This article is a US Government work and, as such, is in the public domain in the United States of America.
Article first published online: 11 APR 2003
Published 2003 Wiley-Liss, Inc.
Cytometry Part A
Volume 53A, Issue 1, pages 9–21, May 2003
How to Cite
Price, O. T., Lau, C. and Zucker, R. M. (2003), Quantitative fluorescence of 5-FU–treated fetal rat limbs using confocal laser scanning microscopy and Lysotracker Red. Cytometry, 53A: 9–21. doi: 10.1002/cyto.a.10036
The research described in this article has been reviewed and approved for publication as an E.P.A. document. Approval does not necessarily signify that the contents reflect the views and policies of the Agency, nor does mention of trade names or commercial products constitute endorsement or recommendation for use.
- Issue published online: 11 APR 2003
- Article first published online: 11 APR 2003
- Manuscript Accepted: 5 DEC 2002
- Manuscript Revised: 19 NOV 2002
- Manuscript Received: 19 APR 2002
- confocal microscopy;
- LysoTracker Red;
- fetal rat limbs;
- fluorescence quantification;
- marching cubes;
LysoTracker Red (LT) is a paraformaldehyde fixable probe that concentrates into acidic compartments of cells and tissues. After cell death, a high level of lysosomal activity (acidic enzyme) is expressed in tissues resulting from phagocytosis of apoptotic bodies by neighboring cells. LT was shown previously to be an indicator of cell death in a manner similar to other standard assays (Annexin, terminal dUTP nick end labeling, Nile blue sulfate, neutral red, and acridine orange).
LT fluorescence in fetal rat hindlimbs at gestational day 14 was measured 8 h after administration of the teratogen, 5-fluorouracil (5-FU), with the use of confocal laser scanning microscopy (CLSM). Four dose levels of 5-FU (0, 20, 30, and 40 mg/kg) were studied. The preparation technique involved staining with LT, paraformaldehyde fixation, methanol dehydration, and clearance with benzyl alcohol and benzyl benzoate. After this treatment, the limb was nearly transparent and ready for CLSM analysis.
LT staining was observed in specific regions undergoing apoptosis in normal (control) hindlimbs. After 5-FU treatment, highly fluorescent regions appeared in the progress zone (PZ) of the limb. A dose-dependent response to 5-FU treatment was observed. Compared with controls, hindlimbs treated with 20, 30, and 40 mg/kg of 5-FU exhibited more fluorescence within the highly proliferative PZ. These results showed a dose–response relation between 5-FU exposure and LT uptake.
We found that three-dimensional volumetric regions indicating a high level of fluorescence in the embryonic limb bud can be quantified with three different computer analysis programs. The combination of a sample preparation procedure that clears tissue, a CLSM technique that addresses the equipment variables, and an application of statistical population analysis procedures enabled the visualization and quantification of fluorescence in entire fetal rat hindlimbs that were approximately 500 μm in thickness. Cytometry Part A 53A:9–21, 2003. Published 2003 Wiley-Liss, Inc.
Fluorescence quantification from thick tissues with the use of confocal laser scanning microscopy (CLSM) is a difficult task. Several factors can affect accurate fluorescence quantification, such as sample preparation, CLSM acquisition, and computer analysis. If these parameters are properly controlled, the fluorescent data obtained from tissues with the use of CLSM can be quantified. This work provides a new method by which to quantify three-dimensional (3D) fluorescence intensity in thick tissues and may yield useful data in measuring the dose-dependent effects of a toxicant.
Apoptosis occurs during normal embryonic development (1, 2) and has been shown to increase after exposure to a variety of toxic insults (3–5). This process of programmed cell death was initially defined by morphologic criteria that included the cell nucleus condensing with fragments and membrane-bound “apoptotic bodies” of various sizes being formed and then engulfed and phagocytosed by macrophages or macrophage-like neighboring cells (1). There is an increase in acidic lysosomal enzyme activity and phagosomes accompanying the removal of these apoptotic bodies by adjacent cells.
Apoptosis has been visualized in whole embryos with the use of vital staining dyes (acridine orange, Nile blue sulfate [NBS], or neutral red). These dyes are acidophilic and are concentrated in acidic areas and phagolysosomal areas of apoptosis (5–11). These classic vital dyes have been correlated to Annexin and terminal dUTP nick end labeling (TUNEL) stainings in embryonic and fetal tissues (9–16). The dye LysoTracker Red (LT) was observed to concentrate in the interdigital areas, representing apoptotic cell death in the gestational day (GD) 15 forelimb, which was similar to the staining patterns of classic vital dyes (NBS, neutral red, and acridine orange) used to determine cell death during normal embryologic development (1, 2, 5–7, 17). Sulik and coworkers compared NBS and LT stains and found that they stain identical apoptotic structures in the embryo (11–13). Because the pattern of limb and embryo LT staining is similar to other commonly used methods that reveal cell death, it can be hypothesized that LT also indicates cell death (11–16). LT has the advantages of being fluorescent, stable, and fixable. Although the exact correlation between LT and the extent of apoptosis determined by other methods remains, it is very qualitative in nature. It is currently difficult to quantify the amount of apoptosis in thick tissue with other apoptosis detection methods. This article describes a method to quantify a fluorescent dye in thick tissue that may have physiologic significance in developing embryos and fetuses.
Development of an embryo and a fetus involves inordinately complex processes, thus rendering evaluation of developmental toxicity a daunting task. An emerging approach to organize this risk assessment practice is the construction of quantitative biologically based dose–response (BBDR) models that are designed to capture all the key mechanistic events involved in the expression of toxicity (17–21). In the past several years, an attempt was made in our laboratory to construct such a BBDR model for developmental toxicity by using the chemotherapeutic drug 5-fluorouracil (5-FU) as a prototypic teratogen (18–22). In our model, 5-FU was proposed to inhibit thymidylate synthetase activity, thereby upsetting the intricate balance of intracellular deoxyribonucleotide pools, interfering with DNA synthesis, impeding cell cycle progress, and producing excessive cell death and ultimately anatomic malformations highlighted by limb defects (18–22). Thus, quantitative measurement of each of these parameters will be essential to provide a database for the model construction. The GD 14 rat hindlimb was chosen as a model because it has a defined shape that is relatively flat. The effects of GD 14 toxicity can be easily ascertained as gross malformations at birth.
Although most mechanistic events described in the 5-FU BBDR model are amenable to quantitative assessment, an accurate account for the 5-FU–induced cell death in an embryonic limb bud proved to be a challenge. Most of the conventional methods used to indicate apoptosis are qualitative by design (18). Thus, to evaluate the extent of cell death in the entire limb comprising a thickness of 400 to 700 μm, a specific sample preparation procedure was needed to stain, fix, dehydrate, and clear the tissue (14, 15, 23, 24). The vital dye LT was chosen to stain the tissue, and paraformaldehyde (PF) was chosen to fix the dye into the tissue. LT concentrates into acidic compartments in the tissue, correlating to lysosomal activity and phagocytosis of apoptotic bodies by neighboring cells. It has the advantage of being fixable, and it is more stable than NBS. The stained and fixed specimen is dehydrated in methanol and cleared in a solution of benzyl alcohol and benzyl benzoate (BABB) that provides a refractive index approximately the same as that of the fetal tissue, as previously described (14). The resulting tissue is nearly transparent. By suspending the tissue in BABB, the refraction, reflection, scattering, and absorption factors of laser light excitation are reduced. Without this critical step, the excitation and emission light would be attenuated, and the optical sections located internally would appear less bright. The sample preparation procedure used in this study has effectively increased the laser penetration, minimized light scatter and refraction absorption, and matched the refractive index of the tissue to the suspending medium (25). With this approach, we believe our laboratory has taken a critical step in the process of developing a method to quantify the fluorescence in a solid tissue. The CLSM data were analyzed by three computer-based approaches to quantify the amount of fluorescence detectable in the control and 5-FU–treated fetal rat limbs. The first method (the volume method) determines the amount of fluorescence by measuring the intensity in each element of the data set and enumerating over the entire 3D volume. The second method (the peak method) measures the intensity of the peak signals of fluorescence in the data set and counts only those peak signals above a threshold level. The third method uses the newly released Bitplane software (Imaris/Surpass) that contains the marching cubes algorithm to determine the number of particles and the surface area and volume of the particles derived above a specified threshold value (26, 27). This Surpass method combines features of the peak and volume methods in determining the number of apoptotic particles and the volume of the particles. All three methods were applied to 5-FU–treated limbs to illustrate that fluorescence from solid tissues can be quantified.
MATERIALS AND METHODS
Animal Treatment and Sample Preparation
Timed-pregnant Sprague-Dawley rats were obtained from Charles River Laboratories (Raleigh, NC). The presence of a copulatory plug was designated GD 0. On GD 14, pregnant rats were injected subcutaneously with 20, 30, or 40 mg/kg of 5-FU or saline (control) and killed 8 h later. The pregnant rats were killed by decapitation, the fetuses were removed, and the hindlimbs were dissected immediately and placed in warm phosphate buffered saline (PBS) before incubation. Groups consisting of four hindlimbs each were incubated in 0.5 ml of this medium at 37°C.
Sample Preparation: Staining, Fixation, Dehydration, and Clearing
The procedure to stain embryonic tissue with LT was described previously. Briefly, a vial of LT (Molecular Probes, Eugene, OR) containing 50 μl of a 1 mM solution was added to 10 ml of PBS to make a final concentration of 5 μM. PF (20%; Electron Microscopy Sciences, Fort Washington, PA) was diluted to 4% with PBS and stored frozen at −20°C (14). Limbs were washed twice with PBS after the 30-min staining period, fixed in 4% PF at 4°C overnight, and processed within 24 h. The fetal tissues were washed twice with PBS to remove the fixative and then dehydrated with methanol (MeOH). The limbs were cleared with BABB (1:2 by volume; Sigma, St. Louis, MO) to produce a nearly transparent limb bud (14, 23, 24). The limbs were placed first into a 1:1 solution of MeOH and BABB for a few hours and then into 100% BABB. The limbs were then transferred into an instrument shop made a ⅛-in. aluminum slide containing a ½-in. hole and sealed.
Laser Scanning Confocal Microscopy
The Leica CLSM (TCS-SP1, Leica, Deerfield, IL) consisted of a Leica inverted DMIRBE microscope and an Omnichrome laser emitting at three wavelengths (488, 568, and 647 nm). The 568-nm line using a triple-dichroic beam splitter–excited LT dye with a slit between 580 and 630 nm was used to measure the emitted light. To visualize an entire GD 14 rat hindlimb with sufficient fluorescent intensity, a Zeiss 5× objective (0.25 numerical aperture [NA]) was used. A Dell 420 workstation with two 933-MHz processors and 1 gigabyte of random access memory and an Nvida Gforce2 video 32-megabyte board were used for the analysis.
The CLSM was evaluated to ensure it was stable and produced good resolution, field illumination, and stable laser power as described in two recent publications from our laboratory (28, 29). The limb was optically sectioned into 20 to 35 sections with a constant 20-μm thickness between sections. Each 3D volume data set completely encompassed the entire depth (400–700 μm) of the limb tissue. The 40-mg/kg 5-FU–treated limbs containing the brightest fluorescence was measured first, and all confocal settings were kept constant for the other dose groups. These settings included laser power, photomultiplier (PMT) voltage, PMT offset, frame averaging, and step distances between adjacent sections. The size of each image section was 512 × 512 pixels and occupied a region of 1.9 × 1.9 mm. The fluorescent intensity at each voxel was detected and digitized into TIFF images retaining 8 bits of information, or 256 intensity levels. Thus the file size of a representative 512 × 512 × 30 × 8 volume data set was approximately 8 megabytes.
The following software products were used to analyze the raw data: TCS-SP1 (Leica Lasertechnik, Heidelberg, Germany), VoxBlast (Vaytek, Fairfield, IA), Image-Pro Plus (Media Cybernetics, Silver Springs, MD), and Imaris/Surpass (Bitplane, Zurich, Switzerland).
Each 3D maximum projection image of limbs was segmented into the following regions: whole limb, apical ectodermal ridge (AER), and progress zone (PZ; Fig. 1). The analysis focused on the highly proliferative region of mesenchymal tissue at the front of the limb paddle, the PZ. This region contains proliferating cells that are very reactive to 5-FU and other chemotherapeutic drugs and accumulate most of the LT stain (30–32). The PZ was arbitrarily determined to extend 450 μm from the most distal point of the limb into the interior of the limb from a two-dimensional image. A perpendicular line intersecting this interior endpoint was made, and all regions distal to this line were cropped from the image, as shown in Figure 1. The segmentation process for the AER can be performed manually or with an edge-detection technique (33) to remove it from the quantitative analysis procedures.
The PZ consisted of the proliferative limb tissue that was most sensitive to 5-FU toxicity.
Quantitation: Peak and Volume
The fluorescent frequency histogram of the 3D image data was produced with VoxBlast. This histogram displays the number of voxels in each intensity channel (from 0 to 255). The Reverse cumulative histogram (RCH) is interpreted as the proportion of voxels greater than or equal to each intensity channel. This proportion decreases monotonically from 1 to 0 (34).
A simple method of deconvolution, i.e., the peak method, was attempted in this study. The deconvolution procedure begins by setting a neighborhood size and geometry for the volume of 26 cubic voxels for each voxel. Then each voxel in the volume is investigated to determine whether that voxel has the greatest intensity in its neighborhood. If the algorithm determines that this voxel is a peak voxel in its neighborhood, then the voxel retains its fluorescent intensity value. However, if the voxel is not a peak voxel in its neighborhood, then the intensity level of this voxel is changed to 0. The resulting 3D image contains the distribution of the peak concentrations of fluorescence in the volume.
The Surpass software (Bitplane) was used in conjunction with the 3D Imaris software to visualize 3D objects. Because this Surpass method is inherently easier to use and combines conceptual features of the volume and peak methods, we reanalyzed our data set with the Surpass software. The application of this software to the data set allowed for the quantification of the number of particles, the surface area, and the volume of particles above a specific threshold value in the limb images. The Surpass program uses the marching cube algorithm that creates triangle models of constant density surfaces (26, 27). Originally applied to magnetic resonance data and topography, it was adapted by Bitplane to be used with confocal TIFF data stacks. Briefly, three different intensity thresholds (gray-scale values [GSVs]) were used to evaluate the data on the limbs. These thresholds were determined by measuring the intensity of individual fluorescent particles and the neighboring area and arbitrarily determining three acceptable GSVs (40, 60, and 80). A rectilinear region of interest (ROI) encompassing the front part of the hindlimb containing the PZ was measured, and the limbs were cropped by using rectilinear coordinates of the Surpass program.
By using CLSM, structures inside tissues or cells can be observed. As shown in Figure 2, the visualization of LT in individual optical sections demonstrated sufficient penetration of the LT dye throughout the limb bud to allow visualization of the tissue with a confocal microscope. A comparison of representative 3D reconstructed images (maximum projection) from control and 20-, 30-, and 40-mg/kg 5-FU–treated hind limbs is displayed in Figure 3. The treated limbs, in a dose-dependent manner, exhibit more regions of high intensity fluorescence (illustrated by dark areas) in the PZ than did the control limbs. The dark regions around the edge of the limb comprise the AER, where a rapid up take of dye occurred. The AER was stained more intensely in control limbs than in any of the 5-FU–treated limbs.
To illustrate the two methods of analysis (peak and volume), a maximum projection of a limb AER was used (Fig. 4). The initial image (Fig. 4A) was enlarged at 2× magnification. By enlarging a region of the limb containing the AER (Fig. 4B, 4×; Fig. 4C, 16×), the pixels could be visualized at greater magnification. Figure 4C shows the different intensity levels of the pixels, and Figure 4D shows only the maximum pixel intensities after eliminating the nearest neighbors surrounding the region. Thus the difference between the two methods is illustrated in Figure 4C (volume), where all pixels were counted, and in Figure 4D (peak), where only the maximum peaks in the data set were counted.
Although the image is shown in two dimensions, the actual analysis was conducted in three dimensions.
A modified RCH, previously described in the flow cytometry literature to differentiate samples by fluorescence intensity differences (34), was applied in a similar manner to differentiate fluorescent intensity differences from a stack of confocal images. The values of all pixels in the stack of TIFF images were added, normalized to 1, and displayed as a relative percentage according to the GSV (from 0 to 255). The first step is to count all the voxels with a given intensity value. Then, to generate the RCH, the voxels at or greater than a threshold intensity value are accumulated. These counts are then normalized to the proportion of total voxels. At a threshold of 1, all voxels are contained between that threshold and 255, so proportion of voxels is 0. At any other threshold, the relative proportion of voxels between that threshold and 255 are displayed. The RCHs from the volume or peak method of quantification are presented on a semi-log scale for improved visualization. In all situations, the 5-FU–treated fetal limbs had greater concentrations of LT and thus more voxel intensity in the higher channels than did the controls, whereas the controls had higher voxels intensities and less fluorescence at the very low channels.
Quantitation: Volume and Peak
The quantitation procedures were applied to the PZ region, shown in Figure 1, by using the unprocessed volume (Fig. 5A) and processed peak (Fig. 5B) data. All RCH distributions are presented on a semi-log scale to better visualize the entire histogram. The volume (Fig. 5A) and peak (Fig. 5B) methods of quantification showed increased fluorescence in the higher GSV regions of the treated limbs compared with control limbs. These differences also can be visualized as channel-by-channel ratios between treated and control limbs (Fig. 5C,D). The data displayed in this manner showed intensity differences of each dose relative to the control. The ratio between the two histograms when using both methods of analysis denotes the difference between the treated and control data at each possible threshold intensity level. In all cases, fluorescence in the PZ in the upper GSV regions was greater in the treated limbs (in a dose-dependent fashion) than in the control limbs.
To determine the relative activity, a window of intensity channels was chosen between a background (channel 32) and a saturation (channel 224) level. The relative activity between channels 32 and 224 was integrated with the three doses, with four fetuses per dose, to yield the following relative mean values: 1 for the control limbs, 10 for limbs treated with 20 mg/kg of 5-FU, 45 for limbs treated with 30 mg/kg of 5-FU, and 120 for limbs treated with 40 mg/kg of 5-FU (Fig. 6).
The Surpass software was applied to limbs by using three different thresholds (40, 60, and 80 on a 256 scale) to determine the particle number and the volume and area of the particles. The threshold level was determined by measuring the intensity of the fluorescent pixel, representing a highly fluorescent region, and a pixel in the adjacent surrounding background area and then choosing three GSVs to measure. The AER contained an area of brightly fluorescent particles that is usually connected to the proper threshold level after using the Surpass program. The 10 largest particles in this AER region were eliminated from analysis, which does not greatly change the particle count but does change the surface area and volume. The smallest particles (consisting of fewer than 10 triangles) were excluded from analysis. The four different treatment groups (control and 20, 30, and 40 mg/kg of 5-FU) containing fluorescent particles as defined by Surpass are shown in Figure 7. The volumes of particles in the limbs from the four treatment groups are shown in Figure 8.
The application of different threshold levels (GSV) in the analysis resulted in different values being measured. At a high GSV, the number of particles measured is reduced, with the particles being well separated. At lower threshold levels, the original particles become larger and thus occupy more surface area and volume, and there is an increase in particle number. In certain cases, these larger particles become connected, which increases the area and volume but not the particle count. Figures 7 and 8 show quantitative differences in LT fluorescence similar to that seen in Figure 3 with the three treatment doses.
This study described a quantitative method using CLSM and the fixable dye LT to assess the toxic effects of 5-FU treatment on GD 14 fetal limbs. Our laboratory previously developed an assay that uses LT to reveal an increased fluorescence in embryonic tissue after toxic insult. This increased fluorescence has been correlated to acidification of tissues and to other standard toxicity and apoptotic assays in embryos (9–16).
Characterization of Embryonic Cell Death
LT appears to be connected with the initial apoptotic phase and the clean-up phagocytotic phase, both resulting in acidification of the tissue (11–15). The LT dye was concentrated in the interdigital areas, representing apoptotic cell death in the GD 15 forelimb (data not shown). This stain pattern correlated to the spatial location of cell death staining with classic vital dyes (NBS, neutral red, acridine orange) and other apoptotic detection methods used to determine cell death during normal embryologic development or embryologic toxicology (11–15). However, the exact correlation between LT and other apoptotic probes is very much qualitative in nature because there are no specific known methods that measure the amount of apoptosis in thick tissue. Because there is an unpredictable nature of apoptotic body formation and its subsequent engulfment by neighboring cells, efforts to quantify apoptosis as a function of apoptotic body numbers are difficult and likely inaccurate. It can be concluded that the extent of cell death shown in Figure 3 was unmistakably greater in 5-FU–exposed fetuses than in control fetuses. However, a direct correlation between the various probes is not possible because all methods cannot accurately quantify apoptosis.
Many methods have been used for the detection and localization of programmed cell death (PCD) in embryos, including NBS, LT, and TUNEL. Each offers specific advantages. NBS, a vital dye used since the 1960s to detect PCD within the developing embryo (5–8, 11–13), stains acidic compartments of cells such as lysosomes and correlates with locations of apoptotic bodies in the tissues (9, 10). Although NBS is useful for quick, inexpensive analysis of large numbers of specimens, the use of vital stains is problematic due to penetration, retention, and visualization constraints. Similar to NBS, LT localizes to membrane-bound acidic intracellular compartments and is used to monitor the amount of phagolysosomal activity resulting from the engulfment of apoptotic bodies by adjacent cells. This probe has the advantages of being fixable, penetrating tissue extremely well, and being highly fluorescent even after tissue processing (dehydration and clearing). Correlations between LT staining and cell death labeling with the use of other lysosomotropic dyes (NBS, acridine orange, and neutral red) and apoptosis-specific biochemical assays (TUNEL and Annexin V) have been reported (11–15). Whereas NBS and LT were used for the morphologic identification of apoptotic bodies, TUNEL was employed to provide a biochemically based detection of apoptosis (10–12). Morphologic and biochemical approaches have indeed confirmed the induction of cell death after chemical exposure during fetal development.
We observed specific differences in the amount of cell death in control and 5-FU–exposed specimens. Historically, studies in the field of teratology have provided only limited quantification of embryonic apoptosis after teratogen exposure (4, 5, 35). Staining with LT provides a viable alternative to evaluate the extent of cell death in a solid tissue. The present results indicated a dose-dependent increase in fluorescence signals at the highly proliferative PZ of the embryonic limb after an acute exposure to the teratogen 5-FU. Although the exact correlation between LT fluorescence intensity in the fetal limb and the extent of apoptosis indicated by other methods (e.g., TUNEL, Annexin, and apoptotic bodies activation of caspases) remains to be established, our findings are consistent with the known actions of teratogens and the digital effects that are specific to damage in the PZ. Hence, it is likely that the differential display of LT fluorescence signal among the various 5-FU dose groups reflect the different extents of cell death induced by the teratogen, thereby providing a quantitative measure of its developmental toxicity.
The advantage of using cleared LT-stained specimens is that they are transparent, which allows for optimal confocal imaging throughout the entire embryo due to the reduction of light scattering effects and increased penetration of laser light. Despite limitations of lengthy sample preparation when using LT, increased technical difficulties, and cost associated with confocal imaging, the LT method of apoptosis detection provides an invaluable 3D view of apoptosis in the embryo. Sulik and collaborators used this approach and published three papers reporting that LT is similar to other probes in detecting PCD in embryonic and fetal tissue (11–13). Zucker and collaborators also found that LT reveals morphologic structures in normal and treated embryos that are suggestive of cell death (14–16). From these studies, we concluded that LT is a good probe to indicate the apoptotic process.
Apical Ectodermal Ridge
The AER is at the edge of the limb and is believed to be responsible for controlling limb growth and development (17, 31, 32). The AER appeared to be brightest in normal limbs and had relatively reduced intensity in limbs treated with 30 and 40 mg/kg of 5-FU. The high AER fluorescence in control limbs was hypothesized to be due to the highly responsive cells in this region that may be influenced by handling and the culture conditions. Alternatively, the fact that the AER area is very thin, with increased surface area, may enable it to acquire more label than the rest of the limb paddle (the edging effect). The attenuation of LT staining in the AER of the high 5-FU dose groups was quite puzzling. One possible explanation is that 5-FU inhibits AER activity and signaling capacity preferentially, and the cells in this highly reactive region of the drug-treated limbs are not growing properly and not generating the proper signals as compared with the rest of the limb. Because of this unusual finding, the AER variable was segmented and removed from the data set.
To begin the accurate quantification of fluorescence from a thick tissue sample with the use of low power objectives, there are a number of technical factors that must be understood and controlled (36–39). These factors are consistent sample preparation, uniform control of CLSM acquisition, and reliable data analysis with the use of valid mathematical and statistical approaches (14, 28–30).
Many instrument performance and acquisition parameters may affect the quality of the data obtained from the thick hindlimb samples. Zucker and Price (28–30) and other groups (39–41) have published a series of instrument performance tests to ensure that the equipment is working correctly. These tests include laser power, laser stability, field illumination, spectral registration, lateral resolution, axial Z resolution, lens cleanliness, lens functionality, and Z-drive reproducibility. The CLSM variables that appeared to be the most critical in affecting the quantification of fluorescence intensity in this study were field illumination, PMT stability, and laser stability.
Each sample was positioned and oriented in the same manner in the middle of the field to minimize the effects of non-uniform field illumination. The laser power should be relatively constant throughout the experiment in making quantifiable measurements. An increase or decrease in laser power would manifest itself as an overall increase or decrease in GSV throughout the limb. Fortunately, the laser source did not change greatly during the acquisition of our data set, and normalization values to adjust for laser instability were not used. Laser power of the system was checked for power fluctuations before the experiment, as previously described, and the power was monitored throughout the experiment.
The amount of variability due to noise in the PMT detection process can be improved by averaging consecutive frames (29, 30). However, excessive averaging can result in sample bleaching, thereby decreasing the fluorescent intensity and increasing the time of acquisition. In our study, we found that the bleaching of the LT in PF-fixed samples suspended in BABB is minimal when the ROI is averaged between four and eight times.
Although light attenuation within a Z stack can affect the fluorescence intensity, it was not a major factor in these samples. The fluorescent intensity between successive TIFF sections of a cleared limb did not decease greatly with depth of analysis. Therefore, mathematical treatment of the data to adjust for signal attenuation was not necessary (42, 43). It should be emphasized that proper sample clearing procedures are essential to observe throughout thick tissue and acquire signals that can be quantified (12–14, 23, 24).
The use of low power objectives (5× or 10×) with low NAs is advantageous because the laser light is not as sensitive to optical variations such as spherical aberrations, scattering, and absorption as with the use of high power objectives. Low power objectives also have the advantages of being able to view a large area of tissue with even field illumination and being able to view deeply into the tissue because of the very large working distance with reduced light attenuation. Their major disadvantage is a lack of resolution due to the inherent low NA of the objective. Thus the image consists of regions of fluorescence instead of fluorescent structures in the interior of individual cells. In summary, the use of a low power objective yields a lack of resolution, an increased depth of field, and a larger viewing area.
In sample preparation it is important to handle the samples in a consistent manner by controlling the concentration of LT and the length of incubation. The fixation procedure should use a fresh, electron microscopic grade PF solution diluted to 4%. The sample preparation and image acquisition procedures attempted to minimize the blurring due to poor PSF blurring (14, 28–30).
Quantification and Data Analysis
The image preprocessing routine known as the peak method was developed to isolate the peak concentrations of fluorescence within the volume data set. This method assumes that fluorescent light is detected at locations where cell death may have occurred. The program eliminates staining variable problems and acquisition variables. The use of the RCH allows an objective method for evaluating the groups of fluorescence histograms without limiting the analysis to one threshold intensity level (34). In a similar manner, Surpass uses different threshold values to quantify particles and volume fluorescence at a specific GSV. Fluorescence quantification based on voxel frequency histograms is a unique application inspired by flow cytometric fluorescence analysis procedures (34, 44, 45), The Surpass quantitative technique is based on the marching cubes algorithm initially developed for the magnetic nuclear resonance image data (26, 27).
After this study was completed and the manuscript prepared, new software was released from Bitplane that included characteristics of the peak and volume methods of analysis. Because this package is inherently easier to use, we have included data in Figures 7 and 8 that support the other findings and concepts presented in the text. Object-counting algorithms used to quantify fluorescent objects in digital images have been applied to 3D data sets. The marching cubes algorithm contained in Surpass is one such algorithm to evaluate the particles, volume, and area of fluorescence (26, 27). A similar procedure has been described to measure the volume and surface area of chondrocytes when using geometric modeling of confocal serial sections (27). Similar to the peak and volume methods, a number of factors that can affect the accuracy of the data should be addressed and optimized. These factors include sample preparation (11–15, 46) and instrument optimization (28–30, 39–41). In addition, the data are affected by these other factors: ROI, GSV, size of triangles in the marching cubes algorithm, and elimination of AER particle fluorescence.
Biologically Based Dose and Response
One goal of this study was to provide a quantitative method to acquire data that should reflect the extent of cell death induced by a chemical treatment and to apply the data in the construction of a BBDR model for developmental toxicity. In our BBDR model of fetal development, there was a relation between the administered dose and the number of intermediate steps that occur and result in limb malformations (17–22). Exposure to 5-FU inhibits cell proliferation, which is manifested by an alteration in cell cycle histogram when using flow cytometry (17). After this perturbation there is an increase in apoptosis and a subsequent rise in phagocytosis. It is useful for each step to be quantified to make an accurate working model. In this limb study, the lysosomal activity was measured with the fixable dye, LT. The exact connection between 5-FU treatment and the number of dead cells resulting from such treatment is not known. However, there appeared to be a direct relation between the amount of 5-FU administered and the amount of LT fluorescent staining in fetal tissues; moreover, because this LT staining pattern seemed to be correlative to other probes suggesting embryonic and fetal cell death, it can be inferred that we probably were measuring cell death. A correlation between LT staining patterns in the proliferative zones of GD 14 limbs after 5-FU treatment (Figs. 6–8) with abnormal limb development at birth can now be made (17, 21). The fluorescent data shown in Figures 6 through 8 can now be incorporated into a proposed BBDR mathematical model for evaluating developmental toxicity.
This study demonstrated the use of low power optics on a CLSM to quantify a fluorescent probe incorporated into living tissue. This protocol controlled the variables of staining, fixation, confocal acquisition, and computer analysis of the data set. LT is a PF-fixable probe that concentrates into acidic compartments in the tissue and was shown to be similar to other probes used to monitor apoptosis. Our results demonstrated a dose-dependent increase of LT staining in the hindlimb of rats exposed to 5-FU. The application of three computer analysis procedures provided numerical measurements of 5-FU–induced tissue damage. This information will be useful in toxicologic studies that require 3D visualization and quantification of fluorescent staining.
We thank Earl Puckett of the U.S. E.P.A. machine shop for constructing the aluminum slides to contain the limbs and the power detector holder that attached to the Leica confocal microscope stage. We also thank Dr. Robert Kavlock for his support and encouragement.
- 5In vitro approaches to the study of embryonic cell death in developmental toxicity. In: KimmelGL, KochharDM, editors. In vitro methods in developmental toxicology: use in defining mechanisms and risk parameters. Boca Raton: CRC Press; 1992. p 129–142..
- 7Vital fluorochroming as a tool for embryonic cell death research. In: PersaudTVN, editor. Advances in the study of birth defects. Volume 3. Abnormal embryogenesis. Baltimore: University Park Press; 1979. p 219–241., , .
- 10The study of cell death by the use of cellular and developmental models. In: LockshinRA, ZakeriZ, TillyJL, editors. When cells die: a comprehensive evaluation of apoptosis and programmed cell death. New York: Wiley-Liss; 1998. p 97–130..
- 20Biologically based risk assessment models for developmental toxicity. In: TuanRS, LoCW, editors. Developmental biology protocols. Volume 1. Totowa: Humana Press; 1999. p 271–281., .
- 23Confocal immunofluorescence microscopy of microtubules in amphibian oocytes and eggs. Methods Cell Biol 1993; 38: 231–264..
- 26Marching cubes: a high resolution 3D surface construction algorithm. Comput Graph 1987 21 163–169, .
- 33The image processing handbook. 3rd ed. Boca Raton: CRC Press; 1995..
- 36Quantitative fluorescence microscopy. Cambridge: Cambridge University Press; 1991. p 1–55..
- 37Fundamental limits in confocal microscopy. In: PawleyJ, editor. Handbook of biological confocal microscopy. 2nd ed. New York: Plenum Press; 1995. p 19–36..
- 39Tutorial on Practical confocal microscopy and use of the confocal test specimen. In: PawleyJ, editor. Handbook of biological confocal microscopy. 2nd ed. New York: Plenum Press; 1995. p 559–567., .
- 40Practical considerations for collecting confocal images. In: PaddockS, editor. Methods in molecular biology. Volume 122. Confocal microscopy methods and protocols. Totowa: Humana Press; 1999. p 35–57..