Pathologists examining tissue samples judge the status of cells by their morphologic appearance and, in the case of immuno-histopathology (IHP), by the presence and arrangement of single specific molecules that act as markers for disease. Proper evaluation of these molecules requires reliable, reproducible staining and the pathologist's expertise.
Our approach uses spectral histopathology (SHP),1 where we record spatially resolved vibrational spectra of a tissue (from the colon or bladder, for example) using either an IR2 or a Raman microscope.3 A vibrational spectrum reflects the biochemical status of a sample based on IR and Raman active molecular vibrations, creating a characteristic biochemical fingerprint for the cell status. Therefore, we can gather cumulative and localized information about proteins (the proteome), metabolized molecules (the metabolome), and nucleotides (the genome).
SHP is compatible with standard pathology procedures. Applicable to both freshly frozen or paraffin-fixated/deparaffinized tissue slices, a 5–10μm-thick section is represented by more than 10 million recorded IR spectra, to produce a single image of 15×15mm2 tissue. To build a reference database, we applied computational image analysis to cluster the spectra by similarity.4 Expert pathologists examined the resulting false-color images for spatial similarity with the classically stained slice, obtained after spectral acquisition. Thus, SHP procedures are label- and marker-free. Combining pathology, biophysics, and bioinformatics, we set up a reliable algorithm using a random forest (a classification method using decision trees). Initially, we set up databases for the annotation of colorectal and urinary bladder tissue classes.
We allocated each newly acquired spectrum to the tissue types in the reference database. Every type can be represented by a specific color, for example, red for carcinoma and white for muscle (see Figure 1). Combining all types, we obtained a highly accurate, spatially resolved, and annotated image of the section. For colorectal tissue, we obtained a specific discrimination of 14 tissue classes, including cancerous lesions and morphologically intact pre-cancer states (see Figure 1).2
Figure 1. Spectral histopathology: Fourier transform IR (FTIR) and Raman imaging analysis of colorectal carcinoma tissue sections (colon tumor). Image magnification increases from the top down. A classic hematoxylin-eosin (H&E)–stained section is shown on the right. FTIR imaging (left and center) achieves a resolution of 5μm. Spectral histopathology reveals alterations in the crypts of the colon (red), which are confirmed by an immunohistochemical fluorescent stain (P53). Raman imaging depicts the crypt alterations in more detail (bottom image). Altered nuclei (gray) in the altered crypts (red) were identified in comparison with the P53 stain (green).
Our group also compared SHP with IHP methods, in which tumors were identified using fluorescent labels.2, 3 The SHP identified cancerous areas widely colocalized with p53-activated cells (those subject to an immunohistochemical fluorescent stain: see Figure 1). Sensitivity and specificity of the IR SHP on colorectal tissues exceeded 95%.
Microscopy with medium-IR radiation is resolution-limited by the applied wavelengths (4000–600cm−1, corresponding to 2.5–16.7μm). Using Raman microspectroscopy, we were able to overcome this limit, to obtain a deeper understanding of the affected cells. The enhanced resolution enabled automatic recognition of single cells and intracellular structures, based on specific biochemical signatures,3 including tumor characteristic nuclei. This technique has potential applications in monitoring distribution of a drug within a single cell.
Analysis of thin tissue slices requires considerable time for preparation, scanning, and evaluation of the millions of spectra, and in the operating theater, time is crucial. Modern developments in fiber-optic light guides enable flexible positioning of a probe head on tissues of interest, and there are designs for dedicated IR and Raman probes for sample surface analysis. Raman probes for depth profiling, which are currently under development, would enable a true minimally invasive analysis of deep tissue areas within a minimal time frame.
Nonetheless, analysis of resected tissue is an invasive procedure, which causes patient discomfort (in colonoscopy for colorectal biopsies, or cystoscopy for biopsies of bladder tissue, for example). Applying spectroscopic analysis to body fluids, by drawing a blood sample, for instance, would be a far less invasive diagnostic process. Unfortunately, spectroscopic body fluid analysis is limited to only one spectrum per sample to represent the patient's health status. Furthermore, the disease-associated changes are usually well masked in the biological variance of the fluid itself, so that reproducible spectrum acquisition is challenging to achieve. The key to a successful identification of those subtle marker band contributions is an automated Fourier transform IR (FTIR) spectroscopic system, which also avoids the user accidentally influencing the data prior to classification.5Using this system and dedicated bioinformatics, we identified characteristic marker patterns for an accurate, sensitive, and specific discrimination of urinary bladder cancer patients within a risk group.6 Therefore, a vibrational spectroscopic analysis of body fluids may provide minimally invasive, fast, and cost-efficient diagnostics for specific medical questions.
In conclusion, SHP with vibrational biospectroscopy is a promising label-free automated diagnostic technique. The method offers applications for pathology, surgery, and endoscopy. Applied microscopically, SHP annotations of diseased tissues and cells are very detailed and precise. A far less invasive body fluid analysis, when combined with fiber optics, provides early and immediate information about a patient's status, demonstrating a versatile, easy-to-use, objective, and reproducible diagnostic tool for personalized medicine. In future, we aim to bring vibrational spectroscopy into clinical application.
Klaus Gerwert, Frederik Großeruschkamp, Julian Ollesch
Department of Biophysics
Klaus Gerwert received his PhD in 1985 from the University of Freiburg, Germany. He worked at the Max Planck Institute in Dortmund from 1986 until 1989, and was a Heisenberg fellow from 1990 to 1993. He then became chair of the Department of Biophysics at Ruhr-University Bochum, and was appointed as a Max Planck fellow in 2008. He investigates mechanisms of membrane proteins using FTIR, and vibrational imaging of tissues and cells as diagnostic tools.
Frederik Großerüschkamp received his PhD at Ruhr-University Bochum in 2013. His studies include the behavior of water in the hydrolysis reaction of small GTPase Ras proteins with time-resolved FTIR. He also conducts data preprocessing and analysis for FTIR-imaging projects.
Julian Ollesch received his PhD from Ruhr-University Bochum in 2006. In 2010, after postdoctoral fellowships in the United States and at Halle/Saale, Germany, he joined the Protein Research Unit Ruhr within Europe in the Department of Biophysics. He investigates spectral disease markers for cancer and neurodegenerative conditions.
1. B. Bird, M. S. Miljković, S. Remiszewski, A. Akalin, M. Kon, M. Diem, Infrared spectral histopathology (SHP): a novel diagnostic tool for the accurate classification of lung cancer, Lab. Invest.
92(9), p. 1358-1373, 2012. doi:10.1038/labinvest.2012.101
2. A. Kallenbach-Thieltges, F. Großerüschkamp, A. Mosig, M. Diem, A. Tannapfel, K. Gerwert, Immunohistochemistry, histopathology, and infrared spectral histopathology of colon cancer tissue sections, J. Biophoton.
6(1), p. 88-100, 2013. doi:10.1002/jbio.201200132
3. L. Mavarani, D. Petersen, S. F. El-Mashtoly, A. Mosig, A. Tannapfel, C. Kötting, K. Gerwert, Spectral histopathology of colon cancer tissue sections by Raman imaging with 532 nm excitation provides label free annotation of lymphocytes, erythrocytes, and proliferating nuclei of cancer cells, The Analyst
138(14), p. 4035-4039, 2013. doi:10.1039/c3an00370a
4. Q. Zhong, C. Yang, F. Großerüschkamp, A. Kallenbach-Thieltges, P. Serocka, K. Gerwert, A. Mosig, Similarity maps and hierarchical clustering for annotating FT-IR spectral images, BMC Bioinform.
14(1), p. 333, 2013. doi:10.1186/1471-2105-14-333
5. J. Ollesch, S. L. Drees, H. M. Heise, T. Behrens, T. Brüning, K. Gerwert, FTIR spectroscopy of biofluids revisited: an automated approach to spectral biomarker identification, Analyst
138(14), p. 4092-4102, 2013. doi:10.1039/c3an00337j
6. J. Ollesch, M. Heinze, H. M. Heise, T. Behrens, T. Brüning, K. Gerwert, It's in your blood: spectral biomarker candidates for urinary bladder cancer from automated FTIR spectroscopy, J. Biophoton. (Accepted.)