Detection of Pre-Invasive Lung Cancer: Technical Aspects of the LIFE Project

Abstract

Lung cancer is the leading cause of cancer deaths in both men and women in the United States. The LIFE (Light Induced Fluorescence Endoscopy) Project was initiated at the University of North Carolina Medical Center in November, 1999, for the dual purposes of (1) detecting pre-invasive lung cancer in high-risk patients and (2) studying the molecular biology of pre-invasive lesions of the bronchus for possible development of molecular biomarkers. Of the 47 patients enrolled, all were current or former tobacco smokers, except for 1. Fluorescence endoscopy was utilized, in addition to white light bronchoscopy, to increase the detection of intraepithelial lesions. Adjacent biopsies were submitted for permanent and frozen sections, respectively, from four predetermined sites as well as from any abnormal areas. The snap-frozen specimens were cryostat sectioned, and the mucosal epithelial cells laser capture microdissected for DNA analysis. The great majority of specimens yielded sufficiently abundant and intact DNA to accomplish the molecular objectives. Histologic concordance of adjacent permanent and frozen sections was equivalent to the concordance of adjacent permanent sections, suggesting that frozen section diagnosis was adequate for the research purpose of correlating histology with molecular analysis.

Introduction

Lung cancer is the leading cause of cancer deaths in both men and women in the United States. Moreover, the number of deaths from lung cancer is greater than that of deaths from colon, breast, and prostate cancer combined. It is estimated that in the United States in 2006, 31% of all cancer deaths in men and 26% of all cancer deaths in women will be due to lung cancer (Jemal et al., 2006).

Since the reporting of 2 retrospective studies in 1950 (Doll and Hill, 1950; Wynder and Graham, 1950), it has been recognized that cigarette smoking is causally related to lung cancer, and numerous subsequent epidemiologic, pathologic, and experimental studies have convincingly shown that the great majority of lung cancers are secondary to tobacco smoking. Thus the major cause of cancer deaths in this country is preventable, and educational campaigns by both the government and private organizations have been successful in reducing the number of smokers from approximately one half to one quarter of the population. Former smokers remain at risk for the development of lung cancer, however, and thus of the population at risk today, approximately half are current smokers and half are former smokers.

Because of the reduction in number of male smokers, the number of deaths from lung cancer has been declining in males in the United States since the early 1990s. Lung cancer mortality among women, however, has continued to increase since the popularity of cigarette smoking among women peaked about 20 years later than the peak for men. During the period from 1930 to 1997 the death rate from lung cancer in U.S. women increased 600%, and lung cancer moved from the seventh most common to the most common cancer cause of death of women (Patel et al., 2004), surpassing breast cancer for the first time in 1987.

One of the primary reasons that lung cancer is so deadly is that three quarters of the patients already have metastatic spread to regional or distant sites at the time of diagnosis (Mulshine and Sullivan, 2005). Thus, the 5-year survival for all patients presenting with lung cancer is only about 15%. On the other hand, those patients who are discovered with Stage I tumors (tumors limited to the lung) have a 60% chance of living 5 years. For this reason large scale screening programs of high-risk patients have been designed in an effort to identify lung cancer in the early, treatable stage. However, 5 major studies utilizing chest X-ray and/or sputum cytology found that screening with these methods detected small cancers, but did not reduce mortality (Bach et al., 2003).

More recent studies utilizing low-dose computed tomography (CT) have proven more sensitive in identifying Stage I cancers, but with the drawback of a high false positive rate, often leading to anxiety and costly follow-up procedures (Abramowicz, 2005). In 2002 the National Cancer Institute initiated the National Lung Cancer Screening Trial to determine if CT screening with multidetector-row scanners would lead to a significant reduction in mortality associated with lung cancer, and a similar program is also underway in Europe (Mulshine and Sullivan, 2005).

The Life (Light Induced Fluorescence Endoscopy) Project: Utilizing the Advantage of Fluorescence Bronchoscopy

Although radiographic diagnosis of Stage I lung cancer would clearly improve the prognosis, the ultimate goal would be to identify neoplasia in the pre-invasive stage when there is the greatest chance for cure. With this goal in mind the LIFE Project was initiated in November, 1999, as a collaborative effort between the National Institute of Environmental Health Sciences (NIEHS) and the University of North Carolina Medical Center in Chapel Hill (UNC). The dual purposes of this project were to detect pre-invasive lung cancer in high-risk patients and to study the molecular biology of early disease for possible development of molecular biomarkers. The clinical, clinicopathologic, and molecular results of this project are currently undergoing analysis and will be the subject of future reports. In this article we will focus primarily upon the technical aspects of the study, and in particular the methods employed in obtaining and processing of the tissue specimens, as well as the rationale for selection of those methods.

High-Risk Patients

It is known that patients with a strong tobacco smoking history are at 10- to 20-fold greater risk for developing lung cancer than the general population, and that patients with a previous history of lung cancer or upper respiratory cancer are at increased risk for developing a second cancer in the lung. Thus, of the 47 patients recruited from the UNC medical and surgical clinic population for the LIFE Project, all either had a history of previous lung or upper respiratory cancer, or were current or former smokers with at least a 15 pack-year history. In addition, all but 1 of the patients in the group with a history of previously treated respiratory tract cancer were also smokers.

Fluorescence Bronchoscopy

With traditional white-light bronchoscopy, it is estimated that only about 40% of the cases of carcinoma in situ can be detected, and that lesions <5 mm are usually not visible (Travis et al., 2004). In addition, in 1 study (Lam et al., 2000) slightly more than half of the dysplastic lesions measured ≥1.5 mm in greatest dimension (the size of a bronchial biopsy with a standard biopsy forceps), and these lesions usually exhibited no visible abnormality on white-light bronchoscopy. Because of the limitations of white-light bronchoscopy, the Light Induced Fluorescence Endoscope, or LIFE bronchoscope (Xillix Technologies Corp., Richmond, BC, Canada) was developed in the early 1990s in an effort to enhance the sensitivity of detection of dysplasia and carcinoma in situ (Hung et al., 1991). Subsequent studies have reported that the addition of fluorescence bronchoscopy can increase the detection of pre-invasive lesions from 40% to an average of 80% (Lam et al., 2000), and that lesions as small as 0.5 mm can be localized by this method (Travis et al., 2004).

The LIFE technology, also referred to as Laser Induced Fluorescence Endoscopy, exploits the difference in autofluorescent properties between normal mucosa and mucosa altered by dysplasia, carcinoma in situ, or invasive carcinoma (Hung et al., 1991). When bronchial mucosa is exposed to the helium-cadmium laser of the LIFE bronchoscope at a wavelength of 442 nm, the intensity of the autofluorescence emitted by carcinoma in situ lesions is greatly reduced in comparison to normal bronchus tissue, especially in the green region of the spectrum. As a result, normal bronchial mucosa exhibits a bright green autofluorescence while carcinoma in situ, and to a lesser extent dysplastic lesions, exhibit a brownish-red appearance. The chemical or structural basis for the decrease in green autofluorescence by dysplastic or malignant tissues is not known, but it has been suggested that the reduced concentration of riboflavin in tumor tissues might account for the difference in autofluorescent intensity (Hung et al., 1991).

Because of the enhanced sensitivity of fluorescence bronchoscopy, the LIFE bronchoscope was employed routinely in the LIFE project, in conjunction with white-light bronchoscopy.

Collection of Bronchoscopic Biopsies

Each of the 47 patients underwent bronchoscopy as part of the initial screening procedure. In addition, follow-up bronchoscopies were performed in 22 of the patients, either for repeat 2-year screening of those with normal histology at the first bronchoscopy, for detection of progression/regression in patients with dysplasia, or for detection of any residual disease following treatment for carcinoma in situ. Bronchoscopic biopsies were obtained from 4 predetermined sites (main carina, left upper lobe, right upper lobe, and right lower lobe) as well as from any abnormal areas, and bronchial washings were also collected for cytologic examination. Ordinarily, 2 adjacent biopsies were taken from each site with one of these placed in neutral-buffered formalin for histologic diagnosis, and the other snap-frozen for molecular analysis (Figure 1). In abnormal areas more than 2 biopsies were sometimes taken from the same site.

In preparation for snap-freezing of biopsies, a layer of O.C.T. freezing medium (Tissue-Tek, Sakura Finetek, Torrance, CA) was placed in the bottom of a plastic cryomold (Tissue-Tek, Sakura Finetek, Torrance, CA). When the biopsy specimen was obtained, the tissue was transferred with sterile forceps from the biopsy instrument to the O.C.T. in the bottom of the cryomold, and the remainder of the cryomold then filled with O.C.T. In this manner, the tissue was protected on both sides by the O.C.T. and yet was close to 1 surface of the block. It should be noted that an approach to be avoided is one in which the cryomold is largely filled with O.C.T. and the biopsy then inserted with the tweezers into the bottom of the cryomold, as this may lead to trailing of pieces of the specimen through the O.C.T., requiring many sections to examine the entire specimen.

The cryomold containing the biopsy and O.C.T. was then snap-frozen in isopentane within a cryo-preserve tank at −63°C (Shandon Lipshaw Histo Bath 2 Low Temperature Freezing Bath, Thermo Electron Corporation, Pittsburgh, PA). Each frozen cryomold was labeled and individually placed in a plastic zip-lock bag to avoid the possibility of misidentification should one of the frozen blocks pop out of the cryomold. The frozen cryomolds were transported to the laboratory at NIEHS on dry ice and then stored in a −80°C freezer until ready to section.

Laser Capture Microdissection (LCM) of Frozen Specimens

The Use of Frozen Tissue for Optimization of Molecular Analysis

Although DNA can be and has been extracted from formalin-fixed, paraffin-embedded tissue, much of the DNA may be degraded. One of the major mechanisms of formalin-induced DNA degradation appears to be the cross-linking of proteins to DNA (Crisan and Mattson, 1993), which makes the DNA molecule rigid and susceptible to mechanical shearing (Mies, 1994). The extensive cross-linking of proteins to DNA as well as the formation of formalin-induced protein complexes also necessitate a prolonged digestion time with proteinase K, and not all of these chemical modifications may be overcome and reversed (Wu et al., 1990). For polymerase chain reactions (PCR) performed on formalin-fixed tissues, it is therefore recommended that the PCR target be less than 400 base pairs (bp) and optimally between 80 and 170 bp (Shibata, 1994). In addition, one report has noted a high frequency of nonreproducible sequence alterations, or artifactual mutations, in the DNA of formalin-fixed material, while no nonreproducible sequence alterations were noted in frozen tissue from the same tumor (Williams et al., 1999).

To avoid these potential problems with the molecular analysis of formalin-fixed tissue, some of the biopsy specimens were snap-frozen at the time of bronchoscopy since this was felt to be optimal for research purposes and collection could be prospectively arranged. In addition, optimal material for molecular analysis was necessary in view of the anticipated low cellular yield of microdissected samples from small biopsies, and the recognition that more cells per PCR reaction would probably have been needed for DNA recovered from formalin-fixed tissue.

Cryostat Sectioning

The cryostat sectioning and staining protocol was designed to accomplish 3 objectives with the frozen biopsies: (1) obtain sufficient material for microdissection while preserving the DNA integrity, (2) produce hematoxylin and eosin (H&E) stained slides at intermittent levels for diagnostic correlation with molecular studies, and (3) conserve as much of the tissue as possible to serve as a resource for future studies.

During the major part of the study, a Slee Benchtop Cryostat Type MTC (Marston Technical Services, Inc., Cincinnati, Ohio) was used for cutting frozen sections; in the latter part of the project, sectioning was performed on a Leica CM 1850 Cryostat (Leica Microsystems, Inc., Bannockburn, IL). With each cryostat the temperature was maintained between −20°C and −22°C since this range was found to be optimal for these bronchial biopsy tissues. Disposable Sakura Accu-Edge^® Low Profile Blades (Sakura Finetek, Torrance, CA) were used for sectioning, and these blades were kept in the cryostat to achieve the equivalent cryostat temperature.

Several precautions were employed to avoid the possibility of specimen cross-contamination, which would have invalidated the molecular analysis. Before sectioning each day, the inner surfaces of the cryostat, particularly around the cutting area, were cleansed with gauze pads soaked in absolute ethanol. A new disposable cutting blade was inserted for each biopsy, and the anti-roll plate was not utilized, to prevent carryover. Small finepoint brushes were sometimes needed to prevent curling of frozen sections, but the brushes were cleaned with absolute ethanol between biopsies. Non-powdered, disposable gloves were worn at all times.

Because of the length of time required to cut the frozen sections and to examine the diagnostic frozen sections for adequacy, one or more days—depending on the number of biopsies taken from the individual patient—were reserved for this purpose. Two separate H&E staining sets were necessary, one of which was modified for the laser capture microdissection slides and the other designed to yield optimally stained slides for diagnostic purposes. On the morning of sectioning fresh H&E staining solutions were prepared for the diagnostic staining set. The cryomold tissue blocks were transferred from the −80°C freezer to the laboratory in a styrofoam box of dry ice and were kept in the dry ice until ready for cryostat sectioning.

In an effort to produce quality frozen sections for diagnostic purposes as well as optimal sections for LCM, steps were taken to minimize some of the technical problems which are frequently experienced with frozen sections. Initially, when the cryomold with the tissue block was transferred from the dry ice to the cryostat, the cryomold was allowed to equilibrate to the cryostat sectioning temperature (~20 minutes) (Callis, 2004). After removing the tissue block from the cryomold, the block was placed on a chuck containing a small amount of O.C.T. The angle of the cutting blade was maintained between 5 and 6 degrees. Cryostat sections were cut at 4–5 μ, and were picked up on slides as soon as any tissue was noted in the sections being cut in order to conserve tissue.

For the H&E diagnostic slides, only 1 section was placed on a slide and the slide was fixed immediately in Shandon Rapid-Fixx^™ (Thermo Electron Corporation, Pittsburgh, PA) to avoid drying artifact. Immediate fixation was critical to prevent loss of nuclear detail as well as the enlargement of nuclei associated with drying. If 2 sections were placed on 1 slide, for instance, then the first of the 2 sections usually exhibited drying artifact, and sometimes both sections would be suboptimal. On the other hand, because the nuclear detail of the tissue on the uncoverslipped slides used for laser capture microdissection was not of major importance, 2 or 3 sections were picked up on each slide intended for LCM. Thus, the routine that we adopted for our initial sections on these biopsies included cuts 1, 8, 15, and 19 with 1 section on each slide for diagnostic H&E stains, and the cuts in between (2–7, 9–14, and 16–18) with 2 or 3 sections on each slide for LCM purposes. Sections cut for LCM were placed in a slide box on dry ice as they were cut, and then stored in a −80°C freezer until ready for microdissection.

Before removing the tissue block from the specimen holder, the diagnostic H&E slides from the initial 4 levels were examined. If the tissue in these 4 levels appeared to contain enough epithelial cells for molecular analysis, the diagnosis was clear, and there were no atypical epithelial changes that were just beginning to appear, then the block was removed, and sectioning of the next block could begin. However, approximately 50% of the time one of those conditions was not met, and it was necessary to cut additional sections, often up to 30 sections, sometimes as many as 60, and rarely over 100. After removal of each frozen block from the specimen holder, a thin layer of O.C.T. was spread over the cut surface of the block and allowed to freeze to prevent drying of the tissue in the −80°C freezer (Callis, 2004).

Occasionally, additional sections were desired after removing the tissue block from the cryostat and returning the block to the −80°C freezer. Attempts to recut frozen blocks for diagnostic purposes after removal from the specimen holder, whether stored in the −80°C freezer or not, were seldom successful in demonstrating areas of interest because the orientation of the block in the specimen holder could not be precisely returned to that of the original orientation, resulting in trimming of the block with loss of tissue. For this reason, every effort was made in the initial sectioning to obtain the necessary sections of each frozen block prior to removal of the block from the cryostat specimen holder. In addition, diagnostic staining of the unfixed frozen slides that had been cut for LCM purposes also proved of little value because of the drying artifact related to lack of immediate fixation. Thus, the inability to reliably recut the block and the lack of serial sections (tissue ribbons) were clearly 2 of the limiting factors in our attempts to accurately categorize these biopsies for diagnostic purposes.

Laser Capture Microdissection

With the advent of a variety of microdissection instruments during the last 10 years, it has become possible to selectively remove cells of interest from heterogeneous tissues. Consequently, molecular analysis of these pure populations of cells is not only more accurate, but is also more sensitive for detection of nucleic acid alterations because there is little or no dilution by other cell types in the tissue. For instance, it has been reported that the admixture of normal cells in a DNA preparation resulted in homozygously deleted regions appearing as heterozygous (Cairns et al., 1994).

Although the bronchial endoscopic biopsies were relatively thin and superficial, they usually contained underlying submucosal fibrous tissue with fibroblasts and capillaries. Frequently submucosal mucoserous glands were present, and occasionally small fragments of hyaline cartilage were also attached. A few scattered lymphocytes were not uncommonly seen in the fibrous stroma, and in some cases there were inflammatory infiltrates variably composed of lymphocytes, plasma cells, eosinophils, and neutrophils. In light of the coexistence of these cellular elements of the lamina pro-pria and submucosa in the biopsy specimens, laser capture microdissection of the desired surface epithelial cell component was essential to avoid compromise of the molecular studies.

Staining of Slides for Laser Capture Microdissection

Regardless of the type of microdissection instrument used, it is necessary to choose a histochemical stain that will not interfere with molecular procedures such as the polymerase chain reaction. It has been reported, for instance, that tissue sections stained with hematoxylin and eosin (H&E) in a routine manner will yield unsatisfactory PCR amplification (Medintz et al., 1997; Burton et al., 1998). Further investigation has shown that it is the hematoxylin rather than the eosin that inhibits PCR amplification (Serth et al., 2000), possibly due to the dye binding to DNA and interfering with proteinase digestion, or by influencing divalent cation (Mg⁺⁺) concentration important in maintaining Taq polymerase activity (Murase et al., 2000). On the other hand, some investigators have found that when tissue was sampled by LCM rather than by manual scraping of entire sections, and the staining time with Mayer’s hematoxylin was reduced to 30 seconds, no interference was noted with DNA retrieval and amplification (Ehrig et al., 2001; Huang et al., 2002). The explanation advanced for this finding is that the small size of LCM samples per volume of sample lysis buffer results in concentrations of the inhibitory ingredients of hematoxylin that are too low to produce any effect on tissue digestion or DNA amplification (Ehrig et al., 2001; Huang et al., 2002).

For the purposes of this study we elected to use an abbreviated H&E stain for the LCM slides, rather than an alternative such as methyl green, because of our familiarity with H&E staining, both from the technical and the histologic standpoint. In addition, we felt that the color contrast provided by even an abbreviated H&E stain would be helpful with small bronchial biopsy specimens that were often fragmented and frequently denuded with strips of detached epithelial cells lying within mucus or blood clot. However, to reduce the likelihood of adversely affecting either the efficiency of cell transfer from slide to LCM cap or the downstream molecular procedures, we decreased the staining time in Mayer’s hematoxylin to 15 seconds. The complete staining protocol that we employed for our slides prepared for microdissection was: (1) fixation in 70% ethanol, 1 minute; (2) distilled water, 15 seconds; (3) Mayer’s hematoxylin (Sigma-Aldrich, Atlanta, GA, filtered before each use), 15 seconds; (4) 0.5% lithium carbonate in distilled water, 30 seconds; (5) distilled water, 30 seconds; (6) 70% ethanol, 30 seconds; (7) eosin (alcoholic eosin Y with phloxine (Prophet et al., 1994), filtered before each use), 30 seconds; (8) 95% ethanol, 15 seconds; (9) 100% ethanol, 30 seconds; (10) 100% ethanol, 30 seconds; (11) xylene, 1 minute; (12) xylene, 1 minute. The slides were then air-dried for 2–4 minutes. All alcohols and xylenes were changed daily, and a small amount of 4 Å molecular sieves (Sigma-Aldrich, Atlanta, GA) was added to each xylene to reduce hydration with repeated use. The hematoxylin and eosin were changed weekly and filtered daily.

The Arcturus Pixcell Laser Capture Microdissection System

The major portion of the microdissection for this study was performed with the Arcturus Pixcell^® II instrument (Molecular Devices, Arcturus, Sunnyvale, CA). During the final year of the project, the Pixcell II was updated to the Pixcell IIE, which was then used for the remainder of the study. The basic operation of this instrument is relatively straightforward, and there are numerous publications describing both the instrument and its applications (Emmert-Buck et al., 1996; Bonner et al., 1997; Simone et al., 1998). Despite the simplicity of operation of this instrument, however, there are many fine points and potential pitfalls that must be recognized if successful results are to be achieved. At the NIEHS our experience with this technology in a core lab facility since 1997 has enabled us to develop guidelines to cope with some of these issues.

The Importance of Dehydration for Efficient Cell Transfer

The basic principle of the Arcturus LCM instrument is the activation of a focused, short-duration pulse from an infrared laser, producing mild heating and melting of a precise spot on a transfer polymer film, which results in flow of the polymer into the interstices of the targeted cells and bonding of the polymer with those cells, thus permitting the chosen cells to be selectively removed from the rest of the tissue. For cell transfer to occur from the slide to the thermoplastic polymer, the bond between the laser-melted plastic polymer and the targeted cells must be greater than the bond between the cells and the glass slide.

One of the major factors necessary for the adherence and bonding of the polymer to the cells of interest is thorough dehydration or desiccation of the tissue on the slide. As noted in the LCM staining protocol previously described, our slides were routinely passed through 2 containers of xylene, each for 1 minute, at the end of the staining protocol, and then allowed to air dry in the fume hood for 2–4 minutes. This was usually sufficient for desiccation of the tissue on the slide, but on occasion cell transfer was not adequate and a longer xylene rinse (10–30 minutes) and/or prolonged drying of the slide in the fume hood was necessary. Another suspected cause of incomplete desiccation, when multiple slides were being stained, was hydration of the 100% ethanol, requiring preparation of fresh solutions. The reason for the failure of the polymer to bond with incompletely desiccated tissue is uncertain. Some have alluded to the porosity of desiccated tissue and the necessity for the molten polymer to fill the air spaces of the targeted tissue (Goldstein et al., 1998). Alternatively, it may be that the polymer is not miscible with water, or that the high surface tension of water repels the polymer, thus blocking its adherence to, or impregnation of, the tissue.

It is also important that the caps to which the thermoplastic polymer is attached are free of moisture. For this reason, we have routinely maintained our caps in a desiccator until ready for use. Room conditions will also impact cell transfer (Cornea and Mungenast, 2002); we have noted that when the room temperature or humidity are high that poor cell transfer can be expected.

Other Factors Influencing Cell Transfer

While bonding of the polymer with the tissue is essential, it is equally important that adjacent nontargeted tissue not lift off of the slide when the cap is raised at the end of the microdissection session. In order that this does not occur, the bond between the tissue and the slide must be stronger than the tissue-shearing strength (Goldstein et al., 1998). This consideration is said to become increasingly important as the laser spot size is reduced (Goldstein et al., 1998). In our experience, it also becomes an issue when the dimensions of the tissue on the slide are reduced, particularly with small biopsies. Thus, in dealing with these small bronchoscopic biopsies, it became apparent early in the study that the entire biopsy specimen might be pulled off of the slide by the Arcturus cap if the tissue was attached to a regular, uncharged slide. On the other hand, when fully charged slides were used, the bond between the slide and the tissue was so strong that the removal of targeted cells was impeded.

An effective compromise developed by one of the authors (PS) was to open the box of charged slides, and then replace the lid on the box and allow the box of slides to sit on the countertop for 2 to 3 weeks before using. This technique allowed the charge to dissipate partially but not completely, and resulted in successful cell transfer without unintended lifting of the entire biopsy. If the slides were allowed to remain in this condition for a month or more, however, the charge was so dissipated that lifting of nontargeted tissue could not be prevented.

Efficient cell transfer from the slide to the cap also requires a flat tissue surface. If there are folds in the tissue, then the cap cannot make direct contact with the tissue close to the fold, and another site must be chosen for the cell capture. Likewise, the smoothness of the surface of the cap will also affect cell transfer. As more cells are progressively added to the cap during the microdissection, cell capture becomes more difficult because the cap is lifted slightly off of the tissue surface by the captured cells and polymer, and thus cannot make direct contact with the tissue. For this reason, we would routinely start the microdissection with the slides containing the best epithelial cell material, usually from the deeper cuts of the block, to ensure the maximum capture of those cells.

Microdissection Parameters

The microdissection parameters were adjusted as the session would proceed since an increase in power and duration was often necessary for effective transfer because of the elevation of the Arcturus cap above the slide surface as more cells were captured. The laser power used for this study never exceeded 50 mW, and ranged from 25 mW early in the microdissection to 50 mW later in the session. Similarly, the laser duration ranged from 450 μsec for the 7.5 μ laser spot size and 1.5 msec for the 15 μ laser spot size at the beginning of the session to 600 μsec for the 7.5 μ laser spot size and 4 msec for the 15 μ laser spot size by the end of the microdissection.

Since the epithelial cells from multiple slides were collected on one cap, slides containing the best epithelial cell clusters were microdissected first to ensure abundant space on the cap for the best material. Cell transfer was often not as efficient with the 7.5 μ laser spot size, sometimes necessitating double laser pulses for the same cell. However, the same part of the cap was never used for capturing cells from different areas of the tissue, as this will entomb the cells in the polymer because of the additional melting and render those cells unavailable for extraction by the proteinase K buffer.

Harvesting of Cells

The total cell capture from these biopsies was estimated to be about 3000 to 7000 cells for most of the adequate specimens. In some of the biopsies the epithelium was particularly abundant, requiring the use of 2 caps to collect all material. Since a minimum of 1000 cells was required to perform loss of heterozygosity (LOH) analyses at 4 loci, microdissection was not performed if the cell yield for a particular biopsy was estimated to be less than 1000 cells. In addition, it was important that the microdissected sample contain a sufficiently large number of cells to be representative of each biopsy, so that misleading LOH results were not obtained as a result of a small sample size (Slebos et al., 2004).

In actuality, the true number of cells captured cannot be determined since the laser spot size may be smaller or larger than the diameter of the cell, the thickness of the cell represented on the slide (4–5 μ) is less than the thickness of a complete cell, and the efficiency of cell transfer is seldom 100%. For these reasons, the Pixcell IIE expresses the total capture in terms of estimated dissection volume (cubic microns), with the calculation based on a cylindrical transfer using the beam diameter, tissue thickness, and estimated percent of transfer per laser pulse. Our transfer efficiency was usually estimated to be between 75 and 90% (Figure 2C). The PicoGreen^® ds-DNA Quantitation Reagent (Molecular Probes, Inc., Eugene, Oregon) was also used in some cases to estimate the amount of DNA present in the microdissected specimen.

To prevent the possibility of DNA contamination of either the slide or the captured cells, gloves were worn by the operator at all times, and the microscope stage and unload platform were cleaned with absolute ethanol between biopsies.

Digital images of the coverslipped H&E map slide, the uncoverslipped slide before and after microdissection, and the cap with attached microdissected cells were obtained on each specimen for documentation (Figure 2). If there appeared to be significant nontargeted cellular transfer on the cap when examined microscopically, the cap would be blotted once or twice on a CapSure^™ Pad for removal of the less adherent nontargeted material.

Upon the completion of microdissection, each cap was inserted into the top of a microcentrifuge tube containing 50 μl of proteinase K buffer (0.1 μg/ml proteinase K, 0.2% Tween-20 in 50 mM TRIS at pH 8.0) (Slebos et al., 1992). The tube was then inverted and placed in a 55°C incubator for 3 hours. Following removal from the incubator, the tube was placed right side up, and centrifuged briefly in a microcentrifuge to sediment the material from the cap. The cap was then discarded, the microcentrifuge tube containing the lysed sample sealed with its own cap, and the tube stored at −20°C until analysis.

In the event that there was insufficient time on the first day to microdissect all of the stained slides, the remaining slides would be placed in the desiccator overnight and then microdissected the following day, using a separate Arcturus cap for the capture.

Of the initial 169 biopsy specimens received, 146 (86%) were judged to be adequate by histologic exam, and were submitted for microdissection. Despite the small size of these bronchoscopic biopsies, molecular analysis was successful in all but 5 (3%) of the first 146 samples examined. These results attest to the adequacy of cell capture, as well as to the maintenance of DNA integrity in the harvested cells. Failure to obtain adequate material in the 5 samples could have occurred for a variety of reasons, including insufficient cell capture, failure to invert the microcentrifuge tube during incubation so that the buffer did not contact the cells, or failure to obtain a good seal between the Arcturus cap (with the attached cells) and the microcentrifuge tube allowing leakage during incubation.

Evaluation of Diagnostic Frozen Sections

As previously noted, 2 biopsies were frequently taken from adjacent sites in the bronchus (Figure 1), with one of these placed in formalin for diagnostic purposes and the other snap-frozen for research purposes (laser microdissection and molecular analysis). However, because of the possibility of mosaic variation in mucosal pathology and because dysplastic lesions of the bronchial mucosa are often small (Lam et al., 2000), a few diagnostic cryostat sections were also prepared from the snap-frozen tissue for the purpose of correlating the frozen block histology with the molecular findings. During the nearly 6-year course of this study (Nov., 1999–Sept., 2005), 390 snap-frozen biopsies were collected from 106 bronchoscopic procedures. From these biopsies, nearly 4000 frozen section slides (average of about 10 diagnostic slides per biopsy) were prepared for histologic examination.

Technical Issues

It is well known that permanent section histomorphology is superior to that of frozen sections, and that special efforts must be taken with cryostat sections if one is to approach the quality of permanent (formalin fixed, paraffin embedded) sections. In addition, small biopsies such as bronchoscopic biopsies pose a particular challenge, both technically and morphologically. Two problems that we encountered frequently with the cryostat sections were folding of the mucosal epithelium upon itself and epithelial denudation with dispersal of cells. The folding of epithelium often resulted in tangential sections of mucosal basal and parabasal cell nuclei, complicating interpretation. Epithelial denudation occurred in both normal respiratory epithelium and squamous metaplastic or dysplastic epithelium, the normal respiratory cells often dispersing individually and the squamous metaplastic epithelium detaching in cohesive clusters of cells. Since the orientation of detached cell clusters is often uncertain, diagnostic interpretation is restricted. Both problems may have been related to the manipulation required to place the delicate, flexible mucosal biopsy specimens in a viscous medium such as O.C.T., as opposed to placing tissue in liquid formalin with rapid hardening and fixation in the intact state. With experience in processing and with gentle handling, these issues were partially alleviated, although not fully overcome. Recutting of multiple sections from the frozen block prior to removal from the specimen holder was sometimes helpful in interpretation of tangentially cut epithelium.

To obtain optimum staining of cryostat sections for diagnostic purposes, a separate staining protocol from that used for the LCM slides was required. Fresh staining solutions were prepared each day to avoid weak staining and/or the hazy appearance associated with incomplete clearing due to hydrated alcohols and xylenes. Thin sections (4–5 μ) were selected for diagnostic purposes, whereas occasional sections recognized as being thick were used for microdissection. Immediate fixation was imperative to prevent drying with the associated nuclear enlargement and loss of detail. For this purpose, we found the Shandon Rapid-Fixx^™, consisting of methanol (75%), formaldehyde (7.4%), and acetic acid (5.0%), to be an effective combination that worked well with these biopsies. Modified Harris hematoxylin (Richard-Allan Scientific, Kalamazoo, MI), rather than Mayers, was used for superior differentiation and for simulation of permanent section staining. The complete staining protocol used for the diagnostic slides was: (1) Shandon Rapid-Fixx^™, 15 seconds; (2) tap water, 15 seconds; (3) Modified Harris hematoxylin (filtered), 30 seconds; (4) tap water, 15 seconds; (5) 1.0% Automation buffer, 15 seconds; (6) 95% ethanol, 15 seconds; (7) alcoholic eosin (filtered), 15 seconds; (8) 95% ethanol, 15 seconds; (9) 100% ethanol, 15 seconds; (10) 100% ethanol, 15 seconds; (11) xylene, 1 minute; (12) xylene, 1 minute. A small amount of 4 Å molecular sieves (Sigma-Aldrich) was added to each xylene to reduce hydration with repeated use.

Interpretation of Frozen Sections

Frozen sections of bronchial tissue have been used in the clinical laboratory primarily for margin checks at the time of lung cancer resection (Ghiribelli et al., 1999; Hofmann et al., 2002; Maygarden et al., 2004; Thunnissen and den Bakker, 2005), in which setting the primary emphasis has been upon the detection of infiltrating carcinoma or carcinoma in situ. For research purposes of the current study, an attempt was made on these frozen sections to more precisely categorize a range of histopathologic changes encompassing metaplastic and dysplastic lesions as well as carcinoma in situ, in order that correlations could be made with the molecular findings. Histologic changes on the frozen sections were classified according to World Health Organization (W.H.O.) criteria (Travis et al., 2004), in the same manner as used for permanent sections. Histopathological coding of lesions was based upon the scheme previously reported by Lam et al. (1998), in which numbers were assigned as follows: (1) normal; (2) inflammation; (3) hyperplasia/squamous metaplasia; (4) mild dysplasia; (5) moderate/severe dysplasia; (6) carcinoma in situ; (7) microinvasive carcinoma; (8) invasive carcinoma; and (9) specimen unsatisfactory.

The 2 most common and difficult issues to resolve in the interpretation of the frozen sections were the separation of reactive changes from dysplastic changes, and the grading of those lesions judged to be dysplastic. Of course, the same issues are also problematic in the interpretation of permanent sections, but they are more challenging with frozen sections for the technical and morphologic reasons previously described. Some of the more important issues and limitations were the nuclear enlargement and hyperchromasia (relative to permanent sections) frequently seen in frozen sections, the lack of serial section ribbons, and the difficulty in recutting a frozen block once removed from the cryostat specimen holder.

After observing a wide spectrum of nuclear and cytologic changes in these biopsies, a cautious approach was adopted because of the recognized difficulty in clearly separating some of these changes into either reactive or dysplastic categories. For example, cytomegaly and nucleomegaly with hyperchromasia were sometimes seen in the respiratory epithelium in the absence of squamous metaplasia, but rather in association with inflammation and/or mild epithelial hyperplasia. Another common, problematic finding was that observed in squamous metaplasia of the mature type, characterized by cells having abundant, eosinophilic cytoplasm and sometimes indistinct cell borders, in which frequently some of the cells would exhibit nucleomegaly (Figure 3). In each instance an attempt was made to restrict the diagnosis of dysplasia to lesions which fulfilled W.H.O. criteria, generally in association with squamous metaplasia, while designating more equivocal changes as atypia or atypia, favor reactive. On occasion, squamous metaplasia of immature type could also enter the dysplasia differential, but nucleomegaly was less often associated with this form of metaplasia.

Grading of dysplastic lesions on frozen sections was no less challenging than distinguishing low-grade dysplasia from reactive changes. Particular attention was paid to separation of mild and moderate dysplasia since it was felt that moderate dysplasia is more likely to be a true preneoplastic dysplasia rather than a reactive lesion on morphologic grounds, because it is categorized along with severe dysplasia as high-grade dysplasia by some authors (Lam et al., 2000), and because in this study it was coded together with severe dysplasia as code 5 and was considered to be a histologically positive finding as opposed to mild dysplasia.

Since the dysplasias present as a morphologic continuum, however, in some cases the separation of mild and moderate dysplasia was problematic and necessitated examination of additional sections, if such could be obtained from the frozen block. Severe dysplasia was diagnosed in a few biopsies exhibiting prominent nuclear pleomorphism extending into the upper third of the epithelium but falling short of full thickness involvement of the epithelium. Carcinoma in situ presented somewhat less of a problem in diagnosis because of the marked nuclear pleomorphism, lack of cellular organization, and full thickness involvement of the epithelium (Figure 4).

Variations in degree of dysplasia, as well as nondysplastic areas of squamous metaplasia, were often present in the same biopsy. This mosaic variation in epithelial pattern was also evident when examining 2 adjacent biopsies from the same bronchial segment, and it is consistent with the previously reported observation of the small size of dysplastic lesions frequently noted with fluorescence endoscopy (Lam et al., 2000).

Because of the mosaic variation in histologic pattern and the small size of many dysplastic lesions, it is difficult to assess the validity of this attempt to grade dysplastic lesions on frozen sections by comparing the frozen section and permanent section diagnoses. Interobserver variation also enters into the comparison of diagnoses since many of the frozen section and permanent section slides were examined by different pathologists. Previous studies of observer variation in the grading of cervical dysplasias have shown that significant variation may exist among pathologists in the interpretation of intraepithelial lesions (Ismail et al., 1989; de Vet et al., 1992). Similar interobserver variation was reported in a study of the reproducibility of the WHO grading system for pre-invasive squamous lesions of the bronchus, in which interobserver agreement of 55% was found (Nicholson et al., 2001).

In light of these issues of mosaic variation, small size of dysplastic lesions, interobserver variation, and the use of 2 different types of tissue preparation (frozen and permanent sections), it is worthy of note that the concordance (55%) of frozen section and permanent section diagnoses which were rendered on biopsies taken from adjacent sites in the same bronchial segment at the same time was similar to that of the concordance (53%) of 2 or more permanent section diagnoses rendered on biopsies taken from adjacent sites at the same time (Table 1). In determining these concordance values, code 9 (inadequate) biopsies were not included, and code 1 (normal) biopsies were considered equivalent to code 2 (inflammation only) biopsies since no epithelial abnormality was present. Otherwise, disagreements between frozen section and permanent section diagnoses for all other codes were considered discordant diagnoses.

These concordance values suggest that the histologic categorization of intraepithelial lesions of the bronchus on frozen section, despite the many limitations of that methodology, are reasonably accurate for the research purpose of correlation with the molecular findings from microdissected biopsy specimens.

Technical Conclusions

1. Collection of bronchial biopsies by snap-freezing was optimal for molecular studies since it avoided the cross-linking induced by formalin fixation, thus enhancing the recovery of intact DNA.

2. Virtually pure epithelial cell populations were obtained from bronchial biopsies by laser capture microdissection. Furthermore, the DNA derived from the microdissected cells in this study was both sufficiently abundant and preserved to allow molecular analyses in the great majority of biopsies.

3. The morphologic features of the bronchial mucosa were preserved by snap freezing. With utilization of optimal cryostat sectioning and staining, it was felt that intraepithelial lesions of the bronchus could be reasonably categorized for research purposes of correlation with molecular analyses of microdissected cells.