Raman Chemical Imaging: Histopathology of Inclusions in Human

a 50 MHz 80486-based computer (Gateway 2000) is used for Raman image collection and processing. ..... Citation data is made available by participa...
1 downloads 0 Views 218KB Size
Anal. Chem. 1996, 68, 1829-1833

Raman Chemical Imaging: Histopathology of Inclusions in Human Breast Tissue Michael D. Schaeberle,† Victor F. Kalasinsky,‡ James L. Luke,‡ E. Neil Lewis,§ Ira W. Levin,§ and Patrick J. Treado*,†

Department of Chemistry, University of Pittsburgh, Pittsburgh, Pennsylvania 15260, Department of Environmental and Toxicologic Pathology, Armed Forces Institute of Pathology, Washington, D.C. 20306-6000, and Laboratory of Chemical Physics, National Institute of Diabetes, Digestive and Kidney Diseases, National Institutes of Health, Bethesda, Maryland 20892

High-definition Raman chemical imaging microscopy is applied to the histopathological characterization of biopsied human breast tissue containing foreign polymer inclusions. The polymer material is found in a patient with a history of silicone implant breast reconstructive surgery. Silicone implants are, on occasion, anchored to the soft tissues of the chest with polyester patches. In the case studied here, the polyester anchors were incorporated into the patient’s fibrous tissue surrounding the implant. High-definition Raman chemical imaging provides molecular-specific image contrast in the complex breast tissue matrix without the use of stains or dyes. This is the first example in which Raman spectroscopic imaging microscopy has been applied to pathology. A goal of this and future studies is to complement traditional histopathologic diagnoses of disease states utilizing vibrational spectroscopic imaging techniques. Pathology is employed in biomedicine to identify and characterize the disease state of human tissue, using well-established optical microscopy techniques.1 Often, disease state determination involves the analysis of thinly sectioned, stained biopsied tissue in order to visualize various disease-specific pathologic markers. Although these examinations are typically made by highly trained practitioners, the potential exists to make the determinations more quantitative and less reliant on subjective observations by integrating the efficacy of vibrational spectroscopy with optical microscopy. This article describes the first application of Raman chemical imaging microscopy to pathology studies of foreign inclusions in human breast capsular tissue. The field of chemical imaging is an emerging area of chemistry which combines analytical spectroscopic methods such as Raman, infrared (IR), or fluorescence spectroscopy with high-definition imaging. These techniques are capable of rapidly probing a sample’s chemical heterogeneity at high spatial resolution with definitive molecular specificity.2 Our use of Raman microscopy to image foreign matter in human tissue is the first step in a broader research effort. We †

University of Pittsburgh. Armed Forces Institute of Pathology. § National Institute of Diabetes, Digestive and Kidney Diseases. (1) Lillie, R. D.; Fullmer, H. M. Histopathologic Technic and Practical Histochemistry, 4th ed.; McGraw-Hill: New York, 1976. (2) Treado, P. J.; Morris, M. D. In Spectroscopic and Microscopic Imaging of the Chemical State; Morris, M. D., Ed.; Marcel Dekker: New York, 1992; Chapter 3. ‡

S0003-2700(95)01245-5 CCC: $12.00

© 1996 American Chemical Society

are investigating the use of Raman and IR chemical imaging for natural disease state determination and, ultimately, for in vivo use. Raman spectroscopy has been successfully demonstrated in a variety of biological and medical applications,3-13 including the identification of inclusions in lung tissue,4 the examination of colored particles in cancerous organs,5 the characterization of breast tissue,6-8 the biocompatibility of prosthetic materials,9 and the detection of implant-related silicone in lymph nodes.10,11 Raman imaging has been employed in the study of human white blood cells,12 while confocal Raman imaging has been used to investigate the spatial distribution of antitumor drugs in living cancer cells.13 The utility of visible absorption microscopy and fluorescence microscopy in biological analysis is well established.14,15 However, potentially invasive dyes must be employed to generate the specificity and sensitivity that traditional pathology imaging techniques enjoy. Thus, it is desirable to obtain spatially resolved chemical information without the use of dyes, tags, or stains in order to reduce sample preparation time, or eliminate sample preparation entirely, and to minimize sample modifications. The reduction of sample degradation by eliminating staining is anticipated to be an essential step in performing quantitative analyses on living cells (either in vivo or in vitro). Raman chemical imaging microscopy is a noninvasive technique because it can provide image contrast without the use of stains or dyes. Raman microscopy is widely applicable because (3) Ozaki, Y. Appl. Spectrosc. Rev. 1988, 24, 259-312. (4) Buiteveld, H.; De Mul, F. F.; Mud, J.; Greve, J. Appl. Spectrosc. 1984, 38, 304-306. (5) Huong, V.; Plouvier, S. R. J. Mol. Struct. 1984, 115, 489-492. (6) Redd, D. C. B.; Feng, Z. C.; Yue, K. T.; Gansler, T. S. Appl. Spectrosc. 1993, 47, 787-791. (7) Frank, C. J.; Redd D. C. B.; Gansler, T. S. McCreery, R. L. Anal. Chem. 1994, 66, 319-326. (8) Frank, C. J.; McCreery, R. L.; Redd, D. C. B. Anal. Chem. 1995, 67, 777783. (9) Bertoluzza, A.; Fagnano, C.; Tinti, A.; Morelli, M. A.; Tosi, M. R.; Maggi, G.; Marchetti, P. G. J. Raman Spectrosc. 1994, 25, 109-114. (10) Frank, C. J.; McCreery, R. L.; Redd, D. C. B.; Gansler, T. S. Appl. Spectrosc. 1993, 47, 387-390. (11) Abraham, J. L.; Etz, E. S. Science 1979, 206, 716-718. (12) Puppels G. J.; Bakker Schut, T. C.; Sijtsema, N. M.; Grond, M.; Maraboeuf, F.; de Grauw C. G.; Figdor, C. G.; Greve, J. J. Mol. Struct. 1995, 347, 477484. (13) Sharonov, S.; Chourpa, I.; Morjani, H.; Nabiev, I.; Manfait, M.; Feofanov, A. Anal. Chim. Acta 1994, 290, 40-47. (14) Wang, Y. L., Taylor, D. L., Eds. Methods in Cell Biology; Academic Press: New York, 1989; Vol. 29. (15) Ross, M. H.; Reith, E. J. Histology: A Text and Atlas; Harper & Row: New York, 1985; Chapter 1.

Analytical Chemistry, Vol. 68, No. 11, June 1, 1996 1829

image contrast is based on the material’s intrinsic vibrational spectroscopic signature. Although Raman microscopy does not provide sensitivity comparable to that of fluorescence microscopy and lacks the specificity of immunolabeling, the advantage of the technique arises from the ability to correlate the high informational content of Raman spectra with the details of sample composition. In contrast to Raman methods, the more prevalent fluorescence microscopy techniques rely on broad featureless spectra to monitor sample characteristics, including the differentiation between normal and diseased tissue.16 Several approaches to Raman imaging have been demonstrated.2,17-29 Of these, the use of tunable filters, including AOTFs23-27 and LCTFs,20,28,29 holds the most promise. In general, tunable filter methods employ wide field illumination, in conjunction with two-dimensional detection. The two spatial dimensions of the image are recorded directly by a multidimensional camera, while the multispectral information is acquired, under computer control, by capturing images at wavelengths selected by the tunable filter. In this manner, it is possible to create an image data set with a Raman spectrum at each pixel. Image fidelity is limited primarily by the number of pixels in the camera, and the use of high-definition detectors allows the efficient collection of high-definition images. The Raman imaging microscope we have developed uses an acousto-optic tunable filter (AOTF) for wavelength selection. The AOTF is a solid-state device that is capable of functioning from the UV to the mid-IR, depending on the choice of the filter’s crystal material. We employ a TeO2 AOTF optimized for red wavelengths, in conjunction with a krypton ion excitation laser (647.1 nm). AOTFs provide high optical throughput (diffraction efficiencies ∼40%), moderate spectral resolution (50 cm-1), wide tuning range (0.4-5.5 µm for TeO2 crystals), rapid tuning ability (25 µs), random accessibility, and an angular field of view of (5°. AOTF principles are well established and have been described in detail.30-32 In this article, we describe the application of an AOTF Raman chemical imaging system to the study of polymer inclusions in a biopsied section of human breast tissue. The polymer inclusions (16) Andersonengels, S.; Ankerst J.; Johansson J.; Svanberg, K.; Svanberg, S. Photochem. Photobiol. 1993, 57, 978-983. (17) Delhaye, M.; Dhamelincourt, P. J. Raman Spectrosc. 1975, 3, 33-43. (18) Bowden, M.; Gardiner, D. J.; Rice, G.; Gerrard, D. L. J. Raman Spectrosc. 1990, 21, 37-41. (19) Treado, P. J.; Govil, A.; Morris, M. D.; Sternitzke, K. D.; McCreery, R. L. Appl. Spectrosc. 1990, 44, 1270-1275. (20) Christensen, K. A.; Bradley, N. L.; Morris, M. D.; Morrison, R. V. Appl. Spectrosc. 1995, 49, 1120-1125. (21) Puppels, G. J.; Grond, M.; Greve, J. Appl. Spectrosc 1993, 47, 1256-1267. (22) Batchelder, D. N.; Cheng, C.; Pitt, G. D. Adv. Mater. 1991, 3, 566-568. (23) Treado, P. J.; Levin, I. W.; Lewis, E. N. Appl. Spectrosc. 1992, 46, 12111216. (24) Treado, P. J.; Levin, I. W.; Lewis, E. N. Appl. Spectrosc. 1992, 46, 553559. (25) Lewis, E. N.; Treado, P. J.; Levin, I. W. Appl. Spectrosc. 1993, 47, 539543. (26) Schaeberle, M. D.; Turner, J. F., II; Treado, P. J. Proc. SPIE-Int. Soc. Opt. Eng. 1994, 2173, 11-20. (27) Schaeberle, M. D.; Karakatsanis, C. G.; Lau, C. J.; Treado, P. J. Anal. Chem. 1995, 67, 4316-4321. (28) Morris, H. R.; Hoyt, C. C.; Treado, P. J. Appl. Spectrosc. 1994, 48, 857866. (29) Morris, H. R.; Hoyt, C. C.; Miller, P.; Treado, P. J. Submitted to Appl. Spectrosc. (30) Chang, I. C. Appl. Phys. Lett. 1974, 25, 370-372. (31) Goultzoulis, A. P.; Pape, D. R. Design and Fabrication of Acousto-Optic Devices; Marcel Dekker: New York, 1994; Chapter 4. (32) Tran, C. D. Anal. Chem. 1992, 64, 971A-981A.

1830

Analytical Chemistry, Vol. 68, No. 11, June 1, 1996

were believed to be Dacron polyester based on the patient’s medical history and previous Raman and IR data.33 Dacron polyester patches are occasionally employed to attach breast implants to the chest wall during reconstructive and cosmetic surgery. The problems associated with silicone implants, including rupture and release of silicone gel contained in these implants, have been well publicized in the past several years.34,35 The goal of this study is to visualize the distribution of polyester in the thin section prepared from tissue biopsies. No sample preparation beyond conventional microtomy is required for the AOTF Raman imaging study. EXPERIMENTAL SECTION The AOTF Raman chemical imaging system used in this study has been described in detail.23,27,28 Briefly, a 50 MHz 80486-based computer (Gateway 2000) is used for Raman image collection and processing. Camera control is provided by commercial software (Princeton Instruments, WinView 1.3B). Software written in the C programming language (Borland, C++ 3.1) is used to integrate image acquisition and AOTF control. Raman chemical image visualization and processing are performed using Windows software (ChemIcon, ChemImage 1.0) written to store and enhance the 32-bit floating point spectral image file format (SPIFF) data files. For publication, the Raman images are processed on a Silicon Graphics workstation employing multispectral image rendering software (Jet Propulsion Laboratory, LinkWinds 2.0)36 and are printed on a dye sublimation printer (Tektronics Phaser SDX). Raman microprobe spectroscopy and chemical imaging of polymer inclusions in human breast implant capsular tissue were performed on a sectioned (5 µm) biopsy mounted on an aluminumcoated microscope slide. The methodology used for sample analysis has been described previously.27 The Raman chemical image analysis initially consists of optical microscopy inspection of the sample, followed by the use of nonimaging Raman microspectroscopy to determine sample heterogeneity. From the Raman microspectroscopy, spectral bands are identified that are best suited to generate Raman images. Microspectroscopy is followed by preliminary (high spectral resolution/low spatial resolution) AOTF chemical image data set acquisition. Once optimal imaging parameters have been identified, high-fidelity Raman image acquisition is performed, followed by chemical image processing, which has been described previously.27 RESULTS Brightfield microscopy is used in conjunction with Raman microprobing to substantiate the heterogeneous nature of the biopsied section of human breast implant capsular tissue. A brightfield image of the sample is shown in Figure 1. The white box corresponds to the area sampled through the AOTF with a 20× objective. Although clusters of round bodies are observed in the brightfield image, it is not possible using conventional optical microscopy alone to identify the molecular composition of the structures. Raman chemical imaging provides the means to determine, with molecular specificity, the identity and distribution of the structures in question. (33) Kalasinsky, V. F.; Centeno, J. A.; Johnson, F. B.; Luke, J. L. 1994 Pittsburgh Conference, 2\28\94; Paper 294. (34) Kessler, D. A. N. Engl. J. Med. 1992, 326, 1713-1715. (35) Bridges, A. J.; Vasey, F. B. Arch. Intern. Med. 1993, 153, 2638-2643. (36) Jacobson, A. S. LinkWinds 2.0; Jet Propulsion Laboratory, Pasadena, CA.

Figure 1. Brightfield reflectance image of 5 µm thin-section breast implant capsular tissue sample; 10× (NA ) 0.30) objective. The white box corresponds to the area sampled through the AOTF by a 20× (NA ) 0.46) objective.

Raman microspectra of the tissue biopsy acquired with a conventional dispersive spectrometer are shown in Figure 2. The spectrum in Figure 2A is collected by focusing the laser within a suspected inclusion. Features at 1287, 1615, and 1727 cm-1 correlate well with the reference spectrum of Dacron polyester37 and clearly indicate the presence of the polymer in the tissue. The spectrum in Figure 2B is collected from the surrounding tissue and indicates a lack of the polymer. The lipid or peptide features that might be expected from the tissue seem to be masked by the sample’s fluorescent background. Raman microanalysis confirms the heterogeneous nature of the sample and identifies unique vibrational spectroscopic features. As a result, a Raman chemical image data set is acquired by defocusing the laser source and illuminating a region on the sample that matches the AOTF field of view, the limiting aperture of the microscope. An exposure time of 3 min is employed to collect each Raman image frame. Image contrast can readily be observed in traditional Raman imaging experiments. More challenging, however, is differentiating Raman image contrast that has molecular specificity from existing, nonspecific fluorescence. A Raman spectrum associated with each pixel in the image enables a straightforward compositional analysis, even in the presence of a fluorescent background. Figure 3 illustrates this phenomenon in detail. Figure 3, parts A and B, are taken from the unprocessed Raman image data set and appear almost identical. Figure 3A is collected at 1615 cm-1, the Dacron Raman peak maximum, and Figure 3B is collected at 1670 cm-1, a background spectral region. The intensity distribution in Figure (37) Hendra, P. J.; Agbenyega, J. K. The Raman Spectra of Polymers; John Wiley & Sons: New York, 1993; C:66.

Figure 2. In situ Raman microspectra of the Dacron polyester in breast implant tissue; 20× (NA ) 0.46) objective: (A) Dacron polyester tissue inclusions and (B) normal tissue.

3A arises from the sum of the Raman signal and the background, while Figure 3B contains no Raman scattering and arises only Analytical Chemistry, Vol. 68, No. 11, June 1, 1996

1831

Figure 4. AOTF Raman spectra of the biopsy sample showing Raman emission superimposed on a fluorescent background: (A) Raman spectrum of Dacron polyester inclusions and (B) the surrounding tissue. Spectra are collected from the points labeled in Figure 3C.

Figure 3. AOTF Raman images of breast implant capsular tissue; 20× (NA ) 0.46) objective: (A) raw image taken at the 1615 cm-1 peak maximum of the Dacron polyester spectrum; (B) raw background image acquired at 1670 cm-1; (C) ratio of peak maximum to background (A/B). The labels in C correspond to the spectra in Figure 4.

from the fluorescent background of the sample. Image ratioing is employed to enhance the contrast by minimizing the contribution of the fluorescence. Figure 3A is ratioed by Figure 3B to produce the Raman image shown in Figure 3C. Contrast in Figure 3C arises exclusively from the Dacron polyester inclusion Raman signal. In addition to removing the fluorescence contribution from the Raman image, image ratioing minimizes the instrument response and image artifacts that are not related to the sample, such as nonuniform illumination, dust on the optics, and detector nonuniformity. 1832 Analytical Chemistry, Vol. 68, No. 11, June 1, 1996

The images in Figure 3 are extracted from a three-dimensional Raman image data set consisting of 128 pixels × 128 pixels × 87 images. Vector projections at a given pixel (xi,yi) comprise Raman spectra. The spectra are useful in verifying that the contrast observed Figure 3C is due to Raman scattering and not to contrast generation mechanisms, such as fluorescence or Rayleigh scattering. Figure 4 shows representative spectral vectors from the unprocessed image data set. The spectra in Figure 4 are Raman spectra plotted through the inclusion (A) and tissue (B) regions. Figure 4A clearly identifies that Dacron polyester is present in local concentrations within the tissue and is responsible for the contrast observed in Figure 3C. Figure 4B reveals the fluorescence background upon which the Dacron Raman scattering is superimposed. Figure 5 illustrates high-fidelity Raman chemical imaging. A magnified brightfield reflectance image of the biopsied tissue area is shown in Figure 5A. A Raman image of Dacron, having 230 × 230 pixels, is shown in Figure 5B at high magnification. Figure 5B is the ratio of the Dacron image at 1615 cm-1 divided by the background at 1670 cm-1. The integration time is 10 min/image, which produces a significantly higher signal-to-noise ratio than is seen in Figure 3C. The round bodies apparent in the brightfield

integration time used to generate Figure 2), 6 days would be required to map a 230 × 230 pixel image with visual clarity comparable to that shown in Figure 5. Raman images collected with a line scanning approach can be acquired more rapidly than with point scanning, as only 230 line scans of the sample would be required to construct a Raman image having 230 × 230 pixels. However, line scanning still requires considerable time. For example, it would require 7.7 h to construct a 230 × 230 pixel image if each laser line scan position was acquired in 2 min, which is typical.

Figure 5. High-definition images of Dacron polyester in human breast implant capsular tissue: (A) brightfield reflectance image and (B) background ratioed Raman image (1615 cm-1/1670 cm-1); 10 min integration, 20× (NA ) 0.46) objective.

image correlate well with the bright regions in the Raman image. The inclusions are ∼15 µm in diameter. The AOTF approach requires a total of 20 min to provide a high-definition, background-corrected Raman image having 52 900 total pixels. In comparison with more conventional technologies, such as point or line scanning Raman microprobes, the AOTF approach provides a significant time advantage. For example, point scanning would involve focusing the laser to a small spot (∼1 µm diameter) on the sample and rastering it in the X and Y dimensions to construct a high-definition image. Assuming it would take 10 s to acquire each point spectrum (the same

CONCLUSIONS AND FUTURE DIRECTIONS Raman chemical imaging has great potential to impact quantitative pathology determinations by visualizing the spatial distribution of foreign particulate material in pathologic tissue specimens. Specifically, high-fidelity Raman chemical images of Dacron polyester inclusions in human breast implant capsular tissue have been acquired. These analyses provide a unique strategy for characterizing two-dimensional sample morphology, in terms of the size, shape, and distribution of foreign bodies within sectioned tissue. Our initial analysis did not provide a perspective of the threedimensional architecture of the tissue specimens, which could provide an assessment of the mechanism by which the foreign inclusions were incorporated into the breast tissue. Reconstruction of the three-dimensional architecture of this system would be relatively straightforward, by first imaging multiple contiguous thin sections and then digitally reassembling the tissue volume architecture on a computer. In this work, we have used Raman chemical imaging to study objects embedded in human tissue. The same methodology is currently being employed to analyze quantitatively the natural diseased states of various tissues. Since these studies require the identification of unique spectroscopic signatures for disease states, efforts are underway to identify these vibrational spectroscopic markers. Raman chemical imaging is anticipated to play a central role in the successful application of spectroscopic studies to research in pathology. ACKNOWLEDGMENT We acknowledge partial financial support of the work from the Society of Analytical Chemists of Pittsburgh (SACP) and the Central Research and Development Fund, University of Pittsburgh. Received for review December 29, 1995. Accepted March 21, 1996.X AC951245A X

Abstract published in Advance ACS Abstracts, April 15, 1996.

Analytical Chemistry, Vol. 68, No. 11, June 1, 1996

1833