Molecular mechanisms that regulate cellular metabolism, development, differentiation, and survival are essentially shared by most animal species. Recent evidence suggests that the last common ancestor of vertebrates and invertebrates, the so-called urbilaterian, already possessed a centralized nervous system that contained the precursors of brain structures and neurosecretory organs of extant protostomes and deuterostomes (1, 2). In this respect, homologous structures have been identified between insect and mammalian brains, such as the mushroom bodies and the pallium or cortex (2), the central complex and the basal ganglia (3), the pars intercerebralis/pars lateralis/corpora cardiaca system and the hypothalamus–pituitary axis (4, 5), the corpora allata, and the adenohypophysis (anterior pituitary) (6). Moreover, genes implicated in human disease and the generation of social behaviors have well-conserved homologs in insects and other invertebrates (7–9).

Studies on the fruit fly, Drosophila melanogaster, have successfully contributed to the characterization of molecular pathways underlying human nervous system diseases, such as Rett syndrome, Parkinson’s disease, Alzheimer’s disease, and others (10–13). More recently, Drosophila has even been used to study the mechanistic basis of neuro-developmental diseases, such as fragile X syndrome (14, 15), autism spectrum disorders (ASDs) (16), and schizophrenia (17–19).

Drosophila may serve as a suitable organism to study the basis of these diseases since it performs elaborate social interactions, such as courtship (20, 21) and aggression with the establishment of social dominance (22, 23), uses intraspecific acoustic communication (24), establishes long-term memory in classical and operant learning paradigms (25), and performs sensory-motor tasks with great precision (26). Drosophila offers molecular and genetic tools to identify the functions of individual genes and proteins, their interaction partners within cellular/molecular pathways [recent summary (27)], and their impact on physiology and behavioral performance. Homologs of disease-related genes can be mutated globally or in particular tissues or cell types, and transgenic flies may express human genes with or without characteristically disease-related mutations. Whether, and how, such genetic alterations impact Drosophila social behavior needs to be assessed in a quantifiable manner.

Neuroligins (Nlgs) are postsynaptic adhesion molecules that typically associate with presynaptic neurexins to form bidirectional signaling complexes required for the correct formation, maturation, and functional adjustment of chemical synaptic connections between neurons (28–31). Additional neurexin-independent synaptic functions have also been reported [reviewed by Reissner et al. (32)]. In mammalian nervous systems, different Nlgs are differentially expressed at different types of synapses [reviewed in Ref. (33–35)]. Nlg1 is predominantly expressed at excitatory glutamatergic synapses, fostering the accumulation of postsynaptic density proteins and ionotropic and metabotropic glutamate receptors (36). Nlg2 is selectively expressed at inhibitory synapses, where it associates with gephyrin and recruits GABA or glycine receptors (37). Nlg3 and Nlg4 appear at both excitatory and inhibitory synapses with preferences of Nlg3 for GABAergic and Nlg4 for glycinergic synapses (38–42). Alterations in nlg genes were found in patients affected by ASDs (40–43) and mutations of the same genes caused autism-like phenotypes in rodent model organisms (35, 44). ASD represent neuro-developmental disorders that cause impairments in social interaction and communication accompanied by restricted and repetitive behaviors. Especially mutations in nlg3 and nlg4, mutations in genes encoding direct interaction partners of Nlgs, such as neurexins and shank, and alterations of other proteins involved in synaptic mechanisms are directly associated with ASD (44–46). Based on the differential expression of Nlgs at different types of synapses, it was hypothesized that ASD phenotypes may result from disturbed balance of excitatory and inhibitory synaptic transmission in brain regions controlling respective behaviors and functions (47). Supporting evidence for this hypothesis derived from both ASD patients [review by Dickinson et al. (48)] and studies on rodent models for ASD (46, 49–52).

Much like in mammals, insects present multiple neuroligin genes (38), whereby the four genes found in D. melanogaster (nlg1–nlg4) show differential expression at central and peripheral nervous synapses. None of these genes has a particular similarity to the four mammalian nlg genes and the designations (vertebrate nlg1–4 vs Drosophila nlg1–4) do not imply phylogenetic relatedness. Previous studies have shown that Dnlg2 is predominantly expressed by excitatory postsynapses (53), while Dnlg4 is abundant at inhibitory synapses (54). Immunostaining with a Dnlg2 antibody in the adult Drosophila brain shows that the protein is abundant in the mushroom body and the central complex (W. Xie, personal communication). Both these brain structures are involved in the control of diverse behaviors like short-term courtship memory, center avoidance, olfactory learning, sleep regulation, and spatial orientation (55–59). Antibody staining against Dnlg4 revealed high expression in the lateral clock neurons (LNvs), possibly explaining the abnormal sleep behavior found in dnlg4-mutant flies, as well as expression in the central complex (54). Earlier studies (16) revealed that deletion of the dnlg2 gene alters social behavior in Drosophila. While retaining intact sensory perception, Dnlg2-deficient flies display reduced social interactions with respect to male–female courtship and male–male agonistic behavior, produce altered acoustic communication signals, and often fail to terminate behavior upon context changes. In the present study, we subjected flies deficient in Dnlg2 (typically expressed at excitatory synapses) or Dnlg4 (typically expressed at inhibitory synapses) and wild-type D. melanogaster (wt) to a series of behavioral assays that assess their social interactions (courtship, aggression, group formation, reaction to conspecific songs) and, in addition, analyzed their acoustic communication signals. We also tested for motor and sensory defects by analyzing locomotion, open space avoidance, circadian activity, and the sound sensitivity or their hearing organs. Our results show that both Dnlg2 and Dnlg4 are implicated in Drosophila social behavior.

Heads of 50 Dnlg2-deficient and 50 Dnlg4-deficient flies were subjected to qPCR, revealing reduced levels of the respective dlng transcripts by 27 and 40% compared to wild type (see Table S1 in Supplementary Material), respectively. We subjected these flies to various tests to assess their behaviors. To identify general defects in mobility, we first monitored their spontaneous locomotion in a plain, circular arena. The locomotion trajectories were categorized using unsupervised k-means clustering as reported in Ref. (71, 72). The resulting movement categories were two “forward-sideways movements,” two “fast yaw turns,” and “resting” [see also Ref. (61)]. The distribution of these three categories showed no significant differences between wild type and the two mutant stains (Figure 2). Therefore, locomotion and its assembly from movement components seemed to be uncompromised by mutations in dlng2 and dlng4 and mutant defects in other behaviors are unlikely to result from general locomotion impairments.

The avoidance of central (open) areas of an arena is called centrophobia (73, 74). In Drosophila, impaired centrophobia has been associated with defects of the mushroom bodies (55). In contrast to wild-type flies that nearly exclusively circled around the edge of the arena, both dnlg-mutants often traversed the central part of the arena (Figure 3). Both mutant strains seemed to avoid the immediate vicinity of the walls, resulting in median radial positions that are significantly closer to the center than in wt (Figure 3B, two-sided Kolmogorov–Smirnov test p > 0.01). Dnlg4-deficient flies even displayed a weak tendency for preferred occupation of central regions.

A previous study demonstrated alterations of sleep in dnlg4^LL01874/Def-mutant flies (54). We, thus, analyzed activity patterns and found Dnlg4-deficient flies have longer, but fewer episodes of sleep compared to wild-type and Dnlg2-deficient flies (see Figures 4D,E). Dnlg2-deficient flies, however, seemed to show an opposite phenotype with more but shorter sleep episodes than the wild type. Wild-type and both Dnlg-deficient fly strains displayed activity peaks associated with, or slightly preceding, the dark-to-light and light-to-dark switches (Figure 4A). In dnlg4^LL01874/Def, these peaks lasted longer than in wild type and dnlg2^KO17-mutants (Figure 4A); their overall activity was significantly reduced compared to both other strains during light periods (Figure 4B) and compared to wt during dark periods (Figure 4C). Compared to wt, Dnlg2-deficient flies displayed a slight but not significant increase of center crossing activity during light phases (Figure 4B) and no difference during dark periods (Figure 4C). In summary, overall activity is reduced in dnlg4^LL01874/Def flies compared to wild-type flies (Fisher’s permutation test corrected with Benjamini–Hochberg fdr; p-value 0.00015) and dnlg2 (p = 0.00008), while dnlg2^KO17 flies show a tendency for increased activity that fails significance (p = 0.09102) at least during periods of light.

Previously, Dnlg2-deficient flies reportedly displayed abnormal social and mating behavior (16). In order to study altered attraction to conspecifics, we recorded the location data of 24 animals (per replicate) that were allowed to successively enter a circular arena from two opposite sides in 90-s intervals. To assess group formation, we used their positional data and subjected it to a hierarchical agglomerative clustering. While wild-type males remained single with a probability of 39%, this probability was significantly increased in Dnlg2-deficient flies (60%) and reduced to 24% in Dnlg4-deficient flies [Figure 5C; Fisher’s exact test p-value wt vs dnlg2^KO17 = 0.0065; wt vs dnlg4^LL01874/Defp = 0.0304; dnlg2^KO17 vs dnlg4^LL01874/Defp = 5.6103 × 10⁻⁷; N(wt) = 104, N(dnlg2^KO17) = 94, N(dnlg4^LL01874/Def) = 119]. Wild-type flies and dnlg2^KO17-mutants associated in groups of up to seven individuals whereas group size shifted to larger values in dnlg4^LL01874/Def-mutants with up to 15 animals per group [Figure 5A; differences in group sizes tested with a two-sided Kolmogorov–Smirnov test (wt vs dnlg2^KO17p = 0.0198; wt vs dnlg4^LL01874/Defp = 3.2577 × 10⁻⁴; dnlg2^KO17 vs dnlg4^LL01874/Defp = 2.5584 × 10⁻⁹; N(wt) = 104, N(dnlg2^KO17) = 94, N(dnlg4^LL01874/Def) = 119; p-values corrected after (70))]. Chaining behavior was excluded from this analysis, because this phenomenon was caused by an improper termination of courtship and not a direct effect on aggregation behavior. We also calculated the average distances of animals to their nearest neighbor within such groups. Compared to wild-type and dnlg2^KO17-mutant males that maintained similar interindividual distances, the distance was reduced in Dnlg4-deficient flies (Figure 5D). Hence, although interindividual distances are only changed in dnlg4^LL01874/Def-mutants, Dnlg2-deficient males have a lower tendency to form groups, while dnlg4^LL01874/Def-mutants show an increased tendency to aggregate (Figure 5B).

Courting D. melanogaster produce two types of songs, a sine song and a pulse song. These acoustic communication signals play important roles in driving female mating decisions. By comparing songs between the three strains, we found that mutations in both dnlg2^KO17 and dnlg4^LL01874/Def affect the songs. The major frequency component of sine songs was quite variable in wild type, ranging between 120 and 160 Hz (Figure 6B). While the dominant sine song frequency was slightly lowered in Dnlg2-deficient flies, Dnlg4-deficient flies produced songs with higher sine song frequencies (dnlg2^KO17 vs dnlg4^LL01874/Defp = 0.00625; Fisher’s permutation test, corrected by Benjamini–Hochberg fdr). Analysis of courtship pulse songs revealed no differences in amplitude and shape (number of oscillations per pulse) between wild-type males and the two dnlg-mutants, suggesting that the neuromuscular components that generate acoustic communication signals were not compromised by the mutations in Dnlgs. Interpulse intervals had median durations of 38 ms in wild types and 37 ms in dnlg4^LL01874/Def but were significantly longer in CSs of dnlg2^KO17-mutants (Figure 6C).

We, next, analyzed male-chaining behavior, whereby males follow each other with one extended wing (75). The probability of wild-type males to engage in male-directed courtship was generally low and remained below 5% even when the arena was filled with larger numbers of individuals (Figures 6D,E). Chaining, however, was entirely absent in Dnlg2-deficient flies and thereby significantly lower compared to wt (p-value > 0.001; Kolmogorv–Smirnov). In contrast, dnlg4-mutants formed courtship chains that included up to 17 animals, and the probability of individual flies to engage in chaining increased significantly (p-value > 0.001; Kolmogorv–Smirnov), reaching more than 60% as the number of animals in the arena was increasing (Figure 6D). In summary, absence of Dnlg2 and Dnlg4 not only affects male CSs, but also chain formation, a male-directed courtship behavior.

Hahn et al. (16) reported reduced social interactions in Dnlg2-deficient flies in competitive courtship assays where two males switched between male-directed agonistic and female-directed behavior. A recent study (76) further reported that AS promote aggression, while CSs inhibit aggressive interactions between Drosophila males. In order to test whether Dnlg2- and Dnlg4-deficient flies react appropriately to sound signals, we extended the competitive courtship paradigm by continuous stimulation with WN, CS, or aggression sounds. Prior to the experiments, we assayed hearing organ function in the flies, revealing that auditory mechanics and sound-induced auditory nerve response in both mutant strains resemble those of wild-type flies (Figure 6A). During stimulation with WN and CSs wt males spent more time courting the female than displaying aggression against the other male (see Figure 7). During stimulation with aggression sounds, the latter male-directed aggression was increased [Figure 7A; Fisher’s exact test corrected with Benjamini–Hochberg fdr; p-value wt(white noise) vs wt(aggression sounds) = 0.0237]. Neither Dnlg2- or Dnlg4-deficient males altered the frequencies of aggressive and courtship behavior during stimulation with aggressive sounds. Stimulation with CSs slightly increased wt courtship behavior, whereas the opposite effect was seen in Dnlg2- and Dnlg4-deficient males, which reduced their courtship significantly and increased aggression (wt vs dnlg2^KO17p = 0.0237; wt vs dnlg4^LL01874/Defp = 0.0237). Hence, both Dnlg2- and Dnlg4-deficient flies seem to fail to respond appropriately to sounds.

The presented behavioral data suggest that the trans-synaptic adhesion molecules Dnlg2 and Dnlg4 may play a prominent role in the neuronal regulation of Drosophila’s social interactions. dnlg2 and dnlg4 are both expressed at central nervous synapses, but Dnlg2 is also present at neuromuscular synapses (53). Of the other two Drosophila Nlgs, dnlg1 is expressed at neuromuscular postsynapses (77) and dnlg3 is expressed in neuromuscular junctions and the central nervous system (78). Similar to Nlgs in mammalian central nervous systems, Dnlg1, Dnlg2, and Dnlg3 seem important for synaptic maturation and functional maintenance (studied at the neuromuscular junction) rather than being crucial for synaptogenesis.

In order to assess the requirements of Nlgs for the proper functionality of neural circuits, we analyzed the effects of mutations in dlng2 and dlng4 on walking, hearing, sound production, and social behavior. Electrophysiological recordings from antennal auditory nerves detected no differences in auditory sensitivity between wild type and mutant flies. Intact chemosensation can also be assumed, since males of both mutants correctly addressed females with courtship and males with agonistic behaviors in competitive courtship assays. In Drosophila, sex recognition and the assessment of female reproductive state has been demonstrated to largely depend on the detection of sex- and state-specific surface hydrocarbons (79, 80). In addition, dnlg2^KO17- and dnlg4^LL01874/Def-mutants, like wild-type flies, maintained peaks of locomotor activity when lights were switched on or off [(16, 54), this study]. Nonetheless, the sleep rhythms of Dnlg2- and Dnlg4-deficient flies were altered in an opposing manner, with Dnlg4-deficient flies sleeping more often and shorter than wt, consistent with previous studies (54), and Dnlg2-deficient flies sleeping less often with longer duration (Figure 4). Notwithstanding seemingly normal sensory functions, both dnlg2^KO17- and dnlg4^LL01874/Def-mutants thus show opposing deficits in sleep.

Unlike other insect species, such as locusts and cockroaches, that contain both excitatory glutamatergic and inhibitory GABAergic motor neurons, Drosophila only possesses excitatory neuromuscular synapses (81). Synaptic expression of dnlgs 1, 2, and 3 and consequences for synaptic transmission resulting from the lack of these Nlgs at the Drosophila neuromuscular junction have been described in previous studies (53, 77, 78, 82). The dnlg2^KO17-mutants used in our experiments displayed no defects that could be attributed to compromised neuromuscular transmission. Their walking behavior included all typical movement components (saccades, straight translation, and intermittent rest) that were combined in a coordinated fashion resembling that of wild-type flies (61). Moreover, CS pulses contained similar numbers of oscillations of the extended wing in dnlg2^KO17-mutants and wild-type flies, indicating normal neuromuscular function. Altered repetition rates of courtship pulses and altered sine song frequencies in dnlg2^KO17 point to alterations of central nervous circuits that determine the rhythm of wing movements. Along this line, Clyne and Miesenböck (83) demonstrated that initiation of sine and pulse song is triggered by descending brain neurons, whose activity synchronizes the intrinsic activity of thoracic pattern generators to a faster central clock. Avoidance of open areas is regarded as a measure for anxiety in various animal models (84–87). Wild-type Drosophila exhibit an obvious centrophobism that critically depends on the functionality of the mushroom bodies (55). While total ablation of mushroom bodies reduced centrophobism, specific inactivation of mushroom body γ-lobes increased centrophobistic behavior (55). Both, dnlg-mutant strains used in this study showed reduced centrophobic behavior (Figure 3) and dnlg2 is expressed in mushroom bodies along with dlng4 (unpublished immunocytochemical data by W. Xie), in the central complex (54). Both mushroom bodies and central complex are involved in the regulation of higher order social behaviors (55–59), so expression patterns seem consistent with the observed behavioral defects.

During an unsupervised and data-derived group formation analysis, 20 mm emerged as the distance threshold for group interactions. This is nearly identical to the distance at which a conspecific fly is only detected by a single ommatidium in the complex eye of Drosophila (61, 88, 89). Even though courtship and potentially other social contexts depend also on other sensory modalities [e.g., olfaction (90)], vision seems to be the most accurate and direct sense to judge the interindividual distance. Therefore, it is not surprising that group formation might be limited by the visual acuity of Drosophila. Aggregation of individuals and the formation of groups is a basic feature of social interaction, as for example oviposition in female Drosophila is dependent on group size (91). Compared to wild-type males, dnlg2^KO17-mutants displayed a reduced tendency to form groups but those who accumulated in groups assumed similar minimal interindividual distances (Figure 5). In contrast, dnlg4^LL01874/Def-mutants had a lower tendency to remain single, formed larger groups, and assumed closer positions to other group members.

Furthermore, this closer interindividual distance might have led to an increased formation of courtship chains in Dnlg4-deficient flies (Figure 6). Dnlg2-deficient flies showed a significantly lower chance of male–male courtship, which might be caused by their larger interindividual distance (Figure 6). It has been shown that CSs stimulate the formation of courtship chains (92, 93) and especially the inter-pulse interval of CSs has been identified as a critical factor for species recognition and attractiveness (20, 94), an alteration in song production might also lead to the chaining phenotype. Since dnlg2^KO17-mutant males produced songs with significantly prolonged inter-pulse intervals (Figure 6C), complete absence of chaining behavior could also have been a consequence of less attractive songs. Thus, the chaining phenotypes might be epiphenomena of the altered interindividual distance and CS production.

Absence of Dnlg2 and Dnlg4 altered social interactions of Drosophila males, without causing obvious impairments of sensory functions and execution of movements [this study, (16)]. Deficiency of Dnlg2 and Dnlg4, which seem to be differentially expressed at excitatory and inhibitory synapses, induced opposing deviations from wild-type behaviors in some behavioral paradigms, such as sleep rhythm, male chaining, group size, and interindividual distance. Other behavioral paradigms, such as center avoidance and stimulation with wt CSs, revealed equally altered behavior in a non-opposing fashion. Thus, Dnlg2 and Dnlg4 may play different roles in the regulation of synaptic transmission within brain neuropils implicated in the social behavior of D. melanogaster.

Like fly Nlgs, mammalian ones are differentially expressed at different types of synapses, and deletion or overexpression of particular Nlgs resulted in altered proportions of excitatory vs inhibitory transmission in brain neuropils (28, 51). Both overrepresentation of excitation and overrepresentation of inhibition have been associated with ASD phenotypes (48) and have also been observed in mouse models of ASD (95–97). Targeted manipulation of Drosophila Dnlg2 and Dnlg4 functions in specific brain regions might help to identify the neural circuits that regulate social behaviors and to assess the role of Nlgs in the balance between neuronal excitation and inhibition.

All authors designed the study. KC, AH, RK, IG, NH, RH, and BG collected and analyzed the data, HG and BG designed the behavioral setups and data acquisition. RH and BG wrote the manuscript, and all authors edited and approved the manuscript. KC and AH contributed equally.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.