Caenorhabditis elegans N2 Bristol worms were maintained on 35 mm agar plates with nematode growth medium lite (Sun and Lambie, 1997) and Escherichia coli OP50 lawn at 20°C (Brenner, 1974). Crowded populations where OP50 bacteria were still present were chunked onto 94 mm agar plates seeded with OP50 and cultured for seven days at 20°C. Most worms developed into the dauer stage due to overcrowding and starvation. Worms were washed off from plates with M9 buffer and treated with 1% SDS solution for 10 min (Cassada and Russell, 1975). Five washing steps with M9 buffer were applied. To remove the liquid, worms were centrifuged at 2,000 g first. Then, worms were pipetted at the edge of a fresh unseeded agar plate for recovery and incubated for about 1–2 h. Dauer and some L1 larvae survived the SDS treatment and spread out over the plate. The agar areas with dead worms were removed with a spatula.

The dauer and L1 larvae that survived SDS were washed off with M9 buffer, which was then exchanged with 20% bovine serum albumin in M9 buffer. Worms were pipetted into high pressure freezing planchettes with 100 μm recesses and high pressure frozen with an EM HPM100 machine (Leica) (Schieber et al., 2017). Worms were freeze substituted according to our published protocol in an EM AFS2 machine (Leica), infiltrated with Durcupan resin, and then minimal resin embedded (Schieber et al., 2017).

Minimal resin embedded worms were further prepared for FIB-SEM imaging (Schieber et al., 2017). The data sets were acquired with a FIB-SEM Auriga 60 or Crossbeam 540 (Carl Zeiss Company) using ATLAS 3D software (part of Atlas5 from Fibics) by milling about 8 nm layers with the ion beam and imaging the block face with 5 nm pixel size. The data sets were acquired at 1.5 kV with the energy-selective back-scattered electron (EsB) detector (grid voltage 1,100 V), at 700 pA. The anterior ends of three dauer larvae (Dauer^E1, Dauer^E2, Dauer^E3) were imaged in transversal sections starting from posterior in the worm body orienting at specific anatomical landmarks of the worm. For Dauer^E1, a larger body segment, starting at the amphid commissures was imaged. The anterior end of an L1 larva (L1^E4) was imaged longitudinally.

Images were aligned using TrakEM2 in Fiji (Cardona et al., 2012; Schindelin et al., 2012). Image stack orientation regarding rotation (transversal), flipping, and stack recursion (longitudinal) as well as cropping, brightness, and contrast were adjusted either with TrakEM2 or Fiji. The image stack of L1^E4 was transformed first from a longitudinal to a transverse stack using the reslicing function of Fiji and then scaled to the respective pixel size by enlargement using a bilinear interpolation. In Dauer^E2 and Dauer^E3, the VSNR denoising tool was used to reduce stationary artifacts (Fehrenbach et al., 2012). L1^E4 was denoised applying a Gaussian blur (sigma 1) in Fiji. The full volumes of Dauer^E1 and Dauer^E3 were prepared and exported with TrakEM2 and then uploaded to CATMAID (Saalfeld et al., 2009; Cardona et al., 2012).

For stage identification, we imaged a transverse section in the middle of the body of each worm to reveal cuticle, alae, and intestine morphology using the FIB-SEM. For additional investigation of the L1 larva, the remaining body fragments were re-embedded in Epon resin and longitudinally sectioned with an ultramicrotome. The sections were contrasted and imaged with a JSM-7500F SEM (JEOL) with our standard settings (Markert et al., 2017).

Images in figures and videos were adjusted and the layout was carried out as described in the Supplementary Material.

We identified the larval stage and sex of the three analyzed dauer larvae and the starved L1 larva by anatomical features listed in Supplementary Table 1 and shown in Supplementary Figure 1. Left and right orientation were concluded from the known imaging direction and the other two body axes. This was validated for all data sets by analyzing the asymmetrical ventral nerve cord and asymmetry of certain neurons in the nerve ring in Dauer^E1. In general, cells and other structures were identified with an orientation provided by published maps (Ward et al., 1975; Albert and Riddle, 1983; White et al., 1986; Hall, 2008; Doroquez et al., 2014; Witvliet et al., 2021). All cells forming sensilla were identified in general by their position and morphology in the anterior end for all four data sets. Investigated neurons that are not part of sensilla (BAG, FLP, URX, URY, AVM, RIP) were identified by skeletal tracing up to the nerve ring region in Dauer^E1 first. Some of these neurons (BAG, URX, URY, RIP) with a characteristic morphology were identified in Dauer^E2, Dauer^E3, and L1^E4 by comparison of their anterior endings with those in Dauer^E1. This was not reliably possible for others (FLP, AVM) as the morphology of their anterior endings was not characteristic in Dauer^E1. RIP neurons endings were not part of the image volume of Dauer^E2. As the dendrites of AM CNs (ASE, ADF, ASG, ASH, ASI, ASJ, ASK, ADL) could not be traced back to their nerve ring entry, they were identified for Dauer^E1 comparing their entry-order into AMsh cell as well as their location of the distal tips in the AMso cell with the literature (Ward et al., 1975; Albert and Riddle, 1983; Hall, 2008; Doroquez et al., 2014). These neurons could not be reliably identified in the other three data sets. Anterior projections of SAB and SAA in the ventral sub-lateral cords in Dauer^E1 could not be traced back to their nerve ring entry or somas, and thus, their identity was not differentiated.

Cells of interest in the most anterior 2,000 images of Dauer^E1 were volumetrically reconstructed (Dauer^E1v) with the 3dmod package of the IMOD software (Kremer et al., 1996). The same cells were reconstructed in Dauer^E2 and L1^E4. In general, cell contours were drawn manually at regular section intervals and contours of skipped sections were interpolated with the 3dmod interpolator plugin. Some cells, like sheath cells, were encapsulating other cells or fusing with themselves, forming inward-facing surfaces which we did not reconstruct as the inner morphology was already defined by the ensheathed cells. Relevant cells in the data volume of Dauer^E1 were skeleton traced (Dauer^E1s) using CATMAID. The tracing nodes were set into the centroid of the cells. This was performed for Dauer^E3 as well (Dauer^E3s). In addition, cells of special interest were volumetrically reconstructed (Dauer^E3v). Measurements were performed in this model as well. We traced some of the IL2Q dendrites retrogradely from the midlines to their 1° dendrite due to resolution limits.

The color code for cells was adapted from literature (Doroquez et al., 2014). The colors of the left and right neurons were assigned different gradings for visual differentiation in some cases (AWA, AWB, AWC, FLP, IL2L, RIP). The cuticle of a larger volume of Dauer^E1s shown in Figure 1B was automatically reconstructed with 3dmod applying the imodauto function to an image stack rendering mask. This mask was created by the application of filters and manual editing in Fiji, 3dmod, and GIMP (GIMP.org Team, 2021) functions. Automatically created contours were manually corrected in 3dmod. The cuticle of Dauer^E1v shown in Figure 2 was manually reconstructed as described above.

Distances and lengths were measured with 3dmod by tracing the respective cell segment with an open contour and getting its length with Edit > Contour > Info for all four data sets (Kremer et al., 1996). The exact number of the dendritic endings of AWA was determined by drawing points on each end in 3dmod and counting the total number of points. Only endings distal to the ciliary base were counted. The length of the AWC wings was measured by a contour from end to end. In addition, the contour length was measured as flat contour applying Edit > Object > Flatten. The span of the AWC wings was measured end-to-end as follows. A point was set at each end and one in the center of the worm body in 3dmod. Then, a snapshot facing the three points in parallel perspective (front view) in the 3dmod model view was taken. The snapshot was imported into Inkscape (Inkscape.org Team, 2021). The two endpoints were connected with the center-point each by two straight lines. The angle between the two lines was measured with the Inkscape measurement tool. This method was adapted from previous measurements of the wing span of AWC endings (Albert and Riddle, 1983). The number of AFD microvilli endings was determined in the same manner as the number of AWA endings.

To get a near-to-native isotropic high-resolution 3D-insight into the anterior sensory apparatus of C. elegans dauer larvae we used high-pressure freezing followed by freeze substitution, minimal resin embedding, and FIB-SEM acquisition as previously described (Schieber et al., 2017).

We analyzed the anterior regions of sensilla of three dauer hermaphrodites by FIB-SEM image stacks, Dauer^E1, Dauer^E2, and Dauer^E3 (with E = EM data set). The number that follows the letter denotes different individuals, and the letter that follows the number denotes different reconstruction techniques (v = volumetric; s = skeletal). We applied volumetric reconstruction (Dauer^E1v, Dauer^E2v, Dauer^E3v) (Figures 1A,B,D) or skeleton tracing (Dauer^E1s, Dauer^E3s) (Figures 1C,E). The Dauer^E1v data set was restricted to the anterior sensilla region (Figure 1B and Supplementary Video 1) while Dauer^E1s was traced to the amphid commissures (Figure 1C and Supplementary Video 2). Only relevant cells of Dauer^E3 were volumetrically reconstructed (Dauer^E3v) (Supplementary Figures 2B, 4D,E, 5C). For comparative reasons we also acquired a data set of the anterior region of the sensilla of a starved L1 hermaphrodite that we reconstructed volumetrically (L1^E4v) (Figure 1F). Dauer^E1, Dauer^E2, Dauer^E3, and L1^E4 larva were maintained under the same conditions. Features of these larvae and properties of these image stacks can be found in section Materials and Methods and Supplementary Tables 1, 2 and Supplementary Figure 1.

For Dauer^E1v, we present volumetric reconstructions of six IL, four CEP, six OL, two amphid sensilla, and sensory endings of URX, URY, FLP, and BAG (Figures 2A,B). 3D reconstructions of the endings of somatic sensory neuron AVM and interneurons (RIP) are included. The spatial organization of all sensilla and sensory neurons is shown in a transverse EM section (Figures 2C,D). Cells are labeled by the same color code used for the EM reconstruction of the anterior sensilla of the adult C. elegans (Doroquez et al., 2014) (Figure 2E) to facilitate comparisons. The full reconstruction of Dauer^E1v and additional details are shown in Supplementary Video 1.

Due to the improved Z-resolution and near-to-native sample preparation, we could identify several features that have been previously either not reported or misreported, which we now can present and clarify in detail.

The crown-like dendrites of the inner labial sensilla IL2L cells are a specific feature of dauer larvae and their gross anatomy has been reported by fluorescence microscopy (Schroeder et al., 2013). Yet, the ultrastructural context of the dauer crown-like dendrites of IL2L and their location in relation to anterior sensilla is still unknown. We investigated and observed an overall similar anatomy for IL2L in all three dauer larvae. The location, morphology, and cellular structure of the crown-like dendrites of IL2L are shown for Dauer^E1v (Figures 3C,I) and Dauer^E3v (Supplementary Figures 2A,E).

For each IL2L neuron (Figure 3A, left), a small 2° dendrite was branching off the end of the 1° dendrite (Figures 3C,E–G), oriented in parallel to the anterior-lateral midline. These dendrites branched again to form a 3° dendrite, which we referred to as the crown-like dendrites (Figures 3C,D,H). The origin of the 2° dendrite was located posterior to the ciliary base (Supplementary Material). Crown-like dendrites had varicosities (Figures 3A,D) and in some cases (e.g., Dauer^E3v), vesicular organelles (Supplementary Figure 2A). They projected to both dorsal and ventral sides, each around the circumference of the worm head, mostly peripherally of amphid, CEP, and OL sensilla (Figures 3C,D,H). Crown-like dendrites were peripherally only in contact with the Hyp syncytium which produces the cuticle (Figure 3C). We also observed crown-like dendrites to overlap in some cases (Figures 3D,H and Supplementary Figure 2B). There were minor differences in IL2L crown-like dendrites among dauer samples as they extended 4° dendrites in Dauer^E3v (Supplementary Figures 2B–D) but not in other dauers (Figure 3H and Supplementary Material).

As described previously (Schroeder et al., 2013), we observed several dendritic branches on IL2Q dendrites (Figures 4A,J and Supplementary Video 2) following the body wall (body wall dendrites) (Figures 4C,D) or along the midline (midline dendrites) into both directions (Figures 4B,G–I). Both types originated mostly from a 2° radial dendrite which projected into the respective midline between BWM and the Hyp cell (Figures 4C,F). Some body wall dendrites originated from midline dendrites (Figure 4B). Body wall dendrites projected further from the midline along the body wall of the respective side, still between BWM and Hyp cell (Figure 4D). Some of them reached SAB neurons in the case of dorsal IL2Q dendrites (Figures 4D,E). Ventral IL2Q dendrites reached neurons in the sub-lateral cords as well but in this study, the identity of ventral SAB vs. SAA projections could not be clarified (Figure 4G). We were able to identify several IL2Q branches that were very clearly visible (Figure 4F). Some IL2Q midline dendrites ended in the midline (Figure 4G) and some were bundled with AVM (Supplementary Material). Dendritic IL2Q branches showed varicosities in most cases (Figure 4D).

The two branches of the ILLso cells interacted with non-IL sensilla cells (URX and BAG). One branch was facing into posterior direction originating from the socket ending (Figure 3A, right) and was enclosed by a certain part of the URX neuron (Figures 5A,I). This part of the URX neuron was then again enclosed by the AMso cell at the level where the AMso cell formed the channel of the AM CNs (Figures 5B,D,H, 6B). The AMso cell ended at the amphid opening of the cuticle and was enclosed there by the OLLso cell (Figure 5C). URX endings were enclosed by AMso cells in all dauers. A similar but less extended ILLso-URX-AMso sandwich structure was observed across all dauers, and in some cases (Dauer^E2v) ILLso was in contact but not enclosed by URX. The described ILLso branch was missing in one case (Dauer^E3s). URX endings were not enclosed by AMso cells in L1^E4v but were enclosing respective branches of ILLso endings (Supplementary Figures 3E,F).

Compared to the branch of ILLso described above, its second branch was originating from the cell rod and enclosed by BAG (Supplementary Figures 3B,C), which was facing into the anterior direction (Figure 3A, right), just below the cuticle surface (Supplementary Figures 3B,C). We observed this in all three dauer larvae and to a lesser extent in L1^E4v where BAG enclosed a swelling at the respective ILLso cell rod posterior to its socket (Supplementary Material).

We observed microtubules in the URX endings, suggesting that those are ciliated (Figures 5E–G). We also observed convoluted whorls of membranes and large vesicles, enclosed either by URX or the AMso cell (Supplementary Figures 3A,G). We found similar membranous structures enclosed by BAG or ILLso cells of which some originated from Hyp cells or URX (Supplementary Figure 3B). However, we could not determine their origin exactly.

Each FLP neuron had one 1° dendrite following the lateral labial bundle to the anterior of which each extended two 2° dendrites following dorsal or ventral labial bundles, respectively (Supplementary Figure 3D). These anterior projections were almost unbranched. Their endings did not appear to be branched in Dauer^E1v (Supplementary Figure 3C). We could not find any cilia microtubules in FLP. During the investigation of the ventral midline of Dauer^E1s, we followed the AVM neuron and found its end anterior in the dauer at approximately 5 μm from the tip of the worm (Supplementary Figures 3C,D) reaching as far as the respective BWM cell.

We determined the exact number of dendritic endings at the sensory tip of AWA neurons in dauer (Table 1 and Figures 6A,D). Branches of AWA neurons in Dauer^E3v widely overlapped (Supplementary Figures 4D,G) which was not the case in Dauer^E1v and just partly in Dauer^E2v. We also counted AWA endings in L1^E4v (Table 1). One AWA neuron showed a wing-like expansion in one branch in L1^E4v (Supplementary Figure 4F).

Notably, we observed variability in the branching patterns of AWB as the two branches of the AWB neuron endings extended into the dorsal and ventral direction in dauer (Figure 6A) while in one case of AWBL of individual Dauer^E3v both branches extended into ventral direction only (Supplementary Figure 4E).

The prominent wing-like endings of AWC neurons were Y-shaped (Figure 6A) in dauers. The AWC wings overlapped on the dorsal and ventral sides (Figure 6C) and intertwined in some cases. We measured the span and length of AWC wings from end to end across all dauer larvae (Table 1). The wing of AWCR had a smaller span in Dauer^E2v (Table 1). This wing fused with itself enclosing AMsh cytoplasm at two positions (Supplementary Figures 4A–C) which was only seen in this data set. The AWC wings in L1^E4v were less expanded and did not overlap (Table 1).

We determined the exact number of AFD microvilli endings (Figure 6A and Table 1). In Dauer^E2v, some posterior microvilli might be missing, because they were cut off in the data set. Therefore, we provide in this study the number of cut off microvilli in the reduced data set in addition, being aware that this is an underestimate of the total number (Table 1). We also counted AFD microvilli endings in L1^E4v (Table 1).

We found that in dauers, RIP neurons (Figures 7A,G) maintain their projections to the pharyngeal nervous system (Figure 7F). They entered through the pharyngeal basal lamina (Figures 7B,D), similarly in L1^E4v (Supplementary Figures 5E,F). Posterior to the position where the endings migrated into the pharynx, they featured a bouton containing electron dense vesicles in both Dauer^E1v and Dauer^E3v (Figure 7E and Supplementary Figures 5A,C) and to a lesser extent in L1^E4v (Supplementary Figure 5D). Inside the pharyngeal nervous system, RIP endings of Dauer^E1v and Dauer^E3v formed such a bouton-like shape as well (Figure 7F and Supplementary Figure 5C) filled with electron dense vesicles (Figure 7C), which were also present in L1^E4v but again, to a lesser extent (Supplementary Figure 5E). The bouton of RIPL, located posterior to the entry into the pharynx in Dauer^E3v, was enclosed by the respective BWM cell (Supplementary Figures 5A,C). We observed a readily identifiable active zone-like structure inside the bouton of RIPL of Dauer^E3v opposing the BWM, which suggests that it is a neuromuscular junction (Supplementary Figure 5B).

We confirm a distinct difference in the morphology between IL2L and IL2Q neuron branching types in dauer previously revealed by a light-microscopy study (Schroeder et al., 2013). These differences in their arborization pattern are mirrored at the molecular level by distinct expressions of transcription factors (Schroeder et al., 2013). Furthermore, these two neuron types differ in their synaptic connectivity in non-dauers (White et al., 1986; Witvliet et al., 2021). These differences suggest a functional difference between IL2L (Figure 3A) and IL2Q (Figure 3B) neurons that is morphologically manifested in dauer.

IL2 neurons are hypothesized to play roles in chemosensation as their endings are exposed to the external environment in non-dauers (Ward et al., 1975; Ware et al., 1975; Lewis and Hodgkin, 1977; Albert and Riddle, 1983). In dauers, sensory endings of IL2 project less far anterior in relation to IL1 endings. This is opposite to non-dauers (Albert and Riddle, 1983). Potential retraction of IL2 sensory endings suggests that chemosensory functions in dauers may be partially replaced by increased mechanosensory or temperature perceptions mediated by dauer-extended branches important for nictation (Lee et al., 2011; Schroeder et al., 2013; Yang et al., 2020). Consistent with this idea, the crown-like dendrites at IL2L endings are located peripherally to other sensilla and are, therefore, in direct contact with the Hyp cell and are in prime position to sense mechanical forces impacting on the anterior surface of the worm. Their location suggests sensing of nose touch. Furthermore, the midline dendrites of IL2QVs (Figure 4A) shared the trajectory of AVM what contributes to gentle touch responses (Chalfie and Sulston, 1981), suggesting similar mechanosensitive functionalities for IL2Qs which are slightly different from IL2Ls. Previously, dauer IL2 neurons were proposed to play similar roles as mechano- and temperature-sensitive FLP neurons (Kaplan and Horvitz, 1993; Chatzigeorgiou and Schafer, 2011; Liu et al., 2012) in non-dauers as they build dendritic branches in a similar receptive field (Albeg et al., 2011; Schroeder et al., 2013; Androwski et al., 2020). The morphological appearance of IL2L crown-like dendrites is fairly stable across dauers, with only slight differences in the 4° dendrites in one individual (Supplementary Figures 2B–D). It would be interesting to identify specific functional roles of the crown-like dendrites. Their overlap may indicate signaling feedback.

We found that dauer-specific dendritic endings of IL2Qs reach the anterior projections of SAA or SAB (Figure 4A). The anterior projections of SAAs originate from their axons (White et al., 1986) but are devoid of synapses (Witvliet et al., 2021). They are proposed to possess stretch receptors and were modeled to be required for periodic bending of the anterior body (White et al., 1986; Sakata and Shingai, 2004; Vidal et al., 2015). An alternative model suggests that IL1 dendrites serve this role instead (Hart et al., 1995; Sakata and Shingai, 2004; Kindt et al., 2007). In non-dauers, anterior projections of SABs make neuromuscular junctions (Witvliet et al., 2021). It would be interesting to examine whether BWM cells receive synaptic input from SABs in dauers and whether SAB is required for nictation. Our current findings let us speculate that IL2Q may acquire mechanosensory functions and may regulate nictation via dauer-specific synapses with SAA or SAB (White et al., 1986; Witvliet et al., 2021).

Being a diapause state, dauer larvae exhibit remarkably quick responses and the fast forward movement induced by mechanical stimuli (Cassada and Russell, 1975; Gaglia and Kenyon, 2009). Our hypotheses that dauer shows a higher degree of plasticity and is more mechanosensitive is further supported by finding a dauer-specific anterior extension of AVM dendrite in Dauer^E1. AVM ends more posterior in non-dauers, approximately at the level of the pharyngeal metacorpus (Chalfie and Sulston, 1981; White et al., 1986). In the future, these observations should be further quantified by analyzing more individuals with imaging techniques that allow much larger sampling such as fluorescence microscopy with specific reporter lines (Ch'ng et al., 2003; Androwski et al., 2020).

Gas sensing neurons BAG interact with ILLso cells in C. elegans adults. Therefore, ILLso is hypothesized to provide structural stability (Doroquez et al., 2014; Cebul et al., 2020). Contrary to previous observations (Doroquez et al., 2014), gas sensing neurons URX interact with ILLso as well (Cebul et al., 2020). Similar observations were made for BAG and URX cell homologs in P. pacificus adults (Hong et al., 2019). We observed similar interaction in C. elegans dauers. ILQso does not form structures to physically interact with BAG or URX. Therefore, ILso cells also consist of two principally different categories similar to IL2 neurons.

It would be interesting to address if the morphological difference within ILso cells is also functionally relevant. ILso cells express receptors required for nose touch sensitivity mediated by IL1 (Hart et al., 1995; Kindt et al., 2007; Han et al., 2013). Such sensitivity may be further functionally enhanced in dauer. Moreover, we discovered that in dauers the AMso cell encloses the URX neuron. In some cases, URX additionally encloses ILLso at the same position, forming a sandwich structure that was not observed in C. elegans (Doroquez et al., 2014) and P. pacificus (Hong et al., 2019) adults. Our investigation also revealed that AMso cells did not enclose URX neuron endings in the starved L1^E4v larva (Supplementary Figures 3E,F). This is likely a dauer-specific adaptation that can be further investigated by examining the morphology of AMso and URX endings with fluorescent reporter lines previously described (Kim and Li, 2004; Low et al., 2019; Cebul et al., 2020; Fung et al., 2020).

We confirm that URX endings are ciliated in dauer (Figures 5F,G), as is the case in adults (Kazatskaya et al., 2020) and P. pacificus (Hong et al., 2019). We are aware of the resolution limits of our FIB-SEM approach. To further clarify precise microtubule arrangement and fine structure, a higher lateral resolution (<4 nm/pixel) by transmission or scanning EM would be necessary. Surprisingly, we did not observe FLP endings to interact with ILLso or to be ciliated, as previously reported for non-dauers (Ward et al., 1975; Ware et al., 1975; Perkins et al., 1986; White et al., 1986; Doroquez et al., 2014) and dauers (Albert and Riddle, 1983).

We observed convoluted whorls of membrane enclosed by or originating from either URX, BAG, AMso, ILLso, or Hyp cells (Supplementary Figures 3A,B). They resemble what has been described as exophers (Melentijevic et al., 2017). Exophers are formed by sensory neurons to overcome neurotoxic stress (Melentijevic et al., 2017). As dauers are in a constant state of physiological stress due to the absence of food uptake (Cassada and Russell, 1975), it is not surprising to find exopher-like structures. Hyp cells are involved in the cellular recycling of exophers (Melentijevic et al., 2017). As AMso and ILLso have hypodermal-like properties (White et al., 1986; Low et al., 2019; Cebul et al., 2020) and interact with URX or BAG, they could take on this task. The formation of exophers reinforces the sensitivity of neurons (Melentijevic et al., 2017). These findings further support that sensory activity is increased in dauers.

Carbon dioxide and temperature can be perceived by AFD neurons (Bretscher et al., 2011; Liu et al., 2012). Both parameters are important for dauers (Golden and Riddle, 1984; Hallem et al., 2011; Yang et al., 2020). We found a range between 82 and 110 microvilli per AFD neuron in dauers (Table 1), two times more than the previous estimation (Albert and Riddle, 1983). AWA terminal branches are between 12 and 24 in dauers (Table 1). Adults possess ~80 branches (Doroquez et al., 2014), implicating a developmental increase as reported for FLP (Albeg et al., 2011). We observed a wing-like swelling in one AWA branch in L1^E4v. This may be an effect of starvation as similar starvation-dependent morphological changes for AWB neurons in adults have been described by Mukhopadhyay et al. (2008). In non-dauers, AWC is rather T-shaped (Ward et al., 1975; Doroquez et al., 2014). In dauers, we found them more as a Y-shape, and the radial span of AWC wings in our data sets are larger (up to 343°, Table 1) than previously reported by Albert and Riddle (1983).

Morphological differences in sensory structures between individuals may be triggered by the external environment or intrinsic influences. In contrast to developmental plasticity where the relevance of environmental influences is clear, here we prefer to use the term variability, as the environmental impact has not been investigated in detail yet. There is variability in the morphology of amphid sensory endings among dauers. AWC in Dauer^E2 showed cupola-shaped structures in parts of the wing. In addition, the branch direction of AWBL in Dauer^E3 was different from others. We hypothesize that phenotypic variability among dauers might be a preadaptation for the dauer population to find and adapt to new habitats.

Pharyngeal pumping is inactivated in dauers (Cassada and Russell, 1975; Albert and Riddle, 1983). We found that RIP, the only direct neuronal connection between the pharyngeal and somatic nervous system, projects in a similar manner as in non-dauers (Ward et al., 1975). RIP projection terminals contain synaptic varicosities filled with electron dense vesicles in our observations which often contain neuropeptides (Ann et al., 1997; Salio et al., 2006). Similar varicosities were not found in starved L1^E4 and have not been described in other non-dauers (Ward et al., 1975; Albertson and Thomson, 1976). In dauers, RIP may also function as motor neuron (Supplementary Figure 5B).

In non-dauers, RIP regulates inhibition of pharynx pumping triggered by light touch perceived by other neurons (Chalfie et al., 1985; Avery and Thomas, 1997). RIP could be responsible for pharynx inhibition in dauers. As IL1 are mechanosensitive (Hart et al., 1995; Kindt et al., 2007) and IL2 are hypothesized to be so with high activity, both may dauer-specifically enhance inhibition of the pharynx via RIP similarly (White et al., 1986).