|
|
|
|
 |
|||
|
|||
Volume 53 (6): 753-762, 2005
Copyright ©The Histochemical Society, Inc.
Quantifying Estrogen and Progesterone Receptor Expression in Breast Cancer by Digital Imaging
Department of Medicine (JRF,LMW), the Department of Family and Preventive Medicine (JMM,LN), the Rebecca and John Moores University of California, San Diego, Cancer Center (MKS,LC,SM,JMM,LN,AS,JRF,LMW), and San Diego Supercomputer Center (CLC), University of California, San Diego, La Jolla, California; and San Diego State University, San Diego, California (AS)
Correspondence to: Linda M. Wasserman, MD, PhD, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093. E-mail: lwasserman{at}ucsd.edu
 |
Summary |
|---|
 Top Summary IntroductionMaterials and MethodsResultsDiscussionLiterature Cited |
|---|
Developments in digital imaging and fluorescent microscopy provide a new method and opportunities for quantification of protein expression in human tissue. Archived collections of paraffin-embedded tumors can be used to study the relationship between quantitative differences in protein expression in tumors and patient outcome. In this report we describe the use of a DeltaVision Restoration deconvolution microscope, combined with fluorescent immunohistochemistry, to obtain reproducible and quantitative estimates of protein expression in a formalin-fixed paraffin-embedded tissue. As proof of principle, we used antibodies to the estrogen and progesterone receptors in a hormone receptor–positive breast cancer specimen. We provide guidelines for control of day-to-day variability in camera and microscope performance to ensure that image acquisition leads to reproducible quantitative estimates of protein expression. We show that background autofluorescence related to formalin fixation can be controlled and that for proteins that are expressed in nearly every cell, multiplexing two primary antibodies on the same slide does not significantly affect the results obtained. We demonstrate that for proteins whose expression varies markedly from cell to cell, data reproducibility, as assessed by imaging successive tissue sections, is more difficult to determine.
(J Histochem Cytochem 53:753–762, 2005)
Key Words: digital imaging • estrogen receptor • progesterone receptor • protein quantification
|
Introduction |
|---|
|
TopSummaryIntroductionMaterials and MethodsResultsDiscussionLiterature Cited |
|---|
RECENT IMPROVEMENTS in confocal fluorescent microscopy, camera systems, and associated software enable the use of fluorescent immunohistochemistry to quantify light emitted from each cell and to use these quantitative estimates as a biomarker for protein expression. These developments can be harnessed to quantify and compare protein expression in different types of cells in human tissue sections. The availability of archived, formalin-fixed, paraffin-embedded human tumors can be combined with this technology and with antibodies to proteins whose expression is altered in tumorigenesis to evaluate differences in protein expression across tumors. Quantitative analyses of protein expression can be related to important clinical parameters such as disease outcome.
We used a DeltaVision Restoration confocal microscope (Applied Precision, LLC; Issaquah, WA) and associated SoftWorx software (Applied Precision) to compare and quantify the expression of the estrogen receptors (ER) and progesterone receptors (PR) in a series of formalin-fixed, paraffin-embedded sections of breast cancers. For many years, it has been thought that formalin fixation of tissue, with its inherent autofluorescence, precluded the use of fluorescent secondary antibodies (Mote et al. 1999
; Chwirot et al, 2001; Ermert et al. 2001). We have developed and optimized methods for fluorescent immunohistochemical staining of formalin-fixed, paraffin-embedded tissue, image acquisition, and cell-by-cell analysis to obtain reproducible quantitation of ER and PR expression in breast cancers to compare the relationship between their levels of expression and inherited differences in estrogen metabolism.
We have developed and describe here methods to control and minimize formalin autofluorescence and to standardize image acquisition procedures to insure that day-to-day fluctuations in the microscope do not compromise the data reproducibility. We compare data obtained from multiplexed and uniplexed immunostaining for the two antibodies and from repeated imaging of sequential sections of a breast cancer. We demonstrate that fluorescent digital microscopy can be used to evaluate and quantify intratumoral and intertumoral differences in protein expression in breast cancer.
|
Materials and Methods |
|---|
|
TopSummaryIntroductionMaterials and MethodsResultsDiscussionLiterature Cited |
|---|
Antibodies and Materials
Rabbit polyclonal antibodies to the human ER (clone 18-0174) and monoclonal antibodies to the human PR (clone PR-2C5) were purchased from Zymed Laboratories, Inc. (South San Francisco, CA). Normal sera from goat and donkey, whole affinity purified anti-mouse IgG-conjugated to rhodamine Tritc and anti-rabbit antisera conjugated to fluorescein FITC were purchased from Jackson ImmunoResearch Laboratories, Inc. (Westgrove, PA). Tetraspeck fluorescent microspheres were purchased from Molecular Probes (Eugene, OR). Tetraspeck microspheres facilitate the adjustment and calibration of microscopes and can be imaged with each filter (McNally et al. 1999
). Tetraspeck beads have minimal fading and are consistent in intensity because of prestaining of one lot. Fluorescent intensity readings of the Tetraspeck beads were used to assess variability in microscope and camera settings (Swedlow et al. 2002). Cover slips (12-544C), glycine, and various reagents were purchased from Fisher Scientific (Pittsburgh, PA). Sodium borohydride was purchased from ICN Biomedicals (Aurora, CO). DAPI, DABCO, and other reagents were purchased from Sigma-Aldrich (St. Louis, MO). Target Retrieval Solution was purchased from Dako (Carpinteria, CA).
Positive and Negative Controls
To determine the reproducibility of quantitative data obtained and the parameters affecting it, we used a single breast cancer specimen as a control and reference point. With permission of the University of California, San Diego (UCSD) Human Subject Research Protection Program, a breast cancer tissue specimen was obtained from the Moores UCSD Cancer Center Molecular Pathology Shared Resource human tissue repository. The specimen had previously been shown by conventional immunohistochemistry to express both ER and PR. The breast cancer tissue specimen was divided into multiple fragments, each approximately 8 mm3. Each fragment was fixed in formalin and embedded in paraffin. Multiple, consecutive 5-µm sections were made from the paraffin tissue blocks.
Several negative controls were used to evaluate background fluorescence. One negative control was a slide of the breast tumor that received only secondary antibody, which established the level of background autofluorescence attributable only to formalin fixation (Mote et al. 1999). Background fluorescent readings were also obtained on slides receiving both primary and secondary antibodies from tissue areas not expressing ER or PR. At least two background readings were taken on each slide and served as additional negative controls.
Fluorescent Immunohistochemistry Protocol
Paraffin sections were dewaxed in xylene and rehydrated in decreasing concentrations of alcohol. All slides were treated with 2% sodium borohydride for 30 min followed by incubation in 0.1 M glycine to quench endogenous formalin-related fluorescence (Jagla et al. 2000). The Dako Black and Decker Steamer protocol was used for antigen retrieval with a steam incubation time of 75 min. Slides were washed with several changes of water followed by washing in standard PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.4 mM KH2PO4, and 100 mM HCl) (1x PBS) and stored at 4C until staining.
Before placement of the first primary antibody, nonspecific antibody binding was blocked with 20% normal goat and 20% normal donkey serum for 1 hr at RT. A second 20-min incubation in 20% normal goat and 20% normal donkey serum preceded incubation with the second primary antibody. Primary antibodies were diluted 1:50 in 10% normal donkey serum and 1x PBS and incubated for 1 hr at RT. Following all antibody incubations, all slides were washed three times in 1x PBS. Secondary antibodies were diluted 1:100 and also incubated for 1 hr at RT in a humidified chamber in the dark (Mote et al. 1999). Nuclei were stained with DAPI (0.4 µg/ml in 1x PBS) for 30 min. Following three water rinses, slides were air dried and mounted in gelvatol (with DABCO) using #1.5 (0.16–0.19 mm) cover slips (Fisher Scientific).
Microscope Parameters
Images were captured with a DeltaVision Restoration microscope system using a Photometrics CH350L liquid-cooled charged-coupled device (CCD) camera (Applied Precision) (500 kHz, 12-bit, 1024 x 1024 pixels), with an intensity range of 0 to 4095 counts, attached to an inverted, wide-field deconvolution fluorescent microscope (Nikon TE-200) incorporated into a DeltaVision System (Applied Precision). Optical sections were acquired using a 10x (Nikon, PlanApo, NA 0.45 lens) or 40x (Nikon, Plan Fluor, NA 1.4) objective in 0.5-µm steps in the z-axis using the attached Applied Precision motorized stage.
The fluorescently labeled secondary antibodies for ER (FITC) and PR (rhodamine) were excited with a standard mercury arc lamp attached to a fiberoptic illumination scrambler providing even illumination. Fluorescence was detected using a DeltaVision standard DAPI, FITC, Texas Red, CY-5 filter set. The camera was set to a gain of 1. The photo sensor (PSR) settings, with our preferred settings in parentheses, included frequency response (Hz) settings (DC, 3K) and gain factor dials (1 x 104). The photo sensor measures illumination intensity by sampling a small percentage (1%) of the light from the arc lamp. The photo sensor signal is recorded by the DeltaVision Controller (log files) and then used to correct for variations in arc lamp intensity. The generated log files containing the recorded signal were monitored for daily changes in illumination. The CCD camera was set to a gain factor of 1 to maintain a linear relationship between exposure time and fluorescence. No neutral density filter was used.
Because the level of expression of ER and PR in breast cancers is likely to vary, a conservative exposure time was chosen for each fluorophore to ensure that pixels in each acquired image would not saturate (i.e., reach the maximum possible intensity of the camera). By using shorter exposure times, it was possible to image each slide within the camera's linear range of fluorescent intensity, thereby enabling comparison of differences in fluorescent intensity of ER and PR expression across slides. The shorter exposure time also reduced photobleaching of fluorophores.
Image Acquisition
The microscope and camera system was shared among a group of investigators and was in constant use. As a result, we noted significant day-to-day variability in the ER and PR image intensity of the positive control breast cancer. To acquire reproducible quantitative data from slides imaged on different days, we developed a method to compensate for the daily variability in microscope and camera performance.
We repeatedly imaged the positive control breast cancer using slides stained on different days and images acquired at varying periods after immunostaining. We identified specific areas within the center of the tissue that could be followed through successive sections. These areas included invasive ductal carcinoma, ductal carcinoma in situ, and normal breast ducts. By repeatedly imaging the same tissue areas, we developed a frame of reference to identify and then control for fluctuations in the camera and microscope.
We developed a checklist of microscope and camera settings, including each parameter that could vary and affect the fluorescent intensity readings. To obtain the appropriate intensity reading for each fluorophore, microscope and camera settings were checked using the distinctive and centrally placed tissue areas of the positive control breast cancer as reference, each time that images were acquired. Settings were adjusted when necessary. The variables checked daily included light bulb intensity, illumination alignment, exposure time, filter degradation, and artificial light from outside source. Monitoring the log files for daily changes of the PSR accounted for any inconsistency in fluorophore intensities.
The slide containing only the Tetraspeck microspheres was imaged first. The data obtained from the daily Tetraspeck check was essential to evaluate variability in camera bulb intensity, filter status, or illumination misalignment. The microspheres are stained with four separate fluorophores that emit light at excitation/emission wavelengths comparable to DAPI, FITC, rhodamine, and CY-5. To control for any variability within the Tetraspeck microspheres, all Tetraspeck calibrations used microspheres obtained from the same dye lot. Normative values were established for these molecules through repeated measures. A field containing 5 to 10 beads was imaged with each set of controls and verified for intensity changes. To minimize photobleaching of the microspheres, Tetraspeck slides were changed after every 4 days of imaging. Because staining variability is not a factor in imaging the Tetraspeck slide, the resulting fluorescent intensity reading provides information regarding changes in apparent image intensity due to changes in PSR.
The negative control slide was then imaged to determine average background autofluorescence attributable to formalin fixation of tissue. Finally, the positive control tissue slide was imaged to adjust microscope settings and camera exposure times to minimize variability in readings due to instrument fluctuations. Each day two positive control slides were imaged: the control slide stained concurrently with the new series of cases to be imaged and the control slide that had been imaged most recently.
By repeatedly imaging approximately the same areas within the positive control tissue, we developed a reference range for expected fluorescent intensity readings for each fluorophore. The usual fluorescence intensity for each antibody and fluorophore was the criterion used to adjust camera variables. Image intensities of the previously imaged control slide and the new control slide were compared to determine consistency and adjust exposure times. By comparing similar fields on slides of the same positive control tissue imaged on different days, changes in microscope and camera performance could be detected.
Acquired images were transferred and archived onto a Silicon Graphics Octane workstation (SGI; Mountain View, CA).
Image Processing
Before processing images, the photo sensor log sheets generated by SoftWorx for each image were verified to ensure all the parameters (i.e., gain, ND filters, exposures, and PSR) were set correctly at time of imaging. DeltaVision raw 12-bit data images were processed for collection and analysis of multidimensional microscopy data.
Raw 10x data images were saved, processed, and analyzed on attached Silicon Graphics workstations (O2, Octane) using the DeltaVision software package SoftWorx (version 2.50). Raw rather than deconvolved data were used to maintain scalar consistency from image to image.
The variation in fluorescent intensities of successive z sections was less than 3%. To quantify fluorophore intensities of in-focus cells, only one z section was selected from each image to be resaved as a DeltaVision image. A 24-bit RGB color-tagged image file format (tiff) was saved to show the relative intensity and intracellular localization of each fluorophore in the various tissue components of the image. An 8-bit grayscale tiff file of the DAPI z section was saved for evaluation of tissue morphology within the section. Using the SoftWorx Data Inspector, a three-dimensional (3-D) graph flattened image was generated for each DeltaVision image and saved as a tiff. The Data Inspector shows the intensity values in a rectangular region of the image as both color tables and graphs.
All tiff files were imported into Adobe Photoshop 7.0 (Adobe Systems, San Jose, CA).
Quantification of Antibody Fluorescent Intensities
The select box analysis program from SoftWorx Data Inspector was used to quantify the fluorescent intensity of ER and PR expression within each nucleus within each of the tissue areas examined. Using SoftWorx Data Inspector, a 6 x 6 pixel box (pixel size = 0.6680) was carefully centered within each nucleus. This was the largest box that could fit entirely within the nucleus of most normal breast ductal cells. Because breast cancer nuclei are larger than normal ductal nuclei, boxes were placed approximately in the center of breast cancer nuclei. For each box, the fluorescent intensity attributable to FITC and to rhodamine was calculated by the software. For each box, summary measures (N, minimum/maximum, mean, and SD) of the pixel intensities were calculated for each selected cell and saved to a cumulative table. The localized statistics for each cell initially were gathered using the SoftWorx analysis program for DeltaVision microscopy data sets.
To measure the mean background fluorescence intensity for each selected area, two boxes were placed in stromal areas in which there was no binding by primary antibody. These background fluorescence values were compared with the values obtained from the negative control slide and were subtracted later from the total mean intensity of each nucleus (Trinkle-Mulcahy et al. 2003).
The feature set of the SoftWorx program proved inadequate for data sets containing a large number of cells. Intensity data could be acquired for only one fluorophore at a time. The process of reselecting the cells to assess a second fluorophore was both tedious and error prone. A custom program (available at http://www.msg.ucsf.edu/IVE/) was written to perform statistical operations identical to that of SoftWorx on each of the local centers of the cells in the data set, including all wavelengths simultaneously. In this utility, both the format that SoftWorx uses to describe these nuclei and the interface with which SoftWorx accesses the data were taken into account.
The program was developed using the Priism data interface (Macromolecular Structure Group; University of California, San Francisco, CA) (Chen et al. 1992; Chen et al. 1996). Although Priism and SoftWorx share a common origin in that they both are based on David Agard's and John Sedat's image analysis libraries, there are significant differences in the format each uses to describe relative locations in the data set (Agard and Sedat 1983; Agard et al. 1989). Nuclei are selected using a graphical interface in SoftWorx, and the location of the nuclei for one wavelength and their statistics are then exported to a simple file. This file is then used as input to our custom software, which reads only the locations of the nuclei, formats them into relative locations compatible with Priism, and recalculates the summary statistics for each center in all the wavelengths present in the data set. The software then outputs the summary statistics and the locations in the exact format as the input data, with entries added for the additional wavelengths. All data were saved as text files and imported into Excel spreadsheets (Microsoft Excel for Windows, XP2000; Microsoft, Redmond, WA) for preliminary evaluation (e.g., numerical data inspection from region of interest).
Data Analysis
Fluorescent intensities were obtained simultaneously from filters specific for FITC and for rhodamine. In our paradigm, differences in fluorescent intensity across cells are presumed to reflect differences in the amount of binding of primary antibody to antigen, which in turn reflects the amount of protein present in the cells. However, variation in fluorescent intensity may also reflect nonspecific factors such as autofluorescence secondary to formalin fixation, nonspecific binding of either primary or secondary antibodies, or overlap in the emitted light spectra of the two fluorophores.
We evaluated the contribution of background fluorescence by comparing fluorescent intensities obtained from slides receiving only secondary antibodies with background values obtained from slides receiving both primary and secondary antibodies. Paired t-tests were computed to compare these background fluorescence intensities for each fluorophore.
To evaluate the contribution of nonspecific binding of primary antibodies or overlapping fluorescent spectra to obtained cell fluorescent intensities, we compared data obtained from slides that received only one primary antibody with data from slides receiving both primary antibodies in a multiplexed format. We reasoned that if there were nonspecific antibody binding or overlapping spectra effects, we would find a difference in the relative ratio of brightly to weakly staining tumor and normal ductal cells as a function having been exposed either to one or to two primary antibodies. We assumed that nonspecific effects would increase observed fluorescent intensities so that multiplexing would be associated with a higher frequency of brightly staining tumor or normal ductal cells.
To compare the relative frequencies of brightly and weakly staining tumor and normal ductal cells as a function of staining condition, we first calculated the ranges of mean cell intensities for all tumor and normal ductal cells for each fluorophore for each condition. We then divided each of the four distributions of mean cell intensities (ER only, PR only, ER multiplexed, PR multiplexed) into quintiles. We compared the frequency of normal ductal and tumor cells falling within each quintile for ER-only, PR-only, and ER- or PR-multiplexed immunostaining conditions using nonparametric statistics.
We evaluated the effect of elapsed time between immunostaining and image acquisition on mean cell intensities of ER with FITC as the fluorophore in tumor cells in a series of 10 slides imaged between 5 and 61 days following immunostaining. Differences in mean cell intensities were compared by paired t-tests.
A p-value of less than 0.05 for all statistical tests was considered significant. All statistical tests were computed using StatView software (SAS; Cary, NC).
|
Results |
|---|
|
TopSummaryIntroductionMaterials and MethodsResultsDiscussionLiterature Cited |
|---|
Image Analysis
The positive control breast cancer specimen had been previously shown by conventional immunohistochemistry to express both ER and PR. Sections of this tumor demonstrated widespread in situ and invasive ductal cancer (Figure 1A).
|
|
Figures 1 and 2 show a progression of images of a representative area from the positive control breast cancer specimen. The area contains both invasive and in situ breast carcinoma and is shown following hematoxylin and eosin staining (Figure 1A) and following multiplexed staining with antibodies to ER and PR (Figures 1B–1D).
In this breast tumor, cell to cell variability in the expression of the PR was pronounced in comparison to expression of the ER. Expression of the ER is seen in nearly every tumor cell and is relatively uniform (Figure 1C). In contrast, PR expression varies dramatically from cell to cell and is most pronounced within cells localized to the interface between tumor and stroma (Figure 1D). Because of this cell-to-cell variability in PR expression, we had determined in preliminary experiments that PR-positive cells detected by secondary antibodies conjugated to rhodamine were easier to visualize than if PR was detected by secondary antibodies conjugated to FITC.
In Figure 2, a small area of tumor from Figure 1 has been selected for additional magnification to demonstrate the method of box analysis used to quantify ER and PR expression pixel by pixel and nucleus by nucleus. The relative homogeneity of ER expression within each nucleus and among neighboring cells (Figures 2A and 2C) contrasts with pronounced inter- and intranuclear PR expression (Figures 2B and 2D). Whereas most neighboring tumor cells typically have some level of expression of ER, a cell with intense expression of PR can be surrounded by cells with no detectable PR expression.
Background Fluorescence
Background fluorescence attributable to formalin fixation of tissue and to nonspecific binding of antibodies was assessed in two ways. Readings were made within tissue areas of negative control slides that received only secondary antibodies. Background fluorescence was measured as well on slides receiving both primary and secondary antibodies in stromal areas where there was no antibody binding. Fluorescent intensities of least two such fields were obtained for each section. Fluorescent background values were compared by paired t-tests for negative control sections and for sections stained only with ER or only with PR and for slides multiplexed with both ER and PR (Table 1).
|
For FITC, mean background fluorescence increased progressively, but not significantly, as primary antibodies were added. Background fluorescence for slides receiving no primary antibody was not significantly different from that of slides receiving either only ER as primary antibody (p=0.94) or ER multiplexed with PR (p=0.09). There was no difference in mean background fluorescence in slides receiving ER only vs ER multiplexed with PR (p=0.16).
Considering rhodamine as the fluorophore, background fluorescence did not demonstrate a consistent relationship with addition of primary antibodies. Background fluorescence of negative control slides was not different from the background fluorescence for multiplexed slides (p=0.42). However, background fluorescence for slides receiving only PR as primary antibody was significantly less than either the negative control (p=0.02) or multiplexed slides (p=0.03).
Comparison of Uniplexed and Multiplexed Immunostaining for Estrogen and Progesterone Receptors
To assess whether multiplexed immunostaining, using two primary antibodies on the same slide, affected the ratio of brightly to weakly staining ER- and PR-positive cells, three sequential sections of the breast cancer control tissue were stained on the same day. One section received only ER as primary antibody, one section received only PR as primary antibody, and one section received multiplexed immunostaining with both ER and PR as primary antibodies. For evaluation we selected three microscopic fields in the center of each slide whose distinctive morphology could be followed through successive sections.
Using the box analysis method, fluorescent intensities for ER and PR were obtained within nuclei of all normal ductal and invasive breast tumor cells within these microscopic fields on the three successive sections. We evaluated 657 cells in the ER and PR multiplex condition, 520 cells in the ER-only condition, and 408 cells in the PR-only condition. Quintile frequencies of the mean cell intensities, corrected for background, were determined for each fluorophore for each staining condition. The frequency of normal ductal and tumor cells falling within each quintile for the ER-only, for the PR-only, and for the ER and PR multiplexed immunostaining conditions were calculated and compared by chi-square (Figure 3).
|
For ER expression, multiplexing ER with PR did not significantly alter the ratio of weakly staining to brightly staining tumor cells (
2 = 3.44, df = 4, p=NS) or normal ductal cells (2 = 8.52, df = 4, p
0.10). For PR expression, however, the ratios of brightly staining to weakly staining normal ductal and invasive tumor cells were significantly different in the two staining conditions. For normal ductal cells, multiplexing PR with ER antibodies was associated with a greater number of brightly staining PR-positive cells (2 = 70.7, df = 4, p0.001). For tumor cells, multiplexing was associated with a greater number of weakly staining PR-positive cells (2 = 113.2, df = 4, p0.001).
Reproducibility of Quantitative Analyses of Estrogen Receptor (FITC) Expression as a Function of Time between Immunostaining and Image Acquisition
Our ultimate research aim is to compare ER and PR expression in a series of breast cancers. Images of the series of breast cancer cases being studied are acquired on different days and at varying lengths of time following immunostaining. We wished to determine whether differences in the length of time between immunostaining and image acquisition affects fluorescent intensities.
Because ER expression in tumor and ductal cells within the positive control tumor was relatively uniform compared with PR expression, we evaluated the effect of time on average ER intensity in tumor cells in a series of sequential sections of the positive control breast tumor. Each slide received multiplexed ER and PR immunostaining according to our protocol. An area of distinctive tumor morphology was identified in the central area of each slide so that cells within that area could be found and imaged through sequential sections. Before image acquisition, immunostained slides had been stored in the dark at RT. Data from the first imaging of each slide were compared to avoid photobleaching from repeated fluorescent illumination of the same field.
Data obtained from this series of experiments are presented in Figure 4. Eight slides were imaged from 1 to 61 days following immunostaining. The corrected mean intensity of ER-expressing tumor cells was relatively consistent when the elapsed time between immunostaining and image acquisition varied from 1 to 19 days, averaging 1396 ± 75 for the four slides imaged within that time period. Four of the slides were not imaged until 3 weeks or more had elapsed after immunostaining. For these four slides, a progressive decrement in mean FITC fluorescent intensity was noted. The average ER fluorescent intensity for the same tumor area dropped to 922 ± 250. A paired t-test comparing the mean intensities of tumor cells imaged less than 21 days after immunostaining with the mean intensities of tumor cells imaged 22 or more days after immunostaining was significant at p=0.03. We concluded that FITC fluorescent intensity remains relatively stable for approximately 3 weeks after immunostaining, but there is loss of fluorophore stability after a 3-week delay between staining and image acquisition, with a decrement in staining intensity as time elapses.
|
|
Discussion |
|---|
|
TopSummaryIntroductionMaterials and MethodsResultsDiscussionLiterature Cited |
|---|
Progress in fluorescent digital imaging microscopy and associated software enables quantitation of the expression of proteins in tissue sections. Quantitation of protein expression in formalin-fixed, paraffin-embedded sections of human tumors can enable correlation with important clinical data such as age at diagnosis, response to treatment, and survival. However, autofluorescence inherent in formalin-fixation of tissue and the day-to-day variability in microscope performance can hinder the acquisition of accurate and reproducible quantitative data.
Presented here is a protocol for fluorescent immunostaining of formalin-fixed, paraffin-embedded tissue, image acquisition, and software analysis of protein expression using a DeltaVision Restoration deconvolution microscope. Because of our own research interests, we have evaluated expression of the ER and PR in breast cancer. The protocol presented here, however, could be adapted to quantitative analysis of the expression of any protein that can be recognized by primary antibodies active in formalin-fixed, paraffin-embedded tissue.
Historically, background autofluorescence associated with formalin-fixed, paraffin-embedded tissue has been problematic (Mote et al. 1999; Chwirot et al. 2001; Ermert et al. 2001). In addition, background fluorescence secondary to nonspecific binding of antibodies must be controlled. We have presented data showing that background fluorescence can be kept to a minimum, with fluorescent intensities that are relatively consistent day to day. Particularly when FITC was used as the fluorophore, there was no difference in background fluorescence for slides receiving no, one, or two primary antibodies. When rhodamine was considered as the fluorophore, there was no difference in background fluorescence for slides receiving either no or two primary antibodies. An unexpected finding for rhodamine in this series of experiments was that slides receiving only one primary antibody had significantly lower background fluorescence than either the negative control slides receiving no primary antibody or the slides receiving both ER and PR as primary antibodies. There is no obvious explanation for this finding. Because rhodamine is an intrinsically brighter fluorophore than FITC, background fluorescence values with rhodamine are greater and can be more variable.
Critical features of the immunostaining protocol that assisted in the control of background fluorescence were careful deparaffinization of slides, pretreatment of the slides with sodium borohydride and glycine, and inclusion of both normal goat and normal donkey serum to block nonspecific antibody binding. An additional blocking step with normal goat and normal donkey serum before the addition of the second primary antibody further reduced background fluorescence for multiplexed slides. Donkey serum was included because both the secondary antibodies were made in donkey; however, we found that addition of goat serum in addition to donkey serum reduced background fluorescence even further.
The antibodies we chose to study, ER and PR, show marked differences in cell-to-cell expression in this breast cancer specimen, as in many of the cases we have analyzed. ER expression was relatively uniform from cell to cell, whereas PR expression demonstrated significant variability among neighboring cells. Because of the great intercellular variability in PR expression, we chose to evaluate the reproducibility of quantitative data by focusing primarily on the consistency of ER expression, because some degree of expression was seen in nearly every tumor and normal ductal cell. We considered consistency of expression both as a function of time and of staining condition. With our image acquisition and analysis methods, we have demonstrated that quantitation of expression of ER-positive normal and malignant breast ductal cells is not significantly affected by multiplexing ER and PR primary antibodies on the same slide or by delays of up to 19 days between immunostaining and image acquisition. With respect to the stability of FITC over time, we observed consistent mean cell intensities for ER expression in tumor cells on slides imaged within 19 days following immunostaining. After 19 days, there was a gradual and continuing decrement in mean cell intensities. Our data thus demonstrate fluorophore stability for FITC for about 3 weeks for slides stored in the dark at RT.
PR expression demonstrated large differences in nuclear localization and in level of expression within and between cells. It was not uncommon for intensely positive cells to be adjacent to completely negative cells. When protein expression is as variable from cell to cell as PR was in this specimen, assessment of the reproducibility of measurements of its expression, either as a function of staining condition or of elapsed time, becomes difficult. The risks of photobleaching from repeated imaging of the same fields on a slide must be balanced against the use of sequential sections that, by necessity, will expose somewhat different cell populations to antibody binding and subsequent quantitative analysis. Not surprisingly, we found that the distributions of brightly staining to weakly staining PR tumor and normal ductal cells were significantly different in the two staining conditions. Multiplexing was associated with relatively more brightly staining PR-positive normal ductal cells and fewer brightly staining PR-positive tumor cells. This difference may reflect a nonspecific effect of multiplexing or, equally probable, the evaluation of different sets of highly variable cells through sequential sections.
Critical features of this protocol to ensure data reproducibility are the use of Tetraspeck beads for initial adjustment of camera settings and daily and consistent checks of PSR, light bulb intensity, illumination alignment, exposure time, filter degradation, and artificial light from outside sources to monitor daily fluctuations in the microscope. Raw rather than deconvolved data were used to maintain scalar consistency from image to image.
Especially when the microscope and camera are in frequent use or are used by different investigators, controlling variability in quantitative data requires vigilance with respect to every manipulable camera and microscope parameter. In our experience, the goal of many investigators using high-resolution fluorescent microscopes and camera systems are to take a dramatic and beautiful picture from a single slide to demonstrate the expression of the protein(s) of interest. Adjustments made to the microscope and camera to produce the most beautiful and dramatic picture can be detrimental to ensuring reproducible quantitative data.
Few investigators have used fluorescent microscopy to quantitate and compare protein expression over a series of tumors. Ermert et al. (2001) compared quantitative estimates of CD-45 expression in rat leukocytes obtained by immunofluorescence using Texas Red as fluorophore, immunogold-silver epipolarization microcopy, and alkaline phosphatase-based Vector Red absorbance. They determined the linear range of Vector Red absorbance and fluorescence as a function of time and of antibody concentration. They documented photobleaching of all fluorophores examined and emphasized the difficulties in controlling day-to-day variability to obtain reproducible measurements. They did not compare CD-45 expression in a series of animals.
The inherent autofluorescence of melanocytes after formalin fixation and paraffin embedding was used diagnostically by Chwirot et al. (2001). Digital imaging was used to quantitate and compare epifluorescence of a series of malignant melanomas, basal cell carcinomas, and benign pigmented lesions. The authors document the variability in epifluorescent intensity of melanocytes as a function of location in the epidermis or dermis and the similarity in intensity for melanomas compared with benign lesions. Nevertheless, using a cut-off score, they could correctly classify 74% of the melanomas. The group did not evaluate melanocytes by immunohistochemistry.
A fluorophore-based analysis of expression of ER and PR in formalin-fixed, paraffin-embedded breast cancers, using flow cytometry, has been reported (Redkar and Krishan 1999). Cell suspensions of tumor cells prepared by enzymatic digestion were compared for DNA content and ER and PR expression. The parameter of interest was the frequency of ER- and PR-positive cells. The authors demonstrated a concordance of 70% for ER and 55% for PR expression comparing the results of flow cytometry and enzyme immunoassay.
In summary, developments in digital imaging, fluorescent microscopy, and associated software now enable quantitative estimates of protein expression in formalin-fixed, paraffin-embedded tissue. The availability of archived collections of paraffin-embedded tumors, which can be analyzed using these methods, can enable studies of the relationship between protein expression in tumors and clinical and outcome data. We have described immunostaining, image acquisition, and analytic methods that can be used to obtain reproducible quantitative data from such tumors. We have demonstrated that background fluorescence can be controlled. We have shown that for proteins that are expressed in nearly every cell, albeit to varying degrees, multiplexing two primary antibodies on the same slide does not significantly affect the results obtained. When FITC is the fluorophore, we have shown that time delays between immunostaining and image acquisition varying between 1 and 19 days can yield comparable results. We have also demonstrated that for proteins whose expression varies markedly from cell to cell, reproducibility of data obtained is more difficult to assess. We hope that the suggestions provided here with respect to the operation of microscope and camera and the use of associated software will enable other investigators to use these powerful tools in their own research.
|
Acknowledgments |
|---|
Support was provided by a grant from the Susan G. Komen Foundation (LMW) and a gift to the Moores University of California, San Diego Cancer Center from the Avon Foundation.
Digital imaging and analyses were performed in the Digital Imaging Shared Resource of the Moores UCSD Cancer Center.
Perspective views and movies were made by Alex Decastro at the San Diego Supercomputer Center VisLab using National Partnership for Advanced Computational Infrastructure Scalable Visualization Tools, in particular the Multi-Mesh Viewer.
Carolan Buckmaster provided valuable advice regarding immunohistochemistry protocols.
|
Footnotes |
|---|
Received for publication December 8, 2004; accepted December 20, 2004
|
Literature Cited |
|---|
|
TopSummaryIntroductionMaterials and MethodsResultsDiscussionLiterature Cited |
|---|
Agard D, Hiraoka Y, Shaw P, Sedat J (1989) Fluorescence microscopy in three dimensions. Meth Cell Bio 30:353–377
Agard D, Sedat J (1983) Three-dimensional architecture of a polytene nucleus. Nature 302:676–681[CrossRef][Medline]
Chen H, Clyborne W, Sedat J, Agard D (1992) PRIISM: an integrated system for display and analysis of three-dimensional microscope images. Proc Intl Soc Optic Engin 1660:784–790
Chen H, Hughes D, Chan T-A, Sedat J, Agard D (1996) IVE (image visualization environment): a software platform or all three-dimensional microscopy applications. J Struc Biol 116:56–60[CrossRef][Medline]
Chwirot BW, Sypniewska N, Swiatlak J (2001) Diagnostic potential of fluorescence of formalin-fixed paraffin-embedded malignant melanoma and pigmented skin lesions: quantitative study of fluorescence intensity using fluorescence microscope and digital imaging. Melanoma Res 11:569–576[Medline]
Ermert L, Hocke A, Duncker H-R, Seeger W, Ermert M (2001) Comparison of different detection methods in quantitative microdensity. Am J Pathol 158:407–417
Jagla W, Wiede A, Dietzmann K, Rutkowski K, Hoffmann W (2000) Co-localization of TFF3 peptide and oxytocin in the human hypothalamus. FASEB J 14:1126–1131
McNally JG, Karpova T, Cooper J, Conchello JA (1999) Three-dimensional imaging by deconvolution microscopy. Methods 19:373–385[CrossRef][Medline]
Mote PA, Balleine RL, McGowan EM, Clarke CL (1999) Colocalization of progesterone receptors A and B by dual immunofluorescent histochemistry in human endometrium during the menstrual cycle. J Clin Endocrinol Metab 84:2963–2971
Redkar AA, Krishan A (1999) Flow cytometric analysis of estrogen, progesterone receptor expression and DNA content in formalin-fixed paraffin-embedded human breast tumors. Cytometry 38:61–69[CrossRef][Medline]
Swedlow JR, Hu K, Andrews PD, Roos DS, Murray JM (2002) Measuring tubulin content in Toxoplasma gondii: a comparison of laser-scanning confocal and wide-field fluorescence microscopy. Proc Natl Acad Sci USA 99:2014–2019
Trinkle-Mulcahy L, Andrews PD, Wickramasinghe S, Sleeman J, Prescott A, Lam YW, Lyon C, et al. (2003) Time-lapse imaging reveals dynamic relocalization of PP1 gamma throughout the mammalian cell cycle. Mol Biol Cell 14:107–117
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  R A Walker Immunohistochemical markers as predictive tools for breast cancer J. Clin. Pathol., June 1, 2008; 61(6): 689 - 696. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Guidelines | Subscriptions | About | exPRESS - Current - Archive | Business Information | Contact |