Automated Platform for High-Resolution Tissue Imaging Using

Sep 7, 2012 - Chemical and Materials Sciences Division, Pacific Northwest National Laboratory, Richland, Washington 99352. ‡. Environmental and ...
0 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/ac

Automated Platform for High-Resolution Tissue Imaging Using Nanospray Desorption Electrospray Ionization Mass Spectrometry Ingela Lanekoff,† Brandi S. Heath,† Andrey Liyu,‡ Mathew Thomas,§ James P. Carson,§ and Julia Laskin*,† †

Chemical and Materials Sciences Division, Pacific Northwest National Laboratory, Richland, Washington 99352 Environmental and Molecular Sciences Laboratory, Pacific Northwest National Laboratory, Richland, Washington 99352 § Biological Sciences Division, Pacific Northwest National Laboratory, Richland, Washington 99352, United States ‡

S Supporting Information *

ABSTRACT: An automated platform has been developed for acquisition and visualization of mass spectrometry imaging (MSI) data using nanospray desorption electrospray ionization (nano-DESI). The new system enables robust operation of the nano-DESI imaging source over many hours by precisely controlling the distance between the sample and the nano-DESI probe. This is achieved by mounting the sample holder onto an automated XYZ stage, defining the tilt of the sample plane, and recalculating the vertical position of the stage at each point. This approach is useful for imaging of relatively flat samples such as thin tissue sections. Custom software called MSI QuickView was developed for visualization of large data sets generated in imaging experiments. MSI QuickView enables fast visualization of the imaging data during data acquisition and detailed processing after the entire image is acquired. The performance of the system is demonstrated by imaging rat brain tissue sections. Low background noise enables simultaneous detection of lipids and metabolites in the tissue section. High-resolution mass analysis combined with tandem mass spectometry (MS/MS) experiments enabled identification of the observed species. In addition, the high dynamic range (>2000) of the technique allowed us to generate ion images of low-abundance isobaric lipids. A high-spatial resolution image was acquired over a small region of the tissue section revealing the distribution of an abundant brain metabolite, creatine, on the boundary between the white and gray matter. The observed distribution is consistent with the literature data obtained using magnetic resonance spectroscopy.

T

fully hydrated tissue samples.19−25 In addition, laser ablation coupled to an atmospheric pressure flowing afterglow ionization26 and infrared laser ablation metastable-induced chemical ionization27 have been used for tissue imaging. Despite significant advances in ambient ionization MSI, considerable efforts are directed toward improving the spatial resolution and the collection efficiency of the analyte molecules. Nanospray desorption electrospray ionization (nano-DESI) is a new ambient technique28 that enables sensitive imaging of fully hydrated biological materials with high spatial resolution.29 Nano-DESI utilizes localized desorption of analyte molecules from surfaces into a liquid bridge between two fused silica capillaries, followed by nanospray ionization of the molecules.28,30 In our previous study, we reported first proof-ofprinciple experiments demonstrating the potential of nanoDESI for high-spatial resolution imaging of tissue samples.29 Precise control of the distance between the nano-DESI probe and the sample was identified as a key parameter controlling the stability of the ion signal and the spatial resolution of the technique. Because high-resolution imaging MS experiments

issue imaging using mass spectrometry (MS) plays an important role in clinical research and drug discovery.1−10 A vast majority of imaging mass spectrometry studies utilize matrix assisted laser desorption ionization (MALDI)11 for ionizing proteins, lipids, and metabolites directly from tissues.1,2,7,12 MALDI is a soft ionization technique that generates intact ions of molecules present in the tissue sample, which is critically important for their identification. Recent improvements in the sensitivity and spatial resolution of MALDI imaging enabled molecular histology studies providing unique molecular signatures of diseased tissues that cannot be obtained using traditional approaches.2,10,13 In addition, detailed spatial mapping of different classes of molecules in healthy tissue sections is essential for understanding biochemical processes in living systems. More recently threedimensional MALDI imaging has been utilized for obtaining ion images of whole organs for comparison with magnetic resonance imaging (MRI) data.14 Ambient ionization mass spectrometry imaging (MSI) is undergoing rapid development because it enables chemical analysis of biological samples in their native environment without special sample pretreatment.15−18 For example, desorption electrospray ionization (DESI) and laser ablation electrospray ionization mass spectrometry (LAESI) have been used for 2D and 3D mapping of lipids and drug metabolites in © 2012 American Chemical Society

Received: July 9, 2012 Accepted: September 7, 2012 Published: September 7, 2012 8351

dx.doi.org/10.1021/ac301909a | Anal. Chem. 2012, 84, 8351−8356

Analytical Chemistry

Article

field gradient necessary for nanospray ionization was created by applying +3 kV potential to the stainless steel union housing the primary capillary. The high voltage presents an electrical shock hazard. The heated capillary inlet was held at 30 V and 250 °C. The region between the capillaries and the sample was monitored using two Dino-Lite digital microscopes (8, SunriseDino, Flushing, NY). Positive-mode high-resolution mass spectra (m/Δm = 60 000 at m/z 412) were acquired in the Orbitrap. Imaging experiments were performed by scanning the sample under the nano-DESI probe at a constant velocity while acquiring mass spectra in a line. The scan rate was 10 μm/s and 20 μm/s for the high- and low-resolution imaging, respectively. After the completion of the line acquisition, the sample was positioned at the same initial x-coordinate and a new y-coordinate. The step between the lines was selected based on the desired spatial resolution.

generate large data files, development of new and improved data visualization and analysis tools is also critically important.4 In this study, we present the design and performance of an automated platform for nano-DESI imaging of biological tissue samples combined with novel software that enables on-the-fly visualization of the imaging data.

■

EXPERIMENTAL SECTION The rat brain tissue sample was provided by Drs. Chuck Timchalk and Jordan Smith (PNNL). All animal procedures were conducted in accordance with the guidelines for the care and use of laboratory animals in the NIH/NRC Guide and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee (IACUC) of Battelle, Pacific Northwest Division. Male Sprague−Dawley rats were humanely euthanized using CO2 as an asphyxiant and brains rapidly removed. The frozen midbrain was sectioned to 10 μm on a cryo-microtome (model CM 1900, Leica Microsystems, Buffalo Grove, IL) thaw-mounted on a glass microscope slide and stored at −80 °C. The samples were brought to room temperature but not dried prior to analysis. Imaging mass spectrometry experiments were performed using an LTQ/ Orbitrap mass spectrometer (Thermo Electron, Bremen, Germany) equipped with a custom-designed nano-DESI source shown in Figure 1. Tissue samples attached to glass slides (1)

■

RESULTS AND DISCUSSION Automated Control of the Sample Position. A previous study by van Berkel and co-workers31 utilized an image analysis approach for controlling the distance between the surface sampling probe32 and the sample. In that study, the distance between the probe and the surface was controlled using a closed circuit camera combined with image analysis software.31 The working distance was around 500 μm. In contrast, in this study the distance between the nano-DESI probe and the sample was around 10 μm. Maintaining such a small distance between the sample and the probe requires a different approach. We note that tissue sections examined by mass spectrometry are typically only 5−20 μm in thickness and are expected to be reasonably flat. To confirm this assertion, we examined the roughness of the tissue sections using contact profilometry. Figure 2 shows a typical result of height profiling

Figure 1. Photographs of the nano-DESI imaging setup: (A) top view and (B) side view. (1) sample holder with a glass slide holding four serial tissue sections, (2) motorized XYZ stage, (3) mass spectrometer inlet, (4) primary capillary, (5) micromanipulator for positioning the primary capillary, (6) nanospray capillary, (7) miniature XYZ stage for positioning the nanospray capillary, (8) Dino-lite cameras.

were positioned using a high-resolution XYZ stage (2) composed of three MFA series miniature linear stages (Newport, Corporation, Irvine, CA). The stage control was accomplished using a custom-designed Labview interface. During imaging experiments, the program enables control of the motorized stage in the XY plane by scanning along one axis and stepping along the second axis. The z-coordinate of the stage determines the distance between the tissue sample and the nano-DESI probe. An approach for controlling of the zcoordinate of the stage developed in this study will be described in detail in the next section. The nano-DESI probe was comprised of two fused silica capillaries (50 μm i.d., 150 μm o.d., Polymicro Technologies, L.L.C., Phoenix) and positioned in front of the mass spectrometer inlet (3). The primary capillary (4) was positioned using a high-resolution micromanipulator (5, XYZ500TIM, Quater Research and Development, Bend, OR). The 20 mm long nanospray capillary (6) was mounted in a 1/16 in. o.d. capillary PEEK tubing (Upchurch Scientific, Oak Harbor) and positioned using a T12 miniature XYZ stage (7, Thorlabs, Newton, NJ). A 9:1 (v:v) methanol−water solution was infused using a syringe pump at 0.5 μL/min. The

Figure 2. Variation in the height of the tissue sample measured using contact profilometer.

of a tissue sample mounted on a glass slide using a contact profilometer (Dektak 150, Veeco, Plainview, NY). In this experiment, the profilometer was allowed to scan across the sample starting from a point located on the sample using a stylus force of 3 mg. An average variation in height of 1.5 ± 0.6 μm observed in multiple experiments indicates that the tissue sample is very flat. On the basis of these results, we decided to use an approach described in the following paragraph for controlling the distance between the sample and the nanoDESI probe to within better than 10 μm. 8352

dx.doi.org/10.1021/ac301909a | Anal. Chem. 2012, 84, 8351−8356

Analytical Chemistry

Article

In this approach, the stage position along the z-axis is automatically varied to compensate for the tilt of the sample. First, the user specifies (x,y,z) coordinates of three points on the substrate thereby defining the plane in which the glass slide holding the sample resides. Next, the initial z-coordinate of the sample used for imaging, z0, is defined in one point to account for the sample thickness. The (x,y,z) coordinates of the three points defining the plane and one point of a tissue sample are set by bringing the nano-DESI probe to the surface using manual control of the stage and by monitoring the distance between the probe and the surface using two CDD cameras. During the XY scan, the Labview program automatically controls the z-coordinate of the stage to maintain the initial distance between the sample and the nano-DESI probe. In addition, the program enables remote control of the Xcalibur software by providing “contact closure” signal in the beginning of each line scan. This initiates an automated data acquisition of individual lines. Data Visualization Using MSI QuickView. Lines acquired by the Xcalibur software were subsequently processed using custom software developed in our laboratory called MSI QuickView. This software enables visualization during data acquisition and subsequent off-line analysis of the MSI data. The program was written in MATLAB (MATLAB R2011b, Mathworks, Inc., Natick, MA) and compiled using the MATLAB Compiler. A copy of the program is available upon request. Figure 3 shows the typical workflow of MSI QuickView during data acquisition. First, the LTQ/Orbitrap saves a data

The converted data are subsequently visualized using the MSI QuickView graphical user interface (GUI) (Figure 4). For each data file acquired, the GUI can display 2-dimensional (2D) ion images and extracted ion chronograms (timedependent signal for a selected m/z range without chromatographic separation) for up to 6 different m/z values, or ranges, chosen by the user. By displaying each data file as it is acquired, the GUI assists in continuously validating the progress of the MSI experiment. After the entire MS image is acquired, MSI QuickView can be used for further visualization, processing, and analysis of the massive data set without any limitation on the number of ion images that can be created. There is no manual manipulation of the data files required by the user to visualize the data. The user can select time ranges and scroll through the spectra for different lines in the data set without having to manually load the data. To create ion images the user can simply pick peaks, or ranges, in the spectrum using the GUI. To process the images, the aspect ratios can be changed to fit the experimental values and the images can be displayed using existing or userdefined color maps. Optionally, ion images can be normalized to the total ion signal or a selected ion signal. The ion images can be saved in .tif format or exported as pixel intensity values in an ASCII matrix. For further analysis of the obtained images, line scans over the image can be performed in the GUI. The user chooses the path of the line and the GUI plots the intensity in the line scan as a function of location. Both the image with the drawn line and the plot can be saved in .tif format, and the data in the plot can further be exported to Excel. The LTQ/Orbitrap automated spectral gain control was utilized in these experiments, resulting in variations in the time required to acquire a complete spectrum during the image acquisition. The effect of this is a variation in both the width of each pixel and the number of pixels per line during image acquisition. In order to account for these pseudorandom changes in the pixel size, an alignment procedure is incorporated into MSI QuickView. This alignment procedure detects the maximum number of pixels per line for the entire data set and, for each line, interpolates the spectral values between every pixel based on the time at which they were acquired. MSI QuickView has been optimized for fast visualization and efficient handling of the large data sets which results from combining high spatial resolution and high spectral resolution. To avoid running out of memory, MSI QuickView keeps only the required information in memory by reading one data file at a time. The example data set shown in Figure 5 was visualized within 12 min using MSI QuickView. This data set consists of 31 500 total pixels (35 lines with up to 900 pixels per line) and an average of 13 636 data points (intensity vs m/z) per pixel (across a range of 7 × 106 different data points in total). The data set size on disk is 1.6 GB in CDF format. Imaging of Rat Brain Tissue Samples. Figure 5A shows a single-pixel high-resolution mass spectrum of the rat brain tissue sample. In this experiment, the tissue sample was scanned under the nano-DESI probe at 20 μm/s. The spectral acquisition time was ∼1.7 s/spectrum. Under these conditions, the pixel size is ∼35 μm. In this study, we analyzed a fairly large 16.5 mm × 6.8 mm area of the tissue sample with a 200 μm step size between the lines along the y-coordinate. The total acquisition time was 7.5 h. Comparison of spectral features observed in the gray matter at different times during image

Figure 3. MSI QuickView workflow.

file for each line acquisition onto the 32 bit Windows computer running the Xcalibur software. Because of data size, subsequent analysis is performed on a 64 bit Windows computer with the data directory shared between the two computers. MSI QuickView checks for new data files in this shared directory. Next, MSI QuickView utilizes the Xcalibur libraries to automatically convert the RAW files into an easily accessible binary Common Data Format (CDF, http://cdf.gsfc.nasa.gov/ html/FAQ.html#intro). 8353

dx.doi.org/10.1021/ac301909a | Anal. Chem. 2012, 84, 8351−8356

Analytical Chemistry

Article

Figure 4. Screenshot of the MSI Quickview data visualization software showing the mass spectrum (top), two ion images (bottom), and six selected ion chronograms (right) displayed during data acquisition.

acquisition indicated that the sample did not undergo significant chemical alteration. The optical image of the 10.9 mm × 6.8 mm portion of the analyzed tissue section after nanoDESI analysis is shown in Figure 5B. The darker color corresponds to the area sampled by the nano-DESI probe. The spatial resolution defined as the average line width measured with an optical microscope is 120 μm. Accurate mass measurement combined with MS/MS experiments and database searching were used for identification of metabolites and lipids observed in our imaging experiments. In this study, database searching was performed using Metlin (http://metlin. scripps.edu/) and Lipid Maps (http://www.lipidmaps.org/). In addition to abundant lipid peaks in the m/z 700−900 range, we observed low-molecular weight metabolites. High salt concentrations in the sample result in abundant Na+ and K+ adducts of metabolites and lipids in the spectrum; protonation was observed as a minor ionization pathway. For example, [M + K]+ of creatine is the most abundant metabolite at m/z 170.0327 while the most abundant lipid peak at m/z 798.5412 corresponds to the [M + K]+ ion of glycerophosphocholine with a total of 34 carbon atoms and 1 double bond in the fatty acid tail groups, hereafter abbreviated as PC 34:1. MS/MS spectra of the peaks labeled in Figure 5 are shown in the Supporting Information. In addition, we observed sodium and potassium adducts of glucose, glutamine and glutamate salts, taurine, hypoxantine, and several amino acids. These metabolites were tentatively assigned based on the accurate mass. Creatine, glucose, and glutamate are highly abundant metabolites in the brain.33,34 Both creatine and glucose are important in energy metabolism while glutamate is the most common excitatory neurotransmitter in the brain and is significant for learning and memory. Ion images of a solvent peak at m/z 429.2405, [M + K]+ ion of PC 34:1, and [M + K]+ ion of creatine are shown in Figures 5C−E. Note that the images are shown without normalization

Figure 5. (A) High-resolution spectrum obtained by averaging the signal along one acquired line; (B) optical image of the 10.9 mm × 6.8 mm portion of the scanned tissue section. The lines are the traces left on the tissue by the nano-DESI probe. Note that we only observe discoloration but no scratching of the sample by the probe. (C−E) Ion images obtained for selected mass-to-charge ratios: (C) solvent peak at m/z 429.2405, (D) [M + K]+ of PC 34:1 at m/z 798.5412; (E) [M + K]+ of creatine at m/z 170.0327. The scale bar is 2 mm. 8354

dx.doi.org/10.1021/ac301909a | Anal. Chem. 2012, 84, 8351−8356

Analytical Chemistry

Article

or smoothing. Solvent peaks are typically enhanced outside the tissue section making it easy to distinguish them from analyte peaks. The spatial distribution observed for PC 34:1 is consistent with previous studies.35−37 This lipid is observed in lower abundance in the corpus callosum and in the dorsal third ventricle. In contrast, creatine is evenly distributed across the tissue section. This observation is consistent with the results reported by Nemes et al. using LAESI imaging.24 The signal-to-noise ratio of 3000 was obtained in a single spectrum for the most abundant peak at m/z 798.5412 corresponding to [M + K]+ ion of PC 34:1. The ratio of the peak intensity of the most abundant PC 34:1 peak to the lowest-abundance peaks for which we could create 2D images (the dynamic range) was greater than 2000. High-mass resolution combined with high dynamic range allowed us to detect both abundant and low-abundance metabolites and lipids in the tissue sample. Figure 6 shows an example of three

Figure 7. (A) Higher spatial-resolution ion image of the [M + K]+ ion of creatine at m/z 170.0327 acquired over a 2 mm × 2 mm area of the tissue section overlaid with the optical image of the sample. The length of the white scale bar is 200 μm. (B) Line scan along dotted blue line shown in panel a showing the suppression of the creatine signal. The creatine signal increases and decreases over ∼20 μm distance, which defines the spatial resolution of this experiment.

Figure 6. (A) Region of the high-resolution spectrum shown in Figure 5 showing three isobaric low-abundance peaks. Ion images obtained for (B) [M + K]+ of PE (O-40:7) at m/z 812.5006, (C) [M + Na]+ of PC (O-38:6) at m/z 812.5578, (D) the second 13C isotope of the abundant [PC 36:1 + Na]+ peak at m/z 810.5987. The scale bar is 2 mm.

low-abundance (