Diverse Roles of Axonemal Dyneins in Drosophila Auditory Neuron Function and Mechanical Amplification in Hearing

Much like vertebrate hair cells, the chordotonal sensory neurons that mediate hearing in Drosophila are motile and amplify the mechanical input of the ear. Because the neurons bear mechanosensory primary cilia whose microtubule axonemes display dynein arms, we hypothesized that their motility is powered by dyneins. Here, we describe two axonemal dynein proteins that are required for Drosophila auditory neuron function, localize to their primary cilia, and differently contribute to mechanical amplification in hearing. Promoter fusions revealed that the two axonemal dynein genes Dmdnah3 (=CG17150) and Dmdnai2 (=CG6053) are expressed in chordotonal neurons, including the auditory ones in the fly’s ear. Null alleles of both dyneins equally abolished electrical auditory neuron responses, yet whereas mutations in Dmdnah3 facilitated mechanical amplification, amplification was abolished by mutations in Dmdnai2. Epistasis analysis revealed that Dmdnah3 acts downstream of Nan-Iav channels in controlling the amplificatory gain. Dmdnai2, in addition to being required for amplification, was essential for outer dynein arms in auditory neuron cilia. This establishes diverse roles of axonemal dyneins in Drosophila auditory neuron function and links auditory neuron motility to primary cilia and axonemal dyneins. Mutant defects in sperm competition suggest that both dyneins also function in sperm motility.

neurons of Johnston's organ in the antenna's proximal part 8 . Each neuron bears one primary cilium, and the neurons actively amplify antennal vibrations on a cycle-by-cycle basis, documenting their motile properties 4,9,10 . Biophysically, the source of this amplification in hearing might reside in (i) the interplay between force-gated ion channels and associated motor protein, (ii) the collective behavior of cells or motor proteins, or (iii) ionically driven conformational changes of force-gated ion channels or other proteins, without the involvement of ATP-consuming motors 11 . Modelling studies revealed that the first scenario, the interplay between force-gated channels and motors, might be realized in the Drosophila ear 10 , and amplification was shown to require the NOMPC (= TRPN1) transient receptor potential (TRP) channel 12,13 , which localizes to the tips of auditory neuron cilia 14,15 and is gated by force 16,17 . The gain of amplification is negatively controlled by the TRPV channel subunits Nan and Iav 12 , which form a heteromeric Nan-Iav channel complex in the proximal ciliary region that presents dynein arms 18,19 . The mere presence of these arms suggests that Drosophila auditory neurons might use axonemal dyneins to drive mechanical amplification, a possibility that seems supported by genetic evidence: both mechanical amplification and the dynein arms are disrupted by mutations in genes that are implicated in axonemal dynein arm assembly, including for example fd3f (Ref. 20), tilB (Refs 4,21), zmynd10 (Refs 22,23), dyx1c1 (Refs 22,24), and hmw (Ref. 25). Although these mutant defects suggest that axonemal dyneins might power mechanical amplification, genetic evidence demonstrating that this amplification involves axonemal dynein genes has hitherto not been reported 26 .
Axonemal dynein arms can be categorized into outer and inner ones that differ in their molecular composition: each arm represents a multi-protein complex that consists of several axonemal dynein heavy, intermediate, and light chain subunits 27 . The sequenced Drosophila genome includes eleven axonemal dynein heavy chain genes 28,29 , and twelve genes encoding WD-repeat, dynein intermediate chain proteins (Figs. S2,3). Some proteins of each type seem specific to the sperm flagellum: for instance, three of the axonemal heavy chain genes are on the Y chromosome and so are normally present only in males, and several subunits encoded on other chromosomes are expressed at high levels only in testis 30 . Here we describe two axonemal dynein subunits, an inner arm heavy chain and a WD-repeat intermediate chain, that are expressed in both males and females, in chordotonal sensory neurons including those of Johnston's organ. Using mutations of these genes, we tested whether and, if so, how axonemal dyneins and ciliary motility contribute to Drosophila auditory neuron function and mechanical amplification in fly hearing.

Results
DmDNAH3 is a monomeric dynein heavy chain. CG17150 is one of eleven axonemal dynein heavy chain genes in Drosophila 28,29 (Fig. S1). It falls into the IAD-3 (inner-arm dynein-3) subgroup, described as inner arm-associated and single-headed (monomeric), together with the Drosophila Dhc62B/CG15804 and Dhc36C/CG5526 and human DNAH3 and DNAH7 gene products 28,29 . Among the human dynein heavy chains, CG17150 is most similar (50% amino acid identity) to DNAH3, and we name the gene Dmdnah3 for this orthology. The IAD-3 heavy chain subgroup has several representatives in each ciliated species, but information about their function is scant. A mutation of Chlamydomonas that deletes most of an IAD-3 heavy chain (DHC9) lacks a dynein inner arm and shows reduced flagellar beat frequency, especially in more viscous media, suggesting that this motor is important for motility under load 31 . DmDNAI2 is an outer arm IC2 protein. CG6053 encodes a WD-repeat, dynein intermediate chain (IC) (Fig. S1), and is one of twelve dynein IC genes in Drosophila . When compared to the dynein ICs in Chlamydomonas 29 , CG6053 protein is most similar to ODA6 (36.7% amino acid identity), placing it with human DNAI2 (44.7% identity) in the IC2 subgroup (Figs. S2-4), and we name it DmDNAI2 for its human ortholog. In both Chlamydomonas flagella and human cilia, loss-of-function mutations in IC2 proteins prevent the assembly of the entire outer arm complex: mutant cilia lack outer arms and their constituent heavy chain proteins 32,33 , while partial ODA6 revertants have specific effects on flagellar beating 34 . Structural labeling locates the protein near the base of the outer dynein arm, where it controls the activity of both the outer and inner arm 35 . IC2 proteins are thus conserved across ciliated eukaryotes, and required both for outer dynein arm assembly and for normal motility of the assembled axoneme.
Two other Drosophila IC proteins, encoded by CG1571 and CG10859, also fall within the IC2 subclass, but are not as similar to their algal and human homologs as is CG6053/Dmdnai2. RNA-seq expression data for both genes 30 show high transcript levels in testis and little or no expression in females, suggesting that they function in the sperm flagellum. In contrast, CG6053 transcripts, though less abundant, are expressed and more broadly distributed in both males and females 30 , consistent with a function outside the male germline. and neurons were counterstained with the monoclonal anti-Futsch antibody 22c10 (Ref. 36). Labelling induced by the Dmdnah3 and Dmdnai2 enhancer/promoter regions was observed in Johnston's organ, the chordotonal auditory sensory organ in the fly's antenna (Fig. 1A). Within this organ, anti-GFP and 22c10 staining superimposed, documenting that virtually all its 500 sensory neurons express Dmdnah3 and Dmdnai2 (Fig. 1A). Apart from Johnston's organ neurons, expression of Dmdnah3 and Dmdnai2 was also observed in other chordotonal neurons, including those of the femoral chordotonal organ (FCO) in the fly's leg and those of the larval pentascolopidial organ (lch5) (Fig. 1B). No expression was seen in the central nervous system or ciliated chemoreceptors and mechanosensory bristle neurons whose cilia reportedly lack dynein arms 7 . Chordotonal sensory neurons thus seem to be the only Drosophila neurons that express Dmdnah3 and Dmdnai2.
To gain insights into the subcellular localization of axonemal dyneins, we generated transgenic flies carrying a UAS-Dmdnai2-YFP construct, in which DmDNAI2 is tagged with yellow fluorescent protein (YFP). Expression of UAS-Dmdnai2-YFP construct was targeted to Johnston's organ neurons using Dmdnai2-GAL4, and the subcellular localization of DmDNAI2-YFP protein was assessed after enhancing YFP fluorescence with an anti-GFP antibody that recognizes YFP. DmDNAI2-YFP fluorescence was observed in Johnston's organ neuron somata and dendrites (Fig. 1C). Within the dendrites, fluorescence signals extended into the mechanosensory cilia, where the signals were confined to the proximal ciliary region that harbors both Nan-Iav TRPV channels and dynein arms 6 . Counterstaining with an anti-Iav antibody confirmed that DmDNAI2-YFP co-localizes with Iav in the proximal ciliary region but is absent from the distal tips of the cilia that lack Iav protein and dynein arms 6 . Hence, within Drosophila auditory neuron cilia, at least DmDNAI2 occurs in that region that presents dynein arms.

Mutations in Dmdnah3 and Dmdnai2 affect auditory neuron function and mechanical amplification in the ear.
To determine whether axonemal dyneins are required for the motility of Johnston's organ neurons, we next tested for mutant alterations in mechanical amplification. Flies carrying Minos (Mi) transposon insertions in Dmdnah3 and Dmdnai2 were used as mutants, i.e. Mi{ET1}CG17150 MB05004 (hereafter named Dmdnah3 1 ) and Mi{ET1}CG6053 MB06262 (hereafter named Dmdnai2 1 ). PCR confirmed the genomic positions of the two Mi insertions (Fig. S1), and both Dmdnah3 1 and Dmdnai2 1 seemed to be null mutations as no transcripts were detected by RT-PCR (Fig. S5). To assess mechanical amplification, we exposed the flies to pure tones at the individual mechanical best frequencies of their antennae and simultaneously monitored the resulting antennal displacement and electrical compound action potentials of Johnston's organ neurons that were recorded extracellularly from the antennal nerve. Mechanical best frequencies of the antennae were deduced from power spectra of their mechanical free fluctuations in the absence of acoustic stimulation, and tone intensities were measured as the sound particle velocity at the position of the flies 12 .
In w 1118 genetic background controls, sound particle velocities exceeding ca. 50 μ m/s elicited robust electrical compound responses that, increasing sigmoidally with the sound intensity, reached maximum potential amplitudes of ca. 40 μ V ( Fig. 2A,B). These electrical sound responses of Johnston's organ neurons were virtually abolished in both Dmdnah3 1 and Dmdnai2 1 mutants, whereby the latter mutants retained some residual responses (maximum potential amplitudes around 2 μ V) to intense sound stimuli (particle velocities > ca. 5 mm/s) ( Fig. 2A,B).
The loss of sensitive electrical sound responses in Dmdnai2 1 mutants was associated with a loss of mechanical amplification: in control flies, antennal displacements showed the characteristic nonlinear intensity scaling ( Fig. 2A) that, arising from motile responses of Johnston's organ neurons 10,12 , amplified the ear's mechanical input with a gain of approximately 10 ( Fig. 2B). This mechanical amplification was completely abolished in Dmdnai2 1 mutants ( Fig. 2A), as witnessed by amplificatory gains of around one (Fig. 2B). Contrasting with this loss of motility, motility persisted and became excessive in Dmdnah3 1 mutants ( Fig. 2A), with amplification gains around 20 (Fig. 2B). Hence, in addition to affecting electrical auditory neuron responses, Dmdnai2 1 abolishes active amplification whereas Dmdnah3 1 facilitates this amplification, documenting that auditory neuron motility requires Dmdnai2 and must be negatively regulated by Dmdnah3.
To determine whether these mutant phenotypes arise from the Mi insertions, we generated precise excisions by mobilizing the respective Mi elements. Both Dmdnai2 ex1 and Dmdnah3 ex1 revertants displayed normal sound-evoked electrical and mechanical responses that resembled those of the controls ( Fig. 2A,B). A full restoration of normal sound-evoked responses was also observed in Dmdnah3 1 and Dmdnai2 1 mutants when we expressed respective genomic rescue fragments in the respective mutant backgrounds ( Fig. 2A,B). In Dmdnai2 1 mutants, normal hearing was also restored when we targeted the expression of UAS-Dmdnai2-YFP to chordotonal neurons via F-Gal4 ( Fig. 2A), which expresses Gal4 under the control of the nan enhancer/promoter 37 . Accordingly, the diverse auditory phenotypes observed in Dmdnah3 1 and Dmdnai2 1 mutants can all be ascribed to the mutations in the respective dynein genes.
Dmdnah3 regulates motility together with Nan-Iav TRPV channels. Mechanical hyper-amplification and virtual loss of electrical sound responses, as observed in Dmdnah3 1 mutants, also characterizes flies lacking Nan-Iav TRPV channels (Ref. 12,18). To assess whether Dmdnah3 and Nan-Iav might operate in the same signaling pathway, we generated double mutants carrying and normalized amplitude of the associated nerve response (bottom) as a function of the sound particle velocity of the tone. Tone frequencies were adjusted to match the mechanical best frequency of the antenna 12 , and lines indicate linear antennal mechanics. In w 1118 controls, the antenna's displacement displays a compressive nonlinearity that aligns with the dynamic range of the nerve response and arises from mechanical amplification by JO neurons 12 (data from N ≥ 5 flies/antennae per strain). (B) Corresponding sensitivity gain due to mechanical amplification, sound particle velocity thresholds of the nerve responses, and maximum amplitudes of the nerve responses extracted from the data in (A) (means ± SD). CS: Canton S wild-type flies. X: not accessible because the nerve response is entirely lost. *significant difference (p < 0.05) from w 1118 controls: (Kruskal-Wallis test followed by two-tailed Mann-Whitney U-tests with Bonferroni correction). Nerve responses were measured as compound action potentials (CAP). Mechanical amplification is enhanced in Dmdnah3 1 mutants, where the nerve response is largely abolished. Both the nerve response and normal amplification are restored by precise excision of the responsible Minos insertion and by a genomic rescue of Dmdnah 3 . Dmdnai 2 mutants lack both the antennal nonlinearity and the nerve response, which are restored by excising the respective Mi insertion, genomic rescue, and by targeting the expression of UAS-Dmdnai2-YFP to Johnston's organ neurons via the chordotonal neuron driver F-GAL4. along with nan dy5 -a nan null allele that abolishes both Iav and Nan proteins from auditory neuron cilia 18 . Dmdnah3 1 , nan dy5 double mutants entirely lacked sound-evoked electrical potentials, same as single Dmdnah3 1 and nan dy5 mutants (Fig. 3A). Mechanical amplification due to ciliary motility was excessive in the double mutants, with amplification gains of around 25 (Fig. 3B). This gain was substantially lower than that of single nan dy5 mutants (gains of around 50) but closely resembled that of single Dmdnah3 1 mutants (gains of around 20). This establishes an epistatic relation between DmDNAH3 and Nan-Iav, placing DmDNAH3 downstream of Nan-Iav in a regulatory pathway that controls the mechanical amplification gain.
We also generated double mutants carrying Dmdnai2 1 and iav 1 -an iav null allele that abolishes Nan and Iav from auditory neuron cilia 18 , same as nan dy5 . Like single Dmdnai2 1 and iav 1 mutants Dmdnai2 1 , iav 1 double mutants lacked sound-evoked nerve responses. Mechanical amplification was excessive in single iav 1 mutants, but entirely abolished in Dmdnai2 1 , iav 1 double mutants, as in single Dmdnai2 1 mutants (Fig. 3A,B). Hence, DmDNAI2 is also placed downstream of Nan-Iav in the regulatory pathway that controls amplification, analogous to the respective placement of the NOMPC TRP channel 12 , which also operates downstream of Nan-Iav in amplificatory gain control 12 and is essential for mechanical amplification 13 . Axonemal dynein arms in auditory neuron cilia require DmDNAI2. To test whether axonemal dyneins are required for the anatomical integrity of Johnston's organ neurons, we analyzed their cellular morphologies in Dmdnah3 1 and Dmdnai2 1 mutant flies. No gross morphological defects were seen when we visualized the neurons with an anti-horseradish peroxidase (HRP) antibody (Fig. S6A), and antibody staining against NOMPC and Iav documented a normal ciliary localization of these TRPs (Fig. S6B). Inspecting the ciliary axonemes with transmission electron microscopy revealed that the dynein arms in the proximal region of the cilium persist and seem structurally uncompromised in Dmdnah3 1 mutants, but that the outer arms are selectively lost in Dmdnai2 1 mutant flies (Fig. 3C). In the latter mutants, the outer dynein arms were restored upon precise excision of the respective Mi element (Fig. 3C), documenting that the outer dynein arms require DmDNAI2.
DmDNAH3 and DmDNAI2 impair sperm competition. In Drosophila, sperm and auditory chordotonal neurons are the only ciliated cells whose cilia generate motility with axonemal dynein arms. Driving UAS-EGFP via Dmdnah3-GAL4 or Dmdnai2-GAL4 revealed that both Dmdnah3 and Dmdnai2 are expressed in sperm (Fig. 4A), corroborating the reported presence of DmDNAH3 in sperm revealed by mass spectrometry 38 . We thus analyzed male fertility in Dmdnah3 1 and Dmdnai2 1 mutants. Homozygous Dmdnah3 1 mutants were behaviorally able to mate, but no offspring were recovered, and testis squashes consistently showed sperm that were not motile. Homozygous Dmdnai2 1 males, by contrast, if they achieved mating, produced as many males as heterozygous controls, and testis squashes from all males showed motile sperm. Offspring was also obtained from Dmdnah3 1 mutants when the Dmdnah3 1 mutation was uncovered by the deficiency Df(3L)BSC371, suggesting that the loss of male fertility and sperm motility in the homozygous mutants -but not their hearing defects (Fig. S7)-arise from a secondary mutation. To characterize contributions of DmDNAH3 and DmDNAI2 to the function of sperm, we performed sperm competition assays in which w 1118 females were first mated to w; FRT 40A, neo FRT G13, w+ males (P1) and then to balanced Dmdnah3 1 /TM3 or Dmdnai2 1 /TM6 flies and mutant Dmdnah3 1 /Df(3L) BSC371 or Dmdnai2 1 /Dmdnai2 1 males (P2). Scoring the offspring revealed that sperm from control males displaced 80-90% of P1 sperm, but sperm from Dmdnah3 1 /Df(3L)BSC371 males were almost completely unable to compete, and Dmdnah2 1 homozygotes showed a highly significant reduction in sperm competition (Fig. 4B). This extends the roles of axonemal dyneins in sperm competition 39 to DmDNAH3 and DmDNAI2 and shows that these two proteins, in addition to their auditory requirements, contribute to sperm function and motility.

Discussion
Much like vertebrate hair cells, Drosophila auditory neurons are motile and mechanically amplify the vibrations they transduce 4,10 . Whereas mechanical amplification by hair cells involves prestin molecules and, presumably, myosin motors 40,41 , amplification by the fly's auditory chordotonal neurons is now shown to involve dyneins: according to our results, axonemal dyneins are expressed in Drosophila chordotonal neurons and are required for their mechanosensory function and mechanical amplification in the ear. DmDNAH3 controls the mechanical amplification gain together with Nan-Iav TRPV channels, and DmDNAI2 seems to be an important constituent of outer dynein arms in auditory neuron primary cilia that is essential for mechanical amplification. That both DmDNAH3 and DmDNAI2 participate in ciliary motility is supported by mutant defects in sperm competition. This establishes the presence of specific axonemal dynein proteins in insect chordotonal neuron cilia and links mechanical amplification by Drosophila auditory neurons to primary cilium motility and axonemal dynein motor components.
Judging from the present and previous 10,13,16,17 findings, it now seems that mechanical amplification by Drosophila auditory neurons might arise from the interplay between the force-gating of NOMPC channels and associated active movements of axonemal dynein motors. If so, NOMPC and axonemal dyneins would have to signal along the cilium because they reside in distinct ciliary compartments 6,11 , an indirect interaction that might be mediated by the microtubule axoneme: both NOMPC and axonemal dyneins bind to microtubules [42][43][44][45] and can be activated by force 16,45 . Being mechanically coupled through the axoneme, gating movements of the channels thus might directly trigger motor movements, and vice versa, allowing for the fast channel-motor interactions that are required to explain the cycle-by-cycle amplification of vibrations in the Drosophila ear, which operates at frequencies above 100 Hz 10 . The above scenario seems consistent with an amplification model in which the interplay between motors and force-gated channels drives amplification 10 . According to this model, motor movements power amplification by promoting transducer adaptation, whereby the motors actively reclose the mechanotransduction channels when forcing is maintained 10 . Besides NOMPC, Nan-Iav has been surmised to mediate transduction in JO neurons 26,46 , yet recent evidence suggests that Nan-Iav cannot be a mechanotransduction channel because it is not mechanosensitive 19 . Correlates of NOMPC-dependent force-gating and channel adaptation have been observed in the fly's antennal mechanics 10,17,47 , suggesting that dyneins might drive amplification by promoting adaptation of NOMPC. We now found that these correlates are entirely abolished from the antennal mechanics of Dmdnai2 1 mutants (Fig. S8), documenting that forces no longer gate the channels, possibly because the channels can no longer adapt and, thus, maintain mechanosensitivity. Recent work has shown that NOMPC requires direct interactions with microtubules for mechanogating 44 . Future work will be needed to test whether NOMPC adaptation requires microtubule-bound dyneins.
The different effects on dynein arm morphology and amplification resulting from loss of the DmDNAI2 and DmDNAH3 proteins are consistent with the different functions of their homologs in other species. The IC2 intermediate chain protein encoded by DmDNAI2 is not a force-generating subunit, but IC2 proteins in Chlamydomonas and in humans also function in assembly of the outer arms 32,33,48,49 , which are required for flagellar and ciliary motility. In contrast, monomeric inner arm dynein proteins such as DmDNAH3 are not absolutely required for motility: Chlamydomonas mutants lacking one  ) were tested for their ability to displace sperm in previously mated females. Sperm displacement was measured as the proportion of offspring from the second male (P2/P1 + P2). Bars represent means, error bars represent SEM. Number of males measured for each genotype is shown in or above the bar. **p < 0.01; ****p < 0.0001 (Kruskal-Wallis test followed by two-tailed Mann-Whitney U tests). Data were included only if at least one offspring from the second male was recovered. This conservative approach was to eliminate any possible effect of the mutations on the ability of males to achieve copulation. The numbers of males that were thus not taken into consideration were 0 for Dmdnah3 1 /TM3 but 23 for Dmdnah3 1 /Df, and 1 for Dmdnai2 1 /TM6 male but 4 for Dmdnai2 1 /Dmdnai2 1 . such protein show reduced flagellar beating only under viscous load 31 , consistent with our finding that DmDNAH3 mutant sperm function defects are only revealed in competition with normal sperm flagella. The Drosophila motor that includes DmDNAH3 may modulate antennal motility in response to TRPV channel activity, mechanical loading, or both. Candidates for the force-generating proteins that drive auditory motility in Drosophila are the outer arm heavy chains encoded by CG9492 and Dhc93AB, which both have been detected in the fly's auditory organ 22 . Rigorously testing their roles in amplification must involve genetic rescue experiments as presented here for Dmdnai2 and Dmdnah3, and demonstrating that they drive amplification and ion channel adaptation will require targeted manipulations of their active motor properties 50,51 .

Nomenclature.
A convention to date has been to identify autosomal Drosophila dynein heavy chain (Dhc) genes by their chromosomal location. CG17150, however, is located in the same cytogenetic interval (64C) as the cytoplasmic dynein gene Dhc64C, so following convention in this case could cause confusion between these distinct dynein species.
Flies. w 1118 was used as genetic background control and Canton-S as wild-type strain. w 1118 ;

Promoter fusions and reporter constructs. Genomic DNA was extracted from WT-flies using
Qiagen DNeasy Blood and Tissue kit. To generate Dmdnai2-GAL4 promoter fusions, the putative promoter gene sequences were amplified using the primers 5′ -CGAATTCAAATCAAACCAGCTCTTGTAGTTACC -3′ (forward) and 5′ -CGGATCCGAGTTCTCGGTGAACACCACCT-3′ (reverse) and inserted into pPTGAL vector 53 . To generate UAS-Dmdnai2 rescue constructs, a coding sequence (cds) clone of CG6053 (IP13643, DGRC, Vienna) was used to amplify the CG6053 cDNA region using the 5′ -CCGAATTCAAATATTTTGCTAAGTTTCCGATTGAAATGGAA-3′ (forward) and 5′ -GATCTAGACAGCCTCCTCCGCATCCTCTAC-3′ (reverse) primers. The amplicon was inserted into an UAS-attP vector obtained from the Konrad Bassler lab at the University of Zurich. The same construct was tagged with YFP to generate the UAS-Dmdnai2-YFP reporters. To generate a genomic rescue construct for Dmdnai2, the BACPAC clone pBAC70G22 (P(acman) resource centre, http://www. pacmanfly.org/), which contains a 20 kb region spanning 11683101 to 11705106 of chromosome arm 3L that includes the complete genomic region of CG6053 (which spans from 11690438 to 11692363), was inserted into the w 1118 background. Microinjections were performed by BestGene Inc (http://www. thebestgene.com/). RNA extraction and cDNA preparation. RNA was extracted from whole flies using the ZR Tissue & Insect RNA Microprep Kit (Zymo Research, Irvine, CA) following the manufacturer's instructions, and cDNA was synthetized using the QuantiTect Reverse Transcription Kit (Qiagen, Venlo, Netherlands).
Sound responses and correlates of mechanotransduction. The methods to assess sound-evoked antennal displacements and nerve responses as well as mechanical correlates of force-gating and adaptation have been described 17,47,54 . To compare auditory sensitivities across flies, nerve response amplitudes were plotted against the sound particle velocity and the increasing slope was fitted with a Hill equation. The sound particle velocity corresponding to 10% of the maximum amplitude assumed by the Hill fit was defined as the threshold intensity 22 . Immunohistochemistry. The following primary antibodies were used: mouse anti-Futsch (22c10) For antibody staining, fly heads (Johnston's organ) were washed with 0.1% PBT (0.2% bovine serum albumin, 0.1% Triton-X in phosphate buffered saline) and fixed in 4% paraformaldehyde (PFA) for one hour. Heads were embedded in gelatin-albumin and sectioned at 40 to 50 μ m with a vibratome (Leica, Oberkochen, Germany). Expression in the pentascolopidial organ was examined in larval fillets that were fixed for one hour in 4% PFA.
Upon sample preparation, samples were washed three times for 15 minutes in 1% PBT and blocked in 0.25% bovine serum albumin (BSA) and 10% normal goat serum (NGS) for one hour at room Scientific RepoRts | 5:17085 | DOI: 10.1038/srep17085 temperature. Antibodies (1%) were added to the samples, and were incubated at 4 °C temperature overnight. The samples were washed three times for 5 minutes per wash in PBT and treated with anti-rabbit and -mouse fluorophore conjugated secondary antibodies at a 1:300 concentration in PBS-T for 3 hours at room temperature. After washing the samples three times for 15 minutes, they were mounted in DABCO (Sigma-Aldrich, St. Lewis, MI). Staining was visualized with a Leica SP8 confocal microscope, and images were analyzed with ImageJ.
Transmission electron microscopy. To maintain the structural integrity of the samples, we used high pressure freezing. A 200 μ m deep aluminium specimen carrier (Leica, 16770141 Type A) was filled with 20% Polyvinylpyrrolidone (Sigma-Aldrich) in fly ringer (70 mM NaCl, 5 mM KCl, 10 mM NaHCO 3 , 1,5 mM CaCl 2 (*2H2O), 4 mM MgCl 2 , 5 mM trehalose, 115 mM saccharose, 5 mM HEPES). Dissected antennae were transferred into the platelet, closed with a second specimen carrier (16770142 Type B) and immediately frozen using a Leica EM HPM 100. Freeze substitution was carried out in a Leica EM AFS at − 90 °C for 100 h in 0.1% tannic acid and another 40 h in 2% OsO 4 (each mass/volume (w/v) in dry acetone + 1% water) while slowly increasing temperature 55 . Antennae were then infiltrated with Durcupan (Fluka, 30% for 3 h, 70% for 3 h, 90% overnight, 2 × 100% for 3 h and thin embedded between two glass slides for 48 h at 70 °C. For microscopic examination, 70 nm sections were cut using a Reichert Ultracut E ultramicrotome and transferred onto Formvar-coated copper mesh grids (PLANO G2405C). Sections were post-stained for 40 min with 4% (w/v) uranyl acetate in water and for 2 min with lead citrate 56 . Micrographs were taken in a JEOL electron microscope (JEM 1011, JEOL, Eching, Germany) with a Gatan Orius 1200A camera (Gatan, Munich, Germany).

Sperm competition.
Females (w 1118 ) were crossed to w; FRT 40A,neo FRT G13 w+ males in groups of 10 pairs per vial. These first males were removed after 48 hrs. After an additional 48 hrs, the females were crossed individually to second males of the following four genotypes: Dmdnah3 1 /TM3 and their sibs Dmdnah3 1 /Df(3L)BSC371, Dmdnai2 1 /TM6C and their sibs Dmdnai2 1 /Dmdnai2 1 . After 60 hrs, the second males were discarded and the females were transferred to fresh vials for 5 days, and subsequently discarded as well. All offspring from both vials for each second male were scored for white (P2) or orange (P1) eyes. For each second male, sperm displacement was calculated as P2/(P1 + P2). Statistical significance was determined by Mann-Whitney U tests.