Contribution of sensory feedback during locomotion in vertebrates
Sensory feedback as a driver of locomotor activity: spinal reflexes vs. half-center models
Studies from the mid-19th century reported that spinal birds and mammals could use the body parts innervated by spinal circuits below the site of transection to produce patterns of locomotion (Flourens, 1824; Freusberg, 1874; Unzer, 1771). These observations led to further studies demonstrating that inputs from sensory afferents elicited by mechanical or electrical skin stimulations were sufficient to produce rhythmic contraction of hindlimb flexor and extensor muscles, thus recapitulating scratching in spinal cats and dogs (Sherrington, 1906a, 1910). At the time, it was suggested that the sequential activation of muscles during these evoked movements might be directly driven by the feedback from muscle contraction itself. Sherrington and his peers formulated the hypothesis that rhythmic move-ments such as stepping in cats were generated by a succession of these so-called “reflexes” in the ab-sence of descending “activating” inputs from the brain (Sherrington, 1906a, 1906b; 1910; reviewed in McCrea, 2001; Hultborn, 2006). Although it was clear that inputs from cutaneous or proprioceptive afferent fibers (Table 1, Figure I.3, reviewed in Rossignol et al., 2006) could provide powerful excita-tion to spinal centers and enable the contraction of limb muscles, it was later shown that spinal cats with deafferented hindlimbs could still present rhythmic patterns of motor output (Brown, 1911, 1914). This observation followed by the work of many other groups suggested that sensory feedback was not necessary for the generation of rhythmic contraction of hindlimb muscles (Brown, 1911, 1914; Brown and Sherrington, 1912; reviewed in Delcomyn, 1980; Stuart and Hultborn; 2008; McCrea and Rybak, 2008). As a matter of fact, Sherrington himself had gathered similar evidence when he found that spinal dogs with deafferented limbs could still perform scratching, a movement that resembles stepping (Sherrington, 1906b).
Altogether this series of pioneer studies gave rise to the idea that the spinal cord contains autonomous “half-center” circuits composed of limb and muscle-specific excitatory neurons that are responsible for the rhythmic activation of limb muscles. These circuits were later defined as single-level locomotor central pattern generators by Wilson and Wyman from their work in the locust (Wilson and Wyman, 1965) and by Grillner and colleagues following their work in the lamprey (Grillner, 1969; reviewed in Grillner, 2003; McCrea and Rybak, 2008; Grillner and Jessell, 2009; Grillner and El Manira, 2015). This large body of work also led to the general understanding that locomotion, while initiated by su-praspinal brain centers, is generated and maintained in the spinal cord.
Even though sensory inputs are not necessary for the production of motor output, they may be im-portant for muscle coordination and for the modulation of circuit activity, notably by interacting with descending circuits and by tuning the excitability of local spinal circuits (Grillner and Rossignol, 1978; Duysens and Pearsons, 1980; Hiebert et al., 1996; Lam and Pearson, 2001). Eliminating inputs from sensory feedback pathways has dramatic effects on locomotion and postural control (Akay, 2014; Takeoka; 2014), as depicted in the example of the disembodied lady from Oliver Sacks, a patient who may have suffered from a rare form of sensory neuritis targeting dorsal root ganglia.
Integration of sensory inputs in spinal circuits
The work from Sherrington, Brown, and colleagues revealed that spinal circuits underlying locomo-tion and sensory pathways emerging from muscles and tendons had complex interactions during movement. Because of the lack of genetic tools and techniques for recording individual neurons, it was not technically possible to probe the cellular organization of spinal circuits receiving these senso-ry signals. This limitation prevented them from understanding to what extent sensory and descending inputs are dynamically integrated at the spinal neuron level in order to generate different types of mo-tor output. The development of intracellular recording techniques allowed researchers to further dis-sect the architecture of afferent feedback circuits and the function of proprioception during locomotion (Brette and Destexhe, 2012; reviewed in Jankoswka and Hammar, 2002; Hultborn, 2006).
Table 1. Proprioceptors in limbed vertebrates. Proprioceptors are located in muscles, tendons, joint lig-aments and in joint capsules. There are no specialized sensory receptor cells for body proprioception. In skeletal (striated) muscle, there are two types of encapsulated proprioceptors, muscle spindles (Ia and II) and Golgi tendon organs (Ib), as well as numerous free nerve endings (III). Within the joints, there are encapsulated end-ings similar to those in skin, as well as numerous free nerve endings. Neuroscience online, an electronic textbook for neurosciences, The University of Texas Medical School at Houston.
Figure 1.3. Location of proprioceptors in the body. (A) The Golgi tendon organ is located at the junction of muscle and tendon. Afferent terminal fibers are intertwined with collagenous fibers of the tendon and the entire organ is encapsulated in a fibrous sheath. (B) A muscle spindle with its sensory and motor innervation. The primary muscle spindle afferent terminates as annulospiral endings in the central area of the intrafusal mus-cles whereas the secondary muscle spindle afferent terminates as flower spray endings in more polar regions of intrafusal muscles. (C) The joint receptors are free nerve endings encapsulated in the joint capsule and joint ligaments. Adapted from Dougherty, Neuroscience online, an electronic textbook for neurosciences, The Univer-sity of Texas Medical School at Houston.
Of particular relevance, the work of many labs, including the labs of Sir John Eccles, Anders Lundberg, and Elzbieta Jankowska revisited the concept of spinal half-centers developed a few dec-ades earlier by T. G Brown, notably by identifying spinal interneurons forming rhythmic circuits that would also be part of reflex circuits (reviewed in Jankowska, 2001; McCrea, 2001; Hultborn, 2006; Jankowska 2008; Stuart and Hultborn, 2008).
Descending pathways reconfigure the output of sensory afferents
After the identification of monoaminergic terminals in the spinal cord (Carlsson et al., 1964), Lundberg and colleagues probed the influence of monoaminergic descending inputs on the regulation of spinal processing by performing intravenous injections of the noradrenergic precursor L-3,4-dihydroxyphenyl-alanine (L-DOPA) and 5-hydroxy-tryptophan (5-HTP) in spinal cats. They found that these compounds modified the pattern of motor activity elicited by stimulation of sensory afferent fibers (Anden et al., 1963, 1964; Jankoswka, et al., 1965). Instead of measuring a short-latency depo-larization of flexor motor neurons, at it is the case for typical reflex responses, Jankowska and col-leagues observed delayed, long-lasting and rhythmic depolarizing volleys in flexor motor neurons, which were reminiscent of a locomotor-like state (Jankoswka, et al., 1967a, 1967b). These results sug-gest that inputs from sensory pathways have different effects on motor pools when supraspinal de-scending neurons are active compared to when they are not (Figure I.4). This observation led to the idea that inputs from descending neurons may either directly modulate the activity of sensory affer-ent’s target motor neurons or activate sets of spinal neurons that would in turn modify the recruitment or excitability of motor neurons in response to sensory stimulations. Alternatively, Jankowska and colleagues found that flexor motor neurons did receive direct reciprocal inhibition when contralateral flexor motor neurons were active. This pioneering work identified the first diagram of connectivity of spinal interneurons with ipsilateral excitatory interneurons recruited by the combination of descending and sensory inputs. These interneurons were proposed to be responsible for the rhythmic activation of motor neurons, while inputs from inhibitory interneurons were similarly recruited to ensure alternation of activity from one side of the body to the other (Jankowska, 1967a, 1967b, reviewed in Hultborn, 2006; McCrea and Rybak 2008).
Figure I.4. Schematic representation of neu-ronal pathways transmitting excitatory action to motoneurones (Mn) from flexor reflex af-ferents (FRA).
Excitatory terminals are indicated by two branches, inhibitory by filled circles. Termination of inhibitory interneurones on a cell body merely indicates inhibi-tory action and there is no commitment as to whether the inhibition is postsynaptic on cell bodies or pre-synaptic on terminals of interneurones. Pathway A is activated in the acute spinal cat (without DOPA) and inhibits pathway B so that no effect is transmitted via this route. After DOPA, transmission in pathway A is partially or completely inhibited (by liberation of transmitter from a descending noradrenergic path-way) thereby removing the inhibition of pathway B through which the late longlasting EPSP is evoked in the motoneurone. From Jankowska et al., 1967a.
Figure I5. Slow excitation of extensor moto-neurons by extensor group I afferents. Rever-sal of the influence of group I afferents from plantaris (PL) on a motor neuron supplying either the lateral gastrocnemius or soleus muscle (LCS) following the administration of L-DOPA in an acute spinal cat. Before L-DOPA, the response was predominantly hyperpolarizing due to summation of disynaptic group lb IPSPs. After L-DOPA, the response was depolarizing due to the opening of an additional ex-citatory pathway. Adapted from Gossard et al., 1994 and from Pearson, 1995.
Figure I.6. Diagrams illustrating convergence from primary afferents (I, prim aff) and de-scending pathways (II, desc) onto common INs projecting to MNs. Neuronal circuits shown on the left with (A) showing excitatory convergence from both sources whereas (B) shows excitation from primary afferents and inhibition from a descending pathway. INs (open circles) represent populations of these cells with similar convergence. Traces in the right-side diagrams show idealized IC records from a single MN after a test stimulus (I, prim aff), a condi-tioning (cond) stimulus (II, desc) and their combina-tion (I+II). From Stuart and Hultborn, 2008.
Along the same line, stimulation of Ib fibers led to different outputs on motor neurons with or without L-DOPA (Gossard et al., 1994). While stimulating group Ib fibers at rest was associated with large volleys of inhibition in recorded target motor neurons, the same stimulation after addition of L-DOPA in the preparation triggered large depolarization of the same motor neurons (Figure I.5, Gossard et al., 1994; Pearson and Collins, 1993; Whelan et al., 1995; reviewed in Pearson, 1995). This phenomenon, termed reflex reversal, demonstrates that sensory afferents are fully integrated within rhythmic spinal circuits and that the effect of their activation depends on the state of spinal circuit’s activation and as a consequence the nature of active spinal neurons at the time of the stimulation. Conversely, these re-sults also indicate that neuromodulators such as dopamine (and additionally for noradrenaline and serotonin) can dramatically modify the output of both spinal circuits and sensory feedback pathways. This body of work is important because it gave rise to some of the most fundamental concepts in the field of sensorimotor integration. First, these results showed that descending circuits interact with sen-sory afferents and can change the output of motor neurons following sensory stimulation. Then, it brought to light the architecture of sensorimotor connectivity by revealing the existence of premotor circuits, which are not active when L-DOPA is absent but seems nonetheless responsible for the rhythmicity observed in these experiments. Finally, it showed that descending commands can modu-late the configuration of active circuits during sensory processing by adding or derecruiting spinal interneurons (Figure I.4). Altogether, these data and a series of follow-up papers (described in Lundberg, 1975 and Burke, 1985 and a long series of papers from Jankowska and colleagues) demon-strated that descending and sensory inputs converge onto spinal microcircuits with the ability to modi-fy their output in a context-dependent manner (Stuart and Hultborn, 2008; Figure I.6). The latter con-cept is particularly important for the work presented in this manuscript.
State-dependent modulation of motor output by sensory feedback
Using L-DOPA to induce locomotor-like states in decerebrate and acute spinal cats allowed research-ers to investigate the effects of activating sensory pathways during different contexts, either at rest or during fictive locomotion. It was demonstrated that the timing of Ia or group II afferents activation would determine whether these sensory inputs could entrain the rhythm if they are elicited at rest (without L-DOPA) or reset the rhythmicity of spinal neurons in L-DOPA (Figure I.7, Conway et al., 1987; Kriel-laars et al., 1994; Guertin et al.,1995). Thus, sensory feedback can participate in controlling the transi-tion from one phase to the next during the step cycle. This work led to the concept that sensory feed-back and local circuits in the spinal cord interact in a complex and state-dependent manner to control the timing of muscle coordination during complex motor sequences (McCrea, 2001). One interpreta-tion for these findings is that sensory feedback could compensate for perturbations occurring in the locomotor environment or correct non-linear mechanics during movement (Stuart, 1999; McCrea, 2001).
Table of contents :
I. Contribution of sensory feedback during locomotion in vertebrates
I.1. Sensory feedback as a driver of locomotor activity: spinal reflexes vs. half-center models
I.2. Integration of sensory inputs in spinal circuits
I.2.a. Descending pathways reconfigure the output of sensory afferents
I.2.b. State-dependent modulation of motor output by sensory feedback
I.2.c. Modulation of sensory signaling by presynaptic inhibition
I.3. Connectivity between primary afferent neurons and spinal premotor interneurons
I.4. Cellular and molecular investigation of proprioception in mammals
I.5. Sensory feedback in fish and amphibians
II. Anatomical, genetic, and functional organization of locomotor central pattern generators
II.1. Basic organization of spinal circuits controlling the pattern and rhythm of locomotion
II.2. Inductive signals control the fate of spinal neuron specification during development
II.3. Contribution of genetically identified spinal neurons to locomotion
II.3.a. V0 neurons
II.3.b. V1 neurons
II.3.c. V2a neurons
II.3.c. V3 neurons
II.3.d. Sensory-motor interactions between locomotor CPGs and sensory afferents
III. Cerebrospinal fluid contacting neurons are polymodal sensory neurons in the ventral spinal cord
III.1. Identification and characterization of sensory neurons lining the central canal in the spinal cord
III.2. Molecular analysis of CSF-cNs across species reveals specific markers and a double developmental origin
III.3. Sensory modalities underlying the recruitment of CSF-cNs
III.4. Experimental strategies and optogenetic approaches to unravel the function of CSF-cNs Key concepts of part III
IV. Aims of the thesis
Chapter 1 – Functional connectivity mapping between CSF-cNs and spinal premotor neurons controlling slow
Predictions regarding the connectivity of CSF-cNs
Building the experimental paradigm
Highlights of the findings described in this chapter
Graphical abstract of the results
Published article: State-Dependent Modulation of Locomotion by GABAergic Spinal
Figures and Supplemental Information
Discussion and perspectives
On the specificity of CSF-cNs connectivity
CSF-cNs target V3 interneurons, a second class of ventral glutamatergic interneurons
On the modulation of bout generation
The problem of chloride homeostasis in developing neuronal networks
On the role of CSF-contacting neurons during slow swimming
Can CSF-cNs modulate distinct locomotor behaviors using circuit-specific mechanisms?
Analysis of the contribution of CSF-cNs to evoked fast locomotion
Chapter 2- Modulation of spinal circuits controlling the frequency of locomotion by cerebrospinal fluid-contacting neurons
BoTxBLC-mediated silencing of V2a interneurons confirms their critical role in fast locomotion
Silencing of V2a interneurons decreases the locomotor frequency during spontaneous slow swimming
Topographic organization of CSF-cNs inputs onto V2a interneurons
Discussion and perspectives
Experimental strategy to probe the modulation of V2a neurons by CSF-cNs
Some assembly required: building a global picture of CSF-cNs-mediated modulation of spinal central pattern generators
Are V2a interneurons at the core of a global mechanosensory feedback loop?
Conclusions and future directions