High-Resolution, 2- and 3-Dimensional Imaging of Uncut, Unembedded Tissue Biopsy Samples (original) (raw)

Automated histology laboratory instrumentation has significantly improved the ability of pathology laboratories to process tissue samples, particularly biopsy samples, in a relatively rapid and consistent manner. These efforts have also reduced somewhat the dependence on skilled histology personnel and improved the quality of diagnostic material. Similarly, with all its limitations, the current evolution of slide-scanning technology has begun to make remote viewing and digital storage of tissue samples a reality. However, despite the continuing long-term success of paraffin-embedded, microtome-cut, hematoxylin-eosin (H&E)–stained slices for routine pathologic evaluation, there are aspects of these tried-and-true techniques that limit our ability to make more significant advances in the speed, quality, and completeness of tissue biopsy evaluation.

Many alternative tissue processing and imaging approaches have been proposed to address limitations of traditional processing techniques. More recent ones include high-resolution x-ray computed tomography1,2 and optical coherence tomography.3,4 They have the advantages of being applicable to unprocessed fresh tissue and allowing complete 3-dimensional visual examination while leaving tissue unaltered and amenable to further characterization. While efforts in these fields continue, neither technique is able to produce images of sufficient resolution and contrast for adequate routine pathology evaluation.

Multiphoton microscopy (MPM), on the other hand, has the ability to provide images with excellent cellular detail and is a popular, powerful method for analysis of research samples. Use of short-pulse laser light also permits concurrent mapping of second-harmonic generation (SHG), making it possible to simultaneously produce quantifiable images of repeating asymmetric protein structures such as collagen and amyloid. Unfortunately, although the long wavelengths used in MPM can image deeper into tissue than confocal microscopy, traditional methods can only achieve clear images at depths of at most 50 μm with formalin-fixed specimens. Previous attempts to use MPM for imaging through fixed tissue have used serial sectioning5 or serial tissue ablation,6 both of which result in the destruction of the tissue specimen during the course of imaging, making them nonviable for routine clinical use.

As we and others have recently reported in animal tissue,7–9 this significant limitation as it pertains to clinical evaluation of biopsy samples can be overcome by combining MPM with optical clearing. Optical clearing is a method that increases the depth and clarity of MPM imaging by replacing water with a solvent with higher refractive index, matching closely that of proteins and organelles, thus drastically reducing light scattering and enabling multiphoton imaging depths of millimeters instead of micrometers. Traditional pathology processing typically uses a partial clearing step in that xylene, which is primarily used for its ability to solubilize molten paraffin wax, has a higher refractive index than water and partially clears tissue. Several other clearing agents have been used in tissue, with variable success. Our use of benzyl alcohol/benzyl benzoate (BABB) is based on prior demonstration of its superior clearing ability as a result of its high index of refraction, low cost, and overall safety.8

In this article, we demonstrate the potential for clinical use of multiphoton microscopy, combined with simultaneous measurement of SHG, on clarified human tissue specimens with and without nuclear staining. Our goal was to illustrate how digital, H&E-quality, high-resolution images of entire specimens can be obtained in uncut, unembedded human biopsy samples while maintaining the ability to further process tissue with traditional methods.

METHODS

Tissue Clearing and Staining

Human tissue specimens were obtained from discarded pathologic tissue of liver, kidney, breast, and prostate resections under approval from the Yale University Institutional Review Board (New Haven, Connecticut) for human investigation. Samples had been fixed in 4% formaldehyde solution before clearing for a variable period of time ranging from hours to days. Random tissue sections of approximately 1 cm × 5 mm × 2 mm were rinsed with phosphate-buffered saline and then dehydrated via a phosphate-buffered saline/methanol gradient of increasing methanol concentration: 50%, 70%, 95%, and 100%. Incubations were 30 minutes each and performed at room temperature. Optical clearing was performed by following previously described protocols.7,8 Briefly, dehydrated tissue specimens were incubated at room temperature in 24-hour steps, first with 50% methanol and 50% BABB in a 1:2 ratio and then with 100% BABB. Total processing time was 50 hours. For nuclear staining, specimens were incubated with SYTOX Green (Molecular Probes, Eugene, Oregon) at 10 nM during the methanol gradient steps and the first BABB/methanol step.

Imaging

Multiphoton images, including intrinsic fluorescence, nuclear fluorescent staining, and SHG, were obtained by using a custom home-built microscope based on a tunable 80-MHz–pulsed Ti:Sapphire laser (Mai Tai, Spectra-Physics, Mountain View, California), a 3-axis motorized microscope stage (ASI Imaging, Eugene, Oregon), and an Olympus BX51 upright microscope head fitted with an ×5 Nikon objective with a numerical aperture of 0.5 (AZ-Plan Fluor 5x, Nikon Corp, Tokyo, Japan).

Both intrinsic and nucleic acid dye fluorescence were generated by using 740-nm incident light with a pulse width of 100 fsec. A 500-nm wavelength dichroic mirror separated intrinsic from exogenous fluorescence, both detected by using photomultiplier tubes (H7422, Hamamatsu, Bridgewater, New Jersey). The microscope head incorporates a modified optical collection filter box to accommodate the photomultiplier tubes. Second-harmonic generation was collected in transmission by using a 370/20 bandpass filter (Chroma Technologies, Rockingham, Vermont). An adjustable 3-axis mount (New Focus, Santa Clara, California) was used to manually position the SHG photomultiplier tube (Hamamatsu HC-125-02).

Control and image collection were performed with the use of ScanImage software (Howard Hughes Medical Institute, Janelia Farm Research Campus, Ashburn, Virginia).10 Focusing was done at 512×512 resolution with 1 millisecond per line scan times, giving a frame rate of approximately 2 frames per second. Image resolution at collection was 2056×2056 or 1024×1024 at a zoom factor of 1 to 6, depending on desired magnification. Between 4 and 8 frames were averaged per slice for a total acquisition time of 20 seconds per slice. Incident laser intensity was manually adjusted via a Pockels cell in the excitation pathway. Stacks were collected in 1- to 5-μm steps at 16-bit depth and processed by using ImageJ software (developed at the National Institutes of Health, Bethesda, Maryland). Total imaging time for 1-mm cube reconstructions was approximately 6 hours. Postimage processing involved conversion to 8-bit, image inversion, manual global contrast adjustment using the built-in “brightness/contrast” plug-in on a random sample section, and application of the built-in “smooth” function.

Pseudocoloring was performed by inverting the matrix conversion process presented by Ruifrok and Johnston.11 Briefly, intensity values from intrinsic fluorescence and nuclear stains were converted to optical densities in red, green, and blue channels according to the published matrix values for H&E by using MATLAB (MathWorks, Natick, Massachusetts). Intrinsic fluorescence intensity was assigned to the eosin channel (E) while nucleic acid stains were assigned to the hematoxylin channel (H). Following image normalization and scaling to achieve adequate contrast, the red (R), green (G), and blue (B) channel values for the combined pseudocolored images were calculated as follows:

Traditional Histology

Paraffin embedding, sectioning, and H&E staining were performed by using established techniques with a Tissue-Tek VIP tissue processor (Sakura, Torrance, California). Immunohistochemical stains for cytokeratin (CK) 7 and CK20 used commercial antibodies and were performed with standard commercial immunohistochemistry equipment (Dako, Glostrup, Denmark).

RESULTS

Clearing allowed imaging with excellent cellular and nuclear resolution of SYTOX Green or acridine orange–stained specimens more than 500 μm deep into formalin-fixed human prostate, liver, breast, and kidney samples (Figures 1, a through d; 2, a through d). Multiphoton images showed readily identifiable features, comparable to cut slices of H&E-stained tissue, that were amenable to visual pathologic diagnosis without additional morphology training. Normal prostatic glandular structure was readily visualized with adequate nuclear detail (Figure 2, a). Similarly, high-power views of liver tissue produced recognizable chromatin patterns and cytoplasmic detail (Figure 2, b). Breast virtual sectioning showed distinguishable tubular and glandular organization (Figure 2, c). Kidney samples showed clear glomerular structure and visible nuclear and cellular detail in adjacent tubules (Figure 2, d). Clearing was most complete in less cellularly dense tissues such as breast and prostate, but with the BABB clearing protocol even kidney and liver cleared sufficiently to show good morphologic detail 1 mm deep into formalin-fixed tissue.

Figure 1. Clearing example. Formalin-fixed tissue sections of breast and liver before (a and b, respectively) and after (c and d, respectively) a benzyl alcohol/benzyl benzoate clearing protocol. Note near-complete transparency of breast tissue specimen and translucency of liver specimen with some remaining pigment. Grid line spacing is 0.9 cm.Figure 1. Clearing example. Formalin-fixed tissue sections of breast and liver before (a and b, respectively) and after (c and d, respectively) a benzyl alcohol/benzyl benzoate clearing protocol. Note near-complete transparency of breast tissue specimen and translucency of liver specimen with some remaining pigment. Grid line spacing is 0.9 cm.Figure 1. Clearing example. Formalin-fixed tissue sections of breast and liver before (a and b, respectively) and after (c and d, respectively) a benzyl alcohol/benzyl benzoate clearing protocol. Note near-complete transparency of breast tissue specimen and translucency of liver specimen with some remaining pigment. Grid line spacing is 0.9 cm.

Figure 1. Clearing example. Formalin-fixed tissue sections of breast and liver before (a and b, respectively) and after (c and d, respectively) a benzyl alcohol/benzyl benzoate clearing protocol. Note near-complete transparency of breast tissue specimen and translucency of liver specimen with some remaining pigment. Grid line spacing is 0.9 cm.

Citation: Archives of Pathology and Laboratory Medicine 138, 3; 10.5858/arpa.2013-0094-OA

Figure 2. Multiphoton microscopy images of clarified normal human tissue. Specimens were stained either with SYTOX Green or acridine orange nucleic acid dyes during dehydration steps. a, Prostate, medium power. b, Liver, high power. c, Breast, medium power. d, Kidney, medium power. Images are from depths ranging from 200 to 500 μm. Morphologic detail was comparable at 1 mm in depth.Figure 2. Multiphoton microscopy images of clarified normal human tissue. Specimens were stained either with SYTOX Green or acridine orange nucleic acid dyes during dehydration steps. a, Prostate, medium power. b, Liver, high power. c, Breast, medium power. d, Kidney, medium power. Images are from depths ranging from 200 to 500 μm. Morphologic detail was comparable at 1 mm in depth.Figure 2. Multiphoton microscopy images of clarified normal human tissue. Specimens were stained either with SYTOX Green or acridine orange nucleic acid dyes during dehydration steps. a, Prostate, medium power. b, Liver, high power. c, Breast, medium power. d, Kidney, medium power. Images are from depths ranging from 200 to 500 μm. Morphologic detail was comparable at 1 mm in depth.

Figure 2. Multiphoton microscopy images of clarified normal human tissue. Specimens were stained either with SYTOX Green or acridine orange nucleic acid dyes during dehydration steps. a, Prostate, medium power. b, Liver, high power. c, Breast, medium power. d, Kidney, medium power. Images are from depths ranging from 200 to 500 μm. Morphologic detail was comparable at 1 mm in depth.

Citation: Archives of Pathology and Laboratory Medicine 138, 3; 10.5858/arpa.2013-0094-OA

Multiphoton laser excitation and use of a fluorescent nuclear dye also allowed isolation of cytoplasmic, nuclear, and collagen components of the specimens. For example, as illustrated with kidney images presented in Figure 3, intrinsic fluorescence corresponded to the cellular structure and allowed clear evaluation of the glomerular vasculature and tubular cellular organization (Figure 3, a). The nucleic acid stain channel allowed independent evaluation of nuclei (Figure 3, b). And combining these 2 channels with the SHG by intervening collagen strands (pseudocolored in red) allowed clear visualization of the low-level collagen banding that is present in normal human kidney (Figure 3, c). It was also possible to replicate H&E-type coloration on fluorescently stained sample images obtained with MPM (Figure 4). The multiple channels could be individually matched to corresponding hues that mimic the effect of H&E.

Figure 3. Multichannel data for a kidney section. a, Intrinsic fluorescence dominated by signal from cell cytoplasm. b, Inverted nucleic acid fluorescence channel highlighting predominantly nuclear DNA and some cytoplasmic RNA. c, Combined intrinsic fluorescence and nuclear fluorescence (gray scale) with second-harmonic generation channel in red showing distribution of collagen fibers around a normal glomerulus.Figure 3. Multichannel data for a kidney section. a, Intrinsic fluorescence dominated by signal from cell cytoplasm. b, Inverted nucleic acid fluorescence channel highlighting predominantly nuclear DNA and some cytoplasmic RNA. c, Combined intrinsic fluorescence and nuclear fluorescence (gray scale) with second-harmonic generation channel in red showing distribution of collagen fibers around a normal glomerulus.Figure 3. Multichannel data for a kidney section. a, Intrinsic fluorescence dominated by signal from cell cytoplasm. b, Inverted nucleic acid fluorescence channel highlighting predominantly nuclear DNA and some cytoplasmic RNA. c, Combined intrinsic fluorescence and nuclear fluorescence (gray scale) with second-harmonic generation channel in red showing distribution of collagen fibers around a normal glomerulus.

Figure 3. Multichannel data for a kidney section. a, Intrinsic fluorescence dominated by signal from cell cytoplasm. b, Inverted nucleic acid fluorescence channel highlighting predominantly nuclear DNA and some cytoplasmic RNA. c, Combined intrinsic fluorescence and nuclear fluorescence (gray scale) with second-harmonic generation channel in red showing distribution of collagen fibers around a normal glomerulus.

Citation: Archives of Pathology and Laboratory Medicine 138, 3; 10.5858/arpa.2013-0094-OA

Figure 4. Demonstration of pseudocolorization. Prostate section obtained with multiphoton microscopy on cleared tissue with SYTOX Green at depth of approximately 500 μm (same as in Figure 2), processed to mimic hematoxylin-eosin section.Figure 4. Demonstration of pseudocolorization. Prostate section obtained with multiphoton microscopy on cleared tissue with SYTOX Green at depth of approximately 500 μm (same as in Figure 2), processed to mimic hematoxylin-eosin section.Figure 4. Demonstration of pseudocolorization. Prostate section obtained with multiphoton microscopy on cleared tissue with SYTOX Green at depth of approximately 500 μm (same as in Figure 2), processed to mimic hematoxylin-eosin section.

Figure 4. Demonstration of pseudocolorization. Prostate section obtained with multiphoton microscopy on cleared tissue with SYTOX Green at depth of approximately 500 μm (same as in Figure 2), processed to mimic hematoxylin-eosin section.

Citation: Archives of Pathology and Laboratory Medicine 138, 3; 10.5858/arpa.2013-0094-OA

Clearing and fluorescent staining did not have any detectable effect on the subsequent paraffin embedding, sectioning, and H&E staining of the tissues. The same specimens shown in Figure 2 were further processed by traditional histologic techniques and showed no identifiable morphologic adverse effects (Figure 5, a through d). In addition, immunohistochemical staining of kidney tissue for CK7 and CK20 showed the expected specificity of CK7 for descending medullary renal tubules without binding of CK20 (Figure 6, a and b). Thus, the sensitivity and specificity of these antibodies are clearly maintained after the use of BABB as a clearing agent and SYTOX Green or acridine orange as a fluorescent nuclear stain.

Figure 5. Hematoxylin-eosin (H&E)–stained images post multiphoton microscopy (MPM) of clarified tissue. Sample sections from the same specimens depicted in Figure 2, including prostate (a), liver (b), breast (c), and kidney (d), show no perceptible degradation or other visual change with traditional wax embedding, cutting, and H&E staining after clarification of tissue and MPM imaging (original magnifications ×20 [a, c, and d] and ×50 [b]).Figure 5. Hematoxylin-eosin (H&E)–stained images post multiphoton microscopy (MPM) of clarified tissue. Sample sections from the same specimens depicted in Figure 2, including prostate (a), liver (b), breast (c), and kidney (d), show no perceptible degradation or other visual change with traditional wax embedding, cutting, and H&E staining after clarification of tissue and MPM imaging (original magnifications ×20 [a, c, and d] and ×50 [b]).Figure 5. Hematoxylin-eosin (H&E)–stained images post multiphoton microscopy (MPM) of clarified tissue. Sample sections from the same specimens depicted in Figure 2, including prostate (a), liver (b), breast (c), and kidney (d), show no perceptible degradation or other visual change with traditional wax embedding, cutting, and H&E staining after clarification of tissue and MPM imaging (original magnifications ×20 [a, c, and d] and ×50 [b]).

Figure 5. Hematoxylin-eosin (H&E)–stained images post multiphoton microscopy (MPM) of clarified tissue. Sample sections from the same specimens depicted in Figure 2, including prostate (a), liver (b), breast (c), and kidney (d), show no perceptible degradation or other visual change with traditional wax embedding, cutting, and H&E staining after clarification of tissue and MPM imaging (original magnifications ×20 [a, c, and d] and ×50 [b]).

Citation: Archives of Pathology and Laboratory Medicine 138, 3; 10.5858/arpa.2013-0094-OA

Figure 6. Example of compatibility of benzyl alcohol/benzyl benzoate clearing and SYTOX Green staining with traditional immunohistochemistry on human renal tissue. a, Cytokeratin (CK) 7 stain of normal kidney showing expected pattern of transition to positive staining on descending medullary cords. b, Appropriate lack of staining of same renal tissue with CK20 (original magnifications ×4 [a and b]).Figure 6. Example of compatibility of benzyl alcohol/benzyl benzoate clearing and SYTOX Green staining with traditional immunohistochemistry on human renal tissue. a, Cytokeratin (CK) 7 stain of normal kidney showing expected pattern of transition to positive staining on descending medullary cords. b, Appropriate lack of staining of same renal tissue with CK20 (original magnifications ×4 [a and b]).Figure 6. Example of compatibility of benzyl alcohol/benzyl benzoate clearing and SYTOX Green staining with traditional immunohistochemistry on human renal tissue. a, Cytokeratin (CK) 7 stain of normal kidney showing expected pattern of transition to positive staining on descending medullary cords. b, Appropriate lack of staining of same renal tissue with CK20 (original magnifications ×4 [a and b]).

Figure 6. Example of compatibility of benzyl alcohol/benzyl benzoate clearing and SYTOX Green staining with traditional immunohistochemistry on human renal tissue. a, Cytokeratin (CK) 7 stain of normal kidney showing expected pattern of transition to positive staining on descending medullary cords. b, Appropriate lack of staining of same renal tissue with CK20 (original magnifications ×4 [a and b]).

Citation: Archives of Pathology and Laboratory Medicine 138, 3; 10.5858/arpa.2013-0094-OA

The acquisition of digital images also allowed 3-dimensional reconstruction of 1-mm-thick blocks of tissue. Full-scale 3-D reconstructions of intrinsic fluorescence of liver provide a more complete perspective on normal tissue growth, as illustrated in the liver reconstruction presented in Figure 7, a. The potential for accurate evaluation of neoplastic growth margins is apparent. Nuclear fluorescence scans taken every 4 μm also allowed visualization of the arborizing structure of breast glands as noted in Figure 7, b. The transparency of collagen fibers and fat to the nuclear dye wavelengths facilitated these large set reconstructions, which could be easily rotated and manipulated with the ImageJ 3-D volume-rendering plug-in on a standard 64-bit laptop computer. In addition, 3-dimensional reconstructions of SHG signal in liver and kidney (Figure 7, c) demonstrate the ability to perform complete specimen quantitative analysis of fibrosis without the need for additional tissue processing. As expected, SHG was brighter toward the portion of the tissue distal to laser excitation (closest to detector), but produced resolved collagen detail throughout 1-mm-thick tissue.

Figure 7. Representative large block 3-D reconstructions of normal human tissue. a, Approximately 1-mm cubic section of normal human liver obtained by multiphoton microscopy on cleared tissue without staining (intrinsic fluorescence only, low power). b, Similar sized block of normal human breast tissue, which has been fixed, cleared, and stained with the nucleic acid dye SYTOX Green (low power). c, Perspective image of 3-D reconstruction of collagen from normal human kidney (approximately 200×200×40 μm).Figure 7. Representative large block 3-D reconstructions of normal human tissue. a, Approximately 1-mm cubic section of normal human liver obtained by multiphoton microscopy on cleared tissue without staining (intrinsic fluorescence only, low power). b, Similar sized block of normal human breast tissue, which has been fixed, cleared, and stained with the nucleic acid dye SYTOX Green (low power). c, Perspective image of 3-D reconstruction of collagen from normal human kidney (approximately 200×200×40 μm).Figure 7. Representative large block 3-D reconstructions of normal human tissue. a, Approximately 1-mm cubic section of normal human liver obtained by multiphoton microscopy on cleared tissue without staining (intrinsic fluorescence only, low power). b, Similar sized block of normal human breast tissue, which has been fixed, cleared, and stained with the nucleic acid dye SYTOX Green (low power). c, Perspective image of 3-D reconstruction of collagen from normal human kidney (approximately 200×200×40 μm).

Figure 7. Representative large block 3-D reconstructions of normal human tissue. a, Approximately 1-mm cubic section of normal human liver obtained by multiphoton microscopy on cleared tissue without staining (intrinsic fluorescence only, low power). b, Similar sized block of normal human breast tissue, which has been fixed, cleared, and stained with the nucleic acid dye SYTOX Green (low power). c, Perspective image of 3-D reconstruction of collagen from normal human kidney (approximately 200×200×40 μm).

Citation: Archives of Pathology and Laboratory Medicine 138, 3; 10.5858/arpa.2013-0094-OA

COMMENT

Traditional techniques of fixation with formalin with physical wax embedding and microtome sectioning for histology have been successfully used in routine pathology evaluation for more than a century. Part of their success can be attributed to the ease of use and forgiving nature of formalin fixation, coupled with the compatibility of wax embedding with a range of simple and inexpensive staining techniques. Other important factors for the continued success of formalin-fixed, wax-embedded slides have been the long-term preservation that formalin fixation affords and the cumulative experience of pathologists, which increases the accuracy and consistency of interpretation.

Nonetheless, there remain considerable limitations associated with current specimen processing methods and the evaluation of these by pathologists. For biopsies, these include the limited amount of tissue that is typically directly visualized, a function of both the desire to preserve tissue for ancillary testing and the time required to inspect multiple slides. Not infrequently, additional tissue evaluation is needed, but requests for recuts and levels delay diagnoses. Also, they usually still result in sampling only a portion of the tissue while reducing tissue availability for increasingly important immunostaining and molecular analysis. In addition, imperfect embedding results in tissue waste and can hamper interpretation, and the cutting process itself produces artifacts that may hinder evaluation. Imaging of unembedded and uncut tissue addresses these traditional processing limitations. It provides the opportunity to visualize entire biopsy specimens, reducing the likelihood of missing important features owing to incomplete sampling, and to preserve tissue for ancillary tests.

Other advantages of analyzing uncut, unembedded specimens relate to savings in time and effort. Embedding, cutting, and staining are some of the most time-consuming and manual steps in tissue processing,12–14 requiring personnel with significant expertise. While automation and microwave-based tissue processing have allowed some sites to begin to offer same-day diagnosis for some biopsy samples, the postdehydration and clearing steps are an impediment in satisfying an increasing need on the part of providers and patients for rapid turnaround of morphologic evaluation.

The approach we present here allows complete visualization of biopsy-sized specimens without the need for the postclearing steps. Thus, the potential for considerably reducing the time from biopsy to morphologic assessment exists. We anticipate that cleared biopsy specimens could be either provided to pathologists for direct visualization or completely scanned for image distribution. Primary diagnosis could potentially be rendered on these images and subsequent studies ordered if needed.

Principal hurdles that need to be overcome before such a system becomes practical are the cost of instrumentation, the speed of clearing, and the speed of scanning. However, all 3 of these barriers are imminently surmountable with the rapid advances in the field of computing and optics. For instance, multiphoton imaging approaches exist that incorporate the use of inexpensive diode lasers, significantly reducing the overall cost of an MPM system to that of a modern motorized microscope.15,16 Alternatives to galvanometer-based single-point laser raster scanning—including multibeam scanning systems,17 spatiotemporal multiplexing,18 and temporal focusing19,20—promise to improve scanning times by an order of magnitude or more, making real-time video imaging a possibility. Finally, there are several potential approaches to speed up the otherwise diffusion-limited process of clearing, including the use of established microwave technology for histologic processing. The clearing process is not reaction dependent and is akin to dehydration, such that a total processing time for biopsy specimens of less than half an hour, similar to current rapid processing methods, is theoretically possible. Also, since there is no cutting, and tissue clearing does not intrinsically require fixation, even more rapid processing protocols may be attainable. High-refractive-index, water-soluble agents introduced into cells with low-level fixation, as used for frozen sections, is one potential approach.

Another aspect of interest in tissue biopsy evaluation that we explored is the visualization of 3-dimensional structure. Diagnostic aspects that may benefit from a 3-D perspective include identification of low-grade abnormalities in glandular cell growth, such as with prostate and breast neoplasia; evaluation of depth of invasion of tumors, such as for determining depth of muscle invasion in bladder biopsies; and the more complete quantitative evaluation of fibrosis, of particular significance in kidney and liver biopsies. Previous attempts at creating large 3-D data sets from tissue have used methods producing poor contrast, poor depth penetration, or that successively remove tissue as the 3-D volume is imaged.3–6 Thus, past 3-D reconstruction techniques fail to improve upon the most important limitations of traditional histology. But 3-D reconstructions of clarified tissue using MPM showed excellent cellular contrast; sufficient depth, such that entire biopsy specimens could be imaged; and compatibility with subsequent traditional processing, including preservation of immunostain capability with the few antibodies tested. The prospect of nondestructive 3-D reconstruction can open the door to needed comprehensive studies designed to better evaluate the benefit of 3-D visualization in a variety of clinical settings.

We consider these latter points particularly advantageous for the MPM/clearing method. A critical barrier to clinical adoption of new imaging technologies is resistance from pathologists who have spent years honing their skills on a specific set of specialty stains and the desire to maintain full compatibility with established clinical practice. The MPM/clearing approach has produced images that have resolution and fields of view similar to those of current routine practice and provide contrast similar to that obtained with commonly used histologic stains, but that allow subsequent traditional processing without apparent adverse effects. The multichannel method presented here also allowed straightforward pseudocolorization that represents morphology in a way analogous to traditional stains, allowing pathologists to easily recognize salient histologic features. In our estimation, alternative approaches to current methods that do not present visual images analogous to those obtained with current traditional stains or that do not allow simultaneous or subsequent access to morphologic, immunohistochemical, or molecular data currently accepted as the gold standard face a significant uphill battle in becoming universally accepted for routine clinical care.

In summary, the combination of clearing agents and fluorescent dyes for clinical application of multiphoton imaging of complete biopsy specimens, along with added informational content from SHG, shows great promise. Excellent cellular contrast is achievable from both endogenous fluorescence and with extrinsic nucleic acid dyes. Multichannel imaging facilitated a pseudocolorization process that mimics the appearance of traditional stains. Three-dimensional reconstructions of MPM imaging from clarified tissue are possible on complete biopsy-sized tissue specimens and can also be used to produce quantifiable characterization of collagen fibrosis. The preservation of traditional histology and compatibility with immunostaining, if confirmed with follow-up studies, reduces considerably the risk of adopting this new technology. While studies that demonstrate a specific advantage of complete imaging of uncut, unembedded biopsy specimens over conventional processing are needed, the results obtained thus far merit further development of technologies that can facilitate the practical implementation of this approach in clinical practice.

Financial support for this research comes in part from National Science Foundation CAREER Award DBI-0953902.

References