NEWS AND VIEWS

Brain’s immune cells put the brakes on neurons

Microglia are the brain’s immune cells. A previously unknown role for microglia has now been uncovered: providing negative feedback to active neurons, to help the brain process information.
Thomas Pfeiffer is in the Department of Neuroscience, Physiology and Pharmacology, University College London, London WC1E 6BT, UK.

Search for this author in:

David Attwell is in the Department of Neuroscience, Physiology and Pharmacology, University College London, London WC1E 6BT, UK.
Contact

Search for this author in:

Neural circuits in the brain rely on neuronal excitation (a positive change in the electrical potential across the cell membrane), combined with delayed inhibition (Fig. 1). Inhibition is crucial for keeping neuronal activity in the optimal range for encoding information, minimizing the brain’s energy use and computing useful neuronal outputs. It has conventionally been thought that inhibition is mediated by a neuronal subtype called interneurons that release neurotransmitter molecules (such as the amino acid GABA) to make the membrane potential of the downstream neuron more negative — although neurotransmitter release from non-neuronal cells called astrocytes can also contribute1. Writing in Nature, Badimon et al.2 extend this repertoire of inhibitory influences to include microglia, the resident immune cells of the brain. The authors’ work raises fascinating questions about the role of microglia in information processing.

Figure 1

Figure 1 | Inhibition of active neurons by microglial cells. a, A generic neuronal circuit, centred on a principal neuron (PN). The PN and an excitatory input to the circuit both release the excitatory neurotransmitter molecule glutamate (Glu). Interneurons (IN) release the inhibitory neurotransmitter GABA. Neurotransmitters derived from cells called astrocytes fine-tune the neuronal circuits (these signals are not shown). The circuit is also inhibited by the molecule adenosine (ADO), which Badimon et al.2 show is generated, in part, by microglial cells. b, When the input to the circuit is increased, GABA-mediated inhibition decreases the output on a rapid timescale. Microglia-derived ADO adds a slower component to the inhibition.

Badimon and colleagues took advantage of the fact that blocking activation of the growth-factor receptor protein CSF1R in mice leads to a lack of microglia3. The authors found that if they gave neurostimulants to animals that lacked microglia, the drugs produced long-lasting epileptic seizures, indicative of hyperactive neuronal excitation. Seizures were not observed in wild-type animals receiving the same drugs, indicating that microglia normally exert a brake on neuronal activity. This result echoes and extends two previous studies4,5. Microglial processes are attracted to the cell bodies (housing the nucleus) of active neurons by the release of ATP molecules. There, the processes decrease neuronal activity, both normally4 and in pathological5 conditions.

Whereas these previous studies focused on cell bodies, Badimon and colleagues focused on the synaptic junctions between neurons, which also release ATP to attract microglial processes. The microglial enzyme CD39 converts ATP into ADP (and then into AMP); ADP activates P2Y12 receptor proteins found only on microglia (go.nature.com/3iuewxa and go.nature.com/33hwjft; Fig. 2). Blocking P2Y12 receptors has been shown to inhibit the attraction of microglia to cell bodies and synapses5, and Badimon et al. found that such a block also reduces neuronal inhibition by microglia in response to neurostimulants.

Figure 2

How might microglia–neuron interactions inhibit the electrical activity of neurons? The authors found that deleting microglia decreased extracellular levels of the molecule adenosine (ADO). Pharmacologically blocking CD39 or the downstream enzyme CD73 (which converts AMP into ADO; Fig. 2) also lowered ADO levels. Furthermore, blocking the activity of CD39 increased the susceptibility of mice to seizures in response to neurostimulants. Together, these observations implicate ADO as the microglia-derived factor that dampens neuronal activity.

It is well known that ADO lowers neuronal excitability6. Indeed, the reason that coffee makes us more alert is that caffeine blocks ADO’s inhibitory effects. ADO lowers excitability by acting on what are called A1 receptors, which (by lowering the concentration of the intracellular messenger molecule cyclic AMP) decrease the release of the excitatory neurotransmitter glutamate, and reduce its effects on the downstream neuron that receives the neurotransmitter. A1 receptors also activate potassium ion channels in neuronal membranes to keep their membrane potential negative (and so keep the neurons unexcited). Thus, Badimon et al. have uncovered a previously unknown feedback loop for neuronal regulation mediated by microglia, which, when attracted to active synapses, generate ADO to inhibit excessive neuronal activity (Fig. 2).

The authors showed that this negative feedback operates in a region-specific manner. Deletion of microglia in the brain’s grey matter (where neurons have their cell bodies and synapses) caused the hyperactive neuronal response to mild excitation. By contrast, deleting microglia in white matter (where long-range neuronal connections run) did not cause hyperactivity. In addition, deletion of microglia in specific regions of grey matter affected only those regions, rather than causing excessive activity across the whole brain.

Just how spatially and temporally specific might the feedback mechanism be? Two factors should slow its activity. First, there will be a lag between release of ATP by a synapse and the production of local ADO after microglial processes are drawn to that synapse. Second, it is unclear whether the enzymes CD39 and CD73 are close enough spatially for rapid ADO production. Although microglia express CD39 highly, they only weakly express CD73, which is expressed more in other brain cells, such as neurons and cells of the ‘oligodendrocyte’ lineage (go.nature.com/3iuewxa and go.nature.com/33hwjft). Another enzyme, non-tissue-specific alkaline phosphatase, can also convert AMP into ADO7, but this is largely expressed in astrocytes (go.nature.com/3iuewxa and go.nature.com/33hwjft). Thus, after microglial CD39 has converted ATP into ADP and AMP, the AMP molecule might have to diffuse some distance, to a different cell type, to be converted into ADO. This would lengthen the duration of the feedback loop compared with that of conventional GABA-mediated synaptic inhibition, which operates within about 50 milliseconds of neuron stimulation (Fig. 1b). The ADO feedback loop might have longer-lasting effects, and also be less spatially specific; whereas synaptic inhibition involves direct contacts with target neurons, diffusion of ADO precursors implies that the microglial mechanism would act on multiple neurons in an area.

Consistent inhibition of neuronal synapses can cause a decrease in the strength of the connection between neurons. Synapses that are weakened in this way are sometimes removed by microglia or astrocytes in a process called pruning8. It will be interesting to determine whether ADO-mediated weakening of synapses triggers this pruning mechanism.

Another question is to what extent the inhibitory influence of microglia depends on the amount of neuronal excitation occurring. Badimon et al. used neurostimulants that affect many neurons. It remains to be seen whether ADO-mediated inhibition also operates (to a lesser extent) when there is a small amount of excitation. In other words, is this system an emergency brake for extreme situations, or does it act proportionally for all levels of excitation? Inhibitory interneurons have increased influence as neuronal excitation increases — this enables neural circuits to respond differentially to a wider range of input strengths9. Microglia-facilitated ADO production might similarly enhance the coding range of neural circuits.

ADO derived from ATP released by astrocytes is proposed to regulate sleep onset10. Badimon et al. found that the extracellular level of ADO in a brain region called the striatum was reduced by 85% in anaesthetized mice lacking microglia, compared with control mice that had microglia. This suggests that the build-up of extracellular ADO that generates sleep pressure might largely be derived from the activity of microglial CD39. Thus, microglia-facilitated negative-feedback control of neuronal activity could be a side effect of the evolution of a system to induce sleep (or vice versa).

There are also hints that this feedback system might contribute to neurological or psychiatric disease. As Badimon and colleagues show, epileptic seizures can result if microglia-mediated negative feedback is absent. In less extreme situations, both P2Y12 receptors and CD39 are downregulated in a range of diseases in which the immune-defence role of microglia is activated, including Alzheimer’s disease and Huntington’s disease, and after injection of the bacterial-coat protein lipopolysaccharide to mimic bacterial infection (as summarized in Extended Data Fig. 10 of the paper). All of these conditions can also involve increases in neuronal activity. Conversely, upregulation of CD39 can lead to depression-like behaviour11.

Going forward, it will be crucial to define the mechanisms of ATP release from neurons, and the spatial and temporal scales on which ADO acts. It also remains to be seen whether there is any role for ADO’s lower-affinity A2 receptors in microglia-mediated neuronal inhibition. Finally, do circadian-rhythm and disease-related factors modulate these mechanisms? How these immune cells regulate information processing is just beginning to be unravelled.

Nature 586, 366-367 (2020)

References

  1. 1.

    Araque, A. et al. Neuron 81, 728–739 (2014).

  2. 2.

    Badimon, A. et al. Nature 586, 417–423 (2020).

  3. 3.

    Ginhoux, F. et al. Science 330, 841–845 (2010).

  4. 4.

    Li, Y., Du, X.-F., Liu, C.-S., Wen, Z.-L. & Du, J.-L. Dev. Cell 23, 1189–1202 (2012).

  5. 5.

    Cserép, C. et al. Science 367, 528–537 (2020).

  6. 6.

    Haas, H. L. & Selbach, O. Naunyn-Schmiedeberg’s Arch. Pharmacol. 362, 375–381 (2000).

  7. 7.

    Zhang, D. et al. PLoS ONE 7, e39772 (2012).

  8. 8.

    Chung, W.-S., Welsh, C. A, Barres, B. A. & Stevens, B. Nature Neurosci. 18, 1539–1545 (2015).

  9. 9.

    Pouille, F., Marin-Burgin, A., Adesnik, H., Atallah, B. V. & Scanziani, M. Nature Neurosci. 12, 1577–1585 (2009).

  10. 10.

    Halassa, M. M. et al. Neuron 61, 213–219 (2009).

  11. 11.

    Cui, Q.-Q. et al. EMBO Rep. 21, e47857 (2020).

  12. 12.

    Di Angelatonio, S. et al. Front. Cell. Neurosci. 9, 409 (2015).

Download references

Nature Briefing

An essential round-up of science news, opinion and analysis, delivered to your inbox every weekday.