TL;DR: A 2026 mouse study in Translational Psychiatry found deleting connexin 36, an electrical-coupling protein between neurons, weakened several schizophrenia-relevant brain-wave readouts.
Key Findings
- Sleep architecture mostly held; the rhythm on top of it broke: Cx36 knockout mice did not lose normal sleep-wake organization. They lost the fast coordination riding inside it.
- Spindle density preserved, spindle quality degraded: The network still generated spindle-like events, but amplitude and duration were reduced — counting events would have missed the phenotype.
- Ketamine gamma response blunted: A direct pharmacology angle. Cx36 loss reduced the cortical gamma response to ketamine — a drug widely used to probe excitation-inhibition balance.
- 40 Hz auditory steady-state and mismatch negativity reduced: Two of the most durable EEG biomarkers in schizophrenia research, both moved by removing a single gap-junction protein.
- Social behavior and gamma activity decoupled together: Impaired social habituation paired with reduced investigation-induced gamma — tying the EEG abnormality to active behavior, not only passive recording.
- Electrical synapses earn a seat at the psychiatric circuit table: The disorder-relevant biomarker family changed when one connexin protein was deleted — before any chemical synapse was touched.
Source: Translational Psychiatry (2026) | McNally et al.
Psychiatric neuroscience pays a lot of attention to neurotransmitters, receptors, and chemical synapses — partly because most psychiatric drugs act there. Neural timing also depends on a quieter system: electrical synapses formed by gap-junction proteins like connexin 36.
They let neurons coordinate at millisecond resolution, and they are central to how inhibitory networks synchronize fast rhythms. This mouse study tested what happens when that timing machinery is removed from mature GABAergic circuits.
Sleep Stage Was Less Damaged Than Sleep Rhythm
The most informative result was a dissociation: Cx36 knockout mice did not lose normal sleep-stage structure.
They entered NREM and REM. They cycled. The abnormalities appeared in the rhythms riding on top of that architecture — the gamma, beta, sigma surges, and spindle features that act as network-timing findings.
That dissociation makes the study easier to interpret. Cx36 is less about whether the animal enters a sleep state and more about how coordinated the thalamocortical network is inside that state.
The knockout did not erase sleep. It degraded the fast coordination that lets thalamic and cortical circuits communicate with millisecond precision.
Connexin 36 Loss Reduced Spindle Strength Without Reducing Spindle Count
Sleep spindles are often treated as countable events — this many spindles per minute, this density per epoch. This study shows why counting alone misses the phenotype.
Spindle density was preserved. Spindle amplitude and duration were reduced.
The network could still generate spindle-like events. The events were just weaker or shorter.
For neuropsychiatric EEG biomarkers, that distinction changes interpretation: a normal event count can hide abnormal event quality. Studies that report spindles as a tally may be underestimating relevant abnormalities, especially when the biology lives in oscillation precision rather than oscillation presence.
Ketamine Gamma Response Linked Connexin 36 to Schizophrenia-Relevant Physiology
The ketamine result gives the mouse experiment its translational hook. Ketamine is widely used experimentally to perturb cortical excitation-inhibition balance, and its effects on gamma oscillations are heavily studied in psychiatric neuroscience.
Cx36 knockout mice showed blunted gamma responses to ketamine — meaning electrical synapses help shape how the cortex responds when NMDA-receptor signaling is disrupted.
The 40 Hz auditory steady-state response and mismatch negativity findings point in the same direction. Both are durable EEG biomarkers in schizophrenia research, and both moved when a single gap-junction protein was deleted.
The results sort cleanly into three categories:
- Spontaneous rhythm problem: gamma, beta, sigma surges, and spindle features all degraded.
- Evoked-response problem: ketamine gamma, 40 Hz auditory response, mismatch negativity, and ERP amplitudes all disrupted.
- Behavior-linked timing problem: social investigation triggered less gamma activity — tying the EEG abnormality to active behavior, not only passive recording.
Mismatch Negativity Linked Connexin 36 to Thalamocortical Synchrony
Mismatch negativity is one of the more durable EEG findings in schizophrenia research. It reflects the brain’s automatic response when an expected sound pattern is violated.
Reduced mismatch negativity has been linked to impaired predictive processing and to functional outcomes — how patients actually fare in life beyond symptom checklists.
The Cx36 knockout mice showed reduced mismatch negativity, attenuated ERP amplitudes, and altered evoked power. That does not turn a mouse knockout into schizophrenia, and researchers do not claim otherwise.
What it does show is that disrupting a single electrical-synapse protein can disturb a whole biomarker family human psychiatry already uses for circuit-level diagnosis. The same logic applies to the 40 Hz auditory steady-state response.
If Cx36 changes that rhythm in mice, electrical synapses belong in the search for circuit-level causes of abnormal sensory processing in humans.
Connexin 36 Knockout Mice Also Showed Behavioral Changes
The abnormalities were not only electrophysiological. Cx36 knockout mice also showed impaired social habituation and reduced investigation-induced gamma activity, linking the synchrony defect to a measurable behavior rather than only an EEG trace.
The standard caveat applies: mouse social investigation is not a direct equivalent of human social dysfunction. The right framing is more modest — the timing abnormality appears during behavior, not only in passive recordings, and the behavioral readout moves with the rhythm signal in a coherent direction.
Connexin 36 Mouse Data Still Need Human Schizophrenia Validation
Researchers propose that Cx36 may be a therapeutic target for dysfunctional network activity, but the evidence is preclinical. The study does not show that altering Cx36 treats schizophrenia, bipolar disorder, autism, or any specific psychiatric condition.
What it does show is that removing Cx36 disrupts thalamocortical synchrony in ways that resemble well-established neuropsychiatric biomarkers. That puts electrical synapses much closer to the center of psychiatric circuit research than they have typically lived.
Many psychiatric symptoms may not come from circuits being absent — they may come from circuits being present but mistimed.
Cx36 sits exactly at that timing level. The animals still slept, heard sounds, and investigated other mice. The timing signatures attached to those states just got weaker.
If future studies can modulate Cx36 function rather than delete it, researchers could test whether restoring electrical coupling improves the EEG phenotypes without broadly suppressing the network.
That is the path from biomarker mechanism to potential therapy: first showing that a single gap-junction protein shapes psychiatric-relevant EEG timing more than the field has usually recognized.
Citation: DOI: 10.1038/s41398-026-04018-1; McNally et al; Loss of connexin 36 elicits abnormalities in thalamocortical network activity relevant to neuropsychiatric disorders; Translational Psychiatry; 2026.
Study Design: Preclinical mouse study assessing sleep-wake regulation, spontaneous and evoked EEG activity, ketamine responses, and behavior after Cx36 deletion.
Sample: Cx36 knockout mice and littermate controls.
Key Result: Cx36 loss preserved sleep architecture but impaired gamma/beta activity, reduced spindle amplitude and duration, blunted ketamine gamma response, disrupted 40 Hz auditory steady-state response, and reduced mismatch negativity.
Caveat: Constitutive knockout model; mouse social and EEG findings are not direct equivalents of human disorders.
