As part of a study on graft-versus-host disease in the gut mucosa, Kohta Miyawaki, PhD, Physician-Scientist at Kyushu University partnered with Aaron Mayer, Chief Scientific Officer at Enable Medicine to perform multiplex imaging of gut GvHD samples using the CODEX platform. They discussed the results from their study in a recent webinar. In this blog post, we summarize some of the key points from their presentation.
What is GvHD?
Bone marrow transplantation is sometimes the only treatment option for patients with blood cancers like leukemia and lymphoma. In some cases, the donor cells attack a patient’s healthy tissues in a phenomenon known as graft-versus-host disease (GvHD).
In gut GvHD, the donor cells attack the patient’s gut mucosa, causing severe inflammation. More severe cases of gut GvHD have a survival rate of only 25%. The standard therapy for gut GvHD is the administration of glucocorticoids, but a substantial number of patients relapse and their prognosis is extremely poor. Determining who will respond well to such therapy is critical to enabling healthcare providers to design optimal treatment strategies.
In recent years, we have begun to understand the types of immune reactions that take place in the gut, noted Dr. Miyawaki. Various immune cells interact with each other in response to gut microbiota, playing a critical role in gut health. It is still unknown, however, which immune reactions cause the effects of gut GvHD.
Before describing the methods of the study, Dr. Miyawaki explained some of the limitations they faced with their samples. Most gut GvHD samples are obtained via endoscopic biopsy from severely ill patients, making it difficult to obtain a large number of samples. They are also commonly archived as formalin-fixed, paraffin-embedded (FFPE) tissues.
RNA extracted from FFPE is usually severely degraded, making it unsuitable for whole transcriptome analysis. Because tissues are already fixed, they cannot be used to generate single-cell suspensions, which also rules out global immune profiling through single-cell analysis methods like single-cell sequencing or flow cytometry.
With these issues in mind, Dr. Miyawaki’s team instead used two different technologies: probe-based gene expression profiling and the CODEX® system for ultra-high multiplex immunofluorescence, which can be used in both fresh frozen and FFPE tissues.
Using probe-based gene expression profiling in gut GvHD samples, the team observed that mast cell-related genes were upregulated in patients responsive to steroidal therapy, while macrophage-related genes were upregulated in non-responders. “These results motivated us to see whether we could stratify patients by mast cell or macrophage signatures and associate them with prognosis”, said Dr. Miyawaki.
“We performed multiplex imaging and analysis on CODEX…we stained the gut GvHD tissues with 38 antibodies and performed high-plex spatial analysis.”
In order to determine the mechanisms behind these observations, the team decided to perform multiplex imaging analysis with CODEX. To do so, they collaborated with Aaron Mayer at Enable Medicine, a service provider which specializes in high-dimensional spatial analysis methods. They stained gut GvHD samples with 38 antibodies and performed high-plex spatial analysis using the CODEX platform to validate and expand on the gene expression profiling results.
Multiplex imaging analysis with CODEX
“CODEX was particularly well suited for providing insights into these questions due to its ability to simultaneously look at a variety of cell types, their functional states, and interactions.”
After Dr. Miyawaki’s introduced the study and their initial findings, Dr. Mayer elaborated on how CODEX was used to perform further spatial analysis. CODEX, an ultra-high plex immunohistochemistry platform, enables researchers to label tissues with over 40 antibodies with single-cell, spatial resolution in situ.
The Enable Medicine team worked with Dr. Miyawaki to develop a CODEX antibody panel which included tissue markers, markers of cell types and cellular functional states.
|Marker||Cell Type||Marker||Cell Type|
|CD45RO||T cells, Memory||CD107a||Degranulation, NK cells, T cells|
|CD4||Helper T cells, Macrophage||CD121||Cytokine receptor|
|CD8||Cytotoxic T cells||CD274||PDL1, Immune checkpoint, Suppression|
|FoxP3||Regulatory T cells||CD278||PDL1, Immune checkpoint, Activation, Exhaustion|
|PanCK||Epithelium||Mast Cell Tryptase||Mast cells|
|DEC205||Dendritic cells||CD56||NK cells|
|CD11c||Dendritic cells||ECadherin||Cell adhesion, epithelium|
|Mast Cell Chymase||Mast cells||Chromogranin A||Endocrine cells|
CODEX images were acquired for all patients from the study’s GvHD cohort. This allowed the team to qualitatively observe how the expression of markers varied as a function of disease state. The image below from Dr. Mayer’s presentation highlights increased GvHD score in the gut from left to right.
While these images qualitatively reveal an increase in inflammation, breakdown of tissue architecture, and changes in biomarker expression, the goal of the analysis was to quantify the data to discover features that correlate with prognosis, patient outcomes, and disease progression. “CODEX was particularly well suited for providing insights into these questions,” said Dr. Mayer, “due to its ability to simultaneously look at a variety of cell types, their functional states, and interactions.”
Using cell segmentation algorithms, the team was able to segment out all the single cells from the gut biopsies and create a high-dimensional single-cell data set while preserving tissue samples. Each cell was represented mathematically by its expression of various biomarkers and its spatial coordinates.
In order to get an unbiased view into the cell types present in the gut GvHD samples, the Dr. Mayer’s team ran unsupervised clustering on all the single cells segmented from the images. Clusters were then identified and labeled based on their biomarker expression. Because CODEX retains the spatial coordinates for each cell, the team was able to map the cell types back on the tissue. They also mathematically indexed each cell and computed its interactions with its neighbors.
Dr. Miyawaki was able to use the results of this in-depth analysis to test his hypotheses about patient outcomes in gut GvHD.
Summary of the findings
Dr. Miyawaki ended the webinar by summarizing the key results from their study. His team determined that the frequencies of mast cells and macrophages are associated with GvHD prognosis. Macrophages are abundant in non-responders, while mast cells display high frequency in responders.
The team also analyzed cell to cell interactions based on proximity and aimed to understand which cellular interactions are most associated with prognosis. They found evidence that mast cell interactions may directly suppress immune cell activation in gut GvHD. Interactions between mast cells and lamina propria macrophages, CD11c dendritic cells, and CD8 T cells were strongly associated with favorable clinical outcomes. On the other hand, interactions between macrophages (lamina propria and CD11B) and CD8 T cells were associated with poor prognosis.
By combining CODEX data with gene expression data, Dr. Miyawaki’s team discovered that macrophage interactions are associated with inflammatory gene expression. Interaction between macrophages and lymphatic endothelial cells is tightly linked to ATG7 expression. ATG7 is an autophagy protein and it has been suggested that it regulates endothelial cell inflammation and permeability. High ATG7 expression was observed in the non-responder group — in these cases, macrophages and lymphatic endothelial cells were also spatially associated with one another. This data suggests that ATG7 expression induced through interactions with macrophages could contribute to inflammation in macrophage-rich gut GvHD patients.