Main Menu (Mobile)- Block

Main Menu - Block

Hantman Lab

janelia7_blocks-janelia7_fake_breadcrumb | block
Hantman Lab
node:field_slogan | entity_field
The Hantman lab investigates how top-down feedback affects sensory processing. In particular, we are interested how cortical output influences proprioceptive streams in the brain. 
node_title | node_title
Hantman Lab
node:field_slogan | entity_field
The Hantman lab investigates how top-down feedback affects sensory processing. In particular, we are interested how cortical output influences proprioceptive streams in the brain. 
node_body | node_body

Voluntary movement results from a goal-based decision and a process that converts the goal into relevant motor commands. Planning and executing appropriate motor commands depends on learning the relationships between motor commands and physical consequences. Once command-consequence relationships are learned they can be used to predict the sensory consequences of the movement (forward model). Such predictions can be utilized before execution for planning purposes, and during execution can be compared with sensory information to determine differences between intended and actual movement. The resulting error signals can then serve to inform corrective actions and to train the forward model. By enabling planning and performance-tuning, predictions allow the brain to achieve skilled movements. While strong evidence exists that motor control (and perhaps aspect of cognition) depends on predictions, the network and cellular mechanisms of prediction generation and utilization are largely unknown. To reveal the logic of how predictions are made and used to control movement, the Hantman lab is developing the capability to identify, monitor, and selectively manipulate elements of motor control circuitry.

A central framework of our lab is that predictions of the consequences of motor commands are used in conjunction with sensory information to achieve motor control (Figure 1). Predictions are necessary for the following reasons. First, when preparing a sequential action, the brain needs to estimate the outcomes of earlier steps to plan later components of the behavior. Second, because peripheral pathways impose temporal delays that outdate sensory information, the current position of a moving animal can only be estimated internally. Lastly execution errors can be calculated by comparing predictions against the actual (albeit delayed) sensory information related to the movement. Predictions of current and future body positions are constructed by converting copies of motor messages (corollary discharges) into sensory consequences using learned forward models. Learning necessarily occurs in brain areas where motor commands and sensory information converge. We are dissecting the neural circuitry that generates, carries, and utilizes sensory predictions relevant for motor control. 

Psychometric studies have made predictions an accepted component of the current understanding of motor control. However, specifically testing the contributions of predictions to musculoskeletal motor control has not been examined. The mechanisms by which motor-related information are transformed into sensory coordinates to make predictions is also unknown. Even the error computations that compare predictions and actual sensory information are anatomically and physiologically uncharacterized.

Forward models must be built in areas of the brain where motor-related information is fed back onto sensory pathways. While any movement-related sensory system could be examined, we concentrated on proprioception because of its obligatory relationship with motor output to the musculoskeletal system. The anatomy described below led us to investigate the cerebellum as a focus of the convergence of proprioceptive and motor-related information. The cerebellum receives nearly every proprioceptive modality through mossy fiber boutons (large specialized axon terminations) originating from precerebellar neurons. The external cuneate nucleus (ECN) receives forelimb proprioceptive inputs directly from the periphery and delivers this relatively unprocessed sensory information to the cerebellum. The cerebellum also receives information related to motor commands. The basal pontine nucleus (BPN) is the main conduit of motor-related messages into the cerebellum, constituting over half of the total mossy fiber input to the cerebellum. The BPN receives inputs from motor control centers including cerebral cortex, superior colliculus, and red nucleus. What was unknown was whether BPN motor-related information and ECN proprioceptive information converged within cerebellar circuitry. Our lab showed that proprioceptive-related regions of the cerebellum receive information from the BPN and that individual granule cells (receive on average only four excitatory inputs) receive convergent inputs from these two modalities of information. Therefore, the cerebellum represents a major convergence site of relatively unprocessed proprioceptive and motor-related information.

Convergence of BPN and sensory inputs in cerebellum suggests a model of current position prediction as shown in Figure 1b. In our working hypothesis, a movement generates sensory and motor-related inputs that activates a subpopulation of granule cells. Downstream circuits learn this granule cell activity pattern and use it to represent that movement. If the BPN is sufficient to recreate the same downstream effect, these motor-related inputs will generate a representation that would predict the sensory consequence of that action. Since these BPN-dependent representations are not dependent on sensory information and thus do not incur peripheral delays, they could provide an accurate estimate of the current or future positions, useful for planning actions.

BPN inputs to proprioceptive regions of cerebellum may also be involved in error detection. In our model, BPN inputs generate a representation of the sensory consequences of intended movements in the cerebellum. Since errors are differences between intended and actual movements, these BPN-dependent representations could be used to cancel consequent proprioceptive information. Non-canceled inputs would be preserved and interpreted as errors or unrelated to the motor output.

The Hantman lab is testing if the BPN provides the motor-related information necessary to generate predictions and detect errors. To test this hypothesis, we have developed a multifaceted experimental system that takes advantage of the genetic toolbox available in the mouse. With this platform–centered on a multi-step reaching task–we seek to answer three questions.

1) Is the BPN necessary for motor control and planning? We, for the first time, selectively perturb the BPN during a motor control task.

2) How do cerebellar circuits integrate BPN and sensory information to predict sensory consequences and estimate errors? We are following BPN information though cerebellar circuitry.

3) What is the nature of information carried by the BPN? We are characterizing BPN inputs and their transformation through the BPN during behavior.