research synopsis

I encourage those interested in the lab to contact me.


Modern molecular genetics is providing an ever clearer view of how individual neurons are born, grow, function, and form modifiable synapses. Complimentary advances in imaging technology are disclosing operational processes of the intact brain. The next challenge is to examine the mechanisms that link the computations by elemental micro-circuits to the highly complex behaviors they regulate.

The defining characteristics of our lab is the integration of biological and engineering techniques to investigate multi-sensory processing and sensory-motor integration at genetic, cellular, cell-system, and organ-system levels of organization. Making extensive use of state-of-the-art electronic flight and walking simulators (“virtual reality” for flies), we quantify the influence of visual and olfactory stimuli on behavioral dynamics under conditions in which the animals actively control their sensory experience. We assess performance deficits resulting from targeted and conditional gene expression (e.g., the pGal4-UAS system) to reversibly silence specific neuronal circuits. We hope to reveal specific structural circuits and general functional algorithms by which the multi-sensory world is processed, integrated, and transformed into the biomechanics of complex behavior.

Research Program

Visual computations, circuits, and behavior

To stabilize gaze, seeing animals ranging from insects to primates reflexively move their eyes or body to follow panoramic image motion. My lab uses this powerful reflex in flies to assess how visual input is transformed into motor output. Flies offer a unique advantage here in that we can integrate broadly from photoreceptor input to whole-animal behavior. Our experiments are performed with individual flies suspended in either of two flight simulator configurations. The compelling aspect of our approach is that we can precisely measure and manipulate sensory conditions such as the spatial, temporal, and contrast structure of visual motion cues, yet the fly actively controls its own sensory experience. For example, in one configuration the fly is fixed in place, but its wing motions are optically tracked and fed back to control the visual panorama such that an attempted left turn is met with a rightward rotation of the visual display, as expected during free-flight. A second configuration fixes the visual panorama in place, and instead a magnetic field allows the fly to turn on a pivot and freely orient toward interesting features of the arena. Both arenas are equipped with laminar-flow odor systems that allow precise manipulation of spatial and temporal properties of olfactory stimuli. Our techniques therefore offer a high degree of experimental control while at the same time providing vast output degrees of freedom for a behaving animal.

During my postdoc with Michael Dickinson, I helped disclose a novel form of visual reflex in flies. Simply stated, flies are highly sensitive to patterns of optic flow generated by the apparent expansion of approaching features. In my own lab I wanted to know whether this reflex operates through a dedicated neural “channel” or rather is it is part of a well-studied visual response to panoramic rotation? Though it is a critical computational task, it is unknown in any visual system how rotation and translation patterns of optic flow are functionally segregated. We systematically varied the spatial, temporal and contrast properties for each of two primary forms of optic flow – visual expansion and rotation – and examined the motor control of steering during tethered flight. Owing to substantial neurophysiological, behavioral and modeling literature, we were armed with quantitative predictions of what to expect if (i) the two types of optic flow (expansion and rotation) are emergent properties of one motion processing system, or (ii) they comprise two independent control pathways. All indications support the latter conclusion. The results were published in The Journal of Comparative Physiology, and Journal of Experimental Biology. Motivated in part by these findings, we analyzed the patterns optic flow experienced by freely flying animals exploring a one-meter laboratory arena. By manipulating the structure of the visual surroundings we showed that flight speed and collision-avoidance maneuvers are strongly influenced by the spatial structure of optic flow.

Complex patterns of optic flow as well as simple motion stimuli generate spatiotemporal variations in luminance – for example a bright bug crawling across dark tree bark. But humans and non-human primates also detect motion generated by variations in contrast or texture instead of luminance. In primates, luminance based “first order” motion and contrast and texture based “second order” motion cues are thought to be processed by two parallel cortical streams. Imagine a butterfly in flight - the body motion generates first-order motion whereas the fluttering wings generate second-order cues. These types of visual features might be relevant to non-primate animals as well. We tested whether flies can see second-order motion cues by presenting high-resolution motion cues during tethered flight. Remarkably, flies robustly track several forms of second-order motion stimuli, even when these cues are competing with first-order signals. This is surprising because the standard implementation of a well-known computational model for elementary motion detection tracks first-order cues and is completely blind to second-order motion.

Gamma amino butyric acid (GABA) is a major inhibitory neurotransmitter in both mammals and insects. In flies, GABA signaling has been shown to play a vital role in both elementary motion detection and also higher-order feature tracking. We are interested in developing neurogenetic and behavioral methods for further examining GABA signaling for visual processing. In collaboration with Dr. David Krantz at UCLA and Julie Simpson at JFRC, we combined genetics and quantitative real-time behavioral analysis to behaviorally investigate the role of inhibitory synaptic signaling for complex visual motion processing. The Krantz lab characterized the Drosophila ortholog of the Vesicular GABA Transporter (VGAT), mapped the distribution of the protein in the nervous system, and characterized the phenotype of a lethal mutant allele that is unable to release GABA. A similar analysis has been performed using knock-out mice, but the mice die as embryos. Using genetic techniques, the developmental lethality of the dVGAT mutant was conditionally rescued, enabling us to perform visuo-motor behavioral tests. Our early results show that dVGAT mutants show surprisingly normal gaze stabilizing visual reflexes, but are deficient in object tracking.

Ultimately, we are interested in identifying the neural circuits responsible for the sensory-motor regulation of complex behavior. As part of this broader effort, we recently developed a high-throughput real-time walking simulator (a sort of virtual reality “fly stampede”) that allows us to examine several independent and subtle visual-motor behavioral phenotypes in a large group of walking flies under reversible inactivation of specific brain microcircuits. We initially focused our first efforts on the biological implementation of the elementary motion detector (EMD), which has been a highly successful conceptual model in biology, explaining the electrophysiological properties of visual circuits and the behavioral dynamics of freely-moving animals with widely differing eye design. However, excepting photoreceptor inputs and their immediate postsynaptic targets, much of the neural substrate for the EMD remains enigmatic. We genetically manipulated the function of peripheral visual processing circuits, and quantified the resultant dynamics of visually evoked walking responses in freely-moving animals. Flies with non-functional photoreceptors R1-R6 are motion-blind but retain attraction to light, confirming an early segregation between light detection and motion computation. Flies with genetically deactivated peripheral visual circuits that interconnect neighboring visual columns are completely insensitive to motion cues, but display fully intact phototaxis. One member (lamina neuron L4) of the inactivated cell ensemble synaptically interconnects optical columns, and the connection pattern matches the functional axes of the EMD across the eye. Thus, our results strongly implicate L4 as an inter-column comparator that participates in elementary motion computations.

Taken together our results highlight the immense computational complexity as well as experimental tractability offered by flies for dissecting the underlying computational algorithms and neural circuits of high-order visual processing. Further vision research projects currently underway include using white noise system identification methods to mathematically characterize the transfer of visual input signals into motor output behavior. In the parlance of the field, our results have disclosed the first-order filter kernels for translation, rotation, divergence and deformation flow fields. We are also performing genetic screens for the neural circuits mediating parallel first-order and second-order motion processing.


Olfactory computations and behavior

Recent technical advances have disclosed the molecular-genetic and physiological bases of early olfactory coding. However, our understanding diminishes for the steps between primary sensory function and odor-mediated behavioral responses. Applying our unique quantitative experiments with intact animals, we are systematically mapping the physiological activation patterns of olfactory receptors onto odor mediated flight behavior. We developed a “virtual odor arena” in which a tethered fly is suspended in a magnetic field that allows it to voluntarily steer in the horizontal (yaw) plane. The apparatus is suspended inside a wrap-around visual display and several small nozzles allow for the controlled delivery of narrow odor plumes. As a fly steers either in response to the odor plumes or in response to controlled visual motion, its body orientation is tracked with digital video. The spatiotemporal properties of visual and olfactory stimuli can be varied in real-time.

Under these conditions, a hungry fly steers back-and-forth with high-angle turns termed saccades for their functional analogy to our own gaze-stabilizing eye movements. Upon contacting an appetitive odor plume, saccades are suppressed. However, our data indicate that flies continue to generate small-angle turns back-and-forth across the odor plume. We are examining the possibility that these small turns are functionally analogous to our own visual microsaccades, which circumvent the corrupting influence of sensory adaptation in photoreceptors. We suspect that flight microsaccades refresh either visual or olfactory signals (or both) without causing the fly to lose the plume altogether. We are currently testing this hypothesis with genetic reagents such as a transient receptor potential (trp) calcium channel mutant that shows reduced sensory adaptation to odors, and by introducing temporal modulations to visual and olfactory stimuli in order to mitigate adaptation and thereby diminish flight ‘microsaccades’. Additional analyses under way include genetic manipulation of olfactory sensory receptors to examine how the spatial code of activation in the early olfactory system is represented in active odor tracking behavior across a range of stimulus intensities and identities. These issues are important because whereas the temporal characteristics of early olfactory pathways have been thoroughly examined, further temporal transformations by higher brain centers into olfactory behavior have yet to be disclosed in any system.

We are also investigating the role of bilateral antennal intensity cues for spatial gradient tracking. We have recent evidence that during flight, flies make bilateral comparisons of odor intensity to actively steer into an odor plume. Osmotropotaxis has been demonstrated in walking flies, but the much increased locomotion speed and complex fluid dynamics associated with flight would appear to preclude meaningful bilateral comparisons. Nevertheless, we show that unlike normal flies, those with one occluded antenna fail to track a directional gradient, and steer toward the un-occluded side only in the presence of an appetitive odor.


Visual and olfactory cross-modal integration

Despite having received broad interest in recent years, the mechanisms and algorithms by which the brain integrates multisensory signals remain enigmatic. As a postdoc, I discovered that during free-flight flies require visual feedback to successfully locate the source of an attractive odor. The physiological mechanisms of cross-modal integration in flies could offer general insight into how information from one sensory modality enhances stimulus detection by another modality. Recently, we examined how varying ambient visual conditions influence a hungry fly’s ability to stably track an attractive odor plume. For a magnetically-tethered fly that is free to steer within an odor plume, we showed that both olfactory and visual stimuli influence the production of body saccades. In addition, when the visual panorama is composed of high-contrast patterns, generating strong visual motion cues, the robustly flies stabilize their heading within an attractive plume with remarkable accuracy. However, reducing the contrast of the visual background diminishes plume tracking ability such that flies behave more like a no-odor control, orienting randomly in the arena. Flies require visual feedback to stabilize an odor plume. We showed that the cross-modal integration is context specific in that it requires panoramic visual motion cues but not small-field object cues. Our results support the exciting conclusion that olfactory signals directly activate visual stabilization reflexes, and the specificity of action implicates wide-field, not small-field visual processing pathways.

Do olfactory signals enhance the sensitivity to visual rotation specifically, or do they enhance the sensitivity of visual motion more generally? We compared visual responses before and after presentation of an attractive odor within a flight simulator projecting two optic flow fields: rotation and expansion. The results are remarkable: first, olfactory signals directly enhance the salience of visual cues such that flies “pay better attention” to visual stimuli in the presence of odor. Second, odor increases flies’ sensitivity to visual rotation and decreases sensitivity to expansion. Thus, in addition to a general increase in the perceptual salience of visual motion cues, odor activation of visual motor reflexes shows context-specificity with respect to the same parallel visual processing channels described above.

We are currently pursuing the neural circuits mediating visuo-olfactory fusion. Olfactory projection neurons selectively innervate both learning integration centers and also reflexive memory-independent processing regions of the fly brain. Genetic markers are allowing us to asses the function of circuits putatively involved in either rapid reflex integration, or learning dependent goal directed movements. In particular, we are focusing on the central complex, an anatomically sophisticated brain region thought to integrate spatial memory and coordinate complex pre-motor command patterns.


Arousal-enhanced sensory-motor integration

During our walking optomotor analyses, we discovered that distributing flies into multiple small groups or eliminating a pre-trial startle stimulus results in weaker optomotor responses. This is interesting since the same flies placed together in one large group show a highly robust visually-evoked “stampede” response. Thus there appears to be a strong “group effect” on the strength of optomotor reflexes that is enhanced by a mechanical startle stimulus. We have evidence suggesting that the apparent social-influence is mediated by a pheromone-like signal known to be released by stressed flies, which activates the olfactory system. So-called “Drosophila stress odorant”, composed largely of CO2, has recently been suggested to represent an innate aversive cue in flies. Our results with the stampede assay not only confirm that CO2 produces an innate avoidance response, but also indicate that this olfactory signal enhances the time course and strength of optomotor behaviors. We are currently conducting a series of genetic manipulations of olfactory circuits (e.g. olfactory and gustatory double-mutants and selective receptor rescues) to nail down the signaling mechanisms underlying stress-induced multisensory enhancement in flies. We suspect that the enhancement functions across sensory modalities. To test for multimodal activation we have adapted the visual stampede technique to a star-shaped walking arena in which multiple odors are released simultaneously form the corners, and a large group of flies is monitored for movement into the preferred quadrant.


Exploratory behavior and active search

In the fruit fly, visual-olfactory fusion presumably evolved for a single purpose: to enable an adult to disperse and locate patchy food resources within widely varying visual environments. This is a “needle in a haystack” problem. We recently tested the hypothesis that flies use something other than a Brownian motion-like search pattern when seeking distant odor sources. Our results suggest that in free-flight the reiterated pattern of straight flight and saccades that characterize Drosophila flight paths constitute a mathematically optimal spatially scale-free search strategy known as a Lévy flight. This is somewhat reminiscent of the fractal pattern on a snowflake in that it appears similar whether viewed up close or from a distance. Similar search patterns have been revealed in foraging birds, monkeys, zooplankton, and even human hunter-gatherers. However, our work represents the first demonstration of a search strategy that is both scale-free and intermittent in that the Lévy search is terminated in favor of a diffusive Brownian random walk very near the odor source. Regardless the specific algorithm employed for this behavior, we are keenly interested in the mechanisms by which sensory-independent distant exploratory behavior switches to sensory-dependent active local search behavior. Work currently underway is aimed at understanding how the switch to active local search might be influenced by both the time course of starvation and transcription factors that regulate carbohydrate metabolism by insulin.


View several posters that highlight our work:

UCLA media coverage:
UCLA newsroom 1
UCLA Newsroom 2

Yan_SFN_2006 (<1MB)

Brian_SFN_2006 (12MB)

Mark_NSIDP (<2MB)

Funding to support our research is provided by The Howard Hughes Medical Institute , The National Science Foundation , The Whitehall Foundation, The Alfred P. Sloan Foundation, the W.M. Keck Foundation, and the UCLA Council on Research

movie gallery