How the Brain Removes Its Own Connections: Part One: The Cell That Prunes

Abstract

The brain does not only build connections. It removes them, deliberately, using cells that belong to the immune system — and it does this throughout life, not only during development.

This is the first of four parts, one per publication. It follows a 2026 review in Current Neuropharmacology to establish what microglia are and what they do to synapses when nothing is wrong.

The parts that follow take the molecular signal that marks a synapse for removal, the discovery that microglia are not the only cell doing the removing, and the closest thing this field has to human evidence.

1.What Microglia Are

Microglia are resident immune cells of the central nervous system and account for approximately 10 per cent of cells in it. They are not neurons, they are not supporting scaffolding, and for most of the twentieth century they were treated as janitorial — cells that cleared debris after injury.

The review by Xie and colleagues describes them differently: highly motile, extensively branched cells that continuously monitor the microenvironment, and which under physiological conditions promote synapse formation and maturation as well as removing connections.

They are also highly heterogeneous, owing to what the authors call high transcriptional plasticity. Single-cell profiling of microglia isolated from different regions of the human brain after death identified four distinct clusters with different regional abundance — one specific to the thalamus and subventricular zones, others concentrated in temporal and frontal lobes.

This matters for how the field's findings should be read. "Microglia do X" is a claim about a population that differs by brain region, by sex — genes in the type I interferon pathway are more expressed in female microglia — and by age, with circadian clock gene expression suppressed in microglia from older brains.

2.Pruning as a Normal Process

Synapses are the principal mode of communication between neurons, and they change in number and strength throughout life. That turnover is the physical substrate of learning and memory.

Circuit assembly therefore requires a balance between two opposing processes: synapse formation and synapse elimination. Too little pruning leaves a network cluttered with connections that carry no useful signal; too much strips out connections that were doing something.

Microglia participate in both directions. They recognise and phagocytose redundant synapses through specific phagocytic receptors, and they regulate synaptic plasticity by releasing effector molecules.

That second function is easy to overlook and it changes the picture. A cell that both removes connections and secretes factors influencing the strength of remaining ones is not performing maintenance. It is participating in the computation.

3.Why This Was a Genuine Surprise

The reason this counts as one of the more consequential findings in recent neuroscience is that it broke a boundary the field had assumed.

The brain was understood as immune-privileged — separated from the immune system by the blood–brain barrier, running on its own rules. Immune involvement in the brain meant pathology: infection, injury, disease.

What this literature established is that immune machinery is part of normal brain development and normal brain function. The molecules doing the tagging in Part Two are complement proteins, borrowed from the system that marks bacteria for destruction, and repurposed to mark a synapse.

Evolution did not invent a new mechanism for sculpting neural circuits. It used the one that was already there for identifying things to be eaten.

4.Where It Goes Wrong

The review's second half concerns disease, and the list is long: Alzheimer's disease, Parkinson's disease, ischaemic stroke, cerebral haemorrhage, traumatic brain injury, multiple sclerosis and epilepsy.

The framing the authors use is worth quoting for its balance: highly heterogeneous microglia exhibit diverse functions in these diseases and participate in progression by exacerbating or inhibiting synaptic dysfunction.

Both directions, in the same cell type, in the same disease. This is why the therapeutic picture is complicated and why nobody has a drug that simply turns microglia down.

Synaptic dysfunction and microglial activation are common features across these conditions. Whether the activation drives the dysfunction, responds to it, or does both at different stages is the question Parts Three and Four run into from two directions.

5.How to Read Claims in This Area

Three things follow from the review that are worth carrying into the rest of the series.

Pruning is not damage. A finding that microglia eliminated synapses is not, on its own, a finding that something bad happened. Establishing harm requires showing that the wrong synapses went, or too many.

Microglial activation is not one state. The heterogeneity described above means that measurements of activation markers report something coarse about a population that is doing several different things.

And nothing in this literature yet supports intervention in healthy people. The therapeutic targets the review identifies are for defined neurological disease, in preclinical models.

Editorial Comment

MindHeaven® makes no claim relating to microglia, synaptic pruning, neuroinflammation, or the immune system of the brain, and no ingredient in our formulations is proposed to act on any of them.

We publish this series because it is foundational to how the adult brain changes, which underlies everything else in this section — and because "neuroinflammation" has become a marketing word attached to products with no evidence of touching any of the mechanisms described here.

Part Two takes the most specific question in the field: given that a microglial cell must choose which synapses to remove, what physically marks one for removal?