Mechano-dependent signaling by Latrophilin/CIRL quenches cAMP in proprioceptive neurons

Adhesion-type G protein-coupled receptors (aGPCRs), a large molecule family with over 30 members in humans, operate in organ development, brain function and govern immunological responses. Correspondingly, this receptor family is linked to a multitude of diverse human diseases. aGPCRs have been suggested to possess mechanosensory properties, though their mechanism of action is fully unknown. Here we show that the Drosophila aGPCR Latrophilin/dCIRL acts in mechanosensory neurons by modulating ionotropic receptor currents, the initiating step of cellular mechanosensation. This process depends on the length of the extended ectodomain and the tethered agonist of the receptor, but not on its autoproteolysis, a characteristic biochemical feature of the aGPCR family. Intracellularly, dCIRL quenches cAMP levels upon mechanical activation thereby specifically increasing the mechanosensitivity of neurons. These results provide direct evidence that the aGPCR dCIRL acts as a molecular sensor and signal transducer that detects and converts mechanical stimuli into a metabotropic response. DOI: http://dx.doi.org/10.7554/eLife.28360.001


Introduction
Sensory strategies for the perception of mechanical cues are essential for survival. However, our understanding of the underlying molecular mechanisms is far from complete. G protein-coupled receptors (GPCRs) hand over stimulus-induced conformational changes to metabotropic signaling outlets that carry the signal to intracellular destinations.
Adhesion-type G protein-coupled receptors (aGPCRs) display structural characteristics that distinguish them as a separate family within the GPCR superfamily . Remarkably, as opposed to the majority of GPCRs, aGPCRs interact through their N-termini with membrane-tethered or ECM-fixed partner molecules rather than soluble compounds indicating that their function requires positional fixation outside the receptor-bearing cell .
Several aGPCRs have recently been linked to mechanosensitive functions (Petersen et al., 2015;Scholz et al., 2015;White et al., 2014). These examples collectively suggest that processing of mechanical stimuli may be a common feature of this receptor family (Langenhan et al., 2016). However, while elemental signaling properties of aGPCRs have recently become available , a molecular model of their signal transduction strategy is at large.
By combining genomic engineering with electrophysiological recordings, super-resolution microscopy and optogenetics, we have determined the critical steps that are required to transduce a mechanical stimulus into an intracellular response by an individual aGPCR, Drosophila Latrophilin/ dCIRL. We have taken advantage of the functional modulation of mechanosensory neurons by dCIRL and the accessibility of this system for physiological interrogation in vivo. Our results show that dCIRL is located in the neuronal dendrites and cilia of chordotonal organs (ChOs), the sites of ionotropic mechanotransduction (Ranade et al., 2015). dCIRL specifically shapes the generation of mechanically-gated receptor currents but is dispensible for normal membrane excitability of ChO neurons. Lengthening dCIRL's N-terminal fragment (NTF) gradually reduces mechanosensory neuronal responses. This is consistent with a model in which mechanical tension applied to the receptor determines the extent of its activity. In contrast, autoproteolysis of the GAIN domain is not essential for dCIRL activity, which instead requires an intact Stachel sequence. Finally, we show that mechanical stimuli effect a dCIRL-dependent decrease of cAMP levels in ChO neurons.

Results dCIRL is located in dendrites and cilia of mechanosensory neurons
To precisely determine the expression of dCirl in larval mechanosensory chordotonal organs (ChOs), we used a dCirlp GAL4 promoter element to drive the nuclear reporter UAS-GFP::nls and analyzed immunohistochemical stainings against GFP and HRP, a comarker of ChO neuron structure. In the larval pentascolopidial ChO (lch5) only the five neuronal nuclei were marked (Figure 1a), showing that dCirl is a neuronal gene. To obtain a translational expression profile of dCIRL, we constructed a genomic transgene that contains an mRFP cassette inserted into an exon encoding part of the extracellular domain (ECD) of the receptor at a position where its folding and trafficking should not be affected (dCirl N-RFP ; Figure 1-figure supplement 1) (Scholz et al., 2015). The dCIRL N-RFP fusion protein could be observed in the lch5 at the level of the dendrite and cilia ( Figure 1b). Next, we employed super-resolution imaging by structured illumination microscopy (SIM) to resolve the subcellular arrangement of dCIRL in greater detail (Gustafsson, 2000). SIM images depicted a patchy distribution of dCIRL N-RFP at the membrane of the lch5 dendrite and cilium, where it localized near the TRP channel TRPN1/NompC (Yan et al., 2013;Zhang et al., 2015) (Figure 1c). This demonstrates that dCIRL resides at the location where ionotropic mechanosensation operates.
The ultrastructure of dCirl KO chordotonal organs is unaffected As dCIRL possesses molecular characteristics of adhesion molecules, we performed ultrastructural analyses to ascertain that removal of dCirl does not affect the complex architecture and structural integrity of ChOs. Scanning electron microscopy uncovered no structural anomalies in dCirl KO mutants (dCirl Rescue : n = 11 ChOs from 5 larvae; dCirl KO : n = 11 from 5 larvae; Figure 1d,e). Additionally, the ultrastructure, cell-cell and cell-matrix contacts of distal inner dendrites and cilia appeared unaltered in transmission electron microscopy (dCirl Rescue : n = 9 ChOs from 6 larvae; dCirl KO : n = 15 ChOs from 8 larvae; Figure 1-figure supplement 2). Thus, all ChO cell types and their serial interconnections are present in the mutant demonstrating that removal of dCirl does not interrupt the complex architecture and cytology of the larval lch5. This corroborates earlier findings, based on fluorescence microscopy of molecular markers (Scholz et al., 2015), that dCirl is not involved in the structural specialization of ChOs.

Optogenetic stimulation of chordotonal neurons bypasses dCIRLdependence
Two qualitatively different forms of electrical activity mediate signal transduction and transformation in primary sensory neurons, such as the bipolar nerve cells of ChOs. During transduction, stimulus encounter by sensory receptors is converted into current flow through ion channels to generate the receptor potential. This membrane depolarization is then transformed into a train of action potentials by voltage-gated ion channels to carry the sensory signal along the axon. dCIRL increases the mechanically-induced firing frequency of ChO neurons (Scholz et al., 2015). We reasoned that the light-gated cation channel Channelrhodopsin-2 (Nagel et al., 2003) [ChR2; retinal-bound channelopsin-2 (Chop2)] could be used to distinguish whether this effect was exerted at the level of mechanosensory transduction or transformation. Because ChOs are also thermoresponsive (Liu et al., 2003), this strategy necessitated an efficient ChR variant to limit the heat generated by the required light intensities. We therefore screened for a ChR2 version that combines high photostimulation efficiency (Dawydow et al., 2014) with good temporal precision. The D156H mutant displayed very high expression in Xenopus oocytes upon inspection by confocal microscopy (Figure 2a)     favorable kinetic properties, especially after short light pulses (10 ms: t off1 = 11 ± 1.2 ms SD, t off2 = 1.1 ± 0.13 s SD; Figure 2b), and over ten-fold larger photocurrents than the wildtype version (ChR2-wt; Figure 2c). We therefore named the ChR2 D156H variant ChR2-XXM (extra high expression and medium open state). Imaging, electrophysiological recordings and in vivo assays confirmed the utility of ChR2-XXM at the neuromuscular junction (NMJ; ok6-GAL4; Figure  To examine whether dCirl supports the initiation of action potentials in mechanosensory neurons, we recorded from the Ich5 axon bundle during photostimulation via ChR2-XXM. Photoinduced action current frequencies were indistinguishable in control and dCirl KO animals over the entire irradiance spectrum ( Figure 2g). Thus, by bypassing the receptor potential, this optogenetic approach demonstrates that dCIRL does not promote membrane excitability per se to help initiate and propagate action potentials in the sensory neuron.

Chordotonal organs sense temperature changes independently of dCIRL
Because ChOs respond to temperature changes (Liu et al., 2003) we tested whether dCIRL also processes this non-mechanical stimulus. Action current frequencies in lch5 afferents gradually increased with rising temperature, roughly doubling from 15˚C to 30˚C (Figure 3a Figure 3. dCIRL shapes mechanosensory signal transduction. (a) Recordings of wildtype lch5 action currents at 15˚C and 30˚C without and during mechanical vibration at 900 Hz applied to the cap cell. (b) Quantification of action current frequencies without (dashed line) and with (solid line) mechanical stimulation in control (black) and dCirl KO larvae (gray). Asterisk denotes p 0.05 comparing event frequency at 20˚C with a Student's t-test. Data are presented as mean ± SEM, n = 8 animals per genotype. (c) Current recordings from lch5 neurons during 900 Hz mechanical stimulation in the presence of TTX (average of 10 sweeps). The wildtype (black) receptor current displays phasic (yellow shaded area) and tonic (gray area) components, both of which are strongly reduced after removal of dCirl (gray). (d) Quantification of phasic and (e) tonic current amplitudes across a stimulation range from 100 to 1500 Hz. Data are presented as mean ± SEM, n = 8 per genotype. Asterisks denote comparisons of current amplitude with a Mann-Whitney U test (*p 0.05, **p 0.01). DOI: 10.7554/eLife.28360.008 20˚C and was partially compensated by low and high temperatures (Figure 3b). These findings demonstrate that dCIRL plays a mechano-specific role in this sensory organ.

dCIRL increases mechanically triggered receptor currents
Next, we blocked voltage-gated sodium channels with tetrodotoxin (TTX) to isolate mechanosensory receptor currents. As a result, the initiation of action potentials is prevented and isolated receptor currents can be assessed. Both phasic and tonic current components were strongly reduced in dCirl KO neurons (Figure 3c-e), providing direct evidence that dCIRL modulates the receptor potential evoked by mechanical stimulation.
We observed that a diminished yet graded receptor current profile persisted upon increasing vibrational cues even in the absence of dCirl. This feature further attests to the fact that dCIRL controls the sensitivity of mechanosensory neurons towards mechanostimulation rather than the neurons' principal ability to respond to mechanical challenge.

dCIRL NTF length determines mechanosensitivity of chordotonal neurons
Characteristic of aGPCRs, dCIRL possesses a long extracellular N-terminus with adhesive properties that anchors the receptor to the extracellular matrix or to opposed cell surfaces via cognate ligands. By applying mechanical tension to the ECD this setting may facilitate the reliable transmission of mechanical deformation to the receptor. We sought to test this hypothesis by relaxing dCIRL's extracellular region via gradual elongation of the ECD through the insertion of spacer elements. All transgenic constructs were expressed from the genomic dCirl locus (Figure 1-figure supplement 1) (Scholz et al., 2015) and a small Bungarotoxin binding site fused to a hemagglutinin tag (dCirl BBS:: HA ) served as an insertion site control. Action current frequencies of dCirl BBS::HA neurons were comparable to wildtype indicating that cassette insertion did not interfere with structure or expression of the receptor (Figure 4a,b). Elongating the ECD through an mRFP cassette (dCirl N-RFP ), which adds at least 2 nm, blunted the response at 900 Hz and a substantial length increase by the 3xCD4 spacer marked with poly-V5 tags (dCirl 3xCD4 ; Figure 4a,c), which adds approximately 20 nm, flattened the activity profile across the entire stimulation range (Figure 4b). We therefore hypothesize that ECD length and tensile properties may adjust dCIRL's response towards mechanical challenge ( Figure 4d).

Autoproteolytic processing is dispensable for dCIRL activity
All aGPCRs contain a juxtamembrane GPCR autoproteolysis inducing (GAIN) domain (Araç et al., 2012), which catalyzes receptor cleavage in N and C-terminal fragments (NTF, CTF) and maintains the two non-covalently affixed (Gray et al., 1996). This unusual property may be required for protein folding and trafficking  or to expose the receptor's tethered agonist (Stachel), which begins at the GPCR proteolysis site (GPS; Figure 5a) (Krasnoperov et al., 1997;Lin et al., 2004) and can potently stimulate receptor activity (Liebscher et al., 2014;Stoveken et al., 2015). To test this assumption, we abolished autoproteolytic activity of the GAIN domain in two sets of dCirl alleles by mutating the À2 (dCirl H>A ) or +1 (dCirl T>A ) position of the GPS (H À2 L À1 #T +1 ; Figure 5a,b) (Prö mel et al., 2012), notably the latter within the Stachel sequence. In the first set, the GPS mutations were inserted into the RFP-tagged receptor background (dCirl N-RFP/H>A , dCirl N-RFP/T>A ), and in the second set, the unmodified dCirl template was mutated (dCirl H>A , dCirl T>A ). We prepared protein extracts from dCirl N-RFP/H>A and dCirl N-RFP/T>A flies and immunoblotted against the RFP tag. Both mutant proteins were detected as a full-length band of ca. 218 kDa ( Figure 5b). In contrast, the 106 kDa band, which corresponds to the RFP-tagged dCIRL NTF, was not present ( Figure 5b). This shows that both GPS mutations abrogated the autoproteolytic activity of the dCIRL GAIN domain.
SIM images of immunostained mechanosensory neurons revealed that autoproteolysis is not required for membrane targeting of dCIRL to dendritic and ciliary compartments ( Figure 5c). Interestingly, however, mechanically-induced receptor currents (Figure 5d,e) were differently affected by the two mutations. Whereas dCirl H>A neurons displayed wildtype responses, the dCirl T>A mutant delivered a null phenotype. These results demonstrate that dCIRL activation in vivo depends on an intact tethered agonist, but that NTF-CTF disruption is dispensable.

Mechanostimulation of dCIRL decreases the cAMP concentration in mechanosensory neurons
To interrogate intracellular signaling by dCIRL we chose an optogenetic approach by utilizing the photoactivated adenylyl cyclase bPAC (Stierl et al., 2011) (iav-GAL4>UAS-bPAC). Photoinduced cAMP elevation in wildtype lch5 quenched neuronal activity to the level observed in dCirl KO mutants, while bPAC activation in the dCirl KO background did not further decrease action current frequencies  Protein extracts from animals (10 per genotype) were blotted and immunostained with an a-V5 antiserum specifically detecting the elongated NTF of dCIRL 3xCD4 (ca. 177 kDa) bestowed with poly-V5-tags (arrowhead). Consistent with previous results on the high efficiency of GAIN-mediated dCIRL autoproteolysis (Scholz et al., 2015), no full-length receptor was found. a-Tubulin staining was used as loading control (circle significantly (Figure 6a-c). Conversely, pharmacological inhibition of adenylyl cyclase activity specifically rescued dCirl KO neuron function (Figure 6d). These observations indicate that increased cAMP levels attenuate the mechanosensory response and suggest that dCIRL modulates neuronal activity by suppressing cAMP production. Next, we employed the FRET-based cAMP sensor Epac1-camps (Maiellaro et al., 2016;Nikolaev et al., 2004) to directly visualize neuronal cAMP dynamics during mechanical stimulation

Discussion
Here we demonstrate how a GPCR can specifically shape mechanotransduction in a sensory neuron in vivo. This study thus serves a two-fold purpose. It delineates pivotal steps in the activation paradigm of aGPCRs and sheds light on the contribution of metabotropic signals to the physiology of neuronal mechanosensation.  While there is ongoing discussion whether metabotropic pathways are suitable to sense physical or chemical stimuli with fast onset kinetics, due to the supposed inherent slowness of second messenger systems (Knecht et al., 2015;Wilson, 2013), our results demonstrate that the aGPCR dCIRL/Latrophilin is necessary for faithful mechanostimulus detection in the lch5 organ of Drosophila larvae. Here, dCIRL contributes to the correct setting of the neuron's mechanically-evoked receptor potential. This is in line with the location of the receptor, which is present in the dendritic membrane and the single cilium of ChO neurons, one of the few documentations of the subcellular location of an aGPCR in its natural environment. The dendritic and ciliary membranes harbor mechanosensitive Transient Receptor Potential (TRP) channels that elicit a receptor potential in the mechanosensory neuron by converting mechanical strain into ion flux (Cheng et al., 2010;Kim et al., 2003;Zhang et al., 2015). Moreover, two mechanosensitive TRP channel subunits, TRPN1/NompC and TRPV/Nanchung, interact genetically with dCirl (Scholz et al., 2015). The present study further specifies this relationship by showing that the extent of the mechanosensory receptor current is controlled by dCirl. This suggests that the activity of the aGPCR directly modulates ion flux through TRP channels, and highlights that metabotropic and ionotropic signals may cooperate during the rapid sensory processes that underlie primary mechanosensation.
The nature of this cooperation is yet unclear. Second messenger signals may alter force-response properties of ion channels through post-translational modifications to correct for the mechanical setting of sensory structures, e.g. stretch, shape or osmotic state of the neuron, before acute mechanical stimuli arrive. Indeed, there are precedents for such a direct interplay between GPCRs and channel proteins in olfactory (Connelly et al., 2015) and cardiovascular contexts (Chachisvilis et al., 2006;Mederos y Schnitzler et al., 2011;Zou et al., 2004).
ChOs are polymodal sensors that can also detect thermal stimuli (Liu et al., 2003). We show that dCIRL does not influence this thermosensory response (between 15˚C and 30˚C) emphasizing the mechano-specific role of this aGPCR. Replacing sensory input by optogenetic stimulation supports this conclusion, as ChR2-XXM evoked normal activity in dCirl KO larvae.
Turning to the molecular mechanisms of dCIRL activation, we show that the length of the extracellular tail instructs receptor activity. This observation is compatible with an extracellular engagement of the dCIRL NTF with cellular or matricellular protein(s) through its adhesion domains. Mammalian latrophilins were shown to interact with teneurins (Silva et al., 2011), FLRTs (O'Sullivan et al., 2014 and neurexins 1b and 2b  suggesting that the receptors are anchored to opposed cell surfaces through their ligands. However, FLRTs do not exist in Drosophila and an engagement of dCIRL with the other two candidate partners could not be detected to date (N.S. and T.L., unpublished observations) indicating that other interactors may engage and mechanically affix dCIRL. Our data support a model where the distance between ligand-receptor contact site and signaling 7TM unit determines the mechanical load onto the receptor protein and its subsequent signal output. This scenario bears similarity to the role of the cytoplasmic ankyrin repeats of NompC, which provide a mechanical tether to the cytoskeleton of mechanosensory cells, and are essential for proper mechanoactivation of this ionotropic sensor (Zhang et al., 2015).
aGPCR activation occurs by means of a tethered agonist (Stachel) (Liebscher et al., 2014;Monk et al., 2015;Stoveken et al., 2015), which encompasses the last b-strand of the GAIN domain. Structural concerns imply that after GAIN domain cleavage a substantial part of the Stachel remains enclosed within the GAIN domain and should thus be inaccessible to interactions with the 7TM domain (Araç et al., 2012;Prö mel et al., 2013). These considerations beg the question how the tethered agonist gets exposed to stimulate receptor activity, and how this process relates to the mechanosensitivity of aGPCRs. Two models account for the elusive link between these critical features Liebscher et al., 2013). Mechanical challenge to the receptor causes: (1) physical disruption of the heterodimer at the GPS thereby exposing the tethered agonist. In this scenario, GPS cleavage is absolutely essential for receptor activity; (2) Allosteric changes of the GAIN domain, e.g. through isomerization of the tethered agonist-7TM region, that allow for the engagement of the Stachel with the 7TM. In this situation, GPS cleavage and disruption of the NTF-CTF receptor heterodimer are not necessary for receptor activity. We found that autoproteolytic cleavage is not required for the perception and transduction of vibrational mechanical stimuli by dCIRL.
We further uncovered that the concomitant disruption of Stachel and autoproteolysis disables dCIRL's mechanosensory function in ChO neurons. Thus, the tethered agonist concept  pertains to aGPCRs in Drosophila. Notably, these findings also demonstrate that classical GPS mutations have similar biochemical but different physiological effects in vivo.
Finally, we interrogated intracellular signaling by dCIRL. In contrast to previously described Ga s coupling of rat and nematode latrophilins (Müller et al., 2015), the mechanosensory response of ChO neurons was decreased by optogenetic augmentation of adenylyl cylcase activity, and the mechanosensory deficit of dCirl KO mutants was rescued by pharmacological inhibition of adenylyl cyclase. FRET measurements also directly demonstrated that mechanical stimulation reduces the cAMP concentration in the sensory neurons, and that this mechano-metabotropic coupling depends on dCIRL. Thus, dCIRL converts a mechanosensory signal into a drop of cAMP levels. This suggests that the Drosophila latrophilin entertains a cascade that inhibits adenylyl cyclases or stimulates phosphodiesterases in ChO neurons, and that G-protein coupling pathways by latrophilin homologs may depend on species and/or cell type.
Members of the aGPCR family are associated with a vast range of physiological processes extending beyond canonical neuronal mechanosensation. For example, dysfunction of ADGRG1/GPR56 causes polymicrogyria (Piao et al., 2004), ADGRF5/GPR116 controls pulmonary surfactant production (Bridges et al., 2013), genetic lesions in many aGPCR loci are associated with a roster of cancer types (Kan et al., 2010;O'Hayre et al., 2013) and ADGRE2/EMR2 regulates mast cell degranulation (Boyden et al., 2016). Intriguingly, a point mutation in the GAIN domain of ADGRE2 sensitizes the receptor to mechanical stimuli in kindreds of patients suffering from vibratory urticaria. Our results now provide a basis to test the generality of the concept that aGPCRs are metabotropic mechanosensors also outside classical mechanosensory structures, and aid in understanding the contribution of ailing aGPCR signaling in diseased tissues.
pMN9: T>A GPS cleavage-deficient dCirl was created with QuikChange site-directed mutagenesis of pTL370 using primers mn_12F/13R containing the altered GPS sequence.
pMN9: T>A GPS cleavage-deficient dCirl was created with QuikChange site-directed mutagenesis of pTL370 using primers mn_12F/13R containing the altered GPS sequence.
pMN10: T>A GPS cleavage-deficient dCirl N-RFP containing the extracellular mRFP cassette was created with QuikChange site-directed mutagenesis of pMN4 using primers mn_12F/13R containing the altered GPS sequence.
pMN38: H>A GPS cleavage-deficient dCirl N-RFP containing the extracellular mRFP cassette was created with QuikChange site-directed mutagenesis of pMN4 using primers mn_38F/39R containing the altered GPS sequence. pMN44: H>A GPS cleavage-deficient dCirl was created with QuikChange site-directed mutagenesis of pTL370 using primers mn_38F/39R containing the altered GPS sequence.
pTL512: The cDNA of the dCirl E splice variant was amplified from EST clone RE25258 obtained from the Drosophila Genomics Resource Center using primers tl_508F/509R and cloned into pCR-BluntII-TOPO (Thermo Fisher Scientific). A 150 bp fragment encoding the signal peptide of human GPR56 and a HA-tag was amplified with primers tl_514F/515R from a template vector and inserted into the plasmid via ApaI/EcoRV generating pTL506. A 5.1 kb BglII/SpeI fragment was released from pTL506 and inserted into the pcDps backbone generating pTL512.
pTL518: A 0.2 kb fragment was amplified off pTL370 (Scholz et al., 2015) with primers tl_540F/ 549R, cut with EcoRV and inserted into the EcoRV site of pTL506 to complete the RBL domain coding region.
pTL535: A 0.15 kb fragment encoding the signal peptide of the mouse ADGRL1/LPHN1 receptor was amplified off pSP113 (Müller et al., 2015), cut with EcoRI and BglII and inserted into pTL526.
pTL536: A 2.2 kb SpeI/AfeI-fragment of pTL507 was ligated with a 6.3 kb SpeI/AfeI-fragment of pTL521. A 0.15 kb fragment, amplified from pSP113 with primers tl_550F/551R, was cut with EcoRI and BglII and inserted into the resultant plasmid.
pTL564: To generate the dCirl length sensor control construct, which includes a single Bungarotoxin binding site and hemagglutinin-tag in the RBL-HRM connecting region, a 3.5 kb MluI/PacI fragment was released from pTL555 (subclone of exons 3-6 of dCirl tagged with Bungarotoxin-HA-tag in pMCS5 backbone) and inserted into pTL393 (attB-flanked genomic dCirl wild-type construct).
pTL697: A 2.9 kb fragment was amplified off pTL526 using primers tl_730F/728R. A second 3.4 kb fragment was amplified off pSA3 using primers tl_729F/696R. Both fragments were fused using the Gibson cloning kit (NEB).
All QuikChange-based PCRs were performed with pfu polymerase (Agilent). All amplicons were validated by restriction analyses followed by sequencing of the entire amplified exonic region.

SIM
SIM images were recorded and processes with a commercial inverted SIM microscope (Zeiss Elyra) equipped with an oil-immersion objective (Plan-Apochromat 63x, NA 1.4 Oil Dic M27). Standard laser illumination at 488 nm, 561 nm and 642 nm was used for excitation of Alexa Fluor-488, Cy3 and Cy5-conjugated antibodies, respectively. Stacks of at least 5 planes were recorded with structured illumination from 5 rotational and 5 phase variations and processed with standard Elyra settings.

Transmission electron microscopy
Third instar larvae were dissected in ice-cold Ca 2+ -free HL3 (Stewart et al., 1994) and prepared for transmission electron microscopy essentially as previously described (Wagh et al., 2006;Wagner et al., 2015). Briefly, after dissection, the larval filets were fixed in 2.5% glutaraldehyde and 2.5% paraformaldehyde in either 0.  ) were carried out at 4˚C. Larvae were washed in 0.05 M CB and postfixed in 2% osmiumtetroxide in the same buffer for 1.5-2 hr followed by contrasting with 0.5% aqueous uranyl acetate (UA) overnight, washing in dH2O and dehydrating in ethanol. After dehydration, all preparations were transferred to Epon via propylene oxide as intermedium, flat embedded in Epon, ultrathin sectioned (~80 nm), and contrasted with uranyl acetate (UA) and lead citrate according to standard protocols. Ultrathin sections were analyzed using a LEO 912 AB transmission electron microscope (Zeiss). Both fixation protocols gave similar results, with slightly better ultrastructure preservation using Fix I. Digitally recorded electron micrographic images were composed and adjusted for brightness and contrast using Photoshop (Adobe).

Immunoblots
Fly heads were collected in standard radioimmunoprecipitation assay buffer (RIPA buffer; 150 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris [pH 8.0]) supplemented with protease inhibitor cocktail (1:1000; Sigma-Aldrich) and immediately frozen in liquid nitrogen. Next, heads were homogenized and supplemented with SDS-based protein buffer (Li-cor) and 2mercaptoethanol (Merck). Next, samples were centrifuged for 5 min at 13,000 rpm (4˚C), incubated for 10 min at 55˚C, subjected to electrophoresis on a 4-12% Tris-Glycin SDS gel (Invitrogen) and blotted onto 0.2 mm nitrocellulose membrane (AmershamProtran). The membrane was blocked for 1 hr using Odyssey Blocking buffer (Li-cor) diluted 1:8 with 1 x PBS. For dCIRL 3xCD4 detection ten fly heads of each genotype were collected and immediately frozen using liquid nitrogen. Subsequently, 20 ml 2% SDS was added and a glas stirrer was used to grind the heads before 8 ml of 4x Sample buffer (Li-cor) and 2 ml of 10% Triton X-100 was supplemented. Samples were cooked for 5 min at 95˚C and centrifuged for 15 min at 13,000 rpm at RT. Gel electrophoresis was done using 4-12% Tris Glycine gels (Invitrogen). Protein was blotted onto 0.2 mm nitrocellulose membrane (Li-cor), blocked for 1 hr using Odyssey Blocking buffer (Li-cor) diluted 1:1 with 1x PBS.

Chordotonal neurons
Electrophysiological measurements were essentially carried out as previously described (Scholz et al., 2015). In brief, activity of lch5 neurons was recorded from the axon bundle using a suction electrode coupled to an EPC 10 USB amplifier (HEKA Instruments) and analyzed in Clampfit 10.2 (Molecular Devices). Mechanical stimulation was applied through a piezo-actuated, fire-sealed glass electrode placed on the muscle covering the cap cells. Spontaneously active neurons were stimulated optogenetically or at the indicated sine wave frequencies (three cycles of 1 s stimulation preceded by 1 s rest for each frequency). Data were sampled at 10 kHz and a notch filter was used to remove the specific stimulation frequency from the current trace. Pharmacological inhibition of adenylyl cyclase activity followed a full series of mechanical stimulation. Preparations were then incubated for 10 min with 100 mM SQ22536 (Merck) to inhibit adenylyl cyclase activity (Gao and Raj, 2001) before applying a second set of mechanical stimulation.
In order to isolate receptor currents, 4 mM TTX was added to the bath to block action potentials. For each frequency, either ten (Figure 2j-l) or three stimulation cycles (Figure 3g,h) were applied (1 s stimulation preceded by 1 s rest). Traces were low-pass filtered at 30 Hz before measuring the amplitudes of phasic (peak response) and tonic current components (average of last 200 ms).
Genotypes were blinded for all electrophysiological recordings of ChOs.

Optogenetics in vivo Chordotonal neurons
Larvae expressing ChR2-XXM::tdTomato in mechanosensory neurons (iav-Gal4>UAS-chop2 XXM :: tdTomato; 100 mM retinal food supplementation) were placed in a petri dish (10 cm diameter, filled with 1% agar) and recorded under infrared illumination. In each set of experiments, seven larvae were analyzed for 30 s before and during illumination with blue LEDs (440 nm,~3 mW/mm 2 ). During light stimulation, the head swinging phase was defined as the time interval between repeated lateral movements of the anterior segment and two complete crawling sequences in forward direction.

NMJ
Light from a mercury lamp passed through a GFP excitation band-pass filter was used to photostimulate crawling larvae expressing tagged or untagged ChR2-XXM in motoneurons (ok6-Gal4 driver; 100 mM retinal food supplementation unless indicated otherwise). Measurements denote the time between light-induced immobilization and resumed movement (defined as anterior displacement of posterior end) during ongoing irradiation. Adult flies were transferred to a vertically positioned Petri dish (10 cm diameter) and stimulated with blue LEDs (440 nm) for 10 s. After 5 s, the dish was tapped and the immobilized individuals were counted.

FRET-based cAMP measurements
Ratiometric FRET imaging was performed using an upright epifluorescence microscope (Axio Observer, Zeiss) equipped with a water-immersion objective (63x, NA 1.1), a xenon lamp coupled to a monochromator (VisiView, VisiChrome), filters for CFP (436/20, 455LP dichroic) and YFP (500/20, 515LP dichroic) excitation, a beam splitter (DualView, Photometrics) with a 505LP dichroic mirror, emission filters for CFP (480/30) and YFP (535/40), and an electron-multiplied charge coupled device camera (Evolve 512, Photometrics). CFP and YFP images upon CFP excitation were captured every 5 s with 100 ms illumination time. FRET was monitored in real-time with the MetaFluor 5.0 software (Molecular Devices) as the ratio between YFP and CFP emission. The YFP emission was corrected for direct excitation of YFP at 436 nm and the bleedthrough of CFP emission into the YFP channel as previously described (Bö rner et al., 2011). Larval preparations expressing Epac1-camps in lch5 neurons (iav-GAL4>UAS-Epac1-camps) were imaged at RT and stimulated with FSK (0.5 or 1 mM) at the beginning of the experiment to accumulate cAMP and decrease the FRET signal to a plateau phase (low forskolin response). 0.5 mM and 1 mM FSK elicited the same amplitude of FRET changes and the results were pooled accordingly. The amplitude of the low forskolin response was calculated by averaging five data points immediately before the stimulation and at the plateau phase. The difference was expressed as a percentage of maximal FRET response, obtained by application of IBMX (100 mM) followed by additional forskolin stimulation (10 mM). Piezo-actuated stimulation was performed only during the plateau phase (10 sweeps of 3 Â 1 s 900 Hz stimulation separated by 1 s rest, 1 s inter-sweep interval).
The amplitude of the piezo-induced FRET change was calculated by averaging five data points immediately before and at the end of the mechanical stimulation block. The difference was expressed as a percentage of the low FSK response. Two quality criteria were used to assess cell health and failure to meet these resulted in exclusion of samples from further analysis: (1) stimulation with low FSK concentrations produced a FRET change and (2) did not saturate the sensor (i.e. subsequent stimulation with 10 mM FSK and 100 mM IBMX further decreased the FRET signal).

G protein coupling assays Peptide synthesis
Peptides were synthesized using standard Fmoc-chemistry on an automated peptide synthesizer MultiPep (Intavis AG). Final side chain deprotection and cleavage from the solid support was achieved using TFA, water and thioanisole (95:2.5:2.5 vol%). Peptides were subsequently purified to >95% purity by preparative RP-HPLC (Shimadzu LC-8) equipped with a 300 Â 25 mm PLRP-S column (Agilent). For both analytical and preparative use, the mobile phases were water or acetonitrile, respectively, each containing 0.1% TFA. Samples were eluted with a linear gradient of 5-90% acetonitrile in water: 30 min for analytical runs and 90 min for preparative runs. Peptide characterization by analytical HPLC (Agilent 1100) and MALDI-MS (Bruker Microflex) yielded the expected [M+H]+ mass peaks. Peptides were dissolved in DMSO to 100 mM and stored at 4˚C until use.

In vitro expression analysis and functional assays
For expression analyses and functional assays, transiently transfected COS-7 cells were used. COS-7 cells were cultivated in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum, 100 U/ml penicillin and 100 mg/ml streptomycin at 37˚C and 5% CO 2 in a humidified atmosphere. For enzyme-linked immunosorbent assays (ELISA) to determine cell surface expression, cells were split into 48-well plates (3.8 Â 10 4 cells/well), for total ELISA into 6-well plates (3 Â 10 5 cells/well) and for cAMP accumulation or IP assays into 96-well plates (2 Â 10 4 cells/well). After 24 hr cells were transfected with 0.5 mg/well receptor-encoding plasmid DNA for detecting cell surface expression, 1 mg/well for detecting total expression and 0.2 mg/well for analyzing response to peptides in functional assays using Lipofectamine 2000 (Invitrogen) according to manufacturer's protocol.
For an estimation of total and cell surface expression, receptors carrying an N-terminal HA were analyzed with a rat anti-HA-peroxidase antibody (Roche) in indirect cellular ELISA as described previously (Schö neberg et al., 1998).
To determine cAMP accumulation, COS-7 cells were washed 48 hr post transfection for 5 min with serum-and phenol red-free DMEM containing 1 mM IBMX. For analysis of agonistic peptides transfected cells were treated with 1 mM peptide in this cell medium.
Incubation was stopped by aspirating medium and lysing cells in LI buffer (PerkinElmer Life Sciences). Samples were frozen at À20˚C and thawed for detection of cAMP concentrations using the AlphaScreen cAMP assay kit (PerkinElmer Life Sciences) according to manufacturer's protocol and the Fusion AlphaScreen multilabel reader (PerkinElmer Life Sciences).
For IP accumulation assays, the IP-One HTRF assay kit (CisBio) was used according to manufac-turer´s protocol. In brief, transfected COS-7 cells were washed 48 hr post transfection with PBS and subsequently stimulated with 1 mM peptide in stimulation buffer (CisBio) for 30 min at 37˚C. Incubation was terminated by lysing cells in lysis buffer on ice for 10 min and subsequent freezing at À20˚C. Cell lysates were defrosted and subject to IP measurements in a 384-well format using the EnVision multilabel reader (PerkinElmer Life Sciences).

Statistics
Data were analyzed in Prism 5.0 (GraphPad). Group means were compared by two-tailed Student's t-test. Where the assumption of normality of the sample distribution was violated as indicated by the D'Agostino and Pearsons omnibus normality test, group means were compared by two-tailed Mann-Whitney U test. Where indicated in figures asterisks denote the level of significance: *p 0.05, **p 0.01, ***p 0.001.