Ambient ionization mass spectrometric analysis of human surgical specimens to distinguish renal cell carcinoma from healthy renal tissue
Touch spray-mass spectrometry (TS-MS) is an ambient ionization technique (ionization of unprocessed samples in the open air) that may find intraoperative applications in quickly identifying the disease state of cancerous tissues and in defining surgical margins. In this study, TS-MS was performed on fresh kidney tissue (∼1–5 cm3), within 1 h of resection, from 21 human subjects afflicted by renal cell carcinoma (RCC). The preliminary diagnostic value of TS-MS data taken from freshly resected tissue was evaluated. Principal component analysis (PCA) of the negative ion mode (m/z 700–1000) data provided the separation between RCC (16 samples) and healthy renal tissue (13 samples). Linear discriminant analysis (LDA) on the PCA-compressed data estimated sensitivity (true positive rate) and specificity (true negative rate) of 98 and 95 %, respectively, based on histopathological evaluation. The results indicate that TS-MS might provide rapid diagnostic information in spite of the complexity of unprocessed kidney tissue and the presence of interferences such as urine and blood. Desorption electrospray ionization-MS imaging (DESI-MSI) in the negative ionization mode was performed on the tissue specimens after TS-MS analysis as a reference method. The DESI imaging experiments provided phospholipid profiles (m/z 700–1000) that also separated RCC and healthy tissue in the PCA space, with PCA-LDA sensitivity and specificity of 100 and 89 %, respectively. The TS and DESI loading plots indicated that different ions contributed most to the separation of RCC from healthy renal tissue (m/z 794 [PC 34:1 + Cl]− and 844 [PC 38:4 + Cl]− for TS vs. m/z 788 [PS 36:1 − H]− and 810 [PS 38:4 − H]− for DESI), while m/z 885 ([PI 38:4 − H]−) was important in both TS and DESI. The prospect, remaining hurdles, and future work required for translating TS-MS into a method of intraoperative tissue diagnosis are discussed.
KeywordsTouch spray ionizationDesorption electrospray ionizationMass spectrometryMultivariate statisticsSurgical tissue analysisCancer
Kidney cancers cause a significant fraction of cancer-related fatalities in the USA, with 14,080 deaths and 61,560 new cases estimated in 2015 [1, 2]. Approximately 90 % of kidney cancers arise from the renal parenchyma and are termed renal cell carcinomas (RCC) . RCC has been differentiated from normal tissue based on epigenetic [3, 4], genetic , proteomic [6–8], metabolomic [9, 10], and lipidomic markers [11, 12]. While these techniques are highly informative, analysis time is generally incompatible with intraoperative application, which is the long-term objective of this study. Ambient ionization mass spectrometry (MS), the MS analysis of samples in their native environment with little or no sample preparation, may improve surgical outcomes by providing surgeons with information that helps determine tissue disease state and identifies the tumor margin on the time scale of surgery. Previous intraoperative and/or ex vivo results from several investigators for brain cancer [13–16] and colon, lung, and liver cancers [17, 18] support this expectation. Moreover, it is known in several cancers that phospholipid profiles differ between tumor and normal tissues [19–21]. Lipids have been explored for recognizing cancerous tissue using a variety of ambient ionization techniques such as touch spray (TS) , probe electrospray ionization (PESI) , rapid evaporative ionization MS (REIMS) , and desorption electrospray ionization (DESI)-MS . Lipids are of interest because of their roles in cell signaling, cell-cell recognition, immune response, energy metabolism, and malignant transformation of cells [25–27] and are expected to be reflective of disease state.
TS-MS involves the physical sampling of tissue using a metallic probe with subsequent ionization occurring directly from the probe upon the addition of solvent (1–2 μL) and high voltage (4–5 kV) (see Electronic Supplementary Material Fig. S1) [22, 28]. Free fatty acids and glycerophospholipids, including phosphatidylcholines (PCs) and phosphatidylinositols (PIs), are typically detected when performing TS-MS on frozen tissue sections . TS-MS was previously used to analyze frozen sections of prostate cancer specimens from 18 human subjects, differentiating between normal prostate and prostate cancer . However, TS-MS has not been used to study fresh in vitro tissue in a way that emulates possible intraoperative use. Like TS-MS, PESI-MS is an ambient ionization technique that uses a conductive metallic probe to sample tissue by puncturing it and then creating ions directly from the probe by application of high voltage [23, 29]. Yoshimura et al. analyzed fresh human RCC specimens with PESI in the positive ion mode and observed differences from normal tissue in phosphatidylcholine and triacylglycerol profiles . PESI has also been used to analyze chemically induced hepatocellular carcinoma in a live mouse mounted on a sampling stage positioned in front of a mass spectrometer . PESI and TS are both invasive techniques, but only small amounts of material (10–100 μg) are required to generate data. Both these techniques balance speed of analysis, degree of invasiveness, and amount of sample preparation such that they are amenable to intraoperative applications. An alternative technique, REIMS, samples ions directly from the surgical smoke produced by electrocauterization [17, 18, 31]. REIMS is a fast method because the surgical removal of tissue and mass spectrometric sampling are performed nearly simultaneously; however, the data are harder to correlate with traditional histopathology, the gold standard for diagnosis.
DESI-MS is a microprobe experiment in which a charged aerosol is sprayed onto a sample (e.g., a tissue section or smear) with a spot size of around 200 μm. The resulting thin liquid film formed on the sample surface dissolves some compounds, and subsequent droplet impacts produce splashes of charged analyte containing microdroplets. Gas-phase ions are generated via electrospray-like processes when the secondary microdroplets are sampled by the mass spectrometer . DESI-MS imaging (DESI-MSI) experiments are performed by rastering the spot across the sample in a controlled fashion. An advantage of DESI-MSI when using an appropriate solvent is that the analyzed tissue can be stained with hematoxylin and eosin (H&E) subsequent to MS analysis and evaluated by histopathology, enabling direct correlations to be drawn between molecular and morphological information . DESI-MSI has been applied to analyze several malignancies [20, 21, 34, 35], including kidney cancers [12, 36], and demonstrated that the detected ions were reliably indicative of disease state (viz., distinction between tumor and normal tissues) via multivariate statistics. Finally, although not an ambient ionization technique, matrix-assisted laser desorption ionization (MALDI)-mass spectrometry has been performed in an imaging format, and it can also be used to profile tissue specimens, but the technique requires additional sample manipulation and it takes longer to analyze biopsy specimens [37–39].
In this study, TS-MS was used to analyze fresh healthy renal tissue and RCC specimens. As a reference method, DESI-MSI was performed following TS-MS analysis because DESI has been studied in more detail (DESI was first used to analyze tissue sections in 2005 , and TS was developed in 2014 ) and because measured DESI-MS lipid profiles can be precisely correlated with histopathological information by staining the analyzed tissue after DESI-MSI. Previous DESI-MSI studies of RCC in the negative ionization mode indicated the ability to differentiate RCC and healthy renal tissue with multivariate statistics, including the ability to differentiate clear cell renal cell carcinoma (ccRCC) and papillary RCC . Based on this previous work, we chose to perform the data acquisition in the negative ionization mode. This study focuses on the rapid analysis of fresh renal tissue using TS-MS with supporting DESI-MSI analysis.
Materials and methods
All tissue samples were acquired from patients undergoing nephrectomy (partial or radical) for the treatment of RCC and were handled in accordance with approved Institutional Review Board (IRB) protocols at Indiana University School of Medicine (1205008669) and Purdue University (1203011967). A small portion (approximately 3 × 3 × 3 mm) of tumor and adjacent normal kidney tissue was taken from the biopsy in the pathology lab (University Hospital, Indianapolis, IN) and analyzed on-site using TS-MS. TS-MS analysis was performed while the tissue was fresh (not frozen or fixed). Subsequent to TS-MS analysis, the biopsy specimens were snap frozen in liquid N2 and transported back to Purdue University on dry ice for DESI-MSI analysis. Tissue samples were obtained from 21 subjects over the period of 1 year. Of the 21 cancer subjects, there were 15 with clear cell RCC, 2 with papillary RCC, 3 with unclassified RCC, and 1 with oncocytoma (see Table S1).
TS-MS analysis of freshly resected kidney tissue
On each day of surgery, the pathology laboratory at University Hospital notified the Purdue researchers when tissue was acquired from the operating room. The researchers transported the tissue specimens from the pathology laboratory to the linear trap quadrupole (LTQ) mass spectrometer (Thermo Scientific, San Jose, CA) located in an adjacent research building. TS-MS was performed on the fresh (not frozen or fixed) tissue specimens using a metal teasing needle purchased from Fisher Scientific (Pittsburgh, PA) to scrape small amounts of cellular material from the bulk tissue . The probe was positioned in front of the mass spectrometer with the metal tip pointing at the inlet (∼5–8 mm), a high voltage (4 kV) was applied, and a 1-μL aliquot of methanol was added to initiate ionization. Data were collected over the range m/z 200–1000 in the negative ionization mode. A chronogram showing the time dependence of a typical TS-MS signal recorded from a single solvent addition is shown in Electronic Supplementary Material Fig. S2. After TS-MS analysis at Indiana University School of Medicine (IUSM), the tissue specimens were snap frozen in liquid nitrogen to preserve morphology and halt biochemical processes. The samples were then transported to Purdue for subsequent cryosectioning and DESI-MSI analysis. Parameters for MS/MS experiments which used an LTQ mass spectrometer and exact mass measurements using an orbitrap (Thermo Exactive, San Jose, CA), as well as additional methods on data handling, are provided in the Electronic Supplementary Material.
DESI-MS analysis of frozen tissue sections
At Purdue, the samples from each case were embedded in an optimal cutting temperature polymer (Sakura Finetek, Torrance, CA) and sectioned on a cryotome FSE to 10 μm thickness and thaw mounted on glass slides (Gold Seal UltraFrost Frosted Glass Slides, Thermo Scientific, San Jose, CA). Tumor and normal specimens (as determined by gross analysis performed in the pathology lab at IUSM) from each case were thaw mounted adjacent to each other on the same glass slide. A custom-built DESI source, similar to the DESI 2D source available from Prosolia Inc. (Indianapolis, IN), was used to image the tissue . DESI spray solvent of dimethylformamide (DMF)-acetonitrile (ACN) 1:1 was used to minimize alteration of tissue morphology . Additional DESI imaging and sample handling parameters are provided in the Electronic Supplementary Material. MS data were recorded over the range m/z 200–1000 in the negative ion mode, automatic gain control (AGC) was turned off, and data were acquired in the profile mode. The tissue sections were stained with H&E after DESI-MSI. A pathologist (L.C.) determined regions of healthy renal tissue and RCC. The DESI imaging data were compiled into a data cube using an in-house program. Data format was compatible with BioMap software (free to download at http://www.imzml.org/index.php?option=com_content&view=article&id=14&Itemid=32).BioMap and was used to extract mass spectra from regions of interest (ROIs) representing RCC and healthy renal tissue based on the histopathological examination (see Fig. S3).
Histological and pathological data
The pathology lab at IUSM performed histopathological analyses on the remaining tissue after providing the specimens to researchers as discussed. The pathological status of the adjacent tissue as well as the histological subtype, both determined by the pathology lab at IUSM, were used to correlate disease status with TS-MS lipid profiles. De-identified clinical and pathological data are provided in Electronic Supplementary Material Table S1. In addition, the tissue sections analyzed by DESI-MSI were subsequently stained with H&E and interpreted by a pathologist (L.C.) to determine regions of healthy tissue and RCC.
Statistical analyses of mass spectrometry data and data visualization
Principal component analysis (PCA), using the non-linear iterative partial least squares algorithm, on the TS-MS and DESI-MSI data was performed using MATLAB (MathWorks, Natick, MA) and in-house routines which were described in previous work . Data were normalized using standard normal variate transformations to correct for baseline shifts and global variation in signal intensities. While data were recorded over the range m/z 200–1000, only the range m/z 700–1000 was utilized for PCA and subsequent linear discriminant analysis (LDA), as this mass range provides better diagnostic information . LDA on the PCA-compressed data was performed in MATLAB to discriminate between healthy renal tissue and RCC . PCA-LDA was used to calculate sensitivity (true positive rate) and specificity (true negative rate) by means of cross validation (using n = 5 deletion groups, Table S2). This procedure is described in detail elsewhere . The dispersion in relative ion intensity for a few ions was plotted as box and whisker plots using OriginPro 2016 (OriginLab, Northampton, MA). Unpaired two-sample t tests of unequal variance and Kruskal-Wallis non-parametric tests were performed in OriginPro 2016 using the raw, unnormalized signal intensities (Table S3).
Results and discussion
Histopathological subtypes of the specimens were provided by the IUSM pathology lab, and additional PCA score and loading plot as well as average mass spectra are provided in Electronic Supplementary Material Fig. S8. The number of samples in each subtype was too small to comment on the ability of TS-MS to delineate RCC subtypes. Subtyping is a topic worth exploring further but is beyond the scope of this communication.
The reported data suggest that TS-MS has the potential to yield diagnostic chemical information from fresh healthy renal tissue and RCC tissue. Prior to undertaking a large-scale study, a TS-MS ionization source equipped with microcontrollers for precise probe manipulation and an automated solvent delivery system should be designed and optimized to increase data quality and reproducibility. While the TS-MS experiments were conducted on tissue after resection from a human subject, the analysis occurred on the intraoperative time scale. The results support larger-scale studies of ex vivo surgical tissue using TS-MS to generate a comprehensive database of RCC specimens.
Additional aspects worth investigating are tumor heterogeneity and tumor margins. Ideally, the short analysis time and minimal invasiveness of TS-MS suggest that it might be used to profile various regions within the tumor mass itself and in the surrounding parenchyma. It would be interesting to investigate how the TS-MS lipid profiles (e.g., molecular information) vary within the tumor, as it has been shown that the extent of tumor heterogeneity is underestimated because of limitations in the number of samples that can be obtained [45, 46]. Additionally, TS-MS may aid in delineation of the tumor-parenchyma margin (i.e., tumor margin) in partial nephrectomies, as the utility of intraoperative frozen section analysis remains controversial . This would be beneficial because negative parenchymal surgical margins of any size (e.g., less than 1 mm) for localized, low-grade (T1) RCC have been shown to result in virtually 0 % recurrence of the disease [48, 49].
This preliminary study demonstrated the feasibility of TS-MS analysis on a small number of fresh ex vivo surgical tissue specimens, as an intermediate step between frozen tissue analysis and intraoperative analysis. TS-MS differentiation of specimens using PCA and LDA was demonstrated, and the ions of greatest importance for separation were tentatively identified as complex membrane phospholipids m/z 794 (PC 16:0–18:1), 885 (PI 38:4), and 844 (PC 38:4). Differentiation of RCC and normal specimens was readily achieved by TS-MS and DESI-MS, although different ions were important for the separation between the two methods. Future work aims to expand the number of specimens in order to validate diagnostic capabilities, and further study the analytical figures of merit of the TS-MS methodology. The applicability of the technique is likely not exclusive to renal cell carcinomas, and it would be fruitful to investigate TS-MS in other cancers. In addition, refinements to the TS-MS methodology should be pursued to increase the ruggedness, ease of use, accuracy, and reproducibility of the measurements.
The research reported in this publication was supported by the National Institute of Biomedical Imaging and Bioengineering of the National Institutes of Health under award number R21EB015722. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. K. A. Kerian gratefully recognizes the funding support from the Purdue University Center for Cancer Research SIRG Graduate Research Assistantship Award that funded him for this study.
Compliance with ethical standards
The research involving human subjects was conducted in compliance with the ethical guidelines of the approved Institutional Review Board protocols at the Indiana University School of Medicine (study # 1205008669R004) and Purdue University (study # 1203011967). The renal cell carcinoma samples were obtained from 21 human subjects after they provided written informed consent to participate in the research study.
Conflict of interest
The authors declare that they have no conflict of interest.