Investigating T-Cell Communities in Immune Tissues with StrataQuest 8
Introduction
Tonsils are complex immune tissues that habor a number of various cell populations. Warranting powerful immune responses against pathogens, they are often used as a model tissue for investigating how our immune system ensures the homeostasis and what role immune cells play in it. Tonsils also contain germinal centers – highly organized units, providing necessary conditions for B-cell activation and differentiation. T cells are another important immune cell population, represented by CD4+ T helper cells, which aid in activation of B cells, and CD8+ cytotoxic T cells, responsible for eliminating diseased cells. Dissection of the spatial relationships of these cell populations remains an open task, as they are complex and require sophisticated methods to determine their quantity and distribution. These solutions should capture the dynamics of the native tissue microenvironment while precisely detecting and phenotyping the cells, ensuring consistency and enhancing data quality.
In the following application note, we present a project conducted in cooperation with Prof. Sarah Ellis, Head of the ACRF Centre for Imaging the Tumour Environment (CITE), Australia. The image analysis of multiplex fluorescence staining of tonsil tissue was performed using StrataQuest, contextual image analysis software from TissueGnostics. The focus was on the quantification and spatial localization of T-cell subsets within and around germinal centers.
Immunofluorescence multiplexing techniques provide a significant amount of data and enable the investigation of multiple markers, which is important in immunological studies, as cellular phenotypes often require 2-3 markers for accurate identification. However, the main disadvantage that may appear with a high number of markers in one staining round is autofluorescence and channel bleed-through, which can make the phenotyping process more error-prone and lead to the bias in the results. Using TissueFAXS SPECTRA for multispectral imaging, which enables up to 8 markers in a single imaging round, in combination with StrataQuest image analysis software—including a spectral reference database and a user-friendly spectral unmixing engine — supports the unmixing of channels and produces a clean final image. Additionally, it includes several optional pre-processing steps, such as background substraction or image smoothing, to optimize the image for analysis.
Analysis of T-cell populations
The workflow begins with nuclei segmentation, based on the DAPI channel. This is a critical step which requires accuracy of the algorithm, as subsequent steps depend on it. StrataQuest offers workflow adaptability, allowing multiple parameters to be fine-tuned to the project, saved, and reused for consistency.
Figure 1. From upper left to lower right: original image, unmixed image, DAPI grey channel, nuclei segmentation mask on DAPI grey channel.
In addition to DAPI, the tonsil sample was stained for five markers: CD8 for cytotoxic T cells, CD4 for T helper cells, CD45RO for a memory-like subset (mainly in T cells), FoxP3 for regulatory T cells, and PanCK for epithelial cells, as shown in the Figure 2 below. In the following study we considered CD45RO+ T cells as memory T cells.
Figure 2. Overlay and individual marker images.
The next step focuses on setting up scattergrams for the given phenotypes, a process similar to gating in flow cytometry. To correctly identify mutually exclusive phenotypes, it is recommended to calculate a ratio between the overlapping signals and use it as input for further gating, as shown in Figure 3. Here, two ratios were used: one to differentiate overlapping signals from CD4 and CD8 markers versus PanCK, and another to differentiate CD4 signal from the CD8 signal. The cutoffs were set to allow the best possible differentiation, and corresponding quadrants were used as input to define the phenotypes. For example, events with a high CK ratio were used as input to define PanCK-positive epithelial cells.
Figure 3. Defining phenotypes with a gating strategy.
The phenotype scattergrams are usually gated based on the marker signal intensity and nuclei size using cutoffs. The forward connection allows the user to track the cell of interest by clicking on it, which then highlights the cell across the available scattergrams. The backward gating permits the user to check the results by selecting a quadrant and viewing the events associated with it on the image. The user can change the cutoffs and the image will be updated automatically in real time. Figure 4 shows CD8+CD45RO+ T cells selected from upper-right quadrant of the scattergram. A cell, marked by a red square, is shown on the same scattergram.
The Phenotype Diagram tool simplifies the gating process: the user selects whether positive (right quadrant) or negative (left quadrant) events should be included or excluded, eliminating the need to navigate through multiple scattergrams.
Figure 4. Forward-backward connection and Phenotype Diagram.
The end result of cellular phenotypes can be visualized as color-coded cellular masks over the overlay image. This makes it easy to verify the correspondence between the output to the actual distribution of a given phenotype. Figure 5 shows the representative images of the identified cellular phenotypes: T cells (CD4 and CD8-positive) in red, cytotoxic T cells (CD8+) in turquoise, helper T cells (CD4+) in green, memory T helper cells (CD4+CD45RO+) in yellow, memory T killer cells (CD8+CD45RO+) in orange, regulatory T cells/Tregs (CD4+FoxP3+) in pink and epithelial cells (PanCK+) in violet.
Figure 5. Identified immune cell phenotypes.
The numerical data such as the percentage of cells of a given phenotype, is indicated on the scattergram but can also be exported. Using StrataQuest’s statistical report tool, we calculated the percentages of several T-cell subpopulations in the tonsil to describe their relative constitution.
In this tonsil specimen, we assessed that T helper cells are the biggest population among T-cell subsets, comprising about 35% of all cells, followed by T killer cells (17.6%). Interestingly, CD45RO+ T helper cells constitute slightly more than 28% of the T helper cell population, and the ratio of memory T killer cells to the overall T killer population is about the same. Regulatory T cells make up only 0.9% of total cell count, but almost 80% of them are also positive for CD45RO. Therefore, StrataQuest allows for the straightforward extraction of numerical data, which can serve as a solid basis for a scientific study.
To deeper characterize the T-cell neighborhoods, we took advantage of StrataQuest’s new Phenotype Interaction tool, which allows users to select two phenotypes of interest and the maximal desired distance between them – in this example, 20 µm. The algorithm outputs the image, displayed cells connected by lines (orange for all in the given distance, yellow for closest neighbor), as in Figure 6. A corresponding scattergram can be generated which represents the amount of connections for a single cell.
In this example, we explored T-cell neighborhoods near epithelial cells and found that 18% of T helper cells are within a 20 µm range of a neighboring epithelial cell. Furthermore almost 60% of these cells are in the vicinity of 1-4 epithelial cells. On the contrary, 9% of CD8+ T cells are found within the same range, with most of them located near 1-4 epithelial cells as well.
Figure 6. Phenotype interactions.
Another way to explore the spatial distribution of the cells can be through tissue classification and distance maps based on it. We defined several tissue classes for dissecting the tonsil tissue: germinal center (red), mantle zone (green), epithelium (yellow), and stroma including T cell zones (grey). A distance map was generated from the follicle (germinal center + mantle zone) into ranges with 25 µm step.
To calculate the percentage of phenotypes in each distance range, a scattergram was created where the ranges are gated and color-coded. This way, we could identify that almost 30% of T helper cells are found inside the follicle, and about 17% can be found close to the follicle (up to 50 µm). Around 40% of total T helper cells can be found further than 100 µm from the follicle. T killer cells, on the other side, are much less abundant inside the follicle – solely 15% - while about half of their population is dispersed further than 100 µm from the follicle.
Figure 7. Phenotypes on the Distance Map.
A different perspective on the composition of T-cell populations in tonsil can be derived from 3D diagrams and violin plots, which are directly available in StrataQuest. An example of a 3D diagram in Figure 8 shows the distribution of color-coded T-cell subpopulations along three axes representing mean intensities of CD4, CD45RO and FoxP3. It is also possible to select a cell and track it across all diagrams. Violin plots allow to compare the marker intensity across multiple phenotypes. One plot shows CD4 expression in the T helper cell population and while another shows CD45RO across all T-cell subpopulations.
Figure 8. Composition of T cell populations in tonsil.
Lastly, manifold learning techniques such as SONG, t-SNE and UMAP provide unique insights into the data structure that other algorithms may not capture. They can be generated directly in StrataQuest and immediately visualize the distribution of various T cell subpopulations, as shown in Figure 9. This way, these graph types can reveal patterns, clusters, or relationships between various subpopulations, which might not be immediately apparent through traditional analysis methods.
Figure 9. Dimensionality reduction plots in StrataQuest.
Images, figures and numerical data can be exported in multiple file formats. Particularly, StrataQuest allows users to specify the figure resolution and few other parameters, streamlining the process of figure preparation for publication. It also outputs raw statistics, such as event count and percentage, but this data can be exported separately for statistical analysis.
To conclude, StrataQuest is a powerful tool for spatial phenotyping, including neighborhood analysis between various cell types, tissue classification and comprehensive quantification of subpopulations in their spatial context. For more information about StrataQuest, feel free to contact our scientific team.
