Orientation-dependent changes in single motor neuron activity during adaptive soft-bodied locomotion.

Original Paper Brain Behav Evol 2015;85:47–62 DOI: 10.1159/000369372

Received: April 4, 2014 Returned for revision: May 14, 2014 Accepted after revision: October 24, 2014 Published online: March 6, 2015

Orientation-Dependent Changes in Single Motor Neuron Activity during Adaptive Soft-Bodied Locomotion Cinzia Metallo a Barry A. Trimmer b a

Neuroscience Program, Sackler School of Biomedical Sciences, Tufts University School of Medicine, Boston, Mass., and b Biology Department, Tufts University, Medford, Mass., USA

Abstract Recent major advances in understanding the organizational principles underlying motor control have focused on a small number of animal species with stiff articulated skeletons. These model systems have the advantage of easily quantifiable mechanics, but the neural codes underlying different movements are difficult to characterize because they typically involve a large population of neurons controlling each muscle. As a result, studying how neural codes drive adaptive changes in behavior is extremely challenging. This problem is highly simplified in the tobacco hawkmoth Manduca sexta, which, in its larval stage (caterpillar), is predominantly soft-bodied. Since each M. sexta muscle is innervated by one, occasionally two, excitatory motor neurons, the electrical activity generated by each muscle can be mapped to individual motor neurons. In the present study, muscle activation patterns were converted into motor neuron frequency patterns by identifying single excitatory junction potentials within recorded electromyographic traces. This conversion

© 2015 S. Karger AG, Basel 0006–8977/15/0851–0047$39.50/0 E-Mail [email protected] www.karger.com/bbe

was carried out with single motor neuron resolution thanks to the high signal selectivity of newly developed flexible microelectrode arrays, which were specifically designed to record from M. sexta muscles. It was discovered that the timing of motor neuron activity and gait kinematics depend on the orientation of the plane of motion during locomotion. We report that, during climbing, the motor neurons monitored in the present study shift their activity to correlate with movements in the animal’s more anterior segments. This orientation-dependent shift in motor activity is in agreement with the expected shift in the propulsive forces required for climbing. Our results suggest that, contrary to what has been previously hypothesized, M.sexta uses central command timing for adaptive load compensation. © 2015 S. Karger AG, Basel

Introduction

All animals, whether vertebrates or invertebrates, must appropriately coordinate their bodies to navigate different environments. Depending on their morphology and the challenges posed by the environment in which they live, different species have evolved different solutions to locomotion and movement control. In verteCinzia Metallo Tufts Advanced Technology Laboratory 200 Boston Avenue, Suite 2600 Medford, MA 02155 (USA) E-Mail Cinzia.Metallo @ tufts.edu

Downloaded by: Kungliga Tekniska Hogskolan 198.143.54.65 - 8/14/2015 12:15:45 AM

Key Words Motor control · Motor neuron · Motor pattern · Adaptive behavior · Arthropod · Electromyography · Electrophysiology

48

Brain Behav Evol 2015;85:47–62 DOI: 10.1159/000369372

challenging to match muscle activation patterns to specific motor tasks [Gutfreund et al., 1998; Sumbre et al., 2005], complicating the acquisition of the neural codes that drive adaptive changes in behavior. Despite this challenge, the neural coding underlying behavior in the soft-bodied animal Manduca sexta can be monitored with high precision because both the neural and the muscular components are easily accessible. In the larval stage of M. sexta (caterpillars), movements are achieved by coordinating several concatenated segments, each containing approximately 50 longitudinal and oblique muscles [Peterson, 1912; Taylor and Truman, 1974; Levine and Truman, 1985]. Despite the large number of muscles potentially involved in any movement, the CNS is composed of only a few hundred motor neurons, each innervating one (occasionally two) muscle(s) [Taylor and Truman, 1974; Levine and Truman, 1985]. Furthermore, there is no complicating influence of inhibitory activity. As a result, by correlating EMG events to motor neuron activity, muscle activation patterns can be converted directly into specific motor neuron firing patterns. By exploiting the high signal selectivity of newly developed EMG electrode arrays [Metallo et al., 2011], this conversion was carried out with a single motor neuron resolution in the case of the dorsal internal medial (DIM) muscle (fig. 1a), which is one of the few dually innervated M. sexta muscles. Contrary to conventional implantable electrodes [Cooley and Vanderwolf, 1978; Whelan, 2003; Ahn et al., 2006], the high signal selectivity and sensitivity of the multichannel EMG electrodes used here is due to two main factors: (1) their intrinsic flexibility, which enables the electrodes to conform to the muscle surface, reducing motion artifacts and enhancing the signal-to-noise ratio (SNR), and (2) the ability to customize the size and the arrangement of their recording sites to match the anatomy of the DIM muscle [Metallo et al., 2011]. Using these electrode arrays, it was therefore possible to simultaneously monitor different regions of DIM and extract fine-scale electrical information from the composite EMG signals, considerably simplifying signal decomposition. As a result, the firing activity of each DIM motor neuron was correlated with crawling in different orientations and adaptive changes in the corresponding central neural commands were characterized. With its 19 muscle fibers, DIM is the largest M. sexta muscle [Levine and Truman, 1985] and spans the entire space between the dorsal midline and the line of the spiracles in each half body segment (fig. 1a). Despite its size, the function of DIM has never been studied in detail. Together with the ventral internal lateral (VIL) muscle, Metallo/Trimmer


brates and most adult arthropods, movement is based on muscles acting on a rigid skeleton via joints and tendons [Biewener, 2003]. A jointed skeleton not only provides a system of levers and mechanical support during movement, but it also acts as a structural constraint to limit the range of possible movements. In contrast, soft-bodied animals do not possess a stiff jointed skeleton. They have highly flexible and deformable bodies that can be twisted and bent in complex ways and in almost any possible direction. To move in such a high-dimensional workspace, most soft-bodied animals rely on muscles transmitting forces through pressurized fluids (hydrostatic skeleton) or through densely packed tissues (muscular hydrostats) [Kier, 2012]. However, the movements of some soft animals, such as larval insects, do not always depend on pressure changes but instead exploit the dynamic properties of their soft tissues to distribute forces through elastic storage and release [Lin and Trimmer, 2010; Lin et al., 2011]. The evolutionary origins of soft-bodied behavior [Grasso and Basil, 2009] and hydrostatic motor control [Mather and Kuba, 2013] are probably extremely ancient and diverse, with key structural and biomechanical features arising multiple times in different lineages [Van Leeuwen et al., 2000]. Regardless of body structure, the transmission of information from the central nervous system (CNS) to the muscles is encoded in the firing patterns and identities of neurons. An accurate understanding of this information flow depends on knowing the relationship between neural activity and motor output. Because of easily quantifiable mechanics, the study of the organizational principles underlying motor control has largely focused on model systems with stiff articulated skeletons. Animals such as stick insects [Bassler and Büschges, 1998; Büschges et al., 2008], cockroaches [Sponberg and Full, 2008; Büschges, 2012], fish [Tytell et al., 2010], and quadrupeds [Biewener, 1990, 2003; Pearson, 2004] have been studied using a variety of implantable electrodes [Ahn and Full, 2002; Keating and Gerstein, 2002; Whelan, 2003]. Unfortunately, not only it is rarely possible to record the activity of multiple neurons in freely moving animals but, since a large number of motor neurons typically controls each muscle, it is also rarely possible to decompose electromyographic (EMG) signals with sufficient accuracy to identify the contribution of single motor units to the recorded signals. Because of a reduced number of motor neurons controlling each muscle, motor control strategies can be more easily studied in invertebrates [Belanger, 2005], including soft-bodied animals [Bullock and Horridge, 1965]. However, the vast range of movements available to soft animals makes it

Color version available online

which is the second largest M. sexta muscle, DIM is believed to be a major propulsive muscle during locomotion. Because of its size, DIM is also expected to significantly contribute to body stiffening during proleg movements [Mezoff et al., 2004]. Horizontal and upward vertical crawling have been selected to study adaptation to changes in the orientation to the plane of motion for two main reasons: (1) they represent two natural M. sexta locomotor behaviors and (2) they represent two dynamically different conditions characterized by different force requirements. By acquiring the DIM neural codes in these two conditions, it was possible to directly test the hypothesis that soft-bodied animals exploit self-adaptive properties of muscles to simplify neural control when presented with environmental changes such as crawling orientation. In fact, as part of the idea that complex behaviors emerge from the dynamic interplay between the CNS, the body, and the environment [Pfeifer and Iida, 2005; Nishikawa et al., 2007], it has been proposed that, in soft-bodied animals, muscles and their associated structures might be able to mechanically compensate for changes in the environment and do not necessitate a precise neural control [Dorfmann et al., 2007, 2008; Woods et al., 2008]. This purely mechanical compensatory mechanism would significantly reduce the number of control variables that the CNS needs to monitor in order to produce well-coordinated movements and would explain why caterpillars are able to control all the degrees of freedom that are available to them using only a few hundred motor neurons. Indeed, measurements of muscle force and length changes using isometric, isotonic, and dynamic work loop protocols have shown that the effective stiffness of muscles varies with the magnitude, rate, and timing of loading [Josephson, 1993, 1999], suggesting that, in response to environmental perturbations, muscles may be able to adapt to operate in a different region of their force-length relationship without the need for direct motor commands. While it is generally believed that insects cannot transition between behaviors (and gaits) without the descending influence of neurons in the central complex [Ritzmann et al., 2005], it has been reported that de-brained caterpillars (i.e. caterpillars whose central complex has been removed) do not change their gait kinematics when transitioning from crawling to climbing [unpubl. data], suggesting that caterpillars use the same motor programs in both orientations. Furthermore, although most animals adopt distinct gait kinematics to adapt to changes in slope [Biewener, 1990, 2004], some animals, such as geckos and cockroaches [Biewener, 1990; Biewener et al., 2004; Ritzmann et al., 2005], use similar vertical and horizontal

Fig. 1. Position of the DIM muscle and intracellular experimental setup. a In a caterpillar, DIM is located between the dorsal midline and the line of the spiracles in each abdominal body segment (A6 to A3). b A cross-sectional representation of a caterpillar body shows the location of DIM. Different colors (online version only) indicate different muscle groups. c Schematic representation of the reduced preparation used for intracellular experiments. The arrow on the left shows the location of the stimulating suction electrode along the dorsal nerve, represented by the horizontal line originating from the ganglion. The arrow on the right shows the location of the recording suction electrode. P = Posterior; A = anterior. Muscle acronyms: D = dorsal; V = ventral; L = lateral; M = medial; I = internal; O = oblique; E = external.

Single Motor Neuron Activity during Adaptive Soft-Bodied Locomotion


A6

A5

A4

A3

Spiracles

Prolegs

D P

DIM fibers

a

A V

DIM DIL D V

c

Nerve cord

VIL

DIL

DIM

VIM Spiracle

VEO

V

D

49


Nerve cord

Ganglion

b

Spiracle VIL VIM

*

*

1s 0.5 mV 500 ms

0.2 mV

b

c

Averaged nerve APs

Averaged nerve APs

MN1+MN2 MN2

al Ve n

0.2 mV

f

l

e

r sa

d

5.3 ms

Averaged EJPs 9.1 ms

Do

1 mV

Averaged EJPs

er

a

lat

100 ms

tr o

0.2 mV

g

Fig. 2. Simultaneous recordings of nerve APs and intracellular EJPs from DIM fibers. a Spontaneous EJPs from a dually innervated fiber. Two sizes of EJPs are visible. The smaller EJPs have been marked with an asterisk. b Spontaneous EJPs from a singly innervated fiber. Only large EJPs are visible. c Nerve APs recorded from the dorsal nerve (top trace) and intracellular EJPs simultaneously recorded from a DIM muscle fiber (bottom trace). d Small and large EJPs were identified by threshold analysis and then over-

lapped in time (bottom trace). In this way, the corresponding nerve APs also overlap and can be easily identified (top trace). e, f The latency between the APs and the onset of the EJPs was calculated from their average values. g By recording from every fiber, a map of the spatial distribution of singly and dually innervated DIM fibers was generated. The most ventrolateral side of DIM is innervated by two motor neurons (MN1 and MN2). The most dorsal fibers of DIM are singly innervated by MN2.

kinematics. However, past studies have focused on animals with a rigid skeleton and little is known about how soft-bodied animals adapt their gait and their motor control strategies to changes in orientation. Here, we report for the first time that motor patterns underlying horizontal and upward vertical crawling in caterpillars do not preserve the timing of peak muscle activity as previously hypothesized. Not only were changes in crawling kinematics identified, but the timing of muscle activity was also found to be highly dependent on the orientation of the plane of motion, suggesting a more direct involvement of the CNS in M. sexta adaptive behavior.

(caterpillar) has a soft cylindrical body with 3 major anatomical divisions (fig. 1a): the head, 3 thoracic segments that bear thin thoracic legs, and 8 abdominal segments (A1 to A8, from the most anterior to the most posterior). Segments A3 to A6 bear bilaterally symmetrical pairs of prolegs, which are used by the animal to grip substrates.

Materials and Methods Animals Fifth day instar M. sexta larvae were used in all the experiments described here. Following the rearing protocol developed by Bell and Joachim [1978], animals were kept at 27 ° C on a fixed 17: 7hour light-dark cycle and fed an artificial diet. An M. sexta larva

50


Electrode Arrays Flexible microelectrode arrays were fabricated as previously described [Metallo et al., 2011]. Briefly, each array consists of 2 flexible insulating layers of parylene C that protect a chrome/gold conductive layer defining the contacts and the interconnection traces (fig. 3b). Each device is 20 μm thick and is provided with 12 recording electrodes. Pairs of electrodes were differentially connected to yield 6 recording channels (fig. 3a). The diameter of the electrodes (100 μm) was selected to be slightly smaller than the average width of the DIM muscle fibers to enable single fiber recordings. Furthermore, a longitudinal interelectrode distance of 200 μm allows the recording channels to be distributed on distinct DIM muscle fibers and monitor the whole muscle uniformly. The transverse spacing between electrodes was selected to be 100 μm to minimize the overall size of the devices, facilitate insertion, and minimize muscle damage.

Metallo/Trimmer


*


*

Ch6

Ch6

Ch5

Ch5

Ch4

Ch4

Ch3

Ch3

Ch2

Ch2

Ch1

Ch1

a

b


1s

Ch6 Ch5 Ch2 Dorsal

Ch1

Ventrolateral Ventrolateral

HC(p)

c

d

Spiracle

VC(p)

A4 up – down Crawl duration

A4 up – down Crawl duration

MN1 HC(p) –2

–1

0

1

2

3

4

5

–1

0

1

2

3

4

5

–1

0

1

2

3

4

5

–1

0

1 2 Time (s)

3

4

5

MN1 VC(p) –2 MN2 HC(p) –2 MN2 VC(p)

e

1s

f

Fig. 3. In vivo raw EMG recordings. a Schematic representation of

the array configuration used in all the in vivo experiments. Each array has 12 recording electrodes (black circles). Pairs of electrodes were differentially connected to yield 6 recording channels (Ch1– 6). b A postmortem dissection shows an electrode array underneath DIM muscle fibers, which run horizontally in the figure. The recording sites face upward. The arrows highlight single DIM fibers. c An electrode array is implanted in the dorsal side of the animal to target the DIM muscle (top). A sketch, corresponding to the dotted area on the caterpillar body, illustrates the position of the electrode array after implantation (bottom). d A typical, raw EMG recording from the DIM muscle. Four of the 6 channels of the array are shown. The top two traces represent signals recorded from the most dorsal side of DIM (channels 6 and 5), while the bottom two traces represent signals recorded from the most ven-

Reduced Preparation A reduced preparation was used for all the intracellular experiments [Weeks and Truman, 1985]. Briefly, after deeply chilling (anesthetizing) an M. sexta larva, an incision was made on the cuticle from the horn to the head capsule along the lateral spiracles to expose dorsal muscles (fig. 1a, 2b). The larva was then pinned down, with the inner side up, to a Sylgard dish in cold modified Miyazaki saline [Trimmer and Weeks, 1989]. Finally, after removing the gut, fat tissue and tracheal tubes were gently dissected away to expose the nerve cord and the underlying musculature (fig. 1b, c). Intracellular Muscle Recordings and Nerve Action Potentials Thin-wall glass microelectrodes with a resistance of 10–15 MΩ were pulled using a horizontal puller (model P-87; Sutter Instru-


g

–2

trolateral side of DIM (channels 1 and 2). e Raw EMG recordings from the A4 DIM muscle during 3 progressive horizontal crawls in one animal, 1 crawl per channel. f Three progressive vertical crawls in the same animal, 1 crawl per channel. The two vertical bars represent the beginning and the end of the A4 swing phase (i.e. A4 up – A4 down). The horizontal line at the bottom represents the average crawl duration. g Motor neuron (MN) firing patterns corresponding to progressive horizontal and vertical crawls. The activity of MN1 is shown in the top two traces. The activity of MN2 is shown in the bottom two traces. Six progressive horizontal crawls and 6 progressive vertical crawls performed by one animal were considered. The horizontal line represents the average crawl duration. The zero on the x-axes represents the beginning of the A4 swing phase (i.e. A4 up). HC = Horizontal crawl; VC = upward vertical crawl.

ment Co., Novato, Calif., USA). After being filled with a 2 M KCl solution, the electrodes were mounted on a flexible silver wire that had been previously dipped in hypochlorite-based bleach to allow an AgCl coat to form. Excitatory junction potentials (EJPs) were recorded from single fibers of the DIM muscle (fig. 3b, arrows) in 2 animals. To elicit muscle activity in DIM fibers, a custom-made suction electrode was used to stimulate the dorsal nerve emerging from the ganglion located in the same segment as the target DIM muscle (fig. 1c, left arrow). A second suction electrode was used to record nerve action potentials (APs) from the dorsal nerve (fig. 1c, right arrow). All the experiments were conducted at room temperature.


51


1s

Extracellular Muscle Recordings: Experimental Setup and Electrode Placement A treadmill was specifically designed for caterpillar locomotion using the 3-D CAD design software SolidWorks (Waltham, Mass., USA). The treadmill was 3-D printed using a fused deposition modeling printer (Dimension 1200; Stratasys) and placed in a grounded metal cage in an effort to minimize noise. The treadmill can rotate 360° and is connected by two adjustable arms to a printed circuit board holding the electrode array. This arrangement minimizes electrode displacement when the treadmill is rotated. By rotating the treadmill about the horizontal axis by 0 and 90°, it was possible to select two crawling orientations, i.e. horizontal and vertical upwards, respectively. During each experiment, the animals crawled on a circular nitrile rubber band with a cross-sectional diameter of 5 mm that was used as the treadmill belt. In the in vivo experiments presented here, the electrode arrays were implanted on the dorsal side of the abdominal segment A4 to record from the DIM muscle (also called A4 DIM) in 3 animals. The position of the implant is illustrated in figure 3c. First, the animals were deeply chilled (anesthetized) on ice and then secured on the treadmill belt using a paper clip. Next, a small incision was made along the dorsal midline of the caterpillar body and the array was fed into the cut [Metallo et al., 2011]. The array was slid perpendicular to the cut to cover the entire length of the DIM muscle, with the electrode pads facing outward (fig. 3a–c). To limit hemolymph seepage, great care was taken to make sure that the width of the cut on the cuticle was as close as possible to that of the implanted electrode. A small drop of rubber cement (Elmer’s Products Inc.) was distributed along the cut to keep the electrode in place. To guide insertion, the attachment points of DIM were carefully mapped to external features of the caterpillar’s body. Lastly, the tip of the animal’s horn was cut off and a silver ground wire was inserted and secured to the animal using a small amount of rubber cement. Once the anesthesia wore off and the animal could clearly grip the substrate, the paper clip was removed. Data were acquired using the PowerLab acquisition system (ADInstruments). EMG signals were amplified at ×10,000, low-pass filtered at 10 Hz, and high-

52


pass filtered at 10 kHz. A video camera (Cannon, 15 fps) was used to record the motion of the caterpillar throughout each experiment. To synchronize the video with the EMG recordings, the output of an LED light connected to a Grass S9 Stimulator (AstroMed) was fed to a recording channel of the acquisition system so that a flash of light in the video would correspond to a spike in that channel. Extracellular Muscle Recordings: Video Analysis The videos acquired during each EMG recording session were analyzed using VirtualDub (www.virtualdub.org), a video-editing program that allows frame-by-frame analysis. The start and the end times of the swing phase of each proleg were determined with a single frame resolution by defining the beginning of the swing phase as the first frame when the upward motion of each proleg could be detected and the end of the swing phase as the first frame when the proleg motion could no longer be detected. Stepping Pattern Analysis Manduca moves in a stereotypical anterograde wavelike motion consisting of a series of steps that are initiated by the most posterior proleg [Trimmer and Issberner, 2007]. Here, a crawl was defined as the time from the start of the swing phase (i.e. the moment the proleg breaks contact with the substrate) in the sixth abdominal segment (A6) to the end of the swing phase (i.e. when the proleg comes in contact with the substrate) in the third abdominal segment (A3) (fig. 4a). The swing phase of a single proleg was defined as the elapsed time between lift off (proleg up) and recontact with the substrate (proleg down). A stepping pattern is the sequence of proleg steps (i.e. up/down proleg movements) within a crawl. A stepping pattern was defined as progressive (p) when two main phases were present: (1) each proleg is lifted in succession, from A6 to A3, and (2) each proleg is brought back down to contact the substrate in the same order, from A6 to A3. Any pattern differing from the progressive pattern was called nonprogressive (np). Nonprogressive patterns were further classified according to the sequence of up and down proleg movements that characterized them (fig. 4c) as follows: (1) np1 pattern = A6 up, A5 up, A4 up, A6 down, A3 up, A5 down, A4 down, and A3 down; (2) np2 pattern = A6 up, A5 up, A6 down, A4 up, A5 down, A3 up, A5 down, A4 down, and A3 down, and (3) np3 pattern = A6 up, A5 up, A4 up, A6 down, A5 down, A3 up, A4 down, and A3 down. To visually illustrate differences in stepping patterns, the swing phases of the A5, A4, and A3 prolegs were color coded and the swing phases of the A5 and A3 prolegs were broken down into two segments: (1) Ax up/down – A4 up, and (2) Ax up/down – A4 down, where x = 3, 5 (fig. 5h). Since recordings were acquired from the DIM muscle located in the fourth abdominal segment (A4), only the two adjacent segments (A3 and A5) were considered. Throughout the text, the terms ‘progressive pattern’ and ‘progressive crawl’ will be used interchangeably to describe a crawl characterized by a progressive stepping pattern. The same convention will be used for the nonprogressive patterns. Extracellular Muscle Recordings: Signal Analysis A subset of crawls was selected from 3 animals and for each orientation based on three criteria: (1) Signal: Because the conversion of EJPs into motor neuron spike activity relies on the ability to precisely identify EJPs within the recorded signals, only crawls whose corresponding EMG signals

Metallo/Trimmer


Intracellular Muscle Recordings: Data Analysis Since the dorsal nerve used to record nerve APs innervates DIM and all the other dorsal muscles, the number of recorded APs exceeded the number of intracellular EJPs recorded simultaneously from single DIM fibers (fig. 2c). To identify the nerve APs triggering specific EJPs, threshold analysis was first applied to identify intracellular EJPs of different amplitudes. The two identifiable types of EJPs were classified according to their amplitude as (1) large EJPs (i.e. EJPs with the largest amplitude) or (2) small EJPs (i.e. EJPs with a smaller amplitude). Next, the identified EJPs were overlapped in time, causing the corresponding nerve APs to overlap as well (fig. 2d), similarly to spike trigger averaging. Lastly, by taking the average of both EJPs and nerve APs, the latency between the APs and the onset of the EJPs was measured (fig. 2e, f). Furthermore, by varying the stimulus intensity applied to the nerve branch innervating the DIM muscles, two different activation thresholds were identified: the motor neuron with the lowest threshold was named MN1; the motor neuron with the highest threshold was named MN2. Small EJPs were associated with the lowest threshold and thus were classified as evoked by MN1. On the other hand, large EJPs were associated with the highest threshold and thus were classified as evoked by MN2.

80 60

Time

40

8

HC

d

np3 A6 up

A5 up

A5 up

A5 up

A5 up

A4 up

A4 up

A6 D

A4 up

A3 up

A6 D

A4 up

A6 D

A6 D

A3 up

A5 D

A5 D

A5 D

A5 D

A3 up

A3 up

A4 D

A4 D

A4 D

A4 D

A3 D

A3 D

A3 D

A3 D

100

HC VC

5 4 3 2 0

A6 A5 A4 A3

80

6

60 40 20

1 Up Down

np2 A6 up

c

VC

Swing phase (%)

Crawl duration (s)

7

a

np1 A6 up

20 0

b

p A6 up

p

np1

np2

Fig. 4. Types of stepping patterns during horizontal and upward vertical crawling. a A typical progressive stepping sequence. b Percentage of progressive and nonprogressive stepping patterns observed during horizontal and vertical crawling. c List of the ob-

np3

e

0

HC(p)

VC(p)

VC(np1) VC(np2) VC(np3)

served stepping pattern types and the corresponding sequences of up and down proleg movements. d Relationship between stepping

pattern type (p, np1, np2, np3) and crawl duration. e The time that each proleg spends off the ground (swing phase) was plotted in the case of different stepping patterns in both orientations. All the percentages were calculated from 3 animals. Twelve crawls were considered for each animal and each orientation. D = Down; HC = horizontal crawl; VC = upward vertical crawl.

were artifact-free were selected. Motion artifacts mainly arose when the treadmill belt was pushed forward, masking the EJPs within the recorded traces. All the recordings reported here are from 3 animals in which the SNR was high, so that signal decomposition could be accomplished with fidelity. (2) Electrode position: to ensure meaningful comparisons, recordings were also selected from animals in which the electrode position on the DIM muscle could be accurately determined following postmortem dissections. (3) Crawls: in an effort to isolate EJPs exclusively related to crawling, only bouts of straight-line crawling were analyzed, avoiding crawls that included other movements. Additionally, only crawls that were sufficiently isolated from one another were included in the analysis to avoid any overlap in muscle activity due to closely consecutive crawls. Successive steps in a crawl can be coanalyzed because Manduca crawling is considered quasi-static (i.e. there is no significant inertial component transferred from one step to the next) [Trimmer and Issberner, 2007; Lin and Trimmer, 2010]. Here, 12 crawls were considered for each animal and each orientation. Since transients with different amplitudes were clearly discernible, voltage thresholds were used to identify EJP spike events within each channel. By capitalizing on the fact that different channels

detect EJPs with amplitudes that are proportional to the distance between a specific channel and the muscle fiber that generates the EJPs, the events identified by threshold analysis were cross-correlated in time and space to identify subgroups of spikes and attribute them to different motor units. Single EJPs were classified according to their amplitude as (1) large EJPs (i.e. EJPs with the largest amplitude), (2) small EJPs (i.e. EJPs with a smaller amplitude), and (3) noise and background activity from neighboring muscles (i.e. EJPs with the smallest detectable amplitude). Motor neuron firing activity was determined by considering single EJP spikes as single APs originating from the motor neuron that innervates the muscle fibers from which the EJPs were recorded. The instantaneous firing frequency was then plotted as a function of time for each motor neuron and for each type of stepping pattern in both crawling orientations. Data analysis was performed with DataView (http:// www.st-andrews.ac.uk/∼wjh/dataview/) and Matlab (Mathworks). A moving average was used to highlight trends in the instantaneous frequency plots. Five time points were considered, that is 2 points before and 2 points after each time point. The moving average was then fitted using a Gaussian fit. Both the moving average and its fit were calculated using Origin 9.0 (OriginLab).



53


Stepping pattern type (%)

p np


100

A6 A5 A4 A3

HC

160

p Avg Fit

120

140 120

120

80

80

80

60

60

60

40

40

40

20

20

20

0

0

0

0

1

2

3

4

160

5

p Avg Fit

140 120

b

–2

–1

0

1

2

3

4

160

c

5

120

–2

100

80

80

80

60

60

60

40

40

40

20

20

20

0

0

0

Moving Avg peak (s)

4

g

–1

0

1 2 Time (s)

MN2

3

4

5

e

–2

–1

0

1 2 Time (s)

3

4

2

3

4

5

f

5

np1 np3 Avg Avg

–2

–1

0

1 2 Time (s)

3

4

5

HC VC

MN1

Shorter crawl – (p)

3 A5 up

2

A4 up

A3 up

A3 down

Longer crawl – (np)

A4 down A5 down

1 0

1

120

100

–2

0

140

100

d

–1

160

p Avg Fit

140

np1 np3 Avg Avg

140 100

–1

VC

160

p Avg Fit

100

–2

MN2 Frequency (Hz)

VC

160

A4 swing

100

a

p

p np1 np3

p

p np1 np3

h

Fig. 5. Instantaneous frequency plots and moving averages. a Motor neuron (MN) 1 instantaneous firing frequency during progressive horizontal crawls. The moving average corresponding to one representative crawl and the Gaussian fit of the moving average are both shown. b MN1 instantaneous firing frequency during progressive vertical crawls. c MN1 instantaneous firing frequency during np1 and np3 vertical crawls. d MN2 instantaneous firing frequency during progressive horizontal crawls. e MN2 instantaneous firing frequency during progressive vertical crawls. f MN2 instantaneous firing frequency during np1 and np3 vertical crawls. For the sake of clarity, in c and f, the lines corresponding to the moving average and the Gaussian fit have the same color as the dots representing either np1 or np3 crawls (colors refer to the online version only). The horizontal line corresponding to A5 swing is not shown in c and f due to lack of space. The vertical dotted

54

A5 swing


lines represent the average time at which the crawls begin (i.e. A6 up). g The time values corresponding to the peaks of the fitted moving average shown in a–f are plotted as a function of stepping pattern types for each crawling orientation and for each motor neuron. h Schematic representation of the color-coded proleg swing phases (colors refer to the online version only). The swing phases of A5 and A3 have been broken down into two segments: (1) Ax up/down – A4 up; and (2) Ax up/down – A4 down, where x = 3, 5. The two panels on the right show the differences in swing phases between shorter crawls, such as progressive crawls, and longer crawls, such as np crawls. All crawls are from the same animal as figure 3. In the horizontal orientation, 6 progressive crawls were considered. In the vertical orientation, 6 progressive crawls, 3 np1 crawls, and 3 np3 crawls were considered. HC = Horizontal crawl; VC = upward vertical crawl; Avg = average.

Metallo/Trimmer


MN1 Frequency (Hz)

140

A4 swing


A3 swing

The role of the DIM muscle during adaptive locomotion in M. sexta was studied with a combination of intracellular and extracellular muscle recordings. Since DIM is one of the few caterpillar muscles innervated by two motor neurons, intracellular recordings from single fibers were used to determine the spatial distribution map of dually and singly innervated fibers. Intracellular EJPs from DIM fibers are characterized by two different amplitudes. The two types of EJPs were called small and large EJPs. Some fibers exhibit only large EJPs, while other fibers exhibit both small and large EJPs. Figure 2a and b shows spontaneous intracellular EJPs recorded from two DIM fibers. The fiber in figure 2a exhibited both large and small EJPs, while the fiber in figure 2b exhibited only large EJPs. By recording nerve APs and intracellular EJPs simultaneously, it was possible to match the recorded EJPs to their corresponding nerve APs and verify that the two types of EJPs originate from two distinct motor neurons. Figure 2c shows simultaneous nerve APs (top trace) and intracellular EJPs (bottom trace) from a DIM fiber. Since the recording suction electrode was positioned on a nerve branch that innervates DIM and several other muscles, the number of recorded nerve APs exceeded the number of recorded EJPs. As described in Intracellular Muscle Recordings: Data Analysis, the recorded EJPs (and thus the corresponding nerve APs) were temporally overlapped and then averaged (fig. 2d). The latency measured in a fiber that exhibits large EJPs is shown in figure 2e. The latency measured in a fiber that exhibits small EJPs is shown in figure 2f. Since the distance of the suction electrode from the two fibers is much greater than the distance between the two fibers, the two latency values (5.3 ms in fig. 2e and 9.1 ms in fig. 2f) strongly suggest that large and small EJPs originate from two physiologically distinct motor neurons. Since we did not record directly from the DIM motor neurons, MN1 and MN2 could not be unambiguously matched to the two DIM motor neurons that have been previously mapped in the M. sexta CNS [Levine and Truman, 1985]. However, by recording intracellular EJPs from DIM fibers, it was possible to produce a spatial distribution map of dually and singly in-

nervated fibers. The results are summarized in figure 2g. The most ventrolateral DIM fibers are dually innervated (fig. 2g, left), while the most dorsal fibers are innervated only by MN2 (fig. 2g, right), which is the motor neuron with the highest activation threshold. Even though the number of DIM fibers varies slightly in different animals (with an average of 19 fibers), the difference in the innervation pattern of ventrolateral and dorsal fibers was consistent in all the preparations. Because of the presence of dually innervated fibers, it is not possible to consider DIM as two functionally distinct muscles. However, by converting the EJPs that characterize DIM activation patterns into motor neuron firing patterns, it is possible to correlate the firing properties of MN1 and MN2 to behavior. To acquire DIM motor patterns during adaptive locomotion, a flexible microelectrode array specifically designed to record high-resolution muscle activity from Manduca muscles was employed [Metallo et al., 2011]. Each array is composed of 12 recording electrodes. Pairs of electrodes are differentially connected to yield 6 recording channels (fig. 3a). Motor patterns underlying horizontal and upward vertical crawling were acquired from the A4 DIM muscle in 3 animals (fig. 1a). After implantation, the recording channels of the array were parallel to the DIM muscle fibers (fig. 3b). While the outermost channels of the array recorded from the dorsal DIM fibers, the innermost channels recorded from the ventrolateral fibers, as illustrated in figure 3c. Figure 3d shows 4 channels of a typical 6-channel recording from the DIM muscle of one animal. In the two most dorsal channels (top two traces in fig. 3d), high-amplitude EJPs are visible. In the most ventrolateral channels (bottom two traces in fig. 3d), smaller-amplitude EJPs were also detected. Accounting for small amplitude variations, the EJPs detected by the array can be grouped into two main clusters: larger-amplitude EJPs and smaller-amplitude EJPs. The spatial segregation of the two types of EJPs across the recording channels is consistent with the spatial innervation map determined by recording intracellular EJPs from single DIM fibers (fig. 2g). Larger EJPs appear to be correlated with the most dorsal channels of the array (top two traces, fig. 3d) and thus with the most dorsal DIM fibers. On the other hand, smaller EJPs appear to be correlated with the most ventrolateral channels (bottom two traces, fig. 3d) and thus with the most ventrolateral DIM fibers. Typical raw EMG signals recorded from the A4 DIM muscle of the same animal during a series of horizontal and vertical crawls are shown in figure 3e and f. Each channel represents a progressive crawl, for a total of 3



Results

55


Postmortem Dissections Postmortem dissections were performed at the end of each experiment to determine the location of the electrode array during the EMG recording sessions. Reduced preparations were used with a minimal quantity of saline to minimize electrode displacement during the dissection steps.

56


last two proleg movements (A4 down and A3 down) are never altered. Most importantly, in nonprogressive patterns, fewer prolegs are simultaneously in swing in the middle of a crawl. Different stepping patterns translate into different crawl durations, as shown in figure 3d. The progressive pattern produces the fastest crawls in both orientations. The np1 pattern involves only one proleg step change and, as such, is the fastest of the np patterns. Additionally, as the sequence of up/down proleg movements is altered within a crawl, the time that individual prolegs spend off the ground (swing phase) is also altered (fig. 4e). For instance, in the case of the np3 pattern, the A3 proleg is lifted towards the end of a crawl, only after the A5 and A6 prolegs have touched the ground. As a result, its swing phase is significantly shorter than any other pattern (fig. 4e). The swing phases in figure 4e were expressed as percentages of the crawl duration. To better understand the differences between the firing patterns of MN1 and MN2 and to compare the motor neuron activity underlying different stepping patterns in the two orientations, firing patterns were converted into instantaneous frequency plots (fig. 5). Color-coded swing phases were used in combination with frequency plots to visualize the stepping patterns underlying specific frequency profiles and allow more direct comparisons. For the sake of clarity, firing patterns were grouped according to their underlying stepping pattern type, and comparisons among groups were carried out. Horizontal and Vertical Crawls with Progressive Stepping Patterns The instantaneous frequency profiles of the motor neuron firing activity underlying several horizontal and vertical crawls of one animal during a progressive stepping pattern are presented in figure 5. The instantaneous frequency plots for MN1 and MN2 during horizontal progressive crawling are shown in figure 5a and d, respectively. As shown by a moving average, the firing pattern of MN2 is characterized by a sharp increase in frequency at the beginning of the A5 and A4 swing phases, which largely overlap. This trend is visible in the frequency profile of MN1 as well, although in this case the firing activity is not confined within the average crawl duration. The instantaneous frequency plots for MN1 and MN2 during vertical progressive crawling are shown in figure 5b and e, respectively. In this case, the moving average shows an increase in MN2 frequency towards the end of A4 swing and during A3 swing, which partly overlaps with A4 swing. In particular, the initial increase in MN2 frequency is correlated with the part of A3 swing that goes from A3 up to A4 down. Metallo/Trimmer


progressive crawls in each orientation. The two vertical bars represent the beginning and the end of the A4 swing phase. The horizontal grey line at the bottom represents the overall crawl duration. During horizontal crawling, a large burst of activity can be observed at the beginning of A4 swing, starting before the A4 proleg is even lifted. On the other hand, during vertical crawling, a large burst of activity occurs towards the end of A4 swing and persists after the A4 proleg is back on the substrate. There are two main EJP amplitudes visible in each channel of figure 3e and f, consistent with the dual innervation of DIM. Because of the high signal selectivity of the electrode array, it was possible to identify single EJPs within the composite EMG signals and correlate them to either MN1 or MN2 according to (1) their amplitude and (2) their spatial distribution across the recording channels. In this way, the firing activity of single motor neurons was determined. The firing patterns of the two motor neurons MN1 and MN2 are shown in figure 3g for 6 progressive horizontal crawls and 6 progressive vertical crawls. The activity of motor neuron MN1 is shown in the top two traces, while the activity of MN2 is shown in bottom two traces. The 3 horizontal crawls and the 3 vertical crawls presented in figure 3e and f as raw EMG data are among those presented in figure 3g as motor neuron firing activity. The horizontal grey lines in figure 3g represent the average crawl duration, allowing a direct comparison with figure 3e and f. Even just by looking at the raw data, a higher firing frequency can be observed before the onset of A4 swing in both MN2 and MN1 in the horizontal orientation, in agreement with figure 3e. In the vertical orientation, the activity of MN2 is concentrated towards the end of the crawls, in agreement with figure 3f. To investigate whether different stepping patterns result in different motor neuron firing patterns, the stepping patterns exhibited by the 3 animals considered here were first identified and then classified (fig. 4a). A careful stepping pattern analysis revealed that, while the vast majority of horizontal crawls are progressive, caterpillars adopt alternative patterns 52% of the time during climbing (fig. 4b). In the horizontal orientation, the only nonprogressive pattern that was observed is np1, which differs from the progressive pattern in one step: A3 is lifted before A6 touches the substrate (fig. 4c). In the vertical orientation, 3 stepping patterns were identified: np1, np2, and np3 (fig. 4c). Despite a high interanimal variability, the np1 pattern constitutes 50% of all the nonprogressive vertical patterns, with the np2 and np3 patterns making up the remaining 50%. In all the nonprogressive patterns, the first two proleg movements (A6 up and A5 up) and the

A3 up

A3 down A4 down A5 down

a

b

80

60

60

40

40

20

20

0

HC(p)

VC(p)

A4 up + 1 s (VC) A4 up + 1 s (HC)

VC(np1) VC(np2, 3)

0

c

2

3 4 5 A5 up – A3 down – time (s)

6

Fig. 6. Average spike count for the MN2 motor neuron. a The two time windows in which the spike count was carried out: (A4 up + 1 s) and (A4 down – 1 s). b Percentage of spikes that fall in the two time windows plotted as a function of the orientation and the type of stepping pattern. c Percentage of spikes in the (A4 up + 1 s) in-

terval plotted as a function of the average crawl duration in the two orientations. The spike counts were averaged for the 3 animals. Ten crawls were considered for each animal. HC = Horizontal crawl; VC = upward vertical crawl.

Vertical Crawls with np1 and np3 Stepping Patterns The same animal that performed the progressive vertical crawls shown in figure 5b and e also performed a series of nonprogressive vertical crawls with np1 and np3 stepping patterns (np2 patterns were not performed by this particular animal). The instantaneous frequency profiles characterizing the activity of MN1 during these crawls and the corresponding moving average are shown in figure 5c, and those of MN2 are shown in figure 5f. As in the case of the progressive vertical crawls, the MN2 instantaneous frequency increases towards the end of A4 swing and during A3 swing. In the case of the np3 stepping pattern, the A4 swing duration is 1.6 s longer than that of the np1 stepping pattern (see bars at the bottom of fig. 5c, f). Furthermore, during the np3 crawls, a more sustained MN2 activity can be observed during A3 swing. Figure 5h summarizes the color-coded scheme used for the swing phases, with the middle panel illustrating how the A5 and A3 swing phases largely overlap during shorter crawls, as in the case of progressive crawls, and how they tend to move away from each other as the overall crawl duration increases, as in the case of nonprogressive crawls. Figure 5g illustrates the time values corresponding to the peak of the fitted moving averages shown in figure 5a–f as a function of crawl type (p, np1, and np3 as this particular animal did not perform np2 crawls) and orientation for both motor neurons. In the horizontal orientation, the moving average peaks close to zero (t = 0.2 s for

MN1, and t = 0.4 s for MN2), that is when the A4 proleg goes up. In the vertical orientation, the moving average peaks at a later time for both progressive (t = 2.1 s for MN1, and t = 1.9 s for MN2) and nonprogressive stepping patterns (t = 1.7 s for MN1 and np1, and t = 2.9 s for MN1 and np3; t = 2.4 s for MN2 and np1, and t = 3.2 s for MN2 and np3). To better highlight this time shift, an average spike count was performed. Differences in motor neuron activity were summarized by counting spikes in two time intervals: (1) 1 s after the A4 proleg is lifted (i.e. A4 up + 1 s) and (2) 1 s before the A4 proleg comes back in contact with the ground (i.e. A4 down – 1 s), as shown in figure 6a. The two intervals were chosen because they approximate the (A4 up – A5 down) and the (A3 up – A4 down) intervals, that is the last part of A5 swing and the first part of A3 swing. Furthermore, 1 s is long enough to carry out a spike count for every crawl, but most importantly short enough to limit the overlap of the A5 and A3 swing phases. Spike counts were computed and averaged for all the animals. The percentage of spikes that fall in each selected interval was plotted as a function of the orientation and the type of stepping pattern, as shown in figure 6b and c. In figure 6b, only the spikes originating from the MN2 motor neuron were considered since this is the motor neuron whose activity is more closely correlated with the overall crawl duration. In the horizontal orientation, there is a higher percentage of spikes in the (A4 + 1 s) interval than in the (A4 down – 1 s) interval, indicating that



57


A4 up A5 up

80

A4 down – 1 s

Spikes (%)

A4 up + 1 s

100

A4 up + 1 s (VC) A4 up + 1 s (HC) A4 down – 1 s (HC) A4 down – 1 s (VC)


100

HC

160

MN2 Frequency (Hz)

c

100

80

80

60

60

40

40

20

20

0

0 –2

–1

0


1

2

3

4

5 b

160

160 140

A5 before A4

120

120

100

100

80

80

60

60

40

40

20

20

0

0 –2

–1

0

1 2 Time (s)

3

4

5

d

No A3 swing

–2

140

the firing activity of MN2 peaks at the end of A5 swing (that is the beginning of A4) and subsequently decreases. In the vertical orientation, the situation is reversed: the firing activity of MN2 is higher in the (A4 down – 1 s) interval, suggesting that the increase in MN2 firing activity is delayed and occurs at the beginning of A3 swing (that is the end of A4 swing). This shift is more pronounced in the nonprogressive vertical patterns np2 and np3 probably due to a longer crawl duration. In figure 6c, the percentage of spikes in the (A4 + 1 s) time window was plotted as a function of the average crawl duration. MN2 activity in the (A4 + 1 s) interval decreases as the overall crawl duration increases, with the longest crawls being the nonprogressive crawls (fig. 4d). In addition to the change in the proportion of spikes in each phase of the crawl cycle, the overall spike frequency during a crawl was also analyzed for each animal and combination of orientation and stepping patterns (data not shown). There were no significant differences between the average firing frequency of MN2 and MN1, which was below 50 Hz for all animals and for all the combinations of stepping patterns and orientations. Further58

140 120

100

a

VC

160

A5 before A4

120

A5 swing

–1

0

1

2

3

4

5

No A3 swing

–2

–1

0

1 2 Time (s)

3

4

5

more, the average firing frequency during horizontal and vertical crawling did not change significantly, suggesting that the timing of muscle activation is more important than the firing frequency (data not shown). Unusual Crawls: Early A5 Swing and No A3 Swing To better understand the contribution of the A3 and A5 swing phases to the firing frequency patterns underlying vertical and horizontal crawling, two unusual types of crawls were analyzed: (1) 3 consecutive horizontal crawls in which the A5 swing phase preceded the A4 swing phase so that the two do not overlap (shown in fig. 7a and c for MN1 and MN2, respectively), and (2) 1 incomplete vertical crawl in which the A3 proleg did not swing (shown in fig. 7b and d for MN1 and MN2, respectively). Since the 2 animals that performed these crawls went on to behave normally and were able to complete numerous crawls, no morphological defects appear to explain these unusual stepping patterns. When the A5 proleg completes its swing phase before the A4 proleg is lifted during horizontal crawling, there is an increase in the instantaneous frequency before the onset of A4 swing (fig. 7a, dotted line), Metallo/Trimmer


MN1 Frequency (Hz)

140

Fig. 7. Instantaneous frequency plots of unusual horizontal and vertical crawls. a Instantaneous frequency profile of 3 consecutive horizontal crawls with A5 swing preceding A4 swing in the case of motor neuron (MN) 1 and MN2 (c). The dotted vertical line indicates the end of A5 swing and the beginning of A4 swing. Unlike standard crawls, A5 and A4 swing phases do not overlap. b Instantaneous frequency profile of an incomplete vertical crawl in which the A3 proleg fails to lift (arrow) in the case of MN1 and MN2 (d). HC = Horizontal crawl; VC = upward vertical crawl.

A4 swing


A3 swing

Discussion

Multisite Recordings Reveal a Functional Difference in DIM Motor Neurons. The high signal selectivity of the multichannel electrodes used here enabled recordings of muscle activity in freely behaving animals with single motor neuron resolution. The firing patterns of the two DIM motor neurons (MN1 and MN2) were isolated and correlated with single crawl cycles during horizontal and vertical crawling on a stiff nitrile rubber-based substrate. The MN1 motor neuron, which innervates the ventrolateral side of DIM, was found to fire more tonically than the MN2 motor neuron. Its activity was also less correlated with the duration of a crawl. On the other hand, activity in the MN2 motor neuron, which singly innervates the dorsal side of DIM and dually innervates the ventrolateral side together with MN1, was more closely correlated with the overall crawl duration. This result suggests that the lower threshold motor neuron MN1 is responsible for maintaining the DIM muscle in a state of tonic contraction for postural control and slow movements not associated with crawling. The higher threshold motor neuron MN2 would be recruited during movements such as crawling to increase the overall tension in the muscle. This finding is reminiscent of a suggestion by Barth [1937], who proposed that external M. sexta muscles control body turgor, while internal ones generate major movements. It was assumed that the smaller external muscles connect the intersegmental folds with the intrasegmental body wall to resist outward bulging when the large internal muscles contract. This is a reasonable suggestion but there is no evidence for a functional distinction between M. sexta internal and external muscles. Our results indicate that the dual innervation of Single Motor Neuron Activity during Adaptive Soft-Bodied Locomotion

large longitudinal muscles could fulfill these two roles. However, we have only identified distinct roles for motor neurons MN1 and MN2 during locomotion. It is possible that they each have different roles in other behaviors. Our findings are analogous to what has been reported in vitro for the dually innervated proleg retractor muscle of the silkmoth larva [Cox, 1989], where the synergistic action of two motor neurons increases the range of possible muscle contractions. The tonic activity noted here is also consistent with the decrease in body wall stiffness caused by cutting peripheral nerves, which results in body wall bulging through the loss of muscle tension [Kopec, 1919; Holst, 1934]. A possible experiment to quantify the tension generated by the two DIM motor neurons during crawling would be to directly stimulate the motor neurons with the firing patterns identified in vivo and simultaneously record the mechanical responses produced by the muscle fibers in a reduced preparation. The firing patterns of the two DIM motor neurons are correlated not only with the overall crawl duration but also with the swing phase of particular prolegs. During horizontal crawling, the activity of the MN2 motor neuron innervating the DIM muscle located in the nth abdominal segment is correlated with the swing phase of the proleg belonging to the (n – 1)th abdominal segment. On the other hand, during vertical crawling, the activity of the same motor neuron is correlated with the swing phase of the proleg belonging to the (n + 1)th abdominal segment, as shown in figure 8. In other words, the A4 DIM muscle is active during A5 swing in the case of horizontal crawling but it becomes more active during A3 swing in the case of climbing. The fact that the timing of muscle activity is so strongly dependent on the orientation of the plane of motion suggests that the CNS issues different sequences of motor commands to compensate for changes in orientation. However, it does not mean that all the adaptive changes are mediated by switching to a new motor program; it is more likely that this change represents a shift in the timing of muscle activation relative to its strain cycle. This is known to generate different forcelength relationships (work loops) in M. sexta muscles and to alter the relative contributions of positive and dissipative work outputs [Woods et al., 2008]. This could explain why there is a shift in the timing of muscle activation during crawling in the two orientations even when the stepping patterns are the same [Finlayson and Lowenstein, 1958; Grueber et al., 2001; Simon and Trimmer, 2009; van Griethuijsen and Trimmer, 2009] (fig. 8).


59


suggesting that the initial frequency increase in the motor activity of A4 DIM is correlated with the reextension of the A5 proleg and not with the release of the A4 proleg. The slight frequency increase (>30 Hz) observed after the end of A3 swing (>1.5 s) in figure 7a and c is due to body movements that the caterpillar performed at the end of the 3 horizontal crawls presented here. In the second case, when the A3 proleg failed to lift during climbing, there was no increase in firing frequency at the end of the A4 swing phase, just a continued decrease (

Developmental Changes in Hippocampal CA1 Single Neuron Firing and Theta Activity during Associative Learning.

Single motor unit H-reflex in motor neuron disorders.

Phasic modulation of vestibulospinal neuron activity during fictive locomotion in lampreys.

Human temporal cortical single neuron activity during working memory maintenance.

Human Temporal Cortical Single Neuron Activity during Language: A Review.

Context-dependent changes in motor control and kinematics during locomotion: modulation and decoupling.

Rewarding imperfect motor performance reduces adaptive changes.

Plastic changes in the spinal cord in motor neuron disease.

Adaptive changes in the motor cortex during and after longterm forelimb immobilization in adult rats.

Adaptive and non-adaptive changes in activity-deprived presynaptic terminals.

Prefrontal Single-Neuron Responses after Changes in Task Contingencies during Trace Eyeblink Conditioning in Rabbits.

Motor cortical activity during drawing movements: single-unit activity during sinusoid tracing.

Abnormal single motor unit behavior in the upper motor neuron syndrome.

Extending the Cortical Grasping Network: Pre-supplementary Motor Neuron Activity During Vision and Grasping of Objects.

Persistent Single-Neuron Activity during Working Memory in the Human Medial Temporal Lobe.

Functional Diversification of Motor Neuron-specific Isl1 Enhancers during Evolution.

Structured Variability in Purkinje Cell Activity during Locomotion.

Motor activity changes in scrapie-affected mice.

Neuroinflammation in motor neuron disease.

Postradiation motor neuron syndrome.

Motor neuron disease.

Motor neuron disease model.

Single-neuron activity and eye movements during human REM sleep and awake vision.

Activity of interneurons mediating reciprocal 1a inhibition during locomotion.