Research in the Jessell lab explores the developmental wiring and mature function of neural circuits that provide mammals with the ability to act on demand, through the neural control of movement. Studies focus on the neural circuits that control two forms of motor behavior that rely on limb musculature: locomotion and goal-directed reaching. In one approach, we aim to define the cellular rules and molecular mechanisms that direct the intricate wiring of these circuits. In parallel, we have used insights into the molecular origins of neuronal identity to devise more precise genetic methods to monitor and manipulate the activity of defined neuronal classes, permitting us to delve into the design of circuits and systems responsible for the planning and execution of movement.
Wiring Sensory-Motor Circuits
The precision that characterizes patterns of connectivity in the mammalian central nervous system has its origins in the diversification of developing neurons into distinct functional classes. The strategies and mechanisms used to translate neuron-subtype identity into selective connectivity remain unclear, however. We address this general issue by analyzing the feedback connections formed between proprioceptive sensory neurons, interneurons, and their target motor neurons. Some 50 or so muscle groups endow the mammalian limb with its versatile biomechanics, and each muscle is innervated by a dedicated pool of motor neurons. Generating the profile of connections that direct limb movement requires that the sensory neurons that convey feedback from individual limb muscles form strong connections with self–motor neurons and weaker connections with motor neurons that innervate synergist muscles and avoid nonself–motor neurons that innervate muscles with irrelevant or opponent functions. The exquisite specificity of sensory-motor connections appears to be hardwired and is thought to emerge from the acquisition of motor and sensory neuron–subtype identities.
But how are these connections established? Recent findings in our lab indicate that intricate patterns of sensory-motor connectivity derive from neuron-subtype identities in two sequential stages. One relies on the settling position of neurons and the other on the surface labels they express. In the spinal cord, the positioning of motor neuron cell bodies has long been known to exhibit a remarkable spatial register with the limb muscles they innervate. The motor neurons that innervate an individual limb muscle are clustered into spatially coherent columels and pools, which occupy stereotypic locations along the dorsoventral and mediolateral axes of the spinal cord. Our studies have shown that incoming sensory axons are directed to appropriate dorsoventral tiers in the ventral spinal cord independently of motor neuron surface labels. This tier-targeting strategy greatly limits potential motor pool targets and thus simplifies the eventual task of motor neuron recognition by sensory axons. Appreciation of the key role of neuron-settling position in sensory-motor connectivity begins to make sense of the finding that neuron-subtype identity is revealed as much by distinctions in settling position as by surface label.
Within the constraints imposed by this early neuron-positioning system programs of sensory-motor recognition appear to operate. As one example, expression of a repellent ligand, semaphorin 3E (Sema 3E), by selected motor pools is complemented by proprioceptive sensory expression of its cognate receptor, Plexin D1. Ectopic expression of Sema 3E in motor neurons markedly reduces the incidence of inputs from Plexin D1–marked sensory afferents, and conversely, genetic elimination of Sema 3E–Plexin D1 signaling permits illicit sensory-motor connections. Thus the potential for complementary matching of surface labels and resultant repellent recognition helps to determine the fine pattern of proprioceptive sensory inputs to motor pools. More generally, these findings establish that both neuronal position and surface label have roles in converting neuron-subtype identity into precise patterns of sensory-motor connectivity.
The diversification of proprioceptors into discrete pool subclasses is thought to direct sensory neurons to their motor neuron targets with unerring specificity. Our studies have shown that two proprioceptor transcription factors, Etv1 and Runx3, direct aspects of proprioceptive sensory character. Altering the level of Runx3 and Etv1 activity in proprioceptive sensory neurons changes the dorsoventral termination zone of their axon collaterals in the developing spinal cord, suggesting that the level of activity of these two factors helps to specify the distinction between sensory neurons innervating muscle spindles and Golgi tendon organs, and possibly between sensory pool subclasses. In addition, the transcriptional activity of Etv1 is linked to muscle-derived neurotrophic factor signaling, which may serve to regulate the differentiation of proprioceptor subclasses in advance of the challenge of connecting with appropriate motor neuron targets.
Diversity and Microcircuitry of Premotor Interneurons
Local spinal interneurons serve crucial roles in refining sensory-motor connections and establishing motor rhythm, but whether they follow the rules of sensory connectivity remain unclear. We have used transcriptional identity as an entry point to define interneuron subtype diversity and principles of motor connectivity, focusing on V1 inhibitory interneurons. The V1 population can be fractionated into more than 30 subpopulations on the basis of differential expression of nearly two dozen distinct transcription factors. Most of these V1 interneuron subsets settle in discrete dorsoventral and mediolateral domains of the spinal cord, establishing a complex spatial matrix of interneuron–motor neuron positions.
Our studies have shown that V1 interneuron–settling position has a defining role in establishing patterns of sensory input connectivity. As one example, Renshaw interneurons—the mediators of motor neuron recurrent inhibition—settle in an extreme ventral position and, accordingly, receive proprioceptive sensory input from afferents projecting to ventral but not dorsal motor neuron pools. Similarly, Renshaw interneurons fail to provide input to the most dorsally positioned motor pools that innervate foot muscles. Thus interneuronal circuits with motor neurons are constructed in a highly individualized, motor pool–selective manner, presumably to accommodate the distinct biomechanical demands of muscles controlling different limb joints.
One particularly informative instance of the cellular targeting of inhibitory synapses is found in primary sensory systems, where sensory terminals serve both as presynaptic structures that innervate recipient central nervous system neurons and as the postsynaptic target of local inhibitory interneurons at axo-axonic synapses. Sensory terminals in the ventral spinal cord represent the sole target of GABApre neurons, implying stringent recognition specificity in the assembly and organization of this specialized inhibitory microcircuit. We have found that sensory expression of the contactin family protein NB2and the contactin-associated protein Caspr4 are required to establish the normal high-density accumulation of studding of GABApre-derived synaptic boutons on proprioceptive sensory terminals. In a complementary manner, two members of the L1 Ig family, CHL1 and NrCAM, are expressed by GABApre neurons, and their function is required for the formation of high-density GABApre synapses with sensory terminals. These findings pinpoint a molecular recognition system that helps to direct the formation of presynaptic inhibitory synapses.
Manipulating Circuits for Skilled Reach
Skilled forelimb movements have been refined over the course of evolution to the point that they constitute some of the more impressive accomplishments of the mammalian motor system. One cornerstone of the logic of motor systems is the idea that the transition from central plan to smooth performance involves online validation by feedback signals which update the motor system about the fidelity of motor action. One inherent problem in supplying feedback information through the periphery is the temporal delay incurred—information arrives at motor planning centers too late to be of much use in online updating. The motor system appears to have invented numerous tricks to overcome the delay problem—notably the relaying of internal copies of motor commands to sensory processing centers. In addition, such delays adversely impact the gain of feedback signals, invoking the introduction of sensory filtering systems that reduce the likelihood of gain deregulation. Our functional studies aim to probe the neural circuitry that accommodates feedback delay and the wide dynamic range of sensory input.
We are using genetic methods in mice to probe the neural basis and circuit logic of skilled reach. Cervical propriospinal neurons (PNs) represent one class of spinal interneuron implicated in the control of forelimb behavior. PNs serve as intermediary relays for descending motor commands and exhibit a bifurcated axonal output: one branch projects to the cervical motor neurons that control forelimb muscles, and the other projects to neurons in the lateral reticular nucleus, which serves as a precerebellar relay centercorrect The duality of PN axonal projections provides a potential anatomical substrate for copying premotor signals and raises the issue of whether information relayed by this internal copy branch has any impact on forelimb motor output. We reasoned that molecular delineation could provide a genetic means of manipulating the internal copy projections of PNs. We have found that one major population of excitatory PNs is contained in the V2a interneuron class—one of the cardinal subtypes of ventral interneurons implicated in locomotor control. Eliminating cervical V2a neurons elicits a reach-specific defect in forelimb movement, revealed by quantitative three-dimensional kinematics. Conversely, selective activation of the internal PN branch activates a rapid cerebellar feedback loop that excites motor neurons and degrades the fidelity of reaching movements. Thus excitatory PNs form one neural element of an internal feedback circuit for mammalian skilled reaching.
We have also explored the small subset of GABAergic interneurons that form axo-axonic contacts with the terminals of sensory afferents. Despite the occurrence of axo-axonic contacts on most sensory terminals, the predominance of postsynaptic inhibition has left unresolved the motor behavioral significance of this presynaptic control system. One theoretical analysis of sensory-motor transformation has proposed that the gain of proprioceptive sensory feedback needs to be tightly constrained to prevent motor instability. In principle, presynaptic inhibition provides an effective means of imposing gain control, but without a means of selectively manipulating the relevant spinal inhibitory interneurons, it has not been possible to address whether or how presynaptic inhibition influences motor behavior.
We have used the gene encoding the GABA synthetic enzyme GAD2 as a genetic entry point for activating and eliminating presynaptic inhibitory interneurons and assessing their role in motor behavior. We showed that GAD2 interneurons mediate presynaptic inhibition at sensory-motor synapses and that genetic stripping of presynaptic inhibitory boutons from sensory terminals uncovers a pronounced motor oscillation during goal-directed reaching. Our studies identified the GABApre neuron as a neural substrate for suppression of a sensory-motor oscillation that, when unleashed, undermines steady goal-directed forelimb movement. These findings provide insight into the mechanisms for scaling of sensory gain across a wide and dynamic range of afferent firing frequencies, as well as clues about the grain of neural circuitry that empowers presynaptic inhibitory control during sensory-motor transformation. The motor task selectivity of GABApre neuronal recruitment, taken together with the diversity of sensory modalities influenced by presynaptic inhibition, hints at the existence of discrete GABApre neuron subtypes, each devoted to its own sensory feedback system.
Locomotion is another prominent instance of limb motor behavior. The simple repetitive movements that underlie locomotion are regulated by localized neural networks known as central pattern generators. The organization of the locomotor central pattern generator circuit in walking mammals has remained elusive, in part because of the difficulty in identifying and manipulating its intrinsic interneuronal components. We therefore used profiles of transcription factor expression to define distinct sets of ventral interneurons, each with a different intraspinal projection pattern and target connectivity. We have defined classes of interneurons with key roles in establishing left-right alternation in motor activity, in facilitating motor burst activity, and in the core circuitry that establishes locomotor rhythm. As one example, we identified an ipsilaterally projecting excitatory interneuron population, marked by the expression of the transcription factor Shox2. Studies performed with the Ole Kiehn lab (Karolinska Institute, Stockholm) have shown that most Shox2 interneurons are rhythmically active during locomotion and that optogenetic silencing or blocking of the synaptic output of Shox2 interneurons in transgenic mice perturbs rhythm without an effect on pattern generation. These findings identify this genetically marked set of excitatory interneurons as a neural component of the rhythm-generating kernel for locomotion.
Work in the Jessell lab is supported by grants from the National Institute of Neurological Disorders and Stroke, Project ALS, the Harold and Leila Mathers Foundation and the Tow Foundation.
As of March 4, 2014