Pseudotime Analysis Identifies Receptor Tyrosine Kinase Ligands involved in the Temporal Development of Alzheimer’s Disease Associated Microglia

Introduction: Alzheimer’s disease (AD) is the most common cause of dementia globally and lacks robust treatment options. One component of the disease is inflammation in the brain, largely driven by the immune-like glial cells, microglia. In AD these cells transition their phenotype from homeostatic to a potentially destructive disease associated phenotype (DAM), worsening pathology. Prior research identified the ERK1/2 protein as hyperactivated in acutely isolated microglia from an AD mouse model and a correlation between ERK1/2 phosphorylation and diminished function. ERK1/2 is differentially activated by the binding of upstream receptor tyrosine receptors (RTK) with their corresponding ligands depending on the specific ligand-receptor pair.

DAM development does not occur instantaneously, rather it is a progressive and temporally regulated process. Prior studies have identified a two-stage development process, identified by marker genes, and regulated by receptor signaling. As the disease progresses there is enrichment in the later stage DAM and a reduction in homeostatic state microglia, underscoring a temporal development process. In this study, computational methods were applied to an AD mouse model single nuclear dataset to identify RTK ligands driving DAM development.

Materials and

Methods: A single nucleus RNA sequencing dataset from 5xFAD amyloid beta mouse model of AD was obtained from the Gene Expression Omnibus (GEO) database (GSE140511). 3 wildtype (WT) and 3 5xFAD mice samples were used for this work. The transcriptomes from these samples were processed per a standard Seurat (v.5.0.3) workflow. This preprocessing included an integration step to minimize sample to sample variance. The microglial population of cells was identified using the marker genes Hexb, Csf1r, C1qa, and P2ry12. This population was loaded into the computational platform NicheNetR (v2.1.0) to identify ligands suspected of driving the differential gene expression found in the microglia from 5xFAD relative to WT. This platform uses interaction and regulatory databases to estimate ligand regulatory potential and predict the ligands driving target cell gene expression in a single nuclear transcriptome.

The microglia population was removed from the larger cell population and re-clustered. The homeostatic and DAM population of cells were identified using the homeostatic markers, P2ry12 and Tmem119; stage 1 DAM markers, Trem2 and Apoe; and stage 2 DAM markers, Axl and Csf1. A single nuclear trajectory was identified for these cells using Monocle3 (v1.3.7), where the homeostatic population was labeled as the pseudotime origin and stage 2 DAM was confirmed as the end point, enabling the evaluation of ligand expression in pseudotime.

Results, Conclusions, and Discussions: NicheNet analysis identified several RTK-targeting ligands in the top ligands suspected of driving the gene expression in 5xFAD microglia cells (Fig. 1A), implicating them as potential drivers of the DAM phenotype. They include Csf1, Tgfb1, Pdgfb, and Gas6, which are largely expressed by microglia. Notably, each of these ligands, with the exception of Pdgfb, have been established as altering microglia behavior outside of an AD context. Of note, NicheNet does not distinguish between positive and negative regulatory potential for each of these ligands on their target genes.

The Monocle3 trajectory of DAM development was then used to identify the stage in DAM development that the ligand may regulate. Only Csf1 and Tgfb1 changed as a function of pseudotime, shown in Figure 1B. Csf1 signaling began early in DAM development and plateaued later. Alternatively, Tgfb1 expression declined early in DAM development and further declined in the late stage. These trends align with what is known about each ligand in microglia, where Csf1 is typically proinflammatory and Tgfb1 is typically anti-inflammatory. The other ligands, Pdgfb and Gas6, were generally expressed at the same level over time. These ligands are still suspected of regulating DAM, just not at a specific point of development.

This analysis has identified four ligands for RTKs that are potential drivers of DAM. These results will guide the identification of causal RTKs driving the dysfunctional DAM phenotype and train a computational model to control and promote protective microglia function using agonists and antagonists targeting these causal RTKs.

Acknowledgements (Optional): The authors would like to thank funding provided by the NSF CAREER 1944053 and NIA R01 NS114130 Awards.