrosurgical unit, Bovie A1250S-V (475 kHz, 80 W), (Apyx Medical, Antioch, TN, USA) was used for comparative analysis. Testo 875 thermal (testo, Lenzkirch, Germany) imager was used for the tissue's temperature measurement. The power supplied by the QMR generator to the scalpel was 40 W. The voltage on the electrospray source was set to 0 V, while the cap electrode of the mass spectrometer was set to −4000 V. The distance between the edge of the sprayer and the mass spectrometer inlet was about 0.5 cm ("a" in Figure 1A). The measurements were carried out in the range of m/z 100-1500 in positive-ion mode. The EESI solvent was methanol/water (90/10, v/v).

Results and Discussion
Chicken meat samples were used to study the properties and possibilities of the method under development and for the performance tuning of the experimental setup. Vesalius QMR surgical generator was used to provide desorption in the current study. This surgical instrument is considered as a low-temperature one. Indeed, the temperature in the electrosurgery unit probe region, measured by thermal imager, was about 61.4 °C on the chicken skin ( Figure 2A). The temperature of the same sample was about 180 °C when the low-frequency electrosurgery unit was used ( Figure 2B). The low temperature in the heating zone is a big advantage, both for the patient and for mass spectrometric analysis. Significant advantages include reducing tissue damage due to less heating, avoiding the formation of soot particles, preserving the native form of most biomolecules, and more precise positioning due to a reduction in the area from which ions are generated. However, because of low temperature, the efficiency of ion formation in the process of cutting or coagulation decreases [29]. Post-ionization techniques, which has already been shown as an efficient approach [26], can be implemented to enhance the ion yield in such cases. Chicken meat from a grocery store was used as sample for the test and method optimization. The breast cancer experiments were carried out on biopsy materials from patients treated at the National Medical Research Center for Obstetrics, Gynecology and Perinatology named after academician V.I. Kulakov of the Ministry of Healthcare of the Russian Federation (Moscow, Russia). All clinical investigations were conducted according to the principles expressed in the Declaration of Helsinki. All patients read and signed informed consents approved by the ethical committee of the National Medical Research Center for Obstetrics, Gynecology and Perinatology named after academician V.I. Kulakov. The samples were sliced for the histological study, and the rest were frozen in liquid nitrogen and stored under −75 • C until the investigation. Microscopic samples were investigated by optical microscopy using an Olympus MX51 light microscope (Olympus Corporation, Tokyo, Japan). A slice of a sample adjacent to the histologically investigated region was cut, thawed, and fixed on the holder in the ion source for the analysis. Biopsy samples of healthy tissues and malignant tumors separated by a histologist were taken from 50 patients for the development of a classification model. Five samples of tissues with tumor and bordered normal tissues characterized by the histologist were taken for testing the classification model and to calculate sensitivity and specificity of the model. Mass spectra were recorded from about 20 different points of each of the five samples. Thus, a set of 100 mass spectra obtained from 5 tissue samples with both normal and cancer regions were used to validate the classification model.
The resulting mass spectrometric data were transformed into a table of peak intensities using a set of functions developed in the R language, with columns corresponding to m/z values of the detected peaks and rows corresponding to individual samples. Subsequently, the tabulated data was aligned by the average value, scaled by the Pareto method, and analyzed using the orthogonal partial least squares discriminant analysis (OPLS-DA) method. The data on the contribution of each peak obtained from the OPLS-DA model was then used to calculate the "estimated parameter" of tissue pathology at a separately analyzed point.
The lipid nomenclature used throughout the paper is in accordance with LIPID MAPS [32] terminology and shorthand notation summarized in [33].

Results and Discussion
Chicken meat samples were used to study the properties and possibilities of the method under development and for the performance tuning of the experimental setup. Vesalius QMR surgical generator was used to provide desorption in the current study. This surgical instrument is considered as a low-temperature one. Indeed, the temperature in the electrosurgery unit probe region, measured by thermal imager, was about 61.4 • C on the chicken skin ( Figure 2A). The temperature of the same sample was about 180 • C when the low-frequency electrosurgery unit was used ( Figure 2B). The low temperature in the heating zone is a big advantage, both for the patient and for mass spectrometric analysis. Significant advantages include reducing tissue damage due to less heating, avoiding the formation of soot particles, preserving the native form of most biomolecules, and more precise positioning due to a reduction in the area from which ions are generated. However, because of low temperature, the efficiency of ion formation in the process of cutting or coagulation decreases [29]. Post-ionization techniques, which has already been shown as an efficient approach [26], can be implemented to enhance the ion yield in such cases. For the soft ionization of aerosol desorbed during LTEU desorption, we used EESI. In EESI, the neutral aerosol is crossed by electrospray plume to produce analyte ions in front of a mass spectrometer ( Figure 1A). This design allows the analysis of complex biological samples with minimal pretreatment [30,34].
The spectrum in Figure 3A shows the absence of an informative signal when an EESI solvent was not supplied, demonstrating that ions were not generated from a tissue in the LTEU desorption process. The peaks in this mass spectrum can be identified as common mass spectrometry contaminants, such as ammonium adducts of polysiloxanes (within 10 ppm accuracy) [35]. A good-quality spectrum of tissue was only achieved when the EESI solution was supplied ( Figure  3B). The most intensive peaks were grouped in two clusters. The first cluster was in the m/z range of 600-800 and identified as phospholipids according to accurate mass within 10 ppm (lysophosphatidylethanolamines (LPE), phosphatidylethanolamines (PE), phosphatidylcholines (PC)). The second cluster was in the m/z range of 800-1000 and identified as triacylglycerols according to accurate mass within 10 ppm (Table S1). The difference between the LTEU-EESI-MS spectrum and the mass spectrum obtained with high-temperature electrosurgery unit is evident from Figure 3B,C. Peaks in the higher mass range are less intensive in Figure 3C and more intensive in lower masses, which can be a consequence of thermal degradation. For the soft ionization of aerosol desorbed during LTEU desorption, we used EESI. In EESI, the neutral aerosol is crossed by electrospray plume to produce analyte ions in front of a mass spectrometer ( Figure 1A). This design allows the analysis of complex biological samples with minimal pretreatment [30,34].
The spectrum in Figure 3A shows the absence of an informative signal when an EESI solvent was not supplied, demonstrating that ions were not generated from a tissue in the LTEU desorption process. The peaks in this mass spectrum can be identified as common mass spectrometry contaminants, such as ammonium adducts of polysiloxanes (within 10 ppm accuracy) [35]. A good-quality spectrum of tissue was only achieved when the EESI solution was supplied ( Figure 3B). The most intensive peaks were grouped in two clusters. The first cluster was in the m/z range of 600-800 and identified as phospholipids according to accurate mass within 10 ppm (lysophosphatidylethanolamines (LPE), phosphatidylethanolamines (PE), phosphatidylcholines (PC)). The second cluster was in the m/z range of 800-1000 and identified as triacylglycerols according to accurate mass within 10 ppm (Table S1). The difference between the LTEU-EESI-MS spectrum and the mass spectrum obtained with high-temperature electrosurgery unit is evident from Figure 3B,C. Peaks in the higher mass range Signal from a set of lipids, including PC 34:2, PC 34:1, PC 34:0, TG 52:3, TG 54:4, and TG 57:5, was used for the LTEU-EESI-MS ion source optimization. Several important parameters, including the EESI solvent composition (the optimal ratio is methanol/water 9/1, v/v), power of the surgical probe (35-40W), and the proper geometry of the setup were determined for subsequent analyses during the experiments with chicken meat samples ( Figure S1).
Breast cancer samples were used for method validation after its optimization. The possibility to distinguish between normal and cancer tissues was tested.
The mass spectra in Figure 4 demonstrate differences not only in the intensity of peaks but also in the presence of some peaks unique for the pathology. Signal from a set of lipids, including PC 34:2, PC 34:1, PC 34:0, TG 52:3, TG 54:4, and TG 57:5, was used for the LTEU-EESI-MS ion source optimization. Several important parameters, including the EESI solvent composition (the optimal ratio is methanol/water 9/1, v/v), power of the surgical probe (35-40W), and the proper geometry of the setup were determined for subsequent analyses during the experiments with chicken meat samples ( Figure S1).
Breast cancer samples were used for method validation after its optimization. The possibility to distinguish between normal and cancer tissues was tested.
The mass spectra in Figure 4 demonstrate differences not only in the intensity of peaks but also in the presence of some peaks unique for the pathology.  The most profound difference between the mass spectra was observed in the m/z range of 600-1000. These peaks were mostly identified as glycero-and phospholipids according to accurate mass within 10 ppm. Supervised orthogonal projections to latent structures discriminant analysis (OPLS-DA) was applied to the mass spectrometric data on histologically validated samples to create a classification model. The value of −1 was set for normal tissue and 1 for cancer. The model confirmed significant distinction between normal and cancer LTEU-EESI-MS profiles. Data points corresponding to these two groups of tissues were clustered separately on the OPLS-DA score plot ( Figure 5). The developed statistical model showed good potential for classification between normal and cancer tissues scored with Q 2 = 0.65. The analysis of the variable influence on projection (VIP) of the OPLS-DA model was used to reveal features that were the most significant for the classification. A list of the lipids with VIP >1, a measure that can be used as cancer biomarkers, is summarized in Table S2. The most profound difference between the mass spectra was observed in the m/z range of 600-1000. These peaks were mostly identified as glycero-and phospholipids according to accurate mass within 10 ppm. Supervised orthogonal projections to latent structures discriminant analysis (OPLS-DA) was applied to the mass spectrometric data on histologically validated samples to create a classification model. The value of −1 was set for normal tissue and 1 for cancer. The model confirmed significant distinction between normal and cancer LTEU-EESI-MS profiles. Data points corresponding to these two groups of tissues were clustered separately on the OPLS-DA score plot ( Figure 5). The developed statistical model showed good potential for classification between normal and cancer tissues scored with Q 2 = 0.65. The analysis of the variable influence on projection (VIP) of the OPLS-DA model was used to reveal features that were the most significant for the classification. A list of the lipids with VIP >1, a measure that can be used as cancer biomarkers, is summarized in Table S2. corresponding to these two groups of tissues were clustered separately on the OPLS-DA score plot ( Figure 5). The developed statistical model showed good potential for classification between normal and cancer tissues scored with Q 2 = 0.65. The analysis of the variable influence on projection (VIP) of the OPLS-DA model was used to reveal features that were the most significant for the classification. A list of the lipids with VIP >1, a measure that can be used as cancer biomarkers, is summarized in Table S2.  The estimated scores for several analyzed tissue points are plotted in Figure 7. Data points with negative ordinate correspond to normal-tissue profile. Data points with positive ordinate correspond to cancer-tissue MS profile. The bigger absolute values of the score are related to more pronounced tissue type. The switch between negative and positive scores corresponds to a margin between normal and cancer regions. Results obtained by OPLS-DA classification of the LTEU-EESI-MS data gave 97% sensitivity and 94% specificity as related to histological analysis. The mass spectra were obtained with 1 mm spatial resolution. Therefore, the estimated accuracy of tissue margin determination was not worse than 1 mm in the present study. The estimated scores for several analyzed tissue points are plotted in Figure 7. Data points with negative ordinate correspond to normal-tissue profile. Data points with positive ordinate correspond to cancer-tissue MS profile. The bigger absolute values of the score are related to more pronounced tissue type. The switch between negative and positive scores corresponds to a margin between normal and cancer regions. Results obtained by OPLS-DA classification of the LTEU-EESI-MS data gave 97% sensitivity and 94% specificity as related to histological analysis. The mass spectra were obtained with 1 mm spatial resolution. Therefore, the estimated accuracy of tissue margin determination was not worse than 1 mm in the present study. correspond to cancer-tissue MS profile. The bigger absolute values of the score are related to more pronounced tissue type. The switch between negative and positive scores corresponds to a margin between normal and cancer regions. Results obtained by OPLS-DA classification of the LTEU-EESI-MS data gave 97% sensitivity and 94% specificity as related to histological analysis. The mass spectra were obtained with 1 mm spatial resolution. Therefore, the estimated accuracy of tissue margin determination was not worse than 1 mm in the present study.  The figures are colored according to histological results, with blue color corresponding to normal tissue and red color corresponding to cancer tissue. The red line on the graph was determined by statistical model and separates "normal region" from "cancer region".

Conclusions
The LTEU-EESI-MS method provided fast and reliable differentiation of breast cancer and normal tissue with 97% sensitivity, 94% specificity, and spatial accuracy of at least 1 mm. Such high accuracy and spatial resolution of analysis were mainly achieved owing to the low temperature of the surgical tool and thus the small area of tissue thermal destruction. The combination of low-temperature electrosurgical unit and extractive electrospray ionization mass spectrometry is a promising method for intraoperative tissue characterization because it provides high molecular specificity and high stability and robustness of operation.
Supplementary Materials: The following are available online, Figure S1: Optimization of the LTEU-EESI-MS ion source parameters, Table S1: Tentatively identified peaks based on accurate mass within 10 ppm, Table S2. Tentatively identified peaks with VIP ≥ 1. The identification is based on accurate mass within 10 ppm.

Conflicts of Interest:
The authors declare no conflict of interest.