Serotonin induces DOWN states across the anesthetized mouse forebrain

Serotonin, often simplified as a mere "mood stabilizer," is actually a powerful neuromodulator that orchestrates massive shifts in brain-wide activity. In the context of sleep and anesthesia, the brain undergoes rhythmic cycles of intense activity (UP states) and profound silence (DOWN states). This process is known as slow-wave oscillations. For years, the scientific community has been divided on how serotonin influences these cycles. Some imaging studies suggested it suppresses brain activity globally. Older electrical stimulation studies claimed it could actually trigger active UP states.

This discrepancy left a fundamental question unanswered: does serotonin act as a precise local tuner, or does it serve as a master switch? A new study published by the International Brain Laboratory addresses this. It bridges the gap between broad, low-resolution imaging and highly localized electrical recordings. By combining optogenetic control with large-scale, single-cell recordings, the researchers demonstrate that serotonin acts as a global driver of neural silence. It even reaches regions that lack direct connections to the serotonin-producing centers.

The disconnect between scale and precision

The difficulty in mapping serotonin's influence stems from a fundamental trade-off in neuroscientific measurement. Functional MRI (fMRI) provides a wide-angle lens. It observes blood-oxygen-level-dependent (BOLD) signals—changes in blood flow that act as a proxy for neural activity—across the entire brain. However, fMRI lacks the temporal resolution to see millisecond-scale transitions between UP and DOWN states. Conversely, traditional electrophysiology offers exquisite temporal precision. It measures the electrical spikes of individual neurons but is typically restricted to a tiny, localized patch of tissue.

Because of this, the literature became fractured. Researchers using fMRI saw widespread suppression of activity. Meanwhile, those using electrical stimulation of the dorsal raphe nucleus (DRN)—the brain's primary serotonergic hub—occasionally observed the induction of UP states. This mismatch suggested that the effect of serotonin might depend heavily on how it was delivered. The authors of this paper sought to resolve this by using Neuropixel probes. These tools allow for the simultaneous recording of thousands of neurons across multiple distant brain regions. This provides both the breadth of fMRI and the precision of electrophysiology.

Mapping the serotonergic switch

To move beyond the limitations of previous studies, the researchers employed a sophisticated multi-stage experimental design. First, they utilized optogenetics—a technique using light to control genetically modified neurons—to selectively activate only the serotonergic neurons in the DRN. This avoided the "messy" activation of neighboring excitatory neurons that occurs with traditional electrical stimulation. Such non-selective activation likely caused the contradictory UP-state results seen in prior literature.

The mechanism of their investigation followed three distinct layers: 1. Large-scale Recording: They implanted Neuropixel probes across seven distinct brain regions. These included the cortex, striatum, thalamus, and amygdala. They captured real-time spiking activity in mice under light anesthesia. 2. State Quantification: To move from raw spikes to meaningful brain states, they applied a two-state Hidden Markov Model (HMM). This statistical tool identifies the most likely underlying state (either UP or DOWN) based on observed firing rates. It essentially filters the "noise" of individual neurons to find the "signal" of the network state. 3. Computational Validation: Finally, they built a multi-area computational model. This model used biologically realistic connectivity data from the Allen Brain Atlas. It simulated how an inhibitory signal in one area might propagate through the rest of the network.

As shown in, the optogenetic stimulation of the DRN led to a rapid and consistent decrease in spiking activity.

This suppression occurred across nearly all recorded regions.

Evidence of a coordinated descent

The core finding is that serotonin stimulation does not just quiet individual neurons. It forces the network into a DOWN state. The authors report that in regions with natural bi-stability—such as the cortex, striatum, and thalamus—serotonin stimulation rapidly induced a transition from an active UP state to a silent DOWN state [Figure 2d]. They further demonstrated that serotonin actually lengthened existing DOWN states and forced abrupt transitions from UP states [Figure 2e, f].

Perhaps most striking was the behavior of the visual cortex. Unlike the frontal cortex, the visual cortex does not receive significant direct projections from the DRN. Yet, the authors found that the visual cortex still transitioned into a DOWN state following stimulation. Through their computational model, they revealed the mechanism. This is an emergent property of the network. When the primary targets of serotonin are suppressed, the loss of excitatory drive causes the visual cortex to "fall into line" with the rest of the brain through inter-area synchronization [Figure 3h].

Regarding the molecular drivers, the researchers found that the magnitude of suppression did not correlate with the physical density of serotonin projections to a specific area. Instead, they found significant correlations with the expression of specific receptor subtypes, namely 5-HT1f and 5-HT2a [Figure S3]. This suggests that the "volume" of the serotonin signal is controlled more by the sensitivity of the receiving neurons than by the sheer number of incoming wires.

Limits of the global silencing model

While the study provides a compelling case for serotonin as a global state switch, several caveats remain. First, the researchers note that the induced DOWN states in the amygdala and hippocampus might be qualitatively different from the "natural" DOWN states seen in the cortex. Because these regions do not typically exhibit UP-DOWN oscillations, the HMM may simply be detecting a generic suppression of activity. Future research must determine if there is a mechanistic difference between these types of states.

Second, the computational model utilized an "open-loop" architecture. This means it modeled serotonin flowing from the DRN to the brain, but did not account for feedback from the brain back to the DRN. This limitation implies that future studies should prioritize closed-loop models to understand full regulatory dynamics. Finally, the use of isoflurane anesthesia is a proxy for natural sleep. While the UP-DOWN dynamics are similar, the chemical environment of anesthesia is not identical to the natural neurochemistry of non-REM sleep.

The verdict on serotonergic control

The evidence presented here strongly supports the view of serotonin as a high-level modulator of global brain states. By resolving the conflict between fMRI and electrophysiology, the authors have demonstrated that serotonin's ability to induce DOWN states is a widespread, network-level phenomenon. The discovery that regions like the visual cortex can be "switched off" via indirect, network-mediated synchronization is a significant leap in our understanding of how neuromodulators coordinate distributed systems.

Whether this mechanism holds true in the fully awake brain remains to be seen. However, the paper establishes a clear hierarchy. Serotonin does not just talk to neurons; it talks to the network.

Figures from the paper