To characterize the osmosensitive response of aqp-8 expression more precisely, we performed qRT-PCR on animals raised on four different NaCl concentrations chronically applied. This experiment showed a linear and significant increase in aqp-8 mRNA levels in correlation to the applied salt concentration (Figure 1B). The results of a western blot analysis confirmed this mRNA results. Protein levels increased in the same degree as the mRNA concentration (Figure 1B, Supplementary Figure 1D). In the absence of a suitable AQP-8 antibody, we used the strain AKR-14 +/+; akaIs2 [K02G10.7(translational)::GFP + pCeh361]. The strain does express a AQP-8::GFP fusion protein under the control of the endogenous promotor for aqp-8. We measured the changes in the AQP-8::GFP fusion protein concentrations using a GFP antibody. The antibody was evaluated for this use and we were able to detect a single band with the predicted molecular weight of 54.45 kDa (Supplementary Figures 1A–C). In summary, we uncovered a linear correlation between aqp-8 mRNA and protein abundance with the surrounding osmolarity (Figure 1B).

To test whether the aqp-8 mRNA expression is specifically regulated by osmotic stress conditions, we assayed its expression in animals that were heat-shocked (1/2 h at 35°C) or underwent oxidative stress (1 h, 200 mM paraquat in NGM). However, only acute and chronic osmotic stress conditions significantly changed the abundance of aqp-8 mRNA (Figure 1C).

Loss of any of these three osm proteins in the cuticle causes an activation of the cuticular osmotic stress response pathway (Rohlfing et al., 2011). Based on a microarray, we had previously identified increased aqp-8 mRNA levels in the three osr mutants even under isotonic conditions (Rohlfing et al., 2010). We now confirmed this observation by qRT-PCR (Figure 1D). Similarly, AQP-8 protein levels were elevated in the osm-8 strain (Figure 1E). Again transgenic lines expressing AQP-8::GFP were used to measure AQP-8 protein levels using the established GFP antibody. Taken together, these results demonstrate that aqp-8 expression is specifically induced by the hyper-osmotic stress response and that activation can occur physiologically (hyper-osmotic stress) or constitutively (mutant).

Since aqp-8 has two predicted isoforms, K02H10.7a (1022 nt) and K02H10.7b (1134 nt; Supplementary Figure 2A), we also tested by qRT-PCR whether the truncated isoform a is present and affected by the osmotic stress response. However, agarose gel analyses of PCR amplified cDNA samples from wild type strain N2[Bristol] and osm-8(n1518) animals (Supplementary Figure 2B) revealed that only the full-length isoform b was present in our samples, Subsequent sensitive melting curve analysis confirmed this result (Supplementary Figure 2C). In coherence with the above described constitutive activation of aqp-8 in an osm-8(n1518) background, the amount of PCR product in the osm-8(n1518) sample is visibly increased compared to N2[BRISTOL] by equal amounts of cDNA used in the PCR reaction.

To test whether hypo-osmotic stress has an opposite effect on aqp-8 mRNA expression, we placed young adult wild type animals, reared at control conditions (50 mM NaCl in NGM), in deionized water (60 min treatment). In line with a role of aqp-8 regulation in osmotic stress response, acute exposure to hypo-osmotic conditions induced a significant decrease in aqp-8 mRNA levels within 1 h upon treatment (Figure 1F). This finding was consistent with the regulation of aqp-8 upon recovery from hyper-osmotic stress (Figure 1G). Within 3 h of recovery, aqp-8 mRNA levels had decreased to control levels. Hence, aqp-8 expression is negatively regulated under different hypo-osmotic stress conditions.

To investigate the effect of osmotic stress conditions on the localization of AQP-8 in the adult excretory cell, we characterized AQP-8::GFP localization using standard fluorescence microscope techniques. Under normal osmolarity conditions, AQP-8::GFP localized within the excretory cell soma and throughout the entire anterior excretory canal. Within the posterior excretory canal, the signal intensity decreased from the soma toward the tail region (Figure 2A), although the canal is fully extended throughout the whole body in adults (Khan et al., 2013; Kolotuev et al., 2013). In most animals, the signal was only visible along ~80% of the length of the posterior canal. Under acute hyper-osmotic stress conditions (8 h, 200 mM NaCl in NGM), the AQP-8::GFP signal expanded from the soma throughout the posterior canal (Figures 2A,B). Animals raised from L1 stage under chronic hyper-osmotic stress conditions (200 mM NaCl in NGM), showed a bright fluorescence signal within the entire excretory canal (Figure 2B). The body length was not significantly different between control and treated animals and therefore was not taken into account in our measurements (Supplementary Figure 3). In summary, the expansion of the AQP-8::GFP signal toward the tip of the canal was significantly increased by acute and chronic stress conditions (Figure 2B).

When larvae are exposed to hyper-osmotic stress conditions, pearl-string like varicosities have been reported to appear during excretory cell development (Khan et al., 2013; Kolotuev et al., 2013). These swellings represent extensions of the endoplasmic reticulum and Golgi apparatus into the posterior canal, which supplies vacuoles and canaliculi (Kolotuev et al., 2013). Similarly, we observed the appearance of AQP-8::GFP positive varicosities in the posterior canal (Figure 2A, asterisk). In our experiments, 33% of control animals developed one or rarely two varicosities within the posterior canal (Figure 2C). This number increased significantly to 77% (p = 0.0003) with one to four varicosities per canal when animals were exposed to acute osmotic stress conditions (8 h/200 mM NaCl). The number of varicosities in chronically treated animals (starting from L1) was lower but still significantly elevated (56%; p = 0.0113) compared to control conditions (Figure 2C). In an osm-8(n1518) background the amount of varicosities is also significantly increased (64%; p = 0.0007) under isotonic growth conditions (Figure 2C), which indicates an involvement of the osmotic stress response pathway in the formation of varicosities. Taken together, these results suggest that the excretory cell, when challenged with osmotic stress, uses its entire length and organelle supply for osmoregulation.

To further characterize how the osmotic stress response pathway and aqp-8 activation are linked, we functionally tested the involvement of glycerol synthesis in this regulation. gpdh-1 expression is activated by the transcription factor elt-2 within the intestine and, to a negligible degree, by elt-3 within the hypodermis (Rohlfing et al., 2011). gpdh-1 is the final target gene in the osmotic response pathway and produces the protective osmolyte glycerol (Figure 5; Rohlfing et al., 2011). Conversely, the patch-related receptor ptr-23 is an upstream suppressor of the signal cascade, possibly receiving direct input from an osmosensor in the cuticle (Figure 5; Rohlfing et al., 2011). An intermediate pathway has not been uncovered yet. We therefore asked, whether aqp-8 and gpdh-1 could partially share a common signaling pathway downstream of ptr-23, or whether the internal glycerol concentration itself could be a signal for aqp-8 expression.

To address this question, we correlated gpdh-1 expression and onset glycerol accumulation to aqp-8 activation. Under moderate osmotic stress condition (200 mM NaCl in NGM) applied at day 1 of adulthood, gpdh-1 mRNA levels were elevated within half an hour after the onset of the moderate osmotic stress reaching maximum levels after 2 h (Figure 3A). gpdh-1 and glycerol levels have been shown to be tightly linked, glycerol levels start to rise as soon as the enzyme is present (Lamitina et al., 2004). Under our test conditions the osmolyte glycerol should have started to accumulate between half an hour to one hour after the onset of the stress. The aqp-8 mRNA levels started to increase after a lag period of ~3 h. After 6 h, a significant increase in aqp-8 mRNA was observed compared to control conditions (50 mM NaCl; Figure 3A, Supplementary Figure 4). At the onset of aqp-8 mRNA expression, a significant amount of glycerol should be present. Therefore, glycerol accumulation could be a possible trigger for aqp-8 mRNA expression.

Next, we tested whether the inhibition of glycerol accumulation would affect aqp-8 mRNA expression. We used the gpdh-1 loss-of-function allele ok1558 and applied acute osmotic stress (6 h, 200 mM NaCl in NGM) at day 1 of adulthood. The loss of glycerol accumulation in the mutant did not impair with a significant activation of aqp-8 mRNA expression (Figure 3B). In correspondence to this results, the RNAi knock-down of the gpdh-1 transcription factor elt-2 did not affect the constitutive activation aqp-8 expression the osm-8(n1518) mutant (Supplementary Figure 5). Taken together, glycerol accumulation does not seem to be the trigger for aqp-8 expression.

To investigate the direct impact of ptr-23 on AQP-8 protein concentrations, we performed ptr-23 RNAi knock-down in wild type and osm-8(n1518) backgrounds. Western blot analysis revealed that the constitutive high AQP-8 protein concentration in osm-8(n1518) mutants is reduced to normal values upon loss of ptr-23 via RNAi knock-down (Figure 3C). However, the physiological-induced increase in AQP-8 protein concentration by hyper-osmotic stress in wild type animals was only partially affected by ptr-23 RNAi knock-down (Figure 3C). Hence, additional pathways, besides the ptr-23-mediated activation, could be present to induce of aqp-8 expression.

To identify ptr-23-independent pathways involved in AQP-8 regulation, we performed a genome-wide RNAi screen using the established Ahringer RNAi library on our AQP-8::GFP reporter strain AKR-14 grown under standard conditions (NGM, 50 mM NaCl). Changes in AQP-8::GFP transgene expression levels were monitored using a standard fluorescence stereomicroscope.

In total, we identified 687 genes with an effect on AQP-8::GFP expression out of 19,762 tested genes. Many of these genes affect basic cellular processes including ribosome function or metabolism. Surprisingly, 63 genes (9.17% of all hits) were connected to neuronal function or neural development. Subsequently, the RNAi efficacy for these most interesting candidates was re-examined and quantified by Western blot. AQP-8 protein expression was reduced by at least 20% upon loss of 17 genes that are connected to neuronal function tested in the secondary screen while the loss of another neuronal gene activated AQP-8 expression by 1.5-fold (Supplementary Table 1). The list includes among others various G-protein coupled receptors, two FMRF-like peptides, a glutamate gated chloride channel and an ionotropic glutamate receptor. Since neuronally-expressed genes have previously been described to be refractory to RNAi (Timmons et al., 2001), we also analyzed the aqp-8 mRNA expression in corresponding mutants of some interesting candidates, that had failed to pass the secondary RNAi screen. This approach further confirmed the seven transmembrane G-protein coupled olfactory receptor gene odr-10 (Figure 3D). odr-10 has been linked to chemosensation by amphid neurons (Sengupta et al., 1996; Troemel et al., 1997; Zhang et al., 1997). Using qRT-PCR techniques in the odr-10 mutant allele ky32, we detected a significant 0.45-fold reduction of aqp-8 mRNA expression (Figure 3D). The secondary screen also identified the G-protein encoding gene gpa-6, which may function in amphid neuron chemosensation and is co-expressed with odr-10 (Lans et al., 2004). Western blot analysis revealed a reduction in AQP-8 levels by 45% upon RNAi knock-down of gpa-6 (Supplementary Table 1). We confirmed the effect of gpa-6 by measuring the aqp-8 mRNA concentration in the mutant gpa-6(pk480). The qRT-PCR results revealed a significant reduction of aqp-8 expression (Figure 3D).

Two other interesting candidates from our genome wide RNAi screen are the FMRF-like peptides flp-25 and flp-27. Especially flp-27 RNAi significantly suppresses aqp-8 expression levels compared to control levels (Figure 3E). In the kidney of vertebrates aquaporin location and expression are regulated by the peptide hormone vasopressin (Borgnia et al., 1999). C. elegans exhibits a vasopressin homolog called nematocin (ntc-1) and two nematocin receptors ntr-1 and ntr-2 (Beets et al., 2012 and Garrison et al., 2012). Although, none of these receptors is expressed in the excretory cell, we tested nematocin for a regulatory effect on aqp-8 expression. RNAi against nematocin failed to affect aqp-8 RNA expression under isotonic and acute hyper-osmotic stress conditions, but did alter the expression of other C. elegans aquaporins which are susceptible to ntr-1 and ntr-2 signaling under the mentioned conditions (data not shown). The FMRF-like peptide flp-27 may have a modulatory function within the neuronal network or it could have taken over the role of nematocin in the excretory cell and may constitute the connection between neuronal osmosensing and tissue specific effects. All four candidates, odr-10, gpa-6, flp-25, and flp-27 establish a connection between osmosensing in the amphid neurons and the physiological response to osmotic stress.

osm-9 is a member of the TRPV channel family and plays major roles in transduction and regulation of signals in several sensory neurons. The channel is important for processes such as water-soluble chemotaxis, volatile chemotaxis, and osmotic avoidance (Colbert et al., 1997; Bargmann, 2006). However, contrary to other osm genes, osm-9 has not been associated with the osmotic stress response or glycerol accumulation. Recently, the osm-9 mutant allele ok1677 has been shown to promote proteostasis and induce the expression of genes that manage protein damage (Lee et al., 2016). Since protein damage occurs in C. elegans upon exposure to osmotic stress conditions (Burkewitz et al., 2011), osm-9 might be linked to another branch of the osmotic stress response independent of glycerol activation. osm-9 was not identified in our genome wide screen, but osm-9 is co-expressed with gpa-6 and odr-10 in some amphid neurons and all three genes potentially encode components of a common signaling pathway regulating aqp-8 expression (Bargmann, 2006). Furthermore, its role in osmosensing and proteostasis described above made osm-9 an even more interesting candidate. osm-9 RNAi proved to be refractory. Hence, we tested aqp-8 mRNA levels in an osm-9(ok1677) background and detected a 2.81-fold increase under isotonic conditions (Figure 3F). For proper localization in the sensory cilia of the amphid neurons and chemosensation, the OSM-9 TRPV channel requires co-expression of the TRPV channel OCR-2. OSM-9 forms a heterotetramer with OCR-2, which functions in chemosensation, but not in other functions such as e.g., sensory adaptation (Tobin et al., 2002). Consistent with an involvement of OCR-2 in osmoregulation, the RNAi knock-down of ocr-2 in the AKR-14 background caused a significant AQP-8 protein overexpression similar to osm-9(ok1677) overexpression (Figure 3G). These results provide further evidence for an involvement of the OSM-9/OCR-2 TRPV channel in aqp-8 regulation. Further experiments have to be performed to explore the cross talk between osm-9/ocr-2, gpa-6 and odr-10 but potentially all four proteins could function as components of a consecutive pathway in the AWA sensory neuron, in which they are co-expressed.

Standard techniques for C. elegans breeding were applied during this study (Stiernagle, 2006). NGM media plates seeded with the E. coli strain OP50 containing 50 mM NaCl were used to maintain C. elegans. All strains were raised and maintained at 20°C. Standard genetic crossing methods were used to create double mutants. Verification of the mutant genotypes was performed by standard PCR analysis and/or DNA sequencing.

The following strains were used: N2[BRISTOL]^*1/AKR-14 +/+;akaIs2[K02G10.7(translational)::GFP +pCeh361]//LGI:gpdh-1(ok1558)//LGII: osm-8(n1518)/AKR-19osm-8(n1518);akaIs2[K02G10.7(translational)::GFP + pCeh361]//LGIII: osm-7(n1515)/LGX: aqp-8(ok2800)^*1,aqp-8(tm1919)^*2, osm-11(n1604).

^*1 The strains were obtained from the C. elegans Genetic Stock Center (University of Minnesota, USA). ^*2 The strain was received from the National Bioresource Project for the Experimental Animal “Nematode C. elegans” (Japan).

Single RNAi clones from the MRC library were used for RNAi treatment of C. elegans strains. Standard 10 cm NGM plates containing 1 mM ITPG (Carl Roth, Laborbedarf, Germany) were used for RNAi activation and seeded with 500 μl of RNAi bacteria (Rohlfing et al., 2011). C. elegans were hypochlorite treated, synchronized over night at 20°C and ~250 L1 animals were spotted onto each NGM plate. Animals were grown at 20°C for 4 days. At the first day of adulthood animals were screened for phenotypes or harvested for further analysis in M9 buffer, frozen in liquid nitrogen and stored at −80°C.

Chronic hyper-osmotic stress conditions were induced by elevation of the NaCl concentration of the NGM (100, 150, 200 mM; Lamitina et al., 2004; Rohlfing et al., 2010, 2011). Approximately 250 hypochlorite treated and over night synchronized L1 animals were placed on the test plates and grown at 20°C. NGM plates (OP50 or RNAi bacteria) containing 50 mM NaCl were used as control conditions. Animals were harvested for further analysis at the first day of adulthood in M9 buffer, frozen in liquid nitrogen and stored at −80°C.

To induce acute hyper-osmotic stress conditions ~250 hypochlorite treated and synchronized animals were grown on standard 10 cm NGM (50 mM NaCl; OP50 or RNAi bacteria) at 20°C until first day of adulthood. The animals were harvested in M9 buffer and placed on 10 cm NGM plates (OP50 or RNAi bacteria) containing 200 mM NaCl. Again NGM plates containing 50 mM NaCl were used as control conditions. Animals were harvested in M9 buffer at the indicated time after exposure, frozen in liquid nitrogen and stored at −80°C until used for further analysis.

Prior to the assays synchronized L1 animals were grown until first day of adulthood at 20°C on NGM plated with OP50 bacteria. To induce oxidative stress ~250 young adult animals were placed on 10 cm NGM OP50 plates containing 200 mM paraquat. After 1 h exposure animals were harvested for further analysis in M9 buffer, frozen in liquid nitrogen and stored at −80°C. To heat shock C. elegans, young adult animals were placed on 10 cm NGM OP50 plates pre-heated to 35°C. After half an hour animals were harvested in M9 buffer, frozen in liquid nitrogen and stored at -80°C until further use.

Animals were anesthetized for microscopy in 10 mM sodium azide solved in M9 medium (Stiernagle, 2006) with the appropriate salt concentrations to maintain the test conditions. Microscopy analyses were performed using a laser-scanning microscope 510 (LSM510, Carl Zeiss, Jena, Germany). The obtained images were analyzed using the ZEN2008 software (Carl Zeiss, Jena, Germany). Animal length and the extent of AQP-8::GFP fluorescence were measured with the appropriate tools provided by the ZEN2008 software. Varicosities were defined as pearl-like extensions of the posterior excretory canal and counted manually.

RNA was purified from samples harvested from treated and untreated animals (see above) using TRIzol®;(Invitrogen) and the RNeasy RNA purification kit (Qiagen). The purified RNA was converted to cDNA (SensiFAST cDNA Synthesis Kit, Bioline). Quantitative PCR was carried out on an ABI7500fast qPCR System (Applied Biosysthems) using the SYBR green method (KAPA SYBR Fast qPCR Mastermix, PeqLab). The 7500 Software 2.0.1 (Applied Biosysthems) was used to calculate CT values and time fold changes as well as melting curve analysis of the product. The housekeeping gene act-2 (T04C12.5, C. elegans) was used as normalizer. Primer pairs were designed to span an exon-exon boundary to minimize the risk of amplification of residual genomic DNA in the analyzed cDNA probes. All primer pairs were validated prior to use to secure copy number-dependent amplification and single products. Wildtype cDNA was analyzed for the presence of aqp-8 isoforms by PCR, using standard PCR protocols (Taq DNA Polymerase, Roboklon; PCR cycler, Biometra). The PCR product was separated in 1% agarose gel and the result documented (Figure 2E). All primer pairs used in this study are listed in the dataset (Supplementary Table 2).

For the protein purification the samples harvested from treated and untreated animals (see above) were thawed on ice. Subsequent the samples were mixed with 300 μl HEPES-buffered saline (Stiernagle, 2006) containing proteinase inhibitors and grinded for 5 min on ice. The solution was centrifuged for 2 min at 4.000 rpm and the supernatant was used for further analysis. The protein concentration was determined using a standard Bradford assay (Bradford, 1976). SDS electrophoresis of the samples through 10% polyacrylamide gels, blotting onto nitrocellulose membranes and antibody detection were performed as described previously (Rohlfing et al., 2005). The detection of AQP-8::GFP was performed using 1:20,000 dilution of a mixture of two monoclonal Anti-GFP antibodies (Mouse IgG, clones 7.1 and 13.1, Cat. No. 11814460001, Roche) and a HRP-coupled goat anti-mouse (1:20,000, Jackson ImmunoResearch) as secondary antibody. Testing samples from N2[BRISTOL] animals not carrying akaIs2[aqp-8::GFP] results in the loss of the band labeled as specific signal (Supplementary Figure 1B). A SYPRO Ruby staining of the blotting membrane according to the manual (Molecular Probes, Invitrogen) was performed after blotting and before the antibody staining. The SYPRO Ruby staining of the total protein was used as loading control and normalizer to calculate the time fold change in protein abundance (Supplementary Figures 1A,C; Aldrige et al., 2008). The protein expression captured by western blot and SYPRO Ruby staining was analyzed using the densitometry measurement tool of the ZEN2008 software (Carl Zeiss, Jena, Germany).

The genome wide RNAi screen was performed using the MRC feeding RNAi library and the supplement set (Source Bioscience, UK) covering ~87% of the C. elegans genes in 19,726 individual RNAi bacteria feeding strains. For the screen preparation the clones were stamped from the frozen stocks onto 96 well plates containing LB_Carb/Tet Agar and grown for 24 h at 37°C. From these plates 96 well plates containing 100 μl LB_Carb/Tet Medium were inoculated and grown for maximum 16 h at 37°C. Twenty-four well plates containing 2 ml NGM with 50 mM NaCl and 1 mM ITPG were seeded with 50 μl culture per well, each well representing another RNAi bacteria strain. The seeded plates were incubated over night at room temperature and stored at 4°C. Plates were used within 2 days after preparation. RNAi strains containing GFP RNAi vector and the empty RNAi vector L4440 were grown and seeded as described for the library strains. Those 24 well RNAi plates were used as control for the RNAi efficiency. AKR-14 +/+;akaIs2[K02G10.7(translational)::GFP + pCeh361] animals were used as test strain. Animals were hypochlorite treated and synchronized over night at 20°C. Approximately 25 synchronized L1 animals were seeded onto each well of the prepared RNAi plates and grown at 20°C for 4 days until first day of adulthood. Animals were screened for GFP expression using a fluorescence-equipped stereo dissecting microscope (Discovery V8, Zeiss, Germany). RNAi strains suppressing or enhancing the GFP expression of AKR-14 in the primary screen were further analyzed for effects on the AQP-8::GFP protein concentration by western blot analysis (see above). The inserts of effective RNAi suppressor or enhancer strains were verified by DNA sequencing.

For statistical analysis and graphical presentation the program prism6 (GraphPad Software, LaJolla, USA) has been used. If not indicated otherwise mean and SE are plotted and Students t-tests were performed.

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.