Microglia as builders, not only pruners
Microglia are best known for removing excess synapses during development. This study asks the opposite question for the nucleus accumbens, a region the authors describe as important for emotional regulation and motivated behavior: do microglia also help excitatory synapses form, and does losing them early change adult behavior?
The main tool is a mouse line, Csf1rΔFIRE/ΔFIRE (called KO here), that lacks microglia entirely because a microglial survival enhancer for the CSF1 receptor is deleted. Controls were Csf1r+/+ wild-type and Csf1rΔFIRE/+ heterozygous littermates; the authors note heterozygotes have no detectable difference in microglia number compared with wild type, so they were sometimes pooled as microglia-positive controls. Because this model removes microglia for the whole lifespan and across the whole brain, the authors later add a second, region-specific depletion experiment to test whether the NAc and a specific developmental window matter.
Fewer excitatory events in the second and third weeks
The central measurement is whole-cell patch clamp of NAc medium spiny neurons, recording miniature excitatory postsynaptic currents (mEPSCs) at P8-10, P14-16, P20-22 and 2.5-3 months. mEPSC frequency is used as an index of how many functional excitatory synapses are releasing transmitter onto the recorded cell; amplitude reflects the size of individual events.
In KO mice, mEPSC frequency was reduced relative to controls, with the deficit most prominent at P14-16 and P20-22, while amplitude did not differ significantly by genotype. Inhibitory currents (mIPSCs) also showed no significant genotype difference in frequency at P20-22 or adulthood. The authors therefore describe the effect as blunted excitatory synaptogenesis rather than a general loss of all synapses, and they place it in the second and third postnatal weeks, when NAc microglial density normally rises to nearly double adult levels.
Tissue proteomics implicates synaptic machinery
To look for a molecular correlate, the authors performed LC-MS/MS proteomics on microdissected NAc tissue from wild-type and KO mice at P20-22, the age when the mEPSC frequency deficit was clearest. They detected more than 3,700 proteins, of which 229 (about 10%) differed significantly by genotype.
Expected microglial proteins such as P2ry12 and Hexb were downregulated in KO tissue, confirming the loss of microglia, even though microglia are only roughly 5% of NAc cells. The authors report changes in proteins involved in synapse structure, trans-synaptic signaling and presynaptic function. Notably, detected astrocyte-derived synaptogenic cues (Sparcl1, Gpc4, Gpc6) and several adhesion molecule families did not differ significantly, and astrocyte morphology and mEPSC/mIPSC decay kinetics were preserved. The authors interpret this as evidence that the reduced synapse number is not explained simply by missing astrocyte or adhesion support, while flagging that tissue-level proteomics cannot resolve which cell type supplies each protein.
Adult synapses largely recover, behavior does not
By 2.5-3 months, the mEPSC frequency difference between KO and control mice was no longer significant, and the authors describe the electrophysiological changes as largely resolving by adulthood. This matters for interpretation: the adult behavioral phenotype is not simply a continuation of an obvious adult synaptic deficit.
Adult KO mice nevertheless differed in threat avoidance assays. In the platform-mediated avoidance (PMA) task, mice received nine tone-shock pairings on training day and six tones without shock on retrieval day. KO mice showed increased avoidance during training, and the paper reports persistent changes in threat avoidance behavior and underlying NAc neuronal activity in adulthood. Fiber photometry recordings of NAc calcium signal during the task, plus Fos immunostaining after retrieval, indicated altered NAc activity in KO mice, with fewer Fos+ NAc neurons than controls. These are associations between behavior and measured activity, not proof that the activity change alone causes the behavior.
A second-week depletion reproduces the core deficit
The lifespan KO cannot say when microglia matter. To test timing and region, the authors injected clodronate disodium salt (CDS) or saline bilaterally into the NAc of wild-type mice at P10, when depletion is strongest about five days later, around P15. CDS-treated mice had large microglia-depleted zones at P15, with microglia fully tiling the NAc again by P21 and no depleted zones at P21 or 3 months.
At P20-22, CDS-treated mice showed reduced mEPSC frequency, matching the KO result, with no significant change in amplitude, rise time or decay. As adults, they showed anxiety-like behavior in the elevated zero maze and increased avoidance during PMA training. However, they did not differ in PMA retrieval, unlike the lifespan KO mice. The authors use this mismatch to argue that the retrieval/memory phenotype in KO mice may depend on the acute absence of NAc microglia in adulthood, absence during a different developmental window, or microglial loss in another brain region. The transient experiment therefore supports a causal role for second-week NAc microglia in synapse formation and some adult avoidance behavior, but not for every KO phenotype.
What the two depletion models jointly support
Read together, the two models separate what the paper can and cannot claim. The lifespan KO shows that lacking microglia changes NAc excitatory synapse development and adult avoidance behavior. The transient, NAc-localized CDS experiment shows that removing microglia from the NAc during the second postnatal week is sufficient to blunt excitatory synaptogenesis at P20-22 and to produce some adult avoidance changes, which strengthens a causal, region- and window-specific interpretation for those endpoints.
The comparison also sets the boundary. The CDS approach involves invasive stereotaxic injection and possible effects of clodronate on vasculature and other glial cells, as the authors note, and neither model gives precise temporal control over microglial functional states. The KO lacks regional and temporal specificity, while CDS depletion is transient and local. The paper's own framing is that candidate proteins may mediate microglial support of NAc synaptogenesis and merit dedicated follow-up functional studies, so the proteomic changes are leads rather than established mechanisms.
The developmental window is the key variable
The most decision-relevant distinction in this paper is not simply whether microglia are present or absent, but when and where they are absent. Lifelong absence and a second-postnatal-week NAc depletion both blunt excitatory synapse formation and alter adult avoidance, yet only the lifelong KO shows the retrieval-phase difference. For readers comparing microglial depletion studies, that split argues against treating all microglial loss models as interchangeable and against assuming that adult synaptic recovery implies adult behavioral recovery.
The paper's strongest supported claim is narrower than a general statement that microglia build synapses: in the mouse NAc, microglia support excitatory synapse formation during the second and third postnatal weeks, and disrupting that support is associated with lasting changes in NAc circuit activity and threat avoidance behavior. Whether the same window and mechanism apply to other brain regions, other cell types or human development is not tested here.
Sources used in this answer
Microglia regulate nucleus accumbens synaptic development and circuit function underlying threat avoidance behaviors in mice
Mice lacking microglia show blunted NAc excitatory synapse formation in the second and third postnatal weeks, with adult electrophysiological changes largely resolved but persistent changes in threat avoidance behavior and NAc activity; transient second-week NAc microglial depletion reproduces the synaptic deficit and some adult avoidance effects.