J Neurophysiol 90: 531–538, 2003;
10.1152/jn.00338.2003.
review
Intersegmental Coordination of Rhythmic Motor Patterns
Andrew A.V. Hill, Mark A. Masino, and Ronald L. Calabrese
Biology Department, Emory University, Atlanta, Georgia 30322
Submitted 7 April 2003; accepted in final form 22 April 2003
INTRODUCTION
The study of the neural basis of motor patterns that underlie
wavelike behaviors such as undulatory swimming has contributed to our general understanding of coordination in the nervous system. Numerous studies have shown that the isolated
nervous system is capable of producing rhythmic motor output
in the absence of sensory feedback (Marder and Calabrese
1996). In addition, in many cases these rhythmic motor patterns are similar to those required for specific behaviors in the
intact animal. For example, in the lamprey forward swimming
is accomplished by means of side-to-side undulations that
travel from anterior to posterior along the length of the body
(Wallén and Williams 1984). In the lamprey, as in many other
animals, wavelike motor patterns arise from a network of
neurons that is distributed longitudinally along the neural axis.
This type of neural network consists of a chain of coupled
segmental oscillators, local neural networks that are capable of
independently generating rhythmic output (Skinner and Mulloney 1998a). The appropriate phase relationships between
these segmental oscillators arise as an emergent property of the
segmental oscillators and the coupling between them. Although this distributed organization is found in many different
animals, there are large differences in terms of the properties of
the segmental oscillators, the strength and symmetry of coupling, and the importance of sensory feedback. In this review,
we will discuss some of the variations in design found in three
animals: the lamprey, leech, and crayfish.
PHASE CONSTANCY IS NECESSARY FOR
UNDULATORY SWIMMING
The leech and the lamprey both swim in an undulatory
fashion with the body forming approximately one full wave at
any given time during swimming. To maintain this mode of
swimming as an animal changes its swim cycle period, the
phase between the muscle contractions in different segments
must remain constant. In the leech, there are 18 body segments
that are actively used for swimming. Thus the phase lag between consecutive body segments is about 20° in the swimming animal (Kristan et al. 1974). Intersegmental phase lags
that are nearly independent of cycle period are also observed in
isolated chains of the leech nerve cord consisting of as few as
two ganglia (Pearce and Friesen 1985). However, in contrast to
the intact animal, the phase lag per segment within a long chain
of ganglia in vitro is only about 8°.
In the leech swim network, the segmental oscillators are not
Address for reprint requests: R. L. Calabrese (E-mail: RCalabre@
Biology.EMORY.EDU).
www.jn.org
uniform in their ability to generate rhythmic output. Isolated
ganglia from the anterior end to the midpoint of the ventral
nerve cord are capable of generating swim-like oscillations, but
not individual posterior ganglia (Hocker et al. 2000; cf. Tunstall et al. 2002). In addition, there is a U-shaped gradient of
cycle period. Isolated mid-cord ganglia produce oscillations
with a shorter cycle period than either individual anterior
ganglia or short chains of posterior ganglia.
A segmental oscillator of the leech swim network consists of
motor neurons and oscillator interneurons within a single ganglion that are organized into two bilaterally symmetric hemisegmental circuits. Oscillations originate within a single ganglion from the local circuit formed by these oscillator interneurons, which are connected almost exclusively by inhibitory
synapses (Friesen and Pearce 1993). The hemisegments within
a single ganglion oscillate synchronously based on strong
electrical and chemical coupling across the midline (Friesen
and Hocker 2001). This activity pattern is appropriate since the
leech swims by dorsoventral undulations, which require synchronous activation of muscles on the left and right sides of the
body. Because the two hemisegments are equivalent, in studying intersegmental coordination it is justified to consider only
the circuitry on one side of the animal. The segmental oscillators are coupled by ascending and descending projections of
the oscillator interneurons, which synapse directly with oscillator interneurons in other ganglia (Fig. 1).
In principle, the generation of phase lags appropriate for
forward swimming can be explained by a system consisting of
a chain of coupled oscillators in which the anterior oscillators
have shorter inherent periods than the posterior oscillators
(Ikeda and Wiersma 1964; Matsushima and Grillner 1992).
When a system such as this is coupled, all segmental oscillators
of the network share the same period, but the faster oscillators
will lead in phase. In the case of the leech swimming, this
explanation is not sufficient to explain forward swimming
because the segmental oscillators with the shortest cycle period
lie at the mid point of the ventral nerve cord. Experimental and
modeling work has provided evidence that forward swimming
most likely arises from asymmetries in the coupling between
the segmental oscillators (Cang and Friesen 2002).
Intersegmental coupling, which spans six segments in both
the ascending and the descending directions, is approximately
equal in functional strength in either direction (Friesen and
Hocker 2001; Pierce and Friesen 1985). However, at the level
of specific interconnections there are many asymmetries. The
oscillator interneurons are active at three different phases sepThe costs of publication of this article were defrayed in part by the payment
of page charges. The article must therefore be hereby marked ‘‘advertisement’’
in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
0022-3077/03 $5.00 Copyright © 2003 The American Physiological Society
531
532
A.A.V. HILL, M. A. MASINO, AND R. L. CALABRESE
FIG. 1. Intersegmental coupling between two hemisegmental oscillators of
the leech swim network. Intersegmental synaptic connections between oscillator interneurons are shown (intrasegmental connections have been omitted
for simplicity). The interneurons in each ganglion are arranged in rows
according to the phase of their activity. Each of the interneurons shown occurs
as a bilateral pair except for the unpaired interneuron 208. The filled circles
represent inhibitory synapses. The T symbols represent excitatory synapses.
Question marks indicate that certain targets are unknown. Figure adapted from
Cang and Friesen (2002).
arated by 120° (Fig. 1). The interneurons in the 0° phase group
all project their axons in the descending direction, whereas the
interneurons of the 120° and 240° groups only project in the
ascending direction (Friesen 1989). In addition, the sole excitatory interneuron projects in the descending direction, whereas
inhibitory interneurons project in both directions (Brodfuehrer
et al. 1995). Finally, the synaptic targets are asymmetric;
ascending and descending projection neurons project to targets
with different activity phases (Friesen 1989). A computer
model that incorporates these asymmetries but not the inherent
period differences produced a 8 –10° phase lag similar to that
seen in long chains of isolated nerve cords as well as phase
constancy with changing cycle period (Cang and Friesen
2002).
Although the model replicated the behavior of the isolated
nerve cord, the 8° phase lag between segments is clearly too
small to produce a traveling wave of one wavelength along the
body. To test the idea that sensory feedback from stretch
receptors embedded in the body wall and mechanical coupling
between muscles in neighboring segments may contribute to
the phase lag observed in the intact animal, these mechanisms
were added to the computer model (Cang and Friesen 2000; Yu
et al. 1999). The resulting computer model produced a phase
lag of 17° per segment, which is similar to the value in the
intact animal (20°), suggesting that sensory feedback and mechanical coupling are indeed important for appropriate coordination (Cang and Friesen 2002; Friesen and Cang 2001).
Experiments in which the nervous system was transected
(Wallén 1982; Yu et al. 1999) have shown that mechanical
coupling and sensory feedback are more important for intersegmental coordination in the leech than in the lamprey, which
is discussed in the following text. Perhaps a high degree of
J Neurophysiol • VOL
peripheral feedback is necessary in an animal such as the leech,
which has a hydrostatic skeleton.
In the lamprey as in the leech, forward swimming results
from a traveling wave of one body length that begins at the
anterior end of the animal. The lamprey has about 100 body
segments; therefore, the phase difference between body segments is about 1% in the intact animal. An identical phase
difference occurs in the isolated spinal cord when rhythmic
activity is induced with the NMDA receptor agonist, D-glutamate (Wallén and Williams 1984).
Although it is not possible to identify individual interneurons as in the leech swim network, two classes of interneurons,
excitatory (E) and inhibitory (C), appear to be essential for
burst generation (Grillner et al. 1995). The E interneurons
synapse ipsilaterally with other E interneurons and with C
interneurons (Fig. 2). The mutual excitation between E interneurons appears to support the generation of bursts in hemisegments (Hellgren-Kotaleski et al. 1999b). The C interneurons
project contralaterally and inhibit other C interneurons as well
as E interneurons (Buchanan 1982). These interneurons are
responsible for producing alternating activity between the right
and left sides of the spinal cord, which is necessary for sideto-side undulatory swimming (Hellgren-Kotaleski et al.
1999a).
In the first model of a segmental oscillator in this system,
burst termination was governed by reciprocal inhibition and the
hemisegments were incapable of independent bursting. This
network generated oscillations with a duty cycle (burst duration divided by cycle period) of 50%, a characteristic property
of a half-center oscillator (Skinner et al. 1994; Wallén et al.
1992). In contrast, in the biological system, the duty cycle is
30 – 40% and when inhibitory synapses are pharmacologically
blocked, the hemisegments burst endogenously (Hagevik and
McClellan 1994; Wallén and Williams 1984).
Recent work has confirmed that reciprocal inhibition is not
FIG. 2. A segmental oscillator of the lamprey swim network. The segmental oscillator consists of 2 bilaterally symmetric hemisegmental circuits. The
interneurons (E, C, and L) generate oscillations. E interneurons excite all
neurons within a hemisegmental oscillator. C interneurons inhibit all neurons
of the contralateral hemisegmental oscillator. L interneurons inhibit ipsilateral
C interneurons but are not critical for ipsilateral burst termination (Hellgren et
al. 1992). The motor neurons (M) receive synaptic input from the E and C
interneurons. The stretch receptors (triangles) excite ipsilateral and inhibit
contralateral neurons. E and L interneurons project ipsilaterally in both directions. C interneurons project in the descending direction and cross the midline.
Open circles indicate excitatory synapses, and black circles indicate inhibitory
synapses. Synapse symbols impinging on the circle around a hemisegmental
oscillator indicate that synapses are made with all of the neurons within the
circle. Figure adapted from Bem et al. (2003).
90 • AUGUST 2003 •
www.jn.org
INTERSEGMENTAL COORDINATION OF RHYTHMIC MOTOR PATTERNS
the primary factor that terminates bursts, but rather the activation of two types of Ca2⫹-dependent K⫹ channels (KCa). One
type, activated by Ca2⫹ that enters during action potentials,
causes postspike afterhyperpolarizations that lead to spike frequency adaptation (Tegnér et al. 1998). The other type, activated by Ca2⫹ that enters through open NMDA channels
during each burst, produces a more gradual hyperpolarization
(Brodin et al. 1991). The incorporation of these mechanisms
into a model of a segmental oscillator results in hemisegments
that can oscillate independently and a duty cycle similar to that
of the biological system (Hellgren-Kotaleski et al. 1999b). The
addition of a mechanism in which the amount of adaptation
(based loosely on KCa activation) increases as the cycle frequency increases resulted in a model that was capable of
oscillating with a constant duty cycle over roughly a 10-fold
frequency range, which would be necessary for the animal to
swim at different speeds (Ullström et al. 1998).
Important early theoretical work (Kopell and Ermentrout
1988) indicated that appropriate phase lags for lamprey swimming may be generated by two mechanisms: asymmetries in
the coupling between segmental oscillators and differences in
inherent segmental periods. Evidence for coupling asymmetries comes from a variety of experiments. Movements imposed on the isolated spinal cord can entrain the activity of the
entire swim network by activating intra-spinal stretch receptors, which provide sensory feedback to the interneurons (Fig.
2) (Grillner et al. 1981). Interestingly, movement applied to the
caudal end of the spinal cord can entrain the system to a greater
range of frequencies than movements applied to the rostral end
(Williams et al. 1990). Additionally, split-bath experiments in
which the rostral and caudal halves of the spinal cord were
bathed in pools with different concentrations of D-glutamate
demonstrated that the rostral spinal cord dominates the frequency of the coupled network (Sigvardt and Williams 1996).
In another study, physiological and computer modeling results
suggested that longitudinal coupling is due primarily to ipsilateral excitatory coupling that is stronger in the descending
direction than in the ascending direction (Hagevik and McClellan 1994). There is also evidence of anatomical coupling
asymmetries. For example, whereas the E interneurons project
symmetrically over a few segments both rostrally and caudally,
the C interneurons project 14 –20 segments caudally but have
only short rostral projections (Buchanan 1982; Dale 1986).
Evidence for, or against, the existence of a gradient of
segmental oscillator frequency is more limited than for asymmetric coupling. Pharmacologically induced frequency gradients can alter and even reverse the normal rostrocaudal phase
lags in the isolated spinal cord (Matsushima and Grillner 1992;
Tegnér et al. 1993). Although such experiments cannot demonstrate that such a gradient exists naturally, they show that at
least in principle a gradient could produce the appropriate
phase lags. Presumably in the intact system a gradient of
oscillator frequencies could be produced by a corresponding
gradient of descending synaptic drive as has been found in the
swim network of Xenopus embryos (Tunstall and Roberts
1994). In contrast to these results, surgically isolated sections
of the spinal cord do not vary in frequency in a systematic way
(Cohen 1987). One problem with this study, however, is that
the rhythm was induced pharmacologically rather than by
normal descending spinal pathways. To address this problem,
a study was conducted in which fictive swimming was induced
J Neurophysiol • VOL
533
by pharmacological microstimulation of the brain. By blocking
local rhythmic activity in either the rostral or the caudal half of
the spinal cord with a low Ca2⫹ saline, it was possible to
measure the activity of independent sections of the spinal cord
(Hagevik and McClellan 1999). This study revealed faster
oscillations in the rostral spinal cord than in the caudal spinal
cord.
The extent of longitudinal coupling between segmental oscillators in the lamprey is doubtless another important factor
for determining phase lags. Functional coupling that determines phase appears to extend over a much more limited range
than the full projection range of 30 –50 segmental reported for
some interneurons (Rovainen 1974). In one study, the lamprey
spinal cord was placed in a chamber with partitions allowing
the rostral, middle, and caudal sections to be bathed in different
solutions (Miller and Sigvardt 2000). Local synaptic activity
was blocked in the middle section with a solution containing
low-Ca2⫹ and high-Mg2⫹ without affecting spike conduction
in axons spanning the middle compartment. The results indicated that although the maximal functional length of propriospinal coupling is 16 –20 segments, phase was controlled by
short-range coupling that spans only 4 – 6 segments (see also
McClellan and Hagevik 1999). It has been proposed that the
interneurons that comprise a segmental oscillator act not only
to generate the local rhythm but also project to, and influence,
rhythm generation in other segments. The feasibility of this
idea, which has been termed “synaptic spread,” has been demonstrated in modeling studies in which the synaptic contacts
made locally by an interneuron are also made with similar
targets in neighboring segments but with lower synaptic
strengths (Buchanan 1992; Ekeberg 1993; Wadden et al. 1997;
Williams 1992). Recent models incorporating cellular and synaptic properties of the swim network can replicate many of the
observed properties of the swimming animal such as a reversal
of phase lag necessary for backward swimming (Ullström et al.
1998).
At first glance, sensory feedback appears to be less important for swimming in the lamprey than in the leech since the
motor output of the isolated spinal cord closely resembles the
pattern of the intact animal (Wallén and Williams 1984). To
assess the contribution of sensory feedback, a model was
created that included the swim network, muscle activation,
body mechanics, counteracting water forces, and sensory feedback (Ekeberg 1993; Ekeberg et al. 1999). In this model,
stretch receptors had very little effect on swimming movements in still water. However, when the virtual lamprey swam
in a cross current, a model incorporating stretch receptors
performed much better than a model lacking them. In the latter,
transversely flowing water forced the head to one side, eventually resulting in a complete change of direction. In the model
with stretch receptors, the sensory feedback counteracted the
perturbing effects of the current and allowed the simulated
lamprey to swim straight. Therefore although sensory feedback
does not contribute strongly to phase lag as in the leech, it is
necessary to counteract environmental perturbations.
PHASE FLEXIBILITY IN THE LEECH HEARTBEAT
NETWORK
The timing network that paces the leech heartbeat differs in
many ways from the two systems discussed so far. In this
90 • AUGUST 2003 •
www.jn.org
534
A.A.V. HILL, M. A. MASINO, AND R. L. CALABRESE
system, the interaction between segmental oscillators is characterized by phase flexibility rather than phase constancy. The
heartbeat network contains two segmental oscillators located in
the 3rd and 4th ganglia of the ventral nerve cord (Fig. 3A)
(Peterson 1983a,b). The output of these two segmental oscillators paces and coordinates the activity of a network of interneurons and motor neurons that control the peristaltic contractions of two lateral heart tubes that run the length of the body
(Calabrese et al. 1995).
The phase between these oscillators in isolated nerve cords
is stable for the duration of an experiment but varies between
preparations from about ⫺12 to ⫹20% (the median phase
difference is 6%; a positive phase lag indicates that the G4
segmental oscillator leads in phase) (Masino and Calabrese
2002a). A combination of physiological and modeling work
has demonstrated that this wide phase range can be accounted
for by the specific network configuration combined with intrinsic period differences between the segmental oscillators.
At the core of each segmental oscillator is a pair of heart
interneurons (HN cells) that form reciprocally inhibitory synapses across the ganglion midline and constitute a half-center
oscillator (Fig. 3A). For example, the heart interneuron on the
left side of the 3rd ganglion [HN(L,3)] is reciprocally inhibitory with its contralateral homologue. Similar to the hemisegments in the lamprey spinal cord, these interneurons are capa-
ble of endogenous oscillations when inhibitory synapses are
blocked (Cymbalyuk et al. 2002). However, physiological and
modeling data show that although the individual oscillator
interneurons are capable of endogenous bursting, the halfcenter configuration makes the oscillations more robust, ensures bilateral alternation of bursting, alters period substantially from the inherent period, and fixes duty cycle near 50%
(Cymbalyuk et al. 2002). The two segmental oscillators are
coupled by coordinating interneurons that have cell bodies in
the first and second ganglia but have spike initiation sites and
form reciprocally inhibitory synapses with ipsilateral oscillator
interneurons in the 3rd and 4th ganglia (Fig. 3A) (Masino and
Calabrese 2002a; Peterson 1983b).
To understand how phase and cycle period are controlled in
this system, alternative models that incorporated different assumptions about the system were created (Hill et al. 2002).
Based on anatomical and physiological data, one possible
configuration of the network is a “symmetric” model in which
the oscillator neurons of both segmental oscillators equally
inhibit the ipsilateral coordinating interneurons (Fig. 3B). This
model assumes that the G3 spike initiation sites of the coordinating interneurons are silent and that all spikes originate at the
G4 sites (Fig. 3A). In this model, the activity of a coordinating
interneuron is completely inhibited when either the G3 or the
G4 oscillator interneuron on the same side of the body is
FIG. 3. The timing network of the leech heartbeat neural network. A: the timing network contains 4 pairs of bilaterally
symmetric interneurons that have cell bodies in the first 4 midbody ganglia. There are 2 segmental oscillators located in the 3rd
and 4th ganglia. The coordinating interneurons of the first 2 ganglia are functionally equivalent and are therefore combined in
representation. Open circles represent cell bodies; open squares represent sites of spike initiation; small filled circles represent
inhibitory synapses. B: the symmetric model. Circles represent oscillator interneurons of the 3rd and 4th ganglia. Squares represent
the active neurites of the coordinating interneurons, which initiate spikes and make reciprocally inhibitory synapses with the
oscillator interneurons. C: simulated activity of heart interneurons in the symmetric model. The model interneurons are labeled HN
and are indexed by body side and midbody ganglion number. The model neurons contain Hodgkin–Huxley-style voltage-dependent
conductances. The maximal conductance of the hyperpolarization-activated current (g៮ h) was set to different values within the 2
pairs of oscillator interneurons (g៮ h equals 4.0 nS in the G3 oscillator interneurons and 5.4 nS in G4 oscillator interneurons). An
increase in g៮ h leads to faster oscillations (Hill et al. 2001). Symbols above each voltage trace indicate the occurrence of the median
spike within each burst. The yellow rectangles show the windows of time in which the coordinating interneurons were active. Figure
adapted from Hill et al. (2002).
J Neurophysiol • VOL
90 • AUGUST 2003 •
www.jn.org
INTERSEGMENTAL COORDINATION OF RHYTHMIC MOTOR PATTERNS
active. Therefore a coordinating interneuron only fires in the
window of time when neither the ipsilateral G3 nor the G4
interneuron is active (yellow rectangles in Fig. 3C).
In a model in which the two segmental oscillators had the
same intrinsic period, the phase difference was zero and the
duty cycle of the coordinating interneurons was large (40%,
not shown). Under conditions where the intrinsic period of one
segmental oscillator was shorter than the other, the faster
segmental oscillator led in phase and the duty cycle of the
coordinating interneurons was reduced (30%, Fig. 3C). In the
symmetric model, either segmental oscillator can lead in phase
and the period of the system is equal to the period of the faster
segmental oscillator (Hill et al. 2002). The range of phase
values (⫺20 to ⫹20%) is greater in the negative direction than
in the biological system (⫺12 to ⫹20%), perhaps reflecting a
tendency of the real G4 oscillator to be inherently faster than
the G3 oscillator (Masino and Calabrese 2002b).
How are stable phase lags created in this system and why is
the range of stable phase lags so great? On a cycle-by-cycle
basis the faster segmental oscillator will lead in phase due to its
shorter intrinsic period. As a result, the oscillator interneurons
of the faster segmental oscillator will terminate the bursts of
the ipsilateral coordinating interneurons. Thus during each
cycle there is a brief interval in which an interneuron of the
slower segmental oscillator only receives inhibition from its
contralateral partner. The faster oscillator interneuron “removes” inhibition that would normally fall late during the
inhibited phase of the slower oscillator interneuron’s cycle. In
this way, the faster segmental oscillator speeds the slower
segmental oscillator to its own period. The greater the phase
difference, the greater the removal of inhibition and the stronger the speeding effect on the slower segmental oscillator. At
the limit the faster segmental oscillator can accelerate the
coupled system to the slower oscillator’s half-center oscillator
period—the period that it expresses in the absence of inhibition
from the coordinating interneurons. Beyond this point, the two
oscillators do not share the same period (Hill et al. 2002).
The properties of this symmetric network depend not only
on the details of the network configuration, but also on the
intrinsic properties of the interneurons themselves. For this
reason many voltage-dependent currents have been incorporated in the model oscillator interneurons (Hill et al. 2001). For
example, the bursts are supported by several currents including
a low-threshold, slowly inactivating Ca2⫹ current (ICaS). ICaS is
important because it inactivates during a burst, causing a slow
decline in the membrane potential. This decline leads to a
reduction in spike frequency, which helps to release the contralateral interneuron from synaptic inhibition. Simultaneously,
Ih becomes activated in the contralateral interneuron, helping it
to escape from inhibition and begin to burst. Late in the
inhibited phase, small changes in the amount of inhibition can
have large effects on the timing of the next burst (Hill et al.
2002). Therefore the removal of inhibition can effectively
speed an oscillator interneuron that lags in phase. Similar
results, showing that inhibitory input can cause either phase
advances or delays depending on its timing, have been found in
a modeling study of intersegmental coordination in the tadpole
swim network (Tunstall et al. 2002).
The predictions of this simple symmetric model are largely
in agreement with physiological data (Masino and Calabrese
2002a,b). In mutual entrainment experiments in which the two
J Neurophysiol • VOL
535
segmental oscillators were reversibly uncoupled by blocking
spike conduction in the connective between the 3rd and 4th
ganglia, the period of the coupled system was equal to that of
the faster oscillator regardless of which oscillator was faster
(G3 or G4). In addition, in split-bath experiments the application of pharmacological agents that accelerated or slowed the
period of a segmental oscillator allowed for a reversal in the
phase relationships between the oscillators (Masino and Calabrese 2002b). For example, in a preparation in which the G4
oscillator originally led in phase, a decrease in the intrinsic
period of the G3 segmental oscillator allowed the latter to lead
in phase.
The agreement between the model and the biological data
breaks down, however, in experiments in which repetitive
pulses of current were used to drive one of the segmental
oscillators to periods faster or slower than that of the mutually
entrained system (analogous to the forcing experiments described in the lamprey) (Masino and Calabrese 2002c). The
symmetric model predicts that the driven oscillator cannot slow
the follower segmental oscillator by lagging in phase because
there is no mechanism to add inhibition to the follower oscillator. In contrast, in the biological system the driven segmental
oscillator may lag in phase and consequently slow the system.
A new generation of the model, which incorporates details
such as spike frequency adaptation of the coordinating interneurons, has been able to account for this property (Jezzini et
al. 2000).
In addition, contrary to the assumptions of the symmetric
model, the driving experiments have revealed that the system
can behave asymmetrically. The driven G3 oscillator can entrain the network over a broader range of cycle periods than the
G4 oscillator (Masino and Calabrese 2002c). This result shows
that the leech heartbeat network changes dynamically depending on the experimental conditions. The network behaves symmetrically under conditions of mutual entrainment but asymmetrically when driven by external input. The ability of the
system to switch between these two modes has been explored
in a new model in which coordinating interneurons have two
spike initiation zones (Jezzini et al. 2000).
The functional role of the large variations in phase observed
in the mutually entrained system is not known. The experiments and modeling to date have been concerned with short
isolated chains of ganglia. In experiments in which phase was
measured between segmental motor neurons in long, de-afferented chains, intersegmental phase differences were consistent
between preparations and were independent of cycle period
(Wenning et al. 2000). The study of semi-intact preparations in
which afferent input is preserved may yield further information
about the normal phase relations in this system.
A LARGE AND CONSTANT PHASE LAG IS
NECESSARY FOR THE BEATING OF CRAYFISH
SWIMMERETS
Intersegmental coordination has also been studied in the
crayfish swimmeret system (Mulloney et al. 1993, 1998). The
crayfish swims forward by beating four pairs of swimmerets
located ventrally on the abdomen. Normally the two swimmerets located on a single abdominal segment beat synchronously in a cycle consisting of a power-stroke followed by a
return-stroke (Hughes and Wiersma 1960). Together the swim-
90 • AUGUST 2003 •
www.jn.org
536
A.A.V. HILL, M. A. MASINO, AND R. L. CALABRESE
merets beat in a wavelike pattern that begins with the posteriormost pair and then spreads anteriorly with a phase lag of 25%
between neighboring abdominal segments. As in the leech and
lamprey swim networks, the phase lag between segments is
independent of cycle frequency (Braun and Mulloney 1993).
The isolated abdominal nerve cord of the crayfish can produce a rhythm that is nearly identical to that in the intact animal
(Braun and Mulloney 1993, 1995; Ikeda and Wiersma 1964).
In addition, a single isolated ganglion can produce a normal
and robust motor pattern (Murchison et al. 1993). Similar to the
lamprey swim network, a segmental oscillator consists of two
hemisegmental pattern-generating circuits that are capable of
independent oscillation, although in the crayfish the hemisegments are active in phase as would be necessary for synchronous beating of swimmeret pairs (Murchison et al. 1993).
Each hemisegment contains all the circuitry necessary to
control a single swimmeret (Murchison et al. 1993). In addition
to two groups of antagonistic motor neurons and a set of
primary afferent neurons, each hemisegment contains four
local, nonspiking interneurons. Two of these interneurons (1A
and 1B) depolarize during the return-stroke portion of each
cycle while the other two interneurons (2), which are identical,
depolarize during the power-stroke phase (Fig. 4) (Paul and
Mulloney 1985a,b). Alternating oscillations between these two
groups of interneurons may be based on reciprocal inhibition
(Fig. 4) (Skinner and Mulloney 1998b).
It was originally proposed that the wavelike activation of the
FIG. 4. Pattern of intersegmental coupling between hemisegmental oscillators of the crayfish swimmeret system predicted by a model. A circuit of
nonspiking interneurons that are mutually inhibitory generates the oscillatory
rhythm. Interneurons 1A and 1B are active in phase with the return-stroke. A
pair of identical interneurons (2) is active in phase with the power-stroke.
There are 3 coordinating interneurons in each hemisegment. ASCL and ASCE
project in the ascending direction, whereas DSC projects in the descending
direction. The coordinating interneurons were not explicitly modeled but were
simply represented as connections directly between oscillator interneurons.
The output of the descending interneuron (DSC) can presumably be split
among more than 1 target and can have different signs, since the coordinating
interneurons act through intermediary nonspiking commissural interneurons
(Mulloney and Hall 2002). The solid circles represent inhibitory synapses;
open triangles represent excitatory synapses. Figure adapted from Jones et al.
(2003).
J Neurophysiol • VOL
swimmerets arises from a gradient of intrinsic segmental oscillator periods (Ikeda and Wiersma 1964). However, no systematic difference in segmental oscillator periods was subsequently found between isolated ganglia (Mulloney 1997). In
contrast, the pattern of intersegmental connectivity is asymmetric. Three types of coordinating interneurons have been
identified that are necessary and sufficient for intersegmental
coordination (Namba and Mulloney 1999). In each hemisegment, there are two interneurons that project in the ascending
direction (ASCL and ASCE), and one interneuron that projects
in the descending direction (DSC) (Fig. 4). The ASCL and
ASCE interneurons fire bursts in phase with the power-stroke
in their home ganglion, whereas the DSC interneuron fires
bursts in phase with return-stroke. These coordinating interneurons have little direct effect on activity in their home
ganglion but affect the timing of motor output in their target
ganglia where they make spike-mediated synapses with local,
nonspiking commissural interneurons, which in turn entrain
hemisegmental activity (Mulloney and Hall 2002; Namba and
Mulloney 1999).
Chains of only two ganglia exhibited normal phase relationships, leading to the original assumption that intersegmental
coordinating interneurons project only to neighboring ganglia
(Skinner and Mulloney 1998b). However, when the rhythmic
activity of an individual ganglion of a chain was blocked with
a low-Ca2⫹, high-Mg2⫹ saline, the segmental oscillators on
either side maintained their normal phase difference, demonstrating that coordinating interneurons project to targets at least
two ganglia away (Tschuluun et al. 2001). In agreement with
the synaptic spread model proposed for lamprey swimming
these distant connections are weaker than those between neighboring ganglia. In contrast, the appropriate phase lag of about
50% is maintained between the two ganglia on either side of
the inactive ganglion, suggesting that the connections in distant
ganglia are not identical to those made in neighboring ganglia.
A variety of models have been used to help understand how
intersegmental coordination is accomplished in the crayfish
swimmeret system. In one model, the swimmeret network was
represented as a system of phase-coupled oscillators (Skinner
et al. 1997). Fitting this model to the experimental data resulted
in a number of predictions: coupling is asymmetric; ascending
and descending coupling are about equal in strength, and either
ascending or descending coupling alone can generate a phase
difference of 25%. To understand what these asymmetries
mean in synaptic terms, a cellular model was created in which
the segmental oscillators were modeled using Morris–Lecartype equations to represent the nonspiking interneurons (Skinner and Mulloney 1998b). A number of alternative circuits,
constrained by experimental results and the predictions of the
phase-coupled model, were tested. One circuit exhibited at
constant phase of 25% over a range of oscillation frequencies.
As predicted by the phase-coupled oscillator model, either the
ascending or the descending coordinating interneurons alone
could produce a 25% phase lag. However, only the full circuit
model showed phase constancy. This cellular-based model
shows that the asymmetric coupling may be embodied in
differences in phases of coordinating interneuron activity,
choices of intersegmental targets, and signs of intersegmental
connections.
The cellular model predicts that certain network configurations produce appropriate phase lags but cannot explain why
90 • AUGUST 2003 •
www.jn.org
INTERSEGMENTAL COORDINATION OF RHYTHMIC MOTOR PATTERNS
these phase lags are stable. In a study based on the assumptions
of weakly coupled oscillator theory, coupling functions were
calculated for individual intersegmental connections in the
cellular model (Jones et al. 2003). These coupling functions are
curves that predict stable phase lags for individual connections
and when added together can predict stable phase lags within
a full network. For example, in the cellular model a stable
phase lag of 25% was found with a specific pair of ascending
excitatory and inhibitory connections. By examining the coupling functions of these two connections, it was possible to see
that although neither connection on its own produces an appropriate phase, together they yield a stable phase lag of 25%.
Furthermore, this analysis explained why bidirectional connections are required for phase constancy. With only either ascending or descending connections, the predicted phase lags
shifted systematically with oscillation frequency. In a model
with bidirectional coupling, however, these phase shifts cancelled out, resulting in a constant phase lag over a range of
frequencies.
CONCLUSIONS
This review has focused on intersegmental coordination in a
few well-studied preparations. Discovering how motor patterns
are coordinated in these model systems may help us to understand how coordination is accomplished in more complicated
networks such as those responsible for terrestrial limbed locomotion (Bem et al. 2003; Butt et al. 2002). At the most
fundamental level of rhythm generation there are clear similarities between the model systems we have discussed. For
example, in the lamprey swim network and in the leech heartbeat system, hemisegments are capable of endogenous bursting
even though they are embedded in a half-center oscillator
circuit. However, at the network level, the differences are
perhaps more striking than the similarities. For example, the
swimming behaviors of leech and lamprey are very similar yet
the two underlying neural circuits are very different in terms of
their architecture and the role of sensory feedback.
A further complexity is that a given network may operate
differently depending on experimental conditions. For example, the leech heartbeat network behaves nearly symmetrically
under conditions of mutual entrainment but asymmetrically
during driving experiments, demonstrating an ability to change
dynamically that is similar to network reconfiguration in the
crustacean stomatogastric nervous system (Marder and Calabrese 1996; Masino and Calabrese 2002b,c).
The future of this field will clearly continue to involve
modeling work because the dynamic nature of these oscillatory
networks makes comprehension based on static circuit diagrams impossible. The use of models with different levels of
detail, as in the study of the crayfish swimmeret system, may
be the best approach. Abstract models provide a general indication of the important features of a system, while more
detailed models allow for a direct comparison with the biological system.
REFERENCES
Bem T, Cabelguen JM, Ekeberg Ö, and Grillner S. From swimming to
walking: a single basic network for two different behaviors. Biol Cybern 88:
79 –90, 2003.
J Neurophysiol • VOL
537
Braun G and Mulloney B. Cholinergic modulation of the swimmeret motor
system in crayfish. J Neurophysiol 70: 2391–2398, 1993.
Braun G and Mulloney B. Coordination in the crayfish swimmeret system:
differential excitation causes changes in intersegmental phase. J Neurophysiol 73: 880 – 885, 1995.
Brodin L, Tråvén HG, Lansner A, Wallén P, Ekeberg Ö, and Grillner S.
Computer simulations of N-methyl-D-aspartate receptor-induced membrane
properties in a neuron model. J Neurophysiol 66: 473– 484, 1991.
Brodfuehrer PD, Debski EA, O’Gara BA, and Friesen WO. Neuronal
control of leech swimming. J Neurobiol 27: 403– 418, 1995.
Buchanan JT. Identification of interneurons with contralateral, caudal axons
in the lamprey spinal cord: synaptic interactions and morphology. J Neurophysiol 47: 961–975, 1982.
Buchanan JT. Neural network simulations of coupled locomotor oscillators in
the lamprey spinal cord. Biol Cybern 66: 367–374, 1992.
Butt SJ, Lebret JM, and Kiehn O. Organization of left-right coordination in
the mammalian locomotor network. Brain Res Brain Res Rev 40:107–117,
2002.
Calabrese RL, Nadim F, and Olsen ØH. Heartbeat control in the medicinal
leech: a model system for understanding the origin, coordination, and
modulation of rhythmic motor patterns. J Neurobiol 27: 390 – 402, 1995.
Cang J and Friesen WO. Sensory modification of leech swimming: rhythmic
activity of ventral stretch receptors can change intersegmental phase relationships. J Neurosci 20: 7822–7829, 2000.
Cang J and Friesen WO. Model for intersegmental coordination of leech
swimming: central and sensory mechanisms. J Neurophysiol 87: 2760 –
2769, 2002.
Cohen AH. Effects of oscillator frequency on phase-locking in the lamprey
central pattern generator. J Neurosci Methods 21: 113–125, 1987.
Cymbalyuk GS, Gaudry Q, Masino MA, and Calabrese RL. Bursting in
leech heart interneurons: cell-autonomous and network-based mechanisms.
J Neurosci 22: 10580 –10592, 2002.
Dale N. Excitatory synaptic drive for swimming mediated by amino acid
receptors in the lamprey. J Neurosci 6: 2662–2675, 1986.
Ekeberg Ö. A combined neuronal and mechanical model of fish swimming.
Biol Cybern 69:363–374, 1993.
Ekeberg Ö and Grillner S. Simulations of neuromuscular control in lamprey
swimming. Phil Trans R Soc Lond 354: 895–902, 1999.
Friesen WO. Neuronal control of leech swimming movements. In: Cellular
and Neuronal Oscillators, edited by Jacklet JW. New York: Marcel Dekker,
1989, p. 269 –316.
Friesen WO and Cang J. Sensory and central mechanisms control intersegmental coordination. Curr Opin Neurobiol 11: 678 – 683, 2001.
Friesen WO and Hocker CG. Functional analyses of the leech swim oscillator. J Neurophysiol 86: 824 – 835, 2001.
Friesen WO and Pearce RA. Mechanisms of intersegmental coordination in
leech locomotion. Semin Neurosci 5: 41– 47, 1993.
Grillner S, Deliagina T, Ekeberg Ö, El Manira A, Hill RH, Lansner A,
Orlovsky GN, and Wallén P. Neural networks that co-ordinate locomotion
and body orientation in lamprey. Trends Neurosci 18: 270 –279, 1995.
Grillner S, McClellan A, and Perret C. Entrainment of the spinal pattern
generators for swimming by mechano-sensitive elements in the lamprey
spinal cord in vitro. Brain Res 217: 380 –386, 1981.
Hagevik A and McClellan AD. Coupling of spinal locomotor networks in
larval lamprey revealed by receptor blockers for inhibitory amino acids:
neurophysiology and computer modeling. J Neurophysiol 72: 1810 –1829,
1994.
Hagevik A and McClellan AD. Coordination of locomotor activity in the
lamprey: role of descending drive to oscillators along the spinal cord. Exp
Brain Res 128: 481– 490, 1999.
Hellgren J, Grillner S, and Lansner A. Computer simulation of the segmental neural network generating locomotion in lamprey by using populations of
network interneurons. Biol Cybern 68: 1–13, 1992.
Hellgren-Kotaleski J, Grillner S, and Lansner A. Neural mechanisms potentially contributing to the intersegmental phase lag in lamprey. I. Segmental oscillations dependent on reciprocal inhibition. Biol Cybern 81: 317–330,
1999a.
Hellgren-Kotaleski J, Lansner A, and Grillner S. Neural mechanisms potentially contributing to the intersegmental phase lag in lamprey. II. Hemisegmental oscillations produced by mutually coupled excitatory neurons.
Biol Cybern 81: 299 –315, 1999b.
Hill AA, Lu J, Masino MA, Olsen OH, and Calabrese RL. A model of a
segmental oscillator in the leech heartbeat neuronal network. J Comput
Neurosci 10: 281–302, 2001.
90 • AUGUST 2003 •
www.jn.org
538
A.A.V. HILL, M. A. MASINO, AND R. L. CALABRESE
Hill AA, Masino MA, and Calabrese RL. Model of intersegmental coordination in the leech heartbeat neuronal network. J Neurophysiol 87: 1586 –
602, 2002.
Hocker CG, Yu X, and Friesen WO. Functionally heterogeneous segmental
oscillators generate swimming in the medical leech. J Comp Physiol [A]
186: 871– 883, 2000.
Hughes GM and Wiersma CAG. The co-ordination of swimmeret movements in the crayfish, Procambarus clarkii. J Exp Biol 37: 657– 670, 1960.
Ikeda K and Wiersma CAG. Autogenic rhythmicity in the abdominal ganglion of the crayfish: the control of swimmeret movements. Comp Biochem
Physiol 12: 107–115, 1964.
Jezzini SH, Hill AAV, and Calabrese RL. Dynamic activity of a coordinating
fiber with two initiation sites. Abst Soc Neurosci 164.1, 2000.
Jones SR, Mulloney B, Kaper TJ, and Kopell N. Coordination of cellular
pattern-generating circuits that control limb movements: the sources of
stable differences in intersegmental phases. J Neurosci 23: 3457–3468,
2003.
Kopell N and Ermentrout GB. Coupled oscillators and the design of central
pattern generators. Math Biosci 90: 87–109, 1988.
Kristan WB Jr, Stent GS, and Ort CA. Neuronal control of swimming in the
medicinal leech. I. Dynamics of the swimming rhythm. J Comp Physiol 94:
97–119, 1974.
Marder E and Calabrese RL. Principles of rhythmic motor pattern generation. Physiol Rev 76: 687–717, 1996.
Masino MA and Calabrese RL. Phase relationships between segmentally
organized oscillators in the leech heartbeat pattern generating network.
J Neurophysiol 87: 1572–1585, 2002a.
Masino MA and Calabrese RL. Period differences between segmental oscillators produce intersegmental phase differences in the leech heartbeat
timing network. J Neurophysiol 87: 1603–1615, 2002b.
Masino MA and Calabrese RL. A functional asymmetry in the Leech
Heartbeat Timing Network is revealed by driving the network across various
cycle periods. J Neurosci 22: 4418 – 4427, 2002c.
Matsushima T and Grillner S. Neural mechanisms of intersegmental coordination in lamprey: local excitability changes modify the phase coupling
along the spinal cord. J Neurophysiol 67: 373–388, 1992.
McClellan AD and Hagevik A. Coordination of spinal locomotor activity in
the lamprey: long-distance coupling of spinal oscillators. Exp Brain Res 126:
93–108, 1999.
Miller WL and Sigvardt KA. Extent and role of multisegmental coupling in
the lamprey spinal locomotor pattern generator. J Neurophysiol 83: 465–
476, 2000.
Mulloney B. A test of the excitability-gradient hypothesis in the swimmeret
system of crayfish. J Neurosci 17: 1860 –1868, 1997.
Mulloney B and Hall WM. Local commissural interneurons are targets of
swimmeret coordinating axons. Abst Soc Neurosci 270.5, 2002.
Mulloney B, Murchison D, and Chrachri A. Modular organization of pattern-generating circuits in a segmental motor system: the swimmerets of
crayfish. Semin Neurosci 5: 49 –57, 1993.
Mulloney B, Skinner FK, Namba H, and Hall WM. Intersegmental coordination of swimmeret movements: mathematical models and neural circuits. Ann NY Acad Sci 860: 266 –280, 1998.
Murchison D, Chrachri A, and Mulloney B. A separate local patterngenerating circuit controls the movements of each swimmeret in crayfish.
J Neurophysiol 70: 2620 –2631, 1993.
Namba H and Mulloney B. Coordination of limb movements: three types of
intersegmental interneurons in the swimmeret system and their responses to
changes in excitation. J Neurophysiol 81: 2437–2450, 1999.
Paul DH and Mulloney B. Local interneurons in the swimmeret system of the
crayfish. J Comp Physiol [A] 156: 489 –502, 1985a.
Paul DH and Mulloney B. Nonspiking local interneuron in the motor pattern
generator for the crayfish swimmeret. J Neurophysiol 54: 28 –39, 1985b.
Pearce RA and Friesen WO. Intersegmental coordination of the leech swimming rhythm. II. Comparison of long and short chains of ganglia. J Neurophysiol 54: 1460 –1472, 1985.
J Neurophysiol • VOL
Peterson EL. Generation and coordination of heartbeat timing oscillation in
the medicinal leech. I. Oscillation in isolated ganglia. J Neurophysiol 49:
611– 626, 1983a.
Peterson EL. Generation and coordination of heartbeat timing oscillation in
the medicinal leech. II. Intersegmental coordination. J Neurophysiol 49:
627– 638, 1983b.
Rovainen CM. Synaptic interactions of identified nerve cells in the spinal cord
of the sea lamprey. J Comp Neurol 154: 189 –206, 1974.
Sigvardt KA and Williams TL. Effects of local oscillator frequency on
intersegmental coordination in the lamprey locomotor CPG: theory and
experiment. J Neurophysiol 76: 4094 – 4103, 1996.
Skinner FK, Kopell N, and Marder E. Mechanisms for oscillation and
frequency control in reciprocally inhibitory model neural networks. J Comput Neurosci 1: 69 – 87, 1994.
Skinner FK, Kopell N, and Mulloney B. How does the crayfish swimmeret
system work? Insights from nearest-neighbor coupled oscillator models.
J Comput Neurosci 4: 151–160, 1997.
Skinner FK and Mulloney B. Intersegmental coordination in invertebrates
and vertebrates. Curr Opin Neurobiol 8: 725–732, 1998a.
Skinner FK and Mulloney B. Intersegmental coordination of limb movements during locomotion: mathematical models predict circuits that drive
swimmeret beating. J Neurosci 18: 3831–3842, 1998b.
Tegnér J, Matsushima T, El Manira A, and Grillner S. The spinal GABA
system modulates burst frequency and intersegmental coordination in the
lamprey: differential effects of GABA(A) and GABA(B) receptors. J Neurophysiol 69: 647– 657, 1993.
Tegnér J, Lansner A, and Grillner S. Modulation of burst frequency by
calcium-dependent potassium channels in lamprey locomotor system: dependence of the activity level. J Comp Neurosci 5: 121–140, 1998.
Tschuluun N, Hall WM, and Mulloney B. Limb movements during locomotion: Tests of a model of an intersegmental coordinating circuit. J Neurosci 21: 7859 –7869, 2001.
Tunstall MJ and Roberts A. A longitudinal gradient of synaptic drive in the
spinal cord of Xenopus embryos and its role in co-ordination of swimming.
J Physiol 474: 393– 405, 1994.
Tunstall MJ, Roberts A, and Soffe SR. Modelling inter-segmental coordination of neuronal oscillators: synaptic mechanisms for uni-directional
coupling during swimming in Xenopus tadpoles. J Comput Neurosci 13:
143–158, 2002.
Ullström M, Hellgren-Kotaleski J, Tegnér J, Aurell E, Grillner S, and
Lansner A. Activity-dependent modulation of adaptation produces a constant burst proportion in a model of the lamprey spinal locomotor generator.
Biol Cybern 79: 1–14, 1998.
Wadden T, Hellgren-Kotaleski J, Lansner A, and Grillner S. Intersegmental co-ordination in the lamprey—simulations using a continuous network
model. Biol Cybern 76: 1–9, 1997.
Wallén P. Spinal mechanisms controlling locomotion in dogfish and lamprey.
Acta Physiol Scand Suppl 503: 1– 45, 1982.
Wallén P, Ekeberg Ö, Lansner A, Brodin L, Tråvén H, and Grillner S. A
computer-based model for realistic simulations of neural networks. II. The
segmental network generating locomotor rhythmicity in the lamprey. J Neurophysiol 68: 1939 –1950, 1992.
Wallén P and Williams TL. Fictive locomotion in the lamprey spinal cord in
vitro compared with swimming in the intact and spinal animal. J Physiol
347: 225–239, 1984.
Wenning A, Hill AAV, and Calabrese RL. Regulation of blood pressure in
the leech: interaction of peripheral neuromodulation and motor pattern. Abst
Soc Neurosci 164.4, 2000.
Williams TL. Phase coupling by synaptic spread in chains of coupled neuronal
oscillators. Science 258: 662– 665, 1992.
Williams TL, Sigvardt KA, Kopell N, Ermentrout GB, and Remler MP.
Forcing of coupled nonlinear oscillators: studies of intersegmental coordination in the lamprey locomotor central pattern generator. J Neurophysiol
64: 862– 871, 1990.
Yu X, Nguyen B, and Friesen WO. Sensory feedback can coordinate the
swimming activity of the leech. J Neurosci 19: 4634 – 4643, 1999.
90 • AUGUST 2003 •
www.jn.org