Abstract
Studies in which 5-bromo-2′-deoxyuridine (BrdU) is used to quantify rates of cell proliferation are conducted prospectively. Therefore, the opportunity exists to select conditions that optimize detection of the BrdU epitope. The objective of this study was to quantify the extent to which the BrdU epitope was masked by formalin vs methacarn fixation in the assessment of cell proliferation. Mammary carcinomas from animals pulse-labeled with BrdU were trisected. A portion was frozen and the remaining two portions were fixed in 10% neutral buffered formalin or methacarn for 24 hr, processed, embedded in paraffin, and sections stained for incorporated BrdU using a peroxidase immunohistochemical staining technique. Antigen retrieval techniques also were applied to formalin-fixed sections. Fixation in methacarn gave the highest labeling index (16.4%), which was comparable to that observed in unfixed frozen sections (17.5%). Formalin fixation alone dramatically suppressed the labeling index (0.3%), which was only partially recovered using various antigen retrieval techniques (2.1-8.1%). Methacarn fixation is recommended for prospective studies in which BrdU detection is planned because of the quantitative recovery of epitope and the simplicity of the approach.
Keywords
BrdU is a halogenated nucleotide analogue of thymidine that is incorporated into DNA during the S-phase of the cell cycle. Pulse-labeling of DNA with BrdU and subsequent immunohistochemical (IHC) detection of labeled nuclei is increasingly being used to study the rates of cell proliferation in normal and malignant cells in vivo and in vitro. In fact, a key word search of the Medline database for the past 5 years revealed 2123 citations containing the term “BrdU” as a text word cross-referenced to cell proliferation in comparison to 430 citations containing the term “tritiated thymidine” similarly cross-referenced. The BrdU method has the advantage, compared to the autoradiographic detection of [3H]-thymidine incorporated into DNA, of speed and convenience of analysis, and it obviates the need to use radioisotopes. The BrdU method is similar in specificity and sensitivity to the autoradiographic detection of tritiated thymidine labeled nuclei (Thornton et al. 1988; Sapino et al. 1990).
As with all IHC methods, a persistent concern is choosing the correct fixative and duration of fixation that will provide minimal loss of antigenicity with maximal preservation of tissue morphology. This issue is particularly relevant to the detection of the BrdU epitope because studies in which BrdU is used are conducted prospectively. Therefore, the opportunity exists to select conditions that optimize epitope detection.
Many factors have been reported to influence the immunoreactivity of the BrdU epitope; prominent among these is the method of tissue fixation (Mitchell et al. 1985; Schutte et al. 1987). Formalin, which is arguably the most widely used fixative, is a noncoagulant fixative that crosslinks proteins (Puchtler and Meloan 1985). Although formalin provides excellent preservation of tissue architecture, it can extensively mask tissue antigenicity, thereby reducing the staining intensity of certain IHC stains (Carson 1990). Whereas extensive masking of the BrdU epitope occurs in formalin-fixed tissue, enzyme digestion before immunostaining has been reported to be effective in unmasking BrdU epitopes (Schutte et al. 1987; Hayashi et al. 1988). However, the required duration of enzyme digestion for unmasking varies according to the length of fixation and this approach can be accompanied by an increase in nonspecific staining, resulting in false-positive staining (Bak and Panos 1997). As an alternative to enzymatic digestion, heat-induced epitope retrieval (HIER) is useful in recovering immunoreactivity of BrdU epitopes in formalin-fixed tissue (Shi et al. 1991; Lan et al. 1995). Compared to enzyme digestion, HIER pretreatment provides equivalent or superior enhancement of immunoreactivity in a shorter period of time and is applicable to more antigens than enzyme digestion (Gown et al. 1993; Dover and Patel 1994). However, many different approaches are used in HIER, requiring interlaboratory standardization of the HIER method if quantitative comparisons of results are intended, and vigilant attention to consistent application of a particular HIER protocol within a laboratory is essential to prevent assay drift.
Alternatives to formalin fixation are available. Alcohol-based coagulant fixatives do not crosslink proteins and therefore tend to reduce antigenic masking (Puchtler et al. 1970), although not all alcohol-based fixatives are effective in preserving antigenic epitopes. Carnoy's- and ethanol-based fixatives are inferior compared to methanol or methacarn for demonstration of IHC stains (Burford-Mason et al. 1994; Shetye et al. 1996). The use of ethanol or methanol alone can cause over-hardening and tissue shrinkage, which results in poor preservation of tissue morphology (Puchtler et al. 1970). Addition of acetic acid to the alcohol decreases the shrinkage artifact, softens the tissue, and aids in the preservation of nucleoproteins (Carson 1990). In an effort to achieve a balance between the advantages and limitations of alcohol fixation, the use of mixtures of methanol, acetic acid, and chloroform (methacarn) has been proposed to achieve both excellent preservation and increased immunoreactivity (Mitchell et al. 1985).
Despite the fact that most studies in which BrdU is used are prospective and therefore can be designed to maximize its detection by IHC, it appears that many investigators do not take advantage of methacarn fixation but rather rely on formalin fixation with variable approaches to unmasking the BrdU epitope. One reason for this may be the lack of information in the literature providing a quantitative assessment of the magnitude of the differences that can result from different approaches to tissue fixation and pretreatment. With this in mind, the experiments we report here were designed to examine quantitatively the degree to which formalin fixation with and without pretreatment involving enzyme digestion (pepsin vs protease XXIV) or HIER vs methacarn fixation influenced the detection of incorporated BrdU in rat mammary adenocarcinomas. Methacarn was chosen as the comparison alcohol-based fixative for the reasons delineated above. To eliminate field selection bias, we implemented a random sampling method in conjunction with computer assisted image analysis (CAIA) to identify labeled cells. The use of this approach is discussed.
Materials and Methods
Twenty-day-old female Sprague-Dawley rats were obtained from Taconic Farms (Germantown, NY). Rats were injected IP with 50 mg/kg 1-methyl-1-nitrosourea (MNU) at 21 days of age as previously described (Thompson et al. 1995). Tumor-bearing rats were sacrificed 35 days post carcinogen. BrdU was administered to each animal (50 mg/kg body weight) 3 hr before sacrifice. Tumors were excised at necropsy and trisected. A portion of each tumor was snap-frozen in liquid nitrogen, another portion fixed in 10% neutral buffered formalin and a corresponding portion fixed in methacarn (60% methanol v/v, 30% chloroform v/v, and 10% glacial acetic acid v/v). Tumors were fixed for 18-24 hr, which has been reported to be the optimal length of fixation (Burford-Mason et al. 1994; James and Hauer-Jensen 1999). Tissues were processed according to the schedule listed in Table 1, using an Autotechnicon (American Optical; Buffalo, NY) tissue processor, and embedded in paraffin.
Specimen Selection
H&E-stained sections of each tumor were evaluated microscopically. Eight confirmed mammary adenocarcinomas were chosen for the staining comparison. Sections were cut at 5 μm from paraffin blocks of both formalin- and methacarn-fixed carcinomas. Sections were floated onto a tissue flotation bath and mounted on 3-aminopropyltriethoxysilane (APES)-treated slides. Sections were heat-immobilized in an 80C oven for 20 min, deparaffinized in three changes of xylene, hydrated in a series of graded ethanols, and rinsed in several changes of distilled water (DW).
Tissue processing
Enzymatic Digestion
Enzymatic digestion was performed on serial sections of formalin-fixed tumors only. One set of sections was incubated using a modified version of the pepsin digestion protocol established by Hayashi et al. (1988): 0.4% pepsin (Fisher Scientific; Pittsburg, PA) in 0.01 N HCl for 15 min at 37C. A second set of sections was incubated with 0.1% protease XXIV (equivalent to protease VII manufactured after 1975) (Sigma; St Louis, MO), pH 7.4, for 5 min at room temperature (RT) (Hayashi et al. 1988). The sections were then rinsed in several changes of DW.
HIER
HIER was performed on serial sections of formalin-fixed tumors only. The HIER slides were double spaced in a Tissue-Tek (Sakura Finetek; Torrance, CA) 24-slide rack and immersed in a polypropylene Tissue-Tek staining dish filled with 200 ml of 10 mM sodium citrate buffer, pH 6.0. The dish containing the buffer and slides was placed in an 1100-W microwave oven and microwaved for 5 min on high power. The fluid level was checked after 5 min and an additional 50 ml of DW was added if necessary. The sections were microwaved for an additional 5 min at high power, followed by a cooling period of 30 min at RT before immunostaining.
Frozen Sections
Three cryostat sections were cut at 6 μm from each of the available frozen tumors (n = 3) and were allowed to air-dry for 30 min. One set was fixed in 10% neutral buffered formalin for 20 min, a second set was fixed in methanol for 20 min, and a third set remained unfixed. Unfixed sections were mounted on Superfrost Plus Gold slides (Erie Scientific; Portsmouth, NH) to minimize tissue detachment during the staining procedure. Fixed sections were rinsed in several changes of DW and unfixed sections were hydrated in DW before immunostaining.
Immunohistochemistry
The sections were immersed in 2 N HCl for 90 min to hydrolyze the DNA. The sections were rinsed in several changes of DW and acid hydrolysis was neutralized by immersion in 0.1 M sodium borate for 5 min, followed by several rinses of DW. Endogenous peroxidase activity was quenched by immersing sections in 3% H2O2 for 5 min. Sections were rinsed in several changes of DW, followed by PBS, pH 7.4, three times for 5 min. All subsequent steps were performed in a humidity chamber. Primary antibody, mouse anti-BrdU (Beckton-Dickinson; San Jose, CA) 1:40, was applied to positive sections and incubated for 1 hr. Negative control sections were held in PBS during the primary incubation. Sections were rinsed in PBS, three times for 5 min. The secondary antibody, biotinylated rabbit anti-mouse IgG (Dako; Carpinteria, CA) 1:200 in 10% normal rabbit serum was applied to all sections and incubated for 30 min. Sections were rinsed in PBS three times for 5 min. The enzyme label, horseradish peroxidase-conjugated streptavidin (Dako) 1:1000, was applied to all sections and incubated for 30 min. Sections were rinsed in PBS three times for 5 min. 3,3′-diamobenzidine (DAB) (Sigma) was applied to all sections and incubated for 10 min. Sections were rinsed in several changes of DW, counterstained in dilute Harris hematoxylin (1:10) for 2 min, rinsed in DW, and blued in Scott's water for 1 min. Sections were rinsed in DW, dehydrated in a series of graded ethanols, cleared in xylene, and mounted using a synthetic resin.
Computer-assisted Image Analysis
BrdU-stained sections were analyzed using a CAS-200 image analysis system (Beckton-Dickinson). The CAS quantitative nuclear antigen software version 3.0 was used to assess proliferation index. The software controls a single CCD camera with two optical bandpass filters operating at wavelengths of 500 nm and 620 nm, respectively (Bacus and Grace 1987). DAB is optimally detected at 500 nm, whereas the hematoxylin counterstain is poorly detected. Conversely, the hematoxylin is detected well at 620 nm, whereas the DAB shows only moderate detection. The use of two different wavelengths enables the system to discriminate between positive DAB-stained nuclei and negative hematoxylin-stained nuclei. The optical density (OD) thresholds for both the DAB (antibody threshold) image and the hematoxylin (nuclear threshold) image are manipulated independently. For comparison purposes, the nuclear and antibody thresholds were set at 0.08 and 0.36 OD, respectively. Any object with an OD value lower than the threshold was excluded from analysis. The antibody threshold was set at a higher level to exclude weak DAB-stained nuclei or to compensate for any nonspecific staining that might cause false-positive results. All images were captured at a magnification of × 400. Definitions for each of the variables obtained by CAIA are provided in Table 2.
Hematoxylin was used as the conterstain instead of methyl green. Quantitatively, methyl green is considered an optimal counterstain to DAB chromogen for image analysis. Qualitatively, methyl green provides poor contrast of tissue elements in light microscopy. We chose to use a dilute solution of Harris hematoxylin that gave good contrast of nuclei without staining cytoplasmic or extraneous tissue elements and was comparable in absorbance to that of methyl green. The nuclear area estimated using Harris hematoxylin vs methyl green differed by only 2% (unpublished data).
Definition of parameters assessed using CAIA
To validate the automated counting method relative to the manual counting procedure that has routinely been used to assess labeling index, methacarn sections were evaluated both by the manual counting method and by CAIA to determine how close the two methods actually are. The “sequence record” software option on the CAS 200 system was used to capture all image fields counted by CAIA from each metha-carn-fixed tumor sample. Hard copies of each image field were made and the nuclei were counted manually to determine labeling index.
Random Sampling
We used an easily implemented method for random sampling of microscopic fields, which reduced the likelihood of field sampling bias in specimen analysis. The slide was moved to a position at which the lower left corner of the tumor was in the field of view. Stage coordinates were initialized to zero in both × and Y directions. The slide was moved in the × direction spanning the length of the tumor. The slide was then moved in the Y direction spanning the width of the tumor. The × and Y coordinate values were entered into a spreadsheet macro that was designed to output 100 random × and Y coordinates within the tumor area.
Field Count
In a related study performed by our laboratory (unpublished data), 10 mammary adenocarcinomas were subjected to census counting for detection of BrdU-labeled nuclei. These data were used for statistical modeling to compare methods of counting (census vs random) and to determine statistical power as a function of the number of microscopic fields counted. In addition, the number of tumors to be evaluated based on the magnitude of difference anticipated between treatment conditions was assessed. These data provided the basis for using the random method of field selection, for counting 20 high-powered microscopic fields per specimen, and for the number of tumors evaluated per treatment condition.
Statistics
Proportion data were rank-transformed as recommended by Conover (1971). Differences among treatment conditions in positive area, positive stain, and heterogeneity were evaluated by analysis of variance of the rank-transformed data, which were normally distributed, with post hoc comparisons among groups using the Bonferroni method (Sokal and Rohlf 1981).
Results
Qualitative Assessment of Differences in Staining in Paraffin-embedded Tissues
Photomicrographs of BrdU-stained sections from four of the treatments evaluated showed remarkably different staining patterns (Figures 1A-1D). Qualitatively, the formalin-fixed sections that underwent no pretreatment other than acid denaturation exhibited very weak to no staining for BrdU (Figure 1A). Pretreatment of formalin-fixed tissue with pepsin digestion, as described in Materials and Methods, appeared to have little effect in recovering BrdU epitopes. Staining intensity was similar to that observed in formalin-fixed sections without treatment (not shown). Protease XXIV digestion of the formalin-fixed sections aided in partial recovery of BrdU epitopes (Figure 1B). However, enzymatic digestion using protease XXIV had an adverse effect on tissue morphology; several nuclei were digested to the point of being unrecognizable. This made the process of image thresholding using CAIA extremely difficult. HIER treatment of the formalin-fixed sections aided in partial recovery of BrdU epitopes. Moderate staining was exhibited, with a well-circumscribed nuclear membrane and diffuse staining within the nuclei (Figure 1C). The methacarn-fixed sections, which received no pretreatment other than the acid denaturation to which all tissue sections were subjected, displayed the highest staining intensity, with sharply defined nuclear membranes and a more homogeneous staining pattern within the nucleus (Figure 1D).
Quantitative Assessment of Differences in Staining
Validation of CAIA Automated Counting Procedure. The BrdU labeling index was determined on eight mammary adenocarcinomas using a manual vs an automated counting technique as described in Materials and Methods. The results of this comparative analysis are shown in Table 3. Both methods gave essentially identical results. The average labeling index (%) determined by the manual and automated method, 16.25 ± 1.55 and 16.44 ± 1.58, respectively, differed by less than 2%, a difference that was not statistically signficant (p=0.93). Given that both methods yielded comparable results, the automated method was used for the remainder of the analyses.
BrdU staining comparison: formalin, formalin/protease, formalin/HIER, and methacarn fixation. Proliferating cells are marked by dark brown nuclei and counterstained with light hematoxylin. (
Comparison of Fixation and Pretreatment: Paraffin-embedded Tissue. Paraffin-embedded tissues sections were used to assess the effects of tissue fixative and pretreatment on the BrdU labeling index, staining intensity, and staining heterogeneity using the criteria outlined in Table 2. The results are shown in Table 4. Fixation with formalin alone resulted in little detectable staining; the average labeling index for the eight tumors evaluated was 0.3%. This quantitative analysis is consistent with the qualitative observations shown in Figure 1. In contrast, qualitatively, pepsin digestion appeared to have little effect on BrdU immunoreactivity. However, the data in Table 4 show an appreciable increase in the labeling index (2.1%), and in comparison to fixation in formalin without pretreatment the increase was statistically significant (p<0.01). Pepsin digestion also caused a numerical increase in staining intensity (p=0.058) and a reduction in heterogeneity of staining (p = 0.055) in comparison to fixation in formalin alone. Further improvements in BrdU immunoreactivity, as reflected by an increase in the labeling index, were observed with protease XXIV digestion (5.5%) or HIER (8.1%), and these increases were accompanied by increases in staining intensity (22.7% and 35.2%, respectively) and decreases in staining heterogeneity (72.5% and 68.3%, respectively). Relative to increasing the labeling index in formalin-fixed tissue, the rank order of effectiveness was HIER > protease XXIV > pepsin > formalin alone. Statistically, HIER was superior to formalin alone or pepsin in increasing labeling index (p<0.01) but was not different from protease XXIV (p = 0.32).
Methacarn-fixed tissue had the highest labeling index and staining intensity and the lowest staining heterogeneity. These differences were statistically significant in comparison to all other fixation and pretreatments that were examined (p<0.01).
Comparison of labeling indices in paraffin-embedded tissue sections: manual method vs CAIA a
Nuclei from 20 high-power microscopic fields were analyzed by manual counting method and CAIA (n=8).
Values are mean ± SEM and were not statistically different; p>0.05.
Values are total number of BrdU-positive nuclei divided by total number of nuclei.
Values are BrdU-positive area in μm2 divided by total area in μm2.
Effect of fixation on BrdU immunohistochemistry in paraffin-embedded tissue sections a
Nuclei from 20 high-power microscopic fields were analyzed from tissues subjected to different fixatives or IHC pretreatments (n = 8).
Values are the mean ± SEM. Values within a column without superscripts were statistically different; p<0.05.
Values were not statistically different from one another; p>0.05.
Immunohistochemical Staining of Frozen Tissue Sections
To further explore the meaning of the results from paraffin-embedded tissue, a subset of these tumors for which there was an adequate amount of frozen tissue was used to prepare frozen sections that were subsequently probed for the BrdU epitope by IHC. The results of these analyses are shown in Table 5. Although tissue was available from only three of the tumors evaluated, it can be seen that the frozen unfixed tissue or the frozen sections fixed with methanol had labeling indices (17.5% and 16.0%, respectively) that were similar to those observed with methacarn-fixed, paraffin-embedded tissue (16.9%). As expected, fixation of the frozen sections with formalin resulted in a loss of BrdU immunoreactivity, as indicated by a lower labeling index (9.7%).
Comparison of labeling indices of tissue prepared as frozen or paraffin sections a
Nuclei from 20 high-power microscopic fields were analyzed by manual counting method and CAIA (n=3).
Values are mean ± SEM and were not statistically different; p>0.05.
Values are mean ± SEM and were statistically different; p<0.05.
Discussion
The series of experiments reported here was initiated on the premise that, all other things being equal, simple techniques are preferable to complicated ones. We reasoned that, if methacarn fixation could be used to achieve better results in a prospective study using BrdU than a more complicated approach such as formalin fixation and antigen retrival, even if an optimal approach could be developed to reach the same goal, the simpler approach would be preferred. As discussed below, our results support the use of methacarn fixation because of the quantitative recovery of epitope and the simplicity of the approach.
The effect of fixative and pretreatments on the IHC detection of BrdU epitopes was evaluated in rat mammary carcinomas. Tissue fixed in formalin alone exhibited a labeling index of less than 1%, whereas the same tissue fixed in a coagulant fixative, methacarn, had a labeling index of 16.4%. To our knowledge, the magnitude of the difference between the labeling index determined on formalin- vs methacarn-fixed tissue has not been previously quantified. Equally informative were experiments showing that three procedures commonly used to regain BrdU immunoreactivity in formalin-fixed tissue varied in their effectiveness and gave labeling indices 50.6-87.2% lower than the labeling index determined in methacarn-fixed tissue. These findings have several implications. First, an increasing number of laboratories are using pulse-labeling techniques with BrdU to estimate rates of cell proliferation. Clearly, rate estimates can be significantly underestimated depending on choice of fixative and/or pretreatment. The potential magnitude of this effect has not been previously reported. Second, given that studies in which BrdU is used are conducted prospectively, there is not a requirement to use formalin as a fixative. The data reported in Table 4 show the advantages of using methacarn relative to measuring the labeling index, and this is the most convenient and economical of the approaches evaluated. Moreover, methacarn provides excellent preservation of tissue architecture. Finally, the results show that, if formalin must be used as the fixative, HIER treatment gives results numerically better than enzyme digestion, and staining intensity is greater and staining heterogeneity reduced. Both factors improve the sensitivity of the assay. HIER has the dual advantage, in comparison to enzyme digestion, of speed and economy, although vigilance in adherence to the HIER protocol must be exercised to ensure reproducibility of results and to prevent protocol drift. We recommend that serious consideration be given to methacarn fixation for the reasons outlined above. This recommendation is reinforced by the data presented in Table 5 showing that labeling indices determined using methacarn-fixed, paraffin-embedded tissue are essentially identical to those values obtained on unfixed frozen sections of the same tissue.
Many issues must be considered in deciding whether to limit cell counting to high-labeling fields of cells or to use a random field selection approach so as to obtain an unbiased estimate of the overall rate of cell proliferation. Some of the issues to be considered are reviewed in Siitonen et al. (1993). In this article we present an easily applied approach for random field selection, and we use this approach in conjunction with CAIA to facilitate the determination of labeling index. As shown in Table 3, this semiautomated approach gave results that were within 2% of the values determined by manual counting. The CAIA approach is convenient and time-saving, allowing a large sampling of tissue area (20 high-powered fields or approximately 2500-3000 cells; Table 3) in a short time frame (30 min vs 2 hr manual method) and with the objectivity of computer assisted analysis of uniformly acquired digital images. This, plus the easily implemented approach for random field selection, removes potential sources of technical operator-induced variation and bias from the quantification of rates of cell proliferation. The process also reduces the fatigue that can accompany the counting of large numbers of cells manually. These advantages are obtained without loss of accuracy, as shown by the data presented in Table 3.
Other than providing a quantitative assessment of the potential impact of fixative and pretreatments on estimates of the BrdU labeling index, the data in Table 4 show the advantage of methacarn fixation in improving conditions that increase the accuracy of CAIA as applied to the quantification of BrdU signal. These factors can also affect the ability to count labeled cells manually. Specifically, the staining intensity and heterogeneity data shown in Table 4 indicate that fixation with methacarn gives the greatest signal-to-noise ratio, i.e., stained cells are darker and there is less variation in staining intensity among the randomly selected fields that were assessed. Both features make it easier to identify and count labeled cells. The practical implications are easily recognized on examination of the photomicrographs presented in Figure 1.
In summary, fixation of tissue with a coagulant fixative such as methacarn offers significant advantages in the assessment of the rate of cell proliferation if BrdU is used to label cells going through the S-phase of the cell cycle. Fixation with methacarn gave the least biased estimate of the rate of cell proliferation. If a crosslinking fixative such as formalin must be used, then HIER gives better recovery of BrdU immunoreactivity than digestion with either pepsin or protease XXIV. For investigators seeking to apply CAIA to the analysis of rates of cell proliferation, methacarn fixation substantially improves staining intensity and decreases staining heterogeneity. It is also simple and convenient. Using an easily implemented random approach to field selection, CAIA values for the labeling index were comparable to the values obtained using the routine manual counting method.
Footnotes
Acknowledgments
Supported by PHS grants NCI-CN-45029-72 and NCI-CN-75011-72.
We thank Ms Pam Wolfe (AMC Biostatistician) for her contributions to this manuscript.