Temporal derivative computation in the dorsal raphe network revealed by an experimentally driven augmented integrate-and-fire modeling framework
Curation statements for this article:-
Curated by eLife
eLife assessment
Harkin and colleagues explore functional properties of dorsal raphe serotonin neurons using the approach called a generalized integrate-and-fire [aGIF] model, which incorporates a relatively small number of salient biophysical properties of a specific neuron type, and whose parameters are optimized based on voltage dynamics obtained experimentally. The authors make an interesting finding that after-hyperpolarization and A-type potassium currents, in combination with heterogeneous feedforward inhibition from local GABA neurons, give rise to a derivative-like input-output relationship in serotonin neurons.
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
- Neuroscience (eLife)
Abstract
By means of an expansive innervation, the serotonin (5-HT) neurons of the dorsal raphe nucleus (DRN) are positioned to enact coordinated modulation of circuits distributed across the entire brain in order to adaptively regulate behavior. Yet the network computations that emerge from the excitability and connectivity features of the DRN are still poorly understood. To gain insight into these computations, we began by carrying out a detailed electrophysiological characterization of genetically identified mouse 5-HT and somatostatin (SOM) neurons. We next developed a single-neuron modeling framework that combines the realism of Hodgkin-Huxley models with the simplicity and predictive power of generalized integrate-and-fire models. We found that feedforward inhibition of 5-HT neurons by heterogeneous SOM neurons implemented divisive inhibition, while endocannabinoid-mediated modulation of excitatory drive to the DRN increased the gain of 5-HT output. Our most striking finding was that the output of the DRN encodes a mixture of the intensity and temporal derivative of its input, and that the temporal derivative component dominates this mixture precisely when the input is increasing rapidly. This network computation primarily emerged from prominent adaptation mechanisms found in 5-HT neurons, including a previously undescribed dynamic threshold. By applying a bottom-up neural network modeling approach, our results suggest that the DRN is particularly apt to encode input changes over short timescales, reflecting one of the salient emerging computations that dominate its output to regulate behavior.
Article activity feed
-
-
-
eLife assessment
Harkin and colleagues explore functional properties of dorsal raphe serotonin neurons using the approach called a generalized integrate-and-fire [aGIF] model, which incorporates a relatively small number of salient biophysical properties of a specific neuron type, and whose parameters are optimized based on voltage dynamics obtained experimentally. The authors make an interesting finding that after-hyperpolarization and A-type potassium currents, in combination with heterogeneous feedforward inhibition from local GABA neurons, give rise to a derivative-like input-output relationship in serotonin neurons.
-
Reviewer #1 (Public Review):
Harkin and colleagues present a very interesting study in utilizing cell-specific excitability properties of identified 5-HT raphe neurons and SOM-interneurons to specify computational integrate and fire neuronal models. In turn, they explore the resulting properties of these biophysically augmented GIF (aGIF) populations. While their electrophysiological characterization of firing properties is mainly confirmatory and their characterization of relevant conductances limited to fast-inactivating A-type currents, the identified features of the aGIF models are highly relevant for a better understanding of neuromodulatory systems. In particular the link between the strong spike firing adaptation of 5-HT neurons and the associated ability to detect transient changes in synaptic input are important. However, the …
Reviewer #1 (Public Review):
Harkin and colleagues present a very interesting study in utilizing cell-specific excitability properties of identified 5-HT raphe neurons and SOM-interneurons to specify computational integrate and fire neuronal models. In turn, they explore the resulting properties of these biophysically augmented GIF (aGIF) populations. While their electrophysiological characterization of firing properties is mainly confirmatory and their characterization of relevant conductances limited to fast-inactivating A-type currents, the identified features of the aGIF models are highly relevant for a better understanding of neuromodulatory systems. In particular the link between the strong spike firing adaptation of 5-HT neurons and the associated ability to detect transient changes in synaptic input are important. However, the biophysical mechanisms of adaptation (e.g. SK channels) and its variability across 5-HT neurons has not been experimentally explored.
-
Reviewer #2 (Public Review):
This study used electrophysiological data acquired from neurons in the dorsal raphe to model 5-HT output in response to extrinsic excitatory inputs based on the intrinsic properties of 5-HT neurons and local network connectivity with GABAergic neurons. Specifically, general and modified integrate-and-fire single cell models, together with local network models among 5-HT neurons and local GABAergic neurons providing feedforward inhibition (FFI), are used to simulate the firing output of 5-HT neurons in response to transient and prolonged depolarizations. The conclusions are as follows. 1) 5-HT neurons display prominent spike frequency adaptation, resulting from afterhyperpolarization potentials and change in firing threshold, and inactivating K current characteristic of A-type K current (I-A). These two …
Reviewer #2 (Public Review):
This study used electrophysiological data acquired from neurons in the dorsal raphe to model 5-HT output in response to extrinsic excitatory inputs based on the intrinsic properties of 5-HT neurons and local network connectivity with GABAergic neurons. Specifically, general and modified integrate-and-fire single cell models, together with local network models among 5-HT neurons and local GABAergic neurons providing feedforward inhibition (FFI), are used to simulate the firing output of 5-HT neurons in response to transient and prolonged depolarizations. The conclusions are as follows. 1) 5-HT neurons display prominent spike frequency adaptation, resulting from afterhyperpolarization potentials and change in firing threshold, and inactivating K current characteristic of A-type K current (I-A). These two features cause the rapid decline in firing responses at the onset of depolarizing input. 2) Heterogeneous FFI due to heterogeneous electrophysiological properties of local GABA neurons lead to divisive inhibition of 5HT neuron firing (i.e., change in the slope of input-output function) in the network model. 3) Using a ramp depolarization, the authors found that 5-HT neurons encode the temporal derivative of depolarization, i.e., the slope of ramp depolarization. This property can be ascribed to the prominent spike-frequency adaptation observed in 5-HT neurons. Overall, this study provides new insights into the control of 5-HT output by single cell and network mechanisms.
The conclusions are well supported by combination of rigorous brain slice electrophysiological recordings of the two types of neurons in the dorsal raphe, i.e., 5-HT neurons and somatostatin-positive GABA neurons, which are identified by the usage of transgenic mice where these neurons are fluorescently labeled, and the application of single cell and network models.
As the authors state, the most striking finding of this study is that 5-HT neurons encode temporal derivative of excitatory inputs, as it may relate to reinforcement learning models. Here, this feature is captured using a ramp depolarization and is solely modeled with intrinsic property of 5-HT neurons, i.e., spike-frequency adaptation. Instead of using a ramp depolarization, using repetitive brief depolarizations with changing intervals/frequency will be more informative. Further, incorporating the network model with FFI, in particular the delay in inhibition following excitation associated with FFI when same inputs (single and repetitive) feed into 5-HT neurons and GABA neurons, may be more relevant to the reinforcement learning algorithms (e.g., see Fig. 6a in J. Neurosci. 2008, 28: 9619-9631).
-
Reviewer #3 (Public Review):
The authors do an excellent job of producing relatively simple model neurons to characterize the key properties of two different cell classes (feedforward Somatostanin, SOM, interneurons and serotonergic, 5-HT, output neurons) in the dorsal raphe nucleus. They make a strong case for the role of an A-type potassium current and strong spike-threshold adaptation in the 5-HT neurons in reproducing the measured single-cell responses to stimuli and characterize well the interplay of these two currents in producing a transient measure of the derivative of inputs under some circumstances. While the demonstration of rapid adaptation leading to an output that resembles a derivative of the input under some circumstances is far from novel, the authors here have succeeded in matching such a computation to known …
Reviewer #3 (Public Review):
The authors do an excellent job of producing relatively simple model neurons to characterize the key properties of two different cell classes (feedforward Somatostanin, SOM, interneurons and serotonergic, 5-HT, output neurons) in the dorsal raphe nucleus. They make a strong case for the role of an A-type potassium current and strong spike-threshold adaptation in the 5-HT neurons in reproducing the measured single-cell responses to stimuli and characterize well the interplay of these two currents in producing a transient measure of the derivative of inputs under some circumstances. While the demonstration of rapid adaptation leading to an output that resembles a derivative of the input under some circumstances is far from novel, the authors here have succeeded in matching such a computation to known properties of neurons in known circuitry. However, a little more care is needed in how the authors describe the output of serotonergic neurons being linear in the derivative of the inputs, because in general that is not the case. For example, the sustained response of the neurons depends on the net input current when the derivative of the input is zero in all cases. A neuron that really follows a derivative would, for example, respond in a manner independent of baseline current and produce a cosine output to a sinusoidal input. Rather, it is a transient output response that in some limited ranges of constant baseline input and ramping rate of the input, has a peak that is linearly dependent on the ramping rate. Also, only positive slopes were considered. Thus, it seems unlikely that the serotonergic output in the model is very close to the derivative of a general input signal, nor does it appear likely to operate on the long temporal timescale needed for reinforcement learning, as suggested.
-