The insect central complex

Transcription

1 Primer The insect central complex Daniel. Turner-Evans and Vivek Jayaraman Hordes of tourists flock to Washington, D.C. every spring to see the cherry trees blossom. Once in the city, they must find their way to the Tidal asin where the Japanese trees grow. Fortunately, a number of visual landmarks can help them to navigate. In 191, the United States Congress passed The Height of uildings ct, limiting the elevation of commercial and residential structures in D.C. to 13 feet. Thus, the 555-foottall Washington Monument often looms large against the horizon, serving as an anchor point to help set the tourists sense of direction. Once their heading is set, they can lose sight of the monument behind buildings or groups of tall Scandinavian visitors and still use their internal compass to navigate to the asin. This compass keeps track of their paces and turns and updates their sense of where they are and where they need to go. Yet while their heading informs their actions, it does not dictate them. Tourists who have been to D.C. in the past can, for example, use remembered views to alter their routes to avoid crowds. On an even finer scale, their leg movements also depend on their current state they might increase the frequency and length of their strides if hunger pangs compete with their desire to see cherry blossoms, for example. The way in which these disparate cues and motivations influence exploration is a neuroscience mystery across creatures large and small. Navigation, whether by a human tourist or a roving insect, requires an exquisite coupling between the senses and the body s movements. This coupling, also referred to as sensorimotor integration, is flexible enough to allow animals to incorporate past experience and current context into the decision-making process. In insects, compelling behavioral and physiological evidence suggests that a highly conserved brain region called the central complex is key to such adaptive sensorimotor processing. 2 µm This region is linked to sun-compass navigation in Monarch butterflies and locusts, to the use of polarized moonlight to maintain straight paths by nocturnal dung beetles, to visual landmark orientation and place learning in fruit flies, and to antenna-guided obstacle navigation in cockroaches. Thus, the central complex provides a fascinating and accessible substrate for understanding some of the fundamental integrative neural processes underlying navigation. Herein, we provide an overview of anatomical, behavioral and physiological results that suggest how this brain region links sensation and action in insects. Conserved complexity The central complex is aptly named. Central to the arthropod brain, it unites inputs from the right and left in a few central structures, in relative contrast to the more widely separated, paired structures of the rest of the brain and it is indeed intriguingly complex! The structures of the central complex are woven together by dense layers and columns of neurons. These neural threads wind their way through the four heavily interconnected structures shown in Figure 1: the protocerebral bridge, ellipsoid body, fan-shaped body, and the paired noduli. The ellipsoid body and the fan-shaped body F/CU E/CL NO F/CU together are termed the central body, with the ellipsoid body occasionally referred to as the central body lower, and the fan-shaped body as the central body upper. The regions are connected to several paired support areas including the lateral accessory lobe and the bulb. The lateral accessory lobe is innervated by many neurons, including descending neurons that carry signals down to the ventral nerve cord, which controls the legs and wings. The four central complex structures are found across an array of arthropods, and the conservation of these structures across 5 million years of arthropod evolution speaks to their importance. Though their form varies from species to species, individual neurons that innervate the structures are conserved to a remarkable degree. Figure 1 shows one such example, a protocerebral bridge fan-shaped body lateral accessory lobe neuron found in the locust, and its likely counterpart in the fruit fly. While the roles of the individual regions are not yet well understood, an increasing number of exciting studies are shedding light on how they work together to take in sensory information from the world, integrate it with the current state and past experience of the animal, and output motor commands to direct the animal s movements. E 1 µm E 2 µm F F Figure 1. Similarity of the anatomy of the central complex across species. () Overview of the central complex and support structures: the protocerebral bridge (), fanshaped body (F) or central body upper (CU), ellipsoid body (E) or central body lower (CL), noduli (NO), and lateral accessory lobe () in the locust (left, modified with permission from Pfeiffer and Homberg, 214 by nnual Reviews, top, from M. Müller et al with permission of Springer) and cockroach (bottom, with permission from R. Loesel et al. 22). Several key anatomical features of the central complex are conserved across insects and crustaceans. () columnar F neuron (in red) from the locust (top, from Heinze and Homberg, 27, reprinted with permission of S) and fruit fly (bottom, courtesy T. Wolff and Y. so). 26, R445 R46, June 6, Elsevier Ltd. R453

2 C spikes/s imp/s 27 Maplike representation 18 Light polarization 1 steps Movement after antennal contact 18 Turning angle Compass inputs from land and sky efore making a decision about where to move in its environment, an animal usually gathers information about its surroundings and gets its bearings. Monarch butterflies rely on the position of the sun to maintain their heading while migrating thousands of kilometers, from eastern North merica to an oyamel fir forest in Mexico. In both the Monarch butterfly and another longrange flier, the desert locust, intracellular recordings have revealed neurons throughout the central complex that respond to the polarization direction of a global light source (Figure 2), with polarization patterns that match those of scattered sunlight evoking the strongest responses. Further, the responses of 9 Motor control 27 Motor adjustment Time (s) Pre lesion Post lesion Rotational velocity (deg/s) E D Spatial orientation memory 5 mm Visual place learning Probe trials (cold zone removed) Hot 1. Front, back bars visible Side appears, others gone 3. Side disappears Cold zone some protocerebral bridge neurons create a vector-specific polarization map across the protocerebral bridge, with each of its columns representing a preferred electric field vector orientation. Many of the same neurons also respond to the direction of a strong, singular light source, supporting the idea that the central complex enables sun-compass navigation in these species. Maintaining a straight course is no less vital for nocturnal dung beetles. These beetles must roll a ball of dung from a pile as quickly as possible to avoid potential competitors. They get their bearings using polarized light patterns on a dim night, but are flexible enough to use the angular position of the sun if operating in daylight. This 3.? Figure 2. Several orienting and adaptive navigation behaviors in insects require the central complex. () Intracellular recordings from the F neurons of the locust show tuning to different orientations of polarized light E-vectors across the protocerebral bridge. These neurons thus form a map of polarization directions across the structure. Recording shown is from neuron in Figure 1. (From Heinze and Homberg, 27; reprinted with permission from S.) () When a cockroach s antennae come in contact with a wall ( ), the animal turns away (thick black traces). When lesions are made to the central complex, the roach no longer reliably turns directly away from the wall (thin purple traces) (dapted with permission from the Journal of Experimental iology, Harley and Ritzmann, 21.) (C) Sample neural recording from the cockroach central complex showing activity (orange trace) that precedes and predicts turns (blue trace). (dapted from Martin et al. 215.) (D) In a task called the uridan paradigm, fruit flies are allowed to walk back and forth between two vertical landmarks. When they disappear, and another landmark transiently appears to the side, flies alter their course. When this landmark also disappears, the flies turn back towards the remembered position of the now invisible landmark that was their previous destination. Flies with specific ellipsoid body defects do not demonstrate this short-term orientation memory. (Reprinted from Neuser et al. copyright 28.) (E) Modeled on the Morris water maze for rodents, this task places fruit flies inside a visual arena on an inhospitably hot floor with a single cool spot, whose position is locked relative to the visual scene. Flies learn to use surrounding visual landmarks to find the safe spot after several training trials in which both the spot and the visual scene are rotated together. In the probe trial, there is no cool spot, but flies selectively explore the quadrant containing its expected location. When specific ellipsoid body neurons are silenced, flies are unable to recall the location of the cool spot in the probe trial. (Reprinted Ofstad et al. copyright 211.) (D,E) With permission from Macmillan Publishers Ltd. adaptability is mirrored by its central complex neurons, which, intracellular recordings show, are sensitive to the azimuthal position of a single light source in bright conditions, but switch to being polarization-sensitive in dim conditions. Overall, these observations demonstrate the evolutionary flexibility of the central complex, and suggest that central complex activity in these insects allows them to use a sky compass to determine their orientation in the environment. Navigation does not have to be wholly celestially guided, however. Terrestrial cues can also help an insect figure out where it wants to go, and many insects likely use both, shifting more towards one or the other set of cues depending on their situation. The fruit fly, which can use a sky compass to maintain a constant heading during flight, may rely more on terrestrial cues when it is walking around. Some fly neurons that pass from the bulb to the ellipsoid body process the fly s visual surroundings, with each neuron responding preferentially to specific visual features within a small area of the fly s view. s a group, the neurons tile the fly s visual field and show a bias for vertically oriented features around the fly. Thus, the fly s central complex seems well equipped to process terrestrial landmarks like tree trunks and the stalks of plants, fitting perhaps for a generalist insect that makes trips long and short in varied visual surroundings. Finally, while visual cues provide important information about the layout of the world, other sensory cues can also provide directional information. In the cockroach, which is known to use antennal touches to navigate around walls and under or over obstacles, responses of some central complex neurons are correlated with antennal contact. The blowfly central complex also takes in antennal information, with some neurons showing responses to air puffs delivered to the antennae. In short, the central complex receives multisensory input, which enables insects to orient in their surroundings. nd action? Having gathered information about the insect s surroundings, the brain must then select an action. n array of evidence points to the central complex s ability to serve in this role, incorporating R454 26, R445 R46, June 6, 216

3 sensorimotor information to direct the insect. Perhaps the most direct evidence of central complex involvement in motor control comes from the cockroach. Lesions in central complex structures in the cockroach lead to more variable turning and difficulties in navigating shelf-like objects and climbing. If, for instance, the ellipsoid body is damaged, the cockroaches have difficulty turning away from a wall that they brush with their antennae, as exemplified in Figure 2. While this response could indicate alterations to either sensory integration or motor output, when the activity of central complex neurons was compared to various gait parameters, as shown in Figure 2C, it was predictive of the roach s movements by hundreds of milliseconds. This predictive signal suggests that the central complex may be directly involved in the generation of motor commands and the initiation of movement. Other recordings of the activity of cockroach central complex neurons have shown strong correlations with turning and changes in step frequency. Locomotor deficiencies are also seen in fruit flies that lack subsets of central complex neurons. Damage to the protocerebral bridge, for example, leads to shortened walking bouts or difficulties in crossing gaps, while flies with incomplete ellipsoid bodies seem unable to track a stripe while flying. Further, a number of modulatory neurons infiltrate the structures of the central complex and have been linked to, among other things, the amount of locomotor activity. Dopaminergic neurons with processes in dorsal fan shaped body layers mediate sleep and arousal in fruit flies, suggesting a role for the central complex in gating motor activity. Thus, the central complex seems to have the necessary elements to direct an insect s movements: sensory inputs and predictive motor outputs. ut is it simply a hardwired switchboard linking sensation to action? Integrating sensory and motor, past and present If the central complex were a switchboard, its role would be limited to rigidly transforming specific sensory information into particular actions. To navigate more flexibly, however, an animal needs to combine external sensory information with information about its own current state and past experiences. The central complex has been linked to many such behaviors, suggesting that it is much more than a hardwired router. In fruit flies, for instance, the ellipsoid body has been implicated in the fly s ability to learn to discriminate between different visual patterns during tethered flight, and to employ short-term memory for landmark orientation while walking (Figure 2D). dditionally, in an assay that is reminiscent of place learning tasks in mammals, only flies with intact ellipsoid bodies can use visual information to learn and later recall a safe cool spot within an otherwise aversive thermal environment (Figure 2E). While these studies strongly point to the involvement of the fly central complex in incorporating remembered visual cues during navigation, they give little insight into the underlying machinery. Some hints of a possible mechanism have been observed with calcium imaging in the central complex of tethered flies walking in a simple visual virtual reality environment. population of neurons in the ellipsoid body of these flies maintains a bump of activity, shown in Figure 3, wherein only a local subset of neurons is active at a given time. The ellipsoid body takes the shape of a doughnut in fruit flies, and the bump rotates around it as the fly s orientation to its surroundings changes, with different groups of neurons becoming active sequentially. The activity bump moves much like a compass needle, tracking the fly s orientation in darkness and in the presence of visual landmarks in virtual reality. Landmark tracking by the bump follows sudden angular rotations in the world and adjusts to different visual surroundings and to altered gain between the fly s rotation and the virtual world s movement. Further, the persistence of this representation of angular orientation in darkness suggests that this compass must also integrate angular rotations using information about the insect s selfmotion, whether through proprioceptive feedback or an efference copy of motor commands to its limbs. The bump seems to provide the fly with an abstract representation of its orientation, but it is as yet unclear how this representation is shaped by past experience. So how is the bump formed, maintained and moved? While these questions are still open, the answers may lie in the circuitry of the central complex, illustrated in Figure 3C. The protocerebral bridge consists of a series of dendritic and axonal columns, termed glomeruli, split evenly between the left and the right hemispheres. The total number of glomeruli may vary by species, with, for example, a total of 16 in the locust and 18 in the fruit fly. Most of the neurons coming and going from individual glomeruli cross the midline and make synaptic connections within corresponding columns of either the fan shaped body or the ellipsoid body before continuing on to the other structures, such as the noduli, gall, rubus or lateral accessory lobe. Neurons from matched glomeruli on either side of the protocerebral bridge meet to form the fan shaped body and ellipsoid body columns, leading to roughly half as many columns in the ellipsoid body and fan shaped body as there are glomeruli in the protocerebral bridge. The well-organized loops that seem to connect the protocerebral bridge, ellipsoid body, and fan-shaped body seductively hint at the implementation of a so-called ring attractor within the central complex. The ring attractor, a theoretical concept that is often used to describe the mammalian head direction system, allows for the stable formation of one localized bump of activity within a population of neurons. This bump can move around with sensorimotor input, allowing an animal to maintain a single, continuously updated sense of direction. Figure 3 shows a schematic of a sample network that could implement a ring attractor. This compass network features two main elements: nodes whose activation represents different heading directions (the gray circles), and a specific pattern of recurrent connections between those nodes (the weights). External inputs push the bump of activity around the ring, allowing it to move between nodes. n exciting possibility that needs experimental confirmation is that the neurons that connect the central complex structures may act as ring attractor nodes (Figure 3D). Further, connections between the neurons within the structures may provide the necessary weighted connections 26, R445 R46, June 6, 216 R455

4 of orientation to select and initiate situation-appropriate actions. C Nodes Protocerebral bridge Glomeruli E Weights Inhibitory connection Excitatory connection Neural activity D External inputs, neuromodulators Fan shaped body Ellipsoid body Rubus Gall E Noduli scending signals Predominantly spine-like (putative inputs) Predominantly bouton-like (putative outputs) Descending motor signals Figure 3. Possible ring-attractor-like network in the central complex. () In the fly, visual and motor information is integrated by a population of neurons with dendrites that project to single wedges of the ellipsoid body (see Figure 3D, left, for two examples of such neurons). The dendritic activity of this population is localized to a single bump that moves around the ellipsoid body in concert with the fly s turns. The activity bump may serve as a compass needle, giving the fly its heading. (With permission from Macmillan Publishers Ltd: Nature, Seelig and Jayaraman copyright 215.) () Schematic of a sample ring attractor network, which consists of nodes (gray circles) topologically arranged in a ring with distance-dependent connection strengths between them (red and green colored lines depict weights of the connections). The connectivity of such a compass-like network localizes activity into a single bump, which moves around the ring depending on an animal s rotational movements, much as has been observed in the central complex (see, for example, panel ). (C) Schematic of fly central complex that displays a ubiquitous feature of the region: the organization of most of its substructures into stereotypical layers and columns. Many neurons innervating the protocerebral bridge send projections to specific columns of the fan shaped body or sectors of the ellipsoid body, but no known columnar neurons directly connect the latter two structures. Color- and hatching-matched compartments depict connection rules obeyed by some of the neuron classes that connect subsets of these structures, for example, top row in panel D. Protocerebral bridge columns in gray denote areas untouched by these classes of neurons, but see panel E and bottom row in panel D for examples of other types of neurons. Compartmentalization is reduced in the noduli and gall, which receive projections from neurons innervating the contralateral protocerebral bridge, and appears to be entirely lost in the lateral accessory lobe, a likely output structure along with the posterior slope (not shown). External inputs from many regions, including the bulb and protocerebrum (neither shown), influence the activity of the central complex. (D) Sample neurons that link the protocerebral bridge to the ellipsoid body in stereotyped ways may provide some of the recurrent connections necessary to create a ring-attractor-like network within the central complex. (E) Sample of an interneuron that selectively innervates a single structure (here, the protocerebral bridge). Large asterisks denote putative outputs; small asterisk indicates that the arbor on far side of cell body is often sparse. Such neurons likely also play an important role in shaping the activity dynamics of the network. for a ring attractor network, allowing activity to be localized and moved like a compass needle around a ring of nodes. Such connectivity might be shaped by the many interneurons that weave through the protocerebral bridge, fan shaped body, and ellipsoid body (see, for example, Figure 3E) and neuromodulatory inputs to these R456 structures (Figure 3C) might allow contextual and experience-dependent modification of the connections. Such a network could allow the creation, maintenance and movement of an activity bump that represents the animal s orientation in a variety of environments. The insect could then use this sensorimotor percept 26, R445 R46, June 6, 216 The path forward Many questions remain. The full extent and adaptability of the central complex compass still is not clear. How are the central complex inputs and architecture adapted to incorporate what seem like a diversity of different reference points the sun, polarization patterns in the sky, terrestrial landmarks for the region s function in different species? Do the different structures of the central complex have distinct modular roles or are they all just bit players in a network that maintains the animal s whereabouts? nts and honeybees show sophisticated path integration capabilities, which allow them to keep track of distance as well as direction as they make foraging trips from and back to their homes. Whether and how the central complex combines rotation and translation is an open question. Further, while the inputs to the central complex are certainly incompletely understood, the outputs of the central complex are practically unknown. What do the animals do with their orientation vector? How exactly do they convert this abstract information into action? Does the central complex control the fine structure of an insect s motor output, or does it provide high-level direction and leave the details to thoracic networks? n insect s locomotion is clearly influenced by many other descending pathways when does the central complex take center stage? It is also not yet clear how salience is added to an insect s world to weight available actions. Interestingly, some ellipsoid body input neurons have been linked to flies ability to identify the nutritional content of food, which could provide further motivation to remember or explore particular locations. Perturbations to the central complex have also resulted in alterations in courtship in the fruit fly and in singing in the cricket, which suggests that the drive to find a mate may also influence central complex sensorimotor function. The number of open questions is both exciting and dizzying. Fortunately, the tractable number of neurons within the insect central complex readily allows for dissection and understanding of the underlying mechanisms. Electron microscopy and functional connectivity

5 can help to establish a wiring diagram and guide rigorous theoretical models. The ability to examine a complex sensorimotor integration center down to the level of individual synaptic connections is incredibly powerful. In conclusion, the central complex is a fascinating sensorimotor hub that offers the tantalizing hope of fully understanding the neural workings of aspects of navigation that are relevant across species. It takes in sensory and motor input, tracks heading, possibly with an attractor network, and generates predictive motor outputs. Learning, memory, and context also help shape the activity in this network, adding further flexibility to the workings of this region. It is no wonder that this powerful set of neurons is conserved across insects and crustaceans, and it will be fascinating to learn how the structure has evolved to meet each species specific needs. FURTHER REDING el Jundi,., Warrant, E.J., yrne, M.J., Khaldy, L., aird, E., Smolka, J., and Dacke, M. (215). Neural coding underlying the cue preference for celestial orientation. Proc. Natl. cad. Sci. US 112, Harley, C.M., and Ritzmann, R.E. (21). Electrolytic lesions within the central complex neuropils of the cockroach brain affect negotiation of barriers. J. Exp. iol. 213, Heinze, S., and Homberg, U. (27). Maplike representation of celestial E-vector orientations in the brain of an insect. Science 315, Loesel, R., Nässel, D.R., and Strausfeld, N.J. (22). Common design in a unique midline neuropil in the brain of arthropods. rthropod Struct. Dev. 31, Martin, J.P., Guo, P., Mu, L., Harley, C.M., and Ritzmann, R.E. (215). Central-complex control of movement in the freely walking cockroach. Curr. iol. 25, Müller, M., Homberg, U., and Kühn,. (1997). Neuroarchitecture of the lower division of the central body in the brain of the locust. Cell Tissue Res. 288, Neuser, K., Triphan, T., Mronz, M., Poeck,., and Strauss, R. (28). nalysis of a spatial orientation memory in Drosophila. Nature 453, Ofstad, T.., Zucker, C., and Reiser, M. (211). Visual place learning in Drosophila melanogaster. Nature 474, Pfeiffer, K., and Homberg, U. (214). Organization and functional roles of the central complex in the insect brain. nnu. Rev. Entomol. 59, Seelig, J.D., and Jayaraman, V. (215). Neural dynamics for landmark orientation and angular path integration. Nature 521, Strausfeld, N. (212). The rain within the rain in rthropod rains (Cambridge, M: The elknap Press). Wolff, T., Iyer, N., and Rubin, G. (214). Neuroarchitecture and neuroanatomy of the Drosophila central complex: GL4-based dissection of protocerebral bridge neurons and circuits. J. Comp. Neurol. 523, Janelia Research Campus, Howard Hughes Medical Institute, 197 Helix Drive, shburn, V 2147, US. vivek@janelia.hhmi.org Correspondence Dendrogramma is a siphonophore Timothy D. O Hara 1,, ndrew F. Hugall 1, Hugh MacIntosh 1, Kate M. Naughton 1, lan Williams 2, and dnan Moussalli 1 Dendrogramma was the iconic deep-sea animal of 214, voted among the top-ten new species described that year [1]. The two species described are mushroom shaped animals, diploblastic, with an apparent gastrovascular system that extends from the base of the stalk to bifurcating canals that radiate through the flat disc [2]. The authors could not assign the new genus to any known animal group with certainty, leading to numerous media reports that it belonged to an entirely new phylum. Here we use phylogenomic data from newly collected specimens to show that Dendrogramma is a cnidarian, specifically a benthic siphonophore in the family Rhodaliidae. lthough an entire Dendrogramma colony has not been found, we hypothesise that the mushroom-like bodies are bracts, possibly used to aid buoyancy or as defensive appendages to protect feeding gastrozooids or gonads. In November 215, an expedition by the ustralian research vessel R/V Investigator to the continental slope off South ustralia recovered 85 new specimens of Dendrogramma (Figure 1,). We preserved some specimens in RNLater and after the voyage successfully sequenced a partial transcriptome from extracted RN. These are the first genetic data available for these species, as the original specimens were collected and preserved in formalin in 1986, which hindered extraction of DN [2]. We targeted exons used in a recent metazoan phylogenomic study [3] as well as the ribosomal RN genes (18S, 28S and 16S) used in older siphonophore-focused investigations [4,5]. Phylogenetic analyses of the exon data placed Dendrogramma as sister to two physonect siphonophores (galma elegans and Nanomia bijuga, 1% bootstrap support; Figure 1C, Supplemental information). The ribosomal LST and phylogenetic analyses identified the benthic siphonophore Stephalia dilata (family Rhodaliidae) as the nearest sequenced taxon to Dendrogramma (16S d xy =.13; Supplemental information). Siphonophores are bizarre pelagic colonial cnidarians in the class Hydrozoa. They are complex elongate or spherical organisms with specialised locomotive and feeding zooids, and a net of tentacles that can be extended to catch prey or attach to the seafloor [6]. There are 175 described species, living in a range of habitats from the sea surface (e.g., Physalia physalis, the Portuguese Man O War) to the deep-sea [6]. Larger, more delicate species have been found mainly in the non-turbulent mesopelagic (3 1 m) or bathypelagic zones (1 3 m). typical siphonophore has a pneumatophore for floatation, a short nectosome with nectophores (swimming bells) used for propulsion, and a siphosomal stem with repeating units called cormidia. Cormidial units have gastrozooids with long feeding tentacles, gonodendra for reproduction and bracts used for floatation or protection [6,7]. We hypothesise that the mushroomshaped Dendrogramma is a cormidial bract. These can take various shapes, but notably rhodaliid genera such as Tridensia, rancialia and rchangelopsis have detachable mushroom-like bracts with a central canal up the stalk that can branch through the disc on some species [7,8] (Supplemental information). Dromalia has quite complex branching, although the bracts are elongate [7]. While bract morphology is only available for ten of the fourteen species of the Rhodaliidae, bracts are typically small appendages that reach 2 6 mm in size. In contrast our largest specimen of Dendrogramma is 2 mm in diameter with bracteal canals that can branch at least five times, almost to the edge of the rounded disc. The canals are a pink or orange colour in life (Supplemental information). The mouth of Dendrogramma is the attachment surface that articulates with a triangular bracteal lamella on the cormidial stem [8]. We also found an orange gasfilled object at two separate sites (Supplemental information), which is similar in appearance to a rhodaliid pneumatophore and the base of the siphosomal stem, denuded of all nectophores and cormidial units. Ribosomal DN (16S, 28S) from one of these objects was identical to that found from the Dendrogramma bracts from the same site. 26, R445 R46, June 6, 216 Crown Copyright 216 Published by Elsevier Ltd. R457