Quantitation of Spatially-Localized Proteins in Tissue Samples using

Article pubs.acs.org/ac

Quantitation of Spatially-Localized Proteins in Tissue Samples using MALDI-MRM Imaging Elizabeth J. Clemis,† Derek S. Smith,† Alexander G. Camenzind,† Ryan M. Danell,‡ Carol E. Parker,† and Christoph H. Borchers*,†,§ †

University of Victoria-Genome BC Proteomics Centre, University of Victoria, 3101-4464 Markham Street, Victoria, BC, Canada, V8Z 7X8 ‡ Danell Consulting, 3717 Willow Run Drive, Greenville, North Carolina 27858, United States § Department of Biochemistry & Microbiology, University of Victoria, Petch Building Room 207, 3800 Finnerty Road Victoria, BC V8P 5C2, Canada S Supporting Information *

ABSTRACT: MALDI imaging allows the creation of a “molecular image” of a tissue slice. This image is reconstructed from the ion abundances in spectra obtained while rastering the laser over the tissue. These images can then be correlated with tissue histology to detect potential biomarkers of, for example, aberrant cell types. MALDI, however, is known to have problems with ion suppression, making it difficult to correlate measured ion abundance with concentration. It would be advantageous to have a method which could provide more accurate protein concentration measurements, particularly for screening applications or for precise comparisons between samples. In this paper, we report the development of a novel MALDI imaging method for the localization and accurate quantitation of proteins in tissues. This method involves optimization of in situ tryptic digestion, followed by reproducible and uniform deposition of an isotopically labeled standard peptide from a target protein onto the tissue, using an aerosol-generating device. Data is acquired by MALDI multiple reaction monitoring (MRM) mass spectrometry (MS), and accurate peptide quantitation is determined from the ratio of MRM transitions for the endogenous unlabeled proteolytic peptides to the corresponding transitions from the applied isotopically labeled standard peptides. In a parallel experiment, the quantity of the labeled peptide applied to the tissue was determined using a standard curve generated from MALDI time-of-flight (TOF) MS data. This external calibration curve was then used to determine the quantity of endogenous peptide in a given area. All standard curves generate by this method had coefficients of determination greater than 0.97. These proof-of-concept experiments using MALDI MRM-based imaging show the feasibility for the precise and accurate quantitation of tissue protein concentrations over 2 orders of magnitude, while maintaining the spatial localization information for the proteins.

■

INTRODUCTION To better understand the biological functions of different cellular regions within a tissue, it is highly advantageous to determine the location and abundance of the constituent proteins. Mass spectrometry (MS) has proven invaluable for the absolute quantitation of proteins in solutions such as plasma and tissue homogenates; however, important anatomical information is lost when tissue samples are homogenized before protein quantitation. It would therefore be beneficial to have a quantitation method that would maintain this spatial information. MALDI imaging is a technique introduced over 10 years ago that allows 2-dimensional spatial resolution of proteins, peptides, and small molecules in tissue sections.1 In MALDI imaging, a spectrum is acquired at each of a regular series of positions across a section of tissue. Each spectrum contains molecular weight and abundance information representative of the analytes present at that position. Specific ion images can © 2012 American Chemical Society

then be generated from a plot of the abundance of any ion measured, as a function of individual pixel location. MALDI imaging has many useful applications, for example, in cancer research it can be used to identify tumor type, grade, stage, and surgical margins.2 Visualization of aberrant molecular pathology between healthy and diseased tissue can be used for biomarker discovery.3 In addition, it is possible to apply this technology to the imaging of drugs and their metabolites simultaneously, in order to determine the site of action and to track the distribution of the analytes over a period of time.4,5 MALDI imaging provides a qualitative picture of analyte distribution. However, even when the application of the matrix onto the tissue is homogeneous, the formation of the matrix crystals is not. These crystals can differ in density per unit Received: October 31, 2011 Accepted: February 21, 2012 Published: February 21, 2012 3514

dx.doi.org/10.1021/ac202875d | Anal. Chem. 2012, 84, 3514−3522

Analytical Chemistry

Article

Figure 1. Schematic of the MALDI-MRM quantitation method. (A) Peptide image of endogenous myelin basic protein. A thin slice of flash frozen tissue is thaw mounted onto a conductive ITO coated glass slide. Trypsin, labeled peptide, and matrix are applied separately, using aerosol deposition. Ion images are created from the MRM data obtained on a MALDI-QTRAP mass spectrometer. The ion displayed is from the endogenous tryptic peptide. (B) Absolute quantitation of heavy-arginine labeled peptide by aerosol deposition. The labeled peptide is deposited onto the conductive ITO coated glass slide. A 0.5 μL aliquot of the synthetic unlabeled version of the sprayed peptide in matrix is spotted onto the slide in increasing concentrations. Ion images of the spots are obtained in reflector positive mode of a MALDI time-of-flight (TOF) mass spectrometer. The area of each spot is calculated to determine the amount of unlabeled peptide per ablation area. This is then plotted against the average ion signal intensity ratio of the unlabeled to labeled peptide over the area of the spot in order to generate a standard curve. (C) Schematic representation of continuous rastering acquisition on the MALDI-QTRAP. The 1000 Hz laser is pulsed without pause while the plate is moved continuously. The continuous ablation results in an effectively constant stream of ions into the mass spectrometer. Each transition is detected as the laser is traveling a discrete distance across the tissue (45 μm). With a scan speed of 0.9 μm/ms and a dwell time of 50 ms, 50 laser shots are summed as the laser (diameter = 200 μm) travels over 45 μm of tissue. There is a 5 ms settling time between transitions during which the laser moves an additional 4.5 μm. The laser travels across 180 μm of tissue to collect the data for each set of 4 transitions during which time analyte signals are detected and, then, an additional 45 μm for the detection of the alignment line transition. This makes a 275 ms duty cycle, which then repeats.

with respect to a quantitative method. For example, the whole analysis can be rendered less sensitive because large signal intensities are the greatest contributors to the TIC normalization of spectra. Often these large signals fluctuate dramatically from sample to sample and can, at times, even be saturated. Furthermore, if intense signals from ions that are not clinically relevant are more common in one clinical group than another, then TIC normalization will lead to overtraining during classification and will inappropriately suppress all signal intensities in the group.8 In addition to the above-mentioned matrix effects, the ion signals from competing endogenous species, such as lipids and other blood compounds, may cause significant suppression of the signal from the analyte of interest. Accurate quantitation in MALDI is extremely difficult without the use of labeled

surface, size, quality, and analyte extraction efficiency, depending on the local physical properties within the tissue section. The inconsistencies in matrix crystal formation can result in errors in determining the abundances of the analytes, leading to incorrect conclusions about their distributions. The standard approach for mitigating these matrix effects is normalization based on the total ion current (TIC).6,7 Total ion current normalization is performed by summing all of the intensities in each spectrum and obtaining a total intensity value for each spectrum. These totals are then averaged. Each intensity value within a spectrum is then multiplied by the ratio of its total intensity to the average total intensity. This normalizes the spectrum because, after this process, the total ion intensity for each spectrum now equals the average total intensity. This normalization method has several limitations 3515



Article

Figure 2. Protein and peptide identification. (A) Distribution of ions detected at m/z 14097.4 and 18531.3 and a summed ion spectrum analyzed in linear positive mode, using SA as the matrix. (B) Summed ion spectrum analyzed in linear positive mode, using CHCA as the matrix. (C) Spectrum that was obtained from a section following an in situ trypsin digestion and subsequent application of CHCA matrix using the Bruker ImagePrep station. The data was obtained from an analysis in the linear positive mode on a portion of the tissue where myelin basic protein (m/z 14097.4 and 18351.3) expression was observed. (D) Ion images corresponding to signals with m/z 726.5, m/z 1336.7, and m/z 1503.0 from tissue that was analyzed in reflector positive mode. These ion images were selected because they display a similar morphology to the m/z 14097.4 and 18351.3 ion images following in situ tryptic digestion. The base-peak (*) at m/z 1090.6 has been previously identified as a tryptic peptide from PEP-19 (AAVAIQSQFR), by Groseclose et al.13 (E) One of the three MS/MS spectra acquired directly from the tissue following tryptic digestion and CHCA matrix coverage and analyzed in LIFT mode. (F) MS/MS sequence data from all three ions which was searched against the Rattus norvegicus database using BLASTp. The m/z 14097.4 and 18351.3 ion signals were identified as myelin basic protein isoforms 4 and 2, respectively. To distinguish between the distribution of these two isoforms, ions unique to each isoform could have been selected.

However, it is necessary to normalize against a reference standard that has identical ionization properties to the analyte of interest. In MALDI imaging, there is the additional challenge that it is also necessary for the standard to be homogenously distributed on the surface of the tissue section. If a uniform and homogeneous distribution can be achieved, successful quantitation can be performed through the use of an isotopically labeled version of the analyte of interest due to their identical ionization efficiencies. The Applied Biosystems 4000 QTrap mass spectrometer equipped with a vacuum MALDI source is able to provide data for both the detection and the accurate quantitation of targeted

standards, not only because of matrix and ion suppression effects, but also because of compound-dependent ionization efficiencies. The abundance of an ion signal in a mass spectrum is a poor indicator of the amount of analyte present; however, differences in the ion signal intensity between an analyte and an isotopically coded standard of the same analyte accurately reflect differences in the abundance. Additionally, when the amount of labeled standard is known, the quantity of the endogenous analyte can be accurately determined based on the ratio, regardless of the suppression and matrix effects.8 In solution-based analyses, normalization against a reference peak can correct for matrix effects on the analyte signal. 3516



Article

room temperature. The tissue was rinsed in a series of alcohol and water solutions1 × 15 s in 70% ethanol and 1 × 15 s in 90% ethanol: 9% acetic acid (v/v). For the protein discovery analysis, the tissue was coated with 20 mg/mL sinapinic acid: 70% ACN:0.1% TFA (v/v). For peptide imaging, the tissue sections were digested with a trypsin solution of 46 μg/mL in 100 mM ammonium bicarbonate and coated with 10 mg/mL CHCA:70% ACN:0.1% TFA (v/v). Isotopically labeled arginine and unlabeled arginine versions of a myelin basic protein tryptic peptide were synthesized and quantitated.11,12 In all tissue imaging experiments, the matrix, trypsin, and heavy-labeled peptide solutions were deposited on the tissue using the Bruker Daltonics ImagePrep station (Bremen, Germany). The ImagePrep station was run in manual mode with the following settings for the deposition of a 2.0 pmol/μL solution of labeled peptide: cycles = 15, power = 30, modulation = 0, spray = 1 s, incubation = 100 s, and dry = 100 s. The ImagePrep station was also run in manual mode for the trypsin application with the following settings: cycles = 30, power = 20, modulation = 10, spray = 1 s, incubation = 60 s, and dry = 100 s. The application of the CHCA matrix used the same settings as the trypsin application; however, 15 more spray cycles were added. This method repeatedly sprays small amounts of the trypsin solution onto the tissue and incubates for 80 min at room temperature (∼22 °C) in a humid environment. The sprayed trypsin solution appeared to dry completely between each successive application. For the generation of the standard curve to determine the amount of labeled peptide sprayed per area, a series of dilutions of the synthesized unlabeled myelin basic protein tryptic peptide in CHCA matrix solution (0.05, 0.1, 0.2, 0.4, 0.6, 0.8, and 1.0 pmol/μL) were prepared from a 10 pmol/ μL stock solution. A 0.5 μL aliquot of each solution was spotted onto a glass slide that had been previously been sprayed with the labeled peptide. Imaging Mass Spectrometry. All data acquired by the Bruker Ultraflex III was analyzed using the Bruker FlexControl, FlexImaging, FlexAnalysis, and Biotools software. Spectra were recorded in the linear positive mode for protein analysis or in the reflector positive mode for peptide analysis. The MS/MS spectrum was obtained using the LIFT mode on the Bruker Ultraflex III. The spectra were recorded in MRM mode using the transitions 726.5/324.3 and 736.5/324.3. All data acquired by the AB Sciex QTRAP with the attached MALDI source was analyzed and then exported from the Analyst and M3Q server software. The exported MRM traces were imported into our custom-built imaging software for processing. This software was written in LabVIEW (National Instruments, Austin, TX) and automatically detects peaks within alignment traces which are placed on the sample substrate. The alignment traces allow the data to be unfolded into a 2-D image. Image contrast enhancement, dimensioning, and pixel-by-pixel analysis are all available within this custom software package.

analytes in tissues. This instrumentation has the ability to perform MALDI imaging with the additional capability of performing multiple reaction monitoring (MRM) analysis. Two stages of mass filtering are employed by a triple-quadrupole mass spectrometer in MRM mode, which allows the user to achieve high levels of both sensitivity and specificity in the quantitation of analytes from a complex sample. MRM-based MALDI imaging will have increased specificity through the precise selection of a targeted mass in the first quadrupole, and a specific fragment ion mass in the third quadrupole. In situ trypsin digestion allows the resulting peptide fragments to remain close to their point of origin in the two-dimensional space of a tissue slice. In MRM mode, the resulting tryptic peptides along with the uniformly deposited stable isotopelabeled peptide analogs can be monitored for the accurate quantitation of the protein of interest. Previously, MRM based MALDI imaging has been used to map the localization of the drug Moxifloxacin and a reference standard in infected rabbit lung biopsies to determine the penetration of the drug into granulomas. The reference standard used in this study was another fluoroquinolone compound that was applied by hand with a TLC sprayer. While this study did not attempt to quantitate the analyte of interest, they were able to show improved sensitivity and selectivity with the MRM method that enabled detailed drug localization with the granuloma that could not be determined by alternative methods.9 The quantitative imaging method used in this study is outlined in Figure 1 and involves an in situ trypsin digestion, followed by uniform spraying of the tissue sample with an isotopically coded peptide that is representative of the protein of interest using an automated spray device, matrix application, and MALDI MS analysis. The labeled peptide acts as an internal standard for the endogenous tryptic peptide. We present here a new MALDI-MRM based imaging MS approach for the absolute quantitation of proteins present in thin tissue sections after in situ enzymatic digestion.

■

EXPERIMENTAL SECTION Chemicals and Reagents. The rat brain tissue used for these experiments was purchased from Pel-Freez (Rogers, Arkansas). HPLC-grade ethanol, acetonitrile, and water were purchased from Sigma-Alrich (St. Louis, MO). Trifluoroacetic acid (TFA) was purchased from Thermo Fisher Scientific (Rockford, IL). Alpha-cyano-4-hydroxycinnamic acid (CHCA) was purchased from Sigma-Aldrich (St. Louis, MO) and recrystallized. Sinapinic acid (SA) was also purchased from Sigma-Aldrich and was used without further purification. Trypsin was purchased from Promega (Madison, WA) and was diluted with 100 mM ammonium bicarbonate to a final trypsin concentration of 46 μg/mL. A tryptic peptide from rat myelin basic protein was synthesized with a stable-isotopically coded arginine residue (+10 Da) (HGFLPR) at the University of Victoria−Genome BC Proteomics Centre, Victoria, Canada, according to our previously described protocols.10,11 The peptides were diluted in 30% acetonitrile and 0.1% formic acid (FA), also purchased from Sigma-Aldrich. Sample Preparation. All rat brain tissue samples were snap frozen and sectioned (12 μm sections for experiments shown in Figure 2 and 10 μm sections for the experiments shown in Figure 5) on a cryostat and thaw mounted onto conductive ITO coated glass slides. The slides were placed in a vacuum for 60 min to allow the tissue sections to dry and equilibrate to

■

RESULTS AND DISCUSSION The quantitative imaging method we have developed, shown schematically in Figure 1, involves an in situ tryptic digestion followed by uniform application of a “normalization peptide”, similar in concept to the addition of a stable-isotope-labeled internal standard peptide (a SIS peptide) in solution-based quantitation. This peptide was synthesized with an isotopic label on the C-terminal arginine. The sample is then coated in 3517



Article

Figure 3. Reproducibility of the aerosol application of peptides by the ImagePrep Matrix Deposition device. The myelin basic protein tryptic peptide HGFLPR was synthesized with and without an isotopic label. A 2 pmol/μL solution of the heavy labeled peptide was deposited onto a conductive ITO coated glass slide using the ImagePrep station in 15 1 s spray cycles. The same procedure was then followed for the subsequent application of the unlabeled peptide. The slide was then coated in CHCA matrix and an area 1.7 × 1.6 cm2 was analyzed on the MALDI-TOF/TOF in reflector positive mode.

A rat brain section slice, 12 μm thick, was washed and coated with sinapinic acid matrix. MS data acquisition was performed by MALDI-TOF MS in linear mode with a pixel resolution of 150 μm and 200 shots per pixel. From this tissue section, ion images with m/z values of 14 097 and 18 351 are displayed in Figure 2A along with a summed ion spectrum. Indeed, the predominant isoforms of MBP in adult rats are the 14 and 18.5 kDa isoforms. For quantitation based on stable-isotope-labeled peptides, the protein must be converted to peptides and the quantity of the peptides must accurately reflect the amount of protein in the tissue. It is therefore desirable to achieve as complete a digestion as possible of the proteins of interest in the tissue . For the imaging of the intact proteins (Figure 2), SA was used as the matrix, because it produces higher sensitivity for proteins. A spectrum from rat brain using SA as the matrix is shown in Figure 2A, with the masses of two of the MBP isoforms colorcoded to match the respective images. To demonstrate the effectiveness of the digestion, we also show the spectrum of the brain tissue with CHCA as the matrix, before and after digestion (Figure 2B and C, respectively). The spectrum shown in Figure 2C was obtained from a tissue section following in situ tryptic digestion and subsequent application of CHCA matrix using spray deposition. The spectrum was obtained from an analysis in the linear mode on a portion of the tissue where myelin expression is observed. The ion signals corresponding to the intact myelin basic protein isoforms that are observed in the nondigested tissue sample (Figure 2A and B) are greatly reduced in the trypsin-digested tissue (Figure 2C and Supporting Information Figures 2 and 3; note the change in scale of the intensity axis). In addition, the effectively complete digestion of the proteins was confirmed by comparison of the spectra from the spray deposition method and an in situ method using 15 μL spots containing trypsin solution (46 μg/mL) applied to the tissue and incubated for 16 h at 37 °C in a humidity controlled device (data not shown).

CHCA matrix and analyzed in MRM mode using the AB Sciex MALDI-QTRAP. In order to quantitate based on the ratio of the endogenous to the labeled peptide, it is necessary to determine the amount of this peptide that is coating the tissue. To do this, the labeled peptide is applied to a blank indium tin oxide (ITO) coated glass slide using the same ImagePrep spray settings, immediately following the aerosol application of the labeled peptide onto the tissue sample. Increasing concentrations of the synthetic unlabeled version of the sprayed peptide in CHCA matrix is then spotted onto the slide. The spots are imaged using a MALDI-TOF/TOF in MS reflector positive mode. The average signal intensity for the unlabeled peptide is divided by the average signal intensity for the labeled peptide over the entire spot and plotted on the Y-axis to generate a standard curve. The X-axis of the standard curve is the amount of spotted unlabeled peptide per area, which is calculated from the known concentration of peptide over the determined area of the imaged spot. The amount of sprayed labeled peptide per area can be calculated from the regression line. The amount of peptide is expressed in amount per ablation area. The ablation area represents the area volatized by the laser during each point of data acquisition. To demonstrate the proof-of-principle for a method that begins with protein identification and ends with the quantitation of the discovered protein, we performed the following imaging experiments for the already well-characterized myelin basic proteins (MBPs).13 MBPs are the major protein component of myelin, the electrically insulating material that forms a layer around the axon of a neuron. There are several MBP isoforms that all arise from differential transcription initiation and splicing from the Golli (“genes of the oligodendrocyte lineage”) gene complex during different stages of neural development.14 The MBP isoforms, which range from 14 to 21 kDa, are encoded from the major alternatively spliced MBP mRNAs.15,16 3518



Article

Figure 4. Absolute quantitation of heavy-arginine labeled peptide by aerosol deposition. Following the application of the labeled peptide, 0.5 μL of the synthesized unlabeled version of the m/z 726.5 myelin peptide was spotted onto a conductive ITO coated glass slide in increasing concentrations and imaged using the Bruker Ultraflex III. (A) Ion images of the spots obtained with a 125 μm spatial resolution inset into a zoomed in region of the total ion chromatogram. Part A shows an image of the 0.5 μL spot with the corresponding signals from the light and heavy peptide averaged over the entire area of the spot. Six different concentrations of the light peptide were spotted onto the slide. Each pixel is the sum of 300 shots, and there were approximately 250−300 pixels per spot image. (B) Total area of each spot which was calculated in order to determine the amount of unlabeled peptide per ablation area. This was then plotted against the ion signal intensity ratio of the unlabeled to labeled peptide to generate a standard curve. The amount of heavy labeled peptide deposited by the ImagePrep during the application was determined to be 93.0 ± 0.1 amol/ablation area. The laser diameter is 50 μm; therefore, the ablation area represents 1963.5 μm2 on the slide.

device in droplets less than 50 μm in size.17 Because of the small droplet size, the resulting peptides maintain the location of the origin protein. Following the application of matrix, data acquisition (in the reflector positive ion mode with a spatial resolution of 125 μm and 200 shots per pixel) and image generation, ions were selected which displayed a similar distribution to the 14 and 18 kDa target proteins (Figure

On a 12 μm thick slice of tissue from an adjacent portion of the brain, an in situ protein digestion was performed to confirm the identify the 14 and 18 kDa analytes based on the peptide molecular weights and sequences obtained from regions that displayed a similar spatial morphology. First, trypsin was deposited onto the tissue in 30 1 s spray cycles over a period of 1.7 h, followed by the application of CHCA. The trypsin solution is deposited onto the tissue by the aerosol generating 3519



Article

ablation area represents 1963.5 μm2. An independent set of experiments that tested the reproducibility in individual spots such as those used to generate the standard curve obtained CV values of less than 13% (see Supporting Information Table 1). Using the equation of the regression line, as determined from the standards, the amount of labeled peptide per area deposited by aerosol deposition was calculated to be 93.0 amol/ablation area with a standard deviation of ±0.1, which was calculated from the sum of squared errors of prediction (SSE).21 Although we have demonstrated the reproducibility of Bruker ImagePrep station for sequential peptide applications, the extent of the variability of the spray application over time is evident from the results displayed in Table 1. Three standard curves were made

2C). In this case, ion images are shown that correspond to m/z 726.5, 1336.7, and 1503.0. The MS/MS spectra acquired was uploaded to Mascot’s MS/ MS search engine for sequence identification18,19 (a sample Mascot result file is shown in Supporting Information Figure 1). The sequences of the three ions shown in Figure 2C were searched against the Rattus norvegicus database using BLASTp to determine if they were unique within the rat proteome to the myelin basic protein isoforms.20 One of the three MS/MS spectra acquired directly from the tissue following tryptic digestion and CHCA matrix coverage is shown in Figure 2D, and the ion at m/z 726.5 was shown to have the amino acid sequence HGFLPR. All of these three ions were identified as tryptic peptides from myelin basic protein (Figure 2E). The external quantitation method is based on the known quantity of labeled peptide per area on the blank glass slide and extrapolating this quantity to the area of the tissue. For this assumption to be valid, it must be demonstrated that the Bruker ImagePrep station is capable of reproducible spray applications. Figure 3 shows two sequential spray applications of peptide onto a blank glass slide. The unlabeled tryptic myelin protein (m/z 726.5) was sprayed, followed by the isotopically labeled version of the same peptide (m/z 736.5) using the same spray settings. Following the application of matrix, data was collected from a portion of the slide by MALDI-TOF MS in reflector positive mode. The larger rectangle in the center is a colormerge of unlabeled (red) and the labeled (green) ion signals. In the center rectangle, the most intense green color comes from parts of the slide where there was an excess of unlabeled peptide. Likewise, the red color represents an area where there was an excess of labeled peptide. The orange color is representative of areas where the peak areas were close to a one-to-one ratio. Although the uniformity is not perfect, signal intensities of close to 1:1 were observed for the majority of the spectra acquired, and CVs of less than 13% were obtained within an experimentally determined region of the slide where the spray is uniform (see the Supporting Information). This demonstrates that the spray deposition is quite reproducible when the isotopically labeled standard and the matrix are applied in succession. However, we recognize that this region might be different with different sprayers or spotters and recommend that this area of homogeneity be determined by each laboratory. The overall results are good, but we think that we can get peak area ratios that are even closer to 1:1 and an even more homogeneous distribution, by increasing the number of spray cycles and using a less-concentrated peptide solution. This is something we will be examining in future experiments. The method described in Figure 1 was applied to a 10 μm thick coronal section of rat brain tissue in order to generate a MALDI-MRM image showing the distribution of myelin basic protein. Alongside the tissue sample, a blank portion of the ITO coated glass slide was also sprayed with the labeled peptide. Increasing concentrations of the synthetic unlabeled peptide in matrix were spotted onto the slide and imaged by MALDI-TOF MS in the reflector mode. The images of these spots and their corresponding ion signal spectra are shown in Figure 4a. The ratios between the signal intensities of the unlabeled (m/z 726.5) and labeled (m/z 736.5) peptides were averaged over the entire area of these spot and plotted against the amount of unlabeled peptide per ablation area to generate the standard curve shown in Figure 4b. The TOF instrument has a nominal ablation spot diameter of 50 μm; therefore, each

Table 1. Results from the Generation of Three Separate Standard Curves over a 1 month Perioda

a

amount of labeled peptide per area (amol/ ablation area)

coefficient of determination (R2)

43.7 93.0 100.0

0.97775 0.98980 0.97385

Each ablation area represents 1963.5 μm2 of the slide.

over a 1 month period. The spray settings on the ImagePrep station were kept constant in all three experiments. This method consistently produces a linear calibration curve, as shown by the R2 values in Table 1, even though the quantity of peptide sprayed varies over extended periods of time. This demonstrates the necessity for a standard curve to be generated along with each the MALDI-MRM imaging experiments when quantitation is required. The selectivity of the MRM method greatly increases the sensitivity due to the removal of background ions by the first quadrupole. The MRM ion pairs used to acquire the data needed to construct the image shown in Figure 5 were m/z 726.5/324.3 and 736.5/324.3, which have the peptide sequence HGFLPR. The m/z 726.5/324.3 transition represents the endogenous tryptic peptide of the myelin basic protein isoforms from the in situ digestion, and the m/z 736.5/324.3 transition represents the labeled peptide that was used as an external standard. The image on the right has been normalized to the sprayed standard peptide. Because the ratio of the endogenous and the SIS peptide is used for quantitation, the externally applied reference standard compensates for any nonhomogeneity in the matrix application, as well as any differences in ion suppression due to the heterogeneity of the tissue. This method has the capability of giving the amount of endogenous peptide at any location within the tissue by clicking on the image generated by the software. The amount of peptide is determined differently when using the MALDI-quadrupole mass spectrometer compared to the MALDI-time-of-flight mass spectrometer. With the TOF instrument, the laser rasters across the tissue in discrete spots. The laser in the MALDI source attached to the MALDI-quadrupole instrument fires continuously while moving in a serpentine pattern over the tissue. In this experiment, the dwell time was 50 ms and the laser moved across the tissue with a velocity of 0.9 mm/s; therefore, each ablation area for this method represents the volume of peptides volatized by the laser as it travels across 45 μm of a 10 μm thick tissue section. The diameter of the laser is 200 μm and travels over 45 μm of tissue during each dwell time; therefore, the 3520



Article

Figure 5. MALDI images created from MRM data. The image shows the endogenous tryptic peptide from myelin basic protein from a coronal section of rat brain tissue. The image on the right has been normalized to the sprayed labeled peptide that had been determined to be distributed on the tissue at 93.0 ± 0.1 amol/1963.5 μm2 (see Figure 4). The externally applied reference standard reduces the matrix localization effects as well as the compound specific ion suppression due to the heterogeneity of the brain tissue. On the basis of the MRM data acquired, the amount of endogenous peptide per area was calculated from the ratio of the endogenous to the labeled peptides and then multiplied by 426.3 ± 0.45 amol/ ablation area.

ablation area is 9000 μm2. On the basis of the MRM data acquired, the amount of endogenous peptide per area was calculated from the ratio of the endogenous to the labeled peptides and then multiplied by 426.3 ± 0.45 amol/ablation area. It is important to note that the quantities of protein determined by this method come from tissue sections that were 10 μm thick and, therefore, need to be considered in three dimensions. As in other methods that use peptides as surrogates for proteins, this imaging method quantifies peptide signals resulting from a tryptic digestion step; therefore, if any modified forms of the parent protein are present, this information may be lost. For example, Figure 1 shows two isoforms of MBP (14 and 18 kDa), as identified by Crecelius et al.22 in mouse brain. Because the peptides used in our calculations come from the portions of these highly homologous protein isoforms that are identical, the amount of peptide calculated from the MRM MALDI data corresponds to the tissue concentrations of the sum of the protein isoforms that contain these peptides. However, if it is necessary to quantitate a specific isoform, this limitation can be overcome by using a peptide that is unique to the isoform and is detectable by MALDI.23 This study represents the first on tissue absolute quantitation of MBP; therefore, a comparison of the results with another method is not possible. However, we can calculate the average MBP quantity per cubic micrometers from a whole brain and compare this to our results. Using data from a study on MBP in mouse brain,24 we calculated the amount of MBP protein to be 0.045 amol/μm3. From the data collected in our study, we determined the quantity of MBP/μm3 to be in a range between 0.001 and 0.071 amol/μm3. While it is one of the most abundant proteins in brain tissue, MBP is not expressed homogenously throughout the brain; therefore, an average calculated over the entire brain is only an estimate. Furthermore, the results from our study are from rat brain tissue, which could be slightly different of those from mice. The strategy employed here is a targeted imaging approachit is not a nontargeted strategy like traditional MALDI imaging methods, but the goal here is to determine tissue concentrations, not simply to determine relative signal intensities. Furthermore, as in classical in-solution MRM quantitative assays, this method also requires that each peptide used in the quantitation protocol be synthesized in its labeled and unlabeled form. Currently our instrument is limited to five transitions (see Figure 1C). In these proof-of-principle experiments, we used

four of these transitions: one transition for the endogenous peptide, one transition for the SIS, one transition to blank the mass spectrometer between peptide transitions, and one for the alignment. We have demonstrated this method for the quantification of one peptide per tissue slice, although other SIS peptides could be added. In these proof-of-concept experiments, we chose to use one peptide from MBP instead of two because our initial focus was on method development and the feasibility of the approach. Future experiments will use two peptides per protein (with one transition per peptide) for a more confident quantitative result.25 For the results shown in this paper, we used an automated aerosol-generating device for the application of trypsin and matrix solutions onto tissue sections. Alternative matrix deposition technologies deposit trypsin and matrix onto tissue sections in discrete spots using very small volumes (80−150 pL),26 either using technology similar to that used in inkjet printers27 or by the ejection of droplets from a fluid interface in a reservoir by acoustic energy.28,29 While the application of solutions using the spotting technologies is more controlled and homogeneous, the distance between sample spots is higher than when sprayed droplet deposition is used. The spots deposited by the spotters range from 100 to 200 μm in diameter with a distance of 300 μm between the center of one spot to the center of the next. This limits the achievable image resolution.30 This is important because, although the diameter of the laser on the MALDI source attached to the QTRAP is 200 μm, the ions are detected as the laser travels over distance increments of 45 μm (Figure 1C). We have shown in this study that with an aerosol spraying system homogeneous deposition of the stable-isotope-labeled internal standard is achievable. Because the endogenous and SIS peptides have the same physicochemical properties, and the ratio of these two ion signals is used for quantitation, the presence of this internal standard in each MALDI-MS scan allowed the accurate quantitation of the target tissue protein over 2 orders of magnitude in concentration, while maintaining the localization information on the protein in the tissues.

■

ASSOCIATED CONTENT

S Supporting Information *

Figures S1−S4 and Table S1. This material is available free of charge via the Internet at http://pubs.acs.org. 3521



■

Article

(23) Zulak, K. G.; Lippert, D. N.; Kuzyk, M. A.; Domanski, D.; Chou, T.; Borchers, C. H.; Bohlmann, J. Plant J. 2009, 60, 1015−1030. (24) Barbarese, E.; Carson, J. H.; Braun, P. E. J. Neurosci. Res. 1972, 31, 779−782. (25) Domanski, D.; Percy, A. J.; Yang, J.; Chambers, A. G.; Hill, J. S.; Cohen Freue, G. V.; Borchers, C. H. Proteomics 2012. (26) Végvári, A.; Fehniger, T. E.; Gustavsson, L.; Nilsson, A.; Andrén, P.; Kenne, K.; Nilsson, J.; Laurell, T.; Marko-Varga, G. J. Proteomics 2010, 73, 1270−1278. (27) Nokihara, K.; Mihara, H. Tanpakushitsu Kakusan Koso 2002, 47, 626−632. (28) Baluya, D. L.; Garrett, T. J.; Yost, R. A. Anal. Chem. 2007, 79, 6862−6867. (29) Aerni, H.-R.; Cornett, D. S.; Caprioli, R. M. Anal. Chem. 2006, 78, 827−834. (30) Végvári, A.; Fehniger, T. E.; Gustavsson, L.; Nilsson, A.; Andrén, P.; Kenne, K.; Nilsson, J.; Laurell, T.; Marko-Varga, G. J. Proteomics 2010, 73, 1270−1278.

AUTHOR INFORMATION

Corresponding Author

*Tel.: (250) 483-3221. Fax: (250) 483-3238. E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors would like to thank Dr. Pierre Chaurand of the Department of Chemistry at the Université de Montréal for helpful discussions. We would also like to thank Genome Canada, Genome BC, and the Western Economic Diversification for providing funding for the University of Victoria− Genome BC proteomics facility.

■

REFERENCES

(1) Caprioli, R. M.; Farmer, T. B.; Gile, J. Anal. Chem. 1997, 69, 4751−4760. (2) Chaurand, P.; Sanders, M. E.; Jensen, R. A.; Caprioli, R. M. Am. J. Pathol. 2004, 165, 1057−1068. (3) Meistermann, H.; Norris, J. L.; Aerni, H.-R.; Cornett, D. S.; Friedlein, A.; Erskine, A. R.; Augustin, A.; De Vera Mudry, M. C.; Ruepp, S.; Suter, L.; Langen, H.; Caprioli, R. M.; Ducret, A. Mol. Cell. Proteomics 2006, 5, 1876−1886. (4) Troendle, F. J.; Reddick, C. D.; Yost, R. A. J. Am. Soc. Mass Spectrom. 1999, 10, 1315−1321. (5) Reyzer, M. L.; Hsieh, Y.; Ng, K.; Korfmacher, W. A.; Caprioli, R. M. J. Mass Spectrom. 2003, 38, 1081−1092. (6) Alfassi, Z. B. J. Am. Soc. Mass Spectrom. 2004, 15, 385−387. (7) Rasmussen, G. T.; Isenhour, T. L. J. Chem. Inf. Comput. Sci. 1979, 19, 179−186. (8) Duncan, M. W.; Roder, H.; Hunsucker, S. W. Brief. Funct. Genom. Proteomics 2008, 7, 355−370. (9) Prideaux, B.; Dartois, V.; Staab, D.; Weiner, D. M.; Goh, A.; Via, L. E.; Barry, C. E.; Stoekli, M. Anal. Chem. 2011, 83, 2112−2118. (10) Bordeerat, N. K.; Georgieva, N. I.; Klapper, D. G.; Collins, L. B.; Cross, T. J.; Borchers, C. H.; Swenberg, J. A.; Boysen, G. Proteomics 2009, 9, 3939−3944. (11) Kuzyk, M. A.; Parker, C. E.; Borchers, C. H. In Methods in Molecular Biology; Backvall, H., Ed.; Humana Press, 2012, in press. (12) Kuzyk, M. A.; Smith, D.; Yang, J.; Cross, T. J.; Jackson, A. M.; Hardie, D. B.; Anderson, N. L.; Borchers, C. H. Mol. Cell. Proteomics 2009, 8, 1860−1877. (13) Groseclose, M. R.; Andersson, M.; Hardesty, W. M.; Caprioli, R. M. J. Mass Spectrom. 2007, 42, 254−262. (14) Campagnoni, A. T.; Campagnoni, C. In In Myelin biology and disorders; Lazzarini, R. A., Griffin, J. W., Lassman, H., Nave, K.-A., Miller, R. H., Trapp, B. D., Eds.; Elsevier Academic Press: San Diego, 2004; pp 387−400. (15) Kamholz, J.; Toffenetti, J.; Lazzarini, R. A. J. Neurosci. Res. 1988, 21, 62−70. (16) Roth, H. J.; Kronquist, K. E.; Kerlero de Rosbo, N.; Crandall, B. F.; Campagnoni, A. T. J. Neurosci. Res. 1987, 17, 321−328. (17) Kaletaş, B. K.; van der Wiel, I. M.; Stauber, J.; Lennard, J. D.; Güzel, C.; Kros, J. M.; Luider, T. M.; Heeren, R. M. Proteomics 2009, 9, 2622−2633. (18) Mascot www.matrixscience.com (Accessed October 3, 2011). (19) Perkins, D. N.; Pappin, D. J.; Creasy, D. M.; Cottrell, J. S. Electrophoresis 1999, 20, 3551−3567. (20) Altschul, S. F.; Gish, W.; Miller, W.; Myers, E. W.; Lipman, D. J. J. Mol. Biol. 1990, 215, 403−410. (21) Cheng, B.; Tong, H. Stochastic Process. their Appl. 1993, 48, 157−174. (22) Crecelius, A. C.; Cornett, D. S.; Caprioli, R. M.; Williams, B.; Dawant, B. M.; Bodenheimer, B. J. Am. Soc. Mass Spectrom. 2005, 16, 1093−1099. 3522


Quantitation of Spatially-Localized Proteins in Tissue Samples using

Recommend Documents