Making the leap from petri dish to living organism

In neuroscience, it is tempting to think that understanding how individual cells work will tell us how the brain works. But the brain is more than the sum of its parts: neurons constantly communicate with each other and with the cells that support them. Sometimes, the real story only emerges when we look at the whole network.

In 2020, our group published a study showing how DNA methylation can regulate the function of parvalbumin-positive (PV) interneurons in the cerebral cortex (https://doi.org/10.1093/cercor/bhaa009), a specialised group of inhibitory neurons that is crucially important to balance cortical activity. We found that DNMT1, an enzyme carrying out DNA methylation, regulates genes associated with clathrin-mediated endocytosis – a process cells use to recycle material at their membranes, including components needed for communication between neurons. Removing DNMT1 altered this machinery and increased synaptic activity of cortical PV interneurons in cultured brain slices.

But cells in a dish are not a brain. In isolation, we could observe what happened to individual neurons, but not how the surrounding network would respond. Would the effects remain local, or would the brain find a way to compensate?

In our new study, published in Communications Biology (https://doi.org/10.1038/s42003-026-11079-x), we took the next step: Using mice in which the Dnmt1gene was deleted specifically in PV interneurons, we confirmed that these neurons fired more frequently in the cortex of awake mice in vivo, too. More importantly, we discovered that the consequences extended far beyond the cells in which the gene had been deleted.

More activity does not mean more inhibition

PV interneurons release the inhibitory neurotransmitter GABA and help limit the activity of other neurons. They also coordinate neuronal firing, supporting the precise timing needed for cortical information processing.

It might therefore seem reasonable to expect that more active PV interneurons would produce stronger inhibition. Our results showed a different picture.

Despite increased PV interneuron firing, surrounding cortical neurons were also more active. At the same time, activating PV interneurons was less effective at suppressing activity in the wider network.

One possible explanation is that the network had adapted to the persistent change in inhibitory signalling. Other neurons may have adjusted their responsiveness, reducing the effectiveness of PV interneuron-mediated inhibition. This remains an interpretation rather than an established mechanism.

The consequences for sensory responses were nevertheless clear. When we presented visual patterns or stimulated the whiskers, cortical responses were weaker, delayed and less precisely timed. Gamma oscillations were also strongly reduced.

Thus, increased activity was accompanied by poorer coordination—not by more effective cortical processing.

Looking for changes in behaviour

Given these findings, we expected to see impairments in sensory processing or learning. Yet the mice performed comparably to controls in the tasks we examined, including spatial learning and a more complex visual decision-making task.

This was an important result in its own right. Pronounced changes in cortical activity do not necessarily produce an obvious deficit in every behaviour that depends on the cortex. Some functions may be maintained through compensation, although our experiments do not establish how this was achieved.

There was, however, another difference: the knockout mice were heavier than controls, despite being housed under the same conditions and showing no obvious increase in food intake.

When we examined their behaviour more closely, we found that they moved less. They showed no clear deficit in basic motor coordination, but spontaneous activity and voluntary wheel running were reduced. Nest building and grooming were also decreased, and additional tests revealed anxiety-related behaviour.

Taken together, these observations were consistent with an apathetic, depression-like behavioural profile. Mouse behaviour cannot be equated directly with human depression, but the combination of findings prompted us to look more closely at changes beyond neuronal firing.

Looking beyond the targeted neurons

Using single-nucleus RNA sequencing, we examined gene expression across cortical cell populations. We also analysed chromatin accessibility to investigate changes in gene-regulatory programmes.

Although Dnmt1 had been deleted specifically in PV interneurons, the molecular changes were not restricted to these cells. We found alterations in excitatory neurons, astrocytes and oligodendroglial cells, suggesting a broader reorganisation of the cortical environment.

The oligodendrocyte lineage was particularly interesting. These cells are best known for producing myelin, but their functions extend beyond insulating nerve fibres. Certain populations are associated with neuronal cell bodies and contribute to the surrounding extracellular environment.

This environment includes perineuronal nets, specialised extracellular structures surrounding many PV interneurons. These nets help stabilise synaptic connections and support the precise firing properties of the neurons they surround.

Our analyses suggested that communication between PV interneurons and oligodendroglial cells had changed. In particular, knockout mice had fewer newly formed oligodendrocytes expressing Pdgfc, a population associated with the perineuronal environment. Mathematical modelling also suggested a reduced transition towards this population.

Importantly, we found a corresponding change in the tissue: perineuronal-net staining was reduced. Overall oligodendroglial cell numbers and cortical myelination, however, remained largely unchanged.

The findings therefore pointed to a specific alteration in the cellular environment around PV interneurons, rather than a general loss of oligodendroglial cells.

What holds the network together?

Our results suggest that the effects of losing DNMT1 in PV interneurons extend beyond changes within these neurons. Altered inhibitory function was accompanied by changes in other cell populations, less pronounced perineuronal nets, disrupted cortical synchrony and altered behaviour.

We cannot yet place all these observations into a direct causal sequence. Further experiments will need to establish whether changes in oligodendroglial cells contribute to the reduction in perineuronal nets, and whether restoring either can improve network function or behaviour.

For us, the study highlights why it matters to follow a molecular finding beyond the cell in which it first appears. Understanding PV interneuron function requires understanding the cells and extracellular structures that support it.

What began as a question about DNA methylation in inhibitory neurons ultimately became a question about how this wider cellular network is maintained—and what happens when its relationships change.