Induction of a common microglia gene expression signature by aging and neurodegenerative conditionsa co-expression meta-analysis

For decades, the field of neuroinflammation operated on a simple binary. Microglia, the brain's resident immune cells, were either activated into an angry, destructive state, or they remained quiet. Researchers borrowed a framework from peripheral immunology and referred to these opposing states as M1 and M2: M1 typically triggered by lipopolysaccharide or interferon-gamma, and M2 by interleukins like interleukin-4 or interleukin-13. It was neat, experimentally manageable, and it dominated conversations for years. Then, Holtman and colleagues actually examined the gene expression of microglia in aging brains and in mouse models of Alzheimer's disease and amyotrophic lateral sclerosis, or ALS — and neither state appeared. Instead, they discovered a third profile, conserved across all those conditions, that lacked a clear name and a molecular map. Their paper aimed to construct one. Microglia are the brain's sentinels. Under normal conditions, they continuously survey neural tissue, clearing debris and supporting homeostasis. When damage occurs, they alter their behavior by releasing cytokines, engulfing dying cells, and remodeling tissue. What Holtman and colleagues describe as "primed" microglia exist in a unique position: at baseline, they are not releasing inflammatory molecules, but they are hypersensitive. When exposed to a pro-inflammatory stimulus, they overreact — releasing significantly more cytokines, chemokines, and reactive molecules than normal microglia would. This exaggerated response is thought to exacerbate neuronal damage in chronic disease. The challenge was that no one had defined what this state actually looks like at the molecular level, or whether it is consistent across different conditions. Those are the questions this paper addresses. The analytical approach was intentionally cross-conditional. Rather than asking which specific genes change, Holtman and colleagues investigated which groups of genes move together. They employed Weighted Gene Co-expression Network Analysis, or WGCNA, which clusters genes into modules based on how closely their expression profiles correlate across samples. Modules reflect co-regulated biological programs, rather than isolated findings. The datasets included pure ex vivo microglia from aged mice, an accelerated aging model known as Ercc1, an Alzheimer's model termed App-Ps1, and an ALS model called Sod1, plus brain tissue datasets, including rTg4510 and ME7 prion infection. Microglia were isolated through cell sorting and profiled across multiple platforms — Illumina beadchips, Agilent and Affymetrix arrays, and published RNA sequencing data. To ensure platform comparability, the team consolidated everything to gene symbols and focused on the intersection across all datasets, resulting in seven thousand five hundred twelve genes in the pure microglia analysis. Topological overlap matrices were computed for each model, scaled to a common quantile, and combined by taking the minimum overlap value across models — a conservative method that retains only the most consistent signal. Modules needed a minimum of one hundred genes to qualify, with strict statistical thresholds maintained throughout. The central finding was a consensus module of two hundred ninety-five upregulated genes, and a consensus module of two hundred five downregulated genes, shared across all four models. The consistency was remarkable: pairwise overlap p-values between the four individual upregulated modules ranged from approximately one times ten to the negative seventy-ninth to one times ten to the negative one hundred forty-sixth. These are not borderline results. Furthermore, the signal persisted in brain tissue — the primed microglia module significantly overlapped with App-Ps1 brain tissue, aging brain tissue, and ME7 prion infection datasets. What functions does this conserved program actually serve? Functional annotation pointed to pathways involving phagosomes and lysosomes, antigen presentation, oxidative phosphorylation, and Alzheimer's disease signaling. In simpler terms, phagosomes and lysosomes are the cellular machinery microglia utilize to engulf and digest debris; antigen presentation involves displaying protein fragments to other immune cells; and oxidative phosphorylation and mitochondrial signatures signify a shift in energy metabolism. Holtman and colleagues interpret this as microglia existing in a state of heightened phagocytic readiness — prepared to clear debris, but also capable of degrading healthy synapses and perpetuating degeneration if that readiness is chronically sustained. The hub genes — the most highly connected nodes in the network, and the candidates most likely to drive or represent this state — included apolipoprotein E, Axl, Clec7a, Itgax, also known as CD11c, and Lgals3, also referred to as Galectin-3. Four of these hubs that are unique to the primed network have well-documented roles in microglial proliferation, activation, and phagocytosis. Insulin-like growth factor one signals through Galectin-3; without Galectin-3, microglia become insensitive to insulin-like growth factor one, less active, and ischemic lesions grow larger. Colony stimulating factor one drives microglial proliferation, and blocking its receptor completely depletes microglia. Axl promotes the phagocytosis of apoptotic cells and myelin, and Axl knockout mice display increased inflammation and delayed myelin clearance. These are not abstract gene names — they act as functional nodes in a network that can be examined and potentially targeted. The team validated several of them by quantitative reverse transcription polymerase chain reaction in App-Ps1 and Ercc1 microglia, confirming the upregulation of Axl, Cybb, apolipoprotein E, Clec7a, and Cox6a at the RNA level. Immunostaining revealed Iba1 and Lgals3 double-positive cells in aged, Ercc1, and App-Ps1 brains, but not in controls. Now, here's where the comparison sharpens everything. Holtman and colleagues didn't just characterize primed microglia — they performed the same analysis on microglia from mice that received an acute intraperitoneal lipopolysaccharide injection, which is the standard laboratory trigger for acute inflammation. The acute network formed its own distinct module. When the team compared the two, the separation was clear. The acute module was strongly enriched for NF-kappa B signaling — NF-kappa B is the master transcriptional switch of classical inflammation, activating genes for cytokines, adhesion molecules, and survival factors — as well as toll-like receptor and NOD-like receptor signaling. In contrast, the primed module showed none of that enrichment. Instead, it emphasized lysosomes, phagosomes, oxidative phosphorylation, and Alzheimer's signaling. While there was some overlap between the primed and acute modules, it was much weaker than the overlap among the primed modules themselves. The overlap between the primed module and the aging microglia profile reached p-values between approximately one times ten to the negative twenty-fifth and one times ten to the negative forty-fourth. In contrast, the overlap with the acute module was around one times ten to the negative thirteenth — significant, but of a different magnitude. The authors are straightforward about the implication: priming is not a milder version of acute inflammation. It represents a qualitatively different state. This distinction carries immediate consequences for how we approach therapeutic targeting — strategies aimed at NF-kappa B or classical inflammatory pathways may miss primed microglia entirely. Beneath the shared primed core, the paper also identified disease-specific layers. Aging microglia, considering both physiological aging and the Ercc1 accelerated model, exhibited specific enrichment for ribosome activity and interferon alpha and beta signaling. The Alzheimer's App-Ps1 model demonstrated a disease-specific reduction in neurotrophin signaling, with a p-value of approximately three times ten to the negative fifth, and included hub genes linked to amyloid-beta clearance, notably the low-density lipoprotein receptor and CD14. The ALS Sod1 model featured a specific upregulated module associated with cell division and organelle organization. This layered architecture, comprising a shared primed core plus condition-specific overlays, is significant for the translational question. It distinguishes broadly conserved microglial responses from pathology-specific ones, providing both common targets and condition-tailored entry points. What are the implications of all this for our understanding of the aging and diseased brain? Holtman and colleagues present a compelling argument in their discussion. Since the primed state does not rely on NF-kappa B, does not resemble M1 or M2, and features enriched pathways focused on phagocytic readiness and metabolic rewiring rather than cytokine storms, strategies based on classical neuroinflammation frameworks may overlook the most relevant biology. The enrichment of phagosomes and lysosomes indicates that microglia in aging and degenerating brains are consistently mobilized for clearance — which sounds beneficial, but in reality may mean they are also at chronic risk of clearing things they shouldn't, including functional synapses. The hub genes Axl, colony stimulating factor one, Lgals3, and insulin-like growth factor one are not mere markers — they are nodes in pathways with known pharmacological ramifications. Precisely naming them within a coherent molecular map is essential for any effective application of that pharmacology. This paper's contribution is significant: it takes the concept of primed microglia, which existed only as a functional description, and provides it with a molecular architecture. A conserved transcriptional signature, validated across multiple models and platforms, includes specific hub genes and shows a clear distinction from acute inflammation. The therapeutic conversation about neuroinflammation in aging and disease can now begin with something concrete. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.