R E V I E W

0166-2236/00/$ – see front matter © 2000 Elsevier Science Ltd. All rights reserved. PII: S0166-2236(99)01535-0 TINS Vol. 23, No. 5, 2000 199

AMONG THE SENSORY–MOTOR LOOPS that havebeen described, the stretch reflex illustrates the

simplest and the most-widespread proprioceptivefeedback system in both vertebrates1 and invertebrates(called the resistance reflex in this group2–4): when amuscle is stretched, sensory feedback causes an acti-vation of the motoneurons innervating that muscle.Functionally, this postural negative feedback helps tomaintain a given position, and for the past decade,this reflex has been studied in a number of vertebrateand invertebrate species. Recent information hasshown that it can be modulated at different locationswithin the neural circuits. Indeed, during the pastdecade, a number of studies have pointed out the rela-tionship between sensory–motor pathways and cen-trally generated rhythmic motor activities (for a reviewsee Ref. 5). The concept of a central pattern generatorwas initially proposed to describe populations of neur-ons that elaborate a basic rhythmic motor-pattern ac-tivity in the absence of any movement-related sensoryfeedback. First demonstrated in locust6, central patterngenerators have subsequently been described in manyinvertebrate and vertebrate models (for a review seeRef. 7). In addition to controlling motor output, centralpattern generators also exert control over sensory–motorpathways in many systems. For example, during rhyth-mic motor activities such as walking, the ‘stretch’ reflexis modulated not only in intensity, but also in sign (fora review see Ref. 8). Finally, long-term changes haverecently been described for this reflex9,10, making it apossible target for motor learning. A striking feature ofthese studies is that the central control of this reflex ap-pears to share common properties in different vertebrateand invertebrate animal models.

The stretch reflex in the crayfish walking system hasbeen studied for more than ten years, and a wealth ofinformation has accumulated on the diverse sites andmechanisms involved in reflex modulation, from thecellular to the network level. This short article gives anoverview of all these central mechanisms that control

this reflex in the crayfish walking system, and comparesthese findings with those described in other systems.In addition, it presents an analysis of how these vari-ous mechanisms operate simultaneously or alternatelyin different states of the locomotor system, in order toensure different behavioral functions. Modulation of thestretch reflex will be considered successively at threelevels: in sensory afferents, in interneuronal pathwaysand in motoneurons. In addition to these immediateregulatory mechanisms, long-term modifications willbe also considered.

Organization of the ‘stretch’ reflex

The stretch reflex is a negative-feedback system,involving proprioceptors that detect and code for thegeometry and changes in position of joints. Althoughvertebrates, insects and crustaceans use different types ofproprioceptors (Fig. 1), this negative-feedback systemobeys similar principles: proprioceptive organs, whichlie in parallel with skeletal muscles, activate musclesthat counteract the imposed movement.

In vertebrates, the proprioceptors involved in the re-flex are the muscle spindles (Fig. 1a), which lie withinskeletal muscles, parallel to the muscle fibers, wherethey respond to stretch of the muscle. Muscle spindlesconsist of intrafusal fibers surrounded in their centralregion by two types of sensory endings (primary andsecondary). There is usually only one primary endingin each spindle, consisting of a single group-Ia afferentaxon, and only one secondary ending that consists ofthe branches of a single group-II afferent axon. Thegroup-Ia afferent neurons excite the motoneurons ofthe same muscle monosynaptically1 (Fig. 1a).

In insects and crustaceans, the proprioceptors in-volved in negative feedback are mainly chordotonalorgans (Fig. 1b), which consist of an elastic strand thatcrosses the joint. There is often just one chordotonalorgan in each joint, located outside the skeletal musclesbut parallel to one of them (see Box 1). In the elasticstrand of the chordotonal organ, tens of sensory neurons

Central control components of a ‘simple’ stretch reflexFrançois Clarac, Daniel Cattaert and Didier Le Ray

The monosynaptic stretch reflex is a fundamental feature of sensory–motor organization inmost animal groups. In isolation, it serves largely as a negative feedback devoted to posturalcontrols; however, when it is involved in diverse movements, it can be modified by centralcommand circuits. In order to understand the implications of such modifications, a modelsystem has been chosen that has been studied at many different levels: the crayfish walkingsystem. Recent studies have revealed several levels of control and modulation (for example, atthe levels of the sensory afferent and the output synapse from the sensory afferent, and viachanges in the membrane properties of the postsynaptic neuron) that operate complex andhighly adaptive sensory–motor processing. During a given motor task, such mechanisms reshape the sensory message completely, such that the stretch reflex becomes a part of thecentral motor command.Trends Neurosci. (2000) 23, 199–208

François Claracand Didier Le Rayare at theLaboratoire deNeurobiologie etMouvements, UPR9011 du CentreNational de laRechercheScientifique,Institut Fédératif deRecherche ‘Sciencesdu Cerveau’,13402 MarseilleCedex 20, France,and DanielCattaert is at theLaboratoire deNeurobiologie desRéseaux, UMR5816 du CentreNational de laRechercheScientifique,Université deBordeaux 1,33405 Talencecedex, France.

F. Clarac et al. – A ‘simple’ stretch reflexRE V I E W

200 TINS Vol. 23, No. 5, 2000

coding for the different parameters of positions andmovements are divided into two functionally distinctgroups: some respond to stretch and others respond torelease. However, as is the case for vertebrate group-Iaafferents, each of these sensory groups excites mono-synaptically the motoneurons of the muscle thatcounteract the imposed movement2,4 (Fig. 1b).

In vertebrates, the contractile polar regions of intra-fusal fibers are innervated by g motoneurons. Whenactivated, these motoneurons change the sensitivity ofthe sensory endings to stretch (for a review see Ref. 11).In arthropods, a similar arrangement exists in one typeof proprioceptive organ, the muscle-receptor organ; how-ever, in contrast to chordotonal organs, muscle-receptororgans are not present in all leg joints12. Although thistype of control is also important at some joints, theefferent control of proprioceptors will not be discussed.

The direct convergence of sensory fibers onto eachmotoneuron is a shared feature of the resistance reflexin arthropods (insects2, crustacea3,4) and of the stretchreflex in vertebrates13. For example, in crayfish, whereeight of the twelve depressor motoneurons are involvedin the stretch reflex, each of these motoneurons isexcited monosynaptically by two to five of the 20 pro-prioceptive neurons from the coxo–basipodite chordo-tonal organ (CBCO) that respond to levation of thebasipodite (Fig. 2). A much larger proportion of moto-neurons (up to 100%) seem to be involved in the stretchreflex in the cat spinal cord, where a substantial numberof single Ia afferent fibers have been demonstrated toproject onto most (65–80%) of the homonymous moto-neurons15. However, the details of the connections arebetter known in crayfish, in which all the motoneuronsinnervating a given muscle can be recorded successivelyin the same experiment. In this case, it appears that, inaddition to convergence of multiple sensory fibers ontoeach motoneuron, there is also considerable, but highlyheterogeneous, divergence of sensory neurons onto themotoneurons. A single CBCO fiber16, for example, canproject onto from one to eight motoneurons.

Ia inhibitoryinterneuron

Inhibitoryinterneuron

Ia afferent

Spindle

Chordotonalorgan

FF

EE

(a) (b)

trends in Neurosciences

Fig. 1. Comparison of the organization of the ‘stretch’ reflex in vertebrates and arthropods.(a) In vertebrates, proprioceptors involved in the stretch reflex are spindle muscles, which consistof intrafusal fibers connected to primary afferent neurons. In the two antagonistic muscles represented, a sensory neuron (Ia afferent) monosynaptically excites motoneurons of thehomonymous muscle (F, flexor; E, extensor). When an extension movement is imposed on the joint(green arrow), the flexor muscle and its spindles are stretched. This information is conveyed by thecorresponding Ia afferent (blue) that excites the flexor motoneuron (gray). Subsequently, thismotoneuron commands the contraction of the flexor muscle and opposes the imposed movement(gray arrow). In addition to the excitation of the homonymous motoneurons, the antagonisticmotoneurons are inhibited (reciprocal inhibition) by a disynaptic pathway involving a Ia inhibitoryinterneuron (the orange pathway represents the flexor muscle Ia afferent. (b) In arthropods,chordotonal organs are the proprioceptors involved in the ‘stretch’ reflex. In contrast to musclespindles of vertebrates, chordotonal organs are located outside the skeletal muscles. They consistof an elastic strand (gray bar) that crosses the joint. This strand is stretched when the joint opens,and released when the joint closes. In the strand, two populations of sensory neurons (blue andorange circles) code for these two directions of movement and monosynaptically connect the setof motoneurons (flexors, F; extensors, E) that antagonize the movement (resistance feedback).When an extension movement is imposed on the joint (green arrow) only one population ofsensory neurons is activated (blue). These sensory neurons activate monosynaptically the moto-neurons of the flexor muscle that opposes (gray arrow) the imposed movement. In addition tothe excitation of the flexor motoneurons, the antagonistic extensor motoneurons are inhibited(reciprocal inhibition) by a pathway involving an inhibitory interneuron.

Unlike vertebrates, in which proprioceptive coding of movements ismediated by stretch-sensitive fusorial organs inserted into each antag-onistic muscle, a single proprioceptor (chordotonal organ) codes forboth directions of angular joint movement in crustaceansa. At the baseof the leg, the second joint (coxo–basipodite) is responsible for theupward and downward movements driven by levator and depressormuscles, respectively. This joint, which is involved in both the swingand the stance phases, is essential for walking activity of the animal.The coxo–basipodite chordotonal organ (CBCO) is composed of anelastic strand that crosses the coxo–basipodite joint. Sensory bipolarcells are inserted in this strand. Some are sensitive to the stretch ofthe strand, which occurs during downward leg movements, as thelevator muscles are stretched; others are sensitive to the release of thestrand, which occurs during upward leg movements, as the depressormuscles are stretched.

An in vitro preparation of the region of the CNS that commandsleg movements has been developedb. In this preparation, the CBCOis dissected and pinned out in a Petri dish in such a position that amechanical puller attached to the distal end of the CBCO imposesstretches and releases to the strand in exactly the same way as doesmovement of the intact leg during locomotion. Using this prepara-tion, which allows paired intracellular recording of identified moto-neurons and CBCO sensory terminals, it is possible to record intra-

cellularly from sensory neurons and to study the monosynapticconnections they make with identified motoneurons.

The network studied using this in vitro preparation consists of about80 neurons: the CBCO is composed of 40 bipolar cells, and the outputconsists of 12 depressor motoneurons and 19 levator motoneurons.In addition, a dozen or so interneurons are involved in polysynapticpathways between CBCO neurons and these motoneurons. AmongCBCO sensory neurons, 20 respond when the strand is stretched,and the remaining 20 respond when the strand is released. The reflexresponses to imposed leg movements can be studied either whenthe network is in a tonic state or during rhythmic activity induced byoxotremorine, a muscarinic ACh-receptor agonistc. Moreover, by usinga high-Ca21 and high-Mg21 saline that raises the threshold for spiking,monosynaptic sensory–motor connections can be specifically studiedd.

Referencesa Bush, B.M.H. and Laverack, M.S. (1982) The Biology of Crustacea (Atwood,

H.L. and Sandeman, D.C., eds), pp. 397–467, Academic Pressb El Manira, A. et al. (1991) Monosynaptic connections mediate resistance reflex

in crayfish (Procambarus clarkii) walking legs. J. Comp. Physiol. 168, 337–349c Cattaert, D. et al. (1994) Nicotinic and muscarinic activation of motoneurons

in the crayfish locomotor network. J. Neurophysiol. 72, 1622–1633d Berry, M.S. and Pentreath, V.W. (1976) Criteria for distinguishing between

monosynaptic and polysynaptic transmission. Brain Res. 105, 1–20

Box 1.The in vitro preparation of the crayfish locomotor network

Other components of thesensory–motor system

Although most of the sensory–motor connections are mediatedby classical chemical synapses,electrical coupling has also beendescribed in some sensory–motorconnections, at least in some ani-mal groups. For example, in crus-tacea, in the thoraco–coxal muscle-receptor organ, a proprioceptivestructure that codes movementsand positions of the first leg joint,the two sensory fibers are coupledelectrically to motoneurons17. Simi-lar electrical connections also existbetween Ia afferents and moto-neurons in the frog18.

In parallel with monosynapticchemical and electrical connec-tions, the negative-feedback reflexalso involves polysynaptic pathwaysthrough spiking interneurons. Invertebrates, a disynaptic pathwayinvolving Ia inhibitory inter-neurons inhibits the antagonistmotoneurons and is responsiblefor the reciprocal innervation de-scribed by Sherrington (see Fig. 1a).In addition, recent studies suggestthat muscle–spindle afferents canactivate the excitatory interneuronsof disynaptic pathways to homony-mous motoneurons, and activateother interneurons involved in thehalf-center responsible for loco-motion5 (see below). In arthropods,polysynaptic pathways also involvenon-spiking interneurons, whichare the premotor elements thatcontrol motoneuron activity inlocust2,19 and in stick insects20,21,22.Relative to that of vertebrates, thewiring of the reciprocal innervationdemonstrated in the locust23 is more complex, andinvolves both spiking and non-spiking interneurons.

The polysynaptic pathways are partially responsiblefor the fact that the negative-feedback reflex is not rigidin arthropods24 and vertebrates25. Indeed, this reflex canbe modified not only in intensity but also in sign, whenthe resistance (negative feedback) reflex is reversed intoan assistance (positive feedback) reflex. This reflex rever-sal involves both presynaptic inhibition of primaryafferents in both vertebrates26 and invertebrates27, andchanges in the activation of spiking and non-spikinginterneurons of the polysynaptic pathways in insects28

and crustacea14.

The sensory terminal as a sensory-processingstructure

It has been known for many years that the afferentmessage can be modified by presynaptic inhibitionwithin sensory axons in vertebrates29. More-recentresults indicate that, at least in crustaceans, two othermechanisms might modify the function of sensoryneurons: (1) the presence of electrical synapses betweensensory afferent terminals30, and (2) the possibility of

modifying the sensitivity of proprioceptive neuronsby neuromodulatory substances31,32.

Nearly 40 years ago, primary afferents were shownto be the site of presynaptic modulation in both verte-brates and invertebrates. In 1957, Frank and Fuortes re-ported the first evidence of presynaptic inhibition in catgroup-I afferent fibers29. During fictive locomotion inthe cat, these afferent fibers display rhythmic bursts ofPADs (primary afferent depolarizations) that are phase-locked with the locomotor rhythm33. Similar PADs ofcentral origin have been reported in crayfish. In insects,in addition to PADs that originate centrally, chordotonalafferents are presynaptically inhibited by PADs gener-ated by other sensory afferents of the same sense organ34.However, such presynaptic inhibition of sensory originhas never been observed in chordotonal terminals ofthe crayfish. Recent studies have demonstrated that themechanisms that underlie PADs are similar in verte-brates and invertebrates26,27,35,36 (with some specific dif-ferences in the case of insects, see below). However, inspite of the similarities in the mechanisms underlyingPADs, there are differences in the mechanisms by whichthe PADs produce presynaptic inhibition.

Fig. 2. The sensory–motor system controlling the crayfish second leg joint. (a) and (b) Arrangement of the chordotonalorgan (CBCO) at the coxo–basipodite joint (a) commanded by levator and depressor muscles for upward and downwardmovements, respectively (b). (c) In vitro arrangement of the ventral nerve cord of the crayfish, together with the motorand sensory nerves to the fifth leg. The CBCO proprioceptor can be stimulated mechanically to mimic leg movements.(d) A CBCO sensory terminal (CBCO terminal) and a depressor motoneuron in the fifth thoracic ganglion. (e) Paired intra-cellular recordings from a CBCO terminal coding for upward leg movement and a depressor motoneuron show the re-sponse of the motoneuron (excitatory postsynaptic potential) to the CBCO sensory spike. (f) Responses of proprioceptiveneurons and motoneurons to leg movements. At the bottom are two classes of CBCO terminals that respond to downwardleg movements (CBCO terminal 1) and to upward leg movements (CBCO terminal 2), respectively. The monosynapticresponses of all 12 depressor motoneurons (successive intracellular recordings in the same experiment, in a high-Ca21 andhigh-Mg21 saline) during downward (left column) and upward (right column) ramp movements imposed on the CBCOstrand are presented above. CBCO sensory neurons activated by upward movements monosynaptically activate eightdepressor motoneurons, whereas CBCO sensory neurons activated by downward movements monosynaptically activateone depressor assistance motoneuron. Three depressor motoneurons are not monosynaptically connected to either groupof CBCO neurons. (f) adapted, with permission, from Ref. 14.

TINS Vol. 23, No. 5, 2000 201

F. Clarac et al. – A ‘simple’ stretch reflex RE V I E W

F. Clarac et al. – A ‘simple’ stretch reflexRE V I E W

202 TINS Vol. 23, No. 5, 2000

Recent studies of crayfish have provided new insightinto the mechanisms of presynaptic inhibition37. Duringoxotremorine-induced rhythmic activity, intracellularrecordings from CBCO terminals in the thoracic gan-glion that commands the corresponding leg revealedthe presence of PADs occurring in phase with rhythmicdepressor motor bursts (Fig. 3a). PADs were demon-strated to be inhibitory in experiments that involvedsimultaneous intracellular recordings from a CBCO ter-minal and a postsynaptic motoneuron (Fig. 3b). WhenPADs occurred, the amplitude of both the orthodromicspike and the corresponding excitatory postsynapticpotential in the postsynaptic motoneuron were reducedproportionally to PAD amplitude27. Similar effects werereproduced by direct pressure-application of GABAonto the CBCO terminals27. PAD-mediated presynapticinhibition in crayfish CBCO afferents is produced byinhibitory interneurons (the PADIs – not yet identified)using GABA as their neurotransmitter, and activating aGABAA-like receptor38 associated with a Cl2 conductance,the reversal potential of which is around 235 mV(which explains why PADs are depolarizing). The gen-eration of PADs in cat primary afferent neurons appearsto involve the same mechanisms: the depolarizationsare mediated by GABA, and involve Cl2 conductanceswith similar reversal potentials39. By contrast, the rever-sal potential for Cl2 is much closer to the resting mem-brane potential in insect sensory neurons36, and theamplitude of PADs, therefore, never exceeds a few milli-volts. The mechanisms of GABA-mediated presynaptic

inhibition are therefore likely to be different in thethree groups. Although in insects this inhibition prob-ably results from a shunting effect, in the cat, owingto the large amplitude of PADs, inactivation of Na1

channels probably has a major role40. Crayfish shouldrepresent an intermediate situation in which small PADsare exclusively shunting, whereas large ones exert botha shunting effect and inactivate Na1 channels37.

In addition to these two mechanisms (shunting andinactivation of Na1 channels), a third mechanism,which acts at the level of the sensory neuron, seems tobe involved in the phasic modulation of the reflex inboth cat and crayfish. Owing to the very depolarizedvalue of the equilibrium potential for Cl2, the ampli-tude of rhythmic PAD bursts might be large enough toelicit antidromic spikes during rhythmic motor activityin the cat33 and in crayfish41. The analysis of antidromicdischarges in CBCO neurons has shown that they arerelated directly to the GABA-mediated increase of Cl2

conductance41. A recent study demonstrated that anti-dromic discharges exerted a powerful direct inhibitionon peripheral sensory coding by CBCO neurons at thesite of mechanotransduction itself42. Increasing the fre-quency and the duration of the antidromic burst resultsin a reduction in frequency of the sensory dischargeobserved in CBCO neurons that code for position. Athigh frequency (50–100 Hz), antidromic trains generallyresult in the cessation of sensory input activity that canoutlast the antidromic train for up to 500 ms. Thus,depending on the level of activation of the Cl2 channel

Fig. 3. Sensory processing in primary afferents. (a) During oxotremorine-induced rhythmic activity, monitored by the rhythmic bursts of spikesrecorded from the nerve that innervates the depressor muscle, an intracellular recording from a chordotonal organ (CBCO) terminal displays bursts ofprimary afferent depolarizations (PADs) time-locked with the depressor bursts. (b) During each PAD burst, the amplitude of sensory spikes is reduced,as is the amplitude of the corresponding excitatory postsynaptic potentials recorded from a postsynaptic levator motoneuron. Paired intracellularrecordings from a CBCO terminal and a levator motoneuron were performed. The two superimposed recordings were obtained in the absence (black)and in the presence (gray) of a PAD. The amplitudes of the sensory spike in the CBCO terminal and the corresponding excitatory postsynapticpotential in the levator motoneuron are reduced (cf. black with gray). In the inset, the excitatory postsynaptic potentials obtained in these two situa-tions have been enlarged to show more clearly the reduction of excitatory postsynaptic potential amplitude when a PAD was present in the CBCOterminal . (c) and (d) Electrical coupling between CBCO sensory neurons. Two coupled CBCO terminals (CBCO terminal 1, CBCO terminal 2) have beenstained with Lucifer Yellow (intracellularly injected into only one terminal) and analyzed with the confocal microscope (c). Physiological evidence forelectrical coupling is shown (d). Depolarizing and hyperpolarizing current pulses were injected into one terminal and the corresponding responseswere recorded in the electrically coupled terminal (terminal 1 and terminal 2 are represented in black and gray, respectively). (e) 5-HT enhancesthe response to stretch movement (down) of an intracellularly recorded CBCO terminal. Scale bar, 20 mm. (b) adapted, with permission, fromRef. 27, (c) and (d) adapted, with permission, from Ref. 30, and (e) adapted, with permission, from Ref. 32.

TINS Vol. 23, No. 5, 2000 203

associated with the GABAA-type receptor in their pri-mary afferents, crayfish, and possibly vertebrates, canmodulate the proprioceptive message in three ways:(1) small PADs can finely adjust the level of synaptictransmission of the sensory message to the moto-neurons by a local shunting mechanism (see Fig. 3b);(2) larger PADs can inactivate the Na1 channel over alarge distance40 (and therefore result in a completeblock of transmission of the proprioceptive signal); and(3) during larger amplitude PADs, antidromic burstscan block the sensory activity of the proprioceptiveneurons for a longer duration. The latter two effectswould prevent co-contraction of antagonistic musclesduring ‘active’ programmed movements.

In addition to presynaptic inhibition, proprioceptiveafferents in crayfish might use electrical connectionsbetween sensory fibers in order to achieve complex dataprocessing30. Confocal analysis of dye-coupled axonshas revealed the existence of large zones of close appo-sitions between Lucifer-Yellow-stained axons (Fig. 3c).By increasing the amount of transmitter released30, weakelectrical coupling (Fig. 3d) serves to enhance trans-mission of sensory signal from afferents that share thesame coding properties onto depressor motoneurons.Stronger electrical coupling can induce spike trigger-ing in the coupled CBCO afferent, which results in anincreased number of active afferents. This mechanismcould act as a coincidence detector43,44 and result in amore-precise and effective resistance reflex.

Neuromodulation constitutes a third level of controlof the proprioceptive message. It is generally achievedby neuroactive substances that are either released byneurons, or are present in the blood. In the presenceof such substances, the properties of the neurons arechanged (gating effects). Sometimes, these changes per-sist after the neuroactive substance has been removed(trigger effects). Modulatory effects on sensory neuronswere first described in the lobster oval organ31 (sensoryelement of the ventilatory system): although 5-HTinhibits the firing activity of this mechanoreceptor,octopamine and proctolin enhance it. Similarly, in cray-fish, 5-HT modulates both sensory coding32 and recip-rocal inhibition between antagonistic motoneurons45.In fact, 5-HT exerts a dual effect on the coding charac-teristics of CBCO sensory neurons that depends on itsconcentration. At low concentrations (1029 to 1026

M),the sensory neurons (phaso–tonic and phasic) increasetheir discharge (Fig. 3e), resulting in an enhancementof the stretch reflex. However, at higher doses (1024

M),the effect of 5-HT is reversed, and the discharge fre-quency diminishes32. The mechanisms by which 5-HTmodifies the coding of movement parameters are stillunknown. It should be noted that neuromodulatoryeffects can affect only specific sensory–motor pathways.For example, in the stick insect, octopamine inhibitsthe resistance reflex pathways, but does not affect posi-tive feedback in active animals46; in the locust,octopamine enhances the position (tonic) componentresponse of the femur chordotonal organ, but not themovement (phasic) component47.

Modulation of polysynaptic pathways

Although phasic locomotor-related presynaptic inhib-ition has been described in the lamprey48 and in thecat33, the mechanisms by which reflex reversal occurs arenot yet known in vertebrates. In parallel to presynapticmodulation of proprioceptive inputs during fictive loco-

motion, reflex reversal involves changes of the poly-synaptic pathway interneurons. This level is much moreaccessible in invertebrates; hence most of the datadescribed below are from studies on invertebrates.

In insects, non-spiking interneurons are local inter-neurons that are involved in the polysynaptic pathwaysfrom proprioceptors to motoneurons2. Such pathwayswere characterized first in the locust (Fig. 4a, see Ref. 2for a review). Their participation in reflex reversal wasstudied primarily in the stick insect20,28,49, where it hasbeen shown that the sign of the reflex is the result of abalance between excitatory and inhibitory non-spikinginterneurons (NSIs). The level of activity of each typeof non-spiking interneuron is dependent on the ‘state’of the preparation, that is, the motor program beingengaged. In addition to these central control pathwaysin the locust23 and in the stick insect20, some non-spikinginterneurons are monosynaptically excited or disynapti-cally inhibited, or both, by proprioceptive inputs. Forexample, in the stick insect, spiking interneurons (SINs),which are activated by proprioceptive afferents, in-hibit the non-spiking interneurons20. Such pathwaysmight serve to regulate the positive-feedback reflex.When an active movement occurs, positive feedbackautomatically increases the excitation of the moto-neurons that command this movement, whose velocitythus increases continuously. However, at this stage, thepresence of the disynaptic inhibitory pathway, whichconveys velocity information, limits the velocity ofmovement. In crayfish (Fig. 4c), during the assistancereflex, a group of non-spiking interneurons, the assis-tance reflex interneurons (ARIN), receive monosynapticexcitatory postsynaptic potentials from CBCO neurons,and connect directly to the motoneurons that will helpthe ongoing movement (positive feedback)14. Assistancereflex interneurons are activated strongly by movement-sensitive CBCO neurons. Furthermore, at least in thecase of the assistance reflex interneurons that reinforcedepressor motoneuron activity during downwardmovements of the leg, up to eight velocity-codingdownward-movement-sensitive CBCO sensory neuronsconverge onto a single assistance-reflex interneuron.However, without any regulatory system, such positivefeedback could have dangerous consequences becauseit can engender instability. The existence of such aregulatory system is indicated by the following obser-vation. Whereas low velocity (0.05 mm/s) movementsimposed on the CBCO elicit only compound excitatorypostsynaptic potentials in assistance reflex interneurons,during high velocity (0.25 m/s) movements, the excit-atory response is blocked rapidly by a compoundinhibitory postsynaptic potential. The inhibition ob-served in assistance reflex interneurons has been attrib-uted to an assistance-reflex-controlling interneuron14

(ARCIN). Assistance-reflex-controlling interneurons arehighly dependent on the velocity of joint movement:the faster the movement, the more strongly they inhibitthe assistance reflex interneurons. This gain controlmechanism could have an essential role in limitingthe positive feedback loop and, thus, preventing thevelocity of the movement from becoming excessive.

In the cat, polysynaptic pathways also exist in par-allel with monosynaptic connections between group-Iafferent fibers and motoneurons (Fig. 4d). During loco-motion, the stance phase can be facilitated by spindleand tendon organ afferents of extensors via three path-ways50: a monosynaptic pathway from group-I afferent

F. Clarac et al. – A ‘simple’ stretch reflex RE V I E W

F. Clarac et al. – A ‘simple’ stretch reflexRE V I E W

204 TINS Vol. 23, No. 5, 2000

fibers, a disynaptic pathway from group-Ia and -Ib affer-ents, and a polysynaptic pathway via the extensor half-center from group-Ia and -Ib afferents. At the same time,spindle afferents from flexor muscles are activatedduring the stance phase, and consequently shortenthis phase, most probably by inhibiting activity in theextensor half-center5.

In contrast to the simple cases pre-sented above for both invertebrateand vertebrate sensory-motor syn-apses, interactions between sensoryterminals and postsynaptic moto-neurons can be very complex. Thisis the case for the thoraco–coxalmuscle receptor organ17 (Fig. 4e), acrustacean proprioceptive organconsisting of two non-spiking sensory neurons, a static (S) fiberand a dynamic (T) fiber, which arestretched when the leg moves back-ward (remotion). At rest, the nega-tive-feedback reflex is due primarilyto the monosynaptic activation ofpromotor motoneurons by T-fibersduring remotion. During rhythmicactivity, a reversal of the reflex occurs,and the T fiber excites remotormotoneurons during remotion.Note that within the group of pro-motor motoneurons, some are ex-cited by the T fiber, whereas othersare inhibited (probably via an inter-neuron). Therefore, the sign of thereflex depends on the balance be-tween excitatory and inhibitory in-fluences produced by T fibers thatsynapse with promotor moto-neurons, a situation resembling thatdescribed for non-spiking inter-neurons in insects (Fig. 4a,b).

Implication of motoneurons inthe regulation of proprioceptivereflexes

Are motoneurons passive outputelements or do they participateactively in shaping the reflex re-sponse? Increasing data in both ver-tebrates and invertebrates indicatethat motoneurons can display ac-tive membrane properties and makeoutput connections onto otherneurons in the central network.Active properties in motoneuronsresult from the existence of voltage-dependent conductances in theirmembranes. However, such non-linear and oscillatory membraneproperties of motoneurons arerarely spontaneously expressed51,but are, in most cases, seen only in the presence of pharmacologi-cal substances such as 5-HT52,53,NMDA (Ref. 57) or agonists of mus-carinic ACh receptors54. For exam-ple, in vertebrates, NMDA-inducedtetrodotoxin-resistant voltage oscil-

lations in the membrane potential of motoneuronshave been observed in lamprey55, Rana tadpole56, neo-natal rat57 and turtle58 motoneurons. Thus, the statusof motoneurons in at least some vertebrate groups isnot very different from that previously shown to existin many invertebrate motor networks (see Ref. 59 fora review).

trends in Neurosciences

NSIs

NSIs

SINsSINs

LvMNs

DepMNsSNs

CoxopoditeBasipodite

Levation

ARIN

PADIs

PADIs

ARCIN

Crayfish CBCO CPG

SINs

SINs

SNs NSIs

SINsSINs

ExMNs

PADIs

PADIsFlexion

Femur

Tibia

Stick insect femoralchordotonal organ CPG

CPG

?

(b)(a)

(c)

T

S

Coxopodite

Rem

Pro

Crayfish thoraco–coxalmuscle receptor organ

Remotion

(e)E

F

INs

INs

INs

ExMN

FlxMN

Ia

Ib

Ia/II

Stance

Cat muscle afferents

CPG

(d)

NSIs

NSIs

NSIs

ExMN

FlxMN

SNs

SINsSINs

ExtensionTibia

Femur

Proprioceptiveafferents

Proprioceptiveafferents

Proprioceptiveafferents

Flexor group-Ia/IIproprioceptive

afferents

Extensor group-Iproprioceptive

afferents

Locust femoralchordotona organ

SINs

Fig. 4. Comparison of the pathways involved in modulation of the stretch reflex in different animals. (a) and (b)In the locust (a) and in the stick insect (b), proprioceptive afferents are presynaptically inhibited via primary afferentdepolarizations (PADs) by at least two groups of spiking interneurons (SINs) known as PADIs (PAD interneurons – notyet identified). Some PADIs (the blue SINs) are activated by other proprioceptive fibers from the same chordotonal organ,and exert an online automatic gain control; other PADIs (red SINs) are activated by the central pattern generator (CPG)through unknown pathways (broken line), and are involved in the phasic modulation of the sensory–motor synapse duringthe walking cycle. In the locust (a), parallel monosynaptic and disynaptic (via a non-spiking interneuron) pathways fromproprioceptive afferents coding for the extension of the tibia activate the flexor motoneurons (resistance reflex). As is the casein vertebrates, the antagonistic motoneurons (extensors) are inhibited whereas the muscles that oppose the imposedmovement (flexors) are activated. However, in the case of the locust, this inhibition is achieved via at least four pathwaysinvolving spiking (SINs) and non-spiking interneurons (NSIs). In the stick insect (b), in parallel with monosynaptic con-nections, flexion-sensitive proprioceptive neurons are involved in polysynaptic pathways via a group of NSIs that makeexcitatory and inhibitory connections on extensor motoneurons (Ex MNs). Depending on the balance between these twoeffects (which are controlled by the CPG), a positive or a negative feedback is elicited. A third class of SINs (yellow) regulates the gain of the disynaptic pathway. (c) In crayfish, proprioceptive afferents are presynaptically inhibited by a typeof SIN (red), the PADIs (not yet identified), that exerts a phasic presynaptic modulation on the sensory–motor synapseduring the walking cycle. The organization of the sensory–motor pathway involves monosynaptic excitatory connectionsthat support negative feedback (levation of the leg activates depressor motoneurons), whereas the reversal of the reflex(levation of the leg activates levator motoneurons) involves disynaptic pathways via non-spiking interneurons (NSIs,orange) named ARINs (assistance-reflex interneurons). In addition, spiking interneurons (yellow SINs) named ARCINs(assistance-reflex controlling interneurons) and activated by velocity-coding proprioceptive neurons, inhibit the NSIs of thepositive feedback circuit and thus prevent the positive feedback reflex from becoming unstable. (d) In the cat, Ia afferentsfrom muscle spindles of the extensor muscle make monosynaptic excitatory contacts on the extensor motoneurons.However, polysynaptic pathways are involved during stepping in the regulation of stance to swing phases. Feedback fromspindle- and tendon-organ afferents of extensors facilitates extensor activities via at least three pathways. Some are mono-synaptic, others are disynaptic, and yet others are polysynaptic via the extensor half-center. (e) The complex relationshipsthat exist between thoraco–coxal muscle-receptor organ (composed of a dynamic fiber, T, and a static fiber, S, with differ-ent coding properties), and motoneurons (Pro, promotor; Rem, remotor) that control the first leg joint during backwardmovements in crayfish). Key: filled circles, inhibitory connections; open triangles, excitatory connections; resistor symbols,electrical connections; diode symbols, rectifying electrical synapse; broken lines, assumed pathways not yet identified.(a) adapted, with permission, from Ref. 23, (b) adapted, with permission, from Ref. 20, (d) adapted, with permission,from Ref. 5, (e) adapted, with permission, from Ref. 17.

In crayfish, it has been shown that active membraneproperties of motoneurons are involved in the reversalof the reflex. Indeed, the reflex reversal is not a simplesign inversion of the sensory–motor pathways. Threemain changes are observed: (1) the firing frequencywithin motor bursts is increased substantially; (2) therelationship between motoneuron bursts and imposedmovements is less precise; and (3) some previously activemotoneurons become silent, whereas previously silentmotoneurons can become active. In the preceding para-graphs, we have shown that both presynaptic inhibitionof primary afferent and changes in the activation levelof polysynaptic pathways are involved in reflex reversal.In addition, a large part of the observed changes in motoroutput is due to changes in motoneuron properties.

When the locomotor generator is activated, the moto-neurons themselves have a role in the suppression ofthe resistance reflex and in the increased activity ofmotoneurons in the assistance response. At the sametime that presynaptic inhibition blocks the negative-feedback pathways, other branches or other proprio-ceptive afferents activate polysynaptic positive feedbackpathways via interneurons. Moreover, some of the moto-neurons involved in this positive feedback now expressactive plateau properties, which can likewise be inducedby agonists of muscarinic ACh receptors54 (compareFig. 5a with Fig. 5c). Consequently, the intensity of dis-charge of depressor motoneurons is much higher thanit is during resistance reflex responses elicited at rest(compare Fig. 5b with Fig. 5d). Direct reciprocal inhibi-tion between antagonistic motoneurons60,61 shouldthen result in only the more-depolarized group beingactive, whereas the antagonistic group is inhibited(Fig. 5e). Since the polysynaptic pathways involved inthe assistance-reflex response produce larger excitatorypostsynaptic potentials in motoneurons than do themonosynaptic resistance reflex pathways (partly becauseof presynaptic inhibition in the monosynaptic resist-ance reflex pathway), only motoneurons involved inthe assistance-reflex response will be allowed to fire.They will, thus, massively inhibit the antagonisticmotoneurons. In this situation, the active propertiesin motoneurons are partly responsible for the massiveblocking of the resistance-reflex responses.

In addition to possessing active membrane proper-ties, motoneurons in most vertebrate and invertebratesystems (except insects) share another characteristic ofneurons that are involved in central pattern genera-tion: the ability to influence other neurons of the cen-tral network through output synapses. The recurrentinhibition mediated by the Renshaw cell constitutes awell-known example of an output synapse from a moto-neuron onto central neurons in mammals (Fig. 6a). Itis now known that the Renshaw cells inhibit Ia inter-neurons, thereby allowing the motoneurons to exertcontrol over the segmental sensory–motor pathways ofthe reciprocal inhibition circuit62. However, Renshawcells are also controlled by descending excitatory andinhibitory inputs that could, therefore, adjust the ex-citability of all motoneurons around a joint. The most-spectacular example of a motoneuron controlling itspresynaptic neuron terminals was demonstrated in thestomatogastric system of the crab63 (Fig. 6b). A moto-neuron named LG (lateral gastric) makes an inhibitorysynapse and an electrical synapse onto a presynapticterminal (SNAX) of an interneuron (modulatory com-missural neuron 1: MCN1). When MCN1 fires tonically,

it induces rhythmic activity in postsynaptic neurons.However, the terminal branches of MCN1 behave verydifferently from the remainder of the neuron. Thesynapses between LG and MCN1, together with themembrane properties of the LG neuron and its recip-rocal inhibition with a particular interneuron, trans-forms the tonic firing of MCN1 into a bursting activityof the terminal that then entrains the other elementsof the network. It is interesting to note that in thiscase, a part of a neuron (the SNAX terminal) is used bya postsynaptic motoneuron to generate a rhythmicactivity in the gastric network. Such functioning prin-ciples would be difficult to decipher in more-complex

trends in Neurosciences

Motornerve

Motoneuron

Motoneuron

Intracellularcurrentinjection

CBCO nervestimulation

10 mV

DepMN

ARINProprioceptive afferent

Presynapticinhibition Lv

MN

200 ms

Control

(a)

(b)

(c)

(d)

(e)

20 mV

1s

Oxotremorine 10–5M

Small reflex response

Passive properties

Large reflex response

Plateau properties

Fig. 5. Changes in the activity of motoneurons induced by muscarinic ACh receptor agonists.(a) and (b) In the absence of muscarinic-receptor activation, motoneurons display passiveelectrical properties: the motoneuron depolarizes in response to a pulse of depolarizing current,and repolarizes as soon as the current injection ceases (a). In this condition the motoneurondisplays small monosynaptic responses to electrical stimulation of the chordotonal organ (CBCO)nerve (b). (c) and (d) When perfused with muscarinic-receptor agonist (oxotremorine 1025

M),plateau properties are induced in motoneurons (c): a pulse of depolarizing current elicits adepolarization of the motoneuron that persists after the current pulse (plateau potential); theplateau can be stopped by injection of a pulse of hyperpolarizing current. In this condition, thereflex responses to CBCO nerve stimulation are much larger, owing to plateau properties (d).(e) During walking, motoneuron activities are enhanced by the activation of muscarinic AChreceptors. Hence, when stimulated (open triangles) in assistance mode by the specialized inter-neuron (ARIN), their response is enhanced by the presence of a plateau. In addition, owing toreciprocal inhibitory connections (filled circles) between antagonistic motoneurons, the activationof the agonistic motoneurons (here, the levator motoneurons, Lv MN) in assistance mode largelyinhibits the antagonistic motoneuron (here, the depressor motoneuron, Dep MN). This recip-rocal inhibitory control, together with the presynaptic inhibition of the monosynaptic afferentsignals and active properties in motoneurons contributes to perform a powerful block of theresistance reflex pathway.

TINS Vol. 23, No. 5, 2000 205

F. Clarac et al. – A ‘simple’ stretch reflex RE V I E W

F. Clarac et al. – A ‘simple’ stretch reflexRE V I E W

206 TINS Vol. 23, No. 5, 2000

systems because they consider the neuron not as awhole, but consisting of separate compartments thatmight not be accessible to recording techniques.

If the motoneurons involved in locomotion in cray-fish are considered, then, similarly, they can exert adirect inhibitory control over their proprioceptive affer-ents64 (Fig. 6c). Such a mechanism has not yet beendescribed in other sensory–motor systems. In intracellu-lar recordings from CBCO terminals, two types of PADsare observed during fictive locomotion. In addition tothe large GABA-mediated PADs described above, small-amplitude slowly developing PADs (sdPADs), which canalso be produced by antidromic motoneuron stimu-lation, are also observed. These sdPADs persist in thepresence of picrotoxin, and therefore do not involvethe classical Cl2 channel associated with GABAA recep-tors. In contrast, sdPADs are produced by a glutamatereceptor that activates a mixed K1 and Na1 conductancewith a reversal potential of 255 mV. The decrease inmembrane input resistance during the activation ofthis glutamate receptor indicates that sdPADs exert apresynaptic inhibition on the proprioceptive messageby a purely shunting mechanism. In contrast to the gaincontrol mechanism studied in the assistance reflex,

the sdPAD gain control is based onthe postsynaptic motoneuron ac-tivity, and is activated only whenmotoneurons are very active.

With the exception of those ininsects, motoneurons make centralelectrical connections with othermotoneurons in most locomotorsystems, as was recently demon-strated in the frog65. In mammals,widespread electrotonic couplingis transient during development.However, it persists in retina, infe-rior olive, hippocampus, striatum66,neocortex67 and postnatally insome spinal motoneurons. In cray-fish, electrotonic connections are awidespread feature and have beenstudied extensively68. As for CBCOfibers, such connections weredemonstrated between motoneuronsin crayfish, using both anatomicaland electrophysiological tech-niques. Depolarizing current in-jected into one motoneuron canactivate up to four motoneurons ofthe same functional group. Simi-larly, injection of Lucifer Yellowinto one depressor motoneuron re-veals a group of four to five stainedmotoneurons. However, as is thecase for CBCO fibers, these electri-cal connections are heterogeneousamong the depressor motoneuronpopulation, the efficacy of the cou-pling (generally weak and alwaysless than 10%) and the number ofmotoneurons connected (betweennone and five) varying betweenmotoneurons. Consequently, elec-trical coupling seems to definesub-groups of motoneurons withsynchronized activity. In this way,

a single input from a CBCO afferent would tend topropagate to the motoneurons of the same sub-group.This is likely to be the case during activation of thecentral-pattern generator.

For reasons of efficacy and adaptability, each of thecontrol mechanisms we have described so far needs tobe adapted to the behavioral requirement state of thewhole animal, which can change with time, owing togrowth or seasonal rhythms. Two slowly modulatingmechanisms will be considered that allow a regulationof sensory–motor circuit performance over long periodsof time.

Long-term modifications of the proprioceptivefeedback system

Neural networks behave as highly nonlinear systems,and the integration of proprioceptive feedback intocentral processing is therefore a dynamic task. In suchsystems, control parameters need to be fitted in order toadapt neural-network machinery to behavioral require-ments. These changes operate over a much longer time-scale (hours) than the simple online control mecha-nisms described above, which generally operate overmillisecond timescales.

trends in Neurosciences

(d)

(a)

(c)

(b)

EP

SP

mea

n am

plitu

de(%

of c

ontr

ol)

Time (min)

100

0

150

200

250

0 50 100 150

MCN1SNAX

LG

Int1

DG

Ia afferent

Ia afferent

RCIaIN

IaIN

MN

MN

CBCO terminal 1

CBCO terminal 2

MN

MN

100 ms

Induction

30 mV

CBCOterminal

MN

10 ms30 mV

10 ms1 mV

Fig. 6. Retrocontrol of afferent message by postsynaptic motoneurons. (a) In mammals, Renshaw cells (RC) exert aninhibitory control on Ia interneurons of the segmental polysynaptic sensory–motor pathways responsible for reciprocalinhibition). Two antagonistic motoneurons, each with its Ia afferent, are represented. Reciprocal inhibition is achieved viaIa interneurons (for clarity only one of the pathways has been represented). The Renshaw cell, activated by a motoneuron,is responsible for the recurrent inhibition of this same motoneuron. In addition, it inhibits the Ia inhibitory interneuronthat synapses on the antagonistic motoneuron. (b) In the stomatogastric system of the crab, MCN1 (modulatory com-missural neuron 1) activates the LG (lateral gastric neuron from the gastric network) and the DG (dorsal gastric neuron)via slow chemical excitatory synapses, and interneuron 1 (Int 1) via a fast chemical excitatory synapse. In turn, the LGinhibits the stomatogastric nerve axon (SNAX) terminals of MCN1, to which it is also electrically coupled. (c) In the cray-fish walking network, motoneurons exert a dual control on their chordotonal organ (CBCO) sensory afferents. They elicitshunting presynaptic inhibition of the sensory terminal by activating a glutamate-receptor channel (filled circle), andLTP in the same sensory afferent via a metabotropic glutamate receptor (parallel T-bars). (d) Long-term potentiation ofthe CBCO–motoneuron synapse induced by activation of the postsynaptic motoneuron (gray bar). In this experiment,intracellular recordings of a CBCO sensory terminal and a postsynaptic motoneuron were performed simultaneously. Thetimecourse of LTP is expressed as relative excitatory postsynaptic potential mean amplitude [each point represents anaverage over 5 min (6SEM)]. After the activation of the motoneuron [injection of 10 Hz depolarizing pulses eliciting twospikes each, for 10 min (see top inset)], the excitatory postsynaptic potential amplitude was increased dramatically. Right-hand insets display paired intracellular recordings of the CBCO terminal and the postsynaptic motoneuron before andafter induction of LTP, respectively. Each inset shows eight superimposed traces. Data on the graph are from a singletrial. (a) adapted, with permission, from Ref. 62, (b) adapted, with permission, from Ref. 63.

TINS Vol. 23, No. 5, 2000 207

In rats, monkeys and humans, the stretch reflex canbe operantly conditioned69–71. The site and nature ofthis plasticity lies in the spinal cord itself9. In contrastto LTD, which seems to result directly from changes inthe motoneuron properties, LTP probably results froma decrease in the disynaptic inhibition that involvesgroup-I afferent fibers9.

In invertebrates, LTP also exists in motor systems, asdemonstrated in motoneurons involved in jumping inthe locust72. More recently, in the crayfish, LTP hasbeen demonstrated in CBCO afferents–motoneuronsynapses10. In contrast to the plastic changes reportedabove, this LTP is intrinsic to the two-neuron (sensory-afferent–motoneuron) connection and is initiated bythe postsynaptic motoneuron activity (Fig. 6c). TheCBCO–motoneuron synapse generally remains stablefor hours in the absence of postsynaptic activity.However, the intracellular stimulation of the postsy-naptic motoneuron (10 Hz, 10 min) generally inducesa significant (up to 300%) and long-lasting (a fewhours; Fig. 6d) increase in the amplitude of the mono-synaptic excitatory postsynaptic potential elicited bythe CBCO sensory spike. The mechanisms that under-lie this LTP have recently been analyzed10: it is medi-ated by glutamate, the motoneuron neurotransmitterin arthropods. A quantal analysis of unitary excitatorypostsynaptic potentials before and after induction ofLTP demonstrated that the increase in excitatory post-synaptic potential amplitude resulted from a large in-crease in the probability of release from the presynapticneuron, without changes either in the quantal size orthe total number of quanta. This result, which is asso-ciated with the absence of any long-term effects ofglutamate upon the motoneuron, strongly suggests thesynaptic changes that lead to LTP are purely presynaptic.Moreover, the results of pharmacological studies indi-cate that a glutamate metabotropic receptor located onthe presynaptic sensory neuron terminal is likely to beinvolved. Note that in this system, motoneurons havetwo effects on their presynaptic sensory neurons: theyinhibit them presynaptically and induce LTP in them.However, those two phenomena support different func-tions and would occur in different states of the net-work. Retrograde glutamatergic presynaptic inhibitionrequires high level of motor activity, and is thereforea ‘protective’ mechanism that limits the activity of thepostsynaptic motoneuron. By contrast, LTP seems tobe more of an ‘arousal’ process that is observed only invery quiet motor systems, where, via positive feedback,it permits reinforcement of the input synapses control-ling motor activity. When a motoneuron is recruited,its sensory pathways are also reinforced, owing to thisLTP mechanism. This finding reinforces the idea thatsensory–motor units, rather than motoneurons, repre-sent the real basic components of motor commands.

Concluding remarks

The data presented in this article demonstrate thatsensory–motor connections are much more complexthan was initially thought. Most of the levels of con-trol reported here are fast regulatory systems (involvingpresynaptic inhibition mediated by GABA, histamineand glutamate, as well as amine-mediated neuromodu-lation), but long-term changes (involving retrogradeglutamate-mediated motoneuron control) are also pres-ent. In the crayfish model, most of the adjusting levelsare intrinsic to the locomotor network itself, and are

therefore activity dependent. It is striking that similarmechanisms are used in different animal models. Forexample, PADs involve a GABA-induced activation ofCl2 channels, which could signify that presynaptic in-hibition of primary afferents is a fundamental featurethat appeared very early in evolution. However, sub-stantial differences also exist. For example, in primaryafferents of insects, the equilibrium potential for Cl2 isclose to the resting membrane potential, whereas it ismuch more depolarized in vertebrate and crustaceanprimary afferents. The role of non-spiking interneuronsconstitutes another substantial difference between ver-tebrates and invertebrates. Non-spiking interneuronsare very suitable premotor elements because they canexert a graded control over the motoneurons, andachieve independent local processing in their differentbranches2. Are such elements really absent in the verte-brate spinal cord? Or are they simply not accessible tointracellular recordings? More generally, a particularlyinteresting question concerns the compartmentaliz-ation of neuron processing. For example, it seems thatthe different branches of vertebrate primary afferentsare differentially affected by presynaptic inhibition73.Future investigations should address these issues. Inaddition, there are still many unanswered questions thatconcern the control of the sensory–motor pathways bydescending interneurons, and, more generally, howsensory–motor pathways operate in a real behavioralcontext. The answer to this question will be an impor-tant challenge in the future, as we know that, in verte-brates and in mammals in particular, cephalization hasresulted in increased control of local circuits by superiorstructures, and thus probably masks such intrinsicadjustments.

Selected references1 Lloyd, D.P.C. (1943) Conduction and synaptic transmission of

reflex arcs. J. Neurophysiol. 6, 317–3262 Burrows, M. (1992) Local circuits for the control of leg movements

in an insect. Trends Neurosci. 15, 226–2323 Blight, A.R. and Llinás, R. (1980) The non-impulsive stretch-receptor

complex of the crab: a study of depolarization-release coupling ata tonic sensorimotor synapse. Philos. Trans. R. Soc. London Ser. B290, 219–276

4 El Manira, A. et al. (1991) Monosynaptic connections mediateresistance reflex in crayfish (Procambarus clarkii) walking legs. J. Comp. Physiol. 168, 337–349

5 Pearson, K.G. and Ramirez, J-M. (1997) Sensory modulation ofpattern-generating circuits. In Neurons, Networks and Motor Behavior(Stein, P.S.G. et al., eds), pp. 225–235, MIT Press

6 Wilson, D.M. (1961) The central nervous control of flight in locust.J. Exp. Biol. 38, 471–490

7 Stein, P.S.G. et al. (1997) Neurons, Networks and Behavior, MIT Press8 Pearson, K.G. (1995) Reflex reversal in the walking systems of

mammals and arthropods. In Neural Control of Movement. (Ferrel,W.R. and Proske, U., eds), Plenum Press

9 Wolpaw, J.R. (1997) The complex structure of a simple memory.Trends Neurosci. 20, 588–594

10 Le Ray, D. and Cattaert, D. (1999) Active motor neurons potentiatetheir own sensory inputs via glutamate-induced long-termpotentiation. J. Neurosci. 19, 1473–1483

11 Hulliger, M. (1984) The mammalian muscle receptors. Rev. Physiol.Biochem. Pharmacol. 101, 1–110

12 Skorupski, P. and Bush, B.M.H. (1992) Parallel reflex and centralcontrol of promotor and receptor motoneurons in crayfish. Proc.R. Soc. London B Biol. Sci. 249, 7–12

13 Eccles, J.C. et al. (1957) The convergence of monosynaptic excitatoryafferents on to many different species of alpha motoneurones. J. Physiol. 22–50

14 Le Ray, D. and Cattaert, D. (1997) Neural mechanisms of reflexreversal in coxo–basipodite depressor motor neurons of the crayfish.J. Neurophysiol. 77, 1963–1978

15 Scott, J.G. and Mendell, L.M. (1976) Individual EPSPs producedby single triceps surae Ia afferent fibers in homonymous andheteronymous motoneurons. J. Neurophysiol. 39, 679–692

F. Clarac et al. – A ‘simple’ stretch reflex RE V I E W

F. Clarac et al. – A ‘simple’ stretch reflexRE V I E W

208 TINS Vol. 23, No. 5, 2000

16 Le Ray, D. et al. (1997) Functional analysis of the sensory motorpathway of resistance reflex in crayfish. I. Multisensory codingand motor neuron monosynaptic responses. J. Neurophysiol. 78,3133–3143

17 Skorupski, P. (1992) Synaptic connections between nonspikingafferent neurons and motor neurons underlying phase-dependentreflexes in crayfish. J. Neurophysiol. 67, 664–679

18 Shapovalov, A.I. and Shiriaev, B.I. (1978) Electrical couplingbetween primary afferents and amphibian motoneurons. Exp.Brain Res. 33, 299–312

19 Burrows, M. (1987) Parallel processing of proprioceptive signalsby spiking local interneurons and motor neurons in the locust. J. Neurosci. 7, 1064–1080

20 Bässler, U. and Büschges, A. (1998) Pattern generation for stickinsect walking movements. Multisensory control of a locomotorprogram. Brain Res. Rev. 27, 65–68

21 Wolf, H. and Büschges, A. (1995) Nonspiking local interneuronsin insect leg motor control. II. Role of nonspiking local interneuronsin the control of the leg swing during walking. J. Neurophysiol. 73,1861–1875

22 Kittmann, R. et al. (1996) Premotor interneurons in generation of adaptive leg reflexes and voluntary movements in stick insects.J. Neurobiol. 31, 512–532

23 Burrows, M. et al. (1988) Proprioceptive inputs to nonspiking localinterneurons contribute to local reflexes of a locust hindleg. J. Neurosci. 8, 3085–3093

24 DiCaprio, R.A. and Clarac, F. (1981) Reversal of a walking leg reflexelicited by a muscle receptor. J. Exp. Biol. 90, 197–203

25 Duysens, J. et al. (1990) Gating and reversal of reflexes in anklemuscles during human walking. Exp. Brain Res. 82, 351–358

26 Rudomin, P. et al. (1998) Presynaptic Inhibition and Neural Control,Oxford University Press

27 Cattaert, D. et al. (1992) Direct evidence for presynaptic inhibitorymechanisms in crayfish sensory afferents. J. Neurophysiol. 67,610–624

28 Büschges, A. and Wolf, H. (1995) Nonspiking local interneuronsin insect leg motor control I. Common layout and species-specificresponse properties of femur–tibia joint control pathways in stickinsect and locust. J. Neurophysiol. 73, 1843–1859

29 Frank, K. and Fuortes, M.G.F. (1957) Presynaptic and postsynapticinhibition of monosynaptic reflexes. Fed. Proc. Fed. Am. Soc. Exp.Biol. 16, 39–40

30 El Manira, A. et al. (1993) Electrical coupling of mechanoreceptorafferents in the crayfish: a possible mechanism for enhancementof sensory signal transmission. J. Neurophysiol. 69, 2248–2251

31 Pasztor, V.M. and Bush, B.M.H. (1987) Peripheral modulation ofmechanosensitivity in primary afferent neurons. Nature 326, 793–795

32 Rossi-Durand, C. (1993) Peripheral proprioceptive modulation incrayfish walking leg by serotonin. Brain Res. 632, 1–15

33 Gossard, J-P. et al. (1991) An intracellular study of muscle primaryafferents during fictive locomotion in the cat. J. Neurophysiol. 65,914–926

34 Burrows, M. and Laurent, G. (1993) Synaptic potentials in thecentral terminals of locust proprioceptive afferents generated byother afferents from the same sense organ. J. Neurosci. 13, 808–819

35 Rudomin, P. (1990) Presynaptic inhibition of muscle spindle andtendon organ afferents in the mammalian spinal cord. TrendsNeurosci. 13, 499–505

36 Burrows, M. and Matheson, T. (1994) A presynaptic gain controlmechanism among sensory neurons of a locust leg proprioceptor.J. Neurosci. 14, 272–282

37 Cattaert, D. and El Manira, A. (1999) Shunting versus inactivation:analysis of presynaptic inhibitory mechanisms in primary afferentsof the crayfish. J. Neurosci. 19, 6079–6089

38 El Manira, A. and Clarac, F. (1991) GABA-mediated presynapticinhibition in crayfish primary afferents by non-A, non-B GABAreceptors. Eur. J. Neurosci. 3, 1208–1218

39 Gallagher, J.P. et al. (1978) Characterization and ionic basis ofGABA-induced depolarizations recorded in vitro from cat primaryafferent neurones. J. Physiol. 275, 263–282

40 Lamotte d’Incamps, I. et al. (1998) Reduction of presynaptic actionpotentials by PAD: model and experimental study. J. Comput.Neurosci. 5, 141–156

41 Cattaert, D. et al. (1994) Chloride conductance produces bothpresynaptic inhibition and antidromic spikes in primary afferents.Brain Res. 666, 109–112

42 Bévengut, M. et al. (1997) Antidromic modulation of a proprioceptorsensory discharge in crayfish. J. Neurophysiol. 78, 1180–1183

43 Marder, E. (1998) Electrical synapses: beyond speed and synchronyto computation. Curr. Biol. 8, R795–R797

44 Edwards, D.H. et al. (1998) Neuronal coincidence detection byvoltage-sensitive electrical synapses. Proc. Natl. Acad. Sci. U. S. A.95, 7145–7150

45 Pearlstein, E. et al. (1998) Neuromodulation of reciprocalglutamatergic inhibition between antagonistic motoneurons by5-hydroxytryptamine (5-HT) in crayfish walking system. Neurosci.Lett. 241, 37–40

46 Büschges, A. et al. (1993) Octopamine effects mimick state-dependent changes in a proprioceptive feedback system. J.Neurobiol. 24, 598–610

47 Matheson, T. (1997) Octopamine modulates the responses andpresynaptic inhibition of proprioceptive sensory neurones in thelocust Schistocerca gregaria. J. Exp. Biol. 200, 1317–1325

48 El Manira, A. et al. (1997) Locomotor-related presynaptic modulationof primary afferents in the lamprey. Eur. J. Neurosci. 9, 696–705

49 Bässler, U. (1993) The femur-tibia control system of stick insect –a model system for the study of the neural basis of joint control.Brain Res. Rev. 18, 207–226

50 Pearson, K.G. (1995) Proprioceptive regulation of locomotion.Curr. Opin. Neurobiol. 5, 786–791

51 Hancox, J.C. and Pitman, R.M. (1993) Plateau potentials in aninsect motoneurone can be driven by synaptic stimulation. J. Exp.Biol. 176, 307–310

52 Kiehn, O. (1991) Plateau potentials and active integration in the‘final common pathway’ for motor behaviour. Trends Neurosci. 14,68–73

53 Hounsgaard, J. and Kiehn, O. (1989) Serotonin-induced bistabilityof turtle motoneurones caused by a nifedipine-sensitive calciumplateau potential. J. Physiol. 414, 265–282

54 Cattaert, D. et al. (1994) Nicotinic and muscarinic activation ofmotoneurons in the crayfish locomotor network. J. Neurophysiol.72, 1622–1633

55 Wallen, P. and Grillner, S. (1987) N-methyl-D-aspartate receptor-induced, inherent oscillatory activity in neurons active duringfictive locomotion in the lamprey. J. Neurosci. 7, 2745–2755

56 Sillar, K.T. and Simmers, A.J. (1994) Electrical coupling and intrinsicneuronal oscillations in Rana temporaria spinal cord. Eur. J. Morphol.32, 293–298

57 Hochman, S. et al. (1994) TTX-resistant NMDA receptor-mediatedvoltage oscillations in mammalian lumbar motoneurons. J. Neurophysiol. 72, 2559–2562

58 Guertin, P.A. and Hounsgaard, J. (1998) NMDA-induced intrinsicvoltage oscillations depend on L-type calcium channels in spinalmotoneurons of adult turtles. J. Neurophysiol. 80, 3380–3382

59 Selverston, A.I. et al. (1997) Shared features of invertebrate centralpattern generators. In Neurons, Networks and Motor Behavior (Stein,P.S.G. et al., eds), pp. 105–117, MIT Press

60 Pearlstein, E. et al. (1994) Inhibitory effects of L-glutamate oncentral processes of crustacean leg motoneurons. Eur. J. Neurosci.6, 1445–1452

61 Pearlstein, E. et al. (1998) Inhibitory connections betweenantagonistic motor neurones of the crayfish walking legs. J. Comp.Neurol. 399, 241–254

62 Katz, R. and Pierrot-Deseilligny, E. (1998) Recurrent inhibition inhumans. Prog. Neurobiol. 57, 325–355

63 Coleman, M.J. et al. (1995) A switch between two modes of synaptictransmission mediated by presynaptic inhibition. Nature 378,502–505

64 Cattaert, D. and Le Ray, D. (1998) Direct glutamate-mediatedpresynaptic inhibition of sensory afferents by the postsynapticmotor neurons. Eur. J. Neurosci. 10, 3737–3746

65 Perrins, R. and Roberts, A. (1995) Cholinergic contribution toexcitation in a spinal locomotor central pattern generator in Xenopusembryos. J. Neurophysiol. 73, 1013–1019

66 O’Donnell, P. and Grace, A.A. (1997) Cortical afferents modulatestriatal gap junction permeability via nitric oxide. Neuroscience76, 1–5

67 Nadarajah, B. et al. (1996) Gap junctions in the adult cerebral cortex:regional differences in their distribution and cellular expressionof connexins. J. Comp. Neurol. 376, 326–342

68 Chrachri, A. and Clarac, F. (1989) Synaptic connections betweenmotor neurons and interneurons in the fourth thoracic ganglionof the crayfish, Procambarus clarkii. J. Neurophysiol. 62, 1237–1250

69 Evatt, M.L. et al. (1989) Modification of human spinal stretchreflexes: preliminary studies. Neurosci. Lett. 105, 350–355

70 Chen, X.Y. and Wolpaw, J.R. (1995) Operant conditioning of H-reflex in freely moving rats. J. Neurophysiol. 73, 411–415

71 Wolf, S.L. et al. (1995) Contralateral and long latency effects ofhuman biceps brachii stretch reflex conditioning. Exp. Brain Res.107, 96–102

72 Parker, D. (1995) Long-lasting potentiation of a direct centralconnection between identified motor neurons in the locust. Eur.J. Neurosci. 7, 1097–1106

73 Lomeli, J. et al. (1998) Local control of information flow insegmental and ascending collaterals of single afferents. Nature 395,600–604

Acknowledgments This work was

supported by theCentre National de

la RechercheScientifique (CNRS).

The authors thankJ. Simmers and

P. Dickinson forproviding valuable

comments on themanuscript.