Ionotropic Receptor-dependent moist and dry cells control hygrosensation in Drosophila

Insects use hygrosensation (humidity sensing) to avoid desiccation and, in vectors such as mosquitoes, to locate vertebrate hosts. Sensory neurons activated by either dry or moist air (‘dry cells’ and ‘moist cells’) have been described in many insects, but their behavioral roles and the molecular basis of their hygrosensitivity remain unclear. We recently reported that Drosophila hygrosensation relies on three Ionotropic Receptors (IRs) required for dry cell function: IR25a, IR93a and IR40a (Knecht et al., 2016). Here, we discover Drosophila moist cells and show that they require IR25a and IR93a together with IR68a, a conserved, but orphan IR. Both IR68a- and IR40a-dependent pathways drive hygrosensory behavior: each is important for dry-seeking by hydrated flies and together they underlie moist-seeking by dehydrated flies. These studies reveal that humidity sensing in Drosophila, and likely other insects, involves the combined activity of two molecularly related but neuronally distinct hygrosensing systems. DOI: http://dx.doi.org/10.7554/eLife.26654.001


Introduction
Hygrosensation is a critical sensory modality for all animals, particularly for insects, whose small bodies and large surface area to volume ratios make dehydration a constant threat. Environmental humidity levels are therefore a key determinant of where and when a given species of insect will thrive (Chown et al., 2011). Hygrosensory cues are also fundamental for host-seeking by diseasetransmitting mosquitoes (Brown, 1966). Despite its importance, the molecular and cellular basis of hygrosensation in insects has remained unclear. In a recent eLife paper (Knecht et al., 2016)-in parallel with a related study (Enjin et al., 2016)-we reported that members of the Ionotropic Receptor (IR) family of variant ionotropic glutamate receptors (Benton et al., 2009;Rytz et al., 2013) are critical for hygrosensing in Drosophila melanogaster (Knecht et al., 2016). We identified a set of dry-air-activated neurons ('dry cells') in the sacculus, an internal sensory structure within the antenna, and demonstrated that their capability to sense changes in humidity, as well as the behavioral responses of flies in humidity gradients, require three highly conserved receptors: IR25a, IR93a and IR40a (Knecht et al., 2016).
Electrophysiological studies in larger insects suggest that hygrosensation also involves sensory neurons activated by high humidity levels ('moist cells') (Tichy and Gingl, 2001). In this study, we identify moist cells in the D. melanogaster sacculus and show that their hygrosensitivity requires IR25a and IR93a as well as a previously uncharacterized but evolutionarily conserved receptor, IR68a. We further show that hygrosensory behavior is driven by a combination of IR68a-dependent moist sensing and IR40a-dependent dry sensing in a manner that varies with the hydration state of the animal.

Results and discussion
An Ir68a reporter is expressed in candidate moist cells in the Drosophila sacculus To identify cells and receptors involved in moist sensing, we hypothesized that this modality, like dry sensing, involves a conserved IR for which no chemical ligand had been identified. IR68a was an excellent candidate, as this receptor has been conserved across~350 million years of insect evolution (Rytz et al., 2013). Moreover, previous RT-PCR studies (Croset et al., 2010) as well as transcriptomic analyses (Menuz et al., 2014;Shiao et al., 2013) detected Ir68a expression in the antenna.
As moist and dry cells are housed in the same sensilla in other species (Altner and Loftus, 1985), we anticipated that moist cells in Drosophila should be located in the antennal sacculus adjacent to the IR40a-expressing dry cells. While attempts to generate IR68a antisera were unsuccessful, we found that an Ir68a-Gal4 transgene, containing the putative Ir68a promoter, drives expression in a population of neurons that innervate chamber II of the sacculus (Figure 1a-b, Figure 1-figure supplement 1). Importantly, these Ir68a-Gal4-expressing cells are intermingled with, but distinct from Ir40a-expressing neurons (Figure 1c), consistent with a role as moist cells in the sensilla of chamber II.
IR40a-positive neurons express two co-receptors, IR25a and IR93a (Knecht et al., 2016). Immunostaining revealed that IR25a and IR93a proteins are also present in Ir68a-Gal4-expressing neurons (Figure 1d-e). The cell bodies of sacculus neurons show heterogeneous levels of IR25a and IR93a, with generally lower levels detected in Ir68a-Gal4-positive cells; this might reflect differences in the overall IR expression between cells or the efficiency of IR transport to sensory processes.
In the brain, Ir68a-Gal4-labeled neurons project to the antennal lobe, terminating in a discrete region near its ventrolateral edge ( Figure 1f). This region is distinct from the one innervated by IR40a-positive neurons (Figure 1g-j) and does not appear to correspond to any previously characterized glomerulus (Grabe et al., 2015;Münch and Galizia, 2016), consistent with a novel sensory function for these neurons.

Ir68a-Gal4 neurons are activated by moist air
To examine the physiological sensitivity of Ir68a-Gal4-expressing neurons to humidity changes, we performed calcium imaging in their axon termini using GCaMP6m (Figure 2a-b). We observed robust, non-adapting increases in GCaMP fluorescence upon switching from low to high humidity air (7% to 90% relative humidity [RH]), and robust, non-adapting decreases in fluorescence upon switching from high to low humidity air (Figure 2b-d). These humidity-dependent calcium responses are opposite of those of IR40a-expressing dry cells (Enjin et al., 2016;Knecht et al., 2016), indicating that Ir68a-Gal4 neurons correspond to moist cells.
Moist cell responses require IR68a, IR25a and IR93a, but not IR40a To examine the function of IR68a, we obtained Ir68a mutants (Figure 1-figure supplement 1). Calcium imaging in this mutant background revealed a complete loss of sensitivity of Ir68a neurons to humidity changes (Figure 2c-d), which was restored by a wild-type Ir68a transgene (Figure 2c-d, Figure 1-figure supplement 1). IR68a is therefore essential for hygrosensory transduction in moist cells.
We previously showed that IR25a and IR93a are required for hygrosensing by dry cells (Knecht et al., 2016). Consistent with the expression of these receptors in Ir68a-Gal4 neurons, moist cell responses were also eliminated in Ir25a and Ir93a mutants, and these defects could also be restored by the corresponding rescue constructs (Figure 2c-d). By contrast, IR40a, which is required for dry cell responses (Knecht et al., 2016), was dispensable for the responses of moist cells (Figure 2c-d). Similarly, loss of IR68a had no effect on dry cell responses to humidity changes (Figure 2e-f). Together these data define two distinct classes of hygrosensory neurons in the sacculus: IR68a-dependent moist cells and IR40a-dependent dry cells.

IR68a is required for hygrosensory behavior
The role of IR68a in behavioral responses to humidity differences was assessed by quantifying the distribution of flies in a humidity gradient (~67% to~96% RH) ( Figure 3a) (Knecht et al., 2016). Consistent with our previous report (Knecht et al., 2016), wild-type flies exhibit robust dry-seeking behavior, and loss-of-function mutations in Ir25a, Ir93a and Ir40a significantly reduce this response ( Figure 3b). Loss-of-function mutations in Ir68a caused a similar decrease in dry preference, a defect rescued by the introduction of a wild-type Ir68a transgene ( Figure 3b). Thus, IR68a is critical for behavioral responses to humidity. Notably, because IR40a-dependent dry cell responses persist in Ir68a mutants (Figure 2e-f), dry-sensing neurons appear to be insufficient to guide hygrotaxis, at least in this assay. Similarly, moist cell function alone is insufficient to support normal hygrotaxis, because Ir40a mutants display hygrosensory behavioral impairment (Figure 3b), despite having physiologically-active moist cells (Figure 2c-d). Flies lacking both Ir68a and Ir40a displayed defects similar to the single mutants (as well as Ir25a or Ir93a mutants in which both moist and dry pathways are eliminated) (Figure 3b). Together these results indicate that intact IR40a-and IR68a-dependent pathways are required for wild-type hygrotaxis.
Hydration state alters the impact of IR-dependent moist and dry sensing on behavior The behavioral requirement for both IR40a and IR68a raised the question of whether their combined activity was an obligate feature of the hygrosensory system or whether the function of a single pathway suffices under some conditions. Hydration state dramatically alters insect responses to humidity (Chown et al., 2011), and Drosophila prefer significantly moister environments when dehydrated (Perttunen and Erkkila, 1952). Consistent with those observations, animals previously subjected to desiccation stress became strongly moist-seeking in our assay (reflected in the shift of Dry Preference to negative values) (Figure 4a-b). Null mutations in either Ir25a or Ir93a (which disrupt both dry and moist cell functions) abolished this moist preference, indicating that dehydrated flies still rely on IR-dependent hygrosensing (Figure 4b). By contrast, moist-seeking behavior persisted in dehydrated Ir40a and Ir68a single mutants, although it was slightly reduced in Ir40a mutants ( Figure 4b). However, in Ir40a;Ir68a double mutant animals, moist-seeking was completely abolished (Figure 4b). Thus, while hydrated flies are dependent on both Ir40a-and Ir68a-dependent signaling, flies experiencing desiccation stress exhibit significant moist-seeking as long as one pathway is operative.
It is notable that the same IRs mediate hygrotaxis regardless of the hydration state of the fly, even though dehydration switches the valence of behavioral responses in a humidity gradient. These observations suggest that the moist and dry hygrosensory pathways do not simply promote attraction or aversion when activated. We propose that the information conveyed by these peripheral neurons is combined in the brain with signals from internal hydration sensors to determine how the  Figure 3. Ir68a is required for hygrosensory behavior. (a) Schematic of hygrotaxis assay.~67% to~96% RH gradients were generated as described (Knecht et al., 2016). Dry preference was quantified by counting flies on either side of chamber midline. 25-35 flies were used per assay. (b) Dry preference of hydrated flies. Asterisks denote values that are statistically distinct from wild type (**p<0.01, Steel with control). wild type (n = 12 assays). Ir25a 2 (n = 12). Ir93a MI05555 (n = 11). Ir40a 1 (n = 15). Ir68a c04139 (n = 16). Ir68a MB05565 (n = 14). Ir68a MB05565 + rescue (Ir68a + rescue transgene(II);Ir68a MB05565 ) (n = 12). Ir40a 1 ;Ir68a MB05565 (n = 15). Source data for summary graph are provided in Figure 3-source data 1. DOI: 10.7554/eLife.26654.006 The following source data is available for figure 3: Source data 1. Hygrotaxis behavior data. DOI: 10.7554/eLife.26654.007 animal responds to moisture. Internal osmolarity-sensing neurons that influence water consumption have recently been identified in Drosophila (Jourjine et al., 2016), and similar classes of internal sensors of water balance seem likely to be involved in setting the preference for moist versus dry environments. Future central mapping of the pathways that influence water-seeking behaviors will begin to reveal how the animal monitors and adjusts its hydration state to maintain optimal concentrations of this essential biological solvent. . Ir40a 1 (n = 12). Ir68a MB05565 (n = 11). Ir40a 1 ;Ir68a MB05565 (n = 12). Source data for summary graph are provided in Figure 4-source data 1. DOI: 10.7554/eLife.26654.008 The following source data is available for figure 4: Source data 1. Hygrotaxis behavior data. DOI: 10.7554/eLife.26654.009

Immunohistochemistry
Whole mount antennal stainings ( Figure 1b) were performed as previously described (Knecht et al., 2016). Immunostaining of antennal cryosections (Figure 1c-e) was performed largely as described (Saina and Benton, 2013). Male and female 1-to 3-week-old flies were mounted in OCT (Sakura #4583), 12 mm or 16 mm frozen sections were cut and then fixed in 4% paraformaldehyde at room temperature for 7-10 min. For brain stainings (Figure 1f-i), female flies were dissected in 1XPBS before either~2 min of fixation and immediate mounting, or 5 min of fixation and staining in primary antibody. All samples were mounted in Vectashield (Vector Labs) for confocal microscopy (Leica SP5 or Zeiss LSM710). The following antibodies were used: Figure

Calcium imaging
Calcium imaging was performed as described (Knecht et al., 2016). Data were processed largely as described (Knecht et al., 2016) but using a different custom Matlab script (source code file provided as Source code 1-Calcium Imaging). For analysis, the average pixel intensity of a nearby background region (BGR) was subtracted from the average pixel intensity of a polygon drawn around the labeled glomerulus (ROI). The first 20 frames (5 s) were used to define baseline fluorescence (F 0 ). F/ F 0 was calculated using F i /F 0 (frame i) = (ROI i -BGR i )/F 0 . As quantified imaging data did not conform to normal distribution (assessed by Shapiro-Wilk test, p<0.01), statistical comparisons were performed by Steel-Dwass test using JMP11 (SAS). Samples unresponsive to humidity change were subsequently depolarized by bath application of 0.7 mM KCl to confirm their physiological integrity; animals unresponsive to this positive control were excluded from analysis. Animals were also excluded if movement artifacts could not be corrected using Stackreg in ImageJ (Schneider et al., 2012).

Behavior
Hygrosensory behavior was assayed as previously described (Knecht et al., 2016). Desiccation prior to analysis was performed using a modification of (Lin et al., 2014). Flies were sorted and placed in tubes containing a strip of filter paper soaked in 3% sucrose and let to dry. The vial stopper was pushed down below the vial lip,~0.5 g Drierite spread over it, and Parafilm was placed over the top to seal the vial. Vials were kept in an incubator (25˚C, 70% RH) for 6 hr before hygrosensory behavior was assessed as described above. Humidity preference data in Figure 3 did not conform to normal distributions (assessed by Shapiro-Wilk test, p<0.01) and were analyzed by Steel with control test using JMP11 (SAS). Humidity preference data in Figure 4 did conform to normal distributions and were analyzed using Tukey HSD using JMP11 (SAS).