Delete Every Microglial Cell, Then Ask What Changed

Block one receptor and more than 90% of the microglia in a mouse brain are dead within a week. Not impaired — gone. Stop the block and they come back, repopulating from the survivors within another week. No other cell type in the central nervous system can be deleted and restored that cleanly, and CSF1R is the reason: it is not a growth factor receptor the cell can do without, it is a continuous survival requirement. That single fact is what turned microglia from a descriptive subject into an experimental one, because it lets you ask the only question that matters — is this phenotype caused by microglia, or does it merely happen while they are present? Click any protein for the matching Assay Genie ELISA kit or In Vivo antibody.

The microglial cell end to end: the survival and identity signals that hold the state in place, the neuronal brakes, the TREM2–DAP12 arm the Alzheimer's genetics keeps pointing at, complement-mediated pruning, the disease-associated state and the secretome it produces. Open the interactive version to click any protein for its role and the matching validated reagent.

Microglial identity is a state, not a lineage trait

Microglia are yolk-sac derived and self-renewing, and it is tempting to treat them as a fixed cell type with fixed markers. They are not. P2RY12, TMEM119 and high CX3CR1 are a transcriptional state, held in place by CNS-specific signals — chiefly IL-34 acting on CSF1R, and TGF-β. Culture a microglial cell and it starts losing that signature within hours. Put a monocyte into the brain and it acquires part of it.

The practical consequence is a trap that appears constantly in the literature: a marker-negative cell is not necessarily a monocyte. P2RY12 and TMEM119 both fall on activation, so in exactly the disease context where you most want to distinguish resident from recruited cells, the markers are least reliable. Fate mapping or parabiosis settles it; a single stain does not. Note too that the ligand split matters — IL-34 dominates in brain while M-CSF dominates elsewhere, so a rising M-CSF in CNS tissue is usually telling you about infiltrating myeloid cells rather than resident ones.

The neurons hold the brake, and the brake is real

Two inhibitory axes run from neuron to microglial cell. Fractalkine is made by neurons and tethered to the membrane until cleaved, and its receptor CX3CR1 sits at the top of the homeostatic signature. CD200 engages CD200R1, which recruits Dok2 and RasGAP and shuts down ITAM signalling at the level of SYK.

These are not modulatory niceties. Loss of CD200R alone is enough to leave microglia constitutively activated, which tells you the resting state is actively maintained rather than passively default. That framing changes how you read an experiment: an activated phenotype can mean a signal was added, or it can mean a brake was removed, and those have different upstream causes. CD200 is one of the few nodes here that is blockable in vivo, so the brake is testable rather than merely describable.

Complement prunes synapses — and disease re-runs the programme

In the developing brain, C1q tags weak synapses, C3 is deposited on them, and microglial CR3 — CD11b paired with CD18 — engulfs them. This is normal, necessary and how the visual system refines its connections. The uncomfortable finding is that ageing, Alzheimer's models and glaucoma models reactivate the same cascade against healthy synapses, and that C3 knockout preserves synapses and cognition in those models.

Synapse loss, not plaque burden, is what correlates with cognitive decline. That makes this cascade arguably the most therapeutically interesting thing on the map. Two cautions when interpreting it: C1q is made largely by microglia themselves, so this is an autocrine tagging system, and blocking C3 removes both the opsonic arm and the chemotactic C3aR arm at once — a C3 knockout is not a clean single-mechanism experiment.

Efferocytosis runs on a separate track. MERTK, bridged to phosphatidylserine by GAS6, clears apoptotic cells largely without inflammation. Which receptor did the eating determines whether the cell responds at all, which is why "phagocytic" and "inflammatory" are not synonyms and why a high lysosomal load is not by itself evidence of an inflammatory state.

TREM2, DAP12 and where the genetics actually point

TREM2 is a lipid and lipoprotein sensor with no signalling tail of its own; it borrows DAP12, an ITAM adaptor, which recruits SYK and then PLCγ2. One of its ligands is ApoE, which means the two largest genetic signals in Alzheimer's disease meet on this axis.

The genetics are unusually informative here because they point in both directions. Rare loss-of-function variants in TREM2 raise Alzheimer's risk substantially, and complete loss causes Nasu-Hakola disease. A hypermorphic variant in PLCG2 — the same pathway, one step down — is protective. A risk allele and a protective allele on the same signalling arm is about as strong an argument for causality as human genetics offers.

That arm licenses the disease-associated microglia (DAM) state: osteopontin, Dectin-1, CD11c, LPL and high CD68. Two things are worth saying plainly about DAM. First, the shift is metabolic before it is inflammatory — the lipid-handling genes move early. Second, whether DAM is protective or harmful remains contested, and the defensible position is that it is probably protective early and harmful late, which is precisely why timing matters more than presence when you design the intervention.

What each readout actually tells you

The In Vivo angle — one big tool and an endotoxin warning

Nine of the 37 nodes carry a functional-grade In Vivo antibody, which is fewer than most maps in this library. It is worth being straight about that: the sensome and the TREM2 arm are measurable, not blockable. But the one that is blockable is the one that matters most. CSF1R (CD115) is available in low and ultra-low endotoxin, and it is the depletion tool the whole field runs on. Around it sit CD11b, CD68, the CD200 brake and the output layer — TNF, IL-1β, IL-1R1, IL-10 and GM-CSF.

Endotoxin is not a formality on this pathway. LPS is the canonical microglial activator; it is what people use deliberately to produce the phenotype. An antibody carrying residual endotoxin therefore activates the exact cells you are counting, and the result is a beautifully activated brain and an uninterpretable experiment. Ultra-low endotoxin grades are the difference between a readable result and a wasted cohort, and the same applies to any recombinant protein you inject alongside.

On the measurable side, the inflammasome arm is where most of the interpretive errors live. NLRP3 is activated by fibrillar amyloid in AD models and its loss reduces pathology, so microglia are not passive responders to plaque. But caspase-1 has to cleave pro-IL-1β for any of it to matter, and a total IL-1β measurement that ignores the cleaved fraction tells you very little. IL-6 needs the same care in the other direction: microglia are a major CNS source, but CNS IL-6 and blood IL-6 are different measurements and should not be pooled.

The natural design: deplete with anti-CSF1R, then run neurofilament light and GFAP as a time course. If neither moves once the microglia are gone, the phenotype was never microglial — and you have learned that in one experiment rather than three. Add CCL2 to see whether the monocyte compartment is stepping in to replace them.

The one-line version

Identity is maintained by the environment, the resting state is actively held by neuronal brakes, complement prunes synapses in development and again in disease, and the TREM2 arm decides whether the cell enters the disease-associated state. CSF1R is the switch that lets you test any of it.

For the complement cascade in full, see the complement cascade and its regulators; for the don't-eat-me side of phagocytosis, see macrophage phagocytosis and CD47–SIRPα; and for the inflammasome arm, see the NLRP3 inflammasome and pyroptosis.

For research use only. Not for use in diagnostic or therapeutic procedures.