Wooden tips used in this study were common toothpicks purchased from the local supermarket in Guangzhou. The wooden toothpicks are made of natural wood without chemical modification on the surface during manufacturing, as shown in Figure 1A. Before use, wooden toothpicks were washed by methanol and water to clean the surface contaminants. Pure HPLC-grade methanol was purchased from Sigma (St. Louis, USA). All water used in this study is Milli-Q water.

Clinical thyroid tissue samples are collected from 23 patients who hospitalized at the Second Affiliated Hospital of Guangzhou University of Chinese Medicine (Guangzhou, China) and have signed informed consents to allow the collection of clinical samples (i.e., cancer thyroid tissue and normal thyroid tissue) for this study. The size of tissue samples at ca 1.0 × 1.0 mm. Routine analytical procedures for normal tissue (Figure 1B) and cancer tissue (Figure 1C) were performed by pathological diagnosis of paraffin section to diagnose thyroid cancer.

To perform wooden-tip MS analysis, the wooden tip was directly mounted on a commercial nanoESI device (Thermo Fisher Scientific, Bremen, Germany) followed previous studies. By placing nanoESI capillary, wooden tip was placed in the front the inlet of mass spectrometer at distances of 1.0 cm horizontal. Pure methanol (5.0 μL), which is commonly used in spray ionization in ambient MS plays a significant role in ambient ESI (McBride et al., 2019), was directly loaded onto wooden tip to extract analytes and generate spray ionization. By applying a high voltage (3.5 kV) onto wooden tip, spray ionization could be generated from wooden tip-end to acquire a characteristic mass spectrum. The capillary temperature was set at 200°C. The high voltage of wooden tip was supplied from the Orbitrap-QE mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). To perform MS/MS experiments, precursor ions were selected with an isolation window at 0.4 Da and a collision energy at 40%.

This study was approved by the Ethics Committee of the Second Affiliated Hospital of Guangzhou University of Chinese Medicine (approval number: ZE 2022-371).

All the mass spectral data acquisition and instrumental control were conducted by using Xcalibur 3.0 software (Thermo Fisher Scientific, Bremen, Germany). The acquisition speed was 5.4 scans/sec. Typically, signal duration from the first 1 min were averaged to obtain the high-resolution mass spectra. Principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were conducted on using Simca-p software (Umetrics, Sweden) as described previously (Li W. et al., 2019). Briefly, for MS data of each sample, the intensities of those monoisotopic peaks at the mass range from m/z 700 to 1,100 were normalized. The MS signal intensities higher than 1.0% were input to the software for the PCA and PLS-DA analysis. The variable importance in projection (VIP) value of interest ions that are larger than 1.0 were considered the special biomarkers in this work.

In order to improve identification of chemical compositions in thyroid tissues, the high-resolution mass spectra of extract of thyroid tissue were observed using WT-ESI-MS in this work. Figure 2A shows mass spectrum of normal thyroid tissue obtained by adding pure methanol. Many peaks were observed from m/z 700 to m/z 1,100 in the mass spectrum where the peak at m/z 782.5678, m/z 798.5416, and m/z 824.5574 are dominated. The base peak at m/z 798.5416 was also used as indicator to evaluate the reproductivity of direct tissue analysis, it was found that the coefficient of variation (CV) of six measurements of normal tissues is 16.3%, which is comparable to WT-ES-MS analysis of other raw complex samples under ambient conditions, showing acceptable reliability for ambient MS analysis (Hu and Yao, 2022). At this mass range, the mass spectrum indicates that there are abundant lipids extracted from of normal thyroid tissue. However, relative abundances of lipids peaks at m/z 897.7269, m/z 881.7533, m/z 923.7423 in the mass spectrum of thyroid cancer tissue were significantly raised, while the relative abundances of lipids such as the peak at m/z 782.5678, m/z 798.5416, and m/z 824.5574 were significantly reduced (Figure 2B). The change of these lipids is believed due to the cancerization of thyroid tissues, and therefore, characteristic spectra of lipid metabolites could be used for diagnosis of disease. It is reported that lipid profile is a type of biomarkers in thyroid cancer (Li D. et al., 2019; Jiang et al., 2021). The study of lipid metabolism of thyroid cancer represents a new field of research that has recently received increasing interest in cancer research. The correlation between lipid profile and cancerization can be changed with metabolic changes. It is reported that a high fatty acid turnover rate is required to offer the energetic and synthetic requirements for the growth of tumor tissue (Maan et al., 2018). In addition, previous studies also reported that lipid profile of patients, including lipid molecules are associated with various types of carcinoma (Tamura et al., 2019). Several metabolic pathways linking it to lipid metabolism are candidate biomarkers to discriminate between normal and cancer cells (Khatami et al., 2019). To the best of our knowledge, only a few studies have investigated the lipid profile of thyroid tissue as a diagnostic tool based ambient MS.

To further validate the typical lipid profile in a large population. The typical lipids at m/z 782.5690, m/z 798.5353, m/z 824.5544, and m/z 913.7456, which are dominated in normal and cancer tissue, respectively, were further investigated. Typical lipids were selected to conduct the MS/MS experiments, as shown in Figure 3. According to the high-resolution mass spectrum, the peak at m/z 798.5416 should be coincided with calculated mass of potassium ions of phosphatidylcholine (PC) [PC(34:1)+K]⁺ (calculated m/z 798.5415) This peak could be assigned to PC adduct within an excellent mass error as low as less than 1 ppm (0.1 ppm). Usually, the mass error less than 5 ppm is excellent for identification of metabolites. For example, upon the collision-induced dissociation (CID), the dissociation of the potassium adducts ([PC(34:1)+K]⁺) led to the detection of different fragment ions in comparison to that of the other adducts. In MS/MS spectrum, the most abundant product ion at m/z 739.4663 is corresponding to the loss of the choline moiety of the head-group (N(CH₃)₃) (calculated 59.0735 Da) from the precursor ions, as shown in Figure 3A. A minor peak at m/z 615.4786 relates to the loss of the intact zwitterionic PC head-group (HPO₄CH₂CH₂N(CH₃)₃) (calculated 183.0660 Da), and the ions at m/z 577.5187 indicates the further loss of potassium (calculated 38.9637 Da). A fragment ion was detected at m/z 162.9555, which is indicative of the potassium adduct of the phosphate moiety of the head-group ([HPO₄CH₂CH₂+K]⁺) (calculated m/z 162.9563).

Similarly, the peak at m/z 782.5678 should be coincided with sodium adduct of phosphatidylcholine (PC) [PC(34:1)+Na]⁺ (calculated m/z 782.5676). Again, there is an ultra-accuracy mass spectrum with a mass error at 0.25 ppm. Upon MS/MS experiment of [PC(34:1)+Na]⁺, as shown in Figure 3B, the peak at m/z 723.4949 and m/z 599.4998 are corresponding to the natural loss of N(CH₃)₃ and HPO₄CH₂CH₂N(CH₃)₃ from precursor ions, respectively. The peak at 577.5187 is attribute to the further loss of sodium form m/z 599.4998. The fragment at m/z 146.9814 is generated from the sodium adduct of the phosphate moiety of the head-group ([HPO₄CH₂CH₂+Na]⁺). Although there are different relative abundances of fragment ion form different adducts, the same fragments are very useful for identifying the lipid. Therefore, in the MS/MS spectra of PC ions, the fragment ions at m/z 146.9814 and m/z 162.9554 could indicate the sodium adduct and potassium adduct, respectively, while neutral loss of 59 Da indicates the N(CH₃)₃ of PC head group. Therefore, the MS/MS spectrum of the ions at m/z 824.5541 (Figure 3C) could be assigned to the adduct of [PC(36:3)+K]⁺. These identified lipids are good with the lipid database of LIPID MAPS^®. Owing to high-resolution mass spectrum obtained in this study, chemical compositions of the major peaks of lipids were observed in the averaged mass spectrum were proposed using the accurate mass. According to LIPID MAPS^®, these results of lipids such as phosphoinositol (PI) and PC are matched in Table 1 with low mass error, which are useful for biomarkers screening of thyroid tissue.

To validate mass spectral fingerprints and specificity of lipid profile, the multivariate variable analysis of MS data was performed to investigate thyroid tissue from different patients. Monitoring change of thyroid tissue would be useful for biomarker discovery. The PCA plot (Figure 4A) show that normal tissue (n = 23) and cancer tissue (n = 23) were well separated. In general, it was found that the clusters of normal and cancer tissues are centralized. Although the tissue samples from different individuals, these data indicates that characteristic ions from normal tissue and cancer tissue are significantly dominated due their special lipid metabolites.

To further seek the cancer-induced biomarkers, PLS-DA plots of normal and cancer tissue samples were generated (Figure 4B), showing that normal and cancer thyroid tissues samples are well-separated into two different groups. VIP scores derived from PLS-DA analysis are useful for multivariate analysis for identification of biomarkers. VIP scores >1.0 were considered significant potential candidates of biomarkers, while VIP scores larger than 1.2 are considered for depicting the most significant biomarkers contributing to the cluster group in the PLS-DA model analysis. Therefore, all the most significant biomarkers those VIP scores >1.2 in this work are summarized in Supplementary Table S1. These results showed that lipid profile and potential biomarkers of thyroid cancer can be successfully obtained by WT-ESI-MS, showing a potential tool for rapid diagnosis of thyroid cancer.