SEARCH

SEARCH BY CITATION

Keywords:

  • histographic mapping;
  • lymph node;
  • immunofluorescence;
  • LSC;
  • quantitative histology;
  • multiple thresholding;
  • tissue cytometry

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. MANUAL VERSUS AUTOMATED SCORING AND EXAMPLES
  6. DISCUSSION
  7. CONCLUSIONS
  8. LITERATURE CITED

Background

In lymphatic organs, the quantitative analysis of the spatial distribution of leukocytes by tissue cytometry would give relevant information about alterations during diseases (leukemia, HIV, AIDS) and their therapeutic regimen, as well as in experimental settings.

Methods

We have developed a semiautomated analysis method for laser scanning cytometry (LSC) termed “multiple thresholding,” which is suitable for archived or fresh biopsy material of human lymph nodes and tonsils. Sections are stained with PI for nuclear DNA and up to four antigens using direct or indirect immunofluorescence staining. Measurement is triggered on DNA-fluorescence (argon laser, Ar) or on specific cell labeling. Due to the heterogeneity of cell density, measurements are performed repeatedly at different threshold levels (low threshold: regions of low cellular density, germinal center; high threshold: dense regions, mantle zone). Data are acquired by single- (Ar) or dual-laser excitation (Ar-HeNe) in order to analyze single- (FITC) up to four-color (FITC/PE/PECy5/APC) stained specimen.

Results

Percentage and cellular density of cell-subsets is quantified in different microanatomical regions of the specimen. These data were highly correlated with manual scoring of identical specimens (r2 = 0.96, P < 0.0001). With LSC, semiautomated operator-independent immunophenotyping in tissue sections of lymphatic organs with up to three antibodies simultaneously is possible.

Conclusions

We expect this tissue cytometric approach to yield new insight into processes during diseases and help to quantify the success of therapeutic interventions. © 2004 Wiley-Liss, Inc.

Cytomics as the approach to performing standardized high-content analysis of biological samples is of key importance to gaining a deeper insight into biological processes in health and disease (1). In this context, analysis of cells in their natural environment within a tissue is of particularly high value. However, quantitative analysis of cell subsets in tissue sections is difficult to perform and time-consuming, especially if lymphatic organs have to be investigated. The lymphoid organs are highly cellular and a large proportion of the cell content is composed of small lymphocytes. Due to the high cellular density, cell overlap is common, and this can render the analysis of the cell of interest difficult. Furthermore, the characterization of lymphocyte subsets, e.g., activated memory-type CD4+ or CD8+ T cells, often requires multicolor immunofluorescence labeling, which raises serious problems, especially if an antigen is dimly expressed.

Polychromatic analysis of different markers can be achieved on single cells by flow cytometry (FCM) and slide-based cytometry (SBC) on instruments such as the laser scanning cytometer. Routine FCM provides a rapid, sensitive, and quantitative means to simultaneously measure up to four biologic properties of individual cells, even with weak antigen expression. High-end flow cytometers equipped with two or more lasers permit assessment of nine (or even more) markers on each cell (2). Slide-based methods such as laser scanning cytometry (LSC) (3), laser scanning microscopy (LSM) (4, 5), and scanning fluorescent microscopy (6) offer the opportunity to analyze fluorochrome-labeled specimens. These methods are applied mainly to single cells from blood, tissues, or cell cultures (7–9), and are widely used in research and clinical diagnosis.

A major disadvantage of those methods that require cell suspension is that they sacrifice the microanatomy of the tissue. Thus, information on the localization of a given cell type within the tissue, and on the relation of one cell type to another cell type with which it may interact, is lost during the analysis. Such information is especially important when a lymphoid organ has to be investigated. The secondary lymphoid tissue (lymph node, spleen, and mucosa-associated lymphoid tissue) is the anatomic site where immune responses are generated.

With the use of LSC, six or more fluorescence markers can simultaneously be measured on cytological slides (3, 10, 11). However, only a few publications address the use of this method on tissue sections (12, 13). These authors (12, 13) analyzed nuclear markers in solid tumors composed of large cells with a broad rim of cytoplasm that allowed a distinct demarcation of the immunhistochemical signals.

The aim of this study was to adapt LSC for quantitative phenotyping of human lymphatic tissue. Such assay would extend our knowledge of protective and pathologic immune responses in vivo and could serve as a tool to monitor immune reconstitution of lymphoid tissue of patients receiving antiretroviral therapy.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. MANUAL VERSUS AUTOMATED SCORING AND EXAMPLES
  6. DISCUSSION
  7. CONCLUSIONS
  8. LITERATURE CITED

Sampling and Specimen Preparation

Five palatine tonsils and two lymph node samples were retrieved from the files of the Department of Pathology of the Bernhard-Nocht Institute for Tropical Medicine. The biopsies were performed for diagnostic purposes and showed follicular hyperplasia. Frozen sections of various thicknesses (5, 10, and 15 μm) were cut with a cryostat and fixed in pure acetone for 10 min.

For the detection of proliferative activity, sections were incubated with MIB-1 (1:200; Dianova, Hamburg, Germany) at room temperature for 30 min. Slides were rinsed in PBS, fixed in 4% paraformaldehyde for 10 min, and were rinsed in PBS again. Sections were then incubated with isotype-matched FITC-labeled secondary antibody (1:50; Dianova, Hamburg, Germany) at room temperature for 30 min. Slides were rinsed in PBS, and were incubated with anti-FITC-Alexa 488 (1:100; Molecular Probes, Leiden, The Netherlands) in the dark at 4°C for 30 min. Slides were rinsed in PBS again.

With or without previous MIB-1-staining, sections were incubated with single antibodies or different combinations of the following antibodies: CD1a-FITC, CD1a-PE, CD8-APC, CD62L-PECy5 (Caltag Laboratories Int., Hamburg, Germany), or CD45RO-PE (Dako, Carpinteria, CA). Indirect labeling by streptavidin-Alexa 488 (Molecular Probes, Leiden, The Netherlands) for 30 min was performed for biotinylated anti-CD4-stained or anti-CD8-stained (Caltag Laboratories Int., Hamburg, Germany) sections after blocking endogenous biotin by incubation wit the Biotin Blocking System (Dako, Carpinteria, CA) for 10 min in some cases.

In order to visualize DNA, e.g., for total histographic mapping (see below), sections were incubated with RNase (10 mg RNase A; Boehringer Mannheim, Mannheim, Germany, in 100 ml 2× SSC) at 37°C for 30 min, rinsed in PBS, and stained with propidium iodide (PI, 5 μg/ml; Molecular Probes, Leiden, The Netherlands) in the dark at room temperature for 5 min. Slides were again rinsed in PBS, covered with anti-fading medium (Dako, Carpinteria, CA), and mounted.

Laser Scanning Cytometry (LSC)

The instrumentation and the software of the laser scanning cytometer (CompuCyte, Cambridge, MA), as well as its clinical applications, were described elsewhere in detail (14, 15). On this slide-based instrument, in principle any sample immobilized on a microscope slide can be analyzed: the laser scanning cytometer used in this study was equipped with an Argon (Ar)-laser (488 nm) and a Helium-Neon-laser (633 nm), and with four photomultipliers with standard optical filters and mirrors for detection of FITC, PE, PI, and APC/Cy5 (3).

Measurement by LSC

The 40× objective (NA: 1.0) was used for all analyses. The region of interest on the slide was set manually by the operator and was scanned stepwise by the two lasers. The primary signals were displayed as a pixel-to-pixel map and were treated similar to image analysis; objects of interest are triggered by operator-defined settings (see below). For every triggered object, several parameters for the fluorescence channels, as well as the x- and y-coordinates of its exact position, were recorded. For each object, the background fluorescence was measured and subtracted from the signal; for surface markers, manual background subtraction was used, and for nuclear staining and markers, automated background subtraction was applied. This calculation is important, because background staining may be more pronounced and inconstant on glass slides. Automatic cell-by-cell background subtraction, however, is problematic in tissues. If labeled cells are well separated from each other, then background subtraction is a useful tool. If, however, labeled cells are densely packed in a tissue compartment, then stained adjacent cells will be regarded as background and the background subtracted will be too high.

In order to detect cells in a specimen, the instrument needs one or more signals to allow the software to regard this object as a cell or cell section of interest and to acquire its data. These signals are referred to as triggering signals. Light scatter signal is well suited for cell detection in single-cell preparations, but its application is not feasible in sections. In order to detect all cells on a slide (“total histographic mapping”), the best triggering signal is that of the nuclear DNA. However, scanning and analysis with just one fixed threshold level on the nuclear DNA will trigger only on a minor portion of all cells within a lymph node section. We, therefore, developed the technique of “multiple thresholding,” to take into account that 1) nuclear cross-sections with different diameters are present, and that 2) cell density varies across the slide according to microanatomic compartments within the tissue (16). To this end, we made use of a tool included in the proprietary WinCyte® (Compucyte Corp., Cambridge, MA) software called “merge files”: sections were analyzed repeatedly with step-wise increases in the threshold level of the triggering parameter (i.e., for PI, minimal area: 15 μm2 for total histographic mapping of a large section, or 5 μm2 for single follicle analysis). This approach leads to the analysis of cells in different microanatomic regions of the tissue, depending on the applied threshold level. As schematically displayed in Figure 1a, at low threshold levels, cells in loosely-packed regions are acquired, whereas at high threshold levels, cells in dense regions are predominantly detected. Tissue section analysis by multiple thresholding results in a datafile for each separate analysis at a given threshold level (Fig. 1b), stored (as .fcs files) in flow cytometry standard (FCS 3.0) format. Each of these files contains data for counted cells in morphologically different areas and, therefore, yields a different morphological and microanatomical display. In addition to the PI-channel, the appropriate fluorescence channels for the CD-labels are activated.

thumbnail image

Figure 1. a: Schematic model of multiple thresholding for quantitative scoring of cells in tissue sections. The top row shows four nuclei (A, B, C, D) at different distances from one another. In the graphics below, the fluorescence intensity distribution along the straight line is displayed. Three different threshold levels are needed to discriminate each of the four nuclei. b: Histographic mapping of a single tonsil section at different threshold levels. The x- and y-coordinates of the cells contoured by applying different threshold levels represent spatial maps of a lymph node section. One specimen stained for DNA by PI was scanned repeatedly at the threshold levels indicated. Note that the cells detected at each of the different threshold levels are located in different regions of the section. By this setup, all nuclei within this lymph node are included in the analysis; single nuclei triggered in more than one analysis are excluded in the final data analysis by the merging process. In this section, at a threshold level of 3,000–4,000, cells within the germinal centers are triggered, whereas at threshold levels above 6,000, cells within the mantle zones are triggered. In total, 150,000 nuclei were measured in this section.

Download figure to PowerPoint

Following these iterative analyses, all datafiles acquired at the different threshold levels were merged; objects triggered at the same x-y-coordinates were condensed to a single object. If a nucleus was detected several times, i.e., at different threshold level settings (nucleus A in Fig. 1a), then it will only be represented as one object after the merging process. For total histographical mapping, objects with identical (merged), as well as different (unmerged), x-y-coordinates were further analyzed, to avoid overrepresentation of some cells that were triggered several times at different threshold levels (see Figs. 1b–4a).

thumbnail image

Figure 2. Histographic mapping of a tonsil section segment. Triggering at different threshold levels detects different microanatomical regions of the lymphatic tissue. A section of a tonsil from a HIV-negative patient stained for DNA by PI and CD8-ALEXA488 was analyzed sequentially with different threshold levels (1,000–8,000). Data from the scans with low thresholds (1,000–4,000) and those with high thresholds (5,000–8,000) were merged into a low-threshold data file and a high-threshold data file, respectively. In the LSC scans (right), each dot represents one nucleus. Comparison with the micrograph of the identical area (left) reveals that low threshold levels trigger cells within the germinal center and the extrafollicular tissue, whereas high threshold levels trigger cells within the mantle zone and dense regions of the extrafollicular tissue. For further details see text. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com].

Download figure to PowerPoint

thumbnail image

Figure 3. a: Detection and histographic mapping of CD8+ (row A), CD4+ (row B), and CD1a+ cells (row C), in tonsil sections by LSC. Parallel sections of the identical tonsil were stained with PI and the anti-CD antibody as indicated, using ALEXA488 as fluorochrome. The specimens were analyzed by the multiple thresholding method. The left row shows the distribution of the maximum pixel values of ALEXA488. The CD1a labeling was used as negative control; all events brighter than the cutoff in C were regarded as CD8+ (A) and CD4+ (B), respectively. On the histographic maps, all nuclei are shown on the left (data obtained by merging all different threshold data; see Fig. 1b), positively-immunolabeled cells are shown in the center (cells brighter than the cutoff in the histograms), and negative cells are shown on the right. Note that cell distribution of positively-labeled cells is substantially different between the different stainings. b: Micrograph versus histographic mapping of CD8+ cells in tonsil sections. The left composite image of a human tonsil section shows the CD8+ specific fluorescence in green. The identical section was analyzed by the LSC. The histographic map of the cell nuclei (white) and CD8+ positive cells (blue) is displayed on the right. Note the striking similarity of the distribution of positive cells on the micrograph and in the LSC analysis. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com].

Download figure to PowerPoint

thumbnail image

Figure 4. a: Micrograph versus histographic mapping of CD8+ cells in secondary follicles. The micrographs show the identical follicles using red and green optical filters to detect PI and ALEXA488 fluorescence, respectively. The identical follicles were analyzed by the LSC. The cell distribution measured at low or high threshold values, as well as the distribution of CD8+ or CD4+ cells, is shown. Note the similar distribution of positive cells in the micrograph and in the LSC analysis and the lack of CD8+ cells in the germinal center. b: Four-color analysis of a single follicle by LSC. The schematic shows a longitudinal section through a secondary follicle and the respective cell regions. The dot plot below shows a real measurement of a four-color labeled follicle from a HIV-patient (MIB-1 FITCC, CD45RO-PE, CD62L-PECy5, CD8-APC) with identical orientation. In the x-y display, the regions with high (MIB-1+) and low proliferation (MIB-1 low) can be distinguished. In the dot plots below, the CD45RO/CD62L pattern of the cells acquired in these two regions is shown. (CD8+ cells were rare and were excluded from these analyses. See text for details.). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com].

Download figure to PowerPoint

For mapping of an immunologically well-defined subset, the triggering parameter was set to the respective fluorescent label, and the minimal area of an object was set to 10 μm2. Only one .fcs file was generated. With this approach, selected microanatomical regions were analyzed, such as single secondary follicles (Fig. 4a and b).

Slides were also analyzed by manual counting, in order to determine the number of nuclear sections per germinal center and the number of CD8+ cells per follicle. To this end, micrographs taken from secondary follicles previously analyzed by the laser scanning cytometer were enlarged. Using image analysis software, the germinal center corresponding to the gate used for the LSC analysis was encircled (see Fig. 2). Within this region, all nuclear sections and all CD8-positive cells, respectively, were manually counted by an independent observer. These data were compared with the results of LSC on the same slides (see Fig. 5).

thumbnail image

Figure 5. Quantitative histology. Secondary follicles in lymph nodes from HIV– and HIV+ patients stained for CD8-FITC and PI were analyzed by LSC triggering on PI with multiple thresholding. Upper row: The identical germinal centers of follicles were counted manually and by LSC. Manual and LSC scoring of total cell count per germinal center (GC) were highly correlated (upper left, n = 16 germinal centers). Also, the number of CD8+ cells per GC was similar (upper right, mean ± SD, n = 20). Lower row: Percentage distribution (left) and cell density (right) of CD8+ cells in GC and mantel zones (MZ) of secondary follicles in HIV– and HIV+ patients. The median (central lines), 25/75 percentile (boxes), 5/95 percentile (error bars), and maximal/minimal values (dots) are shown. Plots represent data from 20 and nine follicles (HIV– and HIV+, respectively). (Note that in the mantle zones, percentages and cell densities of CD8+ cells are unexpectedly high. See text for explanation.)

Download figure to PowerPoint

Data Display by LSC

Data from the .fcs files were analyzed and displayed by the proprietary software WinCyte™. A dot plot showing the x- and y-coordinate of every object was generated; this dot plot was gated on subsets of cells with specific expression patterns of antigens, as appropriate. The total number of nuclear sections or cell subsets per unit area, as well as the percentage of cell subsets, was calculated by statistical analysis (SigmaPlot 2000, SPSS Knowledge Dynamics, Canyon Lake, TX).

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. MANUAL VERSUS AUTOMATED SCORING AND EXAMPLES
  6. DISCUSSION
  7. CONCLUSIONS
  8. LITERATURE CITED

In initial experiment, we studied sections of different thickness. As expected, we found that 5 μm-thick sections were best suited for analysis by LSC and provided the clearest discrimination between different immunofluorescent signals. This is due to the fact that the thicker a section, the higher the probability that cells of different phenotype overlay one another, resulting in phenotypic “spillover” by adjacent cells; this is known as the Holmes effect. As a next step, we tested nuclear markers for cell recognition. We found that although the spectral properties of 7-aminoactinomycin-D would be best suited, since it does not interfere with PE, its dependence on DNA-condensation rendered it insufficient as a reliable marker for nuclei in the sections. In our hands, PI at 5 μg/ml concentration was the best nuclear marker for analyzing tissue sections. At higher concentrations (up to 10 μg/ml), the signals became insufficient for proper cell recognition. We also analyzed the sections with different magnifications using 10×, 20×, and 40× objectives. A reliable discrimination of cells in areas that are densely packed with small lymphocytes (e.g., T-dependent zone or mantle zone of secondary follicles) could not be achieved by low power fields. Therefore, for all measurements, the slides were scanned using the 40× objective.

Multiple Thresholding for Histographic Mapping by LSC

As regions of interest, we have chosen parts of germinal centers with mantle zone or with extrafollicular lymphoid tissue. These areas were selected because of their different cellular densities. With the approach of multiple thresholding, an automated analysis of large tissue sections is possible (Fig. 1b). Different structural entities within the section are analyzed by each of the threshold levels. In the example given in Figures 1b and 2, at threshold level 1,000, regions outside secondary follicles are acquired. At threshold level 3,000 to 4,000, denser regions including cells within the germinal centers of secondary follicles are analyzed; note that at these levels, regions with higher nuclear density, such as the mantle zones of secondary follicles, are not measured. At threshold levels of 5,000 and higher, nuclei in the most dense regions, such as the mantle zone, are acquired. Merging all these data into one composite datafile shows that in this example, per tissue section data of 150,000 or more nuclear sections were acquired (Fig. 3).

Aided by imaging, microanatomic regions in the section, such as secondary follicles, are defined. Inside these regions, the cellular density of labeled and unlabeled cells and their percentages can be quantified objectively. This is demonstrated in Figure 2. In a micrograph of a segment of a tonsil, secondary follicles with their germinal centers and mantle zones are well visible (Fig. 2, top left). In the same image (Fig. 2, bottom left), some of the follicles are highlighted by regions (note the differences in thickness of the mantle zones; see further explanation in Fig. 4b). The identical area was analyzed by the laser scanning cytometer using the multiple thresholding method (minimal nuclear area: 15 μm2). Then two composite files were generated; one for all nuclei in the low density regions up to a threshold value of 4,000 (low threshold), and one for all nuclei above 4,000 (high threshold). The data are displayed in the same magnification as the images and were overlaid by the regions from Figure 2 (left bottom image), indicating that by appropriate threshold setting, microanatomical regions can be selected (Fig. 2, right). This is shown in more detail in Figure 4a for secondary follicles that have been analyzed by multiple thresholding. In these samples, however, the minimum areas for nuclear density were set to a lower value (5 μm) than for the whole tissue section, in order to detect most of the nuclear cross-sections in the mantle zone. This reduction of the minimum area was important in order to acquire more nuclei in the very densely-packed region of the mantle zone with quiescent B-cells that have only a marginal cytoplasm. If the minimal nuclear area selected is set to a high value (Fig. 2), then the number of nuclei counted in these regions will be low, and many cells will be discarded from the analysis. If the minimal area is lowered, then data of more cells will be included (Fig. 4a).

Note the differences in sectioning planes of the follicles in Figure 4a. Whereas the sectioning plane is longitudinal in the first, second, and third row (similar to that displayed in the scheme in Fig. 4b), sectioning of the follicle in the fourth row was not perfectly longitudinal.

Multicolor Analysis by LSC

PI-stained specimens can be counterstained and analyzed using green emitting dyes, such as FITC or ALEXA488, and red excitable dyes like APC. An important issue for using immunofluorescence labeling is to discriminate stained from unstained objects. To this end, control samples have to be analyzed that represent “background” fluorescence in the respective color. In contrast to control samples for single-cell analysis, unstained specimens or staining in the absence of the primary antibody was useless as control, because these samples were completely negative. Therefore, using this cutoff all cells appeared as positive in a sample with antibodies that had actually stained only a subset of the cells in the tissue. In contrast, control staining with normal species-matched control IgG produced a background too high for application in multicolor analysis. In our hands, to set the cutoff in multicolor analysis, it was best to use a specific staining against an irrelevant CD-marker as an internal control, such as the marker CD1a, which is not expressed by lymph node T- and B-cells or on most interdigitating and follicular dendritic cells (Fig. 3a). As shown in Figure 3, staining of parallel sections with anti-CD8 or anti-CD4 antibodies allowed us to discriminate well between positive and negative cells. In this case, it proved to be helpful to display the location of positive or negative cells within the tissue. This demonstrates that CD8+ and CD4+ cells are differentially localized, as expected. Figure 3b is an overlay of all nuclei and CD8+ labeled cells, showing that the latter are detected by LSC in microanatomical regions where they can be found on the micrograph of the whole tissue section (Fig. 3b, left). In more detail for selected secondary follicles, Figure 4a shows that CD8+ cells are mostly localized outside the germinal centers, but CD4+ cells exist inside the germinal centers.

It is noteworthy, that immunofluorescent labeled cells are best scored in the resulting .fcs datafile using the maximum pixel values of the respective fluorescence, instead of the integral fluorescence values. Integral fluorescence values were useful for the analysis of nuclear fluorescence, but did not yield data that allowed the discrimination of immunolabeled cells. This is due to the fact that the areas of cell cross-sections in thin tissue sections are heterogeneous, and thereby the standard deviation of the integral fluorescence values between cells substantially increases.

Four-Color Analysis

In order to demonstrate the applicability of multicolor staining in tonsil and lymph node sections, we applied different antibodies. Figure 4b schematically depicts the structure of a longitudinally-sectioned secondary follicle. Note that the light zone of the germinal center is located at the thick part of the mantle zone and contains less proliferating cells, whereas the dark zone contains more proliferating cells and is next to the thin part of the mantle zone. The data in Figure 4b demonstrate the feasibility of four-color analysis. Using MIB-1 FITC as a nuclear proliferation marker, CD45RO-PE, CD62L-PECy5, and CD8-APC were analyzed. Data acquisition was triggered on MIB-1 and the threshold was set to a low value in order to allow acquisition of high MIB-1-expressing (proliferating), but also low/negative MIB-1-expressing (nonproliferating) cells. Using color coding, the x-y plot shows that high MIB-1-expressing nuclei are located at one part (dark zone) of the germinal center, and low MIB-1-expressing nuclei are at the other part (light zone). These zones also correspond to the thickness of the mantle zone (thin and broad, respectively). In these analyses CD8+ cells were rare and excluded from the analysis (not shown). Based on the dark and the light zone as determined by MIB-1-expression, in this display two dot-plots were created (bottom) showing the CD45RO/CD62L expressions. As expected, cells in the dark zone are mostly CD45ROlow and CD62Lhigh, but in the light zone, many CD45ROhigh/CD62Lneg cells are present.

As demonstrated here, PE labeling with directly fluorochrome-conjugated antibodies is applicable in tissue sections, although it is not as useful as FITC or APC due to its substantially higher photosensitivity (15). Importantly, labeling with a directly fluorochrome-conjugated antibody is feasible in the case of highly-expressed surface antigens such as CD8, CD45RO on T-cells, or CD14 on monocytes, among others (17). This further simplifies staining and facilitates polychromatic analysis, as it also could be achieved with quantum dot labeled reagents (18).

MANUAL VERSUS AUTOMATED SCORING AND EXAMPLES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. MANUAL VERSUS AUTOMATED SCORING AND EXAMPLES
  6. DISCUSSION
  7. CONCLUSIONS
  8. LITERATURE CITED

How well does quantitative tissue analysis agree with manual scoring? We found a very high correlation of manual versus LSC-based scoring of nuclear cross-sections in the germinal centers of secondary follicles (r2 = 0.93, slope: 0.945, P < 0.0003; Fig. 5, top left). Accordingly, counting CD8+ cells as positive nuclear sections per follicle by LSC and by manual scoring yielded values in the similar range (Fig. 5, top right), although by LSC, the scores were slightly higher than by manual scoring. In the latter example, the specific fluorescence for CD8 was used as trigger signal in the LSC.

We used multiple thresholding for scoring of all nucleated cells in the mantle zone and in the germinal center of lymph nodes from HIV-positive individuals (Fig. 5, bottom). The nuclear density of all cells within the germinal centers and in the mantle zones did not significantly differ between HIV+ and HIV– patients (not shown). However, the percentage of CD8+ cells among all cells, as well as their areal density was dramatically increased in HIV-positive as compared to HIV-negative individuals, as expected (all P < 0.0001). Nevertheless, about 60% of CD8+ cells in the germinal centers of HIV-positive individuals seem to have very high values. This may be due to the fact that the sectioning plane of the follicles in these patients was not longitudinal and, therefore, a region was analyzed where CD8+ cells were overrepresented. Germinal centers of HIV-positive patients were in general larger and contained lower numbers of CD4+ cells (not shown). Surprisingly, the frequency and cell density of CD8+ cells in the mantle zone was higher than expected by optical judgement (virtually none). This observation error with the LSC may be due to several factors:

1. In the mantle zone, very densely packed cells such as B-cells will still not be separated as single cells and discarded from the statistics. Therefore, the percentage of CD8+ cells is overestimated.2. For the analysis of the mantle zones, positioning of the microanatomical gates based on the x-y displays is difficult because there is no direct morphological correlate for exact gate positioning. Therefore, it can not be ruled out that germinal center and extrafollicular tissue fell into this gate as well, which could determine the unusually high density of CD8+ cells in the germinal center of HIV-positive patients.

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. MANUAL VERSUS AUTOMATED SCORING AND EXAMPLES
  6. DISCUSSION
  7. CONCLUSIONS
  8. LITERATURE CITED

Up to now, tissue sections were diagnosed purely based on the subjective experience of the pathologist interpreting the specimen. Whereas the importance of visual evaluation of a section by the experienced pathologists is beyond doubt, the reproducibility of quantitative analysis has often be questioned (19). The development of slide-based cytometric instruments that are able to analyze any specimen mounted on a slide opens the opportunity to objectively perform quantitative histological analyses of tissue sections. Quantitative histology is of special interest when more than one parameter can be measured. Here, we present two different ways (multiple thresholding, and triggering on isolated well-defined cell subsets) to perform analyses of up to four parameters simultaneously in tissue sections on the slide. Both approaches make use of multicolor fluorescent labeling of antigens. Our developments show the feasibility of standardized semiautomated and polychromatic analysis of lymphatic tissue.

In multiple thresholding, the number of different threshold levels that are needed to measure virtually all cells of a section differs between preparations, but is consistent for a certain specimen. In most of our experiments, five to eight different threshold levels were used. However, this approach is very time consuming when measuring large specimens such as a tonsil. This is due to the fact that, in the LSC, sample analysis has to be performed at each threshold setting. Alternatively, multiple thresholding could be performed by off-line analysis of virtual slides of the tissue sections. To this end the section could be mosaicked into detail images that then are analyzed by slide-based cytometry algorithms as in the LSC. This could be done by present instruments such as the scanning fluorescence microscope (6).

If the measurement is limited to selected microanatomical compartments (e.g., secondary follicles), it takes only 2–5 min to measure all thresholds. The different .fcs files are merged into one .fcs file after completing the analyses, and then data are interpreted as if they were obtained by a single run. Merging will reduce the number of events, since some cells will be triggered more than once at different threshold levels. This is in contrast to the initial intention of that software tool—its main application is the combination of data from different measurements on the same cells (20). The proprietary software of the LSC, WinCyte™, also offers a second tool that can be used for the analysis of tissue sections, called “phantom contouring” (21). This tool generates predefined trigger contours in a random pattern across the scan area and treats the pixels within these contours as if they represent a cell. However, the data only represent the mean fluorescence at the covered arbitrary trigger area. This approach is not applicable for the quantification of specific cell subsets by fluorescent labels; if a single cell stained brightly for a given antigen was covered by three or more phantoms, the corresponding cell subset would be overestimated several times. Therefore, phantom contouring yields information about the tissue architecture on a microanatomical level, but not on a cellular level.

Based on the immunofluorescences, the spatial distribution of labeled cells within the section is documented by electronic gating. As an example, a great proportion of cells within the germinal centers of a secondary follicle are CD4+, whereas CD8+ cells are hardly found there. However, CD4+ cells are rare in the mantle zone of the follicles. CD8+ cells rather locate to the extrafollicular tissue. It can be tested whether the selected antibodies stain cells in microanatomical regions where the respective cells are known to be located; the observed distribution of the respective cell subsets corresponded to that known from their normal distribution in biological tissue, as expected. This demonstrates that the detection of immunophenotypically different cells by the LSC is specific and is not due to spectral overlap or other optical phenomena.

There are, however, still some improvements to be made. First of all, it is virtually impossible to judge whether in fact all cells have been analyzed in a section, since it is impossible to directly correlate the acquired data with the section. Second, if cells within defined microanatomical regions are to be analyzed, the accurate positioning of gates based on the x-y data display as a surrogate is not always possible to do as precisely as necessary (in particular, this is obvious for the mantle zones and leads to the overestimation of the CD8+ cells shown in Fig. 5). These minor obstacles could be overcome by some software improvements, such as allowing measured data to overlay with the image of the analyzed parts of a section and manual interference of the observer. Third, no tissue section is infinitely thin, therefore, in finitely thick sections, due to the Holmes effect (22), cells or cell structures will overlap and the tissue cytometric measurement will overestimate the frequency of stained cells (e.g., Fig 5, top left). A solution for the Holmes effect in tissue cytometric analyses was proposed by Irinopoulou et al. (23).

Cytomics analysis is the integral high content analysis of various cell constituents on the single-cell level (1). It requires the data per cell to be determined for the cell as a whole. However, this is not fully achieved by the method proposed in this work, since tissue sections have to be as thin as possible in order to adequately allow the software to discriminate single cells in dense regions such as the mantle zones. Therefore, discrimination of immunolabeled cells is best achievable by using the maximum pixel value, not by the integral fluorescence intensity value. Future developments may provide solutions that combine both cell discrimination in dense regions and high content single cell analysis, for example, by applying confocal acquisition, 3D reconstruction, and 3D cytometric image analysis. This could pave the way to tissue cytometry as a new cytomic technology, i.e., quantitatively analyzing cells and their interactions within their given environment. Considering blood as a special tissue consisting of single cells, flow cytometry could be judged as the first tissue cytometric approach in this regard.

CONCLUSIONS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. MANUAL VERSUS AUTOMATED SCORING AND EXAMPLES
  6. DISCUSSION
  7. CONCLUSIONS
  8. LITERATURE CITED

LSC provides a versatile tool for the analysis of the spatial distribution of different cell subsets within tissues. We established a semiautomated operator-independent assay for the analysis of complex tissues such as lymph nodes and tonsils, which are among the most densely packed organs. Counting cells of special subsets by hand can take up to days for one follicle, but only minutes with the LSC. The assay shown here has high accuracy and reproducibility in terms of spatial cell position and relative subset distribution within spatial compartments of the organ. This technique of quantitative histology should prove to be a powerful tool for investigating disease progress in various entities (e.g., infectious diseases, malignant disease), to better quantify the success of therapeutic interventions, and to improve prognostication.

LITERATURE CITED

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. MANUAL VERSUS AUTOMATED SCORING AND EXAMPLES
  6. DISCUSSION
  7. CONCLUSIONS
  8. LITERATURE CITED