Faster, More Reproducible DESI-MS for Biological Tissue Imaging
- 2k Downloads
A new, more robust sprayer for desorption electrospray ionization (DESI) mass spectrometry imaging is presented. The main source of variability in DESI is thought to be the uncontrolled variability of various geometric parameters of the sprayer, primarily the position of the solvent capillary, or more specifically, its positioning within the gas capillary or nozzle. If the solvent capillary is off-center, the sprayer becomes asymmetrical, making the geometry difficult to control and compromising reproducibility. If the stiffness, tip quality, and positioning of the capillary are improved, sprayer reproducibility can be improved by an order of magnitude. The quality of the improved sprayer and its potential for high spatial resolution imaging are demonstrated on human colorectal tissue samples by acquisition of images at pixel sizes of 100, 50, and 20 μm, which corresponds to a lateral resolution of 40–60 μm, similar to the best values published in the literature. The high sensitivity of the sprayer also allows combination with a fast scanning quadrupole time-of-flight mass spectrometer. This provides up to 30 times faster DESI acquisition, reducing the overall acquisition time for a 10 mm × 10 mm rat brain sample to approximately 1 h. Although some spectral information is lost with increasing analysis speed, the resulting data can still be used to classify tissue types on the basis of a previously constructed model. This is particularly interesting for clinical applications, where fast, reliable diagnosis is required.
KeywordsDESI-MS Mass spectrometry imaging Fast DESI Desorption electrospray ionization
Since its first publication in 2004 , desorption electrospray ionization (DESI) mass spectrometry (MS) has seen a large rise in popularity and extension into a wide field of applications . One such application is the MS imaging of biological tissue specimens, it being particularly well suited for the spatial mapping of lipids (fatty acids, phospholipids, and diglycerides and triglycerides) . By acquisition of the MS profile of different tissue types, these tissues can be characterized and, with use of either characteristic compounds or multivariate statistical tools, classified [4, 5, 6, 7, 8, 9]. This makes DESI a promising method for molecular pathology applications, whereby tissue types are characterized by their biochemical composition as opposed to morphological features as in classic histopathology. To ensure the accuracy of these methods, high quality and repeatability are desirable for DESI data. Two recent studies have looked at the repeatability and reproducibility of DESI-MS. In the first, a Versailles Project on Advanced Materials and Standards (VAMAS) interlaboratory study performed by the National Physical Laboratory , the absolute intensity repeatability was measured with glass slides coated with rhodamine B and with double-sided adhesive tape. The spectral shape, given by the relative intensity of peaks in selected regions, was compared with use of the tape. Absolute intensity repeatability varied largely between the participating laboratories from 14% to 140% (average 51%) for rhodamine B and from 14% to 58% for adhesive tape samples (average 31%). The spectral shape showed an average repeatability of 11%. Day-to-day reproducibility of the spectral shape was 31%. In the second study, by Abbassi-Ghadi et al. , the repeatability and reproducibility of DESI were assessed by the imaging of human esophageal cancer tissue sections. They were calculated from the variation in normalized intensities of 25 lipid species in the m/z range from 600 to 900. Repeatability was calculated as 22% and reproducibility was calculated as 20%.
In addition to geometrical issues, laboratory-built sprayer capillaries are usually cut by hand, which means that the edge can contain small imperfections that can have an impact on the quality of the electrospray. A stainless steel solvent capillary could also be used; however, this can lead to capacitance effects (charging and discharging to the gas nozzle), which can affect ionization. It can also lead to corona discharge and arcing between the sprayer and other metal components of the DESI source.
Here, we examine redesigning the DESI sprayer to provide better stability of solvent capillary positioning and thus better reproducibility. This includes the use of a TaperTip™ emitter. Tapered fused-silica emitters were used previously for DESI by Van Berkel et al.  in 2005 to improve the spatial resolution of DESI. However, without accurate positioning, the reproducibility of the primary electrospray is limited. In an initial experiment, the fused-silica solvent capillary of the laboratory-built sprayer was replaced to increase its wall thickness and, thereby, it stiffness. The use of machine-cut TaperTip™ emitters was also examined both for reproducibility and for improved spatial resolution. The results of these experiments led to a new sprayer design with a 20-μm inner diameter (ID) tapered fused-silica emitter to allow a stable primary electrospray at very low flow rates, and comprising a device to improve positioning accuracy of the emitter and reproducibility.
Solvents and Samples
Methanol and water were liquid chromatography–MS grade and were purchased from Sigma-Aldrich (St Louis, MO, USA). Rhodamine B coated slides were kindly provided by the National Physical Laboratory (Teddington, UK), and were prepared as in the recently published VAMAS interlaboratory study . Pork liver samples were obtained from a local supermarket. Rat cerebellum was from a control animal used in a study by collaborators at AstraZeneca. Details can be found in Swales et al. . The colorectal cancer samples were analyzed within the scope of a larger clinical study, which received full ethics approval (14/EE/0024). All tissue samples were fresh frozen on dry ice and stored at -80 °C until processing. For DESI-MS, samples were cryosectioned at 10-μm thickness and thaw-mounted onto SuperFrost Plus microscope slides (Thermo Fisher Scientific, Waltham, MA, USA). Pork liver was used for optimization of DESI parameters. Three adjacent colorectal cancer tissue sections were used to test different pixel sizes. Five adjacent rat cerebellum sections were used to assess high MS scan speeds and DESI imaging velocities. All sections were subsequently submitted to the Research Histology Facility at the Imperial College South Kensington Campus for hematoxylin and eosin (H&E) staining and subsequent histological examination.
The traditional laboratory-built sprayers were built from a 1/16-in. stainless steel tee (Swagelok, Solon, OH, USA) with a fused-silica gas capillary [363-μm outer diameter (OD), 220-μm ID], and a fused-silica solvent capillary (150-μm OD, 50-μm ID; both were obtained from SGE Analytical Science, Milton Keynes, UK), held in place with NanoTight™ sleeves (Idex, Lake Forest, IL, USA). To assess the impact of solvent capillary stiffness, altered sprayers with a fused-silica gas capillary of 700-μm OD and 530-μm ID and a fused-silica solvent capillary of 360-μm OD and 50-μm ID (both obtained from SGE Analytical Science, Milton Keynes, UK) were built. A further variant, where the solvent capillary was replaced with a TaperTip™ emitter (360-μm OD, 50-μm ID, from New Objective, Woburn, MA, USA) was also tested. A schematic for these sprayers can be found in Fig. S1.
All experiments were performed with a Xevo G2-XS QTof instrument (Waters, Milford, MA, USA) operated in sensitivity mode and equipped with a two-dimensional DESI stage from Prosolia (Indianapolis, IN, USA) and a custom-built inlet capillary heated to 490 °C. DESI parameters were optimized for the best signal intensity on tissue and were as follows: spray voltage, 4.5 kV; solvent, 95:5 methanol–water; flow rate, 1.5 or 0.5 μl/min for pixel sizes of 100 μm or 50 and 20 μm respectively; nebulizing gas, nitrogen; gas pressure, 7 and 4.5 bar for 1.5 and 0.5 μl/min flow rate respectively; sprayer incidence angle, 75°; collection angle, 10°; sprayer-to-inlet distance, 10 mm; sprayer-to-sample distance, 1.5 mm. MS parameters were as follows: scan time (unless specified otherwise), 1 s; source temperature, 120 °C; sampling cone voltage, -40 V; source offset, -80 V.
For the comparison of spectral pattern repeatability, five traditional laboratory-built sprayers, six altered laboratory-built sprayers (three with blunt capillaries, three with TaperTip™ capillaries), three commercially available (Prosolia) sprayers, and three novel sprayers were compared with use of adjacent sections from the same pork liver sample analyzed in randomized order in negative ion mode. Three sections were analyzed with each sprayer. Five lines were acquired across each section at a velocity of 100 μm/s and with a line-to-line spacing of 500 μm to avoid overlap. All tissue imaging data acquired with the novel sprayer were acquired with HDImaging version 1.3.5 in combination with MassLynx version 4.1 (Waters, Milford, MA, USA). The colorectal sample was acquired at a scan speed of one scan per second, and the stage speed was adapted accordingly to achieve pixel sizes of 100, 50, and 20 μm. For the rat cerebellum samples, the scan speed was increased from one scan per second to 5, 10, 20, and 30 scans per second. The stage scan speed was adapted accordingly to maintain a constant pixel size of 50 μm.
Line scan raw files were converted into mzML files with the ProteoWizard msConvertGUI (Vanderbilt University, Nashville, TN, USA), and then into an imzML image file with imzML Converter version 1.3 . The imzML files were imported into a MATLAB environment for subsequent analysis. The spectra for each line of a tissue section were averaged. Spectral pattern reproducibility was calculated according to the method of Dill et al. . Briefly, the overall mean spectrum is calculated and all spectra are converted into n-dimensional vectors (with n being the number of spectral features) and normalized to a unit length. Each spectrum is then multiplied by the mean spectrum to obtain a similarity score. The relative standard deviation (RSD) of this score corresponds to the spectral pattern reproducibility.
Imaging data were analyzed directly in HDImaging or concatenated raw files were centroided and lock mass corrected with MassLynx version 4.1. They were imported into MATLAB, co-registered with the optical image of the H&E-stained section , and annotated by a histopathologist. For principal component analysis, the data were normalized by median fold change normalization and log transformed for variance stabilization (for details see ). We performed k-means clustering on nonnormalized, nontransformed data in two steps: first, 2-means clustering was used to separate the tissue from the background. Then k-means clustering was performed again on the tissue area. Clustering was performed with a squared Euclidean distance measure. The number of spectral features for each image was measured by our picking 25 pixels each for white matter and gray matter in roughly the same region in all images and extracting the spectral information for those pixels from the unprocessed raw data.
The tissue prediction potential of the high scan rate data was evaluated. With use of the histopathologist’s annotations for the one scan per second tissue section, annotated pixels were normalized to the total ion count (TIC) and used as the training set in a leave-one-class-out recursive maximum margin criterion (MMC) discriminatory model . For all but the one scan per second tissue section, peak alignment was performed to match spectral variables across the five tissue sections. The m/z vector from each section was mapped to the m/z vector of the one scan per second section with a maximum permitted m/z shift of 50 ppm. All pixels were TIC normalized and predicted against the MMC linear discriminant analysis model.
For each of the four annotated classes (white matter, gray matter 1, gray matter 2, background) a classification probability between 0 and 1 was derived for each pixel. Pixels with a single high probability (e.g., greater than 0.8) can be confidently assigned, those with multiple high probabilities can be ambiguously classified (e.g., at a tissue intersection), and those with multiple low probabilities cannot be classified.
Results and Discussion
Spectral pattern repeatability (relative standard deviations) for pork liver in negative ion mode for different sprayer designs
Spectral pattern repeatability (%)
Conventional laboratory-built sprayer (n = 5)
0.39–1.81 (average 0.82)
Prosolia sprayer (n = 3)
0.49–2.43 (average 1.28)
Altered laboratory-built sprayer (blunt capillary) (n = 3)
0.02–0.08 (average 0.05)
Altered laboratory-built sprayer (TaperTip™ emitter) (n = 3)
0.02–0.16 (average 0.06)
Novel sprayers (n =3)
0.05–0.25 (average 0.14)
The results from these experiments were then used to design a new sprayer. The fused-silica gas capillary was replaced with a metal cone for robustness. A 20-μm-ID (360-μm-OD) TaperTip™ emitter was used, as this provides good stiffness, a machine-cut tip, and flow stability at low flow rates, which is required to improve spatial resolution. The aperture in the gas cone was set to 400 μm, as a small aperture was found to improve the focusing of the primary electrospray (and thus, potentially, spatial resolution). A capillary positioning device, composed of a stainless steel emitter guide and a centering disc with concentric holes for the gas flow, was inserted (Figure 2) to fix the position of the emitter and to facilitate the insertion of the TaperTip™ through the gas cone aperture. The disc was coated with an insulating material to avoid electrical discharge and charging of the cone. Three of these novel sprayers were then also compared for spectral pattern repeatability. All three sprayers performed excellently, with a spectral pattern reproducibility of 0.25% or below per sprayer and 0.40% between them. Overall, the data demonstrate that fixing the emitter position using a stiffer capillary and an emitter guide leads to a more reliable geometry and thus increased spectral reproducibility.
Very similar results were obtained with 2-means clustering of the tissue area at scan speeds of 1, 5, and 10 scans per second, with a clear separation of white and gray matter. At higher scan speeds, this separation could be achieved only through introduction of a third cluster. This suggests that some spectral information is lost with an increase in scan speed. However, a separation of tissue areas is still easily achievable. Examination of the raw data shows that the overall number of spectral features decreases as the scan speed increases, although in some regions it then starts rising again (see Fig. S3). This rise is likely to be an artifact due to a decrease in signal intensity, which eventually leads to peak shape deterioration and peak splitting (see Fig. S4).
The new sprayer presented here provides a viable alternative to the traditional laboratory-built design. It delivers high-quality lipid spectra and good stability over time with improved resolution. The use of a thick-walled, machined-tapered fused-silica emitter with an emitter guide to fix its position eliminates some of the major factors thought to contribute to DESI variability—namely, the surface qualities and the positioning of the emitter within the gas capillary or nozzle. The present sprayer provides high reproducibility, sensitivity, and spatial resolution, making possible fast and accurate DESI analysis. Acquisition of an image at a nominal pixel size of 20 μm allowed us to clearly resolve features as small as 50 μm across. Although some spectral information is lost at very high scan speed, different tissue areas can still be separated. Use of a high-quality training set allows rough tissue type prediction even in samples acquired at very high speed. This sprayer is therefore very promising for tissue analysis in a clinical context where a fast classification is needed. In practice, a database could be built from a reference sample set with classic scan parameters. Test samples could then be analyzed at high speed and classified with the database, allowing high-throughput diagnostics, which can be crucial in a surgical setting. In-depth analysis could then also be performed with traditional scan speeds.
The authors thank James Alexander and Anna Mroz of Imperial College for the provision and sectioning of the colorectal samples. The novel DESI sprayer was developed in cooperation with and manufactured by Waters Corporation (Wilmslow, UK). This research was funded by Waters Corporation and supported by National Physical Laboratory Strategic Research Programme 116301 through the Stratigrad program, the European Union Seventh Framework Programme (under grant agreement no. 297499/305940), and the European Research Council Consolidator Grant (grant agreement no. 617896).
- 4.Veselkov, K.A., Mirnezami, R., Strittmatter, N., Goldin, R.D., Kinross, J., Speller, A.V.M., Abramov, T., Jones, E.A., Darzi, A., Holmes, E., Nicholson, J.K., Takats, Z.: Chemo-informatic strategy for imaging mass spectrometry-based hyperspectral profiling of lipid signatures in colorectal cancer. Proc. Natl. Acad. Sci. U. S. A. 111(3), 1216–1221 (2014)CrossRefGoogle Scholar
- 6.Eberlin, L.S., Tibshirani, R.J., Zhang, J., Longacre, T.A., Berry, G.J., Bingham, D.B., Norton, J.A., Zare, R.N., Poultsides, G.A.: Molecular assessment of surgical-resection margins of gastric cancer by mass-spectrometric imaging. Proc. Natl. Acad. Sci. U. S. A. 111(7), 2436–41 (2014)CrossRefGoogle Scholar
- 7.Guenther, S., Muirhead, L.J., Speller, A.V., Golf, O., Strittmatter, N., Ramakrishnan, R., Goldin, R.D., Jones, E., Veselkov, K., Nicholson, J., Darzi, A., Takats, Z.: Spatially resolved metabolic phenotyping of breast cancer by desorption electrospray ionization mass spectrometry. Cancer Res. 75(9), 1828–37 (2015)CrossRefGoogle Scholar
- 9.Wiseman, J.M., Puolitaival, S.M., Takáts, Z., Cooks, R.G., Caprioli, R.M.: Mass Spectrometric profiling of intact biological tissue by using desorption electrospray ionization. Angew. Chem. 117(43), 7256–7259 (2005)Google Scholar
- 11.Abbassi-Ghadi, N., Jones, E.A., Veselkov, K.A., Huang, J.Z., Kumar, S., Strittmatter, N., Golf, O., Kudo, H., Goldin, R.D., Hanna, G.B., Takats, Z.: Repeatability and reproducibility of desorption electrospray ionization-mass spectrometry (DESI-MS) for the imaging analysis of human cancer tissue: a gateway for clinical applications. Anal. Methods. 7(1), 71–80 (2015)CrossRefGoogle Scholar
- 12.Tillner, J., McKenzie, J.S., Jones, E.A., Speller, A.V., Walsh, J.L., Veselkov, K.A., Bunch, J., Takats, Z., Gilmore, I.S.: Investigation of the impact of desorption electrospray ionization sprayer geometry on its performance in imaging of biological tissue. Anal. Chem. 88(9), 4808–4816 (2016)CrossRefGoogle Scholar
- 24.Swales, J.G., Tucker, J.W., Strittmatter, N., Nilsson, A., Cobice, D., Clench, M.R., Mackay, C.L., Andren, P.E., Takats, Z., Webborn, P.J., Goodwin, R.J.: Mass spectrometry imaging of cassette-dosed drugs for higher throughput pharmacokinetic and biodistribution analysis. Anal. Chem. 86(16), 8473–80 (2014)CrossRefGoogle Scholar
- 27.Veselkov, K.A., Vingara, L.K., Masson, P., Robinette, S.L., Want, E., Li, J.V., Barton, R.H., Boursier-Neyret, C., Walther, B., Ebbels, T.M., Pelczer, I., Holmes, E., Lindon, J.C., Nicholson, J.K.: Optimized preprocessing of ultra-performance liquid chromatography/mass spectrometry urinary metabolic profiles for improved information recovery. Anal. Chem. 83(15), 5864–72 (2011)CrossRefGoogle Scholar
- 28.Li, H.F., Jiang, T., Zhang, K.H.: Efficient and robust feature extraction by maximum margin criterion. Adv. Neural. Inf. Process. Syst. 16, 97–104 (2004)Google Scholar
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.