MALDI Mass Spectrometry for Direct Tissue Analysis: A New Tool for

---

MALDI Mass Spectrometry for Direct Tissue Analysis: A New Tool for Biomarker Discovery† Michelle L. Reyzer and Richard M. Caprioli* Mass Spectrometry Research Center, Vanderbilt University School of Medicine, Nashville, Tennessee 37232 Received April 7, 2005

The direct analysis of tissue sections by MALDI mass spectrometry holds tremendous potential for biomarker discovery. This technology routinely allows many hundreds of proteins to be detected over a mass range of ∼2000-70 000 Da while maintaining the spatial localization of the proteins detected. This technology has been applied to a wide range of tissue samples, including human glioma tissue and human lung tumor tissue. In many cases, biostatistical analyses of the resulting protein profiles revealed patterns that correlated with disease state and/or clinical endpoints. This work serves as a review of recent applications and summarizes the current state of technology. Keywords: MALDI MS • imaging • tissues • comparative proteomics

Introduction Matrix-assisted laser desorption/ionization (MALDI)1,2 mass spectrometry (MS) is a powerful discovery tool that has been widely applied to the study of biological systems. MALDI and electrospray ionization (ESI) allow the advantages of mass spectrometric analysis, including high sensitivity, high mass accuracy, and rapid analysis time, to be applied to high molecular weight species, including peptides and proteins. More recently, the use of MALDI mass spectrometry as an imaging tool has further allowed the technology to be applied to a broad array of samples, facilitating the direct analysis of tissue sections and whole cells.3-5 While traditional protein analysis methods generally require homogenization of whole tissues and purification of the protein(s) of interest, direct tissue analysis requires much less sample manipulation and maintains the spatial integrity of the sample. Thus, a single thin (∼5-20 µm) section of a frozen tissue sample may be interrogated not only for what proteins may be present but also for where they are present, without prior knowledge of the specific proteins being analyzed. This approach has led to improvements in the ability to perform comparative proteomic experiments as well as in the ability to generate high-resolution images with molecular specificity. Protein profiles generated from a tissue section reflect the overall status of the tissue. Thus analyses of tissues in different states (i.e., cancerous vs normal, drug treated vs nontreated, grade 4 tumor vs grade 2 tumor, etc.), may reveal differences in the expression of proteins that could not be predicted beforehand. From the full complement of changes in the protein profiles between disease states, a panel of biomarkers may be discovered which holds diagnostic and/or prognostic value. Additionally, any of the individual proteins that are demonstrated to have differential expression may hold insight into the mechanism of action of the disease or treatment. †

Part of the Biomarkers special issue.

1138

Journal of Proteome Research 2005, 4, 1138-1142

Published on Web 07/06/2005

The direct analysis of tissue sections by MALDI MS holds tremendous potential. This technology routinely allows many hundreds of proteins to be detected over a mass range of ∼2000-70 000 Da. To date, analyses have been performed on a diverse set of tissues including mouse brain,5,6 mouse transgenic breast tumor,7 mouse colon,4 mouse epididymis,8 human glioma xenografts,5 human lung tumor,9 and human glioma.6,10 In many cases, biostatistical analyses of the resulting protein profiles revealed patterns that correlated with disease state, patient prognosis, and drug resistance. Several of these applications will be discussed in detail below. Sample Preparation. While generally simpler than traditional techniques that require homogenization, careful preparation of the sample for direct tissue analysis is required so that spatial and molecular integrity is maintained. A tutorial describing the practical aspects of sample preparation for direct tissue analysis has been published, so only a brief description will be given here.11 Dissection of the tissue should be done in a way that preserves the native shape of the tissue. Important reference points (such as the tumor margins) should be recorded as well. The tissue should be frozen in liquid nitrogen immediately after dissection and may remain frozen at -80 °C with no significant degradation for at least a year. Tissue sectioning is performed in a cryostat at approximately -10 to -30 °C, depending upon the composition of the tissue (fatty tissues tend to require lower temperatures). Thin tissue sections, ∼5-20 µm, are obtained and thaw-mounted onto either metal or conductive glass MALDI target plates. The glass target plates are useful for performing histological staining, using MS-compatible stains, and MALDI MS analysis on the same section. A comparison of histological staining techniques and their compatibility with MALDI MS analysis was recently published, and it was reported that cresyl violet and methylene blue were suitable both for morphological discrimination and protein profiling.6 Sections on gold plates are not as amenable to visualization by histological staining. In most cases where staining is necessary for morphological identification, serial 10.1021/pr050095+ CCC: $30.25

 2005 American Chemical Society

MALDI MS for Direct Tissue Analysis

Figure 1. General scheme for profiling and imaging MS of mammalian tissue sections. (Reproduced from ref 13 with permission).

sections are obtained on conventional glass slides and stained as needed (for example, with hematoxylin and eosin). The tissue sections are then placed in a desiccator for up to 24 h prior to matrix application. The MALDI matrix used in MS experiments is typically a small organic molecule that absorbs light at the wavelength of the laser and facilitates desorption and ionization of the analyte. The matrix is usually prepared as a solution and then deposited on the sample, where it cocrystallizes with the analyte. For direct tissue analysis, to be effective a matrix solution must: (1) effectively extract the analytes from the tissue, (2) successfully cocrystallize with the analyte, (3) form crystals over all areas of a tissue surface, and (4) efficiently promote ionization of the analytes. There are over 20 MALDI matrixes commercially available and theoretically an infinite combination of matrix, solvent mixture, concentration, and additive (acid) can be used to obtain MALDI spectra. In practice, sinapinic acid (20 mg/ mL in 50:50 acetonitrile:0.1% TFA in water) tends to give the best combination of crystal coverage and signal quality for direct tissue protein analysis.11 The matrix deposition technique depends on the type of experiment being performed. These experiments can be broadly classified as profiling experiments and imaging experiments, and a general scheme is shown in Figure 1. In general, profiling experiments consist of analyzing several spots per tissue section from large numbers of samples in order to make comparisons between groups. Thus, for profiling experiments, discrete drops of small volumes of matrix (∼pL to nL) are applied to specific morphological areas on each tissue section, and each spot is analyzed by the mass spectrometer to generate a protein profile. On the other hand, imaging experiments are run on one or two prototypical samples in order to obtain highresolution two-dimensional images showing the localization of

reviews the proteins detected by the mass spectrometer. For imaging experiments, the matrix needs to be homogeneously distributed over the entire tissue section, and mass spectra are acquired in a raster pattern across the tissue surface. Coating the tissue surface with matrix in a homogeneous fashion is typically accomplished by utilizing a hand-held glass nebulizer to spray coat the tissue surface. It should be noted that this technique generates mass spectra from the tissue, where each signal corresponds to a unique peptide or protein. While the pattern of protein expression and/ or the protein localization patterns may hold diagnostic potential, individual proteins cannot be conclusively identified by molecular weight alone. Additional analyses are required to identify proteins of interest. Well-established methods exist for protein identification and generally include tissue homogenization, fractionation, purification by either liquid chromatography or gel electrophoresis, enzymatic digestion, peptide mass fingerprinting, and MS sequencing, followed by database searching. In addition, new mass spectrometry-based tools, including tandem MS/MS on TOF/TOF instruments allow some proteins to be identified based on sequence information generated directly from the signal (m/z) of interest. Many of the signals that have been shown to be discriminatory in tissue analyses have subsequently been identified using these techniques. The resulting identifications may further be confirmed by traditional biochemical techniques, including immunohistochemistry and Western blotting. Profiling Tissues by MALDI Mass Spectrometry. Initial experiments have shown that protein profiles obtained from comparable morphological regions on a tissue sample are quite similar, both in terms of the number and relative intensities of signals obtained. Conversely, protein profiles obtained from tissue regions displaying different morphologies reflect these differences. For example, the study by Schwartz et al.10 illustrates this using sections from resected human brain tumor tissue. As shown in Figure 2A, two serial sections of tissue from a single, homogeneous anaplastic astrocytoma tumor were acquired and spotted with sinapinic acid. Four individual spectra obtained from different spots on the two sections show remarkable similarity over the mass range shown. Another tissue section was obtained that had two distinct morphological regions: cortex tissue with infiltrative tumor cells surrounded by normal white matter. Spectra obtained from both areas of normal white matter again showed significant similarity, while the spectrum obtained from the cortex was visibly different (Figure 2B). Thus, the spectra are not only reflective of the different protein compositions present in different morphological areas of the tissue, they are representative of the pathological state of the tissue. Interrogating these profiles further may lead to diagnostic and/or prognostic information not readily available by other means. For example, lung cancer is the leading cause of death in the United States for both men and women. While progress has been made in understanding the causes and biology underlying this cancer, few advances have been made in improving the outcome of patients. MALDI MS imaging technology was thus evaluated for its ability to classify tumor subsets in nonsmall-cell lung cancer (NSCLC) as well as its ability to provide prognostic information.9 In this study, a set of 79 lung tumors, including adenocarcinoma, squamous-cell carcinoma, large-cell carcinoma, and metastases to the lung from other sites, was evaluated, along with a set of 14 normal lung tissues. Protein profiles were obtained from frozen tissue Journal of Proteome Research • Vol. 4, No. 4, 2005 1139

reviews

Reyzer and Caprioli

Figure 3. Mass spectral analysis of MMTV/HER2 sections from tumors harvested 20 h after a variable single dose of OSI-774 reveals several proteomic changes induced from the 100 mg/kg dose, including down-regulation of thymosin β 4 (m/z 4965) and ubiquitin (m/z 8565) and up-regulation of m/z 4794. (Reproduced from ref 7 with permission).

Figure 2. (A) Multiple sites analyzed from a single, homogeneous tumor (anaplastic astrocytoma) specimen; spectra a-c are from one section, d is from the middle portion of the tumor section, 12 µm deeper. As shown, the spectra from 4 distinct areas on 2 unique sections of the tumor tissue display a high degree of homogeneity. (B) MALDI MS analysis of brain with infiltrative tumor cells. A definite shift from a homogeneous, normal white matter (WM) pattern (A), to brain with infiltrative tumor cells in a glioblastoma (B), back to histologically normal white matter (C) can be observed. (Reproduced from ref 10, with permission).

sections as described above. The protein profiles were then subjected to statistical analysis, with approximately half the samples used to generate a class-prediction model. Using 82 differentially expressed MS signals, this model was able to classify with 100% accuracy normal lung (n ) 8) vs all lung tumor (n ) 42). Similarly, 100% classification between normal lung vs primary NSCLC and primary NSCLC vs nonprimary lung tumor was achieved using 91 and 23 differentially expressed signals, respectively. The model was equally effective at classifying subsets of NSCLC (adenocarcinoma vs squamouscell, adenocarcinoma vs large-cell, and squamous-cell vs largecell). The class-prediction model was then applied to a blinded test cohort, and was able to correctly classify samples with 100% accuracy as normal lung vs lung tumor, normal lung vs primary NSCLC, and primary NSCLC vs nonprimary lung tumor. Further subclassification of the tumors was also achieved with >94% accuracy. A significant advantage to a molecular analysis of tissue samples is that classifications may be made that are difficult or impossible to perform by histology or other types of analyses. For example, the ability to distinguish primary NSCLC from 1140

Journal of Proteome Research • Vol. 4, No. 4, 2005

lung metastases is often difficult in clinical practice, yet the prediction model based on the molecular profiles was able to perfectly differentiate 32 primary tumors from 5 metastatic tumors in the testing cohort. Further, protein profiles were used to distinguish patterns associated with survival trends. The authors identified a proteomic pattern comprised of 15 distinct MS signals that divided patients with primary resected NSCLC into a group with poor prognosis (median survival 6 months, n ) 25) and a group with good prognosis (median survival 33 months, n ) 41, p < 0.0001). This association held even after adjustment for other prognostic indicators, including nodal status, stage, and grade. Another application of the technology is the assessment of therapeutic efficacy. Genomic and proteomic advances have led to a diverse array of targeted therapies for diseases. Receptor tyrosine kinase inhibitors (RTKI’s) are a promising group of anti-cancer compounds that specifically inhibit the action of various receptor tyrosine kinases, including the epidermal growth factor receptor (EGFR), vascular endothelial growth factor receptor (VEGFR), and platelet derived growth factor receptor (PDGFR).12 These RTK’s are overexpressed in many different cancers and therefore present a reasonable target for therapy. One concern with many of these compounds is that while they appear to be useful, their effectiveness is limited to a subset of the targeted population. Currently, there is no straightforward way to distinguish in advance which patients will be receptive to a targeted therapy and which will obtain no benefit. In a recent study utilizing a mouse model of breast cancer, we have shown that protein profiles display significant differences in tumor tissue from mice that have received a physiological dose (100 mg/kg) of Tarceva (erlotinib), compared to tumor tissue from untreated mice or mice that had received a nonphysiological dose (10 or 30 mg/kg).7 These protein changes correlate with physiological measures of therapeutic efficacy, including significant reduction in tumor volume and a significant increase in tumor cell apoptosis at the physiological dose. Figure 3 shows two selected regions of the acquired mass range that display differential protein expression. As shown, thymosin β 4 (m/z 4965) and ubiquitin (m/z 8565) are significantly decreased in the 100 mg/kg dosed mice, while the signal at

MALDI MS for Direct Tissue Analysis

reviews

Figure 5. High-resolution imaging MS of mouse cauda epididymis. (a) Optical image of a 12 µm section mounted on a sample plate prior to matrix application. (b-d) Ion density maps obtained for m/z 4965 (thymosin β 4), m/z 6227, and m/z 26 830 (CRISP-1). (Reproduced from ref 8 with permission).

Figure 4. Imaging MS of a mouse brain with a tumor. (A) Optical image of the section before matrix application. The area that contains the tumor is outlined in red. (B-L) Ion density maps obtained at different m/z values with an imaging resolution of 110 µm. The ion density maps are depicted as pseudocolor images with white representing the highest protein concentration and black the lowest. (Reproduced from ref 13 with permission).

m/z 4794 is significantly increased. These protein profiles were generated 20 h after a single oral dose of Tarceva. Early proteomic changes were also discovered that correlated with the synergistic effect of Tarceva in combination with Herceptin, and that correlated with the resistance of certain tumors to Herceptin.7 Imaging Tissues by MALDI Mass Spectrometry. One of the most compelling facets of direct tissue analysis is the ability to spatially localize proteins over a tissue surface. Examining the protein distribution in a given tissue may help uncover complex biochemical mechanisms, for example by revealing proteins that co-localize in the same area. This highlights the discovery aspect of tissue imaging, because proteins which may be of interest need not be known in advance. One area in which this imaging technology has been employed is in the analysis of cancer. Not only have protein profiles been analyzed for potential diagnostic and prognostic biomarkers, but images of protein distributions in cancer tissue have been acquired from tissues including mouse models of brain cancer13 and breast cancer;7 and resected tumor tissue or biopsies from human glioma,14 breast cancer, and sarcoma patients. For example, a recently published review of the technology highlighted images generated from a mouse brain that developed a tumor in the right hemisphere two weeks after the injection of GL261 cancer cells.13 The images generated are reproduced here in Figure 4. As shown, the distribution of 11 different proteins in the 12 µm thick coronal section are quite different. However, a subset of proteins, including acyl CoAbinding protein (m/z 9910), histone H4 (m/z 11 307 and m/z

11 348), histone H2B1 (m/z 13 804), histone H3 (m/z 15 350), and two unidentified proteins at m/z 6924 and m/z 24 807, are localized in the tumor tissue compared to the surrounding normal brain tissue. Other proteins, such as cytochrome C (m/z 12 134) are present throughout the brain tissue, while an unidentified protein at m/z 18 412 is localized to the corpus callosum. These images are only a fraction of the images that could be presented, because any protein that was detected in the mass spectra can be plotted as a two-dimensional image. Thus, one can readily assess a large number of proteins to determine which molecules localize primarily in the tumor areas. These proteins are natural targets for identification and further biochemical analysis, and will hopefully generate insight into tumor growth and/or development in the future. Another application of the imaging technology has been the study of mammalian physiology and development. For example, an integrated proteomics approach was used to investigate the mouse epididymis, a long tubular organ where spermatozoa mature and complete their differentiation by interacting with different proteins present in the epididymal fluid.8 Protein profiling, protein identification, immunohistochemistry, and in situ hybridization were utilized in addition to imaging mass spectrometry to further elucidate the patterns of protein expression along the length of the epididymal tubule. High resolution MS images (50 µm spot-to-spot distance) were acquired for two regions of the epididymis, the caput and the cauda. The images acquired for the cauda are reproduced in Figure 5. Two proteins, one unidentified at m/z 6227 and one at m/z 26 830 identified as cysteine-rich secretory protein 1 (CRISP-1), are shown to be localized within the epididymal tubule. This is in contrast to thymosin β 4 (m/z 4965), which is shown to be localized within the cells surrounding the epididymal tubule and not within the tubule itself. The localization of CRISP-1 in the cauda was also examined by immunohistochemical staining and in situ hybridization. Both techniques localized CRISP-1 within the epididymal tubule, which is in agreement with the MS image. This not only validates MS imaging as a viable approach to visualizing protein Journal of Proteome Research • Vol. 4, No. 4, 2005 1141

reviews localization, it also demonstrates how various biochemical techniques can complement each other. A new dimension, literally, being explored for mass spectrometric imaging is the development of three-dimensional images. Analogous to the mouse and rat brain atlases that exist based on histology,15 a molecular atlas detailing the localization of proteins throughout a mouse or rat brain would be of great interest. Correlating proteomic data with anatomical structures would provide a more comprehensive understanding of healthy and pathological brain functions. This is a challenging endeavor, but preliminary work focusing on the corpus callosum of a mouse brain as a model has been encouraging.16 Optical images of 264 consecutive 20 µm thick sections spanning the corpus callosum (Bregma -3.34 to 1.94 mm) were acquired from a single mouse brain. Ten of these sections, in 400-500 µm steps throughout the tissue volume, were subjected to MALDI MS analysis followed by Nissl staining to reference the sections to a stereotaxic coordinate system.15 These MS images were then registered to each other and to stained serial sections. A computer program was used to generate a 3D volume for the 264 optical images, and the ten MALDI images for a given protein were inserted into the 3D volume. Thus for any signal of interest, its localization in the corpus callosum could be represented in three dimensions. This methodology could also be used to correlate protein distributions with physiological and structural information observed by in vivo imaging techniques, such as computer tomography (CT), positron emission tomography (PET) and magnetic resonance imaging (MRI). Examining changes to local protein distributions caused by diseases, such as cancer or Alzheimer’s disease, with both invasive and noninvasive imaging techniques, will provide a more informative and complete understanding of these diseases.

Summary The examples given throughout this review demonstrate the usefulness of MALDI mass spectrometry as a powerful tool for biomarker discovery. However, the full potential of this technique has yet to be attained. Many of these initial studies need to be reproduced in sample sets large enough to give statistical power to the results. Additionally, the technology is continuing to evolve and improve over time. Higher-frequency lasers allow for faster generation of data, and improvements in both sample preparation and instrumentation will enlarge the subset of proteins currently accessible. Advances in bioinformatics will enable the acquisition of increasingly meaningful information from the data obtained. As these advances are made, it will only strengthen the ability of MALDI mass spectrometry to generate meaningful biomarkers with translational clinical potential.

1142

Journal of Proteome Research • Vol. 4, No. 4, 2005

Reyzer and Caprioli

References (1) Karas, M.; Bachmann, D.; Bahr, U.; Hillenkamp, F. Matrix-assisted ultraviolet laser desorption of nonvolatile compounds. Int. J. Mass Spectrom. Ion Proc. 1987, 78, 53-68. (2) Karas, M.; Hillenkamp, F. Laser desorption ionization of proteins with molecular masses exceeding 10,000 daltons. Anal. Chem. 1988, 60(20), 2299-2301. (3) Caprioli, R. M.; Farmer, T. B.; Gile, J. Molecular Imaging of Biological Samples: Localization of Peptides and Proteins Using MALDI-TOF MS. Anal. Chem. 1997, 69(23), 4751-4760. (4) Chaurand, P.; Stoeckli, M.; Caprioli, R. M. Direct Profiling of Proteins in Biological Tissue Sections by MALDI Mass Spectrometry. Anal. Chem. 1999, 71(23), 5263-5270. (5) Stoeckli, M.; Chaurand, P.; Hallahan, D. E.; Caprioli, R. M. Imaging mass spectrometry: A new technology for the analysis of protein expression in mammalian tissues. Nat. Med. 2001, 7(4), 493496. (6) Chaurand, P.; Schwartz, S. A.; Billheimer, D.; Xu, B. J.; Crecelius, A.; Caprioli, R. M. Integrating Histology and Imaging Mass Spectrometry. Anal. Chem. 2004, 76, 1145-1155. (7) Reyzer, M. L.; Caldwell, R. L.; Dugger, T. C.; Forbes, J. T.; Ritter, C. A.; Guix, M.; Arteaga C. L.; Caprioli, R. M. Early Changes in Protein Expression Detected by Mass Spectrometry Predict Tumor Response to Molecular Therapeutics. Cancer Res. 2004, 64, 9093-9100. (8) Chaurand, P.; Fouchecourt, S.; DaGue, B. B.; Xu, B. J.; Reyzer, M. L.; Orgebin-Crist, M.-C.; Caprioli, R. M. Profiling and imaging proteins in the mouse epididymis by imaging mass spectrometry. Proteomics 2003, 3, 2221-2239. (9) Yanagisawa, K.; Shyr, Y.; Xu, B. J.; Massion, P. P.; Larsen, P. H.; White, B. C.; Roberts, J. R.; Edgerton, M.; Gonzalez, A.; Nadaf, S.; Moore, J. H.; Caprioli, R. M.; Carbone, D. P. Proteomic patterns of tumour subsets in nonsmall-cell lung cancer. Lancet 2003, 362, 433-439. (10) Schwartz, S. A.; Weil, R. J.; Johnson, M. D.; Toms, S. A.; Caprioli, R. M. Protein Profiling in Brain Tumors Using Mass Spectrometry: Feasibility of a New Technique for the Analysis of Protein Expression. Clin. Cancer Res. 2004, 10, 981-987. (11) Schwartz, S. A.; Reyzer, M. L.; Caprioli, R. M. Direct tissue analysis using matrix-assisted laser desorption/ionization mass spectrometry: practical aspects of sample preparation. J. Mass Spectrom. 2003, 38, 699-708. (12) Laird, A. D.; Cherrington, J. M. Small molecule tyrosine kinase inhibitors: clinical development of anticancer agents. Expert Opin. Invest. Drugs 2003, 12(1), 51-64. (13) Chaurand, P.; Schwartz, S. A.; Caprioli, R. M. Profiling and Imaging Proteins in Tissue Sections by MS. Anal. Chem. 2004, 86A-93A. (14) Chaurand, P.; Schwartz, S. A.; Reyzer, M. L.; Caprioli, R. M. Imaging Mass Spectrometry: Principles and Potentials. Toxicol. Path. 2005, 33, 92-101. (15) Paxinos, G.; Franklin, K. B. J. The Mouse Brain in Stereotaxic Coordinates; Academic Press: New York, 2001. (16) Crecelius, A. C.; Williams, B.; Cornett, D. S.; Li, X.; Dawant, B. M.; Bodenheimer, R. E.; Lepage, M.; Niermann, K. J.; Caprioli, R. M. Reconstructing and Visualizing Protein Distributions in 3-D by MALDI Imaging Mass Spectrometry. Proceedings of the 52nd ASMS Conference on Mass Spectrometry and Allied Topics, 2004. Nashville, TN.

PR050095+