complexity affects the response of a looming-sensitive neuron to object motion 1 2

17 An increasing number of studies show how stimulus complexity affects the responses of 18 looming sensitive neurons across multiple animal taxa. Locusts contain a well described, 19 descending motion sensitive pathway that is preferentially looming sensitive. However, the 20 lobula giant movement detector/descending contralateral movement detector (LGMD/DCMD) 21 pathway responds to more than simple objects approaching at constant, predictable trajectories. 22 In this study, we presented Locusta migratoria with a series of complex 3-dimensional visual 23 stimuli presented while simultaneously recording DCMD activity extracellularly. In addition to a 24 frontal looming stimulus, we used a combination of compound trajectories (non-looming 25 transitioning to looming) presented at different velocities and onto a simple, scattered, or 26 progressive flow field background. Regardless of stimulus background, DCMD responses to 27 looming were characteristic and related to previously described effects of azimuthal approach 28 angle and velocity of object expansion. However, increasing background complexity caused 29 reduced firing rates, delayed peaks, shorter rise phases and longer fall phases. DCMD responded 30 to transitions to looming with a characteristic drop in a firing rate that was relatively invariant 31 across most stimulus combinations and occurred regardless of stimulus background. Spike 32 numbers were higher in the presence of the scattered background and reduced in the flow field 33 background. We show that DCMD response time to a transition depends on unique expansion 34 parameters of the moving stimulus irrespective of background complexity. Our results show how 35 background complexity shapes DCMD responses to looming stimuli, which is explained within 36 behavioural context. 37 38 Vison, optic flow, looming, Locusta migratoria, DCMD 39 Detection of object motion in complex stimulus environments 3 INTRODUCTION 40 Flying animals are continually challenged with different forms of visual motion, 41 including self-produced flow field motion (self motion in a stationary environment) and motion 42 produced from an object moving within a stationary environment or in a direction opposite to a 43 predicted one. The discrimination between these types of motion is paramount to an animal’s 44 survival, for example predator avoidance. Flying through complex visual environments requires 45 the detection of relevant salient visual cues for successful navigation. Looming objects, for 46 example, provide critical information regarding an oncoming collision or perhaps an approaching 47 predator. While stationary, an approaching visual stimulus is clearly interpreted as noxious, thus 48 detection, and subsequent avoidance behaviour, may be relatively straightforward. However, 49 while generating self-generated optic flow during movement or, in the case of swarming animals, 50 surrounded by conspecifics moving at often unpredictable velocities and directions, detection of 51 noxious stimuli is challenging. 52 The migratory locust, Locusta migratoria, is an established neuroethological model 53 system for studying collision avoidance due to its long research history, easily tractable nervous 54 system, and well-identified looming sensitive neurons (LSNs). During flight, a locust’s visual 55 environment is dynamic. Within a swarm, individual locusts may fly ~3 m/s and in close 56 proximity with each other, while maintaining flight elevations from 1 – 1000 m above ground 57 (Uvarov 1977). Neighbouring locusts approach from different angles and at different velocities 58 while land geography changes below. However challenging the environment, flying locusts are 59 capable of avoiding collisions with conspecifics (Waloff 1972) and aerial attacks of diving birds 60 while swarming (Santer et al. 2012). Indeed, experimental studies on free flying (Dawson et al. 61 2004) and loosely tethered flying locusts (Chan and Gabbiani 2013; McMillan et al. 2013) show 62 Detection of object motion in complex stimulus environments 4 that locusts use a relatively unpredictable range of avoidance behaviours in response to noxious 63 stimuli. Successful navigation within such a complex environment is, in part, related to a well64 developed visual network of movement sensitive neurons, specifically and most widely studied 65 are the lobula giant movement detector (LGMD) and its postsynaptic partner the descending 66 contralateral movement detector (DCMD). 67 Excitation of the LGMD begins when movement within a locust’s visual field stimulate 68 retinotopically arranged fibers within the ommatidia, which produce excitatory input to one of 69 the three large dendritic fields of the LGMD (Rind 1984). During a looming approach, the 70 number of spikes produced by the LGMD is directly related to an approaching object’s angular 71 velocity and subtense angle and is thus referred to as an angular threshold detector (Gabbiani et 72 al. 1999). As an object approaches the retina, the LGMD firing rate increases to a peak and then 73 decays once object motion stops and before a collision would have occurred (Gabbiani et al. 74 1999, 2001, 2002; Gray 2005; Guest and Gray 2006; McMillan and Gray 2012; Dick and Gray 75 2014). Presynaptic lateral inhibition and postsynaptic feed-forward inhibition from the other two 76 dendritic fields control excitation and thus define the peak firing rate of the LGMD (Rowell et al. 77 1977; Gabbiani et al. 1999, 2001). Each LGMD synapses onto a DCMD within the 78 protocerebrum, generating a one-to-one spike ratio (O'Shea and Williams 1974); for the ease of 79 access, many studies record from the DCMD axon within the contralateral side of the ventral 80 nerve cord. 81 The LGMD/DCMD pathway is part of a relay system that tracks the approach and signals 82 an impending collision of visual objects to motor centers within the thoracic ganglia (Simmons 83 1980). Phases of an avoidance jump have been linked to phases of this pathway’s firing rate 84 (Fotowat and Gabbiani 2007) in addition to a possible role in modifying wing beat rhythm 85 Detection of object motion in complex stimulus environments 5 during flight (Santer et al. 2006). While the DCMD habituates to repetitive stimuli (Horn and 86 Rowell 1968; Palka 1967; Gray 2005), it remains sensitive to a simple looming stimuli following 87 translatory motion within a locust’s field of view and also responds to the transition to and from 88 a looming trajectory (McMillan and Gray 2012; Dick and Gray 2014). Although it remains 89 unknown if the DCMD is responsible for avoidance behaviours in complex environments (such 90 as those found while flying in a swarm), these studies suggest that the DCMD is capable of 91 responding to important aspects of a complex visual environment (see Rind and Simmons 1992). 92 The interest in answering the question of complexity reaches beyond neurobiology and into 93 robotics. MAVs and other robotic control systems are often engineered based on the physiology 94 and circuitry of insect models. 95 To understand how visual neurons respond in natural environments, it is important to 96 balance quantifiable stimulus parameters (i.e. object motion) with aspects of complex scenes (i.e. 97 optic flow). In addition to a simple looming stimulus, we presented a combination of bilaterally 98 paired non-looming and looming stimuli (i.e. compound trajectories) at varying velocities. All 99 stimuli were presented in a 3-dimensional environment on a specialized dome projection screen 100 using either a simple white background, a scattered background with hundreds of randomly 101 translating dots to represent a “swarm”, or a progressive flow field background representing 102 optic flow produced byforward motion. Consistent with previous work using compound 103 approach trajectories (McMillan and Gray 2012; Dick and Gray 2014), the DCMD responded to 104 a transition to looming with a quantifiable drop in firing rate (valley) that was relatively 105 consistent for all trajectory types, velocities of object approach, and background environment. 106 Response time from transition to DCMD valley was also consistent with previous work 107 (McMillan and Gray 2012; Dick and Gray 2014) and remained relatively invariant across all 108 Detection of object motion in complex stimulus environments 6 stimulus combinations. Moreover, within each stimulus background the DCMD firing rate was 109 capable of tracking and responding to the motion of each stimulus. However, many of the 110 measured response parameters of the DCMD response differed depending on the type of 111 trajectory, velocity, and background. We show that although background complexity affects 112 DCMD responses to looming and transitions to looming stimuli, the collision associated peak 113 response is, in general, remarkably invariant. 114 Detection of object motion in complex stimulus environments 7 MATERIALS AND METHODS 115 116 Animals 117 We used 14 adult male Locusta migratoria for experimentation. All animals, at least 3 118 weeks past the imaginal molt, were obtained from a crowded colony maintained in the 119 Department of Biology at the University of Saskatchewan (25-28°C, 12hr:12hr light:dark). 120 Experiments were carried out at ~25°C during similar times of the animals’ light cycle to avoid 121 potential variations in responsiveness when locusts fly at night (Gaten et al. 2012). 122 123 Preparation 124 After the legs were removed and the wings were clipped, a rigid tether was attached to 125 the ventral surface of the thorax using low melting point bee wax. A small patch of ventral 126 cervical cuticle was removed to expose the underlying paired connectives of the ventral nerve 127 cord anterior to the prothoracic ganglia. Locust saline (147·mmol NaCl, 10·mmol KCl, 128 4·mmol·CaCl 2 , 3·mmol·NaOH, 10·mmol Hepes, pH·7.2) was applied to the exposed tissue and 129 the preparation was moved to the recordi