Electrolytic lesions within central complex neuropils of the cockroach brain affect negotiation of barriers

Cockroaches are agile insects that readily negotiate obstacles within their natural environment. When a cockroach climbs over a substantial barrier at slow speeds, it first evaluates the object. The distance from the obstacle, its height, and whether or not alternative routes exist all affect how an insect responds (Blaesing and Cruse, 2004; Harley et al., 2009; Pick and Strauss, 2005; Watson et al., 2002a). It then uses that information to redirect leg movements or to make postural adjustments. All of these changes require neural adjustments at least for the initiation of newly directed movement. Sensory information from the environment must be integrated with proprioceptive information about body position, and then used to make motor changes resulting in targeted movements.

Changes in movement direction can be collectively referred to as ‘transitional behaviors’ (Watson et al., 2002a). Transitional behaviors can be thought of as changes from straight, horizontal walking to another behavior as a result of changes in body posture, ground reaction forces, changes in gait or direction of leg movement (Dürr and Ebeling, 2005; Jindrich and Full, 1999; Mu and Ritzmann, 2005; Watson et al., 2002b; Watson et al., 2002a). These transitions may be subtle, occurring as a consequence of simply bumping into or stepping onto an object. For example, a small block placed in front of a cockroach only generates postural changes after the cockroach presses its limb down onto the raised surface (Watson et al., 2002a). More complex obstacles, such as those presented here, require less subtle changes in locomotor behavior. For example, climbing taller blocks requires an initial postural adjustment before the swing of the front leg will reach the top of the object.

The height of such postural adjustments is correlated with obstacle height (Watson et al., 2002a). Thus, the cockroach appears to evaluate the object and then actively adjust its posture, presumably through descending commands interacting with thoracic circuits that control leg movements. In cockroaches, a large part of this evaluation involves information provided by the insects' antennae (Harley et al., 2009). Indeed, faced with a shelf, the manner in which the antennae contact the object determines whether the cockroach will climb over or tunnel under the barrier. Antennal contact from above typically results in climbing whereas contact from below leads to tunneling. However, this process is also affected by ambient light (Harley et al., 2009). Under light conditions, the cockroach will tunnel three times as often as it will climb, while in the dark climbing and tunneling are equally likely. Here, contextual information from ambient lighting, sensed by the ocelli, and antennal information are combined to guide the appropriate motor response.

Which regions of the insect brain bring together the appropriate sensory information and then use it to formulate or modify appropriate descending commands? The central complex (CC) is a region within the insect brain thought to have a role in the modification of motor programs and sensori-motor integration. Indeed, it has been suggested that this highly ordered brain region may ‘supervise’ locomotion (Strausfeld, 1999). The CC is composed of several interconnected neuropils: the fan-shaped body (FB), ellipsoid body (EB), protocerebral bridge (PB) and paired nodules that are located in the protocerebrum of virtually all insects (Strausfeld, 1999). Nevertheless, little is known about how the CC might exert its influence on motor patterns. It has been suggested that this structure translates processed sensory information from the mushroom bodies or other loci into commands that excite the circuits of the thoracic ganglia (Huber, 1962). Recordings of activity in the PB of locusts revealed neurons that are responsive to specific orientations of polarized light and are arranged in a manner consistent with a topographic map of the sky (Heinze and Homberg, 2008; Heinze and Homberg, 2009). Multi-unit recordings in the FB and EB described numerous multi-sensory units that are responsive to antennal movement as well as to changes in ambient light (Ritzmann et al., 2008). However, these structures are not solely sensory, as the FB and EB receive connections from motor reporter neurons arising from the thoracic ganglia (Homberg, 1994), and stimulation in the CC can produce leg movements or calling song in crickets (Bender et al., 2010; Heinrich et al., 2001; Huber, 1960; Weinrich et al., 2008). Furthermore, the CC and the lateral accessory lobe (LAL), which receives projections from the CC, send axons to the ventral body (a pre-motor region of the brain) (Homberg, 1994; Okada et al., 2003; Strausfeld, 1999).

Manipulation of the CC does have motor consequences. Lesions to the CC led to inhibition of walking and acoustic behavior in crickets (Huber, 1960; Huber, 1962). In Drosophila, CC structural mutants show defects in walking speed, leg coordination and directional control of walking and flight (Ilius et al., 1994; Strauss, 2002; Strauss and Heisenberg, 1993). Indeed, the central complex appears to play a key role in motor planning (Heinze and Homberg, 2009) and units have been recorded in the CC that change their firing rate along with, and in some cases before, changes in stepping frequency (Bender et al., 2010). However, not much is known about the role of individual CC structures in obstacle negotiation. In cockroaches, large-scale lesions within the CC revealed heterogeneous deficits in turning behavior. Some individuals failed to turn, turned the wrong direction, or showed delayed turning, whereas others with damage within the CC exhibited normal turning behavior (Ridgel et al., 2007). The heterogeneous consequences of these large-scale lesions suggest a compartmentalization within the CC, with various regions playing different roles in controlling individual behaviors. This arrangement may even be task specific, as different obstacles require different kinds of sensory information and motor changes and thus, may be susceptible to damage in different brain regions.

Here we report on the behavioral effects of discrete electrolytic lesions within the various regions of the CC and surrounding areas of the cockroach brain. Although previous studies typically examined only a single behavior, here we present brain-lesioned cockroaches with several different types of obstacles, each requiring different sensory cues and motor adjustments. This strategy allowed us to compare the potential role of different regions within the CC in climbing, tunneling, turning and transitions between walls and horizontal surfaces. Behavioral responses to our experimental obstacles have been examined previously in both intact cockroaches and in individuals with more profound lesions. For example, shelf and block climbing were described previously in detail as quantitative ethograms composed of specific sequences of sub-elements (Harley et al., 2009). Because that study demonstrated significant differences in behavior under light and dark conditions, we examined behaviors of lesioned individuals under both lighting conditions. Our results begin to show how the various CC neuropils, or regions within individual neuropils, influence transitional movements as the cockroach negotiates objects in its environment.

We examined the effect of discrete electrolytic brain lesions in 47 Blaberus discoidalis (Serville 1839) individuals. These lesions were located in six different regions of the brain; four of which are elements of the CC. The behavior of these individuals was tested on up to four different obstacles both before and after the lesion. The number of individuals with lesions in each location was as follows: mushroom bodies, 4; antenno-cerebral tract, 7; protocerebral bridge, 5; fan-shaped body, 11; ellipsoid body, 4; lateral accessory lobe, 7 (see Figs S1B and S2B in the supplementary material). Nine lesions were located outside all of these structures, categorized as ‘other’ (see Figs S1C and S2C in the supplementary material).

We were able to further divide lesions in the fan-shaped body into medial and lateral components. The fan-shaped body was separated into four equal quadrants. Lesions in the middle two quadrants were determined as being ‘medial’. Lesions in the outer two quadrants were determined as being ‘lateral’. Under this criterion, four of the FB lesions were designated as medial and seven as lateral.

Cockroaches were first anesthetized using cold (−16°F). Once they stopped moving, a ligature was placed around the neck and the head was opened as described by Ritzmann et al. (Ritzmann et al., 2008). Most individuals moved normally after the ligature was removed. If an individual dragged its maxillary palps on the ground or exhibited an inability to support its head with its neck muscles, it was discarded as these may be symptoms of damage to the neck connectives. A small slice was made in the sheath surrounding the brain. We found that a damaged 16-channel silicon-based probe (NeuroNexus Technologies, Ann Arbor, MI, USA) provided a sharp and effective knife for this purpose. Once this was done a ground wire was placed into the top of the head capsule. Then an eligoy steel probe with a 1–2 μm tip (Micro Probes, Inc., Gaithersburg, MD, USA), which had first been dipped in calligraphy ink (Windsor and Newton, Piscataway, NJ, USA), was placed into the brain. Current was applied through the probe into the brain. The amount of current was kept constant by taking into account the resistance of the probe. The resulting lesion size was related to current duration. In the majority of cases current was applied for 30 s, which resulted in spherical lesions damaging a region averaging 90±8 μm in diameter for a sample of 10 lesions. For scale, in these individuals the fan-shaped body of Blaberus discoidalis measured 712.5±127 μm from edge to edge. On average this lesion resulted in damage that would span 12.8±5% (N=10) of the diameter of the fan shaped body, which is equivalent to one to two columns in that structure or a similar area in other regions of the brain (see Fig. S1B,C in the supplementary material). The calligraphy ink was deposited in the brain tissue as the probe was inserted allowing for better visualization of the lesion site histologically (see Fig. S1A in the supplementary material).

Once the lesion was made, the probe and ground wire were removed from the cockroach. Dental wax was used to re-affix the removed head cuticle. Special care was taken to prevent the wax from contacting the antennae or eyes of the cockroach. Once the head cuticle was back in place, the ligature around the neck was removed. The cockroach was then freed from the tube, and allowed to recover overnight (at least 12 h) before further testing. The entire lesion procedure typically took 10–15 min to complete.

We cannot state for sure that the only damage from this process is located within the region identified as the lesion. The probe itself could damage tissue during insertion. To assess the extent of this damage, we inserted the probe into five individuals without running the lesioning current. The location of the probe was determined by the presence of calligraphy ink. In four of the five individuals, ink from the probe was found in the FB, in the other it was found in the mushroom body. All these individuals did show some behavioral deficits; however, those were similar to individuals with lesions to regions that the probe passed through. Those individuals in which the probe contacted the FB exhibited delayed turning in 20.8% of trials and failure to turn in 8.3% as compared with 55.9% and 7.1%, respectively, with the electrolytic lesions. Neither of these behaviors was present in the individual in which the probe contacted the mushroom body. During block climbing behavior, despite the fact that these individuals showed the same number of attempts within 5 mm of the block as their lesioned counterparts, there were differences in the success of these attempts. These individuals exhibited a higher success rate than their lesioned counterparts (72% versus 64%). Since these individuals exhibited less abnormal behavior than their lesioned counterparts, we concluded that although the insertion of the probe probably did cause damage, it was less extensive than that caused by the addition of the electrical current.

Although the rest of the surgical procedure may also lead to behavioral deficits, we did not observe any differences between shams and those individuals with lesions outside the major neuropils, despite these insects having undergone the same procedures as those with lesions in the CC. Furthermore, a previous study found that just opening the insect's head (as was done in our procedure) did not increase abnormalities in turning behavior over that seen prior to the lesions (Ridgel et al., 2007).

After behavioral testing was completed, brains were removed and prepared for histological examination as previously described in Ridgel et al. (Ridgel et al., 2007). A light background stain was used so that the calligraphy ink that transferred from the probe would still be high contrast (see Fig. S1A in the supplementary material). The lesion was identified by locating a hole in the tissue of approximately 90 μm in diameter that spanned 8 or more 12 μm sections and was aided by additional contrast from the calligraphy ink (when present; see Fig. S1A1 in the supplementary material). Some variation in brain structure and lesion sizes was present because of variation in the histological processing of the brains.

The locations of the lesions are illustrated in Fig. S2 in the supplementary material. For clarity we have separated the lesions into those within the CC (see Fig. S2B in the supplementary material) and those outside (see Fig. S2C in the supplementary material). The placement and size of dots representing lesions was done by aligning a scaled photograph of the lesion with a diagram of the brain (see Fig. S1A in the supplementary material). Thus the dots are both indicative of the relative size and location of the lesion. In subsequent figures, the locations of the brain lesions are color coded to reflect the behavioral change relative to the pre-lesion trials. In these data, the proportion of trials in which a specific behavior was exhibited was determined for each individual pre-lesion and post-lesion. The change in behavior in each individual was calculated. The significance of this change was determined using a χ²-test on raw data in a 2×2 contingency table [similar to Ridgel et al. (Ridgel et al., 2007)]. Once determined, the significance level was used to color code the lesion sites on the diagrams. The behavior could increase by one standard deviation (orange), two standard deviations (red), it could decrease by one or two standard deviations (green and blue, respectively), or it could stay the same (black).

It should be noted that the brain structures and lesions are three dimensional. Most lesions only damaged a single brain structure and were contained within that structure. However, there were cases in which the lesion exited a structure. In most cases this did not involve damage to any other known brain structures. In the small number of individuals where more than one brain structure was damaged, lesion location was assigned based on the structure receiving the majority of the damage. No lesions were found that extended beyond the brain.

Behavioral tests were performed under two lighting conditions: referred to simply as ‘light’ and ‘dark’ (see Harley et al., 2009). In the ‘light’ condition, the experimental arena was illuminated to 350 lx incident light (2800 lx reflected light; Gossen Luna-Pro light meter, Nürnberg, Germany) by fluorescent lights and two infrared (IR) strobe lights, which were synchronized to video cameras (Infrared Strobe II, AOS Technologies AG, Baden, Switzerland). This lighting condition approximated an overcast day. In the ‘dark’ condition the room was only illuminated to 0.17 lx, approximating light levels during a quarter moon (Falkenberg and Clarke, 1999). Cockroaches were given 1 h to adapt to room and lighting conditions before testing under these conditions began. Although cockroaches are nocturnal insects that are adapted to see in low-light conditions, we do not believe that our ‘light’ condition is so bright as to blind them. However, it is entirely possible that they were better able to see in the ‘dark’ condition than in the ‘light’ condition. Animals were entrained to a 12 h:12 h light:dark cycle at 27°C for a minimum of 48 h prior to testing. Tests under light conditions were performed during the last 3 h of the light cycle and in the dark during the first 4 h of the dark cycle, as cockroaches are most active during this time (Gunn, 1940; Tobler and Neuner-Jehle, 1992).

Cockroaches faced up to four different obstacles, which were put in random order at the beginning of the experimental period (see Fig. S3 in the supplementary material). To accurately assess abnormalities, individuals performed multiple trials on each obstacle during both pre-lesion tests and post-lesion tests. At the beginning of each trial the cockroach was placed in a release cage measuring 5 cm wide×5 cm high×9 cm long. The release cage was opened to allow the cockroach access to the arena. Once the trial was finished, the cockroach was returned to the release cage.

The block (50 mm×11.7 mm) was a simple obstacle that the cockroach would climb over (see Fig. S3A in the supplementary material). Analysis of behavior at the block was performed using methods previously described by Harley et al. (Harley et al., 2009). Trials on this obstacle were completed when both of the cockroach's front feet reached the top of the block. Each cockroach performed this task a minimum of twice and a maximum of four times under each lighting condition.

‘Climbing attempts’ were counted as easily identifiable pronounced front leg movements that were directed toward the top of the block. Since the behavior stopped once a successful attempt was made, the number of ‘climbing attempts’ reflects the success rate of the behavior. Climbing strategies have been described and classified previously by Harley et al. (Harley et al., 2009).

The distance from the block was measured as the horizontal distance from the front of the pronotum (the cuticular shield on the dorsal surface of the cockroach's thorax) to the edge of the block.

The shelf provided an obstacle that the cockroach could either climb over or tunnel under (see Fig. S3B in the supplementary material) (Harley et al., 2009). Trials on this obstacle were completed when both of the cockroach's ‘feet’ reached the top of the obstacle (climb) or when the individual's thorax passed under the shelf (tunneling) (Harley et al., 2009). Each cockroach performed this behavior three to four times under each lighting condition.

Analysis of behavior was performed using methods previously described by Harley et al. (Harley et al., 2009). To categorize the behaviors as an increase or decrease they had to show a change of more than two standard deviations from the pre-lesion value for the population. The significance of new behaviors was not tested as they had never occurred in pre-lesion trials during this or previous studies (Harley et al., 2009).

The turning arena was a U-shaped track similar to that used by Ridgel et al. (Ridgel et al., 2007) (see Fig. S3C in the supplementary material). This arena exploits the fact that cockroaches will turn away from the point of antennal contact with a wall. The outer walls of the U were 18 cm each, with the inner walls placed such that a channel 6.5 cm wide was present allowing the cockroach to move freely. This particular arena can be put over a piece of glass with a mirror located underneath it at a 45 deg angle. This provided a view of the cockroach's underside, which enabled us to measure the turn angle without moving the cameras. Cockroaches performed at least three turns in each direction. In cases where it was deemed necessary, more turns were recorded to ensure accuracy of behavior classification.

Analysis of turn angle was performed by marking the cockroach's position when its antenna first touched the wall. Then the body axis (a line from the cockroach's head to its posterior) was compared with the original position to obtain an angle (see Fig. S4 in the supplementary material). This angle was assessed at each step with the middle leg following antennal contact. This allowed us to avoid discrepancies in the data due to walking speed. Middle legs were used because timing of their movements was consistent during turns. Front legs, however, can be used for searching or occasionally in an attempt to climb or dig under the walls of the arena. During these behaviors, middle legs would not move, making it easy to exclude them from our analysis. In pre-lesion tests, nearly all individuals reached a 30 deg turn in eight middle-leg steps. Thus, a behavior was labeled ‘failure to turn’ if the cockroach did not turn at least 30 deg within eight steps of antennal contact (see Fig. S4D in the supplementary material). Although cockroaches usually turn away from antennal contact with a wall, occasionally, they turned in the opposite direction. We defined ‘turning in the wrong direction’ as a turn of 30 deg or greater toward the wall (see Fig. S4C in the supplementary material). Whether or not the mean turning direction was significantly different was determined using the Watson–Wheeler F-test in the Oriana software package (Kovach Computing Services, Anglesey, Wales, UK).

The arena used in this test was a large Styrofoam block (19 cm×19 cm; height × length) (see Fig. S3D in the supplementary material). The cockroach would climb up the wall (vertical surface) and then when it reached the top, it would transition to walking on the horizontal surface. With this obstacle, we focused upon the cockroach's ability to transition from the vertical to horizontal surfaces. Plexiglas walls (placed 6 cm apart) limited lateral movement to keep the cockroach in focus during filming. Cockroaches surmounted this wall three times in each lighting condition. Before the lesion one cockroach failed to transition, resulting in it falling off of the wall. As that behavior was extremely rare before the lesion (one out of 282 pre-lesion trials) we examined this deficit in terms of individuals that exhibited it in one or more trials instead of by its overall prevalence in trials.

For the block, turn, and shelf tasks, the prevalence of an abnormality was calculated by examining the number of trials in which it occurred for each individual and then averaging those values over the entire population. This number is represented in box plots throughout the results section.

Each behavior was recorded using two high-speed digital video cameras (or one for the turning behaviors; PhotronTM, San Diego, CA, USA). One of these cameras was located on the side of the arena. The other was located above the arena for all obstacles except the turn. All videos were taken at 60 frames per second. Video records were saved directly to a PC computer hard drive for subsequent analysis with Winanalyze motion analysis software (Mikromak, Berlin, Germany). This software enabled us to determine the distance of the cockroaches from the block, in addition to changes in body angle.

Electrolytic brain lesions were performed on 47 cockroaches that were subsequently examined for behavioral deficits. The behaviors that were examined were chosen because several had been described in detail previously [i.e. block and shelf climbing (Harley et al., 2009), turning (Jindrich and Full, 1998; Mu and Ritzmann, 2005)] and for most behaviors, the effects of more gross lesions that implicated the CC in control had been described previously [i.e. effect of circumoesophageal connective lesion on block climbing and transition from a taller wall (Ritzmann et al., 2005) and effect of gross lesions within the CC on turning (Ridgel et al., 2007)]. To account for variation among individuals, behaviors after the lesion were only compared with observations on the same individuals prior to generating the lesions.

In naïve cockroaches, the point at which rearing movements commence has important ramifications for climbing ability (Harley et al., 2009). Climbing attempts can either be successful, where the ‘foot’ reaches the top of the block, or a miss, when the ‘foot’ fails to reach the top of the block. Before the lesion, 72% of climbing attempts were successful. This decreased to 53% after the lesion. In a previous study, all misses occurred at distances greater than 5 mm from the block and surgically shortening the antennae actually led to an increased success rate, as climbs were initiated closer to the obstacle (Harley et al., 2009). In this study, initial climbing movements in pre-lesion trials occurred at an average distance of 8.3 mm from the block. After the lesion, climbing commenced closer to the block at an average distance of 5.3 mm (Fig. 1A,B). Despite the fact that one might expect this post-lesion decrease in distance from the block to be accompanied by an increase in success, after the brain lesions, the benefit of being close to the block was lost. Successful attempts and misses were equally likely to occur at each distance from the block (Fig. 1B). Thus, brain lesion resulted in more attempts close to the block, but, unlike antennal shortening, success rate decreased.

This observation led us to assess whether or not this decrease in success was associated with lesions to particular regions of the brain. We used the average number of climbing attempts per lesion as a measure of climbing success. A positive number indicates an increase in the number of attempts per trial, i.e. a decrease in climbing success (Fig. 1C). We found a significant change only in individuals with lesions in the lateral accessory lobe (LAL; P<0.005, χ²-test). A small increase in attempts per trial [1 standard deviation of the pre-test population (s.d.= 0.45)] occurred for individuals with lesions to the ACT. However, that was not statistically significant.

Even though the number of attempts occurring at less than 5 mm from the block (a region where 96% of pre-lesion climbing attempts were successful) increased from 24% to 53%, it was not met with an increase in success. This led us to look closer at attempts occurring at less than 5 mm from the block. When these data were broken down by the location of the lesions within the brain, there was a significant increase in attempts at or below 5 mm for individuals with damage to the ACT (P<0.001), MB (P<0.05), FB (P<0.005) and LAL (P<0.001 ANOVA; Fig. 1D). Lesions to EB or other brain regions did not result in a significant change.

After the lesion, many individuals failed to respond to antennal contact with the block, resulting in them hitting their head on the obstacle. This behavior suggests that the subjects have difficulty detecting the block, which could result from sensory processing deficits. Our previous study demonstrated that evaluation of blocks by cockroaches is carried out primarily by antennal contact, but is influenced by ambient light conditions (Harley et al., 2009). We, therefore, tested the intact and lesioned cockroaches on this parameter under both light and dark conditions. The incidence of head contact increased in individuals with lesions to the FB (P<0.001) and LAL (P<0.005), regardless of ambient light conditions. Similar delayed responses have been described previously as an effect of lesions to the CC in Drosophila (Ilius et al., 1994). By contrast, individuals with lesions in the PB or ACT only showed a significant increase in head contact when tested in the light (P<0.05; Fig. 2). Individuals with lesions in the EB or MB showed no significant change under either lighting condition.

Cockroaches can use a number of strategies to climb a block (Harley et al., 2009; Watson et al., 2002a). The most common strategy used by intact cockroaches is controlled rearing (70% of pre-lesion attempts) (Fig. 3A). During this type of climbing, the cockroach raises the front of its body so that when it swings its front legs, the tarsi will land on the top of the block. In post-lesion tests, controlled rearing was only used in 33% of attempts. This decrease was significant in individuals with lesions to the ACT (P<0.05), FB (P<0.05) and LAL (P<0.001; Fig. 3B). After the lesion, controlled rearing gave way to a large increase in elevator and brute force strategies (Fig. 3A). During the elevator strategy, the ‘foot’ misses, often reaching the face of the block. It then swings progressively higher until it reaches the top of the block. Before the lesion this strategy only accounted for 11.6% of attempts. Afterwards, its use nearly doubled to 21.5%. The increase in elevator strategy was statistically significant in individuals with LAL and FB lesions (P<0.01; P<0.05, respectively; Fig. 3C). Another rare strategy, the brute force strategy occurs when the cockroach pushes forward after contacting the block until its body slides up and over the block. This strategy was hardly present in pre-lesion trials (0.5%), but its use increased after the lesion to 27.3% of attempts. In a previous study, similar increases were found in cockroaches after antennectomy (Harley et al., 2009). This suggests an inability to receive or respond to antennal information. This increase in the use of brute force was statistically significant in individuals with lesions to the ACT (P<0.01), MB (P<0.05), FB (P<0.01) and LAL (P<0.01; Fig. 3D).

The shelf provides a complex behavioral task; the cockroach can either climb over or tunnel under it. In naïve cockroaches the primary determinant is the manner in which the antennae contact the shelf. Climbing occurs when both antennae touch the top of the shelf, and tunneling occurs when both antennae touch the underside (Harley et al., 2009). However, in the light, tunneling predominates; whereas in the dark climbing and tunneling have an equal probability of occurring. Our pre-lesion results were consistent with that earlier study. However, after the lesion the overall incidence of climbing decreased, so that there was no longer a significant difference between the two light conditions (Fig. 4A).

The role of antennae in the choice to tunnel or climb was established through the development of ethograms that described the transitions among various sub-behaviors (Harley et al., 2009). To assess whether or not similar transitions occurred in individuals with brain lesions, we created four new ethograms of this behavior for pre-lesion and post-lesion trials run under the two light conditions. Despite the striking similarity between the pre-lesion shelf ethogram and that of naïve individuals [compare Fig. 4B,D with figure 2B,E in Harley et al. (Harley et al., 2009)], changes were noted in the ethogram after the brain lesion (Fig. 4C,E). These changes included the addition of new transitions (Fig. 4C,E shown in red), increases in rare transitions (shown in yellow) and decreases in other transitions (shown in blue). This breakdown in the normal sequence of behaviors at a shelf may be indicative of changes in planned movement and/or changes in the ability to sense the shelf. Individuals exhibiting these behaviors all had lesions to the CC or the ACT (Fig. 4F,G,H).

Similar to the response to a shelf, where antennal contact directs the decision to climb or tunnel, antennal contact with a wall induces turning. Furthermore, antennal contact also influences turning direction, as cockroaches will typically turn away from antennal contact with a wall. Before the lesion most cockroaches consistently turned in response to antennal contact with a wall (98% of pre-lesion trials), but this failed to occur in 40% of the post-lesion trials. This failure resulted in the cockroach hitting its head on the wall, after which the cockroach would execute a turn. Such ‘delayed turning’ behavior increased significantly under both lighting conditions in individuals with lesions to the MB (P<0.001), PB (P<0.001), FB (P<0.001), EB (P<0.001) and LAL (P<0.001) (Fig. 5A,B). Only two individuals with FB lesions failed to show this behavioral change. Both of which had lesions to the center of the FB (Fig. 5C,D).

Delayed turning is not the only turning abnormality noticed after the lesion. Some individuals completely failed to turn or turned in the wrong direction. During failed turns, unlike in delayed turning, after hitting their head on the wall individuals never made a turn greater than 30 deg. In pre-lesion trials, all individuals turned at least 30 deg over the course of eight steps (Fig. 6A), whereas post-lesion, some individuals failed to turn at least 30 deg in either direction (Fig. 6). The increase in failure to turn was present in individuals with LAL lesions (P<0.001; Fig. 7A).

Some individuals managed to turn at least 30 deg, but did so in the wrong direction. This behavior was significant in individuals with FB lesions (P<0.05; Fig. 7B). More specifically, individuals with lesions in the lateral regions of the FB often turned in the wrong direction (P<0.05), whereas those with lesions to the middle of the FB did not (Fig. 6D and Fig. 7B inset).

The combination of these turning abnormalities resulted in quantifiable changes to the mean turn angle (Fig. 6). Individuals with lesions outside the CC showed no significant change in mean turning angle. By contrast, individuals with LAL lesions (P<0.001) exhibited a significant change. This change approached, but did not reach, significance for individuals with FB lesions. However, when the FB was separated into its medial and lateral components, changes in turn angle were significant in individuals with lesions to the lateral, but not the medial FB (P<0.05; P>0.2, respectively).

These directional turning deficits may be limited to a single direction. For this reason we examined whether or not turning abnormalities occurred when turning in the direction ipsilateral to the side of the lesion or contralateral to it. In other words, we examined whether an individual with a lesion on the right side of a given neuropil had deficits that were restricted to right (ipsilateral) and/or left (contralateral) turns. Individuals with LAL lesions exhibited significant changes in mean turn angle for both ipsilateral and contralateral directions (Fig. 8B; P<0.0001, P<0.0001). These mean turn directions were not significantly different from one another. By contrast, individuals with lateral FB lesions only exhibited significant changes in mean turn angle during contralateral turns (Fig. 8A; P<0.0001). The turning angle for ipsilateral turns was not significantly different from pre-lesion trials (Fig. 8A; P<0.2). Furthermore, all but one of the abnormal turns was in the contralateral direction. Thus, an individual with an FB lesion on the right side would exhibit abnormalities when turning left but show no abnormalities when turning to the right (Fig. 8). Similarly, individuals with PB lesions also exhibited abnormalities only in contralateral turns. By contrast, individuals with EB or LAL lesions showed abnormalities in both directions. Interestingly, individuals with LAL lesions that exhibited turn abnormalities in a single direction only did so during ipsilateral turns (Fig. 8C).

Although individuals with lateral FB lesions exhibited turning abnormalities in the direction contralateral to the lesion, many contralateral turns were normal (Fig. 8A). Is this because certain individuals have an inability to turn contralaterally whereas others turn normally? Or is it that each individual turns abnormally in some trials, but normally in others? Of the seven individuals with lesions to the lateral FB, five exhibited turning abnormalities in contralateral turns and two turned normally in all trials (Fig. 9). One of those (cyan) turned in the correct direction on each trial, but in one trial turned an unusually large angle. One of the five abnormal individuals (dark green) failed to execute any normal turns in the contralateral direction (Fig. 9C). However, most of these individuals produced both normal and abnormal turns.

When climbing tall obstacles, cockroaches must transition from a vertical surface to a horizontal surface. This behavior requires a degree of downward body flexion such that the front ‘feet’ will reach the horizontal surface (Ritzmann et al., 2004) (Fig. 10A). Cockroaches with bilateral circumoesophageal connective lesions no longer generate this flexion (Ritzmann et al., 2005). Similarly, if this flexion is prevented by a brace, this task becomes difficult and may result in falling.

Before the lesion, only one out of our 47 individuals fell off this obstacle (represented in pink in Fig. 10B,C). After the lesion, this number increased to five (yellow bars in Fig. 10B). Four additional individuals displayed moderate behavioral deficits resulting from delayed flexion (white bars in Fig. 10B). Individuals that fell off of this obstacle while attempting to transition had lesions to the FB, EB or LAL (Fig. 10C). Those with moderate deficits had lesions within the same region, with the exception of one individual that had a lesion to the ACT.

As animals move through complex natural environments, they must be able to alter motor patterns as they encounter barriers in their path. Even small changes in terrain may require postural adjustments (Watson et al., 2002a) which are associated with changes in the pattern of motor activity in the legs (Watson et al., 2002b). When negotiating small barriers, an animal may simply place its front legs on top of the object through its normal swing movements. Increases in motor activity would then occur through reflexes limited to the thoracic ganglia (Watson et al., 2002b; Watson et al., 2002a). When running at high speeds, insects are able to negotiate more substantial objects without active adjustments (Koditschek et al., 2004). However, during most of a cockroach's life, it deals with barriers at slow speeds when a larger barrier requires an active adjustment of posture and motor activity.

Directed changes in movement typically take advantage of sensory structures located on the head and require connection with brain regions that process that information and use it to formulate appropriate descending commands that ultimately alter movement patterns. Although much of the walking motor program in insects exists in the thoracic ganglia, removal of descending influences results in the inability to modify behavior for complex tasks (Graham, 1979; Ridgel and Ritzmann, 2005; Ritzmann et al., 2005; Roeder, 1937). It has been suggested previously that the central complex (CC) is involved in motor planning (Huber, 1960; Okada et al., 2003; Strausfeld, 1999). Indeed, insects that have had large lesions of the CC show difficulty turning, climbing, adjusting step length, creating calling song, and in performing flight control (Ilius et al., 1994; Ridgel et al., 2007; Strauss and Heisenberg, 1993; Weinrich et al., 2008). Within this study, we noted striking changes in response to a shelf which may be a further indication of such changes in motor planning (Fig. 4). However, the involvement of individual structures within the CC in complex obstacle negotiation tasks was unknown. Here we examined deficits in obstacle negotiation behaviors in individuals with lesions to individual neuropils within the CC and surrounding areas. We compared the involvement of these structures in several different behaviors, each of which had previously been shown to have deficits associated with gross lesion of the CC (Pick and Strauss, 2005; Ridgel et al., 2007). In addition, we were able to investigate how different regions of a single structure, the FB, are involved in these behaviors.

In the results section, we described our data according to the behaviors tested. Here, we will re-examine these effects but now from the standpoint of the damaged neuropil. In so doing, we will take advantage of the neuronal organization previously described in detail for polarized light processing in the CC of locusts (Heinze and Homberg, 2009; Homberg, 1987; Homberg, 2004b). It is probable that this organization may be generalized for processing other sensory modalities in both locusts and other insects that rely less on polarized light, such as the nocturnal cockroach. Thus, this organization can serve as a reasonable basis for examining other modalities in various behavioral contexts.

The PB receives visual information and has recently been shown in locusts to possess a topographically organized map of polarized light orientation (Heinze and Homberg, 2007; Homberg, 2004a). However, this structure is still found in blind cave beetles (Ghaffar et al., 1984), suggesting that it is not solely involved in processing visual information. When this structure was lesioned, individuals exhibited behavioral deficits that could be associated with problems obtaining sensory information or problems interpreting it. Examples of this include delayed turning (Fig. 5) and head contact with the block (Fig. 2). The latter only occurred in the light indicating that it is possible that these individuals may have been able to compensate in the dark, perhaps because their eyes are better adapted to seeing under those conditions. They also exhibited all three abnormalities when presented with a shelf (Fig. 4). The response to a shelf is a task that is directed by the manner in which the antennae contact it (Harley et al., 2009). The behavioral deficits noted here are consistent with previous observations that the PB is involved in sensory processing. The insects remained capable of the physical tasks of climbing and turning, but exhibited problems in climbing accuracy and in localizing obstacles.

Although it seems likely that the PB is involved in sensory processing, how it is involved is a bigger question. The response to a shelf and turning behaviors require integration of information from the right and left antennae. It is possible that after the PB lesions this integration is faulty, leading to these abnormalities. Interestingly, PB lesions resulted in deficits in contralateral turns (Fig. 8), supporting anatomical evidence that sensory information is crossing over upon entering this brain region (Heinze and Homberg, 2009). A similar involvement of the PB in the integration of information from the hemispheres of the brain was noted with visual stimuli (Bausenwein et al., 1994). Such connections may explain why some lesions to this structure result in behavioral abnormalities, whereas others do not. Alternatively, it is possible that observed sensory information within the PB is used in motor planning. This would result in the occasional generation of wrong behaviors or delayed onset of behavior. Further study is needed to examine this possibility.

Individuals with FB lesions exhibited a wide variety of behavioral deficits when presented with the various obstacles (Fig. 11). During climbing and turning tests, they often showed a delayed response to the obstacle, resulting in collisions (Figs 2, 5). A similar delay in behavioral response to obstacles was previously noted in genetic lesions of the CC in Drosophila (Ilius et al., 1994). After collision with the block, FB-lesioned cockroaches proceeded to climb; however, this was done in a less successful manner than in the pre-lesion trials (Fig. 1). Furthermore, brute force and elevator climbing strategies both lack targeting of the limb toward the top of the block (Fig. 3). In the case of the elevator strategy the limb is eventually targeted toward the top of the block, but only after several re-targeting events. This suggests that these attempts are not guided in the same manner as controlled rearing attempts – they may represent a trial-and-error strategy. By contrast, after their collision with the turning arena some individuals failed to turn or turned in the wrong direction, whereas others exhibited relatively normal turning behavior (Figs 6, 7).

The extent of turning deficits in these individuals was related to the lesion location. Individuals with lesions to the center of the FB exhibited no or slight turning deficits while those with lateral FB lesions often turned in the wrong direction or failed to turn. Interestingly, these deficits were mostly limited to an inability to turn in the direction contralateral to the lesion (Fig. 8). This observation means that individuals with lesions to the right side of the FB showed an inability to turn left, and vice versa, but typically had little problem turning toward the side with the lesion.

However, only a subset of these contralateral turns was abnormal (Fig. 9). If the FB or a small part of it completely controlled turning behavior, then we would expect abnormalities on nearly every contralateral turning attempt. Whereas, if it modulates turning behavior, individuals will only be affected under a certain subset of conditions resulting in a few but significant number of abnormal turns per individual.

It does appear that the lateral and medial FB have different roles in these behaviors. During transitions on the wall barrier (Fig. 10) and turning (Fig. 7), the instance of abnormality was higher in individuals with laterally located FB lesions than in those with medial lesions. This difference between individuals with medial and lateral lesions was not present in block or shelf climbing (Figs 1, 2, 3, 4). This suggests that different regions within the same structure can serve unique roles in different behaviors. In one sense, turning and climbing represent extreme differences in motor control. Turning requires a switch from symmetrical leg movements to decidedly asymmetrical movements as a result of an asymmetrical sensory event (e.g. contact by one antenna) (Mu and Ritzmann, 2005). By contrast, climbing involves bilaterally symmetrical changes in leg movement (Watson, 2002a). Right and left legs still perform the same movements but in a different posture (inclined upward). Likewise, climbing typically arises from symmetrical sensory cues (e.g. contact with both antennae). Thus, our results suggest that asymmetrical inputs and resulting asymmetrical outputs are better controlled by lateral columns of the FB whereas the medial columns are receiving symmetrical inputs from sensors on both sides of the insect's head, and generating similar motor outputs on both sides. This distinction is further supported by the observation that FB lesions only affected turning to the side contralateral to the affected region, another asymmetry. Interestingly, comparative studies have revealed that insects that turn with asymmetrical leg movements (e.g. cockroaches) have more highly developed CC neuropils than those that maintain symmetry during turning (e.g. water striders) (Strausfeld, 1999).

It does not appear that this same gradient is present along the EB or the PB; however, our data set contains fewer lesions in those regions and more data is needed to state that definitively. The regionalization of the FB is further suggested by neuroanatomical data (supplementary material Fig. S5). Each FB column receives a connection from the left and the right protocerebral bridge. Lateral FB columns receive inputs from one lateral PB column and one medial PB column, whereas medial FB columns only receive inputs from medial PB columns. However, this connectivity pattern alone cannot explain all of the abnormalities that are present. It is probable that the FB is integrating information from multiple brain regions and is involved in modification or formation of motor commands rather than simply sensory information processing. This situation would result in the same individual only showing abnormality in a subset of trials. In addition, this modification of ongoing behavior would probably be the result of integrated sensory and motor information resulting in abnormalities in both aspects of behavior, as was seen here.

Although, individuals with EB lesions exhibited no behavioral deficits on the block (Figs 1, 2, 3), they did when turning (Figs 6, 7, 8). In contrast to individuals with FB lesions, these individuals showed bi-directional turning deficits. It is possible that behavioral differences are related to how these brain regions connect to the PB. In locust, each FB column receives information from columns located on the right and left side of the PB. By contrast, odd numbered EB columns are connected to the left columns of the PB, and the even numbered columns to the right PB (Heinze and Homberg, 2009; Muller et al., 1997; Strausfeld, 1999; Williams, 1975). This pattern of connections suggests that lesions to the EB result in destruction of large amounts of information from both the right and left columns (see Fig. S5 in the supplementary material). With no additional connections between the specific area of the PB and another region of the EB, this information is lost completely. This hypothesis could also explain abnormalities in response to a shelf (Fig. 4).

This connectivity could also explain the variability between individuals with brain lesions. The EB and PB are connected by a multitude of cell types, which differ in where they connect the structures and in their degree of crossover. This means that lesions to the upper part of the EB may have a different effect to lesions to the lower regions of the EB (Heinze and Homberg, 2009). Future work is necessary to determine whether this is so. Regardless of this connectivity, it appears that the EB and FB have different roles in these behaviors.

Individuals with LAL lesions exhibited striking abnormalities in all obstacle negotiation behaviors. This region contains connections leading to and from the CC as well as the ventral nerve cord (Heinze and Homberg, 2009; Homberg, 1994). Lesions to this region would essentially cut off downstream connections (see Fig. S5 in the supplementary material) from the CC and would, therefore, be expected to result in deficits that are similar to those of an individual with large scale lesions to the CC (Ridgel et al., 2007; Ridgel and Ritzmann, 2005). As a linkage between the CC and thoracic ganglia, it is not surprising that individuals with LAL lesions have profound deficits in most if not all obstacle negotiation behaviors. Consistent with this notion, deficits associated with LAL lesions were more absolute than those associated with other neuropils. For example, rather than making wrong turns, LAL-lesioned cockroaches failed to turn (Fig. 7A).

There is not a simple answer to how the brain controls these complex behaviors, but rather, the control varies for each obstacle-negotiation task. For example, whereas lesions to the EB had little impact on block climbing behavior, they were shown to have a profound effect on all aspects of turning behavior (Figs 5, 6, 7). Furthermore, the control of these obstacle-negotiation behaviors seems to have some specificity for different elements of behavior. Individuals with lesions to the ACT, FB or LAL were close to the block when they started climbing (Fig. 1) and employed different climbing strategies (Fig. 3), but success of climbing attempts was only affected in individuals with FB or LAL lesions (Fig. 1). In some cases, this specificity even extended to ambient lighting conditions. This was the case in individuals with lesions to the ACT, PB or EB that only exhibited a significantly higher rate of collisions with the block in the light (Fig. 2). Despite the similarity between this behavior and the collision during delayed turning, the latter was not affected by ambient lighting conditions (Fig. 5). It is possible that this effect was amplified by the failure of some individuals to turn (Fig. 7), whereas failure to climb did not occur.

It should be noted that, even where lesions generated profound deficits, the cockroaches were still capable of making correct adjustments. For example, individuals with lateral FB lesions still produced normal directional turns in two thirds of the contralateral turning trials. This may suggest that the role of the FB in turning is more modulatory than directive. That is, even with lateral FB damage, the cockroach is still capable of using antennal contact information to direct a turn. One explanation would be that tactile stimulation of an antenna is sufficient to direct a turn in the opposite direction through pathways that may pass through the LAL. However, given numerous additional factors processed within the CC (such as ambient light conditions), the FB could alter that action to better match immediate conditions. Damage in that region could result in inappropriate alterations but only in some trials. One lesion that was well inside the lateral edge of the FB did generate wrong turns in all three trials, but that was an exception. This lesion may have been in a more critical locus or more extensive than the others. Definitive conclusions will require a larger set of FB lesions.

Lesion experiments have inherent limitations. It is possible that the surgical process created unknown damage. However, a previous study which employed the same technique to open and close the head found that the procedure itself did not increase turning abnormalities over levels seen in pre-lesion tests (Ridgel et al., 2007). Furthermore, none of our nine lesions that were outside the CC, mushroom bodies and ACTs resulted in significant behavioral changes. Another possibility is that insertion of the electrode caused damage. This is supported by the observation that individuals that had had the probe inserted without electrical current exhibited similar abnormalities to those with lesions to the same brain region. Most of these abnormalities were less severe than they were in lesioned individuals. However, we cannot rule out the possibility that the probe passing through the brain caused additional damage, resulting in behavioral abnormalities. Severe deficits such as inability to turn, difficulty climbing when close to the block, and falling off the climbing wall were rare in the entire lesioned population. However, these abnormalities are consistent in individuals with lesions to certain brain regions. Together, this suggests that these behavioral deficits are probably a result of the lesions, rather than a complication of the technique.

To fully understand the role of the CC in behavior we must use a variety of techniques. No one experimental method is without consequence. Neurogenetic lesions have become a powerful tool for studying the neural basis of behavior (Strauss, 2002) but these procedures can also have unspecified effects. The unknown consequences are probably different for each technique, suggesting that by combining studies that employ genetics, mechanical and chemical lesions, reversible blocks, as well as intracellular and extracellular recording methods, we may overcome the drawbacks of any single technique.

We would like to thank Yidi Xu, whose undergraduate project at CWRU provided inspiration for this project. Additionally we would like to thank Angela Ridgel and Alan Pollack for valuable technical assistance, Mark Willis, Debra Wood, Hillel Chiel, Roger Quinn, Michael Branicky for their comments on the original manuscript, as well as Solomon Awe and Brittany English who assisted with data acquisition and analysis. The work reported here is part of a study supported by the National Science Foundation under grant no. IOS-0845417 and the AFOSR under grant FA9550-07-1-0149 both to R.E.R., and NSF IGERT training grant DGE 9972747 that provided graduate support to C.M.H. We thank the Company of Biologists for a travel grant to C.M.H. to present this work at the Society for Neuroscience annual meeting.