Mass Cytometry and Topological Data Analysis Reveal Immune Parameters Associated with Complications after Allogeneic Stem Cell Transplantation

A fascinating study taking a big data approach (systems-level analysis) to a very complex process (immune system regeneration after stem cell transplantation). Allogenic stem cell transplantation is used to reboot the immune system of leukemia patients and induce a graft-versus-leukemia response. The regeneration process takes a long time and involves the production of many different types of immune cells, as well as a complex network of serum protein interactions. The immune system regeneration fails in a significant number of cases, resulting in major complications (e.g. serious infections and leukemia relapse), but much of the detailed biology involved remains to be fully understood.

This study aimed to take a systems-level approach by looking simultaneously at many immune cell-types (by mass cytrometry) and potential protein markers (using the Olink INF panel). They did this by longitudinally monitoring 26 patients sampled at multiple time points between 1-12 months after allogenic stem cell transplants, looking for multi-parameter states associated with clinical outcome and complications such as acute graft versus host disease (aGvHD) and infections with cytomegalovirus (CMV).

Patients fell into 2 groups: group 1 (n = 10) regenerated a fully functional donor immune system and were cured of their disease without any serious complications. Group 2 (n = 16) had an unfavourable immune reconstitution, leading to multiple complications that were often fatal. Analysis of the different cell types indicated that in contrast to what is expected in healthy individuals, there were profound variations in immune cell composition in patients undergoing stem cell transplantation. These compositions varied greatly between patients, however, making it difficult to draw detailed conclusions from the cell data alone. For the proteins, 4 temporal patterns were evident, with either a transient increase or decrease at 3 months, or a gradual increase or decrease over the whole 12 months. These patterns showed functional clustering of proteins, e.g. many of those exhibiting the transient reduction at 3 months are involved in stress and wound responses, or general immune responses. Some proteins (e.g. CCL23 & TRAIL) also associated with the group 1 patients who had successful outcomes without complications, and may have potential as predictive markers for outcome (although the informative samples were very few in this pilot study).

The most informative insights were gained by combining all cellular and protein data together (~22 000 “immune measurements” in all) and using a topological data analysis (TDA) approach to get a big picture view. They therefore looked at the “collective states” associated with outcomes and complications. This enabled them to model combinations of immune cell composition and protein profiles that could distinguish patients well between clinical outcomes, and identify the specific complications of aGvHD and CMV infection, even with their relatively modest sample size. They propose that this type of high-resolution systems analysis could be very valuable in the development of future immunotherapies.