Abstract
Behavior depends on discrimination of selective sensory representations. These are modified dynamically by changes in behavioral state, facilitating context-dependent selection of behavior. These are mediated by brain signals carried by noradrenergic input in mammals, or octopamine (OA) in insects. To understand the circuit mechanisms of this signaling, we characterized the function of two OA neurons, in the input region to the memory center, the mushroom body calyx, in larval Drosophila. Here they target multiple neurons, including olfactory projection neurons (PNs), the inhibitory neuron APL, a pair of extrinsic output neurons, but relatively few mushroom body intrinsic neurons, Kenyon cells. The OA receptor Oamb, a Drosophila α1-adrenergic receptor ortholog, localized to PN terminals, and optogenetic activation of OA neurons both potentiated PN activity, and compromised odor discrimination behavior. Our results suggest that OA neurons gate odor signals for sensory processing via extrinsic inputs at the input to the olfactory learning circuit.
Introduction
Behavioral choices depend on discrimination among “sensory objects”, which are neural representations of multiple coincident sensory inputs, across a range of sensory modalities. For example, “odor objects” (Gottfried, 2009; Wilson and Sullivan, 2011; Gire et al., 2013) are represented in sparse ensembles of neurons, that are coincidence detectors of multiple parallel inputs from odor quality channels. This principle is used widely in animals, including in mushroom bodies (MBs), the insect center for associative memory (Masuda-Nakagawa et al. 2005; Honegger et al. 2011), and in the piriform cortex (PCx) of mammals (Stettler and Axel, 2009; Davison and Ehlers, 2011).
The selectivity of sensory representations can be modulated dynamically by changes in behavioral state, allowing an animal to learn and respond according to perceptual task. In mammals, the noradrenergic system originating in the locus coeruleus (LC) is implicated in signaling behavioral states such as attention, arousal and motivation (Sara, 2009; Sara and Bouret, 2012; Aston-Jones and Waterhouse, 2016). Noradrenergic modulation of the olfactory bulb enhances sensitivity of odor detection and discrimination, dependent on a-1-adrenergic receptors (Escanilla et al., 2010, 2012), and LC activation enhances the excitability of PCx pyramidal neurons (Bouret and Sara, 2002), and LC activation coincides with reward anticipation in appetitive odor discrimination (Bouret and Sara, 2004). However, the circuit and synaptic mechanisms of modulation by NA are not well understood.
In insects, octopamine (OA), chemically and functionally homologous to noradrenalin (NA) in mammals, can mediate changes in behavioral state that often promote activity, for example: sensitization of reflex actions in locust (Sombati and Hoyle, 1984); aggressive state in crickets (Stevenson et al., 2005); initiation and maintenance of flight state (Brembs et al., 2007; Suver et al. 2012); and enhanced excitability of Drosophila motion detection neurons during flight (Strother et al. 2018). Another potential role of OA is as a reward signal: although this role is at face value distinct from a role in promoting activity or preparedness, OA appears to encode unconditioned stimuli in honeybee (Hammer, 1993; Hammer and Menzel, 1998; Menzel 2012), cricket (Matsumoto et al. 2015), and Drosophila (Schwaertzel et al. 2003) appetitive learning paradigms. However, taken together, the OA system in insects is heavily involved in signaling behavioral states, comparable to NA in mammals.
Behavioral states facilitate the processing of visual sensory input associated to behavioral relevance (Gilbert and Li, 2013). At the circuit level, two models have been proposed, increase in response gain (Treue and Martinez Trujillo, 1999; Reynolds and Chelazzi, 2004), and changes in tuning properties of cortical visual neurons, reflecting sensory driven stimuli and behavioral context (Li et al. 2004). Although behavioral studies are extensive, the circuit and synaptic mechanisms of these signaling inputs are not well understood. We aimed to dissect the role of the OA circuitry in sensory processing. We used the simple sensory “cortex” of larval Drosophila, the calyx, which is the sensory input region of the mushroom bodies (MBs), the insect memory center. Here, each MB neuron (Kenyon cell, KC) typically arborizes in several glomeruli, most of which are organized around the terminus of an olfactory projection neuron (PN); KCs thus combinatorially integrate multiple sensory input channels (Masuda-Nakagawa et al., 2005; Eichler et al., 2017). Inhibitory feedback (Lin et al., 2014; Masuda-Nakagawa et al. 2014), helps to maintain KC sparse responses and odor selectivity (Honegger et al., 2011), analogous to inhibition in the mammalian PCx (Poo and Isaacson, 2009; Settler and Axel, 2009; Gire et al. 2013).
The larval MB calyx is also innervated by two OA neurons, sVUMmd1 and sVUMmx1, ventral unpaired medial neurons with dendritic fields originating in the mandibular and maxillary neuromeres respectively of the subesophageal zone (SEZ) (Selcho et al., 2014). These sVUM1 neurons also innervate the olfactory neuropiles of the antennal lobe (AL) and lateral horn. This pattern of innervation is conserved in other insects, for example by dorsal unpaired median (DUM) neurons in locust (Bräunig, 1991), the VUMmx1 neuron in honeybee (Hammer 1993; Schröter et al. 2007), and OA-VUMa2 neurons in adult Drosophila (Busch et al. 2009). It also resembles the NA innervation of mammalian forebrain by LC neurons originating in the brainstem. Therefore, the network organization and coding principles in the larval MB calyx appear widely conserved, but with few neurons and genetically tractable tools, it provides many advantages to understand how individual neurons are integrated to generate a concerted neural activity pattern.
We therefore characterized the innervation pattern and synaptic partners of OA neurons in the calyx, the localization of the OA receptor Oamb in the calyx circuit, and the impact of OA calyx innervation on sensory input activity, and on behavioral odor discrimination. We find that OA neurons innervate all the major classes of calyx neuron to some degree, although only a small fraction of KCs. Activating OA neurons can affect both PN activity, and odor discrimination during learning and memory, without affecting underlying olfactory learning and memory ability.
Our findings reveal a synaptic mechanism “gating” at the second synapse in the olfactory pathway, that has an impact on odor discrimination in learning.
Results
sVUM1 neurons in the calyx and their polarity
Two OA neurons innervate throughout the calyx without obvious regional preference
The larval calyx is innervated by two classes of OA neurons, sVUMmd1, and sVUMmx1, originating from the mandibular and maxillary neuromeres, respectively, in the subesophageal zone (SEZ), and labeled by the Tdc2-GAL4 line (Selcho et al. 2014). To visualize the innervation pattern of both sVUM1 neurons in the calyx together, we used the Multicolor FlpOut technique (Nern et al. 2015). Flies of genotype pBPhsFlp2::PEST(attP3); HA_V5_FLAG_OLLAS were crossed to flies of Tdc2-Gal4, and single cell clones were generated by heat shock. Each sVUM1 neuron ramified throughout the calyx, and we only ever found a single sVUMmd1 or sVUMmx1 neuron labeled. When both sVUM1 neurons were labeled, they ramified through the calyx in a non-overlapping pattern, without obvious regional preference (Fig. 1).
OA neurons are presynaptic in the calyx
To visualize the polarity of the calyx-innervating sVUM1 neurons, we used Tdc2-GAL4 to express either plasma membrane, presynaptic or dendritic markers (Fig. 2). The sVUM1 projections visualized by the plasma membrane marker CD4::tdTom showed dense ramification throughout the calyx, with bouton-like enlargements among glomeruli and in the core of the calyx (Fig. 2A). These enlargements contained OA, and were therefore presynaptic boutons; we observed 89 ± 7 (n= 12) OA-positive boutons per calyx. The presynaptic nature of the Tdc2-GAL4-expressing calyx boutons was further supported by the localization of the presynaptic markers nSyb::GFP (Fig. 2A) and Syt::GFP (Fig. 2B) in puncta throughout the calyx in these neurons, prominently between glomeruli or in the non-glomerular core of the calyx. The dendritic marker DenMark::mCherry is not detectable in the calyx (Fig. 2B), although it is strongly localized in the SEZ region (Fig. 2C), which may include other Tdc2-GAL4-expressing neurons as well as the sVUM1 neurons.
Identifying calyx OA neuron partners by GRASP
To visualize synaptic contacts between sVUM1 neurons and other calyx neurons, we used Tdc2-LexA, along with GAL4 lines expressing in other calyx neurons, to drive the expression of the GRASP constructs LexAop-CD4::spGFP11 and UAS-CD4::spGFP1-10 (Fig. 3). We labeled olfactory PNs using NP225-GAL4 (Tanaka et al., 2009), and KCs using MB247-GAL4 (Zars et al. 2000). We also tested for GRASP between sVUM1 neurons and two other classes of extrinsic calyx neurons. First, we labeled the larval APL using NP2361-GAL4 (Masuda-Nakagawa et al., 2014), Second, the “Odd” class of neurons shows dendritic arborizations throughout the calyx, with presynaptic terminals in the vicinity of the MB lobes (Slater et al., 2015), and appear to be the same as those designated as MBON-a1 and MBON-a2 by Eichler et al (2017); we identified a GAL4 line, OK263-GAL4, which expresses in these neurons.
GRASP fluorescence
We detected GRASP using GFP fluorescence as widely distributed puncta in the calyx, suggestive of specific synaptic connections, between the sVUM1 neurons on the one hand, and PNs, KCs, the Odd, and APL neurons on the other (Fig. 3A). These findings suggest that the sVUM1 neurons may form synapses with all the neuronal classes that innervate throughout the calyx: PNs, KCs, the APL, and Odd neurons.
To test whether GRASP signals represented synaptic contacts of the sVUM1 neurons, we also immunolabeled brains with anti-OA. GRASP signals were identified, using a criterion that each GRASP signal was observed in at least two consecutive confocal sections, and discounting occasional GFP-positive axonal tracts that were negative for OA.
OA termini synapse with PNs
Using Tdc2-LexA and NP225-GAL4 to express the GRASP constructs, we found GRASP signal at 49±3% of 80±4 (n=5) OA-positive boutons. OA-positive GRASP signals (Fig. 3B) were found in the core of the calyx away from glomeruli, in interglomerular spaces, and along the periphery of glomeruli. They are likely contacts between sVUM1 termini and PN axons. Around 13.5% (n=5) of total GRASP signals did not overlap with OA, and we cannot conclude that these are synaptic contacts. However, PNs are commonly postsynaptic to the termini of sVUM1 neurons, on their axonal or presynaptic processes, making axon-axon synapses.
OA termini synapse with KCs in the calyx
Using Tdc2-LexA and MB247-GAL4 to express the GRASP constructs, we found GRASP signals at 51 ± 4 % of 185 ± 34 (n=3) OA boutons (Fig. 3C).
OA termini synapse with Odd neurons in the calyx
Using Tdc2-LexA and OK263-GAL4 to express the GRASP constructs, we found GRASP signals at 51 ± 0.3 % of 123 ± 4.2 (n=3) OA labeled boutons. There were OA spots that were apposed to GFP (Fig. 3D).
OA termini synapse with the APL
Using Tdc2-LexA and NP2361-GAL4 to express the GRASP constructs, we found GRASP signal at 77±4% of 74±12 (n=4) OA terminals, indicating that sVUM1 neurons are presynaptic to the larval APL. Most of these GRASP signals were found between glomeruli, and more abundant towards the ventral calyx (Fig. 3E top panels).
Around 28.3 ± 1 GFP signals, 32.8 % ± 2.6 (n=3) GRASP signals, did not overlap with OA, and therefore do not represent synapses of the sVUM1 neurons onto the APL. Some of these appear to be tracts that lead to OA positive spots, rather than the characteristic round shape of synaptic GRASP signals, and might therefore be due to proximity of the sVUM1 axons with APL axons. Others could be due to synapses of the APL onto OA neuron axons. GABA labeling of brains of the same genotype, showed some GABA termini in close proximity to GRASP (Fig. 3E, bottom panels), suggesting that some contacts between APL and sVUM1 neurons could potentially be inputs from APL terminals onto the sVUM1 neurons. Labeling of Tdc2-GAL4>mCD8::GFP calyces with anti-GABA also showed some apposition of sVUM1 boutons to GABAergic termini of the APL (Fig. 3F).
Single cell GRASP
While the above GRASP experiments reveal the partners of the sVUM1 neurons in the calyx, they do not reveal whether the sVUMmd1 and sVUMmx1 have different partners. We therefore performed single-cell GRASP to label the contacts of each sVUM1 randomly, using Brp::mCherry as a presynaptic marker to verify whether GRASP signals have a synaptic localization. We distinguished the two sVUM1 neurons by the positions of their cell bodies in the SEZ, using local neuropil landmarks revealed by anti-Dlg labelling (Fig. 4A), and using single cell clones, we could identify sVUMmd1 and sVUMmx1 individually (Fig. 4B).
GRASP signals were detected between the odd-expressing neurons and both sVUMmd1 and sVUMmx1 (Fig. 4C), and similarly between the larval APL, and both sVUMmd1 and sVUMmx1 (Fig. 4D). Compared to standard GRASP, single cell GRASP signals were fewer, but clearly present and overlapping with the presynaptic marker Brp::mCherry. (Fig. 4D). Therefore, at least as judged by the APL and Odd neurons, both sVUM1 neurons appeared to have similar targets in the calyx.
Localization of a genomic Oamb::GFP fusion in PN terminals in calyx
To further understand how and where OA might act in the calyx, we investigated the localization of OA receptors in the calyx. Drosophila has a number of OA receptor classes defined both by sequence comparisons and pharmacology. Octopamine receptor in mushroom bodies (Oamb, also known as Dmoa1A or CG3856), an ortholog of human a1-adrenergic receptor (Roeder et al., 2003; Bauknecht and Jékely, 2017), is enriched in the MBs (Han et al.,1998). Drosophila also has three OctβR receptors, which stimulate cAMP levels (Maqueira et al., 2005, Balfanz et al., 2005).
To detect the expression and subcellular localization of Oamb, we used recombinase-mediated cassette exchange (RMCE) with a MiMIC insertion (Venken et al., 2011), MI12417, in the third coding-region intron of Oamb, to tag endogenous Oamb with an exonic EGFP- FlAsH-StrepII-TEV-3xFlag fusion (Supp Figs. 1-4). Insertion of an EGFP-encoding exon here should tag all known splice variants of the Oamb protein in their third cytoplasmic loop, downstream of transmembrane (TM) domain 5 (Supp. Fig. 5,6); this includes the alternative TM6-TM7 regions encoded by two alternative groups of C-terminal exons (Supp Figs 4-6). Therefore, a protein trap generated from the MI12417 insertion will not disrupt any transmembrane domains.
Six recombinant Oamb::EGFP stocks were recovered with the EGFP-encoding exon inserted in the same orientation as the Oamb transcript (Supp Fig. 7). One of these was designated as Mi{PT-GFSTF.1}OambMI12417-GFSTF.1, or Oamb(MI12417)::EGFP.1 or Oamb::EGFP for short. Both the original MI12417 MiMIC insertion, and Oamb(MI12417)::EGFP stocks were homozygous infertile, as expected from the egg-laying defects of Oamb mutants (Deady and Sun, 2015), suggesting that the Oamb::EGFP fusion might not be a functional Oamb protein. However, Oamb::EGFP was localized to glomeruli in the larval calyx (Fig. 5), implying that the protein folded normally and was not degraded by the ER unfolded protein response. Expression of UAS-RFP in the olfactory PN line NP225-GAL4 showed localization of Oamb::EGFP in all PN termini labeled with the GAL4 line, as well as in some calyx glomeruli not labeled by NP225-GAL4, which may be either sites of non-olfactory sensory input, or olfactory glomeruli not labeled by NP225-GAL4 (Masuda-Nakagawa et al. 2010) (Fig. 5A). The restriction of Oamb::EGFP to specific glomeruli implies that it is unlikely to be expressed in KC dendrites in the calyx, which arborize through all glomeruli. We also found no overlap of Oamb::EGFP with GABAergic APL terminals in the calyx (Fig. 5B), implying that it was not expressed in the larval APL. OK263-GAL4 calyx projections also showed little or no overlap with Oamb::EGFP (Fig. 5C), suggesting that Oamb is not expressed in the Odd neuron calyx dendrites.
We found no detectable localization of GFP-tagged DmOctβR receptors to the calyx (data not shown). Octβ1R::EGFP (CG6919) was expressed weakly in a few ventral and medial AL glomeruli, but was not detectable in the calyx. Octβ2R::EGFP (CG6989), was not detectable in either the calyx or AL, although it was expressed in a number of adjacent cell bodies that did not colocalize with PNs as labeled by NP225-GAL4 driving RFP expression. We could not detect any expression of Octβ3R::EGFP.
Stimulation of OA neurons enhances odor-evoked responses in calyx PN terminals
Since Oamb is in the α1-adrenergic receptor family and activates release of Ca2+ from intracellular stores via Gq (Morita et al., 2006; Lee et al., 2009), we tested whether activation of OA neurons could increase odor-evoked responses in calyx PN terminals. We therefore measured calyx PN terminal responses, using NP225-GAL4 and UAS-JRCaMP1b, in the presence of a LexAop-ChR2(XXL) construct either in the absence or presence of the OA driver line Tdc2-LexA. We measured responses in individual calyces, to (sequentially) (i) an odor pulse (“odor-only”); (ii) an odor pulse delivered 0.5 seconds after the end of a blue light pulse (“light+odor”); (iii) a pulse of blue light to activate ChR2(XXL) if expressed (“light-only”). We corrected the light+odor response first by subtracting any light-only response in that calyx, and second a “bleach correction” for fluorescence decay between the odor-only and odor+light responses in control larvae lacking Tdc2-LexA. In larvae lacking Tdc2-LexA (and hence lacking specific ChR2(XXL) expression, we found no significant effect of prior light exposure on odor-evoked responses; in the presence of Tdc2-LexA, to drive ChR2(XXL) expression in OA neurons, we found a small but significant enhancement of odor-evoked responses in PN calyx terminals (Fig. 6). Therefore we conclude that OA innervation of PNs can enhance their presynaptic evoked Ca2+ responses, and hence likely increase neurotransmitter release.
Activating an OA neuron subset including sVUM1 neurons impairs behavioral odor discrimination
Since the calyx is a site where MB neurons integrate sensory information that comprises conditioned stimuli (CS) in associative learning, we reasoned that modulating the processing of this information might affect the ability of the brain to discriminate among CSs while learning, but without affecting its underlying learning ability. In particular, increased activity of PN inputs to the calyx might reduce the ability of the calyx to discriminate among odors. Therefore, to test whether OA innervation of the calyx affected odor discrimination during learning, we developed an assay that could distinguish odor discrimination assay from learning ability (Fig. 7). The rationale of this assay was developed in the honeybee by Stopfer et al. (1997); molecularly similar odors are more difficult to discriminate than molecularly dissimilar odors, giving a measure of discriminability. In this case similar odor pairs are generated by varying the proportions of two dissimilar odors in two different odor mixes. By combining odor choice with an appetitive learning paradigm (Scherer et al. 2003), we tested the effect on behavioral odor discrimination of optogenetic activation of OA neurons in third instar larvae, using the long-wavelength absorbing channelrhodopsin, CsChrimson (Klapoetke et al., 2013).
Since we did not have GAL4 or LexA drivers completely specific for sVUM1 neurons, we used an intersectional approach to restrict the expression of CsChrimson to a small subset of OA neurons including the sVUM1 neurons. We could use LexAop-FLP and a GAL80 cassette flanked by two FRT sites to express UAS-CsChrimson only in neurons that expressed both GMR34A11-GAL4 (which labels some VUM neurons in addition to non-OA neurons), and Tdc2-LexA. We thus expressed CsChrimson in only five OA neurons in the SEZ of the larval brain: two (1.8 ± 0.39 (n=12)) in the mandibular neuromere (including sVUMmd1), two (2 ± 0.43 (n=12)) in the maxillary neuromere (including sVUMmx1), and one (1 ± 0 (n=12)) in the labial neuromere (Fig. 8). The second neuron labeled in each neuromere could be sVUM2 or sVUM3, however, sVUM1 neurons are the only neurons of this subset to innervate the AL and calyces.
Activation of these OA neurons, during conditioning but not during testing, had no effect on the ability of larvae to discriminate odors in an appetitive learning assay using a dissimilar odor pair, but abolished their ability to discriminate a similar odor pair, implying that odor discrimination is affected by activation of these neurons, but not underlying learning ability (Fig. 8).
Discussion
OA neurons target extrinsic neurons within the calyx circuitry
Two OA neurons originating in the SEZ, sVUMmd1 and sVUM mx1, innervate the same brain neuropiles, with postsynaptic processes in the SEZ and presynaptic processes in the antennal lobe and MB calyces (Fig. 2). Using GRASP, we found contacts of sVUM1 presynaptic terminals with KCs, PNs, Odd and APL neurons in the calyx, suggesting that sVUM1 neurons can potentially regulate these neurons. However, most KCs cannot be directly affected by sVUM1 innervation: the tens of contacts between the sVUM1 neurons and KCs is less than the 300-plus KCs present in third instar larvae (Pauls et al., 2010), and Eichler et al. (2017) also find inputs of either sVUM1 neuron (which they call OAN-a1 and OAN-a2) into fewer than 10-15% of KCs in first instar larvae. Therefore, context-dependent signaling by OA in the calyx must principally affect MB activity via its other input neurons, not by direct action on KC dendrites.
Connectomic analysis of a six-hour first-instar larva (Eichler et al. 2017) shows that the sVUM1 neurons, with only 36 active zones in a left brain, and 48 in a right brain, have a qualitatively similar but less extensive calyx innervation pattern than we observe in third instar, with 89 ± 7 [n=12] OA-positive boutons per calyx, and even more active zones, assuming multiple active zones per bouton. Consistent with our findings, Eichler et al (2017) also report substantial presynaptic contacts of sVUM1 neurons with Odd neuron dendrites, and some synapses of sVUM1 neurons with KCs, but not with most KCs; they do not comment on synapses with PNs or the APL.
OA modulation of PN activity
GRASP analyses suggest that PNs are postsynaptic to the sVUM1 neurons (Fig. 3). Indeed, an Oamb::eGFP exon-trap fusion is localized on PN presynaptic terminals, albeit more widely than GRASP puncta (Fig. 5). Similarly, the honeybee Oamb ortholog AmOA1 is found widely in the calyx (Sinakevitch et al., 2011), although these authors do not distinguish between PN terminals or KC dendrites. Much aminergic neurotransmission acts via extrasynaptic receptors (Bentley et al., 2016) and this may be the case for Oamb in PN terminals, Oamb is a GPCR that signals apparently through Gq, to release Ca2+ from intracellular stores (Balfanz et al., 2005; Morita et al, 2006); it may also elevate cAMP (Han et al., 1998), although this effect appears smaller (Balfanz et al., 2005). Consistent with this, optogenetic activation of sVUM1 neuron terminals in the calyx led to potentiation of odor-evoked [Ca2+] in PN terminals (Fig. 6), and should therefore enhance acetylcholine release from PN terminals. This effect could be serving as a gating mechanism to increase signal to noise ratio; at lower stimulus intensities this would facilitate the detection of subthreshold signals under behavioral states that promote OA release. Such a role for NA is also found in the mammalian olfactory bulb: here, LC input improves the detection of peri-threshold stimuli (Jiang et al., 1996), and facilitates the detection of near threshold odors, when rewarded (Escanilla et al., 2012), via an increase in mitral cell excitability mediated by NA action on a1-adrenergic receptor (Ciombor et al. 1999; Hayar et al. 2001). OA has comparable roles in the Drosophila central visual system, for example increasing the gain of motion-sensitive neurons in response to behavioral states such as walking or flight. (Chiappe et al. 2010; Maimon et al., 2010; Suver et al., 2012), sometimes via Ca2+ responses elicited by OA inputs in visual system neurons (Strother et al., 2018).
Roles of APL and Odd neurons in calyx activity
sVUM1 presynaptic termini also make many contacts with both the APL and Odd neurons (Fig. 3, 4). Odd neurons ramify throughout the calyx and receive input from PNs generally, thus forming a channel for non-selective odor processing that is parallel to the main MB odor-specific processing through KCs. A similar division of processing is seen in the visual pathway downstream of area V1, the primary sensory area in the cortex, with the segregation of the “what” pathway, for object recognition, from the dorsal “where” pathway that is involved in motion detection, visually guided movements and attentional control (Livingston and Hubel, 1988, Gilbert, 2013). sVUM1 neurons have extensive contacts with Odd neurons and could potentially change their gain or tuning properties, to signal changes in behavioural state that guide odor-driven choice behaviours. In fact Odd neurons have been implicated in chemotactic behaviour (Slater et al., 2015).
The APL mediates a negative feedback loop from KC outputs in the MB lobes to KC inputs in the calyx, thus potentially both limiting the duration of KC activity and improving their odor discrimination (Masuda-Nakagawa et al., 2014; Lin et al., 2014). sVUM1 synapses onto the APL in the calyx could therefore potentially regulate this feedback loop. This could increase signal-to-noise ratio, in a context-dependent manner, by sharpening odor representations in the calyx via APL inhibitory feedback, similar to the “gain control” mechanism with enhancement of behaviorally relevant responses and suppression of non-relevant ones, in monkey visual system, (Treue and Martinez Trujillo, 1999, Gilbert and Li, 2013). Similarly, songbirds possess a similar mechanism, with a gate open to the unique birdsong only during vocalization, in a behavioral state-dependent manner (Schmidt and Konishi, 1998). In addition, since we observed some GRASP signals adjacent to GABA termini (Fig. 3E), APL feedback could also inhibit OA release from sVUM1 termini, further increasing the complexity of interactions between OA innervation and the KC/APL negative feedback loop. NA regulation of inhibitory neurons is also a feature of the mammalian olfactory circuitry. In olfactory bulb, disinhibition of mitral cell (equivalent to PN) activity by NA regulation of inhibitory granule cells has been proposed (Nai et al. 2009). In mammalian PCx, feedforward and feedback inhibition are postulated to enhance cortical representation of strong inputs (Stokes and Isaacson, 2010), and the PCx receives extensive NA innervation from the LC, although its role in modulating inhibition has not been investigated.
Odor discrimination learning
A role for OA as a reinforcer in appetitive associative learning has been shown in honeybees and flies. The honeybee VUMmx1 neuron has properties of a reinforcer; its depolarization could replace sugar reinforcement in appetitive learning (Hammer, 1993), and injection of OA into the MB calyx induced memory consolidation (Hammer and Menzel, 1998). Furthermore, increased density of calyx microglomeruli is observed in long term memory (Hourcade et al. 2010) and KC claws showed increased responses after appetitive conditioning (Szyszka et al. 2008). It has been proposed that VUMmx1 learns about the value of the odor, and acts as a prediction error signal in appetitive learning (Menzel, 2012). However, associative plasticity in the MBs in Drosophila, is thought to reside mainly in the lobes rather than the calyx, for both appetitive (Schwaertzel et al. 2003; Schroll et al. 2006; Liu et al. 2012) and aversive learning (Aso et al., 2012, 2014). OA as a reinforcer in appetitive learning appears to act via Oamb expressed in PAM dopamine neurons that target the MB lobes, and a set of OA neurons that includes OA-VUMa2, the equivalent of larval VUMmx1, did not induce and is not required for appetitive learning in adult Drosophila (Burke et al. 2012). Furthermore, Oamb is required in KCs for adult appetitive learning (Kim et al., 2013), suggesting some direct input of an OA- encoded appetitive signal into KCs. Taken together, OA action as a reinforcer in associative learning might occur via unidentified inputs into dopaminergic neurons or KC lobes – a separate role from gating selected sensory information, which is favored by our imaging and behavioral data, and is also suggested as a role of NA (Berridge and Waterhouse, 2003).
Here we observed that the discrimination of similar odors was compromised, in a reciprocal associative learning paradigm, by VUM1 activation. One explanation is that VUM1 activation enhanced the gain of stimulus-driven PNs, increasing the magnitude and number of KCs responding, and therefore representations by KCs of CS+ and CS- increased in overlap, and lower discriminability of similar odors. Furthermore the effect was specific to the 5 neurons, including VUM1s targeted by our experiments.
Perspectives
Sensory representations are dynamically modified by higher brain signaling, according to behavioral state such as attention, expectation, and behavioral task; and LC (locus coeruleus) activation in mammals and OA activation in insects, correlate with changes in behavioral states. Mammalian olfactory neuropiles are densely innervated by noradrenergic input, similar to the dense octopaminergic sVUM1 neurons innervation of AL and calyx in insects. Although behavioral data is abundant, understanding of the circuit and synaptic mechanisms of NA/OA requires the identification of the synapses and receptors that they target. The innervation of sVUM1 neurons throughout the calyx, and their potential synaptic connections to PNs, KCs, APL, and odd-like neurons, suggest that OA induces a network level switch both via gating input afferent activity, and by interacting with the KC/APL feedback loop, and thus also affecting the output activity from the calyx – not only through KCs but also potentially via the Odd neurons. Behavioral demands would determine the balance between sensitivity and discrimination via OA; whether to escape from a predator at all cost, or the need to recognize closely similar smells associated to food or danger. The effect of brain signals on sensitivity and discrimination is also observed in the human brain; in hallucinations in schizophrenia (Dierks 1999). NA/OA are teaching signals downstream higher order brain commands that compute cognitive processes, future work is necessary to understand the circuits upstream NA/OA and how they regulate the dynamics of bottom-up sensory circuits.
Materials and Methods
Genetics and molecular biology
Fly Stocks
Flies were raised on standard cornmeal medium at 25°C and subjected to a 12 hour day/night cycle. Stocks used are listed in Table 1.
MultiColor FlpOut
MultiColor FlpOut was performed according to Nern et al. (2015). Females of genotype pBPhsFlp2::PEST(attP3); +; HA_V5_FLAG_OLLAS (“MCFO-2”) were crossed with male Tdc2-Gal4 flies. Parents were left in vials to lay eggs for 24-hour intervals. After another 24 hours, larval progeny were heat shocked at 35-37°C by immersion in a heated circulated waterbath for 15-30 minutes, thus ensuring larvae were aged 24-48 hours after egg-laying (AEL) at the time of heat shock.
GRASP
Standard GRASP was according to Gordon and Scott (2009). Reconstituted GFP was detected using rat monoclonal anti-GFP. This did not detect either of the GRASP components GFP1-10 or GFP11, when Gal4 or LexA drivers were used alone. GRASP signals had to meet a criterion of occurring in two consecutive 0.5-µm confocal sections. For single cell GRASP (Karuppudurai et al., 2014), we generated larvae carrying P{hsFLP}12, appropriate GAL4 and LexA combinations, and a recombinant chromosome with insertions LexAOp2-IVS>stop>spGFP11::CD4::HA-T2A-Brp::mCherry (attP2), UAS-spGFP1-10::CD4, and UAS- HRP::CD2. To generate labelled single cells, parents were allowed to lay eggs initially for 24-hour intervals, then for 6-hour intervals in vials containing half the amount of food. At 0-24h, 24-48h, or later at 12-18 h, 18-24 h, or 24-30 h AEL, progeny were heat shocked as above for 10-50 minutes at 37 °C. Progeny were incubated from RT until dissection of non-tubby wandering third instar larvae.
Generation of an EGFP-tagged Oamb line
The Mi{MIC}OambMI12417 insertion in coding intron 3 of Oamb at 3R:20697059, (BDSC 57940; henceforth referred to as MI12417) was verified by PCR using primers MI12417-5F/MiMIC-5R1 for the 5’ end and MI12417-3R/MiMIC- 3F1 for the 3’ end (Table 2; Supplementary Fig. 1). Sequencing of these PCR products and alignment with the Drosophila genome sequence using BLASTN (Altschul et al., 1990; https://blast.ncbi.nlm.nih.gov/) showed insertion of MiMIC at the recorded site of 3R 20697058- 9 (Supplementary Figs. 2,3). The location of the MI12417 insertion site relative to Oamb coding exons was determined by using Oamb-B sequences for BLASTN and TBLASTN queries of the Drosophila genome assembly (http://flybase.org/blast/; Supplementary Fig. 4). TMHMM (Sonnhammer et al., 1998; http://www.cbs.dtu.dk/services/TMHMM/) was used to predict the amino-acid coordinates of Oamb transmembrane (TM) domains (Supplementary Figs. 5,6).
To insert an EGFP-encoding exon into the MI12417 insertion by RMCE, we chose the splice phase-1 version of EGFP-FlAsH-StrepII-TEV-3xFlag plasmid (DGRC 1306; Venken et al., 2011) as recommended by the Baylor Gene Disruption Project (http://flypush.imgen.bcm.tmc.edu/pscreen/rmce/rmce.php?entry=RM00888). This was co-injected with a helper phiC31-integrase plasmid (Venken et al., 2011) by the Drosophila microinjection facility (Department of Genetics, University of Cambridge). Injected embryos were left to hatch into adult flies and crossed to a y w double balancer stock. RMCE events were identified by loss of the MiMIC yellow+ marker in F1 progeny. Four PCR reactions were carried out to determine the orientation of the EGFP cassette in each recombinant Oamb::EGFP stock (Table 2) as described in Venken et al. (2011).
OctβR EGFP fusions
For Octβ1R and Octβ3R, we used lines Octβ1RMI05807-GFSTF.2 and Octβ3RMI06217-GFSTF.1 (Bloomington stocks 60236 and 60245), respectively, as Octβ1R::EGFP and Octβ3::EGFP exon traps. For Octβ2R, we used Mi{MIC}Octβ2RMI13416 to generate Octβ2RMI13416-GFSTF.2 using a similar approach as that described above for Oamb::EGFP. We also generated a second Octβ3::EGFP exon trap, Octβ3RMI06217-GFSTF.0, in a different reading frame from Bloomington stock 60245, to allow for the possibility of the EGFP exon being spliced as in transcripts RJ or RK (frame 0, with only RJ able to encode all seven TM domains), rather than RF or RG (frame 1). Positions of each insertion were confirmed by PCR and sequencing similarly to Oamb, using primers as described in Table 2.
Molecular methods
Genomic DNA was extracted from 15-30 flies (1-7 days after eclosion) and homogenized in 100 mM Tris-HCl, 8.5; 80 mM NaCl, (Sigma 31434); 5% Sucrose (Sigma S0389); 0.5% SDS (Sigma L4509); 50 mM Na-EDTA (Sigma ED2SS), pH 8.0. The homogenate was incubated with RNase A (Roche 10109142001) for 1 hour at 37°C followed by Proteinase K (Roche 03115887001) for 1 hour at 50°C, and purified with phenol-chloroform (Sigma 77617) and chloroform (Sigma C2432). DNA was precipitated with 0.6 volumes of isopropanol (Sigma, 59304) and washed with 75% ethanol (Sigma E7023), dried overnight at room temperature and re-suspended in 10 mM Tris-HCl pH 8.0 (Sigma T6066).
PCR reactions (20 μl) contained 0.4 μl or 1 μl genomic DNA, 1 μl of each 10 μM primer (Sigma), 0.08 μl of 5 U/μl HotStarTaq DNA polymerase (Qiagen 203203) and 15.1 μl or 14.5 μl milliQ water. PCR cycling in a G-Storm Thermal Cycler (GS4) was: 15 minutes at 95°C; 40 cycles of: denaturation at 94°C for 30s, annealing at 60°C for 30s and elongation at 72°C for 1 min; and a final elongation step at 72°C for 10 minutes. PCR products were loaded with 6X DNA gel-loading dye (ThermoFisher R0611) on a 1% Agarose Gel (LifeTech 16500500; 1X TBE buffer, LifeTech 16500500) with GelRed (Biotium 41003-T) for gel electrophoresis. 100 bp DNA ladder was used as a marker (LifeTech 15628019). PCR products were purified using the Qiaquick PCR Purification Kit (Qiagen, 28104), and sequenced at the Department of Biochemistry Sequencing Facility (University of Cambridge).
Immunohistochemistry and Confocal imaging
Third instar wandering larval brains (144-176 hours AEL) were dissected in cold PBS (Sigma P4417), fixed in 4% Formaldehyde (Polysciences 18814) / PEM buffer (0.1 M PIPES, Sigma P1851; 2 mM EGTA, Sigma E3889; 1 mM MgSO4; NaOH) for 2 hours at 4°C, washed for 3×10 minutes (or 4×15 minutes) in 0.3% Triton-X (Sigma T8787) in PBS (PBT) and incubated in 10% NGS (Normal goat serum; Vector Labs S-1000) in 0.3% PBT for 1 hour at room temperature. Brains were incubated in primary antibody in 10% NGS-0.3% PBT at 4°C for 2-3 days on a mini disk rotor (Biocraft, BC-710), washed for 3×15 minutes with 0.3% PBT and further incubated in secondary antibody in 10% NGS at 4°C for 2-3 days again on the mini disk rotor. Brains were finally washed 1×15 minutes with PBT, followed by 3×10 minutes with PBS, and left in 50% Glycerol/PBS at 4°C for at least one overnight prior to imaging.
Primary and secondary antibodies are listed in Table 3. Brains were incubated in primary antibody at 4°C for 2-3 nights, washed three times in PBT for 15-minutes, and incubated in secondary antibody for 2-3 more nights. To reduce background with the polyclonal chicken anti-GFP (Abcam, Ab13970), it was pre-incubated with MI12417 larval brains which do not express GFP. Fifty MI12417 larval brains were incubated in 1:20 chicken anti-GFP in 10% NGS in 0.3% PBT at 4°C for overnight. A further 50 MI12417 larval brains were added and further incubated at 4°C over 2 nights. Mounting and orientation of brains for image acquisition was as described in supplemental information in Masuda-Nakagawa et al. (2009). Imaging was carried out using a Zeiss LSM710 Confocal Microscope with a 40X NA1.3 oil objective. Images were processed using ImageJ software (https://imagej.nih.gov/ij/download.html).
Optogenetics and live imaging
All crosses were performed on cornmeal-yeast-agar medium supplemented with 100 µM all-trans-Retinal (Sigma, R2500). Crosses were kept at 23°C in the dark wrapped in tinfoil, and when necessary handled under dim amber light (591 nm).
Combined optogenetics and activity imaging was performed by crossing LexAop-ChR2-XXL; UAS-jRCaMP1b/TM6B virgin females to stocks expressing relevant GAL4 and LexA insertions. ChR2-XXL function was confirmed by crossing these females to nSyb-LexA/CyO::GFP males and testing undissected, non-CyO::GFP larval progeny for light-induced body contraction under imaging conditions. Driver expression in all GAL4 and LexA combination lines was tested by imaging RFP and GFP expression in the larval progeny of a cross between males of each line to virgin females of a UAS-RFP LexAop-GFP double reporter line (Bloomington stock 32229). As a positive control, virgins of genotype LexAop-ChR2-XXL; UAS-jRCaMP1b/TM6B were crossed either to NP2631-GAL4; MB247-LexA/TM6B to activate KCs and image APL responses, or to NP2631-GAL4 males as a negative control. To activate OA neurons and image PNs, LexAop-ChR2-XXL; UAS-jRCaMP1b/TM6B virgins were crossed to NP225-GAL4; Tdc2-LexA/TM6B males, or to NP225-GAL4; MKRS/TM6B males as a negative control, and only non-TM6B (non-Tb) larvae were imaged.
Humidified odors were presented through valves controlled by a Master-8-cp controller (AMPI, Jerusalem, Israel) as described (Masuda-Nakagawa et al., 2009). The concentration of EA (Ethyl Acetate) was reduced to 1/10,000 so that only a 1-3 PN terminals in the calyx responded to a 2-second odor presentation in any given optical slice.
Wandering stage L3 Drosophila were dissected and mounted for imaging as previously described (Masuda-Nakagawa et al., 2009) under an Olympus BX50-WI microscope with a Zeiss W Plan-Apochromat 40x/1.0 DIC M27 objective. Dissection was performed under dim amber (591 nm) light and the condenser light of the BX50-WI was passed through an ET632/60 M emission filter. A 470nm LED filtered through a ET470/40x (Chroma, VT, USA) emission filter and regulated by a Dual LED Power Supply (Cairn Research, Faversham, UK) was used to illuminate the sample through the objective via a Cairn OptoLED LED mount (Cairn Research) onto the Olympus BX50-WI BX-FLA vertical illuminator. The vertical illuminator aperture was minimized to a diameter of approximately twice that of the calyx (from a dorsal view) and centered within the camera’s field of view. A T495lpxr dichroic mirror (Chroma, VT, USA) directed the LED onto the sample while preventing 470nm light reflected off the sample that reliably evoked contractions in the body wall of mounted LexAop-ChR2-XXL/nSyb-LexA; UAS-jRCaMP1b/+ larvae.
jRCaMP1b fluorescence was imaged using an Andor iXon+ DU-888E-CO-#BV EM-CCD camera (Andor, Belfast, UK) attached to a CSU22 spinning disc confocal (Yokogawa Electric Corporation) via a Cairn Research Optosplit II (Faversham, UK), all mounted on the above Olympus BX50-WI. A 561nm excitation laser was used for all imaging, at 20% laser power; at this power, LexAop-ChR2-XXL/nSyb-LexA; UAS-jRCaMP1b/ MKRS larvae showed no movement in response to the laser. The room was kept at 23°C. The Master-8-cp controller synchronized the timing of imaging, LED illumination, and odor delivery, while images were recorded using Micro-Manager (Edelstein et al., 2010).
To image the effect of OA neuron depolarization on odor-evoked responses of PN axon termini, we identified either the left or right calyx of each brain and centered it within the field of view. All preparations were left to settle for at least two minutes before experiments began, and any still moving at 15 minutes were discarded. We performed four sequential imaging paradigms on each preparation; each acquisition lasted 12 seconds at 5 frames per second with a 100- ms exposure time, with a gap of at least 30 seconds between acquisitions. The first paradigm was an odor-only control, a 2-second odor pulse presented after 4 seconds. If no response was found, we repeated the presentation on a different z-section of the calyx. If no response was observed after 4 attempts, we discarded the preparation. The second paradigm was light+odor; we first presented a 2-second light pulse from the 470nm LED after 1.5 seconds, and then presented odor as in the first paradigm, i.e. 0.5 seconds after the end of the 2-second light pulse. Calyx glomeruli were used as landmarks to refocus as necessary. The third paradigm was a light-only control, delivered as in paradigm 2. The fourth and final paradigm was the same as the first, to control for preparation degradation over time; if no odor response was observed, the preparation was discarded.
To quantify activity in PN terminals, the experimenter was blinded to the genotype of each time-series of images. Each image of each time series was motion corrected using the moco plugin (Dubbs et al., 2016). Regions of interest (ROIs) for the best responding glomerulus in the odor-only paradigm (judged by eye during acquisition), and for background fluorescence outside the calyx, were drawn manually on the frame with peak fluorescence. In each consecutive imaging paradigm, the ROI was redrawn around the glomerulus to compensate for drift of the preparation. The average intensity for each ROI was calculated from each frame, and background fluorescence was subtracted. Each acquisition consisted of 60 frames of background-subtracted intensity data acquired over 12 seconds. To smoothen out frame-to-frame variation, we used the overlapping moving averages of three sequential frames, leaving 58 frames of smoothed data.
To determine the effect of OA neuron activation on PN terminal activity, we performed two comparisons: the first was a control comparison, of odor-evoked PN responses in the absence (Paradigm 1) or presence (Paradigm 2) of prior optogenetic activation of OA neurons, in control brains of genotype LexAop-ChR2-XXL/ NP225-GAL4; UAS-jRCaMP1b/ MKRS. The second comparison was similar, but in brains of LexAop-ChR2-XXL/ NP225-GAL4; UAS-jRCaMP1b/ Tdc2-LexA, in which ChR2-XXL expression is targeted to OA neurons. To calculate the odor-evoked response to OA neuron activation, we applied two corrections to the Light+Odor response (Paradigm 2): we first subtracted the Light-Only response in the same brain (Paradigm 3); we then applied a bleach correction factor to each timepoint in each preparation, since the maximal responses were slightly lower in Paradigm 3 than in Paradigm 1. The bleach correction factor was calculated by (i) averaging the responses for each frame, from all preparations in the odor-only (paradigm 1) and light+odor (paradigm 2) responses; (ii) summing these averaged responses from the start of the odor pulse (frame 21) until the end of the acquisition (frame 59), to obtain a measure of the total strength of the response over the acquisition period; (iii) dividing the summed average response for odor-only by the summed average response for light+odor, to achieve a correction factor of 1.103. The timecourse of average responses in the absence and presence of light matched closely for the control larval genotype, but light activation give responses that were higher than without light consistently across the whole timecourse of the odor response. After corrections, we measured the total evoked response of each calyx, integrated over 6 seconds from the end of the odor pulse to the end of the acquisition, and assayed the effect of OA neuron activation in each individual calyx by comparing responses between the absence (paradigm 1) and presence (paradigm 3, corrected as above) of light. We used Prism (Graphpad Software, San Diego, CA, USA) to generate graphs, and to perform non-parametric Wilcoxon matched-pairs signed rank tests to compare the two conditions within control calyces lacking Tdc2-LexA, and again compare the two conditions within calyces expressing ChR2-XXL under control of Tdc2-LexA. Representative panels for display were chosen so that the difference between corrected “light+odor” and “odor-only” responses was close to the median value, and to have visually similar resting fluorescence levels.
Behavioral assay
Larval culture
Males of genotype w; Tdc2-LexA; GMR34A11-GAL4/TM6B were crossed to females of genotype Tub84B(FRT-GAL80)1, w; LexAop-FLP; UAS-Chrimson::mVenus/TM6B to generate F1 larvae in which UAS-Chrimson::mVenus could be expressed only in cells expressing both Tdc2-LexA and GMR34A11-GAL4, in which LexA-dependent FLP expression had removed the GAL4 inhibitor GAL80. Larvae were allowed to develop in food vials containing 100 µM all-trans-retinal, in the dark at 21 degrees. For both the retinal and non-retinal food vials, transfer of adults into new vials happens both in the morning and in the evening, to then collect them after 108-120 hours in the case of non-retinal vials at 25°C, and after 132-144 hours for those kept in retinal vials at 23°C.
Behavioral arena
Agarose petri dishes were prepared the day before use, by using 100 ml of distilled water with 0.9% agarose (Sigma A9539). Fructose petri dishes were prepared similarly, but containing 35% fructose (Sigma-47740). Petri dishes had perforated lids, to facilitate creation of odorant gradients within the dish, by sucking air from a benchtop fume extractor (Sentry Air Systems (SAS), SS-200-WSL) at the back of the assay platform. Odorants were diluted in 10-µl volumes of paraffin oil (Sigma-Aldrich 76235), held in custom-built Teflon containers with pierced lids (7 holes), on the surface of agarose plates.
Light apparatus
The light apparatus contained a BK Precision pulse generator attached to a DC power pack, driving amber light LEDs (591 nm), constructed as described by deVries and Clandinin (2013), by the Psychology Department workshop of the University of Cambridge. One cycle of pulses consisted of 10-ms pulses at 10Hz for 30s, followed by 30s without pulses. This cycle was repeated 5 times, making a conditioning step of 5 minutes. This pulse frequency and width were chosen to replicate the activity of the only recorded sVUM1 neuron, the honeybee sVUMmx1, when activated by a sucrose reward (Hammer et al., 1993).
Behavior conditioning
Third-instar larvae were collected from vials using a metal sieve, and washed with tap water. Larvae were washed twice by transferring through a drop of water, and then placed on the conditioning agarose plate (35% fructose). A Teflon container with 10 μl of odor A to be reinforced was placed on the left side, and another containing paraffin oil (neutral) symmetrically on the right side, at positions marked on a black background, used to facilitate visualization of the larvae (Fig. 7). Larvae were conditioned on the fructose plate with odor A for 5 minutes under weak blue light. Larvae were then transferred to a water droplet using a brush, and again to a second water droplet to ensure no fructose traces remained, and then to an agarose plate lacking fructose, on which a Teflon container with 10 μl of odor B (non-reinforced) was placed on the left side and a container with paraffin oil (neutral) on the right side. Larvae were conditioned for 5 minutes under weak blue light as above. This conditioning procedure on both plates was repeated for three cycles. For experiments using activation of OA neurons, the entire conditioning cycles were carried out under amber light (Fig. 7).
Odor dilutions
To avoid bias based on innate preferences of odors, odor concentrations were balanced to achieve a response index of around zero. Dilutions of ethyl acetate (EA) at 1:2000 and pentyl acetate (PA) at 1:500 were used. The dissimilar odor pair was ethyl acetate (EA) (Sigma-Aldrich, cat no. 319902) diluted 1:2000, and pentyl acetate (PA) (Sigma-Aldrich, cat no. 109584) diluted 1:500. The similar odor pair was mixtures of EA and PA at dilutions of 1:2000, and 1:500, respectively, at proportions of 1:4 EA:PA, and 4:1 of EA:PA.
Testing
Larvae were tested by placing them on an agarose plate carrying a container with odor A on one side, and a container with odor B on the other. Test was under blue light for 5 minutes. Larvae were counted on the side of odor A (conditioned), odor B (unconditioned) or in the neutral zone in the middle. In some experiments, larvae were transferred during the above procedures under white fluorescent ceiling lighting rather than weak blue light, but we detected no difference in learning scores between these conditions using ANOVA, as described in the Results section. Performance Index (PI) was calculated as: Learning Index (LI) was calculated after both a conditioning with odor A and a reciprocal conditioning with odor B (with a different sample of larvae), using the formula:
Statistical analysis
Planned comparisons were performed using t-tests. A four-factor ANOVA using SPSS software was used to test whether learning index was affected by food type (retinal vs. non-retinal), light (blue vs. amber), odor pairs (similar pair vs. dissimilar pair), or experimenter (either of two).
Author contributions
Experimental design: LMMN, CJO’K
Performed experiments: LMMN, ADM, BAW, JYHW, MM, SWZ.
Data analysis: LMMN, CJO’K, ADM, BAW.
Paper writing: LMMN
Paper editing: CJO’K
Acknowledgements
We thank T. Awasaki, S. Certel, K. Ito, M Landgraf, T. Lee, C.H.-Lee, B Pfeiffer, M. Ramaswami, K. Scott, J Truman, and the Bloomington Drosophila Stock Center (BDSC) for numerous fly stocks, and the Developmental Studies Hybridoma Bank (DSHB), University of Iowa, for antibodies. We thank Tom Bland for helping set up the larval behavior assay, and M Morgan for help in building optogenetic illumination apparatus.
JYHW was supported by a Medical Research Council studentship and MM was supported by a UK Genetics Society “Genes and Development” summer scholarship and an award from the Bedford Fund, King’s College Cambridge. This work was supported by BBSRC Grants BB/I022651/1 and BB/N007948/1 and an Isaac Newton Trust award to LMM-N and CJO’K.
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.
- 19.
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.
- 35.↵
- 36.↵
- 37.↵
- 38.
- 39.↵
- 40.↵
- 41.↵
- 42.
- 43.
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.
- 57.↵
- 58.↵
- 59.
- 60.↵
- 61.
- 62.
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.
- 73.↵
- 74.
- 75.↵
- 76.↵
- 77.
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.
- 83.
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.
- 90.↵
- 91.↵
- 92.↵
- 93.↵