- Record: found
- Abstract: found
- Article: found
cis-regulatory analysis of the Drosophila pdm locus reveals a diversity of neural enhancers
research-article
Read this article at
There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.
Abstract
Background
One of the major challenges in developmental biology is to understand the regulatory events that generate neuronal diversity. During Drosophila embryonic neural lineage development, cellular temporal identity is established in part by a transcription factor (TF) regulatory network that mediates a cascade of cellular identity decisions. Two of the regulators essential to this network are the POU-domain TFs Nubbin and Pdm-2, encoded by adjacent genes collectively known as pdm. The focus of this study is the discovery and characterization of cis-regulatory DNA that governs their expression.
Results
Phylogenetic footprinting analysis of a 125 kb genomic region that spans the pdm locus identified 116 conserved sequence clusters. To determine which of these regions function as cis-regulatory enhancers that regulate the dynamics of pdm gene expression, we tested each for in vivo enhancer activity during embryonic development and postembryonic neurogenesis. Our screen revealed 77 unique enhancers positioned throughout the noncoding region of the pdm locus. Many of these activated neural-specific gene expression during different developmental stages and many drove expression in overlapping patterns. Sequence comparisons of functionally related enhancers that activate overlapping expression patterns revealed that they share conserved elements that can be predictive of enhancer behavior. To facilitate data accessibility, the results of our analysis are catalogued in cisPatterns, an online database of the structure and function of these and other Drosophila enhancers.
Conclusions
These studies reveal a diversity of modular enhancers that most likely regulate pdm gene expression during embryonic and adult development, highlighting a high level of temporal and spatial expression specificity. In addition, we discovered clusters of functionally related enhancers throughout the pdm locus. A subset of these enhancers share conserved elements including sequences that correspond to known TF DNA binding sites. Although comparative analysis of the nubbin and pdm-2 encoding sequences indicate that these two genes most likely arose from a duplication event, we found only partial evidence of sequence duplication between their enhancers, suggesting that after the putative duplication their cis-regulatory DNA diverged at a higher rate than their coding sequences.
Related collections
Most cited references46
- Record: found
- Abstract: found
- Article: not found
Exploiting position effects and the gypsy retrovirus insulator to engineer precisely expressed transgenes.
A major obstacle to creating precisely expressed transgenes lies in the epigenetic effects of the host chromatin that surrounds them. Here we present a strategy to overcome this problem, employing a Gal4-inducible luciferase assay to systematically quantify position effects of host chromatin and the ability of insulators to counteract these effects at phiC31 integration loci randomly distributed throughout the Drosophila genome. We identify loci that can be exploited to deliver precise doses of transgene expression to specific tissues. Moreover, we uncover a previously unrecognized property of the gypsy retrovirus insulator to boost gene expression to levels severalfold greater than at most or possibly all un-insulated loci, in every tissue tested. These findings provide the first opportunity to create a battery of transgenes that can be reliably expressed at high levels in virtually any tissue by integration at a single locus, and conversely, to engineer a controlled phenotypic allelic series by exploiting several loci. The generality of our approach makes it adaptable to other model systems to identify and modify loci for optimal transgene expression.
- Record: found
- Abstract: found
- Article: not found
Phage integrases: biology and applications.
C Groth, Michele Calos (2004)
Phage integrases are enzymes that mediate unidirectional site-specific recombination between two DNA recognition sequences, the phage attachment site, attP, and the bacterial attachment site, attB. Integrases may be grouped into two major families, the tyrosine recombinases and the serine recombinases, based on their mode of catalysis. Tyrosine family integrases, such as lambda integrase, utilize a catalytic tyrosine to mediate strand cleavage, tend to recognize longer attP sequences, and require other proteins encoded by the phage or the host bacteria. Phage integrases from the serine family are larger, use a catalytic serine for strand cleavage, recognize shorter attP sequences, and do not require host cofactors. Phage integrases mediate efficient site-specific recombination between two different sequences that are relatively short, yet long enough to be specific on a genomic scale. These properties give phage integrases growing importance for the genetic manipulation of living eukaryotic cells, especially those with large genomes such as mammals and most plants, for which there are few tools for precise manipulation of the genome. Integrases of the serine family have been shown to work efficiently in mammalian cells, mediating efficient integration at introduced att sites or native sequences that have partial identity to att sites. This reaction has applications in areas such as gene therapy, construction of transgenic organisms, and manipulation of cell lines. Directed evolution can be used to increase further the affinity of an integrase for a particular native sequence, opening up additional applications for genomic modification.
- Record: found
- Abstract: found
- Article: not found
Feature detection and orientation tuning in the Drosophila central complex
Johannes Seelig, Vivek Jayaraman (2013)
Many animals, including insects, are known to use visual landmarks to orient in their environment. In Drosophila melanogaster, behavioral genetics studies have identified a higher brain structure called the central complex as being required for the fly’s innate responses to vertical visual features 1 and its short- and long-term memory for visual patterns 2–4 . But whether and how neurons of the fly central complex represent visual features is unknown. We used two-photon calcium imaging in head-fixed walking and flying flies to probe visuomotor responses of ring neurons—a class of central complex neurons that have been implicated in landmark-driven spatial memory in walking flies 2,3 and memory for visual patterns in tethered flying flies 5 . We found that dendrites of ring neurons are visually responsive and arranged retinotopically. Ring neuron receptive fields comprise both excitatory and inhibitory subfields, resembling those of simple cells in mammalian primary visual cortex. Ring neurons show strong and, in some cases, direction-selective orientation tuning, with a notable preference for vertically oriented features similar to those that evoke innate responses in flies 1,2 . Visual responses were diminished during flight, but, in contrast with the hypothesized role of the central complex in the control of locomotion 6 , not modulated during walking. Taken together, these results suggest that ring neurons represent behaviorally relevant visual features in the fly’s environment, enabling downstream central complex circuits to produce appropriate motor commands 6 . More broadly, this study opens the door to mechanistic investigations of circuit computations underlying visually guided action selection in the Drosophila central complex. Flies display a variety of visual pattern- and position-dependent behaviors, including stripe fixation 2 , short-term orientation memory 2 , pattern learning 4 , and place learning 3 . Common to these behaviors is a need to detect and respond to specific features in the insect’s visual surroundings. In addition, all these behaviors require the central complex 1–5 , a deep brain region that has also been implicated in motor control 6 . We used two-photon calcium imaging in genetically targeted populations of central complex input neurons in behaving flies to investigate their potential visuomotor role. We focused on the dendritic responses of ring neurons—neurons that connect the lateral triangle (LTr) to the ellipsoid body (EB) 7–9 (Fig. 1a), and that have been specifically implicated in visuomotor memory 2,3 . Electron microscopy in the locust has shown that dendrites of ring neuron analogs arborize in specialized structures in the LTr called microglomeruli, where they are contacted by axonal projections from visual areas 10 . Confocal images of the Drosophila LTr labeled with GFP under the control of a pan-neuronal driver line, R57C10 11 , revealed a similar dense microglomerular substructure in the region (Fig. 1b, Supplementary Videos 1–4). Do LTr microglomeruli respond to visual input? We used two-photon imaging with the calcium indicator GCaMP expressed pan-neuronally to record neural activity in the LTr of head-fixed Drosophila placed at the center of a visual arena (Fig. 1c,d). Flies were presented with small bright vertical bars moving horizontally at different elevations, and we recorded LTr calcium transients from multiple planes of focus on one or both sides of the brain in a single experiment (see Methods). Calcium transients showed strong temporal correlations at the spatial scale expected of LTr microglomeruli (Fig. 1e, Supplementary Video 5; Extended Data Fig. 1a). Visual stimuli evoked robust calcium transients in a subset of microglomeruli, but only when the localized stimuli were in specific spatial locations around the fly (Fig. 1f). We computed receptive fields (RFs) for responsive microglomeruli, and found that a majority of RFs are centered in the ipsilateral visual hemifield (Fig. 1g–i, Extended Data Fig. 1b and Methods). Finally, LTr microglomeruli are clustered retinotopically, and principal component analysis based on RF centers suggests that they form a spatial map with axes that are almost parallel to the fly’s visual field (Fig. 1j,k, Extended Data Fig. 1c–e). We next examined the anatomical relationship between the LTr and individual classes of ring neurons that send arbors to the region 8,12 . We studied dendritic arborization patterns of ring neurons targeted by EB1-GAL4, which labels R2 ring neurons required for pattern memory 5 , and c232-GAL4, which labels R3/R4d neurons required for spatial memory 2 (Supplementary Video 1,2 for R3/R4d; Supplementary Video 3,4 for R2). In agreement with past anatomical work 9 , different ring neuron classes arborize in specific contiguous parts of the LTr (Extended Data Fig. 2a,b). Each ring neuron in these classes extends dendrites into a single microglomerulus in the LTr, and sends axonal arbors throughout a class-specific ring of the EB (Fig. 2a, Extended Data Fig. 2c–e, see also 7,12 ). To understand whether different types of ring neurons have distinctive visual response properties, we mapped RFs for dendritic microglomeruli of R3/R4d and R2 ring neurons (Extended Data Fig. 3a–c, and Extended Data Fig. 3d–f respectively). We found visual responses in ~7/40 c232-GAL4-labeled microglomeruli—corresponding to ~7/20 R4d microglomeruli (Extended Data Fig. 3a)—and ~14/20 of R2 microglomeruli labeled by EB1-GAL4. RFs for R2 and R4d neurons cover large parts of the visual field, with highest density near the midline of the ipsilateral visual field (Extended Data Fig. 3g,h). In summary, R4d and R2 microglomeruli appear to have similar visual response properties and overlapping RFs, but with peak sensitivity in different parts of the visual field (Extended Data Fig. 3i–k). We next probed the fine structure of microglomerular RFs using sparse white noise stimuli (Fig. 2b, see Methods). Reverse correlation of microglomerular responses revealed prominent inhibitory subfields in the RFs (Fig. 2c for R2, Fig. 3a for R4d). The spatial scales of RF structure we observe is within the range for visual features that evoke strong innate responses in flies, and that are used for visual pattern learning in tethered flies 2,3 . To test the validity of these white-noise-based RFs, we used them to predict responses to novel bar stimuli (Fig. 2d, see Methods). The predicted responses captured much of the temporal and spatial variation in the data (Fig. 2e), with high correlations between estimated and actual responses (Extended Data Fig. 4a,b). Noting that the RF structure of ring neuron inputs resembles those of simple cells in mammalian primary visual cortex 13 , we next asked if these neurons share other response properties. Indeed, when we presented flies with a series of moving oriented bars, we found strong orientation tuning in microglomerular response patterns (Fig. 2f,g). As expected for RF structures with both excitatory and inhibitory lobes, microglomeruli also showed orientation tuning when presented with bars of opposite contrast, i.e., dark bars on a bright background (Extended Data Fig. 5a,b), as are often used in fly behavioral studies 1,2 . Examining orientation tuning across the population of microglomeruli, we observed a strong preference for vertically oriented bars (Fig. 2h and Extended Data Fig. 4h,i for R2, Extended Data Fig. 4c,k,l for R4d, Fig. 2k and Extended Data Fig. 4g,n,o for pan-neuronal line). RFs tuned to vertical orientations are distributed across the visual field (Fig. 2i,j for R2, Extended Data Fig. 4c,d for R4d, Extended Data Fig. 4f for pan-neuronal line). A significant fraction of neurons also shows direction-selectivity (Fig. 2j and Extended Data Fig. 4j for R2, Extended Data Fig. 4e,m for R4d, Fig. 2l, Extended Data Fig. 4p for pan-neuronal line). Are response properties of ring neuron dendrites stereotyped? We found strong correlations across flies in RF structure for R4d (Fig. 3a–b, Extended Data Fig. 6,8), and R2 (Fig. 3c, Extended Data Fig. 7,8). Ring neurons and the EB have often been ascribed a role in complex visuomotor tasks 2,3 . We examined the possible motor function of ring neurons by assessing potential correlations between neural activity and locomotion in tethered flies walking on an air-supported ball 14 or flying 15 , in darkness or in the presence of a bright stripe moving left or right in front of the fly (Fig. 4a, see also Methods and Supplementary Video 6,7). Although some R3/R4d neurons did show occasional correlations with locomotion when flies walked in the dark, and responses from visually stimulated animals showed occasional modulation during walking, the changes were within the expected variability of visual responses in stationary flies (Fig. 4b–d). Responses were also insensitive to walking direction. Overall, LTr visual responses could be modeled accurately without taking walking state into account (Extended Data Fig. 9a–e and 10). By contrast, responses to visual stimuli were consistently diminished during tethered flight (Fig. 4e,f, Extended Data Fig. 9f–h, Supplementary Video 7), but showed no obvious correlations with flight direction as assessed by differences in wingbeat amplitude envelope. Thus, ring neuron LTr responses were modulated by motor state, but not in a manner consistent with a direct role in motor coordination, and in a markedly different manner than in the optic lobe 15,16 . Behavioral genetics studies in Drosophila have suggested that the central complex is required for a wide range of important sensorimotor functions. However, in the absence of physiological recordings from the region in flies, it has been challenging to constrain its role in the diversity of behaviors that it has been implicated in. We studied the visuomotor responses of ring neurons, which provide input from the LTr to the ellipsoid body of the central complex. Analogous neurons in other insects respond to polarized and unpolarized light 17–19 , and to mechanosensory stimulation 20 , one of the sensory modalities that we did not explore but that may partly account for unresponsive neurons in our study. We found that R2 and R4d ring neurons are visually responsive, and these responses are not significantly modulated by walking state. Although visual responses are diminished during flight, they do not vary systematically with wingbeat patterns associated with turns. It is possible that outputs of these ring neurons within the EB rings could be more sensitive to motor actions, but our physiological results in their dendrites during behavior are inconsistent with a major role for these ring neurons in motor coordination in the fly. Further, the high degree of stereotypy that we observe in ring neuron receptive field characteristics across flies suggests that, rather than directly performing motor coordination, these neurons likely provide downstream central complex circuits with similar behaviorally relevant visual feature sets on which to base motor decisions. As a striking example, the strong vertical tuning preference we observe in the LTr may partly underlie the tendency of flies to fixate on vertical edges during both flight and walking 1,2,21 . Recent work has suggested that vertical stripe fixation during walking largely relies on a hypothesized position system sensitive to local luminance changes that can operate independently of neural circuits involved in optomotor responses to widefield motion 21,22 . The response properties of R2 and R4D neurons is consistent with a role in such a position system 22 . The retinotopic bias, structure of excitatory and inhibitory subfields, orientation tuning and direction selectivity we see are reminiscent of those seen in calcium imaging studies in simple cells in mammalian visual cortex 13 , providing an interesting example of how evolutionarily distant visual systems with different types of eyes nonetheless use similar feature sets to process visual scenes. From Hubel and Wiesel’s findings several decades ago 23 to the present, significant progress has been made on identifying neural representations used at different stages in the mammalian visual system. However, mechanisms underlying simple cell responses are not yet fully understood 24 . With its array of genetic tools, Drosophila melanogaster may allow us to uncover how spatiotemporal interactions of excitatory and inhibitory inputs might produce similar orientation tuning and direction selectivity in the LTr 25,26 . There is considerable evidence for spatially tuned visual responses in the lobula complex and anterior optic tubercle in other insects 27–30 , suggesting that components of ring neuron input response properties may also arise from selective averaging of weaker and more broadly distributed tuning preferences in such areas. Overall, our findings lay the groundwork for future research into how this genetic model organism’s small brain uses feature and pattern information for visual orientation and navigation. METHODS Fly stocks All experiments were performed with female flies, with ages chosen based on expression levels of relevant fluorescent proteins. At least six animals were used for any single condition tested – specific sample sizes are noted for each set of experiments and were chosen based on the level of variability observed in initial experiments. Flies were randomly picked from their housing vials for all experiments. We used all data that passed a quality threshold based on the observed health of the fly during an experiment and the signal-to-noise ratio of the imaging signal. Calcium imaging experiments to measure RFs were performed with UAS-GCaMP3; c232-GAL4, UAS-GCaMP3;EB1-GAL4, and pJFRC7-20XUAS-GCaMP5.003 (VK00005)/R57C10-GAL4 flies. We used pGP-JFRC7-20XUAS-IVS-GCaMP6f 15.693(attP40)/R57C10-GAL4 flies for pan-neuronal orientation tuning experiments (2–5 days old) and pGP-JFRC7-20XUAS-IVS-GCaMP6s 15.641 (attP40)/c232-GAL4 for orientation tuning experiments (1–2 weeks old) with dark bars. c232-GAL4 and EB1-GAL4 were gifts from M. Heisenberg and T. Lee, respectively. To label dendrites and axons in different ring neurons, EB1-GAL4 and c232-GAL4 were each crossed to pJFRC67-3XUAS-IVS-Syt-GFP(attP18) 31,32 , pJFRC118-10XUAS-DenMark(attP40) 32,33 , and pJFRC119-10XUAS-IVS-myr::TopHat2(VK00005) (gift of B.D. Pfeiffer, unpublished). To compare the expression patterns of EB1-GAL4 and c232-GAL4 to the pan-neuronal line R57C10 in the same fly, EB1-GAL4 and c232-GAL4 were each crossed to pJFRC19-13XLexAop2-IVS-myr::GFP (su(Hw)attP8) 32 , pJFRC21-10XUAS-IVS-mCD8::RFP (attP18) 32 ; Sco/Cyo; R57C10-LexA::p65 (attP2) 34 . For stochastic single cell labeling of EB1-GAL4 and c232-GAL4 with three colors a “flip-out”-based approach 35 (Nern et al., in preparation) was used. In brief, heat-shock induced expression of FLP recombinase was used to excise FRT-flanked interruption cassettes from UAS reporter constructs carrying HA, V5 and FLAG epitope tags, and stained with epitope-tag specific antibodies. This results in labeling of a subset of the cells in the expression pattern with a stochastic combination of the three labels. Anatomy: fly dissections, immunohistochemistry and confocal imaging Confocal stacks were recorded with a 40x or 63x objective on a Zeiss confocal microscope. Dissections and staining were performed as previously described 36,37 . The primary antibody mixture consisted of 1:1000 sheep anti-GFP (AbD Serotec), 1:1000 rabbit anti-DsRed (Clontech), 1:100 rat anti-HA (for Syt/DenMark/HA staining; Roche), and 10% normal donkey serum (Jackson ImmunoResearch) in PBS-TX. Fly preparation for RF mapping RF measurements were performed using a preparation described previously 14 , but with the behavioral apparatus removed to maximize the fly’s visual field. The fly was briefly anesthetized on ice and transferred to a cold plate held at 4°C. The proboscis of the fly was fixed by either pressing it into the head and fixing it with wax, or by stretching it with a pair of tweezers mounted on a micromanipulator and then fixing it with a mixture of wax and colophony. We additionally removed either the front legs or all legs for RF measurements. The fly was glued to a pin and positioned in the holder using a micromanipulator and fixed in the holder with UV gel 14 . An opening was cut into the head to obtain optical access to the brain. To stop brain movement due to pulsation of muscle M16, we cut the muscle or the nerves innervating the muscles with dissection needles. The fly holder (including the micromanipulator) was then transferred to the microscope and mounted using magnetic mounts. Flies were dark adapted for 5 to 10 minutes before recordings started. Fly preparation for imaging during walking For walking experiments, an air-supported ball was positioned under the fly with a three-axis micromanipulator as described previously 14 and the walking velocity of the fly was monitored using a camera system. Fly preparation for imaging during tethered flight For flying fly experiments a holder similar to those described previously 14,15,38 was used. The holder was made out of two pieces of stainless steel shim (thickness 12.7 µm). The shim is cut using a laser mill and then folded into its pyramidal shape. One piece of the holder is glued onto a liquid chamber similar to the one used for walking behavior using epoxy 14 . After removing the front legs of the fly to prevent it from grabbing the holder, the fly is glued to a pin and inserted into the holder with a micromanipulator. The second piece of the holder is then inserted to close the pyramidal shape around the fly’s head. The setup for combined imaging and flight behavior was similar to the one described in 15 . The fly was illuminated with IR light from below. The wing beat of the fly was recorded using a mirror placed beneath the fly and a camera (Basler 602f, operating at 100 or 150 Hz frame rate) 15 . Two-photon imaging Calcium imaging was performed using a custom-built two photon microscope controlled with ScanImage 39 . Fluorescence was detected using a photomultiplier tube (H7422PA-40, Hamamatsu). We used an Olympus 40x objective (LUMPlanFl/IR, NA 0.8) and typically adjusted the power to below 20 mW at the back aperture of the objective. We imaged at a frame rate of 6.7 Hz. Focal planes were selected based on the anatomy and visual responses of microglomeruli. We focused on microglomeruli because this approach allowed us to distinguish different neurons in a labeled population (such distinctions are difficult to make in the EB, where axons of different neurons arborize in close proximity). Visual stimulation LED arena Visual stimuli were presented using a curved visual display 40 that was covered with a color filter to prevent cross-talk between fluorescence detection and visual stimulation as described 14 . Additionally, to avoid reflections of stimuli from the curved surface of the display, we covered the display with a diffuser (tracing paper). Under such low-contrast conditions, we found that the signal-to-noise ratio (SNR) of calcium responses when stimulated by dark-on-bright-background stimuli (e.g., dark bar on bright surround) was low. This motivated our preference for bright-on-dark-background stimuli, which produced higher SNR calcium responses (comparison in Extended Data Fig. 5). For RF measurements the display spanned 270° in azimuth and 120° in elevation. The top left and top right corners of the display (three square panels in each corner, each with a size of 30° by 30°) could not be seen by the fly (since they were occluded by the fly holder) and were excluded from the display. For behavioral experiments, a similar but smaller display was used, spanning 210° in azimuth and 90° in elevation. Flashing dots for RF mapping Excitatory RFs for c232-GAL4 and EB1-GAL4 flies were measured with stationary bright square dots (7.5° by 7.5°) appearing randomly in the visual field of the fly for 1s followed by a dark period of 1s. The measurements were repeated until the entire visual field covered by the display was stimulated once. The display was sampled with a spatial resolution of the stimulus (7.5°). Measurements were performed in 8 blocks of 140s each, presenting a total of 468 stimuli covering the entire display. Horizontally moving bars for RF mapping We used bright bars (15° in elevation, 7.5° in azimuth) that moved at a speed of 30 °/s left and right (in azimuthal direction) in the fly’s field of view. Two repetitions with stimulation in one direction were followed by two repetitions of movement in the opposite direction. To map responses over the entire visual field spanned by the display the stimulus was shifted in steps of 15° in elevation from the top most part of the visual display towards the bottom part and the left and right moving stimulation was repeated in each row. White noise For white-noise stimulation 37,41 we subdivided the display into squares of 11.25° by 11.25°. Each stimulus frame consisted of 20 randomly selected bright squares while the remaining squares remained dark (in a few trials for one fly we used 30 squares). The stimulus appeared for one second followed by a dark period of one second. We presented 60 frames of random stimuli in a trial of 140 s. For c232-GAL4 flies we used an average of 33 +/− 20 trials in 7 flies (59, 61, 29, 21, 21, 14, and 26 trials, respectively). For EB1-GAL4 flies we used an average of 41 +/− 13 trials in 6 flies, (38, 25, 39, 63, 35, and 47 trials, respectively). The color scale (Fig. 2, Extended Data Fig. 8) was adjusted to give equal weight to excitatory and inhibitory subfields (mean set to 0). To validate white noise based responses we stimulated the fly with bright bars of 56.25 degrees by 18.75 degrees oriented either vertically, horizontally, or rotated by 45 or −45 degrees with respect to the vertical direction. To stimulate the receptive fields of all microglomeruli in an unbiased way, the bar stimuli were presented at a random position on the display and the display was sampled with a resolution of 11.25° in both elevation and azimuth. Visual stimulation for measuring orientation tuning curves For orientation tuning curve measurement we used a stimulus that extended across the entire display, from 120° in azimuth for vertical orientation (90°) to 270° azimuth for horizontal orientation. The width of the bar was 15° and it was moved at a speed of 75°/s. The angle was changed incrementally in steps of 11.25° starting with either horizontally or vertically oriented bars for one direction of movement, and then repeated in the opposite direction of movement. To obtain tuning curves with dark stimuli we inverted the contrast of bright and dark, and removed the diffuser. Visual stimulation during walking and flying experiments For walking and flying experiments we used a vertical bright bar spanning 90° in elevation and 15° in azimuth. The bar moved horizontally at a velocity of 15 °/s. The bar stayed stationary for 10 s after moving in one direction for 17 s and then resumed moving in the opposite direction for 17 s. Data analysis All data analysis was performed in MATLAB (MathWorks, Inc., Natick, MA). All errors and error bars shown are standard deviation (s.d.). All p-values shown are based on t-tests, unless otherwise noted. Frame alignment and movement correction Data recorded using two-photon imaging were aligned in the XY plane on a frame-by-frame basis. Data were thresholded to distinguish arborizations in the lateral triangle from background. All above-threshold pixels were set to the same value. Frames were aligned by cross-correlating each thresholded frame to a single frame at the beginning of the measurement. Multiple trials were aligned by cross-correlation of trial-averaged frames thresholded as above. Calculation of fluorescence changes The baseline for calculating ΔF/F was selected by averaging over the 10% of frames with lowest intensity in each trial or by using the baseline at the beginning of the experiment. Due to the low baseline intensity of GCaMP6, background fluorescence was not subtracted in measurements with GCaMP6. Calcium traces recorded from behaving flies were smoothed with a third order Savitzky-Golay filter over 7 frames for comparisons with behavioral data. ROI selection For RF measurements in stationary flies, ROIs corresponding to microglomeruli were selected manually in videos of ΔF/F. Overlapping parts of ROIs were excluded from further analysis. For experiments in behaving flies, ROIs were selected using visually supervised k-means clustering (Extended Data Fig. 10a–d). We used a subset of three trials (2820 frames) for clustering-based selection of ROIs. We then used correlation-based k-means clustering between the calcium traces in all thresholded pixels. The number of clusters was selected based on an estimate of the number of microglomeruli. If not all microglomeruli could be separated into different clusters, the number of clusters was increased in a second run. We then set a threshold to remove clusters that were smaller than a certain number of pixels (60). We additionally removed clusters that had an average cross-correlation value lower than a threshold (0.2). We further split anatomically disconnected regions of the same cluster and again removed those parts that were smaller than a size threshold (30 pixels). In a final check, the remaining ROIs were overlaid with the frames of the calcium video that showed the largest response in this ROI, and ROIs that did not correspond to microglomeruli were removed manually. RF mapping RFs were smoothed using a Gaussian filter (4 pixels with a standard deviation of 1 pixel by default; 5 pixels with a standard deviation of 4 pixels for white noise measurements). For RF measurements (e.g., in Fig. 1), we combined responses to left- and right-moving stimuli (each averaged over two trials). The onset of the calcium response was correlated with the bar’s movement into the RF. Due to the size of the RF, a bar moving from the right side of the fly towards the left side entered the RF from the right side and induced a calcium onset starting at the right side of the (excitatory part of the) RF. Similarly, a bar entering the receptive field from the left side induced a calcium onset starting at the left side of the RF. Due to the faster on-response than off-response of GCaMP the onset of the RF was better defined than the offset. To find the center of the excitatory RF measured with moving bars, we defined the RF center as the weighted centroid of the average of responses to thresholded left- and right-moving stimuli (Extended Data Fig. 1 b). This was equivalent to delimiting the RF by its calcium onset response. This procedure made the location of computed RF centers invariant to the kinetics of the calcium indicator. We used the following parameters to characterize the excitatory parts of RFs measured with single stationary dots (measured at 50% of the maximum ΔF/F response of each RF) (see Extended Data Figure 7c): area, major axis (of an ellipse that has the same normalized second central moments as the region as determined with the MATLAB ‘regionprops’ function), minor axis, eccentricity, orientation (the angle between the horizontal x axis and the major axis of the ellipse, 0 corresponding to horizontal orientation), retinotopic correlation (the correlation coefficient between the center of the RF, determined as the weighted centroid of the RF area, and the center of the corresponding ROI). RF display To display microglomerular RFs, we colored LED arena positions in proportion to the ΔF/F response elicited by the stimulus presented in that position. For moving stimuli, the calcium response in each pixel was determined as the average over all response frames that were recorded while the stimulus was at that position. Calcium responses were interpolated to account for mismatches between frame rates and movement of stimuli. For stimulation with stationary stimuli, the ΔF/F values shown in RF plots correspond to the maximum ΔF/F during the stimulation period. To prevent cross talk between sequentially presented stimulus frames due to the slow decay of the calcium response (which extended beyond stimulus presentation) we only considered the calcium increase and not the calcium decay in assigning ΔF/F values to stimulus frames. Retinotopy To assess retinotopy, we calculated the correlation coefficient between microglomerulus centers and centers of their RFs. This was compared to the correlation coefficient obtained after randomly shuffling correspondences. Additionally, we performed principal component analysis (PCA) on the (x,y) values of RF centers. The first PC gave us the direction of maximum variation of RF centers, which we consider to be the primary retinotopic axis (in the fly’s visual field) for the RF population. Orientation tuning To measure orientation tuning (Fig. 2), single 15°-wide bars spanning the entire visual display were presented, and their orientation changed in steps of 11.25°. Responses from multiple trials were then averaged, and tuning curves were fit using the sum of two circular Gaussians 42,43 . a 1 · e − k 1 · ( cos ( θ − θ 1 ) − 1 ) + a 2 · e − k 2 · ( cos ( θ − θ 2 ) − 1 ) , where, a1 and a2 are amplitudes, θ 1 and θ 2 are the maximum angles and k1 and k2 are width parameters. The preferred orientation for each microglomerulus was the maximum of its fitted tuning curve. Orientation selectivity The orientation selectivity index 44 , OSI, was computed as the difference between the response in the preferred direction, ΔF/Fmax, and the direction orthogonal to it (preferred direction +/− 90°), ΔF/Fortho, normalized by the sum of the responses in the two directions: O S I = ( Δ F / F max − Δ F / F ortho ) / ( Δ F / F max + Δ F / F ortho ) . Direction selectivity The direction selectivity index 42 , DI, was calculated as the difference between responses in the preferred direction, ΔF/Fmax, and anti-preferred direction (preferred direction+180°), ΔF/Fopposite, normalized by their sum: D I = ( Δ F / F max − Δ F / F opposite ) / ( Δ F / F max + Δ F / F opposite ) . The sign of the DI was defined as positive for front-to-back movement and negative for back-to-front movement. White-noise-based RFs and response predictions RFs were reconstructed by thresholding calcium traces at 30% ΔF/F and averaging over all frames that induced a response larger than this threshold, weighted by peak calcium response. Only the rising slope of the calcium response was considered. The displayed RFs are the weighted averages of stimulus frames and the mean value (background) is subtracted. Max is the maximum of the weighted average after subtracting the mean value, and min is the minimum of this average. The mean value is set to 0. To predict responses to oriented bars (Fig. 2), we multiplied the mean-subtracted white-noise-based RF with stimulus values in each pixel and summed over all pixels. We convolved the result with a calcium response function. The amplitude and time constants of the calcium response function were fitted using responses to bars of one orientation and then used in the prediction of responses to the remaining three orientations. The calcium response function, crf, used in predicting responses to oriented stimuli and in the analysis of calcium responses recorded during behavior (see below) is given by c r f ~ ( 1 − ( e − ( t − t 0 ) t on ) ) · ( e − ( t − t 0 ) t off ) , where t0 is the onset time, and ton and toff , are the rise and decay times, of the calcium indicator. Walking behavior analysis Ball movement was recorded with a sampling rate of 4 kHz and velocities were calculated with a time base of 250 ms. Velocities were then averaged over all velocity values in each two-photon imaging frame. Average velocities were calculated by detecting epochs in which the fly was moving and then averaged over this period. Due to strong walking activity in behaving animals, there were few recordings that allowed us to compare microglomerular responses in walking and stationary conditions (Extended Data Fig. 10e–n). Flight behavior analysis The wing angle was detected by first manually identifying the wing hinge in both wings. The wing was then detected by first subtracting the background recorded while the fly was not flying, smoothing with a mean filter, and setting a threshold for detecting the wings in two ROIs surrounding the wings. The wing angle was then defined as the angle between the wing hinge and the tip of the wing. Flight was typically intermittent (Extended Data Fig. 10o,p). Model for fitting responses of visual microglomeruli during behavior For describing the calcium responses of visual microglomeruli during walking behavior we fitted the responses with a model consisting of a single excitatory and inhibitory Gaussian function. Model fitting was only performed for microglomeruli that responded with a single peak during one passage of the stimulus on the display, not for bilateral receptive fields. The initial position of each Gaussian, the (common) standard deviation and the amplitude were used as fit parameters. Since the inhibitory responses could not be directly observed in our calcium signals, we set all negative values to zero. r f ~ ( e − ( x − x 0 ) 2 / ( 2 σ x 2 ) − e − ( x − x 1 ) 2 / ( 2 σ x 2 ) ) , r f = 0 if r f < 0 , where, x0 and x1 are the location of the excitatory and inhibitory receptive fields, respectively, and σx is the width of the receptive field. The calcium signal s (ΔF/F) was then modeled as the convolution of the receptive field with the calcium response function, crf, and a constant offset, bg, s ~ c r f ⊗ r f + b g . Since the response in flying flies depended on the state of the fly we used a model for the receptive field that additionally included the left and right wing angles as parameters r f flying ~ behavior · ( e − ( x − x 0 ) 2 / ( 2 σ x 2 ) − e − ( x − x 1 ) 2 / ( 2 σ x 2 ) ) , r f = 0 if r f < 0 where, behavior = a · lw + b · rw + c depends linearly, with the parameters a, b and c, on the left wing angle, lw, and the right wing angle, rw. Supplementary Material 1 8 Extended Data Figure 7. Stereotypy of excitatory RFs across flies in R2 neurons; Summary of receptive field properties a, Same as Extended Data Fig. 6a, measured in R2 neurons in 8 flies. b, Cross-correlation coefficient between fly 1 and all other RFs in each column for R2 neurons shown in a (mean = 0.75 ± 0.12). c, Table showing characteristics of excitatory receptive fields for R4d and R2 ring neurons measured with single dot stimulation calculated at half maximum ΔF/F. The retinotopic correlation coefficients are significantly different from the control distributions, with p = 0.0283 for R4d (n = 8 flies, 159 RFs), and p = 0.0069 for R2 (n = 12 flies, 219 RFs). 9 Extended Data Figure 8. Stereotypy of excitatory/inhibitory RFs across flies in R2 neurons a, RFs measured using white-noise stimulation in 6 flies aligned according to similarity. We used an average of 41 ± 13 trials (6 flies, number of trials: 38, 25, 39, 63, 35, and 47). Each trial consisted of presentation of 60 stimulus frames (see Methods). b, RFs measured using white-noise stimulation in 7 flies aligned by similarity (subset of this data is shown in Fig. 3). We used an average of 33 ± 20 trials (7 flies, number of trials: 59, 61, 29, 21, 21, 14, and 26). 10 Extended Data Figure 9. Ring neuron visual responses in the LTr are significantly diminished during flight, but not walking a, Microglomeruli selected by visually assisted k-means clustering in R2 neurons, a subset of which are shown in Fig. 4b. b, Average over n = 24 trials of responses of R2 microglomeruli to a bright bar (15°×90°) moving at 30 °/s shown during a period when the fly is walking more than 50 % of the time (blue) and less than 50% of the time (red) during rightward moving stimulation (envelopes show s.d.). c, Correlation coefficient (mean = 0.72 ± 0.25) between model (see Methods) and data for R2 microglomeruli recorded in n = 8 flies, and, d, for R4d microglomeruli (n = 14 flies, mean 0.7 ± 0.25) that have unilateral RFs. e, Concatenation of seven epochs during walking behavior with EB1-GAL4, UAS-GCaMP3 flies during stimulation with a left and right moving bar: data (blue) and model (red) fit (see Methods). f, Microglomeruli in R4d selected using visually assisted k-means clustering. g, Visual responses during flying and non-flying behavior during stimulation with a bar moving from left to right (averaged over n = 14 trials for flying, n = 14 trials for non-flying), and during stimulation with a bar moving from right to left (n = 17 for flying, and n = 11 trials for non-flying) for the fly shown in Fig. 4e (envelopes show s.d.). h, Correlation coefficient between model (see Methods) and data for R4d neurons with a linear dependence of the visual response on both wing angles (mean = 0.65 ± −0.27) recorded in n = 14 flies. 11 Extended Data Figure 10. Selection of microglomeruli using visually guided k-means clustering (a–d); Walking and flight activity during behaving fly experiments (e–p) a, Cross-correlation matrix between all pixels above threshold (see Methods for details). b, Cross-correlation matrix sorted according to clusters found by k-means clustering. c, Resulting regions of interest. d, Overlay of resulting regions of interest and selected frames of calcium video. Regions of interest that do not correspond to microglomeruli (based on comparison with the calcium video) are removed. e, Forward, f, side, and, g, rotational velocities for all c232-GAL4, UAS-GCaMP3 flies recorded during walking behavior. h, Percentage of walking for all EB1-GAL4, UAS-GCaMP3 flies recorded. i, Forward, j, sidewise, k, and rotational velocities, and l, percentage of walking for all EB1-GAL4, UAS-GCaMP3 flies recorded during walking behavior. m, Number of 130s epochs recorded for each c232-GAL4, UAS-GCaMP3 walking fly. n, Number of 130s epochs recorded for each EB1-GAL4, UAS-GCaMP3 walking fly. o, Percentage of time flying during recording for all c232-GAL4, UAS-GCaMP3 flies recorded during flight behavior. p, Number of 130s epochs recorded for each c232-GAL4, UAS-GCaMP3 flying fly. 2 Extended Data Figure 1. Pan-neuronal RFs for LTr microglomeruli a, Selected two-photon imaging frames showing calcium response in the LTr with pan-neuronal expression of GCaMP5. Highlighted in red are the microglomeruli that were selected for RF reconstruction. Scale bar = 5 µm. b, Schematic of visual stimulation used, RFs (two-trial averages) measured with left-and right-moving visual stimuli and their intersection labeled at FWHM intensity (white ellipse) with the weighted centroid (white star) of the distribution. c, Center of each RF plotted against the center of its corresponding microglomerulus. d, Correlation coefficient of data in d (r = 0.39, p = 0) and correlation coefficient of same data after randomly permuting the microglomeruli (n = 3000 permutations, mean = (4.6 ± 0.076) · 10−4). e, Principal component axes of RF centers (see Methods). 3 Extended Data Figure 2. Ring neurons make dendritic projections to LTr microglomeruli, and send axonal arbors to the rings of the ellipsoid body a, A frame of a confocal stack of the LTr labeled with antibody staining against green fluorescent protein (GFP) using a pan-neuronal driver line, R57C10-LexA and antibody staining against red fluorescent protein (RFP) of EB1-GAL4. b, A section of a confocal stack of the LTr labeled with GFP using a pan-neuronal driver line, R57C10-LexA and RFP labeling of c232-GAL4. c, Multicolor FLP-out labeling of c232-GAL4 (see Methods). d, Antibody staining of c232-GAL4 expressing the dendritic marker, DenMark (red) and an axonal label, synaptotagmin-tagged GFP (green). Entire cell is outlined with membrane-tagged HA (blue) (see Methods). Top panel is merge of all colors. e, Same as d for EB1-GAL4. 4 Extended Data Figure 3. Ring neurons respond with strong calcium transients to single stationary bright dots; RFs of ring neuron populations are spread non-uniformly across the largely ipsilateral visual field a, Overlay of all regions of interest of R4d microglomeruli recorded in a fly during visual stimulation over an average of all frames in a calcium video (940 frames), and selected frames of the calcium video showing responses of individual microglomeruli. Scale bar = 5 µm. b, Top: An example of the single dot stimulus appearing at the specified azimuth and elevation for 1s followed by a dark period of 1s (colored dot indicates both the stimulus and the subsequent off period). Bottom: ΔF/F for microglomerulus 3 in a responding to the stimulus. c, Excitatory RFs for all microglomeruli shown in a. d, Overlay of all regions of interest selected among R2 microglomeruli recorded during visual stimulation over an average of all frames (n = 940) in a calcium video, and selected frames of the calcium video showing responses of individual microglomeruli. Scale bar = 5 µm. e, Top: Azimuth and elevation of the single bright dot presented for 1s followed by a 1s interval without stimulus. Bottom: ΔF/F for microglomerulus 1 in d responding to the visual stimulus. f, Excitatory RFs for all microglomeruli shown in d. g, Overlapping RFs of R4d neurons. Ellipses have the same normalized second central moments as the RFs shown in c thresholded at 30% ΔF/F. Intensity scale indicates how many RFs overlap in a given region. h, Same as g for R2 neurons shown in f. i, Average RFs measured across all flies with single dot stimulation for R4d neurons (green, n = 8 flies, 159 RFs), and j, for R2 neurons (purple, n = 12 flies, 219 RFs), and k, for both (overlay). Recordings from the right lateral triangle were flipped to the contralateral side for averaging. Note: Although a large fraction of R2 and R4d ring neurons we imaged were visually responsive, there were ring neurons that did not respond in our experiments. Such neurons may have RFs in areas of the fly’s visual field that we did not sample (for example, very lateral areas), be tuned to visual dimensions that we did not explore (such as color or polarization or complex shapes), or be selectively responsive to other sensory modalities (such as mechanosensation or thermosensation). 5 Extended Data Figure 4. Ring neuron orientation preference and selectivity, direction selectivity, and RF center positions a, Correlation coefficients between predicted and measured responses for R2 neurons (histogram over all trials and RFs) (n = 6 flies). b, Same as a, for R4d neurons (n = 6 flies). c, Microglomeruli of R4d neurons colored according to orientation preference and orientation selectivity (six-trial averages). d, Azimuth and elevation of the center of the excitatory RFs of the neurons shown in c (two-trial averages, measured using moving bars as described in Extended Data Fig. 1 and Methods). e, Direction selectivity for neurons in c (six-trial averages). f, Microglomeruli shown in Fig. 2k,l labeled according to the azimuth and elevation of their excitatory RF center (measured using moving stimuli as in Extended Data Fig. 1 and, g, their orientation selectivity (four-trial averages). h–p, Distributions of preferred orientation, orientation selectivity and direction index for R2 (n = 10 flies), R4d (n = 12) and pan-neuronal (n = 11 flies) microglomeruli. 6 Extended Data Figure 5. Comparison of orientation tuning curves measured using dark-on-bright and bright-on-dark bars a, Orientation tuning curves (GCaMP6s) of three R4d microglomeruli measured using (top row) bright bars on a dark background (as in Fig. 2f) and dark bars on a bright background (exact contrast inversion of bright-bar stimulus). As suggested by excitatory-inhibitory structure of RFs (Fig. 2c, Extended Data Fig. 8), tuning curves measured with the two stimuli are qualitatively similar, but bright bars produce stronger responses (with our low-contrast, filtered LED array display, see Methods). b, Fitted tuning curves measured using dark and bright bars are highly correlated (mean = 0.648 ± 0.26, n = 7 flies). 7 Extended Data Figure 6. Stereotypy of excitatory RFs across flies in R4d neurons a, Excitatory RFs measured using single dot stimulation in one fly (fly1), and RFs recorded in n=12 other flies aligned by similarity. b, Cross-correlation coefficient between fly 1 and all other RFs in each column (mean = 0.74 ± 0.2).
Author and article information
Contributors
Jermaine Ross: rossje@ninds.nih.gov
Alexander Kuzin: kuzina@ninds.nih.gov
Thomas Brody: brodyt@ninds.nih.gov
Ward F. Odenwald:
ORCID: http://orcid.org/0000-0001-5791-9729
odenwaldw@ninds.nih.gov
Journal
Journal ID (nlm-ta): BMC Genomics
Journal ID (iso-abbrev): BMC Genomics
Title:
BMC Genomics
Publisher:
BioMed Central
(London
)
ISSN
(Electronic):
1471-2164
Publication date
(Electronic):
16
September
2015
Publication date PMC-release: 16
September
2015
Publication date Collection: 2015
Volume: 16
Issue: 1
Electronic Location Identifier: 700
Affiliations
Author information
Ward F. Odenwald
http://orcid.org/0000-0001-5791-9729
Article
Publisher ID: 1897DOI: 10.1186/s12864-015-1897-2
PMC ID: 4574355
PubMed ID: 26377945
SO-VID: 43ddd310-c5ae-4908-bef7-eff7dfc295a1
Copyright © © Ross et al. 2015
License:
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License ( http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
History
Date
received
: 22
June
2015
Date
accepted
: 2
September
2015
Categories
Subject:
Research Article
Custom metadata
issue-copyright-statement © The Author(s) 2015
ScienceOpen disciplines: Genetics