Thermoresponsive motor behavior is mediated by ring neuron circuits in the central complex of Drosophila

Insects are ectothermal animals that are constrained in their survival and reproduction by external temperature fluctuations which require either active avoidance of or movement towards a given heat source. In Drosophila, different thermoreceptors and neurons have been identified that mediate temperature sensation to maintain the animal’s thermal preference. However, less is known how thermosensory information is integrated to gate thermoresponsive motor behavior. Here we use transsynaptic tracing together with calcium imaging, electrophysiology and thermogenetic manipulations in freely moving Drosophila exposed to elevated temperature and identify different functions of ellipsoid body ring neurons, R1-R4, in thermoresponsive motor behavior. Our results show that warming of the external surroundings elicits calcium influx specifically in R2-R4 but not in R1, which evokes threshold-dependent neural activity in the outer layer ring neurons. In contrast to R2, R3 and R4d neurons, thermogenetic inactivation of R4m and R1 neurons expressing the temperature-sensitive mutant allele of dynamin, shibireTS, results in impaired thermoresponsive motor behavior at elevated 31 °C. trans-Tango mediated transsynaptic tracing together with physiological and behavioral analyses indicate that integrated sensory information of warming is registered by neural activity of R4m as input layer of the ellipsoid body ring neuropil and relayed on to R1 output neurons that gate an adaptive motor response. Together these findings imply that segregated activities of central complex ring neurons mediate sensory-motor transformation of external temperature changes and gate thermoresponsive motor behavior in Drosophila.


Results
Increasing temperature causes thermoresponsive motor behavior in Drosophila. To investigate the role of EB ring neurons in thermoresponsive motor behavior, we made use of earlier observations which showed that an increase of surround temperature from the preferred 24-31 °C evokes goal-directed locomotion 2 .
To quantify behavioral activity, we measured locomotion using an open-field assay, employing video-assisted motion tracking to record freely moving Drosophila 45 . Flies were recorded for 60 min in 35 mm diameter arenas either at 25 °C or 31 °C, with a short pulse of mechanical stimulation applied after 30 min ( Fig. 1D and methods) to assess their arousal 47 . Based on these recordings, we generated raster plots and trajectories for each individual fly and calculated their overall activity. We extracted parameters detailing motor behavior including the average walking speed, the number of initiated activity bouts (action initiation), the length of walking bouts and the response to sensory stimulation ( Supplementary Fig. S1). In addition, we measured the duration of pauses (interbout interval, IBI) and determined a Weibull distribution and its shape factor κ that calculates the distribution of IBIs over time as a measure for random walks or burstiness 45 (Supplementary Fig. S2).
In response to 31 °C surround temperature, control flies more than doubled their motor activity throughout the recording period (from 34 to 77%, Fig. 1E); they walked faster (from 6.2 to 10.4 mm/s) and initiated more (from 0.3 to 1.1 initiations/s) and longer (from 1.6 to 3.2 s) walking bouts ( Supplementary Fig. S1). Interboutintervals became shorter (IBI, from 3.2 to 0.9 s), with the cumulative distribution of IBIs fitted on a Weibull function revealing a higher shape factor κ (from 0.27 to 0. 45), demonstrating that IBIs were more randomly distributed ( Supplementary Fig. S2). In response to a mechanical stimulus, flies increased their walking speed in both 25 °C (2.4 mm/s) and 31 °C (1.7 mm/s) recording conditions, compared to before the stimulus. However, this response was less pronounced, but not significantly so at 31 °C, likely because flies were already walking faster before the stimulus was applied (Supplementary Fig. S1). Consistent with earlier reports 2 , these data Figure 1. Increasing temperature causes thermoresponsive motor behavior and Ca 2+ influx in R1-R4 ring neurons in Drosophila. (A) Cartoon of adult Drosophila brain showing central complex (CX) ground pattern (PB, protocerebral bridge; FB, fan-shaped body; EB, ellipsoid body; NO, noduli; LAL, lateral accessory lobes-MB, mushroom bodies are shown for orientation). (B) Two major types of projection neurons characterize the circuit architecture of the CX, columnar neurons (example shown, Ellipsoid body-Protocerebral bridge-Gall neuron, E-PG in green) and tangential neurons (example shown, EB ring neuron, R in red); CX neuropils depicted as in A, in addition to gall and bulb. (C) The EB forms a toroidal or ring-like neuropil that resembles a ' closed arch' . Projections from tangential R neurons divide it into at least 4 different layers (white numbers 1-4), R1-R4. Sensory associations from PB and FB converge onto EB modules that can be subdivided according to PB-FB input into segments of sensory space, spanning the left and right hemispheres. (D) Raster plots showing sequences of activity (black bar) and inactivity (white spaces in-between) of 138 wild-type flies recorded either at 25 °C (left panel surrounded by blue frame) or 31 °C (right panel, red frame); vertical dashed orange line indicates mechanical stimulation. Above, trajectories shown for 6 exemplary flies recorded for the first 10 min either at 25 °C (in blue) or 31 °C (in red). (E) Flies show a significant increase of activity in response to elevated temperature across the whole 60 min recording of the experiment, including the response to mechanical stimulation given after 30 min (orange dashed line). Solid line, mean; shaded area, SEM; box and whiskers, median, IQR, Tukey. (F,G) Fluorescence after photoconversion of R2/R4 (EB1-Gal4), R3/R4d (c232-Gal4) and R1 (c105-Gal4) EB ring neuropil and neurons expressing CaMPARI at 25 °C (G, top row) and 31 °C (bottom row) respectively and (F) quantification of the red/green ratio shows a Ca 2+ increase for R2/R4 and R3/R4d only. Scale bar, 50 µm, mean ± SD, n ≥ 6. **p < 0.01, ***p > 0.001, Mann-Whitney test for (E), one-way ANOVA with Sidak's post hoc test for (F). See also Supplementary Figs. S1 and S2. www.nature.com/scientificreports/ demonstrate that an increase in environmental temperature to 31 °C evokes thermoresponsive motor behavior in freely moving Drosophila.
Increasing temperature causes ring neuron subtype-specific calcium influx. Next, we investigated whether EB ring neurons are involved in processing integrated temperature information. Since none of the known thermal receptors are expressed in the EB 5,6 , we first asked whether integrated temperature information is received and processed by EB ring neurons. For this, we used EB1-, c232-and c105-Gal4 drivers to express UAS-transgenes for functional imaging and thermogenetic manipulations targeted to R2/R4, R3/R4d and R1 neurons, respectively. We tested for a temperature-related functional response and utilised the fluorescent ratiometric calcium (Ca 2+ ) sensor CaMPARI 48 . For this, animals were kept at either 25 °C or 31 °C for 30 min prior to photoconversion (see methods). The resulting red-to-green ratio showed different baselines at 25 °C for the different EB ring neuron subtypes, probably caused by different amounts of free intracellular Ca 2+ indicating different neuronal activity (Fig. 1F,G). The lowest ratio was recorded for R2/R4 neuron-specific EB1 > CaMPARI with 16%, while the higher values for R1 neuron-specific c105 > CaMPARI (42%) and R3/R4d neuron-specific c232 > CaMPARI (55%) suggest higher amounts of free intracellular Ca 2+ in these structures caused by more active neurons. Photoconversion after 30 min at 31 °C showed a significant 38% increase in R2/R4 neurons when compared to 25 °C, followed by a 20% increase in R3/R4d, whereas R1 neurons only showed a 2.5% increase suggesting they did not respond to elevated temperature (Fig. 1F). These results demonstrate that an increase in surround temperature to 31 °C evokes a strong Ca 2+ response in EB1-specific R2/R4 neurons and a somewhat weaker response in R3/R4d neurons, but no response in R1 neurons. These data suggest that EB ring neurons process temperature information in a graded, layer-specific outside-in response. In comparison to genetic controls (Gal4/+ and UAS/+), recordings at 31 °C revealed a reduction in activity of R2/R4-EB1 > shibire TS flies to 49%, close to levels seen at 25 °C ( Fig. 2A and see Fig. 1E). These flies also walked slower (8.0 mm/s), initiated fewer (0.5 starts/s) and shorter (1.8 s) walking bouts at longer intervals (2.1 s) with a reduced shape factor κ (0.22) while the response to stimulation was only slightly reduced (1.9 mm/s), as shown in Supplementary Figs. S1 and S2. We then tested R3/R4d ring neurons for their involvement in thermoresponsive motor behavior. However, c232-Gal4 mediated expression of UAS-shibire TS did not affect the overall activity levels of R3/R4d-c232 > shibire TS flies recorded at 31 °C (Fig. 2B), except for a slightly decreased walking speed (8.7 mm/s) and a decreased shape factor κ (0.26) (Supplementary Figs. S1 and S2). In contrast, analysis of the temperature-induced locomotor behavior of R1-c105 > shibire TS flies recorded at 31 °C (Fig. 2C) revealed a significant reduction in activity to 57%, a decreased walking speed (6.8 mm/s), fewer initiations (1.1 s −1 ), shorter (1.3 s) and more spaced (0.9 s) activity bouts with a smaller κ (0.27) but unaffected startle response (2.4 mm/s) (Supplementary Figs. S1 and S2). Together these data demonstrate that R2/R4 as well as R1, but not R3/R4d ring neurons mediate thermoresponsive motor behavior. R2/R4 ring neurons integrate increasing temperature cues. Ring neuron-specific Ca 2+ influx has been used as a proxy to measure their neuronal activity in response to an external sensory cue 35 . Thus, we carried out ex vivo activity-related GCaMP Ca 2+ -imaging 54 of ring neurons in response to a graded temperature ramp from 20 to 35 °C and back to 20 °C (Fig. 2D-F and Supplementary Fig. S3). Similar to the observations with the fluorescent ratiometric Ca 2+ sensor CaMPARI, GCaMP imaging revealed a response for R2/ R4-EB1 > GCaMP6f and for R3/R4d-c232 > GCaMP6f only. R3/R4d neurons responded like R2/R4 neurons to increasing temperature but reached peak fluorescence at the maximum imposed 35 °C when signals of R2/R4 neurons were already decreasing ( Supplementary Fig. S3). With 23%, the average peak change of R3/R4d was merely half of that for R2/R4 (49%). In contrast, R1-c105 > GCaMP6f did not respond to temperature changes (0%, Supplementary Fig. S3), consistent with the CaMPARI data (see Fig. 1F). Thus, R2/R4 neurons quickly and strongly respond to an increase in surround temperature, with R3/R4d neurons responding less strongly and later, while R1 neurons do not respond.

R2
Together these findings suggest that segregated ring neuron activity mediates sensory-motor transformation of integrated temperature cues to gate thermoresponsive motor behavior. To gain further insights into the role of ring neuron subtypes in this process, we carried out GCaMP imaging in response to increasing temperature, while simultaneously expressing either UAS-shibire TS to block synaptic output or UAS-TrpA1 to activate neurons 7,55 .
Increasing temperature evokes neural activity in R2/R4 ring neurons. In order to further examine the kinetics of the temperature response, we studied the response of R2/R4-EB1 > GCaMP6f brains in more detail (Fig. 3A). The GCaMP signal of an exemplary brain revealed a saccade of small peaks starting around 23 °C, indicative of moderate activation of individual neurons, which by > 30 °C coalesced in a fluorescence spike in both R2/R4 cell somas and ring neuropil. Spiking occurred across the whole population with individual EB1 > GCaMP6f labelled neurons responding at slightly different temperatures and with varying intensity above 27 °C ( Fig. 3B and Supplementary Video SV1). GCaMP fluorescence intensity increased 4-to 7-fold above baseline for ring neuropil and cell bodies, respectively, and rapidly declined over 2 min, followed by inactivity even at noxious 35 °C (Fig. 3A). www.nature.com/scientificreports/ To corroborate these findings and to gain further insights into the physiological response of R2/R4 neurons, we carried out ex vivo electrophysiological whole-cell patch recordings of EB1 > mCD8::GFP-labelled R2/R4 neurons. This revealed resting membrane potentials of − 52.9 ± 4.0 mV (mean ± standard deviation, SD) and input resistances of 1185 ± 407 MΩ (n = 20). Injection of increasing amounts of depolarising current (up to + 40 pA) evoked repetitive firing with successively more spikes triggered by higher currents, up to a maximum frequency of 80 Hz (Fig. 3C), suggesting that spike frequency increased proportional to stimulus strength. We then measured the response of EB1-specific R2/R4 neurons to a temperature increase from 20 to 30 °C. These recordings demonstrated a physiological response that mimicked the activity-related Ca 2+ -imaging for three out of four tested neurons (Fig. 3D, compare with Fig. 3A). Detailed observation of an exemplary EB1-targeted ring neuron showed it was silent at rest but responded to increasing temperature, first by a few individual spikes corresponding to the small Ca 2+ peaks seen at lower temperature; second by more frequent spiking while the neuron depolarised, likely coinciding with a massive increase in intracellular Ca 2+ as seen by the corresponding GCaMP spike; third by a period of depolarised block where no individual spikes were generated; and finally by a return to resting potential without further spiking (Fig. 3D). Thus, both whole-cell recordings and Ca 2+ -imaging demonstrate a robust response of R2/R4 neurons to increasing temperature. R2/R4 ring neurons are inhibitory GABAergic. Despite the fact that both Ca 2+ -imaging and electrophysiological recordings of R2/R4 neurons evoked neural activity in a time and temperature-related manner, neither EB1 > GCaMP6f; shibire TS nor EB1 > GCaMP6f; TrpA1 ex vivo imaging showed Ca 2+ responses upon temperature increase. These observations were remarkable in two ways. First, a GCaMP response could be expected in EB1 > GCaMP6f; TrpA1 flies because thermogenetic activation of TrpA1 cation channels can lead to increased intracellular Ca 2+ 7,55 ; however, higher Ca 2+ concentrations have also been shown to rapidly inactivate TrpA1 56 . Second, activation of shibire TS impairs post-synaptic neurons 52 , however EB1 > shibire TS not only resulted in impaired behavioral output but also caused an abolished temperature-related GCaMP response in EB1 > GCaMP6f; shibire TS neurons themselves (Fig. 2D). These data suggest that R2/R4-EB1-Gal4 targets a group of functionally distinct neurons that are connected by reciprocal inhibition 57 , thereby affecting each other's activity 58,59 .
To investigate both possibilities, we first performed electrophysiological recordings to investigate whether R2/R4 neurons might be inhibitory GABAergic, as already suggested be earlier studies 20,60,61 . Following bathapplication of picrotoxin (250 µM), which is known to block GABA-mediated inhibition 62 , all of the six recorded neurons of EB1 > mCD8::GFP-labelled brains exhibited individual spikes followed by bursting activity that lasted between 10 and 40 s (Fig. 3E, n = 6). Bursts (0.5-10 Hz frequency) comprised up to 60 individual spikes, with an instantaneous spike frequency ranging from 30 to 100 Hz, followed by strong depolarisation which suppressed further activity. The pattern and time course of this picrotoxin response resembled the activity seen with Ca 2+ -imaging and electrophysiological recordings in response to increasing temperature (Fig. 3A,B,D), as well as for the response to increasing amounts of depolarising current (Fig. 3C).
We then examined the connectivity pattern of EB1-Gal4 targeted R2/R4 neurons by utilizing the trans-Tango technique based on anterograde transsynaptic tracing 63 . EB1 > trans-Tango flies identified R2/R4 ring neurons and layer-specific presynaptic projections to the EB ring neuropil (Fig. 4A, top in green and 4C). Postsynaptic trans-Tango labelling revealed connections to all ring neuron subtypes and neuropil layers (Fig. 4A, middle in magenta and 4B) which in many cases showed overlapping GFP and RFP immunolabelling (Fig. 4A, bottom). Of note, postsynaptic RFP-only patterns were detected for parts of the bulb (Fig. 4A, asterisks) as well as for columnar neurons and the protocerebral bridge (Fig. 4D), suggesting that R2/R4 activity reverberates with EB ring and columnar wedge neurons. These results suggest that R2/R4 neurons are inhibitory GABAergic interneurons that are connected to themselves and other R1-R4 neuron subtypes. Together with whole-cell recordings and Ca 2+ -imaging, these findings indicate that R2/R4 neurons assess sensory information, such as temperature increase, by threshold-dependent neuronal activity.

R4m ring neurons mediate temperature integration and thermoresponsive motor behavior.
Reciprocal inhibition between GABAergic R2/R4 neurons could explain the abolished GCaMP response in EB1 > GCaMP6f; shibire TS and EB1 > GCaMP6f; TrpA1 neurons at increased temperature, by way of inhibiting their own output either by disinhibition (with shibire TS ) or enhanced inhibition (with TrpA1). In addition, EB1-Gal4 may target a group of functionally distinct neurons, thereby affecting each other's activity 58,59 . To test this second possibility and to further disambiguate between R2 and R4 neurons, we made use of the Janelia Gal4 collection 64 and utilized two different Gal4 lines, R59B10 specific to R4m neurons 65 and R78B06 specific to R2 neurons 66 . To determine their role in sensory-motor transformation of increasing temperature, we performed behavioral experiments and GCaMP imaging together with trans-Tango labelling to investigate their connectivity pattern (Figs. 5, 6).
In comparison to genetic controls (Gal4/+ and UAS/+), recordings at 31 °C revealed a reduction in locomotor activity of R4m-R59B10 > shibire TS flies to 59% (Fig. 5A). These flies also walked slower (9.0 mm/s), showed fewer (1.3 initiations/s) and shorter walking bouts (1.7 s) at longer intervals (0.8 s) resulting in a smaller κ (0.37), however their response to a mechanical stimulus (− 1.2 mm/s) was largely unaltered (Supplementary Figs. S1 and S2). Ca 2+ -imaging revealed a 43% increased peak fluorescence temperature response for R4m similar in magnitude to EB1-targeted R2/R4 neurons, however, individual brain responses were short and peaking at different temperatures causing a wider spread of GCaMP peak ΔF/F0 fluorescence (Fig. 5B, compare to Fig. 2D and Supplementary Fig. S3). This peak fluorescence response was further reduced to 15% when coexpressing UAS-shibire TS . In contrast, UAS-TrpA1 expression resulted in an increase to 45% ΔF/F0 fluorescence in R59B10 > GCaMP6f; TrpA1 preparations over a wide range of temperature (Fig. 5B) www.nature.com/scientificreports/ identified R4m ring neurons and their layer-specific presynaptic projections to the EB ring neuropil (Fig. 5C, left), with postsynaptic trans-Tango labelling detectable in all ring neuron subtypes and neuropil layers (Fig. 5C, middle and right, and D, F), as well as in columnar wedge neurons and the protocerebral bridge (Fig. 5E).
In contrast to R59B10-specific R4m neurons, we did not detect any significant changes in Ca 2+ -imaging or thermoresponsive motor behavior when testing R2 neurons using R78B06-Gal4. Overall locomotor activity of R78B06 > shibire TS flies (Fig. 6A) was unaffected, however they walked slightly slower (5.4 mm/s) and for shorter periods (1.2 s) with a reduced response to mechanical stimulation (0.8 mm/s). Interestingly, GCaMP imaging of R2 neurons showed no temperature response (0% fluorescence change) and only a small 8% increase in R78B06 > GCaMP6f; TrpA1 preparations ( Fig. 6B and Supplementary Fig. S3). Of note, R78B06 > GCaMP6f; shibire TS flies could not be tested as they did not produce viable offspring. R78B06 > trans-Tango identified R2 ring neurons and their layer-specific presynaptic projections to the EB ring neuropil (Fig. 6C, left), with postsynaptic trans-Tango labelling detectable in R2 and R4m ring neuron subtypes and neuropil layers (Fig. 6C, middle and right, and D, F), as well as pre-and postsynaptic labelling in specific layers of the FB (Fig. 6E). Together these data demonstrate that R4m but not R2 neurons respond to increasing temperature and mediate thermoresponsive motor behavior in Drosophila.

Discussion
Our results implicate ring neuron circuits of the central complex (CX) as part of the regulatory network underlying thermoresponsive motor behavior in Drosophila. The presented findings suggest that specific EB ring neurons receive and integrate external temperature changes above 25 °C. Their subtype-specific functions indicate that segregated ring neuron circuit activities mediate sensory-motor transformation that gates adaptive motor output in Drosophila.

Temperature increase is registered by ring neuron sub-circuit activity. Previous studies showed
that temperature changes > 25 °C are detected by peripheral Hot Cell (HC) neurons in the arista which mediate rapid (< 1 min) warmth avoidance in adult flies 11,67 . Anterior Cell (AC) neurons in the adult brain mediate long term warmth avoidance by responding to smaller thermal gradients 7 . In both cases, however, it has remained elusive how these sensory cues of warming-related temperature changes elicit thermoresponsive movements to relocate to more ambient surroundings fitting the thermal preference of Drosophila 2,68 . In flies, the mushroom bodies and related dopaminergic projections of the Protocerebral Posterior Lateral (PPL1) cluster neurons have been implicated in thermosensory behaviors 14 www.nature.com/scientificreports/ ond and third order neurons relay thermosensory information from the arista to the so called lateral accessory calyx 71 . These data suggest a neural network for long-term warmth avoidance and conditioning 11,32,69,72 including the mushroom bodies that have been shown to negatively regulate locomotor activity 73 ; it is unknown, however, how these networks mediate sensory-motor transformation for thermoresponsive motor behavior. Our results demonstrate that ellipsoid body (EB) R2/R4 (EB1-Gal4) and R3/R4d (c232-Gal4) ring neurons respond to temperature changes of > 25 °C by calcium influx, which we detected by both CaMPARI and GCaMP imaging that was not seen for R1 neurons targeted by c105-Gal4. The response of R3/R4d ring neurons was temporarily delayed to that of R2/R4 ring neurons, suggesting that warming above 25 °C causes increased neural activity in these EB ring neuron subtypes, but not in R1. While it has been shown that R1 neurons are characterized by high tonic activity 74 , the lack of a significant calcium response to increased temperature suggests either such sensory input is not relayed onto R1 neurons or their high activity precludes detectable changes with calcium imaging. In contrast, our physiological analysis of EB1-targeted ring neurons indicates that increasing temperature evokes neural activity in a time and temperature-related manner. The further dissection of R2 and R4m neurons revealed that R4m neurons directly respond to warming, which is further supported by the fact that simultaneous expression of UAS-shibire TS largely abolished the temperature-related GCaMP signal. Furthermore, an increase in GCaMP signal was also observed for R3/R4d ring neurons, suggesting that R3-R4 neurons register temperature changes above 25 °C with enhanced neural activity.
Ring neurons, however, normally do not express warm-sensing thermoreceptors 7,[11][12][13]67,75 . AC neurons that sense warming of > 25 °C have been shown to project to the superior lateral protocerebrum, the antennal lobe and the subesophageal ganglion 7 . Anatomical studies revealed neuronal connections between the superior lateral protocerebrum and the central complex 76 , this raises the possibility that integrated temperature changes are relayed onto ring neurons via such a network. Moreover, recent studies identified connections between AC www.nature.com/scientificreports/ neurons and PDF-expressing neurons 16 that are themselves connected to EB ring neurons 77 . We used ex vivo preparations and Ca 2+ -imaging to show that R4m/d and R3 but not R1 neurons specifically respond to warming between 20 and 35 °C. These data together with known connectomics indicate that warm-sensing AC neurons relay temperature changes of > 25 °C via the superior lateral protocerebrum and/or PDF-expressing neurons to R4m/d and R3 neurons, the exact network of which was not revealed by previous studies 78 nor by the recent hemibrain connectome analysis 79 , and thus remains to be determined.

Segregated ring neuron activity mediates thermoresponsive motor behavior. The observed
warming-related calcium influx in R4m/d and R3 neurons suggests they are involved in sensory-motor transformation for thermoresponsive motor behavior. Surprisingly, however, only R4m neurons but not R2 or R4d and R3 neurons showed impaired thermoresponsive motor behavior at 31 °C when they expressed the temperature-sensitive mutant allele of dynamin. shibire TS has been shown to affect synaptic vesicle endocytosis 52 , thus acting on postsynaptic targets by inhibiting synaptic transmission which in turn can modify behavior 53 .
Our observations suggest that R2 as well as R3/R4d ring neurons are either not involved in thermoresponsive behavior or their connectivity network precludes gating of adaptive motor output. trans-Tango mediated transsynaptic tracing 63 identifies R3/R4d ring neurons that connect only onto themselves and to columnar wedge neurons ( Fig. 7A-C), and despite the fact that they show warming-related calcium influx, R4d and R3 neurons do not mediate thermoresponsive motor behavior (Fig. 2). Moreover, R2 neurons did not show a significant warming-related calcium influx, nor did expression of shibire TS at 31 °C impair thermoresponsive motor behavior of R78B06 > shibire TS flies (Fig. 6A,B). trans-Tango mediated transsynaptic tracing identifies R2 neurons that connect onto themselves and onto R4m neurons. However, in contrast to R2, R4m neurons showed warmingrelated calcium influx and impaired thermoresponsive motor behavior at 31 °C when simultaneously expressing shibire TS . Together these data indicate that R3/R4d neurons integrate warming-related temperature changes above 25 °C but themselves do not directly mediate thermoresponsive motor behavior. Instead, we observed a strong impairment of warming-related motor output in R1-c105 > shi TS flies that was also observed in R2/R4-EB1 > shi TS as well as in R4m-R59B10 > shi TS flies. trans-Tango analysis of these Gal4 lines revealed that R2/R4 ring neurons as well as R4m neurons connected onto themselves and to all other ring www.nature.com/scientificreports/ neuron layers, whereas R1 neurons are connected to themselves, R3 and the LAL (Fig. 7D-G), consistent with electron-microscopy (EM) based connectome data 79 . However, our data reveal that R3 neurons did not trigger impairment of warming-related motor output of c232 > shi TS flies at 31 °C, which was detected in R1-c105 > shi TS flies. Moreover, the fact that R1 neurons did not show alterations in Ca 2+ influx upon temperature changes above 25 °C, suggests that sensory-motor transformation is distributed among EB ring neurons. The dynamics of the observed calcium influx suggest R4m as the input layer where temperature changes are registered and reverberated with connected R3/R4d layers, as indicated by the temporal delay in Ca 2+ -imaging of c232 > GCaMP6f when compared to R2/R4-EB1 > GCamP6f or R4m-R59B10 > GCamP6f (Supplementary Fig. S3). Thus, Trans-Tango related connectivity patterns and shibire TS -triggered behavioral changes suggest that integrated sensory information of warming-related temperature changes is conveyed to R1 as the bona fide output layer gating thermoresponsive motor behavior in Drosophila (Fig. 8).
Ring neuron circuits are anatomical substrates for sensory-motor transformation. Theoretical considerations and CX-related neural network simulations hypothesized that the EB expresses a certain degree of directionality by specialized input and output layers 58,59 . Our findings further suggest specialized layers of the EB ring neuropil that express a concentric directionality from the outer rim as input layer to the R1 center that serves to gate parts of the lateral accessory lobe to control descending pathways. This hypothesis is also sup- www.nature.com/scientificreports/ ported by EM based connectome data which identify R1 neurons (e.g. ID1291099014) and their postsynaptic connections onto the LAL 79 . Our GCaMP imaging together with electrophysiological recordings revealed a saccade of small peaks starting around 23 °C which by > 30 °C coalesced into a large fluorescent spike (Fig. 3A,D), suggesting that temperature increase triggers ring neuron subtype-specific activity in a threshold-dependent manner. Based on these observations, we speculate that an all-or-nothing response mediates sensory-motor transformation in thermoresponsive motor behavior: Once warming rises above ~ 28 °C, spiking of outer rim layer neurons is relayed onto R1 neurons that gate a motor response. Given the recently identified ring neuron sublayers and shells 79 , it remains to be shown whether segregated ring neuron activity is further distributed into sublayer and/or shell activities. The observed ring neuron functions in sensory-motor transformation are not restricted to thermoresponsive motor behavior. A number of studies utilizing different behavioral paradigms but targeting subtypes similar to our study showed that ring neuron circuits are involved in a wide range of adaptive behaviors, from visual pattern 33 and spatial memory formation in a flight simulator 31,44 , to ethanol-related sensitivity 80 and ethanolinduced locomotion 28 , as well as to startle-induced arousal where flies responded to repeated air puffs 29 . Of note, outer layer ring neurons are known to be inhibitory GABAergic (Refs. 57,60,81 and Fig. 3) and R1 neurons receive direct GABAergic inhibition 74 . The resulting center-surround inhibition within the ring neuron network is a central feature of ring attractor dynamics underlying spatial navigation in Drosophila 82 . Such networks exert winner-takes-all functionality whereby reciprocal inhibition leads to the selection of only one of many signals 83 , which in turn can convey directed output to the next network level 59,84 . The recently identified inhibitory projections between EB and LA 74 could reverberate the selection process also during persistent stimulus input, such as elevated temperature. As a node of convergence situated several synapses downstream of sensory neurons and several synapses upstream of motor ganglia, the position and connectivity of the EB-LAL interface implies a direct role in the translation of sensory representations into motor representations. Indeed, experiments in locusts 85 and silk moths 86 identified the LAL as a premotor command centre projecting to descending neurons that innervate central pattern generators executing motor actions 40 . This conceptual framework is consistent with recent connectome data 79 and together with our findings indicate that ring neurons of the central complex integrate multiple sensory modalities, including warming-related temperature changes, to gate adaptive motor behavior in Drosophila.  CaMPARI. Response of EB ring neurons to elevated temperature in intact animals was measured using the fluorescent ratiometric calcium sensor CaMPARI 48 . Single flies were kept in either 25 °C or 31 °C for 30 min prior to the experiment. Flies were then quickly beheaded, their brains dissected in prewarmed Ca 2+ free external solution and placed in a preheated recording chamber. For photoconversion, 20 pulses of 5 s duration and with a 10 s interval were delivered to the whole brain with a 365 nm LED (3.96 mW/cm 2 ) using a 20 × lens. Images of the green (470 nm, 1.41 mW/cm 2 ) and red channel (555 nm, 2.55 mW/cm 2 ) were acquired immediately using an optiMOS camera (QImaging, Surrey, BC, Canada) with exposure times adjusted depending on the staining intensity ranging from 1 to 3 s and kept the same for both channels. A stack of 2.5 µm thick virtual slices encompassing the whole central complex including ring structure and cell bodies was used to produce a maximum intensity projection and the ratio between the two channels for a region of interest drawn around the ring was used for analysis (Zeiss ZEN and Fiji 2.1.0 88 software).

Methods
GCaMP. To test the temporal sequence of the Ca 2+ response to elevated temperature in the central complex we utilised GCaMP6f 54 . Whole brains were gradually heated from 20 to 35 °C and cooled back down to 20 °C and the calcium fluorescence signal obtained using a CCD camera (Zeiss Axiocam), a 470 nm LED light source (3.04 mW/cm 2 ) and a 20 × water immersion lens. Images were acquired at 4 fps with 15-50 ms exposure, recorded with ZEN (Zeiss) and plotted with Microsoft Excel. Baseline fluorescence (F 0 ) was taken as the mean fluorescence of the first 5 images before any temperature change. The change in fluorescence relative to baseline (ΔF/F 0 ) was recorded and the peak change used as a metric of the transient Ca 2+ increase.
Whole-cell recordings. Current clamp recordings were performed as described 87 using glass electrodes with 8-15 MΩ resistance filled with intracellular solution (in mM: 102 K-gluconate, 17 NaCl, 0.94 EGTA, 8.5 HEPES, 0.085 CaCl 2 , 1.7 MgCl 2 , pH 7.2) and an Axon MultiClamp 700B amplifier, digitised with an Axon DigiData 1440A (sampling rate: 20 kHz; filter: Bessel 10 kHz) and recorded using pClamp 10 (Molecular Devices, CA, USA). After cleaning the brain, a small incision was made over the position of the EB1 neurons in order to give easier access for the recording electrode. Brains were placed ventral side up in the temperature-controlled recording chamber, secured using a custom-made anchor and continuously perfused with aerated (95% O 2 , 5% CO 2 ) saline solution. Picrotoxin (Sigma, 250 µM in extracellular solution) was bath applied through the perfusion system. The liquid junction potential was calculated as 13 mV and subtracted from all membrane voltages. www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.