European Journal of Neuroscience, pp. 1–9, 2014
doi:10.1111/ejn.12731
REVIEW In search of the holy grail of fly motion vision Alexander Borst Department of Circuits, Computation, Models, Max-Planck-Institute of Neurobiology, Am Klopferspitz 18, 82152 Martinsried, Germany Keywords: calcium imaging, insect vision, neural computation, neurogenetics
Abstract Detecting the direction of image motion is important for visual navigation as well as predator, prey and mate detection and, thus, essential for the survival of all animals that have eyes. However, the direction of motion is not explicitly represented at the level of the photoreceptors: it rather needs to be computed by subsequent neural circuits, involving a comparison of the signals from neighbouring photoreceptors over time. The exact nature of this process as implemented at the neuronal level has been a longstanding question in the field. Only recently, much progress has been made in Drosophila by genetically targeting individual neuron types to block, activate or record from them. The results obtained this way indicate that: (i) luminance information from fly photoreceptors R1–6 is split into two parallel motion circuits, specialized to detect the motion of luminance increments (ON-Channel) and decrements (OFF-Channel) separately; (ii) lamina neurons L1 and L2 are the primary input neurons to these circuits (L1 ? ON-channel, L2 ? OFF-channel); and (iii) T4 and T5 cells carry their output signals (ON ? T4, OFF ? T5).
Introduction Determining the direction in which something is moving seems like a trivial task. However, while the direction of motion is certainly implicit in an image sequence, it is not explicitly encoded at the level of single photoreceptors. Making it explicit requires computation. As depicted in Fig. 1A, when a bright bar moves from left to right and back again in front of a fly’s eye, the electrical response of a single photoreceptor would look just the same. Thus, from the response of an individual photoreceptor, one cannot tell the direction of local image motion. The fly’s optic lobe consists of four different neuropiles called lamina, medulla, lobula and lobula plate. All these neuropiles are built from repetitive columns, arranged in a retinotopic way. At the level of the lobula plate, large tangential cells are found that cover many hundreds of columns with their dendrites. If the electrical signals from such neurons are recorded during the same stimulation protocol as above, a very different response is found: during bar motion to the right (their ‘preferred direction’); these cells depolarize, during bar motion to the left (their ‘null direction’), they hyperpolarize. In contrast to the photoreceptor signal, these cells encode the direction of image motion in their membrane potential. Obviously, somewhere between the photoreceptor terminals and the dendrites of the lobula plate tangential cells, the direction of motion has been computed. This process is called ‘elementary motion detection’ and is the central topic of interest here.
The Hassenstein–Reichardt detector Analysing the turning tendency of the beetle Chlorophanus viridis walking on a spherical Y-maze, Hassenstein & Reichardt (1956)
Correspondence: Dr A. Borst, as above. E-mail: [email protected] Received 16 July 2014, revised 19 August 2014, accepted 20 August 2014
proposed a specific model of elementary motion detection that could account for their observations in a quantitative way (Fig. 1B). The idea was that the beetle’s nervous system contains many hundreds of such elementary units, which, collectively, cover the whole visual field, each extracting locally the direction of image motion. The algorithmic model for such an elementary motion detector consists of two subunits, which are mirror-symmetrical to each other (Fig. 1B, left; Reichardt, 1961, 1987; Borst & Egelhaaf, 1989). Each subunit reads the luminance values measured in two adjacent image pixels. These values pass an input filter before they are fed through temporal filters with different characteristics (low-pass, high-pass). The differential filtering is an essential property as it creates a timelag between the signals. Next, the signals become multiplied (M). This is the first stage at which a directionally selective signal arises. When the output values of both subunits are finally subtracted, a fully opponent response is obtained: now, the response is positive for motion in one direction and negative for motion in the opposite direction. The value of this model for research in the field of motion vision can hardly be overestimated. The reason for this is that this model makes predictions that are both quantitative and counterintuitive. When, for example, a periodic grating is moving at a constant velocity, the model predicts, in contrast to a simple speedometer, a speed optimum at which the response is maximal. Furthermore, this optimum speed is proportional to the pattern wavelength. All this has indeed been demonstrated in house flies (Fermi & Reichardt, 1963; Eckert, 1973), blow flies (Haag et al., 2004) and fruit flies (Goetz, 1964, 1965; Joesch et al., 2008; Schnell et al., 2010), in behavioural experiments as well as at the level of lobula plate tangential cells. The Hassenstein–Reichardt detector also makes a number of other predictions relating to the response dynamics (Borst & Bahde, 1986; Borst et al., 2003, 2005). Again, these predictions were verified in recordings from lobula plate tangential cells of blow flies as well as of Drosophila (Egelhaaf & Borst, 1989; Brenner
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
2 A. Borst A
B
Fig. 1. (A) From non-directional to directional responses. Left: schematic of the optic lobe of Drosophila. The retina (red) built from repetitive facets is followed by four layers of neuropile, called lamina, medulla, lobula and lobula plate. Within the lobula plate, three tangential cells of the horizontal system (HS-cells) are shown as reconstructed from stained tissue. Right: schematic signals recorded from a single photoreceptor (top) and from a tangential cell (bottom) during motion of a bar. While the photoreceptor signal is non-directional, the signal of the tangential cell is strongly directional. (B) The Hassenstein–Reichardt detector of elementary motion detection (left), together with their inventors Bernhardt Hassenstein (top) and Werner Reichardt (bottom) and the behavioural set-up they used (right). The model consists of two mirror-symmetrical subunits sharing the same two inputs after passing a filter. Within each subunit, the signal from one input is processed by a temporal low-pass filter (LP) and subsequently multiplied (M) with the high-pass filtered signal (HP) derived from the neighbouring input. The signals from both subunits are subtracted ( ), giving rise to a directionally selective output.
et al., 2000; Reisenman et al., 2003; Borst et al., 2005; Joesch et al., 2008). Interestingly, the fly motion-detection system seems to adapt some of its parameters. During ongoing motion stimulation, the responses as recorded in tangential cells can be best described by a model where the time-constant of the high-pass filter shortens (Borst & Egelhaaf, 1987; Borst et al., 2003; Safran et al., 2007). Furthermore, the response of the tangential cells changes depending on the behavioural state of the animal, i.e. whether the fly is actively walking or flying or whether the fly is at rest (Longden & Krapp, 2009; Chiappe et al., 2010; Maimon et al., 2010; Rosner et al., 2010; Jung et al., 2011). Beside the filter time-constants, experimental evidence suggests the sampling base of the detector also to be variable in the fly visual system: while at high luminance levels, nearestneighbour interactions dominate (Goetz, 1964, 1965; Buchner, 1976), wide-range interactions come into play at low luminance conditions (Pick & Buchner, 1979; Schuling et al., 1989).
Neural implementation of the Hassenstein–Reichardt detector Given the evidence for motion computation according to the Hassenstein–Reichardt detector in the fly optic lobes, the question
about its neural implementation naturally arises: which neurons form the Hassenstein–Reichardt detector? Here, an almost complete catalogue of all columnar neurons between the photoreceptors and the lobula plate tangential cells has been worked out. Starting with the work of Cajal & Sanchez (1915), most of the columnar cell types of the lamina, medulla, lobula and lobula plate have meanwhile been identified and described on the basis of Golgi impregnations in the house fly (Strausfeld, 1976) as well as in Drosophila (Fischbach & Dittrich, 1989). Each lamina column (or cartridge, as it is usually referred to) contains a set of 12 cell types, connected to the photoreceptors either directly or indirectly (Meinertzhagen & O’Neil, 1991). These lamina neurons connect the photoreceptors to specific layers of the medulla (Takemura et al., 2008, 2011, 2013). In the medulla, a single column houses more than 60 different cell types. They can be roughly grouped into four groups: medulla intrinsic (‘Mi’) neurons connect different layers of the medulla to each other; transmedulla (‘Tm’) neurons connect specific layers of the medulla to various layers in the lobula; trans-medulla Y (‘TmY’) neurons connect specific layers of the medulla to various layers in the lobula and lobula plate; and bushy T4 cells connect medulla layer 10 to the four layers of the lobula plate. In addition, a similar group of bushy T5 cells connect the posterior layer of the lobula to the four layers of the lobula plate. However, the small size of all these
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
In search of the holy grail of fly motion vision 3 neurons prohibited electrophysiological recordings in most cases. Therefore, while a rather complete map of all columnar neurons has been available for a while, the visual response properties of most of them were completely unknown until recently. It is for this reason that the problem of the neural implementation of the Hassenstein– Reichardt detector has been in the field for over half a century, becoming the ‘holy grail of fly motion vision’, with only little progress for many decades. Only recently, the advent of sophisticated neurogenetic methods in Drosophila has allowed for elucidating the circuits for elementary motion detection. These techniques are all based on a two-component expression system where a so-called ‘driver line’ defining the neurons in which a certain effector gene is expressed is crossed with another line, the so-called ‘effector line’, defining what gene is expressed (Brand & Perrimon, 1993). Today, thousands of different driver lines are available, many of which reveal a high degree of specificity for expression in individual cell types of the optic lobe (Pfeiffer et al., 2008; Tuthill et al., 2013). Furthermore, many effector lines have been developed that allow for blocking the synaptic output of neurons, for activating neurons optogenetically or via temperature shift, as well as for recording from neurons via genetically encoded calcium indicators (for review, see Borst, 2009; Venken et al., 2011). Applying these techniques to the problem of local motion detection revealed: (i) that luminance information from fly photoreceptors R1–6 is split into two parallel motion circuits, specialized to detect the motion of luminance increments (ON-Channel) and decrements (OFF-Channel) separately; (ii) that lamina neurons L1 and L2 are the primary input neurons to these circuits (L1 ? ON-channel, L2 ? OFF-channel); and (iii) that T4 and T5 cells carry their output signals (ON ? T4, OFF ? T5). The Golgi shapes of the most important cellular constituents of the motion pathway in Drosophila are shown in Fig. 2.
Lamina input A single facet of the fly eye comprises two sets of photoreceptors: the peripheral photoreceptors R1–6; and the central R7,8. Early work on different mutants where either the central or the peripheral ones were defective indicated that motion vision relies primarily on R1–6, and not on R7,8 (Heisenberg & Buchner, 1977). Within the lamina, five different monopolar cells distribute the photoreceptor signal to five different layers of the medulla. While monopolar cells L1–3 are directly postsynaptic to R1–6, L4 and L5 receive polysynaptic input: L4 through L2, and L5 via an amacrine cell (Meinertzhagen & O’Neil, 1991). Amongst the monopolar cells, the two most prominent neurons are L1 and L2. This suggested that they convey the input signal to downstream motion-detection circuits. The first evidence along these lines was provided by experiments where either behavioural responses (Rister et al., 2007) or electrophysiological recordings from lobula plate tangential cells were used as a read-out (Joesch et al., 2010): blocking both L1 and L2 led to a complete abolition of the responses to gratings moving along any direction. The individual contributions of L1 and L2 were discovered in flies where either L1 or L2 was blocked and, instead of gratings, edges of different contrast polarity were moved to probe the responses of lobula plate tangential cells (Joesch et al., 2010): while control flies responded to moving brightness increments (ON-edges) and decrements (OFF-edges) with about the same amplitude, flies with lamina neurons L1 blocked completely lost their response to moving ON-edges, while their response to moving OFF-edges was largely intact. The opposite was found for flies with lamina neuron L2 blocked: here, the response to moving ON-edges was roughly as
Fig. 2. Golgi gestalt of important neurons of the motion pathway in Drosophila (Fischbach & Dittrich, 1989).
large as the one of control flies, whereas the response to moving OFF-edges was severely reduced. These results were later confirmed in a study on behavioural responses of walking fruit flies (Clark et al., 2011). More recent evidence from blocking experiments as well as anatomy demonstrates that lamina neurons L3 (Silies et al., 2013; Shinomiya et al., 2014) and L4 (Takemura et al., 2011; Meier et al., 2014) provide additional contribution to the OFF-pathway. Together, these experiments suggest that the photoreceptor input from R1–6 is split into parallel channels depending on the contrast polarity of the incoming signal: while neurons postsynaptic to L1 specifically transmit information about brightness increments to downstream motion circuits, neurons postsynaptic to L2–4 preferentially allow information about brightness decrements to be passed onto further motion-processing units. Technically, such signal processing is referred to as ‘half-wave rectification’.
Two parallel ON- and OFF-motion pathways The finding about the specific contribution of L1 and L2 to motion processing of brightness increments and decrements immediately suggests that in Drosophila, motion is detected in several separate circuits. Theoretically, there could be four such channels comprising all possible combinations between ON- and OFF-signals. However, apparent motion experiments in blow flies indicated that only two such channels exist, one for the interaction of ON-signals and one for the interaction of OFF-signals between neighbouring image points (Riehle & Franceschini, 1984). In these experiments, pairs of
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
4 A. Borst single photoreceptors were stimulated by light-ON and light-OFF flashes while recording from a motion-sensitive lobula plate neuron H1. Recordings from lobula plate tangential cells in Drosophila arrived at the same conclusion: only sequences of light pulses of the same contrast polarity (ON–ON and OFF–OFF) elicited significant responses, while light pulses of differing contrast polarity (ON–OFF and OFF–ON) failed to evoke any responses in lobula plate tangential cells (Eichner et al., 2011). Interestingly, using brightness steps instead of brightness pulses led to different results: while again, ON–ON and OFF–OFF sequences along the preferred direction of the cell led to positive responses, ON–OFF and OFF–ON sequences elicited negative responses. While this so-called ‘reverse phi’ phenomenon (Anstis & Rogers, 1975) is usually taken as evidence for an interaction between ON and OFF stimuli (Egelhaaf & Borst, 1992), as in the original Hassenstein–Reichardt detector, careful simulations of such circuits as well as experiments with flies with either lamina neurons L1 or L2 blocked revealed that such an inversion of the response is not necessarily indicative of an interaction between ON- and OFF-channels. Rather, it can be explained by the residual sustained response component of the lamina neurons taking into account the contribution of motion detectors at the edge of the stimulated areas in the visual field of the fly (Eichner et al., 2011; Joesch et al., 2013). In summary, thus, the Hassenstein–Reichardt detector is implemented in the visual system of the fly in two parallel circuits, one dealing with brightness increments (ON-channel), the other one with brightness decrements (OFF-channel). What could be the advantage of such a doubling of wiring expenditure? Without splitting and half-wave rectification, i.e. with just a single motion detector dealing with brightness going positive as well as negative, a multiplicative interaction (according to the sign rule of multiplication) would imply the output being positive when both input signals go positive as well as when both go negative. Such a processing rule seems hard to implement biophysically and indeed is unheard of in cellular neuroscience. If, however, the input signals are split according to their polarity with the negative ones being sign-inverted, subsequent circuits only deal with positive input signals reducing the problem of multiplication to a supra-linear input–output characteristic as can easily be implemented by a voltage-gated ion channel. Interestingly, neurobiology is not the only area where such problems arise: analogue electronics faced the same problem of multiplying positive as well as negative signals, and the solution was what is called a ‘4-quadrant-multiplier’. There, the input signals, prior to multiplication, were half-wave rectified, split into parallel ON- and OFFchannels carrying positive signals only, and subsequently multiplied in four separate multipliers. Their output signals were then added with the appropriate sign, with outputs from ON–ON and OFF–OFF multipliers having positive, and the ones from ON–OFF and OFF– ON units having negative sign (Hassenstein & Reichardt, 1956). This looks similar to what we find in the fly, except that the fly seems to discard two of the four units, reducing a true multiplication to a ‘2-quadrant-multiplier’. In the context of motion detection, however, this does not have any functional consequences: because motion of rigid objects gives rise to spatial correlations between time-delayed signals at subsequent image pixels, an ON-event at one location will always correlate with an ON-event at a later point in time at a neighbouring location, and the same holds true for OFF-events. In other words: no correlations between ON- and OFFevents at neighbouring locations will be indicative for physically moving objects. Consequently, a comparison of a 4- and a 2-quadrant detector reveals no significant differences in response to realistic visual motion (Eichner et al., 2011).
Motion-sensitive small-field neurons Having identified lamina neurons L1 and L2–4 as the decisive input elements to two parallel motion-detector circuits allowed anatomy to guide the next steps. Indeed, two parallel processing streams had previously been postulated in the fly visual system, based on careful investigation of co-stratification of Golgi-stained columnar cells (Bausenwein et al., 1992), as well as cell-unspecific activity labelling using the deoxy-glucose method (Bausenwein & Fischbach, 1992). These studies indicated that an L1 pathway should indirectly lead to columnar T4 cells, and an L2 pathway to columnar T5 cells, with both T4 and T5 cells projecting into the lobula plate. Given that four different subtypes of T4 and T5 cells exist terminating in four different layers of the lobula plate (Fischbach & Dittrich, 1989), and that deoxy-glucose labelling indicated activity in one of the four layers according to the direction of the stimulus (Buchner et al., 1984), T4 and T5 cells were prime candidates for local elementary motion detectors and for representing the output signals of the ON- and the OFF-motion pathway, respectively. A recent study (Maisak et al., 2013) used driver lines specific for T4 and T5 cells (Fig. 3A), and combined them with an effector line expressing a high-sensitivity genetically encoded calcium indicator GCaMP5 (Akerboom et al., 2012). Using 2-photon calcium imaging as a proxy for recording their membrane potential (Egelhaaf & Borst, 1995; Haag & Borst, 2000) and stimulating the flies with grating motion in four cardinal directions (front-to-back, back-to-front, upwards, downwards), Maisak and colleagues recorded directionally selective activity from T4 and T5 cells in each one of the four lobula plate layers depending on the direction in which the grating was moving (Fig. 3B). To assess the particular contribution of T4 and T5 cells to the signals observed in the above experiments, driver lines specific for T4 or T5 cells were used, respectively. Applying the same stimulus protocol and data evaluation as before, identical results were obtained for both the T4- as well as the T5-specific driver line. Further experiments revealed similar response properties for T4 and T5 cells with respect to their orientation and velocity tuning (Maisak et al., 2013). If, however, instead of gratings, moving edges with either positive or negative contrast polarity were used as visual stimuli, T4 cells were found to strongly and selectively respond to moving ON-edges, with little or no responses to moving OFF-edges, while T5 cells selectively responded to moving OFF-edges and mostly failed to respond to moving ON-edges (Fig. 3C,D). To investigate whether the specific response properties of T4 and T5 cells are visible in postsynaptic lobula plate cells and visually driven behaviour, T4 and T5 cells were genetically blocked and flies subsequently tested. Blocking both T4 and T5 cells led to a complete loss of the motion response in lobula plate tangential cells (Schnell et al., 2012) and of the optomotor response of tethered walking flies (Bahl et al., 2013). Blocking T4 cells specifically led to selective loss of the responses to moving ON-edges, in the electrical signal of tangential cells as well as in optomotor behaviour. Conversely, blocking T5 cells led to a loss of the responses to moving OFF-edges, again both in the electrical signal of tangential cells as well as in optomotor behaviour (Maisak et al., 2013). In summary, the selective defects of T4 block flies for ON-edges and of T5 block flies for OFF-edges not only corroborate the above findings about the selective preference of T4 and T5 cells for different contrast polarities, but also demonstrate that the signals of T4 and T5 cells are indeed the major, if not exclusive, inputs to downstream circuits and motion-driven behaviours.
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
In search of the holy grail of fly motion vision 5 A
C
B
D
Fig. 3. T4 and T5 cells are local, motion-sensitive neurons of the fly visual system. (A) Confocal image of a driver line giving rise to expression in both T4 and T5 cells, shown as a horizontal cross-section (modified from Schnell et al., 2012). Neurons are marked in green (Kir2.1-EGFP labelled), revealing clear labelling (in green) in the medulla (T4 cell dendrites), in the lobula (T5 cell dendrites), as well as in four distinct layers of the lobula plate, representing the terminal arborizations of the four subpopulations of both T4 and T5 cells. The neuropile is stained in red by an antibody against the postsynaptic protein Dlg. The image within the yellow square is from a 2-photon microscope and contains the area shown enlarged to the right. (B) 2-Photon calcium imaging in the lobula plate reveals directionally selective signals. Following stimulation of the fly with grating motion along four cardinal directions (front-to-back, back-to-front, upwards and downwards), activity is confined to one of the four layers, depending on the direction in which the grating is moving (modified from Maisak et al., 2013). (C and D) Responses of T4 (C) and T5 (D) cells to ON- and OFF-edges moving along all four cardinal directions. ON (white) and OFF (black) responses within each layer are significantly different from each other with P < 0.005, except for layers 3 and 4 in T5 cells where P < 0.05 (from Maisak et al., 2013).
Medulla interneurons As an immediate hypothesis derived from the above results, the dendrites of T4 and T5 cells could be the place where the signals from neighbouring facets converge, giving rise to T4 and T5 cells direction selectivity. In the case of T4 cells, these cells have recently been described anatomically: using large-scale serial sectioning transmission electron microscopy and computer-aided reconstruction, Takemura and colleagues found two neurons, Mi1 and Tm3, to connect the output terminals of L1 neurons to the dendrites of T4 cells (Takemura et al., 2013). They furthermore observed an asymmetry of the dendrites of T4 cells in their wiring to Tm3 and Mi1 going along with the lobula plate layer they project to, i.e. their direction selectivity. This raised the hypothesis that these two cells represent the two input lines to the multiplier of the Hassenstein–Reichardt detector one cell carrying the low-pass, the other the high-pass filtered signal (Takemura et al., 2013). In a similar approach as in Takemura et al. (2013), the input neurons to T5 cells have also recently been identified anatomically. Here, four different medulla neurons, Tm1, Tm2, Tm4 and Tm9, were identified to provide the major input to the dendrites of T5 cells (Shinomiya et al., 2014). Obviously, there is no immediate correspondence between such a scenario and the Hassenstein–Reichardt model. Physiological experiments demonstrated that one of these input cells, Tm2, is absolutely necessary for OFF-responses of the lobula plate tangential cells: with Tm2 cells blocked, tangential cells no longer respond to moving OFF-edges, while their responses to moving ON-edges were only mildly affected (Meier et al., 2014).
In order to clarify the exact functional role of each of the anatomically identified input neurons to T4 and T5 cells, their visual response properties are of importance as these should have pronounced consequences for the outcome of subsequent processing. Here, the first question concerns their response to contrast changes. From all of the above, it would be expected that neurons of the ON-pathway should respond to brightness increments with an increased activity, while neurons of the OFF-pathway should be activated by brightness decrements. This indeed has been shown for four of the anatomically identified interneurons studied so far: Mi1, Tm1, Tm2 and Tm3. Using genetically encoded calcium indicators, two studies found that indeed Mi1 is excited by brightness increments, while Tm1 and Tm2 respond exclusively to brightness decrements (Meier et al., 2014; Strother et al., 2014). Furthermore, a recent study reports successful whole-cell patch recording from the somata of medulla interneurons (Behnia et al., 2014). The authors corroborate the selectivity of Mi1, Tm1 and 2 for ON- vs. OFF-signals and extend it to Tm3: as suggested by the calcium signals, the membrane potential of the cells shows a pronounced depolarization of Mi1 and Tm3 in response to ON-; and of Tm1 and Tm2 in response to OFF-signals with only a slight hyperpolarization for the opposite stimulus polarity (Fig. 4). Thus, as postulated previously based on L1 vs. L2 silencing experiments (Joesch et al., 2010; Eichner et al., 2011), interneurons of the ON- and OFF-pathway go in exact counter-phase and exhibit a pronounced half-wave rectification. Looking carefully at the dynamics of these four cells revealed an overall low-pass characteristic with slight differences in the timeconstants between Mi1 and Tm3, and Tm1 and Tm2, respectively,
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
6 A. Borst A
B
Fig. 4. Visual response properties of medulla interneurons postsynaptic to L1 and L2 (modified from Behnia et al., 2014). (A) Changes in membrane potential recorded from Mi1 (top) and Tm3 (bottom) in response to fieldflicker stimuli. (B) Changes in membrane potential recorded from Tm1 (top) and Tm2 (bottom) in response to field-flicker stimuli.
in the range of 15 ms (Behnia et al., 2014). The authors propose that Mi1 and Tm3 perform critical processing of the delayed and non-delayed input channels of the correlator responsible for the detection of light edges, while Tm1 and Tm2 play analogous roles in the detection of moving dark edges (Behnia et al., 2014). However, the delay might alternatively be implemented in the dendritic membrane receptors of T4 and T5 cells: as shown by RNA-profiling, T4 and T5 cells express both nicotinic and muscarinic acetylcholine receptors that could mediate a fast ionotropic, and a slow metabotropic signal (Shinomiya et al., 2014). Further differences between the different cells exist with respect to their receptive field properties: while Mi1 is strongly sensitive to
full-field brightness changes (‘field-flicker’; Strother et al., 2014), Tm1 and Tm2 receive a pronounced surround inhibition reducing their responses to such stimuli (Meier et al., 2014; Strother et al., 2014). An antagonistic centre-surround organization has also been observed in the terminals of L2 cells (Freifeld et al., 2013). An immediate question concerns the mechanism by which the different preference for contrast polarity in the ON- and the OFF-pathway arises. Because both L1 and L2 neurons hyperpolarize in response to light, with a rebound excitation at the end of the stimulus (Laughlin et al., 1987), medulla neurons of the ON-pathway should be inhibited by L1, while those of the OFF-pathway should receive an excitatory signal from L2–4 (Takemura et al., 2011). Calcium measurements from the axon terminals of L1 neurons demonstrate a change in calcium of equal magnitude to brightness increments and decrements (Clark et al., 2011), while axon terminals of L2 neurons show no or little decrease in calcium in response to brightness increments, but a pronounced increase in calcium in response to brightness decrements (Reiff et al., 2010; Clark et al., 2011). At the level of medulla neurons, a strong response selectivity for either ON- or OFF-signals is observed (Behnia et al., 2014; Meier et al., 2014; Strother et al., 2014). This suggests that half-wave rectification in the OFF-pathway starts already at the output of L2, whereas postsynaptic mechanisms seem to be involved in the ON-pathway.
Implementing motion opponency As mentioned in the Introduction, the final processing stage in the Hassenstein–Reichardt detector comprises the subtraction of oppositely tuned motion detectors. If both subunits are mirror-images of
A
B
C
D
Fig. 5. Probing synaptic connectivity between T4/T5 and lobula plate tangential cells. (A) Schematic to illustrate the anatomical layout of fly optic lobe. One tangential cell of the vertical system (VS) is shown in green with a recording electrode; the dendrites arborize in layer 4 of the lobula plate. Examples of two T4 and two T5 cells are depicted in red that receive input onto their dendrites in the medulla and lobula, respectively. Individual terminals providing synaptic input to the lobula plate are located either in layer 3 or 4. Equivalent cells innervating layers 1 and 2 are omitted. A, anterior; P, posterior; M, medial; L, lateral. (B) View on a preparation from posterior onto the back of the head (right hemisphere). T4/T5 cells express mCherry-tagged ChR2–H134R (red). A VS cell is filled via a patch electrode with a fluorescent dye (green). (C) Single horizontal confocal image of Drosophila expressing ChR2–H134R-mCherry using a driver line for T4 and T5 cells. (D) Average synaptic tangential cell response to optogenetic T4/T5 cell stimulation. Recordings were done in blind flies homozygously carrying the norpA7 mutation (N = 7; with SD as shaded area). All data from Mauss et al. (2014). © 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
In search of the holy grail of fly motion vision 7 each other with identical parameters, the subtraction results in a ‘fully opponent’ detector response with identical time course and amplitude, but different signs for opposite directions of motion. How is such a subtraction implemented in biophysical terms? Here, evidence from experiments both in blow flies and in fruit flies suggests that this subtraction is realized by a synaptic push–pull organization with local motion detectors of opposite preferred direction providing excitatory and inhibitory input onto the dendrites of lobula plate tangential cells (Borst & Egelhaaf, 1990; Borst et al., 1995; Joesch et al., 2008). As one prerequisite for such a linear processing, the reversal potential of both the excitatory and inhibitory conductance needs to be sufficiently different from the resting membrane potential of the tangential cell. Furthermore, the conductance changes need to be relatively small compared with the resting conductance. Both these requirements seem to be met in the fly lobula plate (Borst & Egelhaaf, 1990). However, at present it is not known whether and how individual T4 and T5 cells can provide, at the same time, excitation to one and inhibition to the adjacent lobula plate layer. One possibility could be that there exists a separate set of local, inhibitory neurons, like inhibitory twins of the T4/T5 cells. This, however, seems unlikely as blocking of T4 and T5 cells abolishes both the preferred as well as the null direction response of lobula plate tangential cells in Drosophila (Schnell et al., 2012). As an alternative explanation, a recent study suggests that inhibition is transferred from excitatory T4/T5 signals within one layer of the lobula plate to the adjacent one via local inhibitory interneurons: when T4/T5 cells are optogenetically activated, a fast excitatory postsynaptic potential is recorded in tangential cells, followed by a delayed inhibitory postsynaptic potential (Fig. 5; Mauss et al., 2014). Current experiments aim to identify such local inhibitory interneurons and to test whether they indeed provide the inhibition to tangential cells during null direction motion.
Conclusion and outlook Local motion detection in Drosophila has long been shown to follow the predictions of the Hassenstein–Reichardt detector. However, only the last few years have witnessed a major advance in our understanding of its neural implementation. Making use of genetic targeting of individual cell types and expressing proteins that either block synaptic transmission, signal neural activity or allow for neural activation by light, together with detailed analysis of serial electron microscopic image stacks, led to the following picture (Fig. 6). Starting with L1 and L2–4, the visual input is split into separate ON- and OFF-pathways, with medulla neurons Mi1 and Tm3 participating in the ON-pathway, and Tm1, Tm2, Tm4 and Tm9 in the OFF-pathway. Motion along all four cardinal directions is computed separately within each pathway. The output signals of these eight different motion detectors are then sorted such that ON (T4) and OFF (T5) motion detectors with the same directional tuning converge in the same layer of the lobula plate, jointly providing the input to downstream circuits and motion-driven behaviours. These findings are in striking similarity to motion processing in the mammalian retina, where the photoreceptor output is split into ON- and OFF-pathways as well, and ON–OFF direction-selective ganglion cells are found in four variants with each group responding preferentially to one of the four cardinal directions (for review, see Borst & Euler, 2011). Nevertheless, the core computation that leads from non-directional signals to direction selectivity is not yet understood, neither in the fly optic lobe nor in the mammalian retina. In the mammalian retina, so-called ‘starburst amacrine cells’ with radially symmetric dendrites
Fig. 6. Circuit diagram of the neurons involved in the fly elementary motion detection. Visual input from photoreceptors R1–6 is split into parallel pathways at the level of the lamina. The ON-pathway (to the right) is shown to involve lamina neuron L1 and two postsynaptic cells, Mi1 and Tm3, in the medulla. These cells contact the dendrites of T4 cells. The OFF-pathway (to the left) involves more neurons. Here lamina cells L2 and L4 synapse onto medulla neurons Tm1, Tm2 and Tm4. In addition, lamina cell L3 synapses onto Tm9. All four medulla neurons contact the dendrites of T5 cells. Directionally selective signals are carried via T4 and T5 cells to the four layers of the lobula plate where T4 and T5 cells with the same preferred direction converge again on the dendrites of the tangential cells (in yellow). Inhibition is conveyed via hypothetical, local interneurons from one layer to the adjacent one (in red).
are the first neurons to show direction selectivity. However, in contrast to T4 and T5 cells, each of which exist four subgroups according to their preferred direction, only two groups of starburst amacrine cells exist, one in the ON- and one in the OFF-layer, each one representing all directions of motion on their individual dendrites with centrifugal as their preferred direction (Euler et al., 2002). With the distal tip of their dendrites, starburst amacrine cells form selective, inhibitory contacts onto the four groups of ON–OFF ganglion cells (Briggman et al., 2011), giving rise to their direction selectivity (Yonehara et al., 2013). A recent connectomic study suggests that starburst amacrine cells receive input from different types of bipolar cells on their proximal and their distal dendrites (Kim et al., 2014). Because another study had previously shown that these two types of bipolar cells exhibit different dynamics in their calcium signal upon visual stimulation (Baden et al., 2013), the authors propose that those two types of bipolar cells correspond to the slow and fast input to the multiplier of the Hassenstein–Reichardt detector, similar to the idea that Mi1 and Tm3, and Tm1 and Tm2 represent the two differentially filtered input signals to the multiplier in the fly visual system (Behnia et al., 2014). It will be most interesting to test these ideas experimentally in both systems in order to see whether, indeed, the similarities in motion sensing between flies and
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
8 A. Borst mice even extend to the core computation of direction selectivity. In addition, the analysis of direction selectivity in Drosophila might be regarded as a case study, where methods are developed that might be applied in the future to other brain regions and behaviours as well.
Acknowledgements The author is thankful to all lab members, present and past, for sharing enthusiasm for fly motion vision and contributing so much to the progress, and to Alex Mauss, Alexander Arenz, Juergen Haag and Aljoscha Leonhardt for carefully reading the manuscript.
References Akerboom, J., Chen, T.W., Wardill, T.J., Tian, L., Marvin, J.S., Mutlu, S., Calderón, N.C., Esposti, F., Borghuis, B.G., Sun, X.R., Gordus, A., Orger, M.B., Portugues, R., Engert, F., Macklin, J.J., Filosa, A., Aggarwal, A., Kerr, R.A., Takagi, R., Kracun, S., Shigetomi, E., Khakh, B.S., Baier, H., Lagnado, L., Wang, S.S.-H., Bargmann, C.I., Kimmel, B.E., Jayaraman, V., Svoboda, K., Kim, D.S., Schreiter, E.R. & Looger, L.L. (2012) Optimization of a GCaMP calcium indicator for neural activity imaging. J. Neurosci., 32, 13819–13840. Anstis, S.M. & Rogers, B.J. (1975) Illusory reversal of visual depth and movement during changes of contrast. Vision Res., 15, 957–961. Baden, T., Berens, P., Bethge, M. & Euler, T. (2013) Spikes in the mammalian bipolar cells support temporal layering of the inner retina. Curr. Biol., 23, 48–52. Bahl, A., Ammer, G., Schilling, T. & Borst, A. (2013) Object tracking in motion-blind flies. Nat. Neurosci., 16, 730–738. Bausenwein, B. & Fischbach, K.F. (1992) Activity labeling patterns in the medulla of Drosophila melanogaster caused by motion stimuli. Cell Tissue Res., 270, 25–35. Bausenwein, B., Dittrich, A.P.M. & Fischbach, K.F. (1992) The optic lobe of Drosophila melanogaster. II. Sorting of retinotopic pathways in the medulla. Cell Tissue Res., 267, 17–28. Behnia, R., Clark, D.A., Carter, A.G., Clandinin, T.R. & Desplan, C. (2014) Processing properties of ON and OFF pathways for Drosophila motion detection. Nature, 512, 427–430. Borst, A. (2009) Drosophila’s view on insect vision. Curr. Biol., 19, R36–R47. Borst, A. & Bahde, S. (1986) What kind of movement detector is triggering the landing response of the housefly? Biol. Cybern., 55, 59–69. Borst, A. & Egelhaaf, M. (1987) Temporal modulation of luminance adapts time constant of fly movement detectors. Biol. Cybern., 56, 209–215. Borst, A. & Egelhaaf, M. (1989) Principles of visual motion detection. Trends Neurosci., 12, 297–306. Borst, A. & Egelhaaf, M. (1990) Direction selectivity of fly motion-sensitive neurons is computed in a two-stage process. Proc. Natl. Acad. Sci. USA, 87, 9363–9367. Borst, A. & Euler, T. (2011) Seeing things in motion: models, circuits, and mechanisms. Neuron, 71, 974–994. Borst, A., Egelhaaf, M. & Haag, J. (1995) Mechanisms of dendritic integration underlying gain control in fly motion-sensitive interneurons. J. Comput. Neurosci., 2, 5–18. Borst, A., Reisenman, C. & Haag, J. (2003) Adaptation of response transients in fly motion vision. II: model studies. Vision Res., 43, 1309–1322. Borst, A., Flanagin, V. & Sompolinsky, H. (2005) Adaptation without parameter change: dynamic gain control in motion detection. Proc. Natl. Acad. Sci. USA, 102, 6172–6176. Brand, A.H. & Perrimon, N. (1993) Targeted gene expression as a means of altering cell fates and generating dominant phenotypes. Development, 118, 401–415. Brenner, N., Bialek, W. & de Ruyter van Steveninck, R. (2000) Adaptive rescaling maximizes information transmission. Neuron, 26, 695–702. Briggman, K.L., Helmstaedter, M. & Denk, W. (2011) Wiring specificity in the direction-selectivity circuit of the retina. Nature, 471, 183–188. Buchner, E. (1976) Elementary movement detectors in an insect visual system. Biol. Cybern., 24, 86–101. Buchner, E., Buchner, S. & B€ulthoff, I. (1984) Deoxyglucose mapping of nervous activity induced in Drosophila brain by visual movement. J. Comp. Physiol. A., 155, 471–483.
Cajal, S.R. & Sanchez, D. (1915) Contribucion al conocimiento de los centros nerviosos de los insectos. Imprenta de Hijos de Nicholas Moja, Madrid. Chiappe, M.E., Seelig, J.D., Reiser, M.B. & Jayaraman, V. (2010) Walking modulates speed sensitivity in Drosophila motion vision. Curr. Biol., 20, 1470–1475. Clark, D.A., Bursztyn, L., Horowitz, M.A., Schnitzer, M.J. & Clandinin, T.R. (2011) Defining the computational structure of the motion detector in Drosophila. Neuron, 70, 1165–1177. Eckert, H. (1973) Optomotorische Untersuchungen am visuellen System der Stubenfliege Musca domestica L. Kybernetik, 14, 1–23. Egelhaaf, M. & Borst, A. (1989) Transient and steady-state response properties of movement detectors. J. Opt. Soc. Am. A., 6, 116–127. Egelhaaf, M. & Borst, A. (1992) Are there separate on- and off-channels in fly motion vision? Visual Neurosci., 8, 151–164. Egelhaaf, M. & Borst, A. (1995) Calcium accumulation in visual interneurons of the fly: stimulus dependence and relationship to membrane potential. J. Neurophysiol., 73, 2540–2552. Eichner, H., Joesch, M., Schnell, B., Reiff, D.F. & Borst, A. (2011) Internal structure of the fly elementary motion detector. Neuron, 70, 1155–1164. Euler, T., Detwiler, P.D. & Denk, W. (2002) Directionally selective calcium signals in dendrites of starburst amacrine cells. Nature, 418, 845–852. Fermi, G. & Reichardt, W. (1963) Optomotorische Reaktionen der Fliege Musca domestica. Kybernetik, 2, 15–28. Fischbach, K.F. & Dittrich, A.P.M. (1989) The optic lobe of Drosophila melanogaster. I. A golgi analysis of wild-type structure. Cell Tissue Res., 258, 441–475. Freifeld, L., Clark, D.A., Schnitzer, M.J., Horowitz, M.A. & Clandinin, T.R. (2013) GABAergic lateral interactions tune the early stages of visual processing in Drosophila. Neuron, 78, 1075–1089. Goetz, K.G. (1964) Optomotorische Untersuchung des visuellen Systems einiger Augenmutanten der Fruchtfliege Drosophila. Kybernetik, 2, 77–92. € Goetz, K.G. (1965) Die optischen Ubertragungseigenschaften der Komplexaugen von Drosophila. Kybernetik, 2, 215–221. Haag, J. & Borst, A. (2000) Spatial distribution and characteristics of voltage-gated calcium currents within visual interneurons. J. Neurophysiol., 83, 1039–1051. Haag, J., Denk, W. & Borst, A. (2004) Fly motion vision is based on Reichardt detectors regardless of the signal-to-noise ratio. Proc. Natl. Acad. Sci. USA, 101, 16333–16338. Hassenstein, B. & Reichardt, W. (1956) Systemtheoretische analyse der Zeit-, Reihenfolgen- und Vorzeichenauswertung bei der Bewegungsperzeption des R€ usselk€afers Chlorophanus. Z. Naturforsch., 11b, 513–524. Heisenberg, M. & Buchner, E. (1977) The role of retinula cell types in visual behavior of Drosophila melanogaster. J. Comp. Physiol. A., 117, 127–162. Joesch, M., Plett, J., Borst, A. & Reiff, D.F. (2008) Response properties of motion-sensitive visual interneurons in the lobula plate of Drosophila melanogaster. Curr. Biol., 18, 368–374. Joesch, M., Schnell, B., Raghu, S., Reiff, D.F. & Borst, A. (2010) ON and OFF pathways in Drosophila motion vision. Nature, 468, 300–304. Joesch, M., Weber, F., Eichner, H. & Borst, A. (2013) Functional specialization of parallel motion detection circuits in the fly. J. Neurosci., 33, 902–905. Jung, S.N., Borst, A. & Haag, J. (2011) Flight activity alters velocity tuning of fly motion-sensitive neurons. J. Neurosci., 31, 9231–9237. Kim, J.S., Greene, M.J., Zlateski, A., Lee, K., Richardson, M., Turaga, S.C., Purcaro, M., Balkam, M., Robinson, A., Behabadi, B.F., Campos, M., Denk, W., Seung, H.S. & the EyeWirers (2014) Space-time wiring specificity supports direction selectivity in the retina. Nature, 509, 331–336. Laughlin, S.B., Howard, J. & Blakeslee, B. (1987) Synaptic limitations to contrast coding in the retina of the blowfly Calliphora. P. Roy. Soc. BBiol. Sci., 231, 437–467. Longden, K.D. & Krapp, H.G. (2009) State-dependent performance of opticflow processing interneurons. J. Neurophysiol., 102, 3606–3618. Maimon, G., Straw, A.D. & Dickinson, M.H. (2010) Active flight increases the gain of visual motion processing in Drosophila. Nat. Neurosci., 13, 393–399. Maisak, M.S., Haag, J., Ammer, G., Serbe, E., Meier, M., Leonhardt, A., Schilling, T., Bahl, A., Rubin, G.M., Nern, A., Dickson, B.J., Reiff, D.F., Hopp, E. & Borst, A. (2013) A directional tuning map of Drosophila elementary motion detectors. Nature, 500, 212–216. Mauss, A., Meier, M., Serbe, E. & Borst, A. (2014) Optogenetic and pharmacologic dissection of feedforward inhibition in Drosophila motion vision. J. Neurosci., 34, 2254–2263.
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
In search of the holy grail of fly motion vision 9 Meier, M., Serbe, E., Maisak, M.S., Haag, J., Dickson, B.J. & Borst, A. (2014) Neural circuit components of the Drosophila OFF motion vision pathway. Curr. Biol., 24, 385–392. Meinertzhagen, I.A. & O’Neil, S.D. (1991) Synaptic organization of columnar elements in the lamina of the wild type in Drosophila melanogaster. J. Comp. Neurol., 305, 232–263. Pfeiffer, B.D., Jenett, A., Hammonds, A.S., Ngo, T.-T.B., Misra, S., Murphy, C., Scully, A., Carlson, J.W., Wan, K.H., Laverty, T.R., Mungall, C., Svirskas, R., Kadonaga, J.T., Doe, C.Q., Eisen, M.B., Celniker, S.E. & Rubin, G.M. (2008) Tools for neuroanatomy and neurogenetics in Drosophila. Proc. Natl. Acad. Sci. USA, 105, 9715–9720. Pick, B. & Buchner, E. (1979) Visual movement detection under light- and dark-adaptation in the fly, Musca domestica. J. Comp. Physiol., 134, 45–54. Reichardt, W. (1961) Autocorrelation, a principle for the evaluation of sensory information by the central nervous system. In Rosenblith, W.A. (Ed.), Sensory Communication. The M.I.T. Press and John Wiley & Sons, New York, London, pp. 303–317. Reichardt, W. (1987) Evaluation of optical motion information by movement detectors. J. Comp. Physiol. A., 161, 533–547. Reiff, D.F., Plett, J., Mank, M., Griesbeck, O. & Borst, A. (2010) Visualizing retinotopic half-wave rectified input to the motion detection circuitry of Drosophila. Nat. Neurosci., 13, 973–978. Reisenman, C., Haag, J. & Borst, A. (2003) Adaptation of response transients in fly motion vision. I: experiments. Vision Res., 43, 1291–1307. Riehle, A. & Franceschini, N. (1984) Motion detection flies: parametric control over ON-OFF pathways. Exp. Brain Res., 54, 390–394. Rister, J., Pauls, D., Schnell, B., Ting, C.Y., Lee, C.H., Sinakevitch, I., Morante, J., Strausfeld, N.J., Ito, K. & Heisenberg, M. (2007) Dissection of the peripheral motion channel in the visual system of Drosophila melanogaster. Neuron, 56, 155–170. Rosner, R., Egelhaaf, M. & Warzecha, A.-K. (2010) Behavioural state affects motion-sensitive neurones in the fly visual system. J. Exp. Biol., 213, 331–338. Safran, M., Flanagin, V., Borst, A. & Sompolinsky, H. (2007) Adaptation and information transmission in fly motion detection. J. Neurophysiol., 98, 3309–3320. Schnell, B., Joesch, M., Forstner, F., Raghu, S., Otsuna, H., Ito, K., Borst, A. & Rieff, D.F. (2010) Processing of horizontal optic flow in three visual interneurons of the Drosophila brain. J. Neurophysiol., 103, 1646–1657.
Schnell, B., Raghu, S., Nern, A. & Borst, A. (2012) Columnar cells necessary for motion responses of wide-field visual interneurons in Drosophila. J. Comp. Physiol. A., 198, 389–395. Schuling, F.H., Mastebroek, H.A.K., Bult, R. & Lenting, B.P.M. (1989) Properties o elementary movement detectors in the fly Calliphora erythrocephala. J. Comp. Physiol. A., 165, 179–192. Shinomiya, K., Karuppudurai, T., Lin, T.-Y., Lu, Z., Lee, C.-H. & Meinertzhagen, I.A. (2014) Candidate neural substrates of Off-edge motion detection in Drosophila. Curr. Biol., 24, 1062–1070. Silies, M., Gohl, D.M., Fisher, Y.E., Freifeld, L., Clark, D.A. & Clandinin, T.R. (2013) Modular use of peripheral input channels tunes motion-detecting circuitry. Neuron, 79, 111–127. Strausfeld, N.J. (1976) Atlas of an Insect Brain. Springer, Berlin, Heidelberg. Strother, J.A., Nern, A. & Reiser, M.B. (2014) Direct observation of ON and OFF pathways in the Drosophila visual system. Curr. Biol., 24, 976–983. Takemura, S.Y., Lu, Z. & Meinertzhagen, I.A. (2008) Synaptic circuits of the Drosophila optic lobe: the input terminals to the medulla. J. Comp. Neurol., 509, 493–513. Takemura, S.Y., Karuppudurai, T., Ting, C.-Y., Lu, Z., Lee, C.-H. & Meinertzhagen, I.A. (2011) Cholinergic circuits integrate neighboring visual signals in a Drosophila motion detection pathway. Curr. Biol., 21, 2077– 2084. Takemura, S.Y., Bharioke, A., Lu, Z., Nern, A., Vitaladevuni, S., Rivlin, P.K., Katz, W.T., Olbris, D.J., Plaza, S.M., Winston, P., Zhao, T., Horne, J.A., Fetter, R.D., Takemura, S., Blazek, K., Chang, L.-A., Ogundeyi, O., Saunders, M.A., Shapiro, V., Sigmund, C., Rubin, G.M., Scheffer, L.K., Meinertzhagen, I.A. & Chklovskii, D.B. (2013) A visual motion detection circuit suggested by Drosophila connectomics. Nature, 500, 175–181. Tuthill, J.C., Nern, A., Holtz, S.L., Rubin, G.M. & Reiser, M.B. (2013) Contributions of the 12 neuron classes in the fly lamina to motion vision. Neuron, 79, 128–140. Venken, K.J.T., Simpson, J.H. & Bellen, H.J. (2011) Genetic manipulation of genes and cells in the nervous system of the fruit fly. Neuron, 72, 202–230. Yonehara, K., Farrow, K., Ghanem, A., Hillier, D., Balint, K., Teixeira, M., Juettner, J., Noda, M., Neve, R.L., Conzelmann, K.-K. & Roska, B. (2013) The first stage of cardinal direction selectivity is localized to the dendrites of retinal ganglion cells. Neuron, 79, 1078–1085.
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
doi:10.1111/ejn.12731
REVIEW In search of the holy grail of fly motion vision Alexander Borst Department of Circuits, Computation, Models, Max-Planck-Institute of Neurobiology, Am Klopferspitz 18, 82152 Martinsried, Germany Keywords: calcium imaging, insect vision, neural computation, neurogenetics
Abstract Detecting the direction of image motion is important for visual navigation as well as predator, prey and mate detection and, thus, essential for the survival of all animals that have eyes. However, the direction of motion is not explicitly represented at the level of the photoreceptors: it rather needs to be computed by subsequent neural circuits, involving a comparison of the signals from neighbouring photoreceptors over time. The exact nature of this process as implemented at the neuronal level has been a longstanding question in the field. Only recently, much progress has been made in Drosophila by genetically targeting individual neuron types to block, activate or record from them. The results obtained this way indicate that: (i) luminance information from fly photoreceptors R1–6 is split into two parallel motion circuits, specialized to detect the motion of luminance increments (ON-Channel) and decrements (OFF-Channel) separately; (ii) lamina neurons L1 and L2 are the primary input neurons to these circuits (L1 ? ON-channel, L2 ? OFF-channel); and (iii) T4 and T5 cells carry their output signals (ON ? T4, OFF ? T5).
Introduction Determining the direction in which something is moving seems like a trivial task. However, while the direction of motion is certainly implicit in an image sequence, it is not explicitly encoded at the level of single photoreceptors. Making it explicit requires computation. As depicted in Fig. 1A, when a bright bar moves from left to right and back again in front of a fly’s eye, the electrical response of a single photoreceptor would look just the same. Thus, from the response of an individual photoreceptor, one cannot tell the direction of local image motion. The fly’s optic lobe consists of four different neuropiles called lamina, medulla, lobula and lobula plate. All these neuropiles are built from repetitive columns, arranged in a retinotopic way. At the level of the lobula plate, large tangential cells are found that cover many hundreds of columns with their dendrites. If the electrical signals from such neurons are recorded during the same stimulation protocol as above, a very different response is found: during bar motion to the right (their ‘preferred direction’); these cells depolarize, during bar motion to the left (their ‘null direction’), they hyperpolarize. In contrast to the photoreceptor signal, these cells encode the direction of image motion in their membrane potential. Obviously, somewhere between the photoreceptor terminals and the dendrites of the lobula plate tangential cells, the direction of motion has been computed. This process is called ‘elementary motion detection’ and is the central topic of interest here.
The Hassenstein–Reichardt detector Analysing the turning tendency of the beetle Chlorophanus viridis walking on a spherical Y-maze, Hassenstein & Reichardt (1956)
Correspondence: Dr A. Borst, as above. E-mail: [email protected] Received 16 July 2014, revised 19 August 2014, accepted 20 August 2014
proposed a specific model of elementary motion detection that could account for their observations in a quantitative way (Fig. 1B). The idea was that the beetle’s nervous system contains many hundreds of such elementary units, which, collectively, cover the whole visual field, each extracting locally the direction of image motion. The algorithmic model for such an elementary motion detector consists of two subunits, which are mirror-symmetrical to each other (Fig. 1B, left; Reichardt, 1961, 1987; Borst & Egelhaaf, 1989). Each subunit reads the luminance values measured in two adjacent image pixels. These values pass an input filter before they are fed through temporal filters with different characteristics (low-pass, high-pass). The differential filtering is an essential property as it creates a timelag between the signals. Next, the signals become multiplied (M). This is the first stage at which a directionally selective signal arises. When the output values of both subunits are finally subtracted, a fully opponent response is obtained: now, the response is positive for motion in one direction and negative for motion in the opposite direction. The value of this model for research in the field of motion vision can hardly be overestimated. The reason for this is that this model makes predictions that are both quantitative and counterintuitive. When, for example, a periodic grating is moving at a constant velocity, the model predicts, in contrast to a simple speedometer, a speed optimum at which the response is maximal. Furthermore, this optimum speed is proportional to the pattern wavelength. All this has indeed been demonstrated in house flies (Fermi & Reichardt, 1963; Eckert, 1973), blow flies (Haag et al., 2004) and fruit flies (Goetz, 1964, 1965; Joesch et al., 2008; Schnell et al., 2010), in behavioural experiments as well as at the level of lobula plate tangential cells. The Hassenstein–Reichardt detector also makes a number of other predictions relating to the response dynamics (Borst & Bahde, 1986; Borst et al., 2003, 2005). Again, these predictions were verified in recordings from lobula plate tangential cells of blow flies as well as of Drosophila (Egelhaaf & Borst, 1989; Brenner
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
2 A. Borst A
B
Fig. 1. (A) From non-directional to directional responses. Left: schematic of the optic lobe of Drosophila. The retina (red) built from repetitive facets is followed by four layers of neuropile, called lamina, medulla, lobula and lobula plate. Within the lobula plate, three tangential cells of the horizontal system (HS-cells) are shown as reconstructed from stained tissue. Right: schematic signals recorded from a single photoreceptor (top) and from a tangential cell (bottom) during motion of a bar. While the photoreceptor signal is non-directional, the signal of the tangential cell is strongly directional. (B) The Hassenstein–Reichardt detector of elementary motion detection (left), together with their inventors Bernhardt Hassenstein (top) and Werner Reichardt (bottom) and the behavioural set-up they used (right). The model consists of two mirror-symmetrical subunits sharing the same two inputs after passing a filter. Within each subunit, the signal from one input is processed by a temporal low-pass filter (LP) and subsequently multiplied (M) with the high-pass filtered signal (HP) derived from the neighbouring input. The signals from both subunits are subtracted ( ), giving rise to a directionally selective output.
et al., 2000; Reisenman et al., 2003; Borst et al., 2005; Joesch et al., 2008). Interestingly, the fly motion-detection system seems to adapt some of its parameters. During ongoing motion stimulation, the responses as recorded in tangential cells can be best described by a model where the time-constant of the high-pass filter shortens (Borst & Egelhaaf, 1987; Borst et al., 2003; Safran et al., 2007). Furthermore, the response of the tangential cells changes depending on the behavioural state of the animal, i.e. whether the fly is actively walking or flying or whether the fly is at rest (Longden & Krapp, 2009; Chiappe et al., 2010; Maimon et al., 2010; Rosner et al., 2010; Jung et al., 2011). Beside the filter time-constants, experimental evidence suggests the sampling base of the detector also to be variable in the fly visual system: while at high luminance levels, nearestneighbour interactions dominate (Goetz, 1964, 1965; Buchner, 1976), wide-range interactions come into play at low luminance conditions (Pick & Buchner, 1979; Schuling et al., 1989).
Neural implementation of the Hassenstein–Reichardt detector Given the evidence for motion computation according to the Hassenstein–Reichardt detector in the fly optic lobes, the question
about its neural implementation naturally arises: which neurons form the Hassenstein–Reichardt detector? Here, an almost complete catalogue of all columnar neurons between the photoreceptors and the lobula plate tangential cells has been worked out. Starting with the work of Cajal & Sanchez (1915), most of the columnar cell types of the lamina, medulla, lobula and lobula plate have meanwhile been identified and described on the basis of Golgi impregnations in the house fly (Strausfeld, 1976) as well as in Drosophila (Fischbach & Dittrich, 1989). Each lamina column (or cartridge, as it is usually referred to) contains a set of 12 cell types, connected to the photoreceptors either directly or indirectly (Meinertzhagen & O’Neil, 1991). These lamina neurons connect the photoreceptors to specific layers of the medulla (Takemura et al., 2008, 2011, 2013). In the medulla, a single column houses more than 60 different cell types. They can be roughly grouped into four groups: medulla intrinsic (‘Mi’) neurons connect different layers of the medulla to each other; transmedulla (‘Tm’) neurons connect specific layers of the medulla to various layers in the lobula; trans-medulla Y (‘TmY’) neurons connect specific layers of the medulla to various layers in the lobula and lobula plate; and bushy T4 cells connect medulla layer 10 to the four layers of the lobula plate. In addition, a similar group of bushy T5 cells connect the posterior layer of the lobula to the four layers of the lobula plate. However, the small size of all these
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
In search of the holy grail of fly motion vision 3 neurons prohibited electrophysiological recordings in most cases. Therefore, while a rather complete map of all columnar neurons has been available for a while, the visual response properties of most of them were completely unknown until recently. It is for this reason that the problem of the neural implementation of the Hassenstein– Reichardt detector has been in the field for over half a century, becoming the ‘holy grail of fly motion vision’, with only little progress for many decades. Only recently, the advent of sophisticated neurogenetic methods in Drosophila has allowed for elucidating the circuits for elementary motion detection. These techniques are all based on a two-component expression system where a so-called ‘driver line’ defining the neurons in which a certain effector gene is expressed is crossed with another line, the so-called ‘effector line’, defining what gene is expressed (Brand & Perrimon, 1993). Today, thousands of different driver lines are available, many of which reveal a high degree of specificity for expression in individual cell types of the optic lobe (Pfeiffer et al., 2008; Tuthill et al., 2013). Furthermore, many effector lines have been developed that allow for blocking the synaptic output of neurons, for activating neurons optogenetically or via temperature shift, as well as for recording from neurons via genetically encoded calcium indicators (for review, see Borst, 2009; Venken et al., 2011). Applying these techniques to the problem of local motion detection revealed: (i) that luminance information from fly photoreceptors R1–6 is split into two parallel motion circuits, specialized to detect the motion of luminance increments (ON-Channel) and decrements (OFF-Channel) separately; (ii) that lamina neurons L1 and L2 are the primary input neurons to these circuits (L1 ? ON-channel, L2 ? OFF-channel); and (iii) that T4 and T5 cells carry their output signals (ON ? T4, OFF ? T5). The Golgi shapes of the most important cellular constituents of the motion pathway in Drosophila are shown in Fig. 2.
Lamina input A single facet of the fly eye comprises two sets of photoreceptors: the peripheral photoreceptors R1–6; and the central R7,8. Early work on different mutants where either the central or the peripheral ones were defective indicated that motion vision relies primarily on R1–6, and not on R7,8 (Heisenberg & Buchner, 1977). Within the lamina, five different monopolar cells distribute the photoreceptor signal to five different layers of the medulla. While monopolar cells L1–3 are directly postsynaptic to R1–6, L4 and L5 receive polysynaptic input: L4 through L2, and L5 via an amacrine cell (Meinertzhagen & O’Neil, 1991). Amongst the monopolar cells, the two most prominent neurons are L1 and L2. This suggested that they convey the input signal to downstream motion-detection circuits. The first evidence along these lines was provided by experiments where either behavioural responses (Rister et al., 2007) or electrophysiological recordings from lobula plate tangential cells were used as a read-out (Joesch et al., 2010): blocking both L1 and L2 led to a complete abolition of the responses to gratings moving along any direction. The individual contributions of L1 and L2 were discovered in flies where either L1 or L2 was blocked and, instead of gratings, edges of different contrast polarity were moved to probe the responses of lobula plate tangential cells (Joesch et al., 2010): while control flies responded to moving brightness increments (ON-edges) and decrements (OFF-edges) with about the same amplitude, flies with lamina neurons L1 blocked completely lost their response to moving ON-edges, while their response to moving OFF-edges was largely intact. The opposite was found for flies with lamina neuron L2 blocked: here, the response to moving ON-edges was roughly as
Fig. 2. Golgi gestalt of important neurons of the motion pathway in Drosophila (Fischbach & Dittrich, 1989).
large as the one of control flies, whereas the response to moving OFF-edges was severely reduced. These results were later confirmed in a study on behavioural responses of walking fruit flies (Clark et al., 2011). More recent evidence from blocking experiments as well as anatomy demonstrates that lamina neurons L3 (Silies et al., 2013; Shinomiya et al., 2014) and L4 (Takemura et al., 2011; Meier et al., 2014) provide additional contribution to the OFF-pathway. Together, these experiments suggest that the photoreceptor input from R1–6 is split into parallel channels depending on the contrast polarity of the incoming signal: while neurons postsynaptic to L1 specifically transmit information about brightness increments to downstream motion circuits, neurons postsynaptic to L2–4 preferentially allow information about brightness decrements to be passed onto further motion-processing units. Technically, such signal processing is referred to as ‘half-wave rectification’.
Two parallel ON- and OFF-motion pathways The finding about the specific contribution of L1 and L2 to motion processing of brightness increments and decrements immediately suggests that in Drosophila, motion is detected in several separate circuits. Theoretically, there could be four such channels comprising all possible combinations between ON- and OFF-signals. However, apparent motion experiments in blow flies indicated that only two such channels exist, one for the interaction of ON-signals and one for the interaction of OFF-signals between neighbouring image points (Riehle & Franceschini, 1984). In these experiments, pairs of
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
4 A. Borst single photoreceptors were stimulated by light-ON and light-OFF flashes while recording from a motion-sensitive lobula plate neuron H1. Recordings from lobula plate tangential cells in Drosophila arrived at the same conclusion: only sequences of light pulses of the same contrast polarity (ON–ON and OFF–OFF) elicited significant responses, while light pulses of differing contrast polarity (ON–OFF and OFF–ON) failed to evoke any responses in lobula plate tangential cells (Eichner et al., 2011). Interestingly, using brightness steps instead of brightness pulses led to different results: while again, ON–ON and OFF–OFF sequences along the preferred direction of the cell led to positive responses, ON–OFF and OFF–ON sequences elicited negative responses. While this so-called ‘reverse phi’ phenomenon (Anstis & Rogers, 1975) is usually taken as evidence for an interaction between ON and OFF stimuli (Egelhaaf & Borst, 1992), as in the original Hassenstein–Reichardt detector, careful simulations of such circuits as well as experiments with flies with either lamina neurons L1 or L2 blocked revealed that such an inversion of the response is not necessarily indicative of an interaction between ON- and OFF-channels. Rather, it can be explained by the residual sustained response component of the lamina neurons taking into account the contribution of motion detectors at the edge of the stimulated areas in the visual field of the fly (Eichner et al., 2011; Joesch et al., 2013). In summary, thus, the Hassenstein–Reichardt detector is implemented in the visual system of the fly in two parallel circuits, one dealing with brightness increments (ON-channel), the other one with brightness decrements (OFF-channel). What could be the advantage of such a doubling of wiring expenditure? Without splitting and half-wave rectification, i.e. with just a single motion detector dealing with brightness going positive as well as negative, a multiplicative interaction (according to the sign rule of multiplication) would imply the output being positive when both input signals go positive as well as when both go negative. Such a processing rule seems hard to implement biophysically and indeed is unheard of in cellular neuroscience. If, however, the input signals are split according to their polarity with the negative ones being sign-inverted, subsequent circuits only deal with positive input signals reducing the problem of multiplication to a supra-linear input–output characteristic as can easily be implemented by a voltage-gated ion channel. Interestingly, neurobiology is not the only area where such problems arise: analogue electronics faced the same problem of multiplying positive as well as negative signals, and the solution was what is called a ‘4-quadrant-multiplier’. There, the input signals, prior to multiplication, were half-wave rectified, split into parallel ON- and OFFchannels carrying positive signals only, and subsequently multiplied in four separate multipliers. Their output signals were then added with the appropriate sign, with outputs from ON–ON and OFF–OFF multipliers having positive, and the ones from ON–OFF and OFF– ON units having negative sign (Hassenstein & Reichardt, 1956). This looks similar to what we find in the fly, except that the fly seems to discard two of the four units, reducing a true multiplication to a ‘2-quadrant-multiplier’. In the context of motion detection, however, this does not have any functional consequences: because motion of rigid objects gives rise to spatial correlations between time-delayed signals at subsequent image pixels, an ON-event at one location will always correlate with an ON-event at a later point in time at a neighbouring location, and the same holds true for OFF-events. In other words: no correlations between ON- and OFFevents at neighbouring locations will be indicative for physically moving objects. Consequently, a comparison of a 4- and a 2-quadrant detector reveals no significant differences in response to realistic visual motion (Eichner et al., 2011).
Motion-sensitive small-field neurons Having identified lamina neurons L1 and L2–4 as the decisive input elements to two parallel motion-detector circuits allowed anatomy to guide the next steps. Indeed, two parallel processing streams had previously been postulated in the fly visual system, based on careful investigation of co-stratification of Golgi-stained columnar cells (Bausenwein et al., 1992), as well as cell-unspecific activity labelling using the deoxy-glucose method (Bausenwein & Fischbach, 1992). These studies indicated that an L1 pathway should indirectly lead to columnar T4 cells, and an L2 pathway to columnar T5 cells, with both T4 and T5 cells projecting into the lobula plate. Given that four different subtypes of T4 and T5 cells exist terminating in four different layers of the lobula plate (Fischbach & Dittrich, 1989), and that deoxy-glucose labelling indicated activity in one of the four layers according to the direction of the stimulus (Buchner et al., 1984), T4 and T5 cells were prime candidates for local elementary motion detectors and for representing the output signals of the ON- and the OFF-motion pathway, respectively. A recent study (Maisak et al., 2013) used driver lines specific for T4 and T5 cells (Fig. 3A), and combined them with an effector line expressing a high-sensitivity genetically encoded calcium indicator GCaMP5 (Akerboom et al., 2012). Using 2-photon calcium imaging as a proxy for recording their membrane potential (Egelhaaf & Borst, 1995; Haag & Borst, 2000) and stimulating the flies with grating motion in four cardinal directions (front-to-back, back-to-front, upwards, downwards), Maisak and colleagues recorded directionally selective activity from T4 and T5 cells in each one of the four lobula plate layers depending on the direction in which the grating was moving (Fig. 3B). To assess the particular contribution of T4 and T5 cells to the signals observed in the above experiments, driver lines specific for T4 or T5 cells were used, respectively. Applying the same stimulus protocol and data evaluation as before, identical results were obtained for both the T4- as well as the T5-specific driver line. Further experiments revealed similar response properties for T4 and T5 cells with respect to their orientation and velocity tuning (Maisak et al., 2013). If, however, instead of gratings, moving edges with either positive or negative contrast polarity were used as visual stimuli, T4 cells were found to strongly and selectively respond to moving ON-edges, with little or no responses to moving OFF-edges, while T5 cells selectively responded to moving OFF-edges and mostly failed to respond to moving ON-edges (Fig. 3C,D). To investigate whether the specific response properties of T4 and T5 cells are visible in postsynaptic lobula plate cells and visually driven behaviour, T4 and T5 cells were genetically blocked and flies subsequently tested. Blocking both T4 and T5 cells led to a complete loss of the motion response in lobula plate tangential cells (Schnell et al., 2012) and of the optomotor response of tethered walking flies (Bahl et al., 2013). Blocking T4 cells specifically led to selective loss of the responses to moving ON-edges, in the electrical signal of tangential cells as well as in optomotor behaviour. Conversely, blocking T5 cells led to a loss of the responses to moving OFF-edges, again both in the electrical signal of tangential cells as well as in optomotor behaviour (Maisak et al., 2013). In summary, the selective defects of T4 block flies for ON-edges and of T5 block flies for OFF-edges not only corroborate the above findings about the selective preference of T4 and T5 cells for different contrast polarities, but also demonstrate that the signals of T4 and T5 cells are indeed the major, if not exclusive, inputs to downstream circuits and motion-driven behaviours.
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
In search of the holy grail of fly motion vision 5 A
C
B
D
Fig. 3. T4 and T5 cells are local, motion-sensitive neurons of the fly visual system. (A) Confocal image of a driver line giving rise to expression in both T4 and T5 cells, shown as a horizontal cross-section (modified from Schnell et al., 2012). Neurons are marked in green (Kir2.1-EGFP labelled), revealing clear labelling (in green) in the medulla (T4 cell dendrites), in the lobula (T5 cell dendrites), as well as in four distinct layers of the lobula plate, representing the terminal arborizations of the four subpopulations of both T4 and T5 cells. The neuropile is stained in red by an antibody against the postsynaptic protein Dlg. The image within the yellow square is from a 2-photon microscope and contains the area shown enlarged to the right. (B) 2-Photon calcium imaging in the lobula plate reveals directionally selective signals. Following stimulation of the fly with grating motion along four cardinal directions (front-to-back, back-to-front, upwards and downwards), activity is confined to one of the four layers, depending on the direction in which the grating is moving (modified from Maisak et al., 2013). (C and D) Responses of T4 (C) and T5 (D) cells to ON- and OFF-edges moving along all four cardinal directions. ON (white) and OFF (black) responses within each layer are significantly different from each other with P < 0.005, except for layers 3 and 4 in T5 cells where P < 0.05 (from Maisak et al., 2013).
Medulla interneurons As an immediate hypothesis derived from the above results, the dendrites of T4 and T5 cells could be the place where the signals from neighbouring facets converge, giving rise to T4 and T5 cells direction selectivity. In the case of T4 cells, these cells have recently been described anatomically: using large-scale serial sectioning transmission electron microscopy and computer-aided reconstruction, Takemura and colleagues found two neurons, Mi1 and Tm3, to connect the output terminals of L1 neurons to the dendrites of T4 cells (Takemura et al., 2013). They furthermore observed an asymmetry of the dendrites of T4 cells in their wiring to Tm3 and Mi1 going along with the lobula plate layer they project to, i.e. their direction selectivity. This raised the hypothesis that these two cells represent the two input lines to the multiplier of the Hassenstein–Reichardt detector one cell carrying the low-pass, the other the high-pass filtered signal (Takemura et al., 2013). In a similar approach as in Takemura et al. (2013), the input neurons to T5 cells have also recently been identified anatomically. Here, four different medulla neurons, Tm1, Tm2, Tm4 and Tm9, were identified to provide the major input to the dendrites of T5 cells (Shinomiya et al., 2014). Obviously, there is no immediate correspondence between such a scenario and the Hassenstein–Reichardt model. Physiological experiments demonstrated that one of these input cells, Tm2, is absolutely necessary for OFF-responses of the lobula plate tangential cells: with Tm2 cells blocked, tangential cells no longer respond to moving OFF-edges, while their responses to moving ON-edges were only mildly affected (Meier et al., 2014).
In order to clarify the exact functional role of each of the anatomically identified input neurons to T4 and T5 cells, their visual response properties are of importance as these should have pronounced consequences for the outcome of subsequent processing. Here, the first question concerns their response to contrast changes. From all of the above, it would be expected that neurons of the ON-pathway should respond to brightness increments with an increased activity, while neurons of the OFF-pathway should be activated by brightness decrements. This indeed has been shown for four of the anatomically identified interneurons studied so far: Mi1, Tm1, Tm2 and Tm3. Using genetically encoded calcium indicators, two studies found that indeed Mi1 is excited by brightness increments, while Tm1 and Tm2 respond exclusively to brightness decrements (Meier et al., 2014; Strother et al., 2014). Furthermore, a recent study reports successful whole-cell patch recording from the somata of medulla interneurons (Behnia et al., 2014). The authors corroborate the selectivity of Mi1, Tm1 and 2 for ON- vs. OFF-signals and extend it to Tm3: as suggested by the calcium signals, the membrane potential of the cells shows a pronounced depolarization of Mi1 and Tm3 in response to ON-; and of Tm1 and Tm2 in response to OFF-signals with only a slight hyperpolarization for the opposite stimulus polarity (Fig. 4). Thus, as postulated previously based on L1 vs. L2 silencing experiments (Joesch et al., 2010; Eichner et al., 2011), interneurons of the ON- and OFF-pathway go in exact counter-phase and exhibit a pronounced half-wave rectification. Looking carefully at the dynamics of these four cells revealed an overall low-pass characteristic with slight differences in the timeconstants between Mi1 and Tm3, and Tm1 and Tm2, respectively,
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
6 A. Borst A
B
Fig. 4. Visual response properties of medulla interneurons postsynaptic to L1 and L2 (modified from Behnia et al., 2014). (A) Changes in membrane potential recorded from Mi1 (top) and Tm3 (bottom) in response to fieldflicker stimuli. (B) Changes in membrane potential recorded from Tm1 (top) and Tm2 (bottom) in response to field-flicker stimuli.
in the range of 15 ms (Behnia et al., 2014). The authors propose that Mi1 and Tm3 perform critical processing of the delayed and non-delayed input channels of the correlator responsible for the detection of light edges, while Tm1 and Tm2 play analogous roles in the detection of moving dark edges (Behnia et al., 2014). However, the delay might alternatively be implemented in the dendritic membrane receptors of T4 and T5 cells: as shown by RNA-profiling, T4 and T5 cells express both nicotinic and muscarinic acetylcholine receptors that could mediate a fast ionotropic, and a slow metabotropic signal (Shinomiya et al., 2014). Further differences between the different cells exist with respect to their receptive field properties: while Mi1 is strongly sensitive to
full-field brightness changes (‘field-flicker’; Strother et al., 2014), Tm1 and Tm2 receive a pronounced surround inhibition reducing their responses to such stimuli (Meier et al., 2014; Strother et al., 2014). An antagonistic centre-surround organization has also been observed in the terminals of L2 cells (Freifeld et al., 2013). An immediate question concerns the mechanism by which the different preference for contrast polarity in the ON- and the OFF-pathway arises. Because both L1 and L2 neurons hyperpolarize in response to light, with a rebound excitation at the end of the stimulus (Laughlin et al., 1987), medulla neurons of the ON-pathway should be inhibited by L1, while those of the OFF-pathway should receive an excitatory signal from L2–4 (Takemura et al., 2011). Calcium measurements from the axon terminals of L1 neurons demonstrate a change in calcium of equal magnitude to brightness increments and decrements (Clark et al., 2011), while axon terminals of L2 neurons show no or little decrease in calcium in response to brightness increments, but a pronounced increase in calcium in response to brightness decrements (Reiff et al., 2010; Clark et al., 2011). At the level of medulla neurons, a strong response selectivity for either ON- or OFF-signals is observed (Behnia et al., 2014; Meier et al., 2014; Strother et al., 2014). This suggests that half-wave rectification in the OFF-pathway starts already at the output of L2, whereas postsynaptic mechanisms seem to be involved in the ON-pathway.
Implementing motion opponency As mentioned in the Introduction, the final processing stage in the Hassenstein–Reichardt detector comprises the subtraction of oppositely tuned motion detectors. If both subunits are mirror-images of
A
B
C
D
Fig. 5. Probing synaptic connectivity between T4/T5 and lobula plate tangential cells. (A) Schematic to illustrate the anatomical layout of fly optic lobe. One tangential cell of the vertical system (VS) is shown in green with a recording electrode; the dendrites arborize in layer 4 of the lobula plate. Examples of two T4 and two T5 cells are depicted in red that receive input onto their dendrites in the medulla and lobula, respectively. Individual terminals providing synaptic input to the lobula plate are located either in layer 3 or 4. Equivalent cells innervating layers 1 and 2 are omitted. A, anterior; P, posterior; M, medial; L, lateral. (B) View on a preparation from posterior onto the back of the head (right hemisphere). T4/T5 cells express mCherry-tagged ChR2–H134R (red). A VS cell is filled via a patch electrode with a fluorescent dye (green). (C) Single horizontal confocal image of Drosophila expressing ChR2–H134R-mCherry using a driver line for T4 and T5 cells. (D) Average synaptic tangential cell response to optogenetic T4/T5 cell stimulation. Recordings were done in blind flies homozygously carrying the norpA7 mutation (N = 7; with SD as shaded area). All data from Mauss et al. (2014). © 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
In search of the holy grail of fly motion vision 7 each other with identical parameters, the subtraction results in a ‘fully opponent’ detector response with identical time course and amplitude, but different signs for opposite directions of motion. How is such a subtraction implemented in biophysical terms? Here, evidence from experiments both in blow flies and in fruit flies suggests that this subtraction is realized by a synaptic push–pull organization with local motion detectors of opposite preferred direction providing excitatory and inhibitory input onto the dendrites of lobula plate tangential cells (Borst & Egelhaaf, 1990; Borst et al., 1995; Joesch et al., 2008). As one prerequisite for such a linear processing, the reversal potential of both the excitatory and inhibitory conductance needs to be sufficiently different from the resting membrane potential of the tangential cell. Furthermore, the conductance changes need to be relatively small compared with the resting conductance. Both these requirements seem to be met in the fly lobula plate (Borst & Egelhaaf, 1990). However, at present it is not known whether and how individual T4 and T5 cells can provide, at the same time, excitation to one and inhibition to the adjacent lobula plate layer. One possibility could be that there exists a separate set of local, inhibitory neurons, like inhibitory twins of the T4/T5 cells. This, however, seems unlikely as blocking of T4 and T5 cells abolishes both the preferred as well as the null direction response of lobula plate tangential cells in Drosophila (Schnell et al., 2012). As an alternative explanation, a recent study suggests that inhibition is transferred from excitatory T4/T5 signals within one layer of the lobula plate to the adjacent one via local inhibitory interneurons: when T4/T5 cells are optogenetically activated, a fast excitatory postsynaptic potential is recorded in tangential cells, followed by a delayed inhibitory postsynaptic potential (Fig. 5; Mauss et al., 2014). Current experiments aim to identify such local inhibitory interneurons and to test whether they indeed provide the inhibition to tangential cells during null direction motion.
Conclusion and outlook Local motion detection in Drosophila has long been shown to follow the predictions of the Hassenstein–Reichardt detector. However, only the last few years have witnessed a major advance in our understanding of its neural implementation. Making use of genetic targeting of individual cell types and expressing proteins that either block synaptic transmission, signal neural activity or allow for neural activation by light, together with detailed analysis of serial electron microscopic image stacks, led to the following picture (Fig. 6). Starting with L1 and L2–4, the visual input is split into separate ON- and OFF-pathways, with medulla neurons Mi1 and Tm3 participating in the ON-pathway, and Tm1, Tm2, Tm4 and Tm9 in the OFF-pathway. Motion along all four cardinal directions is computed separately within each pathway. The output signals of these eight different motion detectors are then sorted such that ON (T4) and OFF (T5) motion detectors with the same directional tuning converge in the same layer of the lobula plate, jointly providing the input to downstream circuits and motion-driven behaviours. These findings are in striking similarity to motion processing in the mammalian retina, where the photoreceptor output is split into ON- and OFF-pathways as well, and ON–OFF direction-selective ganglion cells are found in four variants with each group responding preferentially to one of the four cardinal directions (for review, see Borst & Euler, 2011). Nevertheless, the core computation that leads from non-directional signals to direction selectivity is not yet understood, neither in the fly optic lobe nor in the mammalian retina. In the mammalian retina, so-called ‘starburst amacrine cells’ with radially symmetric dendrites
Fig. 6. Circuit diagram of the neurons involved in the fly elementary motion detection. Visual input from photoreceptors R1–6 is split into parallel pathways at the level of the lamina. The ON-pathway (to the right) is shown to involve lamina neuron L1 and two postsynaptic cells, Mi1 and Tm3, in the medulla. These cells contact the dendrites of T4 cells. The OFF-pathway (to the left) involves more neurons. Here lamina cells L2 and L4 synapse onto medulla neurons Tm1, Tm2 and Tm4. In addition, lamina cell L3 synapses onto Tm9. All four medulla neurons contact the dendrites of T5 cells. Directionally selective signals are carried via T4 and T5 cells to the four layers of the lobula plate where T4 and T5 cells with the same preferred direction converge again on the dendrites of the tangential cells (in yellow). Inhibition is conveyed via hypothetical, local interneurons from one layer to the adjacent one (in red).
are the first neurons to show direction selectivity. However, in contrast to T4 and T5 cells, each of which exist four subgroups according to their preferred direction, only two groups of starburst amacrine cells exist, one in the ON- and one in the OFF-layer, each one representing all directions of motion on their individual dendrites with centrifugal as their preferred direction (Euler et al., 2002). With the distal tip of their dendrites, starburst amacrine cells form selective, inhibitory contacts onto the four groups of ON–OFF ganglion cells (Briggman et al., 2011), giving rise to their direction selectivity (Yonehara et al., 2013). A recent connectomic study suggests that starburst amacrine cells receive input from different types of bipolar cells on their proximal and their distal dendrites (Kim et al., 2014). Because another study had previously shown that these two types of bipolar cells exhibit different dynamics in their calcium signal upon visual stimulation (Baden et al., 2013), the authors propose that those two types of bipolar cells correspond to the slow and fast input to the multiplier of the Hassenstein–Reichardt detector, similar to the idea that Mi1 and Tm3, and Tm1 and Tm2 represent the two differentially filtered input signals to the multiplier in the fly visual system (Behnia et al., 2014). It will be most interesting to test these ideas experimentally in both systems in order to see whether, indeed, the similarities in motion sensing between flies and
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
8 A. Borst mice even extend to the core computation of direction selectivity. In addition, the analysis of direction selectivity in Drosophila might be regarded as a case study, where methods are developed that might be applied in the future to other brain regions and behaviours as well.
Acknowledgements The author is thankful to all lab members, present and past, for sharing enthusiasm for fly motion vision and contributing so much to the progress, and to Alex Mauss, Alexander Arenz, Juergen Haag and Aljoscha Leonhardt for carefully reading the manuscript.
References Akerboom, J., Chen, T.W., Wardill, T.J., Tian, L., Marvin, J.S., Mutlu, S., Calderón, N.C., Esposti, F., Borghuis, B.G., Sun, X.R., Gordus, A., Orger, M.B., Portugues, R., Engert, F., Macklin, J.J., Filosa, A., Aggarwal, A., Kerr, R.A., Takagi, R., Kracun, S., Shigetomi, E., Khakh, B.S., Baier, H., Lagnado, L., Wang, S.S.-H., Bargmann, C.I., Kimmel, B.E., Jayaraman, V., Svoboda, K., Kim, D.S., Schreiter, E.R. & Looger, L.L. (2012) Optimization of a GCaMP calcium indicator for neural activity imaging. J. Neurosci., 32, 13819–13840. Anstis, S.M. & Rogers, B.J. (1975) Illusory reversal of visual depth and movement during changes of contrast. Vision Res., 15, 957–961. Baden, T., Berens, P., Bethge, M. & Euler, T. (2013) Spikes in the mammalian bipolar cells support temporal layering of the inner retina. Curr. Biol., 23, 48–52. Bahl, A., Ammer, G., Schilling, T. & Borst, A. (2013) Object tracking in motion-blind flies. Nat. Neurosci., 16, 730–738. Bausenwein, B. & Fischbach, K.F. (1992) Activity labeling patterns in the medulla of Drosophila melanogaster caused by motion stimuli. Cell Tissue Res., 270, 25–35. Bausenwein, B., Dittrich, A.P.M. & Fischbach, K.F. (1992) The optic lobe of Drosophila melanogaster. II. Sorting of retinotopic pathways in the medulla. Cell Tissue Res., 267, 17–28. Behnia, R., Clark, D.A., Carter, A.G., Clandinin, T.R. & Desplan, C. (2014) Processing properties of ON and OFF pathways for Drosophila motion detection. Nature, 512, 427–430. Borst, A. (2009) Drosophila’s view on insect vision. Curr. Biol., 19, R36–R47. Borst, A. & Bahde, S. (1986) What kind of movement detector is triggering the landing response of the housefly? Biol. Cybern., 55, 59–69. Borst, A. & Egelhaaf, M. (1987) Temporal modulation of luminance adapts time constant of fly movement detectors. Biol. Cybern., 56, 209–215. Borst, A. & Egelhaaf, M. (1989) Principles of visual motion detection. Trends Neurosci., 12, 297–306. Borst, A. & Egelhaaf, M. (1990) Direction selectivity of fly motion-sensitive neurons is computed in a two-stage process. Proc. Natl. Acad. Sci. USA, 87, 9363–9367. Borst, A. & Euler, T. (2011) Seeing things in motion: models, circuits, and mechanisms. Neuron, 71, 974–994. Borst, A., Egelhaaf, M. & Haag, J. (1995) Mechanisms of dendritic integration underlying gain control in fly motion-sensitive interneurons. J. Comput. Neurosci., 2, 5–18. Borst, A., Reisenman, C. & Haag, J. (2003) Adaptation of response transients in fly motion vision. II: model studies. Vision Res., 43, 1309–1322. Borst, A., Flanagin, V. & Sompolinsky, H. (2005) Adaptation without parameter change: dynamic gain control in motion detection. Proc. Natl. Acad. Sci. USA, 102, 6172–6176. Brand, A.H. & Perrimon, N. (1993) Targeted gene expression as a means of altering cell fates and generating dominant phenotypes. Development, 118, 401–415. Brenner, N., Bialek, W. & de Ruyter van Steveninck, R. (2000) Adaptive rescaling maximizes information transmission. Neuron, 26, 695–702. Briggman, K.L., Helmstaedter, M. & Denk, W. (2011) Wiring specificity in the direction-selectivity circuit of the retina. Nature, 471, 183–188. Buchner, E. (1976) Elementary movement detectors in an insect visual system. Biol. Cybern., 24, 86–101. Buchner, E., Buchner, S. & B€ulthoff, I. (1984) Deoxyglucose mapping of nervous activity induced in Drosophila brain by visual movement. J. Comp. Physiol. A., 155, 471–483.
Cajal, S.R. & Sanchez, D. (1915) Contribucion al conocimiento de los centros nerviosos de los insectos. Imprenta de Hijos de Nicholas Moja, Madrid. Chiappe, M.E., Seelig, J.D., Reiser, M.B. & Jayaraman, V. (2010) Walking modulates speed sensitivity in Drosophila motion vision. Curr. Biol., 20, 1470–1475. Clark, D.A., Bursztyn, L., Horowitz, M.A., Schnitzer, M.J. & Clandinin, T.R. (2011) Defining the computational structure of the motion detector in Drosophila. Neuron, 70, 1165–1177. Eckert, H. (1973) Optomotorische Untersuchungen am visuellen System der Stubenfliege Musca domestica L. Kybernetik, 14, 1–23. Egelhaaf, M. & Borst, A. (1989) Transient and steady-state response properties of movement detectors. J. Opt. Soc. Am. A., 6, 116–127. Egelhaaf, M. & Borst, A. (1992) Are there separate on- and off-channels in fly motion vision? Visual Neurosci., 8, 151–164. Egelhaaf, M. & Borst, A. (1995) Calcium accumulation in visual interneurons of the fly: stimulus dependence and relationship to membrane potential. J. Neurophysiol., 73, 2540–2552. Eichner, H., Joesch, M., Schnell, B., Reiff, D.F. & Borst, A. (2011) Internal structure of the fly elementary motion detector. Neuron, 70, 1155–1164. Euler, T., Detwiler, P.D. & Denk, W. (2002) Directionally selective calcium signals in dendrites of starburst amacrine cells. Nature, 418, 845–852. Fermi, G. & Reichardt, W. (1963) Optomotorische Reaktionen der Fliege Musca domestica. Kybernetik, 2, 15–28. Fischbach, K.F. & Dittrich, A.P.M. (1989) The optic lobe of Drosophila melanogaster. I. A golgi analysis of wild-type structure. Cell Tissue Res., 258, 441–475. Freifeld, L., Clark, D.A., Schnitzer, M.J., Horowitz, M.A. & Clandinin, T.R. (2013) GABAergic lateral interactions tune the early stages of visual processing in Drosophila. Neuron, 78, 1075–1089. Goetz, K.G. (1964) Optomotorische Untersuchung des visuellen Systems einiger Augenmutanten der Fruchtfliege Drosophila. Kybernetik, 2, 77–92. € Goetz, K.G. (1965) Die optischen Ubertragungseigenschaften der Komplexaugen von Drosophila. Kybernetik, 2, 215–221. Haag, J. & Borst, A. (2000) Spatial distribution and characteristics of voltage-gated calcium currents within visual interneurons. J. Neurophysiol., 83, 1039–1051. Haag, J., Denk, W. & Borst, A. (2004) Fly motion vision is based on Reichardt detectors regardless of the signal-to-noise ratio. Proc. Natl. Acad. Sci. USA, 101, 16333–16338. Hassenstein, B. & Reichardt, W. (1956) Systemtheoretische analyse der Zeit-, Reihenfolgen- und Vorzeichenauswertung bei der Bewegungsperzeption des R€ usselk€afers Chlorophanus. Z. Naturforsch., 11b, 513–524. Heisenberg, M. & Buchner, E. (1977) The role of retinula cell types in visual behavior of Drosophila melanogaster. J. Comp. Physiol. A., 117, 127–162. Joesch, M., Plett, J., Borst, A. & Reiff, D.F. (2008) Response properties of motion-sensitive visual interneurons in the lobula plate of Drosophila melanogaster. Curr. Biol., 18, 368–374. Joesch, M., Schnell, B., Raghu, S., Reiff, D.F. & Borst, A. (2010) ON and OFF pathways in Drosophila motion vision. Nature, 468, 300–304. Joesch, M., Weber, F., Eichner, H. & Borst, A. (2013) Functional specialization of parallel motion detection circuits in the fly. J. Neurosci., 33, 902–905. Jung, S.N., Borst, A. & Haag, J. (2011) Flight activity alters velocity tuning of fly motion-sensitive neurons. J. Neurosci., 31, 9231–9237. Kim, J.S., Greene, M.J., Zlateski, A., Lee, K., Richardson, M., Turaga, S.C., Purcaro, M., Balkam, M., Robinson, A., Behabadi, B.F., Campos, M., Denk, W., Seung, H.S. & the EyeWirers (2014) Space-time wiring specificity supports direction selectivity in the retina. Nature, 509, 331–336. Laughlin, S.B., Howard, J. & Blakeslee, B. (1987) Synaptic limitations to contrast coding in the retina of the blowfly Calliphora. P. Roy. Soc. BBiol. Sci., 231, 437–467. Longden, K.D. & Krapp, H.G. (2009) State-dependent performance of opticflow processing interneurons. J. Neurophysiol., 102, 3606–3618. Maimon, G., Straw, A.D. & Dickinson, M.H. (2010) Active flight increases the gain of visual motion processing in Drosophila. Nat. Neurosci., 13, 393–399. Maisak, M.S., Haag, J., Ammer, G., Serbe, E., Meier, M., Leonhardt, A., Schilling, T., Bahl, A., Rubin, G.M., Nern, A., Dickson, B.J., Reiff, D.F., Hopp, E. & Borst, A. (2013) A directional tuning map of Drosophila elementary motion detectors. Nature, 500, 212–216. Mauss, A., Meier, M., Serbe, E. & Borst, A. (2014) Optogenetic and pharmacologic dissection of feedforward inhibition in Drosophila motion vision. J. Neurosci., 34, 2254–2263.
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
In search of the holy grail of fly motion vision 9 Meier, M., Serbe, E., Maisak, M.S., Haag, J., Dickson, B.J. & Borst, A. (2014) Neural circuit components of the Drosophila OFF motion vision pathway. Curr. Biol., 24, 385–392. Meinertzhagen, I.A. & O’Neil, S.D. (1991) Synaptic organization of columnar elements in the lamina of the wild type in Drosophila melanogaster. J. Comp. Neurol., 305, 232–263. Pfeiffer, B.D., Jenett, A., Hammonds, A.S., Ngo, T.-T.B., Misra, S., Murphy, C., Scully, A., Carlson, J.W., Wan, K.H., Laverty, T.R., Mungall, C., Svirskas, R., Kadonaga, J.T., Doe, C.Q., Eisen, M.B., Celniker, S.E. & Rubin, G.M. (2008) Tools for neuroanatomy and neurogenetics in Drosophila. Proc. Natl. Acad. Sci. USA, 105, 9715–9720. Pick, B. & Buchner, E. (1979) Visual movement detection under light- and dark-adaptation in the fly, Musca domestica. J. Comp. Physiol., 134, 45–54. Reichardt, W. (1961) Autocorrelation, a principle for the evaluation of sensory information by the central nervous system. In Rosenblith, W.A. (Ed.), Sensory Communication. The M.I.T. Press and John Wiley & Sons, New York, London, pp. 303–317. Reichardt, W. (1987) Evaluation of optical motion information by movement detectors. J. Comp. Physiol. A., 161, 533–547. Reiff, D.F., Plett, J., Mank, M., Griesbeck, O. & Borst, A. (2010) Visualizing retinotopic half-wave rectified input to the motion detection circuitry of Drosophila. Nat. Neurosci., 13, 973–978. Reisenman, C., Haag, J. & Borst, A. (2003) Adaptation of response transients in fly motion vision. I: experiments. Vision Res., 43, 1291–1307. Riehle, A. & Franceschini, N. (1984) Motion detection flies: parametric control over ON-OFF pathways. Exp. Brain Res., 54, 390–394. Rister, J., Pauls, D., Schnell, B., Ting, C.Y., Lee, C.H., Sinakevitch, I., Morante, J., Strausfeld, N.J., Ito, K. & Heisenberg, M. (2007) Dissection of the peripheral motion channel in the visual system of Drosophila melanogaster. Neuron, 56, 155–170. Rosner, R., Egelhaaf, M. & Warzecha, A.-K. (2010) Behavioural state affects motion-sensitive neurones in the fly visual system. J. Exp. Biol., 213, 331–338. Safran, M., Flanagin, V., Borst, A. & Sompolinsky, H. (2007) Adaptation and information transmission in fly motion detection. J. Neurophysiol., 98, 3309–3320. Schnell, B., Joesch, M., Forstner, F., Raghu, S., Otsuna, H., Ito, K., Borst, A. & Rieff, D.F. (2010) Processing of horizontal optic flow in three visual interneurons of the Drosophila brain. J. Neurophysiol., 103, 1646–1657.
Schnell, B., Raghu, S., Nern, A. & Borst, A. (2012) Columnar cells necessary for motion responses of wide-field visual interneurons in Drosophila. J. Comp. Physiol. A., 198, 389–395. Schuling, F.H., Mastebroek, H.A.K., Bult, R. & Lenting, B.P.M. (1989) Properties o elementary movement detectors in the fly Calliphora erythrocephala. J. Comp. Physiol. A., 165, 179–192. Shinomiya, K., Karuppudurai, T., Lin, T.-Y., Lu, Z., Lee, C.-H. & Meinertzhagen, I.A. (2014) Candidate neural substrates of Off-edge motion detection in Drosophila. Curr. Biol., 24, 1062–1070. Silies, M., Gohl, D.M., Fisher, Y.E., Freifeld, L., Clark, D.A. & Clandinin, T.R. (2013) Modular use of peripheral input channels tunes motion-detecting circuitry. Neuron, 79, 111–127. Strausfeld, N.J. (1976) Atlas of an Insect Brain. Springer, Berlin, Heidelberg. Strother, J.A., Nern, A. & Reiser, M.B. (2014) Direct observation of ON and OFF pathways in the Drosophila visual system. Curr. Biol., 24, 976–983. Takemura, S.Y., Lu, Z. & Meinertzhagen, I.A. (2008) Synaptic circuits of the Drosophila optic lobe: the input terminals to the medulla. J. Comp. Neurol., 509, 493–513. Takemura, S.Y., Karuppudurai, T., Ting, C.-Y., Lu, Z., Lee, C.-H. & Meinertzhagen, I.A. (2011) Cholinergic circuits integrate neighboring visual signals in a Drosophila motion detection pathway. Curr. Biol., 21, 2077– 2084. Takemura, S.Y., Bharioke, A., Lu, Z., Nern, A., Vitaladevuni, S., Rivlin, P.K., Katz, W.T., Olbris, D.J., Plaza, S.M., Winston, P., Zhao, T., Horne, J.A., Fetter, R.D., Takemura, S., Blazek, K., Chang, L.-A., Ogundeyi, O., Saunders, M.A., Shapiro, V., Sigmund, C., Rubin, G.M., Scheffer, L.K., Meinertzhagen, I.A. & Chklovskii, D.B. (2013) A visual motion detection circuit suggested by Drosophila connectomics. Nature, 500, 175–181. Tuthill, J.C., Nern, A., Holtz, S.L., Rubin, G.M. & Reiser, M.B. (2013) Contributions of the 12 neuron classes in the fly lamina to motion vision. Neuron, 79, 128–140. Venken, K.J.T., Simpson, J.H. & Bellen, H.J. (2011) Genetic manipulation of genes and cells in the nervous system of the fruit fly. Neuron, 72, 202–230. Yonehara, K., Farrow, K., Ghanem, A., Hillier, D., Balint, K., Teixeira, M., Juettner, J., Noda, M., Neve, R.L., Conzelmann, K.-K. & Roska, B. (2013) The first stage of cardinal direction selectivity is localized to the dendrites of retinal ganglion cells. Neuron, 79, 1078–1085.
© 2014 Federation of European Neuroscience Societies and John Wiley & Sons Ltd European Journal of Neuroscience, 1–9
